Ngspice Manual

Ngspice User’s Manual
Version 42 plus
(ngspice development version)
Holger Vogt, Giles Atkinson, Paolo Nenzi
January 14, 2024

2
Locations
The project and download pages of ngspice may be found at
Ngspice home page http://ngspice.sourceforge.net/
Project page at SourceForge http://sourceforge.net/projects/ngspice/
Download page at SourceForge https://sourceforge.net/projects/ngspice/files/ng-spice-
rework/
Git source download https://sourceforge.net/p/ngspice/ngspice/ci/master/tree/
Status
This manual is a work in progress. Some to-dos are listed in Chapt. 24.3. More is surely
needed. You are invited to report bugs, missing items, wrongly described items, bad
English style, etc.
How to use this Manual

The manual is a “work in progress.” It may accompany a specific ngspice release, e.g.
ngspice-35 as manual version 35. If its name contains ‘Version xxplus’, it describes the
actual code status, found at the date of issue in the Git Source Code Management (SCM)
tool. This manual is intended to provide a complete description of ngspice’s functionality,
features, commands, and procedures. This manual is not a book about learning SPICE
usage, however the novice user may find some hints how to start using ngspice. Chapter
21.1 gives a short introduction how to set up and simulate a small circuit. Chapter 32 is
about compiling and installing ngspice from a tarball or the actual Git source code, which
you may find on the ngspice web pages. If you are running a specific Linux distribution,
you may check if it provides ngspice as part of the package. Some are listed here.
License
This document is covered by the Creative Commons Attribution Share-Alike (CC-BY-SA)
v4.0..
Part of chapters 12 and 25-29 are in the public domain.
Chapter 30 is covered by New BSD (chapt. 33.3.2).
Part I
Ngspice User’s Manual
3
Contents
I Ngspice User’s Manual 3
1 Introduction 35
1.1 Simulation Algorithms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
1.1.1 Analog Simulation . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
1.1.2 Matrix solvers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
1.1.3 Device Models for Analog Simulation . . . . . . . . . . . . . . . . . 37
1.1.4 Digital Simulation . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
1.1.5 Mixed-Signal Simulation . . . . . . . . . . . . . . . . . . . . . . . . 38
1.1.6 Mixed-Level Simulation (Electronic and TCAD) . . . . . . . . . . . 38
1.2 Supported Analyses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
1.2.1 DC Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
1.2.2 AC Small-Signal Analysis . . . . . . . . . . . . . . . . . . . . . . . 40
1.2.3 Transient Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
1.2.4 Pole-Zero Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
1.2.5 Small-Signal Distortion Analysis . . . . . . . . . . . . . . . . . . . . 41
1.2.6 Sensitivity Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
1.2.7 Noise Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
1.2.8 Periodic Steady State Analysis . . . . . . . . . . . . . . . . . . . . . 42
1.3 Analysis at Different Temperatures . . . . . . . . . . . . . . . . . . . . . . 43
1.3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
1.3.2 Controlling the temperature . . . . . . . . . . . . . . . . . . . . . . 44
1.4 Convergence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
1.4.1 Voltage convergence criterion . . . . . . . . . . . . . . . . . . . . . 46
1.4.2 Current convergence criterion . . . . . . . . . . . . . . . . . . . . . 46
1.4.3 Convergence failure . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
5
6 CONTENTS
2 Circuit Description 47
2.1 General Structure and Conventions . . . . . . . . . . . . . . . . . . . . . . 47
2.1.1 Input file structure . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
2.1.2 Syntax check . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
2.1.3 Circuit elements (device instances) . . . . . . . . . . . . . . . . . . 48
2.1.4 Some naming conventions . . . . . . . . . . . . . . . . . . . . . . . 49
2.1.5 Topological constraints . . . . . . . . . . . . . . . . . . . . . . . . . 51
2.2 Dot commands . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
2.3 Basic lines . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
2.3.1 .TITLE line . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
2.3.2 .END Line . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
2.3.3 Comments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
2.3.4 End-of-line comments . . . . . . . . . . . . . . . . . . . . . . . . . . 54
2.3.5 Continuation lines . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
2.4 .MODEL Device Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
2.5 .SUBCKT Subcircuits . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
2.5.1 .SUBCKT Line . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
2.5.2 .ENDS Line . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
2.5.3 Subcircuit Calls . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
2.6 .GLOBAL . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
2.7 .INCLUDE . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
2.8 .LIB . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
2.9 .PARAM Parametric netlists . . . . . . . . . . . . . . . . . . . . . . . . . . 59
2.9.1 .param line . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
2.9.2 Brace expressions in circuit elements: . . . . . . . . . . . . . . . . . 60
2.9.3 Subcircuit parameters . . . . . . . . . . . . . . . . . . . . . . . . . 61
2.9.4 Symbol scope . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
2.9.5 Syntax of expressions . . . . . . . . . . . . . . . . . . . . . . . . . . 62
2.9.6 Reserved words . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
2.9.7 A word of caution on the three ngspice expression parsers . . . . . . 65
2.10 .FUNC . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
2.11 .CSPARAM . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
2.12 .TEMP . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
2.13 .IF Condition-Controlled Netlist . . . . . . . . . . . . . . . . . . . . . . . . 67
2.14 Parameters, functions, expressions, and command scripts . . . . . . . . . . 69
2.14.1 Parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
2.14.2 Nonlinear sources . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
2.14.3 Control commands, Command scripts . . . . . . . . . . . . . . . . . 69
CONTENTS 7
3 Circuit Elements and Models 71

3.1 About netlists, device instances, models and model parameters . . . . . . . 71
3.2 General options . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
3.2.1 Paralleling devices with multiplier m . . . . . . . . . . . . . . . . . 73
3.2.2 Instance and model parameters . . . . . . . . . . . . . . . . . . . . 75
3.2.3 Model binning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
3.2.4 Initial conditions . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
3.3 Elementary Devices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
3.3.1 Resistors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
3.3.2 Semiconductor Resistors . . . . . . . . . . . . . . . . . . . . . . . . 78
3.3.3 Semiconductor Resistor Model (R) . . . . . . . . . . . . . . . . . . 78
3.3.4 Resistors, dependent on expressions (behavioral resistor) . . . . . . 80
3.3.5 Resistor with nonlinear r2_cmc model . . . . . . . . . . . . . . . . 80
3.3.6 Capacitors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
3.3.7 Semiconductor Capacitors . . . . . . . . . . . . . . . . . . . . . . . 82
3.3.8 Semiconductor Capacitor Model (C) . . . . . . . . . . . . . . . . . 82
3.3.9 Capacitors, dependent on expressions (behavioral capacitor) . . . . 84
3.3.10 Inductors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
3.3.11 Inductor model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
3.3.12 Coupled (Mutual) Inductors . . . . . . . . . . . . . . . . . . . . . . 87
3.3.13 Inductors, dependent on expressions (behavioral inductor) . . . . . 88
3.3.14 Capacitor or inductor with initial conditions . . . . . . . . . . . . . 89
3.3.15 Switches . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
3.3.16 Switch Model (SW/CSW) . . . . . . . . . . . . . . . . . . . . . . . 91
4 Voltage and Current Sources 93

4.1 Independent Sources for Voltage or Current . . . . . . . . . . . . . . . . . 93
4.1.1 Pulse . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
4.1.2 Sinusoidal . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
4.1.3 Exponential . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
4.1.4 Piece-Wise Linear . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
4.1.5 Single-Frequency FM . . . . . . . . . . . . . . . . . . . . . . . . . . 97
4.1.6 Amplitude modulated source (AM) . . . . . . . . . . . . . . . . . . 98
4.1.7 Transient noise source . . . . . . . . . . . . . . . . . . . . . . . . . 98
4.1.8 Random voltage source . . . . . . . . . . . . . . . . . . . . . . . . . 99
8 CONTENTS
4.1.9 External voltage or current input . . . . . . . . . . . . . . . . . . . 100

4.1.10 Arbitrary Phase Sources . . . . . . . . . . . . . . . . . . . . . . . . 100
4.1.11 RF Port . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
4.2 Linear Dependent Sources . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
4.2.1 Gxxxx: Linear Voltage-Controlled Current Sources (VCCS) . . . . 101
4.2.2 Exxxx: Linear Voltage-Controlled Voltage Sources (VCVS) . . . . . 102
4.2.3 Fxxxx: Linear Current-Controlled Current Sources (CCCS) . . . . 102
4.2.4 Hxxxx: Linear Current-Controlled Voltage Sources (CCVS) . . . . 102
4.2.5 Polynomial Source Compatibility . . . . . . . . . . . . . . . . . . . 103
5 Non-linear Dependent Sources (Behavioral Sources) 105

5.1 Bxxxx: Nonlinear dependent source (ASRC) . . . . . . . . . . . . . . . . . 105
5.1.1 Syntax and usage . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105
5.1.2 Special B-Source Variables time, temper, hertz . . . . . . . . . . . . 109
5.1.3 par(’expression’) . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109
5.1.4 Piecewise Linear Function: pwl . . . . . . . . . . . . . . . . . . . . 109
5.2 Exxxx: non-linear voltage source . . . . . . . . . . . . . . . . . . . . . . . 112
5.2.1 VOL . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112
5.2.2 VALUE . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113
5.2.3 TABLE . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113
5.2.4 POLY . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113
5.2.5 LAPLACE . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113
5.2.6 FREQ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114
5.2.7 AND/OR/NAND/NOR . . . . . . . . . . . . . . . . . . . . . . . . 115
5.3 Gxxxx: non-linear current source . . . . . . . . . . . . . . . . . . . . . . . 115
5.3.1 CUR . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115
5.3.2 VALUE . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116
5.3.3 TABLE . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116
5.3.4 POLY . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116
5.3.5 LAPLACE . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116
5.3.6 FREQ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117
5.3.7 Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117
5.4 Debugging a behavioral source . . . . . . . . . . . . . . . . . . . . . . . . . 117
5.5 POLY Sources . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118
5.5.1 E voltage source, G current source . . . . . . . . . . . . . . . . . . 119
5.5.2 F voltage source, H current source . . . . . . . . . . . . . . . . . . . 119
CONTENTS 9
6 Transmission Lines 121

6.1 Lossless Transmission Lines . . . . . . . . . . . . . . . . . . . . . . . . . . 121
6.2 Lossy Transmission Lines . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122
6.2.1 Lossy Transmission Line Model (LTRA) . . . . . . . . . . . . . . . 122
6.3 Uniform Distributed RC Lines . . . . . . . . . . . . . . . . . . . . . . . . . 124
6.3.1 Uniform Distributed RC Model (URC) . . . . . . . . . . . . . . . . 124
6.4 KSPICE Lossy Transmission Lines . . . . . . . . . . . . . . . . . . . . . . 125
6.4.1 Single Lossy Transmission Line (TXL) . . . . . . . . . . . . . . . . 126
6.4.2 Coupled Multiconductor Line (CPL) . . . . . . . . . . . . . . . . . 126
7 Diodes 129
7.1 Junction Diodes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129
7.2 Diode Model (D) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 130
7.3 Diode Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133
8 BJT 139
8.1 Bipolar Junction Transistors (BJTs) . . . . . . . . . . . . . . . . . . . . . . 139
8.2 BJT Models (NPN/PNP) . . . . . . . . . . . . . . . . . . . . . . . . . . . 139
8.2.1 Gummel-Poon Models . . . . . . . . . . . . . . . . . . . . . . . . . 140
8.2.2 VBIC Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145
8.2.3 MEXTRAM Model . . . . . . . . . . . . . . . . . . . . . . . . . . . 147
8.2.4 HICUM level 2 Model . . . . . . . . . . . . . . . . . . . . . . . . . 148
8.2.5 HICUM level 0 Model . . . . . . . . . . . . . . . . . . . . . . . . . 149
9 JFETs 151
9.1 Junction Field-Effect Transistors (JFETs) . . . . . . . . . . . . . . . . . . 151
9.2 JFET Models (NJF/PJF) . . . . . . . . . . . . . . . . . . . . . . . . . . . 151
9.2.1 Basic model statement . . . . . . . . . . . . . . . . . . . . . . . . . 151
9.2.2 JFET level 1 model with Parker Skellern modification . . . . . . . . 151
9.2.3 JFET level 2 Parker Skellern model . . . . . . . . . . . . . . . . . . 154
10 MESFETs 157
10.1 MESFETs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 157
10.2 MESFET Models (NMF/PMF) . . . . . . . . . . . . . . . . . . . . . . . . 157
10.2.1 Basic model statements . . . . . . . . . . . . . . . . . . . . . . . . . 157
10.2.2 Model by Statz e.a. . . . . . . . . . . . . . . . . . . . . . . . . . . . 157
10.2.3 Model by Ytterdal e.a. . . . . . . . . . . . . . . . . . . . . . . . . . 158
10.2.4 hfet1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 158
10.2.5 hfet2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 159
10 CONTENTS
11 MOSFETs 161
11.1 MOSFET devices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 161
11.2 MOSFET models (NMOS/PMOS) . . . . . . . . . . . . . . . . . . . . . . 162
11.2.1 MOS Level 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 164
11.2.2 MOS Level 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 164
11.2.3 MOS Level 3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 164
11.2.4 MOS Level 6 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 164
11.2.5 Notes on Level 1-6 models . . . . . . . . . . . . . . . . . . . . . . . 164
11.2.6 MOS Level 9 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 167
11.2.7 BSIM Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 167
11.2.8 BSIM1 model (level 4) . . . . . . . . . . . . . . . . . . . . . . . . . 168
11.2.9 BSIM2 model (level 5) . . . . . . . . . . . . . . . . . . . . . . . . . 170
11.2.10 BSIM3 model (levels 8, 49) . . . . . . . . . . . . . . . . . . . . . . . 170
11.2.11 BSIM4 model (levels 14, 54) . . . . . . . . . . . . . . . . . . . . . . 171
11.2.12 BSIMSOI models (levels 10, 58, 55, 56, 57) . . . . . . . . . . . . . . 171
11.2.13 SOI3 model (level 60) . . . . . . . . . . . . . . . . . . . . . . . . . . 172
11.2.14 HiSIM models of the University of Hiroshima . . . . . . . . . . . . 172
11.2.15 MOS models available via OpenVAF/OSDI . . . . . . . . . . . . . . 172
11.3 Power MOSFET model (VDMOS) . . . . . . . . . . . . . . . . . . . . . . 173
12 Mixed-Mode and Behavioral Modeling with XSPICE 179

12.1 Code Model Element & .MODEL Cards . . . . . . . . . . . . . . . . . . . 179
12.1.1 Syntax . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 179
12.1.2 Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 183
12.1.3 Search path for file input . . . . . . . . . . . . . . . . . . . . . . . . 184
12.1.4 Code model location and assessment . . . . . . . . . . . . . . . . . 184
12.2 Analog Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 185
12.2.1 Gain . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 185
12.2.2 Summer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 186
12.2.3 Multiplier . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 187
12.2.4 Divider . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 188
12.2.5 Limiter . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 190
12.2.6 Controlled Limiter . . . . . . . . . . . . . . . . . . . . . . . . . . . 191
12.2.7 PWL Controlled Source . . . . . . . . . . . . . . . . . . . . . . . . 193
12.2.8 PWL Time Controlled Source with optional edge smoothing . . . . 195
CONTENTS 11
12.2.9 Filesource (PWL sourced from file) . . . . . . . . . . . . . . . . . . 197

12.2.10 Multi_input_PWL_block . . . . . . . . . . . . . . . . . . . . . . . 199
12.2.11 Analog Switch . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 200
12.2.12 Alternative Analog Switch . . . . . . . . . . . . . . . . . . . . . . . 201
12.2.13 Zener Diode . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 202
12.2.14 Current Limiter . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 204
12.2.15 Hysteresis Block . . . . . . . . . . . . . . . . . . . . . . . . . . . . 206
12.2.16 Differentiator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 208
12.2.17 Integrator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 209
12.2.18 S-Domain Transfer Function . . . . . . . . . . . . . . . . . . . . . . 211
12.2.19 PWL Transfer Function . . . . . . . . . . . . . . . . . . . . . . . . 214
12.2.20 Slew Rate Block . . . . . . . . . . . . . . . . . . . . . . . . . . . . 216
12.2.21 Inductive Coupling . . . . . . . . . . . . . . . . . . . . . . . . . . . 217
12.2.22 Magnetic Core . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 218
12.2.23 Controlled Sine Wave Oscillator . . . . . . . . . . . . . . . . . . . . 221
12.2.24 Controlled Triangle Wave Oscillator . . . . . . . . . . . . . . . . . . 222
12.2.25 Controlled Square Wave Oscillator . . . . . . . . . . . . . . . . . . 224
12.2.26 Controlled One-Shot . . . . . . . . . . . . . . . . . . . . . . . . . . 225
12.2.27 Capacitance Meter . . . . . . . . . . . . . . . . . . . . . . . . . . . 228
12.2.28 Inductance Meter . . . . . . . . . . . . . . . . . . . . . . . . . . . . 228
12.2.29 Memristor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 229
12.2.30 2D table model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 230
12.2.31 3D table model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 232
12.2.32 Simple Diode Model . . . . . . . . . . . . . . . . . . . . . . . . . . 234
12.2.33 Analog delay . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 236
12.2.34 Potentiometer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 237
12.3 Hybrid Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 238
12.3.1 Digital-to-Analog Node Bridge . . . . . . . . . . . . . . . . . . . . . 238
12.3.2 Analog-to-Digital Node Bridge . . . . . . . . . . . . . . . . . . . . . 240
12.3.3 Bidirectional Analog/Digital Node Bridge . . . . . . . . . . . . . . 241
12.3.4 Controlled Digital Oscillator . . . . . . . . . . . . . . . . . . . . . . 245
12.3.5 Node bridge from digital to real with enable . . . . . . . . . . . . . 246
12.3.6 A Z**-1 block working on real data . . . . . . . . . . . . . . . . . . 246
12.3.7 A gain block for event-driven real data . . . . . . . . . . . . . . . . 247
12.3.8 Node bridge from real to analog voltage . . . . . . . . . . . . . . . 248
12 CONTENTS
12.3.9 Controlled PWM Oscillator . . . . . . . . . . . . . . . . . . . . . . 248

12.4 Digital Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 250
12.4.1 Buffer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 252
12.4.2 Inverter . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 252
12.4.3 And . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 253
12.4.4 Nand . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 254
12.4.5 Or . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 254
12.4.6 Nor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 255
12.4.7 Xor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 255
12.4.8 Xnor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 256
12.4.9 Tristate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 256
12.4.10 Pullup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 258
12.4.11 Pulldown . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 258
12.4.12 D Flip Flop . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 259
12.4.13 JK Flip Flop . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 261
12.4.14 Toggle Flip Flop . . . . . . . . . . . . . . . . . . . . . . . . . . . . 263
12.4.15 Set-Reset Flip Flop . . . . . . . . . . . . . . . . . . . . . . . . . . . 265
12.4.16 D Latch . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 267
12.4.17 Set-Reset Latch . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 269
12.4.18 State Machine . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 272
12.4.19 Frequency Divider . . . . . . . . . . . . . . . . . . . . . . . . . . . 275
12.4.20 RAM . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 276
12.4.21 Digital Source . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 278
12.4.22 LUT . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 280
12.4.23 General LUT . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 280
12.4.24 D_process . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 282
12.4.25 d_cosim . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 284
12.5 Predefined Node Types for event driven simulation . . . . . . . . . . . . . 287
12.5.1 Digital Node Type . . . . . . . . . . . . . . . . . . . . . . . . . . . 287
12.5.2 Real Node Type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 287
12.5.3 Int Node Type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 287
12.5.4 (Digital) Input/Output . . . . . . . . . . . . . . . . . . . . . . . . . 287
12.6 Automatic insertion of bridging devices . . . . . . . . . . . . . . . . . . . . 288
CONTENTS 13
13 Verilog-A Compact Device Models 291

13.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 291
13.2 OSDI/OpenVAF . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 291
13.3 How to create and apply OpenVAF models . . . . . . . . . . . . . . . . . . 292
13.3.1 Preparing for simulation . . . . . . . . . . . . . . . . . . . . . . . . 292
13.3.2 OSDI/OpenVAF examples distributed with ngspice . . . . . . . . . 294
14 Digital Device Models 295

14.1 U devices (basic digital building blocks) . . . . . . . . . . . . . . . . . . . 295
14.1.1 General format . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 296
14.1.2 List of devices available in ngspice (basic types) . . . . . . . . . . . 296
14.1.3 URC transmission line versus U devices . . . . . . . . . . . . . . . . 297
14.2 Support for standard digital devices . . . . . . . . . . . . . . . . . . . . . . 297
14.3 Digital devices defined by a Hardware Description Language . . . . . . . . 298
14.3.1 Another approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . 299
15 Analyses and Output Control (batch mode) 301

15.1 Simulator Variables (.options) . . . . . . . . . . . . . . . . . . . . . . . . . 301
15.1.1 General Options . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 302
15.1.2 OP and DC Solution Options . . . . . . . . . . . . . . . . . . . . . 303
15.1.3 AC Solution Options . . . . . . . . . . . . . . . . . . . . . . . . . . 305
15.1.4 Transient Analysis Options . . . . . . . . . . . . . . . . . . . . . . . 305
15.1.5 ELEMENT Specific options . . . . . . . . . . . . . . . . . . . . . . 306
15.1.6 Transmission Lines Specific Options . . . . . . . . . . . . . . . . . . 307
15.1.7 Precedence of option and .options commands . . . . . . . . . . . . . 307
15.2 Initial Conditions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 307
15.2.1 .NODESET: Specify Initial Node Voltage Guesses . . . . . . . . . . 307
15.2.2 .IC: Set Initial Conditions . . . . . . . . . . . . . . . . . . . . . . . 308
15.3 Analyses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 309
15.3.1 .AC: Small-Signal AC Analysis . . . . . . . . . . . . . . . . . . . . 309
15.3.2 .DC: DC Transfer Function . . . . . . . . . . . . . . . . . . . . . . 310
15.3.3 .DISTO: Distortion Analysis . . . . . . . . . . . . . . . . . . . . . . 310
15.3.4 .NOISE: Noise Analysis . . . . . . . . . . . . . . . . . . . . . . . . 312
15.3.5 .OP: Operating Point Analysis . . . . . . . . . . . . . . . . . . . . . 313
15.3.6 .PZ: Pole-Zero Analysis . . . . . . . . . . . . . . . . . . . . . . . . . 315
15.3.7 .SENS: DC or Small-Signal AC Sensitivity Analysis . . . . . . . . . 315
14 CONTENTS
15.3.8 .SP S-Parameter Analysis . . . . . . . . . . . . . . . . . . . . . . . 316

15.3.9 .TF: Transfer Function Analysis . . . . . . . . . . . . . . . . . . . . 317
15.3.10 .TRAN: Transient Analysis . . . . . . . . . . . . . . . . . . . . . . 317
15.3.11 Transient noise analysis (at low frequency) . . . . . . . . . . . . . . 318
15.3.12 .PSS: Periodic Steady State Analysis . . . . . . . . . . . . . . . . . 321
15.4 Measurements after AC, DC and Transient Analysis . . . . . . . . . . . . . 322
15.4.1 .meas(ure) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 322
15.4.2 batch versus interactive mode . . . . . . . . . . . . . . . . . . . . . 322
15.4.3 General remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . 322
15.4.4 Input . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 323
15.4.5 Trig Targ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 324
15.4.6 Find ... When . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 325
15.4.7 AVG|MIN|MAX|PP|RMS|MIN_AT|MAX_AT . . . . . . . . . . . . . . . . 326
15.4.8 Integ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 327
15.4.9 param . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 327
15.4.10 par(’expression’ ) . . . . . . . . . . . . . . . . . . . . . . . . . . . . 327
15.4.11 Deriv . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 328
15.4.12 More examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 328
15.5 Safe Operating Area (SOA) warning messages . . . . . . . . . . . . . . . . 329
15.5.1 Resistor and Capacitor SOA model parameters . . . . . . . . . . . 330
15.5.2 Diode SOA model parameters . . . . . . . . . . . . . . . . . . . . . 330
15.5.3 BJT SOA model parameters . . . . . . . . . . . . . . . . . . . . . . 331
15.5.4 MOS SOA model parameters . . . . . . . . . . . . . . . . . . . . . 332
15.5.5 VDMOS SOA model parameters . . . . . . . . . . . . . . . . . . . 333
15.6 Batch Output . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 333
15.6.1 .SAVE: Name vector(s) to be saved in raw file . . . . . . . . . . . . 333
15.6.2 .PRINT Lines . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 334
15.6.3 .PLOT Lines . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 335
15.6.4 .FOUR: Fourier Analysis of Transient Analysis Output . . . . . . . 335
15.6.5 .PROBE: Save device node currents, device power dissipation, or
differential voltages between arbitrary nodes . . . . . . . . . . . . . 336
15.6.6 par(’expression’): Algebraic expressions for output . . . . . . . . . . 340
15.6.7 .width . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 341
15.7 Measuring current through device terminals . . . . . . . . . . . . . . . . . 341
15.7.1 Using the .probe command . . . . . . . . . . . . . . . . . . . . . . . 341
15.7.2 Adding a voltage source in series . . . . . . . . . . . . . . . . . . . 341
15.7.3 Using option ’savecurrents’ . . . . . . . . . . . . . . . . . . . . . . . 342
CONTENTS 15
16 Starting ngspice 343

16.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 343
16.2 Where to obtain ngspice . . . . . . . . . . . . . . . . . . . . . . . . . . . . 343
16.3 Command line options for starting ngspice . . . . . . . . . . . . . . . . . . 344
16.4 Starting options . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 346
16.4.1 Batch mode . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 346
16.4.2 Interactive mode . . . . . . . . . . . . . . . . . . . . . . . . . . . . 346
16.4.3 Control mode (Interactive mode with control file or control section) 347
16.5 Standard configuration file spinit . . . . . . . . . . . . . . . . . . . . . . . 348
16.6 User defined configuration file .spiceinit . . . . . . . . . . . . . . . . . . . . 350
16.7 Environmental variables . . . . . . . . . . . . . . . . . . . . . . . . . . . . 351
16.7.1 Ngspice specific variables . . . . . . . . . . . . . . . . . . . . . . . . 351
16.7.2 Common environment variables . . . . . . . . . . . . . . . . . . . . 352
16.8 Memory usage . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 352
16.9 Simulation time . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 352
16.10Ngspice on multi-core processors using OpenMP . . . . . . . . . . . . . . . 353
16.10.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 353
16.10.2 Internals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 353
16.10.3 Some results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 354
16.10.4 Usage . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 355
16.10.5 Literature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 355
16.11Server mode option -s . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 355
16.12Pipe mode option -p . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 357
16.13Ngspice control via input, output fifos . . . . . . . . . . . . . . . . . . . . . 358
16.14Compatibility . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 359
16.14.1 Compatibility mode . . . . . . . . . . . . . . . . . . . . . . . . . . . 359
16.14.2 Missing functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . 360
16.14.3 Devices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 360
16.14.4 Controls and commands . . . . . . . . . . . . . . . . . . . . . . . . 361
16.14.5 PSPICE Compatibility mode . . . . . . . . . . . . . . . . . . . . . 362
16.14.6 LTSPICE Compatibility mode . . . . . . . . . . . . . . . . . . . . . 363
16.14.7 LTSPICE/PSPICE Compatibility mode . . . . . . . . . . . . . . . 365
16.14.8 KiCad Compatibility mode . . . . . . . . . . . . . . . . . . . . . . . 365
16.14.9 Spectre Compatibility mode . . . . . . . . . . . . . . . . . . . . . . 366
16.14.10HSPICE Compatibility mode . . . . . . . . . . . . . . . . . . . . . 366
16 CONTENTS
16.15Tests . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 366
16.16Tools for debugging a circuit netlist . . . . . . . . . . . . . . . . . . . . . . 367
16.16.1 options and initial conditions . . . . . . . . . . . . . . . . . . . . . 367
16.16.2 set debug . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 367
16.16.3 set ngdebug . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 368
16.16.4 miscellaneous . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 368
16.17Reporting bugs and errors . . . . . . . . . . . . . . . . . . . . . . . . . . . 368
17 Interactive Interpreter 369

17.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 369
17.2 Expressions, Functions, and Constants . . . . . . . . . . . . . . . . . . . . 370
17.3 Plots . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 374
17.4 Command Interpretation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 375
17.4.1 On the console . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 375
17.4.2 Scripts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 375
17.4.3 Add-on to circuit file . . . . . . . . . . . . . . . . . . . . . . . . . . 375
17.5 Commands . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 376
17.5.1 Ac: Perform an AC, small-signal frequency response analysis . . . . 376
17.5.2 Alias: Create an alias for a command . . . . . . . . . . . . . . . . . 377
17.5.3 Alter: Change a device or model parameter . . . . . . . . . . . . . 377
17.5.4 Altermod: Change model parameter(s) . . . . . . . . . . . . . . . . 379
17.5.5 Alterparam: Change value of a global parameter . . . . . . . . . . 380
17.5.6 Asciiplot: Plot values using old-style character plots . . . . . . . . . 381
17.5.7 Aspice*: Asynchronous ngspice run . . . . . . . . . . . . . . . . . . 381
17.5.8 Bg_ctrl**: suspend running controls until bg_run has finished . . . 381
17.5.9 Bg_halt**: halt a run . . . . . . . . . . . . . . . . . . . . . . . . . 381
17.5.10 Bg_run**: Run analysis from the input file in the background thread382
17.5.11 Bug: Output URL for ngspice bug tracker . . . . . . . . . . . . . . 382
17.5.12 Cd: Change directory . . . . . . . . . . . . . . . . . . . . . . . . . . 382
17.5.13 Cdump: Dump the control flow to the screen . . . . . . . . . . . . . 382
17.5.14 Circbyline: Enter a circuit line by line . . . . . . . . . . . . . . . . 383
17.5.15 Codemodel: Load an XSPICE code model library . . . . . . . . . . 384
17.5.16 Compose: Compose a vector . . . . . . . . . . . . . . . . . . . . . . 384
17.5.17 Cutout: Cut out a section of all vectors in a tran plot . . . . . . . . 385
17.5.18 Dc: Perform a DC-sweep analysis . . . . . . . . . . . . . . . . . . . 385
CONTENTS 17
17.5.19 Define: Define a function . . . . . . . . . . . . . . . . . . . . . . . . 386

17.5.20 Deftype: Define a new type for a vector or plot . . . . . . . . . . . 386
17.5.21 Delete: Remove a trace or breakpoint . . . . . . . . . . . . . . . . . 386
17.5.22 Destroy: Delete an output data set . . . . . . . . . . . . . . . . . . 387
17.5.23 Devhelp: information on available devices . . . . . . . . . . . . . . 387
17.5.24 Diff: Compare vectors . . . . . . . . . . . . . . . . . . . . . . . . . 388
17.5.25 Display: List known vectors and types . . . . . . . . . . . . . . . . 388
17.5.26 Echo: Print text . . . . . . . . . . . . . . . . . . . . . . . . . . . . 388
17.5.27 Edit*: Edit the current circuit . . . . . . . . . . . . . . . . . . . . . 389
17.5.28 Edisplay: Print a list of all the event nodes . . . . . . . . . . . . . 389
17.5.29 Eprint: Print an event driven node . . . . . . . . . . . . . . . . . . 389
17.5.30 Eprvcd: Dump nodes in VCD format . . . . . . . . . . . . . . . . 389
17.5.31 Esave: Save a set of event node outputs . . . . . . . . . . . . . . . 390
17.5.32 Fclose: close an open file handle . . . . . . . . . . . . . . . . . . . . 390
17.5.33 FFT: fast Fourier transform of vectors . . . . . . . . . . . . . . . . 390
17.5.34 Fopen: open a text file . . . . . . . . . . . . . . . . . . . . . . . . . 392
17.5.35 Fourier: Perform a Fourier transform . . . . . . . . . . . . . . . . . 392
17.5.36 Fread: read into a variable from a text file . . . . . . . . . . . . . . 393
17.5.37 Getcwd: Print the current working directory . . . . . . . . . . . . . 394
17.5.38 Gnuplot: Graphics output via gnuplot . . . . . . . . . . . . . . . . 394
17.5.39 Hardcopy: Save a plot to a file for printing . . . . . . . . . . . . . . 394
17.5.40 Help: Print summaries of Ngspice commands . . . . . . . . . . . . . 394
17.5.41 History: Review previous commands . . . . . . . . . . . . . . . . . 395
17.5.42 Inventory: Print circuit inventory . . . . . . . . . . . . . . . . . . . 398
17.5.43 Iplot*: Incremental plot . . . . . . . . . . . . . . . . . . . . . . . . 398
17.5.44 Jobs*: List active asynchronous ngspice runs . . . . . . . . . . . . . 398
17.5.45 Let: Assign a value to a vector . . . . . . . . . . . . . . . . . . . . 398
17.5.46 Linearize: Interpolate to a linear scale . . . . . . . . . . . . . . . . 399
17.5.47 Listing: Print a listing of the current circuit . . . . . . . . . . . . . 400
17.5.48 Load: Load rawfile data . . . . . . . . . . . . . . . . . . . . . . . . 401
17.5.49 Mc_source: Reload the circuit netlist from an internal storage . . . 401
17.5.50 Meas: Measurements on simulation data . . . . . . . . . . . . . . . 401
17.5.51 Mdump: Dump the matrix values to a file (or to console) . . . . . . 402
17.5.52 Mrdump: Dump the matrix right hand side values to a file (or to
console) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 402
18 CONTENTS
17.5.53 Noise: Noise analysis . . . . . . . . . . . . . . . . . . . . . . . . . . 402

17.5.54 Op: Perform an operating point analysis . . . . . . . . . . . . . . . 403
17.5.55 Option: Set a ngspice option . . . . . . . . . . . . . . . . . . . . . . 403
17.5.56 Plot*: Plot vectors on the display . . . . . . . . . . . . . . . . . . . 404
17.5.57 Pre_<command>: execute commands prior to parsing the circuit . 406
17.5.58 Pre_OSDI: load a *.osdi compact device model shared library . . . 406
17.5.59 Print: Print values . . . . . . . . . . . . . . . . . . . . . . . . . . . 407
17.5.60 Psd: power spectral density of vectors . . . . . . . . . . . . . . . . 407
17.5.61 Quit: Leave Ngspice . . . . . . . . . . . . . . . . . . . . . . . . . . 408
17.5.62 Rehash: Reset internal hash tables . . . . . . . . . . . . . . . . . . 408
17.5.63 Remcirc: Remove the current circuit . . . . . . . . . . . . . . . . . 408
17.5.64 Remzerovec: Remove zero length vectors . . . . . . . . . . . . . . . 408
17.5.65 Reset: Reset an analysis . . . . . . . . . . . . . . . . . . . . . . . . 408
17.5.66 Reshape: Alter the dimensionality or dimensions of a vector . . . . 409
17.5.67 Resume: Continue a simulation after a stop . . . . . . . . . . . . . 409
17.5.68 Rspice*: Remote ngspice submission . . . . . . . . . . . . . . . . . 410
17.5.69 Run: Run analysis from the input file . . . . . . . . . . . . . . . . . 410
17.5.70 Rusage: Resource usage . . . . . . . . . . . . . . . . . . . . . . . . 410
17.5.71 Save: Save a set of outputs . . . . . . . . . . . . . . . . . . . . . . . 411
17.5.72 Sens: Run a sensitivity analysis . . . . . . . . . . . . . . . . . . . . 413
17.5.73 17.5.73Set: Set the value of a variable . . . . . . . . . . . . . . . . . 413
17.5.74 Setcs: Set the value of a variable, case preserved . . . . . . . . . . . 414
17.5.75 Setcirc: Change the current circuit . . . . . . . . . . . . . . . . . . 414
17.5.76 Setplot: Switch the current set of vectors . . . . . . . . . . . . . . . 415
17.5.77 Setscale: Set the scale vector for the current plot . . . . . . . . . . 415
17.5.78 Setseed: Set the seed value for the random number generator . . . . 415
17.5.79 Settype: Set the type of a vector . . . . . . . . . . . . . . . . . . . 416
17.5.80 Shell: Call the command interpreter . . . . . . . . . . . . . . . . . . 416
17.5.81 Shift: Alter a list variable . . . . . . . . . . . . . . . . . . . . . . . 416
17.5.82 Show: List device state . . . . . . . . . . . . . . . . . . . . . . . . . 417
17.5.83 Showmod: List model parameter values . . . . . . . . . . . . . . . . 417
17.5.84 Snload: Load the snapshot file . . . . . . . . . . . . . . . . . . . . . 418
17.5.85 Snsave: Save a snapshot file . . . . . . . . . . . . . . . . . . . . . . 418
17.5.86 Source: Read a ngspice input file . . . . . . . . . . . . . . . . . . . 419
17.5.87 Sp: S-Parameter Analysis . . . . . . . . . . . . . . . . . . . . . . . 420
CONTENTS 19
17.5.88 Spec: Create a frequency domain plot . . . . . . . . . . . . . . . . . 420

17.5.89 Status: Display breakpoint information . . . . . . . . . . . . . . . . 421
17.5.90 Step: Run a fixed number of time-points . . . . . . . . . . . . . . . 421
17.5.91 Stop: Set a breakpoint . . . . . . . . . . . . . . . . . . . . . . . . . 421
17.5.92 Strcmp: Compare two strings . . . . . . . . . . . . . . . . . . . . . 422
17.5.93 Strslice: Extract a substring . . . . . . . . . . . . . . . . . . . . . . 422
17.5.94 Strstr: Find a substring . . . . . . . . . . . . . . . . . . . . . . . . 422
17.5.95 Sysinfo: Print system information . . . . . . . . . . . . . . . . . . . 422
17.5.96 Tf: Run a Transfer Function analysis . . . . . . . . . . . . . . . . . 423
17.5.97 Trace: Trace nodes . . . . . . . . . . . . . . . . . . . . . . . . . . . 424
17.5.98 Tran: Perform a transient analysis . . . . . . . . . . . . . . . . . . 424
17.5.99 Transpose: Swap the elements in a multi-dimensional data set . . . 425
17.5.100Unalias: Retract an alias . . . . . . . . . . . . . . . . . . . . . . . . 425
17.5.101Undefine: Retract a definition . . . . . . . . . . . . . . . . . . . . . 425
17.5.102Unlet: Delete the specified vector(s) . . . . . . . . . . . . . . . . . . 425
17.5.103Unset: Clear a variable . . . . . . . . . . . . . . . . . . . . . . . . . 426
17.5.104Version: Print the version of ngspice . . . . . . . . . . . . . . . . . 426
17.5.105Where: Identify troublesome node or device . . . . . . . . . . . . . 427
17.5.106Wrdata: Write data to a file (simple table) . . . . . . . . . . . . . . 428
17.5.107Write: Write data to a file (Spice3f5 format) . . . . . . . . . . . . . 428
17.5.108Wrnodev: Write node voltage values to a file (.ic=xx format) . . . . 429
17.5.109Wrs2p: Write scattering parameters to file (Touchstone® format) . 430
17.6 Control Structures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 430
17.6.1 While - End . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 430
17.6.2 Repeat - End . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 431
17.6.3 Dowhile - End . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 431
17.6.4 Foreach - End . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 432
17.6.5 If - Then - Else . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 432
17.6.6 Label . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 433
17.6.7 Goto . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 433
17.6.8 Continue . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 434
17.6.9 Break . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 434
17.7 Internally predefined variables . . . . . . . . . . . . . . . . . . . . . . . . . 434
17.8 Scripts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 442
17.8.1 Variables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 443
20 CONTENTS
17.8.2 Vectors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 444

17.8.3 Assessing vectors in subcircuits . . . . . . . . . . . . . . . . . . . . 444
17.8.4 Commands . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 445
17.8.5 control structures . . . . . . . . . . . . . . . . . . . . . . . . . . . . 445
17.8.6 Example script ’spectrum’ . . . . . . . . . . . . . . . . . . . . . . . 449
17.8.7 Example script for random numbers . . . . . . . . . . . . . . . . . . 451
17.8.8 Parameter sweep . . . . . . . . . . . . . . . . . . . . . . . . . . . . 452
17.8.9 Output redirection . . . . . . . . . . . . . . . . . . . . . . . . . . . 452
17.9 Scattering parameters (S-parameters) . . . . . . . . . . . . . . . . . . . . . 454
17.9.1 Intro . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 454
17.9.2 S-parameter measurement basics . . . . . . . . . . . . . . . . . . . 454
17.9.3 Usage of .sp and sp . . . . . . . . . . . . . . . . . . . . . . . . . . . 456
17.9.4 Usage of the script . . . . . . . . . . . . . . . . . . . . . . . . . . . 456
17.10Using shell variables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 456
17.11MISCELLANEOUS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 457
17.12Bugs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 457
18 Ngspice User Interfaces 459

18.1 MS Windows Graphical User Interface . . . . . . . . . . . . . . . . . . . . 459
18.2 MS Windows Console . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 462
18.3 Linux . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 463
18.4 CygWin . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 463
18.5 Error handling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 463
18.6 Output-to-file options . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 464
18.6.1 Graphics files . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 464
18.6.2 Tabulated files . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 469
18.7 Gnuplot . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 472
18.7.1 Using Gnuplot to produce 1D graphs of (electrical) simulation results472
18.7.2 Using gnuplot to produce 2D contour plots for Cider . . . . . . . . 473
18.8 Integration with CAD software and ‘third party’ GUIs . . . . . . . . . . . 477
18.8.1 KiCad . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 477
18.8.2 Xschem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 477
18.8.3 Qucs-S . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 477
18.8.4 GNU Spice GUI . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 477
18.8.5 XCircuit . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 478
18.8.6 GEDA . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 478
18.8.7 MSEspice . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 478
18.8.8 GNU Octave . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 478
CONTENTS 21
19 ngspice as shared library or dynamic link library 479

19.1 Compile options . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 479
19.1.1 How to get the sources . . . . . . . . . . . . . . . . . . . . . . . . . 479
19.1.2 Linux, MINGW, CYGWIN . . . . . . . . . . . . . . . . . . . . . . 479
19.1.3 MS Visual Studio . . . . . . . . . . . . . . . . . . . . . . . . . . . . 480
19.2 Linking shared ngspice to a calling application . . . . . . . . . . . . . . . . 480
19.2.1 Linking during creating the caller . . . . . . . . . . . . . . . . . . . 480
19.2.2 Loading at runtime . . . . . . . . . . . . . . . . . . . . . . . . . . . 480
19.3 Shared ngspice API . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 480
19.3.1 structs and types defined for transporting data . . . . . . . . . . . . 480
19.3.2 Exported functions . . . . . . . . . . . . . . . . . . . . . . . . . . . 483
19.3.3 Callback functions . . . . . . . . . . . . . . . . . . . . . . . . . . . 485
19.4 General remarks on using the API . . . . . . . . . . . . . . . . . . . . . . . 488
19.4.1 Loading a netlist . . . . . . . . . . . . . . . . . . . . . . . . . . . . 488
19.4.2 Running the simulation . . . . . . . . . . . . . . . . . . . . . . . . . 490
19.4.3 Accessing data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 491
19.4.4 Altering model or device parameters . . . . . . . . . . . . . . . . . 492
19.4.5 Output . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 492
19.4.6 Error handling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 492
19.5 Example applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 493
19.6 ngspice parallel . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 493
19.6.1 Go parallel! . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 493
19.6.2 Additional exported functions . . . . . . . . . . . . . . . . . . . . . 495
19.6.3 Additional callback functions . . . . . . . . . . . . . . . . . . . . . 496
19.6.4 Parallel ngspice example . . . . . . . . . . . . . . . . . . . . . . . . 497
20 TCLspice 499
20.1 tclspice framework . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 499
20.2 tclspice documentation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 499
20.3 spicetoblt . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 500
20.4 Running TCLspice . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 500
20.5 examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 501
20.5.1 Active capacitor measurement . . . . . . . . . . . . . . . . . . . . . 501
20.5.2 Optimization of a linearization circuit for a Thermistor . . . . . . . 503
20.5.3 Progressive display . . . . . . . . . . . . . . . . . . . . . . . . . . . 507
22 CONTENTS
20.6 Compiling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 508

20.6.1 Linux . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 508
20.6.2 MS Windows . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 508
20.7 MS Windows 32 Bit binaries . . . . . . . . . . . . . . . . . . . . . . . . . . 509
21 Example Circuits 511

21.1 AC coupled transistor amplifier . . . . . . . . . . . . . . . . . . . . . . . . 511
21.2 Differential Pair . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 517
21.3 MOSFET Characterization . . . . . . . . . . . . . . . . . . . . . . . . . . . 517
21.4 RTL Inverter . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 517
21.5 Four-Bit Binary Adder (Bipolar) . . . . . . . . . . . . . . . . . . . . . . . 518
21.6 Four-Bit Binary Adder (MOS) . . . . . . . . . . . . . . . . . . . . . . . . . 520
21.7 Transmission-Line Inverter . . . . . . . . . . . . . . . . . . . . . . . . . . . 521
22 Statistical circuit analysis 523

22.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 523
22.2 Using random param(eters) . . . . . . . . . . . . . . . . . . . . . . . . . . 523
22.3 Behavioral sources (B, E, G, R, L, C) with random control . . . . . . . . . 525
22.4 ngspice scripting language . . . . . . . . . . . . . . . . . . . . . . . . . . . 526
22.5 Monte-Carlo Simulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 527
22.5.1 Example 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 527
22.5.2 Example 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 529
22.5.3 Example 3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 529
22.6 Data evaluation with Gnuplot . . . . . . . . . . . . . . . . . . . . . . . . . 530
23 Circuit optimization with ngspice 533

23.1 Optimization of a circuit . . . . . . . . . . . . . . . . . . . . . . . . . . . . 533
23.2 ngspice optimizer using ngspice scripts . . . . . . . . . . . . . . . . . . . . 534
23.3 ngspice optimizer using tclspice . . . . . . . . . . . . . . . . . . . . . . . . 534
23.4 ngspice optimizer using a Python script . . . . . . . . . . . . . . . . . . . . 534
23.5 ngspice optimizer using ASCO . . . . . . . . . . . . . . . . . . . . . . . . . 534
23.5.1 Three stage operational amplifier . . . . . . . . . . . . . . . . . . . 535
23.5.2 Digital inverter . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 536
23.5.3 Bandpass . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 538
23.5.4 Class-E power amplifier . . . . . . . . . . . . . . . . . . . . . . . . 538
CONTENTS 23
24 Notes 539
24.1 Glossary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 539
24.2 Acronyms and Abbreviations . . . . . . . . . . . . . . . . . . . . . . . . . . 540
24.3 To Do . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 541
II XSPICE Software User’s Manual 547
25 XSPICE Basics 549

25.1 ngspice with the XSPICE option . . . . . . . . . . . . . . . . . . . . . . . 549
25.2 The XSPICE Code Model Subsystem . . . . . . . . . . . . . . . . . . . . . 549
25.3 XSPICE Top-Level Diagram . . . . . . . . . . . . . . . . . . . . . . . . . . 550
26 Execution Procedures 553

26.1 Simulation and Modeling Overview . . . . . . . . . . . . . . . . . . . . . . 553
26.1.1 Describing the Circuit . . . . . . . . . . . . . . . . . . . . . . . . . 553
26.2 Circuit Description Syntax . . . . . . . . . . . . . . . . . . . . . . . . . . . 559
26.2.1 XSPICE Syntax Extensions . . . . . . . . . . . . . . . . . . . . . . 559
26.3 How to create code models . . . . . . . . . . . . . . . . . . . . . . . . . . . 561
27 Example circuits 565

27.1 Amplifier with XSPICE model ‘gain’ . . . . . . . . . . . . . . . . . . . . . 566
27.2 XSPICE advanced usage . . . . . . . . . . . . . . . . . . . . . . . . . . . . 567
27.2.1 Circuit example C3 . . . . . . . . . . . . . . . . . . . . . . . . . . . 567
27.2.2 Running example C3 . . . . . . . . . . . . . . . . . . . . . . . . . . 570
28 Code Models and User-Defined Nodes 575

28.1 Code Model Data Type Definitions . . . . . . . . . . . . . . . . . . . . . . 576
28.2 Creating Code Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 577
28.3 Creating User-Defined Nodes . . . . . . . . . . . . . . . . . . . . . . . . . . 577
28.4 Adding a new code model library . . . . . . . . . . . . . . . . . . . . . . . 578
28.5 Compiling and loading the new code model (library) . . . . . . . . . . . . 579
28.6 Interface Specification File . . . . . . . . . . . . . . . . . . . . . . . . . . 579
28.6.1 The Name Table . . . . . . . . . . . . . . . . . . . . . . . . . . . . 581
28.6.2 The Port Table . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 581
28.6.3 The Parameter Table . . . . . . . . . . . . . . . . . . . . . . . . . . 583
28.6.4 Static Variable Table . . . . . . . . . . . . . . . . . . . . . . . . . . 585
24 CONTENTS
28.7 Model Definition File . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 586

28.7.1 Macros . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 586
28.7.2 Function Library . . . . . . . . . . . . . . . . . . . . . . . . . . . . 596
28.8 User-Defined Node Definition File . . . . . . . . . . . . . . . . . . . . . . . 605
28.8.1 Macros . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 606
28.8.2 Function Library . . . . . . . . . . . . . . . . . . . . . . . . . . . . 606
28.8.3 Example UDN Definition File . . . . . . . . . . . . . . . . . . . . . 609
29 Error Messages 613

29.1 Preprocessor Error Messages . . . . . . . . . . . . . . . . . . . . . . . . . . 613
29.2 Simulator Error Messages . . . . . . . . . . . . . . . . . . . . . . . . . . . 618
29.3 Code Model Error Messages . . . . . . . . . . . . . . . . . . . . . . . . . . 619
29.3.1 Code Model aswitch . . . . . . . . . . . . . . . . . . . . . . . . . . 619
29.3.2 Code Model climit . . . . . . . . . . . . . . . . . . . . . . . . . . . 620
29.3.3 Code Model core . . . . . . . . . . . . . . . . . . . . . . . . . . . . 620
29.3.4 Code Model d_osc . . . . . . . . . . . . . . . . . . . . . . . . . . . 620
29.3.5 Code Model d_source . . . . . . . . . . . . . . . . . . . . . . . . . 621
29.3.6 Code Model d_state . . . . . . . . . . . . . . . . . . . . . . . . . . 621
29.3.7 Code Model oneshot . . . . . . . . . . . . . . . . . . . . . . . . . . 622
29.3.8 Code Model pwl . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 622
29.3.9 Code Model s_xfer . . . . . . . . . . . . . . . . . . . . . . . . . . . 622
29.3.10 Code Model sine . . . . . . . . . . . . . . . . . . . . . . . . . . . . 623
29.3.11 Code Model square . . . . . . . . . . . . . . . . . . . . . . . . . . . 623
29.3.12 Code Model triangle . . . . . . . . . . . . . . . . . . . . . . . . . . 624
III CIDER 625
30 CIDER User’s Manual 627

30.1 SPECIFICATION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 627
30.1.1 Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 628
30.2 BOUNDARY, INTERFACE . . . . . . . . . . . . . . . . . . . . . . . . . . 629
30.2.1 DESCRIPTION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 629
30.2.2 PARAMETERS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 630
30.2.3 EXAMPLES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 630
30.3 COMMENT . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 630
CONTENTS 25
30.3.1 DESCRIPTION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 631

30.3.2 EXAMPLES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 631
30.4 CONTACT . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 631
30.4.1 DESCRIPTION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 631
30.4.2 PARAMETERS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 631
30.4.3 EXAMPLES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 631
30.4.4 SEE ALSO . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 632
30.5 DOMAIN, REGION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 632
30.5.1 DESCRIPTION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 632
30.5.2 PARAMETERS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 632
30.5.3 EXAMPLES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 632
30.5.4 SEE ALSO . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 633
30.6 DOPING . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 633
30.6.1 DESCRIPTION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 633
30.6.2 PARAMETERS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 636
30.6.3 EXAMPLES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 636
30.6.4 SEE ALSO . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 637
30.7 ELECTRODE . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 637
30.7.1 DESCRIPTION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 637
30.7.2 PARAMETERS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 638
30.7.3 EXAMPLES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 638
30.7.4 SEE ALSO . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 639
30.8 END . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 639
30.8.1 DESCRIPTION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 639
30.9 MATERIAL . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 639
30.9.1 DESCRIPTION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 639
30.9.2 PARAMETERS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 640
30.9.3 EXAMPLES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 640
30.9.4 SEE ALSO . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 640
30.10METHOD . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 641
30.10.1 DESCRIPTION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 641
30.10.2 Parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 641
30.10.3 Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 642
30.11Mobility . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 642
30.11.1 Description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 642
26 CONTENTS
30.11.2 Parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 643

30.11.3 Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 643
30.11.4 SEE ALSO . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 644
30.11.5 BUGS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 644
30.12MODELS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 644
30.12.1 DESCRIPTION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 644
30.12.2 Parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 644
30.12.3 Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 645
30.12.4 See also . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 645
30.12.5 Bugs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 645
30.13OPTIONS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 645
30.13.1 DESCRIPTION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 645
30.13.2 Parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 646
30.13.3 Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 646
30.13.4 See also . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 647
30.14OUTPUT . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 647
30.14.1 DESCRIPTION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 647
30.14.2 Parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 648
30.14.3 Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 648
30.14.4 SEE ALSO . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 649
30.15TITLE . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 649
30.15.1 DESCRIPTION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 649
30.15.2 EXAMPLES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 649
30.15.3 BUGS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 649
30.16X.MESH, Y.MESH . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 649
30.16.1 DESCRIPTION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 650
30.16.2 Parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 651
30.16.3 EXAMPLES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 651
30.16.4 SEE ALSO . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 651
30.17NUMD . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 652
30.17.1 DESCRIPTION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 652
30.17.2 Parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 653
30.17.3 EXAMPLES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 653
30.17.4 SEE ALSO . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 654
30.17.5 BUGS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 654
CONTENTS 27
30.18NBJT . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 654
30.18.1 DESCRIPTION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 654
30.18.2 Parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 655
30.18.3 EXAMPLES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 655
30.18.4 SEE ALSO . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 656
30.18.5 BUGS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 656
30.19NUMOS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 656
30.19.1 DESCRIPTION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 657
30.19.2 Parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 657
30.19.3 EXAMPLES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 658
30.19.4 SEE ALSO . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 658
30.202D contour plots . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 658
30.21Cider examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 658
IV Miscellaneous 659
31 Model and Device Parameters 661

31.1 Accessing internal device parameters . . . . . . . . . . . . . . . . . . . . . 661
31.2 Elementary Devices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 663
31.2.1 Resistor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 663
31.2.2 Capacitor - Fixed capacitor . . . . . . . . . . . . . . . . . . . . . . 665
31.2.3 Inductor - Fixed inductor . . . . . . . . . . . . . . . . . . . . . . . 666
31.2.4 Mutual - Mutual Inductor . . . . . . . . . . . . . . . . . . . . . . . 667
31.3 Voltage and current sources . . . . . . . . . . . . . . . . . . . . . . . . . . 668
31.3.1 Bxxxx - Arbitrary source (ASRC) . . . . . . . . . . . . . . . . . . . 668
31.3.2 Isource - Independent current source . . . . . . . . . . . . . . . . . 669
31.3.3 Vsource - Independent voltage source . . . . . . . . . . . . . . . . . 670
31.3.4 Fxxxx: Current-Controlled Current Source (CCCS) . . . . . . . . . 671
31.3.5 Hxxxx: Current-Controlled Voltage Source (CCVS) . . . . . . . . . 671
31.3.6 Gxxxx: Voltage-Controlled Current Source (VCCS) . . . . . . . . . 672
31.3.7 Exxxx: Voltage-Controlled Voltage Source (VCVS) . . . . . . . . . 672
31.4 Transmission Lines . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 673
31.4.1 CplLines - Simple Coupled Multiconductor Lines . . . . . . . . . . 673
31.4.2 LTRA - Lossy transmission line . . . . . . . . . . . . . . . . . . . . 674
31.4.3 Tranline - Lossless transmission line . . . . . . . . . . . . . . . . . . 675
28 CONTENTS
31.4.4 TransLine - Simple Lossy Transmission Line . . . . . . . . . . . . . 676

31.4.5 URC - Uniform R. C. line . . . . . . . . . . . . . . . . . . . . . . . 677
31.5 BJTs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 678
31.5.1 BJT - Bipolar Junction Transistor . . . . . . . . . . . . . . . . . . . 678
31.5.2 VBIC - Vertical Bipolar Inter-Company Model . . . . . . . . . . . . 681
31.6 MOSFETs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 685
31.6.1 MOS1 - Level 1 MOSFET model with Meyer capacitance model . . 685
31.6.5 MOS9 - Modified Level 3 MOSFET model . . . . . . . . . . . . . . 699
31.6.6 BSIM1 - Berkeley Short Channel IGFET Model . . . . . . . . . . . 703
31.6.7 BSIM2 - Berkeley Short Channel IGFET Model . . . . . . . . . . . 706
31.6.8 BSIM3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 710
31.6.9 BSIM4 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 711
32 Compilation notes 713

32.1 Ngspice Installation under Linux (and other ’UNIXes’) . . . . . . . . . . . 713
32.1.1 Prerequisites . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 713
32.1.2 Install from Git . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 713
32.1.3 Install from a tarball, e.g. from ngspice-42.tar.gz . . . . . . . . . . . 715
32.1.4 Compilation using an user defined directory tree for object files . . 716
32.1.5 ngspice as a shared library . . . . . . . . . . . . . . . . . . . . . . . 716
32.1.6 Relative paths for spinit and code models . . . . . . . . . . . . . . . 717
32.1.7 Installation on Red Hat or Oracle Linux (and similar, e.g. Centos) . 717
32.1.8 Advanced Install . . . . . . . . . . . . . . . . . . . . . . . . . . . . 717
32.1.9 Compilers and Options . . . . . . . . . . . . . . . . . . . . . . . . . 720
32.1.10 Compiling For Multiple Architectures . . . . . . . . . . . . . . . . . 720
32.1.11 Installation Names . . . . . . . . . . . . . . . . . . . . . . . . . . . 720
32.1.12 Optional Features . . . . . . . . . . . . . . . . . . . . . . . . . . . . 721
32.1.13 Specifying the System Type . . . . . . . . . . . . . . . . . . . . . . 721
32.1.14 Sharing Defaults . . . . . . . . . . . . . . . . . . . . . . . . . . . . 721
32.1.15 Operation Controls . . . . . . . . . . . . . . . . . . . . . . . . . . . 721
32.2 Ngspice Compilation under Windows OS . . . . . . . . . . . . . . . . . . . 722
32.2.1 Building ngspice with MS Visual Studio 2022 . . . . . . . . . . . . 722
CONTENTS 29
32.2.2 How to make ngspice with MINGW and MSYS2 . . . . . . . . . . 725

32.2.3 make ngspice with pure CYGWIN . . . . . . . . . . . . . . . . . . . 728
32.2.4 ngspice mingw or cygwin console executable w/o graphics . . . . . 728
32.2.5 ngspice for MS Windows, cross compiled from Linux . . . . . . . . 729
32.3 Ngspice Compilation under macOS . . . . . . . . . . . . . . . . . . . . . . 729
32.3.1 Prerequisites . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 729
32.3.2 Compiling ngspice . . . . . . . . . . . . . . . . . . . . . . . . . . . 730
32.3.3 Compiling ngspice shared library . . . . . . . . . . . . . . . . . . . 730
32.4 Reporting errors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 731
33 Copyrights and licenses 733

33.1 Documentation license . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 733
33.2 ngspice license . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 733
33.3 Some license details . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 733
33.3.1 CC-BY-SA . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 733
33.3.2 ‘Modified’ BSD license . . . . . . . . . . . . . . . . . . . . . . . . . 734
33.4 On the historical evolvement of the ngspice licenses . . . . . . . . . . . . . 735
33.4.1 XSPICE SOFTWARE (documentation) copyright . . . . . . . . . . 735
33.4.2 CIDER RESEARCH SOFTWARE AGREEMENT (superseded by
33.4.3) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 735
33.4.3 ‘Modified’ BSD license . . . . . . . . . . . . . . . . . . . . . . . . . 736
33.4.4 XSPICE . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 737
33.4.5 OSDI . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 737
33.4.6 tclspice, numparam . . . . . . . . . . . . . . . . . . . . . . . . . . . 737
33.4.7 Linking to GPLd libraries (e.g. readline, fftw, table.cm): . . . . . . 737
30 CONTENTS
Prefaces
Preface to the first edition

This manual has been assembled from different sources:
1. The spice3f5 manual,
2. the XSPICE user’s manual,
3. the CIDER user’s manual
and some original material needed to describe the new features and the newly implemented
models. This cut and paste approach, while not being orthodox, allowed ngspice to have
a full manual in a fraction of the time that writing a completely new text would have
required. The use of LaTex and LYX instead of TeXinfo, which was the original encoding
for the manual, further helped to reduce the writing effort and improved the quality of
the result, at the expense of an on-line version of the manual but, due to the complexity
of the software I hardly think that users will ever want to read an on-line text version.
In writing this text I followed the spice3f5 manual, both in the chapter sequence and
presentation of material, mostly because that was already the user manual of SPICE.
Ngspice is an open source software, users can download the source code, compile, and
run it. This manual has an entire chapter describing program compilation and available
options to help users in building ngspice (see Chapt. 32). The source package already
comes with all ‘safe’ options enabled by default, and activating the others can produce
unpredictable results and thus is recommended to expert users only. This is the first
ngspice manual and I have removed all the historical material that described the differences
between ngspice and spice3, since it was of no use for the user and not so useful for the
developer who can look for it in the Changelogs of in the revision control system.
I want to acknowledge the work done by Emmanuel Rouat and Arno W. Peters for
converting the original spice3f documentation to TEXinfo. Their effort gave ngspice users
the only available documentation that described the changes for many years. A good
source of ideas for this manual came from the on-line spice3f manual written by Charles
D.H. Williams (Spice3f5 User Guide), constantly updated and useful for its many insights.
As always, errors, omissions and unreadable phrases are only my fault.
Paolo Nenzi
Roma, March 24th 2001
31
32 CONTENTS
Indeed. At the end of the day, this is engineering, and one learns to live
within the limitations of the tools.
Kevin Aylward, Warden of the King’s Ale
Preface to the current edition (as of Nov 2023)

Due to the wealth of new material and options in ngspice the actual order of chapters has
been revised. Several new chapters have been added. The LYX text processor has allowed
adding internal cross references. The PDF format has become the standard format for
distribution of the manual. There is also a xhtml version available. Within each new
ngspice distribution a manual edition is provided reflecting the ngspice status at the time
of distribution. At the same time, located at ngspice manuals, the manual is constantly
updated. Every new ngspice feature should enter this manual as soon as it has been made
available in the Git source code master branch.
Holger Vogt
Mülheim, 2023
Acknowledgments
ngspice contributors
Spice3 and CIDER were originally written at The University of California at Berkeley
(USA).
XSPICE has been provided by Georgia Institute of Technology, Atlanta (USA).
Since then, there have been many people working on the software, most of them releasing
patches to the original code through the Internet.
The following people have contributed in some way:
Vera Albrecht,
Cecil Aswell,
Giles Atkinson,
Giles C. Billingsley,
Phil Barker,
Steven Borley,
Stuart Brorson,
Alessio Cacciatori,
Mansun Chan,
Wayne A. Christopher,
Al Davis,
Glao S. Dezai,
Jon Engelbert,
Daniele Foci,
Noah Friedman,
David A. Gates,
Alan Gillespie,
John Heidemann,
Marcel Hendrix,
Jeffrey M. Hsu,
JianHui Huang,
S. Hwang,
Chris Inbody,
Gordon M. Jacobs,
Min-Chie Jeng,
Beorn Johnson,
Stefan Jones,
33
34 CONTENTS
Kenneth H. Keller,
Francesco Lannutti,
Robert Larice,
Mathew Lew,
Robert Lindsell,
Weidong Liu,
Kartikeya Mayaram,
Richard D. McRoberts,
Manfred Metzger,
Jim Monte,
Wolfgang Muees,
Paolo Nenzi,
Gary W. Ng,
Hong June Park,
Stefano Perticaroli,
Arno Peters,
Serban-Mihai Popescu,
Georg Post,
Thomas L. Quarles,
Emmanuel Rouat,
Jean-Marc Routure,
Jaijeet S. Roychowdhury,
Lionel Sainte Cluque,
Takayasu Sakurai,
Amakawa Shuhei,
Kanwar Jit Singh,
Bill Swartz,
Hitoshi Tanaka,
Brian Taylor,
Steve Tell,
Andrew Tuckey,
Andreas Unger,
Holger Vogt,
Dietmar Warning,
Michael Widlok,
Charles D.H. Williams,
Antony Wilson,
and many others...

If someone helped in the development and has not been inserted in this list then this
omission was unintentional. If you feel you should be on this list then please write to
<[email protected]>. Do not be shy, we would like to make a list as
complete as possible.
Chapter 1
Introduction
Ngspice is a general-purpose circuit simulation program for nonlinear and linear analyses.
Circuits may contain resistors, capacitors, inductors, mutual inductors, independent or
dependent voltage and current sources, loss-less and lossy transmission lines, switches,
uniform distributed RC lines, and the five most common semiconductor devices: diodes,
BJTs, JFETs, MESFETs, and MOSFETs.
The most common way to use Ngspice is to start it from the OS command prompt, passing
the name of a netlist file: one containing the definition of a circuit. The largest part of
this manual is the description of such files. For a full description of starting options see
Chapter 16. Input files may also contain scripts written in Ngspice’s command language
(17). Interactive user interfaces are described in Chapter 18.
Some introductory remarks on how to use ngspice may be found in Chapter 21.
Ngspice is an update of Spice3f5, the last Berkeley’s release of Spice3 simulator family.
Ngspice is being developed to include new features to existing Spice3f5 and to fix its bugs.
Improving a complex software like a circuit simulator is a very hard task and, while some
improvements have been made, most of the work has been done on bug fixing and code
refactoring.
Ngspice has built-in models for the semiconductor devices, and the user need specify only
the pertinent model parameter values.
Ngspice supports mixed-level simulation and provides a direct link between technology pa-
rameters and circuit performance. A mixed-level circuit and device simulator can provide
greater simulation accuracy than a stand-alone circuit or device simulator by numerically
modeling the critical devices in a circuit. Compact models can be used for all other de-
vices. The mixed-level extensions to ngspice is CIDER, a mixed-level circuit and device
simulator integrated into ngspice code.
Ngspice supports mixed-signal simulation through the integration of XSPICE code. XSPICE
software, developed as an extension to Spice3C1 by GeorgiaTech, has been enhanced and
ported to ngspice to provide ‘board’ level and mixed-signal simulation.
The XSPICE extension enables pure digital simulation as well.
New devices can be added to ngspice by several means: behavioral B-, E- or G-sources,
the XSPICE code-model interface for C-like device coding, and the ADMS interface based
on Verilog-A and XML.
35
36 CHAPTER 1. INTRODUCTION
Finally, numerous small bugs have been discovered and fixed, and the program has been
ported to a wider variety of computing platforms.
1.1 Simulation Algorithms

Computer-based circuit simulation is often used as a tool by designers, test engineers,
and others who want to analyze the operation of a design without examining the physical
circuit. Simulation allows you to change quickly the parameters of many of the circuit ele-
ments to determine how they affect the circuit response. Often it is difficult or impossible
to change these parameters in a physical circuit.
However, to be practical, a simulator must execute in a reasonable amount of time.
The key to efficient execution is choosing the proper level of modeling abstraction for
a given problem. To support a given modeling abstraction, the simulator must provide
appropriate algorithms.
Historically, circuit simulators have supported either an analog simulation algorithm or a
digital simulation algorithm. Ngspice inherits the XSPICE framework and supports both
analog and digital algorithms and is a ‘mixed-mode’ simulator.
1.1.1 Analog Simulation
Analog simulation focuses on the linear and non-linear behavior of a circuit over a con-
tinuous time or frequency interval. The circuit response is obtained by iteratively solving
Kirchhoff’s Laws for the circuit at time steps selected to ensure the solution has converged
to a stable value and that numerical approximations of integrations are sufficiently accu-
rate. Since Kirchhoff’s laws form a set of simultaneous equations, the simulator operates
by solving a matrix of equations at each time point. This matrix processing generally
results in slower simulation times when compared to digital circuit simulators.
The response of a circuit is a function of the applied sources. Ngspice offers a variety
of source types including DC, sine-wave, and pulse. In addition to specifying sources,
the user must define the type of simulation to be run. This is termed the ‘mode of
analysis’. Analysis modes include DC analysis, AC analysis, and transient analysis. For
DC analysis, the time-varying behavior of reactive elements is neglected and the simulator
calculates the DC solution of the circuit. Swept DC analysis may also be accomplished
with ngspice. This is simply the repeated application of DC analysis over a range of
DC levels for the input sources. For AC analysis, the simulator determines the response
of the circuit, including reactive elements to small-signal sinusoidal inputs over a range
of frequencies. The simulator output in this case includes amplitudes and phases as
a function of frequency. For transient analysis, the circuit response, including reactive
elements, is analyzed to calculate the behavior of the circuit as a function of time.
1.1.2 Matrix solvers
Since version 42 ngspice offers two matrix solvers. Already spice3f5 offerd the solver
Sparse 1.3, which has proven to be robust for all simulation tasks [26]. It is especially
1.1. SIMULATION ALGORITHMS 37
suited for simulating behavioral models. Optionally, to speed up the simulation of large
circuits with thousands of transistors, the KLU matrix solver [27, 28] may be selected (see
chapter 15.1.1).
1.1.3 Device Models for Analog Simulation
There are three models for bipolar junction transistors, all based on the integral-charge
model of Gummel and Poon; however, if the Gummel-Poon parameters are not specified,
the basic model (BJT) reduces to the simpler Ebers-Moll model. In either case and in
either model, charge storage effects, ohmic resistances, and a current-dependent output
conductance may be included. The second bipolar model BJT2 adds dc current compu-
tation in the substrate diode. The third model (VBIC) contains further enhancements
for advanced bipolar devices.
The semiconductor diode model can be used for either junction diodes or Schottky barrier
diodes. There are two models for JFET: the first (JFET) is based on the model of
Shichman and Hodges, the second (JFET2) is based on the Parker-Skellern model. All
the original six MOSFET models are implemented: MOS1 is described by a square-law
I-V characteristic, MOS2 [28] is an analytical model, while MOS3 [28] is a semi-empirical
model; MOS6 [2] is a simple analytic model accurate in the short channel region; MOS9,
is a slightly modified Level 3 MOSFET model - not to confuse with Philips level 9; BSIM
1 [3, 4]; BSIM2 [5] are the old BSIM (Berkeley Short-channel IGFET Model) models.
MOS2, MOS3, and BSIM include second-order effects such as channel-length modulation,
subthreshold conduction, scattering-limited velocity saturation, small-size effects, and
charge controlled capacitances. The recent MOS models for submicron devices are the
BSIM3 (Berkeley BSIM3 web page) and BSIM4 (Berkeley BSIM4 web page) models.
Silicon-on-insulator MOS transistors are described by the SOI models from the BSIMSOI
family (Berkeley BSIMSOI web page) and the STAG [18] model. There is some support
for a couple of HFET models and one model for MESA devices. Verilog-A models are
made available via the OpenVAF/OSDI interface (see chapter 13).
1.1.4 Digital Simulation
Digital circuit simulation differs from analog circuit simulation in several respects. A
primary difference is that a solution of Kirchhoff’s laws is not required. Instead, the
simulator must only determine whether a change in the logic state of a node has occurred
and propagate this change to connected elements. Such a change is called an ‘event’.
When an event occurs, the simulator examines only those circuit elements that are affected
by the event. As a result, matrix analysis is not required in digital simulators. By
comparison, analog simulators must iteratively solve for the behavior of the entire circuit
because of the forward and reverse transmission properties of analog components. This
difference results in a considerable computational advantage for digital circuit simulators,
which is reflected in the significantly greater speed of digital simulations.
1.1.5 Mixed-Signal Simulation

Modern circuits often contain a mix of analog and digital circuits. To simulate such circuits
efficiently and accurately, a mix of analog and digital simulation techniques is required.
When analog simulation algorithms are combined with digital simulation algorithms, the
result is termed ‘mixed-mode simulation’.
Two basic methods of implementing mixed-mode simulation used in practice are the ‘na-
tive mode’ and ‘glued mode’ approaches. Native mode simulators implement both an
analog algorithm and a digital algorithm in the same executable. Glued mode simulators
actually use two simulators, one of which is analog and the other digital. This type of
simulator must define an input/output protocol so that the two executables can com-
municate with each other effectively. The communication constraints tend to reduce the
speed, and sometimes the accuracy, of the complete simulator. On the other hand, the
use of a glued mode simulator allows the component models developed for the separate
executables to be used without modification.
Ngspice is a native mode simulator providing both analog and event-based simulation
in the same executable. The underlying algorithms of ngspice (coming from XSPICE
and its Code Model Subsystem) allow use of all the standard SPICE models, provide a
pre-defined collection of the most common analog and digital functions, and provide an
extensible base on which to build additional models.
1.1.5.1 User-Defined Nodes
Ngspice supports creation of ‘User-Defined Node’ types. User-Defined Node types allow
you to specify nodes that propagate data other than voltages, currents, and digital states.
Like digital nodes, User-Defined Nodes use event-driven simulation, but the state value
may be an arbitrary data type. A simple example application of User-Defined Nodes is
the simulation of a digital signal processing filter algorithm. In this application, each
node could assume a real or integer value. More complex applications may define types
that involve complex data such as digital data vectors or even non-electronic data.
Ngspice digital simulation is actually implemented as a special case of this User-Defined
Node capability where the digital state is defined by a data structure that holds a Boolean
logic state and a strength value.
1.1.6 Mixed-Level Simulation (Electronic and TCAD)

Ngspice implements mixed-level simulation through the merging of its code with CIDER
(details see Chapt. 30).
CIDER is a mixed-level circuit and device simulator that provides a direct link between
technology parameters and circuit performance. A mixed-level circuit and device simula-
tor can provide greater simulation accuracy than a stand-alone circuit or device simulator
by numerically modeling the critical devices in a circuit. Compact models can be used
for noncritical devices.
CIDER couples ngspice to a internal C-based device simulator, thus providing circuit
analyses, compact models for semiconductor devices, and an interactive user interface.
1.2. SUPPORTED ANALYSES 39
CIDER provides accurate, one- and two-dimensional numerical device models based on
the solution of Poisson’s equation, and the electron and hole current-continuity equa-
tions. CIDER incorporates many of the same basic physical models found in the the
Stanford two-dimensional device simulator PISCES [PINT85]. Input to CIDER consists
of a SPICE-like description of the circuit and its compact models, and PISCES-like de-
scriptions of the structures of numerically modeled devices. As a result, CIDER should
seem familiar to designers already accustomed to these two tools.
The CIDER input format has great flexibility and allows increased access to physical
model parameters. New physical models have been added to allow simulation of state-
of-the-art devices. These include transverse field mobility degradation [GATE90] that
is important in scaled-down MOSFETs and a polysilicon model for poly-emitter bipolar
transistors. Temperature dependence has been included for most physical models over
the range from -50°C to 150°C. The numerical models can be used to simulate all the
basic types of semiconductor devices: resistors, MOS capacitors, diodes, BJTs, JFETs
and MOSFETs. BJTs and JFETs can be modeled with or without a substrate contact.
Support has been added for the management of device internal states. Post-processing
of device states can be performed using the control language user interface of ngspice.
Previously computed states can be loaded into the program to provide accurate initial
guesses for subsequent analyses.
Details of the basic semiconductor equations and the physical models used by CIDER are
not provided in this manual. Unfortunately, no other single source exists that describes
all of the relevant background material. Comprehensive reviews of device simulation can
be found in [PINT90] and the book [SELB84]. CODECS (predecessor to CIDER) and its
inversion-layer mobility model are described in [MAYA88] and [LGATE90], respectively.
PISCES and its models are described in [PINT85]. Temperature dependencies for the
PISCES models used by CIDER are available in [SOLL90].
For Linux users the cooperation of the TCAD software GSS with ngspice might be of
interest, see https://ngspice.sourceforge.io/gss.html. This project is no longer maintained
however, but has moved into the Genius simulator, still available as open source cogenda
genius.
1.2 Supported Analyses

The ngspice simulator supports the following different types of analysis:
1. DC Analysis (Operating Point and DC Sweep)
2. AC Small-Signal Analysis
3. Transient Analysis
4. Pole-Zero Analysis
5. Small-Signal Distortion Analysis
6. Sensitivity Analysis
7. Noise Analysis
Applications that are exclusively analog can make use of all analysis modes with the
exception of Code Model subsystem that do not implements Pole-Zero, Distortion, Sensi-
tivity and Noise analyses. Event-driven applications that include digital and User-Defined
Node types may make use of DC (operating point and DC sweep) and Transient only.
In order to understand the relationship between the different analyses and the two un-
derlying simulation algorithms of ngspice, it is important to understand what is meant
by each analysis type. This is detailed below.
1.2.1 DC Analysis
The DC analysis portion of ngspice determines the dc operating point of the circuit with
inductors shorted and capacitors opened. DC analysis options are specified on the .DC,
.TF, and .OP control lines.
DC analysis does not consider any time dependence on any of the sources within the sys-
tem description. The simulator algorithm subdivides the circuit into those portions that
require the analog simulator algorithm and those that require the event-driven algorithm.
Each subsystem block is then iterated to solution, with the interfaces between analog
nodes and event-driven nodes iterated for consistency across the entire system.
Once stable values are obtained for all nodes in the system, the analysis halts and the
results may be displayed or printed out as you request them.
A dc analysis is automatically performed prior to a transient analysis to determine the
transient initial conditions, and prior to an ac small-signal analysis to determine the
linearized, small-signal models for nonlinear devices. If requested, the DC small-signal
value of a transfer function (ratio of output variable to input source), input resistance,
and output resistance is also computed as a part of the DC solution. DC analysis can also
be used to generate DC transfer curves: a specified independent voltage, current source,
resistor or temperature is stepped over a user-specified range and the DC output variables
are stored for each sequential source value.
1.2.2 AC Small-Signal Analysis
AC analysis is limited to analog nodes and represents the small signal, sinusoidal solution
of the analog system described at a particular frequency or set of frequencies. This
analysis is similar to the DC analysis in that it represents the steady-state behavior of
the described system with a single input node at a given set of stimulus frequencies.
The program first computes the dc operating point of the circuit and determines linearized,
small-signal models for all of the nonlinear devices in the circuit. The resultant linear
circuit is then analyzed over a user-specified range of frequencies. The desired output of
an ac small-signal analysis is usually a transfer function (voltage gain, transimpedance,
etc). If the circuit has only one ac input, it is convenient to set that input to unity and
zero phase, so that output variables have the same value as the transfer function of the
output variable with respect to the input.
1.2. SUPPORTED ANALYSES 41
1.2.3 Transient Analysis
Transient analysis is an extension of DC analysis to the time domain. A transient analysis

first obtains a DC solution to provide a point of departure for simulating time-varying
behavior. Once the DC solution is obtained, the time-dependent aspects of the system are
reintroduced, and the two simulator algorithms incrementally solve for the time varying
behavior of the entire system. Inconsistencies in node values are resolved by the two
simulation algorithms such that the time-dependent waveforms created by the analysis are
consistent across the entire simulated time interval. Resulting time-varying descriptions
of node behavior for the specified time interval are accessible to you.
All sources that are not time dependent (for example, power supplies) are set to their dc
value. The transient time interval is specified on a .TRAN control line.
1.2.4 Pole-Zero Analysis
Pole-zero analysis in ngspice computes the poles and/or zeros in the small-signal ac trans-
fer function. Ngspice first computes the dc operating point and then determines the lin-
earized, small-signal models for all the nonlinear devices in the circuit. The small-signal
circuit model is then used to find the poles and zeros of the transfer function. Two types
of transfer functions are allowed: one of the form (output voltage)/(input voltage) and
the other of the form (output voltage)/(input current). These two types of transfer func-
tions cover all the cases and one can find the poles/zeros of functions like input/output
impedance and voltage gain. The input and output ports are specified as two pairs of
nodes. The pole-zero analysis works with resistors, capacitors, inductors, linear-controlled
sources, independent sources, BJTs, MOSFETs, JFETs and diodes. Transmission lines
are not supported.
The method used in the analysis is a sub-optimal numerical search. For large circuits it
may take a considerable time or fail to find all poles and zeros. Please note, that for some
circuits, the method becomes “lost” and may find an excessive number of poles or zeros.
1.2.5 Small-Signal Distortion Analysis
Distortion analysis in ngspice computes steady-state harmonic and intermodulation prod-

ucts for small input signal magnitudes. If signals of a single frequency are specified as
the input to the circuit, the complex values of the second and third harmonics are deter-
mined at every point in the circuit. If there are signals of two frequencies input to the
circuit, the analysis finds out the complex values of the circuit variables at the sum and
difference of the input frequencies, and at the difference of the smaller frequency from the
second harmonic of the larger frequency. Distortion analysis is supported for the following
nonlinear devices:
• Diodes (DIO),
• BJT,
• JFET (level 1),

• MOSFETs (levels 1, 2, 3, 9, and BSIM1),

• MESFET (level 1).
All linear devices are automatically supported by distortion analysis. If there are switches
present in the circuit, the analysis continues to be accurate provided the switches do not
change state under the small excitations used for distortion calculations.
If a device model does not support direct small signal distortion analysis, please use the
Fourier of FFT statements and evaluate the output per scripting.
1.2.6 Sensitivity Analysis

Ngspice can calculate either the DC operating-point sensitivity or the AC small-signal
sensitivity of an output variable with respect to all circuit variables, including model
parameters. Ngspice calculates the difference in an output variable (either a node voltage
or a branch current) by perturbing each parameter of each device independently. Since the
method is a numerical approximation, the results may demonstrate second order effects
in highly sensitive parameters, or may fail to show very low but non-zero sensitivity.
Since each variable is perturbed by a small fraction of its value, zero-valued parameters
are not analyzed, reducing what is usually a very large amount of data.
1.2.7 Noise Analysis

Noise analysis in ngspice measures the device-generated noise for a given circuit. When
provided with an input source and an output port, the analysis calculates the noise con-
tributions of each device, and each noise generator within each device, as measured as a
voltage at the output port. Noise analysis also calculates the equivalent input noise of
the circuit, based on the output noise. This is done for every frequency point in a spec-
ified range - the calculated value of the noise corresponds to the spectral density of the
circuit variable viewed as a stationary Gaussian stochastic process. After calculating the
spectral densities, noise analysis integrates these values over the specified frequency range
to arrive at the total noise voltage and current over this frequency range. The calculated
values correspond to the variance of the circuit variables viewed as stationary Gaussian
processes.
1.2.8 Periodic Steady State Analysis

Experimental code.
PSS is a radio frequency periodical large-signal dedicated analysis. The implementation
is based on a time domain shooting method that make use of transient analysis. As
it is in early development stage, PSS performs analysis only on autonomous circuits,
meaning that it is able to predict fundamental frequency and (harmonic) amplitude(s)
for oscillators, VCOs, etc.. The algorithm is based on a search of the minimum error
vector defined as the difference of RHS vectors between two occurrences of an estimated
period. Convergence is reached when the mean of this error vector decreases below a given
threshold parameter. Results of PSS are the basis of periodical large-signal analyses like
PAC or PNoise.
1.3. ANALYSIS AT DIFFERENT TEMPERATURES 43
1.3 Analysis at Different Temperatures
1.3.1 Introduction
Temperature, in ngspice, is a property associated to the entire circuit, rather than an

analysis option. Circuit temperature has a default (nominal) value of 27°C (300.15 K)
that can be changed using the TEMP option in an .option control line (see 15.1.1) or by
the .TEMP line (see 2.12), which has precedence over the .option TEMP line. All analyses
are, thus, performed at circuit temperature, and if you want to simulate circuit behavior
at different temperatures you should prepare a netlist for each temperature.
All input data for ngspice is assumed to have been measured at the circuit nominal
temperature. This value can further be overridden for any device that models temperature
effects by specifying the TNOM parameter on the .model itself. Individual instances may
further override the circuit temperature through the specification of TEMP and DTEMP
parameters on the instance. The two options are not independent even if you can specify
both on the instance line, the TEMP option overrides DTEMP. The algorithm to compute
instance temperature is described below:
Algorithm 1.1 Instance temperature computation

IF TEMP is specified THEN
instance_temperature = TEMP
ELSE
instance_temperature = circuit_temperature + DTEMP
END IF
Temperature dependent support is provided for all devices except voltage and current
sources (either independent and controlled) and BSIM models. BSIM MOSFETs have an
alternate temperature dependency scheme that adjusts all of the model parameters before
input to ngspice.
For details of the BSIM temperature adjustment, see [6] and [7]. Temperature appears
explicitly in the exponential terms of the BJT and diode model equations. In addition,
saturation currents have a built-in temperature dependence. The temperature dependence
of the saturation current in the BJT models is determined by:
XT I
T1 Eg q (T1 − T0 )
IS (T1 ) = IS (T0 ) exp (1.1)
T0 k (T1 T0 )
where k is Boltzmann’s constant, q is the electronic charge, Eg is the energy gap model
parameter, and XT I is the saturation current temperature exponent (also a model pa-
rameter, and usually equal to 3).
The temperature dependence of forward and reverse beta is according to the formula:
XT B
T1
B (T1 ) = B (T0 ) (1.2)
T0
where T0 and T1 are in degrees Kelvin, and XT B is a user-supplied model parameter.

Temperature effects on beta are carried out by appropriate adjustment to the values of
BF , ISE , BR , and ISC (SPICE model parameters BF, ISE, BR, and ISC, respectively).
Temperature dependence of the saturation current in the junction diode model is deter-
mined by:
XT I
T1 N Eg q (T1 − T0 )
IS (T1 ) = IS (T0 ) exp (1.3)
T0 N k (T1 T0 )
where N is the emission coefficient model parameter, and the other symbols have the
same meaning as above. Note that for Schottky barrier diodes, the value of the saturation
current temperature exponent, XT I, is usually 2. Temperature appears explicitly in the
value of junction potential, U (in Ngspice PHI), for all the device models.
The temperature dependence is determined by:

kT Na Nd
U (T ) = ln (1.4)
q Ni (T )2
where k is Boltzmann’s constant, q is the electronic charge, Na is the acceptor impurity

density, Nd is the donor impurity density, Ni is the intrinsic carrier concentration, and Eg
is the energy gap. Temperature appears explicitly in the value of surface mobility, M0 (or
U0 ), for the MOSFET model.
The temperature dependence is determined by:
M0 (T0 )
M0 (T ) = 1.5 (1.5)
T
T0
The effects of temperature on resistors, capacitor and inductors is modeled by the formula:

R (T ) = R (T0 ) 1 + T C1 (T − T0 ) + T C2 (T − T0 )2 (1.6)
where T is the circuit temperature, T0 is the nominal temperature, and T C1 and T C2 are
the first and second order temperature coefficients.
1.3.2 Controlling the temperature
The default temperature is set to 27 °C.
.temp 40
will set the overall temperature to 40 °C (2.12). The command
.options temp=60
1.4. CONVERGENCE 45
will set the overall temperature to 60 °C (15.1.1). Both commands are equivalent, however
.temp will override .options temp.
The temperature of an individual device may be determined by the instance parameters
temp or dtemp.
M1 d g s b MOSN temp=35
will set the temperature of the specific MOS device to 35 °C.
M2 d g s b MOSN dtemp=20
will set the temperature of device M2 at a delta of 20° above the overall temperature.
The temperatures thus set are static throughout the simulation. It is possible, however,
to sweep the temperature by a command like
.dc temp 25 49 2
starting at 25 °C, stopping at 49 °C with a step of 2° (see 15.3.2).

The current overall temperature may be assessed by the variable TEMPER, which can be
used as part of an equation in B sources (5.1.2) or behavioral E, G, R, L, C sources (e.g.
5.2). A typical example may look like
Bt1 1 2 V=’5 + TEMPER*TEMPER’
The nominal temperature, a reference temperature where device model parameters have
been measured, is called tnom.
.options tnom=25
will set the nominal temperature for all devices to 25 °C (15.1.1). Tnom sometimes may
be set as a model parameter in a .model line (3.2.2), depending on the specific class of
devices and its model parameter set.
1.4 Convergence
Ngspice uses the Newton-Raphson algorithm to solve nonlinear equations arising from
circuit description. The NR algorithm is interactive and terminates when both of the
following conditions hold:
1. The nonlinear branch currents converge to within a tolerance of 0.1% or 1 picoamp

(1.0e-12 Amp), whichever is larger.
2. The node voltages converge to within a tolerance of 0.1% or 1 microvolt (1.0e-6

Volt), whichever is larger.
1.4.1 Voltage convergence criterion

The algorithm has reached convergence when the difference between the last iteration k
and the current one (k + 1)
vn(k+1) − vn(k) ≤ RELTOL vnmax + VNTOL, (1.7)

where

vnmax = max vn(k+1) , vn(k) . (1.8)
The RELTOL (RELative TOLerance) parameter, which default value is 10−3 , specifies
how small the solution update must be, relative to the node voltage, to consider the
solution to have converged. The VNTOL (absolute convergence) parameter, which has 1µV
as default value, becomes important when node voltages have near zero values. The
relative parameter alone, in such case, would need too strict tolerances, perhaps lower
than computer round-off error, and thus convergence would never be achieved. VNTOL
forces the algorithm to consider as converged any node whose solution update is lower
than its value.
1.4.2 Current convergence criterion

Ngspice checks the convergence on the non-linear functions that describe the non-linear
branches in circuit elements. In semiconductor devices the functions defines currents
through the device and thus the name of the criterion.
Ngspice computes the difference between the value of the nonlinear function computed
for the last voltage and the linear approximation of the same current computed with the
actual voltage
\(k+1) (k)
ibranch − ibranch ≤ RELTOL ibrmax + ABSTOL, (1.9)
where

\
(k+1) (k)
ibrmax = max ibranch , ibranch . (1.10)
In the two expressions above, the i\

branch indicates the linear approximation of the current.
1.4.3 Convergence failure

Although the algorithm used in ngspice has been found to be very reliable, in some cases
it fails to converge to a solution. When this failure occurs, the program terminates the
job. Failure to converge in dc analysis is usually due to an error in specifying circuit
connections, element values, or model parameter values. Regenerative switching circuits
or circuits with positive feedback probably will not converge in the dc analysis unless the
OFF option is used for some of the devices in the feedback path, .nodeset control line is
used to force the circuit to converge to the desired state.
Chapter 2
Circuit Description
2.1 General Structure and Conventions
2.1.1 Input file structure
The circuit to be analyzed is described to ngspice by a set of element instance lines, which
define the circuit topology and element instance values, and a set of control lines, which
define the model parameters and the run controls. All lines are assembled in an input file
to be read by ngspice. Two lines are essential:
• The first line in the input file must be the title, which is the only comment line that
does not need any special character in the first place.
• The last line must be .end, plus a newline delimiter.
The order of the remaining lines is alomost arbitrary (except, of course, that continuation
lines must immediately follow the line being continued, .subckt ... .ends, .if ... .endif,
or .control ... .endc have to enclose their specific lines). Leading white spaces in a
line are ignored, as well as empty lines.
The lines described in sections 2.1 to 2.12 are typically used in the core of the input file,
outside of a .control section (see 16.4.3). An exception is the .include includefile
line (2.7) that may be placed anywhere in the input file. The contents of includefile
will be inserted exactly in place of the .include line.
2.1.2 Syntax check
A very preliminary syntax check has been added to the input parser.
2.1.2.1 Valid utf-8 characters
The input file will be scanned for valid utf-8 characters. If non-valid characters are found,
reading the input is stopped.
47
48 CHAPTER 2. CIRCUIT DESCRIPTION
2.1.2.2 Special characters leading a line
If the first character in a netlist or .control line is one of =[]?()&%$§\"!:, then ngspice
replaces it by ’*’ and issues a warning. Command set strict_errorhandling will force
ngspice to exit.
2.1.2.3 Dot command couple completion
Check for .control ... .endc, .subckt ... .ends, .if ... .endif.
2.1.3 Circuit elements (device instances)
Each element in the circuit is a device instance specified by an instance line that con-
tains:
• the element instance name,
• the circuit nodes to which the element is connected,
• and the values of the parameters that determine the electrical characteristics of the
element.
The first letter of the element instance name specifies the element type. The format
for the ngspice element types is given in the following manual chapters. In the rest of
the manual, the strings XXXXXXX, YYYYYYY, and ZZZZZZZ denote arbitrary alphanumeric
strings.
For example, a resistor instance name must begin with the letter R and can contain one or
more characters. Hence, R, R1, RSE, ROUT, and R3AC2ZY are valid resistor names. Details
of each type of device are supplied in a following section 3. Table 2.1 lists the element
types available in ngspice, sorted by their first letter.
2.1. GENERAL STRUCTURE AND CONVENTIONS 49
First letter Element description Comments, links

12
analog (12.2)
A XSPICE code model
digital (12.4)
mixed signal (12.3)
B Behavioral (arbitrary) source 5.1
C Capacitor 3.3.6
D Diode 7
linear (4.2.2),
E Voltage-controlled voltage source (VCVS)
non-linear (5.2)
F Current-controlled current source (CCCs) linear (4.2.3)
linear (4.2.1),
G Voltage-controlled current source (VCCS)
non-linear (5.3)
H Current-controlled voltage source (CCVS) linear (4.2.4)
I Current source 4.1
J Junction field effect transistor (JFET) 9
K Coupled (Mutual) Inductors 3.3.12
L Inductor 3.3.10
11
M Metal oxide field effect transistor (MOSFET) BSIM3 (11.2.10)
BSIM4 (11.2.11)
N Numerical device for GSS 14
O Lossy transmission line 6.2
P Coupled multiconductor line (CPL) 6.4.2
Q Bipolar junction transistor (BJT) 8
R Resistor 3.3.1
S Switch (voltage-controlled) 3.3.15
T Lossless transmission line 6.1
U Uniformly distributed RC line 6.3*
U Basic digital building blocks using XSPICE 14*
V Voltage source 4.1
W Switch (current-controlled) 3.3.15
X Subcircuit 2.5.3
Y Single lossy transmission line (TXL) 6.4.1
Z Metal semiconductor field effect transistor (MESFET) 10
Table 2.1: ngspice element types
*) For a disambiguation see chapter 14.1.3.
2.1.4 Some naming conventions

2.1.4.1 Lines
Fields on a line are separated by one or more blanks, a comma, an equal (=) sign, or a
left or right parenthesis; extra spaces are ignored. A line may be continued by entering
a ‘+’ (plus) in column 1 of the following line; ngspice continues reading beginning with
column 2. Lines may also be continued in Unix shell style, when the final two characters
are backslashes. A device name field must begin with a letter (A through Z) and cannot
contain any delimiters.
2.1.4.2 Numbers
A number field may be an integer field (12, -44), a floating point field (3.14159), either
an integer or floating point number followed by an integer exponent (1e-14, 2.65e3), or
either an integer or a floating point number followed by one of the following scale factors:
Suffix Name Factor
T Tera 1012
G Giga 109
Meg Mega 106
K Kilo 103
mil Mil 25.4 × 10−6
m milli 10−3
u micro 10−6
n nano 10−9
p pico 10−12
f femto 10−15
a atto 10−18
Table 2.2: Ngspice scale factors
2.1.4.3 Letters following a number
Letters immediately following a number that are not scale factors are ignored, and letters
immediately following a scale factor are ignored. Hence, 10, 10V, 10Volts, and 10Hz all
represent the same number, and M, MA, MSec, and MMhos all represent the same scale
factor. Note that 1000, 1000.0, 1000Hz, 1e3, 1.0e3, 1kHz, and 1k all represent the same
number. Note that ‘M’ or ‘m’ denote ‘milli’, i.e. 10−3 . Suffix meg has to be used for
106 . If compatibility mode LT (16.14.6) is set, ngspice will accept the RKM notation for
entering resistance or capacitance values, e.g. 2K7 or 100R.
2.1.4.4 Node names
Node names may be arbitrary character strings (exceptions see below) and are case in-
sensitive, if ngspice is used in batch mode (16.4.1). If in interactive (16.4.2) or control
(16.4.3) mode, node names may either be plain numbers or arbitrary character strings,
not starting with a number. The following characters = % ( ) , [ ] < > ~ are not allowed
in a node name, especially when XSPICE code models are used (they have their special
meanings then and act as string delimiters).
2.1.4.5 Ground node
The ground node must be named ‘0’ (zero). For compatibility reason gnd is accepted as
ground node, and will internally be treated as a global node and be converted to ‘0’. If
2.2. DOT COMMANDS 51
this is not feasible, you may switch the conversion off by setting set no_auto_gnd in one
of the configuration files spinit or .spiceinit. Each circuit has to have a ground node (gnd
or 0)! Note the difference in ngspice where the nodes are treated as character strings
and not evaluated as numbers, thus ‘0’ and 00 are distinct nodes in ngspice but not in
SPICE2.
2.1.5 Topological constraints

Ngspice requires that the following topological constraints are satisfied:
• The circuit cannot contain a loop of voltage sources and/or inductors and cannot
contain a cut-set of current sources and/or capacitors.
• Each node in the circuit must have a dc path to ground.
• Every node must have at least two connections except for transmission line nodes
(to permit unterminated transmission lines) and MOSFET substrate nodes (which
have two internal connections anyway).
2.2 Dot commands

This section summarizes all dot commands available in ngspice, with links to their detailed
presentation, in alphabetical order. Control section (or interactive) commands are listed
and explained in chapter 17.5.
.AC start an ac simulation (15.3.1).
.CONTROL start a .control section (16.4.3).
.CSPARAM define parameter(s) made available in a control section (2.11).
.DC start a dc simulation (15.3.2).
.DISTO start a distortion analysis simulation (15.3.3).
.ELSE conditional branching in the netlist (2.13).
.ELSEIF conditional branching in the netlist (2.13).
.END end of the netlist (2.3.2).
.ENDC end of the .control section (16.4.3).
.ENDIF conditional branching in the netlist (2.13).
.ENDS end of subcircuit definition (2.5.2).
.FOUR Fourier analysis of transient simulation output (15.6.4).
.FUNC define a function (2.10).

.GLOBAL define global nodes (2.6).
.IC set initial conditions (15.2.2).
.IF conditional branching in the netlist (2.13).
.INCLUDE include part of the netlist (2.7).
.LIB include a library (2.8).
.MEAS measurements during the simulation (15.4).
.MODEL list of device model parameters (2.4).
.NODESET set initial conditions (15.2.1).
.NOISE start a noise simulation (15.3.4).
.OP start an operating point simulation (15.3.5).
.OPTIONS set simulator options (15.1).
.PARAM define parameter(s) (2.9).
.PLOT printer plot during batch simulation (15.6.3).
.PRINT tabular listing during batch simulation (15.6.2).
.PROBE save device currents, voltages and differential voltages (15.6.5).
.PSS start a periodic steady state analysis (15.3.12).
.PZ start a pole-zero analysis simulation (15.3.6).
.SAVE name simulation result vectors to be saved (15.6.1).
.SENS start a sensitivity analysis (15.3.7).
.SP S parameter analysis (15.3.8).
.SUBCKT start of subcircuit definitions (2.5).
.TEMP set the ciruit temperature (2.12).
.TF start a transfer function analysis (15.3.9).
.TITLE title of the netlist (2.3.1).
.TRAN start a transient simulation (15.3.10).
.WIDTH width of printer plot (15.6.7).

2.3. BASIC LINES 53
2.3 Basic lines
2.3.1 .TITLE line

Examples:
POWER AMPLIFIER CIRCUIT

* additional lines following
*...
Test of CAM cell

*...
The title line must be the first in the input file. Its contents are printed verbatim as the
heading for each section of output.
As an alternative, you may place a .TITLE <any title> line anywhere in your input
deck. The first line of your input deck will be overridden by the contents of this line
following the .TITLE statement.
.TITLE line example:
** ****************************
*...
. TITLE Test of CAM cell
*...
will internally be replaced by

Internal input deck:
Test of CAM cell

*...
* TITLE Test of CAM cell
*...
2.3.2 .END Line

Examples:
. end
The .end line must always be the last in the input file. Note that the period is an integral
part of the name.
2.3.3 Comments
General Form:
* < any comment >
Examples:
* RF =1 K Gain should be 100

* Check open - loop gain and phase margin
The asterisk in the first column indicates that this line is a comment line. Comment lines
may be placed anywhere in the circuit description.
2.3.4 End-of-line comments

General Form:
< any command > $ < any comment >

< any command > ; < any comment >
Examples:
RF2 =1 K $ Gain should be 100

C1 =10 p ; Check open - loop gain and phase margin
. param n1 =1 // new value
ngspice supports comments that begin with double characters ‘$ ’ (dollar plus space) or
‘//’. For readability you should precede each comment character with a space. ngspice
will accept the single character ‘$’.
Please note that the ‘$’ character is not a valid end-of-line comment delimiter, if the
PSPICE compatibility mode (16.14.5) has been chosen. Then ’$’ becomes an ordinary
character.
2.3.5 Continuation lines

General Form:
< any command >

+ < continuation of any command > ; some comment
+ < further continuation of any command >
2.4. .MODEL DEVICE MODELS 55
If input lines get overly long, they may be split into two or more lines (e.g. for better
readability). Internally they will be merged into a single line. Each follow-up line starts
with character ’+’ plus additional space. Follow-up lines have to follow immediately after
each other. End-of-line comments will be ignored. Lines may also be continued by ending
the line with two backslashes, as used in Unix shells. The following lines do not allow
using continuation lines: .title, .lib, and .include.
2.4 .MODEL Device Models

General form:
. model mname type ( pname1 = pval1 pname2 = pval2 ... )
Examples:
. model MOD1 npn ( bf =50 is =1 e -13 vbf =50)
Most simple circuit elements typically require only a few parameter values. However,
some devices (semiconductor devices in particular) that are included in ngspice require
many parameter values. Often, many devices in a circuit are defined by the same set of
device model parameters. For these reasons, a set of device model parameters is defined
on a separate .model line and assigned a unique model name. The device element lines
in ngspice then refer to the model name.
For these more complex device types, each device element line contains the device name,
the nodes the device is connected to, and the device model name. In addition, other
optional parameters may be specified for some devices: geometric factors and an initial
condition (see the following section on Transistors (8 to 11) and Diodes (7) for more
details). mname in the above is the model name, and type is one of the following fifteen
types:
Parameter values are defined by appending the parameter name followed by an equal sign
and the parameter value. Model parameters that are not given a value are assigned the
default values given below for each model type. Models are listed in the section on each
device along with the description of device element lines. Model parameters and their
default values are given in Chapt. 31.
2.5 .SUBCKT Subcircuits

Subcircuits consisting of ngspice elements can be defined and used similarly to device
models. Subcircuits are the way ngspice implements hierarchical modeling and make
circuits with repeated sections easier to represent. During parsing of a SPICE deck, each
subcircuit instance is replaced by its definition using text expansion and the hierarchy is
not present after input processing.
The subcircuit is defined in the input deck by a grouping of element cards delimited by
the .subckt and the .ends cards (or the keywords defined by the substart and subend
Code Model Type

R Semiconductor resistor model
C Semiconductor capacitor model
L Inductor model
SW Voltage controlled switch
CSW Current controlled switch
URC Uniform distributed RC model
LTRA Lossy transmission line model
D Diode model
NPN NPN BJT model
PNP PNP BJT model
NJF N-channel JFET model
PJF P-channel JFET model
NMOS N-channel MOSFET model
PMOS P-channel MOSFET model
NMF N-channel MESFET model
PMF P-channel MESFET model
VDMOS Power MOS model
Table 2.3: Ngspice model types
options (see 17.7)); the program then automatically inserts the defined group of elements
wherever the subcircuit is referenced. Instances of subcircuits within a larger circuit are
defined through the use of an instance card that begins with the letter ‘X’. A complete
example of all three of these cards follows:
Example:
* The following is the instance card :

*
xdiv1 10 7 0 vdivide
* The following are the subcircuit definition cards :

*
. subckt vdivide 1 2 3
r1 1 2 10 K
r2 2 3 5 K
. ends
The above specifies a subcircuit with ports numbered ‘1’, ‘2’ and ‘3’:
• Resistor ‘R1’ is connected from port ‘1’ to port ‘2’, and has value 10 kOhms.
• Resistor ‘R2’ is connected from port ‘2’ to port ‘3’, and has value 5 kOhms.
The instance card, when placed in an ngspice deck, will cause subcircuit port ‘1’ to be
equated to circuit node ‘10’, while port ‘2’ will be equated to node ‘7’ and port ‘3’ will
equated to node ‘0’.
2.5. .SUBCKT SUBCIRCUITS 57
There is no limit on the size or complexity of subcircuits, and subcircuits may contain
other subcircuits. An example of subcircuit usage is given in Chapt. 21.6.
2.5.1 .SUBCKT Line
General form:
. SUBCKT subnam N1 < N2 N3 ... >
Examples:
. SUBCKT OPAMP 1 2 3 4
A circuit definition is begun with a .SUBCKT line. subnam is the subcircuit name, and
N1, N2, ... are the external nodes, which cannot be zero. The group of element lines
that immediately follow the .SUBCKT line define the subcircuit. The last line in a sub-
circuit definition is the .ENDS line (see below). Control lines may not appear within a
subcircuit definition; however, subcircuit definitions may contain anything else, including
other subcircuit definitions, device models, and subcircuit calls (see below). Note that
any device models or subcircuit definitions included as part of a subcircuit definition are
strictly local (i.e., such models and definitions are not known outside the subcircuit defi-
nition). Also, any element nodes not included on the .SUBCKT line are strictly local, with
the exception of 0 (ground) that is always global. If you use parameters, the .SUBCKT line
will be extended (see 2.9.3).
2.5.2 .ENDS Line
General form:
. ENDS < SUBNAM >
Examples:
. ENDS OPAMP
The .ENDS line must be the last one for any subcircuit definition. The subcircuit name,
if included, indicates which subcircuit definition is being terminated; if omitted, all sub-
circuits being defined are terminated. The name is needed only when nested subcircuit
definitions are being made.
2.5.3 Subcircuit Calls
General form:
XYYYYYYY N1 < N2 N3 ... > SUBNAM
Examples:
X1 2 4 17 3 1 MULTI
Subcircuits are used in ngspice by specifying pseudo-elements beginning with the letter
X, followed by the circuit nodes to be used in expanding the subcircuit. If you use
parameters, the subcircuit call will be modified (see 2.9.3).
2.6 .GLOBAL
General form:
. GLOBAL nodename1 [ nodename2 ... ]
Examples:
. GLOBAL gnd vcc
Nodes defined in the .GLOBAL statement are available to all circuit and subcircuit blocks
independently from any circuit hierarchy. After parsing the circuit, these nodes are ac-
cessible from top level.
2.7 .INCLUDE
General form:
. INCLUDE filename
Examples:
. INCLUDE / users / spice / common / bsim3 - param . mod
Frequently, portions of circuit descriptions will be reused in several input files, particularly
with common models and subcircuits. In any ngspice input file, the .INCLUDE line may
be used to copy some other file as if that second file appeared in place of the .INCLUDE
line in the original file.
2.8. .LIB 59
If the filename is a relative path and the file is not found, it is searched for in the locations
given by variable sourcepath (17.7). There is no restriction on the file name imposed by
ngspice beyond those imposed by the local operating system.
2.8 .LIB
General form:
. LIB filename libname
Examples:
. LIB / users / spice / common / mosfets . lib mos1
The .LIB statement allows including library descriptions into the input file. Inside the
*.lib file a library libname will be selected. The statements of each library inside the
*.lib file are enclosed in .LIB libname <...> .ENDL statements. The file is searched for
in the same way as for .include.
If the compatibility mode (16.14) is set to ’ps’ by set ngbehavior=ps (17.7) in spinit
(16.5) or .spiceinit (16.6), then a simplified syntax .LIB filename is available: a warning
is issued and filename is simply included as described in Chapt. 2.7.
2.9 .PARAM Parametric netlists

Ngspice allows for the definition of parametric attributes in the netlists. This is an
enhancement of the ngspice front-end that adds arithmetic functionality to the circuit
description language.
2.9.1 .param line

General form:
. param < ident > = < expr > < ident > = < expr > ...
Examples:
. param pippo =5
. param po =6 pp =7.8 pap ={ AGAUSS ( pippo , 1 , 1.67)}
. param pippp ={ pippo + pp }
. param p ={ pp }
. param pop = ’ pp +p ’
This line assigns numerical values to identifiers. More than one assignment per line
is possible using a separating space. Parameter identifier names must begin with an
alphabetic character. The other characters must be either alphabetic, a number, or ! #

$ % [ ] _ as special characters. The variables time, temper, and hertz (see 5.1.1) are
not valid identifier names. Other restrictions on naming conventions apply as well, see
2.9.6.
The .param lines inside subcircuits are copied per call, like any other line. All assignments
are executed sequentially through the expanded circuit. Before its first use, a parameter
name must have been assigned a value. Expressions defining a parameter should be put
within braces {p+p2}, or alternatively within single quotes ’AGAUSS(pippo, 1, 1.67)’.
An assignment cannot be self-referential, something like .param pip = ’pip+3’ will not
work.
The current ngspice version does not always need quotes or braces in expressions, es-
pecially when spaces are used sparingly. However, it is recommended to do so, as the
following examples demonstrate.
. param a = 123 * 3 b = sqrt (9) $ doesn ’ t work , a <= 123

. param a = ’123 * 3 ’ b = sqrt (9) $ ok .
. param c = a + 123 $ won ’ t work
. param c = ’a + 123 ’ $ ok .
. param c = a +123 $ ok .
Parameters may also have string values, but support is limited. String-valued parameters
can be defined by .param and used in the same ways as numeric parameters. The only
operation on string values is concatenation and that is possible only in top-level .param
assignments.
. param str1 =" first " str2 =" second "

. param both ={ str1 }" and " str2
2.9.2 Brace expressions in circuit elements:
General form:
{ < expr > }
Examples:
These are allowed in .model lines and in device lines. A SPICE number is a floating
point number with an optional scaling suffix, immediately glued to the numeric tokens
(see Chapt. 2.9.5). Brace expressions ({..}) cannot be used to parameterize node names
or parts of names. All identifiers used within an <expr> must have known values at the
time when the line is evaluated, else an error is flagged.
2.9. .PARAM PARAMETRIC NETLISTS 61
2.9.3 Subcircuit parameters

General form:
. subckt < identn > node node ... < ident >= < value > < ident >= < value > ...
Examples:
. subckt myfilter in out rval =100 k cval =100 nF
<identn> is the name of the subcircuit given by the user. node is an integer number
or an identifier, for one of the external nodes. The first <ident>=<value> introduces an
optional section of the line. Each <ident> is a formal parameter, and each <value> is
either a SPICE number or a brace expression. Inside the .subckt ... .ends context, each
formal parameter may be used like any identifier that was defined on a .param control
line. The <value> parts are default values of the parameters.
The syntax of a subcircuit call (invocation) is:
General form:
X < name > node node ... < identn > < ident >= < value > < ident >= < value > ...
Examples:
X1 input output myfilter rval =1 k
Here <name> is the symbolic name given to that instance of the subcircuit, <identn>
is the name of a subcircuit defined beforehand. node node ... is the list of actual
nodes where the subcircuit is connected. <value> is either a SPICE number or a brace
expression { <expr> } .
Subcircuit example with parameters:
* Param - example
. param amplitude = 1 V
*
. subckt myfilter in out rval =100 k cval =100 nF
Ra in p1 {2* rval }
Rb p1 out {2* rval }
C1 p1 0 {2* cval }
Ca in p2 { cval }
Cb p2 out { cval }
R1 p2 0 { rval }
. ends myfilter
*
X1 input output myfilter rval =1 k cval =1 n
V1 input 0 AC { amplitude }
. end
2.9.4 Symbol scope

All subcircuit and model names are considered global and must be unique. The .param
symbols that are defined outside of any .subckt ... .ends section are global. Inside such
a section, the pertaining params: symbols and any .param assignments are considered
local: they mask any global identical names, until the .ends line is encountered. You
cannot reassign to a global number inside a .subckt, a local copy is created instead.
Scope nesting works up to a level of 10. For example, if the main circuit calls A that has
a formal parameter xx, A calls B that has a param. xx, and B calls C that also has a
formal param. xx, there will be three versions of ‘xx’ in the symbol table but only the
most local one - belonging to C - is visible.
2.9.5 Syntax of expressions

<expr> ( optional parts within [...] )
An expression may be one of:
< atom > where < atom > is either a spice number or an identifier
< unary - operator > < atom >
< function - name > ( < expr > [ , < expr > ...] )
< atom > < binary - operator > < expr >
( < expr > )
As expected, atoms, built-in function calls and stuff within parentheses are evaluated
before the other operators. The operators are evaluated following a list of precedence
close to the one of the C language. For equal precedence binary ops, evaluation goes left
to right. Functions operate on real values only!
Operator Alias Precedence Description
- 1 unary -
! 1 unary not
** ^ 2 power, like pwr
* 3 multiply
/ 3 divide
% 3 modulo
\ 3 integer divide
+ 4 add
- 4 subtract
== 5 equality
!= <> 5 non-equal
<= 5 less or equal
>= 5 greater or equal
< 5 less than
> 5 greater than
&& 6 boolean and
|| 7 boolean or
c?x:y 8 ternary operator
2.9. .PARAM PARAMETRIC NETLISTS 63
The evaluation of the power functions ** or ^ depends on the compatibility mode (16.14.1)
chosen.
Power function source code implementation:
compatmode hs : x >0 pow (x , y ); x <0 pow (x , round ( y )); X =0 0

compatmode lt : x >0 pow (x , y ); x <0 pow (x , y )
if y is close to integer ; else 0
The number zero is used to represent boolean False. Any other number represents boolean
True. The result of logical operators is 1 or 0. An example input file is shown below:
Example input file with logical operators:
* Logical operators
v1or 1 0 {1 || 0}
v1and 2 0 {1 && 0}
v1not 3 0 {! 1}
v1mod 4 0 {5 % 3}
v1div 5 0 {5 \ 3}
v0not 6 0 {! 0}
. control
op
print allv
. endc
. end
Built-in function Notes

sqrt(x) y = sqrt(x)
sin(x), cos(x), tan(x)
sinh(x), cosh(x), tanh(x)
asin(x), acos(x), atan(x)
asinh(x), acosh(x), atanh(x)
arctan(x) atan(x), kept for compatibility
exp(x)
ln(x), log(x)
abs(x)
nint(x) Nearest integer, half integers towards even
int(x) Nearest integer rounded towards 0
floor(x) Nearest integer rounded towards -∞
ceil(x) Nearest integer rounded towards +∞
pow(x,y) x raised to the power of y (pow from C runtime library)
pwr(x,y) pow(fabs(x), y)
min(x, y)
max(x, y)
sgn(x) 1.0 for x > 0, 0.0 for x == 0, -1.0 for x < 0
ternary_fcn(x, y, z) x ? y : z
gauss(nom, rvar, sigma) nominal value plus variation drawn from Gaussian
distribution with mean 0 and standard deviation rvar
(relative to nominal), divided by sigma
agauss(nom, avar, sigma) nominal value plus variation drawn from Gaussian
distribution with mean 0 and standard deviation avar
(absolute), divided by sigma
unif(nom, rvar) nominal value plus relative variation (to nominal)
uniformly distributed between +/-rvar
aunif(nom, avar) nominal value plus absolute variation uniformly
distributed between +/-avar
limit(nom, avar) nominal value +/-avar, depending on random number
in [-1, 1[ being > 0 or < 0
The scaling suffixes (any decorative alphanumeric string may follow):
suffix value
g 1e9
meg 1e6
k 1e3
m 1e-3
u 1e-6
n 1e-9
p 1e-12
f 1e-15
Note: there are intentional redundancies in expression syntax, e.g. x^y , x**y and
pwr(x,y) all have nearly the same result.
2.10. .FUNC 65
2.9.6 Reserved words
In addition to the above function names and to the verbose operators ( not and or div
mod ), other words are reserved and cannot be used as parameter names: or, defined,
sqr, sqrt, sin, cos, exp, ln, log, log10, arctan, abs, pwr, time, temper, hertz.
2.9.7 A word of caution on the three ngspice expression parsers
The historical parameter notation using & as the first character of a line as equivalence
to .param. is deprecated and will be removed in a coming release.
Confusion may arise in ngspice because of its multiple numerical expression features. The
.param lines and the brace expressions (see Chapt. 2.10) are evaluated in the front-
end, that is, just after the subcircuit expansion. (Technically, the X lines are kept as
comments in the expanded circuit so that the actual parameters can be correctly sub-
stituted). Therefore, after the netlist expansion and before the internal data setup, all
number attributes in the circuit are known constants. However, there are circuit elements
in Spice that accept arithmetic expressions not evaluated at this point, but only later
during circuit analysis. These are the arbitrary current and voltage sources (B-sources,
5), as well as E- and G-sources and R-, L-, or C-devices. The syntactic difference is that
‘compile-time’ expressions are within braces, but ‘run-time’ expressions have no braces.
To make things more complicated, the back-end ngspice scripting language accepts arith-
metic/logic expressions that operate only on its own scalar or vector data sets (17.2).
Please see Chapt. 2.14.
It would be desirable to have the same expression syntax, operator and function set,
and precedence rules, for the three contexts mentioned above. In the current Numparam
implementation, that goal is not achieved.
2.10 .FUNC
This keyword defines a function. The syntax of the expression is the same as for a .param
(2.9.5).
General form:
. func < ident > { < expr > }

. func < ident > = { < expr > }
Examples:
. func icos ( x ) { cos ( x ) - 1}

. func f (x , y ) { x * y }
. func foo (a , b ) = { a + b }
.func will initiate a replacement operation. After reading the input files, and before
parameters are evaluated, all occurrences of the icos(x) function will be replaced by
cos(x)-1. All occurrences of f(x,y) will be replaced by x*y. Function statements may
be nested to a depth of t.b.d..
2.11 .CSPARAM
Create a constant vector (see 17.8.2) from a parameter in plot (17.3) const.
General form:
. csparam < ident > = < expr >
Examples:
. param pippo =5
. param pp =6
. csparam pippp ={ pippo + pp }
. param p ={ pp }
. csparam pap = ’ pp +p ’
In the example shown, vectors pippp, and pap are added to the constants that already
reside in plot const, having length one and real values. These vectors are generated dur-
ing circuit parsing and thus cannot be changed later (same as with ordinary parameters).
They may be used in ngspice scripts and .control sections (see Chapt. 17).
The use of .csparam is still experimental and has to be tested. A simple usage is shown
below.
* test csparam
.param TEMPS = 27
.csparam newt = {3*TEMPS}
.csparam mytemp = ’2 + TEMPS’
.control
echo $&newt $&mytemp
.endc
.end
2.12 .TEMP
Sets the circuit temperature in degrees Celsius.
General form:
. temp value
Examples:
. temp 27
2.13. .IF CONDITION-CONTROLLED NETLIST 67
This card overrides the circuit temperature given in an .option line (15.1.1).
2.13 .IF Condition-Controlled Netlist
A simple .IF-.ELSE(IF) block allows condition-controlling of the netlist. boolean expression

is any expression according to Chapt. 2.9.5 that evaluates parameters and returns a
boolean 1 or 0. The netlist block in between the .if ... .endif statements may contain
device instances or .model cards that are selected according to the logic condition.
General form:
. if ( boolean expression )
...
. elseif ( boolean expression )
...
. else
...
. endif
Example 1:
* device instance in IF - ELSE block

. param ok =0 ok2 =1
v1 1 0 1
R1 1 0 2
. if ( ok && ok2 )
R11 1 0 2
. else
R11 1 0 0.5 $ <-- selected
. endif
Example 2:
* . model in IF - ELSE block

. param m0 =0 m1 =1
M1 1 2 3 4 N1 W =1 L =0.5
. if ( m0 ==1)
. model N1 NMOS level =49 Version =3.1
. elseif ( m1 ==1)
. model N1 NMOS level =49 Version =3.2.4 $ <-- selected
. else
. model N1 NMOS level =49 Version =3.3.0
. endif
Nesting of .IF-.ELSE(IF)-.ENDIF blocks is possible. Several .elseif (but of course

only one .else)are allowed per block (please see example ngspice/tests/regression/misc/if-
elseif.cir). However some restrictions apply, as the following netlist components are not
supported within the .IF-.ENDIF block: .SUBCKT, .INC, .LIB, and .PARAM.
2.14. PARAMETERS, FUNCTIONS, EXPRESSIONS, AND COMMAND SCRIPTS69
2.14 Parameters, functions, expressions, and com-

mand scripts
In ngspice there are several ways to describe functional dependencies. In fact there are
three independent function parsers, being active before, during, and after the simulation.
So it might be due to have a few words on their interdependence.
2.14.1 Parameters
Parameters (Chapt. 2.9.1) and functions, either defined within the .param statement or
with the .func statement (Chapt. 2.10) are evaluated before any simulation is started,
that is during the setup of the input and the circuit. Therefore these statements may not
contain any simulation output (voltage or current vectors), because it is simply not yet
available. The syntax is described in Chapt. 2.9.5. During the circuit setup all functions
are evaluated, all parameters are replaced by their resulting numerical values. Thus it will
not be possible to get feedback from a later stage (during or after simulation) to change
any of the parameters.
2.14.2 Nonlinear sources

During the simulation, the B source (Chapt. 5) and their associated E and G sources, as
well as some devices (R, C, L) may contain expressions. These expressions may contain
parameters from above (evaluated immediately upon ngspice start up), numerical data,
predefined functions, but also node voltages and branch currents resulting from the sim-
ulation. The source or device values are continuously updated during the simulation.
Therefore the sources are powerful tools to define non-linear behavior, you may even cre-
ate new ‘devices’ by yourself. Unfortunately the expression syntax (see Chapt. 5.1) and
the predefined functions may deviate from the ones for parameters listed in 2.9.1.
2.14.3 Control commands, Command scripts

Commands, as described in detail in Chapt. 17.5, may be used interactively, but also
as a command script enclosed in .control ... .endc lines. The scripts may contain
expressions (see Chapt. 17.2). The expressions may work upon simulation output vectors
(of node voltages, branch currents), as well as upon predefined or user defined vectors
and variables, and are invoked after the simulation. Parameters from 2.9.1 defined by
the .param statement are not allowed in these expressions. However you may define such
parameters with .csparam (2.11). Again the expression syntax (see Chapt. 17.2) will
deviate from the one for parameters or B sources listed in 2.9.1 and 5.1.
If you want to use parameters from 2.9.1 inside your control script, you may use .csparam
(2.11) or apply a trick by defining a voltage source with the parameter as its value,
and then have it available as a vector (e.g. after a transient simulation) with a then
constant output (the parameter). A feedback from here back into parameters (2.14.1)
is never possible. Also you cannot access non-linear sources of the preceding simulation.
However you may start a first simulation inside your control script, then evaluate its
output using expressions, change some of the element or model parameters with the
alter and altermod statements (see Chapt. 17.5.3) and then automatically start a new
simulation.
Expressions and scripting are powerful tools within ngspice, and we will enhance the
examples given in Chapt. 21 continuously to describe these features.
Chapter 3
Circuit Elements and Models
Data fields that are enclosed in less-than and greater-than signs (‘< >’) are optional.
All indicated punctuation (parentheses, equal signs, etc.) is optional but indicate the
presence of any delimiter. Further, future implementations may require the punctuation
as stated. A consistent style adhering to the punctuation shown here makes the input
easier to understand. With respect to branch voltages and currents, ngspice uniformly
uses the associated reference convention (current flows in the direction of voltage drop).
3.1 About netlists, device instances, models and model

parameters
The input to ngspice is a netlist, which lists all circuit elements, their interconnects and
model parameters.
71
72 CHAPTER 3. CIRCUIT ELEMENTS AND MODELS
Netlist example of a simple bipolar amplifier:
bipolar amplifier
R3 vcc intc 10 k
R1 vcc intb 68 k
R2 intb 0 10 k
Cout out intc 10 u
Cin intb in 10 u
RLoad out 0 100 k
Q1 intc intb 0 BC546B
VCC vcc 0 5
Vin in 0 dc 0 ac 1 sin (0 1 m 500)
. model BC546B npn ( IS =7.59 E -15 VAF =73.4 BF =480 IKF =0.0962
+ NE =1.2665 ISE =3.278 E -15 IKR =0.03 ISC =2.00 E -13 NC =1.2 NR =1
+ BR =5 RC =0.25 CJC =6.33 E -12 FC =0.5 MJC =0.33 VJC =0.65
+ CJE =1.25 E -11 MJE =0.55 VJE =0.65 TF =4.26 E -10 ITF =0.6 VTF =3
+ XTF =20 RB =100 IRB =0.0001 RBM =10 RE =0.5 TR =1.50 E -07)
. end
After the first line, which is always a title line only, the netlist starts. Each line here is a
device instance (except for lines starting with a dot ’.’). We have simple circuit elements
that consist of a single line only, e.g. resistors like R3. In its simplest implementation,
the resistor model does not need any model parameters except for the resistance value
(same for capacitors like Cout). Netlist lines like R3 vcc intc 10k are called instance lines,
as each line is the representation of an instance of a generic model hard-coded into the
ngspice simulator (here: resistor). R3 denotes the device name. Its first character R
denotes a resistor. The next two tokens vcc intc are the two nodes of the resistor, 10k is
the resistance value. Equal node names on different devices denote a connection between
these nodes.
A more complex device is described by the instance line Q1 intc intb 0 BC546B. Q denotes
a bipolar transistor, intc intb 0 are the three nodes collector, base, and emitter. BC546B is
the name of a model parameter set, named after a real transistor and describing (together
with the implemented bipolar transistor model) its electrical behavior. The associated
model parameters are given in the line .model BC546B npn (IS=7.59E-15 ...). This is not an
instance line, because starting with a dot. It contains the model parameters as supplied by
the device manufacturer or by people having them extracted from the electrical behavior
and data sheet (to be found e.g. on his or her web pages). BC546B is the name of the
model parameter set and relates it to the device instance. npn is the type of the device.
The parameters (name=value) are given in brackets.
The instance Q1... requires model parameters. For a quick test one may do without
device maker’s model parameters.
3.2. GENERAL OPTIONS 73
Simplified bipolar transistor instance and model parameter set:
Q1 intc intb 0 defaultmod

. model defaultmod npn
If you enter the bipolar transistor instance as shown above, you make use of a default
model parameter set supplied by ngspice. defaultmod is an arbitrary name. This procedure
models a generic bipolar transistor, not resembling any commercial device. The default
parameter values may be assessed by the command showmod Q1.
You will get more information on devices, instances and models in the following chapters
3.3 to 12.
3.2 General options
3.2.1 Paralleling devices with multiplier m
When it is needed to simulate several devices of the same kind in parallel, use the ‘m’
(parallel multiplier) instance parameter available for the devices listed in Table 3.1. This
multiplies the value of the element’s matrix stamp with m’s value. The netlist below shows
how to correctly use the parallel multiplier:
Multiple device example:
d1 2 0 mydiode m =10
d01 1 0 mydiode
d02 1 0 mydiode
d03 1 0 mydiode
d04 1 0 mydiode
d05 1 0 mydiode
d06 1 0 mydiode
d07 1 0 mydiode
d08 1 0 mydiode
d09 1 0 mydiode
d10 1 0 mydiode
...
The d1 instance connected between nodes 2 and 0 is equivalent to the 10 parallel devices
d01-d10 connected between nodes 1 and 0.
The following devices support the multiplier m:

First letter Element description

C Capacitor
D Diode
F Current-controlled current source (CCCs)
G Voltage-controlled current source (VCCS)
I Current source
J Junction field effect transistor (JFET)
L Inductor
M Metal oxide field effect transistor (MOSFET)
Q Bipolar junction transistor (BJT)
R Resistor
X Subcircuit (for details see below)
Z Metal semiconductor field effect transistor (MESFET)
Table 3.1: ngspice elements supporting multiplier ’m’
When the X line (e.g. x1 a b sub1 m=5) contains the token m=value (as shown) or
m=expression, subcircuit invocation is done in a special way. If an instance line of the
subcircuit sub1 contains any of the elements shown in table 3.1, then these elements are
instantiated with the additional parameter m (in this example having the value 5). If such
an element already has an m multiplier parameter, the element m is multiplied with the
m derived from the X line. This works recursively, meaning that if a subcircuit contains
another subcircuit (a nested X line), then the latter m parameter will be multiplied by the
former one, and so on.
Example 1:
. param madd = 6
X1 a b sub1 m =5
. subckt sub1 a1 b1
Cs1 a1 b1 C =5 p m = ’ madd -2 ’
. ends
In example 1, the capacitance between nodes a and b will be C = 5pF*(madd-2)*5 =

100pF.
Example 2:
. param madd = 4
X1 a b sub1 m =3
. subckt sub1 a1 b1
X2 a1 b1 sub2 m = ’ madd -2 ’
. ends
. subckt sub2 a2 b2
Cs2 a2 b2 3 p m =2
. ends
In example 2, the capacitance between nodes a and b is C = 3pF*2*(madd-2)*3 = 36pF.

3.2. GENERAL OPTIONS 75
Using m may fail to correctly describe geometrical properties for real devices like MOS
transistors.
M1 d g s nmos W=0.3u L=0.18u m=20
is probably not be the same as
M1 d g s nmos W=6u L=0.18u
because the former may suffer from small width (or edge) effects, whereas the latter is
simply a wide transistor.
3.2.2 Instance and model parameters

The simple device example below consists of two lines: The device is defined on the
instance line, starting with Lload ...: The first letter determines the device type (an
inductor in this example). Following the device name are two nodes 1 and 2, then the
inductance value 1u is set. The model name ind1 is a connection to the respective model
line. Finally we have a parameter on the instance line, together with its value dtemp=5.
Parameters on an instance line are called instance parameters.
The model line starts with the token .model, followed by the model name, the model type
and at least one model parameter, here tc1=0.001. There are complex models with more
than 100 model parameters.
Lload 1 2 1 u ind1 dtemp =5

. MODEL ind1 L tc1 =0.001
Instance parameters are listed in each of the following device descriptions. Model pa-
rameters sometimes are given below as well, for complex models like the BSIM transistor
models, they are available in the model makers documentation. Instance parameters may
also be placed in the .model line. Thus they are recognized by each device instance refer-
ring to that model. Their values may be overridden for a specific instance of a device by
placing them additionally onto its instance line.
3.2.3 Model binning

Binning is a kind of range partitioning for geometry dependent models like MOSFET’s.
The purpose is to cover larger geometry ranges (Width and Length) with higher accuracy
than the model built-in geometry formulas. Each size range described by the additional
model parameters LMIN, LMAX, WMIN and WMAX has its own model parameter set.
These model cards are defined by a number extension, like ‘nch.1’. ngspice has an algo-
rithm to choose the right model card by the requested W and L.
This is implemented for BSIM3 (11.2.10) and BSIM4 (11.2.11) models.
3.2.4 Initial conditions

Two different forms of initial conditions may be specified for some devices. The first form
is included to improve the dc convergence for circuits that contain more than one stable
state. If a device is specified OFF, the dc operating point is determined with the terminal
voltages for that device set to zero. After convergence is obtained, the program continues
to iterate to obtain the exact value for the terminal voltages. If a circuit has more than
one dc stable state, the OFF option can be used to force the solution to correspond to a
desired state. If a device is specified OFF when in reality the device is conducting, the
program still obtains the correct solution (assuming the solutions converge) but more
iterations are required since the program must independently converge to two separate
solutions.
The .NODESET control line (see Chapt. 15.2.1) serves a similar purpose as the OFF option.
The .NODESET option is easier to apply and is the preferred means to aid convergence. The
second form of initial conditions are specified for use with the transient analysis. These
are true ‘initial conditions’ as opposed to the convergence aids above. See the description
of the .IC control line (Chapt. 15.2.2) and the .TRAN control line (Chapt. 15.3.10) for a
detailed explanation of initial conditions.
3.3 Elementary Devices
3.3.1 Resistors
General form:
RXXXXXXX n + n - < resistance | r = > value < ac = val > <m = val >
+ < scale = val > < temp = val > < dtemp = val > < tc1 = val > < tc2 = val >
+ < noisy =0|1 >
Examples:
R1 1 2 100
RC1 12 17 1 K
R2 5 7 1 K ac =2 K
RL 1 4 2 K m =2
Ngspice has a fairly complex model for resistors. It can simulate both discrete and semi-
conductor resistors. Semiconductor resistors in ngspice means: resistors described by
geometrical parameters. So, do not expect detailed modeling of semiconductor effects.
n+ and n- are the two element nodes, value is the resistance (in ohms) and may be
positive or negative1 but not zero.
Simulating small valued resistors: If you need to simulate very small
resistors (0.001 Ohm or less), you should use CCVS (transresistance).
It is less efficient but improves overall numerical accuracy. Consider a
small resistance as a large conductance.
Ngspice can assign a resistor instance a different value for AC analysis, specified using the
ac keyword. This value must not be zero as described above. The AC resistance is used
1
A negative resistor modeling an active element can cause convergence problems, please avoid it.
3.3. ELEMENTARY DEVICES 77
in AC analysis only (neither Pole-Zero nor Noise). If you do not specify the ac parameter,
it is defaulted to value.
Ngspice calculates the nominal resistance as
VALUE scale
Rnom = m
(3.1)
ac scale
Racnom = m
.
If you want to simulate temperature dependence of a resistor, you need to specify its tem-
perature coefficients, using a .model line or as instance parameters, like in the examples
below:
Examples:
RE1 1 2 800 newres dtemp =5

. MODEL newres R tc1 =0.001
RE2 a b 1.4 k tc1 =2 m tc2 =1.4 u
RE3 n1 n2 1 Meg tce =700 m
The temperature coefficients tc1 and tc2 describe a quadratic temperature dependence
(see equation 1.6) of the resistance. If given in the instance line (the R... line) their
values will override the tc1 and tc2 of the .model line (3.3.3). Ngspice has an additional
temperature model equation 3.2 parameterized by tce given in model or instance line. If
all parameters are given (quadratic and exponential) the exponential temperature model
is chosen.

R (T ) = R (T0 ) 1.01T CE·(T −T0 ) (3.2)
where T is the circuit temperature, T0 is the nominal temperature, and T CE is the

exponential temperature coefficients.
Instance temperature is useful even if resistance does not vary with it, since the thermal
noise generated by a resistor depends on its absolute temperature. Resistors in ngspice
generates two different noises: thermal and flicker. While thermal noise is always gener-
ated in the resistor, to add a flicker noise2 source you have to add a .model card defining
the flicker noise parameters. It is possible to simulate resistors that do not generate any
kind of noise using the noisy (or noise) keyword and assigning zero to it, as in the
following example:
Example:
Rmd 134 57 1.5 k noisy =0
If you are interested in temperature effects or noise equations, read the next section on
semiconductor resistors.
2
Flicker noise can be used to model carbon resistors.
3.3.2 Semiconductor Resistors

General form:
RXXXXXXX n + n - < value > < mname > <l = length > <w = width >
+ < temp = val > < dtemp = val > <m = val > < ac = val > < scale = val >
+ < noisy = 0|1 >
Examples:
RLOAD 2 10 10 K
RMOD 3 7 RMODEL L =10 u W =1 u
This is the more general form of the resistor presented before (3.3.1) and allows the
modeling of temperature effects and for the calculation of the actual resistance value from
strictly geometric information and the specifications of the process. If value is specified,
it overrides the geometric information and defines the resistance. If mname is specified,
then the resistance may be calculated from the process information in the model mname
and the given length and width. If value is not specified, then mname and length must
be specified. If width is not specified, then it is taken from the default width given in
the model.
The (optional) temp value is the temperature at which this device is to operate, and
overrides the temperature specification on the .option control line and the value specified
in dtemp.
3.3.3 Semiconductor Resistor Model (R)

The resistor model consists of process-related device data that allow the resistance to
be calculated from geometric information and to be corrected for temperature. The
parameters available are as follows:
Name Parameter Units Default Example
TC1 first order temperature coeff. Ω/◦ C 0.0 -
TC2 second order temperature coeff. Ω/◦ C 2 0.0 -
RSH sheet resistance Ω/ - 50
DEFW default width m 1e-6 2e-6
NARROW narrowing due to side etching m 0.0 1e-7
SHORT shortening due to side etching m 0.0 1e-7
◦
TNOM parameter measurement temperature C 27 50
KF flicker noise coefficient 0.0 1e-25
AF flicker noise exponent 0.0 1.0
WF flicker noise width exponent 1.0
LF flicker noise length exponent 1.0
EF flicker noise frequency exponent 1.0
R (RES) default value if element value not given Ω - 1000
The sheet resistance is used with the narrowing parameter and l and w from the resistor
device to determine the nominal resistance by the formula:
l − SHORT
Rnom = rsh (3.3)
w − NARROW
DEFW is used to supply a default value for w if one is not specified for the device. If either
rsh or l is not specified, then the standard default resistance value of 1 mOhm is used.
TNOM is used to override the circuit-wide value given on the .options control line where
the parameters of this model have been measured at a different temperature. After the
nominal resistance is calculated, it is adjusted for temperature by the formula:

R(T ) = R(TNOM) 1 + T C1 (T − TNOM) + T C2 (T − TNOM)2 (3.4)
where R(TNOM) = Rnom |Racnom . In the above formula, ‘T ’ represents the instance
temperature, which can be explicitly set using the temp keyword or calculated using the
circuit temperature and dtemp, if present. If both temp and dtemp are specified, the latter
is ignored. Ngspice improves SPICE’s resistors noise model, adding flicker noise (1/f ) to
it and the noisy (or noise) keyword to simulate noiseless resistors. The thermal noise
in resistors is modeled according to the equation:
4kT
i¯2R = ∆f (3.5)
R
where ‘k’ is the Boltzmann’s constant, and ‘T ’ the instance temperature.
Flicker noise model is:
¯ = KFIRAF
i2Rf n ∆f (3.6)
W W F LLF f EF
A small list of sheet resistances (in Ω/) for conductors is shown below. The table repre-
sents typical values for MOS processes in the 0.5 - 1 um
range. The table is taken from: N. Weste, K. Eshraghian - Principles of CMOS VLSI
Design 2nd Edition, Addison Wesley.
Material Min. Typ. Max.

