A 100G link was designed, installed, tested, and failed by 0.4 dB. The installer retested, then cleaned every end face, then replaced two patch cords before anyone went back to the design sheet and found the problem: the budget had been calculated with three connector pairs when the route actually crossed seven.
Nothing was defective. The design simply counted the wrong things, and by the time the loss showed up there was no room left to absorb it.
What a Loss Budget Actually Is
A loss budget is an estimate of how much optical power a link will lose between transmitter and receiver, built from typical component losses — the fiber, the connectors, the splices, and any passive components such as splitters. Some standards call it the attenuation allowance. FOA (the Fiber Optic Association) defines it as an estimate calculated from typical component losses for each part of the cable plant, used to judge whether the measured insertion loss is reasonable.
That second half matters more than the first. The loss budget is not the pass/fail limit; it is the number you compare test data against to decide whether a cable plant was installed properly. If measured loss exceeds the calculated budget by a significant amount, the link should be tested segment by segment to find the cause.
The budget also feeds the power budget of the link — the difference between what the transmitter launches and what the receiver can detect. A cable plant that consumes more of that budget than expected leaves the transceivers with nothing to work with.
For network, SAN and InfiniBand equipment, that allowance can also depend on the gap between the receiver’s sensitivity and the transceiver’s output into the fiber. Two ports at the same nominal speed can therefore carry different budgets, which is why the transceiver datasheet matters as much as the cabling standard.
The Components You Add Up
A link is the path between transmitter and receiver, and the budget is built from the losses along it.
Component | What to count | Typical figure |
|---|
Fiber attenuation | Every meter, at the operating wavelength | Depends on fiber type (below) |
|---|
Connector pairs | Every mated pair in the route | 0.3 dB typical; 0.75 dB standard-grade maximum |
|---|
Splices | Every fusion or mechanical splice | 0.15 dB single-mode fusion; 0.3 dB multimode mechanical |
|---|
Splitters, WDMs | Passive components in the path | Per component specification |
|---|
Margin | Headroom for aging, contamination, rework | Design to 70 percent of the allowance |
|---|
The failure mode is almost always the connector count. Every patch panel, cassette, adapter and equipment port adds a mated pair, and a route that looks like three connections on a riser diagram is often seven in the ceiling. On small budgets, that gap decides whether the link comes up.
Attenuation by Fiber Type
Fiber loss is the one component most people get roughly right, because the numbers are published.
Fiber | Wavelength | Typical loss | Maximum per TIA |
|---|
Multimode OM3/OM4 | 850 nm | 3 dB/km | 3.5 dB/km |
|---|
Multimode OM3/OM4 | 1300 nm | 1 dB/km | 1.5 dB/km |
|---|
Single-mode OS2 | 1310 nm | 0.5 dB/km | 1.0 dB/km premises, 0.5 dB/km outside plant |
|---|
Single-mode OS2 | 1550 nm | 0.4 dB/km | 1.0 dB/km premises, 0.5 dB/km outside plant |
|---|
In premises terms, multimode at 850 nm costs roughly 0.1 dB per 30 m, and single-mode at 1310 nm roughly 0.1 dB per 200 m. That asymmetry is why a data center budget is dominated by connectors and a campus budget is dominated by distance.
For reference, OM4 attenuation per TIA-568 at 850 nm is 3.5 dB/km, and a calculator working on a maximum-distance link often uses 0.4 dB as the distance contribution. On shorter links that figure can be tuned down, which returns budget to the rest of the design.
Connector Loss: Three Different Numbers
This is where fiber loss budgets go wrong, because three credible numbers circulate and they are not interchangeable.
