Use of Autospectrum vs FFT Magnitude

[271828]

I have what is probably a stupid question, but I'll ask anyway.

I do quite a bit of floor vibration acceleration measurement. Our Siglab software allows one to view the autospectrum or the FFT magnitude.

Folks often view and report the autospectrum. What's the advantage of doing this instead of reporting the FFT magnitude? My best guess is that the autospectrum is available because it was used in the FRF estimator, so the FFT was discarded to save resources, but I'm wondering if there's another reason.

Thank you very much.
2.71828

 

[dolsoi]

My best guess is that
FFT is a fast Fourier transform of a time sequence,
autospectrum is a spectral estimator (requiring averaging) using emsembles of FFTs. FFTs are useful for real time monitoring the trend, but an autospectrum is a stastistcal vaule for the supposed-to-be time variant signal events. Cheers, dolsoi

 

[hacksaw]

pre-FFT, electronic methods, based on tuned filters were used, and were called "autospectrum".

during the transition to FFT availability the common practice was to apply trhe FFT to the autocorrelation function of the data. in principle the result should be the same as the FFT of the raw data, as long as dc levels and linear trends are removed (or filtered out). at the same time, the raw spectra id pretty noisey, so averaging several spectra (i.e. previous post) or filtering the raw data, or using various band-pass filters methods help.

 

[271828]

Thanks guys.

Correct me if I'm wrong, but it's possible to compute the autospectrum of a single record, right? I hope so, because we do this often!

 

[dolsoi]

yes.