If you're
just talking to 1 or 2 peripherals (say you want to hook up a
GPIB plotter, or you want to get the data off an HP disk drive) you can
get away without the special buffers. You can use open-collector TTL
gates to drive the bus, and TTL schmitt triggers ('14s, for example) as
receivers. Like many things I'd not do this on a 'production' board, but
it works as a hack.
Interesting... I had not thought of substitute line drivers, but it makes
sense... unless you are trying to build a gargantuan PET system (the kind
nobody could afford "back in the day"), most people had one disk drive box
(2040/4040/2031/8050...) and _maybe_ a printer. These days, though, the
most common thing for Commodore owners would be to hook up a disk drive
to a modernish machine to extract floppies... using O.C. drivers and
Schmitt trigger receivers would work just fine for that.
Indeed it does. I built an IEEE-488 interface for a P850 like that
(that's where I used a couple of state machines to do the talker and
listener handshakes). In fact I used it to link up a Commodore 8050 drive
-- it was a little worrying that the drive had more memory and processor
power than the mini...
One other thing. The GPIB on Commodore machines, being software driven,
was fairly slack about timing. Some 3rd party manufacturers 'got away
with murder' here -- One classic example was combinatorial logic only,
and totally ignored timing requiremnets. Worked fine on a PET, failed
totally on everything else.
I have plenty
of GPIB-capable machines from handheld calculators up to
workstations (e.g. HP41 + HPIL + HP82169 translator in the first case,
PERQ in the second). I even have the Acorn interface for the Beeb.
Nothing spare, though
With one exception, all of my GPIB hardware is C= or third-party for
C= CPUs... I do have one HP 9-track tape drive with a GPIB interface,
but I've never even powered it up. I should Google for some information
on it when I get home... I have no idea how much buffer space it has
onboard, or if you could drive it at the incredibly slow software-driven
C= data rates, but it'd be cool to put a 9-track on a PET! I'm sure
FWIW, I have a Dylon 9 track controller with a GPIB host interface, and
that does have enough buffer RAM on-board to work with a slow controller
(the example programs in the manual are in HPL for the 9825 of all
things). I've used that on the HP71 + HPIL + HP82169..
-tony