Use numerical integration on specific areas of graph

Hello

I'm doing some double pulse testing, and I'm trying to post process the results. I can get the rise and fall times of the signal relatively easily from the signal processing add-on, but I am struggling to understand how to calculate the switching losses easily. The graph I end up with for the power loss (voltage*current) is looking like this:

and I would like to integrate the power versus time only in the spots the spikes appear. For example in this region while zooming in :

Right now, I am brushing and exporting data, which I'm sure is inefficient. Does someone know a fast way to achieve these calculations?

Many thanks in advance.

[Walter Roberson]

findpeaks()?

[Greg]

Many thanks for your reply; the function itself is relevant and useful. It's a bit tricky to get it to work due to the noise (since it's an oscilloscope measurement and also has circuit ringing) , but playing with the min peak height and distance I'm getting there.

My only issue left is, I'd ideally get the width of the pulse so I can script the locations in width to do numberical integration, when I'm trying this on MATLAB I seem to only be able to choose between half prominence or half height. Ideally I could set this to start counting the width from a lower spot?

[Walter Roberson]

Please show your complete call to findpeaks()

[Walter Roberson]

You do not need to provide any particular option to get findpeaks() to return the width: you just have to have at least 3 outputs from the findpeaks() call.

[Greg]

My apologies, I have worded this wrongly. This is my function call that is incorrect:

[pks,locs,w,p] = findpeaks(power,time,'MinPeakHeight',100,'MinPeakDistance',50,'WidthReference',0)

The thing is, I have found the width is underestimated compared to the real graph. This is because the width is calculated from around the "mid-point" of the pulse's height. Unfortunately it seems there is no straightforward way to set this width to be calculated from a lower point and be more accurate?

[Walter Roberson]

You could take the peak location and search for the first point to the left and right that are <= 0 . But if you have phase on your peaks then it is common for the pulse to go below 0.

If you

outside = find(data <= 0);
before = interp1(outside, outside, PEAK_LOCATIONS, 'prev');
after = interp1(outside, outside, PEAK_LOCATIONS, 'next');

then you can code the searching in a small number of lines.

[Greg]

Thank you @Walter Roberson, This has now been fixed according to your suggestions.