Ghosting — those repeating ripples that echo surface features across your print — is almost always a belt tension problem. Too loose and the print head overshoots during direction changes; too tight and the belt rings like a guitar string, transmitting vibration into the printed surface. The sweet spot is narrow: roughly 50-70Hz resonant frequency for a 6mm GT2 belt on a typical Cartesian printer. Here’s how to measure it, set it, and verify the result.
Why Belt Tension Matters More Than Most Upgrades
A loose belt creates backlash — a dead zone between the motor changing direction and the print head actually moving. Backlash produces layer misalignment, rounded corners, and dimensional inaccuracy. A tight belt creates ringing — the belt’s natural resonance is excited by the stepper motor’s direction changes, and that vibration couples into the print head.
The physics: a GT2 belt under tension behaves like a guitar string. Its resonant frequency is f = (1 / 2L) × √(T / μ), where L is the free span length, T is tension, and μ is the belt’s linear density. For a typical Ender 3 with a 350mm X-axis belt span, 50-70Hz corresponds to roughly 1.5-2.5kg of tension. This is well below the stepper motor’s radial load limit (~10kg for a NEMA 17) but enough to eliminate mechanical backlash.
Belt Tension Quick Reference
| Printer Type | Axis | Belt Type | Typical Span | Target Frequency | Tension (approx.) |
|---|---|---|---|---|---|
| Ender 3 (bed-slinger) | X | 6mm GT2 | 350mm | 55-70Hz | 1.5-2.5kg |
| Ender 3 (bed-slinger) | Y | 6mm GT2 | 400mm | 45-60Hz | 1.2-2.0kg |
| CoreXY (Voron, RatRig) | A/B | 6mm or 9mm GT2 | 300-450mm | 50-65Hz | 2.0-3.5kg |
| Prusa i3 style | X | 6mm GT2 | 320mm | 55-65Hz | 1.5-2.2kg |
| Prusa i3 style | Y | 6mm GT2 | 350mm | 50-60Hz | 1.3-2.0kg |
| Delta printer | All | 6mm GT2 | 600-900mm | 35-50Hz | 1.5-2.5kg |
How to Measure Belt Tension by Frequency
You don’t need a dedicated tension meter. A smartphone with a spectrum analyzer app (Spectroid for Android, SpectrumView for iOS, or the iOS “Audio Spectrum Analyzer”) can measure belt frequency with surprising accuracy.
Procedure: Pluck the belt at its midpoint like a guitar string. Hold the phone’s microphone 5-10cm from the belt. The spectrum app will show a peak — that’s your belt’s resonant frequency. On bed-slingers, the Y-axis belt is harder to reach because it runs under the bed — pluck from underneath or use a long tool to strum it.
A belt that reads below 40Hz on a 350mm span is too loose — tighten until you reach 50-60Hz. Above 90Hz, the belt is dangerously tight and will accelerate bearing wear on the stepper motor and idler pulleys. Back off to 65-75Hz.
If you have two parallel belts (CoreXY’s A and B belts), they must match within 10Hz. Unequal tension on the A and B belts produces skew — the tool head doesn’t move in a true straight line because one belt stretches more than the other under acceleration. Measure both, adjust the looser one until frequencies match.
Common Mistakes
Mistake 1: Tightening until the belt “feels right” by hand. Human tactile sense for belt tension is about as accurate as guessing bed temperature by touching it. On my last tune, a belt that “felt tight” measured 35Hz — well into backlash territory. A belt that “felt about right” was at 82Hz, in the ringing zone. Use the spectrum analyzer method. It takes 30 seconds and is reproducible across tuning sessions.
Mistake 2: Unequal tension across parallel belts on CoreXY. The A and B belts on a CoreXY printer form a kinematic chain — unequal tension produces non-orthogonal motion. Your calibration cube will measure square on two axes but off on the diagonal. The fix is measuring both belts independently and adjusting until they match within 5-10Hz.
Mistake 3: Over-tensioning as a band-aid for a worn belt. A belt with worn teeth or a stretched section won’t be fixed by cranking the tension higher. If the belt has been on the printer for 1000+ print hours and you’re chasing ghosting that won’t go away, replace the belt. A genuine Gates GT2 belt costs $5-8 and is worth replacing during any major maintenance window. Generic unbranded belts often have inconsistent tooth pitch that produces periodic artifacts regardless of tension.
Mistake 4: Neglecting the idler bearing. The idler pulley bearing is in the belt path and experiences the same tension as the belt. A gritty or seized idler bearing creates periodic resistance that produces a ghosting pattern at the idler’s rotation frequency — different from the belt’s natural frequency ghosting. Spin the idler by hand with the belt off. It should rotate freely and silently. Replace any idler that clicks, grinds, or has radial play.
Mistake 5: Aligning the belt by eye instead of with a reference edge. A belt running at an angle across the idler or motor pulley will ride up the pulley flange, generating friction that shows up as inconsistent motion. Use a straightedge to verify the belt runs parallel to the extrusion channel, centered on the idler, and perpendicular to the motor shaft. A belt tracking correctly will stay centered on the pulley without rubbing the flanges.
Safety and Compliance
⚠️ Regulatory Notice: The build and modification recommendations in this article should be followed with attention to mechanical safety. Over-tensioned belts place radial load on stepper motor bearings and idler bearings beyond their design specifications. Bearing failure can result in belt derailment, loss of position, and potential print head crash. Always follow manufacturer torque specifications for belt tension and inspect bearings during routine maintenance. Fire safety: a seized stepper motor can overheat the driver — ensure thermal runaway protection is enabled in firmware.
Belt tension is a maintenance parameter, not a set-it-and-forget-it setting. Check belt frequency monthly. GT2 belts relax slightly over their first 50-100 hours of use as the rubber matrix settles — re-tension after the break-in period, then monitor every 3-6 months.
Related Resources
Once belt tension is dialed in, our input shaping guide covers the firmware-based vibration compensation that eliminates residual ringing at higher print speeds. Belt tension and input shaping work together — input shaping can compensate for some ringing, but it can’t fix backlash from a loose belt. For the full print-quality stack, check our first layer calibration guide — belt tension, bed level, and first-layer Z-offset are the three fundamentals that determine print quality before any advanced tuning.
For belt replacement, genuine Gates GT2 belts are the only choice that guarantees consistent tooth pitch. The generic “GT2” belts included with budget printers often have 0.02-0.05mm tooth pitch variation that produces a subtle herringbone pattern on vertical walls — visible under glancing light and impossible to eliminate through tuning alone.