Date: Tue, 16 Oct 2012 16:26:30 -0400
From: David Riley <fraveydank at gmail.com>
Reply-To: "General Discussion: On-Topic and Off-Topic Posts"
<cctalk at classiccmp.org>
To: "General Discussion: On-Topic and Off-Topic Posts" <cctalk at
Subject: Re: unibus to modern disk interface?
On Oct 16, 2012, at 3:48 PM, Seth Morabito wrote:
* On Sun, Oct 14, 2012 at 07:47:45PM +0100, Rob
Jarratt <robert.jarratt at ntlworld.com> wrote:
I have heard the problem with driver chips before. I don't know enough
electronics really, but could it be done with discrete components?
It could be, yes, at the expense of a lot of board space.
Given the size of modern transistors, less than you might think (if you
have a single transistor for output, you're talking about just about 1cm
above the fingers, really).
One of the biggest issues with the DEC busses
nowadays is that they use
open collector drivers. What that means is that the signal lines in the
bus are normally held at something like 3.4V (a logical "on") by
resistors at both ends of the bus that terminate the bus and form a
voltage divider. The chips are all TTL, so if you were to just directly
connect a TTL-level driver to the bus and assert a "high" logic level,
it would try to pull the bus to +5V and you would be in a world of
trouble. The open collector drivers don't ever do that, they can either
pull the line down to ground, or they can act like an open circuit, but
that's it. That's how signaling on the DEC bus works - a bunch of open
collector drivers share the same line, and any of them can pull the line
down to ground, safely.
That's not really the problem. There are plenty of open-collector
TTL chips; the ones that I've found come closest to what you'd want
(for QBUS, at least) are the 74AS640 and 74AS760. They're both
heavy-hitting octal open-collector bus drivers (one is unidirectional,
the other bidirectional). The primary attractive feature to me is
the 64mA of current handling, which comes pretty close to the 70mA
max I've seen in all the QBUS-related material so far. Their min
fall time is a little too short, though. And their thresholds
aren't quite where the DEC chips were, though they're probably close
enough for most systems out there. As Allison pointed out, plenty
of third-party manufacturers used '244s and '245s with success.
The old chips that were designed for this purpose
also had other
electrical requirements, especially the amount of current they could
sink when pulling the bus to ground.
There are some open collector drivers still made,
but not as many as you
would think. And certainly nothing I know of that is in a single
convenient package like the DEC or National or Signetics chips were.
There are some off-the-shelf things that are close but not close enough.
See above. My major issue with them is that they pull down too
fast (QBUS has a minimum of 5 ns).
Maybe the best bet would be to use a CPLD with
open collector outputs
and +5V tolerant inputs on the other side? Though you'd have to be
careful about current sinking requirements.
A CPLD can't sink nearly enough current. A 2N4401, on the other hand,
should do nicely (assuming you keep it from saturating).
Anyway, it's not as easy as I wish it were :)
Believe me, I'd love a
QBus card with a CompactFlash or SDCard that knew how to speak MSCP. But
if anyone wants to actually build it, it'll be a lot of work.
Well, yes. :-)
A bipolar transistor will turn off too slowly for this use, but a little
SOT-23 MOSFET will work just fine (and a series resistor in the gate can
tailor the on/off slew rates) You can get ones with less than 5 pF output
capacitance that will easily sink 100 mA with less than .7V VOL
but discretes would add to the assembly cost...