pmpse - estimate power spectral density and autocorrelation of a signal 

estimate power spectral density and autocorrelation of a signal 

The PMPSE program from the IEEE Programs for Digital Signal Processing
is no longer being maintained.  The same result can now be obtained with
the spect program by using the -a flag to average the spectrum over a
number of finite overlapping sections of an input signal.  This is a
good idea when the signal is noisy because noise will tend to be 
uncorrelated in successive sections.  Hence, the averaging process will
smooth the noise spectrum while enhancing whatever deterministic 
components may be present.  The tradeoff is that each section should be
as long as possible to maximize spectral resolution; but as many segments
as possible should be averaged to reduce variance.  Since the signal
(or at least the portion of interest) is presumably of finite length
(and since the overlapping should always be by 50%), this produces a
fundamental conflict.

