CNC Accuracy Factors: What Actually Moves Your Tolerance
This page breaks down the CNC accuracy factors that decide whether a part lands at ±0.005 mm or drifts out of spec. It is written for design engineers, manufacturing engineers and buyers who need to judge a process before releasing a drawing. By the end you should be able to tell which factors you can control in the design and which ones belong to the shop.
Where accuracy is won and lost
Accuracy is not a single machine spec. It is the sum of six groups of variables, and the weakest one sets your real tolerance.
Machine geometry, calibration and wear
The machine sets the ceiling. Axis straightness, squareness between axes, spindle runout and the condition of the ball screws all show up directly in the part. A machine that is geometrically off by 0.01 mm per 300 mm cannot be programmed around that error, no matter how good the CAM output is.
Calibration is the part most buyers never see. Axis alignment, spindle truth, backlash compensation and probe calibration need to be checked on a schedule, not once at installation. Thermal growth from a spindle running for hours is a real offset, and shops that hold tight work re-check and re-compensate during long runs.
Wear is gradual and easy to miss. Guideways, ball screw nuts and spindle bearings degrade over years of cutting. A machine can still move to position and repeat poorly, which is worse than an obvious fault because the first article may pass while part 400 does not.
- 1Axis alignmentStraightness and squareness between X, Y and Z, checked with a square and dial indicator or laser interferometer.
- 2Spindle truthRunout at the taper and at the tool tip after clamping, measured under load.
- 3Backlash and pitch errorCompensated in the control, but the compensation is only valid while the screw is healthy.
- 4Probe calibrationA drifting probe quietly shifts every work offset on the machine.
Cutting tool, fixturing and material behavior
The tool is the direct contact point with the workpiece. Runout of 0.01 mm on a small end mill cuts an effective diameter larger than nominal, so a slot comes out undersized on one pass and oversized on the next. Tool overhang makes this worse: the same cutter held 20 mm out versus 60 mm out behaves like two different tools.
Fixturing decides whether the part moves. Thin walls, long slender parts and unsupported sections deflect under clamping force before the cut even starts. In-process support, correct jaw pressure and sequenced roughing that leaves material for support are what keep a flexible part in tolerance.
Material behavior is not noise. Aluminium 6061 and 7075 cut cleanly and hold size well; 316L and titanium TC4 work-harden at the cut and push back on the tool. Plastics such as POM and PEEK move with temperature and spring back after the cutter passes, so a bore measured hot may not match the same bore at 20 °C.
- 1Tool runoutKeep total indicated runout low; it multiplies on small-diameter cutters.
- 2OverhangRigid setups cut more accurately than long reach, so reach is a real cost.
- 3Clamping forceEnough to hold, not enough to deform a thin wall.
- 4Material stabilityAluminium and brass are predictable; titanium, Inconel and PEEK need slower passes.
CAM strategy, thermal drift and in-process control
CAM decides how the machine is asked to move. Aggressive stepovers, climb versus conventional direction, entry ramps and toolpath smoothing all change the load on the cutter and the deflection at the wall. A conservative finishing pass with a small radial engagement is often what makes a tight tolerance achievable, not a higher feed rate.
Thermal effects run across hours, not seconds. The spindle, the coolant, the shop air and the workpiece all move together. On long runs we leave roughing stock, let the part stabilize, and take the finishing cuts after the temperature has settled. On a part with a 0.005 mm tolerance, that pause is cheaper than scrap.
Measurement closes the loop. First article inspection, in-process checks on critical features and a final inspection before shipment catch drift while it is still correctable. Reports and dimensional data are available on request, including material certificates.
- 1Rough then finishLeave stock, let the part cool, then finish in a separate operation.
- 2Light finishing passesReduce radial engagement to cut deflection at the wall.
- 3Coolant controlFlood and through-tool coolant both remove heat and change the thermal picture.
- 4In-process probesMeasure on the machine and adjust offsets before the part comes off.
How each factor shows up on the part
Use this when you are deciding which requirement to relax and which to hold.
| Factor | Typical effect | Where to check |
|---|---|---|
| Axis squareness | 0.005–0.02 mm over 300 mm | Square and indicator, laser check |
| Spindle runout | 0.003–0.01 mm at the tip | Dial indicator on a test bar |
| Tool overhang | Deflection grows with the cube of length | Shortest tool that reaches |
| Clamping force | Thin walls bow before cutting | Jaw pressure, support blocks |
| Thermal drift | 0.005–0.03 mm over hours | Stabilize, then finish cut |
| Material springback | 0.01–0.05 mm on plastics | Measure at 20 °C, not hot |
Questions engineers ask about CNC accuracy factors
Can you hold ±0.005 mm on every feature?
±0.005 mm is achievable on critical features with the right machine and setup. It is not realistic to apply one blanket tolerance to every dimension on a part.
Features far from the datum, deep bores, thin walls and long unsupported sections all lose accuracy faster than a short, well-supported face. Tell us which dimensions actually matter and we will tell you what is achievable.
Does a five-axis machine give better accuracy than a three-axis?
Not automatically. Five-axis helps because one setup replaces several, and each re-fixturing step adds error. For a part with features on several faces, that reduction in setups is the real accuracy gain.
For a simple prismatic part, a well-maintained three-axis machine can be just as accurate and cheaper to run.
How much does temperature matter in a real shop?
More than most drawings assume. A spindle that has run for two hours is at a different temperature from a cold one, and the offset shows up in the part.
We run roughing and finishing as separate operations on tight work, and let the workpiece return to room temperature before the final cut. On plastics, we measure after stabilization because the part moves as it cools.
What tolerance should I put on a prototype drawing?
Put tight tolerances only where function requires them. A drawing with ±0.01 mm on every dimension forces extra setups and slower passes without improving the part.
Mark the critical features, give a general tolerance for the rest, and let us flag anything that looks expensive to hold.
How do you verify accuracy before shipment?
Every order goes through raw material check, in-process monitoring and final inspection before it ships. Critical dimensions are measured on the finished part, not on the machine.
Inspection reports and material certificates are available on request. Uploads and drawings stay confidential, and an NDA can be signed on request.
Can accuracy be improved on an existing design without redesign?
Sometimes. Adding a support rib, opening a deep pocket to a shallower profile or moving a tight tolerance from a thin wall to a thicker section can all help.
Other times the fix is process, not geometry: a different setup order, a softer finishing pass or a different tool. We review the drawing during DFM analysis and tell you which lever applies.
Send a drawing and we will review the tolerance before quoting
We check the critical features, flag anything hard to hold, and come back with a quote and DFM notes within 12 hours.
12-hour quote±0.005 mm100% inspectionNDA on request
