Request an intervention

Operational notes Engineering

Different mode field, different splice: when the OTDR lies in one direction only

7 min read

Dozens of optical fibre strands hanging in the dark, each with its own lit tip, black and white photograph
Every strand carries its own light down to the tip: even two fibres of the same designation never have exactly the same mode field.

A technician closes an above-ground closure: on one side, the G.652.D backbone another operator laid eight years ago; on the other, the G.657.A2 access run his own crew just pulled. The splicer reads 0.04 dB — a near-perfect joint. Next day’s test measures the same fusion splice from the opposite end: the value becomes −0.08 dB, a gain that physics doesn’t allow. Nobody touched the splice between the two measurements. The question a specification rarely asks: what is the true loss of a splice between two fibres that, though both compliant with the recommendation cited, never have exactly the same mode field?

Mode field diameter: same designation, different tolerances

Mode field diameter (MFD) is not the physical core diameter: it is the width of the guided mode’s power distribution, the true optical size a splice actually sees. ITU-T G.652 (08/2024), clause 6.1, requires every fibre to declare a nominal value and a tolerance at 1310 nm: “Both a nominal value and tolerance about that nominal value shall be specified at 1310 nm.” It is not a datasheet footnote: it is a compliance requirement.

The numbers in clause 8 are not the same across categories. The legacy G.652.B fibre declares a nominal range of 8.6–9.5 µm with a tolerance of ±0.6 µm. G.652.D, the fibre most backbones are built with today, has a narrower range: 8.6–9.2 µm with ±0.4 µm. ITU-T G.657 (08/2024) aligns subcategory A (A1 and A2) to exactly those same two figures, and is — in the Recommendation’s own words — “fully compliant with the ITU-T G.652 single-mode fibres”. Subcategory B3, the one for the tightest radii, in dense patch panels or close to a building, declares today the same range, but remains only “system compatible with ITU-T G.657.A and ITU-T G.652.D fibres in access networks”: compatible, not compliant — it introduces, the Recommendation states, “negligible system impairment or deployment issues”, without matching the same specification point for point.

Why the same splice loses one way and seems to gain the other

An OTDR does not measure a splice’s loss: it measures the light Rayleigh-backscattered along the fibre, and infers the loss from the step it sees at the splice point. If the two fibres facing each other have a different backscatter coefficient — and two fibres with different mode fields almost always do — that step does not match the true loss. ITU-T G.650.3 (08/2017), Appendix I, states it plainly: “Where the fibre after the splice scatters more light than the fibre before the splice, the apparent loss is reduced and may even show up as an apparent gain.” If the ratio reverses, the same appendix names the event “a ‘gainer’ (apparent negative splice loss) when measured from the opposite end” — a splice that, measured from one end alone, looks as if it generates power instead of consuming it.

This is not an instrument fault: it is a geometric artefact of one-way measurement. The same Recommendation, clause 6.1.3, resolves it with a plain requirement — “accurate splice loss measurement must be based on the bidirectional OTDR test” — and a formula spelled out in the text: the true loss is the average of the value measured from one end and from the opposite end, (αA→B + αB→A) / 2. A gainer read in one direction only is not proof of an error: it is the missing half of a calculation that anyone reading an OTDR report should always demand in full.

What a mode field mismatch costs, in numbers — and their limits

How much can a mode field mismatch cost, in decibels, within the declared tolerance alone — with no cleave error or misalignment? G.650.3 itself, Appendix III, does the sum for G.652.B: starting from the clause 8 figures — range 8.6–9.5 µm, tolerance ±0.6 µm — the Recommendation states that “it is sufficient to consider the MFD range 8.0 µm to 10.1 µm” in the worst case. That is a value written in the Recommendation, not one of our own calculations.

Applying the same method ourselves to G.652.D and G.657.A/B3 (tolerance ±0.4 µm on 8.6–9.2 µm), the worst case narrows to 8.2–9.6 µm. Using the Gaussian mode-overlap formula common in fibre-optic engineering literature for loss from mode field mismatch alone — not given as such in G.650.3, and here the calculation is ours — the G.652.D/G.657 worst case works out to about 0.11 dB; the same sum on the G.652.B worst case, 8.0 against 10.1 µm, works out to about 0.23 dB. These are illustrative figures that isolate mode field mismatch alone: a real splice almost always adds a cleave misalignment, which costs more. On splicing fibres of different substance, G.652 itself, clause 6.7.1, notes: “Care may be needed in fusion splicing fibres of different substances. Provisional results indicate that adequate splice loss and strength can be achieved when splicing different high-silica fibres.”

What to write when two suppliers’ runs meet

When a G.652.D backbone from supplier A meets a G.657.A2 access run from supplier B — the most common case in a network built in successive lots — a specification cannot stop at the recommendation’s name. It must require:

  1. the exact subcategory declared for each lot meeting at the splice, not just compliant with G.652 or compliant with G.657;
  2. the lot’s nominal MFD value, from the manufacturer’s datasheet, for both sides of every splice between different suppliers — not just the Recommendation’s range;
  3. bidirectional OTDR measurement on every mixed splice, flagged as such in the as-built, with loss calculated as the average of both directions;
  4. a written acceptance threshold per splice, the same criteria already set for fusion and mechanical splicing, applied specifically to splices between suppliers;
  5. an isolated, unaveraged gainer, treated as non-compliant: a negative value in one direction alone does not close out testing — it must always come with the bidirectional average.

Where the proof stands, today

The proof that a splice between different suppliers was measured properly sits in four places, not one conformity statement: the fibre lot’s datasheet, with the nominal MFD the manufacturer declares for both sides; the native OTDR trace in both directions, not just the average value the software has already computed; the as-built, with every splice between different suppliers flagged as such, not confused with an ordinary splice within the same lot; and the test report, with the average calculation shown next to the value, not a single figure with no measurement direction attached.

One clear limit, worth stating: in none of the Recommendations we read did we find a numerical mode field mismatch threshold beyond which a specification must reject a splice. G.650.3 offers the measurement method and the worked example for the G.652.B worst case, not a pass/fail value for site acceptance. That threshold, like the maximum loss per splice, is written only by whoever signs the specification.

The bottom line

An isolated gainer is not good news dressed up as an instrument fault: it is proof that someone measured in one direction only. And a mode field mismatch within tolerance — even between two fibres both compliant with the same recommendation — is not a manufacturing error: it is physics the specification must plan for, not ignore. It is the check we run on every mixed splice when we certify a run, and the clause we write when two suppliers meet in a specifications and compliance engagement. Every mixed splice, its bidirectional average and the MFD of both lots feed into the same network map — not scattered job files — where an AI flags the anomalous deviation and one of our teams steps in, for an operator, a data centre, an industrial site, a public administration body, a healthcare facility or a defence site. Where an AI is needed on network data, it runs within the client’s own perimeter — on-premise or on dedicated cloud with a data centre in Italy — alongside CSIDIA, the group’s other company.

Do you have a run where fibres from different suppliers meet, or a splice whose test value doesn’t add up? Talk to an engineer: the site visit is at no cost, and the bidirectional re-measurement is run with the same OTDR used for testing, before anyone has to reopen a closure.

Sources