The Splitter That Was Not in the Drawing: GPON Power Budget and OTDR
A GPON FTTH network serves a housing colony from a Class B+ OLT port: transmit power +3.0 dBm at 1490 nm, ONT receiver sensitivity −28.0 dBm. The as-built drawing shows a 20.2 km run, made up of 18.0 km of G.652D feeder to the splitter cabinet plus 2.2 km of distribution, with two fusion splices, three connectors (OLT patch panel, splitter tray, ONT), and a 1:8 planar splitter of 10.5 dB insertion loss. Feeder attenuation is taken as 0.35 dB/km at 1490 nm, splices 0.10 dB, and connectors 0.50 dB. The operator's design margin is 3 dB. A new subscriber at the far end syncs but reports frequent loss of signal and occasional ONT resets. The technician's power meter reads −23.5 dBm at the ONT. An OTDR run on the 18.0 km feeder shows a non-reflective event of 2.8 dB at 12.4 km and a reflective event of 0.4 dB at 18.0 km. The splitter in the cabinet is marked 1:16 on its body, although the drawing shows 1:8.
1. Calculate the expected receive power and the design margin, then compare with the measured −23.5 dBm.
Read the worked answer
Losses at 1490 nm:
- Fibre: 20.2 km × 0.35 dB/km = 7.07 dB
- Splices: 2 × 0.10 = 0.20 dB
- Connectors: 3 × 0.50 = 1.50 dB
- Splitter (1:8): 10.50 dB
- Total: 19.27 dB
Expected receive power = +3.0 − 19.27 = −16.3 dBm. Margin to sensitivity = −16.3 − (−28.0) = 11.7 dB, comfortably above the 3 dB design rule.
The measured −23.5 dBm is 7.2 dB worse than the design predicts and leaves only 4.5 dB to sensitivity. That is still above the 3 dB rule on paper, yet the link is failing. The lesson is that a 3 dB rule protects a new, clean, room-temperature link; it does not protect a real one with dirty connectors, tight bends, and a mis-installed splitter. Four and a half decibels of measured margin with an unexplained 7.2 dB gap is a fault report, not a pass.
2. Classify the two OTDR events and say which needs physical intervention.
Read the worked answer
The 2.8 dB non-reflective event at 12.4 km is a splice or a bend. A good fusion splice measures 0.05–0.15 dB, so this event carries roughly 2.7 dB of excess loss. Non-reflective means no air gap, which argues for a poor fusion splice (wrong arc, contaminated fibre, bad cleave), a mechanical splice, or a tight bend rather than a broken or dirty connector. This one needs the cabinet or joint opened and the splice re-made.
The reflective event of 0.4 dB at 18.0 km is the splitter-tray connector. A reflection indicates an air interface, and 0.4 dB is inside the 0.5 dB connector budget, so it is acceptable. Clean and inspect it, but do not cut it out.
Accounting for the gap: about 2.7 dB from the splice, plus 3.3 dB because a 1:16 splitter (13.8 dB) is fitted where the drawing calls for 1:8 (10.5 dB), leaves roughly 1.2 dB unexplained in the distribution and drop section. Suspect a tight bend at the ONT tray or a dirty ONT connector. The arithmetic matters: if you cannot account for the loss, you have not finished fault-finding.
3. The link is technically inside the design margin, so the project manager wants no action. Give your recommendation and the cost of doing nothing.
Read the worked answer
Recommend action anyway, prioritised by decibels recovered per rupee:
- Re-make the splice at 12.4 km: about 2.6 dB recovered, one joint, one visit.
- Resolve the splitter discrepancy. Either fit the specified 1:8 or accept 1:16 as a deliberate design change with the budget recalculated and the as-built corrected: 3.3 dB recovered in the first case, and in the second case a documented budget that no longer lies about the plant.
- Clean the ONT connector and check the bend radius in the tray: about 1.2 dB recovered.
After these, predicted receive power is roughly −17.5 dBm with a margin near 10.5 dB.
Doing nothing means the link runs on 4.5 dB of margin before ageing, temperature swing, and connector dirt. Loss of signal and ONT resets here are also consistent with poor optical return loss rather than pure attenuation, and the 2.8 dB event makes the budget temperature-sensitive, which is exactly what a marginal link cannot tolerate.
There is also a plant-integrity reason: a 1:16 splitter in a cabinet documented as 1:8 means every subscriber on that cabinet is running 3.3 dB poorer than the records claim. Audit all ports and correct the as-built now, while you already have a crew on site.