The signal is too strong: when a fibre run needs an optical attenuator

A new ONT, a run of a few hundred metres from the exchange, very high received power — and the link will not come up, or drops packets intermittently. The technician measures, finds a very strong signal, and looks for the fault elsewhere: a connector, the electronics. The signal itself is the problem: every receiver has a minimum threshold below which it reads nothing, but also a maximum one above which it goes into overload. On that run, nobody can say whether a fixed optical attenuator was ever specified, removed, or never there at all.

A component almost no specification names

The reference is ITU-T L.401, Optical fibre attenuators — since 15 February 2016 the number for the former ITU-T L.31, approved by the WTSC in Geneva on 18 October 1996. The ITU listing notes: «Former ITU-T L.31 renumbered as ITU-T L.401 on 2016-02-15 without further modification and without being republished» — same text, new number, status in force. Its scope covers single-mode fibre only, «because this fibre is mostly used in present telecommunication systems» — because it is the one most used today.

Clause 3 lists three typical applications, and the first is exactly the field case above: the attenuator serves «to assure the linear behaviour of optical fibre receivers avoiding optical power overloading». The second is different and just as concrete: «to balance the optical power into Passive Optical Network (PON) branches» — balancing power across a PON’s branches, where the split ratio never hands every user the same loss. The third is a bench use, not a network one: «to make measurements on an optical telecommunication system».

Fixed or variable, patch-cord or adapter

The Recommendation separates fixed attenuators, with a single value, from variable ones, with a — almost always mechanical — control that slides the value across a range. By configuration: an attenuator with permanently attached pigtails is an attenuating patch-cord, which can also be spliced straight into the line; one without pigtails is an attenuating adapter, female-female or male-female (an optical pad). «The most common types of attenuator that are permanently installed in optical fibre plant are fixed», and among these the male-female is the most widespread because «it is the only configuration that permits the connection or disconnection into the optical link, avoiding any other modification of the plant» — the only one that can be inserted or removed without touching anything else in the plant. It is also why, years after installation, an attenuator a technician added to close a complaint can stay there with the original design never knowing about it.

The numbers: where the standard states them, and where it does not

For the optical parameters, L.401 points to a table in ITU-T G.671; in the edition currently in force — 9.0, approved on 29 November 2025 — the optical attenuator is clause 6.4, no longer the 6.3 that the 1996 text pointed to: a specification that cites table 6.3/G.671 in 2026 is citing a clause that, in today’s G.671, is something else.

Table 6.4 sets two absolute thresholds: maximum reflectance −40 dB and maximum polarisation-dependent loss (PDL) 0.3 dB, with operating windows of 1260–1360 nm and 1480–1580 nm. It does not set a universal value for insertion loss or attenuation accuracy: the convention in G.671’s clause 5 explains that «“sba” (specified by application) is sometimes used as a parameter value in clause 6 to indicate that this parameter for this component must be determined from the application in the relevant transmission system Recommendation» — it has to come from the applicable system recommendation, not from G.671 itself. A specification that only says attenuator compliant with G.671 has not stated the nominal value or its tolerance: someone still has to write those in.

On input power the two sources, thirty years apart, say almost the same thing. The 1996 L.31 requires a guaranteed linear response «at least for input power up to + 15 dBm and up to + 20 dBm will be allowed in some cases». The 2025 G.671 leaves the parameter «for further study» and adds, at Note 2: «The maximum value of allowable input power is under discussion. A value of +20 dBm is considered a starting point» — still a starting point, not a fixed limit. L.31 also sets an ageing margin: during mechanical and environmental tests «a maximum variation of 10% on the actual value of attenuation […] should be tolerated» — the same principle as the design margin in an optical budget, here applied to a single component. And it must withstand «a cycle of 1000 repeated matings», with attenuation measured every 25 matings: a figure that matters when the same attenuating adapter is unplugged and replugged repeatedly over successive maintenance visits.

Even the test method changed number without changing subject: L.31 points to «IEC 869-1 (2nd ed.) Generic specification for optical fibre attenuators»; today’s G.671 cites in its place IEC 60869-1:2018, «Fibre optic passive power control devices – Part 1: Generic specification» — the same document, renumbered, but under a broader title: no longer for attenuators, but for passive power-control devices in general.

Where it is written that a run has an attenuator

The specification may call for one, or not. The as-built record may mark its position and value, or show the run as a single unbroken length of fibre and connectors. The acceptance report may state the measured accuracy — the gap between declared and real value — as a line of its own, or fold it into the run’s overall insertion loss. The label on the component, where one exists, states the nominal value in dB; where it does not, an attenuating adapter in a patch tray is indistinguishable, to the naked eye, from an ordinary coupler. The field crew’s worksheet may record that an attenuator was added to close a complaint — or may not record it at all, remaining the only unwritten proof that the run is no longer what the design says. The maintenance contract, finally, rarely calls for a check when the terminal equipment changes: an ONT more sensitive, or less tolerant of overload, can turn a value that was right for years into the wrong one overnight.

Six different places, none answering on its own: is there an attenuator, what value, where, and why.

What we don’t know

Compliance with ITU-T Recommendations is voluntary: G.671 itself says so, «Compliance with this Recommendation is voluntary. However, the Recommendation may contain certain mandatory provisions […] and compliance with the Recommendation is achieved when all of these mandatory provisions are met» — it binds only as far as a specification names it, with its own numbers. The IEC standards cited in the text — 61300-3-2, 61300-3-4, 61300-3-6, 61300-3-7 and 60869-1 — are paid documents: we have not read them in full, and what we report comes only from what G.671 and L.401 quote in their own free text. We have not verified any specific Italian obligation on optical attenuators, and make no claim either way.

Applied to this run

First axis: the specification line does not say “attenuator where necessary.” It states the expected nominal value, the accuracy tolerance, the type — fixed or variable, patch-cord or optical pad — and its position on the run; and at acceptance the difference between nominal and actual is measured, not simply declared.

Second axis: presence, position and value of the attenuator become, with CSIDIA, the other company in the group, part of the single network map, alongside the splitter and every other passive component on the run. The diagnosis that matters is the comparison between today’s measurement and yesterday’s, tied to the same identifier: an AI flags when a value no longer matches what was expected — attenuator gone, added, or no longer right for the equipment now at the other end — and the crew steps in. Within the client’s own perimeter: on-premise on autonomous machines, or a dedicated cloud with a data centre in Italy, always with shared management.

Could you say, run by run, whether an optical attenuator is specified, where it sits and what value was measured at acceptance? Talk to an engineer: the site visit is at no cost, and out of it comes the list of your runs with, for each, the answer to this question — blank boxes included. It stays yours even if we do not go on together.