Real-Time DSP

ECE 5655/4655 Lecture Notes

© 1998 – 2014
Mark A. Wickert
Portions "Courtesy of Texas Instruments"

ECE 5655/4655 Real-Time DSP i

ii ECE 5655/4655 Real-Time DSP
Table of Contents
Overview of Real-Time Digital Signal Processing
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1–1
A Brief Description of the Course 1–1
Background Requirements 1–4
A Brief History of DSP in the Context of Real-Time Processing . . . . . . . 1–4
Great Moments in DSPS History 1–6
Who Are the Players? 1–7
DSP Integration (Frantz) 1–7
Communications and Wireless 1–8
DSP Hardware Design Issues . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1–11
DSP System Design Flow (Marven & Ewers). . . . . . . . . . . . . . . . . . . . . 1–12
Course Laboratory Foundations. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1–13
Software 1–14
Test Equipment 1–14

The TMS320C6x Family: Hardware and Software

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2–1
C62/64 Product Specifications 2–2
C67 Product Specifications 2–3
C6x Family Feature Overview. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2–4
C6x Architecture Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2–5
A High-Level Look 2–5
C64xTM Enhancements Include: 2–6
A Detailed Look at the C6x Architecture: 2–7
Functional Unit Operations 2–8
C62/67 and C64 Data Paths 2–9
The Internal Bus Structure 2–10
C6x Internal Buses for 6201 and 6701 2–10
Internal Memory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2–11
C6x11 Double Level Memory 2–11
Cache versus RAM 2–12
Internal Memory Summary 2–13
External Memory Interface (EMIF). . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2–13
SDRAM 2–14
SBSRAM 2–14
Asynchronous Memory 2–14

ECE 5655/4655 Real-Time DSP iii

 Host Port and Expansion Bus (HPI/XB) . . . . . . . . . . . . . . . . . . . . . . . . . 2–14
Host Port 2–14
Expansion Bus 2–15
Serial Port (McBSP). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2–15
DMA and Enhanced DMA (EDMA) . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2–16
Timer/Counter . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2–17
Laboratory Hardware Targets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2–18
The C6201 and C6701 EVM Block Diagram 2–19
EVM Codec Block Diagram 2–20
The C6711 DSP Starter Kit (DSK) Board Layout 2–21
AD535 Telephony Codec on the 6711 DSK 2–22
The C6713 DSK Block Diagram 2–23
AIC23 Codec on the 6713 DSK 2–24
6713 DSK Board Layout 2–25
Memory Mapping for the EVM and DSK’s . . . . . . . . . . . . . . . . . . . . . . 2–27
The C713 and C6416 DSK Memory Maps 2–28
Software Development Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2–29
C6x Code Generation Overview 2–30
Code Development Flow Chart 2–31
DSP/BIOS 2–32
DSP/BIOS Summary for Now 2–33

TMS320C6x Programming
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3–1
Programming Alternatives 3–1
Introduction to Assembly Language Programming . . . . . . . . . . . . . . . . . . 3–2
A Dot Product Example 3–2
Instruction Set Summary by Category 3–10
C62xx and C67xx Instruction Set Summary by Unit 3–11
Introduction to the Pipeline . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3–12
Pipelined and Non-Pipelined 3–13
Program Fetch Stage 3–13
Decode Stage 3–14
Execute Stage 3–14
Summary of Pipeline Phases 3–15
Sending Code Through the Pipeline 3–16
Pipeline Code Example 3–16
Use of Parallel Instructions 3–22
C67x Exceptions 3–26
C Programming . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3–27
Debug options 3–29
Optimize Options 3–29

iv ECE 5655/4655 Real-Time DSP

 Code Size 3–30
Assembler Options 3–30
Linker Options 3–31
Summary of Popular Options 3–31
Embedded Systems with C . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3–32
Initialization Under C 3–34
Compiler Sections 3–36
Memory Management 3–39
Linker Options 3–42
Calling Assembly with C . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3–44
Linear Assembly and Assembly Optimization . . . . . . . . . . . . . . . . . . . . 3–48
Calling from Linear Assembly 3–51
Linear Assembly Compiler Settings 3–51
Example: Vector Norm Squared . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3–51
C Version 3–52
Assembly Version 3–55
The Linear Assembly Version 3–60

