Anycubic Bed Leveling: Paper Method and Z Offset Steps

Bed leveling on an Anycubic printer is three separate jobs wearing one name. Tramming makes the plate parallel to the plane the nozzle moves in. Mesh compensation records how far the plate deviates from flat so the toolhead can follow the real surface. Z offset decides how far the nozzle sits above that surface when the firmware believes it is at zero. A sheet of paper participates in the first job and, indirectly, the third. It has nothing to do with the second.

Confusing the three is why leveling gets repeated week after week without the first layer ever improving. Levelling the corners again cannot fix a Z offset that is two hundredths of a millimetre too low, and no amount of Z offset will rescue a plate that sits visibly tilted.

What the paper is actually measuring

Ordinary office paper at 80 gsm is close to 0.1 mm thick. Common stock runs from about 60 gsm to 120 gsm, which spans roughly 0.07 mm to 0.15 mm. That range is the first source of error: the same procedure with a different sheet produces a different gap.

The second source of error is that the paper is not measuring distance at all. What gets judged is friction. The sheet is slid back and forth and the knob is turned until there is a light, consistent drag. How much drag counts as “light” depends on who is doing it, how fast the sheet is moving, and how hard a finger is pressing on the plate while the other hand turns the knob. Pressing on a spring-mounted bed deflects it, and the gap that was set under finger pressure opens up again when the pressure is released.

Read the paper method for what it is: a fast way to get every corner of the plate to within a tenth of a millimetre of the same height, using a reference that is consistent within one session. It puts the machine in the right neighbourhood. The final number comes from looking at extruded plastic, not from feeling paper.

The paper method on a Kobra bed

The sequence matters more than the technique.

Bring the nozzle and the bed to the temperatures the first layer will actually print at. A hot nozzle sits lower than a cold one, and a heated plate is not the same shape as a cold plate. Levelling cold and printing hot bakes that difference into every print.

Wipe the nozzle clean before touching anything. A cooled bead of filament clinging to the tip adds its own thickness to every measurement, and because that bead is easily as thick as the gap being set, it is a common reason a careful leveling pass still produces a bad first layer. Use a brass brush or a folded piece of cotton cloth with the nozzle hot, and mind the heat.

Clean the plate with isopropyl alcohol. Finger oils cause adhesion failures that look exactly like a leveling problem, and chasing them with the knobs makes things worse.

Home the machine, then disable the motors or use the firmware’s manual leveling routine so the toolhead can be positioned without the steppers fighting. Work the four corners in a circuit, sliding the sheet under the nozzle at each one and adjusting until the drag matches. Then do the circuit a second time. Tightening one corner tilts the plate and changes the other three, so a single pass is never converged. Finish with a check in the centre; a centre that is noticeably tighter or looser than four matching corners means the plate is bowed rather than tilted, which is a job for mesh compensation.

On machines with Anycubic’s LeviQ automatic leveling, the corner adjustment still matters. The probe builds a correction mesh on top of whatever plane the hardware presents. If that plane is badly tilted, the mesh spends its entire correction range on the tilt and has nothing left for the actual surface shape. A mesh with a large range is a signal to go back to the screws.

The paper method on a Photon build plate

The resin side uses the same sheet for a different purpose. A Photon build plate is not being levelled against a heated bed; it is being made parallel to the masking LCD, because the LCD is the surface that defines layer zero.

The routine is to loosen the plate’s clamping screws so it can pivot freely, lay the supplied leveling card or a sheet of paper flat over the screen to protect it, home the Z axis so the plate descends onto the card, press the plate down evenly so it settles flat against the screen, then tighten the screws in a cross pattern rather than in a circle. Cross-pattern tightening keeps the plate from walking as each screw takes up load. With the screws tight, the card should be gripped firmly but still be extractable. Then set the Z zero position in the menu.

Two cautions apply to the resin procedure specifically. Never drive the plate down onto a bare masking screen without the card in place, because the plate is steel and the screen is glass over an LCD panel. And never run the leveling routine with resin in the vat, because the vat has to come out for the plate to reach the screen at all.

