I have lost more hours to bed levelling than to any actual modelling, slicing, or printing combined. If you have a printer, you know the ritual. You drag a folded sheet of paper under the nozzle at four corners, twiddle four thumbwheels, and convince yourself it's flat. Then the first layer goes down like a drunk drawing a square and you start again. This is a post about how I stopped doing that, mostly.
The thing nobody tells you when you start is that "level" is the wrong word. You don't want the bed level with respect to gravity. You want it parallel to the plane the nozzle moves in, at a fixed gap, everywhere. Those are very different problems, and the second one is the one that actually matters. A bed can be perfectly level on a spirit level and still give you a terrible first layer because the gantry isn't square to it, or the bed itself has a dip in the middle.
the paper trick, and why it lies
The paper method gives you a feel, not a measurement. A sheet of 80gsm paper is about 0.1mm thick, which is conveniently close to a first layer height, so when you feel "slight drag" you're roughly there. The problem is "slight drag" is not repeatable between people, between days, or even between corners if you've had a coffee. You're also only checking four points and pretending the bed is a perfect plane between them. It usually isn't.
What changed things for me was accepting that the bed is not flat and never will be, and that I should measure it rather than wish it flat.
measuring instead of guessing
The first real improvement cost nothing: a sheet of borosilicate glass off a cheap kitchen chopping board, clipped over the original aluminium bed. The aluminium had a visible crown in the middle, maybe 0.3mm, which is enormous at 0.2mm layer heights. Glass is flat to a tolerance I will never out-print, and it gives a lovely smooth bottom surface into the bargain. That one swap fixed more print failures than any amount of thumbwheel theology.
The second improvement was a sensible levelling routine instead of corner-chasing. The trick is to level diagonally and iterate, because adjusting one corner tilts the whole plane and changes the others. I settled on this:
- Home the printer and disable steppers so the head moves freely.
- Move to each corner in turn, about 5mm in from the edge, and set the paper drag.
- Go round again. It will have drifted. Go round a third time. By the third pass the changes are tiny and you're done.
Three passes, not one. Every time I got impatient and did one pass, I paid for it.
mesh bed compensation, when you've got it
If your board runs a recent Marlin and you have a probe, mesh levelling is the proper answer. It probes a grid of points, builds a height map of the actual bed, and adjusts Z live during the first few layers to compensate for the dips. You stop fighting the bed's shape and let the firmware ride over it.
The commands I use to build and save a mesh, for a manual mesh on a printer without an inductive probe:
G29 ; run the mesh probing/levelling routine
M500 ; save the mesh to EEPROM so it survives a power cycle
The one gotcha: a saved mesh only stays valid if nothing physical moves. The day you re-tension a belt, swap the glass, or nudge the bed springs, the mesh is a lie and you reprobe. I keep a sticky note on the side of the printer reminding me of exactly that, because I forgot it once and spent an evening blaming the slicer.
the bit that actually matters
After all of this, here's what I'd tell my past self. Bed levelling is not a one-off calibration you nail and forget. It's maintenance, like belt tension and lubrication. A glass bed and a three-pass routine got me 90% of the way; mesh compensation got most of the rest. The remaining 10% is just paying attention: a clean bed, a sensible first-layer height, and the discipline to slow down and reprobe when something moves.
I won't pretend the war is over. But I'm not losing every battle any more, and the first layer mostly goes down looking like a square drawn by someone sober. That'll do.