29 Jun
2001
29 Jun
'01
1:01 a.m.
Why not see if you can find a copy of the old (and still occasionally available)
Motorola MECL System Design Handbook. It consists, mostly, of a discussion of
signal integrity issues from the standpoint you're concerned with in this case.
It does focus on MECL, but the principles are the same irrespective of the
technology. The author, Blood, is the authority behind most of the current
generation of signal integrity software in use today, theough he's seldom given
credit.
Dick
----- Original Message -----
From: "Chuck McManis" <cmcmanis(a)mcmanis.com>
To: <classiccmp(a)classiccmp.org>
Sent: Thursday, June 28, 2001 11:06 PM
Subject: Re: Unibus / Qbus bus drivers and receivers
At 09:30 PM 6/28/01 -0700, you wrote:
Well, the question is "how worried about
ringing *should* I be in a
single backplane configuration if I use bus drivers which have a
rise/fall time of 9 ns, when the bus spec says that the rise/fall
time of bus drivers must not be *less* than 10 ns?"
I'm guessing that it will probably be OK, but I really want this
thing to "work first time" and be reliable so I feel a little
uncomfortable cutting corners like this right at the outset.
I take it you are worried about violating the AC spec that reads in part:
Rise/Fall Times (from 10% to 90% and 90% to 10%)
must be no faster than 10ns.
This parameter relates to the 120 ohm bus termination and indicates that
the R/C termination constant will critically damp signals that switch at
10ns. If you drive it faster you will get a bit of bounce in your edges as
the termination network isn't quite fast enough to absorb the energy and
thus some of your signal gets reflected back. Anyway, you could do a couple
of things. You could load the capacitance on your bus drivers a bit and
slow them down, you could run them through some other gates to slow them
down, or you could just use them and look at it with a scope once you're
running. If you want to "check ahead of time" then kludge together a card
with some on it, switch them on and off and watch the ring.
Now how much should you be worried about? If the signal overshoots by more
than about 10% of Vcc I'd worry about it or under shoots by a similar
amount. Or if the ground bounce induced by switching everything is more
than about .2 volts.
Clearly you are a worthy designer unlike a PC guy who would said, "Hey if
it breaks they'll blame the BIOS or something so no worries!"
--Chuck
Then again, I've not done transmission line design in about 10yrs so if I'm
completely wrong you get your money back :-)