Cannot get arc-ok in Mode 0

Hello!

I am trying to set up plasma machine. I have Mesa 7i96 and THCAD-300.
Customer wants to use their existing plasma source - Telwin Enterprise Plasma 160HF
The thing is that it is plasma source for manual cutting, so it does not have any inputs or outputs for cnc controls.
I put a relay in cabinet and connected it in parallel with torch switch for "torch on" and connected two wires to actual work leads of plasma torch cable and earth clamps for THCAD board.
My current problem is "arc-ok" signal.
I had no idea, how to do that in a safe and reliable manner until literally yesterday I found out about PlasmaC (although I am using LinuxCNC since 2010, but somehow had missed this component) and possibility to determine arc closing from plasma voltage.
What is puzzling me is this:
The measured voltage value is strange. First of all, particular value does not make sense to me considering the encoder scale etc values. Second of all, the measured value decreases, when voltage is applied to input pins of THCAD board.
Here are settings in PlasmaC:
PlasmaC settings
In HAL file for encoder module I have:
setp hm2_7i96.0.encoder.00.scale -1
setp hm2_7i96.0.encoder.00.filter 1
setp hm2_7i96.0.encoder.00.counter-mode 1
On THCAD board the sticker of calibration results say 120.5kHz for 0V input and 961.5kHz for 300V input
The jumper is set for 1/32 frequency division.
So basically with no voltage applied to THCAD, I should get 120.5/32 = 3,766 kHz output. With encoder scale = -1, encoder velocity should be around 3766. That is exactly what I see here (please note that I am probing plasmac.arc-voltage-out pin, that lower region of the curve is when PlasmaC tried to turn the torch on - it actually did turn on, but "arc-ok" state was not determined):
Torch off

Notice the placement of mouse pointer and the value at the bottom 3720.

Here is the value of measured volts, when the torch is on and the arc is established:
Torch on

Yes, when torch is off, there is some noise that seems to be a stable frequency. I tried to play with plasmac.lowpass-frequency, but am not sure I did set the value correctly (I tried to determine the frequency from the HalScope plot by calculating the time between the spikes and then trying to convert that to frequency value. It might be pretty good, because it looks way better than encoder.00.velocity graph.
Here are some pics of how encoder.00.velocity pin looks - the noise is when plasma is turned off, while the flat line is when plasma is actually turned on and the arc is established.

encoder.00.velocity
encoder.00.velocity

This is zoomed in on the noisy part. Seems pretty uniform to me:
EDIT: these 2 links will be posted separately, because the site complains that it is too much links here already...

The weird part is that with torch off the encoder velocity value of 3766 is more or less expected. What I do not understand, how the velocity value can be lower, if there is voltage applied to THCAD inputs. 3,76kHz output should be the lowest that card will do. And then suddenly I can get 3,05kHz with torch on. I can get something similar by applying 12VDC from PC PSU to THCAD inputs - the value also decreases, which makes absolutely no sense to me (yes, I tripple checked the polarity and also double checked that card is set for unipolar measurement).
Next weirdness that I did not find explanation for is the fact that on Axis GUI the "Arc voltage" is around 740, when the torch is off.
Encoder velocity is at 3760, plasmac voltage scale = 1, plasmac.arc-voltage out is 3760, but in GUI it shows approximately 740 (it is fluctuating due to the noise), which is approximately 5x lower. Where did this difference come from?
But do not worry, there is more weirdness that I do not understand. When the torch is turned on, encoder.00.velocity pin value drops to 3051. Plasmac.arc-voltage-out drops to 3051. Considering previously mentioned 5x difference between On Axis GUI "Arc voltage" should drop to 5x lower value of something around 610. But guess what - it was 51.
And how do I know, what values should I put in "OK High volts" and in "OK Low volts"? Which one of them should I try to catch? 610 or 51? Obviously 51 was not the right answer, because it falls inbetween 10 and 600, but "arc-ok" was not triggered.

I just would like someone to pat me on the shoulder and tell me, if I am doing this correctly or am I misunderstanding some basic concept of setting this machine up.

This is zoomed in on the noisy part.
Close view
Closer view

You need to set the Voltage Scale and Voltage Offset according to these instructions:
linuxcnc.org/docs/2.8/html/plasma/plasma...ng_the_divider_ratio

You can use an online calculator once you know your divider ratio:
jscalc.io/calc/NTr5QDX6WgMThBVb

With regard the noise, I have had similar issues which were solved by careful attention to a star ground to every motor and over every sliding linear rail AND the addition of a RFI/EMI filter on the mains power input. I bought a cheap USB oscilloscope so I could see what was going on and found most noise was entering via the mains power input.

So the other comments about HF (you must connect your thcad before the HF start module) and scaling the output still apply.

You could also encourage your customer to upgrade to a later TELWIN model with a CNC port. It is interesting because it also has an input allowing current adjustment via 0-10 volt PWM from memory.

tommylight wrote:

Would you please remove those links and upload the pictures here on the forum???

Ok, I attached the pictures

tommylight wrote:

viesturs.lacis wrote:

I am trying to set up plasma machine. I have Mesa 7i96 and THCAD-300.
Customer wants to use their existing plasma source - Telwin Enterprise Plasma 160HF

From the THCAD manual, also mentioned several times here :

The 300V version of the THCAD has all input divider resistor on board and can connect directly to the plasma voltage of touch start type torches only.

viesturs.lacis wrote:

and connected two wires to actual work leads of plasma torch cable and earth clamps for THCAD board.

And that is the worst place to wire a THC as it will suffer through 20000V or more every time the torch is fired.

Well, that is not good. I specifically looked at the manual to check, if 300V version will do and my conclusion was "yes", but now I have no idea, how did I get that or where did I look, because yes, I just looked at the manual again and it says your quoted sentence in the first page...
Ok, I will be there again tomorrow and I will take some pictures of the plasma source.

rodw wrote:

With regard the noise, I have had similar issues which were solved by careful attention to a star ground to every motor and over every sliding linear rail AND the addition of a RFI/EMI filter on the mains power input. I bought a cheap USB oscilloscope so I could see what was going on and found most noise was entering via the mains power input.

I already have the filter installed and have grounded the main frame, control cabinet and shields of all cables. I will post some pictures tomorrow.

rodw wrote:

You could also encourage your customer to upgrade to a later TELWIN model with a CNC port. It is interesting because it also has an input allowing current adjustment via 0-10 volt PWM from memory.

Unfortunately that is not going to happen. We already had the discussion and they want to go as cheap as possible so using existing plasma source is a must.

viesturs.lacis wrote:

Unfortunately that is not going to happen. We already had the discussion and they want to go as cheap as possible so using existing plasma source is a must.

Thats the problem. Really the words CNC Plasma cutting and as cheap as possible don't really go together.

rodw wrote:

viesturs.lacis wrote:

Unfortunately that is not going to happen. We already had the discussion and they want to go as cheap as possible so using existing plasma source is a must.

Thats the problem. Really the words CNC Plasma cutting and as cheap as possible don't really go together.

I disagree, cheap as possible is the reason I came to LinuxCNC originally.