Rural GPON power budget: 3.7 dB of unexplained excess loss
A rural electric cooperative is building FTTH to 640 premises in three villages. The design places one 1:32 GPON splitter in the central office for each village and runs a single 14.8 km feeder to the farthest ONT. The OLT uses Class B+ optics rated for a 28 dB ODN budget. At commissioning, the nine most distant ONTs register, but a power-meter test through the link reads 27.9 dB loss at 1490 nm, and those ONTs deregister on hot afternoons when the cabinet reaches 41 °C. The design calculation assumed 0.35 dB/km at 1490 nm, 17.0 dB splitter loss, four 0.3 dB connectors, and eight 0.1 dB splices. The contractor proposes three remedies: move the 1:32 splitter to a remote FDT 11 km out, replace the OLT with Class C+ optics, or re-split the village as 1:16. The co-op has 20 PON ports and 40 spare feeder fibers into the village. Capital is tight, and 640 customers are waiting.
1. Compute the expected ODN loss and compare it with the measured 27.9 dB. What does the gap mean, and what would you measure before approving any design change?
Read the worked answer
Expected loss:
- Fiber: 14.8 km × 0.35 dB/km = 5.18 dB
- Splitter: 17.0 dB
- Connectors: 4 × 0.3 = 1.2 dB
- Splices: 8 × 0.1 = 0.8 dB
- Total: 24.18 dB
The measurement is 27.9 dB, so 3.7 dB is unexplained. At a 28 dB budget, the link has only 0.1 dB of headroom - far below the 3 dB planning margin normally reserved for aging, temperature, repairs, and measurement uncertainty. A 3.7 dB discrepancy is too large to dismiss as meter tolerance.
Before changing architecture, localize the excess. Measure with a calibrated power meter at the OLT, at the FDT, and at several ONT ports. Run OTDR traces from the CO at 1490 nm and 1550 nm with a launch fiber, and compare near and far ONTs. If every port on the splitter shows the excess, suspect the splitter, patch panel, or OLT connector. If only distant ONTs show it, suspect feeder splices, a tight bend in the cabinet, or a damaged cable section. Inspect endfaces, verify splitter port loss against its data sheet, and check splice losses using bidirectional averages. Correct the fault first.
2. Evaluate the three proposed remedies. Which would you choose, and in what order?
Read the worked answer
Move the splitter to a remote FDT: this does not reduce loss. The fiber path remains 14.8 km, and a single 1:32 splitter still contributes about 17 dB. A remote 1:32 enclosure adds cost and a powered cabinet without optical gain. A remote 1:8 with four feeder fibers would save roughly 6.5 dB, but needs four PON ports per village - 40 ports total, exceeding the 20 available.
Re-split to 1:16: saves about 3.5 dB but requires 40 PON ports and 40 feeder fibers (640 ÷ 16). The port count is not available without a second OLT card.
Class C+ optics: raises the budget to 32 dB, a 4 dB gain, and is the cheapest optical fix if both OLT and ONT support C+ and the upstream 1310 nm budget also closes. Its weakness is that it masks a 3.7 dB fault that may grow.
Order of work: find and repair the excess first. If the corrected loss is 24.2 dB, the margin is 3.8 dB - acceptable with a 3 dB aging allowance. If the excess is inherent, for example extra splices from a long bore, then C+ is preferred. Re-splitting is the last resort because it consumes scarce PON ports.
3. If the corrected loss is 24.2 dB, what acceptance records and ongoing checks would you require for the nine distant ONTs?
Read the worked answer
Record bidirectional OTDR traces for the feeder, power-meter readings at the OLT, FDT, and each ONT, splitter port assignments, and a splice-loss table with bidirectional averages. Include the calculated and measured loss for each of the nine links, the test wavelength and meter calibration date, and the ORL.
Because the cabinet reaches 41 °C, repeat the power measurement at peak temperature and require the ONT receive level to remain above sensitivity with at least 1 dB margin. Set an alarm threshold based on measured baseline, not on the nominal sensitivity limit. Document the spare port count and fiber assignments so a future technician does not disturb the working path. Finally, require a 24-hour error-counter check on all nine ONTs to confirm stability before closing the job.