Seven Ways to Detect the Positioning Accuracy of CNC Machine Tools

• 1
  Key takeaways
• 2
  What the positioning accuracy of CNC machine tools really measures
• 3
  Dial indicator and granite square checks
• 4
  Laser interferometer and ballbar testing
• 5
  Test cuts, spindle probes, and roundness checks
• 6
  Step by step: running the seven checks in order
• 7
  Which test to use for which problem

Positioning accuracy is the difference between the coordinate the control commands and the position the axis actually reaches. On a linear axis this is a distance error, reported in micrometers or in parts of a millimeter per meter of travel. It is not the same as repeatability, which describes how tightly the axis returns to the same point. A machine can repeat to 2 μm while sitting 25 μm away from the commanded position.

The error comes from several places at once. Ballscrew pitch error, thermal growth, servo lag, guideway straightness, and coupling wind-up all add up along the travel. Because of that, a single number for the whole machine is misleading. Accuracy changes with axis direction, position along the stroke, and how long the machine has been running.

For most production work the useful figure is the working tolerance, meaning the total error budget the process can absorb. If a drawing calls for ±0.02 mm and the machine drifts 15 μm over a warm-up cycle, the process is already at risk before any tool wear is counted. That is why we measure rather than assume.

The seven methods below run from quick checks any operator can do to full laser interferometer calibration. You do not need all seven on every machine. Match the method to the tolerance you have to hold and to how much downtime you can accept.

• 1
  Positioning errorCommanded point versus actual point, measured along the axis.
• 2
  RepeatabilitySpread of results when the axis returns to the same point.
• 3
  BacklashLost motion when the axis reverses direction.
• 4
  Thermal driftSlow position change as the structure warms during running.

The dial indicator test is the fastest way to catch gross errors. Mount a 0.001 mm resolution indicator on the spindle nose, bring the tip against a granite square or a known reference block, and command a series of moves of 10 mm, 50 mm, and 100 mm. Record the reading at each stop and compare it with the commanded distance. A consistent offset points to calibration; a random spread points to mechanical looseness.

Keep the indicator stem perpendicular to the surface. A 10° tilt on a 0.01 mm reading already introduces about 0.15 μm of cosine error, which sounds small until you are chasing 5 μm. Use a magnetic base with a fine adjustment, and let the spindle sit still for 30 seconds before taking the first reading so the servo settles.

The granite square adds squareness information that a linear check cannot give. Place the square on the table, indicate one face parallel to X, then sweep the perpendicular face while moving Y. Any deviation over a 300 mm sweep is the squareness error between the two axes. On a well-set machine this should stay inside 0.01 mm over 300 mm.

Both checks need a warmed machine. Run the spindle at a moderate speed for 20–30 minutes, or run the axis warm-up program the builder supplied. Cold readings on a machine that has been idle overnight can be 10–20 μm away from steady-state values, and that difference will follow you into production.

These two methods will not find servo mismatch or small cyclic errors from the ballscrew. For that you need the tests in the next sections. Use them as a first filter, not as final proof.

• 1
  Indicator resolution0.001 mm or finer for any check below 20 μm.
• 2
  Move sizesStep through 10 mm, 50 mm, and 100 mm increments.
• 3
  Squareness targetUnder 0.01 mm deviation over a 300 mm sweep.
• 4
  Warm-up20–30 minutes of spindle and axis motion before reading.

A laser interferometer measures linear position directly against the wavelength of light. The optics mount on the table and the spindle, and the system records position every few millimeters along the full stroke. This gives you the complete error map: pitch error, reversal error, and linear positioning deviation, all in one pass. It is the reference method for machine acceptance and for annual calibration.

Set up matters more than the instrument. Align the beam to within 1 mm over the full travel, keep the optics clean, and compensate for air temperature, pressure, and humidity if your system supports it. Uncompensated air changes shift readings by roughly 1 ppm per degree Celsius, which is 1 μm per meter. On a 1,000 mm axis that is enough to matter.

The ballbar test tells a different story. It commands a circular path, usually 100–300 mm in diameter, at a feed rate matched to the machine's normal cutting speed, and records the radial deviation. The resulting polar plot shows squareness error as an oval, servo mismatch as a diagonal lobe, and backlash as a step where the axis reverses.

Read the plot by shape, not just by the peak number. A round plot with one small bump usually means a local guideway issue. A clean oval means squareness. A plot that changes shape with feed rate points to servo tuning or to the acceleration limits in the control.

Both tests need the machine warmed and the work area clear of chips and coolant. Run them at the same time of day if you are building a trend, because shop temperature swings will show up in the data.

• 1
  Laser resolutionTypically 0.001 μm with environmental compensation on.
• 2
  Environmental effectAbout 1 ppm per °C without compensation.
• 3
  Ballbar diameter100–300 mm, run at production feed rates.
• 4
  Best useAcceptance testing, annual calibration, and trend tracking.

A test cut measures the machine under load, which no optical instrument can do. Cut a stepped block or a circular pocket in aluminium 6061 with a sharp, known tool, then measure the result on a coordinate measuring machine. The difference between the programmed geometry and the measured geometry is the machine's working accuracy for that operation.

Keep the variables tight. Use the same tool, the same coolant, and the same feed and speed for every repeat. A 12 mm three-flute carbide end mill at 8,000 rpm and 1,500 mm/min in aluminium is a reasonable starting point. Measure the part after it cools to room temperature, not straight off the table.

A spindle probe or touch probe gives a fast in-process check. Probing a certified ring gauge or a set of gauge blocks tells you how the machine reads a known feature. If the probe reads 0.008 mm small on a certified gauge, that offset applies to every part you measure with it. Recheck after any crash or tool change.

Roundness and circularity checks close the loop. Cut a circular pocket or turn a shaft, then measure roundness on a roundness tester or a CMM. Errors that appear as a two-lobe or four-lobe pattern usually trace back to spindle runout or to guideway preload, while a consistent spiral points to interpolation or servo tuning.

Run the test cut at the start of a production run and again after a few hours. If the numbers move, the machine is drifting, and the drift is what will put parts out of tolerance at the end of the shift.

• 1
  Test materialAluminium 6061 keeps cutting forces low and repeatable.
• 2
  Reference tool12 mm three-flute carbide, 8,000 rpm, 1,500 mm/min.
• 3
  Measure coolLet the part reach room temperature before CMM work.
• 4
  Probe offsetApply the same correction to every part measured.