Machining Optical Components: Mounts, Tolerances, and Keeping Light on Axis

An optical mount looks like a bracket until you put a lens in it. Then it stops being a bracket and becomes half of an optical system, and the machine work that was invisible before is suddenly the difference between a clean image and one that is soft at the edges, tilted, or drifting off axis. These parts are not harder to machine than anything else, but they are judged by a different number: where the light lands. That changes what matters, from the concentricity of a lens barrel to the flatness of a mirror seat, and it changes the way you fixture, the order you cut, and the finish you leave behind.

Why optical parts are not ordinary precision parts

Most machined parts are asked to hold a dimension so two things fit together. An optical part is asked to hold a dimension so a beam of light goes where it should, which is a stricter and stranger requirement. A bracket can be a few microns off and still assemble and work. A lens seat that is a few microns out of concentricity shifts the image and changes the focus, and the error is not caught by the mating part the way a bad fit would be — it shows up later, as a system that will not align or a beam that will not focus to the same spot every time.

The parts themselves are usually unremarkable on the surface: lens barrels, mirror cells, filter housings, and the plates and posts that hold optics in place. What makes them optical is the stack of tolerances that controls where the optical axis sits relative to the mechanical features. A barrel needs its bore concentric with its threads, its seat face square to the axis, and its flange parallel to the seat, all at once. Miss any one of those and the lens tilts or the spacing between elements shifts, and the whole assembly falls out of collimation. The machining is the easy part; the hard part is keeping every feature referenced to the same axis through every setup.

Centration and concentricity are the whole game

The single most important requirement on a lens barrel or an optical mount is centration — that the optical axis of the glass lines up with the mechanical axis of the housing. If the bore the lens sits in is not concentric with the thread that holds it, the glass cants, and a canted lens produces a tilted image plane. The practical control is to machine all the features that locate the optics in a single setup, so they share one axis by construction rather than by luck. A barrel bored and threaded in the same chucking has its seat and its thread concentric to the spindle axis by definition; bore it in one setup and thread it in another, and the error between the two is whatever the fixture can hold, which is never zero.

This is why so much optical work ends up on lathes and turn-mill machines even when the part has milled features. The axis is the point. When a part has to be moved between setups, the way to preserve the axis is to machine it on a shared datum — a turned spigot or a reference face that every later operation locates from — and to inspect concentricity the way the light will see it, as a runout measurement on the locating surface rather than as a wall-thickness reading that hides the tilt.

Concentricity also has to survive the finish. Anodizing builds thickness on a bore and a thread at the same time, but not evenly on corners and flats, so a barrel that was concentric bare can shift slightly when it comes back from the coater. The fix is to machine to the pre-finish dimension with the coating thickness already accounted for, and to mask or protect the locating surfaces where the drawing allows, so the axis set in the machine is the axis that ships.

Flatness and the surface that touches the glass

The face a lens sits on has to be flat, and flat in a way that does not distort the glass. A lens element is stiff in the center and flexible at the edge, and if its seat is not flat the element rocks or is pushed into a slight bend when the retaining ring is tightened, which shows up as a loss of sharpness across the field. The seat is usually a small annular face, and getting it flat is less about the cutter and more about how the part is held while the face is cut. Clamp a thin-walled barrel hard and the bore goes slightly out of round; cut the seat in that state and the face comes out flat only while the clamp is on, and springs back when the part is released.

The same rule as any thin-walled work applies: hold the part lightly, at the faces it will be bolted on in service, and take the finish cut on the seat with a sharp tool and a light pass so the cut does not push the wall around. The surface finish on the seat matters too, because a rough face concentrates the retaining-ring load on the high spots and stresses the glass unevenly, while a smooth, clean face spreads the load. The burr left by the cutter is a third kind of problem — a raised edge around the seat will tip the lens exactly as if the seat were tilted, so edge-breaking the seat and the bore is part of holding flatness, not just an appearance detail.

There is also the quiet requirement that the seat not be shiny where it should not be. Stray light is the enemy in an optical path, and a machined surface that catches a reflection can put a ghost image where the designer did not expect one. A controlled matte finish, or a clean anodized surface, is often the difference between a mount that works and one that adds glare. Here the machining and the finishing have to agree: the finish spec on an optical seat is as much an optical property as a mechanical one.

Threads and spacing that survive the coating

Optical assemblies are full of threads — retaining rings, barrel caps, filter rings, and the focus threads that move one group relative to another — and these threads carry more of the optical job than threads usually do. A focus thread sets the spacing between elements, so its pitch and its concentricity directly control focus. A retaining-ring thread has to run true so the ring lands square and presses the lens evenly instead of on one side. Cut these threads in the same setup as the bore they are concentric to, and the whole locating stack stays on one axis.

