This isn't a war thread so please just post experiences and facts and not
stuff like "[Because my mother uses ADI] I use ADI and thats why TI sucks"
kinda crap.
My project is a real time audio effects device that is either 24-bit @ 96khz
or 192khz(depending on how much "room" I have for my algorithms).
The two processors I've been looking at that seem similar are
http://www.analog.com/en/epProd/0,,ADSP-21367,00.html
http://focus.ti.com/docs/prod/folders/print/tms320c6722.html
(
http://www.analog.com/en/epProd/0,,ADSP-BF561,00.html
has 3*MMACS for about the same price)
I'm not sure which to go with. Each one seems to have there pro's and con's
but it seems to be mostly balanced?
What are others experiences with these two?
One thing that I don't really like is that ADI doesn't seem to easily give
out samples for there DSP chips. This is going to make it really hard to
experiment with ;/
I've worked with both TI DSP and ADI DSP; but my experiences are more
from writing software than from hardware design. The TI chips I
worked with are the 'C30 and 'C40 and the ADI chips were in the
ADSP-21xx line -- most recently, the ADSP-218x line. It's been only a
few months since I last did something with the ADI DSP, but it's been
probably close to 8-10 years since I worked on the TI DSPs. Also, my
experiences with ADI DSPs go back to 1990, or so, and have been
somewhat continuous since then. So take that difference of experience
into account.
For the most part, I enjoyed working on both of them. The only issue
that really bothered me about the TI DSPs regarded a problem in
timing. I had crafted a 7-cycle loop which took 11 cycles, instead.
This code was running from cache and should NOT have depended upon
some external memory interface on the actual application board by a
non-TI designer, but should have entirely depended upon the TI chip
and its technical specs, including register interlock waits (which I
also checked thoroughly.) In the end, the 3rd party board designer
team and some technical folks at TI and I were on a conference call
over this with the result that TI, after attempting to suggest (and
failing to show) various areas where I may not have read their docs
well enough, agreed that they had no idea why. The timing was easy to
demonstrate, they had my source code and were able to duplicate it,
and still could not explain it. In short, we never did find out why
and I never did hear from TI about correcting their docs, either. Not
a confidence lifting experience. By comparison, I have NEVER over the
years found anything like that in the ADI DSPs -- they work as spec'd
or else I always get some effort at ADI to explain the details. On
the other side here, when ADI was changing FABs on one line (ADSP-2111
chip), we started getting something like a 40% failure on our own
custom CPU-checkout software (we qualified the chips as they arrived.)
They asked for, and got, a copy of my software for this and they
confirmed the problems and cleaned them up in a few months' time.
Most recently, my experience with TI is with the MSP430 and it's been
a good experience, by and large.
I don't know about these two chips. I would tend to expect ADI's DSP
design to be fairly well-thought out. I would tend to look a little
more closely at TI's design to make sure there weren't holes in it.
But that doesn't mean that TI's design would have holes, only that
because of my experiences earlier I'd be a little less willing to
trust it all to fate and would want to spend a little more time
carefully checking various algorithm ideas against it to be sure they
had captured the more important details well. (But I expect they have
learned well over the years, too.)
