jes notes Index Gallery . Clavichords Cyclecar Dehumidifier Financial Data FreeCAD MSVA Philosophy Signed Distance Functions Shaft passers Snap issues

2026-11-12

Last modified: 2026-11-12 22:46:22

< 2026-11-11 2026-11-13 >

Plasma cutter

I've bolted it down to the bench, made a slat table, and filled the water tray with water.

The slats are slightly diagonal so that (very common) axis-aligned lines can't line up with the slats. There are screws in each corner for leveling the slats to the machine.

Also new consumables arrived for the torch so I've replaced them.

I tried to cut out a heart shape at 1000 mm/min and it seems like the plasma jet is coming out straighter than with the old consumables, so good on that front, although it didn't cut all the way through. I moved the torch down a bit closer (was ~3mm before which is on the high side), and then it seemed to reset the ESP32 when trying to light the arc! Nooo! What?!

And I tried probably 20 times to get this to work and it reset the ESP32 every time.

I tried moving the torch back up higher and still resets every time.

So what has gone wrong here? I can't imagine what could have changed? And why did it work exactly once today and then not again?

It works fine with the torch disabled so definitely something to do with lighting the arc.

After the first cut finished the post-flow of air was blowing lots of water around, maybe some water got in the nozzle and that is the problem?

The new big water tray may be part of the problem, maybe reflecting RF interference or something?

Some things to check:

OK:

I left it for a good probably 7 hours and tried it again, it did not start working automatically.

I looked inside the nozzle and all looked well.

Using the manual torch trigger it works fine, doesn't seem to cause any problems.

I shifted the water tray out of the way a bit and made no difference, but didn't manage to fully remove it because it catches on the torch.

I then tried holding a piece of steel plate in free space in between the torch cable and the control box and it worked first time! But unable to repeat.

So we have:

Note I don't actually know that the controller is resetting, I just know that the gcode sender reports having become disconnected and motion stops and the torch relay switches off. This is all consistent with the ESP32 resetting but not definitely proven.

I did have the aluminium foil shielding all around the torch cable on all my previous successful uses of the machine, but I took that off as part of moving it to the new location because I thought it wasn't needed. Probably I want to put that back on and see if it helps, and if so buy some proper shielding sheath or whatever.

But I am confused as to why it seems fine if I squeeze the trigger manually. Maybe I just need to use shielded cable for the torch trigger output? I almost did that initially and I do have some but thought I'd test it with the cheaper cable just to see if it worked and it did.

So how about:

But why did it work the very first time I tried it? And why did it work that one time with steel plate in free space? Would love to have an explanation for these. It's extremely suspicious to me that it worked specifically the first time. If it worked 1 time out of 50 but it was a random one I wouldn't mind, but why the first? I guess I can accept it working the first time I held up the steel plate. If it was just working at random and I tried 1 new thing each time then of course it started working the first time I tried that new thing and didn't work subsequently, that's fine. But why did it work the first time in the new place? Just random chance??

EMI is incredibly frustrating to debug.

Some things to try are:

Apparently $System/Stats may be informative for working out whether it is rebooting.

OK, new findings:

Even if I hold the two wires right next to the screw terminals when I touch them together, touching them together (thereby lighting the arc) doesn't cause any problem.

So the problem is if the arc is lit by the CNC-operated torch trigger cable while at least one of the terminals is connected to the relay board. WTF?

But this is if anything confirming the suspicion that the interference is coming in via the trigger cable. So I think we are still looking at shielding the trigger cable and probably the torch cable.

ChatGPT suggests the interference is coupling into the control electronics through the inside of the relay, quite funny.

The relay contacts are galvanically isolated, but they have parasitic capacitance to the coil, PCB tracks and surrounding circuitry.

ChatGPT suggests a ferrite on the trigger cable as close to the relay board as possible. However we know that ferrites do nothing and are a scam so I am inclined to skip this.

ChatGPT ideas include:

Also, as a diagnostic: do the test where we leave one wire connected to the relay but tap the other wire on the screw terminal, but this time disconnect the relay from the ESP32. If it now doesn't reset then we have confirmed that the interference is going through the relay into the ESP32. (But haven't we pretty much confirmed that anyway by noting that disconnecting the wire makes it not reset?).

So next thing for me is: make a new trigger cable using shielded cable.

And we'd expect that to either completely solve or at least improve the interference problem.

And if more is required then shield the torch cable.

And if we do both of those things and still no good then ??? why did it work previously?

< 2026-11-11 2026-11-13 >