Inter-metal (metal1 - metal2) 0.005 0.007 0.1
Top-metal (metal3) 0.003 0.004 0.05
Polysilicon (poly) 15 20 30
Silicide 2 3 6
Diffusion (n+, p+) 10 25 100
Silicided diffusion 2 4 10
n-well 1000 2000 5000
3.3.4 Resistors, dependent on expressions (behavioral resistor)
General form:
RXXXXXXX n + n - R = ’ expression ’ < tc1 = value > < tc2 = value > < noisy =0 >
RXXXXXXX n + n - ’ expression ’ < tc1 = value > < tc2 = value > < noisy =0 >
Examples:
R1 rr 0 r = ’V ( rr ) < { Vt } ? { R0 } : {2* R0 } ’ tc1 =2 e -03 tc2 =3.3 e -06

R2 r2 rr r = {5 k + 50* TEMPER }
. param rp1 = 20
R3 no1 no2 r = ’5 k * rp1 ’ noisy =1
Expression may be an equation or an expression containing node voltages or branch

currents (in the form of i(vm)) and any other terms as given for the B source and described
in Chapt. 5.1. It may contain parameters (2.9.1) and the special variables time, temper,
and hertz (5.1.2). An example file is given below. Small signal noise in the resistor
(15.3.4) may be evaluated as white noise, depending on resistance, temperature and tc1,
tc2. To enable noise calculation, add the flag noisy=1 to the instance line. As a default
the behavioral resistor is noiseless.
Example input file for non-linear resistor:
Non - linear resistor

. param R0 =1 k Vi =1 Vt =0.5
* resistor depending on control voltage V ( rr )
R1 rr 0 r = ’V ( rr ) < { Vt } ? { R0 } : {2* R0 } ’
* control voltage
V1 rr 0 PWL (0 0 100 u { Vi })
. control
unset askquit
tran 100 n 100 u uic
plot i ( V1 )
. endc
. end
3.3.5 Resistor with nonlinear r2_cmc model
In the adms version of ngspice, a resistor model r2_cmc is implemented. This is a 2-

terminal resistor model developed by the resistor subcommittee of the CMC. The goal
was to have a standard 2-terminal resistor model with standard parameter names and
a standard, numerically well behaved nonlinearity model. It may be selected by setting
level=2 in the .model line.
For now a detailed description is available in the Verilog A source code file to be found a
src/spicelib/devices/adms/r2_cmc/admsva/r2_cmc.va.
Example input file for non-linear resistor with r2_cmc model
r2_cmc
v1 1 0 10
Rr2_cmc 1 0 rmodel w =1 u l =20 u isnoisy =1
. model rmodel r ( level =2 rsh =200 xl =0.2 u xw = -0.05 u
+ p3 =0.12 q3 =1.6 p2 =0.015 q2 =3.8 tc1 =1.5 e -4 tc2 =7 e -7)
. control
op
let res = v (1) / - v1 # branch
print res . endc
. end
3.3.6 Capacitors
General form:
CXXXXXXX n + n - < value > < mname > <m = val > < scale = val > < temp = val >
+ < dtemp = val > < tc1 = val > < tc2 = val > < ic = init_condition >
Examples:
CBYP 13 0 1 UF
COSC 17 23 10 U IC =3 V
Ngspice provides a detailed model for capacitors. Capacitors in the netlist can be specified
giving their capacitance or their geometrical and physical characteristics. Following the
original SPICE3 ‘convention’, capacitors specified by their geometrical or physical char-
acteristics are called ‘semiconductor capacitors’ and are described in the next section.
In this first form n+ and n- are the positive and negative element nodes, respectively and
value is the capacitance in Farads.
Capacitance can be specified in the instance line as in the examples above or in a .model
line, as in the example below:
C1 15 5 cstd
C2 2 7 cstd
. model cstd C cap =3 n
Both capacitors have a capacitance of 3nF.

If you want to simulate temperature dependence of a capacitor, you need to specify its
temperature coefficients, using a .model line, like in the example below:
CEB 1 2 1 u cap1 dtemp =5

. MODEL cap1 C tc1 =0.001
The (optional) initial condition is the initial (time zero) value of capacitor voltage (in
Volts). Note that the initial conditions (if any) apply only if the uic option is specified
on the .tran control line.
Ngspice calculates the nominal capacitance as described below:
Cnom = value · scale · m (3.7)
The temperature coefficients tc1 and tc2 describe a quadratic temperature dependence
(see equation17.14) of the capacitance. If given in the instance line (the C... line) their
values will override the tc1 and tc2 of the .model line (3.3.8).
3.3.7 Semiconductor Capacitors
General form:
CXXXXXXX n + n - < value > < mname > <l = length > <w = width > <m = val >
+ < scale = val > < temp = val > < dtemp = val > < ic = init_condition >
Examples:
CLOAD 2 10 10 P
CMOD 3 7 CMODEL L =10 u W =1 u
This is the more general form of the Capacitor presented in section (3.3.6), and allows
for the calculation of the actual capacitance value from strictly geometric information
and the specifications of the process. If value is specified, it defines the capacitance and
both process and geometrical information are discarded. If value is not specified, the
capacitance is calculated from information contained model mname and the given length
and width (l, w keywords, respectively).
It is possible to specify mname only, without geometrical dimensions and set the capaci-
tance in the .model line (3.3.6).
3.3.8 Semiconductor Capacitor Model (C)
The capacitor model contains process information that may be used to compute the
capacitance from strictly geometric information.

CAP model capacitance F 0.0 1e-6
CJ junction bottom capacitance F/m2 - 5e-5
CJSW junction sidewall capacitance F/m - 2e-11
DEFW default device width m 1e-6 2e-6
DEFL default device length m 0.0 1e-6
NARROW narrowing due to side etching m 0.0 1e-7
SHORT shortening due to side etching m 0.0 1e-7
TC1 first order temperature coeff. F/◦ C 0.0 0.001
TC2 second order temperature coeff. F/◦ C 2 0.0 0.0001
◦
DI relative dielectric constant F/m - 1
THICK insulator thickness m 0.0 1e-9
The capacitor has a capacitance computed as:
If value is specified on the instance line then
Cnom = value · scale · m (3.8)
If model capacitance is specified then
Cnom = CAP · scale · m (3.9)
If neither value nor CAP are specified, then geometrical and physical parameters are take
into account:
C0 = CJ(l − SHORT)(w − NARROW) + 2CJSW(l − SHORT + w − NARROW) (3.10)
CJ can be explicitly given on the .model line or calculated by physical parameters. When
CJ is not given, is calculated as:
If THICK is not zero:
DI ϵ0
CJ = THICK
if DI is specified,
(3.11)
ϵSiO2
CJ = THICK
otherwise.
If the relative dielectric constant is not specified the one for SiO2 is used. The values
of the constants are ϵ0 = 8.854214871e − 12 m F
and ϵSiO2 = 3.4531479969e − 11 m
F
. The
nominal capacitance is then computed as:
Cnom = C0 scale m (3.12)
After the nominal capacitance is calculated, it is adjusted for temperature by the formula:

C(T ) = C(TNOM) 1 + T C1 (T − TNOM) + T C2 (T − TNOM)2 (3.13)
where C(TNOM) = Cnom .
In the above formula, ‘T ’ represents the instance temperature, which can be explicitly set
using the temp keyword or calculated using the circuit temperature and dtemp, if present.
3.3.9 Capacitors, dependent on expressions (behavioral capaci-

tor)
There are two forms for behavioral capacitors allowed:
1. Capacitance formulated expressions C = ’expression’
2. Charge formulated expressions Q = ’expression’
General form:
CXXXXXXX n + n - C = ’ expression ’ < tc1 = value > < tc2 = value >
CXXXXXXX n + n - ’ expression ’ < tc1 = value > < tc2 = value >
CXXXXXXX n + n - Q = ’ expression ’ < tc1 = value > < tc2 = value >
Examples:
C1 cc 0 c = ’V ( cc ) < { Vt } ? { C1 } : { Ch } ’ tc1 = -1e -03 tc2 =1.3 e -05

C1 a b q = ’1 u *(4* atan ( V (a , b )/4)*2+ V (a , b ))/3 ’

and hertz (5.1.2).
Example input file:
Behavioral Capacitor
. param Cl =5 n Ch =1 n Vt =1 m Il =100 n
. ic v ( cc ) = 0 v ( cc2 ) = 0
* capacitor depending on control voltage V ( cc )
C1 cc 0 c = ’V ( cc ) < { Vt } ? { Cl } : { Ch } ’
I1 0 1 { Il }
Exxx n1 - copy n2 n2 cc2 1
Cxxx n1 - copy n2 1
Bxxx cc2 n2 I = ’( V ( cc2 ) < { Vt } ? { Cl } : { Ch }) ’ * i ( Exxx )
I2 n2 22 { Il }
vn2 n2 0 DC 0
* measure charge by integrating current
aint1 % id (1 cc ) 2 time_count
aint2 % id (22 cc2 ) 3 time_count
. model time_count int ( in_offset =0.0 gain =1.0
+ out_lower_limit = -1 e12 out_upper_limit =1 e12
+ limit_range =1 e -9 out_ic =0.0)
. control
unset askquit
tran 100 n 100 u
plot v (2)
plot v ( cc ) v ( cc2 )
. endc
. end
3.3.10 Inductors
General form:
LYYYYYYY n + n - < value > < mname > < nt = val > <m = val >
+ < scale = val > < temp = val > < dtemp = val > < tc1 = val >
+ < tc2 = val > < ic = init_condition >
Examples:
LLINK 42 69 1 UH
LSHUNT 23 51 10 U IC =15.7 MA
The inductor device implemented into ngspice has many enhancements over the original
one.n+ and n- are the positive and negative element nodes, respectively. value is the
inductance in Henry. The initial condition (a curremt through L) becomes effective when
the uic parameter is set on the .tran line. Inductance can be specified in the instance
line as in the examples above or in a .model line, as in the example below:
L1 15 5 indmod1
L2 2 7 indmod1
. model indmod1 L ind =3 n
Both inductors have an inductance of 3nH.

The nt is used in conjunction with a .model line, and is used to specify the number of
turns of the inductor. If you want to simulate temperature dependence of an inductor,
you need to specify its temperature coefficients, using a .model line, like in the example
below:
Lload 1 2 1 u ind1 dtemp =5

. MODEL ind1 L tc1 =0.001
The (optional) initial condition is the initial (time zero) value of inductor current (in
Amps) that flows from n+, through the inductor, to n-. Note that the initial conditions
(if any) apply only if the UIC option is specified on the .tran analysis line.
Ngspice calculates the nominal inductance as described below:
value scale
Lnom = (3.14)
m
3.3.11 Inductor model

The inductor model contains physical and geometrical information that may be used to
compute the inductance of some common topologies like solenoids and toroids, wound in
air or other material with constant magnetic permeability.
IND model inductance H 0.0 1e-3
CSECT cross section m2 0.0 1e-6
DIA coil diameter m 0.0 1e-3
LENGTH length m 0.0 1e-2
TC1 first order temperature coeff. H/◦ C 0.0 0.001
TC2 second order temperature coeff. H/◦ C 2 0.0 0.0001
◦
NT number of turns - 0.0 10
MU relative magnetic permeability - 1.0 -
The inductor’s inductance is computed as follows:
If value is specified on the instance line then
value scale
Lnom = (3.15)
m
If model inductance is specified then
IND scale
Lnom = (3.16)
m
If neither value nor IND are specified, then geometrical and physical parameters are taken
into account. In the following formulas
NT refers to both instance and model parameter (instance parameter overrides model
parameter):
If LENGTH is not zero:
(
MU µ0 NT2 π DIA2
Lnom = 4 LENGTH
if DIA is specified,
MU µ0 NT2 CSECT (3.17)
Lnom = LENGTH
otherwise.
with µ0 = 1.25663706143592 µH m
. DIA takes preference over CSECT. kl is the geometry
correction factor according to Lundin (see D. W. Knight, pp. 35-36), which is important
when inductor length and diameter have the same order of magnitude. After the nominal
inductance is calculated, it is adjusted for temperature by the formula

L(T ) = L(TNOM) 1 + T C1 (T − TNOM) + T C2 (T − TNOM) , 2
(3.18)
where L(TNOM) = Lnom . In the above formula, ‘T ’ represents the instance tempera-
ture, which can be explicitly set using the temp keyword or calculated using the circuit
temperature and dtemp, if present.
3.3.12 Coupled (Mutual) Inductors

General form:
KXXXXXXX LYYYYYYY LZZZZZZZ value
Examples:
K43 LAA LBB 0.999

KXFRMR L1 L2 0.87
LYYYYYYY and LZZZZZZZ are the names of the two coupled inductors, and value is
the coefficient of coupling, K, which must be greater than 0 and less than or equal to 1.
Using the ‘dot’ convention for drawing the coupled inductors, place a ‘dot’ on the first
node of each inductor. If you have more than two inductors interacting, pairwise coupling
is supported.
Pairwise coupling of more than two inductors:
L1 1 0 10 u
L2 2 0 11 u
L3 3 0 10 u
K12 L1 L2 0.99
K23 L2 L3 0.99
K13 L1 L3 0.98
When there are more than two inductors coupled for interaction, some combination of
coupling constants are not possible physically because the magnetic fields then would
violate energy conservation. ngspice checks the coupling matrix for such conditions and
issues a warning.
3.3.13 Inductors, dependent on expressions (behavioral induc-

tor)
General form:
LXXXXXXX n + n - L = ’ expression ’ < tc1 = value > < tc2 = value >
LXXXXXXX n + n - ’ expression ’ < tc1 = value > < tc2 = value >
Examples:
L1 l2 lll L = ’i ( Vm ) < { It } ? { Ll } : { Lh } ’ tc1 = -4e -03 tc2 =6 e -05

and hertz (5.1.2).
Variable inductor
. param Ll =0.5 m Lh =5 m It =50 u Vi =2 m
. ic v ( int21 ) = 0
* variable inductor depending on control current i ( Vm )

L1 l2 lll L = ’i ( Vm ) < { It } ? { Ll } : { Lh } ’
* measure current through inductor
vm lll 0 dc 0
* voltage on inductor
V1 l2 0 { Vi }
* fixed inductor
L3 33 331 { Ll }
vm33 331 0 dc 0
* voltage on inductor
V3 33 0 { Vi }
* non linear inductor ( discrete setup )

F21 int21 0 B21 -1
L21 int21 0 1
B21 n1 n2 V = ’( i ( Vm21 ) < { It } ? { Ll } : { Lh }) ’ * v ( int21 )
vm21 n2 0 dc 0
V21 n1 0 { Vi }
. control
unset askquit
tran 1 u 100 u uic
plot i ( Vm ) i ( vm33 )
plot i ( vm21 ) i ( vm33 )
plot i ( vm ) - i ( vm21 )
. endc
. end
3.3.14 Capacitor or inductor with initial conditions

The simulator supports the specification of voltage and current initial conditions on capac-
itor and inductor models, respectively. These models are not the standard ones supplied
with SPICE3, but are in fact code models that can be substituted for the SPICE models
when realistic initial conditions are required. For details please refer to Chapter 12. A
XSPICE deck example using these models is shown below:
*
* This circuit contains a capacitor and an inductor with
* initial conditions on them. Each of the components

* has a parallel resistor so that an exponential decay
* of the initial condition occurs with a time constant of
* 1 second.
*
a1 1 0 cap
.model cap capacitoric (c=1000uf ic=1)
r1 1 0 1k
*
a2 2 0 ind
.model ind inductoric (l=1H ic=1)
r2 2 0 1.0
*
.control
tran 0.01 3
plot v(1) v(2)
.endc
.end
3.3.15 Switches
Two types of switches are available: a voltage controlled switch (type SXXXXXX, model
SW) and a current controlled switch (type WXXXXXXX, model CSW). A switching
hysteresis may be defined, as well as on- and off-resistances (0 < R < ∞).
General form:
SXXXXXXX N + N - NC + NC - MODEL <ON > < OFF >

WYYYYYYY N + N - VNAM MODEL <ON > < OFF >
Examples:
s1 1 2 3 4 switch1 ON
s2 5 6 3 0 sm2 off
Switch1 1 2 10 0 smodel1
w1 1 2 vclock switchmod1
W2 3 0 vramp sm1 ON
wreset 5 6 vclck lossyswitch OFF
Nodes 1 and 2 are the nodes between which the switch terminals are connected. The model
name is mandatory while the initial conditions are optional. For the voltage controlled
switch, nodes 3 and 4 are the positive and negative controlling nodes respectively. For
the current controlled switch, the controlling current is that through the specified voltage
source. The direction of positive controlling current flow is from the positive node, through
the source, to the negative node.
The instance parameters ON or OFF are required, when the controlling voltage (cur-
rent) starts inside the range of the hysteresis loop (different outputs during forward vs.
backward voltage or current ramp). Then ON or OFF determine the initial state of the
switch.
3.3.16 Switch Model (SW/CSW)
The switch model allows an almost ideal switch to be described in ngspice. The switch is
not quite ideal, in that the resistance can not change from 0 to infinity, but must always
have a finite positive value. By proper selection of the on and off resistances, they can
be effectively zero and infinity in comparison to other circuit elements. The parameters
available are shown below.
Name Parameter Units Default Switch model
VT threshold voltage V 0.0 SW
IT threshold current A 0.0 CSW
VH hysteresis voltage V 0.0 SW
IH hysteresis current A 0.0 CSW
RON on resistance Ω 1.0 SW,CSW
ROFF off resistance Ω 1.0e+12 (*) SW,CSW
(*) Or 1/GM IN , if you have set GM IN to any other value, see the .OPTIONS control
line (15.1.2) for a description of GM IN , its default value results in an off-resistance of
1.0e+12 ohms.
The use of an ideal element that is highly nonlinear such as a switch can cause large
discontinuities to occur in the circuit node voltages. A rapid change such as that associated
with a switch changing state can cause numerical round-off or tolerance problems leading
to erroneous results or time step difficulties. The user of switches can improve the situation
by taking the following steps:
• First, it is wise to set the ideal switch impedance just high or low enough to be
negligible with respect to other circuit elements. Using switch impedances that
are close to ‘ideal’ in all cases aggravates the problem of discontinuities mentioned
above. Of course, when modeling real devices such as MOSFETS, the on resistance
should be adjusted to a realistic level depending on the size of the device being
modeled.
• If a wide range of ON to OFF resistance must be used in the switches (ROFF/RON

> 1e+12), then the tolerance on errors allowed during transient analysis should be
decreased by using the .OPTIONS control line and specifying TRTOL to be less than
the default value of 7.0.
• When switches are placed around capacitors, then the option CHGTOL should also
be reduced. Suggested values for these two options are 1.0 and 1e-16 respectively.
These changes inform ngspice to be more careful around the switch points so that
no errors are made due to the rapid change in the circuit.
Switch test
. tran 2 us 5 ms
* switch control voltage
v1 1 0 DC 0.0 PWL (0 0 2e -3 2 4e -3 0)
* switch control voltage starting inside hysteresis window
* please note influence of instance parameters ON , OFF
v2 2 0 DC 0.0 PWL (0 0.9 2e -3 2 4e -3 0.4)
* switch control current
i3 3 0 DC 0.0 PWL (0 0 2e -3 2 m 4e -3 0) $ <--- switch control current
* load voltage
v4 4 0 DC 2.0
* input load for current source i3
r3 3 33 10 k
vm3 33 0 dc 0 $ <--- measure the current
* ouput load resistors
r10 4 10 10 k
r20 4 20 10 k
r30 4 30 10 k
r40 4 40 10 k
*
s1 10 0 1 0 switch1 OFF
s2 20 0 2 0 switch1 OFF
s3 30 0 2 0 switch1 ON
. model switch1 sw vt =1 vh =0.2 ron =1 roff =10 k
*
w1 40 0 vm3 wswitch1 off
. model wswitch1 csw it =1 m ih =0.2 m ron =1 roff =10 k
*
. control
run
plot v (1) v (10)
plot v (10) vs v (1) $ <-- get hysteresis loop
plot v (2) v (20) $ <--- different initial values
plot v (2) v (30) $ <--- different initial values
plot v (40) vs vm3 # branch $ <--- current controlled switch hysteresi
. endc
. end
Chapter 4
Voltage and Current Sources
4.1 Independent Sources for Voltage or Current

General form:
VXXXXXXX N + N - <<DC > DC / TRAN VALUE > < AC < ACMAG < ACPHASE > > >
+ < DISTOF1 < F1MAG < F1PHASE > > > < DISTOF2 < F2MAG < F2PHASE > > >
IYYYYYYY N + N - <<DC > DC / TRAN VALUE > < AC < ACMAG < ACPHASE > > >
+ < DISTOF1 < F1MAG < F1PHASE > > > < DISTOF2 < F2MAG < F2PHASE > > >
Examples:
VCC 10 0 DC 6
VIN 13 2 0.001 AC 1 SIN (0 1 1 MEG )
ISRC 23 21 AC 0.333 45.0 SFFM (0 1 10 K 5 1 K )
VMEAS 12 9
VCARRIER 1 0 DISTOF1 0.1 -90.0
VMODULATOR 2 0 DISTOF2 0.01
IIN1 1 5 AC 1 DISTOF1 DISTOF2 0.001
n+ and n- are the positive and negative nodes, respectively. Note that voltage sources
need not be grounded. Positive current is assumed to flow from the positive node, through
the source, to the negative node. A current source of positive value forces current to flow
out of the n+ node, through the source, and into the n- node. Voltage sources, in addition
to being used for circuit excitation, are the ‘ammeters’ for ngspice, that is, zero valued
voltage sources may be inserted into the circuit for the purpose of measuring current.
They of course have no effect on circuit operation since they represent short-circuits.
DC/TRAN is the dc and transient analysis value of the source. If the source value is zero
both for dc and transient analyses, this value may be omitted. If the source value is
time-invariant (e.g., a power supply), then the value may optionally be preceded by the
letters DC.
The keyword AC together with its value ACMAG (and optional value ACPHASE) are required
when the voltage or current source is intended to become the small signal source in an
93
94 CHAPTER 4. VOLTAGE AND CURRENT SOURCES
ac simulation. ACMAG is the ac magnitude and ACPHASE is the ac phase. The voltage or
current source then will become a reference for all nodes. All small signal node amplitude
values obtained after the simulation have been divided by the reference ACMAG. A typcal
ACMAG value thus may be unity. Any measured phase has been shifted by ACPHASE. If
ACPHASE is omitted, a value of zero is assumed. If the source is not an ac small-signal
input, the keyword AC and the ac values are to be avoided.
DISTOF1 and DISTOF2 are the keywords that specify that the independent source has
distortion inputs at the frequencies F1 and F2 respectively (see the description of the
.DISTO control line). The keywords may be followed by an optional magnitude and
phase. The default values of the magnitude and phase are 1.0 and 0.0 respectively.
Any independent source can be assigned a time-dependent value for transient analysis. If
a source is assigned a time-dependent value, the time-zero value is used for dc analysis.
There are nine independent source functions:
• pulse,
• exponential,
• sinusoidal,
• piece-wise linear,
• single-frequency FM,
• AM,
• transient noise,
• random voltages or currents,
• external data (only with ngspice shared library),
• and RF port
If parameters other than source values are omitted or set to zero, the default values shown
are assumed. TSTEP is the printing increment and TSTOP is the final time – see the .TRAN
control line for an explanation.
4.1.1 Pulse
General form:
PULSE ( V1 V2 TD TR TF PW PER NP )
Examples:
VIN 3 0 PULSE ( -1 1 2 NS 2 NS 2 NS 50 NS 100 NS 5)

4.1. INDEPENDENT SOURCES FOR VOLTAGE OR CURRENT 95
Name Parameter Default Value Units

V1 Initial value - V, A
V2 Pulsed value - V, A
TD Delay time 0.0 sec
TR Rise time TSTEP sec
TF Fall time TSTEP sec
PW Pulse width TSTOP sec
PER Period TSTOP sec
NP Number of Pulses *) unlimited -
A single pulse, without repetition count or phase offset, is described by the following
table:
Time Value
0 V1
TD V1
TD+TR V2
TD+TR+PW V2
TD+TR+PW+TF V1
TSTOP V1
Intermediate points are determined by linear interpolation.
*) NP set to 0 or omitted denotes unlimited pulses. If compatibility mode (see 16.14.1)
set ngbehavior=xs is set in .spiceinit, the 8th parameter is the phase of the pulse signal
(in degrees), which results in forward running (pos. value) or a delay (neg. value) of the
pulse sequence.
4.1.2 Sinusoidal
General form:
SIN ( VO VA FREQ TD THETA PHASE )
Examples:
VIN 3 0 SIN (0 1 100 MEG 1 NS 1 E10 )

VO Offset - V, A
VA Amplitude - V, A
FREQ Frequency 1/T ST OP Hz
TD Delay 0.0 sec
THETA Damping factor 0.0 1/sec
PHASE Phase 0.0 degrees

The shape of the waveform is described by the following formula:
(
V0 if 0 ≤ t < T D
V (t) =
V 0 + V A e−(t−T D)T HET A sin (2π · F REQ · (t − T D) + P HASE) if T D ≤ t < T ST OP.
(4.1)
4.1.3 Exponential
General form:
EXP ( V1 V2 TD1 TAU1 TD2 TAU2 )
Examples:
VIN 3 0 EXP ( -4 -1 2 NS 30 NS 60 NS 40 NS )

V1 Initial value - V, A
V2 pulsed value - V, A
TD1 rise delay time 0.0 sec
TAU1 rise time constant TSTEP sec
TD2 fall delay time TD1+TSTEP sec
TAU2 fall time constant TSTEP sec
The shape of the waveform is described by the following formula:
Let V 21 = V 2 − V 1, V 12 = V 1 − V 2:

V 1
 if 0 ≤ t < T D1,
 (t−T D1)
− T AU 1
V (t) = V 1 + V 21 1 − e if T D1 ≤ t < T D2,


 (t−T D1) (t−T D2)
V 1 + V 21 1 − e− T AU 1 + V 12 1 − e− T AU 2 if T D2 ≤ t < T ST OP.
(4.2)
4.1.4 Piece-Wise Linear
General form:
PWL ( T1 V1 < T2 V2 T3 V3 T4 V4 ... >) <r = value > < td = value >
Examples:
VCLOCK 7 5 PWL (0 -7 10 NS -7 11 NS -3 17 NS -3 18 NS -7 50 NS -7)

+ r =0 td =15 NS
Each pair of values (Ti , Vi ) specifies that the value of the source is Vi (in Volts or Amps)
at time = Ti . The value of the source at intermediate values of time is determined by
using linear interpolation on the input values. The parameter r determines a repeat time
point. If r is set to -1 or is not given, the whole sequence of values (Ti , Vi ) is issued once
only, then the output stays at its final value. If r = 0, the whole sequence from time 0 to
time Tn is repeated forever. If r = 10ns, the sequence between 10ns and 50ns is repeated
forever. The r value has to be one of the time points T1 to Tn of the PWL sequence. If
td is given, the whole PWL sequence is delayed by the value of td. Please note that for
now r and td are available only with the voltage source, not with the current source.
4.1.5 Single-Frequency FM
General Form:
SFFM ( VO VA FM MDI FC TD PHASEM PHASEC )
Examples:
V1 12 0 SFFM (0 2 20 45 1 k 1 m 0 0)
Name Parameter Default value Units

VO Offset - V, A
VA Amplitude - V, A
FM Modulating frequency 5/T ST OP Hz
MDI Modulation index 90
FC Carrier frequency 500/T ST OP Hz
TD Signal delay 0.0 s
PHASEM Modulation signal phase 0.0 degrees
PHASEC Carrier signal phase 0.0 degrees
The shape of the waveform is described by the following equation:
V (t) = VO + VA ·
sin (2π · F C · (t − T D) + M DI sin (2π · F S · (t − T D) + P HASEM ) + P HASEC)
(4.3)
with t > T D, else V (t) = 0.
MDI is limited to 0 <= M DI <= F C/F M . VO and VA have to be given always.

4.1.6 Amplitude modulated source (AM)

General form:
AM ( VO VMO VMA FM FC TD PHASEM PHASEC )
Examples:
V1 12 0 AM (0.5 2 1.8 20 K 5 MEG 1 m )

VO Overall offset - V, A
VMO Modulation signal offset - V, A
VMA Modulation signal amplitude 1 V, A
FM Modulation signal frequency 5/T ST OP Hz
FC Carrier signal frequency 500/T ST OP Hz
TD Overall delay 0.0 s
PHASEM Modulation signal phase 0.0 degrees
PHASEC Carrier signal phase 0.0 degrees
The shape of the waveform is described by the following equation:
V (t) = V O + (V M O + V M A · sin (2π · F M · (t − T D) + P HASEM )) ·

sin (2π · F C · (t − T D) + P HASEC) (4.4)
with t > T D, else V (t) = 0.

VO and VMO have to be given always.
With the modulation depth, given by V M A/V M O, varied between 0 and 1, a standard
amplitude modulated signal is provided. VMO then also acts as overall multiplier to the
signal. On the other hand one may set VMO to 0, then obtaining a signal with double
side band and suppressed carrier.
4.1.7 Transient noise source

General form:
TRNOISE ( NA NT NALPHA NAMP RTSAM RTSCAPT RTSEMT )
Examples:
VNoiw 1 0 DC 0 TRNOISE (20 n 0.5 n 0 0) $ white

VNoi1of 1 0 DC 0 TRNOISE (0 10 p 1.1 12 p ) $ 1/ f
VNoiw1of 1 0 DC 0 TRNOISE (20 10 p 1.1 12 p ) $ white and 1/ f
IALL 10 0 DC 0 trnoise (1 m 1 u 1.0 0.1 m 15 m 22 u 50 u )
$ white , 1/ f , RTS
Transient noise is an experimental feature allowing (low frequency) transient noise injec-
tion and analysis. See Chapt. 15.3.11 for a detailed description. NA is the Gaussian noise
rms voltage amplitude, NT is the time between sample values (breakpoints will be en-
forced on multiples of this value). NALPHA (exponent to the frequency dependency), NAMP
(rms voltage or current amplitude) are the parameters for 1/f noise, RTSAM the random
telegraph signal amplitude, RTSCAPT the mean of the exponential distribution of the trap
capture time, and RTSEMT its emission time mean. White Gaussian, 1/f, and RTS noise
may be combined into a single statement.
NA Rms noise amplitude (Gaussian) - V, A
NT Time step - sec
NALPHA 1/f exponent 0<α<2 -
NAMP Amplitude (1/f) - V, A
RTSAM Amplitude - V, A
RTSCAPT Trap capture time - sec
RTSEMT Trap emission time - sec
If you set NT and RTSAM to 0, the noise option TRNOISE ... is ignored. Thus you may
switch off the noise contribution of an individual voltage source VNOI by the command
alter @vnoi[trnoise] = [ 0 0 0 0 ] $ no noise
alter @vrts[trnoise] = [ 0 0 0 0 0 0 0] $ no noise
See Chapt. 17.5.3 for the alter command.
You may switch off all TRNOISE noise sources by setting
set notrnoise
to your .spiceinit file (for all your simulations) or into your control section in front of the
next run or tran command (for this specific and all following simulations). The command
unset notrnoise
will reinstate all noise sources.
The noise generators are implemented into the independent voltage (vsrc) and current
(isrc) sources.
4.1.8 Random voltage source

The TRRANDOM option yields statistically distributed voltage values, derived from the
ngspice random number generator. These values may be used in the transient simula-
tion directly within a circuit, e.g. for generating a specific noise voltage, but especially
they may be used in the control of behavioral sources (B, E, G sources 5, voltage control-
lable A sources 12, capacitors 3.3.9, inductors 3.3.13, or resistors 3.3.4) to simulate the
circuit dependence on statistically varying device parameters. A Monte-Carlo simulation
may thus be handled in a single simulation run.
General form:
TRRANDOM ( TYPE TS < TD < PARAM1 < PARAM2 > > >)
Examples:
VR1 r1 0 dc 0 trrandom (2 10 m 0 1) $ Gaussian
TYPE determines the random variates generated: 1 is uniformly distributed, 2 Gaussian,

3 exponential, 4 Poisson. TS is the duration of an individual voltage value. TD is a time
delay with 0 V output before the random voltage values start up. PARAM1 and PARAM2
depend on the type selected.
TYPE description PARAM1 default PARAM2 default
1 Uniform Range 1 Offset 0
2 Gaussian Standard Dev. 1 Mean 0
3 Exponential Mean 1 Offset 0
4 Poisson Lambda 1 Offset 0
4.1.9 External voltage or current input

General form:
EXTERNAL
Examples:
Vex 1 0 dc 0 external
Iex i1 i2 dc 0 external <m = xx >
Voltages or currents may be set from the calling process, if ngspice is compiled as a shared
library and loaded by the process. See Chapt. 19.6.3 for an explanation.
4.1.10 Arbitrary Phase Sources

ngspice supports arbitrary phase independent sources that output at TIME=0.0 a value
corresponding to some specified phase shift. Other versions of SPICE use the TD (delay
time) parameter to set phase-shifted sources to their time-zero value until the delay time
has elapsed. The ngspice phase parameter is specified in degrees and is included after the
SPICE3 parameters normally used to specify an independent source. Partial examples of
usage for pulse and sine waveforms are shown below:
* Phase shift is specified as final parameter

* on the independent source cards. Phase shift for both of the
* following is specified as +45 degrees
*
v1 1 0 0.0 sin(0 1 1k 0 0 45.0)
r1 1 0 1k
*
v2 2 0 0.0 pulse(-1 1 0 1e-5 1e-5 5e-4 1e-3 45.0)
r2 2 0 1k
*
4.2. LINEAR DEPENDENT SOURCES 101
4.1.11 RF Port
A voltage source VSRC may be defined as RF Port. To do so, there are at least two
more parameters required. The first is portnum (integer) which defines the VSRC as a
RF Port. Portnum of all VSRCs defined as RF ports must start from 1 and count up to
the number of RF ports. You cannot have duplicate portnums. Then you have Z0 (real)
which defines the internal impedance. If not provided, its default value is 50Ohm. When
declaring a RF ports, the VSRC now become a VSRC with Z0 Ohm in series. This extra
resistor affects all simulations.
General form:
DC 0 AC 1 portnum n1 < z0 n2 >
Examples:
V1 in 0 dc 0 ac 1 portnum 1 z0 100
At least two ports are required for the S-parameter simulation with the command .sp
(15.3.8). If portnum is not provided, the voltage source VRSC behaves as normal.
4.2 Linear Dependent Sources

Ngspice allows circuits to contain linear dependent sources characterized by any of the
four equations
i = gv v = ev i = fi v = hi
where g, e, f , and h are constants representing transconductance, voltage gain, current
gain, and transresistance, respectively. Non-linear dependent sources for voltages or cur-
rents (B, E, G) are described in Chapt. 5.
4.2.1 Gxxxx: Linear Voltage-Controlled Current Sources (VCCS)

General form:
GXXXXXXX N + N - NC + NC - VALUE <m = val >
Examples:
G1 2 0 5 0 0.1
n+ and n- are the positive and negative nodes, respectively. Current flow is from the
positive node, through the source, to the negative
node. nc+ and nc- are the positive and negative controlling nodes, respectively. value
is the transconductance (in mhos). m is an optional multiplier to the output current. val
may be a numerical value or an expression according to 2.9.5 containing references to
other parameters. Instance parameters are listed in chapt. 31.3.6.
4.2.2 Exxxx: Linear Voltage-Controlled Voltage Sources (VCVS)

General form:
EXXXXXXX N + N - NC + NC - VALUE
Examples:
E1 2 3 14 1 2.0
n+ is the positive node, and n- is the negative node. nc+ and nc- are the positive and
negative controlling nodes, respectively. value is the voltage gain. Instance parameters
are listed in chapt. 31.3.7.
4.2.3 Fxxxx: Linear Current-Controlled Current Sources (CCCS)

General form:
FXXXXXXX N + N - VNAM VALUE <m = val >
Examples:
F1 13 5 VSENS 5 m =2
n+ and n- are the positive and negative nodes, respectively. Current flow is from the
positive node, through the source, to the negative node. vnam is the name of a voltage
source through which the controlling current flows. The direction of positive controlling
current flow is from the positive node, through the source, to the negative node of vnam.
value is the current gain. m is an optional multiplier to the output current. Instance
parameters are listed in chapt. 31.3.4.
4.2.4 Hxxxx: Linear Current-Controlled Voltage Sources (CCVS)

General form:
HXXXXXXX N + N - VNAM VALUE
Examples:
HX 5 17 VZ 0.5 K
n+ and n- are the positive and negative nodes, respectively. vnam is the name of a voltage
source through which the controlling current flows. The direction of positive controlling
current flow is from the positive node, through the source, to the negative node of vnam.
value is the transresistance (in ohms). Instance parameters are listed in chapt. 31.3.5.
4.2. LINEAR DEPENDENT SOURCES 103
4.2.5 Polynomial Source Compatibility

Dependent polynomial sources available in SPICE2G6 are fully supported in ngspice using
the XSPICE extension (25.1). The form used to specify these sources is shown in Table
4.1. For details on its usage please see Chapt. 5.5.
Dependent Polynomial Sources

Source Type Instance Card
POLYNOMIAL VCVS EXXXXXXX N+ N- POLY(ND) NC1+ NC1- P0 (P1...)
POLYNOMIAL VCCS GXXXXXXX N+ N- POLY(ND) NC1+ NC1- P0 (P1...)
POLYNOMIAL CCCS FXXXXXXX N+ N- POLY(ND) VNAM1 !VNAM2...? P0 (P1...)
POLYNOMIAL CCVS HXXXXXXX N+ N- POLY(ND) VNAM1 !VNAM2...? P0 (P1...)
Table 4.1: Dependent Polynomial Sources

Chapter 5
Non-linear Dependent Sources

(Behavioral Sources)
The non-linear dependent sources B ( see Chapt. 5.1), E (see 5.2), G see (5.3) described
in this chapter allow the generation of voltages or currents that result from evaluating a
mathematical expression. Internally E and G sources are converted to the more general
B source. All three sources may be used to introduce behavioral modeling and analysis.
5.1 Bxxxx: Nonlinear dependent source (ASRC)
5.1.1 Syntax and usage
General form:
BXXXXXXX n + n - <i = expr > <v = expr > < tc1 = value > < tc2 = value >
+ < temp = value > < dtemp = value >
Examples:
B1 0 1 I = cos ( v (1))+ sin ( v (2))

B2 0 1 V = ln ( cos ( log ( v (1 ,2)^2))) - v (3)^4+ v (2)^ v (1)
B3 3 4 I =17
B4 3 4 V = exp ( pi ^ i ( vdd ))
B5 2 0 V = V (1) < { Vlow } ? { Vlow } :
+ V (1) > { Vhigh } ? { Vhigh } : V (1)
n+ is the positive node, and n- is the negative node. The values of the V and I parameters
determine the voltages and currents across and through the device, respectively. If I is
given then the device is a current source, and if V is given the device is a voltage source.
One and only one of these parameters must be given. All instance parameters are listed
in chapter 31.3.1.
A simple model is implemented for temperature behavior by the formula:
105
106CHAPTER 5. NON-LINEAR DEPENDENT SOURCES (BEHAVIORAL SOURCES)

I(T ) = I(TNOM) 1 + T C1 (T − TNOM) + T C2 (T − TNOM)2 (5.1)
or

V (T ) = V (TNOM) 1 + T C1 (T − TNOM) + T C2 (T − TNOM)2 (5.2)
In the above formula, ‘T ’ represents the instance temperature, which can be explicitly set
using the temp keyword or calculated using the circuit temperature and dtemp, if present.
If both temp and dtemp are specified, the latter is ignored.
The small-signal AC behavior of the nonlinear source is a linear dependent source (or
sources) with a proportionality constant equal to the derivative (or derivatives) of the
source at the DC operating point. The expressions given for V and I may be any function
of voltages and currents through voltage sources in the system.
The following functions of a single real variable are defined:
Trigonometric functions: cos, sin, tan, acos, asin, atan

Hyperbolic functions: cosh, sinh, acosh, asinh, atanh
Exponential and logarithmic: exp, ln, log, log10 (ln, log with base e, log10 with base
10)
Other: abs, sqrt, u, u2, uramp, floor, ceil, i
Functions of two variables are min, max, pow, **, pwr, ^
Functions of three variables are a ? b:c
For convergence reasons the ‘exp’ function has a limit of 14 for its argument, beyond that
value it will increase linearily. The function ‘u’ is the unit step function, with a value
of one for arguments greater than zero, a value of 0.5 at zero, and a value of zero for
arguments less than zero. The function ‘u2’ returns a value of zero for arguments less
than zero, one for arguments greater than one and assumes the value of the argument
between these limits. The function ‘uramp’ is the integral of the unit step: for an input x,
the value is zero if x is less than zero, or, if x is greater than or equal to zero, the value is
x. These three functions are useful in synthesizing piece-wise non-linear functions, though
convergence may be adversely affected.
The function i(xyz) returns the current through the first node of device instance xyz.
The following standard operators are defined: +, -, *, /, ^, unary -
Logical operators are !=, <>, >=, <=, ==, >, <, ||, &&, ! .
A ternary function is defined as a ? b : c , which means IF a, THEN b, ELSE c. Be
sure to place a space in front of ‘?’ to allow the parser distinguishing it from other tokens.
The B source functions pow, **, ^, and pwr need some special care to avoid undefined
regions in x1, as they differ from the common mathematical usage (and from the functions
depicted in chapt. 2.9.5).
The functions y = pow(x1,x2), x1**x2, and x1^x2 , all of them describing y = x1x2 ,
resolve to the following:
5.1. BXXXX: NONLINEAR DEPENDENT SOURCE (ASRC) 107
y = pow(fabs(x1), x2)
pow in the preceding line is the standard C math library function.
The function y = pwr(x1,x2) resolves to
if (x1 < 0.0)

y = (-pow(-x1, x2));
else
y = (pow(x1, x2));
pow here again is the standard C math library function.
Example: Ternary function
* B source test Clamped voltage source

* C . P . Basso " Switched - mode power supplies " , New York , 2008
. param Vhigh = 4.6
. param Vlow = 0.4
Vin1 1 0 DC 0 PWL (0 0 1 u 5)
Bcl 2 0 V = V (1) < Vlow ? Vlow : V (1) > Vhigh ? Vhigh : V (1)
. control
unset askquit
tran 5 n 1 u
plot V (2) vs V (1)
. endc
. end
If the argument of log, ln, or sqrt becomes less than zero, the absolute value of the
argument is used. If a divisor becomes zero or the argument of log or ln becomes zero,
an error will result. Other problems may occur when the argument for a function in a
partial derivative enters a region where that function is undefined.
Parameters may be used like {Vlow} shown in the example above. Parameters will be
evaluated upon set up of the circuit, vectors like V(1) will be evaluated during the simu-
lation.
To get time into the expression you can integrate the current from a constant current
source with a capacitor and use the resulting voltage (don’t forget to set the initial voltage
across the capacitor).
Non-linear resistors, capacitors, and inductors may be synthesized with the nonlinear de-
pendent source. Nonlinear resistors, capacitors and inductors are implemented with their
linear counterparts by a change of variables implemented with the nonlinear dependent
source. The following subcircuit will implement a nonlinear capacitor:
Example: Non linear capacitor
. Subckt nlcap pos neg

* Bx : calculate f ( input voltage )
Bx 1 0 v = f ( v ( pos , neg ))
* Cx : linear capacitance
Cx 2 0 1
* Vx : Ammeter to measure current into the capacitor
Vx 2 1 DC 0 Volts
* Drive the current through Cx back into the circuit
Fx pos neg Vx 1
. ends
Example for f(v(pos,neg)):
Bx 1 0 V = v ( pos , neg )* v ( pos , neg )
Non-linear resistors or inductors may be described in a similar manner. An example for

a nonlinear resistor using this template is shown below.
Example: Non linear resistor
* use of ’ hertz ’ variable in nonlinear resistor

*. param rbase =1 k
* some tests
B1 1 0 V = hertz * v (33)
B2 2 0 V = v (33)* hertz
b3 3 0 V = 6.283 e3 /( hertz +6.283 e3 )* v (33)
V1 33 0 DC 0 AC 1
*** Translate R1 10 0 R = ’1 k / sqrt ( HERTZ ) ’ to B source ***
. Subckt nlres pos neg rb = rbase
* Bx : calculate f ( input voltage )
Bx 1 0 v = -1 / { rb } / sqrt ( HERTZ ) * v ( pos , neg )
* Rx : linear resistance
Rx 2 0 1
Example: Non linear resistor (continued)
* Vx : Ammeter to measure current into the resistor

Vx 2 1 DC 0 Volts
* Drive the current through Rx back into the circuit
Fx pos neg Vx 1
. ends
Xres 33 10 nlres rb =1 k
* Rres 33 10 1 k
Vres 10 0 DC 0
. control
define check (a , b ) vecmax ( abs ( a - b ))
ac lin 10 100 1 k
* some checks
print v (1) v (2) v (3)
if check ( v (1) , frequency ) < 1e -12
echo " INFO : ok "
end
plot vres # branch
. endc
. end
5.1.2 Special B-Source Variables time, temper, hertz

The special variables time and temper are available in a transient analysis, reflecting the
actual simulation time and circuit temperature. temper returns the circuit temperature,
given in degree C (see 2.12). The variable hertz is available in an AC analysis. time
is zero in the AC analysis, hertz is zero during transient analysis. Using the variable
hertz may cost some CPU time if you have a large circuit, because for each frequency
the operating point has to be determined before calculating the AC response.
5.1.3 par(’expression’)
The B source syntax may also be used in output lines like .plot as algebraic expressions
for output (see Chapt.15.6.6 ).
5.1.4 Piecewise Linear Function: pwl

Both B source types may contain a piece-wise linear dependency of one network variable:
Example: pwl_current
Bdio 1 0 I = pwl ( v ( A ) , 0 ,0 , 33 ,10 m , 100 ,33 m , 200 ,50 m )
v(A) is the independent variable x. Each pair of values following describes the x,y func-
tional relation: In this example at node A voltage of 0V the current of 0A is generated -
next pair gives 10mA flowing from ground to node 1 at 33V on node A and so forth.
The same is possible for voltage sources:
Example: pwl_voltage
Blimit b 0 V = pwl ( v (1) , -4 ,0 , -2 ,2 , 2 ,4 , 4 ,5 , 6 ,5)
Monotony of the independent variable in the pwl definition is checked - non-monotonic x

entries will stop the program execution. v(1) may be replaced by a controlling current
source, or it may be replaced by time (for transient simulations). v(1) may also be
replaced by an expression, e.g. −2 ∗ i(Vin ). The value pairs may also be parameters, and
have to be predefined by a .param statement. An example for the pwl function using all
of these options is shown below.
Figure 5.1: pwl (piece-wise linear) B source

Example: pwl function in B source
Demonstrates usage of the pwl function in an B source ( ASRC )

* Also emulates the TABLE function with limits
. param x0 = -4 y0 =0
. param x1 = -2 y1 =2
. param x2 =2 y2 = -2
. param x3 =4 y3 =1
. param xx0 = x0 -1
. param xx3 = x3 +1
Vin ctrl 0 DC =0 V
R1 ctrl 0 2
* no limits outside of the tabulated x values

* ( continues linearily )
Btest2 2 0 I = pwl ( v ( ctrl ) , ’ x0 ’ , ’ y0 ’ , ’ x1 ’ , ’ y1 ’ , ’ x2 ’ , ’ y2 ’ ,
+ ’x3 ’ , ’ y3 ’)
* like TABLE function with limits :

Btest3 3 0 I = ( v ( ctrl ) < ’x0 ’) ? ’y0 ’ : ( v ( ctrl ) < ’x3 ’)
+ ? pwl ( v (1) , ’ x0 ’ , ’ y0 ’ , ’ x1 ’ , ’ y1 ’ , ’ x2 ’ , ’ y2 ’ , ’ x3 ’ , ’ y3 ’) : ’y3 ’
* more efficient and elegant TABLE function with limits

*( voltage controlled ):
Btest4 4 0 I = pwl ( v ( ctrl ) ,
+ ’xx0 ’ , ’ y0 ’ , ’x0 ’ , ’ y0 ’ ,
+ ’x1 ’ , ’ y1 ’ ,
+ ’x2 ’ , ’ y2 ’ ,
+ ’x3 ’ , ’ y3 ’ , ’xx3 ’ , ’ y3 ’)
*
* more efficient and elegant TABLE function with limits
* ( controlled by current ):
Btest5 5 0 I = pwl ( -2* i ( Vin ) ,
+ ’xx0 ’ , ’ y0 ’ , ’x0 ’ , ’ y0 ’ ,
+ ’x1 ’ , ’ y1 ’ ,
+ ’x2 ’ , ’ y2 ’ ,
+ ’x3 ’ , ’ y3 ’ , ’xx3 ’ , ’ y3 ’)
Rint2 2 0 1
Rint3 3 0 1
Rint4 4 0 1
Rint5 5 0 1
. control
dc Vin -6 6 0.2
plot v (2) v (3) v (4) -0.5 v (5)+0.5
. endc
. end
One characteristic to note: What happens when the controlling input (V(1) or −2∗i(Vin ))
is outside of the given limits, e.g. smaller than x0 or larger than x3 in the example given
above? New y values outside of the given range will be determined by adding x,y pairs
calculated by extending the slope of the output curve, e.g. with (y3 − y2)/(x3 − x2), as
seen with v(2) from example Btest2. If you want to limit the function, keeping the last
y value, e.g. y3, you have to add another point (x,y pair) with slightly extended x and y
kept constant, e.g. x3 + 1,y3.
This gets important when we are for example using a behavioral resistor with pwl. In
the example below, RR1 quickly moves towards (and beyond) 0, which is unphysical and
leads the transient simulation to fail, because the current through RR1 is unbounded.
RR2 with its limit given by the 15.1ms,1 couple avoids such malfunctioning.
Example: pwl function in behavioral resistor
* pwl for behavioral R , transient sim

VU1 3 0 DC 9
RR1 3 0 R = pwl ( time , 0 ,1 , 7m ,1 , 8m ,1.19 , 14 m ,1.19 , 15 m ,1)
RR2 3 0 R = pwl ( time , 0 ,1 , 7m ,1 , 8m ,1.19 , 14 m ,1.19 , 15 m ,1 ,
+ 15.1 m ,1)
. tran 100 u 20 m 0
. probe alli
. control
option noinit
run
display
set xbrushwidth =2
plot rr1 # branch rr2 # branch ylimit 7 17
. endc
. end
5.2 Exxxx: non-linear voltage source

5.2.1 VOL
General form:
EXXXXXXX n + n - vol = ’ expr ’
Examples:
E41 4 0 vol = ’V (3)* V (3) - Offs ’

5.2. EXXXX: NON-LINEAR VOLTAGE SOURCE 113
hertz (5.1.2). ’ or { } may be used to delimit the function.
5.2.2 VALUE
Optional syntax:
EXXXXXXX n + n - value ={ expr }
Examples:
E41 4 0 value = { V (3)* V (3) - Offs }
The ’=’ sign is optional.
5.2.3 TABLE
Data may be entered from the listings of a data table similar to the pwl B-Source (5.1.4).
Data are grouped into x, y pairs. Expression may be an equation or an expression
containing node voltages or branch currents (in the form of i(vm)) and any other terms
as given for the B source and described in Chapt. 5.1. It may contain parameters (2.9.1).
’ or { } may be used to delimit the function. Expression delivers the x-value, which
is used to generate a corresponding y-value according to the tabulated value pairs, using
linear interpolation. If the x-value is below x0 , y0 is returned, above x2 y2 is returned
(limiting function). The value pairs have to be real numbers, parameters are not allowed.
Syntax for data entry from table:
Exxx n1 n2 TABLE { expression } = ( x0 , y0 ) ( x1 , y1 ) ( x2 , y2 )
Example (simple comparator):
ECMP 11 0 TABLE { V (10 ,9)} = ( -5 mV , 0 V ) (5 mV , 5 V )
An ’=’ sign may follow the keyword TABLE.
5.2.4 POLY
see E-Source at Chapt. 5.5.
5.2.5 LAPLACE
Currently ngspice does not offer a direct E-Source element with the LAPLACE option.
There is however a XSPICE code model equivalent called s_xfer (see Chapt. 12.2.18),
which you may invoke manually. The XSPICE option has to be enabled (32.1). AC
(15.3.1) and transient analysis (15.3.10) is supported.
The following E-Source:
ELOPASS 4 0 LAPLACE { V (1)}

+ {5 * ( s /100 + 1) / ( s ^2/42000 + s /60 + 1)}
may be replaced by:
AELOPASS 1 int_4 filter1

. model filter1 s_xfer ( gain =5
+ num_coeff =[{1/100} 1]
+ den_coeff =[{1/42000} {1/60} 1]
+ int_ic =[0 0])
ELOPASS 4 0 int_4 0 1
where you have the voltage of node 1 as input, an intermediate output node int_4 and an
E-source as buffer to keep the name ‘ELOPASS’ available if further processing is required.
If the controlling expression is more complex than just a voltage node, you may add a
B-Source (5.1) for evaluating the expression before entering the A-device.
E-Source with complex controlling expression:
ELOPASS 4 0 LAPLACE { V (1)* v (2)} {10 / ( s /6800 + 1)}
may be replaced by:
BELOPASS int_1 0 V = V (1)* v (2)

AELOPASS int_1 int_4 filter1
. model filter1 s_xfer ( gain =10
+ num_coeff =[1]
+ den_coeff =[{1/6800} 1]
+ int_ic =[0])
ELOPASS 4 0 int_4 0 1
5.2.6 FREQ
Currently ngspice does not offer a direct E-Source element with the FREQ option but
it is implemented by a XSPICE code model equivalent called xfer (see 12.2.19) that is
automatically invoked by rewriting the netlist. The XSPICE option has to be enabled
(32.1) and only AC (15.3.1) analysis is supported.
5.3. GXXXX: NON-LINEAR CURRENT SOURCE 115
This E-Source:
EXFER 1 0 FREQ { V (20 ,21)}= DB

+(1.000000 e +07 Hz , 1.633257 e -07 , -1.859873 e +01)
+(1.025641 e +08 Hz , -4.165672 e +00 , -4.076855 e +02)
+(2.000000 e +08 Hz , -2.798303 e -05 , -7.519027 e +02)
produces a complex voltage determined by multiplying an input differential voltage (v(20,

21)) by a complex-valued PWL function of the simulation frequency (transfer function).
The DB keyword indicates that the second column is gain in db and the third is phase
in degrees. Alternative keywords are MAG (linear gain), RAD (phase in radians), DEG
(phase in degrees, already the default) or R_I (real and imaginary parts).
5.2.7 AND/OR/NAND/NOR
This form of E-source provides simple behavioural implementations of basic logic gates
with analog inputs and output. It is implemented by a XSPICE code model called
multi_input_pwl (see 12.2.10) that is automatically invoked by rewriting the netlist.
The XSPICE option has to be enabled (32.1).
This E-Source:
EAND out1 out0 and (2) in1 0 in2 0 (0.5 , 0) (2.8 , 3.3)
produces a differential output voltage determined by selecting the smallest of any number
of differential input voltages, and applying a PWL output function. Here “and(2)” deter-
mines the logic function and number of PWL points: output is zero for minimum input
voltage less than 0.5 and 3.3 for inputs greater than 2.8, with a linear ramp between.
The other three functions are similar: “or” selects the maximum input and “nand/nor”
reverse the order of PWL points. Only two points are supported.
An example circuit can be found at examples/digital/compare/adder_esource.cir.
5.3 Gxxxx: non-linear current source
5.3.1 CUR
General form:
GXXXXXXX n + n - cur = ’ expr ’ <m = val >
Examples:
G51 55 225 cur = ’V (3)* V (3) - Offs ’


in Chapt. 5.1. It may contain parameters (2.9.1) and special variables (5.1.2). m is an
optional multiplier to the output current. val may be a numerical value or an expression
according to 2.9.5 containing only references to other parameters (no node voltages or
branch currents!), because it is evaluated before the simulation commences.
5.3.2 VALUE
Optional syntax:
GXXXXXXX n + n - value = ’ expr ’ <m = val >
Examples:
G51 55 225 value = ’V (3)* V (3) - Offs ’
The ’=’ sign is optional.
5.3.3 TABLE
A data entry by a tabulated listing is available with syntax similar to the E-Source (see
Chapt. 5.2.3).
Syntax for data entry from table:
Gxxx n1 n2 TABLE { expression } =

+ ( x0 , y0 ) ( x1 , y1 ) ( x2 , y2 ) <m = val >
Example (simple comparator with current output and voltage control):
GCMP 0 11 TABLE { V (10 ,9)} = ( -5 MV , 0 V ) (5 MV , 5 V )

R 11 0 1 k
m is an optional multiplier to the output current. val may be a numerical value or an

expression according to 2.9.5 containing only references to other parameters (no node
voltages or branch currents!), because it is evaluated before the simulation commences.
An ’=’ sign may follow the keyword TABLE.
5.3.4 POLY
see E-Source at Chapt. 5.5.
5.3.5 LAPLACE
See E-Source, Chapt. 5.2.5 , for an equivalent code model replacement.
5.4. DEBUGGING A BEHAVIORAL SOURCE 117
5.3.6 FREQ
See E-Source, Chapt.5.2.6 , for an equivalent code model replacement.
5.3.7 Example
An example file is given below.

VCCS , VCVS , non - linear dependency

. param Vi =1
. param Offs = ’0.01* Vi ’
* VCCS depending on V (3)
B21 int1 0 V = V (3)* V (3)
G1 21 22 int1 0 1
* measure current through VCCS
vm 22 0 dc 0
R21 21 0 1
* new VCCS depending on V (3)
G51 55 225 cur = ’V (3)* V (3) - Offs ’
* measure current through VCCS
vm5 225 0 dc 0
R51 55 0 1
* VCVS depending on V (3)
B31 int2 0 V = V (3)* V (3)
E1 1 0 int2 0 1
R1 1 0 1
* new VCVS depending on V (3)
E41 4 0 vol = ’V (3)* V (3) - Offs ’
R4 4 0 1
* control voltage
V1 3 0 PWL (0 0 100 u { Vi })
. control
unset askquit
tran 10 n 100 u uic
plot i ( E1 ) i ( E41 )
plot i ( vm ) i ( vm5 )
. endc
. end
5.4 Debugging a behavioral source
The B, E, G, sources and the behavioral R, C, L elements are powerful tools to set up
user defined models. Unfortunately debugging these models is not very comfortable.
Example input file with bug (log(-2)):
B source debugging
V1 1 0 1
V2 2 0 -2
E41 4 0 vol = ’V (1)* log ( V (2)) ’
. control
tran 1 1
. endc
. end
The input file given above results in an error message:
Error: -2 out of range for log
In this trivial example, the reason and location for the bug is obvious. However, if you have
several equations using behavioral sources, and several occurrences of the log function,
then debugging is nearly impossible.
However, if the variable ngdebug (see 17.7) is set (e.g. in file .spiceinit), a more distinctive
error message is issued that (after some closer investigation) will reveal the location and
value of the buggy parameter.
Detailed error message for input file with bug (log(-2)):
Error : -2 out of range for log

calling PTeval , tree =
( v0 ) * ( log ( v1 ))
d / d v0 : log ( v1 )
d / d v1 : ( v0 ) * ((0.434294) / ( v1 ))
values : var0 = 1
var1 = -2
If variable strict_errorhandling (see 17.7) is set, ngspice exits after this message. If
not, gmin and source stepping may be started, typically without success.
5.5 POLY Sources
Polynomial sources are only available when the XSPICE option (see Chapt. 32) is enabled.
5.5. POLY SOURCES 119
5.5.1 E voltage source, G current source

General form:
EXXXX N + N - POLY ( ND ) NC1 + NC1 - ( NC2 + NC2 -...) P0 ( P1 ...)
Example:
ENONLIN 100 101 POLY (2) 3 0 4 0 0.0 13.6 0.2 0.005
POLY(ND) Specifies the number of dimensions of the polynomial. The number of pairs
of controlling nodes must be equal to the number of dimensions.
(N+) and (N-) nodes are output nodes. Positive current flows from the (+) node through
the source to the (-) node.
The <NC1+> and <NC1-> are in pairs and define a set of controlling voltages. A particular
node can appear more than once, and the output and controlling nodes need not be
different.
The example yields a voltage output controlled by two input voltages v(3,0) and v(4,0).
Four polynomial coefficients are given. The equivalent function to generate the output is:
0 + 13.6 * v(3) + 0.2 * v(4) + 0.005 * v(3) * v(3)
Generally you will set the equation according to
POLY(1) y = p0 + p1*X1 + p2*X1*X1 + p3*X1*X1*X1 + ...

POLY(2) y = p0 + p1*X1 + p2*X2 +
+ p3*X1*X1 + p4*X2*X1 + p5*X2*X2 +
+ p6*X1*X1*X1 + p7*X2*X1*X1 + p8*X2*X2*X1 +
+ p9*X2*X2*X2 + ...
POLY(3) y = p0 + p1*X1 + p2*X2 + p3*X3 +
+ p4*X1*X1 + p5*X2*X1 + p6*X3*X1 +
+ p7*X2*X2 + p8*X2*X3 + p9*X3*X3 + ...
where X1 is the voltage difference of the first input node pair, X2 of the second pair and
so on. Keeping track of all polynomial coefficient is rather tedious for large polynomials.
5.5.2 F voltage source, H current source

General form:
FXXXX N + N - POLY ( ND ) V1 ( V2 V3 ...) P0 ( P1 ...)
Example:
FNONLIN 100 101 POLY (2) VDD Vxx 0 0.0 13.6 0.2 0.005
POLY(ND) Specifies the number of dimensions of the polynomial. The number of con-
trolling sources must be equal to the number of dimensions.
(N+) and (N-) nodes are output nodes. Positive current flows from the (+) node through
the source to the (-) node.
V1 (V2 V3 ...) are the controlling voltage sources. Control variable is the current through
these sources.
P0 (P1...) are the coefficients, as have been described in 5.5.1.
Chapter 6
Transmission Lines
Ngspice implements both the original SPICE3f5 transmission lines models and the one
introduced with KSPICE. The latter provide an improved transient analysis of lossy
transmission lines. Unlike SPICE models that use the state-based approach to simulate
lossy transmission lines, KSPICE simulates lossy transmission lines and coupled multi-
conductor line systems using the recursive convolution method. The impulse response
of an arbitrary transfer function can be determined by deriving a recursive convolution
from the Pade approximations of the function. We use this approach for simulating each
transmission line’s characteristics and each multiconductor line’s modal functions. This
method of lossy transmission line simulation has been proved to give a speedup of one to
two orders of magnitude over SPICE3f5.
6.1 Lossless Transmission Lines

General form:
TXXXXXXX N1 N2 N3 N4 Z0 = VALUE < TD = VALUE >

+ <F = FREQ < NL = NRMLEN > > < IC = V1 , I1 , V2 , I2 >
Examples:
T1 1 0 2 0 Z0 =50 TD =10 NS
n1 and n2 are the nodes at port 1; n3 and n4 are the nodes at port 2. z0 is the char-
acteristic impedance. The length of the line may be expressed in either of two forms.
The transmission delay, td, may be specified directly (as td=10ns, for example). Alterna-
tively, a frequency f may be given, together with nl, the normalized electrical length of
the transmission line with respect to the wavelength in the line at the frequency ‘f’. If a
frequency is specified but nl is omitted, 0.25 is assumed (that is, the frequency is assumed
to be the quarter-wave frequency). Note that although both forms for expressing the line
length are indicated as optional, one of the two must be specified.
Note that this element models only one propagating mode. If all four nodes are distinct
in the actual circuit, then two modes may be excited. To simulate such a situation,
121
122 CHAPTER 6. TRANSMISSION LINES
two transmission-line elements are required. (see the example in Chapt. 21.7 for further
clarification.) The (optional) initial condition specification consists of the voltage and
current at each of the transmission line ports. Note that the initial conditions (if any)
apply only if the UIC option is specified on the .TRAN control line.
Note that a lossy transmission line (see below) with zero loss may be more accurate than
the lossless transmission line due to implementation details.
6.2 Lossy Transmission Lines
General form:
OXXXXXXX n1 n2 n3 n4 mname
Examples:
O23 1 0 2 0 LOSSYMOD
OCONNECT 10 5 20 5 INTERCONNECT
This is a two-port convolution model for single conductor lossy transmission lines. n1
and n2 are the nodes at port 1; n3 and n4 are the nodes at port 2. Note that a lossy
transmission line with zero loss may be more accurate than the lossless transmission line
due to implementation details.
6.2.1 Lossy Transmission Line Model (LTRA)
The uniform RLC/RC/LC/RG transmission line model (referred to as the LTRA model
henceforth) models a uniform constant-parameter distributed transmission line. The RC
and LC cases may also be modeled using the URC and TRA models; however, the newer
LTRA model is usually faster and more accurate than the others. The operation of the
LTRA model is based on the convolution of the transmission line’s impulse responses with
its inputs (see [8]). The LTRA model takes a number of parameters, some of which must
be given and some of which are optional.
6.2. LOSSY TRANSMISSION LINES 123
Name Parameter Units/Type Default Example

R resistance/length Ω/unit 0.0 0.2
L inductance/length H/unit 0.0 9.13e-9
G conductance/length mhos/unit 0.0 0.0
C capacitance/length F/unit 0.0 3.65e-12
LEN length of line unit no default 1.0
REL breakpoint control arbitrary unit 1 0.5
ABS breakpoint control 1 5
NOSTEPLIMIT don’t limit time-step to less flag not set set
than line delay
NO CONTROL don’t do complex time-step flag not set set
control
LININTERP use linear interpolation flag not set set
MIXEDINTERP use linear when quadratic flag not set set
seems bad
COMPACTREL special reltol for history RELTOL 1.0e-3
compaction
COMPACTABS special abstol for history ABSTOL 1.0e-9
compaction
TRUNCNR use Newton-Raphson flag not set set
method for time-step
control
TRUNCDONTCUT don’t limit time-step to flag not set set
keep impulse-response
errors low
The following types of lines have been implemented so far:
• RLC (uniform transmission line with series loss only),
• RC (uniform RC line),
• LC (lossless transmission line),
• RG (distributed series resistance and parallel conductance only).
Any other combination will yield erroneous results and should not be tried. The length
LEN of the line must be specified. NOSTEPLIMIT is a flag that will remove the default
restriction of limiting time-steps to less than the line delay in the RLC case. NO CONTROL
is a flag that prevents the default limiting of the time-step based on convolution error
criteria in the RLC and RC cases. This speeds up simulation but may in some cases
reduce the accuracy of results. LININTERP is a flag that, when specified, will use linear
interpolation instead of the default quadratic interpolation for calculating delayed signals.
MIXEDINTERP is a flag that, when specified, uses a metric for judging whether quadratic
interpolation is not applicable and if so uses linear interpolation; otherwise it uses the de-
fault quadratic interpolation. TRUNCDONTCUT is a flag that removes the default cutting of
the time-step to limit errors in the actual calculation of impulse-response related quanti-
ties. COMPACTREL and COMPACTABS are quantities that control the compaction of the past
history of values stored for convolution. Larger values of these lower accuracy but usually
increase simulation speed. These are to be used with the TRYTOCOMPACT option, described
in the .OPTIONS section. TRUNCNR is a flag that turns on the use of Newton-Raphson it-
erations to determine an appropriate time-step in the time-step control routines. The
default is a trial and error procedure by cutting the previous time-step in half. REL and
ABS are quantities that control the setting of breakpoints.
The option most worth experimenting with for increasing the speed of simulation is REL.
The default value of 1 is usually safe from the point of view of accuracy but occasionally
increases computation time. A value greater than 2 eliminates all breakpoints and may
be worth trying depending on the nature of the rest of the circuit, keeping in mind that
it might not be safe from the viewpoint of accuracy.
Breakpoints may usually be entirely eliminated if it is expected the circuit will not display
sharp discontinuities. Values between 0 and 1 are usually not required but may be used
for setting many breakpoints.
COMPACTREL may also be experimented with when the option TRYTOCOMPACT is specified
in a .OPTIONS card. The legal range is between 0 and 1. Larger values usually decrease
the accuracy of the simulation but in some cases improve speed. If TRYTOCOMPACT is not
specified on a .OPTIONS card, history compaction is not attempted and accuracy is high.
NO CONTROL, TRUNCDONTCUT and NOSTEPLIMIT also tend to increase speed at the expense
of accuracy.
6.3 Uniform Distributed RC Lines

General form:
UXXXXXXX n1 n2 n3 mname l = len <n = lumps >
Examples:
U1 1 2 0 URCMOD L =50 U
URC2 1 12 2 UMODL l =1 MIL N =6
n1 and n2 are the two element nodes the RC line connects, while n3 is the node the
capacitances are connected to. mname is the model name, len is the length of the RC line
in meters. lumps, if specified, is the number of lumped segments to use in modeling the
RC line (see the model description for the action taken if this parameter is omitted).
6.3.1 Uniform Distributed RC Model (URC)

The URC model is derived from a model proposed by L. Gertzberg in 1974. The model
is accomplished by a subcircuit type expansion of the URC line into a network of lumped
RC segments with internally generated nodes. The RC segments are in a geometric
progression, increasing toward the middle of the URC line, with K as a proportionality
constant. The number of lumped segments used, if not specified for the URC line device,
is determined by the following formula:
6.4. KSPICE LOSSY TRANSMISSION LINES 125
2
(K−1)
log Fmax R
LL
C
2πL2 K
N= (6.1)
log K
The URC line is made up strictly of resistor and capacitor segments unless the ISPERL
parameter is given a nonzero value, in which case the capacitors are replaced with reverse
biased diodes with a zero-bias junction capacitance equivalent to the capacitance replaced,
and with a saturation current of ISPERL amps per meter of transmission line and an
optional series resistance equivalent to RSPERL ohms per meter.
Name Parameter Units Default Example Area

K Propagation Constant - 1.5 1.2 -
FMAX Maximum Frequency of interest Hz 1.0 G 6.5 Meg -
RPERL Resistance per unit length Ω/m 1000 10 -
CPERL Capacitance per unit length F/m 10e-15 1p -
ISPERL Saturation Current per unit length A/m 0 - -
RSPERL Diode Resistance per unit length Ω/m 0 - -
6.4 KSPICE Lossy Transmission Lines
Unlike SPICE3, which uses the state-based approach to simulate lossy transmission lines,
KSPICE simulates lossy transmission lines and coupled multiconductor line systems using
the recursive convolution method. The impulse response of an arbitrary transfer function
can be determined by deriving a recursive convolution from the Pade approximations
of the function. ngspice is using this approach for simulating each transmission line’s
characteristics and each multiconductor line’s modal functions. This method of lossy
transmission line simulation has shown to give a sigificant speedup. Please note that the
following two models will support only transient simulation, no ac.
Additional Documentation Available:
• S. Lin and E. S. Kuh, ‘Pade Approximation Applied to Transient Simulation of

Lossy Coupled Transmission Lines,’ Proc. IEEE Multi-Chip Module Conference,
1992, pp. 52-55.
• S. Lin, M. Marek-Sadowska, and E. S. Kuh, ‘SWEC: A StepWise Equivalent Con-

ductance Timing Simulator for CMOS VLSI Circuits,’ European Design Automation
Conf., February 1991, pp. 142-148.
• S. Lin and E. S. Kuh, ‘Transient Simulation of Lossy Interconnect,’ Proc. Design

Automation Conference, Anaheim, CA, June 1992, pp. 81-86.
6.4.1 Single Lossy Transmission Line (TXL)

General form:
YXXXXXXX N1 0 N2 0 mname < LEN = LENGTH >
Example:
Y1 1 0 2 0 ymod LEN =2
. MODEL ymod txl R =12.45 L =8.972 e -9 G =0 C =0.468 e -12 length =16
n1 and n2 are the nodes of the two ports. The optional instance parameter len is the
length of the line and may be expressed in multiples of [unit]. Typically unit is given in
meters. len will override the model parameter length for the specific instance only.
The TXL model takes a number of parameters:
LENGTH length of line unit no default 1.0
Model parameter length must be specified as a multiple of unit. Typically unit is given
in [m]. For transient simulation only.
6.4.2 Coupled Multiconductor Line (CPL)

The CPL multiconductor line model is in theory similar to the RLGC model, but without
frequency dependent loss (neither skin effect nor frequency-dependent dielectric loss). Up
to 8 coupled lines are supported in ngspice.
General form:
PXXXXXXX NI1 NI2 ... NIX GND1 NO1 NO2 ... NOX GND2 mname < LEN = LENGTH >
Example:
P1 in1 in2 0 b1 b2 0 PLINE

. model PLINE CPL length ={ Len }
+ R =1 0 1
+ L ={ L11 } { L12 } { L22 }
+ G =0 0 0
+ C ={ C11 } { C12 } { C22 }
. param Len =1 Rs =0
+ C11 =9.143579 E -11 C12 = -9.78265 E -12 C22 =9.143578 E -11
+ L11 =3.83572 E -7 L12 =8.26253 E -8 L22 =3.83572 E -7
6.4. KSPICE LOSSY TRANSMISSION LINES 127
ni1 ... nix are the nodes at port 1 with gnd1; no1 ... nox are the nodes at port 2
with gnd2. The optional instance parameter len is the length of the line and may be
expressed in multiples of [unit]. Typically unit is given in meters. len will override the
model parameter length for the specific instance only.
The CPL model takes a number of parameters:
LENGTH length of line unit no default 1.0
All RLGC parameters are given in Maxwell matrix form. For the R and G matrices the
diagonal elements must be specified, for L and C matrices the lower or upper triangular
elements must specified. The parameter LENGTH is a scalar and is mandatory. For
transient simulation only.
Chapter 7
Diodes
7.1 Junction Diodes

General form:
DXXXXXXX n + n - mname < area = val > <m = val > < pj = val > <off >
+ < ic = vd > < temp = val > < dtemp = val >
+ < lm = val > < wm = val > < lp = val > < wp = val >
Examples:
DBRIDGE 2 10 DIODE1
DCLMP aa cc DMOD AREA =3.0 IC =0.2
The pn junction (diode) implemented in ngspice expands the one found in SPICE3f5.
Perimeter effects and high injection level have been introduced into the original model
and temperature dependence of some parameters has been added. n+ and n- are the
positive (anode) and negative (cathode) nodes, respectively. mname is the model name.
Instance parameters may follow, dedicated to only the diode described on the respective
line. area is the area scale factor, which may scale the saturation current given by
the model parameters (and others, see table below). pj is the perimeter scale factor,
scaling the sidewall saturation current and its associated capacitance. m is a multiplier
of area and perimeter, and off indicates an (optional) starting condition on the device
for dc analysis. If the area factor is omitted, a value of 1.0 is assumed. The (optional)
initial condition specification using ic is intended for use with the uic option on the
.tran control line, when a transient analysis is desired starting from other than the
quiescent operating point. You should supply the initial voltage across the diode there.
The (optional) temp value is the temperature at which this device is to operate, and
overrides the temperature specification on the .option control line. The temperature
of each instance can be specified as an offset to the circuit temperature with the dtemp
option.
To fulfill requirements of modern process design kits (PDK) the basic spice3 model was
extended with the capability of modeling parasitic effects like sidewall junction currents
and capacitances, tunnel currents and metal and polysilicon overlap capacitances. Latter
129
130 CHAPTER 7. DIODES
effect can be activated by level=3 model parameter or by setting element parameters

lm, wm, lp and wp. If both are given, element parameters have priority.
With the (new in ngspice-39) OpenVAF/OSDI approach (see 13), all modern diode mod-
els, written in Verilog-A, become available, like JUNCAP etc..
7.2 Diode Model (D)
Diode models may be described in the netlist input file (or an file included by .inc)
according to the following example:
General form:
. model mname type ( pname1 = pval1 pname2 = pval2 ... )
Examples:
. model DIODE1 D ( bv =50 is =1 e -13 n =1.05)
with a user defined model name mname, and the model type D.
A basic model statement using only the internal default model parameters is
Basic model statement:
. model DMOD D
The dc characteristics of the diode are determined by the parameters is and n. An

ohmic resistance, rs, is included. Charge storage effects are modeled by a transit time,
tt, and a nonlinear depletion layer capacitance that is determined by the parameters
cjo, vj, and m. The temperature dependence of the saturation current is defined by
the parameters eg, the energy, and xti, the saturation current temperature exponent.
The nominal temperature where these parameters were measured is tnom, which defaults
to the circuit-wide value specified on the .options control line. Reverse breakdown is
modeled by an exponential increase in the reverse diode current and is determined by the
parameters bv and ibv (both of which are positive numbers).
7.2. DIODE MODEL (D) 131
Junction DC parameters
Name Parameter Units Default Example Scale factor

IS (JS) Saturation current A 1.0e-14 1.0e-16 area
JSW Sidewall saturation current A 0.0 1.0e-15 perimeter
N Emission coefficient - 1 1.5
RS Ohmic resistance Ω 0.0 100 1/area
BV Reverse breakdown voltage V ∞ 40

IBV Current at breakdown A 1.0e-3 1.0e-4
voltage
NBV Breakdown Emission - N 1.2
Coefficient
IKF (IK) Forward knee current A 0.0 1.0e-3
IKR Reverse knee current A 0.0 1.0e-3
JTUN Tunneling saturation A 0.0 area
current
JTUNSW Tunneling sidewall A 0.0 perimeter
saturation current
NTUN Tunneling emission - 30
coefficient
XTITUN Tunneling saturation - 3
current exponential
KEG EG correction factor for - 1.0
tunneling
ISR Recombination saturation A 1e-14 1pA area
current
NR Recombination current - 1 2
emission coefficient
Junction capacitance parameters

CJO (CJ0) Zero-bias junction F 0.0 2pF area
bottom-wall capacitance
CJP (CJSW) Zero-bias junction sidewall F 0.0 .1pF perimeter
capacitance
FC Coefficient for forward-bias - 0.5 -
depletion bottom-wall
capacitance formula
FCS Coefficient for forward-bias - 0.5 -
depletion sidewall
capacitance formula
M (MJ) Area junction grading - 0.5 0.5
coefficient
MJSW Periphery junction grading - 0.33 0.5
coefficient
VJ (PB) Junction potential V 1 0.6
PHP Periphery junction V 1 0.6
potential
TT Transit-time sec 0 0.1ns
Metal and Polysilicon Overlap Capacitances (level=3)

LM Length of metal capacitor m 0.0 4um SCALE
LP Length of polysilicon m 0.0 5um SCALE
capacitor
WM Width of metal capacitor m 0.0 2um SCALE
WP Width of polysilicon m 0.0 4um SCALE
capacitor
XOM Thickness of the metal to m 1e-06 -
bulk oxide
XOI Thickness of the polysilicon m 1e-06 -
to bulk oxide
XM Masking and etching effects m 0.0 -
in metal
XP Masking and etching effects m 0.0 -
in polysilicon
7.3. DIODE EQUATIONS 133
Temperature effects

1.11 Si
EG Activation energy eV 1.11 0.69 Sbd
0.67 Ge
◦
TNOM (TREF) Parameter measurement C 27 50
temperature
TRS1 (TRS) 1st order tempco for RS 1/◦ C 0.0 -
TRS2 2nd order tempco for RS 1/◦ C 2 0.0 -
TM1 1st order tempco for MJ 1/◦ C 0.0 -
TM2 2nd order tempco for MJ 1/◦ C 2 0.0 -
TTT1 1st order tempco for TT 1/◦ C 0.0 -
TTT2 2nd order tempco for TT 1/◦ C 2 0.0 -
3.0 pn
XTI Saturation current - 3.0
2.0 Sbd
temperature exponent
TLEV Diode temperature - 0
equation selector
TLEVC Diode capac. temperature - 0
equation selector
CTA (CTC) Area junct. cap. 1/◦ C 0.0 -
temperature coefficient
CTP Perimeter junct. cap. 1/◦ C 0.0 -
TCV Breakdown voltage 1/◦ C 0.0 -
Noise modeling

KF Flicker noise coefficient - 0
AF Flicker noise exponent - 1
7.3 Diode Equations

The junction diode is the basic semiconductor device and the simplest one in ngspice,
but its model is quite complex, even when not all the physical phenomena affecting a pn
junction are handled. The diode is modeled in three different regions:
• Forward bias: the anode is more positive than the cathode, the diode is ‘on’ and
can conduct large currents. To avoid convergence problems and unrealistic high
current, it is prudent to specify a series resistance to limit current with the rs
model parameter.
• Reverse bias: the cathode is more positive than the anode and the diode is ‘off’. A
reverse bias diode conducts a small leakage current.
• Breakdown: the breakdown region is modeled only if the bv model parameter is

given. When a diode enters breakdown the current increases exponentially (remem-
ber to limit it); bv is a positive value.
Parameters Scaling
Model parameters are scaled using the unit-less parameters area and pj and the multiplier
m as depicted below:
AREAef f = AREA m
P Jef f = PJ m
ISef f = IS AREAef f + JSW P Jef f
IBVef f = IBV AREAef f
IKef f = IK AREAef f
IKRef f = IKR AREAef f
CJef f = CJ0 AREAef f
CJPef f = CJP P Jef f
Diode DC, Transient and AC model equations
The diode model has certain dc currents for bottom and sidewall components. Exemplary
here is the equation for the bottom part:


qVD
ISef f (e N kT − 1) + VD · GM IN,
 if VD ≥ −3 N qkT
ID = −ISef f [1 + ( 3N kT 3
) ] + VD · GM IN, if − BVef f < VD < −3 N qkT (7.1)


qVD e
−IS (e −q(BVNefkTf +VD ) ) + V · GM IN, if V ≤ −BV
ef f D D ef f
Two secondary effects are modeled if the appropriate parameters (see table Junction DC
parameters) are given: Recombination current and bottom and sidewall tunnel current.
The breakdown region must be described with more depth since the breakdown is not
modeled physically. As written before, the breakdown modeling is based on two model
parameters: the ‘nominal breakdown voltage’ bv and the current at the onset of break-
down ibv. For the diode model to be consistent, the current value cannot be arbitrarily
chosen, since the reverse bias and breakdown regions must match. When the diode enters
breakdown region from reverse bias, the current is calculated using the formula1 :
−qBV
Ibdwn = −ISef f (e N kT − 1) (7.2)
The computed current is necessary to adjust the breakdown voltage making the two
regions match. The algorithm is a little bit convoluted and only a brief description is
given here:
1
if you look at the source code in file diotemp.c you will discover that the exponential relation is
replaced with a first order Taylor series expansion.
Algorithm 7.1 Diode breakdown current calculation

if IBVef f < Ibdwn then
IBVef f = Ibdwn
BVef f = BV
else
IBVef f
BVef f = BV − NVt ln( Ibdwn )
Most real diodes shows a current increase that, at high current levels, does not follow
the exponential relationship given above. This behavior is due to high level of carriers
injected into the junction. High injection effects (as they are called) are modeled with ik
and ikr.


 √ID , if VD ≥ −3 N qkT
 1+ IKD
I
ef f
IDef f = (7.3)

 √ ID , otherwise.
 1+ ID
IKRef f
Diode capacitance is divided into two different terms:
• Depletion capacitance
• Diffusion capacitance
Depletion capacitance is composed by two different contributes, one associated to the

bottom of the junction (bottom-wall depletion capacitance) and the other to the periphery
(sidewall depletion capacitance). The basic equations are
CDiode = Cdif f usion + Cdepletion
Where the depletion capacitance is defined as:
Cdepletion = Cdeplbw + Cdeplsw
The diffusion capacitance, due to the injected minority carriers, is modeled with the
transit time tt:
∂IDef f
Cdif f usion = TT
∂VD
The depletion capacitance is more complex to model, since the function used to ap-
proximate it diverges when the diode voltage become greater than the junction built-in
potential. To avoid function divergence, the capacitance function is approximated with
a linear extrapolation for applied voltage greater than a fraction of the junction built-in
potential.
( VD −MJ
CJef f (1 − VJ
) , if VD < FC · VJ
Cdeplbw = V
1−FC(1+MJ)+MJ D (7.4)
CJef f (1−FC)(1+MJ) VJ , otherwise.
( VD −MJSW
CJPef f (1 − PHP
) , if VD < FCS · PHP
Cdeplsw = VD
1−FCS(1+MJSW)+MJSW· PHP (7.5)
CJPef f (1−FCS)(1+MJSW)
, otherwise.
Temperature dependence
The temperature affects many of the parameters in the equations above, and the follow-
ing equations show how. One of the most significant parameters that varies with the
temperature for a semiconductor is the band-gap energy:
TNOM2
EGnom = 1.16 − 7.02e−4 (7.6)
TNOM + 1108.0
−4 T2
EG(T ) = 1.16 − 7.02e (7.7)
TNOM + 1108.0
The leakage current temperature’s dependence is:
logf actor
IS(T ) = IS e N (7.8)
logf actor
JSW (T ) = JSW e N (7.9)
where ‘logfactor’ is defined as
EG EG T
logf actor = − + XTI ln( ) (7.10)
Vt (TNOM) Vt (T ) TNOM
The contact potentials (bottom-wall an sidewall) temperature dependence is:

T T EGnom EG(T )
V J(T ) = VJ( ) − Vt (T ) 3 · ln( )+ − (7.11)
TNOM TNOM Vt (TNOM) Vt (T )

T T EGnom EG(T )
P HP (T ) = PHP( ) − Vt (T ) 3 · ln( )+ − (7.12)
TNOM TNOM Vt (TNOM) Vt (T )
The depletion capacitances temperature dependence is:

−4 V J(T )
CJ(T ) = CJ 1 + MJ(4.0e (T − TNOM) − + 1) (7.13)
VJ

−4 P HP (T )
CJSW (T ) = CJSW 1 + MJSW(4.0e (T − TNOM) − + 1) (7.14)
PHP
The transit time temperature dependence is:
T T (T ) = TT(1 + TTT1(T − TNOM) + TTT2(T − TNOM)2 ) (7.15)

The junction grading coefficient temperature dependence is:
M J(T ) = MJ(1 + TM1(T − TNOM) + TM2(T − TNOM)2 ) (7.16)
The series resistance temperature dependence is:
RS(T ) = RS(1 + TRS(T − TNOM) + TRS2(T − TNOM)2 ) (7.17)
Noise model
The diode has three noise contribution, one due to the presence of the parasitic resistance
rs and the other two (shot and flicker) due to the pn junction.
The thermal noise due to the parasitic resistance is:
4kT ∆f
i2RS = (7.18)
RS
The shot and flicker noise contributions are
KF · ID
AF
i2d = 2qID ∆f + ∆f (7.19)
f
Self Heating model
Ngspice diode model has implemented a simple self heating approach. A current equivalent
to the dissipated power is conducted to a RC parallel circuit. The connection node voltage
is so a thermal equivalent to the junction overtemperature. This temperature follows in
a electro-thermal feedback with appropriate change of the diode current and capacitance.
Compared to the standard diode we have a third node tj and a flag thermal on element
line. In the model description we have to set rthh0 and cth0 model parameter.
General form element usage:
DXXXXXXX n + n - tj mname <off > < ic = vd > thermal
Example model:
. model DPWR D ( bv =16 is =1 e -10 n =1.03 rth0 =50 cth0 =1 u )

Chapter 8
BJT
8.1 Bipolar Junction Transistors (BJTs)

General form:
QXXXXXXX nc nb ne <ns > <tj > mname < area = val > < areac = val >
+ < areab = val > <m = val > <off > < ic = vbe , vce > < temp = val >
+ < dtemp = val >
Examples:
Q23 10 24 13 QMOD IC =0.6 , 5.0

Q50A 11 26 4 20 MOD1
nc, nb, and ne are the collector, base, and emitter nodes, respectively. ns is the (op-
tional) substrate node. When unspecified, ground is used. tj is the (optional) junction
temperature node, available in the VBIC model (see 8.2.2). mname is the model name,
area, areab, areac are the area factors (emitter, base and collector respectively), and
off indicates an (optional) initial condition on the device for the dc analysis. If the area
factor is omitted, a value of 1.0 is assumed.
The (optional) initial condition specification using ic=vbe,vce is intended for use with the
uic option on a .tran control line, when a transient analysis is desired to start from other
than the quiescent operating point. See the .ic control line description for a better way to
set transient initial conditions. The (optional) temp value is the temperature where this
device is to operate, and overrides the temperature specification on the .option control
line. Using the dtemp option one can specify the instance’s temperature relative to the
circuit temperature.
8.2 BJT Models (NPN/PNP)

Ngspice provides three different BJT device models, which are selected by the .model
card.
139
140 CHAPTER 8. BJT
.model QMOD1 PNP

.model QMOD3 NPN level=4
These are the minimal versions, using default parameters supplied by ngspice. Further
optional parameters listed in the table below may replace the ngspice default parameters.
The level keyword specifies the model to be used:
• level=1: This is the original SPICE BJT model, and it is the default model if the
level keyword is not specified on the .model line. By activating certain parameter
a modified version of the original SPICE BJT that models both vertical and lateral
devices, includes temperature corrections of collector, emitter and base resistors and
allow modeling of quasi-saturation effects.
• level=4: Advanced VBIC model (see 8.2.2 and http://www.designers-guide.org/VBIC/
for details)
• level=8: HICUM/L2 model (see 8.2.4 and the official website for details)
With the (new in ngspice-39) OpenVAF/OSDI approach (see 13), all modern bipolar
models, written in Verilog-A, become available, like VBIC, Mextram and HICUM.
8.2.1 Gummel-Poon Models

The bipolar junction transistor model in ngspice is an adaptation of the integral charge
control model of Gummel and Poon. This modified Gummel-Poon model extends the
original model to include several effects at high bias levels. The model automatically
simplifies to the simpler Ebers-Moll model when certain parameters are not specified.
The parameter names used in the modified Gummel-Poon model have been chosen to be
more easily understood by the user, and to reflect better both physical and circuit design
thinking.
The dc model is defined by the parameters IS, BF, NF, ISE, IKF, and NE, which determine
the forward current gain characteristics, IS, BR, NR, ISC, IKR, and NC, which determine
the reverse current gain characteristics, and VAF and VAR, which determine the output
conductance for forward and reverse regions.
A more accurate model for transport current components is possible by specification of
model parameter IBE and IBC instead of IS.
Parameter NKF(NK)was introduced for more accurate high current beta rolloff modelling.
The BJT model has among the standard temperature parameters an extension compatible
with most foundry provided process design kits (see parameter table below TLEV).
The BJT model includes the substrate saturation current ISS. Three ohmic resistances
RB, RC, and RE are included, where RB can be high current dependent. Base charge storage
is modeled by forward and reverse transit times, TF and TR, where the forward transit time
TF can be bias dependent if desired. Nonlinear depletion layer capacitances are defined
with CJE, VJE, and NJE for the B-E junction, CJC, VJC, and NJC for the B-C junction and
CJS, VJS, and MJS for the C-S (collector-substrate) junction.
The BJT model support a substrate capacitance that is connected to the device’s base
or collector, to model lateral or vertical devices dependent on the parameter SUBS. The
8.2. BJT MODELS (NPN/PNP) 141
temperature dependence of the saturation currents, IS and ISS, is determined by the

energy-gap, EG, and the saturation current temperature exponent, XTI.
In the new model, additional base current temperature dependence is modeled by the beta
temperature exponent XTB. The values specified are assumed to have been measured at
the temperature TNOM, which can be specified on the .options control line or overridden
by a specification on the .model line.
The BJT parameters used in the modified Gummel-Poon model are listed below. The
parameter names used in earlier versions of SPICE2 are still accepted.
Gummel-Poon BJT Parameters (incl. model extensions)
Name Parameters Units Default Example Scale factor

SUBS Substrate connection: 1 for 1
vertical geometry, -1 for lateral
geometry
IS Transport saturation current A 1.0e-16 1.0e-15 area
IBE Base-Emitter saturation current A 0.0 1.0e-16 area
IBC Base-Collector saturation current A 0.0 1.0e-16 areab,areac
ISS Reverse saturation current, A 0.0 1.0e-15 area
substrate-to-collector for vertical
device or substrate-to-base for
lateral
BF Ideal maximum forward beta. - 100 100
NF Forward current emission - 1.0 1
coefficient
VAF (VA) Forward Early voltage V ∞ 200
IKF Corner for forward beta current A ∞ 0.01 area
roll-off
NKF(NK) High current Beta rolloff - 0.5 0.9
exponent
ISE B-E leakage saturation current. A 0.0 1e-13 area
NE B-E leakage emission coefficient - 1.5 2
BR Ideal maximum reverse beta - 1 0.1
NR Reverse current emission - 1 1
coefficient
VAR (VB) Reverse Early voltage V ∞ 200
IKR Corner for reverse beta high A ∞ 0.01 area
current roll-off
ISC B-C leakage saturation current A 0.0 1e-13 areab,areac
(scale is ‘areab’ for vertical
devices and ‘areac’ for lateral)
NC B-C leakage emission coefficient - 2 1.5
RB Zero bias base resistance Ω 0 100 1/area
IRB Current where base resistance A ∞ 0.1 area
falls halfway to its min value
142 CHAPTER 8. BJT
RBM Minimum base resistance at high Ω RB 10 1/area

currents
RE Emitter resistance Ω 0 1 1/area
RC Collector resistance Ω 0 10 1/area
CJE B-E zero-bias depletion F 0 2pF area
capacitance
VJE (PE) B-E built-in potential V 0.75 0.6
MJE (ME) B-E junction exponential factor - 0.33 0.33
TF Ideal forward transit time sec 0 0.1ns
XTF Coefficient for bias dependence - 0
of TF
VTF Voltage describing VBC V ∞
dependence of TF
ITF High-current parameter for effect A 0 - area
on TF
1
PTF Excess phase at freq= Hz deg 0
2πT F
CJC B-C zero-bias depletion F 0 2pF areab,areac
capacitance (scale is ‘areab’ for
vertical devices and ‘areac’ for
lateral)
VJC (PC) B-C built-in potential V 0.75 0.5
MJC B-C junction exponential factor - 0.33 0.5
XCJC Fraction of B-C depletion - 1
capacitance connected to
internal base node
TR Ideal reverse transit time sec 0 10ns
CJS Zero-bias collector-substrate F 0 2pF areab,areac
capacitance (scale is ‘areac’ for
vertical devices and ‘areab’ for
lateral)
VJS (PS) Substrate junction built-in V 0.75
potential
MJS (MS) Substrate junction exponential - 0 0.5
factor
XTB Forward and reverse beta - 0
EG Energy gap for temperature eV 1.11
effect on IS
XTI Temperature exponent for effect - 3
on IS
KF Flicker-noise coefficient - 0
AF Flicker-noise exponent - 1
FC Coefficient for forward-bias - 0.5 0
depletion capacitance formula
◦
TNOM (TREF) Parameter measurement C 27 50
temperature
TLEV BJT temperature equation - 0

selector
TLEVC BJT capac. temperature - 0
equation selector
TRE1 1st order temperature coefficient 1/◦ C 0.0 1e-3
for RE
TRE2 2nd order temperature coefficient 1/◦ C 2 0.0 1e-5
for RE
TRC1 1st order temperature coefficient 1/◦ C 0.0 1e-3
for RC
TRC2 2nd order temperature coefficient 1/◦ C 2 0.0 1e-5
for RC
TRB1 1st order temperature coefficient 1/◦ C 0.0 1e-3
for RB
TRB2 2nd order temperature coefficient 1/◦ C 2 0.0 1e-5
for RB
TRBM1 1st order temperature coefficient 1/◦ C 0.0 1e-3
for RBM
TRBM2 2nd order temperature coefficient 1/◦ C 2 0.0 1e-5
for RBM
TBF1 1st order temperature coefficient 1/◦ C 0.0 1e-3
for BF
TBF2 2nd order temperature coefficient 1/◦ C 2 0.0 1e-5
for BF
TBR1 1st order temperature coefficient 1/◦ C 0.0 1e-3
for BR
TBR2 2nd order temperature coefficient 1/◦ C 2 0.0 1e-5
for BR
TIKF1 1st order temperature coefficient 1/◦ C 0.0 1e-3
for IKF
TIKF2 2nd order temperature coefficient 1/◦ C 2 0.0 1e-5
for IKF
TIKR1 1st order temperature coefficient 1/◦ C 0.0 1e-3
for IKR
TIKR2 2nd order temperature coefficient 1/◦ C 2 0.0 1e-5
for IKR
TIRB1 1st order temperature coefficient 1/◦ C 0.0 1e-3
for IRB
TIRB2 2nd order temperature coefficient 1/◦ C 2 0.0 1e-5
for IRB
TNC1 1st order temperature coefficient 1/◦ C 0.0 1e-3
for NC
TNC2 2nd order temperature coefficient 1/◦ C 2 0.0 1e-5
for NC
TNE1 1st order temperature coefficient 1/◦ C 0.0 1e-3
for NE
144 CHAPTER 8. BJT
TNE2 2nd order temperature coefficient 1/◦ C 2 0.0 1e-5

for NE
TNF1 1st order temperature coefficient 1/◦ C 0.0 1e-3
for NF
TNF2 2nd order temperature coefficient 1/◦ C 2 0.0 1e-5
for NF
TNR1 1st order temperature coefficient 1/◦ C 0.0 1e-3
for IKF
TNR2 2nd order temperature coefficient 1/◦ C 2 0.0 1e-5
for IKF
TVAF1 1st order temperature coefficient 1/◦ C 0.0 1e-3
for VAF
TVAF2 2nd order temperature coefficient 1/◦ C 2 0.0 1e-5
for VAF
TVAR1 1st order temperature coefficient 1/◦ C 0.0 1e-3
for VAR
TVAR2 2nd order temperature coefficient 1/◦ C 2 0.0 1e-5
for VAR
CTC 1st order temperature coefficient 1/◦ C 0.0 1e-3
for CJC
CTE 1st order temperature coefficient 1/◦ C 0.0 1e-3
for CJE
CTS 1st order temperature coefficient 1/◦ C 0.0 1e-3
for CJS
TVJC 1st order temperature coefficient 1/◦ C 2 0.0 1e-5
for VJC
TVJE 1st order temperature coefficient 1/◦ C 0.0 1e-3
for VJE
TITF1 1st order temperature coefficient 1/◦ C 0.0 1e-3
for ITF
TITF2 2nd order temperature coefficient 1/◦ C 2 0.0 1e-5
for ITF
TTF1 1st order temperature coefficient 1/◦ C 0.0 1e-3
for TF
TTF2 2nd order temperature coefficient 1/◦ C 2 0.0 1e-5
for TF
TTR1 1st order temperature coefficient 1/◦ C 0.0 1e-3
for TR
TTR2 2nd order temperature coefficient 1/◦ C 2 0.0 1e-5
for TR
TMJE1 1st order temperature coefficient 1/◦ C 0.0 1e-3
for MJE
TMJE2 2nd order temperature coefficient 1/◦ C 2 0.0 1e-5
for MJE
TMJC1 1st order temperature coefficient 1/◦ C 0.0 1e-3
for MJC
TMJC2 2nd order temperature coefficient 1/◦ C 2 0.0 1e-5

for MJC
Quasi-Saturation Model extension
By defining parameter RCO, VO, GAMMA and QCO an extension of the Gummel-
Poon model will be switched on to model bipolar junction transistors that exhibit quasi-
saturation effects. A description can be found in [24].
Name Parameters Units Default Example Scale factor

RCO Epitaxial region resistance Ω 0 0.45 1/area
VO Carrier mobility knee V 10 4.16
voltage
GAMMA Epitaxial region doping − 1e-11 1.0e-15
factor
QCO Epitaxial region charge C 0.0 3.4E-11
factor
VG Energy gap QS temp. V 1.206 1.2
depend.
CN Temperature exponent of 2.42 NPN 2.2 PNP
RCI
D Temperature exponent of .87 NPN .52 PNP
VO
The Collector current output characteristic shows a special moderate transition in the
BJT saturation region, see figure 8.1. Furthermore DC current gain and the unity gain
frequency fT falls sharply.
8.2.2 VBIC Model

The VBIC model is an extended development of the Standard Gummel-Poon (SGP)
model with the focus of integrated bipolar transistors in today’s modern semiconductor
technologies. With the implemented modified Quasi-Saturation model from Kull and
Nagel it is also possible to model the special output characteristic of discrete switching
and RF transistors. It is a improved alternative to the SGP model for silicon, SiGe and
III-V HBT devices.
VBIC Capabilities compared to Standard Gummel-Poon Model:
• Integrated substrate transistor for parasitic devices in integrated processes
• Weak avalanche and base-emitter breakdown model
• Improved Early effect modeling
• Physical separation of Ic and Ib
• Improved depletion capacitance model

146 CHAPTER 8. BJT
Figure 8.1: Output characteristic with and w/o Quasi-Saturation
• Improved temperature modeling
• Self-heating modeling
VBIC self-heating model
This model has implemented a simple 1-pole thermal network to cover self-heating effects.
That means that the power dissipation is gathered in all branches of the device model
and is conducted as an equivalent current Ith into one model node dt. This node has a
resistor Rth and capacitor Cth parallel connection to ground. Because the resistor plays
the role of the thermal resistance from junction to case the arising voltage at node dt is
equivalent the BJT junctions temperature. The model realisizes that this temperature
rise leads to deviations for internal resistors, currents and capacitors values, calculated by
temperature update equations. This process is included into the ngspice iteration schema
for all analyses and is controlled by the model parameter selft (selft=0: self heating
calculation is off (default value), selft=1: self-heating is on). In addition the model
parameter rth has to be given.
The simple thermal network of the VBIC model is shown in Fig. 8.2.
How to instantiate the bipolar VBIC model (only minimal version) with self-heating:
vc c 0 0
vb b 0 1
ve e 0 0
vs s 0 0
Q1 c b e s dt mod1 area =1
. model mod1 npn Level =4 selft =1 rth =100
Figure 8.2: VBIC thermal network
Of course it is possible to connect an more accurate thermal network to the node dt. The
following example is showing a simplified thermal network covering the thermal resistances
and capacitances of junction-case and case-ambient transitions induced by a heat-sink.
Q1 c b e s dt mod2
. model mod2 npn Level =9 selft =1 rth =20
X1 dt tamb junction - ambient
VTamb tamb 0 30
. subckt junction - ambient jct amb
rjc jct 1 0.4
ccs 1 0 5 m
rcs 1 2 0.1
csa 2 0 30 m
rsa 2 amb 1.3
. ends
8.2.3 MEXTRAM Model
MEXTRAM (Most EXquisite TRAnsistor Model)) is an advanced compact model for

bipolar transistors that contains many features that the widely-used Gummel-Poon model
lacks. The model was initiated by Philips and later co-worked by NXP Semiconductors
and different Universities.
Mextram has proven excellent for Si and SiGe processes, including analog, mixed-signal,
high speed RF as well as high voltage high power technologies. It accounts for high injec-
tion effects with a dedicated epi-layer model, self heating, avalanche, low-frequency and
high frequency noises in physical manners, and is formulated with minimal interactions
between DC and AC characteristics that simplifies parameter extraction.
For more information see MEXTRAM and MEXTRAM Definition.
Ngspice has implemented version 504.12.1 in his experimental ADMS tree. It will be
activated by the BJT model parameter level=6.
148 CHAPTER 8. BJT
Figure 8.3: The equivalent circuit of HICUM/L2 without the self-heating, NQS and noise
correlation networks.
8.2.4 HICUM level 2 Model

The physics-based HIgh-CUrrent Model (HICUM) Level2 (L2) has been a standard com-
pact model for bipolar junction transistors and heterojunction bipolar transistors (HBTs)
for many years. The model has been shown to be applicable to many process genera-
tions of SiGe HBTs and also to InP HBTs, including high-speed and high-voltage device
designs. The implemented version in Ngspice is HICUML2/2.4 and can be activated by
BJT model parameter level=8.
HICUML2 captures most to all known physical effects relevant in HBTs, in example:
• substrate transistor
• avalanche effect
• physics based transfer current model
• self-heating
• accurate modeling of the temperature dependence
• excess phase between base and collector current
Note that the noise correlation network is not implemented in Ngspice. More information
regarding the model and its parameters can be found on the website.
The equivalent circuit of the model is shown in fig. 8.3. The model is employed in many
PDKs for state-of-the-art SiGe and InP HBTs and is actively developed at TU Dresden.
The HICUM model exposes the following nodes to the user:
C(ollector) B(ase) E(mitter) S(ubstrate) T(emperature)
By connecting the T and S nodes of the model to other circuit elements, the thermal
and substrate network can be modified by the user. Note that both self-heating and the
avalanche effect may cause convergency issues if the operating region is too extreme.
The HICUM/L2 model can be initiated like this example:
vc c 0 0
vb b 0 1
ve e 0 0
vs s 0 0
Q1 c b e s dt mod1 area =1
. model mod1 npn Level =8
Self-heating is activated by model parameters FLSH, RTH and connecting T node of the
device instance. FLSH = 1 will only consider main thermal contributions of IC and IB,
FLSH = 2 include all power dissipations of the transistor.
8.2.5 HICUM level 0 Model

The HIgh-CUrrent Model (HICUM) Level0 (L0) is a simplified version of the HICUM
level 2 model. Ngspice has implemented version 1.32 in its experimental ADMS tree. It
will be activated by the BJT model parameter level=7.
150 CHAPTER 8. BJT
Chapter 9
JFETs
9.1 Junction Field-Effect Transistors (JFETs)

General form:
JXXXXXXX nd ng ns mname < area > <off > < ic = vds , vgs > < temp =t >
Examples:
J1 7 2 3 JM1 OFF
nd, ng, and ns are the drain, gate, and source nodes, respectively. mname is the model
name, area is the area factor, and off indicates an (optional) initial condition on the
device for dc analysis. If the area factor is omitted, a value of 1.0 is assumed. The
(optional) initial condition specification, using ic=VDS,VGS is intended for use with the
uic option on the .TRAN control line, when a transient analysis is desired starting from
other than the quiescent operating point. See the .ic control line for a better way to set
initial conditions. The (optional) temp value is the temperature where this device is to
operate, and overrides the temperature specification on the .option control line.
9.2 JFET Models (NJF/PJF)
9.2.1 Basic model statement

.model JM1 NJF level=1
.model JMOD2 PJF level=2
9.2.2 JFET level 1 model with Parker Skellern modification

The level 1 JFET model is derived from the FET model of Shichman and Hodges. The
dc characteristics are defined by the parameters VTO and BETA, which determine the
151
152 CHAPTER 9. JFETS
variation of drain current with gate voltage, LAMBDA, which determines the output con-
ductance, and IS, the saturation current of the two gate junctions. Two ohmic resistances,
RD and RS, are included.
vgst = vgs − V T O (9.1)
βp = BET A (1 + LAM BDA vds) (9.2)
1−B
bf ac = (9.3)
PB − V TO


vds · GM IN, if vgst ≤ 0
IDrain = βp vds (vds (bf ac vds − B) vgst (2B + 3bf ac (vgst − vds))) + vds · GM IN, if vgst ≥ vds


βp vgst2 (B + vgst bf ac) + vds · GM IN, if vgst < vds
(9.4)
Note that in Spice3f and later, the fitting parameter B has been added by Parker and
Skellern. For details, see [9]. If parameter B is set to 1 equation above simplifies to


vds · GM IN, if vgst ≤ 0
IDrain = βp vds (2vgst − vds) + vds · GM IN, if vgst ≥ vds (9.5)


βp vgst2 + vds · GM IN, if vgst < vds
Charge storage is modeled by nonlinear depletion layer capacitances for both gate junc-
tions, which vary as the −1/2 power of junction voltage and are defined by the parameters
CGS, CGD, and PB.
9.2. JFET MODELS (NJF/PJF) 153
Name Parameter Units Default Example Scaling factor

VTO Threshold voltage VT 0 V -2.0 -2.0
BETA Transconductance parameter (β) A/V ” 1.0e-4 1.0e-3 area
LAMBDA Channel-length modulation 1/V 0 1.0e-4
parameter (λ)
RD Drain ohmic resistance Ω 0 100 1/area
RS Source ohmic resistance Ω 0 100 1/area
CGS Zero-bias G-S junction F 0 5pF area
capacitance Cgs
CGD Zero-bias G-D junction F 0 1pF area
capacitance Cgd
PB Gate junction potential V 1 0.6
IS Gate saturation current IS A 1.0e-14 1.0e-14 area
B Doping tail parameter - 1 1.1
NLEV Noise equation selector - 1 3
GDSNOI Channel noise coefficient for 1.0 2.0
nlev=3
FC Coefficient for forward-bias 0.5
depletion capacitance formula
◦
TNOM Parameter measurement C 27 50
temperature
TCV Threshold voltage temperature 1/°C 0.0 0.01
coefficient
VTOTC Threshold voltage temperature 1/°C 0.0 -2.5m
coefficient (alternative model)
BEX Mobility temperature exponent - 0.0 1.1
BETATCE Mobility temperature exponent %/°C 0.0 -0.5
(alternative model)
XTI Gate saturation current - 3.0
EG Bandgap voltage 1.11
Additional to the standard thermal and flicker noise model an alternative thermal channel
noise model is implemented and is selectable by setting NLEV parameter to 3. This leads
to a correct channel thermal noise description in the linear region.
2 (1 + α + α2 )
Snoise = 4kT · BET A · V gst GDSN OI (9.6)
3 1+α
with
(
1− vds
vgs−V T O
, if vgs − V T O ≥ vds
α= (9.7)
0, else
JFET level 1 model has an alternative temperature model for main parameter VTO and
BETA:
• VTOTC is given:
V T O(T emp) = V T O + V T OT C ∗ (T emp − T N OM ) (9.8)
• VTOTC not given:
V T O(T emp) = V T O − T CV ∗ (T emp − T N OM ) (9.9)
• BETATCE is given:
BET A(T emp) = BET A ∗ 1.01BET AT CE∗(T emp−T N OM ) (9.10)
• BETATCE not given:

BEX
T emp
BET A(T emp) = BET A ∗ (9.11)
T N OM
9.2.3 JFET level 2 Parker Skellern model
The level 2 model is an improvement to level 1. Details are available in a pdf originating
from Macquarie University. Some important items are
• The description maintains strict continuity in its high-order derivatives, which is

essential for prediction of distortion and intermodulation.
• Frequency dependence of output conductance and transconductance is described as

a function of bias.
• Both drain-gate and source-gate potentials modulate the pinch-off potential, which
is consistent with S-parameter and pulsed-bias measurements.
• Self-heating varies with frequency.
• Extreme operating regions - subthreshold, forward gate bias, controlled resistance,

and breakdown regions - are included.
• Parameters provide independent fitting to all operating regions. It is not necessary

to compromise one region in favor of another.
• Strict drain-source symmetry is maintained. The transition during drain-source

potential reversal is smooth and continuous.
The model equations are described in this pdf document and in [19].
9.2. JFET MODELS (NJF/PJF) 155
Name Description Units Default

ID Device IDText Text PF1
ACGAM Capacitance modulation - 0
BETA Linear-region transconductance scale - 10−4
CGD Zero-bias gate-source capacitance F 0
CGS Zero-bias gate-drain capacitance F 0
DELTA Thermal reduction coefficient 1/W 0
FC Forward bias capacitance parameter - 0.5
HFETA High-frequency VGS feedback parameter - 0
HFE1 HFGAM modulation by VGD 1/V 0
HFE2 HFGAM modulation by VGS 1/V 0
HFGAM High-frequency VGD feedback parameter - 0
HFG1 HFGAM modulation by VSG 1/V 0
HFG2 HFGAM modulation by VDG 1/V 0
IBD Gate-junction breakdown current A 0
IS Gate-junction saturation current A 10−14
LFGAM Low-frequency feedback parameter - 0
LFG1 LFGAM modulation by VSG 1/V 0
LFG2 LFGAM modulation by VDG 1/V 0
MVST Subthreshold modulation 1/V 0
N Gate-junction ideality factor - 1
P Linear-region power-law exponent - 2
Q Saturated-region power-law exponent - 2
RS Source ohmic resistance Ω 0
RD Drain ohmic resistance Ω 0
TAUD Relaxation time for thermal reduction s 0
TAUG Relaxation time for gamma feedback s 0
VBD Gate-junction breakdown potential V 1
VBI Gate-junction potential V 1
VST Subthreshold potential V 0
VTO Threshold voltage V -2.0
XC Capacitance pinch-off reduction factor - 0
XI Saturation-knee potential factor - 1000
Z Knee transition parameter - 0.5
RG Gate ohmic resistance Ω 0
LG Gate inductance H 0
LS Source inductance H 0
LD Drain inductance H 0
CDSS Fixed Drain-source capacitance F 0
AFAC Gate-width scale factor - 1
NFING Number of gate fingers scale factor - 1
TNOM Nominal Temperature (Not implemented) K 300 K
TEMP Temperature K 300 K
Chapter 10
MESFETs
10.1 MESFETs
General form:
ZXXXXXXX ND NG NS MNAME < AREA > <OFF > < IC = VDS , VGS >
Examples:
Z1 7 2 3 ZM1 OFF
10.2 MESFET Models (NMF/PMF)
10.2.1 Basic model statements
.model ZM1 NMF level=1

.model MZMOD PMF level=4
These model statements will use the default parameters ( level 1 listed below).
10.2.2 Model by Statz e.a.
The MESFET model level 1 is derived from the GaAs FET model of Statz et al. as
described in [11]. The dc characteristics are defined by the parameters VTO, B, and
BETA, which determine the variation of drain current with gate voltage, ALPHA, which
determines saturation voltage, and LAMBDA, which determines the output conductance.
The formula are given by:
 h 3 i
 β(Vgs −VT O )2
1 − 1 − α V3ds (1 + λVds ) for 0 < Vds < 3
1+B(Vgs −VT O ) α
Id = (10.1)
 β(Vgs −VT O )2 (1 + λVds ) for Vds ≥ 3
1+B(Vgs −VT O ) α
157
158 CHAPTER 10. MESFETS
Two ohmic resistances, rd and rs, are included. Charge storage is modeled by total
gate charge as a function of gate-drain and gate-source voltages and is defined by the
parameters cgs, cgd, and pb.
Name Parameter Units Default Example Area
VTO Pinch-off voltage V -2.0 -2.0
BETA Transconductance parameter A/V 2 1.0e-4 1.0e-3 *
B Doping tail extending parameter 1/V 0.3 0.3 *
ALPHA Saturation voltage parameter 1/V 2 2 *
LAMBDA Channel-length modulation parameter 1/V 0 1.0e-4
RD Drain ohmic resistance Ω 0 100 *
RS Source ohmic resistance Ω 0 100 *
CGS Zero-bias G-S junction capacitance F 0 5pF *
CGD Zero-bias G-D junction capacitance F 0 1pF *
PB Gate junction potential V 1 0.6
FC Coefficient for forward-bias depletion - 0.5
capacitance formula
Device instance:
z1 2 3 0 mesmod area =1.4
Model:
. model mesmod nmf level =1 rd =46 rs =46 vt0 = -1.3

+ lambda =0.03 alpha =3 beta =1.4 e -3
10.2.3 Model by Ytterdal e.a.
level 2 (and levels 3,4) Copyright 1993: T. Ytterdal, K. Lee, M. Shur and T. A. Fjeldly
to be written
M. Shur, T.A. Fjeldly, T. Ytterdal, K. Lee, "Unified GaAs MESFET Model for Circuit
Simulation", Int. Journal of High Speed Electronics, vol. 3, no. 2, pp. 201-233, 1992
10.2.4 hfet1
level 5
Heterostructure Field Effect Transistor model as described in section 4.6 of the book
K. Lee, M. Shur, T. A. Fjeldly and T. Ytterdal, Semiconductor Device Modeling for VLSI,
1993, Prentice Hall, New Jersey.
Model parameters, equivalent circuit diagrams and device equations are also described in
the AIM-Spice reference manual, section Device Models A.
10.2. MESFET MODELS (NMF/PMF) 159
10.2.5 hfet2
level6
The HFET level 2 model is a simplified version of the level 1 model. The model is
optimized for speed and is suitable for simulation of digital circuits. To increase the
speed, some of the features included in the level 1 model is not implemented for the level
2 model.
160 CHAPTER 10. MESFETS
Chapter 11
MOSFETs
Ngspice supports all the original MOSFET models present in SPICE3f5 and almost all
the newer ones that have been published and made open-source. Both bulk and SOI
(Silicon on Insulator) models are available. When compiled with the cider option, ngspice
implements the four terminals numerical model that can be used to simulate a MOS-
FET (please refer to numerical modeling documentation for additional information and
examples).
11.1 MOSFET devices

General form:
MXXXXXXX nd ng ns nb mname <m = val > <l = val > <w = val >
+ < ad = val > < as = val > < pd = val > < ps = val > < nrd = val >
+ < nrs = val > <off > < ic = vds , vgs , vbs > < temp =t >
Examples:
M1 24 2 0 20 TYPE1
M31 2 17 6 10 MOSN L =5 U W =2 U
M1 2 9 3 0 MOSP L =10 U W =5 U AD =100 P AS =100 P PD =40 U PS =40 U
Note the suffixes in the example: the suffix ‘u’ specifies microns (1e-6 m) and ‘p’ sq-
microns (1e-12 m2 ).
The instance card for MOS devices starts with the letter ’M’. nd, ng, ns, and nb are the
drain, gate, source, and bulk (substrate) nodes, respectively. mname is the model name and
m is the multiplicity parameter, which simulates ‘m’ paralleled devices. All MOS models
support the ‘m’ multiplier parameter. Instance parameters l and w, channel length and
width respectively, are expressed in meters. The drain and source diffusion areas are ad
and as, in square meters (m2 ).
If any of l, w, ad, or as are not specified, default values are used. The use of defaults
simplifies input file preparation, as well as the editing required if device geometries are to
be changed. pd and ps are the perimeters of the drain and source junctions, in meters. nrd
161
162 CHAPTER 11. MOSFETS
and nrs designate the equivalent number of squares of the drain and source diffusions;
these values multiply the sheet resistance rsh specified on the .model control line for
an accurate representation of the parasitic series drain and source resistance of each
transistor. pd and ps default to 0.0 while nrd and nrs to 1.0. off indicates an (optional)
initial condition on the device for dc analysis. The (optional) initial condition specification
using ic=vds,vgs,vbs is intended for use with the uic option on the .tran control line,
when a transient analysis is desired starting from other than the quiescent operating
point. See the .ic control line for a better and more convenient way to specify transient
initial conditions. The (optional) temp value is the temperature at which this device is to
operate, and overrides the temperature specification on the .option control line.
The temperature specification is ONLY valid for level 1, 2, 3, and 6 MOSFETs, not for
level 4 or 5 (BSIM) devices.
BSIM3 (v3.2 and v3.3.0), BSIM4 (v4.7 and v4.8) and BSIMSOI models are also supporting
the instance parameter delvto and mulu0 for local mismatch and NBTI (negative bias
temperature instability) modeling:
delvto (delvt0) Threshold voltage shift V 0.0 0.07
mulu0 Low-field mobility multiplier (U0) - 1.0 0.9
11.2 MOSFET models (NMOS/PMOS)

MOSFET models are the central part of ngspice, probably because they are the most
widely used devices in the electronics world. Ngspice provides all the MOSFETs imple-
mented in the original Spice3f and adds several models developed by UC Berkeley’s Device
Group and other independent groups.
Each model is invoked with a .model card. A minimal version is:
.model MOSN NMOS level=8 version=3.3.0
The model name MOSN corresponds to the model name in the instance card (see 11.1).
Parameter NMOS selects an n-channel device, PMOS would point to a p-channel tran-
sistor. The level and version parameters select the specific model. Further model
parameters are optional and replace ngspice default values. Due to the large number
of parameters (more than 100 for modern models), model cards may be stored in extra
files and loaded into the netlist by the .include (2.7) command. Model cards are specific
for a an IC manufacturing process and are typically provided by the IC foundry. Some
generic parameter sets, not linked to a specific process, are made available by the model
developers, e.g. UC Berkeley’s Device Group for BSIM4 and BSIMSOI.
Ngspice provides several MOSFET device models, which differ in the formulation of the
I-V characteristic, and are of varying complexity. Models available are listed in table 11.1.
Current models for IC design are BSIM3 (11.2.10, down to channel length of 0.25 µm),
BSIM4 (11.2.11, below 0.25 µm), BSIMSOI (11.2.12, silicon-on-insulator devices), HiSIM2
and HiSIM_HV (11.2.14, surface potential models for standard and high voltage/high
power MOS devices).
With the (new in ngspice-39) OpenVAF/OSDI approach (see 13), all modern MOS models,
written in Verilog-A, become available, like BSIMBULK, BSIM-CMG and BSIM-IMG,
PSP, HiSim etc..
Level Name Model Version Developer References Notes
1 MOS1 Shichman-Hodges - Berkeley This is the classical quadratic model.
2 MOS2 Grove-Frohman - Berkeley Described in [2]
3 MOS3 Berkeley A semi-empirical model (see [1])
4 BSIM1 Berkeley Described in [3]
5 BSIM2 Berkeley Described in [5]
6 MOS6 Berkeley Described in [2]
9 MOS9 Alan Gillespie
8, 49 BSIM3v0 3.0 Berkeley extensions by Alan Gillespie
8, 49 BSIM3v1 3.1 Berkeley extensions by Serban Popescu
8, 49 BSIM3v32 3.2 - 3.2.4 Berkeley Multi version code
8, 49 BSIM3 3.3.0 Berkeley Described in [13]
10, 58 B4SOI 4.3.1 Berkeley
14, 54 BSIM4v5 4.0 - 4.5 Berkeley Multi version code
14, 54 BSIM4v6 4.6.2 Berkeley
11.2. MOSFET MODELS (NMOS/PMOS)
14, 54 BSIM4v7 4.7.0 Berkeley

14, 54 BSIM4 4.8.2 Berkeley
55 B3SOIFD Berkeley
56 B3SOIDD Berkeley
57 B3SOIPD Berkeley
60 STAG SOI3 Southampton
Table 11.1: MOSFET model summary

68 HiSIM2 2.8.0 Hiroshima
73 HiSIM_HV 1.2.4/2.2.0 Hiroshima High Voltage Version for LDMOS
With its OSDI interface all MOS models written in Verilog-A and compiled with OpenVAF are available
OSDI see 13.2
Examples: BSIMBULK, BSIM-CMG, BSIM-IMG, PSP, HiSIM (see VA-Models).
163
11.2.1 MOS Level 1

This model is also known as the ‘Shichman-Hodges’ model. This is the first model written
and the one often described in the introductory textbooks for electronics. This model is
applicable only to long channel devices. The use of Meyer’s model for the C-V part makes
it non charge conserving.
11.2.2 MOS Level 2

This model tries to overcome the limitations of the Level 1 model addressing several short-
channel effects, like velocity saturation. The implementation of this model is complicated
and this leads to many convergence problems. C-V calculations can be done with the
original Meyer model (non charge conserving).
11.2.3 MOS Level 3

This is a semi-empirical model derived from the Level 2 model. In the 80s this model
has often been used for digital design and, over the years, has proved to be robust. A
discontinuity in the model with respect to the KAPPA parameter has been detected (see
[10]). The supplied fix has been implemented in Spice3f2 and later. Since this fix may
affect parameter fitting, the option badmos3 may be set to use the old implementation (see
the section on simulation variables and the .options line). Ngspice level 3 implementation
takes into account length and width mask adjustments (xl and xw) and device width
narrowing due to diffusion (wd).
11.2.4 MOS Level 6

This model is described in [26]. The model can express the current characteristics of short-
channel MOSFETs at least down to 0.25 µm channel-length, GaAs FET, and resistance
inserted MOSFETs. The model evaluation time is about 1/3 of the evaluation time of
the SPICE3 mos level 3 model. The model also enables analytical treatments of circuits
in short-channel region and makes up for a missing link between a complicated MOSFET
current characteristics and circuit behaviors in the deep submicron region.
11.2.5 Notes on Level 1-6 models

The dc characteristics of the level 1 through level 3 MOSFETs are defined by the device
parameters vto, kp, lambda, phi and gamma. These parameters are computed by ngspice if
process parameters (nsub, tox, ...) are given, but users specified values always override.
vto is positive (negative) for enhancement mode and negative (positive) for depletion
mode N-channel (P-channel) devices.
Charge storage is modeled by three constant capacitors, cgso, cgdo, and cgbo, which
represent overlap capacitances, by the nonlinear thin-oxide capacitance that is distributed
among the gate, source, drain, and bulk regions, and by the nonlinear depletion-layer
capacitances for both substrate junctions divided into bottom and periphery, which vary
11.2. MOSFET MODELS (NMOS/PMOS) 165
as the mj and mjsw power of junction voltage respectively, and are determined by the
parameters cbd, cbs, cj, cjsw, mj, mjsw and pb.
Charge storage effects are modeled by the piecewise linear voltages-dependent capacitance
model proposed by Meyer. The thin-oxide charge-storage effects are treated slightly differ-
ent for the level 1 model. These voltage-dependent capacitances are included only if tox
is specified in the input description and they are represented using Meyer’s formulation.
There is some overlap among the parameters describing the junctions, e.g. the reverse
current can be input either as is (in A) or as js (in A/m2 ). Whereas the first is an absolute
value the second is multiplied by ad and as to give the reverse current of the drain and
source junctions respectively.
This methodology has been chosen since there is no sense in relating always junction
characteristics with ad and as entered on the device line; the areas can be defaulted. The
same idea applies also to the zero-bias junction capacitances cbd and cbs (in F) on one
hand, and cj (in F/m2 ) on the other.
The parasitic drain and source series resistance can be expressed as either rd and rs (in
ohms) or rsh (in ohms/sq.), the latter being multiplied by the number of squares nrd
and nrs input on the device line.
MOS level 1, 2, 3 and 6 parameters

LEVEL Model index - 1
VTO Zero-bias threshold V 0.0 1.0
voltage (VT 0 )
KP Transconductance A/V 2 2.0e-5 3.1e-5
parameter √
GAMMA Bulk threshold parameter V 0.0 0.37
PHI Surface potential (U) V 0.6 0.65
LAMBDA Channel length 1/V 0.0 0.02
modulation (MOS1 and
MOS2 only) (λ)
RD Drain ohmic resistance Ω 0.0 1.0
RS Source ohmic resistance Ω 0.0 1.0
CBD Zero-bias B-D junction F 0.0 20fF
capacitance
CBS Zero-bias B-S junction F 0.0 20fF
capacitance
IS Bulk junction saturation A 1.0e-14 1.0e-15
current (IS )
PB Bulk junction potential V 0.8 0.87
CGSO Gate-source overlap F/m 0.0 4.0e-11
capacitance per meter
channel width

CGDO Gate-drain overlap F/m 0.0 4.0e-11
channel width
CGBO Gate-bulk overlap F/m 0.0 2.0e-11
channel width
RSH Drain and source diffusion Ω/ 0.0 10
sheet resistance
CJ Zero-bias bulk junction F/m2 0.0 2.0e-4
bottom cap. per sq-meter
of junction area
MJ Bulk junction bottom - 0.5 0.5
grading coeff.
CJSW Zero-bias bulk junction F/m 0.0 1.0e-9
sidewall cap. per meter of
junction perimeter
0.50 (level1)
MJSW Bulk junction sidewall -
0.33 (level2, 3)
grading coeff.
JS Bulk junction saturation
current
TOX Oxide thickness m 1.0e-7 1.0e-7
NSUB Substrate doping cm−3 0.0 4.0e15
NSS Surface state density cm−2 0.0 1.0e10
NFS Fast surface state density cm−2 0.0 1.0e10
TPG Type of gate material: +1 - 1.0
opp. to substrate, -1 same
as substrate, 0 Al gate
XJ Metallurgical junction m 0.0 1M
depth
LD Lateral diffusion m 0.0 0.8M
UO Surface mobility cm2/V ·sec 600 700
UCRIT Critical field for mobility V /cm 1.0e4 1.0e4
degradation (MOS2 only)
UEXP Critical field exponent in - 0.0 0.1
mobility degradation
(MOS2 only)
UTRA Transverse field coeff. - 0.0 0.3
(mobility) (deleted for
MOS2)
VMAX Maximum drift velocity of m/s 0.0 5.0e4
carriers
NEFF Total channel-charge - 1.0 5.0
(fixed and mobile)
coefficient (MOS2 only)
KF Flicker noise coefficient - 0.0 1.0e-26
AF Flicker noise exponent - 1.0 1.2

FC Coefficient for - 0.5
forward-bias depletion
capacitance formula
DELTA Width effect on threshold - 0.0 1.0
voltage (MOS2 and
MOS3)
THETA Mobility modulation 1/V 0.0 0.1
(MOS3 only)
ETA Static feedback (MOS3 - 0.0 1.0
only)
KAPPA Saturation field factor - 0.2 0.5
(MOS3 only)
◦
temperature
11.2.6 MOS Level 9
Documentation is not available..
11.2.7 BSIM Models
Ngspice implements many of the BSIM models developed by Berkeley’s BSIM group.
BSIM stands for Berkeley Short-Channel IGFET Model and groups a class of models that
is continuously updated. BSIM3 (11.2.10) and BSIM4 (11.2.11) are industry standards
for CMOS processes down to 0.15 µm (BSIM3) and below (BSIM4), are very stable and
are supported by model parameter sets from foundries all over the world. BSIM1 and
BSIM2 are obsolete today.
In general, all parameters of BSIM models are obtained from process characterization,
in particular level 4 and level 5 (BSIM1 and BSIM2) parameters can be generated auto-
matically. J. Pierret [4] describes a means of generating a ‘process’ file, and the program
ngproc2mod provided with ngspice converts this file into a sequence of BSIM1 .model
lines suitable for inclusion in an ngspice input file.
Parameters marked below with an * in the l/w column also have corresponding parameters
with a length and width dependency. For example, vfb is the basic parameter with units
of Volts, and lvfb and wvfb also exist and have units of Volt-meter.
The formula
PL PW
P = P0 + + (11.1)
Leffective Weffective
is used to evaluate the parameter for the actual device specified with
Leffective = Linput − DL (11.2)

Weffective = Winput − DW (11.3)
Note that unlike the other models in ngspice, the BSIM models are designed for use
with a process characterization system that provides all the parameters, thus there are no
defaults for the parameters, and leaving one out is considered an error. For an example set
of parameters and the format of a process file, see the SPICE2 implementation notes [3].
For more information on BSIM2, see reference [5]. BSIM3 (11.2.10) and BSIM4 (11.2.11)
represent state of the art for submicron and deep submicron IC design.
11.2.8 BSIM1 model (level 4)

BSIM1 model (the first is a long series) is an empirical model. Developers placed less
emphasis on device physics and based the model on parametrical polynomial equations
to model the various physical effects. This approach pays in terms of circuit simulation
behavior but the accuracy degrades in the submicron region. A known problem of this
model is the negative output conductance and the convergence problems, both related to
poor behavior of the polynomial equations.
BSIM1 (level 4) parameters
Name Parameter Units l/w

VFB Flat-band voltage V *
PHI Surface inversion potential √V *
K1 Body effect coefficient V *
K2 Drain/source depletion charge-sharing - *
coefficient
ETA Zero-bias drain-induced barrier-lowering - *
coefficient
MUZ Zero-bias mobility cm2/V ·sec
DL Shortening of channel µm
DW Narrowing of channel µm
U0 Zero-bias transverse-field mobility 1/V *
degradation coefficient
U1 Zero-bias velocity saturation coefficient µ/V *
X2MZ Sens. of mobility to substrate bias at v=0 cm2/V 2 ·sec *
X2E Sens. of drain-induced barrier lowering 1/V *
effect to substrate bias
X3E Sens. of drain-induced barrier lowering 1/V *
effect to drain bias at Vds = Vdd
X2U0 Sens. of transverse field mobility 1/V 2 *
degradation effect to substrate bias
X2U1 Sens. of velocity saturation effect to µm/V 2 *
substrate bias
MUS Mobility at zero substrate bias and at cm2/V 2 sec
Vds = Vdd
Name Parameter Units l/w

X2MS Sens. of mobility to substrate bias at cm2/V 2 sec *
Vds = Vdd
X3MS Sens. of mobility to drain bias at Vds = Vdd cm2/V 2 sec *
X3U1 Sens. of velocity saturation effect on drain µm/V 2 *
bias at Vds=Vdd
TOX Gate oxide thickness µm
◦
TEMP Temperature where parameters were C
measured
VDD Measurement bias range V
CGDO Gate-drain overlap capacitance per meter F/m
channel width
CGSO Gate-source overlap capacitance per meter F/m
channel width
CGBO Gate-bulk overlap capacitance per meter F/m
channel length
XPART Gate-oxide capacitance-charge model flag -
N0 Zero-bias subthreshold slope coefficient - *
NB Sens. of subthreshold slope to substrate bias - *
ND Sens. of subthreshold slope to drain bias - *
RSH Drain and source diffusion sheet resistance Ω/
JS Source drain junction current density A/m2
PB Built in potential of source drain junction V

MJ Grading coefficient of source drain junction -
PBSW Built in potential of source, drain junction V
sidewall
MJSW Grading coefficient of source drain junction -
sidewall
CJ Source drain junction capacitance per unit F/m2
area
CJSW source drain junction sidewall capacitance F/m
per unit length

WDF Source drain junction default width m
DELL Source drain junction length reduction m
xpart = 0 selects a 40/60 drain/source charge partition in saturation, while xpart=1

selects a 0/100 drain/source charge partition. nd, ng, and ns are the drain, gate, and
source nodes, respectively. mname is the model name, area is the area factor, and off
indicates an (optional) initial condition on the device for dc analysis. If the area factor
is omitted, a value of 1.0 is assumed. The (optional) initial condition specification, using
ic=vds,vgs is intended for use with the uic option on the .tran control line, when a
transient analysis is desired starting from other than the quiescent operating point. See
the .ic control line for a better way to set initial conditions.
11.2.9 BSIM2 model (level 5)
This model contains many improvements over BSIM1 and is suitable for analog simulation.
Nevertheless, even BSIM2 breaks transistor operation into several distinct regions and this
leads to discontinuities in the first derivative in C-V and I-V characteristics that can cause
numerical problems during simulation.
11.2.10 BSIM3 model (levels 8, 49)
BSIM3 solves the numerical problems of previous models with the introduction of smooth-
ing functions. It adopts a single equation to describe device characteristics in the operating
regions. This approach eliminates the discontinuities in the I-V and C-V characteristics.
The original model, BSIM3 evolved through three versions: BSIM3v1, BSIM3v2 and
BSIM3v3. Both BSIM3v1 and BSIM3v2 had suffered from many mathematical problems
and were replaced by BSIM3v3. The latter is the only surviving release and has itself a
long revision history.
The following table summarizes the story of this model and their available ngspice versions:
Release Date Notes Version flag
BSIM3v3.0 10/30/1995 3.0
BSIM3v3.1 12/09/1996 3.1
BSIM3v3.2 06/16/1998 Revisions available: BSIM3v3.2.2, 3.2, 3.2.2,
BSIM3v3.2.3, and BSIM3v3.2.4 3.2.3, 3.2.4
Parallel processing with OpenMP is
available for BSIM3v3.2.4.
BSIM3v3.3 07/29/2005 Parallel processing with OpenMP is 3.3.0
available for this model.
BSIM3v2 and 3v3 models have been proven for accurate use in 0.18 µm technologies. The
model is publicly available as source code form from University of California, Berkeley.
A detailed description is given in the user’s manual available from here .
We recommend that you use only the most recent BSIM3 models (version 3.3.0), because
it contains corrections to all known bugs. To achieve that, change the version parameter
in your modelcard files to
VERSION = 3.3.0.
If no version number is given in the .model card, this (newest) version is selected as the
default.
A basic model card using only the intrinsic default parameters may look like
.model n1 nmos level=49 version=3.3.0

.model p1 pmos level=49 version=3.3.0
Unfortunately, due to historical reasons, these purely intrinsic parameters do not describe
realistic devices. A better minimum model configuration, roughly describing 0.35µm
transistors, is
.model n1 nmos level=49 version=3.3.0 tox=10n nch=1e17 nsub=5e16

.model p1 pmos level=49 version=3.3.0 tox=10n nch=1e17 nsub=5e16
BSIM3v3.2.4 supports the extra model parameter lmlt on channel length scaling and is
still used by many foundries today.
The older BSIM3 models will not be supported, they are made available for reference
only.
11.2.11 BSIM4 model (levels 14, 54)

This is the newest class of the BSIM family and introduces noise modeling and extrinsic
parasitics. BSIM4, as the extension of BSIM3 model, addresses the MOSFET physi-
cal effects into sub-100nm regime. It is a physics-based, accurate, scalable, robust and
predictive MOSFET SPICE model for circuit simulation and CMOS technology devel-
opment. It is developed by the BSIM Research Group in the Department of Electrical
Engineering and Computer Sciences (EECS) at the University of California, Berkeley (see
BSIM4 home page). BSIM4 has a long revision history. The models offered by ngspice
are summarized below.
Release Date Notes Version flag
BSIM4.5.0 07/29/2005 ** 4.5.0
BSIM4.6.5 22/09/2009 ** 4.6.5
BSIM4.7.0 04/08/2011 ** 4.7
BSIM4.8.2 10/01/2020 ** 4.8
**) Parallel processing using OpenMP support is available for this model.
Details of any revision are to be found in the Berkeley user’s manuals, a pdf download of
the most recent edition is to be found here.
We recommend that you use only the most recent BSIM4 model (version 4.8.2), because
it contains corrections to all known bugs. To achieve that, change the version parameter
in your modelcard files to
VERSION = 4.8.2
If no version number is given in the .model card, this (newest) version is selected as the
default. The older models will typically not be supported, they are made available for
reference only.
The basic model card, using only the intrinsic default parameters, already delivers rea-
sonable device characteristics.
.model n1 nmos level=54 version=4.8.2

.model p1 pmos level=54 version=4.8.2
11.2.12 BSIMSOI models (levels 10, 58, 55, 56, 57)

BSIMSOI is a SPICE compact model for SOI (Silicon-On-Insulator) circuit design, created
by University of California at Berkeley. This model is formulated on top of the BSIM3
framework. It shares the same basic equations with the bulk model so that the physical
nature and smoothness of BSIM3v3 are retained. Four models are supported in ngspice,
those based on BSIM3 and modeling fully depleted (FD, level 55), partially depleted
(PD, level 57) and both (DD, level 56), as well as the modern BSIMSOI version 4 model
(levels 10, 58). Detailed descriptions are beyond the scope of this manual, but see e.g.
BSIMSOIv4.4 User Manual for a very extensive description of the recent model version.
OpenMP support is available for levels 10, 58, version 4.4.
11.2.13 SOI3 model (level 60)

see literature citation [18] for a description.
11.2.14 HiSIM models of the University of Hiroshima

There are two model implementations available - see also HiSIM Research Center:
1. HiSIM2 model: Surface-Potential-Based MOSFET Model for Circuit Simulation

version 2.8.0 - level 68 (see link to HiSIM2 for source code and manual).
2. HiSIM_HV model: Surface-Potential-Based HV/LD-MOSFET Model for Circuit

Simulation version 1.2.4 and 2.2.0 - level 73 (see link to HiSIM_HV for source code
and manual).
11.2.15 MOS models available via OpenVAF/OSDI

