Condition reset after data goes below a certain value

[swilson8]

I'm developing a peak detection algorithm. I was wondering is it possible to create a condition such that when one point is detected, the data must  go below a certain value before another point can be detected?

[altenbach]

Yes, it is possible.

(Keep the condition in a shift register or feedback node.)

Have you tried the built-in peak detector?

[swilson8]

I have attempted the peak detection, but it picks up a larger peak twice which I would like to stop by saying that the value has to drop below a certain point so it moves to the next peak. I have tried other methods which would also benefit from this condition.

I need to develop another form of algorithm to the built-in vi as I would like to compare different methods and their effectiveness to the built-in detector.

Thank you

[altenbach]

Can you comment on the VI you have attached, because it seems extremely convoluted. What is the purpose of showing us this? Is it your attempt at "peak detection"?

• The leftmost FOR loop does not need anything orange. What's the purpose of creating an array when you never use the output?
• Wiring an array to a Waveform attribute makes no sense unless you are using something undocumented. can you explain.
• Your small loop with the conditional tunnel removes data from the y-array but not from the x-array. Seems lossy.
• Your boxcar averager could be simplified dramatically (image).
• Shouldn't your pbbypt min&max be initialize on the first iteration?
• It seems unlikely that your (large) boolean array to number ever results in a 1. What's the logic behind that?
• Your loop to "remove duplicates" will be inefficient for large arrays. The resulting array of amplitudes with probably have a very different size than the x-array created and the positions are unrelated to the x-values. 
• Why are front panel and diagram maximized to the screen?
• ...

Again, why do you show us this VI. It seems quite unrelated to the problem and looks completely broken in functionality..

[altenbach]

---

@swilson8 wrote:

I have attempted the peak detection, but it picks up a larger peak twice which I would like to stop by saying that the value has to drop below a certain point so it moves to the next peak.

---

It seems to pick exactly 6 peaks. looks fine to me. 😄

[swilson8]

Sorry for the slow response and the state of the vi. I've been testing lots of things at a time so the vi is rather experimental as I was trying to see what would solve the problem.

The Boolean section was an attempt to create a condition that would mean that only values where dx/dt = 0 and d(^2)x/dt(^2) < 0 would be detected. It kept throwing out an issue where the source was Boolean and the sink was Boolean but it wouldn't accept it for the case structure, I was trying to (but failing) to overcome this.

Your peak detection does appear to give nice values, but it still uses the vi which I am unable to for the purposes of this. This is why I was trying to create this new condition where a peak is detected and then the data must drop below a value before a new peak is then detected.

Thank you

[altenbach]

Every derivative emphasizes noise and your first and second derivative criteria are not suitable for noisy and spaced data.

NI has worked hard developing a robust peak detector and I seriously doubt that you could do better. Unless you only have LabVIEW base, that's what should be used. What's the reason you don't?

[cstorey]

Well if you aren't going to use the NI built-in VIs then you should poke around the web for pre-existing solutions.  There are many available in many languages you could translate and test.  One popular one is by Eli Billauer - http://billauer.co.il/peakdet.html

Its quite a clever simple solution to finding peaks/valleys but I haven't tested it on real data to compare to NI VIs.  It works for your data, but probably not all data.

Here's a LabVIEW version of Eli's code, added to altenbach's tidy code for comparison.