My initial reason for getting into this debacle was the search for a universal window size for a Fourier transform of an acoustic time series, in order to mitigate the loss of detail in the time domain.  To put it another way, the aim was to increase the time domain resolution in the frequency domain.