CNC for laboratory equipment: where accuracy actually comes from

• 1
  What makes laboratory equipment parts hard to machine
• 2
  How five-axis machining changes what is possible
• 3
  Tolerances and surface finish: what to specify and why
• 4
  Material selection for chemical and thermal exposure
• 5
  Inspection: how you know the part is correct
• 6
  When CNC is the wrong process for a lab part
• 7
  Choosing a process for laboratory equipment parts

A lab instrument is a measurement chain. Every machined part sits inside that chain, so its error adds to the error of everything else. A grating mount that is 20 μm off center shifts the diffracted beam. A manifold port that is 10 μm out of flat leaks at 10⁻⁶ mbar. Nothing in the instrument absorbs that error. It shows up in the data.

The second constraint is scale. Many features are small: 0.2 mm fluidic channels, 0.5 mm pinholes, 1.5 mm laser-etched labels. Others are large: a 4,000 mm optical bench rail that must stay flat. Few machine shops handle both ends in the same order.

The third constraint is environment. Parts may see acids, solvents, autoclave cycles at 134 °C, liquid nitrogen, or continuous vacuum. Material choice and surface finish stop being cosmetic. They decide whether the part lasts five years or five months.

So the question is not whether a shop owns a CNC machine. It is whether the shop controls the variables that matter for the instrument. That control is what we sell.

A three-axis mill moves the tool in X, Y, and Z. The part stays in one orientation, so every feature must be reachable from one direction. Deep side pockets, angled ports, and curved channels need multiple setups. Each setup adds a re-clamping error, typically 10–30 μm even with good fixturing.

A five-axis machine adds two rotary axes. The tool reaches the part from many directions in one setup. On a 16-machine five-axis fleet, that means one datum for the whole part. For a manifold with ports on four faces, setup count drops from four to one. The tolerance stack drops with it.

The rotary axes also let the tool stay normal to a curved surface. On a spherical mirror housing or a curved microfluidic channel, the cutter engages the surface at a constant angle. Scallop height falls, and the tool marks stay uniform. We can hold Ra 0.8–1.6 μm directly off the machine, and Ra 0.2–0.8 μm after finishing.

There is a limit. Five-axis does not fix a part with a deep, narrow slot that no tool can reach. It does not fix a thin wall that deflects under cutting force. It does not fix a design that needs a feature inside a closed cavity. Those cases need a different process, or a design change.

• 1
  One setup, one datumFewer re-clamps means a smaller error stack on multi-face parts.
• 2
  Constant tool engagementBetter finish on curved and spherical surfaces.
• 3
  Not a cure-allDeep slots and unsupported walls still deflect.

Tolerance is not one number. A drawing that says ±0.005 mm everywhere will cost more than it needs to. Split the drawing into functional groups. Optical mounting surfaces, sealing faces, and bearing bores carry the tight callouts. Clearance holes and cover plates do not.

For optical alignment, position tolerance on the mounting face controls the beam path. We machine these to ±0.005 mm and inspect them on the machine before the part leaves the fixture. For sealed surfaces, flatness matters more than size. A 0.01 mm flatness callout on an O-ring groove face is often enough to hold vacuum, if the finish is fine.

Surface finish interacts with tolerance. A rough face will not seal even if it is flat. Ra 0.2–0.8 μm suits elastomer seals and clean fluid paths. Ra 0.8–1.6 μm suits general mating faces and metal-to-metal contact. As-machined Ra 1.6–3.2 μm is fine for brackets and frames.

The achievable combination depends on part size, geometry, and material. A 30 mm aluminum bracket holds tighter numbers than a 900 mm stainless rail. We confirm the real number after reviewing the drawing, not before.

Start with the exposure, not the strength. For strong acids and solvents, 316L stainless, PEEK, and PTFE-family plastics resist attack better than 6061 aluminum. For autoclave cycles at 134 °C, PEEK and 316L hold up; polycarbonate and ABS do not.

For optical and thermal stability, aluminum 6061-T6 and titanium Ti-6Al-4V are common. Aluminum moves more with temperature but is easier to machine and lighter. Titanium moves less and resists corrosion, but it cuts slowly and costs more.

For vacuum service, avoid porous castings and trapped volatiles. Machined 304, 316L, and 6061 are standard. Electroless nickel plating gives a hard, non-porous surface on aluminum chambers and flanges.

We keep a wide stock list: aluminum 6061, 7075, 2024, 5052, 5083, 6082; stainless 303, 304, 316L, 17-4PH; steels 1018, 1045, 4140, 4340; copper and brass C101, C110, C36000; titanium TA2, TC4; plastics POM, PEEK, PC, PMMA, and carbon fiber. Material choice is a joint decision with your engineering team.

• 1
  Chemical resistance first316L and PEEK for aggressive reagents.
• 2
  Autoclave serviceAvoid PC and ABS; use PEEK or 316L.
• 3
  Vacuum serviceMachined, non-porous stock; plating helps on aluminum.

A tolerance callout is a promise. Inspection is the proof. We check raw material certificates on arrival, monitor dimensions in process, and inspect 100% of parts before shipment. Reports are available on request.

For critical features, we measure on the machine while the part is still in the fixture. If a bore is out, we correct it before unclamping. This avoids the common failure where a good part is re-clamped for inspection and loses its datum.

For sealing faces and optical surfaces, we check flatness and finish with the same instrument set your incoming inspection uses. That keeps the numbers comparable. If your drawing calls for a first article report, we build it against the same drawing revision you hold.

This is also where the ISO 9001:2015, IATF 16949:2016, ISO 13485:2016, and ISO 27001:2022 systems matter. They are not badges. They define how records are kept and how changes are controlled.

CNC is subtractive. It removes material from a solid block. That is ideal for a manifold with internal channels, a one-off optical mount, or a 20-piece run of instrument frames. It is less ideal for a thin-walled enclosure with 0.5 mm walls and complex ribs, where sheet metal or die casting wins on cost.

It is also the wrong choice when the part is a hollow sphere, a closed cavity, or an internal lattice with no tool access. Additive manufacturing handles those shapes. Vacuum casting makes sense for 20–100 urethane parts. Die casting makes sense above a few thousand units in a single alloy.

There is a middle ground. A machined prototype often becomes the master for a cast or molded production part. We machine the prototype, verify the fit, and the same geometry feeds the tooling. That is usually faster than jumping straight to tooling.

The honest rule: if the part needs tight tolerances, a small quantity, or a material that cannot be cast well, CNC is the right call. If it needs thin walls and high volume, it is not.