The ideal interconnection of one fiber to another would have two fibers that are optically and physically identical held by a connector or splice that squarely aligns them on their center axes. However, in the real world, system loss due to fiber interconnection is a factor. Insertion loss is the primary consideration for connector performance. There are three types of insertion loss: fiber-related loss, connector-related loss, and system factors that contribute to loss. Because of the discrepancy between insertion-loss testing and connector performance, users must understand the test methods used to measure insertion loss. The best test results are obtained when lengths of fibers are attached to the source and detector as permanent parts of the test setup. This avoids variations in results that are caused by source and detector interconnection losses from test to test.

There are several components required to test interconnection losses.

1. Light Source: Light sources include lasers, LEDs, broadband sources, or monochromators.

A) Lasers and light-emitting diodes (LEDs) are widely used as sources. Important characteristics include output power, speed, output pattern or numerical aperture (NA), spectral width, fiber-type compatibility, ease of use, lifetime, and cost. Figure 1 illustrates various methods for interfacing a light source to an optical fiber.

Figure 1 - Methods of Interfacing a Source to a Fiber

B) Broadband Source: Once popular but now seldom used, typical incandescent sources include quartz, halogen, or xenon arc lamps with interference filters. If possible, the filter should have a bandpass that approximates the output of the source to be used in the proposed system to better account for wavelength-dependent fiber characteristics such as NA, attenuation, and dispersion.

C) Monochromator: This device isolates narrow portions of light by dispersing light into its component wavelengths. Most commercial monochromators exhibit very low energy on the output side, and they select a very narrow bandwidth.

2. Mode Scramblers: Mode scramblers mix light to excite every possible mode of transmission within the fiber. The easiest to make is a 15-cm tube at least 7 cm in diameter filled with 1-mm lead shot through which the bare fiber passes. Another type uses a row of one-eighth inch diameter brass pins through which the fiber zigzags. The resulting bends in either type cause mode coupling that fills the fiber. A more complex scrambler is a butt-welded (fusion-spliced) length of alternating graded-index, step-index, graded-index fibers. The step-index fiber generally has a length of one meter. The discontinuities that result mix the light; however, butt-welded scramblers are difficult to fabricate and are weak, exhibiting less than 20% of the original fiber’s mechanical strength.

Figure 2 - Mode Scramblers

3. Core Mode Filter: Mandrel wrap core mode filters allow high-order mode signals from the core to be removed. High-order modes traveling through several hundred meters of fiber leak into the cladding and are lost. This results in an exit numerical aperture less than the material NA of the fiber. A fiber that has reached modal equilibrium, along with the reduced NA, is said to exhibit long-launch conditions.

Rather than testing connector loss over several hundred meters of fiber, core mode filters simulate this distance. The standard recommended core mode filter for smaller fibers is 12.5-mm diameter mandrel around which the fiber is wrapped five times under zero tensions. The mandrel wrap reduces the exit NA to about 50% of the fiber’s material NA. The mandrel wrap also reduces the light-emitting area of the core of a graded-index fiber by about 50%. This reduction in the emitting area affects the performance of a connector or a splice during loss measurements.

4. Cladding Mode Stripper: The use of the mandrel wrap described above scrambles the modes or strips the high-order modes. This stripped light has nowhere to go except the cladding. In short fiber runs or in setups where the mandrel wrap occurs at the end of the fiber, this light deflected to the cladding can be substantial. It is necessary to use to remove these modes, a cladding mode stripper, which incorporates a fiber, stripped of its cladding buffer, and covered in Corona Dope (available from TV repair suppliers) or some other liquid with a refractive index higher than the cladding. Corona Dope has advantages: it is low-cost, it has a high refractive index, and the coating is black. Figure 4 illustrates a cladding mode stripper.

Figure 4 - Cladding Mode Stripper

5. Detector System: Optical multimeters, also called optical power meters, read optical power levels. The meter is completely electronic with sensors that plug into the unit. Different sensors are available for use at different power levels and operating wavelengths. Adapters permit bare fibers or a variety of popular connectors to be connected to a sensor. A drawback of the multimeter is that in many applications both ends of the fiber must be available. An optical time-domain reflectometer allows testing when only one end of the fiber is available. This device relies on the backscattering of light that occurs in an optical fiber for detection.

Insertion loss tests will reduce the influence of fiber-related losses. A general test should be reproducible and provide applicable results. Most tests measure the output power (P₁) of a length of fiber. The fiber is then cut in the middle and terminated with a connector or splice. The output power (P₂) is measured again. Insertion loss is given by:

Loss (dB) = 10 • log₁₀ (P₁/P₂)

The length of fiber must be broken perfectly in the middle to produce an identical fiber on each side of the splice. This method purposely eliminates fiber-induced losses in order to evaluate connector performance independently of fiber-related variations. Three sets of launch conditions are of interest.

