Precision CNC machining of Joeburg: a subtle art

• 1
  What CNC machining of Joeburg removes, and why that matters
• 2
  When 3-axis, 4-axis, and 5-axis each make sense
• 3
  How tolerance and surface finish are actually held
• 4
  Material behavior changes the cutting strategy
• 5
  Where the process stops being the right answer
• 6
  Matching the part to the process

Subtractive machining starts from stock that is already close to the final shape. A forged blank, a casting, or a saw-cut bar has the right mass in roughly the right place. The cutting tool then removes what is left. That is the whole mechanism, and most of the engineering judgment sits in one question: how much material must come off before the part is true?

CNC machining of Joeburg is not a single operation. A spindle rotates a tool with a defined number of flutes, the tool advances at a feed rate expressed in mm per tooth, and the machine moves the workpiece along controlled axes. Each pass leaves a scallop pattern. Feed, speed, and depth of cut decide whether those scallops are 2 μm deep or 20 μm deep.

The reason precision matters is stack-up. A single bore held at ±0.005 mm is easy to talk about. Twelve bores that must align with a mating housing are not. Every locating feature, every datum, and every re-clamp adds error. Good process planning keeps the number of re-clamps low and keeps datums consistent from the first op to the last.

So the art is not in one clever cut. It is in sequencing: which face you machine first, where you leave stock for the heat-treat move, and which features you finish after stress has been relieved.

• 1
  Near-net stockCastings and forgings cut cycle time but need a stress-relief step before finishing.
• 2
  Datum disciplineOne datum scheme across all ops keeps position error from compounding.
• 3
  Stock allowanceLeave 0.3–0.5 mm on faces that will be finish-machined after heat treatment.

A 3-axis mill cuts from one direction. It is the cheapest way to make a plate with pockets, slots, and drilled holes, and it holds tight tolerance well because the setup is simple. If your part can be reached from six orthogonal directions, 3-axis work plus a couple of fixtures usually beats anything more complex.

A 4-axis machine adds a rotary table, typically Ø400 mm on our mills. That lets the part index to a new face without being unclamped. Long parts with features on four sides, such as shafts with cross-drilled holes or manifolds with ports on multiple faces, land here. You gain position accuracy between faces and lose almost nothing in rigidity.

A 5-axis center adds a second rotary axis, so the tool can approach the part from almost any angle while the part stays clamped. This matters for two reasons. First, contoured surfaces such as impeller blades, turbine housings, and organic brackets can be cut in one continuous pass instead of dozens of re-clamps. Second, short, stiff tools can reach features that would need a long, flexible tool on a 3-axis machine.

The trade-off is real. Five-axis programming takes longer, and the machine hour costs more. We keep 16 simultaneous 5-axis centers alongside 27 three-axis machines and 12 four-axis mills for that reason. The right answer is the simplest machine that reaches every feature without a fixture stack.

• 1
  3-axisPlates, housings, and any part reachable from six sides. Lowest cost per part.
• 2
  4-axisShafts, manifolds, and long parts with features on multiple faces.
• 3
  5-axisContoured surfaces, deep pockets, and parts that must not be re-clamped.
• 4
  Mill-turnRound parts with milled flats or cross-holes; one machine, one setup.

A tolerance callout is a promise about the whole process, not just the cutter. Holding ±0.005 mm on a 50 mm aluminum bore is routine. Holding the same band on a 300 mm stainless housing with thin walls is not, because thermal growth and cutting force both move the material. On long parts we rough, let the part cool, then finish.

Surface finish follows the same logic. Turning and milling as-machined produce roughly Ra 1.6–3.2 μm. A finishing pass with a smaller stepover and a sharper insert reaches Ra 0.8–1.6 μm. For sealing faces, bearing bores, and optical mounts, we can reach Ra 0.2–0.8 μm, usually by grinding, lapping, or a dedicated finish pass rather than by pushing the same tool harder.

Measurement is where the promise is kept. We check raw material before cutting, monitor dimensions during the run, and inspect 100% of parts before shipment. Reports are available on request, including dimensional layouts and material certificates. If a feature is hard to measure in situ, we plan the inspection method before the first cut, not after.

One boundary worth stating plainly: precision on the drawing only holds if the drawing reflects function. Calling out ±0.005 mm on a clearance hole that never touches anything adds cost with no benefit. Tighten only the fits that matter.

• 1
  As-machinedRa 1.6–3.2 μm. Acceptable for brackets and non-sealing faces.
• 2
  Fine finishRa 0.8–1.6 μm. Standard for bearing bores and sliding fits.
• 3
  Optical finishRa 0.2–0.8 μm. Sealing faces and precision mating surfaces.

Aluminum 6061-T6 cuts fast and stays stable, which is why it is the default for prototypes and low-volume housings. 7075 is stronger but more prone to distortion after heavy stock removal, so we rough, stress-relieve, and finish. 2024 behaves similarly and is common in aerospace brackets.

Stainless 304 and 316 work-harden if the tool rubs instead of cutting. The fix is a positive rake, a feed rate that stays above the work-hardening threshold, and plenty of coolant. 17-4PH in the H900 condition is harder again; we machine it in the solution-treated state where possible, then age it, because aging after machining holds dimensions better than cutting hardened stock.

Titanium Ti-6Al-4V conducts heat poorly, so heat concentrates at the cutting edge. Speeds drop, feeds stay high enough to avoid rubbing, and coolant delivery has to reach the edge, not just the part. Inconel is worse on all counts and is normally reserved for features that cannot be made any other way.

Plastics behave differently again. POM and PEEK machine cleanly but hold heat, so light passes and air blast beat flooding with coolant. Carbon fibre reinforced plastics wear tools quickly and need diamond-coated cutters to keep edge quality.

• 1
  AluminumFast, stable, good for prototypes. 7075 needs stress relief after roughing.
• 2
  StainlessKeep the tool cutting, not rubbing, or the surface work-hardens.
• 3
  TitaniumLower speeds, high feed per tooth, coolant aimed at the cutting edge.
• 4
  PlasticsLight passes and air blast; avoid trapped heat and melting.

Machining is not the cheapest way to make a shape. If a part is hollow, thin-walled, and needed in thousands, die casting or vacuum casting will beat it on unit cost once tooling is amortized. Machining wins when the geometry changes often, when the quantity is low, or when the material cannot be cast to the required density.

Deep pockets are the other boundary. A pocket deeper than about four times the cutter diameter forces a long, slender tool that deflects. The result is chatter, taper, and a finish that fails inspection. If the drawing needs a 6 mm wide slot 40 mm deep in stainless, expect to pay for it, and expect to re-think the design if the budget is fixed.

Thin walls have a similar limit. Below roughly 0.8 mm in aluminum and 1.0 mm in stainless, the wall moves under cutting force. We can support it with fixtures or fill it with a low-melt compound, but the cost climbs and the yield drops. Sometimes the better answer is to machine a thicker wall and remove material later, or to switch to a different process entirely.

The honest guide is this: send the drawing early. A DFM review within 12 hours costs nothing and usually finds one or two features that are expensive for no functional reason.

• 1
  High volume, hollow shapeDie casting or vacuum casting usually wins on unit cost.
• 2
  Deep, narrow pocketsBeyond 4× diameter, tool deflection drives cost and scrap.
• 3
  Very thin wallsUnder 0.8 mm aluminum, expect fixture cost and lower yield.