Analog Input and Output

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4–1
Analog I/O Basics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4–1
Analog I/O on the C6x . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4–3
Interrupt Steps 4–3
Source Selection 4–4
Interrupt Vector Table 4–6
Enabling Interrupts 4–10
Interrupt Service Routines in C 4–13
Serial Port Communication 4–15
On-Board TLC320AD535 Telephony Codec . . . . . . . . . . . . . . . . . . . . . 4–20
A Simple Interrupt Program 4–23
PCM3006 Stereo Daughter Card Codec . . . . . . . . . . . . . . . . . . . . . . . . . 4–29
Daughtercard Physical Layout (Mike Morrow) 4–31
Supplied Software (Mike Morrow) 4–32
MATLAB API (Mike Morrow) 4–42
TI PCM 3003 Audio Daughter Card . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4–48
Default Jumper Settings 4–49
Jumper Descriptions 4–50
Board Schematics 4–52
A Simple Interrupt Program 4–55
6713 DSK with On-Board TLV320AIC23
Stereo Audio Codec . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4–60
AIC23 Portion of the 6713 DSK 4–65

ECE 5655/4655 Real-Time DSP v

 The AIC23 hardware interface on 6713 DSK 4–66
Board Usage Cautions 4–66
A Simple Loop Program 4–67
Appendix: Mike Morrow C6x11 DSK Support Tools. . . . . . . . . . . . . . . 4–72
winDSK6 4–73
winDSK6 Control 4–73
DSK6x MATLAB API 4–74

Fixed-Point Considerations
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5–1
C6x Data Types . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5–2
Q-Format Number Representation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5–2
Multiplicative Overflow. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5–5
Accumulative Overflow. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5–10
Saturating the Result 5–11
Use Guard Bits 5–13
Overflow Allowed by Design 5–14
Floating Point. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5–15
Single Precision 5–15
Double Precision 5–16
Addition 5–16

Code Optimization and Advanced C

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6–1
C Runtime Environment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6–1
Stack Pointer 6–1
Global Pointer 6–2
Memory Management 6–4
Use of Volatile 6–5
Optimizing C Code . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6–6
Memory Aliases 6–6
C Intrinsic Functions 6–6
Intrinsics Reference Guide. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6–9

Real-Time FIR Digital Filters

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7–1
Basics of Digital Filter Design. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7–2
Overview of Approximation Techniques 7–4
Basic FIR Filter Topologies. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7–5

vi ECE 5655/4655 Real-Time DSP

 Overview of FIR Filter Design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7–7
Why or Why Not Choose FIR? 7–7
FIR Design Using Windowing 7–7
Example: Use of the Hanning Window 7–10
Lowpass Design 7–14
Optimum Approximations of FIR Filters 7–17
MATLAB Basic Filter Design Functions . . . . . . . . . . . . . . . . . . . . . . . . 7–19
Windowed FIR Design From Amplitude Specifications 7–20
Using MATLAB’s fdatool . . . . . . . . . . . . . . . . .7–23
Example: Windowed FIR Design 7–23
Example: Equiripple FIR Design 7–28
Writing C Coefficient Files 7–31
Filter Implementation Code Examples . . . . . . . . . . . . . . . . . . . . . . . . . . 7–35
A Simple Fixed-Point Implementation 7–35
A Simple Floating Point Implementation 7–40
Direct Export Into Code Composer 7–42
Circular Addressing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7–45
Circular Buffer in C 7–45
C Circular Buffer 7–46
Performance of Interrupt Driven Programs . . . . . . . . . . . . . . . . . . . . . . . 7–55
Traditional Code Timing Approach 7–56
Using Counters in the Wait Loop 7–57
MATLAB CCS Link Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7–61

Real-Time IIR Digital Filters

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8–1
Basic IIR Filter Topologies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8–1
Direct Form I 8–2
Direct Form II 8–3
Cascade Form 8–5
Parallel Form 8–6
Transposed Forms 8–8
Pole-Zero Lattice 8–8
Digital Sinusoidal Oscillators . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8–8
Overview of IIR Filter Design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8–11
IIR Approximation Approaches 8–11
Converting to 8–13
Classical Design from Analog Prototypes 8–18
Obtaining the Second-Order Section Coefficients 8–19
MATLAB Design Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8–19
MATLAB Filter Design Examples 8–21
Writing C Coefficient Files 8–33

