I H

First Edition
Compiled by the Computer Applications Engineering Department
Author
William R. Blood JI.
Contributors
Jon M. DeLaune
Jerry E. Prioste
Editor
Edmund C. Tynan JI.
@ MOTOROLA Inc.
Circuit diagrams external to Motorola products are included as a means of
illustrating typical semiconductor applications; consequently, complete infor-
mation sufficient for construction purposes is not necessarily given. The
information in this book has been carefully checked and is believed to be entirely
reliable. However, no responsibility is assumed for inaccuracies. Furthermore,
such information does not convey to the purchaser of the semiconductor devices
described any license under the patent rights of Motorola Inc. or others.
MECL, MECL II, MECL III, MECL 10,000, MTTL,
MTTL III, and MDTL are trademarks of Motorola Inc.
First Edition
October, 1971
@MOTOROLA INC., 1971
ii
PREFACE
In response to the demand for higher performance systems, engineers
are looking at digital integrated circuit families which are faster than the
popular TTL and DTL types. Motorola's Emitter Coupled Logic (MECL)
circuits have the characteristics to meet the performance requirements
for present and future systems. MECL 10,000 is ideal for computer and
communications systems, while state-ofthe-art instrumentation equip-
ment uses MECL III.
As circuit speeds increase, wiring rules and system design techniques
must be adjusted accordingly. Designing with MECL is no more difficult
than designing high performance equipment with slower forms of logic.
High performance system design for any form of logic, however, does
require an understanding of the factors which affect system performance.
In fact, many of the MECL features such as transmission line drive capa-
bility, complementary outputs, Wired-OR, and versatile logic functions
can add as much to system performance as the short propagation delays
and high toggle rates.
In the past, several articles and application notes have been written
about MECL circuits and systems. However, there was a need for a book
which would completely define MECL operation. This book has been
written to give the designer the information to establish design rules for
his own high performance systems.
The information in this book is based on equations derived from
electronic theory, laboratory tests, and inputs from MECL users. All of
the rules and tables are for conservative system design with MECL
circuits. It is important to realize that the circuits can operate properly
under conditions much more adverse than suggested in this book.
In addition to the technical contributors, Jon DeLaune and Jerry
Prioste, the author would like to thank Lloyd Maul, Mike Lee, Reg
Hamer, Jim Hively, Howard Gnauden, Don Murray, Tom Balph, and
Colin Crook whose knowledge of MECL has added to the completeness
and accuracy of this book. Finally, great appreciation is due to the
many technicians, engineers, and managers who took their valuable time
to read all or part of this book as it was depeloped.
iii
Table of Contents
Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. vi
What Is MECL? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. VI
History of MECL . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. vi
Why Use MECL? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. vii
The Advantages of MECL . . . . . . . . . . . . . . . . . . . . . . . . . . . .. viii
MECL Areas of Application. . . . . . . . . . . . . . . . . . . . . . . . . .. ix
Purpose of This Book. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . x
CHAPTER 1 - MECL Families . ........................................... .
The Basic MECL Gate ........................................................... .
Noise Margin. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 5
MECL Circuit Types. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 6
MECL Flip-Flops. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
Operation of Flip-Flop. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 8
MECL Family Comparison. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 10
CHAPTER 2 - Using MECL . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 15
MECL II Design Rules. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 15
A. Logic Design Considerations. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 15
B. System Layout Considerations. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 19
C. Circuit Board Layout Techniques. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 20
D. Backplane Wiring. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 21
E. System Considerations. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 22
MECL 10,000 Design Rules. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 23
A. General Considerations. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 23
B. Printed Circuit Card Layout Techniques. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 23
C. Power Supply Bypassing on Circuit Cards. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 24
D. Backplane and Loading Considerations. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 24
E. System Distribution and Grounding. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 25
F. Loading Rules for MECL 10,000 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 25
MECL III Design Rules. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 29
A. Circuit Card Layout. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 29
B. Transmission Line (Microstrip Line) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 30
C. On-Card Clock Distribution via Transmission Lines. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 30
D. Off-Card Clock Distribution. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 31
E. Testing MECL III . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 31
CHAPTER 3 - Printed Circuit Board Connections. . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 35
Transmission Line Geometries. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 37
Basic Transmission Line Operation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 42
Unterminated Lines. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 43
Series Damped and Series Terminated Lines. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 46
Parallel Terminated Lines. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 52
Transmission Line Comparison. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 53
Wire wrapped Cards. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 55
iv
Contents (continued)
CHAPTER 4 - System Interconnections. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 57
Connectors. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
Coaxial Cable ................................................................. , 59
Differential Twisted Pair Lines and Receivers. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 64
Ribbon Cable ................................................................. , 70
Schottky Diode Termination. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . .. 71
Parallel Wire Cables. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 76
Twisted Pair Cable, Driven Single-Ended. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 83
CHAPTER 5 - Power Distribution. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 87
System Power Calculations. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 88
Power Supply Considerations. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
System Power Distribution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
Backplane Power Distribution. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 95
On-Card Power Distribution. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
Vrr Termination Voltage Distribution. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 99
CHAPTER 6 - Thermal Considerations ...................................... 101
MECL Integrated Circuit Heat Transfer .............................................. 102
MECL DC Thermal Characteristics ................................................ , 106
Heat Dissipation Techniques. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 110
Mounting Techniques. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 112
CHAPTER 7 - Transmission Line Theory. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 115
Transmission Line Design Information, With Examples. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 115
Signal Propagation Delay for Microstrip and Strip Lines With
Distributed or Lumped Loads. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 123
Microstrip Transmission Line Techniques, Evaluated Using TDR
Measurements, With Examples. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 126
The Effect of Loading on a Parallel Terminated Transmission Line,
With Examples. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 139
Analysis: Series Terminated Lines Compared to Parallel Terminated
Lines, With Example ........................................................... 146
Analysis of Series Damping Terminations ............................................ 153
Bibliography ....................... , . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 166
CHAPTER 8 - MECL Applications. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 167
Counters. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. .. ............................ 167
Shift Registers. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 182
Adders ...................................................................... , 182
Code Converters .............................................................. , 191
Memories ..................................................................... 196
Oscillators. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 199
One-Shot Multivibrators . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 202
Linear Applications. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 205
Translators. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 206
v
Introduction
What is MECL?
The term MECL identifies Motorola's emitter coupled logic. Emitter coupled
logic is a non-saturating form of digital logic which eliminates transistor storage time
as a speed limiting characteristic, permitting very high speed operation. "Emitter
Coupled" refers to the manner in which the emitters of a differential amplifier with-
in the integrated circuit are connected. The differential amplifier provides high
impedance inputs and voltage gain within the circuit. Emitter follower outputs re-
store the logic levels and provide low output impedance for good line driving and
high fanout capability.
History of MECL
Motorola offers MECL circuits in four logic families: MECL I, MECL II,
MECL III, and MECL 10,000.
The MECL I family was the first digital monolithic integrated circuit line
produced by Motorola. Introduced in 1962, MECL I was considerably beyond the
state-of-the-art at that time. Several years passed before any other form of logic
could equal the 8 ns gate propagation delays and 30 MHz toggle rates of MECL I. As
a result of its reliability and performance, MECL I was designed into many advanced
systems.
Nearly a decade later, MECL I is still being produced by Motorola. It finds wide
usage in existing products. However, several features of the more advanced MECL II,
III, and 10,000 favor their being used in new designs. For example, MECL I requires
a separate bias driver package, MC304/MC354, to be connected to the logic
functions. This means increased package coun t and extra circuit board wiring. Also
the 10-pin packages used for MECL I limit the number of gates per package and the
number of gate inputs. No provision was made for operation of MECL I with
transmission lines, as they were unnecessary with the 8 ns rise and fall times.
In 1966 Motorola introduced the more advanced MECL II. The basic gate
featured 4 ns propagation delays and flip-flop circuits that would toggle at over
70 MHz. MECL II immediately set a new standard for performance that has been
equaled by non-ECL logic only with the introduction of Schottky TTL in 1970.
Motorola continued with the development of MECL II and flip-flop speeds
were increased first to 120 MHz for the MC I 027/MC 1227 JK circuit, and then to
180 MHz for the MC 1 034 type D flip-flop. To drive these high speed flip-flops, high
speed line drivers were introduced with 2 ns propagation delays and 2 ns rise and fall
times. With 2 ns edges, transmission lines could be used to preserve the waveforms
and limit overshoot and ringing on longer lines. Consequently, the MC 1 026 was
designed to drive 50-ohm lines. Because of the significant speed increase of the line
drivers and high speed flip-flops over the basic MECL II parts, these circuits are
commonly called MECL 11-1/2, although they are part of the MECL II family.
vi
MECL II is continuing to gain in popularity and is being designed into many
high speed systems. MECL II circuits have a temperature compensated bias driver
internal to the circuits (except for the MC 1 020/MC 1220 line receiver which requires
no internal bias). The internal bias source simplifies circuit interconnections and
tracks with both temperature and supply voltage to retain noise margin under varied
operating conditions.
Complex functions became available in MECL II when trends shifted toward
more complicated circuits. The family now has adders, data selectors, multiplexers,
decoders and a Nixie* tube decoder/driver. MECL II is a growing line, with new
products currently being designed and introduced.
Motorola's continuing development of ECL made possible an even faster logic
family. As a result, MECL III was introduced in 1968. Its 1 ns gate propagation
delays and greater than 300 MHz flip-flop toggle rates remain the industry leaders.
The I ns rise and fall times require a transmission line environment for all but the
smallest systems. For this reason, all circuit outputs are designed to drive
transmission lines and all output logic levels are specified when driving 50-ohm
loads. Because of MECL TIl's fast edge speeds, multi-layer boards are recommended
above 200 MHz. For the first time with MECL, internal input pull down resistors are
included with the circuits to eliminate the need to tie unused inputs to VEE. The
Hi-Z 50 kn input resistors are used with transmission lines for most applications.
Optional Low-Z 2 kn input resistors can be used in place of pulldown resistors
when the chips are used in a hybrid circuit or when line lengths are very short.
MECL III is gaining in popularity - especially in high speed test and communica-
tions equipment. As a result, Motorola is continuing to expand and develop this
product line.
Trends in large high speed systems have shown the need for an easy to use logic
family with 2 ns propagation delays. To fill this requirement, Motorola introduced
the MECL 10,000 series in 1971. In order to make the circuits comparatively easy to
use, edge speed was slowed to 3.5 ns while the important propagation delay was held
to 2.0 ns. The slow edge speed permits use of wire wrap and standard printed circuit
lines. However, the circuits are specified to drive transmission lines for optimum
performance.
MECL 10,000 is provided with logic levels that are completely compatible with
MECL III to facilitate using both families in the same system. A second important
feature of MECL 10,000 is the significant power reduction. MECL 10,000 gates use
less than one-half the power of MECL III or high speed MECL II gates. Finally, the
low gate power and advanced circuit design techniques have permitted a new level of
MECL complex circuits. For example, complexity of the MCIOl81 four-bit
arithmetic unit compares favorably to that of any bipolar integrated circuit on the
market. MECL 10,000 is the fastest growing ECL family in the industry and
Motorola is designing and introducing many versatile complex functions to expand
the line.
Why Use MECL?
Circuit speed is, of course, an obvious reason for designing with MECL.
MECL III is significantly faster than any other digital logic family. MECL 10,000
offers shorter propagation delays and higher toggle rates than any non-ECL type of
*T.M. - Burroughs Corp.
vii
logic. Equally important to the circuit speed are the characteristics of MECL circuits
which permit entire systems to operate at high speeds.
The ability of the faster MECL families to drive transmission lines becomes
increasingly important in larger and faster systems. While a transmission line
environment imposes some additional design rules and restrictions, the advantages of
longer signal paths, better fanout, improved noise immunity, and faster operation,
often more than compensate for the restrictions.
When using MECL II or MECL 10,000 without transmission lines, their high
input impedances permit the use of series-damping resistors to increase wiring
lengths and to improve waveforms. Unlike non-ECL forms of logic, MECL circuits
have constant power supply requirements, independent of operating frequency. This
simplifies power supply design, since circuit speed need not be considered a variable.
At fast circuit speeds MECL can offer a considerable power saving over the other
types of logic.
In addition to faster operation, the line driving features of MECL circuits can
be exploited to improve system performance. For one, the parts specified to drive
transmission lines will drive coaxial cables over distances limited only by the
bandwidth of the cable. In addition, the shielding in coaxial cable gives good
isolation from external noise.
More economical than using coaxial cable, is the ability of the MECL circuits to
differentially drive and receive signals on twisted pair lines. Using this technique,
signals have been sent over twisted pair lines up to 1000 feet in length.
The complementary outputs and Wired-OR capabilities of MECL circuits result
in faster system operation with reduced package count and a power saving. The
complementary outputs are inherent in the circuit design and both outputs have
equal propagation delay. This eliminates the timing problems associated with using
an inverter to get a complement signal. The logic OR function is obtained by wiring
circuit-outputs together. The propagation delay of the Wired-OR connection is much
less than a gate function and can save power, as only one pulldown resistor or
termination is required per Wired-OR.
Another advantage when designing with MECL is the low noise generated by
the circuits. Unlike totem pole outputs, the emitter follower does not generate a
large current spike when switching logic states, so the power lines stay comparatively
noise free. The low current-switching in signal paths, relatively small voltage swing
(typically 800 mY), and low output impedances, cut down crosstalk and noise.
Generated noise is also reduced by MECL's relatively slow rise and fall times.
For each MECL family the edge speed is equal to or greater than the propagation
delay. The low noise associated with MECL is especially important when the logic
circuits are to be used in a system which contains low level analog or
communications signals.
The flexibility of the MECL line receivers and Schmitt triggers to act as linear
amplifiers leads to many functions that may be performed with standard MECL
circuits. For example, in addition to amplifying low level signals to MECL levels,
these MECL circuits can be used as crystal oscillators, zero crossing detectors, power
buffers, Schmitt triggers, RF and video amplifiers, one-shot multivibrators, etc.
The Advantages of MECL
1. Highest speed IC logic available
2. Low cost
3. Low output impedance
viii
4. High fanout capability
5. Constant supply current as a function of frequency or logic state
6. Very low noise generation
7. Complementary logic outputs save on package count
8. Low crosstalk between signal leads
9. All outputs are buffered
10. Outputs can be tied together giving the Implied-OR function
11. Common mode rejection of noise and supply variations is I V or greater
for differential line receiving
12. Bias supplies are internal, allowing MECL use with a single power
supply
13. Minimal degradation of parameters occurs with temperature variations
14. Large family of devices yields economical designs
15. Power dissipation can be reduced through use of Implied-OR and the
"Series Gating" technique
16. Easy data transmission over long distances by using the balanced twisted
pair technique with standard parts
17. Constant noise immunity versus temperature
18. Best speed-power product available
19. All positive logic functions are available
20. Adapts easily to MSI and LSI techniques
MECL Areas of Application
1. Instrumentation
2. High speed counters
3. Computers
4. Medical electronics
5. Military systems
6. Large real-time computers
7. Aerospace and communication satellite systems
8. Ground support system
9. High speed AID conversion
10. Digital communication systems
11. Data transmission (twisted pair)
12. Frequency synthesizers
13. Phase array radar
14. High speed memories
15. Data delay lines
ix
Purpose of This Book
Rules and guidelines for using the various MECL families comprise the subject
matter of this book. Because of edge speed and bit rate capabilities, each family has
differing system requirements. The family name will therefore be referenced for the
examples and figures in the text, whenever applicable. The information in this book
is meant to apply to MECL II, MECL III, and MECL 10,000. The information about
MECL II will generally apply to MECL I, although the data would be conservative
because of the slower MECL I speed. This book aims at giving the reader an
understanding of the MECL families, as well as the knowledge needed to confidently
design with and use MECL.
Chapter I discusses the operation of MECL circuits and the characteristics of
the various families. It also shows methods for internally connecting the basic gates
to provide efficient complex functions. Of more importance to the user is Chapter 2
- a list of rules providing a condensed reference for using the various MECL
families.
Chapters 3, 4, 5, and 6 elaborate on those rules giving a technical background
for good system design and presenting test results showing MECL circuits in various
modes of operation. Chapter 3 describes circuit-to-circuit interconnections on a
card. Both open wire and transmission line techniques are covered. Chapter 4
expands the wiring techniques to show methods for card-to-card and panel-to-panel
interconnections. Chapter 5 elaborates on power distribution, showing how voltage
drops and power line noise affect noise immunity. Chapter 6 discusses thermal
considerations. Attention is given to the problems of calculating chip temperature,
removing heat from the system, and to the effect of thermal differences on noise
immunity.
Chapter 7 provides background necessary for understanding transmission lines
as they apply to MECL. Derivations of equations are shown, along with test results
correlating with the theoretical analysis. This chapter should be especially useful
when selecting a transmission line impedance and when determining the effect of
fanout or stray capacitance on the line.
Chapter 8 contains application ideas for MECL circuits. Included are methods
for interfacing various logic families with MECL, and numerous useful circuits
designed with MECL for high performance.
x
MECL Families
The Basic MECL Gate
An understanding of the basic circuits used in the construction of a logic family
is important in order to successfully design and trouble-shoot a system which uses
the family. This chapter describes MECL circuits, compares MECL families, and
gives some suggested rules for using MECL circuits in system design.
Figure 1-1 shows a typical MECL gate, - the basic gate circuit for the MECL
10,000 family. (Similar schematics are provided for each other MECL logic family,
in the general information section of the MECL Data Book). The figure shows the
separate functional circuits within the gate.
The differential amplifier section contains the current steering element that
provides the actual logic gating of the circuit. It also provides the voltage gain
necessary for a narrow linear threshold region.
An internal temperature and voltage compensated bias driver supplies a
reference voltage for the differential amplifier. The bias voltage, VBB, is set at the
midpoint of the signal logic swing. With the recommended -5.2 volts supply voltage
11: MECL 10.000 Basic Gate
Internal
Temperature
and Voltage Emitter
Compensated Follower
Differential I nput Amplifier Bias Network Outputs
r ~ - - - - - - - - - - - - - - ~ A ~ - - - - - - - - - - - - - - ~ ~ r r - - - - - - ~ A ~ - - - - - - ~ \
A
\.
Rp
50 k
B
y
Inputs
C
RC1 220
Rp
50 k
D
VCC2
RC2 245
VBB
05
Rp lEI RE
50 k t 779 6.1 k
907
4.98 k
VCC1
OR
"----+--0 0 utput
NOR
Output
Basic MEeL Gate
and 25
0
C ambient temperature, VBB is -1.29 volts dc for either MECL 10,000 or
MECL III, and - 1.175 volts dc for MECL II. The diodes in the voltage divider line,
together with Q6, provide temperature compensation by maintaining a level
consistent with the midpoint of the logic levels despite changing temperatures.
One additional feature of the bias supply is its ability to track supply voltage
changes. Consequently MECL II gates, for example, are specified to operate from a
- 5.2 volt 20% supply. In fact, they are capable of working over a much wider range
(-3.0 to -8.0 volts) although ac performance would be degraded.
The emitter followers are output drivers. They provide level shifting from the
differential amplifier to MECL output levels, and provide a low output impedance
for driving transmission lines. Both MECL 10,000 and MECL III circuits use open
emitter outputs. The reason is that since these circuits are designed for use with
transmission lines, and since the line termination provides an output load, internal
pulldown resistors would be a waste of power.
However MECL II, which is not specified to drive transmission lines, offers
gates both with and without pulldown resistors. In general, the circuits without
pulldown resistors are intended for use in Wired-OR circuit designs rather than for
driving high fanout loads. MECL II circuits with internal output pulldown resistors
use 1.5 kQ values for standard speed parts, and as low as 600 Q values for the high
speed circuits.
MECL 10,000 and MECL III circuit families, designed to drive transmission
lines, have two V CC power voltage inputs. V CC 1 is used to supply current to the
output drivers, while V CC2 supplies the remainder of the circuit. Separate V CC lines
are used to eliminate crosstalk between circuits in a package. More important, the
use of two lines speeds up circuit performance by eliminating a voltage spike which
otherwise would occur on the bias voltage, VBB, caused by the relatively heavy
currents associated with transmission lines. Each V CC pin should be connected to
the system ground by as short a path as possible (all V CC pins are connected to the
same system ground). Standard speed MECL II circuits are not designed to drive
transmission lines and consequently require only a single V CC package pin.
The input pulldown resistors shown in Figure 1-1 are characteristic of MECL
10,000 and MECL III. MECL 10,000 and Hi-Z MECL III use 50 kQ "pinch"
resistors which serve to drain off the input transistor leakage current. These resistors
hold unused inputs at a fixed zero level, so unused inputs are left open. On the other
hand, MECL II without input pulldown resistors, requires unused inputs be tied to
VEE or VOL
The following calculations illustrate the current switching operation of a MECL
10,000 gate. Similar calculations may be performed for the other MECL families by
substituting appropriate resistor values and voltage levels.
When all gate inputs are at a voltage, Yin, equal to a logic f/J level, IVIL
maxi, the input transistors Ql through Q4 in Figure 1-1 will not
be conducting current, because the common emitter point of these four transistors is
at about -2.09 V: i.e., VBB + VBEQ5 -1.29 V + (-0.80 V). This is not
enough forward bias (base to emitter) on Ql through Q4 for conduction. Thus,
current flows through RC2, Q5, and RE- This current, IE(/), is:
IE(/) = VEE - (V:: + VBE) "'" -4.0 rnA.
2
MECL Current Switching
The voltage drop at the collector resistor. RC2, may be calculated as:
VRC2 = IE0RC2 + IBRC2 (-4.0 rnA) (245 D) = -0.98V.
The output transistor base current, IB, is small compared to the switch current,
so the second term above can be ignored.
The OR output is then obtained through an emitter-follower, Q8, which cuts
the output level by one base-emitter drop, giving a voltage level:
VOL OR = VRC2 + VBE,
where: VBE = base to emitter drop on Q8, with logic zero current level (i.e., 6 rnA
thm Q8).
So:
VOL OR -0.98 V + (-0.77 V) -1.75 V,
typical at T A = 25C.
The base of the NOR output emitter-follower, Q7, is at about -0.05 V, yielding
an output of - 0.924 V typical, at an output device current level of 22.5 rnA and
T A = 25C. (These output voltage and current levels assume 50-ohm loads to a
terminating voltage, VTT, of - 2.0 V).
If one or more of the gate inputs is switched to a voltage level, Yin, equal to a
nominal logic 1 level, IVJH minl>IVinl>IVIH maxi, a current lEI flows through
RCI, QI-Q4, and RE. This current is:
where: Yin = -0.924 V
VBE = -0.79 V.
The current flow through RC 1 produces a voltage at the collector nodes of Q I
through Q4:
Finally, the output is obtained through an emitter follower, Q7, which drops
the collector voltage level one base-emitter drop, so that:
VOL NOR = VRCI + VBE (output device at 6 rnA)
-0.98 V + (-0.77 V) = -1.75 V,
typical at T A = 25C.
The transfer curves in Figures 1-2(a) and (b) indicate the behavior of the MECL
gate while switching. Note from the data in Figure 1-3 and from the NOR transfer
characteristic: for Yin increasing from VIL min to VILA max, the output remains
at a high level. When Yin increases from VILA max to VIHA min' the NOR output
3
MECL Logic Levels
1-2: MECL 10,000 Transfer Characteristics and Specification Points
-0.650
-0.750
1\
\ 75C \'
\
"
0/
'd 25 ....
l\ /I
OR Output
75 C I
1\\
f--250C I
I-
-0.850
-0.950
-1.050
-1.150
-1.250 ... 0 C
I

1 0C
nA I
Vout,OUTPUT VOLTAGE -1.350
(VOLTS) -1.450
f-750C
III 1\
75C-25C
r/j "
1\
-1.550
'.J

"" r-2:0C -\
I"-
7 __
NOR Output
cfic
f---T0 OC
I
-1.650
-1.750
-1.850
-1.950
-2.050
-2.0 -1.8 -1.6 -1.4 -1.2 -1.0 -0.8 -0.6 -0.4 -0.2 0
Vin,lNPUT VOLTAGE (VOLTS)
(a) MECL 10,000 Transfer Characteristics
VEE = 5.2 V
Vout,OUTPUT VOLTAGE
(VOLTS)
-0.800
-0.810
-0.960
-0.980
-1.200
___ __
-1.650
-1.800
-1.850
I
-0.800
-1.850 -1.475 -1.105 -0.810
Vin,lNPUT VOLTAGE (VOLTS)
(b) MECL 10,000 Specification Points (25C)
1-3: MECL 10,000 and MECL III Specified Logic Levels and Thresholds
PARAMETER ooc +25
0
C +75
0
C
(VOLTSI MECL 10,000 MECL III MECL 10,000 MECL III MECL 10.000
VIH max & VOH max -0.850
-0.840
-0.810
-0.810 -0.720
VOH min -1.010 -1.000
-0.960
-0.960 -0.900
VOHA min
-1.030 -1.020
-0.980
-0.980 -0.920
VIHA min
-1.155 -1.135
-1.105
-1.095 -1.045
VILA max
-1.485 -1.500
-1.475
-1.485 -1.445
VOLA max
-1.640 -1.615
-1.630
-1.600 -1.600
VOL max
-1.660 -1.635
-1.650
-1.620 -1.620
VI L min & VOL min
-1.870 -1.870
-1.850
-1.850 -1.830
Conditions: Ceramic DI L package stabilized with Ifm air flow.
4
MECL III
-0.720
-0.900
-0.920
-1.035
-1.460
-1.575
-1.595
-1.830
Noise Margin
will switch to a low level. Then, as the input continues more positive than
VIHA min, the output continues more negative with a slope of about -0.24. This is
caused by the collector input node going more negative because of increasing
collcctor current as Yin goes more positive.
If the input continues in the positive direction, saturation will be reached at an
input of about -0.4 volts. Beyond that point, the base-collector junction is forward
biased to saturation and the collector voltage and output will go more positive
with the increasing input level. Since the saturation point is well above VOH max,
operation in this mode will not occur in normal system operation. The OR output
level depends on Q5's collector voltage (cf Fig. 1-1). This output is unaffected by
input levels except in the active transfer region.
Suffice it to say that while the manufacture of MECL circuits is not a primary
concern of the user, nevertheless ease of manufacture does translate directly into
end product cost. Although not as easy to build as some slower logic families
because of smaller transistor geometries, MECL does have some features which
facilitate processing. First, the voltage gain of the basic gate circuit (approximately
4.5 for MECL 10,000) is essentially independent of transistor beta. So transistor
beta can be allowed to vary from a low of about 70 in high speed MECL (40 for
MECL II) to a high in excess of 300, which permits easy processing limits. Second,
the output voltage levels depend on diode drops for a high output, and diode drops
and resistor ratios for a low output. Resistance ratios can be held to within 5% even
though absolute values vary by 20%. Again, this eases processing.
Third, since the transistors used do not saturate, the gold doping which is
normally required to decrease storage time is not required in MECL processing;
therefore yields are better. Fourth, collector-emitter voltages are low, due to circuit
design, again relaxing processing restrictions. Such advantages, together with
Motorola's ability to control processing, permit high volume production of all MECL
circuits. In effect, this means low-cost high performance circuits for the designer.
Noise Margin
Noise margin is a dc voltage specification which measures the immunity of a
circuit to adverse operating conditions. Noise margin is defined as the difference
between the worse case input logic level (V IRA min or VILA max) and the
guaranteed worst case output (V ORA min or VOLA max) for those inputs. Figure
1-3 lists the worst case limits for MECL 10,000 and MECL III. (MECL II levels may
be found in the general information section of the Motorola "MECL Integrated
Circuits Data Book").
High level noise margin is obtained by subtracting (VIHA min) from
(VOHA min); similarly, low level noise margin is (VILA max) minus (VOLA max)
Worst-case noise margin is guaranteed to be at least 0.175 volts for MECL 11,0.125
volts for MECL 10,000 (and MECL III in stud packages), and 0.115 volts for
MECL III in dual in-line packages. Using typical output voltage levels for MECL
circuits, noise margins are usually better than guaranteed - by about 75 millivolts.
A second noise parameter of interest to the designer is obtained by cascading
worst case gates and measuring the minimum "noise" input that will propagate
through the gates. This measurement is more indicative of actual system operation
than dc noise margin, and is often referred to as "noise immunity" or "ac noise
immunity". Testing has shown that this "noise immunity" is typically at least 40
millivolts greater than the dc noise margin specified by voltage levels. However, ac
5
Series Gating
noise immunity is rather difficult to measure. Consequently it is not specified on the
data sheets.
In system design, the user is concerned with noise margin when devices at
different temperatures and different power supply voltages interface with each
other. Figure 1-3 tabulates the worst case change in logic levels as a function of
temperature. Equally important is the change in output levels as a function of
supply voltage (cf Figure 5-2). The logic 1 levels are relatively independent of power
supply voltage, and the change in the output level is typically less than 0.05 of the
VEE change. The change in the ~ level is a function of the resistor ratios in the
current switch and is typically 0.25 of the VEE change. These values illustrate the
rejection of power supply variations that is characteristic of MECL. Detailed
information on noise margin changes due to power supply and temperature
variations is given in Chapters 5 and 6.
MECL Circuit Types
It is possible to connect the basic MECL differential amplifiers together within
a circuit to increase logic flexibility, speed, and power efficiency. Two techniques,
series gating and collector dotting, add the NAND and AND logic functions to the
basic OR and NOR operation of the MECL gate with very little increase in
propagation delay. A third technique, Wired-OR, gives the logic OR function by
tying together two or more emitter-follower transistors. This is used internally in
complex functions to save speed and power and, unlike collector dotting, may also
be used externally by connecting logic outputs together.
Series gating is accomplished by connecting MECL differential amplifiers in a
current-switch "tree", building up from a current source, Q 1, as shown in Figure
1-4. The A input controls the switch, Q2/Q3 through the level shifter ~ and the
1-4: Series Gating
6
Other Logic Techniques
1-5: Collector Dotting
associated resistor diode network. The bias network is modified to provide the
proper voltage level at Q3, a level which is lower than that on Q7 and Q5. The two
upper switch pairs are controlled by inputs Band C. The overall circuit generates the
-- -- -
- - - --
four logic functions: A-B, A'B, A'C, and A-C. MECL circuits use up to three levels of
series gating, permitting up to eight logic functions with one current source.
The propagation delay from an input, to a top current switch is approximately
one gate delay. The propagation delay from an input to a lower level current switch
is slightly longer because of the input level shifter Q0. Typically, the latter takes
about 1.5 gate delays. More specific information is found on the data sheet for a
particular part.
Because of the bias network design, some of the MECL II series gated circuits
require unused inputs to be connected to VOL instead of to the normal VEE
connection_ These circuits are identified by specifying V IL min other than - 5.2 V
on the data sheets.
Series gating is an advantage in MECL logic since it provides the AND or NAND
logic functions. Together with the OR/NOR function of the basic gate, MECL has
the four basic logic functions needed for efficient logic design. Series gating is used
internally in most MECL complex functions and flip flops.
Collector dotting is a second logic technique which is used in the MECL 10,000
series. With it, the logic AND function can be generated by interconnecting
one collector node of separate differential current switches as shown in Figure 1-5.
When connected this way the two 2-input OR gates give the logic function:
Vo = (A + B) (C + D)
7
Sequential Logic: The MECL Flip-Flop
Only one collector resistor (RC) is used for the two transistors Ql and Q2. The
interconnection requires that at least one input to each gate be at a logic 1 level for
the output to be at the logic 1 level. Since it is possible to have both Q 1 and Q2
conducting at the same time (all inputs low), a clamp is used to limit the current in
RC and maintain the output f/J logic level voltage. This clamp consists of R 1 and Q3.
They insure that the Ql/Q2 collector node never goes more negative than
(IBBR 1 + VBEQ3) Propagation delays for all inputs to collector dotted circuits are
equal and are tYPIcally about 20% greater than the basic gate delay.
To allow for temperature variations, the collector-dotted logic functions are
designed to have the same VOL as normal logic gates at TA = 75C when only one
gate has all of its inputs at a logic f/J level. Therefore, when all gates have all their
inputs at a logic f/J level, VOL will be slightly more negative than a normal gate. This
does not limit device operation, but does give an increase in noise immunity for the
logic f/J level.
The collector dot (OR-AND) logic function, series gating, and the Wired-OR
characteristics of MECL combine to provide the means for designing very efficient
and fast complex logic functions.
MECL Flip-Flops
In addition to the basic gate, the flip-flops in a logic line provide a necessary
building block. MECL employs two types of flip-flop circuits; the ac coupled JK
flip-flops found in MECL I and MECL II, and the direct coupled master-slave
flip-flops used in MECL II, MECL 10,000 and MECL III.
The MECL II ac coupled flip-flops are characterized by small capacitors (25 to
30 pF) coupling the J and K inputs to the storage section of the circuit. Because of
this capacitive coupling, the circuit is somewhat sensitive to input rise time.
However, the circuits will trigger on signals of MECL amplitude with edges up to
100 ns long - much longer than any MECL edge. The multiple J and K inputs of the
MC1013/MCI213 and MCI027/MCI227 are especially useful in synchronous
counters or counters with non-binary counting sequences.
A limitation on these flip-flop circuits is their susceptibility to some types of
noise on the input line. For example, a positive going input with a large overshoot
on a Jinput will override a 1 level on another Jinput intended to inhibit that positive
going signal. The overshoot will couple through the capacitor and look like a J input.
A similar condition exists on the K inputs. Fortunately the problem is not a serious
one and its solution requires only that the input signals do not have excessive
ringing. It is recommended that overshoot on the J and K inputs be held to 100
millivolts.
The MECL II type D, and all MECL 10,000 and MECL III flip-flops, use the
direct coupled master-slave circuit as shown in Figure 1-6 for the MC 1670. In each
direct coupled circuit the master is updated while the clock is low, and data is
transferred to the slave on the positive excursion of the clock. This type of circuit
offers better noise protection than the ac coupled circuit and is not susceptible to
overshoot on the inputs. Also, the master-slave flip-flops do not have the rise time
limitations of the ac coupled circuits.
Operation Of Flip-Flop
In the circuit of Figure 1-6 assume that initially Q, CI , C2, R, S, and D are at f/J
levels and that Q is at the 1 level. Since the clocks and the Rand S inputs are low,
8
VCC2
MASTER SLAVE
263
RC4 RC3
100 100 2011
C2

RCl RC2
\
,m is I"
100 100
1010
% R
209 208
;( '00
rt:
1011
1.24
125
UD3 k
'D6
,
D5
125 ,
PD4
"D7 125
Cl
s
\0 R
D v A8
55
102 103
203
2y-
101*-
Ap
SOk
580
272
Ap
Ap


50 k ,
Rp 50 k
700 01
: 50 Ap
100 1.34 k 1.34 k
:> 700 1.34 k
50 k
1'D2
50 k
.
.
Option: Ap = 50 k-ohms or Ap = 2 kohms.
VEE
1-6: MECL III Master-Slave Tvpe 0 Flip-Flop (MC1670)
VCCl

'i-
125
1.34
4

p

13'

C)
"l:$
Flip-Flop Operation
transistors 1 Q3 and 2Q3 are conducting. In the slave section only transistors 2Q6
and 2Q7 are in series with 2Q3. The output of the slave section is fed back to these
two transistors in order to form a latch. Thus, when the clock is low, the output
state of the slave is maintained. In the master section, the curren t path is through
IQ3 and I Q9.
Now assume that the D input goes high. The high-input signal on the base of
I Q4 causes it to conduct, and I Q9 to turn off. The voltage drop across resistor RC 1
causes a low-state voltage on the base of 1 Q II and therefore on the emitter of
I Q 11. Since there is essentially no current flow through RC2, the base of transistor
lQIO is in a high state. This is reflected in the emitter of IQIO, and in turn is
transferred to the base of IQ6. IQ6 is biased for conduction but, since there is no
current path, it does not conduct.
Now assume one of the clocks goes high. As the clock signal rises, transistor
I Q2 turns on and transistor I Q3 turns off. This provides a current path for the
common-emitter transistors IQ5, IQ6, IQ7, and IQ8. Since the bases of all these
devices except 1 Q6 are in the low state, current flow is through 1 Q6. This maintains
the base and emitter of I QII low, and the base and emitter of 1 Q 10 high. The high
state on 1 Q lOis transferred to 2Q4 of the slave section.
As the clock continues to rise 2Q2 begins to turn on and 2Q3 to turn off.
(Reference voltages in the master and slave units are slightly offset to insure prior
clocking of the master section). With transistor 2Q2 conducting and the base of 2Q4
in a high state, the current path now includes 2Q2, 2Q4, and resistor RC3. The
voltage drop across the resistor places a low-state voltage on the base of 2QII, and
therefore on the emitter, of 2QI1. The lack of current flow through RC4 causes a
high-state input to the base of 2Q I O. Finally these states are fed back to the latch
transistors, 2Q6 and 2Q7 and appear on the Q and Q outputs.
As the clock voltage falls, transistor 2Q2 turns off and 2Q3 turns on. This
provides a current path through the latch transistors, "locking in" the slave output.
In the master section, the falling clock voltage turns on transistor I Q3 and
turns off I Q2. This enables the input transistor I Q4 so that the master section will
again track the D input.
A separation of thresholds between the master and slave flip-flops is caused by
R8. The current through this resistor produces an offset between the thresholds of
the transistor pairs 1 Q2/1 Q3 and 2Q2/2Q3. This offset disables the 0 input of the
master flip-flop prior to the enabling of the information transfer from master to
slave via transistors 2Q4 and 2Q9. This disabling operation prevents false
information from being transferred directly from master to slave during the clock
transition, particularly likely if the D input changes at this time. The offsetting
resistor, R8, also allows a relatively slow-rising clock waveform to be used without
the danger of losing information during the transition of the clock.
Both set and reset inputs are symmetrically connected. Therefore, their action
is similar although results are opposite. As a logic I level is applied to the S input
transistor, 1 Q2 begins to conduct because its base is now being driven through I Q 19
which is in turn connected to S. Transistor IQ5 is now on, and the feedback devices
1 Q6 and I Q7 latch this information into the master flip-flop. A similar action takes
place in the slave with transistors 2Q2, 2Q5, 2Q6, and 2Q7.
MECL Family Comparison
A list of MECL circuit characteristics is tabulated in Figure 1-7. The various
families are compared with respect to both features and performance. Because of the
10
The MECL Families
big speed difference between standard MECL II and the higher speed MECL II clock
drivers, many of the figures in the MECL II column do not apply to the complete
line. The following paragraphs describe the MECL characteristics in the order of
Figure 1-7. Differences between standard and high-speed MECL II are poin ted out
when significant.
1-4 ... The first four items in Figure 1-7 are self-explanatory and have been
discussed in the introduction and the preceeding section on the basic gate.
FEATURE
,','" '",-,' - .""
Driver-Sou
PUndqWn ..
";L'W;' ":,_-,,_,,,: _', ,',";'::>"'--"'., """.'::,'


J,.oadirjg Curren'!:
6. Specified DUtputCurrent
1 .... Ma,ximum Qutput Curr/;lnt .
.... $: ..
fliln-Out
.10. Input Capacitance
1.1 .. Oufput Impedance
.
.
,,,,,,-,,,:--,,,,,'" '
.1l:i. G.atePow'er
H"
r1.:

.. Resistors
,1.9, Separate Vee Inputs
20.S.peed-Power Product

emperattJr; .

