One Low ONT on a 1:32 Splitter
You are troubleshooting a GPON serving 32 customers from a single 1:32 PLC splitter. Class B+ optics: OLT launch +1.5 to +5.0 dBm, ONT receiver sensitivity -27.0 dBm at 1490 nm. The feeder is 12 km and the longest drop is 3 km. One customer reports intermittent service. At that ONT you measure -26.8 dBm at 1490 nm with a handheld power meter, and the ONT's own RSSI reads -26.5 dBm. The other 31 ONTs on the same splitter report between -18.9 and -21.3 dBm. The customer's drop was replaced six months ago after a dig-in, and the splice closure at the pole was re-done at that time. The power meter was last calibrated 14 months ago; its certificate gives +/-0.5 dB at 1310/1550 nm and it has no 1490 nm calibration point. There is no PON OTDR on the van, only a 1310/1550 nm OTDR with 34 dB dynamic range, and a visual fault locator.
1. Is -26.8 dBm acceptable? Support your answer with an expected-power calculation.
Read the worked answer
Expected receive power (mid-range launch of about +3 dBm):
- Feeder plus drop fiber: 15 km x 0.25 dB/km at 1490 nm = 3.75 dB
- 1:32 PLC splitter: 10 log(32) = 15.05 dB ideal, plus excess and uniformity, about 16.5 dB typical (17.5 dB worst case)
- Connectors and splices (OLT, feeder ends, splitter ports, ONT, two closures) at about 0.3 dB each, roughly 1.8 dB
- Total: about 22.0 dB, giving about -19.0 dBm
The other 31 ONTs at -18.9 to -21.3 dBm confirm this model. This ONT is 6 to 8 dB below its neighbours.
Against the sensitivity of -27.0 dBm, -26.8 dBm passes by 0.2 dB - which is not a pass in any practical sense. That margin disappears with temperature swing, ageing or an additional dirty connector, and it explains the intermittent service. The complaint is consistent with a real fault on this drop, not with a slow splitter or a weak OLT.
2. How do you isolate the fault with the instruments you have? Address the meter's calibration in your answer.
Read the worked answer
First validate the number. The meter has no 1490 nm calibration point, but 1490 nm is close to 1550 nm, so the wavelength-extrapolation error is small - well under 0.5 dB - and cannot explain a 6-8 dB deficit. More importantly, the ONT's own RSSI (-26.5 dBm) agrees with the meter, so the low power is real, not an instrument artefact.
Then work from the customer inward:
- Inspect and clean the ONT connector and the drop's connector at the ONT; re-measure. A contaminated SC/APC at the ONT is a common several-dB fault.
- Check the fibre route at the ONT for a tight coil or a kink. 1490 nm is more bend-sensitive than 1310 nm, so a bend that was harmless on an older link can cost several dB here.
- Swap the drop onto a spare splitter port. If the low power follows the drop, the fault is in the drop or its closure; if the new port is equally low, the port or splitter leg is suspect. Also record a known-good drop on the original port to clear it.
- With the drop disconnected at the splitter, fire the VFL from the ONT end - a bright glow at a bend or a break is immediately visible, and the re-made closure is the prime suspect.
- Then run the 1550 nm OTDR from the ONT end over the 3 km drop. That path is fully visible because the splitter terminates it; it will show a macrobend, a poor splice in the re-made closure, or a connector event.
3. The supervisor suggests OTDR-testing from the CO through the splitter to find the fault directly. Explain why that will not work here and what the correct PON test approach is.
Read the worked answer
A 1:32 splitter costs about 16.5 dB each way, so an OTDR pulse returning from any drop has lost roughly 33 dB at the splitter alone, before the 3.75 dB of fiber. Against a 34 dB dynamic range instrument, the splitter appears as the end of the fiber: everything beyond it is buried in noise. Even with a higher-dynamic-range instrument, the return from a splitter is the sum of 32 drop fibers - you cannot attribute an event to one customer's drop, and a single bad splice in one drop is diluted by 31 good ones.
Correct approach:
- Test the feeder from the CO with the existing OTDR; the splitter is a legitimate end event, and the feeder is fully characterized that way.
- Test each drop from the ONT end, disconnected at the splitter, so the drop is a simple point-to-point span.
- For in-service testing, use a PON OTDR at 1625 nm (outside the 1490/1310 passbands) with wavelength-selective filters at the OLT and ONT to protect live optics - but the splitter still limits visibility beyond it, so single-drop attribution still comes from testing at the ONT.
After the repair, re-measure at 1490 nm and compare against the other 31 ONTs; the acceptance target is being within a couple of dB of the population, not merely inside the -27 dBm sensitivity limit.