Random Analysis - Cross Spectral Density

[crostolo]

Hi all,

I'm looking for some information about Cross Spectral Density function in order to perform FEM analysis (NASTRAN) with two exictations input on the models. In many case I have to perform random vibration analysis on models (electronic boards for example) linked to the support structure with more than one point, each of them has its own PSD spectrum input. As a conservative approach I use the worst input, but it is difficult then to have a good correlation with tests results.

Is there a way to "create" a cross spectral density of two PSD? In matlab are available many function but only with two signals in the field of time, and not frequency.
I don't think that artificially tranform the frequency domain (PSD) into one of the infinite possibile time domain and then perform the cross spectrum correlation is the right way. O am I wrong?
Any tips?

Thanks in advice
crostolo

 

[SomptingGuy]

No. The PSD has already lost all the phase information of both the input and the output signals, which you cannot recreate.

Steve

 

[crostolo]

Ok, understood.
Thanks steve