Head scratching about bed flatness

[coredump]

So, I have this CoreXY with a bed made of MIC6 plate + steel plate + PEI, and I use the 3 independent Z screws to do bed compensation/leveling. Even so, I get horrible first layers some times. So I started using the mesh compensation and I get this every time I probe:

(I am using a lot of points on that one to give more detailed visuals). The thing is, I put a straight edge from the red area to the blue area, and tried to fit a 0.03 feeler gauge under it, and I can't find ANY gap, what kinda means to me that I am doing something wrong.

Everything is probed using a Precision Piezo probe, and I tested repeated probing and I get consistently:

G32 bed probe heights: -0.045 -0.049 -0.043 -0.046 -0.044 -0.049 -0.043 -0.045 -0.047 -0.046 -0.045 -0.047, mean -0.046, deviation from mean 0.002

So, now I am kinda lost on what to check? Maybe something mechanic on my build?

This is the bed, the lower right spot with filament squished is where I prime the struder at 15,15, so you can orient that with the probed mesh.

Also, this is my current gcode: https://gist.github.com/coredump/b0e76e62d348a3373ffaa67fd2f31052

Am I overlooking something?

[dc42]

If the bed really is flat then the main cause of those height changes is probably twist in the gantry.

[coredump]

@dc42 shouldn't that be compensated by the bed auto compensation thing with the independent z motors?

I read that as bed twist, not gantry twist. How one proceeds to try and identify/fix that?

[dc42]

It could be bed twist, but I think it's likely to be gantry twist. Either way, the 4 corners of the gantry are not all at exactly the same height above the corresponding corner of the bed.

[chapelhill]

What cables and filament supply do you have to the extruder mount as varying tension and twist will affect z height depending on xy position. I have nimble cable which can give a similar effect. It look also like one of the y rods might not be parallel.

[deckingman]

Putting a straight edge between the red and the blue and finding no gap means only that the bed is flat along that line. What happens if you put the straight edge across the front from front left to front right. Do you see the gap in the blue area? Either way, the height map shows that either the bed is twisted or something is amiss with the gantry meaning that the distance from the probe to the bed is inconsistent across the entire surface. You can either fix it mechanically, or use mesh compensation. Does the mesh compensation fix your first layer problem? What exactly do you mean by "I get horrible first layers some times"?

[coredump]

@chapelhill I am using a nimble, yes, with their supplied 95 cm cable. My filament is feed from a shelf up the printer, so I don't think that can put any spin on it.

I've been playing/testing different belt tensions on the d-bot carriage to see if there's something that I can do there. Also a dude on the 3d printing discord chat suggested using a bullseye level to check the gantry and I will do that later today.

@deckingman indeed the straight edge on the front helped, the bed is indeed sagging in the middle. I also found almost half a millimeter drop measuring the left side front to bottom.

Somehow, I was able to bend this MIC6 plate.

[chapelhill]

The nimble puts differing tension on the carriage depending on its curvature (forwards and backwards) and it's twist (left to right as cable leans over). I have now mounted nimble motor with 3d mount but that can also create problems. Suggest you disconnect nimble from extruder carriage and try a leveling run again to see what difference it makes.

Which carriage are you using as I have problems with my piezo reliability and cannot find satisfactory solution.

[coredump]

@chapelhill I am using this https://www.thingiverse.com/thing:2206664 + a thing I made https://www.thingiverse.com/thing:2868944

It is pretty accurate/repeatable.