A point-to-point link has connectors and splices that each cost a few tenths of a dB. A PON has the same, plus one component that costs far more than everything else combined: the splitter.<\/p>\n\n\n\n
Every time the signal is split in two, the power on each branch drops by about 3 dB. A 1×32 splitter is five of those splits, so roughly 15 dB in theory, and more in practice once excess loss is added. One PLC splitter datasheet, for example, lists maximum insertion loss of 10.5 dB for 1×8, 13.5 dB for 1×16 and 16.5 dB for 1×32. Use the datasheet of the splitter actually installed when you check a reading.<\/p>\n\n\n\n
On the trace this shows up as a cliff. Three things follow from it:<\/p>\n\n\n
- \n
- \nDynamic range runs out quickly.<\/strong> The OTDR has to see past the splitter and still have enough signal to show the drop fiber and the far-end connector. A lower-end unit can reach the splitter and show only noise after it.<\/li>\n\n\n\n
- \nThe auto mode often misreads it.<\/strong> Many OTDRs treat a sudden large loss as the end of the fiber. The event table then reports the splitter as “fiber end”, and the link length is wrong.<\/li>\n\n\n\n
- \nThere is a dead zone after it.<\/strong> The receiver needs time to recover after a large event, so faults close to the splitter output are hard to see.<\/li>\n<\/ul>\n\n\n
Which End to Test From<\/strong><\/h2>\n\n\n\n
This is the decision that matters most, and the one most often made wrong.<\/p>\n\n\n
From the OLT side: the branches pile up<\/strong><\/h3>\n\n\n\n
Testing downstream from the OLT, the trace is clean up to the splitter. After the splitter, the pulse travels into all 32 branches at once, and the backscatter from all of them comes back on the same fiber. The OTDR adds it all together.<\/p>\n\n\n\n
The result is a trace that tells you very little about any single drop. Branches of similar length overlap. A broken drop shows up, at best, as a small step that could belong to any of the 32 customers. Each branch only contributes a fraction of the returned light, so a fault on one of them barely moves the combined trace.<\/p>\n\n\n\n
The OLT side is still the right place to check the feeder fiber from the central office to the splitter. It is not the place to troubleshoot a single customer.<\/p>\n\n\n
From the ONT side: one path, one trace<\/strong><\/h3>\n\n\n\n
Testing upstream from the customer’s end, the OTDR sees one drop fiber, the splitter as a single loss step, and the feeder fiber behind it. Everything before the splitter belongs to that one customer, so a bend in their drop or a bad connector at their wall box is clearly visible.<\/p>\n\n\n\n
For fault-finding on a single subscriber, this is the standard approach. The trade-off is that the feeder after the splitter is seen with 15 dB or more less signal, so the dynamic range of the OTDR decides how far past the splitter you can still see detail.<\/p>\n\n\n\n
Test from<\/th> Good for<\/th> Not good for<\/th><\/tr><\/thead> OLT \/ central office side<\/td> Feeder fiber to the splitter, splitter location<\/td> Faults on individual drops<\/td><\/tr> ONT \/ customer side<\/td> Drop fiber, customer connectors, splitter loss for that branch<\/td> Faults on other customers’ branches<\/td><\/tr><\/tbody><\/table><\/figure>\n\n\n Settings That Work on a PON<\/strong><\/h2>\n\n\n
Range and pulse width<\/strong><\/h3>\n\n\n\n
Don’t leave either on auto. Set the range to cover the full path from the ONT to the OLT, plus some margin. The pulse width has to balance two things that pull against each other:<\/p>\n\n\n
- \n
- A short drop (often a few hundred meters) needs a short pulse to separate the wall box connector, the drop splices and the splitter connector.<\/li>\n\n\n\n
- Seeing past a 15–17 dB splitter needs a wider pulse for more energy.<\/li>\n<\/ul>\n\n\n\n
Often one acquisition can’t do both. Take two: a short pulse to look at the drop up to the splitter, and a wider one to look at the feeder beyond it. Save both in the report.<\/p>\n\n\n
End-of-fiber threshold<\/strong><\/h3>\n\n\n\n
Raise the end-of-fiber detection threshold above the expected splitter loss, so the auto analysis does not stop at the splitter. If your OTDR has a PON or FTTH mode, it usually does this for you. Check what it sets before relying on it.<\/p>\n\n\n
Launch cable<\/strong><\/h3>\n\n\n\n
- \nThe auto mode often misreads it.<\/strong> Many OTDRs treat a sudden large loss as the end of the fiber. The event table then reports the splitter as “fiber end”, and the link length is wrong.<\/li>\n\n\n\n
![\"F7](\"https:\/\/www.tfngj.com\/wp-content\/uploads\/2026\/08\/f7-use-1-1024x610.webp\")
<\/figure>\n\n\n\n