Fiber Optical Power Budget Calculation: A Worked PON Example

An optical power budget is the difference between the lowest output power a transmitter guarantees and the lowest input power a receiver can handle without errors. To calculate it, add up every passive loss on the path (fiber, connectors, splices, splitters) and compare the total with the range allowed by the optical class. Keep a margin of typically at least 3 dB for ageing and future repairs.

The sections below give the formulas, typical loss values and a step-by-step GPON example.

Terms and formulas

• Power budget: transmitter minimum output power minus receiver sensitivity. It is the maximum loss the link can tolerate.
• Link loss: the combined loss of every passive element between transmitter and receiver.
• Margin: what is left of the budget after subtracting link loss.
• Overload: the highest input power a receiver accepts without distortion. It becomes a problem on very short, lightly split links.

Power budget (dB)    = Ptx,min (dBm) − Prx,sensitivity (dBm)
Link loss (dB)       = (L × α) + (Nconnectors × Lconnector) + (Nsplices × Lsplice) + ΣLsplitter + Lother
Margin (dB)          = Power budget − Link loss
Received power (dBm) = Ptx − Link loss

Conditions: Margin ≥ 3 dB (typical target)  and  Ptx,max − Link loss ≤ Prx,overload

L is fiber length in km, α is attenuation in dB/km, and each per-element loss is in dB.

For PON systems, vendors usually state the optical class directly (for example GPON B+ = 13–28 dB). In that case the calculated link loss must fall between the class minimum and maximum, with enough margin below the maximum.

Loss contributors and typical values

Fiber attenuation

Single-mode fiber attenuation depends on wavelength. It rises at shorter wavelengths because of Rayleigh scattering. The figures below are typical for G.652.D fiber. On a real project, always design with the cabled maximum from your cable's datasheet.

Connectors, splices and other passives

When counting connectors, include every mated interface between the OLT port and the ONU: ODF adapters, splitter inputs and outputs, the building termination box and the wall outlet.

Splitter loss

An ideal splitter's loss is 10 × log10(N). Real products add excess loss and port-to-port uniformity, so design with the datasheet maximum instead of the theoretical figure.

In cascaded designs (for example 1:4 + 1:8 = 1:32), splitter losses add up and extra connectors appear, so the total can be slightly higher than with a single-stage 1:32. Unbalanced FBT splitters used in linear (bus) topologies, such as 90/10 or 70/30, need each leg calculated separately. Before excess loss, the 10% leg has a theoretical loss of 10 dB and the 90% leg about 0.46 dB.

PON optical classes

Typical GPON B+ module figures:

• OLT transmit power +1.5 to +5 dBm, ONU receiver sensitivity −27 dBm
• ONU transmit power +0.5 to +5 dBm, OLT receiver sensitivity −28 dBm

C+ OLT modules usually launch +3 to +7 dBm, and C+ ONU sensitivity is around −30 dBm. Check your module's datasheet for exact values.

If link loss falls below the class minimum, received power can exceed the overload level (typically −8 dBm for GPON ONUs). The fix is an inline optical attenuator of suitable value.

Worked example: 12 km GPON link with a cascaded 1:32 split

Scenario: the OLT is in the central office. A 10 km feeder runs to a street cabinet with a 1:4 splitter. From there, 2 km of distribution and drop cable reaches a 1:8 splitter at the building entrance, followed by the subscriber's ONU. Total fiber length is 12 km, and the optics are GPON B+.

Upstream (1310 nm)   : 4.2 + 7.2 + 10.3 + 2.0 + 0.8 = 24.5 dB
Downstream (1490 nm) : 3.0 + 7.2 + 10.3 + 2.0 + 0.8 = 23.3 dB

B+ maximum           : 28 dB  → Margin = 28 − 24.5 = 3.5 dB   (OK)
B+ minimum           : 13 dB  → 23.3 dB > 13 dB                (no overload risk)

Power at the ONU     : +1.5 − 23.3 = −21.8 dBm (worst) … +5 − 23.3 = −18.3 dBm (best)
ONU sensitivity      : −27 dBm → 5.2 dB headroom in the worst case (OK)

Upstream is the limiting direction here because fiber attenuation is higher at 1310 nm. A 3.5 dB margin meets the usual target.

