Although the nonlinear effects in optical fibers are small, they accumulate as light passes
through many kilometers of single-mode fiber. The small core diameters, together with
the long transmission distances that may be obtained with these fibers, have enabled the
occurrence of nonlinear phenomena at power levels of a few milliwatts which are well
within the capability of semiconductor lasers. Furthermore, the optical power levels
become much larger when wavelength division multiplexing packs many signal
channels into one single-mode fiber such that the overall power level is the summation
of the individual channel optical powers.
There are two broad categories of nonlinear effects that can be separated based on their
characteristics: namely, scattering and Kerr effects.
Fiber alignment and joint loss
A major consideration with all types of fiber–fiber connection is the optical loss
encountered at the interface. Even when the two jointed fiber ends are smooth and
perpendicular to the fiber axes, and the two fiber axes are perfectly aligned, a small
proportion of the light may be reflected back into the transmitting fiber causing
attenuation at the joint. This phenomenon, known as Fresnel reflection, is associated
with the step changes in refractive index at the jointed interface (i.e. glass–air–glass).
The magnitude of this partial reflection of the light transmitted through the interface
may be estimated using the classical Fresnel formula for light of normal incidence and is
given by
where r is the fraction of the light reflected at a single interface, n1 is the refractive
index of the fiber core and n is the refractive index of the medium between the two
jointed fibers (i.e. for air n =1). However, in order to determine the amount of light
reflected at a fiber joint, Fresnel reflection at both fiber interfaces must be taken into
account. The loss in decibels due to Fresnel reflection at a single interface is given by:
Hence, using the relationships given in Eqs (2.59) and (2.60) it is possible to determine
the optical attenuation due to Fresnel reflection at a fiber–fiber joint.
It is apparent that Fresnel reflection may give a significant loss at a fiber joint even when
all other aspects of the connection are ideal. However, the effect of Fresnel reflection at
a fiber–fiber connection can be reduced to a very low level through the use of an index-
matching fluid in the gap between the jointed fibers. When the index-matching fluid has
the same refractive index as the fiber core, losses due to Fresnel reflection are in theory
eradicated. Unfortunately, Fresnel reflection is only one possible source of optical loss at
a fiber joint. A potentially greater source of loss at a fiber–fiber connection is caused by
misalignment of the two jointed fibers. In order to appreciate the development and
relative success of various connection techniques it is useful to discuss fiber alignment
in greater detail.
Any deviations in the geometrical and optical parameters of the two optical fibers which
are jointed will affect the optical attenuation (insertion loss) through the connection. It is
not possible within any particular connection technique to allow for all these variations.
Hence, there are inherent connection problems when jointing fibers with, for instance
different core and/or cladding diameters;
different numerical apertures and/or relative refractive index differences;
different refractive index profiles;
fiber faults (core ellipticity, core concentricity, etc.).
The losses caused by the above factors together with those of Fresnel reflection are
usually referred to as intrinsic joint losses. The best results are therefore achieved with
compatible (same) fibers which are manufactured to the lowest tolerance.
In this case there is still the problem of the quality of the fiber alignment provided by the
jointing mechanism. Examples of possible misalignment between coupled compatible
optical fibers are illustrated in Figure 2.17. It is apparent that misalignment may occur in
three dimensions: the separation between the fibers (longitudinal misalignment), the
offset perpendicular to the fiber core axes (lateral/radial/ axial misalignment) and the
angle between the core axes (angular misalignment).
Optical losses resulting from these three types of misalignment depend upon the fiber
type, core diameter and the distribution of the optical power between the propagating
modes. Examples of the measured optical losses due to the various types of
misalignment are shown in Figure 2.18. Figure 2.18(a) shows the attenuation
characteristic for both longitudinal and lateral misalignment of a graded index fiber of
50 µm core diameter.
It may be observed that the lateral misalignment gives significantly greater losses per
unit displacement than the longitudinal misalignment. For instance, in this case a lateral
displacement of 10 µm gives about 1 dB insertion loss whereas a similar longitudinal
displacement gives an insertion loss of around 0.1 dB. Figure 2.18(b) shows the
attenuation characteristic for the angular misalignment of two multimode step index
fibers with numerical apertures of 0.22 and 0.3. An insertion loss of around 1 dB is
obtained with angular misalignment of 4° and 5° for the NA =0.22 and NA=0.3 fibers
respectively.
It may also be observed in Figure 2.18(b) that the effect of an index-matching fluid in
the fiber gap causes increased losses with angular misalignment. Therefore, it is clear
that relatively small levels of lateral and/or angular misalignment can cause significant
attenuation at a fiber joint. This is especially the case for fibers of small core diameter
(less than 150 µm) which are currently employed for most telecommunication purposes.
