Tucson CNC machining accuracy

• 1
  What actually sets machining accuracy
• 2
  Why five-axis changes the accuracy picture
• 3
  Thermal drift is the error nobody sees on the drawing
• 4
  Workholding and tooling decide the last 0.01 mm
• 5
  Material behavior and what it does to accuracy
• 6
  How accuracy is verified before shipment
• 7
  How to write tolerances that hold in production
• 8
  The cost curve of tighter tolerance
• 9
  Which process route fits which accuracy demand

Accuracy is the difference between the dimension on your drawing and the dimension on the finished part. Every CNC machine has four error sources stacked on top of each other: geometric error, thermal error, static deflection, and dynamic error. Geometric error comes from the machine itself, squareness of axes, straightness of travel, and spindle runout. Thermal error comes from heat growing the spindle, ballscrews, and the part. Static deflection comes from cutting force bending the tool and the workpiece. Dynamic error comes from vibration.

On a well-kept machine, geometric error is the smallest of the four. Thermal drift and deflection usually dominate, which is why two shops with the same machine model can hold different tolerances. A shop that runs warm-up cycles, controls coolant temperature, and roughs then finishes in separate passes will hold ±0.005 mm on a 50 mm aluminum feature. A shop that skips those steps will drift.

This is the reason a tolerance callout is not a machine spec. It is a process spec. The machine sets the floor. The process decides whether you stay near it.

• 1
  Geometric errorAxis squareness, straightness, spindle runout. Fixed by the machine builder and by regular calibration.
• 2
  Thermal errorSpindle and ballscrew growth. Controlled by warm-up, coolant, and cutting strategy.
• 3
  Static deflectionTool and part bending under cutting force. Controlled by toolpath, tool stick-out, and support.
• 4
  Dynamic errorChatter and vibration. Controlled by spindle speed, depth of cut, and rigidity of the setup.

A three-axis machine moves the tool along X, Y, and Z only. To reach the back of a part, you flip it and reset it. Every reset adds a new stack of errors: fixture location error, probe error, and the small differences between the two setups. On a complex bracket with eight faces, that stack can easily reach 0.03–0.05 mm before you cut anything.

A five-axis machine adds two rotary axes, usually A and B, so the tool can tilt and approach the part from almost any direction in one setup. Fewer setups means fewer datum shifts. On a housing with bores on three sides, five-axis work often holds true position twice as tight as a three-setup three-axis route, and it does so faster.

The gain is not automatic. Rotary axes bring their own error: pivot distance, rotary encoder resolution, and thermal growth in the trunnion. A five-axis machine that is not calibrated and not probed will not beat a well-planned three-axis job. The advantage shows up when the geometry is complex and the datum chain is short.

• 1
  One setup, one datumCuts the error stack from multiple re-clamping operations.
• 2
  Short toolsTilting lets you reach deep pockets with a stubby tool, which reduces deflection.
• 3
  Better surface finishThe tool can stay tangent to curved surfaces instead of scraping with the tip.
• 4
  Not a cure-allRotary axis error and trunnion growth must be calibrated out.

A spindle running at 12,000 rpm grows in length by a few micrometers per minute for the first 30–60 minutes. On a 100 mm bore, that shift can move the diameter by 0.01 mm over a morning. The same happens to the ballscrew: as it warms, the pitch effectively changes and the axis travels slightly long. Aluminum parts grow too. A 200 mm aluminum part rises about 0.0048 mm per °C, so a 5 °C shop swing moves it 0.024 mm.

Shops that hold tight tolerances deal with this in three ways. They run a warm-up program before the first cut. They keep coolant at a controlled temperature so the part does not cycle hot and cold. And they leave a finishing allowance, let the part cool, then take the final pass. That last step is why a roughing and finishing split matters more than raw spindle speed.

When you review a first article, ask what time of day it was machined and whether the machine was warm. A part cut at 7 a.m. on a cold machine and a part cut at 2 p.m. are not the same part.

• 1
  Warm-up cycle15–30 minutes before the first production cut.
• 2
  Coolant controlHolds the part and the spindle near one temperature.
• 3
  Rough, cool, finishRemoves bulk material, then takes the last 0.2–0.5 mm cold.
• 4
  In-process probingMeasures the part on the machine and offsets the tool.

A part can only be as accurate as its grip. If a vise lifts the part 0.01 mm on one corner, every feature cut after that is tilted. Thin-wall parts are worse: a 1.5 mm aluminum wall will deflect away from the cutter and spring back after the clamp is released, so the measured dimension is fine on the machine and wrong on the bench. Soft jaws machined in place, vacuum plates, and low-pressure clamps reduce this.

Tool choice matters just as much. A 20 mm end mill hanging 100 mm out of the holder will deflect far more than a 12 mm tool hanging 30 mm out. The rule we use is simple: use the shortest, stiffest tool that reaches the feature. For deep pockets, that often means a smaller tool with a reduced stepover, not a longer one.

Tool wear is the slow error. A carbide insert that has cut 200 aluminum parts will start to push instead of shear, and the diameter creeps. Tool life monitoring and scheduled replacement keep the last part in a batch as accurate as the first.