ECE 5655/4655 Real-Time DSP vii

 Filter Implementation Code Examples . . . . . . . . . . . . . . . . . . . . . . . . . . 8–35
A Simple Floating-Point Cascade of Biquads Implementation 8–35
Direct Export Into Code Composer 8–39
Coupled Form Oscillator in C 8–41
MATLAB CCS Link Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8–45
Command Line Interfacing 8–45
MATLAB GUI Interfacing 8–52

Adaptive Filters
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9–1
Wiener Filter . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9–2
Adaptive Wiener Filter. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9–4
Least-Mean-Square Adaptation 9–6
Adaptive Filter Variations 9–8
Adaptive Line Enhancement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9–9
C6x Code Examples. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9–13
A Two Channel Input Signal + Signal or Signal + Noise Canceller 9–13
Adaptive Line Enhancer 9–18

Real-Time Fast Fourier Transform

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10–1
DTFT, DFT, and FFT Theory Overview . . . . . . . . . . . . . . . . . . . . . . . . . 10–1
The DTFT 10–2
The DFT and its Relation to the DTFT 10–2
Simple Application Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10–5
Spectral Analysis 10–5
Transform Domain Linear Filtering 10–9
Radix 2 FFT . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10–10
Decimation-in-Time 10–11
Decimation-in-Frequency 10–15
Computing the IDFT 10–15
Frame Processing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10–16
Radix-2 FFT Implementation on the C6x . . . . . . . . . . . . . . . . . . . . . . . 10–17
A C Based Real-Time FFT 10–17
Transform Domain Filtering using a Linear Assembly Based FFT and IFFT 10–21

Real-time Analysis and Scheduling

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11–1
Features . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11–1

viii ECE 5655/4655 Real-Time DSP

 The DSP/BIOS API Modules . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11–3
A Case Study: Audio Player with DTMF . . . . . . . . . . . . . . . . . . . . . . . . 11–4
Run them together under main(): 11–5
A second solution is to use two interrupts under main(): 11–5
Interrupt driven state machine 11–6
The DSP/BIOS Solution 11–7
Getting Started with DSP/BIOS 11–10

Communications Applications
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12–1
Summary of Factors Relating to Implementation Choices 12–2
Transmitting Signals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12–3
Baseband/IF Transmitter 12–3
Complex Baseband Transmitter 12–6
Practical Pulse Shaping 12–8
Receiving Signals. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12–15
Complex Envelope Representation 12–15
Standard I-Q Demodulation 12–16
Using the Hilbert Transform 12–17
System Application: A DSP Based Costas Loop for Coherent Carrier Recov-
ery. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12–23
MATLAB Simulation 12–28
Real-Time Implementation on the C6711 DSK 12–32
Commercial Receiver Architectures . . . . . . . . . . . . . . . . . . . . . . . . . . . 12–42

Reference Frameworks
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13–1

Simulink Embedded Target

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14–1

ECE 5655/4655 Real-Time DSP ix

x ECE 5655/4655 Real-Time DSP
 Chapter
Overview of Real-
Time Digital Signal 1
Processing
Introduction
In this first chapter we provide motivation for the topics to be
addressed in this course. Before going any further let us first give
a short description of the course and the assumed background for
the course

A Brief Description of the Course

• A course in real-time DSP brings together the following:
– Continuous- and discrete-time systems theory (in particu-
lar knowledge from a first DSP course)
– Software engineering concepts
– Microprocessor programming and hardware interfacing
• The interest in doing this stems from the increase in real-time
DSP applications headed for the consumer market, and the
ever improving device VLSI designs for implementing pow-
erful DSP microprocessors

ECE 5655/4655 Real-Time DSP 1–1

Chapter 1 • Overview of Real-Time Digital Signal Processing

• Texas Instruments has created the following long list of DSP

real-time application areas

Automotive Consumer Control

Adaptive ride control Digital radios/TVs Disk drive control
Antiskid brakes Educational toys Engine control
Cellular telephones Music synthesizers Laser printer control
Digital radios Pagers Motor control
Engine control Power tools Robotics control
Navigation and global positioning Radar detectors Servo control
Vibration analysis Solid-state answering machines
Voice commands
Anticollision radar

General-Purpose Graphics/Imaging Industrial

Adaptive filtering 3-D rotation Numeric control
Convolution Animation/digital maps Power-line monitoring
Correlation Homomorphic processing Robotics
Digital filtering Image compression/transmission Security access
Fast Fourier transforms Image enhancement
Hilbert transforms Pattern recognition
Waveform generation Robot vision
Windowing Workstations

