Troubleshooting guide

How to Troubleshoot Part Off Location in CNC Machine

A working procedure for when the parted-off face lands off the drawing, or the cutoff length drifts across a run. Written for setup techs and process engineers who need to separate programming error from mechanical wear before touching anything.

Get Instant Quote Talk to an Engineer

Tolerances to ±0.005 mmMill-turn and Swiss-type1-off to 10,000+ parts

how to troubleshoot part off location in cnc machine

On this page

1
Part off location errors: symptom, likely cause, and first move
2
Program and tool path first
3
Parting tool condition and setup
4
Machine stability and mechanical wear
5
Cutting parameters and coolant delivery
6
Controller, feedback and calibration
7
Step by step: isolate the cause

Start here

Part off location errors: symptom, likely cause, and first move

Match the symptom you see before opening the machine panel.

Symptom	Likely cause	First move
Length drifts down across the run	Thermal growth in spindle and ballscrew	Run 10 warm-up cycles, then cut a test bar
Part-off face sits off center	Parting tool center height wrong	Set tip 0.02–0.05 mm above centerline
Face is square but 0.1 mm short	Z-axis tool offset stale	Re-touch off the insert corner in Z
Face tapers across the diameter	Tool tilted or turret out of alignment	Clock the blade with a dial test indicator
Nub or pip left on the face	Insufficient feed near the end of cut	Reduce feed 30–50% in the last 1 mm
Bore and OD not concentric	Sub-spindle pick-off misaligned	Indicate the sub-spindle to 0.01 mm TIR
Length varies part to part	Bar puller or bar feed slipping	Check pull force and stop position
Face chatters at the cut	Weak setup or worn insert	Shorten overhang, change insert grade

Fix the geometry before you touch the offsets

If the face is not square, an offset change will hide the symptom and keep making scrap. Correct center height, tool tilt and turret repeatability first; only then chase the last few microns with Z offsets.

Get Instant Quote Talk to an Engineer

Check 1

Program and tool path first

Most part off location errors trace back to the program or the offsets it depends on, not to the machine casting. Before you pull a wrench, read the part-off block and confirm the Z value matches the drawing. A stale insert corner radius in the offset will shift the face by the radius difference alone.

Check the workpiece zero. A zero point set from a rough face instead of the finished face moves every downstream feature. We re-confirm zero with a touch probe and hold it to ±0.001 mm on setups that matter.

Look at the cutoff length in the program versus the stock you loaded. Bar feed stop position, pull length and the programmed Z all stack up. If the bar stop is 0.05 mm off, every part in the run is 0.05 mm off.

Finally, confirm the part-off tool number matches the offset you edited. On a 16-tool turret it is easy to touch off station 7 and call station 9.

1
Re-check Z zeroTouch off on the finished face, not the saw cut.
2
Verify tool numberMatch the offset to the station actually in cut.
3
Confirm bar stopMeasure the pull length with a caliper, not by eye.

Check 2

Parting tool condition and setup

The parting tool is the least rigid tool in the turret, so small errors show up there first. Center height is the usual culprit. Set the insert tip 0.02–0.05 mm above the centerline. Too low and the tool pushes off, leaving a pip. Too high and the insert rubs, wears the flank and pulls the face off square.

Inspect the blade and the insert seat. A chip trapped under the insert tilts it by a few tenths. That tilt shows up as a tapered face on a 30 mm diameter.

Check overhang. Every extra millimeter of blade stick-out costs you stiffness. On a Ø20 mm bar, keep overhang under 25 mm where the geometry allows.

Coolant should hit the cutting edge, not the shank. Aim the nozzle at the insert corner and confirm flow before the first cut. High-pressure through-tool coolant helps most on stainless and titanium, where the chip wants to weld to the insert.

1
Center heightTip 0.02–0.05 mm above centerline.
2
Insert seatClean the pocket; one chip is enough to tilt it.
3
Coolant aimNozzle at the insert corner, not the blade.

Check 3

Machine stability and mechanical wear

If the program and the tool are correct, the error is mechanical. Spindle bearing play, ballscrew backlash and turret repeatability all move the cutoff point. A quick test: cut a test bar, measure, index the turret 20 times, cut again, measure. The difference is your repeatability.

Check the Z-axis backlash. More than 0.01 mm on a production machine is worth attention. On a mill-turn or Swiss-type, the guide bushing and the sub-spindle add two more places for error to enter.

Listen to the cut. A change in pitch during the part-off usually means the insert is rubbing or the bar is moving in the collet.

Thermal growth is real. A spindle that has run for one hour is longer than a cold one. On a 200 mm Z travel, 10–20 μm of drift over a shift is normal. That is why warm-up cycles matter before you chase a 0.02 mm error.

1
Turret repeatabilityIndex 20 times and compare the cutoff length.
2
Z backlashAbove 0.01 mm, inspect the thrust bearing.
3
Warm-upRun 10 cycles before measuring a tight part.

