FiberLinkIn a passive optical network the splitter is where most of your light goes. Not the kilometres, not the splices — the splitter. Understanding that one line item is most of what separates a plan that closes from a plan that gets rebuilt.
The physics, briefly
A splitter divides power, it does not amplify it. Split light into two equal paths and each path gets half — 3 dB down — plus a fraction of a decibel lost in the device itself. Every doubling costs another 3 dB and change.
| Ratio | Theoretical | Plan for |
|---|---|---|
| 1:2 | 3.0 dB | 3.5 dB |
| 1:4 | 6.0 dB | 7.0 dB |
| 1:8 | 9.0 dB | 10.5 dB |
| 1:16 | 12.0 dB | 14.0 dB |
| 1:32 | 15.0 dB | 17.5 dB |
| 1:64 | 18.0 dB | 21.0 dB |
Read that bottom row again: a single 1:64 splitter spends more of your budget than 50 km of fiber. Distance is cheap. Splitting is expensive.
Unbalanced taps
Splitters do not have to be even. A 30/70 tap sends 70% of the power down the long run and 30% into a nearby cluster — useful along a linear route where one branch has far more distance ahead of it than the other. Plan roughly 5.5 dB on the 30% leg and 1.8 dB on the 70% leg, plus device loss.
Unbalanced taps are the tool of choice for ribbon-development FTTH: villages strung along a road, apartment risers, industrial parks on a single feeder.
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 →
One stage or two?
Serving 32 homes can be one 1:32 at the cabinet, or a 1:4 at the cabinet feeding eight 1:8s in the street. Same subscriber count, different consequences:
- Single stage costs ~17.5 dB and keeps all the splitting in one accessible box. Simpler to test, cheaper to build, but every drop needs enough fiber back to that one cabinet.
- Two stages cost ~17.5 + 0.8 dB in device penalties but push splitting closer to the customer, which usually means dramatically less distribution cable.
The extra decibel almost always buys itself back in trench metres. Just don't let the cascade run away — three stages is usually a sign the architecture wants revisiting.
Rules of thumb worth keeping
- Fix the ratio last. Set the transmitter, the distances and the joints first, then spend what is left on splitting.
- Leave one spare port at every splitter. Growth is not a hypothetical.
- If a drop lands within 3 dB of ONT sensitivity, the design is finished but fragile. Step the ratio down.
- Record the ratio on the node, not in a separate sheet. The as-built is what the next technician will trust.