OPTICAL COMMUNICATION

UNIT-1
DISADVANTAGES
PROPAGATION WITHIN THE FIBER Principle of operation TOTAL INTERNAL REFLECTION
(TIR).
Diffraction refers to various phenomena which occur when a wave encounters an obstacle. Italian
scientist Francesco Maria Grimaldi coined the word "diffraction" and was the first to record accurate
observations of the phenomenon in 1665.[2][3] In classical physics, the diffraction phenomenon is
described as the apparent bending of waves around small obstacles and the spreading out of waves past
small openings. Similar effects occur when light waves travel through a medium with a varying refractive
index or a sound wave through one with varying acoustic impedance. Diffraction occurs with all waves,
including sound waves, water waves, and electromagnetic waves such as visible light, x-rays and radio
waves. As physical objects have wave-like properties (at the atomic level), diffraction also occurs with
matter and can be studied according to the principles of quantum mechanics.
Richard Feynman [4] says that
"no-one has ever been able to define the difference between interference and diffraction satisfactorily. It
is just a question of usage, and there is no specific, important physical difference between them."
He suggests that when there are only a few sources, say two, we call it interference, as in Young's slits,
but with a large number of sources, the process is labelled diffraction.
While diffraction occurs whenever propagating waves encounter such changes, its effects are generally
most pronounced for waves where the wavelength is roughly similar to the dimensions of the diffracting
objects. If the obstructing object provides multiple, closely-spaced openings, a complex pattern of
varying intensity can result. This is due to the superposition, or interference, of different parts of a wave
that traveled to the observer by different paths (see diffraction grating).
The formalism of diffraction can also describe the way in which waves of finite extent propagate in free
space. For example, the expanding profile of a laser beam, the beam shape of a radar antenna and the
field of view of an ultrasonic transducer can all be analysed using diffraction equations.
Solar glory at the steam from hot springs. A glory is an optical phenomenon produced by light
backscattered (a combination of diffraction, reflection and refraction) towards its source by a cloud of
uniformly-sized water droplets.
The effects of diffraction are often seen in everyday life. The most striking examples of diffraction are
those involving light; for example, the closely spaced tracks on a CD or DVD act as a diffraction grating
to form the familiar rainbow pattern seen when looking at a disk. This principle can be extended to
engineer a grating with a structure such that it will produce any diffraction pattern desired; the hologram
on a credit card is an example. Diffraction in the atmosphere by small particles can cause a bright ring to
be visible around a bright light source like the sun or the moon. A shadow of a solid object, using light
from a compact source, shows small fringes near its edges. The speckle pattern which is observed when
laser light falls on an optically rough surface is also a diffraction phenomenon. All these effects are a
consequence of the fact that light propagates as a wave.
Diffraction can occur with any kind of wave. Ocean waves diffract around jetties and other obstacles.
Sound waves can diffract around objects, which is why one can still hear someone calling even when
hiding behind a tree.[5] Diffraction can also be a concern in some technical applications; it sets a
fundamental limit to the resolution of a camera, telescope, or microscope.
Thomas Young's sketch of two-slit diffraction, which he presented to the Royal Society in 1803
The effects of diffraction of light were first carefully observed and characterized by Francesco Maria
Grimaldi, who also coined the term diffraction, from the Latin diffringere, 'to break into pieces', referring
to light breaking up into different directions. The results of Grimaldi's observations were published
posthumously in 1665.[6][7][8] Isaac Newton studied these effects and attributed them to inflexion of light
rays. James Gregory (16381675) observed the diffraction patterns caused by a bird feather, which was
effectively the first diffraction grating to be discovered. [9] Thomas Young performed a celebrated
experiment in 1803 demonstrating interference from two closely spaced slits. [10] Explaining his results by
interference of the waves emanating from the two different slits, he deduced that light must propagate as
waves. Augustin-Jean Fresnel did more definitive studies and calculations of diffraction, made public in
1815[11] and 1818,[12] and thereby gave great support to the wave theory of light that had been advanced by
Christiaan Huygens[13] and reinvigorated by Young, against Newton's particle theory.

[edit] Mechanism
Photograph of single-slit diffraction in a circular ripple tank
Diffraction arises because of the way in which waves propagate; this is described by the Huygens
Fresnel principle and the principle of superposition of waves. The propagation of a wave can be
visualized by considering every point on a wavefront as a point source for a secondary spherical wave.
The wave displacement at any subsequent point is the sum of these secondary waves. When waves are
added together, their sum is determined by the relative phases as well as the amplitudes of the individual
waves so that the summed amplitude of the waves can have any value between zero and the sum of the
individual amplitudes. Hence, diffraction patterns usually have a series of maxima and minima.
There are various analytical models which allow the diffracted field to be calculated, including the
Kirchhoff-Fresnel diffraction equation which is derived from wave equation, the Fraunhofer diffraction
approximation of the Kirchhoff equation which applies to the far field and the Fresnel diffraction
approximation which applies to the near field. Most configurations cannot be solved analytically, but can
yield numerical solutions through finite element and boundary element methods.
[edit] Systems
It is possible to obtain a qualitative understanding of many diffraction phenomena by considering how the
relative phases of the individual secondary wave sources vary, and in particular, the conditions in which
the phase difference equals half a cycle in which case waves will cancel one another out.
The simplest descriptions of diffraction are those in which the situation can be reduced to a twodimensional problem. For water waves, this is already the case; water waves propagate only on the
surface of the water. For light, we can often neglect one direction if the diffracting object extends in that
direction over a distance far greater than the wavelength. In the case of light shining through small
circular holes we will have to take into account the full three dimensional nature of the problem.
Some of the simpler cases of diffraction are considered below.
[edit] Single-slit diffraction
Main article: Diffraction formalism
Diffraction of red laser beam on the hole
Numerical approximation of diffraction pattern from a slit of width four wavelengths with an incident
plane wave. The main central beam, nulls, and phase reversals are apparent.
Graph and image of single-slit diffraction
A long slit of infinitesimal width which is illuminated by light diffracts the light into a series of circular
waves and the wavefront which emerges from the slit is a cylindrical wave of uniform intensity.
A slit which is wider than a wavelength produces interference effects in the space downstream of the slit.
These can be explained by assuming that the slit behaves as though it has a large number of point sources
spaced evenly across the width of the slit. The analysis of this system is simplified if we consider light of
a single wavelength. If the incident light is monochromatic, these sources all have the same phase. Light
incident at a given point in the space downstream of the slit is made up of contributions from each of
these point sources and if the relative phases of these contributions vary by 2 or more, we may expect to
find minima and maxima in the diffracted light. Such phase differences are caused by differences in the
path lengths over which contributing rays reach the point from the slit.
We can find the angle at which a first minimum is obtained in the diffracted light by the following
reasoning. The light from a source located at the top edge of the slit interferes destructively with a source
located at the middle of the slit, when the path difference between them is equal to /2. Similarly, the
source just below the top of the slit will interfere destructively with the source located just below the
middle of the slit at the same angle. We can continue this reasoning along the entire height of the slit to
conclude that the condition for destructive interference for the entire slit is the same as the condition for
destructive interference between two narrow slits a distance apart that is half the width of the slit. The
path difference is given by so that the minimum intensity occurs at an angle min given bywhere d is the
width of the slit,
min is the angle of incidence at which the minimum intensity occurs, and is the wavelength of the light
A similar argument can be used to show that if we imagine the slit to be divided into four, six,
eight parts, etc., minima are obtained at angles n given by
Where n is an integer other than zero.
There is no such simple argument to enable us to find the maxima of the diffraction pattern.
The intensity profile can be calculated using the Fraunhofer diffraction equation as

Where I() is the intensity at a given angle, I0 is the original intensity, and
1.
the sinc function is given by sinc(x) = sin(x)/(x) if x 0, and sinc(0) = 1
This analysis applies only to the far field, that is, at a distance much larger than the width of the slit.
2-slit (top) and 5-slit diffraction of red laser light
Diffraction of a red laser using a diffraction grating
A diffraction pattern of a 633 nm laser through a grid of 150 slits
[edit] Diffraction grating
Main article: Diffraction grating
A diffraction grating is an optical component with a regular pattern. The form of the light diffracted by a
grating depends on the structure of the elements and the number of elements present, but all gratings have
intensity maxima at angles m which are given by the grating equation
where
i is the angle at which the light is incident,
d is the separation of grating elements, and
m is an integer which can be positive or negative.
The light diffracted by a grating is found by summing the light diffracted from each of the elements, and
is essentially a convolution of diffraction and interference patterns.
The figure shows the light diffracted by 2-element and 5-element gratings where the grating spacings are
the same; it can be seen that the maxima are in the same position, but the detailed structures of the
intensities are different.
A computer-generated image of an Airy disk
Computer generated light diffraction pattern from a circular aperture of diameter 0.5 micrometre at a
wavelength of 0.6 micrometre (red-light) at distances of 0.1 cm 1 cm in steps of 0.1 cm. One can see
the image moving from the Fresnel region into the Fraunhofer region where the Airy pattern is seen.
[edit] Circular aperture
Main article: Airy disk
The far-field diffraction of a plane wave incident on a circular aperture is often referred to as the Airy
Disk. The variation in intensity with angle is given by
where a is the radius of the circular aperture, k is equal to 2/ and J1 is a Bessel function. The smaller the
aperture, the larger the spot size at a given distance, and the greater the divergence of the diffracted
beams.[edit] General aperture
The wave that emerges from a point source has amplitude at location r that is given by the solution of
the frequency domain wave equation for a point source (The Helmholtz Equation),
where is the 3-dimensional delta function. The delta function has only radial dependence, so the Laplace
operator (aka scalar Laplacian) in the spherical coordinate system simplifies to (see del in cylindrical and
spherical coordinates)
By direct substitution, the solution to this equation can be readily shown to be the scalar Green's function,
which in the spherical coordinate system (and using the physics time convention e it) is:
This solution assumes that the delta function source is located at the origin. If the source is located at an
arbitrary source point, denoted by the vector and the field point is located at the point , then we may
represent the scalar Green's function (for arbitrary source location) as:
Therefore, if an electric field, Einc(x,y) is incident on the aperture, the field produced by this aperture
distribution is given by the surface integral:
On the calculation of Fraunhofer region fields where the source point in the aperture is given by the
vector
In the far field, wherein the parallel rays approximation can be employed, the
Green's function, simplifies to

as can be seen in the figure to the right (click to enlarge). The expression for the far-zone (Fraunhofer
region) field becomes

Now, since
and
the expression for the Fraunhofer region field from a planar aperture now becomes,

Letting,
and
the Fraunhofer region field of the planar aperture assumes the form of a Fourier transform

In the far-field / Fraunhofer region, this becomes the spatial Fourier transform of the aperture distribution.
Huygens' principle when applied to an aperture simply says that the far-field diffraction pattern is the
spatial Fourier transform of the aperture shape, and this is a direct by-product of using the parallel-rays
approximation, which is identical to doing a plane wave decomposition of the aperture plane fields (see
Fourier optics).
[edit] Propagation of a laser beam
The way in which the profile of a laser beam changes as it propagates is determined by diffraction. The
output mirror of the laser is an aperture, and the subsequent beam shape is determined by that aperture.
Hence, the smaller the output beam, the quicker it diverges.Diode lasers have much greater divergence
than HeNe lasers for this reason.
Paradoxically, it is possible to reduce the divergence of a laser beam by first expanding it with one
convex lens, and then collimating it with a second convex lens whose focal point is coincident with that
of the first lens. The resulting beam has a larger aperture, and hence a lower divergence.
[edit] Diffraction-limited imaging
Main article: Diffraction-limited system

The Airy disk around each of the stars from the 2.56 m telescope aperture can be seen in this lucky image
of the binary star zeta Botis.
The ability of an imaging system to resolve detail is ultimately limited by diffraction. This is because a
plane wave incident on a circular lens or mirror is diffracted as described above. The light is not focused
to a point but forms an Airy disk having a central spot in the focal plane with radius to first null of
where is the wavelength of the light and N is the f-number (focal length divided by diameter) of the
imaging optics. In object space, the corresponding angular resolution is
where D is the diameter of the entrance pupil of the imaging lens (e.g., of a telescope's main mirror).
Two point sources will each produce an Airy pattern see the photo of a binary star. As the point sources
move closer together, the patterns will start to overlap, and ultimately they will merge to form a single
pattern, in which case the two point sources cannot be resolved in the image.ThRayleighcriterion
specifies that two point sources can be considered to be resolvable if the separation of the two images is

at least the radius of the Airy disk, i.e. if the first minimum of one coincides with the maximum of the
other.
Thus, the larger the aperture of the lens, and the smaller the wavelength, the finer the resolution of an
imaging system. This is why telescopes have very large lenses or mirrors, and why optical microscopes
are limited in the detail which they can see.
[edit] Speckle patterns
Main article: speckle pattern
The speckle pattern which is seen when using a laser pointer is another diffraction phenomenon. It is a
result of the superpostion of many waves with different phases, which are produced when a laser beam
illuminates a rough surface. They add together to give a resultant wave whose amplitude, and therefore
intensity varies randomly.
[edit] Patterns
The upper half of this image shows a diffraction pattern of He-Ne laser beam on an elliptic aperture. The
lower half is its 2D Fourier transform approximately reconstructing the shape of the aperture.Several
qualitative observations can be made of diffraction in general:
The angular spacing of the features in the diffraction pattern is inversely proportional to the dimensions
of the object causing the diffraction. In other words: The smaller the diffracting object, the 'wider' the
resulting diffraction pattern, and vice versa. (More precisely, this is true of the sines of the angles.)The
diffraction anglesare invariant under scaling; that is, they depend only on the ratio of the wavelength to
the size of the diffracting object.When the diffracting object has a periodic structure, for example in a
diffraction grating, the features generally become sharper. The third figure, for example, shows a
comparison of a double-slit pattern with a pattern formed by five slits, both sets of slits having the same
spacing, between the center of one slit and the next.
[edit] Particle diffraction
See also: neutron diffraction and electron diffraction
Quantum theory tells us that every particle exhibits wave properties. In particular, massive particles
caninterfere and therefore diffract. Diffraction of electrons and neutrons stood as one of thepowerful
arguments in favor of quantum mechanics. The wavelength associated with a particle is the de Broglie
wavelength

where h is Planck's constant and p is the momentum of the particle (mass velocity for slow-moving
particles). For most macroscopic objects, this wavelength is so short that it is not meaningful to assign a
wavelength to them. A sodium atom traveling at about 30,000 m/s would have a De Broglie wavelength
of about 50 pico meters.Because the wavelength for even the smallest of macroscopic objects is
extremely small, diffraction of matter waves is only visible for small particles, like electrons, neutrons,
atoms and small molecules. The short wavelength of these matter waves makes them ideally suited to
study the atomic crystal structure of solids and large molecules like proteins.
Relatively larger molecules like buckyballs were also shown to diffract.[14]
[edit] Bragg diffraction
Following Bragg's law, each dot (or reflection), in this diffraction pattern forms from the constructive
interference of X-rays passing through a crystal. The data can be used to determine the crystal's atomic
structure.
For more details on this topic, see Bragg diffraction.
Diffraction from a three dimensional periodic structure such as atoms in a crystal is called Bragg
diffraction. It is similar to what occurs when waves are scattered from a diffraction grating. Bragg
diffraction is a consequence of interference between waves reflecting from different crystal planes. The
condition of constructive interference is given by Bragg's law:
where is the wavelength,d is the distance between crystal planes, is the angle of the diffracted wave.and
m is an integer known as the order of the diffracted beam.Bragg diffraction may be carried out using
either light of very short wavelength like x-rays or matter waves like neutrons (and electrons) whose
wavelength is on the order of (or much smaller than) the atomic spacing. [15] The pattern produced gives

