Fiber Optics: Link Budget, Polarity and OTDR Acceptance

Fiber Optics: Link Budget, Polarity and OTDR Acceptance

By Zoey Zhang, Founder & Lead Sourcing Specialist

Sourcing Guide

Four numbers decide whether a fiber link passes acceptance: the application's maximum channel insertion loss (10GBASE-SR is 2.6 dB on OM3 and 2.9 dB on OM4, 40GBASE-SR4 is 1.5 dB on OM4), cable attenuation at 3.5 dB per km, 0.75 dB per mated connector pair plus 0.3 dB per splice, and the measured loss from a Tier 1 OLTS report. A 400 m OM4 backbone with two connector pairs and two splices comes to 3.5 dB - 0.6 dB over the 10G limit - and only pre-terminated ends bring it back to 2.15 dB.

Last updated: 2026-11-09

Technical China Sourcing Vlog: 2-Day BOQ & QC for CCTV, Barrier Gates, Smart Locks & IP Phones

    Fiber is rarely where a project overspends. It is where a project discovers, on acceptance day, that a link is a few tenths of a dB over budget: traffic still passes, but the 40G module drops, or one cabinet starts losing packets every afternoon. What gets reworked then is not the cable - it is the ceiling, the tray and the wall that has already been closed.

    This page does not repeat the selection logic (Cat6 versus OM4) already covered in structured cabling sourcing. It covers the three things a buyer can actually verify: how the link budget is calculated, which termination to buy, and which two test reports to require.

    Rows of fiber patch cords dressed into a rack, curving through the cable managers on both sides

    1. “It runs 10G” is an arithmetic table, not a promise

    The application standard gives a maximum channel insertion loss. The buying job is to add up your own deductions and see what margin is left:

    ApplicationFiberReach limitChannel insertion loss limit
    1000BASE-SXOM2550 m3.6 dB
    10GBASE-SROM3300 m2.6 dB
    10GBASE-SROM4400 m2.9 dB
    40GBASE-SR4 (8-fiber MPO)OM4150 m1.5 dB

    Three deductions, all at the limit rather than the typical value:

    • Cable attenuation: OM3/OM4 at 850 nm is 3.5 dB/km maximum under IEC 60793-2-10 (TIA-568.3-D calculates at 3.0 dB/km); 1300 nm is at least 1.5 dB/km
    • Connectors: 0.75 dB per mated pair for standard grade, the TIA-568.3-D limit; 0.30 dB for reference grade on multimode
    • Splices: 0.3 dB each as the standard limit, with 0.15 dB common on single-mode fusion splicing

    Worked example: a 400 m OM4 backbone in an office building, one patch panel pair at each end, two splice points in the middle, running 10GBASE-SR.

    DeductionValue
    Cable (0.4 km x 3.5 dB/km)1.40 dB
    Connectors (2 pairs x 0.75)1.50 dB
    Splices (2 x 0.3)0.60 dB
    Total3.50 dB
    Application limit2.9 dB - 0.60 dB over, fails

    The same link with pre-terminated ends (one pair at each end, no splices) totals 2.15 dB with 0.75 dB of margin. That is why backbones move to pre-termination: what you save is not labour, it is 1.35 dB.

    When the whole low-voltage package is bought together, fiber is usually the last leg of the backbone and the hardest to redo; the package scope and interface split is set out in hotel ELV system integration. Below the backbone sits the access layer, where port density and uplink budgets are covered in PoE switch sourcing.

    An SFP+ optical module marked for 850 nm and 10G, with the specification label buyers use to confirm wavelength

    2. Three termination methods: who carries the 0.6 dB

    MethodLoss per pointSite workReworkFits
    Fusion splice with pigtails0.3 dB limitSplicer, splice tray, construction recordMedium (re-splice)Backbone, outdoor, single-mode long reach
    Pre-terminated MPO0.75 dB per pair (standard grade), low-loss grade optionalPlug and test polarityLow (swap module)Comms room to floor, 40G/100G
    Field-assembled connectorDepends on the installer, batch variationNo splicer neededHigh (re-terminate)Temporary links, a few extra points

    3. Fix polarity before the quotation, not on commissioning day

    Duplex links use one of two arrangements (TIA-568.3-D Annex C): consecutive-fiber positioning (A-B at one end, B-A at the other) or reverse-pair positioning (same orientation at both ends, 2,1,4,3 at one end). Array (MPO) links use three methods, and they are not interoperable:

    MethodTrunk type12-fiber mappingTypical use
    AType A, key-up to key-downP1 to P1Most common, with A-to-B and A-to-A jumpers
    BType B, key-up to key-upP1 to P12Parallel links when 10G migrates to 40G/100G
    CType C, pair-wise flippedP1 to P2Duplex backbones with standard A-to-B jumpers

    Later revisions of TIA-568.3-D add Methods U1 and U2, which support multiple duplex signals only and no array links. The buying action is simple: put the method letter and jumper types in the BOM, never just “MPO trunk”. A polarity error will not show up as excess loss - it shows up as every link reversed on the day the network is switched on.

