ï~~Still developers and application engineers hesitate to use
general purpose DSP chips because of their
- high price
- low flexibility in architecture
- well known deficiencies with number Iprecision
- low performance for musical applications
Recent generations of fixed point (16 bit wide data format) DSP
chips, now available on the market, like TMS 320C25 (Texas
Instruments), ADSP 2100 (Analog Devices), DSP-X (Hitachi) improved their performance compared to the first generation of DSP
chips.
Until now the best choice for audio signalprocessing was the
Motorola DSP 56000/56001 with 24 bit wide data format, 7
independent internal busses and integrated peripherals like
direct memory access (DMA), serial interfaces (SSI, SCI).
Besides the 5600X two other DSP chips with floating point
arithmetic ADSP 2100, NEC 77230 offer sufficient number precision
but have disadvantages in speed and architecture.
With floating point arithmetic the now introduced 96001/96002 has
extended memory space, on chip DMA controler and parallel network
links.
Compared to other recently introduced devices (i.e. TMS 320C30)
these features suit best the Motorola chip for multi-signalprocessing in audio applications.
2. Advances of the 9600X Family
In recent developments the DSP 5600X was applied (Snell
1987).
In the following table we compare some of the important properties of the 5600X and the 9600X:
Features DSP 5600X DSP 9600X
Instr. cycle time
Data format
Address size
Internal busses
On chip I/O
Multiplier size
1024 complex FFT
50nsec / 20MHz
12.25 Mips
24 bit data
fixed point
16 bit address
64K byte space
4 data, 3 address
1 DMA port
3x8 bit wide
24 x 24
6 msec
38nsec / 38MHz
13.3 Mips
32 bit data
fixed point and
float, point (IEEE 754)
different rounding modes
32 bit address
4G byte space
5 data, 3 address
2 channel DMA-controller
32 bit wide
32 X 32
2 msec
259
ICMC Proceedings 1988