Figure | What it describes | Source framing |
|---|
0.3 dB | Typical adhesive/polish or fusion splice-on connector | FOA typical component loss |
|---|
0.75 dB | Standard-grade connector maximum per EIA/TIA 568 | Standard limit |
|---|
0.5 dB | Reference-grade mated to standard-grade, per TIA-568.3-E | Tightened test limit |
|---|
0.5 dB (LC), 0.75 dB (MPO) | Per mated pair, per TIA-568 interconnect loss | CABLExpress reading |
|---|
The reconciliation is that these answer three different questions. 0.3 dB is what a good connector actually costs and is the right figure for estimating a typical budget. 0.75 dB is the maximum a standard-grade connector is allowed to cost and is the right figure for a worst-case design. 0.5 dB is the test limit that applies when reference-grade test cords are mated to standard-grade field connectors — TIA-568.3-E tightened the multimode figure to 0.50 dB, with the first and last connector allowance moving from 0.30 dB to 0.50 dB.
Two details sharpen it further. Reference-grade connectors are built to tighter tolerances: a mated multimode reference-grade pair should offer 0.10 dB or less, and single-mode 0.20 dB or less. And multifiber MPO connectors carry a higher allowance than LC — 0.75 dB per mated pair against LC’s 0.50 dB — which is exactly why cassettes and MPO trunks consume budget faster than a duplex design.
The Arithmetic, Worked Through
A basic link of an MPO-to-MPO trunk with two MPO-to-LC cassette modules at each end gives the shape of the problem.
Element | Count | Loss each | Total |
|---|
MPO mated pairs | 2 | 0.75 dB | 1.5 dB |
|---|
LC mated pairs | 2 | 0.50 dB | 1.0 dB |
|---|
Fiber attenuation | short link | — | ~0.1 dB |
|---|
Total | | | ~2.6 dB |
|---|
That lands within a 2.6 dB allowance — and consumes all of it. For 100G the interconnect allowance is tighter still, which is why this same configuration is described as roughly 1 dB over the specification for interconnects supporting 100G.
The trend is unforgiving in one direction only. At 10Gb Ethernet and 8Gb Fibre Channel the totals are already small — 2.6 dB, of which 2.2 dB is interconnect loss. By 256Gb Fibre Channel and 400Gb Ethernet they fall to 1.9 dB, with connector loss between 1.5 dB and 1.0 dB. Supportable multimode distances shrink to between 100 m and 150 m, and most data center links sit inside that — so the budget gets tighter while the cable run stays short. Connectors become the whole design.
Design to 70 Percent, Not 100
The single most useful rule in this article: design the link to only 70 percent of the published maximum loss budget.
The reason is that the published figure assumes new components in perfect condition. Real links accumulate loss from contamination, end-face wear, connector rework, aged index-matching gel in mechanical splices, and transceiver output that drifts as the hardware ages. Designing to 100 percent means the link works on day one and becomes marginal in year three.
Working to 70 percent has a second effect: it forces a component-quality decision early. Products specified at 0.15 to 0.25 dB per LC mated pair and 0.20 to 0.35 dB per MPO mated pair leave enough headroom for the rest of the design, and that headroom is what makes a plant survive a speed upgrade without recabling.
Why OTDR Is Not the Pass/Fail Test
OTDRs are excellent for locating an event — splice loss on a long link with inline splices, or a fault during troubleshooting. They are not the qualification measurement.
All standards require an insertion loss test, with a light source and power meter, to qualify link loss. On multimode fiber an OTDR will underestimate loss considerably — as much as 3 dB on a 10 dB link — and the amount is unpredictable. On long single-mode links the gap narrows, but other uncertainties remain, including connector and splice readings where an OTDR can report a gain rather than a loss.
One consequence catches people out. An OTDR analyzes each event separately, but a connection point built by splicing pigtails is physically two fusion splices plus one connection. Reading each event individually understates what that point costs the link.
Documenting the Budget
The reference method decides whether the end connectors belong in the budget, and it is not optional detail. With a one-cable reference, the connectors on both ends are included in the measurement, so the budget should include both. With a three-cable reference, two connections are already baked into the zero reference, so the budget should exclude the end connectors. With a two-cable reference, only one connector is included.
Whatever method is presumed must be documented when the budget is calculated, because a budget built for one reference method and tested against another will not agree — and the disagreement looks like a bad installation.
