Module 8
The Optical Loss Budget
This is where the whole course comes together. Every splitter, splice, connector, and kilometer of fiber you have met so far costs light. The loss budget is the single spreadsheet that decides whether your link will actually work — and it is the master constraint behind nearly every design choice an operator makes.
What you'll be able to do
- Tell
dBm(absolute power) apart fromdB(a loss/gain ratio) — and never mix them up again. - Compute an available budget, add up total link loss, and find the remaining margin.
- Name what spends the budget and rank the offenders — especially the splitter.
- Pick the right GPON / XGS-PON budget class for a given split and distance.
- Reason about the split-vs-distance tradeoff: why doubling the split roughly halves your reach.
The loss budget in one picture: an equation, a list of spenders, the named ceilings, and the one tradeoff that drives every design.
The master constraint
Back in Module 1 we packed a water bottle for a hike. You leave with a full bottle — the laser's launch power — and must summit with sips to spare, because the receiver has a sensitivity: a floor below which it can no longer read the signal. Every stretch of trail drinks some. The budget answers one question: did you pack enough, with a little extra?
That bottle has an exact size — the available power budget, the light you can afford to lose between transmitter and receiver:
The budget is a bar from launch power to receiver sensitivity. Each component bites off a chunk in dB; whatever survives before the red floor is your margin. Spend it all and the link dies. (This +3/−28 dBm span is an illustrative 31 dB raw budget; it's not the same as the rated 28 dB GPON Class B+ ceiling below — that's a guaranteed class number, not this particular box's span.)
🔢 The three equations that run the link
Available budget (dB) = transmitter launch power (dBm) − receiver sensitivity (dBm)
Total link loss (dB) = fiber attenuation + splices + connectors + splitter
Margin (dB) = available budget − total link loss → must be > 0, ideally ≥ 3 dB
The whole game is keeping margin positive — the few sips left at the summit. A link that arrives at exactly sensitivity has zero margin: it "works" on the bench, then dies the first cold morning or after the first repair splice. So we reserve a cushion (commonly 3 dB) before we start spending.
dBm vs. dB — the one thing people get wrong
These look alike and mean different things. Get them straight and everything below is easy.
- dBm = decibels relative to 1 milliwatt. It is an absolute power level — an actual amount of light.
0 dBm= 1 mW;+3 dBm≈ 2 mW;−28 dBmis a very faint signal. Launch power and receiver sensitivity are both quoted in dBm. - dB = a relative ratio — how much louder or quieter, never an absolute amount. Losses (fiber, splitter, splices) are in dB. A splitter "costs 17 dB" means it divides the light down by that ratio, no matter how bright you started.
💡 The rule that ties them together
Subtract two dBm values and you get a dB. Launch power (+3 dBm) minus sensitivity (−28 dBm) = 31 dB of budget. You spend that budget in dB, and what is left over is your margin — also in dB. dBm is the bottle's size; dB is how fast you drink.
+3 dB always means 2× — that's why each split level (doubling the outputs) costs about 3 dB, and why +3 dBm is double the power.
What spends the budget
Only four kinds of thing drink from the bottle. Notice how lopsided the breakdown is — one component dwarfs the rest.
The splitter is the glutton. On a typical 12 km, 1:32 GPON B+ link it eats ~17.5 dB — more than fiber, connectors, and splices combined. When budgets get tight, look here first.
| Component | Typical loss | dB |
|---|---|---|
| Fiber @ 1310 nm typical | per kilometre of glass | ~0.35 /km |
| Fiber @ 1550 nm typical | per kilometre (less than 1310) | ~0.22 /km |
| Fusion splice | per joint | ~0.1 |
| Connector (mated pair) | per plug-in point | ~0.3–0.5 |
| Splitter 1:32 — the dominant cost | one passive split | ~17 |
| Splitter 1:64 — the dominant cost | one passive split | ~21 |
🧭 The splitter dominates — by a landslide
A single 1:32 splitter (~17 dB) costs more than 48 km of fiber at 1310 nm, or roughly 170 fusion splices. In a real PON, the splitter is usually the single biggest line item by far — often more than everything else combined. When budgets get tight, the splitter is the first place you look.
