Large CNC Machine Tool Ball: How Screw Geometry Sets Positioning Accuracy
A ball screw is the last mechanical link between the servo and the workpiece. On a large CNC machine tool ball screw error shows up directly in the part, and it cannot be corrected in the controller. This page covers preload, stretch, mounting, thermal drift, and when a screw is the wrong choice.
What a Large CNC Machine Tool Ball Screw Actually Does
A ball screw turns rotary motion from the servo motor into linear motion of the table. Steel balls roll between the screw shaft and the nut, so friction stays low and backlash can be almost eliminated by preload. On a small machine the screw is one error source among many. On a large CNC machine tool ball screw error often becomes the dominant one, because the shaft itself is long enough to sag, stretch, and warm up unevenly.
The important number is not the nominal lead. It is lead accuracy over the full travel. A screw rated at ±0.005 mm over 300 mm says very little about what happens over 2,000 mm. Over long distances, cumulative lead error, thermal growth, and shaft sag add together rather than cancel. That is why two machines with the same nominal screw grade can hold very different tolerances on the same part.
Think of the screw as a spring that also has to be a ruler. It carries axial load, it positions the table, and it reacts to heat from its own friction and from the surrounding structure. Every design decision below is a trade between those three jobs. Ignore one and the other two get worse.
A practical check: measure the table position with a laser interferometer at cold start and again after two hours of cycling. If the two curves differ by more than your part tolerance, the screw and its thermal behavior need attention before anything else is changed.
- 1FunctionConverts servo rotation into linear table motion with low friction
- 2Key specLead accuracy over full travel, not over a short sample
- 3Main enemiesThermal growth, shaft sag, mounting misalignment, backlash
- 4Field testCompare cold and warm positioning curves
Preload and Backlash in Long Ball Screws
Preload removes the axial play between the balls and the raceways. Without it, the table can move a few micrometres under a light cutting force before the screw starts to push back. On a 4,000 mm axis that lost motion is not constant along the travel, so it is hard to compensate in software. Preload is usually set by using slightly oversized balls or by offsetting two nuts against each other.
More preload is not better. High preload raises friction torque, which raises heat, which causes the screw to grow. On a long screw the growth can exceed the positioning gain from the preload itself. For most general machining axes, a light to medium preload holds the best balance. For heavy roughing, a firmer preload helps rigidity but needs a cooling or thermal compensation plan.
Backlash also comes from the nut return system and from end bearing play. Check the bearings first. A screw with correct preload but loose angular contact bearings will still show backlash at the table, and the reading changes with direction of approach.
When preload is lost, the symptom is usually a small step at direction reversal, plus a finish mark that appears only on one side of a pocket. Re-preloading is cheaper than replacing the screw, but only if the raceways are still smooth.
- 1Light preloadGeneral milling, lower heat, good for long axes
- 2Medium preloadMixed roughing and finishing, common default
- 3Heavy preloadHeavy cutting, needs thermal control
- 4Check firstEnd bearings and nut return, then the screw
Stretch, Mounting, and Why Alignment Errors Compound
Pre-stretching the screw shaft is standard practice on long axes. The shaft is pulled in tension so that when it warms up and wants to grow, it grows into the stretch instead of pushing the table forward. A typical pre-stretch is small, often a few hundredths of a millimetre over a metre, and it is set at a known reference temperature. Get the reference temperature wrong and the compensation works against you for part of the day.
Mounting errors are the second big source. If the nut housing and the end bearing seats are not coaxial, the screw is forced to bend slightly as it rotates. That adds a once-per-revolution error to the table motion. The classic mistake is to align the screw to a reference surface without checking parallelism to the guide rails over the full stroke.
We correct the nut seat and the end bearing bores from one setup, then verify parallelism to the linear guides along the whole travel. This removes the gap between the chuck and the positioning sleeve that causes radial bias load in the older method. Less radial load means less heat and longer bearing life.
Axial force during assembly matters too. Over-tightening the lock nuts to chase a number can distort the bearing races and cause the screw to bow. Follow the bearing maker's axial force range, then measure, then adjust. Measure again after the first warm-up cycle.
- 1Pre-stretchSet at a known reference temperature, a few hundredths per metre
- 2CoaxialityNut seat and end bearing bores cut in one setup
- 3ParallelismVerify to the guide rails over full stroke, not locally
- 4Axial forceStay inside the bearing maker's range; re-measure warm
Thermal Growth and Long-Travel Error
Steel grows about 11 to 12 μm per metre per degree Celsius. A 4,000 mm screw that warms by 5 °C wants to grow roughly 0.22 mm. That is far larger than a ±0.005 mm tolerance, so thermal behavior has to be managed, not ignored. The heat comes from the nut, the support bearings, the motor, and the surrounding shop air.
Three controls are common. Pre-stretch shifts the growth into tension. Forced cooling of the nut or the shaft pulls heat out. Thermal compensation in the controller models the growth and offsets the commanded position. Most large machines use at least two of these together, because no single method covers the whole day.
The shop matters as much as the hardware. A machine near a loading door sees air temperature swings that the compensation model cannot follow. If a shop is not temperature controlled, plan for a warm-up cycle and a reference measurement before critical work.
Ball screw pitch error also changes with load, though the effect is smaller than thermal growth on a long axis. Under heavy thrust the shaft stretches elastically and the nut compresses slightly. This is why a screw sized for stiffness at the worst-case load behaves better than one sized only for life.