information of the separations of crystallographic planes allowing one to deduce the crystal structure.
Diffraction contrast, in electron microscopes and x-topography devices in particular, is also a powerful
tool for examining individual defects and local strain fields in crystals.[edit] CoherenceMain article:
Coherence (physics)The description of diffraction relies on the interference of waves emanating from the
same source taking different paths to the same point on a screen. In this description, the difference in
phase between waves that took different paths is only dependent on the effective path length. This does
not take into account the fact that waves that arrive at the screen at the same time were emitted by the
source at different times. The initial phase with which the source emits waves can change over time in an
unpredictable way. This means that waves emitted by the source at times that are too far apart can no
longer form a constant interference pattern since the relation between their phases is no longer time
independent.The length over which the phase in a beam of light is correlated, is called the coherence
length. In order for interference to occur, the path length difference must be smaller than the coherence
length. This is sometimes referred to as spectral coherence, as it is related to the presence of different
frequency components in the wave. In the case of light emitted by an atomic transition, the coherence
length is related to the lifetime of the excited state from which the atom made its transition.If waves are
emitted from an extended source, this can lead to incoherence in the transversal direction. When looking
at a cross section of a beam of light, the length over which the phase is correlated is called the transverse
coherence length. In the case of Young's double slit experiment, this would mean that if the transverse
coherence length is smaller than the spacing between the two slits, the resulting pattern on a screen would
look like two single slit diffraction patterns.In the case of particles like electrons, neutrons and atoms, the
coherence length is related to the spatial extent of the wave function that describes the particle.[edit] See
alsoAtmospheric diffraction
Pulse dispersion in a graded index optical fiber is given by

where
is the difference in refractive indices of core and cladding,
is the refractive index of the cladding,
is the length of the fiber taken for observing the pulse dispersion,
is the speed of light, and
is the constant of graded index profile.
Splices n connectors
Numerical of first unit
Unit 2
Losses in optical fibers
Attenuation with graphs
ATTENUATION IN OPTICAL FIBERS
Attenuation and pulse dispersion represent the two most important characteristics of an optical fiber that
determine the information-carrying capacity of a fiber optic communication system. Obviously, the lower
the attenuation (and similarly, the lower the dispersion) the greater can be the required repeater spacing
and therefore the lower will be the cost of the communication system. Pulse dispersion will be discussed
in the next section, while in this section we will discuss briefly the various attenuation mechanisms in an
optical fiber. The attenuation of an optical beam is usually measured in decibels (dB). If an input power
P1 results in an output power P2, the power loss in decibels is given by (dB) = 10 log10 (P1/P2) (712) Thus, if the output power is only half the input power, the loss is 10 log 2 3 dB. Similarly, if the
power reduction is by a factor of 100 or 10, the power loss is 20 dB or 10 dB respectively. If 96% of the
light is transmitted through the fiber, the loss is about 0.18 dB. On the other hand, in a typical fiber
amplifier, a power amplification of about 1000 represents a power gain of 30 dB. Figure 7-10 shows the
spectral dependence of fiber attenuation (i.e., loss coefficient per unit length) as a function of wavelength
of a typical silica optical fiber. The losses are caused by various mechanisms such as Rayleigh scattering,
absorption due to metallic impurities and water in the fiber, and intrinsic absorption by the silica

molecule itself. The Rayleigh scattering loss varies as 1/0 4, i.e., shorter wavelengths scatter more than
longer wavelengths. Here 0 represents the free space wavelength. This is why the loss coefficient
decreases up to about 1550 nm. The two absorption peaks around 1240 nm and 1380 nm are primarily
due to traces of OH ions and traces of metallic ions. For example, even 1 part per million (ppm) of iron
can cause a loss of about 0.68 dB/km at 1100 nm. Similarly, a concentration of 1 ppm of OH ion can
cause a loss of 4 dB/km at 1380 nm. This shows the level of purity that is required to achieve low-loss
optical fibers. If these impurities are removed, the two absorption peaks will disappear. For 0 > 1600 nm
the increase in the loss coefficient is due to the absorption of infrared light by silica molecules. This is an
intrinsic property of silica, and no amount of purification can remove this infrared absorption tail.
Figure 7-10 Typical wavelength dependence of attenuation for a silica fiber. Notice that the lowest
attenuation occurs at 1550 nm [adapted from Miya, Hasaka, and Miyashita].
As you see, there are two windows at which loss attains its minimum value. The first window is around
1300 nm (with a typical loss coefficient of less than 1 dB/km) where, fortunately (as we will see later),
the material dispersion is negligible. However, the loss attains its absolute minimum value of about 0.2
dB/km around 1550 nm. The latter window has become extremely important in view of the availability of
erbium-doped fiber amplifiers.
Calculation of losses using the dB scale become easy. For example, if we have a 40-km fiber link (with
a loss of 0.4 dB/km) having 3 connectors in its path and if each connector has a loss of 1.8 dB, the total
loss will be the sum of all the losses in dB; or 0.4 dB/km 40 km + 3 1.8 dB = 21.4 dB.
Let us assume that the input power of a 5-mW laser decreases to 30 W after traversing through
40 km of an optical fiber. Using Equation 7-12, attenuation of the fiber in dB/km is therefore
[10 log (166.7)]/40 0.56 dB/km.
Mechanisms of attenuation
Light attenuation by ZBLAN and silica fibers
Main article: Transparent materials
Attenuation in fiber optics, also known as transmission loss, is the reduction in intensity of the light beam
(or signal) with respect to distance traveled through a transmission medium. Attenuation coefficients in
fiber optics usually use units of dB/km through the medium due to the relatively high quality of
transparency of modern optical transmission media. The medium is usually a fiber of silica glass that
confines the incident light beam to the inside. Attenuation is an important factor limiting the transmission
of a digital signal across large distances. Thus, much research has gone into both limiting the attenuation
and maximizing the amplification of the optical signal. Empirical research has shown that attenuation in
optical fiber is caused primarily by both scattering and absorption.
[edit] Light scattering
Specular reflection
Diffuse reflection
The propagation of light through the core of an optical fiber is based on total internal reflection of the
lightwave. Rough and irregular surfaces, even at the molecular level, can cause light rays to be reflected
in random directions. This is called diffuse reflection or scattering, and it is typically characterized by
wide variety of reflection angles.
Light scattering depends on the wavelength of the light being scattered. Thus, limits to spatial scales of
visibility arise, depending on the frequency of the incident light-wave and the physical dimension (or
spatial scale) of the scattering center, which is typically in the form of some specific micro-structural
feature. Since visible light has a wavelength of the order of one micrometre (one millionth of a meter)
scattering centers will have dimensions on a similar spatial scale.
Thus, attenuation results from the incoherent scattering of light at internal surfaces and interfaces. In
(poly)crystalline materials such as metals and ceramics, in addition to pores, most of the internal surfaces
or interfaces are in the form of grain boundaries that separate tiny regions of crystalline order. It has
recently been shown that when the size of the scattering center (or grain boundary) is reduced below the
size of the wavelength of the light being scattered, the scattering no longer occurs to any significant
extent. This phenomenon has given rise to the production of transparent ceramic materials.
Similarly, the scattering of light in optical quality glass fiber is caused by molecular level irregularities
(compositional fluctuations) in the glass structure. Indeed, one emerging school of thought is that a glass
is simply the limiting case of a polycrystalline solid. Within this framework, "domains" exhibiting

various degrees of short-range order become the building blocks of both metals and alloys, as well as
glasses and ceramics. Distributed both between and within these domains are micro-structural defects that
provide the most ideal locations for light scattering. This same phenomenon is seen as one of the limiting
factors in the transparency of IR missile domes.[30]
At high optical powers, scattering can also be caused by nonlinear optical processes in the fiber.[31][32]
[edit] UV-Vis-IR absorption
In addition to light scattering, attenuation or signal loss can also occur due to selective absorption of
specific wavelengths, in a manner similar to that responsible for the appearance of color. Primary material
considerations include both electrons and molecules as follows:
1) At the electronic level, it depends on whether the electron orbitals are spaced (or "quantized") such that
they can absorb a quantum of light (or photon) of a specific wavelength or frequency in the ultraviolet
(UV) or visible ranges. This is what gives rise to color.
2) At the atomic or molecular level, it depends on the frequencies of atomic or molecular vibrations or
chemical bonds, how close-packed its atoms or molecules are, and whether or not the atoms or molecules
exhibit long-range order. These factors will determine the capacity of the material transmitting longer
wavelengths in the infrared (IR), far IR, radio and microwave ranges.
The design of any optically transparent device requires the selection of materials based upon knowledge
of its properties and limitations. The Lattice absorption characteristics observed at the lower frequency
regions (mid IR to far-infrared wavelength range) define the long-wavelength transparency limit of the
material. They are the result of the interactive coupling between the motions of thermally induced
vibrations of the constituent atoms and molecules of the solid lattice and the incident light wave radiation.
Hence, all materials are bounded by limiting regions of absorption caused by atomic and molecular
vibrations (bond-stretching)in the far-infrared (>10 m).
Thus, multi-phonon absorption occurs when two or more phonons simultaneously interact to produce
electric dipole moments with which the incident radiation may couple. These dipoles can absorb energy
from the incident radiation, reaching a maximum coupling with the radiation when the frequency is equal
to the fundamental vibrational mode of the molecular dipole (e.g. Si-O bond) in the far-infrared, or one of
its harmonics.
The selective absorption of infrared (IR) light by a particular material occurs because the selected
frequency of the light wave matches the frequency (or an integer multiple of the frequency) at which the
particles of that material vibrate. Since different atoms and molecules have different natural frequencies
of vibration, they will selectively absorb different frequencies (or portions of the spectrum) of infrared
(IR) light.
Reflection and transmission of light waves occur because the frequencies of the light waves do not match
the natural resonant frequencies of vibration of the objects. When IR light of these frequencies strikes an
object, the energy is either reflected or transmitted.
[
[edit] Practical issues
[edit] Optical fiber cables
The propagation constant of an electromagnetic wave is a measure of the change undergone by the
amplitude of the wave as it propagates in a given direction. The quantity being measured can be the
voltage or current in a circuit or a field vector such as electric field strength or flux density. The
propagation constant itself measures change per metre but is otherwise dimensionless.
The propagation constant is expressed logarithmically, almost universally to the base e, rather than the
more usual base 10 used in telecommunications in other situations. The quantity measured, such as
voltage, is expressed as a sinusoidal phasor. The phase of the sinusoid varies with distance which results
in the propagation constant being a complex number, the imaginary part being caused by the phase
change. [edit] Alternative names
The term propagation constant is somewhat of a misnomer as it usually varies strongly with . It is
probably the most widely used term but there are a large variety of alternative names used by various
authors for this quantity. These include, transmission parameter, transmission function, propagation
parameter, propagation coefficient and transmission constant. In plural, it is usually implied that
and are being referenced separately but collectively as in transmission parameters, propagation

parameters, propagation coefficients, transmission constants and secondary coefficients. This last
occurs in transmission line theory, the term secondary being used to contrast to the primary line
coefficients. The primary coefficients being the physical properties of the line; R,C,L and G, from which
the secondary coefficients may be derived using the telegrapher's equation. Note that, at least in the field
of transmission lines, the term transmission coefficient has a different meaning despite the similarity of
name. Here it is the corollary of reflection coefficient.
[edit] Definition
[edit] Attenuation constant
In telecommunications, the term attenuation constant, also called attenuation parameter or
coefficient, is the attenuation of an electromagnetic wave propagating through a medium per unit
distance from the source. It is the real part of the propagation constant and is measured in nepers per
metre. A neper is approximately 8.7dB. Attenuation constant can be defined by the amplitude ratio;

The propagation constant per unit length is defined as the natural logarithmic of ratio of the sending end
current or voltage to the receiving end current or voltage.
[edit] Copper lines
The attenuation constant for copper lines (or ones made of any other conductor) can be calculated from
the primary line coefficients as shown above. For a line meeting the distortionless condition, with a
conductance G in the insulator, the attenuation constant is given by;
however, a real line is unlikely to meet this condition without the addition of loading coils and,
furthermore, there are some frequency dependant effects operating on the primary "constants" which
cause a frequency dependence of the loss. There are two main components to these losses, the metal loss
and the dielectric loss.
The loss of most transmission lines are dominated by the metal loss, which causes a frequency
dependency due to finite conductivity of metals, and the skin effect inside a conductor. The skin effect
causes R along the conductor to be approximately dependent on frequency according to;
Losses in the dielectric depend on the loss tangent (tan) of the material, which depends inversely on the
wavelength of the signal and is directly proportional to the frequency.

[edit] Optical fibre

The attenuation constant for a particular propagation mode in an optical fiber, the real part of the axial
propagation constant.
[edit] Phase constant
In electromagnetic theory, the phase constant, also called phase change constant, parameter or
coefficient is the imaginary component of the propagation constant for a plane wave. It represents the
change in phase per metre along the path travelled by the wave at any instant and is equal to real part of
the angular wavenumber of the wave. It is represented by the symbol and is measured in units of
radians per metre.
From the definition of (angular) wavenumber;

For a transmission line, the Heaviside condition of the telegrapher's equation tells us that the wavenumber
must be proportional to frequency for the transmission of the wave to be undistorted in the time domain.
This includes, but is not limited to, the ideal case of a lossless line. The reason for this condition can be
seen by considering that a useful signal is composed of many different wavelengths in the frequency
domain. For there to be no distortion of the waveform, all these waves must travel at the same velocity so
that they arrive at the far end of the line at the same time as a group. Since wave phase velocity is given
by;

it is proved that is required to be proportional to . In terms of primary coefficients of the line, this
yields from the telegrapher's equation for a distortionless line the condition;
However, practical lines can only be expected to approximately meet this condition over a limited
frequency band.
[edit] Filters
The term propagation constant or propagation function is applied to filters and other two-port networks
used for signal processing. In these cases, however, the attenuation and phase coefficients are expressed
in terms of nepers and radians per network section rather than per metre. Some authors[1] make a
distinction between per metre measures (for which "constant" is used) and per section measures (for
which "function" is used).
The propagation constant is a useful concept in filter design which invariably uses a cascaded section
topology. In a cascaded topology, the propagation constant, attenuation constant and phase constant of
individual sections may be simply added to find the total propagation constant etc.
[edit] Cascaded networks
Three networks with arbitrary propagation constants and impedances connected in cascade. The Z i terms
represent image impedance and it is assumed that connections are between matching image impedances.
The ratio of output to input voltage for each network is given by,[2]
The terms are impedance scaling terms[3] and their use is explained in the image impedance article.
The overall voltage ratio is given by,
Thus for n cascaded sections all having matching impedances facing each other, the overall propagation
constant is given by,
Attenuation
Light power propagating in a fiber decays exponentially with length due to absorption and scattering
losses.
Attenuation is the single most important factor determining the cost of fiber optic telecommunication
systems as it determines spacing of repeaters needed to maintain acceptable signal levels. In the near
infrared and visible regions, the small absorption losses of pure silica are due to tails of absorption bands
in the far infrared and ultraviolet. Impuritiesnotably water in the form of hydroxyl ions are much
more dominant causes of absorption in commercial fibers. Recent improvements in fiber purity have
reduced attenuation losses. State-of-the-art systems can have attenuation on the order of 0.1 dB/km.
Scattering can couple energy from guided to radiation modes, causing loss of energy from the fiber. There
are unavoidable Rayleigh scattering losses from small scale index fluctuations frozen into the fiber when
it solidifies. This produces attenuation proportional to l/4. Irregularities in core diameter and geometry
or changes in fiber axis direction also cause scattering. Any process that
Typical Spectral Attenuation in Silica
Dispersion
As a pulse travels down a fiber, dispersion causes pulse spreading. This limits the distance and the bit rate
of data on an optical fiber. Symbols become unrecognizable 1 0 1
Figure 2Dispersion
6
5
4
3
2
1
0
Attenuation (dB/km) Wavelength (m)
0.7 0.8 0.9 1.0 1.1 1.2 1.3 1.4 1.5 1.6
Figure 3Typical Spectral Attenuation in Silica
Polarization Mode Dispersion.
Polarization Mode Dispersion (PMD) is actually another form of material dispersion. Single-mode fiber
supports one mode, which consists of two orthogonal polarization modes. Ideally, the core of an optical
fiber has an index of refraction that is uniform over the entire cross section, unless the fiber is graded
index. However, mechanical stresses, i.e. bending, can cause slight changes in the index of refraction in
one dimension. This can cause one of the orthogonal polarization modes to travel faster than the other,