Instrumentation Medical Military

Digital filtering Diagnostic equipment Image processing
Function generation Fetal monitoring Missile guidance
Pattern matching Hearing aids Navigation
Phase-locked loops Patient monitoring Radar processing
Seismic processing Prosthetics Radio frequency modems
Spectrum analysis Ultrasound equipment Secure communications
Transient analysis Sonar processing

Telecommunications Voice/Speech
1200- to 33 600-bps modems Faxing Speaker verification
Adaptive equalizers Line repeaters Speech enhancement
ADPCM transcoders Personal communications Speech recognition
Cellular telephones systems (PCS) Speech synthesis
Channel multiplexing Personal digital assistants (PDA) Speech vocoding
Data encryption Speaker phones Text-to-speech
Digital PBXs Spread spectrum communications Voice mail
Digital speech interpolation (DSI) Video conferencing
DTMF encoding/decoding X.25 packet switching
Echo cancellation

1–2 ECE 5655/4655 Real-Time DSP

 Introduction

• This course is about the use of general purpose digital signal

processing microprocessors for solving signal processing
problems in real-time
• The course focus will be on using the Texas Instruments (TI)
C6x family of fixed/floating processors, and in particular
implementing programs written in assembly, linear assembly,
and ANSI C, for the TI OMAP-L138 (ARM Cortex-A9 +
C6748 floating-point DSP) ZOOM™ OMAP-L138 Experi-
menter Kit (TMDSEXPL138 $495 at www.ti-estore.com)
– A lower cost version of this board that does not include the
touch screen is available for ~$150 (where?)
• Other platforms that will be explored this semester include:
– TI C6713 DSK (digital signal processing starter kit avail-
able for ~$395.00 at www.ti-estore.com)
– TI C55x fixed-point family via the VC5505/VC5515
eZDSP USB Stick ($49 at www.ti-estore.com)
– TI OMAP-3530 (ARM Cortex-A8 + C64x+ DSP) Beagle-
Board-xM ($149 from links at http://beagleboard.org/)
• The course will start out considering general signal process-
ing applications of real-time DSP and the associated pro-
gramming issues
– Later in the course, a focus applications area will be wire-
less communication system design using DSP algorithms

ECE 5655/4655 Real-Time DSP 1–3

Chapter 1 • Overview of Real-Time Digital Signal Processing

• In addition to the text book, we will be using TI training

materials from the courses
– TMS320C6000 DSP Workshop (normally a 4 day course)
– TI C6000/OMAL-L138 Teaching Materials DVD v3
• The course meeting time will be used for lecturing and labo-
ratory time (about 60/40) using hardware/software develop-
ment tools, primarily Code Composer Studio (4.2 & 5.1)

Background Requirements
• The required background for all students taking the course is
an introductory graduate or junior/senior level undergraduate
course in DSP and experience programming in ANSI C
• Knowledge of ANSI C is required in order to develop real-
time algorithm in a high level language (HLL)
• TMS320C6x assembly and linear assembly language pro-
gramming will be covered in the course
• Eventually, and in practice, we will be using a combination of
C and assembly, e.g., mixed language programming.
• Programming the PC host in C/C++, or perhaps some other
language, may be useful for developing a user interactive
application for the final project
– A MATLAB API (application programming interface) for
the DSK is also available
• Familiarity with test equipment e.g., signal generators, digital
scopes, and a spectrum/network analyzer would be helpful

1–4 ECE 5655/4655 Real-Time DSP

 Undergraduate
Signals & Prob & Signal Comm
Engineering Process
Systems Statistics Lab
Curriculum Lab

Modern Comm Comm

DSP Fa Sp Sys I Sys II Fa (even)

ECE 5655/4655 Real-Time DSP

Senior/1st Year
Course Perspective

Graduate Signals &

Systems Courses
Real-Time Statistical Inform/
Sp Sp Signal (odd) Coding Fa (odd)
DSP
Process

Random Estim & Spread Optical Comm

Signals Adap Filt Spectrum Comm Networks
Other Graduate
Signals & Systems
Detect/ PLL & Wireless & Satellite Radar
Courses Offered on
Estimation Freq Syn Mobil Com Comm Systems
Demand/Indep. Study

