On Fri, 28 Jul 2006 09:08:20 -0700, Al Williams wrote:
>This cured the glitch problem and I have never understood why?
This is pretty far off the topic of this list, so I'll make my response
First, you are confusing the word "ground" with "signal common." In his
excellent workshops on EMC, Henry Ott talks about "the invisible schematic"
hidden behind the ground symbol. The word "ground" is commonly used as a
substitute for "signal common," but it is a really bad choice, because it has
caused us to falsely believe that the earth has something to do with that
circuit. Ott's excellent book on EMC is considered "the bible," and is well
worth tracking down and studying. Google to find it. Henry is WA2IRQ.
Second, those signals running around your circuit board are broadband (that
is, high frequency) signals (thanks to their fast risetime), and those
digital circuits are subject to HF interference and instability. If you used
a single point return for the signal that doesn't run physically next to the
signal trace, you are adding a lot of L in series, and you are forming large
current loops that can pick up noise.
In essence, you need to treat those digital circuit traces as transmission
lines on your circuit board, both from a circuit operation (waveform, travel
time) point of view and from an EMC (noise immunity, noise generation) point
of view. That's why modern circuit boards are laid out with multiple layers,
one of them being a ground plane. The ground plane functions as a return for
the transmission line -- that is, the "hot" side of the signal path is the
trace, the "return" side of the signal path is a narrow part of the ground
plane lying directly under the signal trace.
Jim Brown K9YC
TowerTalk mailing list