C28x Piccolo Workshop

C2000 Piccolo Workshop
Workshop Guide and Lab Manual
F28xPmdw Revision 1.0 September 2009
Technical Training Organization
Important Notice
Important Notice
Texas Instruments and its subsidiaries (TI) reserve the right to make changes to their products or to discontinue any product or service without notice, and advise customers to obtain the latest version of relevant information to verify, before placing orders, that information being relied on is current and complete. All products are sold subject to the terms and conditions of sale supplied at the time of order acknowledgment, including those pertaining to warranty, patent infringement, and limitation of liability. TI warrants performance of its semiconductor products to the specifications applicable at the time of sale in accordance with TIs standard warranty. Testing and other quality control techniques are utilized to the extent TI deems necessary to support this warranty. Specific testing of all parameters of each device is not necessarily performed, except those mandated by government requirements. Customers are responsible for their applications using TI components. In order to minimize risks associated with the customers applications, adequate design and operating safeguards must be provided by the customer to minimize inherent or procedural hazards. TI assumes no liability for applications assistance or customer product design. TI does not warrant or represent that any license, either express or implied, is granted under any patent right, copyright, mask work right, or other intellectual property right of TI covering or relating to any combination, machine, or process in which such semiconductor products or services might be or are used. TIs publication of information regarding any third partys products or services does not constitute TIs approval, warranty or endorsement thereof.
Copyright 2009 Texas Instruments Incorporated
Revision History
September 2009 Revision 1.0
Mailing Address
Texas Instruments Training Technical Organization 7839 Churchill Way M/S 3984 Dallas, Texas 75251-1903
ii
C2000 Piccolo Workshop - Introduction
Texas Instruments Technical Training

T TO
Technical Training Organization
C2000 and Piccolo are trademarks of Texas Instruments. Copyright 2009 Texas Instruments. All rights reserved.
Introductions
Introductions
Name Company Project Responsibilities DSP / Microcontroller Experience TMS320 Processor Experience Hardware / Software - Assembly / C Interests
iii
C2000 Piccolo Workshop Outline
C2000 Piccolo Workshop Outline

1. Architecture Overview 2. Programming Development Environment Lab: Linker command file 3. Peripheral Register Header Files 4. Reset and Interrupts 5. System Initialization Lab: Watchdog and interrupts 6. Analog-to-Digital Converter Lab: Build a data acquisition system 7. Control Peripherals Lab: Generate and graph a PWM waveform 8. Numerical Concepts and IQ Math Lab: Low-pass filter the PWM waveform 9. Control Law Accelerator (CLA) Lab: Use CLA to filter PWM waveform 10. System Design Lab: Run the code from flash memory 11. Communications 12. DSP/BIOS Lab: Run DSP/BIOS code from flash memory 13. Support Resources
C2000 Experimenter Kit
Piccolo Experimenter Kit
ControlCARD
USB Docking Station
iv
Architecture Overview
Introduction
This architectural overview introduces the basic architecture of the C2000 Piccolo series of microcontrollers from Texas Instruments. The Piccolo series adds a new level of general purpose processing ability unseen in any previous DSP/MCU chips. The C2000 is ideal for applications combining digital signal processing, microcontroller processing, efficient C code execution, and operating system tasks. Unless otherwise noted, the terms C28x, F28x and F2803x refer to TMS320F2803x devices throughout the remainder of these notes. For specific details and differences please refer to the device data sheet and users guide.
Learning Objectives
When this module is complete, you should have a basic understanding of the F28x architecture and how all of its components work together to create a high-end, uniprocessor control system.
Learning Objectives
Review the F28x block diagram and device features Describe the F28x bus structure and memory map Identify the various memory blocks on the F28x Identify the peripherals available on the F28x
C2000 Piccolo Workshop - Architecture Overview
1-1
Module Topics
Module Topics
Architecture Overview.............................................................................................................................. 1-1 Module Topics......................................................................................................................................... 1-2 What is the TMS320C2000?................................................................................................................ 1-3 TMS320C2000 Internal Bussing .................................................................................................... 1-4 F28x CPU ............................................................................................................................................... 1-5 Special Instructions............................................................................................................................. 1-6 Pipeline Advantage............................................................................................................................. 1-7 Memory ................................................................................................................................................... 1-8 Memory Map ...................................................................................................................................... 1-8 Code Security Module (CSM) ............................................................................................................ 1-9 Peripherals .......................................................................................................................................... 1-9 Fast Interrupt Response .........................................................................................................................1-10 F28x Mode .............................................................................................................................................1-11 Reset.......................................................................................................................................................1-12 Summary ................................................................................................................................................1-13
1-2
What is the TMS320C2000?

The TMS320C2000 is a 32-bit fixed point microcontroller that specializes in high performance control applications such as, robotics, industrial automation, mass storage devices, lighting, optical networking, power supplies, and other control applications needing a single processor to solve a high performance application.
TMS320F2803x Block Diagram

Program Bus ePWM Sectored Flash
CLA Bus
RAM
Boot ROM
eCAP eQEP 12-bit ADC Watchdog
32-bit R-M-W 32x32 bit Auxiliary Atomic Multiplier Registers ALU Real-Time JTAG Emulation Register Bus CPU
CLA
PIE Interrupt Manager 3 32-bit Timers
CAN 2.0B I2C SCI SPI LIN
Data Bus GPIO
The F2803x architecture can be divided into 3 functional blocks: CPU and busing Memory Peripherals
1-3
TMS320C2000 Internal Bussing

As with many DSP-type devices, multiple busses are used to move data between the memories and peripherals and the CPU. The F28x memory bus architecture contains: A program read bus (22-bit address line and 32-bit data line) A data read bus (32-bit address line and 32-bit data line) A data write bus (32-bit address line and 32-bit data line)
F28x CPU Internal Bus Structure

Program PC Decoder Program Address Bus (22) Program-read Data Bus (32) Data-read Address Bus (32) Data-read Data Bus (32) Registers ARAU SP DP @X XAR0 to XAR7 Execution MPY32x32 ALU XT P ACC Debug R-M-W Atomic ALU CLA Real-Time JTAG Emulation Data Memory Program Memory
Peripherals
Register Bus / Result Bus Data/Program-write Data Bus (32) Data-write Address Bus (32)
External Interface
The 32-bit-wide data busses enable single cycle 32-bit operations. This multiple bus architecture, known as a Harvard Bus Architecture enables the F28x to fetch an instruction, read a data value and write a data value in a single cycle. All peripherals and memories are attached to the memory bus and will prioritize memory accesses.
1-4
F28x CPU
F28x CPU
The F28x is a highly integrated, high performance solution for demanding control applications. The F28x is a cross between a general purpose microcontroller and a digital signal processor, balancing the code density of a RISC processor and the execution speed of a DSP with the architecture, firmware, and development tools of a microcontroller. The DSP features include a modified Harvard architecture and circular addressing. The RISC features are single-cycle instruction execution, register-to-register operations, and a modified Harvard architecture. The microcontroller features include ease of use through an intuitive instruction set, byte packing and unpacking, and bit manipulation.
F28x and CLA

Program Bus CLA Bus
MCU/DSP balancing code density & execution time

16-bit instructions for improved code density 32-bit instructions for improved execution time
PIE Interrupt Manager CLA 3 32-bit
32-bit Auxiliary Registers
32x32 bit Multiplier
R-M-W Atomic ALU
Register Bus CPU Data Bus
Timers
32-bit fixed-point CPU 32x32 fixed-point MAC, doubles as dual 16x16 MAC Parallel processing Control Law Accelerator (CLA) adds IEEE Single-precision 32-bit floating point math operations CLA algorithm execution is independent of the main CPU Fast interrupt service time Single cycle read-modify-write instructions Unique real-time debugging capabilities
The F28x design supports an efficient C engine with hardware that allows the C compiler to generate compact code. Multiple busses and an internal register bus allow an efficient and flexible way to operate on the data. The architecture is also supported by powerful addressing modes, which allow the compiler as well as the assembly programmer to generate compact code that is almost one to one corresponded to the C code. The F28x is as efficient in DSP math tasks as it is in system control tasks. This efficiency removes the need for a second processor in many systems. The 32 x 32-bit MAC capabilities of the F28x and its 64-bit processing capabilities, enable the F28x to efficiently handle higher numerical resolution problems that would otherwise demand a more expensive solution. Along with this is the capability to perform two 16 x 16-bit multiply accumulate instructions simultaneously or Dual MACs (DMAC). Also, some devices feature a floating-point unit. The, F28x is source code compatible with the 24x/240x devices and previously written code can be reassembled to run on a F28x device, allowing for migration of existing code onto the F28x.
1-5
F28x CPU
Special Instructions
F28x Atomic Read/Modify/Write

Atomic Instructions Benefits:
LOAD READ
Simpler programming Smaller, faster code Mem Uninterruptible (Atomic) More efficient compiler
Registers
CPU
ALU / MPY
WRITE STORE
Standard Load/Store
DINT MOV AND MOV EINT AL,*XAR2 AL,#1234h *XAR2,AL
Atomic Read/Modify/Write
AND *XAR2,#1234h
2 words / 1 cycles
6 words / 6 cycles
Atomics are small common instructions that are non-interuptable. The atomic ALU capability supports instructions and code that manages tasks and processes. These instructions usually execute several cycles faster than traditional coding.
1-6
F28x CPU
Pipeline Advantage
F28x Pipeline
A F1 F2 D 1 D2 R1 R 2 E W B C D E F G H
F1: Instruction Address F2: Instruction Content D1: Decode Instruction D2: Resolve Operand Addr R1: Operand Address R2: Get Operand E: CPU doing real work W: store content to memory
8-stage pipeline
W
F 1 F 2 D1 D 2 R1 R2 E W F1 F2 D 1 D2 R1 R 2 E F 1 F 2 D1 D 2 R1 R2 E W F1 F2 D1 D2 R1 R 2 E W F1 F 2 D1 D2 R1 R2 E W F1 F2 D1 D2 R 1 R 2 R2 W W R E E F1 F 2 D1 D2 R1 R2 R 2 W W R1 E E Protected Pipeline Order of results are as written in source code Programmer need not worry about the pipeline
E & G Access same address
The F28x uses a special 8-stage protected pipeline to maximize the throughput. This protected pipeline prevents a write to and a read from the same location from occurring out of order. This pipelining also enables the F28x to execute at high speeds without resorting to expensive high-speed memories. Special branch-look-ahead hardware minimizes the latency for conditional discontinuities. Special store conditional operations further improve performance.
1-7
Memory
Memory
The memory space on the F28x is divided into program memory and data memory. There are several different types of memory available that can be used as both program memory and data memory. They include the flash memory, single access RAM (SARAM), OTP, and Boot ROM which is factory programmed with boot software routines or standard tables used in math related algorithms.
Memory Map
The F28x CPU contains no memory, but can access memory on chip. The F28x uses 32-bit data addresses and 22-bit program addresses. This allows for a total address reach of 4G words (1 word = 16-bits) in data memory and 4M words in program memory. Memory blocks on all F28x designs are uniformly mapped to both program and data space. This memory map shows the different blocks of memory available to the program and data space.
TMS320F28035 Memory Map

0x000000 0x000400 0x000800 0x000D00
Data Program M0 SARAM (1Kw) M1 SARAM (1Kw)
PIE Vectors (256 w )
0x3D7C80 0x3D8000 0x3E8000 0x3F7FF8 0x3F8000 0x3F8800 0x3FE000 0x3FFFC0 0x3FFFFF
ADC / OSC cal. data reserved FLASH (64Kw)

PASSWORDS (8w)
0x000E00 reserved PF 0 (6Kw) 0x002000 0x006000 PF 1 (4Kw) 0x007000 PF 2 (4Kw) 0x008000 L0 SARAM (2Kw) 0x008800 L1 DPSARAM (1Kw) 0x008C00 L2 DPSARAM (1Kw) 0x009000 L3 DPSARAM (4Kw) 0x00A000 reserved 0x3D7800 User OTP (1Kw) 0x3D7C00 reserved 0x3D7C80
L0 SARAM (2Kw) reserved Boot ROM (8Kw)

BROM Vectors (64w)
Data
Program
Dual Mapped: L0
Dual-Port RAM: L1, L2 & L3 (accessible by CPU & CLA)
CSM Protected: L0, L1, L2, L3, OTP FLASH, ADC CAL, Flash Regs in PF0
1-8
Memory
Code Security Module (CSM)
Code Security Module

Prevents reverse engineering and protects valuable intellectual property
0x008000 0x008800 0x008C00 0x009000 0x00A000 0x3D7800 0x3D7C00 0x3D7C80 0x3D8000 0x3E8000 0x3F7FF8 0x3F8000 0x3F8800 L0 SARAM (2Kw) L1 DPSARAM (1Kw) L2 DPSARAM (1Kw) L3 DPSARAM (4Kw) reserved User OTP (1Kw) reserved ADC / OSC cal. data reserved FLASH (64Kw)
PASSWORDS (8w)
Dual Mapped
L0 SARAM (2Kw)
128-bit user defined password is stored in Flash 128-bits = 2128 = 3.4 x 1038 possible passwords To try 1 password every 8 cycles at 60 MHz, it would take at least 1.4 x 1024 years to try all possible combinations!
Peripherals
The F28x comes with many built in peripherals optimized to support control applications. These peripherals vary depending on which F28x device you choose. ePWM eCAP eQEP Analog-to-Digital Converter Watchdog Timer CLA SPI SCI I2C LIN CAN GPIO
1-9
Fast Interrupt Response
Fast Interrupt Response

The fast interrupt response, with automatic context save of critical registers, resulting in a device that is capable of servicing many asynchronous events with minimal latency. F28x implements a zero cycle penalty to do 14 registers context saved and restored during an interrupt. This feature helps reduces the interrupt service routine overheads.
F28x Fast Interrupt Response Manager

96 dedicated PIE vectors No software decision making required Direct access to RAM vectors Auto flags update Concurrent auto context save Auto Context Save T ST0 AH AL PH PL AR1 (L) AR0 (L) DP ST1 DBSTAT IER PC(msw) PC(lsw)
Peripheral Interrupts 12x8 = 96
PIE module For 96 interrupts INT1 to INT12 96 PIE Register Map 12 interrupts
28x CPU Interrupt logic
IFR
IER
INTM
28x CPU
1 - 10
F28x Mode
F28x Mode
The F28x is one of several members of the TMS320 microcontroller family. The F28x is source code compatable with the 24x/240x devices and previously written code can be reassembled to run on a F28x device. This allows for migration of existing code onto the F28x.
F28x Operating Modes

Mode Type C28x Native Mode C24x Compatible Mode Test Mode (default) Reserved Mode Bits
OBJMODE AMODE
Compiler Option -v28 -v28 m20
1 1 0 0
0 1 0 1
Almost all uses will run in C28x Native Mode The bootloader will automatically select C28x Native Mode after reset C24x compatible mode is mostly for backwards compatibility with an older processor family
1 - 11
Reset
Reset
Reset Bootloader
Reset
OBJMODE = 0 AMODE = 0 ENPIE = 0 INTM = 1
Reset vector fetched from boot ROM 0x3F FFC0
Bootloader sets
OBJMODE = 1 AMODE = 0
YES TRST = 1
Emulator Connected?
The wait boot mode is used and the boot mode is determined by the debugger Boot Mode Wait
Emulation Boot Boot determined by 2 RAM locations:

EMU_KEY and EMU_BMODE
Note: Details of the various boot options will be discussed in the Reset and Interrupts module
TRST = JTAG Test Reset EMU_KEY & EMU_BMODE located in PIE at 0x0D00 & 0x0D01, respectively
1 - 12
Summary
Summary
Summary
High performance 32-bit CPU 32x32 bit or dual 16x16 bit MAC Hardware Control Law Accelerator (CLA) Atomic read-modify-write instructions Fast interrupt response manager 64Kw on-chip flash memory Code security module (CSM) Control peripherals 12-bit ADC module Comparators Up to 44 shared GPIO pins Communications peripherals
1 - 13
Summary
1 - 14
Programming Development Environment

Introduction
This module will explain how to use Code Composer Studio (CCS) integrated development environment (IDE) tools to develop a program. Creating projects and setting building options will be covered. Use and the purpose of the linker command file will be described.
Learning Objectives
Learning Objectives
Use Code Composer Studio to:
Create a Project Set Build Options
Create a user linker command file which:

Describes a systems available memory Indicates where sections will be placed in memory
C2000 Piccolo Workshop - Programming Development Environment
2-1
Module Topics
Module Topics
Programming Development Environment .............................................................................................. 2-1 Module Topics......................................................................................................................................... 2-2 Code Composer Studio ........................................................................................................................... 2-3 Software Development and COFF Concepts...................................................................................... 2-3 Projects ............................................................................................................................................... 2-5 Build Options...................................................................................................................................... 2-6 Creating a Linker Command File ........................................................................................................... 2-9 Sections .............................................................................................................................................. 2-9 Linker Command Files (.cmd) .........................................................................................................2-12 Memory-Map Description .................................................................................................................2-12 Section Placement..............................................................................................................................2-14 Exercise 2...............................................................................................................................................2-15 Summary: Linker Command File ......................................................................................................2-16 Lab 2: Linker Command File.................................................................................................................2-17 Solutions.................................................................................................................................................2-22
2-2
Code Composer Studio

Software Development and COFF Concepts
In an effort to standardize the software development process, TI uses the Common Object File Format (COFF). COFF has several features which make it a powerful software development system. It is most useful when the development task is split between several programmers. Each file of code, called a module, may be written independently, including the specification of all resources necessary for the proper operation of the module. Modules can be written using Code Composer Studio (CCS) or any text editor capable of providing a simple ASCII file output. The expected extension of a source file is .ASM for assembly and .C for C programs.

Build Compile lnk.cmd Code Simulator
eZdsp Asm Link Debug External Emulator MCU Board
Editor
Libraries
Graphs, Profiling
Code Composer Studio includes:

Integrated Edit/Debug GUI Code Generation Tools DSP/BIOS
Code Composer Studio includes a built-in editor, compiler, assembler, linker, and an automatic build process. Additionally, tools to connect file input and output, as well as built-in graph displays for output are available. Other features can be added using the plug-ins capability Numerous modules are joined to form a complete program by using the linker. The linker efficiently allocates the resources available on the device to each module in the system. The linker uses a command (.CMD) file to identify the memory resources and placement of where the various sections within each module are to go. Outputs of the linking process includes the linked object file (.OUT), which runs on the device, and can include a .MAP file which identifies where each linked section is located. The high level of modularity and portability resulting from this system simplifies the processes of verification, debug and maintenance. The process of COFF development is presented in greater detail in the following paragraphs.
2-3
The concept of COFF tools is to allow modular development of software independent of hardware concerns. An individual assembly language file is written to perform a single task and may be linked with several other tasks to achieve a more complex total system. Writing code in modular form permits code to be developed by several people working in parallel so the development cycle is shortened. Debugging and upgrading code is faster, since components of the system, rather than the entire system, is being operated upon. Also, new systems may be developed more rapidly if previously developed modules can be used in them. Code developed independently of hardware concerns increases the benefits of modularity by allowing the programmer to focus on the code and not waste time managing memory and moving code as other code components grow or shrink. A linker is invoked to allocate systems hardware to the modules desired to build a system. Changes in any or all modules, when re-linked, create a new hardware allocation, avoiding the possibility of memory resource conflicts.
Code Composer Studio: IDE
Integrates: edit, code generation, and debug Single-click access using buttons Powerful graphing/profiling tools Automated tasks using GEL scripts and CCS scripting Built-in access to BIOS functions Supports TI and 3 rd party plug-ins
2-4
Projects
Code Composer works with a project paradigm. Essentially, within CCS you create a project for each executable program you wish to create. Projects store all the information required to build the executable. For example, it lists things like: the source files, the header files, the target systems memory-map, and program build options.
The CCS Project

Project (.pjt) files contain:
List of files:
Source (C, assembly) Libraries DSP/BIOS configuration file Linker command files
Project settings:
Build options (compiler, Linker, assembler, and DSP/BIOS) Build configurations
The project information is stored in a .PJT file, which is created and maintained by CCS. To create a new project, you need to select the Project:New menu item. Along with the main Project menu, you can also manage open projects using the right-click popup menu. Either of these menus allows you to Add Files to a project. Of course, you can also drag-n-drop files onto the project from Windows Explorer.
2-5
Build Options
Project options direct the code generation tools (i.e. compiler, assembler, linker) to create code according to your systems needs. When you create a new project, CCS creates two sets of build options called Configurations: one called Debug, the other Release (you might think of as Optimize). To make it easier to choose build options, CCS provides a graphical user interface (GUI) for the various compiler options. Heres a sample of the Debug configuration options.
Build Options GUI - Compiler
GUI has 8 pages of categories for code generation tools Controls many aspects of the build process, such as:
Optimization level Target device Compiler/assembly/link options
There is a one-to-one relationship between the items in the text box and the GUI check and dropdown box selections. Once you have mastered the various options, you can probably find yourself just typing in the options.
2-6
Build Options GUI - Linker

GUI has 3 categories for linking
Specify various link options
.\Debug means the directory called Debug one level below the .pjt file directory $(Proj_dir)\Debug is an equivalent expression
There are many linker options but these four handle all of the basic needs. -o <filename> specifies the output (executable) filename. -m <filename> creates a map file. This file reports the linkers results. -c tells the compiler to autoinitialize your global and static variables. -x tells the compiler to exhaustively read the libraries. Without this option libraries are searched only once, and therefore backwards references may not be resolved.
2-7
Default Build Configurations

For new projects, CCS automatically creates two build configurations:
Debug Release (unoptimized) (optimized)
Use the drop-down menu to quickly select the build configuration
Add/Remove your own custom build configurations using Project Configurations Edit a configuration:
1. 2. 3.
Set it active Modify build options Save project
To help make sense of the many compiler options, TI provides two default sets of options (configurations) in each new project you create. The Release (optimized) configuration invokes the optimizer with o3 and disables source-level, symbolic debugging by omitting g (which disables some optimizations to enable debug).
2-8
Creating a Linker Command File

Sections
Looking at a C program, you'll notice it contains both code and different kinds of data (global, local, etc.).
Sections
Global vars (.ebss) Init values (.cinit)
int int
x = 2; y = 7;
void main(void) { long z; z = x + y; }

Local vars (.stack) Code (.text)
All code consists of different parts called sections All default section names begin with . The compiler has default section names for initialized and uninitialized sections
In the TI code-generation tools (as with any toolset based on the COFF Common Object File Format), these various parts of a program are called Sections. Breaking the program code and data into various sections provides flexibility since it allows you to place code sections in ROM and variables in RAM. The preceding diagram illustrated four sections: Global Variables Initial Values for global variables Local Variables (i.e. the stack) Code (the actual instructions)
2-9
Following is a list of the sections that are created by the compiler. Along with their description, we provide the Section Name defined by the compiler.
Compiler Section Names

Initialized Sections Name Description .text .cinit .econst .switch .pinit code initialization values for global and static variables constants (e.g. const int k = 3;) tables for switch statements tables for global constructors (C++) FLASH FLASH FLASH Link Location FLASH FLASH
Uninitialized Sections Name Description .ebss .stack global and static variables stack space
Link Location RAM low 64Kw RAM RAM
.esysmem memory for far malloc functions
Note: During development initialized sections could be linked to RAM since the emulator can be used to load the RAM
Sections of a C program must be located in different memories in your target system. This is the big advantage of creating the separate sections for code, constants, and variables. In this way, they can all be linked (located) into their proper memory locations in your target embedded system. Generally, theyre located as follows: Program Code (.text) Program code consists of the sequence of instructions used to manipulate data, initialize system settings, etc. Program code must be defined upon system reset (power turn-on). Due to this basic system constraint it is usually necessary to place program code into non-volatile memory, such as FLASH or EPROM. Constants (.cinit initialized data) Initialized data are those data memory locations defined at reset.It contains constants or initial values for variables. Similar to program code, constant data is expected to be valid upon reset of the system. It is often found in FLASH or EPROM (non-volatile memory). Variables (.ebss uninitialized data) Uninitialized data memory locations can be changed and manipulated by the program code during runtime execution. Unlike program code or constants, uninitialized data or variables must reside in volatile memory, such as RAM. These memories can be modified and updated, supporting the way variables are used in math formulas, high-level languages, etc. Each variable must be declared with a directive to reserve memory to contain its value. By their nature, no value is assigned, instead they are loaded at runtime by the program
2 - 10
Placing Sections in Memory

Memory
0x00 0000 0x00 0400
Sections .ebss .stack .cinit .text
M0SARAM (0x400) M1SARAM (0x400)
0x3E 8000
FLASH (0x10000)
Linking code is a three step process: 1. Defining the various regions of memory (on-chip SARAM vs. FLASH vs. External Memory). 2. Describing what sections go into which memory regions 3. Running the linker with build or rebuild
2 - 11
Linker Command Files (.cmd)

The linker concatenates each section from all input files, allocating memory to each section based on its length and location as specified by the MEMORY and SECTIONS commands in the linker command file.
Linking
Memory description How to place s/w into h/w
Link.cmd .obj Linker .out
.map
Memory-Map Description
The MEMORY section describes the memory configuration of the target system to the linker. The format is: Name: origin = 0x????, length = 0x????
For example, if you placed a 64Kw FLASH starting at memory location 0x3E8000, it would read:
MEMORY { FLASH: }
origin = 0x3E8000 , length = 0x010000
Each memory segment is defined using the above format. If you added M0SARAM and M1SARAM, it would look like:
MEMORY { M0SARAM: M1SARAM: }
origin = 0x000000 , length = 0x0400 origin = 0x000400 , length = 0x0400
2 - 12
Remember that the DSP has two memory maps: Program, and Data. Therefore, the MEMORY description must describe each of these separately. The loader uses the following syntax to delineate each of these:
Linker Page
Page 0 Page 1
TI Definition
Program Data
Linker Command File

MEMORY { PAGE 0: FLASH: /* Program Memory */ origin = 0x3E8000, length = 0x10000
PAGE 1: /* Data Memory */ M0SARAM: origin = 0x000000, length = 0x400 M1SARAM: origin = 0x000400, length = 0x400 }
2 - 13
Section Placement
The SECTIONS section will specify how you want the sections to be distributed through memory. The following code is used to link the sections into the memory specified in the previous example:
SECTIONS { .text:> .ebss:> .cinit:> .stack:> }
FLASH M0SARAM FLASH M1SARAM
PAGE PAGE PAGE PAGE
0 1 0 1
The linker will gather all the code sections from all the files being linked together. Similarly, it will combine all like sections. Beginning with the first section listed, the linker will place it into the specified memory segment.
Linker Command File

MEMORY { PAGE 0: FLASH: /* Program Memory */ origin = 0x3E8000, length = 0x10000
PAGE 1: /* Data Memory */ M0SARAM: origin = 0x000000, length = 0x400 M1SARAM: origin = 0x000400, length = 0x400 } SECTIONS { .text:> .ebss:> .cinit:> .stack:> }
PAGE PAGE PAGE PAGE
= = = =
0 1 0 1
2 - 14
Exercise 2
Exercise 2
Looking at the following block diagram, and create a linker command file.
Exercise 2
0x00 0000
M0SARAM (0x400)
0x00 0400
M1SARAM (0x400)
0x00 8000
L0SARAM (0x800)
0x3E 8000
FLASH (0x10000)
Generic F28x device
Create the linker command file for the given memory map by filling in the blanks on the following slide
Fill in the blanks:
Exercise 2 - Command File

MEMORY { PAGE__: _____: ______: _______: _______: _______: } SECTIONS { .text: .ebss: .cinit: .stack: } /* origin /* origin origin origin Program Memory */ = ____ ___, length Data Memory */ = __ ____, length = ____ ___, length = ___ ____, length
= ___ ___ = _____ = _____ = __ __
> > > >
PAGE PAGE PAGE PAGE
= = = =
0 1 0 1
2 - 15
Exercise 2
Summary: Linker Command File

The linker command file (.cmd) contains the inputs commands for the linker. This information is summarized below:
Linker Command File Summary

Memory Map Description
Name Location Size
Sections Description
Directs software sections into named memory regions Allows per-file discrimination Allows separate load/run locations
2 - 16
Lab 2: Linker Command File

Objective
Create a linker command file and link the C program file (Lab2.c) into the system described below.

Memory on-chip memory
0x00 0000 M0SARAM (0x400) M1SARAM (0x400) 0x00 8000 L0SARAM (0x800) 0x00 8800 L1DPSARAM (0x400) 0x00 8C00 L2DPSARAM (0x400) 0x00 9000 L3DPSARAM (0x1000)
0x00 0400
F28035
System Description:
TMS320F28035 All internal RAM blocks allocated
Placement of Sections:
.text into RAM Block L0SARAM on PAGE 0 (program memory) .cinit into RAM Block L0SARAM on PAGE 0 (program memory) .ebss into RAM Block M0SARAM on PAGE 1 (data memory) .stack into RAM Block M1SARAM on PAGE 1 (data memory)
System Description
TMS320F28035 All internal RAM blocks allocated
Placement of Sections:
.text into RAM Block L0SARAM on PAGE 0 (program memory) .cinit into RAM Block L0SARAM on PAGE 0 (program memory) .ebss into RAM Block M0SARAM on PAGE 1 (data memory) .stack into RAM Block M1SARAM on PAGE 1 (data memory)
Procedure
Create a New Project

1. Double click on the Code Composer Studio icon on the desktop. Maximize Code Composer Studio to fill your screen. Code Composer Studio has a Connect/Disconnect feature which allows the target to be dynamically connected and disconnected. This will reset the JTAG link and also enable hot swapping a target board.
C2000 Piccolo Workshop - Programming Development Environment 2 - 17
2. Connect to the target. Click: Debug Connect
The menu bar (at the top) lists File ... Help. Note the horizontal tool bar below the menu bar and the vertical tool bar on the left-hand side. The window on the left is the project window and the large right-hand window is your workspace. 3. A project contains all the files you will need to develop an executable output file (.out) which can be run on the MCU hardware. Lets create a new project for this lab. On the menu bar click: Project New
type Lab2 in the project name field and make sure the save in location is: C:\C28x\Labs\Lab2, then click Finish. This will create a .pjt file which will invoke all the necessary tools (compiler, assembler, linker) to build your project. It will also create a debug folder that will hold immediate output files. 4. Add the C file to the new project. Click: Project Add Files to Project
and make sure youre looking in C:\C28x\Labs\Lab2. Change the files of type to view C source files (*.c) and select Lab2.c and click OPEN. This will add the file Lab2.c to your newly created project. 5. Add Lab2.cmd to the project using the same procedure. This file will be edited during the lab exercise. 6. In the project window on the left, click the plus sign (+) to the left of Project. Now, click on the plus sign next to Lab2.pjt. Notice that the Lab2.cmd file is listed. Click on the plus sign next to Source to see the current source file list (i.e. Lab2.c).
Project Build Options

7. There are numerous build options in the project. The default option settings are sufficient for getting started. We will inspect a couple of the default linker options at this time. Click: Project Build Options
8. Select the Linker tab. Notice that .out and .map files are being created. The .out file is the executable code that will be loaded into the MCU. The .map file will contain a linker report showing memory usage and section addresses in memory. 9. Set the Stack Size to 0x200. 10. Next, setup the compiler run-time support library. In the Libraries Category, find the Include Libraries (-l) box and enter: rts2800_ml.lib. Select OK and the Build Options window will close.
Edit the Linker Command File - Lab2a.cmd

11. To open and edit Lab2.cmd, double click on the filename in the project window.
2 - 18
12. Edit the Memory{} declaration by describing the system memory shown on the Lab2: Linker Command File slide in the objective section of this lab exercise. Place the L0SARAM and L3DPSARAM memory blocks into program memory on page 0. Place the other memory blocks into data memory on page 1. 13. In the Sections{} area, notice that a section called .reset has already been allocated. The .reset section is part of the rts2800_ml.lib, and is not needed. By putting the TYPE = DSECT modifier after its allocation, the linker will ignore this section and not allocate it. 14. Place the sections defined on the slide into the appropriate memories via the Sections{} area. Save your work and close the file.
Build and Load the Project

15. The top four buttons on the horizontal toolbar control code generation. Hover your mouse over each button as you read the following descriptions:
Button 1 2 3 4
Name Compile File Incremental Build Rebuild All Stop Build
Description Compile, assemble the current open file Compile, assemble only changed files, then link Compile, assemble all files, then link Stop code generation
16. Code Composer Studio can automatically load the output file after a successful build. On the menu bar click: Option Customize and select the Program/Project/CIO tab, then check Load Program After Build. Also, Code Composer Studio can automatically connect to the target when started. Select the Debug Properties tab, check Connect to the target at startup, then click OK. 17. Click the Build button and watch the tools run in the build window. Check for errors (we have deliberately put an error in Lab2.c). When you get an error, scroll the build window at the bottom of the Code Composer Studio screen until you see the error message (in red), and simply double-click the error message. The editor will automatically open the source file containing the error, and position the mouse cursor at the correct code line. 18. Fix the error by adding a semicolon at the end of the "z = x + y" statement. For future knowlege, realize that a single code error can sometimes generate multiple error messages at build time. This was not the case here. 19. Rebuild the project (there should be no errors this time). The output file should automatically load. The Program Counter should be pointing to _c_int00 in the Disassembly Window. 20. Under Debug on the menu bar click Go Main. This will run through the C-environment initialization routine in the rts2800_ml.lib and stop at main() in Lab2.c.
2 - 19
Debug Enviroment Windows

It is standard debug practice to watch local and global variables while debugging code. There are various methods for doing this in Code Composer Studio. We will examine two of them here: memory windows, and watch windows. 21. Open a memory window to view the global variable z. Click: View Memory on the menu bar.
Type &z into the address field and then enter. Note that you must use the ampersand (meaning "address of") when using a symbol in a memory window address box. Also note that Code Composer Studio is case sensitive. Set the properties format to Hex 16 Bit TI style at the bottom of the window. This will give you more viewable data in the window. You can change the contents of any address in the memory window by double-clicking on its value. This is useful during debug. 22. Open the watch window to view the local variables x and y. Click: View Watch Window on the menu bar.
Click the Watch Locals tab and notice that the local variables x and y are already present. The watch window will always contain the local variables for the code function currently being executed. (Note that local variables actually live on the stack. You can also view local variables in a memory window by setting the address to SP after the code function has been entered). 23. We can also add global variables to the watch window if desired. Let's add the global variable z. Click the Watch 1 tab at the bottom of the watch window. In the empty box in the Name column, type z and then enter. An ampersand is not used here. The watch window knows you are specifying a symbol. Check that the watch window and memory window both report the same value for z. Trying changing the value in one window, and notice that the value also changes in the other window.
Single-stepping the Code

24. Click the Watch Locals tab at the bottom of the watch window. Single-step through main() by using the <F11> key (or you can use the Single Step button on the vertical toolbar). Check to see if the program is working as expected. What is the value for z when you get to the end of the program? End of Exercise
2 - 20
2 - 21
Solutions
Solutions
Exercise 2 - Solution
MEMORY { PAGE 0: FLASH: PAGE 1: M0SARAM: M1SARAM: L0SARAM: } SECTIONS { .text: .ebss: .cinit: .stack: } /* origin /* origin origin origin Program Memory */ = 0x3E8000, length Data Memory */ = 0x000000, length = 0x000400, length = 0x008000, length
= 0x10000 = 0x400 = 0x400 = 0x800
> > > >
PAGE PAGE PAGE PAGE
= = = =
0 1 0 1
Lab 2: Solution - lab2.cmd

MEMORY { PAGE 0: L0SARAM: L3DPSARAM: PAGE 1: M0SARAM: M1SARAM: L1DPSARAM: L2DPSARAM: } SECTIONS { .text: .ebss: .cinit: .stack: .reset: } /* Program Memory */ origin = 0x008000, length origin = 0x009000, length /* Data Memory */ origin = 0x000000, length origin = 0x000400, length origin = 0x008800, length origin = 0x008C00, length = 0x0800 = 0x1000 = = = = 0x0400 0x0400 0x0400 0x0400
> > > > >
L0SARAM M0SARAM L0SARAM M1SARAM L0SARAM
PAGE PAGE PAGE PAGE PAGE
= = = = =
0 1 0 1 0, TYPE = DSECT
2 - 22
Peripherial Registers Header Files

Introduction
The purpose of the DSP2803x C-code header files is to simplify the programming of the many peripherals on the F28x device. Typically, to program a peripheral the programmer needs to write the appropriate values to the different fields within a control register. In its simplest form, the process consists of writing a hex value (or masking a bit field) to the correct address in memory. But, since this can be a burdensome and repetitive task, the C-code header files were created to make this a less complicated task. The DSP2803x C-code header files are part of a library consisting of C functions, macros, peripheral structures, and variable definitions. Together, this set of files is known as the header files. Registers and the bit-fields are represented by structures. C functions and macros are used to initialize or modify the structures (registers). In this module, you will learn how to use the header files and C programs to facilitate programming the peripherals.
Learning Objectives
Learning Objectives
Understand the usage of the F2803x C-Code Header Files Be able to program peripheral registers Understand how the structures are mapped with the linker command file
C2000 Piccolo Workshop - Peripheral Registers Header Files
3-1
Module Topics
Module Topics
Peripherial Registers Header Files .......................................................................................................... 3-1 Module Topics......................................................................................................................................... 3-2 Traditional and Structure Approach to C Coding .................................................................................. 3-3 Naming Conventions............................................................................................................................... 3-6 F2803x C-Code Header Files ................................................................................................................. 3-7 Peripheral Structure .h File ................................................................................................................. 3-7 Global Variable Definitions File ........................................................................................................ 3-9 Mapping Structures to Memory.........................................................................................................3-10 Linker Command File........................................................................................................................3-10 Peripheral Specific Routines..............................................................................................................3-11 Summary ................................................................................................................................................3-12
3-2
Traditional and Structure Approach to C Coding
Traditional Approach to C Coding

#define ADCCTL1 void main(void) { *ADCCTL1 = 0x1234; *ADCCTL1 |= 0x4000; } //write entire register //enable ADC module (volatile unsigned ... int *)0x00007100
Advantages
- Simple, fast and easy to type - Variable names exactly match register names (easy to remember)
Disadvantages
- Requires individual masks to be generated to manipulate individual bits - Cannot easily display bit fields in Watch window - Will generate less efficient code in many cases
Structure Approach to C Coding

void main(void) { AdcRegs.ADCCTL1.all = 0x1234; AdcRegs.ADCCTL1.bit.ADCENABLE = 1; } //write entire register //enable ADC module
Advantages
- Easy to manipulate individual bits. - Watch window is amazing! (next slide) - Generates most efficient code (on C28x)
Disadvantages
- Can be difficult to remember the structure names (Editor Auto Complete feature to the rescue!) - More to type (again, Editor Auto Complete feature to the rescue)
3-3
The CCS Watch Window using #define
The CCS Watch Window using Structures
3-4
Is the Structure Approach Efficient?

The structure approach enables efficient compiler use of DP addressing mode and C28x atomic operations
C Source Code
// Stop CPU Timer0 CpuTimer0Regs.TCR.bit.TSS = 1; // Load new 32-bit period value CpuTimer0Regs.PRD.all = 0x00010000; // Start CPU Timer0 CpuTimer0Regs.TCR.bit.TSS = 0;
Generated Assembly Code*

MOVW OR MOVL MOVL AND DP, #0030 @4, #0x0010 XAR4, #0x010000 @2, XAR4 @4, #0xFFEF
- Easy to read the code w/o comments - Bit mask built-in to structure
5 words, 5 cycles
You could not have coded this example any more efficiently with hand assembly!
* C28x Compiler v5.0.1 with -g and either -o1, -o2, or -o3 optimization level
Compare with the #define Approach

The #define approach relies heavily on less-efficient pointers for random memory access, and often does not take advantage of C28x atomic operations C Source Code
// Stop CPU Timer0 *TIMER0TCR |= 0x0010; // Load new 32-bit period value *TIMER0TPRD32 = 0x00010000; // Start CPU Timer0 *TIMER0TCR &= 0xFFEF;
Generated Assembly Code*

MOV ORB MOV MOVL MOVL MOVL MOV AND MOV @AL,*(0:0x0C04) AL, #0x10 *(0:0x0C04), @AL XAR5, #0x010000 XAR4, #0x000C0A *+XAR4[0], XAR5 @AL, *(0:0x0C04) @AL, #0xFFEF *(0:0x0C04), @AL
- Hard to read the code w/o comments - User had to determine the bit mask
9 words, 9 cycles
* C28x Compiler v5.0.1 with -g and either -o1, -o2, or -o3 optimization level
3-5
Naming Conventions
Naming Conventions
The header files use a familiar set of naming conventions. They are consistent with the Code Composer Studio configuration tool, and generated file naming conventions
Structure Naming Conventions

The DSP2803x header files define:
All of the peripheral structures All of the register names All of the bit field names All of the register addresses
PeripheralName.RegisterName.all PeripheralName.RegisterName.half.LSW PeripheralName.RegisterName.half.MSW PeripheralName.RegisterName.bit.FieldName // Access full 16 or 32-bit register // Access low 16-bits of 32-bit register // Access high 16-bits of 32-bit register // Access specified bit fields of register
Notes: [1] PeripheralName are assigned by TI and found in the DSP2803x header files. They are a combination of capital and small letters (i.e. CpuTimer0Regs). [2] RegisterName are the same names as used in the data sheet. They are always in capital letters (i.e. TCR, TIM, TPR,..). [3] FieldName are the same names as used in the data sheet. They are always in capital letters (i.e. POL, TOG, TSS,..).
Editor Auto Complete to the Rescue!
3-6
F2803x C-Code Header Files

The C-code header files consists of .h, c source files, linker command files, and other useful example programs, documentations and add-ins for Code Composer Studio.
DSP2803x Header File Package

(http://www.ti.com, literature # SPRC892)
Contains everything needed to use the structure approach Defines all peripheral register bits and register addresses Header file package includes:
\DSP2803x_headers\include \DSP2803x_headers\cmd \DSP2803x_headers\gel \DSP2803x_examples \DSP2803x_examples_ccsv4 \doc .h files linker .cmd files .gel files for CCS CCS3 examples CCS4 examples documentation
A peripheral is programmed by writing values to a set of registers. Sometimes, individual fields are written to as bits, or as bytes, or as entire words. Unions are used to overlap memory (register) so the contents can be accessed in different ways. The header files group all the registers belonging to a specific peripheral. A DSP2803x_Peripheral.gel GEL file can provide a pull down menu to load peripheral data structures into a watch window. Code Composer Studio can load a GEL file automatically. To include fuctions to the standard F28035.gel that is part of Code Composer Studio, add: GEL_LoadGel(base_path/gel/DSP2803x_Peripheral.gel) The GEL file can also be loaded during a Code Composer Studio session by clicking: File Load GEL
Peripheral Structure .h File

The DSP2803x_Device.h header file is the main include file. By including this file in the .c source code, all of the peripheral specific .h header files are automatically included. Of course, each specific .h header file can be included individually in an application that does not use all the header files, or you can comment out the ones you do not need. (Also includes typedef statements).
3-7
Peripheral Structure .h files (1 of 2)

Contain bits field structure definitions for each peripheral register
DSP2803x_Adc.h
// ADC Individual Register Bit Definitions: struct ADCCTL1_BITS { Uint16 TEMPCONV:1; // bits description // 0 Temperature sensor connection
Uint16 VREFLOCONV:1; // 1 VSSA connection
Your C-source file (e.g., Adc.c)

#include "DSP2803x_Device.h" Void InitAdc(void) { /* Reset the ADC m odule */ AdcRegs.ADCCTL1.bit.RESET = 1; /* configure the ADC register */ AdcRegs.ADCCTL1.all = 0x00E4; };
Uint16 INTPULSEPOS:1; // 2 INT pulse generation control Uint16 ADCREFSEL:1; Uint16 rsvd1:1; Uint16 ADCREFPWD:1; Uint16 ADCBGPWD:1; Uint16 ADCPWDN:1; Uint16 ADCBSYCHN:5; Uint16 ADCBSY:1; Uint16 ADCENABLE:1; Uint16 RESET:1; }; // Allow access to the bit fields or entire register: union ADCCTL1_REG { Uint16 }; // ADC External References & Function Declarations: extern volatile struct ADC_REGS AdcRegs; all; struct ADCCTL1_BITS bit; // 3 Internal/external reference select // 4 reserved // 5 Reference buffers powerdown // 6 ADC bandgap powerdow n // 7 ADC powerdown // 12:8 ADC busy on a channel // 13 ADC busy signal // 14 ADC enable // 15 ADC master reset
Peripheral Structure .h files (2 of 2)

The header file package contains a .h file for each peripheral in the device
DSP2803x_Adc.h DSP2803x_Comp.h DSP2803x_Device.h DSP2803x_EPwm.h DSP2803x_I2c.h DSP2803x_PieCtrl.h DSP2803x_Spi.h DSP2803x_BootVars.h DSP2803x_ECan.h DSP2803x_EQep.h DSP2803x_Lin.h DSP2803x_PieVect.h DSP2803x_SysCtrl.h DSP2803x_Cla.h DSP2803x_ECap.h DSP2803x_Gpio.h DSP2803x_NmiIntrupt.h DSP2803x_Sci.h DSP2803x_XIntrupt.h DSP2803x_CpuTimers.h DSP2803x_DevEmu.h
DSP2803x_Device.h
Main include file Will include all other .h files Include this file (directly or indirectly) in each source file:
#include DSP2803x_Device.h
3-8
Global Variable Definitions File

With DSP2803x_GlobalVariableDefs.c included in the project all the needed variable definitions are globally defined.
Global Variable Definitions File

DSP2803x_GlobalVariableDefs.c
Declares a global instantiation of the structure for each peripheral Each structure is placed in its own section using a DATA_SECTION pragma to allow linking to the correct memory (see next slide)
#include "DSP2803x_Device.h" #pragma DATA_SECTION(AdcRegs,"AdcRegsFile"); volatile struct ADC_REGS AdcRegs;
Add this file to your CCS project:

3-9
Mapping Structures to Memory

The data structures describe the register set in detail. And, each instance of the data type (i.e., register set) is unique. Each structure is associated with an address in memory. This is done by (1) creating a new section name via a DATA_SECTION pragma, and (2) linking the new section name to a specific memory in the linker command file.
Linker Command Files for the Structures

DSP2803x_nonBIOS.cmd and DSP2803x_BIOS.cmd
#include "DSP2803x_Device.h" #pragma DATA_SECTION(AdcRegs,"AdcRegsFile"); volatile struct ADC_REGS AdcRegs;
Links each structure to the address of the peripheral using the structures named section non-BIOS and BIOS versions of the .cmd file Add one of these files to your CCS project:
DSP2803x_nonBIOS.cmd or DSP2803x_BIOS.cmd
DSP2803x_Headers_nonBIOS.cmd
MEMORY { PAGE1: ... ADC: ... }
origin=0x007100, length=0x000080
SECTIONS { ... AdcRegsFile: ... }
>
ADC
PAGE = 1
Linker Command File

When using the header files, the user adds the MEMORY regions that correspond to the CODE_SECTION and DATA_SECTION pragmas found in the .h and global-definitons.c file. The user can modify their own linker command file, or use a pre-configured linker command file such as F28035.cmd. This file has the peripheral memory regions defined and tied to the individual peripheral.
3 - 10
Peripheral Specific Routines

Peripheral Specific C functions are used to initialize the peripherals. They are used by adding the appropriate .c file to the project.
Peripheral Specific Examples

Example projects for each peripheral Helpful to get you started
3 - 11
Summary
Summary
Peripheral Register Header Files Summary

Easier code development Easy to use Generates most efficient code Increases effectiveness of CCS watch window TI has already done all the work!
Use the correct header file package for your device:
F2803x F2802x F2833x and F2823x F280x and F2801x F2804x F281x # SPRC892 # SPRC832 # SPRC530 # SPRC191 # SPRC324 # SPRC097
Go to http://www.ti.com and enter the literature number in the keyword search box
3 - 12
Reset and Interrupts

Introduction
This module describes the interrupt process and explains how the Peripheral Interrupt Expansion (PIE) works.
Learning Objectives
Learning Objectives
Describe the C28x reset process List the event sequence during an interrupt Describe the C28x interrupt structure
C2000 Piccolo Workshop - Reset and Interrupts
4-1
Module Topics
Module Topics
Reset and Interrupts ................................................................................................................................. 4-1 Module Topics......................................................................................................................................... 4-2 Reset........................................................................................................................................................ 4-3 Reset - Bootloader .............................................................................................................................. 4-3 Emulation Boot Mode ........................................................................................................................ 4-4 Stand-Alone Boot Mode..................................................................................................................... 4-4 Reset Code Flow Summary ............................................................................................................. 4-5 Interrupts ................................................................................................................................................ 4-6 Interrupt Processing............................................................................................................................ 4-6 Interrupt Flag Register (IFR) .............................................................................................................. 4-7 Interrupt Enable Register (IER).......................................................................................................... 4-7 Interrupt Global Mask Bit (INTM)..................................................................................................... 4-8 Peripheral Interrupt Expansion (PIE) ................................................................................................. 4-8 PIE Interrupt Vector Table ................................................................................................................4-10 Interrupt Response and Latency ........................................................................................................4-11
4-2
Reset
Reset
Reset Sources
Missing Clock Detect Watchdog Timer Power-on Reset Brown-out Reset XRS pin active
Logic shown is functional representation, not actual implementation
C28x core
XRS
To XRS pin
POR Power-On Rest generates a device reset during power-up conditions BOR Brown-Out Reset generates a device reset if the power supply drops below specification for the device Note: Devices support an on-chip regulator (VREG) to generate the core voltage
Reset - Bootloader
Reset Bootloader
Reset
OBJMODE = 0 AMODE = 0 ENPIE = 0 INTM = 1
Reset vector fetched from boot ROM 0x3F FFC0
Bootloader sets
OBJMODE = 1 AMODE = 0
YES TRST = 1
Emulator Connected ?
NO TRST = 0

Stand-alone Boot Boot determined by 2 GPIO pins and 2 OTP locations:

OTP_KEY and OTP_BMODE
TRST = JTAG Test Reset
EMU_KEY & EMU_BMODE located in PIE at 0x0D00 & 0x0D01, respectively OPT_KEY & OTP_BMODE located in OTP at 0x3D78FE & 0x3D78FF, respectively
4-3
Reset
Emulation Boot Mode
Emulation Boot Mode (TRST = 1)

Emulator Connected

If either EMU_KEY or EMU_BMODE are invalid, the wait boot mode is used. These values can then be modified using the debugger and a reset issued to restart the boot process NO
EMU_KEY = 0x55AA ? YES EMU_BMODE = 0x0000 0x0001 0x0002 0x0003 0x0004 0x0005 0x0006 0x0007 0x000A 0x000B other Boot Mode Parallel I/O SCI Wait GetMode SPI I2C OTP CAN M0 SARAM FLASH Wait
Boot Mode Wait
OTP_KEY = 0x55AA ? YES OTP_BMODE = 0x0001 0x0003 0x0004 0x0005 0x0006 0x0007 other
NO
Boot Mode FLASH
Boot Mode SCI FLASH SPI I2C OTP CAN FLASH
Stand-Alone Boot Mode
Stand-Alone Boot Mode (TRST = 0)

Emulator Not Connected
Stand-alone Boot Boot determined by 2 GPIO pins and 2 OTP locations:

OTP_KEY and OTP_BMODE
Note that the boot behavior for unprogrammed OTP is the FLASH boot mode
OTP_KEY = 0x55AA ? GPIO 37 0 0 1 1 GPIO 34 0 1 0 1 YES Boot Mode Parallel I/O SCI Wait GetMode OTP_BMODE = 0x0001 0x0003 0x0004 0x0005 0x0006 0x0007 other
NO
Boot Mode FLASH
Boot Mode SCI FLASH SPI I2C OTP CAN FLASH
4-4
Reset
Reset Code Flow Summary
Reset Code Flow - Summary

0x000000
0x000000
M0 SARAM (1Kw) 0x3D7800

0x3D7800
OTP (1Kw) 0x3E8000 FLASH (64Kw)

0x3F7FF6
0x3FE000
Boot ROM (8Kw) Boot Code

0x3FF7BB

Execution Entry determined by Emulation Boot Mode or Stand-Alone Boot Mode
RESET
0x3FFFC0
BROM vector (64w) 0x3FF7BB
Bootloading Routines (SCI, SPI, I2C, Parallel I/O)
4-5
Interrupts
Interrupts
Interrupt Sources
Internal Sources
TINT2 TINT1 TINT0 ePWM, eCAP, eQEP, ADC, SCI, SPI, I2C, eCAN, LIN, CLA, WD PIE (Peripheral Interrupt Expansion)
C28x CORE
XRS NMI INT1 INT2 INT3 INT12 INT13 INT14
External Sources
XINT1 XINT3 TZx XRS
Interrupt Processing
Maskable Interrupt Processing

Conceptual Core Overview Core Interrupt
INT1 INT2
(IFR) Latch
1 0
(IER) (INTM) Switch Global Switch
C28x Core
INT14
A valid signal on a specific interrupt line causes the latch to display a 1 in the appropriate bit If the individual and global switches are turned on the interrupt reaches the core
4-6
Interrupts
Interrupt Flag Register (IFR)
Interrupt Flag Register (IFR)

15 14 13
INT14
12
INT13
11
INT12
10
INT11
9
INT10
8
INT9
RTOSINT DLOGINT
7
INT8
6
INT7
5
INT6
4
INT5
3
INT4
2
INT3
1
INT2
0
INT1
Pending : Absent :
IFR Bit = 1 IFR Bit = 0
/*** Manual setting/clearing IFR ***/ extern cregister volatile unsigned int IFR; IFR |= 0x0008; IFR &= 0xFFF7; //set INT4 in IFR //clear INT4 in IFR
Compiler generates atomic instructions (non-interruptible) for setting/clearing IFR If interrupt occurs when writing IFR, interrupt has priority IFR(bit) cleared when interrupt is acknowledged by CPU Register cleared on reset
Interrupt Enable Register (IER)
Interrupt Enable Register (IER)

