3D Printer Calibration Order: What to Tune First

Most calibration frustration is not a measurement problem. It is an ordering problem. Someone dials in flow rate on a printer whose e-steps are 6% low, then re-dials it after drying the filament, then re-dials it again after changing the nozzle, and concludes the printer is haunted. Every one of those measurements was correct at the moment it was taken and wrong five minutes later, because a variable underneath it moved.

This is the dependency map. It sets out which calibration depends on which, the order that keeps each measurement valid, and — just as usefully — what you have to redo when something changes. If you have been working from a printed checklist such as Ellis’ Print Tuning Guide or the Teaching Tech calibration site, this is the same philosophy applied to the procedures on this site.

The One Rule Behind the Whole Sequence

Every calibration measures one variable while holding the others still. That only works if the others are already correct. Flow rate measured on wrong-temperature plastic is a temperature reading in disguise. Pressure advance tuned on wet filament is a moisture reading. Retraction tuned before pressure advance is a pressure-advance reading.

So the order is not arbitrary or traditional. It runs from the variables nothing else depends on, up to the variables that depend on everything.

The Order at a Glance

#	Stage	Scope	Redo when
0	Mechanical baseline	Machine	Anything is disassembled
1	Bed level and Z-offset	Machine + surface	Bed, surface, or probe changes
2	Extruder e-steps	Machine	Extruder or motor changes
3	Filament condition	Spool	Every spool, after storage
4	Temperature	Material	New brand, colour, or material
5	Flow rate	Material	New material, new nozzle
6	Pressure advance	Material + hotend	New material, hotend, or nozzle
7	Retraction	Material + hotend	New material or hotend
8	Input shaping and speed	Machine	Mass on the toolhead changes
9	Geometry settings	Per part	Every job, deliberately

Stages 0 to 2 are hardware and you do them once. Stages 3 to 7 are per material and you repeat them for every filament worth keeping. Stage 8 is per machine. Stage 9 is a design decision, not a calibration at all.

Stage 0: Mechanical Baseline

Nothing downstream survives a loose belt. Before any test print, confirm the frame is square and the bolts are tight, belts are tensioned evenly, the wheels or linear rails are adjusted with no rock, the Z lead screws are not binding, and the extruder idler tension is firm enough not to slip but not so tight it flattens the filament. The Klipper project’s config checks page makes the same argument from the firmware side: verify motion direction and endstops before you tune anything about extrusion.

Skipping this stage is how people end up calibrating around a mechanical fault. If one axis is dimensionally wrong and the other two are fine, that is a belt or steps-per-mm problem, and no amount of flow tuning will fix it.

Stage 1: Bed Level and Z-Offset

These are two separate problems that get conflated constantly. Levelling makes the bed plane parallel to the toolhead’s motion plane. Z-offset sets how close the nozzle sits to that plane on the first layer. Automatic bed levelling solves the first and does nothing for the second — the mechanism and its limits are set out in bed levelling explained: manual vs auto, and the full probe-and-offset procedure for the most common probe is in how to level a bed with BLTouch.

The bed surface belongs to this stage too, because changing it changes the Z-offset. PEI, glass, garolite and powder-coated sheets each want a different first-layer approach; the trade-offs are compared in the print bed surface material guide.

Do this before extrusion tuning, not after. A first layer that is too high starves every test print you are about to measure.

Stage 2: Extruder E-Steps

E-steps (or rotation_distance in Klipper) is the hardware truth underneath every extrusion number you will ever set. If the extruder physically moves 97 mm when commanded to move 100 mm, then a 100% flow ratio is a lie and every material profile inherits the same 3% error.

Do this once per extruder, machine-agnostically, using the procedure in the extruder e-steps calibration guide. If you are on a Creality machine and want the exact G-code sequence and firmware commands, how to calibrate e-steps on Ender 3 is the machine-specific walkthrough of the same method.

The correct outcome is that 100% flow means roughly the right amount of plastic for every material. Per-material flow trim then stays in a sane band near 100% instead of drifting to 88% on one profile and 112% on another.

Stage 3: Filament Condition

This is the stage people skip, and it invalidates the four stages after it. Wet filament changes extrusion behaviour, surface finish, stringing and layer bonding — which means a temperature tower, a flow measurement and a pressure-advance test run on a damp spool all produce values that stop being correct the moment the spool dries out.

Dry the spool first, using the per-material times and temperatures in the filament drying protocol. Nylon and PC are non-negotiable here. PLA and PETG are more forgiving but not immune, especially after a humid summer on an open shelf.