- 1Steel growthAbout 11–12 μm per metre per degree Celsius
- 24,000 mm screwRoughly 0.22 mm growth from a 5 °C rise
- 3ControlsPre-stretch, cooling, controller compensation
- 4EnvironmentDoor drafts break the compensation model
What This Means for the Parts We Machine
When a customer sends a long part, the axis behavior and the part geometry interact. A 3,000 mm aluminium extrusion machined on a screw-driven axis will show the axis thermal drift as a taper along its length. The same part on a rack-driven axis shows a different error pattern. Knowing which error dominates helps decide where to put the tight tolerance.
Thin walls and long parts also move after cutting. Aluminium 6061 and 7075 release residual stress when material is removed, so a part that measures correct on the machine can bow overnight. On large work we rough, stress-relieve or rest, then finish. This is separate from screw error but often mistaken for it.
For steel and stainless parts, cutting forces are higher and preload matters more. A screw with light preload may show chatter on a deep shoulder cut, while the same screw is fine on a finishing pass. Matching preload to the heaviest realistic cut is better than matching it to the finish pass only.
Titanium and Inconel raise the heat going into the tool and the part, and sometimes into the screw through the table. Coolant management and dwell time between heavy passes both help. If the axis error grows during a long titanium cycle, check the screw temperature before blaming the toolpath.
- 1Long aluminiumAxis drift shows as taper; rough, rest, then finish
- 2Stress relief6061 and 7075 move after material removal
- 3Steel partsSize preload for the heaviest cut, not the finish pass
- 4TitaniumWatch screw temperature during long cycles
Drive Options for Long Machine Axes
Compare ball screw, rack and pinion, and linear motor axes on the criteria that matter for large travel.
| Criterion | Ball screw | Rack and pinion | Linear motor |
|---|---|---|---|
| Typical travel | Up to 4,000 mm and beyond | Very long, several metres | Long, but cost climbs fast |
| Positioning accuracy | High with preload and stretch | Good with laser calibration | Highest, no mechanical link |
| Rigidity under load | High axial stiffness | Good, depends on pinion mesh | Lower axial stiffness |
| Thermal sensitivity | High, needs compensation | Moderate | Low, but coils heat locally |
| Backlash risk | Low when preloaded | Needs anti-backlash pinion | None mechanically |
| Maintenance | Lubrication and preload checks | Rack lubrication, pinion wear | Clean rails, magnet care |
| Best fit | Precision axes under 4,000 mm | Very long travel, moderate accuracy | Fast, high-accuracy light cuts |
Where the Trade-Off Lands
For a precision axis up to about 4,000 mm, a preloaded and pre-stretched ball screw with controller compensation is the right choice. For travel well beyond that, or where accuracy can be laser calibrated, a rack and pinion drive costs less and drifts less. Pick the screw when axial stiffness and repeatability matter more than pure length.
Questions Engineers Ask
How much pre-stretch should a 4,000 mm screw have?
It depends on the expected temperature rise and the reference temperature used when the screw is set. A few hundredths of a millimetre per metre is a common starting range. The number must be tied to a known reference temperature, otherwise the stretch helps in the morning and hurts in the afternoon.
Set it with the machine at a stable temperature, then verify with a laser interferometer after a two-hour warm-up cycle. Adjust only after you have both curves.
Can controller compensation fix a worn ball screw?
No. Compensation handles repeatable errors such as lead error and thermal growth. A worn raceway produces non-repeatable error that changes with load and direction, and no lookup table can follow it.
If backlash readings vary between cycles, the screw or the bearings need mechanical work first. Compensation is the last step, not the first.
Why does the table position drift after a few hours of cutting?
Thermal growth is the usual cause. The nut, bearings, and motor all add heat, and a long steel shaft expands with it. Pre-stretch and controller compensation reduce the drift but do not remove it completely.
Check whether the drift follows the same curve every day. If it does, the compensation model can be tuned. If it changes with the weather or the shop door, the environment is the problem.
Does a larger screw diameter always improve accuracy?
No. A larger shaft is stiffer and sags less, which helps. But it also has more mass and a larger thermal cross-section, so it reacts more slowly to correction and can hold more heat.
Size the screw for the worst-case thrust and the critical speed, then check the thermal plan. Oversizing without a thermal plan often makes drift worse, not better.
When is a ball screw the wrong choice for a large machine?
When travel goes well past 4,000 mm, when the axis must move very fast for long periods, or when the shop has no temperature control and the part tolerance is tight. In those cases a rack and pinion or linear motor axis is usually the better fit.
The screw still wins where axial stiffness and repeatability matter most, especially on axes that take heavy thrust at moderate speed.
How do we verify screw condition without a full teardown?
Run a laser interferometer along the full travel in both directions, cold and warm. Compare the forward and reverse curves to separate backlash from lead error. A once-per-revolution ripple points to mounting or bearing problems.
Also measure friction torque of the nut and listen for a change in sound at the same table position. Both are quick checks that need no disassembly.
Send Us Your Long Parts and Drawings
We review your geometry and tolerances, flag the features that depend on axis behavior, and quote within 12 hours. Parts ship in 3–5 days.
12-hour quote100% inspectionNo minimum order quantityNDA on request
More Machining Notes
We publish setup notes, tooling trials and inspection data from the factory floor.
