Charlie,
You write:
"So according to the standard, the required minimum number
of hosts in one. I believe that one is sufficient,
especially where you have no idea what computer the end user
will have."
Remember, though, if you select a host that is quieter than
most, you may well fail to qualify in noisier, real-world
ones, and no amount of protesting "But it passed!" will sell
what you are building. Better to use a less than perfect
system for the test. And... be sure to wring out your
peripheral as hard as it will go. If you are too protective
during YOUR testing, it will certainly cause problems later
on during someone else's evaluation. Take it from me -- been
there, done that!
Things to look out for:
Noise on the bus: If your test system is quiet, you may be
unpleasantly surprised later on when you realize that your
card doesn't isolate the bus from your outputs. This can
also be aggravated by riser card design -- one system may
pass, another fail, and the only difference be a riser card.
Direct radiated signal pickup: Computers are NOISY inside,
and you may find the card you design non-compliant in some
customer systems simply because it makes a good antenna.
Expansion slot cover grounding. The IBM design is a whole
debate in itself, but it's enough to say you will need to be
sure you have a good enough ground -- and then don't rely on
finding a system that has one. (Yeah, I know it's
contradictory -- grin.)
This kind of exercise is what makes EMI _fun_!
Cortland
