If you’re still not getting results with averaging of the PSD you need to be clear what the difference is between a good motor and a noisy motor. From what you say the noise is distributed across the spectrum. Even if your motor is running at constant speed I would expect variation in the PSD output from one iteration to the next, which is why I suggested adding averaging to the PSD output. When the test is running how much variation do you get in your Low Peaks summation?
I’ve modified my program example. Where I have ‘Signal’ you will need to feed in your ‘Low Peaks’ value This should prevent the jitter you experience. The difference between the values of 'Threshold High' and 'Threshold Low' must be greater than the variation you get in your ‘Low Peaks’ summation.
I think your program will be better if you have two loops, one for the Data Acquisition and one for your processing.
If you can give examples of several PSD output data sets for good and bad motors it may be possible to suggest a better way to identify your noisy motors.
