GPON and XGS-PON can share one optical distribution network through one of three methods defined in ITU-T G.9805: an external coexistence element, a multi-PON module in the optical line terminal (OLT) port, or an M:N splitter architecture with adequate filtering. The choice turns on three things - where the wavelength isolation sits, where the extra insertion loss lands, and whether every subscriber path passes both systems. Here, passive migration means retaining the existing fiber and splitter tree; the MPM option may still require an OLT line-card change.
Quick path choice. Start with the equipment and first split stage:
- Supported combo PON line card or multi-PON module available: combine wavelengths inside the OLT port and avoid a separate head-end element.
- Existing OLT ports must stay: assess an external coexistence element (CEx), including its insertion loss, patching and isolation.
- First split stage has a spare input: assess the 2×N splitter method only when the receiving ports have documented filtering.
Whichever path is considered, confirm the deployed GPON ONU upstream band and check both systems' loss limits in each direction on every subscriber path.
On an installed GPON ODN, the existing fiber and splitter tree are the assets to retain. The FTTH ODN network solution shows their place in the wider network. The migration design then specifies the hardware and interfaces needed to share that plant.
A coexistence element is a wavelength-selective device placed between two PON systems' OLT ports and the shared ODN, while a multi-PON module is an OLT line module with that same wavelength selection integrated inside the port. The practical difference is that the first adds an outside device, two patched interfaces and their insertion loss to the bill of materials, and the second does not.
When can GPON and XGS-PON share one ODN?
Three conditions have to hold together, and they are not all properties of a single subscriber path. First, the two systems' wavelength plans must be separable at the points where their receivers sit. Second, every receiving port must tolerate the leakage that arrives from the other system - and because that leakage depends on the loss difference between a target path and an interfering path, this condition is set by the worst pairing of paths in the tree, not by any one path on its own. Third, every subscriber path must stay inside both systems' optical limits, in both directions, with the coexistence hardware included; that third condition is the one that is genuinely per path.
ITU-T G.9805 defines exactly three architectures that satisfy the first condition: an external coexistence element (CEx), an OLT multi-PON module (MPM) with integrated WDM, and a splitter-based method using an M:N splitter (ITU-T G.9805 (02/2022), Clause 6). The recommendation states that these allow deployed fiber and splitters to be reused when a legacy PON is evolved to higher capacity, and it also gives the method for calculating the isolation a coexistence element needs. It does not select one method for the operator. Before a clause or appendix number from it is written into a specification, check that number against the current amendments, including Amendment 2 of November 2025.
The second condition needs explicit evidence. Isolation is not a fixed property of a part class; G.9805 Appendix II derives the required isolation from the crosstalk the receiving port can tolerate, the launch-power difference between ONUs, the differential ODN loss, the receive-path isolation, and a bit-rate bandwidth compensation term. The differential ODN loss term is where the pairing matters: what has to be evaluated is the combination of a target ONU's path loss and an interfering ONU's path loss into the same receiver, so a check that walks one path at a time does not cover it. Two datasheets quoting the same isolation figure therefore do not describe the same design unless the assumptions behind each figure - including the pairing they were evaluated on - are stated.
The third condition is arithmetic, and a paper approval can fail after installation if its path-loss inputs are wrong: a shared route has to satisfy the more demanding system on that route, in that direction, for that subscriber.
Compare the three coexistence paths
The choice follows a fixed order even though the three methods differ: confirm the installed equipment and wavelength compatibility first, obtain the isolation evidence for the receiving ports second, check every subscriber path against both systems' loss ranges third, and compare implementation and procurement constraints last. The three methods differ in where the wavelength selection physically sits, and that single difference shapes what a buyer has to order, mount and test. In the external method, the two systems' OLT ports are patched to a coexistence element and the shared ODN leaves its common port; in the multi-PON module, the same combination happens inside the OLT port; in the splitter-based method, the new OLT is patched out to a spare input of the splitter, so the combining happens in the ODN and the filtering has to come from the receiving ports instead.
