Unbalanced Splitter Architecture for PON Tap Chains: Design and Acceptance

What is an unbalanced splitter architecture?

An unbalanced splitter architecture is a passive optical network topology in which each tap sends a defined share of optical power to a local branch and passes the rest to the next node. Approval is route-specific because each downstream endpoint inherits the devices, ports, fiber segments, wavelengths, and directions before it.

For this method, the working document for procurement and construction is a tap-chain route matrix. It links each node ID to the topology revision, device and document revision, common input, through port, local drop port or group, input-to-output loss limits, operating wavelength, transmission direction, downstream dependencies, and acceptance evidence. A missing field keeps the route at feasibility stage. A supplier can match components against the matrix, while the network owner and qualified designer remain responsible for the ODN architecture and acceptance limits.

Design notice: The equations, field lists, and downloadable matrix are workflow aids for reference only. They do not guarantee network performance or project acceptance. Product limits, system classes, design allowances, test methods, and release decisions must come from the project's approved records and qualified technical review.

This guide begins after the team has proposed an unbalanced tap chain. Use the GPON splitter selection guide if the architecture is still being chosen, the fiber optic splitter types guide to normalize product terminology, and the fiber loss budget calculator to calculate one fully defined route. The method below controls the dependencies across all routes in the chain.

Already have a proposed route? Go directly to the tap-chain route-matrix fields. Keep an unknown model, port, loss condition, or acceptance input visibly open instead of guessing it for an early quotation.

On this page

• Model the ordered tap nodes
• Build the tap-chain route matrix
• Calculate every continuing and local path
• Check both sides of the loss window
• Control the recalculation blast radius
• Carry port identity into acceptance
• Approve a proposed tap model
• Move from validation to a controlled purchase
• Prepare the submittal and BWNFiber review
• Read the technical and procurement FAQs

Five checks before design release

1. A nominal ratio does not identify the physical through and drop ports or their finished-device insertion losses.
2. Each local branch and each continuing segment needs its own cumulative path calculation.
3. GPON or XGS-PON review must cover the applicable minimum and maximum optical path-loss window, not only the longest path.
4. A change at one tap can invalidate the local branch and every dependent route downstream.
5. The drawing, BOM, field label, as-built record, and test file use the same node, model revision, and port identity.

Model the network as ordered tap nodes, not one split ratio

The Fiber Broadband Association describes unbalanced splits, or optical taps, as a distributed architecture in which power outputs are adjusted along the route and may feed additional splitters.[1] VIAVI describes the same local and continuing paths: one output feeds a local customer or group, while the other continues to the next tap. Ratios may change along the chain.[2]

A useful abstract record is:

OLT → Segment S1 → Tap T1 → Segment S2 → Tap T2 → Segment S3 → Tap T3

At every tap, record four identities:

Writing "1:32 overall" beside the route may describe capacity, but it cannot approve the design. It hides which input-to-output loss applies at T1, T2, or T3 and which endpoint calculations become stale after a port or component change.

The word tap describes the network function. PLC and FBT describe component technologies. Technology still belongs in the device record, but it does not replace the path, port, wavelength, or loss fields. If technology selection is still open, resolve it on the existing splitter selection page before releasing the tap-chain schedule.

ITU-T G.671 uses tap coupler as a synonym for an asymmetric branching component and distinguishes its main and tap ports.[3] A project may call those ports through and local, but the route matrix must map the project names to the physical labels on the proposed model. Similar terminology is not enough to establish port identity.

Build the tap-chain route matrix before product approval

Build the matrix before the quotation and product submittal are finalized. Give each tap one node row, then link every service endpoint to the complete ordered set of nodes on its route. This exposes missing inputs before a part number is treated as approved.

Keep fiber, cable, enclosure, and splitter evidence separate

The splitter architecture does not select the route cable. ITU-T G.652 (08/2024) covers single-mode fiber and cable, while ITU-T G.657 (08/2024) covers bending-loss-insensitive single-mode fiber and cable; its Category A requirements include A1 and A2 with different macrobending performance.[9][10] These categories can be entered in the segment record, but the actual attenuation and bend limits must come from the approved cable and installation data.

IEC 60794-1-1:2023 establishes generic optical-cable requirements for transmission, material, mechanical, ageing, climatic, and related properties.[11] It does not approve a particular ADSS, duct, direct-burial, armored, drop, loose-tube, or tight-buffered construction for this route. LSZH, PE jacket, UV resistance, water resistance, flame performance, and rodent protection are separate material or performance requirements; one label does not prove the others.

