CNC Lathe Guide: How Turning Actually Removes Metal

• 1
  What Happens Where the Tool Meets the Part
• 2
  Which Parts Belong on a Lathe
• 3
  Chucks, Collets, and Why They Change Your Tolerance
• 4
  Where Turning Reaches Its Limits
• 5
  What to Put on the Drawing
• 6
  Turning Method vs. Part Shape

A lathe spins the workpiece and feeds a single-point tool along its surface. That is the whole idea. The spindle holds the part in a chuck or collet, the turret indexes a tool into position, and the carriage moves in X and Z while the spindle encoder reports angular position. Every turned diameter is a function of tool position relative to the spindle axis, which is why roundness on a lathe is usually better than on a mill.

The tool does not shear metal like a pair of scissors. It pushes a thin layer of material past its shear strength, and that layer breaks away as a chip. Heat leaves with the chip, not with the part, as long as the feed and speed stay in range. Push too slow and the tool rubs instead of cuts, which work-hardens stainless and burns the insert. Push too fast and you get chatter, a poor surface, and short tool life.

Depth of cut sets how much material each pass removes. On aluminum, roughing passes of 2–4 mm per side are normal on a rigid machine. On 17-4PH stainless, 0.5–1.0 mm is more realistic. The machine can take more, but the insert and the setup decide the limit. A part held in a three-jaw chuck with 120 mm of unsupported overhang will deflect long before the tool reaches its capacity.

Surface finish comes from the tool nose radius and the feed per revolution. A 0.8 mm nose radius at 0.15 mm/rev produces a theoretical Ra around 1.6 μm. Halve the feed and the finish improves by roughly four times. That relationship is the reason finish specifications are usually met by slowing the feed, not by buying a different machine.

If the part is mostly a solid of revolution, the lathe wins. Shafts, bushings, spacers, pistons, nozzles, valve bodies, connectors, and threaded studs all fall into this group. The geometry is generated by one rotating axis, so setup is simple and cycle time is short. A 30 mm diameter by 100 mm aluminum shaft with two grooves and a thread can often run in under two minutes on a bar feeder.

Parts that need features on multiple faces are a different story. A rectangular housing with bores on four sides, or a bracket with holes at three angles, will need a mill or a mill-turn center. A standard two-axis lathe cannot reach those faces without a second setup, and each extra setup adds a datum shift. That shift is where tolerance stacks start to fail.

Live tooling changes the boundary. A mill-turn center with C-axis interpolation and driven tools can drill cross holes, mill flats, and cut slots without releasing the part. We run 16 mill-turn centers for exactly this reason. The part stays in one chuck, so concentricity between the turned diameter and the cross hole stays inside ±0.005 mm. On a two-setup job, the same feature pair might land at ±0.02 mm.

Length-to-diameter ratio is the other deciding factor. Above about 4:1, a shaft starts to deflect under cutting force. Above 8:1, you need a steady rest or a tailstock, and even then the finish suffers. If your part is 12 mm diameter and 200 mm long, expect to see taper unless the shop supports it properly.

A three-jaw chuck is fast and forgiving, but it repeats to about ±0.05 mm on a good day. A collet repeats to ±0.01 mm or better because it closes evenly around the bar. If your drawing calls for ±0.005 mm on a turned diameter, the part needs to run in a collet or between centers. This is not a machine limit; it is a workholding limit.

Soft jaws bored in place on the machine are the middle ground. They grip a specific diameter and can hold ±0.02 mm on a second operation. The catch is that they only fit that one diameter. Change the part and you bore new jaws, which adds setup time to the quote.

For thin-walled parts, clamping force matters more than tool pressure. A 1.5 mm wall aluminum tube will ovalize in a three-jaw chuck no matter how sharp the insert is. The usual fix is a expanding mandrel or a pot chuck that supports the bore. If your part has a wall under 2 mm, say so on the RFQ. It changes how the job is planned.

Bar feeders also shape what is economical. A part that can be made from bar stock without a second op is cheap to run. As soon as the part needs to be cut off, flipped, and re-clamped, cost climbs. Designing a part so that all critical features are reachable from one side is the single biggest cost lever on a turned component.

Hard materials slow everything down. Titanium Ti-6Al-4V cuts at roughly one-third the surface speed of 6061 aluminum, and the tool wears faster. Inconel is worse. These materials are still turned every day, but the cycle time and insert cost show up in the price. A part that costs 12 USD in aluminum might be 60 USD in Inconel with the same geometry.

Interrupted cuts are the other hard stop. A shaft with a keyway, a cross hole, or a spline presents a broken surface to the insert. Each impact loads the edge. On hardened steel above 40 HRC, that can chip a carbide insert in a few parts. The usual answer is to turn the part soft, then cut the feature by milling or EDM after heat treatment.

Very small diameters and very deep bores push the same limits from the other side. Below about 1 mm diameter, a turning tool has almost no rigidity and deflection dominates. Deep bores beyond 5× diameter need a boring bar that is long and thin, which chatters. Gun drilling or a specialized boring head is a better route when the bore is central and deep.

None of these limits mean the part cannot be made. They mean the process plan has to change. Knowing which limit you are hitting is what separates a quote that holds from one that slips.

Tolerance should follow function, not habit. If a diameter only locates a bearing inner race, it needs a tight band. If it is a clearance surface, ±0.1 mm is plenty and costs less. Putting ±0.005 mm on every dimension does not make the part better; it makes the quote higher and the inspection slower.

Surface finish works the same way. Ra 0.8–1.6 μm is a normal turned finish and covers most sealing and sliding surfaces. Ra 0.2–0.8 μm needs a slower feed, a sharper insert, or a finishing pass, and it adds time. Ra 1.6–3.2 μm is as-machined and is fine for brackets and covers.

Datums matter more than people expect on turned parts. If the drawing dimensions everything from the left face, the shop will hold the part that way. If it dimensions from the centerline of a cross hole, the setup changes. Pick the datum that matches how the part is assembled, and say so explicitly.

Material callout should include temper or condition. Aluminum 6061 and 6061-T6 machine differently. Stainless 316 and 316L differ in work-hardening behavior. 17-4PH in condition H900 is roughly 44 HRC and will not turn cleanly with the same parameters as the annealed bar. A one-word material note leaves the shop guessing.