Why does my FFT return all zeros for certain sound files?

Hi!

I've been working on a Equaliser and I want it to display a frequency-domain graph from an audio file that is read. I had it working, at least so I thought, so I was experimenting with all the sound files I could find. (The application allows .wav, .flac, .m4a and .mp3 files).

For almost all of the files I have tried the code (shown below) works, as seen in this picture.

Then this is the ouput I get with one of two .mp3 files. The matlab code plays the file, and it clearly reads it correctly due to the top display showing the waveform.

Below is the code that I am using:

                file = app.SelectFileDropDown.Value;
                [y,Fs] = audioread(file);
                n = length(y);
                t = 0:seconds(1/Fs):seconds(n/Fs);
                t = t(1:end-1);
                plot(app.VisualiserDisplay,t, y); % This displays the waveform at the top
                
                N = 2^12;
                j = 0:N-1;
                f = (Fs/N).*j;
                X = fft(y,N); % This is _supposed_ to display the graph at the bottom
                semilogx(app.EqualiserDisplay, f(1:round(20000/(Fs/N))),20*log(abs(X(1:round(20000/(Fs/N))))));

I was playing around to try and determine why it wasn't working and discovered that for the files that don't work the 'double' X is filled with all zeros but y is not! For the working files, X is filled with complex numbers, as it should be. Can anyone suggest why this might be happening? It seems rather random that some files return all zeros and others work as intended?

Thanks in advance!

[Mathieu NOE]

hello

can you share one of your audio file that causes trouble ?

[LJC]

Can do! I did try to attach them to the original post but unfortunately I can't add .mp3s. Here is one in a .zip!

Answer 1

[Mathieu NOE]

Luke - see above

again :

I found where the issue is on the OP code -

  N = 2^12;
  j = 0:N-1;
  f = (Fs/N).*j;
  X = fft(y,N); % This is _supposed_ to display the graph at the bottom

the issue is that this is not an averaged fft for the entire signal duration !!

this takes only the first N samples of data and display the fft of that single data buffer

read the fft help :

  %   FFT(X,N) is the N-point FFT, 
  %  padded with zeros if X has less   than N points and truncated if it has more.

in the case of that specific audio file , the first N samples are zero (indeed the first 10,000 samples are zero) , so the fft = 0 all the same.

if one would do an averaged fft for the entire signal , this is my home made fft function :

  function  [freq_vector,fft_spectrum] = myfft_peak(signal, Fs, nfft, Overlap)
% FFT peak spectrum of signal  (example sinus amplitude 1   = 0 dB after fft).
% Linear averaging
%   signal - input signal, 
%   Fs - Sampling frequency (Hz).
%   nfft - FFT window size
%   Overlap - buffer percentage of overlap % (between 0 and 0.95)
signal = signal(:);

samples = length(signal);
% fill signal with zeros if its length is lower than nfft
if samples<nfft
    s_tmp = zeros(nfft,1);
    s_tmp((1:samples)) = signal;
    signal = s_tmp;
    samples = nfft;
end

% window : hanning
window = hanning(nfft);
window = window(:);

%    compute fft with overlap 
 offset = fix((1-Overlap)*nfft);
 spectnum = 1+ fix((samples-nfft)/offset); % Number of windows
%     % for info is equivalent to : 
%     noverlap = Overlap*nfft;
%     spectnum = fix((samples-noverlap)/(nfft-noverlap));	% Number of windows

    % main loop
    fft_spectrum = 0;
    for i=1:spectnum
        start = (i-1)*offset;
        sw = signal((1+start):(start+nfft)).*window;
        fft_spectrum = fft_spectrum + (abs(fft(sw))*4/nfft);     % X=fft(x.*hanning(N))*4/N; % hanning only 
    end
    fft_spectrum = fft_spectrum/spectnum; % to do linear averaging scaling

% one sidded fft spectrum  % Select first half 
    if rem(nfft,2)    % nfft odd
        select = (1:(nfft+1)/2)';
    else
        select = (1:nfft/2+1)';
    end
fft_spectrum = fft_spectrum(select);
freq_vector = (select - 1)*Fs/nfft;
end

[LJC]

You're a legend! Thank you so much!

[Mathieu NOE]

one modification needed !! forgot to take the "stereo" version of my fft function

so this is the right function for you (to get 2 fft spectra )

Uz = unzip('Airport Zone - 52 Song Project.zip');
[y,Fs] = audioread(Uz{:});

n = length(y);
t = 0:seconds(1/Fs):seconds(n/Fs);
t = t(1:end-1);
plot(t, y); % This displays the waveform at the top

N = 2^12;
[freq_vector,fft_spectrum] = myfft_peak(y, Fs, N, 0.5)
semilogx(freq_vector,20*log(fft_spectrum));

function  [freq_vector,fft_spectrum] = myfft_peak(signal, Fs, nfft, Overlap)
% FFT peak spectrum of signal  (example sinus amplitude 1   = 0 dB after fft).
% Linear averaging
%   signal - input signal, 
%   Fs - Sampling frequency (Hz).
%   nfft - FFT window size
%   Overlap - buffer percentage of overlap % (between 0 and 0.95)

[samples,channels] = size(signal);

% fill signal with zeros if its length is lower than nfft
if samples<nfft
    s_tmp = zeros(nfft,channels);
    s_tmp((1:samples)) = signal;
    signal = s_tmp;
    samples = nfft;
end

% window : hanning
window = hanning(nfft);
window = window(:);

%    compute fft with overlap 
 offset = fix((1-Overlap)*nfft);
 spectnum = 1+ fix((samples-nfft)/offset); % Number of windows
%     % for info is equivalent to : 
%     noverlap = Overlap*nfft;
%     spectnum = fix((samples-noverlap)/(nfft-noverlap));	% Number of windows

    % main loop
    fft_spectrum = 0;
    for i=1:spectnum
        start = (i-1)*offset;
        sw = signal((1+start):(start+nfft),:).*(window*ones(1,channels));
        fft_spectrum = fft_spectrum + (abs(fft(sw))*4/nfft);     % X=fft(x.*hanning(N))*4/N; % hanning only 
    end
    fft_spectrum = fft_spectrum/spectnum; % to do linear averaging scaling

% one sidded fft spectrum  % Select first half 
    if rem(nfft,2)    % nfft odd
        select = (1:(nfft+1)/2)';
    else
        select = (1:nfft/2+1)';
    end
fft_spectrum = fft_spectrum(select,:);
freq_vector = (select - 1)*Fs/nfft;
end

[LJC]

This updated version works perfectly! Thank you so much!

[Mathieu NOE]

you're welcome