The lines are laid down neatly and you can see gaps between them. Each one is round in cross-section, like a length of wire lying on a table, and it has a glossy sheen along the top. The layer looks tidy and comes off the plate almost by itself.
What a correct first layer looks like
Worth stating precisely, because "good adhesion" is not something you can judge from across the room.
- The lines have merged. No gaps, no visible boundary channels — the layer is a continuous sheet.
- The surface is matte, not glossy. A squashed bead has been flattened against the plate's texture and picks up that texture; a round one reflects light.
- There are no ridges between lines. Ridges mean the opposite fault: see first layer too squished.
- You can see the plate's pattern through it on a textured sheet. That is the texture the plastic keyed into, and it is the visual proof the bond was made.
- It resists a fingernail. A correct first layer takes real effort to lift while the plate is warm.
A round-lined first layer fails on all five, and it fails for one reason: nothing pressed it into anything. A round bead touches the plate along a line rather than a face, so the bonded area is a fraction of what it should be, and line-to-line bonding is poor as well — which is why parts printed this way delaminate at the base as easily as they detach.
The adjustment ladder
- Print a single-layer test patch, not a real part. A square about the size of a coaster, one layer thick. It takes a minute and it can be repeated a dozen times in the time one failed print wastes.
- Lower the offset while it prints, in 0.02 mm steps, using the machine's live adjustment ("babystepping" on most firmware). Give each step a full patch to show its effect.
- Stop as soon as the lines merge and the surface goes matte. The correct position is a band, not a knife edge, and the top of that band — the least squash that still merges — is the right place to be.
- Save it. Live adjustment on most machines is temporary unless explicitly stored.
- Confirm at temperature. The offset that is correct on a cold plate is not the offset that is correct at 100 °C.
Why an offset that was right stops being right
Two events account for nearly all of it.
A nozzle change. Nozzles differ in length by more than the tolerance you are adjusting to, and even the same part fitted to a different depth in the block changes the offset. Always re-set it after any hotend work.
A probe that has moved or drifted. A sensor on a bracket that got knocked, a magnet that shifted, or a trigger point that has changed with wear. Probing the same spot repeatedly and comparing the readings tells you whether the sensor is still trustworthy — the procedure is on bed mesh no longer accurate.
Firmware updates deserve a mention too: several have reset stored offsets to zero, and the resulting first layer is dramatic rather than subtle.
When lowering the offset does not fix it
If the layer merges in one region and stays round in another, this is not a single-number problem at all — the plate is tilted or dished, and the fix is on uneven first layer.
If it merges everywhere except where a new sheet has a raised area, check for debris trapped underneath before adjusting anything.
And if the layer looks correct on the patch test but real prints still lift, the offset is fine and the plate is dirty. Wash it with hot water and washing-up liquid — see first layer not sticking for why alcohol is the wrong tool for that particular job.
Do not compensate with flow
Raising the first-layer flow multiplier to fill the gaps between round lines produces a thicker layer that is still not pressed into the plate. It looks better and adheres no more strongly, and it introduces a dimensional error at the base of every part. Move the nozzle, not the volume.
Material notes
PLA is the most tolerant material here and will stick reasonably even from a slightly high first layer, which is exactly why the fault survives unnoticed on PLA machines until someone loads PETG or a polyamide. Those materials need real contact with the plate, and a first layer that PLA got away with will fail immediately. TPU is also unforgiving, because a soft filament does not push hard enough to squash itself into anything.