CNC precision machining: what actually holds a tolerance

• 1
  How cnc precision machining closes the loop
• 2
  Where the stiffness budget goes on a precision part
• 3
  Thermal drift: the tolerance nobody sees on the drawing
• 4
  When 3 axes stop being enough
• 5
  Material behavior and what it does to the cut
• 6
  Feature design rules that keep cnc precision machining predictable
• 7
  Measurement decides whether the part is accepted
• 8
  Choosing the right process for the geometry

A CNC machine does not cut to a drawing. It cuts to a stream of coordinates. The CAM software turns your model into toolpaths, the post-processor turns those into G-code, and the control interpolates axis motion from that code. Accuracy comes from how tightly that chain stays closed, not from the spindle alone.

The closed loop is simple to describe. A rotary encoder or linear scale reports actual position. The control compares it to commanded position and corrects the drive. When the scale is mounted on the table rather than the motor, it sees the real position of the part, which removes most of the ballscrew error.

Everything outside that loop is open. Tool deflection, workpiece deflection, thermal growth and fixture slip sit outside the feedback path. The control cannot see them, so it cannot correct them. That is why two machines with identical resolution can produce different results on the same part.

Practical result: cnc precision machining holds tolerance when the closed loop is tight and the open loop is small. When a feature fails, the cause is almost always in the open loop.

• 1
  Closed loopEncoder or scale feedback correcting commanded vs. actual position.
• 2
  Open loopTool, part, fixture and thermal effects the control never sees.

Stiffness decides whether a cut is quiet or whether it chatters. Every element in the load path has a stiffness, and the weakest one sets the limit. On a typical job the order runs: fixture, then workpiece, then tool, then spindle and machine structure.

A long end mill is a spring. A Ø12 mm carbide end mill at 60 mm gauge length deflects far more than the same tool at 25 mm. Rough the feature with the shortest tool that reaches, then finish with a longer one only for the last pass. This single choice often moves a bore from Ra 3.2 μm to Ra 1.6 μm.

Thin walls are the second spring. Below roughly 2 mm wall thickness in aluminium, cutting force pushes the wall away from the cutter, so the tool cuts less than commanded. The wall springs back after the pass and the measured dimension lands oversize.

The fix is not more spindle speed. It is support: reduce radial depth of cut, climb mill, and leave a finishing allowance that is removed at low load.

• 1
  Shortest tool winsRough with stub length, finish with reach length.
• 2
  Support thin wallsLight radial passes and a small finishing allowance.
• 3
  Check the fixture firstUnsupported overhang is the most common single cause of chatter.

Aluminium expands about 23 μm per meter per degree Celsius. A 500 mm part that warms 5 °C during roughing grows roughly 58 μm. That is ten times a ±0.005 mm tolerance, and the machine did nothing wrong.

Steel is closer to 11 μm per meter per degree, titanium about 8.6, so the same part in 4140 moves about half as much. The number that matters is not the coefficient alone. It is how much heat the process puts in and how fast the part can shed it.

In practice we rough, let the part and the machine settle, then finish. For tight bores we keep the finish allowance small and take the final pass after a measured pause. This is not a trick; it is the only way to separate thermal growth from cutting error.

In-process gauging helps on long runs. Touching off a known datum between passes tells the control what happened, and the offset is corrected before the next part starts.

• 1
  Aluminium 6061About 23 μm/m per °C.
• 2
  Steel 4140About 11 μm/m per °C.
• 3
  Ti-6Al-4VAbout 8.6 μm/m per °C, but poor heat evacuation.

A three-axis machine reaches the top face, and that is it. Every new face means a new setup: unclamp, rotate, re-datum. Each setup adds its own error, typically 10–20 μm on a good vise, and adds time.

A fourth axis rotates the part about X or Y. A fifth adds a rotary table that tilts the tool relative to the part. With simultaneous five-axis motion, the cutter can stay normal to a curved surface, so a sculpted profile is cut in one continuous pass instead of a series of blended three-axis patches.

