THC experimental

We have been using the Mesa THCAD and JT's THC for a while now and the combination works great.
Occasionally the question of using it on corrugated or dramatically bent work comes up.

If you keep your correction factor small all is well on relatively flat work, but the torch height control may react too slowly for rapid changes in height.
If you then increase the correction factor too large the torch will over correct and start oscillating up and down across the cut.
This seems to be because the THC is calling for more and more corrections before the z axis can react and reach it's new commanded position.
So there can be a constant overshoot up and down across the cut.

Here is an experimental edited version of your THC comp.
The hope is to allow larger corrections without oscillating up and down for say cutting corrugated steel.

Currently
THC voltage => lowpass.0 => THCR.comp => lowpass.1 =>
The math is in long form to try to make it easier to read.

The major changes...
Do not make a correction until the Z axis has gotten to within the Z tolerance of the last correction position it was sent to with a possible delay period.

The size of the correction is proportional to the difference in voltage up to the maximum voltage change.
So if the maximum voltage change setting is 5 volts and the z correction setting is .01".
A voltage difference of 2.5 volts would mean a correction of .005".
5 / 2.5 = 2
.01 / 2 = .005
The correction is then averaged with the last correction and sent out.

Several pins have been added to bring out to the panel for testing.
Also the THC correction factor can be brought out to the panel so can be set low for flat work and higher when needed
There is some redundancy in that the original concept was to have the THC self contained, but there seems to be a benefit to using the lowpass.
We have been swamped at work and so have not been in the shop for almost a month now, but plan to clean it up when time allows.

For some testing we cut out 4" x 3" rectangles in light gauge steel.
We then bent them and cut .4" x 3" rectangles.
The bend is roughly 1" change in height over 1 1/2" of Y travel.

Awesome, thanks Rick, a few of us have been waiting for you to post this. (but it will still be a while before I can turn a torch on). I'm going to review in detail.

Morning John,

Thee idea of the z delay was to wait so many cycles between the last correction move and the next voltage reading to give the voltage a chance to stabilize before making the next correction. I just set it to 0 for now. In the little bit of testing I have done it has not proven beneficial to set it higher yet. It just seems to make things a bit lazy.

Rick G

Tom,

Thanks for testing.

If the commanded correction is too great it can still overshoot, Say there is a 2 volt change and a 2 volt change equals .005" and the commanded correction is .010" it will overshoot, but only by one commanded position not possibly several. So if it does hop they should be small hops. Sort of like a narrow band O2 sensor searching for the right mix.
Of course the goal was no hops.
I set up Hal Scope and monitored Z up and Z down to try and find the best settings.
The z axis seems to move smother with a low Z correction which is why I think the Z correction should be brought out to the panel and set low for a smooth quality cut on flat work and higher for bent work so it can follow the contour.
There are several settings to adjust.
Besides the voltage tolerance and the Z correction amount there is "Voltage difference for full correction" which is used to calculate the proportional size of the z correction based on the voltage difference. The low pass gain can also be adjusted.

Rick G

Rick, I've been slaving away at implementing the current THC.comp in master and after implementing corner lock in my GUI, I found that the pin is never used in the current code.

github.com/LinuxCNC/linuxcnc/blob/master.../components/thc.comp

I see your code does use the vel-tol pin. I wonder if there has been a regression in the current master code?

Looks like I'm going to learn how to compile HAL components now...

rodw,

There seems to be a few versions of the THC out there. That one uses a parameter that you can set / change in Hal.
If you are working on your panel you might want to bring the THC correction amount out to the panel also as Tommylight has mentioned.

Rick G