How to Improve the Positioning Accuracy of CNC Sliding Table During Processing
Positioning error on a sliding table rarely comes from one bad part. It builds up from squareness, preload, thermal drift and backlash. This guide shows the checks and settings we use to bring a table back inside ±0.005 mm, and when a rebuild is the only honest answer.
In this article
- 1
- 2
- 3
- 4
- 5
- 6
- 7
Seven things that actually move the number
What limits the positioning accuracy of CNC sliding table
A sliding table moves on linear guides and is pushed by a ball screw or a linear motor. Every link in that chain adds error. The guide rail sets straightness, the screw sets lead error, the nut and bearing set backlash, and the servo loop decides how fast the table settles at the commanded point. When a part comes out tapered or a bore lands off-center, the cause is almost always in one of those links.
Mechanical error and control error behave differently. Mechanical error repeats: run the same move ten times and the table lands in the same wrong spot. Control error scatters: the landing point wanders with speed, load and temperature. You separate them with a simple test. Command 50 mm of travel, measure with a dial indicator, then return and repeat. A consistent offset is mechanical. A spread of values is electrical or thermal.
The positioning accuracy of CNC sliding table also depends on the machine as a whole. A table that is square to itself can still cut crooked if the spindle is out of tram or the bed is twisted. Check the table against the spindle axis, not just against its own travel. That single habit prevents a lot of wasted adjustment time.
- 1MechanicalRail straightness, screw lead error, bearing and nut preload, coupling slip.
- 2ThermalScrew growth from friction, spindle heat, ambient swing across a shift.
- 3ControlServo gain, following error, encoder resolution, backlash compensation values.
- 4SetupFixture clamping force, part rigidity, chip packing under the table.
Squareness and parallelism: the first thing to verify
Squareness is the angle between the table travel and the spindle axis. If it is off, every facing pass cuts a dish and every side mill cuts a taper. On a typical vertical mill, we look for squareness within 0.01 mm over 300 mm of travel. A granite square and a dial indicator on a magnetic base get you there. Sweep the square along X, then along Y, and write both numbers down.
Parallelism is a separate check. Clamp a test bar in the spindle, sweep it across the table at the near end and the far end, and compare. A difference larger than 0.02 mm over 500 mm means the column or the bed needs shimming, not the table. People often blame the sliding table for an error that lives in the column.
Do this check cold, then again after a 30-minute warm-up. If the numbers move, you have a thermal problem, not a geometric one. That distinction changes the whole plan. Shimming a machine that drifts with heat just moves the error around.
Setting preload and backlash on the ball screw and guides
Backlash is lost motion when the axis reverses. On a ball screw, it comes from the nut, the thrust bearings or a loose coupling. Measure it by approaching a point from both directions with a dial indicator and noting the difference. For most machining centers we want backlash under 0.005 mm before compensation, and under 0.002 mm after. Anything above that points to a worn nut or a slipping coupling.
Preload on linear guides is a window. A light preload (around 2–3% of the dynamic load rating) suits fast, lightly loaded tables. A medium preload (5–8%) suits heavy cutting. Push past that and friction heat climbs, the screw grows, and the positioning accuracy of CNC sliding table gets worse, not better. If the rails run hot after an hour, the preload is too high.
Never fix backlash with compensation alone. Compensation hides the number but leaves the wear in place. Fix the nut, the bearings or the coupling first, then let the control apply a small residual value. A compensation table that keeps growing shift after shift is a wear signal, not a tuning problem.
Thermal drift and feedback: holding accuracy across a shift
A ball screw 1,000 mm long grows about 0.012 mm for every 1 °C of temperature rise. Run a machine hard for two hours and that is easily 3–5 °C, so 0.04–0.06 mm of position shift appears without anything being wrong. This is why the same program cuts a good part in the morning and a drifting part after lunch. It is not the operator and it is not the CAM file.
A warm-up routine fixes most of it. Run a programmed cycle for 20–30 minutes before the first critical cut, and keep the spindle and axes at a steady speed. In shops running ±0.005 mm work, we also log the coolant and ambient temperature, because a 2 °C swing in the room moves the part as much as the machine.
Feedback closes the loop. A linear scale reads the table directly, so screw growth no longer shows up in the part. A rotary encoder on the screw only sees the motor, so thermal growth passes straight through. If the job needs ±0.005 mm over a long travel, a linear scale is not optional. It is the cheapest way to hold the positioning accuracy of CNC sliding table across a full shift.
