Plasma made in a barn as cheap as possible :) and it's not going smooth

He, so finally got the thing, almost, running.

Quick presentation, i got for free a chinese laser table, it was outside in the rain for like 6 months
I think the tubes and the HV supplys are shot, it was modded / reparaired over the years, it's from 2006-2008.
Anygot got it for free, so no complaints, it hase some HIWIN rails (15mm though) good enough for the plasma and the original cutting area is 1.2M x1.5M, but from the size and constraints of the plasma gantry/torch it's down to 1.37Mx1.1M.

The stock motors X & Y are 3 phase NEMA 23 steppers (they drove me crazy btw) with belts and a 24/72 reduction.
Drivers are good up to 300kHz, so i'm running them as they were set, and it ends up with a 400 steps/mm and a fast move due to 3 phase.

At the moment the hardest part for me is integrating in my head how Linux and LinuxCNC works coming from a windows environment and no programmer of course

For history, i have those threads solved if they may help some1
1_ forum.linuxcnc.org/9-installing-linuxcnc...inuxcnc-latency-test
2_ forum.linuxcnc.org/plasmac/38708-plasmac-7i96-and-thcad-setup


And now i'm running in an other issue, the THCAD has a permanent voltage even when the torch is OFF, there is a permanent 30V between nozzle and ground, thus being a cheapo chenise plasma no surprise there, and it's a HF start to make things worse. But i have for 5 years now and it still works so no complaint.
So i found this topic that relates exactly to it : forum.linuxcnc.org/plasma-laser/38636-va...mac-override?start=0
But it's lockdown here and i cant find a REED switch anywhere, so i rigged the entry ports of the THCAD to an SSR and i got the arc ok signal ( changed from Mode 0 to Mode 1 btw) but lost the 10v input as the SSR kind of shorts it out. Thinking of re-openeing 1 more time the plasma cutter to connect the SSR before my resistor divider/drop for the 10v on THCAD10. I guess it should work.
With my SSR in place i can run the tool without THC, cuts nice on 4mm up to 6mm but on thinner it really needs the THC.

A couple of mismatched videos and photos taken during the process.



And in this video, i can see a strange behaviour, no THC anebled, :
_ the torch goes down to probe the surface >> goes back up >> pierce >> goes back down touching the plate >> goes back up and starts the move

re the voltage, it has to be a configuration issue. You really need to be methodical in testing each element in the chain.

So start with the THCAD calibration sticker. What is the 0 volt frequency? What is the divider link on the THCAD set to?
Multiply the calibration 0 volt by 1000, then divide it by the divider link setting. Lets say that is a frequency of 3800.

So open halshow and find the encoder.0.velocity and watch the pin. At 0 volts, you should be seeing 3800.

So then 3800 should be your offset setting. If its not agreeing, then sort that out!
Then, consult the Plasma primer and do these calculations
linuxcnc.org/docs/devel/html/plasma/plas..._with_the_mesa_thcad
and take the reciprocal of your frequency per volt and that is your voltage scale for plasmac.

You should be sweet once that is done!

So now bout your torch behaviour. When the torch fires, the voltage rises rapidly to about 130 % of your cutting volts. Depending on material thickness and amperage, this can vary a bit but afterwards this peak should settle to a stable cutting volts without any THC function.

So I think either you have auto volts enabled with too short of a time delay (so it samples that piercing peak) or you have set the voltage too far away from what it really is so the THC makes a rapid adjustment to your set point.

My recommendation would be enable auto volts and use a THC delay of 1.5 seconds. Then the peak will definitely be over and the set point will be exactly on the naturally ocurring cutting volts. Once you get going, you may wish to reduce the THC delay but it would be good t do a few Halscope plots of arc voltage to observe the piercing peak voltage to determine a minimum THC delay you can get away with but 1.5 seconds should work on all materials based on my testing on two different plasmas.

I used this calculator for the gain and offset : jscalc.io/calc/NTr5QDX6WgMThBVb

And I used a voltmeter to check between the nozzle and the ground clamp, there is 30V fluctuating always present, as i'm connected to the THCAD it sees it permanently, and yes there is a 130-150% voltage increase when firing the torch.
I have no clue about how to use HALSCOPE but i used a real oscilloscope to see what was going on. I pikes up to 195V then drops to 90-140V when cutting depending on output setting and thickness.

The auto/enabled THC makes no difference, it fires the torch and stops after a sec with " VALID ARC LOST" same as on this topic : forum.linuxcnc.org/plasma-laser/38636-va...mac-override?start=0
That might be solved if i can connect the SSR on the part that has the feed back loop controling the IGBTs in the cutter, i might use a darlington setup for minimizing the impact of the SSR on the circuitry.

The part where the nozzle touches the the plate after the pierce delay is there even when THC disabled

did you first try to cut without thc
it seams to me that the power for this sheet is way to high
it semas like a 3mm sheet and you can cut this with 25A at 3000mm/min at a fixed hight
also where did you connect your voltige mesuring it needs a thin cable 0,25mm²
so fluctuation is low

so i rigged the entry ports of the THCAD to an SSR and i got the arc ok signal ( changed from Mode 0 to Mode 1 btw) but lost the 10v input as the SSR kind of shorts it out.

You should _Never_ attach any control signals to the analog side of the THCAD
This is a safety and EMI issue

yes i'm doing it without the THC and it's @ 30A around 3000mm/mn

I opened the plasma cutter and soldered 0.25mm² wires on the output of the IGBTs , so it's a direct connection, there is a 2x1MegOhm resistor for dropping the voltage then the THCAD

@_PCW_ i did that to see if it's working properly at detecting the arcOK signal witout the THC.

From what i see with a scope ( real one not HaLscope) the cutter as a permanent 30v (fluctuating) between the output and the ground so the THCAD always sees 30V andwhen the plasma fires it goes up to 180V (spike) sometimes 250V then drops to the normal cutting voltage.
BTW it's a HF start chineesium made.

Well having 30 volt on the THCAD while not cutting is not really a show stopper if the voltage reported in plasmac while cutting agrees with what you have observed on your scope.

Plasmac's synthesised arcOK algorithm has a design fault becasue it specifies a low limit and a high limit for ArcOK and only generates an ArcOK if the voltage falls between that range. In my view, there should not be an upper limit. This can now be worked around by setting the upper limit to 9999. The lower limit should be about 70% of the cutting voltage you have observed in your scope.

But I think it will be best to use Mode 1 and not use the synthesised ArcOK at all. (mode 1). Your SSR has its own troubles.

Instead of the SSR, try adding a time delay linuxcnc.org/docs/devel/html/man/man9/timedelay.9.html
to the plasmac:torch-on signal in your connections HAL file.

Use setp to set an on delay of 0.2 seconds and an off delay of 0 and use the out pin as the AcOK signal. The on delay may need some experimenting as there could be some lag from when Linuxcnc tells the torch to fire and when it actually does. This could be checked by using Halscope and plotting the signal above and the arc voltage.

This still does not address the torch dive but that is related to a mismatch between the observed cutting voltage and te voltage you have set in plasmac. I discussed this earlier.