,-
23.
,;;."il'-;":
...
1-7: MECL Family Comparison
MECLI MECL II MECL 10,000 MECL III
5 ... Maximum input dc loading current is specified on individual circuit data
sheets. The numbers listed here apply to a single input of a basic gate. If a package
input goes to more than one point in a circuit, such as a gate strobe line would,
additional current may be required. Because of higher power in the MECL II clock
drivers, these inputs are specified at 300 pA maximum input current. The input
current difference between the MECL II and MECL 10,000 circuits is due to the 50
kn input resistor used with MECL 10,000.
11
Comments Regarding MECL Features
Calculating the input current, lin, for MECL 10,000 with a worst-case input
resistor value of 30 k,Q (Rin) gives an input resistor current of:
143/-LA
where: VR voltage drop across the input resistor, Rin, with
a logic 1 input,
VEE = -5.2 V supply voltage,
VI = -0.9 V (a typical logic 1 level).
The typical 50 k,Q value will use slightly less current, but either resistance value
is very high compared to the output circuit impedance or the line impedance. The
350 /-LA specification for MECL III applies only to the Hi-Z circuits. The 2 k,Q input
impedance of the Lo-Z circuits requires an additional input current of:
2.15 mAo
6-8 ... Output voltage levels are specified at currents representative of circuit
operation. MECL I and MECL II are normally used with unterminated lines and are
specified for a fanout of 25, i.e., 2.5 mA total output current. Although the
MECL II line driver will drive transmission lines, the output levels are given at
2.5 mA to maintain consistency with the rest of the MECL II family. MECL 10,000
and MECL III are designed to drive 50 ,Q transmission lines terminated to - 2 V dc
(measured from VCC). The current, ITT, required by the line termination is:
ITT = VTT -VI = (-2.0 + 0.9) V
Zo 50 Q
-22 mAo
Consequently, the outputs are specified with 50 ohm loads. The maximum
permissible output currents of 20 mA for MECL I and MECL II, 50 mA for MECL
10,000, and 40mA for MECL III, insure a good safety margin over the specified
currents.
9 ... The dc loading fanouts for MECL 10,000 and MECL III are computed by
dividing the output current by the input current. However, both ac limitations
and current needed in the transmission line termination can be expected to
restrict the system fanout to a smaller number than the one computed.
10 ... Two techniques are used to measure circuit input capacitance. One
method uses an impedance meter, such as the H.P. 48l5A RF Vector Impedance
Meter, to measure impedance and phase angle. The other technique uses a time
domain ref1ectometer (TDR) to measure the effect of capacitance on the impedance
of a transmission line. (The mathematical relationships used to calculate input
capacitance from TDR data are presented later in Chapter 7). Although small, the
12
Comments Regarding MECL Features
input capacitance will affect system rise time and transmission line propagation
delay as a function of fanout at high MECL speeds.
11. .. DC output impedance can be calculated from measurements of the
output voltage as a function of output current: Z = ~ V / 1 The gate output
impedance must be much lower than the line characteristic impedance in order to
provide full MECL signal levels when driving transmission lines. The output
impedance (resistive load) is the parallel value of the output transistor and pulldown
resistor. It should be noted that capacitance charging rate during a negative
transition is limited by current flow through the pulldown circuit.
12-13. .. Gate propagation delay, edge speed, toggle rate, and power
dissipation are standard data sheet information. Propagation delay (tpd) is measured
from the 50% amplitude point on the input signal to the 50% amplitude point on
the output signal. Normally the edge speed given is measured between the 10 and
the 90% amplitude points on the output signal. However, because of the amount of
rounding on the upper 10% of the MECL 10,000 edges, this family is specified with
20 to 80% edge speeds for easier correlation. Nevertheless, 3.5 ns is a typical 10 to
90% figure which can be compared with other families.
14 ... Toggle speeds are minimum rates for the flip-flops in a family. MECL II
has flip-flops with speed figures ranging from 70 to 180 MHz. The 180 MHz figure
listed in the table is for the MC1034, the highest-speed MECL II flip-flop.
15 ... Gate power here is specified for MECL II, MECL 10,000 and MECL III with
open emitter outputs. The MECL I power figure is higher since that family is only
available with output pulldown resistors. MECL II gates with pulldown resistors have
specified powers ranging from 29 to 47 mW per gate depending on the number of
gates per package which share a bias supply, and on the number of complementary
outputs. Gate power for the MECL 10,000 and MECL III gates is specified with open
emitter outputs, as is usual with most ECL product lines. The wide variety of output
loads - both resistors used with transmission lines and pulldown resistors - makes a
power specification under load difficult to define. In a system, the output power is
added to the gate power to find total power.
16-17 ... Open wire length and wire wrap usage are a function of edge speed
and the propagation velocity of the wire. The distances shown are maxima, selected
to give less than 100 mV undershoot at the receiving end of the line with a fanout of
one. Additional information on line driving is found in Chapter 3. Wire wrap may be
used with all families but MECL III. The 1 ns edges associated with MECL III cause
too much reflection from the wire wrap connection to permit practical use. The
open wire maximum line lengths still apply when using wire wraps, unless some form
of resistor damping or line termination is used.
18 ... Damping resistors consist of small resistors (5 to 75 ohms) that are
placed in series with a line at the output of the driving circuit to extend the
permissible line length. The resistor provides a closer match between the line and the
output impedance of the circuit than a direct connection. This match limits
overshoot and ringing, and allows the use of line lengths somewhat greater than
twice the non-damped lengths.
This technique is useful with all MECL circuits except low impedance (Lo-Z)
MECL III. The reason is that the 2.2 rnA input current required by the 2 kn input
resistor would cause a voltage drop which would impose prohibitive loss of noise
margin in the logic 1 state.
19 ... Separate VCC inputs (VCC1, VCC2) are characteristic ofMECL 10,000,
MECLIII, and some of the high speed MECL II circuits. The separate VCC pins are used
13
Comments Regarding MECL Features
to minimize any crosstalk between circuits in a package which might occur with the
high switching currents when driving transmission lines. MECL I and MECL II
circuits, not designed to be used with transmission lines, have no requirement for
separate V CC lines. Separate V CC lines do not affect using the parts and only
require that two package pins be connected to a single ground plane or ground bus.
20. .. Speed-power product is a measure of a logic family's efficiency.
Propagation delay (nanoseconds) is multiplied by the gate power dissipation
(milliwatts) to get a measure of efficiency in terms of energy (picojoules). It is
interesting to note that gate efficiency has improved with each succeeding logic line
introduced. The speed-power product is slightly inaccurate because power figures are
used which do not include output loading (discussed previously). However, TTL
speed-power products can be inaccurate also as they are generally computed for the
circuits operating at a low rate. Such figures would be much worse for circuits
operating near top switching rates.
21. .. Wired-OR is a technique used with all MECL circuits to obtain the logic
OR function by connecting circuit outputs together. With MECL II a maximum of
two output load resistors are recommended per Wired-OR connection to limit
output current. When several (more than 5) circuits are connected with Wired-OR
outputs, it is possible to get a noise spike on the output if all gates are at a 1 output,
and all gates but one are simultaneously changed to a logic f/J. The noise spike is due
to the one gate suddenly having to source the output current previously supplied by
the other circuits. The pulse width is normally less than the gate propagation delay
and of insufficient amplitude to propagate in the system.
22-25 ... The remaining family features are selfexplanatory. Packaging and
temperature range for MECL 10,000 are based on initially introduced circuits. Other
configurations are being investigated to meet future requirements.
14
UsingMECL
The design guidelines presented here are intended to assist the MECL user to
apply MECL families in a system. The rules listed have been tried out in complete
systems with good results. As rise times become less than 3 ns, special design rules
must be followed. For rise times of 1.5 ns or shorter, designing with transmission
lines is necessary.
MECL II, MECL 10,000, and MECL III logic families are treated separately
because of the differences in their capabilities and in design techniques to be used.
Reasons for the rules, methods for applying them, and test data are found in the
following chapters under associated subjects. High speed MECL II clock drivers and
flip-flops are not treated separately and should be used in the manner described for
MECL 10,000. Otherwise the faster edges of these MECL II circuits may cause
sufficient overshoot and ringing to seriously reduce noise immunity or cause false
operation.
1. MECL II Design Rules
The MECL II family of integrated circuits is designed to provide high circuit
speed without special system layout techniques. This feature simplifies design with
this form of emitter coupled logic because most of the techniques used with
saturated logic apply to MECL II. The ability of MECL II to interface with the faster
MECL 10,000 and MECL III families enables MECL II to be used in slower sections of
very high speed systems to gain power economy, eased layout rules, and a large
choice of logic functions.
Since MECL II rise, fall, and propagation delay times are typically each 4 ns,
transmission line techniques are not required. Standard double-sided circuit boards
and backplane wiring are normally used with MECL II.
Because of the wide variety of MECL II system sizes and interfaces, not all
techniques will apply to every system. The designer should use these rules as guides,
modifying them sensibly as required by his particular system.
A. Logic Design Considerations
1. MECL rise, fall, and propagation delay times are a function of fanout and
capacitive loading. Figures 2-1 through 24 show the reduction in speeds with load
for MECL II. Consequently, when MECL II is operating near its upper speed limit,
fanout should be restricted as indicated by the curves. Because of the emitter
follower outputs, fall time and propagation delay to a f/J level is more affected by
capacitive loading than rise time and propagation delay to a 1 level (note that the
curves in Figure 2-4 are steeper than those in Figure 2-3).
2. Fall time and tpd- may be improved by adding a load resistor between the
output and -5.2 Vdc. A 1.8 kn resistor will cut the delay caused by capacitive
15
t
r
RISE TI ME
(ns)
tf. FALL TIME
(ns)
MECL II Parameters versus Loading and Temperature
2-1: Rise Time versus Loading and Temperature
2 0 ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ __ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~
o 10 20 30 40 50 60 70 80
15
5.0
LOAD CAPACITANCE (pF)
o 2 4 6 8 10 12 14 16 18 20
o
FANOUT
2-2: Fall Time versus Loading and Temperature
(Single Pulldown Resistor)
10 20 30 40 50 60 70 80 90 100
LOAD CAPACITANCE (pF)
2 4 6 8 1 0 1 2 1 4 1 6 1 8 20
FANOUT
16
(t
r
is measured between
the 10 and 90% points of
the rising edge)
(tf is measured between
the 10 and 90% po i nts of
the fall ing edge)
tpd+. PROPAGATION
DELAY TIME
(ns)
PROPAGATION
DELAY TIME
(ns)
MEeL II Parameters versus Loading and Temperature
2-3: Propagation Delay tpd+ versus Loading and Temperature
4.0
__ __
o 30 40 50 60 70 100
LOAD CAPACITANCE (pF)
o 2 4 6 8 10 12 14 16 18 20
FANOUT
2-4: Propagation Delay versus Loading and Temperature
2
5
. i,

V ' ..
... down

o 10 20 30 40 50 60 70 100
LOAD CAPACITANCE
o 2 4 6 8 10 12 14 16 18 20
FANOUT
17
(tpd+ is measured
between 50% points
of the rising edges)
is measured
between 50% points
of the fall ing edges)
MECL II Logic Design
loading nearly in half. Load resistors less than 1.8 kn and any load resistors with
fanout greater than 10 should be avoided when using circuits with internal pulldown
resistors. There may be a reduction in noise immunity when the output is in the 1 state,
due to increased output emitter-follower VBE drop.
3. When driving flip-flops at high speed, clock driver circuits should be used.
Such circuits are about twice as fast as the standard MECL II gates. As a result, clock
drivers such as the MC1023 or MC 1026 can provide the bandwidth necessary for
clocking several flip-flops at once as in a shift register or a synchronous counter
operating at high speed.
4. When driving a long string of flip-flops at speeds lower than 50 MHz (clock),
two gates may be operated in parallel for additional drive. The MC I 00 I circuit is
useful in this application since its multiple OR or NOR outputs may be wired
together.
5. The high operating speed of MECL and the effect of loading on propagation
delay must be considered when parallel circuits converge at one point, as shown in
Figure 2-5. Unequal delays along paths A and B can result in momen tary outputs at
point C, each lasting a time equal to the propagation delay difference between A and
B. This can be compensated for by additional timing in the form of a strobe, or by
adjusting the fanouts along A and B. If possible, unused gate inputs can be paralleled
to simulate a larger fanout where required; otherwise a small capacitor can be
substituted for the needed fanout (about 5 pF per gate input is recommended).
2-5: Parallel Signal Paths
A
c
B
Under heavy loading, propagation delay along path A will be less than along
path B because of the use of OR outputs in A as opposed to the NOR outputs in B.
This difference (Figure 2-6) is due to the effect of loading on the fall time, rather
than being due to a timing difference between OR and NOR outputs. As a matter of
fact, under light loading, propagation delays for both NOR and OR outputs are
identical.
When designing clocks for the high speed flip-flops, these timing differences
become increasingly important. For example, the MC I 034 flip-flop can toggle on a
2.5 ns pulsewidth clock. Consequently, timing chain skewing in the order of 2.5 ns
can cause false operation.
18
MECL II System Layout
2-6: Propagation Differences
Input
A ~ ~ ______________
2
3
Input
Propagation Difference
B
4
5
6
B. System Layout Considerations
1. System grounding and propagation delays in interconnecting leads are
factors to be considered before laying out a system. Depending on the type of wire
used, the wiring propagation time of a signal can greatly affect overall system speed.
In normal backplane wiring it is realistic to expect a 2 ns per foot delay. Propagation
delay is less in coaxial cable, but more for signal conductors in a multilayer
circuit board_
2. System sections such as shift registers and synchronous counters should be
on one card. Propagation delay between shift register clocks on separate boards can
cause erroneous operation. Where timing is critical, equal length clock lines (to
shift registers or other circuits on separate cards) should be run from a common
clock to the card connectors. Such lines will also help limit overshoot and ringing
(discussed further in section D, "Backplane Wiring" ).
3. The Wire-OR capability of MECL can be a powerful tool for reducing
power, propagation delay, and package count. However, since the Wire-OR
connection switches current when in operation (2.5 rnA per load resistor), these leads
should be kept as short as possible to avoid crosstalk. It is recommended that Wire-
OR gates be kept within a package or between nearby packages. Wire-OR between
circuit boards should be avoided. An increased propagation delay of 0.5 ns per Wire-
OR should be allowed for.
4. Sections of a system where high fanout may be necessary (such as adders
with look ahead carry) should be kept on one card. Signal path length should be
reduced as fanout is increased to minimize both line delay and reflections.
19
MECL II Circuit Boards
C. Circuit Board Layout Techniques
1. The size of a MECL II system circuit board is not restricted by the logic
family. System requirements should determine card size.
2. Standard single- or double-sided circuit boards with a good ground
distribution may be uscd with MECL II. A low impedance ground is necessary since
any noise on the ground line may be coupled into signal lines. Also, any voltage drop
across ground will subtract from the noise immunity of the MECL II circuits. Ground-
ing techniques are discussed at length in Chapter 5, "Power Distribution".
3. As with TTL, bypass capacitors, between ground and -5.2 V should be used
with MECL. A 1.0 JiF capacitor should be located on the board at the power supply
inputs. Bypass capacitors, 0.0 I JiF to 0.1 JiF, should be connected-in once every four
or five packages. When breadboarding with MECL II without an effective ground
bus, an 0.1 I1F bypass capacitor should be used for every two packages. RF quality
capacitors (low inductance) are recommended because of high circuit speeds.
Unlike TTL, MECL II does not have large current spikes in the circuit during
switching, nor large current changes in the interconnecting circuit board paths
during the transition between 1 and (jJ levels. The function of the bypass capacitors
during switching is to charge the small circuit input capacitance and circuit board
stray capacitance, thus preventing spikes on the power leads.
4. As with any high speed system, signal lines should be kept as short as
possible to minimize ringing and overshoot, as well to simplify timing considerations
arising from the propagation delay of a signal along a conductor. Ringing and
overshoot are due to the intrinsic inductance and capacitance of the line itself, as
wcll as lumped capacitance at the end of the line. Intrinsic inductance and
capacitance are reduced by shortening the lines. A graph of recommended maximum
line length as a function of fanout is shown in Figure 2-7. Since increased fanout
adds capacitance at the end of the line, the line should be shortened as shown by
the following curve.
2-7: Recommended Maximum Line Length versus Fanout
15
12
WIRE
LENGTH 8
(iNCHES)
""

----
--
4
---
0
o 2 4 6 8 10
FANOUT
5. Longer line lengths are possible if a series damping resistor is used. The
resistor is placed at the output of a gate, in series with the signal line as shown in
20
MEeL II Layout Techniques
Figure 2-8. The resistor value depends on the fanout and the required line length.
Resistors under 220 ohms for a fanout of one, or under 100 ohms for fan outs
28: Damping Resistor
greater than five, are normally used for damping. Values larger than these produce rise
and fall time degradation and loss of noise immunity due to IR voltage drop in the
resistor. Resistor values are typically 47 ohms, and can be calculated exactly, if
desired, from the information given in Chapters 3 and 7. Standard 1/4 watt carbon
resistors are commonly used.
6. When driving large fanouts, line lengths can be increased by running parallel
leads as shown in Figure 2-9. The distance between the parallel leads is not critical.
29: Parallel Signal Paths
This technique should be used for shift register clocks, counter resets, and other high
fanout applications. Of course, for synchronous clock lines, clock skew delays
should be matched. Series damping may also be used with parallel signal paths.
7. With most MECL II circuits, undershoot ringing on the logic 1 level is
critical, since it subtracts from the noise immunity. For safe operating margins,
undershoot should be limited to 150 mV. Exceptions to this rule are the MClOl3,
MC1027, and MCI032 ac coupled flip-flops. Because of their ac coupling, overshoot
on the inputs should be limited to 100 mV. Shorter signal leads than shown in
Figure 2-7 may be required when driving these circuits. Master-slave flip-flops, such
as the MCI022 and MC1034, do not have this limitation and can be used with the
same undershoot rules as the basic gates.
D. Backplane Wiring
1. A ground screen is a good means for running a ground in the backplane
wiring. A ground screen is made by connecting heavy bus wires to the connectors in
a grid pattern before wiring the signal lines. The ground screen lines are wired both
parallel to the connectors (tying to the connector pins), and perpendicular to the
connectors (contacting multiple ground pins of each connector). This forms a grid
network (cf Figure 5-6) of approximately 1 to 2 inch squares over which signals lines
are then located.
21
MEeL II Backplanes; System Grounds
2. Ferrite beads may be used in backplane wiring for longer signal runs. The
recommended line lengths discussed for circuit cards also apply to backplane wiring.
A ferrite bead on a wire limits rise and fall time to about 7 ns by attenuating the
high frequency components of the signal. With a bead, lines up to three feet long can
be driven without excessive overshoot.
3. Standard backplane wiring techniques are used with MECL II. Both wire
wrapped and soldered connections perform well. Point-to-point wiring is recom-
mended instead of a laced harness, to lessen line length and reduce crosstalk.
4. For longer signal paths (e.g. between panels or between cabinets) twisted
pair lines are recommended. The twisted pair is connected to the OR and NOR
outputs at the sending end and to an MC 1020 line receiver at the receiving end.
With this technique, long lines (hundreds of feet) have been driven with no system
degradation other than propagation delay down the line.
E. System Considerations
1. A good system ground is required for best performance. All grounds should
be connected to a common ground point - normally near the power supply. All
logic circuits are connected to a circuit ground. All relays, solenoids, motors and
other noise generating devices are wired to a separate ground network connected to
the common ground point. Standard noise suppression techniques should be
employed (i.e. diodes across relays, and capacitors across dc motor brushes).
All mechanical parts such as panels, chassis, and cabinet doors should be
grounded with a third ground. A mounting frame is often used for this if good
conduction can be made at points of contact. If some pieces of equipment in the
system are left ungrounded they may carry transient voltages that will interfere with
the rest of the system. The three separate ground systems connected to a common
Logic Circuits
2-10: Grounding System
Common Point Near Power Supply
Relays
and
Motors
Cabinet Doors
and
Chasses
point will eliminate noise on the signal ground (cf Figure 2-10). Heavy ground leads
should be used on large systems to minimize any voltage drop along the ground line
run.
2. Twisted pair lines and line receivers are normally used between sections of a
system unless line lengths can be kept short. Twisted pair lines should always be
used between sections operating at widely differing temperatures (>40C), or
between sections not connected with a solid ground network.
22
MECL 10,000 Design Considerations
2. MECL 10,000 Design Rules
The MECL 10,000 family is a high speed (2 ns propagation delay), economical
logic family designed to fill the gap between the MECL II (4 ns) and MECL III (I ns)
families and to meet the requirements for future high perfonnance systems. The
family is designed to drive terminated transmission lines with impedances as low as
50 ohms. Also, increased circuit complexity is possible due to high component
densities and very low speed-power products. Finally, the relatively slow edge speeds
of MECL 10,000 minimize wiring constraints on a logic system.
This section contains layout and design guidelines for power distribution,
ground planes, terminations, line lengths, fanout loading, clock distribution, thermal
considerations, and packaging, applicable to MECL 10,000.
A. General Considerations
1. Standard double-sided plated-through-hole printed circuit boards may be
used with the MECL 10,000 family. However multilayer boards will permit a higher
component density for a given board area. As a result interconnect lengths are
reduced, making the highest speed systems possible.
2. Backplane wire wrapping is also acceptable using commercially available
boards. Rules and techniques will be discussed for interconnection lengths and
terminations as a function of loading.
3. Coaxial cable, ribbon cable, or twisted pair line is normally required to inter-
face between drawers and card racks in a large system. Microstrip lines are normally
required for clock distribution with either series or parallel termination. Series
damping resistors can be used to facilitate driving long, untenninated lines.
B. Printed Circuit Card Layout Techniques
1. For double-sided boards, a ground plane is recommended on one side of the
card. This plane provides a stable ground reference for microstrip transmission lines
on the other side of the board. Such transmission lines will have a characteristic
impedance of less than 150 ohms. If a ground plane is not possible, a ground bus
must be used as part of the layout on the board, to provide a low inductance V CC
line.
2. If possible, run the interconnections on one side of the board in the
direction perpendicular to the interconnections on the other side of the board. (This
works nicely for large boards holding 100 or more packages).
3. The ground plane or bus should be connected to 10% of the edge connector
pins spaced equally apart. This reduces the ground impedance, in tum minimizing
crosstalk - since multiple signals do not have to rely on a single ground return path.
4. The V ce I and V CC2 package pins should be connected directly to the
ground plane or bus, as close to the package as possible. Having the two Vce pins and
connecting the collectors of emitter follower outputs to only one V CC pin is designed
to minimize internal crosstalk.
5. The ground for high current devices - relays, lamps, core drivers, etc. -
should be separate from the logic ground. These high current circuits should be
connected to a separate ground bus on the card and in the backplane. The separated
grounds should be connected at the system ground point.
6. Signal interconnection wires between circuits should be kept as short as
possible.
23
MECL 10,000 Power, Backplane, and Loading Guidelines
C. Power Supply Bypassing on Circuit Cards
1. A 1.0 pI< bypass capacitor is used on each board at the power supply inputs.
Decouple every 4 to 5 packages with 0.01 pF to 0.1 pF RF quality capacitors
(- 5.2 V dc to ground).
2. The power supply ground line noise should be limited to less than 50 m V
peak-to-peak.
3. Maintain VEE power supply voltage with less than 100 mV difference
among all the logic cards to which signals must interconnect. (This will limit noise
margin degradation to less than 30 mY).
4. Power supply regulation should be better than 10%.
D. Backplane and Loading Considerations
1. Wire wrapping techniques are acceptable in the backplane as long as the
interconnection rules are followed.
2. A ground screen or ground plane is recommended in the backplane. This
gives backplane wiring a characteristic impedance of approximately 140 ohms. (This
may vary as much as 50% depending on distance from the plane and the route
taken). The capacitance of the wire over the ground screen is about I pF lin and the
inductance is about 20 nH/in. Parallel terminating resistors, as described in Chapters 3
and 7, may be used to increase line lengths in the backplane.
3. 10% of the card edge connector pins should be connected to the ground plane
or screen to reduce card-to-backplane ground impedance. The lowered ground
impedance resulting from many pins paralleled to ground minimizes crosstalk since
several signals do not have to rely on a single ground return path.
4. The optimum choice for backplane wiring (for maximum line impedance
continuity) is the strip line motherboard technique. In such a case, board
interconnections on the motherboard would follow the same rules as the strip line
circuit card. Strip line techniques will be discussed in later sections (Chapter 3 and
Chapter 7).
5. Series damping resistors can be used to series terminate the interconnect
wires as follows:
a. Ten inches of open wire (with a 600 [2 output emitter pulldown resistor
connecting to -5.2 volts) can be driven if a 50 ohm resistor is placed in series at the
sending end. Up to eight loads may be driven using this configuration. Eighteen
inches of line with up to 4 loads may be driven by using a 100 ohm series resistor.
These resistor values will insure that any undershoot will be less than 100 m V.
b. Ten inches of open wire may be driven in series with 10 inches of printed
circuit line (in either order) if a 100 ohm resistor is placed in series at the sending
end. This arrangement can drive up to 4 loads.
6. Three inches of open wiring with a fanout of up to 4 gate loads will produce
less than 100 m V undershoot. A ferrite bead place in the line will increase the open
wiring length to 15 inches.
7. A damping resistor or a combination of series/parallel terminations with
microstrip lines is required when driving nip-flops whenever fanouts exceed 4 and
whenever line lengths are greater than 3 inches.
8. Coaxial cable and twisted pair line are recommended when top speeds and
rock-bottom noise pickup is a "must" for signal paths in a backplane. An alternate
approach, ample for nearly all requirements, is to use strip lines or microstrip lines in
a backplane motherboard as in #4 above.
9. Recommended coaxial cables have characteristic impedances of 50-100 [2,
and time delays of 1.5 ns/ft.
24
MECL 10,000 System Rules
10. Twisted-pair lines may be made of standard hookup wire (A WG 24-28),
twisted about 30 turns per foot. Such twisted pair exhibits a characteristic
impedance of about 110 ohms.
II. When driving coaxial cables, the printed circuit leads from the driver and
receiver (going to the coax) should be kept as short as possible to reduce mismatch
reflections, unless microstrip or strip line is used.
12. When driving 110 ohm twisted pair, the pair line should be terminated
with a 110 ohm resistor across the differential input to a line receiver (MC 10 115 or
Me 1 0 116). A 600 ohm pulldown resistor should be connected to each output of the
gate driving the twisted pair.
13. Twenty feet of twisted line can be driven by a MECL 10,000 OR/NOR
gate at a frequency of 100 MHz, when received by a line receiver.
14. Twisted pair line is recommended for interconnections whenever a
temperature differential of more than 35C exists between sections in a system.
15. Twisted pair lines are recommended when high switching-current lines are
in close proximity to the proposed signal route or when signals run between drawers
or racks. If common mode noise is greater than 1.5 volts, then shielded twisted pair
is recommended.
16. Inductance and overshoot are reduced if parallel lines are used to fanout to
various loads on different circuit boards in the backplane (this also holds for
interconnections on the circuit card). In this way, a parallel fanout of 4 will produce
very little more overshoot than a fanout of one.
17. Twisted pair lines should be used to distribute clock signals to different
logic boards and drawers in a system.
E. System Distribution and Grounding
1. High switching current lines for core drivers, relays, and motors should be
separated physically from logic lines. Ccf Chapter 4, discussion of crosstalk).
2. Avoid bundled parallel runs as much as possible. Signals in bundled cables
produce crosstalk.
3. Signal distribution architecture should minimize wiring delays to permit the
highest possible system clock speed. System clock speeds of greater than 40 MHz
can be obtained in medium size computers.
4. The ground for the high switching curren t circuitry, should be separated
from the logic ground. All separate grounds should, however, be tied together at one
point - the system ground point. In that way, the ground buses will be at the same
potential but current cannot be looped since they are connected at only one point.
5. The cabinet should be strapped to the system ground point to make it serve
as an electrostatic shield.
6. If the system is in a high noise environment, connect the system ground
point to earth ground with a heavy conductor.
7. All the equipment in a system should be grounded.
F. Loading Rules for MECL 10,000
1. A MECL 10,000 gate can drive 25 MECL II gate loads with a 2 kn resistor
connected from the open emitter output to -5.2 volts.
25
MECL 10,000 Loading Effects
2. MECL 10,000 and MECL III fanout rules are the same. Minimum output
pulldown resistor loading is 50 ohms to - 2 volts; 10 gate loads (in addition to the
50.Q) will reduce noise margin by less than 20 mY. Maximum output pulldown
resistor loading is 100 ohms to - 2 volts. Larger resistors result in a loss of logic 0
noise margin. See Figure 2-11 for typical output characteristics of MECL 10,000 as a
function of output load current and the value of the output pulldown resistor.
3. It is recommended that output pulldown resistor values of from 270.Q to
51O.Q (connected to -5.2 V) be used when MECL 10,000 drives MECL 10,000 or
MECL Ill. Under these conditions 25 MECL 10,000 gate loads or 20 MECL III gate
loads (50 k.Q Hi-Z input) may be driven.
4. MECL 10,000 fall time is primarily a function of the load capacitance and
Vout,OUTPUT VOLTAGE
(VOLTS)
2-11: Output Voltage Levels versus DC Loading
-0.2
150 n ;'00 0/75 n/]
V
-0.6
/ /
V
50n
/
"-
II V 1/ V
35 n ,/
/
/ / /
/' /,/
V
/ // V
V
VOH
-1.0
VI V ~ ........-
~ o a d Lines
/ V/ / V
VTT = -2.0 V
-1.4
'I.L
V/",:
V
-1.8
pt-""
VOL -
-2.2
o 4.0 8.0 12 16 20 24 28 32 36 40
lout,OUTPUTCURRENT (rnA)
NOTE:
Load line
impedance is fo r
parallel
terminated
transmission
line.
the emitter pUlldown resistor. If the emitter pulldown is connected to -5.2 V, the
fall time is given by:
tf ;:::; (0.2 RC + 2) ns,
where R is the value of the emitter pulldown resistor (in k.Q) and C is the load
capacitance (in pF). If the emitter pulldown is connected to - 2 V, the fall time is
given by:
tf;:::; 0.1 RC + 2)ns.
5. The propagation delay for the output to go negative is also a function of the
load capacitance and the emitter pulldown resistor. If the emitter pulldown resistor
is connected to -5.2 V, the propagation delay for the output to go negative is:
tpd- ;:::; (0.1 RC + 2) ns.
If the emitter pull down is connected to - 2 V, the formula for the delay is:
tpd- ;:::; (0.47 RC + 2) ns.
26
MECL 10,000 Propagation Delay
6. For computing the signal path delay, the typical gate delay for the basic
gates is 2.0 ns without capacitive loading. With either a 50 n emitter pulldown to - 2
volts, or 270 n to -5.2 V, propagation delay will increase by 0.1 ns per gate load
(assuming 5 pF per gate load).
7. For all MECL 10,000 series devices, the various propagation delays listed in
the data sheets have been measured with a 50 ohm emitter pulldown resistor connected
to -2 volts. Thus, these propagation delays are longer than would occur for a lighter
load condition. Consequently the propagation delays specified on the data sheet are
used to determine maximum delay paths in a system. (Of course as discussed above,
loading will increase the propagation delay and should be allowed for in delay
calcUlations).
8. Emitter dotting is accomplished by tying two or more outputs together.
This produces a logic OR function in positive logic. A logic AND function results if
negative logic is assigned. For either the 50 n or the 270 n pulldown, the
propagation delay will increase by 50 picoseconds per emitter dot. For loading
purposes, each emitter dot may be considered as eq uivalent to 1/2 a gate load (more
precisely, each emitter dot is equivalent to slightly less than 2 pF of capacitive
loading).
9. The MECL 10,000 circuit propagation delay is unaffected by the intrinsic line
capacitance of an unterminated line. However, overshoot at the receiving end could
result in a slightly faster rise time.
10. The MECL 10,000 circuit propagation delay is unaffected by a transmission
line properly terminated at the receiving end. Such lines appear as purely resistive load s.
II. High fanout at the end of a terminated transmission line longer than 1.7 ns
does not increase the propagation delay of a MECL 10,000 circuit driving the line.Fan-
out loading increases the propagation delay of a signal on the line.
12. The delay in signal propagation along a printed circuit line must be taken
into consideration. The basic delay of a signal on either a loaded (resistive loading)
or unloaded printed circuit surface line over a ground plane is about 1.8 ns per foot
or 0.15 ns/in for glass epoxy boards. The exact delay can be calculated using the
tpd = ...J LoCo, where Lo and Co are the intrinsic line inductance and
ca paCltance.
13. The signal propagation delay down the line will increase by a factor of
V I + Cd/Co where Co is the intrinsic line capacitance and Cd is the capacitance
due to loading and stubs off the line:
14. The increase in signal delay due to load capacitances should be calculated
for the particular transmission line characteristics. Lines with low characteristic
impedance are less affected because of their higher intrinsic capacitance per unit length.
15. The characteristic impedance of a transmission line is reduced due to load
capacitances by the fact,or Vi + Cd/Co. So, the formula for the modified
characteristic impedance, Zo , of a transmission line is:
...JI + Cd/Co
where Zo is the original line impedance.
27
MECL 10,000 On-board Line Lengths
16. The maximum line length allowable on the circuit board can be calculated
using the data in Figure 2-12 for printed circuit line resistance.
For lines terminated to -2 Vdc at the receiving end of the line, the signal
voltage drop in the line is:
= \ ~ O H I \
V Sig J (line resistance),
212: Resistance versus Line Width for Printed Circuit Lines
RESISTANCE
PER FT (OHMS)
0.9
0.7
0.5 f---+---f--'
0.3
0.2
0.1
0.07
0.05
0.03
3 5 7 10 20 30 50 70 90
LINE WIDTH (MILS)
where VOH is the logic 1 output voltage and RT is the terminating resistor.
Normally this signal voltage drop is small and need not be calculated. For example, 7
feet of IS-mil wide line will have less than 30 m V drop. The maximum length
allowable will be that for which ~ V remains below about 100 m V.
17. The maximum stub length off terminated lines is 3 inches with a fanout of
four gate loads on the stub, (for <100 m V undershoot). Whenever an open line
(stub) is driven by a pulse, the resultant overshoot and ring are held to about 15% of
the logic swing if the two way delay of the line is less than the rise time of the pulse.
For these conditions the maximum unterminated line length may be calculated:
.Qmax (in.)
where tr is the rise time of the pulse. Here tpd is the propagation delay of I inch of
line (cf #13 above).
28
MECL III Design: Card Layout
18. Up to 3 parallel open lines can be driven by one gate, following the rules
given above. Parallel fanout to loads is recommended when possible, since lead
lengths longer than for a single line may then be used. However, a matched
transmission line should be used for driving loads over lines longer than shown in
Figure 3-13. Note that both stubs and terminated lines can be driven by one and the
same gate.
19. If a ground plane is used, longer lines can be driven than if no ground plane
is used; or else the value of a series damping resistor can be reduced. The best
approach for determining the permissible values of resistance, length of line, and
fanout is from the basic equations that are developed in Chapters 3 and 7.
3. MECL III Design Rules
The MECL III logic family is the fastest standard logic available. This family is
designed to fill the high spccd (1 ns) requirements of the computer, communication,
or instrumentation system designer. MECL III, like MECL 10,000, is designed to
drive terminated transmission lines.
Motorola has successfully met the device/package requirements for a 1 ns logic
family. The ability to manufacture very fine geometry devices with reliable
multilayer metallization results in very compact circuitry and makes LSI possible for
MECL; and so, expansion with complex functions operating at higher data rates, lower
power, and smaller size than any other f0fl11 of logic, has become possible with
MECL. This is a direct result of new processing technologies and the techniques
available to the MECL circuit designer. These techniques include: series gating,
collector dotting, and reducing internal logic swing.
The ability to process data with microelectronic structures at bit rates of over
200 million per second requires a thorough understanding of device circuit design,
system interconnects, packaging, and thermal management. Specifically, the
necessary compromises and possible trade-offs must be understood. A set of layout
ground rules or guidelines will provide a first step toward this goal.
A. Circuit Card Layout
1. Leave maximum possible spacing among all parallel signal leads to reduce
crosstalk. If two signal leads are run parallel at spacings of less than 150 mils, then a
ground lead placed between the parallel wires will reduce crosstalk. Such a ground
shield will reduce crosstalk from 12 to 7 m V for two IS-mil lines spaced 115 mils
with a 3S-mil ground shield centered between. If the ground shield is plated through
to the ground plane every 1/2", the crosstalk will be reduced even further - to
3mV.
2. The choice between two-layer and multilayer printed circuit board depends
upon the maximum operating frequency and the circuit complexity. With clock rates
above 200 MHz, the use of multilayer board is highly recommended. This is due to
the possibility of ground loops caused by the use of ground plane areas as signal paths
on double layer boards. One to three packages, as in a test fixture, may be used
satisfactorily above 200 MHz with two-sided printed circuit board.
29
MECL III Transmission Line
B. Transmission Line (Microstrip Line)
1. Avoid sharp bends in transmission lines, to prevent reflections from abrupt
changes in the characteristic impedance of the line.
2. If two sided board is used, Figure 3-7 may be used to determine Zoo
3. For MECL systems the physical width of microstrip lines used leads to
characteristic impedances usually lying between 50 and 120 ohms. To achieve
impedance values greater than 120 ohms, line widths have to be very narrow. This
promotes two problems. One is that as dc series resistance goes up, signal level at the
receiving end of the line is reduced. The second problem is that "etch-outs" or pin
holes exist after etching narrow lines. As a result of various considerations, it happens
that 68 ohms is a wise choice of impedance.
An impedance of 68 n yields the best trade-off between delay time and power
consumption. A 50 n line would consume more power. A higher impedance would
consume less power, but delay time would increase. As a matter of fact, three
impedance levels can serve most applications: 50, 68, and 100 n. A 68 ohm stripline
is a good choice for on-board uses, while 50 and 100 n are used for single ended or
party line drive (respectively) off the board.
4. Line characteristic impedance, Zo, is inversely proportional to the square
root of the line capacitance. Therefore, known values of gate input capacitance can
be used to modify Zo, i.e.:
Z'
o
(ohms),
where Zo is the new effective characteristic impedance, Cd is the sum of the
capaci tance due to loads distributed along the line (circuit inputs and stray
capacitance), and Co is the intrinsic line capacitance.
The effect of load capacitances on signal propagation delay time, tpd, is:
t
PdR
------ Cd
t' = Z C 1 +_.
pd 0 0 Co
If Cd and Co are in pF, and Zo in k ohms, ti)(.1 will be in ns.
These relationships show that load capacitances increase propagation delay and
will decrease the characteristic impedance. Lines with low characteristic impedance
are least affected due to their higher capacitance per unit length.
The one important advantage of transmission lines with proper termination is
that stubs have little effect on line delay times. With a Zo of 50 n, stubs must be
limited to l" or less to prevent excessive ringing.
C. On-Card Clock Distribution Via Transmission Lines
1. Use of the OR output for gates used as clock buffers is recommended in
developing a clock chain or tree. A small clock skew may result from using both the
OR and NOR outputs in the chain.
2. Use balanced fanouts on the clock drivers in a tree.
3. Overshoot can be reduced by using two parallel drive lines in place of one
30
MECL III Gock Distribution On-and Off-Card
drive line. The effect of this arrangement is to cut the load capacitance per line in
half.
4. To minimize clock skewing problems on synchronous sections of the
system, line delays should be matched to within Ins.
5. It is always good practice to use a buffer when driving long lines off the
card. One instance when a buffer is particularly desireable is when Q or Q outputs
from a high frequency counter are also used within the feedback logic of the
counter.
6. Parallel drive gates are used when high clocking repetition rates are required,
and when driving high capacitance loads. The bandwidth of a MECL III gate may be
extended by paralleling both halves of a dual gate. Approximately 40 or 50 MHz of
bandwidth can be gained with the two (or three) clock driver gates in parallel.
7. Fanout limits should be applied to clock distribution drivers. Four to six
loads should be the maximum load per driver for best high speed performance.
Avoid lumped loads at the end of lines greater than 3 inches long. A lumped load, if
used, should consist of no more than four gate loads.
8. For Wired-OR (emitter dotting), two-way lines are required when connec-
tion distance is greater than 1 inch. A two-way 100 Q transmission line is prod uced
by terminating both ends with 100 Q impedances. Single end termination may be
used when all emitter connections are within I inch of each other.
D. OfFCard Clock Distribution
1. The OR/NOR outputs of an MC1660 may be used to drive into twisted pair
lines. At the far end of the twisted pair, an MC 1692 differential line receiver is used.
The line should be terminated. This may be done with approximately 110 ohms
across the differential receiver input.
Alternatively, a 56-ohm resistor from each of the receiver inputs to - 2 V dc will
provide both line termination and pulldown resistance for the MC 1660 driver. This
latter method not only provides high speed board-to-board clock distribution, but
also yields noise margin advantages for the system. That is, the noise margin from
board-to-board becomes independent of temperature differentials, due to the line
receiver operating with differential inputs.
2. MECL III clock distribution to MECL II logic elements can be done in one
of two ways:
a) Use the OR/NOR outputs or Q/Q outputs to drive twisted pair as above,
receiving differentially with the MECL II line receivers. (MC1020, MCI035, or
MCI065).
b) Use any MECL III single-rail output to drive MECL II logic, but lightly
load the MECL III element (2 kQ to -5.2 V) and maintain the interface lead length
under 1 inch total.
3. MECL III interfaces directly with MECL 10,000. Use the wiring rules for
whichever family drives the line.
E. Testing MECL III
1. Keep all unshielded lead lengths as short as possible, less than 1/4".
2. For stud packages, use a modified Barnes 14-pin flat-pack snap-top socket
(or equivalent). Physically cut into the sides of the socket near the upper lead
31
MECL III Test Fixture
contacts, so that probe points and power may be connected right at the device under
test. (See Figure 2-13).
For dual-in-line packages use AMP 16-pin low profile sockets (or equivalent)
which have no long paths from the device under test to the solder pads on the
bottom of the socket.
213: MECL III Stud Package Test Socket
All dimensions in inches.
Power and Signal Connections __
Directly at Device Under Test
Stud E xtensio n
Brass Stock
r
0.010 Slot, 0.320 Deep
(4 Places)
25/32
1 I
For
._
114 Heat
Sink
Stud Extension i t
Filler Detail I
Nylon Filler
I
1/8 Hole : l
y
16

