CNC 3D Machining: A Guide to Accuracy and Speed

In 3D machining the cutter follows a surface, not a set of flat steps. A ball-nose or bull-nose tool sweeps along toolpaths generated from the CAD model, and each pass leaves a scallop. The stepover between passes sets the height of those scallops, and that height is what the drawing calls surface finish.

This is why surface finish and cycle time move together. Halve the stepover and the scallop height drops by roughly a factor of four, but the number of passes doubles. A mold cavity that runs at Ra 1.6–3.2 μm as machined may need Ra 0.8–1.6 μm after a finer semi-finish pass, and hand polishing after that if the print demands it.

A computer-controlled machine does the cutting, but the CAM programmer decides where the tool goes. Two shops with the same machine can produce different results on the same part because of stepover, lead-in style, and how they handle corners.

The biggest single source is tool deflection. A 12 mm end mill hanging 60 mm out of the holder will bend under cutting force, and the wall it leaves will not match the model. Shorten the gauge length or reduce the radial engagement before you blame the machine.

Thermal drift is the quiet one. A spindle that has run for two hours is longer than a cold one. Shops that hold ±0.005 mm on long runs will warm up the machine and keep the coolant at a stable temperature, because a 1 °C shift across a 500 mm workpiece moves the part by a few micrometers.

Then come the geometric errors: squareness between axes, backlash in the ballscrew, and runout in the tool holder. These are baked into the machine. A well-kept five-axis center can hold ±0.005 mm on a 200 mm part; a worn three-axis bed mill cannot, no matter how careful the operator is.

• 1
  Tool runoutAbove 0.01 mm, one flute does most of the cutting and finish suffers.
• 2
  FixturingA part that moves 20 μm under clamping force cannot be measured reliably.
• 3
  Chip recuttingPoor evacuation rubs the surface and dulls edges fast.

Three-axis machines cut from one direction. For a plate with pockets, holes, and a flat back, they are the fastest and cheapest route, and a well-set three-axis machine holds ±0.005 mm all day. The limit is reach: an undercut or a side face at 40° needs a second setup.

Four-axis adds rotation about the X axis. This lets you cut around a cylindrical part in one setup, which removes the re-fixturing error that comes with flipping a part. Shafts, flanges, and parts with radial features are good candidates.

Five-axis moves the tool in the A and C axes as well. Simultaneous five-axis lets a short, stiff cutter tilt into a deep cavity instead of reaching in from above. That is the real gain: not the extra axes, but the shorter tool. It is also the only practical way to machine a contoured surface in one setup with the tool normal to the surface. GreatLight runs 16 simultaneous 5-axis machining centers, alongside 12 four-axis mills and 27 three-axis machines, so the setup can match the geometry rather than the other way around.

Spindle speed is the least interesting lever. Raising RPM helps aluminium and hurts a long tool in steel, because the limiting factor is chatter, not RPM. The useful changes are shorter tools, better workholding, and toolpaths that keep the radial engagement constant.

High-efficiency toolpaths do the heavy lifting. Instead of a full-width cut at shallow depth, they take a narrow radial step at full depth. The load on the cutter stays even, the tool lasts longer, and the machine can run at a predictable feed. On a 6061 aluminium bracket, this is often where the 30 percent comes from.

Setup count matters more than most people expect. Each extra setup is a new datum and a new chance to lose 20 μm. A part that fits in a five-axis machine and comes off complete in one op will beat a three-axis part that needs four fixtures, even if the spindle is slower.

Aluminium 6061, 7075, and 2024 cut freely and hold tight tolerances well, which is why they dominate prototype and fixture work. Stainless 304 and 316 work-harden at the cut, so heavy radial engagement and light feed will burn the tool. Titanium Ti-6Al-4V and Inconel move the problem to heat: they need lower surface speed, more coolant, and a rigid setup.

Plastics are a different story. POM and PEEK move with temperature, and a part machined at 25 °C may not gauge the same at 20 °C. For tight plastic parts, take a light finish pass and let the part stabilize before final inspection.

Finish is not separate from accuracy. If the print asks for Ra 0.2–0.8 μm on a contoured face, that calls for a fine stepover and a sharp tool, which adds cycle time. Saying which surfaces need the fine finish, and which can stay at Ra 1.6–3.2 μm as machined, saves money on every part.

A tolerance on a quote sheet is a claim. Ask what the shop measures it with. A CMM with a stated uncertainty of 2 μm can support a ±0.005 mm call on a 100 mm part. A caliper cannot.

Ask when the inspection happens. In-process checks catch a drifting tool before a whole batch is scrap. Final inspection catches what slipped through. Both matter, and a shop that only does the second one is telling you something.

At GreatLight, every shipment gets a full inspection, and reports are available on request. Raw material check, in-process monitoring, and final inspection are separate steps, so a problem is caught at the machine rather than at the loading dock.