An optical power budget calculator does one thing: it adds up every decibel of loss between the OLT PON port and the ONU, subtracts a margin for ageing and repairs, and checks what is left against the receiver sensitivity — for a Class B+ link, that means total loss inside roughly 28 dB and a received level between about -8 dBm and -27 dBm.
Rayin (Shenzhen Rayin Technology) is a manufacturer of GPON OLTs and industrial Ethernet switches.
The worksheet is purely additive. Start from the OLT transmit power, subtract splitter insertion loss, splice loss, connector loss and fiber attenuation, subtract a 3 dB design margin, and compare the remainder with the ONU receiver sensitivity.
Industry-typical values when no datasheet is at hand: 1:8 ≈ 10.5 dB, 1:16 ≈ 13.5 dB, 1:32 ≈ 17 dB, 1:64 ≈ 20.5 dB; 0.1 dB per fusion splice; 0.3 dB per mated connector pair; 0.35 dB per km at 1310 nm on ITU-T G.652 fiber.
Run two checks, not one. The absolute check asks whether received power sits between the overload point and the sensitivity. The budget check asks whether loss plus margin fits the 28 dB Class B+ allowance. A design routinely passes the first and fails the second.
Every budget calculation rests on the same definition: a loss budget is the difference between what a transmitter launches and what a receiver can still decode. That difference is what engineers call the GPON power budget, and it is fixed by the optical class, not by the route. ITU-T G.984.2 defines optical classes, and Class B+ is the one most FTTH builds are designed against: roughly 28 dB of allowed loss between the OLT and the ONU, with a Class B+ ONU receiver sensitivity near -27 dBm at 1490 nm and an overload point near -8 dBm.
Every ONU on a tree sees the same GPON splitter loss, because the splitter is passive and has no idea which port is listening. Two directions share that loss, but they do not share the same fiber attenuation. Downstream runs at 1490 nm and upstream at 1310 nm, and G.652 fiber attenuates more at the shorter wavelength — about 0.35 dB per km at 1310 nm against roughly 0.25 dB per km near 1490 nm. Using the 1310 nm figure for both directions keeps the worksheet conservative and lets one calculation cover the whole span.
Note: Every number in this article is an industry-typical value for planning, not a measured parameter of any specific product. Substitute the figures from your own OLT module datasheet, splitter specification and cable certificate as soon as you have them — a real splitter with 0.4 dB of extra loss will quietly consume 13% of a 3 dB margin.
| Item | Industry-typical value | Where it comes from |
|---|---|---|
| OLT PON port transmit power | +2 to +5 dBm | Optical module class, G.984.2; use the datasheet minimum for worst case |
| Splitter 1:8 | ≈ 10.5 dB | PLC splitter specification |
| Splitter 1:16 | ≈ 13.5 dB | PLC splitter specification |
| Splitter 1:32 | ≈ 17 dB | PLC splitter specification |
| Splitter 1:64 | ≈ 20.5 dB | PLC splitter specification |
| Fusion splice | 0.1 dB each (0.02–0.1 dB measured) | ITU-T G.652 field practice |
| Mechanical splice | 0.3–0.5 dB each | Avoid in new builds |
| Mated connector pair | 0.3 dB each | Plus ageing and contamination |
| Fiber attenuation | 0.35 dB per km at 1310 nm | G.652 typical; 0.4 dB per km is the usual specification limit |
| Design margin | 3 dB | Ageing, future repairs, extra splices, temperature |
| ONU sensitivity, Class B+ | ≈ -27 dBm | G.984.2 |
| ONU overload, Class B+ | ≈ -8 dBm | G.984.2 |
The margin row is the one engineers argue about, and 3 dB is the conventional compromise: enough to absorb a repair splice, a replacement patch cord and a decade of connector wear without being so generous that it forces a lower split ratio on paper.
Work in one direction at a time and write every line down.
Fix the transmit power. Use the minimum from the module datasheet, not the typical value. A +5 dBm typical module that guarantees +2 dBm must be planned at +2 dBm.
Add the splitter loss. One number for a single-stage splitter; two numbers plus the fiber between them for a two-stage design.
Count the splices. Feeder cable usually arrives in 2 km to 4 km drum lengths, so a 12 km feeder carries three to six splices before any distribution splices.
