Request an intervention

Operational notes Testing

Fibre already installed: what can still be verified, and what is lost for good?

6 min read

Golden hourglass with sand half fallen, on a wooden table against a dark blurred background, in black and white
The glass in the fibre passes its one and only strength test before it ever reaches site. What happens afterwards cannot be repeated.

The acceptance test the next day passes: attenuation within budget, a clean OTDR trace at two wavelengths, no event out of mask. The record gets signed, the job closes. Nobody will ever know whether, halfway along the run, the cable met sudden friction and for a couple of seconds the tension climbed well above the rated value, then settled back before it reached the far chamber. That spike leaves no mark in the attenuation measured the next day: the glass, however, remembers it — in a way that no acceptance-time OTDR ever sees.

The glass passes one exam, and it passes it before it ever reaches site.

Every metre of fibre that arrives on site has already survived, for a fraction of a second, a tensile test: the proof test. Recommendation ITU-T G.650.1 (01/2024), clause 3.2.1, defines it this way: “the proof test level is the specified value of tensile stress or strain to which a full length of fibre is subjected for a short time period.” Clause 6.7.1.1 states when: “proof testing is performed during fibre manufacturing, online as part of the fibre drawing and coating process, or offline as part of the testing process” — during manufacturing, not afterwards. The sections that fail break there, in the factory; the ones that reach the reel are, by definition, the survivors.

The threshold is not generic. ITU-T G.652 (08/2024), clause 6.7.3 and Tables 1 and 2, sets the minimum proof stress at 0.69 GPa, identical for categories G.652.B and G.652.D — the standard single-mode fibre most backbones are built from. It is the one figure, across the entire life of that fibre, that certifies its real mechanical strength with a destructive test on the weakest population. After cabling, after transport, after installation, that test is never repeated: doing it again on a buried run would mean pulling it until it breaks.

Bend radius is not only an optical limit.

We have already written about how much macrobending costs in decibels. The same G.652, clause 6.6, Note 2, explains why that minimum radius exists regardless of attenuation: “the recommended radius is equivalent to the minimum bend-radius widely accepted for long-term deployment of fibres in practical systems installations to avoid static-fatigue failure.” That is a different mechanism from optical loss: G.650.1, clause 3.2.2, defines the “stress corrosion (susceptibility) parameter n” as the coefficient linking the growth of a surface crack to the stress applied over time, with both a static and a dynamic value, each dependent on temperature and humidity. A fibre bent tighter than its design radius and left that way inside a closure can pass today’s acceptance test — the bend generates no measurable loss — and fail years later from fatigue, exactly where residual moisture inside a closure that does not hold its seal accelerates this very mechanism.

Rated tension is qualified once; actual tension is almost never logged.

We have written about this here: ITU-T L.100 sets the rated tensile load LS and the long-term residual load LL (30% of LS), and Annex A requires unchanged attenuation at LL, not at LS — because LS is a brief event, verified once in a laboratory for each cable type, not run by run on every site. It is a product qualification, not an installation measurement. The dynamometer on site, where one is used at all, often shows only the peak value at the end of the pull — not whether, for two seconds, in a tight bend halfway along, the load exceeded that value and then dropped. Without continuous logging, that data point simply does not exist: it was not lost through negligence, it was never recorded.

What remains verifiable, and what does not.

Once installation is complete, what remains verifiable is: attenuation and the OTDR trace at the operating wavelengths, length, geometric continuity, and — weeks later — a fresh comparison across wavelengths to catch a microbend that settled in after the first acceptance test. What does not remain verifiable, unless it was logged during the operation itself, is the actual tension profile along the whole route, every transient peak exceeded and then relieved, and therefore the static fatigue accumulated at each point of the run. A clean optical acceptance test the day after installation does not prove the absence of these events: it only proves that, on that day, they were not yet severe enough to show.

What to write into the specification.

  1. Continuous logging of pulling tension, not a glance at the dynamometer: a time-stamped trace, delivered as a file together with the as-built record, across the whole length of the run.
  2. A proof stress certificate for the fibre batch supplied, with the manufacturer’s declared value — minimum 0.69 GPa for G.652.B/D under ITU-T G.652 — attached to the delivery, not inferred from the words “compliant with G.652.”
  3. An LS/LL qualification report for the cable type offered, per ITU-T L.100 Annex A, requested from the cable manufacturer, not taken on trust.
  4. Minimum bend radius respected point by point, verified in the as-built record at every anchor, bend and closure — not just declared at design stage — because that radius is also a mechanical-fatigue limit, not only a loss limit.
  5. A second optical measurement after a stated settling period, compared against the first acceptance measurement, on the critical runs.

How it is verified at acceptance.

The handover record must attach four separate documents, not one statement of compliance: the logged tension trace, the proof stress certificate for the fibre batch, the LS/LL qualification report for the cable, and the initial bidirectional optical measurement as a baseline for future comparison. None of these, alone or together, removes the uncertainty around static fatigue: they reduce it to a level that can be managed — so that when a run fails at an unexpected point ten years from now, these documents show whether that point had already shown a sign, or whether the failure really was unforeseeable.

The point.

The acceptance test the day after installation certifies that the fibre works that day, not that it went through installation without future consequences. The difference lies entirely in what gets recorded while the cable is moving, not in what gets measured once it stops. That is why, when we design an installation and write a compliance specification, pulling tension is logged continuously and the proof stress certificate is requested in writing, alongside the optical acceptance test and the as-built record. Tension traces, factory certificates and measurements repeated over time all enter the same single network map — not files scattered across jobs — where an AI flags which runs saw a peak close to the limit or a bend that was never relieved, before it becomes a fault, for an operator, a data centre, an industrial plant, a public body, a healthcare facility or a defence site. Where AI is needed on network data, it runs within the client’s own perimeter — on-premise or in a dedicated cloud with data centres in Italy — together with CSIDIA, the group’s other company.

About to install a long run, or accept one installed by someone else? Talk to an engineer: the site survey is free of charge, and tension logging gets agreed before the first metre of cable goes in, not after the acceptance test.

Sources