22 May
2009
22 May
'09
6:46 a.m.
Scanning wrote:
Thanks for the tip. The guy was delusional if he
thought he could get $300
for that setup. It ( 2920 ) was one ( of many ) of Intel's big screw-ups. As
a DSP chip, all you could do was download your "stuff" into the EPROM and
wiggle the input with an analog input and observe the analog output. No way
to see inside and what was happening inside the chip. Made development and
troubleshooting a real bitch ( which probably speaks volumes for its success
in the market ).
I disagree about that being the problem, as basically all
microcontrollers were that way at the time. In fact, many
microcontrollers didn't even *have* EPROM versions, so development for
those required an expensive in-circuit emulator. The 2920 gave you the
choice: do many iterations with the EPROM programmer and eraser, *or*
buy the ICE.
In my not so humble opinion, the reason that the 2920 didn't catch on
but the TI TMS32010 did is that the 2920 didn't have hardware multiply
and accumulate, so it had very low performance for common DSP tasks such
as digital filters and FFT.
Another problem Intel has had from 1980 to the present is that while
they want to play in multiple markets, it is almost always the case that
they make the most money per wafer by making mainstream processors.
That results in few wafers being devoted to other products, which is why
they seem to have such on-again-off-again support for embedded
processors. On the other hand, that worked to their advantage when DRAM
prices dropped through the floor in the mid-1980s, but the managers
didn't have the balls to drop the DRAM product line. Their normal
method of allocating wafer starts naturally resulted in few wafers for
DRAM, minimizing their losses on it until management had the sense to
drop it entirely.
Eric