Phase 1 · The machine · L8
Thermal and motion trust
The last two things that have to stop moving: the temperature the nozzle actually holds, and the resonance the frame actually has.
Lesson 10 opens Phase 2 with a temperature tower — a test that assumes the number you asked for is the number the nozzle held. If the hotend is oscillating ±8 °C, your tower's bands aren't 5 °C apart, they're smeared, and you will pick a value that is roughly right for a machine that never sits still. The same goes for motion: Pressure Advance in Lesson 12 must be tuned at production accelerations, which means the machine's resonance behaviour has to already be settled. This lesson closes both, and with it Phase 1. After this, the machine is a constant.
Two independent things, one goal. Thermal trust: when the profile says 240 °C, the plastic sees something very close to 240 °C, consistently, for hours. Motion trust: when the profile commands an acceleration, the toolhead performs it without the frame ringing.
Which of these you can actually calibrate depends entirely on your firmware family — and here the split is sharper than in any previous lesson.
1Thermal trust
The thermistor sits in the heater block, not in the plastic. It reads how hot the metal is. The control loop's job is to hold that steady against everything trying to disturb it: the part-cooling fan blasting the nozzle, cold filament arriving at speed, ambient air moving through the room or the chamber.
A wobbling hotend temperature produces something genuinely nasty — variation you cannot see and cannot attribute. Viscosity changes with temperature, so flow, ooze and pressure lag all drift with the wobble. Two identical sections of the same print behave differently. That is exactly the corruption described in Lesson 4: results that look like data.
MK3S · PID calibration, then thermal model calibration
PID tuning calibrates the control loop that holds the hotend at target.
Path: LCD → Calibration → PID calibration.1
Set the temperature the calibration should run at — choose the temperature you print at most
often, because the loop is tuned around that operating point.1
Run PID tuning when you see small fluctuations, around ±5 °C. Fluctuations larger than ±5 °C are not a tuning problem — Prusa point at a fan fault or a thermistor wire/connector problem instead.1 Tuning past a hardware fault just calibrates around the fault.
Thermal model calibration is a separate, MK3-family feature from firmware 3.12.0 onward. It
builds a model of expected temperatures and compares measured against expected during printing; a
mismatch gives a beep and a THERMAL ANOMALY warning, and a persistent one shuts heating
down.2
- Path:
Calibration → Temp Model cal. Takes about 12 minutes and saves the profile automatically.2 - Run it after changing hotend components (third-party parts can trigger false warnings), and after PID tuning if thermal anomaly errors persist.2
And one MK3S-only extra: extruder linearity correction. The Trinamic drivers on the Einsy RAMBo assume a sine-wave motor current; real motors distort it, worst near the zero crossing, which shows up as diagonal moiré lines on printed surfaces.3 Requires firmware 3.2.0 or newer, and works only on the MK3 and MK3S.3
LCD → Settings → Lin.Correction → E correct— disable E-correction.3- Print the calibration G-code tower.
- Find the band with the smoothest surface — usually near the vertical centre — and note its value.
- Re-enable E-correction, enter that value, and reprint to verify.3
On firmware 3.13.0+ the menu shows values 0–200, where 1.03 appears as 030 and 1.11 as
110.3 Only chase this if you actually see
diagonal moiré; it is not a routine step.
MK4S · nothing to tune, something to watch
There is no user-run PID or thermal model calibration here. Prusa's PID tuning and thermal model calibration articles are written for the MK3 family;12 on Buddy firmware, heater behaviour is validated by the heater checks inside the selftest (Lesson 5) and surfaced as numbered error codes rather than exposed as a calibration you run.
So what you do instead is observe. Heat the nozzle to your intended print temperature, let it settle for five minutes, then watch the reading on the display for a full minute — with the part-cooling fan forced to 100 %, because that is the disturbance that matters and an idle hotend never sees it.
- Steady, within a degree or two: thermal trust established. Log it and move on.
- Wandering more than a few degrees: that is a hardware question — fan, thermistor, connector, or a sock that has gone. Re-run the selftest and search the exact error code if one appears.
The observation takes two minutes and gives you a logged baseline. When something goes strange in six months, you'll want a number to compare against.
Core One · as MK4S, but the chamber is a thermal variable
Same as the MK4 family — no user-run PID or thermal model calibration; watch and log instead.
The addition is the chamber. On an open-frame printer, ambient is the room. On the Core One, ambient is whatever the enclosure has accumulated, and it climbs through a long print. That makes the chamber a genuine thermal variable in a way it isn't on the other two machines.
So take your observation twice: once from cold, and once after the machine has been printing for half an hour in the door and vent state you use for that material. Log both, and log which state. If the two differ significantly, that difference is a real property of this machine that you'll want to remember when a Phase 2 measurement drifts between the start and end of a print.
Remember the PLA case from Lesson 2: PLA's glass transition is close enough to achievable chamber temperatures that a hot chamber softens the part and the filament path above it. Vent open for PLA. The community-consensus targets — ≤ 30 °C for PLA, 35–40 °C for PETG — are a starting point to test, not a Prusa specification.
2Motion trust — Input Shaper
Input Shaper cancels the resonances that cause ghosting, by analysing the printer's movements and filtering the input signal so it doesn't excite them — which is what lets the machine run high acceleration without wrecking surface finish. It is coordinated with pressure advance compensation.4
MK3S · doesn't have it, and that's fine
The MK3S has no Input Shaper. Its motion trust comes from belts being in spec (Lesson 6), a solid frame, and accepting that it is the slow machine of the three.