hence causin dispersion of the optical pulse. imposes dimensional irregularities such as microbending
increases scattering and hence attenuation.
Absorption losses
VI. ATTENUATION IN OPTICAL FIBERS
Attenuation and pulse dispersion represent the two most important characteristics of an optical fiber that
determine the information-carrying capacity of a fiber optic communication system. Obviously, the lower
the attenuation (and similarly, the lower the dispersion) the greater can be the required repeater spacing
and therefore the lower will be the cost of the communication system. Pulse dispersion will be discussed
in the next section, while in this section we will discuss briefly the various attenuation mechanisms in an
optical fiber.
The attenuation of an optical beam is usually measured in decibels (dB). If an input power P1 results in an
output power P2, the power loss in decibels is given by
(dB) = 10 log10 (P1/P2)
(7-12)
Thus, if the output power is only half the input power, the loss is 10 log 2 3 dB. Similarly, if the power
reduction is by a factor of 100 or 10, the power loss is 20 dB or 10 dB respectively. If 96% of the light is
transmitted through the fiber, the loss is about 0.18 dB. On the other hand, in a typical fiber amplifier, a
power amplification of about 1000 represents a power gain of 30 dB.
Figure 7-10 shows the spectral dependence of fiber attenuation (i.e., loss coefficient per unit length) as a
function of wavelength of a typical silica optical fiber. The losses are caused by various mechanisms such
as Rayleigh scattering, absorption due to metallic impurities and water in the fiber, and intrinsic
absorption by the silica molecule itself. The Rayleigh scattering loss varies as 1/ 04, i.e., shorter
wavelengths scatter more than longer wavelengths. Here 0 represents the free space wavelength. This is
why the loss coefficient decreases up to about 1550 nm. The two absorption peaks around 1240 nm and
1380 nm are primarily due to traces of OH ions and traces of metallic ions. For example, even 1 part per
million (ppm) of iron can cause a loss of about 0.68 dB/km at 1100 nm. Similarly, a concentration of
1 ppm of OH ion can cause a loss of 4 dB/km at 1380 nm. This shows the level of purity that is required
to achieve low-loss optical fibers. If these impurities are removed, the two absorption peaks will
disappear. For 0 > 1600 nm the increase in the loss coefficient is due to the absorption of infrared light
by silica molecules. This is an intrinsic property of silica, and no amount of purification can remove this
infrared absorption tail.
As you see, there are two windows at which loss attains its minimum value. The first window is around
1300 nm (with a typical loss coefficient of less than 1 dB/km) where, fortunately (as we will see later),
the material dispersion is negligible. However, the loss attains its absolute minimum value of about 0.2
dB/km around 1550 nm. The latter window has become extremely important in view of the availability of
erbium-doped fiber amplifiers.
Example 7-3
Calculation of losses using the dB scale become easy. For example, if we have a 40-km fiber link
(with a loss of 0.4 dB/km) having 3 connectors in its path and if each connector has a loss of
1.8 dB, the total loss will be the sum of all the losses in dB; or
0.4 dB/km 40 km + 3 1.8 dB = 21.4 dB.
Example 7-4
Let us assume that the input power of a 5-mW laser decreases to 30 W after traversing through
40 km of an optical fiber. Using Equation 7-12, attenuation of the fiber in dB/km is therefore
[10 log (166.7)]/40 0.56 dB/km.
Scattering
Bending losses
Leaky modes
Mode coupling losses combined losses in the fiber
Numerical sbased ojn these topics
Dispersion

Effect of dispersion on pulse transmission

PULSE DISPERSION IN STEP-INDEX FIBERS (SIF)
In digital communication systems, information to be sent is first coded in the form of pulses and these
pulses of light are then transmitted from the transmitter to the receiver, where the information is decoded.
The larger the number of pulses that can be sent per unit time and still be resolvable at the receiver end,
the larger will be the transmission capacity of the system. A pulse of light sent into a fiber broadens in
time as it propagates through the fiber. This phenomenon is known as pulse dispersion, and it occurs
primarily because of the following mechanisms:
1. Different rays take different times to propagate through a given length of the fiber. We will discuss this
for a step-index multimode fiber and for a parabolic-index fiber in this and the following sections. In the
language of wave optics, this is known as intermodal dispersion because it arises due to different modes
traveling with different speeds.4 2. Any given light source emits over a range of wavelengths, and,
because of the intrinsic property of the material of the fiber, different wavelengths take different amounts
of time to propagate along the same path. This is known as material dispersion and will be discussed in
Section IX.
3. Apart from intermodal and material dispersions, there is yet another mechanismreferred to as
waveguide dispersion and important only in single-mode fibers. We will briefly discuss this in Section
XI. In the fiber shown in Figure 7-7, the rays making larger angles with the axis (those shown as dotted
rays) have to traverse a longer optical path length and therefore take a longer time to reach the output
end. Consequently, the pulse broadens as it propagates through the fiber (see
4 We will have a very brief discussion about modes in Section XI. We may mention here that the number
of modes in a step index fiber is about V2/2 where the parameter V will be defined in Section XI. When
V < 4, the fiber supports only a few modes and it is necessary to use wave theory. On the other hand, if V
> 8, the fiber supports many modes and the fiber is referred to as a multimoded fiber. When this happens,
ray optics gives accurate results and one need not use wave theory.
Figure 7-11). Even though two pulses may be well resolved at the input end, because of the broadening of
the pulses they may not be so at the output end. Where the output pulses are not resolvable, no
information can be retrieved. Thus, the smaller the pulse dispersion, the greater will be the informationcarrying capacity of the system.
Figure 7-11 Pulses separated by 100 ns at the input end would be resolvable at the output end of 1 km
of the fiber. The same pulses would not be resolvable at the output end of 2 km of the same fiber.
We will now derive an expression for the intermodal dispersion for a step-index fiber. Referring back to
Figure 7-7b, for a ray making an angle with the axis, the distance AB is traversed in time.where c/n1
represents the speed of light in a medium of refractive index n1, c being the speed of light in free space.
Since the ray path will repeat itself, the time taken by a ray to traverse a length L of the fiber would be
tnL
L c cos = 1
The above expression shows that the time taken by a ray is a function of the angle made by
the ray with the z-axis (fiber axis), which leads to pulse dispersion. If we assume that all rays
lying between = 0 and = c = cos1(n2/n1) (see Equation 7-8) are present, the time taken by
The following extreme rays for a fiber of length L would be given by
tnL
min = c1 corresponding to rays at = 0 (7-16)
tnL
max cn = 1
corresponding to rays at =c = cos1(n2/n1) (7-17)
Hence, if all the input rays were excited simultaneously, the rays would occupy a time interval
at the output end of duration
i c
= = LN M
OQ P
ttnLn
max min n 1 1
2
1

or, finally, the intermodal dispersion in a multimode SIF is

i n L
where has been defined earlier [see Equations 7-5 and 7-6] and we have used Equation 7-11. The
quantity i represents the pulse dispersion due to different rays taking different times in propagating
through the fiber, which, in wave optics, is nothing but the intermodal dispersion and hence the subscript
i. Note that the pulse dispersion is proportional to the square of NA. Thus, to have a smaller dispersion,
one must have a smaller NA, which of course reduces the acceptance angle im and hence the lightgathering power. If, at the input end of the fiber, we have a pulse of width 1, after propagating through a
length L of the fiber the pulse will have a width 2 given approximately by
That is, a pulse traversing through the fiber of length 1 km will be broadened by 50 ns. Thus, two pulses
separated by, say, 500 ns at the input end will be quite resolvable at the end of 1 km of the fiber.
However, if consecutive pulses were separated by, say, 10 ns at the input end, they would be absolutely
unresolvable at the output end. Hence, in a 1-Mbit/s fiber optic system, where we have one pulse every
106 s, a 50-ns/km dispersion would require repeaters to be placed every 3 to 4 km. On the other hand, in
a 1-Gbit/s fiber optic communication system, which requires the transmission of one pulse every 109s, a
dispersion of 50 ns/km would result in intolerable broadening even within 50 meters or so. This would be
highly inefficient and uneconomical from a system point of view.
From the discussion in the above example it follows that, for a very-high-information-carrying system, it
is necessary to reduce the pulse dispersion. Two alternative solutions existone involves the use of nearparabolic-index fibers and the other involves single-mode fibers.
Types of dispersion- intermodal and intramodels Intermodal Dispersion As its name implies, intermodal
dispersion is a phenomenon between different modes in an optical fiber. Therefore this category of
dispersion only applies to mulitmode fiber. Since all the different propagating modes have different group
velocities, the time it takes each mode to travel a fixed distance is also different. Therefore as an optical
pulse travels down a multim ode fiber, th e pulses begin to spread, until they eventually spread into one
another. This effect limits both the bandwidth of multimode fiber as well as the distance it can transport
data.
Intramodal Dispersion
Intramodal dispersion, sometimes called material dispersion, is a category of dispersion that occurs
within a single-mode. This dispersion mechanism is a result of material properties of optical fiber and
applies to both single-mode and multimode fibers. There are two distinct types of intramodal dispersion:
chromatic dispersion and polarization mode dispersion.
Chromatic Dispersion. In silica, the index of refraction is dependent upon wavelength. Therefore
different wavelengths will travel down an optical fiber at different velocities. This implies that a pulse
with a wider FWHM will spread more than a pulse with a narrower FWHM. This dispersion limits both
the bandwidth and the distance that information can be supported. This is why for long communications
links, it is desirable to use a laser with a very narrow linewidth. Distributed Feedback (DFB) lasers are
popular for communications because they have a single longitudinal mode with a very narrow linewidth.
Material and waveguide dispersion
MATERIAL DISPERSION
We first define the group index. To do this we return to Equation 7-1 where we noted that the
velocity of light in a medium is given by v = c/n (7-24)
Here n is the refractive index of the medium, which, in general, depends on the wavelength. The
dependence of the refractive index on wavelength leads to what is known as dispersion,
discussed in Module 1.3, Basic Geometrical Optics. In Figure 7-13 we have shown a narrow
pencil of a white light beam incident on a prism. Since the refractive index of glass depends on
the wavelength, the angle of refraction will be different for different colors. For example, for
crown glass the refractive indices at 656.3 nm (orange), 589.0 nm (yellow), and 486.1 nm
(green) are respectively given by 1.5244, 1.5270, and 1.5330. Thus, if the angle of incidence
i = 45 the angle of refraction, r, will be r = 27.64, 27.58, and 27.47 for the orange, yellow,
and blue colors respectively. The incident white light will therefore disperse into its constituent
colorsthe dispersion will become more evident at the second surface of the prism as seen in
Figure 7-13.

Figure 7-13 Dispersion of white light as it passes through a prism

Now, the quantity v defined by Equation 7-24 is usually referred to as the phase velocity.
However, a pulse travels with what is known as the group velocity, which is given by
vg = c/ng (7-25)
where ng is known as the group index and, in most cases its value is slightly larger than n. In
Table 7.1 we have tabulated n and ng for pure silica for different values of wavelength lying
between 700 nm and 1600 nm. The corresponding spectral variation of n and ng for pure silica is
given in Figure 7-14.
Figure 7-14 Variation of n and ng with wavelength for pure silica. Notice that ng has
a minimum value around 1270 nm.
In Sections VIII and IX we considered the broadening of an optical pulse due to different rays
taking different amounts of time to propagate through a certain length of the fiber. However,
every source of light has a certain wavelength spread, which is often referred to as the spectral
width of the source. Thus, a white light source (like the sun) has a spectral width of about
300 nm. On the other hand, an LED has a spectral width of about 25 nm and a typical laser
diode (LD) operating at 1300 nm has a spectral width of about 2 nm or less. Each wavelength
component will travel with a slightly different group velocity through the fiber. This results in
broadening of a pulse. This broadening of the pulse is proportional to the length of the fiber and
to the spectral width of the source. We define the material dispersion coefficient Dm, which is
measured in ps/km-nm. Dm represents the material dispersion in picoseconds per kilometer
length of the fiber per nanometer spectral width of the source. At a particular wavelength, the
value of Dm is a characteristic of the material and is (almost) the same for all silica fibers. The
values of Dm for different wavelengths are tabulated in Table 7.1. A negative Dm implies that
the longer wavelengths travel faster; similarly, a positive value of Dm implies that shorter
wavelengths travel faster. However, in calculating the pulse broadening, only the magnitude
should be considered.
Example 7-6
The LEDs used in early optical communication systems had a spectral width 0 of about 20 nm
around 0 = 825 nm. Using Dm in Table 7.1 (at 850 nm), such a pulse will broaden by
m = Dm L = 84.2 (ps/km-nm) 1 (km) 20 (nm) ~ 1700 ps = 1.7 ns
in traversing a 1-km length of the fiber. It is very interesting to note that, if we carry out a similar
calculation around 0 1300 nm, we will obtain a much smaller value of m; thus
m = Dm L = 2.4 (ps/km-nm) 1 (km) 20 (nm) ~ 0.05 ns
OP T I C A L WA V E G U I D E S A N D F I B E R S
2000 University of Connecticut 271
in traversing 1-km length of the fiber. The very small value of m is due to the fact that ng is
approximately constant around 0 = 1300 nm, as shown in Figure 7-14. Indeed, the wavelength
0 1270 nm is usually referred to as the zero material dispersion wavelength, and it is because of
such low material dispersion that the optical communication systems shifted their operation to
around 0 1300 nm.
Example 7-7
In the optical communication systems that are in operation today, one uses laser diodes (LD) with
0 1550 nm having a spectral width of about 2 nm. Thus, for a 1-km length of the fiber, the
material dispersion m becomes
m = Dm L = 21.5 (ps/km-nm) 1 (km) 2 (nm) ~ 43 ps
the positive sign indicating that higher wavelengths travel more slowly than lower wavelengths.
[Notice from Table 7.1 that, for 0 1300 nm, ng increases with 0.]
Waveguide dispersion
In Section IX we discussed material dispersion that results from the dependence of the
refractive index of the fiber on wavelength. Even if we assume the refractive indices n1 and n2 to
be independent of 0, the group velocity of each mode does depend on the wavelength. This
leads to what is known as waveguide dispersion. The detailed theory is rather involved [see,
e.g., Chapter 10, Ghatak and Thyagarajan]; we may mention here two important points:
1. The waveguide dispersion is usually negative for a given single-mode fiber. The
magnitude increases with an increase in wavelength.

