The link shows only the plots.
This applies to the second plot (of the one-sided Fourier transform of the signal).
First increase the frequency resolution of the fft call that created that plot by zero-padding it using:
NFFT = 2^(nextpow2(L)+2);
and then include the ‘NFFT’ value in the fft call. The frequency vector will need to reflect that, my usual approach being:
Fv = linspace(0, 1, NFFT/2+1)*Fn; % Frequency Vector For Frequency Domain Plot
Iv = 1:numel(Fv); % Index Vector
where ‘Fn’ is the Nyquist frequency for the signal, and ‘Iv’ indexes the fft results appropraitely.
Second use xlim to isolate the x-axis between 3.5 and 5.5 kHz to see the detail. The findpeaks function (on the same signal as being plotted) can be helpful in determining the indices of the peaks, and from them, the frequencies and amplitudes.
The frequency resolution of the fft is important, so increase ‘NFFT’ in powers-of-2 (for efficiency) untli the plot demonstrates adequate separation of the carrier and sidelobes.
.
