how big an FFT can I do with function HarmonicAnalyzer

[rorOttawa]

I am seeing weird results when using an FFT that is 64k or 128k.

Is this function good for this big an FFT

[gwd]

The documentation on FFT.vi itself says it can take data size up to 2^23 which is about 8E6. I've used it up to at least 10000 without any funny business.

However the harmonic analyzer vi has a lot more than an FFT going on.

So what exactly do you mean by "weird"?

[rorOttawa]

I am using Labwindows/CVI 8.01

I am using this sequence:

framesize=8192(works OK) or 131072

  status = ScaledWindowEx(distored_signal, framesize, window_type,0.0, &winConst);
  status = AutoPowerSpectrum (distored_signal, framesize, 1/rate, spectrum,&spec_freq);
  status += SpectrumUnitConversion (spectrum, framesize/2, 0, 0, 0, spec_freq, winConst, conSpectrum, units);
  HarmonicAnalyzer (spectrum, framesize/2, 0, number_of_harmonics, window_type,
     rate, fundamental_freq, harmonic_amplitudes, harmonic_frequencies, &thd,
     &thdNoise);

"weird" = when I use  131072 the freq per bin in conSpectrum=spec_freq but if I multiply bin_number*spec_freq it does not equal the expected frequency in the test signal.  The bin count is low by about 25 bins.

Actuall I see the weird behaviour is in function SpectrumUnitConversion  not  HarmonicAnalyzer