First Layer & Bed Levelling Calculator
Why the first-layer tolerance is one-sided, and why paper-drag levelling is 2.5x coarser than the window it is trying to hit.
What each Z error does to the line
How twitchy each first-layer height is
The tolerance is one-sided, so "when in doubt go lower" isn't taste
Squish converts height into width at constant volume, and the slicer spaces its lines at the nominal width. At zero Z error the line comes out exactly that wide and the lines abut. Push the nozzle 0.02 mm high and it narrows to 0.390 mm against a 0.42 mm spacing — a 0.030 mm void between every line on the plate. Push it 0.02 mm low and it widens to 0.458, which simply overlaps and merges. One direction fails and the other doesn't, so the safe side is a fact about the error function rather than a preference.
A thin first layer is much less forgiving. Width responds to height as volume over height, so sensitivity climbs sharply as the layer thins: 0.01 mm of Z moves the width 0.010 mm at a 0.3 mm layer, 0.017 at 0.2, and 0.038 at 0.1 — a factor of 3.9 across that range. That's the arithmetic behind printing a first layer thicker than the rest, and it's a real effect rather than folklore about adhesion.
And manual levelling is coarser than the window it's aiming at. The paper-drag method repeats to about ±0.05 mm on a good day; the first layer wants ±0.02. That's 2.5:1 the wrong way — the method can't succeed by care alone, which is the quantified case for a probe and a mesh rather than a preference for one. Loosen the window to ±0.05 and paper is adequate again, which is exactly why it works for people happy with a coarser first layer.
The bed flatness this implies is startling: ±0.02 mm across a 220 mm bed is 0.0052 degrees — nine microns per 100 mm of travel. No bed is that flat and no manual adjustment holds it there. Levelling corrects tilt; only a mesh corrects shape, and on a large bed the mesh is doing most of the work.
- Judge from the top of the first layer, not the underside. Visible lines with valleys between them mean too high; a glassy surface with ridges pushed up at the edges means too low; an even, faintly textured sheet with no gaps is right. The underside looks flat either way.
- Set it hot, at the temperatures you print at. A nozzle at 200 °C sits measurably lower than the same nozzle cold, and a bed at 60 °C has bowed from where it was at room temperature. A gap set cold is not the gap you print with.
- One corner right and another wrong is bed shape, not Z offset. No single number fixes it — a spring-levelled bed can have tilt removed but not a dish or a crown. That's what a mesh is for.
- If the same point doesn't probe the same twice, stop levelling. That's a mechanical problem — a loose bed, a flexing gantry, a bad probe — and no amount of adjustment settles it.
How to use
- Set the Z offset with the nozzle and bed at printing temperature, not cold.
- Judge the top surface of the first layer, not the underside.
- If in doubt go lower — the error is only costly in one direction.
- Use a thicker first layer than the rest of the print for margin.
Frequently asked questions
Should my first layer be a little high or a little low?
Low, and this is arithmetic rather than taste. Squish converts height into width at constant volume while the slicer keeps spacing lines at the nominal width, so a nozzle 0.02 mm too high narrows each line to 0.390 mm against a 0.42 mm spacing — a 0.030 mm void between every line. The same error the other way widens it to 0.458, which simply overlaps and merges. One direction fails and the other does not.
How do I tell if the nozzle is too high or too low?
Look at the top of the first layer, not the underside. Visible separate lines with valleys between them mean too high. A glassy surface with ridges pushed up at the edges of the part means too low. An even, faintly textured sheet with no gaps between lines is right. The underside tells you about the build surface and adhesion but almost nothing about Z, because it comes out flat either way.
How precise does the Z offset actually need to be?
About plus or minus 0.02 mm for a first layer with no gaps and no ploughing, on a typical 0.42 mm line at a 0.2 mm layer. That is a genuinely small window — roughly a quarter of the thickness of a sheet of paper — and it is the reason first layers cause so much more trouble than any other part of a print.
Is the paper method good enough for levelling?
Not against a 0.02 mm window. The paper drag repeats to about 0.05 mm even done carefully, because paper thickness varies, the drag you judge as correct varies, and you are setting a cold gap for a hot print. That is two and a half times coarser than the target, so it cannot succeed by care alone. It is adequate if you accept a 0.05 mm window and a coarser first layer, which is why it works for plenty of people.
Should the first layer be thicker than the rest of the print?
Usually yes, and there is a real reason rather than a superstition about adhesion. Width responds to height as volume over height, so sensitivity climbs sharply as the layer thins: 0.01 mm of Z moves the width 0.010 mm at a 0.3 mm layer, 0.017 at 0.2 and 0.038 at 0.1. A 0.1 mm first layer is nearly four times as twitchy as a 0.3 mm one, so a thicker first layer buys margin against exactly the error you are most likely to have.
How flat does the bed have to be?
Startlingly flat. A plus or minus 0.02 mm window across a 220 mm bed is 0.0052 degrees of tilt, which is nine microns per 100 mm of travel. On a 350 mm bed it is 0.0033 degrees. No bed is manufactured that flat and no manual adjustment holds it there, which is the entire reason mesh compensation exists and why it matters much more on a large machine.
Why does levelling not fix one bad corner?
Because that is bed shape rather than bed tilt, and manual levelling can only remove tilt. A plate with a dish or a crown in it stays dished no matter how the springs are adjusted — you can trade one corner against another but not flatten the middle. A probed mesh measures the actual surface and compensates for it point by point, which is the only thing that addresses shape.
Should I level the bed hot or cold?
Hot, at the temperatures the print will use. A nozzle at 200 C sits measurably lower than the same nozzle at room temperature, and a bed at 60 C has bowed from where it was cold. A gap set cold is simply not the gap you print with, and the difference is comparable to the whole tolerance window. Heat both, wait for them to settle, then adjust.
What is first layer squish as a percentage?
It is how much the commanded layer height is reduced by the Z offset. Dropping 0.02 mm on a 0.2 mm layer is 10 per cent squish, which widens the line about 9 per cent — from 0.42 to 0.458 mm. People quote squish as a percentage because it transfers between layer heights better than a raw Z number does, though the resulting width is what actually matters.
Why does my first layer look different in different areas of the bed?
Almost always bed shape, sometimes a warped build plate and sometimes an uneven mounting. Since the whole tolerance is around 0.02 mm, a plate that deviates by a tenth of a millimetre across its width will look perfect in one region and gappy in another with no adjustment able to fix both. That variation is what a mesh is measuring and correcting.
Can I fix first layer problems by changing the flow rate instead?
You can compensate, but you are treating the symptom. Raising flow to fill the gaps left by a nozzle that is too high thickens every subsequent layer too, so the part grows and top surfaces over-extrude. It is worth adjusting flow only after the Z offset is right, and then only by a few per cent. If a large flow increase is needed to make the first layer look correct, the Z is wrong.
Why does the same point probe differently each time?
That is a mechanical fault rather than a levelling problem, and no amount of adjustment settles it. Common causes are a bed that shifts on its mounts, a gantry that flexes under probing force, a loose probe mount, or a magnetic sheet that is not seated the same way each time. Check repeatability before spending any more effort on levelling, because a bed that does not repeat cannot be levelled at all.
🔒 This tool runs entirely in your browser. Nothing you enter is uploaded, logged, or stored.