QBone with multiple driver options

UNIBUS and QBUS signals lines are terminated with resistor pairs of about 200/400 ohms, resulting in a quite low impedance. Normal TTL gates can neither drive this load, nor have the proper input thresholds.
These low-impedance busses allow routing over far distances ... but it was surely a design before "energy-awareness". Driving the bus alone requires several watts, burned-out bus drivers chips are a common reason for a dead PDP-11s.

DEC specs require special line-driver and -receiver chips, among them the DS8641, which is used on UniBone and QBone.

However, DS8641 chips (or its compatibles, as "DL8641" or the russian "КР559ИП3") are harder and harder to acquire. So time to look for alternatives.

qbone bus drivers logo

An often mentioned replacement for the DS8641 is the AM26S10: it is designed to drive/receive similar bus lines as UNIBUS/QBUS, and contains also 4 transmitter/receiver pairs. However the pinout is different.
Thats the reason for new QBone version 2023-04 with parallel sockets for DS8641 and also AM26S10:

qbone bus drivers row empty


A pin-compatible variant of AM26S10 is the 75138 driver. As the AM26S10, it's not fully electrically compatible to the DS8641. Their datasheets differ in the switching speed for example:  DS8641: 30ns. 75138: 20ns. AM26S10: 10ns.
The usual rule "faster is better" does not apply here: beside different receiver voltage thresholds, faster switching can produce more signal noise, so called "ringing".
In theory the 75138 is the better choice, as its specs are closer to DS8641. The AM26S10 however is still available today (2023, mouser.com) as DIP and SMD case.

qbone bus drivers close ds8641

qbone bus drivers close am26s10

qbone bus drivers close 75138

qbone bus drivers schematic

These chips can even be mixed freely on the board ... you won't do that.

Whats the problem with those almost-compatible drivers? It's all digital, right?

Clearly AM26S10 and 75138 have different specs than DS8641 ... a bit.

UNIBUS/QBUS Bus signals have switching times of a few 100ns, with sharp switching edges we reach frequencies near or above 100MHz.

So all this weird high-speed electronic black magic starts to apply, as explained in that classic

high speed black magic

For instance signal pulses are reflected at cable ends, if the resistors there do not meet the impedance of the drivers.
Reflection results in "Ringing", which is an awful effect: a clean pulse generated by a transmitter can be received as multiple pulses on the receiver side.
("receiver" can also be a scope or logic analyzer probe ... see UniProbe/QProbe discussion).
Imagine multiple pulses on the QBUS ADDR/DATA demux BSYNC line: Address lines on BDAL are latched with BSYNC, a bouncing BSYNC could latch invalid bus noise again, overwriting valid addresses latched before.

DEC addresses ringing in their UNIBUS Troubleshooting user's manual .

unibus troubleshooting waveform

Field engineers are advised to measure and suppress ringing by means of different-length backplane connector cables ... hidden in M9202 backplane jumper.

m9202 m9202 detail

Thats why there are several M9202s with different amount of cable inside.

unibus troubleshooting waveform m9202 fix

From these DEC pictures it's also clear why the "input threshold" is important: these signals are not "all digital". The exact properties of UNIBUS/QBUS drivers *matter*. So where can non-DEC drivers be used?


Different use of UNIBUS / QBUS

UNIBUS and QBUS have very similar electrical specs and use the same DS8641 drivers. However both busses are used in different environments.

For a UNIBUS PDP-11 its normal to span multiple backplanes, multiple boxes and often multiple racks ... see here.
Even hobbyists have such setups: For example for tape drives like TU56 DECtape or TU10, the tape controller build into the drive rack, so the white UNIBUS BC11 cable runs between several racks. All the ringing and impedance stuff matters over these long distances and multiple plugs. Thats why UNIBUS is designed asynchronically and with elaborate electrical and timing behaviour.

While QBUS specs allow the same cabling, a typical QBUS system is much more compact, occupying usually only one single backplane.
Personally I've never seen a PDP-11/03/23/73 or uVAX with even two backplanes ... not to mention spanning multiple racks.
So QBUS signal traces are short, impedance failures do not matter much. DEC delivers small QBUS systems not even with the bus-end-terminator build in.

Bottom line:

Alternate bus drivers surely can be used on small UNIBUS and certainly on QBUS systems ... as our test have shown.
To be on the save side, use only DS8641 chips for UNIBUS systems with BC11 cables.