I'm having a problem getting a clean first layer on an Ender 3 with BLTouch auto bed leveling. Thickness seems to fluctuate all over the bed, but in a consistent (repeatable) way. Here's my attempt to print a single layer 5 square bed calibration test:
I printed this several (many) times with slight settings tweaks and it looks pretty much the same every time; the ups and downs aren't random. For example, the center square always is always too low on the left and too high on the right:
The printer is a SainSmart Ender 3 Pro with a BLTouch V3.1 and Creality glass bed, otherwise stock. I flashed a bootloader and Marlin 2.0 using the instructions and pre-compiled firmware from 3D Printing Canada. I'm using the glass bed upside-down on the plain glass side (no coating).
I pre-heated and leveled the four corners manually using the paper method. I auto-homed and then lowered the hot end until it would just catch a piece of paper and used that height to set the Z offset using M851 and saved it with M500. It's currently set at -2.80.
G29 to GCode start in Cura, and it does a 3x3 probe before the print starts. Here's the output when I run
Bilinear Leveling Grid: 0 1 2 0 -0.207 +0.172 +0.162 1 -0.100 -0.160 +0.220 2 -0.118 +0.215 +0.295
Here's what it looks like in the Bed Visualizer plug-in in Octoprint:
If I understand this right (dubious) it's showing that the glass is lower toward the front and left, highest at back-right. But it's only 0.4mm from the lowest to highest points. And the whole point of mesh leveling is to compensate for this anyway.
At Paulster's suggestion I turned off mesh leveling using
M420 S0, leveled manually, and printed again. The result is pretty similar (note that this time I let it run all the way through):
Where should I start looking to diagnose this problem?
I noticed my X-axis belt was a bit loose, so I tightened it up. It seems to have helped with the odd Z slanting. My test print is still not great though, so this may not be the whole problem. Also I've never seen this effect listed as one caused by loose belts, so it's dubious as the cause. Here's the current test print after tightening the belt:
It's flatter, but I'm still getting (I think) under-extrusion and some odd wobbles at the corners.