Tier 1 or Tier 2: which fibre test to specify, and to which standard

“Tier 1 and Tier 2 certification to the applicable standard.” That is the line we find most often in optical network specifications, and it commits nobody to anything. It does not say which instrument measures, how the reference is set, which limit passes or fails a link. Sign an acceptance record against that line and you have accepted the number you were handed, not a number you can check. The two standards that answer the question exist, both date from 2024, and they have to be cited together.

A limit declared up front: these are paid documents and we have not read them. What we attribute to the two standards comes from the publisher’s public pages — title, edition, date, scope, list of significant technical changes. No threshold, no decibel value, no clause number that does not appear there. Mechanisms we explain separately, and qualitatively.

The two standards, with their details.

ISO/IEC 14763-3:2024, edition 3.0, published on 22 May 2024, 89 pages. Title: Information technology — Implementation and operation of customer premises cabling — Part 3: Testing of optical fibre cabling. Declared scope: systems and methods for the inspection and testing of installed optical fibre cabling designed in accordance with premises cabling standards, the ISO/IEC 11801 series included. With one sentence worth more than the rest: “The test methods refer to existing standards-based procedures where they exist.” In other words, 14763-3 says what to test and how to accept it, and for how it is measured it points elsewhere.

IEC 61280-4-2:2024, edition 3.0, published on 6 May 2024, 190 pages. Title: Fibre-optic communication subsystem test procedures — Part 4-2: Installed cabling plant — Single-mode attenuation and optical return loss measurements. Declared scope: measurement of attenuation and optical return loss of an installed optical fibre cabling plant using single-mode fibre.

One standard for testing the cabling, one procedure for measuring on single-mode: naming only one of them leaves half the question open. And 61280-4-2 is single-mode by title and by scope: if the tender also covers multimode links, the matching procedure has to be cited separately.

“Tier 1” and “Tier 2” are not methods.

They are convenient labels. Useful for talking, useless for contracting: a method is a procedure with a standard number, an edition and a year. The two methods have proper names.

Insertion loss measured with a light source and power meter — LSPM in ISO/IEC vocabulary, which lists among the third edition’s changes “addition of end-to-end link LSPM testing”. This is the measurement that certifies: light actually crosses the link, and the number is its total loss.
The reflectometric trace, the OTDR: the measurement that characterises, event by event, inferred from backscattered light. How to read a report and how to choose the pulse we have already written.

The consequence is sharp and almost always ignored: the “total loss” an OTDR prints at the foot of the trace is not an insertion loss measurement. It is an inferred value carrying the instrument’s limits — dead zones at both ends, dependence on the backscatter coefficient of the spliced fibres. Using it as the acceptance result is the commonest shortcut in a rushed handover: the optical budget of the link is verified with the other measurement.

The reference: the line that decides pass or fail.

Here a mechanism needs explaining, and we give it as a mechanism. Before measuring, the instrument is zeroed: source and meter are joined through one or more test cords, and that reading becomes the zero. How many cords sit in the path at the moment of zeroing decides which connections end up inside the measured number and which vanish from it.

With the reference taken over a single cord, the two mated connections at the ends stay inside the measurement: the number is higher and includes both terminations at the patch panel. With the reference taken over three cords — source, intermediate, tail, with the intermediate one then replaced by the link under test — those two connections are already inside the zero and cancel out: the number is lower and describes the link without its ends. A two-cord reference also exists, and includes one of them.

Same link, same instrument, same operator: two different numbers, both legitimate, differing by roughly two mated connections. On a campus backbone that difference disappears into the budget. On a link of a few tens of metres in a data hall, where the fibre contributes almost nothing and nearly all the loss sits in the connectors, that difference is the test result. Which is why the way of zeroing belongs in the specification and has to be repeated in the acceptance record, not left to the instrument’s default setting.

That the cords are part of the method is something the standards say. Among the changes declared for the third edition of 61280-4-2 are “addition of the equipment cord method” and “addition of test limit adjustment related to test cord grades”: the grade of the cord moves the limit. On the 14763-3 side the list includes “introduction of description of reference connectors” and “introduction of new test limits for connector attenuation against reference connector”: the reference connector is described in the text, and a specification can demand it by name.

What else changed in 2024.

From the list of significant technical changes to 14763-3 edition 3.0:

“addition of testing of MPO cabling”: testing of multifibre cabling enters the text. There, loss is not enough anyway, and end-face geometry of the ferrule remains a line item of its own.
“introduction of normative inspection for cleanliness”, aligned to the ISO/IEC 11801 series: inspecting the end face is part of the test method, no longer good practice.
“addition of measurement uncertainty for all measurement methods”: a value with no declared uncertainty is not a measurement, it is a number.
“testing will support SM ranges up to 10 km”: campus and site-to-site links now fall inside the standard’s perimeter.
And one in the negative, “removal of plastic fibre testing”: a specification citing 14763-3 for plastic optical fibre now points at a text that no longer covers it.

What to write in the specification.

Both standards in full, with edition and year: ISO/IEC 14763-3:2024 (ed. 3.0) for inspection and testing, IEC 61280-4-2:2024 (ed. 3.0) for attenuation and return loss on single-mode. Never “to the applicable standard” on its own.
The two methods, each with its own result: insertion loss with source and power meter as the acceptance measurement, the reflectometric trace as mandatory characterisation alongside it. They do not replace one another.
How the reference is set: how many cords sit in the path at zeroing, and therefore which end connections are inside the declared value. The same sentence goes into the acceptance record.
Grade and condition of the test cords, with end-face inspection and cleaning before every zeroing, and the reference connector demanded by name.
Bidirectionality: insertion loss in both directions on the links that matter; on the trace, the average of both directions for every splice, as ITU-T G.650.3 sets out.
Optical return loss measured and reported: it is in the title of 61280-4-2, and it is the line missing from almost every acceptance record we are asked to re-read.
Measurement uncertainty declared for each method, with the instrument, its serial number, calibration date, operator and date.
Native data handed over — traces and readings, not the PDF alone: without them, nobody re-checks anything five years from now.

The point.

“Tier 1 or Tier 2” has a short answer: both, called by their proper names. The real question is a different one — against which reference — and it separates a record that holds from one that merely exists. It is the work we do when we write or audit a specification and when we certify the links: the method declared before the measurement, the measurement before the signature. The evidence archive, native traces included, stays wherever you decide: on-premise in your own environment, or on dedicated cloud — an environment reserved to the single client, dedicated VPN, data centres in Italy, premises staffed directly. So that the choice is not a contractual afterthought, we have set it out in full.

Do you have to draft the testing clause of a tender, or sign an acceptance record you are not convinced by? Talk to an engineer: deciding up front which measurement counts as acceptance, and against which reference, costs one meeting; finding out afterwards costs a dispute.