| Decision factor | External coexistence element | Multi-PON module in the OLT port | M:N splitter architecture | What it means for the ODN and the order |
|---|---|---|---|---|
| Where wavelength selection sits | In a separate passive device between the two OLT ports and the ODN | Inside the OLT line module; no external optical device | Not in a purchased wavelength-selective part - combining happens at the splitter, and filtering is left to the receiving ports | Only the first two make wavelength selectivity something you order; the third leaves it to equipment already in the field, so its filtering has to be evidenced separately |
| What changes in the central office | Two additional patched interfaces per element plus the element's own mounting | The OLT line card or port module changes; the fiber patch to the ODN does not | The line card changes and the new port is patched towards the ODN rather than to a device in the rack | Rack space, patch count and the ODF record change under the first method, not the second; under the third the change is a fiber route and a record, not rack space |
| What changes in the ODN | Nothing | Nothing | A fiber from the new OLT to the spare input of the splitter, and the splitter itself where the installed one has no spare input; the subscriber count does not change | The change is a fiber run and possibly a splitter swap; output count is kept and no split stage is added |
| Where the extra insertion loss lands | Consumed from each system's allowance between the OLT and the ODN, on top of any patch cords and connectors added at both interfaces | Inside the module, defined by the equipment vendor rather than by the ODN bill of materials | A property of the splitter's own construction and of how its inputs are used, so it has to come from the device data; the added fiber run into that input is part of the path as well | An outside-plant budget line under the first method, a vendor figure under the second; under the third it is neither standard nor negligible, so assuming it away leaves unbudgeted loss |
| What must be true at the receiving ports | The element provides the isolation between the systems' ports | The module provides it, within the OLT | Receiving ports must provide sufficient band-to-band filtering, and older OLT receivers may need external wavelength blocking added | This is where the third method stands or falls, and it is verified from equipment specifications, not from the splitter datasheet |
| Third overlay (RF video, OTDR, further PON generations) | Modules exist with separate ports for RF video, OTDR or WDM overlays, and some carry an upgrade port for future services | Determined by the module; check which generations it is specified to combine | Extended by architecture rather than by adding a filter: later amendments to the recommendation add three-generation coexistence methods, and on this path the extension depends on the splitter's input count and on verified receiving-port filtering | A spectral plan expected to grow is an order-time question in every case: it sets the device's port list or the splitter's input count, and neither is added by configuration later |
| Main failure mode to plan for | The external element is a common point for both systems: work on it interrupts both, and its isolation assumptions govern both | Coupled to the OLT vendor's roadmap and module availability | Assumes filtering that may not be documented for the deployed receivers | Name the single point of failure and the evidence owner for each option before choosing |
| Typical fit | Fixed OLT, untouched ODN, documented headroom at the coexistence wavelengths | Replaceable OLT line card; rack space and patching are the binding constraints | A first stage that already has a spare input and receivers whose filtering is documented | The decision is made by what is verifiable about the installed plant, not by which method is newer |
Two arithmetic cautions belong with this table.
The maximum insertion loss on a datasheet may already include the port-variation allowance for that part. Whether adding a separate uniformity figure counts the same effect twice depends on how the datasheet defines the two figures and on the reference plane each one uses, so read the definitions before treating them as independent - and where they do overlap, using both produces a budget stricter than the part requires, which in turn can send the project to a higher OLT class it did not need.
And where a path fails, moving a split point is a design change rather than a remedy. It changes fiber length, splice count, enclosure count and the subscriber count on that feeder, so the per-path check has to be run again from the beginning. The same applies to a change in split ratio. The levers that leave the split topology as it is - a higher OLT class, better connector grades on the interfaces actually in the path, and correction of an identified bend or contaminated connector - avoid a redesign of the feeder, but they still change path loss, so the per-path check has to be re-run against both the upper and the lower limit before the change is accepted. What they save is the remapping of the tree, not the arithmetic.
Confirm the wavelength plan before you select hardware
Wavelength compatibility is the first item in that order, and it is where the usual shorthand misleads. The nominal shorthand - GPON at 1310 nm upstream and 1490 nm downstream (ITU-T G.984.2), XGS-PON at 1270 nm and 1577 nm (ITU-T G.9807.1) - is a centre-value convention, not a compatibility proof. The specifications define ranges, and the ranges are where the engineering sits.
