jnc at mercury.lcs.mit.edu
Wed Jul 22 09:12:53 CDT 2015
> From: Guy Sotomayor
> I haven't looked, but I think the Unibus and QBus are comparable in
> terms of pin count.
Well... the QBUS does have those shared address/data lines, so its 'raw'
count (i.e. number of wired-or broadcast bus lines) is somewhat lower (-16
data pins, but +4 address lines, plus various other differences).
Depending on the details that may, or may not, translate into a lower pin
count to the FPGA. E.g. if one has separate input and output pins for
address/data lines (one needs all 4 functions for doing both incoming slave
mode, and DMA), it can add up.
So, I noticed that on some QBUS cards DEC used a quad transceiver with
tri-state output (on the card side), the AM2908PC. It has separate tri-state
drive enable, and bus drive enable, pins. The FPGA we were looking at
supports bi-directional pins (i.e. tri-state output, plus input), so I
conceived the hack of tying the transceiver's input and output pins together,
and routing them into bi-directional pins on the FPGA. As long as one doesn't
turn on the tri-state drive and the QBUS drive enable at the same time
(forming a feedback loop at the transceiver chip), it _should_ work OK.
(Mandatory observation about theory and practise included at this point...)
So that reduces the number of pins needed by half. (Although I guess one
could pull the same hack with the UNIBUS, _iff_ the 2908 can work as a UNIBUS
transceiver; analog electrically, the two seem to be pretty identical.)
One can further observe that on the QBUS, A and D are shared, so one doesn't
need separate A and D pins. One cannot do that with the UNIBUS, since they
way DATO works, the A and D lines have to be driven at the same time (with
different values, of course).
More information about the cctalk