Layer shifting shows up as a sudden horizontal step in the walls of a print — everything above a certain height is offset from everything below it, usually along the X or Y axis. The mechanism is always the same: the stepper motor’s rotation stopped translating 1:1 into head (or bed) movement. Either the motor lost steps, or the mechanical link between motor and toolhead slipped. Everything else — belts, pulleys, acceleration, drivers, collisions — is just a route to one of those two outcomes.
This guide covers the full set of layer shifting causes and fix procedures, in the order you should actually check them, with real numbers where the physics gives us real numbers. Layer shifting is not the same as ghosting or ringing — those are resonance artefacts (faint echoes near corners), not a hard positional offset. Plenty of guides conflate the two; keep them separate or you’ll chase the wrong fix.
Start here: is it mechanical slip or missed steps?
Before touching a single setting, work out which of the two failure modes you have. It saves hours.
- Large, sudden shift (several mm) that happens once or twice — usually mechanical: a loose pulley grub screw, a belt that jumped, or a physical knock/collision.
- Repeated, smaller shifts that creep in over a long print, often on the same axis — usually a loose belt, rising friction, or a stepper driver overheating and losing torque.
- Shifts clustered at sharp corners or fast direction changes — almost always over-aggressive acceleration/jerk making the stepper skip.
If you’re not sure what you’re looking at, our Diagnose tool reads a photo of the failure and tells you which of these it is, with the exact setting to change. The step-by-step logic is also in how to diagnose a failed print from a photo.
Cause 1 — Loose or wrongly tensioned belts
This is the most common mechanical cause. A belt that has stretched or slackened introduces play, so the pulley turns a fraction before the head actually moves. It’s especially common on the Y axis of bed-slinger machines, because a heavy bed flinging back and forth is exactly the load that works a belt loose. Prusa explicitly warns that over long print times a belt can slightly loosen and produce layer shifting or degraded quality — so re-check tension periodically, not just once.
Over-tightening is real but far rarer. A belt that’s too tight causes binding and accelerates bearing wear, and can itself contribute to ringing. Loose is the problem you’ll meet ten times more often.
Belt tension: use numbers, not the “guitar twang”
The popular “pluck it and listen for a low bass note” test is subjective and unreliable. Do it properly with a free phone spectrum analyser (Spectroid on Android, SpectrumView on iOS): pluck the longest free span between motor pulley and idler, and read the frequency peak.
The trap is that the target frequency depends on the free-span length, so a single “correct Hz” number copied from a forum is worthless unless it’s for your exact machine at your exact carriage position. The governing physics is:
f = (1 / (2 × L)) × √(T / μ) — where L is the free span in metres, T is tension in newtons, and μ is the belt’s linear mass density (0.0083 kg/m for standard 6 mm GT2).
Because span changes as the carriage or bed moves, so does the reading. Here are the verified figures — treat them as span-specific, not universal:
| Machine / span | Free span | Target frequency |
|---|---|---|
| Ender 3, carriage at limit switch | ~302 mm | ~94 Hz |
| Ender 3, bed all the way back | ~253 mm | ~113 Hz |
| Generic 200 mm span guide | ~200 mm | ~80–100 Hz |
| Alt. Ender-3 class guide | 200–250 mm | ~40–60 Hz |
| CoreXY A/B belts | 600–900 mm | 110–140 Hz (below 90 Hz = loose) |
Note how sources disagree even for the same class of printer — that’s exactly why you use the manufacturer’s spec for your machine rather than a borrowed number. As a manufacturer reference point, Gates recommends 6–8 lb of tension for a 6 mm GT2 belt used for registration as in a printer. Worth knowing where that lands mechanically: a NEMA 17 tolerates about 6.7 lb (30 N) of radial force 17 mm from its face, and because the belt wraps the shaft the radial force is roughly double the belt tension — so 6 lb of tension is already ~12 lb on the shaft. That’s why you don’t just crank it.
CoreXY: both belts must match
CoreXY kinematics sum the motion of two belts, so mismatched tension skews geometry. If belt A reads 140 Hz and belt B reads 110 Hz, the gantry pulls harder along one diagonal and a square prints as a parallelogram. Also measure the upper and lower runs of each belt — they should match within about 5 Hz; a difference over 10 Hz points to a binding idler bearing or a belt not tracking straight. Check tension once the frame has warmed to operating temperature, as it drifts slightly with heat.
Cause 2 — Loose pulley grub screw
The pulleys couple the steppers to the belts via one or two grub (set) screws. If a grub screw is loose, the motor shaft can spin while the pulley stays put — or slips intermittently — producing sudden, large shifts. Always check the set screws are tight, and make sure at least one seats squarely on the flat of the motor shaft. This is the single most common cause of a big one-off jump, and it takes thirty seconds to rule out with an Allen key.
