GPON to XGS-PON on the same fiber: check the loss budget before you swap the optics

The extra decibel in class N1 is already spoken for. A five-step, standards-based check for operators reusing a legacy ODN.

Many operators planning an XGS-PON upgrade intend to keep the optical distribution network they built for GPON: the same feeder, the same splitters, the same drops, with new OLT line cards and ONUs and a coexistence element in the central office. ITU-T G.9807.1 was written to make that possible. It sets 1:64 split support as a minimum requirement precisely so that XGS-PON can coexist with the Gigabit PONs operators already run.

But “same fiber” is not the same as “same loss budget.” The upstream moves to a wavelength where fiber loss is higher, a coexistence element is inserted in the path, and the downstream moves to 1577 nm, where tight bends in old drops start to show. On a network that was designed close to the GPON class B+ limit, those small changes add up.

The five checks below use only public standards and your own as-built records. The numbers in the worked example are illustrative planning values, not measurements of any specific network.

Step 1: Compare the right classes and wavelengths

Start by putting the old and new optical path loss (OPL) classes side by side (Table 1).

Table 1. GPON and XGS-PON loss classes and wavelengths

System	Class	Min. loss	Max. loss	Downstream	Upstream	Source
GPON	B+	13 dB	28 dB	1480–1500 nm	1260–1360 nm (regular band)	ITU-T G.984.2, G.984.5
GPON	C+	17 dB	32 dB	1480–1500 nm	as above	ITU-T G.984.2
XGS-PON (basic wavelength set)	N1	14 dB	29 dB	1575–1580 nm (1581 nm outdoor)	1260–1280 nm	ITU-T G.9807.1
XGS-PON	N2	16 dB	31 dB	as above	as above	ITU-T G.9807.1
XGS-PON	E1	18 dB	33 dB	as above	as above	ITU-T G.9807.1
XGS-PON	E2	20 dB	35 dB	as above	as above	ITU-T G.9807.1

At first glance, N1 looks like it gives you 1 dB more than B+. It does not. G.9807.1 explains that the basic wavelength set “aligns with the N1 class of XG-PON for 29 dB, considering the extra loss in the WDM1r for this band.” In other words, the standard has already allocated that decibel to the coexistence filter. Treat it as spent, not as margin.

Two other details in the table matter:

• Minimum loss rises from 13 dB to 14 dB. Short branches close to the OLT that were fine on GPON can fall below the N1 minimum. G.9807.1 says the ODN must then include attenuators to protect the receivers.
• The upstream bands overlap. ITU-T G.984.5 defines a regular GPON upstream band of 1260–1360 nm, a reduced band of 1290–1330 nm and a narrow band of 1300–1320 nm. XGS-PON’s 1260–1280 nm upstream overlaps the regular band. Confirm which option your installed GPON ONUs use before you plan coexistence on the same tree.

Step 2: Re-run the budget at the new wavelengths

Most GPON budgets were calculated at 1310 nm for the upstream, which is usually the harder direction. XGS-PON’s upstream sits at 1270 nm. ITU-T G.652 (2024) gives a simple rule: the G.652.D cable attenuation specification can be extended down to 1260 nm by adding 0.07 dB/km of Rayleigh scattering loss to the 1310 nm value.

Table 2 applies that rule to a hypothetical link: 20 km of G.652.D, one 1:32 PLC splitter, four mated connections and ten fusion splices.

Table 2. Worked example: the same ODN, two upstream budgets (illustrative)

Item	GPON upstream, 1310 nm	XGS-PON upstream, 1270 nm
Fiber, 20 km	7.0 dB (0.35 dB/km)	8.4 dB (0.35 + 0.07 = 0.42 dB/km)
1:32 PLC splitter (catalogue maximum, connectors included)	17.1 dB	17.1 dB (confirm at 1270 nm, see Step 4)
4 mated connections at 0.3 dB	1.2 dB	1.2 dB
10 fusion splices at 0.1 dB	1.0 dB	1.0 dB
Subtotal	26.3 dB	27.7 dB
Class maximum	28 dB (B+)	29 dB (N1)
Room left before the coexistence element	1.7 dB	1.3 dB

Planning values: 0.35 dB/km, 0.3 dB per connection and 0.1 dB per splice are common typical figures; replace them with your cable datasheets and as-built records. The splitter value is a published catalogue maximum. The 0.07 dB/km adjustment follows ITU-T G.652, Table 2, Note 2.

Run the same sum on your own lengths with the fiber loss calculator. The GPON half of this example is also in PLC splitter ratios and the GPON loss budget.

Now subtract the insertion loss of the coexistence element (CEx or WDM1r) port each signal passes through, taken from its datasheet. Two things follow:

• The XGS-PON path starts with less room than the GPON path had, even before the filter is counted.
• Inserting the coexistence element also costs the existing GPON service some margin, because GPON now passes through the filter too.

Whatever is left after the filter is your allowance for repair splices, connector wear and ageing. A common planning practice is to hold about 3 dB. On this example link, neither service would keep that much.

The worst case is sobering. With the TIA-568 maximums of 0.75 dB per mated connection and 0.3 dB per splice, and the G.652.D cable limit of 0.47 dB/km at 1270 nm, the same link adds up to 9.4 + 17.1 + 3.0 + 3.0 = 32.5 dB. That is beyond N2 before the filter is even counted. Real links rarely hit every maximum at once, which is exactly why the upgrade decision should rest on measured loss rather than design assumptions (Step 5).

