How to Check CNC Machine Accuracy

• 1
  Key takeaways
• 2
  Why accuracy checks matter more than the spec sheet
• 3
  Spindle and axis alignment with dial indicator and test bar
• 4
  Ball bar testing for circular motion accuracy
• 5
  Laser interferometry for linear positioning and repeatability
• 6
  Tool setter, probe, and thermal checks
• 7
  CMM validation on the finished part
• 8
  Step by step: how to check CNC machine accuracy
• 9
  Accuracy checks compared

A CNC machine's brochure accuracy is measured under lab conditions: no chips, no thermal load, no operator. After six months of production, those numbers drift. The spindle grows when it warms up, the ballscrew wears unevenly, and a small crash can knock the geometry out by 0.01 mm without any alarm going off. The machine still cuts parts, but the dimensional spread widens.

For anyone sourcing machined parts, the question is not whether a supplier claims ±0.005 mm. It is whether they can prove the machine still holds it. That proof comes from scheduled checks: quick in-house tests every week, ball bar and laser tests every quarter, and part-level CMM reports when the job calls for it.

This article covers how to check CNC machine accuracy in the order a shop actually does it. Start with the spindle and basic geometry, move to circular and linear motion, then validate on a finished part. Each test has a clear pass/fail threshold and a specific failure mode it catches.

• 1
  WeeklyDial indicator on spindle taper, test bar in Z, tool setter repeat check
• 2
  QuarterlyBall bar circular test, laser interferometry on all linear axes
• 3
  After any crashFull geometry check before the machine goes back into production

The cheapest accuracy check is also the most revealing. Mount a 0.002 mm resolution dial indicator on the table and sweep the spindle taper. On a 40-taper spindle, runout at the gauge line should stay under 0.005 mm. At 300 mm below the gauge line, TIR should not exceed 0.010 mm. If it does, the spindle taper has wear, debris, or a bent tool holder.

Next, put a 300 mm test bar in the spindle and sweep it with the indicator mounted on the table. Rotate the spindle by hand and read the bar at two heights. The difference between top and bottom readings tells you the spindle is out of square with the Z axis. A 0.01 mm difference over 300 mm is roughly 0.002 degrees of tilt, which will show up as taper on deep bores.

Check the table surface as well. Sweep the full X and Y travel with the indicator on a magnetic base and a granite parallel. A twist of more than 0.02 mm across 500 mm will make flat parts rock, and no amount of tool offset will fix it.

One common mistake: checking the spindle when it is cold. Run the spindle at 8,000 rpm for 20 minutes first, then re-check. Thermal growth of 0.01–0.02 mm in Z is normal on a 40-taper machine, and you want to measure the machine in the state it runs.

A ball bar is a telescoping bar with a precision ball at each end. One ball sits in the spindle, the other on the table. The machine interpolates a circle, and the bar length changes as it follows the true path. Software plots the deviation in polar coordinates. For a 300 mm radius circle at 1,000 mm/min, a healthy VMC should stay within 5–10 μm of roundness.

The plot shape tells you what is wrong. A single lobe pointing in one direction usually means axis squareness is off. A flat spot at the quadrant changes points to backlash or reversal spikes. A }scalloped} circle with many small bumps suggests servo mismatch between axes, or a loose encoder coupling.

Run the test at two feed rates: 1,000 mm/min and 2,000 mm/min. If roundness error grows sharply with feed, the problem is dynamic: servo tuning, belt tension, or way lubrication. If the error stays constant, it is geometric: squareness, straightness, or backlash.

Ball bar tests are fast and repeatable. A trained operator can set up and run one in 15 minutes. Most shops run them quarterly, or after any spindle or axis repair. Do not skip the warm-up cycle; a cold machine will show a different circle than a warm one.

Laser interferometry measures the actual distance an axis travels against the wavelength of light. It is the reference method for positioning accuracy, repeatability, and straightness. A typical test on a 500 mm axis will reveal positioning errors of 5–20 μm even on a well-maintained machine. The machine's backlash compensation may hide this in normal cutting, but the laser exposes it.

The test also captures pitch, yaw, and roll: angular errors that tilt the axis as it moves. A 0.005 degree yaw over 500 mm translates to a 0.04 mm lateral error at the tool. If you are cutting a long, straight feature and it comes out bowed, angular error is often the cause.

Repeatability matters more than absolute accuracy for most production. If the machine returns to the same position within 2–3 μm over 20 cycles, it can hold tight tolerances with a small offset. If repeatability is poor, no offset will save you.

Laser tests are usually done by an outside service or a metrology team. Schedule them quarterly, and always after a crash, a ballscrew replacement, or a move. The report should list positioning error, repeatability, and angular error for each axis, plus the ambient temperature and the compensation values applied.

Tool setters and spindle probes drift. A tool setter that reads 0.01 mm high will make every tool cut 0.01 mm deeper, and the operator will chase the offset all day. Check it with a gauge block or a known-length tool. Touch off the same tool ten times and look at the spread. A repeatability of ±0.002 mm is good; anything above ±0.005 mm needs cleaning or replacement.

Spindle probes need a similar check. Touch a ring gauge or a known bore five times and compare the average to the certified size. If the probe is out by more than 0.005 mm, check the stylus for runout and the probe head for debris. A bent stylus is the most common cause.

Thermal drift is the quiet killer. Log the shop temperature, spindle temperature, and coolant temperature over a full shift. A 3 °C rise in ambient can push a 500 mm aluminium part by 0.02 mm. The machine may be perfectly accurate at 8 a.m. and out of tolerance by 2 p.m. If your parts drift through the day, thermal compensation or a temperature-controlled room is the fix.

Machine checks confirm the machine is healthy. Part checks confirm the process is capable. A CMM measures the actual geometry of a finished part: hole positions, bore diameters, flatness, and profile. Run a first-article inspection on a new job, then sample every 20–50 parts depending on the tolerance band.

For a part with a ±0.005 mm tolerance, the CMM itself should be at least four times more accurate: a ±0.001 mm volumetric accuracy is typical for a good bridge CMM in a temperature-controlled room. If the CMM is in the same shop as the machines, run a master artifact before and after each session to confirm it has not drifted.

Compare CMM results to the machine's own probe data. A 0.01 mm gap between the two is a warning sign. It could be thermal, it could be probe calibration, or it could be that the part moved during fixturing. Do not assume the machine is wrong; investigate both sides.

For production runs, keep a simple trend chart of the key dimension. If the average shifts by 0.005 mm over 100 parts, the process is drifting and needs adjustment before it goes out of tolerance.