2:Loss Measuring Concepts

Written by Bruce Robertson, 97. Reviewed Dec 2005. © Kingfisher International

 Optical Loss Measuring Concepts

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

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

Typical singlemode 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 ability to achieve such long transmission distances between repeaters or amplifiers is a major factor in it’s success.


Optical Fiber Spectral Attenuation Graph showing water peakOptical Fiber Attenuation / Wavemength Graph
Fig 2.2: Typical attenuation graphs for multimode and single mode loss. Note that the loss is sensitive to wavelengthsFiber Optic Transmission WindowsFig 2.3: Fibre Optic Transmission Windows



The above graphs show typical loss when the glass is mechanically unstressed. Additional loss will be created under stress, and this is termed “microbend loss”. Microbend loss is typically caused by: kinking, tension, heat or cold, crushing, winding onto a drum, excessively sharp bending, defective manufacture, or poor cable design. The above graph also fails to show rapidly increasing attenuation above about 1625 nm.

Multimode loss measuring is inherently more uncertain than singlemode measuring, since there are additional variables as follows:

LED and Laser relative Spatial and Wavelength charactersitics

Fig 2.3: A comparison of the spectral and spatial distribution of Lasers and LEDs.

Single mode loss tends to be more stable than multimode, since multi-mode transmission characteristics create instantaneously changing loss characteristics. This is readily observed during practical measurement, where moving multimode patch leads around creates obvious measurement variations.

The loss budget refers to the calculated allowable loss before a link stops working. 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 is usually much less critical , so this is much easier, or it may be defined in appropriate standards.

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

Single mode loss is more sensitive to fiber bending or mechanical stress at wavelengths above 1480 nm. It is therefore common to perform additional measurements on single mode at 1550 nm, as a means of verifying the installed performance, even if the operational system is to be used at 1310 nm.

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.

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

Component Loss
Single mode 1310 nm 0.33 – 0.35 dB/km
1550 nm 0.17 -0.22 dB/km
Multimode 850 nm 2.5 – 3.0 dB/km
1310 nm 0.7 – 0.8 dB/km
Connectors 0.1 – 0.75 dB
Fusion splices 0.1 – 0.15 dB
Mechanical splices 0.1 – 1 dB
Switches 0.1 – 1.5 dB
Couplers 1 – 21 dB (power splitter)
Isolators 0.5 – 5 dB


These figures are very approximate: the exact value will depend on the equipment selected in your application. 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 by definition will often not be achieved. Connector loss is better understood by using a statistical mean and standard deviation approach.

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 LTS, or source and meter.

End to end loss cannot be reliably measured with optical time domain reflectometry, since an OTDR works by mathematical deduction based on a number of 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:

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:

Effect of Testing 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, eg 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.