0.031 Deep A r
/"'1
III -:::v 3/16
13/32, / I ! ,I
y
32
Threaded
Collet
Heat Sink, 2 x 4 x 0.040
Folded Copper Sheet.
MECL III Test Circuit
3. Use small RF quality parts: 1/8 W carbon composition resistors and 0.01 pF
low inductance disc ceramic capacitors.
4. All input/output connections should be made with good quality miniature
50 n semi-ridge coax, and BNC, GR, or miniature coax fittings.
5. A solid ground plane should be used, with V CC pins 1 and 14 or 16
connected directly to the +2.0 volt plane via the shortest possible path.
6. VCC should be at +2.0 volts with VEE at -3.2 volts. The gate under test
should have its output connected to a 3S5/S2-50n sampling plug-in for a Tektronix
568 sampling system (or equivalent). The arrangement shown in Figure 2-14 is
recommended by Motorola for measuring subnanosecond performance.
2-14: Recommended MECL Test Setup
33
This demonstration circuit is capable of measuring frequencies in excess of 150 MHz and propagation delays with
an accuracy of tenths of a nanosecond. MECL II, MECL III and MECL 10,000 logic functions are used in this
circuitry on standard PC board. .
For measuring frequency, the standard technique of opening a precision window ami counting the input frequency
during that interval is used. A 1 ILsec window is generated from a 100 MHz crystal oscillator and a counter chain
made of MCIOl31 flip-flops. The sampling rate is determined by a unijunction oscillator. Translation and wave-
shaping is accomplished with an MCI0115 buffer. This circuitry also provides the reset pulse between sample
intervals. The display is used to show the count and is driven by MECL II MCI045 decoder drivers. The most
significant digit display s 1 and 0 and is driven by two high voltage transistors.
The propagation delay measurement is more complex since the speed requirements necessitate some form of
averaging technique. The two propagation delay inputs are connected to an MC1674 Exclusive NOR gate. The
output of this gate gives a pulse duty cycle proportional to the propagation delay between the two signals.
MECL III circuits, such as the MC1674, are the only commercially available integrated cIrcuits which can perform
this very high speed task - the Exclusive NOR circuit must operate at twice the input frequency. The waveform
out of the Exclusive NOR gate is integrated, giving a voltage level proportional to propagation delay. A ramp
function is generated and compared to this voltage level. An MCI0115 is used as a comparator, outputting a
pulse width window, proportional to the propagation delay of the inputs. This window is used to gate the 100 MHz
oscillator to the display counter used in frequency measurements.
34
Printed Circuit
Board Connections
Any signal path on a circuit board may be considered a form of transmission
line. If the line propagation delay is short with respect to the rise time of the signal,
any reflections are masked during the rise time and are not seen as overshoot or
ringing. As a result, because of the high ratio of rise time to propagation delay time,
signal lines for most MOS circuits may be several feet long without signal distortion.
However, as edge speeds increase with faster forms of logic, the line lengths must be
shorter in order to retain signal integri ty.
Two techniques can be used to enable high speed circuits to operate over
relatively long lines without serious wave shape deterioration. TTL uses an input
clamp for fast negative edges. The energy of the overshoot is clamped at one diode
drop below ground, and this reduces the amplitude of thc following undershoot.
The slower positive-going edges are allowed to overshoot, but are damped out by the
relatively high output impedance (50 to 80 ohms) of the circuit in the logic 1 state.
Also, greater noise immunity in the 1 state makes any undershoot less critical.
The disadvantages of the TTL technique show up at higher bit rates and faster
edge speeds when fanouts along the line are used. Since the reflections are present in
the lines, they will tend to combine at high bit rates to cause signal distortions and
loss of noise immunity.
Consequently, MECL uses another approach for handling reflection problems:
matching the impedance of the line. In this way, reflections are controlled and signal
integrity is maintained.
This chapter discusses circuit interconnections as transmission lines, with the
open line treated as an unterminated line. Although MECL TIl is the only family
with a strict requirement for a transmission line environment, it is expected that
most MECL 10,000 users will use matched impedance lines to improve intercon-
nection distances and signal purity.
Circuit designers have a choice between transmission lines and conventional
interconnect wiring when the distances between MECL devices are short, less than
the lengths in Figure 1-7, #16 or when the rise times are greater than 3 ns. The
design decision must bc made after thorough analysis of the system requirements.
Incorrect selection of conventional interconnect wiring could result in false system
operation due to a high perccntage of incident pulse reflections and subsequent
lowering of the ac noise immunity.
In many cases where MECL devices are used, transmission line techniques are
advantageous. When using MECL devices with rise times less than 3 ns, transmission
lines are highly recommended. The basic factors which will affect this decision are:
A. System rise time
B. Interconnect distance
C. Capacitive loading (fanout)
D. Resistive loading (line termination)
E. Percentage of undershoot and overshoot permissible
(reduction in ac noise immunity).
35
Overshoot and Undershoot On Open Wire Line
The result of analyses shows that transmission lines should be used if the
percentages (E, above) exceed the acceptable design goal. A general rule of thumb
that can be used is that undershoot should not exceed 10%, and overshoot should
not exceed 35% of the logic swing. The 35% overshoot limit keeps the input out of
saturation and the 10% undershoot is less than 100 m V loss of noise margin.
Actually, most MECL circuits can tolerate up to 30% undershoot.
An examplc of a MECL 10,000 device driving an 8-inch open wire is shown in
Figure 3-1. The oscilloscope traces, for the 8-inch open wire without a ground plane,
3': Overshoot and Undershoot With an Open Wire Line
Input Output
1-
8"-
PULSE
GENERATOR
- ---1 MC10109 j
A ~ DB.
~ 5 1 0 ; ---'--1 f'"
-2.0 Vdc
(a) Test Arrangement
(b) Ground Plane Not Used
Vertical Seale ~ 400 m V Icm
Horizontal Scale ~ 20 ns/em
(e) Ground Plane Added
36
Vertical Scale ~ 400 mV fcm
Horizontal Seale ~ 10 ns/em
Advantages of Transmission Line
taken at points A and B are shown in Figure 3-1 (b). Trace A shows an overshoot
condition of 60% and an undershoot of 40%. It can be seen how this undershoot
condition affects trace B during the low level period of the signal - a small spike is
produced.
By way of contrast to the open wire circuit, a ground plane is added and the
trace shown in Figure 3-1 (c) is obtained. The addition of a plane reduces overshoot
and undershoot to about 40% and 20% respectively.
Figure 3-2(a) shows an 8-inch transmission line correctly terminated. The scope
traces in Figure 3-2(b) indicate the marked advantages of using transmission lines
correctly terminated.
3-2: Matched Transmission Line Waveshapes
I nput Output
PULSE
GENERATOR
Lrd --:
500hm
L-_____ -' TransmiSSion Line
50
-2.0 Vdc -2.0 Vdc
(a) Test Configuration
(b) Input and Output Waveforms
Vertical Scale = 400 mV Icm
Horizontal Scale = 10 ns/cm
Transmission Line Geometries
Figures 3-3 through 3-6 show some of the types of transmission lines than can
be used for interconnecting high speed logic systems. Details concerning each type
are elaborated in the following paragraphs.
37
Types of Transmission Line
Coaxial Cable and Twisted Pair
Figure 3-3 shows a twisted pair line and the cross section of a coaxial cable
transmission line. Some common types of coaxial cable have characteristic
impedances of 50,75,93, or 125 ohms. Twisted pairs can be made from standard
hook-up wire (AWG 24-28) twisted about 30 turns per foot. Such twisted pair has a
characteristic impedance of about 110 ohms. Coaxial cable and twisted pair are
recommended for long line lengths in the backplane.
33: Coaxial Cable. Twisted Pair
Wire Over Ground
Figure 3-4 shows the cross section of a wire over a ground. The characteristic
impedance of the wire is
Z = 60 In(4h),
o ~ d
where lOr is the effective dielectric constant surrounding the wire. The wire over a
ground plane is most useful for breadboard layout and for backplane wiring. The
characteristic impedance of a wire over a ground plane in the backplane is about
120 ohms, although this may vary as much as 40% depending on the distance from
the plane, proximity of adjacent wires, and the configuration of the ground.
idi
. ~
h
3-4: Wire Over Ground
~
V 77777777777 7 J Ground
Micros trip Lines
A microstrip line (Figure 3-5) is a strip conductor (signal line) separated from a
ground plane by a dielectric. If the thickness, width of the line, and the distance
3-5: Microstrip
t "'" 0.0015" for 1 oz. Cu.
0.003" for 2 oz. Cu.
f
Dielectric h
~ ~ ~ ~ ~ ~ ~ ~
r o u ~ /
38
Microstrip Line Parameters
from the ground plane are controlled, the line will exhibit a predictable characteristic
impedance that can be controlled to within 5%.
The characteristic impedance, Zo, of a microstrip line is:
where:
87
In ( 5.98h )
0.8w + t '
e
r
= relative dielectric constant of the board material
(about 5 for G-IO fiber-glass epoxy boards),
w, h, t, = dimensions indicated in Figure 3-5.
The signal line is made by etching away the unwanted copper using photo resist
techniques. The characteristic impedance of microstrip lines for various geometries is
plotted in Figure 3-7. These values were calculated from the mathematical relation
above and closely agree with experimental time domain reflectometer measurements.
In fact, the equation proves to be very accurate for ratios of width to height between
0.1 and 3.0 and for dielectric constants between 1 and 15.
Figure 3-8 shows curves for the microstrip capacitance per foot as a function of
line width and spacing.
The inductance per foot may be calculated using the formula:
where:Z
o
Co
characteristic impedance,
capacitance/ft.
The propagation delay of the line may be calculated by:
ns/ft .
Note that the propagation delay of the line is dependent only on the dielectric
constant and is not a function of line width or spacing. For G-IO fiber-glass epoxy
boards (e
r
=:: 5.0) the propagation delay of the microstrip line is calculated to be
1.77 ns/ft.
Strip Line
A strip line (Figure 3-6) consists of a copper ribbon centered in a dielectric
medium between two conducting planes. If the thickness and width of the line, the
-
w
-
3-6: Stripline
V / L
---
t h
b
[ZZ]
Ground Plane
Strip Line

; t
// /
---
Ground Plane
39
Impedance and Capacitance Data for Micrastrip Line
3-7: Impedance versus Line Width and Dielectric Thickness for Microstrip Lines
Z, IMPEDANCE
(OHMS)
LINE WIDTH (MILS)
3-8: Capacitance versus Line Width and Dielectric Thickness for Microstrip Lines
CAPACITANCE/FT
(pF)
100
80
60
40
20
20 30 40
40
50 60 70 80 90
LINE WIDTH (MILS)
110
Impedance and Capacitance Data for Strip Line
3-9: Impedance versus Wire Width and Spacing for Strip Lines
Z, IMPEDANCE
(OHMS)
LINE WIDTH (MILS)
3-10: Capacitance versus Line Width and Spacing for Strip Lines
CAPACITANCE (pF)
10 15 20 25 30 35 40 45 50 55 60 65 70 75
LINE WIDTH (MILS)
41
Signal Behavior On a Transmission Line
dielectric constant of the medium, and the distance between the ground planes are all
controlled, the line will exhibit a characteristic impedance that can be held constant
within 5%. The characteristic impedance of a strip line is theoretically:
Z =
o
60 I (4b )
n .
This equation proves accurate enough for w/(b-t)<0.35 and t/b<0.25.
Figure 3-9 gives the actual characteristic impedance for various geometries of
stripline. These values were measured with a time domain reflectometer. The
measured results closely parallel those calculated from the above equation.
Figure 3-10 shows curves for the stripline capacitance per foot for various line
widths and spacings. An LC meter was used to to determine the capacitance. 2
The inductance per foot can be calculated using the relation Lo = ZoCo,
while the propagation delay of the line can be found from the relation:
tpd = 1.017 ns/ft.
For G-IO fiber-glass epoxy boards (e
r
==: 5.0), the propagation delay of the
strip lines is 2.26 ns/ft. Again, the propagation delay is not a function of line
width or spacing.
Basic Transmission Line Operation
The behavior of signals on a transmission line is important for understanding
the methods used to terminate MECL lines. Figure 3-11 shows a line with typical
3-11: MECL Transmission Line
- 2.0 Vdc
loads at both ends. For the purpose of discussion the line delay will be long with
respect to the rise time so that reflections will appear at their full amplitude. The
output voltage swing at point A is a function of the internal voltage swing, output
impedance, and line impedance:
Since Ro is small compared to line impedance, the output swing is nearly the same
as the input transition. The internal voltage swing is approximately 900 m V, giving a
typical output swing greater than 800 m V.
42
Signal Behavior On an Open Line (Stub)
This signal propagates down the line and is seen at point B time TD later. The
voltage reflection coefficient at the load end of the line, P
L
, is a function of the line
characteristic impedance and the load impedance:
Clearly, for the ideal case of RL = Zo, there is no reflection. More important, for
any value of RL close to Zo the reflection is quite small. At time 2 TD any reflection
returns to point A and is again reflected, by the sending end reflection coefficient PS:
The reflection continuc:s bouncing back and forth between the ends of the line,
being successively reduced by the reflection coefficients and the resistance in the
line.
Untenninated Lines
Figure 3-12 shows a specific transmission line variously known as an "open
line," an "unterminated line" or a "stub." Behavior of this line is as follows. At
time zero an initial, full MECL signal starts at point A. Time TD later the signal
reaches point B and is reflected by P
L
discussed previously. Since the input
impedance of the driven gate is very high with respect to Zo, a large positive
reflection occurs and signal overshoot results. At time 2 TD the reflection is back at
point A and is reflected by P s- Because of the low value of Ro the reflection is in the
negative direction (refer to equation for PS), resulting in a signal at point B at time
3 TD that is in the opposite direction to the initial signal. This signal at B at 3 TD
and its subsequent reflections produce the undershoot which subtracts from signal
noise immunity. These reflections, successively smaller, cause the condition known
as "ringing" as shown in Figure 3-1(b).
3-12: MECL Unterminated Transmission Line
If the lines are sufficiently short, the signal still will be rising at time TD, and
reflections are part of the rising edge. With longer lines, the rise of the signal will be
completed before a time TD, and reflections will appear as overshoot and
43
Maximum Open Line Lengths: MECL 10, 000
undershoot, For this reason, unterminated or undamped lines have maximum
recommended lengths when used with MECL logic.
The undershoot caused by an unterminated line is held to about 15% of the
logic swing if the two way delay of the line is less than the rise time of the pulse.
The maximum open line length may be calculated by expressing this rule with the
relation:
where: tr = rise time,
tpd = propagation delay of the line per unit length.
(Use tpd when line is loaded, cf equation 11, Chapter 7).
It can be seen that the slower rise times of MECL II and MECL 10,000 are
compatible with open lines, but that line lengths are important for the faster MECL III.
The other variable for line length, tpd, is controlled by the type of line (velocity
factor) and the loading on the line. Increased loading raises the propagation
delay and accounts for the decreasing permissible line length with increasing fanout
(cf Figure 2-7). The analysis of rate of propagation with line loading is covered in
Chapter 7.
Suggested maximum open line lengths for MECL 10,000, high speed MECL II
(i.e. MECL I1-1/2) and MECL III are tabulated in Figures 3-13, 3-14, and 3-15 for
3-13: Maximum Open Line Length for MECL 10.000 (Gate Rise Time = 3.5 ns)
MICROSTRIP
(Propagation
Delay
0148 ns/in.)
STRIPLINE
(Propagation
Delay
0.188 ns/in.)
BACKPLANE {
(Propagation
Delay
0.140 ns/in.)
Zo
(OHMS)
50
68
75
82
90
100
50
68
75
82
90
100
100
140
180
FANOUT - 1
(2.9 pF)
Q MAX (IN)
8.3
7.0
6.9
6.6
6.5
6.3
6.5
5.6
5.3
5.2
5.1
4.9
6.6
5.9
5.2
FANOUT -- 2 FANOUT 4
(58 pFI ( 11.6 pF I
Q MAX (IN) Q MAX (IN)
7.5 6.7
6.2 5 . .0
5.9 4.6
5.7 4.2
5.4 3.9
5.1 3.6
5.9 5.2
4.9 3.9
4.7 3.6
4.4 3.3
4.3 3.1
4.0 2.8
5.4 3.8
4.3 2.8
3.6 2.1
44
FANOUT = 8
(23.2pF)
QMAX (IN)
5.7
4.0
3.6
3.3
3.0
2.6
4.5
3.2
2.8
2.6
2.4
2.1
2.8
1.9
1.3
Maximum Open Line Lengths: MECL II and MECL III
3-14: Maximum Open Line Length for High Speed MECL II (Gate Rise Time = 2 n5)
MICROSTRIP
( Propagation
Delay
0.148 ns/in.)
STRIPLINE
( Propagation
Delay
0.188 ns/in.)
BACKPLANE {
(Propagation
Delay
0.140 ns/in.)
MICROSTRIP
(Propagation
Delay
0.148 ns/in.)
STRIPLINE
(Propagation
Delay
0.188 ns/in.)
BACKPLANE
{
(Propagation
Delay
0.140 ns/in.)
Zo
FANOUT ~ 1 FANOUT ~ 2
(OHMS) (33 pFI (6.6 pF)
Q MAX (IN) Q MAX (IN)
50 3.5 2.8
68 3.2 2.3
75 3.0 2.2
82 2.9 2.0
90 2.8 1.9
100 2.6 1.8
50 2.8 2.2
68 2,5 1.9
75 2:4 1.7
82 2.3 1.6
90 2.2 1.5
100 2.0 1.4
100 2.8 1.8
140 2.4 1.4
180 2.0 1.0
3-15: Maximum Open Line Length for MECL III
Zo
FANOUT ~ 1 FANOUT ~ 2
(OHMS)
(33 pF) (6.6 pF)
Q (MAX (IN) QMAX (IN)
50 1.6 1.1
68 1.4 0.8
75 1.3 0.8
82 1.2 0,7
90 1.1 0.6
100 1.0 0.5
50 1.2 0.8
68 1.1 0.7
75 1.0 0.6
82 0.9 0.6
90 0.9 0.5
100 0.8 0.4
100 1.1 0.6
140 0.8 0.3
180 06 0.2
45
FANOUT ~ 4 FANOUT ~ 8
(13.2 pFI (26.4 pFI
Q MAX(lN) Q MAX (IN)
1.9 1.2
1.5 0.8
1.3 0.7
1.2 0 .. 6
1.0 0,5
0.9 0.4
1.5 1.0
1.2 0.6
1.1 0.6
0.9 0.5
0.8 0.4
0.7 0.3
0.9 0,4
0.5 0.3
0.3 0.1
(Gate Rise Time 1.1 ns)
FANOUT ~ 4 FANOUT ~ 8
(13.2 pF) (26.4pF)
QMAX(lN) Q MAX (IN)
0.7 0.6
0.5 0:4
0';4 03
O ~
0.2
0.3 0.2
0.2 0.1
0.6 0:5
0:4 0.3
0.3 0.2
0.3 0.2
0.2 0.1
0.2 0.1
0.2 0.1
0 0
0 a
Control of Waveshape On Long Lines
various fanouts and line impedances. For these tables, line lengths are chosen to
limit overshoot to 35% of logic swing and undershoot to 12%.
Series Damped and Series Terminated Lines
Overshoot and ringing on longer lines may be controlled by using series
damping or series terminating techniques. Series damping is accomplished by
inserting a small resistor (typically 10-75[2) in series with the output of the gate as
shown in Figure 3-16. This technique can be used with all MECL families and is
associated with lines not defined by a controlled characteristic impedance, (e.g. back-
plane wiring, circuit boards without ground plane, and most wire wrapped
connections).
The series termination is a specialized case of damping in which the resistor
value plus the circuit output impedance is equal to the impedance of the
transmission line. The waveforms in Figure 3-16 and the following description of
operation are for series termination. A similar analysis may be done for any value of
damping resistor and line impedance.
3-16: Driving a Series Terminated Line
B
c ______ __________
The impedance looking back toward the driving gate at point B should be
equal to the characteristic impedance of the transmission line. The dc output
impedance is 5 ohms for a MECL III gate and 7 ohms for a MECL 10,000 gate. AC
output impedance is only slightly higher than the dc impedance values. Therefore, if
Zo is 75 ohms, then the value of RS must be approximately 68 ohms.
At time = 0, the internal voltage in the circuit switches to the low state which
represents a change of 0.9 volts = -0.9 V). The voltage change at point B
can be expressed as:
where Ro is the output impedance of the MECL gate.
Since RS + Ro is made equal to Zo for a series terminated line, then the
voltage change at B is 1/2 the voltage, VINT. It takes the propagation delay time
46
Pros and Cons of Series Termination
of the transmission line, TD, for the waveform to reach point C, where the voltage
doubles due to the near unity reflection coefficient at the end of the line. The
reflected voltage, which is equal to the sending voltage, arrives at point B at a time,
TD, later. No more reflections occur if RS + Ro is equal to ZOo Similar waveforms
occur when the driving gate switches from the low to the high state.
One of the advantages of using series terminated lines is that only the logic
power supply is required. Another advantage is the lower overall power require-
ments. One power supply can also be used with parallel terminated lines described in
the next section, but two resistors must be used for the total termination resistor,
resulting in the need for considerably more power. In addition, when two power
supplies arc used with parallel terminated lines using one termination resistor, an
extra voltage bus or plane is required to supply - 2.0 volts to the termination
resistors.
A disadvantage of series termination is that distributed loading along the line
cannot be used, because of the half-voltage waveform travelling down the line (see
Figure 3-16, waveform B). However there is no limit on the number oflumped loads
that can be placed at the end of the series terminated line imposed by reflections at
the receiving gate, since all the reflections will be absorbed at the source.
Nevertheless, voltage drop across the series terminating resistor due to input current,
limits loading to less than 10.
The distance permitted among the receiving gates at the end of the line can be
found from Figures 3-l3, 3-14, or 3-15. For example, if MECL III were used with 50
ohm microstrip lines, the maximum total separation of four gate loads at the end of
a series terminated line is 0.7 inches (see Figure 3-15).
The disadvantages of slower propagation delay and using only lumped loading
at the end of a series terminated line can be eliminated at the expense of more
transmission lines, as in Figure 3-17. For parallel fanout, n transmission lines can be
=D
A
fRe
VEE
3-17: Parallel Fanout Using Series Ternnination
RS
Zo
B
1
'\IV\, ()
n
-
Ide
I
I
hS=
I
I
I I
: n (Total Number of Lines) I
: 4=>-
I
R-=>-
47
I
I
I
I
4>-
Designing With n Parallel Lines
used. The value chosen for RS should be the same as discussed previously when n
was equal to one.
To determine the value of the emitter pulldown resistor, RE, the following
procedure is recommended.
The value of RE must be smaU enough to supply each transmission line with
the necessary current. If RE is made too large, the output transistor will turn off
when switching from the high to the low voltage state. The maximum value for RE
can be derived hy equating the voltage point at which the output transistor turns off
with the midpoint of the logic swing:
~ VB = ~ V TNT ( Zo ) ,
RS + Ro + Zo
where: ~ V B = one half the logic swing = 400 m V,
So:
~ V N T
VEE - logic I level = (5.2 - 0.8) V = 4,400 mY, (since the
output transistor is turned off, it does not affect the
calculation),
RS series damping resistance,
Ro RE (because the output transistor is turned off) .
RE(max) = 10 Zo - RS'
Finally, when n parallel lines are driven as in Figure 3-17:
10 Zo - RS
RE(max) = n
For n
170 ohms.
75 ohms, and RS = 68 ohms, this relation gIves RE
Figure 3-18 shows a circuit using MC 1 026 logic gates. The driving lines have a
width of 50 mils and a board thickness of 62 mils. This geometry corresponds to a
line impedance of approximately 75 ohms. The length of each line is 8 inches, which
produces a line propagation delay of 1.2 ns. The rise and fall times of the driving
gate are about 2 ns each. Figure 3-19 shows the trace seen on a Tektronix 567
oscilloscope using the high impedance probe. The waveform of the line output when
RE = 180 ohms (close to the value calculated above) shows that the rise time and
overshoot of the rising edge are equal to that of the falling edge. The small overshoot
of about 50 mV is due to the line impedance being slightly larger than 75 ohms. This
does not affect circuit operation in any way. The rise and fall time at the line output
are each 3.3 ns.
Figure 3-20 shows the waveforms when RE = 600 ohms. In this case the value
of RE is much larger than the 170 ohms value calculated. Consequently, the faU
time of the waveform suffers since the output transistor turns off and RE is unable to
supply the proper line current. When the output transistor turns off, the output
impedance of the gate becomes that of the pulldown resistor. Calculating the voltage
48
hYfeets of the Emitter Pulldown Resistor, RE
3-18: Series Termination Test Set-Up Using the MC1026 Gates
PULSE
GENERATOR
A
Vertical Scale = 0.2 V fcm
Horizontal Scale = 10 ns/cm
3-20: Waveforms from Test Set-Up of Figure 3-18
(R E = 600 ohms, n = 4)
49
600
600
600
)0---+-_-0 Gate Output
600
Line Output VEE
3-19: Waveforms from Test Set-Up of Figure 3-18
(RE = 180 ohms, n = 4)
Vertical Scale = 0.2 V fcm
Horizo ntal Scale = 10 nsfcm
Waveforms With Series Damping
change at point B shows a .6 V of:
Zo
.6V
B
.6V
INT
n
Ro
n n
(c4,400) (
19
19) 600 + 17 +
::::: -130mV,
where A V INT is the voltage drop in millivolts across the pulldown resistor when high,
and n is the number of parallel series terminated lines.
When the waveform reaches the end of the line, the voltage will double to - 260
millivolts and a reflection of - 130 millivolts will be sent back toward the driving
gate. Since the driving gate output is turned off, the reflection coefficient at the
source is approximately 0.8. Therefore, after a time of twice the line delay, an
additional - 200 millivolts is received at the load. These reflections continue until the
voltage at the end of the line reaches the logic f/J state.
These steps in voltage can be seen in the falling edge of the line output
waveform (Figure 3-20), in close agreement with the calculations. The fall time
increases by approximately six times the line propagation delay, or 7.2 ns. If the
transmission line had been longer, the voltage step duration would have increased
correspondingly. Note that the gate output at the end of the line also has an
increased rise time and propagation delay.
Figure 3-21 shows the waveforms from the test setup shown in Figure 3-18,
when only one line is driven (n = 1) and with RE = 600 ohms. Using the equation
321: Waveforms from Test SetUp of Figure 318with Only One Line Driven
(RE = 600 ohm, n = 1)
i
Line
Output
Gate
Output
J
V
Vertical Scale = 0.2 Vfcm
Horizontal Scale = 10 nsfem
50
Series Damping: Determining the Resistor Value
for RE(max) gives a value of 680 ohms. Note that the rise and fall times are
approximately equal (2 ns) meaning that the proper pulldown resistor was chosen.
The rise and fall times at the line output are much faster in Figure 3-21 than in
Figure 3-19, due to the lighter load at the gate output and reduced nodal
capacitance at point A.
Analysis of series damping is very similar to that for a series terminated line.
Differences are the line length and the value of the series damping resistor, RS' For
series damping this resistor value is normally smaller than the characteristic
impedance of the line. Accordingly line lengths are permitted which are longer than
the worst-case open line lengths (RS = 0), as defined in Figures 3-13, 3-] 4, and
3-15. The same equations for voltage at point B and maximum RE apply, as did for
series terminated lines. In fact, series damping can be used to extend lines to any
length, while limiting overshoot and undershoot to a predetermined amount. Figures
3-22 and 3-23 give minimum values of RS for various line impedances for MECL
10,000 and MECL III. For these figures, overshoot was limited to 35% of signal
swing and undershoot to 12%. The technique for calculating these RS values is given
in Chapter 7.
Here is an example of how Figure 3-22 and 3-23 can be used. Assume that a
MECL III gate must drive a fanout of 2 (6.6 pF) at the end of 1 foot of line in the
backplane. The characteristic impedance in the backplane is between 100 and 180
3-22: Minimum Values of RS for Any Length of Line with Specified
Limits of Overshoot and Undershoot, Using MECL 10,000
MIN RS (OHMS) UNDERSHOOT OVERSHOOT
Zo (OHMS) FOR Ro = 15 % %
50 9 12 34.6
68 18 12 34.6
75 21 12 34.6
82 25 12 34.6
90 29 12 34.6
100 34 12 34.6
120 43 12 34.6
140 53 12 34.6
160 63 12 34.6
180 72 12 34.6
3-23: Minimum Values of RS for Any Length of Line with Specified
Limits of Overshoot and Undershoot, Using MECL III
MIN RS (OHMS) UNDERSHOOT OVERSHOOT
Zo (OHMS)
FOR Ro = 6 % %
50 18 12 34.6
68 27 12 34.6
75 30 12 34.6
82 34 12 34.6
90 38 12 34.6
100 43 12 34.6
120 52 12 34.6
140 62 12 34.6
160 72 12 34.6
180 81 12 34.6
51
Parallel Terminations
ohms. An open line should not be used because it exceeds the length given in Figure
3-15: 0.6 in. Another method therefore must be used - coax, twisted pair, or series
damped line.
If series damping is used, then from Figure 3-23, a series damping resistor of 81
ohms or larger should be placed in the line at the driving end. The maximum value
of the series damping resistor that should be used is 130 ohms for a fanout of 2,
since there will be a dc level shift of 90 m V (maximum) caused by the series
resistance in the line when the driving gate is in the high state. The 90 m V figure is
based on the maximum input current, linH, of the MECL III (high Z) gate being
350 fJ.A. (V = IR, where R = l30n, and I = 2 x 350 fJ.A).
Both the maximum overshoot and undershoot that can occur are given in the
tables. If the proper value of RS (series damping resistor) is used, as given in the
tables, there is no restriction on line length or capacitance at the end of the line for
the specified undershoot and overshoot. Of course, ohmic line losses and line
propagation delay effects must be considered in the design.
Parallel Terminated Lines
Parallel terminated lines (Figure 3-24) are used for fastest circuit performance
and for driving distributed loads. MECL 10,000 and MECL III are specified to drive
"50 ohm lines." This refers to a line, terminated at the receiving end through a
resistor of the characteristic line impedance to - 2 volts from the V CC supply. With
parallel terminated lines, the line termination supplies the output pulldown.
Consequently no other pulldown resistor is required at the output of the driving
gate.
3-24: Driving a Parallel Terminated Line
~ T D ~
The operation of the parallel terminated line is comparatively simple. The
signal swing at point A is:
Since ,1 VINT is approximately 0.9 volts and the output impedance is low
(Ro ~ Zo), the signal swing at point B is typically greater than 800 millivolts. This
signal propagates down the line, undistorted, in time TD. Since the terminating
resistor equals Zo, there is no reflection and the sequence is ended.
52
Thevenin Equivalent Parallel Termination
An important feature of parallel termination is the undistorted waveform along
the full length of the line. It should be noted that parallel termination can also be
used with wire-wrap and backplane wiring where the characteristic impedance is not
exactly defined. By approximating the characteristic impedance, the reflection
coefficient P
L
will be reasonably small, so overshoot and ringing will be held to
within safe limits.
For large systems where total power is a consideration, the lines are normally
terminated to a - 2 Vdc supply. For power conservation, this is the most efficient
manner of terminating MECL circuits. The drawback, of course, is the requirement
of an additional supply voltage.
An alternate approach is to use two resistors in the way depicted in Figure
3-25. The Thevenin equivalent of these two resistors is one resistor equal to the
325: Parallel Termination with a Single Power Supply
Rl
-5.2
THEVENIN EQUIVALENT RESISTORS
FOR TERMINATION
Zo
R1 R2
(OHMS) (OHMS) (OHMS)
50 81 130
70 113 182
75 121 195
80 130 208
90 146 234
100 162 260
120 194 312
150 243 390
characteristic impedance of the line and terminated to - 2 V dc. R I and R2 may be
obtained as:
Transmission Line Comparison
R2::: 2.6 Zo
R2
RI ::: ---
1.6
Since there are advantages to both series and parallel lines, the decision to use
one or both methods in a system depends on the preference of the designer and on
his system requirements. Figure 3-26 lists typical cases where terminations may be
necessary, along with techniques which may be used.
Parallel terminated lines have the advantage when speed is the main factor.
Loading a long line will not affect the propagation delay of the driving gate nor its
edge speed, but loading does increase the propagation time of the signal down the
line. It will be shown in Chapter 7 that the increase in delay time with loading is
about twice as great for series damped lines as for parallel terminated lines. For short
lines the capacitive load increases the propagation delay of the gate by slowing down
the edges.
As mentioned previously, a big advantage of parallel termination is that the
signal is undistorted along the full length of the line. When driving a large fanout, the
loads may be distributed along the line with short stubs, instead of being lumped at
the end of the line as is done with series termination. On the other hand, series
53
Discussion of Line Options
326: Types of Lines Recommended
SITUATION
1'. Line lengths are shoner than
(Fig.3.13,14,-15t:
.. 2",'Or!ving gate drives 1 .... _._
,._-,,,,,4,.,lettgtb..graater.. than specif.ied .. _
"._ ... ,'. . 1.3.44,_15l,
... , ,,a:;$O-rWlng 'gate drives' 3' or more'w
.. lengths. greater than
5petMied (Fig. 313 14.:15).
4;. Gate.loads must be distributed
transmission line.
are lUmped at . , '"
.. .
... '+lne',
5. Only one power supply isto be
useCfand the LOWEST pOWer
" . ." ..
.. .
. ., . " .... . J;he .. ,lo,west.power.consu.mPti.oll.is" ..
. .
. wire lengths are shorter'
tiler.! specified (Fig, .
,.,.
. ', "".- ..g;. lengthsar.e. k! nget;.
.
. ., .. .
. '" backplane. ,:.,
PARALLEL
TERMINATED LINE
Yes
Yes
Pairs
54
SERIES
TERMINATED LINE
Yes
Yes
OPEN
LINE
Wire wrap
termination has the ability to drive several parallel lines, as long as the drive current
is compensated by the value of the output pulldown resistor. The MECL 10,000 and
MECL III outputs will drive only one 50-ohm parallel terminated, or two lOa-ohm
parallel terminated lines. Exceptions to this rule include the MC 10 110 and
MC 10111 which have multiple gate outputs for driving three parallel 50 ohm lines.
Termination power is lowest for a parallel terminated line terminated to
-2 Vdc. However, similar power savings may be realized by connecting the pulldown
resistor for open wire or series terminated lines to - 2 V dc. Using a single power
supply, the series termination and pulldown resistor uses less power than the
two-resistor parallel termination. Typical power in the terminating resistors for 50
ohm lines for signals with 50% duty cycle is tabulated in Figure 3-27.
Additional information for calculating system power is contained in Chapter 5,
"Power Distribution."
Crosstalk on circuit boards is normally not a problem with MECL, because the
relationship of the signal line to the ground plane minimizes the energy coupled to
adjacent lines. Even so, series terminated lines have less crosstalk than parallel
terminated lines. The reason is that only one-half the logic swing is sent down the
series terminated line. As a result the switched current is only one-half that of the
larger, parallel terminated signal. This smaller signal energy results in less crosstalk.
327: Power Consumption for Various 500hmLine Terminations
RESISTOR
TERMINATION RESISTOR POWER
SCHEME ARRANGEMENT CONSUMPTION
Parallel 50 ohm to -2 Vdc 13 mW
Series 510 ohm to VEE 30 mW
Parallel Combination
82 ohm to Vee, and
144 mW
130 ohm to VEE
Wire wrapped Cards
Wirewrapped cards can be used with both MECL U and MECL 10,000. The fast
edge speeds (1 ns) of MECL III exceed the capabilities of normal wirewrapped
connections. Mismatch at the connections causes a reflection which distorts the
fast signal, reducing noise immunity significantly or causing erroneous operation.
The mismatch remains with MECL II and MECL 10,000 but the distance between
the wirewrap connection and the end of the line is well within the allowable
stub-length distance, so the reflections cause no pro blem.
For lines longer than maximum allowable open line length for MECL 10,000 or
MECL II, either series or parallel termination may be used. The parallel resistors are
relatively high (typically 100 to 150 ohms) and are normally used only with MECL
10,000 because it can supply the output current required by the pull down resistors.
These resistors may be used with MECL II, but there will be a loss of approximately
100 millivolts noise immunity in the logic 1 state. Of course series damping resistors
may be used with wire wrapped lines for both MECL II and MECL 10,000. Twisted
pair lines may be used for longer distances across large wirewrapped cards. The
twisted pair gives a more defined characteristic impedance (than a single wire),
and can be connected either single-ended, or differentially using a line receiver.
55
Wire wrap
Twisted pair line driving is an important feature of MECL circuits and is
discussed in more detail in the next chapter. The recommended wirewrapped circuit
cards have a ground plane on one side and a voltage plane on the other, to insure a
good ground and a stable voltage source for the circuits. In addition, the ground
plane near the wirewrapped lines lowers the impedance of those lines and facilitates
terminating the line. Finally, the ground plane serves to minimize crosstalk between
parallel paths in the signal lines. Point-to-point wire routing is recommended because
crosstalk will be minimized and line lengths will be shortest.
Production area for MECL Integrated Circuits
56
System
Interconnections
Signal connections between logic cards, card panels, and cabinets are important
for obtaining the maximum system performance possible with MECL circuits. To
understand how ringing and crosstalk affect system operation, it is helpful to review
guaranteed noise margins, discussed in Chapter 1.
Noise margin is defined as the difference between a worst case input logic level
and the worst case threshold closest to that logic level. Guaranteed noise margin
(N.M.) for MECL 10,000 is:
NM 1 level VOHA min - VIHA min
-0.980V - (-1.105 V) = 125mV;
NM0 level VILA max - VOLA max
-1.475 V - (-1.630 V) = 155 mY.
Thus, using the worst case design conditions, the circuits have 125 m V to guard
against signal undershoot, and power or thermal disturbances. However, using
typical logic levels of - 0.900 volts and - 1. 700 volts, the circuit noise protection is
typically greater than 200 m V for both the logic 1 and logic 0 levels. Power and
thermal design will be discussed in Chapters 5 and 6.
Good circuit interconnections should allow no more than 100 to 110 m V
undershoot. The overshoot and undershoot waveform conventions are shown in
Figure 4-1.
41: MECL Waveform Terminology
Overshoot {
-----J----\-- 1"---_:------- 1 Level
Undershoot 1 ___ -/-_--'''""'''''-

50% ---i-----------+---VBB

Undershoot I __ ___ I Level
Both overshoot and undershoot are functions of many variables: line length,
capacitive/inductive loading, rise time, and so on. Thus, in general, to maintain
undershoot less than 110 m V requires one or more of the following:
Reduction of system rise times;
Reduction of interconnect line lengths;
Use of matched, terminated transmission lines.
57
System Interconnections: Delay, Attenuation, Crosstalk
Reduction of rise time is easily accomplished by going to a slower MECL
family, but this reduction in rise time may limit the use of the high bit rates and
narrow pulse widths necessary for system performance goals. Interconnection line
lengths are dictated by the system design and are routinely minimized as a matter of
practice. Impedance matching of the interconnection lines remains then, the one
variable which can be exploited for limiting the undershoot and ringing.
When using the faster varieties of MECL circuits, the type of card-to-card
wiring in the system backplane area should be considered carefully. The initial
decision is between two basic methods of board-to-board interconnect:
l. Controlled impedance, e.g., mother-daughter boards using microstrip lines,
coax, ribbon flex, or twisted pair interconnects;
2. Uncontrolled impedance, e.g., open wire backplane wiring - with possible
wide variations in characteristic impedaI1ce.
With MECL III, method I must be used. The entire system must be in a
transmission line environment. While MECL 10,000 is designed to drive transmission
lines, the slow edge speed allows it and MECL II to operate with the more
economical wire over a ground plane layout. Wire over a ground plane or ground
screen often has a characteristic impedance between 100 and 150 ohms, and can be
series damped or parallel terminated for extended open wire lengths in the
backplane area. Both wirewrapped and soldered wire connections are suitable for
connecting wires to card connectors in MECL II or MECL 10,000 systems.
When designing system interconnections, four parameters must be taken into
consideration:
Propagation delay per unit length of line;
Line attenuation;
Crosstalk;
Reflections due to mismatched impedance characteristics of the line,
connectors, and line terminations.
Propagation delay of a line is important because unequal delays in parallel lines
may cause timing errors. Also, for long lines the total delay time will often seriously
affect system speed. Since the propagation delay of one foot of wire is
approximately equal to the propagation delay of a MECL 10,000 gate, line length
must be minimized when total propagation time is important.
Attenuation is also a parameter of a line. It varies with frequency and is seen as
an increase in impedance for an increase in frequency. The effects of attenuation
first appear as a degradation in edge speed. This is followed by a loss of signal
amplitude for high frequencies on long lines. A rounding of the waveform occurs,
since the higher frequency components required to give sharp square waves are
attenuated more than low frequency components. Within a backplane attentuation
is seldom a problem, but it must be taken into consideration when interconnecting
among panels or cabinets.
Crosstalk is the undesired coupling of a signal on one wire to a nearby wire.
Since a coupled pulse in the direction of undershoot results in a reduction of noise
immunity, precautions should be taken to limit crosstalk. A good ground system and
shielding are the best methods for limiting crosstalk. Differential twisted pair line
interconnections can avoid problems caused by crosstalk by virtue of the common
mode rejection of the receivers used with such an arrangement. Crosstalk is discussed
in more detail under the heading "Parallel Wire Cables" later in this Chapter.
58
Connectors and Cable for use with MECL
Ret1ections due to mismatched lines in system interconnections cause the same
loss of noise immunity as discussed in Chapter 3 for printed circuit boards. The
ability to terminate a line effectively is primarily a function of how constant the
impedance is over the length of the line. Because it has high uniformity, coaxial
cable is easier to terminate than open wire. Yet in many cases, twisted pair cable or
ribbon cable may be purchased with specifications on the impedance of the line.
Connectors
There are very few high frequency edge connectors that do not cause
waveshape distortion when rise times are under I ns. The few that don't are of the
"matched impedance" type in which the on-board strip transmission line t10ws right
into and out of the connector, without encountering a mismatch. Unfortunately,
this form of connector is usually expensive and is often difficult to design with.
The only form of MECL logic that requires the use of matched edge connectors
is the MECL III family. With rising edges slower than 2 ns, the MECL II and MECL
10,000 families may utilize conventional edge connectors. With them, very little
mismatch occurs: typically < 20 mY.
Coaxial cable connectors that have near ideal characteristics over the
bandwidths exhibited by MECL logic exist in a variety of types. The most popular
are the BNC type and the subminiature SMA, 5MB, or SMC types. The smaller
miniature types offer direct microstrip to coaxial interconnects with low voltage
standing wave ratio (YSWR), i.e. minimum reflection.
Coaxial Cable
Coaxial cable offers many advantages for distributing high frequency signals.
The well defined and uniform characteristic impedance of the line permits easy
matching. The ground shield on the cable minimizes crosstalk. Low attenuation at
high frequencies makes good coaxial cable very desirable for handling the fast rise
times associated with MECL signals.
so:
and:
The line bandwidths required for optimum MECL use are:
f
f
where: k 0.37*,
tr = rise time;
0.37
370 MHz for MECL III with a 50 n load;
I x 10-
9
0.37 9 = 106 MHz for MECL 10,000 with a 50 n load.
3.5 x 10-
At MECL frequencies, skin effect is a primary cause of attenuation. Dielectric
losses are insignificant below I GHz for the common dielectric materials-
polyethylene or teflon. Attenuation due to skin effect is proportional to the square
root of frequency and so may be plotted conveniently on log-log paper. Figure 4-2
contains data for three cable rypes tested. Maximum cable lengths recommended
*0. Gene Gabbard, "High Speed Digital Logic for Satellite Communications."
Electro-Technology, April 1969, p. 59.
59
Behavior of Cable and Terminating Resistors
with the various MECL logic families can be derived from these plots as the
following example will show.
For maximum signal reductions of 100 m V in the J and f/J levels (i.e. a logic
swing reduction from 800 mV pip to 600 mV pip) the permissible attenuation
would be:

