Arc voltage on piercing

I've shot a video to go with this but here is a plot from today
Thermal Dynamics A120, 1.6mm mild steel 40 amps 2219 mm/min (book speed), 100 mm long cut
5 m/sec/sec acceleration (0.5G)
Plots from THCAD-10 with 300 volt full scale.
100 ms per vertical line (0.1 second)



The three signals of interest are Torch voltage (white), arc-ok (blue) and current velocity (green).

You can see on piercing, the voltage ramps up past 300 volts, beyond the full scale of the THCAD. This causes a plateau but that fine as we are not interested in the arc voltage. at that time.

Arc voltage then falls rapidly then overshoots the natural cut voltage, oscillates a bit, then tracks a stable voltage.
Then the plasma cutter returns an arcOK. This quite different to what I observed on my previous plasma cutter where arcOK started while the voltage was oscillating. And also on earlier plots on this machine with lower acceleration.

The other thing I observed was what chance is there of synthesising an ArcOK with those fluctuations by monitoring arc voltage? I think this plot clearly shows why people run into trouble with synthesised arc OK.

So in this case, it seems I could safely set a 0.1 second (100 ms) THC delay. But I would not go under 0.5 seconds without some testing as on my last machine I needed to wait for 1.5 seconds..


I recommend all plasma builders replicate this test before setting a THC delay and make sure the voltage has stabilised before enabling it. I think you are better off with a longer delay than a minimal THC delay.

Not sure how synthesiszed arc-ok works now but it does seem that the large fast drop
in voltage at arc start could be one part of the detection (via DVDT) then perhaps a
fixed wait = retriggerble one-shot, retriggered by any +- deviations > somelimit

PCW wrote:

Not sure how synthesiszed arc-ok works now but it does seem that the large fast drop
in voltage at arc start could be one part of the detection (via DVDT) then perhaps a
fixed wait = retriggerble one-shot, retriggered by any +- deviations > somelimit

Yes, I agree. The voltage plot is actually very similar to the ones I did of void crossing which could be detected by a DV/DT algorithm.

This earlier plot from an Everlast 50s that is included in the plasma primer shows a very different story.
(The original file at a higher res than the primer)

linuxcnc.org/docs/devel/html/plasma/plas...imer.html#_thc_delay

Here the ArcOK (in pink) is enabled before the voltage settles to its equilibrium point.
Here you need to wait out the hump before enabling thc or sampling volts.
So perhaps the better quality machines have better ArcOK algorithims.

I think we need a few more data points not just one machine to base decisions on.
Maybe some others could repeat this experiment

The voltage plot actually looks pretty similar to the first plot except for the much longer
settling time after the arc was established (this might be material + thickness related)

PCW wrote:

The voltage plot actually looks pretty similar to the first plot except for the much longer
settling time after the arc was established (this might be material + thickness related)

Yes but in the second case, the ArcOK is returned before stable volts is achieved so there is a risk of sampling torch voltage or enabling the THC too early and cutting above the desired height. I think (much) slower acceleration has something to do with that.

So anybody wanting to add to the data points, the halscope tutorial is here
linuxcnc.org/docs/devel/html/hal/tutorial.html#_halscope