THC problem

I am having a very odd problem, I am running QTPlasma and using the built in THC. I can cut parts and sometimes the torch takes a dive at the very end of cutting. The attached images show what I get, if it were not for the break away torch mount I may have broke my torch. 
 
 
 
 

I assume you mean qtplasmac?

Go to parameters and increase your (can't remember exact term) void anti dive. This should help but if not try decreasing (or increasing) your p term in your pid.

My guess is it's a pid issue (or you are cutting too slow)

Takes a bit to get it tuned perfect but when it is.....it's sweet. I can cut for hours on a 8x4ft sheet and never touch the torch to the table

Are those cuts supposed to end like that?
I do not think so, so :
-check the THCAD division jumper is seated properly and does not move
-try to catch the screen when that happens, does the voltage rise a lot?
-tighten the screw terminals on both sides of THCAD cable, just in case
-check the high voltage cables from plasma to THCAD
And, do you get latency warnings when running LinuxCNC?

Here is what I have done, changed the speed PID from 10 to 5, PID-I form 0 to 3 and PID-D from 0 to 3. No real change it cuts some part and fails on others. Jumper on the board is on 32 and unipolar.
Tried to use a Cut Rule in SheetCam to drop the THC about 0.250" before end but QTPlasmaC did not like the G-code see attached.  I do get a latency error but not when doing any operations, I get it when editing the G-code in QTPlasmaC randomly.  Attached a screenshot of running the Latency check test.  the voltage does not seem to bounce around that much, the Z axis seems to move maybe0.030" up and down until it crashes. 



 

A dive towards the work would indicate that the voltage was read as high
It it possible that you have lost the arc and it is not signaled to LinuxCNC?

A halscope plot of the voltage, arc-ok, and velocity when you have the issue would help a lot

I will see what I can do. Not a real expert on LinuxCNC