It started with a first layer that looked a bit off. The printer's first layer was uneven, and I figured I would just re-level the bed. Reasonable. Contained. Ten minutes.
Then I noticed the bed was not flat, so I started mapping it. Bed mesh needs a probe. The probe's offset was off, so I calibrated it. To calibrate it I needed the firmware config in version control, which was a mess, so I rewrote it into a clean config with comments. Then I realised the stepper current was set from an old guess, so I measured it. Then the cable chain was rubbing, so I printed a better chain. On the printer. While it was misaligned.
Result after three days: the first layer is now a perfect, mirror-flat sheet, and the printer is better in 14 ways nobody asked for. The thing I actually wanted to print was a 20-minute part. I have not printed it.
What I learned, for the record:
- A bed mesh compensates for a warped bed, but it cannot fix a bed that is loose or a gantry that is out of square. Fix mechanics first, software second.
- Heat the bed to printing temperature before probing, because the bed warps as it warms.
- Keep the config in git. Every change is a commit with a one-line reason, and when something gets worse, you can bisect it like a program.