The goal is not a good first layer. It is a first layer that is the same on Tuesday as it was on Sunday, on the left of the plate as on the right, and after a plate swap as before one. Anyone can get one good first layer by nudging the offset until it looks right; the interesting problem is making that unnecessary.
Almost everything that makes it repeatable is mechanical, and almost everything people adjust is not.
Three conditions, in order of authority
The distance from nozzle to plate must be the same everywhere. This is geometry — frame square, gantry level, bed carrier not twisted. Nothing downstream can fix a bed that sits at an angle to the motion system.
That distance must not change with temperature. A cold plate and a plate at 100 °C are different shapes. A probe reading taken cold and a print run hot are measuring different machines.
The plate must be clean. Adhesion is a surface-chemistry problem, and it belongs to bed adhesion, surface by surface rather than to this page — but no amount of levelling rescues a fingerprint.
Get those three right and the settings almost do not matter. Get them wrong and no settings will rescue you.
Levelling and meshing are different jobs
They are used interchangeably in conversation and they are not the same operation.
Levelling makes the bed physically parallel to the plane the nozzle moves in. On a machine with adjustment screws that is a manual job; on a machine without them it is a matter of the frame being assembled correctly. Levelling removes the cause.
Meshing measures the residual error at a grid of points and compensates for it in software by moving Z as the head travels. Meshing hides the symptom, and it does so imperfectly: the compensation fades out over the first few millimetres of height, and a large correction distorts the bottom of the part.
The practical rule is that meshing should be correcting fractions of a millimetre. If your mesh shows a large tilt, you have a levelling problem being papered over, and it will come back as an uneven first layer across the bed the moment the mesh drifts.
Probe at printing temperature if your firmware allows it. A mesh taken on a cold bed describes a plate that will not exist by the time the first layer is drawn.
Z-offset is the one number to tune
Everything above sets the shape of the plane. Z-offset sets its height, and it is the only first-layer number that should regularly need attention.
Tune it by looking at the extruded lines rather than at a piece of paper. Print a single-layer patch and inspect it:
- Lines with visible gaps between them, round in section: the nozzle is too high.
- Lines merged into a smooth sheet with ridges at the edges and a lip around the part: too low.
- Lines just touching, flat on top, no ridges: correct.
The paper test gets you close enough to start and is not accurate enough to finish. A single-layer test patch takes two minutes and is unambiguous. If the result is at either extreme, too squished and too high both have their own signatures and fixes.
Settings that genuinely help, and why
A short list, because the list of settings that do not help is much longer.
- First layer height at or slightly above the nozzle's normal layer. A thicker first layer is more forgiving of small plane errors, which is exactly what you want it to be.
- First layer line width around 120%. Wider lines touch each other more reliably and cover more plate.
- First layer speed low. Not because slow extrusion is better, but because a slow pass gives the plastic time to wet the surface and the bed time to conduct heat into it.
- Fan off for the first layers, on every material that tolerates it. Cooling is the enemy of adhesion.
- A brim, not a raft, when a part has a small footprint. A raft adds a whole set of new interfaces and rarely earns them.
Why yesterday's offset is wrong today
The most common frustration is a machine that was perfect last week and is not now. The usual causes are boring and mechanical:
- A different plate. Textured, smooth, and coated sheets are not the same thickness, and the difference is far larger than the tolerance you are working to. Keep a saved offset per plate.
- The plate is not seated. A single crumb of plastic under a magnetic sheet lifts a whole region of it.
- Temperature has changed. A different bed temperature for a different material changes the plate's shape.
- The mesh is stale. Plates and springs settle. Re-probing periodically is normal maintenance, and a mesh that drifts fast enough to notice is telling you something.
- The nozzle has debris on it. Baked-on plastic on the nozzle tip changes the effective offset and pushes the reading around.
A pre-print routine worth making automatic
- Look at the plate. If it has been touched, wash it.
- Check the nozzle tip is clean and cold-pull or brush it if not.
- Confirm the correct plate profile and offset are loaded for the sheet actually on the machine.
- Watch the skirt or brim go down. This is the whole payoff — nearly every failed print is visible at this moment, and stopping now costs seconds.
- Only then walk away.
Step four is the one that separates people who have a low failure rate from people who have theories about why theirs is high. It is also, in money terms, the cheapest intervention available: catching a failure in its first minutes rather than its last hour is worth more than reducing the failure rate itself, as counting the cost of a failed print works out with numbers.