1. Multimode fiber joints
Theoretical and experimental studies of fiber misalignment in optical fiber connections
allow approximate determination of the losses encountered with the various
misalignments of different fiber types. We consider here some of the expressions used to
calculate losses due to lateral and angular misalignment of optical fiber joints.
Longitudinal misalignment is not discussed in detail as it tends to be the least important
effect and may be largely avoided in fiber connection.
Both groups of workers claim good agreement with experimental results, which is
perhaps understandable when considering the number of variables involved in the
measurement. Also, all groups predict higher losses for fibers with larger numerical
apertures, which is consistent with intuitive considerations (i.e. the larger the numerical
aperture, the greater the spread of the output light and the higher the optical loss at a
longitudinally misaligned joint).
Theoretical expressions for the determination of lateral and angular misalignment losses
are by no means definitive, although in all cases they claim reasonable agreement with
experimental results. However, experimental results from different sources tend to vary
(especially for angular misalignment losses) due to difficulties of measurement. It is
therefore not implied that the expressions given in the text are necessarily the most
accurate, as at present the choice appears somewhat arbitrary. Lateral misalignment
reduces the overlap region between the two fiber cores. Assuming uniform excitation of
all the optical modes in a multimode step index fiber, the overlapped area between both
fiber cores approximately gives the lateral coupling efficiency µlat. Hence, the lateral
coupling efficiency for two similar step index fibers may be written as
where n1 is the core refractive index, n is the refractive index of the medium between
the fibers, y is the lateral offset of the fiber core axes, and a is the fiber core radius. The
lateral misalignment loss in decibels may be determined using:
The predicted losses obtained using the formulas given in Eqs (2.61) and (2.62) are
generally slightly higher than the measured values due to the assumption that all modes
are equally excited. This assumption is only correct for certain cases of optical fiber
transmission. Also, certain authors assume index matching and hence no Fresnel
reflection, which makes the first term in Eq. (2.61) equal to unity (as n1/n = 1). This
may be valid if the two fiber ends are assumed to be in close contact (i.e. no air gap in
between) and gives lower predicted losses. Nevertheless, bearing in mind these possible
inconsistencies, useful estimates for the attenuation due to lateral misalignment of
multimode step index fibers may be obtained.
Lateral misalignment loss in multimode graded index fibers assuming a uniform
distribution of optical power throughout all guided modes was calculated by Gloge. He
estimated that the lateral misalignment loss was dependent on the refractive index
gradient α for small lateral offset and may be obtained from:
A further estimate including the leaky modes gave a revised expression for the lateral
misalignment loss given in Eq. (2.64) of 0.75(y/a). This analysis was also extended to
step index fibers (where α = ∞) and gave lateral misalignment losses of 0.64(y/a) and
0.5(y/a) for the cases of guided modes only and both guided plus leaky modes
respectively.
Factors causing fiber–fiber intrinsic losses were listed in previous Section; the major
ones comprising a mismatch in the fiber core diameters, a mismatch in the fiber
numerical apertures and differing fiber refractive index profiles are illustrated in Figure
2.19. Connections between multimode fibers with certain of these parameters being
different can be quite common, particularly when a pigtailed optical source is used, the
fiber pigtail of which has different characteristics from the main transmission fiber.
Moreover, as indicated previously, diameter variations can occur with the same fiber
type.
Assuming all the modes are equally excited in a multimode step or graded index fiber,
and that the numerical apertures and index profiles are the same, then the loss resulting
from a mismatch of core diameters (see Figure 2.19(a)) is given by:
where a1 and a2 are the core radii of the transmitting and receiving fibers respectively. It
may be observed from Eq. (2.66) that no loss is incurred if the receiving fiber has a
larger core diameter than the transmitting one. In addition, only a relatively small loss
(0.09 dB) is obtained when the receiving fiber core diameter is 1% smaller than that of
the transmitting fiber.
When the transmitting fiber has a higher numerical aperture than the receiving fiber,
then some of the emitted light rays will fall outside the acceptance angle of the receiving
fiber and they will therefore not be coupled through the joint. Again assuming a uniform
modal power distribution, and fibers with equivalent refractive index profiles and core
diameters, then the loss caused by a mismatch of numerical apertures (see Figure
2.19(b))
2. Single-mode fiber joints
Misalignment losses at connections in single-mode fibers have been theoretically
considered by Marcuse and Gambling et al. The theoretical analysis which was
instigated by Marcuse is based upon the Gaussian or near-Gaussian shape of the modes
propagating in single-mode fibers regardless of the fiber type (i.e. step index or graded
index). Further development of this theory by Gambling etal. gave simplified formulas
for both the lateral and angular misalignment losses at joints in single mode fibers. In the
absence of angular misalignment Gambling et al. calculated that the loss Tl due to lateral
offset y was given by:
where ω is the normalized spot size of the fundamental mode.* However, the normalized
spot size for the LP01 mode (which corresponds to the HE mode) may be obtained from
the empirical formula:
where is the spot size in μm, a is the fiber core radius and V is the normalized
frequency for the fiber. Alternatively, the insertion loss Ta caused by an angular
misalignment (in radians) at a joint in a single-mode fiber may be given by
wheren1 is the fiber core refractive index and NA is the numerical aperture of the fiber.
