General-Purpose DSP Processors - Signal Processing Systems - page 765

2 Arithmetic Type

DSP processors are divided as fixed-point and floating-point processors based
on the type of arithmetic units in the processor. In the early days of DSP
processors, fixed-point processors with 16-bit word width were used, e.g., Texas
Instruments TMS32010 [ 32 ] andNEC

PD7720 [ 35 ] . While the 16-bit data was
sufficient for speech applications, fairly soon it was realized that some applications
require higher accuracy and 24-bit processors were introduced, e.g., Motorola
DSP56000 [ 26 ] . In addition, floating-point DSP processors were introduced, e.g.,
Hitachi HD61810 [ 17 ] , AT&T DSP32 [ 19 ] , and NEC

μ

PD77230 [ 34 ] .
The floating-point processors contain more complex logic and, therefore, con-
sume more power and are more expensive. However, the floating-point processors
are easier to program as the dynamic range in floating-point representation is larger
and there is no need to scale and optimize the signal levels during intermediate com-
putations. Furthermore, high-level languages have floating-point data types while
integers are the only supported fixed-point data types although signal processing
calls for fractional data types for fixed-point arithmetic.

μ

3 Data Path

The actual signal processing operations in a processor are carried out in various
functional units, such as arithmetic logic units (ALU) or multipliers, and the
collection these units is called as data path or data ALU. In order to store
intermediate results, data path contains also accumulators, registers, or register files.
The data path in DSP processors can be expected to be tailored for computations
inherent in typical DSP algorithms. DSP processors may also be enhanced with
special function units to improve performance for a group of applications. In the
following sections, principal features of fixed-point data paths are studied and
differences in floating-point data paths are discussed.

3.1 Fixed-Point Data Paths

A typical fixed-point data path includes multiplier, ALU, shifters, and registers
and accumulators. Data path is used mainly for signal processing operations, while
computations related to memory access, which call for integer arithmetic, are often
performed in separate ALUs called as address generation units or address arithmetic
units.