Fiber_Optics_Physics_Technology

right in the relevant range there are oscillations in the curve so that a clear reading is not possible. This is caused by the so-called whispering galley modes. These are modes that can propagate in a curved fiber in the cladding; this involves reflections at the outside surface of the fiber. To safeguard against them, in e ect one removes the outside surface by stripping the plastic coating and placing the bare fiber in index-matching gel. When the indices are indeed well matched, cladding light will exit from the fiber after a very short distance and the problem is solved [161].

7.6Optical Time Domain Reflectometry (OTDR)

Fiber technology has given rise to a special tool that can be used to easily assess many properties of fibers, both in the lab and in the field. It is called optical time domain reflectometry or OTDR (Fig. 7.17). It is very similar in spirit to radar: A signal is launched into the fiber; whatever light is reflected or scattered back is collected and evaluated. Pulsed laser diodes are employed as light sources and photodiodes to detect the backscattered light.

The time until an echo is registered is calculated from

where n is the e ective index for the mode and L is the length. The factor of 2 arises because light must travel forth and back before it is registered. Echo strength provides information about the type of condition that causes the echo:

OTDR

splice

detector

Fiber length (km)

Figure 7.17: Optical time domain reflectometry (OTDR). Top: Setup. A light pulse is launched into the fiber under test; the reflected light is recorded as a function of time. Time can be converted to position in the fiber. Bottom: The obtained data, shown here schematically, provide information about various fiber conditions.

Rayleigh scattering gives a continuous background that gently goes down with increasing distance; localized conditions like fiber joints or breakage give sharp peaks.

There is always a certain crosstalk of transmitter light to the receiver, so that the receiver is overloaded for a short initial moment. This creates a dead zone in the short range. However, some devices are constructed to minimize the dead zone and measure even on the shortest distances (millimeters, in some cases even less).

OTDR equipment is o ered by several manufacturers and allows to assess a fiber over many kilometers with access only to one end. This makes OTDR a valuable tool for a wide range of tasks, notably to analyze

fiber loss and its spatial allocation;

loss at fiber joints like connectors or splices;

loss at other localized conditions, e.g., sharp bends or damage;

the location of each of these conditions;

fiber length; and

fiber end reflection.

In commercial installations OTDR devices are therefore indispensable in spite of their cost. Some manufacturers o er plug-in cards for computers with complete OTDR hardware; this reduces the cost because the computer does both the number crunching and the displaying.

the company was split into a central segment and several regional operating companies. Right after divestiture there were not as many cables as telephone service providers so that sometimes the same fiber in the same cable was used in time-sharing agreements by several competitors. Maybe that is why cables with 96 fibers were then laid.

A couple of years later, loss of 0.4 dB/km, repeater distances of 40 km, and data rates of 2 Gb/s became routine. This corresponds to 1,500,000 simultaneous telephone calls. See Chap. 11 for methods to put many calls onto the same fiber without mutual interference and Chap. 11.4 for further development.

When a cable incorporating optical fibers is manufactured, there are a couple of things to observe. Fibers must be protected from adverse environmental influences. In the interest of a long lifetime of the cable, fibers must not experience tensile load even while the cable is bent and pulled. Also, both macroand micro-bend losses must be avoided in the deployed fiber. Several cable designs are in use to meet these objectives; Fig. 8.1 shows examples. There is always a strength member to take care of the tensile load; it may me made of fiberglass, Kevlar fiber, or steel wire. (Fiberglass is what the poles for pole vault are often made of; Kevlar is the fiber used for bulletproof vests.) Typically, fibers are individually placed in tiny tubes where they have some slack and can accommodate some extra length. If the cable is then pulled, the stress is kept away from the fibers. The tubes are filled with a gel which prevents the intrusion of water; it also damps vibrations and movement of the fiber. Sometimes a group of fibers sits in a common, slightly larger tube, again filled with gel. There are also “ribbon” constructions where several fibers are connected in a flat side-by- side structure similar to an electric flat ribbon cable. Ribbons allow to make connections of several fibers e ciently by automated machinery. All fibers in a ribbon can be spliced to another ribbon in one go, rather than handling each fiber individually.

Figure 8.1: Schematic cross-section of di erent cable types. Left : A single fiber sits loosely in a structure which is stabilized by fiberglass and Kevlar. Center : Several fibers are placed around a central steel wire acting as strength member. Right : Several fibers are combined into ribbons. Shown is a cable with several such ribbons; the structure is stabilized by steel wires. From [18].

