Signa processing : Detection of noise pulse on DC voltage line

[Hameuza]

Hello All ,

I am working on a labview testbench and I want to be able to detect automatically the presence of pulse noise ( short in time) in a DC voltage line.

Only short pulse that looks like to amortised sinus interest me. Long time oscillation should not be detected.

According to you, what kind of signal processing tool could I use to help this sort of signal detection.

The difficulty is that the signal is short in time and the duration is just few period. So the basic Fourier Analysis seems not  so helpful.

Thank you by advance of any piece of helpful of information.

[johnsold]

Can you show an image of the noise you want to detect? 

You are correct that short duration signals (less than a few periods of the sinusoidal signal) are not so easily detected by the Fourier transform.

Commonly a periodic undesired signal is called interference and the word noise is used to refer to random disturbances.  There is no universal agreement about the dviding line between the two terms, but I would tend to consider your problem one of interference rather than noise.

How big is the interference compared to the DC voltage? What is the range of frequencies of the interference? What is the duration when the interference is present? What is the interval between occurrences of the interference? Are the durations and intervals regular or random? Do you know what causes the interference?

I suspect that some time domain processing might work to detect the interference but until I see the data it is hard to give useful suggestions. What will you do with the information after you detect the interference?

Lynn

[Hameuza]

Hello Johnsold,

Thank you for your interest.

We drive an electromecanical specific device with an analog switch driver own developped.

When a very particular trouble appears  with  the electromecanical during functionning , this is only visible by analysing appearance of these interference on the DC line.

Unfortunnatelly , without entering in all détails of the application, the DC voltage behaviour reflect perfectly what is going on with this particular problem.

The DC level is 3v3 the pulse can reach 5v with a duration of 10µs approx.

The fundamental frenquency is close to 20 kHz.

These pulse are not predictive and reflect only the apparition of the problem.

Ideally, I try to find a method to detect this particular signature to help us react correctelly before demaging the electromeca.

Thank by advance

[Henrik_Volkers]

here is something to play 🙂

I rebuild your source and a add a pulse source.  

Simulating a 500kSPS DAQ capturing 1024 samples , the signal is highpass filtered and the rms value is used as a pulse detector

Try a bandpass filter with 10k to 40k .... or... or ...

[Hameuza]

Hello Henrik,

Thank you very much for the solution using IIR filters.

Unfortunatelly, the interference is non periodic, it occurs randomly and only if the switching device encouters problems.

However, I will use your solution as a good starting point to find the correct strategie combined with a time domain analysis.

The solution should fit correctly the phenomena without producing to much false detection.

Best regards

[Henrik_Volkers]

If you post some real data we migth help you 🙂

If you have a simple capture vi with some typical  data in a graph, just do a 'Edit-Make current values default', save the vi and post it here.

A spectrogram (STFT) of the signal is a nice tool to find strategies !