CNC lost steps: why the cut shifted and how to fix it
Lost steps mean the motor was told to move and could not, or turned without moving the axis, so everything after that point is offset by the same amount. The cause is nearly always too much cutting load, an obstruction, or something loose in the drive. The shift direction, distance and timing tell you which.
Part way through a job the cut jumps to a new position and everything after that is offset by the same amount in the same direction. Often accompanied by a distinct clunk or grinding noise at the moment it happens.
first
Three questions that cut the list in half
Ask these before you change anything. All three are free, none of them require tools, and each one eliminates whole branches of the list below. People routinely skip them and spend a weekend on the wrong half.
- Did the controller say anything? Check the sender's console or the machine's own screen for an alarm, an error, a lost connection or a reset. A machine that kept cutting without complaining believed it was in the right place the whole time, which puts the fault between the motor and the cutter. A machine that alarmed or dropped its connection puts the fault between the computer and the controller instead, and sends you somewhere else entirely.
- Jog back to your work zero. Does the tool sit exactly on the corner you originally zeroed to? If it does, the machine never lost anything and your stock moved underneath it: go to workpiece moved, which is a different problem with different fixes. If it does not, the machine lost position and you are in the right place.
- Measure the offset. Which axis, and how far, to the nearest millimetre. A shift that lands on a multiple of your belt pitch points hard at a toothed belt skipping, because a belt can only ever slip in whole teeth. An arbitrary distance points at a stall or something turning on a shaft. Both axes shifting together points past the mechanics at the controller.
Most likely causes
- Cutting load exceeded what the motor could pushMost common
Too deep, too fast, or a dull bit. The motor stalls, misses steps, and carries on from the wrong place. Reduce depth per pass. - Something physically in the wayCommon
A clamp, a dust hose caught on the gantry, a cable snagging, or the tool hitting a screw. - Mechanical bindingCommon
Dirty rails, a tight spot in a lead screw, a V-wheel adjusted too tight, or debris in the drive. - Loose motor coupler or pulleyCommon
A grub screw backing off looks exactly like lost steps, because the motor turns and the axis does not. - A drive belt skipped teethCommon
Under shock load, or with a worn belt, tired tensioner or dusty pulley. Screw driven machines are not exempt: plenty of them turn the screw through a short toothed belt from the motor, and that belt skips exactly like a long one. - The workpiece moved, not the machineCommon
Identical symptom, unrelated cause. The jog-to-zero test above separates them in ten seconds. See workpiece moved. - Electrical noiseLess often than blamed
Worth knowing that Sienci and Carbide 3D both document this as over-attributed: noise usually stops a machine rather than silently moving it. It can still take your position indirectly, by falsely triggering a limit switch or resetting the controller. See electrical noise and EMI. - Motor current set too lowOccasional
Not enough torque for the cut. Check the driver settings against the motor's rating. - Motor overheating and losing torqueOccasional
Often from current set too high. A motor too hot to touch is a clue. - Acceleration set beyond what the machine can doOccasional
The machine cannot physically accelerate that hard, so it slips on rapid direction changes.
Diagnostic steps
- Note the axis and the distance of the shift before doing anything else.
- Check for obstructions, including the dust hose and cable routing through their full range of travel.
- Move each axis by hand with the power off. It should move smoothly with no tight spots.
- Check every coupler and pulley grub screw. Mark them with paint so you can see if one moves.
- Repeat the job with a much lighter cut. If it completes, cutting load was the cause.
- Watch when it happens. Always at the same point in the file means geometry or an obstruction. Random means load, or a drive part that slips under shock.
Fixes
- Reduce depth per pass and feed rate to bring the load inside what the machine can do.
- Clear obstructions and route hoses and cables so they cannot snag anywhere in travel.
- Clean and lubricate rails and screws as your machine's documentation specifies.
- Tighten couplers and check pulley grub screws.
- Separate spindle power cable from signal wiring, and ground shielding properly.
- Check motor current settings against the motor rating.
- Reduce acceleration if the machine loses position specifically on rapid direction changes.
When in the job it happened
Timing is evidence, and it is free. Run the same file again if you can bear to, and note whether it fails in the same place.
| When it goes wrong | What that points at |
|---|---|
| The same point in the file, every time | Something at that coordinate. A clamp in the path, a screw, a move that runs an axis into its own limit, or geometry that asks for a cut the machine cannot make. |
| Anywhere, early or late, no pattern | Cutting load, a loose pulley or coupler, or a belt skipping under shock. All of them look random, which is why the paint pen test below matters. |
| Late in a long job, almost never early | Something that accumulates. Driver and motor heat, chips packing into a deep pocket, a dust shoe filling up, a grub screw working its way loose, a belt warming and slackening. |
| Only on rapids or sharp direction changes | Acceleration or jerk set beyond what the machine can physically do. |
Why it happens
A stepper motor has no idea where it is. The controller sends a pulse, the motor is supposed to turn one step, and nobody checks. If the load on the motor exceeds its torque for even one pulse, the rotor does not make it to the next magnetic position, snaps back, and the controller carries on counting as if it had. The same thing from the other side: if the motor turns but a grub screw, a belt tooth or a coupler lets the axis stay put, the count is equally wrong. In both cases the machine is now confidently cutting a perfect part in the wrong place, and it will do so until you stop it. Closed loop steppers catch the first kind and not the second, because their encoder is on the motor rather than on the axis.
Prevention
- Paint pen every pulley, coupler and grub screw across the joint, once. A slip becomes visible at a glance.
- Keep cutting load inside what a removal rate test showed the machine can hold, with margin for a dull bit and a knot.
- Route the dust hose and cables, then jog every axis to both ends of travel and watch them.
- Fit homing switches if the machine lacks them, so a lost job is recoverable rather than scrap.
- Put belt tension, wheel adjustment and grub screws on a maintenance schedule instead of waiting for a symptom.
Recovering a job
If your machine has homing switches, re-home and your work offset is usually still valid, which means you can often restart from a sensible line rather than scrapping the piece. Without homing, the work zero is gone and the part is usually lost. This is the single best argument for adding limit switches to a machine that does not have them.