Spectral Wireless Image

Estimation Networking Processing

1–5
Course Perspective
Chapter 1 • Overview of Real-Time Digital Signal Processing

A Brief History of DSP in the Context of

Real-Time Processing
• From a systems engineering point of view DSP stands for
digital signal processing, but in the world of real-time hard-
ware, systems people often refer to digital signal processing
solutions (DSPS)
• The decades of DSP/DSPS have brought the following1
Table 1.1: Decades of DSP

Decade Characteristic $/MIPS

‘60s University Curiosity $100–$1,000
‘70s Military Advantage $10–$100
‘80s Commercial Success $1–$10
‘90s Consumer Enabler $0.10–$1
Beyond Expected Part of Daily Life? $0.01–$0.10

1. Gene A. Frantz, “TI’s DSPS Future,” Texas Instruments DSPS Fest ‘97,
Houston, TX, July 1997.

1–6 ECE 5655/4655 Real-Time DSP

 A Brief History of DSP in the Context of Real-Time Processing

• Another way of looking at this is (Frantz)

Processing Processors Processing Solutions

1980 1990

Technology Product Technology

What is DSP? How do I create How do I solve

a product? problems?

• We are well beyond the time line shown above, what has
been happening in the last 12 years?
– DSP is definitely mainstream in most everything electronic
– general purpose DSP processors are only part of the pic-
ture; FPGA drives the government/military applications
side, and ARM processors drive the consumer product side
– In the last few years we have seen more multi-core devices
– The open media application platform, OMAP-L138 used
for this course is an example of this

ECE 5655/4655 Real-Time DSP 1–7

Chapter 1 • Overview of Real-Time Digital Signal Processing

Great Moments in DSPS History1

• 1976 – DSP is used to simulate a voice in the educational
product “Speak and Spell”
• 1982 – TI announces details on the TMS32010 which exe-
cutes at 5 MIPS
• 1985 – DSP is used in a modem for the first time
• 1986 – Lotus automotive uses DSP in active suspension and
noise abatement in racing cars
• 1988 – First DSP hearing aid introduced
• 1991 – First TI sponsored educators conference
• 1993 – Cadillac automotive introduces a DSP-based ride con-
trol system
• 1995 – TI implements the On-line DSP LabTM for Web based
testing of DSP applications
• 1997 – 15 years of DSPs for TI and
– The introduction of the TMS320C6x family of DSPs 1,600
MIPS performance (C62x)
– The 1-v barrier in power consumption reached
– In late 1997 the first C6x floating point part is announced
(C67x) with 1 GFLOPS performance
• 2012?

1. “Great Moments in DSPS History,” Integration: An Update on Texas

Instruments Semiconductors, vol. 14, No. 4, June 1997.

1–8 ECE 5655/4655 Real-Time DSP

 A Brief History of DSP in the Context of Real-Time Processing

DSP Vendor Market Share for 2006

Source: Forward
Concepts, 2007

11%
6%
7%

13%
63%

TI Freescale Agere ADI Other

– TI increased by 17% in 2006 over 2005; 17% increase in

cellular shipments and 9% in the DSP market

DSP Markets for 2006

Source: Forward
3.3% Concepts, 2007

4.7%
8.2% 2.8%

8.5%

72.5%

Wireless Multipurpose Consumer Wireline

Computer Automotive

ECE 5655/4655 Real-Time DSP 1–9

Chapter 1 • Overview of Real-Time Digital Signal Processing

DSP Integration (Frantz)

Table 1.2: DSP Integration over three decades.

Typical 1982 Typical 1992 Typical 2002

DSP DSP (97) DSP
Die size 50 mm 50 mm 50 mm
Technology size 3 0.8 (.35) 0.18
MIPS 5 40 (100) 2000
MHz 20 80 (200) 500
RAM (words) 144 1K 16 K
ROM (words) 1.5 K 4K 64 K
Price $150.00 $15.00 $1.50
Pwr. Dissp. 250 mW/ 12.5 mW/ 0.1 mW/MIP
MIP MIP (1)
Transistors 50 K 500 K 5M
Wafer Size 3” (75 mm) 6” (150 mm) 12”
(300 mm)

