Request an intervention

Operational notes Engineering

One fibre, three jobs: traffic, certified time and alarms. What changes in the specification

7 min read

Metal tuning fork with adjustable weights and a wooden mallet lying on a dark speckled worktop, in black and white
A reference is worth nothing unless it arrives intact where the tuning happens. That is the whole problem of time in a network.

On 30 July 2026 Open Fiber and INRiM, Italy’s national metrology research institute, announced a 14-month collaboration to trial three different things on the same optical fibre infrastructure: certified time distribution, quantum communications and distributed fibre sensing.

For anyone who designs networks or signs off specifications, this is not laboratory news. It is the moment when “a fibre pair” stops being a line item and becomes a contested resource, shared between uses that do not coexist on their own.

What was announced, and what was not

The Open Fiber release is dated 30 July and its headline is explicit: “Open Fiber and INRiM experiment with the networks of the future: certified time, quantum security and intelligent fibre”. INRiM published its own item the same day.

The three declared strands are these.

  1. Certified time. Synchronisation using the Precision Time Protocol (PTP) and White Rabbit, described as essential for coordinating systems and services that need extremely accurate synchronisation; the sectors named are financial transactions, energy networks and data centres.
  2. Quantum communications. Development and validation of infrastructure and protocols for quantum key distribution (QKD).
  3. Distributed fibre sensing. Real-time monitoring of physical and environmental conditions along the infrastructure.

INRiM brings expertise built over “more than ten years of research” and a distributed optical fibre infrastructure of more than 1,800 km across the peninsula, from Turin to Matera. Nicola Grassi, Technology Director at Open Fiber, is quoted as saying that fibre is no longer only a connectivity infrastructure but can become a technological platform able to distribute certified time. Davide Calonico, Scientific Director of INRiM, states that the objective is to transfer the results of public research to industry in order to develop Italian innovation. (Both releases are in Italian; the renderings here are ours.)

What the two releases do not say matters before anyone writes a requirement on top of them: which routes are involved, how much capacity is dedicated, what numerical accuracy targets are expected, and whether or when any of this becomes a purchasable service. It is a fourteen-month research collaboration. Reading it as a catalogue is the mistake to avoid.

Time on a network is not an application service: it is a path constraint

This part applies to the specifications you are writing today, whatever the trial produces.

PTP measures the delay between two clocks by exchanging messages in both directions, and it assumes the path is symmetric. If the forward and return directions do not take the same route, the resulting time error is half the difference. With the group index typical of single-mode fibre — around 1.47 — one kilometre of fibre is worth roughly 4.9 microseconds of delay: one kilometre of difference between the two directions produces roughly 2.4 microseconds of error. Stable, invisible to alarms, and not correctable by any clock setting.

Three constraints follow, and all three belong in writing. Where possible, bidirectional transmission on the same fibre; where that is not possible, asymmetry measured and compensated with a declared value, not estimated off a drawing. Automatic protection has to be handled separately: a switch to the alternative route changes the asymmetry instantly, traffic does not notice and time does. And every closure reopened and resealed leaves lengths different from the ones that were tested: if the asymmetry is not re-verified, the specification still carries a number that is no longer true.

Reference documents are cited in full, with the edition, not by acronym:

  • ITU-T G.8275.1 (ed. 02/2026), Precision time protocol telecom profile for phase/time synchronization with full timing support from the network: the profile that applies when every node along the path participates in timing.
  • ITU-T G.8275.2 (ed. 02/2026), the same subject with partial timing support, that is, with nodes that do not participate. It is not the “easy” version of the first: it is a different level of performance and must be declared as such.
  • ITU-T G.8271.1 (ed. 11/2022, Amendment 3 of 05/2025), Network limits for time synchronization in packet networks with full timing support from the network: the yardstick for accepting or rejecting an acceptance test.
  • ITU-T G.8272 (ed. 07/2025), Timing characteristics of primary reference time clocks: the source.
  • In the electricity sector the profile is a different one: IEC/IEEE 61850-9-3:2016, Precision time protocol profile for power utility automation, a profile of IEC 61588:2009 / IEEE 1588-2008 designed to meet the most demanding synchronisation classes of IEC 61850-5 and IEC 61869-9. Substation work does not use the telecom profile.

White Rabbit, which the release names, originated at CERN, claims sub-nanosecond accuracy and picosecond precision, and was incorporated into the IEEE 1588-2019 revision of the standard. It is not a box to tick: it needs compatible equipment along the whole chain, not only at the two ends.

Why an optical carrier matters at all: today the reference almost always comes from a satellite receiver, which means a rooftop antenna and a signal arriving from outside that the network operator does not control. Fibre is a second path with entirely different properties. It does not replace the first one: it makes it comparable.

Fibre that senses: what it actually implies

Distributed fibre sensing injects pulses and reads the backscattered light — the same physics as an OTDR, used to measure not the fibre but what happens around it: vibration, temperature and strain along the whole route.

Three consequences that change a specification.

  1. It needs a fibre, not a channel. The interrogator occupies the fibre. “Spare fibre” stops meaning “fibre free in case of a fault” and starts meaning “fibre committed to a service”: two different things, to be counted separately when the cable is sized.
  2. Position is read in metres of fibre. The alarm says “event at 6,312 metres”, not “in such-and-such a street”. The conversion is done by the as-built record, with the slack coils in the chambers and the real lengths section by section. Without that document kept current, the system produces numbers nobody can turn into an address.
  3. Sensitivity depends on how the cable was installed. A cable coupled to the ground, a cable loose in an empty duct and an aerial cable respond differently to the same vibration. No catalogue figure transfers from one route to another: if the requirement is to detect an excavation within a given distance, it has to be verified in the field on that route, and the test record has to say so.

On the quantum strand the honest answer is short: today there is nothing to buy. Keep it in view as a future constraint on the medium — dedicated channels, coexistence with classical channels on the same fibre — not as a tender item.

What to write now

  1. A reserved, non-switching pair for timing and sensing, identified in the test record by its own code, not as “an available pair”.
  2. A path declared with infrastructure identifiers — duct, cable, closure: “a different route” without identifiers cannot be verified by anyone.
  3. Asymmetry measured at handover, recorded in microseconds, and re-verified after every splice intervention — with an intervention log available to whoever runs synchronisation.
  4. Standards cited in full and with the edition, and the choice between G.8275.1 and G.8275.2 justified by the network you have, not copied from another specification.
  5. In-service measurement designed in from the start: if the route also carries time and sensing, measuring it without taking the customer down is not an optional extra.

The point.

The 30 July announcement says something uncomfortable: installed fibre is worth more than the contracts governing it say. Traffic, time and sensing all sit on the same glass, and none of the three tolerates the vagueness with which a route is usually described. That is how we work: design, measurement and documentation defined together, with the specification written as verifiable requirements — pairs, routes, closures, values — rather than adjectives, from backbones to data halls. And when those data need a system to read them, sensing alarms or synchronisation histories, the usual rule applies: on-premise in your own environment, or on a dedicated cloud — an environment reserved to the single client, dedicated VPN, data centre in Italy, premises staffed directly. Where the hardware and the models sit is a specification decision, not something you find out later.

Do you have a synchronisation requirement to write, or a sensing service to assess on your own network? Talk to an engineer.

Sources