| Band | Wavelength range | Condition that selects it | Source |
|---|---|---|---|
| GPON upstream, regular band | 1260–1360 nm | The general GPON upstream specification for single-fiber working; on its own it is not evidence of separation from XGS-PON | ITU-T G.984.2 (08/2019) |
| GPON upstream, reduced band | 1290–1330 nm | The narrower upstream band a GPON transmitter stays inside when G-PON shares the ODN with a 10G system; confirm the ranges against the current edition of the recommendation when writing the order | ITU-T G.984.5; band categories as reproduced in ITU/IEEE NG-PON workshop material |
| GPON upstream, narrow band | 1300–1320 nm | The narrowest G.984.5 upstream option; used in some coexistence filter variants where the guard band is widened further | ITU-T G.984.5; ITU/IEEE NG-PON workshop material |
| GPON downstream | 1480–1500 nm | Single-fiber GPON downstream; the optional RF video overlay sits in a separate enhancement band | ITU-T G.984.5 |
| XGS-PON upstream, basic plan | 1260–1280 nm | The XGS-PON basic wavelength set, the plan used when XGS-PON coexists with G-PON | ITU-T G.9807.1 (02/2023), Annex B |
| XGS-PON downstream, basic plan | 1575–1580 nm | The XGS-PON basic wavelength set. A G.9805 filter example also covers 1575–1581 nm; confirm the exact optic and coexistence element bands against their approved datasheets (G.9805 Appendix I). | ITU-T G.9807.1, Annex B |
| XGS-PON optional plan | 1480–1500 nm downstream; 1300–1320 nm upstream | The plan intended for coexistence with other 10G and NG-PON2 systems rather than with G-PON: its bands sit inside the G-PON bands, so a live G-PON system cannot share them | ITU-T G.9807.1 |
| Coexistence filter example | GPON port 1290–1330 nm; other port 1260–1280 nm and 1524–1625 nm; example loss below 0.8 dB on the GPON port and below 1.0 dB on the other port, connector loss excluded | One non-normative WDM1r example in the standard's appendix, not a product rating | ITU-T G.9805, Clause 6.1 and Appendix I |
Read the first and fifth rows together and the coexistence problem becomes concrete. A GPON transmitter is specified over 1260–1360 nm (ITU-T G.984.2); the XGS-PON basic upstream band is 1260–1280 nm (ITU-T G.9807.1). The two ranges overlap across 20 nm. What makes coexistence work is not the wavelength names but the reduced GPON band: the coexistence hardware is built with a GPON passband of 1290–1330 nm and a 10G passband of 1260–1280 nm (ITU-T G.9805, Clause 6.1 and Appendix I), which leaves roughly 10 nm between them. Isolation, not nominal wavelength, is the binding requirement.
That has a direct procurement consequence. A coexistence element cannot correct a transmitter that leaves its passband - what happens to that power depends on where it lands. Where an ONU transmits inside the other system's passband, its upstream power leaves through that system's port and arrives at a receiver tuned to receive it, so it is in-band interference that no downstream filter can separate. Where the emission falls outside both passbands, port filtering attenuates it instead, and how much leakage the receiving ports tolerate is exactly what the isolation calculation in G.9805 Appendix II settles. For installations where XGS-PON will join an existing GPON population, this makes "which upstream band do the deployed ONUs actually use?" a pre-order question, not an acceptance-test question. Where the answer is not documented, resolving the ONU population's behaviour belongs to the upgrade design rather than to an acceptance test after it; where the deployed models are confirmed to operate inside the reduced band, the question does not arise.
The opposite direction needs the same treatment. XGS-PON adds downstream power at 1575–1580 nm (ITU-T G.9807.1) to a fiber that legacy GPON ONUs are already listening on. ITU-T G.984.5 addresses ONU interference tolerance, and the 2023 revision of G.9807.1 added an annex specifying out-of-band noise limits on XGS-PON ONUs specifically to reduce the impact on other systems coexisting on the same ODN. The evidence to collect is the ONU vendor's specification for the exact deployed model, at the operating wavelength, rather than the year the ONU was installed.
Can an existing PLC splitter carry both PON bands? A splitter specified for 1260–1650 nm covers the GPON and XGS-PON bands in principle; Glory's 1×8 example lists that operating range. For the installed part, verify the exact model's wavelength range, worst-port insertion loss and connector-inclusive values in the path budget. A 1×N splitter can remain in the shared tree with CEx or MPM. The splitter-based combining method needs a multi-input first stage, typically 2×N, plus documented receiver filtering.
Which system sets the limit?
The shared ODN is limited by whichever system is more demanding on that specific path, in that specific direction. There is no single coexistence budget, and a design that compares one headline loss figure against one class limit is comparing the wrong pair of numbers.
Both systems classify the ODN by the optical path loss it may present, in decibels.
| System | Class | Optical path loss range | Source |
|---|---|---|---|
| GPON | B+ | 13–28 dB | ITU-T G.984.2 |
| GPON | C+ | 17–32 dB | ITU-T G.984.2 |
| XGS-PON | N1 | 14–29 dB | ITU-T G.9807.1 |
| XGS-PON | N2 | 16–31 dB | ITU-T G.9807.1 |
| XGS-PON | E1 | 18–33 dB | ITU-T G.9807.1 |
| XGS-PON | E2 | 20–35 dB | ITU-T G.9807.1 |
Taken at face value, the table suggests that a GPON C+ tree and an XGS-PON N2 OLT sit close together, and the similar figures may look interchangeable during early planning. The resemblance is a starting point and not a conclusion, for four reasons.