(0.8\)
Loss (dB) = 20 log -- = 20 log -- = 2.5 dB.
Vo 0.6
For MECL III with RG58/U the loss at 370 MHz is found to be 12 dB/lOO' from
Figure 4-2. Thus, with the 100 m V restriction:
Max Length = 100 ft . [2.5 dB) = 20.8 ft.
\ 12 dB
Figure 4-3 shows curves giving maximum line length as a function of operating
frequency for the same three cable types used for Figure 4-2. Each curve assumes
2.5 dB permissible loss. It should be noted that a high bandwidth line is necessary to
preserve fast signal edges, regardless of the bit rate of the system.
Figure 4-3 and the preceding calculations assume the coaxial line is properly
terminated with a resistive load equal to the characteristic impedance of the line.
The reactive component of the termination is of increasing importance at the high
MECL III frequencies. At such frequencies, reactive elements can change the
terminating impedance, thus causing reflections on the line. In addition, the
effective inductance or capacitance would distort the output waveform, causing
additional reflection down the line.
Standard carbon resistors were carefully measured at high frequencies to
determine their reactive components. Results are listed in Figure 4-4. The effective
circuit is a resistor with an inductor in series. Carbon resistors display more inductive
ATTENUATION
dB/100Ft
42: Coaxial Cable Attenuation versus Frequency
10
9.0
8.0
7. 0
6. 0
5.0
2. 0 '---""-'-.........
40 60 80 100 200 400 600 800 1000
FREQUENCY (MHz)
60
Cable Reflection
4-3: Coaxial Cable Length versus Operating Frequency:
LENGTH
(Ft)
TEST CONDITION
1/2 W, 51 ohms, 500 MHz
1/2 W, 51 ohms, 300 MHz
1/4 W, 51 ohms, 500 MHz
1/4 W, 51 ohms, 300 MHz
1/8 W, 51 ohms, 500 MHz
1/8 W, 51 ohms, 300 MHz
200
100
80
60
Constant 2.5 dB Loss Curves
4 0 ~ 4 ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ 4 d ~ ~ ~ ~ ~
4 . 0 ~ ~ ~ ~ ~ ~ __ ~ ~ ~ ~ ~ ~ ~ __ ~ __
30 50 70 100 200 400 700 1000
3000
FREQUENCY (MHz)
Z = R + jX
Z-51.8+j15.5
Z = 51.4 + j5.6
Z = 48.8 + j6.1
Z = 49.4 + jO.29
Z = 51.5 + j6.7
Z=51.7+j1.6
4-4: Impedance Characteristics of Carbon Resistors
Measured on a GR Admittance Bridge
reactance as the resistor values become smaller, and display more capacitive
reactance as the values get larger. However, 75 ohm resistors are normally close to
being purely resistive.
The reflection at 300 MHz for a 50 ohm line using a l/2-watt 51 ohm carbon
resistor can be calculated:
where: ZL
so:
p
load impedance,
line impedance;
_ 51.4 + j5.6 - 50
p - 51.4 + j 5.6 + 50 .
Calculations yield:
p = 0.055 /72.8
61
C\
N
-, -0.8 Vdc
L-l.6VdC
4-5: Line Driver Test Circuit
MC1026
Clock NOR
() r to
95ft (RG58/U)
PULSE
GENERATOR
Input
-5.2 Vdc
1
Zo = 50 n
Output A
82

I
Output B
Tr igger
To Scope
Pulse Generator: E-H Model 122 or equivalent.
4-6: Line Driver Test Circuit Input and Outputs, Observed Via High Impedance 'Scope Probes.
(a) OR Output Waveforms
Input
Output .. ..... . ...... 'I
A p--.,I "71 I
Output I J I' 't IA 'j " ' ""k:: I J I'
B
Vertical Scale = 500 mV Icm
Horizontal Scale = 5.0 ns/cm
Input
Output
A
(b) NOR Output Waveforms
Output k ;j,;....."'k 1,:1-4-. I ,I
B
Vertical Scale = 500 mV Icm
Horizontal Scale = 5.0 ns/cm
4-7: Typical Switching Times
'-'n
tpd++
tpd- -
tr
tf
NOR .'
tpd+-
tpd-+
tr
tf
tr
tf
(50% to 50%)
(50% to 50%)
(10%t090%)
(90% to 10%)
(50% to 50%)
(50% to 50%)
(10% to 90%)
(90% to 10%)
(10%t080%)
(10% to 90%)
(80% to 10%)
., '
k
2.6
2.7
2.6
2.0
2.6
2.5
2.4
2.7
3.8
5.4

.....

g


D
CJ-

Using Coaxial Cable
As a result, (O.OSS) (800 mV logic swing) = 44 mY, is reflected back down
the transmission line. Clearly, this is much less than the 300 mV maximum
overshoot recommended for safe MECL usage. With a slow repetition rate in
relation to the propagation delay of the line (time per pulse >3TO) the reflection
appears as a small overshoot at the receiving end of the line. In high frequency
operation the reflection may subtract from the transmitted signal. The amount
would depend on the exact length of the line and the propagation velocity of the
line. Subtracting signals appear to reduce the signal on the line, as if either the signal
were attenuated, or as if the driving gate were bandwidth limited.
Standard carbon 1/8 watt resistors have been found to have good high
frequency characteristics when used with MECL III. Either 1/8 or 1/4 watt resistors
work well with MECL 10,000. When using precision wire wound or film resistors,
care should be taken to determine the high frequency properties of these devices.
Most wire wound and some film resistors become very inductive at high frequencies.
The fanout at the end of a coaxial line should also be limited at high
frequencies because of reactive loading. At 300 MHz the fanout should be limited to
four. The terminating resistor leads and circuit leads should be kept short. In many
cases it is desirable to restrict long interconnecting cables to a fanout of one to
minimize reflections and therefore to maintain a high degree of noise immunity.
The propagation velocity is very high in coaxial cable. Computing the
propagation delay as:
tpd = 1.017 .;e; ,
the delay for solid teflon and polyethylene insulated cables is I.S4 ns per foot
(dielectric constant, e
r
~ 2 . 3 . This compares with 2.2 ns per foot for stripline as
calculated in Chapter 3. For maximum propagation velocity, coaxial cables with
styrofoam or polystyrene beads in air dielectric may be used. However, many of
these cables have high characteristic impedances and are slowed by capacitive
loading. Nonetheless, coaxial cable definitely should be used when sending high
repetition rate MECL signals over long lengths.
Illustrated in Figure 4-S is a circuit used to test the performance of coaxial
cable driven by an MC1026 clock driver. Figures 4-6 (a) and (b) show the waveforms
of the circuit with 9S feet of RGS8/U connecting cable. Output B in each figure
clearly shows the waveform for a skin-effect limited line. Skin effect causes the
waveform to rise sharply for the first SO% of the swing, then taper off during the
remaining portion of the edge. Calculations show that the 10 to 90% waveform
rise time is 30 times greater than the 0 to SO% rise time when the cable is skin effect
limited. The output amplitude of the cable is at least 200 mV pip less than the
input, as would be expected from Figure 4-3.
Figure 4-7 presents the test results for the circuit in Figure 4-S. Notice that the
numerical data show that at the output of the line, the time from 10 to 80% is much
less than 10 to 90% - because of the coaxial skin effect. When operating within the
limits discussed previously in this chapter, MECL signals are transmitted over coaxial
lines with minimum distortion.
63
Twisted Pair Line: Differential Use
Differential Twisted Pair Lines and Receivers
Twisted pair line, differentially driven into an MECL line receiver (Figure 4-8),
provides maximum noise immunity. This is because any noise coupled into a twisted
pair line generally appears equally on both wires (common mode). Because the
receiver responds only to the differences in voltage between the lines, crosstalk noise
is ignored, since it is picked up equally by each of the two lines of the pair.
This holds true up to the common mode noise rejection limit of the receiver. Quad
line receivers, such as the MC 1 020 and MC 1692, have + 1 and - 1. 8 volt common
mode rejection limits before the receiver's output approaches MECL input threshold
levels. Common mode rejection can be improved to a - 2.5 volt limit either by using
a Schmitt trigger circuit (e.g. MC1035 or MC1065), or by using MECL 10,000 line
receivers (e.g. MC 10 115 or MC I 0 116).
With devices such as these four, the constant current source employed in the
emitter node of the differential pair allows the increase in common mode rejection.
This improvement is useful when signals are sent from circuits other than MECL.
The MC 1650 AID Campara tor is also used as a special purpose line receiver and
offers 2.5 volts common mode rejection. However the standard line receivers have
more than adequate common mode noise rejection to handle any crosstalk between
MECL signal lines. If higher voltage signal lines are run in parallel with MECL lines,
shielded twisted pair lines may be used to reduce crosstalk further.
For low frequency operation, line length is limited by the dc resistance of the
wire used and the voltage gain of the MC 1 020 line receiver. In order to determine
line length allowed it is first necessary to examine the required signal at the end of
the line, and the amplification possible with the receiver. The typical differential volt-
age gain of the MC1020 circuit (Figure 4-9) may be calculated ...
Assume Q2 is on, Q3 is off. Then:
RC is known, and:
qO'I
E
gm =
4KT
where: - IE
-VEE -(VBB -- base/emitter dropQ2)
RE
and
= -5.2 + 1.175 + 0.7 = 2.77 rna
1.18 x 10
3
_13_
13 + 1
K Boltzmann constant = 1.38 x 10- 23,
T temperature degrees Kelvin 300,
q = charge of electron = 1.602 x 10- 19.
64
Line Receiver Gain
4-8: Twisted-Pair Line Driver and Receiver
-5.2 Vdc
-5.2 Vdc -5.2 Vdc
4-9: 1/4 MC1020 Schematic
r ~ ~ 0 V C C
Out
'-----+-----..... ----+-__ -0 vEE
In 1 In 2
Proceeding with the numerical calculations gives gm = 2.7 10-2, and
voltage gain = 8.3 V/V.
Allowing for 20% resistor tolerances and temperature variations, the circuit
gives a gain of at least7 V IV. To obtain MECL level outputs the minimum required
input signal is 800 m V-=-7 = 114 m V. Referring to Figure 4-8, the drop across the
input terminating resistor, RT, is:
net voltage fraction across
RT
1 logic level). (RT )
Rl + RT + RLine
(5.2 - 0.8) (100)
560 + 100 + RLine
65
Attenuation in Twisted Pair Line
Setting VT = 114 m V minimum, the maximum line resistance is found to be
3,200 ohms. The resistance of #24 A WG wire averages 26 ohms per 1000 feet, giving
a theoretical permissible length of over 10 miles of wire. Clearly, the ac attenuation
of the line will be the realistic limi ting factor.
Figure 4-10 contains data for the attenuation of a 50 foot twisted pair line
using the circuit of Figure 4-8. Although measurements were carried out to
200 MHz, use of the MCI 020 is limited to around 70 MHz because of the bandwidth
of its cirucit. Driving and receiving the twisted pair line with the MECL III circuits as
shown in Figure 4-1 1 yields the results in Figure 4-12. Comparison of the
410: Attenuation of 50 Ft Twisted Pair Line
with MC1026 Line Driver
FREQUENCY
(MHz)
50
75
100
125
150
175
200
VOLTAGE AT RT
(mV)
320
320
280
240
210
160
140
4-11: MECL III Twisted Pair Line Driver and Receiver
-20 Vdc
51
Twisted Pair Line
MC1660
51
-2.0 Vdc
4-12: Attenuation in a 50 Ft Twisted Pair Line
with a MECL III Driver
FREQUENCY VOLTAGE AT RT
(MHz) (mV)
50 380
75 340
100 320
125 280
150 250
175 180
200 160
225 150
250 140
275 120
300 100
66
413: Attenuation in a 10 Ft Twisted Pair Line
with a MECL III Driver
FREQUENCY VOLTAGE AT RT
(MHz) (mV)
50 440
75 420
100 400
125 390
150 380
175 360
200 300
225 270
250 250
275 240
300 230
Discussion: Long and Short Twisted Pairs
attenuation data in Figures 4-10 and 4-12 shows that the twisted pair line is
bandwidth limited rather than limited by the MECL drivers. Reducing the line
length to 10 feet of course results in less attenuation (cf Figure 4-13).
For very long lines the MECL II MC 1 020 line receiver will operate faster than
the MECL III MC 1692 receiver. The reason is that the voltage gain of the MC 1020 is
slightly higher than that of the MC1692. Driving a 1000 foot twisted pair line, the
MC1020 would operate at 6.5 MHz with 600 mW output, while the MC1692 would
be limited to 6.2 MHz for the same output.
Despite the gain difference, as lines become shorter < 1 000 ft.), the MC 1692
will switch faster, since the MC1020 displays bandwidth limitations when used on
short lines. Figure 4-14 shows the behavior of the circuit of Figure 4-8 when
driving a 50-foot line at 50 MHz. The 4 ns rise and fall times of the receiver can be
seen at the output of the MC 1 020 receiver. The same line driven by the MECL III
circuit of Figure 4-11 yields the waveforms of Figure 4-15. The faster switching
4-14: 50 Ft Twisted Pair Line at 50 MHz with MECL II
Horizontal Scale = 5.0 ns/cm (Both Traces)
Vertical Scale = 500 mV/em (Both Traces)
I lijJut
To
MC1026
Output
Of
MC1020
4-15: 50 Ft Twisted Pair Line at 50 MHz with MECL III
Input
To
MC1660
Output
Of
MC1692
Horizontal Seale = 5.0 ns/em (Both Traces)
Vertical Scale = 500 mV Icm (Both Traces)
67
Party Line Operation with Twisted Pair
times of MECL III are shown on the output of the MC 1692. For both circuits, the
propagation down the line is 70 ns for 50 feet of line. At best, overall circuit
performance is improved only from 76 to 72 ns by using the faster MECL III.
However, the MECL III circuits do offer the big advantage of high rate capa-
bilities on lines shorter than 10 feet. The result of driving twisted pair lines at high
speed is shown by the waveforms in Figure 4-16. The traces were obtained for the
circuit of Figure 4-11 driving a 10-foot line at 350 MHz. Both the attenuation of the
line and the bandwidth of the MC1692 limit the output signal to about 650 mY,
which is indeed still a useful MECL signal.
Party line operation over a single twisted pair line with MECL receivers may be
used for saving space, for reducing connections and wiring, as well as to benefit from
the party line two-way scheme. Figure 4-17 shows a method for using MECL in a
data bus circuit. All driving gates are operating in a Wired-OR configuration
VEE 600
-5.2 Vdc
VSS
RT
-1.29 Vdc
416: 10 Ft Twisted Pair Line Operated at 350 MHz with MECL III
Input
To
MC1660
Output
~ ~ ~ ~ ~ ~ ~ ~ ~ Of
Horizontal Scale ~ 1.0 ns/cm (Soth Traces)
Vertical Scale ~ 500 mV/cm (Soth Traces)
MC1692
4-17: Party Line Operation with MECL Line Drivers and Receivers
Drivers
600
RT
~ v ~
Receivers
68
VEE
-5.2Vdc
VSS
-1.29 Vdc
System Advantages of Twisted Pair Lines
requiring that all drive gate outputs be held low when not sending data. (VBB is
available from the MCI035, MCIOI15, MC10116, and MC1692L and may be
buffered as shown in Figure 4-18 to handle the necessary termination current).
All receivers operate in parallel. When driving long lines party-line style, it
should be remembered that the receivers are operating single-ended. Consequently,
the voltage gain is approximately one-half that for differential operation. Line
attenuation must be computed accordingly. However, the advantages of crosstalk
immunity deriving from common mode rejection of the differential line receiver
remain. When stubbing off the data bus, the rules shown in Figures 3-13,3-14, and
3-15 apply.
418: VBB Generator
Twisted pair lines and line receivers offer several system advantages, when
operating MECL circuits under adverse conditions. The following example is used to
illustrate some of these advantages.
MECL II is specified to have a VaH min of -0.850 volts and a VIH min of
-1.025 volts for a noise immunity of 175m V, with both circuits at 25
0
C ambient
temperature. However, if the driving gate is operating at OOC, VaH min is reduced
to -0.895 volts; if the driven gate is operating at 75
0
C the VIR min is increased
to -0.950 volts; so the resultant noise immunity is reduced to 55 mY. To compound
the worst case conditions, consider temperature and power supply changes also: if
the driving gate is operated at a -5.46 volt supply (+5%) and OOC temperature, the
VaH min is -0.891 volts. With the driven gate operating at a -4.94 volt supply volt-
age (-5%) and 75
0
C temperature, VIH min is -0.920, giving a worst case noise
immunity of 29 mY. Testing has shown that under these extreme conditions, worst
case noise immunity in system usage is about 70 m V and circuits operating at typical
levels have about 170 mV noise immunity.
However, this somewhat unrealistic example does illustrate the reduction in
noise immunity under very adverse conditions. Any noise or voltage drop on the
system ground would add to the loss of noise immunity. The example assumes
MECL II signal levels, but similar conditions hold for MECL 10,000 and MECL III
with the voltages changed to reflect the change in input voltage due to the bias point
level, and the change in output voltage due to loading.
Line drivers operating differentially are not affected by the above conditions.
The reduction of noise immunity to 29 mV (single-ended) is seen by a line driver
69
Ribbon and Triax Cable
operated differentially, only as a small shift in the input levels - well within the
acceptable limits of common mode operation. As a general rule, when operating
MECL circuits at greatly differing temperatures or with differing supply voltage, or
when the circuits are connected by a ground network with noise or voltage drop, the
line driver should be used in the differential mode to retain maximum noise
immunity.
Ribbon Cable
Ribbon cable is often used to interconnect MECL cards and panels. The
advantages of ribbon cable include easy bonding to connectors because of the in-line
arrangement of wires; and flexibility for use with hinged panels which swing open
for servicing.
Two types of ribbon cable have been found to work well with the fast MECL
circuits. One is the flat ribbon composed of several twisted pair lines. This twisted
pair cable is operated differentially and should be received by one of the MECL line
receivers. Single-e11ded operation is also possible by grounding one wire of the
twisted pair. The conventional ribbon cable with side-by-side wires has a defined
characteristic impedance only when every other wire is grounded as shown in Figure
4-19 (a). This cable may be driven either single-ended or differentially as shown in
the examples in Figure 4-19 (b).
419: Ribbon Cable Interconnects
Ground =_S_ig_na_I __ ----,
~ ~ ~
(a) CrossSection
---DlL....-f--l --+-1 [f1>-
Differential
-Di +-= I D-
Single-Ended i ~
(b) Connections
Another type of multiconductor cable is called "triax." As its name suggests,
this is a three-conductor cable with characteristics similar to coaxial cable. Triax has
a flat cross section for flexibility, and may be used with all MEeL families including
MECL III. When using triax type cables, the manufacturer should be consulted for
information about the impedance and attenuation characteristics of a specific cable
type.
70
Matching Line Impedance with Schottky Diodes
Schottky Diode Termination
Under certain board interface conditions, it may be advantageous to use a
termination technique employing Schottky diodes. Several advantages are gained by
the use of diode terminations:
No matched impedance striplines are required;
No line matching termination resistors are required;
All signal overshoot is effectively clamped to the 1 or (/J logic level;
All external noise in excess of 1 or (/J logic levels is clamped at the receiving
gate or load;
The total cost of layout, even though diodes are more expensive than
resistors, may be less because no precise transmission line environment is
necessary;
If ringing is a problem on a drive line during system checkout, diode
termination can be used to improve the waveform;
Where line impedances are not well defined, as in breadboarding or
prototype construction of systems using MECL, use of diode terminations
is convenient and saves time.
The forward conduction characteristic of the Schottky barrier diode is used to
match the line impedance of the signal path. For instance, if a 90 ohm line is used,
the diode impedance equals 90 ohms at a forward voltage of 0.45 volts (0.45 V from
5 rnA
Figure 4-20). Therefore, the line would be terminated with only a small overshoot.
The variable conduction curve of the diode permits terminating line impedances
from 150 n to 50 n.
4-20: Schottky Diode MBD101 Forward Transfer Characteristic
IF. FORWARD
CURRENT
(mA)
0.1 '"'-_'----'-oI._.L1....;...;;.;.....L._-'--'-..;...;;.;..l-_'-----'_-'""_--'
0.2 0.3 0.4 0.5 0.6 0.7
VF, FORWARD VOLTAGE (VOLTS)
71
Using Diode Terminations
In use (cf Figure 4-21), one side of the parallel diode network is biased at the
MECL threshold VBB (-1.29 volts for MECL 10,000 and MECL III). The VBB source
can be either a separate supply or a gate that supplies the required sink and source
Data In
4-21: Diode Termination
MECL Driver
Gate
I nterconnect Line
Rp
MBD101
MECL Receiver
Gate
Data Out
current (cf Figure 4-18). These current requirements can be determined from the
graph in Figure 4-20 as follows:
V D 1 (1 level diode drop)
V
BB
- logic 1 level
= -1.29 -(-0.90) = -0.39 V.
From the graph in Figure 4-20, -0.39 V yields a diode current of approximately
-1.0 rnA (lBBI), and:
V D0 (0 level diode drop) = VBB - logic 0 level
=-1.29-0.70)=0.41 V,
also indicating a current of 1.0 rnA (lBB0).
Thus at the receiving end of the line the power consumed would only be
0.4 mW. The driving device must have an emitter-follower pulldown resistor, Rp, to
provide a current path to VEE and to establish a well-defined output level.
Consider a case when this resistor is 600 ohms, as for MECL 10,000. The power
consumed in this resistor would be about 25 mW. If the resistor were 100 ohms to a
VTT of - 2.0 volts, then the power consumed would only be 7 mWaverage.
72
Improvements Due to the Diodes
A disadvantage of the diode termination scheme is that as many as three
voltages might be required: VEE (-5.2 V), VTT (-2.0 V), and VBB (-1.3 V).
4-22: Circuit #1 - Reduction of Line Ringing by Use of Terminating Diodes
PULSE
GENERATOR
tr = 1.0 ns
50
Without Diodes
MC1026
1
"----24" -----
>--+ __ #26 Wire over a __ ..... __ -I
ground plane.
MBD10l
600
-5.2 Vdc
Vertical Scale = 500 mVfcm
Horizontal Scale = 20 nsfcm
Circuit Rise Time with Diodes
Vertical Scale = 500 mV fem
Horizontal Scale = 2.0 nsfem
73
With Diodes
B
600
-5.2 Vdc
PULSE
GENERATOR
tr 1.0 ns
=
Line Ringing Suppressed by Diodes
4-23: Circuit #2 - Stub, off Diode Terminated Line

#26 wire ground plane "I
600
1
6"
-5.2 Vdc
1
MC1026
Without Diodes on Stub
Vertical Scale 500 mV/cm
Horizontal Scale 20 ns/cm
With Diodes on Stub
Vertical Scale 500 mV/cm
Horizontal Scale 20 no/cm
74
MC1026
MBD101
rO.1 J.l.F
=
Subnanosecond Diode Performance
Offsetting this are the elimination of the transmission line requirement, and the
economical average termination power: (0.4 + 7.0) :::: 7.4 mW.
Figures 4-22 through 4-24 illustrate the performance of the Schottky diodes
(MBD 10 1) and show their unique ability to suppress severe ringing. Both circuit #1
(Figure 4-22) and circuit #2 (Figure 4-23) were evaluated with and without diode
terminations. The 'scope traces show that ringing is reduced to less than 100 mY,
while system rise time remains under 2.0 ns. Circuit #2 is a typical example of loads
being stubbed off along a clock distribution line to provide clocking information to
other parts of a system.
Even when dealing with subnanosecond risetimes ~ 4 0 0 pS), Schottky diodes
perform most satisfactorily - as shown by the waveforms derived from circuit #3 in
Figure 4-24. The conclusion is that even for card-to-card or backplane interconnects,
MECL III logic could be distributed with only a small amount of waveform
degradation when diodes are used.
424: Circuit # 3 - Sub nanosecond Performance of Diode Terminated Line
PULSE
GENERATOR
EH 129
tr ~ 400 ps
=
Line I nput/Output Waveforms
Vertical Scale ~ 500 mV fcm
Horizontal Scale ~ 20 nsfcm
50
MBD101
75
Waveform RiseTimes
Vertical Scale ~ 500 mV fem
Horizontal Scale ~ 1.0 nsfem
Oosstalk Between Parallel Wires in Cables
Parallel Wire Cables
Multiple conductor cables as purchased, or as constructed by lacing inter-
connection wires together, are not normally used with MECL because of crosstalk.
Such crosstalk is due to capacitive and inductive coupling of signals among parallel
lines as symbolized in Figure 4-25.
(a)
(b)
(c)
(d)
(e)
4-25: Crosstalk Coupling in Parallel Lines
IC + IL
IC-IL
-U
- -
A B
RT
em
Lm t
+---------0 VT
I
RT
C
I
D
I
a-x

4-26: Calculated Cross Talk

0.8 V
Pulse On Line CD Beginning From C
i
I--i--T D-----I
I
r-----:--T D + tr I
I
I
-0.24 V
Forward Crosstalk at B
-0.195 V I
Backward Crosstalk at A
1--------2 T D --,..--------1, 2 T D + tr
I
1----------L--+--
3T
D
-----------1
I -0.175 V
I
I
3 TD + tr
I
TD +tr H
_ ...... ______ ---'v !
Signal at D
t 0 TIME --
76
Reflected Backward
Crosstalk at B
Forward and Backward Oosstalk
When a pulse propagating down line CD reaches any arbitrary point X, the
signal is capacitively coupled into line AB. The coupled voltage on AB causes current
(Ie) to flow from the point of coupling to both ends of the line.
Current in the direction of A is called "backward crosstalk" and that toward
B is called "forward crosstalk." Coincident with capacitive coupling, the mutual
inductance of the parallel lines also couples current (I U into line AB in the direction
of backward crosstalk.
The total forward crosstalk is IC - IL at point B. Since the parallel line
coupling is primarily inductive, current flows from point B causing a negative pulse
at that point (cf Figure 4-26). IC and IL are proportional in magnitude to the rate of
change of the signal propagating from C (driving function). Coupling occurs only
during the rise and fall times of the pulse as it propagates along CD.
Since the forward crosstalk propagates along AB at the same rate as the signal
on CD, the result is a pulse at point B lasting for the duration of the rise time of the
driving function. The amplitude of the resulting pulse is a function of the difference
between inductive and capacitive coupling. Normally, the reflection of the backward
crosstalk hides the small pulse at B.
Backward crosstalk current is IC + IL and is a function of line length and
velocity of propagation of the line. Backward crosstalk current starts at point A
simultaneously with the signal at C. The coupling continues for the duration of the
signal (TD) on line CD; at time TD the driving function is at point D, and also
appears coupled to the other line at B. The backward crosstalk then requires another
TD to reach point A. Therefore the duration of backward crosstalk is:
The output impedance of the gate at point A (RS) is low- typically 5 ohms
compared with the line (Zo = 75 to 150 ohms), so the backward crosstalk is
reflected toward point B in proportion to the reflection coefficient.
Since the reflection coefficient for current, PI, is:
if Zo 100 ohms and RS 5 ohms, then:
PI = +0.905 .
The reflected backward crosstalk reaches point B at the same time the driven signal
reaches point D, and is 2 TD in duration and about 0.9 (lC + IL) in current
amplitude. The positive reflection coefficient shows that the reflected current has
the same polarity as the backward crosstalk. This reflection results in the pulse at
point B (Figure 4-25) as shown in Figure 4-26d.
A similar analysis shows that if the gates at A and B were reversed so that the
receiving gate and terminating resistor were at point A, the results would be similar.
A positive crosstalk pulse would begin simultaneously with the driven signal at
point C and have a duration of 2 TD. The forward crosstalk would be reflected from
the gate at point B and would not appear at point A until 2 TD. This reflected signal
is normally not seen as it occurs at the trailing edge of the forward crosstalk.
77
Calculating Crosstalk Amptitude
Crosstalk amplitude V (X,t) may be calculated with the following equation (cf
reference 11, Chapter 7):
V (X, t) = KfXi [V. (t- TD X)l +
dt 111 Q J
[ V. (t TD X) V (t 2 TD + TD X)]
111 Q In Q'
where:
K
f
(forward crosstalk constant) =_ J(Lm C z)
2 Zm 0'
o
Kb (backward crosstalk constant)
= _Q- C Z
(
L )
4TD Zo m 0'
Also note that:
mutual line inductance per unit length,
mutual line capacitance per unit length,
characteristic line impedance,
line length = 10 f1. for the following example,
arbitrary point along line,
arbitrary time,
total one-way line delay.
Co intrinsic line capacitance/unit length,
Lo = intrinsic line inductance/unit length.
Using the measured values Co 1 pF lin, Lo 20 nR/in, C
m
0.446 pF lin,
and Lm 10.3 nR/in for the cable under test, crosstalk can be calculated. The
calculations can then be compared with test data on the cable.
First:
Zo

Co
141 ohms,
and
tpd

0.14 ns/in;
so:
TD
16.8 ns.
Substituting values into the appropriate equations above gives:
and:
Kf = ns/ft,
Kb = 0.244.
78
Forward Oosstalk Calculo.tion
Proceeding now with the calculation of the forward crosstalk,V f,in the line:
Vf(X,t) = [Yin (t - '
dt Q
where Yin (t) may be represented by:
Yin (t) = (fo (t) vet)) - (fo (t - t
r
) V (t - t
r
)
Here V (t), the step function, has the values:
V (t) = 0, for all t < 0,
V (t) = I, for all t > 0.
In this equation fo (t) describes the rising portion of the input pulse (Figure
4-26(a, and since the pulse rises with a slope:
then the first term,
The second term,

