Re: New ][ Hardware - DSP board (soon!)

[kpopple@imp.sim.es.com (Ken Poppleton)]

In article <1mhe1fINN5u3@charnel.ecst.csuchico.edu>, jamesb@ecst.csuchico.edu (James L. Brookes) writes:
> In article <C2zDC7.BHG@javelin.sim.es.com> kpopple@imp.sim.es.com
> (Ken Poppleton) writes:
> 
> >Processor: Motorola DSP56001
> >
> >Clock speed: 20Mhz  (10 MIPS)  possibly a 33 Mhz version? (16Mips)
> 
> I believe the 56ks the NeXT used were prototypes, clocked at 20.48 or so
> Mhz.  The production version of the 56001 is 33mhz, so I'd hope that's
> what you would use...

Ths DSP56001 is supplied in 20, 27 and 33 Mhz versions.  All of these are
production versions.  If the board operates cleanly it will support
any of these clock speeds.  But first I will get the 20 Mhz version
working and then see if a 33 Mhz version will work.    Also note that
the 33 Mhz version will increase the price.  I would prefer to promise
a slower board and then increase its speed to the limit after it is
working.

> 
> >Memory: Total of 9K words (24 bit) divided into:
> >	dsp56001 internal memory 512 words program (P),
> >				 256 words X data
> >				 256 words Y data
> >	external 8k words usable as 8k in P,
> >				    or 4k in each of X or Y
> >	external 8kx8 Eprom or EEprom in P, X, or Y memory
> [possible future things]
> >	Design an expansion memory board.
> 
> Wouldn't it be better to have sockets for SIMMs right on the board?  I
> can think of a lot of applications where going through 1mhz to get to
> the memory would bog the DSP down tremendously.  Having 1 SIMM socket
> capable of 256, 1, or 4mb on board would be a very cool option, IMHO.
> 
> James Brookes

The 1Mhz Apple interface will usually be used for downloading
programs, and then sending commands to the dsp and getting responses.

You are correct that the 1 Mhz interface to the Apple is slow for the
DSP but the onboard RAM (8Kx24) is on the DSP local bus.  This is 35
nsec RAM which is much faster than SIMMs (DRAMS most of which are
around 80 nsec).  So using DRAM would also slow the DSP.  I do not
expect the programs for the DSP to be very large, mostly written
in assembly.  As an example I have written a servo controler for a 3
axis servo system using less than 8k bytes which is 3K words of
memory. (That is nearly 3000 lines of assembly!).  And that took
two years to complete working full time!

For data storage some large memory would be useful, so that is why 
I want to do an expansion memory board.  But if you expect to see
the DSP board this year, it will not have DRAMs on the board.  Maybe
if there is enough interest I could do another DSP board designed
targeted to the Audio signal processing needs (or other uses).

It all depends on what you wish to use a coprocessor for.

-----------------------------------------------------------------------
Ken Poppleton kpopple@imp.sim.es.com
-----------------------------------------------------------------------