Stage 4: Temperature

With dry filament and correct e-steps, find the temperature at which this specific spool actually flows well. A temperature tower gives you a single print that shows layer adhesion, surface finish, overhang quality and stringing across a whole temperature range, and it is the cheapest information-per-print test in the whole sequence.

Temperature comes before flow because the melt temperature changes viscosity, and viscosity changes how much plastic actually leaves a nozzle for a given commanded volume. Measure flow first and you will measure it again after you change temperature.

Stage 5: Flow Rate

Now the trim. With the hardware honest and the plastic at the right temperature, the single-wall test measures how much material genuinely comes out and scales the flow ratio to correct it. The measurement method, the maths, and the reason a calibration cube is a verification rather than a measurement are covered in flow rate calibration for dimensional accuracy.

Expect a value within a few percent of 100%. A value far outside that band is a message about stage 2, not about this spool.

Stage 6: Pressure Advance

Flow rate sets the average amount of plastic. Pressure advance controls the timing of it — the lag between the extruder commanding a change in flow and the nozzle delivering it. That lag is why corners bulge and why the first millimetres after a corner look starved even on a printer whose flow ratio is perfect. The firmware names differ (Klipper calls it pressure advance, Marlin calls it Linear Advance, Bambu Studio calls it flow dynamics) but the physics is one model; the test procedure and typical starting values are in pressure advance calibration for Klipper and Marlin.

Pressure advance must come after flow, because the test pattern is read by eye and over- or under-extrusion makes the pattern unreadable.

Stage 7: Retraction

Retraction last among the extrusion settings, and this ordering is the one most tuning guides get wrong. Pressure advance already removes a large fraction of nozzle pressure at the end of a line, so a printer with pressure advance enabled generally needs less retraction than the same printer without it. Tune retraction first and you will over-retract, then find your carefully chosen distance is wrong the moment pressure advance is turned on.

Starting values for direct-drive and Bowden setups, plus the interaction with travel speed and temperature, are in how to fix stringing on 3D prints.

Stage 8: Input Shaping and Speed

Only once the extrusion chain is correct is it worth pushing speed. Input shaping cancels the frame resonance that produces ringing, which moves the speed ceiling up substantially — but it does not fix a defect that came from bad flow, and it is measured against the mass currently on the toolhead, so it needs redoing after a hotend or fan-duct change. What limits speed, and where the acceleration and flow ceilings actually bind, is explained in print speed vs quality: where input shaping changes the math.

Stage 9: Geometry Settings, Per Job

Nozzle diameter and layer height are not calibrations. They are per-part decisions about the trade between detail, strength and time, and they are chosen together rather than separately — the coupling is set out in the nozzle size and layer height guide. Changing nozzle diameter, however, does send you back to stages 5 and 6, because line width and melt behaviour both change.

Before committing to a large job, the print time and filament cost estimator will tell you what a given layer height, infill and material combination costs in grams, hours and electricity, which is often enough to change the decision.

The Environment Sits Alongside All of It

Two variables run in parallel with the sequence rather than inside it.

Ambient temperature. Some materials need a warm chamber and one common material is actively harmed by it; the per-material targets are tabulated in the chamber temperature reference. Calibrating ABS in an open room and then enclosing the printer changes the results.

Adhesion and warping. If parts lift at the corners, that is a separate failure chain with its own ranked fixes, set out in why 3D prints warp at corners. Do not attempt to solve it with flow or retraction.

What to Redo When Something Changes

You changed	Redo from
Filament spool (same brand and type)	Stage 3, then verify stage 5
Filament brand, colour, or material	Stage 3
Nozzle (same diameter)	Stage 5
Nozzle diameter	Stage 5, and re-check stage 9
Hotend or heat break	Stage 4
Extruder or extruder motor	Stage 2
Bed surface	Stage 1
Toolhead mass (fan duct, hotend, camera)	Stage 8
Belts, rails, or frame work	Stage 0

Choosing the Material Before You Start

All of the above assumes you already know what you are printing. If that is still open, two decisions sit upstream of the entire sequence: whether the part wants the mechanical profile of PETG or PLA, compared from published datasheet figures in the PETG vs PLA strength comparison, and whether the part belongs on a filament printer at all, which is the question the FDM vs resin decision framework answers.

Calibration is cheap when it is done once in the right order, and expensive when it is done five times in the wrong one. Work down the list, record the value each stage produces against the spool and machine it belongs to, and the next material takes twenty minutes instead of an afternoon.