SR CNC Machining Excellence: What Actually Decides Your Tolerance

• 1
  How Five Axes Change SR CNC Machining Excellence
• 2
  Fixturing and Clamping Loads: The Silent Tolerance Killer
• 3
  Thermal Drift During Long Cuts
• 4
  Toolpath Strategy and Surface Finish
• 5
  Measurement: Where SR CNC Machining Excellence Is Proven or Lost
• 6
  When the Process Cannot Reach the Drawing
• 7
  Material Choice Sets the Realistic Tolerance Window
• 8
  Which Setup Route Fits Your Part

A three-axis mill moves the tool in X, Y and Z only. Every face that is not reachable from the top needs a second setup: unclamp, rotate, re-datum, re-clamp. Each of those steps adds stack-up error. On a bracket with four angled faces, that stack-up is usually larger than the machine's own positioning error.

A simultaneous five-axis center adds two rotary axes, typically A and B, or a trunnion with a C table. The tool can now approach the workpiece from an angle instead of only from above. One setup replaces three or four. That is where the accuracy gain really comes from, not from the axes themselves.

The rotary axes are not free. Each one carries its own backlash, its own encoder resolution and its own thermal growth. A rotary table that repeats to ±5 arc-seconds still moves a point 200 mm from center by roughly 0.005 mm. Put the feature 400 mm away and the same angular error doubles.

This is the practical rule. Keep critical features as close to the rotary center as the design allows, and avoid putting your tightest bore at the far end of a long part. If the geometry forces it, expect to trade tolerance for reach.

GreatLight runs 16 simultaneous five-axis machining centers alongside 27 three-axis machines and 12 four-axis mills. The mix matters. Simple prismatic parts run faster and cheaper on three axes, and moving them to a five-axis spindle does not improve the result.

• 1
  One setup, fewer datumsEvery eliminated re-clamp removes a stack-up error source.
• 2
  Angular error scales with radiusSame arc-second error, larger deviation further from center.
• 3
  Match the machine to the partFive axes pay off on angled or contoured features, not on plain blocks.

A part that measures perfectly on the machine can spring out of tolerance the moment it is unclamped. Thin walls, long slender ribs and unsupported floors all deflect under cutting force. The cutter pushes, the material moves away, the tool leaves less material than the program asked for.

The fix is not a stiffer machine. It is a better setup. Support the part underneath the cut, keep the wall thickness above roughly 1 mm where the design permits, and use light finishing passes at 0.1–0.2 mm radial engagement instead of one heavy pass.

Clamping force is the second half of the problem. A vise tightened hard enough to hold a part for roughing will distort it for finishing. On a 2 mm aluminum wall, that distortion can exceed 0.05 mm. Sequence matters: rough, release, re-clamp lightly, then finish.

For thin-wall and near-net parts, we often cut soft jaws from the same material as the workpiece. Matching the jaw material reduces the thermal mismatch and gives a conforming surface that spreads the clamping load.

Parts below about 0.8 mm wall thickness are where this stops being a process question and becomes a design question. At that point the drawing usually needs a rib, a boss or a change in material before the tolerance is realistic.

• 1
  Rough, release, finishRe-clamping lightly before the finishing pass removes locked-in stress.
• 2
  Conforming jawsSoft jaws cut to the part profile spread load and cut distortion.
• 3
  Light finishing passes0.1–0.2 mm radial engagement keeps cutting force low.

Spindles warm up. Ballscrews warm up. So does the workpiece, especially in aluminum at high spindle speeds. On a 4,000 mm part, a 2 °C shift in the aluminum and a 2 °C shift in the machine frame do not cancel out, because the two have different coefficients of expansion and different masses.

The usual symptom is a dimension that reads well at 9 a.m. and drifts by 0.01–0.02 mm by mid-afternoon. The machine is not losing position. The part and the frame are simply at different temperatures than they were when the offset was set.

Practical countermeasures are cheap. Run a warm-up cycle before the first cut. Keep coolant temperature stable. Measure on the machine only after the part has soaked. For a tight bore on a long part, let it sit before the final sizing pass instead of measuring hot.

For steel and stainless, thermal effects are smaller but not zero. For titanium and Inconel, the heat goes into the tool rather than the chip, so tool wear moves the dimension instead of thermal growth. That is a different correction, usually handled by tool-life management rather than offsets.

• 1
  Warm-up cycle firstBring spindle and ballscrew to working temperature before the first cut.
• 2
  Soak before measuringMeasure a stable part, not a hot one.
• 3
  Different materials, different causeAluminum drifts thermally; titanium drifts through tool wear.

Surface finish is a function of feed per tooth, stepover and tool runout. Raise the feed too far and the chip gets thin, the tool rubs, and Ra climbs. Run a finishing pass with a stepover above 0.2 mm on a ball nose and you will see scallop marks you cannot polish out of a tight corner.