    4. End faces: blue and green do not mate

    ItemUPCAPC
    End faceFlat (slightly domed)8 degrees angled
    Return loss floor50 dB or better60 dB or better
    ColourBlueGreen
    FiberSingle-mode and multimodeSingle-mode
    MatingMated with APC the insertion loss reaches 3 dB or more and both ferrules are damaged-

    Cleaning and inspection follow IEC 61300-3-35:2022 (third edition): zone A (core) and zone B (cladding) plus the whole contact area, up to 250 um diameter on cylindrical ferrules. That edition removed the zone C and D requirements and states explicitly that visual inspection does not replace attenuation and return loss measurement. In practice the most expensive mistake is an ODF ordered with APC adapters and patch cords shipped UPC - discovered only when the goods arrive.

    5. Two reports: Tier 1 decides, Tier 2 locates

    TIA-568.3-D Annex E splits field testing into two tiers:

    • Tier 1 (OLTS): loss, length and polarity - this is the pass/fail record; field methods follow IEC 61280-4-1 for multimode and IEC 61280-4-2 for single-mode
    • Tier 2 (OTDR): the event trace, showing which connector, splice or bend carries the loss

    The OTDR has to be bidirectional. Measured from one end only, two connectors with different backscatter produce a negative-loss “gainer” event that reads better than the link really is. A proper report therefore contains traces from both directions plus the event table, not a single screenshot.

    6. At 40G the margin is a few tenths of a dB: MPO grade decides

    On the same 150 m OM4 run for 40GBASE-SR4 (8-fiber MPO uplink), the limit is only 1.5 dB:

    ComponentStandard-grade MPOLow-loss MPO (0.35 dB per pair)
    Cable (0.15 km x 3.5)0.53 dB0.53 dB
    2 connector pairs1.50 dB0.70 dB
    Total2.02 dB - fails1.23 dB - passes

    The difference is neither cable nor distance, only whether the quotation says “low-loss grade”. On links like this, paying slightly more at quotation stage is much cheaper than opening the ceiling at acceptance.

    7. A 500-room hotel backbone: count the fibers first

    SystemPointsFibers (1 per point)With 20% spare
    Cameras120120144
    Access points214214257
    Access control points606072
    Total394394473

    473 fibers is about 20 positions on 24-fiber ODF panels; three risers at two 24-fiber cables each comes to 144 backbone fibers. Fiber count, ODF positions and patch cord quantities are all arithmetic - do not accept “as per experience”. On the cabinet side, U-space and dressing space are broken down in ELV package delivery and handover.

    8. Ten lines for the RFQ

    1. Fiber type (OM3 / OM4 / OS2), core count and number of runs
    2. Cable construction: tight or loose buffer, indoor or outdoor, flame class (CPR class where required)
    3. Termination method (fusion, pre-terminated MPO, field-assembled) and who performs the splicing
    4. Polarity method and jumper types, with the method letter written out
    5. Connector type and grade (LC/SC, UPC/APC, standard or reference grade)
    6. MPO grade (standard or low-loss) and fiber count (8/12/24)
    7. ODF specification: 24-fiber positions, adapter type, splice tray included or not
    8. Test scope: whether Tier 1 includes length and polarity, and whether bidirectional Tier 2 traces are supplied
    9. Delivery documents: attenuation report, polarity table, reel-to-link cross reference
    10. Spare ratio (patch cords, adapters, pigtails), warranty and the non-conformance clause

    9. Five goods-in checks and three site acceptance items

    Goods-in: (1) microscope-inspect the first and last connector of each reel; (2) measure insertion loss and return loss on three random patch cords against their marked grade; (3) metre marks continuous along the jacket with no joints; (4) ODF adapter type matches the BOM - this is the step that catches APC/UPC mix-ups; (5) polarity and labelling match the manufacturer’s polarity table.

    On site: (1) Tier 1 on 100% of links, with anything over budget failing and no sampling; (2) Tier 2 bidirectional traces on a 10% sample, with no unexplained spikes in the event table; (3) a link register (link ID, both end ports, length, measured loss) matching the cabinet labels one for one.

