Marlin Bed Leveling: Deterministic Calibration & First-Layer Mesh Engineering

Marlin Bed Leveling covers calibrating and troubleshooting Marlin 2.x automatic bed leveling systems - Unified Bed Leveling (UBL), Bilinear ABL, and Manual Mesh Bed Leveling (MBL) - to get consistent first-layer adhesion, through correct G-code sequencing, probe offset math, and persistent EEPROM storage.

Its central rule is that G28 (home all axes) silently disables bed leveling compensation by default, so every start G-code must restore it strictly after homing:

G28      ; Home all axes (silently turns leveling OFF)
M420 S1  ; Re-enable bed leveling compensation from EEPROM mesh

UBL calibration is treated as a five-phase state machine: Phase 1 (G29 P1) probes every grid point the probe can physically reach; Phase 2 (G29 P2) uses a paper feeler gauge for points the probe's gantry geometry can't reach; Phase 3 (G29 P3 T0.0) mathematically extrapolates unprobed edge points from adjacent measured slopes; Phase 4 (G26) prints a single-layer validation pattern that can be hand-edited with M421; Phase 5 saves the mesh to an EEPROM slot with G29 S1, activates it with G29 A, and commits it permanently with M500.

A thermal equilibrium rule requires heating the bed to its operating temperature and soaking it for at least five minutes before probing, since probing a cold bed and then printing hot bakes 0.08-0.20mm of error into every mesh point from non-linear thermal expansion. For downward-deploying probes like a BLTouch, the Z-probe offset (M851 Z...) must be negative - a positive value on this probe type drives the nozzle straight into the bed on the first move - and mesh fade height (M420 Z...) should be set around 10mm so compensation fades out smoothly by that print height instead of baking bed warp into the model's top surface.

A pre-emit checklist covers verifying M420 S1 sits strictly after G28, that a thermal soak preceded probing, that probe offset sign is correct for the probe type, that the mesh was committed with G29 S1 followed by M500, that fade height is set to a sane value, that physical bed tramming happened before software compensation, and that probe repeatability (via M48) is within 0.005mm standard deviation.

Use it when calibrating, tuning, or troubleshooting automatic bed leveling on Marlin 2.x printers (Ender 3, CR-10, Prusa clones, custom CoreXY or Cartesian machines) - setting up UBL, fixing a printer that homes and ignores its mesh, calculating a BLTouch Z-probe offset, generating a G26 validation print, or editing warped mesh points with M421. It also covers sizing slicer start G-code (Cura, PrusaSlicer, OrcaSlicer, Bambu Studio) for thermal soak and mesh restoration.

Do not use it for Klipper or RepRapFirmware printers, which use different firmware syntax and mesh systems; for gross mechanical faults like loose rollers, bent leadscrews, or a wobbly gantry, which software leveling cannot fix; or for delta printer kinematics setup, which needs radius and tower endstop adjustment before mesh probing.

Input is the printer's Marlin leveling system (UBL, Bilinear, or MBL), probe type, and bed/printing temperatures. Output is a correctly ordered calibration G-code sequence, a verified negative or positive Z-probe offset value, an EEPROM-committed mesh, and a slicer-ready start G-code block that restores the mesh with the right fade height after homing.

3D-printer owners and technicians calibrating or debugging Marlin 2.x automatic bed leveling - anyone chasing first-layer adhesion problems, corner detachment, or a mesh that silently stops applying after a G28 home.