That matters for two reasons. First, fewer setups means fewer datums stacked on each other. Second, undercut features and angled ports become reachable without special tooling. On a Ø400 mm rotary table we can index around a part that would need six or more setups on a three-axis mill.

Five-axis is not automatically more accurate. It is more accurate when the part has multiple faces, contoured surfaces or undercuts. On a flat plate with a few holes, three axes with a good fixture is faster and just as tight.

• 1
  One setup, many facesRemoves stacked datum error.
• 2
  Tool normal to surfaceBetter surface finish on contoured geometry.
• 3
  Not always the answerFlat prismatic parts are usually cheaper on three axes.

Aluminium 6061 and 7075 cut freely and hold a fine finish, which is why they dominate prototypes. 7075 is stronger but less forgiving at the weld line and more prone to stress relief movement after heavy stock removal.

Stainless 304 work-hardens. If the tool rubs instead of cutting, the surface gets harder and the next pass cuts worse. The answer is a positive rake, a sharp edge and enough feed per tooth to stay under the hardened layer. 303 machines much more easily, and 17-4PH sits in between.

Titanium Ti-6Al-4V and Inconel move heat into the tool rather than the chip. Cutting speeds drop, tool life drops faster, and the process needs rigidity more than it needs rpm. Slow and heavy beats fast and light.

Plastics are the opposite problem. POM and PEEK machine cleanly with sharp, polished flutes, but ABS and PC soften with heat and can gum the cutter. Air blast and a generous feed keep the chip clear.

• 1
  Aluminium6061, 7075, 6082: fast, good finish, watch stress relief.
• 2
  Stainless304 work-hardens; 303 and 17-4PH are easier.
• 3
  Titanium and InconelHeat goes into the tool; prioritize rigidity over speed.
• 4
  PlasticsSharp flutes, air blast, avoid heat buildup.

Every cutting tool has a radius. An internal corner cannot be sharper than that radius, so a 6 mm end mill leaves roughly a 3 mm corner. Drawing a sharp internal corner forces the shop to either use a smaller tool, which is slower and more prone to deflection, or to leave the corner and let you argue about it later.

Deep holes are the second constraint. A depth-to-diameter ratio above about 6:1 needs a long, thin tool or a pecking cycle, both of which drift. Past 10:1 the shop may have to drill from both ends, which means a second setup and a visible mismatch at the joint.

Threads follow the same logic. A thread that stops inside a deep pocket needs a tap holder with the right reach, or a thread mill, which is slower but produces a cleaner root. If a thread can be moved to an open face, the cost drops immediately.

None of this is a reason to weaken the design. It is a reason to talk to the shop before releasing the drawing, so the tolerance goes where it matters and the rest stays loose.

• 1
  Corner radiusNever smaller than the tool radius used to cut it.
• 2
  Hole depthKeep under 6:1 for straight drilling.
• 3
  Tolerance placementTight on functional features, loose elsewhere.

A tolerance is only meaningful with a stated measurement method. A bore measured with a two-point micrometer and the same bore measured with a CMM can differ by more than the tolerance when the bore is out of round.

CMM probing gives a fitted diameter over many points, which is closer to how the part functions. A bore gauge reads the high spots. Both are correct, but they answer different questions, so the drawing should say which one governs.

Temperature at the point of measurement matters too. A part measured straight off the machine at 28 °C and then checked in a 20 °C inspection room will read differently. For tight work we let the part stabilize before final inspection.

We inspect 100% of parts before shipment and keep raw material checks, in-process monitoring and final reports. Reports are available on request, and a first article inspection is standard for new parts.

• 1
  State the methodCMM fit vs. two-point contact changes the number.
• 2
  Let the part settleMeasure at 20 °C where the tolerance is tight.
• 3
  Reports on requestRaw material, in-process and final inspection records.