Problems with signal processing

I have a signal as shown here:

The x axis is just the sample number and the y is the ampltiude of the signal in volts.

There are ten 'bursts' where the amplitude of the signal spikes and dies out before rising again. I would like to obtain the fast fourier transform of each of these 10 bursts individually and plot them on the same figure. I'm not entirely sure where to start and any help would be greatly appreciated.

Answer 1

[Star Strider]

For a time-frequency plot, I would use the pspectrum function with the 'spectrogram' option, at least for an initial approach, although that might be all you need. (The pspectrum results are a bit easier to interpret than the spectrogram results.)

[Dan Lecoutre]

I'm not sure I understand what you mean, how would this function help me?. I am trying to plot a fast fourier transform just of each individual burst. At the moment this is code I am using to plot the fft of the first burst, and I have to manually input the x values for burst1. I'm looking to write some code that will obtain the x values for which each burst starts and ends and then plots the fft of this burst. At the end I would have the fft of each burst on one figure.

[Star Strider]

It displays the Fourier transform of each burst as part of the time-frequency plot.

That will tell you if it is worth analysing further. If so, the stft function would be worth considering.

[Dan Lecoutre]

I have my data vector that is of dimensions 1x24e6

In order to use the pspectrum function on the signal I attached in the first message how would I write the code for the spectrum function?

x=pspectrum(src1.Data);
plot(x);

This is how I am trying to write it but I dont think I'm using the function correctly

[Star Strider]

Ideally, it would be something like this —

[p,f,t] = pspectrum(src1.Data, Fs, 'spectrogram')

figure

waterfall(f,t,p)

or something similar, where ‘Fs’ is the sampling frequency. (Most signal processing functions assume that the sampling intervals are the same throughout the signal.) See the documentation for details.

With a vector of that length, it should be possible to get a fair amount of detail.

If you want to post it here, it would be necessary first to use the resample function to downsample it to the required 5 MB length, or to post only the first 5 MB of it, or some 5 MB subsection of it, assuming that contains at least one spike. It looks like there might be two signals in the original plot image, so choose a representative sample of one or both to attach. Use the paperclip icon in the top toolstrip to do that.

[Dan Lecoutre]

I have attached the first spike in the data, I have the sampling frequency already. The data contains the amplitude in volts for each recorded sample of the first burst. Would the pspsectrum function be able to plot the frequency against the magnitude in volts?

[Star Strider]

Here are three ways of depicting it, demonstrating the time-frequency characteristics of the spike —

LD = load(websave('burst1%20of%20data','https://www.mathworks.com/matlabcentral/answers/uploaded_files/1338969/burst1%20of%20data.mat'));
burst1 = LD.burst1;
Fs = 1;                                             % Assume Default Sampling Frequency (Value NMot Provided)

[p,f,t] = pspectrum(burst1, Fs, 'spectrogram');

figure
waterfall(f,t,p.')
set(gca,XDir="reverse",View=[60 60])
ylabel("Time (s)")
xlabel("Frequency (Hz)")
xlim([0 Fs*0.01])
ylim([0 max(ylim)/3])
colormap(turbo)

figure
surfc(f,t,p.', 'EdgeColor',[1 1 1]*0.5)
set(gca,XDir="reverse",View=[60 30])
ylabel("Time (s)")
xlabel("Frequency (Hz)")
xlim([0 Fs*0.01])
ylim([0 max(ylim)/3])
colormap(turbo)

figure
contourf(f,t,p.')
set(gca,XDir="reverse")
ylabel("Time (s)")
xlabel("Frequency (Hz)")
xlim([0 Fs*0.005])
ylim([max(ylim)/10 max(ylim)/4])
colormap(turbo)

It would help to have the sampling frequency (I did not find it anywhere) however the plots should scale with it.

Ths Z-axis units are in terms of power, not decibels. Use the pow2db function to convert them, if desired.

.