3D CNC machining: a working guide to precision parts

• 1
  Key takeaways
• 2
  What 3D CNC machining changes on the shop floor
• 3
  3+2 indexing versus simultaneous 5-axis
• 4
  Why fewer setups is the real precision gain
• 5
  Holding ±0.005 mm on a contoured surface
• 6
  Material and geometry limits you should design around
• 7
  How a 3D machined part moves through the shop
• 8
  Which machining strategy fits your part

A 2.5D cut removes material at one Z level, then drops to the next. Stairs, pockets, slots and drilled holes come out fine. The tool always approaches from straight above, so every wall must be vertical and every floor flat. Most brackets, plates and housings are still made this way, and for good reason: programming is quick and the part is easy to inspect.

3D CNC machining changes the geometry of the cut itself. The tool tip follows a continuous path through X, Y and Z, so a ball nose cutter can sweep a curved surface rather than approximate it with steps. Add rotary axes and the workpiece also tilts, which keeps the cutting edge in contact at a better angle along the whole pass.

That difference shows up in three places: surface finish, feature access, and the number of setups. A blended fillet or a contoured airfoil face comes off the machine with the shape you modeled, not a stepped version of it. Undercuts, angled ports and features on five faces get cut without the part coming off the table.

It also raises the bar for everything around the cut. Toolpath files get larger, simulation matters more, and the fixturing has to hold the part rigidly through changes in direction. When a shop prices 3D work, most of the number is engineering and setup time, not spindle time.

• 1
  2.5DPrismatic parts, straight walls, flat floors. Fastest to program and inspect.
• 2
  3D surfacingCurved faces, blends and radii cut with a ball or bull nose tool.
• 3
  3+2The table indexes to an angle, then cuts a fixed direction. Rigid and simple.
• 4
  Simultaneous 5-axisAll axes move at once. For contoured surfaces and hard-to-reach features.

Both use a machine with two rotary axes, but they are different tools in practice. In 3+2, the rotary table tilts the part to a new orientation and locks. The cut then behaves like a 3-axis cut in a rotated coordinate system. It is stiff, easy to verify, and it collapses five faces of work into one setup.

Simultaneous 5-axis keeps the rotaries moving while the linear axes cut. The controller has to coordinate five axes to hold the tool vector, which is why the post-processor and machine kinematics matter so much. This is the mode you need for a true swept surface: impeller blades, turbine vanes, organic housings, spiral ports.

The practical rule is simple. If a feature can be reached with the part tilted and locked, use 3+2. It is more rigid and the tool wears more predictably. Reserve simultaneous motion for geometry that genuinely cannot be cut any other way, because programming and cycle time both go up.

GreatLight runs 16 simultaneous 5-axis machining centers alongside 12 four-axis mills, 27 three-axis machines and 16 mill-turn centers. That mix matters when a job has one contoured face and twenty flat ones. Sending the whole part to a five-axis machine wastes money.

One more boundary: a rotary table has a swing limit. The largest indexable table here is Ø400 mm. If the part plus fixture exceeds that, the geometry has to be re-thought or split.

• 1
  Reach3+2 handles angled faces and cross-holes. Simultaneous handles ruled and free-form surfaces.
• 2
  RigidityA locked rotary is far stiffer than five axes in motion.
• 3
  Cost driverSimultaneous toolpaths need longer verification and slower feed rates.

Every time a part leaves the table, it is re-clamped and re-datumed. Each of those moves adds a small error, and those errors stack. A part cut in five setups can be in tolerance at every step and still miss a hole position on the final inspection because the datums drifted.

Cutting five faces in one setup removes that stacking. The relationship between a bore on the front face and a slot on the side face is fixed by the machine, not by a fixture operator. For parts with tight positional callouts between features on different faces, this alone justifies the machine rate.

The trade-off is fixturing. A one-setup part usually needs a dovetail blank, a custom soft jaw or a tombstone. That is extra work up front, and it is only worth it when the part count or the position tolerance demands it. For a single flat bracket, three setups on a 3-axis machine is cheaper.

It also changes how you should dimension the drawing. Position tolerances referenced to a single datum scheme are easier to hold in one setup. Stacked dimensions that chain across faces are harder, because each link depends on a different clamping state.

