Halscope of dv/dt during a kerf crossing

suggestion.... use a kalman filtering algorithm to track the voltage, and estimate / calculate the standard deviation .... anything outside of 1,2,xxx (adjustable) standard deviations is ignored / THC is locked from downward travel.

There is an antidive formula that I've come across in this manual (pdf page 30/.38) but I'm not sure where it came from originally
minithc.com/pub/MiniTHC2_EN.pdf

I am under the pump so don't have much time to ponder this but I do want to direct you to this graph in our newly published Plasma Primer linuxcnc.org/docs/devel/html/plasma/plas...#_void_kerf_crossing

This is real data from a live cut. I captured the data out of halscope and used n = 20 for the moving average in a spreadsheet. The data was adjusted to be +- the cut volts (eg cut height = 0 volts in the graph). Given the sampling time of halscope, I figured that the 20x average should be equivalent to n = 100 readings at the servo thread fperiod.

From this, I went on to capture 16,000 data points using the sampler component at different cut heights. Halsampler needs to be modified to output CSV format as it was time consuming to import the data. Therefore, I only did it to calculate the volts per mm for cutting height via linear regression shown in the graph under this heading
linuxcnc.org/docs/devel/html/plasma/plas...mer.html#_cnc_plasma I think we need to capture som ekerf crossing data in this way.

So the intent of the averaging was to create as slow moving oscillation (red line) around the set point that could be used to calculate the variance for DV/DT. I should revisit the formulas to make sure what I did in the spreadsheet is what the component is doing.

Finally, it would be fantastic if you could publish a plot of a standard cut with the ESAB without any kerf crossing. so we have a base line view