Peak detector VI Analization with code

[Crtomir]

Hello,

 

I have created code to filter a signal and calculate its peaks. The code is doing perfectly good until..... the signal frequency gets lower (about 10 Hz).

 

The problem is as shown in Peak detector VI. My code needs to execute under 100 ms. Because Peak detector VI takes to much time at low frequency, my program fails.

My thinking: Peak detector does a quadratic fit on the points (maybe here hides the time problem ?). I already did FIR filtering, so quadratic fit is not needed?

 

I have added the code for you all to test and use it if you want (try changing frequency at the top of graph and look at how calculation takes more time).

 

Please comment on what i created and if there are some FASTER peak calculation alternatives... for us all to use.

 

PS: you are looking at a test. This is not my program arhitecture (queued state machine)

 

Thank you

forum question peak finder.vi

[drjdpowell]

I suggest you just decimate your data after the filter.

[Crtomir]

I added continuous decimation and already at decimating factor = 2 the speed is 3x faster ! So execution problem is solved

 

Thank you, your answer is excellent.

 

Still i'm curious to know. Would you as an expert, do the same peak detection and filtering as i did?

 

 

forum question peak finder with decimate.vi

[Chris_Collier]

One note: you will want to move your location back to where it should be if you're decimating (multiply by the decimation factor).  There're a couple other logic things which are also redundant.  The biggest time sink would probably be polling your controlls.  Try this version (it's in 2013).

And of course I didn't test my code properly, and it seems to run more slowly than yours... I'll have to dig more deeply into why...

Peak Finder 2013.zip