Match the controls to the deployment scenario

A scenario label does not approve the architecture. It changes which route-matrix fields and failure modes deserve extra review. This avoids the common shortcut of treating every rural route as suitable for taps or every urban route as suitable for a centralized split.

The matrix remains the same tool across these scenarios. The emphasis changes because the source of uncertainty changes.

The BWNFiber fiber optic splitter configuration guide owns the broader model, package, connector, document, and quotation fields. The table above adds only the fields required to control serial dependencies in a tap chain.

Technical-review next step: use the matrix as the specification request. Ask the proposed supplier to return model-matched drawings and optical evidence against the named fields. A quotation that leaves a required field blank remains commercially received but technically on hold.

How do you calculate an unbalanced splitter link budget?

Define the path inputs

Let:

• U(i-1) be the cumulative loss at the through output of tap i-1; set U(0) = 0 at the defined upstream reference plane;
• S(i) be the loss between that upstream point and the common input of tap i; for the first node, S(1) includes every passive loss from the reference plane to tap 1;
• T(i) be the exact input-to-through-port insertion loss of tap i for the defined wavelength and direction;
• D(i) be the exact input-to-local-drop-port insertion loss of tap i for the same condition;
• B(i) be any additional local branch loss after the tap, including a separate balanced splitter when present.

Apply the continuing- and local-path equations

Continuing-path equation

The cumulative continuing-path loss after node i is:

U(i) = U(i-1) + S(i) + T(i)

Local-branch equation

The loss to the local service branch at node i is:

L(i) = U(i-1) + S(i) + D(i) + B(i)

These equations account for path loss; they are not product specifications. Run them as separate cases. The worst-loss case uses the applicable maximum segment and input-to-output losses. The minimum-loss or overload case uses the applicable minimum expected values. If a required minimum or maximum value is missing, that route remains on hold.

Take T(i) and D(i) from the approved limits for the proposed physical ports, wavelength, direction, and model revision. ITU-T G.671 identifies transmission parameters for asymmetric branching components, including main-port and tap-port insertion loss, reflectance, polarization-dependent effects, allowable input power, and directivity. It does not approve a BWNFiber or third-party model or define its installation conditions.[3]

Integrated and discrete nodes need different records

An integrated unequal multi-output PLC assembly may expose one high-power continuing port and a local group within a single device. A discrete node may combine a 1×2 unequal tap with a separate local balanced splitter. They need different records:

• For an integrated device, use the documented loss limits from the common input to every relevant output or output group.
• For a discrete node, list the tap and local splitter separately and count each interface once.
• State whether the published device loss already includes connectorized interfaces. Otherwise, the same interface can be counted twice.

The current BWN-PLC-1X9-80:20 model page is evidence that BWNFiber publishes model-specific unequal configurations. It is not authority to transfer that model's ratios, port identities, losses, wavelength scope, or package data to another device or revision.

Calculate a matrix, not one "worst endpoint"

Calculate every local branch and the final continuing endpoint. Repeat the matrix for each required direction and operating wavelength. The route with the largest loss may change when direction, wavelength, device revision, or local splitter construction changes.

If service continues beyond the last tap, add the final fiber segment, connectors, splices, and endpoint elements after U(n). A calculation that stops at the last through port is not an end-to-end route result.

Use the fiber loss budget calculator for each fully defined route. Save the exported record with the tap-chain matrix so the endpoint result can be traced back to its ordered nodes and ports.

Check both sides of the optical path-loss window

ITU-T G.984.2 is the in-force GPON PMD recommendation and includes optical loss classes.[4] ITU-T G.9807.1 defines XGS-PON ODN optical path-loss classes in terms of minimum and maximum loss over possible paths and operating wavelengths; a 2025 amendment is also available.[5][6] The route matrix needs the class and equipment limits for the deployed optics. A familiar class copied from another project is not a valid input.

A route passes only when all three gates pass. Meeting the maximum-loss limit does not clear an overload risk, and a numeric result cannot release a route with an unresolved model or port identity.

IEC 60875-1:2024 is the generic specification for passive, non-wavelength-selective fibre optic branching devices.[7] It supplies component-class context. Certification, qualification, and project suitability still require the model-level evidence named in the contract.

Define the recalculation blast radius before any change

When a tap, segment, or port changes, the dependency field identifies every calculation that has become invalid. Review that impact set before releasing a substitution, repair, or route revision.

