fdc@materna.uucp (Frank D. Cringle) (07/16/89)
We are looking for chips to drive an interface cable connecting two
sub-systems, with approximately equal numbers of signals running in each
direction.
-------- --------
| | ---- 20 signals --------> | |
| | | |
| | <--- 20 signals --------- | |
-------- --------
<--- about 3 meters ---->
The boxes are chunks of TTL logic. We need to get the signals across
cleanly and with a skew in the low-nanosecond range.
Using TTL single-ended drivers (e.g. 74AS645) and a ground wire paired
with each signal we still get significant crosstalk when a large number
of signals switch at the same time - a familiar problem in many
applications. We also overflow a 64-wire flat cable, which would be
convenient from the point of view of connectors.
The NatSemi Fastbus (or is it Futurebus?) trapezoidal drivers address
the problem, but for a somwhat different application. They are designed
for a bus, we just need a point-to-point connection. Their direction is
switchable, we need dedicated drivers and receivers.
Another answer is to use differential drivers, which can drive long
lines avoiding common-mode ground noise. However, these require twice
the number of signal pins on the chips, effectively doubling the number
of chips, which is too high a price to pay in board space.
Some years ago I worked on an ECL system in which bi-directional
differential drivers were used in a similar application (wide
point-to-point symetrical interface). The drivers were not commodity
chips in those days, but I had expected the idea to catch on and become
available with TTL i/o levels. Maybe it has and I am not looking in the
right catalogues (I've drifted into software in the meantime).
Here is a brief description of the scheme, which sounded like
voodoo-electronics to me the first few times I read the explanation. You
get to transmit two independant signals, in opposite directions, over
the same pair of wires at the same time. The pair connects two identical
chips (actually stages of chips) which each have a driver and a receiver
part. The driver is a current source whose direction depends on the
input signal at that end of the interface. If the logic inputs at both
ends of the interface happen to be the same the currents cancel, if they
are different a current flows (in varying directions depending on the
details, but the direction is not evaluated). The receiver is just a
current threshold detector whose output is exor'ed with the local input
signal, producing a copy of the input at the other end of the interface.
Any information on sources of such chips, or other ideas for solutions
to the problem, would be greatly appreciated.
--
Frank D. Cringle | Tel. +49 231 519 08 20
Dr. Materna GmbH | fdc@materna.uucp
Vosskuhle 38
D-4600 Dortmund 1