• 1
  Good candidateHousing with bores on two or more faces that must stay coaxial.
• 2
  Poor candidateFlat plate with through-holes. One 3-axis setup is enough.
• 3
  Drawing tipUse position callouts to one datum, not chained linear dimensions.

A tolerance number on a drawing is a request, not a property of the machine. ±0.005 mm (±0.0002 in) is achievable here, but it holds under conditions: a stable material, a rigid setup, a controlled temperature, and a feature that can actually be measured with the available metrology.

The measure-and-cut loop matters more than the machine spec. A tool deflection of a few micrometres on a long reach cutter will move the wall, no matter how accurate the ballscrew is. On deep 3D cavities, we often rough, semi-finish, then take a light finishing pass with a smaller stepover and a sharp tool to bring the surface in.

Surface finish and dimensional tolerance are separate specifications. Ra 0.8–1.6 μm is a normal high-quality machined finish, Ra 0.2–0.8 μm needs a dedicated finishing strategy or a secondary operation. Ask for the finish you can inspect, not the finest number on the chart.

On a curved 3D surface, the stepover sets the scallop height, and the scallop height sets the effective finish. A 12 mm ball nose tool with a 0.5 mm stepover leaves a different surface than the same tool at 0.1 mm. Those are programming decisions, so they belong in the quote conversation.

• 1
  Tolerance±0.005 mm on critical features, not on every dimension.
• 2
  Fine finishRa 0.2–0.8 μm. Needs a finishing pass or secondary process.
• 3
  Standard finishRa 1.6–3.2 μm as machined, which suits most functional faces.

Aluminium 6061 and 7075 cut cleanly at high spindle speeds and hold tight tolerances well. Stainless 316 and 17-4PH work-harden, so the tool has to stay in cut rather than rub. Titanium TC4 (Ti-6Al-4V) and Inconel generate heat at the edge, which means slower feeds and more tool changes. Those differences show up directly in cycle time.

Thin walls are the other common failure. A 0.5 mm wall on a 40 mm tall pocket will deflect and chatter. A practical starting point is a wall thickness of at least 1 mm for aluminium and 1.5 mm for stainless, with the height-to-thickness ratio kept under about 10:1. Below that, expect to add support material or accept a slower, lighter cut.

Deep pockets have a similar constraint. The tool has to reach the floor, and a long tool is a flexible tool. A pocket depth more than four times the cutter diameter usually forces a reduced stepdown and a slower feed. Sometimes the better answer is to split the part or open the pocket from the other side.

Plastics behave differently again. POM and PEEK hold good dimensions but move with temperature, so a finishing pass on a warm part may measure differently after it cools. ABS and PC are more forgiving but softer, so clamping marks are the usual defect.

• 1
  Wall thicknessStart at 1 mm aluminium, 1.5 mm stainless, 2 mm titanium.
• 2
  Pocket depthKeep under 4× cutter diameter before accepting a slower cut.
• 3
  Work hardeningStainless and titanium need constant feed, no dwell in the cut.
• 4
  Thermal driftPlastics and thin metals move after cooling. Measure after soak.

It starts with a DFM review. We look at wall thickness, tool reach, datum scheme and whether the tolerances are measurable. A quotation and free DFM analysis come back within 12 hours, and production can start within 24 hours after that. Most parts ship in 3–5 days.

Programming comes next. The CAM engineer picks the tool sequence: face and rough with the largest rigid cutter that fits, semi-finish to a controlled stock allowance, then finish with the smallest tool that reaches the geometry without excessive length. Rest machining cleans the corners the previous tool could not enter.

Then the first article. On a 3D part, the first article is where the setup is proven: datums checked, a test cut on the critical surface measured, and the tool offsets adjusted. All parts get 100% inspection before shipment, with raw material checks, in-process monitoring and a final inspection. Reports are available on request.

Finishing is a separate step. Anodizing, electroless nickel, bead blasting or laser marking all change dimensions slightly, so masking and pre-finish dimensions are agreed before the part is cut. Laser marking has a minimum character height of 1.5 mm, which is worth knowing when you specify a part number on a small boss.

Every part is checked against the drawing, not against the CAM file. The historical late-delivery probability here is below 2%, and the qualification rate is 99.99%.

• 1
  DFM firstWall thickness, tool reach and datum scheme reviewed before quoting.
• 2
  Rough, semi-finish, finishEach pass has its own stock allowance and tool.
• 3
  First articleProves the setup before the run continues.