Illustrative field-review scenario: Assume tap T2 is replaced before the closure is sealed. The spare carries the same nominal ratio, but its drawing uses different main/tap labels and different loss limits. The T1 local branch can remain unchanged if it does not traverse T2. The T2 local branch and every route continuing through T2 stay on hold until the port map, affected calculations, labels, and retest scope are approved. This is a review example, not a BWNFiber customer case or measured result.

The cumulative equations make the rule simple: if a changed value appears in a downstream route, the saved result is stale. Updating the BOM description alone cannot preserve an earlier pass status.

Same-ratio replacement is not automatic equivalence

Two devices can share a nominal ratio and still differ in input-to-output maximum loss, wavelength behavior, connector inclusion, fiber leads, package, labeling, or environmental evidence. A replacement is acceptable only after the controlled requirement and affected route calculations show equivalence for the intended use. If that evidence is incomplete, record the substitution as a hold.

Carry port identity from drawing to field test file

Unbalanced routes create a record-control problem as much as an optical one. Keep one identity chain:

Approved drawing → BOM/model revision → field node and port label → as-built route → test file and approval

At receiving and installation, match the chain against the approved model drawing and labeling schedule. Pigtail position or an assumed color convention is not enough to identify a port.

Name the accountable party for each gate in the project quality plan. The roles below are typical starting points and must be aligned with the contract.

The Fiber Broadband Association notes that unbalanced or cascaded approaches can increase testing, troubleshooting, training, and record complexity.[8] That is why GIS or as-built node identity and an explicit route schedule are operational requirements, not documentation polish.

How should an unbalanced splitter OTDR trace be interpreted?

VIAVI reports that unbalanced splitters can produce a different OTDR event signature and may be misinterpreted when the instrument or application is not configured to recognize it.[2] Treat that as an instrument- and method-dependent risk:

• Define whether OTDR evidence is required and what it must prove.
• Confirm that the instrument and application support the proposed topology and stage definitions.
• Load the expected node order and tap behavior when required by the test method.
• Use the launch and receive arrangements and settings specified in the approved method.
• Investigate a mismatch between the trace and the route matrix. Relabeling an event to force a pass destroys traceability.
• Keep end-to-end loss acceptance and OTDR event analysis as separate results when both are required.

An OTDR event map does not replace the project owner's end-to-end optical path-loss acceptance requirement. Conversely, one acceptable end-to-end power result does not prove that the installed through/drop port sequence matches the drawing.

If the contract requires return-loss or reflectance evidence, keep that result separate from end-to-end attenuation and OTDR topology interpretation. State whether the limit applies to a component or the installed route, then record the wavelength, direction, reference plane, test method, instrument setup, and acceptance limit. ITU-T G.671 references IEC 61300-3-6 for return-loss measurement in this component context.[3]

When to hold or reject the tap-chain architecture

Several mistakes can make an incomplete design look numerically acceptable. Treat each one as a hold condition rather than polishing the spreadsheet around it.

For a linear FTTH route, an unbalanced tap chain is appropriate only when node order, port identity, per-path loss, and downstream change records can be controlled. In the hold conditions above, compare a centralized, balanced-cascade, or revised distributed design using the splitter selection guide. The lowest initial fiber count or component count is not sufficient reason to accept operational ambiguity.

How do you approve a proposed tap model?

Use the same technical approval sequence for a named model, an alternate, or a replacement:

1. Freeze the topology revision, direction, wavelength, endpoint list, and applicable system and equipment limits.
2. Map the project's common, through, and local names to the physical labels on the exact model drawing.
3. Compare maximum and minimum input-to-output losses, wavelength conditions, interface inclusion, package, and construction against the matrix.
4. Recalculate every local and continuing endpoint that depends on the node; include the final segment and endpoint elements.
5. Review cable, enclosure, mechanical, environmental, packaging, and installation evidence as separate requirements.
6. Define the acceptance method, limits, retest scope, evidence owner, and final state: approved, conditional, or hold.

Approval is complete only when the result, its sources, and its responsible owner are recorded. A clean quotation or familiar part description cannot replace a technical disposition.

Use a tap-chain supplier evidence scorecard

Evaluate the supplier's response against the route rather than against marketing breadth.

Price and schedule can be compared after the technical response reaches the required evidence state. A lower quote does not close an unresolved port, loss, traceability, or acceptance gap.