There are several options for laying the cables. On long distances, they are dug into the ground, and in cities they are placed in ducts. In some countries including the USA, the cheap method is preferred in rural and suburban areas: the cables are suspended from utility poles. This, of course, is susceptible to interruptions.

The most frequent sources of damage are by humans (digging, vandalism) and natural causes such as lightning strokes and – down to 2 m below ground – rodents. In the USA, damage by gunshot occurs. Sometimes deployed fibers are subject to temperature extremes: For suspended fibers on poles, one calculates with −25◦C to +65◦C for most of the continental USA; in some areas, one has to design for −40◦C to +75◦C. In the ground this range is limited to 0◦C to +30◦C. In this one respect, undersea cables are in a most benign environment: On the sea floor the temperature is quite constant around 10◦C.

8.2Preparation of Fiber Ends

Before fibers can be used for anything at all, first the fiber end faces must be prepared (Fig. 8.2). It is mandatory that the end face, after the fiber has been cut or cleaved, is perfectly smooth and of optical quality. This is not possible by bending the fiber till it breaks, or by cutting it with scissors. The simplest way for controlled fracture is to scratch the fiber surface manually with a diamond, a tungsten carbide blade, or some other extremely hard material, and then to apply mechanical tension. With some routine one can obtain reasonably good surfaces most of the time: The reliability falls short of 100% but in a pinch may be acceptable, but it is a good idea to check the fiber end with a microscope.

Figure 8.2: Fiber end faces. Left : Here an edge remains. Center : An irregular surface called a hackle zone. Either is a sign of a bad preparation. Right : A good preparation results in a face smooth as a mirror.

It is much better to use specialized equipment; the cost lies anywhere between a few hundred and several thousand euros or dollars. Fiber-breaking devices apply a well-defined longitudinal tension to the fiber while scoring it with a blade which may vibrate at ultrasonic frequency. This results in end faces which are perpendicular to the fiber axis within close tolerances and are smooth every time.

When fibers are inserted in connectors, it is important that the front face is in the same plane as the connector front. If the fiber sticks out, it will su er from damage; if it is recessed, there will be no good match to the other fiber.

Figure 8.3: A typical fiber connector. At the center of the ferrule, one can see the fiber either as a dark or a bright spot, depending on lighting conditions.

be connected are a given, the loss from [i] is unavoidable, while the loss from [ii]–[iv] arises from lack of precision in the connection and can be minimized.

In case of actual physical contact of both fibers the contribution from [v] would vanish, but such contact is problematic because abrasion might damage

Figure 8.4: Theoretical coupling loss between two fibers, after [96]. Shown is the expected transmissivity (Fresnel loss not considered) if (a) there are unequal mode-field radii, (b) there is transverse o set, (c) there is a gap, and (d) there is an angular misalignment. A mode-field radius of a = 5 μm, a cladding refractive index nM = 1.46, and a wavelength λ = 1.5 μm are assumed.

the fibers in repeated operation. Therefore Fresnel losses are usually accepted. The reflection at an interface between a medium with index n1 and a medium with index n2 for perpendicular incidence is given by

where r is the reflectivity for the field amplitude, i.e., the the reflected amplitude normalized to that of the incident wave. R = r2 is the reflected power fraction. For fused silica in the visible and near infrared with n ≈ 1.46, one finds r = 0.19 and R = 3.5%. In a connection between two fibers not in physical contact, we consider two such interfaces: fiber–air-fiber. Na¨ıvely one may expect twice the loss from an individual air–glass interface or 7%. Unfortunately the situation is slightly more complicated than that.

In the case of coherent light the loss may be more or less than 7% because both reflections may add in phase or in opposite phase. Both reflecting surfaces are nearly parallel, and light can bounce back and forth between them. Depending on the gap width-to-wavelength ratio, a resonance condition may been fulfilled (round trip path equals integer multiple of wavelength). The total reflection can vary accordingly between zero and four times the individual reflection or 14%. In e ect, one has a Fabry–Perot interferometer (see Fig. 8.5). If the light is not perfectly coherent and the gap is wider than the coherence

d

Laser

LED

White light

Figure 8.5: Depending on the degree of coherence of the light, there can be more or less obvious Fabry–Perot resonances in the coupling e ciency as the gap width between fibers is varied. The coherence length of laser light always exceeds the gap width. In the case of luminescent diodes (LEDs), the coherence length is often just a couple of wavelengths; the resonances then quickly decay as the gap width is increased. For white light, e.g., from a tungsten filament light bulb, the coherence length is on the order of one central wavelength, and no oscillations of the coupling e ciency are observed. If the fibers are brought into physical contact (gap width zero), Fresnel loss vanishes altogether.

length, resonances are washed out and eventually the na¨ıvely expected value is approached. The coherence length of laser light by far exceeds all reasonable gap widths, and interference needs to be fully taken into account. LEDs have limited coherence length, and only a few resonances occur. White light would avoid resonances but is not what one usually deals with.