15 14 13
INT14
12
INT13
11
INT12
10
INT11
9
INT10
8
INT9
RTOSINT DLOGINT
7
INT8
6
INT7
5
INT6
4
INT5
3
INT4
2
INT3
1
INT2
0
INT1
Enable: Set Disable: Clear
IER Bit = 1 IER Bit = 0
/*** Interrupt Enable Register ***/ extern cregister volatile unsigned int IER; IER |= 0x0008; IER &= 0xFFF7; //enable INT4 in IER //disable INT4 in IER
Compiler generates atomic instructions (non-interruptible) for setting/clearing IER Register cleared on reset
4-7
Interrupts
Interrupt Global Mask Bit (INTM)
Interrupt Global Mask Bit

Bit 0
ST1
INTM
INTM used to globally enable/disable interrupts:

Enable: Disable: INTM = 0 INTM = 1 (reset value)
INTM modified from assembly code only: /*** Global Interrupts ***/ asm( CLRC INTM); asm( SETC INTM); //enable global interrupts //disable global interrupts
Peripheral Interrupt Expansion (PIE)
Peripheral Interrupt Expansion - PIE

PIE module for 96 Interrupts
Peripheral Interrupts 12x8 = 96
Interrupt Group 1 PIEIFR1 PIEIER1 INT1.1 INT1.2 1 0
INT1.x interrupt group INT2.x interrupt group INT3.x interrupt group INT4.x interrupt group INT5.x interrupt group

INT1.8 1
INT1
96
INT6.x interrupt group INT7.x interrupt group INT8.x interrupt group INT9.x interrupt group INT10.x interrupt group INT11.x interrupt group INT12.x interrupt group
INT1 INT12
28x Core Interrupt logic
INTM
12 Interrupts
IER
IFR
28x Core
(TINT1) (TINT2)
4-8
Interrupts
F2803x PIE Interrupt Assignment Table

INTx.8 INT1 INT2 INT3 INT4 INT5 INT6 INT7 INT8 INT9 INT10 INT11 INT12
ADCINT8 CLA1 _INT8 LUF ADCINT7 CLA1 _INT7 LVF ECAN1 INTA ADCINT6 CLA1 _INT6 ECAN0 INTA ADCINT5 CLA1 _INT5 LIN1 INTA ADCINT4 CLA1 _INT4 LIN0 INTA ADCINT3 CLA1 _INT3 I2CINT2A I2CINT1A SCITX INTA ADCINT2 CLA1 _INT2 SCIRX INTA ADCINT1 CLA1 _INT1 XINT3 SPITX INTB SPIRX INTB SPITX INTA WAKEINT
INTx.7
TINT0 EPWM7 _TZINT EPWM7 _INT
INTx.6
ADCINT9 EPWM6 _TZINT EPWM6 _INT
INTx.5
XINT2 EPWM5 _TZINT EPWM5 _INT
INTx.4
XINT1 EPWM4 _TZINT EPWM4 _INT
INTx.3
INTx.2
ADCINT2
INTx.1
ADCINT1 EPWM1 _TZINT EPWM1 _INT ECAP1 _INT EQEP1 _INT SPIRX INTA
EPWM3 _TZINT EPWM3 _INT
EPWM2 _TZINT EPWM2 _INT
PIE Registers
PIEIFRx register
15 - 8 7
(x = 1 to 12)
6 5 4 3 2 1 0
reserved
INTx.8 INTx.7 INTx.6 INTx.5 INTx.4 INTx.3 INTx.2 INTx.1
PIEIERx register
15 - 8 7
(x = 1 to 12)
6 5 4 3 2 1 0
reserved
INTx.8 INTx.7 INTx.6 INTx.5 INTx.4 INTx.3 INTx.2 INTx.1
PIE Interrupt Acknowledge Register (PIEACK)

15 - 12 11 10 9 8 7 6 5 4 3 2 1 0
reserved
PIEACKx
15 - 1 0
PIECTRL register
PIEVECT
ENPIE
#include DSP2803x_Device.h PieCtrlRegs.PIEIFR1.bit.INTx4 = 1; PieCtrlRegs.PIEIER3.bit.INTx2 = 1; PieCtrlRegs.PIEACK.all = 0x0004; //manually set IFR for XINT1 in PIE group 1 //enable EPWM2_INT in PIE group 3 //acknowledge the PIE group 3
PieCtrlRegs.PIECTRL.bit.ENPIE = 1; //enable the PIE
4-9
Interrupts
PIE Interrupt Vector Table
Default Interrupt Vector Table at Reset

Vector RESET INT1 INT2 INT3 INT4 INT5 INT6 INT7 INT8 INT9 INT10 INT11 INT12 INT13 INT14 DATALOG RTOSINT EMUINT NMI ILLEGAL USER 1-12 Offset 00 02 04 06 08 0A 0C 0E 10 12 14 16 18 1A 1C 1E 20 22 24 26 28-3E
Default Vector Table Re-mapped when ENPIE = 1
Memory
0x00 0D00
PIE Vectors 256w
BROM Vectors 0x3F FFC0 64w

ENPIE = 0
0x3F FFFF
PieVectTableInit{ } Used to initialize PIE vectors
PIE Vector Mapping (ENPIE = 1)

Vector Name Reset INT1 INT12 INT13 INT14 DATALOG USER12 INT1.1 INT1.8 INT12.1 INT12.8 PIE Address 0x00 0D00 0x00 0D02 0x00 0D18 0x00 0D1A 0x00 0D1C 0x00 0D1E 0x00 0D3E 0x00 0D40 0x00 0D4E 0x00 0DF0 0x00 0DFE PIE Vector Description Reset fetched from Boot ROM 0x3F FFC0 INT1 remapped to PIE group below INTx remapped to PIE group below INT12 remapped to PIE group below CPU Timer 1 CPU Timer 2 CPU Data Logging Interrupt User Defined Trap PIE INT1.1 Interrupt Vector PIE INT1.8 Interrupt Vector PIE INT12.1 Interrupt Vector PIE INT12.8 Interrupt Vector
Remapped
PIE vector location 0x00 0D00 256 words in data memory RESET and INT1-INT12 vector locations are re-mapped CPU vectors are re-mapped to 0x00 0D00 in data memory
4 - 10
Interrupts
Device Vector Mapping - Summary

RESET <0x3F FFC0>
Reset Vector User Code Start
<0x3F F7BB> = Boot Code < _c_int00 >
Flash Entry Point <0x3F 7FF6 > = LB _c_int00
_c_int00: . . . CALL main() Initialization() { Load PIE Vectors Enable the PIE Enable PIEIER Enable Core IER Enable INTM }
main() { initialization(); . . . }
PIE Vector Table 256 Word RAM 0x00 0D00 0DFF
Interrupt Response and Latency
Interrupt Response - Hardware Sequence

CPU Action
Registers stack 0 IFR (bit) 0 IER (bit) 1 INTM/DBGM Vector PC Clear other status bits
Description
14 Register words auto saved Clear corresponding IFR bit Clear corresponding IER bit Disable global ints/debug events Loads PC with int vector address Clear LOOP, EALLOW, IDLESTAT
Note: some actions occur simultaneously, none are interruptible T AH PH AR1 DP DBSTAT PC(msw) ST0 AL PL AR0 ST1 IER PC(lsw)
4 - 11
Interrupts
Interrupt Latency
Latency
ext. interrupt occurs here Internal interrupt occurs here Assumes ISR in internal RAM cycles
2
Sync ext. signal (ext. interrupt only)
3
F1/F2/D1 of ISR instruction (3 reg. pairs saved)
Recognition Get vector delay (3), SP and place alignment (1), in PC interrupt (3 reg. placed in pairs pipeline saved)
Save D2/R1/R2 of return ISR address instruction
ISR instruction executed on next cycle
Minimum latency (to when real work occurs in the ISR): Internal interrupts: 14 cycles External interrupts: 16 cycles Maximum latency: Depends on wait states, INTM, etc.
4 - 12
System Initialization
Introduction
This module discusses the operation of the OSC/PLL-based clock module and watchdog timer. Also, the general-purpose digital I/O ports, external interrups, various low power modes and the EALLOW protected registers will be covered.
Learning Objectives
Learning Objectives
OSC/PLL Clock Module Watchdog Timer General Purpose Digital I/O External Interrupts Low Power Modes Register Protection
C2000 Piccolo Workshop - System Initialization
5-1
Module Topics
Module Topics
System Initialization.................................................................................................................................. 5-1 Module Topics......................................................................................................................................... 5-2 Oscillator/PLL Clock Module................................................................................................................. 5-3 Watchdog Timer...................................................................................................................................... 5-6 General-Purpose Digital I/O .................................................................................................................5-10 External Interrupts.................................................................................................................................5-13 Low Power Modes..................................................................................................................................5-14 Register Protection ................................................................................................................................5-16 Lab 5: System Initialization ...................................................................................................................5-18
5-2
Oscillator/PLL Clock Module
F2803x Oscillator / PLL Clock Module

(lab file: SysCtrl.c)
WDCLKSRCSEL
Internal OSC 1 (10 MHz)
OSC1CLK
0* WDCLK 1
OSCCLKSRCSEL
Watchdog Module
OSCCLKSRC2
MUX
Internal OSC2CLK OSC 2 (10 MHz)
1 0*
0* 1
OSCCLK (PLL bypass) VCOCLK
DIVSEL
1/n
CLKIN
PLL
XCLKINOFF
C28x Core
SYSCLKOUT
DIV
XCLKIN
0
X1 XTAL
0* 1
XTAL OSC
EXTCLK
TMR2CLKSRCSEL
LOSPCP
LSPCLK SCI, SPI All other peripherals clocked by SYSCLKOUT * = default
SYSCLKOUT
10 11 CPUTMR2CLK 01 00* CPU

Timer 2
X2
The on-chip oscillator and phase-locked loop (PLL) block provide all the necessary clocking signals for the F2803x devices. The two internal oscillators (INTOSC1 and INTOSC2) need no external components.
F2803x PLL and LOSPCP

MUX
OSCCLK (PLL bypass) VCOCLK SysCtrlRegs.PLLSTS.bit.DIVSEL CLKIN
1/n
C28x Core
SYSCLKOUT
PLL
LOSPCP
LSPCLK
SysCtrlRegs.LOSPCP.bit.LSPCLK SysCtrlRegs.PLLCR.bit.DIV DIV 0 00 0 0 00 1 0 01 0 0 01 1 0 10 0 0 10 1 0 11 0 0 11 1 1 00 0 1 00 1 1 01 0 1 01 1 1 10 0 CLKIN OSCCLK / n * (PLL bypass) OSCCLK x 1 / n OSCCLK x 2 / n OSCCLK x 3 / n OSCCLK x 4 / n OSCCLK x 5 / n OSCCLK x 6 / n OSCCLK x 7 / n OSCCLK x 8 / n OSCCLK x 9 / n OSCCLK x 10 / n OSCCLK x 11 / n OSCCLK x 12 / n DIVSEL 0x 10 11 n /4 * /2 /1 LSPCLK 000 001 010 011 100 101 110 111 Peripheral Clk Freq SYSCLKOUT / 1 SYSCLKOUT / 2 SYSCLKOUT / 4 * SYSCLKOUT / 6 SYSCLKOUT / 8 SYSCLKOUT / 10 SYSCLKOUT / 12 SYSCLKOUT / 14
* default Note: /1 mode can only be used when PLL is bypassed
Input Clock Fail Detect Circuitry PLL will issue a limp mode clock (1-4 MHz) if input clock is removed after PLL has locked. An internal device reset will also be issued (XRSn pin not driven).
5-3
The PLL has a 4-bit ratio control to select different CPU clock rates. In addition to the on-chip oscillators, two external modes of operation are supported crystal operation, and external clock source operation. Crystal operation allows the use of an external crystal/resonator to provide the time base to the device. External clock source operation allows the internal (crystal) oscillator to be bypassed, and the device clocks are generated from an external clock source input on the XCLKIN pin. The C28x core provides a SYSCLKOUT clock signal. This signal is prescaled to provide a clock source for some of the on-chip communication peripherals through the low-speed peripheral clock prescaler. Other peripherals are clocked by SYSCLKOUT and use their own clock prescalers for operation.
Clock Control Register

Upper Register:
Watchdog HALT Mode Ignore 0 = automatic turn on/off 1 = ignores HALT Mode Internal Oscillator 1 HALT Mode Ignore 0 = automatic turn on/off 1 = ignores HALT Mode Internal Oscillator 1 Off 0 = on 1 = off
15
14
13
12
11
10
NMIRESET XTAL SEL OSCOFF
XCLKIN OFF
WDHALTI INTOSC2 HALTI
INTOSC2 OFF
INTOSC1 INTOSC1 HALTI OFF
NMI Reset 0 = no delay 1 = delay
Crystal Oscillator Off 0 = on 1 = off
XCLKIN Off 0 = on 1 = off
Internal Oscillator 2 HALT Mode Ignore 0 = automatic turn on/off 1 = ignores HALT Mode
Internal Oscillator 2 Off 0 = on 1 = off
0 = default
5-4
Clock Control Register

Lower Register:
Watchdog Clock Source 0 = internal OSC1 1 = external or internal OSC2
7-5 4-3 2
Oscillator Clock Source 0 = internal OSC1 1 = external or internal OSC2

1 0
TMR2CLKPRESCALE
TMR2CLKSRCSEL
WDCLK SRCSEL
OSCCLK OSCCLK SRC2SEL SRCSEL
CPU Timer 2 Clock Prescale 000 = /1 001 = /2 010 = /4 011 = /8 100 = /16 1xx = reserved
CPU Timer 2 Clock Source 00 = SYSCLKOUT 01 = external 10 = internal OSC1 11 = internal OSC2
Oscillator 2 Clock Source 0 = external 1 = internal OSC2
0 = default
The peripheral clock control register allows individual peripheral clock signals to be enabled or disabled. If a peripheral is not being used, its clock signal could be disabled, thus reducing power consumption.
Peripheral Clock Control Registers

12 11
SysCtrlRegs.PCLKCR0
15 14 13 10 9 8
reserved
7
ECANA ENCLK
6
reserved
5
reserved
4
reserved
3
SCIA ENCLK
2
SPIA ENCLK
1
SPIA ENCLK
0
reserved
15
reserved
14
reserved
13
I2CA ENCLK
12
ADC ENCLK
11
TBCLK SYNC
10
LINA ENCLK
9
HRPWM ENCLK
8
SysCtrlRegs.PCLKCR1
reserved
7
EQEP1 ENCLK
6
reserved
5
reserved
4
reserved
3
reserved
2
reserved
1
ECAP1 ENCLK
0
reserved
15
EPWM7 ENCLK
14
EPWM6 ENCLK
13
EPWM5 ENCLK
12
EPWM4 ENCLK
11
EPWM3 ENCLK
10
EPWM2 ENCLK
9
EPWM1 ENCLK
8
SysCtrlRegs.PCLKCR3
reserved
7
CLA1 ENCLK
6
GPIOIN ENCLK
5
reserved
4
reserved CPUTIMER2 CPUTIMER1 CPUTIMER0 ENCLK ENCLK ENCLK

3 2 1 0
reserved
reserved
reserved
reserved
reserved
COMP3 ENCLK
COMP2 ENCLK
COMP1 ENCLK
Module Enable Clock Bit 1 = enable 0 = disable (default)
5-5
Watchdog Timer
Watchdog Timer
Watchdog Timer
Resets the C28x if the CPU crashes
Watchdog counter runs independent of CPU If counter overflows, a reset or interrupt is triggered (user selectable) CPU must write correct data key sequence to reset the counter before overflow
Watchdog must be serviced or disabled within 131,072 WDCLK cycles after reset This translates to 13.11 ms with a 10 MHz WDCLK
The watchdog timer provides a safeguard against CPU crashes by automatically initiating a reset if it is not serviced by the CPU at regular intervals. In motor control applications, this helps protect the motor and drive electronics when control is lost due to a CPU lockup. Any CPU reset will revert the PWM outputs to a high-impedance state, which should turn off the power converters in a properly designed system. The watchdog timer is running immediately after system power-up/reset, and must be dealt with by software soon after. Specifically, you have 13.11 ms (for a 60 MHz device) after any reset before a watchdog initiated reset will occur. This translates into 131,072 WDCLK cycles, which is a seemingly tremendous amount! Indeed, this is plenty of time to get the watchdog configured as desired and serviced. A failure of your software to properly handle the watchdog after reset could cause an endless cycle of watchdog initiated resets to occur.
5-6
Watchdog Timer
Watchdog Timer Module (lab file: Watchdog.c)

WDPS WDCLK WDOVERRIDE
/512
Watchdog Prescaler
WDDIS
8-bit Watchdog Counter

CLR System Reset WDCHK 2-0
WDRST
Output Pulse
WDINT
55 + AA Detector Watchdog Reset Key Register
Good Key
3 1 0 1 3
/ /
Bad WDCHK Key
Watchdog Period Selection

WDPS Bits 00x: 010: 011: 100: 101: 110: 111: * reset default FRC rollover 1 2 4 8 16 32 64 WD timeout period @ 10 MHz WDCLK 13.11 ms * 26.22 ms 52.44 ms 104.88 ms 209.76 ms 419.52 ms 839.04 ms
Remember: Watchdog starts counting immediately after reset is released! Reset default with WDCLK = 10 MHz computed as (1/10 MHz) * 512 * 256 = 13.11 ms
5-7
Watchdog Timer
Watchdog Timer Control Register

SysCtrlRegs.WDCR (lab file: Watchdog.c)
WD Flag Bit Gets set when the WD causes a reset
Writing a 1 clears this bit Writing a 0 has no effect
15 - 8 7 6 5-3 2-0
reserved
WDFLAG
WDDIS
WDCHK
WDPS
Watchdog Disable Bit
Logic Check Bits Write as 101 or reset immediately triggered
WD Prescale Selection Bits

WDPS 0 00 0 01 0 10 0 11 1 00 1 01 1 10 1 11 WDCLK = OSCCLK / 512 / OSCCLK / 512 / OSCCLK / 512 / OSCCLK / 512 / OSCCLK / 512 / OSCCLK / 512 / OSCCLK / 512 / OSCCLK / 512 / 1 1 2 4 8 16 32 64
Write 1 to disable (Functions only if WD OVERRIDE bit in SCSR is equal to 1)
Resetting the Watchdog

15 - 8 7-0
SysCtrlRegs.WDKEY (lab file: Watchdog.c)
reserved
WDKEY
WDKEY write values:

55h - counter enabled for reset on next AAh write AAh - counter set to zero if reset enabled
Writing any other value has no effect Watchdog should not be serviced solely in an ISR
If main code crashes, but interrupt continues to execute, the watchdog will not catch the crash Could put the 55h WDKEY in the main code, and the AAh WDKEY in an ISR; this catches main code crashes and also ISR crashes
5-8
Watchdog Timer
WDKEY Write Results

Sequential Value Written Result to WDKEY Step 1 2 3 4 5 6 7 8 9 10 11 12 13 14 AAh AAh 55h 55h 55h AAh AAh 55h AAh 55h 23h AAh 55h AAh No action No action WD counter enabled for reset on next AAh write WD counter enabled for reset on next AAh write WD counter enabled for reset on next AAh write WD counter is reset No action WD counter enabled for reset on next AAh write WD counter is reset WD counter enabled for reset on next AAh write No effect; WD counter not reset on next AAh write No action due to previous invalid value WD counter enabled for reset on next AAh write WD counter is reset
System Control and Status Register

SysCtrlRegs.SCSR (lab file: Watchdog.c)
WD Override (protect bit)
Protects WD from being disabled 0 = WDDIS bit in WDCR has no effect (WD cannot be disabled) 1 = WDDIS bit in WDCR can disable the watchdog This bit is a clear-only bit (write 1 to clear) The reset default of this bit is a 1
15 - 3
reserved
WDINTS WDENINT WDOVERRIDE
WD Interrupt Status (read only)

0 = active 1 = not active
WD Enable Interrupt
0 = WD generates a DSP reset 1 = WD generates a WDINT interrupt
5-9
General-Purpose Digital I/O
F2803x GPIO Grouping Overview

(lab file: Gpio.c)
GPIO Port A Mux1 Register (GPAMUX1) [GPIO 0 to 15] GPIO Port A Mux2 Register (GPAMUX2) [GPIO 16 to 31] Internal Bus GPIO Port A Direction Register (GPADIR) [GPIO 0 to 31]
Input Qual
GPIO Port A
Input
GPIO Port B Mux1 Register (GPBMUX1) [GPIO 32 to 44] GPIO Port B Direction Register (GPBDIR) [GPIO 32 to 44]
GPIO Port B
Qual
ANALOG Port
ANALOG I/O Mux1 Register (AIOMUX1) [AIO 0 to 15]
ANALOG Port Direction Register (AIODIR) [AIO 0 to 15]
F2803x GPIO Pin Block Diagram

(lab file: Gpio.c)
Peripheral 1 GPxSET GPxCLEAR GPxTOGGLE GPxDAT Out I/O DAT Bit (R/W) In I/O DIR Bit 0 = Input 1 = Output GPxDIR Peripheral 2 Peripheral 3
10 11 00
01
GPxMUX1 GPxMUX2 MUX Control Bits * 00 = GPIO 01 = Peripheral 1 10 = Peripheral 2 11 = Peripheral 3
GPxPUD Internal Pull-Up 0 = enable (default GPIO 12-44) 1 = disable (default GPIO 0-11)
Input Qualification
(GPIO 0-44)
GPxQSEL1 GPxQSEL2 GPxCTRL
Pin
* See device datasheet for pin function selection matrices
5 - 10
F2803x GPIO Input Qualification

Input Qualification
SYSCLKOUT
pin
to GPIO and peripheral modules
Qualification available on ports A & B (GPIO 0 - 44) only Individually selectable per pin samples taken
no qualification (peripherals only) sync to SYSCLKOUT only qualify 3 samples qualify 6 samples
AIO pins are fixed as sync to SYSCLKOUT
T = qual period
F2803x GPIO Input Qual Registers

GpioCtrlRegs.register (lab file: Gpio.c)
GPAQSEL1 / GPAQSEL2 / GPBQSEL1

31 0
16 pins configured per register
00 = 01 = 10 = 11 =
sync to SYSCLKOUT only * qual to 3 samples qual to 6 samples no sync or qual (for peripheral only; GPIO same as 00)
GPACTRL / GPBCTRL
31 24 16 8 0
QUALPRD3
QUALPRD2
QUALPRD1
QUALPRD0
B: A:
reserved GPIO31-24 00h 01h 02h FFh
reserved GPIO23-16
GPIO44-40 GPIO15-8
GPIO39-32 GPIO7-0
no qualification (SYNC to SYSCLKOUT) * QUALPRD = SYSCLKOUT/2 QUALPRD = SYSCLKOUT/4 QUALPRD = SYSCLKOUT/510

* reset default
5 - 11
F2803x GPIO Control Registers

GpioCtrlRegs.register (lab file: Gpio.c)
Register GPACTRL GPAQSEL1 GPAQSEL2 GPAMUX1 GPAMUX2 GPADIR GPAPUD GPBCTRL GPBQSEL1 GPBMUX1 GPBDIR GPBPUD AIOMUX1 AIODIR Description GPIO A Control Register [GPIO 0 31] GPIO A Qualifier Select 1 Register [GPIO 0 15] GPIO A Qualifier Select 2 Register [GPIO 16 31] GPIO A Mux1 Register [GPIO 0 15] GPIO A Mux2 Register [GPIO 16 31] GPIO A Direction Register [GPIO 0 31] GPIO A Pull-Up Disable Register [GPIO 0 31] GPIO B Control Register [GPIO 32 44] GPIO B Qualifier Select 1 Register [GPIO 32 44] GPIO B Mux1 Register [GPIO 32 44] GPIO B Direction Register [GPIO 32 44] GPIO B Pull-Up Disable Register [GPIO 32 44] ANALOG I/O Mux1 Register [AIO 0 15] ANALOG I/O Direction Register [AIO 0 15]
F2803x GPIO Data Registers

GpioDataRegs.register (lab file: Gpio.c)
Description GPIO A Data Register [GPIO 0 31] GPIO A Data Set Register [GPIO 0 31] GPIO A Data Clear Register [GPIO 0 31] GPIO A Data Toggle [GPIO 0 31] GPIO B Data Register [GPIO 32 44] GPIO B Data Set Register [GPIO 32 44] GPIO B Data Clear Register [GPIO 32 44] GPIO B Data Toggle [GPIO 32 44] ANALOG I/O Data Register [AIO 0 15] ANALOG I/O Data Set Register [AIO 0 15] ANALOG I/O Data Clear Register [AIO 0 15] ANALOG I/O Data Toggle [AIO 0 15] Register GPADAT GPASET GPACLEAR GPATOGGLE GPBDAT GPBSET GPBCLEAR GPBTOGGLE AIODAT AIOSET AIOCLEAR AIOTOGGLE
5 - 12
External Interrupts
External Interrupts
External Interrupts
3 external interrupt signals: XINT1, XINT2 and XINT3 XINT1, XINT2 and XINT3 can be mapped to any of GPIO0-31 XINT1, XINT2 and XINT3 also each have a free-running 16-bit counter that measures the elapsed time between interrupts
The counter resets to zero each time the interrupt occurs
External Interrupt Registers
Interrupt
Pin Selection Register (GpioIntRegs.register) GPIOXINT1SEL GPIOXINT2SEL GPIOXINT3SEL
Configuration Register (XIntruptRegs.register) XINT1CR XINT2CR XINT3CR
Counter Register (XIntruptRegs.register) XINT1CTR XINT2CTR XINT3CTR
XINT1 XINT2 XINT3
Pin Selection Register chooses which pin(s) the signal comes out on Configuration Register controls the enable/disable and polarity Counter Register holds the interrupt counter
5 - 13
Low Power Modes
Low Power Modes
Low Power Modes
Low Power Mode Normal Run IDLE STANDBY HALT
CPU Logic Peripheral Watchdog Clock Clock Logic Clock on off off off on on off off on on on off
PLL / OSC on on on off
See device datasheet for power consumption in each mode
Low Power Mode Control Register 0

SysCtrlRegs.LPMCR0 (lab file: SysCtrl.c)
Watchdog Interrupt wake device from STANDBY 0 = disable (default) 1 = enable
15
Wake from STANDBY GPIO signal qualification *
000000 = 2 OSCCLKs 000001 = 3 OSCCLKs 111111 = 65 OSCCLKS (default)
. . .
. . .
. . .
14 - 8
7-2
1-0
WDINTE
reserved
QUALSTDBY
LPM0
Low Power Mode Entering 1. Set LPM bits 2. Enable desired exit interrupt(s) 3. Execute IDLE instruction 4. The power down sequence of the hardware depends on LP mode
Low Power Mode Selection 00 = Idle (default) 01 = Standby 1x = Halt
* QUALSTDBY will qualify the GPIO wakeup signal in series with the GPIO port qualification. This is useful when GPIO port qualification is not available or insufficient for wake-up purposes.
5 - 14
Low Power Modes
Low Power Mode Exit

Exit Interrupt
RESET
Low Power Mode
GPIO Port A Signal
Watchdog Interrupt
Any Enabled Interrupt
IDLE STANDBY HALT
yes yes yes
yes yes yes
yes yes no
yes no no
GPIO Low Power Wakeup Select

SysCtrlRegs.GPIOLPMSEL
31 30 29 28 27 26 25 24
GPIO31 GPIO30 GPIO29 GPIO28 GPIO27 GPIO26 GPIO25 GPIO24

23 22 21 20 19 18 17 16
GPIO23 GPIO22 GPIO21 GPIO20 GPIO19 GPIO18 GPIO17 GPIO16

15 14 13 12 11 10 9 8
GPIO15 GPIO14 GPIO13 GPIO12 GPIO11 GPIO10 GPIO9

7 6 5 4 3 2 1
GPIO8
0
GPIO7
GPIO6
GPIO5
GPIO4
GPIO3
GPIO2
GPIO1
GPIO0
Wake device from HALT and STANDBY mode (GPIO Port A) 0 = disable (default) 1 = enable
5 - 15
Register Protection
Register Protection
Write-Read Protection
DevEmuRegs.DEVICECNF.bit.ENPROT
Suppose you need to write to a peripheral register and then read a different register for the same peripheral (e.g., write to control, read from status register)? CPU pipeline protects W-R order for the same address Write-Read protection mechanism protects W-R order for different addresses
Peripheral Frame 1 and Peripheral Frame 2 zones protected Write-read protection mode bit ENPROT located in the DEVICECNF register is enabled by default
Peripheral Frame Registers
PF0 eCAN COMP ePWM eCAP eQEP LIN GPIO PF1 System Control SPI SCI Watchdog XINT ADC I2C
Protected address: 0x4000 - 0x7FFF
EALLOW Protection (1 of 2)
EALLOW stands for Emulation Allow Code access to protected registers allowed only when EALLOW = 1 in the ST1 register The emulator can always access protected registers EALLOW bit controlled by assembly level instructions
EALLOW sets the bit (register access enabled) EDIS clears the bit (register access disabled)
EALLOW bit cleared upon ISR entry, restored upon exit
5 - 16
Register Protection
EALLOW Protection (2 of 2)
The following registers are protected:
Device Emulation Flash Code Security Module PIE Vector Table LIN (some registers) eCANA/B (control registers only; mailbox RAM not protected) ePWM1-7 and COMP1-3 (some registers) GPIO (control registers only) System Control
See device datasheet and peripheral users guides for detailed listings
EALLOW register access C-code example:

asm(" EALLOW"); SysCtrlRegs.WDKEY=0x55; asm(" EDIS"); // enable protected register access // write to the register // disable protected register access
5 - 17
Lab 5: System Initialization

Objective
The objective of this lab is to perform the processor system initialization. Additionally, the peripheral interrupt expansion (PIE) vectors will be initialized and tested using the information discussed in the previous module. This initialization process will be used again in all of the lab exercises throughout this workshop. The system initialization for this lab will consist of the following:
Setup the clock module PLL, LOSPCP = /4, low-power modes to default values, enable all module clocks Disable the watchdog clear WD flag, disable watchdog, WD prescale = 1 Setup watchdog system and control register DO NOT clear WD OVERRIDE bit, WD generate a CPU reset Setup shared I/O pins set all GPIO pins to GPIO function (e.g. a "00" setting for GPIO function, and a 01, 10, or 11 setting for a peripheral function.)
The first part of the lab exercise will setup the system initialization and test the watchdog operation by having the watchdog cause a reset. In the second part of the lab exercise the PIE vectors will be added and tested by using the watchdog to generate an interrupt. This lab will make use of the DSP2803x C-code header files to simplify the programming of the device, as well as take care of the register definitions and addresses. Please review these files, and make use of them in the future, as needed.
Procedure
Create Project File

1. Create a new project called Lab5.pjt in C:\C28x\Labs\Lab5 and add the following files to it: CodeStartBranch.asm DelayUs.asm DSP2803x_GlobalVariableDefs.c DSP2803x_Headers_nonBIOS.cmd Gpio.c Lab_5_6_7.cmd Main_5.c SysCtrl.c Watchdog.c
Note that include files, such as DSP2803x_Device.h and Lab.h, are automatically added at project build time. (Also, DSP2803x_DefaultIsr.h is automatically added and will be used with the interrupts in the second part of this lab exercise).
5 - 18

2. We need to setup the search path to include the peripheral register header files. Click: Project Build Options
Select the Compiler tab. In the Preprocessor Category, find the Include Search Path (-i) box and enter: ..\DSP2803x_headers\include This is the path for the header files. 3. Select the Linker tab and set the Stack Size to 0x200. 4. Setup the compiler run-time support library. In the Libraries Category, find the Include Libraries (-l) box and enter: rts2800_ml.lib. Select OK and the Build Options window will close.
Modify Memory Configuration

5. Open and inspect the linker command file Lab_5_6_7.cmd. Notice that the user defined section codestart is being linked to a memory block named BEGIN_M0. The codestart section contains code that branches to the code entry point of the project. The bootloader must branch to the codestart section at the end of the boot process. Recall that the "Jump to M0 SARAM" bootloader mode branches to address 0x000000 upon bootloader completion. Modify the linker command file Lab_5_6_7.cmd to create a new memory block named BEGIN_M0: origin = 0x000000, length = 0x0002, in program memory. You will also need to modify the existing memory block M0SARAM in data memory to avoid any overlaps with this new memory block.
Setup System Initialization

6. Modify SysCtrl.c and Watchdog.c to implement the system initialization as described in the objective for this lab. 7. Open and inspect Gpio.c. Notice that the shared I/O pins have been set to the GPIO function. Save your work and close the modified files.
Build and Load

8. Click the Build button and watch the tools run in the build window. The output file should automatically load. 9. Under Debug on the menu bar click Reset CPU.
5 - 19
10. Under GEL on the menu bar click: EMU_BOOT_SARAM. EMU Boot Mode Select This has the debugger load values into EMU_KEY and EMU_BMODE so that the bootloader will jump to "M0 SARAM" at 0x000000. 11. Under Debug on the menu bar click Go Main. You should now be at the start of Main().
Run the Code Watchdog Reset

12. Place the cursor in the main loop section (on the asm( NOP); instruction line) and right click the mouse key and select Run To Cursor. This is the same as setting a breakpoint on the selected line, running to that breakpoint, and then removing the breakpoint. 13. Place the cursor on the first line of code in main() and set a breakpoint by right clicking the mouse key and select Toggle Software Breakpoint. Notice that line is highlighted with a red dot indicating that the breakpoint has been set. Alternately, you can double-click in the gray field to the left of the code line to set the breakpoint. The breakpoint is set to prove that the watchdog is disabled. If the watchdog causes a reset, code execution will stop at this breakpoint. 14. Run your code for a few seconds by using the <F5> key, or using the Run button on the Run on the menu bar. After a few seconds halt vertical toolbar, or using Debug your code by using Shift <F5>, or the Halt button on the vertical toolbar. Where did your code stop? Are the results as expected? If things went as expected, your code should be in the main loop. 15. Modify the InitWatchdog() function to enable the watchdog (WDCR). This will enable the watchdog to function and cause a reset. Save the file and click the Build button. 16. Reset the CPU by performing the following steps: Click on Debug Reset CPU Next click Debug Go Main 17. Like before, place the cursor in the main loop section (on the asm( NOP); instruction line) and right click the mouse key and select Run To Cursor.. 18. Run your code. Where did your code stop? Are the results as expected? If things went as expected, your code should have stopped at the breakpoint. What happened is as follows. While the code was running, the watchdog timed out and reset the processor. The reset vector was then fetched and the ROM bootloader began execution. Since the device is in emulation boot mode (i.e. the emulator is connected) the bootloader read the EMU_KEY and EMU_BMODE values from the PIE RAM. These values were previously set for boot to M0 SARAM bootmode when we invoked the EMU_BOOT_SARAM GEL function earlier in this lab. Since these values did not change and are not affected by reset, the bootloader transferred execution to the beginning of our code at address 0x000000 in the M0SARAM, and execution continued until the breakpoint was hit in main().
5 - 20
Setup PIE Vector for Watchdog Interrupt

The first part of this lab exercise used the watchdog to generate a CPU reset. This was tested using a breakpoint set at the beginning of main(). Next, we are going to use the watchdog to generate an interrupt. This part will demonstrate the interrupt concepts learned in the previous module. 19. Add the following files to the project: DefaultIsr_5.c PieCtrl_5_6_7_8_9_10.c PieVect_5_6_7_8_9_10.c Check your files list to make sure the files are there. 20. In Main_5.c, add code to call the InitPieCtrl() function. There are no passed parameters or return values, so the call code is simply: InitPieCtrl(); 21. Using the PIE Interrupt Assignment Table shown in the previous module find the location for the watchdog interrupt, WAKEINT. This will be used in the next step. PIE group #: # within group:
22. Modify main() to do the following: - Enable global interrupts (INTM bit) Then modify InitWatchdog() to do the following: - Enable the "WAKEINT" interrupt in the PIE (Hint: use the PieCtrlRegs structure) - Enable the appropriate core interrupt in the IER register 23. In Watchdog.c modify the system control and status register (SCSR) to cause the watchdog to generate a WAKEINT rather than a reset. Save all changes to the files. 24. Open and inspect DefaultIsr_5.c. This file contains interrupt service routines. The ISR for WAKEINT has been trapped by an emulation breakpoint contained in an inline assembly statement using ESTOP0. This gives the same results as placing a breakpoint in the ISR. We will run the lab exercise as before, except this time the watchdog will generate an interrupt. If the registers have been configured properly, the code will be trapped in the ISR. 25. Open and inspect PieCtrl_5_6_7_8_9_10.c. This file is used to initialize the PIE RAM and enable the PIE. The interrupt vector table located in PieVect_5_6_7_8_9_10.c is copied to the PIE RAM to setup the vectors for the interrupts. Close the modified and inspected files.
Build and Load

26. Click the Build button. Next reset the CPU, and then Go Main.
5 - 21
Run the Code Watchdog Interrupt

27. Place the cursor in the main loop section, right click the mouse key and select Run To Cursor. 28. Run your code. Where did your code stop? Are the results as expected? If things went as expected, your code should stop at the ESTOP0 instruction in the WAKEINT ISR. End of Exercise Note: By default, the watchdog timer is enabled out of reset. Code in the file CodeStartBranch.asm has been configured to disable the watchdog. This can be important for large C code projects (ask your instructor if this has not already been explained). During this lab exercise, the watchdog was actually re-enabled (or disabled again) in the file Watchdog.c.
5 - 22
Analog-to-Digital Converter and Comparator

Introduction
This module explains the operation of the analog-to-digital converter and comparator. The ADC system consists of a 12-bit analog-to-digital converter with up to 16 analog input channels. The analog input channels have a full range analog input of 0 to 3.3 volts or VREFHI/VREFLO ratiometric. Two input analog multiplexers are available, each supporting up to 8 analog input channels. Each multiplexer has its own dedicated sample and hold circuit. Therefore, sequential, as well as simultaneous sampling is supported. The ADC system is start-of-conversion (SOC) based where each independent SOCx (where x = 0 to 15) register configures the trigger source that starts the conversion, the channel to convert, and the acquisition (sample) window size. Up to 16 results registers are used to store the conversion values. Conversion triggers can be performed by an external trigger pin, software, an ePWM or CPU timer interrupt event, or a generated ADCINT1/2 interrupt.
Learning Objectives
Learning Objectives
Understand the operation of the Analog-to-Digital converter (ADC) and Comparator Use the ADC to perform data acquisition
C2000 Piccolo Workshop - Analog-to-Digital Converter
6-1
Module Topics
Module Topics
Analog-to-Digital Converter and Comparator ....................................................................................... 6-1 Module Topics......................................................................................................................................... 6-2 Analog-to-Digital Converter................................................................................................................... 6-3 ADC Block and Functional Diagrams ................................................................................................ 6-3 ADC Triggering.................................................................................................................................. 6-4 ADC Conversion Priority ................................................................................................................... 6-5 ADC Clock and Timing...................................................................................................................... 6-7 ADC Converter Registers ................................................................................................................... 6-8 ADC Calibration and Reference........................................................................................................6-13 Comparator............................................................................................................................................6-15 Comparator Block Diagram...............................................................................................................6-15 Comparator Registers ........................................................................................................................6-16 Lab 6: Analog-to-Digital Converter ......................................................................................................6-17
6-2
Analog-to-Digital Converter
ADC Block and Functional Diagrams
ADC Module Block Diagram

ADCINA0 ADCINA1 ADCINA7 ADCINB0 ADCINB1 ADCINB7 ADC full-scale input range is 0 to 3.3V CHSEL ADC EOCx Generation Logic SOCx Signal SOC0 SOC1 SOC2 SOC3 TRIGSEL TRIGSEL TRIGSEL TRIGSEL CHSEL CHSEL CHSEL CHSEL ACQPS ACQPS ACQPS ACQPS ADC ADCINT1-9 Interrupt Logic ADCINT1 ADCINT2 SOCx Triggers MUX A S/H A RESULT0 RESULT1 RESULT2
12-bit A/D Converter

SOC
Result MUX
MUX B
S/H B
MUX
RESULT15
SOC15 TRIGSEL CHSEL ACQPS

SOCx Configuration Registers
Software CPU Timer (0,1,2) EPWMxSOCA (x = 1 to 7) EPWMxSOCB (x = 1 to 7) External Pin

(GPIO/XINT2_ADCSOC)
ADC SOCx Functional Diagram

ADCSOCFRC1
Software Trigger TINT0 (CPU Timer 0) TINT1 (CPU Timer 1) TINT2 (CPU Timer 2) XINT2_ADCSOC (GPIO) SOCA (ePWM1) SOCB (ePWM1) SOCA (ePWM7) SOCB (ePWM7) none ADCINT1 ADCINT2 T r i g g e r
ADCSOCxCTL
ADCINT1
ADCRESULTx
Channel S O C x Select Sample Window Result Register E O C x
ADCINT2 ADCINT3 ADCINT4 ADCINT5 ADCINT6 ADCINT7 ADCINT8 ADCINT9
ReTrigger
INTSELxNy ADCINTSOCSEL1 ADCINTSOCSEL2
This block diagram is replicated 16 times
6-3
ADC Triggering
Example ADC Triggering (1 of 2)

Sample A2 B3 A7 when ePWM1 SOCB is generated and then generate ADCINT1n:
SOC0 SOC1 SOC2 SOCB (ETPWM1)
Channel A2 Channel B3 Channel A7 B0
Sample 7 cycles Sample 10 cycles Sample 4 cycles
Result0 Result1 Result2
no interrupt no interrupt ADCINT1n
As above, but also sample A0

SOCB (ETPWM1) SOC0 SOC1 SOC2 Software Trigger SOC3 SOC4 SOC5
A5 continuously and generate ADCINT2n: Sample 7 cycles Sample 10 cycles Sample 4 cycles Sample 10 cycles Sample 15 cycles Sample 12 cycles Result0 Result1 Result2 Result3 Result4 Result5
Channel A2 Channel B3 Channel A7 Channel A0 Channel B0 Channel A5
ADCINT2n
Example ADC Triggering (2 of 2)

Sample all channels continuously and provide Ping-Pong interrupts to CPU/system:
Software Trigger ADCINT2n SOC2 SOC4 SOC6 SOC8 SOC10 SOC0
Channel A0:B0 Channel A1:B1 Channel A2:B2 Channel A3:B3 Channel A4:B4 Channel A5:B5 A6:B6 A7:B7
Sample 7 cycles Sample 7cycles Sample 7 cycles Sample 7 cycles Sample 7 cycles Sample 7 cycles Sample 7 cycles Sample 7 cycles
Result0 Result1 Result2 Result3 Result4 Result5 Result6 Result7 Result8 Result9 Result10 Result11 Result12 Result13 Result14 Result15
no interrupt no interrupt no interrupt
ADCINT1n no interrupt no interrupt no interrupt
SOC12 Channel SOC14 Channel
ADCINT2n
6-4
ADC Conversion Priority
ADC Conversion Priority

When multiple SOC flags are set at the same time priority determines the order in which they are converted
Round Robin Priority (default)
No SOC has an inherent higher priority than another Priority depends on the round robin pointer
High Priority
High priority SOC will interrupt the round robin wheel after current conversion completes and insert itself as the next conversion After its conversion completes, the round robin wheel will continue where it was interrupted
Conversion Priority Functional Diagram

High Priority SOC0 SOC1 SOC2 SOC3 SOC4 SOC5 SOC6 SOC7 SOC8 SOC9 SOC10 SOC11 SOC12 SOC13 SOC14 SOC15
SOC Priority
Determines cutoff point for high priority and round robin mode SOCPRIORITY
AdcRegs.SOCPRICTL
Round Robin
RRPOINTER
Round Robin Pointer

Points to the last converted round robin SOCx and determines order of conversions
6-5
Round Robin Priority Example

SOCPRIORITY configured as 0; RRPOINTER configured as 15; SOC0 is highest RR priority SOC7 trigger received SOC7 is converted; RRPOINTER now points to SOC7; SOC8 is now highest RR priority SOC2 & SOC12 triggers received simultaneously SOC12 is converted; RRPOINTER points to SOC12; SOC13 is now highest RR priority SOC2 is converted; RRPOINTER points to SOC2; SOC3 is now highest RR priority
SOC 14 SOC 13 SOC 12 SOC 11 SOC 10 SOC 9 SOC 8 SOC 6 SOC 7
SOC 15
SOC 0
SOC 1 SOC 2 SOC 3 SOC 4 SOC 5
RRPOINTER
High Priority Example

SOCPRIORITY configured as 4; RRPOINTER configured as 15; SOC4 is highest RR priority SOC7 trigger received SOC7 is converted; RRPOINTER points to SOC7; SOC8 is now highest RR priority SOC2 & SOC12 triggers received simultaneously SOC2 is converted; RRPOINTER stays pointing to SOC7 SOC12 is converted; RRPOINTER points to SOC12; SOC13 is now highest RR priority
High Priority
SOC 0 SOC 1 SOC 13 SOC 2 SOC 3 SOC 12 SOC 11 SOC 9 SOC 8 SOC 14 SOC 15
SOC 4
SOC 5 SOC 6
RRPOINTER
SOC 7
SOC 10
6-6
ADC Clock and Timing
ADC Clocking Flow

Internal OSC1 (10 MHz) PLLCR PLLSTS
DIV bits
1100b (x12)
DIVSEL bits
10b (/2)
To CPU
SYSCLKOUT (60 MHz)
PCLKCR0.ADCENCLK = 1 To ADC pipeline
ADCCLK (60 MHz) ADCSOCxCTL
ACQ_PS bits
0110b
sampling window
sampling window = (ACQ_PS + 1)*(1/ADCCLK)
ADC Timing Sequential Sampling

Latch
2 Clocks
Sample
7 Clocks
Convert
6 Clocks 7 Clocks
Write
2 Clocks
Generate Early Interrupt
Generate Late Interrupt Start Sampling Next Channel
Max Continuous Sampling:

60 MHz = 4.62 MSPS 13 cycles / 1 sample 40 MHz = 3.08 MSPS 13 cycles / 1 sample
6-7
ADC Timing Simultaneous Sampling

Convert B Channel
6 Clocks 7 Clocks
Write
2 Clocks
Latch
2 Clocks
Sample
7 Clocks
Convert A Channel
13 Clocks
Write
2 Clocks
Generate Early Interrupt A Channel
Generate Late Interrupt A Channel & Generate Early Interrupt B Channel
Generate Late Interrupt B Channel Start Sampling Next Channel
Max Continuous Sampling:

60 MHz = 4.62 MSPS 26 cycles / 2 sample 40 MHz = 3.08 MSPS 26 cycles / 2 sample
ADC Converter Registers
Analog-to-Digital Converter Registers

AdcRegs.register (lab file: Adc.c)
Description Control 1 Register SOC0 to SOC15 Control Registers Interrupt SOC Selection 1 and 2 Registers Sampling Mode Register SOC Flag 1 Register SOC Force 1 Register SOC Overflow 1 Register SOC Overflow Clear 1 Register Interrupt x and y Selection Registers Interrupt Flag Register Interrupt Flag Clear Register Interrupt Overflow Register Interrupt Overflow Clear Register SOC Priority Control Register Reference Trim Register Offset Trim Register Revision Register reserved ADC Result 0 to 15 Registers Register ADCCTL1 ADCSOCxCTL ADCINTSOCSELx ADCSAMPLEMODE ADCSOCFLG1 ADCSOCFRC1 ADCSOCOVF1 ADCSOCOVFCLR1 INTSELxNy ADCINTFLG ADCINTFLGCLR ADCINTOVF ADCINTOVFCLR SOCPRICTL ADCREFTRIM ADCOFFTRIM ADCREV ADCRESULTx
Note: ADCRESULTx is located in AdcResult.register and not in AdcRegs
6-8
ADC Control Register 1

AdcRegs.ADCCTL1
Upper Register:
ADC Module Reset
0 = no effect 1 = reset (set back to 0 by ADC logic)
ADC Busy
0 = ADC busy 1 = ADC available
ADC Busy Channel

When ADCBSY = 0: last channel converted 1: channel currently processing
15 RESET
14 ADCENABLE
13 ADCBSY
12 - 8 ADCBSYCHN
ADC Enable
0 = ADC disable 1 = ADC enable
00h = ADCINA0 01h = ADCINA1 02h = ADCINA2 03h = ADCINA3 04h = ADCINA4 05h = ADCINA5 06h = ADCINA6 07h = ADCINA7
08h = ADCINB0 09h = ADCINB1 0Ah = ADCINB2 0Bh = ADCINB3 0Ch = ADCINB4 0Dh = ADCINB5 0Eh = ADCINB6 0Fh = ADCINB7
ADC Control Register 1

AdcRegs.ADCCTL1
Lower Register:
ADC Power Down
0 = analog circuitry powered down 1 = analog circuitry powered up 7 6
ADC Reference Power Down

0 = reference circuitry powered down 1 = reference circuitry powered up 5 4
ADC Reference Select

0 = internal 1 = external
(VREFHI/VREFLO)
Temperature Sensor Convert

currently not used 0 = only valid setting
3 2 1 0 ADCREF INTPULSE VREFLO TEMP ADCPWN ADCBGPWN ADCREFPWD reserved SEL POS CONV CONV
ADC Bandgap Power Down

0 = bandgap circuitry powered down 1 = bandgap circuitry powered up
INT Pulse Generation Control

0 = beginning of conversion 1 = one cycle prior to result
VREFLO Convert
0 = not connected 1 = connected (B5)
6-9
ADC SOC0 SOC15 Control Registers

AdcRegs.ADCSOCxCTL
SOCx Channel Select
10 reserved 9-6 CHSEL
SOCx Trigger Source Select

15 - 11 TRIGSEL
00h = software 01h = CPU Timer 0 02h = CPU Timer 1 03h = CPU Timer 2 04h = XINT2SOC 05h = ePWM1SOCA 06h = ePWM1SOCB 07h = ePWM2SOCA 08h = ePWM2SOCB 09h = ePWM3SOCA 0Ah = ePWM3SOCB 0Bh = ePWM4SOCA 0Ch = ePWM4SOCB 0Dh = ePWM5SOCA 0Eh = ePWM5SOCB 0Fh = ePWM6SOCA 10h = ePWM6SOCB 11h = ePWM7SOCA 12h = ePWM7SOCB
SOCx Acquisition Prescale (S/H window)

5-0 ACQPS
Sequential S/M (SIMULENx=0) 0h = ADCINA0 1h = ADCINA1 2h = ADCINA2 3h = ADCINA3 4h = ADCINA4 5h = ADCINA5 6h = ADCINA6 7h = ADCINA7 8h = ADCINB0 9h = ADCINB1 Ah = ADCINB2 Bh = ADCINB3 Ch = ADCINB4 Dh = ADCINB5 Eh = ADCINB6 Fh = ADCINB7
Simultaneous S/M (SIMULENx=1) 0h = ADCINA0/B0 1h = ADCINA1/B1 2h = ADCINA2/B2 3h = ADCINA3/B3 4h = ADCINA4/B4 5h = ADCINA5/B5 6h = ADCINA6/B6 7h = ADCINA7/B7 8h Fh = invalid
Sampling Window 00h 05h = invalid 06h = 7 cycles long 07h = 8 cycles long 08h = 9 cycles long 09h = 10 cycles long
3Fh = 64 cycles long
ADC Interrupt Trigger SOC Select Registers 1 & 2

AdcRegs.ADCINTSOCSELx
ADCINTSOCSEL2
15 - 14 SOC15 13 - 12 SOC14 11 - 10 SOC13 9-8 SOC12 7-6 SOC11 5-4 SOC10 3- 2 SOC9 1- 0 SOC8
ADCINTSOCSEL1
15 - 14 SOC7 13 - 12 SOC6 11 - 10 SOC5 9-8 SOC4 7-6 SOC3 5-4 SOC2 3- 2 SOC1 1- 0 SOC0
SOCx ADC Interrupt Select

Selects which, if any, ADCINT triggers SOCx 00 = no ADCINT will trigger SOCx (TRIGSEL field determines SOCx trigger) 01 = ADCINT1 will trigger SOCx (TRIGSEL field ignored) 10 = ADCINT2 will trigger SOCx (TRIGSEL field ignored) 11 = invalid selection
6 - 10
ADC Sample Mode Register

AdcRegs.ADCSAMPLEMODE
15 - 8
reserved
SIMULEN14 SIMULEN12 SIMULEN10 SIMULEN8 SIMULEN6 SIMULEN4 SIMULEN2 SIMULEN0
Simultaneous Sampling Enable

Couples SOCx and SOCx+1 in simultaneous sampling mode 0 = single sample mode for SOCx and SOCx+1 1 = simultaneous sample mode for SOCx and SOCx+1
SOC Priority Control Register

AdcRegs.SOCPRICTL
10 - 5 RRPOINTER 15 - 11 reserved 4-0 SOCPRIORITY
Round Robin Pointer

Points to the last converted round robin SOCx and determines order of conversions
00h = SOC0 last converted, SOC1 highest priority 01h = SOC1 last converted, SOC2 highest priority 02h = SOC2 last converted, SOC3 highest priority 03h = SOC3 last converted, SOC4 highest priority 04h = SOC4 last converted, SOC5 highest priority 05h = SOC5 last converted, SOC6 highest priority 06h = SOC6 last converted, SOC7 highest priority 07h = SOC7 last converted, SOC8 highest priority 08h = SOC8 last converted, SOC9 highest priority 09h = SOC9 last converted, SOC11 highest priority 0Ah = SOC10 last converted, SOC11 highest priority 0Bh = SOC11 last converted, SOC12 highest priority 0Ch = SOC12 last converted, SOC13 highest priority 0Dh = SOC13 last converted, SOC14 highest priority 0Eh = SOC14 last converted, SOC15 highest priority 0Fh = SOC15 last converted, SOC0 highest priority 1xh = invalid selection 20h = reset value (no SOC has been converted)
SOC Priority
Determines cutoff point for high priority and round robin mode
00h = round robin mode for all channels 01h = SOC0 high priority, SOC1-15 round robin 02h = SOC0-1 high priority, SOC2-15 round robin 03h = SOC0-2 high priority, SOC3-15 round robin 04h = SOC0-3 high priority, SOC4-15 round robin 05h = SOC0-4 high priority, SOC5-15 round robin 06h = SOC0-5 high priority, SOC6-15 round robin 07h = SOC0-6 high priority, SOC7-15 round robin 08h = SOC0-7 high priority, SOC8-15 round robin 09h = SOC0-8 high priority, SOC9-15 round robin 0Ah = SOC0-9 high priority, SOC10-15 round robin 0Bh = SOC0-10 high priority, SOC11-15 round robin 0Ch = SOC0-11 high priority, SOC12-15 round robin 0Dh = SOC0-12 high priority, SOC13-15 round robin 0Eh = SOC0-13 high priority, SOC14-15 round robin 0Fh = SOC0-14 high priority, SOC15 round robin 10h = all SOCs high priority (arbitrated by SOC #) 1xh = invalid selection
6 - 11
Interrupt Select x and y Register

AdcRegs.INTSELxNy
Where x/y = 1/2, 3/4, 5/6, 7/8, 9/10 and 10 is reserved 15 14 13 reserved 7 reserved INTyCONT 6 INTxCONT INTyE 5 INTxE 12 - 8 INTySEL 4-0 INTxSEL ADCINTx/y EOC Source Select
ADCINTx/y Continuous Mode Enable

0 = one-shot pulse generated (until flag cleared by user) 1 = pulse generated for each EOC
ADCINTx/y Interrupt Enable

0 = disable 1 = enable
00h = EOC0 is trigger for ADCINTx/y 01h = EOC1 is trigger for ADCINTx/y 02h = EOC2 is trigger for ADCINTx/y 03h = EOC3 is trigger for ADCINTx/y 04h = EOC4 is trigger for ADCINTx/y 05h = EOC5 is trigger for ADCINTx/y 06h = EOC6 is trigger for ADCINTx/y 07h = EOC7 is trigger for ADCINTx/y 08h = EOC8 is trigger for ADCINTx/y 09h = EOC9 is trigger for ADCINTx/y 0Ah = EOC10 is trigger for ADCINTx/y 0Bh = EOC11 is trigger for ADCINTx/y 0Ch = EOC12 is trigger for ADCINTx/y 0Dh = EOC13 is trigger for ADCINTx/y 0Eh = EOC14 is trigger for ADCINTx/y 0Fh = EOC15 is trigger for ADCINTx/y 1xh = invalid value
ADC Conversion Result Registers

AdcResult.ADCRESULTx, x = 0 - 15
MSB 15 14 13 12 11 10 9 8 7 6 5 4 3 2 1 LSB 0
Input Voltage 3.3 1.65 0.00081 0
Digital Result FFFh 7FFh 1h 0h
AdcResult. ADCRESULTx 0000|1111|1111|1111 0000|0111|1111|1111 0000|0000|0000|0001 0000|0000|0000|0000
Sequential Sampling Mode (SIMULENx = 0)

After ADC completes a conversion of an SOCx, the digital result is placed in the corresponding ADCRESULTx register
Simultaneous Sampling Mode (SIMULENx = 1)

After ADC completes a conversion of a channel pair, the digital results are found in the corresponding ADCRESULTx and ADCRUSULTx+1 registers
6 - 12
How Can We Handle Signed Input Voltages?