Two honest caveats from the source data:
- Fiber loss has two right answers: real-world physics (~0.32–0.35 at 1310, ~0.20–0.25 at 1550) and deliberately pessimistic FOA (Fiber Optic Association) acceptance values (0.5 / 0.4 dB/km) for conservative sign-off. We use the physics numbers and flag them typical.
1550 nmalways attenuates less than1310 nm— scattering falls off as ~1/λ⁴ — which is why longer wavelengths are preferred for long reach.
GPON budget classes
Instead of quoting launch power and sensitivity per box, the standards bundle them into named budget classes — a guaranteed number of dB you may spend between OLT and ONT. (Module 2: OLT = the central-office port; ONT = the box at the home.)
GPON ITU-T G.984 (1490 nm down / 1310 nm up) defines two everyday classes:
- Class B+ = 28 dB — launch +1.5 to +5 dBm, sensitivity down to −28 dBm. The historical workhorse.
- Class C+ = 32 dB — launch +3 to +7 dBm, sensitivity down to −32 dBm. The "extra reach / extra split" upgrade.
🔢 What 4 extra dB buys you
C+ gives you just 4 dB more than B+ — but those 4 dB roughly cover the ~3.5 dB cost of doubling the split (1:32 → 1:64), leaving only a little extra for distance; alternatively you can spend them to push reach from ~15 km to ~20 km. Small number, big consequence: that is the leverage of working at the budget's edge.
XGS-PON ITU-T G.9807.1 10G (1577 nm down / 1270 nm up) uses optical-path-loss classes: N1 = 29 dB, N2 = 31 dB, E1 = 33 dB, E2 = 35 dB. "Nominal" (N1/N2) covers ordinary builds; "Extended" (E1/E2) is for long reach or deep splits. You'll meet N2 (31 dB) and E2 (35 dB) most.
⚠️ The budget is a ceiling, not a target
The class number is the maximum loss the link may have, end to end — not a goal to hit. Your job is to land comfortably under it with margin to spare. And remember: published designs usually quote the downstream direction, but the upstream path is often the tighter constraint in practice — the ONT's cheaper, lower-power laser has to talk back up the very same fiber, splitter, and splices the OLT just spoke down.
The split-vs-distance tradeoff
The single most important idea in fiber access design — and why operators agonize over split ratios.
Within one fixed 28 dB B+ budget: a 1:32 split reaches ~17 km; the extra ~3.5 dB to go 1:64 comes straight out of fiber length, cutting reach to ~7 km. You buy split or distance — not both.
🧭 Distance and split ratio compete for the same dB
Inside a fixed budget, every dB you hand to the splitter is a dB you can no longer spend on fiber length — and vice versa. Doubling the split costs about 3–3.5 dB; each kilometre of fiber costs only ~0.2–0.35 dB. So doubling the split ratio roughly halves how far you can reach. You cannot have maximum split and maximum distance on the same budget — you choose.
Walk the numbers on a GPON Class B+ link to see it bite:
🧪 Worked example — 28 dB B+ budget
Start with the whole bottle and pour out the fixed costs first:
- Connectors & splices: 2 connector pairs + 4 splices ≈ (2×0.5) + (4×0.1) = 1.4 dB
- Safety margin reserved up front: 3 dB
- Remaining for splitter + fiber: 28 − 1.4 − 3 = 23.6 dB
Now the splitter takes its (large) cut, and whatever survives buys fiber length at ~0.35 dB/km:
| Design | Splitter | dB left for fiber | Reach |
|---|---|---|---|
| 1:32 split | ~17.5 dB | 6.1 dB | ≈ 17 km |
| 1:64 split | ~21 dB | 2.6 dB | ≈ 7 km |
Going 1:32 → 1:64 doubles the homes served per OLT port but cuts reach from ~17 km to ~7 km — the extra ~3.5 dB came straight out of the fiber-length allowance. To reach 25 km on B+ you'd have to shrink the split (say, to 1:16). Want both a deep split and long reach? That is exactly when you buy a C+ (32 dB) OLT. (Illustrative; real designs also account for the tighter upstream direction.)
