CNC Double-Sided Processing: How One Setup Machines Two Faces

• 1
  What CNC double-sided processing actually means
• 2
  Workholding choices that decide the outcome
• 3
  Where the accuracy actually comes from
• 4
  Material behavior on the second face
• 5
  What double-sided processing cannot fix
• 6
  Two setups versus one-setup double-sided work
• 7
  When one-setup double-sided work is worth it

Double-sided processing cuts features on two opposed faces of the same part without releasing the workpiece from its fixture between operations. The machine indexes the part or the spindle, so the second face is machined in the same coordinate system as the first.

That is the whole idea. It is not a separate machine category, and it is not a coating or finishing step. The benefit comes from the coordinate system staying intact.

Two mechanisms are common on our floor. A trunnion table rotates the part 180° in one continuous program, which suits housings, brackets and manifolds. A mill-turn center grips the part in a sub-spindle instead, so a shaft can be faced and bored on both ends while the bar stock never leaves the chuck.

Either way, the second face inherits the first face's datum. That inheritance is what separates this method from flipping a part by hand on a vise.

The fixture decides whether one-setup work pays off. A trunnion table on a 5-axis center gives the part a clean 180° index, and a Ø400 mm rotary table covers most housings we see. Self-centering vises hold a repeatable position to about 0.01 mm, which is enough for general work but not for tight bores.

Zero-point clamping is the better answer when faces must stay parallel. Pallets repeat to roughly 0.005 mm, and the operator never touches the part between faces. That repeatability is what lets us hold ±0.005 mm across both sides rather than on one face only.

Thin walls need support, not force. A part with 1.5 mm walls will deflect under a standard vise, so we back it with a custom soft jaw or a low-melt fixture. Cutting forces drop with smaller radial engagement, typically 5–8% of tool diameter for finishing passes.

Deep cavities on the second face are the hard case. The tool has to reach past the first face without chattering, so we shorten the gauge length and accept a smaller stepover.

Accuracy in double-sided work is mostly a geometry problem, not a spindle problem. When both faces are cut from one datum, the only errors that survive are the machine's own positioning errors and thermal drift. Re-clamping a part adds its own error on top of those.

Thermal drift matters on long cycles. A 4,000 mm part on a large travel machine can grow more than 0.02 mm over a warm afternoon, so roughing and finishing on both faces are kept close together in the program, not split across shifts.

Tool wear is the second variable. A worn Ø12 mm end mill pushes harder and deflects more, which shows up as taper in a deep bore. We change tools on a count basis rather than waiting for the finish to drift out of band.

Surface finish ties into the same picture. A stable setup holds Ra 1.6–3.2 μm as machined, Ra 0.8–1.6 μm with a finishing pass, and Ra 0.2–0.8 μm when the part needs a fine finish on a bearing face.

Symmetry is the quiet benefit. A part cut from one datum ends up balanced because both faces see the same tool and the same offsets, which matters for anything that rotates.

Aluminium alloys such as 6061, 7075 and 6082 cut cleanly on both faces and hold tight tolerances without much fuss. Heat-treated 7075 can move after the first face is removed, so we leave a 0.3–0.5 mm allowance and take it off in the finishing pass.

Stainless 304 and 316 work-harden if the tool rubs. On a double-sided part the second face often has more stock to remove, so we keep the feed per tooth up and never let the cutter dwell. 17-4PH in the H900 condition behaves more like a hard steel and needs a rigid setup.

Titanium Ti-6Al-4V and Inconel generate heat at the cutting edge. Both faces are cut with high-pressure coolant and reduced speed, and the part is checked for growth between faces rather than only at the end.

Plastics and carbon fibre behave differently again. POM and PEEK hold well; carbon fibre needs diamond tooling and dust extraction, and the second face must be cut before the edges fray.

Mixing materials in one assembly is common. A 6061 housing with a 316L insert can be machined in one setup if the insert is pressed in after the second face is finished.

It cannot rescue a bad design. If one face needs a 0.5 mm wall and the other needs a deep bore, no setup will keep the part straight. The geometry has to allow the tool to reach both faces with a sensible length-to-diameter ratio.

It does not remove the need for a datum. Double-sided work concentrates accuracy into one coordinate system, so if that datum is wrong, both faces are wrong together. We check the raw stock before the first cut for that reason.

Size is a real boundary. Our largest travel is 4,000 × 400 × 150 mm, and medium machines cover 750 × 1,150 × 550 mm. A part beyond those envelopes has to be split into two setups regardless of what the drawing asks for.

It also adds programming time. A double-sided program needs a full simulation of the index move, and collision checks around the fixture take longer than the toolpath itself.