> Thanks for the feedback. I understand the general issues with RF noise. I
> think the thing that surprised me was that I had assumed that it would be
> the rate of execution (i.e. processor speed) that defined the type and
> frequency rather than the actual mix of instructions being executed. I
> suppose I shouldn't be that surprised since presumably different
> instructions cause different groups of gates to operate etc. and so cause
> different amounts of current to flow.

Generally, the most important thing is the clock rate; on the PIC, most of
the noise will be at harmonics of the instruction clock (xtal/4).  For very
tight loops there will be additional noise at harmonics of the loop rate,
but for loops longer than a few cycles the frequency will be low enough that
you don't need to worry about /that/ aspect of RF interference.  On the
other hand, there were some games for the Timex-Sinclair 1000 which used an
AM radio for their sound output; bursts of instructions at regular intervals
could cause identifiable sounds on the radio (supposedly).

In your particular situation, what's probably happening is that the RF noise
at the particular frequency of interest is being affected greatly by the
instruction mix, even though the total amount of RF noise is not.  Unless
you have equipment to measure the spectrum of RF noise emitted, anything you
try to do with the instruction mix will be primarily hocus-pocus and will
have little practical effect on RF emissions.