Move from technical validation to a controlled purchase

An engineering guide should not force every reader into an immediate quote. It should show what decision comes next and what record must survive that decision. For an unbalanced tap chain, the ten-stage path below keeps technical approval and commercial negotiation connected without treating them as the same gate.

Normalize price and lead-time comparisons

Compare commercial offers only after every supplier is pricing the same controlled scope. At minimum, align:

• exact model, revision, physical port map, included connectors or adapters, and permitted alternatives;
• order quantity and price basis, optional sample or first article, and any one-time approval activity;
• document, labeling, serial or lot, test-record, packaging, and language requirements;
• destination, delivery basis, requested date, and the event that starts the quoted lead time;
• drawing or sample approval time, buyer witness points, and responsibility for a rejected submission;
• quotation validity, exclusions, change-notification route, and effect of an approved deviation on price or schedule.

A low unit price is not comparable if another offer includes model-matched test evidence, controlled labeling, or an approved sample step that the first excludes. Likewise, a short lead time is ambiguous until the quotation states whether it begins at inquiry, purchase order, drawing approval, sample approval, or another agreed trigger.

Treat datasheets, samples, and reports as different evidence

A current datasheet defines the proposed model limits. A configuration drawing resolves construction and physical ports. A sample or first article can verify the submitted configuration under an agreed inspection or test plan. A production or batch record addresses the delivered quantity where the contract requires it. None of these records alone proves that the network architecture or every installed route will pass.

If the project requires private labeling, OEM/ODM packaging, bulk scheduling, export documents, or destination-specific marking, add each item to the quotation and document register. Confirm availability, scope, quantity, destination, approval sequence, price, and lead time for the exact proposal; do not infer them from a company description or product-family page.

Localize the evidence package, not the optical theory

The ordered-path equations and model-evidence rules do not change by sales region. The project inputs do. Treat the table below as an RFQ screening matrix, not as a statement that every country in a region uses the same climate, standard, approval, or buying practice.

A destination never proves compliance. The country, network operator, contract, installation site, and authority requirements control. Keep splitter optical evidence, cable evidence, enclosure evidence, export packaging, and commercial documents as separate submittal lines. If a certificate, declaration, approval, or local test report is required, name the exact document, applicable product, edition or scheme, issuing party, language, and acceptance owner in the RFQ.

How do you prepare an unbalanced splitter submittal?

The submittal should add tap-chain dependencies to the normal model and quotation fields:

1. Topology drawing ID, revision, and route direction.
2. OLT and ONU/ONT equipment records and the selected optical path-loss class.
3. Ordered tap-node list with unique node and location IDs.
4. Common, through, and local port identity at every node.
5. Integrated or discrete node construction, including any local balanced splitter.
6. Exact proposed model, current datasheet and configuration-drawing revision, substitution rule, and any sample or first-article verification required by the contract.
7. Maximum and minimum input-to-output loss used for each direction and wavelength.
8. Fiber category, cable construction, installed length, connector pairs, splices, filters, attenuators, route environment, and other passive elements by route.
9. Endpoint route IDs and the downstream dependency set for every tap.
10. Worst-loss, minimum-loss, and evidence-readiness result for every endpoint.
11. Packaging, connector protection where applicable, label, serial/lot, as-built, and test-file naming requirements.
12. Acceptance method, limits, instrument/application requirements, retest scope, and approval owner.
13. Country, operator, and contract requirements for document language, marking, environmental or fire evidence, test-file format, witness points, and release authority.
14. Commercial request covering quantity, destination, required date, packaging, optional sample or first article, quotation validity, approved deviations, and the event that starts lead time.

What can BWNFiber review before quotation?

BWNFiber's published splitter workflow asks buyers to send a BOM, network diagram, loss budget, or reference model when a configuration spans several catalog fields. For a tap-chain project, the route matrix is the clearest review interface: BWNFiber can compare it with published splitter families and model records, identify candidate model options, and flag missing product or document fields. The BWN-PLC-1X9-80:20 page is one model-specific discovery record. Its ratios, losses, port identities, and package data cannot be transferred to another device or revision.

The published BWNFiber project workflow moves from project inputs to engineering review, then to sample and quotation where applicable, followed by production and test after the configuration is approved. That sequence does not promise that every tap model, sample, customization, quantity, document, price, or delivery date is available. Confirm the exact offer and evidence in the written quotation and approved order records.

BWNFiber review principle: Route matrix before model approval. Evidence before price. Revision before release.

