Compensation filter

Started by January 26, 2011
Dear all,
I'm Andrea and this is my first post on DSPRelated.

I have a trouble that I can't solve.

I have a generic system whose output, in the frequency domain, has a module proportional to the fourth power of angular frequency [ |.| proportional to w.^4 ].
I have to build a FIR filter that compensates this phenomena. So the filter is a sort of lowpass filter decaying with an inverse fourth power law.

I build up such a filter with the frequency sampling method (manual, not fir2), but filtering the signal (sampled at 44100 Hz) I encounter some problems:

- filter's numer of taps must be a multiple of 1024 to have linear phase (!!!)
- filtering the signal (with filter command) I obtain some unwanted zeros in the frequency domain corresponding to the secondary lobes. But I can see these secondary lobes only with a very high resolution FFT or FREQZ
- fir2 doesn't work (the ideal freq. response is very different from the obtained one)

What I can do to solve?
Thank you very much,