Example: -1.65 V V in +1.65 V
R R R +
1) Add 1.65 volts to the analog input
Vin 1.65V
R + R
C28x
ADCINx
ADCLO GND
2) Subtract 1.65 from the digital result

#include DSP2803x_Device.h #define offset 0x07FF void main(void) { int16 value; // signed value = AdcResult.ADCRESULT0 offset; }
ADC Calibration and Reference
Built-In ADC Calibration

TI reserved OTP contains device specific calibration data for the ADC and internal oscillators The Boot ROM contains a Device_cal() routine that copies the calibration data to their respective registers Device_cal() must be run to meet the ADC and oscillator specs in the datasheet
The Bootloader automatically calls Device_cal() such that no action is normally required by the user If the bootloader is bypassed (e.g., during development) Device_cal() should be called by the application:
#define Device_cal (void (*)(void))0x3D7C80 void main(void) { (*Device_cal)(); } // call Device_cal()
A GEL function using CCS is also available as part of the Peripheral Register Header Files to accomplish this
6 - 13
Manual ADC Calibration

If the offset and gain errors in the datasheet* are unacceptable for your application, or you want to also compensate for board level errors (e.g., sensor or amplifier offset), you can manually calibrate Offset error
Compensated in analog with the ADCOFFTRIM register No reduction in full-scale range Configure input B5 to VREFLO, set ADCOFFTRIM to maximum offset error, and take a reading Re-adjust ADCOFFTRIM to make result zero ADCOFFTRIM
CH
12-bit ADC
B5
CH
VREFLOCONV
Gain error
VREFLO Compensated in software Some loss in full-scale range Requires use of a second ADC input pin and an upper-range reference voltage on that pin; see TMS320280x and TMS320F2801x ADC Calibration appnote #SPRAAD8 for more information
Tip: To minimize mux-to-mux variation effects, put your most critical signals on a single mux and use that mux for calibration inputs
* +/-15 LSB offset, +/-30 LSB gain. See device datasheet for exact specifications
ADC Reference Selection

AdcRegs.ADCREFSEL The internal reference has temperature stability of ~50 PPM/C* The internal reference (default) will convert an applied input voltage to a fixed scale of 0 to 3.3 V range If this is not sufficient for your application, there is the option to use an external reference*
External reference will scale an input voltage range from VREFLO to VREFHI (ratiometric) The reference value changes the 0 - 3.3 V full-scale range of the ADC
The ADCREFSEL in ADCCTL1 controls the reference choice

15 - 5 4 3 2-0
reserved ADCREFSEL
ADC Reference Selection

0 = internal (default) 1 = external VREFHI/VREFLO pins used for reference generation * See device datasheet for exact specifications and ADC reference hardware connections
6 - 14
MUX
Comparator
Comparator
Comparator Block Diagram
Comparator
A0 B0 A1 B1 A2 B2 A3 B3 A4 B4 A5 B5 A6 B6 A7 B7 AIO6 AIO14 10-bit DAC
COMP3 COMP3OUT
AIO2 AIO10
10-bit DAC
COMP1
COMP1OUT
AIO4 AIO12
10-bit DAC
COMP2
COMP2OUT
ADC
Comparator Block Diagram

COMPDACE SYNCSEL
Input Pin A Input Pin B

VDDA VSSA
+ COMPx 1
V 0
0 1
SYSCLKOUT
10-bit DAC
Sync/ Qual
COMPSOURCE CMPINV QUALSEL
ePWM Event COMPxTRIP Trigger & GPIO 1 MUX 0

COMPSTS
DACVAL
DAC Reference
V= DACVAL * (VDDA VSSA ) 1023
Comparator Truth Table

Voltages Voltage A < Voltage B Voltage A > Voltage B Output 0 1
6 - 15
Comparator
Comparator Registers
Comparator Registers
AdcRegs.COMPCTL Compare Control Register
8 15 - 9 reserved SYNCSEL 7- 3 QUALSEL 2 CMPINV 1 0 COMPSOURCE COMPDACE
Synchronization Select Qualification Invert Period Output before being feed 0 = passed
to ETPWM/GPIO blocks 0 = Asynchronous 1 = Synchronous 0h = passed 1h = 2 clocks 2h = 3 clocks Fh = 15 clocks 1 = inverted
Comparator Comparator/ Source DAC Enable

0 = DAC 1 = pin 0 = disable 1 = enable
AdcRegs.COMPSTS Compare Output Status Register

15 - 1 reserved 0 COMPSTS Logical latched value of the comparator
AdcRegs.DACVAL DAC Value Register

15 - 10 reserved 9-0 DACVAL
DAC Value
Scales output of DAC from 0 1023 Value = 0 3FFh
6 - 16
Lab 6: Analog-to-Digital Converter

Objective
The objective of this lab is to become familiar with the programming and operation of the on-chip analog-to-digital converter. The MCU will be setup to sample a single ADC input channel at a prescribed sampling rate and store the conversion result in a memory buffer. This buffer will operate in a circular fashion, such that new conversion data continuously overwrites older results in the buffer.
Lab 6: ADC Sampling

GND +3.3 V Toggle (GPIO20) (GPIO18)
ADC
connector wire
ADCINA0
ePWM2 triggering ADC on period match using SOCA trigger every 20 s (50 kHz)
View ADC buffer PWM Samples
ePWM2
Recall that there are three basic ways to initiate an ADC start of conversion (SOC):
1. Using software a. SOCx bit (where x = 0 to 15) in the ADC SOC Force 1 Register (ADCSOCFRC1) causes a software initiated conversion Automatically triggered on user selectable conditions a. CPU Timer 0/1/2 interrupt b. ePWMxSOCA / ePWMxSOCB (where x = 1 to 7) - ePWM underflow (CTR = 0) - ePWM period match (CTR = PRD) - ePWM underflow or period match (CTR = 0 or PRD) - ePWM compare match (CTRU/D = CMPA/B) c. ADC interrupt ADCINT1 or ADCINT2 - triggers SOCx (where x = 0 to 15) selected by the ADC Interrupt Trigger SOC Select1/2 Register (ADCINTSOCSEL1/2) Externally triggered using a pin a. ADCSOC pin (GPIO/XINT2_ADCSOC)
2.
3.
One or more of these methods may be applicable to a particular application. In this lab, we will be using the ADC for data acquisition. Therefore, one of the ePWMs (ePWM2) will be
pointer rewind
RESULT0
CPU copies result to buffer during ADC ISR
data memory
...
6 - 17
configured to automatically trigger the SOC A signal at the desired sampling rate (ePWM period match CTR = PRD SOC method 2b above). The ADC end-of-conversion interrupt will be used to prompt the CPU to copy the results of the ADC conversion into a results buffer in memory. This buffer pointer will be managed in a circular fashion, such that new conversion results will continuously overwrite older conversion results in the buffer. In order to generate an interesting input signal, the code also alternately toggles a GPIO pin (GPIO18) high and low in the ADC interrupt service routine. The ADC ISR will also toggle LED LD3 on the ControlCARD as a visual indication that the ISR is running. This pin will be connected to the ADC input pin, and sampled. After taking some data, Code Composer Studio will be used to plot the results. A flow chart of the code is shown in the following slide.
Lab 6: Code Flow Diagram

Start General Initialization PLL and clocks watchdog configure GPIO setup PIE initialization ADC Initialization convert channel A0 on ePWM2 period match send interrupt on every conversion setup a results buffer in memory ePWM2 Initialization clear counter set period register set to trigger ADC on period match set the clock prescaler enable the timer ADC interrupt
Main Loop while(1) { }
ADC ISR read the ADC result write to result buffer adjust the buffer pointer toggle the GPIO pin return from interrupt
return
Notes
Program performs conversion on ADC channel A0 (ADCINA0 pin) ADC conversion is set at a 50 kHz sampling rate ePWM2 is triggering the ADC on period match using SOCA trigger Data is continuously stored in a circular buffer GPIO18 pin is also toggled in the ADC ISR ADC ISR will also toggle the ControlCARD LED LD3 as a visual indication that it is running
6 - 18
Procedure
Project File
1. A project named Lab6.pjt has been created for this lab. Open the project by clicking on Project Open and look in C:\C28x\Labs\Lab6. All Build Options have been configured the same as the previous lab. The files used in this lab are: Adc.c Gpio.c CodeStartBranch.asm Lab_5_6_7.cmd DefaultIsr_6.c Main_6.c DelayUs.asm PieCtrl_5_6_7_8_9_10.c DSP2803x_GlobalVariableDefs.c PieVect_5_6_7_8_9_10.c DSP2803x_Headers_nonBIOS.cmd SysCtrl.c EPwm_6.c Watchdog.c
Setup ADC Initialization and Enable Core/PIE Interrupts

2. In Main_6.c add code to call InitAdc() and InitEPwm() functions. The InitEPwm() function is used to configure ePWM2 to trigger the ADC at a 50 kHz rate. Details about the ePWM and control peripherals will be discussed in the next module. 3. Edit Adc.c to implement the ADC initialization as described above in the objective for the lab. Configure SOC0 for single sample mode, with an acquisition sample window of 7 cycles. Dont use the ADCINT to trigger a SOC0, and have all SOCs handled in round-robin mode. Enable ADCINT1 interrupt with EOC0 as the trigger for ADCINT1. Continuously generate an ADCINT1 pulse for each EOC. 4. Using the PIE Interrupt Assignment Table find the location for the ADC interrupt ADCINT1 (high-priority) and fill in the following information:
PIE group #:
# within group:
This information will be used in the next step. 5. Modify the end of Adc.c to do the following: - Enable the "ADCINT" interrupt in the PIE (Hint: use the PieCtrlRegs structure) - Enable the appropriate core interrupt in the IER register 6. Open and inspect DefaultIsr_6.c. This file contains the ADC interrupt service routine.
Build and Load

7. Save all changes to the files and click the Build button. 8. Reset the CPU, select EMU_BOOT_SARAM, and then Go Main.
6 - 19
Run the Code

9. In Main_6.c place the cursor in the main loop section, right click on the mouse key and select Run To Cursor. 10. Open a memory window to view some of the contents of the ADC results buffer. The address label for the ADC results buffer is AdcBuf. Note: Exercise care when connecting any wires, as the power to the USB Docking Station is on, and we do not want to damage the ControlCARD! 11. Using a connector wire provided, connect the ADCINA0 (pin # ADC-A0) to GND (pin # GND) on the Docking Station. Then run the code again, and halt it after a few seconds. Verify that the ADC results buffer contains the expected value of 0x0000. 12. Adjust the connector wire to connect the ADCINA0 (pin # ADC-A0) to +3.3V (pin # GPIO-20) on the Docking Station. (Note: pin # GPIO-20 has been set to 1 in Gpio.c). Then run the code again, and halt it after a few seconds. Verify that the ADC results buffer contains the expected value of 0x0FFF. 13. Adjust the connector wire to connect the ADCINA0 (pin # ADC-A0) to GPIO18 (pin # GPIO-18) on the Docking Station. Then run the code again, and halt it after a few seconds. Examine the contents of the ADC results buffer (the contents should be alternating 0x0000 and 0x0FFF values). Are the contents what you expected? 14. Open and setup a graph to plot a 50-point window of the ADC results buffer. Click: View Graph Time/Frequency and set the following values: Start Address Acquisition Buffer Size Display Data Size DSP Data Type Sampling Rate (Hz) Time Display Unit Select OK to save the graph options. 15. Recall that the code toggled the GPIO18 pin alternately high and low. (Also, the ADC ISR is toggling the LED LD3 on the ControlCARD as a visual indication that the ISR is running). If you had an oscilloscope available to display GPIO18, you would expect to see a square-wave. Why does Code Composer Studio plot resemble a triangle wave? What is the signal processing term for what is happening here? AdcBuf 50 50 16-bit unsigned integer 50000 s
6 - 20
16. Recall that the program toggled the GPIO18 pin at a 50 kHz rate. Therefore, a complete cycle (toggle high, then toggle low) occurs at half this rate, or 25 kHz. We therefore expect the period of the waveform to be 40 s. Confirm this by measuring the period of the triangle wave using the graph (you may want to enlarge the graph window using the mouse). The measurement is best done with the mouse. The lower left-hand corner of the graph window will display the X and Y axis values. Subtract the X-axis values taken over a complete waveform period.
Using Real-time Emulation

Real-time emulation is a special emulation feature that offers two valuable capabilities: A. Windows within Code Composer Studio can be updated at up to a 10 Hz rate while the MCU is running. This not only allows graphs and watch windows to update, but also allows the user to change values in watch or memory windows, and have those changes affect the MCU behavior. This is very useful when tuning control law parameters on-the-fly, for example. B. It allows the user to halt the MCU and step through foreground tasks, while specified interrupts continue to get serviced in the background. This is useful when debugging portions of a realtime system (e.g., serial port receive code) while keeping critical parts of your system operating (e.g., commutation and current loops in motor control). We will only be utilizing capability #1 above during the workshop. Capability #2 is a particularly advanced feature, and will not be covered in the workshop. 17. Reset the CPU, and then enable real-time mode by selecting: Debug Real-time Mode
A message box may appear. Select YES to enable debug events. This will set bit 1 (DBGM bit) of status register 1 (ST1) to a 0. The DBGM is the debug enable mask bit. When the DBGM bit is set to 0, memory and register values can be passed to the host processor for updating the debugger windows. 18. The memory and graph windows displaying AdcBuf should still be open. The connector wire between ADCINA0 (pin # ADC-A0) and GPIO18 (pin # GPIO-18) should still be connected. In real-time mode, we would like to have our window continuously refresh. Click: View Real-time Refresh Options
and check Global Continuous Refresh. Use the default refresh rate of 100 ms and select OK. Alternately, we could have right clicked on each window individually and selected Continuous Refresh.
6 - 21
Note: Global Continuous Refresh causes all open windows to refresh at the refresh rate. This can be problematic when a large number of windows are open, as bandwidth over the emulation link is limited. Updating too many windows can cause the refresh frequency to bog down. In that case, either close some windows, or disable global refresh and selectively enable Continuous Refresh for individual windows of interest instead. 19. Run the code and watch the windows update in real-time mode. Carefully remove and replace the connector wire from GPIO18. Are the values updating as expected? 20. Fully halting the CPU when in real-time mode is a two-step process. First, halt the processor with Debug Halt. Then uncheck the Real-time mode to take the CPU out of real-time mode (Debug Real-time Mode). 21. So far, we have seen data flowing from the MCU to the debugger in realtime. In this step, we will flow data from the debugger to the MCU. Open and inspect DefaultIsr_6.c. Notice that the global variable DEBUG_TOGGLE is used to control the toggling of the GPIO18 pin. This is the pin being read with the ADC. Highlight DEBUG_TOGGLE with the mouse, right click and select Add to Watch Window. The global variable DEBUG_TOGGLE should now be in the watch window with a value of 1. Run the code in real-time mode and change the value to 0. Are the results shown in the memory and graph window as expected? Change the value back to 1. As you can see, we are modifying data memory contents while the processor is running in real-time (i.e., we are not halting the MCU nor interfering with its operation in any way)! When done, fully halt the CPU.
22. Code Composer Studio includes GEL (General Extension Language) functions which automate entering and exiting real-time mode. Four functions are available: Run_Realtime_with_Reset (reset CPU, enter real-time mode, run CPU) Run_Realtime_with_Restart (restart CPU, enter real-time mode, run CPU) Full_Halt (exit real-time mode, halt CPU) Full_Halt_with_Reset (exit real-time mode, halt CPU, reset CPU) Realtime Emulation Control GEL Function
These GEL functions can be executed by clicking: GEL In the remaining lab exercises we will be using the above GEL functions to run and halt the code in real-time mode. If you would like, try repeating the previous step using the following GEL functions: GEL GEL Realtime Emulation Control Realtime Emulation Control End of Exercise Run_Realtime_with_Reset Full_Halt
6 - 22
Control Peripherals
Introduction
This module explains how to generate PWM waveforms using the ePWM unit. Also, the eCAP unit, and eQEP unit will be discussed.
Learning Objectives
Learning Objectives
Pulse Width Modulation (PWM) review Generate a PWM waveform with the Pulse Width Modulator Module (ePWM) Use the Capture Module (eCAP) to measure the width of a waveform Explain the function of Quadrature Encoder Pulse Module (eQEP)
Note: Different numbers of ePWM, eCAP, and eQEP modules are available on F2803x and F2802x devices. See the device datasheet for more information.
C2000 Piccolo Workshop - Control Peripherals
7-1
Module Topics
Module Topics
Control Peripherals................................................................................................................................... 7-1 Module Topics......................................................................................................................................... 7-2 PWM Review........................................................................................................................................... 7-3 ePWM...................................................................................................................................................... 7-5 ePWM Time-Base Sub-Module ......................................................................................................... 7-6 ePWM Compare Sub-Module ............................................................................................................ 7-9 ePWM Action Qualifier Sub-Module................................................................................................7-11 Asymmetric and Symmetric Waveform Generation using the ePWM..............................................7-16 PWM Computation Example.............................................................................................................7-17 ePWM Dead-Band Sub-Module........................................................................................................7-18 ePWM PWM Chopper Sub-Module..................................................................................................7-21 ePWM Digital Compare Sub-Module ...............................................................................................7-24 ePWM Trip-Zone Sub-Module..........................................................................................................7-27 ePWM Event-Trigger Sub-Module ...................................................................................................7-30 Hi-Resolution PWM (HRPWM) .......................................................................................................7-33 eCAP ......................................................................................................................................................7-34 eQEP......................................................................................................................................................7-40 Lab 7: Control Peripherals....................................................................................................................7-42
7-2
PWM Review
PWM Review
What is Pulse Width Modulation?

PWM is a scheme to represent a signal as a sequence of pulses
fixed carrier frequency fixed pulse amplitude pulse width proportional to instantaneous signal amplitude PWM energy original signal energy
t T
Original Signal
PWM representation
Pulse width modulation (PWM) is a method for representing an analog signal with a digital approximation. The PWM signal consists of a sequence of variable width, constant amplitude pulses which contain the same total energy as the original analog signal. This property is valuable in digital motor control as sinusoidal current (energy) can be delivered to the motor using PWM signals applied to the power converter. Although energy is input to the motor in discrete packets, the mechanical inertia of the rotor acts as a smoothing filter. Dynamic motor motion is therefore similar to having applied the sinusoidal currents directly.
7-3
PWM Review
Why use PWM with Power Switching Devices?

Desired output currents or voltages are known Power switching devices are transistors
Difficult to control in proportional region Easy to control in saturated region
PWM is a digital signal easy for DSP to output

DC Supply DC Supply
?
Desired signal to system
PWM PWM approx. of desired signal
Unknown Gate Signal
Gate Signal Known with PWM
7-4
ePWM
ePWM
ePWM Module Signals and Connections

ePWMx -1
EPWMxSYNCI EPWMxTZINT EPWMxINT EPWMxA
GPIO MUX
eQEP1 SYSCTRL CPU
TZ1 TZ3
PIE GPIO MUX ADC
EQEP1ERR TZ4 CLOCKFAIL TZ5 EMUSTOP TZ6
ePWMx
EPWMxB EPWMxSOCA
COMP
COMPxOUT EPWMxSYNCO
EPWMxSOCB
ePWMx+1
ePWM Block Diagram

Shadowed Shadowed
Clock Prescaler
Compare Register
Compare Register
TBCLK
16-Bit Time-Base Counter

EPWMxSYNCO
Compare Logic
Action Qualifier
Dead Band
EPWMxSYNCI
Period Register
Shadowed
EPWMxA
PWM Chopper
Trip Zone
EPWMxB TZy
SYSCLKOUT
Digital Compare
TZ1-TZ3
COMPxOUT
7-5
ePWM
ePWM Time-Base Sub-Module
ePWM Time-Base Sub-Module

Shadowed Shadowed
Clock Prescaler
Compare Register
Compare Register
TBCLK

EPWMxSYNCO
Compare Logic
Action Qualifier
Dead Band
EPWMxSYNCI
Period Register
Shadowed
EPWMxA
PWM Chopper
Trip Zone
EPWMxB TZy
SYSCLKOUT
Digital Compare
TZ1-TZ3
COMPxOUT
ePWM Time-Base Count Modes

TBCTR TBPRD Asymmetrical Waveform Count Up Mode TBCTR TBPRD Asymmetrical Waveform Count Down Mode TBCTR TBPRD Symmetrical Waveform Count Up and Down Mode
7-6
ePWM
ePWM Phase Synchronization

Ext. SyncIn (optional) Phase =0 En SyncIn EPWM1A EPWM1B To eCAP1 SyncIn EPWM2A EPWM2B =120
o o
CTR=zero o CTR=CMPB o o o X SyncOut Phase =120 SyncIn
. o . o .
En
o o
CTR=zero o CTR=CMPB o o o X SyncOut Phase =240 SyncIn
En
o o
=120 EPWM3A EPWM3B =240
o CTR=zero o CTR=CMPB o o o X
SyncOut
ePWM Time-Base Sub-Module Registers

(lab file: EPwm.c)
Name TBCTL TBSTS TBPHS TBCTR TBPRD
Description Time-Base Control Time-Base Status Time-Base Phase Time-Base Counter Time-Base Period
Structure EPwmxRegs.TBCTL.all = EPwmxRegs.TBSTS.all = EPwmxRegs.TBPHS = EPwmxRegs.TBCTR = EPwmxRegs.TBPRD =
7-7
ePWM
ePWM Time-Base Control Register

EPwmxRegs.TBCTL
Upper Register:
Phase Direction 0 = count down after sync 1 = count up after sync
15 - 14 FREE_SOFT 13 PHSDIR
TBCLK = SYSCLKOUT / (HSPCLKDIV * CLKDIV) 12 - 10 CLKDIV 9- 7 HSPCLKDIV
Emulation Halt Behavior 00 = stop after next CTR inc/dec 01 = stop when: Up Mode; CTR = PRD Down Mode; CTR = 0 Up/Down Mode; CTR = 0 1x = free run (do not stop)
TB Clock Prescale 000 = /1 (default) 001 = /2 010 = /4 011 = /8 100 = /16 101 = /32 110 = /64 111 = /128
High Speed TB Clock Prescale 000 = /1 001 = /2 (default) 010 = /4 011 = /6 100 = /8 101 = /10 110 = /12 111 = /14
(HSPCLKDIV is for legacy compatibility)
ePWM Time-Base Control Register

EPwmxRegs.TBCTL
Lower Register:
Software Force Sync Pulse 0 = no action 1 = force one-time sync
6 SWFSYNC 5-4 SYNCOSEL 3
Counter Mode 00 = count up 01 = count down 10 = count up and down 11 = stop freeze (default)
2 PHSEN 1- 0 CTRMODE
PRDLD
Sync Output Select

(source of EPWMxSYNC0 signal)
00 01 10 11
= EPWMxSYNCI = CTR = 0 = CTR = CMPB = disable SyncOut
Period Shadow Load 0 = load on CTR = 0 1 = load immediately
Phase Reg. Enable 0 = disable 1 = CTR = TBPHS on EPWMxSYNCI signal
7-8
ePWM
ePWM Compare Sub-Module
ePWM Compare Sub-Module

Shadowed Shadowed
Clock Prescaler
Compare Register
Compare Register
TBCLK

EPWMxSYNCO
Compare Logic
Action Qualifier
Dead Band
EPWMxSYNCI
Period Register
Shadowed
EPWMxA
PWM Chopper
Trip Zone
EPWMxB TZy
SYSCLKOUT
Digital Compare
TZ1-TZ3
COMPxOUT
ePWM Compare Event Waveforms

TBCTR TBPRD CMPA CMPB
= compare events are fed to the Action Qualifier Sub-Module
..
..
..
Asymmetrical Waveform
Count Up Mode TBCTR TBPRD CMPA CMPB
.. ..
Count Down Mode
..
Asymmetrical Waveform
TBCTR TBPRD CMPA CMPB
.. .. .. ..
Count Up and Down Mode
Symmetrical Waveform
7-9
ePWM
ePWM Compare Sub-Module Registers

(lab file: EPwm.c)
Name CMPCTL CMPA CMPB
Description Compare Control Compare A Compare B
Structure EPwmxRegs.CMPCTL.all = EPwmxRegs.CMPA = EPwmxRegs.CMPB =
ePWM Compare Control Register

EPwmxRegs.CMPCTL
CMPA and CMPB Shadow Full Flag
(bit automatically clears on load)
0 = shadow not full 1 = shadow full

15 - 10 reserved 9 SHDWBFULL 8 SHDWAFULL 7 reserved
6 SHDWBMODE
5 reserved
4 SHDWAMODE
3- 2 LOADBMODE
1-0 LOADAMODE
CMPA and CMPB Operating Mode 0 = shadow mode; double buffer w/ shadow register 1 = immediate mode; shadow register not used
CMPA and CMPB Shadow Load Mode 00 = load on CTR = 0 01 = load on CTR = PRD 10 = load on CTR = 0 or PRD 11 = freeze (no load possible)
7 - 10
ePWM
ePWM Action Qualifier Sub-Module
ePWM Action Qualifier Sub-Module

Shadowed Shadowed
Clock Prescaler
Compare Register
Compare Register
TBCLK

EPWMxSYNCO
Compare Logic
Action Qualifier
Dead Band
EPWMxSYNCI
Period Register
Shadowed
EPWMxA
PWM Chopper
Trip Zone
EPWMxB TZy
SYSCLKOUT
Digital Compare
TZ1-TZ3
COMPxOUT
ePWM Action Qualifier Actions

for EPWMA and EPWMB S/W Force Time-Base Counter equals: Zero CMPA CMPB TBPRD EPWM Output Actions Do Nothing
SW X SW SW SW T
Z X Z Z Z T
CA X CA CA CA T
CB X CB CB CB T
P X P P P T
Clear Low
Set High
Toggle
7 - 11
ePWM
ePWM Count Up Asymmetric Waveform

with Independent Modulation on EPWMA / B
TBCTR TBPRD
.
Z P X CB X
.
CA Z P X
.
CB X
.
CA Z P X
EPWMA Z EPWMB P X CB CA X Z P X CB CA X Z P X
ePWM Count Up Asymmetric Waveform

with Independent Modulation on EPWMA
TBCTR TBPRD
.
CA
.
CB
.
CA
.
CB
EPWMA Z T EPWMB Z T Z T
7 - 12
ePWM
ePWM Count Up-Down Symmetric Waveform

with Independent Modulation on EPWMA / B
TBCTR TBPRD
. . . .
CA CA
. . . .
CA CA
EPWMA CB EPWMB CB CB CB
ePWM Count Up-Down Symmetric Waveform

with Independent Modulation on EPWMA
TBCTR TBPRD
.
CA
.
CB
.
CA
.
CB
EPWMA Z EPWMB P Z P
7 - 13
ePWM
ePWM Action Qualifier Sub-Module Registers

(lab file: EPwm.c) Name AQCTLA AQCTLB AQSFRC AQCSFRC Description AQ Control Output A AQ Control Output B AQ S/W Force AQ Cont. S/W Force Structure EPwmxRegs.AQCTLA.all = EPwmxRegs.AQCTLB.all = EPwmxRegs.AQSFRC.all = EPwmxRegs.AQCSFRC.all =
ePWM Action Qualifier Control Register

EPwmxRegs.AQCTLy (y = A or B)
Action when CTR = CMPB on UP Count
Action when CTR = CMPA on UP Count
Action when CTR = 0
15 - 12 reserved
11 - 10 CBD
9-8 CBU
7-6 CAD
5-4 CAU
3- 2 PRD
1-0 ZRO
Action when CTR = CMPB on DOWN Count 00 01 10 11
Action when CTR = CMPA on DOWN Count
Action when CTR = PRD
= do nothing (action disabled) = clear (low) = set (high) = toggle (low high; high low)
7 - 14
ePWM
ePWM Action Qualifier S/W Force Register

EPwmxRegs.AQSFRC
One-Time S/W Force on Output B / A 0 = no action 1 = single s/w force event
15 - 8 reserved
7-6 RLDCSF
5 OTSFB
4-3 ACTSFB
2 OTSFA
1-0 ACTSFA
AQSFRC Shadow Reload Options 00 = load on event CTR = 0 01 = load on event CTR = PRD 10 = load on event CTR = 0 or CTR = PRD 11 = load immediately (from active reg.)
Action on One-Time S/W Force B / A 00 = do nothing (action disabled) 01 = clear (low) 10 = set (high) 11 = toggle (low high; high low)
ePWM Action Qualifier Continuous S/W Force Register

EPwmxRegs.AQCSFRC
15 - 4 reserved
3-2 CSFB
1-0 CSFA
Continuous S/W Force on Output B / A 00 = forcing disabled 01 = force continuous low on output 10 = force continuous high on output 11 = forcing disabled
7 - 15
ePWM
Asymmetric and Symmetric Waveform Generation using the ePWM PWM switching frequency:
The PWM carrier frequency is determined by the value contained in the time-base period register, and the frequency of the clocking signal. The value needed in the period register is: Asymmetric PWM:
switching period period register = timer period 1 period register = switching period 2(timer period)
Symmetric PWM:
Notice that in the symmetric case, the period value is half that of the asymmetric case. This is because for up/down counting, the actual timer period is twice that specified in the period register (i.e. the timer counts up to the period register value, and then counts back down).
PWM resolution:
The PWM compare function resolution can be computed once the period register value is determined. The largest power of 2 is determined that is less than (or close to) the period value. As an example, if asymmetric was 1000, and symmetric was 500, then: Asymmetric PWM: approx. 10 bit resolution since 210 = 1024 1000 Symmetric PWM: approx. 9 bit resolution since 29 = 512 500
PWM duty cycle:

Duty cycle calculations are simple provided one remembers that the PWM signal is initially inactive during any particular timer period, and becomes active after the (first) compare match occurs. The timer compare register should be loaded with the value as follows: Asymmetric PWM: TxCMPR = (100% - duty cycle) TxPR Symmetric PWM: TxCMPR = (100% - duty cycle) TxPR Note that for symmetric PWM, the desired duty cycle is only achieved if the compare registers contain the computed value for both the up-count compare and down-count compare portions of the time-base period.
7 - 16
ePWM
PWM Computation Example
Symmetric PWM Computation Example

Determine TBPRD and CMPA for 60 kHz, 25% duty symmetric PWM from a 60 MHz time base clock
fPWM = 60 kHz
Period Compare Counter
.
CA
.
CA
fTBCLK = 60 MHz
PWM Pin
TBPRD =
1 . fTBCLK 1 60 MHz = . = 500 2 fPWM 2 60 kHz
CMPA = (100% - duty cycle)*TBPRD = 0.75*500 = 375
Asymmetric PWM Computation Example

Determine TBPRD and CMPA for 60 kHz, 25% duty asymmetric PWM from a 60 MHz time base clock
fPWM = 60 kHz
Period Compare Counter

fTBCLK = 60 MHz PWM Pin
.
P
CA
TBPRD =
fTBCLK -1= fPWM
60 MHz - 1 = 999 60 kHz
CMPA = (100% - duty cycle)*(TBPRD+1) - 1 = 0.75*(999+1) - 1 = 749
7 - 17
ePWM
ePWM Dead-Band Sub-Module
ePWM Dead-Band Sub-Module

Shadowed Shadowed
Clock Prescaler
Compare Register
Compare Register
TBCLK

EPWMxSYNCO
Compare Logic
Action Qualifier
Dead Band
EPWMxSYNCI
Period Register
Shadowed
EPWMxA
PWM Chopper
Trip Zone
EPWMxB TZy
SYSCLKOUT
Digital Compare
TZ1-TZ3
COMPxOUT
Motivation for Dead-Band

supply rail
gate signals are complementary PWM
to power switching device
Transistor gates turn on faster than they shut off Short circuit if both gates are on at same time!
7 - 18
ePWM
Dead-band control provides a convenient means of combating current shoot-through problems in a power converter. Shoot-through occurs when both the upper and lower gates in the same phase of a power converter are open simultaneously. This condition shorts the power supply and results in a large current draw. Shoot-through problems occur because transistors open faster than they close, and because high-side and low-side power converter gates are typically switched in a complimentary fashion. Although the duration of the shoot-through current path is finite during PWM cycling, (i.e. the closing gate will eventually shut), even brief periods of a short circuit condition can produce excessive heating and over stress in the power converter and power supply.
ePWM Dead-Band Block Diagram

PWMxA
. .
PWMxB
1
S4
Rising Edge Delay

In Out
. .
0 S2 RED
(10-bit counter) Falling Edge Delay

In Out
S1
PWMxA
1
S5 IN-MODE
(10-bit counter)
HALFCYCLE
S3 1
POLSEL
FED
S0 0
OUT-MODE
PWMxB
Two basic approaches exist for controlling shoot-through: modify the transistors, or modify the PWM gate signals controlling the transistors. In the first case, the opening time of the transistor gate must be increased so that it (slightly) exceeds the closing time. One way to accomplish this is by adding a cluster of passive components such as resistors and diodes in series with the transistor gate, as shown in the next figure.
by-pass diode PWM signal
Shoot-through control via power circuit modification The resistor acts to limit the current rise rate towards the gate during transistor opening, thus increasing the opening time. When closing the transistor however, current flows unimpeded from the gate via the by-pass diode and closing time is therefore not affected. While this passive approach offers an inexpensive solution that is independent of the control microprocessor, it is
7 - 19
ePWM
imprecise, the component parameters must be individually tailored to the power converter, and it cannot adapt to changing system conditions. The second approach to shoot-through control separates transitions on complimentary PWM signals with a fixed period of time. This is called dead-band. While it is possible to perform software implementation of dead-band, the C28x offers on-chip hardware for this purpose that requires no additional CPU overhead. Compared to the passive approach, dead-band offers more precise control of gate timing requirements. In addition, the dead time is typically specified with a single program variable that is easily changed for different power converters or adapted on-line.
ePWM Dead-Band Sub-Module Registers

(lab file: EPwm.c)
Name DBCTL DBRED DBFED
Description Dead-Band Control 10-bit Rising Edge Delay 10-bit Falling Edge Delay
Structure EPwmxRegs.DBCTL.all = EPwmxRegs.DBRED = EPwmxRegs.DBFED =
Rising Edge Delay = T TBCLK x DBRED Falling Edge Delay = TTBCLK x DBFED
7 - 20
ePWM
ePWM Dead Band Control Register

EPwmxRegs.DBCTL
Polarity Select 00 = active high 01 = active low complementary (RED) 10 = active high complementary (FED) 11 = active low
5-4 IN_MODE 3-2 POLSEL 1-0 OUT_MODE
Half Cycle Clocking 0 = full cycle clocking (TBCLK rate) 1 = half cycle clocking (TBCLK*2 rate)
15 HALFCYCLE 14 - 6 reserved
In-Mode Control 00 = PWMxA is source for RED and FED 01 = PWMxA is source for FED PWMxB is source for RED 10 = PWMxA is source for RED PWMxB is source for FED 11 = PWMxB is source for RED and FED
Out-Mode Control 00 = disabled (DBM bypass) 01 = PWMxA = no delay PWMxB = FED 10 = PWMxA = RED PWMxB = no delay 11 = RED & FED (DBM fully enabled)
ePWM PWM Chopper Sub-Module
ePWM PWM Chopper Sub-Module

Shadowed Shadowed
Clock Prescaler
Compare Register
Compare Register
TBCLK

EPWMxSYNCO
Compare Logic
Action Qualifier
Dead Band
EPWMxSYNCI
Period Register
Shadowed
EPWMxA
PWM Chopper
Trip Zone
EPWMxB TZy
SYSCLKOUT
Digital Compare
TZ1-TZ3
COMPxOUT
7 - 21
ePWM
Purpose of the PWM Chopper

Allows a high frequency carrier signal to modulate the PWM waveform generated by the Action Qualifier and Dead-Band modules Used with pulse transformer-based gate drivers to control power switching elements
ePWM Chopper Waveform

EPWMxA EPWMxB CHPFREQ
EPWMxA
EPWMxB
Programmable Pulse Width (OSHTWTH)
OSHT
EPWMxA
Sustaining Pulses
With One-Shot Pulse on EPWMxA and/or EPWMxB
7 - 22
ePWM
ePWM Chopper Sub-Module Registers

(lab file: EPwm.c)
Name PCCTL
Description PWM-Chopper Control
Structure EPwmxRegs.PCCTL.all =
ePWM Chopper Control Register

EPwmxRegs.PCCTL
Chopper Clk Freq. 000 = SYSCLKOUT/8 001 = SYSCLKOUT/8 010 = SYSCLKOUT/8 011 = SYSCLKOUT/8 100 = SYSCLKOUT/8 101 = SYSCLKOUT/8 110 = SYSCLKOUT/8 111 = SYSCLKOUT/8
7- 5 CHPFREQ
Chopper Clk Duty Cycle 000 = 1/8 (12.5%) 001 = 2/8 (25.0%) 010 = 3/8 (37.5%) 011 = 4/8 (50.0%) 100 = 5/8 (62.5%) 101 = 6/8 (75.0%) 110 = 7/8 (87.5%) 111 = reserved
15 - 11 reserved 10 - 8
1 2 3 4 5 6 7 8
Chopper Enable 0 = disable (bypass) 1 = enable
4- 1 OSHTWTH
0 CHPEN
CHPDUTY
0000 0001 0010 0011 0100 0101 0110 0111
=1x =2x =3x =4x =5x =6x =7x =8x
One-Shot Pulse Width SYSCLKOUT/8 1000 = 9 SYSCLKOUT/8 1001 = 10 SYSCLKOUT/8 1010 = 11 SYSCLKOUT/8 1011 = 12 SYSCLKOUT/8 1100 = 13 SYSCLKOUT/8 1101 = 14 SYSCLKOUT/8 1110 = 15 SYSCLKOUT/8 1111 = 16
x x x x x x x x
SYSCLKOUT/8 SYSCLKOUT/8 SYSCLKOUT/8 SYSCLKOUT/8 SYSCLKOUT/8 SYSCLKOUT/8 SYSCLKOUT/8 SYSCLKOUT/8
7 - 23
ePWM
ePWM Digital Compare Sub-Module
ePWM Digital Compare Sub-Module

Shadowed Shadowed
Clock Prescaler
Compare Register
Compare Register
TBCLK

EPWMxSYNCO
Compare Logic
Action Qualifier
Dead Band
EPWMxSYNCI
Period Register
Shadowed
EPWMxA
PWM Chopper
Trip Zone
EPWMxB TZy
SYSCLKOUT
Digital Compare
TZ1-TZ3
COMPxOUT
Purpose of the Digital Compare Sub-Module

COMP3)
Comparator module outputs (COMP1, COMP2, and and Trip-Zone inputs (TZ1, TZ2, and TZ3) generate Digital Compare A and B High/Low Signals (DCAH, DCAL, DCBH, and DCBL) DCAH/L and DCBH/L signals trigger events which can be filtered or fed directly to the trip-zone, event-trigger, and time-base sub-modules to:
Generate a trip-zone interrupt Generate an ADC start of conversion Force an event Generate a synchronization event for synchronizing the ePWM module TBCNT
Event filtering can optionally blank the input signal to remove noise
7 - 24
ePWM
Digital Compare Sub-Module Signals

DCAH
TZ1 TZ2
DCAL
Digital Trip Event A1 Compare Digital Trip Event A2 Compare
Time-Base Sub-Module Generate PWM Sync Event-Trigger Sub-Module Generate SOCA Trip-Zone Sub-Module Trip PWMA Output Generate Trip Interrupt
TZ3 COMP1OUT
DCBH
COMP2OUT COMP3OUT
DCBL
Digital Trip Event B1 Compare Digital Trip Event B2 Compare TZDCSEL
Time-Base Sub-Module Generate PWM Sync Event-Trigger Sub-Module Generate SOCB Trip-Zone Sub-Module Trip PWMB Output Generate Trip Interrupt DCACTRL / DCBCTRL
DCTRIPSEL
The Digital Compare sub-module compares signals external to the ePWM module to directly generate events which are then feed to the Event-Trigger, Trip-Zone, and Time-Base sub-modules
ePWM Digital Compare Sub-Module Registers

(lab file: EPwm.c) Name DCACTL DCBCTL DCTRIPSEL DCCAPCTL DCCAP DCFCTL DCFOFFSETCNT DCFWINDOW Description DC A Control DC B Control DC Trip Select Capture Control Counter Capture DC Filter Control Filter Offset Ctr Filter Window Structure EPwmxRegs.DCACTL.all = EPwmxRegs.DCBCTL.all = EPwmxRegs.DCTRIPSEL.all = EPWMxRegs.DCCAPCTL.all = EPwmxRegs.DCCAP = EPwmxRegs.DCFCTL.all = EPwmxRegs.DCOFFSETCNT = EPwmxRegs.DCFWINDOW = EPwmxRegs.DCFWINDOWCNT =
DCFWINDOWCNT Filter Window Ctr
7 - 25
ePWM
ePWM Digital Compare Control Register

EPwmxRegs.DCyCTL (y = A or B)
DCy EVT2 Source Force Sync Signal Select 0 = synchronous 1 = asynchronous
DCyEVT1 SOC Generation 0 = disable 1 = enable
DCy EVT1 Source Force Sync Signal Select 0 = synchronous 1 = asynchronous
15 - 10 9 8 7-4 reserved EVT2FRC EVT2SRC reserved SYNCSEL SEL
3 EVT1 SYNCE
2 EVT1 SOCE
0 1 EVT1FRC EVT1SRC SYNCSEL SEL
DCyEVT2 Source Signal Select 0 = DCyEVT2 signal 1 = DCEVTFILT signal
DCy EVT1 SYNC Generation 0 = disable 1 = enable
DCyEVT1 Source Signal Select 0 = DCyEVT1 signal 1 = DCEVTFILT signal
ePWM Digital Compare Trip Select Register

EPwmxRegs.DCTRIPSEL
Digital Compare B Low Input Source Select
15 - 12 DCBLCOMPSEL 7- 4 DCALCOMPSEL
Digital Compare B High Input Source Select

11 - 8 DCBHCOMPSEL 3-0 DCAHCOMPSEL
Digital Compare A Low Input Source Select
Digital Compare A High Input Source Select
0000 = TZ1 input 0001 = TZ2 input 0010 = TZ3 input 1000 = COMP1OUT input 1001 = COMP2OUT input 1010 = COMP3OUT input other values reserved
7 - 26
ePWM
ePWM Trip-Zone Sub-Module
ePWM Trip-Zone Sub-Module

Shadowed Shadowed
Clock Prescaler
Compare Register
Compare Register
TBCLK

EPWMxSYNCO
Compare Logic
Action Qualifier
Dead Band
EPWMxSYNCI
Period Register
Shadowed
EPWMxA
PWM Chopper
Trip Zone
EPWMxB TZy
SYSCLKOUT
Digital Compare
TZ1-TZ3
COMPxOUT
Trip-Zone Features
Trip-Zone has a fast, clock independent logic path to high-impedance the EPWMxA/B output pins Interrupt latency may not protect hardware when responding to over current conditions or short-circuits through ISR software Supports: #1) one-shot trip for major short circuits or over current conditions #2) cycle-by-cycle trip for current limiting operation
Over Current Sensors COMPxOUT TZ1 TZ2 TZ3 TZ4 TZ5 TZ6
Digital Compare
CPU core
EPWMxTZINT Cycle-by-Cycle Mode
EPWM1A EPWM1B

EPWMxA EPWMxB
P W M O U T P U T S
eQEP1 SYSCTRL CPU
EQEP1ERR CLOCKFAIL EMUSTOP
One-Shot Mode
The power drive protection is a safety feature that is provided for the safe operation of systems such as power converters and motor drives. It can be used to inform the monitoring program of
7 - 27
ePWM
motor drive abnormalities such as over-voltage, over-current, and excessive temperature rise. If the power drive protection interrupt is unmasked, the PWM output pins will be put in the highimpedance state immediately after the pin is driven low. An interrupt will also be generated.
ePWM Trip-Zone Sub-Module Registers

(lab file: EPwm.c)
Name TZCTL TZSEL TZEINT TZDCSEL TZFLG TZCLR TZFRC
Description Trip-Zone Control Trip-Zone Select Enable Interrupt Digital Compare Trip-Zone Flag Trip-Zone Clear Trip-Zone Force
Structure EPwmxRegs.TZCTL.all = EPwmxRegs.TZSEL.all = EPwmxRegs.TZEINT.all = EPWMxRegs.TZDCSEL.all = EPwmxRegs.TZFLG.all = EPwmxRegs.TZCLR.all = EPwmxRegs.TZFRC.all =
ePWM Trip-Zone Control Register

EPwmxRegs.TZCTL
15 - 12 reserved
11 - 10
9-8
7-6
5-4
3-2 TZB
1- 0 TZA
DCBEVT2 DCBEVT1 DCAEVT2 DCAEVT1
Digital Compare Output Digital Compare Output TZ1 to TZ6 Action on Event 2 / 1 Action EPWMxB / EPWMxA Event 2 / 1 Action on EPWMxB on EPWMxA
00 = 01 = 10 = 11 =
high impedance force high force low do nothing (disable)
7 - 28
ePWM
ePWM Trip-Zone Select Register

EPwmxRegs.TZSEL
One-Shot Trip Zone
(event only cleared under S/W control; remains latched)
0 = disable as trip source 1 = enable as trip source

15 14 13 OSHT6 5 CBC6 12 OSHT5 4 CBC5 11 OSHT4 3 CBC4 10 OSHT3 2 CBC3 9 OSHT2 1 CBC2 8 OSHT1 0 CBC1
DCBEVT1 DCAEVT1 7 6
DCBEVT2 DCAEVT2
Cycle-by-Cycle Trip Zone

(event cleared when CTR = 0; i.e. cleared every PWM cycle)
0 = disable as trip source 1 = enable as trip source
ePWM Trip-Zone Enable Interrupt Register

EPwmxRegs.TZEINT
15 - 7
2 OST
1 CBC
0 reserved
reserved DCBEVT2 DCBEVT1 DCAEVT2 DCAEVT1
Digital Compare Output B Event 2 / 1 Enable 0 = disable 1 = enable
Digital Compare Output A Event 2 / 1 Enable 0 = disable 1 = enable
One-Shot Interrupt Enable 0 = disable 1 = enable
Cycle-by-Cycle Interrupt Enable 0 = disable 1 = enable
7 - 29
ePWM
ePWM Trip-Zone Digital Compare Event Select Register

EPwmxRegs.TZDCSEL
15 - 12 reserved
11 - 9 DCBEVT2
8-6 DCBEVT1
5-3 DCAEVT2
2-0 DCAEVT1
Digital Compare Output B Event 2 / 1 Select 000 001 010 011 100 101 11x = event disable = DCBH low, DCBL = DCBH high, DCBL low, DCBH = DCBL = DCBL high, DCBH = DCBL high, DCBH = reserved
Digital Compare Output A Event 2 / 1 Select
dont care dont care dont care dont care low
ePWM Event-Trigger Sub-Module
ePWM Event-Trigger Sub-Module

Shadowed Shadowed
Clock Prescaler
Compare Register
Compare Register
TBCLK

EPWMxSYNCO
Compare Logic
Action Qualifier
Dead Band
EPWMxSYNCI
Period Register
Shadowed
EPWMxA
PWM Chopper
Trip Zone
EPWMxB TZy
SYSCLKOUT
Digital Compare
TZ1-TZ3
COMPxOUT
7 - 30
ePWM
ePWM Event-Trigger Interrupts and SOC

TBCTR TBPRD CMPB CMPA
. . . .
. . . .
EPWMA EPWMB
CTR = 0 CTR = PRD CTR = 0 or PRD CTRU = CMPA CTRD = CMPA CTRU = CMPB CTRD = CMPB
ePWM Event-Trigger Sub-Module Registers

(lab file: EPwm.c) Name ETSEL ETPS ETFLG ETCLR ETFRC Description Event-Trigger Selection Event-Trigger Pre-Scale Event-Trigger Flag Event-Trigger Clear Event-Trigger Force Structure EPwmxRegs.ETSEL.all = EPwmxRegs.ETPS.all = EPwmxRegs.ETFLG.all = EPwmxRegs.ETCLR.all = EPwmxRegs.ETFRC.all =
7 - 31
ePWM
ePWM Event-Trigger Selection Register

EPwmxRegs.ETSEL
Enable SOCB / A 0 = disable 1 = enable Enable EPWMxINT 0 = disable 1 = enable
15 SOCBEN
14 - 12 SOCBSEL
11 SOCAEN
10 - 8 SOCASEL
7-4 reserved
3 INTEN
2-0 INTSEL
EPWMxSOCB / A Select 000 = DCBEVT1 / DCAEVT1 001 = CTR = 0 010 = CTR = PRD 011 = CTR = 0 or PRD 100 = CTRU = CMPA 101 = CTRD = CMPA 110 = CTRU = CMPB 111 = CTRD = CMPB
EPWMxINT Select 000 = reserved 001 = CTR = 0 010 = CTR = PRD 011 = CTR = 0 or PRD 100 = CTRU = CMPA 101 = CTRD = CMPA 110 = CTRU = CMPB 111 = CTRD = CMPB
ePWM Event-Trigger Prescale Register

EPwmxRegs.ETPS
EPWMxSOCB / A Counter
(number of events have occurred)
EPWMxINT Counter
(number of events have occurred)
00 01 10 11
15 - 14 SOCBCNT
= no events = 1 event = 2 events = 3 events

13 - 12 SOCBPRD 11 - 10 SOCACNT 9- 8
00 01 10 11
= no events = 1 event = 2 events = 3 events

7-4 reserved 2-3 1-0
SOCAPRD
INTCNT INTPRD
EPWMxSOCB / A Period
(number of events before SOC)
EPWMxINT Period
(number of events before INT)
00 01 10 11
= disabled = SOC on first event = SOC on second event = SOC on third event
00 01 10 11
= disabled = INT on first event = INT on second event = INT on third event
7 - 32
ePWM
Hi-Resolution PWM (HRPWM)
Hi-Resolution PWM (HRPWM)

