Nothing changed. No new plate, no new nozzle, no new profile. First layers were dependable for months and have become gradually inconsistent — fine some days, patchy on others, and steadily more often the second than the first.
Gradual onset is the signature. A fault that arrives suddenly has a cause you can name; this one is wear, relaxation and drift.
A mesh is a photograph, not a description
When a machine probes its bed it records the plate's shape at that moment, at that temperature, with that sheet on it, with the probe behaving as it did then. Everything after that is compensation applied from a stored snapshot.
Four things in that sentence change over a year.
The plate takes a set. Repeated heat cycling relieves stress in a metal sheet, and plates deform slowly and permanently. A bed that was flat when new is not necessarily flat after five hundred hours at 100 °C.
Springs and mounts relax. Spring-mounted beds settle unevenly. Screws that were snug work loose. Both change the tilt without changing anything you can see.
The probe drifts. Inductive and capacitive sensors change trigger height with temperature, and physical switches wear at the contact. A probe that triggers a few hundredths later than it used to shifts the whole mesh downward uniformly.
The sheet changes. PEI wears thin where you print most, adhesive layers accumulate, and a well-used plate is genuinely a different shape from a new one in the middle where every part goes.
Test the probe before re-probing
This is the step that separates a mesh that needs refreshing from a probe that needs replacing, and it takes five minutes.
Probe the same single point ten times in a row and compare the readings. Most firmware has a command for exactly this. A healthy probe repeats within a few hundredths of a millimetre; a probe whose readings scatter by more is generating a fresh, wrong mesh every time you ask it to, and a re-probe will not help.
Do it hot as well as cold. A probe that repeats perfectly at room temperature and scatters at printing temperature is a thermal drift problem, and the machine will need a warm-up soak or thermal compensation rather than a new sensor.
Refreshing the mesh properly
- Tram the bed mechanically first. Software compensation should be correcting a small residual, not a large tilt.
- Heat the bed and let it soak at the temperature you print at, for several minutes.
- Clean the plate before probing if the probe touches it. A crust of adhesive is part of the shape you are measuring.
- Re-probe with a grid dense enough for the plate size.
- Re-check the probe-to-nozzle Z offset afterwards with a single-layer test patch. The mesh describes shape; the offset describes height, and refreshing one without the other moves the whole plate up or down.
How often is reasonable
There is no universal interval, and the honest answer is that it depends on how the machine is used. A printer that runs daily at high bed temperatures deserves a fresh mesh every month or two and a probe repeatability check when anything feels off. A machine used occasionally at PLA temperatures can go much longer.
What is worth doing more often is the cheap check: print a single-layer patch across the plate before any long or expensive job. It costs a minute of filament and it catches this fault before it costs a night.
Hardware worth changing
Replace bed springs with solid standoffs or silicone spacers. Springs exist so a bed can be adjusted by hand, and every machine with mesh compensation has largely outgrown that need. Solid mounts hold their setting for years.
Fit a new PEI sheet when the print region is visibly worn. Sheets are consumables and a worn one is both a shape problem and an adhesion problem.
Check the probe mounting. A sensor on a bracket that flexes reports the bracket's behaviour, not the plate's.
When the mesh is innocent
If the first layer is wrong the same way in the same places every time and has been since the machine was assembled, that is a static problem — see uneven first layer. If it went wrong abruptly after a specific event, look for what that event changed rather than at drift. And if adhesion is poor everywhere rather than unevenly, wash the plate before touching the mesh at all: a dirty sheet imitates this fault convincingly and costs nothing to rule out.