2. If the core radius a (of a single-mode fiber) is made smaller and the value of is made
larger, the magnitude of the waveguide dispersion increases. Thus we can tailor the
waveguide dispersion by changing the refractive index profile.
The following two examples demonstrate how one can tailor the zero-dispersion wavelength by
changing the fiber parameters.
Example 7-17
We consider the fiber discussed in Example 7-12 for which n2 = 1.447, = 0.003, and a = 4.2 m.
The variations of the waveguide dispersion (w), material dispersion (m), and total dispersion
OP T I C A L WA V E G U I D E S A N D F I B E R S
2000 University of Connecticut 277
(tot = w + m) with 0 are shown in Figure 7-15. From the figure it can be seen that the total
dispersion passes through zero around 0 1300 nm. This is known as zero total-dispersion
wavelength and represents an extremely important parameter.
Figure 7-15 The variations of m,, w, and tot with 0 for a typical conventional single-mode fiber
(CSF) with parameters given in Example 7-17. The total dispersion passes through zero at around
0 1300 nm, known as zero total dispersion wavelength.
Example 7-18
We next consider the fiber discussed in Example 7-13 for which n2 = 1.444, = 0.0075, and
a = 2.3 m. For this fiber, at 0 1550 nm,
w = 20 ps/km-nm, as seen in Figure 7-16.
On the other hand, the material dispersion at this wavelengthper km and per unit wavelength
interval in nmis given by Table 7.1 as
Dm = m = +21 ps/km-nm
We therefore see that the two expressions are of opposite sign and almost cancel each other.
Physically, because of waveguide dispersion, longer wavelengths travel more slowly than shorter
wavelengths. And, because of material dispersion, longer wavelengths travel faster than shorter
wavelengths. So the two effects compensate for each other, resulting in a zero total dispersion
around 1550 nm. Thus we have been able to shift the zero-dispersion wavelength by changing the
fiber parameters. These are known as the dispersion-shifted fibers, the importance of which will be
discussed in the next section. The variations of m, w, and tot with 0 are plotted in Figure 7-16,
showing clearly that tot is near zero at o = 1550 nm.
F U N D AM E N T AL S O F P H O T O N I C S
278 2000 University of Connecticut
Figure 7-16 The variations of m, w, and tot with 0 for a typical dispersion-shifted single-mode
fiber (DSF) with parameters given in Example 7-18. The total dispersion passes through zero at
around 0 1550 nm.
X. DISPERSION AND MAXIMUM BIT RATE
We may mention here briefly that, in a digital communication system employing light pulses,
pulse broadening would result in an overlap of pulses, resulting in loss of resolution and leading
to errors in detection. Thus pulse broadening is one of the mechanisms (other than attenuation)
that limits the distance between two repeaters in a fiber optic link. It is obvious that, the larger
the pulse broadening, the smaller will be the number of pulses per second that can be sent down
a link. Different criteria based on slightly different considerations are used to estimate the
maximum permissible bit rate (Bmax) for a given pulse dispersion. However, it is always of the
order of 1/, where is the pulse dispersion. In one type of extensively used coding (known as
NRZ) we have
Bmax 0.7/ (7-26)
This formula takes into account (approximately) only the limitation imposed by the pulse
dispersion in the fiber. In an actual link the source and detector characteristics would also be
taken into account while estimating the maximum bit rate (see Module 1-8, Fiber Optic
Telecommunication). It should also be pointed out that, in a fiber, the pulse dispersion is caused,
in general, by intermodal dispersion, material dispersion, and waveguide dispersion. However,
waveguide dispersion is important only in single-mode fibers and may be neglected in carrying

out analysis for multimode fibers. Thus (considering multimode fibers), if i and m are the
dispersion due to intermodal and material dispersions respectively, the total dispersion is given
by
=+im
2 2 (7-27)
F U N D AM E N T AL S O F P H O T O N I C S
272 2000 University of Connecticut
Example 7-8
We consider a step-index multimode fiber with n1 = 1.46, = 0.01, operating at 850 nm. For such a
fiber, the intermodal dispersion (for a 1-km length) is
which is usually written as
1 49 ns/km
If the source is an LED with = 20 nm, using Table 7.1, the material dispersion m is 1.7 ns/km
[see Example 7-6]. Thus, in step-index multimode fibers, the dominant pulse-broadening
mechanism is intermodal dispersion and the total dispersion is given by
= i + m =
2 2 49 ns/km
Using Equation 7-26, this gives a maximum bit rate of about
Bmax 0.7/ 14 Mbit-km/s
Thus a 10-km link can at most support 1.4 Mbit/s, since increases by a factor of 10, causing Bmax
to decrease by the same factor.
Example 7-9
Let us now consider a parabolic-index multimode fiber with n1 = 1.46, = 0.01, operating at
850 nm with an LED of spectral width 20 nm. For such a fiber, the intermodal dispersion, using
Equation 7-22, is
im = L n
c
12
2
0.24 ns/km
The material dispersion is again 1.7 ns/km. Thus, in this case the dominant mechanism is material
dispersion rather than intermodal dispersion. The total dispersion is
= 0.242 +1.72 = 1.72 ns/km
This gives a maximum bit rate of about
Bmax 0.7/ 400 Mbit-km/s
giving a maximum permissible bit rate of 20 Mbit/s for a 20-km link.
Example 7-10
If we now shift the wavelength of operation to 1300 nm and use the parabolic-index fiber of the
previous example, we see that the intermodal dispersion remains the same at 0.24 ns/km while the
material dispersion (for an LED of 0 = 20 nm) becomes 0.05 ns/km (see Example 7-6). The
OP T I C A L WA V E G U I D E S A N D F I B E R S
2000 University of Connecticut 273
material dispersion is now negligible in comparison to intermodal dispersion. Thus the total
dispersion and maximum bit rate are respectively given by
0.242 + 0.052 0.25 ns/km; Bmax = 2.8 Gbit-km/s
Example 7-11
If, in Example 7-10, we replace the LED with a laser diode of spectral width 2 nm, the material
dispersion becomes 0.17 ns/km, which is now smaller than the intermodal dispersion. The total
dispersion is
= 0.242 + 0.172 = 0.29 ns/km
giving a maximum bit rate of
Bmax = 0.7/ 2.4 Gbit-km/s
We should reiterate that, in the examples discussed above, the maximum bit rate has been estimated
by considering the fiber only. In an actual link, the temporal response of the source and detector

must also be taken into account.

Chromatic dispersion
Group delay etc. derivation
Total dispersion
Transmission rate
Dispersion-shifted fibers
In Section VI we learned that the attenuation of a silica fiber attains its minimum value of about
0.2 dB/km at around 0 1550 nm. The second- and third-generation optical communication
systems operated around 0 1300 nm, where the dispersion was extremely small but the loss
was about 1 dB/km, and therefore the repeater spacing was limited by the loss in the fiber. Since
the lowest loss lies at around 0 1550 nm, if the zero-dispersion wavelength could be shifted
to the 0 1550-nm region, one could have both minimum loss and very low dispersion. This
would lead to very-high-bandwidth systems with very long (~ 100 km) repeater spacings. Apart
from this, extremely efficient optical fiber amplifiers capable of amplifying optical signals in the
1550-nm band have also been developed. Thus, shifting the operating wavelength from 1310 nm
to 1550 nm would be very advantageous. As discussed in Example 7-18, by reducing the core
size and increasing the value of , we can shift the zero-dispersion wavelength to 1550 nm,
which represents the low-loss window. Indeed, the current fourth-generation optical
communication systems operate at 1550 nm, using dispersion-shifted single-mode fibers with
repeater spacing of about 100 km, carrying about 10 Gbit/s of information (equivalent to about
150,000 telephone channels) through one hair-thin single-mode fiber.
We may mention here that, if one is interested in carrying out accurate calculations for total
dispersion, one may use the software described in Ghatak, Goyal, and Varshney.
DISPERSION
Since optical fiber is a waveguide, light can propagate in a number of modes
If a fiber is of large diameter, light entering at different angles will excite different modes while
narrow fiber may only excite one mode
Multimode propagation will cause dispersion, which results in the spreading of pulses and limits
the usable bandwidth
Single-mode fiber has much less dispersion but is more expensive to produce. Its small size,
together with the fact that its numerical aperture is smaller than that of multimode fiber, makes it
more difficult to couple to light sources
Types of fibers

Both types of fiber described earlier are known as step-index fibers because the index of
refraction changes radically between the core and the cladding
Graded-index fiber is a compromise multimode fiber, but the index of refraction gradually
decreases away from the center of the core
Graded-index fiber has less dispersion than a multimode step-index fiber
Dispersion in fiber optics results from the fact that in multimode propagation, the signal travels
faster in some modes than it would in others
Single-mode fibers are relatively free from dispersion except for intramodal dispersion
Graded-index fibers reduce dispersion by taking advantage of higher-order modes
One form of intramodal dispersion is called material dispersion because it depends upon the
material of the core
Another form of dispersion is called waveguide dispersion
Dispersion increases with the bandwidth of the light source

Losses
Losses in optical fiber result from attenuation in the material itself and from scattering, which
causes some light to strike the cladding at less than the critical angle
Bending the optical fiber too sharply can also cause losses by causing some of the light to meet
the cladding at less than the critical angle
Losses vary greatly depending upon the type of fiber
Plastic fiber may have losses of several hundred dB per kilometer
Graded-index multimode glass fiber has a loss of about 24 dB
per kilometer
Single-mode fiber has a loss of 0.4 dB/km or less

Fiber-Optic Cables
There are two basic types of fiber-optic cable
The difference is whether the fiber is free to move inside a tube with a diameter much
larger than the fiber or is inside a relatively tight-fitting jacket
They are referred to as loose-tube and tight-buffer cables
Both methods of construction have advantages
Loose-tube cablesall the stress of cable pulling is taken up by the cables strength
members and the fiber is free to expand and contract with temperature
Tight-buffer cables are cheaper and generally easier to use

Group and phase velocity

Another consequence of dispersion manifests itself as a temporal effect. The formula above, v = c / n
calculates the phase velocity of a wave; this is the velocity at which the phase of any one frequency
component of the wave will propagate. This is not the same as the group velocity of the wave, which is
the rate that changes in amplitude (known as the envelope of the wave) will propagate. The group
velocity vg is related to the phase velocity by, for a homogeneous medium (here is the wavelength in
vacuum, not in the medium): The group velocity vg is often thought of as the velocity at which energy or
information is conveyed along the wave. In most cases this is true, and the group velocity can be thought
of as the signal velocity of the waveform. In some unusual circumstances, where the wavelength of the
light is close to an absorption resonance of the medium, it is possible for the group velocity to exceed the
speed of light (vg > c), leading to the conclusion that superluminal (faster than light) communication is
possible. In practice, in such situations the distortion and absorption of the wave is such that the value of
the group velocity essentially becomes meaningless, and does not represent the true signal velocity of the
wave, which stays less than c. The group velocity itself is usually a function of the waves frequency.
This results in group velocity dispersion (GVD), which causes a short pulse of light to spread in time as a
result of different frequency components of the pulse travelling at different velocities. GVD is often
quantified as the group delay dispersion parameter (again, this formula is for a uniform medium only):
If D is less than zero, the medium is said to have positive dispersion. If D is greater than zero, the
medium has negative dispersion.If a light pulse is propagated through a normally dispersive medium, the
result is the higher frequency components travel slower than the lower frequency components. The pulse
therefore becomes positively chirped, or up-chirped, increasing in frequency with time. Conversely, if a
pulse travels through an anomalously dispersive medium, high frequency components travel faster than
the lower ones, and the pulse becomes negatively chirped, or down-chirped, decreasing in frequency with
time.
For ordering assistance, please contact Customer Service at 877-529-9114 (sales@fiberoptic.com)
or visit our website at: www.fiberoptic.com
Chromatic Dispersion
Chromatic dispersion (optics)

Knowledgebase at FiberOptic.com
The result of GVD, whether negative or positive, is ultimately temporal spreading of the pulse. This
makes dispersion management extremely important in optical communications systems based on optical
fiber, since if dispersion is too high, a group of pulses representing a bit-stream will spread in time and
merge together, rendering the bit-stream unintelligible. This limits the length of fiber that a signal can be
sent down without regeneration. One possible answer to this problem is to send signals down the optical
fibre at a wavelength where the GVD is zero (e.g. around ~1.3-1.5 m in silica fibres), so pulses at this
wavelength suffer minimal spreading from dispersionin practice, however, this approach causes more
problems than it solves because zero GVD unacceptably amplifies other nonlinear effects (such as four
wave mixing). Another possible option is to use soliton pulses in the regime of anomalous dispersion, a
form of optical pulse which uses a nonlinear optical effect to self-maintain its shape
solitons have the practical problem, however, that they require a certain power level to be maintained in
the pulse for the nonlinear effect to be of the correct strength. Instead, the solution that is currently used
in practice is to perform dispersion compensation, typically by matching the fiber with another fiber of
opposite-sign dispersion so that the dispersion effects cancel; such compensation is ultimately limited by
nonlinear effects such as self-phase modulation, which interact with dispersion to make it very difficult to
undo. Dispersion control is also important in lasers that produce short pulses. The overall dispersion of
the optical resonator is a major factor in determining the duration of the pulses emitted by the laser. A pair
of prisms can be arranged to produce net negative dispersion, which can be used to balance the usually
positive dispersion of the laser medium. Diffraction gratings can also be used to produce dispersive
effects; these are often used in highpower laser amplifier systems. Recently, an alternative to prisms and
gratings has been developed: chirped mirrors.
These dielectric mirrors are coated so that different wavelengths have different penetration lengths, and
therefore different group delays. The coating layers can be tailored to achieve a net negative dispersion.
Dispersion in waveguides
Optical fibers, which are used in telecommunications, are among the most abundant types of waveguides.
Dispersion in these fibers is one of the limiting factors that determine how much data can be transported
on a single fiber.
The transverse modes for waves confined laterally within a waveguide generally have different speeds
(and field patterns) depending upon their frequency (that is, on the relative size of the wave, the
wavelength) compared to the size of the waveguide.
In general, for a waveguide mode with an angular frequency () at a propagation constant (so that the
electromagnetic fields in the propagation direction z oscillate proportional to ei(z t)), the groupvelocity dispersion parameter D is defined as:
where = 2c / is the vacuum wavelength and vg = d / d is the group velocity. This formula
generalizes the one in the previous section for homogeneous media, and includes both waveguide
dispersion and material dispersion. The reason for defining the dispersion in this way is that |D| is the
(asymptotic) temporal pulse spreading t per unit bandwidth per unit distance travelled, commonly
reported in ps / nm km for optical fibers.
A similar effect due to a somewhat different phenomenon is modal dispersion, caused by a waveguide
having multiple modes at a given frequency, each with a different speed. A special case of this is
polarization mode dispersion (PMD), which comes from a superposition of two modes that travel at
different speeds due to random imperfections that break the symmetry of the waveguide.
For ordering assistance, please contact Customer Service at 877-529-9114 (sales@fiberoptic.com)
or visit our website at: www.fiberoptic.com
Chromatic Dispersion
Chromatic dispersion (optics)
Knowledgebase at FiberOptic.com
Dispersion in gemology
In the technical terminology of gemology, dispersion is the difference in the refractive index of a material
at the B
and G Fraunhofer wavelengths of 686.7 nm and 430.8 nm and is meant to express the degree to which a
prism
cut from the gemstone shows fire, or color. Dispersion is a material property. Fire depends on the
dispersion,

the cut angles, the lighting environment, the refractive index, and the viewer
Dispersion in imaging
In photographic and microscopic lenses, dispersion causes chromatic aberration, distorting the image, and
various techniques have been developed to counteract it
Dispersion in pulsar timing
Pulsars are spinning neutron stars that emit pulses at very regular intervals ranging from milliseconds to
seconds.
It is believed that the pulses are emitted simultaneously over a wide range of frequencies. However, as
observed on Earth, the components of each pulse emitted at higher radio frequencies arrive before those
emitted at lower frequencies. This dispersion occurs because of the ionised component of the interstellar
medium, which makes the group velocity frequency dependent. The extra delay added at frequency is
where the dispersion measure DM is is the integrated free electron column density ne out to the pulsar at
a distance d[4].
Of course, this delay cannot be measured directly, since the emission time is unknown. What can be
measured is the difference in arrival times at two different frequencies. The delay T between a high
frequency hi and a low frequency lo component of a pulse will be
and so DM is normally computed from measurements at two different frequencies. This allows
computation of the absolute delay at any frequency, which is used when combining many different pulsar
observations into an integrated timing solution.
Introduction
Telecommunications service providers have to face continuously growing bandwidth demands in all
networks areas, from long-haul to access. Because installing new communication links would require
huge investments, telecommunications carriers prefer to increase the capacity of their existing fiber links
by using dense wavelength-division multiplexing (DWDM) systems and/or higher bit rates systems.
However, most of the installed optical fibers are old and exhibit physical characteristics that may limit
their ability to transmit high-speed signals.
The broadening of light pulses, called dispersion, is a critical factor limiting the quality of signal
transmission over optical links. Dispersion is a consequence of the physical properties of the transmission
medium. Single-mode fibers, used in high-speed optical networks, are subject to Chromatic Dispersion
(CD) that causes pulse broadening depending on wavelength, and to Polarization Mode Dispersion
(PMD) that causes pulse broadening depending on polarization. Excessive spreading will cause bits to
overflow their intended time slots and overlap adjacent bits. The receiver may then have difficulty
discerning and properly interpreting adjacent bits, increasing the Bit Error Rate. To preserve the
transmission quality, the maximum amount of time dispersion must be limited to a small proportion of the
signal bit rate, typically 10% of the bit time.
With optical networks moving from 2.5 Gbps to 10 Gbps and onto 40 Gbps, the acceptable tolerance of
dispersion is drastically reduced. For instance, the amount of acceptable chromatic dispersion decreases
by a factor of 16 when moving from 2.5 to 10 Gbps, and by an additional factor of 16 moving from 10 to
40 Gbps. These tight tolerances of high-speed networks mean that every possible source of pulse
spreading should be addressed. Operating companies need to measure the dispersion of their networks to
assess the possibility of upgrading them to higher transmission speeds, or to evaluate the need for
compensation. This paper presents the causes and effects of dispersion and describes the different ways to
measure it.
Chromatic Dispersion (CD)
CD Definition and Origin
Light within a medium travels at a slower speed than in vacuum. The speed at which light travels is
determined by the mediums refractive index. In an ideal situation, the refractive index would not depend
on the wavelength of the light. Since this is not the case, different wavelengths travel at different speeds
within an optical fiber.
Figure 1: CD in single-mode fiber
Laser sources are spectrally thin, but not monochromatic. This means that the input pulse contains several
wavelength components, traveling at different speeds, causing the pulse to spread. The detrimental effects
of chromatic dispersion result in the slower wavelengths of one pulse intermixing with the faster

wavelengths of an adjacent pulse, causing intersymbol interference.

