How to find occupied bandwidth

Greetings,

I am trying to estimate the occupied bandwidth of the measured signal. After post-processing, I arrived to the spectrum estimate "powerSpectrum.mat" and frequencies "freq.mat" (both files are attached).

    load powerSpectrum.mat powerSpectrum
    load freq.mat          freq

    figure
    obw(powerSpectrum,freq);

The returned occupied bandwidth does not quite make sense to me. Is there anything that I did not do? or did wrong?

I would appreciate your help.

[Star Strider]

How did you calculate ‘powerSpectrum’ and ‘freq’?

[kauerbach]

I applied easyspec.m to the measured waveform:

A spectrum-analyzer like plotter for MATLAB

Answer 1

[Greg Dionne]

You have a fair amount of noise power, which is swamping your measurement. To see what I mean try (assuming your spectrum is a PSD -- you were using the PSD syntax above):

plot(freq/1e9, cumsum(powerSpectrum)./mean(diff(freq)))
xlabel('Freq (GHz)')
ylabel('Cumulative Power (Watts)')

Note the large slope in cumulative power in the noise regions.

If you want to ignore the noise, try restricting the frequency band of interest. Something like:

obw(powerSpectrum, freq, [4.75 9]*1e9, 99)

Something looks awry in the band between 8.5 GHz - 9 GHz. Is this a real world signal?

[kauerbach]

Thank you for your reply.

This was a real signal before denoising. The notch between 8.5 and 9 GHz probably occurred because one of the cut-off frequencies fell in the middle of the ripple. You can see reminiscences of other ripples on the peak itself.