From help-octave-request at bevo dot che dot wisc dot edu Fri Dec  7 10:59:56 2001
Subject: Re: need help filtering a signal
From: John Day <jday at csihq dot com>
To: help-octave at bevo dot che dot wisc dot edu
Date: Fri, 07 Dec 2001 11:57:35 -0500

Philipp,

What you want to do is apply a "notch" filter to your signal, which is the 
converse of a "band pass" filter.

Filtering can be applied in either the time or freq domain. They are 
equivalent in terms of linear transformations.

Your attempt to filter by cutting frequencies out of an FFT is in the 
frequency domain. At first glance this seems ideal, because you seem to be 
able to zero out any arbitrary band of frequencies. But this is an 
illusion. You are only constraining the sampled points in the FFT. The 
points in between are not constrained to zero, so this won't work 
"perfectly", as you might have imagined. You can make it "better" by 
increasing the FFT size, but perfection would demand an infinite number of 
points.

In the time domain, you achieve the same results by convolving a filter 
against the sampled points. Again, a perfect filter does not exist, but is 
the limit as the filter length goes to infinity. A filter to "notch" out a 
single frequency would have to be rather long and it will spill over 
somewhat and attenuate neighboring frequencies. You will have to determine 
how "steep" the walls of your filter will have to be.

Practical time-domain filters can be built with the Remez algorithm. A 
message was posted in Oct detailing an Octave version of this algorithm, 
under the subject:"An octave port of Remez algorithm to design linear phase 
FIR filters".

HTH,
John Day
Computer Science Innovations

At 05:34 PM 12/7/01 +0100, Philipp Schwaha wrote:
>hi!
>
>what i need to do is fiter a signal to remove noise. what i wanted to try
>first was a fft then cut all frequencies except the the important ones (e.g
>100Hz, 200Hz ...), but when i did the fft i did not know how to do this,
>since there are no frequencies associated with the values the function fft
>returns.
>this is the first time i try to do something using an fft.
>then i discovered the function fftfilter, but i could not make it produce the
>results i wanted.
>
>how can i remove all frequencies from a certain signal, except a few selected
>ones?
>
>thanks
>philipp
>
>
>
>-------------------------------------------------------------
>Octave is freely available under the terms of the GNU GPL.
>
>Octave's home on the web:  http://www.octave.org
>How to fund new projects:  http://www.octave.org/funding.html
>Subscription information:  http://www.octave.org/archive.html
>-------------------------------------------------------------



-------------------------------------------------------------
Octave is freely available under the terms of the GNU GPL.

Octave's home on the web:  http://www.octave.org
How to fund new projects:  http://www.octave.org/funding.html
Subscription information:  http://www.octave.org/archive.html
-------------------------------------------------------------