Joonas-
> I try to get a frequency response using
FFT(output)/FFT(input). My input
> signal is white noise. This works pretty good in low frequency. However
> in high frequency (next to the sampling frequency) the gain curve goes
> fast to infinite but it shouldn't. Moreover the phase curve is not
exactly
> correct. It has the right form but it goes about 200 degrees too high. How
> should improve the results?
Near the sampling rate you have practical limitations that could be making
the
results look this way:
-freq response of anti-alias filter used in data acquisition card
-freq response in any subsequent processing, for example sampling
rate conversion
For instance, the closer to "ideal"anti-alias cutoff near the sampling rate
the
higher the cost of the data acq card.
-Jeff