    Untagged cable bundles inside a rack that cannot be traced back to the drawing, a common starting point for fiber acceptance rework

    The link register earns its place in the contract because it decides whether, three years later, anyone can tell which port goes to which floor when a 40G link is added. Without it, a fiber backbone is a pile of patch cords that all look the same. The point counts behind those links are in CCTV camera sourcing and the routing side is in cable tray and containment.

    10. Six traps

    1. Budgeting with typical values: 3.0 dB/km cable and 0.3 dB connectors make a beautiful table that fails on the day acceptance applies the limits
    2. Polarity only on the drawing: no method letter in the BOM or jumper list, so every contractor interprets it differently
    3. APC and UPC mixed: a green connector into a blue adapter starts at 3 dB of insertion loss
    4. One-direction OTDR screenshots: gainers stay invisible and nobody knows which connector carries the loss
    5. MPO without a grade: the two quotations differ in connector grade and nobody can see it at bid evaluation
    6. Fiber count sized to today: three years later one more 40G run finds no spare fiber

    What to send us

    The verifiable part of fiber comes down to four numbers: application limit, cable attenuation, connector and splice deductions, measured loss. Put those four into the RFQ and the acceptance clause, and the rest is who produces the reports and whether they can be read. Send a drawing or BOM and we will return the loss calculation and test scope in writing within 48 hours; the free RFQ template lines up the fields first.

    For the whole package - cameras, access control, wireless, IPTV and backbone in one procurement chain - see hotel ELV system integration, or contact us with the project scope.

    Frequently asked questions

    The supplier says the link 'runs 10G fine' - can that go into the contract?

    No, because reach is not a claim, it is a loss limit set by the application standard: 10GBASE-SR is 300 m at 2.6 dB on OM3 and 400 m at 2.9 dB on OM4, while 40GBASE-SR4 on OM4 is only 150 m at 1.5 dB. Acceptance looks at three rows - measured link loss, length and polarity - against the limit of the application that link will actually run, and the budget is calculated on worst-case connector and splice losses, not on typical values.

    Fusion splice, pre-terminated MPO, or field-assembled connectors?

    It comes down to who carries the 0.6 dB. A spliced joint has a 0.3 dB limit (0.15 dB is common on single-mode with a good splicer), but every splice needs a splice tray and a construction record. Pre-terminated MPO has the connectors made in the factory, so the site only plugs and tests polarity - the loss is guaranteed by the factory, but the 0.75 dB per mated pair at both ends does not disappear. Field-assembled connectors save the most schedule and depend most on the installer's hands. The longer the backbone and the more joints it has, the more it should move towards pre-terminated.

    Why do two quotations for the same fiber differ so much?

    Usually it is not the cable but the connector grade. In TIA-568.3-D a standard-grade mated pair is limited to 0.75 dB, while a reference-grade pair is 0.30 dB on multimode; MPO comes in a standard grade and a low-loss grade. A quotation that says only 'MPO patch cords' cannot be compared with another one. Multimode attenuation also has two conventions: IEC 60793-2-10 uses 3.5 dB/km maximum at 850 nm, TIA-568.3-D calculates at 3.0 dB/km - over 400 m that is 0.2 dB, enough to flip the conclusion on a link with a few tenths of margin.

    The OTDR trace is full of spikes - how do I know if the link passes?

    Separate the two reports. Tier 1 is the OLTS reading of loss, length and polarity, and that is what decides pass or fail. Tier 2 is the OTDR event trace, which shows where the loss sits - connector, splice or bend. The OTDR has to be bidirectional: measured from one end only, two connectors with different backscatter read as a negative-loss 'gainer' event that looks better than reality. The pass criterion is still the Tier 1 loss against the budget; the OTDR locates problems, it does not replace the budget.

    How many fibers should run between the comms room and the floor cabinets?

    Count points, not current usage. A 500-room hotel is roughly 120 cameras, 214 access points and 60 access control points - 394 points, one fiber each, 473 fibers with 20% spare, which is about 20 positions on 24-fiber ODF panels; three risers at two 24-fiber cables each comes to 144 backbone fibers. Adding a riser later costs far more than the 24 fibers you did not buy.

    Next step

    Need a quote or a technical answer?

    Tell us the product and the quantity. Our team in Guangzhou sends back pricing, MOQ and lead time within one business day.

    Guangzhou-based sourcing team · Replies in English within 1 business day

    Share this guide:
    WhatsApp