fourier analysis and slow spectra

[Gerald Pollack]

In article <49cjnl$hnp at sparcserver.lrz-muenchen.de> LEOPOLD at mip.paed.uni-muenchen.de writes:
>From: LEOPOLD at mip.paed.uni-muenchen.de
>Subject: fourier analysis and slow spectra
>Date: 27 Nov 1995 14:59:01 GMT

>In our lab we use the Neuroscan software equipement. For our spontaneous EEG we
>compute the fast fourier transformations.
>The result is a huge amount of (artifical!) theta in our relative power
>spectrum. Who can give an explanation and how to solve the problem.

The problem might be what's known as aliasing. The FFT procedure assigns all 
of the energy in a signal to frequencies between 0 Hz and half the sampling 
frequency (i.e. the digitizing rate of your A-to-D board). If energy is 
present at frequencies greater than half the sampling frequency, then it shows 
up in spectra at lower frequencies. The solution is to use a hardware low-pass 
filter to ensure that there is no signal above half the A-to-D sampling 
frequency.