Precision CNC: Deliver Consistent Quality

• 1
  What deliver consistent quality actually means on a machine
• 2
  The five variables that move a dimension between parts
• 3
  How setup decisions hold a tolerance across a run
• 4
  Measurement and in-process checks that catch drift early
• 5
  When consistency is hard, and what changes the answer
• 6
  Which consistency control matters for which part

A single part that measures well proves the operator had a good morning. Consistency means the tenth, hundredth, and thousandth part land in the same place inside the tolerance band. That difference matters because a tolerance is a window, not a target. A ±0.005 mm window is only 0.01 mm wide in total, and a process that wanders across that window is far riskier than one that sits on the nominal and barely moves.

So consistency has two parts. Accuracy is where the cut lands relative to the drawing. Repeatability is how little that landing point drifts, part after part and day after day. On a well-set-up machine, repeatability is usually better than accuracy, which is why calibration and setup discipline do more for a run than a faster spindle does.

We measure this in real numbers. A finished lot should hold ±0.005 mm (±0.0002 in) on critical features, with fine finishes down to Ra 0.2–0.8 μm where the drawing calls for it. Those numbers are only meaningful when the process holds them across the whole order, not on the first article.

There is also a practical definition the buyer cares about: the same part number, ordered again in three months, fits the same assembly without rework. That is what precision CNC should deliver, and it is mostly a documentation and setup problem, not a machining problem.

Thermal growth is the most underrated one. Aluminium expands roughly 23 μm per meter per °C. A 300 mm aluminium part that is 5 °C warmer than the reference temperature can grow about 35 μm, which is already larger than a ±0.005 mm band on that feature. Steel moves less, about 11–12 μm per meter per °C, but on a 4,000 mm part even a small shift adds up.

Tool wear is second. A carbide end mill loses edge sharpness progressively, so cutting forces rise, deflection rises, and the finished dimension creeps. On long runs we set a wear offset schedule rather than waiting for a dimension to go out of tolerance.

Workholding comes next. Clamping force deforms a thin wall; release it and the part springs back somewhere else. The fix is usually the sequence, not more clamps. Rough, relax, semi-finish, relax, then finish with low clamping pressure.

Spindle and axis positioning contribute a smaller but constant error. Ballscrew pitch error, thermal drift in the spindle, and servo reversal all add a few microns. Modern controllers compensate for most of it through ballbar calibration and pitch error mapping, which is why we calibrate on a schedule rather than once at install.

Finally there is raw material. Different heat lots of the same alloy machine differently. Hardness varies, residual stress varies, and a batch of 6061-T6 from one mill may cut cleaner than the next. When a part is stress-sensitive, stress-relieved stock or a rough-and-aging cycle is worth the extra day.

A repeatable run starts before the first chip. We pick the smallest machine that fits the part envelope, because a 4,000 mm bed cutting a 200 mm part gives up stiffness for no benefit. The 500 × 500 × 450 mm and 500 × 310 × 200 mm travels cover most small and medium parts; the 750 × 1,150 × 550 mm and 600 × 600 × 600 mm machines take the mid-size work, and the 4,000 × 400 × 150 mm travel handles long parts.

Fixture design decides whether the setup repeats. If a part locates on a machined face rather than a raw casting surface, the second setup starts in the same place as the first. We use a Ø400 mm rotary table on the 5-axis centers so four or five faces come off in one setup, which removes the largest single source of stack-up error: re-fixturing.

Five-axis work also changes the tolerance math. When the tool approaches from the correct angle, a contoured surface can be cut with a smaller stepover and a better surface finish, often Ra 0.8–1.6 μm without a separate polishing step. On 3-axis machines the same geometry needs multiple setups and often hand blending, which adds variation between parts.

Cutting data is the last lever. Too light a finishing pass rubs the tool and work-hardens stainless; too heavy a pass deflects a thin wall. On 17-4PH and titanium we run conservative depths with high-pressure coolant, and on plastics we slow the feed to avoid melting the edge. These are per-part decisions, not house rules.

You cannot hold a tolerance you do not measure. Incoming stock gets checked for hardness and dimensions. During a run, the operator measures critical features at a set interval and logs the reading. The interval depends on how fast the process moves: a short cycle with a fresh tool might be every 20 parts, a long cycle with high tool wear might be every 3.

Between machine and inspection room the temperature matters. A part pulled hot off the machine and measured immediately can read differently once it cools. For tight features we let parts stabilize before final inspection, and we measure at a controlled temperature when the drawing demands it.

CMM reports are available on request, and we inspect 100% of parts before shipment. That does not mean every dimension on every part; it means every part is checked against the drawing's critical features, with raw material checks, in-process monitoring, and a final inspection gate. Inspection records stay with the job so a repeat order starts from the same baseline.

When a dimension trends, we correct the process, not just the parts. Adjusting a wear offset without updating the setup sheet hides the problem and it returns on the next run.

Thin walls are the classic hard case. A 0.8 mm wall on a 100 mm aluminium housing will move under any clamping force, and no amount of machine precision fixes it. The answer is usually a different process route: leave more stock, use a fixture that supports the wall, and take the finish pass with minimal load. Sometimes the honest answer is to thicken the wall or add a rib.

Very tight tolerances across a long dimension are also hard. Holding ±0.005 mm on a 20 mm feature is routine. Holding it on a 1,000 mm feature is a different problem because thermal expansion and machine geometry both scale with length. If the drawing allows it, move the tight tolerance to a short datum and reference the rest from there.

Hard materials raise tool wear sharply. Inconel and hardened tool steel wear edges fast, so the run needs shorter tool life limits and more frequent replacement. That is a cost, not a failure, and it should be planned into the quote.

The opposite boundary is a one-off prototype. Consistency over a run of one is not measurable. What matters there is whether the design can be made at all, and whether the process chosen for the prototype is the one that will run production. We offer no minimum order quantity, from one prototype to 10,000+ part runs, but the setup should be decided with the final volume in mind.