• In Spring 2000 TI also announced the C55x family, which

offers 0.05 mW/MIPS and 600–800 MIPS
• Late 2004, C67 at 1350 MFLOPs at $36.50, 800 MFLOPs at
$21.07
• Late 2004, 1GHz C64 which offers fixed-point processing for
communications and imaging applications (eight 32-bit
instructions/cycle, 28 operations/cycle, 8000 MIPS

1–10 ECE 5655/4655 Real-Time DSP

 Microcontrollers (MCUs) ARM®-Based Processors Digital Signal Processors (DSPs)

16-bit ultra- 32-bit 32-bit ARM ARM Ultra

DSP Multi-core
low power real-time Cortex™-M3 Cortex-A8 Low power
MCUs DSP+ARM DSP
MCUs MCUs MPUs DSP

C2000™ Sitara ™ C6000™ C5000™

MSP430™ Stellaris®

ECE 5655/4655 Real-Time DSP

Delfino™ ® DaVinci™ C6000™
ARM® Cortex™-M3 ARM Cortex™-A8 video processors
Piccolo™ & ARM9 Integra™ 24.000
Up to 40MHz to Up to 300MHz to 300MHz to >1Ghz MMACS Up to 300 MHz
25 MHz 300 MHz 100 MHz >1GHz +Accelerator +Accelerator
Flash Flash, RAM Flash Cache, Cache Cache Up to 320KB RAM
1 KB to 256 KB 16 KB to 512 KB 8 KB to 256 KB RAM, ROM RAM, ROM RAM, ROM Up to 128KB ROM
Analog I/O, ADC PWM, ADC, USB, ENET USB, CAN, USB, ENET, SRIO, EMAC USB, ADC
LCD, USB, RF 2
CAN, SPI, I C MAC+PHY CAN, PCIe, EMAC PCIe, SATA, SPI DMA, PCIe McBSP, SPI, I2C
ADC, PWM, SPI
Measurement, Motor Control, Connectivity, Security, Industrial computing, Floating/Fixed Point Telecom test & meas, Audio, Voice
Sensing, General Digital Power, Motion Control, HMI, POS & portable Video, Audio, Voice, media gateways,
Purpose Lighting, Ren. Enrgy Industrial Automation data terminals Security, Conferencing base stations Medical, Biometrics
$0.25 to $9.00 $1.50 to $20.00 $1.00 to $8.00 $5.00 to $20.00 $5.00 to $200.00 $40 to $200.00 $3.00 to $10.00

Software & Dev. Tools

MPUs – Microprocessors
A Brief History of DSP in the Context of Real-Time Processing

1–11
Chapter 1 • Overview of Real-Time Digital Signal Processing

– C67x floating-point family

New OMAP-L138, TMS320C6748/6/2 power-efficient
processors offer unmatched connectivity options for industrial
and communication applications
With these four new processors from TI
based on the OMAP-L13x/C674x platforms,
developers can:

Access flexible networking and interface

options (SATA, uPP, Ethernet MAC and
more)
Optimize their designs for power and
performance utilizing the industry’s
lowest power floating-point DSP
Develop and design products easier with low-cost development tools (starting at $149)
and pin-for-pin and software compatible options.

All the devices feature:

C674x fixed- and floating-point DSP core which provides high-precision and wide
dynamic range as well as higher performance, efficiency and low power
Power management software with DVFS to enable various power down modes
Complementary analog products including the TI TPS65070 power management device
with integrated 10-bit, 4-channel ADC, touch screen controller and battery charger
implements all sequencing and default options and supports the devices’ power modes
Object code compatible with all TMS320C67x+™ and TMS320C64x+™devices so
developers can leverage existing software code
Pin-for-pin compatibility (C6742, C6746, C6748 and OMAP-L138) allows for the ability
to expand an entire product portfolio using the same hardware and software platform

The OMAP-L138 applications processor consists of an ARM9 and C674x fixed and
floating point DSP combined with interface options for SATA, Universal Parallel Port
(uPP), Ethernet MAC, USB 2.0HS, USB1.1 FS, Video Input/Output, LCD Controller and
much more.

The TMS320C6748/6/2 processors consist of a C674x fixed- and floating-point DSP

combined with a variety of networking and interface features including many, if not all,
of those found in the OMAP-L138.