Cause 3 — Missed steps from over-aggressive motion
A stepper misses steps when the motion system asks for more speed or acceleration than it can physically deliver — the motor commutates but the mass doesn’t keep up. Acceleration is the key driver, because acceleration is what forces a rapid change of direction; jerk (the instantaneous velocity change at a corner) is the sharp end of the same problem. On hobby FDM machines this is the most common slicer-side cause of shifting, and it clusters at sharp corners.
Dial the motion back to sane starting points, then work up:
| Parameter | Conservative starting point | Notes |
|---|---|---|
| Print speed | 40–60 mm/s | Drop to 40–50 mm/s if shifts persist |
| Print acceleration | 500–1000 mm/s² | Apply to walls first |
| Travel acceleration | 1500–3000 mm/s² | Reduces snatch on rapid moves |
| Jerk (X/Y) | 5–10 mm/s | Cura default is 8; don’t exceed ~20 on hobby machines |
A smart compromise if you don’t want to lose print time: apply the low 500–1000 mm/s² acceleration to the outer wall only. Most of a print’s time is infill, which can stay fast, but it’s the outer-wall corner overshoot that skips steps. You kill the failure without a big time penalty.
Alignment matters: slicer motion values must sit at or below your firmware’s caps, or the firmware silently clamps them and your carefully chosen numbers do nothing. Re-slice and confirm the two agree. If you want the tool to pick the specific value for your symptom rather than a table, that’s exactly what this troubleshooting workflow is built for.
Cause 4 — Nozzle colliding with the print
If the nozzle physically strikes the model, it can knock the head off position and skip a stepper. The usual triggers are warped corners, a curled edge, or a blob of filament the head catches on the return pass. Polymaker have documented a shift caused purely by the nozzle hitting raised infill and skipping the motor.
This is a material problem as much as a motion one: ABS, ASA, PA, PC and composites lift at the corners as they cool, creating raised edges the passing nozzle clips, which accumulates into incremental X/Y displacement. Fix the warping — enclosure, bed adhesion, first-layer squish — and the “shift” disappears. Our guides on first layer adhesion problems and why prints fail cover the root causes.
Cause 5 — Friction, binding and lubrication
Dust, debris or dry rails add resistance; the motor works harder and eventually skips. Prusa’s official checks are the right routine: confirm nothing obstructs the extruder or heatbed path — a common one is a scrap of filament wrapped around the Y-axis pulley from a previous failure — and check the smooth rods for deep scratches. Clean rods with a paper towel before lubricating.
Lubricant choice is machine-specific, so follow the maker’s spec rather than a generic recommendation. Prusa recommend Prusa Lubricant (a lithium-based general-purpose grease also works) for smooth rods — but explicitly say do not use it on the linear rail of the CORE One, which needs a dedicated rail grease such as Microlube GL 261. “Any lube anywhere” is how you gum up a linear rail.
Cause 6 — Stepper driver current (VREF)
Set VREF too low and the motor doesn’t get enough current, loses torque and skips. Set it too high and the motor and driver run hot — and excessive heat itself reduces torque, so shifts can appear midway through a long print as things warm up. A stepper too hot to comfortably touch, together with recurring shifts, is a strong tell that VREF is off. Adjust it to the value in your printer’s manual, verify with a multimeter, and monitor motor temperature afterwards. Make sure the drivers have adequate cooling, particularly in an enclosure.
Cause 7 — Cable snags and physical knocks
Poor cable management lets a loom snag mid-move. On Prusa machines, a badly positioned cable bundle or zip tie can let the cables hit the frame before the extruder assembly reaches its intended position, so the printer reads an inaccurate end point. Route and secure cables so nothing fouls across the full travel.
Finally, the boring one that’s genuinely common: knocks and vibration. Bumping the machine, an unstable bench, or a vibrating appliance nearby can shove the head. Put the printer on a solid, level surface out of the traffic path, and add anti-vibration feet if the bench is lively.
Frequently asked questions
How do I know if it’s layer shifting or ghosting?
Layer shifting is a hard, permanent offset — everything above a point is displaced sideways. Ghosting/ringing is a faint repeating echo near corners with no positional offset. If you can run a fingernail across a clear step in the wall, it’s a shift; if it’s a surface ripple only, it’s resonance.
What belt frequency should I tune to?
There’s no universal number — it depends on your free-span length. For a ~302 mm Ender 3 span, ~94 Hz is a documented target; CoreXY A/B belts sit around 110–140 Hz. Use a spectrum analyser app, pluck the longest span, and match the figure your printer manufacturer publishes for that span.
Can slowing the print down fix layer shifting on its own?
Often, yes — if the cause is missed steps from aggressive motion. Drop print speed to 40–50 mm/s, acceleration to 500–1000 mm/s² and jerk to 5–10 mm/s. If shifts continue at slow speed, the cause is mechanical (belts, pulleys, friction) and no speed change will fix it.
Why does the shift only appear late in a long print?
Two usual suspects: a belt that loosens over hours of running, or a stepper driver overheating and losing torque as the enclosure warms. Both get worse with time and load, which is why the failure creeps in rather than appearing on layer one.