Step 3: Look for bends that 1490 nm forgave

On straight fiber, the downstream is not the problem. G.9807.1 notes that fiber attenuation at 1577 nm is lower than at 1270 nm, which leaves a downstream margin. It expects that margin to be at least 1 dB on fiber runs longer than 20 km.

Bends are different. Macrobend loss climbs steeply with wavelength. ITU-T G.657 warns that G.652 fibers bent to a 15 mm radius “generally can have macrobending losses of several dB per 10 turns at 1625 nm.” For comparison, G.657 sets these limits for ten turns at a 15 mm radius:

• G.657.A1: 0.25 dB at 1550 nm and 1.0 dB at 1625 nm.
• G.657.A2: 0.03 dB at 1550 nm and 0.1 dB at 1625 nm.

A GPON downstream at 1490 nm may have quietly tolerated an over-tight G.652.D drop, a crowded splice tray or a stapled indoor run. A 1577 nm downstream, or an OTDR monitoring wavelength at 1625 or 1650 nm, will not be as forgiving.

The practical test is to shoot OTDR traces at 1550 nm and 1625 nm and compare events. A bend shows up as a loss event that is noticeably larger at the longer wavelength, while a splice or connector looks about the same at both. Where you find bends, re-dress the fiber, or replace drops and indoor runs with G.657.A2 or B3 fiber. G.657.A fibers meet the G.652.D specification, so they splice into the existing feeder normally. Where each grade belongs is set out in G.652.D vs G.657. Optical fiber and drop and indoor cable are specified by that grade.

Step 4: Check that every passive works across the full band

GPON-era passives were bought to work at 1310 and 1490 nm, and sometimes at 1550 nm for RF video. Check each class of device against the new wavelengths:

• PLC splitters are broadband devices, usually specified from 1260 to 1650 nm. However, catalogue loss tables often list only 1310, 1490 and 1550 nm. Ask for insertion loss and uniformity at 1270 and 1577 nm, or across the full band, for the exact part numbers in your plant.
• Fused biconic (FBT) couplers are more wavelength-dependent than PLC splitters. Where they appear in older or unbalanced-tap designs, check their datasheet window before assuming they pass 1270 and 1577 nm.
• RF video overlay filters need a second look. G.984.5 places video in the 1550–1560 nm enhancement band, close to the XGS-PON downstream. Confirm that the overlay filters and WDM components pass 1575–1581 nm, and that their isolation suits the new plan.
• Older ONUs need checking too. G.984.5 specifies wavelength blocking filters and interference tolerance so that GPON ONUs can coexist with next-generation signals, but ONUs deployed before those requirements were widespread may not have them.

Catalogue maxima for common PLC ratios, including the 17.1 dB used above for a 1:32 with connectors, are listed with PLC splitters. Compare ratios in the splitter loss calculator.

Step 5: Measure, then write it down

Every number in Table 2 is an assumption until it is measured. IEC 61280-4-2 (2024) covers attenuation and optical return loss measurement of installed single-mode cabling, including the passive devices in the path. It uses two methods: a light source and power meter (LSPM) and an OTDR.

• Use both methods, and test in both directions where you can. LSPM measurements per branch give the end-to-end loss that the budget must cover. OTDR traces show where the loss is.
• Inspect and clean every endface before measuring. IEC 61300-3-35 defines the visual inspection criteria. A contaminated connector both inflates the measurement and wastes a margin you cannot spare.
• Record the results per branch in your network inventory, not in a closed-out test report. The next upgrade will need the same data.

Power meters, light sources and OTDRs for those two methods are listed with splicing and test tools.

The decision

With measured, per-branch loss in hand, the choice is straightforward. Here, “room left” means the class maximum minus the measured loss, including the coexistence element.

Measured result on the branch	Action
Room left is at least your repair allowance (for example 3 dB)	Proceed with N1 optics.
Room left is positive but below that allowance	Remediate first: re-terminate the worst connectors, re-make high-loss splices, fix bends. Or move to N2 or E1 optics.
Measured loss is over the class maximum	Re-architect: move the split, use a lower split ratio on long branches, or use an extended-class optics pair.

None of this is exotic. It is the same arithmetic that went into the original GPON design, repeated at the wavelengths the new system actually uses, with the coexistence filter counted honestly. Done before the optics are ordered, it separates the branches that can upgrade on day one from the ones that need a truck roll first.

When the measured loss says which branches can take the new optics, name the class, the split and the fiber grade on the bill of materials. Send the BOM and we match every line and return one priced BOM, usually within 3 to 7 working days. GPON and XGS-PON optics are listed with optical transceivers.

References

• ITU-T G.9807.1 (02/2023), 10-Gigabit-capable symmetric passive optical network (XGS-PON), Annex A (clauses A.8.4 and A.8.5) and Annex B (clause B.6). itu.int
• ITU-T G.984.2 (08/2019), G-PON physical media dependent layer. itu.int
• ITU-T G.984.5 (02/2022), G-PON enhancement band. itu.int
• ITU-T G.652 (08/2024), single-mode optical fibre and cable, Table 2 and Notes. itu.int
• ITU-T G.657 (08/2024), bending-loss insensitive single-mode optical fibre and cable, Table 1. itu.int
• IEC 61280-4-2:2024, installed cabling plant, single-mode attenuation and optical return loss measurement. webstore.iec.ch
• IEC 61300-3-35:2022, visual inspection of fibre optic connectors and fibre-stub transceivers. webstore.iec.ch
• TIA-568 maximum connector and splice loss values, as summarized by The Fiber Optic Association, “Calculating Fiber Optic Loss Budgets.” thefoa.org

Next guide Fiber optic cable prices in 2026: what fiber, cable and FTTH parts cost from China