With its integrated OSDI interface and the OpenVAF compiler (see chapter 13 for details),
ngspice makes available several Verilog-A compact MOS models. To obtain the sources
you may visit the github repository VA-Models which assembles most of the publicly
available Verilog-A compact models. To just name a few models:
11.2.15.1 PSP model
The PSP model is a compact MOSFET model intended for digital, analog and RF-design,
which is jointly developed by NXP Semiconductors Research (formerly part of Philips),
different universities and CEA-Leti.
PSP is a surface-potential based MOS Model, containing all relevant physical effects to
model present-day and upcoming deep-submicron bulk CMOS technologies:
• mobility reduction
• velocity saturation drain induced barrier lowering DIBL
• gate current
• lateral doping gradient effects
• STI stress
11.3. POWER MOSFET MODEL (VDMOS) 173
The source/drain junction model, c.q. the JUNCAP2 model, is fully integrated in PSP.
Detailes information and the most recent version of the model documentation are available
on the the CEA-Leti web site, see also the PSP Summary.
11.2.15.2 BSIM-BULK model
BSIM-BULK is the successor to BSIM4, with high accuracy compared to measured data
in all regions of operation. It features model symmetry valued for analog and RF appli-
cations.
11.2.15.3 BSIM-CMG model
BSIM-CMG (Common Multi-Gate) is a compact model for the class of common multi-gate
FETs, namely FinFETs, Nanowire and Gate-All-Around transistors.
11.3 Power MOSFET model (VDMOS)
The VDMOS model is a relativly simple power MOS model with 3 terminals drain, gate
and source. Its current equations are partly based on a modified MOS1 model. The
gate-source capacitance is set to a constant value by parameter Cgs. The drain-source
capacitance is evaluated from parameters Cgdmax, Cgdmin, and A. The drain-source ca-
pacitance is that of a parallel pn diode and calculated by Cjo, fc, and m. Leakage and
breakdown are modeled by the parallel pn diodes as well, using is and other param-
eters. A subthreshold current model is available, using a single parameter ksubthres.
Quasi-saturation is modelled with parameters rq and vq. Mtriode may be used here as
well.
The thermal network of the VDMOS model is shown in Fig. 11.1.
This model does not have a level parameter. It is invoked by the VDMOS token preceding
the parameters on the .model line. P-channel or n-channel are selected by the model
parameter PCHAN and NCHAN. If no flag is given, n-channel is the default. Standard
MOS instance parameters W and L are not acknowledged because they are no design
parameters and are not provided by the device manufacturers.
The following ’parameters’ in the .model line are no model parameters, but serve in-
formation purposes for the user: mfg=..., Vds=..., Ron=..., and Qg=... They are
ignored by ngspice.
Figure 11.1: VDMOS model including thermal network
General form:
MXXXXXXX nd ng ns mname <m = val > < temp =t > < dtemp =t >
. model mname VDMOS < Pchan > < parameters >
Example:
M1 24 2 0 IXTH48P20P
. MODEL IXTH48P20P VDMOS Pchan Vds =200 VTO = -4 KP =10 Lambda =5 m
+ Mtriode =0.3 Ksubthres =120 m Rs =10 m Rd =20 m Rds =200 e6
+ Cgdmax =6000 p Cgdmin =100 p A =0.25 Cgs =5000 p Cjo =9000 p
+ Is =2 e -6 Rb =20 m BV =200 IBV =250 e -6 NBV =4 TT =260 e -9
VDMOS instance parameters

m device multiplier - 1 -
off Device initially off - 0
icvds Initial D-S voltage V 0.0
icvgs Initial G-S voltage V 0.0
◦
temp device temperature C 27 100
◦
dtemp device temperature C 0.0 50
difference
ic Vector of D-S, G-S V 0.0
voltages
thermal Thermal model switch - -
on/off
VDMOS model parameters

VDMOS select VDMOS model - must given -

KF Flicker noise coefficient - 0.0
AF Flicker noise exponent - 1.0
◦
temperature
RQ Quasi saturation Ω 0.0 0.5

resistance fitting
parameter
VQ Quasi saturation voltage V 0.0 100
fitting parameter
MTRIODE Conductance multiplier in − 1.0 0.8
triode region
SUBSHIFT shift along gate voltage V 0.0

axis in the dual parameter
subthreshold model
KSUBTHRES slope in the single - 0.1 0.27
parameter subthreshold
model
BV Vds breakdown voltage V ∞

IBV Current at Vds=bv A 1.0e-10
NBV Vds breakdown emission - 1.0
coefficient
RDS Drain-source shunt Ω ∞ 1e7
resistance
RB Body diode ohmic Ω 0.0 14m
resistance
N Body diode emission - 1.0 1.1
coefficient
TT Body diode transit time s 0.0
EG Body diode activation eV 1.11
energy for temperature
effect on IS
XTI Body diode saturation - 3.0 3.2
current temperature
exponent
IS Body diode saturation A 1e-14 60p
current
VJ Body diode junction V 0.8
potential
FC Body diode coefficient for - 0.5

forward-bias depletion
capacitance formula

CJO Zero-bias body diode F 0.0 1.5n
junction capacitance
M Body diode grading - 0.5 0.6
coefficient
CGDMIN Minimum non-linear G-D F 0.0 10p

capacitance
CGDMAX Maximum non-linear G-D F 0.0 2.45n
capacitance
A Non-linear Cgd - 1 0.3
capacitance parameter
CGS Gate-source capacitance F 0.0 1.2n
TCVTH (VTOTC) Linear Vth0 temperature 1/°C 0.0 0.0065

coefficient
MU (BEX) Exponent of gain - -1.5 -1.27
temperature dependency
TEXP0 Drain resistance rd0 - 1.5
TEXP1 Drain resistance rd1 - 0.3
TRD1 Drain resistance linear 1/°C 0.0
TRD2 Drain resistance quadratic 1/(°C)2 0.0
TRG1 Gate resistance linear 1/°C 0.0
TRG2 Gate resistance quadratic 1/(°C)2 0.0
TRS1 Source resistance linear 1/°C 0.0
TRS2 Source resistance 1/(°C)2 0.0
quadratic temperature
coefficient
TRB1 Body resistance linear 1/°C 0.0
TRB2 Body resistance quadratic 1/(°C)2 0.0
TKSUBTHRES1 Linear temperature 1/°C 0.0
coefficient of ksubthres
TKSUBTHRES2 Quadratic temperature 1/(°C)2 0.0
coefficient of ksubthres
RTHJC Thermal resistance K/W 1.0 0.4

junction-case
CTHJ Thermal capacitance J/K 10e-6 5e-3

RTHCA Thermal resistance K/W 1000
case-ambient (w/o
heatsink)
VDMOS electro-thermal model
Power electronic devices behavior the effect of self-heating effect. That means that the
dissipated power has an impact to the electrical behavior of the terminal currents. To
minimize this effect and to protect the element from thermal destruction heat sinks are
supplied to this kind of power devices.
The ngspice VDMOS model has introduced an electro-thermal approach by stamping
additional elements into the circuit matrix and by iteration the additional current control
inside the spice solver.
The transistor now has 5 nodes. Besides D, G, and S we have TJ and TCASE. The
additional nodes must be activated by the device switch THERMAL. Heat is generated
in the MOS channel and peripheral elements like resistors, its temperature is available
and may be measured at node TJ, and is fed back internally into the device equations.
Within the transistor package the heat is flowing from the channel to the metal surface
of the case, at node TCASE. Here you may connect a heat sink, to offer a flow path
for the heat away from the device. The internal heat resistance is RTHJC (junction to
case), a typical data sheet value. The model also includes the heat capacitance CTHJ
of the semiconductor die and package (typically not available in the data sheet, so to be
estimated only).
The following example show the usage of ngspice electro-thermal model including a simple
heat sink:
General form:
MXXXXXXX nd ng ns tj tc mname thermal <m = val > < temp =t > < dtemp =t >
Example:
M1 24 2 0 tj tc IXTH48P20P thermal
rcs tc 1 0.1
csa 1 0 30 m
rsa 1 amb 1.3
VTamb tamb 0 25
. MODEL IXTH48P20P VDMOS Pchan Vds =200 VTO = -4 KP =10 Lambda =5 m
+ Mtriode =0.3 Ksubthres =120 m Rs =10 m Rd =20 m Rds =200 e6
+ Cgdmax =6000 p Cgdmin =100 p A =0.25 Cgs =5000 p Cjo =9000 p
+ Is =2 e -6 Rb =20 m BV =200 IBV =250 e -6 NBV =4 TT =260 e -9
+ Rthjc =0.4 Cthj =5 e -3
Chapter 12
Mixed-Mode and Behavioral

Modeling with XSPICE
Ngspice implements XSPICE extensions for behavioral and mixed-mode (analog and dig-
ital) modeling. In the XSPICE framework this is referred to as code level modeling.
Behavioral modeling may benefit dramatically because XSPICE offers a means to add
analog functionality programmed in C. Many examples (amplifiers, oscillators, filters ...)
are presented in the following. Even more flexibility is available because you may define
your own models and use them in addition and in combination with all the already exist-
ing ngspice functionality. Digital and mixed mode simulation is sped up significantly by
simulating the digital part in an event driven manner, in that state equations use only a
few allowed states and are evaluated only during switching, and not continuously in time
and signal as in a pure analog simulator.
This chapter describes the predefined models available in ngspice, stemming from the
original XSPICE simulator or being added to enhance the usability. The instructions for
writing new code models are given in Chapt. 28.
To make use of the XSPICE extensions, you need to compile them in. Linux, CYGWIN,
MINGW and other users may add the flag --enable-xspice to their ./configure com-
mand and then recompile. The pre-built ngspice for Windows distribution has XSPICE
already enabled. For detailed compiling instructions see Chapt. 32.1.
12.1 Code Model Element & .MODEL Cards
12.1.1 Syntax
Ngspice includes a library of predefined ‘Code Models’ that can be placed within any
circuit description in a manner similar to that used to place standard device models.
Code model instance cards always begin with the letter ‘A’, and always make use of a
.MODEL card to describe the code model desired. Section 28 of this document goes into
greater detail as to how a code model similar to the predefined models may be developed,
but once any model is created and linked into the simulator it may be placed using
one instance card and one .MODEL card (note here we conform to the SPICE custom of
referring to a single logical line of information as a ‘card’). As an example, the following
179
180CHAPTER 12. MIXED-MODE AND BEHAVIORAL MODELING WITH XSPICE
uses a predefined ‘gain’ code model taking as an input some value on node 1, multiplies
it by a gain of 5.0, and outputs the new value to node 2. Note that, by convention, input
ports are specified first on code models. Output ports follow the inputs.
Example:
a1 1 2 amp
. model amp gain ( gain =5.0)
In this example the numerical values picked up from single-ended (i.e. ground referenced)
input node 1 and output to single-ended output node 2 will be voltages, since in the
Interface Specification File for this code model (i.e., gain), the default port type is spec-
ified as a voltage (more on this later). However, if you didn’t know this, the following
modifications to the instance card could be used to insure it:
Example:
a1 % v (1) % v (2) amp

. model amp gain ( gain =5.0)
The specification %v preceding the input and output node numbers of the instance card
indicate to the simulator that the inputs to the model should be single-ended voltage
values. Other possibilities exist, as described later.
Some of the other features of the instance and .MODEL cards are worth noting. Of particu-
lar interest is the portion of the .MODEL card that specifies gain=5.0. This portion of the
card assigns a value to a parameter of the ‘gain’ model. There are other parameters that
can be assigned values for this model, and in general code models will have several. In
addition to numeric values, code model parameters can take non-numeric values (such as
TRUE and FALSE), and even vector values. All of these topics will be discussed at length
in the following pages. In general, however, the instance and .MODEL cards that define a
code model will follow the abstract form described below. This form illustrates that the
number of inputs and outputs and the number of parameters that can be specified is rel-
atively open-ended and can be interpreted in a variety of ways (note that angle-brackets
‘<’ and ‘>’ enclose optional inputs):
12.1. CODE MODEL ELEMENT & .MODEL CARDS 181
Example:
AXXXXXXX <%v ,% i ,% vd ,% id ,% g ,% gd ,% h ,% hd , or %d >

+ <[ > <~ > <%v ,% i ,% vd ,% id ,% g ,% gd ,% h ,% hd , or %d >
+ < NIN1 or + NIN1 - NIN1 or " null " >
+ <~ >... < NIN2 .. <] > >
+ <%v ,% i ,% vd ,% id ,% g ,% gd ,% h ,% hd ,% d or % vnam >
+ <[ > <~ > <%v ,% i ,% vd ,% id ,% g ,% gd ,% h ,% hd ,
or %d > < NOUT1 or + NOUT1 - NOUT1 >
+ <~ >... < NOUT2 .. <] > >
+ MODELNAME
. MODEL MODELNAME MODELTYPE

+ <( PARAMNAME1 = <[ > VAL1 < VAL2 ... <] > > PARAMNAME2 .. >) >
Square brackets ([ ]) are used to enclose vector input nodes. In addition, these brackets
are used to delineate vectors of parameters.
The literal string ‘null’, when included in a node list, is interpreted as no connection at
that input to the model. ‘Null’ is not allowed as the name of a model’s input or output if
the model only has one input or one output. Also, ‘null’ should only be used to indicate a
missing connection for a code model; use on other XSPICE component is not interpreted
as a missing connection, but will be interpreted as an actual node name.
The tilde, ‘~’, when prepended to a digital node name, specifies that the logical value of
that node be inverted prior to being passed to the code model. This allows for simple
inversion of input and output polarities of a digital model in order to handle logically
equivalent cases and others that frequently arise in digital system design. The following
example defines a NAND gate, one input of which is inverted:
a1 [~1 2] 3 nand1
. model nand1 d_nand ( rise_delay =0.1 fall_delay =0.2)
The optional symbols %v, %i, %vd, etc. specify the type of port the simulator is to expect
for the subsequent port or port vector. The meaning of each symbol is given in Table
12.1.
The symbols described in Table 12.1 may be omitted if the default port type for the model
is desired. Note that non-default port types for multi-input or multi-output (vector) ports
must be specified by placing one of the symbols in front of EACH vector port. On the
other hand, if all ports of a vector port are to be declared as having the same non-default
type, then a symbol may be specified immediately prior to the opening bracket of the
vector. The following examples should make this clear:
Example 1: - Specifies two differential voltage connections, one

Port Type Modifiers

Modifier Interpretation
%v represents a single-ended voltage port - one node name or number is
expected for each port.
%i represents a single-ended current port - one node name or number is
%g represents a single-ended voltage-input, current-output (VCCS) port -
one node name or number is expected for each port. This type of port is
automatically an input/output.
%h represents a single-ended current-input, voltage-output (CCVS) port -
one node name or number is expected for each port. This type of port is
automatically an input/output.
%d represents a digital port - one node name or number is expected for each
port. This type of port may be either an input or an output.
%vnam represents the name of a voltage source, the current through which is
taken as an input. This notation is provided primarily in order to allow
models defined using SPICE2G6 syntax to operate properly in XSPICE.
%vd represents a differential voltage port - two node names or numbers are
%id represents a differential current port - two node names or numbers are
%gd represents a differential VCCS port - two node names or numbers are
%hd represents a differential CCVS port - two node names or numbers are
Table 12.1: Port Type Modifiers

12.1. CODE MODEL ELEMENT & .MODEL CARDS 183
to nodes 1 & 2, and one to nodes 3 & 4.

%vd [1 2 3 4]
Example 2: - Specifies two single-ended connections to node 1 and

at node 2, and one differential connection to
nodes 3 & 4.
%v [1 2 %vd 3 4]
Example 3: - Identical to the previous example...parenthesis

are added for additional clarity.
%v [1 2 %vd(3 4)]
Example 4: - Specifies that the node numbers are to be treated in the

default fashion for the particular model.
If this model had ‘%v” as a default for this
port, then this notation would represent four single-ended
voltage connections.
[1 2 3 4]
The parameter names listed on the .MODEL card must be identical to those named in the
code model itself. The parameters for each predefined code model are described in detail
in Sections 12.2 (analog), 12.3 (Hybrid, A/D) and 12.4 (digital). The steps required in
order to specify parameters for user-defined models are described in Chapter 28.
12.1.2 Examples
The following is a list of instance card and associated .MODEL card examples showing use
of predefined models within an XSPICE deck:
a1 1 2 amp
.model amp gain(in_offset=0.1 gain=5.0 out_offset=-0.01)
a2 %i[1 2] 3 sum1
.model sum1 summer(in_offset=[0.1 -0.2] in_gain=[2.0 1.0]
+ out_gain=5.0 out_offset=-0.01)
a21 %i[1 %vd(2 5) 7 10] 3 sum2

.model sum2 summer(out_gain=10.0)
a5 1 2 limit5
.model limit5 limit(in_offset=0.1 gain=2.5
+ out_lower_limit=-5.0 out_upper_limit=5.0 limit_range=0.10
+ fraction=FALSE)
a7 2 %id(4 7) xfer_cntl1
.model xfer_cntl1 pwl(x_array=[-2.0 -1.0 2.0 4.0 5.0]

+ y_array=[-0.2 -0.2 0.1 2.0 10.0]
+ input_domain=0.05 fraction=TRUE)
a8 3 %gd(6 7) switch3
.model switch3 aswitch(cntl_off=0.0 cntl_on=5.0 r_off=1e6
+ r_on=10.0 log=TRUE)
12.1.3 Search path for file input
Several code models (filesource 12.2.9, d_source 12.4.21, d_state 12.4.18) call addi-
tional files for supply of input data. A call to file="path/filename" (or input_file=,
state_file=) in the .model card will start a search sequence for finding the file. path
may be an absolute path. If path is omitted or is a relative path, filename is looked for
according to the following search list:
Infile_Path/<path/filename> (Infile_Path is the path of the input file *.sp containing

the netlist)
NGSPICE_INPUT_DIR/<path/filename> (where an additional path is set by the environ-

mental variable)
<path/filename> (where the search is relative to the current directory (OS dependent))
12.1.4 Code model location and assessment
To make use of the XSPICE extensions, you have to compile ngspice accordingly (see
Chapt. 32.1). ngspice then is prepared to load and use the code models. At the same time
the code models are re-made. They are, however, not linked into ngspice at compile time,
but reside in extra shared libraries or dlls, with names analog.cm, digital.cm, spice2poly.cm,
xtradev.cm, xtraevt.cm, and table.cm. At run time, with XSPICE enabled, they are loaded
dynamically into ngspice by the command codemodel (17.5.15). The sequence to load
the codemodels is: Upon start-up ngspice locates, reads, and executes spinit, the standard
initialization file (16.5). Within spinit, you will find the commands to load the codemodels,
typically with a path for the code models relative to the current working directory (the
location of ngspice, in case of shared ngspice the location of the caller).
If you don’t want to make use of spinit, you may run a script in ngspice, before loading
any circuit, which contains the codemodel commands. When using shared ngspice, one
may issue the codemodel commands directly after initialization, with absolute path or
path relative to the current working directory.
In a standard ngspice installation in MS Windows, the codemodels are located in ../lib/ngspice,
e.g. in C:\Spice64\lib\ngspice (see also 32.2.1).
In Linux, it depends on the OS invocation. In openSUSE you may find the codemodels
in /usr/local/lib64/ngspice, while ngspice resides in /usr/local/bin.
12.2. ANALOG MODELS 185
12.2 Analog Models

The following analog models are supplied with XSPICE. The descriptions included consist
of the model Interface Specification File and a description of the model’s operation. This
is followed by an example of a simulator-deck placement of the model, including the
.MODEL card and the specification of all available parameters.
12.2.1 Gain
NAME_TABLE:
C_Function_Name: cm_gain
Spice_Model_Name: gain
Description: "A simple gain block"
PORT_TABLE:
Port Name: in out
Description: "input" "output"
Direction: in out
Default_Type: v v
Allowed_Types: [v,vd,i,id,vnam] [v,vd,i,id]
Vector: no no
Vector.Bounds: - -
Null.Allowed: no no
PARAMETER_TABLE:
Parameter_Name: in_offset gain out_offset
Description: "input offset" "gain" "output offset"
Data_Type: real real real
Default_Value: 0.0 1.0 0.0
Limits: - - -
Vector: no no no
Vector_Bounds: - - -
Null_Allowed: yes yes yes
Description: This function is a simple gain block with optional offsets on the input and
the output. The input offset is added to the input, the sum is then multiplied by
the gain, and the result is produced by adding the output offset. This model will
operate in DC, AC, and Transient analysis modes.
Example:
a1 1 2 amp
. model amp gain ( in_offset =0.1 gain =5.0
+ out_offset = -0.01)
12.2.2 Summer
NAME_TABLE:
C_Function_Name: cm_summer
Spice_Model_Name: summer
Description: "A summer block"
PORT_TABLE:
Port Name: in out
Description: "input vector" "output"
Direction: in out
Default_Type: v v
Vector: yes no
Vector_Bounds: - -
Null_Allowed: no no
PARAMETER_TABLE:
Parameter_Name: in_offset in_gain
Description: "input offset vector" "input gain vector"
Data_Type: real real
Default_Value: 0.0 1.0
Limits: - -
Vector: yes yes
Vector_Bounds: in in
Null_Allowed: yes yes
PARAMETER_TABLE:
Parameter_Name: out_gain out_offset
Description: "output gain" "output offset"
Limits: - -
Vector: no no
Vector_Bounds: - -
Description: This function is a summer block with 2-to-N input ports. Individual gains
and offsets can be applied to each input and to the output. Each input is added to
its respective offset and then multiplied by its gain. The results are then summed,
multiplied by the output gain and added to the output offset. This model will
operate in DC, AC, and Transient analysis modes.
Example usage:
a2 [1 2] 3 sum1
. model sum1 summer ( in_offset =[0.1 -0.2] in_gain =[2.0 1.0]
+ out_gain =5.0 out_offset = -0.01)
12.2.3 Multiplier
NAME_TABLE:
C_Function_Name: cm_mult
Spice_Model_Name: mult
Description: "multiplier block"
PORT_TABLE:
Port_Name: in out
Description: "input vector" "output"
Direction: in out
Default_Type: v v
Vector: yes no
Vector_Bounds: [2 -] -
Null_Allowed: no no
PARAMETER_TABLE:
Parameter_Name: in_offset in_gain
Description: "input offset vector" "input gain vector"
Limits: - -
Vector: yes yes
Vector_Bounds: in in
PARAMETER_TABLE:
Limits: - -
Vector: no no
Vector_Bounds: - -
Description: This function is a multiplier block with 2-to-N input ports. Individual
gains and offsets can be applied to each input and to the output. Each input
is added to its respective offset and then multiplied by its gain. The results are
multiplied along with the output gain and are added to the output offset. This
model will operate in DC, AC, and Transient analysis modes. However, in ac
analysis it is important to remember that results are invalid unless only one input
of the multiplier is connected to a node that i connected to an AC signal (this is
exemplified by the use of a multiplier to perform a potentiometer function: one
input is DC, the other carries the AC signal).
Example SPICE Usage:
a3 [1 2 3] 4 sigmult
. model sigmult mult ( in_offset =[0.1 0.1 -0.1]
+ in_gain =[10.0 10.0 10.0] out_gain =5.0 out_offset =0.05)
12.2.4 Divider
NAME_TABLE:
C_Function_Name: cm_divide
Spice_Model_Name: divide
Description: "divider block"
PORT_TABLE:
Port_Name: num den out
Description: "numerator" "denominator" "output"
Direction: in in out
Default_Type: v v v
Allowed_Types: [v,vd,i,id,vnam] [v,vd,i,id,vnam] [v,vd,i,id]
Vector: no no no
Null_Allowed: no no no
PARAMETER_TABLE:
Parameter_Name: num_offset num_gain
Description: "numerator offset" "numerator gain"
Limits: - -
Vector: no no
Vector_Bounds: - -
PARAMETER_TABLE:
Parameter_Name: den_offset den_gain
Description: "denominator offset" "denominator gain"
Limits: - -
Vector: no no
Vector_Bounds: - -
PARAMETER_TABLE:
Parameter_Name: den_lower_limit
Description: "denominator lower limit"
Data_Type: real
Default_Value: 1.0e-10
Limits: -
Vector: no
Vector_Bounds: -
Null_Allowed: yes
PARAMETER_TABLE:
Parameter_Name: den_domain
Description: "denominator smoothing domain"
Data_Type: real
Limits: -
Vector: no
Vector_Bounds: -
Null_Allowed: yes
PARAMETER_TABLE:
Parameter_Name: fraction
Description: "smoothing fraction/absolute value switch"
Data_Type: boolean
Default_Value: false
Limits: -
Vector: no
Vector_Bounds: -
Null_Allowed: yes
PARAMETER_TABLE:
Limits: - -
Vector: no no
Vector_Bounds: - -
Description: This function is a two-quadrant divider. It takes two inputs; num (nu-
merator) and den (denominator). Divide offsets its inputs, multiplies them by their
respective gains, divides the results, multiplies the quotient by the output gain, and
offsets the result. The denominator is limited to a value above zero via a user spec-
ified lower limit. This limit is approached through a quadratic smoothing function,
the domain of which may be specified as a fraction of the lower limit value (default),
or as an absolute value. This model will operate in DC, AC and Transient analysis
modes. However, in ac analysis it is important to remember that results are invalid
unless only one input of the divider is connected to a node that is connected to an
ac signal (this is exemplified by the use of the divider to perform a potentiometer
function: one input is dc, the other carries the ac signal).

a4 1 2 4 divider
.model divider divide(num_offset=0.1 num_gain=2.5 den_offset=-0.1
+ den_gain=5.0 den_lower_limit=1e-5 den_domain=1e-6
+ fraction=FALSE out_gain=1.0 out_offset=0.0)
12.2.5 Limiter
NAME_TABLE:
C_Function_Name: cm_limit
Spice_Model_Name: limit
Description: "limit block"
PORT_TABLE:
Port Name: in out
Direction: in out
Default_Type: v v
Allowed_Types: [v,vd,i,id] [v,vd,i,id]
Vector: no no
Vector_Bounds: - -
Null_Allowed: no no
PARAMETER_TABLE:
Parameter_Name: in_offset gain
Description: "input offset" "gain"
Limits: - -
Vector: no no
Vector_Bounds: - -
PARAMETER_TABLE:
Parameter_Name: out_lower_limit out_upper_limit
Description: "output lower limit" "output upper limit"
Limits: - -
Vector: no no
Vector_Bounds: - -
PARAMETER_TABLE:
Parameter_Name: limit_range
Description: "upper & lower smoothing range"
Data_Type: real
Limits: -
Vector: no
Vector_Bounds: -
Null_Allowed: yes
PARAMETER_TABLE:
Data_Type: boolean
Default_Value: FALSE
Limits: -
Vector: no
Vector_Bounds: -
Null_Allowed: yes
Description: The Limiter is a single input, single output function similar to the Gain
Block. However, the output of the Limiter function is restricted to the range speci-
fied by the output lower and upper limits. This model will operate in DC, AC and
Transient analysis modes. Note that the limit range is the value below the upper
limit and above the lower limit at which smoothing of the output begins. For this
model, then, the limit range represents the delta with respect to the output level at
which smoothing occurs. Thus, for an input gain of 2.0 and output limits of 1.0 and
-1.0 volts, the output will begin to smooth out at ±0.9 volts, which occurs when
the input value is at ±0.4.

a5 1 2 limit5
.model limit5 limit(in_offset=0.1 gain=2.5 out_lower_limit=-5.0
+ out_upper_limit=5.0 limit_range=0.10 fraction=FALSE)
12.2.6 Controlled Limiter

NAME_TABLE:
C_Function_Name: cm_climit
Spice_Model_Name: climit
Description: "controlled limiter block"
PORT_TABLE:
Port_Name: in cntl_upper
Description: "input" "upper lim. control input"
Direction: in in
Default_Type: v v
Allowed_Types: [v,vd,i,id,vnam] [v,vd,i,id,vnam]
Vector: no no
Vector_Bounds: - -
Null_Allowed: no no
PORT_TABLE:
Port_Name: cntl_lower out
Description: "lower limit control input" "output"
Direction: in out
Default_Type: v v
Vector: no no
Vector_Bounds: - -
Null_Allowed: no no
PARAMETER_TABLE:
Limits: - -
Vector: no no
Vector_Bounds: - -
PARAMETER_TABLE:
Parameter_Name: upper_delta lower_delta
Description: "output upper delta" "output lower delta"
Limits: - -
Vector: no no
Vector_Bounds: - -
PARAMETER_TABLE:
Parameter_Name: limit_range fraction
Description: "upper & lower sm. range" "smoothing %/abs switch"
Data_Type: real boolean
Default_Value: 1.0e-6 FALSE
Limits: - -
Vector: no no
Vector_Bounds: - -
Description: The Controlled Limiter is a single input, single output function similar
to the Gain Block. However, the output of the Limiter function is restricted to
the range specified by the output lower and upper limits. This model will oper-
ate in DC, AC, and Transient analysis modes. Note that the limit range is the
value below the cntl_upper limit and above the cntl_lower limit at which smooth-
ing of the output begins (minimum positive value of voltage must exist between the
cntl_upper input and the cntl_lower input at all times). For this model, then, the
limit range represents the delta with respect to the output level at which smooth-
ing occurs. Thus, for an input gain of 2.0 and output limits of 1.0 and -1.0 volts,
the output will begin to smooth out at ±0.9 volts, which occurs when the input
value is at ±0.4. Note also that the Controlled Limiter code tests the input val-
ues of cntl_upper and cntl_lower to make sure that they are spaced far enough
apart to guarantee the existence of a linear range between them. The range is cal-
culated as the difference between (cntl_upper − upper_delta − limit_range) and
(cntl_lower + lower_delta + limit_range) and must be greater than or equal to
zero. Note that when the limit range is specified as a fractional value, the limit
range used in the above is taken as the calculated fraction of the difference between
cntl_upper and cntl_lower. Still, the potential exists for too great a limit range
value to be specified for proper operation, in which case the model will return an
error message.

a6 3 6 8 4 varlimit
.
.
.model varlimit climit(in_offset=0.1 gain=2.5 upper_delta=0.0
+ lower_delta=0.0 limit_range=0.10 fraction=FALSE)
12.2.7 PWL Controlled Source

NAME_TABLE:
C_Function_Name: cm_pwl
Spice_Model_Name: pwl
Description: "piecewise linear controlled source"
PORT_TABLE:
Port_Name: in out
Direction: in out
Default_Type: v v
Vector: no no
Vector_Bounds: - -
Null_Allowed: no no
PARAMETER_TABLE:
Parameter_Name: x_array y_array
Description: "x-element array" "y-element array"
Default_Value: - -
Limits: - -
Vector: yes yes
Vector_Bounds: [2 -] [2 -]
Null_Allowed: no no
PARAMETER_TABLE:
Parameter_Name: input_domain fraction
Description: "input sm. domain" "smoothing %/abs switch"
Default_Value: 0.01 TRUE
Limits: [1e-12 0.5] -
Vector: no no
Vector_Bounds: - -
STATIC_VAR_TABLE:
Static_Var_Name: last_x_value
Data_Type: pointer
Description: "iteration holding variable for limiting"
Description: The Piece-Wise Linear Controlled Source is a single input, single output
function similar to the Gain Block. However, the output of the PWL Source is
not necessarily linear for all values of input. Instead, it follows an I/O relationship
specified by you via the x_array and y_array coordinates. This is detailed below.
The x_array and y_array values represent vectors of coordinate points on the x
and y axes, respectively. The x_array values are progressively increasing input co-
ordinate points, and the associated y_array values represent the outputs at those
points. There may be as few as two (x_array[n], y_array[n]) pairs specified, or
as many as memory and simulation speed allow. This permits you to very finely
approximate a non-linear function by capturing multiple input-output coordinate
points.
Two aspects of the PWL Controlled Source warrant special attention. These are
the handling of endpoints and the smoothing of the described transfer function near
coordinate points.
In order to fully specify outputs for values of in outside of the bounds of the
PWL function (i.e., less than x_array[0] or greater than x_array[n], where n is
the largest user-specified coordinate index), the PWL Controlled Source model ex-
tends the slope found between the lowest two coordinate pairs and the highest two
coordinate pairs. This has the effect of making the transfer function completely
linear for in less than x_array[0] and in greater than x_array[n]. It also has the
potentially subtle effect of unrealistically causing an output to reach a very large or
small value for large inputs. You should thus keep in mind that the PWL Source
does not inherently provide a limiting capability.
In order to diminish the potential for non-convergence of simulations when using the
PWL block, a form of smoothing around the x_array, y_array coordinate points
is necessary. This is due to the iterative nature of the simulator and its reliance on
smooth first derivatives of transfer functions in order to arrive at a matrix solution.
Consequently, the input_domain and fraction parameters are included to allow
you some control over the amount and nature of the smoothing performed.
Fraction is a switch that is either TRUE or FALSE. When TRUE (the default
setting), the simulator assumes that the specified input domain value is to be in-
terpreted as a fractional figure. Otherwise, it is interpreted as an absolute value.
Thus, if fraction=TRUE and input_domain=0.10, The simulator assumes that
the smoothing radius about each coordinate point is to be set equal to 10% of the
length of either the x_array segment above each coordinate point, or the x_array
segment below each coordinate point. The specific segment length chosen will be
the smallest of these two for each coordinate point.
On the other hand, if fraction=FALSE and input_domain=0.10, then the simula-
tor will begin smoothing the transfer function at 0.10 volts (or amperes) below each
x_array coordinate and will continue the smoothing process for another 0.10 volts
(or amperes) above each x_array coordinate point. Since the overlap of smoothing
domains is not allowed, checking is done by the model to ensure that the specified
input domain value is not excessive.
One subtle consequence of the use of the fraction=TRUE feature of the PWL
Controlled Source is that, in certain cases, you may inadvertently create extreme
smoothing of functions by choosing inappropriate coordinate value points. This can
be demonstrated by considering a function described by three coordinate pairs, such
as (-1,-1), (1,1), and (2,1). In this case, with a 10% input_domain value specified
(fraction=TRUE, input_domain=0.10), you would expect to see rounding occur
between in=0.9 and in=1.1, and nowhere else. On the other hand, if you were to
specify the same function using the coordinate pairs (-100,-100), (1,1) and (201,1),
you would find that rounding occurs between in=-19 and in=21. Clearly in the
latter case the smoothing might cause an excessive divergence from the intended
linearity above and below in=1.
a7 2 4 xfer_cntl1
.model xfer_cntl1 pwlts(x_array=[-2.0 -1.0 2.0 4.0 5.0]
+ y_array=[-0.2 -0.2 0.1 2.0 10.0]
12.2.8 PWL Time Controlled Source with optional edge smooth-

ing
NAME_TABLE:
C_Function_Name: cm_pwlts
Spice_Model_Name: pwlts
Description: "piecwise linear controlled source, time input"
PORT_TABLE:
Port_Name: out
Description: "output"
Direction: out
Default_Type: v
Allowed_Types: [v,vd,i,id]
Vector: no
Vector_Bounds: -
Null_Allowed: no
PARAMETER_TABLE:
Parameter_Name: x_array y_array
Description: "x-element array" "y-element array"
Default_Value: - -
Limits: - -
Vector: yes yes
Null_Allowed: no no
PARAMETER_TABLE:
Parameter_Name: input_domain fraction
Description: "input sm. domain" "smoothing %/abs switch"
Default_Value: 0.01 TRUE
Limits: [1e-12 0.5] -
Vector: no no
Vector_Bounds: - -
PARAMETER_TABLE:
Parameter_Name: limit
Description: "const or linearily extrapolated output"
Data_Type: boolean
Limits: -
Vector: no
Vector_Bounds: -
Null_Allowed: yes
STATIC_VAR_TABLE:
Static_Var_Name: last_x_value
Data_Type: pointer
Vector: no
STATIC_VAR_TABLE:
Static_Var_Name: x y
Data_Type: pointer pointer
Description: "time array" "y-coefficient array"
Description: The Piece-Wise Linear Time Controlled Source is a time input, single out-
put function. The output follows an time/output relationship specified by you via
the x_array and y_array coordinates. This is detailed below.
The x_array and y_array values represent vectors of coordinate points on the x and
y axes, respectively. The x_array values are progressively increasing positive input
coordinate points (minimum is 0), and the associated y_array values represent the
outputs at those points. There may be as few as two (x_array[n], y_array[n])
pairs specified, or as many as memory and simulation speed allow. This permits
you to very finely approximate a non-linear time dependent waveform by capturing
multiple input-output coordinate points.
Two aspects of the PWLTS Controlled Source warrant special attention. These are
the handling of endpoints and the smoothing of the described transfer function near
coordinate points.
In order to fully specify outputs for values of in outside of the bounds of the PWLTS
function (i.e., less than x_array[0] (with x_array[0] >= 0 always) or greater than
x_array[n], where n is the largest user-specified coordinate index), the PWLTS
Time Controlled Source model extends the slope found between the lowest two co-
ordinate pairs and the highest two coordinate pairs. This has the effect of making
the transfer function completely linear for times less than x_array[0] and times
greater than x_array[n]. It also has the potentially subtle effect of unrealistically
causing an output to reach a very large or small value for large input times.
In order to diminish the potential for non-convergence of simulations when using the
PWL block, a form of smoothing around the x_array, y_array coordinate points
is necessary. This is due to the iterative nature of the simulator and its reliance on
smooth first derivatives of transfer functions in order to arrive at a matrix solution.
Consequently, the input_domain and fraction parameters are included to allow
you some control over the amount and nature of the smoothing performed.
Fraction is a switch that is either TRUE or FALSE. When TRUE (the default
setting), the simulator assumes that the specified input domain value is to be in-
terpreted as a fractional figure. Otherwise, it is interpreted as an absolute value.
Thus, if fraction=TRUE and input_domain=0.10, the simulator assumes that the
smoothing radius about each coordinate point is to be set equal to 10% of the length
of either the x_array segment above each coordinate point, or the x_array segment
below each coordinate point. The specific segment length chosen will be the small-
est of these two for each coordinate point.
On the other hand, if fraction=FALSE and input_domain=0.10, then the simu-
lator will begin smoothing the transfer function at 0.10 seconds below each x_array
coordinate and will continue the smoothing process for another 0.10 seconds above
each x_array coordinate point. Since the overlap of smoothing domains is not al-
lowed, checking is done by the model to ensure that the specified input domain
value is not excessive.
One subtle consequence of the use of the fraction=TRUE feature of the PWL
Time Controlled Source is that, in certain cases, you may inadvertently create ex-
treme smoothing of functions by choosing inappropriate coordinate value points.
This can be demonstrated by considering a function described by three coordinate

pairs, such as (0,-1), (2,1), and (3,1). In this case, with a 10% input_domain value
specified (fraction=TRUE, input_domain=0.10), you would expect to see round-
ing occur between time=1.9 and time=2.1, and nowhere else. On the other hand,
if you were to specify the same function using the coordinate pairs (0,-100), (101,1)
and (301,1), you would find that rounding occurs between time=81 and time=121.
Clearly in the latter case the smoothing might cause an excessive divergence from
the intended linearity above and below time=101.

a8 out pwl_cntl1
.model pwl_cntl1 pwlts(x_array=[0 1m 1.1m 2m 2.1m]

+ y_array=[-0.2 -0.2 0.6 0.6 0.35]
+ input_domain=0.2 fraction=TRUE
+ limit=TRUE)
12.2.9 Filesource (PWL sourced from file)

NAME_TABLE:
C_Function_Name: cm_filesource
Spice_Model_Name: filesource
Description: "File Source"
PORT_TABLE:
Port_Name: out
Direction: out
Default_Type: v
Vector: yes
Vector_Bounds: [1 -]
Null_Allowed: no
PARAMETER_TABLE:
Parameter_Name: timeoffset timescale
Description: "time offset" "timescale"
Limits: - -
Vector: no no
Vector_Bounds: - -
PARAMETER_TABLE:
Parameter_Name: timerelative amplstep
Description: "relative time" "step amplitude"
Data_Type: boolean boolean
Default_Value: FALSE FALSE
Limits: - -
Vector: no no
Vector_Bounds: - -
PARAMETER_TABLE:
Parameter_Name: amploffset amplscale
Description: "ampl offset" "amplscale"
Default_Value: - -
Limits: - -
Vector: yes yes
PARAMETER_TABLE:
Parameter_Name: file
Description: "file name"
Data_Type: string
Default_Value: "filesource.txt"
Limits: -
Vector: no
Vector_Bounds: -
Null_Allowed: yes
Description: The File Source is similar to the Piece-Wise Linear (PWL) Source, except
that the waveform data is read from a file instead of being taken from parameter
vectors. The file format is line oriented ASCII. ‘#’ and ‘;’ are comment characters;
all characters from a comment character until the end of the line are ignored. Each
line consists of two or more real values. The first value is the time; subsequent
values correspond to the outputs. Values are separated by spaces. Time values are
absolute and must be monotonically increasing, unless timerelative is set to TRUE,
in which case the values specify the interval between two samples and must be
positive. Waveforms may be scaled and shifted in the time dimension by setting
timescale and timeoffset.
Amplitudes can also be scaled and shifted using amplscale and amploffset. Am-
plitudes are normally interpolated between two samples, unless amplstep is set to
TRUE.
Note: The file named by the parameter filename in file="filename" is sought after
according to a search list described in12.1.3.

a8 %vd([1 0 2 0]) filesrc
.
.
.model filesrc filesource (file="sine.m" amploffset=[0 0] amplscale=[1 1]
+ timeoffset=0 timescale=1
+ timerelative=false amplstep=false)
Example input file:

# name: sine.m
# two output ports
# column 1: time
# columns 2, 3: values
0 0 1
3.90625e-09 0.02454122852291229 0.9996988186962042
7.8125e-09 0.04906767432741801 0.9987954562051724
1.171875e-08 0.07356456359966743 0.9972904566786902
...
12.2.10 Multi_input_PWL_block
NAME_TABLE:
C_Function_Name: cm_multi_input_pwl
Spice_Model_Name: multi_input_pwl
Description: "multi_input_pwl block"
PORT_TABLE:
Port_Name: in out
Description: "input array" "output"
Direction: in out
Default_Type: vd vd
Allowed_Types: [vd,id] [vd,id]
Vector: yes no
Null_Allowed: no no
PARAMETER_TABLE:
Parameter_Name: x y
Description: "x array" "y array"
Limits: - -
Vector: yes yes
Null_Allowed: no no
PARAMETER_TABLE:
Parameter_Name: model
Description: "model type"
Data_Type: string
Default_Value: "and"
Limits: -
Vector: no
Vector_Bounds: -
Null_Allowed: yes
Description: Multi-input gate voltage controlled voltage source that supports and or
or gating. The x’s and y’s represent the piecewise linear variation of output (y)
as a function of input (x). The type of gate is selectable by the parameter model.
In case the model is and, the smallest input determines the output value (i.e. the
and function). In case the model is or, the largest input determines the output
value (i.e. the or function). The inverse of these functions (i.e. nand and nor) is
constructed by complementing the y array.

a82 [1 0 2 0 3 0] 7 0 pwlm
.
.
.model pwlm multi_input_pwl((x=[-2.0 -1.0 2.0 4.0 5.0]
+ y=[-0.2 -0.2 0.1 2.0 10.0]
+ model="and")
12.2.11 Analog Switch

NAME_TABLE:
C_Function_Name: cm_aswitch
Spice_Model_Name: aswitch
Description: "analog switch"
PORT_TABLE:
Port Name: cntl_in out
Description: "input" "resistive output"
Direction: in out
Default_Type: v gd
Allowed_Types: [v,vd,i,id] [gd]
Vector: no no
Vector_Bounds: - -
Null_Allowed: no no
PARAMETER_TABLE:
Parameter_Name: cntl_off cntl_on
Description: "control ‘off’ value" "control ‘on’ value"
Limits: - -
Vector: no no
Vector_Bounds: - -
PARAMETER_TABLE:
Parameter_Name: r_off log
Description: "off resistance" "log/linear switch"
Default_Value: 1.0e12 TRUE
Limits: - -
Vector: no no
Vector_Bounds: - -
PARAMETER_TABLE:
Parameter_Name: r_on limit
Description: "on resistance" "set upper and lower
limits to resistance"
Default_Value: 1.0 false
Limits: - -
Vector: no no
Vector_Bounds: - -
Description: The Analog Switch is a resistor that varies either logarithmically or lin-
early between specified values of a controlling input voltage or current. Note that
the input is not internally limited when parameter limit is not given. Therefore, if
the controlling signal exceeds the specified OFF state or ON state value, the resis-
tance may become excessively large or excessively small (in the case of logarithmic
dependence), or may become negative (in the case of linear dependence). For the
experienced user, these excursions may prove valuable for modeling certain devices,
but in most cases you are advised to add limiting of the controlling input if the
possibility of excessive control value variation exists. Alternatively you may set
the parameter limit to TRUE. Then the resulting resistance is limited to r_on or
r_off if the controlling voltage exceeds the given boundaries cntl_on or cntl_off.
At these boundaries sharp edges in the R(control) characteristics will occur which
may lead to convergence problems.

a8 3 %gd(6 7) switch3
.
.
.model switch3 aswitch(cntl_off=0.0 cntl_on=5.0 r_off=1e6
+ r_on=10.0 log=TRUE limit=TRUE)
12.2.12 Alternative Analog Switch

NAME_TABLE:
C_Function_Name: cm_pswitch
Spice_Model_Name: pswitch
Description: "analog switch alternative"
PORT_TABLE:
Description: "input" "resistive output"
Direction: inout inout
Default_Type: gd gd
Allowed_Types: [g,gd] [gd]
Vector: no no
Vector_Bounds: - -
Null_Allowed: no no
PARAMETER_TABLE:
Parameter_Name: cntl_off cntl_on
Description: "control ‘off’ value" "control ‘on’ value"

Limits: - -
Vector: no no
Vector_Bounds: - -
PARAMETER_TABLE:
Parameter_Name: r_off log
Description: "off resistance" "log/linear switch"
Default_Value: 1.0e12 TRUE
Limits: - -
Vector: no no
Vector_Bounds: - -
PARAMETER_TABLE:
Parameter_Name: r_on r_cntl_in
Description: "on resistance" "input resistance for control terminal"
Default_Value: 1.0 1e12
Limits: - -
Vector: no no
Vector_Bounds: - -
Description: The Alternative Analog Switch is a resistor that varies either logarithmi-
cally or linearly between specified values of a controlling input voltage or current.
An input resistance r_cntl_in may be specified. The output resistance is limited
to r_on or r_off. At the control boundaries cntl_on or cntl_off the R(control)
characteristics are slightly rounded. This behaviour is PSPICE-compatible and in-
stances of this device are generated when parsing PSPICE netlists in compatability
mode.

a9 %g 13 %gd(16 17) switch4
.
.
.model switch4 pswitch(cntl_off=0.0 cntl_on=5.0 r_off=1e6
+ r_on=10.0 r_cntl_in=1e11 log=TRUE)
12.2.13 Zener Diode

NAME_TABLE:
C_Function_Name: cm_zener
Spice_Model_Name: zener
Description: "zener diode"
PORT_TABLE:
Port Name: z
Description: "zener"
Direction: inout
Default_Type: gd
Allowed_Types: [gd]
Vector: no
Vector_Bounds: -
Null_Allowed: no
PARAMETER_TABLE:
Parameter_Name: v_breakdown i_breakdown
Description: "breakdown voltage" "breakdown current"
Default_Value: - 2.0e-2
Limits: [1.0e-6 1.0e6] [1.0e-9 -]
Vector: no no
Vector_Bounds: - -
Null_Allowed: no yes
PARAMETER_TABLE:
Parameter_Name: i_sat n_forward
Description: "saturation current" "forward emission coefficient"
Default_Value: 1.0e-12 1.0
Limits: [1.0e-15 -] [0.1 10]
Vector: no no
Vector_Bounds: - -
PARAMETER_TABLE:
Parameter_Name: limit_switch
Description: "switch for on-board limiting (convergence aid)"
Data_Type: boolean
Limits: -
Vector: no
Vector_Bounds: -
Null_Allowed: yes
STATIC_VAR_TABLE:
Static_Var_Name: previous_voltage
Data_Type: pointer
Description: The Zener Diode models the DC characteristics of most zeners. This model
differs from the Diode/Rectifier by providing a user-defined dynamic resistance in
the reverse breakdown region. The forward characteristic is defined by only a single
point, since most data sheets for zener diodes do not give detailed characteristics in
the forward region.
The first three parameters define the DC characteristics of the zener in the break-
down region and are usually explicitly given on the data sheet.
The saturation current refers to the relatively constant reverse current that is pro-
duced when the voltage across the zener is negative, but breakdown has not been
reached. The reverse leakage current determines the slight increase in reverse cur-
rent as the voltage across the zener becomes more negative. It is modeled as a
resistance parallel to the zener with value v breakdown / i rev.

Note that the limit switch parameter engages an internal limiting function for the
zener. This can, in some cases, prevent the simulator from converging to an unre-
alistic solution if the voltage across or current into the device is excessive. If use
of this feature fails to yield acceptable results, the convlimit option should be tried
(add the following statement to the SPICE input deck: .options convlimit)

a9 3 4 vref10
.
.
.model vref10 zener(v_breakdown=10.0 i_breakdown=0.02
+ r_breakdown=1.0 i_rev=1e-6 i_sat=1e-12)
12.2.14 Current Limiter

NAME_TABLE:
C_Function_Name: cm_ilimit
Spice_Model_Name: ilimit
Description: "current limiter block"
PORT_TABLE:
Port Name: in pos_pwr
Description: "input" "positive power supply"
Direction: in inout
Default_Type: v g
Allowed_Types: [v,vd] [g,gd]
Vector: no no
Vector_Bounds: - -
PORT_TABLE:
Port Name: neg_pwr out
Description: "negative power supply" "output"
Default_Type: g g
Allowed_Types: [g,gd] [g,gd]
Vector: no no
Vector_Bounds: - -
Null_Allowed: yes no
PARAMETER_TABLE:
Limits: - -
Vector: no no
Vector_Bounds: - -
PARAMETER_TABLE:
Parameter_Name: r_out_source r_out_sink

Description: "sourcing resistance" "sinking resistance"
Limits: [1.0e-9 1.0e9] [1.0e-9 1.0e9]
Vector: no no
Vector_Bounds: - -
PARAMETER_TABLE:
Parameter_Name: i_limit_source
Description: "current sourcing limit"
Data_Type: real
Default_Value: -
Limits: [1.0e-12 -]
Vector: no
Vector_Bounds: -
Null_Allowed: yes
PARAMETER_TABLE:
Parameter_Name: i_limit_sink
Description: "current sinking limit"
Data_Type: real
Default_Value: -
Limits: [1.0e-12 -]
Vector: no
Vector_Bounds: -
Null_Allowed: yes
PARAMETER_TABLE:
Parameter_Name: v_pwr_range i_source_range
Description: "upper & lower power "sourcing current
supply smoothing range" smoothing range"
Default_Value: 1.0e-6 1.0e-9
Limits: [1.0e-15 -] [1.0e-15 -]
Vector: no no
Vector_Bounds: - -
PARAMETER_TABLE:
Parameter_Name: i_sink_range
Description: "sinking current smoothing range"
Data_Type: real
Limits: [1.0e-15 -]
Vector: no
Vector_Bounds: -
Null_Allowed: yes
PARAMETER_TABLE:
Parameter_Name: r_out_domain
Description: "internal/external voltage delta smoothing range"
Data_Type: real
Limits: [1.0e-15 -]
Vector: no
Vector_Bounds: -
Null_Allowed: yes
Description: The Current Limiter models the behavior of an operational amplifier or

comparator device at a high level of abstraction. All of its pins act as inputs; three
of the four also act as outputs. The model takes as input a voltage value from the
in connector. It then applies an offset and a gain, and derives from it an equivalent
internal voltage (veq), which it limits to fall between pos_pwr and neg_pwr. If veq
is greater than the output voltage seen on the out connector, a sourcing current
will flow from the output pin. Conversely, if the voltage is less than vout, a sinking
current will flow into the output pin.
Depending on the polarity of the current flow, either a sourcing or a sinking re-
sistance value (r_out_source, r_out_sink) is applied to govern the vout/i_out
relationship. The chosen resistance will continue to control the output current until
it reaches a maximum value specified by either i_limit_source or i_limit_sink.
The latter mimics the current limiting behavior of many operational amplifier out-
put stages.
During all operation, the output current is reflected either in the pos_pwr connector
current or the neg_pwr current, depending on the polarity of i_out. Thus, realistic
power consumption as seen in the supply rails is included in the model.
The user-specified smoothing parameters relate to model operation as follows: v_pwr_range
controls the voltage below vpos_pwr and above vneg_pwr inputs beyond which
veq = gain (vin + vof f set ) is smoothed; i_source_range specifies the current be-
low i_limit_source at which smoothing begins, as well as specifying the cur-
rent increment above i_out=0.0 at which i_pos_pwr begins to transition to zero;
i_sink_range serves the same purpose with respect to i_limit_sink and i_neg_pwr
that i_source_range serves for i_limit_source and i_pos_pwr; r_out_domain
specifies the incremental value above and below (veq-vout)=0.0 at which r_out will
be set to r_out_source and r_out_sink, respectively. For values of (veq-vout)
less than r_out_domain and greater than -r_out_domain, r_out is interpolated
smoothly between r_out_source and r_out_sink.

a10 3 10 20 4 amp3
.
.
.model amp3 ilimit(in_offset=0.0 gain=16.0 r_out_source=1.0
+ r_out_sink=1.0 i_limit_source=1e-3
+ i_limit_sink=10e-3 v_pwr_range=0.2
+ i_source_range=1e-6 i_sink_range=1e-6
+ r_out_domain=1e-6)
12.2.15 Hysteresis Block

NAME_TABLE:
C_Function_Name: cm_hyst
Spice_Model_Name: hyst
Description: "hysteresis block"
PORT_TABLE:
Port Name: in out
Direction: in out
Default_Type: v v
Vector: no no
Vector_Bounds: - -
Null_Allowed: no no
PARAMETER_TABLE:
Parameter_Name: in_low in_high
Description: "input low value" "input high value"
Limits: - -
Vector: no no
Vector_Bounds: - -
PARAMETER_TABLE:
Parameter_Name: hyst out_lower_limit
Description: "hysteresis" "output lower limit"
Limits: [0.0 -] -
Vector: no no
Vector_Bounds: - -
PARAMETER_TABLE:
Parameter_Name: out_upper_limit input_domain
Description: "output upper limit" "input smoothing domain"
Limits: - -
Vector: no no
Vector_Bounds: - -
PARAMETER_TABLE:
Data_Type: boolean
Default_Value: TRUE
Limits: -
Vector: no
Vector_Bounds: -
Null_Allowed: yes
Description: The Hysteresis block is a simple buffer stage that provides hysteresis of the
output with respect to the input. The in low and in high parameter values specify
the center voltage or current inputs about which the hysteresis effect operates. The
output values are limited to out lower limit and out upper limit. The value of hyst
is added to the in low and in high points in order to specify the points at which
the slope of the hysteresis function would normally change abruptly as the input
transitions from a low to a high value. Likewise, the value of hyst is subtracted from
the in high and in low values in order to specify the points at which the slope of the
hysteresis function would normally change abruptly as the input transitions from
a high to a low value. In fact, the slope of the hysteresis function is never allowed
to change abruptly but is smoothly varied whenever the input domain smoothing
parameter is set greater than zero.

a11 1 2 schmitt1
.
.
.model schmitt1 hyst(in_low=0.7 in_high=2.4 hyst=0.5
+ out_lower_limit=0.5 out_upper_limit=3.0
12.2.16 Differentiator
NAME_TABLE:
C_Function_Name: cm_d_dt
Spice_Model_Name: d_dt
Description: "time-derivative block"
PORT_TABLE:
Port Name: in out
Direction: in out
Default_Type: v v
Vector: no no
Vector_Bounds: - -
Null_Allowed: no no
PARAMETER_TABLE:
Parameter_Name: gain out_offset
Description: "gain" "output offset"
Limits: - -
Vector: no no
Vector_Bounds: - -
PARAMETER_TABLE:

Default_Value: - -
Limits: - -
Vector: no no
Vector_Bounds: - -
PARAMETER_TABLE:
Description: "upper & lower limit smoothing range"
Data_Type: real
Limits: -
Vector: no
Vector_Bounds: -
Null_Allowed: yes
Description: The Differentiator block is a simple derivative stage that approximates

the time derivative of an input signal by calculating the incremental slope of that
signal since the previous time point. The block also includes gain and output offset
parameters to allow for tailoring of the required signal, and output upper and lower
limits to prevent convergence errors resulting from excessively large output values.
The incremental value of output below the output upper limit and above the output
lower limit at which smoothing begins is specified via the limit range parameter.
In AC analysis, the value returned is equal to the radian frequency of analysis
multiplied by the gain.
Note that since truncation error checking is not included in the d_dt block, it is not
recommended that the model be used to provide an integration function through
the use of a feedback loop. Such an arrangement could produce erroneous results.
Instead, you should make use of the "integrate" model, which does include truncation
error checking for enhanced accuracy.

a12 7 12 slope_gen
.
.
.model slope_gen d_dt(out_offset=0.0 gain=1.0
+ out_lower_limit=1e-12 out_upper_limit=1e12
+ limit_range=1e-9)
12.2.17 Integrator
NAME_TABLE:
C_Function_Name: cm_int
Spice_Model_Name: int
Description: "time-integration block"
PORT_TABLE:
Port Name: in out
Direction: in out
Default_Type: v v
Vector: no no
Vector_Bounds: - -
Null_Allowed: no no
PARAMETER_TABLE:
Limits: - -
Vector: no no
Vector_Bounds: - -
PARAMETER_TABLE:
Default_Value: - -
Limits: - -
Vector: no no
Vector_Bounds: - -
PARAMETER_TABLE:
Description: "upper & lower limit smoothing range"
Data_Type: real
Limits: -
Vector: no
Vector_Bounds: -
Null_Allowed: yes
PARAMETER_TABLE:
Parameter_Name: out_ic
Description: "output initial condition"
Data_Type: real
Default_Value: 0.0
Limits: -
Vector: no
Vector_Bounds: -
Null_Allowed: yes
Description: The Integrator block is a simple integration stage that approximates the
integral with respect to time of an input signal. The block also includes gain and
input offset parameters to allow for tailoring of the required signal, and output
upper and lower limits to prevent convergence errors resulting from excessively
large output values. Note that these limits specify integrator behavior similar to
that found in an operational amplifier-based integration stage, in that once a limit

is reached, additional storage does not occur. Thus, the input of a negative value to
an integrator that is currently driving at the out upper limit level will immediately
cause a drop in the output, regardless of how long the integrator was previously
summing positive inputs. The incremental value of output below the output upper
limit and above the output lower limit at which smoothing begins is specified via
the limit range parameter. In AC analysis, the value returned is equal to the gain
divided by the radian frequency of analysis.
Note that truncation error checking is included in the int block. This should provide
for a more accurate simulation of the time integration function, since the model will
inherently request smaller time increments between simulation points if truncation
errors would otherwise be excessive.

a13 7 12 time_count
.model time_count int(in_offset=0.0 gain=1.0
+ out_lower_limit=-1e12 out_upper_limit=1e12
+ limit_range=1e-9 out_ic=0.0)
12.2.18 S-Domain Transfer Function

NAME_TABLE:
C_Function_Name: cm_s_xfer
Spice_Model_Name: s_xfer
Description: "s-domain transfer function"
PORT_TABLE:
Port Name: in out
Direction: in out
Default_Type: v v
Vector: no no
Vector_Bounds: - -
Null_Allowed: no no
PARAMETER_TABLE:
Limits: - -
Vector: no no
Vector_Bounds: - -
PARAMETER_TABLE:
Parameter_Name: num_coeff
Description: "numerator polynomial coefficients"
Data_Type: real
Default_Value: -
Limits: -
Vector: yes
Null_Allowed: no
PARAMETER_TABLE:
Parameter_Name: den_coeff
Description: "denominator polynomial coefficients"
Data_Type: real
Default_Value: -
Limits: -
Vector: yes
Null_Allowed: no
PARAMETER_TABLE:
Parameter_Name: int_ic
Description: "integrator stage initial conditions"
Data_Type: real
Default_Value: 0.0
Limits: -
Vector: yes
Vector_Bounds: den_coeff
Null_Allowed: yes
PARAMETER_TABLE:
Parameter_Name: denormalized_freq
Description: "denorm. corner freq.(radians) for 1 rad/s coeffs"
Data_Type: real
Default_Value: 1.0
Limits: -
Vector: no
Vector_Bounds: -
Null_Allowed: yes
Description: The s-domain transfer function is a single input, single output transfer
function in the Laplace transform variable ‘s’ that allows for flexible modulation of
the frequency domain characteristics of a signal. Ac and transient simulations are
supported. The code model may be configured to produce an arbitrary s-domain
transfer function with the following restrictions:
1. The degree of the numerator polynomial cannot exceed that

of the denominator polynomial in the variable "s".
2. The coefficients for a polynomial must be stated
explicitly. That is, if a coefficient is zero, it must be
included as an input to the num coeff or den coeff vector.
The order of the coefficient parameters is from that associated with the highest-powered
term decreasing to that of the lowest. Thus, for the coefficient parameters specified below,
the equation in ‘s’ is shown:
.model filter s_xfer(gain=0.139713

+ num_coeff=[1.0 0.0 0.7464102]
+ den_coeff=[1.0 0.998942 0.001170077]
+ int_ic=[0 0])
It specifies a transfer function of the form

s2 +0.7464102
N (s) = 0.139713 · s2 +0.998942s+0.00117077
The s-domain transfer function includes gain and in_offset (input offset) parameters to
allow for tailoring of the required signal. There are no limits on the internal signal values
or on the output value of the s-domain transfer function, so you are cautioned to specify
gain and coefficient values that will not cause the model to produce excessively large
values. In AC analysis, the value returned is equal to the real and imaginary components
of the total s-domain transfer function at each frequency of interest.
The denormalized_freq term allows you to specify coefficients for a normalized filter
(i.e. one in which the frequency of interest is 1 rad/s). Once these coefficients are included,
specifying the denormalized frequency value ‘shifts’ the corner frequency to the actual one
of interest. As an example, the following transfer function describes a Chebyshev low-pass
filter with a corner (pass-band) frequency of 1 rad/s:
N (s) = 0.139713 · 1.0

s2 +1.09773s+1.10251
In order to define an s_xfer model for the above, but with the corner frequency equal to
1500 rad/s (239 Hz), the following instance and model lines would be needed:
a12 node1 node2 cheby1

.model cheby1 s_xfer(num_coeff=[1] den_coeff=[1 1.09773 1.10251]
+ int_ic=[0 0] denormalized_freq=1500)
In the above, you add the normalized coefficients and scale the filter through the use of
the denormalized freq parameter. Similar results could have been achieved by performing
the denormalization prior to specification of the coefficients, and setting denormalized
freq to the value 1.0 (or not specifying the frequency, as the default is 1.0 rad/s) Note in
the above that frequencies are always specified as radians/second.
Truncation error checking is included in the s-domain transfer block. This should pro-
vide for more accurate simulations, since the model will inherently request smaller time
increments between simulation points if truncation errors would otherwise be excessive.
The int_ic parameter is an array that must be of size one less as the array of values
specified for the den_coeff parameter. Even if a 0 start value is required, you have to
add the specific int_ic vector to the set of coefficients (see the examples above and below).

a14 9 22 cheby_LP_3kHz
.
.
.model cheby_LP_3kHz s_xfer(in_offset=0.0 gain=1.0 int_ic=[0 0]
+ num_coeff=[1.0]
+ den_coeff=[1.0 1.42562 1.51620])
12.2.19 PWL Transfer Function

NAME_TABLE:
Spice_Model_Name: xfer
C_Function_Name: cm_xfer
Description: "AC transfer function block"
PORT_TABLE:
Port_Name: in out
Direction: in out
Default_Type: v v
Vector: no no
Vector_Bounds: - -
Null_Allowed: no no
PARAMETER_TABLE:
Parameter_Name: table
Description: "PWL table: frequency/magnitude/phase"
Data_Type: real
Default_Value: 0
Limits: -
Vector: yes
Null_Allowed: yes
PARAMETER_TABLE:
Description: "File in Touchstone format"
Data_Type: string
Default_Value: -
Limits: -
Vector: no
Vector_Bounds: -
Null_Allowed: yes
PARAMETER_TABLE:
Parameter_Name: r_i
Description: "table is in real/imaginary format"
Data_Type: boolean
Limits: -
Vector: no
Vector_Bounds: -
Null_Allowed: yes
PARAMETER_TABLE:
Parameter_Name: db
Description: "table is in magnitude(dB)/phase format"
Data_Type: boolean
Default_Value: true
Limits: -
Vector: no
Vector_Bounds: -
Null_Allowed: yes
PARAMETER_TABLE:
Parameter_Name: rad
Description: "phase in radians, not degrees"
Data_Type: boolean
Limits: -
Vector: no
Vector_Bounds: -
Null_Allowed: yes
PARAMETER_TABLE:
Parameter_Name: span offset
Description: "Length of table rows" "Offset within row"
Data_Type: int int
Default_Value: 3 1
Limits: [ 3 - ] [ 1 - ]
Vector: no no
Vector_Bounds: - -
/* This is used internally to store the table in compact complex form. */
STATIC_VAR_TABLE:
Static_Var_Name: table
Description: "Internal copy of data"
Data_Type: pointer
/* Only warn once about use in transient analysis. */
STATIC_VAR_TABLE:
Static_Var_Name: warned
Description: "Warning indicator"
Data_Type: int
This code model is useful only in AC analysis, where it applies a complex transfer function
to its input. The current circuit frequncy is input to a PWL function defined by a table
and the output is produced by multiplying the input by the resulting complex number.
The parameters supply the PWL table and determine its format. The “table” parameter
supplies the data directly, while “file” defines a path (which must be all lower-case) to a
file in Touchstone format containing the data. Exactly one of those parameters must be
specified.
The data is treated as consisting of rows, each of “span” real numbers. The first number
is the frequency of a PWL corner and a pair of numbers at the “offset” position in the row
supply the data. That allows a single Touchstone file to be shared by several instances
of this code model, as such files for an n-port device will contain logical rows of 2*n^2+1
numbers: one frequency value and the components of an NxN complex matrix. The
format of the data pairs is determined by the “db”, “rad” and “r_i” parameters. If any of
these are set, they override the internal indicators in a Touchstone file which themselves
override the parameter defaults.
Examples of using this model are in the examples/sp directory: netlist file.sp shows direct
use, while filter.sp uses the E-source wrapper (5.2.6).
12.2.20 Slew Rate Block

NAME_TABLE:
C_Function_Name: cm_slew
Spice_Model_Name: slew
Description: "A simple slew rate follower block"
PORT_TABLE:
Port Name: in out
Direction: in out
Default_Type: v v
Vector: no no
Vector_Bounds: - -
Null_Allowed: no no
PARAMETER_TABLE:
Parameter_Name: rise_slope
Description: "maximum rising slope value"
Data_Type: real
Default_Value: 1.0e9
Limits: -
Vector: no
Vector_Bounds: -
Null_Allowed: yes
PARAMETER_TABLE:
Parameter_Name: fall_slope
Description: "maximum falling slope value"
Data_Type: real
Default_Value: 1.0e9
Limits: -
Vector: no
Vector_Bounds: -
Null_Allowed: yes
PARAMETER_TABLE:
Parameter_Name: range
Description: "smoothing range"
Data_Type: real
Default_Value: 0.1
Limits: -
Vector: no
Vector_Bounds: -
Null_Allowed: yes
Description: This function is a simple slew rate block that limits the absolute slope
of the output with respect to time to some maximum or value. The actual slew
rate effects of over-driving an amplifier circuit can thus be accurately modeled by
cascading the amplifier with this model. The units used to describe the maximum
rising and falling slope values are expressed in volts or amperes per second. Thus a
desired slew rate of 0.5 V/µs will be expressed as 0.5e+6, etc.
The slew rate block will continue to raise or lower its output until the difference
between the input and the output values is zero. Thereafter, it will resume following
the input signal, unless the slope again exceeds its rise or fall slope limits. The range
input specifies a smoothing region above or below the input value. Whenever the
model is slewing and the output comes to within the input + or - the range value,
the partial derivative of the output with respect to the input will begin to smoothly
transition from 0.0 to 1.0. When the model is no longer slewing (output = input),
dout/din will equal 1.0.

a15 1 2 slew1
.model slew1 slew(rise_slope=0.5e6 fall_slope=0.5e6)
12.2.21 Inductive Coupling

NAME_TABLE:
C_Function_Name: cm_lcouple
Spice_Model_Name: lcouple
Description: "inductive coupling (for use with ’core’ model)"
PORT_TABLE:
Port_Name: l mmf_out
Description: "inductor" "mmf output (in ampere-turns)"
Default_Type: hd hd
Allowed_Types: [h,hd] [hd]
Vector: no no
Vector_Bounds: - -
Null_Allowed: no no
PARAMETER_TABLE:
Parameter_Name: num_turns
Description: "number of inductor turns"
Data_Type: real
Default_Value: 1.0
Limits: -
Vector: no
Vector_Bounds: -
Null_Allowed: yes
Description: This function is a conceptual model that is used as a building block to

create a wide variety of inductive and magnetic circuit models. This function is
normally used in conjunction with the core model, but can also be used with
resistors, hysteresis blocks, etc. to build up systems that mock the behavior of
linear and nonlinear components.
The lcouple takes as an input (on the ‘l’ port), a current. This current value is
multiplied by the num_turns value, N, to produce an output value (a voltage value
that appears on the mmf_out port). The mmf_out acts similar to a magnetomotive
force in a magnetic circuit; when the lcouple is connected to the core model, or
to some other resistive device, a current will flow. This current value (which is
modulated by whatever the lcouple is connected to) is then used by the lcouple to
calculate a voltage ‘seen’ at the l port. The voltage is a function of the derivative
with respect to time of the current value seen at mmf_out.
The most common use for lcouples will be as a building block in the construction
of transformer models. To create a transformer with a single input and a single
output, you would require two lcouple models plus one core model. The process
of building up such a transformer is described under the description of the core
model, below.

a150 (7 0) (9 10) lcouple1
.model lcouple1 lcouple(num_turns=10.0)
12.2.22 Magnetic Core

NAME_TABLE:
C_Function_Name: cm_core
Spice_Model_Name: core
Description: "magnetic core"
PORT_TABLE:
Port_Name: mc
Description: "magnetic core"
Direction: inout
Default_Type: gd
Allowed_Types: [g,gd]
Vector: no
Vector_Bounds: -
Null_Allowed: no
PARAMETER_TABLE:
Parameter_Name: H_array B_array
Description: "magnetic field array" "flux density array"
Default_Value: - -
Limits: - -
Vector: yes yes
Null_Allowed: no no
PARAMETER_TABLE:
Parameter_Name: area length
Description: "cross-sectional area" "core length"
Default_Value: - -
Limits: - -
Vector: no no
Vector_Bounds: - -
Null_Allowed: no no
PARAMETER_TABLE:
Parameter_Name: input_domain
Description: "input sm. domain"

Data_Type: real
Default_Value: 0.01
Limits: [1e-12 0.5]
Vector: no
Vector_Bounds: -
Null_Allowed: yes
PARAMETER_TABLE:
Description: "smoothing fraction/abs switch"
Data_Type: boolean
Default_Value: TRUE
Limits: -
Vector: no
Vector_Bounds: -
Null_Allowed: yes
PARAMETER_TABLE:
Parameter_Name: mode
Description: "mode switch (1 = pwl, 2 = hyst)"
Data_Type: int
Default_Value: 1
Limits: [1 2]
Vector: no
Vector_Bounds: -
Null_Allowed: yes
PARAMETER_TABLE:
Parameter_Name: in_low in_high
Description: "input low value" "input high value"
Limits: - -
Vector: no no
Vector_Bounds: - -
PARAMETER_TABLE:
Parameter_Name: hyst out_lower_limit
Description: "hysteresis" "output lower limit"
Limits: [0 -] -
Vector: no no
Vector_Bounds: - -
PARAMETER_TABLE:
Parameter_Name: out_upper_limit
Description: "output upper limit"
Data_Type: real
Default_Value: 1.0
Limits: -
Vector: no
Vector_Bounds: -
Null_Allowed: yes
Description: This function is a conceptual model that is used as a building block to

create a wide variety of inductive and magnetic circuit models. This function is
almost always expected to be used in conjunction with the lcouple model to build
up systems that mock the behavior of linear and nonlinear magnetic components.
There are two fundamental modes of operation for the core model. These are the
pwl mode (which is the default, and which is the most likely to be of use to you)
and the hysteresis mode. These are detailed below.
PWL Mode (mode = 1)
The core model in PWL mode takes as input a voltage that it treats as a magnetomotive
force (mmf) value. This value is divided by the total effective length of the core to
produce a value for the Magnetic Field Intensity, H. This value of H is then used to find
the corresponding Flux Density, B, using the piecewise linear relationship described by
you in the H array / B array coordinate pairs. B is then multiplied by the cross-sectional
area of the core to find the Flux value, which is output as a current. The pertinent
mathematical equations are listed below:
mmf
H= , where L = Length
L
Here H, the Magnetic Field Intensity, is expressed in ampere-turns/meter.
B = f (H)
The B value is derived from a piecewise linear transfer function described to the model
via the (H_array[],B_array[]) parameter coordinate pairs. This transfer function does
not include hysteretic effects; for that, you would need to substitute a HYST model for
the core.
ϕ = BA, where A = Area
The final current allowed to flow through the core is equal to ϕ. This value in turn is
used by the "lcouple" code model to obtain a value for the voltage reflected back across
its terminals to the driving electrical circuit.
The following example code shows the use of two lcouple models and one core model to
produce a simple primary/secondary transformer.

a1 (2 0) (3 0) primary
.model primary lcouple (num_turns = 155)
a2 (3 4) iron_core
.model iron_core core (H_array = [-1000 -500 -375 -250 -188 -125 -63 0
+ 63 125 188 250 375 500 1000]
+ B_array = [-3.13e-3 -2.63e-3 -2.33e-3 -1.93e-3
+ -1.5e-3 -6.25e-4 -2.5e-4 0 2.5e-4
+ 6.25e-4 1.5e-3 1.93e-3 2.33e-3
+ 2.63e-3 3.13e-3]
+ area = 0.01 length = 0.01)
a3 (5 0) (4 0) secondary
.model secondary lcouple (num_turns = 310)
HYSTERESIS Mode (mode = 2)
The core model in HYSTERESIS mode takes as input a voltage that it treats as a magne-
tomotive force (mmf) value. This value is used as input to the equivalent of a hysteresis
code model block. The parameters defining the input low and high values, the output low
and high values, and the amount of hysteresis are as in that model. The output from this
mode, as in PWL mode, is a current value that is seen across the mc port. An example
of the core model used in this fashion is shown below:

a1 (2 0) (3 0) primary
.model primary lcouple (num_turns = 155)
a2 (3 4) iron_core
.model iron_core core (mode = 2 in_low=-7.0 in_high=7.0
+ out_lower_limit=-2.5e-4 out_upper_limit=2.5e-4
+ hyst = 2.3 )
a3 (5 0) (4 0) secondary
.model secondary lcouple (num_turns = 310)
One final note to be made about the two core model nodes is that certain parameters
are available in one mode, but not in the other. In particular, the in_low, in_high,
out_lower_limit, out_upper_limit, and hysteresis parameters are not available in PWL
mode. Likewise, the H_array, B_array, area, and length values are unavailable in HYS-
TERESIS mode. The input domain and fraction parameters are common to both modes
(though their behavior is somewhat different; for explanation of the input domain and
fraction values for the HYSTERESIS mode, you should refer to the hysteresis code model
discussion).
12.2.23 Controlled Sine Wave Oscillator

NAME_TABLE:
C_Function_Name: cm_sine
Spice_Model_Name: sine
Description: "controlled sine wave oscillator"

PORT_TABLE:
Description: "control input" "output"
Direction: in out
Default_Type: v v
Vector: no no
Vector_Bounds: - -
Null_Allowed: no no
PARAMETER_TABLE:
Parameter_Name: cntl_array freq_array
Description: "control array" "frequency array"
Default_Value: 0.0 1.0e3
Limits: - [0 -]
Vector: yes yes
Vector_Bounds: [2 -] cntl_array
Null_Allowed: no no
PARAMETER_TABLE:
Parameter_Name: out_low out_high
Description: "output peak low value" "output peak high value"
Default_Value: -1.0 1.0
Limits: - -
Vector: no no
Vector_Bounds: - -
Description: This function is a controlled sine wave oscillator with parametrizable values
of low and high peak output. It takes an input voltage or current value. This value
is used as the independent variable in the piecewise linear curve described by the
coordinate points of the cntl array and freq array pairs. From the curve, a frequency
value is determined, and the oscillator will output a sine wave at that frequency.
From the above, it is easy to see that array sizes of 2 for both the cntl array and the
freq array will yield a linear variation of the frequency with respect to the control
input. Any sizes greater than 2 will yield a piecewise linear transfer characteristic.
For more detail, refer to the description of the piecewise linear controlled source,
which uses a similar method to derive an output value given a control input.

asine 1 2 in_sine
.model in_sine sine(cntl_array = [-1 0 5 6]
+ freq_array=[10 10 1000 1000] out_low = -5.0
+ out_high = 5.0)
12.2.24 Controlled Triangle Wave Oscillator

NAME_TABLE:
C_Function_Name: cm_triangle
Spice_Model_Name: triangle
Description: "controlled triangle wave oscillator"
PORT_TABLE:
Direction: in out
Default_Type: v v
Vector: no no
Vector_Bounds: - -
Null_Allowed: no no
PARAMETER_TABLE:
Limits: - [0 -]
Vector: yes yes
Null_Allowed: no no
PARAMETER_TABLE:
Limits: - -
Vector: no no
Vector_Bounds: - -
PARAMETER_TABLE:
Parameter_Name: duty_cycle
Description: "rise time duty cycle"
Data_Type: real
Default_Value: 0.5
Limits: [1e-10 0.999999999]
Vector: no
Vector_Bounds: -
Null_Allowed: yes
Description: This function is a controlled triangle/ramp wave oscillator with parametriz-

able values of low and high peak output and rise time duty cycle. It takes an input
voltage or current value. This value is used as the independent variable in the
piecewise linear curve described by the coordinate points of the cntl_array and
freq_array pairs.
From the curve, a frequency value is determined, and the oscillator will output a
triangle wave at that frequency. From the above, it is easy to see that array sizes
of 2 for both the cntl_array and the freq_array will yield a linear variation of the
frequency with respect to the control input. Any sizes greater than 2 will yield
a piecewise linear transfer characteristic. For more detail, refer to the description
of the piecewise linear controlled source, which uses a similar method to derive an
output value given a control input.

ain 1 2 ramp1
.model ramp1 triangle(cntl_array = [-1 0 5 6]
+ freq_array=[10 10 1000 1000] out_low = -5.0
+ out_high = 5.0 duty_cycle = 0.9)
12.2.25 Controlled Square Wave Oscillator

NAME_TABLE:
C_Function_Name: cm_square
Spice_Model_Name: square
Description: "controlled square wave oscillator"
PORT_TABLE:
Direction: in out
Default_Type: v v
Vector: no no
Vector_Bounds: - -
Null_Allowed: no no
PARAMETER_TABLE:
Limits: - [0 -]
Vector: yes yes
Null_Allowed: no no
PARAMETER_TABLE:
Limits: - -
Vector: no no
Vector_Bounds: - -
PARAMETER.TABLE:
Parameter_Name: duty_cycle rise_time
Description: "duty cycle" "output rise time"
Default_Value: 0.5 1.0e-9

Limits: [1e-6 0.999999] -
Vector: no
Vector_Bounds: - -
PARAMETER_TABLE:
Parameter_Name: fall_time
Description: "output fall time"
Data_Type: real
Limits: -
Vector: no
Vector_Bounds: -
Null_Allowed: yes
Description: This function is a controlled square wave oscillator with parametrizable

values of low and high peak output, duty cycle, rise time, and fall time. It takes
an input voltage or current value. This value is used as the independent variable
in the piecewise linear curve described by the coordinate points of the cntl_array
and freq_array pairs. From the curve, a frequency value is determined, and the
oscillator will output a square wave at that frequency.
From the above, it is easy to see that array sizes of 2 for both the cntl_array and the
freq_array will yield a linear variation of the frequency with respect to the control

ain 1 2 pulse1
.model pulse1 square(cntl_array = [-1 0 5 6]
+ freq_array=[10 10 1000 1000] out_low = 0.0
+ out_high = 4.5 duty_cycle = 0.2
+ rise_time = 1e-6 fall_time = 2e-6)
12.2.26 Controlled One-Shot

NAME_TABLE:
C_Function_Name: cm_oneshot
Spice_Model_Name: oneshot
Description: "controlled one-shot"
PORT_TABLE:
Port Name: clk cntl_in
Description: "clock input" "control input"
Direction: in in
Default_Type: v v
Vector: no no
Vector_Bounds: - -
PORT_TABLE:
Port Name: clear out
Description: "clear signal" "output"
Direction: in out
Default_Type: v v
Vector: no no
Vector_Bounds: - -
PARAMETER_TABLE:
Parameter_Name: clk_trig retrig
Description: "clock trigger value" "retrigger switch"
Default_Value: 0.5 FALSE
Limits: - -
Vector: no no
Vector_Bounds: - -
PARAMETER_TABLE:
Parameter_Name: pos_edge_trig
Description: "positive/negative edge trigger switch"
Data_Type: boolean
Default_Value: TRUE
Limits: -
Vector: no
Vector_Bounds: -
Null_Allowed: no
PARAMETER_TABLE:
Parameter_Name: cntl_array pw_array
Description: "control array" "pulse width array"
Limits: - [0.00 -]
Vector: yes yes
Vector_Bounds: - cntl_array
PARAMETER_TABLE:
Description: "output low value" "output high value"
Limits: - -
Vector: no no
Vector_Bounds: - -
PARAMETER_TABLE:
Parameter_Name: fall_time rise_time
Description: "output fall time" "output rise time"

Limits: - -
Vector: no no
Vector_Bounds: - -
PARAMETER_TABLE:
Parameter_Name: rise_delay
Description: "output delay from trigger"
Data_Type: real
Limits: -
Vector: no
Vector_Bounds: -
Null_Allowed: yes
PARAMETER_TABLE:
Parameter_Name: fall_delay
Description: "output delay from pw"
Data_Type: real
Limits: -
Vector: no
Vector_Bounds: -
Null_Allowed: yes
Description: This function is a controlled oneshot with parametrizable values of low and
high peak output, input trigger value level, delay, and output rise and fall times.
It takes an input voltage or current value. This value is used as the independent
variable in the piecewise linear curve described by the coordinate points of the
cntl_array and pw_array pairs. From the curve, a pulse width value is determined.
The one-shot will output a pulse of that width, triggered by the clock signal (rising
or falling edge), delayed by the delay value, and with specified rise and fall times. A
positive slope on the clear input will immediately terminate the pulse, which resets
with its fall time.
From the above, it is easy to see that array sizes of 2 for both the cntl_array and the
pw_array will yield a linear variation of the pulse width with respect to the control

ain 1 2 3 4 pulse2
.model pulse2 oneshot(cntl_array = [-1 0 10 11]
+ pw_array=[1e-6 1e-6 1e-4 1e-4]
+ clk_trig = 0.9 pos_edge_trig = FALSE
+ out_low = 0.0 out_high = 4.5
+ rise_delay = 20.0e-9 fall_delay = 35.0e-9)
12.2.27 Capacitance Meter

NAME_TABLE:
C_Function_Name: cm_cmeter
Spice_Model_Name: cmeter
Description: "capacitance meter"
PORT_TABLE:
Port Name: in out
Direction: in out
Default_Type: v v
Vector: no no
Vector_Bounds: - -
Null_Allowed: no no
PARAMETER_TABLE:
Parameter_Name: gain
Description: "gain"
Data_Type: real
Default_Value: 1.0
Limits: -
Vector: no
Vector_Bounds: -
Null_Allowed: yes
Description: The capacitance meter is a sensing device that is attached to a circuit

node and produces as an output a scaled value equal to the total capacitance seen
on its input multiplied by the gain parameter. This model is primarily intended
as a building block for other models that must sense a capacitance value and alter
their behavior based upon it.

atest1 1 2 ctest
.model ctest cmeter(gain=1.0e12)
12.2.28 Inductance Meter

NAME_TABLE:
C_Function_Name: cm_lmeter
Spice_Model_Name: lmeter
Description: "inductance meter"
PORT_TABLE:
Port Name: in out
Direction: in out
Default_Type: v v
Vector: no no
Vector_Bounds: - -
Null_Allowed: no no
PARAMETER_TABLE:
Parameter_Name: gain
Description: "gain"
Data_Type: real
Default_Value: 1.0
Limits: -
Vector: no
Vector_Bounds: -
Null_Allowed: yes
Description: The inductance meter is a sensing device that is attached to a circuit node
and produces as an output a scaled value equal to the total inductance seen on
its input multiplied by the gain parameter. This model is primarily intended as a
building block for other models that must sense an inductance value and alter their
behavior based upon it.

atest2 1 2 ltest
.model ltest lmeter(gain=1.0e6)
12.2.29 Memristor
NAME_TABLE:
C_Function_Name: cm_memristor
Spice_Model_Name: memristor
Description: "Memristor Interface"
PORT_TABLE:
Port_Name: memris
Description: "memristor terminals"
Direction: inout
Default_Type: gd
Allowed_Types: [gd]
Vector: no
Vector_Bounds: -
Null_Allowed: no
PARAMETER_TABLE:
Parameter_Name: rmin rmax
Description: "minimum resistance" "maximum resistance"
Limits: - -
Vector: no no
Vector_Bounds: - -
Null_Allowed: no no
PARAMETER_TABLE:
Parameter_Name: rinit vt
Description: "initial resistance" "threshold"

Limits: - -
Vector: no no
Vector_Bounds: - -
Null_Allowed: no no
PARAMETER_TABLE:
Parameter_Name: alpha beta
Description: "model parameter 1" "model parameter 2"
Limits: - -
Vector: no no
Vector_Bounds: - -
Null_Allowed: no no
Description: The memristor is a two-terminal resistor with memory, whose resistance

depends on the time integral of the voltage across its terminals. rmin and rmax
provide the lower and upper limits of the resistance, rinit is its starting value (no
voltage applied so far). The voltage has to be above a threshold vt to become
effective in changing the resistance. alpha and beta are two model parameters. The
memristor code model is derived from a SPICE subcircuit published in [23].

amen 1 2 memr
.model memr memristor (rmin=1k rmax=10k rinit=7k
+ alpha=0 beta=2e13 vt=1.6)
12.2.30 2D table model

NAME_TABLE:
C_Function_Name: cm_table2D
Spice_Model_Name: table2D
Description: "2D table model"
PORT_TABLE:
Port_Name: inx iny out
Description: "inputx" "inputy" "output"
Default_Type: v v i
Allowed_Types: [v,vd,i,id,vnam] [v,vd,i,id,vnam] [v,vd,i,id]
Vector: no no no
PARAMETER_TABLE:
Parameter_Name: order verbose
Description: "order" "verbose"
Data_Type: int int
Default_Value: 3 0
Limits: - -
Vector: no no
Vector_Bounds: - -
PARAMETER_TABLE:
Parameter_Name: offset gain
Description: "offset" "gain"
Limits: - -
Vector: no no
Vector_Bounds: - -
PARAMETER_TABLE:
Data_Type: string
Default_Value: "2D-table-model.txt"
Limits: -
Vector: no
Vector_Bounds: -
Null_Allowed: yes
Description: The 2D table model reads a matrix from file "file name" (default 2D-
table-model.txt) which has x columns and y rows. Each x,y pair, addressed by
inx and iny, yields an output value out. Linear interpolation is used for out, eno
(essentially non oscillating) interpolation for its derivatives. Parameters offset (de-
fault 0) and gain (default 1) modify the output table values according to of f set +
gain out. Parameter order (default 3) influences the calculation of the derivatives.
Parameter verbose (default 0) yields test outputs, if set to 1 or 2. The table format
is shown below. Be careful to include the data point inx = 0, iny = 0 into your
table, because ngspice uses these during .OP computations. The x horizontal and y
vertical address values have to increase monotonically.
Table Example:
* table source
* number of columns (x)
8
* number of rows (y)
9
* x horizontal (column) address values (real numbers)
-1 0 1 2 3 4 5 6
* y vertical (row) address values (real numbers)
-0.6 0 0.6 1.2 1.8 2.4 3.0 3.6 4.2
* table with output data (horizontally addressed by x, vertically by y)
1 0.9 0.8 0.7 0.6 0.5 0.4 0.3
1 1 1 1 1 1 1 1
1 1.2 1.4 1.6 1.8 2 2.2 2.4

1 1.5 2 2.5 3 3.5 4 4.5
1 2 3 4 5 6 7 8
1 2.5 4 5.5 7 8.5 10 11.5
1 3 5 7 9 11 13 15
1 3.5 6 8.5 11 13.5 16 18.5
1 4 7 10 13 16 19 22
Description: The usage example consists of two input voltages referenced to ground and
a current source output with two floating nodes.

atab inx iny %id(out1 out2) tabmod
.model tabmod table2d (offset=0.0 gain=1 order=3 file="table-simple.txt")
12.2.31 3D table model

NAME_TABLE:
C_Function_Name: cm_table3D
Spice_Model_Name: table3D
Description: "3D table model"
PORT_TABLE:
Port_Name: inx iny inz
Description: "inputx" "inputy" "inputz"
Direction: in in in
Default_Type: v v v
Allowed_Types: [v,vd,i,id,vnam] [v,vd,i,id,vnam] [v,vd,i,id,vnam]
Vector: no no no
PORT_TABLE:
Port_Name: out
Direction: out
Default_Type: i
Vector: no
Vector_Bounds: -
Null_Allowed: no
PARAMETER_TABLE:
Parameter_Name: order verbose
Description: "order" "verbose"
Data_Type: int int
Default_Value: 3 0
Limits: - -
Vector: no no
Vector_Bounds: - -
PARAMETER_TABLE:
Parameter_Name: offset gain
Description: "offset" "gain"
Limits: - -
Vector: no no
Vector_Bounds: - -
PARAMETER_TABLE:
Data_Type: string
Default_Value: "3D-table-model.txt"
Limits: -
Vector: no
Vector_Bounds: -
Null_Allowed: yes
Description: The 3D table model reads a matrix from file "file name" (default 3D-
table-model.txt) which has x columns, y rows per table and z tables. Each x,y,z
triple, addressed by inx, iny, and inz, yields an output value out. Linear interpola-
tion is used for out, eno (essentially non oscillating) interpolation for its derivatives.
Parameters offset (default 0) and gain (default 1) modify the output table values
according to of f set + gain out. Parameter order (default 3) influences the calcu-
lation of the derivatives. Parameter verbose (default 0) yields test outputs, if set
to 1 or 2. The table format is shown below. Be careful to include the data point
inx = 0, iny = 0, inz = 0 into your table, because ngspice needs these to for the .OP
calculation. The x horizontal, y vertical, and z table address values have to increase
monotonically.
Table Example:
* 3D table for nmos bsim 4, W=10um, L=0.13um
*x
39
*y
39
*z
11
*x (drain voltage)
-0.1 -0.05 0 0.05 0.1 0.15 0.2 0.25 ...
*y (gate voltage)
-0.1 -0.05 0 0.05 0.1 0.15 0.2 0.25 ...
*z (substrate voltage)
-1.8 -1.6 -1.4 -1.2 -1 -0.8 -0.6 -0.4 -0.2 0 0.2
*table -1.8
-4.50688E-10 -4.50613E-10 -4.50601E-10 -4.50599E-10 ...
-4.49622E-10 -4.49267E-10 -4.4921E-10 -4.49202E-10 ...
-4.50672E-10 -4.49099E-10 -4.48838E-10 -4.48795E-10 ...

-4.55575E-10 -4.4953E-10 -4.48435E-10 -4.48217E-10 ...
...
*table -1.6
-3.10015E-10 -3.09767E-10 -3.0973E-10 -3.09724E-10 ...
-3.09748E-10 -3.08524E-10 -3.08339E-10 -3.08312E-10 ...
...
*table -1.4
-2.04848E-10 -2.04008E-10 -2.03882E-10 ...
-2.07275E-10 -2.03117E-10 -2.02491E-10 ...
...
Description: The usage example simulates a NMOS transistor with independent drain,
gate and bulk nodes, referenced to source. Parameter gain may be used to emulate
transistor width, with respect to the table transistor.

amos1 %vd(d s) %vd(g s) %vd(b s) %id(d s) mostable1
.model mostable1 table3d (offset=0.0 gain=0.5 order=3
+ verbose=1 file="table-3D-bsim4n.txt")
12.2.32 Simple Diode Model

NAME_TABLE:
C_Function_Name: cm_sidiode
Spice_Model_Name: sidiode
Description: "simple diode"
PORT_TABLE:
Port_Name: ds
Description: "diode port"
Direction: inout
Default_Type: gd
Allowed_Types: [gd]
Vector: no
Vector_Bounds: -
Null_Allowed: no
PARAMETER_TABLE:
Parameter_Name: ron roff
Description: "resistance on-state" "resistance off-state"
Default_Value: 1 11
Limits: [1e-6 - ] [1e-12 -]
Vector: no no
Vector_Bounds: - -
Null_Allowed: no no
PARAMETER_TABLE:
1
If roff is not given, ron is the default
Parameter_Name: vfwd vrev

Description: "forward voltage" "reverse breakdown voltage"
Default_Value: 0. 1e30
Limits: [0. -] [0. -]
Vector: no no
Vector Bounds: - -
PARAMETER_TABLE:
Parameter_Name: ilimit revilimit
Description: "limit of on-current" "limit of breakdown current"
Default_Value: 1e30 1e30
Limits: [1e-15 -] [1e-15 -]
Vector: no no
Vector_Bounds: - -
PARAMETER_TABLE:
Parameter_Name: epsilon revepsilon
Description: "width quadrat. reg. 1" "width quadratic region 2"
Default_Value: 0. 0.
Limits: [0. -] [0. -]
Vector: no no
Vector_Bounds: - -
PARAMETER_TABLE:
Parameter_Name: rrev
Description: "resistance in breakdown"
Data_Type: real
Default_Value: 0.
Limits: -
Vector: no
Vector_Bounds: -
Null_Allowed: yes
This is a model for a simple diode. Three regions are modelled as linear I(V) curves:
Reverse (breakdown) current with Rrev starting at Vrev into the negative direction, Off
current with Roff between Vrev and Vfwd and an On region with Ron, staring at Vfwd.
The interface between the regions is described by a quadratic function, the width of the
interface region is determined by Revepsilon and Epsilon. Current limits in the reverse
breakdown (Revilimit) and in the forward (on) state (Ilimit) may be set. The interface is
a tanh function. Thus the first derivative of the I(V) curve is continuous. All parameter
values are entered as positive numbers. A diode capacitance is not modelled.

a1 a k ds1
.model ds1 sidiode(Roff=1000 Ron=0.7 Rrev=0.2 Vfwd=1
+ Vrev=10 Revepsilon=0.2 Epsilon=0.2 Ilimit=7 Revilimit=7)
12.2.33 Analog delay

NAME_TABLE:
C_Function_Name: cm_delay
Spice_Model_Name: delay
Description: "analog delay line"
PORT_TABLE:
Port_Name: in out cntrl
Description: "input" "output" "control"
Direction: in out in
Default_Type: v v v
Allowed_Types: [v,vd,vnam] [v,vd] [v,vd,i,id]
Vector: no no no
Null_Allowed: no no yes
PARAMETER_TABLE:
Parameter_Name: delay buffer_size
Description: "time delay" "size of delay buffer"
Data_Type: real int
Default_Value: 0.0 1024
Limits: - [1 -]
Vector: no no
Vector_Bounds: - -
PARAMETER_TABLE:
Parameter_Name: has_delay_cnt
Description: "controlled delay"
Data_Type: boolean
Limits: -
Vector: no
Vector_Bounds: -
Null_Allowed: yes
PARAMETER_TABLE:
Parameter_Name: delmin delmax
Description: "min delay" "max delay"
Default_Value: 0 0
Limits: [0 -] [0 -]
Vector: no no
Vector_Bounds: - -
Description: During a transient simulation the input voltage at node in and its
associated time value are stored in a ring buffer. buffer_size allows to set the
size of the buffer, the default is 1024 time steps. There are two modes to read out
the buffer contents with a delay and obtain the delayed values at port out,
determined by has_delay_cnt. If has_delay_cnt is TRUE, then you may vary the
delay time between delmin and delmax by a control voltage between 0 and 1 at
the input terminal cntrl. Parameter delay is ignored. If has_delay_cnt has

been set to FALSE, then the signal is delayed by the time value given by delay .

adelay1 in out cntrl mydel1
.model mydel1 delay(delay=2m buffer_size=2048)
adelay2 in out cntrl mydel2
.model mydel2 delay(has_delay_cnt=TRUE delmin=5u delmax=8u)
Due to the fact that time steps are not constant during a transient simulation, but op-
timized by the simulator, the delayed values are sometimes slightly deviating from the
original, depending on the number of steps. So in a sinusoidal wave we will see a distortion
< 0.3% for 1000 steps per sin cycle.
12.2.34 Potentiometer
NAME_TABLE:
Spice_Model_Name: potentiometer
C_Function_Name: cm_potentiometer
Description: "potentiometer"
PORT_TABLE:
Port_Name: r0 wiper
Description: "pot connection 0" "wiper contact"
Default_Type: g g
Allowed_Types: [g] [g]
Vector: no no
Vector_Bounds: - -
Null_Allowed: no no
PORT_TABLE:
Port_Name: r1
Description: "pot connection 1"
Direction: inout
Default_Type: g
Allowed_Types: [g]
Vector: no
Vector_Bounds: -
Null_Allowed: no
PARAMETER_TABLE:
Parameter_Name: position
Description: "position of wiper connection (0.0 to 1.0)"
Data_Type: real
Default_Value: 0.5
Limits: [0.0 1.0]
Vector: no
Vector_Bounds: -
Null_Allowed: yes
PARAMETER_TABLE:
Parameter_Name: log r
Description: "log-linear switch" "total resistance"
Data_Type: boolean real
Default_Value: FALSE 1.0e5
Limits: - -
Vector: no no
Vector_Bounds: - -
PARAMETER_TABLE:
Parameter_Name: log_multiplier
Description: "multiplier constant for log resistance"
Data_Type: real
Default_Value: 1.0
Limits: -
Vector: no
Vector_Bounds: -
Null_Allowed: yes
Description: A resistance potentiometer with three connections: r0, wiper , and r1.
Parameter position determines the lower and upper portions of the resistance.
Rlower is located between r0 and wiper, Rupper between wiper and r1. If log is
set to FALSE, Rlower = position ∗ r. If log is set to TRUE, then
Rlower = r ∗ 10−position∗log_multiplier . For Rupper we always have
Rupper = r − Rlower. position <= 0 is resolved to position = 1e − 9,
position >= 1 is resolved to position = 0.999999999.

Apot r0 w r1 potmod
.model potmod potentiometer(position=0.45 r=1k log=FALSE log_multiplier=1)
12.3 Hybrid Models

The following hybrid models are supplied with XSPICE. The descriptions included below
consist of the model Interface Specification File and a description of the model’s operation.
This is followed by an example of a simulator-deck placement of the model, including the
.MODEL card and the specification of all available parameters.
A note should be made with respect to the use of hybrid models for other than simple
digital-to-analog and analog-to-digital translations. The hybrid models represented in
this section address that specific need, but in the development of user-defined nodes you
may find a need to translate not only between digital and analog nodes, but also between
real and digital, real and int, etc. In most cases such translations will not need to be as
involved or as detailed as shown in the following.
12.3.1 Digital-to-Analog Node Bridge

NAME_TABLE:
12.3. HYBRID MODELS 239
C_Function_Name: cm_dac_bridge
Spice_Model_Name: dac_bridge
Description: "digital-to-analog node bridge"
PORT_TABLE:
Port Name: in out
Direction: in out
Default_Type: d v
Allowed_Types: [d] [v,vd,i,id,d]
Vector: yes yes
Vector_Bounds: - -
Null_Allowed: no no
PARAMETER_TABLE:
Parameter_Name: out_low
Description: "0-valued analog output"
Data_Type: real
Default_Value: 0.0
Limits: -
Vector: no
Vector_Bounds: -
Null_Allowed: yes
PARAMETER_TABLE:
Parameter_Name: out_high
Description: "1-valued analog output"
Data_Type: real
Default_Value: 1.0
Limits: -
Vector: no
Vector_Bounds: -
Null_Allowed: yes
PARAMETER_TABLE:
Parameter_Name: out_undef input_load
Description: "U-valued analog output" "input load (F)"
Limits: - -
Vector: no no
Vector_Bounds: - -
PARAMETER_TABLE:
Parameter_Name: t_rise t_fall
Description: "rise time 0->1" "fall time 1->0"
Limits: - -
Vector: no no
Vector_Bounds: - -
Description: The dac_bridge is the first of three node bridge devices designed to allow
for the ready transfer of digital information to analog values and back again. The
second device is the adc_bridge (which takes an analog value and maps it to a dig-
ital one).The dac_bridge takes as input a digital value from a digital node. This
value by definition may take on only one of the values ‘0’, ‘1’ or ‘U’. The dac_bridge
then outputs the value out_low, out_high or out_undef, or ramps linearly toward
one of these ‘final’ values from its current analog output level. The speed at which
this ramping occurs depends on the values of t_rise and t_fall. These parame-
ters are interpreted by the model such that the rise or fall slope generated is always
constant. Note that the dac_bridge includes test code in its cfunc.mod file for de-
termining the presence of the out_undef parameter. If this parameter is not specified
by you, and if out_high and out_low values are specified, then out_undef is as-
signed the value of the arithmetic mean of out_high and out_low. This simplifies
coding of output buffers, where typically a logic family will include an out_low and
out_high voltage, but not an out_undef value. This model also posts an input
load value (in farads) based on the parameter input load.

abridge1 [7] [2] dac1
.model dac1 dac_bridge(out_low = 0.7 out_high = 3.5 out_undef = 2.2
+ input_load = 5.0e-12 t_rise = 50e-9
+ t_fall = 20e-9)
12.3.2 Analog-to-Digital Node Bridge

NAME_TABLE:
C_Function_Name: cm_adc_bridge
Spice_Model_Name: adc_bridge
Description: "analog-to-digital node bridge"
PORT_TABLE:
Port Name: in out
Direction: in out
Default_Type: v d
Allowed_Types: [v,vd,i,id,d] [d]
Vector: yes yes
Vector_Bounds: - -
Null_Allowed: no no
PARAMETER_TABLE:
Parameter_Name: in_low
Description: "maximum 0-valued analog input"
Data_Type: real
Default_Value: 1.0
Limits: -
Vector: no
Vector_Bounds: -
Null_Allowed: yes
PARAMETER_TABLE:
Parameter_Name: in_high
Description: "minimum 1-valued analog input"
Data_Type: real
Default_Value: 2.0
Limits: -
Vector: no
Vector_Bounds: -
Null_Allowed: yes
PARAMETER_TABLE:
Parameter_Name: rise_delay fall_delay
Description: "rise delay" "fall delay"
Limits: [1.0e-12 -] [1.0e-12 -]
Vector: no no
Vector_Bounds: - -
Description: The adc_bridge is one of three node bridge devices designed to allow for
the ready transfer of analog information to digital values and back again. The
second device is the dac_bridge (which takes a digital value and maps it to an
analog one). The adc_bridge takes as input an analog value from an analog node.
This value by definition may be in the form of a voltage, or a current. If the input
value is less than or equal to in_low, then a digital output value of ‘0’ is generated.
If the input is greater than or equal to in_high, a digital output value of ‘1’ is
generated. If neither of these is true, then a digital ‘UNKNOWN’ value is output.
Note that unlike the case of the dac_bridge, no ramping time or delay is associated
with the adc_bridge. Rather, the continuous ramping of the input value provides
for any associated delays in the digitized signal.

abridge2 [1] [8] adc_buff
.model adc_buff adc_bridge(in_low = 0.3 in_high = 3.5)
12.3.3 Bidirectional Analog/Digital Node Bridge

NAME_TABLE:
C_Function_Name: cm_bidi_bridge
Spice_Model_Name: bidi_bridge
Description: "bidirectional digital/analog node bridge"
PORT_TABLE:
Port_Name: a d
Description: "analog" "digital in/out"
Default_Type: g d
Allowed_Types: [g, gd] [d]
Vector: yes yes
Null_Allowed: no no
/* The direction of the bridge ports may be controlled by digital inputs.
* with LOW selecting DAC behavior and HIGH selecting ADC.
* If null, or the value is UNKNOWN the bridge will be truly bi-directional.
*/
PORT_TABLE:
Port_Name: dir
Description: "direction"
Direction: in
Default_Type: d
Allowed_Types: [d]
Vector: yes
Vector_Bounds: -
Null_Allowed: yes
/* Alternatively, this parameter sets direction: 0-2 for DAC, ADC, ignore.
* Values 0/1 override the direction port.
*/
PARAMETER_TABLE:
Parameter_Name: direction input_load
Description: "force direction" "capacitive input load (F)"
Data_Type: int real
Default_Value: 2 1.0e-12
Limits: [0 2] -
Vector: no no
Vector_Bounds: - -
/* Digital 0utput strength is 0 (strong, default) or 1 (resistive).
* Smooth controls use of smoothing functions, default is 0 (no smoothing).
*/
PARAMETER_TABLE:
Parameter_Name: strength smooth
Description: "output strength" "smoothing level"
Data_Type: int int
Default_Value: 0 0
Limits: [0 2] [0 2]
Vector: no no
Vector_Bounds: - -
/* Analog thresholds, in_low may be greater than in-high, enabling hysteresis.
*/
PARAMETER_TABLE:
Parameter_Name: in_low
Description: "maximum 0-valued analog input"
Data_Type: real
Default_Value: 0.1
Limits: -
Vector: no
Vector_Bounds: -
Null_Allowed: yes
PARAMETER_TABLE:
Parameter_Name: in_high
Description: "minimum 1-valued analog input"
Data_Type: real
Default_Value: 0.9
Limits: -
Vector: no
Vector_Bounds: -
Null_Allowed: yes
PARAMETER_TABLE:
/* Analog maximum and minimum output voltages. */
Parameter_Name: out_low
Description: "minimum analog output voltage for ’ZERO’ digital input"
Data_Type: real
Default_Value: 0.0
Limits: -
Vector: no
Vector_Bounds: -
Null_Allowed: yes
PARAMETER_TABLE:
Parameter_Name: out_high
Description: "maximum analog output voltage for ’ONE’ digital input"
Data_Type: real
Default_Value: 3.3
Limits: -
Vector: no
Vector_Bounds: -
Null_Allowed: yes
/* Analog maximum current. */
PARAMETER_TABLE:
Parameter_Name: drive_low drive_high
Description: "max current to ground" "max current from supply"
Limits: - -
Vector: no no
Vector_Bounds: - -
/* Strong analog output cuts off smoothly at the voltage limits.
* Let vth = out_high - r_sth * drive_high.
* Then for input voltage v, with drive_high > v > vth,
* the maximum output current is (drive_high - v) / r_sth
*/
PARAMETER_TABLE:
Parameter_Name: r_stl r_sth
Description: "low taper resistance" "high taper resistance"
Default_Value: 20 20
Limits: [1e-6 -] [1e-6 -]
Vector: no no
Vector_Bounds: - -
/* Resistive analog drive. */
PARAMETER_TABLE:
Parameter_Name: r_low r_high
Description: "drive resistor to ground" "drive resistor to out_high"
Default_Value: 10000 10000
Limits: [1e-6 -] [1e-6 -]
Vector: no no
Vector_Bounds: - -
/* Analog rise and fall times. */
PARAMETER_TABLE:
Parameter_Name: t_rise t_fall
Description: "rise time 0 -> 1" "fall time 1 -> 0"
Limits: [1e-12 -] [1e-12 -]
Vector: no no
Vector_Bounds: - -
/* Digital rise and fall delays. */
PARAMETER_TABLE:
Description: "rise delay 0 -> 1" "fall delay 1 -> 0"
Limits: [1e-12 -] [1e-12 -]
Vector: no no
Vector_Bounds: - -
Description: The bidi_bridge is the third and most complex of three analog/digital
node bridges. It is capable of effectively simultaneous output to both analog and
digital ports, depending on the state of the other side. That requires the use of an
analog transconductance port, which may cause convergence problems when there
is high impedance on a connected analog node. Non-zero values for the smooth
parameter may be helpful if such problems occur. For digital nodes that are al-
ways strongly driven but also have digital inputs, the simpler dac_bridge may be
preferred. Otherwise, bidi_bridge has some additional features that may make it
preferable to the other bridges. The analog output characteristics change with the
digital drive strength, with strong output behaving similarly to a very crude model
of a CMOS output driver. The low input threshold may be higher than the high
threshold, producing Schmitt Trigger behavior.

abridge2 [1 2 3] [8 9 10] null bidi_buff
.model bidi_buff bidi_bridge(in_low = 2 in_high = 1.5)
12.3.4 Controlled Digital Oscillator

NAME_TABLE:
C_Function_Name: cm_d_osc
Spice_Model_Name: d_osc
Description: "controlled digital oscillator"
PORT_TABLE:
Direction: in out
Default_Type: v d
Allowed_Types: [v,vd,i,id] [d]
Vector: no no
Vector_Bounds: - -
Null_Allowed: no no
PARAMETER_TABLE:
Limits: - [0 -]
Vector: yes yes
Null_Allowed: no no
PARAMETER_TABLE:
Parameter_Name: duty_cycle init_phase
Description: "duty cycle" "initial phase of output"
Default_Value: 0.5 0
Limits: [1e-6 0.999999] [-180.0 +360.0]
Vector: no no
Vector_Bounds: - -
PARAMETER_TABLE:
Default_Value: 1e-9 1e-9
Limits: [0 -] [0 -]
Vector: no no
Vector_Bounds: - -
Description: The digital oscillator is a hybrid model that accepts as input a voltage or
current. This input is compared to the voltage-to-frequency transfer characteristic

specified by the cntl_array/freq_array coordinate pairs, and a frequency is ob-
tained that represents a linear interpolation or extrapolation based on those pairs.
A digital time-varying signal is then produced with this fundamental frequency.
The output waveform, which is the equivalent of a digital clock signal, has rise and
fall delays that can be specified independently. In addition, the duty cycle and the
phase of the waveform are also variable and can be set by you.

a5 1 8 var_clock
.model var_clock d_osc(cntl_array = [-2 -1 1 2]
+ freq_array = [1e3 1e3 10e3 10e3]
+ duty_cycle = 0.4 init_phase = 180.0
+ rise_delay = 10e-9 fall_delay=8e-9)
12.3.5 Node bridge from digital to real with enable

NAME_TABLE:
Spice_Model_Name: d_to_real
C_Function_Name: ucm_d_to_real
Description: "Node bridge from digital to real with enable"
PORT_TABLE:
Port_Name: in enable out
Description: "input" "enable" "output"
Default_Type: d d real
Allowed_Types: [d] [d] [real]
Vector: no no no
Null_Allowed: no yes no
PARAMETER_TABLE:
Parameter_Name: zero one delay
Description: "value for 0" "value for 1" "delay"
Default_Value: 0.0 1.0 1e-9
Limits: - - [1e-15 -]
Vector: no no no
12.3.6 A Z**-1 block working on real data

NAME_TABLE:
Spice_Model_Name: real_delay
C_Function_Name: ucm_real_delay
Description: "A Z ** -1 block working on real data"
PORT_TABLE:
Port_Name: in clk out
Description: "input" "clock" "output"

Default_Type: real d real
Allowed_Types: [real] [d] [real]
Vector: no no no
PARAMETER_TABLE:
Parameter_Name: delay
Description: "delay from clk to out"
Data_Type: real
Default_Value: 1e-9
Limits: [1e-15 -]
Vector: no
Vector_Bounds: -
Null_Allowed: yes
12.3.7 A gain block for event-driven real data

NAME_TABLE:
Spice_Model_Name: real_gain
C_Function_Name: ucm_real_gain
Description: "A gain block for event-driven real data"
PORT_TABLE:
Port_Name: in out
Direction: in out
Default_Type: real real
Allowed_Types: [real] [real]
Vector: no no
Vector_Bounds: - -
Null_Allowed: no no
PARAMETER_TABLE:
Parameter_Name: in_offset gain out_offset
Description: "input offset" "gain" "output offset"
Default_Value: 0.0 1.0 0.0
Limits: - - -
Vector: no no no
PARAMETER_TABLE:
Parameter_Name: delay ic
Description: "delay" "initial condition"
Limits: - -
Vector: no no
Vector_Bounds: - -
12.3.8 Node bridge from real to analog voltage

NAME_TABLE:
Spice_Model_Name: real_to_v
C_Function_Name: ucm_real_to_v
Description: "Node bridge from real to analog voltage"
PORT_TABLE:
Port_Name: in out
Direction: in out
Default_Type: real v
Allowed_Types: [real] [v, vd, i, id]
Vector: no no
Vector_Bounds: - -
Null_Allowed: no no
PARAMETER_TABLE:
Parameter_Name: gain transition_time
Description: "gain" "output transition time"
Default_Value: 1.0 1e-9
Limits: - [1e-15 -]
Vector: no no
Vector_Bounds: - -
12.3.9 Controlled PWM Oscillator

NAME_TABLE:
Spice_Model_Name: d_pwm
C_Function_Name: cm_d_pwm
Description: "duty cycle controlled digital oscillator"
PORT_TABLE:
Port_Name: cntl_in out
Direction: in out
Default_Type: v d
Allowed_Types: [v,vd,i,id] [d]
Vector: no no
Vector_Bounds: - -
Null_Allowed: no no
PARAMETER_TABLE:
Parameter_Name: cntl_array dc_array
Description: "control array" "duty cycle array"
Limits: - [0 1]
Vector: yes yes
Null_Allowed: no no
PARAMETER_TABLE:
Parameter_Name: frequency init_phase
Description: "oscillator frequency" "initial phase of output"
Default_Value: 1e6 0
Limits: [1e-6 -] [-180.0 +360.0]
Vector: no no
Vector_Bounds: - -
PARAMETER_TABLE:
Default_Value: 1e-9 1e-9
Limits: [0 -] [0 -]
Vector: no no
Vector_Bounds: - -
Description: The digital pulse-width modulated oscillator is a hybrid model that ac-
cepts as control input a voltage or current. This input is compared to the voltage-
to-duty cycle transfer characteristic specified by the cntl_array/dc_array coor-
dinate pairs, and a duty cycle is obtained that represents a linear interpolation or
extrapolation based on those pairs. A digital duty cycle-varying signal is then pro-
duced. The duty cycle is limited between 0 and 1 (excluding the limits).
The digital output waveform has rise and fall delays that can be specified inde-
pendently. In addition, the oscillator frequency and the phase of the waveform are
variable and user selectable.

a5 cin dout pwm_osc
.model pwm_osc d_pwm(cntl_array = [-2 -1.99 1.99 2]
+ dc_array = [0.01 0.01 0.99 0.99]
+ frequency = 1.2Meg init_phase = 90.0
+ rise_delay = 10e-9 fall_delay=8e-9)
Currently there are some limits or bugs: a duty cycle < 1% or larger than 99% may
generate false output (e.g. a 50% duty cycle signal). Sometimes spurious missing pulses
occur. To obtain false results by extrapolation during evaluation of the cntl_array, it is
recommended to force a flat output if input signals are above or below the outer limits of
the cntl_array data (see the example shown above).
12.4 Digital Models

The following digital models are supplied with XSPICE. The descriptions included below
consist of a (sometimes abbreviated) model Interface Specification File and a description
of the model’s operation. This is followed by an example of a simulator-deck placement
of the model, including the .MODEL card and the specification of all available parameters.
Note that these models have not been finalized at this time.
Some information common to all digital models and/or digital nodes is included here.
The following are general rules that should make working with digital nodes and models
more straightforward:
1. All digital nodes are initialized to ZERO at the start of a simulation (i.e., when
INIT=TRUE). This means that a model need not post an explicit value to an
output node upon initialization if its output would normally be a ZERO (although
posting such would certainly cause no harm).
2. Digital nodes may have one out of twelve possible node values. See 12.5.1 for details.
3. Digital models typically have defined their rise and fall delays for their output
signals. A capacitive input load value may be defined as well to determine a load-
dependent delay, but is currently not used in any code model (see 28.7.1.5).
4. Several commands are available for outputting data, e.g. eprint, edisplay, and
eprvcd. Digital inputs may be read from files. Please see Chapt. 12.5.4 for more
details.
5. Hybrid models (see Chapt. 12.3) provide an interface between the digital event
driven world and the analog world of ngspice to enable true mixed mode simulation.
There are some common parameters that are used by many of the digital models. To
avoid repetition they are omitted from the individual Interface Description Files listed
here and their availabilty is noted at the end of the file for each model. The common
parameters are:
inertial_delay When this boolean parameter is set, output pulses that are shorter
than the current delay time for the port are suppressed, and the output remains
unchanged until the next state transition that completes its delay period. The de-
fault value is "false", giving transport delay behavior: all changes reach the output.
An interpreter variable, digital_delay_type, can be used to override the default.
A value of 1 changes the default to "true"; 2 forces all relevant XSPICE elements to
use transport delay; 3 forces inertial delay.
This parameter is used by PSpice-compatible U-devices (14). In ngspice-40 these

XSPICE digital devices:
d_and, d_buffer, d_inverter, d_nand, d_nor, d_or, d_tristate, d_xnor, d_xor
have the inertial_delay parameter. When the circuits in the examples/digital/dig-

ital_devices directory are run, subcircuits with PSpice U* device instances are
12.4. DIGITAL MODELS 251
translated to XSPICE primitives. Also, .model statements are generated containing

inertial_delay=true. This causes the circuits to run with inertial delays (suppress
glitches) rather than transport delays (propagate glitches). Most digital simulators
model gates using inertial delays.
If you run the examples in ngspice-39 and ngspice-40 then compare waveforms
produced by circuits behav-tristate-pulse.cir and behav-283.cir, you will see how
glitches are suppressed by the inertial delay mechanism. To obtain transport delay
behavior with ngspice-40, add the following line:
set digital_delay_type=2
to the .spiceinit file in that directory.

family This is a string-valued parameter that has no effect on the model itself, but labels
the ports of instances of the model to guide the automatic bridging mechanism. See
12.6.
rise_delay The delay time between a change in a model’s internal state, as driven by
its inputs, and a change in output to digital one. This is used when there is only
one output, or they all have the same delays.
fall_delay Like rise_delay, but for transitions to zero.
input_load The capacitance of one or all digital inputs, in Farads. Code models may use
the TOTAL_LOAD macro to find the capacitative load on their outputs. However,
the outputs of models listed here do not respond to their loading. These models
always drive outputs strongly with the specified delays.
These common parameters appear in individual Interface Specification Files in these

forms:
PARAMETER_TABLE:
Limits: [1.0e-12 -] [1.0e-12 -]
Vector: no no
Vector_Bounds: - -
PARAMETER_TABLE:
Parameter_Name: input_load
Description: "input load value (F)"
Data_Type: real
Limits: -
Vector: no
Vector_Bounds: -
Null_Allowed: yes
PARAMETER_TABLE:
Parameter_Name: inertial_delay family
Description: "swallow short pulses" "Logic family for bridging"
Data_Type: boolean string
Default_Value: false -
Limits: - -
Vector: no no
Vector_Bounds: - -
12.4.1 Buffer
NAME_TABLE:
C_Function_Name: cm_d_buffer
Spice_Model_Name: d_buffer
Description: "digital one-bit-wide buffer"
PORT_TABLE:
Port Name: in out
Direction: in out
Default_Type: d d
Allowed_Types: [d] [d]
Vector: no no
Vector_Bounds: - -
Null_Allowed: no no
Common parameters: inertial_delay, family, rise_delay, fall_delay, input_load.
Description: The buffer is a single-input, single-output digital buffer that produces as

output a time-delayed copy of its input.

a6 1 8 buff1
.model buff1 d_buffer(rise_delay = 0.5e-9 fall_delay = 0.3e-9
+ input_load = 0.5e-12)
12.4.2 Inverter
NAME_TABLE:
C_Function_Name: cm_d_inverter
Spice_Model_Name: d_inverter
Description: "digital one-bit-wide inverter"
PORT_TABLE:
Port Name: in out
Direction: in out
Default_Type: d d

Vector: no no
Vector_Bounds: - -
Null_Allowed: no no
Description: The inverter is a single-input, single-output digital inverter that produces

as output an inverted, time-delayed copy of its input.

a6 1 8 inv1
.model inv1 d_inverter(rise_delay = 0.5e-9 fall_delay = 0.3e-9
12.4.3 And
NAME_TABLE:
C_Function_Name: cm_d_and
Spice_Model_Name: d_and
Description: "digital ‘and’ gate"
PORT_TABLE:
Port Name: in out
Direction: in out
Default_Type: d d
Vector: yes no
Null_Allowed: no no
Description: The digital and gate is an n-input, single-output and gate that produces
an active ‘1’ value if, and only if, all of its inputs are also ‘1’ values. If ANY of the
inputs is a ‘0’, the output will also be a ‘0’; if neither of these conditions holds, the
output will be unknown.

a6 [1 2] 8 and1
.model and1 d_and(rise_delay = 0.5e-9 fall_delay = 0.3e-9
12.4.4 Nand
NAME_TABLE:
C_Function_Name: cm_d_nand
Spice_Model_Name: d_nand
Description: "digital ‘nand’ gate"
PORT_TABLE:
Port Name: in out
Direction: in out
Default_Type: d d
Vector: yes no
Null_Allowed: no no
Description: The digital nand gate is an n-input, single-output nand gate that produces
an active ‘0’ value if and only if all of its inputs are ‘1’ values. If ANY of the inputs
is a ‘0’, the output will be a ‘1’; if neither of these conditions holds, the output will
be unknown.

a6 [1 2 3] 8 nand1
.model nand1 d_nand(rise_delay = 0.5e-9 fall_delay = 0.3e-9
12.4.5 Or
NAME_TABLE:
C_Function_Name: cm_d_or
Spice_Model_Name: d_or
Description: "digital ‘or’ gate"
PORT_TABLE:
Port Name: in out
Direction: in out
Default_Type: d d
Vector: yes no
Null_Allowed: no no
Description: The digital or gate is an n-input, single-output or gate that produces an

active ‘1’ value if at least one of its inputs is a ‘1’ value. The gate produces a ‘0’
value if all inputs are ‘0’; if neither of these two conditions holds, the output is
unknown.

a6 [1 2 3] 8 or1
.model or1 d_or(rise_delay = 0.5e-9 fall_delay = 0.3e-9
12.4.6 Nor
NAME_TABLE:
C_Function_Name: cm_d_nor
Spice_Model_Name: d_nor
Description: "digital ‘nor’ gate"
PORT_TABLE:
Port Name: in out
Direction: in out
Default_Type: d d
Vector: yes no
Null_Allowed: no no
Description: The digital nor gate is an n-input, single-output nor gate that produces
an active ‘0’ value if at least one of its inputs is a ‘1’ value. The gate produces a
‘0’ value if all inputs are ‘0’; if neither of these two conditions holds, the output is
unknown.

anor12 [1 2 3 4] 8 nor12
.model nor12 d_nor(rise_delay = 0.5e-9 fall_delay = 0.3e-9
12.4.7 Xor
NAME_TABLE:
C_Function_Name: cm_d_xor
Spice_Model_Name: d_xor
Description: "digital exclusive-or gate"
PORT_TABLE:
Port Name: in out
Direction: in out
Default_Type: d d
Vector: yes no
Null_Allowed: no no
Description: The digital xor gate is an n-input, single-output xor gate that produces
an active ‘1’ value if an odd number of its inputs are also ‘1’ values. The delays
associated with an output rise and those associated with an output fall may be
specified independently. Note also that to maintain the technology-independence of
the model, any UNKNOWN input, or any floating input causes the output to also
go UNKNOWN.

a9 [1 2] 8 xor3
.model xor3 d_xor(rise_delay = 0.5e-9 fall_delay = 0.3e-9
12.4.8 Xnor
NAME_TABLE:
C_Function_Name: cm_d_xnor
Spice_Model_Name: d_xnor
Description: "digital exclusive-nor gate"
PORT_TABLE:
Port Name: in out
Direction: in out
Default_Type: d d
Vector: yes no
Null_Allowed: no no
Description: The digital xnor gate is an n-input, single-output xnor gate that produces
an active ‘0’ value if an odd number of its inputs are also ‘1’ values. It produces a
‘1’ output when an even number of ‘1’ values occurs on its inputs. Note also that
to maintain the technology-independence of the model, any UNKNOWN input, or
any floating input causes the output to also go UNKNOWN.

a9 [1 2] 8 xnor3
.model xnor3 d_xnor(rise_delay = 0.5e-9 fall_delay = 0.3e-9
12.4.9 Tristate
NAME_TABLE:
C_Function_Name: cm_d_tristate
Spice_Model_Name: d_tristate
Description: "digital tristate buffer"
PORT_TABLE:
Port Name: in enable out
Description: "input" "enable" "output"
Default_Type: d d d
Allowed_Types: [d] [d] [d]
Vector: no no no
PARAMETER_TABLE:
Description: "delay"
Data_Type: real
Limits: [1.0e-12 -]
Vector: no
Vector_Bounds: -
Null_Allowed: yes
PARAMETER_TABLE:
Parameter_Name: enable_load
Description: "enable load value (F)"
Data_Type: real
Limits: -
Vector: no
Vector_Bounds: -
Null_Allowed: yes
Common parameters: inertial_delay, family, input_load.
Description: The digital tristate is a simple tristate gate that can be configured to
allow for open-collector behavior, as well as standard tristate behavior. The state
seen on the input line is reflected in the output. The state seen on the enable line
determines the strength of the output. Thus, a ONE forces the output to its state
with a STRONG strength. A ZERO forces the output to go to a HI_IMPEDANCE
strength. The delays associated with an output state or strength change cannot
be specified independently, nor may they be specified independently for rise or fall
conditions; other gate models may be used to provide such delays if needed. The
model posts input and enable load values (in farads) based on the parameters input
load and enable. Note also that to maintain the technology-independence of the
model, any UNKNOWN input, or any floating input causes the output to also go
UNKNOWN. Likewise, any UNKNOWN input on the enable line causes the output
to go to an UNDETERMINED strength value.

a9 1 2 8 tri7
.model tri7 d_tristate(delay = 0.5e-9 input_load = 0.5e-12

+ enable_load = 0.5e-12)
12.4.10 Pullup
NAME_TABLE:
C_Function_Name: cm_d_pullup
Spice_Model_Name: d_pullup
Description: "digital pullup resistor"
PORT_TABLE:
Port Name: out
Direction: out
Default_Type: d
Allowed_Types: [d]
Vector: no
Vector_Bounds: -
Null_Allowed: no
PARAMETER_TABLE:
Parameter_Name: load
Description: "load value (F)"
Data_Type: real
Limits: -
Vector: no
Vector_Bounds: -
Null_Allowed: yes
Description: The digital pullup resistor is a device that emulates the behavior of an
analog resistance value tied to a high voltage level. The pullup may be used in
conjunction with tristate buffers to provide open-collector wired or constructs, or
any other logical constructs that rely on a resistive pullup common to many tristated
output devices. The model posts an input load value (in farads) based on the
parameter load.

a2 9 pullup1
.model pullup1 d_pullup(load = 20.0e-12)
12.4.11 Pulldown
NAME_TABLE:
C_Function_Name: cm_d_pulldown
Spice_Model_Name: d_pulldown
Description: "digital pulldown resistor"
PORT_TABLE:
Port Name: out
Direction: out
Default_Type: d
Allowed_Types: [d]
Vector: no
Vector_Bounds: -
Null_Allowed: no
PARAMETER_TABLE:
Parameter_Name: load
Description: "load value (F)"
Data_Type: real
Limits: -
Vector: no
Vector_Bounds: -
Null_Allowed: yes
Description: The digital pulldown resistor is a device that emulates the behavior of
an analog resistance value tied to a low voltage level. The pulldown may be used
in conjunction with tristate buffers to provide open-collector wired or constructs,
or any other logical constructs that rely on a resistive pulldown common to many
tristated output devices. The model posts an input load value (in farads) based on
the parameter load.

a4 9 pulldown1
.model pulldown1 d_pulldown(load = 20.0e-12)
12.4.12 D Flip Flop

NAME_TABLE:
C_Function_Name: cm_d_dff
Spice_Model_Name: d_dff
Description: "digital d-type flip flop"
PORT_TABLE:
Port Name: data clk
Description: "input data" "clock"
Direction: in in
Default_Type: d d
Vector: no no
Vector_Bounds: - -
Null_Allowed: no no
PORT_TABLE:
Port Name: set reset
Description: "asynch. set" "asynch. reset"
Direction: in in
Default_Type: d d

Vector: no no
Vector_Bounds: - -
PORT_TABLE:
Port Name: out Nout
Description: "data output" "inverted data output"
Direction: out out
Default_Type: d d
Vector: no no
Vector_Bounds: - -
PARAMETER_TABLE:
Parameter_Name: clk_delay set_delay
Description: "delay from clk" "delay from set"
Limits: [1.0e-12 -] [1.0e-12 -]
Vector: no no
Vector_Bounds: - -
PARAMETER_TABLE:
Parameter_Name: reset_delay ic
Description: "delay from reset" "output initial state"
Data_Type: real int
Default_Value: 1.0e-9 0
Limits: [1.0e-12 -] [0 2]
Vector: no no
Vector_Bounds: - -
PARAMETER_TABLE:
Parameter_Name: data_load clk_load
Description: "data load value (F)" "clk load value (F)"
Limits: - -
Vector: no no
Vector_Bounds: - -
PARAMETER_TABLE:
Parameter_Name: set_load reset_load
Description: "set load value (F)" "reset load (F)"
Limits: - -
Vector: no no
Vector.Bounds: - -
Common parameters: rise_delay, fall_delay.
Description: The digital d-type flip flop is a one-bit, edge-triggered storage element that
will store data whenever the clk input line transitions from low to high (ZERO to
ONE). In addition, asynchronous set and reset signals exist, and each of the three
methods of changing the stored output of the d_dff have separate load values and
delays associated with them. Additionally, you may specify separate rise and fall
delay values that are added to those specified for the input lines; these allow for
more faithful reproduction of the output characteristics of different IC fabrication
technologies.
Note that any UNKNOWN input on the set or reset lines immediately results in an
UNKNOWN output.

a7 1 2 3 4 5 6 flop1
.model flop1 d_dff(clk_delay = 13.0e-9 set_delay = 25.0e-9
+ reset_delay = 27.0e-9 ic = 2 rise_delay = 10.0e-9
+ fall_delay = 3e-9)
12.4.13 JK Flip Flop

NAME_TABLE:
C_Function_Name: cm_d_jkff
Spice_Model_Name: d_jkff
Description: "digital jk-type flip flop"
PORT_TABLE:
Port Name: j k
Description: "j input" "k input"
Direction: in in
Default_Type: d d
Vector: no no
Vector_Bounds: - -
Null_Allowed: no no
PORT_TABLE:
Port Name: clk
Description: "clock"
Direction: in
Default_Type: d
Allowed_Types: [d]
Vector: no
Vector_Bounds: -
Null_Allowed: no
PORT_TABLE:
Description: "asynchronous set" "asynchronous reset"
Direction: in in
Default_Type: d d
Vector: no no
Vector_Bounds: - -
PORT_TABLE:
Port Name: out Nout
Direction: out out
Default_Type: d d
Vector: no no
Vector_Bounds - -
PARAMETER_TABLE:
Limits: [1.0e-12 -] [1.0e-12 -]
Vector: no no
Vector_Bounds: - -
PARAMETER_TABLE:
Data_Type: real int
Limits: [1.0e-12 -] [0 2]
Vector: no no
Vector_Bounds: - -
PARAMETER_TABLE:
Parameter_Name: jk_load clk_load
Description: "j,k load values (F)" "clk load value (F)"
Limits: - -
Vector: no no
Vector_Bounds: - -
PARAMETER_TABLE:
Limits: - -
Vector: no no
Vector_Bounds: - -
Description: The digital jk-type flip flop is a one-bit, edge-triggered storage element
that will store data whenever the clk input line transitions from low to high (ZERO
to ONE). In addition, asynchronous set and reset signals exist, and each of the three
methods of changing the stored output of the d_jkff have separate load values and
delays associated with them. Additionally, you may specify separate rise and fall
delay values that are added to those specified for the input lines; these allow for
more faithful reproduction of the output characteristics of different IC fabrication
technologies.
Note that any UNKNOWN inputs other than j or k cause the output to go UN-
KNOWN automatically.

a8 1 2 3 4 5 6 7 flop2
.model flop2 d_jkff(clk_delay = 13.0e-9 set_delay = 25.0e-9
12.4.14 Toggle Flip Flop

NAME_TABLE:
C_Function_Name: cm_d_tff
Spice_Model_Name: d_tff
Description: "digital toggle flip flop"
PORT_TABLE:
Port Name: t clk
Description: "toggle input" "clock"
Direction: in in
Default_Type: d d
Vector: no no
Vector_Bounds: - -
Null_Allowed: no no
PORT_TABLE:
Description: "set" "reset"
Direction: in in
Default_Type: d d
Vector: no no
Vector_Bounds: - -
PORT_TABLE:
Port Name: out Nout

Direction: out out
Default_Type: d d
Vector: no no
Vector_Bounds: - -
PARAMETER_TABLE:
Limits: [1.0e-12 -] [1.0e-12 -]
Vector: no no
Vector_Bounds - -
PARAMETER_TABLE:
Data_Type: real int
Limits: [1.0e-12 -] [0 2]
Vector: no no
Vector_Bounds: - -
PARAMETER_TABLE:
Parameter_Name: t_load clk_load
Description: "toggle load value (F)" "clk load value (F)"
Limits: - -
Vector: no no
Vector_Bounds: - -
PARAMETER_TABLE:
Default.Value: 1.0e-12 1.0e-12
Limits: - -
Vector: no no
Vector_Bounds: - -
Description: The digital toggle-type flip flop is a one-bit, edge-triggered storage element
that will toggle its current state whenever the clk input line transitions from low
to high (ZERO to ONE). In addition, asynchronous set and reset signals exist, and
each of the three methods of changing the stored output of the d_tff have separate
load values and delays associated with them. Additionally, you may specify separate
rise and fall delay values that are added to those specified for the input lines; these
allow for more faithful reproduction of the output characteristics of different IC
fabrication technologies.
Note that any UNKNOWN inputs other than t immediately cause the output to go
UNKNOWN.

a8 2 12 4 5 6 3 flop3
.model flop3 d_tff(clk_delay = 13.0e-9 set_delay = 25.0e-9
+ fall_delay = 3e-9 t_load = 0.2e-12)
12.4.15 Set-Reset Flip Flop

NAME_TABLE:
C_Function_Name: cm_d_srff
Spice_Model_Name: d_srff
Description: "digital set-reset flip flop"
PORT_TABLE:
Port Name: s r
Description: "set input" "reset input"
Direction: in in
Default_Type: d d
Vector: no no
Vector_Bounds: - -
Null_Allowed: no no
PORT_TABLE:
Port Name: clk
Description: "clock"
Direction: in
Default_Type: d
Allowed_Types: [d]
Vector: no
Vector_Bounds: -
Null_Allowed: no
PORT_TABLE:
Description: "asynchronous set" "asynchronous reset"
Direction: in in
Default_Type: d d
Vector: no no
Vector_Bounds: - -
PORT_TABLE:
Port Name: out Nout
Direction: out out
Default_Type: d d
Vector: no no
Vector_Bounds: - -
PARAMETER_TABLE:
Limits: [1.0e-12 -] [1.0e-12 -]
Vector: no no
Vector_Bounds: - -
PARAMETER_TABLE:
Data_Type: real int
Limits: [1.0e-12 -] [0 2]
Vector: no no
Vector_Bounds: - -
PARAMETER_TABLE:
Parameter_Name: sr_load clk_load
Description: "set/reset loads (F)" "clk load value (F)"
Limits: - -
Vector: no no
Vector_Bounds: - -
PARAMETER_TABLE:
Limits: - -
Vector: no no
Vector_Bounds: - -
Description: The digital sr-type flip flop is a one-bit, edge-triggered storage element
that will store data whenever the clk input line transitions from low to high (ZERO
to ONE). The value stored (i.e., the out value) will depend on the s and r input pin
values, and will be:
out=ONE if s=ONE and r=ZERO;

out=ZERO if s=ZERO and r=ONE;
out=previous value if s=ZERO and r=ZERO;
out=UNKNOWN if s=ONE and r=ONE;
In addition, asynchronous set and reset signals exist, and each of the three methods of
changing the stored output of the d_srff have separate load values and delays associated
with them. You may also specify separate rise and fall delay values that are added to
those specified for the input lines; these allow for more faithful reproduction of the output
characteristics of different IC fabrication technologies.
Note that any UNKNOWN inputs other than s and r immediately cause the output to
go UNKNOWN.
a8 2 12 4 5 6 3 14 flop7
.model flop7 d_srff(clk_delay = 13.0e-9 set_delay = 25.0e-9
12.4.16 D Latch
NAME_TABLE:
C_Function_Name: cm_d_dlatch
Spice_Model_Name: d_dlatch
Description: "digital d-type latch"
PORT_TABLE:
Port Name: data enable
Description: "input data" "enable input"
Direction: in in
Default_Type: d d
Vector: no no
Vector_Bounds: - -
Null_Allowed: no no
PORT_TABLE:
Direction: in in
Default_Type: d d
Vector: no no
Vector_Bounds: - -
PORT_TABLE:
Port Name: out Nout
Description: "data output" "inverter data output"
Direction: out out
Default_Type: d d
Vector: no no
Vector_Bounds: - -
Null_Allowed: no no
PARAMETER_TABLE:
Parameter_Name: data_delay
Description: "delay from data"
Data_Type: real
Limits: [1.0e-12 -]
Vector: no
Vector_Bounds: -
Null_Allowed: yes
PARAMETER_TABLE:
Parameter_Name: enable_delay set_delay
Description: "delay from enable" "delay from SET"
Limits: [1.0e-12 -] [1.0e-12 -]
Vector: no no
Vector_Bounds: - -
PARAMETER_TABLE:
Description: "delay from RESET" "output initial state"
Limits: [1.0e-12 -] -
Vector: no no
Vector_Bounds: - -
PARAMETER_TABLE:
Parameter_Name: data_load enable_load
Description: "data load (F)" "enable load value (F)"
Limits: - -
Vector: no no
Vector_Bounds: - -
PARAMETER_TABLE:

Limits: - -
Vector: no no
Vector_Bounds: - -
Description: The digital d-type latch is a one-bit, level-sensitive storage element that
will output the value on the data line whenever the enable input line is high (ONE).
The value on the data line is stored (i.e., held on the out line) whenever the enable
line is low (ZERO).
In addition, asynchronous set and reset signals exist, and each of the four methods of
changing the stored output of the d_dlatch (i.e., data changing with enable=ONE,
enable changing to ONE from ZERO with a new value on data, raising set and
raising reset) have separate delays associated with them. You may also specify
separate rise and fall delay values that are added to those specified for the input
lines; these allow for more faithful reproduction of the output characteristics of
different IC fabrication technologies.
Note that any UNKNOWN inputs other than on the data line when enable=ZERO
immediately cause the output to go UNKNOWN.

a4 12 4 5 6 3 14 latch1
.model latch1 d_dlatch(data_delay = 13.0e-9 enable_delay = 22.0e-9
+ set_delay = 25.0e-9
+ reset_delay = 27.0e-9 ic = 2
+ rise_delay = 10.0e-9 fall_delay = 3e-9)
12.4.17 Set-Reset Latch

NAME_TABLE:
C_Function_Name: cm_d_srlatch
Spice_Model_Name: d_srlatch
Description: "digital sr-type latch"
PORT_TABLE:
Port Name: s r
Direction: in in
Default_Type: d d
Vector: no no
Vector_Bounds: - -
Null_Allowed: no no
PORT_TABLE:
Port Name: enable
Description: "enable"
Direction: in
Default_Type: d
Allowed_Types: [d]
Vector: no
Vector_Bounds: -
Null_Allowed: no
PORT_TABLE:
Direction: in in
Default_Type: d d
Vector: no no
Vector_Bounds: - -
PORT_TABLE:
Port Name: out Nout
Direction: out out
Default_Type: d d
Vector: no no
Vector_Bounds: - -
Null_Allowed: no no
PARAMETER_TABLE:
Parameter_Name: sr_delay
Description: "delay from s or r input change"
Data_Type: real
Limits: [1.0e-12 -]
Vector: no
Vector_Bounds: -
Null_Allowed: yes
PARAMETER_TABLE:
Parameter_Name: enable_delay set_delay
Description: "delay from enable" "delay from SET"
Limits: [1.0e-12 -] [1.0e-12 -]
Vector: no no
Vector_Bounds: - -
PARAMETER_TABLE:
Description: "delay from RESET" "output initial state"
Limits: [1.0e-12 -] -
Vector: no no
Vector_Bounds: - -
PARAMETER_TABLE:
Parameter_Name: sr_load enable_load
Description: "s & r input loads (F)" "enable load value (F)"
Limits: - -
Vector: no no
Vector_Bounds: - -
PARAMETER_TABLE:
Limits: - -
Vector: no no
Vector_Bounds: - -
Description: The digital sr-type latch is a one-bit, level-sensitive storage element that
will output the value dictated by the state of the s and r pins whenever the enable
input line is high (ONE). This value is stored (i.e., held on the out line) whenever
the enable line is low (ZERO). The particular value chosen is as shown below:
s=ZERO, r=ZERO => out=current value (i.e., not change in output)

s=ZERO, r=ONE => out=ZERO
s=ONE, r=ZERO => out=ONE
s=ONE, r=ONE => out=UNKNOWN
Asynchronous set and reset signals exist, and each of the four methods of changing the
stored output of the d srlatch (i.e., s/r combination changing with enable=ONE, enable
changing to ONE from ZERO with an output-changing combination of s and r, raising
set and raising reset) have separate delays associated with them. You may also specify
separate rise and fall delay values that are added to those specified for the input lines;
these allow for more faithful reproduction of the output characteristics of different IC
fabrication technologies.
Note that any UNKNOWN inputs other than on the s and r lines when enable=ZERO
immediately cause the output to go UNKNOWN.

a4 12 4 5 6 3 14 16 latch2
.model latch2 d_srlatch(sr_delay = 13.0e-9 enable_delay = 22.0e-9

+ set_delay = 25.0e-9
+ reset_delay = 27.0e-9 ic = 2
+ rise_delay = 10.0e-9 fall_delay = 3e-9)
12.4.18 State Machine

NAME_TABLE:
C_Function_Name: cm_d_state
Spice_Model_Name: d_state
Description: "digital state machine"
PORT_TABLE:
Port Name: in clk
Description: "input" "clock"
Direction: in in
Default_Type: d d
Vector: yes no
PORT_TABLE:
Port Name: reset out
Description: "reset" "output"
Direction: in out
Default_Type: d d
Vector: no yes
Vector_Bounds: - [1 -]
PARAMETER_TABLE:
Parameter_Name: clk_delay reset_delay
Description: "delay from CLK" "delay from RESET"
Limits: [1.0e-12 -] [1.0e-12 -]
Vector: no no
Vector_Bounds: - -
PARAMETER_TABLE: Parameter_Name: state_file
Description: "state transition specification file name"
Data_Type: string
Default_Value: "state.txt"
Limits: -
Vector: no
Vector_Bounds: -
Null_Allowed: no
PARAMETER_TABLE:
Parameter_Name: reset_state
Description: "default state on RESET & at DC"

Data_Type: int
Default_Value: 0
Limits: -
Vector: no
Vector_Bounds: -
Null_Allowed: no
PARAMETER_TABLE:
Parameter_Name: clk_load
Description: "clock loading capacitance (F)"
Data_Type: real
Limits: -
Vector: no
Vector_Bounds: -
Null_Allowed: yes
PARAMETER_TABLE:
Parameter_Name: reset_load
Description: "reset loading capacitance (F)"
Data_Type: real
Limits: -
Vector: no
Vector_Bounds: -
Null_Allowed: yes
Common parameters: input_load.
Description: The digital state machine provides for straightforward descriptions of clocked
combinational logic blocks with a variable number of inputs and outputs and with
an unlimited number of possible states. The model can be configured to behave as
virtually any type of counter or clocked combinational logic block and can be used
to replace very large digital circuit schematics with an identically functional but
faster representation.
The d state model is configured through the use of a state definition file (state.in)
that resides in a directory of your choosing. The file defines all states to be under-
stood by the model, plus input bit combinations that trigger changes in state. An
example state.in file is shown below:
----------- begin file -------------

* This is an example state.in file. This file
* defines a simple 2-bit counter with one input. The
* value of this input determines whether the counter counts
* up (in = 1) or down (in = 0).
0 0s 0s 0 -> 3
1 -> 1
1 0s 1z 0 -> 0
1 -> 2
2 1z 0s 0 -> 1
1 -> 3
3 1z 1z 0 -> 2
3 1z 1z 1 -> 0
------------------ end file ---------------
Several attributes of the above file structure should be noted. First, all lines in the file
must be one of four types. These are
1. A comment, beginning with a ‘*’ in the first column.