PWM Period Regular PWM Step (i.e. 16.67 ns)
Device Clock (i.e. 60 MHz)
HRPWM divides a clock cycle into smaller steps called Micro Steps (Step Size ~= 150 ps)
ms
ms
ms
ms
ms
ms
Calibration Logic
Calibration Logic tracks the number of Micro Steps per clock to account for variations caused by Temp/Volt/Process
HRPWM Micro Step (~150 ps)
Significantly increases the resolution of conventionally derived digital PWM Uses 8-bit extensions to Compare registers (CMPxHR), Period register (TBPRDHR) and Phase register (TBPHSHR) for edge positioning control Typically used when PWM resolution falls below ~9-10 bits which occurs at frequencies greater than ~120 kHz (with system clock of 60 MHz) Not all ePWM outputs support HRPWM feature (see device datasheet)
7 - 33
eCAP
eCAP
Capture Module (eCAP)
Timer Trigger pin Timestamp Values
The eCAP module timestamps transitions on a capture input pin
The capture units allow time-based logging of external TTL signal transitions on the capture input pins. The C28x has up to six capture units. Capture units can be configured to trigger an A/D conversion that is synchronized with an external event. There are several potential advantages to using the capture for this function over the ADCSOC pin associated with the ADC module. First, the ADCSOC pin is level triggered, and therefore only low to high external signal transitions can start a conversion. The capture unit does not suffer from this limitation since it is edge triggered and can be configured to start a conversion on either rising edges or falling edges. Second, if the ADCSOC pin is held high longer than one conversion period, a second conversion will be immediately initiated upon completion of the first. This unwanted second conversion could still be in progress when a desired conversion is needed. In addition, if the end-of-conversion ADC interrupt is enabled, this second conversion will trigger an unwanted interrupt upon its completion. These two problems are not a concern with the capture unit. Finally, the capture unit can send an interrupt request to the CPU while it simultaneously initiates the A/D conversion. This can yield a time savings when computations are driven by an external event since the interrupt allows preliminary calculations to begin at the start-of-conversion, rather than at the end-of-conversion using the ADC end-of-conversion interrupt. The ADCSOC pin does not offer a start-of-conversion interrupt. Rather, polling of the ADCSOC bit in the control register would need to be performed to trap the externally initiated start of conversion.
7 - 34
eCAP
Some Uses for the Capture Module

Measure the time width of a pulse Low speed velocity estimation from incr. encoder:
Problem: At low speeds, calculation of speed x -x based on a measured position change at vk k k-1 t fixed time intervals produces large estimate errors Alternative: Estimate the speed using a measured time interval at fixed position intervals
vk
x tk - tk-1
Signal from one quadrature encoder channel
Auxiliary PWM generation
eCAP Module Block Diagram Capture Mode

CAP1POL
Capture 1 Register
Polarity Select 1
CAP2POL
Event Logic
Capture 2 Register 32-Bit Time-Stamp Counter Capture 3 Register

SYSCLKOUT
Polarity Select 2
PRESCALE
CAP3POL
Event Prescale
Polarity Select 3
ECAPx pin
CAP4POL
Capture 4 Register
Polarity Select 4
7 - 35
eCAP
eCAP Module Block Diagram APWM Mode

Shadowed
immediate mode
Period Register (CAP1)
Period shadow Register mode (CAP3)
32-Bit Time-Stamp Counter

SYSCLKOUT
PWM Compare Logic
ECAP pin
immediate mode
Compare Register (CAP2)
Compare Register (CAP4) Shadowed
shadow mode
eCAP Module Registers

(lab file: ECap.c) Name ECCTL1 ECCTL2 TSCTR CTRPHS CAP1 CAP2 CAP3 CAP4 ECEINT ECFLG ECCLR ECFRC Description Capture Control 1 Capture Control 2 Time-Stamp Counter Counter Phase Offset Capture 1 Capture 2 Capture 3 Capture 4 Enable Interrupt Interrupt Flag Interrupt Clear Interrupt Force Structure ECapxRegs.ECCTL1.all = ECapxRegs.ECCTL2.all = ECapxRegs.TSCTR = ECapxRegs.CTRPHS = ECapxRegs.CAP1 = ECapxRegs.CAP2 = ECapxRegs.CAP3 = ECapxRegs.CAP4 = ECapxRegs.ECEINT.all = ECapxRegs.ECFLG.all = ECapxRegs.ECCLR.all = ECapxRegs.ECFRC.all =
7 - 36
eCAP
eCAP Control Register 1

ECapxRegs.ECCTL1
Upper Register:
CAP1 4 Load on Capture Event 0 = disable 1 = enable
15 - 14 FREE_SOFT 13 - 9 PRESCALE 8 CAPLDEN
Emulation Control 00 = TSCTR stops immediately 01 = TSCTR runs until equals 0 1X = free run (do not stop)
Event Filter Prescale Counter 00000 = divide by 1 (bypass) 00001 = divide by 2 00010 = divide by 4 00011 = divide by 6 00100 = divide by 8 11110 = divide by 60 11111 = divide by 62

ECapxRegs.ECCTL1
Lower Register:
Counter Reset on Capture Event 0 = no reset (absolute time stamp mode) 1 = reset after capture (difference mode)
CTRRST4 CAP4POL CTRRST3 CAP3POL CTRRST2 CAP2POL CTRRST1 CAP1POL
Capture Event Polarity 0 = trigger on rising edge 1 = trigger on falling edge
7 - 37
eCAP

ECapxRegs.ECCTL2
Upper Register:
Capture / APWM mode 0 = capture mode 1 = APWM mode
15 - 11 reserved 10 APWMPOL 9 8
CAP_APWM SWSYNC
APWM Output Polarity

(valid only in APWM mode)
0 = active high output 1 = active low output
Software Force Counter Synchronization 0 = no effect 1 = TSCTR load of current module and other modules if SYNCO_SEL bits = 00

ECapxRegs.ECCTL2
Lower Register:
Counter Sync-In 0 = disable 1 = enable
7-6 SYNCO_SEL 5 SYNCI_EN
Re-arm
(capture mode only)
Continuous/One-Shot
(capture mode only)
0 = no effect 1 = arm sequence

4 3 REARM
0 = continuous mode 1 = one-shot mode

2-1 0
TSCTRSTOP
STOP_WRAP CONT_ONESHT
Sync-Out Select 00 = sync-in to sync-out 01 = CTR = PRD event generates sync-out 1X = disable
Time Stamp Counter Stop 0 = stop 1 = run
Stop Value for One-Shot Mode/ Wrap Value for Continuous Mode
(capture mode only)
00 01 10 11
= stop/wrap = stop/wrap = stop/wrap = stop/wrap
after capture after capture after capture after capture
event 1 event 2 event 3 event 4
7 - 38
eCAP
The capture unit interrupts offer immediate CPU notification of externally captured events. In situations where this is not required, the interrupts can be masked and flag testing/polling can be used instead. This offers increased flexibility for resource management. For example, consider a servo application where a capture unit is being used for low-speed velocity estimation via a pulsing sensor. The velocity estimate is not used until the next control law calculation is made, which is driven in real-time using a timer interrupt. Upon entering the timer interrupt service routine, software can test the capture interrupt flag bit. If sufficient servo motion has occurred since the last control law calculation, the capture interrupt flag will be set and software can proceed to compute a new velocity estimate. If the flag is not set, then sufficient motion has not occurred and some alternate action would be taken for updating the velocity estimate. As a second example, consider the case where two successive captures are needed before a computation proceeds (e.g. measuring the width of a pulse). If the width of the pulse is needed as soon as the pulse ends, then the capture interrupt is the best option. However, the capture interrupt will occur after each of the two captures, the first of which will waste a small number of cycles while the CPU is interrupted and then determines that it is indeed only the first capture. If the width of the pulse is not needed as soon as the pulse ends, the CPU can check, as needed, the capture registers to see if two captures have occurred, and proceed from there.
eCAP Interrupt Enable Register

ECapxRegs.ECEINT
CTR = CMP Interrupt Enable
CTR = Overflow Interrupt Enable
Capture Event 3 Interrupt Enable
15 - 8
reserved CTR=CMP CTR=PRD CTROVF CEVT4 CEVT3 CEVT2 CEVT1 reserved
CTR = PRD Interrupt Enable
0 = disable as interrupt source 1 = enable as interrupt source
7 - 39
eQEP
eQEP
What is an Incremental Quadrature Encoder?

A digital (angular) position sensor
photo sensors spaced /4 deg. apart slots spaced deg. apart light source (LED)
/4
Ch. A Ch. B
shaft rotation Incremental Optical Encoder Quadrature Output from Photo Sensors
The eQEP circuit, when enabled, decodes and counts the quadrature encoded input pulses. The QEP circuit can be used to interface with an optical encoder to get position and speed information from a rotating machine.
How is Position Determined from Quadrature Signals?

Position resolution is /4 degrees
(00) (11) (A,B) = (10) (01)
increment counter
10
decrement counter
Ch. A Ch. B
00
Illegal Transitions; generate phase error interrupt
11
01 Quadrature Decoder State Machine
7 - 40
eQEP
eQEP Module Block Diagram

Measure the elapsed time between the unit position events; used for low speed measurement Generate periodic interrupts for velocity calculations
Quadrature Capture
Quadrature clock mode Monitors the quadrature clock to indicate proper operation of the motion control system Direction count mode
EQEPxA/XCLK EQEPxB/XDIR
32-Bit Unit Time-Base
QEP Watchdog
Quadrature Decoder
EQEPxI EQEPxS
SYSCLKOUT
Position/Counter Compare
Generate a sync output and/or interrupt on a position compare match Generate the direction and clock for the position counter in quadrature count mode
eQEP Module Connections

Ch. A
Quadrature Capture
Ch. B
EQEPxA/XCLK 32-Bit Unit Time-Base QEP Watchdog

SYSCL KOUT
EQEPxB/XDIR Quadrature Decoder EQEPxI EQEPxS Position/Counter Compare
Index Strobe
from homing sensor
7 - 41
Lab 7: Control Peripherals

Objective
The objective of this lab is to become familiar with the programming and operation of the control peripherals and their interrupts. ePWM1A will be setup to generate a 2 kHz, 25% duty cycle symmetric PWM waveform. The waveform will then be sampled with the on-chip analog-todigital converter and displayed using the graphing feature of Code Composer Studio. Next, eCAP1 will be setup to detect the rising and falling edges of the waveform. This information will be used to determine the width of the pulse and duty cycle of the waveform. The results of this step will be viewed numerically in a memory window.

ePWM1
TB Counter Compare Action Qualifier
connector wire
ADC
RESULT0
eCAP1
Capture 1 Register Capture 2 Register Capture 3 Register Capture 4 Register
ADCINA0
View ADC buffer PWM Samples
ePWM2
Procedure
Project File
1. A project named Lab7.pjt has been created for this lab. Open the project by clicking on Project Open and look in C:\C28x\Labs\Lab7. All Build Options have been configured the same as the previous lab. The files used in this lab are: Adc.c Gpio.c CodeStartBranch.asm Lab_5_6_7.cmd DefaultIsr_7.c Main_7.c DelayUs.asm PieCtrl_5_6_7_8_9_10.c DSP2833x_GlobalVariableDefs.c PieVect_5_6_7_8_9_10.c DSP2833x_Headers_nonBIOS.cmd SysCtrl.c ECap_7_8_9_10_12.c Watchdog.c EPwm_7_8_9_10_12.c
7 - 42
pointer rewind
data memory
...
Setup Shared I/O and ePWM1

2. Edit Gpio.c and adjust the shared I/O pin in GPIO0 for the PWM1A function. 3. In EPwm_7_8_9_10_12.c, setup ePWM1 to implement the PWM waveform as described in the objective for this lab. The following registers need to be modified: TBCTL (set clock prescales to divide-by-1, no software force, sync and phase disabled), TBPRD, CMPA, CMPCTL (load on 0 or PRD), and AQCTLA (set on up count and clear on down count for output A). Software force, deadband, PWM chopper and trip action has been disabled. (Hint notice the last steps enable the timer count mode and enable the clock to the ePWM module). Either calculate the values for TBPRD and CMPA (as a challenge) or make use of the global variable names and values that have been set using #define in the beginning of Lab.h file. Notice that ePWM2 has been initialized earlier in the code for the ADC lab. Save your work.
Build and Load

4. Save all changes to the files and click the Build button to build and load the project.
Run the Code PWM Waveform

5. Open a memory window to view some of the contents of the ADC results buffer. The address label for the ADC results buffer is AdcBuf. We will be running our code in realtime mode, and will have our window continuously refresh. 6. Using a connector wire provided, connect the PWM1A (pin # GPIO-00) to ADCINA0 (pin # ADC-A0) on the Docking Station.
Realtime 7. Run the code (real-time mode) using the GEL function: GEL Emulation Control Run_Realtime_with_Reset. Watch the window update. Verify that the ADC result buffer contains the updated values.
8. Open and setup a graph to plot a 50-point window of the ADC results buffer. Click: View Graph Time/Frequency and set the following values: Start Address Acquisition Buffer Size Display Data Size DSP Data Type Sampling Rate (Hz) Time Display Unit Select OK to save the graph options. AdcBuf 50 50 16-bit unsigned integer 50000
s
7 - 43
9. The graphical display should show the generated 2 kHz, 25% duty cycle symmetric PWM waveform. The period of a 2 kHz signal is 500 s. You can confirm this by measuring the period of the waveform using the graph (you may want to enlarge the graph window using the mouse). The measurement is best done with the mouse. The lower left-hand corner of the graph window will display the X and Y-axis values. Subtract the X-axis values taken over a complete waveform period (you can use the PC calculator program found in Microsoft Windows to do this).
Frequency Domain Graphing Feature of Code Composer Studio

10. Code Composer Studio also has the ability to make frequency domain plots. It does this by using the PC to perform a Fast Fourier Transform (FFT) of the DSP data. Let's make a frequency domain plot of the contents in the ADC results buffer (i.e. the PWM waveform). Click: View
Graph Time/Frequency and set the following values:
Display Type Start Address Acquisition Buffer Size FFT Framesize DSP Data Type Sampling Rate (Hz) Select OK to save the graph options.
FFT Magnitude AdcBuf 50 50 16-bit unsigned integer 50000
11. On the plot window, left-click the mouse to move the vertical marker line and observe the frequencies of the different magnitude peaks. Do the peaks occur at the expected frequencies? 12. Fully halt the CPU (real-time mode) by using the GEL function: GEL Emulation Control Full_Halt.
Realtime
Setup eCAP1 to Measure Width of Pulse

The first part of this lab exercise generated a 2 kHz, 25% duty cycle symmetric PWM waveform which was sampled with the on-chip analog-to-digital converter and displayed using the graphing feature of Code Composer Studio. Next, eCAP1 will be setup to detect the rising and falling edges of the waveform. This information will be used to determine the period and duty cycle of the waveform. The results of this step will be viewed numerically in a memory window and can be compared to the results obtained using the graphing features of Code Composer Studio. 13. Add the following file to the project:
ECap_7_8_9_10_12.c
7 - 44
Check your files list to make sure the file is there. 14. In Main_7.c, add code to call the InitECap() function. There are no passed parameters or return values, so the call code is simply:
InitECap();
15. Edit Gpio.c and adjust the shared I/O pin in GPIO5 for the ECAP1 function. 16. Open and inspect the eCAP1 interrupt service routine (ECAP1_INT_ISR) in the file DefaultIsr_7.c. Notice that PwmDuty is calculated by CAP2 CAP1 (rising to falling edge) and that PwmPeriod is calculated by CAP3 CAP1 (rising to rising edge). 17. In ECap_7_8_9_10_12.c, setup eCAP1 to calculate PWM_duty and PWM_period. The following registers need to be modified: ECCTL2 (continuous mode, re-arm disable, and sync disable), ECCTL1 (set prescale to divide-by-1, configure capture event polarity without reseting the counter), and ECEINT (enable desired eCAP interrupt). 18. Using the PIE Interrupt Assignment Table find the location for the eCAP1 interrupt ECAP1_INT and fill in the following information:
PIE group #:
# within group:
This information will be used in the next step. 19. Modify the end of ECap_7_8_9_10_12.c to do the following: - Enable the ECAP1_INT interrupt in the PIE (Hint: use the PieCtrlRegs structure) - Enable the appropriate core interrupt in the IER register
Build and Load

20. Save all changes to the files and click the Build button.
Run the Code Pulse Width Measurement

21. Open a memory window to view the address label PwmPeriod. (Type &PwmPeriod in the address box). The address label PwmDuty (address &PwmDuty) should appear in the same memory window. 22. Set the memory window properties format to 32-Bit UnSigned Int. 23. Using the connector wire provided, connect the PWM1A (pin # GPIO-00) to ECAP1 (pin # GPIO-05) on the Docking Station. 24. Run the code (real-time mode) by using the GEL function: GEL Realtime Emulation Control Run_Realtime_with_Reset. Notice the values for PwmDuty and PwmPeriod. 25. Fully halt the CPU (real-time mode) by using the GEL function: GEL Emulation Control Full_Halt.
Realtime
7 - 45
Questions:
How do the captured values for PwmDuty and PwmPeriod relate to the compare register CMPA and time-base period TBPRD settings for ePWM1A? What is the value of PwmDuty in memory? What is the value of PwmPeriod in memory? How does it compare with the expected value?
End of Exercise
7 - 46
Numerical Concepts
Introduction
In this module, numerical concepts will be explored. One of the first considerations concerns multiplication how does the user store the results of a multiplication, when the process of multiplication creates results larger than the inputs. A similar concern arises when considering accumulation especially when long summations are performed. Next, floating-point concepts will be explored and IQmath will be described as a technique for implementing a virtual floatingpoint system to simplify the design process. The IQmath Library is a collection of highly optimized and high precision mathematical functions used to seamlessly port floating-point algorithms into fixed-point code. These C/C++ routines are typically used in computationally intensive real-time applications where optimal execution speed and high accuracy is needed. By using these routines a user can achieve execution speeds considerable faster than equivalent code written in standard ANSI C language. In addition, by incorporating the ready-to-use high precision functions, the IQmath library can shorten significantly a DSP application development time. (The IQmath user's guide is included in the application zip file, and can be found in the /docs folder once the file is extracted and installed).
Learning Objectives
Learning Objectives
Integers and Fractions IEEE-754 Floating-Point IQmath Format Conversion of ADC Results
C2000 Piccolo Workshop - Numerical Concepts
8-1
Module Topics
Module Topics
Numerical Concepts .................................................................................................................................. 8-1 Module Topics......................................................................................................................................... 8-2 Numbering System Basics ....................................................................................................................... 8-3 Binary Numbers.................................................................................................................................. 8-3 Two's Complement Numbers ............................................................................................................. 8-3 Integer Basics ..................................................................................................................................... 8-4 Sign Extension Mode.......................................................................................................................... 8-5 Binary Multiplication.............................................................................................................................. 8-6 Binary Fractions ..................................................................................................................................... 8-8 Representing Fractions in Binary ....................................................................................................... 8-8 Fraction Basics ................................................................................................................................... 8-8 Multiplying Binary Fractions ............................................................................................................. 8-9 Fraction Coding.....................................................................................................................................8-11 Fractional vs. Integer Representation....................................................................................................8-12 Floating-Point........................................................................................................................................8-13 IQmath ...................................................................................................................................................8-15 IQ Fractional Representation.............................................................................................................8-15 Traditional Q Math Approach........................................................................................................8-16 IQmath Approach ..............................................................................................................................8-18 IQmath Library ......................................................................................................................................8-23 Converting ADC Results into IQ Format...............................................................................................8-25 AC Induction Motor Example ................................................................................................................8-26 IQmath Summary ...................................................................................................................................8-32 Lab 8: IQmath & Floating-Point FIR Filter..........................................................................................8-33
8-2
Numbering System Basics

Given the ability to perform arithmetic processes (addition and multiplication) with the C28x, it is important to understand the underlying mathematical issues which come into play. Therefore, we shall examine the numerical concepts which apply to the C28x and, to a large degree, most processors.
Binary Numbers
The binary numbering system is the simplest numbering scheme used in computers, and is the basis for other schemes. Some details about this system are: It uses only two values: 1 and 0 Each binary digit, commonly referred to as a bit, is one place in a binary number and represents an increasing power of 2. The least significant bit (LSB) is to the right and has the value of 1. Values are represented by setting the appropriate 1's in the binary number. The number of bits used determines how large a number may be represented.
Examples:
01102 = (0 * 8) + (1 * 4) + (1 * 2) + (0 * 1) = 610 111102 = (1 * 16) + (1 * 8) + (1 * 4) + (1 * 2) + (0 * 1) = 3010
Two's Complement Numbers

Notice that binary numbers can only represent positive numbers. Often it is desirable to be able to represent both positive and negative numbers. The two's complement numbering system modifies the binary system to include negative numbers by making the most significant bit (MSB) negative. Thus, two's complement numbers: Follow the binary progression of simple binary except that the MSB is negative in addition to its magnitude Can have any number of bits more bits allow larger numbers to be represented
Examples:
01102 = (0 * -8) + (1 * 4) + (1 * 2) + (0 * 1) = 610 111102 = (1 * -16) + (1 * 8) + (1 * 4) + (1 * 2) + (0 * 1) = -210
The same binary values are used in these examples for two's complement as were used above for binary. Notice that the decimal value is the same when the MSB is 0, but the decimal value is quite different when the MSB is 1. Two operations are useful in working with two's complement numbers: The ability to obtain an additive inverse of a value The ability to load small numbers into larger registers (by sign extending)
8-3
To load small two's complement numbers into larger registers:

The MSB of the original number must carry to the MSB of the number when represented in the larger register. 1. Load the small number right justified into the larger register. 2. Copy the sign bit (the MSB) of the original number to all unfilled bits to the left in the register (sign extension). Consider our two previous values, copied into an 8-bit register:
Examples:
Original No. 1. Load low 2. Sign Extend 0 1 1 02 0110 00000110 =4+2=6 = 610 1 1 1 1 02 11110 11111110 = -128 + 64 + ... + 2 = -2 = -210
Integer Basics
Integer Basics
2n-1 2n-1
233 2
222 2
211 2
200 2
Unsigned Binary Integers

0100b = (0*23)+(1*22)+(0*22)+(0*20) = 4 1101b = (1*23)+(1*22)+(0*21)+(1*20) = 13
Signed Binary Integers (2s Complement)

0100b = (0*-23)+(1*22)+(0*22)+(0*20) = 4 1101b = (1*-23)+(1*22)+(0*21)+(1*2 0) = -3
8-4
Sign Extension Mode

The C28x can operate on either unsigned binary or two's complement operands. The Sign Extension Mode (SXM) bit, present within a status register of the C28x, identifies whether or not the sign extension process is used when a value is brought into the accumulator. It is good programming practice to always select the desired SXM at the beginning of a module to assure the proper mode.
What is Sign Extension?

When moving a value from a narrowed width location to a wider width location, the sign bit is extended to fill the width of the destination Sign extension applies to signed numbers only It keeps negative numbers negative! Sign extension controlled by SXM bit in ST0 register; When SXM = 1, sign extension happens automatically 4 bit Example: Load a memory value into the ACC memory
1101 = -23 + 22 + 20 = -3
Load and sign extend
ACC
1111 1101 = -27 + 26 + 25 + 24 + 23 + 22 + 20

= -128 + 64 + 32 + 16 + 8 + 4 + 1 = -3
8-5
Binary Multiplication
Now that you understand two's complement numbers, consider the process of multiplying two two's complement values. As with long hand decimal multiplication, we can perform binary multiplication one place at a time, and sum the results together at the end to obtain the total product. Note: This is not the method the C28x uses in multiplying numbers it is merely a way of observing how binary numbers work in arithmetic processes. The C28x uses 16-bit operands and a 32-bit accumulator. For the sake of clarity, consider the example below where we shall investigate the use of 4-bit values and an 8-bit accumulation:
Integer Multiplication (signed)

0100 x 1101 00000100 0000000 000100 11100 11110100
Accumulator
4 -3
-12
11110100 11110100 ?
Data Memory
In this example, consider the following: What are the two input values, and the expected result? Why are the partial products shifted left as the calculation continues? Why is the final partial product different than the others? What is the result obtained when adding the partial products? How shall this result be loaded into the accumulator? How shall we fill the remaining bit? Is this value still the expected one? How can the result be stored back to memory? What problems arise?
8-6
Note: With twos complement multiplication, the leading 1 in the second multiplicand is a sign bit. If the sign bit is 1, then take the 2s complement of the first multiplicand. Additionally, each partial product must be sign-extended for correct computation.
Note: All of the above questions except the final one are addressed in this module. The last question may have several answers:
Store the lower accumulator to memory. What problem is apparent using this method in this example? Store the upper accumulator back to memory. Wouldn't this create a loss of precision, and a problem in how to interpret the results later? Store both the upper and lower accumulator to memory. This solves the above problems, but creates some new ones: Extra code space, memory space, and cycle time are used How can the result be used as the input to a subsequent calculation? Is such a condition likely (consider any feedback system)?
From this analysis, it is clear that integers do not behave well when multiplied. Might some other type of number system behave better? Is there a number system where the results of a multiplication are bounded?
8-7
Binary Fractions
Binary Fractions
Given the problems associated with integers and multiplication, consider the possibilities of using fractional values. Fractions do not grow when multiplied, therefore, they remain representable within a given word size and solve the problem. Given the benefit of fractional multiplication, consider the issues involved with using fractions: How are fractions represented in two's complement? What issues are involved when multiplying two fractions?
Representing Fractions in Binary

In order to represent both positive and negative values, the two's complement process will again be used. However, in the case of fractions, we will not set the LSB to 1 (as was the case for integers). When one considers that the range of fractions is from -1 to ~+1, and that the only bit which conveys negative information is the MSB, it seems that the MSB must be the negative ones position. Since binary representation is based on powers of two, it follows that the next bit would be the one-halves position, and that each following bit would have half the magnitude again. Considering, as before, a 4-bit model, we have the representation shown in the following example. 1 -1 . 0 1/2 1 1/4 1 1/8 = -1 + 1/4 + 1/8 = -5/8
Fraction Basics
Fraction Basics
-200 -2 2-1 2-1 2-2 2-2 2-3 2-3 2-(n-1) 2-(n-1)
1101b = (1*-20)+(1*2-1)+(0*2-2)+(1*2-3) = -1 + 1/2 + 1/8 = -3/8
Fractions have the nice property that fraction x fraction = fraction
8-8
Binary Fractions
Multiplying Binary Fractions

When the C28x performs multiplication, the process is identical for all operands, integers or fractions. Therefore, the user must determine how to interpret the results. As before, consider the 4-bit multiply example:
Fraction Multiplication
0100 . x 1101 . 00000100 0000000 000100 11100 11110100
Accumulator x
1/2 -3/8
-3/16
11110100 11110100 1110 . -1/4
Data Memory
As before, consider the following: What are the two input values and the expected result? As before, partial products are shifted left and the final is negative. How is the result (obtained when adding the partial products) read? How shall this result be loaded into the accumulator? How shall we fill the remaining bit? Is this value still the expected one? How can the result be stored back to memory? What problems arise?
To read the results of the fractional multiply, it is necessary to locate the binary point (the base 2 equivalent of the base 10 decimal point). Start by identifying the location of the binary point in the input values. The MSB is an integer and the next bit is 1/2, therefore, the binary point would be located between them. In our example, therefore, we would have three bits to the right of the binary point in each input value. For ease of description, we can refer to these as Q3 numbers, where Q refers to the number of places to the right of the point. When multiplying numbers, the Q values add. Thus, we would (mentally) place a binary point above the sixth LSB. We can now calculate the Q6 result more readily.
8-9
Binary Fractions
As with integers, the results are loaded low and the MSB is a sign extension of the seventh bit. If this value were loaded into the accumulator, we could store the results back to memory in a variety of ways: Store both low and high accumulator values back to memory. This offers maximum detail, but has the same problems as with integer multiply. Store only the high (or low) accumulator back to memory. This creates a potential for a memory littered with varying Q-types. Store the upper accumulator shifted to the left by 1. This would store values back to memory in the same Q format as the input values, and with equal precision to the inputs. How shall the left shift be performed? Heres three methods: Explicit shift (C or assembly code) Shift on store (assembly code) Use Product Mode shifter (assembly code)
8 - 10
Fraction Coding
Fraction Coding
Although COFF tools recognize values in integer, hex, binary, and other forms, they understand only integer, or non-fractional values. To use fractions within the C28x, it is necessary to describe them as though they were integers. This turns out to be a very simple trick. Consider the following number lines:
Coding Traditional 16-bit Q15 Fractions

~1 0 - -1 32767 0x7FFF 0x4000 0x0000 0xC000 0x8000
32768 (215)
16384 0 -16384 -32768
Fraction C-code example: y = 0.707 * x
Integer
void main(void) { int16 coef = 32768*707/1000; // 0.707 in Q15 int16 x, y; y = (int16)( (int32)coef * (int32)x ) >> 15); }
By multiplying a fraction by 32K (32768), a normalized fraction is created, which can be passed through the COFF tools as an integer. Once in the C28x, the normalized fraction looks and behaves exactly as a fraction. Thus, when using fractional constants in a C28x program, the coder first multiplies the fraction by 32768, and uses the resulting integer (rounded to the nearest whole value) to represent the fraction. The following is a simple, but effective method for getting fractions past the assembler: 1. Express the fraction as a decimal number (drop the decimal point). 2. Multiply by 32768. 3. Divide by the proper multiple of 10 to restore the decimal position.
Examples:
To represent 0.62: To represent 0.1405: 32768 x 62 / 100 32768 x 1405 / 10000
This method produces a valid number accurate to 16 bits. You will not need to do the math yourself, and changing values in your code becomes rather simple.
8 - 11
Fractional vs. Integer Representation
Fractional vs. Integer Representation
Integer vs. Fractions

Range Integer Fraction determined by # of bits ~+1 to -1 Precision 1 determined by # of bits
Integers grow when you multiply them Fractions have limited range
Fractions can still grow when you add them Scaling an application is time consuming
Are there any other alternatives?

The C28x accumulator, a 32-bit register, adds extra range to integer calculations, but this becomes a problem in storing the results back to 16-bit memory. Conversely, when using fractions, the extra accumulator bits increase precision, which helps minimize accumulative errors. Since any number is accurate (at best) to one-half of a LSB, summing two of these values together would yield a worst case result of 1 LSB error. Four summations produce two LSBs of error. By 256 summations, eight LSBs are noisy. Since the accumulator holds 32 bits of information, and fractional results are stored from the high accumulator, the extra range of the accumulator is a major benefit in noise reduction for long sum-of-products type calculations.
8 - 12
Floating-Point
Floating-Point
IEEE-754 Single Precision Floating-Point

31 30 1 bit sign 23 22 23 bit mantissa (fraction) then v = NaN then v = [(-1)s]*infinity then v = [(-1)s]*[2(e- 127)]*(1.f) then v = [(-1)s]*[2(-1 26)]*(0.f) then v = [(-1)s]*0 0
s eeeeeeee fffffffffffffffffffffff
8 bit exponent if e = 255 and f 0, if e = 255 and f = 0, if 0 < e < 255, if e = 0 and f 0, if e = 0 and f = 0, Case 1: Case 2: Case 3: Case 4: Case 5:
Normalized values
Example: 0x41200000 = 0 100 0001 0 010 0000 0000 ... 0000 b

s e = 130 f = 2-2 = 0.25
Case 3
v = (-10)*2(130-1 27)*1.25 = 10.0
Advantage Exponent gives large dynamic range Disadvantage Precision of a number depends on its exponent
Number Line Insight

Floating-Point:
Non-uniform distribution
Precision greatest near zero Less precision the further you get from zero
8 - 13
Floating-Point
Floating-Point Pros and Cons

Advantages
Easy to write code No scaling required
Disadvantages
Somewhat higher device cost May offer insufficient precision for some calculations due to 23 bit mantissa and the influence of the exponent
What if you dont have the luxury of using a floating-point C28x device?
8 - 14
IQmath
IQmath
Implementing complex digital control algorithms on a Digital Signal Processor (DSP), or any other DSP capable processor, typically come across the following issues: Algorithms are typically developed using floating-point math Floating-point devices are more expensive than fixed-point devices Converting floating-point algorithms to a fixed-point device is very time consuming Conversion process is one way and therefore backward simulation is not always possible
The design may initially start with a simulation (i.e. MatLab) of a control algorithm, which typically would be written in floating-point math (C or C++). This algorithm can be easily ported to a floating-point device, however because of cost reasons most likely a 16-bit or 32-bit fixedpoint device would be used in many target systems. The effort and skill involved in converting a floating-point algorithm to function using a 16-bit or 32-bit fixed-point device is quite significant. A great deal of time (many days or weeks) would be needed for reformatting, scaling and coding the problem. Additionally, the final implementation typically has little resemblance to the original algorithm. Debugging is not an easy task and the code is not easy to maintain or document.
IQ Fractional Representation
A new approach to fixed-point algorithm development, termed IQmath, can greatly simplify the design development task. This approach can also be termed virtual floating-point since it looks like floating-point, but it is implemented using fixed-point techniques.
IQ Fractional Representation
31 0
S IIIIIIII fffffffffffffffffffffff
32 bit mantissa
-2I + 2I-1 + + 21 + 20 2-1 + 2-2 + + 2-Q

I8Q24 Example: 0x41200000 = 0100 0001 . 0010 0000 0000 0000 0000 0000 b = 26 + 20 + 2 -3 = 65.125 Advantage Precision same for all numbers in an IQ format Disadvantage Limited dynamic range compared to floating-point
8 - 15
IQmath
The IQmath approach enables the seamless portability of code between fixed and floating-point devices. This approach is applicable to many problems that do not require a large dynamic range, such as motor or digital control applications.
Number Line Insight

Distributions
Floating-Point: non-uniform distribution (variable precision)
IQ Fractions: uniform distribution (same precision everywhere)
Both floating-point and IQ formats have 232 possible values on the number line Its how each distributes these values that differs
Traditional Q Math Approach
Traditional 32-bit Q Math Approach

y = mx + b
I8 I16 Q48 I8 ssssssssssssssssssI8 Q24 I8 Q24 Q24 Q24
M X B
<< 24
ssssI8 Q48
Align Decimal Point for Add
I16
Q48
>> 24
sssssssssssssssssI16
Align Decimal Point for Store

Q24 I8 Q24
in C:
Y = ((int64) M * (int64) X + (int64) B << Q) >> Q;

Note: Requires support for 64-bit integer data type in compiler
8 - 16
IQmath
The traditional approach to performing math operations, using fixed-point numerical techniques can be demonstrated using a simple linear equation example. The floating-point code for a linear equation would be:
float Y, M, X, B; Y = M * X + B;
For the fixed-point implementation, assume all data is 32-bits, and that the "Q" value, or location of the binary point, is set to 24 fractional bits (Q24). The numerical range and resolution for a 32-bit Q24 number is as follows: Q value Q24
(32-24)
-2
Min Value = -128.000 000 00
(32-24)
Max Value ()24 = 127.999 999 94
Resolution () = 0.000 000 06

24
The C code implementation of the linear equation is:

int32 Y, M, X, B; // numbers are all Q24 Y = ((int64) M * (int64) X + (int64) B << 24) >> 24;
Compared to the floating-point representation, it looks quite cumbersome and has little resemblance to the floating-point equation. It is obvious why programmers prefer using floating-point math. The slide shows the implementation of the equation on a processor containing hardware that can perform a 32x32 bit multiplication, 64-bit addition and 64-bit shifts (logical and arithmetic) efficiently. The basic approach in traditional fixed-point "Q" math is to align the binary point of the operands that get added to or subtracted from the multiplication result. As shown in the slide, the multiplication of M and X (two Q24 numbers) results in a Q48 value that is stored in a 64-bit register. The value B (Q24) needs to be scaled to a Q48 number before addition to the M*X value (low order bits zero filled, high order bits sign extended). The final result is then scaled back to a Q24 number (arithmetic shift right) before storing into Y (Q24). Many programmers may be familiar with 16-bit fixed-point "Q" math that is in common use. The same example using 16-bit numbers with 15 fractional bits (Q15) would be coded as follows:
int16 Y, M, X, B; // numbers are all Q15 Y = ((int32) M * (int32) X + (int32) B << 15) >> 15;
In both cases, the principal methodology is the same. The binary point of the operands that get added to or subtracted from the multiplication result must be aligned.
8 - 17
IQmath
IQmath Approach
32-bit IQmath Approach

y = mx + b
I8 I16 Q48 I8 Q24 Q24
M X
Align Decimal Point Of Multiply

sssssssssssssssssI16
>> 24
Q24
I8
Q24
I8
Q24
I8
Q24
in C:
Y = ((int64) M * (int64) X) >> Q + B;
In the "IQmath" approach, rather then scaling the operands, which get added to or subtracted from the multiplication result, we do the reverse. The multiplication result binary point is scaled back such that it aligns to the operands, which are added to or subtracted from it. The C code implementation of this is given by linear equation below:
int32 Y, M, X, B; Y = ((int64) M * (int64) X) >> 24 + B;
The slide shows the implementation of the equation on a processor containing hardware that can perform a 32x32 bit multiply, 32-bit addition/subtraction and 64-bit logical and arithmetic shifts efficiently. The key advantage of this approach is shown by what can then be done with the C and C++ compiler to simplify the coding of the linear equation example. Lets take an additional step and create a multiply function in C that performs the following operation:
int32 _IQ24mpy(int32 M, int32 X) { return ((int64) M * (int64) X) >> 24; }
The linear equation can then be written as follows:

Y = _IQ24mpy(M , X) + B;
Already we can see a marked improvement in the readability of the linear equation.
8 - 18
IQmath
Using the operator overloading features of C++, we can overload the multiplication operand "*" such that when a particular data type is encountered, it will automatically implement the scaled multiply operation. Lets define a data type called "iq" and assign the linear variables to this data type:
iq Y, M, X, B // numbers are all Q24
The overloading of the multiply operand in C++ can be defined as follows:

iq operator*(const iq &M, const iq &X){return((int64)M*(int64) X) >> 24;}
Then the linear equation, in C++, becomes:

Y = M * X + B;
This final equation looks identical to the floating-point representation. It looks "natural". The four approaches are summarized in the table below: Math Implementations 32-bit floating-point math in C 32-bit fixed-point "Q" math in C 32-bit IQmath in C 32-bit IQmath in C++ Linear Equation Code Y = M * X + B; Y = ((int64) M * (int64) X) + (int64) B << 24) >> 24; Y = _IQ24mpy(M, X) + B; Y = M * X + B;
Essentially, the mathematical approach of scaling the multiplier operand enables a cleaner and a more "natural" approach to coding fixed-point problems. For want of a better term, we call this approach "IQmath" or can also be described as "virtual floating-point".
8 - 19
IQmath
IQmath Approach
Multiply Operation
Y = ((i64) M * (i64) X) >> Q + B;
Redefine the multiply operation as follows:

_IQmpy(M,X) == ((i64) M * (i64) X) >> Q
This simplifies the equation as follows:

Y = _IQmpy(M,X) + B;
C28x compiler supports _IQmpy intrinsic; assembly code generated:

MOVL IMPYL QMPYL LSL64 XT,@M P,XT,@X ; P = low 32-bits of M*X ACC,XT,@X ; ACC = high 32-bits of M*X ACC:P,#(32-Q) ; ACC = ACC:P << 32-Q ; (same as P = ACC:P >> Q) ADDL ACC,@B ; Add B MOVL @Y,ACC ; Result = Y = _IQmpy(M*X) + B ; 7 Cycles
IQmath Approach
It looks like floating-point!
Y, M, X, B;
Floating-Point
float
Y = M * X + B;
Traditional Fix-Point Q IQmath In C IQmath In C++
long Y, M, X, B; Y = ((i64) M * (i64) X + (i64) B << Q)) >> Q; _iq Y, M, X, B;
Y = _IQmpy(M, X) + B; iq Y, M, X, B;
Y = M * X + B;
IQmath code is easy to read!
8 - 20
IQmath
IQmath Approach
GLOBAL_Q simplification
User selects Global Q value for the whole application
GLOBAL_Q
based on the required dynamic range or resolution, for example:
GLOBAL_Q 28 24 20 #define _iq Max 7.999 127.999 2047.999 Val Min Val Resolution 999 996 -8.000 000 000 0.000 000 004 999 94 -128.000 000 00 0.000 000 06 999 -2048.000 000 0.000 001 18 // set in IQmathLib.h file
GLOBAL_Q
Y, M, X, B; // all values are in Q = 18
Y = _IQmpy(M,X) + B;
The user can also explicitly specify the Q value to use:

_iq20 Y, M, X, B; // all values are in Q = 20 Y = _IQ20mpy(M,X) + B;
The basic "IQmath" approach was adopted in the creation of a standard math library for the Texas Instruments TMS320C28x DSP fixed-point processor. This processor contains efficient hardware for performing 32x32 bit multiply, 64-bit shifts (logical and arithmetic) and 32-bit add/subtract operations, which are ideally suited for 32 bit "IQmath". Some enhancements were made to the basic "IQmath" approach to improve flexibility. They are: Setting of GLOBAL_Q Parameter Value: Depending on the application, the amount of numerical resolution or dynamic range required may vary. In the linear equation example, we used a Q value of 24 (Q24). There is no reason why any value of Q can't be used. In the "IQmath" library, the user can set a GLOBAL_Q parameter, with a range of 1 to 30 (Q1 to Q30). All functions used in the program will use this GLOBAL_Q value. For example:
#define GLOBAL_Q 18 Y = _IQmpy(M, X) + B; // all values use GLOBAL_Q = 18
If, for some reason a particular function or equation requires a different resolution, then the user has the option to implicitly specify the Q value for the operation. For example:
Y = _IQ23mpy(M,X) + B; // all values use Q23, including B and Y
The Q value must be consistent for all expressions in the same line of code.
8 - 21
IQmath
IQmath Provides Compatibility Between Floating-Point and Fixed-Point

1) Develop any mathematical function
Y = _IQmpy(M, X) + B;
2) Select math type in IQmathLib.h #if MATH_TYPE == IQ_MATH #if MATH_TYPE == FLOAT_MATH
3) Compiler automatically converts to: Y = (float)M * (float)X + (float)B; Fixed-Point Math Code Floating-Point Math Code
Compile & Run on Fixed-Point F282xx
Compile & Run on Floating-Point F283xx *
All IQmath operations have an equivalent floating-point operation

* Can also compile floating-point code on any floating-point compiler (e.g., PC, Matlab, fixed-point w/ RTS lib, etc.)
Selecting FLOAT_MATH or IQ_MATH Mode: As was highlighted in the introduction, we would ideally like to be able to have a single source code that can execute on a floating-point or fixedpoint target device simply by recompiling the code. The "IQmath" library supports this by setting a mode, which selects either IQ_MATH or FLOAT_MATH. This operation is performed by simply redefining the function in a header file. For example:
#if MATH_TYPE == IQ_MATH #define _IQmpy(M , X) _IQmpy(M , X) #elseif MATH_TYPE == FLOAT_MATH #define _IQmpy(M , X) (float) M * (float) X #endif
Essentially, the programmer writes the code using the "IQmath" library functions and the code can be compiled for floating-point or "IQmath" operations.
8 - 22
IQmath Library
IQmath Library
IQmath Library: Math & Trig Functions

Operation
type constant multiply divide add substract boolean trig and power functions
Floating-Point
float A, B; A = 1.2345 A* B A/ B A+B A-B
>, >=, <, <=, ==, |=, &&, ||
IQmath in C
_iq A, B; A = _IQ(1.2345) _IQmpy(A , B) _IQdiv (A , B) A +B A- B
>, >=, <, <=, ==, |=, &&, ||
IQmath in C++
iq A, B; A = IQ(1.2345) A*B A/ B A+ B AB
>, >=, <, <=, ==, |=, &&, ||
sin(A),cos(A) _IQsin(A), _IQcos(A) IQsin(A),IQcos(A) sin(A*2pi),cos(A*2pi) _IQsinPU(A), _IQcosPU(A) IQsinPU(A),IQcosPU(A) asin(A),acos(A) _IQasin(A),_IQacos(A) IQasin(A),IQacos(A) atan(A),atan2(A,B) _IQatan(A), _IQatan2(A,B) IQatan(A),IQatan2(A,B) atan2(A,B)/2pi _IQatan2PU(A,B) IQatan2PU(A,B) sqrt(A),1/sqrt(A) _IQsqrt(A), _IQisqrt(A) IQsqrt(A),IQisqrt(A) sqrt(A*A + B*B) _IQmag(A,B) IQmag(A,B) exp(A) _IQexp(A) IQexp(A) if(A > Pos) A = Pos if(A < Neg) A = Neg _IQsat(A,Pos,Neg) IQsat(A,Pos,Neg)
saturation
Accuracy of functions/operations approx ~28 to ~31 bits
Additionally, the "IQmath" library contains DSP library modules for filters (FIR & IIR) and Fast Fourier Transforms (FFT & IFFT).
IQmath Library: Conversion Functions

Operation
iq to iqN iqN to iq integer(iq) fraction(iq) iq = iq*long integer(iq*long)
Floating-Point
A A (long) A A (long) A A * (float) B (long) (A * (float) B)
IQmath in C
_IQtoIQN(A) _IQNtoIQ(A) _IQint(A) _IQfrac(A) _IQmpyI32(A,B) _IQmpyI32int(A,B)
IQmath in C++
IQtoIQN(A) IQNtoIQ(A) IQint(A) IQfrac(A) IQmpyI32(A,B) IQmpyI32int(A,B) IQmpyI32frac(A,B) QNtoIQ(A) IQtoQN(A) atoIQ(char) IQtoF(A) IQtoA(A,B,C)
fraction(iq*long) A - (long) (A * (float) B) _IQmpyI32frac(A,B) qN to iq A _QNtoIQ(A) iq to qN A _IQtoQN(A) string to iq atof(char) _atoIQ(char) IQ to float A _IQtoF(A) IQ to ASCII sprintf(A,B,C) _IQtoA(A,B,C)
IQmath.lib IQmathLib.h IQmathCPP.h
> contains library of math functions > C header file > C++ header file
8 - 23
IQmath Library
16 vs. 32 Bits
The "IQmath" approach could also be used on 16-bit numbers and for many problems, this is sufficient resolution. However, in many control cases, the user needs to use many different "Q" values to accommodate the limited resolution of a 16-bit number. With DSP devices like the TMS320C28x processor, which can perform 16-bit and 32-bit math with equal efficiency, the choice becomes more of productivity (time to market). Why bother spending a whole lot of time trying to code using 16-bit numbers when you can simply use 32-bit numbers, pick one value of "Q" that will accommodate all cases and not worry about spending too much time optimizing. Of course there is a concern on data RAM usage if numbers that could be represented in 16 bits all use 32 bits. This is becoming less of an issue in today's processors because of the finer technology used and the amount of RAM that can be cheaply integrated. However, in many cases, this problem can be mitigated by performing intermediate calculations using 32-bit numbers and converting the input from 16 to 32 bits and converting the output back to 16 bits before storing the final results. In many problems, it is the intermediate calculations that require additional accuracy to avoid quantization problems.
8 - 24
Converting ADC Results into IQ Format
Converting ADC Results into IQ Format
Getting the ADC Result into IQ Format

0 000 xx xxx xx xxx xx Do not sign extend 31 15 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 x x x x x x x x x x x x 32-bit long AdcResult. ADCRESULTx
Notice that the 32-bit long is already in IQ12 format

_iq Result; void main(void) { // Convert the ADC result into global IQ format valued between 0.0 and 1.0 Result = _IQ12toIQ( (_iq)AdcResult.ADCRESULT0 ); } // Optional: scale by ADC full-scale range to get 0.0 to 3.3 // (if you prefer to think/scale in terms of voltage) Result = _IQmpy( _iq(3.3), Result); }
As you may recall, the converted values of the ADC are placed in the lower 12 bits of the ADCRESULT0 register. Before these values are filtered using the IQmath library, they need to to be put into the IQ format as a 32-bit long. For uni-polar ADC inputs (i.e., 0 to 3.3 V inputs), a conversion to global IQ format can be achieved with:
IQresult_unipolar = _IQmpy(_IQ(3.3),_IQ12toIQ((_iq) AdcResult.ADCRESULT0));
How can we modify the above to recover bi-polar inputs, for example +-1.65 volts? One could do the following to offset the +1.65V analog biasing applied to the ADC input:
IQresult_bipolar = _IQmpy(_IQ(3.3),_IQ12toIQ((_iq) AdcResult.ADCRESULT0)) - _IQ(1.65);
However, one can see that the largest intermediate value the equation above could reach is 3.3. This means that it cannot be used with an IQ data type of IQ30 (IQ30 range is -2 < x < ~2). Since the IQmath library supports IQ types from IQ1 to IQ30, this could be an issue in some applications. The following clever approach supports IQ types from IQ1 to IQ30:
IQresult_bipolar = _IQmpy(_IQ(1.65),_IQ15toIQ((_iq) ((int16) (AdcResult.ADCRESULT0 ^ 0x8000))));
The largest intermediate value that this equation could reach is 1.65. Therefore, IQ30 is easily supported.
8 - 25
AC Induction Motor Example

One of the more complex motor control algorithms
Sensorless, ACI induction machine direct rotor flux control Goal: motor speed estimation & alpha-axis stator current estimation
The "IQmath" approach is ideally suited for applications where a large numerical dynamic range is not required. Motor control is an example of such an application (audio and communication algorithms are other applications). As an example, the IQmath approach has been applied to the sensor-less direct field control of an AC induction motor. This is probably one of the most challenging motor control problems and as will be shown later, requires numerical accuracy greater then 16-bits in the control calculations. The above slide is a block diagram representation of the key control blocks and their interconnections. Essentially this system implements a "Forward Control" block for controlling the d-q axis motor current using PID controllers and a "Feedback Control" block using back emf's integration with compensated voltage from current model for estimating rotor flux based on current and voltage measurements. The motor speed is simply estimated from rotor flux differentiation and openloop slip computation. The system was initially implemented on a "Simulator Test Bench" which uses a simulation of an "AC Induction Motor Model" in place of a real motor. Once working, the system was then tested using a real motor on an appropriate hardware platform. Each individual block shown in the slide exists as a stand-alone C/C++ module, which can be interconnected to form the complete control system. This modular approach allows reusability and portability of the code. The next few slides show the coding of one particular block, PARK Transform, using floating-point and "IQmath" approaches in C:
8 - 26

Park Transform floating-point C code
#include math.h
#define
TWO_PI
6.28318530717959
void park_calc(PARK *v) { float cos_ang , sin_ang; sin_ang = sin(TWO_PI * v->ang); cos_ang = cos(TWO_PI * v->ang);
v->de = (v->ds * cos_ang) + (v->qs * sin_ang); v->qe = (v->qs * cos_ang) - (v->ds * sin_ang); }
Park Transform - converting to IQmath C code
#include math.h #include #define IQmathLib.h TWO_PI _IQ(6.28318530717959) 6.28318530717959
void park_calc(PARK *v) { float cos_ang , sin_ang; _iq _IQsin(_IQmpy(TWO_PI sin_ang = sin(TWO_PI * v->ang);, v->ang)); _IQcos(_IQmpy(TWO_PI cos_ang = cos(TWO_PI * v->ang);, v->ang)); v->de = _IQmpy(v->ds , cos_ang) + _IQmpy(v->qs , sin_ang); (v->ds * cos_ang) + (v->qs * sin_ang); v->qe = _IQmpy(v->qs , cos_ang) - _IQmpy(v->ds , sin_ang); (v->qs * cos_ang) - (v->ds * sin_ang); }
The complete system was coded using "IQmath". Based on analysis of coefficients in the system, the largest coefficient had a value of 33.3333. This indicated that a minimum dynamic range of 7 bits (+/-64 range) was required. Therefore, this translated to a GLOBAL_Q value of 32-7 = 25 (Q25). Just to be safe, the initial simulation runs were conducted with GLOBAL_Q = 24 (Q24)
8 - 27
value. The plots start from a step change in reference speed from 0.0 to 0.5 and 1024 samples are taken.