The Chromatic Dispersion of a fiber is expressed in ps/(nm*km), representing the differential delay, or
time spreading (in ps), for a source with a spectral width of 1 nm traveling on 1 km of the fiber. It
depends on the fiber type, and it limits the bit rate or the transmission distance for a good quality of
service.
CD of current network spans
As a consequence of its optical characteristics, the Chromatic Dispersion of a fiber can be changed by
acting on the physical properties of the material. To reduce fiber dispersion, new types of fiber were
invented, including dispersion-shifted fibers (ITU G.653) and non-zero dispersion-shifted fiber (ITU
G.655).
The most commonly deployed fiber in networks (ITU G.652), called dispersion-unshifted singlemode
fiber, has a small chromatic dispersion in the optical window around 1310 nm, but exhibits a higher CD
in the 1550 nm region. This dispersion limits the possible transmission length without compensation on
OC- 768/STM-256 DWDM networks.
ITU G.653 is a dispersion-shifted fiber (DSF), designed to minimize chromatic dispersion in the 1550 nm
window with zero dispersion between 1525 nm and 1575 nm. But this type of fiber has several
drawbacks, such as higher polarization mode dispersion than ITU G.652, and a high Four Wave Mixing
risk, rendering DWDM practically impossible. For these reasons, another singlemode fiber was
developed: the Non-Zero Dispersion-Shifted Fiber (NZDSF). NZDSF fibers have now replaced DSF
fibers, which are not used anymore.
The ITU G.655 Non-Zero Dispersion-Shifted Fibers were developed to eliminate non-linear effects
experienced on DSF fibers. They were developed especially for DWDM applications in the 1550 nm
window. They have a cut-off wavelength around 1310 nm, limiting their operation around this
wavelength.
Poly-chromatic
incident light
1
2
Single-Mode Fiber
Refractive index: n()
Figure 2: Chromatic dispersion profiles of different fiber types
CD Limit and Compensation
The chromatic dispersion in fiber causes a pulse broadening and degrades the transmission quality,
limiting
the distance a digital signal can travel before needing regeneration or compensation. For DWDM systems
using DFB lasers, the maximum length of a link before being affected by chromatic dispersion is
commonly
calculated with the following equation:
L is the link distance in km, CD is the chromatic dispersion in ps/(nm * km),
and B is the bit rate in Gbps.
As an example, consider a typical network transmitting data at 10 Gbps on a channel at 1550 nm over a
standard ITU G.652 fiber, having a CD coefficient of 17 ps/(nm*km). In this case, the theoretical
maximum
distance of the link, before adding CD compensators, will be around 61 km, which is close to the typical
node distance in Europe. This length will be divided by 16 when upgrading the network to a 40 Gbps bit
rate,
thus falling under 4 km. This distance is around 20 km for a 40 Gbps transmission through NZDSF. We
can
thus see that CD is a major limiting factor in high-speed transmission.
Fortunately, CD is quite stable, predictable, and controllable. Dispersion Compensation Fiber (DCF),
with its
large negative CD coefficient, can be inserted into the link at regular intervals to minimize its global
chromatic dispersion.
Tx Rx
DC modules

fiber span
delay [ps]
0
L
Figure 3: Chromatic dispersion compensation scheme
While each spool of DCF adequately solves chromatic dispersion for one channel, this is not usually the
case for all channels on a DWDM link. At the extreme wavelengths of a band, dispersion still
accumulates
and can be a significant problem. In this case, a tunable compensation module may be necessary at the
receiver.
Dispersion [ps/nm-km]
1525
1530
1535
1540
1545
1550
1555
1560
SMF
NZDSF
Dispersion
Shifted Fiber
5
0
-5
Wavelength [nm]
10
1550nm DWDM window
1565
Fiber Type 1310 nm 1550 nm
ITU G.652 conventional 0 17
ITU G.653 DSF -15 0
ITU G.655 NZDSF -12 3
Chromatic Dispersion
ps / (nm * km)
delay [ps]
0
1
2
3
Figure 4: Residual chromatic dispersion on a DWDM-compensated link
Therefore, chromatic dispersion measurement is essential in the field to verify the types of installed
fibers.
Such measurements assess if and how the fibers can be upgraded to transmit higher bit rates, verify fiber
zero point and slope for new installations, and carefully evaluate compensation plans.
CD Measurement Methods
In the field, there are three main methods for determining the chromatic dispersion of an optical fiber.
These
are described by three TIA/EIA industry standards: the pulse-delay method (FOTP-168 standard), the
modulated phase-shift method (FOTP-169 standard), and the differential phase shift method (FOTP-175
standard).
These methods all measure first the time delay, in ps, as a function of the wavelength. They then deduce
the chromatic dispersion coefficient, in ps/(nm*km), from the slope of this delay curve and from the
length of

the link. The delay curve plotted by CD analyzers is a fitted trace based on the acquired time delays at
discrete wavelength points. Therefore, the accuracy of the computation of the CD coefficient depends on
the amount of the measured data.
Figure 5: Curve fit to calculate CD coefficient
Phase-shift and differential phase-shift methods are quite similar. In both methods, a modulated source is
injected at the input of the fiber under test. The phase of the sinusoidal modulating signal is analyzed at
the
output of the fiber and compared to the phase of a reference signal, modulated with the same frequency.
In
the phase-shift method, the reference signal has a fixed wavelength, while the other modulated signal is
tuned in wavelengths. In the differential phase-shift method, both signals are tuned in wavelengths, with a
fixed wavelength interval. The analyzed modulated signal, tuned in wavelengths, is compared to a close
reference signal, also tuned in wavelength, but the wavelength gap is constant. The time delay of the link
is
deduced from the phase-shift measurement, using the relationship between the delay (t), the phase (),
and
the modulation frequency (f):
f
t

=
The phase-shift methods assume there is access to both ends of the fiber under test, with a transmitter unit
connected at the input, and a receiver unit at the output of the fiber. The wavelength selection can be
achieved either at the transmitter unit level, using a tunable laser, or at the receiver unit level, with a
broadband source at the input and a wavelength tunable filter in the receiver unit. This filter must be
spectrally thin (FWHM < 1 nm) to ensure low impact on the measurement accuracy. This second
alternative
is commonly used in the differential phase-shift method. One disadvantage of this method is that the
phaseshift
is evaluated comparing two signals with close wavelengths. Increasing the wavelength interval
between the two modulated signals will increase the accuracy of the delay at one individual wavelength
point, but will decrease the number data points that can be acquired to trace the delay curve, leading to a
lower accuracy in the CD coefficient computation. A compromise has to be found between the
wavelength
interval for the differential phase-shift calculation and the number of acquired points to fit the delay
curve,
but achieving this is not straightforward.
Figure 6: Phase-shift principle
Figure 7: Differential phase-shift principle
The phase-shift methods are two-ended solutions, allowing measurement through amplifiers on long-haul
links. They measure the CD with high accuracy, since the wavelengths of modulated signals are known
accurately, and many points of measurements can be acquired, leading to a better fit of the delay curve.
However, these solutions are expensive and thus may not be appropriate for metropolitan applications.
They require operators at both ends of the link, often with bulky instruments.
The pulse-delay methods measure the time that different wavelength carriers travel through the fiber
under
test, either by photon counting (a direct but very complex method), or by measuring the link length with a
multi-wavelength OTDR. The CD-OTDR launches multiple laser pulses into one end of the fiber under
test,
ideally using more than four different wavelengths for better accuracy. It then analyzes the time to return
after a back-reflection from the connector at the other end. The time delay as a function of the wavelength
is

deduced by comparing the times of flight of the laser pulses.

Figure 8: CD-OTDR principle
Broadband source Low freq. oscillator Fiber under test Delay Wavelength Tunable filter transmitter
receiver
1
2
Tunable
laser
Reference
laser
Low freq.
oscillator
Fiber under test
Demux
transmitter receiver Delay
1
2
End
reflection
The CD-OTDR solution is highly cost-effective, primarily because the initial cost of the instrument is far
lower than that of a phase-shift setup. Furthermore, the CD-OTDR is also a complete 3-wavelength
OTDR
module, which is at the heart of any fiber characterization effort. The CD-OTDR combines two test
setups in
a single, small instrument. It is a one-ended test solution, resulting in less manpower to carry out the test,
less training cost, and lower transportation costs. It can be less accurate than a phase-shift tester, but
accuracy better than 5% can be achieved with a 6-wavelength CD-OTDR. This is sufficient to assess a
links chromatic dispersion, which is a stable and predictable parameter. The CD-OTDR is therefore a
good
alternative to phase-shift testing in metro rings and regional networks.
Table 1: Comparison table of CD measurement methods
Polarization Mode Dispersion (PMD)
PMD Definition and Origin
Testing for Polarization Mode Dispersion (PMD) is becoming essential before upgrading a network to a
higher bit rate because PMD can highly degrade the quality of transmission. It is a difficult parameter to
measure, however, because it varies with time and depends on environmental conditions.
PMD is known to stem from the difference in the propagation constants of a fiber due to geometrical
imperfections in the fiber. The term PMD denotes both the physical phenomenon and the associated
temporal delay. PMD also causes a system penalty because of the associated pulse spreading in a
highspeed
digital transmission system.
The physical origin of PMD is essentially linear birefringence due to core eccentricity and ovalization.
These
appear during the manufacturing process or result from external stresses on the fiber, such as bends and
twists, and can be considered constant over a length called the coupling length. The typical value of the
coupling length is several hundred meters and depends on fiber manufacturing parameters. This means
that for distances that are practical for transmission applications, the actual length of the fiber is much
greater than the coupling length.
The PMD phenomenon is characterized by Differential Group Delay (DGD). DGD is the difference in
propagation time between the two polarization eigenstates, which are the states of polarization with
minimum and maximum propagation time for each wavelength.
Figure 9: DGD of PM and random coupling fibers
Phase-shift method CD-OTDR method
application long-haul link metro & access link
measurement constraint two-ended measurement one-ended measurement

accuracy
good accuracy (depending on
number of acquired points) <5% with 6-wavelength CD-OTDR
cost high low + 3-wavelength OTDR testing capability
In case of weak mode coupling (Polarization Maintaining Fiber--short length of ordinary fiber), the light
polarized along the slow axis arrives later than the light traveling along the fast axis (i.e., the fast and
slow axes have different indexes of refraction). In this case, the PMD is equal to the DGD. In other cases
(long fiber lengths), the optical fiber acts like many short birefringent elements stacked together and the
alignment of fast- and slow-axes is random from element to element. Consequently, we speak about
random (or strong) mode coupling. In that case, the DGD varies as a function of wavelength and the
PMD, expressed in ps, is the average value of the DGD spectral distribution.
The average DGD scales as the square root of the length of the fiber. So the PMD coefficient, expressed
in ps/km, is often calculated. In addition, the second-order PMD coefficient, in ps/(nm.km), expresses
the PMD dependency with the wavelength.
PMD needs to be tested on the C&L bands. But, depending on the wavelength transmission window of
the network, there is a need to also test PMD at 1310 nm as PMD values could be different from 1310 nm
to 1550 nm.
The Statistical Nature of PMD
For a practical transmission system, DGD determines the system penalty and depends to a large extent on
the wavelength of operation within the operating wavelength band. But DGD also changes with
environmental conditions over time. The next two traces show DGD as a function of wavelength, for the
same fiber at different times.
Figure 10: DGD versus wavelength, on the same fiber, at two different moments
The graphs show that DGD at a particular wavelength changes with time. The variation can be as much
as 1 ps in a few minutes. The general aspect of the plots is nevertheless the same, and the distribution
statistics remain constant. It can be shown that DGD versus wavelength exhibits a Maxwellian
distribution [1], with a fairly constant mean value over time. The PMD figure is usually taken as the
average of the wavelength distribution. For this fiber, the PMD (average value of DGD) is 0.65 ps.
PMD Limit and Compensation
PMD is inevitable because it is caused by fiber stresses during manufacturing or by environmental
condition changes. But a small PMD level can be tolerated in networks, depending on the data rate that is
intended to be transmitted. Generally, the maximum tolerable PMD is typically 10% of the bit time.
Table 2: Bit rate and time, and PMD limit
Since PMD is an unstable phenomenon, it is not easy to compensate for it. However, Dispersion
Compensation Modules (DCM) have been developed for that purpose. DCMs are placed in front of
receivers on the network and can be tuned in dispersion to compensate for the dispersion measured
continuously on a sample of optical pulses.
Regularly measuring PMD in the field is essential in order to evaluate network capacity and assess the
possibility of upgrading networks for higher bit rate transmission.
PMD Measurement Methods
In the field, there are three main methods for determining the polarization mode dispersion of an optical
fiber, described by three TIA/EIA industry standards: the Fixed Analyzer Method (FOTP-113 standard),
the Jones Matrix Method (FOTP-122 standard), and the Interferometric Method (FOTP-124 standard).
The fixed analyzer method, also called the wavelength scanning method, involves launching
monochromatic polarized light into the fiber under test and measuring the spectral transmission with
another polarizer.
Figure 11: Fixed analyzer setup
The linear birefringence of a short fiber induces a sinusoidal transmission response when scanning the
wavelength.
Figure 12: Wavelength scanning of a short fiber
The calculation of PMD on a short fiber is straightforward: it is the inverse of the fringe spacing (in
frequency) in the wavelength scanning method. But fiber lengths in networks are much longer than the
coupling length. This changes the aspect of the signal obtained, as illustrated in the example below
showing a 50 km fiber with a 0.1 ps/ km specific PMD and a 1 km coupling length.
Figure 13: Wavelength scanning of a long fiber

Broadband
source Monochromator
Polarizer Polarization
FUT controller
The most striking feature compared to a short fiber is that the spectral transmission is no longer
sinusoidal.
Around some wavelengths (1530 nm in our example), the fringe separation is narrower, indicating that
the effect of PMD is higher around these wavelengths. The definition of PMD is not as straightforward as
it was for short fiber. One method uses fringe counting over the wavelength span. A PMD value can be
established this way, although it is not always clear what should be considered as a fringe. For instance,
should the small peak on the graph around 1555 nm be counted or not? Another method uses the Fourier
transform of the transmission spectrum and is technically very similar to interferometry.
The fixed analyzer method is limited to a 60 ps delay range and is very sensitive to the fiber movement,
thus causing uncertainty. Averaging is required for better accuracy, but this significantly increases the
measurement time. Also, as this method requires step-by-step wavelength scanning, it is necessary to
ensure that the wavelength step is small enough with regard to the PMD to be measured.
The Jones matrix method uses a tunable laser source, a polarization controller, and a polarization
analyzer.
At each wavelength, the polarization controller is scanned and a mathematical calculation is performed
using the Jones matrix formalism (hence the name of the method). From this, the Differential Group
Delay -the maximum delay that can be encountered over polarization -- at that wavelength is derived.
Figure 14: Jones matrix method setup
For a short fiber (shorter than the coupling length), Differential Group Delay is independent of
wavelength.
For long fibers, DGD is no longer independent of wavelength, as illustrated in the example below of a 50
km
fiber with a 0.1 ps/ km specific PMD and a 1 km coupling length. The PMD definition must take this
wavelength dependence into account. PMD is computed as the mean value for DGD over a specified
wavelength range, typically 40 nm around an operating wavelength.
Figure 15: Differential Group Delay versus wavelength
(Jones matrix eigenanalysis) for a long fiber
The Jones matrix method gives an indication of the Differential Group Delay dependence with
wavelength.
Using a narrow band tunable laser, the minimum measurable PMD value is not limited.
As in the fixed analyzer method, the Jones matrix eigenanalysis requires that the fibers not be moved
during the measurement procedure, since this would create measurement artifacts. It is therefore more
suitable for laboratory applications. This method also requires step-by-step wavelength scanning. It is
necessary to ensure that the wavelength step is small enough with regard to the PMD to be measured.
The Jones matrix eigenanalysis is a complex method that requires polarization scanning. The time
required
for testing is usually long: about 10 minutes for each measurement. Finally, it is an expensive solution,
limited to the C&L bands, due to the need to use a tunable laser source.
Tunable Laser
Source
Polarization
controller
Polarimeter
Fiber under test
Computer Differential
Group Delay
The Interferometric Method uses a polarized broadband source and a Michelson scanning interferometer.
The interferometer can be seen as a Fourier transform spectrum analyzer, or more simply as a correlator
for
the PMD-delayed wave trains.

