Low-pass filter in MATLAB returning NaN values

Question

I'm doing some audio processing. I've been able to load some audiofiles into MATLAB, play them, mix them and synthesize signals.

Now, however, I want to extract the base similar to an equalizer or a low-pass filter in an audio editing program. I know MATLAB has a lot of filters and built-in filter design tools; I just want to add a simple low-pass filter, though. I found this guide: http://www.aquaphoenix.com/lecture/matlab10/page4.html

f = 44100;
n = [1:f];
note_a = 440;
note_csharp = 554.365;
note_e = 659.255;
chord_a = sin(2*pi*(note_a/f)*n)+sin(2*pi*(note_csharp/f)*n)+sin(2*pi*(note_e/f)*n);
fNorm = 200 / (f/2);
[b,a] = butter(10, fNorm, 'low');
chords_low = filtfilt(b, a, chord_a);

I get no error, but all the values in the output of the filter is NaN. Why is that? The song is in stereo, could that be it?

What are the values of f & fNorm prior to the call to butter()? What is the result of all(isfinite(funky))? — Max
f has value 44100 fNorm has value 0.0091 all(isfinite(funky)) has value [1, 1] — user1661303
I received a downvote. Anybody care to explain what I did wrong so that I can avoid doing it next time? I thought the question was clearly stated but maybe I was wrong. It's also not overly simplistic. I tried googling the answer for half an hour before posting. Please feel free to point out what I did wrong. — user1661303
I think you got a very good answer on meta. If you remove the last part (it's a different question) I'll vote to reopen this. — Stewie Griffin

learnvst learnvst · Accepted Answer · 2014-03-06T16:57:14

The order of your filter is too high for the design type with a cutoff frequency that low, so you get numerical problems. To make a quick fix, reduce the order of your Butterworth filter to something a little lower than 10. An 8th order filer works for your design, or raising the cutoff frequency also results in real valued output.

If you want a filter with the frequency response specified by your design, you need to cascade lower order sections. See a detailed old answer I made for information on how to go about this. The tf2sos function is your friend here.

Low-pass filter in MATLAB returning NaN values

1 Answers