What if the split were 1:64?

Replacing the 1:8 building splitter with a 1:16 (13.5 dB) raises upstream loss to 27.7 dB. With B+ optics that leaves only 0.3 dB of margin, which is not acceptable. The options are C+ optics (32 − 27.7 = 4.3 dB margin), a shorter fiber route, or redistributing the split onto shorter links.

What if the same link moves to XGS-PON?

XGS-PON transmits upstream at 1270 nm, so fiber loss becomes 12 × 0.40 = 4.8 dB. Assume a coexistence element with a typical loss of 1.2 dB is added at the OLT so that GPON can stay on the same fiber:

Upstream (1270 nm)   : 4.8 + 7.2 + 10.3 + 2.0 + 0.8 + 1.2 = 26.3 dB
N1 (29 dB)           : Margin = 2.7 dB   (below target)
N2 (31 dB)           : Margin = 4.7 dB   (OK)

The CEx also raises GPON loss to 24.5 + 1.2 = 25.7 dB, which leaves 2.3 dB with B+. On routes like this, consider Combo PON modules. Their built-in WDM loss is usually included in the module's stated power figures, but confirm this on the datasheet.

How much margin is enough?

The common design target is at least 3 dB. The margin absorbs:

• Connector loss creeping up from repeated mating and contamination
• Extra splices added when a cut cable is repaired
• Temperature variation and gradual laser power decline
• Measurement uncertainty and datasheet tolerances

Leave more margin on aerial routes that get damaged often, or where future expansion is planned.

Verifying the calculation in the field

• Light source and power meter (OLTS): measure the total loss of the link at 1310 and 1550 nm, and at PON wavelengths where possible.
• PON power meter: reads downstream (1490/1577 nm) and upstream (1310/1270 nm) power separately on a live link. It is the quickest way to see how much power actually reaches an ONU.
• OTDR: locates splices and connectors and measures their loss. Judge splice loss by averaging bidirectional traces. To see past a splitter you need a high-dynamic-range PON OTDR, and live networks call for filtered testing at 1625/1650 nm.
• Connector inspection: inspect and clean end faces before testing. A single dirty connector can use up the entire margin.

If measured loss is clearly worse than calculated, the usual causes are dirty connectors, macrobends and poor splices. Macrobends show up far more at 1550 nm than at 1310 nm, so testing at both wavelengths speeds up diagnosis.

WiMap currently covers duct, cable, splice and core matching with BOQ/BOM output. GPON OLT → splitter → ONU design with automatic optical budget calculation is coming soon. Splitters, SFP modules and fiber test equipment are available through wi.com.tr.

Frequently asked questions

Does splitter loss change with wavelength?

PLC splitters are nearly flat across 1260–1650 nm, so GPON, XGS-PON and RF video can share the same splitter. FBT splitters depend more on wavelength. Check their datasheet before adding new wavelengths.

What should I do if calculated loss is below the class minimum?

On short, lightly split links, received power can exceed the overload level and cause bit errors. Add an optical attenuator that makes up the difference. For example, a link with 9 dB of loss needs at least a 4 dB attenuator to reach the B+ minimum of 13 dB.

Why is at least 3 dB of margin recommended?

Link loss does not stay at its day-one value. Repair splices, dirty connectors, temperature swings and laser ageing all push it up over time. 3 dB is a widely used engineering rule that covers most of these effects. Harsher environments justify more.

Can GPON B+ optics support a 1:64 split?

On short routes, yes. A 1:64 splitter alone costs about 20 dB, though, which leaves little budget for fiber, connectors and splices. Beyond a few kilometres, C+ optics are usually required. Decide link by link, based on the calculation.

Why does OTDR-measured loss differ from the calculation?

Calculations use conservative design values, so measurements often come out slightly better. If they are worse, look for connector contamination, macrobends or bad splices. A one-direction OTDR trace can misstate splice loss, and a bidirectional average is more reliable.