Vo
t '
r
Vo
t , for t > 0,
Tr
= 0, for t < 0.
Note that the second term of Yin (t) is zero until t = t
r
. The V function is being
used to "turn on" the first term at t = 0, and bring in the second term only for
t > tr:
Note that after time t
r
, the function Yin (t) remains at a value V 0, for all values of t.
X
Substituting Yin (t) into the equation for V f (X,t), substituting (t - TD"Q) for
tin Yin (t), and evaluating at the end of the line (X = Q), gives:
Vf(Q, t) = Kf (t - T
D
) V (t - T
D
)] -
d
dt
79
Backward Crosstalk Calculation
This gives a pulse at point B equal in duration to the driven line rise time, t
r
, and
starting at time TD as shown in Figure 4-26(b). The amplitude of this pulse is, for a
rise time tr = 2 ns, at point B (t = TD):
(-0.06) (10) (0.8)
2 = -0.24 volt.
The backward crosstalk is calculated as follows:
Vb (X,t) = Kb [Vin - - Vin 2Tn +
For the same ramp function, 1'0 (t)
Taking X = 0 for point A:
Vo
Tr
t, as used with the forward crosstalk.
Vb (tl = Kb [IV (tl - (t - ttl U (t - t,l -
This gives a pulse at point A starting simultaneously with the driving signal at
point C. The leading edge of the backward crosstalk pulse (Figure 4-23(c)) is a ramp
until time t
r
. The pulse levels off until time 2 TD then slopes to the starting point at
time 2 TD + t
r
. The amplitude of this pulse is:
Vb (A) = Kb Vo = 0.244 (0.8) = 0.195 volt.
These calculations assume the 10 foot line is terminated in its characteristic
impedance at points A and B. However, since the gate output at point A of Figure
4-25 is a low impedance, only a small voltage pulse is seen at the MECL gate. As
previously discussed, 90% of the backward crosstalk is reflected to point B. The
amount of crosstalk at point B due to reflected backward crosstalk is calculated to
be: (- 0.9) (0.195) = -0.175 volts as shown in Figure 4-26(d).
Figure 4-27 lists measured crosstalk in a ten foot multiconductor cable for the
test circuit of Figure 4-28. Using all wires in the cable for signal lines causes a
prohibitive amount of crosstalk, as shown.
Several factors contribute to the discrepancy between calculated and measured
crosstalk. The characteristic impedance of the line is comparatively undefined
80
1 or 0
Measuring Crosstalk
4-27: 10 Ft. Multiple Conductor Cable Crosstalk
CROSSTALK AT B
CONDUCTORS AT C (mV)
CD
1 240
2 360
3 420
1
60
2'
80
3*** 100
'With one wire in cable grounded at both ends
'With two wires in cable grounded at both ends
"' ... "'With three wires in cable grounded at both ends
CD Compare with theoretically calculated example.
4-28: Cross Talk Test Circuit
R
(OHMS)
150
75
50
150
75
50
-2.0 Vdc
A
t ~ 10' -----1
'1
150
MC1026
B
MC1026
Bundled Cable
-2.0 Vdc
MC1026
R
MC1026
,-- c
0
=
50
-2.0 Vdc
50
-2.0 Vdc
because there is not a solid ground reference for the cable. Thus, placement of the
cable with respect to the system ground and other cables affects the characteristic
impedance. Tn addition, capacitive and inductive coupling will vary along the cable
due to the relative location of wires with respect to each other. The one other factor
not allowed for in the calculations is attcntuation in the line which damps out the
higher frequency components of the signal, slowing the rise time of the signal as it
propagates along the line.
The 150 ohm terminating resistor gives an approximate impedance match, to
cut down overshoot and ringing. However, residual mismatch causes reflections to
return along the line. Such reflections interfere with the signal by producing
distortion at the receiving gate input, and so limiting high speed operation of the
cable. Serious distortion occurs when the reflected signal coincides with a following
signal, i.e. when the transmitted frequency equals:
I
Frequency = 2T
D
81
Reduction of Crosstalk
For the 10 foot line example just discussed:
1
f=---
2 (16.8)
29.8 MHz.
Test data coincides with the calculated performance to indicate that serious
distortion occurs around 30 MHz in the 10 foot cable. The previously computed
propagation speed of 1.68 ns/ft also closely agrees with the measured time of
1.65 ns/ft.
Crosstalk is reduced by supplying a ground reference in the cable. In a
multiconductor cable this may be done by grounding approximately the same
number of wires in the cable as there are signal lines. This measure reduces crosstalk
by a factor of 4 (cf Figure 4-27). Figures 4-29 and 4-30 show the crosstalk in the
circuit of Figure 4-28, and compare crosstalk of cables with and without one
grounded wire in the cable.
4-29: Crosstalk in Multiconductor Cable with No Grounded Conductor
Horizontal Scale = 20 nsfcm (both traces)
Crosstalk
Vertical Scale = 200 mVfcm
Driving Signal
Vertical Scale = 500 mVfcm
4-30: Crosstalk in Multiconductor Cable with One Grounded Conductor
Horizontal Scale = 20 nsfem (both traces)
82
Crosstalk
Vertical Scale = 200 mVfcm
Driving Signal
Vertical Scale = 500 mVfcm
Twisted Pair Cables: Single-Ended Use
The amplitude of crosstalk is independent of length for "long lines." Defining a
long line as having a propagation delay greater than 1/2 the input rise time gives a
"long line" length of 0.605 ft. for a 2 ns rise time waveshape:
2
Long Line Length > (tpd) (tr)
2
0.605 ft. (for the cable just discussed).
(1.65) (2)
431: Test Results for an 18 I nch Multiple Conductor Cable: Crosstalk
CROSSTALK AT B
CONDUCTORS AT C (mV)
1 240
2 350
3 400
1* 70
2** 80
3*** 100
'With one wire in the cable grou nded at both ends .
'With two wires in the cable grounded at both ends .
'With three wires in the cable grounded at both ends.
R
(OHMS)
150
75
50
150
75
50
Test results (Figure 4-31) show that crosstalk for an 18 inch multiconductor cable is
approximately equal to that for the 10 foot bundled cable shown in Figure 4-27.
However, since reflections damp-out much faster because of the lesser propagation
delay, the shorter cable is useful to 100 MHz.
Multiple conductor cables of this type (bundled) may be used successfully with
MECL II or MECL 10,000 if one-half the wires are grounded at both ends. However,
the 100 m V of crosstalk present with the grounded lines significantly reduces noise
immunity. The cable is also susceptible to external signals coupling to the entire
cable. These cause additional noise on the line. Thus, this cable should be used only
when cost or manufacturing techniques require it. Other cable types - coaxial,
tri-axial, ribbon, or twisted pair - are recommended wherever possible.
Twisted Pair Cable, Driven Single - Ended
Cables formed of twisted pair lines have a more defined characteristic
impedance than parallel wires. So, twisted pair can be terminated more accurately
at the receiving end, reducing reflections. Further, the speed of operation of
twisted pair is limited by attenuation rather than by any significant reflection
interference. Test results show a 10 foot length of twisted pair cable may be used
up to 70 MHz before attenuation reduces noise immunity by 100 m V. Propagation
delay is the same as for parallel lines - 1.65 ns/ft.
83
Measured Data: Crosstalk Between Twisted Pairs
Crosstalk for the twisted pair cable is comparable to that for the parallel wire
cable operated with half the leads grounded. This is because the higher switching
current (due to lower Zo of twisted pair) offsets the better ground of the twisted
pair. Crosstalk magnitude in a twisted pair cable is listed in Figure 4-32 for the test
circuit of Figure 4-33. If shielded twisted pair cable is used, crosstalk is significantly
reduced (compared to unshielded) as shown in Figure 4-34 and Figure 4-35. Both
ends of the shield as well as the second wire of the pair were grounded in the test
whose results are shown in Figure 4-34.
432: Crosstalk for 10 Ft Multiple Twisted Pair Cable
CONDUCTORS AT C CROSSTALK AT B R
(mV) (OHMS)
1 60 75
2
80 39
3 90 27
Differential operation of twisted pair line offers advantages over the standard
multiconductor cable when sending higher frequency signals. When operated
single-ended (as shown in Figure 4-33), twisted pair is still susceptible to noise
external to the cable. Any noise coupled into the entire cable causes a direct
reduction of noise immunity.
4-33: Twisted Pair Crosstalk Test Circuit
-2.0 Vdc
1----- 10' ----1
75
1 or 0
50
-2.0 Vdc
50
-2.0 Vdc
84
Shielded Twisted Pair
Since one-half the wires in a multiconductor cable should be grounded for low
crosstalk (comparable to the twisted pair), cable density is the same for both - two
wires per signal path. The use of shielded twisted pair significantly reduces crosstalk
and should be used in applications where crosstalk could be a problem. Differential
operation of twisted pair lines is definitely preferred over single-ended twisted pairs
for sending MECL signals between sections of a system.
434: Crosstalk for 10 Ft Multiple Shielded Twisted Pair
CONDUCTORS AT C CROSSTALK AT B R
(mV) (OHMS)
1 30 75
2 30 39
3 40 27
435: Crosstalk in a Multiconductor Shielded Twisted Pair Cable
Horizontal Scale = 20 ns/cm (Both Traces)
85
Crosstalk
Vertical Scale = 100 mV/cm
Driving Signal
Vertical Scale = 500 mV/cm
Co""'''tio",' tw"<ided prin'''' "'''''' b, ""d who. d""""", with MEeL 10,000, Shown", d'mo",
stration board which illustrates some of the capabilities of MECL 10,000 series parts.
Th' ""wit ron"'" of, 'log o,"'h'm ,," , ,"rid, by rono"",d by , 12 Inch m'oro""p lin"
Operating frequency is 83 MHz.
With '" 0""'10'001", the doc"it ill""""" pro"",," .. d'hy, olg, ",,00, ,," .yo,,",ono", t11pflop
performance, as well as series, parallel and non-terminated line connections.
86
Power Distribution
Power distribution is an important factor in system design. The loss of noise
margin due to reduced power supply voltage or noise on the power supply lines
means a reduction in the circuit tolerance to crosstalk and ringing as discussed in
Chapters 3 and 4. Points to consider for overall system operation include total
circuit and termination power, voltage drops on the power buses, and noise induced
on the power distribution lines by the circuits and by external sources.
MECL circuits are designed to interface with each other over a wide power
supply voltage range - at least IO% from the nominal ~ 5 2 volts, - without loss of
noise margin (other than that due to reduced signal swing at low voltage). However,
if two circuits are at different supply voltages or on the same power supply with a
voltage offset between circuits, there will be a predictable loss of noise margin.
Figure 5-1 illustrates supply points for two MECL circuits (A and B). The
MECL circuits are most sensitive to voltage differences between V CCS for the two
circuits. Any voltage drop on this power bus causes a direct loss of noise immunity
and should be avoided. Similarly, any noise on the V CC line not common to both
circuits may subtract from noise immunity. For this reason, VCC is normally
made to be the system ground - usually the most stable reference level in the
system.
The main causes of V CC offsets between circuits are:
inadequate power buses to handle the current;
separate supplies with common negative terminals operating at slightly
different voltages (not recommended for system design);
separate positive grounded supplies with inadequate interconnecting ground
bus bars.
5-1: MECL Power Points
87
Logic Level Variations
5-2: Changes in Output Levels and VBB with VEE
VOLTAGE MECL II MECL 10,000 MECL III
L'lVOH/L'lVEE 0.015 0.016 0.033
L'lVOL/L'lVEE 0.230 0.25 0.27
L'lVBB/L'lVEE
0.115 0.148 0.14
A more common problem is with circuits which have a good ground, but which
operate with different VEE voltages. The loss of noise margin can be calculated from
the changes in VOH, VOL, and V BB as [unctions of supply voltage. Figure 5-2
shows the change in these levels as a function of VEE for the MECL families. The
change in the logic I output level is very small compared to the change in logic 0 as a
function of VEE. VBB is designed to change at one-half the logic 0 rate, to stay at
the center of the logic swing.
The foBowir:.g example illustrates the loss of noise margin due to circuits
operating at largely differing voltages. Worst case MECL 10,000 Series logic levels
are used in the example.
If the driving gate is at -5.46 volts (+5% of nominal), the output levels are:
VOHA min (-0.980)
VOLA max = (-1.630)
(0.016) (5.2) (0.05) = -0.984 volts,
(0.25) (5.2) (0.05) = -1.695 volts.
If the receiving gate is at -4.94 volts (-5% of nominal), the input levIes are:
VlHA min
-1.105 + (0.15) (5.2) (0.05) = -1.066 volts,
VILA max
-1.475 + (0.15)(5.2)(0.05) - -1.436 volts.
Worst case noise margin is therefore:
Logic 1: 1.068 - 0.984
Logic 0: 1.695 - 1.436
0.084 volts,
0.259 volts.
In this example, worst case noise margin in the logic I state was reduced from
125 mV to 84 mV by a 10% power supply difference. Although the logic 0 noise
margin here improved, it would in fact have been reduced if the receiving gate were at
the +5% supply voltage. Since the example assumed worst case voltages, an
additional 100 mV protection from noise could be expected in typical system use.
System Power Calculations
The total power required by MECL circuits consists of several parts: current
switch, bias supply (VBB), output emitter follower transistor, and terminating or
pulldown resistor power. Since the output emitter follower power and resistor power
are dependent on the method of termination for MECL 10,000 and MECL III, they
are not included in the specified circuit power and must be added for total system
power. The data sheet power specifications for MECL II parts are representative of
system requirements, because the internal puBdown resistors are included in the
determination of package power.
88
MECL Circuit Power Requirements
Gate power is calculated as the sum of the powers for each of the three sections
illustrated for the basic MECL 10,000 gate (Figure 1-1). The bias driver, which
furnishes -1.29 volts to the base of Q5, dissipates about 5 mW /gate, as may be seen
from the following:
VEE - V
BB
- diode drop (0.79 V)
IBB = = -0.625 rnA
R
BQ6
so:
PBB1 = IBB VEE = 3.24 mW (shared by 2 gates);
and:
VEE - V
BB
= -0.64 rnA,
IQ6 VEE = 3.32 mW ;
PTOTAL BIAS
P
BB1
= -2- + P
BB2
= 4.94 mW ,
per gate in a gate pair. For a single gate, the bias power is not shared, so the total
power for a single gate would be 6.56 mW.
Current switch power can be calculated in a similar fashion:
VEE - V
BB
- diode drop (0.79 V)
= -3.99 rnA;
RE
PQ5 = I
EQ5
VEE = 20.8 mW .
For one gate, the combined power dissipation would be: 20.8 mW + 6.6 mW
= 27.4 mW. However, the actual power dissipation is less than this on a package
basis, because the gates share a common bias driver, which is coupled through
emitter followers for isolation. A quad gate, for example, has a typical per gate
dissipation of 25 mW. Note that this power is constant over the full speed range of
operation. Transistor base currents were omitted from the above calculations as they
are beta dependent and have little effect on package power.
Typical input power may be computed when using a 50 kQ input pulldown
resistor. Input power for a logic I level (- 0.9 volts) on the input is:
(V EE - logic 1)2
P
inl
= = 0.37 mW .
Rp
89
Output Power
Input power for a logic 0 (- 1.7 volts) on the input is:
(V EE - logic 0 )2
Rp
0.25 mW .
Totaling the input and gate power gives a typical 26.4 mW power per gate for a
dual four-input gate, with two inputs high on each gate.
Output power is a function of the load network. It is usually computed for
circuits operating at a 50% duty cycle by calculating the I and (jJ level output powers
and forming their average.
Figure 5-3 shows output transistor powers and load resistor powers for several
of the popular terminations. This power must be added to gate power when
determining system power. Unused outputs draw no power and may be ignored.
5-3: Typical Output Power
TERMINATING P 07 or 08 P RESISTOR P TOTAL
RESISTOR (mW) (mW) (mW)
150 ohms to -2.0 Vdc 5.0 4.3 9.3
100 ohms to -2.0 Vdc 7.5 6.5 14
75 ohms to -2.0 Vdc 10 8.7 18.7
50 ohms to -2.0 Vdc 15 13 28
2.0 k ohms to VE E 2.5 7.7 10.2
1.0kohmtoVEE 4.9 15.4 20.3
680ohmsto VEE 7.2 22.6 29.8
510ohmsto VEE 9.7 30.2 39.9
270 ohms to VEE 18.3 57.2 75.5
82 ohmsto Vee and 15 140 165
130ohmsto VEE
Calculations for power required with an external 510 Q output pulldown
resistor are:
IR (logic I)
So:
PQ7 or Q8
and:
P5 \0 ohm
VEE - logic I level
Rp (output)
= -8.43 mA .
OR logic I) (logic I level) = (-8.43)(-0.9) = 7.6 mW
(Q7 or Q8, depending upon which is connected to the 510 Q
output pulldown);
OR logic I) (VEE - logic 1 level) = 36.3 mW.
90
Design of the Power Supply
Similar calculations for a logic 0 state give PQ7 or Q8
PSI0 ohm = 24 m W . Averaging 0 and 1 level powers, gives:
PSlOn =: 30.2 mW avg.;
PQ7 or Q8
9.7 mW avg.
11.7 mW and
Low impedance MECL III circuits require more input power because of their
2 kn input resistors. An average of 7.7 mW per used input must be added to the
power for the rest of the circuit:
(VEE - logic I leve1)2
Rin
(V EE - logic 0 level)2
P
in
0 =
=: 9.3 mW
6.1 mW.
MECL II normally uses internal output pulldown resistors. Power in these
resistors is part of the gate specification, so no addition of an output power figure is
needed.
Power Supply Considerations
The loose 1 0% power supply regulation specification for MECL circuits gives
the designer some freedom with operating voltage. If greater than the nominal-S.2
volt supply is used there is an increase in noise margin due to the larger signal swing.
The cost is increased power dissipation. Circuit speed is optimized at - S.2 volts but
there is a negligible loss of performance over a S% voltage range.
Power supply output impedance requirements are dependent on the MECL
family used. MECL II, without transmission lines, has very low switching current
(less than 100 J.1A per input) in the signal line and only a 740 I1A current differential
(in the I.S kn pulldown resistor) between logic I and 0 levels. This low current
difference between logic states, along with complementary outputs on many gates
and flip-flops, makes power supply drain essentially constant- independent of logic
state and speed of operation. Power supply requirements are therefore not stringent.
Simple bridge rectifiers with capacitor filters have been used on small MECL II
systems. Larger systems with increased current drain require regulated supplies.
MECL 10,000, used without transmission lines, requires slightly more switching
current (less than 2.0 mA) because of its smaller pull down resistors (typically
S10 [2) and lower input impedance. Even so, worst case fluctuation in current
requirements is less than 12 percent. In system use the fluctuation would normally
be much less than 12% because of complementary outputs and the low probability
of all circuits being in a logic 1 or 0 state at the same time.
Power supply requirements do become more important for MECL 10,000 and
MECL III when they are used with transmission lines. In particular, a 50 ohm
parallel terminated transmission line sinks 22 rnA with a logic 1 output, and 6 rnA
with a logic 0. The 16 rnA differential between the two states can produce a
91
Distributing Power to MECL Circuits
significant power supply current fluctuation. Such an effect should be considered
when specifying the power supply.
The current fluctuations are by no means insurmountable. Brief current
changes are smoothed by bypass capacitors at the circuits. However longer current
changes could cause noise on the supply lines unless a properly regulated supply is
used. Fortunately, the presence of complementary outputs and the typical 50%
distribution of output logic levels minimize current changes.
High frequency noise and ripple from the power supply should be avoided
because they produce, in effect, differences in voltage levels among sections of a
system, and lead to loss of noise margin. As a rule of thumb, noise can be considered
"high frequency" whenever the mean wave length of the noise on the power lines
is not several times greater than the length of the longest power line. It is recom-
mended that for operation with MECL, high frequency supply noise be held to
under 50 m V.
When multiple power supplies are used, the positive terminals should be
connected together with a large bus and the output voltages maintained as equal as
possible. It is desirable to keep the various supply levels within 50 m V of one
another.
System Power Distribution
When designing the system power distribution network, pnmary areas of
concern are:
I. Maintaining a low impedance ground - without voltage drops;
2. Limiting VEE voltage drops;
3. Designing the supply lines to hinder external noise from coupling into the
system.
The following method is used to calculate voltage drops along a voltage bus which
has distributed loads (cf Figure 5-4).
5-4: Voltage Drops Along a Power Bus
I
vo-----':-t--I: v1-t---l_ + , ~
50
I
5 ___
S
POWER
---...,
lJ
SUPPLY -
(R eference)
Bus 10 = nA
111 = (n-1) A 112 = (n -2) A II (n
1
1 2
-=-
3
r r r
92
Where:
and
Calculating Power Bus Voltage Drop
S == average spacing of cards in inches,
r == resistance per inch of bus (ohms/in),
n = number of cards,
A == average card current load (amps),
So = distance from reference to first card (inches),
the voltage drop to the first card will be:
Vo == I R = (n A) (So r).
Between cards I and 2 the voltage drop is:
Likewise:
So:
Example:
let
V I = (n - I) A (S r).
V2 == (n - 2) A (S r),
V3 etc.
V'
n
n A So r +
n - I
n A So r + A S r Ln.
5 inches,
r = 0.0004 ohms/inch,
The voltage drop to package 10,
S = I inch,
A == 500 mA,
n = 10.
V'lO 10- 2 + 2 X 10- 4 (I + 2 + 3 + 4 + 5 + 6 + 7 + 8 + 9)
19 mY.
This type of calculation should be performed for all voltage distribution
systems, and should also include edge connector voltage drops, etc. These
calculations will indicate the results to be expected for a proposed distribution
system, and the consequences of using a high resistance voltage bus are evident. The
equation may be modified to accommodate other conditions - such as unequal
spacing between cards or variations of loads among cards.
Laminated bus bars have advantages for power distribution to larger systems
because they minimize the effects of induced noise. Noise is reduced by the high
intrinsic capacitance of the laminated bus bars. Since each bus layer is separated by a
dielectric, the bus bar appears overall as a very large capacitor. Bus bar design using a
large width to thickness ratio ensures low self inductance. This type of power bus
system is available with various options from many manufacturers.
93
Power Distribution for a MECL System
For large systems, power distribution should avoid ground loops. Figure 5-5
shows power distribution to a typical large system. The flow of power from the
supplies is via main bus bars directly to the ground plane or ground screen of
individual card racks and cards. This method minimizes supply losses which would
otherwise occur with power supplied through a series string of card racks.
In addition to preventing large voltage drops along the supply lines, the power
distribution system must be designed to ward off external noise interference. All
noisy and high power devices such as relays and motors should use a separate power
supply and ground system. The ground systems are connected at the system ground
point which is normally at the power supply. Relays and solenoids should be diode
suppressed and motor brushes should be filtered. Other standard design practices
should also be used to eliminate these sources of noise.
Power Supply
1
+
VCC
5-5: Power Supply System
~ - -
Power Supply
2
+
VCC
Card Track (Gnd Plane)
I n Number
: of Cards
n
Number
of
Power
Supplies
-----
Electronic
System
Ground
~ ~
Power Supply
n
+
VCC
Card Track (Gnd Plane)
I n Nurnbe
: of Cards
Card (Gnd Plane or Screen) Card
Busbar - - - - - - - - - - - - - - - Busbar
Main
Distribution
Busbar
1 n
+Card Rack Busbar 1
I
InN umber of B usbars
I
I
+ Card Rack Busbar n
+
- Card Rack Busbar 1
I
InN umber of B u sbars
I
I - Card Rack Busbar n
94
Backplane Power Distribution: VCCand VEE
The mechanical sections of a system are commonly connected together with
another ground. The frame connecting the panels to the chassis is used for this
ground if good electrical conduction is made at points of mechanical contact. This
hardware ground is also connected at the common ground point (cf Figure 2-10).
Backplane Power Distribution
For systems using MECL II or MECL 10,000 circuits a common hard wired
backplane is often used. The wires are either soldered or wire wrapped to connectors
and are routed over a ground plane or ground screen. The ground plane is often
formed by a large printed circuit board to which the connectors are mounted, or
else, the ground plane is connected to the frame holding the card connectors. The
metal is left on one side of the board and forms the ground plane for the backplane
wiring. Alternatively, metal can be left on both sides of the board, to conduct both
ground and VEE. Ground plane circuit boards are commonly used over a metal
ground plane as a means of isolating the MECL circuit ground system from the
mechanical system component ground.
When a ground plane is not practical, a ground screen should be constructed on
the backplane. A ground screen is made by connecting bus wires (wire size
compatible with connector) to the edge connectors in a grid pattern prior to signal
wiring, as shown in Figure 5-6. About every sixth pin on the card edge connectors is
5-6: Ground Screen Construction
Edge Connectors
Ground Screen
used as a ground, providing connection points for the ground grid. This
interconnection of ground points forms a grid network of approximately 1 inch
squares over which the signal lines are wired. A characteristic impedance for a wire
over ground screen of about 140 ohms can be expected, depending upon the exact
routing and distance from the ground screen. The capacitance of this type line will
be about 1 to 2 pF per inch, and series inductance will be about 20 nH per inch.
95
Distributing Power On a Logic Ozrd
Point-to-point wiring is normally used instead of routing along channels, to shorten
the interconnecting paths and minimize crosstalk which would occur among parallel
signal paths. The system interconnecting methods of Chapter 4 are used in
backplane wiring over a ground screen.
The faster edges of MECL III require a transmission line environment for
connecting among circuit boards. One method is to use coaxial cable for
interconnections, and matched impedance connectors on the boards. Care must be
taken when stubbing off the cable using a connector "T", because of the short stub
length allowed with MECL III. Normally this is avoided in favor of wiring with a
single output per cable.
MECL III works very well with twisted pair lines if these lines are specified to
have a constant, defined impedance. Differentially driven twisted pair lines with an
MC 1692 line receiver should always be used where there may be significant power
supply voltage drops or noise on the ground system. The board connectors used with
the twisted pair lines should be designed to minimize reflection from the interconnect
point. Standard edge connectors with the terminating resistor and line receiver close
to the point where the line meets the connector (within I inch) normally provide
adequate termination points.
Although coaxial cable and twisted pair line do not require a ground plane in
the backplane for impedance matching, the ground plane must be retained in both
cases for a good circuit ground.
Multilayer backplane wiring (motherboard) is commonly used with MECL III.
Striplines and microstrip line interconnects are designed in the circuit board, along
with ground and VEE voltage planes. Matched impedance connectors are available to
permi t interfacing between cards and the backplane motherboard without line
discontinuity. This technique is normally used when a system design is sufficiently
determined to minimize changes in the backplane wiring.
On - Card Power Distribu tion
Just as with backplane wiring, the method for distributing power on cards
depends on the logic family used. Standard double sided circuit boards with a good
ground may be used with MECL II because of relatively slow edge speeds and very
low switching currents in the signal lines. A good ground is necessary to prevent
voltage drops and noise from reducing circuit noise margin.
Here's an example of a circuit board which would work well with MECL II.
The various techniques can be modified to fit specific system requirements.
The majority of interconnecting wires would be on one side of the board. A
ground bus or modified ground plane with any remaining interconnections would be
on the other side. The - 5.2 V dc line is not as critical as the V CC line and may be
routed as necessary. The ground buses would be made of wide circuit board paths on
the card. The width should be kept as large as possible, with at least 0.15 inch of
width for each 10 packages recommended. A modified ground plane is made by
leaving the metal on one side of the board and etching only as necessary to run the
interconnecting leads for devices on the other side. The layout should be planned so
that such interconnecting paths will not cut off a section of the ground plane from
the ground inputs or isolate a section so that it is connected only with a narrow
metal strip to the rest of the ground planc.
For either method, circuit board grounding is simplified if several pins in the
edge connector are used for ground. A standard 22 pin connector could have four or
five evenly spaced pins on the connector allocated to ground.
96
On-Card MECL Design
Power supply bypass capacitors are used on the circuit boards to handle the
small current transients required by signal lines for charging stray capacitances.
Bypass capacitors also lower the supply impedance on the card, reducing noise on
the VEE line. Typically a I to 10 J-lF capacitor is placed on the board at the power
supply inputs, and a 0.1 to 0.0 I J-lF RF type capacitor is connected between ground
and -5.2 Vdc every four or five packages. RF type capacitors are recommended
because of their low inductance. Because of their nearly constant current
requirements, many MECL II systems are built without using bypass capacitors, and
operate perfectly. However, the use of these capacitors will insure cleaner supply
lines, especially at top circuit operating speeds.
MECL 10,000 systems use both standard double sided and multilayered circuit
boards. However, when using MECL 10,000, the ground plane described previously
for MECL II is recommended. Such a ground plane permits low impedance signal
lines (over the ground plane) which may be terminated for optimum performance.
Also, a ground plane gives the solid ground necessary for suppressing the current
transients arising in parallel terminated lines and eliminates possible high frequency
ground loops. Ideally, the ground plane would fully cover one side of the circuit
board. However with MECL 10,000, ground planes covering greater than 70 percent
of the board surface area give good results.
The VCCI and VCC2 pins of MECL 10,000 and MECL III packages should be
connected directly to the ground plane as closely as possible to the package. V CCI
should equal VCC2 for best operation. If VCCI drops below VCC2 by more than
two tenths of a volt, the output devices could saturate and cause additional
propagation delays.
When designing the VEE line, care should be taken to prevent excessive voltage
drop in the line. Figure 5-7 shows the bus resistance per foot for microstrip lines.
This should be taken into consideration when designing large cards with high current
requirements. Use of bypass capacitors with MECL 10,000 is strongly recommended
to handle the current transients occurring when parallel terminated transmission
lines are used. A 1.0 to 10 J-lF capacitor at the power supply inputs and 0.1 to
0.01 J-lF capacitors every four or five packages along the board give a low impedance
supply.
57: Bus. Resistance Per Foot for Microstrip Lines
RESISTANCE/Ft
(OHMS)
20 30 40 50 60 80 10'0
WIRE WIDTH (MILS)
97
Multilayer Boards for MECL III
Multilayer circuit boards are ordinarily used with MECL III. For small systems
(with few packages) or small test circuits, a double sided board with a good ground
plane may be used. With larger systems or systems operating above 200 MHz,
multilayer boards are recommended for two reasons: to eliminate ground loops
caused by the use of what would normally be ground plane areas as signal paths, and
to provide uniform transmission line characteristics. Multilayer boards can be used
for other advantages in MECL III systems - possible higher packing density and
shorter interconnecting lines.
The layout of a typical small MECL III multilayer board is shown in Figure 5-8.
When using multilayer boards, the correct use of ground and voltage planes leads to
specific benefits and eliminates .serious problems. For instance, when adjacent signal
lines are switching, signal line crosstalk may occur. Crosstalk can be reduced by
using a voltage plane to separate successive layers of signal lines. Ground lines,
between parallel lines on a signal plane, connected to the ground plane via
plated-through holes, give additional protection against noise coupling.
58: Typical MECL III Multilayer Board Layout
Board Layer Orientation
I11111111I1111111
Board Layout
98
The Terminating Voltage, VTT
If two successive layers are used for signal interconnects, the use of an
orthogonal system is suggested, i.e. interconnects running on one layer are
perpendicular to those on the other. This will facilitate layout and reduce crosstalk
problems. An associated ground plane can follow below to give ground reference to
the two layers of signal lines. VEE may be a separate plane or may be included with
one of the signal planes.
The multilayer board ground planes provide a non-inductive, capacitive
decoupling function. However, the thickness of the dielectric separating the voltage
planes may be too great to provide sufficient inherent low frequency decoupling. In
such a case, discrete capacitors are needed. These should be 0.1 to 0.01 J.LF in value,
and are to be placed every three to five packages, to minimize voltage transients
between the voltage planes (i.e. ground and VEE).
V
TT
Termination Voltage Distribution
The generation of a separate -2 Vdc termination voltage, common to all termi-
nation resistors, may be advantageous in many system designs. This is an alternate
approach to the Thevenin equivalent resistor termination for each parallel termination,
in which two resistors are needed.
The decision to use a separate -2 volt supply will depend on the system size. If
it is feasible to provide a separate -2 volt supply, then lower termination component
count per termination (one less resistor) and a power saving (up to a factor of 4) will
be achieved. Since the VTT supply is only used to sink current through the termination
resistors, current regulation and ripple are not critical. A good rule to follow is to use
the same design practices for VTT as used for the negative supply, VEE. However,
if the system is small, cost may weigh against the use of a separate -2 volt supply. Also,
the short circuit interconnects of many small systems use only a single pulldown
resistor, and this reduces the need for a separate VTT supply.
99
Complex MECL logic functions are exemplified by this microphotograph of the MClO181 Arithmetic/Logic Unit
die. The array is a member of the MECL 10,000 logic family - whose low power gate is permitting a higher level
of sophistication in the use of emitter coupled logic. (Some applications for this integrated circuit are discussed
in Chapter 8).
100
Thermal
Considerations
The electrical power dissipated in any integrated circuit forms a heat source in
the package. This heat source increases the temperature of the circuit die relative to
some reference point (normally 2S
o
C ambient) in an amount which depends
upon the net thermal resistance between the heat source and the reference point.
Thermal resistance, a, is the difference between the temperature of the junction and
the temperature of the reference point, per unit power dissipation. Thermal
resistance is the primary figure of merit for the power handling capability of any
integrated circuit package. Thermal resistance from "junction to case", a IC, and/or
the thermal resistance from "junction to ambient", a lA, are the thermal parameters
most often specified for integrated circuit packages.
The junction temperature, TJ, for a given junction-to-ambient thermal
resistance a lA, power dissipation PD, and ambient temperature T A, is given by:
If a heat sink with thermal resistance aSA (sink to ambient) is used and the
thermal resistance from junction to case, alC, is given, then:
where aCS is the thermal resistance from the integrated circuit package (case) to the
heat sink. Due to the poor thermal conductivity of still air, the factor eCS may be
significant if air voids exist. When using dual in-line MECL III packages that dissipate
more than 4S0 mW, eCS should be reduced to a usable value by applying a good
thermal paste between the package and the sink.
All integrated circuits, including the high speed MECL family members, have
maximum allowable junction temperature limits. The MECL II family is specified as
having TJ (max) = ISOC in plastic packages and 17SoC in ceramic packages; the
MECL 10,000 family has TJ (max) = ISOC; and the MECL III family has
TJ (max) = 110e. These limits are generally lower than for most other integrated
circuits which may have a TJ (max) of between 17 Sand 200e. With very high
speed MECL circuits, stray die capacitances must be held to an absolute minimum.
To do this, the on-chip interconnect metallization is made narrow. Here the current
density and junction temperature become a significant concern to the integrated
circuit device designer and require a lower junction temperature limit.
Thermal resistance usually is not specified for digital integrated circuits though
maximum power dissipation is generally defined. The maximum ambient temper-
ature rating has been the usual thermal limit of interest to the digital integrated
circuit user. The system designer using MECL should be aware of the device junction
101
Heat Flow
temperature, regardless of what his ambient temperature is. The lower the junction
temperature of a device, the higher the reliability and consequently the life of the
device; thus, system MTBF (mean time between failure) will be increased as junction
temperatures are decreased.
MECL Integrated Circuit Heat Transfer
The electrical power dissipated in an integrated circuit is the heat source for
thermal purposes. That is, the heat flow in watts equals the power dissipation in
watts. The power-dissipating circuit elements are within a very narrow region on the
top of the die (diffusion depths for MECL are shallow). The top of the die remains
isothermal within a few degrees for MECL power dissipation levels.
The major means of heat transfer from the top of the die to the outside
surfaces of the package is by conduction through solids. Heat transfer through
bonding wires from the die to the lead frame is negligible.
Once heat is transmitted to the package, its transfer to ambient depends upon
the package mounting technique and its environment. If the integrated circuit
package is installed in, or attached to a heat sink, then heat is transferred mainly by
conduction to the heat sink, and then by convection and radiation from the heat
sink to ambient.
In the 16-pin dual in-line ceramic package (see Figure 6-lb), used for both
MECL 10,000 and MECL III, the heat flows from the top of the die, through the
6-1: MECL Package Dimensions
MECL III integrated circuits are available in the 14-lead ceramic flat package with a stud, Case 617 (suffix 51,
and in the 16-lead dual in-lin ceramic package, Case 620 (suffix L).
lal
SSUFFIX
CERAMIC PACKAGE
CASE 617
0.011
iDiTiL

\ \
I
0.140

0265 _1_
.... _.J

I
T. P.
14 e. 1
0160
-1 iill5 r 0.150 MIN -
MIN -j

001J L?003 0 109 MAX
0040 0.006 L .__ -.------L
I I
0.063 1 I-
Lead 1 Identified by II lab on that lead.
" Stud diameter controlled and solderability defined Within thiS dimenSIOn
At maximum material condition. leads to be Within 0.010" total of true position
(T.P.) with respect to each other, and withm 0.025" total of true position with
respect to stud
To convert lIlches to millimeters multiply by 25.4
102
(bl
L SUFFIX
CERAMIC PACKAGE
CASE 620
16
J: : : : : : :
0.035 1 I L
__ J 0.055
. 0.065
r-
0
.740 I

0.200
0;'; =:t-T-
0.165 I L II
,..j ri--- 0.135 /
0.100 0.015
T.P. iDl2ii SEATING
PLANE
1_0.3251 MAX
<DThis dimension is measured from the lead centers 0.300
at the seating plane with leads vertical. r ToPol
<]) Lead 1 c.olor dot. notch in lead, CD
or notch In ceramic.
To convert Inches to millimeters multiply by 25.4 I I!
Die Temperature Measurement
chip and gold eutectic die bond, to the ceramic base. The optimum heat sink
location would be in contact with the bottom of the package. Due to the poor
thermal conductivity of glass, only a limited amount of heat is transferred from the
ceramic base out through the lead frame.
The difference between the temperature of the die and some reference point
per unit power dissipation, yields the thermal resistance. The method used to
measure the temperature of MECL devices is "internal temperature sensing" - by a
special MECL integrated circuit. It employs an independent diode diffused on the
chip. It is an easy method to use and calibrate, and has a voltage output that is very
nearly a linear function of temperature.
The sensing diode within the MECL package is calibrated as a function of
temperature by using the circuit shown in Figure 6-2. The forward VBE drop of the
DIODE
FORWARD
VOLTAGE
DROP
(mV)
6-2: Diode Calibration Circuit
MECL
Package
I
I
I
I
L __ +-_","""";,,,
6-3: Typical Thermal Diode Calibration Curve
780
740
700
660
620
580
540
...........
'"
..........
'"
.........
I'---
.......
I"-
.........
"-..
500
o 10 20 30 40 50 60 70 80 90 100
TEMPERATURE (oC)
103
Determining Thermal Resistance
diode is recorded at stabilized oven temperatures between 0 and 1000C with the
diode current held constant at 100 JJ.A. A calibration curve is plotted as shown in
Figure 6-3. This curve, along with the data recorded in the test setup of Figure 6-4,
64: Thermal Evaluation Test Circuit for 16Pin
Dual InLine Ceramic Package
Circuit
Current
Control
Circuit
Voltage
&
Current ~ H
I
I
20-- J
30--
O.l/LF
Control
IE
I
I
__ ..1
2 3
will produce data to plot TJ CC), junction temperature, versus true power (watts).
The slope of the curve is the thermal resistance, 8 J (OCjWatt), of the MECL case.
By recording the ambient temperature (T A in C) during the test, the thermal
resistance from junction to ambient (8 JA in CjW) may be calculated as:
where:
To obtain the thermal resistance from junction to case (8 JC), an infinite heat
sink must be provided. This can be approximated by using a copper bar
104
MECL Packages - Thermal Characteristics
3-3/4" X 1-1/2" X 1/2" laid in thermal contact with the dual in-line l6-pin
ceramic package, or by two 3" X 3" X 1/4" copper planes for the MECL III stud
package. Copper-constantin thermocouples should be placed in holes in the sink
next to the surface of the package. These thermocouples are used to measure the
case temperature. Figure 6-5 shows a sample set-up for an infinite heat sink.
The thermal characteristics of both the l6-pin dual in-line ceramic package and
the l4-pin stud flat package, are listed in Figure 6-6. This data is based on the
following package characteristics:
(1) 14-pin stud package - ceramic package with molybdenum/gold base; die
bond by gold eutectic method. Die well is 0.070 X 0.10 inches. Stud size is 0.105"
length X 0.029" diameter.
(2) l6-pin dual in-line package - ceramic black-pigment alumina package with
goldjglass-frit base; die bond - gold eutectic; lead frame is 42% nickel, 58% iron,
with tin plating; die well is 0.11" X 0.14".
6-5: Sample Infinite Heat Sink
i---L Holes for Thermocouples
Printed Circuit Board _____ r 7 7 ~ T T 7 ~ '-CO",", ,,,
MECL DeVice;= Thermal
Paste
6-6: Typical Thermal Characteristics for MECL Packages
THERMAL
CHARACTERISTIC
16-PIN DIP
NOTES
TYPICAL
VALUE
105
14-PIN STUD
TYPICAL
NOTES VALUE
MECL dc Performance versus Temperature
MECL DC Thermal Characteristics
To fully understand the thermal effects on the characteristics of MECL circuits,
an explanation of the output level tracking and the reference (VBB) tracking will be
presented. Some of the thermal equations offered are used mainly by an MECL
integrated circuit designer, but are presented here to illustrate what parameters are
changing and how they change as a function of temperature. Figure 6-7 shows the
MECL III gate used for equation derivation. For all calculations, an ambient
reference temperature of 25C was chosen. The MECL circuit has the following
basic parameters which influence dc performance: VBE, beta, and resistor variations
with temperature.
The threshold voltage level (VBB) is most important and so an expression for
VBB as a function of VBE, beta, and resistor values in the bias supply is derived
first. Then, to analyze the temperature dependence of VBB, a total derivative with
respect to temperature is found in terms of dVBE/dT, dl3/dT, dR l/dT, dR2/dT, and
dR3/dT.
6-7: Basic High Input Impedance MECL III Gate
VCC2
A D
OR
Output
NOR
Output
~ ~ __________ ~ ________
Y
Inputs
V
CC1
= V
CC
2 = Gnd
VEE = -5.2 Vdc
If loop equations are written for the bias supply, the expression obtained for
VBB is:
RIR313 (VEE - V
BE
+ 21 + RIR3 (VEE - V
BE
+ 21 -
13 (R. R3 + R
2
R
3
) +
106
(1)
VBB and 1 Level Dependence on Temperature
Differentiating with respect to temperature. T:
dV
BB
aV
BB
dR
I
aV
BB
dR
2
aV
BB
dV
BE
aV
BB
dR3
=
--+
----+ +
----+
dT aR
I
dT
aR2
dT aV
BE
dT aR
3
dT
aV
BB

a{3 dT
(2)
Solving equation (1) for VBB at 25C using the parameters,
Rl 0.35 kn VEE = -5.2 volts
R2
= 1.958 kn (3 = 100
R3
= 3.0 kn
VBE =
0.745 volts
<P = 0.80 volts Uunction drop)
we obtain:
VBB
= -1.29 volts. (3)
The partial differential equations for reducing equation (2) will not be solved
here due to their length. However a solution of (2) will show that the change of VBB
with temperature is:
dVBB/dT = +1.11 mV;oC .
(4)
This threshold tracking level will always insure that V BB is centered between the
VOH and VOL output logic levels. As a result, noise immunity can be guaranteed
across the full operating temperature range.
Temperature variations in the two logic levels can be derived from the basic
equations for the MECL gate. The logic 1 level equation is simply a relation of VOH
to the emitter-follower base-emitter voltage drop (VBE) plus some further
dependence upon emitter-follower base current through the current-switch collector
resistor. It can be shown that the contribution by changes in {3EF and RC to the I
logic level output is about 100}1V;oC. These changes subtract from the nominal
dVBE/dT of -1.5 mV;oC.
The basic equation for the 1 logic level is:
Differentiating with respect to temperature and inserting the values discussed:
dV
OH
dT
dV
OH
dT
= (-1) (-1.5mV;oC)- 0.1 mV;oC
= + 1.5 mV;oC - 0.1 mV;oC = + 1.4 mV;oC .
107
(5)
(6)
(7)
Level Dependence on Temperature
The logic f/J level can be calculated by developing the following equation from
Figure 6-7:
VOL (OR)
(8)
where:
VEE = - 5.2 volts,
V
BE
1 = 0.745 volts (bias driver transistor),
V
BE2
0.870 volts (current switch transistor),
V
BE3
0.810 volts (emitter-follower transistor),
<I> = 0.800 volts (bias driver diode drop).
Substituting values yields:
VOL (OR) = - 1.745 volts. (9)
The logic zero level change with temperature can now be calculated from:
(10)
So,
In normal operating temperature environments, the bias voltage shifts in such a
way that it always remains halfway between the logic levels. Figure 6-8 shows the
logic levels as a function of temperature for the MECL III gate.
The effects of temperature on MECL can be illustrated by a specific example.
Assume that within a panel, one card is operating near the inlet airflow duct at
25C, and another interconnected card (remote from the air inlet) is at 35C. Thus a
10C thermal differential exists within the system. The 25C device has a typical
VOH of -0.900 volts and a VOL of -1.700 volts. The 35C device will have the
following typical levels:
108
Effect of Temperature Differentials on MECL
dV
V OH (25C) + (LlT)
(-900mV) + (lAmV;oC)(lOC) - 0.886 volts.
dV
VOL (35C) = VOL (25C) + (LlT)
VOL (35C) = (-1,700 mY) + (0.77 mV;oC) (lOC) =-1.692 volts.
This shows that a shift of only 14 m V took place in the logic 1 level and about
8 mV in the logic 0 level. The overall loss in noise immunity (N.I.) would be even
smaller than these figures - due to the positive threshold-shifL It is recommended
that thermal gradients be limited to on-card differentials under 35C. If differentials
on a card get as great or greater than 35
0
C, then good thermal management has
not been employed. However when a differential of 35
0
C or greater exists
between panels or cabinets, another feature of MECL logic can be used to
advan tage: namely, the availability of complementary output devices and line
receivers allows a differential mode of line driving and receiving which eliminates the
loss of noise immunity between units that have large temperature differences. The
differential transmission of signals on twisted pair is covered in detail in Chapter 4.
The measure of safety, noise margin (N .M.), is defined as the worst case
input threshold voltage (VIHA(min) or VILA(max)) for which the output is still
within specified limits (>VOLA or <VOHA), as was indicated in Chapter 1.
Logic
Level
(Volts)
NI{NM{
68: MECL III Logic Levels versus Temperature
- 0.6 ,----,----,---.----y--,------r---,
-0.8
VIHmin
VOHmax
.J----+-VOHmin
VOHAmin
IHAmin
-1.2 r--i-i--i-i-i--=t==+-
I VBB
- 1.4 r----+-----i---f------j---j-----j---::L.-V I LA max
VOLAmax
-1.6
VOLmax
-1.8 I Lmin
VOLmin
- 2.0 __ -'-_----' __ -'--_----' _ ____'
o 25 50 75
TEMPE RA TU RE c
109
That is:
Worst Case Temperature Effects
N.M.(logicqJlevel) VILA(max) - VOLA(max)
N.M. (logic 1 level) = VOHA (min) - VIHA (min).
As can be seen from the data in Figure 6-9, a worst case noise margin of
125 m V is guaranteed for the stud package, and a worst case noise margin of
115 mV is guaranteed for the dual in-line 16-pin ceramic package (both between a
and 75C with packages at the same ambient temperature, TA). The majority of the
10 mV N.M. difference between the flat and the DIP packages can be attributed to
differences in package thermal resistances. For MECL 10,000, worst case noise
margin is 125 mY, while MECL II is specified at 175 m V.
6-9: MECl III Worst-Case logic levels
ENVIRONMENT TEMPERATURE
VOLTAGE
LEVEL
(VOLTS)
S-PACKAGE
NOTE: Output Load R L is 50 ohms to -2.0 volts.
(JJA equal to 50
o
C/W.
L-PACKAGE
When calculating system noise immunity three factors must be taken into
consideration. These are: loss of immunity due to temperature differentials (as
above); power supply line losses, and power supply regulation (as shown in
Chapter 5); and signal losses due to undershoot and ringing on signal lines (as
described in Chapters 3 and 4). Proper attention must be given both to power
distribution and to thermal factors for any system. The reason is that losses derived
from these two areas directly subtract from the circuit's ability to withstand
external noise, and to function properly despite signal deterioration due to
mismatched lines.
Heat Dissipation Techniques
The majority of MECL users provide some form of air flow cooling in medium
and large size systems. For this reason MECL 10,000 and MECL III output levels are
specified with air flow at 500 linear feet per minute (or greater) across the package.
Many small systems and test circuits do not use forced air flow, but do use
convective cooling with ambient temperature air, or some form of heat conduction
- to avoid large thermal gradients. MECL II output levels are specified for still air,
110
Air Cooling
but many MECL II systems have used forced air cooling to limit air temperatures
and thermal differences between packages.
As air passes over devices on a printed circuit board, it absorbs heat from each
package. Thus the ambient temperature of the air will increase as it flows from inlet
to outlet. The heat gradient from the first package to the last package is a function
of the package density, air flow rate, and the individual package dissipations. The
table in Figure 6-10 lists this gradient at various power levels for an air flow rate of
500 LFPM. These figures show the increase in junction temperature for each of the
16-pin DIPs as the inlet air passes over each device. Although Z-axis air flow
information is given, the figures are similar for air flow 90 from this axis, in the
plane of the PC board.
16PIN DIP
POWER DISSIPATION TJ GRADIENT
(mW) (oC/PACKAGE)
200
250
300
400
Devices mounted on 0.062" PC board
with Z axis spacing of 0.5" Air flow is
500 LFPM in the Z axis.
0.4
610: Junction Temperature Thermal Gradients
0.5
0.63
0.88
From the air flow curve of Figure 6-1 I, the 16-pin ceramic DIP has a e J A of
50C/W (MC1660L dual 4 OR/NOR, loaded 50 ohms to -2 Vdc, mounted on
6-11: Typical Thermal Resistance versus Air Flow
e
JA
, TYPICAL
THERMAL
RESISTANCE
(oC/W)
a 500 750 1000 1250
AIR FLOW (LFPM)
Dissipation Level -:lQOmW.
Air Flow -Z-Ax;s2S
o
C
Measurement Method _ Calibrated 'Diode
Equi\i'!.lernl
/t;2PrintedCirc.u;t Soard

