The bottom of the part is fine. At some height the whole thing steps sideways, and every layer above that keeps the offset — so the finished print looks as though someone slid the top half across.
Why the printer did not notice
Stepper motors are commanded, not measured. The controller sends a number of pulses, assumes the motor turned by that much, and updates its internal coordinates. Nothing on a normal desktop machine reads back where the toolhead actually is.
So when a motor meets more resistance than its torque can overcome, the rotor slips a pole or several. The pulses keep coming, the coordinates keep updating, and from that moment the machine's idea of the origin is wrong by the distance it lost. It cannot recover, because it does not know.
Read the shift before touching a setting
| What the print shows | What it points at |
|---|---|
| A single shift, in one axis, with a scar or knocked-off feature at that exact height | The nozzle hit something |
| Repeated shifts at unrelated heights, always the same axis | The drive on that axis: belt tension, or a pulley slipping on its motor shaft |
| Both axes displaced at the same instant | The toolhead was obstructed, or the machine lost and regained position |
| Shifts only late in long prints, never in short ones | Drivers or motors overheating |
| A shift immediately after the lights flickered | Power interruption — see layer shift after a power blip |
That table is the whole diagnosis for most cases, and it costs one careful look at the failed part.
The collision case, which is the common one
Something rose above the height the machine expected and the nozzle drove into it at travel speed. Usually that something is a corner that lifted off the plate, or the leading edge of an overhang that curled up as it cooled.
If that is what happened, the layer shift is not the fault — it is the consequence. Chasing motion settings will not help, because the machine will hit the same curl on the next attempt. Fix the warping or the curling overhang first, and see nozzle hitting the print for the earlier warning signs, which are audible before they are visible.
The mechanical case
Two components carry the drive and both can let go of it: the belt, if it is slack enough to jump a tooth under hard acceleration, and the pulley, if it is gripping the smooth part of the motor shaft instead of the flat machined into it. Either produces displacement without a collision and without a sound you would notice.
The distinguishing feature is randomness. Where a collision happens at a specific feature, a drive failure happens wherever the toolhead happens to be reversing hardest, so the heights are scattered across the print. Tensioning, tooth wear and the grub-screw check are all covered on belt slipping, which is worth working through in full once this branch is confirmed.
The thermal case
Stepper drivers pass current continuously, and motors hold position with current even when standing still. Drive them harder than they need and the surplus becomes heat; a hot motor produces less torque exactly when a long print is asking for the most. This is why some machines shift only after several hours, and why it is more common on enclosed printers whose electronics live inside the heated volume — see stepper motors running hot.
What to change, and how to know it worked
- Drop acceleration by about 30% and reprint the same file. Acceleration is what demands peak torque; speed alone demands much less. If the shift disappears, the cause was torque, not collision.
- Check both belts have the same tension by plucking them and listening. Two belts on the same axis should sound alike.
- Push the toolhead and the bed through their full travel with the steppers released. Resistance should be even throughout. A stiff patch is a binding rail, and binding steals the torque reserve the machine needs at peak acceleration.
- Confirm the electronics fan is spinning. A driver fan that has stopped fails exactly this way, late and inconsistently.
Two things not to do
Do not simply halve every speed and declare the machine fixed. A mechanical fault that shows up at high acceleration is still there at low acceleration; you have bought quiet time, and it will return on the first ambitious print.
Do not raise stepper driver current to "get more torque" without watching the motor temperature afterwards. Motors that are too hot to hold have already lost more torque than the extra current bought, and cooking a driver is an expensive way to learn that.
Which materials make this more likely
Not the material itself, but its warping behaviour. ABS, ASA and the polyamides lift corners and curl overhangs, and those are the features that get hit. A machine that never shifts on PLA and shifts every time on ABS does not have a motion problem — it has an adhesion problem that presents as one.