S-100 bus termination

From: Allison J Parent <allisonp_at_world.std.com>
Date: Sat Jul 25 09:45:46 1998

< The first S-100 computers did not come with any termination but they
< soon found that the long bus lines were causing ringing and false
< triggering in the circuits. Ringing is caused when a signal reaches the
< of an unterminated (open) line such as a data buss.* The signal reverse
< polarity and travels back in the direction that it came from just like a
< echo. When it reaches another end, it changes polarity and direction aga
< Now you have two positive pulses and one negative pulse where there is o
< supposed to be one pulse! Consider the number of times that each line
< branched off to a card socket and you can understand that now there wher
< dozen or more echos on every line and they were all at different times.
< It's no wonder that the systems where so difficult to make work with al
< the noise on the buss.

This is true. All signals on the bus and off it obey Transmission line
theory. IT's also the one of the reasons why the T-connector for thin
eithernet must be on the back of the machine and not a 3ft jumper away.

One of the design rules that ALtair bused the worst and IMSAI seemed to
have a clue on was what are clalled stubs. What this translates to
is anything going onto and off of the backplane(bus) MUST be buffered
as close to the edge connector as possible.

< board. I believe that the later S-100 systems like the N* Horizon came w
< termination built into the MB. The ringing problem in the early Altair

No it didn't and doesn't need it. They approached it a bit different
by making one end of the bus a permanently resident IO section with
seria, parallel and heartbeat interrupt. Their choice of drivers and
recievers helped some too. Obeying stub theory was another step taken.
It is good enough that even at 8mhz z80 it works.

< was made worse by use of the long wires that ran from the MB to the fron
< panel. Better designs eliminated a lot the sources of the ringing. FW

that was the horror I saw when I built my altair back when. As an
RF/analog engineer I could see that making my day pure hell. It
proved to be true. The solution was a small board that plugged in and
buffered all the signals to and from the front pannel.

< this is EXACTLY the reason that you still have to terminate SCSI busses
< disk drive cabling. There is still an ongoing debate in the SCSI communi
< about the nessecity/benefits of active vs passive termination.

SCSI is an open collector bus so you need both termination (any) and
it must pull up unused lines or it don't work. This is also true for DEC
Omnibus, Unibus and Qbusses.

One of the long standing problems with bus terminations is that they
absorb power. This required stiffer drivers at the source and since
stiffer drivers matched the bus impedence even worse it tended to make
problem grow rather than solve it. it also consumes power contributing
heat to the box. S100 bus was one of those we made it that way but, no
one ever would have engineered it that way. All of the engineered buses
are terminated in some way or designed to perform without the need for it.

Allison
Received on Sat Jul 25 1998 - 09:45:46 BST