Another Bed Leveling Weirdness

  • Hi all
    I got a "funny" situation with an Ender 3 Pro and a BL-Touch running a Duet Ethernet.
    The BL-Touch seems to work fine for homing the axes.
    However, the bed leveling behaves very odd...

    Here's a bed level chart across the whole bed (M557 X41:221 Y10:220 S30):
    0_1557156860878_bed level 01.png

    As you can see, one point is completely off (the bed would have to have a severe and visible dip to justify that picture, but it doesn't). I did manual level the bed before the mesh calibration, so I'm very confident, that the 4 corner points are not off by more than 0.1mm or thereabouts.

    I repeated the same calibration a second time with the same result, so it wasn't a temporal fluke.
    Anything I can do to troubleshoot?

  • Moderator

    What are you using for your M558 and G31 commands in config.g?t it's odd that it happened across two probing runs. The BlTouch is susceptible to magnetic interference and the Ender 3 Pro does have a magnetic bed. I'm not exactly sure of the arrangement, but if the probe lines up with an actua magnet at that point it could be the cause.

    You can improve probe reliability by slightly modifying your M558 command.

    Try M558 P9 H5 F100 T6000 A10 R0.5 S0.003 B1

    If you get the same results with that command, try changing the grid size and spacing slightly to see if you can avoid that point.

    You also have the option of manually editing the resulting heightmap.csv to replace the obviously wrong values.

  • @Phaedrux Thx for the suggestion!

    Here's the current config:
    M558 P9 H5 F500 T6000 R0.2 A5 B1
    G31 P25 X41 Y0 Z2.200

    I've modified M558 according to your suggestion: M558 P9 H5 F100 T6000 R0.5 A10 B1 S0.003

    Here's the new result:
    0_1557160173242_bed level 02.png

    It looks pretty much the same... Now I totally understand, that the bed might warp up toward the back left corner (although my manual leveling is definitely not off by 0.6mm), but that one point at -0.281mm is complete bogus. What could make the BL-Touch measure just one point completely wrong and all others correctly?

  • Now, just for fun, I activated that mesh compensation (with the corner shown at +0.64) and it makes the nozzle be about 0.64mm off the bed, so the bed is pretty level, but the mesh compensation with that rising corner is actually complete bogus...

    I'm at a loss why mesh grid compensation seems to measure such crap?

  • Moderator

    Are there any wires pulling or interfering when the probe is in that area?

  • Nope, checked that, it's actually the corner closest to the wire mount, so with the most slack. No taut wires anywhere...

  • If this is an Ender 3 I would make sure the eccentric nut the hotend has is tight enough. If it's loose the entire hotend assembly/fan/bltouch could move and rock when the wiring and ptfe tube bind up. Also make sure the eccentric nuts on the bed are also tight enough. Gotta make sure the wheels have proper tension on a v-slot type printer.

  • @jamesm Thx for the tip, however the printer is tuned pretty well and has no play in any of the axes.

    Repeatability is excellent (so no slipping or skipping of any steps) and mechanically, everything is as good as it can be.

    I've successfully printed large parts with the printer with very nice first layers, it's only once I introduced the BL-Touch, that I run into this weird problem.

    The printer still works very fine when not doing a mesh bed leveling and just using it with manual corner leveling.

  • Moderator

    I think it may be due to the magnets. Can you try placing something like a thick book on the bed at that corner and probing it to see if the resulting map is now flat or if it still has a weird spike and dip. The book would add distance between the probe and the underlying magnets.

  • @phaedrux Hm... that actually seems to have had a big influence... the level is obviously horrible (books aren't really flat), but there are no outliers.

    Now, the only question is: how do I make the probe less prone to fubar from the magnetic bed? I can't be the only one having issues?

  • @remopini as stated in the second post, you can set a different spacing (more or less points) for your mesh and avoid the spot all together. This will probably be a bit of trail and error.

  • Moderator

    As @Twigz says, you'll have to jiggle your grid around. You;re not the only one encountering this. It's a side effect of the current popularity of magnetic bed surfaces.

    What version of BLTouch are you using? Does it have the metal, or plastic pin?

    What BLTouch mount are you using? Can you put it in a different location?

  • @Twigz @Phaedrux I got a 2.0 (platic pin). The mount is one of those Titan Aero all-in-one mounts, so can't really change the position (I'd have to design my own mount for that). But I can of course change the mesh (or shift it around by some factor). I'll do some trial and error tomorrow to figure out, whether I can find a setup that works. Will update here with any findings.

  • Moderator

    I'll be in a similar situation soon with the Ender 3 and BLTouch. So i'm interested in seeing what you find.

  • @remopini you can also try to turn the bltouch 180 degrees. I had mine too close to the hotend and it acted up. Turning it around meant the board was further away from the hotend and it worked flawless.
    Maybe the all metal aero has some influence?

  • @Twigz The PCB is already turned the "right way", i.e as far away from the hotend as it can (with the mount I have):

    Electrically or magnetically, I can't imagine the hotend influencing the BLTouch too much in that setup (the closest active component is more than 40mm from the BLTouch).

  • Ok, twiddled a little bit with the mesh definition. Moving everything over a few mm did the trick with the one outlier:
    0_1557211516059_bed level 03.png

    However, the systematic measurement error (that back left corner is NOT 0.48 higher than the other corners) persists.

    With mesh compensation enabled, the nozzle is now basically half a mm higher up in that corner than in all others. So, for whatever reason, there is still a significant measurement error introduced in that corner.

    Current ideas:

    • Since that is the corner where all the cables are, could it be interference? But the very regular error kind of suggests otherwise (interference wouldn't usually be regular)
    • Some mechanical problem that only manifests during probing, but not during normal moves. Also not super likely.
    • The bed magnetism is influencing the measurement. Possible, but nothing I could do anything against, so I'd be limited to use the BLTouch for the initial probe in the front and not do any mesh compensation.
    • Something other I can't think of right now...

  • you could try building this.

    due to the optical sensor it is less prone to interference.

  • @Veti I kind of want to try to get the $50 thing to work first before giving up and going with something completely different. BTW: I wouldn't go with an opto-switch solution, they are rather imprecise.

    As a last resort, I can always try to shield the BLTouch somehow, but finding thin ferromagnetic foil is probably difficult.

  • @remopini said in Another Bed Leveling Weirdness:

    I wouldn't go with an opto-switch solution, they are rather imprecise.

    i get a standard deviation of 0.002. I would not call that imprecise.

  • @veti Not based on my experience, but it's great if you get that kind of performance out of it 🙂

Log in to reply