Rods "Spaceship" Scratch built Plasma Cutter build

Here is a video of the machine while cutting 8mm mild steel with these settings. Have a look at the movement of the Z axis (at the stepper motor shaft joiner) and tell me if you think this is too busy as I don't really have a reference. The THC is enabled 1.5 seconds after ArcOK. Open the video full screen and you'll get a good view



This is still the original external offsets plasma config. To me it looks like it needs a way to slow the response down. Having said that, a bit more acelleration would help but thats not really possible under the current laws of physics

Anyway, I think this is the best cut I've ever got with the hpid.comp from external offsets.

I will say the arc voltage signal is a lot cleaner than my Everlast. and we are cutting 3 times as fast. Up around 1800 mm/min so this 150mm cut does not take long!

I did try my separate slower thread idea but it was a fail....

This video is a bit longer as its a 1090mm long cut to tidy up my sheet of steel after all my test cuts.

Hi Rod,

To me the z axis responds nervous cq too fast up and down. It can react 3 times slower. My control way is to step 0.1mm up or down at a time. This i call resolution.

Grotius wrote:

Hi Rod,

To me the z axis responds nervous cq too fast up and down. It can react 3 times slower. My control way is to step 0.1mm up or down at a time. This i call resolution.

Grotius, thanks, thats what I thought. I think that the weakness of the eoffset-pid component. There is no way to tune down the applied offset in its current form as it takes the error from the desired position and multiplies it by a scale to determine the number of counts to apply (say the scale is 0.1mm per count). So if you change the scale to 0.001mm, it will just add more counts to account for the error. I think thats why people have had more luck with the external pid as there are ways to apply a limit. I think the adjustment is always more than what s required.

I have an idea but I don't know how pure it is. That is to modify the component so it has a scale pin that can be >0.0 and < 1.0 so we can turn down the adjustment. eg. allow apply an adjustment x 0.8 or something similar.

andypugh wrote:

What happens if you just turn down the gains?

Andy, I've tried all sorts of things with the P,I,D gains and the current result is by far the best I've got ever got and very close to what I had set for my old machine but much improved. But yes I did try lower gains today and bracketed the settings I settled on with increasingly smaller steps on all of them.

If you look at what skunkworks did with V 2.7 and a PID and offset component, he added a limit3 in there to be able to apply some limits after the PID. I was going to try replicating his work on master branch without Dewey's component and use an external offset to apply the offset and see what happens but I keep coming up with ideas to try on my current config. The cutting here is from a pure Axis EO sim by Dewey with addition of a torch voltage sampling routine so we set our commanded voltage on the actual voltage and a 1.5 second timedelay to enable the THC.

There is a theory that says applying a position offset to a velocity stepgen on the mesa is a stumbling block but I can't comment on that.

I think though as it stands, I can use the machine!

andypugh wrote:

If you have a PID based on arc voltage moving the Z axis then I would be running than axis in position mode, I think.
And I would expect the arc-controlled PID to consist mainly of I-gain. I would start tuning it with P and D at zero and tune only I for the level of responsiveness I wanted.

(NB: I have never even seen a plasma cutter)

Thats a bit of a conundrum (not your knowledge of the subject) but Dewey's docs said

When used to regulate an axis offset, the offset_pid
component servo loop is an outer loop to the LinuxCNC
motion control. For external offsets, this loop is implemented by
a simple single-axis trajectory planner with constrained
maximum velocity and maximum acceleration that are fractions
of the limits available for the axis as set by the ini setting:
[AXIS_L]OFFSET_AV_RATIO. The inner loop effectively contains
an integrator and use of the pin is not expected
to be helpful; some amount of may be required
for stability.

We do cheat with plasma and give eoffsets 100% of the velocity and acceleration. If anything on today's tests more acceleration would help but we ran out!

I do value your input Andy. Fresh thinking is always appreciated.

I might add that my regression analysis of 9000 data points collected with halsampler showed a very close linear correlation between the voltage and torch height (with 99.5% confidence). I might look at the different stepgen modes.

Hi Rod,

At the time i wrote my component based on new version external offset branche and reverse feed branche, i could set the z axis eo speed value. This example lives in the opa config of the Eo branche i think. My current value for thc z axis correction speed is 200 mm/min. If you can not find it. I will show it.