1–12 ECE 5655/4655 Real-Time DSP

 A Brief History of DSP in the Context of Real-Time Processing

– C64x fixed-point multi-core family

– Multi-core versions
TMS320C647x Multicore DSPs

Devices Benchmarks

TMS320C647x™ high-performance DSP

Options of up to 6 integrated cores on a single die

Highest performing multicore DSP with 4.2 GHz performance
Includes industry’s best power efficiency with 3 GHz at 0.15 mW/MIPS
100% code compatible with TI’s single core DSPs based on C64x or C64x+ cores
TI’s SmartReflex™ technology reduces device power consumption by tweaking the
core voltage once the chip is integrated into the system.
Serializer/deserializer (SERDES) interfaces
SGMII Ethernet MAC (EMAC)
Antennae Interface (AIF)
Serial Rapid I/O (SRIO)
Up to 4.8 MB of L1/L2 RAM
TI’s fastest DDR2 memory interface

ECE 5655/4655 Real-Time DSP 1–13

Chapter 1 • Overview of Real-Time Digital Signal Processing

– C66x fixed/float dual-core family

1–14 ECE 5655/4655 Real-Time DSP

 Design considerations: Applications such as:

• Ease of programming shortens • Medical imaging

development time by weeks or • Test and measurement
months resulting in reduced time
to market • Military

ECE 5655/4655 Real-Time DSP

• Audio
• Energy efficient DSP devices as
low as 38mW standby, 467mW • Communications
active power consumption • Industrial
• Floating point instructions for
applications that require high
precision, wide dynamic range

• Connectivity peripherals for BOM

cost reduction, including Ethernet
MAC, UPP, SATA, USB, VPIF, etc.

• Code compatibility across full line

of C6000 DSPs – C64x, C64x+,
C67x and C674x DSPs

• Industrial/automotive temp
support from -40 to 125 deg C
A Brief History of DSP in the Context of Real-Time Processing

1–15
1–16
DSP DSP + ARM DSP/ARM + Video

C645x C6L138/OMAP-L138 DM6467/T

C641x C6L137/OMAP-L137
OMAP3530/25
C674x DM644x
C642x DM3xx
C672x DM643x
C671x DM64x

DSP with focus on DSP with integrated Controller/Host Video and

intensive Processing Imaging
Signal Processing Processors

• Power-efficient fixed/floating pt DSP • Low power (38mW standby; • High performance multi-format
Chapter 1 • Overview of Real-Time Digital Signal Processing

devices; 38mW to 467mW <520mW typical active) fixed video up to 1080p

• High-performance DSP devices; up to and floating point DSP with
8000 MMAC @ 1GHz
• Ideal for Video, Imaging and
ARM core Vision applications
• Excels at communications, industrial,
military, medical, T&M and audio • Ideal for industrial, audio
processing and communications apps

Featured Product

ECE 5655/4655 Real-Time DSP

 A Brief History of DSP in the Context of Real-Time Processing

– C55x fixed-point family

ECE 5655/4655 Real-Time DSP 1–17

Chapter 1 • Overview of Real-Time Digital Signal Processing

Communications and Wireless

• In the past two key application areas, telephone line modems
and cellular voice/data communications, have had a major
impact on DSPS sales
• In 1994 when the V.34 modem standard (28.8 kbps) was
approved fixed point (integer) DSPs became a popular design
route
– The V.34 standard allowed for data rates from 2400 to
28,800 bps
– The modulation scheme at 28,800 is a high-order quadra-
ture amplitude modulation (QAM) employing shell map-
ping, trellis encoding, nonlinear precoding, and soft
decision Viterbi decoding1
– All of the above and backward compatibility (lower rate
schemes) can be performed by one 50 MIPS fixed point
DSP
– In late 1994 these modems cost ~$400, today we know that
28,800 modems (dial-up) are not used much at all
• In the cellular telephone arena fixed point DSPs are widely
used in the portable handsets
– For example the TI C55x series is popular because it has
features such as low power dissipation, integration of

1. Steven Tretter, Communication System Design using DSP Algorithms with

Laboratory Experiments for the TMS320C6713, Plenum Press, New York,
2008.

1–18 ECE 5655/4655 Real-Time DSP

 A Brief History of DSP in the Context of Real-Time Processing

memory and peripherals, integrated Viterbi accelerator,

and small packaging, just to name a few
– Other vendors have tailored their DSP product lines for the
high volume cellular handset markets as well
• DSPs are also utilized in VoIP and cellular basestations
– In a portable device market, such as cell phones, the effi-
ciency of a particular processor can be measured in mW/
algorithm, which is obtained via
mA -
mW - = ------------- MIPS -
---------------------- ---------------------- voltage
algorithm MIPS algorithm
• MIPS/algorithm requirements for an IS95 handset1
Table 1.3: MIPS requirements for an IS95 handset.

