From: jimf@DELETEME.pobox.com (Jim (from Oz))
Subject: Re: Ground loops?
Date: Sun, 15 Sep 2002 04:52:21 GMT
Organization: To a limited extent
X-Newsreader: Forte Free Agent 1.1/32.230
On Sat, 14 Sep 2002 20:39:35 -0700, John Larkin wrote:
>Grounding is more religion than science. So...
>Use only one ground plane on each board. Connect it to other boards as
>many places and ways as possible, and as directly as possible. If you
>have a metal chassis, bolt all the boards - and their ground planes -
>solidly to the chassis, using lots of fat spacers and screws. The idea
>is to make all the grounds on all the boards as equipotential as
>possible. Now, if you have really low-level analog signals, or
>picosecond logic signals, handle them differentially or some other way
>so they are not affected by millivolt differences in grounds.
>What this will do is make sure you have minimal EMI problems and
>minimal digital logic noise.
>You haven't said exactly what you're trying to do, so this suggestion
>might be tweaked in unusual circumstances. But the conventional wisdom
>of star grounding, and isolating analog and digital grounds, is just a
>recipe for trouble in most cases. All it does is ensure that various
>parts of the system have grounds that are all over the place
>electrically at high frequencies, and flailing around relative to the
>enclosure, and who wants that?
I use to use star grounding some years ago, trying to figure out
where the "bad" currents would flow and ensure they were kept
away from the "good" :) currents. Then I designed two very
similar systems at about the same time. On one I did a star ground
and on the other I did as you suggest above. The star ground
had much more noise than the non star unit. Now my approach
is simple; ground wise - connect everything to everything else :)
Jim (from Oz)
Speed doesn't kill - bad driving kills