If two polarization-maintaining fibers are to be joined, there is the additional requirement that the orientation of the birefringent axes must match (see Chap. 4.6.2). There are dedicated versions of connectors which have a special locking pin so that they always lock at the desired angular orientation and cannot rotate.

8.3.2Permanent Connections

Permanent connections are known as splices; the expression comes from sailor’s language where it denotes a way to join two ropes by unravelling the strands, then twisting them together. Fiber splices can be made either by gluing or by fusing. Gluing is a low cost technique; fusion is more durable and has lower and more reproducible loss.

For gluing, both fibers are inserted in some tight guiding tube, which provides some centering of the fibers with respect to each other. One can manually move the fibers somewhat and can try to find the optimum position of lowest loss.

The tube is filled with a transparent fluid adhesive which cures under ultraviolet light. As soon as the desired position is found, one turns on an ultraviolet lamp and hopes that the positions are kept until the adhesive sets. Loss of

0.3dB can be obtained with some routine, and with luck, even better than that. The professional procedure is to fuse the fibers. This involves heating the

glass until it softens. As heat sources various options have been tried, including microscopic gas flames. However, it is now standard to use an electric arc; it has the advantage of being easily controlled by a computer.

Figure 8.6 shows how the splicing procedure goes about. Both fibers are positioned and moved closely together. Then during the so-called premelting a very weak arc discharge, not hot enough to soften the glass, is applied, often with a slight increase of the gap width. Premelting serves to remove possible dirt from the fiber tips. Next is the fusing process proper: Microprocessors control the precise amount of discharge current and arc duration to obtain the best possible result. While the arc is on, the fibers are advanced toward each other, actually beyond the zero position so that they are slightly pushed into each other.

The optical loss in a splice can be discussed in close analogy to that of a connector [96] (see Fig. 8.4); of course, there is no air gap. Transverse o set is also not a major problem because when the fiber tips are molten, surface tension moves the fibers into that position where their outsides connect smoothly. As long as the cores are centered well in the fiber, this automatically means a minimal transverse o set. Fibers usually are well-centered these days.

When two fibers with the same mode profile, i.e., fibers of the same type, are joined, one can obtain losses well below 0.1 dB and with the fanciest fusion splicers down to 0.02 dB. As soon as dissimilar fibers are joined, the mode mismatch creates an additional loss. For multimode fibers, the situation is more complicated because the mode partition is modified; for detail see [100].

Figure 8.6: Schematic representation of splicing: Fibers are positioned in three axes. A premelting (also called prefusing) cleans the fiber tips, then the fibers are fused. Afterward, a good splice is nearly invisible.

8.4Elements for Spectral Manipulation

8.4.1Fabry–Perot Filters

Selective filters can be produced in fiber technology [97]. Figure 8.7 shows an all-fiber Fabry–Perot interferometer which uses partially mirrored end faces. Tuning is accomplished through tiny adjustment of the gap width between the mirrors by means of a piezoceramic transducer.

8.4.2Fiber–Bragg Structures

A very di erent type of in-fiber filters is increasingly used: so-called fiber-Bragg gratings. The underlying idea stems from the observation that a germaniumdoped fiber core can su er lasting changes of its refractive index after irradiation with ultraviolet light. This e ect is cultivated in the following way: The beam of an UV laser is split; both parts are then superimposed at a certain angle. Where they cross there are interference fringes with a certain spatial period given by both the wavelength and the crossing angle. The period can therefore be precisely controlled. The fiber to be treated is positioned in this crossing area. After a certain exposure time, there is a periodic modulation of the core index which can act as a Bragg grating. Depending on the length of the treated zone which may range from millimeters to a few centimeters, one can obtain narrowband or wideband filters with reflectivities at the center wavelength very close to 100%. This is why such Bragg filters can even be used as selective end