An OLT location can look settled while the feeder design is still undecided. Put the equipment in an existing building and the farthest service area may require a long detour; move it into a remote cabinet and the project gains a powered site to maintain. We start with that physical choice because it shapes the network outside the rack.
What OLTs do in a passive optical network
Optical line terminals, or olts, are the powered network endpoints that connect an operator's aggregation network to customer optical network terminals through passive fiber splitters. In a 1:32 PON, one OLT port serves a tree with up to 32 splitter outputs, sharing transmission capacity while controlling upstream access.
The signal path runs from a PON interface through a feeder fiber to a splitter, then outward through distribution fiber and a drop to the customer's ONT. The splitter needs no electrical supply. Both endpoints do. The Fiber Optic Association's reference, FTTH PON: Passive Optical Network, describes this shared optical tree and the splitter's reverse role as a combiner for upstream signals.
Downstream delivery and upstream scheduling
Downstream light reaches the branches of the tree, and the PON system identifies the traffic intended for each endpoint. Upstream transmission needs coordination because ONTs share the return path. Without controlled transmission windows, their bursts could overlap at the OLT receiver. We treat this shared capacity as a design input when grouping customers, especially where an owner expects sustained upstream demand.
An ONT terminates the access connection at the customer side and presents the service interface. The OLT provides the operator-side PON termination and connects toward aggregation through uplinks. These are different jobs. A splitter provides neither job, and a cabinet symbol on a drawing does not establish whether that cabinet contains powered equipment or only passive distribution hardware.
GPON and XGS-PON change the capacity decision
A commonly deployed GPON rate pair is 2.48832 Gbit/s downstream and 1.24416 Gbit/s upstream. The physical-layer reference is ITU-T Recommendation G.984.2, Gigabit-capable Passive Optical Networks (G-PON): Physical Media Dependent (PMD) layer specification, August 2019 edition, approved August 29, 2019. Those are line rates shared by the PON. Subscriber payload capacity is lower after protocol overhead, and a retail service tier does not reserve the entire port for one customer.
XGS-PON uses nominal 10 Gbit/s transmission in both directions. Its reference is ITU-T Recommendation G.9807.1, 10-Gigabit-capable symmetric passive optical network (XGS-PON), February 2023 edition, approved February 22, 2023, with Amendment 1 dated May 2025. We distinguish that symmetric capability from GPON's common asymmetric rate pair before comparing service plans against the expected traffic load.
More line capacity does not automatically extend a feeder. Optical reach still depends on the selected interfaces and passive path, and changing the PON generation does not remove a congested aggregation uplink. We carry both questions into the architecture review. Our GPON network design guide develops the broader relationship between the access tree and the service area.
Where the OLT belongs: central office, hut or cabinet
The OLT site anchors the feeder routes. A central office may offer existing power and maintenance access, but an apparently nearby neighborhood can require a much longer cable route when the available corridor crosses around restricted property. We measure the usable route. Straight-line distance belongs in an early screening map, not in the final optical calculation used to approve the serving boundary.
Centralized equipment and remote equipment
A hut can bring the access electronics closer to a distant service area while keeping them inside a dedicated equipment enclosure. A remote cabinet can serve the same geographic purpose with a smaller physical footprint. Neither placement eliminates site engineering. We want a documented power arrangement and a thermal assessment before a candidate location becomes the origin of a committed feeder design.
Maintenance access deserves its own review. A cabinet that fits beside a roadway may leave no safe working area when its doors are open, or the feeder entrance may conflict with the space needed for power equipment. We check the enclosure arrangement against the actual site. Equipment fit alone cannot resolve those access constraints.
Our preference is to compare feasible serving locations before assigning every address to a PON tree. The limitation of this approach is that it takes more front-end design effort than adopting the first available equipment room. That effort is justified when a site change would otherwise force the owner to redraw feeder boundaries after outside-plant work has already been released.
| Design choice | Centralized OLT | Remote OLT | Required design check |
|---|---|---|---|
| Serving location | Existing central office or equipment room | Hut or powered cabinet nearer demand | Usable site and maintenance access |
| Feeder routing | Routes converge on the central site | Access feeders begin at the remote site | Actual cable paths and corridor availability |
| Port allocation | Ports grouped at fewer locations | Port capacity divided among sites | Growth allowance within each serving area |
| Optical reach | Longer access paths may consume margin | Shorter access paths may recover margin | Loss and distance limits for each interface |
| Operations | Fewer powered access sites to maintain | Additional power and environmental exposure | Backup strategy and restoration access |
The table describes alternatives, not a universal ranking. An existing central site is a strong choice when its routes pass the optical checks and it has room to grow; a remote site earns its place when the access-path improvement outweighs the additional operating responsibility. We compare the backhaul separately because a short PON feeder can still depend on a long transport connection.
