4-Axis CNC Machining: How the Rotary Axis Changes the Cut

• 1
  What 4-axis CNC machining adds to a 3-axis mill
• 2
  Indexing versus simultaneous motion, and when each applies
• 3
  Which part shapes actually justify the fourth axis
• 4
  Accuracy, surface finish, and the limits of the rotary axis
• 5
  Setup planning, workholding, and programming notes
• 6
  3-axis, 4-axis indexed, 4-axis simultaneous: which fits the part

A 3-axis mill moves the tool in X, Y and Z. Add a rotary table and you get 4-axis CNC machining: the workpiece can now turn while the cutter works, instead of being repositioned by hand between setups. That fourth axis is normally A (rotation about X) or B (rotation about Y). Which one you get depends on how the table is mounted on the machine bed.

The rotary table is where the whole story lives. On our 4-axis mills the table is Ø400 mm, which sets the practical envelope for anything you want to spin. A part that swings wider than the table can handle will hit the casting or the enclosure long before the tool does. Check the swing diameter before you check anything else.

The fourth axis does not add reach. It adds orientation. The tool still has to get to the feature, and the part still has to fit inside the machine travels. What changes is that features on several faces can be cut in one setup, with the same work offset and the same tool. That is the real gain, and it shows up as fewer setups rather than faster feed rates.

One more thing worth clearing up early. Four axes means four controlled motions, not four directions of cut. On most work the cutter still approaches from one side at a time. The rotary axis simply presents a new face to that same cutter. Keep that picture in mind and the rest of the trade-offs fall into place.

Indexing is the common case. The table rotates to a position, locks, and the cut starts. Then it rotates to the next position and locks again. This is 3+1 work: three linear axes plus a repositioned fourth. It is accurate, easy to program, and forgiving of a slightly loose rotary table because nothing moves while the tool is in the material.

Simultaneous motion is the harder case. All four axes move under interpolation, so the tool follows a path while the part turns. This is what lets you cut a continuous cam profile, a spiral groove, or a port with a changing radius. The controller has to keep four axes in step, and any backlash in the rotary drive lands directly on the surface finish.

The decision rule is simple. If every feature can be reached by stopping and turning, use indexing. If the geometry only exists as a curve in rotary space, you need simultaneous motion. Mixing both on one part is normal and often the cheapest route: index for the flats and bolt circles, go simultaneous only for the one curved feature that demands it.

Simultaneous work also costs programming time. A four-axis simultaneous toolpath has to be verified for collisions across the full rotation, not just at a few index positions. That verification is where most of the setup hours go on a new part. On a repeat order it costs nothing, which is why the first article is always the expensive one.

Cylindrical parts with features around the circumference are the classic case. Think a shaft with cross-drilled holes at several angles, or a flange with slots every 30°. On a 3-axis mill each angle is a separate setup, and each setup adds a re-clamp error. On a 4-axis mill the table indexes and the holes come off one datum.

Parts that are long and thin also benefit, but for a different reason. A 300 mm shaft held in a vise will deflect under cut. Held between the rotary table and a tailstock, the same shaft is supported at both ends and the deflection drops. This is a rigidity gain, not an orientation gain, and it often matters more than the axis count.

What does not need a fourth axis: flat plates with all features on one face, prismatic housings where every wall is reachable from above, and anything small enough to be flipped by hand in a few seconds. Putting these on a 4-axis machine wastes the rotary table and usually costs more per part.

There is a middle group that is easy to misjudge. A part with features on two opposite faces can be done on a 3-axis mill with a flip, or on a 4-axis mill with a 180° index. The flip needs a second work offset and a re-clamp. The index needs a tailstock and a longer setup. Count the setups and the true position tolerance, then pick. Neither answer is wrong in the abstract.

Linear axes on a well-kept mill hold ±0.005 mm without much drama. The rotary axis is a different animal. Angular error converts to linear error through the radius, so a 20 arc-second error at a 200 mm radius is roughly 0.02 mm of position error. Long parts amplify rotary error; short parts hide it.

Surface finish behaves the same way. Indexed cuts finish like any 3-axis cut, typically Ra 0.8–1.6 μm, and can reach Ra 0.2–0.8 μm with a fine stepover and a sharp tool. Simultaneous cuts leave a scallop pattern whose pitch depends on feed, rotary speed and tool radius. If the finish looks banded, the rotary drive is usually the cause, not the cutter.

Backlash is the usual culprit. A worn rotary worm drive can carry 0.01–0.03° of lost motion, which shows up as a witness mark every time the table reverses direction. The fix is mechanical, not a controller setting: preload the drive or use a direct-drive table. Ask what the machine's rotary repeatability is before you accept a tight true-position callout on an indexed feature.

Thermal drift matters on long cycles. A rotary table that runs for hours warms up, and the centre height creeps. On a part with a ±0.02 mm true position across many indexed features, that creep is measurable. We probe the datum after the first article and re-check it at the end of a long run rather than trusting the initial offset for the whole batch.

Workholding drives the design more than the machine does. A three-jaw chuck, a collet block, a faceplate with soft jaws, or a fixture plate with a tailstock: each one sets how much of the part you can reach and how rigid the cut will be. A part that needs a tailstock loses the free end for through-features. Plan that before you quote.

Balance matters once the part spins. A part with an off-centre boss will want to whip at speed, and the table will let you know. Either add a counterweight or keep the rotary speed low during simultaneous moves. On long, thin parts we often cap rotary speed rather than chase cycle time, because chatter costs more than the minutes saved.

Program from a single datum. The point of the fourth axis is fewer setups, and that only pays off if every indexed face is driven from the same work offset. If the programmer re-datums at each index, you have rebuilt the 3-axis setup problem inside a more expensive machine. One datum, one offset, verify with a probe.

Leave stock for the rotary transition. The first cut after an index often has a small step at the joint because the tool re-enters slightly off. A 0.1–0.2 mm finishing allowance on the affected face lets the finishing pass clean that up. It is a cheap habit that saves a lot of rework on the last operation.