It would be useful to have a control on the
spectrum display for oversampling of the spectrum.
This would be particularly useful when the previous
feature request for zoom/pan on the frequency
axis is implemented.
Ideally you should have two controls which most
every other audio spectrum editor I have used
has (really, I am not making this up):
(1) a window size which is the number of
audio samples to analyze. This data is then
windowed with your anlaysis window (rectangular,
hanning, etc.). The size of the window is arbitrary
and could be a power of two (such as 1024 samples) or
not(such as 1563 samples).
(2) an *independent* transform size which is greater
or equal in size to the data window size. Typically
the transform size is equal to a power of two. Of
course, with FFTW you do are not limited to powers
of two if you don't want, but there isn't much
reason to do non-powers of two. When the transform
size is greater than the window size, then you can
guess what gets filled into the extra slots in the
transform.