>I received e-mail once from an engineer at HP who designed a 40MHz
microprocessor board for one of HP's >printers; the thing was a double-sided
(not 4-layer) board, memory-bus speed was 20MHz, and the thing passed >FCC
without ANY shielding!  Took him quite awhile to get the layout perfect, but
when you're making oodles of >something even $0.50 worth of shielding can
get expensive.


Exactly. I use only two layer myself, in plastic enclosures, with switching
power supplies etc. The latest one had a 300uS 19A @32V pulse requirement.


>The only other point that I'd mention would be that if you're driving shift
registers, etc., especially if the signals are >traveling any significant
distance, you may want to avoid clocking them "as fast as possible" since
even a 4MHz >PIC may be able to clock out data at up to 1MHz (if ports are
laid out nicely), and 1MHz is fast enough that >harmonics can pose real EMI
problems.  But if you can afford to add a few NOP's in the shift loop, you
should be >able to reduce EMI easily.


Always a good idea, not to go faster than you really need to.  Clock lines
should have a resistor at the source end, even if you end up placing a zero
ohm there, if you need to slow the edges, you'll be glad for the pad.  I
hear lots of comments about placing CAPS to GROUND on high speed lines,
which is exactly the WRONG thing to do.
You just make the chip draw MORE high frequency current!