Auto-leveler woes Just fired up cp and got it connected to my tabletop CNC

Auto-leveler woes

Just fired up cp and got it connected to my tabletop CNC (grbl 0.9 controller) and tried to auto level just to get a feel for it.

I start the autoleveler, took me a while to understand that once it makes the first contact, it re-zeros the Z axis. That was cluebar to the head #`1.

However, the system probes, picks up and moves, probes, waits an undetermined amount of time, then moves on to next probe. It can be anywhere from several sec to 25 sec. And the results are most puzzling. Sometimes it finds the surface right about where you expect it, other times it’s 70+mm down, which on a PCB mill would be a big trick.

The pauses and the deep probings don’t seem to be related.

here is a sample of the console output.

JSON serial server is latest on raspi Debian Jessie. CP in browser on Mac. OS X 10.10.5, firefox browser.

Anyone got some clue they can lend me?

Thanks!
Bob

Working on probe for {x:16,y:16} Found lowest Z:0.02500000000000213
Moving to {x:16,y:11}
Working on probe for {x:16,y:11} Found lowest Z:0.06400000000000006
Moving to {x:16,y:6}
Working on probe for {x:16,y:6} Found lowest Z:0.10999999999999943
Moving to {x:16,y:1}
Working on probe for {x:16,y:1} Found lowest Z:0.1460000000000008
Moving to {x:21,y:1}
Working on probe for {x:21,y:1} Found lowest Z:-76.67699999999999
Moving to {x:21,y:6}
Working on probe for {x:21,y:6} Found lowest Z:-71.678
Moving to {x:21,y:11}

What workspace are you using?

And do you have decent noise suppression and debouncing of the probe?

grbl workspace. Decent enough debouncing that I can use it manually without any problems.

Then it sounds slightly like the controller is breaking comms with sjps and being ‘reset’ by sjps such that it starts reporting work coordinates from a new zero from time to time. But even given that im confused why it is reporting -71mm for a touch. By then it should have destroyed your board and the mill below it… Unless somehow the reset is causing machine coords to be reported.

Check for resets by capturing the sjps logs in a directly connected set up.

Yeah, it sounds like a work coordinate system is maybe getting applied only in some cases. Could you perhaps try making sure you’re in G54 and make all offsets 0,0,0 for xyz, then re-run auto-level, just to see if that’s a possible reason why?

Dear Posterity,

Figured it out. Turned out the piece I was probing, an aluminum plate, wasn’t all that conductive. When I switched to a chunk of copper clad PCB, it works just fine.

Glad to hear you got it working. Aluminum is pretty conductive, but i’ve heard some folks jack up their probing to 12v or 24v to get better conductivity.

Yep I got probably a similar issue, and just read your suggestions. I think it has to do with surfacing aluminium oxide. As you can see on that Video, the mill gets 0.2 to 0.3 mm into material until it stops. and when it gets off, it probaply dumbing voltage up and down and before getting enought away from Touch plate.