CNC precision parts manufacturing: how tolerance, setup and inspection actually work

• 1
  What CNC precision parts manufacturing really controls
• 2
  Why five-axis machining changes the tolerance stack
• 3
  Material choice sets the real limits of CNC precision parts manufacturing
• 4
  Inspection and finishing decide whether the part ships
• 5
  Where cost and lead time actually come from
• 6
  Matching the process to the part

CNC precision parts manufacturing is subtractive. A rotating or static cutting tool removes material from a solid blank, and the machine's control loop holds the tool path to a programmed coordinate. Precision comes from three things working together: the stiffness of the machine structure, the thermal stability of the setup, and the metrology used to verify the result. Take away any one of them and the tolerance number on the drawing becomes a wish.

A tolerance is not a single value. It is a band that applies to a specific feature under specific conditions. On a shaft diameter of Ø20 mm, ±0.005 mm is achievable on a good turning center. The same ±0.005 mm across a 400 mm bolt-hole pattern is a different problem, because position error accumulates from fixturing, spindle growth, and tool wear. Engineers who treat tolerance as one global number usually get surprises at first article inspection.

The other half of precision is surface finish, and it is not cosmetic. A Ra 0.8–1.6 μm finish on a sealing face controls leak rate. A Ra 0.2–0.8 μm finish on a bearing journal controls fatigue life. As-machined surfaces at Ra 1.6–3.2 μm are fine for brackets and housings, and chasing a finer finish there only adds cost and cycle time.

One practical rule: decide which features are functional and which are reference. Call out tight tolerance and fine finish only on the functional ones. That single habit removes more cost from a part than any negotiation.

• 1
  Machine stiffnessRigid castings and preloaded ballscrews hold the path under load.
• 2
  Thermal stabilitySpindle and ballscrew growth shift dimensions over a long run.
• 3
  MetrologyIf you cannot measure it, you cannot claim it.

On a 3-axis machine, every new face means a new setup. Each setup adds a locating error and a clamping error. If a part has features on five sides, a 3-axis route might need three or four operations, and the position tolerance between those features is the sum of all the re-fixturing steps. That stack is often larger than the feature tolerances themselves.

A simultaneous 5-axis machining center cuts that stack down. The part stays in one fixture while the tool tilts and rotates around it. Holes, pockets and contours on multiple faces come off the same datum. For a housing with angled ports or a bracket with compound-angle faces, this is the difference between a part that assembles and a part that needs rework.

Five-axis is not automatically better. It is slower to program, and a tilting head has less rigidity than a fixed spindle at full extension. For a simple plate with holes on one face, a 3-axis machine with a good vise runs faster at lower cost. Reach matters too, and the machine's work envelope decides what is possible. Our 5-axis centers cover travels from 500 × 500 × 450 mm up to 4,000 × 400 × 150 mm, with a Ø400 mm rotary table for round and cylindrical work.

The practical question is not how many axes, but how many setups. Count the distinct faces that carry toleranced features. Two or fewer, and 3-axis is usually right. Three or more, especially at compound angles, and 5-axis starts paying for itself.

• 1
  One fixture, many facesFewer datums means a shorter tolerance stack.
• 2
  Deep cavitiesA tilted tool reaches pockets a straight tool cannot.
• 3
  Rigidity trade-offTilting heads deflect more than a fixed spindle.

The material decides how fast the tool can cut, how much heat leaves with the chip, and how much the part moves after the cut. Aluminum 6061 and 7075 machine freely and hold tolerance well, which is why they dominate prototypes and low-volume housings. 7075 gives higher strength but is less weldable and more prone to stress relief movement after heavy stock removal.

Stainless steel is a different conversation. Grades 303 and 304 are common, but 304 work-hardens under the tool, so a light finishing pass with a dull insert raises surface hardness instead of cutting it. Grade 316L is the standard choice for medical and food-contact parts, and 17-4PH (SUS630) gives high strength after heat treatment. On these grades, coolant pressure and feed rate matter more than spindle speed.

Titanium TC4 (Ti-6Al-4V) and Inconel sit at the hard end. They conduct heat poorly, so the cutting edge absorbs it. Tool life drops, cycle time rises, and the cost follows. If a titanium part is not strictly required by strength-to-weight or corrosion needs, aluminum with a hardcoat anodize often performs the same job for less.

Plastics behave differently again. POM and PEEK hold dimensions well but need sharp tools and generous coolant to avoid melting. ABS and PMMA are cheap and fast but soft, so clamping pressure alone can mark them. The tolerance you can hold is a property of the material pair, not just the machine.

• 1
  Free machiningAluminum 6061, brass C36000, stainless 303 cut cleanly.
• 2
  Work hardeningStainless 304 and 316 need constant feed, no rubbing.
• 3
  Heat retentionTitanium and Inconel push heat into the cutting edge.
• 4
  Stress movementHeavy stock removal on 7075 can warp thin walls.

A cut part is not a finished part. Inspection is what turns a dimensional claim into evidence. At minimum, a precision job needs a raw material check, in-process monitoring at defined intervals, and a final inspection before shipment. For a ±0.005 mm feature, that means a coordinate measuring machine or a high-accuracy optical system, not a caliper.

In-process checks matter more than final ones on long runs. Tool wear drifts a dimension gradually, so measuring every twentieth part catches the trend while there is still time to offset the tool. If the first check happens only at the end of a 500-piece run, the whole batch is suspect.

Finishing comes next, and it changes dimensions slightly. Anodizing adds a oxide layer, typically 5–25 μm depending on type, which grows the part. Hardcoat anodize grows more than a clear decorative coat. If a bore is anodized after machining, the pre-plate dimension must be undersized by roughly twice the coating thickness. Electroless nickel and plating behave the same way. Designers who ignore this end up with bores that will not accept a shaft.

Laser marking is the last step and has its own limit: minimum character height is 1.5 mm. Below that, the mark is not reliably legible. If a part needs a serial number, leave room for it on a flat, machined face rather than a cast or curved surface.

• 1
  CMM for tight featuresCalipers cannot resolve ±0.005 mm reliably.
• 2
  Trend before scrapFrequent in-process checks catch tool wear early.
• 3
  Coating growthAnodize and plating add thickness to every surface.
• 4
  Mark size floor1.5 mm minimum character height for laser marking.

Two parts with the same drawing can carry very different prices. The usual drivers are setup count, tolerance density, material cost, and inspection level. A part with one setup and loose tolerances is cheap. A part with five setups and a tolerance callout on every dimension is not, even if it fits in the same machine.

Tolerance density is the most misunderstood driver. If a drawing shows ±0.005 mm on forty dimensions, the shop must plan the process to hold all forty simultaneously, which may force a slower spindle speed, a temperature-controlled room, or a different machine. Moving ten non-functional dimensions to ±0.1 mm can cut cycle time noticeably.

Lead time is mostly about queue and setup, not cutting speed. On standard materials, production can start within 24 hours of a released order, and parts typically ship in 3–5 days. Quotation and a free DFM analysis come back within 12 hours, which is the point where a good shop tells you that a wall is too thin or a corner radius cannot be reached.

Volume matters less than people expect. There is no minimum order quantity here, so a single prototype and a 10,000-part run go through the same first-article discipline. The economics shift around fixturing: above a few hundred pieces, a dedicated fixture pays for itself in reduced setup time per part.

• 1
  Setup countEach additional setup adds cost and error.
• 2
  Tolerance densityTight callouts on non-critical dims inflate price.
• 3
  Inspection levelFull CMM reports cost more than sample checks.