The bed.g file is needed only to do G32 bed probing, not for homing Z.

Have you tested the Z probe and it configuration as described at https://duet3d.dozuki.com/Wiki/Test_and_calibrate_the_Z_probe ?

G30 S-1 just stops the probing move when the prove is triggered and reports the height, but it doesn't set the Z position. Use plain G30, without the S-1.

Thanks, that seems to have fixed it (mostly).

The values should be close to 465 and 535. If you are getting 408 and 439 then I suspect that you have a poor connection between the +3.3V pin of the Z probe connector and the Vcc pin of the sensor.

I posted a question a few posts back, getting the answer to that will help us help you.

@jacotheron said in IR Z Probe not very repeatable, nozzle crashed again:

An unrelated thing that was bugging me (not very important) is due to my starting G-code, the first 10 layers are seen as the first layer in the interfaces (and then layers 2-10 is completed in 0 seconds). Any way to tell the firmware where the first layer is starting, so that it does not estimate so high Time Left?
The part in my starting G-code that cause this is a line where I extrude a bunch of filament while moving slowly in the Y direction, and being 1.5mm above the bed. This is to prime my nozzle, and in stead of putting it all down at the origin (like some printers/slicers do with factory settings), this leaves a simple fat line, which also clears the outside of the nozzle.
Since I use Simplify3D, the first layer is marked with "; layer 1, Z = 0.150" -> I think if it finds anything like this, it should be the a good hook to start layer counting?

Please monitor this thread https://forum.duet3d.com/topic/5288/empty-layers-causing-incorrect-layer-count/8 as your issue sounds similar and that one is already flagged for investigation.

Please note, P9 in the M558 command is only supported in firmware 1.21 and later.

Yes the trigger height of the IR sensor can be affected by scratches on the surface of the PEI. The piezo should give more accurate results.

Interesting, I've not heard of that happening before. Perhaps the black paper wasn't black enough.

The ideal would be to use black glass, but it seems to be difficult to obtain.

When using a bed material that is transparent to IR, the aim is to have as little reflection from the rear side of the bed as possible, so that the reflection from the top surface dominates. That's why I advise using a black surface underneath. If the sheet is glued down to the heated plate, then you need to paint the underside black before applying the glue or adhesive sheet, otherwise the adhesive will reflect too much.

Check the ground connection between the IR sensor and the Duet. Also check visually that the ground (centre) pin of the IR sensor looks properly soldered to the PCB.

To get accurate homing, make sure that the XY position at which you home Z coincides with one of the points in the mesh. This isn't as easy as it sounds, because when you execute G30 to home Z, probing is done at the current location; whereas when you use G29 to generate a height map, the nozzle position is adjusted to allow for the offset of the probe. So in homez.g and homeall.g, choose the XY location that you send the nozzle to before you execute G30 so that it is at the location of a probe point close to the centre of the bed minus the Z probe XY offset that you specified in the G31 command.

@ivang, perhaps you are loading a mesh using G29 S1? If you clear bed compensation by sending G29 S2 then the Z axis should not move when you move X or Y.

Oh, thank you, that's what I will do 🙂

I have been using my IR with hairspray on my glass bed for almost a year now without any issues.

I have a question: how does the output probe signal works, on Duet? It could be used to switch between sensors (using an analog switch, driven by the probe output signal)…

And how the probe input works, in analog mode? What voltage does the Duet expect to trigger? I read that a simple FSR could be used that way; I have one in my stock...

I think I was using either F300 or F600 in that video. The sample interval for analog Z probes is 1ms, so F600 (10mm/sec) would give a theoretical uncertainty of 0.01mm; except that probing speed is reduced by a factor of 3 when the sensor indicates that it is close, so the uncertainty would actually be 0.0033mm.

You may wish to use a M558 S parameter somewhat lower than the default 0.03mm, depending on the Z resolution of your printer.

The mesh that Anycubic puts down causes different heights to be selected. I t gives a result that looks like the bed is uneven. I got rid of my ultrabase as it didnt work with the ir sensor.

Cool thanks