Then there is the part worth internalizing. The budget is an estimate, and the test has measurement error, so neither number is exact. FOA’s worked example makes it concrete: a link with a budget of 8.0 dB that measures 8.2 dB should not be rejected, because the budget uncertainty is on the order of ±0.5 dB and the test uncertainty is similar, putting both figures in overlapping ranges. A link measuring over 9 dB is worth investigating. FOA notes this gap between estimated and measured loss causes more friction between managers and installers than almost any other factor in a cabling project.
Our fiber optic cable range covers single-mode and multimode constructions with the attenuation performance these budgets depend on, and communication cable covers the copper side of the same build. Our patch cord assemblies are tested as complete units, test and inspection parameters are published by measurement, and how Ethernet cables work explains the transmission principles underneath both. If a link is over budget and the components have already been blamed, send us the numbers — the component list, the reference method and the measured loss usually locate the disagreement.
Frequently Asked Questions
What is a fiber loss budget?
An estimate of the total optical loss a link will incur, built from typical losses of the fiber, connectors, splices and passive components. It is compared against insertion loss test results to judge whether the cable plant was installed properly.
How do I calculate a fiber loss budget?
Multiply the connector pair count by the loss per mated pair, add the splice count multiplied by splice loss, add fiber attenuation multiplied by link length, then add margin. Count every mated pair in the route, not just the ones on the riser diagram.
What is an acceptable loss for a fiber link?
Acceptable loss is set by the application standard for the speed being deployed, not by a single universal figure. Then design to only 70 percent of that allowance so the link still works as components age.
How much loss is normal for an LC connector?
0.3 dB is a reasonable typical figure for a good connector. 0.75 dB is the standard-grade maximum, and 0.50 dB is the TIA-568.3-E limit when reference-grade test cords mate with standard-grade field connectors.
Can I qualify a link with an OTDR instead of a power meter?
No. All standards require an insertion loss test with a light source and power meter to qualify link loss. OTDRs locate and characterize events and are used for troubleshooting and outside plant work.
The Short Version
A loss budget is an estimate, not a limit, and it exists to be compared against measured insertion loss. Build it from the fiber, every mated connector pair, every splice, any passive component, and margin.
The numbers to use are published but easy to mix up. Fiber attenuation runs about 3 dB/km for multimode at 850 nm and 0.5 dB/km for single-mode at 1310 nm. A good connector costs 0.3 dB, a standard-grade connector is allowed up to 0.75 dB, and the test limit under TIA-568.3-E is 0.50 dB. MPO pairs cost more than LC pairs, which is where cassette-heavy designs lose their headroom.
Count the connectors honestly, design to 70 percent of the allowance rather than 100, document the reference method the budget assumes, and qualify the link with a light source and power meter rather than an OTDR. And when the measurement lands slightly over the estimate, check the arithmetic before rejecting the installation — both numbers carry uncertainty.
Sources
- The Fiber Optic Association (FOA) — Guidelines On What Loss To Expect When Testing Fiber Optic Cables, covering the statement that to judge whether a fiber optic cable plant is good one does an insertion loss test with a light source and power meter and compares it to an estimate of what is a reasonable loss for that cable plant, that the estimate called a loss budget is calculated using typical component losses for each part of the cable plant being the fiber, splices and or connectors, that if the measured loss exceeds the calculated loss by a significant amount the system should be tested segment by segment to determine the cause of high loss, that a loss budget estimate can also be used to compare results from OTDR testing but the inherent uncertainties of OTDR testing make the estimate less accurate, that some standards refer to the loss budget as the attenuation allowance, that the calculated loss budget is an estimate which assumes the values of component losses and does not take into account the uncertainty of the measurement, that the loss budget of the cable plant is a major factor in the power budget of the fiber optic link and is what one calculates to compare against tested insertion loss, that calculating a loss budget involves estimating all the component losses of fiber, splices and connectors and summing them up, the connector loss guidance that for each connector 0.3 dB loss is usually figured for most adhesive polish or fusion splice-on connectors and that the loss spec for prepolished mechanical