Later Photon Mono machines ship with the plate mounted on a fixed, factory-set interface, which turns leveling from a routine into an occasional check after a crash or a hard knock. The check is the same: a card under the plate at Z zero should be pinched evenly across its whole area.

Where automatic leveling stops and Z offset starts

This is the boundary that costs people the most time. Both major firmware families draw it in the same place.

A probe measures the plate at a grid of points and stores the result as a mesh, which the motion planner then adds to every move so the nozzle traces the real surface instead of an ideal plane. The Klipper bed mesh documentation and Marlin’s auto bed leveling feature page both describe this as compensation for shape.

What neither can determine is the constant term. The probe triggers at some fixed distance from the nozzle tip, and that distance is a property of the hardware. Firmware stores it as a configured value, exposed in Marlin through M851 and in Klipper through the probe’s z_offset, set with PROBE_CALIBRATE. Automatic leveling maps the surface; the offset decides how high above that mapped surface the nozzle rides. It is a number a person has to supply.

That is the whole reason a printer with flawless auto leveling still lays down a first layer that will not stick. The mesh is correct and the offset is wrong.

Setting Z offset by looking at the first layer

Start a print with a large single-layer patch, or use the firmware’s first-layer calibration routine, and watch the extrusion while adjusting the offset live in steps of 0.01 mm to 0.02 mm. Klipper exposes this as SET_GCODE_OFFSET; Marlin exposes it as babystepping in the tune menu.

A correct first layer is a continuous sheet with no valleys between the beads. Each bead is squashed slightly wider than the nozzle diameter and fused into its neighbours, and the surface has a uniform matte sheen with the texture of the plate faintly visible in it.

Too high shows as separate round strands with gaps between them, a surface you can see the plate through, and corners that lift as the part cools. Too low shows as a translucent, ridged film, plough marks where the nozzle has dragged through material it already laid down, a pronounced lip around the outside of the part, and under-extrusion further into the print as the constricted nozzle starves.

Save the value to the machine when it looks right. An offset that lives only in the tune menu disappears at the next power cycle.

Symptoms and what they usually mean

• First layer sticks at the front and lifts at the back: tramming, not offset. Redo the corner circuit twice.
• First layer good at the edges, poor in the middle: a bowed plate. This is what mesh compensation exists for; increase probe point density rather than turning knobs.
• Perfect one day, unusable the next: a dirty plate, a dirty nozzle, or leveling done at a different temperature than printing.
• Mesh values that change between consecutive probing runs: probe repeatability. A probe that varies by more than a couple of hundredths of a millimetre makes the mesh noise rather than data.
• Resin prints that fail only in the first few layers: bottom exposure or plate flatness against the LCD, not the Z offset. Bottom-layer exposure is a separate setting from normal exposure, and the resin exposure calculator gives a starting point for both.

When paper is the wrong tool

A 0.1 mm feeler gauge removes the paper-thickness variable and gives a harder, more repeatable stop than a sheet that compresses. A printed first-layer test patch removes the friction-judgement variable entirely and is the only method that measures what actually matters.

On the resin side, none of this applies to print quality in the same way. Once the plate is flat against the screen, the variable that governs whether a resin print succeeds is exposure time, which depends on the resin rather than the machine. That is covered in the guide to Anycubic resin types and how they differ, and the geometry problems that survive a correct exposure are covered in the guide to checking a sliced file before printing.

Common mistakes

Leveling with a cold nozzle and printing hot. Leaving a bead of filament on the tip. Doing one corner circuit instead of two. Pressing hard on a spring bed while adjusting it. Treating a large mesh range as normal instead of as evidence of tilt. Expecting automatic leveling to supply the Z offset. Adjusting the offset without saving it. Tightening a resin plate in a circle instead of a cross. Assuming a resin print that fails in the first millimetre is a leveling problem when bottom exposure is the more likely cause.

If the machine choice itself is still open, the differences that drive all of this are set out in resin MSLA compared with high-speed FDM.

Related on this site

• Anycubic Kobra 2 Not Extruding Fix: Extruder Fault, Clog or Heat Creep