How to Adjust Backlash on CNC Machine Axes

• 1
  Key takeaways
• 2
  What backlash is and where the lost motion comes from
• 3
  Measure backlash so you know which part to adjust
• 4
  Ball screw, leadscrew, or rack and pinion: which adjustment applies
• 5
  Common mistakes and the limits of DIY backlash adjustment
• 6
  Verify the adjustment with a test cut
• 7
  Step by step: how to adjust backlash on CNC machine axes
• 8
  Match the symptom to the likely source

Backlash is the motion an axis loses when it reverses direction before the table starts moving again. Command a move to the right, stop, then command a move to the left. The servo turns, the screw turns, but the table sits still for a few thousandths of a millimeter. That dead band is backlash, and it shows up as a step in the part profile at every direction change.

The error does not come from one place. On a typical vertical mill you have four contributors in series: the motor-to-screw coupler, the thrust bearings that locate the screw axially, the ball nut or leadscrew nut, and the linear guides or dovetail gibs. Each one adds its own play. If you tighten the nut while the thrust bearing is the real problem, the axis will still lose motion and you will have added drag for nothing.

The size of the problem matters. A ball screw with a light preload might hold 5–15 μm of lost motion when new. A worn leadscrew on a manual-conversion mill can reach 0.10–0.20 mm. On a machine cutting to ±0.005 mm, even 0.02 mm of backlash will fail a bore or a slot, because the cutter deflects differently on climb and conventional passes. Measure, then decide whether you adjust or replace.

• 1
  Thermal growthA 2 m screw running 15 °C above ambient grows roughly 0.36 mm. That is not backlash, but it hides in the same measurement if you test cold.
• 2
  Servo tuningAn under-tuned axis can overshoot and settle, which looks like backlash on a scope but is not mechanical.
• 3
  Cutter deflectionTool push-off mimics axis error in the part but disappears when you measure the table with an indicator.

Mount a dial indicator with 0.001 mm resolution on the table, with the stem against a solid surface such as the spindle nose or a vise jaw bolted to the table. Jog the axis 0.100 mm away from the indicator, then jog back toward it in 0.010 mm increments until the needle just moves. Write down the commanded distance. Do this five times and average. That average is your total lost motion at the table.

Now move the indicator to the screw end. On a mill with a visible screw, put the stem against the end face of the screw or a coupling flange. Repeat the same reversal test. This reading contains only the thrust bearing, locknut, and coupler play. Subtract it from the table reading. What is left belongs to the nut and the linear guides.

Run the test at three positions along the travel: near the motor, mid-travel, and near the far bearing. Backlash that changes with position points to a worn screw or a sagging screw, not a loose nut. If the number is the same everywhere, the cause is at one end and is usually simpler to fix. Log the numbers before and after any adjustment so the next tech has a baseline.

• 1
  Warm the machineRun the axis for 15–20 minutes at normal feed before measuring. Cold readings run 10–20 μm high on long screws.
• 2
  Push in one directionAlways approach the indicator from the same side. Reversing the approach adds indicator hysteresis to your number.
• 3
  Check the servo loadNote the axis load meter during the test. A jump after adjustment means you set preload too tight.

Ball screws use recirculating balls and need a small preload to remove internal clearance. Most come as double-nut or oversized-ball designs. You adjust them with a shim pack or a threaded collar between the two nuts. Turn it in small steps, because ball nut preload is measured in micrometers and a quarter turn can add 40 μm of drag.

Leadscrews, common on older mills, lathes, and hobby conversions, use a split bronze nut with a take-up screw. This type tolerates larger backlash and is meant to be adjusted. Set the nut until the axis turns with light drag along the whole travel. If you need more than about half the available take-up range, the nut is worn and a new one is cheaper than the downtime.

Rack and pinion axes on routers and plasma tables adjust differently. The pinion-to-rack mesh should have 0.05–0.10 mm of clearance, set with an eccentric bearing housing or slotted motor mount. Too tight and the pinion wears a wave into the rack. Check mesh at three points along the rack, because a bowed rack changes the gap. Never set rack mesh to zero.

• 1
  Ball screwPreload 2–5 percent of dynamic load. Measure in micrometers, not by feel.
• 2
  LeadscrewSplit nut take-up. Expect 0.02–0.05 mm residual and accept it.
• 3
  Rack and pinionKeep 0.05–0.10 mm mesh clearance. Zero clearance destroys the rack.

The most common mistake is chasing zero. A ball nut with zero clearance has no oil film and will run hot. Within an hour the screw is at 50 °C and the preload has doubled from thermal expansion. Set preload to a measured lost motion of 0.005–0.010 mm and leave it there. If the part still fails, the problem is elsewhere.

The second mistake is adjusting the wrong end. Techs loosen the nut locknut, turn the collar, and never check the thrust bearing. The bearing was contributing 0.030 mm of the 0.040 mm total. After the nut adjustment the reading barely moves and the machine is now over-preloaded. Always split the measurement first.

Know when to stop. If the screw has visible wear tracks, if the nut needs more than half its take-up range, or if lost motion exceeds 0.050 mm at mid-travel, adjustment will not hold. A worn screw also changes pitch, which no preload setting can correct. On machines cutting to ±0.005 mm, replace the screw and nut as a pair. Backlash compensation in the control can help a worn machine finish a job, but it does not fix the mechanical error and it drifts with temperature.

• 1
  Do not bottom out the take-upA split nut with no gap left has no adjustment range and will seize when warm.
• 2
  Do not skip the test cutA circular pocket reveals axis reversal error that a single-axis indicator check can miss.
• 3
  Do not compensate a worn screwPitch error changes along the travel. Compensation tables go stale as wear continues.

An indicator tells you the table has stopped losing motion. It does not tell you the part will pass. Run a circular interpolation test, 50–80 mm diameter, in aluminium at 0.05 mm depth of cut. A machine with residual backlash shows a flat or a step at the four quadrant points where each axis reverses. Measure the roundness with a micrometer or on a CMM.

Next, cut a slot in both directions and measure the width. If the climb-milled side and the conventional side differ by more than 0.010 mm, either backlash remains or the cutter is deflecting. Check the table with the indicator again before blaming the tool. On a good setup, both sides of the slot come out within 0.005–0.008 mm.

Finish with a repeat run after the machine has been running for 30 minutes. Screws grow with temperature, and a setting that felt right cold can add drag warm. If axis load has climbed and the surface finish has dulled, back the preload off one step. Log everything: date, axis, measured values, parts used, and the test cut result.

• 1
  Circular test50–80 mm diameter, 0.05 mm depth, look for quadrant steps.
• 2
  Bidirectional slotCompare both walls. More than 0.010 mm difference means residual error.
• 3
  Warm re-checkRepeat after 30 minutes of running and compare axis load.