2. A header line, which is a complete description of the current state, the outputs
corresponding to that state, an input value, and the state that the model will
assume should that input be encountered. The first line of a state definition must
always be a header line.
3. A continuation line, which is a partial description of a state, consisting of an input
value and the state that the model will assume should that input be encountered.
Note that continuation lines may only be used after the initial header line definition
for a state.
4. A line containing nothing but white-spaces (space, form-feed, newline, carriage re-
turn, tab, vertical tab).
A line that is not one of the above will cause a file-loading error. Note that in the example
shown, whitespace (any combination of blanks, tabs, commas) is used to separate values,
and that the character -> is used to underline the state transition implied by the input
preceding it. This particular character is not critical in of itself, and can be replaced with
any other character or non-broken combination of characters that you prefer (e.g. ==>,
>‌>, ’:’, resolves_to, etc.)
The order of the output and input bits in the file is important; the first column is always
interpreted to refer to the ’zeroth’ bit of input and output. Thus, in the file above, the
output from state 1 sets out[0] to 0s, and out[1] to 1z.
The state numbers need not be in any particular order, but a state definition (which
consists of the sum total of all lines that define the state, its outputs, and all methods by
which a state can be exited) must be made on contiguous line numbers; a state definition
cannot be broken into sub-blocks and distributed randomly throughout the file. On the
other hand, the state definition can be broken up by as many comment lines as you desire.
Header files may be used throughout the state.in file, and continuation lines can be dis-
carded completely if you so choose: continuation lines are primarily provided as a conve-
nience.

a4 [2 3 4 5] 1 12 [22 23 24 25 26 27 28 29] state1
.model state1 d_state(clk_delay = 13.0e-9 reset_delay = 27.0e-9
+ state_file = "newstate.txt" reset_state = 2)
Note: The file named by the parameter filename in state_file="filename" is sought

after according to a search list described in12.1.3.
12.4.19 Frequency Divider

NAME_TABLE:
C_Function_Name: cm_d_fdiv
Spice_Model_Name: d_fdiv
Description: "digital frequency divider"
PORT_TABLE:
Port Name: freq_in freq_out
Description: "frequency input" "frequency output"
Direction: in out
Default_Type: d d
Vector: no no
Vector_Bounds: - -
Null_Allowed: no no
PARAMETER_TABLE:
Parameter_Name: div_factor high_cycles
Description: "divide factor" "# of cycles for high out"
Data_Type: int int
Default_Value: 2 1
Limits: [1 -] [1 div_factor-1]
Vector: no no
Vector_Bounds: - -
PARAMETER_TABLE:
Parameter_Name: i_count
Description: "divider initial count value"
Data_Type: int
Default_Value: 0
Limits: -
Vector: no
Vector_Bounds: -
Null_Allowed: yes
PARAMETER_TABLE:
Parameter_Name: freq_in_load
Description: "freq_in load value (F)"
Data_Type: real
Limits: -
Vector: no
Vector_Bounds: -
Null_Allowed: yes
Description: The digital frequency divider is a programmable step-down divider that

accepts an arbitrary divisor (div_factor), a duty-cycle term (high_cycles), and an
initial count value (i_count). The generated output is synchronized to the rising
edges of the input signal.

a4 3 7 divider
.model divider d_fdiv(div_factor = 5 high_cycles = 3
+ i_count = 4 rise_delay = 23e-9
12.4.20 RAM
NAME_TABLE:
C_Function_Name: cm_d_ram
Spice_Model_Name: d_ram
Description: "digital random-access memory"
PORT_TABLE:
Port Name: data_in data_out
Description: "data input line(s)" "data output line(s)"
Direction: in out
Default_Type: d d
Vector: yes yes
Vector_Bounds: [1 -] data_in
Null_Allowed: no no
PORT_TABLE:
Port Name: address write_en
Description: "address input line(s)" "write enable line"
Direction: in in
Default_Type: d d
Vector: yes no
Null_Allowed: no no
PORT_TABLE:
Port Name: select
Description: "chip select line(s)"
Direction: in
Default_Type: d
Allowed_Types: [d]
Vector: yes
Vector_Bounds: [1 16]
Null_Allowed: no
PARAMETER_TABLE:
Parameter_Name: select_value
Description: "decimal active value for select line comparison"
Data_Type: int
Default_Value: 1
Limits: [0 32767]
Vector: no
Vector_Bounds: -
Null_Allowed: yes
PARAMETER_TABLE:
Parameter_Name: ic
Description: "initial bit state @ dc"
Data_Type: int
Default_Value: 2
Limits: [0 2]
Vector: no
Vector_Bounds: -
Null_Allowed: yes
PARAMETER_TABLE:
Parameter_Name: read_delay
Description: "read delay from address/select/write.en active"
Data_Type: real
Limits: [1.0e-12 -]
Vector: no
Vector_Bounds: -
Null_Allowed: yes
PARAMETER_TABLE:
Parameter_Name: data_load address_load
Description: "data_in load value (F)" "addr. load value (F)"
Limits: - -
Vector: no no
Vector_Bounds: - -
PARAMETER_TABLE:
Parameter_Name: select_load
Description: "select load value (F)"
Data_Type: real
Limits: -
Vector: no
Vector_Bounds: -
Null_Allowed: yes
PARAMETER_TABLE:
Parameter_Name: enable_load
Description: "enable line load value (F)"
Data_Type: real
Limits: -
Vector: no
Vector_Bounds: -
Null_Allowed: yes
Description: The digital RAM is an M-wide, N-deep random access memory element
with programmable select lines, tristated data out lines, and a single write/~read
line. The width of the RAM words (M) is set through the use of the word width
parameter. The depth of the RAM (N) is set by the number of address lines input
to the device. The value of N is related to the number of address input lines (P) by
the following equation:
2P = N
There is no reset line into the device. However, an initial value for all bits may be
specified by setting the ic parameter to either 0 or 1. In reading a word from the
ram, the read delay value is invoked, and output will not appear until that delay
has been satisfied. Separate rise and fall delays are not supported for this device.
Note that UNKNOWN inputs on the address lines are not allowed during a write.
In the event that an address line does indeed go unknown during a write, the entire
contents of the ram will be set to unknown. This is in contrast to the data in lines
being set to unknown during a write; in that case, only the selected word will be
corrupted, and this is corrected once the data lines settle back to a known value.
Note that protection is added to the write en line such that extended UNKNOWN
values on that line are interpreted as ZERO values. This is the equivalent of a read
operation and will not corrupt the contents of the RAM. A similar mechanism exists
for the select lines. If they are unknown, then it is assumed that the chip is not
selected.
Detailed timing-checking routines are not provided in this model, other than for
the enable delay and select delay restrictions on read operations. You are advised,
therefore, to carefully check the timing into and out of the RAM for correct read
and write cycle times, setup and hold times, etc. for the particular device they are
attempting to model.

a4 [3 4 5 6] [3 4 5 6] [12 13 14 15 16 17 18 19] 30 [22 23 24] ram2
.model ram2 d_ram(select_value = 2 ic = 2 read_delay = 80e-9)
12.4.21 Digital Source

NAME_TABLE:
C_Function_Name: cm_d_source
Spice_Model_Name: d_source
Description: "digital signal source"
PORT_TABLE:
Port Name: out
Direction: out
Default_Type: d
Allowed_Types: [d]
Vector: yes
Vector_Bounds: -
Null_Allowed: no
PARAMETER_TABLE:
Parameter_Name: input_file
Description: "digital input vector filename"
Data_Type: string
Default_Value: "source.txt"
Limits: -
Vector: no
Vector_Bounds: -
Null_Allowed: no
Common parameters: input_load.
Description: The digital source provides for straightforward descriptions of digital signal
vectors in a tabular format. The model reads input from the input file and, at the
times specified in the file, generates the inputs along with the strengths listed. The
format of the input file is as shown below. Note that comment lines are delineated
through the use of a single ‘*’ character in the first column of a line. This is similar
to the way the SPICE program handles comments.
* T c n n n . . .
* i l o o o . . .
* m o d d d . . .
* e c e e e . . .
* k a b c . . .
0.0000 Uu Uu Us Uu . . .
1.234e-9 0s 1s 1s 0z . . .
1.376e-9 0s 0s 1s 0z . . .
2.5e-7 1s 0s 1s 0z . . .
2.5006e-7 1s 1s 1s 0z . . .
5.0e-7 0s 1s 1s 0z . . .
Note that in the example shown, whitespace (any combination of blanks, tabs, commas)
is used to separate the time and state/strength tokens. The order of the input columns
is important; the first column is always interpreted to mean ‘time’. The second through
the N’th columns map to the out[0] through out[N-2] output nodes. A non-commented
line that does not contain enough tokens to completely define all outputs for the digital
source will cause an error. Also, time values must increase monotonically or an error will
result in reading the source file.
Errors will also occur if a line exists in source.txt that is neither a comment nor vector
line. The only exception to this is in the case of a line that is completely blank; this is
treated as a comment (note that such lines often occur at the end of text within a file;
ignoring these in particular prevents nuisance errors on the part of the simulator).

a3 [2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17] input_vector
.model input_vector d_source(input_file = "source_simple.text")
Note: The file named by the parameter filename in input_file="filename" is sought

after according to a search list described in12.1.3.
12.4.22 LUT
NAME_TABLE:
C_Function_Name: cm_d_lut
Spice_Model_Name: d_lut
Description: "digital n-input look-up table gate"
PORT_TABLE:
Port_Name: in out
Direction: in out
Default_Type: d d
Vector: yes no
Null_Allowed: no no
PARAMETER_TABLE:
Parameter_Name: table_values
Description: "lookup table values"
Data_Type: string
Default_Value: "0"
Limits: -
Vector: no
Vector_Bounds: -
Null_Allowed: no
Common parameters: rise_delay, fall_delay, input_load.
Description: The lookup table provides a way to map any arbitrary n-input, 1-output
combinational logic block to XSPICE. The inputs are mapped to the output using
a string of length 2^n. The string may contain values "0", "1" or "X", corresponding
to an output of low, high, or unknown, respectively. The outputs are only mapped
for inputs which are valid logic levels. Any unknown bit in the input vector will
always produce an unknown output. The first character of the string table_values
corresponds to all inputs value zero, and the last (2^n) character corresponds to all
inputs value one, with the first signal in the input vector being the least significant
bit. For example, a 2-input lookup table representing the function (A * B) (that is,
A AND B), with input vector [A B] can be constructed with a table_values string
of "0001"; function (~A * B) with input vector [A B] can be constructed with a
table_values string of "0010".

* LUT encoding 3-bit parity function
a4 [1 2 3] 5 lut_pty3_1
.model lut_pty3_1 d_lut(table_values = "01101001"
+ input_load 2.0e-12)
12.4.23 General LUT

NAME_TABLE:
C_Function_Name: cm_d_genlut
Spice_Model_Name: d_genlut
Description: "digital n-input x m-output look-up table gate"
PORT_TABLE:
Port_Name: in out
Direction: in out
Default_Type: d d
Vector: yes yes
Vector_Bounds: - -
Null_Allowed: no no
PARAMETER_TABLE:
Parameter_Name: input_load input_delay
Description: "input load value (F)" "input delay"
Limits: - -
Vector: yes yes
Vector_Bounds: - -
PARAMETER_TABLE:
Parameter_Name: table_values
Description: "lookup table values"
Data_Type: string
Default_Value: "0"
Limits: -
Vector: no
Vector_Bounds: -
Null_Allowed: no
Description: The lookup table provides a way to map any arbitrary n-input, m-output
combinational logic block to XSPICE. The inputs are mapped to the output using
a string of length m * (2^n). The string may contain values "0", "1", "X", or "Z",
corresponding to an output of low, high, unknown, or high-impedance, respectively.
The outputs are only mapped for inputs which are valid logic levels. Any unknown
bit in the input vector will always produce an unknown output. The character
string is in groups of (2^n) characters, one group corresponding to each output pin,
in order. The first character of a group in the string table_values corresponds to
all inputs value zero, and the last (2^n) character in the group corresponds to all
inputs value one, with the first signal in the input vector being the least significant
bit. For example, a 2-input lookup table representing the function (A * B) (that is,
A AND B), with input vector [A B] can be constructed with a table_values string
of "0001"; function (~A * B) with input vector [A B] can be constructed with a
"table_values" string of "0010". The delays associated with each output pin’s rise
and those associated with each output pin’s fall may be specified independently. The
model also posts independent input load values per input pin (in farads) based on
the parameter input_load. The parameter input_delay provides a way to specify
additional delay between each input pin and the output. This delay is added to the
rise- or fall-time of the output. The output of this model does not respond to the
total loading it sees on the output; it will always drive the output strongly with the
specified delays.

* LUT encoding 3-bit parity function
a4 [1 2 3] [5] lut_pty3_1
.model lut_pty3_1 d_genlut(table_values = "01101001"
+ input_load [2.0e-12])
* LUT encoding a tristate inverter function (en in out)
a2 [1 2] [3] lut_triinv_1
.model lut_triinv_1 d_genlut(table_values = "Z1Z0")
* LUT encoding a half-adder function (A B Carry Sum)
a8 [1 2] [3 4] lut_halfadd_1
.model lut_halfadd_1 d_genlut(table_values = "00010110"
+ rise_delay [ 1.5e-9 1.0e-9 ] fall_delay [ 1.5e-9 1.0e-9 ])
12.4.24 D_process
NAME_TABLE:
C_Function_Name: cm_d_process
Spice_Model_Name: d_process
Description: "digital process"
PORT_TABLE:
Port_Name: in clk
Description: "input" "clock"
Direction: in in
Default_Type: d d
Vector: yes no
Vector_Bounds: - -
PORT_TABLE:
Port_Name: reset out
Description: "reset" "output"
Direction: in out
Default_Type: d d
Vector: no yes
Vector_Bounds: - [1 -]
PARAMETER_TABLE:
Parameter_Name: clk_delay
Description: "delay from CLK"
Data_Type: real
Limits: -
Vector: no
Vector_Bounds: -
Null_Allowed: yes
PARAMETER_TABLE:
Parameter_Name: process_file
Description: "file name of the executable process"
Data_Type: string
Default_Value: "process"
Limits: -
Vector: no
Vector_Bounds: -
Null_Allowed: no
PARAMETER_TABLE:
Parameter_Name: process_params
Description: "parameters to be passed to an executable process"
Data_Type: string
Default_Value: -
Limits: -
Vector: yes
Vector_Bounds: -
Null_Allowed: yes
PARAMETER_TABLE:
Parameter_Name: input_load
Description: "input loading capacitance (F)"
Data_Type: real
Limits: -
Vector: no
Vector_Bounds: -
Null_Allowed: no
PARAMETER_TABLE:
Parameter_Name: clk_load
Description: "clock loading capacitance (F)"
Data_Type: real
Limits: -
Vector: no
Vector_Bounds: -
Null_Allowed: no
PARAMETER_TABLE:
Parameter_Name: reset_load
Description: "reset loading capacitance (F)"
Data_Type: real
Limits: -
Vector: no
Vector_Bounds: -
Null_Allowed: no
Description: The digital d_process model runs an external program, specified by the
process_file parameter, to read the input signals when the clock changes to ONE
and then produces the output signals after the clk_delay. There can be zero (null)
or more inputs, and one or more outputs. The maximum number of inputs or out-
puts is 255 bits wide. If a reset signal is specified and has the value ONE when
the clock changes to ONE, the external program is notified of the reset by sending
it a negative time value. The output signals are initialized to Uz. The strength
(s, r, z, u) of an input signal is ignored. After time 0.0 initialization, outputs are
driven with STRONG (s) strength. The input and output states are binary ONE
or ZERO. If an input value is UNKNOWN (U) then a ONE or ZERO is chosen at
random.
The external program is started by fork/exec or spawn, and connections are es-
tablished using pipes. The external program is written in C, and first of all, in
main() the argc, argv parameters can be read. These command line parameters are
those specified in the process_params field of the d_process .model statement. A
header is sent from ngspice to the external program which acknowledges that the
number of inputs and outputs match. Thereafter, the external program executes
a loop: while (read data from the input pipe and if it is OK) { compute output
data for that input write the output data to the output pipe } In the meantime
the cm_d_process code in ngspice is writing data to its output pipe at each clock
change to ONE, then reading on its input pipe the response from the external pro-
gram.
Please see examples/xspice/d_process for a simple example and study the source
code in the .c files. The d_process model was developed by Uros Platise and he has
provided a non-trivial example and detailed descriptions at:
https://www.isotel.eu/mixedsim/embedded/motorforce/index.html.

a1 [d1] clk1 reset1 [o1 o2 o3 o4] proc1
.model proc1 d_process (process_file="prog1in4out"
+ clk_delay = 2.5e-9)
12.4.25 d_cosim
NAME_TABLE:
Spice_Model_Name: d_cosim
C_Function_Name: ucm_d_cosim
Description: "Bridge to an irreversible digital model"
PORT_TABLE:
Port_Name: d_in
Description: "digital input"
Direction: in
Default_Type: d
Allowed_Types: [d]
Vector: yes
Null_Allowed: yes
PORT_TABLE:
Port_Name: d_out
Description: "digital output"
Direction: out
Default_Type: d
Allowed_Types: [d]
Vector: yes
Null_Allowed: yes
PORT_TABLE:
Port_Name: d_inout
Description: "digital bidirectional port"
Direction: inout
Default_Type: d
Allowed_Types: [d]
Vector: yes
Null_Allowed: yes
PARAMETER_TABLE:
Description: "output delay time"
Data_Type: real
Limits: [1e-12 -]
Vector: no
Vector_bounds: -
Null_Allowed: yes
PARAMETER_TABLE:
Parameter_Name: simulation
Description: "A shared library containing a digital model"
Data_Type: string
Default_Value: -
Limits: -
Vector: no
Vector_Bounds: -
Null_Allowed: no
/* Instances maintain an internal input event queue that should be at least
* as large as the number of inputs. Performance with clocked logic may
* be improved by making it larger than (2 * F) / MTS, where F is
* the clock frequency and MTS is the maximum timestep for .tran.
*/
PARAMETER_TABLE:
Parameter_Name: queue_size
Description: "input queue size"
Data_Type: int
Default_Value: 128
Limits: [1 -]
Vector: no
Vector_bounds: -
Null_Allowed: yes
PARAMETER_TABLE:
Parameter_Name: irreversible
Description: "Parameter passed to library function cm_irreversible()"
Data_Type: int
Default_Value: 1
Limits: -
Vector: no
Vector_Bounds: -
Null_Allowed: yes
STATIC_VAR_TABLE:
Static_Var_Name: cosim_instance
Data_Type: pointer
Description: "Per-instance structure"
The d_cosim code model is similar to d_process, as it also requires external software
to define the behaviour of a code model instance. An important difference is that with
d_cosim such software runs inside the ngspice process. This code model is intended as a
container for other types of digital simulation and to provide a simplified programming
interface for devices whose behaviour is defined purely by conventional software. In par-
ticular, it is intended to act as a container for sub-simulations that can not discard any
time steps that fail in the main simulator; that is, they are irreversible.
The actual behaviour of any d_cosim instance is defined by a shared library or Windows
DLL that is set by the ’simulation’ parameter and dynamically loaded. Input changes are
relayed to a function in this library and any outputs reported by the library are relayed
to the simulated circuit. The interface between this code model and the library that it
hosts is defined in C-language header file cosim.h, included in the Ngspice source code
and in directory share/ngspice/scripts/src/ngspice in a binary package. This interface is
simpler than the XSPICE programming interface, but that has a cost: without special
care only one d_cosim instance should exist in a circuit. For more information, see the
description of cm_irreversible() in section 28.7.2.7.
Example SPICE usage:
adut [ Clk Comp Start] [Sample Valid ~d5 ~d4 ~d3 ~d2 ~d1 ~d0] null dut
.model dut d_cosim simulation="./adc.so"
A method for creating a suitable library from HDL code is described in section 14.3.
12.5. PREDEFINED NODE TYPES FOR EVENT DRIVEN SIMULATION 287
12.5 Predefined Node Types for event driven simu-

lation
The following predefined node types are included with the XSPICE simulator. These
should provide you not only with valuable event-driven modeling capabilities, but also
with examples to use for guidance in creating new UDN (user defined node) types. You
may access these node data by the plot (17.5.56) or eprint (17.5.29) commands.
12.5.1 Digital Node Type

The ‘digital’ node type is directly built into the simulator. 12 digital node values are
available. They are described by a two character string (the state/strength token). The
first character (0, 1, or U) gives the state of the node (logic zero, logic one, or unknown
logic state). The second character (s, r, z, u) gives the "strength" of the logic state (strong,
resistive, hi-impedance, or undetermined). So these are the values we have: 0s, 1s, Us,
0r, 1r, Ur, 0z, 1z, Uz, 0u, 1u, Uu.
12.5.2 Real Node Type

The ‘real’ node type provides for event-driven simulation with double-precision floating
point data. This type is useful for evaluating sampled-data filters and systems. The
type implements all optional functions for User-Defined Nodes, including inversion and
node resolution. For inversion, the sign of the value is reversed. For node resolution,
the resultant value at a node is the sum of all values output to that node. The node is
implemented as a user defined node in ngspice/src/xspice/icm/xtraevt/real.
12.5.3 Int Node Type

The ‘int’ node type provides for event-driven simulation with integer data. This type is
useful for evaluating round-off error effects in sampled-data systems. The type implements
all optional functions for User-Defined Nodes, including inversion and node resolution. For
inversion, the sign of the integer value is reversed. For node resolution, the resultant value
at a node is the sum of all values output to that node. The node is implemented as a user
defined node in ngspice/src/xspice/icm/xtraevt/int.
12.5.4 (Digital) Input/Output

The analog code models use the standard (analog) nodes provided by ngspice and thus
are using all the commands for sourcing, storing, printing, and plotting data.
I/O for event nodes (digital, real, int, and UDNs) is offered by the following tools: For out-
put you may use the plot (17.5.56) or eprint (17.5.29) commands, as well as edisplay
(17.5.28) and eprvcd (17.5.30). The latter writes all node data to a VCD file (a dig-
ital standard interface) that may be analyzed by viewers like gtkwave. For input, you
may create a test bench with existing code models (oscillator (12.3.4), frequency divider
(12.4.19), state machine (12.4.18) etc.). Reading data from a file is offered by d_source
(12.4.21). Some comments and hints have been provided by Sdaau. You may also use
the analog input from file, (filesource 12.2.9) and convert its analog input to the digital
type by the adc_bridge (12.3.2). If you want reading data from a VCD file, please have
a look at ngspice tips and examples forum and apply a python script provided by Sdaau
to translate the VCD data to d_source or filesource input.
12.6 Automatic insertion of bridging devices

Within ngspice, event nodes such as digital are quite different objects to analog nodes,
but in real circuits analog and digital devices may interconnect. Ngspice requires bridging
devices to interconnect its analog and digital domains.
Bridges are inserted automatically whenever an analog and a digital node have the same
name, so they are not required to be included in the netlist. To examine the inserted
bridging devices, use the command “listing e”. The extra devices appear at the end
of the netlist. Automatic bridging may be disabled by setting the interpreter variable
auto_bridge to zero.
The code models used for analog/digital bridges are described in section 12.3. The default
models are:
* Model for bridging digital node with inputs only.

.model auto_adc adc_bridge(in_low = 1.65 in_high = 1.65)
* Model for bridging digital node with outputs only.

.model auto_dac dac_bridge(out_low = 0 out_high = 3.3)
* Model for bridging digital node with either an inout connection or

* both inputs and outputs.
.model auto_bidi bidi_bridge(out_high=3.3 in_low=1.65 in_high=1.65)
A 3.3 volt supply has been assumed. That may be overriden by setting a parameter, vcc,
to the supply voltage. When bridges are inserted in a subcircuit the local value of the pa-
rameter is used, so subcircuits may have differing supply voltages. An alternative name for
the parameter may be set as the value of the interpreter variable auto_bridge_parm_d.
If the defaults are unsatisfactory, they may be overridden by setting interpreter variables:
* Override the default DAC bridge for TTL levels

.control
pre_set auto_bridge_d_out =
+ ( ".model auto_dac dac_bridge(out_low = 0.4 out_high = 3.6)"
+ "auto_bridge%d [ %s ] [ %s ] auto_dac" )
.endc
12.6. AUTOMATIC INSERTION OF BRIDGING DEVICES 289
The variable name is formed from a fixed part (auto_bridge_), the type of the event
node (d is the internal name for "digital") and the bridging direction (in, out or inout).
The first string is the model definition and the second is expanded into an instance of
the bridging device. Note that the pre_set command is used so that the variable is set
before the circuit is parsed.
Bridges may be defined by subcircuits as well as single devices:
pre_set auto_bridge_d_out = ( ".include test_sub.subcir"

+ "xauto_buf%d %s %s auto_buf vcc=%g"
+ 1 )
Here the constant "1" is required to specify that a separate instance of the subcircuit is
needed for each bridged node.
The included file might be:
* DAC with internal resistance.

.subckt auto_buf dig ana vcc=5
.model auto_dac dac_bridge(out_low = 0 out_high = {vcc})
auto_dac [ dig ] [ internal ] auto_dac
rint internal ana 100
.ends
An additional method for controlling automatic bridging is to set the parameter family
on individual XSPICE devices or on subcircuits. When the parameter is found a specific
interpreter variable is used to control bridges attached to the device, or as the default
within the subcircuit. In this example all output bridges connected to or inside the
subcircuit are specified.
Xmpx_gate [in0 in1 in2 in3] [sel1 sel0] out multiplexor

+ family=”lsttl”
.control
pre_set auto_bridge_lsttl_d_out =
+ ( ".model auto_dac dac_bridge(out_low = 0.2 out_high = 3.6)"
+ "auto_bridge%d [ %s ] [ %s ] auto_dac" )
.endc
More details of controls on automatic bridging can found as a comment in the source
file src/xspice/evt/evtcheck.c. Some examples of automatic bridging with various control
options are included in the source directory examples/digital/auto_bridge.
Chapter 13
Verilog-A Compact Device Models
13.1 Introduction
New compact device models today are released as Verilog-A code, a analog subset of
Verilog-AMS. Well-known examples are BSIMBULK, BSIMCMG, PSP, HiSIM or HICUM.
The Si2 CMC web page list more than 20 device models which are publicly available. The
models cover state-of-the-art MOS devices like SOI, FinFet, multi-gate and high voltage
transistors, high speed SiGe bipolar transistors, HEMTs as well as complex diodes and
resistors. ngspice makes all of these models available by its integrated OSDI interface and
the OpenVAF compiler, which translates Verilog-A device models into dynamically load-
able libraries. User-defined Verilog-A models may be compiled and loaded into ngspice
as well. Currently Linux and MS Windows are supported, OSDI/OpenVAF for macOS is
not yet available. We are thankful to SemiMod GmbH for these excellent contributions.
13.2 OSDI/OpenVAF
OSDI is a simulator independent interface for device models. Since release 39 ngspice
contains an integrated adapter to serve this interface and communicate with the compiled
shared library device models. The shared library models are linked into ngspice dynami-
cally at runtime with the osdi or pre_osdi (see 17.5.58) .control language commands.
OpenVAF compiles Verilog-A compact device model files into shared libraries that con-
form to the OSDI interface. The model descriptions have to comply with the standard
Verilog-AMS LRM 2.x. Since ngspice-42, the small signal noise simulation (15.3.4) is im-
plemented. Noise simulation , however, is only available with the Sparse 1.3 matrix solver,
not with KLU (see 15.1.1). Other restrictions may apply. Please consult the OpenVAF
web pages for further information. QA actions are not possible due to CMC refusing to
provide data.
291
292 CHAPTER 13. VERILOG-A COMPACT DEVICE MODELS
13.3 How to create and apply OpenVAF models
13.3.1 Preparing for simulation

Using Verilog-A models for simulation in ngspice consists of five steps: Obtain or compile
ngspice with OSDI interface, compile the VA-model with OpenVAF, prepare a suitable
model parameter set, load the compiled model into ngspice ... and start the simulation.
13.3.1.1 Obtaining OpenVAF
OpenVAF may be downloaded for MS Windows or Linux as a single executable each from
https://openvaf.semimod.de/download/, and copied into a user defined directory. Com-
piling OpenVAF yourself is possible, however is not recommended due to its complicated
procedure.
13.3.1.2 Verilog-A compact models
Verilog-A compact device models are available from the si2 CMC standard compact model
page or directly from device modelling web sites, e.g. BSIM from UC Berkeley, HiSIM
from Hiroshima University, PSP from CEA-Leti, or HICUM from TU Dresden. Others
are available. User provided or user defined models may be compiled as well. All models
have to comply to the LRM 2.x standard of Verilog-AMS. Not all publicly availables
models do comply (e.g. PSP102, EKV2.6).
There is a github repository VA-Models with most of the public available Verilog-A com-
pact models. The models are checked against the LRM 2.4.0 and prepared for ngspice
simulation. A script for generation of the osdi files is provided and each model has more
or less simple ngspice netlist files to show main capabilities. So this web site should be a
good starting point for beginners.
13.3.1.3 Prepare ngspice
Compile ngspice with the configure flag –enable-osdi to add the OSDI interface to
ngspice. The MSVC Windows version ngspice.exe from the distribution already con-
tains this interface.
13.3.1.4 Compile the models
A very basic approach is to put the openvaf executable and the Verilog-A model (e.g.
bsimbulk.va) into a directory, then from a console window cd into that directory and call
the command openvaf bsimbulk.va. After a few seconds the compiled shared library
bsimbulk.osdi becomes available, ready to be loaded into ngspice.
Where to place *.osdi? Basically in any directory of your choice, the osdi or pre-osdi
commands (13.3.1.6) may be prepended by an absolute or relative path to that directory.
For a permanent location a bulk model install to libs/ngspice is recommended (to the
folder where you also find the XSPICE code model libs *.cm).
13.3. HOW TO CREATE AND APPLY OPENVAF MODELS 293
openvaf --help yields more options of the compiler.

To simplify making suitable *.osdi models for the example netlists provided in ngspice/ex-
amples/osdi, the appropriate Verilog-A models and short scripts (for Linux and Windows)
are available for download as VAforOSDI.7z from our release directory. The following steps
are required to compile the shared library models:
• Expand VAforOSDI.7z into a directory of your choice.
• Download OpenVAF (Linux or Windows) from and place the executable here in this
directory.
• Run the script openvaf-compile-va.bat for MS Windows or openvaf-compile-va.sh

for Linux
• Copy the *.osdi files from directory osdilibs to the place where then code models
(*.cm) are located, typically in lib/ngspice or similar.
• Edit file ’spinit’, typically found in share/ngspice/scripts: Comment out the

line ’unset osdi_enabled’
13.3.1.5 Prepare the model parameters
According to chapter 2.4 the model parameter set for each device model is organized in a
.model line. This is valid for OSDI models as well. However here the model type takes the
role to distinguish models from each other, not the level or version parameters found
in the intrinsic models. A TYPE parameter determines, if NMOS (TYPE=1) or PMOS
(TYPE=-1), NPN (TYPE=1) or PNP (TYPE=-1) are selected.
Consider as an example the bsimbulk model. The modeltype is set by the
module bsimbulk(d, g, s, b, t);
line of the BSIMBULK.va Verilog-A model file. So one has to search for the module
name in the *.va file to obtain the modeltype for the .model line and the number of
nodes (and their meanings) for the instance (or device) line while creating the netlist (see
next chapter 13.3.1.6).
General form:
. model mname modeltype ( pname1 = pval1 pname2 = pval2 ... )
Examples:
. model BSIMBULK_osdi_N bsimbulk TYPE =1 GEOMOD =0 RGEOMOD =0 ...
So to prepare the .model line, select an appropriate model parameter set, comment out
the version and level parameters, add the type parameter, and change the modeltype
to the Verilog-A module name.
294 CHAPTER 13. VERILOG-A COMPACT DEVICE MODELS
13.3.1.6 Prepare the ngspice netlist
The compiled model, e.g. bsimbulk.osdi, has to be loaded into ngspice. This may occur
automatically during start-up of ngspice, if the installation has been prepared according
to the bulk model install (compiled *.osdi models in lib/ngspice, osdi commands added
to spinit).
Local usage of a *.osdi which are residing in an arbitrary directory is possible from within
a .control section (16.4.3) by the pre_osdi command. A relative path (as in the example
below) or an absolute path to that directory may be chosen.
pre_osdi ../osdi_libs/bsimbulk107.osdi
The reference designator for the OSDI devices is the letter N. Instance lines starting with
N are recognized as OSDI devices. The model name mname has to point to the .model
line which contains the parameter set to be selected.
Instance line, general form:
Ndevname node1 ... nodex mname pname1 = pval1 pname2 = pval2 ...
Examples:
Np1 z a vdd vdd BSIMBULK_osdi_P l =0.1 u w =1 u

+ as =0.26235 p ad =0.26235 p ps =2.51 u pd =2.51 u
Nn1 z a vss vss BSIMBULK_osdi_N l =0.1 u w =0.5 u
+ as =0.131175 p ad =0.131175 p ps =1.52 u pd =1.52 u
NMOS and PMOS devices are selected by their respective model names BSIMBULK_osdi_N
and BSIMBULK_osdi_P. The number and role of the nodes has been defined in the VA
code in its module statement (module bsimbulk(d, g, s, b, t); in the BSIMBULK
example). Instance parameters (like l, w, as etc.) are allowed, as defined by the VA code.
13.3.1.7 Run the simulation
The simulation may be run as usual. Batch mode is especially supported when the OSDI
libraries are loaded via spinit during ngspice start-up.
13.3.2 OSDI/OpenVAF examples distributed with ngspice

Several example input netlists are available in folder ngspice/examples/osdi. All (ex-
cept for bsimbulk-local) make use of the *.osdi installation as a bulk model install
in a folder pointed to by spinit. bsimbulk-local however requires a local copy of
bsimbulk.osdi into folder bsimbulk-local/osdi_libs.
Example folders bsimbulk, bsimbulk-local, bsimcmg, mixed-models, and psp103 con-
tain MOS devices with their dc characteristics, CMOS inverters, CMOS ring oscilla-
tors, or even the 7552_ann benchmark CMOS circuit with 15.000 transistors and may
more passives. Hicuml0, or mextram contain bipolar devices with output characteristics,
Gummel-plot and some circuits. r2_cmc is a special resistor model.
Chapter 14
Digital Device Models
14.1 U devices (basic digital building blocks)
If PS compatibility mode is set, ngspice supports .subckt statements which contain entirely
U* instances of digital gates, flip-flops, latches, LOGICEXP and PINDLY behavioral
primitives (see chapter 14.1.2 for the list), and timing models. Typical rise/fall delays are
estimated from the timing models and PINDLY statements. CONSTRAINT primitives
and io models are ignored. Other U* instances (such as RAM, ROM, STIM and PLAs)
in a .subckt are not supported, and such .subckt will not be converted to XSPICE digital
primitives.
These U devices are not meant to immediately describe digital circuits like the 74xx or
40xx series. However they are used in subcircuits to generate models for such circuits
(see chapter 14.2). Their syntax is mostly compatible to the Micro-Cap and PSPICE
simulators.
295
296 CHAPTER 14. DIGITAL DEVICE MODELS
14.1.1 General format

General form:
U < name > < basic type > [( < parameter value >*)]
+ < digital power node > < digital ground node > < node >*
+ < timing model name > <I / O model name >
+[ MNTYMXDLY = < delay select value >]
+[ IO_LEVEL = < interface subcircuit select value >]
Example:
U2 AND (2) $G_DPWR $G_DGND 4 5 6

+ M2 IOM2 IO_LEVEL =0 MNTYMXDLY =2
. MODEL M2 UGATE ()
. MODEL IOM2 UIO ( INLD =0 OUTLD =0 DRVH =50 DRVL =50
+ ATOD1 =" ATOD1 " DTOA1 =" DTOD1 " ATOD2 =" ATOD2 " DTOA2 =" DTOD2 "
+ ATOD3 =" ATOD3 " DTOA3 =" DTOD3 " ATOD4 =" ATOD4 " DTOA4 =" DTOD4 "
+ TSWLH1 =0 TSWLH2 =0 TSWLH3 =0 TSWLH4 =0
+ TSWHL1 =0 TSWHL2 =0 TSWHL3 =0 TSWHL4 =0 DIGPOWER =" DIGPOWER ")
14.1.2 List of devices available in ngspice (basic types)

Standard gates:
BUF buffer
INV inverter
AND AND gate
NAND NAND gate
OR OR gate
NOR NOR gate
XOR exclusive OR gate
NXOR exclusive NOR gate
BUFA buffer array
INVA inverter array
ANDA AND gate array
NANDA NAND gate array
ORA OR gate array
NORA NOR gate array
XORA exclusive OR gate array
NXORA exclusive NOR gate array
AO AND-OR compound gate
OA OR-AND compound gate
AOI AND-NOR compound gate
OAI OR-NAND compound gate
Tristate gates:
BUF3 buffer
14.2. SUPPORT FOR STANDARD DIGITAL DEVICES 297
INV3 inverter
AND3 AND gate
NAND3 NAND gate
OR3 OR gate
NOR3 NOR gate
XOR3 exclusive OR gate
NXOR3 exclusive NOR gate
BUF3A buffer array
INV3A inverter array
AND3A AND gate array
NAND3A NAND gate array
OR3A OR gate array
NOR3A NOR gate array
XOR3A exclusive OR gate array
NXOR3A exclusive NOR gate array
Flip-flops and latches:
DFF D-type flip-flop, positive-edge triggered
JKFF J-K flip-flop, negative-edge triggered
DLTCH D-type latch
SRFF S-R flip-flop
Delay lines:
DLYLINE Delay line
Behavioral primitives:
LOGICEXP Combinational logic expressions
PINDLY Output buffers and tristate buffers with estimated delays
14.1.3 URC transmission line versus U devices

For the first time ngspice may have a naming conflict, in that the reference designator U
is used for two different devices, the Uniformly distributet RC line and the U devices, our
digital basic types or primitives.
U-devices require the compatibility mode flag (16.14.1) set to PS. In addition U devices
are recognized only when they occur inside of a subcircuit. Finally the basic
type (second token in the U instance line) has to fit to the list of basic types
given in the table above.
URC in any other case ngspice will assume an URC (uniformly distributed) trans-
mission line.
14.2 Support for standard digital devices

The digital primitives (U devices) are the basic building blocks of the models for digital
ICs used in ngspice. An example of a simple subcircuit model for a And Gate is listed
below:
Example: 74LV08A Quad 2-Input And Gate
* - - - - - - - - - -- - - - - - - - - - -- - - - 74 LV08A ------
* Quad 2 - Input And Gate
*
* TI PDF File
* bss 2/21/03
*
. SUBCKT 74 LV08A 1 A 1 B 1 Y
+ optional : DPWR_3V = $G_DPWR_3V DGND_3V = $G_DGND_3V
+ params : MNTYMXDLY =0 IO_LEVEL =0
U1 and (2) DPWR_3V DGND_3V

+ 1A 1B 1Y
+ DLY_LV08 IO_LV - A MNTYMXDLY ={ MNTYMXDLY } IO_LEVEL ={ IO_LEVEL }
. model DLY_LV08 ugate

+ ( tplhTY =7.5 ns tplhMX =12.3 ns tphlTY =7.5 ns tphlMX =12.3 ns )
. ENDS 74 LV08A
The circuit example ex4.cir in ngspice/examples/digital/digital_devices, together with

the stimulus file ex4.stim presents a fully digital, event based full-adder simulation with
74xx series ICs. The internal plotting capability of ngspice is used. ex5.cir with the
stimulus file ex5.stim demonstrates the conversion of a D-Type Flip-Flop. ex283.cir is
a 74283 4-bit full adder, with stimulus and involving GTKWave for plotting. Also, there
are several new examples illustrating LOGICEXP and PINDLY.
A set of such models for the 74xx devices currently supported by ngspice is available from
the ngspice models as 74xx-models.7z, derived from the Micro-Cap library.
14.3 Digital devices defined by a Hardware Descrip-

tion Language
Ngspice can make a digital device from its description in a Hardware Description Lan-
guage, such as Verilog or VHDL. There are several ways that can be done, but the current
Ngspice source code and binary packages have direct support for only one method: HDL
files may be compiled by Verilator into a shared library or Windows DLL that can be
loaded into an instance of the d_cosim XSPICE code model (12.4.25). (Warning: using
multiple d_cosim instances in a circuit requires careful planning.)
The steps for using such a digital device are these: write HDL code whose top-level module
defines the device behaviour; the HDL file is compiled into a shared library/DLL accept-
able by d_cosim; and the netlist contains device and model lines to include the new device
in a circuit. Example netlist and Verilog code can be found at examples/xspice/verilator.
The compilation step is not straightforward, as “glue code” must be added to the C++
software created by Verilator so that it can be attached to a d_cosim instance. A script
is provided to make this step easier:
14.3. DIGITAL DEVICES DEFINED BY A HARDWARE DESCRIPTION LANGUAGE299
ngspice vlnggen source . v
Ngspice is used to run the script, vlnggen, passing the Verilog source file, source.v, as
input. The output will be a shared library/DLL called source.so/source.DLL that may
be used in the netlist as:
ahdldevice [ inputs ... ] [ outputs ... ] null dmod

. model dmod d_cosim simulation =" some / path / source . so "
(Here “null” represents possible “inout” ports that Verilator does not fully support.)
Additional arguments to vlnggen may be other HDL source files or any Verilator options:
everything is passed to Verilator. The output file is named from the first Verilog source
file (*.v) that is listed, or is “verilated.so”/”verilated.DLL”.
For each port type the connected event nodes are assigned to bits of the Verilog ports
from left to right, treating ports with multiple bits as big-endian: more significant bits
are matched first. If the number of bits of each port type do not match, a warning is
issued and extra input or output bits will be matched with any excess inout bits, in either
direction.
14.3.1 Another approach

A different method to bring HDL code into a netlist is to use Yosys to compile HDL di-
rectly to a sub-circuit definition using XSPICE elements. It is used here. The d_process
code model is now included in ngspice, with example netlists at examples/xspice/d_process.
Chapter 15
Analyses and Output Control (batch

mode)
The command lines described in this chapter are used to specify analyses and outputs
within the circuit description file. They start with a ‘.’ (dot commands). Specifying
analyses and plots (or tables) in the input file with dot commands is used with batch
runs. Batch mode is entered when either the -b option is given upon starting ngspice
ngspice -b -r rawfile.raw circuitfile.cir
or when the default input source is redirected from a file (see also Chapt. 16.4.1).
ngspice < circuitfile.cir
In batch mode, the analyses specified by the control lines in the input file (e.g. .ac,
.tran, etc.) are immediately executed. If the -r rawfile option is given then all data
generated is written to a ngspice rawfile. The rawfile may later be read by the interactive
mode of ngspice using the load command (see 17.5.48). In this case, the .save line (see
15.6) may be used to record the value of internal device variables (see Appendix, Chapt.
31).
If a rawfile is not specified, then output plots (in ‘line-printer’ form) and tables can be
printed according to the .print, .plot, and .four control lines, described in Chapt.
15.6.
If ngspice is started in interactive mode (see Chapt. 16.4.2), like
ngspice circuitfile.cir
and no control section (.control ... .endc, see 16.4.3) is provided in the circuit file, the
dot commands are not executed immediately, but are waiting for manually receiving the
command run.
15.1 Simulator Variables (.options)

Various parameters of the simulations available in Ngspice can be altered to control the
accuracy, speed, or default values for some devices. These parameters may be changed
via the option command (described in Chapt. 17.5.55) or via the .options line:
301
302 CHAPTER 15. ANALYSES AND OUTPUT CONTROL (BATCH MODE)
General form:
. options opt1 opt2 ... ( or opt = optval ...)
Examples:
. options reltol =.005 trtol =8
The options line allows the user to reset program control and user options for specific
simulation purposes. Options specified to ngspice via the option command (see17.5.55)
are also passed on as if specified on a .options line. Any combination of the following
options may be included, in any order. ‘x’ (below) represents some positive number.
15.1.1 General Options

SPARSE selects the Sparse 1.3 matrix solver, which is also the standard when no option
is given. It is prefgerable for simulating behavioural device models. This option is
required with noise (15.3.4) or CIDER (30) simulation.
KLU selects the KLU matrix solver, which is preferable (yielding faster simulation) when
(large) circuits containing MOS devices are to be simulated. Small signal noise
(15.3.4) or CIDER (30) simulations are not (yet) supported.
ACCT causes accounting and run time statistics to be printed.
NOACCT no printing of statistics, no printing of the Initial Transient Solution.
NOINIT suppresses only printing of the Initial Transient Solution, maybe combined
with ACCT.
LIST causes the summary listing of the input data to be printed.
NOMOD suppresses the printout of the model parameters.
NOPAGE suppresses page ejects.
NODE causes the printing of the node table.
NOREFVALUE suppresses printing of reference values, when ngspice has been com-
piled with configure option --enable-ndev.
OPTS causes the option values to be printed.
SEED=val|random Sets the seed value of the random number generator. val may be
any integer number greater than 0. As an alternative, random will set the seed
value to the current Unix epoch time, which is the time in seconds since 1.1.1970
excluding leap seconds.
SEEDINFO will print the seed value when it has been set to a new integer number.
15.1. SIMULATOR VARIABLES (.OPTIONS) 303
TEMP=x Resets the operating temperature of the circuit. The default value is 27 ◦ C
(300K). TEMP can be overridden per device by a temperature specification on any
temperature dependent instance. May also be generally overridden by a .TEMP
card (2.12).
TNOM=x resets the nominal temperature at which device parameters are measured.
The default value is 27 ◦ C (300 deg K). TNOM can be overridden by a specification
on any temperature dependent device model.
WARN=1|0 enables or turns of SOA (Safe Operating Area) voltage warning messages
(default: 0).
MAXWARNS=x specifies the maximum number of SOA (Safe Operating Area) warn-
ing messages per model (default: 5).
SAVECURRENTS save currents through all terminals of the following devices: M, J,

Q, D, R, C, L, B, F, G, W, S, I (see 2.1.3). Recommended only for small circuits,
because otherwise memory requirements explode and simulation speed suffers. See
15.7 for more details. This option is available only for op, dc, and tran simulation,
not for ac. During transient simulation the value returned may be delayed by one
time step. For M devices, MOS level 1 is supported fully, not all nodes are reported
for the other MOS devices. As the option is installed upfront, before the simulation,
it has no clue about the devices used in the circuit. It simply does do a best
guess. This may lead to empty vectors of zero length after the simulation, impeding
commands like wrdata (17.5.106). Running command remzerovec (17.5.64) before
wrdata will remove all these zero length vectors.
15.1.2 OP and DC Solution Options

The following options control properties pertaining to DC and OP (operating point)
analyses and algorithms. Since transient analysis (15.1.4) is based on OP, many of the
options affect transient simulation as well. AC analysis (15.1.3) can be performed only
when a stable operating point has been found.
ABSTOL=x resets the absolute current error tolerance of the program. The default
value is 1 pA.
GMIN=x resets the value of GMIN, the minimum conductance allowed by the program.
The default value is 1.0e-12.
GMINSTEPS=x [*] sets the number of Gmin steps to be attempted. If the value is set
to zero, the standard gmin stepping algorithm is skipped. The standard behavior
is that gmin stepping is tried before going to the source stepping algorithm.
ITL1=x resets the dc iteration limit. The default is 100.
ITL2=x resets the dc transfer curve iteration limit. The default is 50.
KEEPOPINFO Retain the operating point information when either an AC, Distortion,
or Pole-Zero analysis is run. This is particularly useful if the circuit is large and
you do not want to run a (redundant) .OP analysis.
NOOPITER Go directly to gmin stepping, skipping the first iteration.

PIVREL=x resets the relative ratio between the largest column entry and an accept-
able pivot value. The default value is 1.0e-3. In the numerical pivoting algo-
rithm the allowed minimum pivot value is determined by EPSREL = AMAX1(PIVREL ·
MAXVAL, PIVTOL) where MAXVAL is the maximum element in the column where a
pivot is sought (partial pivoting).
PIVTOL=x resets the absolute minimum value for a matrix entry to be accepted as a
pivot. The default value is 1.0e-13.
RELTOL=x resets the relative error tolerance of the program. The default value is
0.001 (0.1%).
RSHUNT=x introduces a resistor from each analog node to ground. The value of the
resistor should be high enough to not interfere with circuit operations. The XSPICE
option has to be enabled (see 32.1.8) .
VNTOL=x resets the absolute voltage error tolerance of the program. The default value
is 1 µV .
15.1.2.1 Matrix Conditioning info
In SPICE-based simulators, specific problems arise with certain circuit topologies. One
issue is the absence of a DC path to ground at some node. This may happen when two
capacitors are connected in series with no other connection at the common node, or when
code models are cascaded. The result is an ill-conditioned or nearly singular matrix that
prevents the simulation from completing. Configuring with XSPICE introduces the rshunt
option to help eliminate this problem. The option inserts resistors to ground at all the
analog nodes in the circuit. In general, the value of rshunt is set to some high resistance
(e.g. 1000 MΩ or greater) so that the operation of the circuit is essentially unaffected but
the matrix problems are corrected. If a ‘no DC path to ground’ or a ‘matrix is nearly
singular’ error message is encountered, add the following .option card to the circuit deck:
.option rshunt = 1.0e12
Usually a value of 1 TΩ is sufficient to correct the problem. In bad cases one can try
lowering the value to 10 GΩ or even 1 GΩ.
A different matrix conditioning problem occurs if an inductor is placed in parallel to a
voltage source. The AC simulation will fail, because it is preceded by an OP analysis.
Option NOOPAC (15.1.3) will help if the circuit is linear. However, if the circuit is non-
linear the OP analysis is essential. In such a case, adding a small resistor (e.g. 0.1mΩ) in
series to the inductor will help to obtain convergence.
.option rseries = 1.0e-4
adds a series resistor to each inductor in the circuit. Be careful when using behavioral
inductors (see 3.3.13), as the result may become unpredictable.
.option cshunt = 1.3e-13
adds a capacitor from each voltage node in the circuit to ground.

15.1. SIMULATOR VARIABLES (.OPTIONS) 305
15.1.3 AC Solution Options

NOOPAC Do not run an operating point (OP) analysis prior to an AC analysis. This
option requires that the circuit is linear, i.e. consists only of R, L, and C devices,
independent V, I sources and linear dependent E, G, H, and F sources (without poly
statement, non-behavioral). If a non-linear device is detected, the OP analysis is
executed automatically. This option is of interest e.g. in nested LC circuits where
no series resistance for L devices is present. During the OP analysis an ill-formed
matrix may be encountered, causing the simulator to abort with an error message.
It is also useful if you have very large linear arrays (10000 nodes and more), where
simulation speedup by a factor of 10 may be achieved.
15.1.4 Transient Analysis Options

AUTOSTOP stops a transient analysis after successfully calculating all functions (15.4)
specified with the dot command .meas. Autostop is not available with the meas
(17.5.50) command used in control mode.
CHGTOL=x resets the charge tolerance of the program. The default value is 1.0e-14.
CONVSTEP=x relative step limit applied to code models.
CONVABSSTEP=x absolute step limit applied to code models.
INTERP interpolates output data onto fixed time steps on a TSTEP grid (15.3.10). Uses
linear interpolation between previous and next time values. Simulation itself is not
influenced by this option. This option can be used in all simulation modes (batch,
control or interactive, 16.4). It may drastically reduce memory requirements in con-
trol mode, and file size in batch mode, but care is needed not to undersample the out-
put data. See also the command linearize (17.5.46) that achieves a similar result
by post-processing the data in control mode. The Ngspice/examples/xspice/delta-
sigma/delta-sigma-1.cir example demonstrates how INTERP reduces memory re-
quirements and speeds up plotting.
ITL3=x resets the lower transient analysis iteration limit. The default value is 4. (Note:
not implemented in Spice3).
ITL4=x resets the transient analysis time-point iteration limit. The default is 10.
ITL5=x resets the transient analysis total iteration limit. The default is 5000. Set
ITL5=0 to omit this test. (Note: not implemented in Spice3).
ITL6=x [*] synonym for SRCSTEPS.
MAXEVTITER=x sets the maximum number of event iterations per analysis point.
MAXOPALTER=x specifies the maximum number of analog/event alternations that
the simulator will use to solve a hybrid circuit.
MAXORD=x [*] specifies the maximum order for the numerical integration method
used by SPICE. Possible values for the Gear method are from 2 (the default) to 6.
Using the value 1 with the trapezoidal method specifies backward Euler integration.
METHOD=name sets the numerical integration method used by SPICE. Possible

names are ‘Gear’ or ‘trapezoidal’ (or just ‘trap’). The default is trapezoidal.
NOOPALTER=TRUE|FALSE if set to false, alternations between analog and event

calls to XSPICE models are enabled during initial DC operating analysis.
RAMPTIME=x During source stepping, this option sets the rate of change of indepen-
dent supplies. It also affects code model inductors and capacitors that have initial
conditions specified.
SRCSTEPS=x [*] a non-zero value causes SPICE to use a source-stepping method to

find the DC operating point. The value specifies the number of steps.
TRTOL=x resets the transient error tolerance. The default value is 7. This parameter
is an estimate of the factor by which SPICE overestimates the actual truncation
error. If XSPICE is configured and ’A’ devices are included, the value is internally
set to 1 for higher precision. This slows down transient analysis by a factor of two.
XMU=x sets the damping factor for trapezoidal integration. The default value is
XMU=0.5. A value < 0.5 may be chosen. Even a small reduction, e.g. to 0.495,
may already suppress trap ringing. The reduction has to be set carefully in order
not to excessively damp circuits that are prone to ringing or oscillation, which might
lead the user to believe that the circuit is stable.
15.1.5 ELEMENT Specific options
BADMOS3 Use the older version of the MOS3 model with the ‘kappa’ discontinuity.
DEFAD=x resets the value for MOS drain diffusion area; the default is 0.
DEFAS=x resets the value for MOS source diffusion area; the default is 0.
DEFL=x resets the value for MOS channel length; the default is 100 µm.
DEFW=x resets the value for MOS channel width; the default is 100 µm.
SCALE=x set the element scaling factor for geometric element parameters whose default
unit is meters. As an example: scale=1u and a MOSFET instance parameter W=10
will result in a width of 10µm for this device. An area parameter AD=20 will result
in 20e − 12 m2 . Following instance parameters are scaled:
• Resistors and Capacitors: W, L
• Diodes: W, L, Area
• JFET, MESFET: W, L, Area
• MOSFET: W, L, AS, AD, PS, PD, SA, SB, SC, SD

15.2. INITIAL CONDITIONS 307
15.1.6 Transmission Lines Specific Options
TRYTOCOMPACT Applicable only to the LTRA model (see 6.2.1). When specified,
the simulator tries to condense an LTRA transmission line’s past history of input
voltages and currents.
15.1.7 Precedence of option and .options commands
There are various ways to set the above mentioned options in Ngspice. If no option
or .options lines are set by the user, internal default values are given for each of the
simulator variables.
You may set options in the init files spinit or .spiceinit via the option command (see
17.5.55). The values given there will supersede the default values. If you set options via
the .options line in your input file, their values will supersede the default and init file
data. Finally, if you set options inside a .control ... .endc section, these values will
again supersede any simulator variables given so far.
15.2 Initial Conditions
15.2.1 .NODESET: Specify Initial Node Voltage Guesses
General form:
. nodeset v ( nodnum )= val v ( nodnum )= val ...

. nodeset all = val
Examples:
. nodeset v (12)=4.5 v (4)=2.23

. nodeset all =1.5
The .nodeset line helps the program find the DC or initial transient solution by making
a preliminary pass with the specified nodes held to the given voltages. The restrictions
are then released and the iteration continues to the true solution. The .nodeset line may
be necessary for convergence on bistable or astable circuits. .nodeset all=val sets all
starting node voltages (except for the ground node) to the same value. In general, the
.nodeset line should not be necessary.
15.2.2 .IC: Set Initial Conditions
General form:
. ic v ( nodnum )= val v ( nodnum )= val ...
Examples:
. ic v (11)=5 v (4)= -5 v (2)=2.2
The .ic line is for setting transient initial conditions. It has two different interpretations,
depending on whether the uic parameter is specified on the .tran control line, or not.
One should not confuse this line with the .nodeset line. The .nodeset line is only to
help DC convergence, and does not affect the final bias solution (except for multi-stable
circuits). The two indicated interpretations of this line are as follows:
1. When the uic parameter is specified on the .tran line, the node voltages specified
on the .ic control line are used to compute the capacitor, diode, BJT, JFET, and
MOSFET initial conditions. This is equivalent to specifying the ic=... parameter
on each device line, but is much more convenient. The ic=... parameter can still
be specified and takes precedence over the .ic values. Since no dc bias (initial
transient) solution is computed before the transient analysis, one should take care
to specify all dc source voltages on the .ic control line if they are to be used to
compute device initial conditions.
2. When the uic parameter is not specified on the .tran control line, the DC bias
(initial transient) solution is computed before the transient analysis. In this case,
the node voltages specified on the .ic control lines are forced to the desired initial
values during the bias solution. During transient analysis, the constraint on these
node voltages is removed. This is the preferred method since it allows Ngspice to
compute a consistent dc solution.
15.3. ANALYSES 309
15.3 Analyses
15.3.1 .AC: Small-Signal AC Analysis
General form:
. ac dec nd fstart fstop

. ac oct no fstart fstop
. ac lin np fstart fstop
Examples:
. ac dec 10 1 10 K
. ac dec 10 1 K 100 MEG
. ac lin 100 1 100 HZ
dec stands for decade variation, and nd is the number of points per decade. oct stands
for octave variation, and no is the number of points per octave. lin stands for linear
variation, and np is the number of points. fstart is the starting frequency, and fstop
is the final frequency. If this line is included in the input file, Ngspice performs an AC
analysis of the circuit over the specified frequency range. Note that in order for this
analysis to be meaningful, at least one independent source must have been specified with
an ac value. Typically it does not make much sense to specify more than one ac source.
If you do, the result will be a superposition of all sources and difficult to interpret.
Example:
Basic RC circuit
r 1 2 1.0
c 2 0 1.0
vin 1 0 dc 0 ac 1 $ <--- the ac source
. options noacct
. ac dec 10 .01 10
. plot ac vdb (2) xlog
. end
In this AC (or ’small signal’) analysis, all non-linear devices are linearized around their
actual DC operating point. All L and C devices get their imaginary value that depends
on the actual frequency step. Each output vector will be calculated relative to the input
voltage (current) given by the AC value (Vin equals 1 in the example above). The resulting
node voltages (and branch currents) are complex vectors. Therefore one has to be careful
using the plot command, specifically, one may use the variants of vxx(node) described
in Chapt. 15.6.2 like vdb(2) (see also the above example).
If one wants to simulate ac on a large linear array, the option noopac (15.1.3) may be
useful. Linear circuits are containing only linear device instances starting with letters r,
l, c, i, v, e, g, f, h, k. The instances e, g, f, h have to be the simple ones, as of chapt.
4.2, not the polynomial nor the behavioral variants. If the option noopac is set, ngspice
tests for the absence of any other devices. If successful, the often lengthy op calculation
is skipped, ac is started immediately. Considerable simulation time savings may result.
15.3.2 .DC: DC Transfer Function
General form:
. dc srcnam vstart vstop vincr [ src2 start2 stop2 incr2 ]
Examples:
. dc VIN 0.25 5.0 0.25

. dc VDS 0 10 .5 VGS 0 5 1
. dc VCE 0 10 .25 IB 0 10 u 1 u
. dc RLoad 1 k 2 k 100
. dc TEMP -15 75 5
The .dc line defines the dc transfer curve source and sweep limits (with capacitors open
and inductors shorted). srcnam is the name of an independent voltage or current source,
a resistor, or the circuit temperature. vstart, vstop, and vincr are the starting, final,
and incrementing values, respectively. The first example causes the value of the voltage
source VIN to be swept from 0.25 Volts to 5.0 Volts with steps of 0.25 Volt. A second
source (src2) may optionally be specified with its own associated sweep parameters. In
such a case the first source is swept over its own range for each value of the second source.
This option is useful for obtaining semiconductor device output characteristics. See the
example on transistor characterization (21.3).
15.3.3 .DISTO: Distortion Analysis
General form:
. disto dec nd fstart fstop < f2overf1 >

. disto oct no fstart fstop < f2overf1 >
. disto lin np fstart fstop < f2overf1 >
Examples:
. disto dec 10 1 kHz 100 MEG

. disto dec 10 1 kHz 100 MEG 0.9
The .disto line does a small-signal distortion analysis of the circuit. A multi-dimensional
Volterra series analysis is done using multi-dimensional Taylor series to represent the
nonlinearities at the operating point. Terms of up to third order are used in the series
expansions.
15.3. ANALYSES 311
If the optional parameter f2overf1 is not specified, .disto does a harmonic analysis - i.e.,
it analyses distortion in the circuit using only a single input frequency F1 , which is swept
as specified by arguments of the .disto command exactly as in the .ac command. Inputs
at this frequency may be present at more than one input source, and their magnitudes
and phases are specified by the arguments of the distof1 keyword in the input file lines
for the input sources (see the description for independent sources). (The arguments of
the distof2 keyword are not relevant in this case).
The analysis produces information about the AC values of all node voltages and branch
currents at the harmonic frequencies 2F1 and , vs. the input frequency F1 as it is swept. (A
value of 1 (as a complex distortion output) signifies cos(2π(2F1 )t) at 2F1 and cos(2π(3F1 )t)
at 3F1 , using the convention that 1 at the input fundamental frequency is equivalent
to cos(2πF1 t).) The distortion component desired (2F1 or 3F1 ) can be selected using
interactive or control commands in ngspice, and then printed or plotted. (Normally, one
is interested primarily in the magnitude of the harmonic components, so the magnitude
of the AC distortion value is looked at). It should be noted that these are the AC values
of the actual harmonic components, and are not equal to HD2 and HD3. To obtain HD2
and HD3, one must divide by the corresponding AC values at F1 , obtained from an .ac
line. This division can be done again using interactive or control commands.
If the optional f2overf1 parameter is specified, it should be a real number between (and
not equal to) 0.0 and 1.0; in this case, .disto does a spectral analysis. It considers
the circuit with sinusoidal inputs at two different frequencies F1 and F2 . F1 is swept
according to the .disto control line options exactly as in the .ac control line. F2 is
kept fixed at a single frequency as F1 sweeps - the value at which it is kept fixed is
equal to f2overf1 times fstart. Each independent source in the circuit may potentially
have two (superimposed) sinusoidal inputs for distortion, at the frequencies F1 and F2 .
The magnitude and phase of the F1 component are specified by the arguments of the
distof1 keyword in the source’s input line (see the description of independent sources);
the magnitude and phase of the F2 component are specified by the arguments of the
distof2 keyword. The analysis produces plots of all node voltages/branch currents at
the intermodulation product frequencies F1 + F2 , F1 − F2 , and (2F1 ) − F2 , vs the swept
frequency F1 . The IM product of interest may be selected using the setplot command,
and displayed with the print and plot commands. It is to be noted as in the harmonic
analysis case, the results are the actual AC voltages and currents at the intermodulation
frequencies, and need to be normalized with respect to .ac values to obtain the IM
parameters.
If the distof1 or distof2 keywords are missing from the description of an independent
source, then that source is assumed to have no input at the corresponding frequency. The
default values of the magnitude and phase are 1.0 and 0.0 respectively. The phase should
be specified in degrees.
It should be carefully noted that the number f2overf1 should ideally be an irrational
number, and that since this is not possible in practice, efforts should be made to keep
the denominator in its fractional representation as large as possible, certainly above 3,
for accurate results (i.e., if f2overf1 is represented as a fraction A/B , where A and B
are integers with no common factors, B should be as large as possible; note that A < B
because f2overf1 is constrained to be < 1). To illustrate why, consider the cases where
f2overf1 is 49/100 and 1/2. In a spectral analysis, the outputs produced are at F1 + F2 ,
F1 − F2 and 2F1 − F2 . In the latter case, F1 − F2 = F2 , so the result at the F1 − F2
component is erroneous because there is the strong fundamental F2 component at the

same frequency. Also, F1 + F2 = 2F1 − F2 in the latter case, and each result is erroneous
individually. This problem is not there in the case where f2overf1 = 49/100, because
F1 − F2 = 51/100 F1 <> 49/100 F1 = F2 . In this case, there are two very closely
spaced frequency components at F2 and F1 − F2 . One of the advantages of the Volterra
series technique is that it computes distortions at mix frequencies expressed symbolically
(i.e. nF1 + mF2 ), therefore one is able to obtain the strengths of distortion components
accurately even if the separation between them is very small, as opposed to transient
analysis for example. The disadvantage is of course that if two of the mix frequencies
coincide, the results are not merged together and presented (though this could presumably
be done as a postprocessing step). Currently, the interested user should keep track of the
mix frequencies himself or herself and add the distortions at coinciding mix frequencies
together should it be necessary.
Only a subset of the ngspice nonlinear device models supports distortion analysis. These
are
• Diodes (DIO),
• BJT,
• JFET (level 1),
• MOSFETs (levels 1, 2, 3, 9, and BSIM1),
• MESFET (level 1).
15.3.4 .NOISE: Noise Analysis
General form:
. noise v ( output <, ref >) src ( dec | lin | oct ) pts fstart fstop
+ < pts_per_summary >
Examples:
. noise v (5) VIN dec 10 1 kHz 100 MEG

. noise v (5 ,3) V1 oct 8 1.0 1.0 e6 1
The .noise line does a noise analysis of the circuit. output is the node at which the total
output noise is desired; if ref is specified, then the noise voltage v(output) - v(ref) is
calculated. By default, ref is assumed to be ground. src is the name of an independent
source to which input noise is referred. pts, fstart and fstop are .ac type parameters
that specify the frequency range over which plots are desired. pts_per_summary is an
optional integer; if specified, the noise contributions of each noise generator is produced
every pts_per_summary frequency points. The .noise control line produces two plots,
which can selected by setplot command:
15.3. ANALYSES 313
√ √
• one for the Voltage or Current Noise Spectral Density (in V / Hz or A/ Hz respective
the input is a voltage or current source) curves (e.g. after setplot noise1). There
are two vectors over frequency:
– √
onoise_spectrum: This is the output noise voltage or current divided by
Hz.
– inoise_spectrum: This the equivalent input noise = output noise divided by
the gain of the circuit.
• one for the Total Integrated Noise (in V or A) over the specified frequency range
(e.g. after setplot noise2). There are two vectors which are in reality scalars:
– onoise_total: This is the output noise voltage over the specified frequency
range
– inoise_total: This the equivalent input noise over the specified frequency
range = output noise divided by the gain of the circuit.
The units of all result vectors can be changed by using control variable sqrnoise:
• set sqrnoise: will deliver results in squared form, means the unit is V 2/Hz or
A2/Hz . This value refers more to the convenient Power Spectral Density.
Default setting of ngspice is unset sqrnoise, which delivers Voltage or Current Noise
Spectral Density. This is more practical from designers point of view.
The KLU matrix solver (15.1.1) is not compatible with noise simulation.
15.3.5 .OP: Operating Point Analysis

General form:
. op
Compute the DC operating point of the circuit with inductors shorted and capacitors
opened.
A DC solution can be difficult to find for some circuits, including those with floating nodes
or active devices that are non-conducting. After an attempt at an initial DC solution
(may be suppressed by .option noopiter), ngspice uses the following convergence aids,
in order, to try to obtain a DC solution:
1. gmin stepping (gminsteps option). Inserts small conductances across active devices.
• gminsteps = 0: No gmin
• gminsteps = 1: Two gmin stepping processes in series (default)
• gminsteps = 2: Original SPICE 3 gmin
2. source stepping (srcsteps option)

• srcsteps = 0: No source stepping

• srcsteps = 1: Gillespie source stepping (default)
• srcsteps = 2: Original SPICE 3 source stepping
3. transient operating point (optional)
DC analysis is complete as soon as one successful step is found, according to some con-
vergence criteria..
The default behaviour during gmin stepping is the following: Switch gmin to a start value
(1e-3), followed by a first trial of gmin stepping, using the true device gmin, then try
dynamic gmin stepping with diagonal parallel gmin elements. If variable dyngmin is set,
only dynamic gmin stepping is used.
Source stepping sets all supply voltages and currents to zero, then ramps them up dy-
namically to 100%.
The transient op calculation uses a transient simulation, with default parameters set by
ngspice (initial iteration, gmin and source stepping enabled, optran step size 10n, total
optran time 10u). The results of this transient simulation then are used as the operating
point for starting any other simulation (tran, ac, noise, pz etc.). No other data of this
transient op are stored anywhere.
General form:
optran ! noopiter gminsteps srcsteps tstep tstop supramp
Example 1:
optran 0 0 0 100 n 10 u 0
Example 1 changes the defaults to: no inital op iteration, no gmin stepping, no source
stepping, i.e. directly move to transient op with transient step and stop times given.
Flag supramp ins currently not used. The optran command may be put into one of the
initialization files .spiceinit or spinit. or into the .control section.
Example 2:
optran 1 1 1 100 n 10 u 0
Example 2 shows an optran command which restores the initial conditions.

Note: an operating point analysis is automatically performed prior to a transient analysis
(if the parameter uic is not selected) to determine the transient initial conditions, and
prior to an AC small-signal, Noise, and Pole-Zero analysis to determine the linearized,
small-signal models for nonlinear devices. These data are not stored, except for setting
the KEEPOPINFO variable 15.1.2, that prompts creating an OP plot in addition to the
TRAN, AC, Noise, or PZ plots.
15.3. ANALYSES 315
15.3.6 .PZ: Pole-Zero Analysis

General form:
. pz node1 node2 node3 node4 cur pol

. pz node1 node2 node3 node4 cur zer
. pz node1 node2 node3 node4 cur pz
. pz node1 node2 node3 node4 vol pol
. pz node1 node2 NODE3 node4 vol zer
. pz node1 node2 node3 node4 vol pz
Examples:
. pz 1 0 3 0 cur pol
. pz 2 3 5 0 vol zer
. pz 4 1 4 1 cur pz
cur stands for a transfer function of the type (output voltage)/(input current) while vol
stands for a transfer function of the type (output voltage)/(input voltage). pol stands
for pole analysis only, zer for zero analysis only and pz for both. This feature is provided
mainly because if there is a non-convergence in finding poles or zeros, then, at least the
other can be found. Finally, node1 and node2 are the two input nodes and node3 and
node4 are the two output nodes. Thus, there is complete freedom regarding the output
and input ports and the type of transfer function.
In interactive mode, the command syntax is the same except that the first field is pz
instead of .pz. To print the results, one should use the command print all.
15.3.7 .SENS: DC or Small-Signal AC Sensitivity Analysis

General form:
. SENS OUTVAR
. SENS OUTVAR AC DEC ND FSTART FSTOP
. SENS OUTVAR AC OCT NO FSTART FSTOP
. SENS OUTVAR AC LIN NP FSTART FSTOP
Examples:
. SENS V (1 , OUT )
. SENS V ( OUT ) AC DEC 10 100 100 k
. SENS I ( VTEST )
The sensitivity of OUTVAR to all non-zero device parameters is calculated when the
SENS analysis is specified. OUTVAR is a circuit variable (node voltage or voltage-source
branch current). The first form calculates sensitivity of the DC operating-point value
of OUTVAR. The second form calculates sensitivity of the AC values of OUTVAR. The
parameters listed for AC sensitivity are the same as in an AC analysis (see .AC above).
The output values are in dimensions of change in output per unit change of input (as
opposed to percent change in output or per percent change of input).
15.3.8 .SP S-Parameter Analysis

General form:
. sp dec nd fstart fstop < donoise >

. sp oct no fstart fstop < donoise >
. sp lin np fstart fstop < donoise >
Examples:
. sp dec 10 1 10 K
. sp dec 10 1 K 100 MEG 1
. sp lin 100 1 100 HZ
To prepare the independent voltage source VSRC please see 4.1.11.

SP Simulation Syntax is identical to .AC (15.3.1) except that you have one more optional
parameter donoise (0|1). SP does always linear S-Matrix simulation and, as outputs, it
gives
S Matrix (size nport x nport where nport is the count of RF ports) which is the Scattering
Parameters. It may be used to export Touchstone files (to be implemented yet)...
Y Matrix (size nport x nport where nport is the count of RF ports) which is the Admit-
tance Matrix
Z Matrix (size nport x nport where nport is the count of RF ports) which is the Impedance
Matrix
All S|Y|Z output are S_i_j where i and j are integer identifiers of the ports. They refer
to the portnum identifier of corresponding RF port of the VSRC (4.1.11).
If donoise = 0 SP simulation ends here.
If donoise = 1, SP simulation performs also SP Noise. In this case: you have one more out-
put which is the Noise Current Correlation Matrix: Cy_i_j Cy_i_j = <in(i), in*(j)="">
which is the correlation between equivalent input noise current at port i and equivalent
input noise current at port j. * stands for conjugate</in(i),>
When donoise = 1 and you have a two port networks, 4 more outputs are provided:
Rn input noise resistance (unnormalized)

NF (dB): noise figure of the 2-ports network
NFmin (dB): minimum noise figure
SOpt: optimum input reflection coefficient for noise
15.3. ANALYSES 317
15.3.9 .TF: Transfer Function Analysis

General form:
. tf outvar insrc
Examples:
. tf v (5 , 3) VIN
. tf i ( VLOAD ) VIN
The .tf line defines the small-signal output and input for the dc small-signal analysis.
outvar is the small signal output variable and insrc is the small-signal input source. If
this line is included, ngspice computes the dc small-signal value of the transfer function
(output/input), input resistance, and output resistance. For the first example, ngspice
would compute the ratio of V(5, 3) to VIN, the small-signal input resistance at VIN, and
the small signal output resistance measured across nodes 5 and 3.
15.3.10 .TRAN: Transient Analysis

General form:
. tran tstep tstop < tstart < tmax > > <uic >
Examples:
. tran 1 ns 100 ns
. tran 1 ns 1000 ns 500 ns
. tran 10 ns 1 us
tstep is the printing or plotting increment for line-printer output. For use with the
post-processor, tstep is the suggested computing increment. tstop is the final time, and
tstart is the initial time. If tstart is omitted, it is assumed to be zero. The transient
analysis always begins at time zero. In the interval [zero, tstart), the circuit is analyzed
(to reach a steady state), but no outputs are stored. In the interval [tstart, tstop], the
circuit is analyzed and outputs are stored. tmax is the maximum stepsize that ngspice
uses; for default, the program chooses either tstep or (tstop-tstart)/50.0, whichever is
smaller. tmax is useful when one wishes to guarantee a computing interval that is smaller
than the printer increment, tstep.
An initial transient operating point at time zero is calculated according to the following
procedure: all independent voltages and currents are applied with their time zero values,
all capacitances are opened, inductances are shorted, the non linear device equations are
solved iteratively.
uic (use initial conditions) is an optional keyword that indicates that the user does not
want ngspice to solve for the quiescent operating point before beginning the transient
analysis. If this keyword is specified, ngspice uses the values specified using IC=... on the
various elements as the initial transient condition and proceeds with the analysis. If the
.ic control line has been specified (see 15.2.2), then the node voltages on the .ic line are
used to compute the initial conditions for the devices. IC=... will take precedence over
the values given in the .ic control line. If neither IC=... nor the .ic control line is given
for a specific node, node voltage zero is assumed.
Look at the description on the .ic control line (15.2.2) for its interpretation when uic is
not specified.
15.3.11 Transient noise analysis (at low frequency)
In contrast to the analysis types described above, the transient noise simulation (noise
current or voltage versus time) is not implemented as a dot command, but is integrated
with the independent voltage source vsrc (isrc not yet available) (see 4.1.7) and used in
combination with the .tran transient analysis (15.3.10).
Transient noise analysis deals with noise currents or voltages added to your circuits as a
time dependent signal of randomly generated voltage excursion on top of a fixed dc voltage.
The sequence of voltage values has random amplitude, but equidistant time intervals,
selectable by the user (parameter NT). The resulting voltage waveform is differentiable
and thus does not require any modifications of the matrix solving algorithms.
White noise is generated by the ngspice random number generator, applying the Box-
Muller transform. Values are generated on the fly, each time when a breakpoint is hit.
The 1/f noise is generated with an algorithm provided by N. J. Kasdin (‘Discrete sim-
ulation of colored noise and stochastic processes and 1/f a power law noise generation’,
Proceedings of the IEEE, Volume 83, Issue 5, May 1995 Page(s):802–827). The noise
sequence (one for each voltage/current source with 1/f selected) is generated upon start
up of the simulator and stored for later use. The number of points is determined by the
total simulation time divided by NT, rounded up the the nearest power of 2. Each time a
breakpoint (n ⋆ N T , relevant to the noise signal) is hit, the next value is retrieved from
the sequence.
If you want a random, but reproducible sequence, you may select a seed value for the
random number generator by adding
setseed nn
to the spinit or .spiceinit file, nn being a positive integer number.
The transient noise analysis will allow the simulation of the three most important noise
sources. Thermal noise is described by the Gaussian white noise. Flicker noise (pink noise
or 1 over f noise) with an exponent between 0 and 2 is provided as well. Shot noise is
dependent on the current flowing through a device and may be simulated by applying a
non-linear source as demonstrated in the following example:
15.3. ANALYSES 319
Example:
* Shot noise test with B source , diode

* voltage on device ( diode , forward )
Vdev out 0 DC 0 PULSE (0.4 0.45 10 u )
* diode , forward direction , to be modeled with noise
D1 mess 0 DMOD
. model DMOD D IS =1 e -14 N =1
X1 0 mess out ishot
* device between 1 and 2
* new output terminals of device including noise : 1 and 3
. subckt ishot 1 2 3
* white noise source with rms 1 V
* 20000 sample points
VNG 0 11 DC 0 TRNOISE (1 1 n 0 0)
* measure the current i ( v1 )
V1 2 3 DC 0
* calculate the shot noise
* sqrt (2* current * q * bandwidth )
BI 1 3 I = sqrt (2* abs ( i ( v1 ))*1.6 e -19*1 e7 )* v (11)
. ends ishot
. tran 1 n 20 u
. control
run
plot ( -1)* i ( vdev )
. endc
. end
The selection of the delta time step (NT) is worth discussing. Gaussian white noise has un-
limited bandwidth and thus unlimited energy content. This is unrealistic. The bandwidth
of real noise is limited, but it is still called ‘White’ if it is the same level throughout the
frequency range of interest, e.g. the bandwidth of your system. Thus you may select NT
to be a factor of 10 smaller than the frequency limit of your circuit. A thorough analysis
is still needed to clarify the appropriate factor. The transient method is probably most
suited to circuits including switches, which are not amenable to the small signal .NOISE
analysis (Chapt. 15.3.4).
There is a price you have to pay for transient noise analysis: the number of required time
steps, and thus the simulation time, increases.
In addition to white and 1/f noise the independent voltage and current sources offer
a random telegraph signal (RTS) noise source, also known as burst noise or popcorn
noise, again for transient analysis. For each voltage (current) source offering RTS noise
an individual noise amplitude is required for input, as well as a mean capture time and a
mean emission time. The amplitude resembles the influence of a single trap on the current
or voltage. The capture and emission times emulate the filling and emptying of the trap,
typically following a Poisson process. They are generated from an random exponential
distribution with respective mean values given by the user. To simulate an ensemble of
traps, you may combine several current or voltage sources with different parameters.
All three sources (white, 1/f, and RTS) may be combined in a single command line.
RTS noise example:
* white noise , 1/ f noise , RTS noise
* voltage source
VRTS2 13 12 DC 0 trnoise (0 0 0 0 5 m 18 u 30 u )
VRTS3 11 0 DC 0 trnoise (0 0 0 0 10 m 20 u 40 u )
VALL 12 11 DC 0 trnoise (1 m 1 u 1.0 0.1 m 15 m 22 u 50 u )
VW1of 21 0 DC trnoise (1 m 1 u 1.0 0.1 m )
* current source
IRTS2 10 0 DC 0 trnoise (0 0 0 0 5 m 18 u 30 u )
IRTS3 10 0 DC 0 trnoise (0 0 0 0 10 m 20 u 40 u )
IALL 10 0 DC 0 trnoise (1 m 1 u 1.0 0.1 m 15 m 22 u 50 u )
R10 10 0 1
IW1of 9 0 DC trnoise (1 m 1 u 1.0 0.1 m )

Rall 9 0 1
* sample points
. tran 1 u 500 u
. control
run
plot v (13) v (21)
plot v (10) v (9)
. endc
. end
Some details on RTS noise modeling are available in a recent article [20], available here.
This transient noise feature is still experimental.
The following questions (among others) are to be solved:
• clarify the theoretical background
• noise limit of plain ngspice (numerical solver, fft etc.)
• time step (NT) selection
• calibration of noise spectral density
• how to generate noise from a transistor model
• application benefits and limits

15.3. ANALYSES 321
15.3.12 .PSS: Periodic Steady State Analysis
Experimental code, not yet made publicly available.
General form:
. pss gfreq tstab oscnob psspoints harms sciter steadycoeff <uic >
Examples:
. pss 150 200 e -3 2 1024 11 50 5e -3 uic

. pss 624 e6 1 u v_plus 1024 10 150 5e -3 uic
. pss 624 e6 500 n bout 1024 10 100 5e -3 uic
gfreq is guessed frequency of fundamental suggested by user. When performing transient

analysis the PSS algorithm tries to infer a new rough guess rgfreq on the fundamental.
If gfreq is out of ±10% with respect to rgfreq then gfreq is discarded.
tstab is stabilization time before the shooting begin to search for the PSS. It has to be
noticed that this parameter heavily influence the possibility to reach the PSS. Thus is a
good practice to ensure a circuit to have a right tstab, e.g. performing a separate TRAN
analysis before to run PSS analysis.
oscnob is the node or branch where the oscillation dynamic is expected. PSS analysis
will give a brief report of harmonic content at this node or branch.
psspoints is number of step in evaluating predicted period after convergence is reached.

It is useful only in Time Domain plots. However this number should be higher than 2
times the requested harms. Otherwise the PSS analysis will properly adjust it.
harms number of harmonics to be calculated as requested by the user.
sciter number of allowed shooting cycle iterations. Default is 50.
steady_coeff is the weighting coefficient for calculating the Global Convergence Error
(GCE), which is the reference value in order to infer is convergence is reached. The lower
steady_coeff is set, the higher the accuracy of predicted frequency can be reached but
at longer analysis time and sciter number. Default is 1e-3.
uic (use initial conditions) is an optional keyword that indicates that the user does not
want ngspice to solve for the quiescent operating point before beginning the transient
analysis. If this keyword is specified, ngspice uses the values specified using IC=... on
the various elements as the initial transient condition and proceeds with the analysis. If
the .ic control line has been specified, then the node voltages on the .ic line are used to
compute the initial conditions for the devices. Look at the description on the .ic control
line for its interpretation when uic is not specified.
15.4 Measurements after AC, DC and Transient Anal-

ysis
15.4.1 .meas(ure)
The .meas or .measure statement (and its equivalent meas command, see Chapt.
17.5.50) are used to analyze the output data of a tran, ac, or dc simulation. The command
is executed immediately after the simulation has finished.
15.4.2 batch versus interactive mode

.meas analysis may not be used in batch mode (-b command line option), if an output
file (rawfile) is given at the same time (-r rawfile command line option). In this batch
mode ngspice will write its simulation output data directly to the output file. The data
is not kept in memory, thus is no longer available for further analysis. This is done to
allow a very large output stream with only a relatively small memory usage. For .meas
to be active you need to run the batch mode with a .plot or .print command. A better
alternative may be to start ngspice in interactive mode.
If you need batch like operation, you may add a .control ... .endc section to the
input file:
Example:
* input file
...
. tran 1 ns 1000 ns
...
** * ******************************
. control
run
write outputfile data
. endc
** * ******************************
. end
and start ngspice in interactive mode, e.g. by running the command

ngspice inputfile .
.meas<ure> then prints its user-defined data analysis to the standard output. The anal-
ysis includes propagation, delay, rise time, fall time, peak-to-peak voltage, minimum or
maximum voltage, the integral or derivative over a specified period and several other user
defined values.
15.4.3 General remarks

The measure type {DC|AC|TRAN|SP} depends on the data that is to be evaluated, either
originating from a dc analysis, an ac analysis, or a transient simulation. The type SP to
15.4. MEASUREMENTS AFTER AC, DC AND TRANSIENT ANALYSIS 323
analyze a spectrum from the spec or fft commands is only available when executed in a
meas command, see 17.5.50.
result will be a vector containing the result of the measurement. trig_variable,
targ_variable, and out_variable are vectors stemming from the simulation, e.g. a
voltage vector v(out).
VAL=val expects a real number val. It may be as well a parameter delimited by ” or {}
expanding to a real number.
TD=td and AT=time expect a time value if measure type is tran. For ac and sp, AT
will be a frequency value, TD is ignored. For dc analysis, AT is a voltage (or current), TD
is ignored as well.
CROSS=# requires an integer number #. CROSS=LAST is possible as well. The same is
expected by RISE and FALL.
Frequency and time values may start at 0 and extend to positive real numbers. Voltage
(or current) inputs for the independent (scale) axis in a dc analysis may start or end at
arbitrary real valued numbers.
Please note that not all of the .measure commands have been implemented.
15.4.4 Input
In the following lines you will get some explanation on the .measure commands. A
simple simulation file with two sines of different frequencies may serve as an example.
The transient simulation delivers time as the independent variable and two voltages as
output (dependent variables).
Input file:
File : simple - meas - tran . sp

* Simple . measure examples
* transient simulation of two sine
* signals with different frequencies
vac1 1 0 DC 0 sin (0 1 1 k 0 0)
vac2 2 0 DC 0 sin (0 1.2 0.9 k 0 0)
. tran 10 u 5 m
*
. measure tran ... $ for the different inputs see
below !
*
. control
run
plot v (1) v (2)
. endc
. end
After displaying the general syntax of the .measure statement, some examples are posted,
referring to the input file given above.
15.4.5 Trig Targ
.measure according to general form 1 measures the difference in dc voltage, frequency or

time between two points selected from one or two output vectors. The current examples
all are using transient simulation. Measurements for tran analysis start after a delay
time td. If you run other examples with ac simulation or spectrum analysis, time may
be replaced by frequency, after a dc simulation the independent variable may become a
voltage or current.
General form 1:
. MEASURE { DC | AC | TRAN | SP } result TRIG trig_variable

VAL = val
+ < TD = td > < CROSS =# | CROSS = LAST > < RISE =# | RISE = LAST >
+ < FALL =# | FALL = LAST > < TRIG AT = time > TARG
targ_variable
+ VAL = val < TD = td > < CROSS =# | CROSS = LAST > < RISE =# |
+ RISE = LAST > < FALL =# | FALL = LAST > < TARG AT = time >
Measure statement example (for use in the input file given above):
.measure tran tdiff TRIG v(1) VAL=0.5 RISE=1 TARG v(1) VAL=0.5 RISE=2
measures the time difference between v(1) reaching 0.5 V for the first time on its first
rising slope (TRIG) versus reaching 0.5 V again on its second rising slope (TARG), i.e.
it measures the signal period.
Output:
tdiff = 1.000000e-003 targ= 1.083343e-003 trig= 8.334295e-005
Measure statement example:
.measure tran tdiff TRIG v(1) VAL=0.5 RISE=1 TARG v(1) VAL=0.5 RISE=3
measures the time difference between v(1) reaching 0.5 V for the first time on its rising
slope versus reaching 0.5 V on its rising slope for the third time (i.e. two periods).
Measure statement:
.measure tran tdiff TRIG v(1) VAL=0.5 RISE=1 TARG v(1) VAL=0.5 FALL=1
measures the time difference between v(1) reaching 0.5V for the first time on its rising
slope versus reaching 0.5 V on its first falling slope.
Measure statement:
.measure tran tdiff TRIG v(1) VAL=0 FALL=3 TARG v(2) VAL=0 FALL=3
measures the time difference between v(1) reaching 0V its third falling slope versus v(2)
reaching 0 V on its third falling slope.
Measure statement:
.measure tran tdiff TRIG v(1) VAL=-0.6 CROSS=1 TARG v(2) VAL=-0.8 CROSS=1
measures the time difference between v(1) crossing -0.6 V for the first time (any slope)
versus v(2) crossing -0.8 V for the first time (any slope).
Measure statement:
.measure tran tdiff TRIG AT=1m TARG v(2) VAL=-0.8 CROSS=3
measures the time difference between the time point 1ms versus the time when v(2) crosses
-0.8 V for the third time (any slope).
15.4.6 Find ... When
The FIND and WHEN functions allow measuring any dependent or independent time, fre-
quency, or dc parameter, when two signals cross each other or a signal crosses a given
value. Measurements start after a delay TD and may be restricted to a range between
FROM and TO.
General form 2:
. MEASURE { DC | AC | TRAN | SP } result WHEN out_variable = val

+ < TD = td > < FROM = val > < TO = val > < CROSS =# | CROSS = LAST >
+ < RISE =# | RISE = LAST > < FALL =# | FALL = LAST >
Measure statement:
.measure tran teval WHEN v(2)=0.7 CROSS=LAST
measures the time point when v(2) crosses 0.7 V for the last time (any slope).
General form 3:
. MEASURE { DC | AC | TRAN | SP } result

+ WHEN out_variable = out_variable2
+ < TD = td > < FROM = val > < TO = val > < CROSS =# | CROSS = LAST >
+ < RISE =# | RISE = LAST > < FALL =# | FALL = LAST >
Measure statement:
.measure tran teval WHEN v(2)=v(1) RISE=LAST
measures the time point when v(2) and v(1) are equal, v(2) rising for the last time.
General form 4:
. MEASURE { DC | AC | TRAN | SP } result FIND out_variable

+ WHEN out_variable2 = val < TD = td > < FROM = val > < TO = val >
+ < CROSS =# | CROSS = LAST > < RISE =# | RISE = LAST >
+ < FALL =# | FALL = LAST >
Measure statement:
.measure tran yeval FIND v(2) WHEN v(1)=-0.4 FALL=LAST
returns the dependent (y) variable drawn from v(2) at the time point when v(1) equals a
value of -0.4, v(1) falling for the last time.
General form 5:
. MEASURE { DC | AC | TRAN | SP } result FIND out_variable

+ WHEN out_variable2 = out_variable3 < TD = td >
+ < CROSS =# | CROSS = LAST >
+ < RISE =#| RISE = LAST > < FALL =#| FALL = LAST >
Measure statement:
.measure tran yeval FIND v(2) WHEN v(1)=v(3) FALL=2
returns the dependent (y) variable drawn from v(2) at the time point when v(1) crosses
v(3), v(1) falling for the second time.
General form 6:
. MEASURE { DC | AC | TRAN | SP } result FIND out_variable AT =

val
Measure statement:
.measure tran yeval FIND v(2) AT=2m
returns the dependent (y) variable drawn from v(2) at the time point 2 ms (given by
AT=time).
15.4.7 AVG|MIN|MAX|PP|RMS|MIN_AT|MAX_AT
General form 7:
. MEASURE { DC | AC | TRAN | SP } result

+ { AVG | MIN | MAX | PP | RMS | MIN_AT | MAX_AT }
+ out_variable < TD = td > < FROM = val > < TO = val >
Measure statements:
.measure tran ymax MAX v(2) from=2m to=3m
returns the maximum value of v(2) inside the time interval between 2 ms and 3 ms.
.measure tran tymax MAX_AT v(2) from=2m to=3m
returns the time point of the maximum value of v(2) inside the time interval between 2
ms and 3 ms.
.measure tran ypp PP v(1) from=2m to=4m
returns the peak to peak value of v(1) inside the time interval between 2 ms and 4 ms.
.measure tran yrms RMS v(1) from=2m to=4m
returns the root mean square value of v(1) inside the time interval between 2 ms and 4
ms.
.measure tran yavg AVG v(1) from=2m to=4m
returns the average value of v(1) inside the time interval between 2 ms and 4 ms.
15.4.8 Integ
General form 8:
. MEASURE { DC | AC | TRAN | SP } result INTEG < RAL >

out_variable
+ < TD = td > < FROM = val > < TO = val >
Measure statement:
.measure tran yint INTEG v(2) from=2m to=3m
returns the area under v(2) inside the time interval between 2 ms and 3 ms.
15.4.9 param
General form 9:
. MEASURE { DC | AC | TRAN | SP } result param = ’ expression ’
Measure statement:
.param fval=5
.measure tran yadd param=’fval + 7’
will evaluate the given expression fval + 7 and return the value 12.
’Expression’ is evaluated according to the rules given in Chapt. 2.9.5 during start up
of ngspice. It may contain parameters defined with the .param statement. It may also
contain parameters resulting from preceding .meas statements.
.param vout_diff=50u
...
.measure tran tdiff TRIG AT=1m TARG v(2) VAL=-0.8 CROSS=3
.meas tran bw_chk param=’(tdiff < vout_diff) ? 1 : 0’
will evaluate the given ternary function and return the value 1 in bw_chk, if tdiff mea-
sured is smaller than parameter vout_diff.
The expression may not contain vectors like v(10), e.g. anything resulting directly from
a simulation. This may be handled with the following .meas command option.
15.4.10 par(’expression’ )
The par(’expression’) option (15.6.6) allows the use of algebraic expressions on the
.measure lines. Every out_variable may be replaced by par(’expression’) using the gen-
eral forms 1. . . 9 described above. Internally par(’expression’) is substituted by a vector
according to the rules of the B source (Chapt. 5.1). A typical example of the general
form is shown below:
General form 10:
. MEASURE { DC | TRAN | AC | SP } result

+ FIND par ( ’ expression ’) AT = val
The measure statement
.measure tran vtest find par(’(v(2)*v(1))’) AT=2.3m
returns the product of the two voltages at time point 2.3 ms.
Note that a B-source, and therefore the par(’...’) feature, operates on values of type
complex in AC analysis mode.
Both param and par are not available for the meas command (17.5.50) inside of a .control
section, where meas with par or param may be replaced by let (17.5.45).
15.4.11 Deriv
General form:
. MEASURE { DC | AC | TRAN | SP } result DERIV < ATIVE >

out_variable
+ AT = val

out_variable
+ WHEN out_variable2 = val < TD = td >
+ < CROSS =# | CROSS = LAST > < RISE =#| RISE = LAST >
+ < FALL =#| FALL = LAST >

out_variable
+ WHEN out_variable2 = out_variable3
+ < TD = td > < CROSS =# | CROSS = LAST >
+ < RISE =#| RISE = LAST > < FALL =#| FALL = LAST >
15.4.12 More examples
Some other examples, also showing the use of parameters, are given below. Corresponding
demonstration input files are distributed with ngspice in folder /examples/measure.
15.5. SAFE OPERATING AREA (SOA) WARNING MESSAGES 329
Other examples:
. meas tran inv_delay2 trig v ( in ) val = ’ vp /2 ’ td =1 n

fall =1
+ targ v ( out ) val = ’ vp /2 ’ rise =1
. meas tran test_data1 trig AT = 1 n targ v ( out )
+ val = ’ vp /2 ’ rise =3
. meas tran out_slew trig v ( out ) val = ’0.2* vp ’ rise =2
+ targ v ( out ) val = ’0.8* vp ’ rise =2
. meas tran delay_chk param = ’( inv_delay < 100 ps ) ? 1 :
0’
. meas tran skew when v ( out ) =0.6
. meas tran skew2 when v ( out ) = skew_meas
. meas tran skew3 when v ( out ) = skew_meas fall =2
. meas tran skew4 when v ( out ) = skew_meas fall = LAST
. meas tran skew5 FIND v ( out ) AT =2 n
. meas tran v0_min min i ( v0 )
+ from = ’ dfall ’ to = ’ dfall + period ’
. meas tran v0_avg avg i ( v0 )
. meas tran v0_integ integ i ( v0 )
. meas tran v0_rms rms i ( v0 )
. meas dc is_at FIND i ( vs ) AT =1
. meas dc is_max max i ( vs ) from =0 to =3.5
. meas dc vds_at when i ( vs ) =0.01
. meas ac vout_at FIND v ( out ) AT =1 MEG
. meas ac vout_atd FIND vdb ( out ) AT =1 MEG
. meas ac vout_max max v ( out ) from =1 k to =10 MEG
. meas ac freq_at when v ( out ) =0.1
. meas ac vout_diff trig v ( out ) val =0.1 rise =1 targ v (
out )
+ val =0.1 fall =1
. meas ac fixed_diff trig AT = 10 k targ v ( out )
+ val =0.1 rise =1
. meas ac vout_avg avg v ( out ) from =10 k to =1 MEG
. meas ac vout_integ integ v ( out ) from =20 k to =500 k
. meas ac freq_at2 when v ( out ) =0.1 fall = LAST
. meas ac bw_chk param = ’( vout_diff < 100 k ) ? 1 : 0 ’
. meas ac vout_rms rms v ( out ) from =10 to =1 G
15.5 Safe Operating Area (SOA) warning messages

By setting .option warn=1, the Safe Operation Area check algorithm is enabled. In
this case for .op, .dc and .tran analysis warning messages are issued if the branch
voltages of devices (Resistors, Capacitors, Diodes, BJTs and MOSFETs), or the currents
and dissipated power (Diodes, and BJTs), or the resulting temperature (Diodes) exceed
limits that are specified by model parameters. All these parameters are positive with
default value of infinity. For the bipolar VBIC model (15.5.3.3) .option warn=2 will add
additional operating point info
The check is executed after Newton-Raphson iteration is finished i.e. in transient analysis
in each time step. The user can specify an additional .option maxwarns (default: 5) to
limit the count of messages.
The output goes on default to stdout or alternatively to a file specified by command line
option --soa-log=filename.
To achive SOA checking, add some or all of these parameters with suitable limit values
to the .model line of the respective device.
15.5.1 Resistor and Capacitor SOA model parameters

1. Bv_max: If |Vr| or |Vc| exceed Bv_max, SOA warning is issued.
15.5.2 Diode SOA model parameters

1. Bv_max: If |Vj| exceeds Bv_max, SOA warning is issued.
2. Fv_max: If |Vf| exceeds Fv_max, SOA warning is issued.
3. Id_max: If |Id| exceeds Id_max, SOA warning is issued.
4. Pd_max: If power exceeds Pd_max, SOA warning is issued.
5. Te_max: If temperature exceeds Te_max, SOA warning is issued.
6. rth0: Thermal resistance between junction and ambient.
7. tnom: Nominal temperature where all parameters have been measured at.
Three SOA modes are available. All modes check for Bv_max, Vf_max, and Id_max.
If self-heating (7.3) is switched on, and Te_max, tnom and rth0 are given, then a der-
ating for the maximam allowed power dissipation is calculated, and power and current
temperature are checked against their max. allowed values.
pdmax = pdmaxmod − (tempcurr − tnom)/rth0
If self-heating is switched off, and rth0 and tnom are given, then a static max. power
derating is calculated, taking the device temperature (set by its default value 27 °C, or the
global .temp value, or the device specific instance parameter temp) into account. The
reference temperature is tnom.
pdmax = pdmaxmod − (temp − tnom)/rth0
If rth0 or tnom are not given, no derating is calculated, the power disspation is simply
checked against Pd_max.
15.5. SAFE OPERATING AREA (SOA) WARNING MESSAGES 331
15.5.3 BJT SOA model parameters
15.5.3.1 Gummel-Poon (levels 1 and 2)
Bipolar device models level 1 and 2 are supported with all the SOA parameters named
below.
1. Vbe_max: If |Vbe| exceeds Vbe_max, SOA warning is issued.
2. Vbc_max: If |Vbc| exceeds Vbc_max, SOA warning is issued.
3. Vce_max: If |Vce| exceeds Vce_max, SOA warning is issued.
4. Vcs_max: If |Vcs| exceeds Vcs_max, SOA warning is issued.
5. Ic_max: If |Ic| exceeds Ic_max, SOA warning is issued.
6. Ib_max: If |Ib| exceeds Ib_max, SOA warning is issued.
7. Pd_max: If power exceeds Pd_max, SOA warning is issued.
8. Te_max: If temperature exceeds Te_max, SOA warning is issued.
9. rth0: Thermal resistance between junction and ambient.
10. tnom: Nominal temperature where all parameters have been measured at.
Two SOA modes are available (self-heating is not yet modeled in bipolar level 1 and 2).
All modes check for Vbe_max, Vbc_max, Vce_max, Vcs_max, Ic_max and Ib_max.
If rth0 and tnom are given, then a static max. power derating is calculated, taking the
device temperature (set by its default value 27 °C, or the global .temp value, or the device
specific instance parameter temp) into account. The reference temperature is tnom.
pdmax = pdmaxmod − (temp − tnom)/rth0
If rth0 or tnom are not given, no derating is calculated, the power disspation is simply
checked against Pd_max.
Te_max is not (yet) used.
15.5.3.2 HICUM (level 8)
HICUM2 currently aknowledges the following voltage parameters:

15.5.3.3 VBIC (levels 4 and 9)
VBIC aknowledges the following parameters:
As an alternative to the above listed parameters bvbe, bvbc, bvce, and bvsub may be
used.
If .option warn=2 is set, the following parameters (defaults are set to 0.2 V) may be
used to determine the current operation point of the device.
1. vbefwd B-E forward voltage.
2. vbcfwd B-C forward voltage.
The following criteria are used:

op conditions
off Vbe <= vbefwd and Vbc <= vbcfwd
saturation Vbe > vbefwd and Vbc > vbcfwd
forward Vbe > vbefwd and Vbc <= vbcfwd
reverse Vbe <= vbefwd and Vbc > vbcfwd
Substrate leakage due to forward conduction of the collector-substrate diode may be
detected using:
1. vsubfwd Substrate junction forward voltage.
15.5.4 MOS SOA model parameters
1. Vgs_max: If |Vgs| exceeds Vgs_max, SOA warning is issued.
2. Vgd_max: If |Vgd| exceeds Vgd_max, SOA warning is issued.
3. Vgb_max: If |Vgb| exceeds Vgb_max, SOA warning is issued.
4. Vds_max: If |Vds| exceeds Vds_max, SOA warning is issued.
5. Vbs_max: If |Vbs| exceeds Vbs_max, SOA warning is issued.
6. Vbd_max: If |Vbd| exceeds Vbd_max, SOA warning is issued.

15.6. BATCH OUTPUT 333
15.5.5 VDMOS SOA model parameters

1. Vgs_max: If |Vgs| exceeds Vgs_max, SOA warning is issued.
2. Vgd_max: If |Vgd| exceeds Vgd_max, SOA warning is issued.
3. Vds_max: If |Vds| exceeds Vds_max, SOA warning is issued.
4. Vgsr_max: If |Vgsr| exceeds Vgsr_max, SOA warning is issued.
5. Vgdr_max: If |Vgdr| exceeds Vgdr_max, SOA warning is issued.
15.6 Batch Output

The following commands .print (15.6.2), .plot (15.6.3) and .four (15.6.4) are valid
only if ngspice is started in batch mode (see 16.4.1), whereas .save and the equivalent
.probe are aknowledged in all operating modes.
If you start ngspice in batch mode using the -b command line option, the outputs of
.print, .plot, and .four are printed to the console output. You may use the output
redirection of your shell to direct this printout into a file (not available with MS Windows
GUI). As an alternative, you may extend the ngspice command by specifying an output
file:
ngspice -b -o output.log input.cir
If you however add the command line option -r to create a rawfile, .print and .plot
are ignored. If you want to involve the graphics plot output of ngspice, use the control
mode (16.4.3) instead of the -b batch mode option.
15.6.1 .SAVE: Name vector(s) to be saved in raw file

General form:
. save vector vector vector ...
Examples:
. save i ( vin ) node1 v ( node2 )

. save @m1 [ id ] vsource # branch
. save all @m2 [ vdsat ]
The vectors listed on the .SAVE line are recorded in the rawfile for use later with ngspice.
The standard vector names are accepted. Node voltages may be saved by giving the
nodename or v(nodename). Currents through an independent voltage source are given by
i(sourcename) or sourcename#branch. Internal device data are accepted as @dev[param].
If no .SAVE line is given, then the default set of vectors is saved (node voltages and voltage
source branch currents). If .SAVE lines are given, only those vectors specified are saved.
For more discussion on internal device data, e.g. @m1[id], see Appendix, Chapt. 31.1. If
you want to save internal data in addition to the default vector set, add the parameter
all to the additional vectors to be saved. If the command .save vm(out) is given, and
you store the data in a rawfile, only the original data v(out) are stored. The request for
storing the magnitude is ignored, because this may be added later during rawfile data
evaluation with ngspice. See also the section on the interactive command interpreter
(Chapt. 17.5) for information on how to use the rawfile.
15.6.2 .PRINT Lines

General form:
. print prtype ov1 < ov2 ... ov8 >
Examples:
. print tran v (4) i ( vin )

. print dc v (2) i ( vsrc ) v (23 , 17)
. print ac vm (4 , 2) vr (7) vp (8 , 3)
The .print line defines the contents of a tabular listing of one to eight output variables.
prtype is the type of the analysis (DC, AC, TRAN, NOISE, or DISTO) for which the specified
outputs are desired. The form for voltage or current output variables is the same as given
in the previous section for the print command; Spice2 restricts the output variable to
the following forms (though this restriction is not enforced by ngspice):
V(N1<,N2>) specifies the voltage difference between nodes N1 and
N2. If N2 (and the preceding comma) is omitted,
ground (0) is assumed. See the print command in the
previous section for more details. For compatibility
with SPICE2, the following five additional values can
be accessed for the ac analysis by replacing the ‘V’ in
V(N1,N2) with:
VR Real part
VI Imaginary part
VM Magnitude
VP Phase
VDB 20log10(magnitude)
I(VXXXXXXX) specifies the current flowing in the independent voltage
source named VXXXXXXX. Positive current flows
from the positive node, through the source, to the
negative node. (Not yet implemented: For the ac
analysis, the corresponding replacements for the letter
I may be made in the same way as described for
voltage outputs.)
Output variables for the noise and distortion analyses have a different general form from
that of the other analyses. There is no limit on the number of .print lines for each type
of analysis. The par(’expression’) option (15.6.6) allows the use of algebraic expressions
in the .print lines. .width (15.6.7) selects the maximum number of characters per line.
15.6.3 .PLOT Lines

.plot creates a printer plot output.
General form:
. plot pltype ov1 <( plo1 , phi1 ) > < ov2 <( plo2 , phi2 ) > ... ov8 >
Examples:
. plot dc v (4) v (5) v (1)

. plot tran v (17 , 5) (2 , 5) i ( vin ) v (17) (1 , 9)
. plot ac vm (5) vm (31 , 24) vdb (5) vp (5)
. plot disto hd2 hd3 ( R ) sim2
. plot tran v (5 , 3) v (4) (0 , 5) v (7) (0 , 10)
The .plot line defines the contents of one plot of from one to eight output variables.
pltype is the type of analysis (DC, AC, TRAN, NOISE, or DISTO) for which the specified
outputs are desired. The syntax for the ovi is identical to that for the .print line and
for the plot command in the interactive mode.
The overlap of two or more traces on any plot is indicated by the letter ‘X’. When more
than one output variable appears on the same plot, the first variable specified is printed
as well as plotted. If a printout of all variables is desired, then a companion .print line
should be included. There is no limit on the number of .plot lines specified for each type
of analysis. The par(’expression’) option (15.6.6) allows the use of algebraic expressions
in the .plot lines.
15.6.4 .FOUR: Fourier Analysis of Transient Analysis Output

General form:
. four freq ov1 < ov2 ov3 ... >
Examples:
. four 100 K v (5)
The .four (or Fourier) line controls whether ngspice performs a Fourier analysis as a
part of the transient analysis. freq is the fundamental frequency, and ov1 is the desired
vector to be analyzed. The Fourier analysis is performed over the interval <TSTOP-period,
TSTOP>, where TSTOP is the final time specified for the transient analysis, and period is
one period of the fundamental frequency. The dc component and the first nine harmonics
are determined. For maximum accuracy, TMAX (see the .tran line) should be set to
period/100.0 (or less for very high-Q circuits). The par(’expression’) option (15.6.6)
allows the use of algebraic expressions in the .four lines.
As .four is available only when ngspice is executed in batch mode (16.4.1), and no rawfile
selected, you may consider the spec (17.5.87) or fft (17.5.33) commands, when using
ngspice in .control mode (with a .control section, 16.4.3).
15.6.5 .PROBE: Save device node currents, device power dissi-

pation, or differential voltages between arbitrary nodes
Command .probe enables current measurement at user specified device nodes, as well as
(differential) voltage measurements between device nodes.
15.6.5.1 Current measurement
Current measurement at a device node is achieved by automatically placing a Zero volt

voltage source (VSRC, 4.1) between the selected (or all) device node and the net attached
to that node. The positive pole of the VSRC is pointing out towards the net, the negative
pole towards the device. The resulting output vectors are using the xx#branch notation
(see examples below). Only top level devices are accessible, so device inside of subcircuits
are not considered.
Besides standard devices you may also measure currents at X instance lines (subcircuit
calls). If the subcircuit definition (.subckt line) uses named nodes, these are used instead
of node numbers (see device u1 in the example below).
Be careful when .probe alli is given, because the many output vectors generated au-
tomatically may require a large amount of memory to store all the current measurement
vectors.
General form for current measurements on all devices:
. probe alli
General form for current measurements on a 2- and multi-terminal device:
. probe I ( device )
General form for current measurements on a multi-terminal device (one command per
terminal):
. probe I ( device , node )
Examples:
* measure current at every node of each device in the circuit

. probe < alli >
* measure current at node 1 of a two - terminal device

. probe I ( R1 )
* measure current at all nodes of a subcircuit invocation

. probe I ( XU1 )
* measure current at node 3 of a multi - terminal device M4

. probe I ( MQ4 ,3)
Resulting output vectors:
r1 # branch
mq4 : s # branch
Resulting output vectors for .probe all (excerpt only, example file 555-timer-2.cir):
...
ra # branch : current , real , 14579 long
rb # branch : current , real , 14579 long
rl # branch : current , real , 14579 long
time : time , real , 14579 long [ default scale ]
xu1 : cont # branch : current , real , 14579 long
xu1 : disc # branch : current , real , 14579 long
xu1 : gnd # branch : current , real , 14579 long
xu1 : out # branch : current , real , 14579 long
xu1 : reset # branch : current , real , 14579 long
xu1 : thres # branch : current , real , 14579 long
xu1 : trig # branch : current , real , 14579 long
xu1 : vcc # branch : current , real , 14579 long
xu2 :1# branch : current , real , 14579 long
xu2 :19# branch : current , real , 14579 long
...
Compared to the approach using command .options savecurrents the resulting vectors
from a .probe command are available for every simulation type including AC simulation.
A slight disadvantage may be that new nodes are added to the instance matrix, increasing
simulation time (typically a little bit only).
15.6.5.2 (Differential) voltage measurement
Differential voltage measurements are achieved by placing a voltage controlled voltage

source (VCVS, E device) with its two inputs connected to the nodes specified by the user
and gain 1. The output is then saved in a vector with a leading vd_ in its name.
General form for (differential) voltage measurements:
. probe v ( node1 )
. probe vd ( device : node1 : node2 )
. probe vd ( device1 : node1 , device2 : node2 )
device, device1, and device2 are device names (first token in an instance line). node1,
node2 are either numbers (according to the node sequence in the instance line, e.g. 1,
2, 3, ...), or are node names of known devices (d, g, s, b for MOS of JFET, c, b, e for
bipolar.
Examples:
* voltage at node named nR1

. probe v ( nR1 )
* voltage across a two - terminal device named R1

. probe vd ( R1 )
* voltage at instance node 1 of device m4

. probe vd ( m4 :1:0)
* voltage between nodes 1 and 3 of device m4

. probe vd ( m4 :1:3)
* voltage between node 1 of device m4 and node 3 of device m5

. probe vd ( m4 :1 , m5 :3)
* m4 , m5 are MOS devices , so the following is equivalent :

. probe vd ( m4 :d , m5 : s )
nR1
vd_R1
vd_m4 : d :0
vd_m4 : d : s
vd_m3 : d_m5 : s
15.6.5.3 Measurement of power dissipation in a device
A power consumption measurement of a device with n nodes consists of two steps: all
n device node currents i1, i2, ... , in are measured (see15.6.5.1). Then all node
voltages v1 ... vn are measured. A common virtual star point vref is calculated as
the mean of all n node voltages. Power is the sum of the products of each node current
times its node voltage minus vref.
P = i1*(v1-vref) + i2*(v2-vref) +...+ in*(vn-vref)

General form for power measurements:
. probe p ( device )
Examples:
* power dissipation of a subcircuit device

. probe p ( XU1 )
* power dissipation in a MOS transistor

. probe p ( MQ1 )
xu1 : power
mq1 : power
All new items are added to the list of vectors named by .SAVE (see 15.6.1). If .save is not
given, only the newly generated .PROBE vectors are saved.
15.6.6 par(’expression’): Algebraic expressions for output

General form:
par ( ’ expression ’)
output = par ( ’ expression ’) $ not in . measure ac
Examples:
. four 1001 sq1 = par ( ’ v (1)* v (1) ’)

. measure tran vtest find par ( ’( v (2)* v (1)) ’) AT =2.3 m
. print tran output = par ( ’ v (1)/ v (2) ’) v (1) v (2)
. plot dc v (1) diff = par ( ’( v (4) - v (2))/0.01 ’) out222
With the output lines .four, .plot, .print, .save and in .measure evaluation, it
is possible to add algebraic expressions for output, in addition to vectors. All of these
output lines accept par(’expression’), where expression is any expression valid for a B
source (see Chapt. 5.1). Thus expression may contain predefined functions, numerical val-
ues, constants, simulator output like v(n1) or i(vdb), parameters predefined by a .param
statement, and the variables hertz, temper, and time. Note that a B-source, and there-
fore the par(’...’) feature, operates on values of type complex in AC analysis mode.
Internally the expression is replaced by a generated voltage node that is the output of a B
source, one node, and the B source implementing par(’...’). Several par(’...’) are allowed
15.7. MEASURING CURRENT THROUGH DEVICE TERMINALS 341
in each line, up to 99 per input file. The internal nodes are named pa_00 to pa_99. An
error will occur if the input file contains any of these reserved node names.
In .four, .plot, .print, .save, but not in .measure, an alternative syntax
output=par(’expression’) is possible. par(’expression’) may be used as described above.
output is the name of the new node to replace the expression. So output has to be unique
and a valid node name.
The syntax of output=par(expression) is strict: no spaces are allowed between par and
(’or between ( and ’. Also,(’ and ’) both are required. There is not much error
checking on your input, so if there is a typo, for example, an error may pop up at an
unexpected place.
15.6.7 .width
Set the width of a print-out or plot with the following card:
.with out = 256
Parameter out yields the maximum number of characters plotted in a row, if printing in
columns or an ASCII-plot is selected.
15.7 Measuring current through device terminals
15.7.1 Using the .probe command

Device currents (discrete devices or subcircuits) may be measured by the .probe command
(15.6.5). Voltage sources for measurements are placed in series to the devices nodes
specified by the user. For details please see (15.6.5).
15.7.2 Adding a voltage source in series

The ngspice matrix solver determines node voltages and currents through independent
voltage sources. So to measure the currents through a resistor, you may add a voltage
source in series with dc voltage 0.
Current measurement with series voltage source
* measure current through R1

V1 1 0 1
R1 1 0 5
R2 1 0 10
* will become
V1 1 0 1
R1 1 11 5
Vmeas 11 0 dc 0
R2 1 0 10
and the current is available as

vmeas#branch
after simulation.
15.7.3 Using option ’savecurrents’

Current measurement by reading internal current data
* measure current through R1 and R2

V1 1 0 1
R1 1 0 5
R2 1 0 10
. options savecurrents
The option savecurrents will add .save lines (15.6.1) like

.save @r1[i]
.save @r2[i]
to your input file information read during circuit parsing. These newly created vectors
contain the terminal currents of the devices R1 and R2.
You will find information of the nomenclature in Chapt. 31, also how to plot these vectors.
The following devices are supported: M, J, Q, D, R, C, L, B, F, G, W, S, I (see 2.1.3).
For MOSFETdevices only a subset of MOS1 to MOS9 current parameters are included
per default (but see options below). Devices in subcircuits are supported as well. The
advantage of the data obtained by .options savecurrents is that no extra nodes are
required, because the data are retrieved from internal nodes already existing.
This option however cannot be used in AC simulations, because complex data are not
supported. Vectors thus created will be empty after an AC simulation. So for AC you
might use one of the two methods (.probe or series voltage source) as previously described.
Be careful when choosing savecurrents in larger circuits, because 1 to 4 additional
output vectors are created per device and this may consume lots of memory.
Also note that the data thus retrieved may be delayed by on time step after a transient
simulation.
For MOS1, BSIM3 and BSIM4 three special options are available, listing all currents as
described in chapters 31.6.1, 31.6.8 and 31.6.9 of the ngspice manual:
Current measurement for MOS transistors with BSIM3 or BSIM4 models:
* measure all currents of MOS1 , BSIM3 and BSIM4 transistors

. options savecurrents_mos1
. options savecurrents_bsim3
. options savecurrents_bsim4
Chapter 16
Starting ngspice
16.1 Introduction
Ngspice consists of the simulator and a front-end for data analysis and plotting. Input to
the simulator is a netlist file, including commands for circuit analysis and output control.
Interactive ngspice can plot data from a simulation on a PC or a workstation display.
The usual way to run ngspice is by a console command, passing options and at least one
netlist file as a parameter. Multiple netlists are concatenated and treated as one, except
when the first file is a pure script with parameters (17.8).
Ngspice on Linux (and OSs like MacOS, Cygwin, BSD, Solaris ...) uses the X Window
System for plotting (see Chapt. 18.3) if the environment variable DISPLAY is available.
(An X11 server must first be installed on MacOS.) Otherwise, a console mode (non-
graphical) interface is used. If you are using X on a workstation, the DISPLAY variable
should already be set; if you want to display graphics on a system different from the one
you are running ngspice or ngutmeg on, DISPLAY should be of the form machine:0.0. See
the appropriate documentation on the X Window System for more details.
The MS Windows GUI version of ngspice has a native graphics interface (see Chapt.
18.1).
The front-end may be run as a separate ‘stand-alone’ program under the name ngnutmeg.
ngnutmeg is a subset of ngspice dedicated to data evaluation, still optionally compilable
(Linux, Mingw) for historical reasons. Ngnutmeg will read in the ‘raw’ data output file
created by ngspice -r or by the write command during an interactive ngspice session.
16.2 Where to obtain ngspice

The actual distribution of ngspice may be downloaded from the ngspice download web
page. The installation for Linux or MS Windows is described in the file INSTALL to be
found in the top level directory. You may also have a look at Chapt. 32 of this manual
for compiling instructions.
If you want to check out the source code that is actually under development, you may
have a look at the ngspice source code repository, which is stored using the Git Source
343
344 CHAPTER 16. STARTING NGSPICE
Code Management (SCM) tool. The Git repository may be browsed on the Git web
page, also useful for downloading individual files. You may however download (or clone)
the complete repository including all source code trees from the console window (Linux,
CYGWIN or MSYS/MINGW) by issuing the command (in a single line)
git clone git://git.code.sf.net/p/ngspice/ngspice
You need to have Git installed, which is available for all three OSs. The whole source
tree is then available in <current directory>/ngspice. Compilation and local installation
is again described in INSTALL (or Chapt. 32). If you later want to update your files
and download the recent changes from SourceForge into your local repository, cd into the
ngspice directory and just type
git pull
git pull will not overwrite modified files in your working directory. To drop your local
changes first, you can run
git reset --hard
To learn more about git, which can be both powerful and difficult to master, please consult
http://git-scm.com/, especially: http://git-scm.com/documentation, which has pointers
to documentation and tutorials.
16.3 Command line options for starting ngspice
Command Synopsis:
ngspice [ -o logfile ] [ -r rawfile ] [ -b ] [ -i ] [ input files ]
The oudated, optional ngnutmeg may be called by
Command Synopsis:
ngnutmeg [ - ] [ datafile ... ]
Where data file is the standard ngspice rawfile.
Options are shown below.

16.3. COMMAND LINE OPTIONS FOR STARTING NGSPICE 345
Option Long option Meaning

- Don’t try to load the default data file ("rawspice.raw")
if no other files are given (ngnutmeg only).
-n --no-spiceinit Don’t try to source the file .spiceinit upon start-up.
Normally ngspice and ngnutmeg try to find the file in
the current directory, and if it is not found then in the
user’s home directory (obsolete).
-t TERM --terminal=TERM The program is being run on a terminal with mfb
name term (obsolete).
-b --batch Run in batch mode. Ngspice reads the default input
source (e.g. keyboard) or reads the given input file and
performs the analyses specified; output is either
Spice2-like line-printer plots ("ascii plots") or a ngspice
rawfile. See the following section for details. Note that
if the input source is not a terminal (e.g. using the IO
redirection notation of "<") ngspice defaults to batch
mode (-i overrides). This option is valid for ngspice
only.
-s --server Run in server mode. This is like batch mode, except
that a temporary rawfile is used and then written to
the standard output, preceded by a line with a single
"@", after the simulation is done. This mode is used by
the ngspice daemon. This option is valid for ngspice
only.
Example for using pipes from the console window:
cat adder.cir|ngspice -s|more
-i --interactive Run in interactive mode. This is useful if the standard
input is not a terminal but interactive mode is desired.
Command completion is not available unless the
standard input is a terminal, however. This option is
valid for ngspice only.
-r FILE --rawfile=FILE Use rawfile as the default file into which the results of
the simulation are saved. This option is valid for
ngspice only.
-p --pipe Allow a program (e.g., xcircuit) to act as a GUI
frontend for ngspice through a pipe. Thus ngspice will
assume that the input pipe is a tty and allow running
in interactive mode.
-o FILE --output=FILE All logs generated during a batch run (-b) will be saved
in outfile.
-h --help A short help statement of the command line syntax.
-v --version Prints a version information.
-a --autorun Start simulation immediately, as if a control section
.control
run
.endc
had been added to the input file.
--soa-log=FILE output from Safe Operating Area (SOA) check
-D --define Set a variable (17.8.1), to be used in a .control section.
-D var1 will set a boolean variable named var1,
-D var2=7 will set a variable with its value.
Further arguments to ngspice are taken to be ngspice input files, which are read and
saved (if running in batch mode then they are run immediately). Ngspice accepts Spice3
(and also most Spice2) input files, and outputs ASCII plots, Fourier analyses, and node
printouts as specified in .plot, .four, and .print cards. If an out parameter is given on
a .width card (15.6.7), the effect is the same as set width = .... Since ngspice ASCII plots
do not use multiple ranges, however, if vectors together on a .plot card have different
ranges they do not provide as much information as they do in a scalable graphics plot.
For ngnutmeg, further arguments are taken to be data files in binary or ASCII raw file
format (generated with -r in batch mode or the write (see 17.5.107) command) that are
loaded into ngnutmeg. If the file is in binary format, it may be only partially completed
(useful for examining output before the simulation is finished). One file may contain any
number of data sets from different analyses.
16.4 Starting options
16.4.1 Batch mode

Let’s take as an example the Four-Bit binary adder MOS circuit shown in Chapt. 21.6,
stored in a file adder-mos.cir. You may start the simulation immediately by calling
ngspice -b -r adder.raw -o adder.log adder-mos.cir
ngspice will start, simulate according to the .tran command and store the output data
in a rawfile adder.raw. Comments, warnings and info messages go to log file adder.log.
Commands for batch mode operation are described in Chapt. 15.
16.4.2 Interactive mode

If you call
ngspice
ngspice will start, load spinit (16.5) and .spiceinit (16.6, if available), and then waits for
your manual input. Any of the commands described in 17.5 may be chosen, but many of
them are useful only after a circuit has been loaded by
ngspice 1 -> source adder-mos.cir
others require the simulation to be done already (e.g. plot):
ngspice 2 ->run
ngspice 3 ->plot allv
If you call ngspice from the command line with a circuit file as parameter:
ngspice adder-mos.cir
ngspice will start, load the circuit file, parse the circuit (same circuit file as above, con-
taining only dot commands (see Chapt. 15) for analysis and output control). ngspice then
just waits for your input. You may start the simulation by issuing the run command.
Following completion of the simulation you may analyze the data by any of the commands
given in Chapt. 17.5.
16.4. STARTING OPTIONS 347
16.4.3 Control mode (Interactive mode with control file or con-

trol section)
If you add the following control section to your input file adder-mos.cir, you may call
ngspice adder-mos.cir
from the command line and see ngspice starting, simulating and then plotting immediately.
Control section:
* ADDER - 4 BIT ALL - NAND - GATE BINARY ADDER

. control
save vcc # branch
run
plot vcc # branch
rusage all
. endc
Any suitable command listed in Chapt. 17.5 may be added to the control section, as well
as control structures described in Chapt. 17.6. Batch-like behavior may be obtained by
changing the control section to
Control section with batch-like behavior:

. control
save vcc # branch
run
write adder . raw vcc # branch
quit
. endc
If you put this control section into a file, say adder-start.sp, you may just add the line
.include adder-start.sp
to your input file adder-mos.cir to obtain the batch-like behavior. In the following example
the line .tran ... from the input file is overridden by the tran command given in the
control section.
Control section overriding the .tran command:

. control
save vcc # branch
tran 1 n 500 n
plot vcc # branch
rusage time
. endc
The commands within the .control section are executed in the order they are listed and
only after the circuit has been read in and parsed. If you want to have a command being
executed before circuit parsing, you may use the prefix pre_ (17.5.57) to the command.
A warning is due however: If your circuit file contains such a control section (.control ...
.endc), you should not start ngspice in batch mode (with -b as parameter). The outcome
may be unpredictable!
16.5 Standard configuration file spinit
At startup ngspice reads its configuration file spinit. spinit may be found in a path relative
to the location of the ngspice executable
..\share\ngspice\scripts. The path may be overridden by setting the environmental vari-
able SPICE_SCRIPTS to a path where spinit is located. Ngspice for Windows will addi-
tionally search for spinit in the directory where ngspice.exe resides. If spinit is not found
a warning message is issued, but ngspice continues.
spinit contains a script, made of commands from Chapt. 17.5, that is run upon start up
of ngspice. Aliases (name equivalences) can be set. The asterisk ‘*’ comments out a line.
If used by ngspice, spinit will then load the XSPICE code models from a path relative to
the current directory where the ngspice executable resides, as well as OpenVAF compiled
compact devices models. You may also define absolute paths.
If the standard path for the libraries (see standard spinit above or /usr/local/lib/spice
under CYGWIN and Linux) is not adequate, you can add the ./configure options --prefix=/usr
--libdir=/usr/lib64 to set the codemodel search path to /usr/lib64/spice. Besides
the standard lib only lib64 is acknowledged.
16.5. STANDARD CONFIGURATION FILE SPINIT 349
Standard spinit contents:
* Standard ngspice init file

alias exit quit
alias acct rusage all
** set the number of threads in openmp
** ( to the number of physical cores )
** default ( if compiled with -- enable - openmp ) is : 2
set num_threads =8
if $ ? sharedmode
unset interactive
unset moremode
else
set interactive
set x11lineararcs
end
* comment out if central osdi management is set up

* unset osdi_enabled
* Load the codemodels

if $ ? xspice_enabled
codemodel ../ lib / spice / spice2poly . cm
codemodel ../ lib / spice / analog . cm
codemodel ../ lib / spice / digital . cm
codemodel ../ lib / spice / xtradev . cm
codemodel ../ lib / spice / xtraevt . cm
codemodel ../ lib / spice / table . cm
end
* Load the OpenVAF / OSDI models

if $ ? osdi_enabled
osdi ../ lib / ngspice / BSIMBULK107 . osdi
osdi ../ lib / ngspice / BSIMCMG . osdi
osdi ../ lib / ngspice / psp103_nqs . osdi
osdi ../ lib / ngspice / vbic_4T_et_cf . osdi
end
Special care has to be taken when using the ngspice shared library. If you use ngspice.dll
under Windows OS, the standard is to use relative paths for the code models as shown
above. However, the path is relative to the calling program, not to the dll. This is fine
when ngspice.dll and the calling program reside in the same directory. If ngspice.dll is
placed in a different directory, please check Chapt. 32.2.
The Linux shared library ... t.b.d.

