A loss budget is one long subtraction. You start with the power the transmitter puts into the fiber, take away everything the light meets on the way, and check that what arrives is still above what the receiver can hear. If the answer is positive with margin to spare, the link closes. If it isn't, something upstream has to change.
The arithmetic is simple. What makes it error-prone is the bookkeeping — a GPON tree with two splitter stages and eleven splices has a lot of small numbers, and they are usually written on the back of a printout.
The five things that eat your decibels
- Fiber attenuation. Distance × loss per km. Budget roughly 0.35 dB/km at 1310 nm and 0.25 dB/km at 1550 nm for standard single-mode, and use the figure on your cable spec sheet when you have it.
- Splices. A good fusion splice costs about 0.1 dB. Plan 0.15 dB if the crew is working in a chamber in the rain.
- Connectors. Around 0.5 dB per mated pair. Count patch panels, not just the ends.
- Splitters. The big one. A 1:2 costs about 3.5 dB, a 1:32 about 17.5 dB. See splitter ratios and dBm.
- Margin. Keep 3 dB back for ageing, repairs and the splice someone will add in year four.
A worked example
Take a real-shaped FTTH distribution area: 8.2 km of route, a headend transmitter at +17.0 dBm, one 1:2 split at the street cabinet, then a 1:32 splitter serving the block. The ONT needs −27 dBm to work — the Class B+ sensitivity figure.
| Element | Detail | dB |
|---|---|---|
| Transmitter output | Headend EDFA T-1 | +17.0 |
| Fiber, feeder | 6.7 km × 0.35 | −2.35 |
| Fiber, distribution | 1.5 km × 0.25 | −0.38 |
| Connectors | 4 mated pairs × 0.5 | −2.00 |
| Splices | 11 × 0.1 | −1.10 |
| Splitter, 1:2 | street cabinet | −3.50 |
| Splitter, 1:32 | block splitter | −17.50 |
| Received power at ONT | vs. −27.0 dBm sensitivity | −9.83 |
That leaves 17.2 dB of headroom over the ONT's −27 dBm sensitivity — comfortable. Take the 3 dB planning margin off and you still have 14 dB, which means this area could take a longer feeder or a deeper second split if the subscriber count grows.
Do this on your phone instead.
FiberLink runs the whole subtraction for you as you build the tree — every node shows its own line-by-line math, and the finished plan exports as a PDF. Get it free on Google Play →
Where budgets usually go wrong
Three mistakes account for most failed links:
- Counting the splitter once when it splits twice. Cascaded splitters multiply subscribers and add their losses. A 1:4 followed by a 1:8 is not a 1:32 — it is about 0.8 dB worse.
- Forgetting connectors inside cabinets. Every patch panel adds a mated pair. Two panels quietly cost 1 dB.
- Using route length instead of cable length. Slack loops, risers and service coils add 5–10%. Measure the fiber, not the trench.
Check it against the meter
A budget is a prediction. When the crew shoots the link with a power meter, the two numbers should agree within about 1 dB. A bigger gap means a bad splice, a dirty connector or a bend — and the value of having the calculation node by node is that you can bisect the tree and find it in one visit rather than three.