splice connectors or multifiber connectors like MPOs will be higher at 0.75 maximum per EIA/TIA 568, that when testing cable plants per OFSTP-14 double ended connectors on both ends of the cable should be included when using the 1-cable reference and when testing per FOTP-171 single ended only one connector should be included, the splice loss guidance that for each splice 0.3 dB should be figured for multimode mechanical splices at 0.3 maximum per EIA/TIA 568 and 0.15 dB for singlemode fusion splices, the fiber loss guidance that for multimode fiber the loss is about 3 dB per km for 850 nm sources and 1 dB per km for 1300 nm with 3.5 and 1.5 dB/km maximum per EIA/TIA 568, which roughly translates into a loss of 0.1 dB per 100 feet or 30 m for 850 nm and 0.1 dB per 300 feet or 100 m for 1300 nm, and that for singlemode fiber the loss is about 0.5 dB per km for 1310 nm sources and 0.4 dB per km for 1550 nm with 1.0 dB/km for premises and 0.5 dB/km at either wavelength for outside plant maximum per EIA/TIA 568, which roughly translates into a loss of 0.1 dB per 600 feet or 200 m for 1310 nm and 0.1 dB per 750 feet or 250 m for 1300 nm, the summary formula that the estimated loss of a cable plant is calculated as 0.5 dB times the number of connectors plus 0.2 dB times the number of splices plus fiber attenuation times the total length of cable, the OTDR guidance that OTDRs are used for verifying individual events like splice loss on long links with inline splices or for troubleshooting and that all standards require an insertion loss test for qualification of the link loss, that in multimode fibers the OTDR will underestimate the loss considerably by as much as 3 dB in a 10 dB link but the amount is unpredictable, that in long distance singlemode links the difference may be less but there are other measurement uncertainties like connector or splice loss where the OTDR can show a gain, that for insertion loss testing the loss contributors are simply summed up to get a total for the cable run while in the case of an OTDR each event is analyzed, and that where both fibers were terminated with spliced-on pigtails the event should be analyzed as the sum of 2 fusion splices and one connection rather than each individually, the note that a patchcord termination would be two connection losses plus splices if the termination was by splicing on pigtails, the note on including connectors on the ends covering that they should be included for two reasons being that when the cable plant is connected up to communications equipment with patchcords the connections to the patchcords will have loss and that when the cable plant is tested the reference cables will mate with those connectors on the ends and their loss will be included in the measurements depending on the method used to set the 0dB reference, that testing standards often include 3 different ways of setting the 0dB reference, that all three of these methods are approved in many standards but they will give different loss readings due to the connections included when making the 0dB reference measurement, that with a 1 cable reference the connectors on both ends of the cable will be included in the loss so the loss budget should include both connectors, that with a 3 cable reference the loss budget should not include the connectors on the ends because two connections are included in setting the reference, that with a 2 cable reference only one connector is included, and that whatever test method is presumed it must be documented when the loss budget is calculated, the statement that short fiber optic premises cabling networks are generally tested in three ways being connector inspection and cleaning with a microscope, insertion loss testing with a light source and power meter or optical loss test set, and polarity data, and that outside plant testing adds OTDR testing to the connector inspection, insertion loss and polarity testing, the statement that none of these are absolute numbers and that the loss budget created early in the design phase estimates the loss of the cable plant based on estimates of component loss and therefore is not an absolute number but an estimate to be used to compare to test data, that test data is created by instruments which have measurement errors, the worked example that a loss budget for each fiber of 8.0 dB against a measured loss of 8.2 dB should not lead to rejection because the uncertainty of the loss budget is probably plus or minus 0.5 dB providing a range of 7.5 to 8.5 dB and the uncertainty of the loss test is probably in the same range so the actual loss is in the range of 7.7 to 8.7 dB and there is considerable overlap with no reason to reject the fiber, that if one fiber is testing at over 9 dB there is reason to double check tests to determine if it is acceptable, and that in the FOA experience those two factors cause more stress between managers and installers than just about any other factor in a cable plant project — https://www.thefoa.org/tech/loss-est.htm
