27 Feb
2013
27 Feb
'13
11:57 a.m.
On Feb 27, 2013, at 2:24 PM, Guy Sotomayor <ggs at shiresoft.com> wrote:
On Feb 27, 2013, at 12:41 AM, Philipp Hachtmann
wrote:
I'm not sure any but the last National Semiconductor transceivers
were actual trapezoidal drivers. That requires an integrator on
the output driver. My impression from the datasheets of the
earlier DEC chips is that they're binned for a particular max
slew rate, which is really all you need. By the time the true
trapezoidal transceivers came out (like the DS3662), the silicon
was both cheaper and faster, necessitating the rate limiting.
The current sink is also designated to be a maximum of 70mA by
DEC standards, but this is presumably a momentary sink for
driving the bus load capacitance low; the maximum sustained
current would be about 28mA (5v / 180 ohms) plus the relatively
small amount contributed by driver/receiver leakages (another
few mA if all the other bus participants are following the same
rules). There are also rules about the minimum voltage when
driving zero at 70mA, which many modern FETs meet quite handily.
There are a lot of discrete or paired FETs that should do the job
of the driver quite adequately; the receiver is usually best done
with a high-speed comparator, which unfortunately tend not to be
particularly cheap. The best one I've found for modern devices
is the MAX9108, which fits the required 35ns propagation delay;
it has TTL outputs (not great for 3.3v logic, but you can use
glue) and costs about $1 per gate ($.50 per in >= 100 qty).
That's what I have planned for my QBUS board, whenever it may
happen. If anyone happens to find a suitable comparator that
is cheaper, I'm all ears (don't forget about input leakage and
capacitance constraints!).
Ah, really? Of course NOT! The original driver
was DEC8881 which is a relabelled 7439. The 7438 I used has very similar specifications.
As the DEC8881 and the 7438 are not available, I have chosen the 7438.
And it does not make a big difference if the machine is more or less loaded. The load on
the drivers is always the same. It's an open collector bus.
The 8881 is *not* a relabeled 7439. Yes, it's open collector but the current sinking
capabilities are pretty high (80ma?). It's also what's called a
"trapezoidal" driver which has a controlled slew rate (rise/fall times) to
minimize reflections on the bus. The drivers are optimized to drive into a bus terminated
by a 180ohm resistor to +5 and a 390ohm resistor to ground. The receivers are also specialized and are not quite
at the "standard" TTL levels for greater noise immunity and minimal input
leakage.
The Unibus interface chips (that were also used on Omnibus) are:
DS8640 quad bus receivers
DS8641 quad bus transceivers
DS8837 hex bus receivers
DS8881 quad bus drivers
There were older versions but these are pretty much what everyone used for Unibus and
Omnibus systems.
There are newer ones as well, with true trapezoidal drivers, such
as the DS3662 (I think John Wilson said he had a large quantity
of some member of that family that he was happy to sell). None,
of course, are in production anymore.
- Dave