[comp.dsp] Processor choice & Thanks

bointon@kcl-cs.UUCP (Marcus Barnaby Bointon) (10/17/90)

  Thanks to all those that replied to my plea!

I am faced with the question of which processor to use for my final year DSP
board. I seem to have narrowed the choice to TMS320x0 or MC56000 series.
Which are better/easier to use etc? I will be using it for audio with the
required converters, and would want to use it for housekeeping on the board
too (i.e. rs232) so I need a reasonable address space (>64k).
I only have docs for the TMS's so far- what can anyone tell me about the
56k series? (just an outline!)
Are there any other alternatives? Would I benefit in using the 56k as I
have a 68k background?

Thanks again...

<<<<<<<<<<<<<<<<<<<<<<<<<<<<<=Marcus Bointon=>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>
<<  Consultant, Campaign for the Reduction of Entropy (C.R.E. Group(U.K.))  >>
<<    Campaign HQ: JANET%ZDAC810@UK.AC.KCL.CC.OAK or bointon@kcl.cs.UUCP    >>
<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>
<< Disclaimer: This message has been treated with flame-resistant chemicals >>
<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>

todd@ivucsb.sba.ca.us (Todd Day) (10/19/90)

bointon@kcl-cs.UUCP (Marcus Barnaby Bointon) writes:

%I am faced with the question of which processor to use for my final year DSP
%board. I seem to have narrowed the choice to TMS320x0 or MC56000 series.

Well, I've used the TMS32010 and the DSP 56001.  This is not really a fair
comparison, since the 32010 is clearly a first generation chip, and the
56001 is probably third or fourth.  Anyway, I hated using the 32010.  If
you must use TI, use one of their later generation chip.  I found the
command structure very difficult to use (it would make you write inefficient
code).

%Would I benefit in using the 56k as I
%have a 68k background?

Yes, most definitely.  This is one of the things I like about the chip.
I really enjoy programming all of Motorola's line, from the 6800 series
to the 68000 series to the 56000.

-- 
Todd Day |   todd@ivucsb.sba.ca.us   |  ucsbcsl!ivucsb!todd
    Pleasure leaves a fingerprint, As surely as mortal pain
    In memories they resonate, And echo back again. -- Rush

stambaug@hpspkla.spk.hp.com (Mark A. Stambaugh) (10/24/90)

>I am faced with the question of which processor to use for my final year DSP
>board. I seem to have narrowed the choice to TMS320x0 or MC56000 series.
>Which are better/easier to use etc? I will be using it for audio with the
>required converters, and would want to use it for housekeeping on the board
>too (i.e. rs232) so I need a reasonable address space (>64k).
>I only have docs for the TMS's so far- what can anyone tell me about the
>56k series? (just an outline!)
>Are there any other alternatives? Would I benefit in using the 56k as I
>have a 68k background?

Your large address space requirement rules out the MC56000 series unless
you use paging techniques (yech!).   The 56000 only has a 64K word address 
space.   I believe that you will have to go to the 96000 or the TMS320C30 
generation to go beyond 64K words of program space.

Even if you can live with 64K of program space, I would not recommend
using the 56000 due to its development system.   It sounds like you
will be writing alot of code, so this factor should be considered.
The 56000's C compiler isn't too bad, but there is a significant learning curve
to figure out how to get around its eccentricities.   Contrary to Motorola's
claim, it is NOT a full implementation of K&R C.   The 56000 does have 
an ADS board that acts as a crude in-circuit emulator.   Unfortunately it 
only has a monitor program, so you are forced to single step through
a program segment to see what's happening.   An HP 16500 logic analyzer can
be used to inverse assemble traces taken from the external bus, but
all internal transfers are hidden.   Any connectivity between the two system
must be implemented by you.  

However, the MC56000 shines when doing typical audio processing due to its
24 bit words.   As long as you aren't writing more than about 4K words of 
program space, this is probably a good choice.   With larger programs, you
will be fighting the development system too much.   I should probably 
mention that I am using the PC version of their tools, which might be the
cause of many of my peeves. 

The TI processors seem to have alot more public domain code available.
The software tools from TI seem to have been thought out more than the 
tools from Motorola, and there are very good development systems available 
for the series.    

Finally, I don't think you will be able to leverage your 68000 knowledge
significantly more by using the 56000 vs the TI product.   The 68K and 56K
processors are very different animals....totally seperate instruction sets,
different word widths, and different math.   The 56K uses fractional math
which requires a little getting used to at first.   

Disclaimer: These opinions are my own and do not necessarily reflect the
opinion of my employer.

Mark Stambaugh
stambaug@hpspkla.hp.com

golds@fjc.GOV (Rich Goldschmidt) (10/30/90)

> Your large address space requirement rules out the MC56000 series unless
> you use paging techniques (yech!).   The 56000 only has a 64K word address 
> space.   I believe that you will have to go to the 96000 or the TMS320C30 
> generation to go beyond 64K words of program space.
> Mark Stambaugh
> stambaug@hpspkla.hp.com

I have been considering using the 96000 as a fast floating point coprocessor
for some general purpose computing which is floating point intensive.  Does
this make sense, and does the compiler support the needs of general purpose 
computing with standard libraries, including I/O (like printf and file 
access) and math functions?  Any suggestions on sources for coprocessor
boards and alternatives to the 96000 for general floating point?  Thanks.


-- 
Rich Goldschmidt: uunet!fjcp60!golds or golds@fjc.gov
Commercialization of space is the best way to escape the zero-sum game.