Count every mated connector pair. Patch panels at the OLT, at the splitter and at the ONU, plus any field-connectorised drop.
Multiply the route length by the attenuation coefficient. Include the drop cable, which is easy to forget and commonly 0.3 km to 0.8 km on its own.
Subtract the 3 dB margin.
Compare with the sensitivity and with the class budget.
The last step is two separate questions. The received level must be above roughly -27 dBm, and the total loss plus the margin must fit inside roughly 28 dB. A link can satisfy the first and still be a bad design.
A 32-subscriber cluster five kilometres from the OLT.
| Line item | Calculation | Loss (dB) |
|---|---|---|
| OLT transmit power | datasheet | +3.0 dBm |
| Splitter 1:32 | single stage | 17.0 |
| Fiber, 5 km at 0.35 dB per km | 5 × 0.35 | 1.75 |
| Splices, 6 at 0.1 dB | 6 × 0.1 | 0.6 |
| Connectors, 4 at 0.3 dB | 4 × 0.3 | 1.2 |
| Total span loss | 20.55 | |
| Received power at the ONU | 3.0 − 20.55 | −17.55 dBm |
| Headroom to −27 dBm sensitivity | 9.45 dB | |
| Budget check, loss plus 3 dB margin | 20.55 + 3.0 | 23.55 dB (inside 28 dB) |
This design has room to spare. Nine and a half decibels of headroom means the link survives two extra repair splices, a dirty connector and a decade of ageing before anyone has to think about it again.
The same topology pushed to 64 subscribers and fifteen kilometres.
| Line item | Calculation | Loss (dB) |
|---|---|---|
| OLT transmit power | datasheet minimum | +2.0 dBm |
| Splitter 1:64 | single stage | 20.5 |
| Fiber, 15 km at 0.35 dB per km | 15 × 0.35 | 5.25 |
| Splices, 10 at 0.1 dB | 10 × 0.1 | 1.0 |
| Connectors, 6 at 0.3 dB | 6 × 0.3 | 1.8 |
| Total span loss | 28.55 | |
| Received power at the ONU | 2.0 − 28.55 | −26.55 dBm |
| Headroom to −27 dBm sensitivity | 0.45 dB | |
| Budget check, loss plus 3 dB margin | 28.55 + 3.0 | 31.55 dB (outside 28 dB) |
This is the classic near-miss. The absolute check passes by less than half a decibel, which means the ONU will register and the customer will get service — on a clean day, with clean connectors, at 20 °C. It fails the budget check by 3.55 dB, which means the GPON power budget is already spent before a single repair happens. The practical consequences are predictable: every repair splice is an outage, one contaminated patch cord takes the subscriber off the network, and the installation team spends its first year chasing intermittent drops.
Before committing to any of them, confirm which optical classes your Rayin PON module supports, because the class sets the allowance the whole worksheet is measured against.
The fixes, in order of preference: split the far cluster onto its own 1:32 tree and add a PON port, shorten the route, replace the two mechanical splices hiding in that count, or move to Class C+ optics where the 32 dB allowance buys back the deficit.
A 1:4 primary splitter feeding eight 1:16 secondaries — an effective 1:64 ratio spread over a wider area.
| Line item | Calculation | Loss (dB) |
|---|---|---|
| OLT transmit power | datasheet | +3.0 dBm |
| Primary splitter 1:4 | stage one | 7.2 |
| Secondary splitter 1:16 | stage two | 13.5 |
| Splitter subtotal | 7.2 + 13.5 | 20.7 (0.2 dB worse than one 1:64) |
| Fiber, 8.5 km total at 0.35 dB per km | 8.5 × 0.35 | 3.0 |
| Splices, 9 at 0.1 dB | 9 × 0.1 | 0.9 |
| Connectors, 8 at 0.3 dB | 8 × 0.3 | 2.4 |
| Total span loss | 27.0 | |
| Received power at the ONU | 3.0 − 27.0 | −24.0 dBm |
| Headroom to −27 dBm sensitivity | 3.0 dB | |
| Budget check, loss plus 3 dB margin | 27.0 + 3.0 | 30.0 dB (outside 28 dB) |
Two-stage splitting costs a little more than single-stage for the same ratio — here 0.2 dB in the splitters plus two extra mated connector pairs at 0.3 dB each, so about 0.8 dB in total — and it buys flexibility in return: one feeder cable serves a wide area and capacity is rebalanced by moving drop cables between secondaries.