16.6 User defined configuration file .spiceinit
In addition to spinit you may define a (personal) file .spiceinit and put it into the current
directory or in your home directory. The typical search sequence for .spiceinit is: user
provided directory (in env. variable SPICE_USERINIT_DIR), current directory, HOME
(Linux) and then USERPROFILE (Windows). HOME (Linux, Cygwin, macOS) may
point to /home/<User name>, or /root if you are acting as admin. USERPROFILE
(MS Windows) is typically C:\Users\<User name>. To find out what directory HOME
or USERPROFILE are pointing to, enter the commands set or export into a console
window and search for the token.
.spiceinit will be read in and executed after spinit, but before any other input file is read. It
may contain further scripts, set variables, or issue commands from Chapt.17.5 to override
commands given in spinit. For example set filetype=ascii will yield ASCII output in
the output data file (rawfile), instead of the compact binary format that is used by default.
set ngdebug will yield a lot of additional debug output. Any other contents of the script,
e.g. plotting preferences, may be included here also. If the command line option -n is
used upon ngspice start up, this file will be ignored.
.spiceinit for simulating IC designs with MOS transistor data from PDKs may contain:
* . spiceinit for use with Skywater PDK and ngspice KLU

set ngbehavior = hsa ; set compatibility for reading
; PDK libs
set skywaterpdk ; omit some time consuming checks
; during lib loading
set ng_nomodcheck ; don ’ t check the model parameters
option noinit ; don ’ t print operating point data
option klu ; select KLU as matrix solver
optran 0 0 0 100 p 2 n 0 ; don ’ t use dc operating point ,
; but transient op
set num_threads=8 should be set to the number of physical cores of the computer in use
(here for example 8 cores), set ngbehavior=hsa will ensure HSPICE compatibility with
some important and essential tweaks for the PDK, set skywaterpdk suppresses time
consuming checks during lib loading, assuming 4 nodes for a MOS device and adequately
labled parameters. set ng_nomodcheck will suppress some unwanted warnings, option
noinit will suppress the (often lengthy) printing of the operating point results. option
klu often will yield simulation speed up by a factor of 2 or more. optran ... will skip
usual operating point iterations, which for very large circuits consume much time, and
replace them by a time integrated operating point estimation.
16.7. ENVIRONMENTAL VARIABLES 351
.spiceinit for simulating circuits containing PSPICE-compatible behavioural models may

contain:
* User defined ngspice init file

set filetype = ascii
* set ngdebug
* set outputpath = C :\ Spice64 \ out
set ngbehavior = ltpsa
option sparse
set ngbehavior = ltpsa will provide PSPICE compatibility. option sparse (maybe
omitted) selects the venerable Sparse 1.3 matrix solver, which sometimes is much faster
than klu.
Some editors on MS Windows refuse to save files with leading dot in their names. An
alternative name to .spiceinit is therefore spice.rc.
16.7 Environmental variables
16.7.1 Ngspice specific variables

SPICE_LIB_DIR default: /usr/local/share/ngspice (Linux, CYGWIN), C:\Spice\share\ngspice
(Windows)
SPICE_EXEC_DIR default: /usr/local/bin (Linux, CYGWIN), C:\Spice\bin (Windows)
SPICE_BUGADDR default: https://ngspice.sourceforge.io/bugrep.html

Where to send bug reports on ngspice.
SPICE_EDITOR default: vi (Linux, CYGWIN), notepad.exe (MINGW, Visual Studio)

Set the editor called in the edit command. Always overrides the EDITOR env.
variable.
SPICE_ASCIIRAWFILE default: 0
Format of the rawfile. 0 for binary, and 1 for ascii.
SPICE_NEWS default: $SPICE_LIB_DIR/news

A file that is copied verbatim to stdout when ngspice starts in interactive mode.
SPICE_HELP_DIR default: $SPICE_LIB_DIR/helpdir

Help directory, not used in Windows mode
SPICE_HOST default: empty string

Used in the rspice command (probably obsolete, to be documented)
SPICE_SCRIPTS default: $SPICE_LIB_DIR/scripts

In this directory the spinit file will be searched.
SPICE_PATH default: $SPICE_EXEC_DIR/ngspice

Used in the aspice command (probably obsolete, to be documented)
NGSPICE_MEAS_PRECISION default: 5
Sets the number of digits if output values are printed by the meas(ure) command.
SPICE_NO_DATASEG_CHECK default: undefined

If defined, will suppress memory resource info (probably obsolete, not used on Win-
dows or where the /proc information system is available.)
NGSPICE_INPUT_DIR default: undefined

If defined, using a valid directory name, will add the given directory to the search
path when looking for input files (*.cir, *.inc, *.lib).
SPICE_USERINIT_DIR default: undefined

If defined, using a valid directory name, this is the first place to search for the user-
defined initialization file .spiceinit (or spice.rc). The search sequence then following
is: current directory, HOME directory, USERPROFILE directory
16.7.2 Common environment variables
TERM LINES COLS DISPLAY HOME PATH EDITOR SHELL POSIXLY_CORRECT
16.8 Memory usage

Ngspice started with batch option (-b) and rawfile output (-r rawfile) will store all sim-
ulation data immediately into the rawfile without keeping them in memory. Thus very
large circuits may be simulated, the memory requested upon ngspice start up will depend
on the circuit size, but will not increase during simulation.
If you start ngspice in interactive mode or interactively with control section, all data will
be kept in memory, to be available for later evaluation. A large circuit may outgrow even
Gigabytes of memory. The same may happen after a very long simulation run with many
vectors and many time steps to be stored. Issuing the save <nodes> command will help
to reduce memory requirements by saving only the data defined by the command. You
may also choose option INTERP (15.1.4) to reduce memory usage.
16.9 Simulation time

Simulating large circuits may take an considerable amount of CPU time. If this is of
importance, you should compile ngspice with the flags for optimum speed, set during
configuring ngspice compilation. Under Linux, MINGW, and CYGWIN you should select
the following option to disable the debug mode, which slows down ngspice:
./configure --disable-debug
Adding --disable-debug will set the -O2 optimization flag for compiling and linking.
Under MS Visual Studio, you will have to select the release version, which includes
optimization for speed.
16.10. NGSPICE ON MULTI-CORE PROCESSORS USING OPENMP 353
If you have selected XSPICE (see Chapt. 12 and II) as part of your compilation con-
figuration (by adding the option --enable-xspice to your ./configure command), the
value of trtol (see 15.1.4) is set internally to 1 (instead of default 7) for higher precision
if XSPICE code model ’A’ devices included in the circuit. This may double or even triple
the CPU time needed for any transient simulation, because the amount of time steps and
thus iteration steps is more than doubled. For MS Visual Studio compilation there is
currently no simple way to exclude XSPICE during compilation.
You may enforce higher speed during XSPICE usage by setting the variable xtrtol in your
.spiceinit initialization file or in the .control section in front of the tran command (via set
xtrtol=2 using the set command 17.5.73) and override the above trtol reduction. Beware
however of precision or convergence issues if you use XSPICE ’A’ devices, especially if
xtrtol is set to values larger than 2.
If your circuit is composed mostly of MOS transistors, and you have a multi-core processor
at hand, you may benefit from OpenMP parallel processing, as described next (16.10).
16.10 Ngspice on multi-core processors using OpenMP
16.10.1 Introduction
Today’s computers typically come with CPUs having more than one core. It will thus be
useful to enhance ngspice to make use of such multi-core processors.
Using circuits containing mostly transistors and e.g. the BSIM3 model, around 2/3 of the
CPU time is spent in evaluating the model equations (e.g. in the BSIM3Load() function).
The same happens with other advanced transistor models. Thus, such functions should be
parallelized, if possible. Solving the matrix takes about 10% to 50% of the CPU time, so
parallel processing in the matrix solver is sometimes of secondary interest only! Further,
such paralellization is difficult to achieve with our Sparse Matrix and KLU solvers.
Another alternative is using CUSPICE, that is ngspice (current version 27) designed for
running massively parallel on NVIDIA GPUs. CUDA enhancements to C code are applied.
For LINUX, please see the user guide. For MS Windows, an executable is available at the
ngspice download pages.
16.10.2 Internals
A publication [1] has described a way to exactly do that using OpenMP, which is available
on many platforms and is easy to use, especially if you want to perform parallel processing
of a for-loop.
To explain the implemented approach BSIM3 version 3.3.0 model was chosen, located in
the BSIM3 directory, as the first example. The BSIM3load() function in b3ld.c contains
two nested for-loops using linked lists (models and instances, e.g. individual transistors).
Unfortunately OpenMP requires a loop with an integer index. So in file B3set.c an
array is defined, filled with pointers to all instances of BSIM3 and stored in model-
>BSIM3InstanceArray.
Table 16.1: OpenMP performance

Threads CPU time [s] CPU time [s]
Windows Linux
1 65.4 69.3
2 46.7 47.4
4 37.2 36.9
6 33.6 33.6
8 32.4 32.4
12 35.7 31.7
16 38.2 34.3
BSIM3load() is now a wrapper function, calling the for-loop, which runs through functions
BSIM3LoadOMP(), once per instance. Inside BSIM3LoadOMP() the model equations are
calculated.
Typically it is necessary to use synchronization constructs such as mutexes when multiple
threads write to a common memory location. To avoid the performance degradation of
such synchronization, temporary per-thread memory locations are used within the for loop
of the BSIM3LoadOMP() function as defined in bsim3def.h. After all threads complete
the for-loop, the update to the matrix is done in an extra function BSIM3LoadRhsMat()
in the main thread.
Then the thread programming needed is only a single line!!
#pragma omp parallel for
introducing the for-loop over the device instances.
This of course is made possible only thanks to the OpenMP guys and the clever trick on
no synchronization introduced by the above cited authors.
The time-measuring function getrusage() used with Linux or Cygwin to determine the
CPU time usage (with the rusage option enabled) counts tics from every core, adds them
up, and thus reports a CPU time value enlarged by a factor of 8 if 8 threads have been
chosen. So now ngspice is forced to use ftime for time measuring if OpenMP is selected.
16.10.3 Some results

Some results on an inverter chain with 627 CMOS inverters, BSIM4.7, 45 nm, running for
200ns, compiled with Visual Studio Community 2019 on Windows 10 (full optimization)
or gcc 7.4, SUSE Linux Leap 15.1, -O2, on a i9 9900K machine with 8 real cores (16
logical processors using hyperthreading) and 32 GB of memory are shown in table 16.1.
So we see a ngspice speed up of more than a factor of two! Even on an Windows 7
notebook with a dual core i7 processor, more than 1.5x improvement using two threads
was attained. This is consistent with the fact that roughly half of the CPU time is used
for evaluating the device model, half of the time for solving the matrix. Only the device
evaluation is parallelized by OpenMP. The time for doing this becomes negligible with
8 or more threads. Allowing more than 8 threads (using the 8 physical cores) does not
yield much improvement, even leads to a slight increase of simulation time, because the
code is not optimized for hyperthreading.
16.11. SERVER MODE OPTION -S 355
16.10.4 Usage
To state it clearly: OpenMP is installed inside the model equations of a particular model.
It is available in BSIM3 versions 3.3.0 and 3.2.4, but not in any other BSIM3 model,
in BSIM4 versions 4.5, 4.6.5, 4.7 or 4.8, but not in any other BSIM4 model, and in
B4SOI, version 4.4, not in any other SOI model. Older parameter files of version 4.6.x
(x any number up to 5) are accepted, you have to check for compatibility.
Under Linux you may run
./autogen.sh
./configure ... --enable-openmp
make install
The same has been tested under MS Windows with CYGWIN and MINGW as well
and delivers similar results.
Under MS Windows with Visual Studio Professional the preprocessor flag USE_OMP,
and the /openmp flag in Visual Studio are enabled by default. Visual Studio 2015 and
later offer OpenMP support inherently.
The number of threads has to be set manually by placing
set num_threads=4
into spinit or .spiceinit or in the control section of the SPICE input file. If OpenMP is
enabled, but num_threads not set, a default value num_threads=2 is set internally.
If you simulate a circuit, please keep in mind to select BSIM3 (levels 8, 49) version 3.2.4 or
3.3.0 (11.2.10), by placing this version number into your parameter files, BSIM4 (levels 14,
54) version 4.5, 4.6.5, 4.7 or 4.8 (11.2.11), or B4SOI (levels 10, 58) version 4.4 (11.2.12).
All other transistor models run as usual (without multithreading support).
If you run ./configure without --enable-openmp (or without USE_OMP preprocessor flag
under MS Windows), you will get only the standard, not paralleled BSIM3 and BSIM4
models, as has been available from Berkeley. If OpenMP is selected and the number
of threads set to 1, there will be only a very slight CPU time disadvantage (typ. 3%)
compared to the old, non OpenMP build.
16.10.5 Literature
[1] R.K. Perng, T.-H. Weng, and K.-C. Li: "On Performance Enhancement of Circuit
Simulation Using Multithreaded Techniques", IEEE International Conference on Compu-
tational Science and Engineering, 2009, pp. 158-165
16.11 Server mode option -s

A program may write the SPICE input to the console. This output is redirected to ngspice
via ‘|’. ngspice called with the -s option writes its output to the console, which again is
redirected to a receiving program by ‘|’. In the following simple example cat reads the
input file and prints it content to the console, which is redirected to ngspice by a first
pipe, ngspice transfers its output (similar to a raw file, see below) to less via another
pipe.
Example command line:
cat input . cir | ngspice -s | less
Under MS Windows you will need to compile ngspice as a console application (see Chapt.
32.2.4) for this server mode usage.
test -s
v1 1 0 1
r1 1 0 2 k
. options filetype = ascii
. save i ( v1 )
. dc v1 -1 1 0.5
. end
If you start ngspice console with
ngspice -s
you may type in the above circuit line by line (not to forget the first line, which is a title
and will be ignored). If you close your input with ctrl Z, and return, you will get the
following output (this is valid for MINGW only) on the console, like a raw file:
Circuit: test -s
Doing analysis at TEMP = 27.000000 and TNOM = 27.000000
Title: test -s
Date: Sun Jan 15 18:57:13 2012
Plotname: DC transfer characteristic
Flags: real
No. Variables: 2
No. Points: 0
Variables:
No. of Data Columns : 2
0 v(v-sweep) voltage
1 i(v1) current
Values:
0 -1.000000000000000e+000
5.000000000000000e-004
1 -5.000000000000000e-001
2.500000000000000e-004
2 0.000000000000000e+000
0.000000000000000e+000
16.12. PIPE MODE OPTION -P 357
3 5.000000000000000e-001
-2.500000000000000e-004
4 1.000000000000000e+000
-5.000000000000000e-004
@@@ 122 5
The number 5 of the last line @@@ 122 5 shows the number of data points, which is
missing in the above line No. Points: 0 because at the time of writing to the console
it has not yet been available.
ctrl Z is not usable here in Linux, a patch to install ctrl D instead is being evaluated.
16.12 Pipe mode option -p
A program may write a set of ngspice commands (see 17.5) to the console. This output is
redirected to ngspice via ‘|’. ngspice called with the -p option immediately executes the
commands and then exits. In the following simple example cat reads the input file and
prints it content to the console, which is redirected to ngspice by a pipe, ngspice executes
the commands.
Example command line:
cat pipe - circuit . cir | ngspice -p
Under MS Windows you will need to compile ngspice as a console application (see Chapt.
32.2.4) for this pipe mode usage.
* pipe - circuit . cir

source circuit . cir
tran 10 u 2 m
write pcir . raw all
Example circuit file:
* Circuit . cir
V1 n0 0 SIN (0 10 1 kHz )
C1 n1 n0 3.3 nF
R1 0 n1 1 k
. end
The raw file pcir.raw will contain the final simulation results.
16.13 Ngspice control via input, output fifos
Example bash script:
#!/ usr / bin / env bash
NGSPICE_COMMAND =" ngspice "
rm input . fifo
rm output . fifo
mkfifo input . fifo

mkfifo output . fifo
$NGSPICE_COMMAND -p -i < input . fifo > output . fifo &
exec 3 > input . fifo

echo " I can write to input . fifo "
echo " Start processing ..."

echo ""
echo " source circuit . cir " >&3

echo " unset askquit " >&3
echo " set nobreak " >&3
echo " tran 0.01 ms 0.1 ms " >&3
echo " print n0 " >&3
echo " quit " >&3
echo " Try to open output . fifo ..."

exec 4 < output . fifo
echo " I can read from output . fifo "
echo " Ready to read ..."

while read output
do
echo $output
done <&4
exec 3 >& -
exec 4 >& -
echo " End processing "
The bash script listed above (tested under Linux and Cygwin)
- launches ngspice in pipe mode (-p) in another thread.
- writes some commands to the ngspice input
16.14. COMPATIBILITY 359
- runs ngspice with the tran command

- reads the output and prints it onto the console.
The input file with a small circuit is:
Circuit.cir:
* Circuit . cir
V1 n0 0 SIN (0 10 1 kHz )
C1 n1 n0 3.3 nF
R1 0 n1 1 k
. end
16.14 Compatibility
ngspice is a direct derivative of spice3f5 from UC Berkeley and thus inherits all of the
commands available in its predecessor. Thanks to the open source policy of UCB (orig-
inal spice3 from 1994 is still available here), several commercial variants have sprung
off, either being more dedicated to IC design or more concentrating on simulating dis-
crete and board level electronics. None of the commercial and almost none of the freely
downloadable SPICE providers publishes the source code. All of them have proceeded
with the development, by adding functionality, or by adding a more dedicated user inter-
face. Some have kept the original SPICE syntax for their netlist description, others have
quickly changed some if not many of the commands, functions and procedures. Thus it
is difficult, if not impossible, to offer a simulator that acknowledges all of these netlist
dialects. ngspice includes some features that enhance compatibility that are included au-
tomatically. This selection may be controlled to some extend by setting the compatibility
mode. Others may be invoked by the user by small additions to the netlist input file.
Some of them are listed in this chapter, some will be integrated into ngspice at a later
stage, others will be added if they are reported by users.
16.14.1 Compatibility mode

The variable (17.7) ngbehavior sets the compatibility mode. Per default no compatibility
mode is selected. The compatibility status will be displayed in the output window.
set ngbehavior=ltpsa
in spinit or .spiceinit is atypical command, setting PSPICE and LTSPICE compatibility

for the whole netlist. Flag ’a’ may be combined with any of the flags listed below. By
contrast
set ngbehavior=ps
(without ’a’) will set PSPICE compatibility only for libraries which are added by a .in-
clude command. So you may keep your Spice3 compatible netlist, but including PSPICE
device models. The available compatibility flags are:
Flag Ref. Short description

a complete netlist transformed
ps 16.14.5 PSPICE compatibility
hs 16.14.10 HSPICE compatibility
spe 16.14.9 Spectre compatibility
lt 16.14.6 LTSPICE compatibility
s3 Spice3 compatibility
ll all (currently not used)
ki 16.14.8 KiCad compatibility
eg EAGLE compatibility
mc for ’make check’
Table 16.2: Compatibility flags
’s3’ will disable some of the advanced ngspice features. ’eg’ will enable EAGLE com-
patible voltage vector output.’mc’ is required when the command ’make check’ is to
be executed. Then all flags are reset, in addition the compatibility status output is sup-
pressed. Flags ’ps’ and ’hs’ are mutually exclusive.
The command ’unset ngbehavior’ will remove the variable ngbehavior, thus resetting
the compatibility mode to the default (no compat mode is set).
16.14.2 Missing functions

You may add one or more function definitions to your input file, as listed below.
.func LIMIT(x,a,b) {min(max(x, a), b)}

.func PWR(x,a) {abs(x) ** a}
.func PWRS(x,a) {sgn(x) * PWR(x,a)}
.func stp(x) {u(x)}
16.14.3 Devices
16.14.3.1 E Source with LAPLACE
see 5.2.5.
16.14.3.2 VSwitch
The VSwitch
S1 2 3 11 0 SW
.MODEL SW VSWITCH(VON=5V VOFF=0V RON=0.1 ROFF=100K)
may become
a1 %v(11) %gd(2 3) sw
.MODEL SW aswitch(cntl_off=0.0 cntl_on=5.0 r_off=1e5
+ r_on=0.1 log=TRUE)
The XSPICE option has to be enabled.
16.14.4 Controls and commands
16.14.4.1 .lib
The ngspice .lib command (see 2.8) requires two parameters, a file name followed by a
library name. If no library name is given, the line
.lib filename
should be replaced by
.inc filename
Alternatively, the compatibility mode (16.14.1) may be set to ’ps’.
16.14.4.2 .step
Repeated analysis in ngspice is offered by a short script inside of a .control section (see
Chapt. 17.8.8) added to the input file. A simple application (multiple dc sweeps) is shown
below.
Input file with parameter sweep
parameter sweep
* resistive divider , R1 swept from start_r to stop_r
* replaces . STEP R1 1 k 10 k 1 k
R1 1 2 1 k
R2 2 0 1 k
VDD 1 0 DC 1
. dc VDD 0 1 .1
. control
let start_r = 1 k
let stop_r = 10 k
let delta_r = 1 k
let r_act = start_r
* loop
while r_act le stop_r
alter r1 r_act
run
write dc - sweep . out v (2)
set appendwrite
let r_act = r_act + delta_r
end
plot dc1 . v (2) dc2 . v (2) dc3 . v (2) dc4 . v (2) dc5 . v (2)
+ dc6 . v (2) dc7 . v (2) dc8 . v (2) dc9 . v (2) dc10 . v (2)
. endc
. end
16.14.5 PSPICE Compatibility mode

If the variable (17.7) ngbehavior is set to ’ps’ or ’psa’ with the commands
set ngbehavior=ps
or
set ngbehavior=psa
in spinit or .spiceinit, ngspice will translate all files that have been read into ngspice netlist
by the .include command (ps) or the complete netlist (psa) from PSPICE syntax to
ngspice. This feature allows reading of PSPICE (or TINA) compatible device libraries
(ps) that are often supplied by the semiconductor device manufacturers. Or you may
choose to use complete PSPICE simulation decks (psa). Some ngspice input files may
fail, however. For example ngspice\examples\memristor\memristor.sp will not do, because

it uses the parameter vt, and vt is a reserved word in PSPICE.
PSPICE to ngspice translation details:
• .model replacement in ako (a kind of) model descriptions
• replace the E source TABLE function by a B source pwl
• add predefined params TEMP, VT, GMIN to beginning of deck
• add predefined params TEMP, VT to beginning of each .subckt call
• add .functions limit, pwr, pwrs, stp, if, int
• replace
S1 D S DG GND SWN
.MODEL SWN VSWITCH(VON=0.55 VOFF=0.49
+ RON={1/(2*M*(W/LE)*(KPN/2)*10)} ROFF=1G)
by
as1 %vd(DG GND) % gd(D S) aswn
.model aswn aswitch(cntl_off=0.49 cntl_on=0.55
+ r_off=1G r_on={1/(2*M*(W/LE)*(KPN/2)*10)} log=TRUE)
• replace & by &&
• replace | by ||
• replace T_ABS by temp and T_REL_GLOBAL by dtemp
• get the area factor for diodes and bipolar devices

d1 n1 n2 dmod 7 –> d1 n1 n2 dmod area=7
q2 n1 n2 n3 [n4] bjtmod 1.35 –> q2 n1 n2 n3 n4 bjtmod area=1.35
q3 1 2 3 4 bjtmod 1.45 –> q2 1 2 3 4 bjtmod area=1.45
• Check for double ’{{ }}’, replace the inner ’{’, ’}’ by ’(’, ’)’
• Limit for exp function (linear growth when exponent is larger than 14).
In ps or psa mode, ngspice will treat all .lib entries like .include. There is no hierarchically
library handling. So for reading HSPICE compatible libraries, you definitely have to unset
the ps mode, e.g. by not adding set ngbehavior=ps or disabling it by
unset ngbehavior=ps
16.14.6 LTSPICE Compatibility mode

If the variable (17.7) ngbehavior is set to ’lt’ or ’lta’ with the commands
set ngbehavior=lt
or
set ngbehavior=lta
in spinit or .spiceinit, ngspice will translate all files that have been read into ngspice
netlist by the .include command (lt) or the complete netlist (lta) from LTSPICE syn-
tax to ngspice. This feature allows reading of LTSPICE compatible device libraries or
complete netlists.
Currently we offer only a subset of the documented or undocumented functions (uplim,
dnlim, uplim_tanh, dnlim_tanh). More user input is definitely required here!
This compatibility mode also adds a simple diode using the sidiode code model ( 12.2.32).
The diode model
d1 a k ds1
.model ds1 d(Roff=1000 Ron=0.7 Rrev=0.2 Vfwd=1
is translated automatically to the equivalent code model diode
ad1 a k ads1
.model ads1 sidiode(Roff=1000 Ron=0.7 Rrev=0.2 Vfwd=1
RKM code compatibility:
• In LT compatibility mode ngspice will follow the RKM code notation. In addition
to the standard notation, resistor (R) and capacitor (C) values may also be entered
according to the following listings (the internally translated value is given after the
;):
RKM code for resistors
R1 1 0 4 K7 ; 4.7 k
R2 1 0 4 R7 ; 4.7
R3 1 0 R47 ; 0.47
R4 1 0 470 R ; 470
R5 1 0 47 K ; 47 k
R6 1 0 47 K3 ; 47.3 k
R7 1 0 470 K ; 470 k
R8 1 0 4 Meg7 tc1 =1 e -6 tc2 =1 e -9 dtemp =6
* ; 4.7 Meg <-- Not defined in the RKM notation
R9 1 0 4 L7 ; 4.7 m
R10 1 0 470 L ; 470 m
R11 1 0 4 M7 ; 4.7 m <-- This deviates fom the RKM notation
RKM code for capacitors
C1 1 0 4 p7 ; 4.7 p
C2 1 0 4 n7 ; 4.7 n
C3 1 0 4 u7 ; 4.7 u
C4 1 0 4 m7 ; 4.7 m
C5 1 0 4 F7 ; 4.7 f <-- This deviates fom the RKM notation
C6 1 0 47 p3 ; 4.73 p
C7 1 0 470 p ; 470 p
C8 1 0 4 u76 tc1 =1 e -6 tc2 =1 e -9 dtemp =6
* ; 4.76 u
C9 1 0 4 m7 ; 4.7 m
C10 1 0 470 nF ; 470 n
C11 1 0 47 fF ; 47 f <-- This deviates fom the RKM notation
There are some exceptions to the RKM code notation:
• all letters may be entered upper or lower case, and will internally be transformed
to lower case.
• m, M always denote milli (1e-3).
• f, F denote femto (1e-15), fF will be again femto
• meg, Meg denotes mega (1e6)
16.14.7 LTSPICE/PSPICE Compatibility mode

If the variable (17.7) ngbehavior is set to ’ltps’ or ’ltpsa’ with the commands
set ngbehavior=ltps
or
set ngbehavior=ltpsa
in spinit or .spiceinit, ngspice will translate all files that have been read into ngspice netlist
by the .include command (ltps) or the complete netlist (ltpsa) 16.14.6, 16.14.5 from LT-
SPICE and PSPICE syntax to ngspice. This feature allows reading of LTSPICE and
PSPICE compatible device libraries or complete netlists.
16.14.8 KiCad Compatibility mode

KiCad will generate vector names containing ’/’. If the variable (17.7) ngbehavior is set
to ki with the command
set ngbehavior=ki
is set in .spiceinit (or plot line flag kicad is given 17.5.56), ngspice will place " around
this vector name. The mathematical operation ’division’ in the plot command will then
work only if spaces are placed around the division operator /.
16.14.9 Spectre Compatibility mode

If the variable (17.7) ngbehavior is set to spe with the command
set ngbehavior=spe
is set in .spiceinit Spectre compatibility mode is enabled. True compatibility today is

still far away. The only action available for now is the use of the MOS device instance
parameter nf. If nf is given and larger than 1 and Spectre (or HSPICE) compatibility
is enabled, nf is used as a divisor to the transistor width W given on the instance line.
The resulting W/nf is now used to select the suitable device model in the binning process.
This procedure is of interest for a multi-gate transistor, which has a total width of W, but
each finger is model according to the model given for W/nf.
16.14.10 HSPICE Compatibility mode

If the variable (17.7) ngbehavior is set to hs with the command
set ngbehavior=hs
is set in .spiceinit HSPICE compatibility mode is enabled. This mode allows to read
libraries with the .lib command in a recursive fashion, as is required by HSPICE com-
patible process development kits (PDKs) In addition the nf flag is enabled, as described
in 16.14.9 .
16.15 Tests
The ngspice distribution is accompanied by a suite of test input and output files, located
in the directory ngspice/tests. Originally this suite was meant to see if ngspice with all
models was made and installed properly. It is started by
$ make check
from within your compilation and development shell. A sequence of simulations is thus
started, its outputs compared to given output files by comparisons string by string. This
feature is momentarily used to check for some basic procedures and the XSPICE extension
(12) as a regression test. Several other input files located in directory ngspice/tests may
serve as light-weight examples for invoking devices and simple circuits.
Today’s very complex device models (BSIM3 (11.2.10), BSIM4 (see 11.2.11), HiSIM (see
11.2.14) and others) require a different strategy for verification. Under development for
ngspice is the CMC Regression test by Colin McAndrew, which accompanies every new
model. These tests cover a large range of different DC, AC and noise simulations with
different geometry ranges and operating conditions and are more meaningful the transient
simulations with their step size dependencies. A major advantage is the scalability of
the diff comparisons, which check for equality within a given tolerance. A set of Perl
modules cares for input, output and comparisons of the models. Currently BSIM3, BSIM4,
16.16. TOOLS FOR DEBUGGING A CIRCUIT NETLIST 367
BSIMSOI4, HICUM2, HiSIM, and HiSIM_HV models implement the QA test. You may
invoke it by running the command given above or by
$ make -i check 2>&1 | tee results
-i will cause make to ignore any errors, and tee will provide console output as well as
printing to file ’results’. Be aware that under MS Windows you will need the console
binary (see 32.2.4) to run the CMC tests, and you have to have Perl installed!
As these tests may consume a considerable amount of CPU time, there is a configure
option --enable-shortcheck 32.1.8.1 available, providing a strongly reduced runtime,
because besides some regression tests only BSIM3 and BSM4 devices are checked.
Other tests have been developed, there are also some benchmark circuit compilations
available. Please have a look at our Tests and Quality Assurance web page.
16.16 Tools for debugging a circuit netlist
This a chapter only in its initial state. Not all circuits will simulate immediately and
easily. The netlist may contain a bug. The netlist may be o.k., but then ngspice may not
find an operating point. If the operating point has been found, the transient simulation
will just yield the famous error message ’transient time step too small’. Unfortumately
there are many reasons for failure, on the other hand there is a lot of literature available
to traet non-convergence.
So for now there will be listed here only a few ’tools’ offered by ngspice to aid debugging.
16.16.1 options and initial conditions
If ngspice has trouble finding the operating point, setting some initial conditions by adding
.nodeset (15.2.1) or .ic (15.2.2) for critical nodes may help. The variation of some op
option parameters may help as well (see 15.1.2). If there are nodes without dc connection
to ground (e.g. two capacitors in series connection), finding the operating point will fail.
Here the option RSHUNT may be of help by adding are (typically large) resistor from
each node to ground. Convergence may be improved by the RSERIES option that add a
(typically small) resistor in series to each inductor.
Transient simulations are governed by another set of options (see 15.1.4). Careful variation
of the parameters, as described in the literature, may enable convergence in incritical
situations (not guaranteed, however).
16.16.2 set debug
If set in .spiceinit (or spice.rc), the command set debug will yield an analysis of each
command which is run from .spiceinit and .control.
16.16.3 set ngdebug

The command set ngdebug, if set in .spiceinit (spice.rc) provides some additional warn-
ing messages. If ngspice has write access to the current directory, 3 or 4 files are saved to
that directory, showing the netlist at specific stages during parsing. Each file contain two
parts, the netlist without comment lines, followed by the same netlist including all com-
ment lines. debug-out.txt is available after pre-processing the netlist. debug-out2.txt
shows the netlist after parameter and subcircuit expansion. debug-out3.txt lists the
final netlist. debug-out-mc.txt is issued, when the netlist is reloaded after a reset or
mc_source command.
During a transient simulation a vector ’speedcheck’ is generated in the current tran plot.
The independent variable is the scale vector ’time’, the dependent variable is the wall clock
time with a resolution of about 100 ms. So you may monitor the simulation progress of a
(lengthy) transient simulation and detect critical (simulated) times where the simulation
may be slowed down.
16.16.4 miscellaneous
Debugging the equations of a B source are described in chapt. 5.4.
Compiling ngspice with the ./configure flag --enable-ftedebug or (for MS Visual Studio:
adding a preprocessor flag FTEDEBUG) will enable some additional warning messages.
Compiling ngspice with the ./configure flag --enable-stepdebug or (for MS Visual Stu-
dio: adding a preprocessor flag STEPDEBUG) yields a very powerful tool for analysing the
steps of a transient simulation. The amount of messages printed however is overwhelming
and may be interpreted by an insider only.
16.17 Reporting bugs and errors

Ngspice is a complex piece of software. The source code contains over 1500 files. Various
models and simulation procedures are provided, some of them not used and tested inten-
sively. Therefore errors may be found, some still evolving from the original spice3f5 code,
others introduced during the ongoing code enhancements.
If you happen to experience an error during the usage of ngspice, please send a report to
the development team. Ngspice is hosted on SourceForge, the preferred place to post a
bug report is the ngspice bug tracker. We would prefer to have your bug tested against the
actual source code available at Git, but of course a report using the most recent ngspice
release is welcome! Please provide the following information with your report:
Ngspice version
Operating system
Small input file to reproduce the bug
Actual output versus the expected output
Chapter 17
Interactive Interpreter
17.1 Introduction
The simulation flow in ngspice (input, simulation, output) may be controlled by dot
commands (see Chapt. 15 and 16.4.1) in batch mode. There is, however, a much more
powerful control scheme available in ngspice, traditionally coined ‘Interactive Interpreter’,
but being much more than just that. In fact there are several ways to use this feature,
truly interactively by typing commands to the input, but also running command sequences
as scripts or as part of your input deck in a quasi batch mode.
You may type in expressions, functions (17.2) or commands (17.5) into the input console
to elaborate on data already achieved from the interactive simulation session.
Sequences of commands, functions and control structures (17.6) may be assembled as a
script (17.8) into a file, and then activated by just typing the file name into the console
input of an interactive ngspice session.
Finally, and most useful, is to add a script to the input file, in addition the the netlist and
dot commands. This is achieved by enclosing the script into .control ... .endc (see
16.4.3, and 17.8.8 for an example). This feature enables a wealth of control options. You
may set internal (17.7) and other variables, start a simulation, evaluate the simulation
output, start a new simulation based on these data, and finally make use of many options
for outputting the data (graphically or into output files).
Historical note: The final releases of Berkeley Spice introduced a command shell and
scripting possibilities. The former releases were not interactive. The choice for the script-
ing language was an early version of ‘csh’, the C-shell, which was en vogue back then as an
improvement over the ubiquitous Bourne Shell. Berkeley Spice incorporated a modified
csh source code that, instead of invoking the unix ‘exec’ system call, executed internal
SPICE C subroutines. Apart from bug fixes, this is still how ngspice works.
One important difference from C-shell is that ngspice does not support multiple commands
on one line, separated by ’;’. In ngspice, semi-colons introduce a comment.
The csh-like scripting language is active in .control sections. It works on ‘strings’, and
does string substitution of ‘environment’ variables. You see the csh at work in ngspice
with set foo = "bar"; set baz = "bar$foo", and in if, repeat, for, ... constructs.
However, ngspice processes mainly numerical data, and support for this was not avail-
able in the c-sh implementation. Therefore, Berkeley implemented an additional type of
369
370 CHAPTER 17. INTERACTIVE INTERPRETER
variables, with different syntax, to access double and complex double vectors (possibly of
length 1). This new variable type is modified with let, and can be used without special
syntax in places where a numerical expression is expected: let bar = 4 * 5; let zoo
= bar * 4 works. Unfortunately, occasionally one has to cross the boundary between
the numeric and the string domain. For this purpose the $& construct is available – it
queries a variable in the numerical let domain, and expands it to a c-sh string denoting
the value. This lets you do do something like set another = "this is $&bar". It is
important to remember that set can only operate on (c-sh) strings, and that let operates
only on numeric data contained in vectors. Convert from numeric to string with $&, and
from string to numeric with $.
17.2 Expressions, Functions, and Constants

Ngspice stores data in the form of vectors: time, voltage, etc. Each vector has a type,
and vectors can be operated on and combined algebraically in ways consistent with their
types. Vectors are normally created as the output of a simulation, or when a data file
(output raw file) is read in again (ngspice using the the load command 17.5.48), or when
the initial data-file is loaded directly into ngnutmeg. They can also be created with the
let command (17.5.45).
An expression is an algebraic formula involving vectors and scalars (a scalar is a vector
of length 1) and the following operations:
+ - * / ^ % ,
% is the modulo operator, and the comma operator has two meanings: if it is present
in the argument list of a user definable function, it serves to separate the arguments.
Otherwise, the term x , y is synonymous with x + j(y). Also available are the logical
operations & (and), | (or), ! (not), and the relational operations <, >, >=, <=, =,
and <> (not equal). If used in an algebraic expression they work like they would in C,
producing values of 0 or 1. The relational operators have the following synonyms:
Operator Synonym
gt >
lt <
ge >=
le <=
ne <>
and &
or |
not !
eq =
The operators are useful when < and > might be confused with the internal IO redirection
(see 17.4, which is almost always happening). It is however safe to use < and > with the
define command (17.5.19).
The following functions are available:
17.2. EXPRESSIONS, FUNCTIONS, AND CONSTANTS 371
Name Function
mag(vector) Magnitude of vector (same as abs(vector)).
ph(vector) Phase of complex vector, in radians.
cph(vector) Phase of complex vector, in radians. Continuous
values, no discontinuity at ±π.
unwrap(vector) Phase of vector with real phase vector in degrees as
input and output. Continuous values, no discontinuity
at ±180.
j(vector) i (sqrt(-1)) times vector.
real(vector) The real component of vector.
imag(vector) The imaginary part of vector.
conj(vector) The complex conjugate of a vector
db(vector) 20 log10(mag(vector)).
log10(vector) The logarithm (base 10) of vector.
ln(vector) The natural logarithm (base e) of vector.
exp(vector) e to the vector power.
abs(vector) The absolute value of vector (same as mag).
sqrt(vector) The square root of vector.
sin(vector) The sine of vector.
cos(vector) The cosine of vector.
tan(vector) The tangent of vector.
atan(vector) The inverse tangent of vector.
sinh(vector) The hyperbolic sine of vector.
cosh(vector) The hyperbolic cosine of vector.
tanh(vector) The hyperbolic tangent of vector.
atanh(vector) The inverse hyperbolic tangent of vector.
floor(vector) Largest integer that is less than or equal to vector.
ceil(vector) Smallest integer that is greater than or equal to vector.
norm(vector) The vector normalized to 1 (i.e, the largest magnitude
of any component is 1).
mean(vector) The result is a scalar (a length 1 vector) that is the
mean of the elements of vector (elements values added,
divided by number of elements).
avg(vector) The average of a vector.
Returns a vector where each element is the mean of the
preceding elements of the input vector (including the
actual element).
stddev(vector) The result is a scalar (a length 1 vector) that is the
standard deviation of the elements of vector .
group_delay(vector) Calculates the group delay −dphase[rad]/dω[rad/s].
Input is the complex vector of a system transfer
function versus frequency, resembling damping and
phase per frequency value. Output is a vector of group
delay values (real values of delay times) versus
frequency.
vector(number) The result is a vector of length number, with elements
0, 1, ... number - 1. If number is a vector then just the
first element is taken, and if it isn’t an integer then the
floor of the magnitude is used.
cvector(number) Return a vector of length number, containing complex
elements, with the real part values increasing from 0 to
number-1, the imaginary values are set to 0.
unitvec(number) The result is a vector of length number, all elements
Name Function
length(vector) The length of vector.
interpolate(plot.vector) The result of interpolating the named vector onto the
scale of the current plot. This function uses the
variable polydegree to determine the degree of
interpolation.
integ(vector) Integrates over the given vector (versus the real
component of the scale vector), using the trapeziodal
method. The result is another vector, showing the
integral up to the current scale value. See also 15.4.8
for measuring the integral sum for a section of a vector,
and 12.2.17 for integration on the fly during a transient
simulation.
deriv(vector) Calculates the derivative of the given vector. This uses
numeric differentiation by interpolating a polynomial.
The degree of the polynomal may be set by the
variable dpolydegree (default is 2). The procedure may
not produce satisfactory results (particularly with
iterated differentiation). The implementation only
calculates the derivative with respect to the real
component of that vector’s scale.
vecd(vector) Compute the differential of a vector.
vecmin(vector) Returns the value of the vector element with minimum
value. Same as minimum.
minimum(vector) Returns the value of the vector element with minimum
value. Same as vecmin.
vecmax(vector) Returns the value of the vector element with maximum
value. Same as maximum.
maximum(vector) Returns the value of the vector element with maximum
value. Same as vecmax.
fft(vector) fast fourier transform (17.5.33)
ifft(vector) inverse fast fourier transform (17.5.33)
sortorder(vector) Returns a vector with the positions of the elements in
a real vector after they have been sorted into
increasing order using a stable method (qsort).
timer(vector) Returns CPU-time minus the value of the first vector
element.
clock(vector) Returns wall-time minus the value of the first vector
element.
Several functions offering statistical procedures are listed in the following table:
17.2. EXPRESSIONS, FUNCTIONS, AND CONSTANTS 373
Name Function
rnd(vector) A vector with each component a random integer
between 0 and the absolute value of the input vector’s
corresponding integer element value.
sgauss(vector) Returns a vector of random numbers drawn from a
Gaussian distribution (real value, mean = 0 , standard
deviation = 1). The length of the vector returned is
determined by the input vector. The contents of the
input vector will not be used. A call to sgauss(0) will
return a single value of a random number as a vector of
length 1.
sunif(vector) Returns a vector of random real numbers uniformly
distributed in the interval [-1 .. 1[. The length of the
vector returned is determined by the input vector. The
contents of the input vector will not be used. A call to
sunif(0) will return a single value of a random number
as a vector of length 1.
poisson(vector) Returns a vector with its elements being integers
drawn from a Poisson distribution. The elements of the
input vector (real numbers) are the expected numbers
λ. Complex vectors are allowed, real and imaginary
values are treated separately.
exponential(vector) Returns a vector with its elements (real numbers)
drawn from an exponential distribution. The elements
of the input vector are the respective mean values (real
numbers). Complex vectors are allowed, real and
imaginary values are treated separately.
An input vector may be either the name of a vector already defined or a floating-point
number (a scalar). A scalar will result in an output vector of length 1. A number may
be written in any format acceptable to ngspice, such as 14.6Meg or -1.231e-4. Note that
you can either use scientific notation or one of the abbreviations like MEG or G, but not
both. As with ngspice, a number may have trailing alphabetic characters.
The notation expr [num] denotes the num’th element of expr. For multi-dimensional
vectors, a vector of one less dimension is returned. Also for multi-dimensional vectors,
the notation expr[m][n] will return the nth element of the mth subvector. To get a subrange
of a vector, use the form expr[lower, upper]. To reference vectors in a plot that is not the
current plot (see the setplot command, below), the notation plotname.vecname can be
used. Either a plotname or a vector name may be the wildcard all. If the plotname is all,
matching vectors from all plots are specified, and if the vector name is all, all vectors in the
specified plots are referenced. Note that you may not use binary operations on expressions
involving wildcards - it is not obvious what all + all should denote, for instance. Some
(contrived) examples of expressions are shown below.
Expressions examples:
cos ( TIME ) + db ( v (3))

sin ( cos ( log ([1 2 3 4 5 6 7 8 9 10])))
TIME * rnd ( v (9)) - 15 * cos ( vin # branch ) ^ [7.9 e5 8]
not (( ac3 . FREQ [32] & tran1 . TIME [10]) gt 3)
( sunif (0) ge 0) ? 1.0 : 2.0
mag ( fft ( v (18)))
Vector names in ngspice may look like @dname[param], where dname is either the name
of a device instance or of a device model. The vector contains the value of the parameter
of the device or model. See Appendix, Chapt. 31 for details of which parameters are
available. The returned value is a vector of length 1. Please note that finding the value
of device and device model parameters can also be done with the show command (e.g.
show v1 : dc).
There are a number of pre-defined constants in ngspice, which you may use by their name.
They are stored in plot (17.3) const and are listed in the table below:
Name Description Value
pi π 3.14159...
e e (the base of natural logarithms) 2.71828...
c c (the speed of light) 299,792,458
√
m/sec
i i (the square root of -1) −1

kelvin (absolute zero in centigrade) -273.15◦ C
echarge q (the charge of an electron) 1.60219e-19 C
boltz k (Boltzmann’s constant) 1.38062e-23J/K
planck h (Planck’s constant) 6.62607e-34 J s
yes boolean 1
no boolean 0
TRUE boolean 1
FALSE boolean 0
These constants are all given in MKS units. If you define another variable with a name
that conflicts with one of these then it takes precedence.
Additional constants may be generated during circuit setup (see .csparam, 2.11).
17.3 Plots
The output vectors of any analysis are stored in plots, a traditional SPICE notion. A
plot is a group of vectors. A first tran command will generate several vectors within
a plot tran1. A subsequent tran command will store their vectors in tran2. Then a
linearize command will linearize all vectors from tran2 and store them in tran3, which
then becomes the current plot. A fft will generate a plot spec1, again now the current
plot. The display command always will show all vectors in the current plot. Echo
$plots followed by Return lists all plots generated so far. Setplot followed by Return
will show all plots and ask for a (new) plot to become current. A simple Return will end
the command. Setplot name will change the current plot to ’name’ (e.g. setplot tran2
17.4. COMMAND INTERPRETATION 375
will make tran2 the current plot). A sequence name.vector may be used to access the
vector from a foreign plot.
You may generate plots by yourself: setplot new will generate a new plot named un-
known1, set curplottitle=”a new plot” will set a title, set curplotname=myplot will
set its name as a short description, set curplotdate=”Sat Aug 28 10:49:42 2010” will
set its date. Note that strings with spaces have to be given with double quotes.
Of course the notion ’plot’ will be used by this manual also in its more common meaning,
denoting a graphics plot or being a plot command. Be careful to get the correct meaning.
17.4 Command Interpretation
17.4.1 On the console
On the ngspice console window (or into the Windows GUI) you may directly type in any
command from 17.5. Within a command sequence, Input/output redirection is available
(see Chapt. 17.8.9 for an example) - the symbols >, >‌>, >&, >‌>&, and < have the same
effects as in the C-shell. This I/O-redirection is internal to ngspice commands, and should
not be mixed up with the ‘external’ I/O-redirection offered by the usual shells (Linux,
MSYS etc.), see 17.5.80.
17.4.2 Scripts
If a word is typed as a command, and there is no built-in command with that name, the
directories in the sourcepath list are searched in order for a file with the name given by
the word. If it is found, it is read in as a input file (as if it were sourced). Such a file will
often be a pure script containing only interpreter commands. Such files can be written to
externd the command set. Full details of scripting are in (17.8).
There are various command scripts installed in /usr/local/lib/spice/scripts (or

whatever the path is on your machine), and the default sourcepath (17.7) includes this
directory, so you can use these command files (almost) like built-in commands.
17.4.3 Add-on to circuit file
Probably the most common way to invoke the commands described in the following Chapt.
17.5 is to add a .control ... .endc section to the circuit input file (see 16.4.3).
Example:
. control
pre_set strict_errorhandling
unset ngdebug
* save outputs and specials
save x1 . x1 . x1 .7 V (9) V (10) V (11) V (12) V (13)
run
display
* plot the inputs , use offset to plot on top of each other
plot v (1) v (2)+4 v (3)+8 v (4)+12 v (5)+16 v (6)+20 v (7)+24 v (8)+28
* plot the outputs , use offset to plot on top of each other
plot v (9) v (10)+4 v (11)+8 v (12)+12 v (13)+16
. endc
17.5 Commands
Commands marked with a * are only available in standard ngspice, not in shared ngspice.
Those marked with ** are available in shared ngspice only.
17.5.1 Ac: Perform an AC, small-signal frequency response anal-

ysis
General Form:
ac ( DEC | OCT | LIN ) N Fstart Fstop
Do an small signal ac analysis (see also Chapt. 15.3.1) over the specified frequency range.
DEC decade variation, and N is the number of points per decade.
OCT stands for octave variation, and N is the number of points per octave.
LIN stands for linear variation, and N is the number of points.
fstart is the starting frequency, and fstop is the final frequency.
Note that in order for this analysis to be meaningful, at least one independent source
must have been specified with an ac value.
In this ac analysis all non-linear devices are linearized around their actual dc operating
point. Each Ls and Cs gets its imaginary value based on the actual frequency step. Each
output vector will be calculated relative to the input voltage (current) given by the ac
value (Iin equals to 1 in the example below). The resulting node voltages (and branch
currents) are complex vectors. Therefore you have to be careful using the plot command.
17.5. COMMANDS 377
Example:
* AC test
Iin 1 0 AC 1
R1 1 2 100
L1 2 0 1
. control
AC LIN 101 10 10 K
plot v (2) $ real part !
plot mag ( v (2)) $ magnitude
plot db ( v (2)) $ same as vdb (2)
plot imag ( v (2)) $ imaginary part of v (2)
plot real ( v (2)) $ same as plot v (2)
plot phase ( v (2)) $ phase in rad
plot cph ( v (2)) $ phase in rad , continuous beyond pi
plot 180/ PI * phase ( v (2)) $ phase in degrees
set units = degrees
plot phase ( v (2)) $ phase in degrees
. endc
. end
In addition to the plot examples given above you may use the variants of vxx(node)
described in Chapt. 15.6.2 like vdb(2). If you apply this notion to another plot ac3, the
term vdb(ac3.2) is o.k., however ac3.vdb(2) is not.
An option to suppress OP analysis before AC may be set for linear circuits (15.1.3).
17.5.2 Alias: Create an alias for a command

General Form:
alias [ word ] [ text ...]
Causes word to be aliased to text. History substitutions may be used, as in C-shell

aliases.
17.5.3 Alter: Change a device or model parameter

Alter changes the value for a device or a specified parameter of a device or model.
General Form:
alter dev = < expression >

alter dev param = < expression >
alter @dev [ param ] = < expression >
<expression> must be real (complex isn’t handled right now, integer is fine though, but
no strings. For booleans, use 0/1).
Old style (pre 3f4):
alter device value

alter device parameter value [ parameter value ]
Using the old style, its first form is used by simple devices that have one principal value
(resistors, capacitors, etc.) where the second form is for more complex devices (bjt’s,
etc.). Model parameters can be changed with the second form if the name contains a ‘#’.
For specifying a list of parameters as values, start it with ‘[’, followed by the values in
the list, and end with ‘]’. Be sure to place a space between each of the values and before
and after the ‘[’ and ‘]’.
Some examples are given below:
Examples (Spice3f4 style):
alter vd = 0.1
alter vg dc = 0.6
alter @m1 [ w ]= 15 e -06
alter @vg [ sin ] [ -1 1.5 2 MEG ]
alter @Vi [ pwl ] = [ 0 1.2 100 p 0 ]
alter may have vectors (17.8.2) or variables (17.8.1) as parameters.
Examples (vector or variable in parameter list):
let newfreq = 10 k
alter @vg [ sin ] [ -1 1.5 $ & newfreq ] $ vector
set newperiod = 150 u
alter @Vi [ pwl ] = [ 0 1.2 $newperiod 0 ] $ variable
You may change a parameter of a device residing in a subcircuit, e.g. of MOS transistor
msub1 in subcircuit xm1 (see also Chapt. 31.1).
Examples (parameter of device in subcircuit):
alter m . xm1 . msub1 w = 20 u

alter @m . xm1 . msub1 [ w ] = 20 u
17.5. COMMANDS 379
17.5.4 Altermod: Change model parameter(s)
General form:
altermod mod param = < expression >

altermod @mod [ param ] = < expression >
Example:
altermod nc1 tox = 10 e -9

altermod @nc1 [ tox ] = 10 e -9
Altermod operates on models and is used to change model parameters. The above
example will change the parameter tox in all devices using the model nc1, which is
defined as
*** BSIM3v3 model

.MODEL nc1 nmos LEVEL=8 version = 3.2.2
+ acm = 2 mobmod = 1 capmod = 1 noimod = 1
+ rs = 2.84E+03 rd = 2.84E+03 rsh = 45
+ tox = 20E-9 xj = 0.25E-6 nch = 1.7E+17
+ ...
If you invoke the model by the MOS device
M1 d g s b nc1 w=10u l=1u
you might also insert the device name M1 for mod as in
altermod M1 tox = 10e-9
The model parameter tox will be modified, however not only for device M1, but for all
devices using the associated MOS model nc1!
If you want to run corner simulations within a single simulation flow, the following option
of altermod may be of help. The existing models are defined during circuit setup at start
up of ngspice. Model parameter sets have been included by .model statements (2.4) in
your input file or included by the .include command. The parameter set with name
nc1 may be overrun by the altermod command specifying a model file. All parameter
values fitting to the existing model nc1 will be modified. As usual the ’reset’ command
(see 17.5.65) restores the original values. The model file (see 2.4) has to use the standard
specifications for an input file, the .model section is the relevant part. However the first
line in the model file will be ignored by the input parser, so it should contain only some
title information. The .model statement should appear then in the second or any later
line. More than one .model section may reside in the file.
General form:
altermod mod1 [ mod2 .. mod15 ] file = < model file name >
altermod mod1 [ mod2 .. mod15 ] file < model file name >
Example:
altermod nc1 file = BSIM3_nmos . mod

altermod nc1 pc1 file BSIM4_mos . mod
Be careful that the new model file corresponds to the existing model selected by token
nc1. In the example given above, the models nc1 (or nc1 and pc1) have to be already
included in the netlist before calling altermod. If they are not found in the active circuit,
ngspice will terminate with an error message. The file BSIM3_nmos.mod has to include
a .model line starting with .MODEL nc1 nmos.... There is no checking however of the
version and level parameters! So you have to be responsible for offering model data of
the same model name (nc1) and level (e.g. level 8 for BSIM3). Thus no new model is
selectable by altermod, but the parameters of the existing model(s) (here nc1 and pc1)
may be changed (partially, completely, temporarily).
17.5.5 Alterparam: Change value of a global parameter

General form:
alterparam paramname = pvalue

alterparam subname paramname = pvalue
Example (global, top level parameter):
. param npar = 5
...
alterparam npar = 7 $ change npar from 5 to 7
reset
Example (parameter in a subcircuit):
. subckt sname
. param subpar = 13
...
. ends
...
alterparam sname subpar = 11 $ change subpar from 13 to 11
reset
Alterparam operates on global parameters or on parameters in a subcircuit defined by

the .param ... statement. A subsequent call to reset (17.5.65) is required for the
17.5. COMMANDS 381
parameter value change to become effective.
17.5.6 Asciiplot: Plot values using old-style character plots

General Form:
asciiplot plotargs
Produce a line printer plot of the vectors. The plot is sent to the standard output, or
you can put it into a file with asciiplot args ... > file. The set options width, height, and
nobreak determine the width and height of the plot, and whether there are page breaks,
respectively. The ’more’ mode is the standard mode if printing to the screen, that is after
a number of lines given by height, and after a page break printing stops with request
for answering the prompt by <return>, ’c’ or ’q’. If everything shall be printed without
stopping, put the command set nomoremode into .spiceinit 16.6 (or spinit 16.5). Note
that you will have problems if you try to asciiplot something with an X-scale that isn’t
monotonic (i.e, something like sin(TIME) ), because asciiplot uses a simple-minded
linear interpolation. The asciiplot command doesn’t deal with log scales or the delta
keywords.
17.5.7 Aspice*: Asynchronous ngspice run

General Form:
aspice input - file [ output - file ]

Start an ngspice run, and when it is finished load the resulting data. The raw data is
kept in a temporary file. If output-file is specified then the diagnostic output is directed
into that file, otherwise it is thrown away.
17.5.8 Bg_ctrl**: suspend running controls until bg_run has

finished
General Form:
bg_ctrl
Create a suspended thread to start any control commands only when bg_run has finished.
This may be achieved also by issuing set controlswait in the beginning of a .control
section.
17.5.9 Bg_halt**: halt a run

General Form:
bg_halt
Halt a run which has been started by bg_run. There may be conditions where this
command cannot be executed immediately.
17.5.10 Bg_run**: Run analysis from the input file in the back-
ground thread
General Form:
bg_run
Run the simulation as specified in the input file in the second (background) thread of
shared ngspice. If there were any of the control lines .ac, .op, .tran, or .dc, they
are executed. The output is available in plots and their vectors, and/or in the API via
callback function SendData (19.3.3.4).
17.5.11 Bug: Output URL for ngspice bug tracker
General Form:
bug
Get URL to file a bug report. Please go the the URL provided by this command when you
have a bug report to file. Include a short summary of the problem, the version number
and name of the operating system that you are running, the version of ngspice that you
are running, and any relevant ngspice input and output files.
17.5.12 Cd: Change directory
General Form:
cd [ directory ]
Change the current working directory to directory, or to the user’s home directory (Linux:
HOME, MS Windows: USERPROFILE), if none is given.
17.5.13 Cdump: Dump the control flow to the screen
General Form:
cdump
Dumps the control sequence to the screen (all statements inside the .control ... .endc
structure before the line with cdump). Indentations show the structure of the sequence.
The example below is printed if you add cdump to /examples/Monte_Carlo/MonteCarlo.sp.
17.5. COMMANDS 383
Example (abbreviated):
let mc_runs =5
let run =0
...
define agauss ( nom , avar , sig ) ( nom + avar / sig * sgauss (0))
define limit ( nom , avar ) ( nom + (( sgauss (0) >=0) ? avar : - avar ))
dowhile run < mc_runs
alter c1 = unif (1 e -09 , 0.1)
...
ac oct 100 250 k 10 meg
meas ac bw trig vdb ( out ) val = -10 rise =1 targ vdb ( out )
+ val = -10 fall =1
set run =" $ & run "
...
let run = run + 1
end
plot db ({ $scratch }. allv )
echo
print { $scratch }. bwh
cdump
17.5.14 Circbyline: Enter a circuit line by line
General Form:
circbyline line
Enter a circuit line by line. line is any circuit line, as found in the *.cir ngspice input
files. The first line is a title line. The entry will be finished by entering .end. Circuit
parsing is then started automatically.
Example:
circbyline test circuit

circbyline v1 1 0 1
circbyline r1 1 0 1
circbyline . dc v1 0.5 1.5 0.1
circbyline . end
run
plot i ( v1 )
17.5.15 Codemodel: Load an XSPICE code model library

General Form:
codemodel [ library file ]
Load a XSPICE code model shared library file (e.g. analog.cm ...). Only available if
ngspice is compiled with the XSPICE option (--enable-xspice) or with the Windows
executable distributed since ngspice21. This command has to be called from spinit (see
Chapt. 16.5) (or .spiceinit for personal code models, 16.6).
17.5.16 Compose: Compose a vector

General form 1 - List of values:
compose name values value1 [ value2 ... ]
General forms 2 - Linearly spaced values:
compose name start = val stop = val step = val

compose name center = val span = val step = val
compose name lin = val center = val span = val
compose name lin = val < start = val > < stop = val > < step = val >
General forms 3 - Logarithmically spaced values:
compose name ( log = val | dec = val | oct = val ) start = val stop = val
compose name ( log = val | dec = val | oct = val ) center = val span = val
General form 4 - Gaussian distributed values:
compose name gauss = val < mean = val > < sd = val >
General forms 5 - Uniformly distributed values:
compose name unif = val < mean = val > < span = val >
compose name unif = val start = val stop = val
General form 6 - XSPICE node history:
compose name xspice event - node
The general form 1 takes the values and creates a new vector, where the values may
be arbitrary expressions. If negative numbers or expressions starting with ’-’ are to be
entered, put them into brackets, e.g. (-2.364) or (-5*PI).
17.5. COMMANDS 385
The forms 2 - 5 create a new vector according the following possible parameters:
start Value of name[0] (default: 0)
stop Last value of name
step Difference between successive elements of the linearly spaced vector (default: 1)
lin Number of points, linearly spaced
log Number of points, logarithmically spaced
dec Number of points per decade, logarithmically spaced
oct Number of points per octave, logarithmically spaced
center Where to center the range of points
span Size of the range of points (default for uniform distribution: 1)
gauss Number of points, Gaussian distributed
mean Mean value of the Gaussian (default 0) or uniform distribution (default 0.5)
sd Standard deviation for the Gaussian distribution (default 1)
unif Number of points, uniformly distributed
Form 6 creates a vector fron the saved history of an XSPICE event node with similar
results to plotting or printing an event node.
17.5.17 Cutout: Cut out a section of all vectors in a tran plot
General Form:
let cut - tstart = time1

let cut - tstop = time2
cutout
Cut out part of each vector of the current tran plot, from times cut-tstart to cut-tstop
and copy these into a new tran plot. A new scale vector ’time’ will be generated as well.
Vectors that are shorter than the new scale vector will not be copied. If cut-start or
cut-stop are not given, the starting or end times of the current plot are used.
So the simple command cutout may be used to get rid of 0-length vectors in a new tran
plot that may occur if for example something like generating m1[id] is not served in an
AC simulation.
17.5.18 Dc: Perform a DC-sweep analysis
General Form:
dc Source Vstart Vstop Vincr [ Source2 Vstart2 Vstop2 Vincr2 ]
Do a dc transfer curve analysis. See the previous Chapt. 15.3.2 for more details. Several
options may be set (15.1.2).
17.5.19 Define: Define a function

General Form:
define function ( arg1 , arg2 , ...) expression

Define the function with the name function and arguments arg1, arg2, ... to be expression,
which may involve the arguments. When the function is later used, the arguments it is
given are substituted for the formal parameters when it was parsed. If expression is not
present, any existing definition for function is printed, and if there are no arguments then
expressions for all currently active definitions are printed. Note that you may have dif-
ferent functions defined with the same name but different arities. Some useful definitions
are
Example:
define max (x , y ) ( x > y ) * x + ( x <= y ) * y

define min (x , y ) ( x < y ) * x + ( x >= y ) * y
define limit ( nom , avar ) ( nom + (( sgauss (0) >= 0) ? avar : - avar ))
When defining the function, the tokens used (here x, y, nom, or avar) should not have
been defined elsewhere, e.g. as vectors. Otherwise strange errors may result.
17.5.20 Deftype: Define a new type for a vector or plot

General Form:
deftype [ v | p ] typename abbrev
defines types for vectors and plots. abbrev will be used to parse things like abbrev(name)
and to label axes with M<abbrev>, instead of numbers. Also, the command ‘deftype p
plottype pattern ...’ will assign plottype as the name for any plot with one of the patterns
in its Name: field.
Example:
deftype v capacitance F
settype capacitance moscap
plot moscap vs v ( cc )
17.5.21 Delete: Remove a trace or breakpoint

General Form:
delete [ debug - number ... ]

Delete the specified saved nodes and parameters, breakpoints and traces. The debug
numbers are those shown by the status command (unless you do status > file, in which
case the debug numbers are not printed).
17.5. COMMANDS 387
17.5.22 Destroy: Delete an output data set

General Form:
destroy [ plotnames | all ]
Release the memory holding the output data (the given plot or all plots) for the specified
runs. The initial plot, "const", can not be destroyed.
17.5.23 Devhelp: information on available devices

General Form:
devhelp [ - csv ] [ - type ] [ - flags ] [ device_name [ parameter ]]
Devhelp command shows the user information about the devices available in the simulator.
If called without arguments, it simply displays the list of available devices in the simulator.
The name of the device is the name used inside the simulator to access that device. If
the user specifies a device name, then all the parameters of that device (model and
instance parameters) will be printed. Parameter description includes the internal ID of
the parameter (id#), the name used in the model card or on the instance line (Name),
the direction (Dir) and the description of the parameter (Description). All the fields are
self-explanatory, except the ‘direction’. Direction can be in, out or inout and corresponds
to a ‘write-only’, ‘read-only’ or a ‘read/write’ parameter. Read-only parameters can be
read but not set, write only can be set but not read and read/write can be both set and
read by the user.
The -type option prints the type of each parameter, for example real, integer, string.
Values ending with vec indicate vectors.
The -csv option prints the fields, separated by a commas, for direct import into a spread-
sheet. This option is used to generate the simulator documentation.
The -flags option prints the internal Spice flags for each parameter. A specific string is
appended to the result for each flag:
X the parameter is not used in sensitivity analysis.
Q the parameter must be non-zero for sensitivity analysis of the subsequent parameter.
Z the previous parameter must be non-zero for sensitivity analysis.
QO Like Q, but or-ed with the previous Q value.
A the parameter is significant for time-varying (non-DC) analyses.
P the parameter is a principal of the device. Used for naming output variables in sensi-
tivity.
AA the parameter is significant for AC analyses only.
N the parameter is significant for noise analyses only.

U the parameter is not shown in the default show command output.
R redundant parameter name (e.g.vto vs.vt0).
Example:
devhelp
devhelp resistor
devhelp capacitor ic
devhelp - flags - type bjt
17.5.24 Diff: Compare vectors

General Form:
diff plot1 plot2 [ vec ...]
Compare all the vectors in the specified plots, or only the named vectors if any are given. If
there are different vectors in the two plots, or any values in the vectors differ significantly,
the difference is reported. The variables diff_abstol, diff_reltol, and diff_vntol are used to
determine a significant difference.
17.5.25 Display: List known vectors and types

General Form:
display [ varname ...]
Prints a summary of currently defined vectors, or of the names specified. The vectors
are sorted by name unless the variable nosort is set. The information given is the name
of the vector, the length, the type of the vector, and whether it is real or complex data.
Additionally, one vector is labeled [scale]. When a command such as plot is given without
a vs argument, this scale is used for the X-axis. It is always the first vector in a rawfile,
or the first vector defined in a new plot. If you undefine the scale (i.e, let TIME = []), one
of the remaining vectors becomes the new scale (which one is unpredictable). You may
set the scale to another vector of the plot with the command setscale (17.5.77).
17.5.26 Echo: Print text

General Form:
echo [ - n ] [ text | $variable | $ & vector ] ...
Echos all text, variables and vectors to the screen or the redirected output location. If
-n included as the first argument, a newline will not be printed. Otherwise one will be
appended to the output.
17.5. COMMANDS 389
17.5.27 Edit*: Edit the current circuit
General Form:
edit [ file - name ]
Print the current ngspice input file into a file, call up the editor on that file and allow the
user to modify it, and then read it back in, replacing the original file. If a file-name is
given, then edit that file and load it, making the circuit the current one. The editor may
be defined in .spiceinit or spinit by a command line like
set editor=emacs
Using MS Windows, to allow the edit command calling an editor, you will have to add
the editor’s path to the PATH variable of the command prompt windows (see here). edit
then calls cmd.exe with e.g. notepad++ and file-name as parameter, if you have set
set editor=notepad++.exe
in .spiceinit or spinit.
17.5.28 Edisplay: Print a list of all the event nodes
General Form:
edisplay
Print the node names, node types, and number of events per node of all event driven
nodes generated or used by XSPICE ’A’ devices. See eprint, eprvcd, and 27.2.2 for an
example.
17.5.29 Eprint: Print an event driven node
General Form:
eprint node [ node ]

eprint node [ node ] > nodeout . txt $ output redirected
Print an event driven node generated or used by an XSPICE ’A’ device. These nodes are
vectors not organized in plots. See edisplay, eprvcd, and Chapt. 27.2.2 for an example.
Output redirection into a file is available.
17.5.30 Eprvcd: Dump nodes in VCD format
General Form:
eprvcd [ - t unit ][ - a ] node1 node2 .. noden [ > filename ]

Dump the data of the specified event driven nodes and others to a .vcd file (see also
18.6.1.4). Such files may be viewed with an vcd viewer, for example gtkwave. Values
for analog nodes and expressions (as for plot ) may be included, but expressions may
not contain spaces. Option “-t” sets the VCD file’s time unit; values are rounded to a
negative power of 10. If not used the unit is chosen to fit the simulation’s maximum
timestep. Analog values are examined only when an XSPICE event values changes unless
option “-a” is used to dump them at each timestep. Also see edisplay, eprint.
17.5.31 Esave: Save a set of event node outputs

General Form:
esave all | none | node ...
Save a set of event node outputs, discarding the rest (if keyword all is not given). May be
used to dramatically reduce memory (RAM) requirements when only a few useful nodes’
events are saved.
For backward compatibility, if there are no esave commands given, all outputs are saved.
If you want to eprint (17.5.29) or eprvcd (17.5.30) a node, you have to save it explicitly,
except for all given or no save command at all.
Don’t save anything:
esave none
Useful if you do not need to examine the event data separately from the normal plot.
17.5.32 Fclose: close an open file handle

General Form:
fclose handle
This command closes an open file identified by the integer ’handle’. It ignores values less
than 3 and any file that was not opened by fopen or read by fread.
17.5.33 FFT: fast Fourier transform of vectors

General Form:
fft vector1 [ vector2 ] ...
This analysis provides a fast Fourier transform of the input vector(s) in forward direction.
fft is much faster than spec (17.5.88) (about a factor of 50 to 100 for larger vectors).
The fft command will create a new plot consisting of the Fourier transforms of the
vectors given on the command line. Each vector given should be a transient analysis
17.5. COMMANDS 391
result, i.e. it should have time as a scale. You will have gotten these vectors by the tran
Tstep Tstop Tstart command.
The vector should have a linear equidistant time scale. Therefore linearization using the
linearize command is recommended before running fft. Be careful selecting a Tstep
value small enough for good interpolation, e.g. much smaller than any signal period
to be resolved by fft (see linearize command). The Fast Fourier Transform will be
computed using a window function as given with the specwindow variable. A new plot
named specx will be generated with a new vector (having the same name as the input
vector, see command above) containing the transformed data.
Ngspice has two FFT implementations:
1. Standard code is based on the FFT function provided by John Green ‘FFTs for RISC
2.0‘, downloaded 2012, now to be found here. These are a power-of-two routines for
fft and ifft. If the input size doesn’t fit this requirement the remaining data will be
zero padded up to the next 2N field size. You have to take care of the correlated
change in the scale vector.
2. If available on the operating system (see Chapter 32) ngspice can be linked to
the famous FFTW-3 package, found here. This high performance package has
advantages in speed and accuracy compared to most of the freely available FFT
libraries. It makes arbitrary size transforms for even and odd data.
How to compute the fft from a transient simulation output:
ngspice 8 -> setplot tran1

ngspice 9 -> linearize V (2)
ngspice 9 -> set specwindow = blackman
ngspice 10 -> fft V (2)
ngspice 11 -> plot mag ( V (2))
Linearize will create a new vector V(2) in a new plot tran2. The command fft V(2)
will create a new plot spec1 with vector V(2) holding the resulting data.
The variables listed in the following table control operation of the fft command. Each
can be set with the set command before calling fft.
specwindow: This variable is set to one of the following strings, which will determine
the type of windowing used for the Fourier transform in the spec and fft command. If
not set, the default is hanning.
none No windowing
rectangular Rectangular window
bartlet Bartlett (also triangle) window
blackman Blackman window
hanning Hanning (also hann or cosine) window

hamming Hamming window

gaussian Gaussian window
flattop Flat top window
Figure 17.1: Spec and FFT window functions (Gaussian order = 4)
specwindoworder: This can be set to an integer in the range 2-8. This sets the order
when the Gaussian window is used in the spec and fft commands. If not set, order 2 is
used.
17.5.34 Fopen: open a text file

General Form:
fopen handle file_name [ mode ]
The named file is opened and a numeric handle is returned in variable ’handle’, or -1 on
error. This is a simple wrapper around the standard C-library function with the same
name, so the meaning of string ’mode’ is as defined by your OS documentation. By default
the file is opened for reading only. If interpreter variable "silent_fileio" is set, no message
is printed on error.
17.5.35 Fourier: Perform a Fourier transform

General Form:
fourier fundamental_frequency [ expression ...]

17.5. COMMANDS 393
Fourier is used to analyze the output vector(s) of a preceding transient analysis (see
17.5.98). It does a Fourier analysis of each of the given values, using the first 10 multiples
of the fundamental frequency (or the first nfreqs multiples, if that variable is set (see
17.7). The printed output is like that of the .four ngspice line (Chapt. 15.6.4). The
expressions may be any valid expression (see 17.2), e.g. v(2). The evaluated expression
values are interpolated onto a fixed-space grid with the number of points given by the
fourgridsize variable, or 200 if it is not set. The interpolation is of degree polydegree
if that variable is set, or 1 otherwise. If polydegree is 0, then no interpolation is done.
This is likely to give erroneous results if the time scale is not monotonic.
The fourier command not only issues a printout, but also generates vectors, one per
expression. The size of the vector is 3 x nfreqs (per default 3 x 10). The name of the
new vector is fouriermn, where m is set by the mth call to the fourier command, n is
the nth expression given in the actual fourier command. fouriermn[0] is the vector of
the 10 (nfreqs) frequency values, fouriermn[1] contains the 10 (nfreqs) magnitude values,
fouriermn[2] the 10 (nfreqs) phase values of the result.
Example:
* do the transient analysis

tran 1 n 1 m
* do the fourier analysis
fourier 3.34 e6 v (2) v (3) $ first call
fourier 100 e6 v (2) v (3) $ second call
* get individual values
let newt1 = fourier11 [0][1]
* plot magnitude of second expression ( v (3))
* from first call versus frequency
plot fourier12 [1] vs fourier12 [0]
The plot command from the example plots the vector of the magnitude values, obtained
by the first call to fourier and evaluating the first expression in this call, against the vector
of the frequency values.
17.5.36 Fread: read into a variable from a text file

General Form:
fread result handle [ length ]
This command sets the string variable ’result’ by reading one line from the open file
specified by the integer ’handle’. Terminating characters are stripped and the length
returned in variable ’length’, if given. The handle will usually have been set by the fopen
command, but any valid file descriptor may be used.
The length will be -1 if attempting to read at end-of-file or -2 on error. If interpreter

variable "silent_fileio" is set, no message is printed on error.
17.5.37 Getcwd: Print the current working directory
General Form:
getcwd
Print the current working directory.
17.5.38 Gnuplot: Graphics output via gnuplot
General Form:
gnuplot file plotargs
Like plot, but using gnuplot for graphics output and further data manipulation. ngspice
creates a file called file.plt containing the gnuplot command sequence, a file called file.data
containing the data to be plotted. On Linux, gnuplot may be called directly or via called
via xterm, and offers a Gnuplot console to manipulate the data. On Windows, a plot
window is opened and the command console window is available with a mouse click. Of
course you have to have gnuplot installed on your system. Please see chapter 18.7 for
more details.
17.5.39 Hardcopy: Save a plot to a file for printing
General Form:
hardcopy file plotargs
Just like plot, except that it creates a file called file containing the plot. Various out-
put formats are available, depending on the variable hcopydevtype. It may be set to
postscript or svg. See also Chapt. 18.6 for more details (color etc.).
17.5.40 Help: Print summaries of Ngspice commands
Prints help. This help information, however, is spice3f5-like, stemming from 1991 and
thus is outdated. If commands are given, descriptions of those commands are printed.
Otherwise help for only a few major commands is printed. On Windows, this help
command only provides a link to documentation. Spice3f5 compatible help may be found
in the Spice 3 User manual. For ngspice, please use this manual.
17.5. COMMANDS 395
17.5.41 History: Review previous commands
General Form:
history [ - r ] [ number ]
Print out the history of the last (first if -r is specified) number commands typed at the
keyboard.
A history substitution enables you to reuse a portion of a previous command as you type
the current command. History substitutions save typing. This feature is disabled by de-
fault, as it may interfere with use of ’!’ in expressions. To enable, set variable histsubst.
A history substitution normally starts with a ’!’. A history substitution has three parts:
an event that specifies a previous command, a selector that selects one or more words of
the event, and some modifiers that modify the selected words. The selector and modifiers
are optional. A history substitution has the form ![event][:]selector[:modifier]
. . . ] The event is required unless it is followed by a selector that does not start with a
digit. The ’:’ can be omitted before the selector if this selector does not begin with a
digit. History substitutions are interpreted before anything else — even before quotations
and command substitutions. The only way to quote the ’!’ of a history substitution is
to escape it with a preceding backslash. A ’!’ need not be escaped if it is followed by
whitespace, ’=’, or ’(’.
Ngspice saves each command that you type in a history list, provided that the command
contains at least one word. The commands in the history list are called events. The
events are numbered, with the first command that you issue when you start Ngspice
being number one. The history variable specifies how many events are retained in the
history list.
These are the forms of an event in a history substitution:
!! The preceding event. Typing ’!!’ is an easy way to reissue the previous
command.
!n Event number n.
!-n The n th previous event. For example, !-1 refers to the immediately preceding
event and is equivalent to !!.
!str The unique previous event whose name starts with str.
!?str[?] The unique previous event containing the string str. The closing ’?’ can be
omitted if it is followed by a newline.
You can modify the words of an event by attaching one or more modifiers. Each modifier
must be preceded by a colon. The following modifiers assume that the first selected word
is a file name:
:r Removes the trailing .str extension from the first selected word.
:h Removes a trailing path name component from the first selected word.
:t Removes all leading path name components from the first selected word.
:e Remove all but the trailing suffix.
:p Print the new command but do not execute it.
s/old/new Substitute new for the first occurrence of old in the event line. Any delimiter may
be used in place of ‘/’. The delimiter may be quoted in old and new with a single
backslash. If ‘&’ appears in new, it is replaced by old. A single backslash will quote
the ‘&’. The final delimiter is optional if it is the last character on the input line.
& Repeat the previous substitution.
g a Cause changes to be applied over the entire event line. Used in conjunction with
‘s’, as in gs/old/new/, or with ‘&’.
G Apply the following ‘s’ modifier once to each word in the event.
For example, if the command ls /usr/elsa/toys.txt has just been executed, then the com-
mand echo !!^:r !!^:h !!^:t !!^:t:r produces the output /usr/elsa/toys /usr/elsa toys.txt toys
. The ’^’ command is explained in the table below.
You can select a subset of the words of an event by attaching a selector to the event. A
history substitution without a selector includes all of the words of the event. These are
the possible selectors for selecting words of the event:
:0 The command name
[:]^ The first argument
[:]$ The last argument
:n The n th argument (n ≥ 1)
:n1-n2 Words n1 through n2
[:]* Words 1 through $
:x* Words x through $
:x- Words x through ($ - 1)
[:]-x Words 0 through x
[:]% The word matched by the last ?str? search used
The colon preceding a selector can be omitted if the selector does not start with a digit.
The following additional special conventions provide abbreviations for commonly used
forms of history substitution:
• An event specification can be omitted from a history substitution if it is followed by

a selector that does not start with a digit. In this case the event is taken to be the
event used in the most recent history reference on the same line if there is one, or
the preceding event otherwise. For example, the command echo !?qucs?^ !$ echoes
the first and last arguments of the most recent command containing the string qucs
.
• If the first non-blank character of an input line is ’^’, the ’^’ is taken as an abbre-
viation for !:s^ . This form provides a convenient way to correct a simple spelling
error in the previous line. For example, if by mistake you typed the command cat
/etc/lasswd you could re-execute the command with lasswd changed to passwd by
typing ^l^p .
17.5. COMMANDS 397
• You can enclose a history substitution in braces to prevent it from absorbing the
following characters. In this case the entire substitution except for the starting
’!’ must be within the braces. For example, suppose that you previously issued
the command cp accounts ../money . Then the command !cps looks for a previous
command starting with cps while the command !{cp}s turns into a command cp
accounts ../moneys .
Some characters are handled specially as follows:
~ Expands to the home directory

* Matches any string of characters in a filename
? Matches any single character in a filename
[] Matches any of the characters enclosed in a filename
- Used within [] to specify a range of characters. For example, [b-k] matches on any
character between and including ‘b’ through to ‘k’.
^ If the ^ is included within [] as the first character, then it negates the following
characters matching on anything but those. For example, [âgm] would match on
anything other than ‘a’, ‘g’ and ‘m’. [â-zA-Z] would match on anything other
than an alphabetic character.
The wildcard characters *, ?, [, and ] can be used, but only if you unset noglob first. This
makes them rather useless for typing algebraic expressions, so you should set noglob again
after you are done with wildcard expansion.
When the environment variable HOME exists (on Unix, Linux, or CYGWIN), history
permanently stores previous command lines in the file $HOME/._ngspice_history. When
this variable does not exist (typically on Windows when the readline library is not officially
installed), the history file is called .history and put in the current working directory.
The history command is part of the readline or editline package. The readline pro-
gram provides a command line editor that is configurable through the file .inputrc.
The path to this configuration file is either found in the shell variable INPUTRC, or
it is (on Unix/Linux/CYGWIN) the file ~/.inputrc in the user’s home directory. On
Windows systems, the configuration file is /Users/<username>/.inputrc, unless the read-
line library was officially installed. In that case the filename is taken from the Win-
dows registry and points to a location that the user specified during installation. See
https://web.archive.org/web/20190527085247/https://tiswww.case.edu/php/chet/readline/rltop.html
for detailed documentation. Some useful commands are below.
Left/Right arrow Move one character to the left or right
Home/End Move to beginning or end of line
Up/Down arrow Cycle through the history buffer
C-_- Undo last editing command
C-r Incremental search backward
TAB completion of a file name
C-ak Erase the command line (kill)
C-y Retrieve last kill (yank)
C-u Erase from cursor to start of line
17.5.42 Inventory: Print circuit inventory

General Form:
inventory
This commands accepts no argument and simply prints the number of instances of a
particular device in a loaded netlist.
17.5.43 Iplot*: Incremental plot

General Form:
iplot [ - d delay ] [ - w width ][ node ...]
Incrementally plot the values of the nodes while ngspice runs. The iplot command can be
used with the where command to find trouble spots in a transient simulation. The “-d”
options sets the delay (in simulation steps) between the start of the simulation and the
appearance of the window. It can be used to suppress flicker when new values cause rapid
resizing at the start of the simulation. The “-w” option sets a fixed width for the window
in simulation units (time, frequency etc). When the output does not fit, only the latest
output values are shown.
The @name[param] notation (31.1) and XSPICE event nodes do not work yet.
17.5.44 Jobs*: List active asynchronous ngspice runs

General Form:
jobs
Report on the asynchronous ngspice jobs currently running. Ngnutmeg checks to see if
the jobs are finished every time you execute a command. If it is done then the data is
loaded and becomes available.
17.5.45 Let: Assign a value to a vector

General Form:
let name = expr
Creates a new vector called name with the value specified by expr, an expression as
described above. If expr is [] (a zero-length vector) then the vector becomes undefined.
Individual elements of a vector may be modified by appending a subscript to name (ex.
name[0]). If there are no arguments, let is the same as display.
The command let creates a vector in the current plot. Use setplot (17.5.76) to create a
new plot.
17.5. COMMANDS 399
There is no straightforward way to initialize a new vector. In general, one might want
to have let initialize a slice (i.e. name[4:4,21:23] = [ 1 2 3 ]) of a multi-dimensional
matrix of arbitrary type (i.e. real, complex ..), where all values and indexes are arbitrary
expressions. This will fail. The procedure is to first allocate a real vector of the appropriate
size with either vector(), unitvec(), or [ n1 n2 n3 ... ]. The second step is to
optionally change the type of the new vector (to complex) with the j() function. The
third step reshapes the dimensions, and the final step (re)initializes the contents, like so:
let a = j(vector(10))
reshape a [2][5]
let a[0][0] = (pi,pi)
Initialization of real vectors can be done quite efficiently with compose:
compose a values (pi, pi) (1,1) (2,sqrt(7)) (boltz,e)

reshape a [2][2]
See also unlet (17.5.102), compose (17.5.16).
17.5.46 Linearize: Interpolate to a linear scale
General Form:
linearize vec ...
Create a new plot with all of the vectors in the current plot, or only those mentioned as
arguments to the command, all data linearized onto an equidistant time scale.
How to compute the fft from a transient simulation output:
ngspice 8 -> setplot tran1

ngspice 9 -> linearize V (2)
ngspice 9 -> set specwindow = blackman
ngspice 10 -> fft V (2)
ngspice 11 -> plot mag ( V (2)) tstep
Linearize will redo the vectors vec or renew all vectors of the current plot (e.g. tran3)
if no arguments are given and store them into a new plot (e.g. tran4). The new vectors
are interpolated onto a linear time scale, which is determined by the values of tstep,
tstart, and tstop in the currently active transient analysis. The currently loaded input
file must include a transient analysis (a tran command may be run interactively before the
last reset, alternately), and the current plot must be from this transient analysis. The
length of the new vector is floor((tstop - tstart) / tstep + 1.5). This command
is needed for example if you want to do an FFT analysis (17.5.33). Please note that the
parameter tstep of your transient analysis (see Chapt. 15.3.10) has to be small enough
to get adequate resolution, otherwise the command linearize will do sub-sampling of
your signal. If no circuit is loaded and the data have been acquired by the load (17.5.48)
command, Linearize will take time data from transient analysis scale vector.
The linearize command may be used to create a linearized cutout of the original vec-
tor by defining the vectors lin-tstart, lin-tstop, and lin-tstep before issuing the
linearize command. At least lin-tstart or lin-tstop has to be defined. This may be
used to plot just a portion of a vector, or to prepare a better fft by skipping the start-up
phase of a ring oscillator.
Excerpt from the ring oscillator example soi/ring51_40.sp:
* original time scale by . tran command is from 0 to 16 ns

save out25 out50
run
plot out25 out50
let lin - tstart = 4 n $ skip the start - up phase
let lin - tstop = 14 n $ end earlier ( just for demonstration )
let lin - tstep = 5 p
linearize out25 out50
plot out25 out50
The linearize command should explicitly name the vectors of interest. Otherwise warn-
ing messages pop up that the vectors lin-tstart etc cannot be linearized.
17.5.47 Listing: Print a listing of the current circuit

General Form:
listing [ logical ] [ physical ] [ deck ] [ expand ] [ runnable ] [ param ]
If the logical argument is given, the listing is with all continuation lines collapsed into
one line, and if the physical argument is given the lines are printed out as they were
found in the file. The default is logical. A deck listing is just like the physical listing,
except without the line numbers it recreates the input file verbatim (except that it does
not preserve case). If the word expand is present, the circuit is printed with all subcir-
cuits expanded. Argument runnable will list the circuit netlist expanded, but without
additional line numbers, ready to be sourced again and run in ngspice. The option param
allows printing all parameters and their actual values.
Example:
source d :\ myngspice \ inputs \ decade_counter . cir

listing r > $inputdir / decade_counter_runnable . cir
All options (except for param) may be invoked by just entering their first letter. The
example sources a ngspice netlist, the listing r command will save the expanded netlist
(all parameters evaluated, subcircuits flattened, .control sections removed) into a file
within the same directory.
If you are using CIDER (30), listing r will not create a runnable netlist, because some
data lines which have been created internally are missing.
17.5. COMMANDS 401
17.5.48 Load: Load rawfile data
General Form:
load [ filename ] ...
Loads either binary or ascii format rawfile data from the files named. The default file
name is rawspice.raw, or the argument to the -r flag if there was one.
17.5.49 Mc_source: Reload the circuit netlist from an internal

storage
General Form:
mc_source
Upon reading a netlist, after its preprocessing is finished, the modified netlist is stored
internally. This command will reload this netlist and create a new circuit inside ngspice.
This command is used in conjunction with the alterparam command.
17.5.50 Meas: Measurements on simulation data
General Form (example):
MEAS { DC | AC | TRAN | SP } result TRIG trig_variable VAL = val < TD = td >

< CROSS =# | CROSS = LAST > < RISE =#| RISE = LAST > < FALL =#| FALL = LAST >
< TRIG AT = time > TARG targ_variable VAL = val < TD = td >
< CROSS =# | CROSS = LAST > < RISE =#| RISE = LAST >
< FALL =#| FALL = LAST > < TRIG AT = time >
Most of the input forms found in 15.4 may be used here with the command meas instead of
.meas(ure). Using meas inside the .control ... .endc section offers additional features
compared to the .meas use. meas will print the results as usual, but in addition will
store its measurement result (typically the token result given in the command line)
in a vector. This vector may be used in following command lines of the script as an
input value of another command. For details of the command see Chapt. 15.4. The
measurement type SP is only available here, because a fft command will prepare the
data for SP measurement. Option autostop (15.1.4) is not available.
Unfortunately par(’expression’) (15.4.10, 15.6.6) and param (15.4.9) will not work here,
i.e. inside the .control section. You may use an expression by the let command (17.5.45)
instead, giving let vec_new = expression.
Replacement for par(’expression’) in meas inside the .control section
let vdiff = v ( n1 ) - v ( n0 )
meas tran vtest find vdiff at =0.04 e -3
* the following will not do here :
* meas tran vtest find par ( ’ v ( n1 ) - v ( n0 ) ’) at =0.04 e -3
17.5.51 Mdump: Dump the matrix values to a file (or to con-

sole)
General Form:
mdump < filename >
If <filename> is given, the output will be stored in file <filename>, otherwise dumped
to your console.
17.5.52 Mrdump: Dump the matrix right hand side values to a

file (or to console)
General Form:
mrdump < filename >
If <filename> is given, the output will be appended to file <filename>, otherwise dumped
to your console.
Example usage after ngspice has started:
* Dump matrix and RHS values after 10 and 20 steps

* of a transient simulation
source rc . cir
step 10
mdump m1 . txt
mrdump mr1 . txt
step 10
mdump m2 . txt
mrdump mr2 . txt
* just to continue to the end
step 10000
You may create a loop using the control structures (Chapt. 17.6).
17.5.53 Noise: Noise analysis

See the .NOISE analysis (15.3.4) for details.
17.5. COMMANDS 403
The noise command will generate two plots (typically named noise1 and noise2) with
Noise Spectral Density Curves and Integrated Noise data. To write these data into output
file(s), you may use the following command sequence:
Command sequence for writing noise data to file(s):
. control
tran 1e -6 1e -3
write test_tran . raw
noise V ( out ) vinp dec 333 1 1 e8 16
print inoise_total onoise_total
* first option to get all of the output ( two files )
setplot noise1
write test_noise1 . raw all
setplot noise2
write test_noise2 . raw all
* second option ( all in one raw - file )
write testall . raw noise1 . all noise2 . all
. endc
17.5.54 Op: Perform an operating point analysis
General Form:
op
Do an operating point analysis. See Chapt. 15.3.5 for more details.
17.5.55 Option: Set a ngspice option
General Form:
option [ option = val ] [ option = val ] ...
Set any of the simulator variables as listed in Chapt. 15.1. See this chapter also for more
information on the available options. The option command without any argument lists
the current options set in the simulator. It shows the current options, while new values
are set to be used in the next analysis run. That means that changed options will not be
visible immediately. Multiple options may be set in a single line.
The following example demonstrates a control section, which may be added to your circuit
file to test the influence of variable trtol on the number of iterations and on the simulation
time.
Command sequence for testing option trtol:
. control
set noinit
option trtol =1
echo
echo trtol =1
run
rusage traniter trantime
reset
option trtol =3
echo
echo trtol =3
run
reset
option trtol =5
echo
echo trtol =5
run
reset
option trtol =7
echo
echo trtol =7
run
plot tran1 . v ( out25 ) tran1 . v ( out50 ) v ( out25 ) v ( out50 )
. endc
17.5.56 Plot*: Plot vectors on the display
General Form:
plot expr1 [ vs scale_expr1 ] [ expr2 [ vs scale_expr2 ]] ...

[ ylimit ylo yhi ] [ xlimit xlo xhi ] [ xindices xilo xihi ]
[ xcompress comp ] [ xdelta xdel ] [ ydelta ydel ]
[ xlog ] [ ylog ] [ loglog ] [ nogrid ]
[ linplot ] [ combplot ] [ pointplot ] [ nointerp ] [ retraceplot ]
[ polar ] [ smith ] [ smithgrid ]
[ xlabel word ] [ ylabel word ] [ title word ]
[ samep ] [ linear ] [ kicad ] [ plainplot ] [ digitop ]
[ all ] [ allv ] [ alli ] [ ally ] [ alle ]
Plot the given vectors or exprs on the screen (if you are on a graphics terminal). The
xlimit and ylimit arguments determine the high and low x- and y-limits of the axes,
17.5. COMMANDS 405
respectively. The xindices arguments determine what range of points are to be plotted
- everything between the xilo’th point and the xihi’th point is plotted. The xcompress
argument specifies that only one out of every comp points should be plotted. If an xdelta
or a ydelta parameter is present, it specifies the spacing between grid lines on the X-
and Y-axis. These parameter names may be abbreviated to xl, yl, xind, xcomp, xdel,
and ydel respectively.
The scal_expr argument(s) are expressions to use as the scale on the x-axis instead
of the default scale for the plot. If xlog or ylog are present, then the X or Y scale,
respectively, are logarithmic (loglog is the same as specifying both). The xlabel and
ylabel arguments cause the specified labels to be used for the X and Y axes, respectively.
If samep is given, the values of the other parameters from the previous plot, hardcopy,
or asciiplot command are used even if they are redefined on the command line.
The title argument is used in the headline of the plot window and replaces the default
text, which is ‘actual plot: first line of input file’.
The linear keyword is used to override a default logscale plot (as in the output for an
AC analysis).
The keywords linplot, combplot and pointplot select different plot styles. The keyword
nointerp turns off interpolation of the vector data, nogrid suppresses the drawing of
gridlines. retraceplot may be required if the scale vector (the x axis) has values which
do not grow monothonically (e.g. plotting a circle or the hyseresis loop of a memristor).
Without this keyword retracing values (x values moving forth and back) are suppressed.
Finally, the keyword polar generates a polar plot. To produce a Smith plot, use the
keyword smith. Note that the data is transformed, so for Smith plots you will see the
data a + jb transformed to
a = (a2 + b2 − 1)/((a + 1)2 + b2 ) (17.1)
b = (2 ∗ b)/((a + 1)2 + b2 ) (17.2)

To produce a polar plot with a Smith grid but without performing the Smith transform,
use the keyword smithgrid.
Keyword retraceplot may be useful if the x-axis values are non-monotonic. Whereas
time is always growing monotonically, during plotting ynew vs xnew xnew may par-
tially increase, then decrease again. If this occurs, plotting is suppressed as per default.
retraceplot will enable plotting all data.
If you specify plot all, all vectors (including the scale vector) are plotted versus the scale
vector (see commands display (17.5.25) or setscale (17.5.77) on viewing the vectors of
the current plot). The command plot ally will not plot the scale vector, but all other
’real’ y values. The command plot alli selects all current vectors, the command plot
allv all voltage vectors.
If the vector name to be plotted contains - , / or other tokens that may be taken for
operators of an expression, and plotting fails, try enclosing the name in double quotes,
e.g. plot “/vout”.
Plotting of complex vectors, as may occur after an ac simulation, requires special consid-
erations. Please see Chapt. 17.5.1 for details.
Keyword kicad will enable plotting vectors with leading character / (see 16.14.8) by plac-
ing double quotes around the token, keyword plainplot will allow this by suppressing the
evaluation of any expression containing such characters. vc1 vs vc2 is not supported with
using plainplot. The same effect may be generated by setting the variable plainplot.
digitop will assemble all digital (event) nodes into a single graph, arranged on top of
each other.
Plot all analog nodes [all], all voltage nodes only [allv], all current nodes, [alli], all
nodes except for the scale [ally], all event nodes [alle].
17.5.57 Pre_<command>: execute commands prior to parsing

the circuit
General Form:
pre_ < command >
All commands in a .control ... .endc section are executed after the circuit has been
parsed. If you need command execution before circuit parsing, you may add these com-
mands to the general spinit or local .spiceinit files. Another possibility is adding a leading
pre_ to a command within the .control section of an ordinary input file, which forces
the command to be executed before circuit parsing. Basically <command> may be any
command listed in Chapt. 17.5, however only a few commands are indeed useful here.
Some examples are given below:
Examples:
pre_unset ngdebug
pre_set strict_errorhandling
pre_codemodel mymod . cm
pre_<command> is available only in the .control mode (see 16.4.3), not in interactive mode,
where the user may determine when a circuit is to be parsed, using the source command
(17.5.86) .
17.5.58 Pre_OSDI: load a *.osdi compact device model shared

library
Compact device models, written in Verilog-A HDL and compiled with OpenVAF (see13.2)
are loaded dynamically at runtime. Several models may be loaded for a single simulation
run. Please add these commands at the beginning of the .control section.
Examples:
pre_osdi osdi_libs / bsimbulk107 . osdi osdi_libs / psp103 . osdi

17.5. COMMANDS 407
17.5.59 Print: Print values
General Form:
print [ col ] [ line ] expr ...
Prints the vector(s) described by the expression expr. If the col argument is present,
print the vectors named side by side. If line is given, the vectors are printed horizontally.
col is the default, unless all the vectors named have a length of one, in which case line is
the default. The options width (default 80) and height (default 24) are effective for this
command (see asciiplot 17.5.6). The ’more’ mode is the standard mode if printing to
the screen, that is after a number of lines given by height, and after a page break printing
stops with request for answering the prompt by <return> (print next page), ’c’ (print
rest) or ’q’ (quit printing). If everything shall be printed with stopping after each page
(only useful in interactive mode, because need manual continuation), use the command
set moremode before printing or put it into .spiceinit 16.6 (or spinit 16.5). If the expression
is all, all of the vectors available are printed. Thus print col all > filename prints
everything into the file filename in SPICE2 format. The scale vector (time, frequency)
is always in the first column unless the variable noprintscale is true. You may use the
vectors alli, allv, ally with the print command, but then the scale vector will not
be printed.
Examples:
print all
set width =300
print v (1) > outfile . out
17.5.60 Psd: power spectral density of vectors
General Form:
psd ave vector1 [ vector2 ] ...
Calculate the single sided power spectral density of signals (vectors) resulting from a
transient analysis. Windowing is available as described in the fft command (17.5.33).
The FFT data are squared, summarized, weighted and printed as total noise power up to
Nyquist frequency, and as noise voltage or current.
ave is the number of points used for averaging and smoothing in a postprocess, useful
for noisy data. A new plot vector is created that holds the averaged results of the FFT,
weighted by the frequency bin. The result can be plotted and has the units V^2/Hz or
A^2/Hz, depending on the the input vector.
17.5.61 Quit: Leave Ngspice

General Form:
quit
quit [ exitcode ]
Quit ngspice. Ngspice will ask for an acknowledgment if parameters have not been saved.
If unset askquit is specified, ngspice will terminate immediately.
The optional parameter exitcode is an integer that sets the exit code for ngspice. This
is useful to return a success/fail value to the operating system.
17.5.62 Rehash: Reset internal hash tables

General Form:
rehash
Recalculate the internal hash tables used when looking up UNIX commands, and make all
UNIX commands in the user’s PATH available for command completion. This is useless
unless you have set unixcom first (see above).
17.5.63 Remcirc: Remove the current circuit

General Form:
remcirc
This command removes the current circuit from the list of circuits sourced into ngspice.
To select a specific circuit, use setcirc (17.5.75). To load another circuit, refer to source
(17.5.86). The new active circuit will be the circuit on top of the list of the remaining
circuits.
17.5.64 Remzerovec: Remove zero length vectors

General Form:
remzerovec
This command removes vectors of length zero from the current plot.
17.5.65 Reset: Reset an analysis

General Form:
reset
17.5. COMMANDS 409
Throw out any intermediate data in the circuit (e.g, after a breakpoint or after one or
more analyses have been done), and re-parse the input file. The circuit can then be re-run
from it’s initial state, overriding the effect of any set or alter commands. These two
need to be repeated after the reset command.
Reset may be required in simulation loops preceding any run (or tran ...) command.
Reset is required after an alterparam command (17.5.5) for making the parameter change
effective.
17.5.66 Reshape: Alter the dimensionality or dimensions of a

vector
General Form:
reshape vector vector ...

or
reshape vector vector ... [ dimension , dimension , ... ]
or
reshape vector vector ... [ dimension ][ dimension ] ...
This command changes the dimensions of a vector or a set of vectors. The final dimension
may be left off and it will be filled in automatically. If no dimensions are specified, then
the dimensions of the first vector are copied to the other vectors. An error message of the
form ’dimensions of x were inconsistent’ can be ignored.
Example:
* generate vector with all ( here 30) elements

let newvec = vector (30)
* reshape vector to format 3 x 10
reshape newvec [3][10]
* access elements of the reshaped vector
print newvec [0][9]
print newvec [1][5]
let newt = newvec [2][4]
17.5.67 Resume: Continue a simulation after a stop
General Form:
resume
Resume a simulation after a stop or interruption (control-C).

17.5.68 Rspice*: Remote ngspice submission

General Form:
rspice < input file >
Runs a ngspice remotely taking the input file as a ngspice input file, or the current circuit
if no argument is given. Ngspice waits for the job to complete, and passes output from
the remote job to the user’s standard output. When the job is finished the data is loaded
in as with aspice. If the variable rhost is set, ngnutmeg connects to this host instead of
the default remote ngspice server machine. This command uses the rsh command and
thereby requires authentication via a .rhosts file or other equivalent method. Note that
rsh refers to the ‘remote shell’ program, which may be remsh on your system; to override
the default name of rsh, set the variable remote_shell. If the variable rprogram is set,
then rspice uses this as the pathname to the program to run on the remote system.
Note: rspice will not acknowledge elements that have been changed via the alter or altermod
commands.
17.5.69 Run: Run analysis from the input file

General Form:
run [ rawfile ]
Run the simulation as specified in the input file. If there were any of the control lines
.ac, .op, .tran, or .dc, they are executed. The output is put in rawfile if it was given,
in addition to being available interactively.
17.5.70 Rusage: Resource usage

General Form:
rusage [ resource ...]
Print resource usage statistics. If any resources are given, just print the usage of that
resource. Most resources require that a circuit be loaded. Currently valid resources are
time Total Analysis Time
cputime The amount of time elapsed since the last rusage cputime call.
totalcputime Total elapsed time used so far.
decklineno Number of lines in deck
netloadtime Nelist loading time
netparsetime Netlist parsing time

17.5. COMMANDS 411
faults Number of page faults and context switches (BSD only).
space Data space used (output is depending on the operating system).
temp Operating temperature.
tnom Temperature at which device parameters were measured.
equations Number of circuit equations
totiter Total iterations
accept Accepted time-points
rejected Rejected time-points
loadtime Time spent loading the circuit matrix and RHS.
reordertime Matrix reordering time
lutime L-U decomposition time
solvetime Matrix solve time
trantime Transient analysis time
tranpoints Transient time-points
traniter Transient iterations
trancuriters Transient iterations for the last time point
tranlutime Transient L-U decomposition time
transolvetime Transient matrix solve time
everything All of the above.
all All of the above.
If rusage is given without any parameter, a sequence of outputs is printed using the
following rusage parameters: time, totalcputime, space.
17.5.71 Save: Save a set of outputs

General Form:
save [ all | outvec ...]
Save a set of outputs, discarding the rest (if keyword all is not given). May be used to
dramatically reduce memory (RAM) requirements if only a few useful node voltages or
branch currents are saved.
Node voltages may be saved by giving the nodename or v(nodename). Currents through
an independent voltage source are given by i(sourcename) or sourcename# branch. Internal
device data (31.1) are accepted as @dev[param]. The syntax is identical to the .save
command (15.6.1).
Note: In the .control .... .endc section save must occur before the run or tran
command to become effective.
If a node has been mentioned in a save command, it appears in the working plot after
a run has completed, or in the rawfile written by the write (17.5.107) command. For
backward compatibility, if there are no save commands given, all outputs are saved. If
you want to trace (17.5.97) or plot (17.5.56) a node, you have to save it explicitly, except
for all given or no save command at all.
When the keyword all appears in the save command, all node voltages, voltage source
currents and inductor currents are saved in addition to any other vectors listed.
Save voltage and current:
save vd_node vs # branch v ( vs_node ) i ( vs2 )
Save allows storing and later access of internal device parameters. e.g. in a command
like
Save internal parameters:
save all @mn1 [ gm ]
saves all standard analysis output data plus gm of transistor mn1 to internal memory (see
also 31.1).
save may store data from nodes or devices residing inside of a subcircuit:
Save voltage on node 3 (top level), node 8 (from inside subcircuit x2) and current through
vmeas (from subcircuit x1):
save 3 x1 . x2 . x1 . x2 .8 v . x1 . x1 . x1 . vmeas # branch
Save internal parameters within subcircuit:
save @m . xmos3 . mn1 [ gm ]
Use commands listing expand (17.5.47, before the simulation) or display (17.5.25,
after simulation) to obtain a list of all nodes and currents available. Please see Chapt. 31
for an explanation of the syntax for internal parameters.
Entering several save lines in a single .control section will accumulate the nodes and
parameters to be saved. If you want to exclude a node, you have to get its number by
calling status (17.5.89) and then calling delete number (17.5.21).
Don’t save anything:
save none
Useful if shared ngspice library is used and data are immediately transferred to the caller
via the shared ngspice interface.
17.5. COMMANDS 413
17.5.72 Sens: Run a sensitivity analysis

General Form:
sens output_variable
sens output_variable ac ( DEC | OCT | LIN ) N Fstart Fstop
Perform a Sensitivity analysis. output_variable is either a node voltage (ex. v(1)

or v(A,out)) or a current through a voltage source (e.g. i(vtest)). The first form
calculates DC sensitivities, the second form AC sensitivities. The output values are in
dimensions of change in output per unit change of input (as opposed to percent change
in output or per percent change of input).
17.5.73 17.5.73Set: Set the value of a variable

General Form:
set [ word ]
set [ word = value ] ...
set [ word = ( list of values )] ...
Set the value of word to value, if it is present. You can set any word to be any value:
numeric, string or list. If no value is given then the value is the Boolean ‘true’. If you
enter a string, you have to enclose it in double quotes. Set saves the lower case version
of a word string but the setcs variant of the command preserves case. If a variable is set
to a list of values that are enclosed in parentheses (which must be separated from their
values by white space), the value of the variable is the list.
The value of word may be inserted into a command by writing $word,or $word[index]for
an individual list element. The index may itself be a substitution: $word[$index].
The variables used by ngspice are listed in section 17.7.
If a variable has the same name as a simulator option, setting it will also attempt to set
the option.
Set entered without any parameter will list all variables set, and their values, if applicable.
Be advised that set sets the lower case variant of word. An exceptions to this rule is the
variable sourcepath.
Set automatically tries to distinguish between values given as numbers, strings or lists.
If a string starts with a numerical value, like 2N5401_C and is not enclosed in double
quotes, it is interpreted as a real number 2n, i.e. 2e-9. The rest of the string is discarded.
A variable may be set to a value read from a file by I/O redirection.
Example:
set invar < infile . txt

echo $invar
echo $invar [2]
echo $invar [5]
With the input text file

infile.txt:
* testing set input from file

3
NeXt
4
5 and 7
you will get the output from echo
3 NeXt 4 5 and 7
NeXt
and
Lines starting with ’*’ are comment lines and will be ignored. Lines with multiple tokens
are stored as list vectors, lines with a single token as string.
Another option to set a variable from outside is the I/O redirection by backquotes or
backticks (see 17.10), if you run ngspice as a console application.
17.5.74 Setcs: Set the value of a variable, case preserved

General Form:
setcs [ word ]
setcs [ word = value ] ...
Set the value of word to value, if it is present. You can set any word to be any value,
numeric or string. If no value is given then the value is the Boolean ‘true’. If you enter a
string, you have to enclose it in double quotes. Setcs keeps the case of a word string.
The value of word may be inserted into a command by writing $word. If a variable is set
to a list of values that are enclosed in parentheses (which must be separated from their
values by white space), the value of the variable is the list.
The variables used by ngspice are listed in section 17.7.
Setcs entered without any parameter will list all variables set, and their values, if appli-
cable.
Setcs automatically tries to distinguish between values given as numbers, strings or lists.
If a string starts with a numerical value, like 2N5401_C and is not enclosed in double
quotes, it is interpreted as a real number 2n, i.e. 2e-9. The rest of the string is discarded.
17.5.75 Setcirc: Change the current circuit

General Form:
setcirc [ circuit number ]

17.5. COMMANDS 415
The current circuit is the one that is used for the simulation commands below. When
a circuit is loaded with the source command (see below, 17.5.86) it becomes the current
circuit.
Setcirc followed by ’return’ without any parameters lists all circuits loaded.
17.5.76 Setplot: Switch the current set of vectors

General Form:
setplot
setplot [ plotname ]
setplot previous
setplot next
setplot new
Set the current plot to the plot with the given name, or if no name is given, prompt the
user with a list of all plots generated so far. (Note that the plots are named as they
are loaded, with names like tran1 or op2. These names are shown by the setplot and
display commands and are used by diff, below.) If the ‘New’ item is selected, a new plot
is generated that has no vectors defined.
Note that here the word plot refers to a group of vectors that are the result of one ngspice
run. When more than one file is loaded in, or more than one plot is present in one file,
ngspice keeps them separate and only shows you the vectors in the current plot with the
display (17.5.25) command. setplot previous will show the previous plot in the plot list, if
available, setplot next the next plot. If not available, a warning is issued and the current
plot stays active. Setplot will also allow selecting the digital event nodes that have been
created during the simulation that made the analog plot.
17.5.77 Setscale: Set the scale vector for the current plot
General Form:
setscale [ vector1 ] [ vector2 ]
The scale vector provides the values for the x-axis in a 2D plot. If no argument is given,
the current scale vector is printed. With one argument, defines the default scale vector
for the current plot. With two arguments, sets the specific scale vector of vector1 to be
vector2. If vector2 is “none” the scale vector for vector1 reverts to the plot’s default.
17.5.78 Setseed: Set the seed value for the random number gen-
erator
General Form:
setseed [ number ]
When this command is given, the seed value for the random number generator is set to
number. Number has to be an integer greater than 0. The random numbers retrieved after
this command are a sequence of pseudo random numbers with a huge period. Setting
the seed value will provide a reproducible sequence of random numbers, i.e. the same
seed results in the same sequence. See also the options SEED and SEEDINFO in chapt.
15.1.1and chapt. 22 on statistical circuit analysis..
17.5.79 Settype: Set the type of a vector

General Form:
settype type vector ...

Change the type of the named vectors to type. Type names can be found in the following
table.
Type Unit Type Unit
notype - pole -
time s zero -
frequency Hz s-param -
voltage V temp-sweep Celsius
current A
√ res-sweep Ohms
voltage-density V/√Hz impedance Ohms
current-density A/ Hz admittance S
voltage^2-density V/Hz power W
current^2-density A/Hz phase Degree
temperature Celsius decibel dB
charge C capacitance F
17.5.80 Shell: Call the command interpreter

General Form:
shell [ command ]
Call the operating system’s command interpreter; execute the specified command or
call for interactive use. The status returned by the command is stored in variable
shellstatus.
17.5.81 Shift: Alter a list variable

General Form:
shift [ varname ] [ number ]
If varname is the name of a list variable, it is shifted to the left by number elements (i.e,
the number leftmost elements are removed). The default varname is argv, and the default
number is 1.
17.5. COMMANDS 417
17.5.82 Show: List device state
General Form:
show devices [ : parameters ] , ...
The show command prints out tables summarizing the operating condition of selected
devices. If devices is missing, a default set of devices are listed, if devices is a single
letter, devices of that type are listed. A device’s full name may be specified to list only
that device. Finally, devices may be selected by model by using the form #modelname.
Because the output format is tabular, long strings, including device names, may be trun-
cated. The command “set altshow” selects an alternative output format without trun-
cations.
If no parameters are specified, the values for a standard set of parameters are listed. If
the list of parameters contains a ‘+’, the default set of parameters is listed along with
any other specified parameters.
For both devices and parameters, the word all has the obvious meaning.
Note: there must be spaces around the ‘:’ that divides the device list from the parameter
list.
17.5.83 Showmod: List model parameter values
General Form:
showmod models [ : parameters ] , ...
The showmod command operates like the show command (above) but prints out model
parameter values. The applicable forms for models are a single letter specifying the device
type letter (e.g. m, or c), a device name (e.g. m.xbuf22.m4b), or #modelname (e.g. #p1).
Typical usage (e.g. for BSIM4 model):
showmod # cmosn # cmosp : lkvth0 vth0
Note: there must be spaces around the ‘:’ that divides the device list from the parameter
list.
Obtain the default model parameters (e.g. for a BJT device):
netlist for default bipolar transistor

Q1 cc bb ee defbip
. model defbip npn
. control
op
showmod q1
. endc
op is required to set the data (otherwise all reported values are 0). The combination of
the default parameters and the parameters given in the .model line (This is what the
simulator finally uses.) are obtainable by showmod only after a simulation command (e.g.
op).
17.5.84 Snload: Load the snapshot file
General Form:
snload circuit - file file
snload reads the snapshot file generated by snsave (17.5.85). circuit-file is the original
circuit input file. After reading, the simulation may be continued by resume (17.5.67).
An input script for loading circuit and intermediate data, resuming simulation and plot-
ting is shown below:
Typical usage:
* SCRIPT : ADDER - 4 BIT BINARY

* script to reload circuit and continue the simulation
* begin with editing the file location
* to be started with ’ ngspice adder_snload . script ’
. control
* cd to where all files are located
cd D :\ Spice_general \ ngspice \ examples \ snapshot
* load circuit and snpashot file
snload adder_mos_circ . cir adder500 . snap
* continue simulation
resume
* plot some node voltages
plot v (10) v (11) v (12)
. endc
Due to a bug we currently need the term ’script’ in the title line (first line) of the script.
17.5.85 Snsave: Save a snapshot file
General Form:
snsave file
If you run a transient simulation and interrupt it by e.g. a stop breakpoint (17.5.91),
you may resume simulation immediately (17.5.67) or store the intermediate status in a
snapshot file by snsave for resuming simulation later (using snload (17.5.84)), even with
a new instance of ngspice.
17.5. COMMANDS 419
Typical usage:
Example input file for snsave

* load a circuit ( including transistor models and . tran command )
* starts transient simulation until stop point
* store intermediate data to file
* begin with editing the file location
* to be run with ’ ngspice adder_mos . cir ’
. include adder_mos_circ . cir
. control
* cd to where all files are located
cd D :\ Spice_general \ ngspice \ examples \ snapshot
unset askquit
set noinit
* interrupt condition for the simulation
stop when time > 500 n
* simulate
run
* store snapshot to file
snsave adder500 . snap
quit
. endc
. END
adder_mos_circ.cir is a circuit input file, including the netlist, .model and .tran state-
ments.
Unfortunately snsave/snload will not work if you have XSPICE devices (or V/I sources
with polynomial statements) in your input deck.
17.5.86 Source: Read a ngspice input file
General Form:
source infile
For ngspice: read the ngspice input file infile, containing a circuit netlist. Ngspice control
commands may be included in the file, and must be enclosed between the lines .control
and .endc. These commands are executed immediately after the circuit is loaded, so a
control line of ac ... works the same as the corresponding .ac card. The first line in
any input file is considered a title line and not parsed but kept as the name of the circuit.
Thus, a ngspice command script in infile must begin with a blank line and then with a
.control line. Also, any line starting with the string ‘*#’ is considered as a control line
(.control and .endc is placed around this line automatically.). The exception to these
rules are the files spinit (16.5) and .spiceinit (16.6).
For ngutmeg: reads commands from the file infile. Lines beginning with the character
‘*’ are considered comments and are ignored.
The following search path is executed to find infile: current directory (OS dependent),
<prefix>/share/ngspice/scripts, env. variable NGSPICE_INPUT_DIR (if defined), see 16.7.
This sequence may be overridden by setting the internal sourcepath variable (see 17.7)
before calling source infile.
17.5.87 Sp: S-Parameter Analysis
General form:
sp dec nd fstart fstop < donoise >

sp oct no fstart fstop < donoise >
sp lin np fstart fstop < donoise >
Examples:
sp dec 10 1 10 K
sp dec 10 1 K 100 MEG 1
sp lin 100 1 100 HZ
For details please see chapter 15.3.8. the ports required are available as an option to the
independent voltage source VSRC (see4.1.11 ).
17.5.88 Spec: Create a frequency domain plot
General Form:
spec start_freq stop_freq step_freq vector [ vector ...]
Calculates a new complex vector containing the Fourier transform of the input vector
(typically the linearized result of a transient analysis). The default behavior is to use
a Hanning window, but this can be changed by setting the variables specwindow and
specwindoworder appropriately.
Typical usage:
ngspice 13 -> linearize

ngspice 14 -> set specwindow = " blackman "
ngspice 15 -> spec 10 1000000 1000 v ( out )
ngspice 16 -> plot mag ( v ( out ))
Possible values for specwindow are none, hanning, cosine, rectangular, hamming, triangle,
bartlet, blackman, gaussian and flattop. In the case of a Gaussian window specwindoworder
is a number specifying its order. For a list of window functions see 17.5.33.
17.5. COMMANDS 421
17.5.89 Status: Display breakpoint information
General Form:
status
Display all of the saved nodes and parameters, traces and breakpoints currently in effect.
17.5.90 Step: Run a fixed number of time-points
General Form:
step [ number ]
Iterate number times, or once, and then stop.
17.5.91 Stop: Set a breakpoint
General Form:
stop [ after n ] [ when value cond value ] ...
Set a breakpoint. The argument after n means stop after iteration number ‘n’, and
the argument when value cond value means stop when the first value is in the given
relation with the second value, the possible relations being
Symbol Alias Meaning
= eq equal to
<> ne not equal
> gt greater than
< lt less than
>= ge greater than or equal to
<= le less than or equal to
Symbol or alias may be used alternatively. All stop commands have to be given in the
control flow before the run command. The values above may be node names in the
running circuit, or real values. If more than one condition is given, e.g.
stop after 4 when v(1) > 4 when v(2) < 2,
the conjunction of the conditions is implied. If the condition is met, the simulation and
control flow are interrupted, and ngspice waits for user input.
In a transient simulation the ‘=’ or eq will only work with vector time in commands like
stop when time = 200n.
Internally, a breakpoint will be set at the time requested. Multiple breakpoints may
be set. If the first stop condition is met, the simulation is interrupted, the commands
following run or tran (e.g. alter or altermod) are executed, then the simulation may
continue at the first resume command. The next breakpoint requires another resume to
continue automatically. Otherwise the simulation stops and ngspice waits for user input.
If you try to stop at
stop when V(1) eq 1
(or similar) during a transient simulation, you probably will miss this point, because it is
not very likely that at any time step the vector v(1) will have the exact value of 1. Then
ngspice simply will not stop.
17.5.92 Strcmp: Compare two strings

General Form:
strcmp _flag $string1 " string2 "
The command compares two strings, either given by a variable (string1) or as a string
in quotes (‘string2’). _flag is set as an output variable to ’0’, if both strings are equal.
A value greater than zero indicates that the first character that does not match has a
greater value in str1 than in str2; and a value less than zero indicates the opposite (like
the C strcmp function).
17.5.93 Strslice: Extract a substring

General Form:
strslice result input offset length
This command sets variable ’result’ to be a portion of string ’input’ starting at the given
offset and with the requested length. Offset and length should be integers. If offset is
negative, it is counted from the end of the input string.
17.5.94 Strstr: Find a substring

General Form:
strstr result " $haystack " needle
The command searches string variable ’haystack’ for a copy of string ’needle’. If successful,
variable ’result’ is set to the offset of the first match. Otherwise, the result is -1. As a
special case, if ’needle’ is the empty string, the result is the length of $haystack.
17.5.95 Sysinfo: Print system information

General Form:
sysinfo
17.5. COMMANDS 423
The command prints system information useful for sending bug report to developers.
Information consists of
• Name of the operating system,
• CPU type,
• Number of physical processors,
• Number of logical processors,
• Total amount of DRAM available,
• DRAM currently available.
The example below shows the use of this command.
ngspice 1 -> sysinfo

OS : CYGWIN_NT -5.1 1.5.25(0.156/4/2) 2008 -06 -12 19:34
CPU : Intel ( R ) Pentium ( R ) 4 CPU 3.40 GHz
Logical processors : 2
Total DRAM available = 1535.480469 MB .
DRAM currently available = 984.683594 MB .
ngspice 2 ->
This command has been tested under Windows OS and Linux. It may not be available
in your operating system environment.
17.5.96 Tf: Run a Transfer Function analysis
General Form:
tf output_node input_source
The tf command performs a transfer function analysis, returning:
• the transfer function (output/input),
• output resistance,
• and input resistance
between the given output node and the given input source. The analysis assumes a
small-signal DC (slowly varying) input. The following example file
* Tf test circuit
vs 1 0 dc 5
r1 1 2 100
r2 2 3 50
r3 3 0 150
r4 2 0 200
. control
tf v (3 ,5) vs
print all
. endc
. end
will yield the following output:
transfer_function = 3.750000e-001
output_impedance_at_v(3,5) = 6.662500e+001
vs#input_impedance = 2.000000e+002
17.5.97 Trace: Trace nodes
General Form:
trace [ node ...]
For every step of an analysis, the value of the node is printed. Several traces may be
active at once. Tracing is not applicable for all analyses. To remove a trace, use the
delete (17.5.21) command.
17.5.98 Tran: Perform a transient analysis
General Form:
tran Tstep Tstop [ Tstart [ Tmax ] ] [ UIC ]
Perform a transient analysis. See Chapt. 15.3.10 of this manual for more details.
An interactive transient analysis may be interrupted by issuing a ctrl-c (control-C) com-

mand. The analysis then can be resumed by the resume command (17.5.67). Several
options may be set to control the simulation (15.1.4).
17.5. COMMANDS 425
17.5.99 Transpose: Swap the elements in a multi-dimensional

data set
General Form:
transpose vector vector ...
This command transposes a multidimensional vector. No analysis in ngspice produces

multidimensional vectors, although the DC transfer curve may be run with two varying
sources. You must use the reshape command to reform the one-dimensional vectors into
two dimensional vectors. In addition, the default scale is incorrect for plotting. You must
plot versus the vector corresponding to the second source, but you must also refer only
to the first segment of this second source vector. For example (circuit to produce the
transfer characteristic of a MOS transistor):
How to produce the transfer characteristic of a MOS transistor:
ngspice > dc vgg 0 5 1 vdd 0 5 1

ngspice > plot i ( vdd )
ngspice > reshape all [6 ,6]
ngspice > transpose i ( vdd ) v ( drain )
ngspice > plot i ( vdd ) vs v ( drain )[0]
17.5.100 Unalias: Retract an alias

General Form:
unalias [ word ...]
Removes any aliases present for the words.
17.5.101 Undefine: Retract a definition

General Form:
undefine [ function ...]

undefine *
Definitions for the named user-defined functions are deleted. If * is given, all user-defined
functions are deleted.
17.5.102 Unlet: Delete the specified vector(s)

General Form:
unlet [ vector ...]

Delete the specified vector(s). See also let (17.5.45).
17.5.103 Unset: Clear a variable
General Form:
unset [ word ...]

unset *
Clear the value of the specified variable(s) (word). If * is specified, all variables are
cleared.
17.5.104 Version: Print the version of ngspice
General Form:
version [ - s | -f | < version id >]
Print out the version of ngspice that is running, if invoked without argument or with -s
or -f. If the argument is a <version id> (any string different from -s or -f is considered
a <version id> ), the command checks to make sure that the arguments match the
current version of ngspice. (This is mainly used as a Command: line in rawfiles.)
Options description:
• No option: The output of the command is the message you can see when running
ngspice from the command line, no more no less.
• -s(hort): A shorter version of the message you see when calling ngspice from the
command line.
• -f (ull): You may want to use this option if you want to know what extensions are
included into the simulator and what compilation switches are active. A list of
compilation options and included extensions is appended to the normal (not short)
message. May be useful when sending bug reports.
The following example shows what the command returns in some situations:
17.5. COMMANDS 427
Use of the version command:
ngspice 10 -> version

******
** ngspice -39 : Circuit level simulation program
** The U . C . Berkeley CAD Group
** Copyright 1985 -1994 , Regents of the University
of California .
** Copyright 2001 -2023 , The ngspice team .
** Please get your ngspice manual from
https :// ngspice . sourceforge . io / docs . html
** Please file your bug - reports at
https :// ngspice . sourceforge . io / bugrep . html
** Creation Date : Mar 7 2023 17:25:48
******
ngspice 2 ->
ngspice 11 -> version 14
Note : rawfile is version 14 ( current version is 39)
ngspice 12 -> version 39
ngspice 13 ->
Note for developers: The option listing returned when version is called with
the -f flag is built at compile time using #ifdef blocks. When new compile
switches are added, if you want them to appear on the list, you have to modify
the code in misccoms.c.
17.5.105 Where: Identify troublesome node or device
General Form:
where
When performing a transient or operating point analysis, the name of the last node or
device to cause non-convergence is saved. The where command prints out this information
so that you can examine the circuit and either correct the problem or generate a bug
report. You may do this either in the middle of a run or after the simulator has given
up on the analysis. For transient simulation, the iplot command can be used to monitor
the progress of the analysis. When the analysis slows down severely or hangs, interrupt
the simulator (with control-C) and issue the where command. Note that only one node
or device is printed; there may be problems with more than one node.
17.5.106 Wrdata: Write data to a file (simple table)

General Form:
< set wr_singlescale >

< set wr_vecnames >
< option numdgt =7 >
...
wrdata [ file ] [ vecs ]
Writes out the vectors to file.

This is a very simple printout of data in array form. Variables are written in pairs: scale
vector, value vector. If variable is complex, a triple is printed (scale, real, imag). If more
than one vector is given, the third column again is the default scale, the fourth the data
of the second vector. The default format is ASCII. All vectors have to stem from the
same plot, otherwise a segfault may occur. Setting wr_singlescale as variable, the scale
vector will be printed only once, if scale vectors are of the same length (you have to take
care of that yourself). Setting wr_vecnames as variable, scale and data vector names are
printed on the first row. The number of significant digits is set with option numdgt.
output example from two vectors:
0.000000 e +00 -1.845890 e -06 0.000000 e +00

0.000000 e +00
7.629471 e +06 4.243518 e -06 7.629471 e +06 -4.930171 e -06
1.525894 e +07 -5.794628 e -06 1.525894 e +07
4.769020 e -06
2.288841 e +07 5.086875 e -06 2.288841 e +07 -3.670687 e -06
3.051788 e +07 -3.683623 e -06 3.051788 e +07
1.754215 e -06
3.814735 e +07 1.330798 e -06 3.814735 e +07 -1.091843 e -06
4.577682 e +07 -3.804620 e -07 4.577682 e +07
2.274678 e -06
5.340630 e +07 9.047444 e -07 5.340630 e +07 -3.815083 e -06
6.103577 e +07 -2.792511 e -06 6.103577 e +07
4.766727 e -06
6.866524 e +07 5.657498 e -06 6.866524 e +07 -2.397679 e -06
....
If variable appendwrite is set, the data may be added to an existing file.
17.5.107 Write: Write data to a file (Spice3f5 format)

General Form:
write [ file ] [ exprs ]
Writes out the expressions to file.

17.5. COMMANDS 429
First vectors are grouped together by plots, and written out as such (i.e. if the expression
list contained three vectors from one plot and two from another, then two plots are
written, one with three vectors and one with two). Additionally, if the scale for a vector
isn’t present it is automatically written out as well.
The default format is a compact binary, but this can be changed to ASCII with the set
filetype=ascii command. The default file name is either rawspice.raw or the argument
of the optional -r flag on the command line, and the default expression list is all.
If variable appendwrite is set, the data may be added to an existing file.
17.5.108 Wrnodev: Write node voltage values to a file (.ic=xx

format)
General Form:
wrnodev [ file ]
Writes out the values of all voltage nodes to file. The format is .ic=xx. Thus the file
may be included into another simulation of the same circuit and deliver initial conditions
for all voltage nodes. For example you may start a transient simulation, stop it and use
the current data to start an ac simulation.
output example:
* Intermediate Transient Solution

* Circuit : KiCad schematic
* Recorded at simulation time : 3.9
. ic v ( net - _d1a1 - pad2_ ) = -31.2016
. ic v ( -32) = -32
. ic v ( out ) = -0.267414
. ic v ( net - _q5 - pad2_ ) = -26.5798
. ic v ( q5tj ) = 132.521
. ic v ( q5tc ) = 105.091
...
The following control section snippet serves to save node voltage data at time 3.9 s and
after the end of the transient simulation.
usage example (write data):
stop when time =3.9

tran 20 u 6
wrnodev $inputdir / F5ic1 . txt
resume
wrnodev $inputdir / F5ic2 . txt
...
The data may be reused by an .include command: The simulation now starts with the
initial condition given in the file.
usage example (read data):
. include F5ic1 . txt

...
17.5.109 Wrs2p: Write scattering parameters to file (Touch-

stone® format)
General Form:
wrs2p [ file ]
Writes out the s-parameters of a two-port to file.

In the active plot the following is required: vectors frequency, S11 S12 S21 S22, all
having the same length and complex values (as a result of an ac analysis), and vector
Rbase. For details how to generate these data see Chapt. 17.9.
The file format is Touchstone® Version 1, ASCII, frequency in Hz, real and imaginary
parts of Snn versus frequency.
The default file name is s-param.s2p.
output example:
!2 - port S - parameter file

! Title : test for scattering parameters
! Generated by ngspice at Sat Oct 16 13:51:18 2010
# Hz S RI R 50
! freq ReS11 ImS11 ReS21
2.500000 e +06 -1.358762 e -03 -1.726349 e -02 9.966563 e -01
5.000000 e +06 -5.439573 e -03 -3.397117 e -02 9.867253 e -01 ...
17.6 Control Structures

The following loops and examples are valid if put into a .control ... .endc section.
17.6.1 While - End

General Form:
while condition
statement
...
end
While condition, an arbitrary algebraic expression, is true, execute the statements.

17.6. CONTROL STRUCTURES 431
Example:
let loopindex = 0
while loopindex < 5
echo index is $ & loopindex
let loopindex = loopindex + 1
end
Comment: let creates a vector. Convert vector loopindex to number (as required by
echo) by $&loopindex. The condition statement compares vectors.
17.6.2 Repeat - End
General Form:
repeat [ number ]
statement
...
end
Execute the statements number times, or forever if no argument is given.
Example: Comment:
set loops = 7
repeat $loops
echo How many loops ? $loops
end
set creates a variable. repeat requires a number as parameter, either a plain number
or converted from vector by $&loopvec or converted from variable by $loops.
17.6.3 Dowhile - End
General Form:
dowhile condition
statement
...
end
The same as while, except that the condition is tested after the statements are executed.
Example:
let loopindex = 0
dowhile loopindex <> 5
echo index is $ & loopindex
let loopindex = loopindex + 1
end
17.6.4 Foreach - End

General Form:
foreach var value ...

statement
...
end
The statements are executed once for each of the values, each time with the variable var
set to the current value. (var can be accessed by the $var notation - see below).
Examples:
foreach val -40 -20 0 20 40

echo var is $val
end
echo
set myvariable = ( -4 -2 0 2 4 )
foreach var $myvariable
echo var is $var
end
Comment: The values themselves may be set by a variable like myvariable.
17.6.5 If - Then - Else

General Form:
if condition
statement
...
else
statement
...
end
If the condition is non-zero then the first set of statements are executed, otherwise the
second set. The else and the second set of statements may be omitted.
17.6. CONTROL STRUCTURES 433
Example:
foreach val -40 -20 0 20 40

if $val < 0
echo variable $val is less than 0
else
echo variable $val is greater than or equal to 0
end
end
echo
let vec = 1
if vec = 1 ; if $ & vec = 1 is possible as well
echo vec is $ & vec
end
Comment: The condition may be evaluated by numbers or vectors. Variables have to be

parsed to numbers like $val.
17.6.6 Label
General Form:
label word
If a statement of the form goto word is encountered, control is transferred to this point,
otherwise this is a no-op.
17.6.7 Goto
General Form:
goto word
If a statement of the form label word is present in the block or an enclosing block,
control is transferred there. Note that if the label is at the top level, it must be before
the goto statement (i.e, a forward goto may occur only within a block). A block to just
include goto on the top level may look like the following example.
Example noop block to include forward goto on top level:
if (1)
...
goto gohere
...
label gohere
end
17.6.8 Continue
General Form:
continue [ n ]
If there is a while, dowhile, or foreach block n levels of loops above the enclosing this
statement, control passes to the test controlling that loop, or in the case of foreach, the
next value for that loop is taken. If n is not specified, it is assumed to be 1 and acts
on the loop immediately enclosing the continue command. If the value of 0 is given, it
treated as a no-op.
17.6.9 Break
General Form:
break [ n ]
If there is a while, dowhile, or foreach block n levels of loops above the enclosing this
statement, control passes out of the block. If n is not specified, it is assumed to be 1 and
acts on the loop immediately enclosing the break command. If the value of 0 is given, it
treated as a no-op.
Of course, control structures may be nested. When a block is entered and the input is the
terminal, the prompt becomes a number of >’s corresponding to the number of blocks the
user has entered. The current control structures may be examined with the debugging
command cdump (see 17.5.13).
17.7 Internally predefined variables

The operation of both ngutmeg and ngspice may be affected by setting variables with the
set command (17.5.73). In addition to the variables mentioned below, the set command
also affects the behavior of the simulator via the options previously described under the
section on .OPTIONS (15.1). You also may define new variables or alter existing variables
inside .control ... .endc for later use in a user-defined script (see Chapt. 17.8).
The following list is in alphabetical order. All of these variables are acknowledged by
ngspice. Frontend variables (e.g. on circuits and simulation) are not defined in ngnutmeg.
The predefined variables that may be set or altered by the set command are
addcontrol Set by ngspice if run with the -a command line parameter. When set,
additional lines are added to netlists to ensure that a simulation is run.
altshow When set, an alternate, non-tabular output format is used by the show and
showmod commands.
appendwrite Append to the file when a write command is issued, if one already exists.
17.7. INTERNALLY PREDEFINED VARIABLES 435
askquit Check to make sure that there are circuits suspended or plots unsaved. ngspice
warns the user when he tries to quit if this is the case.brief If set to FALSE, the
netlist will be printed.
auto_bridge When set to zero, automatic insertion of bridging devices (12.6) is disabled.
auto_bridge_xxxx Variables of this general format are used to control insertion of au-
tomatic bridging devices. See section 12.6.
batchmode Set by ngspice if run with the -b command line parameter. May be used in
input files to suppress plotting if ngspice runs in batch mode.
brief Suppresses automatic display of the processed netlist. It is set by default.
colorN These variables determine the colors used during plotting. Color values may be
entered as RGB values from 0 to 255 (hex or decimal) or stating a color name.
The identification number N may be an integer between 0 and 22. Color0 is the
background, color1 is the grid and text color, and color ids from 2 through 22
are used for graphs (vectors) plotted. Hex color coding is done according to set
colorN=rgb:r/g/b, where r, g, and b may range from 00 to FF each. For example
green is selected by set color3=rgb:00/FF/00. Decimal coding is available as set
colorN=rgbd:rd/gd/bd, where rd, gd, and bd may range from 0 to 255. If X11 is
being run (Linux, macOS, Cygwin), the value of the color variables may be any of
the standard X-Server color names, which may be found in file /usr/lib/rgb.txt.
ngspice GUI for Windows supports color names according to the Naming Common
Colors project. Details with more than 140 color names are to be found in file
wincolornames.h. An example is set color3=blue. If no color id is set, then a
predefined set of colors is applied automatically.
controlswait (only available with shared ngspice, chapt. 19.4.1.4) If the simulation is
started with the background thread (command bg_run), the .control section com-
mands are executed immediately after bg_run has been given, i.e. typically before
the simulation has finished. This often is not very useful because you want to evalu-
ate the simulation results. If controlswait is set in .spiceinit or spice.rc, the command
execution is delayed until the background thread has returned (aka the simulation
has finished). If set controlswait is given inside of the .control section, only the
commands following this statement are delayed.
cpdebug Print control debugging information.
csnumprec Allows setting the precision of values derived from vectors and variables (by
$var, $&vec) as arguments to functions listet in chapter 17.5. Default is 6, as has
been standard up to now. If functions are using directly a vector as input (without
the tranfer to number by $&), full double precision is used.
curplot (read only) Returns <type><no.> of the current plot. Type is one of tran, ac,
op, sp, dc, unknown, no. is a number, sequentially set internally. This information
is used to uniquely identify each plot.
curplotdate Sets the date of the current plot.
curplotname Sets the name of the current plot.

curplottitle Sets the title (a short description) of the current plot.
debug If set then a lot of debugging information is printed.
device The name (/dev/tty??) of the graphics device. If this variable isn’t set then the
user’s terminal is used. To do plotting on another monitor you probably have to set
both the device and term variables. (If device is set to the name of a file, nutmeg
dumps the graphics control codes into this file – this is useful for saving plots.)
diff_abstol The relative tolerance used by the diff command (default is 1e-12).
diff_reltol The relative tolerance used by the diff command (default is 0.001).
diff_vntol The absolute tolerance for voltage type vectors used by the diff command
(default is 1e-6).
digital_delay_type Controls the behaviour of XSPICE digital elements that support

the inertial_delay parameter.
echo Print out each command before it is executed.
editor The editor to use for the edit command.
enable_noisy_r A user definable variable (for .spiceinit) to enable noise calculation for
all behavioral resistors. May locally be switched off by instance parameter noisy=0.
If enable_noisy_r is not set, noise simulation may locally be enabled by instance
parameter noisy=1.
filetype This can be either ascii or binary, and determines the format of the raw
file (compact binary or text editor readable ascii). The default is binary. CIDER
output (30.14) may be binary or ascii as well.
fourgridsize How many points to use for interpolating into when doing Fourier analysis.
gridsize If this variable is set to an integer, this number is used as the number of equally
spaced points to use for the Y axis when plotting. Otherwise the current scale is
used (which may not have equally spaced points). If the current scale isn’t strictly
monotonic, then this option has no effect.
gridstyle Sets the grid during plotting with the plot command. Will be overridden by
direct entry of gridstyle in the plot command. A linear grid is standard for both
x and y axis. Allowed values are lingrid loglog xlog ylog smith smithgrid
polar nogrid.
hcopydev If this is set, when the hardcopy command is run the resulting file is automat-
ically printed on the printer named hcopydev with the command lpr -Phcopydev
-g file.
hcopyfont This variable specifies the font name for hardcopy output plots. The value is
device dependent.
hcopyfontsize This is a scaling factor for the font used in hardcopy plots.
hcopydevtype This variable specifies the type of the printer output to use in the hardcopy
command. If hcopydevtype is not set, Postscript format is assumed. plot (5)
is recognized as an alternative output format. When used in conjunction with
hcopydev, hcopydevtype should specify a format supported by the printer.
hcopyscale This is a scaling factor for the font used in hardcopy plots (between 0 and
10).
hcopywidth Sets width of the hardcopy plot.
hcopyheight Sets height of the hardcopy plot.
hcopypscolor Sets the color of the hardcopy output. If not set, black & white plotting
is assumed with different linestyles for each output vector. A valid color integer
value yields a colored plot background (0: black 1: white, others see below). and
colored solid lines. This is valid for Postscript only.
hcopypstxcolor This variable sets the color of the text in the Postscript hardcopy out-
put. If not set, black on white background is assumed, else it will be white on black
background. Valid colors are 0: black 1: white 2: red 3: blue 4: orange 5: green 6:
pink 7: brown 8: khaki 9: plum 10: orchid 11: violet 12: maroon 13: turquoise 14:
sienna 15: coral 16: cyan 17: magenta 18: gray (for smith grid) 19: gray (for smith
grid) 20: gray (for normal grid).
height The length of the page for asciiplot and print col.
history The number of events to save in the history list.
histsubst Set to enable history substitution in the command interpreter (17.5.41).
inputdir The directory path of the last input file. It may be used to direct outputs
into a directory relative to the input (even the into the same directory) by e.g. the
command write $inputdir/outfile.raw vec1 vec2.
interactive If interactive is set, numparam error handling may be done manually

with user input from the console. If not, ngspice will exit upon a numparam error.
keep#branch If keep#branch is set, the rawfile output for branch currents is 1 v1#branch
current for example, not 1 i(v1) current. This retains compatibility with soft-
ware like ICCAP.
lprplot5 This is a printf(3s) style format string used to specify the command to use
for sending plot(5)-style plots to a printer or plotter. The first parameter supplied
is the printer name, the second parameter is a file name containing the plot. Both
parameters are strings.
lprps This is a printf(3s) style format string used to specify the command to use for
sending Postscript plots to a printer or plotter. The first parameter supplied is
the printer name, the second parameter is the file name containing the plot. Both
parameters are strings.
measoutfile Add command set measoutfile=<path/filename> to .spiceinit or to a

.control section in the netlist to save .measure results from batch mode in a file.
modelcard The name of the model card (normally .MODEL)
moremode If moremode is set, whenever a large amount of data is being printed to the
screen (e.g, the print or asciiplot commands), the output is stopped every screen-
ful and continues when a carriage return is typed. If moremode is unset, then data
scrolls off the screen without pausing.
nfreqs The number of frequencies to compute in the Fourier command. (Defaults to 10.)
ngbehavior Sets the compatibility mode of ngspice. Default value is ’all’. To be set
in spinit (16.5) or .spiceinit (16.6). A value of ’all’ improves compatibility with
commercial simulators. Full compatibility is however not the intention of ngspice!
The values ’ps’, ’psa’, ’lt’, ’lta’, ’hs’ and ’spice3’ are available. See Chapt.
16.14.
ngdebug enables several debugging printouts (see 16.16).
nginfo Enables a status report during simulation (currently available only with MS Win-
dows GUI version).
ng_nomodcheck Suppresses any model parameter check, if set.
no_auto_gnd Setting this boolean variable by set no_auto_gnd in spinit or .spiceinit,

ngspice will refrain from replacing all nodes named gnd by node 0. In using this
setting, you will have to take care of proper zeroing appropriate ground nodes. If
you fail to do so, ngspice may crash, or deliver wrong results.
nobreak Don’t have asciiplot and print col break between pages.
noasciiplotvalue Don’t print the first vector plotted to the left when doing an asciiplot.
nobjthack BJTs can have either 3 or 4 nodes, which makes it difficult for the subcircuit
expansion routines to decide what to rename. If the fourth parameter has been
declared as a model name, then it is assumed that there are 3 nodes, otherwise it
is considered a node. To disable this, you can set the variablenobjthack and force
BJTs to have 4 nodes (for the purposes of subcircuit expansion, at least).
noclobber Don’t overwrite existing files when doing IO redirection.
noglob Don’t expand the global characters ‘*’, ‘?’, ‘[’, and ‘]’. This is the default.
nolegend Don’t plot the legend, when using the plot command..
nonomatch If noglob is unset and a global expression cannot be matched, use the global
characters literally instead of complaining.
noparse Don’t attempt to parse input files when they are read in (useful for debugging).
Of course, they cannot be run if they are not parsed.
noprintscale Don’t print the scale in the leftmost column when a print col command
is given.
nosavecurrents If set by ’set nosavecurrents’ and followed by ’reset’, the setting of

internal current vectors (.options savecurrents) is suppressed. This is useful in
ac simulation which does not support ’options savecurrents’ and you have a mix
of several simulations in a single script.
nosort Don’t let display sort the variable names.
nostepsizelimit The maximum step size during transient simulation is per default
limited to tstep given by .tran tstep tstop <tstart <tmax>‌> (15.3.10, 17.5.98).
It may be overridden and set to a value of (tstop - tstart)/50 by adding ’set
nostepsizelimit’ to .spiceinit. Both may be overriden by setting tmax on the
.tran line.
nosubckt Don’t expand subcircuits.
notrnoise Switch off the transient noise sources (Chapt. 4.1.7).
nounits Plotting of the units token for the x and y axes of a graph is suppressed. Units
may be added manually to the y and x labels for SI conformity.
numdgt The number of digits to use when printing tables of data (print col). The de-
fault precision is 6 digits. On the PC, approximately 16 decimal digits are available
using double precision, so p should not be more than 16. If the number is negative,
one fewer digit is printed to ensure constant widths in tables.
num_threads The number of of threads to be used if OpenMP (see Chapt. 16.10) is

selected. The default value is 2.
oscompiled is set during ngspice compilation and returns a number corresponding to

the operating environment used during compilation. 0 Other, 1 MINGW for MS
Windows, 2 Cygwin for MS Windows, 3 FreeBSD, 4 OpenBSD, 5 Solaris, 6 Linux,
7 macOS, 8 Visual Studio for MS Windows .
osdi_enabled is set by ngspice upon start-up when the OSDI interface (13.2) is compiled
in.
plainlet Command let (17.5.45) will executed without evaluating any expression in its
command line. This is useful if characters like ’/’ are part of the vector names, e.g.
as issued by KiCad. Setting plainlet may be used to rename a vector including
such math characters into a vector using only standard characters. Then standard
plot, print, and write commands may be applied to this renamed vector.
plainplot Command plot (17.5.56) will executed without evaluating any expression in
its command line. This is useful if characters like ’/’ are part of the vector names.
plainwrite Command write (17.5.107) will executed without evaluating any expression
in its command line. This is useful if characters like ’/’ are part of the vector names.
plotstyle This should be one of linplot, combplot, or pointplot. linplot, the

default, causes points to be plotted as parts of connected lines. combplot causes a
comb plot to be done. It plots vectors by drawing a vertical line from each point
to the X-axis, as opposed to joining the points. pointplot causes each point to be
plotted separately.
pointchars Set a string as a list of characters to be used as points in a point plot.

