Application Notes

Here Kingfisher's experienced engineers share their experience in best practices and procedures for fiber optic testing related mostly to installation and maintenance. We hope that by sharing our knowledge, we will help grow our industry. Please enjoy & pass on these notes. Alternatively, browse Standards related Test Procedures, Equipment & Reporting

Optical loss & testing overview

For standards compliant test procedures & equipment, please refer to Standards Center. For more background theory, please read on...

Understanding Optical Loss & testing concepts in fiber systems requires a general understanding of the following major components:

Glass fiber used for data communications comes in 2 general types:

Single mode fiber

Used to transmit 1270 - 1625 nm light over long distances and high data rates, most commonly at 1310 and 1550 nm. DWDM (Dense Wavelength Division Multiplexing) systems operate in the C, S and L bands in the region of 1450 - 1625 nm, and CWDM (Corse Wavelength Division Multiplexing) systems operate over 1270 - 1610 nm. A single ray of light travels down the fiber core, with a mode field diameter of about 9.5 µm. Single mode fiber comes in varying grades typically designed to optimize cost, general loss, bending loss, chromatic dispersion, polarization mode dispersion, and with varying levels of water peak performance at 1383 nm.

Cross section of multimode and singlemode fiber

Fig 2.1: Cross-section of typical single mode and multimode fibers. In both cases, the glass diameter is 125 µm, and the diameter including a protective acrylate plastic coating is 250 µm.

Typical single mode loss is 0.35 dB / Km at 1310 nm, which with a typical link loss of 20 dB, gives a maximum link length of 57 Km.

The lowest loss wavelngth region is around 1550 nm. Best performance is achieved with for example Corning SMF-28® ULL with <0.16 dB/Km. With a typical link loss of 20 dB, this gives a maximum link length of 125 Km The ability to achieve such long transmission distances between repeaters or optcal amplifiers is a major factor in the success of fiber optics.

Typical single mode fiber operates from 1270 - 1650 nm. Wavelengths below the "cut off wavelength" of 1260 nm are transmitted as multimode light by standard 9.5 µm core fiber, and specialist small-core fiber is required to achieve single mode operation below 1270 nm. Wavelengths above 1625 nm are more heavily attenuated.

Bend Insensitive Fiber offers improved bend tolerance for installation intensive FTTX applications.

Optical Fiber Spectral Attenuation Graph showing water peak

Fig 2.2: Typical attenuation graphs for single mode fiber.
Note that the loss is sensitive to wavelength.

Multimode fiber

Is commonly used to transmit light for shorter distances and moderate data rates usually at 850 nm, and less commonly at 1300 nm. Multiple rays of light travel down the core, which has a diameter of 62.5 µm or 50 µm. Modern 50u core types have higher bandwidth and couple well with VCSEL lasers at 850 nm. Older 62.5 µm core types give better coupled power from an LED, but significantly lower bandwidth. Multimode fiber comes in varying grades typically designed to optimize bandwidth or pulse dispersion. The core has a graded refractive index structure, to reduce pulse dispersion of the various light rays. It is therefore more complicated to manufacture than single mode fiber.

Optical Fiber Attenuation / Wavemength Graph

Fig 2.3: Typical attenuation graphs for multimode fiber.
Note that the loss is sensitive to wavelength.

Fiber Optic Transmission Windows
Fig 2.4: Fibre Optic Transmission Windows

Fiber cable loss discussion

The above graphs show typical loss when the glass is mechanically unstressed. Additional loss will be created under stress, and this is termed "micro bend loss". Micro bend loss is typically caused by: kinking, tension, heat or cold, crushing, winding onto a drum, excessively sharp bending, defective manufacture, poor cable design, or defective cable manufacture. Micro bend loss for any given fiber stress, usually increases with increasing wavelength, particularly above 1550 nm. The above graph also fails to show rapidly increasing attenuation above about 1625 nm.

