Search All of the Math Forum:

Views expressed in these public forums are not endorsed by NCTM or The Math Forum.

Notice: We are no longer accepting new posts, but the forums will continue to be readable.

Topic: Q: psd.m (Welch's method)
Replies: 1   Last Post: Nov 18, 1996 3:57 PM

 Messages: [ Previous | Next ]
 Michael M. Bech Posts: 7 Registered: 12/7/04
Q: psd.m (Welch's method)
Posted: Nov 18, 1996 3:57 PM

Dear Matlab Friends !

I use the psd.m function from the Signal Toolbox to estimate the power spectral density. psd.m
uses Welch's averaged periodogram method.

The algorithm implemented in psd.m returns a vector of frequencies at the same size as the size
of the estimated PSD. For real data we have

length(psd-estimated)
= length(frequency vector)
= nfft/2 + 1 (n even)

Therefore, the function psd.m uses zero-padding of each (possibly overlapping) section to length
NFFT. Hence, the frequency resolution is equal to Fsample/nfft.

Now the questions:

1) Is the variance of the psd estimates decreased if we do not zero-pad each section before
doing the FFT ? (i.e., length(section) = length(window) )

2) Has somebody (including The MathWorks, Inc. !) rewrote the psd.m routine to include the
option described in 1) and to return the corresponding frequency vector ?

Best regards,

Michael M. Bech
e-mail: mmb@iet.auc.dk

Date Subject Author
11/18/96 Michael M. Bech
11/18/96 Thomas P. Krauss