1. Short-launch, short-receive: Represented by short fibers with no mandrel wrap on the transmitting or receiving ends. Short-launch, short-receive conditions exhibit losses that increase with the slightest mechanical offset of the connection. Lateral misalignment is a critical parameter under these conditions.
2. Long-launch, short-receive: A mandrel wrap is on the transmitting end but not on the receiving end. This condition reduces the exit numerical aperture of the transmitting fiber, and end-separation losses are smaller. Since all of the receiving core can be used, greater separation of the fibers can be tolerated.
3. Long-launch, long-receive: This condition is created by using a mandrel wrap at both the transmit and receive end and shows greater sensitivity to lateral misalignment than the other two conditions. Because the effective core area of both fibers is reduced, any offset increases loss more significantly.

The insertion loss test assumed that two pieces of identical fiber were used. However, if two different types of fibers are connected, then NA mismatch loss and diameter mismatch loss must be accounted for.

NA mismatch loss occurs when the numerical aperture of the transmitting fiber (t) is larger than that of the receiving fiber (r). NA mismatch loss is illustrated in Figure 5.

Figure 5 - NA-Mismatch

The calculated loss for numerical aperture mismatch is approximated by:

Loss_NA = 10 • log₁₀(NA_r/NA_t)

As illustrated in Figure 6, core diameter mismatch occurs when the core diameter of the transmitting fiber (t) is larger than the core diameter of the fiber at the receiving end (r). Cladding diameter mismatch is similar to core diameter mismatch loss except the cladding of the transmitting fiber differs in diameter from the cladding of the receiving fiber. Either mismatch prevents the cores from aligning.

Figure 6 - Core-Diameter Mismatch

Both types of diameter mismatch loss are approximated by:

Loss_dia = 10 • log₁₀ (dia_r/dia_t)²

This equation is only accurate if all of the modes in the fiber are excited. When only low-order modes are excited, the loss is greatly reduced and may not be present at all.

Concentricity occurs because the core may not be perfectly centered in the cladding. Ellipticity or ovality describes the fact that the core or cladding may be elliptical rather than circular. The alignment of the two elliptical cores will vary depending on how the fibers are brought together. These forms of connector loss are illustrated in Figure 7.

Figure 7 - Concentricity and Ellipticity

Connector-related loss can also result from the mechanical misalignment of the optical fiber cores. There are several types of misalignment loss: lateral displacement, angular misalignment, and end separation.

A connector should align the fibers on their center axes, but when one fiber’s axis does not coincide with the other fiber’s axis, lateral displacement occurs. A displacement of only 10% of the core axis diameter results in a loss of about 0.5 dB. The ends of mated fibers should be perpendicular to the fibers’ axes and to each other. Failure to be perpendicular is called angular misalignment. Figure 8 illustrates angular misalignment, and Figure 9 illustrates lateral misalignment.

Figure 8 - Angular Misalignment

Figure 9 - Lateral Misalignment

Some connectors hold the two fibers slightly apart to prevent the fibers from rubbing against each other and damaging their end polishes. Fresnel reflection loss or end separation loss is caused by the difference in the refractive indices of the two fibers and the air that fills the gap between the two fibers. Some connector manufacturers believe the use of index-matching gel in the gap reduces Fresnel reflection loss, but others do not recommend using index-matching gel. This gap may collect small flecks of abrasive contaminates that will damage the end finishes, and the addition of index-matching gel could compound this contamination.

In a single-mode interconnection with a flat end finish, Fresnel reflection loss can be as much as -11 dB, a level sufficient to disrupt the operation of most lasers. This loss can be reduced by rounding the fiber end of one fiber during polishing (called a PC or physical contact finish). While it would seem practical to use a flat finish and butt the ends, getting two perfectly smooth, flat finishes is nearly impossible. With a rounded finish, fibers always touch on a high point near the light-carrying fiber core.

Figure 10 - Fiber End Face Finishes

System-related factors in connector loss involve the launch and receive conditions. These conditions result from the mode distribution in the fibers. The performance of the connector depends on modal conditions and the connector’s position in the system. 

These launch conditions must be controlled in order to provide repeatable measurements. Long-launch conditions are generally preferred. Long-launch or receive conditions mean that equilibrium mode distribution (EMD), illustrated in Figure 11, exists in the fiber. The Electronic Industry Association (EIA) recommends a 70/70 launch: 70% of the fiber core diameter and 70% of the fiber NA should be filled. This recommendation corresponds to the EMD in a graded-index fiber. EMD can be reached three ways: by the optical approach, filtering, or long fiber length. In general, connector losses under long-launch conditions range from 0.4-0.5 dB. Under short-launch conditions, losses are in the 1.3-1.4 dB range.

Figure 11 - Equilibrium Mode Distribution