CNC Machine Accuracy Explained: Where Tolerance Comes From

• 1
  Tolerance, accuracy and repeatability are three different things
• 2
  Where error enters the cut
• 3
  How many axes you need before accuracy improves
• 4
  What ±0.005 mm actually requires
• 5
  How accuracy gets verified, not assumed
• 6
  Which process fits which accuracy demand

Tolerance is a print requirement. It is the band a dimension is allowed to fall inside. A bore called out at 10.00 mm ±0.01 mm is accepted anywhere from 9.99 mm to 10.01 mm. Nothing about the machine decides that number. The designer decides it, and the shop has to hit it.

Accuracy is how close the machine lands to the commanded position. Repeatability is how close it lands to the same spot on the next cycle. A mill can be accurate but poorly repeatable, or repeatable but offset. Repeatability matters more for production runs, because a stable offset can be compensated in the program. Random scatter cannot.

So when a buyer asks whether a shop can hold ±0.005 mm, the real question is about the whole system: spindle, ballscrew, thermal state, fixturing, tool, and metrology. The machine nameplate is only the starting point.

• 1
  ToleranceA drawing requirement, not a machine property.
• 2
  AccuracyDistance from commanded position to actual position.
• 3
  RepeatabilitySpread of results across repeated cycles.

Positioning error starts at the ballscrew. As the screw turns, it heats and grows. On a 1,000 mm axis, a 1 °C rise moves the tool roughly 12 μm. That is already larger than a ±0.005 mm band. Good shops let spindles and axes warm up, then measure and compensate before the first finishing pass.

Tool deflection is the second source. A long 12 mm end mill reaching 80 mm deep will bend under cutting force. The wall comes out tapered, and the bottom of the pocket is not where the program said. Short tools and light radial engagement fix most of this.

Fixture and workpiece stiffness matter just as much. A thin aluminum plate clamped at four corners will vibrate and spring. Clamp it across the full face or support it underneath. The machine may be fine, but the setup is not.

Then there is measurement. A part that is 20 °C in the shop and measured at 20 °C in a clean room reads differently if the two rooms disagree by 3 °C. Aluminum grows about 23 μm per meter per °C, so a 300 mm part moves 7 μm across that gap.

• 1
  Thermal growthBallscrew and spindle expansion during the shift.
• 2
  Tool deflectionLong reach tools bend and taper the wall.
• 3
  Setup stiffnessWeak clamping lets the part move and ring.

A 3-axis mill moves X, Y and Z. Complex parts with features on five sides need multiple setups. Every setup re-establishes a datum, and every datum adds stack-up error. Two setups can easily add 20–40 μm of positional shift between features.

A 5-axis center tilts the tool and the part at the same time. Features on angled faces come off one setup, so the datum stays fixed. That is the main accuracy gain, not the axis count itself. Fewer setups means fewer chances to introduce error.

For parts under 100 mm with tight bore-to-bore relationships, a mill-turn center is often the better pick. Turning and milling happen without releasing the part, so concentricity stays tight. For long shafts, a dedicated turn between centers still wins.

The rule of thumb we use: if the part needs more than two setups on a 3-axis machine, move it to 5-axis. If the part is mostly round and needs cross-drilled holes, move it to mill-turn.

Holding ±0.005 mm on a 50 mm aluminum bracket is routine work. Holding it on a 400 mm steel plate is a different job. The longer the part, the more thermal and deflection effects scale with size.

Surface finish and tolerance travel together. A Ra 0.2–0.8 μm finish usually means a finishing pass with a small stepover and a sharp tool. If the finish callout is Ra 1.6–3.2 μm, the tolerance is often looser and the cycle is faster.

Material matters too. 6061-T6 cuts clean and holds size well. 316L work-hardens and pushes the tool, so finishing passes need to be light. Titanium Ti-6Al-4V generates heat at the edge and will move the part if coolant and feeds are not controlled.

We inspect 100% of parts before shipment, and we check raw material, in-process dimensions and final geometry. Reports are available on request. That does not change the physics, but it catches drift before the parts ship.

• 1
  Small parts±0.005 mm is standard on aluminum under 100 mm.
• 2
  Large partsThermal and deflection error scale with length.
• 3
  Hard alloysTool wear moves the last pass; plan a spring pass.

A machine that has not been checked is a guess. We use ballbar tests to see circularity and backlash, laser interferometer checks for linear positioning, and artifact cutting for real-world performance. Each method catches something the others miss.

Ballbar testing traces a circle and shows whether the machine pulls out of round. It reveals backlash, servo mismatch and squareness error quickly. A short test on a new machine tells you more than a spec sheet.

Laser interferometry measures actual axis travel against a known wavelength. It finds pitch error and scale error across the full stroke. Compensation tables can then be loaded into the control.

Artifact cutting is the final check. We cut a test part with bores, slots and faces, then measure it on a CMM. If the artifact holds, the machine is ready for production. If it drifts, we know which axis is responsible.