[comp.dsp] TMS32031 - vs - Motorola56001, information requested

moorthy@saturn.ucsc.edu (Harihara Moorthy) (03/08/91)

Hi:

We are considering developing a board for voice applications, and have
narrowed down the choice of the dsp processor to TMS32031 and Motorola56001.
We are inclining towards 32031 for the following reasons: faster, both
floating and integer arithmetic, more on-chip memory, and possibly
cheaper.  We would like the chip to reasonably meet our future needs which
might include voice recognition, fax transmission etc.  I would appreciate
your input on how these two chips compare.

Also, can someone identify applications where floating point dsp processor
will be esspecially needed.

Finally, what are the advantages and disadvantages of floating point
calculations vs fixed point calculations?

Thanks.

Please email your responses (or post) to moorthy@saturn.ucsc.edu
If there is enough interest, I will summarize.

H. Moorthy

--
 ___    
/ (_)      o     o        ___
| .   _   ___   ___     _|_  \ Harihara SV Moorthy
| |  | | | |  _|_|_   _|_|_  | Univ. of California, Santa Cruz

aep@world.std.com (Andrew E Page) (03/20/91)

In article <13192@darkstar.ucsc.edu> moorthy@saturn.ucsc.edu (Harihara Moorthy) writes:
>
>Hi:
>
>We are considering developing a board for voice applications, and have
>narrowed down the choice of the dsp processor to TMS32031 and Motorola56001.
>We are inclining towards 32031 for the following reasons: faster, both
>floating and integer arithmetic, more on-chip memory, and possibly
>cheaper.  We would like the chip to reasonably meet our future needs which
>might include voice recognition, fax transmission etc.  I would appreciate
>your input on how these two chips compare.

  I don't know about cheaper.  The last time I was princing C30 chips 
they were running $270.  in quantity.  However, the C31-27 could be 
significantly cheaper.  Call your local reps about that.  

   My preference is towards the C3x as well.  Much of my DSP expertise
was cut with this device.  IN addition I like the feature of the dual
busses in the same memory address space.  This saves a lot of hassles
in coding and design.    
    I'm not sure, but I do not think that the 56K supports the following 
features that the C30 does:

	a)  parallel instructions
        b)  serial ports (2 for the C30, 1 for the C31)


>
>Also, can someone identify applications where floating point dsp processor
>will be esspecially needed.

  Graphics, particularly maldebrot fractals.  




-- 
Andrew E. Page (Warrior Poet)   |   Decision and Effort The Archer and Arrow
Concepts Engineering            |     The difference between what we are
CIS:70202,234  BIX:page1        |           and what we want to be.

bhanafee@deimos.ads.com (Brian Hanafee) (03/20/91)

In article <1991Mar19.183918.6460@world.std.com> aep@world.std.com (Andrew E Page) writes:

>In article <13192@darkstar.ucsc.edu> moorthy@saturn.ucsc.edu (Harihara Moorthy) writes:
>    I'm not sure, but I do not think that the 56K supports the following 
>features that the C30 does:
>
>   a)  parallel instructions
The 56k basically allows a parallel data move instruction.  It's
mainly useful for setting up registers in advance;  the parallel moves
often get bottlenecked by the data bus anyway.

>   b)  serial ports (2 for the C30, 1 for the C31)

The 56k has 2 serial ports and a parallel port.  It can generate baud
rates internally.  Actually, I think the interfacing capabilities are
one of its strong points.  The Motorola documentation of those
capabilities is not one of its strong points.


Brian Hanafee

carroll@ssc-vax (Jeff Carroll) (03/20/91)

In article <1991Mar19.183918.6460@world.std.com> aep@world.std.com (Andrew E Page) writes:
>
>    I'm not sure, but I do not think that the 56K supports the following 
>features that the C30 does:
>
>	a)  parallel instructions
>        b)  serial ports (2 for the C30, 1 for the C31)

a)	the 56000/56001 supports "parallel moves", i.e., the move
	instructions can under  some circumstances be executed in parallel
	with ALU ops.

b)	the 56000/56001 has three reconfigurable I/O ports, two of which
	can be used as serial ports.



-- 
Jeff Carroll
carroll@ssc-vax.boeing.com

svoboda@motcid.UUCP (David Svoboda) (03/21/91)

From article <1991Mar19.183918.6460@world.std.com>, by aep@world.std.com (Andrew E Page):
>    My preference is towards the C3x as well.  Much of my DSP expertise
> was cut with this device.  IN addition I like the feature of the dual
> busses in the same memory address space.  This saves a lot of hassles
> in coding and design.    

If I follow you right, I think you are mistaken on this about the 56K.
To save silicon, you can do as many of us do and use a single memory 
space addressed by X, Y, and P memory.

>     I'm not sure, but I do not think that the 56K supports the following 
> features that the C30 does:
> 
> 	a)  parallel instructions

Yes, it most definitely does.  By this I presume you mean allowing an ALU or
multiply-accumulate instruction in parallel with two memory reads or writes.
This it will do.

>         b)  serial ports (2 for the C30, 1 for the C31)

The DSP56001 has two serial ports and a host interface port for use in
communicating with a master processor.  One serial port is asynch, the
other synch.  The host port supports addressable, networked DSPs.

Note: I work for Motorola Cellular, not DSP Operations,
 so I only USE the 56K.

           Dave Svoboda, Motorola CID, RTSG, Arlington Heights, IL
               uucp => {uunet|mcdchg|att}!motcid!svoboda
  Don't feel paranoid; when we're out to get you, you won't have the vaguest 
    idea what's going on.  Corollary:  When you feel safe, that's when you 
                             should feel paranoid.