ONE THIRD OCTAVE STUFF

[CESSNA1]

HELLO: I am analyzing some accelerometer vibration data and have to put it in a one-third octave format. My question has to do with noise reduction. In a normal analysis the noise can be reduced in the frequency domain by successively averaging the data blocks of the PSD. My software(DaDisp) does the accelerometer to octave conversion, but where do I do the noise reduction. Can I do it in the one third octave domain or, take the PSD, do the reduction, and then invert the resulting PSD and then do the octave analysis?

Your input is appreciated
Regards
Dave

 

[GregLocock]

You can convert to third octaves at any stage in the proceedings, ie before or after summing the magnitudes^2.

I always found it helpful to think about where the energy from each frequency was going.

There is a gotcha, are you doing a triggered, gated or phase locked analysis? If not you should be OK.

If you are attempting to get transfer functions I'd do the conversion to third octaves last, given the choice, as phase and third octaves are a funny mix.





Cheers

Greg Locock