GLOBAL_Q = 24, system stable
IQmath: speed
IQmath: current
Floating-Point: speed
Floating-Point: current
The speed eventually settles to the desired reference value and the stator current exhibits a clean and stable oscillation. The block diagram slide shows at which points in the control system the plots are taken from.
Equal Precision in the Computation Region

Floating-Point:
Whats Happening Here?
+ I8Q24 Fractions:
0
Same precision as I8Q24
In the region where these particular computations occur, the precision of single-precision floating-point just happens to equal the precision of the I8Q24 format. So, both produce similar results!
8 - 28

GLOBAL_Q = 27, system unstable
IQmath: speed
IQmath: current

GLOBAL_Q = 16, system unstable
IQmath: speed
IQmath: current
8 - 29
With the ability to select the GLOBAL_Q value for all calculations in the "IQmath", an experiment was conducted to see what maximum and minimum Q value the system could tolerate before it became unstable. The results are tabulated in the slide below:

Q stability range
Q range
Q31 to Q27 Q26 to Q19 Q18 to Q0
Stability Range
Unstable
(not enough dynamic range)
Stable Unstable
(not enough resolution, quantization problems)
The developer must pick the right GLOBAL_Q value!
The above indicates that, the AC induction motor system that we simulated requires a minimum of 7 bits of dynamic range (+/-64) and requires a minimum of 19 bits of numerical resolution (+/0.000002). This confirms our initial analysis that the largest coefficient value being 33.33333 required a minimum dynamic range of 7 bits. As a general guideline, users using IQmath should examine the largest coefficient used in the equations and this would be a good starting point for setting the initial GLOBAL_Q value. Then, through simulation or experimentation, the user can reduce the GLOBAL_Q until the system resolution starts to cause instability or performance degradation. The user then has a maximum and minimum limit and a safe approach is to pick a midpoint. What the above analysis also confirms is that this particular problem does require some calculations to be performed using greater then 16 bit precision. The above example requires a minimum of 7 + 19 = 26 bits of numerical accuracy for some parts of the calculations. Hence, if one was implementing the AC induction motor control algorithm using a 16 bit fixed-point DSP, it would require the implementation of higher precision math for certain portions. This would take more cycles and programming effort. The great benefit of using GLOBAL_Q is that the user does not necessarily need to go into details to assign an individual Q for each variable in a whole system, as is typically done in conventional fixed-point programming. This is time consuming work. By using 32-bit resolution and the "IQmath" approach, the user can easily evaluate the overall resolution and quickly implement a typical digital motor control application without quantization problems.
8 - 30

Performance comparisons
C28x C floating-point std. RTS lib (150 MHz) 401 421 2336 2757 36.8% Benchmark C28x C floating-point fast RTS lib (150 MHz) 401 371 792 1163 15.5% C28x C IQmath v1.4d (150 MHz) 625 403 1011 1414 18.9%
B1: ACI module cycles B2: Feedforward control cycles B3: Feedback control cycles Total control cycles (B2+B3) % of available MHz used (20 kHz control loop)
Notes: C28x compiled on codegen tools v5.0.0, -g (debug enabled), -o3 (max. optimization) fast RTS lib v1.0beta1 IQmath lib v1.4d
Using the profiling capabilities of the respective DSP tools, the table above summarizes the number of cycles and code size of the forward and feedback control blocks. The MIPS used is based on a system sampling frequency of 20 kHz, which is typical of such systems.
8 - 31
IQmath Summary
IQmath Summary
IQmath Approach Summary

IQmath + fixed-point processor with 32-bit capabilities =
Seamless portability of code between fixed and floatingpoint devices
User selects target math type in IQmathLib.h file
#if MATH_TYPE == IQ_MATH #if MATH_TYPE == FLOAT_MATH
One source code set for simulation vs. target device Numerical resolution adjustability based on application requirement
Set in IQmathLib.h file
#define GLOBAL_Q 18
Explicitly specify Q value

_iq20 X, Y, Z;
Numerical accuracy without sacrificing time and cycles Rapid conversion/porting and implementation of algorithms
IQmath library is freeware - available from TI DSP website http://www.ti.com/c2000
The IQmath approach, matched to a fixed-point processor with 32x32 bit capabilities enables the following: Seamless portability of code between fixed and floating-point devices Maintenance and support of one source code set from simulation to target device Adjustability of numerical resolution (Q value) based on application requirement Implementation of systems that may otherwise require floating-point device Rapid conversion/porting and implementation of algorithms
8 - 32
Lab 8: IQmath & Floating-Point FIR Filter

Objective
The objective of this lab is to become familiar with IQmath programming. In the previous lab, ePWM1A was setup to generate a 2 kHz, 25% duty cycle symmetric PWM waveform. The waveform was then sampled with the on-chip analog-to-digital converter. In this lab the sampled waveform will be passed through an FIR filter and displayed using the graphing feature of Code Composer Studio. The filter math type is selected in the IQmathLib.h file.
Lab 8: IQmath FIR Filter

ePWM1
ADCINA0
ADC
RESULT0
FIR Filter
connector wire
data memory
ePWM2
pointer rewind
Display using CCS
...
Procedure
Project File
1. A project named Lab8.pjt has been created for this lab. Open the project by clicking on Project Open and look in C:\C28x\Labs\Lab8. All Build Options have been configured the same as the previous lab. The files used in this lab are: Adc.c Filter.c CodeStartBranch.asm Gpio.c DefaultIsr_8.c Lab_8.cmd DelayUs.asm Main_8.c DSP2803x_GlobalVariableDefs.c PieCtrl_5_6_7_8_9_10.c DSP2803x_Headers_nonBIOS.cmd PieVect_5_6_7_8_9_10.c ECap_7_8_9_10_12.c SysCtrl.c EPwm_7_8_9_10_12.c Watchdog.c
8 - 33

2. Setup the include search path to include the IQmath header file. Open the Build Options and select the Compiler tab. In the Preprocessor Category, find the Include Search Path (-i) box and add to the end of the line (preceeded with a semicolon to append this directory to the existing search path): ;..\IQmath\include 3. Setup the library search path to include the IQmath library. Select the Linker tab. a. In the Libraries Category, find the Search Path (-i) box and enter: ..\IQmath\lib b. In the Include Libraries (-l) box add to the end of the line (preceeded with a semicolon to append this library to the existing library): ;IQmath.lib Then select OK to save the Build Options.
Include IQmathLib.h
4. In the CCS project window left click the plus sign (+) to the left of the Include folder. Edit Lab.h to uncomment the line that includes the IQmathLib.h header file. Next, in the Function Prototypes section, uncomment the function prototype for IQssfir(), the IQ math single-sample FIR filter function. In the Global Variable References section uncomment the two _iq references. Save the changes and close the file.
Inspect Lab_8.cmd
5. Open and inspect Lab_8.cmd. First, notice that a section called IQmath is being linked to L0SARAM. The IQmath section contains the IQmath library functions (code). Second, notice that a section called IQmathTables is being linked to the IQTABLES with a TYPE = NOLOAD modifier after its allocation. The IQmath tables are used by the IQmath library functions. The NOLOAD modifier allows the linker to resolve all addresses in the section, but the section is not actually placed into the .out file. This is done because the section is already present in the device ROM (you cannot load data into ROM after the device is manufactured!). The tables were put in the ROM by TI when the device was manufactured. All we need to do is link the section to the addresses where it is known to already reside (the tables are the very first thing in the BOOT ROM, starting at address 0x3FE000). Close the inspected file.
8 - 34
Select a Global IQ value

6. Use File Open to open c:\C28x\Labs\IQmath\include\IQmathLib.h. Confirm that the GLOBAL_Q type (near beginning of file) is set to a value of 24. If it is not, modify as necessary: #define GLOBAL_Q 24
Recall that this Q type will provide 8 integer bits and 24 fractional bits. Dynamic range is therefore -128 < x < +128, which is sufficient for our purposes in the workshop. Notice that the math type is defined as IQmath by: #define Close the file. MATH_TYPE IQ_MATH
IQmath Single-Sample FIR Filter

7. Open and inspect DefaultIsr_8.c. Notice that the ADCINT_ISR calls the IQmath single-sample FIR filter function, IQssfir(). The filter coefficients have been defined in the beginning of Main_8.c. 8. Open and inspect the IQssfir() function in Filter.c. This is a simple, non-optimized coding of a basic IQmath single-sample FIR filter. Close the inspected files.
Build and Load

9. Click the Build button to build and load the project.
Run the Code Filtered Waveform

10. Open a memory window to view some of the contents of the filtered ADC results buffer. The address label for the filtered ADC results buffer is AdcBufFiltered. Set the Format to 16-Bit Unsigned Integer. We will be running our code in real-time mode, and will have our window continuously refresh. Note: For the next step, check to be sure that the jumper wire connecting PWM1A (pin # GPIO-00) to ADCINA0 (pin # ADC-A0) is in place on the Docking Station.
11. Run the code in real-time mode using the GEL function: GEL Realtime Emulation Control Run_Realtime_with_Reset, and watch the memory window update. Verify that the ADC result buffer contains updated values. 12. Open and setup a dual-time graph to plot a 50-point window of the filtered and unfiltered ADC results buffer. Click: View Graph Time/Frequency and set the following values:
8 - 35
Display Type
Dual Time
Start Address upper display AdcBufFiltered Start Address lower display AdcBuf Acquisition Buffer Size Display Data Size DSP Data Type Sampling Rate (Hz) Time Display Unit Select OK to save the graph options. 13. The graphical display should show the generated FIR filtered 2 kHz, 25% duty cycle symmetric PWM waveform in the upper display and the unfiltered waveform generated in the previous lab exercise in the lower display. Notice the shape and phase differences between the waveform plots (the filtered curve has rounded edges, and lags the unfiltered plot by several samples). The amplitudes of both plots should run from 0 to 4095. 14. Open and setup two (2) frequency domain plots one for the filtered and another for the unfiltered ADC results buffer. Click: View Graph Time/Frequency and set the following values: GRAPH #1 Display Type Start Address Acquisition Buffer Size FFT Framesize DSP Data Type Sampling Rate (Hz) FFT Magnitude AdcBufFiltered 50 50 16-bit unsigned integer 50000 GRAPH #2 FFT Magnitude AdcBuf 50 50 16-bit unsigned integer 50000 50 50 16-bit unsigned integer 50000 s
Select OK to save the graph options. 15. The graphical displays should show the frequency components of the filtered and unfiltered 2 kHz, 25% duty cycle symmetric PWM waveforms. Notice that the higher frequency components are reduced using the Low-Pass FIR filter in the filtered graph as compared to the unfiltered graph.
8 - 36
16. Fully halt the CPU (real-time mode) by using the GEL function: GEL Emulation Control Full_Halt. End of Exercise
Realtime
8 - 37
Lab 8 Reference: Low-Pass FIR Filter
Bode Plot of Digital Low Pass Filter Coefficients: [1/16, 4/16, 6/16, 4/16, 1/16] Sample Rate: 50 kHz
8 - 38
Control Law Accelerator

Introduction
This module explains the operation of the control law accelerator (CLA). The CLA is an independent, fully programmable, 32-bit floating-point math processor that enables concurrent execution into the C28x family. This extends the capabilities of the C28x CPU by adding parallel processing. The CLA has direct access to the ADC result registers, and all ePWM, HRPWM and comparator registers. This allows the CLA to read ADC samples just-in-time and significantly reduces the ADC sample to output delay enabling faster system response and higher frequency operation. Utilizing the CLA for time-critical tasks frees up the CPU to perform other system and communication functions concurrently.
Learning Objectives
Learning Objectives
Explain the purpose and operation of the Control Law Accelerator (CLA) Describe the CLA initialization procedure Review the CLA registers, instruction set, and programming flow
C2000 Piccolo Workshop - Control Law Accelerator
9-1
Module Topics
Module Topics
Control Law Accelerator .......................................................................................................................... 9-1 Module Topics......................................................................................................................................... 9-2 Control Law Accelerator (CLA) ............................................................................................................. 9-3 CLA Block Diagram........................................................................................................................... 9-3 CLA Memory and Register Access .................................................................................................... 9-4 CLA Tasks.......................................................................................................................................... 9-4 Control and Execution Registers ........................................................................................................ 9-5 CLA Registers .................................................................................................................................... 9-6 CLA Initialization............................................................................................................................... 9-8 CLA Task Programming .................................................................................................................... 9-9 CLA Instruction Set...........................................................................................................................9-10 CLA Addressing Modes ....................................................................................................................9-11 CLA Code Example...........................................................................................................................9-11 CLA Code Debugging .......................................................................................................................9-12 Lab 9: CLA Floating-Point FIR Filter...................................................................................................9-13
9-2
Control Law Accelerator (CLA)

ADC & CMP CLA C28x CPU PWM
CLA is an independent 32-bit floatingpoint math accelerator Executes algorithms independently and in parallel with the main CPU Direct access to ePWM / HRPWM, ADC result and comparator registers Responds to peripheral interrupts independently of CPU Frees-up CPU for other tasks (communications and diagnostics)
CLA Block Diagram
CLA Block Diagram

Task Triggers
(Peripheral Interrupts)
ADCINT1 or EPWM1_INT ADCINT7 or EPWM7_INT ADCINT8 or CPU Timer 0

CLA Program Bus MPERINT1-8
CLA
Control & Execution Registers
CLA_INT1-8 LVF, LUF
PIE
INT11 C28x INT12 CPU
CLA Data Bus
Prog RAM
Data RAM0
Data RAM1
MSG RAMs CPU to CLA CLA to CPU
Periph. Regs ADC Results ePWM HRPWM Comparator
9-3
CLA Memory and Register Access
CLA Memory and Register Access

CLA Program Memory
Contains CLA program code Mapped to the CPU at reset Initialized by the CPU
Message RAMs
Used to pass data between the CPU and CLA Always mapped to both the CPU and CLA
L3 DPSARAM
L1 DPSARAM
L2 DPSARAM
PF0
PF0 & PF1
Prog RAM
Data RAM0
Data RAM1
MSG RAMs CPU to CLA CLA to CPU
Periph. Regs ADC Results ePWM HRPWM Comparator
CLA Data Memory

Contains variables and coefficients used by the CLA program code Mapped to the CPU at reset Initialized by CPU
Peripheral Reg Access

ADC Results Regs ePWM (all regs) HRPWM (all regs) Comparator (all regs)
CLA Tasks
CLA Tasks
Task Triggers
(Peripheral Interrupts)
ADCINT1 or EPWM1_INT ADCINT7 or EPWM7_INT ADCINT8 or CPU Timer 0

MPERINT1-8
CLA
Control & Execution Registers
CLA_INT1-8 LVF, LUF
PIE
A Task is similar to an interrupt service routine CLA supports 8 Tasks (Task1-8) A task is started by a peripheral interrupt trigger
Triggers are enabled in the MPISRCSEL1 register
When a trigger occurs the CLA begins execution at the associated task vector entry (MVECT1-8) Once a task begins it runs to completion (no nesting)
A task is terminated with an MSTOP instruction
9-4
Software triggering a task

Tasks can also be started by a software trigger using the CPU
Method #1: Write to Interrupt Force Register (MIFRC) register
15 - 8 7 6 5 4 3 2 1 0
reserved asm(" EALLOW");
INT8
INT7
INT6
INT5
INT4
INT3
INT2
INT1
// enable protected register access
Cla1Regs.MIFRC.bit.INT4 = 1; // start task 4 asm(" EDIS"); // disable protected register access
Method #2: Use IACK instruction

asm(" IACK #0x0008"); // set bit 4 in MIFR to start task 4
More efficient does not require EALLOW

Note: Use of IACK requires Cla1Regs.MCTL.bit.IACKE = 1
Control and Execution Registers
CLA Control and Execution Registers

MIFR
ADCINT1 ePWM1_INT S/W trigger 1
MIER
CLA_INT1-8 LVF, LUF
CLA Core
0
PIE
ADCINT8 CPU Timer 0 S/W trigger
MR0 MPISRCSEL1 MPC MVECT1-8 MR1 MR2 MR3

CLA Data Bus
MAR0 MAR1 Data Memory
Program Memory
CLA Program Bus
MMEMCFG MPISRCSEL1 Peripheral Interrupt Source Select (Task 1-8) MVECT1-8 Task Interrupt Vector (MVECT1/2/3/4/5/6/7/8) MMEMCFG Memory Map Configuration (RAM1E, RAM0E, PROGE) MPC 12-bit Program Counter (initialized by appropriate MVECTx register) MR0-3 CLA Floating-Point 32-bit Result Registers MAR0-1 CLA Auxiliary Registers
9-5
CLA Registers
Cla1Regs.register (lab file: Cla.c)
CLA Registers
Description
Register MCTL MMEMCFG MPISRCSEL1 MIFR MIER MIFRC MICLR MIOVF MICLROVF MIRUN MVECTx MPC MARx MRx MSTF
Control Register Memory Configuration Register Peripheral Interrupt Source Select 1 Register Interrupt Flag Register Interrupt Enable Register Interrupt Force Register Interrupt Flag Clear Register Interrupt Overflow Flag Register Interrupt Overflow Flag Clear Register Interrupt Run Status Register Task x Interrupt Vector (x = 1-8) CLA 12-bit Program Counter CLA Auxiliary Register x (x = 0-1) CLA Floating-Point 32-bit Result Register (x = 0-3) CLA Floating-Point Status Register
CLA Control Register

Cla1Regs.MCTL
IACK Enable 0 = CPU IACK instruction ignored 1 = CPU IACK instruction triggers a task Hard Reset 0 = no effect 1 = CLA reset (registers set to default state)
1 0
15 - 3 reserved
IACKE SOFTRESET HARDRESET
Soft Reset 0 = no effect 1 = CLA reset (stop current task)
9-6
CLA Memory Configuration Register

Cla1Regs.MMEMCFG
CLA Program Space Enable 0 = mapped to CPU program and data space 1 = mapped to CLA program space
15 - 6 reserved
5 RAM1E
4 RAM0E
2- 1 reserved
0 PROGE
CLA Data RAM1 / RAM0 Enable 0 = mapped to CPU program and data space 1 = mapped to CLA data space
CLA Peripheral Interrupt Source Select 1 Register

Cla1Regs.MPISRCSEL1
Task 8 Peripheral Interrupt Input 000 = ADCINT8 010 = CPU Timer 0 xx1 = no source
31 - 28 PERINT8SEL 15 - 12 PERINT4SEL
Task 7 Peripheral Interrupt Input 000 = ADCINT7 010 = ePWM7 xx1 = no source
27 - 24 PERINT7SEL 11 - 8 PERINT3SEL
23 - 20 PERINT6SEL 7-4 PERINT2SEL
19 - 16 PERINT5SEL 3-0 PERINT1SEL
000 = Default
Note: select xx1 (no source) if task is generated by software
9-7
CLA Interrupt Enable Register

Cla1Regs.MIER
15 - 8 reserved
7 INT8
6 INT7
5 INT6
4 INT5
3 INT4
2 INT3
1 INT2
0 INT1
0 = task interrupt disable (default) 1 = task interrupt enable
#include DSP2803x_Device.h Cla1Regs.MIER.bit.INT2 = 1; Cla1Regs.MIER.all = 0x0028; //enable Task 2 interrupt //enable Task 6 and 4 interrupts
CLA Initialization
CLA Initialization
CLA initialization is performed by the CPU in C code (typically done with the Peripheral Register Header Files) 1. 2. 3. Copy CLA task code from flash to CLA program RAM Initialize CLA data RAMs, as needed
Populate with data coefficients, constants, etc.
Configure the CLA registers

Enable the CLA clock (PCLKCR3 register) Populate the CLA task interrupt vectors (MVECT1-8 registers) Select the desired task interrupt sources (PERINT1SEL register) If desired, enable IACK to start task using software (avoids EALLOW) Map CLA program RAM and data RAMs to CLA space
4. 5. 6.
Configure desired CLA task completion interrupts in the PIE Enable CLA tasks triggers in the MIER register Initialize the ePWM and/or ADC to trigger the CLA tasks
Data is passed between the CLA and CPU via message RAMs
9-8
Enabling CLA Support in CCS
Note: You must be using a C28x Piccolo device that has the Control Law Accelerator! In the project build options, select: cla0 (From Device Type 0) This is required in order to assemble CLA code CLA support requires codegen tools v5.2.0 or later
CLA Task Programming
CLA Task Programming

CLA tasks are written in assembly code Same instruction format as the C28x and C28x+FPU
Destination operand is always on the left Same mnemonics as C28x+FPU but with a leading M
CPU: FPU: CLA: MPY MPYF32 MMPYF32 ACC, T, loc16
R0H, R1H, R2H MR0, MR1, MR2

Source Operands
Destination
9-9
CLA Instruction Set
CLA Instruction Overview

Type Load (Conditional) Store Load with Data Move Store/Load MSTF Compare, Min, Max Absolute, Negative Value Unsigned Integer to Float Integer to Float Float to Integer & Round Float to Integer Multiply, Add, Subtract 1/X (16-bit Accurate) 1/Sqrt(x) (16-bit Accurate) Integer Load/Store Load/Store Auxiliary Register Branch/Call/Return Conditional Delayed Integer Bitwise AND, OR, XOR Integer Add and Subtract Integer Shifts Write Protection Enable/Disable Halt Code or End Task No Operation MMOV32 MMOV32 MMOVD32 MMOV32 MCMPF32 MABSF32 MUI16TOF32 MI32TOF32 MF32TOI16R MF32TOI32 MMPYF32 MEINVF32 MEISQRTF32 MMOV16 MMOV16 MBCNDD MAND32 MSUB32 MLSR32 MEALLOW MSTOP MNOP Example MRa,mem32{,CONDF} mem32,MRa MRa,mem32 MSTF,mem32 MRa,MRb MRa,MRb MRa,mem16 MRa,mem32 MRa,MRb MRa,MRb MRa,MRb,MRc MRa,MRb MRa,MRb MRa,mem16 MAR,mem16 16bitdest {,CNDF} MRa,MRb,MRc MRa,MRb,MRc MRa,#SHIFT Cycles 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1-7 1 1 1 1 1 1
CLA Parallel Instructions

Parallel bars indicate a parallel instruction Parallel instructions operate as a single instruction with a single opcode and performs two operations
Example: Add + Parallel Store MADDF32 MR3, MR3, MR1 MMOV32 @_Var, MR3
||
Instruction Multiply & Parallel Add/Subtract Multiply, Add, Subtract & Parallel Store Multiply, Add, Subtract, MAC & Parallel Load
Example MMPYF32 MRa,MRb,MRc || MSUBF32 MRd,MRe,MRf MADDF32 MRa,MRb,MRc || MMOV32 mem32,MRe MADDF32 MRa,MRb,MRc || MMOV32 MRe, mem32
Cycles 1 1 1
Both operations complete in a single cycle
9 - 10
CLA Addressing Modes
CLA Addressing Modes

CLA has two addressing modes
Both modes can access the low 64Kw of memory:
All of the CLA data space Both message RAMs Shared peripheral registers
There is no stack pointer or data page pointer
Direct Addressing Mode:

Populates opcode field with 16-bit address of the variable Example 1: Example 2: MMOV32 MMOV32 MR1, @_VarA MR1, @_EPwm1Regs.CMPA.all
Indirect Addressing with 16-bit Post Increment:

Uses the address in MAR0 or MAR1 to access memory After the read or write MAR0/MAR1 is incremented by #Imm16 Example 1: Example 2: MMOV32 MMOV32 MR0, *MAR0[2]++ MR1, *MAR1[-2]++
CLA Code Example
CLA Code Example (1 of 2)

ClaTasks.asm .cdecls "Lab.h" .sect "Cla1Prog" _Cla1Prog_Start _Cla1Task1: ; FIR filter
.cdecls directive used to include the C header file in the CLA assembly file .sect directive used to place CLA assembly code in its own section C Peripheral Register Header File references can be used in CLA assembly code MSTOP instruction used at the end of the task CLA assembly and C28 C-code reside in the same project
MUI16TOF32 MR2, @_AdcResult.ADCRESULT0 MMPYF32 MR2, MR1, MR0 MADDF32 MR3, MR3, MR2
MF32TOUI16 MR2, MR3 MMOV16 @_ClaFilteredOutput, MR2 MSTOP _Cla1Task2: MSTOP ;------------------------------------_Cla1Task3: MSTOP ; End of task
;-------------------------------------
9 - 11
CLA Code Example (2 of 2)

Lab.h #include "DSP2803x_Device.h" extern Uint32 Cla1Prog_Start; extern Uint32 Cla1Task1; extern Uint32 Cla1Task2; extern Uint32 Cla1Task8;
DSP2803x_Device.h defines register bit field structures Symbols in header file that are defined in the CLA assembly file are made global (by the .cdecls in Cla.asm) and are usable in C
Cla.c #include "Lab.h" // Symbols used to calculate vector address Cla1Regs.MVECT1 = (Uint16)((Uint32)&Cla1Task1 (Uint32)&Cla1Prog_Start); Cla1Regs.MVECT2 = (Uint16)((Uint32)&Cla1Task2 (Uint32)&Cla1Prog_Start);
CLA Code Debugging
CLA Code Debugging

The CLA can halt, single-step and run independently from the CPU Both the CLA and CPU are debugged from the same JTAG port
1. 2. 3.
Insert a breakpoint in CLA code

Insert MDEBUGSTOP instruction to halt CLA and then rebuild/reload
Enable CLA breakpoints

Enable CLA breakpoints in the debugger
Start the task

Done by peripheral interrupt, software (IACK) or MIFRC register CLA executes instructions until MDEBUGSTOP MPC will the have address of MDEBUGSTOP instruction
4.
Single step the CLA code

Once halted, single step the CLA code Can also run to the next MDEBUGSTOP or to the end of task If another task is pending it will start at end of previous task
5.
Disable CLA breakpoints, if desired

CLA single step CLA pipeline is clocked only one cycle and then frozen
CPU single step CPU pipeline is flushed for each single step
9 - 12
Lab 9: CLA Floating-Point FIR Filter

Objective
The objective of this lab is to become familiar with operation of the CLA. In the previous lab, the CPU was used to filter the ePWM1A generated 2 kHz, 25% duty cycle symmetric PWM waveform. In this lab, the PWM waveform will be filtered using the CLA. The CLA will directly read the ADC result register and a task will run a low-pass FIR filter on the sampled waveform. The filtered result will be stored in a circular memory buffer. Note that the CLA is operating concurrently with the CPU. As an operational test, the filtered and unfiltered waveforms will be displayed using the graphing feature of Code Composer Studio.

ePWM1
ADCINA0
ADC
RESULT0
CLA
_Cla1Task1 _Cla1Task2 _Cla1Task8
connector wire
data memory
ePWM2
pointer rewind
Display using CCS
...
Procedure
Project File
1. A project named Lab9.pjt has been created for this lab. Open the project by clicking Open and look in C:\C28x\Labs\Lab9. All Build Options on Project have been configured the same as the previous lab. The files used in this lab are: Adc.c EPwm_7_8_9_10_12.c Cla_9.c Filter.c ClaTasks.asm Gpio.c CodeStartBranch.asm Lab_9.cmd DefaultIsr_9_10.c Main_9.c DelayUs.asm PieCtrl_5_6_7_8_9_10.c DSP2803x_GlobalVariableDefs.c PieVect_5_6_7_8_9_10.c DSP2803x_Headers_nonBIOS.cmd SysCtrl.c ECap_7_8_9_10_12.c Watchdog.c
9 - 13
Enabling CLA Support in CCS

2. Open the Build Options and select the Compiler tab. In the Basic Category set the Specify CLA Support to cla0 (From Device Type 0). This is needed to assemble CLA code. Then select OK to save the Build Options.
Inspect Lab_9.cmd
3. Open and inspect Lab_9.cmd. Notice that a section called Cla1Prog is being linked to L3DPSARAM. This section links the CLA program tasks (assembly code) to the CPU memory space. This memory space will be remapped to the CLA memory space during initialization. Also, notice the two message RAM sections used to pass data between the CPU and CLA.
Setup CLA Initialization

During the CLA initialization, the CPU memory block L3DPSARAM needs to be configured as CLA program memory. This memory space contains the CLA Task routines, which are coded in assembly. The CLA Task 1 has been configured to run an FIR filter. The CLA needs to be configured to start Task 1 on the ADCINT1 interrupt trigger. The next section will setup the PIE interrupt for the CLA. 4. Open ClaTasks.asm and notice that the .cdecls directive is being used to include the C header file in the CLA assembly file. Therefore, we can use the Peripheral Register Header File references in the CLA assembly code. Next, notice Task 1 has been configured to run an FIR filter. Within this code special instructions have been used to convert the ADC result integer (i.e. the filter input) to floating-point and the floatingpoint filter output back to integer. 5. Edit Cla_9.c to implement the CLA operation as described in the objective for this lab exercise. Configure the L3DPSARM memory block to be mapped to CLA program memory space. Set Task 1 peripheral interrupt source to ADCINT1 and set the other Task peripheral interrupt source inputs to no source. Enable CLA Task 1 interrupt. 6. Open Main_9.c and add a line of code in main() to call the InitCla() function. There are no passed parameters or return values. You just type InitCla(); at the desired spot in main().
Setup PIE Interrupt for CLA

Recall that ePWM2 is triggering the ADC at a 50 kHz rate. In the previous lab exercise, the ADC generated an interrupt to the CPU, and the CPU implemented the FIR filter in the ADC ISR. For this lab exercise, the ADC is instead triggering the CLA, and the CLA will directly read the ADC result register and run a task implementing an FIR filter. The CLA will generate an interrupt to the CPU, which will store the filtered results to a circular buffer implemented in the CLA ISR.
9 - 14
7. Edit Adc.c to comment out the code used to enable ADCINT1 interrupt in PIE group 1. This is no longer being used. The CLA interrupt will be used instead. 8. Using the PIE Interrupt Assignment Table find the location for the CLA Task 1 interrupt CLA1_INT1 and fill in the following information: PIE group #: # within group:
This information will be used in the next step. 9. Modify the end of Cla_9.c to do the following: - Enable the "CLA1_INT1" interrupt in the PIE (Hint: use the PieCtrlRegs structure) - Enable the appropriate core interrupt in the IER register 10. Open and inspect DefaultIsr_9_10.c. Notice that this file contains the CLA interrupt service routine. Save and close all modified files.
Build and Load

11. Click the Build button to build and load the project.
Run the Code Test the CLA Operation

Note: For the next step, check to be sure that the jumper wire connecting PWM1A (pin # GPIO-00) to ADCINA0 (pin # ADC-A0) is in place on the Docking Station.
12. Run the code in real-time mode using the GEL function: GEL Realtime Emulation Control Run_Realtime_with_Reset, and watch the memory window update. Verify that the ADC result buffer contains updated values. 13. Setup a dual-time graph of the filtered and unfiltered ADC results buffer. Click: View Graph Time/Frequency and set the following values:
9 - 15
Display Type
Dual Time
Start Address upper display AdcBufFiltered Start Address lower display AdcBuf Acquisition Buffer Size Display Data Size DSP Data Type Sampling Rate (Hz) Time Display Unit 50 50 16-bit unsigned integer 50000 s
14. The graphical display should show the filtered PWM waveform in the upper display and the unfiltered waveform in the lower display. You should see that the results match the previous lab exercise. 15. Fully halt the CPU (real-time mode) by using the GEL function: GEL Emulation Control Full_Halt. End of Exercise Realtime
9 - 16
System Design
Introduction
This module discusses various aspects of system design. Details of the emulation and analysis block along with JTAG will be explored. Flash memory programming and the Code Security Module will be described.
Learning Objectives
Learning Objectives
Emulation and Analysis Block Flash Configuration and Memory Performance Flash Programming Code Security Module (CSM)
C2000 Piccolo Workshop - System Design
10 - 1
Module Topics
Module Topics
System Design ...........................................................................................................................................10-1 Module Topics........................................................................................................................................10-2 Emulation and Analysis Block ...............................................................................................................10-3 Flash Configuration and Memory Performance ....................................................................................10-6 Flash Programming ...............................................................................................................................10-9 Code Security Module (CSM) ..............................................................................................................10-11 Lab 10: Programming the Flash..........................................................................................................10-14
10 - 2
Emulation and Analysis Block
JTAG Emulation System

(based on IEEE 1149.1 Boundary Scan Standard) System Under Test
H E A D E R
SCAN IN TMS320C2000 SCAN OUT
Emulator Pod
Some Available Emulators XDS510 CLASS BlackHawk: USB2000 Signum System: JTAGjet-TMS-C2000 Spectrum Digital: XDS510LC XDS100 CLASS BlackHawk: USB100 Olimex: TMS320-JTAG-USB Spectrum Digital: XDS100 TI: TMDSEMU100U-14T These emulators are C2000 specific, and are much lower cost than emulators that support all TI MCU/DSP platforms (although those can certainly be used) These emulators are much slower than the ones listed above, but are also available at a lower cost than XDS510 class and are NOT C2000 specific
Emulator Connections to the Device

Vcc (3.3 V) GND TMS320F2803x 13 14 TRST TMS TDI TDO TCK 2 1 3 7 11 9 Emulator Header EMU0 EMU1 TRST TMS TDI TDO TCK TCK_RET GND 4 GND 6 GND GND GND 8 10 12 GND PD 5 Vcc (3.3 V)
= If distance between device and header is greater than 6 inches
10 - 3
On-Chip Emulation Analysis Block: Capabilities

Two hardware analysis units can be configured to provide any one of the following advanced debug features: Analysis Configuration
2 Hardware Breakpoints
Debug Activity
Halt on a specified instruction (for debugging in Flash) A memory location is getting corrupted; halt the processor when any value is written to this location Halt program execution after a specific value is written to a variable Halt on a specified instruction only after some other specific routine has executed
2 Address Watchpoints
1 Address Watchpoint with Data
1 Pair Chained Breakpoints
On-Chip Emulation Analysis Block: Hardware Breakpoints

Symbolic or numeric address
Mask value for specifying address ranges
Chained breakpoint selection
10 - 4
On-Chip Emulation Analysis Block: Watchpoints

Symbolic or numeric address
Mask value for specifying address ranges
Bus selection
Address with Data selection
On-Chip Emulation Analysis Block: Online Stack Overflow Detection

Emulation analysis registers are accessible to code as well! Configure a watchpoint to monitor for writes near the end of the stack Watchpoint triggers maskable RTOSINT interrupt Works with DSP/BIOS and non-DSP/BIOS
See TI application report SPRA820 for implementation details
Region of memory occupied by the stack
Stack grows towards higher memory addresses Monitor for data writes in region near the end of the stack Data Memory
10 - 5
Flash Configuration and Memory Performance
Basic Flash Operation

Flash is arranged in pages of 128 words Wait states are specified for consecutive accesses within a page, and random accesses across pages OTP has random access only Must specify the number of SYSCLKOUT wait-states; Reset defaults are maximum value (15) Flash configuration code should not be run from the Flash memory
15 12 11 8 7 4 3 0
FlashRegs.FBANKWAIT
15
reserved
PAGEWAIT
reserved
5 4
RANDWAIT
0
FlashRegs.FOTPWAIT
reserved
OTPWAIT
*** Refer to the F2803x datasheet for detailed numbers *** For 60 MHz, PAGEWAIT = 2, RANDWAIT = 2, OTPWAIT = 3
Speeding Up Code Execution in Flash

Flash Pipelining (for code fetch only)
16 16 or 32 dispatched C28x Core decoder unit 2-level deep fetch buffer Flash Pipeline Enable
0 = disable (default) 1 = enable
64 64 Aligned 64-bit fetch
FlashRegs.FOPT.bit.ENPIPE = 1;
15 1 0
reserved
ENPIPE
10 - 6
Code Execution Performance

Assume 60 MHz SYSCLKOUT, 16-bit instructions
(80% of instructions are 16 bits wide Rest are 32 bits)
Internal RAM: 60 MIPS

Fetch up to 32-bits every cycle 1 instruction/cycle * 60 MHz = 60 MIPS
Flash (w/ pipelining): 60 MIPS

RANDWAIT = 2 Fetch 64 bits every 3 cycles, but it will take 4 cycles to execute them 4 instructions/4 cycles * 60 MHz = 60 MIPS RPT will increase this; PC discontinuity will degrade this Benchmarking in control applications has shown actual performance of about 54 MIPS
Data Access Performance

Assume 60 MHz SYSCLKOUT
Memory 16-bit access (words/cycle) Internal RAM Flash 1 0.33 32-bit access (words/cycle) 1 0.33
RANDWAIT = 2 Flash is read only!
Notes
Internal RAM has best data performance put time critical data here Flash performance usually sufficient for most constants and tables Note that the flash instruction fetch pipeline will also stall during a flash data access
10 - 7
Other Flash Configuration Registers

FlashRegs.name
Address 0x00 0A80 0x00 0A82 0x00 0A83 0x00 0A84 0x00 0A85 0x00 0A86 0x00 0A87 Name FOPT FPWR FSTATUS FSTDBYWAIT FACTIVEWAIT FBANKWAIT FOTPWAIT Description Flash option register Flash power modes registers Flash status register Flash sleep to standby wait register Flash standby to active wait register Flash read access wait state register OTP read access wait state register
FPWR: Save power by putting Flash/OTP to Sleep or Standby mode; Flash will automatically enter active mode if a Flash/OTP access is made FSTATUS: Various status bits (e.g. PWR mode) FSTDBYWAIT, FACTIVEWAIT: Specify # of delay cycles during wake-up from sleep to standby, and from standby to active, respectively. The delay is needed to let the flash stabilize. Leave these registers set to their default maximum value.
See the TMS320x2803x Piccolo System Control and Interrupts Reference Guide, SPRUGL8, for more information
10 - 8
Flash Programming
Flash Programming
Flash Programming Basics

The DSP CPU itself performs the flash programming The CPU executes Flash utility code from RAM that reads the Flash data and writes it into the Flash We need to get the Flash utility code and the Flash data into RAM
FLASH Flash Utility Code CPU
Emulator
JTAG RAM
RS232
SCI
ROM Bootloader
SPI Flash Data I2C CAN GPIO
TMS320F2803x
Flash Programming Basics

Sequence of steps for Flash programming:
Algorithm
1. Erase 2. Program 3. Verify
Function
- Set all bits to zero, then to one - Program selected bits with zero - Verify flash contents
Minimum Erase size is a sector (4Kw or 8Kw) Minimum Program size is a bit! Important not to lose power during erase step: If CSM passwords happen to be all zeros, the CSM will be permanently locked! Chance of this happening is quite small! (Erase step is performed sector by sector)
10 - 9
Flash Programming
Flash Programming Utilities

JTAG Emulator Based
Code Composer Studio Plug-in BlackHawk Flash utilities (requires Blackhawk emulator) Elprotronic FlashPro2000 Spectrum Digital SDFlash JTAG (requires SD emulator) Signum System Flash utilities (requires Signum emulator)
SCI Serial Port Bootloader Based

Code-Skin (http://www.code-skin.com) Elprotronic FlashPro2000
Production Test/Programming Equipment Based

BP Micro programmer Data I/O programmer
Build your own custom utility

Can use any of the ROM bootloader methods Can embed flash programming into your application Flash API algorithms provided by TI
* TI web has links to all utilities (http://www.ti.com/c2000)
Code Composer Studio Flash Plug-In
10 - 10

Access to the following on-chip memory is restricted:
0x000A80 0x008000 0x008800 0x008C00 0x009000 0x00A000 0x3D7800 0x3D7C00 0x3D7C80 0x3D8000 0x3E8000 0x3F7FF8 0x3F8000 0x3F8800 Flash Registers L0 SARAM (2Kw) L1 DPSARAM (1Kw) L2 DPSARAM (1Kw) L3 DPSARAM (4Kw) reserved User OTP (1Kw) reserved ADC / OSC cal. data reserved FLASH (64Kw)
PASSWORDS (8w )
Dual Mapped
L0 SARAM (2Kw)
Data reads and writes from restricted memory are only allowed for code running from restricted memory All other data read/write accesses are blocked:
JTAG emulator/debugger, ROM bootloader, code running in external memory or unrestricted internal memory
CSM Password
0x3E8000
FLASH (64Kw)
0x3F7FF8
128-Bit Password
CSM Password Locations (PWL) 0x3F7FF8 - 0x3F7FFF
128-bit user defined password is stored in Flash 128-bit KEY registers are used to lock and unlock the device
Mapped in memory space 0x00 0AE0 0x00 0AE7 Registers EALLOW protected
10 - 11
CSM Registers
Key Registers accessible by user; EALLOW protected Address Name Description 0x00 0AE0 KEY0 Low word of 128-bit Key register 0x00 0AE1 KEY1 2nd word of 128-bit Key register 0x00 0AE2 KEY2 3rd word of 128-bit Key register 0x00 0AE3 KEY3 4th word of 128-bit Key register 0x00 0AE4 KEY4 5th word of 128-bit Key register 0x00 0AE5 KEY5 6th word of 128-bit Key register 0x00 0AE6 KEY6 7th word of 128-bit Key register 0x00 0AE7 KEY7 High word of 128-bit Key register 0x00 0AEF CSMSCR CSM status and control register PWL in memory reserved for passwords only Address Name Description 0x3F 7FF8 PWL0 Low word of 128-bit password 0x3F 7FF9 PWL1 2nd word of 128-bit password 0x3F 7FFA PWL2 3rd word of 128-bit password 0x3F 7FFB PWL3 4th word of 128-bit password 0x3F 7FFC PWL4 5th word of 128-bit password 0x3F 7FFD PWL5 6th word of 128-bit password 0x3F 7FFE PWL6 7th word of 128-bit password 0x3F 7FFF PWL7 High word of 128-bit password
Locking and Unlocking the CSM

The CSM is always locked after reset To unlock the CSM:
Perform a dummy read of each PWL (passwords in the flash) Write the correct password to each KEY register
Passwords are all 0xFFFF on new devices

When passwords are all 0xFFFF, only a read of each PWL is required to unlock the device The bootloader does these dummy reads and hence unlocks devices that do not have passwords programmed
10 - 12
CSM Caveats
Never program all the PWLs as 0x0000
Doing so will permanently lock the CSM
Flash addresses 0x3F7F80 to 0x3F7FF5, inclusive, must be programmed to 0x0000 to securely lock the CSM Remember that code running in unsecured RAM cannot access data in secured memory
Dont link the stack to secured RAM if you have any code that runs from unsecured RAM
Do not embed the passwords in your code!

Generally, the CSM is unlocked only for debug Code Composer Studio can do the unlocking
CSM Password Match Flow

Yes
Start
Is PWL = all 0s? No
Device permanently locked
Flash device secure after reset or runtime
Is PWL = all Fs? No
Yes
Do dummy reads of PWL 0x3F 7FF8 0x3F 7FFF
Write password to KEY registers 0x00 0AE0 0x00 0AE7 (EALLOW) protected
Correct password? No
Yes
Device unlocked User can access onchip secure memory
10 - 13
Lab 10: Programming the Flash

Objective
The objective of this lab is to program and execute code from the on-chip flash memory. The TMS320F28035 device has been designed for standalone operation in an embedded system. Using the on-chip flash eliminates the need for external non-volatile memory or a host processor from which to bootload. In this lab, the steps required to properly configure the software for execution from internal flash memory will be covered.

ePWM1
ADCINA0
ADC
RESULT0
CLA
connector wire
data memory
pointer rewind
ePWM2
Objective: Program system into Flash Memory Learn use of CCS Flash Plug-in DO NOT PROGRAM PASSWORDS
Display using CCS
...
Procedure
Project File
1. A project named Lab10.pjt has been created for this lab. Open the project by Open and look in C:\C28x\Labs\Lab10. All Build clicking on Project Options have been configured the same as the previous lab. The files used in this lab are: Adc.c Filter.c Cla_10_12.c Flash.c ClaTasks.asm Gpio.c CodeStartBranch.asm Lab_10.cmd DefaultIsr_9_10.c Main_10.c DelayUs.asm Passwords.asm DSP2803x_GlobalVariableDefs.c PieCtrl_5_6_7_8_9_10.c DSP2803x_Headers_nonBIOS.cmd PieVect_5_6_7_8_9_10.c ECap_7_8_9_10_12.c SysCtrl.c EPwm_7_8_9_10_12.c Watchdog.c
10 - 14
Link Initialized Sections to Flash

Initialized sections, such as code and constants, must contain valid values at device power-up. Stand-alone operation of an F28035 embedded system means that no emulator is available to initialize the device RAM. Therefore, all initialized sections must be linked to the on-chip flash memory. Each initialized section actually has two addresses associated with it. First, it has a LOAD address which is the address to which it gets loaded at load time (or at flash programming time). Second, it has a RUN address which is the address from which the section is accessed at runtime. The linker assigns both addresses to the section. Most initialized sections can have the same LOAD and RUN address in the flash. However, some initialized sections need to be loaded to flash, but then run from RAM. This is required, for example, if the contents of the section needs to be modified at runtime by the code. 2. Open and inspect the linker command file Lab_10.cmd. Notice that a memory block named FLASH_ABCDEFGH has been been created at origin = 0x3E8000, length = 0x00FF80 on Page 0. This flash memory block length has been selected to avoid conflicts with other required flash memory spaces. See the reference slide at the end of this lab exercise for further details showing the address origins and lengths of the various memory blocks used. 3. Edit Lab_10.cmd to link the following compiler sections to on-chip flash memory block FLASH_ABCDEFGH: Compiler Sections .text .cinit .const .econst .pinit .switch 4. In Lab_10.cmd notice that the section named IQmath is an initialized section that needs to load to and run from flash. Previously the IQmath section was linked to L0SARAM. Edit Lab_10.cmd so that this section is now linked to FLASH_ABCDEFGH. Save your work and close the file.
Copying Interrupt Vectors from Flash to RAM

The interrupt vectors must be located in on-chip flash memory and at power-up needs to be copied to the PIE RAM as part of the device initialization procedure. The code that performs this copy is located in InitPieCtrl(). The C-compiler runtime support library contains a memory copy function called memcpy() which will be used to perform the copy.
10 - 15
5. Open and inspect InitPieCtrl() in PieCtrl_5_6_7_8_9_10.c. Notice the memcpy() function used to initialize (copy) the PIE vectors. At the end of the file a structure is used to enable the PIE.
Initializing the Flash Control Registers

The initialization code for the flash control registers cannot execute from the flash memory (since it is changing the flash configuration!). Therefore, the initialization function for the flash control registers must be copied from flash (load address) to RAM (run address) at runtime. The memory copy function memcpy() will again be used to perform the copy. The initialization code for the flash control registers InitFlash() is located in the Flash.c file. 6. Add Flash.c to the project. 7. Open and inspect Flash.c. The C compiler CODE_SECTION pragma is used to place the InitFlash() function into a linkable section named secureRamFuncs. 8. The secureRamFuncs section will be linked using the user linker command file Lab_10.cmd. Open and inspect Lab_10.cmd. The secureRamFuncs will load to flash (load address) but will run from L0SARAM (run address). Also notice that the linker has been asked to generate symbols for the load start, load size, and run start addresses. While not a requirement from a MCU hardware or development tools perspective (since the C28x MCU has a unified memory architecture), historical convention is to link code to program memory space and data to data memory space. Therefore, notice that for the L0SARAM memory we are linking secureRamFuncs to, we are specifiying PAGE = 0 (which is program memory). 9. Open and inspect Main_10.c. Notice that the memory copy function memcpy() is being used to copy the section secureRamFuncs, which contains the initialization function for the flash control registers. 10. Add a line of code in main() to call the InitFlash() function. There are no passed parameters or return values. You just type InitFlash(); at the desired spot in main().
Code Security Module and Passwords

The CSM module provides protection against unwanted copying (i.e. pirating!) of your code from flash, OTP memory, and the L0, L1, L2 and L3 RAM blocks. The CSM uses a 128-bit password made up of 8 individual 16-bit words. They are located in flash at addresses 0x3F7FF8 to 0x3F7FFF. During this lab, dummy passwords of 0xFFFF will be used therefore only dummy reads of the password locations are needed to unsecure the CSM. DO NOT PROGRAM ANY REAL PASSWORDS INTO THE DEVICE. After development, real passwords are typically
10 - 16
placed in the password locations to protect your code. We will not be using real passwords in the workshop. The CSM module also requires programming values of 0x0000 into flash addresses 0x3F7F80 through 0x3F7FF5 in order to properly secure the CSM. Both tasks will be accomplished using a simple assembly language file Passwords.asm. 11. Add Passwords.asm to the project. 12. Open and inspect Passwords.asm. This file specifies the desired password values (DO NOT CHANGE THE VALUES FROM 0xFFFF) and places them in an initialized section named passwords. It also creates an initialized section named csm_rsvd which contains all 0x0000 values for locations 0x3F7F80 to 0x3F7FF5 (length of 0x76). 13. Open Lab_10.cmd and notice that the initialized sections for passwords and csm_rsvd are linked to memories named PASSWORDS and CSM_RSVD, respectively.
Executing from Flash after Reset

The F28035 device contains a ROM bootloader that will transfer code execution to the flash after reset. When the boot mode selection is set for Jump to Flash mode, the bootloader will branch to the instruction located at address 0x3F7FF6 in the flash. An instruction that branches to the beginning of your program needs to be placed at this address. Note that the CSM passwords begin at address 0x3F7FF8. There are exactly two words available to hold this branch instruction, and not coincidentally, a long branch instruction LB in assembly code occupies exactly two words. Generally, the branch instruction will branch to the start of the Cenvironment initialization routine located in the C-compiler runtime support library. The entry symbol for this routine is _c_int00. Recall that C code cannot be executed until this setup routine is run. Therefore, assembly code must be used for the branch. We are using the assembly code file named CodeStartBranch.asm. 14. Open and inspect CodeStartBranch.asm. This file creates an initialized section named codestart that contains a long branch to the C-environment setup routine. This section needs to be linked to a block of memory named BEGIN_FLASH. 15. In the earlier lab exercises, the section codestart was directed to the memory named BEGIN_M0. Edit Lab_10.cmd so that the section codestart will be directed to BEGIN_FLASH. Save your work and close the opened files. On power up the reset vector will be fetched and the ROM bootloader will begin execution. If the emulator is connected, the device will be in emulator boot mode and will use the EMU_KEY and EMU_BMODE values in the PIE RAM to determine the bootmode. This mode was utilized in an earlier lab. In this lab, we will be disconnecting the emulator and running in stand-alone boot mode (but do not disconnect the emulator yet!). The bootloader will read the OTP_KEY and OTP_BMODE values from their locations in the OTP. The behavior when these values have not been programmed (i.e., both 0xFFFF) or have been set to invalid values is boot to flash bootmode.
10 - 17
Initializing the CLA

Previously, the named section Cla1Prog containing the CLA program tasks was linked directly to the CPU memory block L3DPSARAM for both load and run purposes. At runtime, all the code did was map the L3DPSARAM block to the CLA program memory space during CLA initialization. For an embedded application, the CLA program tasks are linked to load to flash and run from RAM. At runtime, the CLA program tasks must be copied from flash to L3DPSARAM. The memory copy function memcpy() will once again be used to perform the copy. After the copy is performed, the L3DPSARAM block will then be mapped to CLA program memory space as was done in the earlier lab. 16. Open and inspect Lab_10.cmd. Notice that the named section Cla1Prog will now load to flash (load address) but will run from L3DPSARAM (run address). The linker will also be used to generate symbols for the load start, load size, and run start addresses. 17. Open Cla_10_12.c and notice that the memory copy function memcpy() is being used to copy the CLA program code from flash to L3DPSARAM using the symbols generated by the linker. Just after the copy the Cla1Regs structure is used to configure the L3DPSARAM block as CLA program memory space. Close the inspected files.
Build Lab.out
18. At this point we need to build the project, but not have CCS automatically load it since CCS cannot load code into the flash (the flash must be programmed)! On the menu bar click: Option Customize and select the Program/Project CIO tab. Uncheck Load Program After Build. CCS has a feature that automatically steps over functions without debug information. This can be useful for accelerating the debug process provided that you are not interested in debugging the function that is being stepped-over. While single-stepping in this lab exercise we do not want to step-over any functions. Therefore, select the Debug Properties tab. Uncheck Step over functions without debug information when source stepping, then click OK. 19. Click the Build button to generate the Lab.out file to be used with the CCS Flash Plug-in.
CCS Flash Plug-in

20. Open the Flash Plug-in tool by clicking: Tools F28xx On-Chip Flash Programmer
21. A Clock Configuration window may open. If needed, in the Clock Configuration window set OSCCLK (MHz): to 10, DIVSEL: to /2, and PLLCR Value: to 12. Then click OK. In the next Flash Programmer Settings window confirm that the selected DSP device to program is F28035 and all options have been checked. Click OK.
10 - 18
22. The CCS Flash Programmer uses the Piccolo 10 MHz internal oscillator as the device clock during programming. Confirm the Clock Configuration in the upper left corner has the OSCCLK set to 10 MHz, the DIVSEL set to /2, and the PLLCR value set to 12. Recall that the PLL is divided by two, which gives a SYSCLKOUT of 60 MHz. 23. Confirm that all boxes are checked in the Erase Sector Selection area of the plug-in window. We want to erase all the flash sectors. 24. We will not be using the plug-in to program the Code Security Password. Do not modify the Code Security Password fields. They should remain as all 0xFFFF. 25. In the Operation block, notice that the COFF file to Program/Verify field automatically defaults to the current .out file. Check to be sure that Erase, Program, Verify is selected. We will be using the default wait states, as shown on the slide in this module. The selection for wait-states only affects the verify step, and makes little noticeable difference even if you reduce the wait-states. 26. Click Execute Operation to program the flash memory. Watch the programming status update in the plug-in window. 27. After successfully programming the flash memory, close the programmer window.
Running the Code Using CCS

28. In order to effectively debug with CCS, we need to load the symbolic debug information (e.g., symbol and label addresses, source file links, etc.) so that CCS knows where everything is in your code. Click: File Load Symbols Load Symbols Only
and select Lab10.out in the Debug folder. 29. Reset the CPU. The program counter should now be at 0x3FF8A1, which is the start of the bootloader in the Boot ROM. 30. Under GEL on the menu bar click: EMU_BOOT_FLASH. EMU Boot Mode Select This has the debugger load values into EMU_KEY and EMU_BMODE so that the bootloader will jump to "FLASH" at 0x3F7FF6. 31. Single-Step <F11> through the bootloader code until you arrive at the beginning of the codestart section in the CodeStartBranch.asm file. (Be patient, it will take about 125 single-steps). Notice that we have placed some code in CodeStartBranch.asm to give an option to first disable the watchdog, if selected. 32. Step a few more times until you reach the start of the C-compiler initialization routine at the symbol _c_int00. 33. Now do Debug Go Main. The code should stop at the beginning of your main() routine. If you got to that point succesfully, it confirms that the flash has been
10 - 19
programmed properly, that the bootloader is properly configured for jump to flash mode, and that the codestart section has been linked to the proper address. 34. You can now RUN the CPU, and you should observe the LED on the ControlCARD blinking. Try resetting the CPU, select the EMU_BOOT_FLASH boot mode, and then hitting RUN (without doing all the stepping and the Go Main procedure). The LED should be blinking again. 35. HALT the CPU.
Running the Code Stand-alone Operation (No Emulator)

36. Close Code Composer Studio. 37. Disconnect the USB cable (emulator) from the Docking Station (i.e. remove power from the ControlCARD). 38. Re-connect the USB cable to the Docking Station to power the ControlCARD. The LED should be blinking, showing that the code is now running from flash memory. End of Exercise
10 - 20
Lab 10 Reference: Programming the Flash
Flash Memory Section Blocks

origin = 0x3E 8000
FLASH
length = 0xFF80 page = 0
Lab_10.cmd
SECTIONS {
0x3F 7F80
CSM_RSVD
length = 0x76 page = 0
codestart csm_rsvd }
:> BEGIN_FLASH, PAGE = 0 :> CSM_RSVD, PAGE = 0
passwords :> PASSWORDS, PAGE = 0
0x3F 7FF6
BEGIN_FLASH
0x3F 7FF8
PASSWORDS
Startup Sequence from Flash Memory

0x3E 8000 _c_int00
4
FLASH (64Kw)
rts2800_ml.lib
0x3F 7FF6
LB _c_int00 Passwords (8w)
0x3F E000

0x3F F8A1
user code sections main ( ) { }
{SCAN GPIO} BROM vector (32w) 0x3F F8A1 0x3F FFC0

1
RESET
10 - 21
10 - 22
Communications
Introduction
The TMS320C28x contains features that allow several methods of communication and data exchange between the C28x and other devices. Many of the most commonly used communications techniques are presented in this module. The intent of this module is not to give exhaustive design details of the communication peripherals, but rather to provide an overview of the features and capabilities. Once these features and capabilities are understood, additional information can be obtained from various resources such as documentation, as needed. This module will cover the basic operation of the communication peripherals, as well as some basic terms and how they work.
Learning Objectives
Learning Objectives
Serial Peripheral Interface (SPI) Serial Communication Interface (SCI) Local Interconnect Network (LIN) Inter-Integrated Circuit (I2C) Enhanced Controller Area Network (eCAN)
Note: Up to 2 SPI modules (A/B), 1 SCI module (A), 1 LIN module (A), 1 I2C module (A), and 1 eCAN module (A) are available on the F2803x devices
C2000 Piccolo Workshop - Communications
11 - 1
Module Topics
Module Topics
Communications.......................................................................................................................................11-1 Module Topics........................................................................................................................................11-2 Communications Techniques .................................................................................................................11-3 Serial Peripheral Interface (SPI) ...........................................................................................................11-4 SPI Registers .....................................................................................................................................11-7 SPI Summary.....................................................................................................................................11-8 Serial Communications Interface (SCI) .................................................................................................11-9 Multiprocessor Wake-Up Modes.....................................................................................................11-11 SCI Registers ...................................................................................................................................11-14 SCI Summary ..................................................................................................................................11-15 Local Interconnect Network (LIN) .......................................................................................................11-16 LIN Message Frame and Data Timing ............................................................................................11-17 LIN Summary..................................................................................................................................11-18 Inter-Integrated Circuit (I2C)..............................................................................................................11-19 I2C Operating Modes and Data Formats .........................................................................................11-20 I2C Summary...................................................................................................................................11-21 Enhanced Controller Area Network (eCAN) .......................................................................................11-22 CAN Bus and Node .........................................................................................................................11-23 Principles of Operation....................................................................................................................11-24 Message Format and Block Diagram...............................................................................................11-25 eCAN Summary ..............................................................................................................................11-26
11 - 2
Communications Techniques
Communications Techniques
Several methods of implementing a TMS320C28x communications system are possible. The method selected for a particular design should reflect the method that meets the required data rate at the lowest cost. Various categories of interface are available and are summarized in the learning objective slide. Each will be described in this module.
Synchronous vs. Asynchronous

Synchronous
Short distances (onboard) High data rate Explicit clock
Asynchronous
longer distances Lower data rate ( 1/8 of SPI) Implied clock (clk/data mixed) Economical with reasonable performance
C28x Port U2
C28x
Port Destination
PCB
PCB
Serial ports provide a simple, hardware-efficient means of high-level communication between devices. Like the GPIO pins, they may be used in stand-alone or multiprocessing systems. In a multiprocessing system, they are an excellent choice when both devices have an available serial port and the data rate requirement is relatively low. Serial interface is even more desirable when the devices are physically distant from each other because the inherently low number of wires provides a simpler interconnection. Serial ports require separate lines to implement, and they do not interfere in any way with the data and address lines of the processor. The only overhead they require is to read/write new words from/to the ports as each word is received/transmitted. This process can be performed as a short interrupt service routine under hardware control, requiring only a few cycles to maintain. The C28x family of devices have both synchronous and asynchronous serial ports. Detailed features and operation will be described next.
11 - 3
Serial Peripheral Interface (SPI)

The SPI module is a synchronous serial I/O port that shifts a serial bit stream of variable length and data rate between the C28x and other peripheral devices. During data transfers, one SPI device must be configured as the transfer MASTER, and all other devices configured as SLAVES. The master drives the transfer clock signal for all SLAVES on the bus. SPI communications can be implemented in any of three different modes: MASTER sends data, SLAVES send dummy data MASTER sends data, one SLAVE sends data MASTER sends dummy data, one SLAVE sends data
In its simplest form, the SPI can be thought of as a programmable shift register. Data is shifted in and out of the SPI through the SPIDAT register. Data to be transmitted is written directly to the SPIDAT register, and received data is latched into the SPIBUF register for reading by the CPU. This allows for double-buffered receive operation, in that the CPU need not read the current received data from SPIBUF before a new receive operation can be started. However, the CPU must read SPIBUF before the new operation is complete of a receiver overrun error will occur. In addition, double-buffered transmit is not supported: the current transmission must be complete before the next data character is written to SPIDAT or the current transmission will be corrupted. The Master can initiate a data transfer at any time because it controls the SPICLK signal. The software, however, determines how the Master detects when the Slave is ready to broadcast.
SPI Data Flow

Simultaneous transmits and receive SPI Master provides the clock signal
SPI Device #1 - Master

shift SPI Shift Register SPI Shift Register clock
SPI Device #2 - Slave

shift SPI Shift Register
11 - 4
SPI Block Diagram

C28x - SPI Master Mode Shown
SPISIMO RX FIFO_0
RX FIFO_15 SPIRXBUF.15-0
MSB
SPIDAT.15-0
LSB
SPISOMI
SPITXBUF.15-0 TX FIFO_0
TX FIFO_15 LSPCLK baud rate clock polarity clock phase SPICLK
SPI Transmit / Receive Sequence

1. Slave writes data to be sent to its shift register (SPIDAT) 2. Master writes data to be sent to its shift register (SPIDAT or SPITXBUF) 3. Completing Step 2 automatically starts SPICLK signal of the Master 4. MSB of the Masters shift register (SPIDAT) is shifted out, and LSB of the Slaves shift register (SPIDAT) is loaded 5. Step 4 is repeated until specified number of bits are transmitted 6. SPIDAT register is copied to SPIRXBUF register 7. SPI INT Flag bit is set to 1 8. An interrupt is asserted if SPI INT ENA bit is set to 1 9. If data is in SPITXBUF (either Slave or Master), it is loaded into SPIDAT and transmission starts again as soon as the Masters SPIDAT is loaded
11 - 5
Since data is shifted out of the SPIDAT register MSB first, transmission characters of less than 16 bits must be left-justified by the CPU software prior to be written to SPIDAT. Received data is shifted into SPIDAT from the left, MSB first. However, the entire sixteen bits of SPIDAT is copied into SPIBUF after the character transmission is complete such that received characters of less than 16 bits will be right-justified in SPIBUF. The non-utilized higher significance bits must be masked-off by the CPU software when it interprets the character. For example, a 9 bit character transmission would require masking-off the 7 MSBs.
SPI Data Character Justification

Programmable data length of 1 to 16 bits Transmitted data of less than 16 bits must be left justified
MSB transmitted first
SPIDAT - Processor #1
11001001XXXXXXXX 11001001XXXXXXXX
Received data of less than 16 bits are right justified User software must mask-off unused MSBs
SPIDAT - Processor #2
XXXXXXXX11001001 XXXXXXXX11001001
11 - 6
SPI Registers
SPI Baud Rate Register

SpixRegs.SPIBRR
Need to set this only when in master mode!

