I am playing around with mathdotnet and am experimenting with the different windows for use in FFTs. I am specifically curious about the Hann window (though i'm sure this applies to the others as well). When I generate a Hann window of, say, 2048 samples wide, I expect to get a set of variables gradually increasing to a peak and then gradually decreasing at the same rate. What i get instead is this:
[0] 0 double
[1] 2.3553948388377322E-06 double
[2] 9.421557163713512E-06 double
[3] 2.11984204002702E-05 double
[4] 3.768587359187503E-05 double
[5] 5.8883761400674306E-05 double
[6] 8.4791884109036708E-05 double
[7] 0.0001154099976216072 double
[8] 0.00015073781346730541 double
[9] 0.0001907749988023788 double
[10]0.00023552117641323367 double
....
[2039] 0.00015073781346730541 double
[2040] 0.00011540999762155169 double
[2041] 8.4791884109036708E-05 double
[2042] 5.8883761400674306E-05 double
[2043] 3.768587359187503E-05 double
[2044] 2.11984204002702E-05 double
[2045] 9.421557163713512E-06 double
[2046] 2.3553948388377322E-06 double
[2047] 0 double
As you can see, the values at the beginning and end of the hann window seem to be completely meaningless especially considering that anything over 1 doesn't make sense...
Generating a window with much less samples (I tried 20) works just fine.
As I don't feel I know enough about FFTs or the library itself, I don't feel comfortable posting a bug report before getting your opinion.
The numbers look correct to me. Note that 2.3553948388377322 E-06 is just the scientific notion of writing 0.0000023553948388377322.
Note that when using this for FFTs, you may want to use Window.HannPeriodic instead of Window.Hann, which is available since v3.14.0-beta01.