The first is direction and wavelength. Upstream and downstream use different transmitters, different receivers and different wavelengths on both systems, and each direction has its own loss limit. A path is acceptable only if it passes both systems in both directions, so the comparison has to be kept per direction rather than collapsed into one figure.
The second is bend sensitivity. The XGS-PON downstream wavelength is longer than the GPON downstream wavelength, and fiber is more sensitive to macrobending at longer wavelengths. A route that passes GPON can therefore fail XGS-PON with no visible damage and no change in the plant - which is why the acceptance figure has to be measured at the coexistence wavelengths rather than inferred from an earlier measurement at the GPON wavelengths.
The third is the lower edge of the range. Every class in the table is bounded at both ends, and a coexistence upgrade is a reason to look at the low end as well: a short branch near the head end of a split tree can present less loss than the receiver expects, and an overloaded receiver fails in a way that a calculation aimed at the maximum loss does not predict. Where a path falls below the minimum, an attenuator is the conventional remedy. Note the direction of travel: improving the loss of an existing path - by changing connector grades, for example - moves a passing path toward that lower boundary as readily as it moves a failing path toward compliance.
The fourth is the margin convention.
Fix the convention before calculating. Where the reserved design margin is already inside the class limit being compared against, the pass condition is . Where the margin is held separately, the condition is . State which one the project uses. A budget that mixes the two is not a budget.calculated path loss ≤ class limitcalculated path loss + margin ≤ class limit
Four kinds of figure appear in this comparison, and they are not interchangeable: the class limits published in the recommendations; the values in a specific device's datasheet, which describe one part; the design values in a project's route record; and the measured attenuation of the installed plant. A new build is checked by calculation from design values; an already-built ODN is checked against measured attenuation at the wavelengths of interest. In both cases the comparison is between the path value and the class limit - one kind of figure against another - and substituting a design value for an as-built measurement, or comparing two figures of the same kind, proves nothing about the installed plant.
The reference plane says exactly where a loss figure starts and ends. Each figure also has a reference plane, so whether a splitter's or a coexistence element's insertion loss includes the connectors at its own interfaces, and whether a route measurement runs from the OLT port to the ONU port or only to the enclosure, has to be stated before any of these numbers can be added to another.
Verify the ODN before you issue the RFQ
Two checks are needed, and they do not have the same shape. Loss is verified per subscriber path; isolation is verified on the worst-case pairing of an interfering path with a target path. An ODN is not "XGS-PON ready" or "not ready" as a whole; it is ready for the specific paths that pass, on the specific wavelengths that will be used, with the specific coexistence hardware that will be installed.
- Map the paths.One row per subscriber path from the OLT port to the ONU, naming every splitter stage with its ratio and package, every enclosure, and a count of the splices and connector interfaces in the path. A count derived from a single-line diagram is a design figure; say so.
- Fix direction and wavelength per row.Record each system's upstream and downstream band for that row, and the receiver limits that apply to it. A path can pass in one direction and fail in the other; a row that carries one figure cannot show that.
- Use measured loss where the plant exists - and record how it was measured.For an in-service ODN, the input is the measured attenuation at the wavelengths of interest, and the record has to carry the test points (OLT port to ONU port, or OLT port to enclosure), the wavelengths, the state of service during the test, and the reference plane the figure belongs to. For a new build, only design figures exist at order time - state that explicitly, and make the post-construction measurement the evidence that closes the record. An as-built measurement cannot be a pre-order input, because it does not exist yet.
- Put the coexistence loss where it belongs.An external element's insertion loss between its common port and each system's OLT port is consumed from that system's allowance at the head end, and patch cords and connectors added at both interfaces are separate items unless the datasheet states that its figure already includes them. Adding all of it in one place, or omitting the patching, both produce a wrong answer.
- Close the isolation evidence on the worst-case pairing, not path by path.For every receiving port, name the pair of paths the isolation figure was evaluated on - the combination that gives the worst ratio of interference to wanted signal at that receiver - and record the assumptions the figure rests on. Then, separately, record the deployed GPON ONU population's actual upstream band and its blocking-filter configuration by model. Neither is inferred from the PON type.
