Basic Knowledge of CNC Lathes

A CNC lathe rotates the part and moves the tool. A mill does the reverse: the cutter spins while the part stays clamped. That difference sets the rule for process selection. Features built around a single centerline turn faster and cheaper. A pocket, a slot, or a hole off the rotation axis pushes the job toward milling.

Most turned parts are not pure. A shaft with a cross hole, a flange with bolt patterns, a fitting with milled flats — these need both operations. That is where mill-turn centers earn their place. We run 16 mill-turn centers that turn and mill in one setup, which removes a second fixture and the position error that comes with it.

The practical question is not lathe or mill. It is how many setups the part needs. Every extra setup adds a datum, a clamp mark, and time. If your part can be finished in one turning setup plus one milling setup, the cost is predictable. If it needs four, ask why.

A basic 2-axis lathe moves the tool in X and Z. It turns outside diameters, faces, bores, and threads. This covers a large share of shafts, bushings, and fittings. It is also the cheapest way to make them.

Adding a Y axis and a second spindle turns the machine into something closer to a machining center. Live tooling lets the turret drill and mill while the part is still chucked. Sub-spindle machines pass the part from the main spindle to the back spindle and finish the second face without a human touching it. On high-volume work this cuts cycle time and operator handling.

The trade-off is setup and programming time. A 5-axis turn-mill center takes longer to dial in than a 2-axis lathe. For a run of 20 simple pins, that overhead does not pay back. For a run of 2,000 hydraulic manifolds with cross ports, it pays back quickly.

A lathe holds the part by gripping it. That grip is also the main source of error. A three-jaw chuck is fast but repeats to about 0.025 mm unless it is bored in place. A collet holds better and repeats closer, but it only fits one bar diameter. Between centers with a face driver, the part is located on its own centers, which is the most repeatable method for shafts.

Thin-wall parts are the hard case. Chuck pressure squeezes the wall into an oval, the tool cuts a round bore in a deformed part, and the bore springs back oval when the jaws open. The fix is not tighter tolerance on the machine. It is lower clamping force, a piejaw or expanding mandrel, and lighter passes.

We hold ±0.005 mm on turned features when the setup supports it. That number is not a property of the lathe alone. It comes from the collet, the material, the wall thickness, and the pass strategy together. A 0.5 mm wall in aluminium will not hold the same tolerance as a solid 40 mm shaft.

Turning uses single-point tools, so the tool nose radius shows up directly in the surface. A larger nose radius leaves a smoother finish at the same feed but pushes radial force up, which matters on slender parts. A small radius cuts freer but leaves a rougher surface and wears faster.

Feed rate is the main lever on finish. At a given nose radius, doubling the feed roughly quadruples the theoretical peak-to-valley height. If a drawing calls for Ra 0.8–1.6 μm, a light finish pass with a wiper insert usually gets there in aluminium and mild steel. Hardened 17-4PH or Inconel needs a slower speed, a more rigid setup, and sometimes a ground finish to reach the same number.

Insert grade matters as much as geometry. Coated carbide covers most steel and stainless. Aluminium wants a polished, uncoated insert to stop built-up edge. Titanium wants sharp edges, low speed, and plenty of coolant. The wrong grade shows up as chatter, poor finish, or short tool life, not as a clean cut.

A turning drawing lives or dies on datums. If the bore and the outside diameter are both called out from the same end face, the machinist can hold both in one setup. If they reference opposite ends, the part needs a flip and the tolerance stack grows. Put the functional datum first and let the rest follow from it.

Call out roundness and concentricity only where they matter. A blanket ±0.005 mm on every dimension raises cost and does not improve function. A shaft that only needs to fit a bearing should hold tight tolerance on the bearing seat and normal tolerance elsewhere. We see drawings that specify the whole part to a level the assembly does not need.

Threads, radii, and chamfers are often left to the machinist. That is usually fine, but if a thread class matters, state it. A 1/4-20 UNC 2B is not the same as a generic 1/4-20. The same goes for surface finish: Ra 0.8–1.6 μm on a seal face is a real requirement, and Ra 3.2 μm on a bracket is not worth paying for.