The practical consequence for Phase 2: don't try to make the MK3S hit MK4S numbers. Around 45–60 mm/s perimeters is the sensible operating region, and quality comes from staying there rather than from fighting ringing at 120 mm/s. Its bedslinger Y axis is carrying your part; there is no filter for mass.
The MK3S's motion-quality lever is instead the extruder linearity correction in section 1 — the one genuinely MK3S-specific surface-finish calibration, and the one to reach for if you see diagonal moiré that belt tension didn't fix.
MK4S · verify rather than calibrate
Calibration is usually unnecessary on a stock printer.4 The shipped values are correct for the shipped hardware. Prusa offer an optional accelerometer set for recalibration after hardware modifications;4 if your machine is stock, you don't need it.
Input Shaper frequency values should sit between 35 Hz and 70 Hz. A value too low or too high is Prusa's signal to go back and check belt tightness and the assembly.4 In other words, this number is a second, independent check on Lesson 6 — and it costs nothing to read. If it's outside the band, you have a mechanical problem masquerading as a motion setting.
One limitation worth knowing: crash detection is not compatible with Input Shaper, because of the high accelerations involved.4 Don't go looking for a crash-detection feature that isn't there.
What to verify, practically: print a tall thin object with a sharp feature — the same 40 mm cube with ribs from Lesson 6 works — at your intended production acceleration, and look downstream of the feature. Clean means motion trust is established at that acceleration. Ghosting means: belts first (Lesson 6), assembly second, and only then Input Shaper.
Remember what this machine is: a bedslinger. Its Y resonance is a function of the part — a tall heavy print changes the moving mass that Input Shaper was calibrated against. Judge ghosting on a part like the ones you actually print, not on an empty bed.
Core One · same verification, different resonant structure
Identical position to the MK4S: calibration is usually unnecessary on a stock printer,4 the accelerometer set is for after hardware modifications,4 and crash detection is not compatible with Input Shaper.4
The same 35–70 Hz band applies, and the same instruction if you fall outside it: go back and check belt tightness and the assembly.4 On this machine that check is Lesson 6's pair of belt frequencies — a CoreXY with mismatched belts can produce an Input Shaper result that looks acceptable while the two axes disagree.
What to verify, practically: the same tall ribbed cube at production acceleration. What differs is where to look. The CoreXY head is light and the part stands still, so ghosting here points at the gantry and the belt pair rather than at the mass of your print — and it will not get worse as the part gets taller, which is itself a useful discriminator against the MK4S.
One enclosure note: run this verification with the door in the state you actually print in. A warm chamber changes belt tension slightly, and this test is downstream of belt tension.
3Sign Phase 1 off
The point of a sign-off isn't ceremony. It is that from here on, when a Phase 2 measurement comes out strange, you can say with evidence "the machine was in this state, on this date, and these were its numbers" — instead of re-opening the machine as a suspect every time.
Re-run this sign-off after any of: a nozzle change, a belt adjustment, a sheet change, a firmware update, moving the printer, or a run of prints that went wrong for reasons you couldn't name. A machine that has been re-verified is a machine you can keep treating as a constant. One that hasn't is a suspect again.
4Check yourself
From memory. These interleave the whole of Phase 1.
Sign off Phase 1 on one printer
Finish the checklist above for the machine you've carried through Lessons 5–7, and write the whole
thing into tuning-log.md under a dated ## Phase 1 sign-off heading. Include:
- Every number you measured, with its target next to it.
- MK3S: the PID calibration temperature you chose, and why that temperature. MK4S / Core One: your one-minute temperature observation at 100 % fan, and the Input Shaper frequency.
- A photo of a 40 mm cube printed at your intended production acceleration.
- One sentence: "This machine is now a constant with respect to ___, and still a variable with respect to ___."
→ Paste it back to me. That last sentence is the real assessment — if you can name what Phase 1 didn't pin down, you understand what Phase 2 is actually measuring.
What you can now say
- An unstable hotend produces unattributable variation — the worst kind, because it still looks like data.
- PID tuning is for ±5 °C wobble. Bigger swings are a fan, thermistor or connector fault, and tuning around them hides them.
- PID and thermal model calibration are MK3-family features; on the MK4S and Core One you observe and log rather than calibrate.
- Input Shaper on a stock machine is verified, not calibrated — and its frequency landing outside 35–70 Hz sends you back to belts.
- Phase 1 is a repeatable sign-off, not a one-off. Any hardware change makes the machine a suspect again.
§References
- Prusa Research. PID tuning. The
Calibration → PID calibrationpath, the instruction to calibrate at the temperature you print at most often, and the ±5 °C threshold above which the cause is a fan or thermistor wire/connector fault rather than tuning. Link. - Prusa Research. Thermal model calibration. Applies to the MK3, MK3S and MK3S+ from
firmware 3.12.0; the
Calibration → Temp Model calpath; ~12 minutes; when to re-run it. Link. - Prusa Research. Extruder linearity correction calibration. MK3 and MK3S only, firmware
3.2.0+, Trinamic drivers on the Einsy RAMBo; diagonal moiré from current-waveform distortion; the
Settings → Lin.Correction → E correctpath and the tower procedure. Link. - Prusa Research. Input Shaper. What it cancels, that calibration is usually unnecessary on a stock printer, the optional accelerometer set for post-modification recalibration, the 35–70 Hz expected band and what a value outside it means, and the incompatibility with crash detection. Link.
Lesson L8 of the prusa-profile-tuning course · all lessons · The Calibration Card · glossary