if you have a brazillion samples/sec they are close enough together so the peak of the samples would be near the peak of the analog signal.
But say you are sampling at 44.1k
for a 44 cps wave you are getting 1000 samples
but on the higher end at say a 20kcps sine wave used for testing
you are only getting 2.2 samples – barely enough to meet nyquist
depending on how the samples line up with the signal
you could get sample values from near zero to near peak
so
wouldnt that mean that the dBFS would not be meaningful
as it depends on the sample position wrt the original signal
i must be missing something obvious here
did nyquist have some caveat about an infinite sequence
and we approximate it (and hope it works) by a smaller set of samples ??