It must be noted that the formulas given in Eqs (2.69) and (2.70) assume that the spot
sizes of the modes in the two coupled fibers are the same. Gambling et al. also derived a
somewhat complicated formula which gave a good approximation for the combined
losses due to both lateral and angular misalignment at a fiber joint. However, they
indicate that for small total losses (less than 0.75 dB) a reasonable approximation is
obtained by simply combining Eqs (2.68) and (2.70).
Assuming that no losses are present due to the extrinsic factors, the intrinsic coupling
loss is given by
where 01 and 02 are the spot sizes of the transmitting and receiving fibers
respectively. Equation (2.71) therefore enables the additional coupling loss resulting
from mode-field diameter mismatch between two single-mode fibers to be calculated.
Fiber splices
A permanent joint formed between two individual optical fibers in the field or factory is
known as a fiber splice. Fiber splicing is frequently used to establish long-haul optical
fiber links where smaller fiber lengths need to be joined, and there is no requirement for
repeated connection and disconnection. Splices may be divided into two broad
categories depending upon the splicing technique utilized. These are fusion splicing or
welding and mechanical splicing.
Fusion splicing is accomplished by applying localized heating (e.g. by a flame or an
electric arc) at the interface between two butted, prealigned fiber ends causing them to
soften and fuse. Mechanical splicing, in which the fibers are held in alignment by some
mechanical means, may be achieved by various methods including the use of tubes
around the fiber ends (tube splices) or V-grooves into which the butted fibers are placed
(groove splices). All these techniques seek to optimize the splice performance (i.e.
reduce the insertion loss at the joint) through both fiber end preparation and alignment of
the two joint fibers. Typical average splice insertion losses for multimode fibers are in
the range 0.1 to 0.2 dB which is generally a better performance than that exhibited by
demountable connections.
It may be noted that the insertion losses of fiber splices are generally much less than the
possible Fresnel reflection loss at a butted fiber–fiber joint. This is because there is no
large step change in refractive index with the fusion splice as it forms a continuous fiber
connection, and some method of index matching (e.g. a fluid) tends to be utilized with
mechanical splices
A requirement with fibers intended for splicing is that they have smooth and square end
faces. In general this end preparation may be achieved using a suitable tool which
cleaves the fiber as illustrated in Figure 2.20. This process is often referred to as scribe
and break or score and break as it involves the scoring of the fiber surface under tension
with a cutting tool (e.g. sapphire, diamond, tungsten carbide blade). The surface scoring
creates failure as the fiber is tensioned and a clean, reasonably square fiber end can be
produced.
Figure 2.20 illustrates this process with the fiber tensioned around a curved mandrel.
However, straight pull, scribe and break tools are also utilized, which arguably give
better results.
1. Fusion splices
The fusion splicing of single fibers involves the heating of the two prepared fiber ends to
their fusing point with the application of sufficient axial pressure between the two
optical fibers. It is therefore essential that the stripped (of cabling and buffer coating)
fiber ends are adequately positioned and aligned in order to achieve good continuity of
the transmission medium at the junction point. Hence the fibers are usually positioned
and clamped with the aid of an inspection microscope.
Flame heating sources such as microplasma torches (argon and hydrogen) and
oxhydricmicroburners (oxygen, hydrogen and alcohol vapor) have been utilized with
some success. However, the most widely used heating source is an electric arc. This
technique offers advantages of consistent, easily controlled heat with adaptability for use
under field conditions.
A schematic diagram of the basic arc fusion method is given in Figure 2.21(a)
illustrating how the two fibers are welded together. Figure 2.21(b) shows a development
of the basic arc fusion process which involves the rounding of the fiber ends with a low-
energy discharge before pressing the fibers together and fusing with a stronger arc.
This technique, known as prefusion, removes the requirement for fiber end preparation
which has a distinct advantage in the field environment. It has been utilized with
multimode fibers giving average splice losses of 0.09 db.
Fusion splicing of single-mode fibers with typical core diameters between 5 and 10 µm
presents problems of more critical fiber alignment (i.e. lateral offsets of less than 1 µm
are required for low loss joints). However, splice insertion losses below 0.3 dB may be
achieved due to a self-alignment phenomenon which partially compensates for any
lateral offset.
