An optical communication housing is asked to do two things that pull in different directions: fit a module mechanically, and hold an optical reference that does not move with temperature. Most housings pass the first requirement and quietly fail the second, because the second one is not on the drawing.
The two requirements are not the same requirement
Mechanical requirements are dimensional: bore diameter, mounting hole pattern, overall envelope. They are checked with a calliper or a CMM and they are either in or out.
Optical requirements are relational: the alignment between a fibre axis, a lens seat and a detector. They are usually expressed as an angular tolerance or a positional tolerance over a short lever, and they are destroyed by small errors in features nobody flagged.
A 0.01 mm shift across a 15 mm lever is about 0.04°. On a single-mode coupling that is the difference between -0.3 dB and -3 dB. The dimensional drawing is satisfied.
Flatness is the quiet killer, and anodising finishes the job
Mounting face flatness is the single most under-specified characteristic on optical housings. Two things make it worse than it looks:
- Machining flatness is achievable; preserving it is not automatic. A face machined flat to 0.005 mm will not stay there through anodising. The oxide grows into the surface and builds faster on edges, so a sharp edge can lift by more than the flatness tolerance.
- The fix is process, not tolerance. Mask the face, specify hard anodising only where wear matters, or finish-machine the face after coating. Changing 0.01 to 0.005 without changing the process changes nothing.
The same logic applies to a plated face: specify the finish of the face after coating, not before.
Parallelism between the bore and the mounting face
This is the characteristic that escapes dimensional inspection most often, because it cannot be checked with a gauge. It requires a CMM alignment: establish the mounting face as a datum plane, then measure the bore axis over its full length.
Typical specification is 0.01 mm over the bore length. On a 20 mm bore that is 0.03°, which is usually acceptable for a receptacle but not for a collimator seat.
The measurement trap: a bore measured at two circles near its ends will under-report a bowed bore. Four circles along the length will show it.
CTE mismatch, in numbers
When an aluminium housing carries a glass or ceramic subassembly, the two expand differently. Typical coefficients:
| Material | CTE (µm/m·K) |
|---|---|
| Invar 36 | 1.2 |
| Kovar / Alloy 42 | 5.1–5.3 |
| TC4 titanium | 8.6 |
| 316 stainless | 16.0 |
| 304 stainless | 17.3 |
| 6061-T6 aluminium | 23.6 |
| 7075-T6 aluminium | 23.4 |
| PEEK | ~47 |
Over a 30 °C excursion and a 25 mm span, aluminium moves about 18 µm relative to Kovar. If the interface is rigid and the adhesive is stiff, that displacement goes somewhere — into the joint, into the glass, or into a loss of alignment. Over a 60 °C range it doubles.
That is why laser and photodiode packages use Kovar despite its terrible machinability: matching the glass feedthrough's expansion matters more than the machining cost.
Sealing faces and the order of operations
A housing with an O-ring groove has three specifications that interact:
- Groove depth — controls squeeze. Plating inside the groove changes the depth by the coating thickness.
- Groove surface finish — too rough and the seal leaks past the asperities; too smooth and there is nothing for the elastomer to bite.
- Face flatness around the groove** — a bowed face gives uneven squeeze, and the leak appears on one side only.
The sequence that avoids all three problems: machine with plating allowance, mask the sealing face, plate, then finish the sealing face and groove after coating.
Cleaning, because optical surfaces are unforgiving
Machining residue in a housing becomes contamination on a lens. Two practical points:
- Trapped coolant in blind holes and grooves migrates for days. Ultrasonic cleaning with a documented sequence, not a rinse.
- Outgassing matters in sealed packages. PEEK and POM are common housing materials precisely because they outgas less than cheaper plastics, but machining oils trapped in porous surfaces defeat that. Bake-out before sealing is standard for a reason.
Where tolerance stack analysis actually pays
For a housing, the useful stack is not the full-dimensional one. It is the three or four characteristics that chain into optical alignment:
- Mounting face flatness.
- Bore-to-face parallelism.
- Axial position of the lens seat relative to the mounting face.
- Concentricity of the bore to the ferrule seat.
Those four, taken together, predict whether the housing works. Everything else on the drawing predicts whether it fits.
The short version
- Optical housings have a mechanical requirement and an optical one; the second is usually under-specified.
- Flatness of the mounting face is the highest-value characteristic, and anodising can destroy it.
- Parallelism bore-to-face needs a CMM alignment, not a gauge.
- Aluminium against Kovar moves ~18 µm over 30 °C on a 25 mm span.
- Finish sealing faces after coating, not before.
Housing geometries, materials and process sequence are covered on our optical communication housing manufacturer page. Aluminium and Kovar housings, ±0.005 mm on bore and interface dimensions, 0.01 mm flatness on mounting faces.
If you are fighting a coupling loss that changes with temperature, post the housing material and the subassembly material in the comments — the CTE pair usually explains it.
Top comments (0)