Madisetti Georgia Institute of Technology 77 Introduction to the TMS320 Family of Digital Signal Processors Panos Papamichalis Introduction • Fixed-Point Devices: TMS320C25 Architecture
Trang 1DSP Software
and Hardware
Vijay K Madisetti
Georgia Institute of Technology
77 Introduction to the TMS320 Family of Digital Signal Processors Panos Papamichalis
Introduction • Fixed-Point Devices: TMS320C25 Architecture and Fundamental Features •
TMS320C25 Memory Organization and Access •TMS320C25 Multiplier and ALU•Other
Ar-chitectural Features of the TMS320C25 •TMS320C25 Instruction Set•Input/Output Operations
of the TMS320C25 •Subroutines, Interrupts, and Stack on the TMS320C25•Introduction to the
TMS320C30 Digital Signal Processor •TMS320C30 Memory Organization and Access•Multiplier
and ALU of the TMS320C30 •Other Architectural Features of the TMS320C30•TMS320C30
Instruction Set •Other Generations and Devices in the TMS320 Family
78 Rapid Design and Prototyping of DSP Systems T Egolf, M Pettigrew, J Debardelaben,
R Hezar, S Famorzadeh, A Kavipurapu, M Khan, Lan-Rong Dung, K Balemarthy, N Desai, Yong-kyu Jung, and V Madisetti
Introduction •Survey of Previous Research•Infrastructure Criteria for the Design Flow•The
Executable Requirement •The Executable Specification•Data and Control Flow Modeling•
Ar-chitectural Design •Performance Modeling and Architecture Verification•Fully Functional and
Interface Modeling and Hardware Virtual Prototypes •Support for Legacy Systems•Conclusions
THE PRIMARY TRAITS OF EMBEDDED signal processing systems that distinguish them
from general purpose computer systems are their predictable reactions to real-time1stimuli
from the environment, their form- and cost-optimized design, and their compliance with
required or specified modes of response behavior and functionality [1].
1 Real-time indicates behavior related to wall-clock time and does not necessarily imply a quick response.
Trang 2Other traits that they share with other forms of digital products include the need for reliability, fault-tolerance, and maintainability, to name just a few An embedded system usually consists
of hardware components such as memories, application-specific ICs (ASICs), processors, DSPs, buses, analog-digital interfaces, and also software components that provide control, diagnostic, and application-specific capabilities required of it In addition, they often contain electromechanical (EM) components such as sensors and transducers and operate in harsh environmental conditions Unlike general purpose computers they may not allow much flexibility in support of a diverse range
of programming applications, and it is not unusual to dedicate such systems to specific application Embedded systems, thus, range from simple, low-cost sensor/actuator systems consisting of a few tens of lines of code and 8/16-bit processors (CPU) (e.g., bank ATM machines) to sophisticated high-performance signal processing systems consisting of runtime operating system support, tens of x86-class processors, digital signal processing (DSP) chips, interconnection networks, complex sensors, and other interfaces (e.g., radar-based tracking and navigational systems) Their lack of flexibility may be apparent when one considers that an ATM machine cannot be easily programmed to support
additional image processing tasks, unless upgraded in terms of resources Finally, embedded systems
typically do not support directuser interaction in terms of higher order programming languages
(HOLs) such as Fortran or C, but allow users to provide inputs that are sensor- or menu-driven The debug and diagnostic interfaces, however, support HOLs and other lower level software and hardware programmability
Embedded systems in general may be classified into one of the following four general categories of products The prices are indicative of the multi-billion dollar marketplace in 1996, and their relative magnitudes are more significant than their actual values The relationship of the categories to dollar cost is intentional and is an early harbinger of the fact that underlying cost and performance tradeoffs motivate and drive most of the system design and prototyping methodologies
Commodity DSP Products: High-volume market and valued at less than $ 300 a piece These include
CD players, recorders, VCRs, facsimile and answering machines, telemetry applications, simple signal processing filtering packages, etc., primarily aimed at the highly competitive mass-volume consumer market
Portable DSP Products: High-volume market and valued at less than $ 800 These include
portable and hand-held low-power electronic products for man-machine communications such as DSP boards, digital audio, security systems, modems, camcorders, industrial controllers, scanners, communications equipment, and others
Cost-Performance DSP Products: High-volume market, and valued at less than $ 3000 These
products trade off cost for performance, and include DSP products such as video teleconferenc-ing equipment, laptops, audio, telecommunications switches, high-performance DSP boards and coprocessors, and DSP CAD packages for hardware and software design
High-Performance Products: Low-to-moderate volume market, and valued at over $8000 These
products include high-end workstations with DSP coprocessors, time signal processors, real-time database processing systems, digital HDTV, radar signal processor systems, avionics and military systems, sensor and data processing hardware and software systems This class of products contains
a significant amount of software compared to the earlier classes, which often focus on large volume, low-cost, hardware-only solutions
It may be useful to classify high-performance products further into three categories
• Real-time embedded control systems: These systems are characterized by the following features:
interrupt driven, large numerical processing requirements, small databases, tight real-time constraints, well-defined user interface, requirements and design driven by performance re-quirements Examples include an aircraft control system, or a control system for a steel plant
• Embedded information systems: These systems are characterized by the following features:
transaction-based, moderate numerical/DSP processing, flexible time constraints, complex
Trang 3user interfaces, requirements and design driven by user interface Examples include accounting and inventory management systems
• Command, control, communication, and intelligence (C4I) systems: These systems are
charac-terized by large numerical processing, large databases, moderate to tight real-time constraints, flexible and complex user interfaces, requirements and design driven by performance and user interface Examples include missile guidance systems, radar-tracking systems, and inventory and manufacturing control systems
These four categories of embedded systems can be further distinguished in terms of other met-rics such as computing speed (integer or floating point performance), input/output transfer rates, memory capacities, market volume, environmental issues, typical design and development budgets, lifetimes, reliability issues, upgrades, and other lifecycle support costs Another interesting fact is that the higher the software value in a product, the greater its profitability margin Recent studies
by Andersen Consulting have shown that profit margin pressures are increasing due to increasing semiconductor content in systems’ sales’ values In 1985, silicon represented 9.5 percent of a system’s value By 1995, that had shot up to 19.1 percent The higher the silicon content, the greater the pressure on margins resulting in lower profits In PCs, integrated circuit components represent 30
to 35 percent of the sales value and the ratio is steadily increasing More than 50 percent of value of the new network computers (NCs) is expected to be in integrated circuits In the area of DSPs, we estimate that this ratio is about 20 percent
In this section, the chapter “Introduction to the TMS320 Family of Digital Signal Processors" by Panos Papamichalis, outlines the programmable DSP families developed by Texas Instruments, the leading organization in this area In, “Rapid Design and Prototyping of DSP Systems", T Egolf,
M Pettigrew, J Debardelaben, R Hezar, S Famorzadeh, A Kavipurapu, M Khan, L.-R Dung, K Balemarthy, N Desai, Y Jung, and V Madisetti, discuss how signal processing systems are designed and integrated using a novel top down design approach developed as part of DARPA’s RASSP program
References
[1] Madisetti, V K.,VLSI Digital Signal Processors, IEEE Press, Piscataway, NJ, 1995.