Note the verdict. The absolute check passes with exactly 3 dB of headroom, which is precisely the margin, so the design has nothing left for a repair. Changing the secondaries from 1:16 to 1:8 drops the split subtotal to 17.7 dB, brings total loss to 24.0 dB, and puts the budget check at 27.0 dB — inside the 28 dB allowance, with a working 1:32 effective ratio and 32 subscribers per tree.
Usually a short one, because the numbers are close but not identical.
The splitter loss is identical in both directions — a passive splitter is reciprocal — so the difference is entirely in the fiber and in the endpoints. Upstream runs at 1310 nm, where G.652 attenuates about 0.35 dB per km, against roughly 0.25 dB per km downstream at 1490 nm. On a 15 km route that is a 1.5 dB difference before anything else. At the endpoints, a Class B+ ONU launches roughly +0.5 dBm at minimum and the OLT receiver sensitivity at 1310 nm is about -28 dBm, which gives a similar allowance to the downstream direction. The practical rule: if the downstream calculation passes with less than 2 dB of margin, run the upstream numbers explicitly rather than assuming symmetry.
A calculation that has never been compared with a measurement is a hypothesis.
Measure at the ONU end with a calibrated power meter set to 1490 nm, with the OLT transmitting — downstream light is always present, so the reading is stable and does not depend on grants. Compare three numbers: the computed value, the measured value, and the received power the ONU reports back through its diagnostics. If the measured figure is more than about 1 dB worse than the calculation, the worksheet is missing something, and the usual suspects are a contaminated end face, which costs 0.5 dB to 1 dB on its own, a bend radius violation on a patch cord, a mechanical splice nobody recorded, or a splitter whose label does not match the cassette inside.
Rayin specifies the optical parameters of the PON modules supplied with its OLTs, and those datasheet figures should replace the typical values in this worksheet the moment you have them. When you evaluate a unit for a site, ask for transmit power, receiver sensitivity and the supported optical class in writing, because the whole calculation is only as good as its first line.
What is a safe GPON power budget to design against?
For Class B+ optics the reference allowance is roughly 28 dB between the OLT and the ONU, with an ONU sensitivity near -27 dBm and an overload point near -8 dBm. Reserve 3 dB of that for ageing and repairs, so plan the span loss at 25 dB or less.
How much loss does a GPON splitter add?
A planar lightwave circuit splitter adds about 10.5 dB at 1:8, 13.5 dB at 1:16, 17 dB at 1:32 and 20.5 dB at 1:64. Those are industry-typical values; the cassette in front of you may differ by a few tenths of a decibel either way.
Why does my measured power disagree with my calculation?
Most gaps come from uncounted loss: a dirty end face costs 0.5 dB to 1 dB, a mechanical splice up to 0.5 dB, and an unrecorded patch panel adds 0.3 dB. Recount the connectors and splices, verify the splitter ratio, and re-measure with a clean adapter.
Can received power be too high on a PON?
Yes. A Class B+ ONU overloads at about -8 dBm, so a subscriber a few hundred metres from the OLT on a short feeder can receive too much light after a 1:2 split. Short-reach designs sometimes need an attenuator to bring the level back into the window.
Should I calculate the upstream direction separately?
Worth doing. Both directions share the same splitter loss, but 1310 nm upstream attenuates at about 0.35 dB per km against roughly 0.25 dB per km downstream at 1490 nm, so the upstream path is usually the one that runs out first on long routes.
An optical power budget calculator is simple arithmetic applied honestly: launch power, minus splitters, splices, connectors and fiber, minus a margin, compared with sensitivity and with the class allowance. Run both checks, use datasheet minimums rather than typical values, and the three worked examples above cover most of what a real FTTH design will throw at you.
About the author: Sara — Sara is a Customer Manager at Rayin with over 10 years of experience in the communications field. In her free time, she enjoys badminton and swimming.
About Rayin: Shenzhen Rayin Technology Co., Ltd. — Company Profile
