[chuck-users] spectrum magnitudes

[eduard aylon]

I've been examining a bit more the output of the fft and I am not  
sure about the spectrum magnitudes obtained after analysis. For  
instance, analysing the peaks of a SinOsc with unity gain in a  
similar manner as in examples/analysis/tracking/pitch-track.ck will  
print out values of max (maximum peak) around 0.26  instead of values  
near 1 (depending on the window used and the frequency. I've used a  
hamming window and a freq proportional to sample_rate/fft_size). Why  
is this? I've forgotten some of the implications of performing an FFT  
but I believe the values of the magnitudes obtained after an FFT are  
half the original value, but then, in the example above, one should  
obtain peak values near 0.5. What is the cause for this? Could it be  
that windowing is not normalised? I'd like to be sure of the values I  
should expect in order to set thresholds above which spectral peaks  
will be taken into account.

thanks in advance,

eduard