Bed_Mesh on Delta Printers

Printer: Anycubic Predator
Board: BTT SKR1.4 Turbo
.log: klippy.log (2.1 MB)

Hi,
I’m still struggling with my printer, getting it working with Klipper. Currently, I’m stuck to get a working Bed_Mesh and noticed something. With a delta printer, it probes a round bed, but the bed mesh shows me a rectangular bed, with probe points which were never probed, and are outside the bed. And one of those probes is a weird spike with I think is messing with the actually taken probe point values.

As you can see the probe point, shown in the picture, were never taken, and the coordinates are outside the print bed.
I changed my effector and rod connectors to make sure that my self printed ones are not at fault, causing effector tilt or something. I also installed springs onto my print bed to adjust it heights, so everywhere the nozzle has the same distance to the bed. I tried various Bed_Mesh settings, but the bed mesh always have this shape, and this spike.

Has anyone idea what causes it, and how to solve it?

First thing is not to define a relative_reference_index.
How well round beds are supported in the visualization of the webif, I cannot tell.

I tried with and without relative_reference_index, both gave me the same results.

Well, look at the raw values of the probing in the log:

  • Min: 50.202511
  • Max: 50.739744

This is an absolute deviation of 0.537 mm, which is quite significant and could match to what you are seeing in the visualization.

I know, but what the bed mesh visually shows me is that the most of this deviation is coming from the spike shown in the screenshot, which is a point that never been probed. I can’t even reach this point with my tool head. But doesn’t the bed_mesh uses the probe points values to create a bed_mesh.
A Spike like that could “pull” up my mesh on that area, making it unusable for me.

You should go by your klippy.log. This is what Klipper has probed and what Klipper is using.
When looking at the visualization then the first noteworthy fact is that it apparently cannot deal with round beds. Maybe there are other issues as well.
This is something you better would rise at the respective UI you are using.

Ah, I see, so I should not rely too much on the visualized bed_mesh.
I can not figure out why the bed_mesh is such a mess. When using it the nozzle is all over the place, currently absolutely unusable …going to try Bed_tilt instead.

Generally the probing of a round bed looks a bit weird. Ref to Bed Mesh - Klipper documentation. I would not go by the visualization but by the quality of your first layer. If this is working, then apparently your mesh is working.

bed_mesh and bed_tilt are fundementally different:

  • bed_tilt treats your entire bed as one plane and tries to compensate for a tilt. This means each point is treated the same in relation to the tilt.
  • bed_mesh cuts the bed in little segments and tries to compensate for bumpiness / unevenness location specific. This means one point can go up, the next down.

Right now I am at work, but have some free time, digging on the internet about my issue. Found a post from 4 years ago where it turns out Klipper having no working support for bed_mesh_leveling round beds. Don’t know if it still the case today, but feels like.

Since I have an Anycubic Ultrabase glass bed, I assume by bed being very flat, so trying out bed_tilt be an option.

As I said, my nozzle is all over the place with the bed_mesh, and without my first layer is crap. I need some sort of compensation.
What I also noticed so far, the web seems full with people having same issues like me, bed_mesh not working and stuff.

Just an assumption, maybe unless your bed is already almost dead flat, bed_mesh is just unreliable. But I know that’s probably not the case.

I tried Bed_Tilt by now, and it didn’t help either.
I noticed my effector having a little tilt, wasn’t an issue under Marlin, maybe Klipper can’t handle it, trying to figure out where the tilt comes from, have to replace some stuff and test it out.