"Bob Cain" wrote in message


That's some excellent hand waving, Arny. But it doesn't
explain where all those other frequencies disappear to.
There's nothing in an FFT of one cycle of the low frequency
that would average them away. That's nonsense.

Since the modulating frequency is the low frequency note, a FFT that covers
an entire low frequency cycle would include instances of all possible
frequencies that the sidebands might have.

This begs the question as to what would happen if one measured the position
of the sidebands in two adjacent 0.01 second periods.

If the sidebands are in fact in motion @50 Hz, then their frequencies can be
expected to differ most of the time. Actual measurements of an actual
measured wave, processed to vastly reduce all AM distortion, does show this
effect.

If in contrast the positions of the sidebands in two adjacent 0.02 second
periods are measured, they should be the same, and this is what one
observes.

0.01 seconds corresponds to 410 samples at 44.1 KHz, so in this experiment,
the FFT should be based on 410 samples or less to avoid overlapping. 512
sample FFTs will overlap a bit, but are required to make the sideband
structure clear enough to comment on.