15-7 reserved 6-0 SPI BIT RATE
LSPCLK (SPIBRR + 1)
SPIBRR = 3 to 127
SPICLK signal =
LSPCLK 4 , SPIBRR = 0, 1, or 2
Baud Rate Determination: The Master specifies the communication baud rate using its baud rate register (SPIBRR.6-0):
For SPIBRR = 3 to 127:
SPI Baud Rate =
LSPCLK bits/sec ( SPIBRR + 1)

LSPCLK bits/sec 4
For SPIBRR = 0, 1, or 2:
SPI Baud Rate =
From the above equations, one can compute Maximum data rate = 25 Mbps @ 100 MHz Character Length Determination: The Master and Slave must be configured for the same transmission character length. This is done with bits 0, 1, 2 and 3 of the configuration control register (SPICCR.3-0). These four bits produce a binary number, from which the character length is computed as binary + 1 (e.g. SPICCR.3-0 = 0010 gives a character length of 3).
11 - 7
Select SPI Registers

Configuration Control SpixRegs.SPICCR
Reset, Clock Polarity, Loopback, Character Length
Operation Control SpixRegs.SPICTL

Overrun Interrupt Enable, Clock Phase, Interrupt Enable Master / Slave Transmit enable
Status SpixRegs.SPIST
RX Overrun Flag, Interrupt Flag, TX Buffer Full Flag
FIFO Transmit SpixRegs.SPIFFTX FIFO Receive SpixRegs.SPIFFRX

FIFO Enable, FIFO Reset FIFO Over-flow flag, Over-flow Clear Number of Words in FIFO (FIFO Status) FIFO Interrupt Enable, Interrupt Status, Interrupt Clear FIFO Interrupt Level (Number of Words in FIFO)
Note: refer to the reference guide for a complete listing of registers
SPI Summary
SPI Summary
Synchronous serial communications
Two wire transmit or receive (half duplex) Three wire transmit and receive (full duplex)
Software configurable as master or slave

C28x provides clock signal in master mode
Data length programmable from 1-16 bits 125 different programmable baud rates
11 - 8
Serial Communications Interface (SCI)

The SCI module is a serial I/O port that permits Asynchronous communication between the C28x and other peripheral devices. The SCI transmit and receive registers are both double-buffered to prevent data collisions and allow for efficient CPU usage. In addition, the C28x SCI is a full duplex interface which provides for simultaneous data transmit and receive. Parity checking and data formatting is also designed to be done by the port hardware, further reducing software overhead.
SCI Pin Connections

(Full Duplex Shown)
TX FIFO_0 TX FIFO_0
TX FIFO_15 Transmitter-data buffer register 8 Transmitter shift register SCITXD SCITXD
TX FIFO_15 Transmitter-data buffer register 8 Transmitter shift register
Receiver shift register 8 Receiver-data buffer register RX FIFO_0
SCIRXD
SCIRXD
Receiver shift register 8 Receiver-data buffer register RX FIFO_0
RX FIFO_15
RX FIFO_15
SCI Device #1
SCI Device #2
11 - 9
SCI Data Format

NRZ (non-return to zero) format
Start LSB 2 3 4 5 6 7 MSB Addr/ Parity Stop 1 Stop 2 Data
This bit present only in Address-bit mode
Communications Control Register (ScixRegs.SCICCR)

7 Stop Bits 6 Even/Odd Parity 5 Parity Enable 4 3 Loopback Addr/Idle Enable Mode 2 SCI Char2 1 SCI Char1 0 SCI Char0
0 = 1 Stop bit 1 = 2 Stop bits 0 = Odd 1 = Even
0 = Disabled 1 = Enabled
0 = Idle-line mode 1 = Addr-bit mode
# of data bits = (binary + 1) e.g. 110b gives 7 data bits
0 = Disabled 1 = Enabled
The basic unit of data is called a character and is 1 to 8 bits in length. Each character of data is formatted with a start bit, 1 or 2 stop bits, an optional parity bit, and an optional address/data bit. A character of data along with its formatting bits is called a frame. Frames are organized into groups called blocks. If more than two serial ports exist on the SCI bus, a block of data will usually begin with an address frame which specifies the destination port of the data as determined by the users protocol. The start bit is a low bit at the beginning of each frame which marks the beginning of a frame. The SCI uses a NRZ (Non-Return-to-Zero) format which means that in an inactive state the SCIRX and SCITX lines will be held high. Peripherals are expected to pull the SCIRX and SCITX lines to a high level when they are not receiving or transmitting on their respective lines. When configuring the SCICCR, the SCI port should first be held in an inactive state. This is done using the SW RESET bit of the SCI Control Register 1 (SCICTL1.5). Writing a 0 to this bit initializes and holds the SCI state machines and operating flags at their reset condition. The SCICCR can then be configured. Afterwards, re-enable the SCI port by writing a 1 to the SW RESET bit. At system reset, the SW RESET bit equals 0.
11 - 10
SCI Data Timing

Start bit valid if 4 consecutive SCICLK periods of zero bits after falling edge
Majority vote taken on 4th , 5 th, and 6 th SCICLK cycles

Majority Vote SCICLK (Internal) SCIRXD
Start Bit
Falling Edge Detected
LSB of Data
Note: 8 SCICLK periods per data bit
Multiprocessor Wake-Up Modes
Multiprocessor Wake-Up Modes

Allows numerous processors to be hooked up to the bus, but transmission occurs between only two of them Idle-line or Address-bit modes Sequence of Operation
1. Potential receivers set SLEEP = 1, which disables RXINT except when an address frame is received 2. All transmissions begin with an address frame 3. Incoming address frame temporarily wakes up all SCIs on bus 4. CPUs compare incoming SCI address to their SCI address 5. Process following data frames only if address matches
11 - 11
Idle-Line Wake-Up Mode

Idle time separates blocks of frames Receiver wakes up when SCIRXD high for 10 or more bit periods Two transmit address methods Deliberate software delay of 10 or more bits Set TXWAKE bit to automatically leave exactly 11 idle bits
Idle periods of less than 10 bits SCIRXD/ SCITXD
Last Data SP ST Addr
Block of Frames
SP ST
Data
SP ST Last Data
SP
ST
Addr
SP
Idle Address frame Period follows 10 bit 10 bits or greater idle or greater
1st data frame
Idle Period 10 bits or greater
Address-Bit Wake-Up Mode

All frames contain an extra address bit Receiver wakes up when address bit detected Automatic setting of Addr/Data bit in frame by setting TXWAKE = 1 prior to writing address to SCITXBUF
Block of Frames
SCIRXD/ SCITXD
Last Data 0
SP
ST
Addr
1 SP ST
Data
0 SP ST Last Data 0 SP ST
Addr
1 SP
Idle Period length of no significance
First frame within 1st data frame block is Address. ADDR/DATA bit set to 1
no additional idle bits needed beyond stop bits
11 - 12
The SCI interrupt logic generates interrupt flags when it receives or transmits a complete character as determined by the SCI character length. This provides a convenient and efficient way of timing and controlling the operation of the SCI transmitter and receiver. The interrupt flag for the transmitter is TXRDY (SCICTL2.7), and for the receiver RXRDY (SCIRXST.6). TXRDY is set when a character is transferred to TXSHF and SCITXBUF is ready to receive the next character. In addition, when both the SCIBUF and TXSHF registers are empty, the TX EMPTY flag (SCICTL2.6) is set. When a new character has been received and shifted into SCIRXBUF, the RXRDY flag is set. In addition, the BRKDT flag is set if a break condition occurs. A break condition is where the SCIRXD line remains continuously low for at least ten bits, beginning after a missing stop bit. Each of the above flags can be polled by the CPU to control SCI operations, or interrupts associated with the flags can be enabled by setting the RX/BK INT ENA (SCICTL2.1) and/or the TX INT ENA (SCICTL2.0) bits active high. Additional flag and interrupt capability exists for other receiver errors. The RX ERROR flag is the logical OR of the break detect (BRKDT), framing error (FE), receiver overrun (OE), and parity error (PE) bits. RX ERROR high indicates that at least one of these four errors has occurred during transmission. This will also send an interrupt request to the CPU if the RX ERR INT ENA (SCICTL1.6) bit is set.
11 - 13
SCI Registers
SCI Baud Rate Registers

LSPCLK (BRR + 1) x 8 , BRR = 1 to 65535
SCI baud rate =

LSPCLK 16 , BRR = 0
Baud-Select MSbyte Register (ScixRegs.SCIHBAUD)

7 BAUD15 (MSB) 6 BAUD14 5 BAUD13 4 BAUD12 3 BAUD11 2 BAUD10 1 BAUD9 0 BAUD8
Baud-Select LSbyte Register (ScixRegs.SCILBAUD)

7 BAUD7 6 BAUD6 5 BAUD5 4 BAUD4 3 BAUD3 2 BAUD2 1 BAUD1 0 BAUD0 (LSB)
Baud Rate Determination: The values in the baud-select registers (SCIHBAUD and SCILBAUD) concatenate to form a 16 bit number that specifies the baud rate for the SCI. For BRR = 1 to 65535: SCI Baud Rate =
LSPCLK bits/sec ( BRR + 1) 8
For BRR = 0:
SCI Baud Rate =
LSPCLK bits/sec 16
Max data rate = 6.25 Mbps @ 100 MHz Note that the CLKOUT for the SCI module is one-half the CPU clock rate.
11 - 14
Select SCI Registers

Control 1 ScixRegs.SCICTL1
Reset, Transmitter / Receiver Enable TX Wake-up, Sleep, RX Error Interrupt Enable
Control 2 ScixRegs.SPICTL2
TX Buffer Full / Empty Flag, TX Ready Interrupt Enable RX Break Interrupt Enable
Receiver Status ScixRegs.SCIRXST

Error Flag, Ready, Flag Break-Detect Flag, Framing Error Detect Flag, Parity Error Flag, RX Wake-up Detect Flag
FIFO Transmit ScixRegs.SCIFFTX FIFO Receive ScixRegs.SCIFFRX

FIFO Enable, FIFO Reset FIFO Over-flow flag, Over-flow Clear Number of Words in FIFO (FIFO Status) FIFO Interrupt Enable, Interrupt Status, Interrupt Clear FIFO Interrupt Level (Number of Words in FIFO)
Note: refer to the reference guide for a complete listing of registers
SCI Summary
SCI Summary
Asynchronous communications format 65,000+ different programmable baud rates Two wake-up multiprocessor modes
Idle-line wake-up & Address-bit wake-up
Programmable data word format

1 to 8 bit data word length 1 or 2 stop bits even/odd/no parity
Error Detection Flags

Parity error; Framing error; Overrun error; Break detection
Transmit FIFO and receive FIFO Individual interrupts for transmit and receive
11 - 15
Local Interconnect Network (LIN)

Compliant to the LIN2.0 protocol Specification Package Module based on SCI (core) with added hardware features for LIN compatibility:
Error detector Mask filter Synchronizer Multi-buffered receiver/transmitter
Standard is based on SCI (UART) serial data link format Communication concept is single-master/multiple-slave with message identification for multi-cast transmission between any network nodes Module can be used in LIN mode or SCI (UART) mode
LIN Block Diagram

7 7 7 7 7 7 7 7
RD7 RD6 RD5 RD4 RD3 RD2 RD1 RD0
0 0 0 0 0 0 0 0
Synchronizer
Checksum Calculator Parity Calculator Bit Monitor
LINRX/ SCIRX LINTX/ SCITX
Mask Filter
SCIRXSHF SCITXSHF
7 7 7 7 7 7 7 7
TD0 TD1 TD2 TD3 TD4 TD5 TD6 TD7
0 0 0 0 0 0 0 0
11 - 16
LIN Message Frame and Data Timing
LIN Message Frame

Message Frame Master Header Response 1 to 8 Data Fields
Sync Sync ID Break Field Field
Data Field
Data Field
Data Field
Data Field
Data Check Sum Field Field
In-Frame Space
Interbyte Spaces
Sync Break beginning of a message Sync Field bit rate information ID Field content of a message Data Field consists of 1 data byte, 1 start bit, and 1 stop bit (10 bits total) Checksum Field consists of 1 checksum byte, 1 start bit and 1 stop bit (10 bits total) In-Frame & Interbyte Spaces can be 0
LIN Data Timing

To make a determination of the bit value, 16 samples of each bit are taken with majority vote on samples 8, 9, and 10
Majority Vote
Majority Vote
Majority Vote
LM_CLK (Internal) LINRX
LIN module is clocked at the CPU clock (SYSCLKOUT)
11 - 17
LIN Summary
LIN Summary
Functionally compatible with standalone SCI of C28x devices Identification masks for filtering Automatic master header generation 228 programmable transmission rates Automatic wakeup support Error detection (bit, bus, no response, checksum, synchronization, parity) Multi-buffered receive/transmit units
11 - 18
Inter-Integrated Circuit (I2C)

Philips I2C-bus specification compliant, version 2.1 Data transfer rate from 10 kbps up to 400 kbps Each device can be considered as a Master or Slave Master initiates data transfer and generates clock signal Device addressed by Master is considered a Slave Multi-Master mode supported Standard Mode send exactly n data values (specified in register) Repeat Mode keep sending data values (use software to initiate a stop or new start condition)
VDD Pull-up Resisters
..
Serial Data (SDA) Serial Clock (SCL)
..
. . . . . . . .
I2C EPROM 28xx I2C
28xx I2C
I2C Controller
I2C Block Diagram
I2CXSR
I2CDXR TX FIFO
SDA RX FIFO I2CRSR I2CDRR
SCL
Clock Circuits
11 - 19
I2C Operating Modes and Data Formats
I2C Operating Modes

Operating Mode Slave-receiver mode Description Module is a slave and receives data from a master (all slaves begin in this mode) Slave-transmitter mode Module is a slave and transmits data to a master (can only be entered from slave-receiver mode) Master-receiver mode Module is a master and receives data from a slave (can only be entered from master-transmit mode) Master-transmitter mode Module is a master and transmits to a slave (all masters begin in this mode)
I2C Serial Data Formats

7-Bit Addressing Format
1 7 1 1 n 1 n 1 1
Slave Address
R/W ACK
Data
ACK
Data
ACK P
10-Bit Addressing Format

1 7 1 1 8 1 n 1 1
11110AA
R/W ACK AAAAAAAA ACK
Data
ACK P
Free Data Format

1 n 1 n 1 n 1 1
Data
ACK
Data
ACK
Data
ACK P
R/W = 0 master writes data to addressed slave R/W = 1 master reads data from the slave n = 1 to 8 bits S = Start (high-to-low transition on SDA while SCL is high) P = Stop (low-to-high transition on SDA while SCL is high)
11 - 20
I2C Arbitration
Arbitration procedure invoked if two or more mastertransmitters simultaneously start transmission
Procedure uses data presented on serial data bus (SDA) by competing transmitters First master-transmitter which drives SDA high is overruled by another master-transmitter that drives SDA low Procedure gives priority to the data stream with the lowest binary value
SCL Data from device #1 Data from device #2 SDA

1 1 1 0 0 0 0 0 1 1 0 0 1 1
Device #1 lost arbitration and switches to slavereceiver mode Device #2 drives SDA
I2C Summary
I2C Summary
Compliance with Philips I2C-bus specification (version 2.1) 7-bit and 10-bit addressing modes Configurable 1 to 8 bit data words Data transfer rate from 10 kbps up to 400 kbps Transmit FIFO and receive FIFO
11 - 21
Enhanced Controller Area Network (eCAN)
Controller Area Network (CAN)

A Multi-Master Serial Bus System CAN 2.0B Standard High speed (up to 1 Mbps) Add a node without disturbing the bus (number of nodes not limited by protocol) Less wires (lower cost, less maintenance, and more reliable) Redundant error checking (high reliability) No node addressing (message identifiers) Broadcast based signaling A B C
CAN does not use physical addresses to address stations. Each message is sent with an identifier that is recognized by the different nodes. The identifier has two functions it is used for message filtering and for message priority. The identifier determines if a transmitted message will be received by CAN modules and determines the priority of the message when two or more nodes want to transmit at the same time.
11 - 22
CAN Bus and Node
CAN Bus
Two wire differential bus (usually twisted pair) Max. bus length depend on transmission rate
40 meters @ 1 Mbps
CAN NODE A
CAN_H 120 CAN_L
CAN NODE B
CAN NODE C
120
The MCU communicates to the CAN Bus using a transceiver. The CAN bus is a twisted pair wire and the transmission rate depends on the bus length. If the bus is less than 40 meters the transmission rate is capable up to 1 Mbit/second.
CAN Node
Wired-AND Bus Connection
CAN_H 120 CAN_L 120
CAN Transceiver (e.g. TI SN65HVD23x)
TX
RX CAN Controller (e.g. TMS320F28035)
11 - 23
Principles of Operation
Principles of Operation
Data messages transmitted are identifier based, not address based Content of message is labeled by an identifier that is unique throughout the network
(e.g. rpm, temperature, position, pressure, etc.)
All nodes on network receive the message and each performs an acceptance test on the identifier If message is relevant, it is processed (received); otherwise it is ignored Unique identifier also determines the priority of the message
(lower the numerical value of the identifier, the higher the priority)
When two or more nodes attempt to transmit at the same time, a non-destructive arbitration technique guarantees messages are sent in order of priority and no messages are lost
Non-Destructive Bitwise Arbitration

Bus arbitration resolved via arbitration with wired-AND bus connections
Dominate state (logic 0, bus is high) Recessive state (logic 1, bus is low)
Start Bit
Node A Node B Node C CAN Bus Node B loses arbitration Node C loses arbitration
Node A wins arbitration
11 - 24
Message Format and Block Diagram
CAN Message Format

Data is transmitted and received using Message Frames 8 byte data payload per message Standard and Extended identifier formats
Standard Frame: 11-bit Identifier (CAN v2.0A)
Arbitration Field S O F Control Field Data Field E O F
11-bit R I Identifier T D r0 DLC R E
08 Bytes Data
CRC
ACK
Extended Frame: 29-bit Identifier (CAN v2.0B)

Arbitration Field S O F Control Field Data Field E O F
R 11-bit S I 18-bit T r1 r0 DLC Identifier R D Identifier R R E
08 Bytes Data
CRC
ACK
The MCU CAN module is a full CAN Controller. It contains a message handler for transmission and reception management, and frame storage. The specification is CAN 2.0B Active that is, the module can send and accept standard (11-bit identifier) and extended frames (29-bit identifier).
eCAN Block Diagram

eCAN0INT eCAN1INT
Address 32
Data
Mailbox RAM (512 bytes) 32 Mailboxes (4 x 32-bit words) A message mailbox

Identifier MID Control MCF Data low MDL Data high - MDH
Memory Management Unit
eCAN Memory (512 bytes)
32
CPU Interface, Receive Control Unit Timer Management Unit
32
Register and Message Object Control
32
Receive Buffer Transmit Buffer Control Buffer Status Buffer
SN65HVD23x 3.3-V CAN Transceiver
CAN Bus
11 - 25
The CAN controller module contains 32 mailboxes for objects of 0 to 8-byte data lengths: configurable transmit/receive mailboxes configurable with standard or extended indentifier
The CAN module mailboxes are divided into several parts: MID contains the identifier of the mailbox MCF (Message Control Field) contains the length of the message (to transmit or receive) and the RTR bit (Remote Transmission Request used to send remote frames) MDL and MDH contains the data
The CAN module contains registers which are divided into five groups. These registers are located in data memory from 0x006000 to 0x0061FF. The five register groups are: Control & Status Registers Local Acceptance Masks Message Object Time Stamps Message Object Timeout Mailboxes
eCAN Summary
eCAN Summary
Fully compliant with CAN standard v2.0B Supports data rates up to 1 Mbps Thirty-two mailboxes
Configurable as receive or transmit Configurable with standard or extended identifier Programmable receive mask Uses 32-bit time stamp on messages Programmable interrupt scheme (two levels) Programmable alarm time-out
Programmable wake-up on bus activity Self-test mode
11 - 26
DSP/BIOS
Introduction
This module discusses the basic features of using DSP/BIOS in a system. Scheduling threads, periodic functions, and the use of real-time analysis tools will be demonstrated, in addition to programming the flash with DSP/BIOS.
Learning Objectives
Learning Objectives
Introduction to DSP/BIOS DSP/BIOS Configuration Tool Scheduling DSP/BIOS threads Periodic Functions Real-Time Analysis Tools
C2000 Piccolo Workshop - DSP/BIOS
12 - 1
Module Topics
Module Topics
DSP/BIOS..................................................................................................................................................12-1 Module Topics........................................................................................................................................12-2 Introduction to DSP/BIOS .....................................................................................................................12-3 DSP/BIOS Configuration Tool...............................................................................................................12-4 Scheduling DSP/BIOS threads...............................................................................................................12-9 Periodic Functions...............................................................................................................................12-14 Real-Time Analysis Tools.....................................................................................................................12-15 Lab 12: DSP/BIOS...............................................................................................................................12-16
12 - 2
Introduction to DSP/BIOS
Introduction to DSP/BIOS
What is DSP/BIOS?
A full-featured, scalable real-time kernel
System configuration tools Preemptive multi-threading scheduler Real-time analysis tools
Why Use DSP/BIOS?

Helps Manage complex system resources
no need to develop or maintain a home-brew kernel faster time to market
Efficient debugging of real-time applications

Real-Time Analysis
Create robust applications

industry proven kernel technology
Reduce cost of software maintenance

code reuse and standardized software
Integrated with Code Composer Studio IDE

requires no runtime license fees fully supported by TI
Uses minimal Mips and Memory (2-8Kw)

scalable use only what is needed easily fits in limited memory space
12 - 3
DSP/BIOS Configuration Tool

The DSP/BIOS Configuration Tool (often called Config Tool or GUI Tool or GUI) creates and modifies a system file called the Text Configuration File (.tcf). If we talk about using .tcf files, were also talking about using the Config Tool.
DSP/BIOS Configuration Tool (file .tcf)

System Setup Tools
Handles memory configuration (builds .cmd file), run-time support libraries, interrupt vectors, system setup and reset, etc.
Real-Time Analysis Tools

Allows application to run uninterrupted while displaying debug data
Real-Time Scheduler
Preemptive tread manager kernel configures DSP/BIOS scheduling
Real-Time I/O
Allows two way communication between threads or between target and PC host
The GUI (graphical user interface) simplifies system design by: Automatically including the appropriate runtime support libraries Automatically handles interrupt vectors and system reset Handles system memory configuration (builds .cmd file) When a .tcf file is saved, the Config Tool generates 5 additional files: Filename.tcf Filenamecfg_c.c Filenamecfg.s28 Filenamecfg.cmd Filenamecfg.h Filenamecfg.h28 Text Configuration File C code created by Config Tool ASM code created by Config Tool Linker command file header file for *cfg_c.c header file for *cfg.s28
When you add a .tcf file to your project, CCS automatically adds the C and assembly (.s28) files and the linker command file (.cmd) to the project under the Generated Files folder.
12 - 4
1. Creating a New Memory Region (Using MEM)

First, to create a specific memory area, open up the .tcf file, right-click on the Memory Section Manager and select Insert MEM. Give this area a unique name and then specify its base and length. Once created, you can place sections into it (shown in the next step).
Memory Section Manager (MEM)

Generates the main linker command file for your code project
Create memories Place sections
To create a new memory area:

Right-click on MEM and select insert memory Enter your choice of a name for the memory Right-click on the memory, and select Properties
fill in base, length, space
12 - 5
2. Placing Sections MEM Manager Properties

The configuration tool makes it easy to place sections. The predefined compiler sections that were described earlier each have their own drop-down menu to select one of the memory regions you defined (in step 1).
Memory Section Manager Properties

To place a section into a memory area:
Right-click on MEM and select Properties Select the desired tab (e.g. Compiler) Select the memory you would like to link each section to
12 - 6
3. PIE Interrupts HWI Interrupts

The configuration tools is also used to assign the interrupt vectors. The vectors are placed into a section named .hwi_vec. The memory manager (MEM) links this section to the proper location in memory.
Hardware Interrupt Manager (HWI)

Config Tool used to assign interrupt vectors Vectors are placed in the section .hwi_vec Use MEM manager to link .hwi_vec to the proper memory
12 - 7
4. Running the Linker

Creating the Linker Command File (via .tcf) When you have finished creating memory regions and allocating sections into these memory areas (i.e. when you save the .tcf file), the CCS configuration tool creates five files. One of the files is BIOSs cfg.cmd file a linker command file.
Files Created by the Configuration Tool

Config tool generates five different files .cmd file is generated from your MEM settings Vectors put into *cfg_c.c
*cfg.cmd *cfg.h *.tcf

save to compiler
*cfg_c.c *cfg.h28 *cfg.s28
This file contains two main parts, MEMORY and SECTIONS. (Though, if you open and examine it, its not quite as nicely laid out as shown above.) Running the Linker The linkers main purpose is to link together various object files. It combines like-named input sections from the various object files and places each new output section at specific locations in memory. In the process, it resolves (provides actual addresses for) all of the symbols described in your code. The linker can create two outputs, the executable (.out) file and a report which describes the results of linking (.map). Note: The linker gets run automatically when you BUILD or REBUILD your project.
12 - 8
Scheduling DSP/BIOS threads
DSP/BIOS Thread Types

HWI
Hardware Interrupts Used to implement urgent part of real-time event Triggered by hardware interrupt HWI priorities fixed in hardware Use SWI to perform HWI follow-up activity SWI's are posted by software Multiple SWIs at each of 15 priority levels Use TSK to run different programs concurrently under separate contexts TSK's enabled by posting semaphore (a signal) Runs when no service routines are pending Runs as an infinite loop, like traditional while loop All BIOS data transfers to host occur here
SWI Priority
Software Interrupts
TSK
Tasks
IDL
Background
Enabling DSP/BIOS in main()

void main(void) { //*** Initialization . . . //*** Enable global interrupts // asm( CLRC INTM); //*** Main Loop // while(1); } //end of main()
BIOS will enable global interrupts for you Must delete the endless loop at end of main()
main() returns to BIOS and goes to the IDLE thread, allowing BIOS to schedule events, transfer data to the host, etc. An endless loop in main() will keep BIOS from running
12 - 9
Using Hardware Interrupts - HWI
Interrupt priority fixed by hardware
The HWI Dispatcher

For non-BIOS code, use the interrupt keyword to declare an ISR
tells the compiler to perform context save/restore interrupt void MyHwi(void) { }
For DSP/BIOS code, use the Dispatcher to perform the save/restore

Remove the interrupt keyword from the MyHwi() Check the Use Dispatcher box when you configure the interrupt vector in the DSP/BIOS configuration tool This is necessary if you want to use any DSP/BIOS functionality inside the ISR
12 - 10
Using Software Interrupts - SWI

Make each algorithm an independent software interrupt
SWI scheduling is handled by DSP/BIOS HWI function triggered by hardware SWI function triggered by software e.g. a call to SWI_post()
Why use a SWI?

No limitation on number of SWIs, and priorities for SWIs are user-defined SWI can be scheduled by hardware or software event(s) Defer processing from HWI to SWI
SWI Properties
12 - 11
Managing SWI Priority

Drag and Drop SWIs to change Drag and Drop SWIs to change priority priority Equal priority SWIs run in the Equal priority SWIs run in the order that they are posted order that they are posted
Priority Based Thread Scheduling

post3 rtn HWI 2
(highest)
SWI_post(&swi2); post2 rtn
HWI 1 post1 SWI 3 int2 SWI 2 rtn SWI 1 rtn MAIN int1 IDLE
(lowest)
rtn rtn
User sets the priority...BIOS does the scheduling
12 - 12
Using Tasks (TSK)

SWI vs. TSK
SWI
SWI_post start
must run to completion
TSK end
SEM_pend
SEM_post
start end
Similar to hardware interrupt, but triggered by SWI_post() SWIs must run to completion All SWI's use system stack faster context switching smaller code size
Pause (blocked state)
SEM_post() readies the TSK which pends on an event TSKs can be terminated by S/W Each TSK has its own stack slower context switching larger code size
12 - 13
Periodic Functions
Periodic Functions
Using Periodic Functions - PRD

tick DSP/BIOS CLK period LED LED LED
Periodic functions are a special type of SWI that are triggered by DSP/BIOS Periodic functions run at a user specified rate: - e.g. LED blink requires 0.5 Hz Use the CLK Manager to specify the DSP/BIOS CLK rate in microseconds per tick Use the PRD Manager to specify the period (for the function) in ticks Allows multiple periodic functions with different rates
Creating a Periodic Function

tick DSP/BIOS CLK period func1 func1 func1
12 - 14
Built-in Real-Time Analysis Tools

Gather data on target (3-10 CPU cycles) Send data during BIOS IDL (100s of cycles) Format data on host (1000s of cycles) Data gathering does NOT stop target CPU
Execution Graph
Software logic analyzer Debug event timing and priority
CPU Load Graph

Shows amount of CPU horsepower being consumed
Built-in Real-Time Analysis Tools

Statistics View
Profile routines w/o halting the CPU
Message LOG
Send debug msgs to host Doesnt halt the DSP Deterministic, low DSP cycle count More efficient than traditional printf()
LOG_printf(&trace, LedSwiCount = %u", LedSwiCount++);
12 - 15
Lab 12: DSP/BIOS
Lab 12: DSP/BIOS

Objective
The objective of this lab is to become familiar with DSP/BIOS. In this lab exercise, we will make use of the DSP/BIOS configuration tool, implement a software interrupt (SWI) and periodic function (PRD), program the DSP/BIOS project into the flash, and explore the built-in real-time analysis tools. The DSP/BIOS configuration tool creates a text configuration file (*.tcf) and generates a linker command file (*cfg.cmd). This generated linker command file is functionally equivalent to the linker command file previously used. The memory area of the lab linker command file will be deleted; however, part of the sections area will be used to link sections that are not part of DSP/BIOS. In the lab files we will change the CLA HWI (CLA1_INT1_ISR) to a SWI and replace the LED blink routine with a periodic function. The steps required to properly configure the software for execution from internal flash memory will be covered. Features of the real-time analysis tools, such as the CPU Load Graph, Execution Graph, Message Log, and RTA Control Panel will be demonstrated.
Lab 12: DSP/BIOS

ePWM1
ADCINA0
ADC
RESULT0
CLA
connector wire
data memory
pointer rewind
ePWM2
Objective: Use DSP/BIOS Configuration Tool to:

Handle system memory & interrupt vectors Create a .tcf file
...
Change CLA CLA1_INT1_ISR HWI to SWI Replace LED blink routine with a Periodic Function Program system into Flash Memory
Display using CCS
Procedure
Project File
1. A project named Lab12.pjt has been created for this lab. Open the project by Open and look in C:\C28x\Labs\Lab12. All Build clicking on Project Options have been configured the same as the previous lab. The files used in this lab are:
12 - 16
Lab 12: DSP/BIOS
Adc.c Cla_10_12.c ClaTasks.asm CodeStartBranch.asm DefaultIsr_12.c DelayUs.asm DSP2803x_GlobalVariableDefs.c DSP2803x_Headers_BIOS.cmd ECap_7_8_9_10_12.c EPwm_7_8_9_10_12.c
Filter.c Flash.c Gpio.c Lab_12.cmd Main_12.c Passwords.asm PieCtrl_12.c SysCtrl.c Watchdog.c
Edit Lab.h File

2. Edit Lab.h to uncomment the line that includes the labcfg.h header file. This is the DSP/BIOS generated include file, and is needed to allow code to access the DSP/BIOS functions and data structures. Next, comment out the line that includes the DSP2803x_DefaultIsr.h ISR function prototypes. DSP/BIOS will supply its own ISR function prototypes. 3. In our lab setup, we are running the ADC at a 50 kHz interrupt rate. Such a high frequency interrupt would typically be handled directly in the HWI, as SWIs and TSKs have some overhead associated with them and lauching them this frequently can cause very large processing loads on the CPU. DSP/BIOS is flexible in this way. You can have some interrupts processed directly in the HWI, and others delegated to SWIs or TSKs. For purposes of this lab however, we would like to illustrate how to code a SWI. Therefore, we will convert the ADC ISR into a SWI. To reduce the CPU load, we are going to reduce the frequency of the ADC sample rate by half to 25 kHz. In Lab.h modify the constant definition for the ADC sample rate as follows:
#define ADC_SAMPLE_PERIOD 2399 // 25 KHz sampling
Save and close the file.
Remove rts2800_ml.lib and Inspect Lab_12.cmd

4. The DSP/BIOS configuration tool supplies its own RTS library. Open the Build Options and select the Linker tab. In the Libraries Category, find the Include Libraries (-l) box and delete: rts2800_ml.lib. 5. Select the Compiler tab. As the project is now configured, we would get a warning at build time stating that the typedef name has already been declared with the same type. This is because it has been defined twice; once in the header files and again in the include file generated by DSP/BIOS. To suppress the warning select Diagnostics Category and find the Suppress Diagnostic <n> (-pds): box. Type in code number 303. Select OK and the Build Options window will close. 6. We will be using the DSP/BIOS configuration tool to create a linker command file. Open and inspect Lab_12.cmd. Notice that the linker command file does not have a memory
12 - 17
Lab 12: DSP/BIOS
area and includes only a limited sections area. These sections are not part of DSP/BIOS and need to be included in a user linker command file. Close the inspected file.
Using the DSP/BIOS Configuration Tool

7. The text configuration file (*.tcf) created by the DSP/BIOS configuration tool controls a wide range of CCS capabilities. The .tcf file will be used to automatically create and perform memory management. Create a new .tcf file for this lab. On the menu bar click: File New DSP/BIOS Configuration
A dialog box appears showing a number of available .tcf seed files. The seed files are used to configure many objects specific to the processor and will be invoked as the first item in your own .tcf file. On the C2xxx tab select the ti.platforms.control28035 template and click OK. A configuration window will open. 8. Save the configuration file by selecting: File Save As
and name it Lab.tcf in C:\C28x\Labs\Lab12 then click Save. Close the configuration window and select YES to save changes to Lab.tcf. 9. Add the configuration file to the project. Click: Project Add Files to Project
Make sure youre looking in C:\C28x\Labs\Lab12. Change the files of type to view All Files (*.*) and select Lab.tcf. Click OPEN to add the file to the project. 10. In the project window left click the plus sign (+) to the left of DSP/BIOS Config. Notice that the Lab.tcf file is listed. 11. Next, add the generated linker command file Labcfg.cmd to the project. After the file has been added you will notice that it is listed under the source files.
Create New Memory Sections Using the TCF File

12. Open the Lab.tcf file by double clicking on Lab.tcf. In the configuration window, left click the plus sign next to System and the plus sign next to MEM. By default, the Memory Section Manager has combined the memory space L1, L2 and L3DPSARAM into a single memory block called DPSARAM. It has also combined M0 and M1SARAM into a single memory block called MSARAM. 13. Next, we will add some of the additional memory sections that will be needed for the lab exercises in this module. To add a memory section: Right click on MEM Memory Section Manager and select Insert MEM. Rename the newly added memory section to BEGIN_FLASH. Repeat the process and add the following memory sections: CLAMSGRAM1, CLAMSGRAM2, CSM_RSVD, IQTABLES, L3DPSARAM, and PASSWORDS. Double check and see that all seven memory sections have been added.
12 - 18
Lab 12: DSP/BIOS
14. Modify the base addresses, length, and space of each of the memory sections to correspond to the memory mapping shown in the table below. To modify the length, base address, and space of a memory section, right click on the memory in the configuration tool, and select Properties.
Memory BEGIN_FLASH
Base
Length
Space code data data code
0x3F 7FF6 0x0002 0x0080 0x0080 0x0076
CLAMSGRAM1 0x00 1480 CLAMSGRAM2 0x00 1500 CSM_RSVD IQTABLES L3DPSARAM PASSWORDS 0x3F 7F80
0x3F E000 0x0B50 code 0x00 9000 0x1000 code code
0x3F 7FF8 0x0008
15. Modify the base addresses, length, and space of each of the memory sections to avoid memory conflicts with the newly added memory sections as shown in the table below.
Memory
Base
Length
Space
BOOTROM 0x3F F27C 0x0D44 code DPSARAM FLASH 0x00 8800 0x3E 8000 0x0800 0xFF80 data code
Link Uninitialized Sections to RAM

16. Right click on MEM Memory Section Manager and select Properties. Select the Compiler Sections tab and link the following uninitialized sections into the MSARAM memory block via the pull-down boxes.
MSARAM .bss .ebss
12 - 19
Lab 12: DSP/BIOS
Link Initialized Sections to Flash

All initialized sections must be linked to the on-chip flash memory. Each initialized section has two addresses associated with it. First, it has a LOAD address which is the address to which it gets loaded at load time (or at flash programming time). Second, it has a RUN address which is the address from which the section is accessed at runtime. The linker assigns both addresses to the section. Most initialized sections can have the same LOAD and RUN address in the flash. However, some initialized sections need to be loaded to flash, but then run from RAM. This is required, for example, if the contents of the section needs to be modified at runtime by the code. 17. This step assigns the RUN address of those sections that need to run from flash. Using the MEM Memory Section Manager in the DSP/BIOS configuration tool link the following sections to on-chip flash memory via the pull-down boxes:
BIOS Data tab .gblinit
BIOS Code tab .bios .sysinit .hwi .rtdx_text
Compiler Sections tab .text .switch .cinit .pinit .econst / .const .data / .cio
18. This step assigns the LOAD address of those sections that need to load to flash. Again using the MEM Memory Section Manager in the DSP/BIOS configuration tool select the Load Address tab and check the Specify Separate Load Addresses box. Then set all entries to the FLASH memory block. 19. Click the BIOS Data tab and notice that the .stack section has been linked into memory. Click OK to close the window. 20. The section named IQmath is an initialized section that needs to load to and run from flash. This section is not linked using the DSP/BIOS configuration tool (because it is neither a standard compiler section nor a DSP/BIOS generated section). Instead, this section is linked with the user linker command file (Lab_12.cmd). Open and inspect Lab_12.cmd. Previously the IQmath section was linked to L0SARAM. Notice that this section is now linked to FLASH.
12 - 20
Lab 12: DSP/BIOS
Set the Stack Size in the TCF File

Recall in the previous lab exercise that the stack size was set using the CCS project Build Options. When using the DSP/BIOS configuration tool, the stack size is instead specified in the .tcf file. First we need to remove the stack size setting from the project Build Options. 21. Click: Project Build Options and select the Linker tab. Delete the entry of 0x200 in the Stack Size box. Select OK to close the Build Options window. 22. Using the MEM Memory Section Manager select the General tab. Set the Stack Size to 0x100. The stack size needs to be reduced from 0x200 to 0x100 because of the limited amount of available RAM on the device when using DSP/BIOS. Click OK to close the window.
Copying .hwi_vec Section from Flash to RAM

The DSP/BIOS .hwi_vec section contains the interrupt vectors. This section must be loaded to flash (load address) but run from RAM (run address). The code that performs this copy is located in InitPieCtrl(). The linker command file generated by the DSP/BIOS configuration tool generates global symbols that can be accessed by code in order to determine the load address, run address, and length of the .hwi_vec section. The RTS library contains a memory copy function called memcpy() which will be used to perform the copy. 23. Open and inspect InitPieCtrl() in PieCtrl_12.c. Notice the memcpy() function and the symbols used to initialize (copy) the .hwi_vec section.
Copying the .trcdata Section from Flash to RAM

The DSP/BIOS .trcdata section is used by CCS and DSP/BIOS for certain real-time debugging features. This section must be loaded to flash (load address) but run from RAM (run address). The linker command file generated by the DSP/BIOS configuration tool generates global symbols that can be accessed by code in order to determine the load address, run address, and length of the .trcdata section. The memory copy function memcpy() will again be used to perform the copy. The copying of .trcdata must be performed prior to main(). This is because DSP/BIOS modifies the contents of .trcdata during DSP/BIOS initialization, which also occurs prior to main(). The DSP/BIOS configuration tool provides a user initialization function which will be used to perform the .trcdata section copy prior to both main() and DSP/BIOS initialization. 24. In the DSP/BIOS configuration file (Lab.tcf) and select the Properties for the Global Settings. Check the box Call User Init Function and enter the UserInit() function name with a leading underscore: _UserInit. This will cause the function UserInit() to execute prior to main(). Click OK to close the window. 25. Open and inspect the file Main_12.c. Notice that the function UserInit() is used to copy the .trcdata section from its load address to its run address before main().
12 - 21
Lab 12: DSP/BIOS
Initializing the Flash Control Registers

The initialization code for the flash control registers cannot execute from the flash memory (since it is changing the flash configuration!). Therefore, the initialization function for the flash control registers must be copied from flash (load address) to RAM (run address) at runtime. The memory copy function memcpy() will again be used to perform the copy. The initialization code for the flash control registers InitFlash() is located in the Flash.c file. 26. Open and inspect Flash.c. The C compiler CODE_SECTION pragma is used to place the InitFlash() function into a linkable section named secureRamFuncs. 27. Since the DSP/BIOS configuration tool does not know about user defined sections, the secureRamFuncs section will be linked using the user linker command file Lab_12.cmd. Open and inspect Lab_12.cmd. The secureRamFuncs will load to flash (load address) but will run from LSARAM (run address). Also notice that the linker has been asked to generate symbols for the load start, load size, and run start addresses. 28. Open and inspect Main_12.c. Notice that the memory copy function memcpy() is being used to copy the section secureRamFuncs, which contains the initialization function for the flash control registers. Close all the inspected files.
Setup PIE Vectors for Interrupts in the TCF File

Next, we will setup all of the PIE interrupt vectors that will be needed for the lab exercises in this module. This will include all of the vectors used in the previous lab exercises. (Note: the PieVect.c file is not used since DSP/BIOS generates the interrupt vector table). 29. Modify the configuration file Lab.tcf to setup the PIE vector for the watchdog interrupt. Click on the plus sign (+) to the left of Scheduling and again on the plus sign (+) to the left of HWI Hardware Interrupt Service Routine Manager. Click the plus sign (+) to the left of PIE INTERRUPTS. Locate the interrupt entry for the watchdog at PIE_INT1_8. Right click, select Properties, and type _WAKEINT_ISR (with a leading underscore) in the function field. Click OK to save. 30. Setup the PIE vector for the ADC interrupt. Locate the interrupt entry for the ADC at PIE_INT1_1. Right click, select Properties, and type _ADCINT1_ISR (with a leading underscore) in the function field. Click OK to save. 31. Setup the PIE vector for the ECAP1 interrupt. Locate the interrupt entry for the ECAP1 at PIE_INT4_1. Right click, select Properties, and type _ECAP1_INT_ISR (with a leading underscore) in the function field. Click OK to save. 32. Setup the PIE vector for the CLA Task 1 interrupt. Locate the interrupt entry for the CLA Task 1 at PIE_INT11_1. Right click, select Properties, and type _CLA1_INT1_ISR (with a leading underscore) in the function field. Click OK to save. Close the configuration window and select YES to save changes to Lab.tcf.
12 - 22
Lab 12: DSP/BIOS
Prepare main() for DSP/BIOS

33. Open Main_12.c and delete the inline assembly code from main() that enables global interrupts. DSP/BIOS will enable global interrupts after main(). 34. In Main_12.c, remove the endless while() loop from the end of main(). When using DSP/BIOS, you must return from main(). In all DSP/BIOS programs, the main() function should contain all one-time user-defined initialization functions. DSP/BIOS will then take-over control of the software execution. Save and close the file.
Configuring DSP/BIOS Global Settings

35. Open the configuration file Lab.tcf and click on the plus sign (+) to the left of System. Right click on Global Settings and select Properties. Confirm that the DSP Speed in MHz (CLKOUT) field is set to 60 so that it matches the processor speed. Click OK to save the value and close the configuration window. This value is used by the CLK manager to calculate the register settings for the on-chip timers and provide the proper time-base for executing CLK functions.
Create a SWI
36. Open Main_12.c and notice that at the end of main() two new functions have been added Cla1Swi() and LedBlink(). We moved part of the CLA1_INT1_ISR() routine from DefaultIsr_12.c to this space in Main_12.c. 37. Open DefaultIsr_12.c and locate the CLA1_INT1_ISR() routine. The entire contents of the CLA1_INT1_ISR() routine was moved to the Cla1Swi() function in Main_12.c with the following exceptions: The instruction used to acknowledge the PIE group interrupt The GPIO pin (LED) toggle code
Comment: In almost all appplications, the PIE group acknowledge code is left in the HWI (rather than move it to a SWI). This allows other interrupts to occur on that PIE group even if the SWI has not yet executed. On the other hand, we are leaving the GPIO toggle code in the HWI just as an example. It illustrates that you can post a SWI and also do additional operations in the HWI. DSP/BIOS is extremely flexible! 38. Delete the interrupt key word from the CLA1_INT1_ISR. The interrupt keyword is not used when a HWI is under DSP/BIOS control. A HWI is under DSP/BIOS control when it uses any DSP/BIOS functionality, such as posting a SWI, or calling any DSP/BIOS function or macro.
Post a SWI
39. Still in DefaultIsr_12.c add the following SWI_post to the CLA1_INT1_ISR(), just after the structure used to acknowledge the PIE group:
12 - 23
Lab 12: DSP/BIOS
SWI_post(&CLA1_swi);
// post a SWI
This posts a SWI that will execute the CLA1_swi() code that was moved to the Cla1Swi() function in Main_12.c. In other words, the CLA1 interrupt still executes the same code as before. However, most of that code is now in a posted SWI that DSP/BIOS will execute according to the specified scheduling priorities. Save and close the modified files.
Add the SWI to the TCF File

