Spectral Estimation by Means of Overlapped Fast Fourier Transform Processing of Windowed Data
NAVAL UNDERWATER SYSTEMS CENTER NEWPORTRI
Pagination or Media Count:
An investigation of power-density autospectrum estimation by means of overlapped Fast Fourier Transform FFT processing of windowed data is conducted for four candidate spectral windows with good side-lobe behavior. A comparison of the four spectral windows is made on the basis of equal half-power resolution bandwidths. The criteria for comparison are statistical stability of the spectral estimates, leakage side lobes of the spectral windows, number of FFTs number of overlapped pieces required, and size of each FFT required. Some striking invariances are discovered in the ultimate variance-reduction capabilities and in the number of FFTs required to realize 99 percent or less of the maximum EDF. The only trade-off among the four windows is that those with better side lobes require larger-size FFTs.