ABSTRACT
In Section 6.4, we describe FFT algorithms that oer computational eciency in calculating DFTs, including auto-and cross power spectrograms. Because all DFTs and FFTs are computed from nite-length data, the innite-length, sampled, continuous signals can be considered to be multiplied with a nite-width, 0,1, rectangular, data win dow. e presence of this data window alters the shape of the resultant spectrogram from what it would be if an ideal, innite-width data record were used. In Section 6.3, it will be shown that there is a trade-o between the width of spectral peaks resulting from two or more coherent (e.g., sinusoidal) components in a waveform and the amount of ripple between peaks. Many data windows (other than rectangular) have been devised that attempt to reduce ripple and still obtain high spectral resolution in the spectrogram.