• 1
  Soft jawsMachined in place so the grip matches the part contour.
• 2
  Minimize stick-outShort tools deflect less; deep features need smaller diameter.
• 3
  Support thin wallsUse sacrificial material or low-pressure fixturing.
• 4
  Track tool lifeReplace before the wear land changes the dimension.

Aluminum 6061 and 7075 cut clean and hold tight dimensions well, but they move after machining. A 7075 plate released from a vise can spring 0.02–0.05 mm over a 300 mm length because of residual stress. Stress-relieved stock and a roughing pass before finishing reduce that. Stainless 304 and 17-4PH work-harden, so a dull tool rubs and pushes the wall instead of cutting it.

Titanium TC4 (Ti-6Al-4V) has low thermal conductivity. Heat goes into the tool and the part, not the chip, so the part grows during the cut and shrinks after. Inconel is worse. On these materials, accuracy comes from low cutting speeds, generous coolant, and a finishing pass with a sharp tool and small depth of cut.

Plastics are their own case. POM and PEEK have high thermal expansion and low stiffness. A 100 mm POM part can move 0.1 mm over a 20 °C swing. Clamping force alone can deform it. For plastics, light passes, sharp tools, and a cool-down before final measurement are the difference between a good part and a scrap part.

• 1
  AluminumStress relief and a roughing pass keep flat parts flat.
• 2
  StainlessSharp tools and constant feed avoid work-hardening.
• 3
  Titanium and InconelHeat stays in the part; plan for cool-down before final cuts.
• 4
  PlasticsLow clamp force, light passes, measure after cooling.

A number on a certificate means nothing without the method behind it. At GreatLight, inspection runs in three stages: raw material check on arrival, in-process monitoring during cutting, and final inspection before shipment. CMM reports, roundness and roughness traces, and first article inspection reports are available on request. Every part is inspected, not sampled.

The measurement method has to match the tolerance. A caliper reads to about ±0.02 mm, so it cannot verify a ±0.005 mm bore. That needs a bore gauge, an air gauge, or a CMM with a calibrated probe. Surface finish needs a profilometer; the eye cannot tell Ra 0.8 μm from Ra 1.6 μm.

Temperature at inspection matters too. A part measured on a cold granite table in a 18 °C room reads differently than the same part at 25 °C. For tight work, we let the part stabilize before the final measurement and record the room temperature on the report.

• 1
  Raw material checkGrade and condition confirmed before cutting.
• 2
  In-process monitoringProbing and spot checks during the run.
• 3
  Final inspection100% of parts measured before shipment.
• 4
  Reports on requestCMM, FAIR, and roughness data supplied with the batch.

The most common mistake on a drawing is applying a tight general tolerance to every dimension. That forces the shop to treat a cosmetic edge like a bearing bore, which raises cost and lead time without adding function. Put tight tolerances only on the features that matter: mating diameters, bores for bearings, and datums used for assembly.

Second mistake: tolerancing a feature that cannot be reached. A ±0.01 mm callout on the floor of a deep pocket is hard to measure and hard to cut. If the function allows, move the tolerance to a feature the probe can touch. Third: mixing datums between drawing views, which forces the shop to guess the setup sequence.

A practical rule set: use ±0.1 mm as the default on non-critical dimensions, hold ±0.02–0.05 mm on mating features, and reserve ±0.005 mm for the few dimensions that drive the assembly. Add a note about which surfaces are cosmetic. That single change cuts quotes and rework on almost every project.

• 1
  Tolerance only what mattersGeneral ±0.1 mm, tight only on functional features.
• 2
  Keep datums consistentOne datum scheme across all views.
• 3
  Tolerance reachable featuresIf the probe cannot touch it, the shop cannot verify it.
• 4
  Flag cosmetic vs functionalPrevents over-machining of non-critical faces.

Accuracy is not linear in cost. Going from ±0.1 mm to ±0.05 mm is usually a modest change: better tools, a finishing pass, and a probe check. Going from ±0.05 mm to ±0.01 mm means more machine time, more inspection, and more scrapped parts. Going below ±0.005 mm often means grinding, lapping, or a temperature-controlled room, and the price per part can double or triple.

The question to ask is what the assembly actually needs. If a bearing sits in a housing with a 0.02 mm clearance, holding the bore to ±0.005 mm buys nothing. If two parts stack and the stack-up must stay under 0.03 mm, then the individual tolerances have to be tight, and the cost is justified.

The cheapest accuracy gain is usually in the design, not the machining. Shorten the tolerance stack by reducing the number of parts in the chain. Use a slip fit instead of a press fit where function allows. Those changes cost nothing and remove the need for extreme tolerance.

• 1
  ±0.1 to ±0.05 mmStandard CNC process, modest cost change.
• 2
  ±0.05 to ±0.01 mmMore machine time, probing, and inspection.
• 3
  Below ±0.005 mmGrinding or lapping, controlled temperature.
• 4
  Design firstFewer parts in the stack beats tighter tolerances.