Implement
Algorithm MIPS
ation
Correlator Hardware 5
Automatic frequency control (AFC) Hardware 5
Automatic gain control (AGC) Hardware 5
Transmit filter Hardware 30
128-pt FFT Software 1
Viterbi decoder (length 9, rate 1/2) Software 6
Vocoder (8-Kbps Qualcomm code Software 20
excited linear prediction (QCELP))
Vocoder (enhanced variable rate coder Software 30
(EVRC))

ECE 5655/4655 Real-Time DSP 1–19

Chapter 1 • Overview of Real-Time Digital Signal Processing

DSP Hardware Design Issues1

• Hardware alternatives, FPGA, ASIC
• Using a general purpose DSP
• Fixed-Point DSPS
– Fixed-point arithmetic
– Quantization effects and scaling in fixed point
– Typically a lower power design
– Typically a lower cost design
• Floating Point DSPS
– A more rapid design due to the ease of floating point algo-
rithm implementation
– With the powerful C6x family a few floating point calcula-
tions can easily be software emulated on the C62x or C64x
– With the C67x the floating point units are available, so
efficient floating point code can be easily written
• Accessing memory resources
• Integration of peripheral devices
• A very practical alternative today is to consider a hybrid
FPGA/general purpose DSP architecture

1. John Groe and Lawrence Larsen, CDMA Mobile Radio Design, Artech
House, Boston, MA, 2000.
1. Craig Marven and Gillian Ewers, A Simple Approach to Digital Signal Pro-
cessing, John Wiley, 1996. ISBN 0-471-15243-9.

1–20 ECE 5655/4655 Real-Time DSP

 DSP System Design Flow (Marven & Ewers)

DSP System Design Flow (Marven & Ewers)

Application

Define System
Requirements

Select DSP
Device
Software Hardware

Code Writing Schematic

Debug Code Prototype

System
Integration

System test
and Debug

ECE 5655/4655 Real-Time DSP 1–21

Chapter 1 • Overview of Real-Time Digital Signal Processing

• Application Design
– Software design
– Generating assembler source and using C where possible
– Testing the code
– Hardware design
– System integration

Course Laboratory Foundations

• Of primary focus is the ZOOM OMAP-L138/C6748 DSK
(open multimedia application platform = OMAP)
• TI TMS320C6713 DSP Starter Kit (DSK)
– In the past we have used the C6713 DSK with CCS 3.3;
this target is supported by the text, and can be programmed
using CCS 4.2 (in theory CCS 5.3 but have not tried yet)
• In the fixed-point arena, we will also spend some time with
the C55x platform on CCS 5.3, specifically either the C5505
or C5515
• The OMAP3530 based Beagleboard is available for the final
project, for those interested in embedded Linux development
• In the future we may have a second daughter card which is
suitable for software radio work involving in-phase/quadra-
ture (I/Q) sampling and/or zero intermediate frequency (IF)
processing

1–22 ECE 5655/4655 Real-Time DSP

 Course Laboratory Foundations

• Zero IF processing is becoming more popular in wireless

chip sets, so this front-end processing approach and others
will hopefully be part of this class in the future

Software
• TI integrated development environment Code Composer Stu-
dio; CCS 5.3 Platinum (all cores) which supports TI DSPs
and MSP430 devices
– The above IDE is based Eclipse
• MATLAB release 2012b with the following enhancement for
DSP
– Signal processing toolbox
– DSP System Toolbox, (includes Filter Design Toolbox)
– Fixed-Point Toolbox
– The text supported MATLAB Real-Time Interface (docu-
mentation to be released very soon)
• GoldWave or Audacity shareware for .wav audio file manipu-
lation

Test Equipment
• Agilent two-channel 300 MHz digital scope (2 Gsps)
• Agilent function/arbitrary waveform generators; 15 MHz and
80 MHz models
• Agilent 10 Hz – 500 MHz spectrum/vector network analyzers

ECE 5655/4655 Real-Time DSP 1–23

Chapter 1 • Overview of Real-Time Digital Signal Processing

1–24 ECE 5655/4655 Real-Time DSP