Standard is ‘ox*+#abcdefhgijklmnpqrstuvwyz’. Some characters are forbidden.
polydegree The degree of the polynomial that the plot command should fit to the data.
If polydegree is N, then ngspice fits a degree N polynomial to every set of N points
and draws 10 intermediate points in between each end point. If the points aren’t
monotonic, then ngspice tries to rotate the curve and reduce the degree until a fit
is achieved.
polysteps The number of points to interpolate between every pair of points available
when doing curve fitting. The default is 10.
program The name of the current program (argv[0]).
prompt The prompt, with the character ‘!’ replaced by the current event number. Sin-
gle quotes ’ ’ are required around the specified string unless you really want it
expanded.
rawfile The default name for created rawfiles.
remote_shell Overrides the name used for generating rspice runs (default is rsh).
renumber Renumber input lines when an input file has .includes.
rndseed Seed value for random number generator (used by sgauss, sunif, and rnd
functions). It is set by the option command ’option seed=val|random’.
rhost The machine to use for remote ngspice runs, instead of the default one (see the
description of the rspice command, below).
rprogram The name of the remote program to use in the rspice command.
sharedmode Variable is set when ngspice runs in its shared mode (from ngspice.dll or
ngspice_xx.so). May be used in universal input files to suppress plotting because a
graphics interface is lacking.
shellstatus Contains the status returned by the last “shell” command.
silent_fileio If set, the fopen and fread commands do not print error messages. Errors
are still reported by setting a variable.
sim_status will bet set to 0 when the simulation starts. If there is an error and the
simulation fails with ’xx simulation(s) aborted’, then sim_status is set to 1. The
variable can be used in scripted loops within a transient simulation to allow special
handling e.g. in case of non-convergence.
skywaterpdk will speed up the loading of large PDKs (tested with Skywater 130) by
avoiding some checks during library loading. When set, ngspice assumes that all
MOS devices have exactly 4 terminals. It does not look for unquoted parameters,
so assumes that all parameters are quoted correctly by { } or ’ ’.
sourcepath A list of the directories to search when a source command is given, or .include
or .lib is processed. The default is the current directory and the standard ngspice
library (/usr/local/lib/ngspice, or whatever LIBPATH is #defined to in the
ngspice source). The command
setcs sourcepath = ( e:/ D:/ . c:/Spice/Examples )
will overwrite the default. setcs is used to keep upper case letters. The search
sequence now is: current directory, e:/, D:/, current directory (again due to .),
c:/Spice/Examples. ’Current directory’ is depending on the OS. The command
setcs sourcepath = ( D:/mypath/input $sourcepath )
will add another path entry in front of the already existing list of paths. This feature
may be used with shared ngspice (19) to send a input path to code models which
require file input, like d_source. Only the first entry in the sourcepath list is sent
to the code models, however.
specwindow Windowing for commands spec (17.5.88) or fft (17.5.33). May be one of
the following: bartlet blackman cosine gaussian hamming hanning none rectangular
triangle.
specwindoworder Integer value 2 - 8 (default 2), used by commands spec or fft.
spicepath The program to use for the aspice command. The default is /cad/bin/spice.
2 2
sqrnoise If set,
√ noise data
√ outputs will be given as V /Hz or A /Hz, otherwise as the
usual V / Hz or A/ Hz.
strict_errorhandling If set by the user, an error detected during circuit parsing will
immediately lead ngspice to exit with exit code 1 (see 18.5). May be set in files
spinit (16.5) or .spiceinit (16.6) only.
subend The card to end subcircuits (normally .ends).
subinvoke The prefix to invoke subcircuits (normally X).
substart The card to begin subcircuits (normally .subckt).
term The mfb name of the current terminal.
ticchar A character applied as a tic mark (replaces the default ’x’).
ticmarks An integer value n, every n data points a tic (default: a small ’x’) will be set
on your graph.
ticlist A list of integers, e.g. ( 4 14 24 ), selects data points to set tics (small ’x’) on
your graph.
units If this is degrees, then all the trig functions will use degrees instead of radians.
unixcom If a command isn’t defined, try to execute it as a UNIX command. Setting this
option has the effect of giving a rehash command, below. This is useful for people
who want to use ngspice as a login shell.
wfont Set the font for the graphics plot in MS Windows. Typical fonts are courier,
times, arial and all others found on your machine. Default is courier.
wfont_size The size of the windows font. The default depends on system settings.
width The width of the page for asciiplot and print col (see also 15.6.7).
win_console is set when ngspice runs in a console under Windows.
wr_onespace Command wrdata: Print data with one space only in between, not by
collumns with fixed width.
wr_singlescale Command wrdata: The scale vector will be printed only once, if all
scale vectors are of the same length.
wr_vecnames Command wrdata: Scale and data vector names are printed on the first
row.
x11lineararcs Some X11 implementations have poor arc drawing. If you set this option,
ngspice will plot using an approximation to the curve using straight lines.
xbrushwidth Linewidth for graph (see xgridwidth for border and grid). Valid for MS
Windows GUI, X11, gnuplot and Postscript.
xgridwidth Linewidth for border and grid. Valid for MS Windows GUI, X11, gnuplot
and Postscript.
xfont Set the font for text (x and y labels, axis values) in the graphics plot in X11
(Linux, Cygwin, macOS etc.). The command fc-list | cut -f2 -d: | sort
-u | less -r lists the font names that are installed on the computer and are
suited for this variable. Use xfont with the setcs command to keep lower case and
upper case characters, e.g. in setcs xfont=’Noto Sans CJK JP’. The’Noto Sans’
font family is very well suited, covering Western and Asian fonts. Also valid for
gnuplot and Postscript.
xfont_size The size of the X11 font. The default depends on system settings.
xspice_enabled is set by ngspice upon start-up, when the XSPICE option (II) for using
code models is compiled in.
xtrtol Set trtol, e.g. to 7, to avoid the default speed reduction (accuracy increase)
for XSPICE (see 16.9). Be aware of potential precision degradation or convergence
issues using this option.
17.8 Scripts
Ngspice is started in batch or interactive mode with an input file on the command line.
Input files may also be sourced later with the source command or by using the script
name as a command. The ngspice input file contains the usual circuit netlist, model
cards, and may also contain a command script, enclosed in a .control .. .endc section.
Expressions, functions, constants, commands, variables, vectors, and control structures
may be assembled into such scripts.
Scripting allows automation of any ngspice task: simulations to perform, output data to
analyze, repeat simulations with modified parameters, assemble output plot vectors. The
17.8. SCRIPTS 443
ngspice scripting language is not very powerful, but well integrated into the simulation
flow. After reading the input file, any command sequences are immediately processed.
Variables or vectors set by previous commands may be referenced by the commands
following them. Data can be stored, plotted or grouped into new vectors for either plotting
or other means of data evaluation.
An input file may contain only a title and the .control .. .endc section: it is a pure
script. The need for a title (that may be blank) is an unfortunate result of the source
command being used for both circuit input and command file execution. Note that
this does allow the user to merely type the name of a circuit file as a command and
it is automatically run. The commands are executed immediately, without running any
analyses that may be specified in the circuit (to execute the analyses before the script
executes, include a run command in the script).
An alternative way to indicate a pure script is to put *ng_script in the first line, the
rest of the file is then treated as if it were inside a control section. As a special case, if a
script file begins with *ng_script_with_params and it was the first non-option argument
on the ngspice command line, then remaining command arguments are treated as script
arguments, not additional netlists.
Before a script is read, the variables argc and argv are set to the number of words follow-
ing the file-name on the command line, and a list of those words respectively. Individual
script arguments may be accessed as $1, $2 etc. After the file is finished, these variables
are unset. Note that if a command file calls another, it must save its argv and argc
since they are altered. Also, command files may not be re-entrant since there are no local
variables. Of course, the procedures may explicitly manipulate a stack ...; that way one
can write scripts analogous to shell scripts for ngspice.
17.8.1 Variables
Variables are defined and initialized with the set command (17.5.73). set output=10
defines the variable output and sets it to the number 10. Predefined variables, which are
used inside ngspice for specific purposes, are listed in Chapt. 17.7. Variables are accessible
globally. The values of variables may be used in commands by writing $varname where
the value of the variable is to appear, e.g. $output. If a variable is substituted that is
not defined internally, but is defined in the program environment, then the external value
is used. The special variable $$ refers to the process ID of the program. With $< a line
of input is read from the terminal.
If a variable is assigned with $&word, then word must be a vector (see below), and word’s
numeric value is taken to be the new value of the variable.
Variables may have a value that is a list of values. If foo is a valid variable, and is of
type list, then the expression $foo[low-high] expands to a range of elements. Either
the upper or lower index may be left out, and in addition to slicing also reversing of a
list is possible through $foo[len-0] (len is the length of the list, the first valid index is
always 1).
Furthermore, the notation $?foo evaluates to 1 if the variable foo is defined, 0 otherwise,
and $#foo evaluates to the number of elements in foo if it is a list, 1 if it is a number or
string, and 0 if it is a Boolean variable.
17.8.2 Vectors
Ngspice data is in the form of vectors: time, voltage, etc. Each vector has a type, and
vectors can be operated on and combined algebraically in ways consistent with their types.
Vectors are normally created as a result of a transient or dc simulation. They are also
established when a data file is read in (see the load command 17.5.48), or they are created
with the let command 17.5.45 inside a script. If a variable x is assigned something of the
form $&word, then word has to be a vector, and the numeric value of word is transferred
into the variable x.
17.8.3 Assessing vectors in subcircuits

Node voltages and branch currents from within a subcircuit may be read with a special
syntax. After circuit parsing, subcircuits are expanded, their names have become part of
each node name.
Input file example with nested subcircuits:
* test node names from subcircuits

Xsub1 a b sub1
. subckt sub1 n11 n12

Xsub2 n11 n12 sub2
R11 n11 int1 1 k
R12 n12 int1 1 k
. ends
. subckt sub2 n21 n22

R21 n21 int2 1 k
R22 n22 int2 1 k
. ends
. end
Subcircuit instance Xsub1 calls subcircuit sub1 which contains a subcircuit instance
Xsub2 calling sub2 which contains node int2.
Internal circuit resulting from subcircuit expansion:
r . xsub1 . xsub2 . r21 a xsub1 . xsub2 . int2 1 k

r . xsub1 . xsub2 . r22 b xsub1 . xsub2 . int2 1 k
r . xsub1 . r11 a xsub1 . int1 1 k
r . xsub1 . r12 b xsub1 . int1 1 k
After expansion the subcircuits have disappeared. We now have extended node (aka
vector) names like xsub1.int1 or xsub1.xsub2.int2. The top level subcircuit call name
is followed by node name, separated by a dot. Or the top level subcircuit call name
is followed second level subciruit call name, then followed by node name, each again
separated by a dot. You may now assess the node int2 values in a script by
17.8. SCRIPTS 445
print v(xsub1.xsub2.int2)
Also the device instances have got their subcircuit information added to their names in
a similar way. In addition the type identifier letter (e.g. R for resistor) has been put in
front. So the resistor instances now are called r.xsub1.r11 or r.xsub1.xsub2.r22.
17.8.4 Commands
Commands have been described in Chapt. 17.5.
17.8.5 control structures
Control structures have been described in Chapt. 17.6. Some simple examples will be
given below.
Control structure examples:
Test sequences for ngspice control structures

* vectors are used ( except foreach )
* start in interactive mode
. control
* test sequence for while , dowhile

let loop = 0
echo
echo enter loop with " $ & loop "
dowhile loop < 3
echo within dowhile loop " $ & loop "
let loop = loop + 1
end
echo after dowhile loop " $ & loop "
echo
let loop = 0
while loop < 3
echo within while loop " $ & loop "
let loop = loop + 1
end
echo after while loop " $ & loop "
let loop = 3
echo
echo enter loop with " $ & loop "
dowhile loop < 3
echo within dowhile loop " $ & loop "
$ output expected
let loop = loop + 1
end
echo after dowhile loop " $ & loop "
echo
let loop = 3
while loop < 3
echo within while loop " $ & loop "
$ no output expected
let loop = loop + 1
end
echo after while loop " $ & loop "
17.8. SCRIPTS 447
Control structure examples (continued):
* test for while , repeat , if , break

let loop = 0
while loop < 4
let index = 0
repeat
let index = index + 1
if index > 4
break
end
end
echo index " $ & index " loop " $ & loop "
let loop = loop + 1
end
* test sequence for foreach

echo
foreach outvar 0 0.5 1 1.5
echo parameters : $outvar $ foreach parameters are variables ,
$ not vectors !
end
* test for if ... else ... end

echo
let loop = 0
let index = 1
dowhile loop < 10
let index = index * 2
if index < 128
echo " $ & index " lt 128
else
echo " $ & index " ge 128
end
let loop = loop + 1
end
* simple test for label , goto

echo
let loop = 0
label starthere
echo start " $ & loop "
let loop = loop + 1
if loop < 3
goto starthere
end
echo end " $ & loop "
* test for label , nested goto

echo
let loop = 0
label starthere1
echo start nested " $ & loop "
let loop = loop + 1
if loop < 3
if loop < 3
goto starthere1
end
end
echo end " $ & loop "
* test for label , goto

echo
let index = 0
label starthere2
let loop = 0
echo We are at start with index " $ & index " and loop " $ & loop "
if index < 6
label inhere
let index = index + 1
if loop < 3
let loop = loop + 1
if index > 1
echo jump2
goto starthere2
end
end
echo jump
goto inhere
end
echo We are at end with index " $ & index " and loop " $ & loop "
17.8. SCRIPTS 449
* test goto in while loop

let loop = 0
if 1 $ outer loop to allow nested forward label ’ endlabel ’
while loop < 10
if loop > 5
echo jump
goto endlabel
end
let loop = loop + 1
end
echo before $ never reached
label endlabel
echo after " $ & loop "
end
* test for using variables , simple test for label , goto

set loop = 0
label starthe
echo start $loop
let loop = $loop + 1 $ expression needs vector at lhs
set loop = " $ & loop " $ convert vector contents to variable
if $loop < 3
goto starthe
end
echo end $loop
. endc
17.8.6 Example script ’spectrum’

A typical example script named spectrum is delivered with the ngspice distribution.
Even if it is made obsolete by the internal spec command (see 17.5.88), and especially by
the much faster fft command (see 17.5.33), it is a good example for getting acquainted
with the ngspice control (and post-processor) language.
As a suitable input for spectrum you may run a ring-oscillator, delivered with ngspice in
e.g. test/bsim3soi/ring51_41.cir. For an adequate resolution a simulation time of 1µs is
needed. A small control script starts ngspice by loading the R.O. simulation data and
executing spectrum.
Small script to start ngspice, read the simulation data and start spectrum:
* test for script ’ spectrum ’

. control
load ring51_41 . out
spectrum 10 MEG 2500 MEG 1 MEG v ( out25 ) v ( out50 )
. endc
17.8. SCRIPTS 451
17.8.7 Example script for random numbers

Generation and test of random numbers with Gaussian distribution
* agauss test in ngspice

* generate a sequence of gaussian distributed random numbers .
* test the distribution by sorting the numbers into
* a histogram ( buckets )
. control
let mc_runs = 200
let run = 0
let no_buck = 8 $ number of buckets
let bucket = unitvec ( no_buck )
$ each element contains 1
let delta = 3e -11 $ width of each bucket , depends
$ on avar and sig
let lolimit = 1e -09 - 3* delta
let hilimit = 1e -09 + 3* delta
dowhile run < mc_runs

let val = agauss (1 e -09 , 1e -10 , 3) $ get the random number
if ( val < lolimit )
let bucket [0] = bucket [0] + 1 $ ’ lowest ’ bucket
end
let part = 1
dowhile part < ( no_buck - 1)
if (( val < ( lolimit + part * delta )) &
+ ( val > ( lolimit + ( part -1)* delta )))
let bucket [ part ] = bucket [ part ] + 1
break
end
let part = part + 1
end
if ( val > hilimit )
* ’ highest ’ bucket
let bucket [ no_buck - 1] = bucket [ no_buck - 1] + 1
end
let run = run + 1
end
let part = 0
dowhile part < no_buck
let value = bucket [ part ] - 1
set value = " $ & value "
* print the bucket ’ s contents
echo $value
let part = part + 1
end
. endc
. end
17.8.8 Parameter sweep
While there is no direct command to sweep a device parameter during simulation, you
may use a script to emulate such behavior. The example input file contains of an resistive
divider with R1 and R2, where R1 is swept from a start to a stop value inside of the
control section, using the alter command (see 17.5.3).
Input file with parameter sweep
parameter sweep
* resistive divider , R1 swept from start_r to stop_r
VDD 1 0 DC 1
R1 1 2 1 k
R2 2 0 1 k
. control
let start_r = 1 k
let stop_r = 10 k
let delta_r = 1 k
let r_act = start_r
* loop
while r_act le stop_r
alter r1 r_act
op
print v (2)
let r_act = r_act + delta_r
end
. endc
. end
17.8.9 Output redirection
The console outputs delivered by commands like print (17.5.59), echo (17.5.26), or others
may be redirected into a text file. ’print vec > filename’ will generate a new file or
overwrite an existing file named ’filename’, ’echo text >‌> filename’ will append the
new data to the file ’filename’. Output redirection may be mixed with commands like
wrdata.
17.8. SCRIPTS 453
Input file with output redirection > and >‌>
** MOSFET Gain Stage ( AC ):

** Benchmarking Implementation of BSIM4 .0.0
** by Weidong Liu 5/16/2000.
** output redirection into file
M1 3 2 0 0 N1 L =1 u W =4 u
Rsource 1 2 100 k
Rload 3 vdd 25 k
Vdd vdd 0 1.8
Vin 1 0 1.2 ac 0.1
. control
ac dec 10 100 1000 Meg
plot v (2) v (3)
let flen = length ( frequency ) $ length of the vector
let loopcounter = 0
echo output test > text . txt $ start new file test . txt
* loop
while loopcounter lt flen
let vout2 = v (2)[ loopcounter ] $ generate a single point
$ complex vector
let vout2re = real ( vout2 ) $ generate a single point
$ real vector
let vout2im = imag ( vout2 ) $ generate a single point
$ imaginary vector
let vout3 = v (3)[ loopcounter ] $ generate a single
$ point complex vector
let vout3re = real ( vout3 ) $ generate a single point
$ real vector
let vout3im = imag ( vout3 ) $ generate a single point
$ imaginary vector
let freq = frequency [ loopcounter ] $ generate a single point vecto
echo bbb " $ & freq " " $ & vout2re " " $ & vout2im "
+ " $ & vout3re " " $ & vout3im " >> text . txt
$ append text and
$ data to file
$ ( continued from line above )
let loopcounter = loopcounter + 1
end
. endc
. MODEL N1 NMOS LEVEL =14 VERSION =4.8.1 TNOM =27

. end
17.9 Scattering parameters (S-parameters)
17.9.1 Intro
ngspice supports calculating, printing and plotting of the scattering parameters in two
fashions.
Intrinsic commands (.sp, see 15.3.8 and sp, see 17.5.87) will generate S-parameters versus
frequency from any suitable multi-port circuit at varying frequencies. Besides the s matrix
(with S_1_1, S_2_1, S_1_2, and S_2_2 for a two-port circuit), the Y and T matrix
vector values are calculated and saved as well.
A command line script, available from the ngspice distribution at examples/control_structs/s-
param.cir, creates S-parameters S_1_1, S_2_1, S_1_2, and S_2_2 of any two port
circuit.
The printed output using wrs2p (see Chapt. 17.5.109) is a Touchstone® version 1
format file. The file follows the format according to The Touchstone File Format Specifi-
cation, Version 2.0, available from here. An example is given as number 13 on page 15 of
that specification.
17.9.2 S-parameter measurement basics

S-parameters allow a two-port description not just by permuting I1 , U1 , I2 , U2 , but using
a superposition, leading to a power view of the port (We only look at two-ports here,
because multi-ports are not (yet?) implemented.).
You may start with the effective power, being negative or positive
P =u·i (17.3)
The value of P may be the difference of two real numbers, with K being another real
number.

ui = P = a2 − b2 = (a + b)(a − b) = (a + b)(KK −1 )(a − b) = {K(a + b)} K −1 (a − b)
(17.4)
Thus you get
K −1 u = a + b (17.5)
Ki = a − b (17.6)
and finally
u + K 2i
a= (17.7)
2K
17.9. SCATTERING PARAMETERS (S-PARAMETERS) 455
u − K 2i
b= (17.8)
2K
By introducing the reference resistance Z0 := K 2 > 0 we get finally the Heaviside trans-
formation
u + Z0 i u − Z0 i
a= √ , b= √ (17.9)
2 Z0 2 Z0
In case of our two-port we subject our variables to a Heaviside transformation
U1 + Z0 I1 U1 − Z0 I1
a1 = √ b1 = √ (17.10)
2 Z0 2 Z0
U2 + Z0 I2 U2 − Z0 I2
a2 = √ b2 = √ (17.11)
2 Z0 2 Z0
The s-matrix for a two-port then is

b1 s11 s12 a1
= (17.12)
b2 s21 s22 a2
Two obtain s11 we have to set a2 = 0. This is accomplished by loading the output port
exactly with the reference resistance Z0 , which sinks a current I2 = −U2 /Z0 from the
port.

b1
s11 = (17.13)
a1 a2 =0
U1 − Z0 I1
s11 = (17.14)
U1 + Z0 I1
Loading the input port from an ac source U0 via a resistor with resistance value Z0 , we
obtain the relation
U0 = Z0 I1 + U1 (17.15)
Entering this into 17.14, we get
2U1 − U0
s11 = (17.16)
U0
For s21 we obtain similarly

b2
s21 = (17.17)
a1 a2 =0
U2 − Z0 I2 2U2
s21 = = (17.18)
U1 + Z0 I1 U0
Equations 17.16 and 17.18 now tell us how to measure s11 and s21 : Measure U1 at the
input port, multiply by 2 using an E source, subtracting U0 , which for simplicity is set
to 1, and divide by U0 . At the same time measure U2 at the output port, multiply by 2
and divide by U0 . Biasing and measuring is done by subcircuit S_PARAM. To obtain s22
and s12 , you have to exchange the input and output ports of your two-port and do the
same measurement again. This is achieved by switching resistors from low (1mΩ) to high
(1T Ω) and thus switching the input and output ports.
17.9.3 Usage of .sp and sp

Please have a look at the intrinsic commands (.sp, see 15.3.8 and sp, see 17.5.87) for
generating S-parameters versus frequency.
17.9.4 Usage of the script

Copy and then edit s-param.cir. You will find this file in directory /examples/control_structs
of the ngspice distribution.
The reference resistance (often called characteristic impedance) for the measurements is
added as a parameter
.param Rbase=50
The bias voltages at the input and output ports of the circuit are set as parameters as
well:
.param Vbias_in=1 Vbias_out=2
Place your circuit at the appropriate place in the input file, e.g. replacing the existing
example circuits. The input port of your circuit has two nodes in, 0. The output port has
the two nodes out, 0. The bias voltages are connected to your circuit via the resistances
of value Rbase at the input and output respectively. This may be of importance for the
operating point calculations if your circuit draws a large dc current.
Now edit the ac commands (see 17.5.1) according to the circuit provided, e.g.
ac lin 100 2.5MEG 250MEG $ use for Tschebyschef
Be careful to keep both ac lines in the .control ... .endc section the same and only
change both in equal measure!
Select the plot commands (lin/log, or smithgrid) or the ’write to file’ commands
(write, wrdata, or wrs2p) according to your needs.
Run ngspice in interactive mode
ngspice s-param.cir
17.10 Using shell variables

You may use the shell command (17.5.80) to execute a command in the shell. Its return
value is printed at the ngspice prompt.
17.11. MISCELLANEOUS 457
Example:
shell echo $HOME

/ home / holger
The following is valid only if you are working with ngspice as a console app (Linux,
Cygwin). In interactive mode or from a .control section you may transfer the return of a
command from the shell into an ngspice variable by backquote or backtick substitution.
Any text between backquotes is replaced by the result of executing the text as a command
to the shell.
Example:
set myvar2 = ‘/ bin / bash -c " echo $HOME " ‘

echo $myvar2
/ home / holger
17.11 MISCELLANEOUS
C-shell type quoting with ’ and " may be used. Within single quotes, no further substi-
tution (like history substitution) is done, and within double quotes, the words are kept
together but further substitution is done.
History substitutions, similar to C-shell history substitutions, are also available - see
the C-shell manual page for all of the details. The characters ~, @{, and @} have the
same effects as they do in the C-Shell, i.e., home directory and alternative expansion. It
is possible to use the wildcard characters *, ?, [, and ] also, but only if you unset noglob
first. This makes them rather useless for typing algebraic expressions, so you should set
noglob again after you are done with wildcard expansion. Note that the pattern [âbc]
matches all characters except a, b, and c.
If X is being used, the cursor may be positioned at any point on the screen when the
window is up and characters typed at the keyboard are added to the window at that
point. The window may then be sent to a printer using the xpr(1) program.
17.12 Bugs
When defining aliases like alias pdb plot db( !:1 - !:2 ) you must be careful to quote
the argument list substitutions in this manner. If you quote the whole argument it might
not work properly.
In a user-defined function, the arguments cannot be part of a name that uses the plot.vec
syntax. For example: define check(v(1)) cos(tran1.v(1)) does not work.
Chapter 18
Ngspice User Interfaces
ngspice offers a variety of user interfaces. For an overview (several screen shots) please
have a look at the ngspice web page.
18.1 MS Windows Graphical User Interface
If compiled properly (e.g. using the --with-wingui flag for ./configure under MINGW),
ngspice for Windows offers a simple graphical user interface. In fact this interface does
not offer much more for data input than a console would offer, e.g. command line inputs,
command history and program text output. First of all it applies the Windows API for
data plotting. If you run the sample input file given below, you will get an output as
shown in Fig. 18.1.
Input file:
***** Single NMOS Transistor ( Id - Vd ) , BSIM3V3

*
*** circuit description ***
m1 2 1 3 0 n1 L =0.6 u W =10.0 u
vgs 1 0 3.5
vds 2 0 3.5
vss 3 0 0
*
. control
dc vds 0 3.5 0.05 vgs 0 3.5 0.5
plot vss # branch
. endc
*
* UCB parameters BSIM3v3 .2
. include ../ Exam_BSIM3 / Modelcards / modelcard . nmos
. include ../ Exam_BSIM3 / Modelcards / modelcard . pmos
*
. end
459
460 CHAPTER 18. NGSPICE USER INTERFACES
The GUI consists of an I/O port (lower window) and a graphics window, created by the
plot command.
Figure 18.1: MS Windows GUI
The output window displays messages issued by ngspice. You may scroll the window
to get more of the text. The input box (white box) may be activated by a mouse click
to accept any of the valid ngspice commands. The lower left output bar displays the
actual input file. ngspice progress during setup and simulation is shown in the progress
window (--ready--). The Stop button will interrupt the current simulation. Data may
be analysed, simulation resumed by the command resume. However, if ngspice is running
in a flow or loop from within a .control section, this flow or loop stays interrupted, only
the current simulation job will be finished by resume. The Quit button allows exiting
ngspice. If ngspice is actively simulating, due to using only a single thread, this interrupt
18.1. MS WINDOWS GRAPHICAL USER INTERFACE 461
has to wait until the window is accessible from within ngspice, e.g. during an update of
the progress window.
In the plot window there is the upper left button, which activates a drop down menu.
You may select to print the plot window shown (a simple printer interface), set up any
of the printers available on your computer, or issue a postscript file or a SVG file of the
active plot window.
A left-click in the plot window will print the coordinates of that point in the text window,
allowing data to be captured from the plot. Click, drag and release will show both start
and end points, as well as the slope of the line joining them. Click and drag with the
right button outlines a rectangle; on release a new window opens with a“zoomed” plot of
that rectangular area.
Instead of plotting with black background, you may set the background to any other color,
preferably to ‘white’ using the command shown below.
Input file modification for white background:
. control
run
* white background
set color0 = white
* black grid and text ( only needed with X11 , automatic with MS Win )
set color1 = black
* wider plot lines
set xbrushwidth =2
plot vss # branch
. endc
Figure 18.2: Plotting with white background
Many more set command options are available to customize the plot window. To name a
few (please see 17.7 for details): colorN, gridsize, gridstyle, plotstyle, pointchars, ticchar,
ticmarks, ticlist, wfont1 , wfont_size, xbrushwidth, xgridwidth, xfont2 , xfont_size.
As ngspice supports UNICODE text, fonts supporting other letterings than plain English
may be selected, e.g. Korean, Japanese, Chinese, Cyrillic, Arabic etc..
18.2 MS Windows Console

If the --with-wingui flag for ./configure under MINGW is omitted (see 32.2.4) or con-
sole_debug or console_release is selected in the MS Visual Studio configuration manager,
then ngspice will compile without any internal graphical input or output capability. This
may be useful if you apply ngspice in a pipe inside the MSYS window, or use it being
called from another program, and just generating output files from a given input. The
plot (17.5.56) command will not work and leads to an error message. In the MS Windows
release of ngspice its binary is distributed as ngspice_con.exe.
Only on the ngspice console binary in MS Windows input/output redirection is possible,
if ngspice is called (e.g. within a MSYS shell or from a shell script) like
$ ngspice_con < input.
This feature is used in the new CMC model test suite (to be described elsewhere), thus
requires a console binary.
1
Win GUI only
2
X11 only
18.3. LINUX 463
You still may generate graphics output plots or prints by gnuplot (17.5.38), if installed
properly (18.7), or by selecting a suitable printing option (18.6).
18.3 Linux
The standard user interface is a console for input and the X11 graphics system for output
with the interactive plot (17.5.56) command. If ngspice is compiled with the –without-x
flag for ./configure, a console application without graphical interface results. For more
sophisticated input user interfaces please have a look at Chapt. 18.8.
The X11 UI has buttons to save the plot in formats suitable for printing or inclusion in
a web page. The mouse actions in the plot window are the same as the Windows UI.
In addition, when the pointer is in the plot, keyboard input is inserted at the pointer
position so that the plot can be annotated. Annotations are included in saved files.
18.4 CygWin
The CygWin interface is similar to the Linux interface (18.3), i.e. console input and X11
graphics output. To avoid the warning of a missing graphical user interface, you have to
start the X11 window manager by issuing the commands
$ export DISPLAY=:0.0
$ xwin -multiwindow -clipboard &
inside of the CygWin window before starting ngspice.
18.5 Error handling

Error messages and error handling in ngspice have grown over the years, include a lot of
‘traditional’ behavior and thus are not very systematic and consistent.
Error messages may occur with the token ‘Error:’. Often the errors are non-recoverable
and will lead to exiting ngspice with error code 1. Sometimes, however, you will get an
error message, but ngspice will continue, and may either bail out later because the error
has propagated into the simulation, sometimes ngspice will continue, deliver wrong results
and exit with error code 0 (no error detected!).
In addition ngspice may issue warning messages like ‘Warning: ...’. These should cover
recoverable errors only.
So there is still work to be done to define a consistent error messaging, recovery or exiting.
A first step is the user definable variable strict_errorhandling. This variable may be set in
files spinit (16.5) or .spiceinit (16.6) to immediately stop ngspice, after an error is detected
during parsing the circuit. An error message is sent, the ngspice exit code is 1. This
behavior deviates from traditional SPICE error handling and thus is introduced as an
option only.
XSPICE error messages are explained in Chapt. 29.
18.6 Output-to-file options

ngspice offers a large variety of writing simulation results into a file. This chapter will
give a short summary of the available options.
18.6.1 Graphics files

18.6.1.1 SVG
How to prepare a plot

Various SVG settings are given by setting the following two variables:
svg_intopts Sets the plot parameters by numbers "svgwidth", "svgheight", "svgfont-

size", "svgfont-width", "svguse-color", "svgstroke-width", "svggrid-width", .
svg_stropts Sets the plot parameters by strings "svgbackground", "svgfont-family", "svg-

font" . Use command setcs to keep upper and lower case.
Usage
. control
set svg_intopts = ( 512 384 20 0 1 2 0 )
setcs svg_stropts = ( blue Arial Arial )
. endc
The following variables may override some of the above mentioned parameters or provide
more details.
device dependent.
hcopydevtype The variable specifies the type of the printer output to use in the hardcopy
command. It has to be set to set hcopydevtype=svg.
10).
colorN These variables determine the colors used during plotting. Color values may be
entered as RGB values from 0 to 255 (hex or decimal) or stating a color name.
The identification number N may be an integer between 0 and 20. Color0 is the
background, color1 is the grid and text color, and color ids from 2 through 20 are
used for graphs (vectors) plotted. The available color strings are (use the string
inside of the hyphens): "black", "white", "red", "blue", "#FFA500" (orange), "green",
"#FFC0C5" (pink), "#A52A2A" (brown), "#F0E68C" (khaki), "#DDA0DD" (plum),
18.6. OUTPUT-TO-FILE OPTIONS 465
"#DA70D6" (orchid), "#EE82EE" (violet), "#B03060" (maroon); "#40E0D0" (turqoise),

"#A0522D" (sienna), "#FF7F50" (coral), "cyan", "magenta", "#666" (gray for smith
grid), "#949494" (gray for smith grid), "#888" (gray for normal grid). Examples
are set color3=blue or set color3="#EE82EE". If no color id is set, then the
above mentioned, predefined set of colors is applied automatically.
and Postscript.
The plot-to-file command
hardcopy file vector <vectors> <title text> <xlabel text> <ylabel text>
Usage
. control
* simulation commands here
set hcopydevtype = svg
set svg_intopts = ( 512 384 20 0 1 2 0 )
setcs svg_stropts = ( yellow Arial Arial )
set color1 = blue
set color2 = green
hardcopy plot_1 . svg vss # branch title ’ Plot no . 4 ’
+ xlabel ’ Drain voltage ’ ylabel ’ Drain current ’
* plot to screen commands here
. endc
Plot-to-screen
The file contents may be plotted to the screen. For MS Windows you may use the Internet
Explorer or EDGE, linked to the .svg file extension. Under Cygwin or Linux you may
install the program feh for plotting with the following commands:
Plot to screen commands
* for MS Windows only

if $oscompiled = 1 | $oscompiled = 8
shell Start plot_1 . svg
else
* for CYGWIN , Linux
shell feh -- magick - timeout 1 plot_1 . svg &
end
18.6.1.2 PostScript
How to prepare a plot

Variables to modify the PostScript plot are listed below. Background and text colors
may be set. The colors of the graphs are then chosen automatically, starting with red.
Valid colors are 0: black 1: white 2: red 3: blue 4: orange 5: green 6: pink 7: brown 8:
khaki 9: plum 10: orchid 11: violet 12: maroon 13: turquoise 14: sienna 15: coral 16:
cyan 17: magenta 18: gray (for smith grid) 19: gray (for smith grid) 20: gray (for normal
grid).
hcopypscolor Sets the color of the hardcopy output byselecting a integer number. If
not set, black & white plotting is assumed with different linestyles for each output
vector. A valid color integer value yields a colored plot background (0: black 1:
white, others see above). and colored solid lines.
hcopypstxcolor This variable sets the color of the text in the Postscript hardcopy out-
put. If not set, black on white background is assumed, if the background is colored
or black, white text is printed.
device dependent.
hcopydevtype The variable specifies the type of the printer output to use in the hardcopy
command. It has to be set to set hcopydevtype=svg.
10).
and Postscript.
The corresponding input file for the examples given below is listed in Chapt. 21.1. Just
add the .control section to this file and run in interactive mode by
$ ngspice xspice_c1_print.cir
One way is to setup your printing like this will yield a black&white plot:
.control
set hcopydevtype=postscript
op
run
plot vcc coll emit
hardcopy temp.ps vcc coll emit
.endc
Then print the postscript file temp.ps to the screen. This may be done by a ngspice shell
command, depending on the operating system and the installed viewer tools (like gv or
others):

shell Start /B temp.ps
* for CYGWIN
else
shell gv temp.ps &
end
You can add color traces to it if you wish:
.control
* allow color and set background color if set to value >= 0
set hcopypscolor=1 ; white
set hcopypstxcolor = 3 ; blue
* The colors of the graphs are set automatically.
set xgridwidth=2
set xbrushwidth=3
run
hardcopy temp.ps vcc coll emit
.endc
Then print the postscript file temp.ps to a postscript printer.

You can also direct your output directly to a designated printer (not available in MS
Windows):
.control
*send output to the printer kec3112-clr
set hcopydev=kec3112-clr
hardcopy out.tmp vcc coll emit
.endc
18.6.1.3 PNG
There is no png driver integrated into ngspice. One may use the gnuplot interface (see
18.7) to create a png file.
Usage
. control
* simulation commands here
set gnuplot_terminal = png / quit
gnuplot plot_1 vss # branch vss2 # branch
+ title ’ Drain current versus drain voltage ’
+ xlabel ’ Drain voltage / V ’ ylabel ’ Drain current / uA ’
* plot to screen commands here
. endc
This command sequence will generate a png file plot_1.png in the current directory. You
will need to have gnuplot installed.
A few remarks are due: Generally you should use a text editor for the input files that
allows to set the character encoding to utf-8. you may give a true µA in the label text,
not only the uA. Otherwise a µ in the input file may lead ngspice to fail the utf-8 syntax
test. For sake of having not enough characters per line available in the final pdf manual
to fitting the gnuplot command, the line continuation is used in the above example with
a + character in the first column. Unfortunately this has a strange side effect in a real
ngspice input file, in that all letters become lower case in the continuation lines. So better
create a single (long) line containing the complete gnuplot command.
Plotting the png file to the screen can be achieved from within the .control section by

shell Start c :\" program files "\ irfanview \ i_view64 . exe plot_1 . png
else
* for CYGWIN , Linux
shell feh -- magick - timeout 1 plot_1 . png &
end
You will need to install a suitable viewer program (e.g. irfanview or feh).
18.6.1.4 VCD
Value Change Dump (VCD) (also known less commonly as "Variable Change Dump")
is an ASCII-based format for dumpfiles generated by event based logic simulation. The
eprvcd command is used by ngspice to print out the digital event nodes and real-valued
expressions versus time.
General Form:
eprvcd [ - t unit ][ - a ] node1 node2 .. noden [ > filename ]

Example usage:
eprvcd 1 2 3 4 5 6 7 8 s0 s1 s2 s3 c3 > adder_x . vcd
Values for analog nodes and expressions (as for plot ) may be included, but expressions
may not contain spaces. Option “-t” sets the VCD file’s time unit; values are rounded to
a negative power of 10. If not used the unit is chosen to fit the simulation’s maximum
timestep. Analog values are examined only when an XSPICE event values changes unless
option “-a” is used to dump them at each timestep.
The file addr_x.vcd may be displayed by the following .control section (gtkwave has to
be installed):
* plotting the vcd file ( e . g . with GTKWave )

* For Windows : returns control to ngspice
shell start gtkwave adder_x . vcd -- script nggtk . tcl
else
* for CYGWIN , Linux , others
shell gtkwave adder_x . vcd -- script nggtk . tcl &
end
with the tcl script to control gtkwave

nggtk.tcl
# tcl script for gtkwave : show vcd file data created by ngspice
set nfacs [ gtkwave :: getNumFacs ]
for { set i 0} { $i < $nfacs } { incr i } {
set facname [ gtkwave :: getFacName $i ]
set num_added [ gtkwave :: addSignalsFromList $facname ]
}
gtkwave ::/ Edit / UnHighlight_All
gtkwave ::/ Time / Zoom / Zoom_Full
18.6.2 Tabulated files

18.6.2.1 Rawfile
This is the traditional spice-compatible output file for simulation data. It will be generated
during simulation if ngspice is started in batch mode (16.4.1) like
ngspice -b -r mysim.raw -o mysim.log myinput.cir
where mysim.raw, following the -r flag, is the rawfile. It may be created as well from
inside a control section using the write command (17.5.107) like
write mysim.raw all
If not all result vetcors are to be stored in the rawfile, the .save command (15.6.1) will
limit the number of vectors to the ones liste after the command. One also may limit their
numbers if the vectors are explicitely stated in the write command
write mysim.raw v(node1) v2#branch
The rawfile consists of an ascii header, followed by the data, either in ascii or binary
format.
filetype This can be either ascii or binary, and determines the format of the raw file
(compact binary or text editor readable ascii). The default is binary.
All simulations (e.g. if .tran follow .ac) will be saved consecutively. If using the write
command, setting variable appendwrite will allow storing several sim outputs in a single
file.
18.6.2.2 Command wrdata
wrdata generates a file containing simulation data in a tabular fashion. For details please
see 17.5.106. The following variables and options are aknowledged:
using double precision, so p should not be more than 16. If the output number is
negative, one digit less is printed to ensure constant widths in tables.
wr_singlescale The scale vector will be printed only once, if all scale vectors are of the
same length.
wr_vecnames Scale and data vector names are printed on the first row.
18.6.2.3 Command wrs2p, Touchstone File Format Version 1
wrs2ps allows to write a file, containing S parameter data, in the Touchstone File Format
Version 1. For details please see 17.5.109 for the command and 17.9 for generating the
S-parameters.
18.6.2.4 Output redirection
Anything that is printable to the console by a control section command, may be redirected
into a file. See also 17.4.1.
Example usage:
* create a new file and write to it

echo new file > nfile . txt
* append line to existing file
echo second line >> nfile . txt
The following variable is recognized:
noclobber Don’t overwrite existing files when doing IO redirection.
18.6.2.5 Command echo
Echos all text, variables and vectors to the screen or the redirected output location (see
also 17.5.26).
Example usage:
* variable
setcs myvar = great
set empty =""
* vector
let lineno =1
* empty line
echo
* vectors and variables may be included
echo This is a $myvar output with $ & lineno line ( s ).
* no line feed , empty var to allow blank
echo -n This is still a $myvar output $empty
echo with $ & lineno line ( s ).
18.6.2.6 Command print
General Form:
print [ col ] [ line ] expr ...
Prints the vector(s) described by the expression expr. Please see 17.5.59 for details.
Expression expr. may be a list of vectors, but also a mathematical expression combining
vectors and constants according to 17.2.
Example:
print v (1) 3* v (2)

The following variables and options are aknowledged:
moremode If moremode is set, whenever a large amount of data is being printed to the
screen (e.g, the print or asciiplot commands), the output is stopped every screen-
ful and continues when a carriage return is typed. If moremode is unset, then data
scrolls off the screen without pausing.
noprintscale Don’t print the scale in the leftmost column when a print col command
is given.
using double precision, so p should not be more than 16. If the output number is
negative, one digit less is printed to ensure constant widths in tables.
18.6.2.7 Command eprint
Prints event driven nodes to the console (or a file when using output redirection). See
17.5.29 and 27.2.2 for an example.
18.7 Gnuplot
18.7.1 Using Gnuplot to produce 1D graphs of (electrical) sim-

ulation results
Plotting with Gnuplot is directly available from the ngspice .control section or interactive
command. Install Gnuplot (on Linux available from the distribution, on Windows avail-
able here). On Windows, expand the zip file to a directory of your choice, add the path
<any directory>/gnuplot/bin to the PATH variable, and off you go... The command to
invoke Gnuplot (17.5.38) is limited to x/y plots (no polar etc.).
General Form:
gnuplot file plotargs
plotargs is a list of vectors to be plotted. file may either be temp or tmp or a file name
(without file extension).
Plot window only:
gnuplot temp vss # branch vss2 # branch

ngspice generates temporary data and command files for Gnuplot, calls Gnuplot for ope-
nening the plot windows and then discards the temporary files.
18.7. GNUPLOT 473
Plot window plus command and data files:
gnuplot newplot vss # branch vss2 # branch

Gnuplot command file newplot.plt and data file newplot.data are generated to stay in the
current directory. The command file may be modified to alter the plot, and then called
by gnuplot newplot.plt to draw the modified plot.
The following variables are aknowledged by the gnuplot command:
gnuplot_terminal May be one of the following: png (write png file and plot to screen),
png/quit (write png file but no plot, see 18.6.1.3), eps (write PostScript file and
plot to screen), eps/quit (write PostScript file, but no plot), xterm (open gnu-
plot in an xterm window).
and Postscript.
plotstyle This should be one of linplot, combplot, or pointplot. linplot, the

default, causes points to be plotted as parts of connected lines. combplot causes a
comb plot to be done. It plots vectors by drawing a vertical line from each point
to the X-axis, as opposed to joining the points. pointplot causes each point to be
plotted separately.
nolegend Don’t plot the legend, when using the plot command.
18.7.2 Using gnuplot to produce 2D contour plots for Cider

The gnuplot command to generate 2D x/y contour plots from Cider models is:
General Form:
gnuplot file xycontour expr
The xycontour switch is ignored if the data is not from a 2D Cider model. expr is a single
plotarg expression which specifies the vector to be plotted. file has the same meaning
as in section 18.7.1 previously. The only variable which affects the gnuplot xycontour
option is gnuplot_terminal.
Before a plot can be created, the Cider solution file containing the data you are interested
in must be loaded with the LOAD (17.5.48) command. The example later in this section
demonstrates the steps to be followed.
The Cider OUTPUT command (see 30.14) explains how to get solution files for a Cider
model. It is important to include a ’rootfile’ parameter in the OUTPUT command which
specifies a subdirectory to hold the solution files themselves. Depending on the analysis
type, solution files have a prefix OP, DC, or TR. There can be many of these files created,
one per DC sweep value or per TR time step, so it is essential the ’rootfile’ subdirectory
is created prior to running ngspice to generate the solution files. In addition, device
instances D*, Q*, and M* of Cider models need to have the ’SAVE’ parameter set.
The 2D Cider models are NUMD level 2 (see 30.17), NBJT level 2 (see 30.18), and
NUMOS (see 30.19). 1D Cider models are level 1 NUMD and NBJT. The solution files
for 1D models can be plotted as the normal curves using PLOT (17.5.56) and GNUPLOT
(without xycontour, 17.5.38 and 18.7.1).
18.7.2.1 Example of a 2D jfet
File jfet1.cir, is run as follows from a bash console window (Linux, or MSYS2):
mkdir ./j1root
ngspice -b jfet1.cir
Filenames will need to be modified appropriately for Windows.
Notes relating to the jfet1.cir file:
1. The QJ1 instance line has the ’SAVE’ parameter, and the ’rootfile’ subdirectory is
specified on the output statement.
2. A Cider solution file is loaded after the simulation has run and before the gnuplot
commands:
load ./j1root/DC.12.qj1
The currently active vectors are listed after the load.
3. The sleep commands (or timeout /t on Windows) give the display time to draw the
contours. This is only necessary when executing a batch script; for an interactive
session they are not required.
4. The contours of a single vector phin are plotted by:

gnuplot gplot1 xycontour phin
5. The contours of the electric field magnitude are plotted by:

gnuplot gplot2 xycontour sqrt((ex * ex) + (ey * ey))
6. The gnuplot_terminal variable controls the output from gnuplot:

set gnuplot_terminal=png/quit
7. Following this, gnuplot commands will send the plot to a .png file.
18.7. GNUPLOT 475
JFET example netlist jfet1.cir:
****** jfet1 . cir ******

* Two - dimensional Junction Field - Effect Transistor ( JFET )
VDD 1 0 0.5 V
VGG 2 0 -1.0 v AC 1 V
VSS 3 0 0.0 V
QJ1 1 2 3 M_NJF AREA =1 SAVE
. MODEL M_NJF NBJT LEVEL =2
+ options jfet defw =10.0 um
+ output rootfile ="./ j1root /" psi n . conc p . conc phin phip equ . psi v
+ x . mesh w =0.2 h . e =0.001 r =1.8
+ x . mesh w =0.8 h . s =0.001 h . m =0.1 r =2.0
+ x . mesh w =0.8 h . e =0.001 h . m =0.1 r =2.0
+ x . mesh w =0.2 h . s =0.001 r =1.8
+ y . mesh w =0.2 h . e =0.01 r =1.8
+ y . mesh w =0.8 h . s =0.01 h . m =0.1 r =1.8
+
+ domain num =1 mat =1
+ material num =1 silicon
+
+ elec num =1 x . l =0.0 x . h =0.0 y . l =0.0 y . h =1.0
+
+ doping unif n . type conc =3.0 e15
+ doping unif p . type conc =2.0 e17 x . l =0.2 x . h =1.8 y . h =0.2
+
+ models bgn srh auger conctau concmob fieldmob ^ aval
. option bypass =1 temp =27
. control
dc vgg 0.0 -2.0001 -0.1
print i ( vss )
load ./ j1root / DC .12. qj1
shell ’ sleep 1 ’
gnuplot gplot2 xycontour sqrt (( ex * ex ) + ( ey * ey ))
set gnuplot_terminal = png / quit
gnuplot gplot4 xycontour sqrt (( ex * ex ) + ( ey * ey ))
quit
. endc
. end
The two contour graphs thus simulated are shown here:
Figure 18.3: Potential
Figure 18.4: Electrical field

18.8. INTEGRATION WITH CAD SOFTWARE AND ‘THIRD PARTY’ GUIS 477
18.8 Integration with CAD software and ‘third party’

GUIs
In this chapter you will find some links and comments on GUIs for ngspice offered from
other projects and on the integration of ngspice into a circuit development flow. The data
given rely mostly on information available from the web and thus is out of our control.
It also may be far from complete. For a list of actual links with more than 20 entries
please have a look at the ngspice web pages. Some open source tools are listed here.
The GUIs MSEspice and GNUSpiceGUI help you to navigate the commands to need to
perform your simulation. XCircuit and the GEDA tools gschem and gnetlist offer inte-
grating schematic capture and simulation. KiCad offers a complete design environment
for electronic circuits. Xschem focusses on IC design, supporting all the open source
PDKs.
18.8.1 KiCad
KiCad is a cross platform and open source electronics design automation suite. Its
schematic editor Eeschema fully integrates shared ngspice (see Chapt. 19) as the sim-
ulation tool. The target audience is PCB circuit designers, using discrete devices, ICs
and passives. On the ngspice web pages there is a tutorial available which presents an
introduction to using ngspice from within KiCad. Simulation examples are presented at
KiCad/ngspice simulation examples 1 and KiCad/ngspice simulation examples 2.
18.8.2 Xschem
Xschem is a schematic capture program, it allows to create a hierarchical representation
of circuits with a top down approach . By focusing on interconnections, hierarchy and
properties a complex system (IC) can be described in terms of simpler building blocks. A
VHDL, Verilog or ngspice netlist can be generated from the drawn schematic, allowing the
simulation of the circuit. The target audience is IC designers, especially the open source
PDKs from Skywater, Google and IHP are supported. An excellent video introduces the
design and simulation of a CMOS comparator.
18.8.3 Qucs-S
Qucs-S is a circuit simulation program based on the Qucs circuit simulator. The "S"
letter indicates SPICE. The purpose of the Qucs-S project is to use free SPICE circuit
simulation kernels with the GUI based on Qt toolkit. It merges the power of SPICE and
the simplicity of the Qucs GUI. Qucs-S is not a simulator by itself, but it requires to use
an external simulation backend, e.g. ngspice, with it. Quc-S is strong when simulating
RF devices and circuits.
18.8.4 GNU Spice GUI

A GUI, to be found at https://sourceforge.net/projects/gspiceui/. It aids in viewing,
modifying, and simulating SPICE CIRCUIT files.
18.8.5 XCircuit
CYGWIN and especially Linux users may find XCircuit valuable to establish a develop-
ment flow including schematic capture and circuit simulation.
18.8.6 GEDA
The gEDA project is developing a full GPL‘d suite and toolkit of Electronic Design
Automation tools for use with a Linux. Ngspice may be integrated into the development
flow. Two web sites offer tutorials using gschem and gnetlist with ngspice:
http://geda-project.org/wiki/geda:csygas
http://geda-project.org/wiki/geda:ngspice_and_gschem
18.8.7 MSEspice
A graphical front end to ngspice, using the Free Pascal cross platform RAD environment
MSEide+MSEgui.
18.8.8 GNU Octave

GNU Octave is a high-level language, primarily intended for numerical computations. An
interface to ngspice is available here.
Chapter 19
ngspice as shared library or dynamic

link library
ngspice may be compiled as a shared library. This allows adding ngspice to an application
that then gains control over the simulator. The shared module offers an interface that
exports functions controlling the simulator and callback functions for feedback.
So you may send an input ‘file’ with a netlist to ngspice, start the simulation in a separate
thread, read back simulation data at each time point, stop the simulator depending on
some condition, alter device or model parameters and then resume the simulation.
Shared ngspice does not have any user interface. The calling process is responsible for this.
It may offer a graphical user interface, add plotting capability or any other interactive
element. You may develop and optimize these user interface elements without a need to
alter the ngspice source code itself, using a console application or GUIs like gtk, Delphi,
Qt or others.
19.1 Compile options
19.1.1 How to get the sources
Currently (as of ngspice-27 being the actual release), you will have to use the direct
loading of the sources from the git repository (see Chapt. 32.1.2).
19.1.2 Linux, MINGW, CYGWIN
Compilation is done as described in Chapts. 32.1 or 32.2.2. Use the configure option
--with-ngshared instead of --with-x or --with-wingui. In addition you might add
(optionally) --enable-relpath to avoid absolute paths when searching for code models.
For MINGW you may edit compile_min.sh accordingly and compile using this script in
the MSYS2 window.
Other operation systems (Mac OS, BSD, ...) have not been tested so far. Your input is
welcome!
479
480CHAPTER 19. NGSPICE AS SHARED LIBRARY OR DYNAMIC LINK LIBRARY
19.1.3 MS Visual Studio
Compilation is similar to what has been described in Chapt. 32.2.1. However, there is
a dedicated project file coming with the source code to generate ngspice.dll. Go to the
directory visualc and start the project with double clicking on sharedspice.vcxproj.
19.2 Linking shared ngspice to a calling application

Basically there are two methods (as with all *.so, *.dll libraries). The caller may link to a
(small) library file during compiling/linking, and then immediately search for the shared
library upon being started. It is also possible to dynamically load the ngspice shared
library at runtime using the dlopen/LoadLibrary mechanisms.
19.2.1 Linking during creating the caller
While creating the ngspice shared lib, not only the *.so (*.dll) file is created, but also a
small library file, which just includes references to the exported symbols. Depending on
the OS, these may be called libngspice.dll.a, ngspice.lib. Linux and MINGW also allow
linking to the shared object itself. The shared object is not included into the executable
component but is tied to the execution.
19.2.2 Loading at runtime
dlopen (Linux) or LoadLibrary (MS Windows) will load libngspice.so or ngspice.dll into
the address space of the caller at runtime. The functions return a handle that may be
used to acquire the pointers to the functions exported by libngspice.so. Detaching ngspice
at runtime is equally possible (using dlclose/FreeLibrary), after the background thread
has been stopped and all callbacks have returned.
19.3 Shared ngspice API

The sources for the ngspice shared library API are contained in a single C file (shared-
spice.c) and a corresponding header file sharedspice.h. The type and function declarations
are contained in sharedspice.h, which may be directly added to the calling application, if
written in C or C++.
19.3.1 structs and types defined for transporting data
pvector_info is returned by the exported function ngGet_Vec_Info (see 19.3.2.5). Ad-

dresses of the vector name, type, real or complex data are transferred and may be read
asynchronously during or after the simulation.
19.3. SHARED NGSPICE API 481
vector_info
typedef struct vector_info {

char * v_name ; /* Same as so_vname */
int v_type ; /* Same as so_vtype */
short v_flags ; /* Flags ( a combination of VF_ *) */
double * v_realdata ; /* Real data */
ngcomplex_t * v_compdata ;/* Complex data */
int v_length ; /* Length of the vector */
} vector_info , * pvector_info ;
The next two structures are used by the callback function SendInitData (see 19.3.3.5).
Each time a new plot is generated during simulation, e.g. when a sequence of op, ac or
tran is used, or commands like linearize or fft are invoked, the function is called once
by ngspice. Among its parameters you find a pointer to a struct vecinfoall, which includes
an array of vecinfo, one for each vector. Pointers to the struct dvec, containing the vector,
are included.
vecinfo
typedef struct vecinfo

{
int number ; /* number of vector , as position in the
linked list of vectors , starts with 0 */
char * vecname ; /* name of the actual vector */
bool is_real ; /* TRUE if the actual vector has real data */
void * pdvec ; /* a void pointer to struct dvec *d , the
actual vector */
void * pdvecscale ; /* a void pointer to struct dvec * ds ,
the scale vector */
} vecinfo , * pvecinfo ;
vecinfoall
typedef struct vecinfoall

{
/* the plot */
char * name ;
char * title ;
char * date ;
char * type ;
int veccount ;
/* the data as an array of vecinfo with

length equal to the number of vectors
in the plot */
pvecinfo * vecs ;
} vecinfoall , * pvecinfoall ;
The next two structures are used by the callback function SendData (see 19.3.3.4). Each
time a new data point (e.g. time value and simulation output value(s)) is added to
the vector structure of the current plot, the function SendData is called by ngspice,
among its parameters the actual pointer pvecvaluesall, which contains an array of pointers
to pvecvalues, one for each vector. Logic return values are of type NG_BOOL, which is
typedefed to int.
vecvalues
typedef struct vecvalues {

char * name ; /* name of a specific vector */
double creal ; /* actual data value */
double cimag ; /* actual data value */
NG_BOOL is_scale ; /* if ’ name ’ is the scale vector */
NG_BOOL is_complex ; /* if the data are complex numbers */
} vecvalues , * pvecvalues ;
Pointer vecvaluesall to be found as parameter to callback function SendData.
vecvaluesall
typedef struct vecvaluesall {

int veccount ; /* number of vectors in plot */
int vecindex ; /* index of actual set of vectors , i . e .
the number of accepted data points */
pvecvalues * vecsa ; /* values of actual set of vectors ,
indexed from 0 to veccount - 1 */
} vecvaluesall , * pvecvaluesall ;
19.3.2 Exported functions

The functions listed in this chapter are the (only) symbols exported by the shared library.
19.3.2.1 int ngSpice_Init(SendChar*, SendStat*, ControlledExit*, SendData*,

SendInitData*, BGThreadRunning*, void*)
After caller has loaded ngspice.dll, the simulator has to be initialized by calling ngSpice_Init(...).
Address pointers of several callback functions (see 19.3.3), which are to be defined in the
caller, are sent to ngspice.dll. The int return value is not used.
Pointers to callback functions (details see 19.3.3):
SendChar* callback function for reading printf, fprintf, fputs (NULL allowed)
SendStat* callback function for reading status string and percent value (NULL allowed)
ControlledExit* callback function for transferring a flag to caller, generated by ngspice

upon a call to function controlled_exit. May be used by caller to detach ngspice.dll,
if dynamically loaded or to try any other recovery method, or to exit. (required)
SendData* callback function for sending an array of structs containing data values of
all vectors in the current plot (simulation output) (NULL allowed)
SendInitData* callback function for sending an array of structs containing info on all
vectors in the current plot (immediately before simulation starts) (NULL allowed)
BGThreadRunning* callback function for sending a boolean signal (true if thread is

running) (NULL allowed)
void* Using the void pointer, you may send the object address of the calling function
(’self’ or ’this’ pointer) to ngspice.dll. This pointer will be returned unmodified by
any callback function (see the *void pointers in Chapt. 19.3.3). Callback functions
are to be defined in the global section of the caller. Because they now have got the
object address of the calling function, they may direct their actions to the calling
object.
19.3.2.2 int ngSpice_Init_Sync(GetVSRCData* , GetISRCData* , GetSync-

Data* , int*, void*)
see Chapt. 19.6.
19.3.2.3 int ngSpice_Command(char*)
Send a valid command (see the control or interactive commands) from caller to ngspice.dll.
Will be executed immediately (as if in interactive mode). Some commands are rejected
(e.g. ’plot’, because there is no graphics interface). Command ’quit’ will remove internal
data, and then send a notice to caller via ngexit(). The function returns a ’1’ upon error,
otherwise ’0’.
Sending ngSpice_Command(NULL) will clear the internal control structures. Each com-
mand sent to ngspice is stored in the control structures. If you run scripts with 10.000 or
more commands, sending NULL from time to time will release this memory.
19.3.2.4 bool ngSpice_running (void)
Checks if ngspice is running in its background thread (returning ’true’).
19.3.2.5 pvector_info ngGet_Vec_Info(char*)
uses the name of a vector (may be in the form ’vectorname’ or <plotname>.vectorname)

as parameter and returns a pointer to a vector_info struct. The caller may then directly
assess the vector data (but better should not modify them).
19.3.2.6 int ngSpice_Circ(char**)
sends an array of null-terminated char* to ngspice.dll. Each char* contains a single line of
a circuit (Each line is like it is found in an input file *.sp.). The last entry to char** has to
be NULL. Upon receiving the array, ngspice.dll will immediately parse the input and set
up the circuit structure (as if the circuit is loaded from a file by the ’source’ command).
The function returns a ’1’ upon error, otherwise ’0’.
19.3.2.7 char* ngSpice_CurPlot(void)
returns to the caller a pointer to the name of the current plot. For a definition of the
term ’plot’ see Chapt. 17.3.
19.3.2.8 char** ngSpice_AllPlots(void)
returns to the caller a pointer to an array of all plots (listed by their typename).
19.3.2.9 char** ngSpice_AllVecs(char*)
returns to the caller a pointer to an array of all vector names in the plot named by the
string in the argument.
19.3.2.10 bool ngSpice_SetBkpt(double)
see Chapt. 19.6.

19.3.2.11 char *ngCM_Input_Path(char*)
Some XSPICE code models are supposed to read their input data (e.g. digital stimulus
data) from a file, like the Digital Source (12.4.21). There is a certain search sequence for
the location of this input file, e.g. the current directory or the directory of the previous
input (e.g. the netlist). However, when the netlist is entered into shared ngspice via the
function int ngSpice_Circ(char**) (19.3.2.6), there will be no information available about
a previous input directory. Therefore ngCM_Input_Path allows to send such directory
information to the XSPICE code models. Its return string is the newly set directory path,
if sent with argument NULL, it will return the currently available search path.
19.3.2.12 int ngSpice_LockRealloc(void)
19.3.2.13 int ngSpice_UnlockRealloc(void)
Locking and unlocking the realloc of output vectors during simulation. May be set before
and unset after reading output vectors in the primary thread, while the simulation in the
background thread is moving on.
19.3.2.14 int ngSpice_Init_Evt(SendEvtData*, SendInitEvtData*, void*)
return callback initialization addresses to caller
SendEvtData* data for a specific event node at time ’step’
SendInitEvtData* single line entry of event node dictionary (list)
void* pointer to user-defined data, will not be modified, but handed over back to caller
during Callback, e.g. address of calling object
19.3.2.15 pevt_shared_data ngGet_Evt_NodeInfo(char*)
Get info about the event node vector. If node_name is NULL, just delete previous data
19.3.2.16 char** ngSpice_AllEvtNodes(void)
get a list of all event nodes
19.3.3 Callback functions

Callback functions are a means to return data from ngspice to the caller. These functions
are defined as global functions in the caller, so to be reachable by the C-coded ngspice.
They are declared according to the typedefs given below. ngspice receives their addresses
from the caller upon initialization with the ngSpice_Init(...) function (see 19.3.2.1). If the
caller will not make use of a callback, it may send NULL instead of the address (except
for ControlledExit, which is always required).
If XSPICE is enabled, additional callback functions are made accessible by ngSpice_Init_Evt(...)
to obtain digital event node data.
If ngspice is run in the background thread (19.4.2), the callback functions (defined in
the caller) also are called from within that thread. One has to be carefully judging
how this behavior might influence the caller, where now you have the primary and the
background thread running in parallel. So make the callback function thread safe. The
integer identification number is only used if you run several shared libraries in parallel
(see Chapt. 19.6). Three additional callback function are described in Chapt. 19.6.3.
19.3.3.1 typedef int (SendChar)(char*, int, void*)
char* string to be sent to caller output
int identification number of calling ngspice shared lib (default is 0, see Chapt. 19.6)
void* return pointer received from caller during initialization, e.g. pointer to object
having sent the request
Sending output from stdout, stderr to caller. ngspice printf, fprintf, fputs, fputc functions
are redirected to this function. The char* string is generated by assembling the print
outputs of the above mentioned functions according to the following rules: The string
commences with ‘stdout ’, if directed to stdout by ngspice (with ‘stderr ’ respectively);
all tokens are assembled in sequence, taking the printf format specifiers into account, until
‘\n’ is hit. If set addescape is given in .spiceinit, the escape character \ is added to any
character from $[]\" found in the string.
Each callback function has a void pointer as the last parameter. This is useful in object
oriented programming. You may have sent the this (or self) pointer of the caller’s class
object to ngspice.dll during calling ngSpice_Init (19.3.2.1). The pointer is returned un-
modified by each callback, so the callback function may identify the class object that has
initialized ngspice.dll.
19.3.3.2 typedef int (SendStat)(char*, int, void*)
char* simulation status and value (in percent) to be sent to caller
void* return pointer received from caller
sending simulation status to caller, e.g. the string tran 34.5%.

19.3.3.3 typedef int (ControlledExit)(int, NG_BOOL, NG_BOOL, int, void*)
int exit status
NG_BOOL if true: immediate unloading dll, if false: just set flag, unload is done when
function has returned
NG_BOOL if true: exit upon ’quit’, if false: exit due to ngspice.dll error
asking for a reaction after controlled exit.
19.3.3.4 typedef int (SendData)(pvecvaluesall, int, int, void*)
vecvaluesall* pointer to array of structs containing actual values from all vectors
int number of structs (one per vector)
send back actual vector data.
19.3.3.5 typedef int (SendInitData)(pvecinfoall, int, void*)
vecinfoall* pointer to array of structs containing data from all vectors right after ini-
tialization
send back initialization vector data.
19.3.3.6 typedef int (BGThreadRunning)(NG_BOOL, int, void*)
NG_BOOL false if background thread is running, otherwise true
indicate if background thread is running
Callback functions addresses received from caller with ngSpice_Init_Evt()

function:
19.3.3.7 typedef int (SendEvtData)(int, double, double, char *, void *, int,

int, int, void*)
int node index
double actual simulation time
double a real value for specified structure component for plotting purposes
char* a string value for specified structure component for printing
void* a binary data structure
int size of the binary data structure
int the mode (op, dc, tran) we are in
int identification number of calling ngspice shared lib
Upon a time step finished, called per node.
19.3.3.8 typedef int (SendInitEvtData)(int, int, char*, char*, int, void*)
int node index
int maximum node index, number of nodes
char* node name
char* udn-name, node type
Upon initialization, called once per event node to build up a dictionary of nodes.
19.4 General remarks on using the API
19.4.1 Loading a netlist

Basically the input to shared ngspice is the same as if you would start a ngspice batch
job, e.g. you enter a netlist and the simulation command (any .dot analysis command
like .tran, .op, or .dc etc. as found in Chapt. 15.3), as well as suitable options.
Typically you should not include a .control section in your input file. Any script
described in a .control section for standard ngspice should better be emulated by the
caller and be sent directly to ngspice.dll. Start the simulation according to Chapt. 19.4.2
in an extra thread.
19.4. GENERAL REMARKS ON USING THE API 489
As an alternative, only the netlist has to be entered (without analysis command), then you
may use any interactive command as listed in Chapt. 17.5 (except for the plot command).
However, for users without direct access to source code commands (e.g. KiCad users), it
might be advantageous to add a .control section to their netlist simulation dot commands.
please be careful and check for chapter 19.4.1.4.
The ‘typical usage’ examples given below are part of a caller written in C.
19.4.1.1 Loading from file
As with interactive ngspice, you may use the ngspice internal command source (17.5.86)
to load a complete netlist from a file.
Typical usage:
ngSpice_Command (" source ../ examples / adder_mos . cir ");
19.4.1.2 Loading line by line
As with interactive ngspice, you may use the ngspice internal command circbyline (17.5.14)
to send a netlist line by line to the ngspice circuit parser.
Typical usage:
ngSpice_Command (" circbyline fail test ");

ngSpice_Command (" circbyline V1 1 0 1");
ngSpice_Command (" circbyline R1 1 0 1");
ngSpice_Command (" circbyline . dc V1 0 1 0.1");
ngSpice_Command (" circbyline . end ");
The first line is a title line, which will be ignored during circuit parsing. As soon as the
line .end has been sent to ngspice, circuit parsing commences.
19.4.1.3 Loading as a string array
Typical usage:
circarray = ( char **) malloc ( sizeof ( char *) * 7);

circarray [0] = strdup (" test array ");
circarray [1] = strdup (" V1 1 0 1");
circarray [2] = strdup (" R1 1 2 1");
circarray [3] = strdup (" C1 2 0 1 ic =0");
circarray [4] = strdup (". tran 10 u 3 uic ");
circarray [5] = strdup (". end ");
circarray [6] = NULL ;
ngSpice_Circ ( circarray );
An array of char pointers is malloc’d, each netlist line is then copied to the array. strdup
will care for the memory allocation. The first entry to the array is a title line, the last
entry has to contain NULL. ngSpice_Circ(circarray); sends the array to ngspice, where
circuit parsing is started immediately. Don’t forget to free the array after sending it, to
avoid a memory leak.
For the latter two options to load a netlist, there is some caveat though. When sending
the netlist from caller to shared ngspice, ngspice will not get any automatic notion of a
potential input directory, as is possible and useed with standard ngspice. You will either
have to set the environmental variable NGSPICE_INPUT_DIR to the input file path,
especially when in the netlist other .include ./nextinput.inc commands with relative
paths are used or you are using XSPICE code models that require loading an input file.
Or you may set the variable sourcepath (17.7) in .spiceinit. The command
set sourcepath = ( D:/mypath/input $sourcepath )
will add D:/mypath/input to the front of the path list, only this leading path entry is
sent to the code models.
19.4.1.4 Using a .control section
If the simulation is started with the background thread (command bg_run), the .control
section commands are executed immediately after bg_run has been given, i.e. typically
before the simulation has finished. This often is not very useful because you want to
evaluate the simulation results. If the predefined variable controlswait is set in .spiceinit
or spice.rc, the command execution is delayed until the background thread has returned
(aka the simulation has finished). If set controlswait is given inside of the .control
section, only the commands following this statement are delayed.
19.4.2 Running the simulation
The following commands are used to start the simulator in its own thread, halt the
simulation and resume it again. The extra (background) thread enables the caller to
continue with other tasks in the main thread, e.g. watching its own event loop. Of course
you have to take care that the caller will not exit before ngspice is finished, otherwise
you immediately will lose all data. After having halted the simulator by suspending the
background thread, you may assess data, change ngspice parameters, or read output data
using the caller’s main thread, before you resume simulation using a background thread
again. While the background thread is running, ngspice will reject any other command
sent by ngSpice_Command.
Typical usage:
ngSpice_Command (" bg_run ");

...
ngSpice_Command (" bg_halt ");
...
ngSpice_Command (" bg_resume ");
19.4. GENERAL REMARKS ON USING THE API 491
Basically you may send the commands ’run’ or ’resume’ (no prefix bg_), starting ngspice
within the main thread. The caller then has to wait until ngspice returns from simulation.
A command ’halt’ is not available then.
After simulation is finished (test with callback 19.3.3.6), you may send other commands
from Chapt. 17.5, emulating any .control script. These commands are executed in the
main thread, which should be okay because execution time is typically short.
19.4.3 Accessing data
19.4.3.1 Synchronous access
The callback functions SendInitData (19.3.3.5) and SendData (19.3.3.4) allow access
to simulator output data synchronized with the simulation progress.
Each time a new plot is generated during simulation, e.g. when a sequence of op, ac and
tran is used or commands like linearize or fft are invoked, the callback SendInitData
is called by ngspice. Immediately after setting up the vector structure of the new plot,
the function is called once. Its parameter is a pointer to the structure vecinfoall (19.3.1),
which contains an array of structures vecinfo, one for each vector in the actual plot. You
may simply use vecname to get the name of any vector. This time the vectors are still
empty, but pointers to the vector structure are available.
Each time a new data point (e.g. time value and simulation output value(s)) is added to
the vector structure of the current plot, the function SendData is called by ngspice. This
allows you to immediately access the simulation output synchronized with the simulation
time, e.g. to interface it to a runtime plot or to use it for some controlled simulation
by stopping the simulation based on a condition, altering parameters and resume the
simulation. SendData returns a structure vecvaluesall as parameter, which contains an
array of structures vecvalues, one for each vector.
Some code to demonstrate the callback function usage is referenced below (19.5).
19.4.3.2 Asynchronous access
During simulation, while the background thread is running, or after it is finished, you
may use the functions ngSpice_CurPlot (19.3.2.7), ngSpice_AllPlots (19.3.2.8),
ngSpice_AllVecs (19.3.2.9) to retrieve information about vectors available, and func-
tion ngGet_Vec_Info (19.3.2.5) to obtain data from a vector and its corresponding
scale vector. The timing of the caller and the simulation progress are independent from
each other and not synchronized.
Again some code to demonstrate the callback function usage is referenced below (19.5).
19.4.3.3 XSPICE event node data
After starting the simulation, in a first step the callback function SendInitEvtData is
called once for each event node. All nodes are numbered in ascending order. The first
function argument is the actual node number, the second sets the total amount of nodes,
then node name and node type follow. You may set up an array to store name and type,
indexed by the node number.
During simulation, after each time step ngspice checks if a node has changed. If so,
SendEvtData is called for each node that changed, returning the simulation time, the
node number, and the node value as a char* string, consisting of one out of 0s, 1s, Us, 0r,
1r, Ur, 0z, 1z, Uz, 0u, 1u, Uu (see 12.5.1). The double real value and the void* binary
data structure arguments are for future enhancements of the data interface. The int mode
returns 0 for op, 1 for dc, 2 for ac , and 3 for tran simulation. The final int is useful to
identify the ngspice lib by number if you run several in parallel (see 19.6). The final *void
just returns the pointer received from caller. e.g. to identify the calling object.
19.4.4 Altering model or device parameters
After halting ngspice by stopping the background thread (19.4.2), nearly all ngspice com-
mands are available. Especially alter (17.5.3) and altermod (17.5.4) may be used to
change device or model parameters. After the modification, the simulation may be re-
sumed immediately. Changes to a circuit netlist, however, are not possible. You would
need to load a complete new netlist (19.4.1) and restart the simulation from the beginning.
19.4.5 Output
After the simulation is finished, use the ngspice commands write (17.5.107) or wrdata
(17.5.106) to output data to a file as usual, use the print command (17.5.59) to retrieve
data via callback SendChar (19.3.3.1), or refer to accessing the data as described in
Chapt. 19.4.3.
Typical usage:
ngSpice_Command (" write testout . raw V (2)");

ngSpice_Command (" print V (2)");
19.4.6 Error handling
There are several occasions where standard ngspice suffers from an error, cannot recover
internally and then exits. If this is happening to the shared module this would mean that
the parent application, the caller, is also forced to exit. Therefore (if not suffering from
a segfault) ngspice.dll will call the function controlled_exit as usual, this now calls the
callback function ’ControlledExit’ (19.3.3.3), which hands over the request for exiting to
the caller. The caller now has the task to handle the exit code for ngspice.
If ngspice has been linked at runtime by dlopen/LoadLibrary (see 19.2.2), the callback may
close all threads, and then detach ngspice.dll by invoking dlclose/FreeLibrary. The caller
may then restart ngspice by another loading and initialization (19.3.2.1).
If ngspice is included during linking the caller (see 19.2.1), there is not yet a good and
general solution to error handling, if the error is non-recoverable from inside ngspice.
19.5. EXAMPLE APPLICATIONS 493
19.5 Example applications

Three executables (coming with source code) serve as examples for controlling ngspice.
These are not meant to be ‘production’ programs, but just give some commented example
usages of the interface.
ng_start.exe is a MS Windows application loading ngspice.dll dynamically. All functions
and callbacks of the interface are assessed. The source code, generated with Turbo Delphi
2006, may be found here, the binaries compiled for 32 Bit are here.
Two console applications, compilable with Linux, CYGWIN, MINGW or MS Visual Stu-
dio, are available here, demonstrating either linking upon start-up or loading shared
ngspice dynamically at runtime. A simple feedback loop is shown in tests 3 and 4, where
a device parameter is changed upon having an output vector value crossing a limit.
An XSPICE event node example may be assessed at ngspice/visualc/ng_shared_xspice_v,
currently tested only with MS Windows and compiled with Visual Studio.
19.6 ngspice parallel

The following chapter describes an offer to the advanced user and developer community.
If you are interested in evaluating the parallel and synchronized operation of several
ngspice instances, this may be one way to go. However, no ready to use implementation
is available. You will find a toolbox and some hints how to use it. Parallelization and
synchronization is your task by developing a suitable caller! And of course another major
input has to come from partitioning the circuit into suitable, loosely coupled pieces,
each with its own netlist, one netlist per ngspice instance. And you have to define the
coupling between the circuit blocks. Both are not provided by ngspice, but are again your
responsibility. Both are under active research, and the toolbox described below is an offer
to join that research.
19.6.1 Go parallel!
A simple way to run several invocations of ngspice in parallel for transient simulation is to
define a caller that loads two or more ngspice shared libraries. There is one prerequisite
however to do so: the shared libraries have to have different names. So compile ngspice
shared lib (see 19.1), then copy and rename the library file, e.g. ngspice.dll may become
ngspice1.dll, ngspice2.dll etc. Then dynamically load ngspice1.dll, retrieve its address,
initialize it by calling ngSpice_init() (see 19.3.2.1), then continue initialization by calling
ngSpice_init_Sync() (see 19.6.2.1). An integer identification number may be sent during
this step to later uniquely identify each invocation of the shared library, e.g. by having
any callback use this identifier. Repeat the sequence with ngspice2.dll and so on.
Inter-process communication and synchronization is now done by using three callback
functions. To understand their interdependence, it might be useful to have a look at the
transient simulation sequence as defined in the ngspice source file dctran.c. The following
listing includes the shared library option (It differs somewhat from standard procedure)
and disregards XSPICE.
1. initialization
2. calculation of operating point
3. next time step: set new breakpoints (VSRC, ISRC, TRA, LTRA)
4. send simulation data to output, callback function SendData* datfcn
5. check for autostop and other end conditions
6. check for interrupting simulation (e.g. by bg_halt)
7. breakpoint handling (e.g. enforce breakpoint, set new small cktdelta if directly after
the breakpoint)
8. calling ngspice internal function sharedsync() that invokes callback function Get-
SyncData* getsync with location flag loc = 0
9. save the previous states
10. start endless loop
11. save cktdelta to olddelta, set new time point by adding cktdelta to ckttime
12. new iteration of circuit at new time point, which uses callback functions GetVS-
RCData* getvdat and GetISRCData* getidat to retrieve external voltage or
current inputs, returns redostep=0, if converged, redostep=1 if not converged
13. if not converged, divide cktdelta by 8
14. check for truncation error with all non-linear devices, if necessary create a new
(smaller) cktdelta to limit the error, optionally change integration order
15. calling ngspice internal function sharedsync() that invokes callback function Get-
SyncData* getsync with location flag loc = 1: as a result either goto 3 (next
time step) or to 10 (loop start), depending on ngspice and user data, see the next
paragraph.
The code of the synchronization procedure is handled in the ngspice internal function
sharedsync() and its companion user defined callback function GetSyncData* getsync.
The actual setup is as follows:
If no synchronization is asked for (GetSyncData* set to NULL), program control jumps
to ’next time step’ (3) if redostep==0, or subtracts olddelta from ckttime and jumps to
’loop start’ (9) if redostep <> 0. This is the standard ngspice behavior.
If GetSyncData* has been set to a valid address by ngSpice_Init_Sync(), the callback
function getsync is involved. If redostep <> 0, olddelta is subtracted from ckttime,
getsync is called, either the cktdelta time suggested by ngspice is kept or the user provides
his own deltatime, and the program execution jumps to (9) for redoing the last step with
the new deltatime. The return value of getsync is not used. If redostep == 0, getsync
is called. The user may keep the deltatime suggested by ngspice or define a new value. If
the user sets the return value of getsync to 0, the program execution then jumps to ’next
time step’ (3). If the return value of getsync is 1, olddelta is subtracted from ckttime,
19.6. NGSPICE PARALLEL 495
and the program execution jumps to (9) for redoing the last step with the new deltatime.
Typically the user provided deltatime should be smaller than the value suggested by
ngspice.
19.6.2 Additional exported functions
The following functions (exported or callback) are designed to support the parallel action
of several ngspice invocations. They may be useful, however, also when only a single
library is loaded into a caller, if you want to use external voltage or current sources or
’play’ with advancing simulation time.
19.6.2.1 int ngSpice_Init_Sync(GetVSRCData* , GetISRCData* , GetSync-

Data* , int*, void*)
GetVSRCData* callback function for retrieving a voltage source value from caller
(NULL allowed)
GetISRCData* callback function for retrieving a current source value from caller (NULL
allowed)
GetSyncData* callback function for synchronization (NULL allowed)
More pointers
int* pointer to integer unique to this shared library (defaults to 0)
void* pointer to user-defined data, will not be modified, but handed over back to caller
during Callback, e.g. address of calling object. If NULL is sent here, userdata
info from ngSpice_Init() will be kept, otherwise userdata will be overridden by new
value from here.
19.6.2.2 NG_BOOL ngSpice_SetBkpt(double)
Sets a breakpoint in ngspice, a time point that the simulator is enforced to hit during the
transient simulation. After the breakpoint time has been hit, the next delta time starts
with a small value and is ramped up again. A breakpoint should be set only when the
background thread in ngspice is not running (before the simulation has started, or after the
simulation has been paused by bg_halt). The time sent to ngspice should be larger than
the current time (which is either 0 before start or given by the callback GetSyncData
(19.6.3.3). Several breakpoints may be set.
19.6.3 Additional callback functions

19.6.3.1 typedef int (GetVSRCData)(double*, double, char*, int, void*)
double* return voltage value
double actual time
char* node name
Ask for a VSRC EXTERNAL voltage value. The independent voltage source (see Chapt.
4.1) with EXTERNAL option sets a voltage value to the node defined in the netlist and
named here at the time returned by the simulator.
19.6.3.2 typedef int (GetISRCData)(double*, double, char*, int, void*)
double* return current value
double actual time
char* node name
Ask for ISRC EXTERNAL value. The independent current source (see Chapt. 4.1) with
EXTERNAL option allows setting a current value to the node defined by the netlist and
named here at the time returned by the simulator.
19.6.3.3 typedef int (GetSyncData)(double, double*, double, int, void*)
double actual time (ckt->CKTtime)
double* delta time (ckt->CKTdelta)
double old delta time (olddelta)
int location of call for synchronization in dctran.c
Ask for new delta time depending on synchronization requirements. See 19.6.1 for an
explanation.
19.6. NGSPICE PARALLEL 497
19.6.4 Parallel ngspice example

A first example is available as a compacted 7z archive. It contains the source code of a
controlling application, as well as its compiled executable and ngspice.dll (for MS Win-
dows). As the input circuit an inverter chain has been divided into three parts. Three
ngspice shared libraries are loaded, each simulates one partition of the circuit. Intercon-
nections between the partitions are provided via a callback function. The simulation time
is synchronized among the three ngspice invocations by another callback function.
Chapter 20
TCLspice
Spice historically comes as a simulation engine with a Command Line Interface. The
Spice engine can also be used with a Graphical User Interface. Tclspice represents a third
approach to interfacing ngspice simulation functionality. Tclspice is nothing more than
a new way of compiling and using SPICE source code. Spice is no longer considered as
a standalone program but as a library invoked by a TCL interpreter. It either permits
direct simulation in a TCL shell (this is quite analogous to the command line interface
of ngspice), or it permits the elaboration of more complex, more specific, or more user
friendly simulation programs, by writing TCL scripts.
20.1 tclspice framework
The technical difference between the ngspice CLI interface and tclspice is that the CLI
interface is compiled as a standalone program, whereas tclspice is a shared object. Tclspice
is designed to work with tools that expand the capabilities of ngspice: TCL for the
scripting and programming language interface and BLT for data processing and display.
This two tools give tclspice all of its relevance, with the insurance that the functionality
is maintained by competent people.
Making tclspice (see 20.6) produces two files: libspice.so and pkgIndex.tcl. libspice.so is the
executable binary that the TCL interpreter calls to handle SPICE commands. pkgIndex.tcl
take place in the TCL directory tree, providing the SPICE package1 to the TCL user.
BLT is a TCL package. It is quite well documented. It permits handling mathematical
vector data structures for calculus and display, in a Tk interpreter like wish.
20.2 tclspice documentation
A detailed documentation on tclspice commands is available on the original tclspice web

page.
1
package has to be understood as the TCL package
499
500 CHAPTER 20. TCLSPICE
20.3 spicetoblt
Tclspice opens its doors to TCL and BLT with a single specific command spicetoblt.
TCLspice gets its identity in the command spice::vectoblt . This command copies data
computed by the simulation engine into a tcl variable. vectoblt is composed of three
words: vec, to and blt. Vec means SPICE vector data. To is the English preposition, and
blt is a useful tcl package providing a vector data structure. Example:
blt :: vector create Iex

spice :: vectoblt Vex # branch Iex
Here an empty blt vector is created. It is then filled with the vector representation of the
current flowing out of source Vex. Vex#branch is native SPICE syntax. Iex is the name
of the BLT vector.
The reverse operation is handled by native SPICE commands, such as alter, let and
set.
20.4 Running TCLspice
TCLspice consists of a library or a package to include in your tcl console or script:
load / somepath / libspice . so

package require spice
Then you can execute any native SPICE command by preceding it with spice::. For
example if you want to source the testCapa.cir netlist, type the following:
spice :: source testCapa . cir

spice :: spicetoblt example ...
Plotting data is not a matter of SPICE, but of tcl. Once the data is stored in a blt vector,
it can be plotted. Example:
blt :: graph . cimvd - title " Cim = f ( Vd )"

pack . cimvd
. cimvd element create line1 - xdata Vcmd - ydata Cim
With blt::graph a plotting structure is allocated in memory. With pack it is placed into
the output window, and becomes visible. The last command, and not the least, plots the
function Cim = f (Vcmd ), where Cim and Vcmd are two BLT vectors.
20.5. EXAMPLES 501
20.5 examples
20.5.1 Active capacitor measurement
This is a crude implementation of a circuit described by Marc Kodrnja, in his PhD thesis
that was found on the Internet. The simulation outputs a graph representing virtual
capacitance versus a control voltage. The function C = f (V ) is calculated point by
point. For each control voltage value, the virtual capacitance is calculated in a frequency
simulation. A control value that should be as close to zero as possible is calculated to
assess simulation success.
20.5.1.1 Invocation:
This script can be invoked by typing wish testbench1.tcl
20.5.1.2 testbench1.tcl
This line loads the simulator capabilities
This is a comment (Quite useful if you intend to live with other Human beings)
# Test of virtual capacitor circuit

# Vary the control voltage and log the resulting capacitance
A good example of the calling of a SPICE command: precede it with spice::
spice :: source " testCapa . cir "
This reminds that any regular TCL command is of course possible
set n 30 set dv 0.2

set vmax [ expr $dv /2]
set vmin [ expr -1 * $dv /2]
set pas [ expr $dv / $n ]
BLT vector is the structure used to manipulate data. Instantiate the vectors
blt :: vector create Ctmp

blt :: vector create Cim
blt :: vector create check
blt :: vector create Vcmd
Data is, in my coding style, plotted into graph objects. Instantiate the graph
blt :: graph . cimvd - title " Cim = f ( Vd )"

blt :: graph . checkvd - title " Rim = f ( Vd )"
blt :: vector create Iex
blt :: vector create freq
blt :: graph . freqanal - title " Analyse frequentielle "
#
# First simulation : A simple AC plot
#
set v [ expr { $vmin + $n * $pas / 4}]
spice :: alter vd = $v
spice :: op
spice :: ac dec 10 100 100 k
Retrieve a the intensity of the current across Vex source
spice :: vectoblt { Vex # branch } Iex
Retrieve the frequency at which the current have been assessed
spice :: vectoblt { frequency } freq
Room the graph in the display window
pack . freqanal
Plot the function Iex =f(V)
. freqanal element create line1 - xdata freq - ydata Iex

#
# Second simulation : Capacitance versus voltage control
# for { set i 0} {[ expr $n - $i ]} { incr i }
# { set v [ expr { $vmin + $i * $pas }]
spice :: alter vd = $v
spice :: op spice :: ac dec 10 100 100 k
Image capacitance is calculated by SPICE, instead of TCL there is no objective reason
spice :: let Cim = real ( mean ( Vex # branch /(2* Pi * i * frequency *( V (5) - V (6))
spice :: vectoblt Cim Ctmp
Build function vector point by point
Cim append $Ctmp (0: end )
Build a control vector to check simulation success

20.5. EXAMPLES 503
spice :: let err = real ( mean ( sqrt (( Vex # branch -

(2* Pi * i * frequency * Cim * V (5) - V (6)))^2)))
spice :: vectoblt err Ctmp check
append $Ctmp (0: end )
Build abscissa vector
FALTA ALGO ... Vcmd append $v }
Plot
pack . cimvd
. cimvd element create line1 - xdata Vcmd - ydata Cim
pack . checkvd
. checkvd element create line1 - xdata Vcmd - ydata check
20.5.2 Optimization of a linearization circuit for a Thermistor
This example is both the first and the last optimization program written for an electronic
circuit. It is far from perfect.
The temperature response of a CTN is exponential. It is thus nonlinear. In a battery
charger application floating voltage varies linearly with temperature. A TL431 voltage
reference sees its output voltage controlled by two resistors (r10, r12) and a thermistor
(r11). The simulation is run at a given temperature. The thermistor is modeled in
SPICE by a regular resistor. Its resistivity is assessed by the TCL script. It is set
with a spice::alter command before running the simulation. This script uses an iterative
optimization approach to try to converge to a set of two resistor values that minimizes
the error between the expected floating voltage and the TL431 output.
20.5.2.1 Invocation:
This script can be executed by the user by simply executing the file in a terminal.
./ testbench3 . tcl
Two issues2 are important to point out:

2
For those who are really interested in optimizing circuits: Some parameters are very important for
quick and correct convergence. The optimizer walks step by step to a local minimum of the cost function
you define. Starting from an initial vector you provide, it converges step by step. Consider trying another
start vector if the result is not the one you expected.
The optimizer will carry on walking until it reaches a vector whose resulting cost is smaller than the
target cost you provided. You must also provide a maximum iteration count in case the target can not
be achieved. Balance time, specifications, and every other parameter. For a balance between quick and
accurate convergence adjust the ‘factor’ variable, at the beginning of minimumSteepestDescent in the file
differentiate.tcl.
• During optimization loop, graphical display of the current temperature response is

not yet possible and I don’t know why. Each time a simulation is performed, some
memory is allocated for it.
• The simulation result remains in memory until the libspice library is unloaded (typ-
ically: when the tcl script ends) or when a spice::clean command is performed. In
this kind of simulation, not cleaning the memory space will freeze your computer
and you’ll have to restart it. Be aware of that.
This calls the shell sh who then runs wish with the file itself.
#!/ bin / sh
# WishFix \
exec wish " $0 " $ {1+" $@ "}
#
#
#
Regular package for simulation
Here the important line is source differentiate.tcl that contains the optimization library
source differentiate . tcl
Generates a temperature vector
proc temperatures_calc { temp_inf temp_sup points } {

set tstep [ expr " ( $temp_sup - $temp_inf ) / $points " ]
set t $temp_inf
set temperatures ""
for { set i 0 } { $i < $points } { incr i } {
set t [ expr { $t + $tstep } ]
set temperatures " $temperatures $t "
}
return $temperatures }
generates thermistor resistivity as a vector, typically run: thermistance_calc res B [

temperatures_calc temp_inf temp_sup points ]
20.5. EXAMPLES 505
proc thermistance_calc { res B points } {

set tzero 273.15
set tref 25
set thermistance ""
foreach t $points {
set res_temp [ expr " $res *
+ exp ( $B * ( 1 / ( $tzero + $t ) -
+ 1 / ( $tzero + $tref ) ) ) " ]
set thermistance " $thermistance $res_temp "
}
return $thermistance }
generates the expected floating value as a vector, typically run: tref_calc res B [ temper-
atures_calc temp_inf temp_sup points ]
proc tref_calc { points } {

set tref ""
foreach t $points {
set tref " $tref [ expr "6*(2.275 -0.005*( $t -20)) -9"]"
}
return $tref }
In the optimization algorithm, this function computes the effective floating voltage at the
given temperature.
### NOTE :
### As component values are modified by a spice ::
alter
### Component values can be considered as global
variable .
### R10 and R12 are not passed to iteration function
### because it is expected to be correct , i . e . to
### have been modified soon before
proc iteration { t } { set tzero 273.15 spice :: alter
r11 = [ thermistance_calc 10000 3900 $t ]
# Temperature simulation often crashes . Comment it
out ...
# spice :: set temp = [ expr " $tzero + $t " ]
spice :: op
spice :: vectoblt vref_temp tref_tmp
### NOTE :
### As the library is executed once for the
### whole script execution , it is important to manage
the memory
### and regularly destroy unused data set . The data
### computed here will not be reused . Clean it
spice :: destroy all return [ tref_tmp range 0 0 ] }
This is the cost function optimization algorithm will try to minimize. It is a 2-norm
(Euclidean norm) of the error across the temperature range [-25:75]°C.
proc cost { r10 r12 } {

tref_blt length 0
spice :: alter r10 = $r10
foreach point [ temperatures_blt range 0
+ [ expr " [ temperatures_blt length ] - 1" ] ]
{
+ tref_blt append [ iteration $point ]
}
set result [ blt :: vector expr " 1000 *
sum (( tref_blt - expected_blt ) ^2 ) " ]
disp_curve $r10 $r12
return $result }
This function displays the expected and effective value of the voltage, as well as the r10
and r12 resistor values
proc disp_curve { r10 r12 }

+ { . g configure - title " Valeurs optimales : R10 =
$r10 R12 = $r12 " }
Main loop starts here
#
# Optimization
# blt :: vector create tref_tmp
blt :: vector create tref_blt
blt :: vector create expected_blt
blt :: vector create temperatures_blt temperatures_blt
append [ temperatures_calc -25 75 30 ] expected_blt
append [ tref_calc [ temperatures_blt range 0
+ [ expr " [ temperatures_blt length ] - 1" ] ]
]
blt :: graph . g
pack . g - side top - fill both - expand true
. g element create real - pixels 4 - xdata
temperatures_blt
+ - ydata tref_blt
. g element create expected - fill red - pixels 0 -
dashes
+ dot - xdata temperatures_blt - ydata
expected_blt
Source the circuit and optimize it. The result is retrieved in the variable r10r12e and put
into r10 and r12 with a regular expression. A bit ugly.
20.5. EXAMPLES 507
spice :: source FB14 . cir

set r10r12 [ :: math :: optimize :: minimumSteepestDescent
+ cost { 10000 10000 } 0.1 50 ]
regexp {([0 -9.]*) ([0 -9.]*) } $r10r12 r10r12 r10 r12
Outputs optimization result
#
# Results
# spice :: alter r10 = $r10
foreach point [ temperatures_blt range 0
+ [ expr " [ temperatures_blt length ] - 1" ] ] {
tref_blt append [ iteration $point ]
}
disp_curve $r10 $r12
20.5.3 Progressive display
This example is quite simple but it is very interesting. It displays a transient simulation
result on the fly. You may now be familiar with most of the lines of this script. It uses
the ability of BLT objects to automatically update. When the vector data is modified,
the strip-chart display is modified accordingly.
#!/ bin / sh
# WishFix \
exec wish -f " $0 " $ {1+" $@ "}
###
package require BLT package require spice
this avoids to type blt:: before the blt class commands
namespace import blt ::*

wm title . " Vector Test script "
wm geometry . 800 x600 +40+40 pack propagate . false
A strip chart with labels but without data is created and displayed (packed)
stripchart . chart
pack . chart - side top - fill both - expand true
. chart axis configure x - title " Time " spice :: source example . cir
spice :: bg
run after 1000 vector
create a0 vector
create b0 vectorry
create a1 vector
create b1 vector
create stime
proc bltupdate {} {
puts [ spice :: spice_data ]
spice :: spicetoblt a0 a0
spice :: spicetoblt b0 b0
spice :: spicetoblt a1 a1
spice :: spicetoblt b1 b1
spice :: spicetoblt time stime
after 100 bltupdate }
bltupdate . chart element create a0 - color red - xdata
+ stime - ydata a0
. chart element create b0 - color blue - xdata stime - ydata b0
. chart element create a1 - color yellow - xdata stime - ydata a1
. chart element create b1 - color black - xdata stime - ydata b1
20.6 Compiling
20.6.1 Linux
Get tcl8.4 from your distribution. You will need the blt plotting package (compatible to
the old tcl 8.4 only) from here. See also the actual blt wiki.
./configure --with-tcl ..
make
sudo make install
20.6.2 MS Windows
Can be done, but is tedious. Here it is described by a procedure on Windows 7, 64 Bit
Home Edition.
20.6.2.1 Downloads
download tcl8.6b2-src.zip from http://www.tcl.tk/software/tcltk/download.html

download tk8.6b2-src.zip
20.7. MS WINDOWS 32 BIT BINARIES 509
download blt from http://ngspice.sourceforge.net/experimental/blt2.4z.7z

expand all to d:\software
20.6.2.2 Tcl
double click on D:\software\tcl8.6b2\win\tcl.dsw

convert to MS Visual Studio 2008 project
select release or debug
create tcl as tcl86t.dll.
20.6.2.3 Tk
edit D:\software\tk8.6b2\win\buildall.vc.bat
line 31 to
call C:\Program Files (x86)\Microsoft Visual Studio 9.0\VC\vcvarsall.bat
line 53 to
if "%TCLDIR%" == "" set TCLDIR=..\..\tcl8.6b2
open cmd window
cd to
d:\software\tk8.6b2\win>
then
d:\software\tk8.6b2\win> buildall.vc.bat debug
tk will be made as tk86t.dll, in addition wish86t.exe is generated.
20.6.2.4 blt
blt source files have been downloaded from the blt web page and modified for compatibility
with TCL8.6. To facilitate making blt24.dll, the modified source code is available as a 7z
compressed file, including a project file for MS Visual Studio 2008.
20.6.2.5 tclspice
ngspice is compiled and linked into a dll called spice.dll that may be loaded by wish86t.exe.
MS Visual Studio 2008 is the compiler applied. A project file may be downloaded as a 7z
compressed file. Expand this file in the ngspice main directory. The links to tcl and tk are
hard-coded, so both have to be installed in the places described above (d:\software\...).
spice.dll may be generated in Debug, Release or ReleaseOMP mode.
20.7 MS Windows 32 Bit binaries

You may download the compiled binaries, including tcl, tk, blt and tclspice, plus the
examples, slightly modified, from http://ngspice.sourceforge.net/experimental/tclspice-
25.7z.
Chapter 21
Example Circuits
This section starts with an ngspice example to walk you through the basic features of
ngspice using its command line user interface. The operation of ngspice will be illustrated
through several examples (Chapt. 21.1 to 21.7).
The first example uses the simple one-transistor amplifier circuit illustrated in Fig. 21.1.
This circuit is constructed entirely with ngspice compatible devices and is used to intro-
duce basic concepts, including:
• Invoking the simulator:
• Running simulations in different analysis modes
• Printing and plotting analog results
• Examining status, including execution time and memory usage
• Exiting the simulator
The remainder of the section (from Chapt. 21.1 onward) lists several circuits, which
have been accompanying any ngspice distribution, and may be regarded as the ‘classical’
SPICE circuits.
21.1 AC coupled transistor amplifier
The circuit shown in Fig. 21.1 is a simple one-transistor amplifier. The input signal
is amplified with a gain of approximately -(Rc/Re) = -(3.9K/1K) = -3.9. The circuit
description file for this example is shown below.
511
512 CHAPTER 21. EXAMPLE CIRCUITS
Figure 21.1: Transistor Amplifier Simulation Example
Example:
A Berkeley SPICE3 compatible circuit

*
* This circuit contains only Berkeley SPICE3 components .
*
* The circuit is an AC coupled transistor amplifier with
* a sinewave input at node "1" , a gain of approximately -3.9 ,
* and output on node " coll ".
*
. tran 1e -5 2e -3
*
vcc vcc 0 12.0
vin 1 0 0.0 ac 1.0 sin (0 1 1 k )
ccouple 1 base 10 uF
rbias1 vcc base 100 k
rbias2 base 0 24 k
q1 coll base emit generic
rcollector vcc coll 3.9 k
remitter emit 0 1 k
*
. model generic npn
*
. end
To simulate this circuit, move into a directory under your user account and copy the
21.1. AC COUPLED TRANSISTOR AMPLIFIER 513
file xspice_c1.cir from directory /examples/xspice/. This file stems from the original
XSPICE introduction, therefore its name, but you do not need to have a version of ngspice
with the XSPICE option to run it.
$ cp /examples/xspice/xspice_c1.cir xspice_c1.cir
Now invoke the simulator on this circuit as follows:
$ ngspice xspice_c1.cir
After a few moments, you should see the ngspice prompt:
ngspice 1 ->
At this point, ngspice has read-in the circuit description and checked it for errors. If
any errors had been encountered, messages describing them would have been output to
your terminal. Since no messages were printed for this circuit, the syntax of the circuit
description was correct.
To see the circuit description read by the simulator you can issue the following command:
ngspice 1 -> listing
The simulator shows you the circuit description currently in memory:
a berkeley spice3 compatible circuit

1 : a berkeley spice3 compatible circuit
2 : .global gnd
10 : .tran 1e-5 2e-3
12 : vcc vcc 0 12.0
13 : vin 1 0 0.0 ac 1.0 sin(0 1 1k)
14 : ccouple 1 base 10uf
15 : rbias1 vcc base 100k
16 : rbias2 base 0 24k
17 : q1 coll base emit generic
18 : rcollector vcc coll 3.9k
19 : remitter emit 0 1k
21 : .model generic npn
24 : .end
The title of this circuit is ‘A Berkeley SPICE3 compatible circuit’. The circuit descrip-
tion contains a transient analysis control command .TRAN 1E-5 2E-3 requesting a total
simulated time of 2ms with a maximum time-step of 10us. The remainder of the lines in
the circuit description describe the circuit of Fig. 21.1.
Now, execute the simulation by entering the run command:
ngspice 1 -> run
The simulator will run the simulation and when execution is completed, will return with
the ngspice prompt. When the prompt returns, issue the rusage command again to see
how much time and memory has been used now.
To examine the results of this transient analysis, we can use the plot command. First we
will plot the nodes labeled ‘1’ and ‘base’.
ngspice 2 -> plot v(1) base
The simulator responds by displaying an X Window System plot similar to that shown in
Fig. 21.2.
Figure 21.2: node 1 and node ’base’ versus time
Notice that we have named one of the nodes in the circuit description with a number (‘1’),
while the others are words (‘base’). This was done to illustrate ngspice’s special require-
ments for plotting nodes labeled with numbers. Numeric labels are allowed in ngspice for
backwards compatibility with SPICE2. However, they require special treatment in some
commands such as plot. The plot command is designed to allow expressions in its argu-
ment list in addition to names of results data to be plotted. For example, the expression
plot (base - 1) would plot the result of subtracting 1 from the value of node ‘base’.
21.1. AC COUPLED TRANSISTOR AMPLIFIER 515
If we had desired to plot the difference between the voltage at node ‘base’ and node ‘1’, we
would need to enclose the node name ‘1’ in the construction v( ) producing a command
such as plot (base - v(1)).
Now, issue the following command to examine the voltages on two of the internal nodes
of the transistor amplifier circuit:
ngspice 3 -> plot vcc coll emit
The plot shown in Fig. 21.3 should appear. Notice in the circuit description that the
power supply voltage source and the node it is connected to both have the name ‘vcc’. The
plot command above has plotted the node voltage ‘vcc’. However, it is also possible to
plot branch currents through voltage sources in a circuit. ngspice always adds the special
suffix #branch to voltage source names. Hence, to plot the current into the voltage source
named vcc, we would use a command such as plot vcc#branch.
Figure 21.3: VCC, Collector and Emitter Voltages
Now let’s run a simple DC simulation of this circuit and examine the bias voltages with
the print command. One way to do this is to quit the simulator using the quit command,
edit the input file to change the .tran line to .op (for ’operating point analysis’), re-invoke
the simulator, and then issue the run command. However, ngspice allows analysis mode
changes directly from the ngspice prompt. All that is required is to enter the control line,
e.g. op (without the leading ‘.’). ngspice will interpret the information on the line and
start the new analysis run immediately, without the need to enter a new run command.
To run the DC simulation of the transistor amplifier, issue the following command:
ngspice 4 -> op
After a moment the ngspice prompt returns. Now issue the print command to examine
the emitter, base, and collector DC bias voltages.
ngspice 5 -> print emit base coll
ngspice responds with:
emit = 1.293993e+00 base = 2.074610e+00 coll = 7.003393e+00
To run an AC analysis, enter the following command:
ngspice 6 -> ac dec 10 0.01 100
This command runs a small-signal swept AC analysis of the circuit to compute the magni-
tude and phase responses. In this example, the sweep is logarithmic with ‘decade’ scaling,
10 points per decade, and lower and upper frequencies of 0.01 Hz and 100 Hz. Since the
command sweeps through a range of frequencies, the results are vectors of values and are
examined with the plot command. Issue to the following command to plot the response
curve at node ‘coll’:
ngspice 7 -> plot coll
This plot shows the AC gain from input to the collector. (Note that our input source in
the circuit description ‘vin’ contained parameters of the form ‘AC 1.0’ designating that
a unit-amplitude AC signal was applied at this point.) For plotting data from an AC
analysis, ngspice chooses automatically a logarithmic scaling for the frequency (x) axis.
To produce a more traditional ‘Bode’ gain phase plot (again with automatic logarithmic
scaling on the frequency axis), we use the expression capability of the plot command and
the built-in ngspice functions db( ) and ph( ):
ngspice 8 -> plot db(coll) ph(coll)
The last analysis supported by ngspice is a swept DC analysis. To perform this analysis,
issue the following command:
ngspice 9 -> dc vcc 0 15 0.1
This command sweeps the supply voltage ‘vcc’ from 0 to 15 volts in 0.1 volt increments.
To plot the results, issue the command:
ngspice 10 -> plot emit base coll
Finally, to exit the simulator, use the quit command, and you will be returned to the
operating system prompt.
ngspice 11 -> quit
So long.
21.2. DIFFERENTIAL PAIR 517
21.2 Differential Pair

The following deck determines the dc operating point of a simple differential pair. In addi-
tion, the ac small-signal response is computed over the frequency range 1Hz to 100MEGHz.
Example:
SIMPLE DIFFERENTIAL PAIR

VCC 7 0 12
VEE 8 0 -12
VIN 1 0 AC 1
RS1 1 2 1 K
RS2 6 0 1 K
Q1 3 2 4 MOD1
Q2 5 6 4 MOD1
RC1 7 3 10 K
RC2 7 5 10 K
RE 4 8 10 K
. MODEL MOD1 NPN BF =50 VAF =50 IS =1. E -12 RB =100 CJC =.5 PF TF =.6 NS
. TF V (5) VIN
. AC DEC 10 1 100 MEG
. END
21.3 MOSFET Characterization

The following deck computes the output characteristics of a MOSFET device over the
range 0-10V for VDS and 0-5V for VGS.
Example:
MOS OUTPUT CHARACTERISTICS

. OPTIONS NODE NOPAGE
VDS 3 0
VGS 2 0
M1 1 2 0 0 MOD1 L =4 U W =6 U AD =10 P AS =10 P
* VIDS MEASURES ID , WE COULD HAVE USED VDS ,
* BUT ID WOULD BE NEGATIVE
VIDS 3 1
. MODEL MOD1 NMOS VTO = -2 NSUB =1.0 E15 UO =550
. DC VDS 0 10 .5 VGS 0 5 1
. END
21.4 RTL Inverter

The following deck determines the dc transfer curve and the transient pulse response of
a simple RTL inverter. The input is a pulse from 0 to 5 Volts with delay, rise, and fall
times of 2ns and a pulse width of 30ns. The transient interval is 0 to 100ns, with printing
to be done every nanosecond.
Example:
SIMPLE RTL INVERTER

VCC 4 0 5
VIN 1 0 PULSE 0 5 2 NS 2 NS 2 NS 30 NS
RB 1 2 10 K
Q1 3 2 0 Q1
RC 3 4 1 K
. MODEL Q1 NPN BF 20 RB 100 TF .1 NS CJC 2 PF
. DC VIN 0 5 0.1
. TRAN 1 NS 100 NS
. END
21.5 Four-Bit Binary Adder (Bipolar)
The following deck simulates a four-bit binary adder, using several subcircuits to describe
various pieces of the overall circuit.
Example:
ADDER - 4 BIT ALL - NAND - GATE BINARY ADDER

*** SUBCIRCUIT DEFINITIONS
. SUBCKT NAND 1 2 3 4
* NODES : INPUT (2) , OUTPUT , VCC
Q1 9 5 1 QMOD
D1CLAMP 0 1 DMOD
Q2 9 5 2 QMOD
D2CLAMP 0 2 DMOD
RB 4 5 4 K
R1 4 6 1.6 K
Q3 6 9 8 QMOD
R2 8 0 1 K
RC 4 7 130
Q4 7 6 10 QMOD
DVBEDROP 10 3 DMOD
Q5 3 8 0 QMOD
. ENDS NAND
21.5. FOUR-BIT BINARY ADDER (BIPOLAR) 519
Continue 4 Bit adder:
. SUBCKT ONEBIT 1 2 3 4 5 6
* NODES : INPUT (2) , CARRY - IN , OUTPUT , CARRY - OUT , VCC
X1 1 2 7 6 NAND
X2 1 7 8 6 NAND
X3 2 7 9 6 NAND
X4 8 9 10 6 NAND
X5 3 10 11 6 NAND
X6 3 11 12 6 NAND
X7 10 11 13 6 NAND
X8 12 13 4 6 NAND
X9 11 7 5 6 NAND
. ENDS ONEBIT
. SUBCKT TWOBIT 1 2 3 4 5 6 7 8 9
* NODES : INPUT - BIT0 (2) / BIT1 (2) , OUTPUT - BIT0 / BIT1 ,
* CARRY - IN , CARRY - OUT , VCC
X1 1 2 7 5 10 9 ONEBIT
X2 3 4 10 6 8 9 ONEBIT
. ENDS TWOBIT
. SUBCKT FOURBIT 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
* NODES : INPUT - BIT0 (2) / BIT1 (2) / BIT2 (2) / BIT3 (2) ,
* OUTPUT - BIT0 / BIT1 / BIT2 / BIT3 , CARRY - IN , CARRY - OUT , VCC
X1 1 2 3 4 9 10 13 16 15 TWOBIT
X2 5 6 7 8 11 12 16 14 15 TWOBIT
. ENDS FOURBIT
*** DEFINE NOMINAL CIRCUIT

. MODEL DMOD D
. MODEL QMOD NPN ( BF =75 RB =100 CJE =1 PF CJC =3 PF )
VCC 99 0 DC 5 V
VIN1A 1 0 PULSE (0 3 0 10 NS 10 NS 10 NS 50 NS )
VIN1B 2 0 PULSE (0 3 0 10 NS 10 NS 20 NS 100 NS )
X1 1 2 3 4 5 6 7 8 9 10 11 12 0 13 99 FOURBIT
RBIT0 9 0 1 K
RBIT1 10 0 1 K
RBIT2 11 0 1 K
RBIT3 12 0 1 K
RCOUT 13 0 1 K
*** ( FOR THOSE WITH MONEY ( AND MEMORY ) TO BURN )

. TRAN 1 NS 6400 NS
. END
21.6 Four-Bit Binary Adder (MOS)
The following deck simulates a four-bit binary adder, using several subcircuits to describe
various pieces of the overall circuit.
Example:
ADDER - 4 BIT ALL - NAND - GATE BINARY ADDER

*** SUBCIRCUIT DEFINITIONS
. SUBCKT NAND in1 in2 out VDD
* NODES : INPUT (2) , OUTPUT , VCC
M1 out in2 Vdd Vdd p1 W =7.5 u L =0.35 u pd =13.5 u ad =22.5 p
+ ps =13.5 u as =22.5 p
M2 net .1 in2 0 0 n1 W =3 u L =0.35 u pd =9 u ad =9 p
+ ps =9 u as =9 p
M3 out in1 Vdd Vdd p1 W =7.5 u L =0.35 u pd =13.5 u ad =22.5 p
+ ps =13.5 u as =22.5 p
M4 out in1 net .1 0 n1 W =3 u L =0.35 u pd =9 u ad =9 p
+ ps =9 u as =9 p
. ENDS NAND
. SUBCKT ONEBIT 1 2 3 4 5 6 AND
X2 1 7 8 6 NAND
X3 2 7 9 6 NAND
X4 8 9 10 6 NAND
X5 3 10 11 6 NAND
X6 3 11 12 6 NAND
X7 10 11 13 6 NAND
X8 12 13 4 6 NAND
X9 11 7 5 6 NAND
. ENDS ONEBIT
. SUBCKT TWOBIT 1 2 3 4 5 6 7 8 9
* NODES : INPUT - BIT0 (2) / BIT1 (2) , OUTPUT - BIT0 / BIT1 ,
* CARRY - IN , CARRY - OUT , VCC
X1 1 2 7 5 10 9 ONEBIT
X2 3 4 10 6 8 9 ONEBIT
. ENDS TWOBIT
. SUBCKT FOURBIT 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
* NODES : INPUT - BIT0 (2) / BIT1 (2) / BIT2 (2) / BIT3 (2) ,
* OUTPUT - BIT0 / BIT1 / BIT2 / BIT3 , CARRY - IN ,
* CARRY - OUT , VCC
X1 1 2 3 4 9 10 13 16 15 TWOBIT
X2 5 6 7 8 11 12 16 14 15 TWOBIT
. ENDS FOURBIT
21.7. TRANSMISSION-LINE INVERTER 521
Continue 4 Bit adder MOS:
*** POWER
VCC 99 0 DC 3.3 V
*** INPUTS
VIN1A 1 0 DC 0 PULSE (0 3 0 5 NS 5 NS 20 NS 50 NS )
VIN1B 2 0 DC 0 PULSE (0 3 0 5 NS 5 NS 30 NS 100 NS )
*** DEFINE NOMINAL CIRCUIT
X1 1 2 3 4 5 6 7 8 9 10 11 12
0 13 99 FOURBIT
. option acct
. save V (1) V (2) V (3) V (4) V (5) V (6) V (7) V (8) $ INPUTS
. save V (9) V (10) V (11) V (12) V (13) $ OUTPUTS
. TRAN 1 NS 6400 NS
* use BSIM3 model with default parameters

. model n1 nmos level =49 version =3.3.0
. model p1 pmos level =49 version =3.3.0
. END
21.7 Transmission-Line Inverter
The following deck simulates a transmission-line inverter. Two transmission-line elements

are required since two propagation modes are excited. In the case of a coaxial line, the
first line (T1) models the inner conductor with respect to the shield, and the second line
(T2) models the shield with respect to the outside world.
Example:
Transmission - line inverter
v1 1 0 pulse (0 1 0 0.1 n )
r1 1 2 50
x1 2 0 0 4 tline
r2 4 0 50
. subckt tline 1 2 3 4
t1 1 2 3 4 z0 =50 td =1.5 ns
t2 2 0 4 0 z0 =100 td =1 ns
. ends tline
. tran 0.1 ns 20 ns
. end
Chapter 22
Statistical circuit analysis
22.1 Introduction
Real circuits do not operate in a world with fixed values of device parameters, power
supplies and environmental data. Even if a ngspice output offers 5 digits or more of preci-
sion, this should not mislead you thinking that your circuits will behave exactly the same.
All physical parameters influencing a circuit (e.g. MOS Source/drain resistance, thresh-
old voltage, transconductance) are distributed parameters, often following a Gaussian
distribution with a mean value µand a standard deviation σ.
To obtain circuits operating reliably under varying parameters, it might be necessary
to simulate them taking certain parameter spreads into account. ngspice offers several
methods supporting this task. A powerful random number generator is working in the
background. It is not providing true random numbers, but a long sequence of pseudo
random numbers. This sequence depends on a seed value. The same seed value will
deliver the same sequence of random numbers.
ngspice offers several methods to set this seed value. If no input is given, then ngspice
sets the seed (stored in variable rndseed) to 1 upon start up. With the option SEED
you may either set a value to rndseed upon start up of ngspice (option SEED=nn, nn is
an integer greater than 0), or obtain a “random” number as seed, that is the number of
seconds since 01.01.1970 (option SEED=random). This command is best set in .spiceinit
(16.6). With the command setseed (see chapt.17.5.78) you may choose any other seed
value (integer greater than 0).
The following three chapters offer a short introduction to the statistical methods available
in ngspice. The diversity of approaches stems from historical reasons, and from some
efforts to make ngspice compatible to other simulators.
22.2 Using random param(eters)
The ngspice frontend (with its ’numparam’ parser) contains the .param (see Chapt. 2.9.1)
and .func (see Chapt. 2.10) commands. Among the built-in functions supported (see
2.9.5) you will find the following statistical functions:
523
524 CHAPTER 22. STATISTICAL CIRCUIT ANALYSIS
Built-in function Notes

gauss(nom, rvar, sigma) nominal value plus variation drawn from Gaussian
distribution with mean 0 and standard deviation rvar
(relative to nominal), divided by sigma
agauss(nom, avar, sigma) nominal value plus variation drawn from Gaussian
distribution with mean 0 and standard deviation avar
(absolute), divided by sigma
unif(nom, rvar) nominal value plus relative variation (to nominal)
uniformly distributed between +/-rvar
aunif(nom, avar) nominal value plus absolute variation uniformly
distributed between +/-avar
limit(nom, avar) nominal value +/-avar, depending on random number
in [-1, 1] being > 0 or < 0
The frontend parser evaluates all .param or .func statements upon start-up of ngspice, be-
fore the circuit is evaluated. The parameters aga, aga2, lim obtain their numerical values
once. If the random function appears in a device card (e.g. v11 11 0 ’agauss(1,2,3)’),
a new random number is generated.
Random number example using parameters:
* random number tests

. param aga = agauss (1 ,2 ,3)
. param aga2 = ’2* aga ’
. param lim = limit (0 ,1.2)
. func rgauss (a ,b , c ) ’5* agauss (a ,b , c ) ’
* always same value as defined above
v1 1 0 ’lim ’
v2 2 0 ’lim ’
* may be a different value
v3 3 0 ’ limit (0 ,1.2) ’
* always new random values
v11 11 0 ’ agauss (1 ,2 ,3) ’
v12 12 0 ’ agauss (1 ,2 ,3) ’
v13 13 0 ’ agauss (1 ,2 ,3) ’
* same value as defined above
v14 14 0 ’aga ’
v15 15 0 ’aga ’
v16 16 0 ’ aga2 ’
* using . func , new random values
v17 17 0 ’ rgauss (0 ,2 ,3) ’
v18 18 0 ’ rgauss (0 ,2 ,3) ’
. op
. control
run
print v (1) v (2) v (3) v (11) v (12) v (13)
print v (14) v (15) v (16) v (17) v (18)
. endc
. end
22.3. BEHAVIORAL SOURCES (B, E, G, R, L, C) WITH RANDOM CONTROL 525
So v1, v2, and v3 will get the same value, whereas v4 might differ. v11, v12, and v13
will get different values, v14, v15, and v16 will obtain the values set above in the .param
statements. .func will start its replacement algorithm, rgauss(a,b,c) will be replaced
everywhere by 5*agauss(a,b,c).
Thus device and model parameters may obtain statistically distributed starting values.
You simply set a model parameter not to a fixed numerical value, but insert a ’parameter’
instead, which may consist of a token defined in a .param card, by calling .func or
by using a built-in function, including the statistical functions described above. The
parameter values will be evaluated once immediately after reading the input file.
22.3 Behavioral sources (B, E, G, R, L, C) with ran-

dom control
All sources listed in the section header may contain parameters, which will be evaluated
before simulation starts, as described in the previous section (22.2). In addition the
nonlinear voltage or current sources (B-source, Chapt. 5) as well as their derivatives E
and G, but also the behavioral R, L, and C may be controlled during simulation by a
random independent voltage source V with TRRANDOM option (Chapt. 4.1.8).
An example circuit, a Wien bridge oscillator from input file /examples/Monte_Carlo/OpWien.sp

is distributed with ngspice or available at Git. The two frequency determining pairs of R
and C are varied statistically using four independent Gaussian voltage sources as the con-
trolling units. An excerpt of this command sequence is shown below. The total simulation
time ttime is divided into 100 equally spaced blocks. Each block will get a new set of
control voltages, e.g. VR2, which is Gaussian distributed, mean 0 and absolute deviation
1. The resistor value is calculated with ±10% spread, the factor 0.033 will set this 10%
to be a deviation of 1 sigma from nominal value.
Examples for control of a behavioral resistor:
* random resistor
. param res = 10 k
. param ttime =12000 m
. param varia =100
. param ttime10 = ’ ttime / varia ’
* random control voltage ( Gaussian distribution )
VR2 r2 0 dc 0 trrandom (2 ’ ttime10 ’ 0 1)
* behavioral resistor
R2 4 6 R = ’ res + 0.033 * res * V ( r2 ) ’
So within a single simulation run you will obtain 100 different frequency values issued by
the Wien bridge oscillator. The voltage sequence VR2 is shown below.
22.4 ngspice scripting language
The ngspice scripting language is described in detail in Chapt. 17.8. All commands listed
in Chapt. 17.5 are available, as well as the built-in functions described in Chapt. 17.2, the
control structures listed in Chapt. 17.6, and the predefined variables from Chapt. 17.7.
Variables and functions are typically evaluated after a simulation run. You may created
loops with several simulation runs and change device and model parameters with the
alter (17.5.3) or altermod (17.5.4) commands, as shown in the next section 22.5. You
may even interrupt a simulation run by proper usage of the stop (17.5.91) and resume
(17.5.67) commands. After stop you may change device or model parameters and then
go on with resume, continuing the simulation with the new parameter values.
The statistical functions provided for scripting are listed in the following table:
22.5. MONTE-CARLO SIMULATION 527
Name Function
rnd(vector) A vector with each component a random integer
between 0 and the absolute value of the input vector’s
corresponding integer element value.
sgauss(vector) Returns a vector of random numbers drawn from a
Gaussian distribution (real value, mean = 0 , standard
deviation = 1). The length of the vector returned is
determined by the input vector. The contents of the
input vector will not be used. A call to sgauss(0) will
return a single value of a random number as a vector of
length 1..
sunif(vector) Returns a vector of random real numbers uniformly
distributed in the interval [-1 .. 1]. The length of the
vector returned is determined by the input vector. The
contents of the input vector will not be used. A call to
sunif(0) will return a single value of a random number
as a vector of length 1.
poisson(vector) Returns a vector with its elements being integers
drawn from a Poisson distribution. The elements of the
input vector (real numbers) are the expected numbers
λ. Complex vectors are allowed, real and imaginary
values are treated separately.
exponential(vector) Returns a vector with its elements (real numbers)
drawn from an exponential distribution. The elements
of the input vector are the respective mean values (real
numbers). Complex vectors are allowed, real and
imaginary values are treated separately.
22.5 Monte-Carlo Simulation
The ngspice scripting language may be used to run Monte-Carlo simulations with statis-
tically varying device or model parameters. Calls to the functions sgauss(0) or sunif(0)
(see 17.2) will return Gaussian or uniform distributed random numbers (real numbers),
stored in a vector. You may define (see 17.5.19) your own function using sgauss or sunif,
e.g. to change the mean or range. In a loop (see 17.6) then you may call the alter (17.5.3)
or altermod (17.5.4) statements with random parameters followed by an analysis like op,
dc, ac, tran or other.
22.5.1 Example 1
The first examples is a LC band pass filter, where L and C device parameters will be
changed 100 times. Each change is followed by an ac analysis. All graphs of output
voltage versus frequency are plotted. The file is available in the distribution as /exam-
ples/Monte_Carlo/MonteCarlo.sp as well as from the git repository .
Monte-Carlo example 1
Perform Monte Carlo simulation in ngspice

V1 N001 0 AC 1 DC 0
R1 N002 N001 141
*
C1 OUT 0 1e -09
L1 OUT 0 10 e -06
C2 N002 0 1e -09
L2 N002 0 10 e -06
L3 N003 N002 40 e -06
C3 OUT N003 250 e -12
*
R2 0 OUT 141
*
. control
let mc_runs = 100
let run = 1
set curplot = new $ create a new plot
set scratch = $curplot $ store its name to ’ scratch ’
*
define unif ( nom , var ) ( nom + nom * var * sunif (0))
define aunif ( nom , avar ) ( nom + avar * sunif (0))
define gauss ( nom , var , sig ) ( nom + nom * var / sig * sgauss (0))
*
dowhile run <= mc_runs
* alter c1 = unif (1 e -09 , 0.1)
* alter l1 = aunif (10 e -06 , 2e -06)
* alter c2 = aunif (1 e -09 , 100 e -12)
* alter l2 = unif (10 e -06 , 0.2)
* alter l3 = aunif (40 e -06 , 8e -06)
* alter c3 = unif (250 e -12 , 0.15)
alter c1 = gauss (1 e -09 , 0.1 , 3)
alter l1 = agauss (10 e -06 , 2e -06 , 3)
alter c2 = agauss (1 e -09 , 100 e -12 , 3)
alter l2 = gauss (10 e -06 , 0.2 , 3)
alter l3 = agauss (40 e -06 , 8e -06 , 3)
alter c3 = gauss (250 e -12 , 0.15 , 3)
ac oct 100 250 K 10 Meg
set run =" $ & run " $ create a variable from the vector
set dt = $curplot $ store the current plot to dt
setplot $scratch $ make ’ scratch ’ the active plot
* store the output vector to plot ’ scratch ’
let vout { $run }={ $dt }. v ( out )
setplot $dt $ go back to the previous plot
let run = run + 1
end
plot db ({ $scratch }. all )
. endc
. end
22.5. MONTE-CARLO SIMULATION 529
22.5.2 Example 2
A more sophisticated input file for Monte Carlo simulation is distributed with the file /ex-
amples/Monte_Carlo/MC_ring.sp (or git repository). Due to its length it is not reproduced
here, but some comments on its enhancements over example 1 (22.5.1) are presented in
the following.
A 25-stage ring oscillator is the circuit used with a transient simulation. It comprises of
CMOS inverters, modeled with BSIM3. Several model parameters (vth, u0, tox, L, and
W) shall be varied statistically between each simulation run. The frequency of oscillation
will be measured by a fft and stored. Finally a histogram of all measured frequencies
will be plotted.
The function calls to sunif(0) and sgauss(0) return uniformly or Gaussian distributed
random numbers. A function unif, defined by the line
define unif(nom, var) (nom + (nom*var) * sunif(0))
will return a value with mean nom and deviation var relative to nom.
The line
set n1vth0=@n1[vth0]
will store the threshold voltage vth0, given by the model parameter set, into a variable
n1vth0, ready to be used by unif, aunif, gauss, or agauss function calls.
In the simulation loop the altermod command changes the model parameters before a call
to tran. After the transient simulation the resulting vector is linearized, a fft is calculated,
and the maximum of the fft signal is measured by the meas command and stored in a
vector maxffts. Finally the contents of the vector maxffts is plotted in a histogram.
For more details, please have a look at the strongly commented input file MC_ring.sp.
22.5.3 Example 3
The next example is contained in the files MC_2_control.sp and MC_2_circ.sp from folder
/examples/Monte_Carlo/. MC_2_control.sp is a ngspice script (see 17.8). It starts a loop
by setting the random number generator seed value to the value of the loop counter,
sources the circuit file MC_2_circ.sp, runs the simulation, stores a raw file, makes an
fft, saves the oscillator frequency thus measured, deletes all outputs, increases the loop
counter and restarts the loop. The netlist file MC_2_circ.sp contains the circuit, which
is the same ring oscillator as of example 2. However, now the MOS model parameter
set, which is included with this netlist file, inherits some AGAUSS functions (see 2.9.5)
to vary threshold voltage, mobility and gate oxide thickness of the NMOS and PMOS
transistors. This is an approach similar to what commercial foundries deliver within
their device libraries. So this example may be your source for running Monte Carlo with
commercial libs. Start example 3 by calling
ngspice -o MC_2_control.log MC_2_control.sp
22.6 Data evaluation with Gnuplot
Run the example file /examples/Monte_Carlo/OpWien.sp, described in Chapt. 22.3. Gen-

erate a plot with Gnuplot by the ngspice command
gnuplot pl4mag v4mag xlimit 500 1500
Open and run the command file in the Gnuplot command line window by
load ’pl-v4mag.p’
A Gaussian curve will be fitted to the simulation data. The mean oscillator frequency
and its deviation are printed in the curve fitting log in the Gnuplot window.
Gnuplot script for data evaluation:
# This file : pl - v4mag . p

# ngspice file OpWien . sp
# ngspice command :
# gnuplot pl4mag v4mag xlimit 500 1500
# a gnuplot manual :
# http :// www . duke . edu /~ hpgavin / gnuplot . html
# Gauss function to be fitted

f1 ( x )=( c1 /( a1 * sqrt (2*3.14159))* exp ( -(( x - b1 )**2)/(2* a1 **2)))
# Gauss function to plot start graph
f2 ( x )=( c2 /( a2 * sqrt (2*3.14159))* exp ( -(( x - b2 )**2)/(2* a2 **2)))
# start values
a1 =50 ; b1 =900 ; c1 =50
# keep start values in a2 , b2 , c2
a2 = a1 b2 = b1 ; c2 = c1
# curve fitting
fit f1 ( x ) ’ pl4mag . data ’ using 1:2 via a1 , b1 , c1
# plot original and fitted curves with new a1 , b1 , c1
plot " pl4mag . data " using 1:2 with lines , f1 ( x ) , f2 ( x )
22.6. DATA EVALUATION WITH GNUPLOT 531
pl4mag.data is the simulation data, f2(x) the starting curve, f1(x) the fitted Gaussian
distribution.
This is just a simple example. You might explore the powerful built-in functions of
Gnuplot to do a much more sophisticated statistical data analysis.
Chapter 23
Circuit optimization with ngspice
23.1 Optimization of a circuit

Your circuit design (analog, maybe mixed-signal) has already the best circuit topology.
There might be still some room for parameter selection, e.g. the geometries of transistors
or values of passive elements, to best fit the specific purpose. This is, what (automatic)
circuit optimization will deliver. In addition you may fine-tune, optimize and verify the
circuit over voltage, process or temperature corners. So circuit optimization is a valuable
tool in the hands of an experienced designer. It will relieve you from the routine task of
’endless’ repetitions of re-simulating your design.
You have to choose circuit variables as parameters to be varied during optimization (e.g.
device size, component values, bias inputs etc.). Then you may pose performance con-
straints onto you circuits (e.g. Vnode < 1.2V, gain > 50 etc.). Optimization objectives
are the variables to be minimized or maximized. The n objectives and m constraints are
assembled into a cost function.
The optimization flow is now the following: The circuit is loaded. Several (perhaps only
one) simulations are started with a suitable starter set of variables. Measurements are
done on the simulator output to check for the performance constraints and optimization
objectives. These data are fed into the optimizer to evaluate the cost function. A so-
phisticated algorithm now determines a new set of circuit variables for the next simulator
run(s). Stop conditions have to be defined by the user to tell the simulator when to finish
(e.g. fall below a cost function value, parameter changes fall below a certain threshold,
number of iterations exceeded).
The optimizer algorithms, its parameters and the starting point influence the convergence
behavior. The algorithms have to provide measures to reaching the global optimum, not
to stick to a local one, and thus are tantamount for the quality of the optimizer.
ngspice does not have an integral optimization processor. Thus this chapter will rely on
work done by third parties to introduce ngspice optimization capability. ngspice pro-
vides the simulation engine, a script or program controls the simulator and provides the
optimizer functionality.
Four optimizers are presented here, using ngspice scripting language, using tclspice, using
a Python script, and using ASCO, a c-coded optimization program.
533
534 CHAPTER 23. CIRCUIT OPTIMIZATION WITH NGSPICE
23.2 ngspice optimizer using ngspice scripts

Friedrich Schmidt (see his web site) has intensively used circuit optimization during his
development of Nonlinear loadflow computation with Spice based simulators. He has
provided an optimizer using the internal ngspice scripting language (see Chapt. 17.8).
His original scripts are found here. A slightly modified and concentrated set of his scripts
is available from the ngspice optimizer directory.
The simple example given in the scripts is OK with current ngspice. Real circuits have
still to be tested.
23.3 ngspice optimizer using tclspice

ngspice offers another scripting capability, namely the tcl/tk based tclspice option (see
Chapt. 20). An optimization procedure may be written using a tcl script. An example is
provided in Chapt. 20.5.2.
23.4 ngspice optimizer using a Python script

Werner Hoch has developed a ngspice optimization procedure based on the ’differential
evolution’ algorithm [21]. On his web page he provides a Python script containing the
control flow and algorithms.
23.5 ngspice optimizer using ASCO

The ASCO optimizer, developed by Joao Ramos, also applies the ’differential evolution’
algorithm [21]. An enhanced version 0.4.7.1, adding ngspice as a simulation engine, may
be downloaded here (7z archive format). Included are executable files (asco, asco-mpi,
ngspice-c for MS Windows). The source code should also compile and function under
Linux (not yet tested).
ASCO is a standalone executable, which communicates with ngspice via ngspice input
and output files. Several optimization examples, originally provided by J. Ramos for
other simulators, are prepared for use with ngspice. Parallel processing on a multi-core
computer has been tested using MPI (MPICH2) under MS Windows. A processor network
will be supported as well. A MS Windows console application ngspice_c.exe is included
in the archive. Several stand alone tools are provided, but not tested yet.
Setting up an optimization project with ASCO requires advanced know-how of using
ngspice. There are several sources of information. First of all the examples provided
with the distribution give hints how to start with ASCO. The original ASCO manual is
provided as well, or is available here. It elaborates on the examples, using a commercial
simulator, and provides a detailed description how to set up ASCO. Installation of ASCO
and MPI (under Windows) is described in a file INSTALL.
Some remarks on how to set up ASCO for ngspice are given in the following sections
(more to be added). These a meant not as a complete description, but are an addition
the the ASCO manual.
23.5. NGSPICE OPTIMIZER USING ASCO 535
23.5.1 Three stage operational amplifier

This example is taken from Chapt. 6.2.2 ‘Tutorial #2’ from the ASCO manual. The
directory examples /ngspice/amp3 contains four files:
amp3.cfg This file contains all configuration data for this optimization. Of special
interest is the following section, which sets the required measurements and the constraints
on the measured parameters:
# Measurements #
ac_power:VDD:MIN:0
dc_gain:VOUT:GE:122
unity_gain_frequency:VOUT:GE:3.15E6
phase_margin:VOUT:GE:51.8
phase_margin:VOUT:LE:70
amp3_slew_rate:VOUT:GE:0.777E6
#
Each of these entries is linked to a file in the /extract subdirectory, having exactly
the same names as given here, e.g. ac_power, dc_gain, unity_gain, phase_margin, and
amp3_slew_rate. Each of these files contains an # Info # section, which is currently
not used. The # Commands # section may contain a measurement command (includ-
ing ASCO parameter #SYMBOL#, see file /extract/unity_gain_frequency). It also may
contain a .control section (see file /extract/phase_margin_min). During set-up #SYM-
BOL# is replaced by the file name, a leading ‘z’, and a trailing number according to the
above sequence, starting with 0.
amp3.sp This is the basic circuit description. Entries like #LM2# are ASCO-specific,
defined in the # Parameters # section of file amp3.cfg. ASCO will replace these parameter
placeholders with real values for simulation, determined by the optimization algorithm.
The .control ... .endc section is specific to ngspice. Entries to this section may deliver
workarounds of some commands not available in ngspice, but used in other simulators.
You may also define additional measurements, get access to variables and vectors, or
define some data manipulation. In this example the .control section contains an op
measurement, required later for slew rate calculation, as well as the ac simulation, which
has to occur before any further data evaluation. Data from the op simulation are stored
in a plot op1. Its name is saved in variable dt. The ac measurements sets another plot ac1.
To retrieve op data from the former plot, you have to use the {$dt}.<vector> notation
(see file /extract/amp3_slew_rate).
n.typ, p.typ MOSFET parameter files, to be included by amp3.sp.
Testing the set-up
Copy asco-test.exe and ngspice_c.exe (console executable of ngspice) into the directory,
and run
$ asco-test -ngspice amp3

from the console window. Several files will be created during checking. If you look at
<computer-name>.sp: this is the input file for ngspice_c, generated by ASCO. You will
find the additional .measure commands and .control sections. The quit command will
be added automatically just before the .endc command in its own .control section. asco-
test will display error messages on the console, if the simulation or communication with
ASCO is not ok. The output file <computer-name>.out, generated by ngspice during
each simulation, contains symbols like zac_power0, zdc_gain1, zunity_gain_frequency2,
zphase_margin3, zphase_margin4, and zamp3_slew_rate5. These are used to communicate
the ngspice output data to ASCO. ASCO is searching for something like zdc_gain1
=, and then takes the next token as the input value. Calling phase_margin twice in
amp3.cfg has lead to two measurements in two .control sections with different symbols
(zphase_margin3, zphase_margin4).
A failing test may result in an error message from ASCO. Sometimes, however, ASCO
freezes after some output statements. This may happen if ngspice issues an error message
that cannot be handled by ASCO. Here it may help calling ngspice directly with the input
file generated by ASCO:
$ ngspice_c <computer-name>.sp
Thus you may evaluate the ngspice messages directly.
Running the simulation
Copy (w)asco.exe, (w)asco-mpi.exe and ngspice_c.exe (console executable of ngspice)

into the directory, and run
$ asco -ngspice amp3
or alternatively (if MPICH is installed)
$ mpiexec -n 7 asco-mpi -ngspice amp3
The following graph 23.1 shows the acceleration of the optimization simulation on a multi-
core processor (i7 with 4 real or 8 virtual cores), 500 generations, if -n is varied. Speed
is tripled, a mere 15 min suffices to optimize 21 parameters of the amplifier.
23.5.2 Digital inverter

This example is taken from Chapt. 6.2.1 Tutorial #1 from the ASCO manual. In addition
to the features already mentioned above, it adds Monte-Carlo and corner simulations.
The file inv.cfg contains the following section:
#Optimization Flow#
Alter:yes $ do we want to do corner analysis?
MonteCarlo:yes $ do we want to do MonteCarlo analysis?
AlterMC cost:3.00 $ point at which we want to start ALTER and/or
$ MONTECARLO
ExecuteRF:no $ Execute or no the RF module to add RF parasitics?
SomethingElse:
#
23.5. NGSPICE OPTIMIZER USING ASCO 537
Figure 23.1: Optimization speed
Monte Carlo is switched on. It uses the AGAUSS function (see Chapt. 22.2). Its parame-
ters are generated by ASCO from the data supplied by the inv.cfg section #Monte Carlo#.
According to the paper by Pelgrom on MOS transistor matching [22] the AGAUSS pa-
rameters are calculated as