Figure 16: Interferometric method setup

On a short fiber, the interferogram envelope shows a single central peak (which corresponds to the
autocorrelation
peak of the source) and two satellite peaks, as represented in the figure below. The
interferometer unbalance between the central peak and the satellite peaks results in the PMD. The
interferogram can also be seen as the Fourier transform of the transmitted spectrum of the wavelength
scanning method, taken over the spectral range of the source.
Figure 17: Interferogram envelope for a short fiber
For longer randomly mode-coupled fibers, the interferogram is close to a Gaussian function multiplied by
a
noise after removing the auto-correlation peak.
PMD
Figure 18: Interferogram envelope for a long fiber
From a mathematical point of view, the signal is the envelope of the Fourier transform of the transmission
spectrum resulting from the wavelength scanning method. PMD is given by the width of the envelope.
The auto-correlation peak does not contain any information related to PMD measurement. It can be
removed either by computation, or by adding a quarter-wave-plate inside the interferometer to create
destructive interferences when both arms of the interferometer have the same length. This second method
enables a higher dynamic range, resulting in the ability to test longer links, a higher accuracy, and better
repeatability.
Movable
Mirror
Fixed
Mirror
/4
PMD Base Module
Fiber Under Test
This method is well suited for field measurement requirements. The test instruments are rugged and
portable, and can operate on batteries. The interferometric method is insensitive to fiber moves. It
provides
accurate measurements in a few seconds, and does not require averaging. High power broadband sources
can be used to increase the dynamic range, allowing measurement of long links up to 250 km. This
method
also allows PMD measurements through multiple amplifiers such as EDFAs.
The minimum measurable PMD value is limited in this method, depending on the spectral width of the
broadband source. However, sources with FWHM >50 nm will allow PMD measurement down to 0.05
ps,
which is quite acceptable in the field.
The interferometric method has been standardized in two versions: the TINTY (traditional
interferometry)
version, as described above, and the GINTY (generalized interferometry) version. GINTY differs from
TINTY by adding polarization scramblers and an analyzer at the input and the output of the fiber under
test.
This is another way to remove the auto-correlation peak from the interferogram and then measure long
links
having EDFAs. However, because scramblers are used, averaging a large number of measurements is
necessary to achieve the full benefit of this method. As was explained above, technical solutions also
exist
in the TINTY method to remove this central peak using methods other than computation.
Table 3: Comparison table of PMD measurement methods
Conclusion
Dispersion in optical fibers limits the quality of signal transmission. Both CD and PMD must be
measured to
assess the potential of upgrading networks to higher transmission speeds, or to evaluate the need for
compensations. Various testing solutions exist, each having advantages and drawbacks. As an aid to

finding the most appropriate solution, this paper described th

Introduction
Telecommunications service providers have to face continuously growing bandwidth demands in all
networks areas, from long-haul to access. Because installing new communication links would require
huge
investments, telecommunications carriers prefer to increase the capacity of their existing fiber links by
using
dense wavelength-division multiplexing (DWDM) systems and/or higher bit rates systems. However,
most of
the installed optical fibers are old and exhibit physical characteristics that may limit their ability to
transmit
high-speed signals.
The broadening of light pulses, called dispersion, is a critical factor limiting the quality of signal
transmission
over optical links. Dispersion is a consequence of the physical properties of the transmission medium.
Single-mode fibers, used in high-speed optical networks, are subject to Chromatic Dispersion (CD) that
causes pulse broadening depending on wavelength, and to Polarization Mode Dispersion (PMD) that
causes pulse broadening depending on polarization. Excessive spreading will cause bits to overflow
their
intended time slots and overlap adjacent bits. The receiver may then have difficulty discerning and
properly
interpreting adjacent bits, increasing the Bit Error Rate. To preserve the transmission quality, the
maximum
amount of time dispersion must be limited to a small proportion of the signal bit rate, typically 10% of
the bit
time.
With optical networks moving from 2.5 Gbps to 10 Gbps and onto 40 Gbps, the acceptable tolerance of
dispersion is drastically reduced. For instance, the amount of acceptable chromatic dispersion decreases
by
a factor of 16 when moving from 2.5 to 10 Gbps, and by an additional factor of 16 moving from 10 to 40
Gbps. These tight tolerances of high-speed networks mean that every possible source of pulse spreading
should be addressed. Operating companies need to measure the dispersion of their networks to assess the
possibility of upgrading them to higher transmission speeds, or to evaluate the need for compensation.
This paper presents the causes and effects of dispersion and describes the different ways to measure it.
Chromatic Dispersion (CD)
CD Definition and Origin
Light within a medium travels at a slower speed than in vacuum. The speed at which light travels is
determined by the mediums refractive index. In an ideal situation, the refractive index would not depend
on
the wavelength of the light. Since this is not the case, different wavelengths travel at different speeds
within
an optical fiber.
Figure 1: CD in single-mode fiber
Laser sources are spectrally thin, but not monochromatic. This means that the input pulse contains several
wavelength components, traveling at different speeds, causing the pulse to spread. The detrimental effects
of chromatic dispersion result in the slower wavelengths of one pulse intermixing with the faster
wavelengths of an adjacent pulse, causing intersymbol interference.
The Chromatic Dispersion of a fiber is expressed in ps/(nm*km), representing the differential delay, or
time
spreading (in ps), for a source with a spectral width of 1 nm traveling on 1 km of the fiber. It depends on
the
fiber type, and it limits the bit rate or the transmission distance for a good quality of service.
CD of current network spans

As a consequence of its optical characteristics, the Chromatic Dispersion of a fiber can be changed by
acting on the physical properties of the material. To reduce fiber dispersion, new types of fiber were
invented, including dispersion-shifted fibers (ITU G.653) and non-zero dispersion-shifted fiber (ITU
G.655).
The most commonly deployed fiber in networks (ITU G.652), called dispersion-unshifted singlemode
fiber,
has a small chromatic dispersion in the optical window around 1310 nm, but exhibits a higher CD in the
1550 nm region. This dispersion limits the possible transmission length without compensation on OC768/STM-256 DWDM networks.
ITU G.653 is a dispersion-shifted fiber (DSF), designed to minimize chromatic dispersion in the 1550 nm
window with zero dispersion between 1525 nm and 1575 nm. But this type of fiber has several
drawbacks,
such as higher polarization mode dispersion than ITU G.652, and a high Four Wave Mixing risk,
rendering
DWDM practically impossible. For these reasons, another singlemode fiber was developed: the Non-Zero
Dispersion-Shifted Fiber (NZDSF). NZDSF fibers have now replaced DSF fibers, which are not used
anymore.
The ITU G.655 Non-Zero Dispersion-Shifted Fibers were developed to eliminate non-linear effects
experienced on DSF fibers. They were developed especially for DWDM applications in the 1550 nm
window.
They have a cut-off wavelength around 1310 nm, limiting their operation around this wavelength.
Poly-chromatic
incident light
1
2
Single-Mode Fiber
Refractive index: n()
Figure 2: Chromatic dispersion profiles of different fiber types
CD Limit and Compensation
The chromatic dispersion in fiber causes a pulse broadening and degrades the transmission quality,
limiting
the distance a digital signal can travel before needing regeneration or compensation. For DWDM systems
using DFB lasers, the maximum length of a link before being affected by chromatic dispersion is
commonly
calculated with the following equation:
2
104,000
CD B
L

=
L is the link distance in km, CD is the chromatic dispersion in ps/(nm * km),
and B is the bit rate in Gbps.
As an example, consider a typical network transmitting data at 10 Gbps on a channel at 1550 nm over a
standard ITU G.652 fiber, having a CD coefficient of 17 ps/(nm*km). In this case, the theoretical
maximum
distance of the link, before adding CD compensators, will be around 61 km, which is close to the typical
node distance in Europe. This length will be divided by 16 when upgrading the network to a 40 Gbps bit
rate,
thus falling under 4 km. This distance is around 20 km for a 40 Gbps transmission through NZDSF. We
can
thus see that CD is a major limiting factor in high-speed transmission.
Fortunately, CD is quite stable, predictable, and controllable. Dispersion Compensation Fiber (DCF),
with its
large negative CD coefficient, can be inserted into the link at regular intervals to minimize its global

chromatic dispersion.
Tx Rx
DC modules
fiber span
delay [ps]
0
L
Figure 3: Chromatic dispersion compensation scheme
While each spool of DCF adequately solves chromatic dispersion for one channel, this is not usually the
case for all channels on a DWDM link. At the extreme wavelengths of a band, dispersion still
accumulates
and can be a significant problem. In this case, a tunable compensation module may be necessary at the
receiver.
Dispersion [ps/nm-km]
1525
1530
1535
1540
1545
1550
1555
1560
SMF
NZDSF
Dispersion
Shifted Fiber
5
0
-5
Wavelength [nm]
10
1550nm DWDM window
1565
Fiber Type 1310 nm 1550 nm
ITU G.652 conventional 0 17
ITU G.653 DSF -15 0
ITU G.655 NZDSF -12 3
Chromatic Dispersion
ps / (nm * km)
delay [ps]
0
1
2
3
Figure 4: Residual chromatic dispersion on a DWDM-compensated link
Therefore, chromatic dispersion measurement is essential in the field to verify the types of installed
fibers.
Such measurements assess if and how the fibers can be upgraded to transmit higher bit rates, verify fiber
zero point and slope for new installations, and carefully evaluate compensation plans.
CD Measurement Methods
In the field, there are three main methods for determining the chromatic dispersion of an optical fiber.
These
are described by three TIA/EIA industry standards: the pulse-delay method (FOTP-168 standard), the
modulated phase-shift method (FOTP-169 standard), and the differential phase shift method (FOTP-175
standard).

These methods all measure first the time delay, in ps, as a function of the wavelength. They then deduce
the chromatic dispersion coefficient, in ps/(nm*km), from the slope of this delay curve and from the
length of
the link. The delay curve plotted by CD analyzers is a fitted trace based on the acquired time delays at
discrete wavelength points. Therefore, the accuracy of the computation of the CD coefficient depends on
the amount of the measured data.
Figure 5: Curve fit to calculate CD coefficient
Phase-shift and differential phase-shift methods are quite similar. In both methods, a modulated source is
injected at the input of the fiber under test. The phase of the sinusoidal modulating signal is analyzed at
the
output of the fiber and compared to the phase of a reference signal, modulated with the same frequency.
In
the phase-shift method, the reference signal has a fixed wavelength, while the other modulated signal is
tuned in wavelengths. In the differential phase-shift method, both signals are tuned in wavelengths, with a
fixed wavelength interval. The analyzed modulated signal, tuned in wavelengths, is compared to a close
reference signal, also tuned in wavelength, but the wavelength gap is constant. The time delay of the link
is
deduced from the phase-shift measurement, using the relationship between the delay (t), the phase (),
and
the modulation frequency (f):
f
t

=
The phase-shift methods assume there is access to both ends of the fiber under test, with a transmitter unit
connected at the input, and a receiver unit at the output of the fiber. The wavelength selection can be
achieved either at the transmitter unit level, using a tunable laser, or at the receiver unit level, with a
broadband source at the input and a wavelength tunable filter in the receiver unit. This filter must be
spectrally thin (FWHM < 1 nm) to ensure low impact on the measurement accuracy. This second
alternative
is commonly used in the differential phase-shift method. One disadvantage of this method is that the
phaseshift
is evaluated comparing two signals with close wavelengths. Increasing the wavelength interval
between the two modulated signals will increase the accuracy of the delay at one individual wavelength
point, but will decrease the number data points that can be acquired to trace the delay curve, leading to a
lower accuracy in the CD coefficient computation. A compromise has to be found between the
wavelength
interval for the differential phase-shift calculation and the number of acquired points to fit the delay
curve,
but achieving this is not straightforward.
Figure 6: Phase-shift principle
Figure 7: Differential phase-shift principle
The phase-shift methods are two-ended solutions, allowing measurement through amplifiers on long-haul
links. They measure the CD with high accuracy, since the wavelengths of modulated signals are known
accurately, and many points of measurements can be acquired, leading to a better fit of the delay curve.
However, these solutions are expensive and thus may not be appropriate for metropolitan applications.
They require operators at both ends of the link, often with bulky instruments.
The pulse-delay methods measure the time that different wavelength carriers travel through the fiber
under
test, either by photon counting (a direct but very complex method), or by measuring the link length with a
multi-wavelength OTDR. The CD-OTDR launches multiple laser pulses into one end of the fiber under
test,

ideally using more than four different wavelengths for better accuracy. It then analyzes the time to return
after a back-reflection from the connector at the other end. The time delay as a function of the wavelength
is
deduced by comparing the times of flight of the laser pulses.
Figure 8: CD-OTDR principle
Broadband
source
Low freq.
oscillator
Fiber under test
Delay
Wavelength
Tunable filter
transmitter receiver
1
2
Tunable
laser
Reference
laser
Low freq.
oscillator
Fiber under test
Demux
transmitter receiver Delay
1
2
End
reflection
The CD-OTDR solution is highly cost-effective, primarily because the initial cost of the instrument is far
lower than that of a phase-shift setup. Furthermore, the CD-OTDR is also a complete 3-wavelength
OTDR
module, which is at the heart of any fiber characterization effort. The CD-OTDR combines two test
setups in
a single, small instrument. It is a one-ended test solution, resulting in less manpower to carry out the test,
less training cost, and lower transportation costs. It can be less accurate than a phase-shift tester, but
accuracy better than 5% can be achieved with a 6-wavelength CD-OTDR. This is sufficient to assess a
links chromatic dispersion, which is a stable and predictable parameter. The CD-OTDR is therefore a
good
alternative to phase-shift testing in metro rings and regional networks.
Table 1: Comparison table of CD measurement methods
Polarization Mode Dispersion (PMD)
PMD Definition and Origin
Testing for Polarization Mode Dispersion (PMD) is becoming essential before upgrading a network to a
higher bit rate because PMD can highly degrade the quality of transmission. It is a difficult parameter to
measure, however, because it varies with time and depends on environmental conditions.
PMD is known to stem from the difference in the propagation constants of a fiber due to geometrical
imperfections in the fiber. The term PMD denotes both the physical phenomenon and the associated
temporal delay. PMD also causes a system penalty because of the associated pulse spreading in a
highspeed
digital transmission system.
The physical origin of PMD is essentially linear birefringence due to core eccentricity and ovalization.
These
appear during the manufacturing process or result from external stresses on the fiber, such as bends and
twists, and can be considered constant over a length called the coupling length. The typical value of the

