I did some more testing. I probed the +5V output of the power regulator on my Arduino Mega (ATmega 2560, 16 MHz clock). Using the FFT function on the Rigol itself I don't see anything, but with your program I do.http://img404.imageshack.us/img404/521/arduino2560.png
I see a peak at 32 MHz, with harmonics at 64, 96, etc and some kind of 'echo' (not the right term) at 16 and 48 MHz (peaks have roughly the same level). This was while powered from USB, but when powered from a seperate DC supply I got pretty much the same thing.
It surprises me I get a peak at 32 MHz, not 16 MHz. As far as I know the ATmega doesn't internally double its clock or something like that. The ATmega may have been running some code that has a fast PWM configured (been experimenting with DDS), but as far as I know that PWM runs at a maximum of 16 MHz (= system clock), not 32.
Anyway, the result intrigues me; your software definately can extract more from the data than the FFT function on the Rigol itself.
Edit: hmm, if I would look at a 16 MHz square wave, I would only get odd-numbered harmonics, which might explain why I see a peak at 32, 64 and 96 but not at 48, 80, etc. However, there is something there at 48 MHz; after that things disappear a bit in the general noise.