11 I
A Typical Thermal Calculation
printed circuit board with SOO LFPM air flow). This is a fairly standard air flow rate
for cooling a moderate size system. In this case the first device in the 2SoC air flow
stream would have a junction temperature of:
TJ fJJAPD + TA
TJ (SOC/W) (O.ISS W) + 2SoC = 34.2C.
At an average power level of 200 mW/package, the heat gradient for junction
temperature increase is OAoC per package. For example, the tenth package in an air
flow path would have a junction temperature of 37.SoC.
The following is a typical thermal calculation for the amount of heat sinking to
use with a dual in-line ceramic 16-pin MECL III counter circuit dissipating 900 mW.
The maximum allowable junction temperature. TJ, is 1100C and the operating
ambient temperature, T A, is 25C. These calculations are based on still air:
fJJA max
It is known that:
where:
therefore,
TJ max - TA max
PD max
110C - 2SoC
0.9 watt
thermal resistance, junction to case,
fJCS thermal resistance, case to heat sink,
fJSA = thermal resistance, sink to ambient;
fJSA = fJJA - (fJJC + 8CS)
In still air, the thermal resistance, fJJC, for the l6-pin DILis 2SoC/W. If thermal
paste (Dow Corning 340, or equivalent) is used, the case to sink thermal resistance,
fJCS, would be SOCjW. Thus the thermal resistance, Os A, can be calculated:
This is a worst allowable value. The heat sinking method actually chosen should
have less thermal resistance than this, to insure the j unction temperature (T J) does
not exceed 110C.
The above calculations assume all package heat is dissipated through a heat
sink. However, from Figure 6-11 it is seen that fJ JA for the package on a circuit
board in still air is 7 SC(W. Since this is below the 94.S0CjW previously calculated,
it is possible to use the MECL III counter in still air at 2S
o
C and maintain the die
temperatures within rated limits.
However, to allow for increased ambient temperatures and tight packaging, it is
recommended that MECL III systems be designed with air flow, and that high
dissipation MSI parts have additional heat sinking.
Mounting Techniques
Mounting techniques are particularly important with MECL III because of its
higher package power dissipation. Some of the more complex MSI MECL III
112
Package Mounting Techniques
functions, such as the MC 1678 decade counter, dissipate up to 900 mW and do
require special cooling. MECL II and MECL 10,000 dissipate much less package
power, so standard mounting is normally more than adequate for most systems. For
this reason, most of the mounting techniques discussed in this section apply
primarily to MECL III.
The main concern when designing component mounting is to provide not only
a good circuit ground path, but also to provide a conductive thermal path away from
the package case. The MECL stud flat package presents a problem since the stud,
like the collector on some power transistors, is not at ground potential. This means
that methods must be used to electrically insulate the stud from ground, while
providing a good thermal path to the heat sinking medium.
The use of a multilayer printed circuit layout is the easiest way to provide both
a ground plane and a solid thermal path on the VEE (-5.2 volt) plane. Previous
discussions about strip lines (cf Chapter 3), have pointed out that a solid ground
with no discontinuities is desirable below every micros trip line. Thus it is
recommended that the VEE plane (thermal plane) be the bottom layer of the
multilayer board, and that 3 ounce copper be used to conduct heat from the stud.
However, if the number of MECL III devices used per board is less than 6 packages,
then the normal 1- or 2-ounce copper clad may be used.
Two sided printed circuit boards may be used on layouts where the board
dimensions and the package count are small. Here the most difficult problem is to
maintain a good ground environment between interconnecting signal points. In the
two sided layout, where signal lines may have to be placed on both sides of the
board, the VEE thermal plane can be used as a pseudo ground plane. This pseudo
ground plane becomes the ac ground reference under the signal lines placed on the
same side as the V CC ground plane, thus maintaining a microstrip signal line
environment.
A commercially available heat sink that was developed for use with the 16-pin
dual in-line ceramic MECL III package is illustrated in Figure 6-12. This heat sink is
612: 16 Pin DIP Heat Sink Mounting
113
Screws
Retainer Clip
IERC DC000080S
(Or Equivalent)
Heat Dissipator
IERC LlC14A2CB
(Or Equivalent)

-.If Hole
./ "
Multi-Layer
PC Board
Conduction Cooling via Printed Channels
small in size and will not affect lead interconnections. The overall package is very
efficient in removing heat from the case as indicated by the graph in Figure 6-13.
This type sink is recommended for the MC 167 8L decade counter complex function,
or for that matter, any dual in-line package dissipating> 500 mW. Another suitable
heat sink is made by the Thermalloy Corporation.
6-13: Thermal Curves for 16-Pin DIP Heat Sink
I ERC-L1C 14A2CB (Or Equivalent)
PD (mW)
Printed channeling is a useful technique for conducting heat away from the
MECL DIP package when the device is soldered into a printed circuit board and
thennal paste is used between the package and channel. As illustrated in Figure 6-14,
this heat dissipation surface could also serve as a VEE voltage distribution or ground
bus. The channels should terminate into channel strips at each side or rear of a
plug-in type printed circuit board. Then, by means of wipers that come into thermal
contact with the edge channels, the heat can be removed from the circuit board into
the cabinet or board slide-rack. This same technique can be used with the MECL
stud flat package.
Printed
Channel
(May be used
for VEE or as
a ground bus.)
Double Sided PC Board
6-14: Channel/Wiper Heat Sinking
The importance of thermal management cannot be overemphasized. Proper
design in this area can result in excellent system performance and increased reliability,
especially in MECL III systems where higher power dissipation is encountered.
114
Transmission
Line Theory
Understanding the operation of transmission lines used in conjunction with
high speed MECL circuits is necessary in order to be able to completely characterize
system operation. While it is not expected that every system interconnection will be
fully evaluated by a designer, the information in this chapter will be especially useful
for setting up system design rules. This chapter describes transmission lines with
respect to both line reflections and propagation delay times. Discussed will be the
use of the Time Domain Reflectometer (TDR) for measuring transmission line
characteristics.
Transmission Line Design Information
A transmission line as used with high speed MECL is a signal path that exhibits
a characteristic impedance. Coaxial cables and twisted pair lines have a defined
characteristic impedance and are commonly referred to as transmission lines.
Equally important, printed circuit fabrication of microstrip and striplines (as
discussed in Chapter 3) results in closely-controlled transmission-line impedance.
The equations for voltage and current along a transmission line are fairly
universal and may be found in reference 4. These equations show the voltage and
current transmitted along a transmission line using the differential equations based
on a point along the line.
Transmission lines may be approximated by the lumped constant representa-
tion shown in Figure 7-1. The effect of the resistance, Ro' of the line on the
characteristic impedance, Zo, is negligible, but it will cause some loss in voltage at
the receiving end of long lines. The inductance and capacitance of the line in the
presence of a ground plane are a function of the dielectric medium, the thickness
and width of the line, and the spacing from the ground plane. The inductance and
capacitance of the line can be measured using an LC meter.
7-1: Equivalent Circuit of a Transmission Line
115
Transmission Line, Driven by MECL Gate
Microstrip and strip lines may be treated as operating in the transverse
electro-magnetic (TEM) mode. Although microstrip propagation is not purely TEM
because of non-uniform dielectrics, for all practical purposes it can be treated as
TEM. The characteristic impedance of the line is Zo = JLo/C
o
and the propagation
delay is tpd = JLoC
o
= ZoC
o
' Reference 1 shows that for a homogeneous medium
the propagation delay is also equal to tpd = file = J J10 J1r eo c
r
, where J1 is the
permeability and e is the permittivity of the medium. In transmission lines, the
relative permeability (J1r) is unity, J10 = 41T X 10-
7
Henry/meter, and eo =
8.85 X 10- 12 Farad/meter. Therefore, tpd = 1.017 Fr ns/ft, as was discussed
earlier in Chapter 3 (er is the relative dielectric constant). For microstrip lines on
glass epoxy boards e
r
= 3.0, and for strip lines e
r
= 5.0 (see reference I).
From transmission line theory for a losslcss line, it can be shown that a signal
sent down a line of constant characteristic impedance will travel along the line
without distortion. However, when the signal reaches the end of the line, a reflection
will occur if the line is not properly terminated with the characteristic impedance of
the line.
Figure 7-2 shows a MECL gate driving a transmission line terminated in a load
resistor, RL. A negative-going transition on the input to the gate will result in a
positive-going transition at the NOR output. The MECL gate is essentially a VHF
linear differential amplifier with a bandwidth of 0.37/t
r
(MHz), where tr is the rise
time of the gate in nanoseconds. The effect of the capacitance of the transmission
line will not decrease the bandwidth or affect the rise time at the MECL gate output.
Howcver, the signal at the end of a long transmission line may be attenuated due to
band width limitations in the particular type of transmission line used. For the
purposes of this discussion a long line is defined as a line having a propagation delay
larger than the rise time of the driving circuit divided by two: TO > t
r
/2 .
The circuit of Figure 7-2 can be redrawn as shown in Figure 7-3 to include the
equivalent circuit of the MECL gate. The resistor, R
o
, is the output source
-=-
MECL
Sate
VCC = +2.0 Vdc
VEE = -3.2 Vdc
7-2: MECL Gate Driving a Transmission Line
7-3: Equivalent MECL Gate Output. Driving a Transmission Line

VOL = +0.32 V --II--
-=-
116

A B
VCC = +2.0 Vdc
VEE = -3.2 Vdc
Derivation of the Total Line Voltage
impedance (for MECL 10,000 it is 7 ohms, and MECL III it is 5 ohms). According to
theory, the rise time of the driving voltage source is not affected by the capacitance
of the transmission line. Except for skin effect and dielectric losses, the signal will
remain undistorted until it reaches the load. The equation representing the voltage
waveform going down the line as a function of distance and time can be written as:
V I (X, t) = VA (t) U (t - Xtpd), for t < T D '
(1)
where:
V A(t) = ES (t)(z Zo ),
o + Ro
VA voltage at point A,
X the distance to an arbitrary point on the line,
Q total line length,
tpd = propagation delay of the line in ns/unit distance,
TD Q tpd'
U (t) a unit step function occurring at t = 0, and
ES (t) the source voltage at the sending end of the line.
When the incident voltage V I reaches the end of the long line, a reflected voltage,
V'I' will occur if RL =1= Zoo The reflection coefficient at the load, PL, can be
obtained by applying Ohm's Law.
The voltage at the load is V I + V'I which must be equal to (II + II) RL. But
II = VI/Zo, and 1'1 == -V'I/Zo (the minus sign is due to V'l travelling toward the
source). Therefore,
By definition,
V'
P = reflected voltage __ 1
L incident voltage VI
(2)
Solving for Vi/V 1 in equation 2, and substituting in the relation for PL results in:
(3)
Similarly, the reflection coefficient at the source is:
(4)
117
Lattice Method for Finding Total Line Voltage
By summing the incident voltage, V I (eq. 1), together with similar voltage
contributions from the various orders of reflection (due to PL and PS), a general
equation for total line voltage can be written, and used to develop practical design
information:
(5)
Note that as time progresses, the U step function brings successively higher order
reflection coefficient terms into V (X, t). Successive terms may be positive or
negative, depending on the resulting sign, and so damped ringing can occur. Equation
5 expresses the voltage at any point on the line, X, for any time, t. The equation can
be used graphically with a lattice diagram (as explained in References 5 and 6), to
find V (X, t).
Example 1. Figure 7-4 will be used to illustrate the lattice diagram method for
7-4: Lattice Diagram for a Typical Reflection Example
VOH = +1.22 ~
VOL = +0.32 V _--I ~ _
V A = 0.30 Vdc
IS = 4.6 rnA
VA = 1.11 Vdc
IS = 20.8 mA
t = 0
VA=1.13Vdc t=
2T
O
IS = 17.0 rnA
A
Zo = 50 ohms
Vee = +2.0 Vdc
VEE = -3.2 Vdc
118
s
t = TO
Ps = -0.82
PL=+0.13
Vs= 0.30 Vdc
IL = 4.6 rnA
Vs= 1.215 Vdc
IL =18.7mA
Vs= 1.12 Vdc
IL =17.2mA
Typical Reflection Example: Lattice Diagram Discussion
finding V (X, t) and the use of equation 5. The source impedance of the MECL III
gate is 5 ohms, resulting in a reflection coefficient at the source of -0.82 for a line
impedance of 50 ohms.
The load resistor is arbitrarily chosen to be 30 percent greater (65 ohms) than
the characteristic impedance (50 ohms) so that reflections will occur. The resulting
reflection coefficient at the load is PL = +0.13. Two vertical lines are drawn to
represent the input of the line, point A, and the output of the line, point B. A line is
drawn from point A to point B before t = 0 to represent the steady state
conditions. Note that for V CC = +2 V and VEE = - 3.2 volts, the nominal logic
levels are approximately logic f/J = 0.3 volts, and logic 1 = 1.14 volts. (These power
supply conditions are used to permit convenient measurements when output
resistors are returned directly to ground). For steady state conditions, the line looks
like a short line with a resistance equal to Rdc. It can be assumed that Rdc is
negligible for this example.
The voltage and current at points A and B are the same initially, as shown in
the diagram. At t = 0, the voltage at the source switches from a logic f/J to a logic 1
level. The voltage term, V A (t), in equation 1 is:
where:
internal voltage swing in the circuit.
Therefore, at time t = 0 a voltage waveform, V I = 0.81 volt, and a current,
II = 16.2 rnA, travel down the line as shown in the diagram by the line from
t = 0 to t = TD (TD is the time it takes for the wavefront to travel down the
length of line, Q). Next, a line is drawn from t = TD to t = 2TD. Voltage and
current values are indicated. Note that here the reflected current is negative,
indicating the current is flowing back toward the source; the reflection coefficient
for the current is a minus one times the reflection coefficient for the voltage.
To find the voltage at point B for t = TD all the voltages arriving at and
leaving from this point are summed. The same is done to determine the load current.
The process continues until the voltage at the load approaches the new steady state
condition in the example, when t = 3TD. (The steady state logic 1 voltage is
actually 1.13 volts).
This example indicates that for a case in which the load resistor is 30% higher
than the characteristic impedance, 85 m V of overshoot and 10 ill V of undershoot
would occur. Generally, as far as noise immunity is concerned, only the undershoot
need be considered. The typical noise immunity (or noise margin) for a MECL
circuit is greater than 200 mY. Since the undershoot in this example was 10 mY, the
typical noise immunity would exceed 190 m V. In actual system design, typically more
than 100 m V of undershoot can be tolerated. Regarding overshoot, 300 m V can
be tolerated, except in some early ac coupled flip-flops (MECL I and II). This
restriction insures that saturation of the input transistor does not occur (if it did, the
gate would slow down). If a 100 ohm load resistor were used in Figure 7-4, the
resulting overshoot would be about 220 mV and the undershoot, about 80 mY. In
effect then, if the load resistor is twice the characteristic impedance, the noise
margin is typically 120 mV which is more than acceptable for MECL circuits.
119
Effect of the Termination Resistor
A slightly different situation can exist when the output of the MECL gate
switches from a logic I to a logic 0. The output of the MECL gate will tum off if the
termination resistor, RL, is somewhat larger than the characteristic impedance of the
line. For the conditions in Figure 7-4, the output transistor of the MECL gate will
turn off at t = 0 for the negative going transition, when RL > 70 ohms.
An equation for the value of RL at which the gate will turn off can be derived
as follows. The maximum voltage change at point A, Figure 7-4, (due to turning off
the output transistor) is the product of the dc current in the line and the
characteristic impedance of the line:
The voltage at point A is also dependent on the internal resistance of the driving gate
Ro and the internal logic swing:
Equating the two and solving for RL:
(6)
Thus for the conditions given in Figure 7-4, the output transistor will tum off at
1.22 (5 + 50) .
t = 0 when RL = - 5 = 70S"! IS exceeded.
0.9
The case for which the MECL output turns off is not in itself a serious
problem, although it makes a thorough analysis more difficult. Two reflection
coefficients must be used at the sending end, and a piecewise approach used in
determining the voltage reflections.
Example 2. The condition for a negative-going transition will now be analyzed
(cf Figure 7-5.) The steady state high logic level current is:
= 11.6 rnA.
For the conditions shown in Figure 7-5, the use of equation 6 shows that the
load resistor is indeed larger than required to turn off the output transistor during a
negative transition.
To determine the voltage VI at t = 0, the following equation results from the
application of Ohm's Law to the circuit:
(7)
120
Lattice DiD.gram Method for a Negative-Going Transition
For the example shown, let RE = 00, then:
(8)
Solving equation 8, V I = -0.58 volt. The implication of this result is that stubbing
off the line with gate loads in a distributed fashion is not recommended, due to the
reduced intital voltage swing. However, it would be acceptable to lump the loads at
the end of the line (as will be shown).
Since the value of the load resistor is greater than the characteristic impedance,
the voltage swing at the load resistor is greater than V 1 by the amount of PL V 1 (in
this example, 193 mY). When t = TD + Tl, the voltage at B is equal to 0.387 volt;
so 82 m V of undershoot occurs. Undershoot on the falling edge is defined as the
amount of voltage step above the nominal logic (/) level of 0.305 volt. Overshoot in
the low logic state is defined as the amount of voltage change below the logic (/) leveL

IS 11.6 mA
VA 0.58 Vdc
IS = 0
VA = 0.31 Vdc
IS = 2.31 mA
75: Lattice Diagram for NegativeGoing Voltage Transition
+2.0 Vdc
-3.2 Vdc
-3.2 Vdc
A
PS1 = -0.82 When output transistor
is on.
PS2 = + 1.0 When transistor is off.
PL = +0.333
R02 = 00
Ro1 = 5.0 ohms
20 = 50 ohms
B
t = TD
121
VB = 1.16 Vdc
IL = 11.6 mA
VB = 0.387 Vdc
IL = 3.87 mA
VB = 0.283 Vdc
IL = 2.83 mA
Voltage Waveforms as a Function of Time
In Figure 7-6, the voltage waveforms at points A and B of this example are
shown as a function of time. To be more realistic, the waveform in the figure is
shown to be a negative-going ramp rather than an abrupt step function. The term,
T 1, is the amount of time it takes for the waveform at A to switch to the level at
which the output transistor turns off. The fall time of the signal would have been
longer by an amount equal to T'l
(1.16 - 0.305)
------- TI, if the termination resistor
0.16 - 0.58)
had been 70 ohms or less.
The reflected voltage waveform leaving point B at t = TD arrives at point A at
t = 2TD. The source impedance is very high initially (Ps = + 1.0), with the output
transistor being in the off condition until the voltage at A falls to 0.32 volt. Then,
the source impedance changes to 5 ohms (Ps = -0.82). The following formula may
be used to determine the point at which the transistor turns on:
(valid prior to transistor
conduction),
(9)
where V 1 is now the incident voltage approaching the source and Ll V source is the
change in voltage at the source necessary to turn the transistor on.
76: Voltage Waveforms for Points A and B in Example 2
1.2
' ~ ' ' : ' - 7 .
VA 1.0
VOLTAGE
ATA
(VOLTS) 0.8
0.6
0.4
VB 0.8..1-"_",
VOLTAGE
AT 8
(VOLTS) 0.6
0.4
0.2
t=TO To+T1 t=2TO 2TO+T1 t=3To 3TO+T1
2TO+ 0.67 T1 3To + 0.67 T1
122
Propagation Delay Calculations
In this example the actual voltage change for conduction to occur is:
~ V source = 0.32 - 0.58 = -0.26 volt. Therefore, the voltage waveform approach-
ing the source (193 m V) can be broken into two signals, VII = -0.13, and V 12 =
-0.063 volt. The reflected voltage due to Vll is V'll = -0.13 volt, and for Vl2,
the reflected voltage is V'12 = (-0.82) (-0.063) = +0.052 volt. The two reflected
voltages of opposite polarity at point A going toward point B are the reason for the
increased overshoot of short duration at point B, when t = 3T D + 0 ~ / : 3 ) T 1 (see
Figure 7 -6).
The steady state voltage reflection that occurs after t = 2TD + Tl is the sum
of -0.13 volt and +0.052 volt, equal to -78 mV as shown in Figure 7-5. The steady
state voltage reflection can be calculated using the relation:
V'
(10)
Equation 10 may be illustrated by solving for the steady state reflection voltage
at t = 2TD + T 1 :
V'
(+ 1.0) (0.32 - 0.58) (
1
+2 ~ ) + (-0.82) lO.193
(
1 + ~ ) ~
(0.32 - 0.58) 2 00 ~ = 78 mV .
From the analysis of Figure 7-5, it is concluded that the MECL gate can safely
drive the transmission line (Zo = 50 ohms) with a 100 Q load resistor and with the
gate loads lumped at the end of the line, since less than 100 m V of undershoot
occurs. The remaining noise margin will be typically greater than 100 m V.
Signal Propagation Delay for Microstrip and Strip Lines with
Distributed or Lumped Loads
The propagation delay, t d, has been shown in Chapter 3 to be 1.77 ns/ft for
microstrip lines and 2.26 ns/lf for strip lines, when a glass epoxy dielectric is the
surrounding medium. The propagation delay time of the line will increase with gate
loading and the altered delay can be derived as follows. The unloaded propagation
delay for a transmission line is tpd = J LoCo . If a lumped load, Cd, is placed
along the line, then the propagation delay will be modified to tpd:
123
Effect of Capacitances on Propagation Delay
where Lo and Co are the intrinsic line inductance and capacitance per unit length.
Therefore, the signal propagation down the line will increase by the factor of:
g
d
1+-
C .
o
A MECL gate input should be considered to have 5 pF of capacitance for ac
loading considerations (includes stray capacitance). If 4 gate loads are placed on a
I foot signal line, then the distributed capacitance, Cd, is equal to 20 pF/ft or
1.67 pF/in. As an example, assume that it is desired to find the propagation delay
increase for a 50-ohm microstrip line on a glass epoxy board. From Figure 3-7
assume that the line width is chosen to be 25 mils; then the dielectric material
should have a thickness of 15 mils to yield Zo = son. From Figure 3-8, the
capacitance of the line is 35 pF /ft. Therefore, the modified propagation delay would
be:
tpd 1.77 ns/ft VI + = 2.21 ns/ft .
For a 50-ohm strip line on a glass epoxy board with a 15 mil spacing between
the strip line and ground plane, a 12 mil width would be required (cf Figure 3-9).
From Figure 3-10, the strip line would exhibit a capacitance of 41 pF /ft.
The modified propagation delay for such a strip line would be:
tpd = 2.26 ns/ft VI + = 2.75 ns/ft .
Notice that the propagation delay forthe strip line and the microstrip line change by
approximately the same factor when the separation between the line and ground
plane, and the characteristic impedance are the samc. Howevcr the line width of the
strip line is less (by a factor of 2) than the microstrip line for the same characteristic
impedance.
It should be noted that to obtain the minimum change and lowest propagation
delay as a function of gate loading, the lowest characteristic impedance line should
be used. This will result in the largest intrinsic line capacitance. With MECL 10,000
the lowest impedance that can be used is (l bout 35 ohms (VTT = - 2.0 volts,
RTT = 35 ohms).
According to theory (Reference 1), whenever an open line (stub) is driven by a
pulse, the resultant overshoot and ring are held to about IS percent of the logic
swing if the two way delay of the line is less than the rise time of the pulse. The
maximum line length, max, may be calculated using the eq uality:
where tr is the rise time of the pulse in nanoseconds, and tpd is the modified
propagation delay in nanoseconds/inch from equation 11.
124
Maximum Line Length Calculations
A quadratic equation for maximum line length for G-l 0 fiber glass epoxy micro-
strip conductors may be written in terms of Cd, Co and tr as:
Cd
Q2 max + -C Qmax - 11.1 tt = 0, (for microstrip lines). (12)
'0
An equation for maximum open line length for a strip line (using G-IO fiber
glass epoxy material) can be written in a similar fashion. The result is:
2 Cd 2 _
Q max + C Qmax - 7.1 tr - 0, (for strip lines). (13)
o
Using the lattice diagram, it has been found that the rule of thumb used to
derive equations 12 and 13 should be modified for an open line because the incident
voltage doubles at the end of the line. This results in a faster rise time at the
receiving end of an unloaded line than at the driving end. An approximate value of
maximum open line length can be generated from equations 12 and 13 if the rise
time that is substituted into the equations is multiplied by an adjustment factor,
0.75. This maintains an approximate overshoot and undershoot ofless than 35% and
12% respectively.
To demonstrate how equations 12 and 13 may be used, the maximum open
line length will be computed for a 50 ohm line with a fanout of one MECL 10,000 gate.
Using the equation tpd = ZoC
o
, the line capacitance, Co, is found to be Co = 2.96
pF/in for micro strip, and Co = 3.76 pF/in for strip line. For a fanout of one,
Cd is equal to 5 pF when the device is in a socket. The rise time for MECL 10,000
is 3.5 ns which means that a value of tr = 0.75 X 3.5 = 2.6 ns should be used
in the equations. Solving equations 12 and 13, Qmax = 7.9 inches for a 50 ohm
microstrip line and Qmax = 6.2 inches for a 50 ohm strip line.
Equations 12 and 13 can be very useful in finding the approximate maximum
line length under various conditions. However if overshoot or undershoot differing
from the above values is specified, equations 95 and 103 (derived later in this
chapter) should be used for defining maximum line length. The exact voltage at the
end of an open line with loading is also derived later in this chapter, and leads to
equation 87. Using that equation, a computer program has been written in which the
maximum line length is calculated when maximum overshoot and undershoot are
specified. Figures 3-13,3-14, and 3-15 show the results of the program. Note that
the tables give the maximum line lengths for fanouts of 1, 2, 4, and 8 for various
types of lines with a wide range of characteristic impedances.
The maximum line lengths are also given for various characteristic impedances
in the backplane. The characteristic impedance of the backplane should be between
100 and 180 ohms if a ground screen is used. For MECL 10,000 from Table 3-13,
5.9 inches of open backplane wiring can be driven for a fanout of one.
It should be remembered that these line lengths are based on 100 mV maxi-
mum undershoot, and are not absolute maximum lengths with which MECL circuits
will operate. It is possible to use longer unterminated lines thap shown - the trade-
off being an associated loss of noise immunity due to increased ringing.
From these calculations, it can be concluded that lower impedance lines result
in longer line lengths before termination is required. The lower impedance lines are
preferred over higher impedance lines because longer open lines are possible, and the
propagation delay down the line is reduced. In addition, more stubbed-off gate loads
can be driven with a terminated line due to its higher capacitance per unit length.
125
Time Domain Ref/ectometer Measurements of Micro strip Line
Microstrip Transmission Line Techniques,
Evaluated Using TDR Measurements
The time domain reflectometer (TDR) employs a step generator and an
oscilloscope in a system which might be described as "closed-loop radar" (cf Figure
7-7). In operation, a voltage step is propagated down the transmission line under
investigation. Both the incident and reflected voltage waves are monitored on the
oscilloscope at a particular point on the line.
OSCI LLOSCOPE
STEP
GENERATOR
7-7: Time Domain Reflectometer (et Reference 7)
Tee
The incident voltage step, Ei, is a positive edge with an amplitude of 1 volt and
a rise time of 30 ps. It is generated by a tunnel diode, which has a source impedance
of 50 ohms (HP 1817 A sampler, or equivalent). Also, the output edge has very little
overshoot (less than 5%).
This TDR technique reveals the characteristic impedance of the line under test.
It shows both the position and the nature (resistive, inductive, or capacitive) of each
discontinuity along the line, and signifies whether losses in a transmission system are
series losses or shunt losses. All of this information is immediately available from the
oscilloscope's display (cf Reference 7). An example of a microstrip line evaluated
with TDR techniques is shown below.
TDR Example 1. Board material: Norplex Type G-IO;
Dielectric thickness: h
Copper thickness: t
Dielectric constant: e
r
0.062 inch;
0.0014 inch;
5.3 .
The formula for the characteristic impedance given in Chapter 3 was:
87 ( 5.98 h )
Je
r
+ 1.41 in 0.8 w + t .
(14)
For a line width, w = 0.1 inch, the characteristic impedance of the line is calculated
to be 51 ohms. A board was fabricated as shown in Figure 7-8(a) to the dimensions
specified above. Figures 7-8(b) and 7-8(c) show the incident and reflected
126
Measurements: Impedance
7-8: TDR Determination of line Characteristic Impedance
Input Connector
Ground
P/DIV = 0.2
(b)
(e)
50-ohm Reference
t
Connector
Vertical Scale = 200 mV/div
Horizontal Scale = 0.6 ns/div
Line
Horizontal Scale = 0.4 ns/div
127
for Zo = 50 ohms
50-ohm Termination
Crosscheck with Calculated Impedance
waveforms observed with the TDR. The vertical scale is calibrated both in terms of
the voltage and the reflection coefficient, P. Equation 3 can be rearranged to
determine the characteristic impedance of the line:
(
1 + P)
Zline =""l=P Zreference
(15)
where: characteristic impedance of the line under test,
Zreference
impedance of the known line.
The 50 ohm reference point is shown in Figure 7-8( c). The mean level of the
reflected waveform due to the line has a P = +0.01. Substituting values into
equation 15 permits calculation of the line impedance:
(
1 + 0.01)
1 _ 0.01 50 ohms
51 ohms ,
which agrees closely with the calculated value.
The reflected voltage due to the connector is 40 mY. The line reflects a
voltage of 25 mV due to variations in the characteristic impedance of the line. The
reflection of 88 mV shown for the termination resistor (P = 0.088) is due to the
inductance of the resistor. It was calculated (by methods to be shown later) that the
inductance of the resistor was less than 0.9 nR.
In these experiments, the input waveform comes from a tunnel diode generator
which has a rise time of 28 ps. There is some attenuation of the signal noticeable as
it reaches the termination resistor (tr = 80 ps at the load). When driving the line
with a MECL III gate with a rise time of 1 ns, the reflection due to the inductance of
the resistor would be much less (about 10 mY).
TDR Example 2: An equation can be derived to determine the maximum reflection
voltage due to the inductance of the resistor leads. The circuit shown in Figure 7-9
will be used in the derivation.
79: Circuit for Determining the Maximum Reflected Voltage
Due to the Inductance of the Resistor Leads
Ej
Ej
Erefl{t)
--

E
out
r
20

r-- T 1
2L(s) RL + sL
t 0
ej(t)
I
s js the LaPlace
RL
notation for j w
1

-=-
128
Derivation: Maximum Reflection Voltage due to Resistor Inductance
The reflection coefficient at the load is:
RL - Zo
ZL -
Zo
(RL + sL) - Zo
s +
L
P
L
(s)
ZL + Zo
(RL + sL) + Zo
RL +Zo
( 16)
s +
L
where s is the LaPlace notation for jw. The driving voltage will be represented as:
(17)
where U(t) is a step function occurring at t = O. Taking the LaPlace transform of
equation 17 gives:
m ( T s)
Ei (s) = J I - e - I . (18)
The reflected voltage at the load is then the product of the driving voltage and the
reflection coefficient (both in the transformed plane):
s +
RL-Zo
L
Taking the inverse LaPlace transform yields:
129
(
-T s)
-ml-e 1. (19)
(RL + zo)tj
L mU(t) -
e
(20)
TDR Measurement of Resistor Inductance Effect
The maximum reflection voltage occurs at t = T 1. Then, for R = Zo:
_ mL - -T
l
(
2Zo)
E
refl
(t = T I) = E
refl
max = 2Zo 1 - e L .
(21 )
This equation relates the maximum reflected voltage, which can be measured by
TDR, and the inductance, which can then be calculated for the circuit of Figure
7-9.
TDR Example 3. This example indicates how to measure the effect of resistor leads
using the TDR. Figure 7-11 (a) shows the construction of a microstrip board used for
(b)
710: Effects Due to Termination Resistor Leads
(a)
i'",""",,?,,'
" ,I,,,'
Ground Plane
Reflection Due to the Inductance
of the Resistor's Leads
PIDIV = 0.2 t

..
..
Vertical Scale = 200 mV/div
Horizontal Scale = 0.6 ns/div
..
Zo = 50 ohms
..
(c) Equivalent Circuit
130
L=?
RL = 50 ohms
TDR Measurement of Ground Plane Effects
determining the effects of a resistor with 1" lead lengths. The reflected voltage
determined from the TDR measurement is 480 mV (see Figure 7-1O(b)). The rise
time at the input to the line is 28 ps but it is lengthened to about 80 ps as the
wavefront reaches the termination resistor.
The time, T 1, associated with the slope of the input voltage rise at the
terminating resistor can be approximated as:
tr
T} ~ 0.80 = 100 ps.
(22)
The inductance can be computed by using equation 21, giving L = 6 nH. Additional
information can be obtained from the decay of the reflection shown in Figure
7-1O(b). The decay lasts about 0.3 ns, implying a time constant of about
0.3 ns/5 = 60 ps (using 5 time constants as a decay time). The calculated time
constant for an inductance of 6 nR is: L/2Zo = 60 ps. The two results agree closely.
When driving the line with a MECL III gate - rise time = } ns - the reflection
would be only 50 m V. Most carbon resistor types will have less than 10 nR of
inductance. This inductance gives a reflection < 75 mV when the line is driven by a
MECL III gate. Note that the reflection is positive, indicating that the noise
immunity of a MECL gate connected at the load would be unchanged.
TDR Example 4. Experiments have also been performed to determine the effects of
a ground plane on the characteristic impedance of micros trip lines. Figure 7-11
711: Effects of Ground Plane Discontinuities
Ground Plane Epoxy Glass Only
Termination
Resistor ~ 50 ohms
Line Under Test ~ 50 ohms
Zc for Q 2.5"
P/DIV ~ 0.05!
~ ~ ~ ~ ~ - - - ~ - - - ~ ~ ~ ~ ~ ~
50-ohm Reference
Connector
Horizontal Scale = 0.4 ns/div
Vertical Scale ~ 50 mV/div
131
Termination Resistor
Implications for Ground Plane Design
illustrates what happens when the ground plane width under the transmission line
abruptly drops to the width of an active line. The TDR waveform shows that a 12%
reflection occurs due to this discontinuity in the ground plane.
as:
Using equation 15 the impedance of the 2-1/2 inch-long strip can be calculated
+ 0.12 50
- 0.12
68 ohms.
Figure 7-12 shows a curve that approximates the change in the characteristic
impedance of the line for various ratios of ground plane width to active line width.
Note that when the ground width is greater than 3 times the line width, the
characteristic impedance is constant according to equation 14.
712: Variation of Microstrip Impedance as a Function of
Ground Width";' Line Width
Ground Width
Line Width
A related experiment was performed to find the reflection due to a ground
plane near the active line, but not directly under it. The test configuration and test
results are shown in Figure 7-13. As indicated by the TDR measurement, the
reflection is about 36%. Again using equation 15, the impedance of the 2-1/2 inch
strip can be calculated:
+ 0.36 50
1 - 0.36
106 ohms.
The reason for the reflection is the change in the characteristic impedance
along the line resulting from the ground plane not being under part of the active
line. In such a region, capacitance of the line to ground decreases while the
inductance of the line increases, the net result being a higher characteristic
impedance.
It must be remembered that the TDR input waveform has a rise time of 28 ps.
Consequently, in a real logic circuit situation where, perhaps, a MECL III gate with a
1 ns rise time is driving the line, the reflection would actually be less than 27%, not
36% as in this example. This can be determined by scaling the value of P found with
132
Another Ground Plane Discontinuity
the TDR waveshape in Figure 7-13(b), with a 1 ns rise time. When the length of the
ground plane discontinuity is less then the distance travelled by the signal during its
rise time, then the reflection coefficient can also be calculated as:
where:
pI
the propagation delay time of the line in ns/in.
tr = the rise time of the signal in ns,
Q the length of the discontinuity in inches,
P = the reflection coefficient for 2Qtpd/tr > I
(in this case the value found with the TDR waveshape
with tr = 28 ns).
7-13: Effects of Ground Plane Discontinuity
(23)
Input Connector Termination
Resistor = 50 ohms
Ground Plane
P/OIV
= 0.2
t
(b)
E1
./
Line Under Test = 50 ohms
(a)
r--
Ii: t'\
L
'V\
....
r
Vertical Scale = 200 mV/div
Horizontal Scale = 0.8 ns/d iv
133
-TIME
TDR Observation of Hybrid Divider Reflections
For a discontinuity in the ground plane of 2.5 inches length, a propagation
delay of the line of 0.15 ns/in, and a MECL III gate with I ns rise time, the percent
reflected voltage can be calculated. From Figure 7-13(b), P is found to be 0.36.
Using equation 23,
P'
2(0.36) (2.5) (0.15)
(I)
0.27 .
Therefore, the reflection would be 27%. For a MECL 10,000 series gate, with a rise
time of 3.5 ns, the reflection would only be 7.7%.
TDR Example 5. Another measurement was performed, as shown in Figure 7 -14, to
observe the reflections due to the use of a hybrid divider. The construction of the
G round Plane
PIDIV - 0 2 t
-
Ibl E
1
L
714: Hybrid Divider
Zo = 100 ohms
r-------2" ------1
lal
2 3
.-.
"
Horizontal Scale = 200 mV/div
Vertical Scale = 0.4 ns/d iv
134
e
r
= 5.3
t = 0.0014"
h = 0.062"
R L = 100 ohms
No Mismatches Appear
Due to the Crosstalk
Between the Lines
-TIME
Reflection Due to Crosstalk
microstrip board used is shown in the figure. Note that the 50 ohm line branches
out into two 100 ohm lines. A reflection of 4 percent is observed at point 2 where
the junction occurs. Notice that the resistor exhibits a reflection of -8%, due to
capacitance of the resistor.
Previously it was found that the 50 ohm resistor was inductive. Both results
agree with Reference 8 in which it is stated that the lower values of resistors 75 n)
exhibit inductance, while the higher values behave capacitively. These effects are
also shown in the data in Figure 4 of Chapter 4. Note that no mismatch appears due
to crosstalk between the two 100 ohm branches, because of their wide separation.
Figure 7-15(b) shows the reflection due to the construction of Figure 7-15(a)
where the two 100 ohm lines have been brought close together. The reflection at
point 2 is now equal to 8% arising from the cross coupling of the two lines. Crosstalk
is discussed in References 5, 9, 10, and 11.
715: Hybrid Divider With Crosstalk Problem
I nput Connector
G round Plane
Zo ~ 50 ohms
PIDIV ~ 0 2 t
(b)
E1 I
Zo ~ 100 ohms
(a)
2 3
- r
......
.Ac \
I-'"
I
I
I
..
I
i
I
I
I
-+_.
I
+
I
I
Horizontal Scale ~ 200 mV/div
Vertical Scale ~ 0.4 ns/d iv
135
i
R L ~ 100 ohms
Er ~ 5.3 t ~ 0.0014"
h ~ 0.062" s ~ 0.08"
I
~ T I M E
Impedance In a Qrcuit With Crosstalk
Even mode or odd mode characteristic impedance (Zoe or Zoo) can be
considered to exist in a circuit with crosstalk. One, Zoe, is due to the strips being at
the same potential and carrying equal currents in the same direction. The other,
Zoo, is due to the strips being at equal but opposite potentials and carrying equal
currents in opposite directions. The backward crosstalk voltage, VB, on a passive line
is given in Reference 10 as:
E
I '
(24)
where E I is the signal propagating down the active line. Formulas are given in
References 9 and 12 for calculating Zoe and Zoo. The backward crosstalk voltage
shown in Figure 7-15(b) at point 2 is equal to 8% of the incident voltage EI. Since
both lines are active, the crosstalk due to one active line is 4% of E I for a spacing of
80 mils. Reference 5 should be consulted if information concerning crosstalk on
microstrip lines is desired. There, curves are given from which the backward
crosstalk can be predicted. (For example, Figure lOin Reference 5 may be used to
predict the backward crosstalk for Figure 7-15(a) as 11%).
Crosstalk is not ordinarily a problem when using MECL III on microstrip or
strip line circuit boards, when line spacings are greater than 30 mils. Crosstalk theory
is well described in Reference II. In it, the mutual inductance and capacitance
between two lines are used to determine the crosstalk coefficient. Crosstalk theory is
presented in some detail in this handbook in Chapter 4, "System Interconnections".
Forward crosstalk is normally much smaller than the backward crosstalk on
microstrip lines- except for very long lines (>5 feet). Forward crosstalk does not
exist at all on strip lines, since they are made with a homogeneous medium, so that
the inductively and capacitively induced currents cancel (Reference 10).
The backward crosstalk coefficients for various types of microstrip lines on
glass epoxy boards are shown in Figure 7-16 (cf also Reference 5). The backward
crosstalk coefficient is equal to:
(25)
where: LM = the inductive coupling,
C
M
the capacitive coupling,
tpd the propagation delay of the line per unit length.
TDR Example 6. The graph data in Figure 7-16 will be used to determine the
percent of crosstalk coupling for the circuit of Figure 7-15. From the dimensions of
the lines given in Figure 7-15(a), KB is found to be 0.055 from the graph. This
means that if one line (the active line) were driven with a signal, the other line
(passive) would have a coupled signal of 5.5% of the amplitude on the active line, in
a direction opposite to that of the driving signal. Since both 100 ohm lines are active
l36
KB'
CROSSTALK
CONSTANT
Data for Determining Microstrip Crosstalk
7-16: Backward Crosstalk Coefficient for Microstrip
Lines on Glass Epoxy Boards (G-10 Material)
1.0

0.1

i'..
"
'\..
0.01
0.001
10


""
"
"-
'\.
\
\
\
h = Dielectric Thicknes
(MILS)
I'..
"-
'\. to\.
f'..r-.. i\
r-...f\
'\ h = 59.0
\
\ 43.5
r\
\
1\30.5
1\
25.6
\
\
:\
\
8.3
100
S, LINE SPACING (MILS)
1000
simultaneously, the reflection observed on the TDR is twice as much, or 11 %. From
Figure 7-15, the actual crosstalk can be seen to be about 8%.
In very high speed systems, the exact shape of a line can be important, if
reflections are to be kept to a minimum. The arrangement shown in Figure 7-17(a)
has been used to investigate the behavior of two different line shapes. For one line,
corners are sharp. This permits the width of the line to be larger at corners than
elsewhere. Figure 7-l7(b) shows that a -7.5% reflection occurs at point 6 due to the
lowered characteristic impedance at the corner. For the other line, the corners are
rounded to produce a constant line width. Figure 7-17(c) shows that a constant line
impedance exists for the second line. Note that an inductive reflection, as discussed
before, does occur at the end of the line due to the inductance of the resistor. In
conclusion, it is desirable to have smooth, rounded line edges and constant line
widths when designing transmission lines for high speed systems. Resistor leads
should be kept short to minimize termination inductance.
137
Effect of Microstrip Line Shapes
7-17: Reflections Caused by Signal-Line Shape Variations
RL = 50 ohms
G round Plane
I nput Con nector
la)
PIDIV = 0.051
2 3 4
IA
Ib)
11 ..... ~ 1 1 ~
,.A
i
j
If\.f
JAt. A..
y'
.,"
"
.If ,
IIV'
til
V
i
V
-TIME
(Relative Scale)
5 6
I--
J
Ie)
.A r ~ 1a,.1\
.. J
.11\
.... ....
',..
V
V ~
""
.... I""'" Vq
'W
_TIME
(Relative Scale)
138
Reflection Due to Loading
The Effect of Loading, on a Parallel Terminated Transmission Line
For designing high speed systems it is useful to understand the effects of
loading a transmission line with MECL circuit inputs. Some tests were performed to
determine the equivalent loading effects of a MECL gate load. The input impedance
of the MECL gate is high and may be assumed to be purely capacitive as far as
reflections are concerned.
Accurate knowledge of gate input capacitance is necessary to develop accurate
loading rules. A test setup, similar to that for a TDR, was used to determine the
amount of reflection that occurred when driving four MECL III gates (MC 1660L -
cf Figure 7-18). The amount of reflection that occurred at the probe was found to
718: Test Setup for Measuring the Reflection From Four MECL III Gate Loads
EH122
PULSE GENERATOR
6dB Pad
a:
CJ
VCC 1.65 V
VEE -3.55 V
Zo 50 ohms