What the order should require. These are items the buyer should specify in the RFQ and require as evidence with the shipment:
- the wavelength plan, stated as XGS-PON basic or optional, and the GPON upstream band - reduced 1290–1330 nm or the 1300–1320 nm version (ITU-T G.984.5 band categories)
- the port list with passband and reject band for each port
- maximum insertion loss per port, with an explicit statement of whether connectors are included
- minimum isolation between the relevant ports, together with the basis for the figure - the tolerated crosstalk it comes from and the pairing of paths it was evaluated on - since the isolation a design needs is derived rather than fixed
- return loss and directivity
- whether the element passes the OTDR maintenance band on the path the operator uses for in-service testing, and the RF video band where that overlay exists
- housing, mounting and port identification, so the element can be entered into the ODN record
- which report accompanies the shipment, whether per piece or per batch, and at which wavelengths it was produced.
A requirement in an RFQ is a question addressed to a supplier, not a statement about one. Requiring per-port loss figures at both the GPON and the 10G wavelengths says nothing about whether the first quotation received will be able to provide them at the reference plane the project uses.
Which path fits which situation?
| If | Then |
|---|---|
| The OLT line card can be replaced, and rack space and patch-cord count are the binding constraints | Start from the multi-PON module: wavelength selection inside the port, no external device, no added outside-plant interfaces |
| The OLT is fixed and the ODN has verified headroom at the coexistence wavelengths | The external coexistence element is the path that leaves the ODN untouched; its insertion loss and its two patched interfaces are the cost |
| The first split stage already has a spare input, and the receiving ports' band filtering is documented | The splitter-based method can be considered; with no documented filtering at the receivers, it cannot be relied on |
| Deployed GPON ONUs are known to transmit outside the reduced upstream band | Bring the ONU population into the upgrade design rather than treating it as a post-installation fix |
| A third overlay is planned, or the spectral plan may grow | Choose coexistence hardware with the corresponding ports now; a later overlay is an equipment change, not a configuration change |
| A path fails the tighter system in one direction only | Resolve that direction's specific item - connector grade, splice, identified bend - before considering a change to the split ratio for the whole feeder |
| A path fails in both directions and the route is already at its limit | The remaining levers are the OLT optics class in the direction that fails and the split ratio. Whether a higher class actually closes the gap depends on both ends' transmit and receive windows in that direction, not on the OLT alone; and a split-ratio change alters the passive plant as well as the order, so neither lever is a free adjustment |
1x8 PLC splitter module
The splitter fixes the shared tree's ratio and package. Because both PON systems use the same tree, its band coverage and its worst-port insertion loss figure are two of the numbers a coexistence budget depends on.
View product details →
SC/APC fiber optic adapters
A head-end coexistence element adds patched interfaces on each system's port. Their connector type and grade are part of the loss the new path introduces, and their reference plane has to be stated before that loss is added to anything.
View product details →
FTTH fiber termination enclosure
A termination enclosure organizes distribution connections and their port labels. Record its connector interfaces and cable entries when auditing each ODN path. A head-end coexistence element needs mounting hardware suited to its own form factor.
View product details →Product photographs are from the linked Glory Optical product pages. They show the component types a GPON-to-XGS-PON migration touches; they are not a statement that a specific model supports any particular coexistence architecture or wavelength plan. The approved datasheet for the exact ordered model governs the order.
Where to go next
The passive migration path decides what has to be added. Because the head-end interfaces are the ones an upgrade creates, the connector and adapter range under fiber optic connections belongs in the same bill of materials as the PLC splitter range that fixes the shared tree's ratio and package, and the fiber boxes are where an added device has to be mounted and recorded.
For the budget arithmetic behind the per-path check, see the worked FTTH GPON loss budget and splitter ratio treatment. For passive assemblies built to a project's own port map, split ratio, connector type and housing, use the OEM/ODM service or send the requirement through the inquiry form. For the split-ratio trade-off across GPON and XGS-PON classes, see the 1:32 vs 1:64 PLC splitter comparison.
Related GLORY pages and technical references
Explore components: FTTH ODN network solution, PLC splitter range, 1x8 ABS module, fiber optic cable, fiber boxes and enclosures, OEM/ODM configuration and request a quote.
Check the coexistence and wavelength references: ITU-T G.9805, G.9805 Amendment 2 (11/2025), ITU-T G.9807.1, ITU-T G.984.2, ITU-T G.984.5, ITU-T SG15 XGS-PON flyer and ITU-T/IEEE NG-PON workshop, G.984.5 band categories.
