Certifying a fibre network: the complete guide.

The optical budget is the maximum loss a link can absorb while staying within the equipment specification. It is calculated item by item before the cable is ordered — fibre attenuation, splices, connector pairs and, on PON networks, splitters — and compared with the optical class of the planned equipment. It is a binary threshold: if the sum of the losses exceeds it, the link does not respond.

The item that is almost always missing is the design margin: the decibel that will be needed in five years, after one repair and a few summers. A design handed over with zero residual margin is compliant on paper and fragile in real life.

The fibre is chosen on the bend radius the route actually imposes, not on a generic designation. G.652.D is the backbone fibre, historically characterised for wide radii, around 30 mm. Where the bend is unavoidable — dense patching, building entries, racks — the G.657 family is needed, and so is the subcategory, not the designation alone.

The difference is not theoretical: on the very same bend the macrobend loss changes by more than twenty times between one subcategory and another, and on an OTDR trace that loss looks like a poorly made splice.

A cable pulled beyond its limit almost never breaks on the spot: it answers later, with a few tenths of a dB that no test campaign can attribute to that day on site. Maximum pulling tension is a design figure and is calculated from the cable weight; the bend radius during installation is measured in cable diameters, not by eye. The duct is proved first: free of debris, not crushed, with the fill ratio checked together with the cables already in place.

On a public road the technique changes the site, not the permits. Low-impact trenching still requires the road-opening concession issued by the authority that owns the road, and the point where the job is won or lost is the road surface reinstatement, which has to be written in two phases.

Fusion and mechanical splicing are not two variants of the same technique. Fusion welds the two glasses and the link becomes a single fibre again; a mechanical splice aligns the two faces in a precision groove and entrusts optical continuity to a gel, which over the years gets contaminated, dries out and suffers thermal cycling — in a manhole or an outdoor cabinet more than anywhere else.

In a specification, “workmanlike execution” is not a requirement: it is a blank delegation. What is needed is the declared technique, the maximum loss per splice, the reflectance limit and the obligation to measure bidirectionally, otherwise someone else makes the choice for you.

Pulse width is the setting that weighs most on an OTDR measurement, and it is almost always the only one the specification never mentions. A long pulse carries energy to the far end of long links but merges close events into one; a short pulse separates events but leaves the trace in the noise. That is why a single acquisition is not a test campaign.

Launch and tail cords are not an accessory: without them the first and last connectors of the link stay inside the dead zone and never appear in the report. The cord is sized on the pulse actually used, not on habit.

In a test report what is missing matters more than the values. Without the pulse width nobody knows what the instrument could see; without the cord lengths the first connector stayed invisible; without the second direction a splice can appear as an impossible gain. Tier 1 and tier 2 answer two different questions, and on the links that matter you need both.

Three questions are enough for whoever hands the document over: which direction did you measure in, with which thresholds, may I have the native files. If the answer to the last one is no, the first two are already answered.

The design says what was meant to be installed, the as-built says what is there: a diversion around an unexpected utility, a splice moved twenty metres, a cabinet rewired at the last minute. Without a point-by-point match between diagram and test traces a report stays an isolated PDF: it says the link worked on the day of measurement, and says nothing at all five years later.

Labelling is the part that gets paid for at the first emergency. It needs a unique code tying splices, drums and patch ports to their position on the diagram, because whoever intervenes in three years will not be whoever installed it.

Since 1 July 2017 “flame-retardant cable, compliant” identifies nothing: Regulation (EU) 305/2011 and the harmonised standard EN 50575 require CE marking and a Declaration of Performance, with seven reaction-to-fire classes and the subclasses for smoke, droplets and acidity. A class is written in full, and “halogen-free” is not a class.

The class is differentiated by intended use: risers, escape routes and crowded rooms do not carry the same requirements as secondary runs. At testing, compliance is demonstrated on the installed cable, checking by sampling that the designation printed on the sheath matches the specification and the Declaration of Performance.