PSD confidence intervals

[TrickyDick]

When a broad band random vibration signal is FFT'd and an average PSD is calculated using an exponential average with a known number of DOF's what is the PSD amplitude uncertainty interval?

eg what is the +/- dB variation within 95% confidence.

Any statisticians out there?

Many thanks ...

 

[GregLocock]

Here is Mr Beranek's take on the matter, for ideal gaussian signals and perfect filters:

SD=mean power/sqrt(B*T)

This is for proper averaging, not exponential. If you can work out an effective T for your averaging process then you ought to be in the right ballpark.

Blackman proposed a 96% confidence limit

12.6*(2/(3*n-1)+or- 1/(sqrt(n-1))) dB

where n=2*B*T

If nobody else comes up with a better answer then I suggest you contact Dr Colin Mercer at

http://www.prosig.com.

He knows everything. Cheers

Greg Locock