Why order matters
I hear from people chasing a new flow number or a different retraction value almost every week, when the actual problem is a belt that’s gone a little loose or filament that’s been sitting open since spring. Flow cannot fix a loose belt. No retraction setting fixes wet filament. Tune the settings that control how a print looks before the machine underneath it is sound, and you’re compensating for a problem you never found, one that comes back the moment anything else changes. Mechanical foundation and first layer first, appearance settings after.
1. Start from a baseline you can trust
Pick a profile and freeze it for the whole session. Select the exact printer and nozzle diameter in the slicer, and load a plain PLA profile, not the exotic filament you’ve been meaning to try, and not a profile someone shared in a forum with their own tweaks baked in. Ordinary PLA is forgiving enough that it won’t mask a mechanical problem the way a fussier material can. Resist updating firmware or slicer software mid-session too: an update between two tests adds a second variable, and now you can’t tell which one caused the result. Save a copy of that baseline profile before you touch a single number, so you have something to return to if a test goes wrong.
2. Look over the motion system while it’s off
With the printer off and cool, check it the way you’d check a bike before a long ride. Confirm the plate seats fully and doesn’t rock, and check the toolhead and X-axis carriage for play in the wheels: a little from new is normal on plenty of budget machines, more than that means something has worked loose. Look at the belts for fraying or a glazed, shiny patch from rubbing, and check that the hot end and bed cables aren’t pinched or catching on the frame at the extremes of travel. What this step isn’t for is guessing at belt tension by feel and cranking it tighter, or popping off covers the manual doesn’t list as user-serviceable. Both tend to create a new problem instead of solving the one you came in with.
3. Build or refresh the bed mesh
Wipe the plate down with isopropyl alcohol while it’s cool. Fingerprints and old glue-stick residue change how a probe reads certain surfaces. Some printers, Bambu Lab’s among them, build the mesh at printing temperature since the plate deforms slightly as it heats, so follow the manufacturer’s procedure instead of assuming cold is fine. Then run the routine and let it finish without interrupting it.
Be clear about what a mesh actually does. It’s a height correction map, small adjustments the firmware applies as the nozzle crosses a plate that isn’t perfectly flat. It cannot fix a plate that rocks on its mounts, a gantry that’s out of square, or a corner sitting meaningfully higher than the rest. Feed a warped plate into the routine and the mesh just compensates for a problem that reappears the moment you swap plates. That’s a mechanical fix, back to step two, not a software one.
4. Set the Z offset with a printed test
Paper is fine for getting into a safe starting range: slide a sheet between a cool nozzle and the plate until you feel slight drag. It isn’t the final answer though, since paper thickness varies between brands, and what feels right cold can be off once everything is up to temperature.
The printed first layer is the real evidence. Print a single-layer patch, a small square is enough, and adjust the Z offset in small steps (0.02 mm is common on printers that allow it) between reprints. On probe-equipped machines, BLTouch-style or Creality’s CR-Touch, that value is usually a probe-to-nozzle offset rather than a mechanical one, and confusing the two sends people chasing the wrong number. Stop when neighboring lines meet cleanly: no ridges, no scraping sound as the nozzle drags plastic sideways, and no glassy, translucent look that means the layer is squished thinner than it should be. That’s easy to miss if you’re only checking for gaps, and it causes problems later in the print.
5. Tune material temperature with a full tower
Start inside the range printed on the filament’s spool label, not a number you saw recommended for a different brand’s PLA. Two spools labeled PLA can want temperatures more than fifteen degrees apart depending on the pigment and additive package, and that gap shows up most with metallic and glow-in-the-dark filaments. Print a temperature tower suited to your slicer, most have a built-in setting for this now, and judge the result as a whole part rather than hunting for a single winning number.
Check layer bonding by flexing or lightly snapping a piece off each band, look at overhangs for sagging, and check the surface for stringing. It’s tempting to pick whichever band strings least and call the tower done, but the lowest-stringing temperature is worthless if it’s also the band where layers separate under light pressure. On a lot of consumer PLA the sweet spot sits a few degrees above where stringing reappears, a reasonable trade since stringing cleans up easier than a delaminated part.
6. Check extrusion consistency before touching flow
Before adjusting anything, confirm the basics are fine: the nozzle clear, no partial clog dragging flow down on one side. The spool should turn freely, not binding or tangled, since a spool fighting the extruder shows up as inconsistent flow that looks like a calibration problem but isn’t. And the extruder motor shouldn’t click or skip during a normal print, usually a sign it’s fighting more resistance than it can grip, not a flow problem.
With those checked, run a repeatable flow test, a thin single-wall cube is the classic version, or your slicer’s built-in flow calibration if it has one. Measure the actual wall thickness against what the slicer intended and adjust the extrusion multiplier from there, in small steps. A swing of a percent or two between spools or colors is normal. If the number needed is more than about ten percent off 100, that usually points to a different problem, an inaccurate e-steps value, a worn nozzle, heat creep, rather than something flow alone should paper over.
7. Tune retraction and pressure advance last
By this point the machine is mechanically sound, the first layer is dialed in, the material is running at the right temperature, and flow is close to accurate. Retraction and pressure advance (or linear advance, depending on your firmware) are the last things worth touching: fine tuning on top of everything else, not a fix for anything more fundamental.
Start from your printer profile’s tested range rather than a number pulled from a forum post about a different machine. A bowden setup and a direct-drive extruder need very different retraction distances, a few tenths of a millimeter for direct drive against several millimeters for bowden. Change one variable at a time, retraction distance, retraction speed, or pressure advance, and compare the same seam and travel move across every sample. It’s easy to convince yourself a change helped when you’re actually looking at a different part of the model than your baseline print.
Calibration checklist
- Mechanical checks completed with the printer cool and unpowered.
- Plate cleaned and bed mesh current.
- Z offset verified with a printed first layer, not just paper.
- Material temperature tested within its documented range.
- Extrusion checked and stable before touching the flow number.
- Each result recorded against the baseline profile.
Two of these steps deserve their own deep dive rather than a paragraph here. Z-offset calibration covers the difference between a probe’s stored offset and a manual mechanical one, and when redoing it is actually necessary versus just paranoia after a firmware update. E-steps calibration covers the firmware-level steps-per-mm value, the number that determines how far the extruder actually moves for every millimeter it’s told to move, and why it’s worth settling before you trust any flow number from step six.