🔁 The budget isn't the only ceiling
For a typical FTTH PON, the loss budget decides reach — these links are loss-limited. But power isn't the only limit: at very high bit-rates or long-haul distances, dispersion (pulses spreading until adjacent bits blur together) caps reach too. See Module 1 → The other reach limit: dispersion.
Put it together: the loss budget calculator
Now play with the master constraint yourself. Drag the sliders and watch the bar fill toward the budget line. Every category stacks up; overflow the budget and the link fails. Defaults pass comfortably — push the split to 1:64, or distance past the line, and watch where it breaks.
🥤 Reading the bar like a bottle
The black line is your bottle's size (the budget class). The coloured blocks are sips you've already committed; the grey block is the cushion you promised to keep. As long as everything fits left of the line, you summit with sips to spare. The moment the bar spills past the line — red — you've run dry before the top.
Go deeper: where the numbers come from optional
The calculator uses 0.1 dB per fusion splice and 0.5 dB per connector pair (the conservative end of the 0.3–0.5 dB range), and the typical splitter insertion losses including excess loss: 1:2 ≈ 3.6, 1:4 ≈ 7.4, 1:8 ≈ 10.75, 1:16 ≈ 14.0, 1:32 ≈ 17.5, 1:64 ≈ 21.0, 1:128 ≈ 24.5 dB. Splitter loss grows ~3–3.5 dB per doubling — close to the ideal 10·log₁₀(N), plus excess loss that climbs with the ratio.
These are working ballpark figures, not a substitute for vendor datasheets and a real OTDR/OLTS sign-off — the subject of the next module. Real designs also reserve margin for the tighter upstream direction and for the connectors built into the splitter's own ports.
Key takeaways
- Budget = launch power − sensitivity (dBm − dBm = dB). You spend it on loss; what's left is margin, which must stay positive — ideally ≥ 3 dB.
- dBm is absolute power; dB is a ratio. Subtract two dBm values to get a dB.
- The splitter is the glutton — 1:32 ≈ 17 dB, 1:64 ≈ 21 dB — usually the biggest single line item by far.
- Budget classes are ceilings: GPON B+ = 28 dB, C+ = 32 dB; XGS-PON N2 = 31 dB, E2 = 35 dB. C+'s extra 4 dB buys 1:32→1:64 or ~15→20 km.
- Split and distance compete for the same dB. Doubling the split (~3.5 dB) roughly halves your reach; you choose one extreme, not both — unless you buy more budget.
- A link that passes at exactly sensitivity has no margin and will fail in the field. Always keep a cushion.
An OLT launches at +4 dBm and the ONT's sensitivity is −28 dBm. What is the available power budget?
Why: Budget = launch − sensitivity = (+4) − (−28) = 32 dB. Subtracting two dBm values yields a dB (a ratio), not a dBm. The result happens to match a GPON C+ class.
In a typical PON, which component usually spends the most of the loss budget?
Why: A single 1:32 splitter (~17 dB) outweighs dozens of kilometres of fiber and hundreds of splices. The splitter is the dominant — usually the largest — single line item.
You have a GPON B+ (28 dB) link running 1:32 to ~17 km that just closes with 3 dB margin. Your boss wants to double the homes per port by going 1:64. With the same OLT, what most likely happens?
Why: Going 1:32→1:64 adds ~3.5 dB of splitter loss, taken straight out of the fiber-length allowance. Within a fixed budget, distance and split trade off — doubling the split roughly halves the reach. To keep both you'd need a bigger class, e.g. C+.
A link computes to exactly 0 dB of margin remaining on the bench. Why is that not good enough to deploy?
Why: Margin must be > 0 and ideally ≥ 3 dB precisely so the link survives aging, cold weather, and the repair splices that real fiber accumulates. Zero margin works once on the bench and dies in the field.