In my case, the problem went away at the “Fast” setting. At the Fast setting, the test cube was pretty much dead on 20mm. When I slowed down for high-quality prints, my models that wee supposed to be 8mm thick were more like 7.5mm thick, which is a double-digit percentage error.
What fixed it for me was a combination of several things:
First, I recalibrated the extruder. It was massively underextruding, which is probably a sign that there’s something wrong with that replacement hot end. I had to do this twice, because the first time, I missed the step where you save the changes permanently, so I thought that it didn’t make any difference.
Second, I removed and reseated the hot end, just in case it wasn’t all the way in.
Third, I fixed an assembly mistake. I had attached the platform’s linear modules using the wrong screw holes, which made the cable end of the bed a fraction of a millimeter higher because it wasn’t inside the indentation. When it was misassembled, the bed leveling started with the tip off the edge of the bed, which I assumed was by design, but apparently it wasn’t, which meant the bed height estimates were probably way off.
I probably should have realized something was wrong when it was knocking the print sheet loose back in my earliest tests, but I just thought I had leveled the bed wrong, and it went away after re-leveling a few times, so I ignored it. But what clued me in was that when all this weird misbehavior started happening recently, I also started seeing scuffing starting from the leading edge of the bed that extended up to a third of the way across the print sheet and went away, which made me suspect that the bed wasn’t level.
In the meantime, the stop switch on the horizontal axis randomly got stuck on me, causing awful noises. That problem went away on its own after about three failures, so I can only assume it was temporary. If it comes back, I might need a new linear module (and they can only be bought in packs of 5?) to replace that one at some point, but it is working for now, so I’m not going to worry about it. Maybe I can just replace the switch if that happens.
Fourth, I did my last two prints without the painter’s tape (which I had just started using). It wasn’t sticking to the bed well enough, and was getting air bubbles during bed heading and peeling at the sides and catching on the print head and peeling off sometimes, and so on. I’m going to switch to masking tape at some point. Either way, that may have caused the print to pull up from the bed a bit, which would have exacerbated the height problems from the massive underextrusion.
Of course, the reason for the tape was to protect the bed from prints that stuck so well that they ripped off the top of the previous bed. I’m still wondering if there’s something wrong with the vertical modules, like slipping sometimes while moving the hardware up, but I suppose it could have been caused by inadequate software correction for the significantly un-level bed.
Fifth, I rewired the cables to fix a problem where moving the assembly up and down was fighting against some twists in the cables, which might have caused some sort of slippage, and thus might also have been the cause of the total destruction of the first print sheet.
Sixth, I recalibrated the bed using five points instead of three.
What’s odd is that everything seemed to be working fine, and then suddenly it started gouging the bed and printing things that stuck so badly that removing them peeled off large parts of the print sheet surface, with no changes in calibration, no moving of the hardware, etc., and that the new hot end behaved so radically differently with no changes other than adjusting the bed height.
Snapmaker should probably add some safety checks in their firmware to ensure that
- Ridiculously off-level beds are detected and throw major warnings in the UI.
- The sensor detecting the board edge being off by two inches throws a major warning in the UI.
- The tip doesn’t drag if someone mis-assembles it in that way.
- The UI warns you if an end switch starts sticking, just in case you weren’t around to hear the growl.
All of these things should be relatively easy fixes that would make the product a lot more robust against new user assembly mistakes and random malfunctions.