Check 4

Cutting parameters and coolant delivery

Part off location errors often appear when the tool is not cutting, but pushing. Feed too low and the insert rubs, work-hardens the material and wanders. Feed too high and the blade deflects, so the face goes off square.

For 304 stainless on a Ø25 mm bar, a starting point is 0.05–0.08 mm/rev with surface speed around 80–120 m/min. For 6061 aluminium, run faster: 200–300 m/min with 0.08–0.12 mm/rev. Adjust from there, not from a chart alone.

Reduce feed 30–50% in the last 1 mm of the cut. That keeps the nub small without stalling the tool. If you still see a pip, the center height is wrong, not the feed.

Coolant pressure matters more than flow on deep cuts. Below 20 bar, chips tend to pack the groove and push the tool sideways.

One more thing: peck parting. On long overhangs, a peck cycle with a 0.5 mm retract clears the chip and holds the face straighter.

1
Feed window0.05–0.08 mm/rev on stainless, 0.08–0.12 on aluminium.
2
Last 1 mmDrop feed 30–50% to control the nub.
3
PressureKeep coolant above 20 bar on deep grooves.

Check 5

Controller, feedback and calibration

The controller is the last place to look, not the first. Servo drift, encoder count loss and a stale pitch-error compensation table all shift the cutoff point. If the machine has a scale on Z, compare the scale reading to the commanded position at two points 100 mm apart.

Check the tool offset library for duplicate entries. Two offsets pointing at the same station will fight each other. On a lathe with a sub-spindle, confirm the pick-off Z offset is separate from the main spindle offset.

Pitch-error compensation drifts as the machine wears. A laser interferometer check once a year is enough for most shops. If the error is under 0.01 mm over 300 mm, leave the table alone.

After any change to the controller, cut three test parts and measure all three. One good part is not a fix.

1
Scale vs commandCompare at two points 100 mm apart.
2
Offset libraryDelete duplicate entries for the same station.
3
Verify with three partsMeasure every one, not just the first.

Procedure

Step by step: isolate the cause

Run these in order. Stop as soon as the error disappears.

1
Measure the error properlyCut five parts, measure the cutoff length on each, and write down the spread. If the spread is under 0.02 mm, the problem is a shift, not a drift. If it grows across the run, it is thermal or mechanical.
2
Compare the face to the drawingCheck squareness with a dial indicator on the face. A face that is square but offset points to Z zero or a stale offset. A tapered face points to tool tilt or center height.
3
Re-touch the part-off toolSet center height 0.02–0.05 mm above centerline, clean the insert pocket, and re-measure the Z offset. Cut one part and re-measure.
4
Run the warm-up cycleRun 10 air passes or warm-up cycles, then cut a test bar and measure. Repeat after 30 minutes of production. If the length moves between the two, you have a thermal problem, not a programming one.
5
Check mechanical repeatabilityIndex the turret 20 times and cut. Compare to a single-index part. A difference above 0.02 mm means the turret or the ballscrew needs service.
6
Adjust parameters lastChange feed and speed only after the geometry is right. Reduce feed 30–50% in the final 1 mm and raise coolant pressure above 20 bar.

FAQs

Frequently asked questions

What causes the cutoff length to drift during a long run?

Thermal growth is the usual answer. The spindle, ballscrew and bed all grow as the machine warms up, and Z moves with them.

Run a warm-up cycle before the first part, then check the length again after 30 minutes of production. If the two do not match, add a warm-up routine to the start of the shift. Do not compensate with a tool offset mid-run unless the error is stable.

Can I fix part off location errors without stopping the machine?

Sometimes, but only for a known, stable shift. If the length is off by a fixed amount and the spread between parts is tight, a Z offset change can carry you to the end of the run.

If the spread is wide, stop. A drifting length means something is moving, and running on produces scrap. On a 10,000-part run, an hour of diagnosis is cheaper than a bin of rejects.

Which materials are more prone to part off problems?

Stainless 304, 316 and 17-4PH work-harden quickly, so a rubbing insert will wander. Titanium Ti-6Al-4V behaves the same way.

Aluminium 6061 cuts cleanly but galls if the coolant misses the edge. Copper and brass part off easily but leave burrs that can look like a location error on the face.

Does the parting tool holder matter?

Yes. A blade with more overhang deflects more, and deflection shows up as a tapered face.

Use the shortest blade that reaches the bar, keep the insert seat clean, and check the blade for bend after a crash. A bent blade will not hold center height no matter how carefully you set it.

How often should I check turret repeatability?

Once a quarter on a production machine, and after any crash.

The test takes 20 minutes: index the turret 20 times, cut a test bar, and compare the cutoff length to a single-index part. If the difference is above 0.02 mm, schedule service.

What should I do first if a part comes off with the wrong length?

Measure three more parts before changing anything. The pattern tells you where to look.

All four wrong by the same amount means an offset or program problem. A growing error means thermal or mechanical. Random lengths mean the bar is moving in the collet or the bar feeder is slipping.