Plotting FFT Amplitude Spectrum signal in Matlab

user3488736 Source

I've trouble with plotting a correct FFT Amplitude Spectrum signal in decibels and Hertz.

Firstly, I just plot the FFT Spectrum signal in Hz and magnitude like this:

figure;
X_mags = abs(fft(signal));
bin_vals = [0 : N-1];
freq_ax_bins = bin_vals*fs/N;
N_2 = ceil(N/2);
plot(freq_ax_bins(1:N_2), X_mags(1:N_2));
title('FFT Spectrum signal 1');
xlabel('Frequency (Hz)')
ylabel('Magnitude');

This results in the expected plot with magnitude always > 0. Finally, I just want to do the same but in decibels:

bin_vals = [0 : N-1];
freq_ax_Hz = bin_vals*fs/N;
N_2 = ceil(N/2);
figure;
plot(freq_ax_Hz(1:N_2), 10*log10(X_mags(1:N_2)));
xlabel('Frequency (Hz)')
ylabel('Amplitude (dB)');

It looks good but the plot is partially drawn into minus dB. Can someone show me how to correctly draw the spectrum in decibels?

matlabfft

Answers

answered 3 years ago crowdedComputeeer #1

For your first plot, I notice you're only plotting the first half of the signal. Another, maybe easier way to do this is with fftshift:

>> Xmag = fftshift(abs(fft(x)));

Now the DC value is in the middle rather than at the beginning. The frequency vector is (-N/2:N/2 - 1)*fs/N

If your signal is not symmetric, then you need to look at the negative values.

For the second, note that dB requires 20*log10 unless you take the square of Xmag. Not a big deal, just a scalar, but thought you'd want to know.

Also, minus dB are defined and expected. The log(-1) is not defined for real numbers, but log(.1) or any other number less than 1 returns a negative answer.

Hope this helps!

comments powered by Disqus