E1 = 0.85 V
Tr = 1.0 ns 10% to 90%
PW = 40 ns
T
1/2"
R
50
1/2 MC1660
(4 Places, In
Sockets)
510
510
510
510
510
510
510
510
be 275m V, in a direction indicating it was due to a terminal capacitance, CT A
formula may be derived so that the amount of this capacitance can be calculated.
Figure 7-19 shows the equivalent circuit which will be used for the derivation of
719: Circuit for Driving the Maximum Reflected Voltage Due
to the Capacitance of the Gate Inputs
139
Zo 50 ohms

R
50
*Total Gate Capacitance
Derivation of Maximum Reflection Amplitude
such a formula.
The reflection coefficient at the load is:
R
- Z
ZL - Zo sRC
T
+
0
(R - Z ) -
sRZoC
T
P
L
(s)
0
ZL + Zo
R (R
+ Zo) + sRZoC
T
sRC
T
+
+ Zo
(26)
From equation 18, the LaPlace transform of the input voltage can be found to be:
(27)
The reflected voltage at the load, in LaPlace notation, is:
m I - e I .
(
-T S)
(28)
Taking the inverse LaPlace transform yields:
(Zo + R) ]
)
2 RZ C t
R t _ ( 2R ZoCT ) e 0 T mU(t) +
R (Z + R)2
o
The maximum reflection occurs at t T I- Then for R = Zo, we obtain:
(30)
Equation 30 exhibits a relation between the maximum reflected voltage and the
effective capacitance causing reflection, CT, in the circuit of Figures 7-18 and 7-19.
The reflected voltage was measured to be - 27 5 m V and from equation 22, T I is
found to be 1.25 ns. Thus the total capacitance, obtained from equation 30, is
CT = 17.2 pF. Since stray capacitance, C
s
, is approximately 4.0 pF, the capacitance
140
Determining Effective Capacitance
due to the gate loads is the difference between CT and C
s
i.e., 13.2 pF, or 3.3 pF per
gate input.
For comparison, an RF vector impedance meter was used to measure the input
capacitance of a similar 4 gate setup at a frequency of 50 MHz. The total
capacitance measured 20 pF. Since the stray capacitance for this configuration
measured 6.5 pF, the capacitance due to the four gate loads is l3.5 pF, or 3.38 pF
per gate input. It is felt that the two methods agreed well enough with each other to
say that the equivalent load of a MECL III input is 3.3 pF.
MECL 10,000 series elements were also tested. It was found that a MECL
10,000 gate input measured 2.9 pF using the RF vector impedance meter. Using the
reflection method of Figure 7-18 and equation 30, the capacitance of a gate input
was found to be 2.7 pF.
If printed circuit cards are used without sockets, 3.3 pF per MECL III gate
input and 2.9 pF per MECL 10,000 gate input should be used. These values will be
used in later calculations.
It was shown in equations 12 and 13 that the maximum length of an unter-
minated line (stub length) is a function of loading. Figures 3-13,3-14, and 3-15 are a
tabulation of some values of permissible lengths versus fanout and logic family.
However, in most designs it becomes necessary to increase the line length beyond
the distances specified in the table. It has been shown that for long lines, 2TO (line)
> tr (pulse), a termination resistor will reduce or eliminate reflections. In a practical
situation, a MECL gate driving a transmission line must feed other gates along that
line. So it is important to be able to determine the effects of individual gate loads on
the line.
There are two ways of placing gates on a parallel terminated transmission line:
one is called "distributed" loading, the other "lumped" loading. Figure 7-20 shows
an example of a parallel terminated line with a lumped load at the end. The term TO
7-20: Driving a Parallel Terminated Line
VTT = -2.0 Vdc
A
B
141
Distributed Loading
represents the delay of the line. Since a full logic swing is available all along the line,
parallel termination permits distributed loading to be placed anywhere along the
line.
The change in characteristic impedance of a line caused by gate loads being
distributed along the line can be calculated. For a lossless line the characteristic
impedance of a transmission line is:
(31 )
where Lo is the intrinsic inductance of the line and Co is the intrinsic capacitance of
the line, both per unit length. The MECL gate has a high input impedance so that
only the capacitive effect need be considered for ac conditions. The characteristic
impedance of a transmission line altered by gate loading, Z;, is:
Z'
o
(32)
where Zo is the original line impedance defined in equation 31 and Cd is the
distributed gate capacitance. The propagation delay per unit length of a lossless
transmission line is:
tpd = J LoCo
Rearranging, and using equation 31 gives:
(33)
(34)
Example. An application of the foregoing relationships and rules can be seen in the
following design problem. Given: a 68 ohm microstrip line 8 inches long. It is
desired to drive four MECL III gate loads spaced equally at 2" intervals along this
line. These loads are, of course, "distributed" loads. The micros trip line is on a glass
epoxy board which has a dielectric constant, e
r
, of 5.0. It is necessary to determine a
value for a parallel terminating resistor which will essentially eliminate reflections on
the line.
First, the propagation delay of the microstrip line can be found using the
relation from Chapter 3:
tpd = 1.017 J0.475 e
r
+ 0.67 ns/ft = 1.77 ns.ft = 0.148 ns/in, (35)
in this case. Using equation 34, the line capacitance, Co is found to be:
0.148
---=
68
2.l8 pF/in.
142
Terminating Distributed Loads
Four MECL III gate loads are equivalent to a load capacitance of 13.2 pF
which is distributed along 8 inches of line. Therefore, Cd == 13.2 pF 18 in. ==
1.65 pF/in. Substituting these values into equation 32 gives:
Z' == 68
o , 11 + 1.65
V 2.18
51.5 ohms.
Thus, a 51 ohm termination resistor would be acceptable for terminating the 8
inch 68 ohm microstrip line with four distributed MECL III gates. The resulting
circuit is shown in Figure 7-21.
The driving gate shown in Figure 7-21, besides driving the long transmission
line, can also drive many lines (no limit) as long as the length of each stub does not
exceed the limits of Figures 3-l3, 3-14, or 3-15. For instance, if a 50 ohm microstrip
line were used with MECL III to connect the driving gate to I gate load in one
direction, and to four gate loads in another direction (in addition to the loads shown
in Figure 7-21), then from Figure 3-15 the maximum permissible stub lengths are
1.6 and 0.7 inches, respectively (of Figure 7-22). It should be noted that the four
A
Driving
Gate
7-21: Example Illustrating Distributed Loading
Zo ~ 68 ohms
~ - t - - 2
51
7-22: MECL III Gate Driving a Long Transmission Line with Distributed
Loads, and Short Stubs at the Driving Source
Driving
Gate
(;
Qmax
1.6 in
/
20 ~ 68 ohms
f-- 2" --1---2" --1---- 2" - l - - - - 2,,----1
50
-2.0 Vdc
20 ~ 50 ohms
143
B
Computing Maximum Load Capacitance
gates on the stub (Qmax = 0.7 inch) could be lumped at the end of that line,
without the need for any other changes.
In order to determine the amount of ref1ection which can be tolerated on a
line, the following development is presented. Ref1ection is, of course, caused by gate
loading which produces a change in the impedance on a section of the transmission
line. The equations to be developed use distributed line theory an approximate
method, but one which gives very accurate results, as verified in Reference 1 1.
The reflection coefficient given in equation 3 can be revised to take into
account the reflection due to the altered characteristic impedance produced by
loading:
P
Substituting the expression for Z; given in equation 32 yields:
H
- 1 +-
C
P
0
(36)
H
+ 1+-
Co
From this equation, it is possible to find the maximum load capacitance that
can be distributed or Jumped on a length of transmission line. Further, the length of
transmission line for distributing loads will be assumed to be the stub length defined
in equations 12 and 13. This length of line will limit reflection discontinuities caused
by differences between distributed and lumped loads. However a rule is needed
which can be stated for a particular value of transmission line, to specify a limit for
the number of gate loads distributed or lumped along an arbitrary length of line.
For a maximum reflection of 20% (P = -0.20) equation 36 may be solved for
the ratio of Cd/Co, giving:
== 1.25 . (37)
Since Cd is the distributed gate load capacitance per unit line length, it may be
written that:
(38)
where Cd is the total gate load capacitance. Substituting into equation 37 yields:
1.25 Qmax .
(39)
144
Maximum Loads Related to Maximum Line Lengths
For a 50 ohm micros trip transmission line, on a glass epoxy board, with
MECL III gates, with Co = 2.96 pF lin, and tr = 1.1 ns, equation 12 may be used
to find Qmax ~ 2.5 inches. Then substituting into equation 39 and solving:
Cd = 9.2 pF. This means that up to 9.2 pF can be distributed or lumped along any
2.5 inches of 50 ohm microstrip line using MECL III. Two MECL III gate loads can
be used along any 1.8 inches of line for a carefully laid out board, or two MECL III
gate loads and 2.6 pF of stray capacitance can be distributed or lumped along any
2.5 inches of line.
For a 50 ohm strip line on a glass epoxy board and using MECL III gates, with
Co = 3.77 pF/in and tr = 1.1 ns, equations 13 and 39 may be used to find
Qmax ~ 2.0 inches. Then substituting into equation 39 and solving, Cd = 9.4 pF.
This means that up to 9.4 pF can be distributed or lumped along any 2.0 inch
portion of 50 ohm strip line when using MECL III gates. If 3.3 pF per gate input is
used, then from equation 39 two gate loads can be lumped or distributed along any
1.4 inch portion of a 50 ohm strip line.
It is seen from these calculations that strip line has an advantage over
microstrip: it can be used for driving more gate loads per unit length than
microstrip, granting the same amount of reflection in each case. This is due to strip
line having a larger capacitance per unit length. Figure 7-23 gives values for the
maximum capacitance that can be lumped or distributed over a length (Qmax) of line
for MECL III, MECL 10,000, and high speed MECL II.
As an example of how Figure 7-23 can be used, suppose 68 ohm microstrip
lines are to be used with the MECL 10,000 series. From the Figure, 21 pF of
capacitance of five gate loads can be lumped or distributed over any 7.7 inch portion
of the line. The rise times shown in the figure are characteristic of the particular
logic family and were used in the calculations to obtain the data.
7-23: Maximum Capacitance That Can Be Lumped or Distributed
Over a Length of Terminated Transmission Line ~ a x
CHARACTERISTIC IMPEDANCE OF TRANSMISSION LINE
COAX
MICROSTRIP (e
r
= 5.0) (e
r
= 2.2)
145
Output from Series or Parallel Terminated Lines
Analysis: Series Terminated Lines Compared to Parallel Terminated Lines
The propagation delay increase due to gate loading when a line is series
terminated is about twice as large as for a comparable parallel terminated line.
Equation 11 gives a fairly close approximation for the propagation delay of a
parallel terminated line with loading. This equation was:
(II)
The output waveform at the end of a series terminated line or at the end of a
parallel terminated line can be derived from an equivalent circuit using Thevenin's
Theorem, assuming the line is long (2TD t
r
). Figure 7-24(a) shows a parallel
terminated transmission line circuit, along with the waveform driving the line.
The equivalent open circuit voltage of the line is twice the input voltage and
the Thevenin resistance is the impedance of the open line at the load looking toward
the source. The Thevenin equivalent for a parallel terminated line is shown in Figure
7-24: Parallel Terminated Transmission Line. and its Thevenin Equivalent
146
General Derivation of Line Output Voltage
7-24(b). Figure 7-25 shows the Thevenin equivalent for a series terminated line. Note
that the impedance (Zo) of the series terminated line is twice as large as that for the
parallel terminated line. A general equation can be derived for the output voltage
assuming the impedance in the circuit to be R. Then a substitution for R will give
the equations for both types of lines.
Writing the equation around either Thevenin equivalent loop:
But also:
I
iR +--
C
r
t
f
o
dt .
725: Thevenin Equivalent of Series Terminated Transmission Line
,---------------l
I I
I Zo I eout(t)
I
t
t
i ej(t) I CT
It-=-
L _______________
Equating the equations and taking the LaPlace transform of both sides gives:
But:
Therefore:
m
')
~
m m -TIS
---e
s2 s2
t
f idt or Eout(s)
o
147
(40)
(41 )
(42)
(43)
(44)
Transmission Line Output Voltage: Series Terminated Line
Solving for Eout(s):
(45)
which reduces to:
(46)
Taking the inverse LaPlace transform yields:
(47)
Equation 47 defines the output voltage for a series terminated transmission
line when R = Zo ; it defines the output voltage for a parallel terminated trans-
mission line when R = Zo/2.
and:
From equation 47 the equation for the series terminated line can be written:
148
E'
I
+ - (t)
TI
-t
ZoC
T
e + E'
1
for t < T 1 '
(48)
for t > T 1 ' (49)
Voltage Output from Parallel Terminated Line; Line Delays
where Ei is defined in Figure 7-24. The equations for the parallel terminated line
can also be written:
( -2t )
_ Ei ZoCT ZoC
T
E'
+ _1 (t)
for t < T 1 . (50)
eout(t) - 2T1 e - 1
T1
and:
( 2T])
-2t
E'l ZoC
T
---
eout(t)
ZoC
T
ZoC
T
E'
for t > T 1 . (51 )
2TI
I - e e +
I
If the input voltage is assumed to be a step function, then the equation for the
output voltage for a series terminated line can be written as:
(52)
and for a parallel terminated line:
(53)
To derive the equation for the additional propagation delay due to gate loading
at the end of the line, equations 52 and 53 will be used. A more exact equation can
be derived using equations 48 through 51 but the analysis is more difficult due to
the complexity of the equations. Letting eout(t) = 0.5 Fi and solving for t, we
obtain a propagation delay time of:
0.7 ZoC
T
for series termination, (54)
and:
tpd = 0.35 ZoC
T
for parallel termination. (55)
These additional line delays should be added to the existing physical line delay
as expressed in equation 34 to obtain total system line delay. To derive the equation
149
Terminated Lines: Output Rise Time
for the output rise time, t
ro
, knowing the input rise time, tri, equations 49 and 51
will be used. The output rise time is defined as the time it takes for the output
voltage to travel from 10 to 90% of its final value. T 1 is defined as:
(56)
Substituting into Equation 49, rearranging, and taking the natural log of both sides,
the output rise time at the end of the transmission line is obtained:
for series termination. (57)
Doing the same thing with equation 51 :
for parallel termination. (58)
Equations 56 and 57 may also be used to solve for the output fall time by
substituting the input fall time, tfi, in place of trio
Example. An example will be shown to illustrate the use of equations 54 through
58 using MECL 10,000 gates. Figure 7-26(a) and (b) shows comparable setups for
series and parallel termination. The load gates at point B were placed in sockets. The
total capacitance at point B is CT = 20 pF which takes into account the gate
capacitances, socket capacitance, as well as interconnect and stray capacitances. The
propagation delay due to the transmission line from A to B is 1.75 ns. The rise and
fall times at point A, due to the MECL 10,000 driving gate, are tri = 3.5 ns, and
tfi = 2.8 ns. From equation 54, the propagation delay increase due to the series
termination is: (0.7) (50 ohms) (20 pF) = 0.7 ns. From equation 55, the propa-
gation delay due to the parallel terminations is half as much, 0.35 ns. Therefore, the
total propagation delay from point A to B is 2.45 ns for series termination, and
2. I ns for parallel termination.
The rise and fall times at point B can be calculated from equations 56, 57, and
58. The mathematical 10 to 90% rise time at point B is 4.65 ns for series
termination, versus 4.1 ns for parallel termination. The fall time at point B is 4 ns
for series termination versl!s 3.35 ns for parallel termination. Measured test results
agreed closely with the calculated values.
Equation II could have been used to calculate the propagation delay from A to
B for the parallel termination arrangement. Solving for tpd:
tpd = 1.75 ns
20 pF
+---
35 pF
2.211s ,
which is very close to the value of 2. I ns that was calculated using equation 54.
150
Hxample: Comparison of Series and Parallel Terminations
7-26: Test Setups for Comparison of Propagation Delays from
A to B. and the Rise and Fall Times at B
Since the propagation delay increase is twice as much for a series terminated
line as for a parallel terminated line, an equation similar to equation 11 can be
derived. The propagation delay of a series terminated line can be written as:
(59)
where tpd is the modified propagation delay of the line, tpd is the original
propagation delay of the line, CT is the total capacitance at the end of the line, and
Co is the intrinsic capacitance or the transmission line.
The effective characteristic impedance of a transmission line decreases with
capacitance at the end of the series terminated line, but only half as much as for a
parallel terminated line (cf equation 32). The characteristic impedance due to
151
Determining the Emitter Pulldown Resistor Value
loading of a series terminated transmission line can be written:
Z'
o
R
T
1+-- +
Co
= RS + Ro (60)
This means that the series terminating resistor should be changed in the proportion
indicated by equation 60 when the capacitance at the end of the line exceeds the
value of capacitance given in Table 7-23. Equation 60 has been verified in the
laboratory as valid for heavy loading conditions. If the resistor, RS,is not altered
according to equation 60, increased propagation delays will result from under-
damping.
Furthermore, if the amount of loading at the end of a series terminated line is
more than that shown in Table 7-23, then the emitter pulldown resistor, RE, should
be lower than the value given in equation 61 for the maximum pulldown resistor
value (Chapter 3). The maximum value of RE as stated in Chapter 3 is:
RE(max)
n
(61 )
The reason that the pulldown resistor should be lowered is that the reflection
returning to the source contains a short positive component associated with the
slower fall rate of the negative-going signal (due to capacitance loading). Thus,
the output transistor will turn off due to this momentary positive reflection if a
value given by equation 61 is used. The value of RE should be chosen so that the
output transistor of the driving gate is furnishing enough current to supply the
maximum reflection without switching off. The maximum reflection on a series
terminated line, due to capacitance, is +0.4 volts. Therefore, the value of RE should
be reduced enough so that the output transistor will be able to supply an additional
current of O.4/Z
o
. A modified emitter pulldown resistor equation can be written as:
RE(max)
3.6
(62)
_
......:3-'-.6=-- 0.4'
+--
RE max Zo
where RE (max) is defined in equation 61.
There is no limit (due to reflections) on the amount of gate loading that can be
placed at the end of a series terminated line, as long as equation 60 is used to
determine the proper series terminating resistor. All reflections will be terminated at
the driving end if the proper value of RE is chosen with equation 61 or 62. This is
true even though the amount of reflection is twice as much for a series terminated
line as it is for a parallel terminated one.
152
Series Damped Lines
The maximum loading for a parallel terminated transmission line is defined in
Figure 7-23. There is no limit to the amount of distributed loading that can be
placed on the line, as long as equation 32 is used to choose the terminating resistor.
In actual practuce, there seems to be no limit to the amount of lumped loading that
is placed at the end of a parallel terminated line as long as the terminating resistor is
chosen this way. This is true as long as there are no other gates stubbed off the line.
For reference purposes, the equations for the reflection that is sent back to the
driving source, with lumped loading at the end of either a series terminated line or a
parallel terminated line, can be derived from equations 48 and 50 as follows:
eout (1) - E'l
percent of maximum reflection = E'
1
at t = T 1 . (63)
Therefore, the maximum reflection due to lumped loading at the end of a series
terminated line is:
percent of maximum reflection (64)
and for lumped loading at the end of a parallel terminated line:
percent of maximum reflection
ZoCT (e ~ ~ -I)
2Tl
(65)
Analysis of Series Damping Terminations
A series damped line is very similar to a series terminated line with the ex-
ceptions being the line length and the value of the series damping resistor, RS. The
resistor is normally much smaller than the characteristic impedance of the line, Zoo
If RS, = 0, then an open line exists for which the maximum length is defined in
Figures 3-13, 3-14, or 3-15. If a small value resistor, RS, is placed in the line, a
longer line length is possible. An example of series damping is shown in Figure 7-27
7-27: Series Damping Termination
VA
VB
Zo
Vc
RS
~ Q m a x ~
m,
Number
of
RE
Gates
Where RS<Zo
VEE
153
Output Voltage from a Series Damped Line
where the voltage change at point B is defined in Chapter 3 as,
(66)
Series damping is primarily used where the characteristic impedance varies - as
in backplane wiring when line lengths must be longer than those specified in Figures
3-13 through 3-15. A disadvantage of series damping is that distributed loading
cannot be used. A propagation delay slightly slower than for parallel termination also
results. Parallel fanout can be used as shown in Figure 3-18.
Figures 3-22, and 3-23, described in Chapter 3, show the minimum values for
RS for any line length, corresponding to specified limits of undershoot and
overshoot. These figures were generated by a computer program based on the
equations and calculations presented in the following pages. These calculations show
how the output voltage from series-damped transmission lines may be derived.
The Thevenin equivalent circuit for a series damped transmission line is shown
in Figure 7-28. Note that the circuit is similar to the equivalent circuit for a series
terminated line except for the amplitude of the voltage waveform.
728: Thevenin Equivalent of a Series Damped Transmission Line
r ~
I I
I Zo I
I
I
I
I
I
I
I = I
L _______________
E'i
Slope = m =-
T1
2ZoE1
Where E'i = --
Ro+Rs+Zo
C
T
= Total Load Capacitance
The equations for the output voltage from a series terminated line were derived
previously (equations 48 and 49). Since the amplitude has been modified by the
factor (2Zo/Ro + RS + Zo), a simple substitution will define the output voltage at
the load. The output voltage for a series damped line can be written as:
for t < T I '
(67)
154
Pro[Xlgation Delay; Exact Output Voltage Derivation
and:
for t > T 1 '
(68)
where El is the voltage change at the base of the output emitter follower in the
driving gate.
If the input voltage is assumed to be a step function, the equation for the
output voltage from a series damped line becomes:
(69)
The additional propagation delay due to gate loading can be found by using
equation 69, letting eout(t) = 0.5 El, and then solving for t. The increase in the
line propagation delay which results, is:
(70)
Thc cxact reflected voltage will now be derived by a method similar to that
used previously in deriving equation 19. The reflection coefficient at the load is:
sCT
Zo
- sCTZ
o
(s - Z o ~ t
P
L
(s)
1 + sCTZ
o
--+
Zo
s +
sCT ZoC
T
(71 )
The input voltage at point B in Figure 7-27 is:
(72)
where:
m
155
Reflection Amplitudes (La Place Notation)
Here Ro is the output impedance of the driving gate; RS is the series damping
resistor; Zo is the characteristic impedance of the transmission line; E 1 is the voltage
change at the base of the output emitter follower of the gate at point A in Figure
7-27; and T1 = 1.2 tr where tr is the 10 to 90% rise time of the voltage at point A.
Taking the LaPlace transform of equation 72 gives:
(73)
The first reflected voltage waveform at the load at time TD (point C in Figure
7-27) due to the input voltage is (in LaPlace notation):
E
refl
I (s)
(74)
The second reflected voltage waveform at the load occurs a time 2TD later and can
be written as:
(s - _1 ) 2 P
s
m (1 - e - TIS) , (75) E
refl
2 (s) E
refl
1 (s) P
s
P
L
(s)
ZoC
T
= =
s2 (s + _1 ) 2
ZoCT
where:
P
s
Ro + RS -
Zo
(76) =
Ro + RS + Zo
The third reflected voltage waveform at the load is:
(77)
156
Output Voltage Amplitudes (La Place Notation)
So the nth reflected voltage waveform at the load is:
E () E P n - 1 p
L
n
- 1 (s)
refl (n) s = refl 1 S
(_ 1 )n (s _ _1_) n P n - 1 m
ZoC
T
S
(78)
The output voltage at the end of the line can be derived by using equations 73
through 78. The output voltage, Eol, due to the input voltage waveform and the
first reflection is:
(79)
It should be noted that equations 79 and 46 are equal (though they were
derived differently), when R = ZOo At first glance it may appear that equation 79 is
twice the value of equation 46, but the apparent discrepancy is resolved by noting
that m has been defined differently for each equation.
The output voltage, E02, due to the first reflected voltage waveform returning
from the driving source will be:
(80)
Similarly, the output voltage, E03, due to the second reflected voltage
waveform returning from the driving source is:
157
nth Output Voltage and Total Output (La Place Notation)
or:
U (t - 4T D)
(81 )
Finally, the nth output voltage waveform will be:
Eon (s)
(_ 1)n - I
(82)
therefore, the general equation for the output voltage at the end of the line (point C
in Figure 7-27) can be formulated as a summation of the individual reflected voltage
components:
Eout (s) = Eo 1 (s) + E02 (s) + E03 (s) + ... + Eon (s) + ...
00
t(_l)n 1 - 1
(s - n
- I
( -TIS)
ZoC
T
S
1 - e
L
s2 (s
I )n
n = I
+ ZoC
T
U
(t 2( n - 1)T D) 1
(83)
The inverse LaPlace transforms for equations 79. 80, and 81 can be found in
standard tables. The inverses can be written to take into account the time delays the
following way.
158
Inversion of the laPlace Voltage Equations
The inverse LaPlace transform of equation 79 is:
eol (t)" I
U(t)
+
- )
(84)
Likewise, the inverse transform of equation 80 becomes:
U(t - 2T D - T 1) .
(85)
159
Result: Output Voltages as Functions of Time
Finally, the inverse transform of equation 81 produces:
e
0
3 (t)
2mPS2ZoCr [-5 +
t - 4TD
(5
4(t - 4T D)
=
+ + +
ZoC
T
ZoC
T
2(t - 4T
D
)2)
- (t - 4T n) ]
ZoC
T
2
e . U(t - 4T D) -
(ZoC
T
)
+
- (t
e (86)
A general form equation can be established for eo (n)(t) as an extension of the
preceeding equations. Such a general equation can then be used to ohtain a general
relationship for eout(t) derived from equation 83.
However, for our purposes, the output voltage needs to be defined only for a
period of time long enough to determine the maximum amount of overshoot and
undershoot. The following can be written for the first three retlections (a sufficient
time interval):
3
e
out
(t)
=L
n
- 1
( - 2T D (n - 1)
1)-
2mp
S
ZoC
T Z C - (2n -
o T
n = 1
( -
t -
2T D (n -
1))
C
- (t - 2TD (n
I))}
+
e
ZoC
T
- l)e
ZoC
T
+ (n - ---- U t - 2(n - l)TD\
- 1)) 2)('t - 2TD(n - 1))2) (
ZoC
T

160
Output Voltage for Three Reflections
3
-L:
- TI \
- (2n - l)J-
n =
(2 (I
2T D (n - I) -
T
1
)
('
2TD (n - I)
.. I] r)]
+ (n - 2)
-
ZoC
T
ZoC
T
U (t - 2(n - 1)T 0 - T I) for t < 6T
D
'
(87)
where:
m
and tr is the 10 to 90% rise time of the voltage at point A in Figure 7-27.
An extension of equation 87 was used in generating Tables 3-22 and 3-23 and
may also be used to determine maximum line length for specified undershoot and
overshoot, instead of using equations 12 and 13.
For C
T
= 0 and t < 6T D' equation 87 reduces to:
eout(t) = 2 mt U(t) - 2m (t - T
I
) U (t - T
I
) + 2mPS'
(t - 2T
D
) U (t - 2TO) - 2mPS (t - 2TD - T
t
) U (t - 2TD - T
t
) +
2mPS2 (t - 4T
D
) u (t - 4TO) - 2mP
S
2 (t - 4TO - T
I
).
(88)
161
Maximum Overshoot
This equation can also be derived by starting from equation 5.
Using a lattice diagram, it is found that the maximum overshoot occurs at
t = TI. Substituting t = TI into equation 88 gives:
(89)
for 2T D < T I < 4 T D ;
and:
2mTl '
for Tl < 2T
D

(90)
In both cases,
ZoEI
m
(R
o
+ Zo +
Rs)TI
,
TI
1.2t
r
,
TD
=
tpd Q,
and tpd line delay in nanoseconds/inch,
Q = line length in inches.
By definition,
(100 + O.S.)
El 100
(91 )
where E 1 is the voltage change at the output of the driving gate, and O.S. is the
percent overshoot based on logic swing level. Thus by substituting into equation 89,
the percent overshoot can be obtained:
162
Percent Overshoot; Line Length Considerations
t
2 Zo
%O.S. ::: -1 + -Zo--R---
+ S + Ro
for 2tpd < T I < 4tpd .
(92)
Equation 92 gives the overshoot (as a percentage of the logic swing) that occurs
for a particular length of line, assuming zero capacitance at the end of the line. If the
two way propagation delay of the line is equal to or greater than T I, then this
particular length of line is:
(93)
If equation 93 is satisfied, then the overshoot reaches a maximum value which can
be solved for by using equation 90:
%o.s. max
(94)
If the length of line is less than that specified in equation 93, the line length
can be found from equation 92, given the permissible overshoot for a given design:
Q
[
2 Zo (l + P ) _ (100 + O.S.) ]
Ro + Zo + RS S 100
(95)
In this relation, it is necessary that:
163
Maximum Line Lengths
The maximum permissible line length with capacitance loading can be
approximated in a manner similar to that llsed to establish a maximum open line
length in a previolls section:
tpdVl
C
T
+
Q CoQmax
(96)
Qmax tpd
Solving for Qmax:
2
(97)
where CT is the total lumped capacitance in pF, Co is the intrinsic line capacitance
in pF jin, and Q is the length of line in inches defined in equation 95.
Using a lattice diagram, it has been found that the maximum undershoot occurs
when t = 2TD + TI. Substituting this information into equation 88 produces the
relation:
for:
(99)
By definition,
E (100 - U.S.)
I 100
(100)
where U.S. IS the percent undershoot. Equation 98 can be used to form a useful
164
Percent Undershoot
relationship, which expresses the undershoot in tenns of circuit parameters:
%v.S. +
4PS2 tpd Q Zo )
I 2 t (Z + R + R) 100
. r 0 S 0
(10 I)
for:
and:
Thus equation 10 1 gives the undershoot as a percentage of the logic swing for a
particular length of line.
If equation 93 is satisfied, then the undershoot reaches a maximum value which
can be found from equation 99 to be:
(
2 Zo )
%V,S'
max
= 100 I - R R (1 + P
S
) .
Zo + S + 0
(102)
On the other hand, if the length of line is less than specified in equation 93, the
line length can be found from equation 101 once the permissible undershoot has
been specified. Solving:
Q [ 2 Zo (1 + P + P 2) _ (100 - u.S.)) .
-\Zo + RS + Ro S S 100
(103)
165
Line Lengths for Specified Undershoot or Overshoot
for:
Equations 95 and 97 can be used to find the maximum open line length
(instead of equations 12 and 13) when the maximum percentage overshoot has been
specified. Equations 97 and 103 can be used when the maximum undershoot is
specified. Of course, a more exact value for the maximum permissible line length is
found when a computer program is used to generate the values using an extension of
equation 87. This was done to generate Figures 3-22 and 3-23. The first seven
reflections (n = 7) were used to accurately define the series damping resistor re-
quired to limit overshoot and undershoot for a wide range of load capacitances.
REFERENCES
1. Kaupp, H. R., "Characteristics of Microstrip Transmission Lines", IEEE
Transactions on Electronic Computers, Vol. EC-16, No.2, April 1967,
pp. 185-193.
2. Cohn, S. B., "Characteristic Impedance of the Shielded Strip Transmission
Line", Transactions IRE, Vol. MTT-2, July 1954, pp. 52-57.
3. Springfield, W. K., "Designing Transmission Lines into Multilayer Circuit
Boards", Electronics, November 1, 1965, pp. 90-96.
4. Skilling, H. H., "Electric Transmission Lines", New York, McGraw-Hill, 1951.
5. DeFalco, J. A., "Reflections and Crosstalk in Logic Circuit Interconnections",
IEEE Spectrum, July 1970, pp. 44-50.
6. Millman, J., and Taub, H., "Pulse, Digital and Switching Waveforms", New
York, McGraw-Hill, 1965, pp. 83-106.
7. "Time Domain Reflectometry", Hewlett-Packard Application Note 62, 1964.
8. Botos, Bob, "Nanosecond Pulse Handling Techniques in I/C Interconnections",
Motorola Application Note AN-270.
9. Schwarzmann, A., "Microstrip Plus Equations Adds Up to Fast Designs",
Electronics, October 2, 1967, pp. 109-1 12.
10. Catt, 1., "Crosstalk (Noise) in Digital o;;ystems", IEEE Transactions on
Electronic Computers, Vol. EC-16, No.6, December 1967, pp. 743-763.
11. Feller, A., H. R. Kaupp, J. J. Digiacoma, "Crosstalk and Reflections in
High-Speed Digital Systems", Proceedings, Fall Joint Computer Conference,
1965, pp. 511-525.
12. Cohn, S. B., "Shielded Coupled-Strip Transmission Line", IRE Transactions-
Microwave Theory and Techniques, October 1955, pp. 29-38.
13. Gabbard, O. G., "High-Speed Digital Logic for Satellite Communications",
Electro-Technology, April 1969, pp. 59-65.
14. Henschen, I. E. and E. M. Reyner II, "Adapting PC Connectors for Impedance
Matching", Proceedings, NEPCON, 1970.
166
MECL Applications
This chapter presents design ideas for using MECL circuits. The majority of
circuits shown here use MECL II parts, as this family was introduced before the
MECL III and MECL 10,000 families. However most of the MECL II circuits may be
built with MECL 10,000 or MECL III parts, giving a corresponding increase in speed.
Conversely, the MECL III designs can use MECL II or MECL 10,000 devices to save
power, when the speed is not required.
Some of the MECL III circuits do not show pulldown resistors. These circuits
are intended only to illustrate the logic connections and either pulldown resistors or
line termination resistors must be incorporated in these designs.
The intent of this chapter is to present various circuits to the MECL user
together with a description of their use and general performance. No attempt is
made to explain the operation of each circuit completely. Since Motorola is
continually expanding the MECL lines, it is expected that some of the following
circuits may be simplified by future complex functions.
Counters
High speed counters are an important part of many system designs. Since
MECL flip-flops are the industry's fastest, MECL counters find widespread use in
high speed computer, communication and instrumentation systems. Figure 8-1
shows the fastest MECL II decade counter. MClO27s are used in the divide-by five
section because the multiple T and K inputs allow use of the BCD code without an
Clock
Input
Frequency
Q
A
MC1034
RT
o Q
81: 100 MHz MECL II Decade Counter
J
B
MC1027
Q
167
J
C
MC1027
R
Q
o
MC1027
Q
0\
oc
Reset '"'-
r-----l
I I I I I
I R I r V 1 R; 1
D Q....---,----< " EE D
I I I
I 510 I I
I nput 1..--, I .----, I
I
1 -:: e Q 2
0
I 1 ) e Q 21
VEE Q--JIIV\
510
S I I rt-------" S I I
I I I I I I -oVEE
r: MC10131 I 510 I MC10131 I 510
I
1 I V:> I I I
Q EE I S Q 22
r--L/ 113 1-1--/ I I
I I 6
2
I I 510
1 0 Q I 510 1 0 Q I
I R I I R
L __ I __ :J L __
VEE
r:x:
I
1/4 Me 101 02
8-2: BCD MECL 10,000 Decade Counter
510
V\f\r--() VEE
o
:;::
;:s
....
""
A
Reset
1
s
-D Q-
..........
A
O--C
MC1670
Oi--
R
1
Counters
8-3: 300 MHz MECll1i Decade Counter
(No Additional Gates Required)
B
1
s s
-D
Q D
B
C
C
MC1670 MC1670
0 -C
R R
1
OUTPUTS
A B C D STATE
1 0 1 1 0
0 0 1 1 1
1 0 0 1 2
0 0 0 1 3
1 1 0 0 4
0 1 0 0 5
1 1 1 0 6
0 1 1 0 7
1 1 1 1 8
0 1 1 1 9
1 0 1 1 0
C D
I
S
Q
D Qr---I-
D
r- C
MC1670
Or--- O-
R
1
additional gate. Package count can be reduced by going to the MC 10 131 dual
flip-flops as shown in Figure 8-2. The gate (1 /4-MCI 01 02) is required for BCD
coding. Typical input clock speed for this circuit is greater than ISO MHz. When
higher speeds arc needed, the MECL III decade counter shown in Figure 8-3 may be
used. The counter does not operate in a BCD sequence, but the 60/40% duty cycle
output may be a better signal if a slower logic family follows the counter. The speed
of this circuit is more than 300 MHz.
Figure 8-4 shows circuits for using MECL D flip-flops to count from 2 through
10. It should be noted that some counts arc not in a natural binary sequence and
some circuits usc the MECL Wired-OR feature. The divide-by-seven arrangement of
Figure 8-4 may be simplified by adding a gate in place of the extra t1ip-flop as shown
in Figure 8-5.
Synchronous counters have a speed advantage when decoding is required or
when propagation delay of the total counter is important. The multiple inputs of the
MCIOl3 and MCI027 flip-flops may be used for synchronous counting as illustrated
in Figure 8-6. Figure 8-7 shows how the counter may be expanded to 8 stages by
using MECL gates in addition to flip-flops. Gate delays to the J and K inputs must
be considered when computing the circuit's top operating speed. Even longer
(Continued on Page 174)
169
Counters
8-4: Dividers Using MECL D Master-Slave Flip-Flops
~ 75
o9:$oCtFTJ
o
o
NOTE: Pulldown Resistors Not Shown
170
Counters
8-5: High Speed Divide-By-Seven Counter
MC1023
S R
C
A B o Output
f+--------j C 1
C2
Clock Input
-5.2 Vdc
Clock
-5.2 Vdc
,-J
K

.-R

.- K
S R
A
MC1027
T K

1 2 4
STATE A B C
1 1 0 0
2 0 1 0
3 1 1 0
4 0 0 1
5 1 0 1
6 0 1 1
7 1 1 1
0-1
8-6: MECL II Divide-By-16 Synchronous Counter
S R S R

S R

.-J
K Q K
Q . K
r-J
B
>--- J
C

0
K
MC1027
'-----
K
MC1027
'---
K
MC1027
0

Q
,-1 0 J

J R
K
J K
K
J K
of---o

171
...-
-.l
N
-5.2 Vd
Clock
c
~
-=L
I I I r-'
Q I J K
D-[
Q I J K
r ~
I J I
MC1032 MC1032
K
I
Q
K I Q
C
I
C C C
1 1 I
>-
r > ~
r 1
I--J
Q I J K
~
JQ I J K
I I Q
MC1032
0.
MC1032
-
K C
I - I -
I
C
K C I C Q
1 1 1 I
8-7: MECL II Divide-By-256 Synchronous Counter
>
I '0""
ock
56
Q
l::
;::

~
o 10
UI--
o U
I
lli
o 10
T
u ~
o U
I
o 10
~ ~
Uf--4
o U
~
>--
o 10
H
Uf----<
o U
~
Q
-'"
()
E.
U
Counters
o 10
U I--
o U
I
o 10
uf----<
o U
I
o 10
U ---<
o U
I
o 10
uf----.
o U
1
Q
173
...
:J
~ o
11
I
o 10
o u
LJ
o 10
o U
I
o 10
o U
I
U f--
U ---4
uf--
Q
..J
CJ
w
2
.,
0>
'"
...
C/l
Clock
D
Reset
Up-Down
Up = True
c s
A
Q
Q
R
Counters
8-9: MECL Synchronous Excess Three Up-Down Counter
c S c s
Q Q
C B C C
c S
D
Q
Q
R
D
synchronous counters are possible by cascading the count sequence to the other C
input, as shown in Figure 8-8. The count may be increased by continuing the
sequence, but the additional series-gate delays must be considered when determining
top speed operation. Circuits for non-binary count sequences are easily designed
with MECL type D flip-flops, because of the extra C input. Examples of such
counters are shown in Figures 8-9, 8-10, and 8-11.
MECL programmable counters are used in high speed phase-locked loop
applications because the higher speeds of such counters eliminate much of the
prescaling required by slower counters. Figure 8-12 is a typical two decade MECL
programmable counter. Early decoding is used to minimize the reset time and speed
up the circuit. A typical BCD decade down counter for the circuit is shown in Figure
8-13. A three decade MECL II programmable counter is shown in Figure 8-14. Worst
case top speed for the circuit is over 50 MHz.
Higher speed programmable counters are possible with MECL HI circuits.
Figure 8-15 shows a counter using the MC1678L decade counter. The circuit, as
shown, has a worst case toggle frequency of 125 MHz and typically runs at more
than 150 MHz. Since the MC1678 counts forward, a nine's complement code is
required on the inputs.
Figures 8-16, 8-17, and 8-18 illustrate typical MECL ring counters. Ring
counters are used for ease of decoding and synchronous operation. The counter in
Figure 8-18 requires more flip-flops (10), but gives a direct one-of-ten output. These
circuits may be of any length needed to give multiphase clocks.
MECL circuit designs with discrete components have been used to obtain even
faster performance than with rcs. Figure 8-19 shows a very high speed divide-by-two
circuit. Speeds greater than 800 MHz have been achieved using the MMT3960
transistors and over I GHz using the MMT 8015 transistors.
(Continued on Page 182)
174
-.l
VI
Ic
~
Clock
Carry In
I
,------L
L--
0
C
Clock
'\.....
C
I
51
0
Qr-
C
1
0 Or----' 1 Yo 01 1 10
Ic 51
,.
Ic 51
~
C C
I I
8-10: Synchronous Divide-By-9 (3x3) Counter
D--
.----.
~ r-
---L)
~
~
-0 0 '--0 0 '--
0
'--- C 5- '---C
0- -c
C C C
1 1 I
8-11: Synchronous Divide-By-10 Counter
~ l
5r--o Clock 79
Car ry Out
1
7
"
0
....
5
....
9
:;:::
;::

fTLJ!J l :11
11 [I
J J.i
iR' 0 'I
-'---' ftl.f.L _".J L.J
S1 S2 S4 S8
BCD I
BCD Decade
I
fin C2 Down Counter R
ric,
Down Counter R I
g
-.l (cf Figure 813)
(cf Figure 813)
l::
0\
;:::
01 02 04 08 01 02 04 08
....