Request an unbalanced splitter architecture evidence-gap review. Send the topology revision, tap-node schedule, ratio-to-port map, direction/wavelength matrix, endpoint calculations, package and interface requirements, quantity, destination, required date, and evidence list through the verified splitter inquiry form. Ask for the current datasheet, configuration drawing, available test-record format, optional sample or first-article path, and a written project quotation where applicable. BWNFiber can return candidate model options and unresolved fields; the project owner and qualified designer retain final responsibility for the architecture, optical budget, acceptance criteria, and release.

Frequently asked questions

How does an unbalanced splitter architecture work?

Each node sends a defined share of optical power to a local branch while its through path continues to the next tap. Optical tap can describe the same network function. The exact device record still needs the technology, physical ports, finished-device loss, package, and wavelength scope.

Is an unbalanced splitter always PLC or always FBT?

No. PLC and FBT identify component technologies; unbalanced describes power distribution, and tap describes a network function. Approve the technology and complete model record that satisfy the project requirement instead of deriving technology from the word "tap."

Can one overall split ratio approve the chain?

No. It may summarize capacity, but it does not identify the ordered through/drop losses or which endpoints depend on each node. Approval requires route-specific calculations.

How do I calculate loss through several optical taps?

Add the upstream segment losses and the exact through-port loss of every earlier tap, then add the exact local-drop loss and local branch elements at the endpoint node. Repeat for every endpoint, direction, and wavelength.

Which output is the continuing port?

Use the approved model drawing and physical label. Pigtail position, color, or a similar product cannot establish the continuing port.

Can multiple unbalanced splitters be cascaded?

They can form a series architecture when every local and continuing path meets the approved system, equipment, operational, and acceptance requirements. A feasible first node does not prove the downstream chain.

Does one calculation cover upstream and downstream?

No. The physical route may be shared, but transmitter, receiver, wavelength, penalty, and component conditions can differ. Save separate, linked records for each required direction.

Why check minimum path loss as well as maximum loss?

The maximum-loss check protects receiver sensitivity and design reserve. The minimum-loss check protects the receiver input/overload boundary and the application's minimum OPL requirement. Both sides must be satisfied where applicable.

Can I replace a tap with the same nominal ratio?

Not automatically. Compare model revision, port identity, minimum and maximum input-to-output loss, wavelength scope, interfaces, package, labels, and required evidence. Then recalculate every affected route.

Can OTDR alone certify an unbalanced PON route?

No universal method should be assumed. Define the project acceptance plan first. OTDR can support event localization and topology verification, but the instrument must handle the expected unbalanced events, and separate end-to-end loss evidence may still be required.

How should an integrated 1×N unequal splitter be recorded?

Record it as one exact device with one common input, one continuing output when applicable, each local output or group, and the documented input-to-output limits. Model it as a discrete 1×2 plus balanced splitter only when that is its approved construction.

What should I send BWNFiber for review?

Send the topology revision, node and port schedule, proposed models or required performance fields, directions and wavelengths, endpoint calculations, package and interfaces, acceptance evidence, quantity, destination, requested date, and required documents. State whether you need a datasheet, drawing, test-record format, sample or first article, technical review, or quotation. Mark missing fields as open; guessed values make the review less reliable.

References and scope

1. Fiber Broadband Association, Introduction to Passive Optical Network Splitter Architectures, architecture terminology and series-tap context.
2. VIAVI, Certifying PON with Unbalanced Splitter Architecture, continuing/local behavior and OTDR interpretation risk.
3. ITU-T G.671 (11/2025), current in-force optical-component transmission recommendation; use for terminology and component transmission context, not model approval.
4. ITU-T G.984.2 (08/2019), in-force GPON PMD recommendation.
5. ITU-T G.9807.1 (02/2023), XGS-PON PMD and ODN optical path-loss classes.
6. ITU-T G.9807.1 Amendment 1 (05/2025), current amendment identified during review.
7. IEC 60875-1:2024, generic specification for passive non-wavelength-selective branching devices.
8. Fiber Broadband Association, Splitter Architecture Part 2, operations, testing, troubleshooting, training, and records context.
9. ITU-T G.652 (08/2024), current in-force characteristics of single-mode optical fiber and cable.
10. ITU-T G.657 (08/2024), current in-force characteristics of bending-loss-insensitive single-mode optical fiber and cable.
11. IEC 60794-1-1:2023, generic requirements for optical fiber cables and cable elements.

Sources were checked on 2026-09-06. Standards context does not establish certification or compliance for a specific product. Final network design and acceptance remain the responsibility of the project owner and qualified system designer.