Multimode loss is inherently more uncertain than single mode loss, with additional variables as follows:

  • Attenuation varies depending on the power distribution (beam geometry) across the fiber core. So there can be mismatch between the measured loss, and the loss experienced by an installed system. Therfore to be meaningful, both loss measuring and system performance specifications should use a defined modal distribution to ensure that results are repeatable across different meters and systems.
  • When transmitting coherent light from lasers, coherent interference between multiple light paths creates instantaneously changing loss characteristics. This is readily seen with a power meter, where moving multimode patch leads around creates obvious power variations. For this reason, LED sources are preferred for loss testing multimode fiber systems, although LEDs are now rarely used on transmission equipment.
  • LED test sources do provide more stable loss readings, but also have other uncertain characteristics. The center wavelength, spectral width and modal power distribution in the coupled fiber can vary significantly, resulting in varying loss measurements. So the characteristics of LED test sources should comply with various standards.
  • Modern multimode systems almost invariably operate at 850 nm with VCSEL transmitters.
LED Laser Spectral Comparison

Fig 2.5: A comparison of the spectral and spatial distribution of Lasers and LEDs.

Connector & other device loss discussion

In medium to long distance applications, it is assumed that fiber attenuation will dominate any losses.

In short distance & PON networks, loss caused by devices such as connectors and optical splitters may dominate or require specific handling depending on the situation. For example, a 32-split PON network will have a splitter with a >15 dB loss, and in a short distance network with MPO connectors, the loss may be almost entirely due to connector losses.

Note in particular that connector losses are quite uncertain: the same pair of connectors will show a randomly varying loss when mated on different occasions. The connector specification will show the average value, which will often not be achieved. Connector loss is better understood by using a statistical mean and standard deviation approach.

Other components in the transmission path will introduce more loss, for example:

Component Loss
Single mode fiber 1310 nm 0.33 - 0.35 dB/km
  1550 nm 0.16 -0.22 dB/km
Multimode fiber 850 nm 2.5 - 3.0 dB/km
  1310 nm 0.7 - 0.8 dB/km
Connector 0.1 - 0.75 dB
Fusion splice 0.1 - 0.15 dB
Mechanical splice 0.1 - 1 dB
Switch 0.1 - 1.5 dB
Coupler 1 - 21 dB (splitter)
Isolator 0.5 - 5 dB

Out of band loss testing

It is common practice to measure "out of band" loss, e.g. at some extra wavelength in addition to expected operating wavelengths. Typical situations include:

  • Testing at extra operational wavelengths that might be used for future bandwidth upgrades.
  • Testing a live system that incorporates blocking filters to specifically block a test wavelength from the data systems. Most commonly, this is used for OTDR testing at 1650 nm.
  • Testing loss on single mode fiber at 1625 nm. This extra wavelength is used to check that the fiber is unstressed, since this wavelength is more sensitive to bending loss. e.g. a newly installed fiber passes at 1550 nm, but fails at 1625 nm due to a stress point.
  • Testing multimode fiber at 1300nm. Modern multimode systems are almost always 850 nm. 1300nm used to be used on legacy systems, however loss at 1300nm in now used to check that the fiber is unstressed, e.g. a newly installed fiber passes at 850 nm, but fails at 1300 nm due to a stress point.

Loss budget

The loss budget refers to the calculated allowable loss before a link will degrade, usually according to equipment performance specifications.

This is usually determined as a function of transmitter power, receiver sensitivity, and a required reserve margin. The expected losses of individual segments of the link are then estimated, along with a nominal margin to allow for degradation and maintenance. On long or high-speed links, estimating these effects for optimum overall lifetime performance, requires considerable skill and experience. On short or low speed links, the link loss may be defined in appropriate standards.

Unlike the more theoretical loss budget, the optical margin usually refers to the change in actual (tested) loss that an operating link can suffer before it stops working. Note this could be either an increase or decrease in loss.

