Bed mesh problems

[real kite]

Hello, since today started to have issues with bed mesh. Tried to do cold one hot one, results are same. Checked probe accuracy results more or less same with standard deviation 0.002. Then printing first layer it is visible that bed mesh is not right, on the side with higher points of bed mesh there is not enough of squish, and with lower - there is too much squish. Maybe someone had something similar and could give me the direction where to search for problems? I think it started after last updates.

[lunar vault]

This might sound silly, but are you doing a QGL before making a bed mesh?

[real kite]

yes

[lunar vault]

What kind of probe are you using?

[real kite]

Inductive, but if that would be probe problem, shouldn't mesh be different every time?

Also I did probe accuracy tests, it looks like it works fine

[rustic crest]

what's the highest and lowest number in mesh? Can you turn on scaling on that graph or show right side of screen with mesh?

[real kite]

Okay, after running TEST_PROBE_ACCURACY it came back to normal... Don't understand how that could be related

did few bed meshes before all were same and now they are okey without any changes anywhere

[real kite]

Okay it looks like there is still something with bed mesh after KAMP. Bed mesh looks ok without it, but after KAMP it becomes something like that and thats why my first layer does not print properly...

Range between highest point and lowest is 0.096

[real kite]

and doing bigger bed mesh at same places variance is 0.017

[real kite]

okay with less probing points it looks better, so it is deffinetily realated to adaptive meshing

[lunar vault]

Do you have any scrap plastic stuck under your bed or something? All KAMP does is shrink the mesh bounds, it doesn’t adjust anything outside of what you have defined in your config already.

Inductive probes are super prone to thermal drift too which can be very hard to diagnose and fix. Maybe you should try a klicky probe and compare? Or maybe go with tap?

Even using an inductive probe with a textured sheet can give weird readings because of all the little peaks and valleys in the texture of the sheet.

.09 of variance is less than half of a single layer. I think the mesh just looks scary because your scaling is high in the mesh visualizer.

[errant crater]

I had a similar problem... I just recalibrate the mesh and it came right. Had locked in a previously stored mesh and would not overifgmmght it for some reason.

[real kite]

There is nothing under bed. I have euclid probe, so will print some parts and check, how it will work with it. But yes, having less probing points made variance better, also doing adaptive meshing in other bed area works normal too, so it looks like there is something specific with those points.

[real kite]

I do not store any bed mesh, I do new one every time

[brave kettle]

Well I don't think that really solves the problem
Although I kinda hope klipper can probe several points with random offsets to combat a bad spot on the bed