Phase 1 · The machine · L6

Belt tension and squareness

Three printers, three completely different measurements — a number on a screen, a feel in your fingers, and a frequency in Hertz.

⏱ ~25 minutes on the machine• 🔧 Tools: 2.5 mm Allen key, phone• Phase 1

◎ Why this lesson

Belt tension is the single mechanical fault most likely to end up encoded in a filament profile. A slack belt produces ghosting, out-of-round circles and dimensional error — all of which look like things you could tune away with speed, acceleration or advance settings, and none of which you can. It also has to be right before Input Shaper means anything: Prusa are explicit that belt adjustment and maintenance are prerequisites for it, not alternatives.1 This is the lesson where you turn the machine into a constant.

A belt is a spring. Too slack and the toolhead lags behind the motor on every direction change, then overshoots — motion error that scales with acceleration. Too tight and you load the bearings and pulleys, add friction, and skew the printer's own axis-length measurements. There is a window, and each of your three machines expresses it in a different unit.

1Measure it — on your machine

Open your machine — only yours. These procedures are not interchangeable and the numbers are on three different scales.

MK3S · read a number off the screen

The MK3S is the easy one: the firmware measures belt tension for you and reports it as a number.

To take a fresh measurement: LCD → Calibration → Belt test. The values also update every time you run the Selftest.2

To read it: LCD → Support → Belt status.2

Targets. The value is inversely proportional to motor load — **higher means looser**.2
Axis	Target	Acceptable	Out of range
X	~250	±15	Under (or close to) 240 → loosen. Over (or close to) 290 → tighten.2
Y	~275	±15

Read the direction rule twice — it inverts what intuition expects. A small number is a tight belt, because a tight belt loads the motor more.

The physical adjustment is at the belt tensioner on the X-carriage and the Y-belt idler. Make small changes, then re-run the belt test — don't adjust twice before measuring. The one-variable rule applies to hardware too.

Cross-check with axis length. If Calibrate XYZ reports an axis-length error, suspect belt tension before suspecting a broken axis — Prusa note that slightly too tight or too loose tension is a common cause.3 Fix the belt, then re-run Calibrate XYZ, then re-read Support → XYZ cal. details for the skew result.

MK4S · adjust by feel, verify by print

The MK4S does not report a belt-status number. There is no screen to read; you adjust the hardware and verify in a print. That makes this the least satisfying of the three procedures and the one most worth doing carefully.

The feel target: plucked at its midpoint, the belt should sound like a low bass string. You should be able to pinch the two runs together between thumb and forefinger, but with noticeable resistance.

X axis — small adjustment4:

Loosen all the motor-holding screws slightly, just enough to allow movement.
Using the ball end of a 2.5 mm Allen key, tighten the rear screw on the X-end-motor.
Check tension after every turn or two — not after ten.
Re-tighten the motor screws when you're satisfied.

X axis — large adjustment: remove the three screws on the X-carriage cover, then move the belt mount by one or two teeth.4

Y axis — small adjustment: behind the LCD module, turn the two M3x10 screws clockwise to close the gaps between the idler and the frame.4

Y axis — large adjustment: lay the printer on its side, loosen the right screw counter-clockwise, turn the centre screw clockwise to tighten the belt, then re-secure the right screw.4

Verify with a print, not with your fingers. Print something with a sharp direction change — a 40 mm cube with a couple of raised ribs is enough — at a deliberately high acceleration. Ghosting downstream of the ribs is your signal. Change one belt, re-print, compare. If ghosting survives correct tension, that's an Input Shaper question, and it belongs to Lesson 8.

Core One · measure a frequency, and keep the gantry square

This is the most precise of the three, and on a CoreXY machine it is also the most consequential — both belts drive both axes, so unequal tension racks the gantry out of square and circles come out oval.

Path: LCD → Settings → Manual Belt Tuning on firmware 6.4.0+.5 Newer firmware adds a guided wizard that excites the belt and gives stroboscopic feedback so you can see the resonance; if your machine doesn't offer it, use the manual strum-and-measure method below.6

Targets for the Core One / Core One+.5
	Target
Both belts within	90–98 Hz
Upper belt	max 98 Hz
Lower belt	min 92 Hz
Difference between the two	≤ 8 Hz

(The Core One L is a different machine with different targets — 85–95 Hz. Check which variant you have before using a number.)

Procedure5:

Disable the motors from the LCD, or power the printer off.
Move the print head to the frontmost position on Y and centre it on X.
Strum the upper belt at its midpoint and measure the frequency with a phone tuner or spectrum-analyser app.
Strum the lower belt and record its frequency.
Adjust the tensioner screws in half-turn increments, alternating between the left screw (upper belt) and the right screw (lower belt).
Measure again after every adjustment.

⚠ The rule that makes this CoreXY-specific

Always tighten both tensioner bolts evenly and alternately. This is what keeps the gantry square while you adjust.5 Tighten one fully and then the other and you will hit your target frequencies with a racked gantry — every number in spec, every circle still oval.

A free phone tuner app works. Strum the belt like a guitar string, mid-span, with the head parked as above — position changes the free length and therefore the frequency, which is why the procedure specifies where to park it.

2Reading the machine from the print

Once belts are in spec, print a test and read it. Different defects point at different mechanical causes, and knowing which is which stops you attributing motion faults to filament.

