Here is the information about DMA issues and the A2630/A2000/A2091. I never could get this combination to be reliable until I added the resistor mentioned in the text. It is a 10 minute 25 cent fix that really makes a HUGE difference. This worked on a Rev 6.2 2000 as well.
-Jeff
Subject: Re: A2630 fix question
> I had lots of problems with my 2000/Rev4.3, A2091, A2630Rev9.2
> setup until someone told me, that there were some resistor packages
> soldered in the wrong orientation on the A2630 (RP104-RP107). I
> looked at the board and I saw, that they had their common pin
> on the other side than the drawing on the board (now I know what
> it is good for:))
That's odd...
> My question: Was there any good reason for turning the RPs or was it
> just an error that happened during the production of the board?
I suppose it might have been intentional. I didn't do it that way, but
I didn't really touch the A2630 after Rev 6. Pretty much everything
done after that was to FCC certification. It's possible that these
resistors, which I believe I had in there for data bus pullups, were
switched around to act as pulldowns. If you lay out the resistor pack
footprint right, you have +5V at pin 1, Ground on the last pin, and so
you can get "proper" behavior with either orientation. I've been been
a big fan of pulldowns on bipolar logic (typically, bipolar output
stages can drive a great deal of current low, but not so much high),
but they do tend to limit the peak voltage, and occasionally over-peak
spikes, when a signal swings high. This is good for FFC certification,
and not bad for typical system operation in most cases. Apparently you
have found one of the cases in which pulldowns can be a problem.
Or perhaps you're trying a complex fix for a simple problem already
solved in most systems. I don't know what you had failing, but if it's
DMA between the A2091 and A2630, there's a known problem on some A2000
motherboards. Certain vendors versions of the 74ALS245 chip used at
U605 on the A2000 motherboard cause a problem, a potential glitch on
the BAS* (buffered address strobe) line of the Zorro II bus, during
DMA turnaround (when the A2091 drops the bus and the A2000 logic picks
it back up). This glitch is nicely shaped into a pulse by the ALS245's
schmitt input, but that pulse on AS* is ignored on the motherboard,
it's not long enough to usually affect the 7MHz logic on the A2000. It
is long enough to mess with the 25MHz logic on the A2630, however. To
fix this, you add a 1K pullup between pins 11 and 20 on U605 (which
you'll find between the CPU and first Zorro slots).
Maybe changing the A2630 resistor packs to pullups helps, too, I never
really looked at the problem. But this resistor fix makes a very big
difference, you might want do it even if your system is working find
now, just to be safe.
-Dave Haynie
