Quick Question On Pulse Measurements

[johnsold]

About 30 years ago I designed and built a weld current monitor for a manufacturer.  It was all analog circuits and I do not remember any of the details.

Can you monitor the current?  Rogowski coils or Hall effect devices would allow sensing the current without modifying the circuit.

What does the spectrum of the sound look like? Post some actual data and I will try a few things.

Lynn

[NateTheGreat]

Yeah I can, I will attach a couple graphs tomorrow morning. 

[DFGray]

As I understand your problem, you are trying to count the big spikes, but some of the bigger ones have small spikes which add extra counts.  A simple peak find followed by a removal of all peaks which are too close together should do the trick.  Run the Peak Detector VI on your data, using a threshold for your lower peaks.  Using the Locations output, remove any peaks that are too close to the previous one.  Take the size of the result.

[NateTheGreat]

Thanks Grey, I didn't know about the peak detector. That makes a lot more sense to use in this situation than the pulse detector.

Oh, and Lynn I can't actually post any data right now. I will see what I can do later. But if you are interested in sound vs current this paper

has some good graphs of sound vs voltage and current. Also it would be much easier to do this using just current like you suggested but I am interested in seeing if I can do it with just sound first.

Here is the power spectrum of the Short Circuit weld like you asked Lynn. I haven't really found it useful yet. This was shorting at 64 hz and I was hoping to see a peak there, but no such luck.

[johnsold]

That spectrum surprises me a bit also.  Apparently the frequency of the pulses varies enough from one cycle to the next to smear the spectrum.

Dr. Gray's suggestions are always worth trying.

Lynn

[NateTheGreat]

I was dissappointed in the spectrum. I was thinking about putting a piezoelectric sensor on the weldpiece to measuring the vibrations. Im pretty sure that would give a better spectrum.

Oh and I did take Gray's advice on using the peak VI instead of the pulse. It works quite a bit better.

[johnsold]

There was a thread a while back where someone was comparing acoustic and acceleration signals from the same source.  The accelerometer picked up less background noise but otherwise there was not much difference.

I was thinking about the spectrum of short time segments. Perhaps about 5 ms long. Then you could possibly distinguish between the various sounds.  The spectrum of a segment containing multiple pulses gest spread out as you have seen because the pulses are not truly periodic.

Lynn

[NateTheGreat]

Hmm, that is good to know. I will see if I can dig up a accelerometer, but if I don't have one lying about I guess I won't worry about it.

I'll see what smaller segments look like, the average shorting period is usually 10 ms'ish long.