I haven't seen the correct answer to this, so...
You need to limit the edge rate to prevent ringing on the bus.
Note, ringing will show up in different forms on EVERY board
in the system, not just the one you are designing. Draw two square
waves on a the top edge of two pieces of paper (one each). Now,
slide the two relative to each other and think about what the
sum of them would look like. Somplaces you get double amplitude,
other places you get negative amplitude. This is ringing. It is
bad.
The best solution is to use an edge rate limited chip. DEC doesn't
make them anymore, but you might be able to remove them from an
old DHV11 that nobody wants (they show up on ebay occasionally
for next to nothing and don't sell).
Otherwise, a small resistor in series with the output of the
gate will slow the transition down.
Here is the right way to do it:
1) Find or calculate the source resistance of the driver.
tau = RC
2.3 * tau is approximately 90% charge (or 90% voltage)
or 2.3 * tau ~= 9nS
so, Rpin ~= (9ns / Cpin) (C in Farads, R in Ohms)
[ Is that right? as C goes up, R goes down... Okay ]
2) Look at the capacitance spec for the bus (I'll call
them Cmin and Cmax), and calculate the new resistor (Rnew)
10ns = 2.3 * (Rpin + Rnew) * (Cmin + Cpin)
Tmax = 2.3 * (Rpin + Rnew) * (Cmax + Cmin)
[ bigger R or bigger C make rise time worse ]
Regards,
Clint
On Thu, 28 Jun 2001, Michael Davidson wrote:
I am just about to embark upon designing a small qbus
interface
card, and am trying to work out what to use for bus drivers and
receivers.
The 8641 quad bus transceiver is still available, but it appears
that the 8640, 8881, 8131 and 8837 are all obsolete and out of
production.
I was tempted to use a pair of DS3862 octal bus transceivers since
I want an internal tri-state bus on the card, but it looks as if
the rise and fall times on the drivers are just a little fast
(nominal 9ns compared to a minimum of 10ns according to the qbus spec).
It would be really convenient to use these parts since it would
help to keep the chip count down - any opinions on whether the faster
rise/fall time would really be a problem in a small single backplane
system?
Also any suggestions for substitutes for bus drivers and receivers?
I seem to remember the 7438 being capable of sinking enough current
to drive a 120 ohm bus line effectively, but I'm not sure.
(or, even better - anyone got a DCK11 chipset they would be willing
to part with?)