CNC processing basics for engineers who specify parts

• 1
  What CNC processing basics look like at the cutting edge
• 2
  Why 3-axis, 4-axis and 5-axis are different problems
• 3
  How material choice moves the whole process window
• 4
  Where tolerance comes from and where it stops
• 5
  What drives cost in a CNC processing job
• 6
  When CNC processing is the wrong answer
• 7
  CNC processing basics: matching the machine to the part

CNC processing is subtractive. A rotating cutting tool removes material along a toolpath that a control reads from a CAM file. Nothing about the shape is imposed by a mold or a die. The geometry lives in the program, so changing a radius or a pocket depth means editing code rather than cutting new tooling. That is why one-off parts and 10,000-part runs use the same machine family.

Every pass has three variables the programmer controls: spindle speed, feed per tooth, and depth of cut. Get them wrong and the tool rubs instead of cutting. Heat builds in the workpiece, edges smear, and the surface tears. Get them right and chips leave the cut cleanly, carrying most of the heat with them.

The tool leaves marks. A sharp insert on a rigid setup leaves a fine, regular pattern. A dull tool on a thin wall leaves chatter, a repeating wave that shows up in the surface finish and in the measured size. Reading those marks tells you more about the setup than the inspection report does.

So CNC processing basics come down to stiffness and heat. Stiffness comes from the machine, the fixture, and the part itself. Heat comes from the cut. Control both and the part lands inside ±0.005 mm. Lose either and no amount of probing will save the run.

A 3-axis mill moves the tool in X, Y, and Z while the part stays still. It handles plates, blocks, and housings with features you can reach from one direction, or from a few setups on a vise. Most brackets, manifolds, and fixture plates never need more.

A 4-axis machine adds rotation around one axis, usually A. The part turns while the tool cuts, so you can reach four sides without re-chucking. Shafts with cross-holes, cylindrical parts with flats, and long thin parts that would deflect under a second setup all benefit.

A 5-axis machine moves the tool and the part together. The tool can stay normal to a curved surface, so ball-nose stepover stays even across a contoured face. Undercuts, deep cavities, and impeller blades become reachable in one setup. Fewer setups means fewer datum shifts and less accumulated error.

That extra capability has a cost. Programming takes longer, the machine hour rate is higher, and the setup needs a skilled operator. If your part fits in a vise and all its features point the same way, 5-axis adds cost without adding value. Match the machine to the geometry, not to the spec sheet.

Aluminum 6061 cuts fast and holds tight tolerances with little effort. It is the default for prototypes, brackets, and heat sinks. 7075 is stronger but gummier, so it needs sharper tools and lighter depths of cut. Both machine well, which is why aluminum dominates quick-turn work.

Stainless 304 and 316 work-harden. If the tool dwells, the surface gets harder than the tool and the next pass fails. Programmers keep the feed high enough to stay under the hardened layer. 17-4PH in the H900 condition is harder again and often needs carbide with a coating and a rigid setup.

Titanium Ti-6Al-4V and Inconel sit at the other end. They conduct heat poorly, so the cutting edge absorbs it and wears fast. Speeds drop, cycle times climb, and tool life becomes the cost driver. A part that takes 20 minutes in 6061 can take two hours in Inconel.

Plastics behave differently again. POM and PEEK cut cleanly but move with temperature. ABS and PC can melt and smear if the feed is too low. The rule is the same across all of them: pick the tool and the parameters for the material, then confirm the result on the first part before running the rest.

Tolerance is not one number for the whole part. It is the sum of machine positioning error, tool wear, thermal drift, and fixture deflection. A machine that holds ±0.005 mm on a cool morning may drift outside that by afternoon if the shop is not temperature controlled.

The feature matters more than the part. A bore measured with a good micrometer can hold ±0.005 mm. A thin wall 0.8 mm thick will deflect under the same cut and move 0.05 mm or more. So the drawing should call tight tolerance only where it functions, not across every dimension.

Surface finish and tolerance pull against each other. A fine Ra 0.2–0.8 μm finish needs a light finishing pass, which takes time and a sharp tool. A general machined finish of Ra 1.6–3.2 μm comes off in one pass with a stronger insert. Specifying a fine finish on a non-sealing face only adds cost.

Inspection closes the loop. We check every part before shipment, with material verification, in-process monitoring, and a final dimensional check. Reports are available on request. If a feature is hard to measure, it is usually hard to machine, so flag it early.

Setup is a fixed cost. Whether you order one part or 500, the machine still needs to be programmed, fixtured, and proven. That is why the first part carries most of the cost and the tenth part is cheap. Ordering one prototype to test the design is normal and we take orders of one.

Cycle time is the variable cost. It scales with material removal volume, material hardness, and the number of tools needed. A part that needs six tools and two setups costs more than a part that needs two tools and one setup, even if the shapes look similar on the drawing.

Post-processing often surprises buyers. Anodizing, plating, bead blasting, and laser marking each add a step, a queue, and a handling risk. Engraved text needs a minimum character height of 1.5 mm to stay legible after coating. Plan the finish before the geometry is frozen.

The most reliable way to cut cost is to loosen what does not matter. Open a tight tolerance to a general one, drop a fine finish on a hidden face, and make sure the tool can reach the feature from the chosen setup direction. Those three moves often beat any negotiation on rate.

CNC processing wins on accuracy, material range, and speed to first part. It loses on unit cost at high volume. Once a part settles into a stable design and the annual volume climbs into the tens of thousands, die casting or injection molding usually beats it on price per piece.

Very thin or very flexible parts are also a poor fit. Sheet metal under 0.5 mm thick will move under clamping force. Rubber-like materials cannot be held in a vise at all. In those cases, laser cutting, waterjet, or vacuum casting does the job with less fighting.

Geometry with deep internal channels, or features that cannot be reached by any rotating tool, may need a different process entirely. Additive manufacturing builds those shapes without tool access. CNC then finishes the critical faces to size, which is a common hybrid route.

The practical test is simple. If the part is stiff, the features are reachable, and the volume is under a few thousand a year, CNC processing is almost always the right call. If any of those three fails, look at the alternatives before you commit to a machining quote.