Re: DSP

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

In article <2c15mj$2qc@ionews.io.org>, taob@ionews.io.org (Brian Tao) writes:
> In article <2brien$c41@cnn.sim.es.com>, Ken Poppleton writes...

> > The DSP processor board will have a Motorola DSP56002 40Mhz processor
>                                                ^^^^^^^^
>     Isn't this the expensive 32-bit model?  I thought you had decided to
> use the 24-bit 56001 to keep costs down?


No - The dsp56002 is a new version of the 56001 with a few added instructions,
a serial boot option, a non-maskable interupt, and a debug port.  It is also
a 40Mhz part, and should get faster in the future.  And on top of all that
it costs less than the 56001.  But it is not pin compatable so I had to 
do some redesign.  I think it is worth it, espically if we end up with a
faster processor in the future!  The 56001 will not get any faster than 33 Mhz!

> > very good for integer math but is about 8 times slower than the 68882
> > floating point coprocessors at doing floating point math.

>     Can you integrate a 68882 socket in the design, or as an add-on
> module where the FP operations could be re-routed?  Or is that way too
> expensive?

No - sorry, as there is not enough room on the board,  I wish I could add
all the options I wanted, like more RAM.  Maybe there would be room on the
interface card to the dsp processor card, espically if the interface card was
a simple memory mapped version without DMA and interupt support.  This is 
why the interface is seperated from the DSP processor board, so anyone can 
make a better interface card (or processor card).


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