Step by step: bring the sliding table back into tolerance
Work through the list in sequence. Skipping a step makes the later numbers meaningless.
- 1Clean and log the baselineWipe chips and coolant from the rails and the screw. Record backlash, squareness and repeat positioning before you change any setting. Target: backlash under 0.005 mm, squareness within 0.01 mm over 300 mm.
- 2Check the coupling and thrust bearingsGrab the screw by hand and rock it. Any play at the coupling or the bearing housing must be fixed first. A loose coupling can add 0.02 mm of lost motion and never appears in a compensation table.
- 3Verify rail straightness and preloadRun a dial indicator along each rail. Straightness within 0.01 mm per 300 mm. Push the carriage by hand: it should move smoothly with light drag. A gritty or stiff feel means the preload is too high or the rail is dry.
- 4Re-check squareness against the spindleSweep a granite square in X and Y with the spindle. Shim or tram as needed. Re-check after a 30-minute warm-up, because a cold reading can be 0.01 mm off.
- 5Tune the servo loop and following errorSet gain so following error during a 4,000 mm/min move stays inside the machine spec. Too much gain causes a hum or a slight overshoot at the end of a move; too little leaves the axis lagging and slow to settle.
- 6Apply residual backlash compensationOnly after the mechanical fix. Keep the value small and stable. If the number climbs each month, stop and inspect the nut, not the parameters.
- 7Confirm with a test cutCut a test part with a known bore and a facing pass. Measure roundness, taper and position. If the part is good and the dial readings are good, the machine is ready.
Symptom, likely cause and what to do
Use this to decide whether the fix is setting, adjustment or replacement.
| Symptom | Likely cause | Fix |
|---|---|---|
| Consistent offset on every part | Squareness or tool offset error | Re-tram and reset the work offset |
| Scattered values, worse when warm | Thermal drift in the ball screw | Warm-up cycle, linear scale, coolant control |
| Lost motion on reversal | Backlash in nut, bearing or coupling | Fix the mechanical play, then compensate |
| Taper on a facing pass | Table not square to spindle axis | Shim the column or bed, not the table |
| Rails run hot after an hour | Preload too high or lubrication dry | Reduce preload, check the lube line |
| Accuracy fades over months | Rail or screw wear | Replace the worn element; tuning will not help |
Questions engineers ask before we touch the machine
How often should I check the positioning accuracy of CNC sliding table?
For parts held to ±0.005 mm, check backlash and squareness monthly, and log the numbers. For general work, a quarterly check is enough.
Any time a part fails inspection, check the machine before you rewrite the program. The program is usually not the problem.
Can backlash compensation replace a worn ball nut?
No. Compensation shifts the commanded position to hide the gap. The gap keeps growing and the table starts to judder on reversal.
Fix the nut or the bearings, then apply a small residual value. If the compensation number keeps rising, replace the part.
Why does the machine cut well in the morning and drift after lunch?
That pattern is thermal. The screw and the bed warm up under load and the position shifts by 0.03–0.06 mm over a few hours.
Run a 20–30 minute warm-up cycle and keep the room temperature steady. A linear scale removes most of the remaining drift.
Do I need a linear scale for ±0.005 mm work?
Over long travel, yes. A rotary encoder on the screw cannot see thermal growth, so the error reaches the part.
A linear scale reads the table position directly and holds the positioning accuracy of CNC sliding table across a full shift.
What preload should I run on the linear guides?
Light preload, around 2–3% of the dynamic rating, for fast moves and light cuts. Medium, around 5–8%, for heavy cutting.
If the rails are hot after an hour of running, the preload is too high. Heat means friction, and friction means drift.
When is a sliding table beyond adjustment?
When rail straightness or screw lead error is out of spec along most of the travel. Shimming helps at the ends, not in the middle.
At that point the honest answer is replacement. We measure before we quote so the decision is based on numbers, not guesswork.
Send us the drawing, get a real quote
Upload your part file and we will return a quotation and a free DFM analysis within 12 hours. No minimum order quantity, from one prototype to 10,000+ parts.
12-hour quote±0.005 mm tolerance100% inspectionNo MOQ
More process notes from the shop floor
We publish setup notes, tooling trials and inspection data from the factory floor.