S R
D 01 Of out
8-12: BCD Programmable Down Counter (N = 1 to 99)
Flip-Flops are MC1034 or MC1670
--:)
--:)
Los
R S
Ql t-- ---< -0
-5.2 Vdc 0-- Cl 01 r--< Cl
C2
Clock
1
510 510
(
-5.2 Vdc

-l -
1
r-J
rr
-
R S R S R
Q2 -----< '-- 0 Q4 t-- -
'--0 Q8
02 1---1 Cl 04 r--< Cl 58
C2 C2 C2
1
510 510 510 510 510 510
)
)
-5.2 Vdc -5.2 Vdc -5.2 Vdc
8-13: BCD Decade Down Counter
I C,oo>+"

!:::
;::s

-..J
00

BCD
Units
:
4
8

BCD
Tens

BCD
Hundreds
: :
;: I-H<
8-14: MECL II 50 MHz + N Counter
Complementary
2:1 : Output
LJ
MC10n
NOTES,
1. Pin 7 = V (VEE)
2. Pin 14"" Gnd (Vee)
Output
Freq.
3. All unused inputs -5.2 V
4. AU unused resets -5.2 V
5. All unused outputs - open
6. Programmable from

$:;:


--l
\0
y. MC1663
y. MC1662
(8 Places)
Y,MC1660
Frequency In 0 I, I + Y,MC1660
so
C
R
00
Sl S2
MC1678L
(2 Places)
S3
53
so
C
Sl S2
MC1678L
S3
03
,---------------+-0 Frequency 0 ut 1
o
S R 01 1 10 S
R a IHIf-------'
MC1670
5 r l------'
C2
MC1670
C1
51 + o FrequencyOut2
8-15: High Speed MECL III BCD Programmable Counter
g
s::
;:::

00
o
'-
MC1034 MC1034 MC1034 MC1034
D
0p-aD QfqD Q ~ "
a
RT
RT ~ T RT ~ T
RT ~ T C
RT
O u t ~ Q
C C
MC1023
In 0>------'
8-16: MECL II Decade Ring Counter
r-K O-K ar- K
o-K
Ol---K
oj
K MC1027 MC1027 MC1027 MC1027 MC1027
f-J Ci-] at---- J Or---- ] Q f-q-- J oe--
C C C C C
I 1 1 I
1/2 MC1023
VEE
~
~
- -
...,
. - _0
p
8-17: MECL 11100 MHz Decade Ring Counter
MC1034
D a
RT
cRT
Q
Typical Characteristics (T A = +25
0
C):
Toggle Frequency = 100 MHz
Propagation Delay/Per Stage = 4.5 ns
~
;:::
;::
~
I
,
Input
......
00
J t
L MC,""OI
R Q - L C
K - I
R Q K Q
Reset
5 - MC1032 Dual JK Flip-Flops
Can Be Used
STATE A B
0 1 0
1 0 1
2 a 0
3 0 0
4 0 0
5 a 0
6 0 0
7 0 0
8 0 0
9 0 0
C D
0 0
0 0
1 0
0 1
0 0
0 0
0 0
0 0
0 0
0 0

B
C


F
G
H o------J
OUTPUTS
E F
0 0
0 0
0 0
0 0
1 0
0 1
0 0
0 0
0 0
0 0
G
0
0
0
0
0
0
1
0
0
0
H L
0 0
0 0
0 0
0 0
0 0
0 0
0 0
1 0
0 1
0 0
8-18: Self Correcting Decade Ring Counter
M
0
0
0
0
0
0
0
0
0
1

I
1
OM
OM
g
:;;:
;:s

Counters
8-19: Functional High-Speed Binary Divider Circuit
Vee
15 15
150
Shift Registers
39
180
72
Output
510
Similar to counters, MECL shift registers make use of the fast flip-flops for high
performance_ These registers are often used as delay lines, with delays controlled to
nanosecond accuracy_ Figure 8-20 shows a register commonly used for serial-to-
parallel and parallel-to-serial conversions_ This circuit uses the additional gates on the
set and reset inputs to permit parallel data entry without previously clearing the
registeL
The shift register in Figure 8-21 takes advantage of the MCIOl41 to form a
16-bit shift register circuiL With a continuing signal on the clock input, the control
lines are used for shift right, shift left, parallel data entry, and inhibit operations_
The parallel entry mode is used in conjunction with the clock line for loading
information into the registeL
Figure 8-22 illustrates a universal register capable of both shift right and shift
left operations_ Additional gating may be added as shown in Figure 8-20 for strobed
parallel data inputs_
Adders
In many applications MECL adder circuits are fast enough to use a standard
ripple-technique addeL The speed of this type adder is dependent on the carry-in to
carry-out time of the adder, as well as the number of bits. When faster add times are
(Continued on Page 188)
182
1

Serial Input
C<l

w
Clock
Clock Inhibit
Parallel Strobe
Parallel Input
I __________________________________________
PO P 1 P2 P3 P4 P5 P6 P7
:I
]
,-
:J
,-
:J
,-
:I
,-
:J
,-
Parallel Output
50 MHz Using MC1022 Flip-Flops and MC1004!MC1010 Gates
150 MHz Using MC1034 Flip-Flops and MC1023 Gates
8-20: 8-Bit Shift Register
]
1-

S;
....
Serial

:J 1
Output

'"
....

'"
00
.j::.
r-
Clock
Control
SO S1
015 014 013 012
! 1 I I
so S1 03 02 01 DO
DR
S.R.4 L
MC10141
C 03 02 01 00
015 014 013 012
SO
- - DR
C
011 010 09 08 07 06 05 04
I I I I I 'I 'I 1
S1 03 02 01 DO SO S1 03 02 01 DO
S.R.3 L
- r- DR
S.R.2 L -
MC10141 MC10141
03 02 01 00 C 03 02 01 00
( ) )
011 010 09 08 07 06 05 04
S1 SO FUNCTION l
0 0 Enter Data
I
0 1 Shift Right
1 0 Shift Left
I
1 1 Stop/Hold
8-21: MECL 10,000 16-Bit Shift Register With End-Around Shift-Left and Shift-Right
SO
- DR
C
03 02 01 DO
If! f
S1 03 02 01 DO
S.R.1 L
MC10141
03 02 01 00
n
03 02 01 00
I
!
r-

S;
'"


0:;'

.....
00
V1
RA
SRA
ORA
SRA
C1A
C1S
LS


I
saaR
01

"p- Me'022
02
C1 C2
I
1
I
C2

02
MC1022
R Q a s D1r-<Jl
,
(
Parallel Inputs/Outputs
.J.

s--; saaR
01 01
L----
MC1022
p- Me'02'
?--->-02C1
02
C2 C1 C2
T I
1 1
I
1
e, d,
I ki-
C2 C1
02 02
Me'022

01
R Q a s R Q a s
y
Parallel Inputs/Outputs
8-22: MECL II Dual 4-Bit Universal Shift Register
,
a a R
p-D,Me'022
C1 C2
I
T
e
l
, K=2-
C2
02
Me'022 D'I-<J::i
R Q a s


LA
C2A
C2s
SRS
ORB
SRS
RS

S;
......

c
",,'

00
0\
A 1 91 A2 B2 A3 63 A4 B4
Add/Sub 0 I. I I, I I, I I, I 0 To Next Group
Add 1
Gin
C
A
MC1019
c
S1


Gin
vP2 PI



P4
B4
Cil
C A 6
MC1019
15
S2
G2 G3
S3
8-23: MECL II Add/Subtract With Lookahead Carry
C A 6
MC1019
C S Sec
r
S4
04
C)
L-..--'U--()151 To Next Group



'"
00
-...J
A1 B1 A2 B2
CinO T lei
MC1059
G = A. B
j = ~
A3 B3 A4 B4
C3
MC1059
8-24: Look-Ahead Carry Block With MEeL II Dual Adders
PG and GG are for
2nd level lookahead.
Cout
GG
PG
~
:
(1)
~
Adders
A1 61 A2 B2 A3 83 A4 84 A5 65 A6 B6 A7 B7 AS B8
j
? ?
_LLj_ j_l
? ?

-'"ii"
MC1059
CO ,-- C1
MC1059
CO
I"
Co
P1 G1 P2 G2 P3 63 P4 (;4 F5 (;5 F6 (;6 P7 Gl F8 G8
I I I I I I I I I I
G1 P2 G2 C P3 G3 G1 P2 G2 C P3 G3
C
in LC8
CO
Cin LCB
P1 P6 PG GG 1'4 P1
P
G GG G4 F4
-----..--J
L_ L-J

1
---
P1 P
G GG P2 G2
Gin Second Level Look-Ahead Carry Block
_
P3 G3 P4 G4
II

-I
r---l
P1 PG GG G4 P4
P1 PG GG G4 P4
Cin LCB CO -1--
ein
LC8 Co
G1 P2 (;2 C P3 G3 G1 P2 (32 C P3 83
I I I
L-"L G
1
" P12
I I I
L ::'}5 P16 G16 F9 G9 P10 G10 P13 G13 P14 G14
-
Cin MC1059 Co
Gin MC1059 Co

Cin MC1059
CO
Gin MC1059 CO
I I I I A,I, I A\2 I A14 I Ai5 I
A9 B9 A10 810
B11
812 813 814 815 B16
A17 A18 A19
A20 A21 A22 A23 A24
I
8r
I
BI8
,---L I BIO
B22
I
Bf3 I Br
Gin
MC1059
CO
r c," Me"" '"
MC1059 CO
I '," '.
P17
-
- - - - -
FT G
I
3 1'24
G17 P18 G18
T"r P20 G20
P21 G21 P22 G22
I I I
I
j
1
G1 P2 G2 C P3 G3 G1 P2 G2 C P3 G3
Cin LCB Co Gin LC8 Co
F1
G4 P4 P1 G4
"4
L..J L- L--.-J L.-
8-25: 24Bit LookAhead Carry Adder
required, a lookahead-carry circuit may be used. Figure 8-23 shows a MECL II four
bit adder block. The propagate function is derived with NOR gates and carry-out is
used for the generate function. Exclusive NOR circuits on the B inputs permit
adding or subtracting.
The generate (G) and propagate (P) outputs of the MC 1059 dual adder simplify
the lookahead carry block as shown in Figure 8-24. The lookahead carry block for
this circuit is similar to the MC 1 179 and the two are interchangeable. The
lookahead carry blocks can be cascaded to produce second levellookahead carry as
shown for the 24-bit adder circuit in Figure 8-25.
The MC 10 181 4-bit arithmetic unit and associated MC I 0179 lookahead carry
block from the MECL 10,000 series are used to cut package count significantly.
Figure 8-26 illustrates some adder circuits along with typical add times for the total
circuit. The typical time of 19 ns for adding two 32-bit words (using only 10
(Continued on Page 191 )
188
00
\0
Cin
MC10179 MC10179
8-26: Arithmetic Operations With MECL 10,000
12 Bits
Typ13ns
16 Bits
Typ 16 ns
Cout
"out
32 Bits
Typ 25 ns
32 Bits
Typ 19 ns


89
A14
Gin
-.
A6 813
A2
Ag
Al0
814
A1 65
810 A13
82
A5
86
Be
A15
80 All
812
Y II
yI
II'
?
B9
'
A ~
"f
L ~ eo A1 e1 A2 e2 A3 ~
~ i
AD 80 AlB 1 A2 62 A3 83 AD 80 Al 81 A2 82 A3 83 AD 80 AlB 1 A2 82 A3 83
Cn
en + 4
Cn
en + 4 D-O en en + 4 en en + 4 D-O
M M M ,-----c M
,----<
SO MC10181
Gp--- ,-----c
SO MC10181 GO-- ,---<: SO MC10181
Gp---
r-'
SO MC10181
GO--
r--< 51
4-BIT ARITHMETIC
r---<: 51
4-BIT ARITHMETIC
r--<
51
4-BIT ARITHMETIC
---< S1
4-BIT ARITHMETIC
LOGIC UNIT LOGIC UNIT LOGIC UNIT LOGIC UNIT
r<
S2 plo- ,-<J
S2 pO--
r-<:
S2 PP- ,--<
52 Pp-
r< S3
F1 F2 F3
rO
S3
F2 F3
r< S3
F3
-< S3
F3
FO FO F1 FO FI F2 FO FI F2
M
SO
S1
......
S2 \0
S3
0
:r...
~
FO FI F2 F3 F4 F5 F6 F7 Fa F9 FlO Fl1 F12 F13 F14 F15
I
1, ,I,
PO GO PI G1 P2 G2 P3 G3
MC10179
G :>--0
Cn
CARRY LOOKAHEAD
P :>--0
en + 2 en + 4
. b C15
- --
8-27: MECL 10,000 16-Bit Full Look-Ahead Carry Arithmetic Logic Unit
Adders
packages) illustrates the performance of MECL adders. Figure 8-27 shows the
interconnections for a complete 16-bit arithmetic circuit with second level
lookahead carry.
For specialized applications, three numbers may be added simultaneously using
a carry/save adder. Figure 8-28 illustrates this type of adder using MECL II circuits.
Both the sum and carry outputs from the first level adder are used as inputs for the
second level of addition.
828: MECL II 3 Number 4-Bit Carry/Save Adder
Xl Xl Yl Zl X2 X2 Y2 Z2 X3 X3 Y3 Z3 X4 X4 Y4 Z4
CAB
C MC1019 C MC1019 C MC1019 C MC1019
Al A2
-0.8 Vdc
MC1059 MC1059
81 82
81 82 S3 84 85
Code Converters
The complementary outputs and Wired-OR capabilities of the MECL circuits
are useful for improving performance and reducing circuit count in many code
converters. Figure 8-29 and 8-30 illustrate a high speed ripple BCD-to-binary
converter. The circuit may be extended to any length by adding the two basic
building blocks of Figure 8-30. Note the use of the Wired-OR connection in the
building blocks.
Figure 8-31 illustrates cascaded data selectors for selecting 1 of 32 lines. This
circuit also makes a good parallel-to-serial converter when the C inputs are
controlled by a counter chain and when parallel information is put on the input data
lines. The converse of this circuit is made possible by using decoder circuits which
give binary-to-one-of-N code conversion (or, with a countcr, a serial-to-parallel
conversion). A point to consider, when cascading data selectors or decoders with
MECL series gated parts, is the uncqual propagation delays of the differing inputs.
Since the lower levels of the series gate are slightly slower than the top level, a spike
may result on some of the outputs while switching the circuit. This spike is normally
much less than a gate delay in duration. It can be nullified, if necessary, by adding a
faster gate in series with the appropriate input leads or by sampling the outputs after
the longest delay-path time. The MECL 10,000 series multiplexers and decoders do
not use series gating and do not exhibit this output characteristic.
Figures 8-32 and 8-33 show conversions betwecn Gray and binary codes. The
Gray-to-binary coding uses both the Wired-OR and complementary gate outputs to
give a maximum of two gate delays for any signal path.
(Continued on Page 196)
19!
Code Converters
8-29: Ripple BCD to Binary Converter
Inputs
r ~ ~ A ~ ~ ~
2,000 800 200 80 20 8 2
Max Ripple Delay for 14-Bitsof BCD to 12 Binary Bits: 140 nsfor MECL II
192
Code Converters
8-30: MECL II BCD to Binary Converter Logic Blocks



193

",,,,,,,,,,,,-v80
"'XJ"'"""''''"'''''-QAO
)--.,.---0 DO
",'"'...,_-'L) CO
BO
'-0
...
C4
C3
C2
C1
co
LO
r
"
r.
r.
L 1 L3 L5 L7 L9 L 11 L 13 L 15 L17L19L21L23 L25 L27 L29 L31
( (
r
L4
T
L8
Lf Lf
T T T
Lr Lf Lr Lr Lr
L30
r r I
MC1038 f-- MC1038 - MC1038 - MC1038 r--
or t-- or - or
-
or f-
Q
MC10164

MC10164
1
MC10164
1
MC10164

i:l..
""
Q
;:s
"

..... I I

"" MC1028
MC10164
Data Output
8-31: 32 Channel Data Selector Array with Cascaded Data Selectors
Code Converters
8-32: MECL II Binary to Gray Code Converter
MC1030
80
GO=80 0 81
G1=81082
G2 = 82 0 83
...
I \ >-----0 G3 = 83 0 84

I
84(}- --'
Typical Propagation Delay = 5 ns
8-33: MECL II Gray to Binary Code Converter
Logic: MC1010, MC1030, MC1031

.------0 80 = (A 0 B) (C 0 D) +
)0---,
(A 0 8) (C Ef) D)
G1D--+-"
.------081 =8 (C 0 D)+
8 (C @ D)

82= C 0 D
Typical Propagation Delay = 10 ns
G3D----......
Where: A = GO
8 = G1
C = G2
D= G3
'--------------------<) 83 = D
195
Memories
Memories
While the cost and size of present MECL memories limit their practical use for
very large storage, the fast access and write times can speed-up system operation
when these circuits are used for scratch-pad or temporary storage memories. Figure
8-34 illustrates the use of the MC I 036 16-bit memory in such a circuit. The circuit
VCC
Y1
Y2
Y3
Y4


I
I
-
r-
'-- MC1036
,.--
-
1
S1 WO W1 SO
v
Bit 0
8-34: MECL II 16-Word by 4-Bit Memory
I
I
.---r--
'-- MC1036
r--
-
1
S1 WO W1 SO
v
Bit 1
I
I
-'-
-
MC1036
,.--
:--
1
S1 WO W1 SO
v
Bit 2

I
I
- -
'--
MC1036
r-
.-
I
S1 WO W1 SO
'----.v----'
Bit 3
X4
X3
X2
X1
VEE
can be expanded to any reasonable size by decoding the input with extra logic. The
larger and faster MECL 10,000 64-bit memory, the MC10140, may be used in
similar circuits for lower package count.
Although not designed as a memory, the gated inputs and strobed outputs of the
MCI040 or MClO133 quad latches allow this circuit to be used in many memory appli-
cations. An example of a memory using the MC llJ40 latches is shown in Figure 8-35.
The word/bit configuration is flexible, depending on how the input and outputs are
selected and how the outputs are wired together.
Small MECL III memories provide the ultimate in high speed performance.
Available as a random access memory (RAM), content addressable memory (CAM),
or both (CARAM), the circuits may be combined for several high speed memory
types. Figure 8-36 illustrates using both the CARAM and RAM memories in a very
high speed buffer memory. When a word is required from the main storage memory,
it is placed in the RAM portion of the buffer for future access. The word's address in
mass storage is placed in a content addressable memory (tied to the random access
section), thereby allowing words to be addressed by their mass storage location
during one cycle time of the buffer memory. As the address of the desired word is
(Continued on Page 199)
196
\0
-J
E
Q13 Q15
QO Ql Q2 Q3 Q4 Q5 Q6 Q7 Q8Q9Ql0Qll Q12 Q14
So
C2 C2 C2 C2
C21
I
(;2
Q4 Q3 Q2 Ql Q2 Ql Q4 Q3 Q4 Q3 Q2 Ql Q2 Ql Q4 Q3 Q2 Ql Q4 Q3 Q4 Q3 Q2 Ql
- r-- -
r-- -
Al A4 A5 A8 B2 B3
MC1040 MC1040 MC1040 MC1040 MC1040 MC1040
r-
0403 02 01
r--
02 01 0403
,---
0403 0201
r--
02 01 04 03 D2 01 0403
-
04 03 02 01
Cl
I
Cl Cl Cl
C11
I
Cl
j
EB
00 01 02 03 04 05 06 07 08 09 010011 012 014
013 015
8-35: Dual 8-Bit 2-Address Latch/32-Bit 6 ns MECL II Memory
Sc
C2
f-
Q2 Ql Q4 Q3
B6
MC1040
C-
02 01 0403
Cl
EC
(;2
Q4 Q3 Q2 Ql
-
B7
MC1040
-
04030201
Cl
So
f---'
r
(
ED


...,
(ii'
'"
\.Q
00
S1 S1 S1 S1 S1 S1 S1 S1
SO SO SO SO SO SO SO SO S a::
MO
Reado
1
MO
WO
MO
WO WO
4
M1
Read 0 iW1
2
M1
3
M1
M1t I It::;:::::'
MC1662
Match

Read
Devices 1 - 8 are MC 1684/1685
Devices 9 - 16 are MC1680/1681
W1 W1
M14
W14
7
M15
W15
W1
00 01 02 03 04 05 06 07
XO X 1 X2 X3 X4 X5 X6 X7
MC1662
XO X1 XO X1
R14
W14
13
R15
W15
00 01
10
14
00 01
8-36: CAM/RAM High Speed Buffer Memory (MECL III) 64-Word by 2-Bit RAM-16-Word by 8-Bit CAM
XO X1 XO X1
11 12
15 16
00 01 00 01


....

Oscillators
presented to the content addressable section, the CAM will indicate (in one cycle
time) whether or not the address is in the CAM, and whether or not the desired
word is available in the buffer. If the word is present, the desired read and/or write
function can be performed at buffer RAM speeds.
Oscillators
Oscillators are easily designed with MECL circuits. The simplest high speed
oscillators are formed by an odd number of NOR gates arranged to form a ring. The
oscillation frequency is determined by the gate propagation delay and the number of
gates. RC timing networks are used for slower oscillators as shown in Figures 8-37
and 8-38. Both of these circuits give complementary outputs and a good range of
oscillation frequencies.
837: MECL II Astable Multivibrator with Gates and RC Timing
100 pF
3k
MC1004
100 pF
8-38: MECL II Astable Multivibrator with JK FlipFlop and RC Timing
1 I I
R ~
J vCC S
R = 3 k OHMS R = 2 k OHMS
.-
JL
I>-- J Q
C f (kHz) c
f (kHz)
I>-- J
O.OlI1F 16.113 0.01 I1F 25.280
I>-- J
tr = 4.8 ns
0.001 I1F 150 0.001 I1F 233.23
MC1013
tf = 8.8 ns
I>--
K
360 pF 405 360 pF 621.85
K
I>---
100 pF 1361.3 100 pF 2025.3
24 pF 6481.8
I>--
K 6.
L..J
10 pF 10055.3
I>--
K VEE
R
I
I
TC ,
R
-5.2 Vdc
199
Oscillators
Crystal controlled oscillators are required when tighter frequency control is
needed than is practical with the simple RC oscillator. Because of high input
impedance and linear operation, any MECL NOR gate may be used in a simple
crystal oscillator circuit. However, line receiver circuits are normally used because
VBB is available for biasing. Figure 8-39 shows a typical crystal oscillator circuit
using the MC1035.
MECL crystal oscillator circuits are often used with TTL designs. By using a
resistor in series with the VCC line of a second gate, a MECL crystal oscillator can be
converted to a TTL-compatible output, as shown in Figure 8-40. The high speed
MECL III circuits are capable of operation beyond practical crystal frequencies. A
839: 10 MHz MECL II Crystal Oscillator with Schmitt Trigger
4-30 pF
1/3 MC1035
113 MC1035
200
300
Vcc = Gnd
0.1 j.lF
~
VEE = -5.2 Volts
=
VEE = 5.2 V 0-----..,
1.6 k
=
1/3 MC1035
8-40: MECL 10,000 Crystal Oscillator With TTL or MECL Output
5.1 k
600
600
-5.2 Vdc
+5.0 Vdc
MC10111
~ - - - - O TTL Output
MECL
Outputs
9
600
-5.2 Vdc
200
82
=
130
-5.2 Vdc
Oscillators
frequency doubler may be used with the crystal oscillator to generate higher
frequencies as shown in Figure 8-41. The Me 1662 is converted to two one-shot
multivibrators by using the proper length delay lines; Wire-ORing the one-shot
outputs gives the frequency multiplication.
In addition to oscillators, pulse generators are often required in system design
and test equipment. Figure 8-42 illustrates a circuit for constant frequency
C1
270
I
PF
_10k 1.8k
0.01 /L
F
I
-=
841: MECL III Crystal Oscillator with Frequency Doubler
Crystal
r----1Df------,
2 foutO---'"
% MC1692
%MC1662
%MC1662 Delay Line Equal
to % Wave Length
of Frequency
8-42: MECL II/MECL 10,000 Variable Pulse Width Generator
1 k
0.01 /LF
100 pF
1.8 k
1.2k
220
220
-5.2 Vdc
,-----+--+---0 - 5.2 V dc
5 k
*1/3 MC1035 or MC10116
**1/2 MC1023 or MC10109
100 pF
-5.2 Vdc
201
"---_--0-5.2 Vdc
-=
~ o Output
(-0.8 to -1.6 Vdc)
Oscillaturs
operation with variable pulse width. Figure 8-43 shows a generator with both
variable frequency and variable pulse width.
8-43: MECL II Pulse Generator: Variable Frequency and Pulse Width
600 300
Vcc
200
Frequency 5 to 13 MHz
Pulse Width 15 to 40 ns
200
2k
1/3 MC1035
1/4 MC1010
300 pF
r
-=
560
200
-VEE
200
One-Shot Multivibrators
The one-shot multivibrator circuits are used in a variety of pulse shaping
applications_ These circuits are used both as pulse stretchers, and as pulse generators
driven by input pulses differing in duration from the output pUlses. Most of the
slower circuits use RC timing to determine pulse width; but the use of delay lines
gives better results for very narrow pulses.
Figures 8-44, 8-45, and 8-46 illustrate three techniq ues for shaping pulses. If the
input pulse is narrower than the time constant setting, the output in Figure 8-46 will
8-44: Monostable Multivibrator: Constant Output Pulse Width
910
50 pF
510

,--L __
-0.7 V r-J r
-1.6 V....J L.J 0-----1
'----'V\I\r----Q - 5.2 V d c
202
One-Shot Multivibrators
8-45: MECL II One-Shot Multivibrator
-0.75 V r-l
-1.6 v---1 L
S Q
MC1014
~
r- R Or-
C
--, r--- -0.75 V
L----...J -1.55 V
If
1\
R
t2 occurs when VR reaches = -1.15 Vdc A at t1' VR = -2.35 Vdc
8-46: MECL II Multivibrator
Fa D MC1010
1 ~ g
Unused Inputs to -5.2 Vdc J: T
C""'220 pF for T= 40 ns
8-47: High Speed Multivibrators
203
One-Shot Multivibrators
follow the input. The one-shot in Figure 8-44 will accept an input of any pulse
width and give a constant output pulse width. Figure 8-45 must have an input
narrower than the output and be used as a pulse stretcher or the flip-flop will reach
an indeterminate state with both set and reset high. This condition can be eliminated
by connecting the input to a] input of an MCIOl3 or similar circuit. Use of the
MC1013 gives a constant pulse width out, regardless of input pulse width.
For high speed applications a delay line is used to set the time duration of
one-shot multivibrators as shown in Figure 8-47. In addition to commercially
available delay lines, lengths of coaxial cable and circuit board microstrip line are
commonly used for the delay path. A variation of the delay line monostable
multivibrator is shown in Figure 8-48. This circuit uses two delay lines of unequal
duration. Either pulse length, or no output pulse, may he selected with the S inputs.
This circuit may be expanded as necessary with additional delay lines and gates.
848: Gated Pulse Width OneShot Multivibrator

Input 0----/
02

01 if 51 = 0
02 if 52 = 0
o if 51 .52 = 1
849: MECL III Retriggerable OneShot Multivibrator
C1

220 pF
In 0---------1
v CC o---jf--_--I
0.01 J.1F
10 k R1

MC1692
r--------------oVEE
510
>-_--<JOut
510
510

Output: Low During Pulse Train
High No Pulse Train
Maximum One Shot Period = 550 ns
Minimum Input Pulse Width At 5 MHz = 5.0 ns
204
Linear Applications
The retriggerable one shot multivibrator is useful for many applications, for
example, to determine the presence or absence of pulse trains. A MECL III
retriggerable one shot is shown in Figure 8-49. This circuit recognizes a pulse train
having pulses spaced up to 550 ns apart. The minimum input pulse width is
dependent on the repetition rate of the incoming signal. The 550 ns delay time is
adjustable with the RC time constant (CI, R 1).
Linear Applications
The differential amplifier input of the basic MECL circuit allows the parts to be
used in many linear applications. Some devices are made in all families with both in-
puts of the differential amplifier available outside the package. These MECL circuits
are called Schmitt triggers or line receivers. Most circuits supply VBB on an
output pin for additional flexibility. Figure 8-50 and 8-51 illustrate two amplifiers
850: MECL II Amplifier/Schmitt Trigger
MC1035
I nput /'-----1
50
100
0.01 /IF
rr-----o VBB
=
8-51: MECL II Video Amplifier
MC1035
V BB 0---------1 """----I
Output
800mvlJ
-,-
L\;;,- 0 V /""'-----1 /"-------j
>-_---<l Output
850 mV (P-p)
5k 5k
5k
5 k 5 k

0.01 /IF I
=
Voltage Gain = 600 V IV.
3 dB Down = 40 MHz @ Gain = 424 V/V.
Single-Ended Gain = 5.0 V/V Per Stage@ BW = 40 MHz.
Double-Ended Gain = 10.0 V IV Per Stage @ BW = 40 MHz.
Negative Feedback or Fewer Stages May Be Employed
to Reduce Gain and Increase Bandwidth.
205
Linear Applications
using the MC 1035. The circuit in Figure 8-50 is useful for shaping and amplifying
low level signals up to MECL signal levels. The Schmitt trigger third stage assures
good edges. The bandwidth is about 40 MHz for the MCI035, but can be increased
by using faster MECL circuits. The MECL 10,000 MCIOll5 has 10 dB of gain per
stage at 100 MHz when used as an amplifier. The MECL III MC1692 is useful
beyond 300 MHz, although care must be taken in circuit layout to avoid oscillation
at these high frequencies.
Additional flexibility of MECL circuits is demonstrated by the MC 10 19 adder,
used as a balanced modulator as shown in Figure 8-52. The three-level series gating
required for this circuit is not available in the faster adders, but the speed of the
MCI019 is adequate for many applications. When properly adjusted, carrier
suppression is quite good.
Introduction of the MECL MC 1650 AID comparator set a new standard for
high speed A-to-D conversion. A 3-bit parallel A-to-D converter is shown in Figure
8-53. The MC1650 is limited to 6-bits conversion because of hysteresis in the circuit.
The MC1650 is also used in serial conversions to save on parts - but with some
sacrifice in speed.
Translators
MECL signal levels are completely compatible within anyone family. In
addition, MECL 10,000 and MECL III parts are designed to interface directly with
each other. However, translators are required to interface MECL with most other
logic families and with signals from other electrical equipment.
MECL II is easily interfaced with MECL 10,000 and MECL III as shown in
Figure 8-54. The higher MECL II output levels are compatible with the faster MECL
10,000 and MECL III inputs when MECL II is heavily loaded with the resistor pair.
While not essential, the resistor combination insures full noise margin in the logic f/J
level. An alternate approach is to use a single 510 n resistor to VEE on the MECL II
output. Conversely, lightly loading the MECL 10,000 or MECL III outputs with a
0.05/-LF
Audio
0---1
O.l/-LF
r
852: MECL II Balanced Modulator
O.l/-LF

10 k
4 5
MC1019
3
13
6
12
10 k
10 k 5k 10 k
_ 50 Carrier fc
206
Ou tputs
So
Audio Input = 220 mV at 3 kHz
Carrier Input = 400 mV at 30 kHz
30 kHz Output = 0.25 mV (Carrier)
27 kHz LSB = 60 mV
33 kHz USB = 60 mV
Analog
Input 0-_-+---1

Gate
Translntors
8-53: MECL III Analog-to-Digital Converter
1
1/4 MC1662 / 1 MSB 1 I

/ /
/ 1
/ 1
/ I NMSB 1 I

1 / 1 1
1 / 1 1
LSB

0/246, ,6421
/ 1 1 1
Analog
I nput Signal
- V ref - /'---'------'---''\,
3-Bit AID Converter
1
I
-
Time
2.0 kSl resistor raises the output logic levels to meet MECL II requirements.
MECL II will operate directly with MECL 10,000 and MECL III, but there is a loss
of 115 mV noise margin at the interface.
MECL and MTTL or MDTL circuits are interfaceable by any of several
methods: When MECL is operated at the recommended - 5.2 volts and TTL at
+5 volts supply, translator circuits such as the MClOl7 and MCI039 may be used.
When using MECL 10,000 or MECL III, a circuit using the MCIOIIO, shown in
Figure 8-SS, makes a good translator. If a -2 volt supply is not available, the two
Thevenin equivalent resistors may be used as shown in Figure 3-2S. The MCI0110
has the advantage of being a dual translator, but the MC1026 or any MECL 10,000
or MECL III gate (as shown in Figure 8-S6) may be used. The MC16S0 AID
comparator is useful for very high speed TTL to MECL translation.
For many small systems it is not practical to have two supplies, with MECL
operating at -S.2 volts and TTL at +5 volts. No integrated circuit translators are
available for interfacing with both circuit types at +5 volts; however translation is
easily accomplished with discrete components as shown in Figure 8-S 7. These
circuits are fast enough for any available TTL.
207
Translators
8-54: Interfacing MEeL " to MEeL III or MEeL 10,000
.< (bl .
,,'" < "
8-55: Dual MEeL to TTL Interface
+5.0 Vdc +5.0 Vdc
16
270
....._----OTTL Out
5
TTL Out
MECL In
MECL In
12
8
50 50
-5.2 Vdc
-2.0 Vdc
-2.0 Vdc
MECL also interfaces readily with MOS. With CMOS operating at +5 volts, any
of the MECL to TTL translators works very well. On the other hand, CMOS will
drive MECL directly when using a common supply. P-channel MOS, operating with a
negative supply, requires simple translators as shown in Figure 8-58.
MECL translators such as the MC 1018 are used with transistor buffers for lamp
driving as shown in Figure 8-59. These circuits are also used for interfacing with
most discrete component circuits when the input must drive the base of a grounded
emitter transistor. The MECL to MOS translator circuit in Figure 8-58 drives a lamp
connected to a negative voltage. These circuits can be further buffered for driving
relays and other equipment with high current requirements.
Important when designing indicator displays is the ability of MECL 10,000 and
MECL III to drive light emitting diodes (LEDs) directly. The LED is connected
between the MECL output and VCC. The MECL I level is adequate to hold the
diode off, and a (/J level is sufficient to allow the diode to conduct. Current in the
light emitting diode is controlled by the size of the pulldown resistor between the
MECL output and VEE.
208


..s:l
""
:;:::

0\
o
N
Translators
8-57: TTL/MECL I nterfaced with Discrete Components, to Permit Common Supply Voltage Operation
+5.0 Vdc
(a) MECL to TTL Transistor
(b) TTL to MECL Transistor
8-58: MECL 10,OOO/MOS Interface
210
Translators
8-59: MECL II Lamp Driver
Vin -0.75 V: Bulb Off
BulbOn
6 + 10 + 11 + 12 + 13 Bulb On
The MECL line receiver and Schmitt trigger circuits are ideal for interfacing low
frequency sine waves and low amplitude signals to MECL. An example of this type
of circuit was shown in Figure 8-50 and described earlier in this chapter under the
heading, "Linear Applications."
A CONCLUDING COMMENT
The wide variety of applications shown in this chapter illustrates the versatility
of MECL integrated circuits. Although used primarily in high performance digital
equipment, these circuits have been designed into all types of electronic equipment.
The circuits exhibited in this section are intended as design ideas for using MECL,
and are not meant to restrict the user to the particular applications shown.
Motorola has incorporated the features necessary for high system performance
into MECL circuits. How well this performance is translated into an operating system
is measured by the ingenuity and imagination of the System Designer.
211
Index of Tabulated Data
Figure Title Page
1-3 Worst-Case Change of Levels As a Function of Temperature, for MECL 10,000
and MECL III . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 4
1-7 MECL Family Comparison .................................................... 11
3-13 Maximum Open Line Length for MECL 10,000 .................................... 44
3-14 Maximum Open Line Length for MECL II ........................................ 45
3-15 Maximum Open Line Length for MECL III ....................................... 45
3-22 Minimum Values of RS for Any Length of Line with Specified Limits of Overshoot
and Undershoot, Using MECL 10,000 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 51
3- 23 Minimum Valucs of RS for Any Length of Line with Specified Limits of Overshoot
and Undershoot, Using MECL III . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 51
3-26 Types of Lines Recommended ................................................. 54
3-27 Power Consumption for Various Line Terminations ................................. 55
4-4 Impedance Characteristics of Carbon Resistors Measured on a GR Admittance Bridge ...... 61
4-7 Typical Switching Times. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 62
4-10 Attenuation of 50 Ft Twisted Pair Line with MCI026 Line Driver ..................... 66
4-12 Attcnuation in a 50 Ft Twisted Pair Line with a MECL III Driver ...................... 66
4-13 Attenuation in a 10 Ft Twisted Pair Line with a MECL III Driver ...................... 66
4-27 10 Ft Multiple Conductor Cable Crosstalk ........................................ 81
4-31 Test Results for an 18 Inch Multiple Conductor Cable: Crosstalk. . . . . . . . . . . . . . . . . . . . .. 83
4-32 Crosstalk for 10 Ft Multiple Twisted Pair Cable. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 84
4-34 Crosstalk for 10 Ft Multiple Shielded Twisted Pair. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 85
5-2 Changes in Output Levels and VBB with VEE ..................................... 88
5-3 Typical Output Power. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 90
6-6 Typical Thermal Characteristics for MECL Packages ................................ 105
6-9 MECL III Worst-Case Logic Levels ............................................. 110
6-10 Junction Temperature Thermal Gradients ........................................ 111
7-23 Maximum Capacitances That Can be Lumped or Distributed Over a Length of
Terminated Transmission Line Qmax- ........................................... 145
212
Index
Attenuation, 58
Backplane:
power, 95
wiring, 24, 25
Bandwidth, 59, 66
Ca pacitance:
distributed, 123, 139, 144
input, 12, 141
Capacitors, bypass; 20, 24
Characteristic impedance, 27,30, 116, 142
Circuit boards, 20, 23, 29, 96
Coaxial cable, 24, 38, 59
Collector dotting, 7
Common mode rejection, 64
Connectors, 59
Cooling:
forced air, 110
heat sink, 112
Crosstalk, 29, 55, 58, 76,135
Current input, 11
Emitter coupled logic, vi
Emitter follower outputs, 2
Fall time, 13, 15, 26, 120
Ferrite beads, 22
Flip-flop:
AC-coupled, 8
master-sla ve, 9
Gain of circuits, 64
Ground, 22, 25,95
Ground plane, 23, 37, 55, 95, 96,128
Ground screen, 21,95
Heat sink, 112
Impedance, output; 13, 116
Junction temperature, 101, 112
Lattice diagram, 118, 162
Line lengths, 43, 124
Line receivers, 64
MECL I, vi
MECL II, vi, 15
MECL III, vii, 29
MECL 10,000, vii, 23
Microstrip line, 38, 128
Noise immunity, 5
Noise margin, 5, 57, 87, 110
Overshoot, 21, 36, 57, 119
Parallel termination, 47, 52,139,146
Party line, 68
Power:
backplane, 95
circuit, 88
213
lines, 92
supply, 24,91
Propagation delay, 13, 15, 26
Propagation of lines, 18, 27, 30,58,63,123
125,150
Pulldown resistor, 47,91,152
Receivers, line; 64
Reflection coefficient, 43,117,128,144
Resistors:
pulldown,47,91,152
series damping, 51, 153
termination, 47, 52,139,146
Ribbon cable, 70
Rise time, 13, 15
Schottky diode termination, 71
Series:
damping, 51,153
gating, 6
termination, 46, 53, 146
Skin effect, 59
Speed-power product, 14
Strip line, 39, 128
Termination:
parallel, 47,52,139,146
resistors, 90
schottky diode, 71
series, 46, 53, 146
voltage, 99
Thermal:
conduction, 113
resistance, 10 1
Time domain reflectometer, 126
Toggle rate, 13, 15
Transmission line:
coaxial, 24, 38, 59
microstrip, 38, 128
ribbon cable, 70
strip line, 39, 128
triaxial cable, 70
twisted pair, 22, 25, 31, 38,64,69,83
Transmission lines, 35,42, 115
Transfer characteristics, 3
Transistor beta, 5
Twisted pair lines, 22, 25, 31, 38, 64, 69,83
Undershoot, 21, 36, 57, 119
Unterminated lines, 43
Voltage:
termination, 99
VBB, 1,88,106
VCC, 13, 87, 97
VEE, 88
Wire cables, 76
Wired-or, 14, 19, 27, 31
Wire-wrap connections, 13, 55
Motorola Integrated Circuits Center
Mesa, Arizona
214