4 point bed levelling question



  • I have a large glass bed which I know is flat as it's come off another printer.
    So, I only need to correct for skew in the z axis (as there isn't any in-built adjustment).
    I am trying just to corner mesh 4 points and then have this used as the tilt correction.
    Using G29 with a defined grid which dimensions this out correctly and I can measure roughly 4 corners ( as per picture).

    I am using an IR probe, which measures differently for different thicknesses of glue etc, so this is the way I want to level so I only have to worry about 4 key corner points and the center when worrying about the coating and differences reading the points etc.

    The issue I am having is that even if I G30 at the bed center and remove all existing calibrations and offsets before doing the G29, when I do the G29 the resulting graph never dissects the center point ( which it should right if this is the z=0 offset that I am starting from? )

    I get this graph ( which does match the skew and is actually less than 1mm across corners so not too bad).
    But it never dissects the measured Z=0 point in the center, so it looks like it will compensate wrongly for the center of the bed?
    Am I missing something obvious ?
    I have repeated this many times and it never hits the center point z=0 calibration G30 I am doing before I run the G29.

    0_1561051333593_88175476-ab79-4d97-8f7c-fb8d676edfd6-image.png


  • administrators

    It looks like it has treated the back left point as home.

    • How are you homing Z ?
    • Which firmware version are you running?
    • Please post your config.g, homez.g and homeall.g files.

    If your bed has levelling screws, see https://duet3d.dozuki.com/Wiki/Using_the_manual_bed_levelling_assistant.



  • Hi,
    It doesn't have levelling screws, it has rails and belts on all four corners. The bed is fixed and if I notch it up the z-sliders, the adjustment is 2mm so it just goes to far the other way. ( Bed moves down from 0 at the top, print head is fixed on the z plane ).

    • M3D Promega.

    For homing, I am/have been using IR probe at centre to set z=0, using G30 single probe.

    • the actual probing works well, I have black card behind flat borosilicate glass (4mm).
      I can repeatedly home at X190 Y160 ( probe at roughly bed centre ). The probe offset is correct and I can move to this position, set Z0.1 and just slide 0.1mm paper underneath etc.....this survives a restart as well and works with home Z and home all, so I am happy this is correct at bed centre
    • just to be clear, I can print ok at the centre of the bed as well if I don't run the G29.

    I edited the config files to zero out any homing by switch at the far end and I just reset z=0 using the IR probe and G30 at bed centre.

    Printer is using duet maestro.
    Running 2.02 release.

    I did try just doing 5 point compensation using G32 and bed.g - but got told it was being deprecated and I should use G29. Should it work this way though, because it doesn't seem to be including the 5th centre point (Z=0) in the calculation like the G32 and bed.g would.

    Everything works fine and prints well, But I am just trying to avoid the need to babystep so much when I have too much glue or different coatings on the bed by controlling the 4/5 main measured points only. As I know the bed is pretty good.

    If you still need configs, I will have to attach tonight, sorry IT problems connecting back home.


  • administrators

    Have you tried using more points on the mesh? For example, you could halve the spacing in M557 to use a 9-point mesh, which would include the centre point.



  • 0_1561122741159_3c65efcb-39f4-4a9c-96bb-5078057b7300-image.png

    Halving the mesh worked better - and shows that the bed center is at z=0 when measured.

    But it now starts adjusting for the slightly different layers of glue etc in those different places.

    • I know I could bed mesh with it totaly clean, but I can't do this every time I turn on the printer, so I am trying to use this as the 'calibration' when turning on.

    The printer design is not tolerant of room heat changes or stretching of belts etc on the z axis, so I am trying to do this regularly at print start / power-on to counter any 'skew' changes, but I already know the bed itself is flat so trying to avoid the errors introduced because my gluestick coating is uneven.

    The maths involved are probably beyond me, but I wonder why when only 4 points are done it doesn't calculate it through the z=0 at the center?



  • Have you tried doing a single detailed probe run with the bed clean and then just loading it when printing rather that rerunning it before each print?



  • @phaedrux said in 4 point bed levelling question:

    Have you tried doing a single detailed probe run with the bed clean and then just loading it when printing rather that rerunning it before each print?

    Hi Phaedrux, Yes this works fine and I have been doing this for some time and it lasts for a fair few prints, but this is a slightly different use case.
    I have learnt over time that the printer is not repeatable over long periods.
    What I mean by that is that the corners can end up in very slightly different places due to factors like heat, stretch of belts, even a little too much pressure removing a print etc, which then throws the bed mesh out very slightly because the corners are tilted differently.

    • So , think of this as a measurement at each corner so I know what angle the bed is at today.
      Actually the bed is near enough perfectly flat - lucky maybe but it doesn't need a mesh to level it out when printing, - I am just trying to account for the un-guessable tilt changes when moving the really heavy bed 370mm up and down frequently.


  • How thick is the glue that it's adjusting for it that much?



  • With the IR probe, depending on age or last cleaning, 1 /2 /3 layers of elmers glue stick randomly. But different thickness in different places over the bed as I am not exactly an artist, it comes off with print etc.
    Seems to make a big enough difference to mean first layer is either too low or too high and needs baby stepping by up to 0.1-0.12mm sometimes.


Log in to reply