Definition and Computation of Power Spectral density with FFT

Dear all,

when computing the power spectral density of a random signal using fft (being the power spectral density the Fourtier tramsfrom of the autocorrelation fucntion) is it necessary to work on the double-sided correlation fucntion? I mean with positive and negative lags? or can I just build it one-sided? Will this affect the shape of the fft fucntion?

I cannot use pwelch or other matlab fucntion for this application, so please just answer my theoretical question...

thank a lot

Emanuele

Answer 1

[Rick Rosson]

Why don't you use MATLAB to experiment with several different approaches on one or more very simple test signals, like pure tone sine waves, exponentials, Gaussian pulses, etc.? Through simple trial-and-error, you should be able to find a method that works well and matches with the theory.

Answer 2

[Emanuele Spinosa]

Dear Rick,

thank you. I'll try to to that, in my case I need to generate a random signal though. I will let you know if I can make it, it shoudln't be diffucult.

Thank you

Emanuele