High bit rate systems may have a quite small working tolerance for the receiver power level. This results in a requirement of tighter absolute power measurement accuracy. For example, low bit rate systems may have a loss budget of 0 - 25 dB, whereas high speed LAN systems may have a loss budget of only 0 - 2.5 dB.

Loss Measurement Techniques

To accurately measure the end to end loss of a system, it is usual to test at the target wavelength using an OLTS, or source and meter.

End to end loss on multiple joined fiber lengths cannot be reliably measured from one end with optical time domain reflectometry, since an OTDR works by mathematical deduction based on several assumptions. To accurately measure a splice or connector point loss with an OTDR, a measurement must be made from each end direction, and then averaged. This is not always simple to do. An OTDR excels at identifying the location of an event and measuring length.

The accuracy of typical loss measurement methods with a source and meter are usually dominated by these factors:

  • Source power levels drift over time, so a short measurement time is preferable.
  • Source center wavelength tolerance may be an accuracy limiting factor over long distances.
  • Optical power meter absolute accuracy is not as good as that for other measurement metrologies such as mass, length or electrical.
  • Somehow a source reference measurement must be known or obtainable, with the power meter many Km from the source, at the other end of the link.
  • Loss in the same link may be different if measured in different directions. This may be the result of mis-matching or incorrect optical or core diameter characteristics. In practice, mismatched fibre cores can be picked up if two correctly selected test cords are used during measurement.
  • Bi-directional loss method is the practical solution for all the above issues.
  • Because of wavelength sensitivity, measurement should be performed at the same wavelength as the operational system. Other wavelengths may be tested as well.

The requirement of better accuracy has made the bi-directional two-wavelength method the preferred and practical solution. With a pair of modern two-way LTS, the complete analysis can be completed in a few seconds. This measurement method requires either two LTS, or two sources and two meters.

To summarize, measurements using two-way averaging have the following benefits:

  • The elapsed time between taking a reference and second measurements may be very short, so only the short-term source stability matters.
  • Meter absolute accuracy is irrelevant.
  • Core diameter or other optical mismatches will always be found.
  • The highest possible accuracy is always achieved.
  • If a sub-standard test cord is used, errors are halved.
  • A low level of operator skill is required.

Effect of Test Wavelength Uncertainty on Loss Measurements

Most test sources have a significant practical wavelength tolerance. For example, many "1310 nm laser sources" may have a wavelength tolerance of ± 30 nm. Add ± 5 nm for temperature variations, resulting in an actual wavelength of 1310 ± 35 nm. If an LED is used, the FWHM spectral width is probably ± 50 nm, resulting in the possibility of light being transmitted anywhere in the band of 1300 ± 85 nm, e.g. from 1215 - 1385 nm. Fiber loss characteristics may vary quite significantly over these extremes.

As a result of source wavelength uncertainty, there may be some variation between the conditions under which a system is tested, and actual in-service conditions. To put this into perspective, @ 1310 nm on a link of say 50 Km, the link attenuation uncertainty caused by a tolerance of ± 30 nm would be about ± 0.95 dB.

In some older systems, or in a research environment, possible effects due to cladding mode transmission may need to be assessed, however with modern cladding mode stripping fiber, these effects can usually be ignored.


  • Loss should always be measured at the same wavelength band as the operational system.
  • Loss is ideally measured from both directions.
  • Loss should be measured with a source and meter, or LTS.
  • Multimode loss measurements are intrinsically inaccurate compared to those on single mode systems. Adherence to relevant standards may be advisable, to ensure repeatability.
  • Cladding modes can occasionally cause measurement errors in unusual circumstances.
  • The effects of wavelength variations within a wavelength window should be evaluated, both on measurement accuracy, and on the installed system.
  • An allowance for future degradation should be made, to ensure reliable performance. This is called the optical margin.

See A10 Practical Measurement Procedure

© Kingfisher International, Australia, ABN 51 007 250 213