For a target of Ra 0.8–1.6 μm, we typically finish aluminum with a sharp carbide tool at 0.05–0.1 mm stepover and a feed per tooth around 0.05–0.1 mm. To reach Ra 0.2–0.8 μm, the stepover drops further and a separate polishing or lapping step may be needed.

Tool runout is the variable most people miss. A holder with 0.02 mm runout makes one flute cut most of the material. The load per tooth doubles, the finish degrades, and the tool wears unevenly. Checking runout with a dial indicator before a finishing pass takes a minute and saves a scrapped part.

Contouring strategies also matter on five-axis work. A tool that stays normal to the surface gives a more consistent chip load than one that leans. The trade-off is more machine motion and a longer cycle, so we use it where the finish is specified, not everywhere.

• 1
  Stepover drives RaSmaller stepover, finer finish, longer cycle.
• 2
  Check runout first0.02 mm runout doubles load on one flute.
• 3
  Normal-to-surface on contoured facesConsistent chip load costs cycle time.

A CMM is only as good as the datum it is given. If the drawing says datum A is a face that is machined in setup two, and the CMM is programmed to a setup-one feature, the numbers will look wrong even when the part is right. Datum strategy belongs in the process plan, not in the inspection room.

GreatLight inspects 100% of parts before shipment, covering raw material check, in-process monitoring and final inspection, with reports available on request. That is not the same as saying every dimension is measured on every part. It means the process is monitored and the final article is verified against the drawing.

For features at ±0.005 mm, the measurement itself has to be budgeted. A caliper is not a ±0.005 mm instrument. A temperature-controlled CMM with a known probe tip diameter is. If your tolerance is tighter than the measurement uncertainty, the inspection result is not meaningful.

The same logic applies to surface finish. A visual check tells you nothing about Ra 0.8 μm. A profilometer reading at three points on the surface does. For medical and aerospace work, that data often has to travel with the part.

• 1
  Datums come from the process planMeasurement follows the setup sequence, not the other way around.
• 2
  Budget for measurement uncertaintyA caliper cannot verify ±0.005 mm.
• 3
  Finish needs a profilometerVisual checks do not quantify Ra.

Some features are not a machining problem at all. A 0.3 mm wide slot 20 mm deep in stainless has an aspect ratio that no end mill survives at production cost. An internal corner with a 0.5 mm radius in a 100 mm deep pocket needs a tool so slender that chatter sets the limit, not the machine.

Sharp inside corners are the most common version of this. A milling cutter always leaves its own radius. If the drawing shows R0, the design really means R smaller than 0.4 mm, which needs EDM or a different feature definition. Catching this in DFM saves a redesign later.

Deep holes with a tight straightness call are another. A 10:1 depth-to-diameter ratio in aluminum is routine. At 20:1 you need a pilot, peck cycles and a gun drill, and the straightness expectation should be relaxed. In titanium the ratio drops further.

For very large parts, 4,000 mm is the maximum processing size on our largest travel configuration. Beyond that, the part has to be split, or the process changes to fabrication. Neither is a failure. It is a different route to the same function.

• 1
  Aspect ratio limitsDeep, narrow features move to EDM or a design change.
• 2
  Sharp internal cornersThe cutter radius sets the smallest possible corner.
• 3
  Size ceiling4,000 mm maximum processing size; beyond that, split the part.

Aluminum 6061-T6 and 7075 machine cleanly and hold ±0.005 mm on a stable setup. 7075 is stiffer and finishes better, but it is more prone to stress-relief movement after heavy material removal. For a part with a lot of stock removed, stress-relieved stock avoids a slow bend over the following days.

Stainless 303 and 304 behave differently. 303 is free-machining and holds tolerance well. 304 work-hardens at the cut, so a light finishing pass with a dull tool will rub instead of cut. 17-4PH holds tight tolerances after heat treatment, but the heat treat itself moves the part, so finishing after HT is usually required.

Titanium Ti-6Al-4V and Inconel are heat-limited. Cutting speed drops, tool life drops, and the heat goes into the tool nose. Holding ±0.005 mm is possible on a rigid setup with fresh tooling, but the cost per part is several times aluminum. Tolerance and material have to be chosen together.

Plastics are the opposite case. POM and PEEK hold decent tolerance but move with temperature and moisture. A PEEK part measured straight off the machine can be 0.05 mm different a day later. For any plastic part, the inspection timeline belongs in the discussion.

• 1
  Aluminum needs stress-relieved stockHeavy stock removal can release internal stress slowly.
• 2
  Finish after heat treat17-4PH and similar alloys move during HT.
• 3
  Plastics settleMeasure after the part has stabilized, not immediately.