Belt too loose

# the signature
ghosting after corners
  → worse at high accel
  → worse on tall parts
circles slightly oval
layer shifts under load
dimensions run small
in one axis only

Belt too tight

# the signature
audible motor whine
bearings run warm
axis-length errors
  → in XYZ calibration
stepper skipping steps
quality fine until it
suddenly isn't

The asymmetry is worth noticing: too loose degrades gracefully and visibly; too tight fails abruptly and quietly. People overtighten because loose is the failure they've been taught to look for. On the MK3S the belt-status number protects you from this — under 240 means loosen. On the MK4S, where you're going by feel, err towards the low-bass-string end rather than the high one.

▲ Why this must precede Input Shaper

Input Shaper works by cancelling the frame's resonant frequencies. It measures those frequencies and filters the motion commands to avoid exciting them. A loose belt doesn't shift a resonance — it adds slop, a mechanical dead zone that no filtering can compensate for. Prusa's own guidance puts belt adjustment, nozzle inspection and debris removal ahead of Input Shaper for exactly this reason.1 And the Input Shaper frequency values themselves are a diagnostic: they should land between 35 and 70 Hz, and a value outside that band is Prusa's signal to go back and check belt tightness and assembly.1 You'll read that number in Lesson 8.

3Diagnose the machine

Click each symptom. Decide first whether it's slack, over-tension, or something that isn't belts at all.

Slack, over-tight, or not a belt problem?

Ghosting that gets dramatically worse at higher accelerationSlack
MK3S reports an axis-length error, belt status reads 232Over-tight
Core One: both belts inside spec, circles still consistently ovalSquareness
Quality is perfect for 40 mm, then degrades steadily upwardNot belts
Motors whine audibly, bearings warm to the touchOver-tight

4Check yourself

From memory. Question 1 is the one people get backwards.

Question 1 of 3

Your MK3S reports belt status Y = 235. What does that mean and what do you do?

Question 2 of 3

Core One: upper belt 97 Hz, lower belt 92 Hz. Both in range. Is this acceptable?

Question 3 of 3

Why can't Input Shaper compensate for a loose belt?

✎ Your task · do this with your tutor

Measure and log belt tension on one printer

Take the machine you started in Lesson 5 and get its belts into spec. Record in tuning-log.md:

The before number. MK3S: belt status X and Y. Core One: upper and lower Hz, and the gap between them. MK4S: describe the feel, and note the pluck pitch if you can measure it.
What you changed, in physical terms — which screw, how many turns, which direction.
The after number, measured the same way as the before.
MK3S only: re-run Calibrate XYZ and log the skew result again. Did it move?
Core One only: the gap between belts, explicitly. Two in-band numbers 9 Hz apart is a fail.

Then print a 40 mm cube at a deliberately high acceleration, photograph a corner, and tell me what you see.

→ Paste it back to me with the photo. I'll tell you whether what you're looking at is ghosting, over-extrusion at the corner, or pressure lag — the three look similar and get confused constantly.

What you can now say

Each machine expresses belt tension in a different unit — a number on the MK3S, a feel on the MK4S, a frequency on the Core One.
On the MK3S, lower means tighter. Under 240, loosen; over 290, tighten.
On the Core One both belts must match within 8 Hz, and the tensioners are turned evenly and alternately so the gantry stays square.
Too loose degrades visibly; too tight fails abruptly. Which is why people overtighten.
Belts come before Input Shaper, and an Input Shaper frequency outside 35–70 Hz is a signal to come back here.

📖

Primary source · read this next

Adjusting belt tension (MK4/S, MK3.9/S, MK3.5/S, MK3/S/+) — Prusa Knowledge Base

The source of every MK3S and MK4S number above, with photographs of each screw. Core One owners want the CORE One article instead — different machine, different procedure, different units.

💬 I'm your tutor — ask me anything. Belt work is the part of this course where a photo is worth more than a paragraph. Show me the tensioner you're about to turn if you're not sure it's the right one, and send the before-and-after numbers even if they look fine. Confused beats stuck.

Linked: Glossary → belt tension, ghosting, Input Shaper, CoreXY, bedslinger, skew, selftest, one-variable rule, tuning log. Keep it close with → The Calibration Card.

§References

Prusa Research. Input Shaper (CORE One, MK4/S, MK3.9/S, MK3.5/S, XL, MINI/+). Belt adjustment and maintenance are prerequisites; Input Shaper frequency values should fall between 35 and 70 Hz, and a value outside that band means checking belt tightness and assembly. Link.
Prusa Research. Adjusting belt tension (MK4/S, MK3.9/S, MK3.5/S, MK3/S/+). MK3S belt status targets (X ≈ 250, Y ≈ 275, ±15), the inverse relationship to motor load, the under-240 / over-290 action thresholds, and the Support → Belt status and Calibration → Belt test paths. Link.
Prusa Research. Check Axis Length X/Y/Z (MK3S). Slightly too tight or too loose belt tension is a common cause of axis-length errors. Link.
Prusa Research. Adjusting belt tension, MK4S section. The X-end-motor rear screw with a 2.5 mm Allen ball end, the X-carriage cover route for large adjustments, and the M3x10 screws behind the LCD module for Y. Link.
Prusa Research. Adjusting belt tension (CORE One). The 90–98 Hz band, the 8 Hz tolerance between belts, the Settings → Manual Belt Tuning path on firmware 6.4.0+, the strum-and-measure procedure, and the instruction to tighten both tensioner bolts evenly and alternately to keep the gantry square. Link.
Prusa Research. Belt tuning wizard. The guided wizard using controlled belt excitation and stroboscopic feedback to visualise belt resonance. Link.

Lesson L6 of the prusa-profile-tuning course · all lessons · The Calibration Card · glossary