- Fluke Networks — New Loss Budget Values for Reference Grade Connectors in ANSI/TIA-568.3-E, covering the update that Versiv 6.1 software implements a change whereby the multimode limits for first and last connector have been changed to 0.50 dB in accordance with the TIA-568.3-E standard, that the latest revision of the standard calls out for tighter test limits when mating reference-grade connectors to standard-grade connectors, that there are several steps to be taken to reduce the uncertainty in loss testing of installed field links with the first step being to use the 1-cord method and in the case of multimode to use an LED source and Encircled Flux compliant Test Reference Cords, that the standard recommends the use of reference grade connectors with clause 7.3.4 referencing ANSI/TIA-526-7 and ANSI/TIA-526-14 which recommend the use of reference-grade connectors, that reference-grade connectors are built to tighter tolerances which results in their having lower attenuation values in a mated connection, that a pair of mated multimode reference-grade connectors should offer an attenuation value equal to or less than 0.10 dB and that single-mode reference-grade connectors should offer an attenuation value less than or equal to 0.20 dB, the Versiv Version 6.1 statement that when mating a reference-grade connector to a standard-grade connector better than the standard-grade loss value of 0.75 dB can be expected and that the TIA-568 standard specifies that the loss limit for a mated pair of reference-grade to standard-grade connectors is less than or equal to 0.50 dB for multimode and less than or equal to 0.50 dB for singlemode, that the limit names changed to TIA-568.3-E and for the multimode REF Grade limits the first and last connector loss allowance changed from 0.30 dB to 0.50 dB, the Versiv Version 6.0 statement that when mating a reference-grade connector to a standard-grade connector better than the standard-grade loss value of 0.75 dB can be expected and that the TIA-568 standard specifies the loss limit for a mated pair of reference-grade to standard-grade connectors as less than or equal to 0.30 dB for multimode and less than or equal to 0.50 dB for singlemode, and the statement that with the release of the Versiv 5.1 firmware limits were added that allow choosing between the use of Reference Grade or Standard Grade connectors and that if using the Test Reference Cords supplied by Fluke Networks the TIA-568.3-E Multimode REF Grade limits which contemplate the use of reference grade connectors may be selected — https://www.flukenetworks.com/knowledge-base/certifiber-pro/new-loss-budget-values-reference-grade-connectors-ansitia-5683-d
- CABLExpress (Todd Kiehn) — Maximizing Fiber Optic Infrastructure Performance through Loss Budgets, covering the statement that the miniscule fiber optic loss budgets for high-speed topologies such as 400Gb Ethernet and 256Gb Fibre Channel are a real challenge for data center managers, the definition that the total loss budget is the total acceptable amount of optical or signal power loss expressed in decibels that a fiber optic link can sustain and still operate error-free, that a link is defined as the path of the cable between the transmitter and the receiver, that link loss budgets are measured based on the amount of attenuation related to the distance and number of connectors in the link, that prior to designing or installing a fiber optic link the path and connections should be calculated for maximum or worst-case loss, that acceptable fiber optic link loss budgets are defined by the various published speed standards from groups like the IEEE and the Fibre Channel Industry Association addressing speeds from 1Gb to 400Gb for Ethernet and 4Gb to 128Gb Fibre Channel, the factors for attenuation list including absorption scattering or impurities in the glass, the distance from end to end of the link, the design of the fiber optic cables, the number of connectors splices and coupler systems in a link, inferior product design, degradation in the transceivers or active components, aging passive components, exceeding bend radii and end-face contamination, the standards history covering IEEE 802.3ba and 802.3bm introducing 100G Ethernet into the data center space, 802.3cd addressing 50 100 and 200Gb, 802.3bs addressing 200Gb and 400Gb Ethernet and Fibre Channel going from Generation 6 to Generation 7 in 2018 with 64Gb speeds on duplex connections to 256Gb on parallel connections, the statement that starting at 10Gb Ethernet and 8Gb for Fibre Channel the loss budgets are already very small at 2.6 dB total link loss with 2.2 dB of that associated with interconnect loss, that as we progress up into the 256 Gb Fibre Channel and 400 Gb Ethernet spectrum those losses drop to 1.9 dB total link loss with the connector loss of 1.5 dB to 1.0 dB, that these higher speeds leave very little margin for error in fiber optic infrastructure components and that insertion loss is a critical performance parameter in data center cabling measurements, that supportable distances for these new high-speed requirements