Request an intervention

Operational notes Engineering

Fusion or mechanical splicing: what to put in the spec

4 min read

Fusion splicer for optical fibre on a workbench, with fibres being prepared
The splice takes thirty seconds to make. Its consequences last twenty years.

In many specifications, splicing optical fibres is a single line: “executed to good workmanship standards”. Then testing day arrives, and you discover what that line actually meant to whoever laid the cable. Fusion and mechanical splicing are not equivalent: they differ in loss, reflectance and lifespan, and the wrong choice is paid for over the entire life of the network. This note sets out the real numbers and — above all — what to write in the spec so that the choice is not made by someone else on your behalf.

Two techniques, two principles.

Fusion splicing welds the two pieces of glass with an electric arc: the fibres are stripped, cleaned, cleaved, aligned — on the core, in the best machines — and fused into a continuous joint, protected by a heat-shrink sleeve. Optically, the link goes back to being a single fibre.

Mechanical splicing welds nothing: it aligns the two fibres in a precision groove and clamps them. Optical continuity is entrusted to an index-matching gel that fills the micro-gap between the two end faces. It works, but a physical discontinuity remains inside the link.

The numbers that matter.

Typical loss. A well-executed fusion splice on single-mode fibre compliant with ITU‑T G.652.D loses 0.02–0.05 dB. A mechanical splice typically loses 0.1–0.3 dB, with far greater variation from one splice to the next: it depends on the quality of the cleave and the condition of the gel.

Reflectance. A fusion splice is normally non-reflective: below −60 dB, often unmeasurable. A mechanical splice does reflect: typical values sit between −35 and −45 dB. On high-speed systems and on PON networks, reflective events are not a cosmetic detail: they disturb transmitters, degrade the signal and complicate OTDR measurements across the whole link.

Lifespan. A fusion splice is welded glass: if the sleeve is fitted properly, it ages like the fibre itself. A mechanical splice depends on the gel, which becomes contaminated, dries out and suffers thermal cycling over the years — especially in a duct chamber or an outdoor cabinet. A loss that was acceptable at commissioning may no longer be so after five summers.

When mechanical splicing makes sense.

Mechanical splicing is not a wrong technique: it is a technique for specific contexts. It makes sense for the emergency restoration of a cut link, pending the definitive fusion splice. It makes sense for a handful of occasional splices where bringing in and amortising a fusion splicer is not justified. It makes sense for the odd temporary termination that will later be dismantled.

It does not make sense as the standard for backbones, telecommunications networks, data centres, or any link that must be certified and maintained for years: there, it becomes the bottleneck of the optical budget and a reserve of future faults.

What to put in the spec.

“To good workmanship standards” is not a requirement: it is a blank cheque. A serious specification writes down the numbers.

  1. Declared splicing technique. Fusion for all permanent splices; mechanical splicing permitted only for temporary restorations, with a mandatory obligation of replacement.
  2. Maximum loss per splice. A sound benchmark for fusion: bidirectional average ≤ 0.1 dB per splice, no splice above 0.15 dB. And a lower link average (≤ 0.05 dB) where the optical budget design requires it.
  3. Reflectance. Non-reflective splices; for connectors, an explicit limit (for example ≤ −50 dB for UPC connectors, ≤ −60 dB for APC).
  4. Compliant and compatible fibre. ITU‑T G.652.D for backbones; G.657.A2 where tight bend radii are required (building risers, FTTH terminations). G.657.A fibres can be spliced to G.652 fibres with no particular penalty: this should be stated explicitly, so no one improvises.
  5. Bidirectional certification. OTDR measurement in both directions at 1310 and 1550 nm in accordance with ITU‑T G.650.3, with splice loss calculated as the average of the two directions, and delivery of the native traces alongside the report.
  6. Installation standards. Cable routing and infrastructure to the EN 50174 series; structured cabling compliant with the EN 50173 series where applicable (building, campus, data centre).
  7. As-built documentation. Final documentation with a link-by-link diagram, the position and rack of every splice, labelling, and precise correspondence with the commissioning traces.

With these seven items, testing stops being a negotiation and becomes a verification: either the numbers are there, or they are not.

The bottom line.

The quality of an optical network is decided before the first cable is laid: in the specification. Whoever splices well has no fear of written values; whoever fears them is telling you something. It is the criterion we work by, from splicing to certification, and it is why our approach always starts from measurable numbers, never from ritual formulas.

Do you need to write or assess a splicing and testing specification? Let’s talk: a review carried out before the tender costs little, a badly spliced network costs for twenty years.

Sources