Request an intervention

Operational notes Engineering

How many dB does a PON splitter lose? The ITU-T G.671 thresholds

4 min read

LC connectors plugged into a fibre-optic patch panel inside a distribution cabinet
Every connector on this panel hides a split ratio that the specification must state as a number, not an adjective.

A specification that reads “PLC splitter compliant with current standards” is not specifying anything. Standards do not set a single number: they fix a range that changes with the split ratio, the wavelength window, and the reach of the network. Whoever designs the ODN — the optical distribution network between the exchange and the user — has to decide which value in that range enters the optical budget, and which stays as margin. There is only one recommendation that sets these numbers: ITU-T G.671.

Two editions in six months.

The current version of ITU-T G.671 — “Transmission characteristics of optical components and subsystems” — was approved on 29 November 2025. It replaces an edition from just six months earlier, dated 14 May 2025: within a single half-year, ITU-T Study Group 15 revised the parameters of passive optical components — splitters included — twice. The clause covering access networks is 6.6, “Optical branching component (wavelength non-selective) for PONs”: it covers dividers with one or two inputs (two inputs are for redundant configurations) and from 2 to 64 outputs.

The numbers the specification must cite.

The table in clause 6.6 sets minimum and maximum insertion loss for every split ratio, in the WR1 operating window (1310/1550 nm) and for “normal” reach (single input):

  • 1:2 — 2.8 to 3.9 dB
  • 1:4 — 5.4 to 7.4 dB
  • 1:8 — 8.2 to 10.6 dB
  • 1:16 — 10.8 to 14.1 dB
  • 1:32 — 13.3 to 17.5 dB
  • 1:64 — 16.1 to 20.9 dB

These are thresholds, not nameplate estimates: a splitter with a declared loss outside this bracket — in excess, or suspiciously below the minimum — does not comply with clause 6.6, whatever the manufacturer’s datasheet says. The optical budget of a link is built by adding these figures to cable, splice and connector losses: a 1:32 splitter alone weighs more than ten times an 8 km run of installed fibre.

Normal reach, extended reach, two windows.

The standard also distinguishes an “extended reach” grade, with different thresholds: for the 1:32 ratio, for example, the maximum allowed loss drops from 17.5 to 17.1 dB in the same WR1 window. The specification must state which of the two applies, because the choice ties to the optical class of the equipment: as we wrote for GPON and XGS-PON, a class C+ or E1 budget absorbs an extended-reach splitter where a class B+ budget cannot.

There is a second distinction that almost no specification records: the WR2 window, wider than the operating WR1, reaches up to 1625 or 1660 nm depending on the sub-category. This is not a footnote for manuals: it is the window needed when the network must remain measurable with a live-fibre OTDR, at the maintenance wavelength, without switching off data traffic to test it. A splitter that only meets WR1 guarantees nothing about in-service measurement.

The cascade changes the sum, not just the final number.

An ODN rarely installs a single 1:32 splitter: more often it places a 1:4 near the exchange and a 1:8 near the user, so the distribution network can be upgraded without redoing the primary link. The sum of the losses of two cascaded stages almost never matches that of a single stage with the same overall ratio: 1:4 (5.4–7.4 dB) plus 1:8 (8.2–10.6 dB) gives a range of 13.6–18.0 dB, slightly worse than the 13.3–17.5 dB of a single-stage 1:32. That is the price of network flexibility, and it belongs in the budget from the design stage, not something discovered during testing.

What to actually write in the specification.

  • The exact split ratio (1:X or 2:X) and whether normal or extended reach is required.
  • The maximum insertion-loss value taken from clause 6.6 of ITU-T G.671, not a “typical” figure from a commercial datasheet.
  • The required wavelength window — WR1 if in-service operation is enough, WR2 if live-fibre OTDR testing is required.
  • The test method: the standard refers to IEC 61300-3-4 and IEC 61300-3-7 for the insertion-loss measurement.
  • Uniformity and polarisation-dependent loss (PDL), if the link feeds sensitive equipment: the same clause tabulates both.

The point.

A splitter is not a passive component in the sense of “no consequences”: it is the item that weighs most in the optical budget of a PON network, and it is the only one whose value is set by an international standard with precise thresholds, not by a commercial catalogue. In the ODN projects we run the split ratio is chosen together with the optical budget, not after the cabinet has already been installed.

Do you need to size a splitting stage, or check whether an existing one still has room to grow? Talk to an engineer: the calculation is done together, on your own network figures.

Sources