28-07-2012, 03:51 PM
TMS320 DSP Development Support Reference Guide
datasheet.pdf (Size: 2.03 MB / Downloads: 85)
Introduction
Since the first TMS320 DSP was introduced in 1982, Texas Instruments has
been dedicated to the advancement of digital signal processing technology and
its applications. TI recognizes that fast time to market, increased productivity,
and design ease are of primary importance in the development of DSP-based
applications. Therefore, TI offers an innovative, comprehensive program of
development support for TMS320 DSPs to facilitate the design process from
system concept to production.
From the Real World: DSP Technology
The natural environment that we wish to interpret is analog. Signals vary continually
with time and may feasibly take any value. The digital domain is fixed
between two values: high and low. Conversion from the analog domain to the
digital domain (and back) is essential to sense signals, manipulate them, and
reintroduce them to the environment. The advantages gained from digital
manipulation justify the conversion. Special devices that perform this conversion
are manufactured to interface with DSPs. These devices are called analogto-
digital (A/D) and digital-to-analog (D/A) converters. A binary code is assigned
to a periodic sample of an analog signal that is in proportion to its magnitude;
this gives a digital representation (see Figure 1–1).
DSP Architecture
All DSPs consist of several fundamental modules: a digital signal processing
core to perform mathematical operations, memory to store data and program
instructions, and possibly a mixed-signal product to converse between the
analog and digital worlds.
As a stored-program machine, the processor must be told what to do every
clock cycle. Typically, a DSP fetches an instruction and some data from
memory, operates on these, and then returns the manipulated data to storage.
The way this is conducted is not the same for all processors. Two different architectures
can be identified: Von Neumann and Harvard (see Figure 1–2).
The DSP application, in addition to the memory and peripheral configuration,
usually governs the type of architecture employed.
The Need for Speed
The main concern for real-time algorithms is the amount of processing that can
be done before a new sample arrives. DSP-type algorithms are generically of
the form involving a multiply and add operation:
A = BC + D
The addition function is quite simple for conventional computers and can be
performed in a single clock cycle. The same is true of subtraction. Most computers
subtract by negating one number and then adding it to the other. Multiply
functions take much longer, especially when considering numbers such as
p. A general-purpose processor may take several hundreds of clock cycles to
implement such a calculation. A machine is needed that can perform a multiply
and an add in just one clock cycle. This requires an architecture molded to the
specific application. DSPs have hardwired units within the processors for completing
a multiply and add within one single clock cycle. Because such multiply
and accumulate (MAC) instructions are the fundamental building blocks for
many DSP applications, efficient execution is imperative.