The most important component affecting loss in a multimode cable plant is the source-coupling light into the fiber. Light sources may be light-emitting diodes (LEDs) or lasers. Lasers may be VCSELs (vertical cavity surface-emitting lasers) or Fabry-Perot lasers (telecom style). Each of these emits light in a different pattern, with LEDs having the broadest beam, Fabry-Perot lasers a very narrow beam, and VCSELs in between. The light coupledfrom the source is transmitted in a multi-mode fiber in many rays or "modes;"hence, the name multimode.
A laser couples light only into modes that travel near the center of the fiber while a LED couples light into practically all the modes. Look closely and you can see the modes near the center of the fiber core (lower-order modes) travel shorter paths than those near the edges of the core (higher-order modes.) The shorter path of the lower-order modes means that they travel through less glass and suffer less loss than the ones traveling in the outside of the core. That means a laser suffers less attenuation (loss per unit length, in dB/km) in the same multimode fiber than a LED.
Furthermore, as light travels down the fiber, the attenuation changes. The light in the outside modes is attenuated, leaving mostly light in the modes near the center. At a kilometer from a LED source, the light in the outer modes is mostly attenuated and the light carried in the fiber looks more like the light launched from a laser. This means the attenuation at that point is less than at the beginning because it's only in lower-order modes.
So, what is the loss of the fiber? The manufacturer's spec for fiber is around 3 dB/km at 850 nm and 1 dB/km at 1300 nm, for a test using a calibrated source that is much closer to the launch of a laser source than a LED. The difference in the attenuation coefficient of a fiber tested with a laser or LED can be 1 to 2 dB/km. With a LED source, the first hundred meters of fiber (representative of a premises network) may have an attenuation of over 4 dB/km.
The same factors hold for connector and spice loss. Most of the loss in connectors is due to misalignment of the two fibers, and the higher-order modes are much more likely to be lost at a connector than are lower-order modes. A connector coupled to a LED source with a short cable could have a loss of 0.5 dB, while if it were coupled to a laser source, or were 1 km away, could have a loss of 0.3 dB.
By now, I suspect your head is swimming. If you still have your wits about you, you may want to know how any standards body can solve this issue. The answer is how everything is solved: Compromise. Create a standard launch condition that is more than a laser but less than a LED, which today isappropriate, since it's more like the VCSELs used in today's gigabit and faster multimode links.
Manufacturers use special lensed sources in their labs that can precisely control the launch conditions. To approximate this launch for field testing, use a LED source and a mode modifier—usually a few turns of the reference launch cable around acylindrical mandrel that filters out thehigher-order modes. Choose the mandrel size according to the fiber and cable typebeing used. These devices are availablefrom many test equipment manufacturers.
This standard source method will produce more consistent test results and provide better reproducibility if you ever have to retest. And the losses measured are going to be lower, so you are less likely to fail good cables.
Even so, the uncertainty of the measurement is likely to be several tenths of a decibel. The uncertainty comes from the coupling of your reference cables to the fiber under test, which includes the quality of the terminations on the reference cables—how clean they are, and how many times they have been used—since they degrade with use.
Some people think everybody usesOTDRs for fiber-optic testing, but that's only for outside plant (OSP) applications. Most OSP installations involve splicing singlemode fiber to get longer runs, and the OTDR allows verifying the quality of the splice. But when that link is finished, it must still betested for insertion loss with a light source, power meter, and reference cables—just like premises cables.
Insertion loss and OTDR testing use different methods. Insertion loss tests just the fiber that will be used, with a source on one end and a detector on the other. So, tested insertion loss should be close to what the communications linkactually will see. OTDRs, however, make an indirectmeasurement, based on fiber scattering—the major source of loss of a fiber. It sends a very powerful pulse down thefiber, and some of the scattering comes back toward theinstrument where it is measured and stored. As the test pulsemoves down the fiber, it takes a "snapshot" of the fiberilluminated by the test pulse from which informationabout the fiber may be implied.