coupling length is several hundred meters and depends on fiber manufacturing parameters. This means
that for distances that are practical for transmission applications, the actual length of the fiber is much
greater than the coupling length.
The PMD phenomenon is characterized by Differential Group Delay (DGD). DGD is the difference in
propagation time between the two polarization eigenstates, which are the states of polarization with
minimum and maximum propagation time for each wavelength.
Figure 9: DGD of PM and random coupling fibers
Phase-shift method CD-OTDR method
application long-haul link metro & access link
measurement constraint two-ended measurement one-ended measurement
accuracy
good accuracy (depending on
number of acquired points) <5% with 6-wavelength CD-OTDR
cost high low + 3-wavelength OTDR testing capability
In case of weak mode coupling (Polarization Maintaining Fiber--short length of ordinary fiber), the light
polarized along the slow axis arrives later than the light traveling along the fast axis (i.e., the fast and
slow
axes have different indexes of refraction). In this case, the PMD is equal to the DGD. In other cases (long
fiber lengths), the optical fiber acts like many short birefringent elements stacked together and the
alignment
of fast- and slow-axes is random from element to element. Consequently, we speak about random (or
strong) mode coupling. In that case, the DGD varies as a function of wavelength and the PMD, expressed
in
ps, is the average value of the DGD spectral distribution.
The average DGD scales as the square root of the length of the fiber. So the PMD coefficient, expressed
in
ps/km, is often calculated. In addition, the second-order PMD coefficient, in ps/(nm.km), expresses the
PMD dependency with the wavelength.
PMD needs to be tested on the C&L bands. But, depending on the wavelength transmission window of
the
network, there is a need to also test PMD at 1310 nm as PMD values could be different from 1310 nm to
1550 nm.
The Statistical Nature of PMD
For a practical transmission system, DGD determines the system penalty and depends to a large extent on
the wavelength of operation within the operating wavelength band. But DGD also changes with
environmental conditions over time. The next two traces show DGD as a function of wavelength, for the
same fiber at different times.
Figure 10: DGD versus wavelength, on the same fiber, at two different moments
The graphs show that DGD at a particular wavelength changes with time. The variation can be as much
as
1 ps in a few minutes. The general aspect of the plots is nevertheless the same, and the distribution
statistics remain constant. It can be shown that DGD versus wavelength exhibits a Maxwellian
distribution
[1], with a fairly constant mean value over time. The PMD figure is usually taken as the average of the
wavelength distribution. For this fiber, the PMD (average value of DGD) is 0.65 ps.
PMD Limit and Compensation
PMD is inevitable because it is caused by fiber stresses during manufacturing or by environmental
condition
changes. But a small PMD level can be tolerated in networks, depending on the data rate that is intended
to
be transmitted. Generally, the maximum tolerable PMD is typically 10% of the bit time.
Table 2: Bit rate and time, and PMD limit
Since PMD is an unstable phenomenon, it is not easy to compensate for it. However, Dispersion
Compensation Modules (DCM) have been developed for that purpose. DCMs are placed in front of
receivers on the network and can be tuned in dispersion to compensate for the dispersion measured

continuously on a sample of optical pulses.

Regularly measuring PMD in the field is essential in order to evaluate network capacity and assess the
possibility of upgrading networks for higher bit rate transmission.
PMD Measurement Methods
In the field, there are three main methods for determining the polarization mode dispersion of an optical
fiber, described by three TIA/EIA industry standards: the Fixed Analyzer Method (FOTP-113 standard),
the
Jones Matrix Method (FOTP-122 standard), and the Interferometric Method (FOTP-124 standard).
The fixed analyzer method, also called the wavelength scanning method, involves launching
monochromatic polarized light into the fiber under test and measuring the spectral transmission with
another
polarizer.
Figure 11: Fixed analyzer setup
The linear birefringence of a short fiber induces a sinusoidal transmission response when scanning the
wavelength.
Figure 12: Wavelength scanning of a short fiber
The calculation of PMD on a short fiber is straightforward: it is the inverse of the fringe spacing (in
frequency) in the wavelength scanning method. But fiber lengths in networks are much longer than the
coupling length. This changes the aspect of the signal obtained, as illustrated in the example below
showing
a 50 km fiber with a 0.1 ps/ km specific PMD and a 1 km coupling length.
Figure 13: Wavelength scanning of a long fiber
Broadband
source Monochromator
Polarizer Polarization
FUT controller
The most striking feature compared to a short fiber is that the spectral transmission is no longer
sinusoidal.
Around some wavelengths (1530 nm in our example), the fringe separation is narrower, indicating that
the
effect of PMD is higher around these wavelengths. The definition of PMD is not as straightforward as it
was
for short fiber. One method uses fringe counting over the wavelength span. A PMD value can be
established this way, although it is not always clear what should be considered as a fringe. For instance,
should the small peak on the graph around 1555 nm be counted or not? Another method uses the Fourier
transform of the transmission spectrum and is technically very similar to interferometry.
The fixed analyzer method is limited to a 60 ps delay range and is very sensitive to the fiber movement,
thus causing uncertainty. Averaging is required for better accuracy, but this significantly increases the
measurement time. Also, as this method requires step-by-step wavelength scanning, it is necessary to
ensure that the wavelength step is small enough with regard to the PMD to be measured.
The Jones matrix method uses a tunable laser source, a polarization controller, and a polarization
analyzer.
At each wavelength, the polarization controller is scanned and a mathematical calculation is performed
using the Jones matrix formalism (hence the name of the method). From this, the Differential Group
Delay -the maximum delay that can be encountered over polarization -- at that wavelength is derived.
Figure 14: Jones matrix method setup
For a short fiber (shorter than the coupling length), Differential Group Delay is independent of
wavelength.
For long fibers, DGD is no longer independent of wavelength, as illustrated in the example below of a 50
km
fiber with a 0.1 ps/ km specific PMD and a 1 km coupling length. The PMD definition must take this
wavelength dependence into account. PMD is computed as the mean value for DGD over a specified
wavelength range, typically 40 nm around an operating wavelength.
Figure 15: Differential Group Delay versus wavelength

(Jones matrix eigenanalysis) for a long fiber

The Jones matrix method gives an indication of the Differential Group Delay dependence with
wavelength.
Using a narrow band tunable laser, the minimum measurable PMD value is not limited.
As in the fixed analyzer method, the Jones matrix eigenanalysis requires that the fibers not be moved
during the measurement procedure, since this would create measurement artifacts. It is therefore more
suitable for laboratory applications. This method also requires step-by-step wavelength scanning. It is
necessary to ensure that the wavelength step is small enough with regard to the PMD to be measured.
The Jones matrix eigenanalysis is a complex method that requires polarization scanning. The time
required
for testing is usually long: about 10 minutes for each measurement. Finally, it is an expensive solution,
limited to the C&L bands, due to the need to use a tunable laser source.
Tunable Laser
Source
Polarization
controller
Polarimeter
Fiber under test
Computer Differential
Group Delay
The Interferometric Method uses a polarized broadband source and a Michelson scanning interferometer.
The interferometer can be seen as a Fourier transform spectrum analyzer, or more simply as a correlator
for
the PMD-delayed wave trains.
Figure 16: Interferometric method setup
On a short fiber, the interferogram envelope shows a single central peak (which corresponds to the
autocorrelation
peak of the source) and two satellite peaks, as represented in the figure below. The
interferometer unbalance between the central peak and the satellite peaks results in the PMD. The
interferogram can also be seen as the Fourier transform of the transmitted spectrum of the wavelength
scanning method, taken over the spectral range of the source.
Figure 17: Interferogram envelope for a short fiber
For longer randomly mode-coupled fibers, the interferogram is close to a Gaussian function multiplied by
a
noise after removing the auto-correlation peak.
PMD
Figure 18: Interferogram envelope for a long fiber
From a mathematical point of view, the signal is the envelope of the Fourier transform of the transmission
spectrum resulting from the wavelength scanning method. PMD is given by the width of the envelope.
The auto-correlation peak does not contain any information related to PMD measurement. It can be
removed either by computation, or by adding a quarter-wave-plate inside the interferometer to create
destructive interferences when both arms of the interferometer have the same length. This second method
enables a higher dynamic range, resulting in the ability to test longer links, a higher accuracy, and better
repeatability.
Movable
Mirror
Fixed
Mirror
/4
PMD Base Module
Fiber Under Test
This method is well suited for field measurement requirements. The test instruments are rugged and
portable, and can operate on batteries. The interferometric method is insensitive to fiber moves. It
provides
accurate measurements in a few seconds, and does not require averaging. High power broadband sources

can be used to increase the dynamic range, allowing measurement of long links up to 250 km. This
method
also allows PMD measurements through multiple amplifiers such as EDFAs.
The minimum measurable PMD value is limited in this method, depending on the spectral width of the
broadband source. However, sources with FWHM >50 nm will allow PMD measurement down to 0.05
ps,
which is quite acceptable in the field.
The interferometric method has been standardized in two versions: the TINTY (traditional
interferometry)
version, as described above, and the GINTY (generalized interferometry) version. GINTY differs from
TINTY by adding polarization scramblers and an analyzer at the input and the output of the fiber under
test.
This is another way to remove the auto-correlation peak from the interferogram and then measure long
links
having EDFAs. However, because scramblers are used, averaging a large number of measurements is
necessary to achieve the full benefit of this method. As was explained above, technical solutions also
exist
in the TINTY method to remove this central peak using methods other than computation.
Table 3: Comparison table of PMD measurement methods
Conclusion
Dispersion in optical fibers limits the quality of signal transmission. Both CD and PMD must be
measured to
assess the potential of upgrading networks to higher transmission speeds, or to evaluate the need for
compensations. Various testing solutions exist, each having advantages and drawbacks. As an aid to
finding the most appropriate solution, this paper described th
phase-shift principle
laser
Optics
Optical Source
Characteristics
Laser Diode LED
Output Power High Low
Switching Speed Fast Slow
Output Pattern Narrow Wide
Spectral Width Narrow Wide
Operating Constraints More constraints Fewer constraints
Expense More Less
Use Single Mode Multimode
Table 3:
Transceiver Characteristics
Output power refers to the amount of power emitted at a specific drive
Transmitters are designed to emit light at one of three wavelengths: 850
nanometers, 1310 nanometers, and 1550 nanometers. These
wavelengths have extremely low attenuation and therefore are a good
choice for fiber optic communications. Attenuation is loss of optical
power and is measured in decibels.
+ dB = -10log10 = output power
input power
Logarithmic measurement. Small changes in the decibel number
represent large changes in power.
Negative sign indicates loss of signal power.
Positive sign indicates gain in power.
For example; -3dB = 50% power loss, 50% of power remains
-10dB = 90% power loss, 10% of power remains.

Attenuation causes:
Absorption of optical energy by tiny impurities in the fiber such as
iron, copper, or cobalt
The scattering of the light beam as it hits microscopic imperfections,
called Rayleigh scattering
Microbending, which is caused by a nick or dent in the fiber that
disrupts the mode
Macrobending occurs when the fiber is bent beyond its minimum
bend radius
A receiver contains three components: a detector, amplifier, and a
demodulator. The detector converts the optical signal into an electrical
signal, the amplifier boosts or increases the signal strength, and
demodulator extracts the original electrical signal.
When evaluating receivers you need to consider sensitivity and dynamic
range. The sensitivity refers to the minimum signal strength that can be
received. It is a measurement of how much light is required to accurately
detect and decode the data. It is expressed in dBm and is usually a
negative number. The smaller the number, the better the receiver
(i.e. -30 dBM is smaller than -20dBm.
Dynamic range is the range of signal strength the receiver can accept. For
example: if the receiver can accept a signal between -30dBm and 10dBm, the dynamic range is 20dB. Signals that arrive at the receiver
out of the dynamic range of the receiver must be amplified or attenuated
before they can be accepted.
Optics Continued...
Receive sensitivity and transmitter power are used to calculate the optical
power budget available for the cable. The first step in evaluating optical
power budget is determining how much light is available for the
electronic devices. This is accomplished by finding the minimum transmit
power and the minimum receive sensibility. These measurements are
obtained from the equipment manufacturer. The minimum transmit power
is the least amount of transmit power guaranteed by the device. Some
vendors will publish an average transmit power. Be careful using an
average because it does not guarantee the products will perform at that
average level.
To calculate the available light, subtract the minimum receive sensitivity
from the minimum transmit power. The minimum receive sensitivity is
usually a negative number, such as -33dBm. Subtracting a negative
number is the same as adding its absolute value. For example, if a
device has a minimum transmit power equal to -10 dBm and a
minimum receive sensitivity of -33 dBm, the available power will be
23 dBm.
Available light = minimum transmit power - minimum receive sensitivity
= -10 dBm - (-33 dBm)
= 23 dBm
When connecting devices from different vendors or different product
models, the available power calculation needs to be determined for both
directions. The smaller of the two calculations should be used for the
amount of available light to ensure performance.
Once the available light has been calculated, all the loss factors need to
be subtracted out. Losses can stem from cable attenuation, connector
loss, and cable splices. Cable attenuation is the most significant loss and
is determined by using the manufacturers worst case loss factor for the
type of cable being installed. This number will range from .22 dB to .5
dB per kilometer. Multiply this number by the number of kilometers. A

fiber with .4dB per kilometer of loss will lose 16 dB over a distance of 40
kilometers.
Fiber over a certain length will require splicing so you'll need to include
additional loss for splicing. Fiber installers provide a worse case loss
number for your calculation. Typically, each splice will introduce .1 dB of
additional loss. Multiply this number times the number of splices in your
fiber.
Light loss for connectors is another loss factor to consider in your
calculation. The exact number of connectors for the network needs to be
determined. Connector loss is provided by the connector manufacturer
and the installer. Multiply the total number of connectors by the loss per
connector.
10 Fiber Optic Basics
Optical Power Budgets
Conclusion
Fiber Optic Basics 11
Each of the factors is subtracted from the original light availability. If the
number is negative there is not enough power to drive the performance of
the network. If the number is positive, you still need to have a buffer for
anticipated repairs (additional splices in the network) and temperature
extremes. This is typically down by using a safety factor in your
calculation. The number differs per organization, but typically a value
approximately 3 dB is used. It acts as a buffer to your power and guards
against unforeseen factors affecting your optical power budget.
The table below contains some typical numbers used to approximate
optical link budget. Real numbers from specific equipment manufactures
and your network should be used if at all possible.
Optical Power Budgets Continued...
TIA Standard for connector loss .75 dB
Typical cable attenuation at 1310 nm .4 dB/km
Typical cable attenuation at 1550 nm .3 dB/km
Typical cable attenuation at 850 n m 3.75 dB/km
Typical splice attenuation .1 dB/km
Typical distance between splices 6 km
Typical safety margin 3 dB
Other factors to include are the quality of the cables, connectors, and the
quality of the splices. Transition Networks provides a link budget
calculator on our website at
www.transition.com/learning/whitepapers/pwrbud_wp.htm
Although the widespread use of fiber began with the push from the
telecommunications industry, today it is commonplace. Many enterprises
take advantage of fiber to increase the capacity and functionality of their
local area networks (LANs) and now metropolitan area networks (MANs).
One issue faced by some enterprises is how to connect legacy equipment
and infrastructure without expensive "forklift" upgrades. By using copper
to fiber media converters or multimode to single mode media converters,
fiber can be connected in almost any legacy environment.
Transition Networks comprehensive line of media conversion products
are designed to ease the migration to fiber, while minimizing cost and
installation issues.
Names given to different rays
We have seen that rays approaching from within the cone of acceptance are
successfully propagated along the fiber.

