Excellent precision CNC machining: what actually decides the tolerance

• 1
  Where the tolerance actually goes
• 2
  Why five-axis changes the accuracy ceiling
• 3
  Part features that decide whether the target is reachable
• 4
  How material choice moves the achievable window
• 5
  Measurement closes the loop
• 6
  Which process fits which part

A ±0.005 mm callout does not sit on the machine alone. It is a budget. Machine geometry, spindle thermal growth, fixture stiffness, tool wear and the measurement method each take a slice. On a well-kept five-axis center, geometry contributes a small share and heat contributes a large one. Shops that ignore the split chase the wrong variable.

Spindle growth is the clearest example. A spindle running at 12,000 rpm for two hours can grow 20–40 μm along Z. That is four to eight times a ±0.005 mm band. The part is not wrong because the machine is bad. It is wrong because the last finishing pass ran on a warmer machine than the first.

Workholding adds the next slice. A thin wall clamped at 2 kN deflects under the clamp and springs back after release. The machine held position; the part moved. For thin ribs and housings, light clamping plus support wax or a sacrificial web usually beats a stiffer fixture.

Material behaviour closes the loop. Aluminium 6061-T6 cuts clean and holds size. Inconel and Ti-6Al-4V push back, generate more heat, and wear tools faster, so the effective window narrows and passes get lighter.

• 1
  Heat firstThermal growth often dominates the error budget.
• 2
  Fixture secondClamp-induced deflection looks like a machine fault.
• 3
  Geometry thirdCalibration and ballbar checks keep this slice small.
• 4
  Measurement lastA warm part measured cold reads undersize.

Three-axis work needs one setup per face. Every extra setup adds a re-clamp error and a new datum stack. A part with features on five sides might need four setups, and each one can shift the result by 10–20 μm even with good fixtures.

Simultaneous five-axis machining removes most of that. The tool tilts and rotates around the part, so undercuts, deep pockets and compound angles are cut in one continuous pass. Fewer setups means fewer datum transfers, and the error budget stops growing with feature count.

The second gain is tool orientation. A ball nose cutter on a sloped surface leaves a scallop whose height depends on the effective radius. Tilting the tool keeps the effective radius larger, so the same surface finish arrives with fewer passes. On a curved aerospace bracket this can cut cycle time by a third and improve Ra at the same time.

Shorter tools help as well. Five-axis access lets us use a stubby cutter instead of a long slender one. Deflection scales with the cube of length, so a tool at half the overhang is roughly eight times stiffer. That shows up directly in wall straightness and floor flatness.

• 1
  One setup, five facesFewer datum transfers, smaller stack-up.
• 2
  Tilted toolLarger effective radius, better scallop height.
• 3
  Stubby toolLess deflection on deep cavities.
• 4
  Not always neededSimple prismatic parts stay cheaper on three-axis.

Aspect ratio is the first check. A pocket deeper than four times its width forces a long tool, and long tools chatter. Beyond roughly 6:1, excellent precision CNC machining becomes a slow, expensive proposition. Redesigning to 3:1 or splitting the pocket often costs less than holding the original number.

Wall thickness matters just as much. Below about 0.8 mm in aluminium, the wall moves during and after cutting. It can be done, but the process shifts to light finishing passes, sequenced material removal and sometimes a stress-relief step between roughing and finishing.

Datum structure is a design decision, not a shop decision. A part with a clear primary face, two locating holes and a defined secondary datum can be probed in-process and corrected. A part with none of those forces the shop to build a fixture that invents datums, and invented datums drift.

Surface finish and tolerance interact. Holding Ra 0.2–0.8 μm on a deep cavity usually needs a separate finishing pass at low feed, and that pass is where thermal drift does the most damage. On tight-finish work we rough, let the part stabilise, then finish.

• 1
  Keep depth under 4:1Deeper pockets need long, flexible tools.
• 2
  Walls above 0.8 mmThinner walls move after unclamping.
• 3
  Define datumsClear datums allow in-process probing.
• 4
  Separate the finish passRough, cool, then take the final cut.

Aluminium is the easiest family. 6061-T6 and 7075 cut fast, hold ±0.005 mm comfortably on a rigid setup, and take anodizing without dimensional surprises if the coating thickness is allowed for. If a design can be aluminium, the accuracy target is rarely the problem.

Stainless sits in the middle. 303 and 304 are common; 17-4PH in the H900 condition is harder and needs slower speeds and more coolant. 316L for medical work machines well but work-hardens if the tool rubs instead of cuts, so feed per tooth has to stay above a floor value.

Titanium and nickel alloys are the demanding end. Ti-6Al-4V conducts heat poorly, so the cutting edge absorbs it. Inconel is worse and abrasive. Both need high-pressure coolant, sharp edges and conservative depths. Tolerances are still achievable, but cycle time and tool cost rise sharply.

Plastics and composites break the usual rules. POM and PEEK move with temperature and moisture, so a part measured right off the machine may not match one measured the next day. Carbon fibre wears tools fast and needs dust extraction. On these materials, tolerance should be agreed with a defined measurement condition.

• 1
  AluminiumFast, stable, anodizing allowance needed.
• 2
  StainlessWatch work hardening on 316L.
• 3
  Titanium and InconelHeat stays in the tool; coolant pressure matters.
• 4
  PlasticsDefine temperature and humidity at measurement.

A tolerance nobody can measure is not a tolerance. If a drawing calls ±0.005 mm on a bore, the shop needs a CMM or a bore gauge with the right resolution and a temperature-controlled room, or at least a stable one. A 5 °C swing across a 100 mm aluminium part moves it about 12 μm.

In-process probing is the practical answer for complex parts. The machine measures the datum and key features between operations and offsets the remaining work. This catches fixture shift and thermal growth while the part is still in the vise instead of after it is off the machine.

Final inspection should match the drawing's intent. A full report on every dimension is expensive and often unnecessary. Critical-to-function features, datums and any dimension that drives assembly get measured and reported; general tolerances get sampled.

Reports are available on request. We run raw material checks, in-process monitoring and a final inspection before shipment, and we can send the data with the parts. If a customer wants a first article inspection layout, that is a separate scope and should be stated at quotation.

• 1
  Match gauge to toleranceResolution should be about a tenth of the band.
• 2
  Probe in-processCorrect before the part leaves the vise.
• 3
  Report the critical featuresNot every dimension needs a number.
• 4
  State FAI scope earlyIt changes the quotation.