Check the cabinet's function: Label the enclosure as an active OLT site or a passive splitter location on the concept plan. That distinction determines whether power and environmental control belong in the site scope before detailed routing begins.
How OLTs turn port count into feeder and splitter decisions
A chassis port count tells us how many PON interfaces can be available, subject to the installed equipment and enabled configuration. It does not tell us how many addresses can receive the promised service. Start with the serving areas. We allocate the optical trees to real routes before treating total splitter outputs as useful network capacity.
Count connected endpoints and reserve growth
Consider an illustrative layout with 8 PON ports and one 1:32 tree per port. Multiplying 8 by 32 gives 256 theoretical splitter outputs before any reserve. That is arithmetic, not a subscriber forecast. A tree may contain unused outputs while an adjacent tree is full, and moving a customer between them may require a different splice path or additional distribution fiber.
We therefore keep homes passed separate from connected ONTs. A passed address can still need a drop and an assigned splitter output before activation, while a connected endpoint consumes a specific place in the operating tree. Our planning record should show the intended growth case without assuming that every spare port can serve every future address.
For an illustrative requirement of 180 connected endpoints at 32 per tree, the arithmetic minimum is 6 PON ports because 180 divided by 32 rounds upward. Geography may require more. We would check whether the addresses can be grouped into those trees without violating optical limits or creating impractical feeder paths before accepting that minimum as a design quantity.
Centralized splitting and cascaded splitting
A centralized splitter keeps its outputs together at one location, which can make patching and reassignment easier to organize. Cascaded splitting divides the optical tree in stages. A 1:4 stage feeding 1:8 stages produces an overall 1:32 split when fully populated. The Fiber Optic Association's Fiber To The Home Network Design reference describes that arrangement and the need to include both stages in the loss calculation.
We choose the arrangement around the distribution geography and the owner's operating method. Cascading can reduce fiber requirements on selected segments, but it also places allocation decisions deeper into the field. A later boundary change may affect several downstream branches. Draw the stages explicitly so that an overall ratio cannot hide where each physical splitter sits.
Higher split counts share the PON capacity among more endpoints and consume more optical power. Neither effect should be evaluated alone. Our FTTH design work connects port assignments to the feeder and distribution layout so that the equipment schedule agrees with the network the construction package actually describes.
Reach starts with the optical path, not a circle on the map
Distance and optical loss are separate limits. A path can satisfy an interface's distance requirement and still lose too much power through its passive components, while a low-loss path can exceed a supported timing or distance limit. We check both ends of the operating range. Very short paths also need review against the receiver's maximum acceptable input power.
Calculate loss at the relevant wavelength
The planning equation adds fiber attenuation to the insertion loss of every component on the path, then retains the owner's required engineering margin. We calculate it for the relevant operating wavelengths and directions. Connector losses belong in the total. So do splices and any coexistence element that the upgrade architecture introduces, using the approved component data rather than an assumed generic allowance.
Splitting has an unavoidable power-division penalty. For an ideal equal splitter, 10 log10(N) gives about 15.05 dB for 32 outputs and 18.06 dB for 64 outputs. Doubling the split therefore adds about 3.01 dB of ideal loss. The Fiber Optic Association's Testing Fiber Optic Couplers, Splitters or Other Passive Devices explains the underlying division of optical power and the additional excess loss found in real devices.
Those ideal values are lower bounds. We use the specified insertion loss of the actual splitter assembly for design, including its connector arrangement, and retain margin for the accepted operating conditions. A second splitter stage must appear in the calculation even if the final drawing labels the tree only by its total ratio.
