Joe,
     That's a pretty wide open question!  Some things to look at are...

Try to identify where it's coming from.  If your tests are showing
spikes in specific bandwidths, try to determine what is operating at
(or it is a harmonic of) and concentrate your changes in that area of
the board.

Look for looping signal structures.  The traces on the board act as
radiating antennas.  If you have a signal which loops around the board
you have a loop antenna - highly directional in the plane of the board.
Remember that you are looking at current paths and that the current path
starts at the signal source goes to the receiving device AND then back
to the ground source.  Many times a structure won't look like an antenna
until you consider the current path through the ground plane.

Keep high freq signals short and away from the edges.  Less signal length
means less metal and less antenna.  At the edge you lose the advantage
the ground plane gives you and increases return current there.

In the ground plane, look for bottlenecks and slots where return and shadow
currents will intensify.  A nice straight up and back current path may be
forced to detour if the ground plane is too perforated.  Especially watch
for high freq lines crossing over a slot in the ground plane - the return
and shadow currents can excite the edges of the slot just like a slot
antenna.

High freq lines (and/or lines with short rise times) may need to be terminated
with a impedance matched load.  If the prop delay becomes large relative to
the wavefront rise time you can get pulse ringing.  This oscillation can
double crosstalk and emission levels.

This is by no means all - some may apply to your situation.  Let me know
if I can help or if you just cannot make sense out of my ramblings :^).

DonW...
[log in to unmask]
Don Walker
Sr. CAD Designer
Austron Inc.

****************************************************************************
* The mail list is provided as a service by IPC using SmartList v3.05      *
**************************************************************************** 
* To unsubscribe from this list at any time, send a message to:            *
* [log in to unmask] with <subject: unsubscribe> and no text. *
****************************************************************************