Self-alignment, illustrated in Figure 2.22, is caused by surface tension effects between
the two fiber ends during fusing. An early field trial of single-mode fiber fusion splicing
over a 31.6 km link gave mean splice insertion losses of 0.18 and 0.12 dB at
wavelengths of 1.3 and 1.55 µm respectively. Mean splice losses of only 0.06 dB have
also been obtained with a fully automatic single-mode fiber fusion splicing machine
weaken the fiber in the vicinity of the splice. It has been found that even with careful
handling, the tensile strength of the fused fiber may be as low as 30% of that of the
uncoated fiber before fusion.
The fiber fracture generally occurs in the heat affected zone adjacent to the fused joint.
The reduced tensile strength is attributed to the combined effects of surface damage
caused by handling, surface defect growth during heating and induced residential
stresses due to changes in chemical composition. It is therefore necessary that the
completed splice is packaged so as to reduce tensile loading upon the fiber in the
vicinity of the splice.
2. Mechanical splices
A number of mechanical techniques for splicing individual optical fibers have been
developed. A common method involves the use of an accurately produced rigid
alignment tube into which the prepared fiber ends are permanently bonded. This snug
tube splice is illustrated in Figure 2.23(a) and may utilize a glass or ceramic capillary
with an inner diameter just large enough to accept the optical fibers. Transparent
adhesive (e.g. epoxy resin) is injected through a transverse bore in the capillary to give
mechanical sealing and index matching of the splice. Average insertion losses as low as
0.1 dB have been obtained with multimode graded index and single-mode fibers using
ceramic capillaries. However, in general, snug tube splices exhibit problems with
capillary tolerance requirements. Hence as a commercial product they may exhibit losses
of up to 0.5 dB.
Mechanical splicing technique which avoids the critical tolerance requirements of the
snug tube splice is shown in Figure 2.23(b). This loose tube splice uses an oversized
square-section metal tube which easily accepts the prepared fiber ends. Transparent
adhesive is first inserted into the tube followed by the fibers. The splice is self-aligning
when the fibers are curved in the same plane, forcing the fiber ends simultaneously into
the same corner of the tube, as indicated in Figure 2.23(b). Mean splice insertion losses
of 0.073 dB have been achieved using multimode graded index fibers with the loose
tube approach.
Other common mechanical splicing techniques involve the use of grooves to secure the
fibers to be jointed. A simple method utilizes a V-groove into which the two prepared
fiber ends are pressed. The V-groove splice which is illustrated in Figure 2.24(a) gives
alignment of the prepared fiber ends through insertion in the groove. The splice is made
permanent by securing the fibers in the V-groove with epoxy resin. Jigs for producing
Vgroove splices have proved quite successful, giving joint insertion losses of around 0.1
dB
V-groove splices formed by sandwiching the butted fiber ends between a V-groove glass
substrate and a flat glass retainer plate, as shown in Figure 2.24(b), have also proved
very successful in the laboratory. Splice insertion losses of less than 0.01 dB when
coupling single-mode fibers have been reported using this technique. However,
reservations are expressed regarding the field implementation of these splices with
respect to manufactured fiber geometry, and housing of the splice in order to avoid
additional losses due to local fiber bending.
A further variant on the V-groove technique is the elastic tube or elastomeric splice
shown in Figure 2.25. The device comprises two elastomeric internal parts, one of which
contains a V-groove. An outer sleeve holds the two elastic parts in compression to ensure
alignment of the fibers in the V-groove, and fibers with different diameters tend to be
centered and hence may be successfully spliced. Although originally intended for
multimode fiber connection, the device has become a widely used commercial product
which is employed with single-mode fibers, albeit often as a temporary splice for
laboratory investigations. The splice loss for the elastic tube device was originally
reported as 0.12 dB or less but is generally specified as around 0.25 dB for the
commercial product. In addition, index-matching gel is normally employed within the
device to improve its performance.
A slightly more complex groove splice known as the Springroove® splice utilized a
bracket containing two cylindrical pins which serve as an alignment guide for the two
prepared fiber ends. The cylindrical pin diameter was chosen to allow the fibers to
protrude above the cylinders, as shown in Figure 2.26(a). An elastic element (a spring)
was used to press the fibers into a groove and maintain the fiber end alignment, as
illustrated in Figure 2.26(b). The complete assembly was secured using a drop of epoxy
resin. Mean splice insertion losses of 0.05 dB were obtained using multimode graded
index fibers with the Springroove splice. This device found practical use in Italy.
An example of a secondary aligned mechanical splice for multimode fiber is shown in
Figure 2.27. This device uses precision glass capillary tubes called ferrules as the
secondary elements with an alignment sleeve of metal or plastic into which the glass
tubed fibers are inserted. Normal assembly of the splice using 50 µm core diameter fiber
yields an average loss of around 0.2 dB.