40. In the configuration file Lab.tcf we need to add and setup the Cla1Swi() SWI. Open Lab.tcf and click on the plus sign (+) to the left of Scheduling and again on the plus sign (+) to the left of SWI Software Interrupt Manager. 41. Right click on SWI Software Interrupt Manager and select Insert SWI. Rename SWI0 to CLA1_swi and click OK. This is just an arbitrary name. We want to differentiate the Cla1Swi() function itself (which is nothing but an ordinary C function) from the DSP/BIOS SWI object which we are calling CLA1_swi. 42. Select the Properties for CLA1_swi and type _Cla1Swi (with a leading underscore) in the function field. Click OK. This tells DSP/BIOS that it should run the function Cla1Swi() when it executes the CLA1_swi SWI. 43. We need to have the PIE for the CLA Task 1 interrupt use the dispatcher. The dispatcher will automatically perform the context save and restore, and allow the DSP/BIOS scheduler to have insight into the ISR. You may recall from an earlier lab that the CLA Task 1 interrupt is located at PIE_INT11_1. Click on the plus sign (+) to the left of HWI Hardware Interrupt Service Routine Manager. Click the plus sign (+) to the left of PIE INTERRUPTS. Locate the interrupt entry for the CLA Task 1: PIE_INT11_1. Right click, select Properties, and select the Dispatcher tab. Check the Use Dispatcher box and select OK. Close the configuration file and click YES to save changes.
Add a Periodic Function

Recall that an instruction was used in the CLA1_INT1_ISR to toggle the LED on the ControlCARD. This instruction has been moved into a periodic function that will toggle the LED at the same rate. 44. Open DefaultIsr_12.c and locate the CLA1_INT1_ISR routine. Notice that the instruction used to toggle the LED was moved to the LedBlink() function in Main_12.c:
GpioDataRegs.GPBTOGGLE.bit.GPIO34 = 1; // Toggle the pin
Also, the code used to implement the interval counter for the LED toggle (i.e., the GPIO32_count++ loop), and the declaration of the GPIO32_count itself from the beginning of CLA1_INT1_ISR() have been deleted. These are no longer needed, as DSP/BIOS will implement the interval counter for us in the periodic function configuration (next step in the lab). Close the inspected files.
12 - 24
Lab 12: DSP/BIOS
45. In the configuration file Lab.tcf we need to add and setup the LedBlink_PRD. Open Lab.tcf and click on the plus sign (+) to the left of Scheduling. Right click on PRD Periodic Function Manger and select Insert PRD. Rename PRD0 to LedBlink_PRD and click OK. Select the Properties for LedBlink_PRD and type _LedBlink (with a leading underscore) in the function field. This tells DSP/BIOS to run the LedBlink() function when it executes the LedBlink_PRD periodic function object. Next, in the period (ticks) field type 500. The default DSP/BIOS system timer increments every 1 millisecond, so what we are doing is telling the DSP/BIOS scheduler to schedule the LedBlink() function to execute every 500 milliseconds. A PRD object is just a special type of SWI which gets scheduled periodically and runs in the context of the SWI level at a specified SWI priority. Click OK. Close the configuration file and click YES to save changes.
DSP/BIOS Real-time Analysis Tools

The DSP/BIOS analysis tools complement the CCS environment by enabling real-time program analysis of a DSP/BIOS application. You can visually monitor an MCU application as it runs with essentially no impact on the applications real-time performance. In CCS, the DSP/BIOS realt-time analysis (RTA) tools are found on the DSP/BIOS menu. Unlike traditional debugging, which is external to the executing program, DSP/BIOS program analysis requires that the target program be instrumented with analysis code. By using DSP/BIOS APIs and objects, developers automatically instrument the target for capturing and uploading real-time information to CCS using these tools. 46. In the next few steps the Log Event Manager will be setup to record the occurrence of an event in real-time while the program executes. We will be using LOG_printf() to write to a log buffer. The LOG_printf() function is a very efficient means of sending a message from the code to the CCS display. Unlike an ordinary C-language printf(), which can consume several hundred CPU cycles to format the data on the MCU before transmission to the CCS host PC, a LOG_printf() transmits the raw data to the host. The host then formats the data and displays it in CCS. This consumes only 10s of cycles rather than 100s of cycles. In Main_12.c notice the following code at the top of the LedBlink() function just before the instruction used to toggle the LED:
static Uint16 LedSwiCount=0; // used for LOG_printf
/*** Using LOG_printf() to write to a log buffer ***/ LOG_printf(&trace, "LedSwiCount = %u", LedSwiCount++);
Close the file. 47. In the configuration file Lab.tcf we need to add and setup the trace buffer. Open Lab.tcf and click on the plus sign (+) to the left of Instrumentation and again on the plus sign (+) to the left of LOG Event Log Manager.
12 - 25
Lab 12: DSP/BIOS
48. Right click on LOG Event Log Manager and select Insert LOG. Rename LOG0 to trace and click OK. 49. Select the Properties for trace and confirm that the logtype is set to circular and the datatype is set to printf. Click OK. Close the configuration file and click YES to save changes.
Build Lab.out
50. At this point we need to build the project, but not have CCS automatically load it since CCS cannot load code into the flash (the flash must be programmed)! On the menu bar click: Option Customize and select the Program/Project CIO tab and confirm that the Load Program After Build is unchecked. Next select the Debug Properties tab and confirm that the Step over functions without debug information when source stepping is unchecked. Then click OK. 51. Click the Build button to generate Lab.out.
CCS Flash Plug-in

52. Open the Flash Plug-in tool by clicking: Tools F28xx On-Chip Flash Programmer
53. A Clock Configuration window may open. If needed, in the Clock Configuration window set OSCCLK (MHz): to 10, DIVSEL: to /2, and PLLCR Value: to 12. Then click OK. In the next Flash Programmer Settings window confirm that the selected DSP device to program is F28035 and all options have been checked. Click OK. 54. The CCS Flash Programmer uses the Piccolo 10 MHz internal oscillator as the device clock during programming. Confirm the Clock Configuration in the upper left corner has the OSCCLK set to 10 MHz, the DIVSEL set to /2, and the PLLCR value set to 12. Recall that the PLL is divided by two, which gives a SYSCLKOUT of 60 MHz. 55. Confirm that all boxes are checked in the Erase Sector Selection area of the plug-in window. We want to erase all the flash sectors. 56. We will not be using the plug-in to program the Code Security Password. Do not modify the Code Security Password fields. They should remain as all 0xFFFF. 57. In the Operation block, notice that the COFF file to Program/Verify field automatically defaults to the current .out file. Check to be sure that Erase, Program, Verify is selected. We will be using the default wait states, as shown on the slide in this module. The selection for wait-states only affects the verify step, and makes little noticeable difference even if you reduce the wait-states.
12 - 26
Lab 12: DSP/BIOS
58. Click Execute Operation to program the flash memory. Watch the programming status update in the plug-in window. 59. After successfully programming the flash memory, close the programmer window.
Running the Code Using CCS

60. In order to effectively debug with CCS, we need to load the symbolic debug information (e.g., symbol and label addresses, source file links, etc.) so that CCS knows where everything is in your code. Click: File Load Symbols Load Symbols Only
and select Lab12.out in the Debug folder. 61. Reset the CPU. The program counter should now be at 0x3FF8A1, which is the start of the bootloader in the Boot ROM. 62. Under GEL on the menu bar click: EMU Boot Mode Select EMU_BOOT_FLASH. This has the debugger load values into EMU_KEY and EMU_BMODE so that the bootloader will jump to "FLASH" at 0x3F7FF6. 63. Single-Step <F11> through the bootloader code until you arrive at the beginning of the codestart section in the CodeStartBranch.asm file. (Be patient, it will take about 125 single-steps). Notice that we have placed some code in CodeStartBranch.asm to give an option to first disable the watchdog, if selected. 64. Step a few more times until you reach the start of the C-compiler initialization routine at the symbol _c_int00. 65. Now do Debug Go Main. The code should stop at the beginning of your main() routine. If you got to that point succesfully, it confirms that the flash has been programmed properly, that the bootloader is properly configured for jump to flash mode, and that the codestart section has been linked to the proper address. 66. You can now RUN the CPU, and you should observe the LED on the ControlCARD blinking. Try resetting the CPU, select the EMU_BOOT_FLASH boot mode, and then hitting RUN (without doing all the stepping and the Go Main procedure). The LED should be blinking again.
Run the Code Real-time Analysis Tools

It will be interesting to investigate the CPU computational burden of the the different pieces of DSP/BIOS real-time analysis tools that we will be using in this lab exercise. The CPU Load Graph feature of DSP/BIOS will provide a quick and easy method for doing this. We will be tabulating these results in the table that follows at various steps throughout the remainder of this lab.
12 - 27
Lab 12: DSP/BIOS
Table 12-1: CPU Computational Burden Results

Case # 1 Description CPU Load %
CLA processing handled in SWI. LED blink handled in PRD. RTA Global Host Enable disabled. Case #1 + LOG_printf in SWI. Case #2 + RTA SWI Logging enabled. Case #3 + RTA SWI Accumulators enabled.
2 3 4
67. Open the RTA Control Panel by clicking DSP/BIOS RTA Control Panel. Uncheck ALL of the boxes. This disables most of the realtime analysis tools. We will selectively enable them in the lab. 68. Open the CPU Load Graph by clicking DSP/BIOS CPU Load Graph. The CPU load graph displays the percentage of available CPU computing horsepower that the application is consuming. The CPU may be running ISRs, software interrupts, periodic functions, performing I/O with the host, or running any user routine. When the CPU is not executing user code, it will be idle (in the DSP/BIOS idle thread). 69. Record the value shown in the CPU Load Graph under Case #1 in Table 12-1. 70. Open the Message Log. On the menu bar, click: DSP/BIOS Message Log
The message log dialog box is displaying the commanded LOG_printf() output, i.e. the number of times (count value) that the LedSwi() has executed. 71. Verify that all the check boxes in the RTA Control Panel window are still unchecked. Then, check the box marked Global Host Enable. This is the main control switch for most of the RTA tools. We will be selectively enabling the rest of the check boxes in this portion of the exercise. 72. Record the value shown in the CPU Load Graph under Case #2 in Table 12-1. 73. Open the Execution Graph. On the menu bar, click: DSP/BIOS Execution Graph
12 - 28
Lab 12: DSP/BIOS
Presently, the execution graph is not displaying anything. This is because we have it disabled in the RTA Control Panel. In the RTA Control Panel, check the top four boxes to enable logging of all event types to the execution graph. Notice that the Execution Graph is now displaying information about the execution threads being taken by your software. This graph is not based on time, but the activity of events (i.e. when an event happens, such as a SWI or periodic function begins execution). Notice that the execution graph simply records DSP/BIOS CLK events along with other system events (the DSP/BIOS clock periodically triggers the DSP/BIOS scheduler). As a result, the time scale on the execution graph is not linear. The logging of events to the execution graph consumes CPU cycles, which is why the CPU Load Graph jumped as you enabled logging. 74. Record the value shown in the CPU Load Graph under Case #3 in Table 12-1. 75. Open the Statistics View window. On the menu bar, click: DSP/BIOS Statistics View
Presently, the statistics view window is not changing with the exception of the statistics for the IDL_busyObj row (i.e., the idle loop). This is because we have it disabled in the RTA Control Panel. In the RTA Control Panel, check the next five boxes (i.e., those with the word Accumulator in their description) to enable logging of statistics to the statistics view window. The logging of statistics consumes CPU cycles, which is why the CPU Load Graph jumped as you enabled logging. 76. Record the value shown in the CPU Load Graph under Case #4 in Table 12-1. 77. Table 12-1 should now be completely filled in. Think about the results. Note: In this lab exercise only the basic features of DSP/BIOS and the real-time analysis tools have been used. For more information and details, please refer to the DSP/BIOS users manuals and other DSP/BIOS related training.
Running the Code Stand-alone Operation (No Emulator)

78. Close Code Composer Studio. 79. Disconnect the USB cable (emulator) from the Docking Station (i.e. remove power from the ControlCARD). 80. Re-connect the USB cable to the Docking Station to power the ControlCARD. The LED should be blinking, showing that the code is now running from flash memory. End of Exercise
12 - 29
Lab 12: DSP/BIOS
Lab 12 Reference: Programming the Flash
Flash Memory Section Blocks

base = 0x3E 8000
FLASH
len = 0xFF80 space = code
Lab_12.cmd
SECTIONS {
0x3F 7F80
CSM_RSVD
len = 0x76 space = code
codestart csm_rsvd }
:> BEGIN_FLASH, PAGE = 0 :> CSM_RSVD, PAGE = 0
passwords :> PASSWORDS, PAGE = 0
0x3F 7FF6
BEGIN_FLASH
0x3F 7FF8
PASSWORDS
BIOS Startup Sequence from Flash Memory

BIOS code Sections _c_int00 BIOS_reset( ) BIOS_init( ) main ( ) BIOS_start( )
7 5
0x3E 8000
FLASH (64Kw)
4
0x3F 7FF6
LB _c_int00 Passwords (8w)
IDL_run( ) rts2800_ml.lib
6
0x3F E000

0x3F F8A1
{SCAN GPIO} BROM vector (32w) 0x3F F8A1 0x3F FFC0

1
user code sections main ( ) { return; }
RESET
12 - 30
Lab 12: DSP/BIOS
Table 12-2: CPU Computational Burden Results (Solution)

Case # 1 Description CPU Load %
CLA processing handled in SWI. LED blink handled in PRD. RTA Global Host Enable disabled. Case #1 + LOG_printf in SWI. Case #2 + RTA SWI Logging enabled. Case #3 + RTA SWI Accumulators enabled.
27.5
2 3 4
27.5 37.0 48.6
12 - 31
Lab 12: DSP/BIOS
12 - 32
Development Support
Introduction
This module contains various references to support the development process.
Learning Objectives
Learning Objectives
TI Workshops Download Site Signal Processing Libraries TI Development Tools Additional Resources
Internet Product Information Center
C2000 Piccolo Workshop - Development Support
13 - 1
Module Topics
Module Topics
Development Support ..............................................................................................................................13-1 Module Topics........................................................................................................................................13-2 TI Support Resources.............................................................................................................................13-3 C28x Signal Processing Libraries......................................................................................................13-3 Experimenters Kits...........................................................................................................................13-4 F28335 Peripheral Explorer Kit.........................................................................................................13-5 C2000 ControlCARD Application Kits.............................................................................................13-5 Product Information Resources .........................................................................................................13-6
13 - 2
TI Support Resources
TI Workshops Download Site

http://www.tiworkshop.com/survey/downloadsort.asp
Login Name: c28xmdw Password: ttoc28
C28x Signal Processing Libraries
C2000 Signal Processing Libraries

Signal Processing Libraries & Applications Software ACI3-1: Control with Constant V/Hz ACI3-3: Sensored Indirect Flux Vector Control ACI3-3: Sensored Indirect Flux Vector Control (simulation) ACI3-4: Sensorless Direct Flux Vector Control ACI3-4: Sensorless Direct Flux Vector Control (simulation) PMSM3-1: Sensored Field Oriented Control using QEP PMSM3-2: Sensorless Field Oriented Control PMSM3-3: Sensored Field Oriented Control using Resolver PMSM3-4: Sensored Position Control using QEP BLDC3-1: Sensored Trapezoidal Control using Hall Sensors BLDC3-2: Sensorless Trapezoidal Drive DCMOTOR: Speed & Position Control using QEP without Index Digital Motor Control Library (F/C280x) Communications Driver Library DSP Fast Fourier Transform (FFT) Library DSP Filter Library DSP Fixed-Point Math Library DSP IQ Math Library DSP Signal Generator Library DSP Software Test Bench (STB) Library C28x FPU Fast RTS Library DSP2803x C/C++ Header Files and Peripheral Examples Literature # SPRC194 SPRC207 SPRC208 SPRC195 SPRC209 SPRC210 SPRC197 SPRC211 SPRC212 SPRC213 SPRC196 SPRC214 SPRC215 SPRC183 SPRC081 SPRC082 SPRC085 SPRC087 SPRC083 SPRC084 SPRC664 SPRC892
Available from TI Website
http://www.ti.com/c2000
13 - 3
Experimenters Kits
C2000 Experimenters Kits

F28027, F28035, F2808, F28335
Experimenter Kits include

F28027, F28035, F2808 or F28335 ControlCARD USB docking station C2000 Applications Software CD with example code and full hardware details Code Composer Studio v3.3 with code size limit of 32KB
TMDXDOCK28027
TMDXDOCK28035
Docking station features

Access to ControlCARD signals Breadboard areas Onboard USB JTAG Emulation
JTAG emulator not required
TMDSDOCK2808
Available through TI authorized distributors and the TI eStore

TMDSDOCK28335
C2834x Experimenters Kits

C28343, C28346
Experimenter Kits include

C2834x ControlCARD Docking station C2000 Applications Software CD with example code and full hardware details Code Composer Studio v3.3 with code size limit of 32KB 5V power supply
TMDXDOCK28343
Docking station features

Access to ControlCARD signals Breadboard areas
JTAG emulator required sold separately
TMDXDOCK28346-168
Available through TI authorized distributors and the TI eStore
13 - 4
F28335 Peripheral Explorer Kit
F28335 Peripheral Explorer Kit

Experimenter Kit includes
F28335 ControlCARD Peripheral Explorer baseboard C2000 Applications Software CD with example code and full hardware details Code Composer Studio v3.3 with code size limit of 32KB 5V DC power supply
Peripheral Explorer features

ADC input variable resistors GPIO hex encoder & push buttons eCAP infrared sensor GPIO LEDs, I2C & CAN connection Analog I/O (AIC+McBSP)
TMDSPREX28335
JTAG emulator required sold separately Available through TI authorized distributors and the TI eStore
C2000 ControlCARD Application Kits
C2000 ControlCARD Application Kits

Digital Power Experimenters Kit
Kits includes
ControlCARD and application specific baseboard Full version of Code Composer Studio v3.3 with 32KB code size limit
Digital Power Developers Kit Resonant DC/DC Developers Kit Renewable Energy Developers Kit AC/DC Developers Kit Dual Motor Control and PFC Developers Kit
Software download includes

Complete schematics, BOM, gerber files, and source code for board and all software Quickstart demonstration GUI for quick and easy access to all board features Fully documented software specific to each kit and application
See www.ti.com/c2000 for more details Available through TI authorized distributors and the TI eStore
13 - 5
Product Information Resources
For More Information . . .

Internet
Website: http://www.ti.com FAQ: http://www-k.ext.ti.com/sc/technical_support/knowledgebase.htm Device information my.ti.com Application notes News and events Technical documentation Training Enroll in Technical Training: http://www.ti.com/sc/training
USA - Product Information Center (PIC)

Phone: 800-477-8924 or 972-644-5580 Email: [email protected] Information and support for all TI Semiconductor products/tools Submit suggestions and errata for tools, silicon and documents
European Product Information Center (EPIC)

Web: http://www-k.ext.ti.com/sc/technical_support/pic/euro.htm Phone: Language
Belgium (English) France Germany Israel (English) Italy Netherlands (English) Spain Sweden (English) United Kingdom Finland (English)
Number
+32 (0) 27 45 55 32 +33 (0) 1 30 70 11 64 +49 (0) 8161 80 33 11 1800 949 0107 (free phone) 800 79 11 37 (free phone) +31 (0) 546 87 95 45 +34 902 35 40 28 +46 (0) 8587 555 22 +44 (0) 1604 66 33 99 +358(0) 9 25 17 39 48
Fax: All Languages Email: [email protected]
+49 (0) 8161 80 2045
Literature, Sample Requests and Analog EVM Ordering Information, Technical and Design support for all Catalog TI Semiconductor products/tools Submit suggestions and errata for tools, silicon and documents
13 - 6
Appendix A Experimenters Kit
C2000 Piccolo Workshop - Appendix A -Experimenter's Kit
A-1
Module Topics
Module Topics
Appendix A Experimenters Kit ...........................................................................................................A-1 Module Topics......................................................................................................................................... A-2 F28035 ControlCARD ............................................................................................................................ A-3 F28035 PCB Outline (Top View).......................................................................................................A-3 LD1 / LD2 / LD3................................................................................................................................A-3 SW1 ....................................................................................................................................................A-3 SW2 ....................................................................................................................................................A-4 SW3 ....................................................................................................................................................A-4 F28335 ControlCARD ............................................................................................................................ A-5 F28335 PCB Outline (Top View).......................................................................................................A-5 LD1 / LD2 / LD3................................................................................................................................A-5 Docking Station....................................................................................................................................... A-6 SW1 / LD1..........................................................................................................................................A-6 JP1 / JP2 .............................................................................................................................................A-6 J1 / J2 /J3 / J8 / J9...............................................................................................................................A-6 F2833x Boot Mode Selection .............................................................................................................A-7 F280xx Boot Mode Selection .............................................................................................................A-7 J3 DB-9 to 4-Pin Header Cable .......................................................................................................A-8
A-2
F28035 ControlCARD
F28035 ControlCARD
F28035 PCB Outline (Top View)
SW1
LD1 LD2 LD3
SW3
SW2
LD1 / LD2 / LD3
SW1
A-3
F28035 ControlCARD
SW2
SW3
A-4
F28335 ControlCARD
F28335 ControlCARD
F28335 PCB Outline (Top View)
LD3 LD2 LD1
LD1 / LD2 / LD3
A-5
Docking Station
Docking Station
2833x Boot J9 J2 J8
J3
JP2 SW1 JP1
280xx Boot
J1
LD1
SW1 / LD1
SW1 USB: Power from USB; ON Power from JP1 LD1 Power-On indicator
JP1 / JP2
JP1 5.0 V power supply input JP2 USB JTAG emulation port
J1 / J2 /J3 / J8 / J9
J1 ControlCARD 100-pin DIMM socket J2 JTAG header connector J3 UART communications header connector J8 Internal emulation enable/disable jumper (NO jumper for internal emulation) J9 User virtual COM port to C2000 device (Note: ControlCARD would need to be modified to disconnect the C2000 UART connection from header J3)
A-6
Docking Station
Note: The internal emulation logic on the Docking Station routes through the FT2232 USB device. By default this device enables the USB connection to perform JTAG communication and in parallel create a virtual serial port (SCI/UART). As shipped, the C2000 device is not connected to the virtual COM port and is instead connected to J3.
F2833x Boot Mode Selection
F280xx Boot Mode Selection
A-7
Docking Station
J3 DB-9 to 4-Pin Header Cable

Note: This cable is NOT included with the Experimenters Kit and is only shown for reference.
DB-9 Male
Pin-Out Table for Both Ends of the Cable: DB-9 female SIL 0.1" female Pin# Pin# -------------------------------------------------------------2 (black) 1 (TX) 3 (red) 4 (RX) 5 (bare wire) 3 (GND) Note: pin 2 on SIL is a no-connect
A-8
Appendix B Addressing Modes

Introduction
Appendix B will describe the data addressing modes on the C28x. Immediate addressing allows for constant expressions which are especially useful in the initialization process. Indirect addressing uses auxiliary registers as pointers for accessing organized data in arrays. Direct addressing is used to access general purpose memory. Techniques for managing data pages, relevant to direct addressing will be covered as well. Finally, register addressing allows for interchange between CPU registers.
Learning Objectives
Learning Objectives
Explain .sect and .usect assembly directives Explain assembly addressing modes Understand instruction formats Describe options for each addressing mode
C2000 Piccolo Workshop - Appendix B - Addressing Modes
B-1
Module Topics
Module Topics
Appendix B Addressing Modes .............................................................................................................B-1 Module Topics......................................................................................................................................... B-2 Labels, Mnemonics and Assembly Directives ......................................................................................... B-3 Addressing Modes................................................................................................................................... B-4 Instruction Formats ................................................................................................................................ B-5 Register Addressing ................................................................................................................................ B-6 Immediate Addressing............................................................................................................................. B-7 Direct Addressing ................................................................................................................................... B-8 Indirect Addressing............................................................................................................................... B-10 Review................................................................................................................................................... B-13 Exercise B.........................................................................................................................................B-14 Lab B: Addressing................................................................................................................................. B-15 OPTIONAL Lab B-C: Array Initialization in C .................................................................................... B-17 Solutions................................................................................................................................................ B-18
B-2
Labels, Mnemonics and Assembly Directives
Labels, Mnemonics and Assembly Directives
Labels and Mnemonics

Labels
Optional for all assembly instructions and most assembler directives Must begin in column 1 The : is not treated as part of the label name Used as pointers to memory or instructions .ref start .sect vectors ;make reset vector address 'start' reset: .long start .def start count .set 9 ;create an array x of 10 words x .usect mydata, 10 .sect C28OBJ MOV next: MOVL MOV loop: MOV BANZ bump: ADD SB start: code ;operate in C28x mode ACC,#1 XAR1,#x AR2,#count *XAR1++,AL loop,AR2-ACC,#1 next,UNC
Mnemonics
Lines of instructions Use upper or lower case Become components of program memory
Assembly Directives
Begin with a period (.) and are lower case Used by the linker to locate code and data into specified sections Directives allow you to: Define a label as global Reserve space in memory for un-initialized variables Initialized memory .ref start .sect vectors ;make reset vector address 'start' reset: .long start .def start count .set 9 ; create an array x of 10 words x .usect mydata, 10 .sect C28OBJ MOV next: MOVL MOV loop: MOV BANZ bump: ADD SB start: code ;operate in C28x mode ACC,#1 XAR1,#x AR2,#count *XAR1++,AL loop,AR2-ACC,#1 next,UNC
Directives
initialized section .sect name
used for code or constants
uninitialized section label .usect name,5

used for variables
B-3
Addressing Modes
Addressing Modes
Addressing Modes
Mode
(register) (constant) (paged) (pointer)
Symbol Purpose Operate between Registers # @ * Constants and Initialization General-purpose access to data Support for pointers access arrays, lists, tables
Register Immediate Direct Indirect
Four main categories of addressing modes are available on the C28x. Register addressing mode allows interchange between all CPU registers, convenient for solving intricate equations. Immediate addressing is helpful for expressing constants easily. Direct addressing mode allows information in memory to be accessed. Indirect addressing allows pointer support via dedicated auxiliary registers, and includes the ability to index, or increment through a structure. The C28x supports a true software stack, desirable for supporting the needs of the C language and other structured programming environments, and presents a stack-relative addressing mode for efficiently accessing elements from the stack. Paged direct addressing offers general-purpose single cycle memory access, but restricts the user to working in any single desired block of memory at one time.
B-4
Instruction Formats
Instruction Formats
Instruction Formats
INSTR dst ,src
INSTR INSTR INSTR INSTR INSTR REG REG,#imm REG,mem mem,REG mem,#imm
Example
NEG MOV ADD SUB MOV AL ACC,#1 AL,@x AL,@AR0 *XAR0++,#25
What is a REG? 16-bit Access = AR0 through AR7, AH, AL, PH, PL, T and SP 32-bit Access = XAR0 through XAR7, ACC, P, XT What is an #imm? an immediate constant stored in the instruction What is a mem? A directly or indirectly addressed operand from data memory Or, one of the registers from REG! loc16 or loc32 (for 16-bit or 32-bit data access)
The C28x follows a convention that uses instruction, destination, then source operand order (INSTR dst, src). Several general formats exist to allow modification of memory or registers based on constants, memory, or register inputs. Different modes are identifiable by their leading characters (# for immediate, * for indirect, and @ for direct). Note that registers or data memory can be selected as a mem value.
B-5
Register Addressing
Register Addressing
Register Addressing
32-bit Registers
XAR0 XAR7 AR0 AR7 AH AL
ACC PL
P T
XT TL DP SP
16-bit Registers
PH
Allows for efficient register to register operation 16-bit and 32-bit Register Address modes Reduces code overhead, memory accesses, and memory overhead
Register addressing allows the exchange of values between registers, and with certain instructions can be used in conjunction with other addressing modes, yielding a more efficient instruction set. Remember that any mem field allows the use of a register as the operand, and that no special character (such as @, *, or #) need be used to specify the register mode.
Register Addressing Example
Format Instruction
MOV Ax,loc16 MOV AH,@AL
MOVL loc32,ACC MOVL @XT,ACC
Format Instruction
MOV loc16,Ax,COND MOV @AR1,AL,GT

User Guide & Dis-assembler use @ for second register
B-6
Immediate Addressing
Immediate Addressing
Immediate Addressing #
one word instruction
OPCODE
8-bit OPERAND
two word instruction
OPCODE 16-bit OPERAND
Fixed value part of program memory instruction Supports short (8-bit) and long (16-bit) immediate constants Long immediate can include a shift Used to initialize registers, and operate with constants
Immediate addressing allows the user to specify a constant within an instruction mnemonic. Short immediate are single word, and execute in a single cycle. Long (16-bit) immediate allow full sized values, which become two-word instructions - yet execute in a single instruction cycle.
Immediate Addressing Example

Short Immediate, 1 Word (ANDB) Long Immediate, 2 Words (AND)
ANDB Ax,#8Bit ANDB Ax #8Bit
AND AND
loc16,#16Bit loc16 #16Bit Ax,loc16,#16Bit Ax loc16
AND automatically replaced by ANDB if IMM value is 8 bits or less
AND AND
#16Bit AND AND ACC,#16Bit,<<0-16 ACC shift
#16Bit
B-7
Direct Addressing
Direct Addressing
Direct addressing allows for access to the full 4-Meg words space in 64 word page groups. As such, a 16-bit Data Page register is used to extend the 6-bit local address in the instruction word. Programmers should note that poor DP management is a key source of programming errors. Paged direct addressing is fast and reliable if the above considerations are followed. The watch operation, recommended for use whenever debugging, extracts the data page and displays it as the base address currently in use for direct addressing.
Direct Addressing @
Data Page 00 0000 0000 0000 00 00 0000 0000 0000 00 00 0000 0000 0000 01 00 0000 0000 0000 01 00 0000 0000 0000 10 00 0000 0000 0000 10 11 1111 1111 1111 11 11 1111 1111 1111 11 Offset 00 0000 11 1111 00 0000 11 1111 00 0000 11 1111 00 0000 11 1111 Data Memory Page 0: 00 0000 00 003F Page 1: 00 0040 00 007F Page 2: 00 0080 00 00BF
Page 65,535: 3F FFC0 3F FFFF
Data memory space divided into 65,536 pages with 64 words on each page Data page pointer (DP) used to select active page 16-bit DP is concatenated with a 6-bit offset from the instruction to generate an absolute 22-bit address Access data on a given page in any order
B-8
Direct Addressing
Direct Addressing Example

Z=X+Y
x .usect .sect MOVW MOV ADD MOV samp,3 code DP,#x AL,@x AL,@y @z, AL 0 0 0 1 F F 0000 0000 0000 0001 1111 1111
DP offset
Data Memory address data Page7[00] 0001C0 0001 64 ... Page7[3D] x: 0001FD 1000 Page7[3E] y: 0001FE 0500 Page7[3F] z: 0001FF 1500 Accumulator - - - - - - - MOV AL,@x 0 0 0 0 1 0 0 0 ADD AL,@y 0 0 0 0 1 5 0 0
MOV @z,AL
DP=0007
variations: MOVW DP,#imm ;2W, 16-bit (4 Meg) MOVZ DP,#imm ;1W, 10-bit (64K) MOV DP,#imm ;DP(15:10) unchanged
Direct Addressing Caveats

(X and Y not on the same page)
Z=X+Y
DP offset
0000 0000 0000 0001 1111 1111 0000 0000 0000 0010 0000 0000 DP=0007 0 0 0 7 0 0 0 7 0 0 0 7 Accumulator - - - - - - - 0 0 0 0 1 0 0 0 0 0 0 0 1 0 0 1
expecting 1500
Page7[00] ... Page7[3F] x: 0001FF 1000 Page8[00] y: 000200 0500
Data Memory address data 0001C0 0001
x .usect .sect MOVW MOV ADD MOV samp,3 code DP,#x AL,@x AL,@y @z, AL
Solution: Group and block variables in ASM file:
x y z
.usect samp,3,1 ;Force all locations to same data .set x+1 ;page (1st hole, else linker error) .set x+2 ;Assign vars within block
B-9
Indirect Addressing
Indirect Addressing
Indirect Addressing *
Data Memory XAR0 XAR1 XAR2 XAR3 XAR4 XAR5 XAR6 XAR7 ARAU
Auxiliary Registers (XARn) used to access full data memory space Address Register Arithmetic Unit (ARAU) used to modify the XARn Access data from arrays anywhere in data memory in an orderly fashion
Any of eight hardware pointers (ARs) may be employed to access values from the first 64K of data memory. Auto-increment or decrement is supported at no additional cycle cost. XAR register formats offer larger 32-bit widths, allowing them to access across the full 4-Giga words data space.
Indirect Addressing Modes

Auto-increment / decrement: *XARn++, *--XARn
Post-increment or Pre-decrement
Offset: *+XARn[AR0 or AR1], *+XARn[3bit]

Offset by 16-bit AR0 or AR1, or 3-bit constant
Stack Relative: *-SP[6bit]

Index by 6-bit offset (optimal for C)
Immediate Direct: *(0:16bit)

Access low 64K
Circular: *AR6%++
AR1(7:0) is buffer size XAR6 is current address
B - 10
Indirect Addressing
Indirect Addressing Example

Autoincrement
y = xn
n =0
x y
.usect samp,6 .set (x + 5) .sect code MOVL XAR2,#x MOV ACC,*XAR2++ ADD ACC,*XAR2++ ADD ACC,*XAR2++ ADD ACC,*XAR2++ ADD ACC,*XAR2++ MOV *(0:y),AL
Data x x0 x1 x2 x3 x4 XAR2
y
*(0:16bit) - 16 bit label - must be in lower 64K - 2 word instruction
Fast, efficient access to arrays, lists, tables, etc.

Indexed addressing offers the ability to select operands from within an array without modification to the base pointer. Stack-based operations are handled with a 16-bit Stack Pointer register, which operates over the base 64K of data memory. It offers 6-bit non-destructive indexing to access larger stack-based arrays efficiently.

Offset
Data
x[2] = x[1] + x[3]
XAR2 [3]
x0 x1 x2 x3 x4
x .usect .samp,5 .sect .code MOVL MOV MOV MOV ADD MOV XAR2,#x AR0,#1 AR1,#3 ACC,*+XAR2[AR0] ACC,*+XAR2[AR1] *+XAR2[2],AL
x .usect .samp,5 .sect .code MOVL MOV ADD MOV XAR2,#x ACC,*+XAR2[1] ACC,*+XAR2[3] *+XAR2[2],AL
16 bit offset
3 bit offset
Allows offset into arrays with fixed base pointer
B - 11
Indirect Addressing

Stack Relative
Data Memory
x2 = x1 + x3
- SP -
0 1 2 0 x3 3 2 0 0 5 0 x2 ? ? ? ? 0 2 0 0 x1 empty empty
Instr. 3
.sect MOV ADD MOV
.code
AL,*-SP[1] AL,*-SP[3] *-SP[2],AL
Accumulator Instr. 1 0 0 0 0 0 2 0 0
Instr. 2
0 0 0 0 0 3 2 0
Useful for stack based operations

Circular
start of buffer
AAAA AAAA
access pointer
AAAA AAAA 0000 0000 XAR6 (32) AAAA AAAA xxxx xxxx
Buffer Size N Element 0

circular buffer range
(align on 256 word boundary)
AAAA AAAA
AR1 Low (16) end of buffer ---- ---N-1

(AR1 Low is set to buffer size 1)
Element N-1
MAC
P,*AR6%++,*XAR7++
LINKER.CMD SECTIONS { Buf_Mem: align(256) { } . . . }
> RAM
PAGE 1
B - 12
Review
Review
Addressing Range Review

0x000000 0x00003F Stack Addressing SP 64K Direct Addressing DP(16+6) 4M 0x00FFFF Indirect Addressing XARn 4G
0x3FFFFF
0xFFFFFFFF
Data memory can be accessed in numerous ways: Stack Addressing: allows a range to 64K Direct Addressing: Offers a 16-bit DP plus a 6-bit offset, allowing a 4M range Indirect Addressing: Offers the full 4G range
B - 13
Review
Exercise B
Exercise B: Addressing
Given: Address/Data (hex) Fill in the table below Src Mode DP = 4000 100030 0025 100031 0120 100032 ACC DP = 4004 100100 0105 100101 0060 100102 0020 DP DP = 4006 100180 0100 100181 0030 100182 0040 AR1 XAR1 AR2 XAR2
Program MOVW DP,#4000h MOVL XAR1,#100100h MOVL XAR2,#100180h MOV AL,@31h ADD AL,*XAR1++ SUB AL,@30h ADD AL,*XAR1++ MOVW DP,#4006h ADD AL,@1 SUB AL,*XAR1 ADD AL,*XAR2 SUB AL,*+XAR2[1] ADD AL,#32 SUB AL,*+XAR2[2] MOV @32h,AL Dir: Direct; Idr: Indirect
Imm: Immediate; Reg: Register;
In the table above, fill in the values for each of the registers for each of the instructions. Three areas of data memory are displayed at the top of the diagram, showing both their addresses and contents in hexadecimal. Watch out for surprises along the way. First, you should answer the addressing mode for the source operand. Then, fill in the change values as the result of the instruction operation.
B - 14
Lab B: Addressing
Lab B: Addressing
Note: The lab linker command file is based on the F28035 memory map modify as needed, if using a different F28xx device memory map.
Objective
The objective of this lab is to practice and verify the mechanics of addressing. In this process we will expand upon the ASM file from the previous lab to include new functions. Additionally, we learn how to run and observe the operation of code using Code Composer Studio. In this lab, we will initialize the vars arrays allocated in the previous lab with the contents of the const table. How is this best accomplished? Consider the process of loading the first const value into the accumulator and then storing this value to the first vars location, and repeating this process for each of the succeeding values. What forms of addressing could be used for this purpose? Which addressing mode would be best in this case? Why? What problems could arise with using another mode?
Procedure
Copy Files, Create Project File

1. Create a new project called LabB.pjt in C:\C28x\Labs\Appendix\LabB and add LabB.asm and Lab.cmd to it. Check your file list to make sure all the files are Build there. Be sure to setup the Build Options by clicking: Project Options on the menu bar. Select the Linker tab. In the middle of the screen select No Autoinitialization under Autoinit Model:. Enter start in the Code Entry Point (-e): field. Next, select the Compiler tab. Note that Full Symbolic Debug (-g) under Generate Debug Info: is selected. Then select OK to save the Build Options.
Initialize Allocated RAM Array from ROM Initialization Table

2. Edit LabB.asm and modify it to copy table[9] to data[9] using indirect addressing. (Note: data[9] consists of the allocated arrays of data, coeff, and result). Initialize the allocated RAM array from the ROM initialization table: Delete the NOP operations from the code section. Initialize pointers to the beginning of the const and vars arrays. Transfer the first value from const to the vars array. Repeat the process for all values to be initialized.
To perform the copy, consider using a load/store method via the accumulator. Which part of an accumulator (low or high) should be used? Use the following when writing your copy routine: - use AR1 to hold the address of table - use AR2 to hold the address of data
B - 15
Lab B: Addressing
3. It is good practice to trap the end of the program (i.e. use either end: end,UNC or end: B start,UNC). Save your work.
Build and Load

4. Click the Build button and watch the tools run in the build window. Debug as necessary. To open up more space, close any open files or windows that you do not need. 5. Load the output file onto the target. Click: File Load Program
If you wish, right click on the LabB.asm source window and select Mixed Mode to debug using both source and assembly. Note: Code Composer Studio can automatically load the output file after a successful build. On the menu bar click: Option Customize and select the Program Load Options tab, check Load Program After Build, then click OK. 6. Single-step your routine. While single-stepping, it is helpful to see the values located in table[9] and data[9] at the same time. Open two memory windows by using the View Memory button on the vertical toolbar and using the address labels table and data. Setting the properties filed to Hex 16 Bit TI style will give you more viewable data in the window. Additionally, it is useful to watch the CPU registers. Open the CPU registers by using the View Registers CPU Registers. Deselect Allow Docking and move/resize the window as needed. Check to see if the program is working as expected. End of Exercise
B - 16
OPTIONAL Lab B-C: Array Initialization in C
OPTIONAL Lab B-C: Array Initialization in C

Objective
The objective of this lab is to practice and verify the mechanics of initialization using C. Additionally, we learn how to run and observe the operation of C code using Code Composer Studio. In this lab, we will initialize the vars arrays with the contents of the const table.
Procedure
Create Project File

1. In Code Composer Studio create a new project called LabB-C.pjt in C:\C28x\Labs\Appendix\LabB\LabB-C and add LabB-C.c and Lab.cmd to it. Check your file list to make sure all the files are there. Open the Build Options and select the Linker tab. Select the Libraries Category and enter rts2800_ml.lib in the Incl. Libraries (-l): box. Do not setup any other Build Options. The default values will be used. In Appendix Lab D exercise, we will experiment and explore the various build options when working with C.
Initialize Allocated RAM Array from ROM Initialization Table

2. Edit LabB-C.c and modify the main routine to copy table[9] to the allocated arrays of data[4], coeff[4], and result[1]. (Note: data[9] consists of the allocated arrays of data, coeff, and result).
Build and Load

3. Click the Build button and watch the tools run in the build window. Debug as necessary. Note: Have Code Composer Studio automatically load the output file after a successful build. On the menu bar click: Option Customize and select the Program Load Options tab, check Load Program After Build, then click OK. 4. Under Debug on the menu bar click Go Main. Single-step your routine. While single-stepping, it is helpful to see the values located in table[9] and data[9] at the same time. Open two memory windows by using the View Memory button on the vertical toolbar and using the address labels table and data. Setting the properties field to Hex 16 Bit TI style will give you more viewable data in the window. Additionally, you can watch the CPU registers. Open the CPU registers by using the View Registers CPU Registers. Deselect Allow Docking and move/resize the window as needed. Check to see if the program is working as expected. End of Exercise
B - 17
Solutions
Solutions
Exercise B: Addressing - Solution

Given: Address/Data (hex) Fill in the table below Src Mode Imm Imm Imm Dir Idr Dir Idr Imm Dir Idr Idr Idr Imm Idr Dir DP = 4000 100030 0025 100031 0120 100032 ACC DP = 4004 100100 0105 100101 0060 100102 0020 DP 4000 100100 100180 120 225 200 260 4006 290 270 370 340 360 320 100101 100102 DP = 4006 100180 0100 100181 0030 100182 0040 XAR1 XAR2
Program MOVW DP,#4000h MOVL XAR1,#100100h MOVL XAR2,#100180h MOV AL,@31h ADD AL,*XAR1++ SUB AL,@30h ADD AL,*XAR1++ MOVW DP,#4006h ADD AL,@1 SUB AL,*XAR1 ADD AL,*XAR2 SUB AL,*+XAR2[1] ADD AL,#32 SUB AL,*+XAR2[2] MOV @32h,AL Dir: Direct; Idr: Indirect
100180 100180 1001B2 0320
Imm: Immediate; Reg: Register;
B - 18
Appendix C Assembly Programming

Introduction
Appendix C discusses the details of programming in assembly. It shows you how to use different instructions that further utilize the advantage of the architecture data paths. It gives you the ability to analyze the instruction set and pick the best instruction for the application.
Learning Objectives
Learning Objectives
Perform simple program control using branch and conditional codes Write C28x code to perform basic arithmetic Use the multiplier to implement sum-of-products equations Use the RPT instruction (repeat) to optimize loops Use MAC for long sum-of-products Efficiently transfer the contents of one area of memory to another Examine read-modify-write operations
C2000 Piccolo Workshop - Appendix C - Assembly Programming
C-1
Module Topics
Module Topics
Appendix C Assembly Programming ...................................................................................................C-1 Module Topics.........................................................................................................................................C-2 Program Control.....................................................................................................................................C-3 Branches .............................................................................................................................................C-3 Program Control Instructions .............................................................................................................C-4 ALU and Accumulator Operations..........................................................................................................C-6 Simple Math & Shift...........................................................................................................................C-7 Multiplier ................................................................................................................................................C-9 Basic Multiplier ................................................................................................................................C-10 Repeat Instruction.............................................................................................................................C-11 MAC Instruction...............................................................................................................................C-12 Data Move.............................................................................................................................................C-13 Logical Operations ...............................................................................................................................C-15 Byte Operations and Addressing ......................................................................................................C-15 Test and Change Memory Instructions.............................................................................................C-16 Min/Max Operations.........................................................................................................................C-17 Read Modify Write Operations .............................................................................................................C-18 Lab C: Assembly Programming............................................................................................................C-20 OPTIONAL Lab C-C: Sum-of-Products in C........................................................................................C-22
C-2
Program Control
Program Control
The program control logic and program address generation logic work together to provide proper program flow. Normally, the flow of a program is sequential: the CPU executes instructions at consecutive program memory addresses. At times, a discontinuity is required; that is, a program must branch to a nonsequential address and then execute instructions sequentially at that new location. For this purpose, the C28x supports interrupts, branches, calls, returns, and repeats. Proper program flow also requires smooth flow at the instruction level. To meet this need, the C28x has a protected pipeline and an instruction-fetch mechanism that attempts to keep the pipeline full.
Branches
Branch Types and Range

3 Branch Types 0x000000
Program Memory
Short Branch offset +127/128 1-word instruction
Branch offset +/-32K 2-word instruction
Long Branch absolute 4M 2-word instruction
PC 0x3FFFFF
The PC can access the entire 4M words (8M bytes) range. Some branching operations offer 8and 16-bit relative jumps, while long branches, calls, and returns provide a full 22-bit absolute address. Dynamic branching allows a run-time calculated destination. The C28x provides the familiar arithmetic results status bits (Zero, oVerflow, Negative, Carry) plus a Test Control bit which holds the result of a binary test. The states of these bits in various combinations allow a range of signed, unsigned, and binary branching conditions offered.
C-3
Program Control
Program Control Instructions
Program Control - Branches

Function Short Branch Fast Short Branch Fast Branch Absolute Branch Dynamic Branch Branch on AR Branch on compare Condition Code
NEQ EQ GT GEQ LT LEQ HI HIS (C) LO (NC) LOS NOV OV NTC TC UNC NBIO
Instruction SB SBF BF LB LB BAR 8bit,cond 8bit,EQ|NEQ|TC|NTC 16bit,cond 16bit,cond 22bit *XAR7 16bit,ARn,ARn,EQ|NEQ
Cycles T/F Size 7/4 4/4 7/4 4/4 4 4 4/2 4/2 1 1 2 2 2 1 2 2
Fast Relative Branch B
BANZ 16bit,ARn--
Condition flags are set on the prior use of the ALU The assembler will optimize B to SB if possible
Program Control - Call/Return

Function Call Code Cycles Return code Cycles
Call Dynamic Call Interrupt Return
LCR LCR
22bit *XARn
4 4
LRETR LRETR IRET Stack Local Var 22-bit old RPC
4 4 8
More Call variations in the user guide are for code backward compatibility New RPC
LCR Func LRETR RPC Old RPC Ret Addr Func Ret Addr
PC
C-4
Program Control
BANZ Loop Control Example

Auxiliary register used as loop counter Branch if Auxiliary Register not zero Test performed on lower 16-bits of XARx only
y = xn
n =0
Data x x0 x1 x2 x3 x4 XAR2
len x y
.set .usect .set .sect MOVL MOV MOV ADD BANZ MOV
5 samp,6 (x+len) code XAR2,#x AR3,#len-2 AL,*XAR2++ AL,*XAR2++ sum,AR3-*(0:y),AL
y AR3 COUNT
sum:
C-5
ALU and Accumulator Operations
ALU and Accumulator

16/32 data mem, Product (32) 16/32 bit registers
MUX
8/16 Imm
ALU and Barrel Shifter
ST0, ST1
ACC AH (31-16) AL (15-0)

AH.MSB AH.LSB AL.MSB AL.LSB
One of the major components in the execution unit is the Arithmetic-Logical-Unit (ALU). To support the traditional Digital Signal Processing (DSP) operation, the ALU also has the zero cycle barrel shifter and the Accumulator. The enhancement that the C28x has is the additional data paths added form the ALU to all internal CPU registers and data memory. The connection to all internal registers helps the compiler to generate efficient C code. The data path to memory allows the C28x performs single atomic instructions read-modify-write to the memory. The following slides introduce you to various instructions that use the ALU hardware. Word, byte, and long word 32-bit operation are supported.
C-6
Simple Math & Shift
Accumulator - Basic Math Instructions

Format
xxx xxxB xxxL ADD ADDB Ax, #16b ;word Ax, #8b ;byte ACC, #32b ;long ACC, #01234h<<4 AL, #34h
xxx = instruction: MOV, ADD, SUB, ... Ax = AH, or AL Assembler will automatically convert to 1 word instruction. Two word instructions with shift option One word instruction, no shift
Ex
Ax = AH or AL Operations
ACC Operations Variation

MOV ADD SUB MOV ADD SUB MOV MOVH
} }
ACC,loc16<<shift from memory (left shift optional) ACC,#16b<<shift 16-bit constant (left shift optional) ;AL ;AH
loc16,ACC <<shift loc16,ACC <<shift
MOV ADD SUB AND OR XOR AND NOT NEG MOV
Ax, loc16 Ax, loc16 Ax, loc16 Ax, loc16 Ax, loc16 Ax, loc16 Ax,loc16,#16b Ax Ax loc16,Ax
Shift the Accumulator

Shift full ACC
LSL SFR LSL SFR ACC <<shift (1-16) ACC >>shift ACC <<T ACC >>T (0-15) SXM C 31 0 ACC 31 0 ACC 0 LSL
C SFR
Shift AL or AH
LSL LSR ASR LSL LSR ASR AX <<shift AX <<shift AX >>shift AX <<T AX <<T AX >>T
15 0 Ax 15 0 Ax 15 0 Ax
LSL
SXM
ASR
LSR
C-7
32 Bit Shift Operations [ACC]

31 0 ACC
Logical Shift Left Long: LSLL 31 0 ACC
Examples: LSLL ACC, T LSRL ACC, T ASRL ACC, T

Note: T(4:0) are used; other bits are ignored
Logical Shift Right Long: LSRL 31 0 ACC
0 or 1
based on SXM
Arithmetic Shift Right Long: ASRL
C-8
Multiplier
Multiplier
Multiply Unit
XT Register T Register 32x32 Multiply Unit 16x16 MUX Data Mem or Register
Prog Mem (16) or Immed (8,16)
P Register (32)
Shift (PM) ACC (32)
Digital signal processors require many multiply and add math intensive operations. The single cycle multiplier is the second major component in the execution unit. The C28x has the traditional 16-bit-by-16-bit multiplier as previous TI DSP families. In-addition, the C28x has a single cycle 32-bit-by-32-bit multiplier to perform extended precision math operations. The large multiplier allows the C28x to support higher performance control systems requirement while maintaining small or reduce code. The following slides introduce instructions that use the 16-bit-by-16-bit multiplier and multiply and add (MAC) operations. The 32-bit-by-32-bit multiplication will be covered in the appendix.
C-9
Multiplier
Basic Multiplier
Multiplier Instructions
Instruction MOV MPY MPY MPYB MPYB MOV ADD SUB T,loc16 ACC,T,loc16 P,T,loc16 ACC,T,#8bu P,T,#8bu ACC,P ACC,P ACC,P T ACC P ACC P ACC ACC ACC Execution = loc16 = T*loc16 = T*loc16 = T*8bu = T*8bu = P += P -= P Purpose Get first operand For single or first product For n th product Using 8-bit unsigned const Using 8-bit unsigned const Move 1st product<<PM to ACC Add nth product<<PM to ACC Sub nth product<<PM fr. ACC
Instruction MOVP T, loc16 MOVA T, loc16 MOVS T, loc16 MPYA MPYA MPYS P, T, #16b P, T, loc16 P, T, loc16
Execution ACC = P<<PM T = loc16 ACC += P<<PM T = loc16 ACC - = P<<PM T = loc16 ACC += P<<PM ACC += P<<PM ACC - = P<<PM then then then P = T*#16b P = T*loc16 P = T*loc16
Sum-of-Products
Y = A*X1 + B*X2 + C*X3 + D*X4
ZAPA MOV MPY MOVA MPY MOVA MPY MOVA MPY ADDL MOVL T,@X1 P,T,@A T,@X2 P,T,@B T,@X3 P,T,@C T,@X4 P,T,@D ACC,P<<PM @y,ACC ;ACC ;T = ;P = ;T = ;P = ;T = ;P = ;T = ;P = ;ACC = P = OVC = 0 X1 A*X1 X2 ;ACC = A*X1 B*X2 X3 ;ACC = A*X1 + B*X2 C*X3 X4;ACC = A*X1 + B*X2 + C*X3 D*X4 = Y
C - 10
Multiplier
32x32 Long Multiplication

X X Y XO * Y0
Integer long multiplication u(long) = u(long) * u(long) Fraction long multiplication: (long) = (long) * (long)
Y1 Z3
X1 Z2 Z1 Z0
(long) 64 = (long) 32 * (long) 32
Accumulator
P-register
IMPYAL QMPYAL IMACL QMACL
P,XT,loc32 ACC,XT,loc32 P,loc32,*XAR7 P,loc32,*XAR7
P = u(XT)*u(loc32) ACC = (XT)*(loc32) ACC += P; P = u(loc32)*u(loc32) ACC += P; P = (loc32)*(loc32)
Repeat Instruction
Repeat Next: RPT

Options: RPT #8bit RPT loc16 up to 256 iterations location loc16 holds count value Example : int x[5]={0,0,0,0,0}; x .usect samp,5 MOV AR1,#x RPT #4 || MOV *XAR1++,#0
Features: Next instruction iterated N+1 times Saves code space - 1 word Low overhead - 1 cycle Easy to use Non-interruptible Requires use of | | before next line May be nested within BANZ loops
Instruction Cycles RPT BANZ

