Summary

  • AFL, Corning, Sumitomo Electric and TeraHop have published a common Version 1.0 specification for a four-core, 2x2 square-lattice fibre intended for short-reach O-band links.
  • The document makes fibre geometry, per-core optical performance and crosstalk measurable. It expressly leaves transceivers out of scope and leaves several assembly and application choices to later specifications and local implementation.

A useful standard does not have to describe an entire system. Often its value lies in removing one ambiguity and refusing to absorb the rest. SDM4 MCF MSA 1.0 does that for the glass at the centre of a possible new data-centre cabling ecosystem.

The four announcing companies call the document a foundation for multi-vendor interoperability. That is fair at the layer it actually covers. A supplier can now make a four-core fibre against shared coordinates, dimensions and optical limits. A buyer can ask whether a sample meets those requirements. A test laboratory can reproduce the relevant measurements under declared conditions. None of those statements yet says that a complete optical path assembled from different vendors will carry a particular Ethernet service, meet a link budget or survive installation.

A square inside the cladding

The common physical picture is unusually specific. Four core centres occupy a square: each sits 20 micrometres from the cladding centre on both axes. Adjacent cores are spaced at 40 ± 1 µm, the radial position is 28 ± 1 µm, and core-position error may not exceed 1 µm. The cladding is 125.0 ± 1.0 µm, preserving the familiar outside dimension of conventional single-mode fibre, while the coloured coating may span 180 to 220 µm.

The optical contract is also concrete. At 1310 nm, cabled-fibre attenuation is limited to 0.4 dB/km and point discontinuity to 0.1 dB. Cable cut-off wavelength must not exceed 1260 nm. Mode-field diameter, dispersion and per-core polarization-mode dispersion all receive bounded values. Both adjacent and diagonal core pairs must keep crosstalk at or below -40 dB when normalised to 1 km at 1310 nm.

These figures turn a vendor description into a testable object. The shared unit is not “four times the network”. It is one passive fibre design whose four guides can be measured separately and whose mutual interference can be bounded.

The measurement already exposes the next boundary

Crosstalk is not observed without a measurement system. The specification permits fan-in/fan-out devices, an offset-spliced single-mode fibre or another suitable alignment technique to excite and receive from individual cores. When a FIFO is used, its own crosstalk sits inside the measured result unless its contribution is considered or de-embedded. Core-dependent loss from fibre, splices, connectors, launch and receive fibres, and test equipment can also make one coupling direction differ from the other.

That detail is more important than it looks. Even a conformance number has a provenance chain: specimen, length, wavelength, bend state, launch condition, receive condition, FIFO, connector, splice, averaging method and correction. The MSA describes acceptable ways to make the fibre measurement meaningful. It does not convert the surrounding apparatus into standardised production equipment.

The same boundary appears in the appendix. Symmetric and asymmetric marker arrangements are both acceptable. Core numbering can depend on whether transmission is unidirectional or bidirectional, on connector and transceiver configuration, and on other system choices. Bare fibres can be joined by aligning rotational references rather than requiring the same printed core number to meet itself. Connectorised assemblies, however, still need a defined angular relationship between the connector key and the fibre orientation reference.

In other words, the fibre can be geometrically compatible while an assembly can still map the wrong core to the wrong port.

What Version 1.0 deliberately does not decide

The most useful sentence in the document is the explicit exclusion: transceiver specifications are out of scope. The fibre may support intensity-modulation/direct-detection or coherent designs and may be used with DR, FR or LR-style O-band interfaces, but those examples do not select a line rate, wavelength plan, modulation, FEC, transmitter launch condition or receiver tolerance.

Version 1.0 also does not supply a complete connector standard, fan-in/fan-out product contract, cable-assembly polarity plan, core-to-lane map, installed-link budget or acceptance procedure. It does not certify a particular splice cassette, patch panel or repair method. It does not say how an operator should record an assembly whose longitudinal orientation changes between sections. Those questions are not defects in the fibre specification. They are the next contracts.

This is why “multi-vendor interoperability” needs a noun after it. The MSA can support interoperability of compliant fibre inputs. End-to-end interoperability requires compliant and mutually compatible assemblies, interfaces and operational procedures as well.

Marketing numbers belong to a different ledger

Corning separately presents multicore fibre as delivering up to four times the optical-path density, with claims of fewer cables and connectors, lower cable mass, faster installation and lower passive-component emissions. Those statements may describe real advantages for particular designs. They are not measurements published in the MSA specification.

A 125-micrometre strand with four cores can increase path density inside the glass. Whether a site removes 75% of physical connections depends on the old and new topology, duplexing, spare policy, breakout points and connector architecture. Whether labour falls depends on assembly preparation, inspection, cleaning, splicing, testing and rework. Whether usable capacity rises fourfold depends on transceivers, lane mapping, crosstalk margin, link loss, FEC and application demand.

The correct conclusion is not that the marketing outcome is false. It is that fibre conformance cannot serve as its receipt.

A minimum common layer, not a finished network

The four suppliers have done something operationally valuable: they have agreed on a small common object before each product ecosystem hardens around incompatible geometry. The specification can reduce supplier lock-in at the fibre layer because a buyer can name a stable physical and optical baseline rather than a proprietary cross-section.

That restraint should be preserved. International standards work may later absorb the fibre requirements; interface bodies may standardise transceivers and assemblies; operators may publish qualification results. Each new layer should earn its authority through implementation and tests at that layer. Publication makes a choice legible. Adoption, installation and observed performance make it real.

SDM4 1.0 therefore deserves neither dismissal nor promotion into a result it does not claim. It has fixed the shape and minimum behaviour of the passive fibre. The link begins where that document stops.

Sources