Thermal Deformation of CNC Turntable: Causes, Error Size, and Compensation Measurements

• 1
  Where the Heat Enters a Rotary Table
• 2
  How Thermal Deformation of CNC Turntable Becomes a Position Error
• 3
  Compensation Measurements That Survive a Production Day
• 4
  When Compensation Stops Working
• 5
  What the Part Designer Can Do About It
• 6
  Step by Step: Building a Thermal Compensation Map
• 7
  Error Size vs. Table Diameter and Part Reach
• 8
  Compensate or Change the Process?

A CNC turntable has three heat paths. The worm or roller-cam drive runs in an oil bath and loses 5–15% of motor power as heat. The bearing set, usually a crossed-roller or an axial-radial pair, generates heat proportional to preload and speed. The third path comes from outside: cutting heat conducts through the fixture, the part, and the table face itself.

The third path is the one people forget. On a 5-axis job where the part sits on a 300 mm fixture plate, 60–70% of the heat reaching the table can come from the cut, not the drive. That heat arrives slowly and leaves slowly, so the table keeps moving after the spindle stops.

Table size sets the thermal mass. A Ø400 mm table with a cast iron body may take 40–60 minutes to reach steady state at 12 rpm. A smaller Ø200 mm table can settle in 15–20 minutes. Neither is better. The larger table drifts less per degree but takes longer to stabilize, which matters when you are chasing a first-article check at 8 a.m.

Thermal deformation of CNC turntable shows up as three separate errors, and they do not scale the same way. First is angular position drift: the table face tilts as one side of the housing grows faster than the other. Second is radial and axial offset of the rotation center. Third is a change in drive preload, which changes backlash.

Angular drift is the expensive one. A 20 μm tilt across a Ø400 mm table face equals roughly 10 arc-seconds of angular error. Put a part 250 mm from the center and that 10 arc-seconds becomes about 12 μm of linear error at the tool tip. Now put a 400 mm long aerospace bracket on the same table and the same tilt moves the far end by nearly 20 μm.

Radial offset is usually smaller, in the 5–15 μm range for a well-built table, but it is harder to compensate because it changes sign as the table rotates. Backlash change is the quiet one. A 2–3 μm preload loss does not show on a single-point check; it shows as a size trend across a batch of parts.

• 1
  Angular driftTable face tilt, roughly 10 arc-seconds per 20 μm across Ø400 mm
• 2
  Center offsetRadial and axial shift of the rotation axis, 5–15 μm typical
• 3
  Preload lossBacklash grows, size trend appears across a run

You cannot compensate what you do not measure, and a cold machine tells you nothing useful. The measurement has to run at thermal steady state, which means the spindle and the table have both been turning for at least 45–60 minutes under a representative load.

The standard approach is a ball or sphere artifact mounted off-center on the table, probed at four or more index positions. On a 5-axis machine, run the same artifact through a rotary cycle and log the center coordinates. The drift between the first and last pass is your thermal signal. A 3 μm drift over 90 minutes is normal on a Ø400 mm table; 15 μm points to a preload or cooling problem.

Log the table housing temperature at the same time. A single thermocouple on the housing is enough to build a drift-per-degree coefficient. Once you have that number, you can predict the offset instead of chasing it. We log housing temperature on every 5-axis setup that holds tighter than ±0.01 mm.

Probing resolution matters more than probe accuracy here. A probe with 1 μm repeatability is fine even if its absolute accuracy is 3 μm, because you are looking at change, not absolute position.

• 1
  Run 45–60 min firstMeasure at steady state, never from cold
• 2
  Use an off-center artifactProbe at 4+ index positions, log center drift
• 3
  Log housing temperatureBuild a drift-per-degree coefficient for prediction

Thermal compensation has a ceiling. It works when the drift is repeatable and slow. It fails when the drift is non-repeatable, which usually means the heat source is outside the machine.

A shop door opening onto a 35 °C yard will swing the table housing by 3–5 °C in ten minutes. No coefficient table keeps up with that. The same applies to a cold-air blow-off aimed at the fixture, or a chiller that cycles on a 6 °C band.

The second limit is part geometry. Compensation assumes a rigid part. A thin-wall aluminum housing that grows 15 μm from its own cutting heat will not follow the table correction, because the part and the table are moving in different directions. For those jobs, the fix is process-side: lighter finishing passes, a coolant soak before the final pass, and a dwell of 3–5 minutes before the finishing cut.

The third limit is measurement time. If your probe cycle takes 4 minutes and the table drifts 2 μm in that window, you are compensating against a moving target. Keep the artifact close to the cutting zone and keep the cycle short.

Tolerances on features machined from different table positions stack. If a hole pattern is drilled at 0° and a bore is finished at 90°, any angular drift between the two positions lands directly in the relationship between them.

The practical move is to group tight-tolerance features into one table position. On a 5-axis part, that may mean re-fixturing once instead of indexing six times. Each index is another chance for the table to be at a different temperature than it was during setup.

Datums matter too. A datum on a surface machined in the same setup as the controlled features removes the table drift from the stack entirely. A datum on a cast surface does not, because the cast surface was never tied to the table position. When we review a drawing for 5-axis work, datum choice is the first thing we flag.

Material choice plays a smaller role than most people expect. Aluminum and steel parts both follow the table because the fixture dominates the thermal path. The exception is a long steel part on a small table, where the part itself can bend as it warms.