Refer to User Guide for more repeatable instructions
1 4 .N
C - 11
Multiplier
Single repeat instruction (RPT) is used to reduce code size and speed up many operations in the DSP application. Some of the most popular operations that use the RPT instruction to perform multiple taps digital filters or perform block of data transfer.
MAC Instruction
Sum-of-Products: RPT / MAC

y = xn a n
n =0 19
x y .usect sample,20 .usect result,2 .sect coefficient a0: .word 0x0101 .word 0x0202 .word 0x2020 .sect code SOP: SPM 0 MOVW DP,#y MOVL XAR1,#x MOVL XAR7,#a0 ZAPA RPT #19 || MAC P,*XAR1++,*XAR7++ ADDL ACC,P<<PM MOVL @y,ACC B SOP,UNC
XAR1++
X0 X1 ... X19
XAR7++
Second operand must use XAR7
A0 A1 ... A19
Zero ACC & P Repeat single Dual operand last ADD
MOV ADD
T,loc16 ACC,P
MOVA T,loc16 MPY P,T,loc16
MAC
ACC+=P T=*ARn++ P=T*(*ARn++)
C - 12
Data Move
Data Move
Data Move Instructions

DATA DATA (4G 64K) MOV MOV loc16, *(0:16bit) *(0:16bit), loc16 DATA PGM (4G 4M) PREAD PWRITE loc16 ,*XAR7 *XAR7, loc16
pointer with a 22-bit program memory address
16-bit address concatenated with 16 leading zeros
32-bit address memory location
.sect START: MOVL MOVL RPT || PREAD ... x .usect .sect TBL: .word len .set
.code XAR5,#x XAR7,#TBL #len-1 *XAR5++,*XAR7 .samp,4 .coeff 1,2,3,4 $-TBL
Optimal with RPT (speed and code size) In RPT, non-mem address is autoincremented in PC
Faster than Load / Store, avoids accumulator Allows access to program memory
Conditional Moves
Instruction Execution (if COND is met)
MOV loc16,AX,COND MOVB loc16,#8bit,COND

Instruction
[loc16] [loc16] [loc32]
= AX = 8bit = AX
Execution (if COND is met)
MOVL loc32,ACC,COND If A<B, Then B=A

A B .usect .set .sect MOVW DP, MOV AL, CMP AL, MOV @B, var,2,1 A+1 code #A @A @B AL, LT
Example
Accumulator 0 0 0 0 0 1 2 0 Data Memory 0 1 2 0 A 0 3 2 0 B Before Data Memory 0 1 2 0 A 0 1 2 0B After
The conditional move instruction is an excellent way to avoid a discontinuity (branch or call) based upon a condition code set prior to the instruction. In the above example, the 1st step is to
C - 13
Data Move
place the contents of A into the accumulator. Once the Ax content is tested, by using the CMP instruction, the conditional move can be executed. If the specified condition being tested is true, then the location pointed to by the loc16 addressing mode or the 8bit zero extended constant will be loaded with the contents of the specified AX register (AH or AL): if (COND == true) [loc16] = AX or 0:8bit; Note: Addressing modes are not conditionally executed. Hence, if an addressing mode performs a pre or post modification, it will execute regardless if the condition is true or not. This instruction is not repeatable. If this instruction follows the RPT instruction, it resets the repeat counter (RPTC) and executes only once. Flags and Modes N - If the condition is true, then after the move, AX is tested for a negative condition. The negative flag bit is set if bit 15 of AX is 1, otherwise it is cleared. Z - If the condition then after the move, AX is tested for a zero condition. The zero flag bit is set if AX = 0, otherwise it is cleared. V - If the V flag is tested by the condition, then V is cleared. C-Example ; if ( VarA > 20 ) ; VarA = 0; CMP @VarA,#20 ; Set flags on (VarA 20) MOVB @VarA,#0,GT ; Zero VarA if greater then
C - 14
Logical Operations
Logical Operations
Byte Operations and Addressing
Byte Operations
MOVB AX.LSB,loc16 MOVB AX.MSB,loc16 MOVB loc16, AX.LSB MOVB loc16, AX.MSB 0000 0000 Byte No change No change Byte No change Byte Byte AX AX loc16 loc16
Byte = 1. Low byte for register addressing 2. Low byte for direct addressing 3. Selected byte for offset indirect addressing
For loc16 = *+XARn[Offset] Odd Offset Even Offset loc16
Byte Addressing
AH.MSB 12 AH.LSB AL.MSB 34 56 AL.LSB 78 16 bit memory 01 78 03 05 34 07 00 56 02 04 12 06 AR2
Example of Byte Un-Packing

MOVL MOVB MOVB MOVB MOVB XAR2, #MemA *+XAR2[1], AL.LSB *+XAR2[2], AL.MSB *+XAR2[5], AH.LSB *+XAR2[6], AH.MSB
Example of Byte Packing

MOVL MOVB MOVB MOVB MOVB XAR2, #MemA AL.LSB,*+XAR2[1] AL.MSB,*+XAR2[2] AH.LSB,*+XAR2[4] AH.MSB,*+XAR2[7]
C - 15
Logical Operations
Test and Change Memory Instructions

The compare (CMPx) and test (Txxx) instructions allow the ability to test values in memory. The results of these operations can then trigger subsequent conditional branches. The CMPx instruction allows comparison of memory with respect to a specified constant value, while the Txxx instructions allow any single bit to be extracted to the test control (TC) field of status register 0. The contents of the accumulator can also be non-destructively analyzed to establish branching conditions, as seen below.
Test and Change Memory

Instruction TBIT TSET TCLR CMPB CMP CMP CMPL loc16,#(0-15) loc16,#(0-15) loc16,#(0-15) AX, #8bit AX, loc16 loc16,#16b ACC, @P Execution ST0(TC) = loc16(bit_no) Test (loc16(bit)) then set bit Test (loc16(bit)) then clr bit Test (AX - 8bit unsigned) Test (AX loc16) Test (loc16 - #16bit signed) Test (ACC - P << PM) Affects TC TC TC C,N,Z C,N,Z C,N,Z C,N,Z
C - 16
Logical Operations
Min/Max Operations
MIN/MAX Operations
Instruction
MAX MIN MAXL MINL MAXCUL (for 64 MINCUL (for 64 ACC,loc16 ACC,loc16 ACC,loc32 ACC,loc32 P,loc32 bit math) P,loc32 bit math) if if if if if if if if if if if if
Execution
ACC < loc16, ACC = loc16 ACC >= loc16, do nothing ACC > loc16, ACC = loc16 ACC <= loc16, do nothing ACC < loc32, ACC = loc32 ACC >= loc32, do nothing ACC > loc32, ACC = loc32 ACC <= loc32, do nothing P < loc32, P = loc32 P >= loc32, do nothing P > loc32, P = loc32 P <= loc32, do nothing
Find the maximum 32-bit number in a table: MOVL ACC,#0 MOVL XAR1,#table RPT || #(table_length 1) MAXL ACC,*XAR1++
C - 17
Read Modify Write Operations

The accumulator (ACC) is the main working register for the C28x. It is the destination of all ALU operations except those, which operate directly on memory or registers. The accumulator supports single-cycle move, add, subtract and compare operations from 32-bit-wide data memory. It can also accept the 32-bit result of a multiplication operation. These one or two cycle operations are referred to as read-modify-write operations, or as atomic instructions.
Read-Modify-Write Instructions
Work directly on memory bypass ACC Atomic Operations protected from interrupts
AND OR XOR ADD SUB
loc16,AX loc16,AX loc16,AX loc16,AX loc16,AX AH, AL
AND OR XOR ADD
loc16,#16b loc16,#16b loc16,#16b loc16,#16b 16- bit constant
SUBR loc16,#16b TSET loc16,#bit TCLR loc16,#bit
SUBR loc16,AX INC loc16 DEC loc16
C - 18
Read-Modify-Write Examples
update with a mem update with a constant update by 1
VarA += VarB
SETC MOV ADD MOV CLRC INTM AL, @VarB AL, @VarA @VarA, AL INTM
VarA += 100
SETC MOV ADD MOV CLRC INTM AL, @VarA AL, #100 @VarA, AL INTM
VarA += 1
SETC MOV ADD MOV CLRC INTM AL, @VarA AL, INTM #1 @VarA, AL
MOV ADD
AL, @VarB @VarA, AL
ADD
@VarA,#100
INC
@VarA
Benefits of Read-Modify-Write Instructions
C - 19
Lab C: Assembly Programming

Objective
The objective of this lab is to practice and verify the mechanics of performing assembly language programming arithmetic on the TMS320C28x. In this exercise, we will expand upon the .asm file from the previous lab to include new functions. Code will be added to obtain the sum of the products of the values from each array. Perform the sum of products using a MAC-based implementation. In a real system application, the coeff array may well be constant (values do not change), therefore one can modify the initialization routine to skip the transfer of this arrays, thus reducing the amount of data RAM and cycles required for initialization. Also there is no need to copy the zero to clear the result location. The initialization routine from the previous lab using the load/store operation will be replaced with a looped BANZ implementation. As in previous lab, consider which addressing modes are optimal for the tasks to be performed. You may perform the lab based on this information alone, or may refer to the following procedure.
Procedure

1. Create a new project called LabC.pjt in C:\C28x\Labs\Appendix\LabC and add LabC.asm and Lab.cmd to it. Check your file list to make sure all the files are Build there. Be sure to setup the Build Options by clicking: Project Options on the menu bar. Select the Linker tab. In the middle of the screen select No Autoinitialization under Autoinit Model:. Enter start in the Code Entry Point (-e): field. Next, select the Compiler tab. Note that Full Symbolic Debug (-g) under Generate Debug Info: is selected. Then select OK to save the Build Options.
Initialization Routine using BANZ

2. Edit LabC.asm and modify it by replacing the initialization routine using the load/store operation with a BANZ process. Remember, it is only necessary to copy the first four values (i.e. initialize the data array). Do you still need the coeff array in the vars section? 3. Save your work. If you would like, you can use Code Composer Studio to verify the correct operation of the block initialization before moving to the next step.
C - 20
Sum of Products using a RPT/MAC-based Implementation

4. Edit LabC.asm to add a RPT/MAC-based implementation to multiply the coeff array by the data array and storing the final sum-of-product value to result.
Build and Load

5. Click the Build button and watch the tools run in the build window. Debug as necessary. To open up more space, close any open files or windows that you do not need. 6. If the Load program after build option was not selected in Code Composer Load Program Studio, load the output file onto the target. Click: File If you wish, right click on the source window and select Mixed Mode to debug using both source and assembly. 7. Single-step your routine. While single-stepping, open memory windows to see the values located in table [9] and data [9]. Open the CPU Registers. Check to see if the program is working as expected. Debug and modify, if needed.
Optional Exercise
After completing the above, edit LabC.asm and modify it to perform the initialization process using a RTP/PREAD rather than a load/store/BANZ. End of Exercise
C - 21
OPTIONAL Lab C-C: Sum-of-Products in C
OPTIONAL Lab C-C: Sum-of-Products in C

Objective
The objective of this lab is to practice and verify the mechanics of performing C programming arithmetic on the TMS320C28x. The objective will be to add the code necessary to obtain the sum of the products of the n-th values from each array.
Procedure
Create Project File

1. In Code Composer Studio create a new project called LabC-C.pjt in C:\C28x\Labs\Appendix\LabC\LabC-C and add LabC-C.c and Lab.cmd to it. Check your file list to make sure all the files are there. Open the Build Options and select the Linker tab. Select the Libraries Category and enter rts2800_ml.lib in the Incl. Libraries (-l): box. Do not setup any other Build Options. The default values will be used. In Appendix Lab D exercise, we will experiement and explore the various build options when working with C.
Sum of Products using a MAC-based Implementation

2. Edit LabC-C.c and modify the main routine to perform a MAC-based implementation in C. Since the MAC operation requires one array to be in program memory, the initialization routine can skip the transfer of one of the arrays, thus reducing the amount of data RAM and cycles required for initialization.
Build and Load

3. Click the Build button and watch the tools run in the build window. Debug as necessary. Note: Have Code Composer Studio automatically load the output file after a successful build. On the menu bar click: Option Customize and select the Program Load Options tab, check Load Program After Build, then click OK. 4. Under Debug on the menu bar click Go Main. Single-step your routine. While single-stepping, open memory windows to see the values located in table [9] and data [9] . (Note: data[9] consists of the allocated arrays of data, coeff, and result). Open the CPU Registers. Check to see if the program is working as expected. Debug and modify, if needed. End of Exercise
C - 22
Appendix D C Programming
Introduction
The C28x architecture, hardware, and compiler have been designed to efficiently support C code programming. Appendix D will focus on how to program in C for an embedded system. Issues related to programming in C and how C behaves in the C28x environment will be discussed. Also, the C compiler optimization features will be explained.
Learning Objectives
Learning Objectives
Learn the basic C environment for the C28x family How to control the C environment How to use the C-compiler optimizer Discuss the importance of volatile Explain optimization tips
C2000 Piccolo Workshop Appendix D C Programming
D- 1
Module Topics
Module Topics
Appendix D C Programming.................................................................................................................D-1 Module Topics.........................................................................................................................................D-2 Linking Boot code from RTS2800.lib ......................................................................................................D-3 Set up the Stack .......................................................................................................................................D-4 C28x Data Types.....................................................................................................................................D-5 Accessing Interrupts / Status Register.....................................................................................................D-6 Using Embedded Assembly .....................................................................................................................D-7 Using Pragma .........................................................................................................................................D-8 Optimization Levels ................................................................................................................................D-9 Volatile Usage ..................................................................................................................................D-11 Compiler Advanced Options ............................................................................................................D-12 Optimization Tips Summary.............................................................................................................D-13 Lab D: C Optimization..........................................................................................................................D-14 OPTIONAL Lab D2: C Callable Assembly...........................................................................................D-17 Solutions................................................................................................................................................D-20
D- 2
Linking Boot code from RTS2800.lib
Linking Boot code from RTS2800.lib
Boot.ASM - Invoked With -C

Reset : PC <- *0x3F FFC0 vectors.asm .ref _c_int00 Reset: .long _c_int00 _c_int00 1. 2. 3. 4. 5. Allocate stack Init SP to top of stack Initialize status bits Copy .cinit to .bss (skip if -cr) Call _main
_main
...
The boot routine is used to establish the environment for C before launching main. The boot routine begins with the label _c_int00 and the reset vector should contain a ".long" to this address to make boot.asm the reset routine. The contents of the boot routine have been extracted and copied on the following page so they may be inspected. Note the various functions performed by the boot routine, including the allocation and setup of the stack, setting of various C-requisite statuses, the initialization of global and static variables, and the call to main. Note that if the link was performed using the "cr" option instead of the "c" option that the global/static variable initialization is not performed. This is useful on RAM-based C28x systems that were initialized during reset by some external host processor, making transfer of initialization values unnecessary. Later on in this chapter, there is an example on how to do the vectors in C code rather than assembly.
D-3
Set up the Stack
Set up the Stack

Data Memory
The Stack
The C/C++ compiler uses a stack to: Allocate local variables Pass arguments to functions Save the processor status Save the function return .stack address Save temporary results The compiler uses the hardware stack pointer (SP) to manage the stack. SP defaults to 0x400 at reset. The run-time stack grows from low addresses to higher addresses.
SP 0x400 (reset)
Callers local vars Arguments passed on stack Return address Function return addr Temp results
64K 4M
The C28x has a 16-bit stack pointer (SP) allowing accesses to the base 64K of memory. The stack grows from low to high memory and always points to the first unused location. The compiler uses the hardware stack pointer (SP) to manage the stack. The stack size is set by the linker.
Setting Up the Stack

Boot.asm sets up SP to point at .stack
Linker command file:

SECTIONS .stack :> ... { RAM align=2 }
The .stack section has to be linked into the low 64k of data memory. The SP is a 16-bit register and cannot access addresses beyond 64K. Stack size is set by the linker. The linker creates a global symbol, --STACK-SIZE, and assigns it a value equal to the size of the stack in bytes. (default 1K words) You can change stack size at link time by using the -stack linker command option.
Note: The compiler provides no means to check for stack overflow during compilation or at runtime. A stack overflow disrupts the run-time environment, causing your program to fail. Be sure to allow enough space for the stack to grow.
In order to allocate the stack the linker command file needs to have align = 2.
D- 4
C28x Data Types
C28x Data Types
C28x C-Language Data Types

Type char int (natural size CPU word) unsigned int short (same as int or smaller) unsigned short long (same as int or larger) unsigned long Bit Value Range 16 16 16 16 16 32 32 Usually 0 .. 255, but can hold 16 bits -32K .. 32K, 16 bits signed 0 .. 64K, 16 bits unsigned same as int same as unsigned int -2M .. 2M, 32 bits signed 0 .. 4M, 32 bits unsigned
float 32 IEEE single precision double 64 IEEE double precision long double 64 IEEE double precision Suggestion: Group all longs together, group all pointers together
Data which is 32-bits wide, such as longs, must begin on even word-addresses (i.e. 0x0, 0x2, etc). This can result in holes in structures allocated on the stack.
D-5
Accessing Interrupts / Status Register

Initialize via C :
extern cregister volatile unsigned int IFR; extern cregister volatile unsigned int IER; . . . IER &= ~Mask; //clear desired bits IER |= Mask; //set desired bits IFR = 0x0000; //clear prior interrupts
Interrupt Enable & Interrupt Flag Registers (IER, IFR) are not memory mapped Only limited instructions can access IER & IFR (more in interrupt chapter) The compiler provides extern variables for accessing the IER & IFR
D- 6
Using Embedded Assembly
Using Embedded Assembly
Embedding Assembly in C
Allows direct access to assembly language from C Useful for operating on components not used by C, ex:
asm ( CLRC #define EINT
INTM
; enable global interrupt ); INTM)
asm ( CLRC
Note: first column after leading quote is label field - if no label, should be blank space. Avoid modifying registers used by C Lengthy code should be written in ASM and called from C main C file retains portability yields more easily maintained structures eliminates risk of interfering with registers in use by C
The assembly function allows for C files to contain 28x assembly code. Care should be taken not to modify registers in use by C, and to consider the label field with the assembly function. Also, any significant amounts of assembly code should be written in an assembly file and called from C. There are two examples in this slide the first one shows how to embed a single assembly language instruction into the C code flow. The second example shows how to define a C term that will invoke the assembly language instruction.
D-7
Using Pragma
Using Pragma
Pragma is a preprocessor directive that provides directions to the compiler about how to treat a particular statement. The following example shows how the DATA_SECTION pragma is used to put a specific buffer into a different section of RAM than other buffers. The example shows two buffers, bufferA and bufferB. The first buffer, bufferA is treated normally by the C compiler by placing the buffer (512 words) into the ".bss" section. The second, bufferB is specifically directed to go into the my_sect portion of data memory. Global variables, normally ".bss", can be redirected as desired. When using CODE_SECTION, code that is normally linked as ".text", can be identified otherwise by using the code section pragma (like .sect in assembly).
Pragma Examples
User defined sections from C :
#pragma CODE_SECTION (func, section name) #pragma DATA_SECTION (symbol, section name)
Example - using the DATA_SECTION Pragma

C source file char bufferA[512]; #pragma DATA_SECTION(bufferB, my_sect) char bufferB[512]; Resulting assembly file .global _bufferA, _bufferB .bss _bufferA,512 _bufferB: .usect my_sect,512
More #pragma are defined in the C compiler UG
D- 8
Optimization Levels
Optimization Levels
Optimization Scope
FILE1.C { { SESE } { } SESE: Single Entry, Single Exit } { . . . } FILE2.C { . . . } ... LOCAL single block FUNCTION across blocks FILE across functions
-o0, -o1
-o2
-o3
-pm -o3
PROGRAM across files
Optimizations fall into 4 categories. This is also a methodology that should be used to invoke the optimizations. It is recommended that optimization be invoked in steps, and that code be verified before advancing to the next step. Intermediate steps offer the gradual transition from fully symbolic to fully optimized compilation. Compiler switched may be invoked in a variety of ways. Here are 4 steps that could be considered: 1st: use g By starting out with g, you do no optimization at all and keep symbols for debug. 2nd: use g o3 The option o3 might be too big a jump, but it adds the optimizer and keeps symbols. 3rd: use g o3 mn This is a full optimization, but keeps some symbols 4th: use o3 Full optimization, symbols are not kept.
D-9
Optimization Levels
Optimization Performance
o0
LOCAL Performs control-flow-graph simplification Allocates variables to registers Performs loop rotation Eliminates unused code Simplifies expressions and statements Expands calls to functions declared inline Performs local copy/constant propagation Removes unused assignments Eliminates local common expressions Default (-o) Performs loop optimizations Eliminates global common sub-expressions Eliminates global unused assignments Removes all functions that are never called Simplifies functions with return values that are never used Inlines calls to small functions Identifies file-level variable characteristics
o1 o2
FUNCTION
o3
FILE
PROGRAM
o3 pm
Optimizer levels zero through three, offer an increasing array of actions, as seen above. Higher levels include all the functions of the lower ones. Increasing optimizer levels also increase the scope of optimization, from considering the elements of single entry, single-exit functions only, through all the elements in a file. The -pm option directs the optimizer to view numerous input files as one large single file, so that optimization can be performed across the whole system.
D- 10
Optimization Levels
Volatile Usage
Optimization Issue: Volatile Variables

Problem: The compiler does not know that this pointer may refer to a hardware register that may change outside the scope of the C program. Hence it may be eliminated (optimized out of existence!) Wrong: Wait loop for a hardware signal Optimizer removes empty loop
unsigned int *CTRL while (*CTRL !=1);

Solution:
No CTRL = 1? Yes empty loop
volatile unsigned int *CTRL while (*CTRL !=1);
When using optimization, it is important to declare variables as volatile when: The memory location may be modifed by something other than the compiler (e.g. its a memory-mapped peripheral register). The order of operations should not be rearranged by the compiler Define the pointer as volatile to prevent the optimizer from optimizing
D - 11
Optimization Levels
Compiler Advanced Options

To get to these options, go to Project In the category, pick Advanced.
The first thing to notice under advanced options is the Auto Inlining Threshold.
Build Options in Code Composer Studio.
- Used with o3 option - Functions > size are not auto inlined
Note: To prevent code size increases when using o3, disable auto inlining with -oi0 The next point we will cover is the Normal Optimization with Debug (-mn). - Re-enables optimizations disabled by g option (symbolic debug) - Used for maximum optimization Note: Some symbolic debug labels will be lost when mn option is used.
Optimizer should be invoked incrementally: -g test -g -o3 test -g -o3 -mn test -o3 test
Symbols kept for debug Add optimizer, keep symbols More optimize, some symbols Final rev: Full optimize, no symbols
[-mf] : Optimize for speed instead of the default optimization for code size [-mi] : Avoid RPT instruction. Prevent compiler from generating RPT instruction. RPT instruction is not interruptible [-mt] : Unified memory model. Use this switch with the unified memory map of the 281x & 280x. Allows compiler to generate the following: -RPT PREAD for memory copy routines or structure assignments -MAC instructions -Improves efficiency of switch tables
D- 12
Optimization Levels
Optimization Tips Summary
Summary: Optimization Tips

Within C functions :
Use const with variables for parameter constants Minimize mixing signed & unsigned ops : SXM changes Keep frames <= 64 (locals + parameters + PC) : *-SP[6bit] Use structures <= 8 words : use 3 bit index mode Declare longs first, then declare ints : minimize stack holes Avoid: long = (int * int) : yields unpredictable results
Optimizing : Use -o0, -o1, -o2, -o3 when compiling

Inline short/key functions Pass inlines between files : static inlines in header files Invoke automatic inlining : -o3 -oi Give compiler project visibility : use -pm and -o3
Tune memory map via linker command file Re-write key code segments to use intrinsics or in assembly
App notes 3rd Parties
The list above documents the steps that can be taken to achieve increasingly higher coding efficiency. It is recommended that users first get their code to work with no optimization, and then add optimizations until the required performance is obtained.
D - 13
Lab D: C Optimization
Objective
The objective of this lab is to practice and verify the mechanics of optimizing C programs. Using Code Composer Studio profile capabilities, different routines in a project will be benchmarked. This will allow you to analyze the performance of different functions. This lab will highlight the profiler and the clock tools in CCS.
Procedure
Create Project File

1. Create a new project in C:\C28x\Labs\Appendix\LabD called LabD.pjt and add LabD.c, Lab.cmd, and sop-c.c to it. (Note that sop-asm.asm will be used in the next part of the lab, and should not be added now). 2. Setup the Build Options. Select the Linker tab and notice that Run-time Autoinitialization under Autoinit Model:is selected. Do not enter anything in the Code Entry Point (-e): field (leave it blank). Set the stack size to 0x200. In the Linker options select the Libraries Category and enter rts2800_ml.lib in the Incl. Libraries (-l): box. Next, select the Compiler tab. Note that Full Symbolic Debug (-g) under Generate Debug Info: in the Basic Category is selected. On the Feedback Category pull down the interlisting options and select C and ASM (-ss). On the Assembly Category check the Keep generated .asm Files (-k), Keep Labels as Symbols (-as) and Generate Assembly Listing Files (-al). The as will allow you to see symbols in the memory window and the al will generate an assembly listing file (.lst file). The listing file has limited uses, but is sometime helpful to view opcode values and instruction sizes. (The .lst file can be viewed with the editor). Both of these options will help with debugging. Then select OK to save the Build Options.
Build and Load

3. Click the Build button and watch the tools run in the build window. Be sure the Load program after build option is selected in Code Composer Studio. The output file should automatically load. The Program Counter should be pointing to _c_int00 in the Disassembly Window.
Set Up the Profile Session

4. Restart the DSP (debug restart) and then Go Main. This will run through the C initialization routine in Boot.asm and stop at the main routine in LabD.c.
D- 14
5. Set a breakpoint on the NOP in the while(1) loop at the end of main() in LabD.c. 6. Set up the profile session by selecting Profiler a session name of your choice (i.e. LabD). Start New Session. Enter
7. In the profiler window, hover the mouse over the icons on the left region of the window and select the icon for Profile All Functions. Click on the + to expand the functions. Record the Code Size of the function sop C code in the table at the end of this lab. Note: If you do not see a + beside the .out file, press Profile All Functions on the horizontal tool bar. (You can close the build window to make the profiler window easier to view by right clicking on the build window and selecting hide). 8. Select F5 or the run icon. Observe the values present in the profiling window. What do the numbers mean? Click on each tab to determine what each displays.
Benchmarking Code
9. Lets benchmark (i.e.count the cycles need by) only a portion of the code. This requires you to set a breakpoint pair on the starting and ending points of the benchmark. Open the file sop-c.c and set a breakpoint on the for statement and the return statement. 10. In CCS, select Profile Setup. Check Profile all Functions and Loops for Total Cycles and click Enable Profiling. Then select Profile viewer. 11. Now Restart the program and then Run the program. The program should be stopped at the first breakpoint in sop. Double click on the clock window to set the clock to zero. Now you are ready to benchmark the code. Run to the second breakpoint. The number of cycles are displayed in the viewer window. Record this value in the table at the end of the lab under C Code - Cycles.
C Optimization
12. To optimize C code to the highest level, we must set up new Build Options for our Project. Select the Compiler tab. In the Basic Category Panel, under Opt Level select File (-o3). Then select OK to save the Build Options. 13. Now Rebuild the program and then Run the program. The program should be stopped at the first breakpoint in sop. Double click on the clock window to set the clock to zero. Now you are ready to benchmark the code. Run to the second breakpoint. The number of cycles are displayed in the clock window. Record this value in the table at the end of the lab under Optimized C (-o3) - Cycles. 14. Look in your profile window at the code size of sop. Record this value in the table at the end of this lab.
Benchmarking Assembly Code

15. Remove sop-c.c from your project and replace it with sop-asm.asm. Rebuild and set breakpoints at the beginning and end of the assembly code (MOVL & LRETR).
D - 15
16. Start a new profile session and set it to profile all functions. Run to the first breakpoint and study the profiler window. Record the code size of the assembly code in the table. 17. Double Click on the clock to reset it. Run to the last breakpoint. Record the number of cycles the assembly code ran. 18. How does assembly, C code, and optimized C code compare on the C28x? C Code Code Size Cycles Optimized C Code (-o3) Assembly Code
End of Exercise
D- 16
OPTIONAL Lab D2: C Callable Assembly

Objective
The objective of this lab is to practice and verify the mechanics of implementing a C callable assembly programming. In this lab, a C file will be used to call the sum-of-products (from the previous Appendix LabC exercise) by the main routine. Additionally, we will learn how to use Code Composer Studio to configure the C build options and add the run-time support library to the project. As in previous labs, you may perform the lab based on this information alone, or may refer to the following procedure.
Procedure

1. Create a new project in C:\C28x\Labs\Appendix\LabD2 called LabD2.pjt and add LabD2.c, Lab.cmd, and sop-c.c to it. 2. Do not add LabC.asm to the project (copy of file from Appendix Lab C). It is only placed here for easy access. Parts of this file will be used later during this lab exercise. 3. Setup the Build Options. Select the Linker tab and notice that Run-time Autoinitialization under Autoinit Model:is selected. Do not enter anything in the Code Entry Point (-e): field (leave it blank). Set the stack size to 0x200. In the Linker options select the Libraries Category and enter rts2800_ml.lib in the Incl. Libraries (-l): box. Next, select the Compiler tab. Note that Full Symbolic Debug (-g) under Generate Debug Info: in the Basic Category is selected. On the Feedback Category pull down the interlisting options and select C and ASM (-ss). On the Assembly Category check the Keep generated .asm Files (-k), Keep Labels as Symbols (-as) and Generate Assembly Listing Files (-al). The as will allow you to see symbols in the memory window and the al will generate an assembly listing file (.lst file). The listing file has limited uses, but is sometime helpful to view opcode values and instruction sizes. (The .lst file can be viewed with the editor). Both of these options will help with debugging. Then select OK to save the Build Options.
Build and Load

4. Click the Build button and watch the tools run in the build window. Be sure the Load program after build option is selected in Code Composer Studio. The output file should automatically load. The Program Counter should be pointing to _c_int00 in the Disassembly Window. 5. Under Debug on the menu bar click Go Main. This will run through the C initialization routine in Boot.asm and stop at the main routine in LabD2.c.
D - 17
Verify C Sum of Products Routine

6. Debug using both source and assembly (by right clicking on the window and select Mixed Mode or using View Mixed Source/ASM). 7. Open a memory window to view result and data. 8. Single-step through the C code to verify that the C sum-of-products routine produces the results as your assembly version.
Viewing Interlisted Files and Creating Assembly File

9. Using File Open view the LabD2.asm and sop-c.asm generated files. The compiler adds many items to the generated assembly file, most are not needed in the Ccallable assembly file. Some of the unneeded items are .func / .endfunc. .sym, and .line. 10. Look for the _sop function that is generated by the compiler. This code is the basis for the C-callable assembly routine that is developed in this lab. Notice the comments generated by the compiler on which registers are used for passing parameters. Also, notice the C code is kept as comments in the interlisted file. 11. Create a new file (File New, or clicking on the left most button on the horizontal toolbar New) and save it as an assembly source file with the name sop-asm.asm. Next copy ONLY the sum of products function from LabC.asm into this file. Add a _sop label to the function and make it visible to the linker (.def). Also, be sure to add a .sect directive to place this code in the code section. Finally, add the following instruction to the end: LRETR ; return statement
12. Next, we need to add code to initialize the sum-of-products parameters properly, based on the passed parameters. Add the following code to the first few lines after entering the _sop routine: (Note that the two pointers are passed in AR4 and AR5, but one needs to be placed in AR7. The loop counter is the third argument, and it is passed in the accumulator.) MOVL MOV SUBB XAR7,XAR5 AR5,AL XAR5,#1 ;XAR7 points to coeff [0] ;move n from ACC to AR5 (loop counter) ;subtract 1 to make loop counter = n-1
Before beginning the MAC loop, add statements to set the sign extension mode, set the SPM to zero, and a ZAPA instruction. Use the same MAC statement as in Lab 4, but use XAR4 in place of XAR2. Make the repeat statement use the passed value of n-1 (i.e. AR5). RPT AR5 ;repeat next instruction AR5 times
D- 18
Now we need to return the result. To return a value to the calling routine you will need to place your 32-bit value in the ACC. What register is the result currently in? Adjust your code, if necessary. 13. Save the assembly file as sop-asm.asm. (Do not name it LabD2.asm because the compiler has already created with that name from the original LabD2.c code).
Defining the Function Prototype as External

14. Note in LabD2.c an extern modifier is placed in front of the sum-of-products function prototype: extern int sop(int*,int*,int); //sop function prototype
Verify Assembly Sum of Products Routine

15. Remove the sop-c.c file from the project and add the new sop-asm.asm assembly file to the project. 16. Rebuild and verify that the new assembly sum-of-products routine produces the same results as the C function.
End of Exercise
D - 19
Solutions
Solutions
Lab D Solutions
C Code Code Size Cycles 27 118
Optimized Assembly C Code Code (-o3) 12 11 32 22
D- 20
Appendix E Control Law Accelerator

Introduction
Appendix E discusses the details of the Piccolo TMS320F2803x Control Law Accelerator (CLA). The floating-point number format and the CLA registers will be discussed. Details of the CLA instruction set and pipeline will be explained. Additionally, system configuration and a comparison to the Delfino floating-point unit (FPU) will be covered.
Learning Objectives
Learning Objectives
Floating-point format CLA registers and execution flow Instructions Pipeline System configuration Summary
C2000 Piccolo Workshop - Appendix E - Control Law Accelerator
E-1
Module Topics
Module Topics
Appendix E Control Law Accelerator ..................................................................................................E-1 Module Topics......................................................................................................................................... E-2 Control Law Accelerator ........................................................................................................................ E-3 Floating-Point Format......................................................................................................................... E-3 CLA Registers and Execution Flow ................................................................................................... E-4 CLA Instructions ................................................................................................................................ E-5 Status Register and Pipeline ............................................................................................................... E-6 CLA System Configuration ................................................................................................................ E-9 CLA Compared to C28x+FPU ......................................................................................................... E-11 Summary .......................................................................................................................................... E-12
E-2

Floating-Point Format
IEEE Single-Precision Floating-Point Format

1 Sign Bit (0 = Positive, 1 = Negative) S E S E 8-bit Exponent (Biased) 23-bit Mantissa (Implicit Leading Bit + Fraction Bits) M M
S 0 0 0 0 0 1 1 1 1 1
E 0 0
M 0 Non-Zero
Value Positive or Negative Zero Denormalized Number
1-254 0-0x7FFFF Positive or Negative Values* 255 (max) 0 Positive or Negative Infinity 255 (max) Non-Zero Not a Number (NaN)
* Normal Positive and Negative Values are Calculated as: ( -1 ) s x 2 (E-127) x 1.M +/- ~1.7 x 10 -38 to +/- ~3.4 x 10 +38 The Normalized IEEE numbers have a hidden 1; thus the equivalent signed integer resolution is the number of mantissa bits + sign + 1
IEEE Single-Precision Floating-Point Format (IEEE 754)

Most widely used standard for floating-point
Standard number formats, Special values (NaN, Infinity) Rounding modes & floating-point operations Used on many CPUs
Simplifications for the C28x floating-point unit

Flags & Compare Operations:
Negative zero is treated as positive zero
Denormalized values are treated as zero Not-a-number (NaN) is treated as infinity Round-to-Zero Mode Supported (truncate) Round-to-Nearest Mode Supported (even)
These formats are commonly handled this way on embedded processors
E-3
CLA Registers and Execution Flow
CLA Register Set

Interrupt/Task Control
MIFR: MICLR: MIFRC: MIOVF: MICLROVF: Flag Clear Force Overflow flag Overflow clear
CLA Configuration Registers MIER MIRUN MIFR MICLR MIFRC MIOVF MICLROVF MPISRCSEL1 MEMCFG MCTL MVECT1 to MVECT8
CLA Configuration Registers:

CSM and EALLOW Protected Main CPU has Read and Write Access
MIER: Interrupt enable/disable MIRUN: Which task is running Interrupt/Task Source Selection MPISRCSEL1:
Task1: ADCINT1 Task2: ADCINT2 . Task7: ADCINT7 Task8: ADCINT8 or EPWM1_INT or EPWM2_INT or EPWM7_INT or CPU Timer 0
Configuration and Control MEMCFG: Memory config MCTL: CLA control Eight Interrupt (Task) Vectors MVECT1 to MVECT8
Offset from the start of CLA Program Memory to the beginning of the task
CLA Execution Registers:

CSM Protected Main CPU has Read Only Access
CLA Execution Registers
MSTF: Status Register

Zero, negative, overflow, underflow Rounding mode RPC: Return PC MEALLOW
Four 32-bit Result Registers MR0 MR3 Two 16-bit Auxiliary Registers MAR0, MAR1
Used for indirect addressing
MR0 (32) MR1 (32) MR2 (32) MR3 (32) MSTF (32) MAR0 MAR1 MPC
MPC: 12-bit Program Counter

Offset from the start of CLA program memor y Indicates instruction in the D2 phase
CLA Execution Flow

Task request is via software or interrupt assigned in MPISRCSEL1: Task1: ADCINT1 or EPWM1_INT Task2: ADCINT2 or EPWM2_INT
No Task Task No Yes Pending? Enabled? (MIFR) (MIER) Yes Clear MIFR.x bit Set MIRUN.x bit MPC == MVECTx Run CLA
x = Highest priority task both enabled and pending Task1:

.. .
Task7: ADCINT7 or EPWM7_INT Task8: ADCINT8 or CPU Timer 0 No Task Request ? Yes MIFR No Set MIFR bit bit (Task Pending) Set? Yes Set MIOVF Bit (Overflow Flagged)
Priority Highest Lowest
Task8:
The task runs to completion (No task nesting)
End of Task? MSTOP Yes
No
The main CPU continues code execution in parallel with the CLA
Note: Software task requests will not set MIOVF
Clear MIRUN.x bit Task x Interrupt to PIE
When a task completes a taskspecific interrupt is sent to the PIE
E-4
CLA Instructions
CLA Parallel Instructions

Parallel bars indicate a parallel instruction Parallel instructions operate as a single instruction with a single opcode and performs two operations
Example: Add + Parallel Store MADDF32 MR3, MR3, MR1 MMOV32 @_Var, MR3
||
Instruction Multiply & Parallel Add/Subtract Multiply, Add, Subtract & Parallel Store Multiply, Add, Subtract, MAC & Parallel Load
Example MMPYF32 MRa,MRb,MRc || MSUBF32 MRd,MRe,MRf MADDF32 MRa,MRb,MRc || MMOV32 mem32,MRe MADDF32 MRa,MRb,MRc || MMOV32 MRe, mem32
Cycles 1 1 1
Both operations complete in a single cycle
Multiply and Store Parallel Instruction

; Before: MR0 = 2.0, MR1 = 3.0, MR2 = 10.0 MMPYF32 MR2, MR1, MR0 ; 1/1 instruction || MMOV32 @_X, MR2 <any instruction> ; After: MR2 = MR1 * MR0 = 3.0 * 2.0 ; @_X = 10.0
Both the math operation and store complete in 1 cycle Parallel Instruction:
MMOV32 uses the value of MR2 before the MMPY32 updates
E-5
Status Register and Pipeline
CLA Status Flags

CLA Status Register MSTF (32-bits)
RPC MEALLOW rsvd RND rsvd F32 TF rsvd ZF NF LUF LVF
LVF LUF ZF NF
Latched Overflow Float math: MMPYF32, MADDF32, 1/x etc. and Underflow Connected to the PIE for debug
Negative and Zero
TF RNDF32 MEALLOW RPC
Test Flag Rounding Mode Write Protection Return Program Counter
Float move operations to registers. Result of compare, min/max, absolute, negative Integer result of integer operations (MAND32, MOR32, SUB32, MLSR32 etc.) MTESTTF Instruction To Zero (truncate) or To Nearest (even) Enable/disable CLA writes to EALLOW protected registers Call and return: MCNDD, MRCNDD Use store/load MSTF instructions to nest calls
CLA Pipeline Stages

Fetch CLA Pipeline
F1 F2
Decode
D1 D2
Read Exe Write

R1 R2 E W
Independent 8 Stage Pipeline Fetch1: Fetch2: Program read address generated Read Opcode via CLA program data bus
Decode1: Decode instruction Decode2: Generate address Conditional branch decision made MAR0/MAR1 update due to indirect addressing post increment Read1: Read2: Data read address via CLA data read address bus Read data via CLA data read data bus
Execute: Execute operation MAR0/MAR1 update due to load operations Write: Write All Instructions are single cycle (except for Branch/Call/Return) Memory conflicts in F1, R1 and W stall the pipeline
E-6
Write Followed-by-Read
Fetch CLA Pipeline
F1 F2
Decode
D1 D2
Read Exe Write

R1 R2 E W
MMOV32 @_Reg1, MR3 MMOV32 MR0, @_Reg2
; Write Reg1 ; Read Reg2
Due to the pipeline order, the read of Reg2 occurs before the Reg1 write This is only an issue if the location written to can affect the location read Some peripheral registers Write to followed by read from the same location Insert 3 other instructions or MNOPs to allow the write to occur first
Note: This behavior is different for the main C28 CPU: The C28x CPU protects write followed by read to the same location Blocks of peripheral registers have write-followed-by read protection
Loading MAR0 and MAR1

Fetch CLA Pipeline
F1 F2
Decode
D1 D2
Read
R1 R2
Exe Write
E W
D2: EXE:
Update to MAR0/MAR1 due to indirect addressing post increment Update to MAR0/MAR1 due to load operation
Assume MAR0 is 50 and #_X is 20 MMOV16 MAR0, #_X MMOV32 MAR1, *MAR0[0]++ MMOV32 MAR1, *MAR0[0]++ <Instruction 4> MMOV32 MAR1, *MAR0[0]++ ; I1 Load MAR0 with 20 ; I2 Uses old MAR0 Value (50) ; I3 Uses old MAR0 Value (50) ; I4 Can not use MAR0 ; I5 Uses new MAR0 Value (20)
When instruction I1 is in EXE instruction I4 is in D2 If I4 uses MAR0, then a conflict will occur and MAR0 will not be loaded
E-7
Branch, Call, Return Delayed Conditional

Fetch CLA Pipeline
F1 F2
Decode
D1 D2
Read Exe Write

R1 R2 E W
D2: Decide whether or not to branch EXE: Branch taken (or not) <Instruction 1> <Instruction 2> <Instruction 3> <Instruction 4> Branch, CND <Instruction 5> <Instruction 6> <Instruction 7> ; I1 Last instruction to affect flags for branch ; I2 ; I3 ; I4 Can not be branch or stop * Do not change flags in time to affect branch
; MBCNDD, MCCNDD or MRCNDD ; I5 ; I6 ; I7 Can not be branch or stop * Always executed whether branch is taken or not
* Can not be MSTOP (end of task), MDEBUGSTOP (debug halt), MBCNDD (branch), MCCNDD (call), or MRCNDD (return)
Optimizing Delayed Conditional Branch

6 instruction slots are executed on every branch Use these slots to improve performance
Cycle count varies depending on delay slot usage
Taken 7 1 4 Not Taken 7 7 4 MCMPF32 MR0,#0.1 MNOP MNOP MNOP MBCNDD Skip1,NEQ MNOP MNOP MNOP MMOV32 MR1,@_Ramp MMOVXI MR2,#RAMP_MASK MOR32 MR1,MR2 MMOV32 @_Ramp,MR1 ... MSTOP Skip1: MCMPF32 MR0,#0.01 MNOP MNOP MNOP MBCNDD Skip2,NEQ MNOP MNOP MNOP MMOV32 MR1,@_Coas t MMOVXI MR2,#COAST _MASK MOR32 MR1,MR2 MMOV32 @_Coast,MR 1 ... MSTOP Skip2: MMOV32 MR3,@_Stea dy MMOVXI MR2,#STEAD Y_MASK MOR32 MR3,MR2 MMOV32 @_Steady,M R3 ... MSTOP
Optimized Code
MCMPF3 2 MCMPF3 2 MTESTT F MNOP MBCNDD MMOV32 MMOVXI MOR32 MMOV32 ... MSTOP MR0, #0.1 MR0, #0.01 EQ Skip 1,NEQ MR1, @_Ramp MR2, #RAMP_MA SK MR1, MR2 @_Ra mp,MR1
Skip1:
MSTOP, MDEBUGSTOP MBCNDD, MCCNDD MRCNDD are not allowed in delay slots
MMOV32 MMOVXI MOR32 MBCNDD MMOV32 MMOVXI MOR32 MMOV32 ... MSTOP MMOV32 ... MSTOP
MR3, @_Steady MR2, #STEADY_ MASK MR3, MR2 Skip 2,NTF MR1, @_Coast MR2, #COAST_M ASK MR1, MR2 @_Co ast,MR1
Skip2:
@_St eady,MR3
E-8
CLA System Configuration
Code Partitioning
CLA and Main CPU communication via shared message RAMs and interrupts Main CPU performs communication, diagnostics, I/O in C
C28 Run Time Code

Go
System initialization by the main CPU in C
Access peripheral registers & memory
C Code Assembly Code
C28 + CLA System Initialization Code
Configure
Peripherals & Memory
Go
CLA concurrently services time-critical control loops
CLA Run Time Code
Access peripheral registers & memory
Just in Time ADC Sampling Using CLA

ADCs early interrupt
ADC Sample Window 7 Cycles (minimum) ADC Conversion
The ADC early interrupt occurs at the end of the sampling window The CLA can read the result register as soon as it is latched
ADC to CLA Interrupt Response Latency 6 Cycles
7 cycles after the early interrupt, the first CLA instruction is in the D2 phase of the pipeline ; I1 Perform pre-calculations using the first 7 instructions (7 cycles) The 8th instruction is just-in-time to read the ADC RESULT register (1 cycle)
I1 in D2
CLA Max Bandwidth = 26 Cycles
RESULT Register Updates After 15 Cycles
<Instruction 1> ...
<Instruction 7> ; I7 MUI16TOF32 MR0,@_AdcRegs.RESULT1
I8 in R2
R ead AD C Reg
RESULT register is latched and ready to be read
Assume 12 instructions 12 cycles
Enables low ADC sample to output delay
MSTOP
; 1 cycle Timing shown for 2803x
Minimum CLA Next Task Response 5 cycles Pre Calc (7 instructions)...
E-9
CLA Interrupts Improved Control Loop Timing

SOCA/B
ePWM1 ePWM7
SOCA/B
ADC
ADCINT1 ADCINT8 ADCINT9 CLA1_INT1
C28x CPU
EPWM1_INT/EPWM1_TZINT
PIE
EPWM7_INT7/EPWM7_TZINT
Piccolo ADC & CLA interrupt structure enables handling of multi-channel systems with different frequencies and/or phases
CLA1_INT8 LUF LVF
EPWM1_INT
CLA
EPWM7_INT
Anatomy of CLA Code

Using a shared C-code header file approach provides easy access to variables and constants in both C28x C and CLA assembly
// F ile: C28 x_Proje ct.h #inc lude DS P2803x_ Device. h #inc lude DS P2803x_ Example s.h // F ile: CLA Shared. h #inc lude DS P28x_Pr oject.h #def ine PERI OD 100. 0 stru ct PI_CT RL { f loat KP ; f loat KI ; f loat I; f loat Re f; } exte rn struc t PI_CT RL PIVa rs; exte rn Uint3 2 Cla1P rog_Sta rt; exte rn Uint3 2 Cla1T ask1; exte rn Uint3 2 Cla1T ask2; etc // Fil e main. c #inclu de CLA Shared. h #pragm a DATA_ SECTION (PIVars ,"CpuToCla1MsgRAM"); struct PI_CTR L PIVar s; .. // Use Symbol s defin ed in t he CLA asm file Cla1R egs.MVE CT1 = ( Uint16) (&Cla1 Task1 \ - &Cla 1Prog_S tart)*s izeof(U int32); // Ini tialize variab les PIVa rs.KP = 1.234 ; PIVa rs.KI = 0.923 67; PIVa rs.Ref = 2048. 0; PIVa rs.I = PIVar s.KP*PI Vars.Re f; .. // Ini tialize Periph erals: Epwm3R egs.PRD = (Uin t16) PE RIOD;
Declare shared constants and variables in C Include DSP2803x_Device.h to define register bitfield structures Assign variables to message RAMs or CLA data memory sections using DATA_SECTION pragma
Add symbols defined in CLA assembly to make them global and usable in C
E - 10
Anatomy of CLA Code

CLA assembly and C28 code reside in the same project
Use .cdecls to include the shared C header file in the CLA assembly file
// File: CLAShared.h #include DSP28x_Project.h #define PERIOD 100.0 struct PI_CTRL { float KP; float KI; float I; float Ref; } extern struct PI_CTRL PIVars; extern Uint32 Cla1Prog_Start; extern Uint32 Cla1Task1; extern Uint32 Cla1Task2; etc ; File: cla.asm ; Include C Header File: .cdecls C,LIST,CLAShared.h ; Add linker directives: Place CLA code .sect Cla1Prog into its own _Cla1Prog_Start: assembly section _Cla1Task2: MDEBUGSTOP ; breakpoint .. ; Read memory or register: MMOV32 MR0,@_PIVars.Ref MUI16TOF32 MR1,@_AdcResult.ADCRESULT0 MSUBF32 MR2,MR1,MR0 .. Use C header file ; Use constants defined in C references in MMPYF32 MR1,MR2,#PERIOD CLA assembly .. ; Write to memory or register MMOV32 @_PIVars.I, MR3 MMOV32 @_EPwm1Regs.CMPA.all, MR2 .. ; End of task Put an MSTOP MSTOP _Cla1Task3:
at the end of the task
CLA Compared to C28x+FPU
CLA Compared to C28x+FPU

Independent 8 Stage Pipeline Single Cycle Math and Conversions No Data Page Pointer; Only uses Direct & Indirect with Post-Increment 4 Result Registers 2 Independent Auxiliary Registers No Stack Pointer or Nested Interrupts Native Delayed Branch, Call & Return Use Delay Slots to Do Extra Work No repeatable instructions Self-Contained Instruction Set Data is Passed Via Message RAMs Supports Native Integer Operations: AND, OR, XOR, ADD/SUB, Shift Programmed in Assembly Single step moves the pipe one cycle
C28x + Floating-Point Unit

F1-D2 Shared with the C28x Pipeline Math and Conversions are 2 Cycle Uses C28x Addressing Modes 8 Result Registers Shares C28x Auxiliary Registers Supports Stack, Nested Interrupts Uses C28x Branch, Call and Return Copy flags from FPU STF to C28x ST0 Repeat MACF32 & Repeat Block Instructions Superset on Top of C28x Pass Data Between FPU and C28x Regs C28x Integer Operations Programmed in C/C++ or Assembly Single step flushes the pipeline
E - 11
Summary
Summary
CLA is an independent 32-bit floating-point math accelerator
robust, self saturating, and easy to program
System and CLA initialization is done by the CPU in C The CLA can directly access:
ADC Result, ePWM+HRPWM and comparator registers
The CLA is interrupt driven and has a low interrupt response time (no nesting of interrupts) By using the ADC early interrupt, the CLA can read the sample Just-in-time
Reduced ADC sample to output delay Faster system response and higher MHz control loops Support for multi-channel loops
E - 12

C28x Piccolo Workshop

Uploaded by

Copyright:

Available Formats

C28x Piccolo Workshop

Uploaded by

Document Information

Original Description:

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

C28x Piccolo Workshop

Uploaded by

Copyright:

Available Formats

C2000 Piccolo Workshop

Workshop Guide and Lab Manual

F28xPmdw Revision 1.0 September 2009

Technical Training Organization

Copyright 2009 Texas Instruments Incorporated

C2000 Piccolo Workshop - Introduction

C2000 Piccolo Workshop

C2000 Piccolo Workshop

C2000 Piccolo Workshop

Texas Instruments Technical Training

C2000 Piccolo Workshop - Introduction

C2000 Piccolo Workshop

C2000 Piccolo Workshop Outline

C2000 Piccolo Workshop Outline

C2000 Experimenter Kit

Piccolo Experimenter Kit

USB Docking Station

C2000 Piccolo Workshop - Introduction

C2000 Piccolo Workshop - Architecture Overview

C2000 Piccolo Workshop - Architecture Overview

What is the TMS320C2000?

What is the TMS320C2000?

TMS320F2803x Block Diagram

eCAP eQEP 12-bit ADC Watchdog

PIE Interrupt Manager 3 32-bit Timers

CAN 2.0B I2C SCI SPI LIN

Data Bus GPIO

C2000 Piccolo Workshop - Architecture Overview

What is the TMS320C2000?

TMS320C2000 Internal Bussing

F28x CPU Internal Bus Structure

C2000 Piccolo Workshop - Architecture Overview

F28x and CLA

MCU/DSP balancing code density & execution time

32-bit Auxiliary Registers

32x32 bit Multiplier

R-M-W Atomic ALU

Register Bus CPU Data Bus

C2000 Piccolo Workshop - Architecture Overview

F28x Atomic Read/Modify/Write

C2000 Piccolo Workshop - Architecture Overview

E & G Access same address

C2000 Piccolo Workshop - Architecture Overview

TMS320F28035 Memory Map

0x3D7C80 0x3D8000 0x3E8000 0x3F7FF8 0x3F8000 0x3F8800 0x3FE000 0x3FFFC0 0x3FFFFF

ADC / OSC cal. data reserved FLASH (64Kw)

L0 SARAM (2Kw) reserved Boot ROM (8Kw)

Dual-Port RAM: L1, L2 & L3 (accessible by CPU & CLA)

C2000 Piccolo Workshop - Architecture Overview

Code Security Module (CSM)

Code Security Module

C2000 Piccolo Workshop - Architecture Overview

Fast Interrupt Response

Fast Interrupt Response

F28x Fast Interrupt Response Manager

Peripheral Interrupts 12x8 = 96

28x CPU Interrupt logic

C2000 Piccolo Workshop - Architecture Overview

F28x Operating Modes

= _ _ = ___ = _ = __