Lidless Fiber Array: Why 1.5 mm Height Matters for Co-Packaged Optics
A lidded fiber array is 2.5 mm tall; a lidless one is 1.5 mm, with identical insertion loss, return loss and pitch tolerance. In co-packaged optics that millimetre often decides whether the interface fits at all.
A lidded fiber array stands 2.5 mm tall. A lidless one stands 1.5 mm. That single millimetre sounds trivial until you are trying to fit an optical interface underneath a heat sink in a co-packaged optics module, at which point it decides whether the design closes.
What the lid is for, and what it costs
In a conventional fiber array the fibers are seated into the V-grooves and a glass lid is bonded over the top. The lid does real work: it holds the fibers captive during bonding and polishing, and it protects the end face afterwards. That is why it is the default.
It also adds a millimetre of stack height and closes off optical and mechanical access from above. In most applications that is an acceptable trade. In co-packaged optics it increasingly is not.
| Standard (lidded) | Lidless | |
|---|---|---|
| Outline height | 2.5 mm | 1.5 mm |
| Channel count | 1 – 128 | 1 – 32 |
| Core pitch | 127 / 250 µm | 127 / 250 µm |
| Pitch tolerance (1–8CH) | ±0.3 µm | ±0.3 µm |
| Insertion loss | ≤ 0.5 dB (typ. 0.3) | ≤ 0.5 dB (typ. 0.3) |
| Return loss | UPC ≥ 45 dB / APC ≥ 55 dB | UPC ≥ 45 dB / APC ≥ 55 dB |
The optical specifications are identical. Width, length and pitch tolerance carry over unchanged — removing the lid does not relax the V-groove accuracy, because the grooves are cut the same way either way. What changes is height and access.
Why height is the binding constraint in CPO
Co-packaged optics moves the optical interface from the faceplate onto the substrate next to the switch ASIC. That eliminates a long electrical run and the power it burns, but it puts the fiber attach point into a region that is already crowded: the ASIC and its heat sink above, the substrate below, and a vertical budget measured in millimetres rather than centimetres.
In that context, a millimetre is not a rounding error. It can be the difference between a fiber array that fits under the thermal solution and one that forces a redesign of the lid, the heat sink, or the module height itself.
Direct access is the second reason
With no lid, the tops of the fibers stay reachable after bonding — optically and mechanically. Where an assembly process needs to inspect, illuminate or reference the fiber tops during or after attach, a lidded array simply cannot offer that. This is usually the reason customers who are not height-constrained still ask for the lidless version.
The trade you are accepting
Without a lid, the fibers are not captured from above and the polished end face is exposed. Neither is a problem in a controlled assembly line, but both need to be planned for:
- End-face protection in transit and handling. Arrays can be shipped with a temporary cover, or in fixtures matched to your process, rather than bare.
- Downstream process compatibility. Cleaning and adhesive steps that assume a lid may need revisiting.
- Channel count. The lidless configuration is specified to 32 channels, against 128 for the standard array.
The constraint that defines the part
Our production drawing for an 8-channel lidless array carries a note that says more about this product than any specification line:
That one sentence is the whole engineering problem. With a lid, adhesive height is uncritical — the lid caps the stack and defines the top surface. Without one, the adhesive is the top surface, and if it wicks above the fiber tops you lose both reasons the part exists: the height advantage disappears, and the fiber tops are no longer accessible.
Controlling that means controlling adhesive volume, viscosity and cure profile per build, not applying a generic bonding recipe. It is also why we run adhesive dispensing on dedicated automated equipment rather than by hand — the tolerance is set by the fiber diameter, and there is no lid to hide behind.
Other details from the same drawing are worth flagging because they are the kind of thing that gets left out of an RFQ and then causes a re-spin: the V-groove substrate is quartz glass, the bonded region is 10 mm long, fiber exits the array bare before transitioning to 0.9 mm loose tube, and the connectors are FC/APC narrow key — not standard key. On a PM build the key type is what your axis reference is measured against, so it is not a detail you want assumed.
How thin it actually goes
The 1.5 mm figure is the specified standard. Substrate outline is one of the parameters we cut to drawing, and shipped configurations include arrays with 1 mm bodies — for example a 16-channel array on 127 µm pitch with a 12 × 6 × 1 mm outline, 12° polish, SMF-28e, FC/APC terminated.
Polish angle is equally open. Across shipped builds we have delivered 0°, 2.62°, 8°, 10°, 12° and 45° end faces — the last for turning the beam rather than suppressing back-reflection. If your interface needs an angle that is not in the standard set, it is a drawing question, not a capability question.
Specifying one
The information that actually determines the build:
- Channel count and pitch (127 or 250 µm, or custom)
- Maximum outline height available in your stack-up
- Polish angle, and whether the end face needs AR coating
- Fiber type — SM, PM or mode-matched
- Whether your process requires access to the fiber tops after bonding
That last point is the one most often left out of an RFQ, and it is the one that decides whether lidless is the right answer at all.