GPON Link That Works at Night but Drops Every Afternoon
An ISP is commissioning a GPON link from its block exchange to a gram panchayat office 24 km away. The OLT PON port launches +3 dBm and the ONT needs at least −27 dBm, giving a 30 dB optical budget; company policy reserves 3 dB for ageing and repair, so design loss must stay below 27 dB. The design assumes 24 km of G.652D fibre at 0.35 dB/km, ten fusion splices at 0.10 dB each, three connectors at 0.5 dB each and one 1:8 splitter at 10.7 dB. After installation the customer reports that the link works at night but drops every afternoon. The OTDR trace taken from the OLT at 1310 nm shows a total loss of 28.4 dB, an event of 4.6 dB at 9.3 km, and average fibre attenuation of 0.43 dB/km. The splicing team says the 9.3 km splice 'looked fine' on the fusion splicer's estimated-loss display, and proposes raising the OLT launch power by 3 dB instead of reopening the splice.
1. Calculate the designed link loss and the actual worst-case received power at the ONT. Does the link meet company policy? Explain why it works at night but fails in the afternoon.
Read the worked answer
Design loss = fibre 24 × 0.35 = 8.4 dB; splices 10 × 0.10 = 1.0 dB; connectors 3 × 0.5 = 1.5 dB; splitter 10.7 dB. Total = 21.6 dB, leaving 27 − 21.6 = 5.4 dB of permitted headroom, so the design is sound.
Actual received power = +3 − 28.4 = −25.4 dBm. This is above the ONT's −27 dBm sensitivity, so the link is not dead - but the margin is only 1.6 dB, below the 3 dB policy reserve, so the link fails policy by 1.4 dB.
That 1.6 dB explains the symptom. In the afternoon the cable and joint closures heat up, fibre microbending and connector losses rise by a decibel or two, and the margin disappears. At night the loss falls back and the ONT re-registers. A link that 'works at night' is not healthy; it is running with no reserve. The 4.6 dB event and the 0.43 dB/km average show real plant defects, so the correct conclusion is that the link is marginal by design and must be repaired, not that the ONT is faulty.
2. Quantify the excess loss, identify its two main contributors, and state the first corrective action. Why is the fusion splicer's estimated-loss reading not proof that the splice is good?
Read the worked answer
Excess loss = measured − design = 28.4 − 21.6 = 6.8 dB.
Breakdown:
- The 9.3 km event should be about 0.10 dB but reads 4.6 dB, contributing roughly 4.5 dB of excess.
- Fibre attenuation is 0.43 dB/km instead of 0.35 dB/km, contributing 24 × 0.08 = 1.9 dB.
- 4.5 + 1.9 = 6.4 dB; the remaining ~0.4 dB is within connector and measurement tolerance.
First action: reopen and re-splice at 9.3 km - fresh cleave with a controlled cleave angle, clean fibre and V-grooves, correct arc parameters, then verify from both directions with the OTDR.
The splicer's 'estimated loss' is derived from fibre alignment and arc conditions, not from a transmission measurement; it can read 0.00 dB on a splice with trapped contamination, a bad cleave angle or a fibre mismatch. Only an OTDR (bidirectional, averaged) or a power-meter loss test proves splice quality. At the same time, inspect the cable at 9.3 km for tight bends or tension, because the elevated 0.43 dB/km suggests route damage, not just one bad joint.
3. The team lead suggests simply raising the OLT launch power by 3 dB. Evaluate that proposal and state what should be done about the 0.43 dB/km fibre loss.
Read the worked answer
Raising launch power to +6 dBm gives −22.4 dBm at the ONT, a 4.6 dB margin, so the numbers would pass. But this is masking, not repair:
- The 4.6 dB splice will keep degrading with temperature and moisture; the 4.6 dB margin will erode and the link will fail again, probably in the same afternoon pattern.
- Higher launch power increases Rayleigh backscatter and reflection effects, and pushes the port beyond the operating class the link was engineered and safety-assessed for.
- It hides the real fault from the next technician, who will see a 'passing' link budget and stop looking.
The 0.43 dB/km figure is 23% above specification over 24 km, which is not normal ageing. Treat the route as suspect: check for tight bend radii at closures and trays, crushed or kinked duct, water ingress in joint closures, and cable damaged during pulling. Ask for a bidirectional OTDR trace section by section, repair the splice and any damaged cable section, then re-measure. Only if the link still falls short of 27 dB after repair should power or splitter ratio be revisited.