MPO polarity: Method A, B or C, and why the specification must name it

“Test report, polarity line: verified.” It is a line that shows up in almost every Tier 1 test on MPO multi-fibre cabling — we have already written about that test — and it almost never says verified against what. On multi-fibre cabling, correct polarity is not a property of the cable or the connector: it is a cabling choice, made once inside the trunk and then inherited by every patch downstream. It has three names — Method A, B, C — and a specification that just says “MPO cabling to standard” has not specified which of the three is needed.

What polarity means on a connector that carries several fibres at once.

On a single-fibre link (LC, SC) the direction is set by the pair of patch cords at each end: get one wrong and it shows immediately, one patch at a time. On MPO multi-fibre connectors — 8, 12 or more fibres in a single connector, used both for parallel-signal transmission from MPO to MPO (the typical case of a 40G, 100G or 400G transceiver) and for duplex breakout to LC patch cords — the direction of each position is set once, inside the trunk cable, at the point where the fibres are aligned to the rectangular ferrule. From there the whole chain inherits it: cassettes, patches, transceivers. A mistake in the trunk is not fixed at a downstream patch panel. It is inherited.

The three methods, with a name.

The naming that the entire supply chain uses — European manufacturers included, on the labels of trunks and cassettes — comes from the American standard ANSI/TIA-568.3-E, Optical Fiber Cabling and Components Standard, current edition published in September 2022, which superseded the earlier 568.3-D from 2016. Let us say it straight away: it is a paid document and we have not read it in full. What follows on the mechanics of the three methods comes from technical documentation published by several independent manufacturers, which cites the standard explicitly and agrees without contradiction — a necessary condition, since without an identical definition everywhere, multi-fibre cabling from two different suppliers would not interoperate at all. No figure, no threshold: only the cabling structure.

On single-row trunks — 8 or 12 fibres, the most common on patching backbones — the logic runs like this:

Method A — “straight-through” trunk: position 1 arrives at position 1 at the far end, with the ferrule key oriented in opposite directions at each end. For duplex breakout this needs an A-to-B patch cord at one end and an A-to-A patch cord at the other (or, to avoid the A-to-A cord, a reverse-polarity cassette and A-to-B patch cords at both ends).
Method B — “reversed” trunk: the entire row flips end to end — position 1 becomes the last position in the row — with the key oriented the same way at both ends. A-to-B patch cords at both ends.
Method C — “pair-flipped” trunk: each transmit-receive pair swaps internally (1↔2, 3↔4…) without reversing the whole row, key oriented opposite as in Method A. A-to-B patch cords at both ends.

On two-row connectors — 16, 24, 32 fibres, the density that the move to 800G has brought into data centres — the same principle holds, but the position mapping gets more complex: key-up-to-key-down mating physically flips the top row onto the bottom row, and “Method A” on a 24-fibre connector no longer means “position 1 to position 1”. Get the exact map in writing from the system supplier — do not infer it by analogy from the single-row case.

The 568.3-E edition adds two methods the D edition did not have, U1 and U2: they apply only to multiple duplex implementations, not to parallel-signal transmission across the whole connector. Anyone still citing “the three MPO methods” as if that were the whole standard is describing the 2016 edition.

Why they do not swap.

The three trunk types are, from the outside, the same connector: same ferrule, same declared insertion loss, same end-face geometry — we have written separately about that measurement. Internally the fibre map differs, and the three are not interchangeable. A Method B trunk installed where the design called for Method A reverses the whole row: on a parallel-signal link the link simply fails to align; on a duplex breakout, every transmit-receive pair ends up crossed end to end. And the fault is not caught by the loss measurement: the source and power meter report how much is lost position by position, not which piece of equipment sits at the far end. A run can pass Tier 1 in full, with loss inside budget and the reference correctly zeroed, and still not work. That is why the “polarity: verified” line on a report needs to say against which method.

Pre-terminated cassettes carry the same problem one level up.

In Base-8, Base-12 and Base-24 systems, the cassette that brings the multi-fibre trunk out to duplex LC patch cords has a type of its own — in practice straight, reverse-polarity, or the two variants that appear in catalogues under manufacturer-specific codes — matched to the adapter that receives the trunk (key-up-to-key-down, or key-up-to-key-up). Cassette and trunk have to be the matching pair: a delivery that arrives with cassettes from one manufacturer and trunks from another, both generically “MPO-compliant”, can have clean end faces, geometry within tolerance, and still have the polarity wrong.

The connector standard does not say it; the polarity standard does — and it is not the same standard.

IEC 61754-7:2008 (edition 3.0, 2008-03-27, withdrawn 2017-12-13) defines “the standard interface dimensions for type MPO family of connectors”: it is the mechanical standard, the one that guarantees two MPO connectors from different manufacturers mate physically. It does not address polarity: the term does not appear in its public description. Cabling “compliant with 61754-7” tells you the connector plugs in, not which way round it carries the signal — the same gap between classifying and prescribing that we have already seen on cable reaction-to-fire classes.

An Italian or European specification more often cites EN 50174-1 (cabling installation and quality assurance, 2018 edition with amendment A1 of 2020) alongside ISO/IEC 11801-1:2017 (general requirements for generic cabling): both include correct polarity among the installation and acceptance requirements, but neither one assigns a name to the methods. A specification that cites only these two has asked for a network that “works the right way round”, without saying by which scheme — and on the scheme, today, the whole supply chain speaks the TIA naming. There is no other one in current use.

What to write in the specification, and what to check before accepting the run.

Name the method — A, B or C, and U1/U2 where the breakout is duplex-only — for every MPO trunk, not “MPO cabling to standard”.
One method for the whole project. Mixing methods between different backbones is the most likely cause of a maintenance job, months later, that reverses polarity without whoever carries it out noticing.
Specify trunk type and cassette type together, on pre-terminated systems: they are a matched pair, not two independent specification lines.
Ask for the type (A/B/C) labelled legibly at both ends of the trunk, and recorded in the as-built — not only on the factory tag that disappears once installed.
Make polarity verification a distinct line on the report, with the expected method written next to the result: “verified” on its own does not say whether it matches the design.
On duplex breakouts, state the patch-cord combination supplied — A-to-B plus A-to-A, or a reverse-polarity cassette with A-to-B everywhere: identical result, different components, different price.

The point.

An MPO run can have loss inside budget and end-face geometry within tolerance and still be cabled the wrong way round: these are independent measurements, and neither of the first two checks the third. Naming the method before ordering trunks and cassettes is part of the work we do when we write or audit a specification for a data centre or for a high-density industrial network; recording it run by run is what goes into the as-built documentation we hand over, so that whoever opens the cabinet in three years does not have to guess.

Do you need to write the specification for a high-density MPO cabling job, or accept one that has already turned up on site? Talk to a technician: the method gets named in one line of a specification, and gets discovered wrong by a link that will not come up.