for Ethernet and Fibre Channel is shrinking to between 100 meters and 150 meters on multimode fiber and most data center links will not exceed this distance, that attenuation of OM4 per the TIA-568 standard at 850nm is 3.5 dB/KM and that 0.4dB can be estimated as standard loss in calculations per the distance spec of the link with this distance loss further tunable in extenuating circumstances if the link is shorter than the maximum standard, the guidance that it is prudent to eliminate any contributing factors that would introduce loss into a link that is not native to the installed fiber optic cabling plant or its connectors splices or any of the components and that even small loss points of 0.1 and 0.2dB add up very quickly, the statement that following the industry standards for interconnect loss per TIA-568 the LC connector has a standard of 0.5dB per mated pair and the MPO has 0.75dB per mated pair, the worked scenario featuring a basic link of two MPO-LC cassette modules between an MPO to MPO fiber optic trunk where the maximum dB loss will equate to -2.5db and the observation that this is 1dB over the specification for interconnects in the link to support 100G and that it does not take many interconnects to significantly be at or above the 40Gb spectrum minimum loss requirements without the hope of ever achieving 400Gb, the recommendation that prior to designing or installing a fiber optic system a link loss budget analysis be conducted of the passive components making up the end-to-end fiber optic links in the system and that the analysis should be conservative, the statement that it is best practice to design the link out to only 70% of the maximum loss budget stated and that this conservative approach will help ensure the connected equipment will work over the installed fiber optic link, the design challenge discussion covering that the standard loss ratings assigned to the components needed require research, that standard ratings alone should not be the final guideline for product selection, that standards organizations work to be universal and unbiased which results in a standard that offers a minimum recommended requirement for performance and a starting point reference, that to ensure a design that leaves headroom in the link budget products should be used that exceed these industry standards, and that higher-quality low dB loss fiber interconnects such as LC and MPO and their components will come at a cost increase because fiber optic termination is a labor intensive process, the statement that proper core termination requires a skilled and knowledgeable workforce and that the process is very unforgiving of mistakes, that most of the defective subset will need to have the connector removed and the process started again, that allowing for higher dB loss will mean less waste and lower costs while higher expectations increase cost for better performing product, and the guidance that for the most flexibility in design you should locate products that offer 0.15-0.25dB for LC and 0.20-0.35 for MPO per mated pair which will allow deploying a manageable fiber infrastructure that will last the next 10 to 15 years and significantly exceed the industry standards, the estimating guidance to first find the allowed link loss budget per speed of transmit and use it to calculate the typical expected total link loss from the length of the fiber optic link and then add up all of the interconnects while keeping in mind that splices will need to be calculated as well but are highly unusual in a data center, the additional considerations covering quality components manufactured for performance and reliability, best practices installation maintaining proper bend radii and efficient cable routing, inspection of the fiber end faces and cleaning as necessary prior to installation, planning an acceptable dB buffer into the calculation and transceiver effectiveness, the note that loss budget allowances for network SAN and InfiniBand equipment may depend on the difference between the sensitivity of the receiver and output of the transceiver into the fiber, that degradation of the sources or transceiver over time is a factor that many forget or are not aware of, that the quality of fiber optic connectors and splice terminations in the infrastructure may also degrade over time with some field-installable connectors and mechanical splices wearing and the index matching gel degrading which can contribute to higher losses, that connectors can get dirty or contaminated even if best practices are followed and that with the fiber diameter of less than a human hair contamination is not easily avoided with consideration of data centers in very low humidity locations where static charges on connector ferrules attract particles and dirt even after cleaning, the statement that even small loss points of 0.1 and 0.2dB add up very quickly, and the example stating that seven interconnects with only 1.25dB of loss is achievable — https://www.cablexpress.com/whitepaper/understanding-fiber-optic-loss-budgets/