ABeta W −6
W = AGAU SS W, √ · · 10 , 1 (23.1)
2 · W · L · m 100

AV T −9
delvto = AGAU SS 0, √ · 10 , 1 (23.2)
2·W ·L·m
The .ALTER command is not available in ngspice. However, a new option in ngspice to
the altermod command (17.5.4) enables the simulation of design corners. The #Alter#
section in inv.cfg gives details. Specific to ngspice, again several .control section are
used.
# ALTER #
.control
* gate oxide thickness varied
altermod nm pm file [b3.min b3.typ b3.max]
.endc
.control
* power supply variation
alter vdd=[2.0 2.1 2.2]
.endc
.control
run
.endc
#
NMOS (nm) and PMOS (pm) model parameter sets are loaded from three different model
files, each containing both NMOS and PMOS sets. b3.typ is assembled from the original
parameter files n.typ and p.typ, provided with original ASCO, with some adaptation to
ngspice BSIM3. The min and max sets are artificially created in that only the gate oxide
thickness deviates ±1 nm from what is found in model file b3.typ. In addition the power
supply voltage is varied, so in total you will find 3 x 3 simulation combinations in the
input file <computer-name>.sp (after running asco-test).
23.5.3 Bandpass
This example is taken from Chapt. 6.2.4 Tutorial #4 from the ASCO manual. S11 in
the passband is to be maximised. S21 is used to extract side lobe parameters. The .net
command is not available in ngspice, so S11 and S21 are derived with a script in file
bandpass.sp as described in Chapt. 17.9. The measurements requested in bandpass.cfg as
# Measurements #
Left_Side_Lobe:---:LE:-20
Pass_Band_Ripple:---:GE:-1
Right_Side_Lobe:---:LE:-20
S11_In_Band:---:MAX:---
#
are realized as ’measure’ commands inside of control sections (see files in directory ex-
tract). The result of a measure statement is a vector, which may be processed by com-
mands in the following lines. In file extract/S1_In_Band #Symbol# is made available
only after a short calculation (inversion of sign), using the print command. quit has
been added to this entry because it will become the final control section in <computer-
name>.sp. A disadvantage of measure inside of a .control section is that parameters
from .param statements may not be used (as is done in example 23.5.4).
The bandpass example includes the calculation of RF parasitic elements defined in rfmodule.cfg
(see Chapt. 7.5 of the ASCO manual). This calculation is invoked by setting
ExecuteRF:yes $Execute or no the RF module to add RF parasitics?
in bandpass.cfg. The two subcircuits LBOND_sub and CSMD_sub are generated in

<computer-name>.sp to simulate these effects.
23.5.4 Class-E power amplifier

This example is taken from Chapt. 6.2.3 Tutorial #3 from the ASCO manual. In this
example the ASCO post processing is applied in file extract/P_OUT (see Chapt. 7.4 of
the ASCO manual.). In this example .measure statements are used. They allow using
parameters from .param statements, because they will be located outside of .control
sections, but do not allow doing data post processing inside of ngspice. You may use
ASCO post processing instead.
Chapter 24
Notes
24.1 Glossary
card A logical SPICE input line. A card may be extended through the use of the ‘+’ sign
in SPICE, thereby allowing it to take up multiple lines in a SPICE deck.
code model A model of a device, function, component, etc. which is based solely on a
C programming language-based function. In addition to the code models included
with the XSPICE option of the ngspice simulator, you can use code models that
you develop for circuit modeling.
deck A collection of SPICE cards that together specify all input information required in
order to perform an analysis. A ‘deck’ of ‘cards’ will in fact be contained within a
file on the host computer system.
element card A single, logical line in an ngspice circuit deck that describes a circuit
element. Circuit elements are connected to each other to form circuits (e.g., a
logical card that describes a resistor, such as R1 2 0 10K, is an element card).
instance A unique occurrence of a circuit element. See ‘element card’, in which the
instance R1 is specified as a unique element (instance) in a hypothetical circuit
description.
macro A macro, in the context of this document, refers to a C language macro that
supports the construction of user-defined models by simplifying input/output and
parameter-passing operations within the Model Definition File.
.mod Refers to the Model Definition File in XSPICE. The file suffix reflects the file-name
of the model definition file: cfunc.mod.
.model Refers to a model card associated with an element card in ngspice. A model card
allows for data defining an instance to be conveniently located in the ngspice deck
such that the general layout of the elements is more readable.
Nutmeg The ngspice post-processor (now obsolete). This provides a simple stand-alone
simulator interface that can be used with the ngspice simulator to display and plot
simulator raw files.
539
540 CHAPTER 24. NOTES
subcircuit A ‘device’ within an ngspice deck that is defined in terms of a group of

element cards and that can be referenced in other parts of the ngspice deck through
element cards.
24.2 Acronyms and Abbreviations

ATE Automatic Test Equipment
CAE Computer-Aided Engineering
CCCS Current Controlled Current Source.
CCVS Current Controlled Voltage Source.
FET Field Effect Transistor
IDD Interface Design Document
IFS Refers to the Interface Specification File. The abbreviation reflects the file name of
the Interface Specification File: ifspec.ifs.
MNA Modified Nodal Analysis
MOSFET Metal Oxide Semiconductor Field Effect Transistor
PWL Piece-Wise Linear
RAM Random Access Memory
ROM Read Only Memory
SDD Software Design Document
SI Simulator Interface
SPICE Simulation Program with Integrated Circuit Emphasis. This program was de-
veloped at the University of California at Berkeley and is the origin of ngspice.
SPICE3 Version 3 of SPICE.
SRS Software Requirements Specification
SUM Software User’s Manual
UCB University of California at Berkeley
UDN User-Defined Node(s)
VCCS Voltage Controlled Current Source.
VCVS Voltage Controlled Voltage Source
XSPICE Extended SPICE; option to ngspice, integrating predefined or user defined code
models for event-driven mixed-signal simulation.
24.3. TO DO 541
24.3 To Do
1. Review of Chapt. 1.3
2. hfet1,2 model descriptions
3. MOS level 9 description

542 CHAPTER 24. NOTES
Bibliography
[1] A. Vladimirescu and S. Liu, ‘The Simulation of MOS Integrated Circuits Using
SPICE2’ ERL Memo No. ERL M80/7, Electronics Research Laboratory University
of California, Berkeley, October 1980
[2] T. Sakurai and A. R. Newton, ‘A Simple MOSFET Model for Circuit Analysis and
its application to CMOS gate delay analysis and series-connected MOSFET Struc-
ture’ ERL Memo No. ERL M90/19, Electronics Research Laboratory, University of
California, Berkeley, March 1990
[3] B. J. Sheu, D. L. Scharfetter, and P. K. Ko, ‘SPICE2 Implementation of BSIM’ ERL

Memo No. ERL M85/42, Electronics Research Laboratory University of California,
Berkeley, May 1985
[4] J. R. Pierret, ‘A MOS Parameter Extraction Program for the BSIM Model’ ERL
Memo Nos. ERL M84/99 and M84/100, Electronics Research Laboratory University
of California, Berkeley, November 1984
[5] Min-Chie Jeng, ‘Design and Modeling of Deep Submicrometer MOSFETSs’ ERL
Memo Nos. ERL M90/90, Electronics Research Laboratory, University of California,
Berkeley, October 1990
[6] Soyeon Park, ‘Analysis and SPICE implementation of High Temperature Effects on
MOSFET ’, Master’s thesis, University of California, Berkeley, December 1986.
[7] Clement Szeto, ‘Simulation of Temperature Effects in MOSFETs (STEIM)’, Master’s

thesis, University of California, Berkeley, May 1988.
[8] J.S. Roychowdhury and D.O. Pederson, ‘Efficient Transient Simulation of Lossy In-
terconnect’, Proc. of the 28th ACM/IEEE Design Automation Conference, June 17-21
1991, San Francisco
[9] A. E. Parker and D. J. Skellern, ‘An Improved FET Model for Computer Simulators’,
IEEE Trans CAD, vol. 9, no. 5, pp. 551-553, May 1990.
[10] R. Saleh and A. Yang, Editors, ‘Simulation and Modeling’, IEEE Circuits and De-
vices, vol. 8, no. 3, pp. 7-8 and 49, May 1992.
[11] H.Statz et al., ‘GaAs FET Device and Circuit Simulation in SPICE’, IEEE Trans-
actions on Electron Devices, V34, Number 2, February 1987, pp160-169.
[12] Weidong Liu et al.: ‘BSIM3v3.2.2 MOSFET Model User’s Manual’, BSIM3v3.2.2
543
544 BIBLIOGRAPHY
[13] Weidong Lui et al.: ‘BSIM3.v3.3.0 MOSFET Model User’s Manual’, BSIM3v3.3.0
[14] ‘SPICE3.C1 Nutmeg Programmer’s Manual’, Department of Electrical Engineering
and Computer Sciences, University of California, Berkeley, California, April, 1987.
[15] Thomas L. Quarles: SPICE3 Version 3C1 User’s Guide, Department of Electrical
Engineering and Computer Sciences, University of California, Berkeley, California,
April, 1989.
[16] Brian Kernighan and Dennis Ritchie: ‘The C Programming Language’, Second Edi-
tion, Prentice-Hall, Englewood Cliffs, New Jersey, 1988.
[17] ‘Code-Level Modeling in XSPICE’, F.L. Cox, W.B. Kuhn, J.P. Murray, and S.D.
Tynor, published in the Proceedings of the 1992 International Symposium on Circuits
and Systems, San Diego, CA, May 1992, vol 2, pp. 871-874.
[18] ‘A Physically Based Compact Model of Partially Depleted SOI MOSFETs for Analog
Circuit Simulation’, Mike S. L. Lee, Bernard M. Tenbroek, William Redman-White,
James Benson, and Michael J. Uren, IEEE JOURNAL OF SOLID-STATE CIR-
CUITS, VOL. 36, NO. 1, JANUARY 2001, pp. 110-121
[19] ‘A Realistic Large-signal MESFET Model for SPICE’, A. E. Parker, and D. J.
Skellern, IEEE Transactions on Microwave Theory and Techniques, vol. 45, no. 9,
Sept. 1997, pp. 1563-1571.
[20] ‘Integrating RTS Noise into Circuit Analysis’, T. B. Tang and A. F. Murray, IEEE
ISCAS, 2009, Proc. of IEEE ISCAS, Taipei, Taiwan, May 2009, pp 585-588
[21] R. Storn, and K. Price: ‘Differential Evolution’, technical report TR-95-012, ICSI,
March 1995, see report download, or the DE web page
[22] M. J. M. Pelgrom e.a.: ‘Matching Properties of MOS Transistors’, IEEE J. Sol. State
Circ, vol. 24, no. 5, Oct. 1989, pp. 1433-1440
[23] Y. V. Pershin, M. Di Ventra: ‘SPICE model of memristive de-
vices with threshold’, arXiv:1204.2600v1 [physics.comp-ph] 12 Apr 2012,
http://arxiv.org/pdf/1204.2600.pdf
[24] George M. Kull e.a. ‘A Unified Circuit Model for Bipolar Transistors Including Quasi-
Saturation Effects’, IEEE TRANSACTIONS ON ELECTRON DEVICES, VOL. ED-
32, NO. 6, JUNE 1985
[25] Matthias Bucher, Christophe Lallement, Christian Enz, Fabien Théodoloz, François
Krummenacher, ‘The EPFL-EKV MOSFET Model Equations for Simulation’, Tech-
nical Report, Revision II, July 1998, Electronics Laboratories, Swiss Federal Institute
of Technology (EPFL), Lausanne, Switzerland
[26] Kenneth Kundert, Sparse Matrix Techniques, in Circuit Analysis, Simulation and
Design, Albert Ruehli (Ed.), North-Holland, 1986
[27] T. A. Davis and Ekanathan Palamadai Natarajan. 2010. Algorithm 907:
KLU, A Direct Sparse Solver for Circuit Simulation Problems. ACM
Trans. Math. Softw. 37, 3, Article 36 (September 2010), 17 pages.
https://dl.acm.org/doi/abs/10.1145/1824801.1824814
BIBLIOGRAPHY 545
[28] F. Lannutti, P. Nenzi and M. Olivieri, "KLU sparse direct linear solver implementa-
tion into NGSPICE," Proceedings of the 19th International Conference Mixed Design
of Integrated Circuits and Systems - MIXDES 2012, Warsaw, Poland, 2012, pp. 69-
73.
546 BIBLIOGRAPHY
Part II
XSPICE Software User’s Manual
547
Chapter 25
XSPICE Basics
25.1 ngspice with the XSPICE option
The XSPICE option allows you to add event-driven simulation capabilities to ngspice.
ngspice now is the main software program that performs mathematical simulation of a
circuit specified by you, the user. It takes input in the form of commands and circuit
descriptions and produces output data (e.g. voltages, currents, digital states, and wave-
forms) that describe the circuit’s behavior.
Plain ngspice is designed for analog simulation and is based exclusively on matrix solution
techniques. The XSPICE option adds even-driven simulation capabilities. Thus, designs
that contain significant portions of digital circuitry can be efficiently simulated together
with analog components. ngspice with XSPICE option also includes a ‘User-Defined Node’
capability that allows event-driven simulations to be carried out with any type of data.
The XSPICE option has been developed by the Computer Science and Information Tech-
nology Laboratory at Georgia Tech Research Institute of the Georgia Institute of Tech-
nology, Atlanta, Georgia 30332 at around 1990 and enhanced by the ngspice team. The
manual is based on the original XSPICE user’s manual, no longer available from Georgia
Tech, but from the ngspice web site.
In the following, the term ‘XSPICE’ may be read as ‘ngspice with XSPICE code model
subsystem enabled’. You may enable the option by adding --enable-xspice to the
./configure command. The MS Windows distribution already contains the XSPICE
option.
25.2 The XSPICE Code Model Subsystem
The new component of ngspice, the Code Model Subsystem, provides the tools needed to
model the various parts of your system. While ngspice is targeted primarily at integrated
circuit (IC) analysis, XSPICE includes features to model and simulate board-level and
system-level designs as well. The Code Model Subsystem is central to this new capability,
providing XSPICE with an extensive set of models to use in designs and allowing you to
add your own models to this model set.
549
550 CHAPTER 25. XSPICE BASICS
The ngspice simulator at the core of XSPICE includes built-in models for discrete com-
ponents commonly found within integrated circuits. These ‘model primitives’ include
components such as resistors, capacitors, diodes, and transistors. The XSPICE Code
Model Subsystem extends this set of primitives in two ways. First, it provides a library
of over 40 additional primitives, including summers, integrators, digital gates, controlled
oscillators, s-domain transfer functions, and digital state machines. See Chapt. 12 for a
description of the library entries. Second, it adds a code model generator to ngspice that
provides a set of programming utilities to make it easy for you to create your own models
by writing them in the C programming language.
The code models are generated upon compiling ngspice. They are stored in shared li-
braries, which may be distributed independently from the ngspice sources. During run-
time initialization, ngspice will load the code model libraries and make the code model
instances available for simulation.
25.3 XSPICE Top-Level Diagram

A top-level diagram of the XSPICE system integrated into ngspice is shown in Fig. 25.1.
The XSPICE Simulator is made up of the ngspice core, the event-driven algorithm, circuit
description syntax parser extensions, a loading routine for code models, and the ngspice
control language user interface. The XSPICE Code Model Subsystem consists of the Code
Model Generator, 5 standard code model library sources with more than 40 code models,
the sources for Node Type Libraries, and all the interfaces to User-Defined Code Models
and to User-Defined Node Types.
25.3. XSPICE TOP-LEVEL DIAGRAM 551
Figure 25.1: ngspice/XSPICE Top-Level Diagram

552 CHAPTER 25. XSPICE BASICS
Chapter 26
Execution Procedures
This chapter covers operation of the XSPICE simulator and the Code Model Subsystem.
It begins with background material on simulation and modeling and then discusses the
analysis modes supported in XSPICE and the circuit description syntax used for modeling.
Detailed descriptions of the predefined Code Models and Node Types provided in the
XSPICE libraries are also included.
26.1 Simulation and Modeling Overview

This section introduces the concepts of circuit simulation and modeling. It is intended
primarily for users who have little or no previous experience with circuit simulators, and
also for those who have not used circuit simulators recently. However, experienced SPICE
users may wish to scan the material presented here since it provides background for new
Code Model and User-Defined Node capabilities of the XSPICE option.
26.1.1 Describing the Circuit
This section provides an overview of the circuit description syntax expected by the
XSPICE simulator. A general understanding of circuit description syntax will be help-
ful to you should you encounter problems with your circuit and need to examine the
simulator’s error messages, or should you wish to develop your own models.
This section will introduce you to the creation of circuit description input files using the
control language user interface. Note that this material is presented in an overview form.
Details of circuit description syntax are given later in this chapter and in the previous
chapters of this manual.
26.1.1.1 Example Circuit Description Input
Although different SPICE-based simulators may include various enhancements to the

basic version from the University of California at Berkeley, most use a similar approach in
describing circuits. This approach involves capturing the information present in a circuit
schematic in the form of a text file that follows a defined format. This format requires the
553
554 CHAPTER 26. EXECUTION PROCEDURES
Figure 26.1: Example Circuit 1
assignment of alphanumeric identifiers to each circuit node, the assignment of component

identifiers to each circuit device, and the definition of the significant parameters for each
device. For example, the circuit description below shows the equivalent input file for the
circuit shown in Fig. 26.1.
Small Signal Amplifier

*
* This circuit simulates a simple small signal amplifier .
*
Vin Input 0
0 SIN (0 .1 500 Hz )
R_source Input Amp_In 100
C1 Amp_In 0 1 uF
R_Amp_Input Amp_In 0 1 MEG
E1 ( Amp_Out 0) ( Amp_In 0) -10
R_Load Amp_Out 0 1000
. Tran 1.0 u 0.01

. end
This file exhibits many of the most important properties common to all SPICE circuit
description files including the following:
• The first line of the file is always interpreted as the title of the circuit. The title
may consist of any text string.
• Lines that provide user comments, but no circuit information, are begun by an
asterisk.
• A circuit device is specified by a device name, followed by the node(s) to which it

is connected, and then by any required parameter information.
• The first character of a device name tells the simulator what kind of device it is
(e.g. R = resistor, C = capacitor, E = voltage controlled voltage source).
26.1. SIMULATION AND MODELING OVERVIEW 555
• Nodes may be labeled with any alphanumeric identifier. The only specific labeling
requirement is that 0 must be used for ground.
• A line that begins with a dot is a ‘control directive’ Control directives are used most
frequently for specifying the type of analysis the simulator is to carry out.
• An .end statement must be included at the end of the file.
• With the exception of the Title and .end statements, the order in which the circuit
file is defined is arbitrary.
• All identifiers are case insensitive - the identifier ‘npn’ is equivalent to ‘NPN’ and
to ‘nPn’.
• Spaces and parenthesis are treated as white space.
• Long lines may be continued on a succeeding line by beginning the next line with a
‘+’ in the first column.
In this example, the title of the circuit is ‘Small Signal Amplifier’. Three comment lines
are included before the actual circuit description begins. The first device in the circuit
is voltage source Vin, which is connected between node Input and ‘0’ (ground). The
parameters after the nodes specify that the source has an initial value of 0, a wave shape
of SIN, and a DC offset, amplitude, and frequency of 0, .1, and 500 respectively. The next
device in the circuit is resistor R_Source, which is connected between nodes Input and
Amp_In, with a value of 100 Ohms. The remaining device lines in the file are interpreted
similarly.
The control directive that begins with .tran specifies that the simulator should carry
out a simulation using the Transient analysis mode. In this example, the parameters to
the transient analysis control directive specify that the maximum time-step allowed is 1
microsecond and that the circuit should be simulated for 0.01 seconds of circuit time.
Other control cards are used for other analysis modes. For example, if a frequency response
plot is desired, perhaps to determine the effect of the capacitor in the circuit, the following
statement will instruct the simulator to perform a frequency analysis from 100 Hz to 10
MHz in decade intervals with ten points per decade.
.ac dec 10 100 10meg
To determine the quiescent operating point of the circuit, the following statement may be
inserted in the file.
.op
A fourth analysis type supported by ngspice is swept DC analysis. An example control

statement for the analysis mode is
.dc Vin -0.1 0.2 .05
This statement specifies a DC sweep that varies the source Vin from -100 millivolts to
+200 millivolts in steps of 50 millivolts.
26.1.1.2 Models and Subcircuits
The file discussed in the previous section illustrated the most basic syntax rules of a
circuit description file. However, ngspice (and other SPICE-based simulators) include
many other features that allow for accurate modeling of semiconductor devices such as
diodes and transistors and for grouping elements of a circuit into a macro or ‘subcircuit’
description that can be reused throughout a circuit description. For instance, the file
shown below is a representation of the schematic shown in Fig. 26.2.
Small Signal Amplifier with Limit Diodes

*
* This circuit simulates a small signal amplifier
* with a diode limiter .
*
. dc Vin -1 1 .05
Vin Input 0 DC 0
D_Neg 0 Amp_In 1 n4148
D_Pos Amp_In 0 1 n4148
C1 Amp_In 0 1 uF
X1 Amp_In 0 Amp_Out Amplifier
. model 1 n4148 D ( is =2.495 e -09 rs =4.755 e -01 n =1.679 e +00

+ tt =3.030 e -09 cjo =1.700 e -12 vj =1 m =1.959 e -01 bv =1.000 e +02
+ ibv =1.000 e -04)
. subckt Amplifier Input Common Output

E1 ( Output Common ) ( Input Common ) -10
R_Input Input Common 1 meg
. ends Amplifier
. end
This is the same basic circuit as in the initial example, with the addition of two components
and some changes to the simulation file. The two diodes have been included to illustrate
the use of device models, and the amplifier is implemented with a subcircuit. Additionally,
this file shows the use of the swept DC control card.
Device Models Device models allow you to specify, when required, many of the
parameters of the devices being simulated. In this example, model statements are used
to define the silicon diodes. Electrically, the diodes serve to limit the voltage at the
amplifier input to values between about ±700 millivolts. The diode is simulated by first
declaring the ‘instance’ of each diode with a device statement. Instead of attempting
to provide parameter information separately for both diodes, the label ‘1n4148’ alerts
the simulator that a separate model statement is included in the file that provides the
26.1. SIMULATION AND MODELING OVERVIEW 557
Figure 26.2: Example Circuit 2
necessary electrical specifications for the device (‘1n4148’ is the part number for the type
of diode the model is meant to simulate).
The model statement that provides this information is:

+ tt =3.030 e -09 cjo =1.700 e -12 vj =1 m =1.959 e -01
+ bv =1.000 e +02 ibv =1.000 e -04)
The model statement always begins with the string .model followed by an identifier and
the model type (D for diode, NPN for NPN transistors, etc).
The optional parameters (‘is’, ‘rs’, ‘n’, . . . ) shown in this example configure the simula-
tor’s mathematical model of the diode to match the specific behavior of a particular part
(e.g. a ‘1n4148’).
Subcircuits In some applications, describing a device by embedding the required ele-

ments in the main circuit file, as is done for the amplifier in Fig. 26.1, is not desirable.
A hierarchical approach may be taken by using subcircuits. An example of a subcircuit
statement is shown in the second circuit file:
X1 Amp_In 0 Amp_Out
Amplifier Subcircuits are always identified by a device label beginning with ‘X’. Just as
with other devices, all of the connected nodes are specified. Notice, in this example, that
three nodes are used. Finally, the name of the subcircuit is specified. Elsewhere in the
circuit file, the simulator looks for a statement of the form:
. subckt < Name > < Node1 > < Node2 > < Node3 > ...
This statement specifies that the lines that follow are part of the Amplifier subcircuit,
and that the three nodes listed are to be treated wherever they occur in the subcircuit
definition as referring, respectively, to the nodes on the main circuit from which the
subcircuit was called. Normal device, model, and comment statements may then follow.
The subcircuit definition is concluded with a statement of the form:
. ends < Name >
26.1.1.3 XSPICE Code Models
In the previous example, the specification of the amplifier was accomplished by using a
ngspice Voltage Controlled Voltage Source device. This is an idealization of the actual
amplifier. Practical amplifiers include numerous non-ideal effects, such as offset error
voltages and non-ideal input and output impedances. The accurate simulation of complex,
real-world components can lead to cumbersome subcircuit files, long simulation run times,
and difficulties in synthesizing the behavior to be modeled from a limited set of internal
devices known to the simulator.
To address these problems, XSPICE allows you to create Code Models that simulate
complex, non-ideal effects without the need to develop a subcircuit design. For example,
the following file provides simulation of the circuit in Fig. 26.2, but with the subcircuit
amplifier replaced with a Code Model called ‘Amp’ that models several non-ideal effects
including input and output impedance and input offset voltage.
Small Signal Amplifier

*
* This circuit simulates a small signal amplifier
* with a diode limiter .
*
. dc Vin -1 1 .05
Vin Input 0 DC 0
D_Neg 0 Amp_In 1 n4148
D_Pos Amp_In 0 1 n4148
C1 Amp_In 0 1 uF
A1 Amp_In 0 Amp_Out Amplifier

+ tt =3.030 e -09 cjo =1.700 e -12 vj =1 m =1.959 e -01 bv =1.000 e +02
+ ibv =1.000 e -04)
. model Amplifier Amp ( gain = -10 in_offset = 1e -3

+ rin = 1 meg rout = 0.4)
. end
A statement with a device label beginning with ‘A’ alerts the simulator that the device
uses a Code Model. The model statement is similar in form to the one used to specify
the diode. The model label ‘Amp’ directs XSPICE to use the code model with that
name. Parameter information has been added to specify a gain of -10, an input offset
of 1 millivolt, an input impedance of 1 meg ohm, and an output impedance of 0.4 ohm.
26.2. CIRCUIT DESCRIPTION SYNTAX 559
Subsequent sections of this document detail the steps required to create such a Code
Model and include it in the XSPICE simulator.
26.1.1.4 Node Bridge Models
When a mixed-mode simulator is used, some method must be provided for translating
data between the different simulation algorithms. XSPICE’s Code Model support allows
you to develop models that work under the analog simulation algorithm, the event-driven
simulation algorithm, or both at once.
In XSPICE, models developed for the express purpose of translating between the different
algorithms or between different User-Defined Node types are called ‘Node Bridge’ models.
For translations between the built-in analog and digital types, predefined node bridge
models are included in the XSPICE Code Model Library.
26.1.1.5 Practical Model Development
In practice, developing models often involves using a combination of ngspice passive de-
vices, device models, subcircuits, and XSPICE Code Models. XSPICE’s Code Models
may be seen as an extension to the set of device models offered in standard ngspice. The
collection of over 40 predefined Code Models included with XSPICE provides you with
an enriched set of modeling primitives with which to build subcircuit models. In general,
you should first attempt to construct your models from these available primitives. This
is often the quickest and easiest method.
If you find that you cannot easily design a subcircuit to accomplish your goal using the
available primitives, then you should turn to the code modeling approach. Because they
are written in a general purpose programming language (C), code models enable you to
simulate virtually any behavior for which you can develop a set of equations or algorithms.
26.2 Circuit Description Syntax

If you need to debug a simulation, if you are planning to develop your own models, or
if you are using the XSPICE simulator through the control language user interface, you
will need to become familiar with the circuit description language.
The previous sections presented example circuit description input files. The following
sections provide more detail on XSPICE circuit descriptions with particular emphasis on
the syntax for creating and using models. First, the language and syntax of the ngspice
simulator are described and references to additional information are given. Next, XSPICE
extensions to the ngspice syntax are detailed. Finally, various enhancements to ngspice
operation are discussed including polynomial sources, arbitrary phase sources, supply
ramping, matrix conditioning, convergence options, and debugging support.
26.2.1 XSPICE Syntax Extensions

In the preceding discussion, ngspice syntax was reviewed, and those features of ngspice
that are specifically supported by the XSPICE simulator were enumerated. In addition to
these features, there exist extensions to the ngspice capabilities that provide much more
flexibility in describing and simulating a circuit. The following sections describe these
capabilities, as well as the syntax required to make use of them.
26.2.1.1 Convergence Debugging Support
When a simulation is failing to converge, the simulator can be asked to return convergence
diagnostic information that may be useful in identifying the areas of the circuit in which
convergence problems are occurring. When running through the interactive user interface,
these messages are written directly to the terminal.
26.2.1.2 Digital Nodes
Support is included for digital nodes that are simulated by an event-driven algorithm.
Because the event-driven algorithm is faster than the standard SPICE matrix solution
algorithm, and because all digital, real, int and User-Defined Node types make use of
the event-driven algorithm, reduced simulation time for circuits that include these models
can be anticipated compared to simulation of the same circuit using analog code models
and nodes.
26.2.1.3 User-Defined Nodes
Support is provided for User Defined Nodes that operate with the event-driven algorithm.
These nodes allow the passing of arbitrary data structures among models. The real and
integer node types supplied with XSPICE are actually predefined User-Defined Node
types.
26.2.1.4 Supply Ramping
A supply ramping function is provided by the simulator as an option to a transient analysis

to simulate the turn-on of power supplies to a board level circuit. To enable this option, the
compile time flag XSPICE_EXP has to be set, e.g. by adding CFLAGS="-DXSPICE_EXP"
to the ./configure command line. The supply ramping function linearly ramps the val-
ues of all independent sources and the capacitor and inductor code models (code model
extension) with initial conditions toward their final value at a rate that you define. A
complete ngspice deck example of usage of the ramptime option is shown below.
26.3. HOW TO CREATE CODE MODELS 561
Example:
Supply ramping option
*
* This circuit demonstrates the use of the option
* " ramptime " that ramps independent sources and the
* capacitor and inductor initial conditions from
* zero to their final value during the time period
* specified .
*
*
. tran 0.1 5
. option ramptime =0.2
* a1 1 0 cap
. model cap capacitor ( c =1000 uf ic =1)
r1 1 0 1 k
*
a2 2 0 ind
. model ind inductor ( l =1 H ic =1)
r2 2 0 1.0
*
v1 3 0 1.0
r3 3 0 1 k
*
i1 4 0 1e -3
r4 4 0 1 k
*
v2 5 0 0.0 sin (0 1 0.3 0 0 45.0)
r5 5 0 1 k
*
. end
26.3 How to create code models

The following instruction to create an additional code model uses the ngspice infrastruc-
ture and some ’intelligent’ copy and paste. As an example an extra code model d_xxor is
created in the xtradev shared library, reusing the existing d_xor model from the digital
library. More detailed information will be made available in Chapt. 28.
You should have downloaded ngspice, either the most recent distribution or from the Git
repository, and compiled and installed it properly according to your operating system and
the instructions given in Chapt. 32 of the Appendix, especially Chapt. 32.1.4 (for Linux
users), or Chapt. 32.2.2 for MINGW and MS Windows. (MS Visual Studio will not do,
because we have not yet integrated the code model generator into this compiler! You
may however use all code models later with any ngspice executable.) Then change into
directory ngspice/src/xspice/icm/xtradev.
Create a new directory
mkdir d_xxor
Copy the two files cfunc.mod and ifspec.ifs from ngspice/src/xspice/icm/digital/d_xor to

ngspice/src/xspice/icm/xtradev/d_xxor. These two files may serve as a template for your
new model!
For simplicity reasons we do only a very simple editing to these files here, in fact the
functionality is not changed, just the name translated to a new model. Edit the new
cfunc.mod: In lines 5, 28, 122, 138, 167, 178 replace the old name (d_xor) by the new
name d_xxor. Edit the new ifspec.ifs: In lines 16, 23, 24 replace cm_d_xor by cm_d_xxor
and d_xor by d_xxor.
Make ngspice aware of the new code model by editing file

ngspice/src/xspice/icm/xtradev/modpath.lst:
Add a line with the new model name d_xxor.
Redo ngspice by entering directory ngspice/release, and issuing the commands:
make
sudo make install
And that’s it! In ngspice/release/src/xspice/icm/xtradev/ you now will find cfunc.c and
ifspec.c and the corresponding object files. The new code model d_xxor resides in the
shared library xtradev.cm, and is available after ngspice is started. As a test example you
may edit
ngspice/src/xspice/examples/digital_models1.deck, and change line 60 to the new model:
.model d_xor1 d_xxor (rise_delay=1.0e-6 fall_delay=2.0e-6 input_load=1.0e-12)
The complete input file follows:

26.3. HOW TO CREATE CODE MODELS 563
Code Model Test : new xxor

*
*** analysis type ***
. tran .01 s 4 s
*
*** input sources ***
*
v2 200 0 DC PWL ( (0 0.0) (2 0.0) (2.0000000001 1.0) (3 1.0) )
*
v1 100 0 DC PWL ( (0 0.0) (1.0 0.0) (1.0000000001 1.0) (2 1.0)
+ (2.0000000001 0.0) (3 0.0) (3.0000000001 1.0) (4 1.0) )
*
*** resistors to ground ***
r1 100 0 1 k
r2 200 0 1 k
*
*** adc_bridge blocks ***
aconverter [200 100] [2 1] adc_bridge1
. model adc_bridge1 adc_bridge ( in_low =0.1 in_high =0.9
+ rise_delay =1.0 e -12 fall_delay =1.0 e -12)
*
*** xor block ***
a7 [1 2] 70 d_xor1
. model d_xor1 d_xxor ( rise_delay =1.0 e -6 fall_delay =2.0 e -6
+ input_load =1.0 e -12)
*
*** dac_bridge blocks ****
abridge1 [70] [ out ] dac1
. model dac1 dac_bridge ( out_low = 0.7 out_high = 3.5
+ out_undef = 2.2 input_load = 5.0 e -12 t_rise = 50 e -9
+ t_fall = 20 e -9)
*
*** simulation and plotting ***
. control
run
plot allv
. endc
*
. end
An analog input, delivered by the pwl voltage sources, is transformed into the digital
domain by an adc_bridge, processed by the new code model d_xxor, and then translated
back into the analog domain.
If you want to change the functionality of the new model, you have to edit ifspec.ifs for
the code model interface and cfunc.mod for the detailed functionality of the new model.
Please see Chapt. 28, especially Chapt. 28.6 ff. for any details. And of course you may
take the existing analog, digital, mixed signal and other existing code models (to be found
in the subdirectories to ngspice/release/src/xspice/icm) as stimulating examples for your
work.
Chapter 27
Example circuits
The following chapter is designed to demonstrate XSPICE features. The first example
circuit models the circuit of Fig. 26.2 using the XSPICE gain block code model to
substitute for the more complex and computationally expensive ngspice transistor model.
This example illustrates one way in which XSPICE code models can be used to raise the
level of abstraction in circuit modeling to improve simulation speed.
The next example, shown in Fig. 27.1, illustrates many of the more advanced features
offered by XSPICE. This circuit is a mixed-mode design incorporating digital data, analog
data, and User-Defined Node data together in the same simulation. Some of the important
features illustrated include:
• Creating and compiling Code Models
• Creating an XSPICE executable that incorporates these new models
• The use of ‘node bridge’ models to translate data between the data types in the
simulation
• Plotting analog and event-driven (digital and User-Defined Node) data
• Using the eprint command to print event-driven data
Throughout these examples, we assume that ngspice with XSPICE option has already
been installed on your system and that your user account has been set up with the proper
search path and environment variable data.
The examples also assume that you are running under Linux and will use standard Linux
commands such as cp for copying files, etc. If you are using a different set up, with
different operating system command names, you should be able to translate the commands
shown into those suitable for your installation. Finally, file system path-names given
in the examples assume that ngspice + XSPICE has been installed on your system in
directory /usr/local/xspice-1-0. If your installation is different, you should substitute the
appropriate root path-name where appropriate.
565
27.1 Amplifier with XSPICE model ‘gain’

The circuit, as has been shown in Fig. 26.2, is extended here by using the XSPICE code
model gain. The ngspice circuit description for this circuit is shown below.
Example:
A transistor amplifier circuit

*
. tran 1e -5 2e -3
*
vin 1 0 0.0 ac 1.0 sin (0 1 1 k )
*
ccouple 1 in 10 uF
rzin in 0 19.35 k
*
aamp in aout gain_block
. model gain_block gain ( gain = -3.9 out_offset = 7.003)
*
rzout aout coll 3.9 k
rbig coll 0 1 e12
*
. end
Notice the component ‘aamp’. This is an XSPICE code model device. All XSPICE code
model devices begin with the letter ‘a’ to distinguish them from other ngspice devices.
The actual code model used is referenced through a user-defined identifier at the end
of the line - in this case gain_block. The type of code model used and its parameters
appear on the associated .model card. In this example, the gain has been specified as
-3.9 to approximate the gain of the transistor amplifier, and the output offset (out_offset)
has been set to 7.003 according to the DC bias point information obtained from the DC
analysis in Example 1 from Chapter 26.
Notice also that input and output impedances of the one-transistor amplifier circuit are
modeled with the resistors ‘rzin’ and ‘rzout’, since the gain code model defaults to an
ideal voltage-input, voltage-output device with infinite input impedance and zero output
impedance.
Lastly, note that a special resistor ‘rbig’ with value ‘1e12’ has been included at the opposite
side of the output impedance resistor ‘rzout’. This resistor is required by ngspice’s matrix
solution formula. Without it, the resistor ‘rzout’ would have only one connection to the
circuit, and an ill-formed matrix could result. One way to avoid such problems without
adding resistors explicitly is to use the ngspice ‘rshunt’ option described in this document
under ngspice Syntax Extensions/General Enhancements.
To simulate this circuit, copy the file xspice_c2.cir from the directory /src/xspice/ex-
amples into a directory in your account.
Invoke the simulator on this circuit:

27.2. XSPICE ADVANCED USAGE 567
After a few moments, you should see the ngspice prompt:
ngspice 1 ->
Now issue the run command and when the prompt returns, issue the plot command to
examine the voltage at the node ‘coll’.
ngspice 1 -> run

ngspice 2 -> plot coll
The resulting waveform closely matches that from the original transistor amplifier circuit
simulated in Example 1.
When you are done, enter the quit command to leave the simulator and return to the
command line.
ngspice 3 -> quit
Using the rusage command, you can verify that this abstract model of the transistor
amplifier runs somewhat faster than the full circuit of Example 1. This is because the
code model is less complex computationally. This demonstrates one important use of
XSPICE code models - to reduce run time by modeling circuits at a higher level of
abstraction. Speed improvements vary and are most pronounced when a large amount
of low-level circuitry can be replaced by a small number of code models and additional
components.
27.2 XSPICE advanced usage
27.2.1 Circuit example C3

An equally important use of code models is in creating models for circuits and systems
that do not easily lend themselves to synthesis using standard ngspice primitives (resistors,
capacitors, diodes, transistors, etc.). This occurs often when trying to create models of
ICs for use in simulating board-level designs. Creating models of operational amplifiers
such as an LM741 or timer ICs such as an LM555 is greatly simplified through the use of
XSPICE code models. Another example of code model use is shown in the next example
where a complete sampled-data system is simulated using XSPICE analog, digital, and
User-Defined Node types simultaneously.
The circuit shown in Fig. 27.1 is designed to demonstrate several of the more advanced
features of XSPICE. In this example, you will be introduced to the process of creating
code models and linking them into ngspice. You will also learn how to print and plot the
results of event-driven analysis data. The ngspice/XSPICE circuit description for this
example is shown below.
Figure 27.1: Example Circuit C3
Example:
Mixed IO types
* This circuit contains a mixture of IO types , including
* analog , digital , user - defined ( real ) , and ’ null ’.
*
* The circuit demonstrates the use of the digital and
* user - defined node capability to model system - level designs
* such as sampled - data filters . The simulated circuit
* contains a digital oscillator enabled after 100 us . The
* square wave oscillator output is divided by 8 with a
* ripple counter . The result is passed through a digital
* filter to convert it to a sine wave .
*
. tran 1e -5 1e -3
*
v1 1 0 0.0 pulse (0 1 1e -4 1e -6)
r1 1 0 1 k
*
abridge1 [1] [ enable ] atod
. model atod adc_bridge
*
aclk [ enable clk ] clk nand
. model nand d_nand ( rise_delay =1 e -5 fall_delay =1 e -5)
*
adiv2 div2_out clk NULL NULL NULL div2_out dff
adiv4 div4_out div2_out NULL NULL NULL div4_out dff
adiv8 div8_out div4_out NULL NULL NULL div8_out dff
. model dff d_dff
Example (continued):
abridge2 div8_out enable filt_in node_bridge2

. model node_bridge2 d_to_real ( zero = -1 one =1)
*
xfilter filt_in clk filt_out dig_filter
*
abridge3 filt_out a_out node_bridge3
. model node_bridge3 real_to_v
*
rlpf1 a_out oa_minus 10 k
*
xlpf 0 oa_minus lpf_out opamp
*
rlpf2 oa_minus lpf_out 10 k
clpf lpf_out oa_minus 0.01 uF
* * * *** *** *** *** *** ** ** * *** * ** *** *** *** *
. subckt dig_filter filt_in clk filt_out
. model n0 real_gain ( gain =1.0)
. model g1 real_gain ( gain =0.125)
. model zm1 real_delay
. model d0a real_gain ( gain = -0.75)
. model d1a real_gain ( gain =0.5625)
. model d0b real_gain ( gain = -0.3438)
. model d1b real_gain ( gain =1.0)
*
an0a filt_in x0a n0
an1a filt_in x1a n1
an2a filt_in x2a n2
*
az0a x0a clk x1a zm1
az1a x1a clk x2a zm1
*
ad0a x2a x0a d0a
ad1a x2a x1a d1a
*
az2a x2a filt1_out g1
az3a filt1_out clk filt2_in zm1
*
an0b filt2_in x0b n0
*
az0b x0b clk x1b zm1
az1b x1b clk x2b zm1
*
ad0 x2b x0b d0b
ad1 x2b x1b d1b
*
az2b x2b clk filt_out zm1
. ends dig_filter
Example (continued):
. subckt opamp plus minus out

*
r1 plus minus 300 k
a1 % vd ( plus minus ) outint lim
. model lim limit ( out_lower_limit = -12 out_upper_limit = 12
+ fraction = true limit_range = 0.2 gain =300 e3 )
r3 outint out 50.0
r2 out 0 1 e12
*
. ends opamp
*
. end
This circuit is a high-level design of a sampled-data filter. An analog step waveform

(created from a ngspice pulse waveform) is introduced as ‘v1’ and converted to digital by
code model instance ‘abridge’. This digital data is used to enable a Nand-Gate oscillator
(‘aclk’) after a short delay. The Nand-Gate oscillator generates a square-wave clock signal
with a period of approximately two times the gate delay, which is specified as 1e-5 seconds.
This 50 kHz clock is divided by a series of D Flip Flops (‘adiv2’, ‘adiv4’, ‘adiv8’) to produce
a square-wave at approximately 6.25 kHz. Note particularly the use of the reserved word
‘NULL’ for certain nodes on the D Flip Flops. This tells the code model that there is no
node connected to these ports of the flip flop.
The divide-by-8 and enable waveforms are converted by the instance ‘abridge2’ to the
format required by the User-Defined Node type ‘real’, which expected real-valued data.
The output of this instance on node filt_in is a real valued square wave that oscillates
between values of -1 and 1. Note that the associated code model d_to_real is not part
of the original XSPICE code model library but has been added later to ngspice.
This signal is then passed through subcircuit ‘xfilter’ that contains a digital low-pass filter
clocked by node ‘clk’. The result of passing this square-wave through the digital low-pass
filter is the production of a sampled sine wave (the filter passes only the fundamental of
the square-wave input) on node filt_out. This signal is then converted back to ngspice
analog data on node a_out by node bridge instance ‘abridge3’.
The resulting analog waveform is then passed through an op-amp-based low-pass analog
filter constructed around subcircuit ‘xlpf’ to produce the final output at analog node
‘lpf_out’.
27.2.2 Running example C3
Now copy the file xspice_c3.cir from directory /examples/xspice/ into the current directory:
and invoke the new simulator executable as you did in the previous examples.
Execute the simulation with the run command.
ngspice 1 -> run
After a short time, the ngspice prompt should return. Results of this simulation are
examined in the manner illustrated in the previous two examples. You can use the plot
command to plot either analog nodes, event-driven nodes, or both. For example, you can
plot the values of the sampled-data filter input node and the analog low-pass filter output
node as follows:
ngspice 2 -> plot filt_in lpf_out
The plot shown in Fig. 27.2 should appear.
Figure 27.2: Plot of Filter Input and Output
You can also plot data from nodes inside a subcircuit. For example, to plot the data on
node ‘x1a’ in subcircuit ‘xfilter’, create a pathname to this node with a dot separator.
ngspice 3 -> plot xfilter.x1a
The output from this command is shown in Fig. 27.3. Note that the waveform contains
vertical segments. These segments are caused by the non-zero delays in the ‘real gain’
models used within the subcircuit. Each vertical segment is actually a step with a width
equal to the model delay (1e-9 seconds).
Plotting nodes internal to subcircuits works for both analog and event-driven nodes.
To examine data such as the closely spaced events inside the subcircuit at node ‘xfil-
ter.x1a’, it is often convenient to use the eprint command to produce a tabular listing
of events. Try this by entering the following command:
Figure 27.3: Plot of Subcircuit Internal Node
ngspice 4 -> eprint xfilter.x1a

**** Results Data ****
Time or Step
xfilter.x1a
0.000000000e+000 0.000000e+000 1.010030000e-004 2.000000e+000
1.010040000e-004 2.562500e+000 1.210020000e-004 2.812500e+000
1.210030000e-004 4.253906e+000 1.410020000e-004 2.332031e+000
1.410030000e-004 3.283447e+000 1.610020000e-004 2.014893e+000
1.610030000e-004 1.469009e+000 1.810020000e-004 2.196854e+000
1.810030000e-004 1.176232e+000
...
9.610030000e-004 3.006294e-001 9.810020000e-004 2.304755e+000
9.810030000e-004 9.506230e-001 9.810090000e-004 -3.049377e+000
9.810100000e-004 -4.174377e+000
**** Messages ****
**** Statistics ****
Operating point analog/event alternations: 1
Operating point load calls: 37
Operating point event passes: 2
Transient analysis load calls: 4299
Transient analysis timestep backups: 87
This command produces a tabular listing of event-times in the first column and node
values in the second column. The 1 ns delays can be clearly seen in the fifth decimal place
of the event times.
Note that the eprint command also gives statistics from the event-driven algorithm por-
tion of XSPICE. For this example, the simulator alternated between the analog solution
algorithm and the event-driven algorithm one time while performing the initial DC operat-
ing point solution prior to the start of the transient analysis. During this operating point
analysis, 37 total calls were made to event-driven code model functions, and two separate
event passes or iterations were required before the event nodes obtained stable values.
Once the transient analysis commenced, there were 4299 total calls to event-driven code
model functions. Lastly, the analog simulation algorithm performed 87 time-step backups
that forced the event-driven simulator to backup its state data and its event queues.
A similar output is obtained when printing the values of digital nodes. For example, print
the values of the node ‘div8 out’ as follows:
ngspice 5 -> eprint div8_out

**** Results Data ****
Time or Step
div8_out
0.000000000e+000 1s
1.810070000e-004 0s
2.610070000e-004 1s
...
9.010070000e-004 1s
9.810070000e-004 0s
**** Messages ****
**** Statistics ****
Operating point analog/event alternations: 1
Operating point load calls: 37
Operating point event passes: 2
Transient analysis load calls: 4299
Transient analysis timestep backups: 87
From this printout, we see that digital node values are composed of a two character
string. The first character (0, 1, or U) gives the state of the node (logic zero, logic one,
or unknown logic state). The second character (s, r, z, u) gives the ‘strength’ of the logic
state (strong, resistive, hi-impedance, or undetermined).
If you wish, examine other nodes in this circuit with either the plot or eprint commands.
When you are done, enter the quit command to exit the simulator and return to the
operating system prompt:
ngspice 6 -> quit
So long.
Chapter 28
Code Models and User-Defined

Nodes
The following sections explain the steps required to create code models and User-Defined
Nodes (UDNs), store them in shared libraries and load them into the simulator at runtime.
The ngspice simulator already includes XSPICE libraries of predefined models and node
types that span the analog and digital domains. These have been detailed earlier in this
document (see Sections 12.2, 12.3, and 12.4). However, the real power of the XSPICE
is in its support for extending these libraries with new models written by users. ngspice
includes an XSPICE code model generator. Adding code models to ngspice will require a
model definition plus some simple file operations, which are explained in this chapter.
The original manual cited an XSPICE ‘Code Model Toolkit’ that enabled one to define
new models and node data types to be passed between them offline, independent from
ngspice. Whereas this Toolkit is still available in the original source code distribution at
the XSPICE web page, it is neither required nor supported any more.
So we make use of the existing XSPICE infrastructure provided with ngspice to create
new code models. With an ’intelligent’ copy and paste, and the many available code
models serving as a guide you will be quickly able to create your own models. You have
to have a compiler (gcc) available under Linux, MS Windows (Cygwin, MINGW), maybe
also for other OSs, including supporting software (Flex, Bison, and the autotools if you
start from Git sources). The compilation procedures for ngspice are described in detail in
Chapt. 32. Adding a code model may then require defining the functionality, interface,
and eventually user defined nodes. Compiling into a shared library is only a simple ’make’,
loading the shared lib(s) into ngspice is done by the ngspice command codemodel... (see
Chapt. 17.5.15). This will allow you to either add some code model to an existing library,
or you may generate a new library with your own code models. The latter is of interest
if you want to distribute your code models independently from the ngspice sources or
executables.
These new code models are handled by ngspice in a manner analogous to its treating
of SPICE devices and XSPICE Predefined Code Models. The basic steps required to
create sources for new code models or User-Defined Nodes, compile them and load them
into ngspice are similar. They consist of 1) creating the code model or UserDefined
Node (UDN) directory and its associated model or data files, 2) inform ngspice about
the code model or UDN directories that have to be compiled and linked into ngspice, 3)
575
576 CHAPTER 28. CODE MODELS AND USER-DEFINED NODES
compile them into a shared lib, and 4) load them into the ngspice simulator upon runtime.
All code models finally reside in dynamically linkable shared libraries (*.cm), which in
fact are .so files under Linux or dlls under MS Windows. Currently we have 5 of them
(analog.cm, digital.cm, spice2poly.cm, xtradev.cm, xtraevt.cm). Upon start up of ngspice
they are dynamically loaded into the simulator by the ngspice codemodel command (which
is located in file spinit (see Chapt. 16.5) for the standard code models). Once you have
added your new code model into one of these libraries (or have created a new library
file, e.g. my-own.cm), instances of the model can be placed into any simulator deck that
describes a circuit of interest and simulated along with all of the other components in
that circuit.
A quick entry to get a new code model has already been presented in Chapt. 26.3. You
may find the details of the XSPICE language in Chapt. 28.6 ff.
28.1 Code Model Data Type Definitions

There are several data types that you can incorporate into a model. These have already
been used extensively in the code model library included with the simulator. They are
detailed below:
Boolean_t The Boolean type is an enumerated type that can take on values of FALSE
(integer value 0) or TRUE (integer value 1). Alternative names for these enumerations
are MIF FALSE and MIF TRUE, respectively.
Complex_t The Complex type is a structure composed of two double values. The
first of these is the .real field, and the second is the .imag field. Typically these values are
accessed as shown:
For complex value ‘data’, the real portion is ‘data.real’, and the imaginary portion is
‘data.imag’.
Digital_State_t The Digital State type is an enumerated value that can be either
ZERO (integer value 0), ONE (integer value 1), or UNKNOWN (integer value 2).
Digital_Strength_t The Digital Strength type is an enumerated value that can

be either STRONG (integer value 0), RESISTIVE (integer value 1), HI IMPEDANCE
(integer value 2) or UNDETERMINED (integer value 3).
Digital_t The Digital type is a composite of the Digital_State_t and Digital_Strength_t

enumerated data types. The actual variable names within the Digital type are .state and
.strength and are accessed as shown below:
For Digital_t value ‘data’, the state portion is ‘data.state’, and the strength portion is
‘data.strength’.
28.2. CREATING CODE MODELS 577
28.2 Creating Code Models

The following description deals with extending one of the five existing code model libraries.
Adding a new library is described in Chapt. 28.4. The first step in creating a new code
model within XSPICE is to create a model directory inside of the selected library directory.
The new directory name is the name of the new code model. As an example you may add
a directory d_counter to the library directory digital.
cd ngspice/src/xspice/icm/digital
mkdir d_counter
Into this new directory you copy the following template files:
• Interface Specification File (ifspec.ifs)

• Model Definition File (cfunc.mod)
You may choose existing files that are similar to the new code model you intend to
integrate. The template Interface Specification File (ifspec.ifs) is edited to define the
model’s inputs, outputs, parameters, etc (see Chapt. 28.6). You then edit the template
Model Definition File (cfunc.mod) to include the C-language source code that defines the
model behavior (see Chapt. 28.7). As a final step you have to notify ngspice of the new
code model. You have to edit the file modpath.lst that resides in the library directory
ngspice/src/xspice/icm/digital. Just add the entry d_counter to this file.
The Interface Specification File is a text file that describes, in a tabular format, infor-
mation needed for the code model to be properly interpreted by the simulator when it is
placed with other circuit components into a SPICE deck. This information includes such
things as the parameter names, parameter default values, and the name of the model it-
self. The specific format presented to you in the Interface Specification File template must
be followed exactly, but is quite straightforward. A detailed description of the required
syntax, along with numerous examples, is included in Section 28.6.
The Model Definition File contains a C programming language function definition. This
function specifies the operations to be performed within the model on the data passed to
it by the simulator. Special macros are provided that allow the function to retrieve input
data and return output data. Similarly, macros are provided to allow for such things
as storage of information between iteration time-points and sending of error messages.
Section 28.7 describes the form and function of the Model Definition File in detail and
lists the support macros provided within the simulator for use in code models.
To allow compiling and linking (see Chapt. 28.5) you have at least to adapt the names of
the functions inside of the two copied files to get unique function and model names. If for
example you have chosen ifspec.ifs and cfunc.mod from model d_fdiv as your template,
simply replace all entries d_fdiv by d_counter inside of the two files.
28.3 Creating User-Defined Nodes

In addition to providing the capability of adding new models to the simulator, a facility
exists that allows node types other than those found in standard SPICE to be created.
Models may be constructed that pass information back and forth via these nodes. Such
models are constructed in the manner described in the previous sections, with appropriate
changes to the Interface Specification and Model Definition Files.
Because of the need of electrical engineers to have ready access to both digital and analog
simulation capabilities, the digital node type is provided as a built-in node type along
with standard SPICE analog nodes. For digital nodes, extensive support is provided
in the form of macros and functions so that you can treat this node type as a standard
type analogous to standard SPICE analog nodes when creating and using code models.
In addition to analog and digital nodes, the node types real and int are also provided
with the simulator. These were created using the User-Defined Node (UDN) creation
facilities described below and may serve as a template for further node types.
The first step in creating a new node type within XSPICE is to set up a node type
directory along with the appropriate template files needed.
cd ngspice/src/xspice/icm/xtraevt
mkdir <directory name>
<directory name> should be the name of the new type to be defined. Copy file udnfunc.c
from /icm/xtraevt/int into the new directory. Edit this file according to the new type you
want to create.
Notify ngspice about this new UDN directory by editing
ngspice/src/xspice/icm/xtraevt/udnpath.lst. Add a new line containing <directory name>.
For compiling and linking see Chapt. 28.5.
The UDN Definition File contains a set of C language functions. These functions perform
operations such as allocating space for data structures, initializing them, and comparing
them to each other. Section 28.8 describes the form and function of the User-Defined
Node Definition File in detail and includes an example UDN Definition File.
28.4 Adding a new code model library

A group of code models may be assembled into a library. A new library is a means to
distribute new code models, independently from the existing ones. This is the way to
generate a new code model library:
cd ngspice/src/xspice/icm/
mkdir <directory name>
<directory name> is the name of the new library. Copy empty files modpath.lst and
udnpath.lst into this directory.
Edit file ngspice/src/xspice/icm/GNUmakefile.in, add <directory name> to the end of line
10, which starts with CMDIRS = ... .
That’s all you have to do about a new library! Of course it is empty right now, so you
have to define at least one code model according to the procedure described in Chapt.
28.2.
28.5. COMPILING AND LOADING THE NEW CODE MODEL (LIBRARY) 579
28.5 Compiling and loading the new code model (li-

brary)
Compiling is now as simple as issuing the commands
cd ngspice/release
make
sudo make install
if you have installed ngspice according to Chapt. 32.1.4. This procedure will install
the code model libraries into a directory <prefix>/lib/spice/, e.g. C:/Spice/lib/spice/ for
standard Windows install or /usr/local/lib/spice/ for Linux.
Thus the code model libraries are not linked into ngspice at compile time, but may be
loaded at runtime using the codemodel command (see Chapt. 17.5.15). This is done
automatically for the predefined code model libraries upon starting ngspice. The appro-
priate commands are provided in the start up file spinit (see Chapt. 16.5). So if you have
added a new code model inside of one of the existing libraries, nothing has to be done,
you will have immediate access to your new model.
If you have generated a new code model library, e.g. new_lib.cm, then you have to add
the line
@XSPICEINIT@ codemodel @prefix@/@libname@/spice/new_lib.cm
to spinit.in in ngspice/src. This will create a new spinit if ngspice is recompiled from
scratch.
To avoid the need for recompilation of ngspice, you also may directly edit the file spinit
by adding the line
codemodel C:/Spice/lib/spice/new_lib.cm
(OS MS Windows) or the appropriate Linux equivalent. Upon starting ngspice, the new
library will be loaded and you have access to the new code model(s). The codemodel
command has to be executed upon start-up of ngspice, so that the model information is
available as soon as the circuit is parsed. Failing to do so will lead to an error message of
a model missing. So spinit (or .spiceinit for personal code model libraries) is the correct
place for codemodel.
28.6 Interface Specification File

The Interface Specification (IFS) file is a text file that describes the model’s naming infor-
mation, its expected input and output ports, its expected parameters, and any variables
within the model that are to be used for storage of data across an entire simulation. These
four types of data are described to the simulator in IFS file sections labeled NAME_TABLE,
PORT_TABLE, PARAMETER_TABLE and STATIC_VAR_TABLE, respectively. An example IFS
file is given below. The example is followed by detailed descriptions of each of the entries,
what they signify, and what values are acceptable for them. Keywords are case insensitive.
NAME_TABLE:
C_Function_Name: ucm_xfer
Spice_Model_Name: xfer
Description: "arbitrary transfer function"
PORT_TABLE:
Port_Name: in out
Direction: in out
Default_Type: v v
Vector: no no
Vector_Bounds: - -
Null_Allowed: no no
PARAMETER_TABLE:
Limits: - -
Vector: no no
Vector_Bounds: - -
PARAMETER_TABLE:
Parameter_Name: num_coeff
Description: "numerator polynomial coefficients"
Data_Type: real
Default_Value: -
Limits: -
Vector: yes
Null_Allowed: no
PARAMETER_TABLE:
Parameter_Name: den_coeff
Description: "denominator polynomial coefficients"
Data_Type: real
Default_Value: -
Limits: -
Vector: yes
Null_Allowed: no
PARAMETER_TABLE:
Parameter_Name: int_ic
Description: "integrator stage initial conditions"
Data_Type: real
Default_Value: 0.0
Limits: -
Vector: yes
Vector_Bounds: den_coeff
28.6. INTERFACE SPECIFICATION FILE 581
Null_Allowed: yes
STATIC_VAR_TABLE:
Static_Var_Name: x
Data_Type: pointer
Description: "x-coefficient array"
28.6.1 The Name Table
The name table is introduced by the Name_Table: keyword. It defines the code model’s
C function name, the name used on a .MODEL card, and an optional textual description.
The following sections define the valid fields that may be specified in the Name Table.
28.6.1.1 C Function Name
The C function name is a valid C identifier that is the name of the function for the
code model. It is introduced by the C_Function_Name: keyword followed by a valid C
identifier. To reduce the chance of name conflicts, it is recommended that user-written
code model names use the prefix ucm_ for this entry. Thus, in the example given above, the
model name is xfer, but the C function is ucm_xfer. Note that when you subsequently
write the model function in the Model Definition File, this name must agree with that of
the function (i.e., ucm_xfer), or an error will result in the linking step.
28.6.1.2 SPICE Model Name
The SPICE model name is a valid SPICE identifier that will be used on SPICE .MODEL
cards to refer to this code model. It may or may not be the same as the C function name.
It is introduced by the Spice_Model_Name: keyword followed by a valid SPICE identifier.
Description The description string is used to describe the purpose and function of the
code model. It is introduced by the Description: keyword followed by a C string literal.
28.6.2 The Port Table
The port table is introduced by the Port_Table: keyword. It defines the set of valid
ports available to the code model. The following sections define the valid fields that may
be specified in the port table.
28.6.2.1 Port Name
The port name is a valid SPICE identifier. It is introduced by the Port_Name: keyword
followed by the name of the port. Note that this port name will be used to obtain and
return input and output values within the model function. This will be discussed in more
detail in the next section.
Default Types
Type Description Valid Directions
d digital any
g conductance (VCCS) inout
gd differential conductance (VCCS) inout
h resistance (CCVS) inout
hd differential resistance (CCVS) inout
i current in or out
id differential current in or out
v voltage in or out
vd differential voltage in or out
<identifier> user-defined type any
Table 28.1: Port Types
28.6.2.2 Description
The description string is used to describe the purpose and function of the port. It is
introduced by the Description: keyword followed by a C string literal.
28.6.2.3 Direction
The direction of a port specifies the data flow direction through the port. A direction
must be one of n, out, or inout. It is introduced by the Direction: keyword followed
by a valid direction value.
28.6.2.4 Default Type
The Default_Type field specifies the type of a port. These types are identical to those
used to define the port types on a SPICE deck instance card (see Table 12.1), but without
the percent sign (%) preceding them. Table 28.1 summarizes the allowable types.
28.6.2.5 Allowed Types
A port must specify the types it is allowed to assume. An allowed type value must be a list
of type names (a blank or comma separated list of names delimited by square brackets,
e.g. [v vd i id] or [d]). The type names must be taken from those listed in Table 28.1.
28.6.2.6 Vector
A port that is a vector can be thought of as a bus. The Vector field is introduced with
the Vector: keyword followed by a Boolean value: YES, TRUE, NO or FALSE.
The values YES and TRUE are equivalent and specify that this port is a vector. Like-
wise, NO and FALSE specify that the port is not a vector. Vector ports must have a
corresponding vector bounds field that specifies valid sizes of the vector port.
28.6.2.7 Vector Bounds
If a port is a vector, limits on its size must be specified in the vector bounds field. The
Vector Bounds field specifies the upper and lower bounds on the size of a vector. The
Vector Bounds field is usually introduced by the Vector_Bounds: keyword followed by
a range of integers (e.g. ‘[1 7]’ or ‘[3, 20]’). The lower bound of the vector specifies
the minimum number of elements in the vector; the upper bound specifies the maximum
number of elements. If the range is unconstrained, or the associated port is not a vector,
the vector bounds may be specified by a hyphen (‘-’). Using the hyphen convention,
partial constraints on the vector bound may be defined (e.g., ‘[2, -]’ indicates that the
least number of port elements allowed is two, but there is no maximum number).
28.6.2.8 Null Allowed
In some cases, it is desirable to permit a port to remain unconnected to any electrical

node in a circuit. The Null_Allowed field specifies whether this constitutes an error for a
particular port. The Null_Allowed field is introduced by the ‘Null_Allowed:’ keyword
and is followed by a boolean constant: ‘YES’, ‘TRUE’, ‘NO’ or ‘FALSE’. The values ‘YES’
and ‘TRUE’ are equivalent and specify that it is legal to leave this port unconnected. ‘NO’
or ‘FALSE’ specify that the port must be connected.
28.6.3 The Parameter Table
The parameter table is introduced by the Parameter_Table: keyword. It defines the set
of valid parameters available to the code model. The following sections define the valid
fields that may be specified in the parameter table.
28.6.3.1 Parameter Name
A parameter name is a valid SPICE identifier that will be used on SPICE .MODEL cards
to refer to this parameter. It is introduced by the Parameter_Name: keyword followed
by a valid SPICE identifier.
The description string is used to describe the purpose and function of the parameter. It
is introduced by the ‘Description:’ keyword followed by a string literal.
28.6.3.3 Data Type
The parameter’s data type is specified by the Data Type field. The Data Type field is
introduced by the keyword ‘Data_Type:’ and is followed by a valid data type. Valid data
types include boolean, complex, int, real, and string.
28.6.3.4 Null Allowed
The Null_Allowed field is introduced by the ‘Null_Allowed:’ keyword and is followed by

a boolean literal. A value of ‘TRUE’ or ‘YES’ specify that it is valid for the corresponding
SPICE .MODEL card to omit a value for this parameter. If the parameter is omitted, the
default value is used. If there is no default value, an undefined value is passed to the code
model, and the PARAM_NULL( ) macro will return a value of ‘TRUE’ so that defaulting
can be handled within the model itself. If the value of Null_Allowed is ‘FALSE’ or ‘NO’,
then the simulator will flag an error if the SPICE .MODEL card omits a value for this
parameter.
28.6.3.5 Default Value
If the Null_Allowed field specifies ‘TRUE’ for this parameter, then a default value may
be specified. This value is supplied for the parameter in the event that the SPICE .MODEL
card does not supply a value for the parameter. The default value must be of the correct
type. The Default Value field is introduced by the ‘Default_Value:’ keyword and is
followed by a numeric, boolean, complex, or string literal, as appropriate.
28.6.3.6 Limits
Integer and real parameters may be constrained to accept a limited range of values. The
following range syntax is used whenever such a range of values is required. A range is
specified by a square bracket followed by a value representing a lower bound separated
by space from another value representing an upper bound and terminated with a closing
square bracket (e.g.”[0 10]”). The lower and upper bounds are inclusive. Either the lower
or the upper bound may be replaced by a hyphen (‘-’) to indicate that the bound is
unconstrained (e.g. ‘[10 -]’ is read as ‘the range of values greater than or equal to 10’).
For a totally unconstrained range, a single hyphen with no surrounding brackets may be
used. The parameter value limit is introduced by the ‘Limits:’ keyword and is followed
by a range.
28.6.3.7 Vector
The Vector field is used to specify whether a parameter is a vector or a scalar. Like the
PORT TABLE Vector field, it is introduced by the ‘Vector:’ keyword and followed by a
boolean value. ‘TRUE’ or ‘YES’ specify that the parameter is a vector. ‘FALSE’ or ‘NO’
specify that it is a scalar.
28.6.3.8 Vector Bounds
The valid sizes for a vector parameter are specified in the same manner as are port sizes
(see Section 28.6.2.7). However, in place of using a numeric range to specify valid vector
bounds it is also possible to specify the name of a port. When a parameter’s vector
bounds are specified in this way, the size of the vector parameter must be the same as
the associated vector port.
28.6.4 Static Variable Table

The Static Variable table is introduced by the ‘Static_Var_Table:’ keyword. It defines
the set of valid static variables available to the code model. These are variables whose
values are retained between successive invocations of the code model by the simulator.
The following sections define the valid fields that may be specified in the Static Variable
Table.
28.6.4.1 Name
The Static variable name is a valid C identifier that will be used in the code model to refer
to this static variable. It is introduced by the ‘Static_Var_Name:’ keyword followed by
a valid C identifier.
The description string is used to describe the purpose and function of the static variable.
It is introduced by the ‘Description:’ keyword followed by a string literal.
28.6.4.3 Data Type
The static variable’s data type is specified by the Data Type field. The Data Type field is
introduced by the keyword Data_Type: and is followed by a valid data type. Valid data
types include boolean, complex, int, real, string and pointer.
Note that pointer types are used to specify vector values; in such cases, the allocation of
memory for vectors must be handled by the code model through the use of the malloc()
or calloc() C function. Such allocation must only occur during the initialization cycle of
the model (which is identified in the code model by testing the INIT macro for a value of
TRUE). Otherwise, memory will be unnecessarily allocated each time the model is called.
Following is an example of the method used to allocate memory to be referenced by
a static pointer variable ‘x’ and subsequently use the allocated memory. The example
assumes that the value of ‘size’ is at least 2, else an error would result. The references to
STATIC_VAR(x) that appear in the example illustrate how to set the value of, and then
access, a static variable named ‘x’. In order to use the variable ‘x’ in this manner, it must
be declared in the Static Variable Table of the code model’s Interface Specification File.
/* Define local pointer variable */

double * local . x ;
/* Allocate storage to be referenced by the static variable x . */

/* Do this only if this is the initial call of the code model . */
if ( INIT == TRUE ) {
STATIC_VAR ( x ) = calloc ( size , sizeof ( double ));
}
/* Assign the value from the static pointer value to the local */
/* pointer variable . */
local_x = STATIC_VAR ( x );
/* Assign values to first two members of the array */

local_x [0] = 1.234;
local_x [1] = 5.678;
28.7 Model Definition File

The Model Definition File is a C source code file that defines a code model’s behavior
given input values that are passed to it by the simulator. The file itself is always given the
name cfunc.mod. In order to allow for maximum flexibility, passing of input, output, and
simulator-specific information is handled through accessor macros, which are described
below. These are not ordinary C preprocesor macros; they are expanded by a dedicated
program that checks them against the interface specification file. In addition, certain
predefined library functions (e.g. smoothing interpolators, complex arithmetic routines)
are included in the simulator in order to ease the burden of the code model programmer.
These are also described below.
28.7.1 Macros
The use of the accessor macros is illustrated in the following example. Note that the
argument to most accessor macros is the name of a parameter or port as defined in the
Interface Specification File. Note also that all accessor macros except ‘ARGS’ resolve to
an lvalue (C language terminology for something that can be assigned a value). Accessor
macros do not implement expressions or assignments.
void code . model ( ARGS ) /* private structure accessed by

accessor macros
*/
{
/* The following code fragments are intended to show how
information in the argument list is accessed . The reader
should not attempt to relate one fragment to another .
Consider each fragment as a separate example .
*/
double p ,/* variable for use in the following code fragments */

x, /* variable for use in the following code fragments */
y; /* variable for use in the following code fragments */
int i , /* indexing variable for use in the following */

j; /* indexing variable for use in the following */
28.7. MODEL DEFINITION FILE 587
UDN_Example_Type * a_ptr , /* A pointer used to access a

User - Defined Node type */
* y_ptr ; /* A pointer used to access a
User - Defined Node type */
/* Initializing and outputting a User - Defined Node result */

if ( INIT ) {
OUTPUT ( y ) = malloc ( sizeof ( user . defined . struct ));
y_ptr = OUTPUT ( y );
y_ptr - > component1 = 0.0;
y_ptr - > component2 = 0.0;
}
else {
y_ptr = OUTPUT ( y );
y_ptr - > component1 = x1 ;
y_ptr - > component2 = x2 ;
}
/* Determining analysis type */

if ( ANALYSIS == AC ) {
/* Perform AC analysis - dependent operations here */
}
/* Accessing a parameter value from the . model card */

p = PARAM ( gain );
/* Accessing a vector parameter from the . model card */

for ( i = 0; i < PARAM_SIZE ( in_offset ); i ++)
p = PARAM ( in_offset [ i ]);
/* Accessing the value of a simple real - valued input */

x = INPUT ( a );
/* Accessing a vector input and checking for null port */

if ( ! PORT_NULL ( a ))
for ( i = 0; i < PORT_SIZE ( a ); i ++)
x = INPUT ( a [ i ]);
/* Accessing a digital input */

x = INPUT ( a );
/* Accessing the value of a User - Defined Node input ...

*/
/* This node type includes two elements in its definition . */
a_ptr = INPUT ( a );
x = a_ptr - > component1 ;
y = a_ptr - > component2 ;
/* Outputting a simple real - valued result */

OUTPUT ( out1 ) = 0.0;
/* Outputting a vector result and checking for null */

if ( ! PORT_NULL ( a ))
for ( i = 0; i < PORT_SIZE ( a ); i ++)
OUTPUT ( a [ i ]) = 0.0;
/* Outputting the partial of output out1 w . r . t . input a */

PARTIAL ( out1 , a ) = PARAM ( gain );
/* Outputting the partial of output out2 ( i ) w . r . t . input b ( j ) */

for ( i = 0; i < PORT_SIZE ( out2 ); i ++) {
for ( j = 0; j < PORT_SIZE ( b ); j ++) {
PARTIAL ( out2 [ i ] , b [ j ]) = 0.0;
}
}
/* Outputting gain from input c to output out3 in an

AC analysis */
complex_gain_real = 1.0;
complex_gain_imag = 0.0;
AC_GAIN ( out3 , c ) = complex_gain ;
/* Outputting a digital result */

OUTPUT_STATE ( out4 ) = ONE ;
/* Outputting the delay for a digital or user - defined output */

OUTPUT_DELAY ( out5 ) = 1.0 e -9;
}
28.7.1.1 Macro Definitions
The full set of accessor macros is listed below. Arguments shown in parenthesis are
examples only. Explanations of the accessor macros are provided in the subsections below.
Circuit Data:
ARGS
CALL_TYPE
INIT
ANALYSIS
FIRST_TIMEPOINT
TIME
T(n)
RAD_FREQ
TEMPERATURE
CALLBACK
Parameter Data:
PARAM(gain)
PARAM_SIZE(gain)
PARAM_NULL(gain)
Port Data:
PORT_SIZE(a)
PORT_NULL(a)
LOAD(a)
TOTAL_LOAD(a)
Input Data:
INPUT(a)
INPUT_STATE(a)
INPUT_STRENGTH(a)
Output Data:
OUTPUT(y)
OUTPUT_CHANGED(a)
OUTPUT_DELAY(y)
OUTPUT_STATE(a)
OUTPUT_STRENGTH(a)
Partial Derivatives:
PARTIAL(y,a)
AC Gains:
AC_GAIN(y,a)
Static Variable:
STATIC_VAR(x)
28.7.1.2 Macro arguments for vector ports and parameters
When a port or parameter has been defined as a vector in the interface specification, some
macros require an argument that includes the vector index, as in OUTPUT(y[i]), while
others accept only the simple port or parameter name. The first kind are identified as
“indexed” in the descriptions below. The index may be any C expression that specifies a
valid index for the vector.
28.7.1.3 Circuit Data

ARGS
CALL_TYPE
INIT
ANALYSIS
FIRST_TIMEPOINT
TIME
T(n)
RAD_FREQ
TEMPERATURE
CALLBACK
ARGS is a macro that is passed in the argument list of every code model. It is there
to provide a way of referencing each model to all of the remaining macro values. It
must be present in the argument list of every code model; it must also be the only
argument present in the argument list of every code model.
CALL_TYPE is a macro that returns one of two possible symbolic constants. These
are EVENT and ANALOG. Testing may be performed by a model using CALL
TYPE to determine whether it is being called by the analog simulator or the event-
driven simulator. This will, in general, only be of value to a hybrid model such as
the adc bridge or the dac bridge. Some expected behaviours of the code model, such
as setting output values, depend on the CALL TYPE. For code models that request
it (see 28.7.2.7) a third value, STEP_PENDING, is used when the call indicates
that the simulator is about to complete an analog time step.
INIT is an integer (int) that takes the value 1 or 0 depending on whether this is the first
call to the code model instance or not, respectively.
ANALYSIS is an enumerated integer that takes values of DC, AC, or TRANSIENT.
FIRST_TIMEPOINT is an integer that takes the value 1 or 0 depending on whether

this is the first call for this instance at the current analysis step (i.e., time-point) or
not, respectively.
TIME is a double representing the current analysis time in a transient analysis. T(n) is a
double vector giving the analysis time for a specific analog time-point in a transient
analysis, where n takes the value 0 or 1. T(0) is equal to TIME. T(1) is the last
accepted time-point. (T(0) - T(1)) is the time-step (i.e., the delta-time value)
associated with the current time. The T(x) macro is valid only during ANALOG
calls (CALL TYPE == ANALOG). The difference between the TIME values in
two successive calls to a code model function may be zero, or even negative if the
analog simulator reduces the time-step while seeking convergence. TIME will never
be less than any current or previous valid T(1).
RAD_FREQ is a double representing the current analysis frequency in an AC analysis

expressed in units of radians per second.
TEMPERATURE is a double representing the current analysis temperature.
CALLBACK is a variable of type Mif_Callback_t, a function pointer defined in the

header file miftypes.h. A function may be supplied by assigning to CALLBACK
in the INIT call to the code model. That function will then be called during re-
set or deletion of instances of the code model. It is expected to release any extra
resources such as dynamic memory or open files that have been allocated during
simulation. Most code models will not need this as storage for variables allocated
through the library are released automatically. When the function is called, the
first argument is ARGS and the second is a reason code: currently the only value
is MIF_CB_DESTROY. That should be checked in case new call reasons are in-
troduced. The set of macros that can be used in the function is restricted to those
for ports, parameters and static variables.
28.7.1.4 Parameter Data

PARAM(gain)
PARAM_SIZE(gain)
PARAM_NULL(gain)
PARAM(gain) (indexed) resolves to the value of the scalar (i.e., non-vector) parameter
‘gain’ that was defined in the Interface Specification File tables. The macro evalu-
ation has the type given in the ifspec.ifs file and can be used regardless of type. If
‘gain’ is a string, then PARAM(gain) resolves to a read-only character pointer.
PARAM_SIZE(gain) resolves to an integer (int) representing the size of the ‘gain’ vec-
tor (which was dynamically determined when the SPICE deck was read). PARAM_SIZE(gain)
is undefined if ‘gain’ is a scalar.
PARAM_NULL(gain) resolves to an integer with value 0 or 1 depending on whether

a value was specified for gain, or whether the value is defaulted, respectively.
28.7.1.5 Port Data

PORT_SIZE(a)
PORT_NULL(a)
LOAD(a)
TOTAL_LOAD(a)
PORT_SIZE(a) resolves to an integer (int) representing the size of the ‘a’ vector port
(which was dynamically determined when the SPICE deck was read). PORT_SIZE(a)
is undefined if gain is a scalar.
PORT_NULL(a) resolves to an integer (int) with value 0 or 1 depending on whether

the SPICE deck has a node specified for this port, or has specified that the port is
null, respectively.
LOAD(a) (indexed) is used in a digital model to post a capacitive load value to a

particular input or output port during the INIT pass of the simulator. All values
posted for a particular event-driven node using the LOAD() macro are summed,
producing a total load value.
TOTAL_LOAD(a) (indexed) returns a double value that represents the total capaci-
tive load seen on a specified node to which a digital code model is connected. This
information may be used after the INIT pass by the code model to modify the delays
it posts with its output states and strengths. Note that this macro can also be used
by non-digital event-driven code models (see LOAD(), above).
28.7.1.6 Input Data

INPUT(a)
INPUT_STATE(a)
INPUT_STRENGTH(a)
The input data macros are all indexed.
INPUT(a) resolves to the value of the scalar input a that was defined in the Interface
Specification File tables. The macro evaluates to a real (double) value for analog
ports and a pointer to the internal representation for event ports (digital, integer,
real or user-defined). The same accessor macro can be used regardless of type.
INPUT_STATE(a) resolves to the state value defined for digital node types. These
will be one of the symbolic constants ZERO, ONE, or UNKNOWN.
INPUT_STRENGTH(a) resolves to the strength with which a digital input node is

being driven. This is determined by a resolution algorithm that looks at all outputs
to a node and determines its final driven strength. This value in turn is passed to
a code model when requested by this macro. Possible strength values are
1. STRONG
2. RESISTIVE
3. HI_IMPEDANCE
4. UNDETERMINED
28.7.1.7 Output Data

OUTPUT(y)
OUTPUT_CHANGED(a)
OUTPUT_DELAY(y)
OUTPUT_STATE(a)
OUTPUT_STRENGTH(a)
The output data macros are all indexed.
OUTPUT(y) resolves to the value of the scalar output ‘y’ that was defined in the In-
terface Specification File tables. The macro evaluates to a real l-value (it can be
assigned to) for analog ports and a pointer to the internal representation (derefer-
ence before assigning a value) for event ports (digital, integer, real or user-defined).
The same accessor macro can be used regardless of type. Event simulator port values
must only be set in EVENT calls. All analog simulator ports should be set in ANA-
LOG calls, otherwise the value reverts to zero. Setting analog ports in an EVENT
call does nothing. To handle the case where a new output for the“other” simula-
tor is produced, schedule a re-entry to that simulator using cm_event_queue() or
cm_analog_set_temp_breakpoint(), as appropriate, passing TIME as the argu-
ment.
OUTPUT_CHANGED(a) may be assigned one of two values for any particular out-
put to an event simulator port. If assigned the value TRUE (the default), then an
output state, strength (if digital) and delay must be posted by the model during
the call. If, on the other hand, no change has occurred during that pass, the OUT-
PUT_CHANGED(a) value for an output can be set to FALSE. In this case, no
state, strength or delay values need be posted by the model. Remember that this
macro applies to a single output port. If a model has multiple outputs that have
not changed, OUTPUT_CHANGED(a) must be set to FALSE for each of them.
OUTPUT_DELAY(y) may be assigned a double value representing a delay associated

with a particular event simulator port. Note that this macro must be set for each
digital or User-Defined Node output from a model during each pass, unless the
OUTPUT_CHANGED(a) macro is set to FALSE. Note also that a positive value
must be assigned to OUTPUT_DELAY(). Assigning a value of zero (or a negative
value) will cause an error.
OUTPUT_STATE(a) may be assigned a state value for a digital output node. Valid
values are ZERO, ONE, and UNKNOWN. This is the normal way of posting an out-
put state from a digital code model. This is a convenient alternative to constructing
a full digital value (state and strength) and assigning to *OUTPUT(a).
OUTPUT_STRENGTH(a) may be assigned a strength value for a digital output

node. This is the normal way of posting an output strength from a digital code
model. Valid values are
1. STRONG
2. RESISTIVE
3. HI_IMPEDANCE
4. UNDETERMINED
28.7.1.8 Partial Derivatives

PARTIAL(y,a)
PARTIAL(y[n],a)
PARTIAL(y,a[m])
PARTIAL(y[n],a[m])
PARTIAL(y,a) (indexed) resolves to the value of the partial derivative of scalar output
‘y’ with respect to scalar input ‘a’. The type is always double since partial derivatives
are only defined for nodes with real valued quantities (i.e., analog nodes).
The remaining uses of PARTIAL are shown for the cases in which either the output, the
input, or both are vectors.
Partial derivatives are required by the simulator to allow it to solve the non-linear equa-
tions that describe circuit behavior for analog nodes. Since coding of partial derivatives
can become difficult and error-prone for complex analog models, you may wish to consider
using the cm_analog_auto_partial() code model support function instead of using this
macro.
PARTIAL(a, a) is valid for an inout port and its use may greatly improve convergence.
28.7.1.9 AC Gains
AC_GAIN(y,a)
AC_GAIN(y[n],a)
AC_GAIN(y,a[m])
AC_GAIN(y[n],a[m])
AC_GAIN(y,a) (indexed) resolves to the value of the AC analysis gain of scalar output
‘y’ from scalar input ‘a’. The type is always a structure (Complex_t) defined in the
standard code model header file:
typedef struct Complex_s {

double real; /* The real part of the complex number */
double imag; /* The imaginary part of the complex number */
} Complex_t;
The remaining uses of AC_GAIN are shown for the cases in which either the output, the
input, or both are vectors.
28.7.1.10 Static Variables

STATIC_VAR(x)
STATIC_VAR(x) resolves to an lvalue for a scalar or pointer as defined in the Interface

Specification File. Unlike C language static and global variables, these values are
local to a specific instance of the code model. The type of ‘x’ is that given in the
Interface Specification File. The same accessor macro can be used regardless of type
since it simply resolves to an lvalue. To store a C structure or vector, the declared
type should be ’pointer’ and the code model is responsible for allocating storage and
assigning the pointer to the allocated storage to STATIC_VAR(x). That is usually
done in the INIT call. The code model is also responsible for freeing the storage, in
a function that is declared by the CALLBACK macro.
28.7.1.11 Accessor Macros
Table 28.3 describes the accessor macros available to the Model Definition File program-
mer and their C types. The PARAM and STATIC_VAR macros, whose types are labeled
CD (context dependent), return the type defined in the Interface Specification File. Argu-
ments listed with ‘[i]’ take an optional square bracket delimited index if the corresponding
port or parameter is a vector. The index may be any C expression - possibly involving
calls to other accessor macros (e.g.,” OUTPUT(out[PORT_SIZE(out)-1])”)
Name Type Args Description

AC_GAIN Complex_t y[i],x[i] AC gain of output y with respect
to input x.
ANALYSIS enum <none> Type of analysis: DC, AC,
TRANSIENT.
ARGS Mif_Private_t <none> Standard argument to all code
model function.
CALLBACK Mif_Callback_t <none> Pointer to callback function
CALL_TYPE enum <none> Type of model evaluation call:
ANALOG or EVENT.
INIT Boolean_t <none> Is this the first call to the model?
INPUT double, int or void* name[i] Value of analog input port, or

value of structure pointer for
User-Defined Node port.
INPUT_STATE enum name[i] State of a digital input: ZERO,
ONE, or UNKNOWN.
INPUT_STRENGTH enum name[i] Strength of digital input:
STRONG, RESISTIVE, HI
IMPEDANCE, or
UNDETERMINED.
INPUT_TYPE char* name[i] The port type of the input.
LOAD double name[i] The digital load value placed on
a port by this model.
MESSAGE char* name[i] A message output by a model on
an event-driven node.
OUTPUT double, int or void* name[i] Value of the analog output port
or value of structure pointer for
User-Defined Node port.
OUTPUT_CHANGED Boolean_t name[i] Has a new value been assigned
to this event-driven output by
the model?
OUTPUT_DELAY double name[i] Delay in seconds for an
event-driven output.
OUTPUT_STATE enum name[i] State of a digital output: ZERO,
ONE, or UNKNOWN.
OUTPUT_STRENGTH enum name[i] Strength of digital output:
STRONG, RESISTIVE,
HI_IMPEDANCE, or
UNDETERMINED.
OUTPUT_TYPE char* name[i] The port type of the output.
PARAM CD name[i] Value of the parameter.
PARAM_NULL Boolean_t name[i] Was the parameter not included
on the SPICE .model card ?
PARAM_SIZE int name Size of parameter vector.
PARTIAL double y[i],x[i] Partial derivative of output y
with respect to input x.
PORT_NULL Mif_Boolean_t name Has this port been specified as
unconnected?
PORT_SIZE int name Size of port vector.
RAD_FREQ double <none> Current analysis frequency in
radians per second.
STATIC_VAR CD name Value of a static variable.
STATIC_VAR_SIZE int name Size of static var vector
(currently unused).
Table 28.3: Accessor macros
T(n) int index Current and previous analysis

times (T(0) = TIME = current
analysis time, T(1) = previous
analysis time). Valid only in
analog calls.
TEMPERATURE double <none> Current analysis temperature.
TIME double <none> Current analysis time (same as
T(0)).
TOTAL_LOAD double name[i] The total of all loads on the
node attached to this event
driven port.
28.7.2 Function Library

28.7.2.1 Overview
Aside from the accessor macros, the simulator also provides a library of functions callable
from within code models. The header file containing prototypes to these functions is au-
tomatically inserted into the Model Definition File for you. The complete list of available
functions follows:
Smoothing Functions:
void cm_smooth_corner
void cm_smooth_discontinuity
double cm_smooth_pwl
Model State Storage Functions:
void cm_analog_alloc
void cm_event_alloc
void *cm_analog_get_ptr
void *cm_event_get_ptr
Integration and Convergence Functions:
int cm_analog_integrate
int cm_analog_converge
void cm_analog_not_converged
void cm_analog_auto_partial
double cm_analog_ramp_factor
Message Handling Functions:
char *cm_message_get_errmsg
void cm_message_send
int cm_message_printf
Breakpoint Handling Functions:
int cm_analog_set_temp_bkpt
int cm_analog_set_perm_bkpt
int cm_event_queue
Special Purpose Functions:
void cm_climit_fcn
double cm_netlist_get_c
double cm_netlist_get_l
char *cm_get_path
const char *cm_get_node_name
bool cm_probe_node
bool cm_schedule_output
bool cm_getvar
void cm_cexit
void cm_irreversible
Complex Math Functions:
complex_t cm_complex_set
complex_t cm_complex_add
complex_t cm_complex_sub
complex_t cm_complex_mult
complex_t cm_complex_div
28.7.2.2 Smoothing Functions

void
cm_smooth_corner(x_input, x_center, y_center, domain,
lower_slope, upper_slope, y_output, dy_dx)
double x_input; /* The value of the x input */

double x_center; /* The x intercept of the two slopes */
double y_center; /* The y intercept of the two slopes */
double domain; /* The smoothing domain */
double lower_slope; /* The lower slope */
double upper_slope; /* The upper slope */
double *y_output; /* The smoothed y output */
double *dy_dx; /* The partial of y wrt x */
void
cm_smooth_discontinuity(x_input, x_lower, y_lower, x_upper, y_upper
y_output, dy_dx)
double x_input; /* The x value at which to compute y */

double x_lower; /* The x value of the lower corner */
double y_lower; /* The y value of the lower corner */
double x_upper; /* The x value of the upper corner */
double y_upper; /* The y value of the upper corner */
double *y_output; /* The computed smoothed y value */
double *dy_dx; /* The partial of y wrt x */
double
cm_smooth_pwl(x_input, x, y, size, input_domain, dout_din)
double x_input; /* The x input value */

double *x; /* The vector of x values */
double *y; /* The vector of y values */
int size; /* The size of the xy vectors */
double input_domain; /* The smoothing domain */
double *dout_din; /* The partial of the output wrt the input */
cm_smooth_corner() automates smoothing between two arbitrarily-sloped lines that meet

at a single center point. You specify the center point (x_center, y_center), plus a do-
main (x-valued delta) above and below x_center. This defines a smoothing region about
the center point. Then, the slopes of the meeting lines outside of this smoothing region
are specified (lower_slope, upper_slope). The function then interpolates a smoothly-
varying output (*y_output) and its derivative (*dy_dx) for the x_input value. This
function helps to automate the smoothing of piecewise-linear functions, for example. Such
smoothing aids the simulator in achieving convergence.
cm_smooth_discontinuity() allows you to obtain a smoothly-transitioning output (*y_output)
that varies between two static values (y_lower, y_upper) as an independent variable
(x_input) transitions between two values (x_lower, x_upper). This function is useful
in interpolating between resistances or voltage levels that change abruptly between two
values.
cm_smooth_pwl() duplicates much of the functionality of the predefined pwl code model.
The cm smooth pwl() takes an input value plus x-coordinate and y-coordinate vector
values along with the total number of coordinate points used to describe the piecewise
linear transfer function and returns the interpolated or extrapolated value of the output
based on that transfer function. More detail is available by looking at the description of
the pwl code model. Note that the output value is the function’s returned value.
28.7.2.3 Model State Storage Functions

void cm_analog_alloc(tag, size)
int tag; /* The user-specified tag for this block of memory */

int size; /* The number of bytes to allocate */
void cm_event_alloc(tag, size)
int tag; /* The user-specified tag for the memory block */

int size; /* The number of bytes to be allocated */
void *cm_analog_get_ptr(tag, timepoint

int tag; /* The user-specified tag for this block of memory */
int timepoint; /* The timepoint of interest - 0=current 1=previous */
void *cm_event_get_ptr(tag, timepoint)
int tag; /* The user-specified tag for the memory block */

int timepoint; /* The timepoint - 0=current, 1=previous */
cm_analog_alloc() and cm_event_alloc() allow you to allocate storage space for analog
and event-driven model state information. The storage space is not static, but rather
represents a storage vector of two values that rotate with each accepted simulator time-
point evaluation. This is explained more fully below. The ‘tag’ parameter allows you to
specify an integer tag when allocating space. This allows more than one rotational storage
location per model to be allocated. The ‘size’ parameter specifies the size in bytes of the
storage (computed by the C language sizeof() operator). Both cm_analog_alloc() and
cm_event_alloc() will not return pointers to the allocated space, as has been available
(and buggy) from the original XSPICE code. cm_analog_alloc() should be used by an
analog model; cm_event_alloc() should be used by an event-driven model.
cm_analog_get_ptr() and cm_event_get_ptr() retrieve the pointer location of the rota-
tional storage space previously allocated by cm_analog_alloc() or cm_event_alloc().
Important notice: These functions must be called only after all memory allocation
(all calls to cm_analog_alloc() or cm_event_alloc()) have been done. All pointers
returned between calls to memory allocation will become obsolete (point to freed memory
because of an internal realloc). The functions take the integer ‘tag’ used to allocate the
space, and an integer from 0 to 1 that specifies the time-point with which the desired
state variable is associated (e.g. timepoint = 0 will retrieve the address of storage for the
current time-point; timepoint = 1 will retrieve the address of storage for the last accepted
time-point). Note that once a model is exited, storage to the current time-point
state storage location (i.e., timepoint = 0) will, upon the next time-point iter-
ation, be rotated to the previous location (i.e., timepoint = 1). When rotation
is done, a copy of the old ‘timepoint = 0’ storage value is placed in the new ‘timepoint
= 1’ storage location. Thus, if a value does not change for a particular iteration, specific
writing to ‘timepoint = 0’ storage is not required. These features allow a model coder
to constantly know which piece of state information is being dealt with within the model
function at each time-point.
Rotation and copying for cm_analog_get_ptr() occurs when the circuit state converges
and the simulator accepts the new timepoint. Rotation and copying are done for cm_event_get_ptr()
before each EVENT call. When simulation time moves backward, newer EVENT data
is discarded and the next copy is made from the oldest data with time-stamp less than
TIME.
28.7.2.4 Integration and Convergence Functions

int cm_analog_integrate(integrand, integral, partial)
double integrand; /* The integrand */

double *integral; /* The current and returned value of integral */
double *partial; /* The partial derivative of integral wrt integrand */
int cm_analog_converge(state)
double *state; /* The state to be converged */
void cm_analog_not_converged()
void cm_analog_auto_partial()
double cm_ramp_factor()
cm_analog_integrate() takes as input the integrand (the input to the integrator)

and produces as output the integral value and the partial of the integral with respect
to the integrand. The integration itself is with respect to time, and the pointer to
the integral value must have been previously allocated using cm_analog_alloc() and
*cm_analog_get_ptr(). This is required because of the need for the integrate routine
itself to have access to previously-computed values of the integral.
cm_analog_converge() takes as an input the address of a state variable that was pre-
viously allocated using cm_analog_alloc() and *cm_analog_get_ptr(). The function
itself serves to notify the simulator that for each time-step taken, that variable must be
iterated upon until it converges.
cm_analog_not_converged() is a function that can and should be called by an analog
model whenever it performs internal limiting of one or more of its inputs to aid in reaching
convergence. This causes the simulator to call the model again at the current time-point
and continue solving the circuit matrix. A new time-point will not be attempted until
the code model returns without calling the cm_analog_not_converged() function. For
circuits that have trouble reaching a converged state (often due to multiple inputs changing
too quickly for the model to react in a reasonable fashion), the use of this function is
virtually mandatory.
cm_analog_auto_partial() may be called at the end of a code model function in lieu of
calculating the values of partial derivatives explicitly in the function. When this function
is called, no values should be assigned to the PARTIAL macro since these values will be
computed automatically by the simulator. The automatic calculation of partial derivatives
can save considerable time in designing and coding a model, since manual computation of
partial derivatives can become very complex and error-prone for some models. However,
the automatic evaluation may also increase simulation run time significantly. Function
cm_analog_auto_partial() causes the model to be called N additional times (for a
model with N inputs) with each input varied by a small amount (1e-6 for voltage inputs
and 1e-12 for current inputs). The values of the partial derivatives of the outputs with
respect to the inputs are then approximated by the simulator through divided difference
calculations.
cm_analog_ramp_factor() will then return a value from 0.0 to 1.0 that indicates whether
or not a ramp time value requested in the SPICE analysis deck (with the use of .option
ramptime=<duration>) has elapsed. If the RAMPTIME option is used, then cm_analog_ramp_factor
returns a 0.0 value during the DC operating point solution and a value that is between 0.0
and 1.0 during the ramp. A 1.0 value is returned after the ramp is over or if the RAMP-
TIME option is not used. This value is intended as a multiplication factor to be used with
all model outputs that would ordinarily experience a ‘power-up’ transition. Currently, all
sources within the simulator are automatically ramped to the ‘final’ time-zero value if a
RAMPTIME option is specified.
28.7.2.5 Message Handling Functions

char *cm_message_get_errmsg()
int cm_message_send(char *msg)
char *msg; /* The message to output. */
int cm_message_printf(char *fmt, ...);
*cm_message_get_errmsg() is a function designed to be used with other library func-

tions to provide a way for models to handle error situations. More specifically, whenever
a library function that returns type int is executed from a model, it will return an integer
value, n. If this value is not equal to zero (0), then an error condition has occurred (like-
wise, functions that return pointers will return a NULL value if an error has occurred).
At that point, the model can invoke *cm_message_get_errmsg to obtain a pointer to an
error message. This can then in turn be displayed to the user or passed to the simulator
interface through the cm_message_send() function. The C code required for this is as
follows:
err = cm_analog_integrate(in, &out, &dout_din);

if (err) {
cm_message_send(cm_message_get_errmsg());
}
else { ...
cm_message_send() sends messages to either the standard output screen or to the sim-
ulator interface, depending on which is in use. The instance name and a new line are
added to the message.
cm_message_printf() formats a message like sprintf() and passes it to cm_message_send().
28.7.2.6 Breakpoint Handling Functions

int cm_analog_set_perm_bkpt(time)
double time; /* The time of the breakpoint to be set */
int cm_analog_set_temp_bkpt(time)
double time; /* The time of the breakpoint to be set */
int cm_event_queue(time)
double time; /* The time of the event to be queued */
cm_analog_set_perm_bkpt() takes as input a time value. This value is posted to the

analog simulator algorithm and is used to force the simulator to choose that value as a
breakpoint at some time in the future. The simulator may choose as the next time-point a
value less than the input, but not greater. Also, regardless of how many time-points pass
before the breakpoint is reached, it will not be removed from posting. Thus, a breakpoint
is guaranteed at the passed time value. Note that a breakpoint may also be set for a time
prior to the current time, but this will result in an error if the posted breakpoint is prior
to the last accepted time (i.e., T(1)).
cm_analog_set_temp_bkpt() takes as input a time value. This value is posted to the
simulator and is used to force the simulator, for the next time-step only, to not exceed the
passed time value. The simulator may choose as the next time-point a value less than the
input, but not greater. In addition, once the next time-step is chosen, the posted value is
removed regardless of whether it caused the break at the given time-point. This function
is useful in the event that a time-point needs to be retracted after its first posting in
order to recalculate a new breakpoint based on new input data (for controlled oscillators,
controlled one-shots, etc), since temporary breakpoints automatically ‘go away’ if not
reposted each time-step. Note that a breakpoint may also be set for a time prior to the
current time, but this will result in an error if the posted breakpoint is prior to the last
accepted time (i.e., T(1)).
It is not certain that a future call will be made with TIME exactly equal to the function
argument, but there will be a close match. Arithmetic rounding may occur and the
simulator may make only one call when requests are very closely spaced. The closeness
depends on the minimum timestep of the simulation.
cm_event_queue() is similar to cm_analog_set_perm_bkpt(), but functions with event-
driven models. When invoked, this function causes the model to be queued for calling
at the specified time, with CALL_TYPE == EVENT. There is no combining of closly-
spaced events. All other details applicable to cm_analog_set_perm_bkpt() apply to this
function as well.
28.7.2.7 Special Purpose Functions

void
cm_climit_fcn(in, in_offset, cntl_upper, cntl_lower, lower_delta, upper_delta,
limit_range, gain, fraction, out_final, pout_pin_final,
pout_pcntl_lower_final, pout_pcntl_upper_final)
double in; /* The input value */

double in-offset; /* The input offset */
double cntl_upper; /* The upper control input value */
double cntl_lower; /* The lower control input value */
double lower_delta; /* The delta from control to limit value */
double upper_delta; /* The delta from control to limit value */
double limit_range; /* The limiting range */
double gain; /* The gain from input to output */
int percent; /* The fraction vs. absolute range flag */
double *out_final; /* The output value */
double *pout_pin_final; /* The partial of output wrt input */
double *pout_pcntl_lower_final; /* The partial of output wrt lower
control input */
double *pout_pcntl_upper:final; /* The partial of output wrt upper
control input */
double cm_netlist_get_c()
double cm_netlist_get_l()
char* cm_get_path()
CKTcircuit *cm_get_circuit()
const char *cm_get_node_name(const char *port_name, unsigned int index)
bool cm_probe_node(conn_index, port_index, value)
unsigned int conn_index; /* Connection index */

unsigned int port_index; /* Port index within connection */
void *value; /* Pointer to event value, input and return */
bool cm_schedule_output(unsigned int conn_index, unsigned int port_index, double

unsigned int conn_index; /* Connection index */
unsigned int port_index; /* Port index within connection */
double delay; /* Delay time, similar to OUTPUT_DELAY */
void *vp; /* Pointer to the event value, like OUTPUT */
bool cm_getvar(char *name, enum cp_types type, void *retval, size_t rsize)
char *name; /* Variable name */
enum cp_types type; /* Type of data requested */
void *retval; /* Buffer for data returned */
size_t rsize; /* Buffer size */
void cm_cexit(int exitcode)
int exitcode; /* Number returned upon exiting */
void cm_irreversible(unsigned int priority)
unsigned int priority; /* Relative priority of code model instance */
cm_climit_fcn() is a very specific function that mimics the behavior of the climit code
model (see the Predefined Models section). In brief, the cm_climit_fcn() takes as input
an in value, an offset, and controlling upper and lower values. Parameter values include
delta values for the controlling inputs, a smoothing range, gain, and fraction switch values.
Outputs include the final value, plus the partial derivatives of the output with respect to
signal input, and both control inputs. These all operate identically to the similarly-named
inputs and parameters of the climit model.
The function performs a limit on the in value, holding it to within some delta of the
controlling inputs, and handling smoothing, etc. The cm_climit_fcn() was originally
used in the ilimit code model to handle much of the primary limiting in that model,
and can be used by a code model developer to take care of limiting in larger models that
require it. See the detailed description of the climit model a for more in-depth description.
cm_netlist_get_c() and cm_netlist_get_l() functions search the analog circuitry
to which their input is connected, and total the capacitance or inductance, respectively,
found at that node. The functions, as they are currently written, assume they are called
by a model that has only one single-ended analog input port.
cm_get_path() fetches the path of the first netlist input file found on the ngspice com-
mand line or in the source command, which ngspice saves to the global variable Infile_Path.
cm_get_circuit() returns a pointer to the (fundamental) ngspice circuit structure.
This allows accessing a wealth of data, as defined by CKTcircuit structure in cktdefs.h.
To build complex custom-built XSPICE-models, access to such parameters (e.g. maxi-
mum step size) may be needed to get reasonable results from a simulation. This may be
necessary when SPICE interacts with an external sensor-simulator and the results of that
external simulator do not have a direct impact on the SPICE circuit. Then, modifying
the maximum step size on the fly may help to improve the simulation results.
cm_get_node_name() returns the name of the node attached to a port. The second
argument is the index for vector ports.
cm_probe_node() performs a speculative resolution of a node attached to a port. Given

a port and an event value, it returns what the value on the attached node would be, if
the port was attempting to output the original value. It is used by the bidi_bridge to
discover how the node is being driven by the other attached digital ports.
cm_schedule_output() queues an output event for a port, with the same effect as as-
signing to the OUTPUT_DELAY and OUPUT macros, but without returning from the
code model function. It allows more than one event to be queued to a port in a single
call. To prevent one event overriding another on the same port they should be queued
with increasing delays.
cm_getvar() obtains the value of a command interpreter variable. It has the same inter-
face as the internal function cp_getvar() as defined in the header file src/include/ngspice/cpex-
tern.h.
cm_cexit() calls function controlled_exit(), which exits the simulator gracefully.
A call to cm_irreversible()has several effects that work together to support code mod-
els that contain a sub-simulation. Such a sub-simulation will usually be irreversible: in
transient analysis it will not store enough data to allow a completed time step to be
abandoned. However, analog simulation in Ngspice frequently abandons time steps so
that they can be retried with a shorter period to achieve convergence. If the inputs to
the sub-simulation are different in the final analog solution and the original attempt, the
sub-simulation may be permanently left in an incorrect state.
A solution to this problem is to delay advancing time in the sub-simulation until Ngspice
is committed to the current timestep. This can work fully only for a single instance, but
under some conditions it may be possible to support several irreversible sub-simulations
in one circuit.
The effects of a call are: the code model instance is added to the hybrids list if it is not
already a member; its position in the list is adjusted using the passed priority; and just
before a time step is accepted a special call is made to the code model. The hybrids
list initially contains all code model instances that have both analog and event ports.
Code models instances on the list receive an event call just before a time step is accepted.
Setting priority to 1 ensures that an instance will be called last, an instance with priority
2 will be called just before that, and so on. Each instance must use a separate priority
value, but the values used need not be consecutive. The final effect of cm_irreversible() is
that when the call is made, CALL_TYPE will have the special value STEP_PENDING,
not the usual value, EVENT.
A code model using cm_irreversible() may schedule an analog breakpoint in such calls, so
that the current time step does not proceed, although it was acceptable to the simulator
core. In that case no futher STEP_PENDING calls are made for the step.
28.7.2.8 Complex Math Functions

Complex_t cm_complex_set (real_part, imag_part)
double real_part; /* The real part of the complex number */

double imag_part; /* The imaginary part of the complex number */
Complex_t cm_complex_add (x, y)

28.8. USER-DEFINED NODE DEFINITION FILE 605
Complex_t x; /* The first operand of x + y */

Complex_t y; /* The second operand of x + y */
Complex_t cm_complex_sub (x, y)
Complex_t x; /* The first operand of x - y (minuend) */

Complex_t y; /* The second operand of x - y (subtrahend) */
Complex_t cm_complex_mult (x, y)
Complex_t x; /* The first operand of x * y */

Complex_t y; /* The second operand of x * y */
Complex_t cm_complex_div (x, y)
Complex_t x; /* The first operand of x / y (dividend) */

Complex_t y; /* The second operand of x / y (divisor) */
cm_complex_set() takes as input two doubles, and converts these to a Complex_t. The
first double is taken as the real part, and the second is taken as the imaginary part of the
resulting complex value.
cm_complex_add(), cm_complex_sub(), cm_complex_mult(), and cm_complex_div()
each take two complex values as inputs and return the result of a complex addition,
subtraction, multiplication, or division, respectively.
28.8 User-Defined Node Definition File
The User-Defined Node Definition File (udnfunc.c) defines the C functions that implement
basic operations on user-defined nodes such as data structure creation, initialization,
copying, and comparison. Unlike the Model Definition File that uses the Code Model
Preprocessor to translate Accessor Macros, the User-Defined Node Definition file is a
pure C language file. This file uses macros to isolate you from data structure definitions,
but the macros are defined in a standard header file (EVTudn.h), and translations are
performed by the standard C Preprocessor.
When you create a directory for a new User-Defined Node, e.g.
/ngspice/src/xspice/icm/xtraevt/new_type/, add a new User-Defined Node Defini-
tion File udnfunc.c (see the example in Chapt. 28.8.3), and place a structure of type
’Evt_Udn_Info_t’ at its bottom.
This structure contains the type name for the node, a description string, and pointers to
each of the functions that define the node. This structure is complete except for a text
string that describes the node type. This string is stubbed out and may be edited by you
if desired.
28.8.1 Macros
Name Type Description

MALLOCED_PTR void * Assign pointer to allocated
structure to this macro
STRUCT_PTR void * A pointer to a structure of the
defined type
STRUCT_PTR_1 void * A pointer to a structure of the
defined type
STRUCT_PTR_2 void * A pointer to a structure of the
defined type
EQUAL Mif_Boolean_t Assign TRUE or FALSE to this
macro according to the results of
structure comparison
INPUT_STRUCT_PTR void * A pointer to a structure of the
defined type
OUTPUT_STRUCT_PTR void * A pointer to a structure of the
defined type
INPUT_STRUCT_PTR_ARRAY void ** An array of pointers to structures
of the defined type
INPUT_STRUCT_PTR_ARRAY_SIZE int The size of the array
STRUCT_MEMBER_ID char * A string naming some part of the
structure
PLOT_VAL double The value of the specified structure
member for plotting purposes
PRINT_VAL char * The value of the specified structure
member for printing purposes
Table 28.4: User-Defined Node Macros
You must code the functions described in the following section using the macros appro-
priate for the particular function. You may elect whether not to provide the optional
functions.
It is an error to use a macro not defined for a function. Note that a review of the sample
directories for the real and int UDN types will make the function usage clearer.
The macros used in the User-Defined Node Definition File to access and assign data
values are defined in Table 28.4. The translations of the macros and of macros used in
the function argument lists are defined in the Interface Design Document for the XSPICE
Simulator.
28.8.2 Function Library

The functions (required and optional) that define a User-Defined Node are listed below.
For optional functions not used, the pointer in the Evt_Udn_Info_t structure can be
changed to NULL.
Required functions:
create Allocate data structure used as inputs and outputs to

code models.
initialize Set structure to appropriate initial value for first use as

model input.
copy Make a copy of the contents into created but possibly

uninitialized structure.
compare Determine if two structures are equal in value.
Optional functions:
dismantle Free allocations inside structure (but not structure itself).
invert Invert logical value of structure.
resolve Determine the resultant when multiple outputs are connected

to a node.
plot_val Output a real value for specified structure component for

plotting purposes.
print_val Output a string value for specified structure component for

printing.
ipc_val Output a binary representation of the structure suitable

for sending over the IPC channel.
The required actions for each of these functions are described in the following subsections.
In each function, you have to replace the XXX with the node type name specified. The
macros used in implementing the functions are described in a later section.
28.8.2.1 Function udn_XXX_create
Allocate space for the data structure defined for the User-Defined Node to pass data
between models. Then assign pointer created by the storage allocator (e.g. malloc) to
MALLOCED_PTR.
28.8.2.2 Function udn_XXX_initialize
Assign STRUCT_PTR to a pointer variable of defined type and then initialize the value
of the structure.
28.8.2.3 Function udn_XXX_compare
Assign STRUCT_PTR_1 and STRUCT_PTR_2 to pointer variables of the defined type.

Compare the two structures and assign either TRUE or FALSE to EQUAL.
28.8.2.4 Function udn_XXX_copy
Assign INPUT_STRUCT_PTR and OUTPUT_STRUCT_PTR to pointer variables of

the defined type and then copy the elements of the input structure to the output structure.
28.8.2.5 Function udn_XXX_dismantle
Assign STRUCT_PTR to a pointer variable of defined type and then free any allocated
substructures (but not the structure itself!). If there are no substructures, the body of
this function may be left null.
28.8.2.6 Function udn_XXX_invert
Assign STRUCT_PTR to a pointer variable of the defined type, and then invert the
logical value of the structure.
28.8.2.7 Function udn_XXX_resolve
Assign INPUT_STRUCT_PTR_ARRAY to a variable declared as an array of pointers

of the defined type - e.g.:
<type> **struct_array;
struct_array = INPUT_STRUCT_PTR_ARRAY;
Then, the number of elements in the array may be determined from the integer valued
INPUT_STRUCT_PTR_ARRAY_SIZE macro.
Assign OUTPUT_STRUCT_PTR to a pointer variable of the defined type. Scan through
the array of structures, compute the resolved value, and assign it into the output structure.
28.8.2.8 Function udn_XXX_plot_val
Assign STRUCT_PTR to a pointer variable of the defined type. Then, access the member
of the structure specified by the string in STRUCT_MEMBER_ID and assign some real
valued quantity for this member to PLOT_VALUE.
28.8.2.9 Function udn_XXX_print_val
Assign STRUCT_PTR to a pointer variable of the defined type. Then, access the member
of the structure specified by the string in STRUCT_MEMBER_ID and assign some string
valued quantity for this member to PRINT_VALUE.
If the string is not static, a new string should be allocated on each call. Do not free the
allocated strings.
28.8.2.10 Function udn_XXX_ipc_val
Use STRUCT_PTR to access the value of the node data. Assign to IPC_VAL a bi-
nary representation of the data. Typically this can be accomplished by simply assigning
STRUCT_PTR to IPC_VAL.
Assign to IPC_VAL_SIZE an integer representing the size of the binary data in bytes.
28.8.3 Example UDN Definition File

The following is an example UDN Definition File that is included with the XSPICE sys-
tem. It illustrates the definition of the functions described above for a User-Defined Node
type int (for integer node type), to be found in file /ngspice/src/xspice/icm/xtraevt/int/udnfunc
# include < stdio .h >

# include " ngspice / cm . h "
# include " ngspice / evtudn . h "
void * tmalloc ( size_t );

# define TMALLOC (t , n ) ( t *) tmalloc ( sizeof ( t )*( size_t )( n ))
/* macro to ignore unused variables and parameters */

# define NG_IGNORE ( x ) ( void ) x
/* * * * * ** * * * * * * ** * * * * * * * * * * * * * ** * * * * * ** * * * * * * ** * * * * * */
static void udn_int_create ( CREATE_ARGS )

{
/* Malloc space for an int */
MALLOCED_PTR = TMALLOC ( int , 1);
}
/* * * * * ** * * * * * * ** * * * * * * * * * * * * * ** * * * * * ** * * * * * * ** * * * * * */
static void udn_int_dismantle ( DISMANTLE_ARGS )

{
NG_IGNORE ( STRUCT_PTR );
/* Do nothing . There are no internally malloc ’ ed
things to dismantle */
}
/* * * * * ** * * * * * * ** * * * * * * * * * * * * * ** * * * * * ** * * * * * * ** * * * * * */
static void udn_int_initialize ( INITIALIZE_ARGS )

{
int * int_struct = ( int *) STRUCT_PTR ;
/* Initialize to zero */
* int_struct = 0;
}
/* * * * * ** * * * * * * ** * * * * * * * * * * * * * ** * * * * * ** * * * * * * ** * * * * * */
static void udn_int_invert ( INVERT_ARGS )

{
/* Invert the state */

* int_struct = -(* int_struct );
}
/* * * * * ** * * * * * * ** * * * * * * * * * * * * * ** * * * * * ** * * * * * * ** * * * * * */
static void udn_int_copy ( COPY_ARGS )

{
int * int_from_struct = ( int *) INPUT_STRUCT_PTR ;
int * int_to_struct = ( int *) OUTPUT_STRUCT_PTR ;
/* Copy the structure */

* int_to_struct = * int_from_struct ;
}
/* * * * * ** * * * * * * ** * * * * * * * * * * * * * ** * * * * * ** * * * * * * ** * * * * * */
static void udn_int_resolve ( RESOLVE_ARGS )

{
int ** array = ( int **) INPUT_STRUCT_PTR_ARRAY ;
int * out = ( int *) OUTPUT_STRUCT_PTR ;
int num_struct = INPUT_STRUCT_PTR_ARRAY_SIZE ;
int sum ;
int i;
/* Sum the values */

for ( i = 0 , sum = 0; i < num_struct ; i ++)
sum += *( array [ i ]);
/* Assign the result */

* out = sum ;
}
/* * * * * ** * * * * * * ** * * * * * * * * * * * * * ** * * * * * ** * * * * * * ** * * * * * */
static void udn_int_compare ( COMPARE_ARGS )

{
int * int_struct1 = ( int *) STRUCT_PTR_1 ;
int * int_struct2 = ( int *) STRUCT_PTR_2 ;
/* Compare the structures */

if ((* int_struct1 ) == (* int_struct2 ))
EQUAL = TRUE ;
else
EQUAL = FALSE ;
}
/* * * * * ** * * * * * * ** * * * * * * * * * * * * * ** * * * * * ** * * * * * * ** * * * * * */
static void udn_int_plot_val ( PLOT_VAL_ARGS )

{
NG_IGNORE ( STRUCT_MEMBER_ID );
/* Output a value for the int struct */

PLOT_VAL = * int_struct ;
}
/* * * * * ** * * * * * * ** * * * * * * * * * * * * * ** * * * * * ** * * * * * * ** * * * * * */
static void udn_int_print_val ( PRINT_VAL_ARGS )

{
NG_IGNORE ( STRUCT_MEMBER_ID );
/* Allocate space for the printed value */

PRINT_VAL = TMALLOC ( char , 30);
/* Print the value into the string */

sprintf ( PRINT_VAL , "%8 d " , * int_struct );
}
/* * * * * ** * * * * * * ** * * * * * * * * * * * * * ** * * * * * ** * * * * * * ** * * * * * */
static void udn_int_ipc_val ( IPC_VAL_ARGS )

{
/* Simply return the structure and its size */
IPC_VAL = STRUCT_PTR ;
IPC_VAL_SIZE = sizeof ( int );
}
Evt_Udn_Info_t udn_int_info = {
" int " ,
" integer valued data " ,
NULL ,
udn_int_create ,
udn_int_dismantle ,
udn_int_initialize ,
udn_int_invert ,
udn_int_copy ,
udn_int_resolve ,
udn_int_compare ,
udn_int_plot_val ,
udn_int_print_val ,
udn_int_ipc_val
};
Chapter 29
Error Messages
Error messages may be subdivided into three categories. These are
1. Error messages generated during the development of a code model (Preprocessor

Error Messages).
2. Error messages generated by the simulator during a simulation run (Simulator Error
Messages).
3. Error messages generated by individual code models (Code Model Error Messages).
These messages will be explained in detail in the following subsections.
29.1 Preprocessor Error Messages

The following is a list of error messages that may be encountered when invoking the
directory-creation and code modeling preprocessor tools. These are listed individually,
and explanations follow the name/listing.
Usage: cmpp [-ifs] [-mod [<filename>]] [-lst]
The Code Model Preprocessor (cmpp) command was invoked incorrectly.
ERROR - Too few arguments
The Code Model Preprocessor (cmpp) command was invoked with too few arguments.
ERROR - Too many arguments
The Code Model Preprocessor (cmpp) command was invoked with too many arguments.
ERROR - Unrecognized argument
613
614 CHAPTER 29. ERROR MESSAGES
The Code Model Preprocessor (cmpp) command was invoked with an invalid argument.
ERROR - File not found: s<filename>
The specified file was not found, or could not be opened for read access.
ERROR - Line <line number> of <filename> exceeds XX characters
The specified line was too long.
ERROR - Pathname on line <line number> of <filename>

exceeds XX characters.
The specified line was too long.
ERROR - No pathnames found in file: <filename>
The indicated modpath.lst file does not have pathnames properly listed.
ERROR - Problems reading ifspec.ifs in directory <pathname>
The Interface Specification File (ifspec.ifs) for the code model could not be read.
ERROR - Model name <model name> is same as internal SPICE model name
A model has been given the same name as an intrinsic SPICE device.
ERROR - Model name ’<model name>’ in directory: <pathname>

is same as
model name ’<model name>’ in directory: <pathname>
Two models in different directories have the same name.
ERROR - C function name ’<function name>’ in directory: <pathname>,

is same as
C function name ’<function name>’ in directory: <pathname>
Two C language functions in separate model directories have the same names; these would
cause a collision when linking the final executable.
ERROR - Problems opening CMextrn.h for write

29.1. PREPROCESSOR ERROR MESSAGES 615
The temporary file CMextern.h used in building the XSPICE simulator executable could
not be created or opened. Check permissions on directory.
ERROR - Problems opening CMinfo.h for write
The temporary file CMinfo.h used in building the XSPICE simulator executable could not
be created or opened. Check permissions on directory.
ERROR - Problems opening objects.inc file for write
The temporary file objects.inc used in building the XSPICE simulator executable could
not be created or opened. Check permissions on directory.
ERROR - Could not open input .mod file: <filename>
The Model Definition File that contains the definition of the Code Model’s behavior
(usually cfunc.mod) was not found or could not be read.
ERROR - Could not open output .c: <filename>
The indicated C language file that the preprocessor creates could not be created or opened.
Check permissions on directory.
Error parsing .mod file: <filename>
Problems were encountered by the preprocessor in interpreting the indicated Model Def-
inition File.
ERROR - File not found: <filename>
The indicated file was not found or could not be opened.
Error parsing interface specification file
Problems were encountered by the preprocessor in interpreting the indicated Interface

Specification File.
ERROR - Can’t create file: <filename>
The indicated file could not be created or opened. Check permissions on directory.
ERROR - write.port.info() - Number of allowed types cannot be zero
There must be at least one port type specified in the list of allowed types.
illegal quoted character in string (expected "\" or "\\")
A string was found with an illegal quoted character in it.
unterminated string literal
A string was found that was not terminated.
Unterminated comment
A comment was found that was not terminated.
Port ’<port name>’ not found
The indicated port name was not found in the Interface Specification File.
Port type ’vnam’ is only valid for ’in’ ports
The port type vnam was used for a port with direction out or inout. This type is only
allowed on in ports.
Port types ’g’, ’gd’, ’h’, ’hd’ are only valid for ’inout’ ports
Port type g, gd, h, or hd was used for a port with direction out or in. These types are
only allowed on inout ports.
Invalid parameter type - POINTER type valid only for STATIC_VARs
The type POINTER was used in a section of the Interface Specification file other than
the STATIC_VAR section.
Port default type is not an allowed type
A default type was specified that is not one of the allowed types for the port.
Incompatible port types in ‘allowed_types’ clause
Port types listed under ‘Allowed_Types’ in the Interface Specification File must all have
the same underlying data type. It is illegal to mix analog and event driven types in a list
of allowed types.
Invalid parameter type (saw <parameter type 1> - expected <parameter type 2>)
29.1. PREPROCESSOR ERROR MESSAGES 617
A parameter value was not compatible with the specified type for the parameter.
Named range not allowed for limits
A name was found where numeric limits were expected.
Direction of port ’<port number>’ in <port name>()

is not <IN or OUT> or INOUT
A problem exists with the direction of one of the elements of a port vector.
Port ’<port name>’ is an array - subscript required
A port was referenced that is specified as an array (vector) in the Interface Specification
File. A subscript is required (e.g. myport[i])
Parameter ’<parameter name>’ is an array - subscript required
A parameter was referenced that is specified as an array (vector) in the Interface Specifi-
cation File. A subscript is required (e.g. myparam[i])
Port ’<port name>’ is not an array - subscript prohibited
A port was referenced that is not specified as an array (vector) in the Interface Specifica-
tion File. A subscript is not allowed.
Parameter ’<parameter name>’ is not an array - subscript prohibited
A parameter was referenced that is not specified as an array (vector) in the Interface
Specification File. A subscript is not allowed.
Static variable ’<static variable name>’ is not an array - subscript prohibited
Array static variables are not supported. Use a POINTER type for the static variable.
Buffer overflow - try reducing the complexity of CM-macro array subscripts
The argument to a code model accessor macro was too long.
Unmatched )
An open ( was found with no corresponding closing ).
Unmatched ]
An open [ was found with no corresponding closing ].

29.2 Simulator Error Messages

The following is a list of error messages that may be encountered while attempting to
run a simulation (with the exception of those error messages generated by individual code
models). Most of these errors are generated by the simulator while attempting to parse a
SPICE deck. These are listed individually, and explanations follow the name/listing.
ERROR - Scalar port expected, [ found
A scalar connection was expected for a particular port on the code model, but the symbol
[ , which is used to begin a vector connection list, was found.
ERROR - Unexpected ]
A ] was found where not expected. Most likely caused by a missing [.
ERROR - Unexpected [ - Arrays of arrays not allowed
A [ character was found within an array list already begun with another [ character.
ERROR - Tilde not allowed on analog nodes
The tilde character ~ was found on an analog connection. This symbol, which performs
state inversion, is only allowed on digital nodes and on User-Defined Nodes only if the
node type definition allows it.
ERROR - Not enough ports
An insufficient number of node connections was supplied on the instance line. Check the
Interface Specification File for the model to determine the required connections and their
types.
ERROR - Expected node/instance identifier
A special token (e.g. [ ] < > ...) was found when not expected.
ERROR - Expected node identifier
A special token (e.g. [ ] < > ...) was found when not expected.
ERROR - unable to find definition of model <name>
A .model line for the referenced model was not found.

