Home / Blog / Slow Axis to Key

Slow Axis to Key: The PM Fiber Array Spec Everyone Skips

September 4, 2026

On a PM array the polarization axis is referenced to the connector key — not to the substrate. Get the key width wrong and the part is built correctly, passes test, and still drifts in your system.

PM ComponentsFiber ArrayExtinction Ratio

12-channel PM fiber arrays packed for shipment with FC/APC connectors

12CH PM460-HP + PM1950 arrays ready to ship. Every connector on both rails is keyed to the same axis reference — that is the part of the build this note is about.

Almost every PM fiber array we ship carries one line in its build spec: slow axis to key. It reads like boilerplate. It is the single most common reason a part measures correctly at our end and still arrives “rotated” at yours.

What the key actually references

A polarization-maintaining fiber has two orthogonal axes with different refractive indices — the slow axis and the fast axis, set by the stress rods either side of the core. Launch light aligned to one of them and the polarization state is preserved down the fiber. Launch it at some angle in between and the two axes beat against each other, and the extinction ratio collapses.

So the array is useless unless you know where the slow axis points. And the answer is not “relative to the V-groove substrate” — it is relative to the connector key. The key is the datum the customer can actually see and mate against.

Standard key orientation: slow axis aligned to the connector K key — Standard key orientation from our 16CH build drawing: the panda stress rods — and therefore the slow axis — line up with the K key.

That is what “slow axis to key” means: when you look into the ferrule with the key pointing up, the stress rods sit on the vertical. Nothing else about the part tells you this. Take the key away and the orientation is unknowable without a polarimeter.

Why a couple of degrees matters so much

Angular error between the launch polarization and the fiber axis sets a hard ceiling on extinction ratio, no matter how good the fiber is. The relationship is pure geometry:

ER_max = −10 log₁₀(tan² θ) where θ is the angle between the launch polarization and the fiber axis.

Axis error	Best achievable PER
0.5°	41 dB
1°	35 dB
2°	29 dB
3°	26 dB
5°	21 dB
10°	15 dB

Read the bottom row again. Ten degrees of rotation caps you at 15 dB, and there is nothing in the fiber, the array or the polish that can recover it. Our alignment accuracy is specified at ≤ ±2°, typically ±1° — which is why the array itself is not usually the limiting term. The rotation that hurts is the one introduced after the part leaves us.

Where that rotation comes from: narrow key vs wide key

FC connectors come with two mutually incompatible key widths, standardised under TIA FOCIS-4:

Narrow key, 2.0 mm (also called Type R or reduced) — the usual choice for FC/APC and PM
Wide key, 2.14 mm (Type N) — the NTT-style standard

They look identical at a glance and they will mate. A narrow-key connector drops happily into a wide-key adapter — and then it can rotate inside the sleeve, because there is 0.14 mm of slack the key was never meant to have. For an ordinary single-mode PC link that is tolerable. For a PM link it means the axis reference you paid for is no longer being enforced by anything.

The failure mode is quiet and intermittent. The part tests fine on a matched adapter, goes into a system with a wide-key sleeve, and the PER wanders with every remating. Nobody suspects the adapter, because the adapter is not on the drawing.

Our drawings call out the key type explicitly — the 16-channel build shown here is FC/APC narrow key, PM1550 Corning fiber, PER ≥ 18 dB. If your bench or system uses wide-key sleeves, tell us before the build, not after.

The rest of the array is ordinary; the reference is not

PM fiber array V-groove structure with fibers seated on pitch — Mechanically a PM array is a standard V-groove array. What separates it is that every fiber must be rotationally aligned before it is bonded.

Look at the drawing and the array is unremarkable: 16 channels on 127 µm pitch, Ø125 µm cladding in a 60° groove, 1.5 mm quartz-glass substrate, 10 mm bonded length. The geometry is the same as any single-mode array.

The difference is invisible in the outline: each fiber has to be rotated to its axis before it is fixed in the groove, and it has to stay there through bonding, curing, polishing and connectorisation. A standard array only needs the core in the right place. A PM array needs the core in the right place and the stress rods pointing the right way, sixteen times over, all referenced back to a connector that gets attached at the other end of a metre of fiber.

16-channel PM fiber array build drawing — 16CH PM fiber array — 127 µm pitch, quartz-glass substrate, FC/APC narrow key, PER ≥ 18 dB.

How to specify one so it arrives right

Four lines in your RFQ remove almost all of the ambiguity:

Which axis — slow axis to key is our default; say so if you need fast axis
Which angle — 0° is standard; 45° and 90° are common in polarisation-splitting and combining setups, and we build to any angle you specify
Which key width — narrow (2.0 mm) or wide (2.14 mm), matched to the adapters actually in your system
What PER you need — and at which wavelength, since the same array can be built in PM460-HP through PM1950 depending on band

The one that gets left out most often is the third. It is also the only one on the list you cannot fix later without re-terminating the assembly.

A note on multi-wavelength builds

The arrays in the photo at the top are a 12-channel build combining PM460-HP and PM1950 fiber in one assembly — roughly the opposite ends of the PM range. Each fiber type has its own mode field, its own stress-rod geometry and its own alignment behaviour, but they share one substrate and one pitch, and every channel still has to end up keyed to the same reference.

If your design needs several wavelengths off one array, that is a drawing question rather than a catalogue one. Send us the layout.