Best LabVIEW methods to compare two signals

[sticyfinger]

Can someone tell me what is the best method(s) to compare two waveform signals (capture from NI DAQ device) using LabVIEW? I am trying to come with method to compare to voltage signals. One of the signals will be the known good sample (the reference waveform). And the second waveform will be the one that is currently being tested. The second signal will be upgrade every time the test is run. The only solution I came up with is to take the integral of the two signals and compare them. Can you provide me with other methods/solution? Also what LabVIEW Vis could help me accomplish this task? Any singal characteristic I can obtain using Labview Vis. Really appreciate it. FYI. The two analog signals are voltage measurements (that represent pressure). Both have the same period and same frequency. Thanks!!!

[johnsold]

It really depends on what characteristics of the signals you need to compare.

Even though the signals have the same frequency, unless they are both generated simultaneously by the same source, there will be frequency errors.

How much phase shift can occur? How much noise is present? How small a difference must be detected (amplitude, frequency, phase, distortion, ...) to declare the test signal bad?

Lynn

[sticyfinger]

Thanks for the fast response.

I am not sure about how much noise we have.  And I do not think we have to worry about the phase shift.

As for what classify as a failure I may need some more time to make that judgment since I am waiting on new hardware to determine what classified UUT (unit under test) as failure.  When it come to this I am basically a novice so  I wanted to see what methods was out there and starts preparing code.  

I figure there were several common techniques I could apply to waveform I have.   The waveform I have measure start and figure of test. My test has two parts. I am mostly concern with the first part.  When I get the new hardware I will be able just get waveform on first part.

Attached are images of the waveforms I am currently getting.  Good result. Bad result. And worse result. 

[johnsold]

I see one obvious difference between good and bad results: the slope of the portion starting at 1.4 V.  If that is consistently related to the failure mechanism of your devices, it would be easy to test.  Find the minimum. Go 1.5 to 2 seconds after the minimum and check the voltage there.  Good is less than 1.6 V, bad is greater than 1.6 V!

If this is the only type of failure you do not need to compare waveforms, only two points.

Lynn

[sticyfinger]

Sorry, I forgot the last type that need to be test. There is a 4^th type, which is harder to determine visual. The fourth type is why I wanted to use a method (or methods) to compare the two signals.

With the new hardware I will be able to focus on first part better.  But just in case, I still cannot tell the different I will need a method to determine how the waveforms are different.  

*FYI: the waveforms I measured show the start (first part) & the finish of my test (2^nd part). 

I was wondering if smoothing out my current data will show the different from the good and the sort of bad waveform data  visually. I found these YouTube series about smoothing below.

My main goals is to use my LabVIEW program to compare two waveforms.   

Youtube series links

series 1 : https://www.youtube.com/watch?v=AxlyqR-vGe0    

series 2 : https://www.youtube.com/watch?v=E3CuIQ9V73Y

series 3 : https://www.youtube.com/watch?v=jq1wPp0o-Fg    

series 4 : https://www.youtube.com/watch?v=JWYliXDiDsM

[johnsold]

Before you work on methods to spearate the good from the bad, you need to carefully and explicitly define what is good and what is bad. What is it about the 4th type which distinguishes it from a good part? Just because you cannot see a difference does not mean that you cannot measure it, but you must decide what you need to measure.

Is it possible that the device is bad but generates a perfectly good waveform under these conditions? For example the waveform might be fine but the device temperture is too high. In that case an addtional measurement - the device temperature - would be required.

I doubt that smoothing will be of any value on the waveforms you have posted.

Lynn

[Henrik_Volkers]

There are many ways that have been used for this kind stuff:

have a look at the waveform monitoring vis:

Limit testing

or

pulse durations, rise/fall times and positions, histogram , integral, DFT

one way is to ask yourself how you decide, but you can extract all the above indicators (and XX more) and try to train a neural network ...  

[sticyfinger]

Thanks for all the suggestions. Once I smart running the test with the newer hardware, I will post this question again (with more details). That will take couple weeks.

[Rihanmalik]

Hi I am working with the same task, as I am very new to the LabVIEW I dont have much idea about signal waveform comparison. Can you please help me how you did it, what technique did you used.

Thanks & Regards

Rihan Malik