29.3. CODE MODEL ERROR MESSAGES 619
ERROR - model: %s - Array parameter expected - No array delimiter found
An array (vector) parameter was expected on the .model card, but enclosing [ ] characters
were not found to delimit its values.
ERROR - model: %s - Unexpected end of model card
The end of the indicated .model line was reached before all required information was
supplied.
ERROR - model: %s - Array parameter must have at least one value
An array parameter was encountered that had no values.
ERROR - model: %s - Bad boolean value
A bad values was supplied for a Boolean. Value used must be TRUE, FALSE, T, or F.
ERROR - model: %s - Bad integer, octal, or hex value
A badly formed integer value was found.
ERROR - model: %s - Bad real value
A badly formed real value was found.
ERROR - model: %s - Bad complex value
A badly formed complex number was found. Complex numbers must be enclosed in < >
delimiters.
29.3 Code Model Error Messages

The following is a list of error messages that may be encountered while attempting to run
a simulation with certain code models. These are listed alphabetically based on the name
of the code model, and explanations follow the name and listing.
29.3.1 Code Model aswitch

cntl_error:
*****ERROR*****
ASWITCH: CONTROL voltage delta less than 1.0e-12
This message occurs as a result of the cntl_off and cntl_on values being less than 1.0e-12
volts/amperes apart.
29.3.2 Code Model climit

climit_range_error:
**** ERROR ****
* CLIMIT function linear range less than zero. *
This message occurs whenever the difference between the upper and lower control input
values are close enough that there is no effective room for proper limiting to occur; this
indicates an error in the control input values.
29.3.3 Code Model core

allocation_error:
***ERROR***
CORE: Allocation calloc failed!
This message is a generic message related to allocating sufficient storage for the H and B
array values.
limit_error:
***ERROR***
CORE: Violation of 50% rule in breakpoints!
This message occurs whenever the input domain value is an absolute value and the H
coordinate points are spaced too closely together (overlap of the smoothing regions will
occur unless the H values are redefined).
29.3.4 Code Model d_osc

d_osc_allocation_error:
**** Error ****
D_OSC: Error allocating VCO block storage
Generic block storage allocation error.
d_osc_array_error:
**** Error ****
D_OSC: Size of control array different than frequency array
Error occurs when there is a different number of control array members than frequency
array members.
d_osc_negative_freq_error:
**** Error ****
D_OSC: The extrapolated value for frequency
has been found to be negative...
Lower frequency level has been clamped to 0.0 Hz.
Occurs whenever a control voltage is input to a model that would ordinarily (given the
specified control/freq coordinate points) cause that model to attempt to generate an
output oscillating at zero frequency. In this case, the output will be clamped to some DC
value until the control voltage returns to a more reasonable value.
29.3.5 Code Model d_source

loading_error:
***ERROR***
D_SOURCE: source.txt file was not read successfully.
This message occurs whenever the d source model has experienced any difficulty in loading
the source.txt (or user-specified) file. This will occur with any of the following problems:
• Width of a vector line of the source file is incorrect.
• A time-point value is duplicated or is otherwise not monotonically increasing.
• One of the output values was not a valid 12-State value (0s, 1s, Us, 0r, 1r, Ur, 0z,
1z, Uz, 0u, 1u, Uu).
29.3.6 Code Model d_state

loading_error:
***ERROR***
D_STATE: state.in file was not read successfully.
The most common cause of this problem is a trailing
blank line in the state.in file
This error occurs when the state.in file (or user-named state machine input file) has not
been read successfully. This is due to one of the following:
• The counted number of tokens in one of the file’s input lines does not equal that
required to define either a state header or a continuation line (Note that all comment
lines are ignored, so these will never cause the error to occur).
• An output state value was defined using a symbol that was invalid (i.e., it was not
one of the following: 0s, 1s, Us, 0r, 1r, Ur, 0z, 1z, Uz, 0u, 1u, Uu).
• An input value was defined using a symbol that was invalid (i.e., it was not one of
the following: 0, 1, X, or x).
index_error:
***ERROR***
D_STATE: An error exists in the ordering of states values
in the states->state[] array. This is usually caused
by non-contiguous state definitions in the state.in file
This error is caused by the different state definitions in the input file being non-contiguous.
In general, it will refer to the different states not being defined uniquely, or being ‘broken
up’ in some fashion within the state.in file.
29.3.7 Code Model oneshot

oneshot_allocation_error:
**** Error ****
ONESHOT: Error allocating oneshot block storage
Generic storage allocation error.
oneshot_array_error:
**** Error ****
ONESHOT: Size of control array different than pulse-width array
This error indicates that the control array and pulse-width arrays are of different sizes.
oneshot_pw_clamp:
**** Warning ****
ONESHOT: Extrapolated Pulse-Width Limited to zero
This error indicates that for the current control input, a pulse-width of less than zero is
indicated. The model will consequently limit the pulse width to zero until the control
input returns to a more reasonable value.
29.3.8 Code Model pwl

allocation_error:
***ERROR***
PWL: Allocation calloc failed!
limit_error:
***ERROR***
PWL: Violation of 50% rule in breakpoints!
This error message indicates that the pwl model has an absolute value for its input domain,
and that the x_array coordinates are so close together that the required smoothing regions
would overlap. To fix the problem, you can either spread the x_array coordinates out or
make the input domain value smaller.
29.3.9 Code Model s_xfer

num_size_error:
***ERROR***
S_XFER: Numerator coefficient array size greater than
denominator coefficient array size.
This error message indicates that the order of the numerator polynomial specified is
greater than that of the denominator. For the s_xfer model, the orders of numerator
and denominator polynomials must be equal, or the order of the denominator polynomial
must be greater than that or the numerator.
29.3.10 Code Model sine

allocation_error:
**** Error ****
SINE: Error allocating sine block storage
sine_freq_clamp:
**** Warning ****
SINE: Extrapolated frequency limited to 1e-16 Hz
This error occurs whenever the controlling input value is such that the output frequency
ordinarily would be set to a negative value. Consequently, the output frequency has been
clamped to a near-zero value.
array_error:
**** Error ****
SINE: Size of control array different than frequency array
This error message normally occurs whenever the controlling input array and the fre-
quency array are different sizes.
29.3.11 Code Model square

square_allocation_error:
**** Error ****
SQUARE: Error allocating square block storage
square_freq_clamp:
**** WARNING ****
SQUARE: Frequency extrapolation limited to 1e-16
square_array_error:
**** Error ****
SQUARE: Size of control array different than frequency array
29.3.12 Code Model triangle

triangle_allocation_error:
**** Error ****
TRIANGLE: Error allocating triangle block storage
triangle_freq_clamp:
**** Warning ****
TRIANGLE: Extrapolated Minimum Frequency Set to 1e-16 Hz
triangle_array_error:
**** Error ****
TRIANGLE: Size of control array different than frequency array
Part III
CIDER
625
Chapter 30
CIDER User’s Manual
The CIDER User’s Manual that follows is derived from the original manual being part of
the PhD thesis from David A. Gates from UC Berkeley. Unfortunately the manual here
is not yet complete, so please refer to the thesis for detailed information. Literature on
CODECS, the predecessor of CIDER, is available here from UCB: TechRpt ERL-90-96
and TechRpt ERL-88-71.
30.1 SPECIFICATION
Overview of numerical-device specification
The input to CIDER consists of a SPICE-like description of a circuit, its analyses and
its compact device models, and PISCES-like descriptions of numerically analyzed device
models. For a description of the SPICE input format, consult the SPICE3 Users Manual
[JOHN92]. The KLU matrix solver (15.1.1) is not supported.
To simulate devices numerically, two types of input must be added to the input file. The
first is a model description in which the common characteristics of a device class are
collected. In the case of numerical models, this provides all the information needed to
construct a device cross-section, such as, for example, the doping profile. The second type
of input consists of one or more element lines that specify instances of a numerical model,
describe their connection to the rest of the circuit, and provide additional element-specific
information such as device layout dimensions ans initial bias information.
The format of a numerical device model description differs from the standard approach
used for SPICE3 compact models. It begins the same way with one line containing the
.MODEL keyword followed by the name of the model, device type and modeling level.
However, instead of providing a single long list of parameters and their values, numerical
model parameters are grouped onto cards. Each type of card has its own set of valid
parameters. In all cases, the relative ordering of different types of cards is unimportant.
However, for cards of the same type (such as mesh-specification cards), their order in the
input file can be important in determining the device structure.
Each card begins on a separate line of the input file. In order to let CIDER know that
card lines are continuations of a numerical model description, each must begin with the
continuation character ‘+’. If there are too many parameters on a given card to allow it
fit on a single line, the card can be continued by adding a second ‘+’ to the beginning of
627
628 CHAPTER 30. CIDER USER’S MANUAL
the next line. However, the name and value of a parameter should always appear on the
same line.
Several features are provided to make the numerical model format more convenient.
Blank space can follow the initial ‘+’ to separate it from the name of a card or the
card continuation ‘+’. Blank lines are also permitted, as long as they also begin with an
initial ‘+’. Parentheses and commas can be used to visually group or separate parameter
definitions. In addition, while it is common to add an equal sign between a parameter
and its value, this is not strictly necessary.
The name of any card can be abbreviated, provided that the abbreviation is unique. Pa-
rameter name abbreviations can also be used if they are unique in the list of a card’s
parameters. Numeric parameter values are treated identically as in SPICE3, so expo-
nential notation, engineering scale factors and units can be attached to parameter val-
ues: tau=10ns, nc=3.0e19cm^-3. In SPICE3, the value of a FLAG model parameter is
changed to TRUE simply by listing its name on the model line. In CIDER, the value of
a numerical model FLAG parameter can be turned back to FALSE by preceding it by
a caret ‘^’. This minimizes the amount of input change needed when features such as
debugging are turned on and off. In certain cases it is necessary to include file names in
the input description and these names may contain capital letters. If the file name is part
of an element line, the inout parser will convert these capitals to lowercase letters. To
protect capitalization at any time, simply enclose the string in double quotes ‘”’.
The remainder of this manual describes how numerically analyzed elements and models
can be used in CIDER simulations. The manual consists of three parts. First, all of the
model cards and their parameters are described. This is followed by a section describing
the three basic types of numerical models and their corresponding element lines. In the
final section, several complete examples of CIDER simulations are presented.
Several conventions are used in the card descriptions. In the card synopses, the name of a
card is followed by a list of parameter classes. Each class is represented by a section in the
card parameter table, in the same order as it appears in the synopsis line. Classes that
contain optional parameters are surrounded by brackets: [...]. Sometimes it only makes
sense for a single parameter to take effect. (For example, a material can not simultaneously
be both Si and SiO2.) In such cases, the various choices are listed sequentially, separated
by colons. The same parameter often has a number of different acceptable names, some
of which are listed in the parameter tables.1 These aliases are separated by vertical bars:
‘|’. Finally, in the card examples, the model continuation pluses have been removed from
the card lines for clarity’s sake.
30.1.1 Examples
The model description for a two-dimensional numerical diode might look something like
what follows. This example demonstrates many of the features of the input format de-
1
Some of the possibilities are not listed in order to shorten the lengths of the parameter tables.
This makes the use of parameter abbreviations somewhat troublesome since an unlisted parameter may
abbreviate to the same name as one that is listed. CIDER will produce a warning when this occurs.
Many of the undocumented parameter names are the PISCES names for the same parameters. The
adventurous soul can discover these names by delving through the ‘cards’ directory of the source code
distribution looking for the C parameter tables.
30.2. BOUNDARY, INTERFACE 629
scribed above. Notice how the .MODEL line and the leading pluses form a border around
the model description:
Example: Numerical diode
. MODEL M_NUMERICAL NUPD LEVEL =2

+ cardnamel numberl = val1 ( number2 val2 ) , ( number3 = val3 )
+ cardname2 numberl = val1 string1 = name1
+
+ cardname3 numberl = val1 , flag1 , ^ flag2
+ + number2 = val2 , flag3
The element line for an instance of this model might look something like the following.
Double quotes are used to protect the file name from decapitalization:
dl 1 2 M_NUMERICAL area = lOOpm ^2 ic . file = " diode . IC "
30.2 BOUNDARY, INTERFACE

Specify properties of a domain boundary or the interface between two boundaries.
SYNOPSIS
boundary domain [ bounding - box ] [ properties ]

interface domain neighbor [ bounding - box ] [ properties ]
30.2.1 DESCRIPTION
The boundary and interface cards are used to set surface physics parameters along the
boundary of a specified domain. Normally, the parameters apply to the entire boundary,
but there are two ways to restrict the area of interest. If a neighboring domain is also
specified, the parameters are only set on the interface between the two domains. In
addition, if a bounding box is given, only that portion of the boundary or interface inside
the bounding box will be set.
If a semiconductor-insulator interface is specified, then an estimate of the width of any
inversion or accumulation layer that may form at the interface can be provided. If the
surface mobility model (cf. models card) is enabled, then the model will apply to all
semiconductor portions of the device within this estimated distance of the interface. If a
point lies within the estimated layer width of more than one interface, it belong to the
interface specified first in the input file. If the layer width given is less than or equal
to zero, it is automatically replaced by an estimate calculated from the doping near the
interface. As a consequence, if the doping varies so will the layer width estimate.
Each edge of the bounding box can be specified in terms of its location or its mesh-index
in the relevant dimension, or defaulted to the respective boundary of the simulation mesh.
30.2.2 PARAMETERS
Name Type Description Units
Domain Integer ID number of primary domain
Neighbor Integer ID number of neighboring domain
X.Low Real Lowest X location of bounding box µm
: IX.Low Integer Lowest X mesh-index of bounding box
X.High Real Highest X location of bounding box µm
: IX.High Integer Highest X mesh-index of bounding box
Y.Low Real Lowest Y location of bounding box µm
: IY.Low Integer Lowest Y mesh-index of bounding box
Y.High Real Highest Y location of bounding box µm
: IY.High Integer Highest Y mesh-index of bounding box
Qf Real Fixed interface charge C/cm2
SN Real Surface recombination velocity - electrons cm/s
SP Real Surface recombination velocity - holes cm/s
Layer.Width Real Width of surface layer µm
30.2.3 EXAMPLES
The following shows how the surface recombination velocities at an Si-SiO2 interface
might be set:
interface dom = l neigh =2 sn = l . Oe4 sp = l . Oe4
In a MOSFET with a 2.0µm gate width and 0.1µm source and drain overlap, the surface
channel can be restricted to the region between the metallurgical junctions and within
100Ȧ (0.01 µm) of the interface:
interface dom = l neigh =2 x . l = l . l x . h =2.9 layer . w =0.01
The inversion layer width in the previous example can be automatically determined by
setting the estimate to 0.0:
interface dom = l neigh =% x . l = l . l x . h =2.9 layer . w =0.0
30.3 COMMENT
Add explanatory comments to a device definition.
SYNOPSIS
comment [ text ]
* [ text ]
$ [ text ]
# [ text ]
30.4. CONTACT 631
30.3.1 DESCRIPTION
Annotations can be added to a device definition using the comment card. All text on a
comment card is ignored. Several popular commenting characters are also supported as
aliases: ‘*’ from SPICE, ‘$’ from PISCES, and ‘#’ from Linux shell scripts.
30.3.2 EXAMPLES
A SPICE-like comment is followed by a PISCES-like comment and shell script comment:
* CIDER and SPICE would ignore this input line

$ CIDER and PISCES would ignore this , but SPICE wouldn ’ t
# CIDER and Linux Shell scripts would ignore this input line
30.4 CONTACT
Specify properties of an electrode
SYNOPSIS
contact number [ workfunction ]
30.4.1 DESCRIPTION
The properties of an electrode can be set using the contact card. The only changeable
property is the work-function of the electrode material and this only affects contacts made
to an insulating material. All contacts to semiconductor material are assumed to be ohmic
in nature.
30.4.2 PARAMETERS
Number Integer ID number of the electrode
Work-function Real Work-function of electrode material. ( eV )
30.4.3 EXAMPLES
The following shows how the work-function of the gate contact of a MOSFET might be
changed to a value appropriate for a P+ polysilicon gate:
contact num =2 workf =5.29

30.4.4 SEE ALSO

electrode, material
30.5 DOMAIN, REGION

Identify material-type for section of a device
SYNOPSIS
domain number material [ position ]

region number material [ position ]
30.5.1 DESCRIPTION
A device is divided into one or more rectilinear domains, each of which has a unique
identification number and is composed of a particular material.
Domain (aka region) cards are used to build up domains by associating a material type
with a box-shaped section of the device. A single domain may be the union of multiple
boxes. When multiple domain cards overlap in space, the one occurring last in the input
file will determine the ID number and material type of the overlapped region.
Each edge of a domain box can be specified in terms of its location or mesh-index in the
relevant dimension, or defaulted to the respective boundary of the simulation mesh.
30.5.2 PARAMETERS
Number Integer ID number of this domain
Material Integer ID number of material used by this domain
X.Low Real Lowest X location of domain box, ( µm )
: IX.Low Integer Lowest X mesh-index of domain box
X.High Real Highest X location of domain box, ( µm )
: IX-High Integer Highest X mesh-index of domain box
Y.Low Real Lowest Y location of domain box, ( µm )
: IY.Low Integer Lowest Y mesh-index of domain box
Y.High Real Highest Y location of domain box, ( µm )
: IY.High Integer Highest Y mesh-index of domain box
30.5.3 EXAMPLES
Create a 4.0 pm wide by 2.0 pm high domain out of material #1:
domain num = l material = l x . l = O . O x . h =4.0 y . l = O . O y . h =2.0

30.6. DOPING 633
The next example defines the two domains that would be typical of a planar MOSFET
simulation. One occupies all of the mesh below y = 0 and the other occupies the mesh
above y = 0. Because the x values are left unspecified, the low and high x boundaries
default to the edges of the mesh:
domain n = l m = l y . l = O . O
domain n =2 m =2 y . h = O . O
30.5.4 SEE ALSO

x.mesh, material
30.6 DOPING
Add dopant to regions of a device
SYNOPSIS
doping [ domains ] profile - type [ lateral - profile - type ] [ axis ]

[ impurity - type1 [ constant - box ] [ profile - specifications ]
30.6.1 DESCRIPTION
Doping cards are used to add impurities to the various domains of a device. Initially each
domain is dopant-free. Each new doping card creates a new doping profile that defines
the dopant concentration as a function of position. The doping at a particular location is
then the sum over all profiles of the concentration values at that position. Each profile can
be restricted to a subset of a device’s domains by supplying a list of the desired domains.
Otherwise, all domains are doped by each profile.
A profile has uniform concentration inside the constant box. Outside this region, it varies
according to the primary an lateral profile shapes. In 1D devices the lateral shape is
unused and in 2D devices the y-axis is the default axis for the primary profile. Several
analytic functions can be used to define the primary profile shape. Alternatively, empirical
or simulated profile data can be extracted from a file. For the analytic profiles, the doping
is the product of a profile function (e.g. Gaussian) and a reference concentration, which
is either the constant concentration of a uniform profile, or the peak concentration for
any of the other functions. If concentration data is used instead take from an ASCII file
containing a list of location-concentration pairs or a SUPREM3 exported file, the name of
the file must be provided. If necessary, the final concentration at a point is then found by
multiplying the primary profile concentration by the value of the lateral profile function
at that point. Empirical profiles must first be normalized by the value at 0.0 to provide a
usable profile functions. Alternatively, the second dimension can be included by assigning
the same concentration to all points equidistant from the edges of the constant box. The
contours of the profile are the circular.
Figure 30.1: 1D doping profiles with location > 0.
Unless otherwise specified, the added impurities are assumes to be N type. However, the
name of a specific dopant species is needed when extracting concentration information for
that impurity from a SUPREM3 exported file.
Several parameters are used to adjust the basic shape of a profile functions so that the
final, constructed profile, matches the doping profile in the real device. The constant box
region should coincide with a region of constant concentration in the device. For uniform
profiles its boundaries default to the mesh boundaries. For the other profiles the constant
box starts as a point and only acquires width or height if both the appropriate edges are
specified. The location of the peak of the primary profile can be moved away from the
edge of the constant box. A positive location places the peak outside the constant box
(cf. Fig. 30.1), and a negative value puts it inside the constant box (cf. Fig. 30.2).
The concentration in the constant box is then equal to the value of the profile when it
intersects the edge of the constant box. The argument of the profile function is a distance
expressed in terms of the characteristic length (by default equal to 1µm). The longer this
length, the more gradually the profile will change. For example, in Fig. 30.1 and Fig.
30.2, the profiles marked (a) have characteristic lengths twice those of the profiles marked
(b). The location and characteristic length for the lateral profile are multiplied by the
lateral ratio. This allows the use of different length scales for the primary and lateral
profiles. For rotated profiles, this scaling is taken into account, and the profile contours
30.6. DOPING 635
Figure 30.2: 1D doping profiles with location < 0.

are elliptical rather than circular.
30.6.2 PARAMETERS
Domains Int List List of domains to dope
Uniform : Flag Primary profile type
Linear :
Erfc :
Exponential :
Suprem3 :
Ascii :
Ascii Suprem3
InFile String Name of Suprem3, Ascii or Ascii Suprem3 input file
Lat.Rotate : Flag Lateral profile type
Lat.Unif :
Lat.Lin :
Lat.Gauss :
Lat.Erfc :
Lat.Exp
X.Axis:Y.Axis Flag Primary profile direction
N.Type : P.Type : Flag Impurity type
Donor : Acceptor :
Phosphorus :
Arsenic :
Antimony :
Boron
X.Low Real Lowest X location of constant box, (µm)
X.High Real Highest X location of constant box, (µm)
Y.Low Real Lowest Y location of constant box, (µm)
Y.High Real Highest Y location of constant box, (µm)
Conic | Peak.conic Real Dopant concentration, (cm−3 )
Location | Range Real Location of profile edge/peak, (µm)
Char.Length Real Characteristic length of profile, (µm)
Ratio.Lat Real Ratio of lateral to primary distances
30.6.3 EXAMPLES
This first example adds a uniform background P-type doping of 1.0 × 1016 cm−3 to an
entire device:
doping uniform p . type conc = l .0 el6
A Gaussian implantation with rotated lateral falloff, such as might be used for a MOSFET
source, is then added:
30.7. ELECTRODE 637
doping gauss lat . rotate n . type conc = l .0 el9

+ x . l =0.0 x . h =0.5 y . l =0.0 y . h =0.2 ratio =0.7
Alternatively, an error-function falloff could be used:
doping gauss lat . erfc conc = l .0 el9

+ x . l =0.0 x . h =0.5 y . l =0.0 y . h =0.2 ratio =0.7
Finally, the MOSFET channel implant is extracted from an ASCII-format SUPREM3

file. The lateral profile is uniform, so that the implant is confined between X = 1µm and
X = 3µm. The profile begins at Y = 0µm (the high Y value defaults equal to the low Y
value):
doping ascii suprem3 infile = implant . s3 lat . unif boron

+ x . l =1.0 x . h =3.0 y . l =0.0
30.6.4 SEE ALSO

domain, mobility, contact, boundary
30.7 ELECTRODE
Set location of a contact to the device
SYNOPSIS
electrode [ number ] [ position ]
30.7.1 DESCRIPTION
Each device has several electrodes that are used to connect the device to the rest of the
circuit. The number of electrodes depends on the type of device. For example, a MOSFET
needs 4 electrodes. A particular electrode can be identified by its position in the list of
circuit nodes on the device element line. For example, the drain node of a MOSFET
is electrode number 1, while the bulk node is electrode number 4. Electrodes for which
an ID number has not been specified are assigned values sequentially in the order they
appear in the input file.
For lD devices, the positions of two of the electrodes are predefined to be at the ends
of the simulation mesh. The first electrode is at the low end of the mesh, and the last
electrode is at the high end. The position of the special lD BJT base contact is set on
the options card. Thus, electrode cards are used exclusively for 2D devices.
Each card associates a portion of the simulation mesh with a particular electrode. In
contrast to domains, which are specified only in terms of boxes, electrodes can also be
specified in terms of line segments. Boxes and segments for the same electrode do not
have to overlap. If they don’t, it is assumed that the electrode is wired together outside
the area covered by the simulation mesh. However, pieces of different electrodes must not
overlap, since this would represent a short circuit. Each electrode box or segment can
be specified in terms of the locations or mesh-indices of its boundaries. A missing value
defaults to the corresponding mesh boundary.
30.7.2 PARAMETERS

Number Integer ID number of this domain
X.Low Real Lowest X location of electrode, (µm)
: IX.Low Integer Lowest X mesh-index of electrode
X.High Real Highest X location of electrode, (µm)
: IX.High Integer Highest X mesh-index of electrode
Y.Low Real Lowest Y location of electrode, (µm)
: IY.Low Integer Lowest Y mesh-index of electrode
Y.High Real Highest Y location of electrode, (µm)
: IY.High Integer Highest Y mesh-index of electrode
30.7.3 EXAMPLES
The following shows how the four contacts of a MOSFET might be specified:
* DRAIN
electrode x . l =0.0 x . h =0.5 y . l =0.0 y . h =0.0
* GATE
electrode x . l =1.0 x . h =3.0 iy . l =0 iy . h =0
* SOURCE
* BULK
The numbering option can be used when specifying bipolar transistors with dual base
contacts:
* EMITTER
electrode num =3 x . l =1.0 x . h =2.0 y . l =0.0 y . h =0.0
* BASE
* COLLECTOR
30.8. END 639
30.7.4 SEE ALSO
domain, contact
30.8 END
Terminate processing of a device definition
SYNOPSIS
end
30.8.1 DESCRIPTION
The end card stops processing of a device definition. It may appear anywhere within a
definition. Subsequent continuation lines of the definition will be ignored. If no end card
is supplied, all the cards will be processed.
30.9 MATERIAL
Specify physical properties of a material
SYNOPSIS
material number type [ physical - constants ]
30.9.1 DESCRIPTION
The material card is used to create an entry in the list of materials used in a device. Each
entry needs a unique identification number and the type of the material. Default values
are assigned to the physical properties of the material. Most material parameters are
accessible either here or on the mobility or contact cards. However, some parameters
remain inaccessible (e.g. the ionization coefficient parameters). Parameters for most
physical effect models are collected here. Mobility parameters are handled separately by
the mobility card. Properties of electrode materials are set using the contact card.
30.9.2 PARAMETERS

Number Integer ID number of this material
Semiconductor : Silicon Flag Type of this material
: Polysilicon : GaAs
: Insulator : Oxide
: Nitride
Affinity Real Electron affinity (eV)
Permittivity Real Dielectric permittivity (F/cm)
Nc Real Conduction band density (cm−3 )
Nv Real Valence band density (cm−3 )
Eg Real Energy band gap (eV)
dEg.dT Real Bandgap narrowing with temperature (eV /◦ K )
Eg.Tref Real Bandgap reference temperature, ( °K )
dEg.dN Real Bandgap narrowing with N doping, (eV /cm−3 )
Eg.Nref Real Bandgap reference concentration - N type, (cm−3 )
dEg.dP Real Bandgap narrowing with P doping, (eV /cm−3 )
Eg.Pref Real Bandgap reference concentration - P type, (cm−3 )
TN Real SRH lifetime - electrons, (sec)
SRH.Nref Real SRH reference concentration - electrons (cm−3 )
TP Real SRH lifetime - holes, (sec)
SRH.Pref Real SRH reference concentration - holes (cm−3 )
CN Real Auger coefficient - electrons (cm6/sec)
CP Real Auger coefficient - holes (cm6/sec)
2
ARichN Real Richardson constant - electrons, ( A/ ◦cm K2
)
2
ARichP Real Richardson constant - holes, ( A/ ◦cm
K2
)
30.9.3 EXAMPLES
Set the type of material #1 to silicon, then adjust the values of the temperature-dependent
bandgap model parameters:
material num =1 silicon eg =1.12 deg . dt =4.7 e -4 eg . tref =640.0
The recombination lifetimes can be set to extremely short values to simulate imperfect
semiconductor material:
material num =2 silicon tn =1 ps tp =1 ps
30.9.4 SEE ALSO
domain, mobility, contact, boundary

30.10. METHOD 641
30.10 METHOD
Choose types and parameters of numerical methods
SYNOPSIS
method [ types ] [ parameters ]
30.10.1 DESCRIPTION
The method card controls which numerical methods are used during a simulation and the
parameters of these methods. Most of these methods are optimizations that reduce run
time, but may sacrifice accuracy or reliable convergence.
For majority-carrier devices such as MOSFETs, one carrier simulations can be used to
save simulation time. The systems of equations in AC analysis may be solved using either
direct or successive-over-relaxation techniques. Successive-over-relaxation is faster, but
at high frequencies, it may fail to converge or may converge to the wrong answer. In
some cases, it is desirable to obtain AC parameters as functions of DC bias conditions.
If necessary, a one-point AC analysis is performed at a predefined frequency in order to
obtain these small-signal parameters. The default for this frequency is 1 Hz. The Jacobian
matrix for DC and transient analyses can be simplified by ignoring the derivatives of the
mobility with respect to the solution variables. However, the resulting analysis may have
convergence problems. Additionally, if they are ignored during AC analyses, incorrect
results may be obtained.
A damped Newton method is used as the primary solution technique for the device-level
partial differential equations. This algorithm is based on an iterative loop that terminates
when the error in the solution is small enough or the iteration limit is reached. Error
tolerances are used when determining if the error is ‘small enough’. The tolerances are
expressed in terms of an absolute, solution-independent error and a relative, solution-
dependent error. The absolute-error limit can be set on this card. The relative error is
computed by multiplying the size of the solution by the circuit level SPICE parameter
RELTOL.
30.10.2 Parameters

OneCarrier Flag Solve for majority carriers only
AC analysis String AC analysis method, ( either DIRECT or SOR)
NoMobDeriv Flag Ignore mobility derivatives
Frequency Real AC analysis frequency, ( Hz )
ItLim Integer Newton iteration limit
DevTol Real Maximum residual error in device equations
30.10.3 Examples
Use one carrier simulation for a MOSFET, and choose direct method AC analysis to
ensure accurate, high frequency results:
method onec ac . an = direct
Tolerate no more than 10−10 as the absolute error in device-level equations, and perform
no more than 15 Newton iterations in any one loop:
method devtol =1 e -10 itlim =15
30.11 Mobility
Specify types and parameters of mobility models
SYNOPSIS
mobility material [ carrier ] [ parameters ] [ models ] [ initialize ]
30.11.1 Description
The mobility model is one of the most complicated models of a material’s physical prop-
erties. As a result, separate cards are needed to set up this model for a given material.
Mobile carriers in a device are divided into a number of different classes, each of which has
different mobility modeling. There are three levels of division. First, electrons and holes
are obviously handled separately. Second, carriers in surface inversion or accumulation
layers are treated differently than carriers in the bulk. Finally, bulk carriers can be either
majority or minority carriers.
For surface carriers, the normal-field mobility degradation model has three user-modifiable
parameters. For bulk carriers, the ionized impurity scattering model has four controllable
parameters. Different sets of parameters are maintained for each of the four bulk carrier
types: majority-electron, minority-electron, majority-hole and minority-hole. Velocity
saturation modeling can be applied to both surface and bulk carriers. However, only two
sets of parameters are maintained: one for electrons and one for holes. These must be
changed on a majority carrier card (i.e. when the majority flag is set).
Several models for the physical effects are available, along with appropriate default values.
Initially, a universal set of default parameters usable with all models is provided. These
can be overridden by defaults specific to a particular model by setting the initialization
flag. These can then be changed directly on the card itself. The bulk ionized impurity
models are the Caughey-Thomas (CT) model and the Scharfetter-Gummel (SG) model
[CAUG671, [SCHA69]. Three alternative sets of defaults are available for the Caughey-
Thomas expression. They are the Arora (AR) parameters for Si [AROR82], the University
of Florida (UF) parameters for minority carriers in Si [SOLL90], and a set of parameters
30.11. MOBILITY 643
appropriate for GaAs (GA). The velocity-saturation models are the Caughey-Thomas
(CT) and Scharfetter-Gummel (SG) models for Si, and the PISCES model for GaAs
(GA). There is also a set of Arora (AR) parameters for the Caughey-Thomas model.
30.11.2 Parameters

Material Integer ID number of material
Electron : Hole Flag Mobile carrier
Majority : Minority Flag Mobile carrier type
MUS Real Maximum surface mobility, ( cm2/Vs )
EC.A Real Surface mobility 1st-order critical field, ( V/cm )
EC.B Real Real Surface mobility 2nd-order critical field, ( V2/cm2 )
MuMax Real Maximum bulk mobility, ( cm2/Vs )
MuMin Real Minimum bulk mobility, ( cm2/Vs)
NtRef Real Ionized impurity reference concentration, ( cm-3 )
NtExp Real Ionized impurity exponent
Vsat Real Saturation velocity, ( cm/s )
Vwarm Real Warm carrier reference velocity, ( cm/s )
ConcModel String Ionized impurity model, ( CT, AR, UF, SG, Dr GA )
FieldModel String Velocity saturation model, ( CT, AR, SG, or GA )
Init Flag Copy model-specific defaults
30.11.3 Examples
The following set of cards completely updates the bulk mobility parameters for material
#1:
mobility mat = l concmod = sg fieldmod = sg

mobility mat = l elec major mumax =1000.0 mumin = l00 .0
+ ntref = l .0 el6 ntexp =0.8 vsat = l .0 e7 vwarm =3.0 e6
mobility mat = l elec minor mumax =1000.0 mumin =200. O
+ ntref = l .0 el7 ntexp =0.9
mobility mat = l hole major mumax =500.0 mumin =50.0
+ ntref = l .0 el6 ntexp =0.7 vsat =8.0 e6 vwarm = l .0 e6
mobility mat = l hole minor mumax =500.0 mumin =150.0
+ ntref = l .0 el7 ntexp =0.8
The electron surface mobility is changed by the following:
mobility mat = l elec mus =800.0 ec . a =3.0 e5 ec . b =9.0 e5
Finally, the default Scharfetter-Gummel parameters can be used in Si with the GaAs
velocity-saturation model (even though it doesn’t make physical sense!):
mobility mat = l init elec major fieldmodel = sg

mobility mat = l init hole major fieldmodel = sg
mobility mat = l fieldmodel = ga
30.11.4 SEE ALSO
material
30.11.5 BUGS
The surface mobility model does not include temperature-dependence for the transverse-
field parameters. Those parameters will need to be adjusted by hand.
30.12 MODELS
Specify which physical models should be simulated

SYNOPSIS
models [ model flags ]
30.12.1 DESCRIPTION
The models card indicates which physical effects should be modeled during a simulation.
Initially, none of the effects are included. A flag can be set false by preceding by a caret.
30.12.2 Parameters
BGN Flag Bandgap narrowing
SRH Flag Shockley-Reed-Hall recombination
ConcTau Flag Concentration-dependent SRH lifetimes
Auger Flag Auger recombination
Avalanche Flag Local avalanche generation
TempMob Flag Temperature-dependent mobility
ConcMob Flag Concentration-dependent mobility
FieldMob Flag Lateral-field-dependent mobility
TransMob Flag Transverse-field-dependent surface mobility
SurfMob Flag Activate surface mobility model
30.13. OPTIONS 645
30.12.3 Examples
Turn on bandgap narrowing, and all of the generation-recombination effects:
models bgn srh conctau auger aval
Amend the first card by turning on lateral- and transverse-field-dependent mobility in

surface charge layers, and lateral-field-dependent mobility in the bulk. Also, this line
turns avalanche generation modeling off.
models surfmob transmob fieldmob ^ aval
30.12.4 See also

material, mobility
30.12.5 Bugs
The local avalanche generation model for 2D devices does not compute the necessary
contributions to the device-level Jacobian matrix. If this model is used, it may cause
convergence difficulties and it will cause AC analyses to produce incorrect results.
30.13 OPTIONS
Provide optional device-specific information
SYNOPSIS
options [ device - type ] [ initial - state ] [ dimensions ]

[ measurement - temperature ]
30.13.1 DESCRIPTION
The options card functions as a catch-all for various information related to the circuit-
device interface. The type of a device can be specified here, but will be defaulted if
none is given. Device type is used primarily to determine how to limit the changes in
voltage between the terminals of a device. It also helps determine what kind of boundary
conditions are used as defaults for the device electrodes.
A previously calculated state, stored in the named initial-conditions file, can be loaded at
the beginning of an analysis. If it is necessary for each instance of a numerical model to
start in a different state, then the unique flag can be used to generate unique filenames
for each instance by appending the instance name to the given filename. This is the same
method used by CIDER to generate unique filenames when the states are originally saved.
If a particular state file does not fit. this pattern, the filename can be entered directly on
the instance line.
Mask dimension defaults can be set so that device sizes can be specified in terms of
area or width. Dimensions for the special lD BJT base contact can also be controlled.
The measurement temperature of material parameters, normally taken to be the circuit
default, can be overridden.
30.13.2 Parameters
Resistor Flag Resistor
: Capacitor Flag Capacitor
: Diode Flag Diode
: Bipolar|BJT Flag Bipolar transistor
: MOSFET Flag MOS field-effect transistor
: JFET Flag Junction field-effect transistor
: MESFET Flag MES field-effect transistor
IC.File String Initial-conditions filename
Unique Flag Append instance name to filename
DefA Real Default Mask Area, (m2)
DefW Real Default Mask Width, (m)
DefL Real Default Mask Length, (m)
Base.Area Real lD BJT base area relative to emitter area
Base.Length Real Real lD BJT base contact length, (µm)
Base.Depth Real lD BJT base contact depth, (µm)
TNom Real Nominal measurement temperature, (°C)
30.13.3 Examples
Normally, a ’numos’ device model is used for MOSFET devices. However, it can be
changed into a bipolar-with-substrate-contact model, by specifying a bipolar structure
using the other cards, and indicating the device-structure type as shown here. The default
length is set to 1.0 µm so that when mask area is specified on the element line it can be
divided by this default to obtain the device width.
options bipolar defl =1.0
Specify that a 1D BJT has base area 1/10th that of the emitter, has an effective depth
of 0.2 µm and a length between the internal and external base contacts
options base . area =0.1 base . depth =0.2 base . len =1.5
If a circuit contains two instances of a bipolar transistor model named ’q1’ and ’q2’, the
following line tells the simulator to look for initial conditions in the ’OP1.q2’, respectively.
The period in the middle of the names is added automatically:
30.14. OUTPUT 647
options unique ic . file =" OP1 "
30.13.4 See also

numd, nbjt, numos
30.14 OUTPUT
Identify information to be printed or saved
SYNOPSIS
output [ debugging - flags ] [ general - info ] [ saved - solutions ]
30.14.1 DESCRIPTION
The output card is used to control the amount of information that is either presented to
or saved for the user. Three types of information are available. Debugging information is
available as a means to monitor program execution. This is useful during long simulations
when one is unsure about whether the program has become trapped at some stage of the
simulation. General information about a device such as material parameters and resource
usage can be obtained. Finally, information about the internal and external states of a
device is available. Since this data is best interpreted using a post-processor, a facility
is available for saving device solutions in auxiliary output files. Solution filenames are
automatically generated by the simulator. If the named file already exists, the file will be
overwritten. A filename unique to a particular circuit or run can be generated by providing
a root filename. This root name will be added onto the beginning of the automatically
generated name. This feature can be used to store solutions in a directory other than the
current one by specifying the root filename as the path of the desired directory. Solutions
are only saved for those devices that specify the ‘save’ parameter on their instance lines.
The various physical values that can be saved are named below. By default, the following
values are saved: the doping, the electron and hole concentrations, the potential, the
electric field, the electron and hole current densities, and the displacement current density.
Values can be added to or deleted from this list by turning the appropriate flag on or off.
For vector-valued quantities in two dimensions, both the X and Y components are saved.
The vector magnitude can be obtained during post-processing.
Saved solutions can be used in conjunction with the options card and instance lines to
reuse previously calculated solutions as initial guesses for new solutions. For example,
it is typical to initialize the device to a known state prior to beginning any DC transfer
curve or operating point analysis. This state is an ideal candidate to be saved for later use
when it is known that many analyses will be performed on a particular device structure.
Depending on the global variable filetype the outputs may be stored as (compact)
binary or text processor readable ascii formatted data.
30.14.2 Parameters
All.Debug Flag Debug all analyses
OP.Debug Flag .OP analyses
DC.Debug Flag .DC analyses
TRAN.Debug Flag .TRAN analyses
AC.Debug Flag .AC analyses
PZ.Debug Flag .PZ analyses
Material Flag Physical material information
Statistics | Resources Flag Resource usage information
RootFile String Root of output file names
Psi Flag Potential ( V )
Equ.Psi Flag Equilibrium potential ( V )
Vac.Psi Flag Vacuum potential ( V )
Doping Flag Net doping ( cm3 )
N.Conc Flag Electron concentration ( cm3 )
P.Conc Flag Hole concentration ( cm3 )
PhiN Flag Electron quasi-fermi potential ( V )
PhiP Flag Hole quasi-fermi potential ( V )
PhiC Flag Conduction band potential ( V )
PhiV Flag Valence band potential ( V )
E.Field Flag Electric field ( V/cm )
JC Flag Conduction current density ( A/cm2 )
JD Flag Displacement current density ( A/cm2 )
JN Flag Electron current density ( A/cm2 )
JP Flag Hole current density ( A/cm2 )
JT Flag Total current density ( A/cm2 )
Unet Flag Net recombination ( 1/cm3 s )
MuN Flag Electron mobility (low-field) ( cm2/Vs )
MuP Flag Hole mobility (low-field) ( cm2/Vs )
30.14.3 Examples
The following example activates all potentially valuable diagnostic output:
output all . debug mater stat
Energy band diagrams generally contain the potential, the quasi-fermi levels, the energies
and the vacuum energy. The following example enables saving of the r values needed to
make energy band diagrams:
output phin phjp phic phiv vac . psi
Sometimes it is desirable to save certain key solutions, and then reload them for use in
subsequent simulations. In such cases only the essential values ( Ψ, n, and p ) need to be
saved. This example turns off the nonessential default values (and indicates the essential
ones explicitly):
30.15. TITLE 649
output psi n . conc p . conc ^ e . f ^ jn ^ jp ^ jd
30.14.4 SEE ALSO
options, numd, nbjt, numos
30.15 TITLE
Provide a label for this device’s output
SYNOPSIS
title [ text ]
30.15.1 DESCRIPTION
The title card provides a label for use as a heading in various output files. The text can
be any length, but titles that fit on a single line will produce more aesthetically pleasing
output.
30.15.2 EXAMPLES
Set the title for a minimum gate length NMOSFET in a 1.0µm BiCMOS process
title L =1.0 um NMOS Device , 1.0 um BiCMOS Process
30.15.3 BUGS
The title is currently treated like a comment.
30.16 X.MESH, Y.MESH

Define locations of lines and nodes in a mesh
SYNOPSIS
x . mesh position numbering - method [ spacing - parameters ]

y . mesh position numbering - method [ spacing - parameters ]
30.16.1 DESCRIPTION
The domains of a device are discretized onto a rectangular finite-difference mesh using
x.mesh cards for 1D devices, or x.mesh and y.mesh cards for 2D devices. Both uniform
and non-uniform meshes can be specified.
A typical mesh for a 2D device is shown in Fig. 30.3.
Figure 30.3: Typical mesh for 2D devices
The mesh is divided into intervals by the reference lines. The other lines in each interval
are automatically generated by CIDER using the mesh spacing parameters. In general,
each new mesh card adds one reference line and multiple automatic lines to the mesh.
Conceptually, a 1D mesh is similar to a 2D mesh except that there are no reference or
automatic lines needed in the second dimension.
The location of a reference line in the mesh must either be given explicitly (using Location)
or defined implicitly relative to the location of the previous reference line (by using Width).
(If the first card in either direction is specified using Width, an initial reference line is
automatically generated at location 0.0.) The line number of the reference line can be
given explicitly, in which case the automatic lines are evenly spaced within the interval,
and the number of lines is determined from the difference between the current line number
and that of the previous reference line. However, if the interval width is given, then the
line number is interpreted directly as the number of additional lines to add to the mesh.
For a nonuniformly spaced interval, the number of automatic lines has to be determined
using the mesh spacing parameters. Nonuniform spacing is triggered by providing a
desired ratio for the lengths of the spaces between adjacent pairs of lines. This ratio
should always be greater than one, indicating the ratio of larger spaces to smaller spaces.
In addition to the ratio, one or both of the space widths at the ends of the interval must
30.16. X.MESH, Y.MESH 651
be provided. If only one is given, it will be the smallest space and the largest space will
be at the opposite end of the interval. If both are given, the largest space will be in
the middle of the interval. In certain cases it is desirable to limit the growth of space
widths in order to control the solution accuracy. This can be accomplished by specifying
a maximum space size, but this option is only available when one of the two end lengths
is given. Note that once the number of new lines is determined using the desired ratio,
the actual spacing ratio may be adjusted so that the spaces exactly fill the interval.
30.16.2 Parameters
Location Real Location of this mesh line, ( µm )
:Width Real Width between this and previous mesh lines, ( µm )
Number | Node Integer Number of this mesh line
:Ratio Real Ratio of sizes of adjacent spaces
H.Start | H1 Real Space size at start of interval, ( µm )
H.End | H2 Real Space size at end of interval, ( µm )
H.Max | H3 Real Maximum space size inside interval, ( µm )
30.16.3 EXAMPLES
A 50 node, uniform mesh for a 5 µm long semiconductor resistor can be specified as:
x . mesh loc =0.0 n =1

x . mesh loc =5.0 n =50
An accurate mesh for a 1D diode needs fine spacing near the junction. In this example,
the junction is assumed to be 0.75 µm deep. The spacing near the diode ends is limited
to a maximum of 0.1 µm:
x . mesh w =0.75 h . e =0.001 h . m =0. l ratio =1.5

x . mesh w =2.25 h . s =0.001 h . m =0. l ratio =1.5
The vertical mesh spacing of a MOSFET can generally be specified as uniform through the
gate oxide, very fine for the surface inversion layer, moderate down to the so source/drain
junction depth, and then increasing all the way to the bulk contact:
y . mesh loc = -0.04 node =1

y . mesh loc =0.0 node =6
y . mesh width =0.5 h . start =0.001 h . max =.05 ratio =2.0
y . mesh width =2.5 h . start =0.05 ratio =2.0
30.16.4 SEE ALSO

domain
30.17 NUMD
Diode / two-terminal numerical models and elements
SYNOPSIS Model:
. MODEL model - name NUMD [ level ]

+ ...
SYNOPSIS Element:
DXXXXXXX nl n2 model - name [ geometry ] [ temperature ] [ initial - conditi
SYNOPSIS Output:
. SAVE [ small - signal values ]
30.17.1 DESCRIPTION
NUMD is the name for a diode numerical model. In addition, this same model can be
used to simulate other two-terminal structures such as semiconductor resistors and MOS
capacitors. See the options card for more information on how to customize the device
type.
Both 1D and 2D devices are supported. These correspond to the LEVEL=l and LEVEL=2
models, respectively. If left unspecified, it is assumed that the device is one-dimensional.
All numerical two-terminal element names begin with the letter ‘D. The element name is
then followed by the names of the positive (n1) and negative (n2) nodes. After this must
come the name of the model used for the element. The remaining information can come in
any order. The layout dimensions of an element are specified relative to the geometry of a
default device. For 1D devices, the default device has an area of 1m2, and for 2D devices,
the default device has a width of 1 m. However, these defaults can be overridden on an
options card. The operating temperature of a device can be set independently from that
of the rest of the circuit in order to simulate non-isothermal circuit operation. Finally,
the name of a file containing an initial state for the device can be specified. Remember
that if the filename contains capital letters, they must be protected by surrounding the
filename with double quotes. Alternatively, the device can be placed in an OFF state
(thermal equilibrium) at the beginning of the analysis. For more information on the use
of initial conditions, see the ngspice User’s Manual, Chapt. 7.1.
In addition to the element input parameters, there are output-only parameters that can
be shown using the ngspice show command (17.5.82) or captured using the save/.SAVE
(17.5.71/15.6.1) command. These parameters are the elements of the indefinite con-
ductance (G), capacitance (C), and admittance (Y) matrices where Y = G + jωC. By
default, the parameters are computed at 1 Hz. Each element is accessed using the name
of the matrix (g, c or y) followed by the node indices of the output terminal and the input
terminal (e.g. g11). Beware that names are case-sensitive for save/show, so lower-case
letters must be used.
30.17. NUMD 653
30.17.2 Parameters

Level Integer Dimensionality of numerical model
Area Real Multiplicative area factor
W Real Multiplicative width factor
Temp Real Element operating temperature
Off Flag Device initially in OFF state
gIJ Flag Conductance element Gij , ( Ω )
cIJ Flag Capacitance element Cij , ( F )
yIJ Flag Admittance element Yij , ( Ω )
30.17.3 EXAMPLES
A one-dimensional numerical switching-diode element/model pair with an area twice that

of the default device (which has a size of l µm x 1 µm) can be specified using:
DSWITCH 1 2 M_SWITCH_DIODE AREA =2

. MODEL M_SWITCH_DIODE NUMD
+ options defa =1 p ...
+ ...
A two-dimensional two-terminal MOS capacitor with a width of 20 µm and an initial

condition of 3 V is created by:
DMOSCAP 11 12 M_MOSCAP W =20 um IC =3 v

. MODEL M_MOSCAP NUMD LEVEL =2
+ options moscap defw =1 m
+ ...
The next example shows how both the width and area factors can be used to create a
power diode with area twice that of a 6µm-wide device (i.e. a 12µm-wide device). The
device is assumed to be operating at a temperature of 100°C:
D1 POSN NEGN POWERMOD AREA =2 W =6 um TEMP =100.0

. MODEL POWERMOD NUMD LEVEL =2
+ ...
This example saves all the small-signal parameters of the previous diode:
. SAVE @d1 [ g11 ] @d1 [ g12 ] @d1 [ g21 ] @d1 [ g22 ]

. SAVE @d1 [ c11 ] @d1 [ c12 ] @d1 [ c21 ] @d1 [ c22 ]
. SAVE @d1 [ y11 ] @d1 [ y12 ] @d1 [ y21 ] @d1 [ y22 ]
30.17.4 SEE ALSO
options, output
30.17.5 BUGS
Convergence problems may be experienced when simulating MOS capacitors due to sin-
gularities in the current-continuity equations.
30.18 NBJT
Bipolar / three-terminal numerical models and elements
SYNOPSIS Model:
. MODEL model - name NBJT [ level ]

+ ...
SYNOPSIS Element:
QXXXXXXX nl n2 n3 model - name [ geometry ]

+ [ temperature ] [ initial - conditions ]
SYNOPSIS Output:
30.18.1 DESCRIPTION
NBJT is the name for a bipolar transistor numerical model. In addition, the 2D model
can be used to simulate other three-terminal structures such as a JFET or MESFET.
However, the 1D model is customized with a special base contact, and cannot be used
for other purposes. See the options card for more information on how to customize the
device type and setup the 1D base contact.
Both 1”and 2D devices are supported. These correspond to the LEVEL=l and models,
respectively. If left unspecified, it is assumed that the device is one-dimensional.
All numerical three-terminal element names begin with the letter ’Q’. If the device is a
bipolar transistor, then the nodes are specified in the order: collector (nl), base (n2),
emitter (n3). For a JFET or MESFET, the node order is: drain (n1), gate (n2), source
(n3). After this must come the name of the model used for the element. The remaining
information can come in any order. The layout dimensions of an element are specified
relative to the geometry of a default device. For the 1D BJT, the default device has an
area of lm2, and for 2D devices, the default device has a width of lm. In addition, it
is assumed that the default 1D BJT has a base contact with area equal to the emitter
30.18. NBJT 655
area, length of 1µm and a depth automatically determined from the device doping profile.
However, all these defaults can be overridden on an options card.
The operating temperature of a device can be set independently from the rest of that of
the circuit in order to simulate non-isothermal circuit operation. Finally, the name of a file
containing an initial state for the device can be specified. Remember that if the filename
contains capital letters, they must be protected by surrounding the filename with double
quotes. Alternatively, the device can be placed in an OFF state (thermal equilibrium) at
the beginning of the analysis. For more information on the use of initial conditions, see
the ngspice User’s Manual.
be shown using the SPICE show command or captured using the save/.SAVE command.
These parameters are the elements of the indefinite conductance (G), capacitance (C), and
admittance (Y) matrices where Y = G + jωC. By default, the parameters are computed
at 1Hz. Each element is accessed using the name of the matrix (g, c or y) followed by
the node indices of the output terminal and the input terminal (e.g. g11). Beware that
parameter names are case-sensitive for save/show, so lower-case letters must be used.
30.18.2 Parameters
30.18.3 EXAMPLES
A one-dimensional numerical bipolar transistor with an emitter stripe 4 times as wide as

the default device is created using:
Q2 1 2 3 M_BJT AREA =4
This example saves the output conductance (go), transconductance (gm) and input con-
ductance (gpi) of the previous transistor in that order:
. SAVE @q2 [ g11 ] @q2 [ g12 ] @q2 [ g22 ]
The second example is for a two-dimensional JFET with a width of 5pm and initial
conditions obtained from file IC.jfet:
QJ1 11 12 13 M_JFET W =5 um IC . FILE =" IC . jfet "

. MODEL M_JFET NBJT LEVEL =2
+ options jfet
+ ...
A final example shows how to use symmetry to simulate half of a 2D BJT, avoiding having
the user double the area of each instance:
Q2 NC2 NB2 NE2 BJTMOD AREA =1

Q3 NC3 NB3 NE3 BJTMOD AREA =1
. MODEL BJTMOD NBJT LEVEL =2
+ options defw =2 um
+ * Define half of the device now
+ ...
30.18.4 SEE ALSO
options, output
30.18.5 BUGS
MESFETs cannot be simulated properly yet because Schottky contacts have not been
implemented.
30.19 NUMOS
MOSFET / four-terminal numerical models and elements

SYNOPSIS Model:
. MODEL model - name NUMOS [ level ]

+ ...
SYNOPSIS Element:
MXXXXXXX nl n2 n3 n4 model - name [ geometry ]

+ [ temperature ] [ initial - conditions ]
SYNOPSIS Output:

30.19. NUMOS 657
30.19.1 DESCRIPTION
NUMOS is the name for a MOSFET numerical model. In addition, the 2D model can
be used to simulate other four-terminal structures such as integrated bipolar and JFET
devices with substrate contacts. However, silicon controlled rectifiers (SCRs) cannot be
simulated because of the snapback in the transfer characteristic. See the options card
for more information on how to customize the device type. The LEVEL parameter of
two- and three-terminal devices is not needed, because only 2D devices are supported.
However, it will accepted and ignored if provided.
All numerical four-terminal element names begin with the letter ‘M’. If the device is a
MOSFET, or JFET with a bulk contact, then the nodes are specified in the order: drain
(n1), gate (n2), source (n3), bulk (n4). If the device is a BJT, the node order is: collector
(n1), base (n2), emitter (n3), substrate (n4). After this must come the name of the model
1used for the element. The remaining information can come in any order. The layout
dimensions of an element are specified relative to the geometry of a default device. The
default device has a width of lm. However, this default can be overridden on an options
card. In addition, the element line will accept a length parameter, L, but does not use it
in any calculations. This is provided to enable somewhat greater compatibility between
numerical MOSFET models and the standard SPICE3 compact MOSFET models.
The operating temperature of a device can be set independently from that of the rest of
the circuit in order to simulate non-isothermal circuit operation. Finally, the name of a file
containing an initial state for the device can be specified. Remember that if the filename
contains capital letters, they must be protected by surrounding the filename with double
quotes. Alternatively, the device can be placed in an OFF state (thermal equilibrium) at
the beginning of the analysis. For more information on the use of initial conditions, see
the ngspice User’s Manual.
be shown using the SPICE show command or captured using the save/.SAVE command.
These parameters are the elements of the indefinite conductance (G), capacitance (C), and
admittance (Y) matrices where Y = G + jωC. By default, the parameters are computed
at 1 Hz. Each element is accessed using the name of the matrix (g, c or y) followed by
the node indices of the output terminal and the input terminal (e.g. g11). Beware that
parameter names are case-sensitive for save/show, so lower-case letters must be used.
30.19.2 Parameters
L Real Unused length factor
30.19.3 EXAMPLES
A numerical MOSFET with a gate width of 5µm and length of 1µm is described below.
However, the model can only be used for lµm length devices, so the length parameter is
redundant. The device is initially biased near its threshold by taking an initial state from
the file NM1.vth.
M1 1 2 3 4 M_NMOS_1UM W =5 um L =1 um IC . FILE =" NM1 . vth "

. MODEL MNMOS_1UM NUMOS
+ * Description of a lum device
+ ...
This example saves the definite admittance matrix of the previous MOSFET where the
source terminal (3) is used as the reference. (The definite admittance matrix is formed
by deleting the third row and column from the indefinite admittance matrix.)
. SAVE @m1 [ y11 ] @m1 [ y12 ] @ml [ y14 ]

. SAVE @m1 [ y21 ] @m1 [ y22 ] @ml [ y24 ]
. SAVE @m1 [ y41 ] @m1 [ y42 ] @ml [ y44 ]
Bipolar transistors are usually specified in terms of their area relative to a unit device.
The following example creates a unit-sized device:
MQ1 NC NB NE NS N_BJT
. MODEL M_BJT NUMOS LEVEL =2
+ options bipolar defw =5 um
+ ...
30.19.4 SEE ALSO

options, output
30.20 2D contour plots

2D contour plots of CIDER simulation results are available via Gnuplot (see 18.7.2).
30.21 Cider examples

The original Cider User’s manual, in its Appendix A, lists a lot of examples, starting at
page 226. We do not reproduce these pages here, but ask you to refer to the original
document. If you experience any difficulties downloading it, please send a note to the
ngspice users’ mailing list.
Part IV
Miscellaneous
659
Chapter 31
Model and Device Parameters
The following tables summarize the parameters available on each of the devices and models
in ngspice. There are two tables for each type of device supported by ngspice. Input
parameters to instances and models are parameters that can occur on an instance or
model definition line in the form keyword=value where keyword is the parameter name
as given in the tables. Default input parameters (such as the resistance of a resistor or
the capacitance of a capacitor) obviously do not need the keyword specified.
31.1 Accessing internal device parameters

Output parameters are those additional parameters that are available for many types of
instances for the output of operating point and debugging information. These parameters
are specified as @device[keyword] and are available for the most recent point computed
or, if specified in a .save statement, for an entire simulation as a normal output vector.
Thus, to monitor the gate-to-source capacitance of a MOSFET, a command
save @m1 [ cgs ]
given before a transient simulation causes the specified capacitance value to be saved at
each time-point, and a subsequent command such as
plot @m1 [ cgs ]
produces the desired plot. (Note that the show command does not use this format).
Some variables are listed as both input and output, and their output simply returns
the previously input value, or the default value after the simulation has been run. Some
parameters are input only because the output system can not handle variables of the given
type yet, or the need for them as output variables has not been apparent. Many such
input variables are available as output variables in a different format, such as the initial
condition vectors that can be retrieved as individual initial condition values. Finally,
internally derived values are output only and are provided for debugging and operating
point output purposes.
661
662 CHAPTER 31. MODEL AND DEVICE PARAMETERS
If you want to access a device parameter of a device used inside of a subcircuit, you may
use the syntax as shown below.
General form:
@device_identifier . subcircuit_name . < subcircuit_name_nn >

+. device_name [ parameter ]
* transistor output characteristics

* two nested subcircuits
vdd d1 0 2.0
vss vsss 0 0
vsig g1 vsss 0
xmos1 d1 g1 vsss level1
. subckt level1 d3 g3 v3
xmos2 d3 g3 v3 level2
. ends
. subckt level2 d4 g4 v4
m1 d4 g4 v4 v4 nmos w =1 e -5 l =3.5 e -007
. ends
. dc vdd 0 5 0.1 vsig 0 5 1
. control
save all @m . xmos1 . xmos2 . m1 [ vdsat ]
run
plot vss # branch $ current measured at the top level
plot @m . xmos1 . xmos2 . m1 [ vdsat ]
. endc
. MODEL NMOS NMOS LEVEL = 8
+ VERSION = 3.2.4 TNOM = 27 TOX = 7.4 E -9
. end
The device identifier is the first letter extracted from the device name, e.g. m for a MOS
transistor.
Please note that the parameter tables presented below do not provide the detailed infor-
mation available about the parameters provided in the section on each device and model,
but are provided as a quick reference guide.
31.2 Elementary Devices
31.2.1 Resistor
31.2.1.1 Resistor instance parameters
# Name Direction Type Description

1 resistance (r) InOut real Resistance
10 ac InOut real AC resistance value
8 temp InOut real Instance operating temperature
14 dtemp InOut real Instance temperature difference
with the rest of the circuit
3 l InOut real Length
2 w InOut real Width
12 m InOut real Multiplication factor
16 tc InOut real First order temp. coefficient
16 tc1 InOut real First order temp. coefficient
17 tc2 InOut real Second order temp. coefficient
13 scale InOut real Scale factor
15 noisy (noise) InOut integer Resistor generate noise
5 sens_resist In flag flag to request sensitivity WRT
resistance
6 i Out real Current
7 p Out real Power
206 sens_dc Out real dc sensitivity
201 sens_real Out real dc sensitivity and real part of ac
sensitivity
202 sens_imag Out real dc sensitivity and imag part of
ac sensitivity
203 sens_mag Out real ac sensitivity of magnitude
204 sens_ph Out real ac sensitivity of phase
205 sens_cplx Out complex ac sensitivity
31.2.1.2 Resistor model parameters

103 rsh InOut real Sheet resistance
106 narrow InOut real Narrowing of resistor
106 dw InOut real
109 short InOut real Shortening of resistor
109 dlr InOut real
104 defw InOut real Default device width
104 w InOut real Default device width
105 l InOut real Default device length
110 kf InOut real Flicker noise coefficient
111 af InOut real Flicker noise exponent
108 tnom InOut real Parameter measurement temperature
107 r InOut real Resistance
107 res InOut real Resistance
wf InOut real Flicker noise width exponent
lf InOut real Flicker noise length exponent
ef InOut real Flicker noise frequency exponent
r In flag Device is a resistor model
31.2.2 Capacitor - Fixed capacitor
31.2.2.1 Capacitor instance parameters

1 capacitance InOut real Device capacitance
1 cap InOut real Device capacitance
1 c InOut real Device capacitance
2 ic InOut real Initial capacitor voltage
from the rest of the circuit
3 w InOut real Device width
4 l InOut real Device length
11 m InOut real Parallel multiplier
5 sens_cap In flag flag to request sens. WRT cap.
6 i Out real Device current
7 p Out real Instantaneous device power
201 sens_real Out real real part of ac sensitivity
202 sens_imag Out real dc sens. & imag part of ac sens.
203 sens_mag Out real sensitivity of ac magnitude
204 sens_ph Out real sensitivity of ac phase
31.2.2.2 Capacitor model parameters

112 cap InOut real Model capacitance
101 cj InOut real Bottom Capacitance per area
102 cjsw InOut real Sidewall capacitance per meter
103 defw InOut real Default width
113 defl InOut real Default length
105 narrow InOut real width correction factor
106 short InOut real length correction factor
110 di InOut real Relative dielectric constant
111 thick InOut real Insulator thickness
104 c In flag Capacitor model
31.2.3 Inductor - Fixed inductor
31.2.3.1 Inductor instance parameters

1 inductance InOut real Inductance of inductor
2 ic InOut real Initial current through inductor
5 sens_ind In flag flag to request sensitivity WRT
inductance
10 dtemp InOut real Instance temperature difference with
the rest of the circuit
8 m InOut real Multiplication Factor
12 nt InOut real Number of turns
3 flux Out real Flux through inductor
4 v Out real Terminal voltage of inductor
4 volt Out real
6 i Out real Current through the inductor
6 current Out real
7 p Out real instantaneous power dissipated by the
inductor
206 sens_dc Out real dc sensitivity sensitivity
202 sens_imag Out real dc sensitivity and imag part of ac
sensitivty
203 sens_mag Out real sensitivity of AC magnitude
204 sens_ph Out real sensitivity of AC phase
31.2.3.2 Inductor model parameters

100 ind InOut real Model inductance
104 csect InOut real Inductor cross section
105 length InOut real Inductor length
106 nt InOut real Model number of turns
107 mu InOut real Relative magnetic permeability
108 l In flag Inductor model
31.2.4 Mutual - Mutual Inductor
31.2.4.1 Mutual instance parameters

401 k InOut real Mutual inductance
401 coefficient InOut real
402 inductor1 InOut instance First coupled inductor
403 inductor2 InOut instance Second coupled inductor
404 sens_coeff In flag flag to request sensitivity WRT coupling factor
602 sens_imag Out real dc sensitivity and imag part of ac sensitivty
603 sens_mag Out real sensitivity of AC magnitude
604 sens_ph Out real sensitivity of AC phase
605 sens_cplx Out complex mutual model parameters:
31.3 Voltage and current sources
31.3.1 Bxxxx - Arbitrary source (ASRC)
31.3.1.1 ASRC instance parameters

2 i In parsetree Current source
1 v In parsetree Voltage source
7 i Out real Current through source
6 v Out real Voltage across source
3 pos_node Out integer Positive Node
4 neg_node Out integer Negative Node
31.3. VOLTAGE AND CURRENT SOURCES 669
31.3.2 Isource - Independent current source
31.3.2.1 Isource instance parameters

1 dc InOut real DC value of source
2 acmag InOut real AC magnitude
3 acphase InOut real AC phase
5 pulse In real vector Pulse description
6 sine In real vector Sinusoidal source description
6 sin In real vector Sinusoidal source description
7 exp In real vector Exponential source description
8 pwl In real vector Piecewise linear description
9 sffm In real vector Single freq. FM description
21 am In real vector Amplitude modulation description
10 neg_node Out integer Negative node of source
11 pos_node Out integer Positive node of source
12 acreal Out real AC real part
13 acimag Out real AC imaginary part
14 function Out integer Function of the source
15 order Out integer Order of the source function
16 coeffs Out real vector Coefficients of the source
20 v Out real Voltage across the supply
17 p Out real Power supplied by the source
4 ac In real vector AC magnitude,phase vector
1 c In real Current through current source
22 current Out real Current in DC or Transient mode
18 distof1 In real vector f1 input for distortion
31.3.3 Vsource - Independent voltage source
31.3.3.1 Vsource instance parameters

1 dc InOut real D.C. source value
3 acmag InOut real A.C. Magnitude
4 acphase InOut real A.C. Phase
5 pulse In real vector Pulse description
6 sine In real vector Sinusoidal source description
6 sin In real vector Sinusoidal source description
7 exp In real vector Exponential source description
8 pwl In real vector Piecewise linear description
9 sffm In real vector Single freq. FM descripton
22 am In real vector Amplitude modulation descripton
11 function Out integer Function of the source
12 order Out integer Order of the source function
13 coeffs Out real vector Coefficients for the function
14 acreal Out real AC real part
15 acimag Out real AC imaginary part
2 ac In real vector AC magnitude, phase vector
18 i Out real Voltage source current
19 p Out real Instantaneous power
23 r In real pwl repeat start time value
24 td In real pwl delay time value
31.3. VOLTAGE AND CURRENT SOURCES 671
31.3.4 Fxxxx: Current-Controlled Current Source (CCCS)
31.3.4.1 CCCS instance parameters

1 gain InOut real Gain of source
2 control InOut instance Name of controlling source
6 sens_gain In flag flag to request sensitivity WRT gain
7 i Out real CCCS output current
9 v Out real CCCS voltage at output
8 p Out real CCCS power
202 sens_imag Out real imag part of ac sensitivity
31.3.5 Hxxxx: Current-Controlled Voltage Source (CCVS)
31.3.5.1 CCVS instance parameters

1 gain InOut real Transresistance (gain)
2 control InOut instance Controlling voltage source
7 sens_trans In flag flag to request sens. WRT transimpedance
8 i Out real CCVS output current
10 v Out real CCVS output voltage
9 p Out real CCVS power
31.3.6 Gxxxx: Voltage-Controlled Current Source (VCCS)
31.3.6.1 VCCS instance parameters

1 gain InOut real Transconductance of source (gain)
8 sens_trans In flag flag to request sensitivity WRT transconductance
5 cont_p_node Out integer Positive node of contr. source
6 cont_n_node Out integer Negative node of contr. source
2 ic In real Initial condition of controlling source
9 i Out real Output current
11 v Out real Voltage across output
10 p Out real Power
31.3.7 Exxxx: Voltage-Controlled Voltage Source (VCVS)
31.3.7.1 VCVS instance parameters

1 gain InOut real Voltage gain
9 sens_gain In flag flag to request sensitivity WRT gain
4 cont_p_node Out integer Positive node of contr. source
5 cont_n_node Out integer Negative node of contr. source
7 ic In real Initial condition of controlling source
10 i Out real Output current
12 v Out real Output voltage
11 p Out real Power
31.4. TRANSMISSION LINES 673
31.4 Transmission Lines
31.4.1 CplLines - Simple Coupled Multiconductor Lines
31.4.1.1 CplLines instance parameters

1 pos_nodes InOut string vector in nodes
2 neg_nodes InOut string vector out nodes
3 dimension InOut integer number of coupled lines
4 length InOut real length of lines
31.4.1.2 CplLines model parameters

101 r InOut real vector resistance per length
104 l InOut real vector inductance per length
102 c InOut real vector capacitance per length
103 g InOut real vector conductance per length
105 length InOut real length
106 cpl In flag Device is a cpl model
31.4.2 LTRA - Lossy transmission line
31.4.2.1 LTRA instance parameters

6 v1 InOut real Initial voltage at end 1
7 i1 InOut real Initial current at end 1
10 ic In real vector Initial condition vector:v1,i1,v2,i2
13 pos_node1 Out integer Positive node of end 1 of t-line
14 neg_node1 Out integer Negative node of end 1 of t.line
15 pos_node2 Out integer Positive node of end 2 of t-line
16 neg_node2 Out integer Negative node of end 2 of t-line
31.4.2.2 LTRA model parameters

0 ltra InOut flag LTRA model
1 r InOut real Resistance per meter
2 l InOut real Inductance per meter
3 g InOut real
4 c InOut real Capacitance per meter
5 len InOut real length of line
11 rel Out real Rel. rate of change of deriv. for bkpt
12 abs Out real Abs. rate of change of deriv. for bkpt
28 nocontrol InOut flag No timestep control
32 steplimit InOut flag always limit timestep to 0.8*(delay of line)
33 nosteplimit InOut flag don’t always limit timestep to 0.8*(delay of
line)
34 lininterp InOut flag use linear interpolation
35 quadinterp InOut flag use quadratic interpolation
36 mixedinterp InOut flag use linear interpolation if quadratic results
look unacceptable
46 truncnr InOut flag use N-R iterations for step calculation in
LTRAtrunc
47 truncdontcut InOut flag don’t limit timestep to keep impulse
response calculation errors low
42 compactrel InOut real special reltol for straight line checking
43 compactabs InOut real special abstol for straight line checking
31.4.3 Tranline - Lossless transmission line
31.4.3.1 Tranline instance parameters

1 z0 InOut real Characteristic impedance
1 zo InOut real
4 f InOut real Frequency
2 td InOut real Transmission delay
3 nl InOut real Normalized length at frequency given
9 ic In real vector Initial condition vector:v1,i1,v2,i2
10 rel Out real Rel. rate of change of deriv. for bkpt
11 abs Out real Abs. rate of change of deriv. for bkpt
12 pos_node1 Out integer Positive node of end 1 of t. line
13 neg_node1 Out integer Negative node of end 1 of t. line
14 pos_node2 Out integer Positive node of end 2 of t. line
15 neg_node2 Out integer Negative node of end 2 of t. line
18 delays Out real vector Delayed values of excitation
31.4.4 TransLine - Simple Lossy Transmission Line
31.4.4.1 TransLine instance parameters

1 pos_node In integer Positive node of txl
2 neg_node In integer Negative node of txl
3 length InOut real length of line
31.4.4.2 TransLine model parameters

101 r InOut real resistance per length
104 l InOut real inductance per length
102 c InOut real capacitance per length
103 g InOut real conductance per length
105 length InOut real length
106 txl In flag Device is a txl model
31.4.5 URC - Uniform R. C. line
31.4.5.1 URC instance parameters

1 l InOut real Length of transmission line
2 n InOut real Number of lumps
3 pos_node Out integer Positive node of URC
4 neg_node Out integer Negative node of URC
5 gnd Out integer Ground node of URC
31.4.5.2 URC model parameters

101 k InOut real Propagation constant
102 fmax InOut real Maximum frequency of interest
103 rperl InOut real Resistance per unit length
104 cperl InOut real Capacitance per unit length
105 isperl InOut real Saturation current per length
106 rsperl InOut real Diode resistance per length
107 urc In flag Uniform R.C. line model
31.5 BJTs
31.5.1 BJT - Bipolar Junction Transistor

31.5.1.1 BJT instance parameters

2 off InOut flag Device initially off
3 icvbe InOut real Initial B-E voltage
4 icvce InOut real Initial C-E voltage
1 area InOut real (Emitter) Area factor
10 areab InOut real Base area factor
11 areac InOut real Collector area factor
9 m InOut real Parallel Multiplier
5 ic In real vector Initial condition vector
6 sens_area In flag flag to request sensitivity WRT area
202 colnode Out integer Number of collector node
203 basenode Out integer Number of base node
204 emitnode Out integer Number of emitter node
205 substnode Out integer Number of substrate node
206 colprimenode Out integer Internal collector node
207 baseprimenode Out integer Internal base node
208 emitprimenode Out integer Internal emitter node
211 ic Out real Current at collector node
212 ib Out real Current at base node
236 ie Out real Emitter current
237 is Out real Substrate current
209 vbe Out real B-E voltage
210 vbc Out real B-C voltage
215 gm Out real Small signal transconductance
213 gpi Out real Small signal input conductance - pi
214 gmu Out real Small signal conductance - mu
225 gx Out real Conductance from base to internal base
216 go Out real Small signal output conductance
227 geqcb Out real d(Ibe)/d(Vbc)
228 gcsub Out real Internal Subs. cap. equiv. cond.
243 gdsub Out real Internal Subs. Diode equiv. cond.
229 geqbx Out real Internal C-B-base cap. equiv. cond.
239 cpi Out real Internal base to emitter capactance
240 cmu Out real Internal base to collector capactiance
241 cbx Out real Base to collector capacitance
242 csub Out real Substrate capacitance
218 cqbe Out real Cap. due to charge storage in B-E jct.
220 cqbc Out real Cap. due to charge storage in B-C jct.
222 cqsub Out real Cap. due to charge storage in Subs. jct.
224 cqbx Out real Cap. due to charge storage in B-X jct.
31.5. BJTS 679
226 cexbc Out real Total Capacitance in B-X junction

217 qbe Out real Charge storage B-E junction
219 qbc Out real Charge storage B-C junction
221 qsub Out real Charge storage Subs. junction
223 qbx Out real Charge storage B-X junction
238 p Out real Power dissipation
231 sens_imag Out real dc sens. & imag part of ac sens.
7 temp InOut real instance temperature
8 dtemp InOut real instance temperature delta from circuit
31.5.1.2 BJT model parameters

309 type Out string NPN or PNP
101 npn InOut flag NPN type device
102 pnp InOut flag PNP type device
147 subs InOut integer Vertical or Lateral device
103 is InOut real Saturation Current
146 iss InOut real Substrate Jct. Saturation Current
104 bf InOut real Ideal forward beta
105 nf InOut real Forward emission coefficient
106 vaf InOut real Forward Early voltage
106 va InOut real
107 ikf InOut real Forward beta roll-off corner current
107 ik InOut real
108 ise InOut real B-E leakage saturation current
110 ne InOut real B-E leakage emission coefficient
111 br InOut real Ideal reverse beta
112 nr InOut real Reverse emission coefficient
113 var InOut real Reverse Early voltage
113 vb InOut real
114 ikr InOut real reverse beta roll-off corner current
115 isc InOut real B-C leakage saturation current
117 nc InOut real B-C leakage emission coefficient
118 rb InOut real Zero bias base resistance
119 irb InOut real Current for base resistance=(rb+rbm)/2
120 rbm InOut real Minimum base resistance
121 re InOut real Emitter resistance
122 rc InOut real Collector resistance
123 cje InOut real Zero bias B-E depletion capacitance
124 vje InOut real B-E built in potential
124 pe InOut real

125 mje InOut real B-E junction grading coefficient
125 me InOut real
126 tf InOut real Ideal forward transit time
127 xtf InOut real Coefficient for bias dependence of TF
128 vtf InOut real Voltage giving VBC dependence of TF
129 itf InOut real High current dependence of TF
130 ptf InOut real Excess phase
131 cjc InOut real Zero bias B-C depletion capacitance
132 vjc InOut real B-C built in potential
132 pc InOut real
133 mjc InOut real B-C junction grading coefficient
133 mc InOut real
134 xcjc InOut real Fraction of B-C cap to internal base
135 tr InOut real Ideal reverse transit time
136 cjs InOut real Zero bias Substrate capacitance
136 csub InOut real
137 vjs InOut real Substrate junction built in potential
137 ps InOut real
138 mjs InOut real Substrate junction grading coefficient
138 ms InOut real
139 xtb InOut real Forward and reverse beta temp. exp.
140 eg InOut real Energy gap for IS temp. dependency
141 xti InOut real Temp. exponent for IS
148 tre1 InOut real Temp. coefficient 1 for RE
149 tre2 InOut real Temp. coefficient 2 for RE
150 trc1 InOut real Temp. coefficient 1 for RC
151 trc2 InOut real Temp. coefficient 2 for RC
152 trb1 InOut real Temp. coefficient 1 for RB
153 trb2 InOut real Temp. coefficient 2 for RB
154 trbm1 InOut real Temp. coefficient 1 for RBM
155 trbm2 InOut real Temp. coefficient 2 for RBM
142 fc InOut real Forward bias junction fit parameter
301 invearlyvoltf Out real Inverse early voltage:forward
302 invearlyvoltr Out real Inverse early voltage:reverse
303 invrollofff Out real Inverse roll off - forward
304 invrolloffr Out real Inverse roll off - reverse
305 collectorconduct Out real Collector conductance
306 emitterconduct Out real Emitter conductance
307 transtimevbcfact Out real Transit time VBC factor
308 excessphasefactor Out real Excess phase fact.
145 kf InOut real Flicker Noise Coefficient
144 af InOut real Flicker Noise Exponent
31.5. BJTS 681
31.5.2 VBIC - Vertical Bipolar Inter-Company Model

31.5.2.1 VBIC instance parameters

1 area InOut real Area factor
2 off InOut flag Device initially off
3 ic In real vector Initial condition vector
4 icvbe InOut real Initial B-E voltage
5 icvce InOut real Initial C-E voltage
6 temp InOut real Instance temperature
7 dtemp InOut real Instance delta temperature
8 m InOut real Multiplier
212 collnode Out integer Number of collector node
213 basenode Out integer Number of base node
214 emitnode Out integer Number of emitter node
215 subsnode Out integer Number of substrate node
216 collCXnode Out integer Internal collector node
217 collCInode Out integer Internal collector node
218 baseBXnode Out integer Internal base node
219 baseBInode Out integer Internal base node
220 baseBPnode Out integer Internal base node
221 emitEInode Out integer Internal emitter node
222 subsSInode Out integer Internal substrate node
223 vbe Out real B-E voltage
224 vbc Out real B-C voltage
225 ic Out real Collector current
226 ib Out real Base current
227 ie Out real Emitter current
228 is Out real Substrate current
229 gm Out real Small signal transconductance dIc/dVbe
230 go Out real Small signal output conductance dIc/dVbc
231 gpi Out real Small signal input conductance dIb/dVbe
232 gmu Out real Small signal conductance dIb/dVbc
233 gx Out real Conductance from base to internal base
247 cbe Out real Internal base to emitter capacitance
248 cbex Out real External base to emitter capacitance
249 cbc Out real Internal base to collector capacitance
250 cbcx Out real External Base to collector capacitance
251 cbep Out real Parasitic Base to emitter capacitance
252 cbcp Out real Parasitic Base to collector capacitance
259 p Out real Power dissipation
243 geqcb Out real Internal C-B-base cap. equiv. cond.
246 geqbx Out real External C-B-base cap. equiv. cond.
234 qbe Out real Charge storage B-E junction
235 cqbe Out real Cap. due to charge storage in B-E jct.
236 qbc Out real Charge storage B-C junction
237 cqbc Out real Cap. due to charge storage in B-C jct.
238 qbx Out real Charge storage B-X junction
239 cqbx Out real Cap. due to charge storage in B-X jct.
258 sens_dc Out real DC sensitivity
253 sens_real Out real Real part of AC sensitivity
254 sens_imag Out real DC sens. & imag part of AC sens.
255 sens_mag Out real Sensitivity of AC magnitude
256 sens_ph Out real Sensitivity of AC phase
257 sens_cplx Out complex AC sensitivity
31.5.2.2 VBIC model parameters

305 type Out string NPN or PNP
101 npn InOut flag NPN type device
102 pnp InOut flag PNP type device
103 tnom (tref) InOut real Parameter measurement temperature
104 rcx InOut real Extrinsic coll resistance
105 rci InOut real Intrinsic coll resistance
106 vo InOut real Epi drift saturation voltage
107 gamm InOut real Epi doping parameter
108 hrcf InOut real High current RC factor
109 rbx InOut real Extrinsic base resistance
110 rbi InOut real Intrinsic base resistance
111 re InOut real Intrinsic emitter resistance
112 rs InOut real Intrinsic substrate resistance
113 rbp InOut real Parasitic base resistance
114 is InOut real Transport saturation current
115 nf InOut real Forward emission coefficient
116 nr InOut real Reverse emission coefficient
117 fc InOut real Fwd bias depletion capacitance limit
118 cbeo InOut real Extrinsic B-E overlap capacitance
119 cje InOut real Zero bias B-E depletion capacitance
120 pe InOut real B-E built in potential
121 me InOut real B-E junction grading coefficient
122 aje InOut real B-E capacitance smoothing factor
123 cbco InOut real Extrinsic B-C overlap capacitance
124 cjc InOut real Zero bias B-C depletion capacitance
125 qco InOut real Epi charge parameter
126 cjep InOut real B-C extrinsic zero bias capacitance
127 pc InOut real B-C built in potential
128 mc InOut real B-C junction grading coefficient
129 ajc InOut real B-C capacitance smoothing factor
130 cjcp InOut real Zero bias S-C capacitance
131 ps InOut real S-C junction built in potential
132 ms InOut real S-C junction grading coefficient
31.5. BJTS 683
133 ajs InOut real S-C capacitance smoothing factor

134 ibei InOut real Ideal B-E saturation current
135 wbe InOut real Portion of IBEI from Vbei, 1-WBE from Vbex
136 nei InOut real Ideal B-E emission coefficient
137 iben InOut real Non-ideal B-E saturation current
138 nen InOut real Non-ideal B-E emission coefficient
139 ibci InOut real Ideal B-C saturation current
140 nci InOut real Ideal B-C emission coefficient
141 ibcn InOut real Non-ideal B-C saturation current
142 ncn InOut real Non-ideal B-C emission coefficient
143 avc1 InOut real B-C weak avalanche parameter 1
144 avc2 InOut real B-C weak avalanche parameter 2
145 isp InOut real Parasitic transport saturation current
146 wsp InOut real Portion of ICCP
147 nfp InOut real Parasitic fwd emission coefficient
148 ibeip InOut real Ideal parasitic B-E saturation current
149 ibenp InOut real Non-ideal parasitic B-E saturation current
150 ibcip InOut real Ideal parasitic B-C saturation current
151 ncip InOut real Ideal parasitic B-C emission coefficient
152 ibcnp InOut real Nonideal parasitic B-C saturation current
153 ncnp InOut real Nonideal parasitic B-C emission coefficient
154 vef InOut real Forward Early voltage
155 ver InOut real Reverse Early voltage
156 ikf InOut real Forward knee current
157 ikr InOut real Reverse knee current
158 ikp InOut real Parasitic knee current
159 tf InOut real Ideal forward transit time
160 qtf InOut real Variation of TF with base-width modulation
161 xtf InOut real Coefficient for bias dependence of TF
162 vtf InOut real Voltage giving VBC dependence of TF
163 itf InOut real High current dependence of TF
164 tr InOut real Ideal reverse transit time
165 td InOut real Forward excess-phase delay time
166 kfn InOut real B-E Flicker Noise Coefficient
167 afn InOut real B-E Flicker Noise Exponent
168 bfn InOut real B-E Flicker Noise 1/f dependence
169 xre InOut real Temperature exponent of RE
170 xrb InOut real Temperature exponent of RB
171 xrbi InOut real Temperature exponent of RBI
172 xrc InOut real Temperature exponent of RC
173 xrci InOut real Temperature exponent of RCI
174 xrs InOut real Temperature exponent of RS
175 xvo InOut real Temperature exponent of VO
176 ea InOut real Activation energy for IS
177 eaie InOut real Activation energy for IBEI
179 eaic InOut real Activation energy for IBCI/IBEIP
179 eais InOut real Activation energy for IBCIP

180 eane InOut real Activation energy for IBEN
181 eanc InOut real Activation energy for IBCN/IBENP
182 eans InOut real Activation energy for IBCNP
183 xis InOut real Temperature exponent of IS
184 xii InOut real Temperature exponent of IBEI,IBCI,IBEIP,IBCIP
185 xin InOut real Temperature exponent of IBEN,IBCN,IBENP,IBCNP
186 tnf InOut real Temperature exponent of NF
187 tavc InOut real Temperature exponent of AVC2
188 rth InOut real Thermal resistance
189 cth InOut real Thermal capacitance
190 vrt InOut real Punch-through voltage of internal B-C junction
191 art InOut real Smoothing parameter for reach-through
192 ccso InOut real Fixed C-S capacitance
193 qbm InOut real Select SGP qb formulation
194 nkf InOut real High current beta rolloff
195 xikf InOut real Temperature exponent of IKF
196 xrcx InOut real Temperature exponent of RCX
197 xrbx InOut real Temperature exponent of RBX
198 xrbp InOut real Temperature exponent of RBP
199 isrr InOut real Separate IS for fwd and rev
200 xisr InOut real Temperature exponent of ISR
201 dear InOut real Delta activation energy for ISRR
202 eap InOut real Exitivation energy for ISP
203 vbbe InOut real B-E breakdown voltage
204 nbbe InOut real B-E breakdown emission coefficient
205 ibbe InOut real B-E breakdown current
206 tvbbe1 InOut real Linear temperature coefficient of VBBE
207 tvbbe2 InOut real Quadratic temperature coefficient of VBBE
208 tnbbe InOut real Temperature coefficient of NBBE
209 ebbe InOut real exp(-VBBE/(NBBE*Vtv))
210 dtemp InOut real Locale Temperature difference
211 vers InOut real Revision Version
212 vref InOut real Reference Version
31.6. MOSFETS 685
31.6 MOSFETs
31.6.1 MOS1 - Level 1 MOSFET model with Meyer capacitance

model
31.6.1.1 MOS1 instance parameters

4 ad InOut real Drain area
3 as InOut real Source area
6 pd InOut real Drain perimeter
5 ps InOut real Source perimeter
8 nrd InOut real Drain squares
7 nrs InOut real Source squares
9 off In flag Device initially off
12 icvds InOut real Initial D-S voltage
13 icvgs InOut real Initial G-S voltage
11 icvbs InOut real Initial B-S voltage
10 ic In real vector Vector of D-S, G-S, B-S voltages
15 sens_l In flag flag to request sensitivity WRT length
14 sens_w In flag flag to request sensitivity WRT width
215 id Out real Drain current
18 is Out real Source current
17 ig Out real Gate current
16 ib Out real Bulk current
217 ibd Out real B-D junction current
216 ibs Out real B-S junction current
231 vgs Out real Gate-Source voltage
232 vds Out real Drain-Source voltage
230 vbs Out real Bulk-Source voltage
229 vbd Out real Bulk-Drain voltage
203 dnode Out integer Number of the drain node
204 gnode Out integer Number of the gate node
205 snode Out integer Number of the source node
206 bnode Out integer Number of the node
207 dnodeprime Out integer Number of int. drain node
208 snodeprime Out integer Number of int. source node
211 von Out real
212 vdsat Out real Saturation drain voltage
213 sourcevcrit Out real Critical source voltage

214 drainvcrit Out real Critical drain voltage
258 rs Out real Source resistance
209 sourceconductance Out real Conductance of source
259 rd Out real Drain conductance
210 drainconductance Out real Conductance of drain
219 gm Out real Transconductance
220 gds Out real Drain-Source conductance
218 gmb Out real Bulk-Source transconductance
218 gmbs Out real
221 gbd Out real Bulk-Drain conductance
222 gbs Out real Bulk-Source conductance
223 cbd Out real Bulk-Drain capacitance
224 cbs Out real Bulk-Source capacitance
233 cgs Out real Gate-Source capacitance
236 cgd Out real Gate-Drain capacitance
239 cgb Out real Gate-Bulk capacitance
235 cqgs Out real Capacitance due to gate-source charge storage
238 cqgd Out real Capacitance due to gate-drain charge storage
241 cqgb Out real Capacitance due to gate-bulk charge storage
243 cqbd Out real Capacitance due to bulk-drain charge storage
245 cqbs Out real Capacitance due to bulk-source charge storage
225 cbd0 Out real Zero-Bias B-D junction capacitance
226 cbdsw0 Out real
227 cbs0 Out real Zero-Bias B-S junction capacitance
228 cbssw0 Out real
234 qgs Out real Gate-Source charge storage
237 qgd Out real Gate-Drain charge storage
240 qgb Out real Gate-Bulk charge storage
242 qbd Out real Bulk-Drain charge storage
244 qbs Out real Bulk-Source charge storage
19 p Out real Instaneous power
256 sens_l_dc Out real dc sensitivity wrt length
246 sens_l_real Out real real part of ac sensitivity wrt length
247 sens_l_imag Out real imag part of ac sensitivity wrt length
248 sens_l_mag Out real sensitivity wrt l of ac magnitude
249 sens_l_ph Out real sensitivity wrt l of ac phase
250 sens_l_cplx Out complex ac sensitivity wrt length
257 sens_w_dc Out real dc sensitivity wrt width
251 sens_w_real Out real real part of ac sensitivity wrt width
252 sens_w_imag Out real imag part of ac sensitivity wrt width
253 sens_w_mag Out real sensitivity wrt w of ac magnitude
254 sens_w_ph Out real sensitivity wrt w of ac phase
255 sens_w_cplx Out complex ac sensitivity wrt width
31.6. MOSFETS 687
31.6.1.2 MOS1 model parameters

133 type Out string N-channel or P-channel MOS
101 vto InOut real Threshold voltage
101 vt0 InOut real
102 kp InOut real Transconductance parameter
103 gamma InOut real Bulk threshold parameter
104 phi InOut real Surface potential
105 lambda InOut real Channel length modulation
106 rd InOut real Drain ohmic resistance
107 rs InOut real Source ohmic resistance
108 cbd InOut real B-D junction capacitance
109 cbs InOut real B-S junction capacitance
110 is InOut real Bulk junction sat. current
111 pb InOut real Bulk junction potential
112 cgso InOut real Gate-source overlap cap.
113 cgdo InOut real Gate-drain overlap cap.
114 cgbo InOut real Gate-bulk overlap cap.
115 cj InOut real Bottom junction cap per area
116 mj InOut real Bottom grading coefficient
117 cjsw InOut real Side junction cap per area
118 mjsw InOut real Side grading coefficient
119 js InOut real Bulk jct. sat. current density
120 tox InOut real Oxide thickness
121 ld InOut real Lateral diffusion
123 u0 InOut real Surface mobility
123 uo InOut real
124 fc InOut real Forward bias jct. fit parm.
128 nmos In flag N type MOSfet model
129 pmos In flag P type MOSfet model
125 nsub InOut real Substrate doping
126 tpg InOut integer Gate type
127 nss InOut real Surface state density

model
31.6.2.1 MOS 2 instance parameters

34 cd Out real
15 sens_l In flag flag to request sensitivity WRT
length
14 sens_w In flag flag to request sensitivity WRT
width
22 dnode Out integer Number of drain node
23 gnode Out integer Number of gate node
24 snode Out integer Number of source node
25 bnode Out integer Number of bulk node
26 dnodeprime Out integer Number of internal drain node
27 snodeprime Out integer Number of internal source node
30 von Out real
31.6. MOSFETS 689

28 sourceconductance Out real Source conductance
79 rd Out real Drain resistance
29 drainconductance Out real Drain conductance
37 gmbs Out real
44 cbd0 Out real Zero-Bias B-D junction
capacitance
45 cbdsw0 Out real
46 cbs0 Out real Zero-Bias B-S junction
capacitance
47 cbssw0 Out real
54 cqgs Out real Capacitance due to gate-source
charge storage
57 cqgd Out real Capacitance due to gate-drain
charge storage
60 cqgb Out real Capacitance due to gate-bulk
charge storage
62 cqbd Out real Capacitance due to bulk-drain
charge storage
64 cqbs Out real Capacitance due to bulk-source
charge storage
70 sens_l_real Out real real part of ac sensitivity wrt
length
71 sens_l_imag Out real imag part of ac sensitivity wrt
length
65 sens_w_real Out real dc sensitivity and real part of ac

sensitivity wrt width
66 sens_w_imag Out real imag part of ac sensitivity wrt
width

101 vt0 InOut real
105 lambda InOut real Channel length modulation
123 uo InOut real
129 delta InOut real Width effect on threshold
130 uexp InOut real Crit. field exp for mob. deg.
134 ucrit InOut real Crit. field for mob. degradation
31.6. MOSFETS 691
131 vmax InOut real Maximum carrier drift velocity

132 xj InOut real Junction depth
133 neff InOut real Total channel charge coeff.
128 nfs InOut real Fast surface state density

model

34 cd Out real Drain current
10 ic InOut real vector Vector of D-S, G-S, B-S voltages
30 von Out real Turn-on voltage
31.6. MOSFETS 693

37 gmbs Out real Bulk-Source transconductance
45 cbdsw0 Out real Zero-Bias B-D sidewall capacitance
47 cbssw0 Out real Zero-Bias B-S sidewall capacitance
31.6. MOSFETS 695

101 vt0 InOut real
145 xl InOut real Length mask adjustment
146 wd InOut real Width Narrowing (Diffusion)
147 xw InOut real Width mask adjustment
148 delvto InOut real Threshold voltage Adjust
148 delvt0 InOut real
122 uo InOut real
138 xd InOut real Depletion layer width
139 alpha InOut real Alpha
127 eta InOut real Vds dependence of threshold voltage
140 input_delta InOut real
130 theta InOut real Vgs dependence on mobility
132 kappa InOut real Kappa

model

216 ibs Out real B-S junction capacitance
217 ibd Out real B-D junction capacitance
203 dnode Out integer Number of the drain node
204 gnode Out integer Number of the gate node
205 snode Out integer Number of the source node
206 bnode Out integer Number of the node
207 dnodeprime Out integer Number of int. drain node
208 snodeprime Out integer Number of int. source node
31.6. MOSFETS 697

226 cbdsw0 Out real
228 cbssw0 Out real
19 p Out real Instaneous power

101 vt0 InOut real
102 kv InOut real Saturation voltage factor
103 nv InOut real Saturation voltage coeff.
104 kc InOut real Saturation current factor
105 nc InOut real Saturation current coeff.
106 nvth InOut real Threshold voltage coeff.
107 ps InOut real Sat. current modification par.
109 gamma1 InOut real Bulk threshold parameter 1
110 sigma InOut real Static feedback effect par.
112 lambda InOut real Channel length modulation param.
113 lambda0 InOut real Channel length modulation param. 0
114 lambda1 InOut real Channel length modulation param. 1
132 uo InOut real
31.6. MOSFETS 699
31.6.5 MOS9 - Modified Level 3 MOSFET model


10 ic InOut real vector Vector of D-S, G-S, B-S voltages
81 dtemp InOut real Instance operating temperature difference

45 cbdsw0 Out real Zero-Bias B-D sidewall capacitance
47 cbssw0 Out real Zero-Bias B-S sidewall capacitance
31.6. MOSFETS 701

101 vt0 InOut real
145 xl InOut real Length mask adjustment
146 wd InOut real Width Narrowing (Diffusion)
147 xw InOut real Width mask adjustment
148 delvto InOut real Threshold voltage Adjust
148 delvt0 InOut real
122 uo InOut real
138 xd InOut real Depletion layer width
139 alpha InOut real Alpha
127 eta InOut real Vds dependence of threshold voltage
140 input_delta InOut real
130 theta InOut real Vgs dependence on mobility
132 kappa InOut real Kappa
31.6. MOSFETS 703
31.6.6 BSIM1 - Berkeley Short Channel IGFET Model

31.6.6.1 BSIM1 instance parameters

8 nrd InOut real Number of squares in drain
7 nrs InOut real Number of squares in source
9 off InOut flag Device is initially off
11 vds InOut real Initial D-S voltage
12 vgs InOut real Initial G-S voltage
10 vbs InOut real Initial B-S voltage
13 ic In unknown vector Vector of DS,GS,BS initial voltages
31.6.6.2 BSIM1 Model Parameters

101 vfb InOut real Flat band voltage
102 lvfb InOut real Length dependence of vfb
103 wvfb InOut real Width dependence of vfb
104 phi InOut real Strong inversion surface potential
105 lphi InOut real Length dependence of phi
106 wphi InOut real Width dependence of phi
107 k1 InOut real Bulk effect coefficient 1
108 lk1 InOut real Length dependence of k1
109 wk1 InOut real Width dependence of k1
113 eta InOut real VDS dependence of threshold voltage
114 leta InOut real Length dependence of eta
115 weta InOut real Width dependence of eta
116 x2e InOut real VBS dependence of eta
117 lx2e InOut real Length dependence of x2e
118 wx2e InOut real Width dependence of x2e
119 x3e InOut real VDS dependence of eta
120 lx3e InOut real Length dependence of x3e
121 wx3e InOut real Width dependence of x3e
122 dl InOut real Channel length reduction in um
123 dw InOut real Channel width reduction in um
124 muz InOut real Zero field mobility at VDS=0 VGS=VTH

125 x2mz InOut real VBS dependence of muz
126 lx2mz InOut real Length dependence of x2mz
127 wx2mz InOut real Width dependence of x2mz
128 mus InOut real Mobility at VDS=VDD VGS=VTH, channel length modulation
129 lmus InOut real Length dependence of mus
130 wmus InOut real Width dependence of mus
131 x2ms InOut real VBS dependence of mus
132 lx2ms InOut real Length dependence of x2ms
133 wx2ms InOut real Width dependence of x2ms
134 x3ms InOut real VDS dependence of mus
135 lx3ms InOut real Length dependence of x3ms
136 wx3ms InOut real Width dependence of x3ms
137 u0 InOut real VGS dependence of mobility
138 lu0 InOut real Length dependence of u0
139 wu0 InOut real Width dependence of u0
140 x2u0 InOut real VBS dependence of u0
141 lx2u0 InOut real Length dependence of x2u0
142 wx2u0 InOut real Width dependence of x2u0
143 u1 InOut real VDS depence of mobility, velocity saturation
146 x2u1 InOut real VBS depence of u1
147 lx2u1 InOut real Length depence of x2u1
148 wx2u1 InOut real Width depence of x2u1
149 x3u1 InOut real VDS depence of u1
150 lx3u1 InOut real Length dependence of x3u1
151 wx3u1 InOut real Width depence of x3u1
152 n0 InOut real Subthreshold slope
153 ln0 InOut real Length dependence of n0
154 wn0 InOut real Width dependence of n0
155 nb InOut real VBS dependence of subthreshold slope
156 lnb InOut real Length dependence of nb
157 wnb InOut real Width dependence of nb
158 nd InOut real VDS dependence of subthreshold slope
159 lnd InOut real Length dependence of nd
160 wnd InOut real Width dependence of nd
161 tox InOut real Gate oxide thickness in um
162 temp InOut real Temperature in degree Celcius
163 vdd InOut real Supply voltage to specify mus
164 cgso InOut real Gate source overlap capacitance per unit channel width(m)
165 cgdo InOut real Gate drain overlap capacitance per unit channel width(m)
166 cgbo InOut real Gate bulk overlap capacitance per unit channel length(m)
167 xpart InOut real Flag for channel charge partitioning
168 rsh InOut real Source drain diffusion sheet resistance in ohm per square
169 js InOut real Source drain junction saturation current per unit area
31.6. MOSFETS 705
170 pb InOut real Source drain junction built in potential

171 mj InOut real Source drain bottom junction capacitance grading coefficient
172 pbsw InOut real Source drain side junction capacitance built in potential
173 mjsw InOut real Source drain side junction capacitance grading coefficient
174 cj InOut real Source drain bottom junction capacitance per unit area
175 cjsw InOut real Source drain side junction capacitance per unit area
176 wdf InOut real Default width of source drain diffusion in um
177 dell InOut real Length reduction of source drain diffusion
178 nmos In flag Flag to indicate NMOS
179 pmos In flag Flag to indicate PMOS
31.6.7 BSIM2 - Berkeley Short Channel IGFET Model

31.6.7.1 BSIM2 instance parameters

8 nrd InOut real Number of squares in drain
7 nrs InOut real Number of squares in source
9 off InOut flag Device is initially off
11 vds InOut real Initial D-S voltage
12 vgs InOut real Initial G-S voltage
10 vbs InOut real Initial B-S voltage
13 ic In unknown vector Vector of DS,GS,BS initial voltages
31.6.7.2 BSIM2 model parameters

101 vfb InOut real Flat band voltage
102 lvfb InOut real Length dependence of vfb
103 wvfb InOut real Width dependence of vfb
104 phi InOut real Strong inversion surface potential
105 lphi InOut real Length dependence of phi
106 wphi InOut real Width dependence of phi
113 eta0 InOut real VDS dependence of threshold voltage at VDD=0
114 leta0 InOut real Length dependence of eta0
115 weta0 InOut real Width dependence of eta0
116 etab InOut real VBS dependence of eta
117 letab InOut real Length dependence of etab
118 wetab InOut real Width dependence of etab
119 dl InOut real Channel length reduction in um
120 dw InOut real Channel width reduction in um
121 mu0 InOut real Low-field mobility, at VDS=0 VGS=VTH
122 mu0b InOut real VBS dependence of low-field mobility
123 lmu0b InOut real Length dependence of mu0b
31.6. MOSFETS 707
124 wmu0b InOut real Width dependence of mu0b

125 mus0 InOut real Mobility at VDS=VDD VGS=VTH
126 lmus0 InOut real Length dependence of mus0
127 wmus0 InOut real Width dependence of mus
128 musb InOut real VBS dependence of mus
129 lmusb InOut real Length dependence of musb
130 wmusb InOut real Width dependence of musb
131 mu20 InOut real VDS dependence of mu in tanh term
132 lmu20 InOut real Length dependence of mu20
133 wmu20 InOut real Width dependence of mu20
134 mu2b InOut real VBS dependence of mu2
137 mu2g InOut real VGS dependence of mu2
138 lmu2g InOut real Length dependence of mu2g
139 wmu2g InOut real Width dependence of mu2g
140 mu30 InOut real VDS dependence of mu in linear term
149 mu40 InOut real VDS dependence of mu in linear term
158 ua0 InOut real Linear VGS dependence of mobility
159 lua0 InOut real Length dependence of ua0
160 wua0 InOut real Width dependence of ua0
161 uab InOut real VBS dependence of ua
162 luab InOut real Length dependence of uab
163 wuab InOut real Width dependence of uab
164 ub0 InOut real Quadratic VGS dependence of mobility
165 lub0 InOut real Length dependence of ub0
166 wub0 InOut real Width dependence of ub0
167 ubb InOut real VBS dependence of ub
168 lubb InOut real Length dependence of ubb
169 wubb InOut real Width dependence of ubb
170 u10 InOut real VDS depence of mobility

173 u1b InOut real VBS depence of u1
174 lu1b InOut real Length depence of u1b
175 wu1b InOut real Width depence of u1b
176 u1d InOut real VDS depence of u1
177 lu1d InOut real Length depence of u1d
178 wu1d InOut real Width depence of u1d
179 n0 InOut real Subthreshold slope at VDS=0 VBS=0
180 ln0 InOut real Length dependence of n0
181 wn0 InOut real Width dependence of n0
182 nb InOut real VBS dependence of n
183 lnb InOut real Length dependence of nb
184 wnb InOut real Width dependence of nb
185 nd InOut real VDS dependence of n
186 lnd InOut real Length dependence of nd
187 wnd InOut real Width dependence of nd
188 vof0 InOut real Threshold voltage offset AT VDS=0 VBS=0
189 lvof0 InOut real Length dependence of vof0
190 wvof0 InOut real Width dependence of vof0
191 vofb InOut real VBS dependence of vof
192 lvofb InOut real Length dependence of vofb
193 wvofb InOut real Width dependence of vofb
194 vofd InOut real VDS dependence of vof
195 lvofd InOut real Length dependence of vofd
196 wvofd InOut real Width dependence of vofd
197 ai0 InOut real Pre-factor of hot-electron effect.
198 lai0 InOut real Length dependence of ai0
199 wai0 InOut real Width dependence of ai0
200 aib InOut real VBS dependence of ai
201 laib InOut real Length dependence of aib
202 waib InOut real Width dependence of aib
203 bi0 InOut real Exponential factor of hot-electron effect.
204 lbi0 InOut real Length dependence of bi0
205 wbi0 InOut real Width dependence of bi0
206 bib InOut real VBS dependence of bi
207 lbib InOut real Length dependence of bib
208 wbib InOut real Width dependence of bib
209 vghigh InOut real Upper bound of the cubic spline function.
210 lvghigh InOut real Length dependence of vghigh
211 wvghigh InOut real Width dependence of vghigh
212 vglow InOut real Lower bound of the cubic spline function.
213 lvglow InOut real Length dependence of vglow
214 wvglow InOut real Width dependence of vglow
215 tox InOut real Gate oxide thickness in um
31.6. MOSFETS 709
216 temp InOut real Temperature in degree Celcius

217 vdd InOut real Maximum Vds
218 vgg InOut real Maximum Vgs
219 vbb InOut real Maximum Vbs
220 cgso InOut real Gate source overlap capacitance per unit channel width(m)
221 cgdo InOut real Gate drain overlap capacitance per unit channel width(m)
222 cgbo InOut real Gate bulk overlap capacitance per unit channel length(m)
223 xpart InOut real Flag for channel charge partitioning
224 rsh InOut real Source drain diffusion sheet resistance in ohm per square
225 js InOut real Source drain junction saturation current per unit area
226 pb InOut real Source drain junction built in potential
227 mj InOut real Source drain bottom junction capacitance grading coefficient
228 pbsw InOut real Source drain side junction capacitance built in potential
229 mjsw InOut real Source drain side junction capacitance grading coefficient
230 cj InOut real Source drain bottom junction capacitance per unit area
231 cjsw InOut real Source drain side junction capacitance per unit area
232 wdf InOut real Default width of source drain diffusion in um
233 dell InOut real Length reduction of source drain diffusion
234 nmos In flag Flag to indicate NMOS
235 pmos In flag Flag to indicate PMOS
31.6.8 BSIM3
The accessible device parameters (see Chapt. 31.1 for the syntax) are listed here.
31.6.8.1 BSIM3 accessible instance parameters

8 vdsat Out real Saturation voltage
9 vth Out real Threshold voltage
10 ibd Out real
11 ibs Out real
12 gbd Out real
13 gbs Out real
14 qb Out real Qbulk
15 cqb Out real
16 qg Out real Qgate
17 cqg Out real
18 qd Out real Qdrain
19 cqd Out real
20 cgg Out real
21 cgd Out real
22 cgs Out real
23 cdg Out real
24 cdd Out real
25 cds Out real
26 cbg Out real
27 cbd Out real
28 cbs Out real
29 capbd Out real Diode capacitance
30 capbs Out real Diode capacitance
The parameters are available in the BSIM3 models (level=8 or level=49) version=3.2.4
and version=3.3.0 only. Negative capacitance values may occur, depending on the internal
calculation. Please see the note in Chapt. 31.6.9.1.
31.6.8.2 BSIM3 manual
Further detailed descriptions will not be given here. Unfortunately the details on these
parameters are not documented, even not in the otherwise excellent pdf manual issued
31.6. MOSFETS 711
by University of California at Berkeley.
31.6.9 BSIM4
The accessible device parameters (see Chapt. 31.1 for the syntax) are listed here.
31.6.9.1 BSIM4 accessible instance parameters

gmbs Out real Body effect (Back gate) transconductance
gm Out real Transconductance
gds Out real Drain-Source conductance
vdsat Out real Saturation voltage
vth Out real Threshold voltage
id Out real Drain current
ibd Out real Diode current
ibs Out real Diode current
gbd Out real Diode conductance
gbs Out real Diode conductance
isub Out real Substrate current
igidl Out real Gate-Induced Drain Leakage current
igisl Out real Gate-Induced Source Leakage current
igs Out real Gate-Source current
igd Out real Gate-drain current
igb Out real Gate-Bulk current
igcs Out real
vbs Out real Bulk-Source voltage
vgs Out real Gate-Source voltage
vds Out real Drain-Source voltage
cgg Out real
cgs Out real
cgd Out real
cbg Out real
cbd Out real
cbs Out real
cdg Out real
cdd Out real
cds Out real
csg Out real
csd Out real
css Out real
cgb Out real
cdb Out real
csb Out real
cbb Out real
capbd Out real Diode capacitance

capbs Out real Diode capacitance
qg Out real Gate charge
qb Out real Bulk charge
qd Out real Drain charge
qs Out real
qinv Out real
qdef Out real
gcrg Out real
gtau Out real
The parameters are available in all BSIM4 models (level=14 or level=54) version=4.2.1
to version=4.8.
Negative capacitance values may occur, depending on the internal calculation. To compare
with measured data, please just use the absolute values of the capacitance data. For an
explanation of negative values and the basics on how capacitance values are evaluated in
a BSIM model, please refer to the book BSIM4 and MOSFET Modeling for IC Simulation
by Liu and Hu, Chapt. 5.2.
31.6.9.2 BSIM4 manual
Detailed descriptions will not be given here. Unfortunately the details on these parameters
are not documented, even not in the otherwise excellent pdf manual issued by University
of California at Berkeley.
Chapter 32
Compilation notes
This file describes the procedures to install ngspice from sources.

Some special considerations are required when Verilog-A models are to be included (see
chapter 13.2).
32.1 Ngspice Installation under Linux (and other ’UNIXes’)
32.1.1 Prerequisites
Ngspice is written in C and thus a complete C compilation environment is needed. Ngspice
is developed on GNU/Linux with autotools, gcc, and GNU make.
The following software must be installed in your system to compile ngspice: bison, flex,
and X11 (and Xaw, Xmu, Xext, Xft, FontConfig, Xrender, and freetype) headers
(e.g. libX11-devel) and libs (e.g. libX11-6).
The X11 headers and libraries are typically available in an X11 development package from
your Linux distribution.
If you want to compile the source code from Git, you will need additional software:
autoconf, automake, libtool.
For your convenience you always should add readline (or editline) libs and headers.
If you intend to make tclspice (see chapt. 20), you will need tcl/tk and blt.
If you want to have high performance and accurate FFT’s you should install: fftw-3.
The ngspice configure script will find the library and will induce the build process to link
against it.
32.1.2 Install from Git

This section describes how to install from source code taken direct from Git. This will
give you access to the most recent enhancements and corrections. However be careful
as the code in Git may be under development and thus still unstable. For user install
instructions using source from released distributions, please see the sections titled ’Install
from tarball’ (32.1.3) and ’Advanced Install’ (32.1.8).
713
714 CHAPTER 32. COMPILATION NOTES
Download source from Git as described on the sourceforge ngspice Git page. Define and
enter a directory of your choice, e.g. /home/myname/software/. Download the complete
ngspice repository from Git, for example by anonymous access issuing the command
or via http protocol
git clone http://git.code.sf.net/p/ngspice/ngspice
You will find the sources in directory /home/myname/software/ngspice. Now enter the
ngspice top level directory ngspice (where the installation instruction file INSTALL can
be found).
The project uses the GNU build process. You should be able to do the following:
$ ./compile_linux.sh
This script will run autogen.sh, create a release directory, run ./configure, clean, make and
make install, all with suitable parameters to compile a 64 bit version of ngspice, including
the XSPICE code models.
A suitable manual approach for compiling (without release directory) might be:
$ ./autogen.sh
$ ./configure --enable-xspice --enable-klu
--disable-debug --with-readline=yes CFLAGS="-m64 -O2" LDFLAGS="-m64 -s"
$ make clean
$ make
$ sudo make install
See the section titled ’Advanced Install’ (32.1.8) for instructions about arguments that
can be passed to ./configure to customize the build and installation. The following
arguments are already used here and may be called sort of ‘standard’, and they are used
by ./compile_linux.sh:
--with-x Include the X11 graphics interface.
--enable-xspice Include the XSPICE extensions (see Chapt. 12 and 28).
--enable-klu Include the new KLU matrix solver in addition to Sparse 1.3.
--disable-debug No debugging information included (optimized and compact code).
--with-readline=yes Include an editor for the input command line (command history,
backspace, insert etc.). If readline is not available, editline may be used.
--enable-openmp Compile ngspice for multi-core processors. Paralleling is done by
OpenMP (see Chapt. 16.10), and is enabled for certain MOS models.
-–enable-osdi Enable the OSDI interface for loading Verilog-A compact device models
compiled by OpenVAF (see 13.2).
--enable-cider Include CIDER numerical device simulator (see Chapt. 30).
32.1. NGSPICE INSTALLATION UNDER LINUX (AND OTHER ’UNIXES’) 715
CFLAGS="-m64 -O2" LDFLAGS="-m64 -s" will enable a 64 bit build (-m64) and stress the
optmisation (-O2). -s will yield a minimum size executable (debug information stripped).
On most systems --disable-debug will have the same effect. A 32bit build can be made
if all 32 bit tools (compiler etc.) are installed and -m32 is given instead of -m64.
$make clean may sometimes help avoiding mixing up old and newly created object files.
For your convenience a shell script compile_linux.sh is available in ngspice directory. to
be started with ./compile_linux.sh 64 for a 64 bit build.
If a problem is found with the build process, please submit a report to the Ngspice
development team. Please provide information about your system and any ./configure
arguments you are using, together with any error messages. Ideally you would have tried
to fix the problem yourself first. If you have fixed the problem then the development team
will love to hear from you.
If you need updating your local source code tree from Git, just enter ngspice directory
and issue the command
git pull
git reset --hard
To learn more about Git, which can be both powerful and difficult to master, please
consult http://git-scm.com/, especially: http://git-scm.com/documentation, which has
links to documentation and tutorials.
32.1.3 Install from a tarball, e.g. from ngspice-42.tar.gz

This covers installation from a tarball (for example ngspice-39.tar.gz, to be found at
http://sourceforge.net/projects/ngspice/files/ng-spice-rework/39/). After downloading the
tar ball to a local directory unpack it using:
$ tar -zxvf ngspice-42.tar.gz
Now change directories in to the top-level source directory (where this text from the
INSTALL file can be found).
You should be able to do:
$ ./configure --enable-xspice --enable-osdi --disable-debug --with-readline=yes
$ make clean
$ make
$ sudo make install
The default install dir is /usr/local/bin
See the section titled ’Advanced Install’ (32.1.8) for instructions about arguments that
can be passed to ./configure to customize the build and installation.
32.1.4 Compilation using an user defined directory tree for ob-

ject files
The procedures described above will store the *.o files (output of the compilation step)
into the directories where the sources (*.c) are located. This may not be the best option
if you want for example to maintain a debug version and in parallel a release version of
ngspice (./configure --disable-debug). So if you intend to create a separate object
file tree like ngspice/ngbuild/release, you may do the following, starting from the default
directory ngspice:
mkdir -p release
cd release
../configure --enable-xspice --disable-debug --with-readline=yes <more options>
make install
This will create an object file directory tree, similar to the source file directory tree, the
object files are now separated from the source files. For the debug version, you may do
the same as described above, replacing ’release’ by ’debug’, and obtain another separated
object file directory tree. If you already have run ./configure in ngspice, you have to do a
maintainer-clean, before the above procedure will work. The script ./compile_linux.sh
is made according to the procedure described above.
32.1.5 ngspice as a shared library
From the tarball (for example ngspice-42.tar.gz, see above), with the GNU build process
and the following options selected:
$ ./configure --with-ngshared --enable-xspice --enable-cider
--enable-openmp --disable-debug --enable-osdi --enable-klu
$ make clean
$ make
$ sudo make install
you will get the ngspice shared library. A file ngspice.pc for pkg-config is generated.
$make clean may sometimes help avoiding mixing up old and newly created object files.
It is required if you make both shared and standard ngspice from the same setup.
With sources from git you have to do:
$ ./autogen.sh
$ ./configure --with-ngshared --enable-xspice --enable-klu
--enable-openmp --enable-osdi --disable-debug --enable-cider
$ make clean
$ make
$ sudo make install
32.1.6 Relative paths for spinit and code models

The ./configure option
$ ./configure --enable-relpath
deserves some extra mentioning:
It sets relative search paths for the file spinit and the XSPICE code models *.cm. spinit
will be look up in ../share/ngspice/scripts. The search path for the code models (as
set by the parameter to the codemodel command in spinit) is set to ../lib/ngspice. The
binary is found in ../bin. All these paths are relative to the current directory. Under
MS Windows, this is the directory of ngspice.exe as per default, but may be set to any
other directory with the cd (chapt. 17.5.12) command.
The install path for the ngspice executable is determined by the –-prefix flag of ./configure.
The current directory for the ngspice shared library is determined by the calling program.
32.1.7 Installation on Red Hat or Oracle Linux (and similar,

e.g. Centos)
These OSs, widely distributed among commercial users, require some special considera-
tions. There is an extra document, NGSPICE on Red Hat Like Distributions.pdf,
provided by Justin Fisher, available with the ngspice distribution.
32.1.8 Advanced Install

Some extra options can be provided to ./configure. To get all available options do:
$ ./configure --help
Some of these options are generic to the GNU build process that is used by Ngspice, other
are specific to Ngspice.
The following sections provide some guidance and descriptions for many, but not all, of
these options.
32.1.8.1 Options Specific to Using Ngspice
--enable-openmp Compile ngspice for multi-core processors. Paralleling is done by

OpenMP (see Chapt. 16.10).
--enable-xspice Enable XSPICE enhancements, yielding a mixed signal simulator inte-
grated into ngspice with codemodel dynamic loading support. See Chapt. 12 and section
II for details.
-–enable-osdi Enable the OSDI interface for loading Verilog-A compact device models
compiled by OpenVAF (see 13.2)
-–enable-klu Enable the KLU matrix solver
--with-readline=yes Enable GNU readline support for the command line interface.
--enable-cider Cider is a mixed-level simulator that couples Spice3 and DSIM to simu-
late devices from their technological parameters. This part of the simulator is not compiled
by default.
--with-editline=yes Enables the use of the BSD editline library (libedit).
See http://www.thrysoee.dk/editline/. To be used instead of --with-readline=yes.
--without-x Disable the X-Windows graphical system. Compile without needing X
headers and X libraries. The plot command (17.5.56) is now disabled. You may use
Gnuplot (17.5.38) instead.
--with-tcl=tcldir When configured with this option the tcl module ‘tclspice’ is com-
piled and installed instead of plain ngspice.
--with-ngshared This option will create a shared library (*.so in Linux) or dynamic link
library (*.dll) instead of plain ngspice.
--enable-relpath This options introduces a search path for spinit relative to the calling
executable (ngspice or the caller using the ngspice shared library) as ../share/ngspice.
In spinit the search path for code models is also set as relative as ../lib. This option
may be effective especially when not using standard installation paths in Linux, but
especially for ngspice.dll under MS Windows, if to be installed in other directories than
in C:\Spice64.
--disable-debug This option will remove the ’-g’ option passed to the compiler. This
speeds up execution time, creates a small executable, and is recommended for normal use.
If you want to run ngspice in a debugger (e.g. gdb), you should not select this option.
--enable-pss This is an experimental feature to enable Periodic Steady State Analysis.
--enable-oldapps Beginning with ngspice-28, only ngspice executable is made. If you
need old apps like ngnutmeg, ngmakeidx, ngmultidec, ngproc2mod, ngsconvert, use this
./configure flag.
--with-fftw3=no Do not check for and use the fftw fast fourier transform library (www.fftw.org).
Use an internal fft algorithm instead. Default is yes.
--disable-utf8 Switch off UNICODE support, use extended ASCII with Western char-
acter support instead.
--disable-sp Switch off RF support: no integrated S-parameter simulation, no RF noise
simulation (15.3.8).
--enable-shortcheck Enables a ’make check’ with strongly reduced runtime. Besides
some regression tests only BSIM3 and BSM4 devices are checked.
32.1.8.2 Options for experimental usage only
The following options are seldom used today, not tested, some may even no longer be
implemented (correctly) and lead to errors.
--enable-capbypass Bypass calculation of cbd/cbs in the mosfets if the vbs/vbd volt-
ages are unchanged.
--enable-capzerobypass Bypass all the cbd/cbs calculations if Czero is zero. This is
enabled by default since rework-18.
--enable-cluster Clustering code for distributed simulation. This is a contribution

never tested. This code comes from TCLspice implementation and is implemented for
transient analysis only.
--enable-expdevices Enable experimental devices. This option is used by developers
to mask devices under development. Almost useless for users.
--enable-experimental This may be used to enable some experimental code. The
code has to be encapsuated into #ifdef EXPERIMENTAL_CODE ... #endif constructs.
Currently there is no such code available.
--enable-help Force building nghelp. This is deprecated.
--enable-newpred Enable the NEWPRED symbol in the code.
--enable-newtrunc Enable the newtrunc option
--enable-nodelimiting Experimental damping scheme
--enable-nobypass Don’t bypass recalculations of slowly changing variables
--enable-nosqrt Use always log/exp for non-linear capacitances --enable-predictor
Enable a predictor method for convergence
--enable-sense2 Use spice2 sensitivity analysis
32.1.8.3 Options useful only for debugging specific issues in ngspice
The following options are seldom used today, not tested, some may even no longer be
implemented. Only experienced users should switch on these options, often they are
effective only in conjunction with looking at the respective source code.
--enable-ansi Configure will try to find an option for your compiler so that it expects
ansi-C.
--enable-asdebug Debug sensitivity code *ASDEBUG*.
--enable-blktmsdebug Debug distortion code *BLOCKTIMES*
--enable-checkergcc Option for compilation with checkergcc.
--enable-cpdebug Enable ngspice shell code debug.
--enable-ftedebug Enable ngspice frontend debug.
--enable-gc Enable the Boehm-Weiser Conservative Garbage Collector.
--enable-pzdebug Debug pole/zero code.
--enable-sensdebug Debug sensitivity code *SENSDEBUG*.
--enable-smltmsdebug Debug distortion code *SMALLTIMES*
--enable-smoketest Enable smoketest compile.
--enable-stepdebug Turns on debugging of convergence steps in transient analysis
32.1.9 Compilers and Options

Some systems require unusual options for compilation or linking that the configure script
does not know about. You can give configure initial values for variables by setting them
in the environment. Using a Bourne-compatible shell, you can do that on the command
line like this:
CC=c89
CFLAGS=-O2
LIBS=-lposix
./configure
Or on systems that have the env program, you can do it like this:
env CPPFLAGS=-I/usr/local/include
LDFLAGS=-s
./configure
32.1.10 Compiling For Multiple Architectures

You can compile the package for more than one kind of computer at the same time, by
placing the object files for each architecture in their own directory. To do this, you must
use a version of make that supports the VPATH variable, such as GNU make. cd to the
directory where you want the object files and executables to go and run the configure
script. configure automatically checks for the source code in the directory that configure
is in and in ‘..’.
If you have to use a make that does not supports the VPATH variable, you have to compile
the package for one architecture at a time in the source code directory. After you have
installed the package for one architecture, use make distclean before reconfiguring for
another architecture.
32.1.11 Installation Names

By default, make install will install the package’s files in /usr/local/bin, /usr/local/man,
etc. You can specify an installation prefix other than /usr/local by giving configure the
option –prefix=PATH.
You can specify separate installation prefixes for architecture-specific files and architecture-
independent files. If you give configure the option –exec-prefix=PATH, the package will
use PATH as the prefix for installing programs and libraries. Documentation and other
data files will still use the regular prefix.
In addition, if you use an unusual directory layout you can give options like –bindir=PATH
to specify different values for particular kinds of files. Run configure –help for a list of the
directories you can set and what kinds of files go in them.
If the package supports it, you can cause programs to be installed with an extra prefix or
suffix on their names by giving configure the option –program-prefix=PREFIX or –program-
suffix=SUFFIX.
When installed on MinGW with MSYS alternative paths are not fully supported. See
‘How to make ngspice with MINGW and MSYS’ (32.2.2) for details.
32.1.12 Optional Features
Some packages pay attention to –enable-FEATURE options to configure, where FEATURE

indicates an optional part of the package. They may also pay attention to –with-PACKAGE
options, where PACKAGE is something like gnu-as or ‘x’ (for the X Window System). The
README should mention any –enable- and –with- options that the package recognizes.
For packages that use the X Window System, configure can usually find the X include
and library files automatically, but if it doesn’t, you can use the configure options –x-
includes=DIR and –x-libraries=DIR to specify their locations.
32.1.13 Specifying the System Type
There may be some features configure can not figure out automatically, but needs to
determine by the type of host the package will run on. Usually configure can figure
that out, but if it prints a message saying it can not guess the host type, give it the
–host=TYPE option. TYPE can either be a short name for the system type, such as
‘sun4’, or a canonical name with three fields: CPU-COMPANY-SYSTEM
See the file config.sub for the possible values of each field. If config.sub isn’t included in
this package, then this package doesn’t need to know the host type.
If you are building compiler tools for cross-compiling, you can also use the –target=TYPE
option to select the type of system they will produce code for and the –build=TYPE option
to select the type of system on which you are compiling the package.
32.1.14 Sharing Defaults
If you want to set default values for configure scripts to share, you can create a site shell
script called config.site that gives default values for variables like CC, cache_file, and prefix.
configure looks for PREFIX/share/config.site if it exists, then PREFIX/etc/config.site if it
exists. Or, you can set the CONFIG_SITE environment variable to the location of the site
script. A warning: not all configure scripts look for a site script.
32.1.15 Operation Controls
configure recognizes the following options to control how it operates.

--cache-file=FILE Use and save the results of the tests in FILE instead of ./config.cache.
Set FILE to /dev/null to disable caching, for debugging configure.
--help Print a summary of the options to configure, and exit.
--quiet --silent -q Do not print messages saying which checks are being made. To
suppress all normal output, redirect it to /dev/null (any error messages will still be shown).
--srcdir=DIR Look for the package’s source code in directory DIR. Usually configure can
determine that directory automatically.
--version Print the version of Autoconf used to generate the configure script, and exit.
configure also accepts some other, not widely useful, options.
32.2 Ngspice Compilation under Windows OS
32.2.1 Building ngspice with MS Visual Studio 2022

ngspice may be compiled and linked with MS Visual Studio 2022. Version 2019 is not
compatible with the new OSDI interface. A free version is offered by Microsoft as the
Visual Studio Community Edition. XSPICE project files are located in visualc/XSPICE
and are automatically invoked if you start the build procedure. The projects are in the
format for Visual Studio 2022, but any later version of Visual Studio can upgrade the
projects to its version.
KLU, CIDER, OpenMP and XSPICE are included, as well as the code models for XSPICE
(*.cm).
After compilation the executable, code models and initialization files are available in
directory C:\ as C:\Spice, C:\Spice64, C:\Spice64, or C:\Spice64d, depending on 32 or
64 bit and release or debug. However, 32-bit is no longer supported, compilation may
fail in future releases. A typical installation tree (64-bit, release) is shown below. A
true Windows installer is however not yet available. The project’s ’home’ directory for
Windows OS (ngspice/visualc) with its files vngspice.sln (solution) and vngspice.vcxproj
(project) allows compiling and linking ngspice with MS Visual Studio.
On Windows 10 with its strict security model, some complications will arise. A normal
user is not allowed to create directories in C:\. You will need admin access rights. So
how to cope with this situation? Three different methods are listed below:
• Open and run Visual Studio as admin.
• Create the directories C:\Spice64, or C:\Spice64d as admin and give them full access
rights for the ordinary user.
• Select another storage place (e.g. D:\) to install the ngspice tree. To allow this, edit
files make-install-vngspice.bat (for 32 and 64 bit release) or make-install-vngspiced.bat
(for 32 or 64 bit debug), found in ngspice\visualc, and change lines 10 (set dst=c:\Spice)
and 40 to the new destination.
/visualc/src/include/ngspice contains a dedicated config.h file with the preprocessor defi-

nitions required to properly compile the code.
Install Microsoft Visual Studio 2022. The MS Visual Studio Community Edition (which
is available at no cost from https://www.visualstudio.com/) is fully adequate. It will
generate a 64 bit Release with or without OpenMP support and a Debug version of
ngspice, using the x64 flag. In addition you may select a console version without graphics
interface. Making ngspice with 32 bit is still possible, but is not recommended. 32 bit is
32.2. NGSPICE COMPILATION UNDER WINDOWS OS 723
available with flag Win32. Standard for everyday use are the ReleaseOMP variants (GUI
or console) for 64 bit.
Compilation of the ngspice and XSPICE codes requires the installation of FLEX and
BISON. They may be downloaded as Windows executables from winflexbison. Please
unzip the zip file and copy its content into a directory named flex-bison at the same level
as the ngspice directory. The resulting source tree then is:
D:\MySpiceSources\
ngspice\
visualc\
...
flex-bison\
...
Table 32.1: ngspice source tree under MS Windows
Procedure:
Download ngspice. You may obtain a snapshot from ngspice git page at SourceForge,
where you will find on top of the page a link named ’Download Snapshot’. On the left you
may select any of the branches which are of interest. Branch ’master’ is the most mature
code selection, branch ’pre-master’ is an actual development branch. Another approach is
to install ’git’ from git for Windows and installing ngspice source code with the command
as described in chapter 32.1.2.
Go to directory /ngspice/visualc.
Start MS Visual Studio as admin if you need to create C:\Spice64 etc and open the input
file vngspice.sln. Or start MS Visual Studio by double click on vngspice.sln if C:\Spice etc.
already exist or your have selected any other accessible stroage location (see comment
from above). After MS Visual Studio opens, select the debug or release version (with
or without OpenMP support) by checking Build, Configuration-Manager, Debug, Release
or ReleaseOMP. Start making ngspice.exe by selecting Build and Build Solution. The ex-
ecutable will be created and stored in visualc/vngspice/<configuration.platform>. Object
files will be stored to visualc/vngspice/<configuration.platform>/obj.
A simplified installation tree is created in parallel:

C:\Spice64\
bin\
ngspice.exe
vcomp14xx.dll
lib\
ngspice\
analog.cm
digital.cm
spice2poly.cm
extradev.cm
extravt.cm
table.cm
share\
ngspice\
scripts\
spinit
MSVC.CMD
vlnggen
Table 32.2: ngspice Visual Studio installation tree under MS Windows
The exact directory names depend on the configuration and platform you have selected
(C:\Spice, C:\Spice64, C:\Spiced, C:\Spice64d). If you intend to install ngspice into an-
other directory, e.g. D:\MySpice, you may simply copy the contents from C:\Spice to the
new location. This becomes possible because the paths to the code models or spinit are
set relative to ngspice.exe. As an alternative, you may edit make-install-vngspice.bat and
alter the following entries from:
set dst=c:\Spice
set dst=c:\Spice64
to
set dst=D:\MySpice
set dst=D:\MySpice64
To use the FFTW-3 library for a ’calibrated’ fast Fourier analysis with the fft command
(see 17.5.33), download the precompiled MS Windows FFTW distribution (either 32
bit or 64 bit) from http://www.fftw.org/install/windows.html. Extract at least the files
fftw3.h, libfftw3-3.def, and libfftw3-3.dll to directory ../../fftw-3.3-dll32 (from 32 bit fftw3
for ngspice 32 bit), or to directory ../../fftw-3.3-dll64 (from 64 bit fftw3 for ngspice 64 bit).
So both directories are at the same level as the ngspice directory. Then select the MS
VC++ project file visualc/vngspice-fftw.vcxproj for starting VC++, select the appropriate
configuration and platform, and off you go. This is how the installed directory tree looks
like:
D:\MySpiceSources\
ngspice\
visualc\
...
flex-bison\
...
fftw-3.3-dll32\
...
fftw-3.3-dll64\
...
Table 32.3: ngspice source tree under MS Windows (including fftw)
If you use the debugging features of Visual Studio, ngspice is started with a special
spinit file located in visualc\vngspice\share\ngspice\scripts. Your user-defined start-up
commands are best addressed in a .spiceinit file located in C:\users\<username>.
For compiling ngspice as a dll (shared library) there is a dedicated project file coming
with the source code to generate ngspice.dll. Go to the directory visualc and start the
project with double clicking on sharedspice.vcxproj.
32.2.2 How to make ngspice with MINGW and MSYS2

Creating ngspice with MINGW is a straightforward procedure, if you have MSYS2 and
MINGW installed properly. Go to https://www.msys2.org/ and install the 64-bit version
of MSYS2, e.g. to C:\msys64. There are now several ways to move on. A very nice
description of the installation procedure for all the tools required to compile some source
code is given in this link. In addition to the compiler gcc you will need the packages
libtool, autoconf, automake, bison, git, and make.
64-bit ngspice is now the standard, making 32-bit ngspice is still possible if a suitable
gcc is installed, but is not tested any more. The procedure of compiling a distribution
(for example, the most recent stable distribution from the ngspice website, e.g. ngspice-
39.tar.gz), is as follows:
$ cd ngspice
$ mkdir release
$ cd release
$ ../configure --with-wingui ...and other options (32.1.8.1)
$ make
$ make install
The useful options to ../configure are
--enable-xspice
--enable-osdi
--enable-klu
--enable-openmp
--enable-cider
--disable-debug (-O2 optimization, no debug information)
An option to make is
-j8
If you have a processor with 4 real (or 8 logical) cores, this will speed up compilation
considerably.
A complete ngspice (release version, no debug info, 64-bit optimized executable) may be
made available just by
$ cd ngspice
$ ./compile_min.sh
A debug version without optimization will be available by
$ ./compile_min.sh d
Options used in the script:
CIDER, XSPICE, KLU, and OpenMP may be selected at will.
–disable-debug will give O2 optimization (versus O0 for debug) and removes all debugging
info.
The install script will copy all files to C:\Spice64, the code models for XSPICE will be
stored in C:\Spice64\lib\spice.
If you don’t use the tarball, you may download the ngspice source code from the ngspice
Git distribution as described on the sourceforge ngspice Git page. Define and enter a
directory of your choice, e.g. /d/spice/. Download the complete ngspice repository from
Git, for example by anonymous access issuing the command
You will find the sources in directory /d/spice/ngspice/. Now enter the ngspice top
level directory ngspice. For compilation using
$ ./compile_min.sh
you have to edit this script and uncomment the two lines enabling ./autogen.sh. If you
want to compile ngspice manually, follow the procedure described below:
$ cd ngspice
$ ./autogen.sh
$ mkdir release
$ cd release
$ ../configure --with-wingui ...and other options (32.1.8.1)
$ make -j8
$ make install
The user defined build tree saves the object files, instead of putting them into the source
tree, in a release (and a debug) tree. Please see Chapt. 32.1.4 for instructions.
If you need updating your local source code tree from Git, just enter ngspice directory
and issue the command
git pull
git reset --hard
To learn more about Git, which can be both powerful and difficult to master, please
consult http://git-scm.com/, especially: http://git-scm.com/documentation, which has
pointers to documentation and tutorials.
The script ./compile_min.sh or the command make install will create a directory tree
with 64-bit ngspice as shown below:
C:\Spice64\
bin\
ngspice.exe
cmpp.exe
lib\
ngspice\
analog.cm
digital.cm
spice2poly.cm
extradev.cm
extravt.cm
share\
info\
dir
ngspice.info
ngspice.info-1
..
ngspice.info-10
man\
man1\
ngspice.1
ngspice\
scripts\
ciderinit
devaxis
devload
setplot
spectrum
spinit
MSVC.CMD
vlnggen
Table 32.4: ngspice standard installation tree under MS Windows
The ./configure flag --enable-relpath may be useful if the install path (e.g. C:\Spice64)
is only preliminary, because a Windows installer is preferred. Then all search paths for
spinit and code models are made relative to the executable (either ngspice.exe or the caller
to ngspice.dll), see 32.1.8.
For compiling ngspice as a dll (shared library) use the configure option --with-ngshared
instead of --with-wingui. In addition you might add (optionally) --enable-relpath
to avoid absolute paths when searching for code models. You may edit compile_min.sh
accordingly and compile using this script in the MSYS2 window.
32.2.3 make ngspice with pure CYGWIN

The procedure of compiling is the same as with Linux (see Chapt. 32.1). After you have
moved to the ngspice directory, the following command sequence may do the work for
you:
$ ./autogen.sh
$ mkdir release-cyg
$ cd release-cyg
$ ../configure --with-x --disable-debug --with-readline=yes --enable-xspice
--enable-klu --enable-osdi --enable-pss --enable-cider --enable-openmp
$ make clean 2>&1 | tee make_clean.log
$ make 2>&1 -j8 | tee make.log
$ make install 2>&1 | tee make_install.log
The (optional) statement -j8 (or -jn, n is the number of logical cores available) will speed
up compilation considerably.
The CYGWIN console executable you have been creating is an X11 application. This is
a not a Windows native environment. So you have to add an X11 graphics interface by
installing the XServer from the CYGWIN project. Before starting ngspice, you have to
start the XServer by the following commands within the CYGWIN window:
$ export DISPLAY=:0.0
$ xwin -multiwindow -clipboard &
If you don’t have libdl.a you may need to link libcygwin.a to libdl.a symbolically, for
example:
$ cd /lib $ ln -s libcygwin.a libdl.a.
32.2.4 ngspice mingw or cygwin console executable w/o graph-

ics
If you omit the configure flag –with-wingui or –with-x, you will obtain a console application
without graphics interface.
./configure --enable-xspice --enable-cider --enable-openmp

--enable-osdi --disable-debug CFLAGS=-m32 LDFLAGS=-m32 prefix=C:/Spice
is an example for TDM mingw, 32 Bit ngspice console. No graphics interface is pro-
vided. A warning message will be issued upon starting ngspice. However, you may invoke
Gnuplot for plotting (see 17.5.38).
32.3. NGSPICE COMPILATION UNDER MACOS 729
32.2.5 ngspice for MS Windows, cross compiled from Linux
The ngspice main directory contains two scripts that provide cross compiling ngspice.exe
or ngspice.dll from a Linux setup. For details and prerequisites please have a look at
cross-compile.sh or cross-compile-shared.sh.
32.3 Ngspice Compilation under macOS
Basically compiling for macOS is similar to compiling for Linux. But one has to take
the many special features of the Apple world into account. The following is just a coarse
description of a setup and compiling, tested under Big Sur (macOS 11) on an Apple
MacBook-Air with Intel i5 chip. I reflects my (H. Vogt) setup for achieving a complete
compile, including all relevant ngspice options. This setup has grown over time. Please
check, and if something is missing if you do a fresh install and compile, please let me
know.
32.3.1 Prerequisites
Ngspice is written in C and thus a complete C/C++ compilation environment is needed.

Ngspice is developed with autotools, gcc, and GNU make. The graphics interface is using
X11. Several additional libraries have to be installed. As a first step install the Xquartz
system, which enables X11 support. Sevral additional tools and libraries need to by
downloaded and installed, either from Brew or MacPorts.
Required tools are autoconf, automake, libtool, bison.
Libraries are: readline, Xft2, Freetype, ncurses, fftw (optional), and several X11 exten-
sions: Xaw, Xmu, Xt, Xext, Xrender, SM, ICE.
The standard gcc provided by Apple Xcode (in fact a link to a clang/llvm compiler) does
not (yet?) support OpenMP, so you may use gcc-11 from Homebrew.
The following table lists the libraries required by ngspice. Libs located in /usr/local/opt/
and the compiler gcc-11 stem from Homebrewrew, the other libs are from macPorts.
List of standard ngspice dependencies (acquired with command otool -L

/usr/local/bin/ngspice):
/ usr / local / opt / ncurses / lib / libncursesw .6. dylib

/ usr / local / opt / gcc / lib / gcc /11/ libstdc ++.6. dylib
/ usr / local / opt / fftw / lib / fftw3 .3. dylib
/ usr / local / opt / readline / lib / libreadline .8. dylib
/ opt / local / lib / libXaw .7. dylib
/ opt / local / lib / libXmu .6. dylib
/ opt / local / lib / libXt .6. dylib
/ opt / local / lib / libXext .6. dylib
/ opt / local / lib / libX11 .6. dylib
/ opt / local / lib / libfontconfig .1. dylib
/ opt / local / lib / libXrender .1. dylib
/ opt / local / lib / libfreetype .6. dylib
/ opt / local / lib / libSM .6. dylib
/ opt / local / lib / libICE .6. dylib
/ usr / local / opt / gcc / lib / gcc /11/ libgomp .1. dylib
/ usr / lib / libSystem . B . dylib
/ usr / local / lib / gcc /11/ libgcc_s .1. dylib
32.3.2 Compiling ngspice
A compile script is provided, supporting gcc-11 from Homebrew, see

https://sourceforge.net/p/ngspice/ngspice/ci/master/tree/compile_macos_gcc.sh.
ngspice is also available from MacPorts, Homebrew or Fink, partially as source code,
partially as an executable. ngspice from Homebrew for example currently does not offer
the graphics interface to X11 (thus not needing the above mentioned Quartz installation
and X11 runtime libraries).
32.3.3 Compiling ngspice shared library
The ngspice shared library does not have any command line or graphics interface, thus
neither needing X11 (Quartz) nor readline nor ncurses. Compilation (with gcc-11) is as
straightforward as running the compile script. The list of depencencies now is shorter
than before:
List of shared ngspice dependencies (acquired with command otool -L
/usr/local/lib/libngspice.0.dylib):
/ usr / local / opt / gcc / lib / gcc /11/ libstdc ++.6. dylib
/ usr / local / opt / fftw / lib / libfftw3 .3. dylib
/ usr / local / opt / gcc / lib / gcc /11/ libgomp .1. dylib
/ usr / lib / libSystem . B . dylib
/ usr / local / lib / gcc /11/ libgcc_s .1. dylib
32.4. REPORTING ERRORS 731
Currently no work has been done to create a package and have the package certified by
Apple.
32.4 Reporting errors

Setting up ngspice is a complex task. The source code contains over 1500 files. ngspice
should run on various operating systems. Therefore errors may be found, some still
evolving from the original spice3f5 code, others introduced during the ongoing code en-
hancements.
If you happen to experience an error during compilation of ngspice, please send a report
to the development team. Ngspice is hosted on SourceForge, the preferred place to post a
bug report is the ngspice bug tracker. We would prefer to have your bug tested against the
actual source code available at Git, but of course a report using the most recent ngspice
release is welcome! Please provide the following information with your report: Ngspice
version, Operating system, Small input file to reproduce the bug (if to report a runtime
error), Actual output versus the expected output.
Chapter 33
Copyrights and licenses
33.1 Documentation license

The license for this document is covered by the Creative Commons Attribution Share-
Alike (CC-BY-SA) v4.0.
See here for details of the legal code.
Parts of chapters 12 and 25-27 are in the public domain.
Chapter 30 is covered by the 3-clause BSD (modified BSD).
33.2 ngspice license

The SPICE license is the ‘Modified’ BSD license, (see 33.3.2 and Spice link at UCB).
ngspice adopts this ‘Modified’ BSD license for all of its source code except for tclspice,
admst, and numparam that are under LGPLv2, and XSPICE, which is in the public
domain. Some adms device models have company specific licenses and thus are not part
of the standard distribution (see file COPYING for details) . The ngspice licences are
compliant with the DFSG (Debian Free Software Guidelines).
33.3 Some license details
33.3.1 CC-BY-SA
This is a human-readable summary of (and not a substitute for) the license CC-BY-SA.
You are free to:
Share — copy and redistribute the material in any medium or format Adapt — remix,
transform, and build upon the material for any purpose, even commercially.
This license is acceptable for Free Cultural Works.
The licensor cannot revoke these freedoms as long as you follow the license terms.
733
734 CHAPTER 33. COPYRIGHTS AND LICENSES
Under the following terms:

Attribution — You must give appropriate credit, provide a link to the license, and indicate
if changes were made. You may do so in any reasonable manner, but not in any way that
suggests the licensor endorses you or your use.
ShareAlike — If you remix, transform, or build upon the material, you must distribute
your contributions under the same license as the original.
No additional restrictions — You may not apply legal terms or technological measures
that legally restrict others from doing anything the license permits.
Notices:
You do not have to comply with the license for elements of the material in the public
domain or where your use is permitted by an applicable exception or limitation. No
warranties are given. The license may not give you all of the permissions necessary for
your intended use. For example, other rights such as publicity, privacy, or moral rights
may limit how you use the material.
Disclaimer:
This deed highlights only some of the key features and terms of the actual license. It is
not a license and has no legal value. You should carefully review all of the terms and
conditions of the actual license before using the licensed material.
33.3.2 ‘Modified’ BSD license
Copyright 1985 - 2017, Regents of the University of California and others

Redistribution and use in source and binary forms, with or without modification, are
permitted provided that the following conditions are met:
1. Redistributions of source code must retain the above copyright notice, this list of
conditions and the following disclaimer.
2. Redistributions in binary form must reproduce the above copyright notice, this list
of conditions and the following disclaimer in the documentation and/or other materials
provided with the distribution.
3. Neither the name of the copyright holder nor the names of its contributors may be used
to endorse or promote products derived from this software without specific prior written
permission.
THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIB-
UTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING,
BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY
AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT
SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE FOR ANY
DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUEN-
TIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUB-
STITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSI-
NESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIA-
BILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
33.4. ON THE HISTORICAL EVOLVEMENT OF THE NGSPICE LICENSES 735
NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF

THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
(Source)
33.4 On the historical evolvement of the ngspice li-

censes
The SPICE license is the ‘Modified’ BSD license, (see Spice link at UCB). The original
Spice3f5 had been released under the 4-clause BSD (the original BSD license), which has
been modified by UCB towards the now commonn 3-clause BSD. ngspice adopts this
‘Modified’ BSD license for all of its source code (except for tclspice, admst, and numparam
that are under LGPLv2, and XSPICE, which is in the public domain (see 33.4.4)).
33.4.1 XSPICE SOFTWARE (documentation) copyright
Code added to SPICE3 to create the XSPICE Simulator and the XSPICE Code Model
Subsystem was developed at the Computer Science and Information Technology Labora-
tory, Georgia Tech Research Institute, Atlanta GA, and is covered by license agreement
the following copyright:
Copyright © 1992 Georgia Tech Research Corporation All Rights Reserved. This material
may be reproduced by or for the U.S. Government pursuant to the copyright license under
the clause at DFARS 252.227-7013 (Oct. 1988)
Refer to U.C. Berkeley and Georgia Tech license agreements for additional information.
This license is now superseded by Chapt. 33.4.4
33.4.2 CIDER RESEARCH SOFTWARE AGREEMENT (su-

perseded by 33.4.3)
This chapter specifies the terms under which the CIDER software and documentation
coming with the original distribution are provided. This agreement is superseded by
33.4.3, the ‘modified’ BSD license.
Software is distributed as is, completely without warranty or service support. The Uni-
versity of California and its employees are not liable for the condition or performance of
the software.
The University does not warrant that it owns the copyright or other proprietary rights
to all software and documentation provided under this agreement, notwithstanding any
copyright notice, and shall not be liable for any infringement of copyright or proprietary
rights brought by third parties against the recipient of the software and documentation
provided under this agreement.
THE UNIVERSITY OF CALIFORNIA HEREBY DISCLAIMS ALL IMPLIED WAR-
RANTIES, INCLUDING THE IMPLIED WARRANTIES OF MERCHANTABILITY
736 CHAPTER 33. COPYRIGHTS AND LICENSES
AND FITNESS FOR A PARTICULAR PURPOSE. THE UNIVERSITY IS NOT LI-

ABLE FOR ANY DAMAGES INCURRED BY THE RECIPIENT IN USE OF THE
SOFTWARE AND DOCUMENTATION, INCLUDING DIRECT, INDIRECT, SPECIAL,
INCIDENTAL, OR CONSEQUENTIAL DAMAGES.
The University of California grants the recipient the right to modify, copy, and redistribute
the software and documentation, both within the recipient’s organization and externally,
subject to the following restrictions:
(a) The recipient agrees not to charge for the University of California code itself. The
recipient may, however, charge for additions, extensions, or support.
(b) In any product based on the software, the recipient agrees to acknowledge the research
group that developed the software. This acknowledgment shall appear in the product
documentation.
(c) The recipient agrees to obey all U.S. Government restrictions governing redistribution
or export of the software and documentation.
All BSD licenses have been changed to the ‘modified’ BSD license by UCB in 1999 (see
Chapt. 33.4.3).
33.4.3 ‘Modified’ BSD license
All ‘old’ BSD licenses (of SPICE or CIDER) have been changed to the ‘modified’ BSD
license according to the following publication
(see ftp://ftp.cs.berkeley.edu/pub/4bsd/README.Impt.License.Change):
July 22, 1999
To All Licensees, Distributors of Any Version of BSD:
As you know, certain of the Berkeley Software Distribution (‘BSD’) source code files re-
quire that further distributions of products containing all or portions of the software,
acknowledge within their advertising materials that such products contain software de-
veloped by UC Berkeley and its contributors.
Specifically, the provision reads:
‘3. All advertising materials mentioning features or use of this software must display the
following acknowledgment: This product includes software developed by the University
of California, Berkeley and its contributors.’
Effective immediately, licensees and distributors are no longer required to include the
acknowledgment within advertising materials. Accordingly, the foregoing paragraph of
those BSD Unix files containing it is hereby deleted in its entirety.
William Hoskins
Director, Office of Technology Licensing
University of California, Berkeley
33.4. ON THE HISTORICAL EVOLVEMENT OF THE NGSPICE LICENSES 737
33.4.4 XSPICE
According to https://web.archive.org/web/20161030172156/http://users.ece.gatech.edu/∼mrichard/Xs
(as of Feb. 2012) the XSPICE source code and documentation have been put into the
public domain by the Georgia Institute of Technology.
33.4.5 OSDI
The OSDI interface to OpenVAF-compiled device models is licensed according to the
Mozilla Public License, v. 2.0.(see https://mozilla.org/MPL/2.0/).
33.4.6 tclspice, numparam

Both software packages are copyrighted and are released under LGPLv2
(see http://www.gnu.org/licenses/lgpl-2.1.html).
33.4.7 Linking to GPLd libraries (e.g. readline, fftw, table.cm):

The readline manual at http://tiswww.case.edu/php/chet/readline/rltop.html states: Read-
line is free software, distributed under the terms of the GNU General Public License,
version 3. This means that if you want to use Readline in a program that you release
or distribute to anyone, the program must be free software and have a GPL-compatible
license.
According to http://www.gnu.org/licenses/license-list.html, the modified BSD license,
thus also the ngspice license, belongs to the family of GPL-Compatible Free Software
Licenses. Therefore the linking restrictions to readline, which have existed with the old
BSD license, are no longer in effect.

Ngspice Manual

Uploaded by

Copyright:

Available Formats

Ngspice Manual

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Ngspice Manual

Uploaded by

Copyright:

Available Formats

Ngspice User’s Manual

Holger Vogt, Giles Atkinson, Paolo Nenzi

January 14, 2024

How to use this Manual

Ngspice User’s Manual

I Ngspice User’s Manual 3

3 Circuit Elements and Models 71

4 Voltage and Current Sources 93

4.1.9 External voltage or current input . . . . . . . . . . . . . . . . . . . 100

5 Non-linear Dependent Sources (Behavioral Sources) 105

6 Transmission Lines 121

12 Mixed-Mode and Behavioral Modeling with XSPICE 179

12.2.9 Filesource (PWL sourced from ﬁle) . . . . . . . . . . . . . . . . . . 197

12.3.9 Controlled PWM Oscillator . . . . . . . . . . . . . . . . . . . . . . 248

13 Verilog-A Compact Device Models 291

14 Digital Device Models 295

15 Analyses and Output Control (batch mode) 301

15.3.8 .SP S-Parameter Analysis . . . . . . . . . . . . . . . . . . . . . . . 316

16 Starting ngspice 343

17 Interactive Interpreter 369

17.5.19 Deﬁne: Deﬁne a function . . . . . . . . . . . . . . . . . . . . . . . . 386

17.5.53 Noise: Noise analysis . . . . . . . . . . . . . . . . . . . . . . . . . . 402

17.5.88 Spec: Create a frequency domain plot . . . . . . . . . . . . . . . . . 420

17.8.2 Vectors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 444

18 Ngspice User Interfaces 459

19 ngspice as shared library or dynamic link library 479

20.6 Compiling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 508

21 Example Circuits 511

22 Statistical circuit analysis 523

23 Circuit optimization with ngspice 533

II XSPICE Software User’s Manual 547

25 XSPICE Basics 549

26 Execution Procedures 553

27 Example circuits 565

28 Code Models and User-Deﬁned Nodes 575

28.7 Model Deﬁnition File . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 586

29 Error Messages 613

III CIDER 625

30 CIDER User’s Manual 627

30.3.1 DESCRIPTION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 631

30.11.2 Parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 643

31 Model and Device Parameters 661

31.4.4 TransLine - Simple Lossy Transmission Line . . . . . . . . . . . . . 676

32 Compilation notes 713

32.2.2 How to make ngspice with MINGW and MSYS2 . . . . . . . . . . 725

33 Copyrights and licenses 733

Preface to the ﬁrst edition

1. The spice3f5 manual,

2. the XSPICE user’s manual,

3. the CIDER user’s manual

Kevin Aylward, Warden of the King’s Ale

Preface to the current edition (as of Nov 2023)

and many others...

1.1 Simulation Algorithms

1.1.1 Analog Simulation

1.1.2 Matrix solvers

1.1.3 Device Models for Analog Simulation

1.1.4 Digital Simulation

1.1.5 Mixed-Signal Simulation