Coating is the complication. Anodizing adds thickness to a thread, and if the thread was cut to its finished class before coating, it will bind or change its fit when the coating goes on. The standard move is to cut the thread slightly oversized or to a controlled pre-coat dimension so the added thickness lands it back on its target class, and to keep the coating thickness consistent so the fit comes out predictable. The same logic applies to every locating surface on the part: machine to the dimension that will be correct after the coating, not the dimension that is correct in bare metal.

The practical sequence is to fix the coating thickness first, machine the locating features to account for it, and then treat the coated part as the finished article that gets its final fit check. A shop that machines a barrel to print in bare aluminum and anodizes it blindly will hand back a part whose lens seats no longer center the glass, even though every bare-metal dimension measured perfectly.

Fixturing optics: holding the part without bending the axis

Because so much of an optical part's value lives in its axis and its flat seats, the fixturing is where the job is won or lost. The part is usually thin-walled and easily distorted, so the clamp has to hold it firmly enough to cut but not hard enough to bend it. Soft jaws bored to the finished diameter grip a barrel around a large area and keep it round, while a hard three-point clamp will spring a thin wall out of round the moment the part is released. A vacuum fixture or a faceplate holds a plate-shaped mount flat without the local distortion that a set of clamps introduces, which matters when the seat being cut is the very face doing the sealing.

Where the part has to be re-fixtured, the rule is to re-locate from a machined datum that was cut concentric to the axis, never from a raw surface or an unrelated feature. A turned reference diameter or a lapped face carried through the setups keeps every operation on the same axis, which is what lets a part move between a lathe and a mill without losing its centration. The inspection follows the same idea: measure runout and squareness relative to that datum, the way the light will, rather than relative to a convenient surface that happens to be reachable.

Temperature deserves a mention, because optical tolerances are tight enough that the part's own temperature matters. A large aluminum barrel grows measurably between a cold start and a warm afternoon, and if the bore is finished at one temperature and measured at another, the reading shifts by more than the tolerance band. The discipline is to let the part stabilize, machine and inspect at a consistent temperature, and keep the hands and the chuck from putting heat into the work in the middle of a finish pass. It is the same rule a metrology lab lives by, applied on the shop floor because the numbers are small enough that a few degrees shows up in the result.

Machining to the light, not just the print

What separates an optical part from a look-alike bracket is not the difficulty of the cuts but the discipline of keeping every feature referenced to one axis and every surface finished for the way it will be used. The print asks for concentricity, flatness, and a finish; the optical system asks for a lens that centers, a seat that does not bend the glass, and a thread that keeps the spacing true after coating. The machinist who gets it right is the one who cuts the locating stack in one setup, machines to the pre-coat dimension instead of the bare-metal dimension, holds the part lightly at its service faces, and inspects runout the way the beam will see it.

None of this is exotic. It is ordinary turning and milling done with an uncommon attention to what the axis is doing and what the coating will change. Get the centration and the flat seats right and the rest of the part is routine careful work; get them wrong and the part measures perfectly and still produces an image that is soft, tilted, or full of ghosts, which is the worst kind of failure because it is invisible until the system is assembled and the light will not behave.

• Machine the locating stack in one setup — bore, thread, and face together so they share one axis by construction.
• Carry a machined datum through every setup — re-locate from the axis, never from a raw surface.
• Hold the part lightly at its service faces — soft jaws or a faceplate, not a hard clamp that springs the wall.
• Cut the seat flat and break its edge — a burr around the seat tips the lens exactly like a tilt.
• Machine to the pre-coat dimension — account for anodizing thickness on bores, threads, and locating faces.
• Finish for the light, not just the fit — control reflection and scatter on surfaces in the optical path.
• Inspect runout relative to the datum — measure the way the beam will see it, not a wall-thickness shortcut.
• Let the part stabilize in temperature — a few degrees moves the bore by more than the tolerance.

"An optical mount is not finished when it measures right; it is finished when the glass sits on axis, the seat does not bend the element, and the thread holds its spacing after the coating goes on. That is a concentricity problem, a flatness problem, and a coating problem wearing the clothes of a machining job, and the part is won or lost in how you keep every feature referenced to one axis."

Optical components reward the shops that treat them as an axis problem rather than a geometry problem. Cut the locating stack in one setup, hold the part without bending it, machine to the post-coating dimension, and finish the seats for the way the light will use them, and the part comes off the machine ready to center its glass. Skip any one of those and the part measures fine and still produces an image that will not align.

If you are machining lens barrels, mirror cells, or optical mounts and the current results are soft, tilted, or drifting off axis, send the drawing over and we will review the setup, the datum, and the coating allowance behind the problem. See how we approach CNC turning for concentric bores and threads and multi-axis milling for optical housings, or get a quote on your part.