Suppose, purely as a design example, an approved interface budget permits 28 dB and the calculated path consumes 24 dB before reserve. The difference is 4 dB. If the owner requires a 3 dB engineering margin, only 1 dB remains beyond that reserve. A route revision must fit within that remaining headroom or trigger a revised design; the numbers are assumptions for this example, not a standard allowance.
Recalculate after a route change: A construction detour changes the installed optical path even when the serving OLT and customer address stay the same. Carry the revised length into the loss budget before accepting the updated route.
Check the farthest and nearest endpoints
The longest route is an obvious candidate for optical review, but it may not have the greatest loss. Another branch can contain more connections or a different splitter path. We evaluate each distinct path configuration. The acceptance plan should also check short, lightly attenuated connections so that the design does not overlook receiver overload while concentrating on the far edge of coverage.
A GPON-to-XGS-PON transition needs its own path review when both technologies share the distribution network. Coexistence requires compatible optical arrangements and endpoints; an old ONT does not gain XGS-PON capability merely because the OLT changes. We check the added passive loss before preserving the existing split ratio. Reusing fiber is useful only when the revised system passes its own operating requirements.
Carry the OLT assignment into the feeder design and operating record
The feeder schedule must identify which fiber leaves each PON port and where that fiber reaches its splitter input. Traceability starts at the rack. We recommend a stable interface identity tied to the physical patching arrangement so that a port reassignment can be reconciled with the outside-plant drawings rather than treated as an isolated equipment change.
In a simple unprotected arrangement, each PON port has its own feeder fiber to the first splitter input. Shared ducts do not change that logical relationship. Protection or coexistence arrangements can alter the physical implementation, so we show the selected topology explicitly instead of applying a blanket rule that every access network uses the same fiber count.
Feeder cable sizing also needs an allowance for the owner's future network plan. Spare fibers can support later trees without replacing the feeder cable, provided their path reaches the area that needs capacity. We reserve by route. A spare strand ending at the wrong enclosure offers little help to a full distribution branch elsewhere in the serving area.
Make the handoff usable during a change
The operating record should connect the OLT interface to a customer endpoint through the installed passive path. Equipment telemetry can identify an associated ONT, but it cannot by itself prove every intermediate splice. Our OLT as-built record guide explains how assignment changes remain connected to the accepted physical baseline after initial network acceptance.
We recommend checking the relationship in both directions: outward from the serving port and back from a selected endpoint. Both traces should agree. If a field change moves a splitter input to a different feeder fiber, the updated schedule must reach operations before a future outage sends technicians to the superseded path.
Testing should retain enough context to identify the path that was measured and the criteria used to accept it. An unexplained pass label is difficult to reuse after a repair. We want the recorded measurement connected to the applicable design budget, with any exception resolved by the owner's acceptance process rather than silently absorbed into spare margin.
Choose the architecture before the equipment order
For an owner extending an existing GPON network: Verify available capacity on the actual serving trees before ordering more chassis capacity. Keep the existing OLT location when the revised paths pass their optical checks and the site supports the growth plan. If only one branch is constrained, redesign that branch before moving an entire serving area.
For a greenfield operator: Compare central and remote locations against feasible feeder corridors before fixing the splitter layout. Select the PON generation from the intended service demand and migration plan, then allocate ports by serving area. Approve the site arrangement early. Otherwise an unresolved power connection can hold up an access design that looks complete on the fiber sheets.
For a team upgrading to XGS-PON: Validate coexistence and endpoint compatibility before assuming that every existing tree can retain its current ratio. Recalculate the loss budget with the intended optical components and check aggregation capacity against the upgraded service load. We would retain usable passive plant, with changes driven by the specific constraints found in that review.
When OLT assignments and feeder drawings are being developed separately, Draftech International can bring them into one coordinated delivery scope. We are a turnkey telecom and power infrastructure company with 100% in-house engineering and full turnkey construction delivered by Draftech-managed subcontract crews. Our partner and delivery overview identifies the team; use our project contact page to discuss the serving-area assumptions that need to be resolved before detailed design.
For an early route that needs those assumptions tested, our free design offer covers engineering the first 20,000 linear feet of a qualifying route at no cost, from feasibility and field survey through permit approval. The owner reviews each request before Draftech commits the package, so the starting point is a defined route and a clear account of the OLT and feeder decisions still open.
