Testing multifibre links: loss is not enough, end-face geometry decides

“Insertion loss within specification.” On a sixteen-fibre cord that line stands for sixteen measurements, and why one is worse than the rest is almost never legible in decibels: it is on the end face. On 27 July 2026 the IEC published edition 3.0 of IEC 61300-3-30, the method for measuring the end-face geometry of rectangular ferrules — the ferrules of multifibre connectors, the MT/MPO families. It replaces edition 2.0 of 14 December 2020, withdrawn on that same 27 July. And it brings, declared among the changes, the tables for 16-, 24- and 32-fibre ferrules.

Let us say it straight away: the standard is a paid document and we have not read it. Everything that follows about its content comes from two public sources: the abstract and the change list on the IEC product page. We attribute to it no figure, no threshold and no requirement that does not appear there: that would be invention. Where a mechanism explains why an attribute matters, we give it qualitatively and keep it separate from the standard.

Why a rectangular ferrule is a different problem.

In single-fibre connections there is a cylindrical ferrule and one fibre: the face can be inspected and cleaned, and we have already written about that. In a rectangular ferrule the fibres are many, lined up in a row inside a single block, and the mating is not done by a cylindrical sleeve but by two guide holes and the pins through them. Physical contact has to happen on every fibre at once, with a single spring force on a single surface. Not a cleanliness problem: a problem of shape.

The IEC interface specification for flat-face rectangular ferrules says as much: it defines dimensional limits “in order to meet specific longitudinal offset requirements for fibre-to-fibre interconnection”. In other words: geometry is there to keep air out from between the two pieces of glass.

The four attributes, one by one.

The abstract states the primary attributes of the measurement: fibre position relative to the end face, end-face angle relative to the guide holes, fibre tip radii, and core dip for multimode fibres. Translated into what happens in the field:

Fibre position relative to the face. The previous edition was more explicit: withdrawal or protrusion. A fibre sitting below the plane leaves an air gap exactly where contact is needed; one standing proud of the others takes the spring force and can stop its neighbours touching. Either way loss and reflectance get worse, differently at every mating.
End-face angle relative to the guide holes. Alignment is imposed by the pins. If the polished plane is tilted relative to the plane defined by the holes, the two ferrules meet askew: contact reaches one end of the row and not the other, and the row splits into good fibres and marginal ones with nobody having done anything wrong.
Fibre tip radius. The tip is not flat, it is a dome: contact happens on the glass, not on the material around it. A radius outside tolerance moves the point where the pair actually touches — too flat, the load spreads where it is not wanted; too pronounced, it concentrates on a minimal area.
Core dip, on multimode fibres. The core is doped and polishes at a different rate from the cladding: it can end up slightly below the surface around it. The two faces touch, visual inspection is clean, and a void is still left above the core. It is the least intuitive of the four, and why it is measured separately.

The change that reaches specifications: 16, 24 and 32 fibres.

Among the changes declared for edition 3.0, two are worth more than the rest: the introduction of the x116 and x132 regions of interest to support MT-16 and MT-32 ferrules, and the preparation of geometry limit (GL) parameter tables for 16-, 24- and 32-fibre ferrules. The region of interest is the area of the face within which the calculation is made: with no region defined for that ferrule type, the number has no boundary.

If those tables appear among the changes, they were not in the previous edition. And 16- and 32-fibre ferrules are no laboratory curiosity: they are the density that the move to 800G has brought into data centres, we have written about it. Anyone installing them today on a specification that refers generically to “61300-3-30”, or accepting a test report issued against the 2020 edition, is relying on tables that did not cover those ferrules. The operational lesson is one line long: write the edition and the year in the specification. Since 27 July 2026, moreover, “61300-3-30” on its own may point at a withdrawn text.

Neighbouring fibres: why a definition matters.

Also among the changes is the clarification of the neighbouring fibres definition when computing adjacent height. It looks like wording. It is not. In a row, the absolute height of each fibre says little: what decides whether contact is uniform is the difference from its neighbours. A fibre standing higher takes the load and leaves its neighbours behind; a lower one never touches, even while within tolerance in absolute terms. Coplanarity across the row has been the heart of this measurement for years: the product page for the 2020 edition lists the GL metric and the minus coplanarity metric among its changes.

But a clarified definition, inside a measurement standard, says one thing only: before it, two laboratories could compute different numbers from the same part. And a criterion that depends on who applies it is not a contractual criterion.

What to ask for, in supply and at acceptance.

An end-face geometry report to IEC 61300-3-30, with the edition and year written out in full — today edition 3.0, 2026.
The fibre count and ferrule type stated in the report. Limit parameters are not the same for a row of 12 and a row of 32: a report that does not say what was measured cannot be read.
The four attributes measured and reported fibre by fibre, not one pass/fail for the whole part.
A part identifier — ferrule or cord serial — in the report and tied to the as-built: without it, the document belongs to nothing.
An explicit distinction between the manufacturer’s certificate and field verification. The first describes the part as it left the factory; the second accounts for transport, installation, termination and matings already made. They do not replace each other.
Geometry as a separate line item from insertion loss and from the OTDR trace, with a price of its own: if it is “included”, nobody does it.

The point.

A multifibre cabling job is handed over with numbers, not with a declaration. Measuring end-face geometry, stating which edition it was measured against and leaving a document still readable in three years is the method with which we write and audit specifications: measurement first, then certification, then the evidence archive, wherever suits you.

A high-density cabling job to specify, or an acceptance test to sign off? Talk to us: deciding up front which attributes get measured, to which edition and with what evidence costs far less than a row of fibres that only touches halfway.