The position and the angle at which the ray strikes the core will determine the
exact path taken by the ray. There are three possibilities, called the skew,
Propagation of light along the fiber
35
Figure 4.29
...and back into
the fiber
Figure 4.30
Absorption
Any impurities that remain in the fiber after manufacture will block some of the
light energy. The worst culprits are hydroxyl ions and traces of metals.
The hydroxyl ions are actually the form of water which caused the large losses
at 1380 nm that we saw in Figure 3.2. In a similar way, metallic traces can cause
absorption of energy at their own particular wavelengths. These small absorption
peaks are also visible.
In both cases, the answer is to ensure that the glass is not contaminated at the
time of manufacture and the impurities are reduced as far as possible. We are
aiming at maximum levels of 1 part in 109 for water and 1 part in 1010 for the
metallic traces.
Now for the second reason, the diversion of the light.
Rayleigh scatter
This is the scattering of light due to small localized changes in the refractive
index of the core and the cladding material. The changes are indeed very localized.
We are looking at dimensions which are less than the wavelength of the
50
6
Losses in optic fibers
light.
There are two causes, both problems within the manufacturing processes.
The first is the inevitable slight fluctuations in the mix of the ingredients. These
random changes are impossible to completely eliminate. It is a bit like making
a currant bun and hoping to stir it long enough to get all the currants equally
spaced.
The other cause is slight changes in the density as the silica cools and solidifies.
One such discontinuity is illustrated in Figure 6.1 and results in light being
scattered in all directions. All the light that now finds itself with an angle of
incidence less than the critical angle can escape from the core and is lost.
However, much of the light misses the discontinuity because it is so small. The
scale size is shown at the bottom.
The amount of scatter depends on the size of the discontinuity compared with
the wavelength of the light so the shortest wavelength, or highest frequency,
suffers most scattering. This accounts for the blue sky and the red of the sunset.
The high frequency end of the visible spectrum is the blue light and this is
scattered more than the red light when sunlight hits the atmosphere. The sky is
only actually illuminated by the scattered light. So when we look up, we see the
blue scattered light, and the sky appears blue. The moon has no atmosphere,
no scattering, and hence a black sky. At sunset, we look towards the sun and
see the less scattered light which is closer to the sun. This light is the lower
frequency red light.
Bending losses
Macrobends
A sharp bend in a fiber can cause significant losses as well as the possibility of
mechanical failure.
It is easy to bend a short length of optic fiber to produce higher losses than a

whole kilometer of fiber in normal use.

The ray shown in Figure 6.4 is safely outside of the critical angle and is therefore
propagated correctly.
Remember that the normal is always at right angles to the surface of the core.
Now, if the core bends, as in Figure 6.5, the normal will follow it and the ray
will now find itself on the wrong side of the critical angle and will escape.
Tight bends are therefore to be avoided but how tight is tight?
The real answer to this is to consult the specification of the fiber optic cable in
use as the manufacturer will consider the mechanical limitations as well as the
bending losses. However a few general indications may not be out of place.
A bare fiber and by this is meant just the core/cladding and the primary buffer
is safe if the radius of the bend is at least 50 mm. For a cable, which is the
bare fiber plus the outer protective layers, make it about 10 _ outside diameter
or 50 mm whichever is the greater (Figure 6.6).
The tighter the bend, the worse the losses. Shown full size, the results obtainedwith
a single sample of bare fiber were shown in Figure 6.7, attached instruments
indicated a loss of over 6 dB before it broke.
The problem of macrobend loss is largely in the hands of the installer.
Losses in optic fibers
53
Introduction to Fiber Optics
54
Figure 6.4
The usual situation
Figure 6.5
Sharp bends are
bad news
Figure 6.6
Minimum safe
bend radius
shown full size
Making use of bending losses
There are many uses of bending losses which are based on either the increase
in the attenuation or on making use of the light which escapes from the optic
fiber.
Microbends
These are identical in effect to the macrobend already described but differ in size and cause. Their radius
is equal to, or less than, the diameter of the bare fiber very small indeed (Figure 6.11). These are
generally a manufacturing problem. A typical cause is differential expansion of the optic fiber and the
outer layers. If the fiber gets too cold, the
Introduction to Fiber Optics
56
Figure 6.10
Is the fiber in use?
outer layers will shrink and get shorter. If the core/cladding shrinks at a slower rate, it is likely to kink
and cause a microbend.
With careful choice of the fiber to be installed, these are less likely to be a problem than the bending
losses caused during installation since fiber optic cables are readily available with a wide range of
operating temperatures from 55C to +85C.
Quiz time
The effect on the data
The data can be corrupted by dispersion. If we send a sequence of ON OFF ON
pulses, it would start its life as an electronic signal with nice sharp edges as in
Figure 7.3.

These pulses are used to switch a light source, usually an LED or a laser and the
resultant pulses of light are launched into the fiber.
Dispersion causes the pulses to spread out and eventually they will blend
together and the information will be lost (Figure 7.4).
We could make this degree of dispersion acceptable by simply decreasing the
transmission frequency and thus allowing larger gaps between the pulses.
This type of dispersion in called intermodal dispersion. At this point, we will
take a brief detour to look at the idea of modes.
Introduction to Fiber Optics
60
Figure 7.1
Ray B will arrive
first
Figure 7.2
The pulse spreads
out
Figure 7.3
Electronic pulses
controlling the
light source
Figure 7.4
Dispersion has
caused the pulses
to merge
Modes
The light traveling down the fiber is a group of electromagnetic (EM) waves
occupying a small band of frequencies within the electromagnetic spectrum, so
it is a simplification to call it a ray of light. However, it is enormously helpful to
do this, providing an easy concept, some framework to hang our ideas on. We
do this all the time and it serves us well providing we are clear that it is only an
analogy. Magnetic fields are not really lines floating in space around a magnet,
electrons are not really little black ball bearings flying round a red nucleus.
Light therefore, is propagated as an electromagnetic wave along the fiber. The
two components, the electric field and the magnetic field form patterns across
the fiber. These patterns are called modes of transmission. Modes means
methods hence methods of transmission. An optic fiber that carries more than
one mode is called a multimode fiber (MM).
The number of modes is always a whole number.
In a given piece of fiber, there are only a set number of possible modes. This is
because each mode is a pattern of electric and magnetic fields having a physical
size. The dimensions of the core determine how many modes or patterns
can exist in the core the larger the core, the more modes.
The number of modes is always an integer, we cannot have incomplete field
patterns. This is similar to transmission of motor vehicles along a road. As the
road is made wider, it stays as a single lane road until it is large enough to
accommodate an extra line of vehicles whereupon it suddenly jumps to a two
lane road. We never come across a 1.15 lane road!
How many modes are there?
The number of modes is given (reasonably accurately) by the formula:
Number of modes =
where NA is numerical aperture of the fiber and _ is the wavelength of the light
source.
Lets choose some likely figures as in Figure 7.5 and see the result.
Method
Find the numerical aperture

Dispersion and our attempts to prevent it

61
Figure 7.5
How many modes
are in the core?
So:
NA = 0.203 (a typical result)
Insert the figures into the formula:
Tap it into a calculator and see what comes out:
number of modes = 703.66
The calculator gave 703.66 but we cannot have part of a mode so we have to round it down. Always
round down as even 703.99 would not be large enough for 704 modes to exist.
Each of the 703 modes could be represented by a ray being propagated at its
own characteristic angle. Every mode is therefore traveling at a different speed
along the fiber and gives rise to the dispersion which we called intermodal
dispersion.
How to overcome intermodal dispersion
We can approach the problem of intermodal dispersion in two ways. We could
redesign the fiber to encourage the modes to travel at the same speed along the
fiber or we can eliminate all the modes except one it can hardly travel at a
different speed to itself! The first strategy is called graded index optic fiber.
Intramodal (or chromatic) dispersion
Unfortunately intermodal dispersion is not the only cause of dispersion.
We know that light of different wavelengths is refracted by differing amounts.
This allows rain drops to split sunlight into the colors of the rainbow. We are
_Diameter of core _ NA _ _
_
_
__2
____
2
Introduction to Fiber Optics
64
really saying that the refractive index, and hence the speed of the light, is determined
to some extent by its wavelength.
A common fallacy is that a laser produces light of a single wavelength. In fact
it produces a range of wavelengths even though it is far fewer than is produced
by the LED, the alternative light source (Figure 7.10).
This is unfortunate as each component wavelength travels at a slightly different
speed in the fiber. This causes the light pulse to spread out as it travels along
the fiber and hence causes dispersion. The effect is called chromatic dispersion.
Actually, chromatic dispersion is the combined effect of two other dispersions
material dispersion and waveguide dispersion. Both result in a change in
transmission speed, the first is due to the atomic structure of the material and
the second is due to the propagation characteristics of the fiber. Any further
investigation will not help us at the moment and is not pursued further.
Dispersion and our attempts to prevent it
65
Figure 7.9
Its easy to see the
difference
Figure 7.10
The laser has a
narrow spectral
width

One interesting feature of chromatic dispersion is shown in Figure 7.11. The

value of the resulting dispersion is not constant and passes through an area of
zero dispersion. This cannot be used to eliminate dispersion altogether because
the zero point only occurs at a single wavelength, and even a laser produces a
range of wavelengths within its spectrum.
By fiddling about with the dimensions of the core and the constituents of the
fiber, we can adjust the wavelength of the minimum dispersion point. This is
called optimizing the fiber for a particular wavelength or window.
Be careful not to muddle intermodal dispersion with intramodal dispersion.
_ inter means between. Intermodal between modes
_ intra means within. Intramodal within a single mode
The alternative name of chromatic, to do with color or frequency, is less confusing.
Although chromatic dispersion is generally discussed in terms of a single mode
fiber it does still occur in multimode fiber but the effect is generally swamped
by the intermodal dispersion.
LeakyMode
In an optical fiber, a mode having a field that decays monotonically for a finite distance in the
transverse direction but becomes oscillatory everywhere beyond that finite distance.
Note: Leaky Modes correspond to leaky rays in the terminology of geometric optics. Leaky modes
experience attenuation, even if the Waveguide is perfect in every respect. Synonym: tunneling
mode.
QOPTICAL FIBER COUPLING LOSS
Ideally, optical signals coupled between fiber optic components are transmitted with no loss of light.
However, there is always some type of imperfection present at fiber optic connections that causes some
loss of light. It is the amount of optical power lost at fiber optic connections that is a concern of system
designers.
The design of fiber optic systems depends on how much light is launched into an optical fiber from an
optical source and how much light is coupled between fiber optic components, such as from one fiber to
another. The amount of power launched from a source into a fiber depends on the optical properties of
both the source and the fiber. The amount of optical power launched into an optical fiber depends on the
radiance of the optical source. An optical source's radiance, or brightness, is a measure of its optical
power launching capability. Radiance is the amount of optical power emitted in a specific direction per
unit time by a unit area of emitting surface. For most types of optical sources, only a fraction of the
power emitted by the source is launched into the optical fiber.
The loss in optical power through a connection is defined similarly to that of signal attenuation through a
fiber. Optical loss is also a log relationship. The loss in optical power through a connection is defined as:

For example, Po is the power emitted from the source fiber in a fiber-to-fiber connection. Pi is the power
accepted by the connected fiber. In any fiber optic connection, P o and Pi are the optical power levels
measured before and after the joint, respectively.
Fiber-to-fiber connection loss is affected by intrinsic and extrinsic coupling losses. Intrinsic coupling
losses are caused by inherent fiber characteristics. Extrinsic coupling losses are caused by jointing
techniques. Fiber-to-fiber connection loss is increased by the following sources of intrinsic and extrinsic
coupling loss:
Reflection losses
Fiber separation
Lateral misalignment
Angular misalignment

Core and cladding diameter mismatch

Numerical aperture (NA) mismatch
Refractive index profile difference
Poor fiber end preparation
Intrinsic coupling losses are limited by reducing fiber mismatches between the connected fibers. This is
done by procuring only fibers that meet stringent geometrical and optical specifications. Extrinsic
coupling losses are limited by following proper connection procedures.

Some fiber optic components are modular devices that are designed to reduce coupling losses between
components. Modular components can be easily inserted or removed from any system. For example, fiber
optic transmitters and receivers are modular components. Fiber optic transmitters and receivers are
devices that are generally manufactured with fiber pigtails or fiber optic connectors as shown in figure 41. A fiber pigtail is a short length of optical fiber (usually 1 meter or less) permanently fixed to the
optical source or detector. Manufacturers supply transmitters and receivers with pigtails and connectors
because fiber coupling to sources and detectors must be completed during fabrication. Reduced coupling
loss results when source-to-fiber and fiber-to-detector coupling is done in a controlled manufacturing
environment. Since optical sources and detectors are pigtailed or connectorized, launching optical power
is reduced to coupling light from one fiber to another. In fact, most fiber optic connections can be
considered fiber-to-fiber.
Figure 4-1. - Pigtailed and connectorized fiber optic devices.

Q.3 Define the loss in optical power through a connection.

Q.4 Fiber-to-fiber coupling loss is affected by intrinsic and extrinsic coupling losses. Can intrinsic
coupling losses be limited by limiting fiber mismatches?
LASER DIODES
A laser is a device that produces optical radiation by the process of stimulated emission. It is necessary to
contain photons produced by stimulated emission within the laser active region.
Figure 6-3 shows an optical cavity formed to contain the emitted photons by placing one reflecting mirror
at each end of an amplifying medium. One mirror is made partially reflecting so that some radiation can
escape from the cavity for coupling to an optical fiber.
Figure 6-3. - Optical cavity for producing lasing.

Only a portion of the optical radiation is amplified. For a particular laser structure, there are only certain
wavelengths that will be amplified by that laser. Amplification occurs when selected wavelengths, also
called laser modes, reflect back and forth through the cavity. For lasing to occur, the optical gain of the

selected modes must exceed the optical loss during one round-trip through the cavity. This process is
referred to as optical feedback.
The lasing threshold is the lowest drive current level at which the output of the laser results primarily
from stimulated emission rather than spontaneous emission. Figure 6-4 illustrates the transition from
spontaneous emission to stimulated emission by plotting the relative optical output power and input drive
current of a semiconductor laser diode. The lowest current at which stimulated emission exceeds
spontaneous emission is the threshold current.
Before the threshold current is reached, the optical output power increases only slightly with small
increases in drive current. However, after the threshold current is reached, the optical output power
increases significantly with small changes in drive currents.
Figure 6-4. - The optical output power as a function of input drive current of a semiconductor laser diode.

Many types of materials including gas, liquid, and semiconductors can form the lasing medium.
However, in this chapter we only discuss semiconductor laser diodes. Semiconductor laser diodes are the
primary lasers used in fiber optics. A laser diode emits light that is highly monochromatic and very
directional.
This means that the LD's output has a narrow spectral width and small output beam angle.
A semiconductor LD's geometry is similar to an ELED with light-guiding regions surrounding the active
region. Optical feedback is established by making the front facet partially reflective. This chapter
provides no diagram detailing LD structures because they are similar to ELEDs in design. The rear facet
is typically coated with a reflective layer so that all of the light striking the facet is reflected back into the
active region. The front facet is typically left uncoated so that most of the light is emitted. By increasing
the drive current, the diode becomes a laser.
At currents below the threshold current, LDs function as ELEDs.
To optimize Frequency response, laser diodes are often biased above this laser threshold. As a result, in
an LD fiber optic system, light is modulated between a high power level and a lower power level, but
never shut off. LDs typically can be modulated at frequencies up to over 2 gigahertz (GHz). Some lasers
are capable of being modulated at frequencies over 20 GHz.
There are several important differences between LDs and LEDs. One is that LEDs usually lack reflective
facets and in some cases are designed to suppress reflections back into the active region. Another is that
lasers tend to operate at higher drive currents to produce light. A higher driver current results in more
complicated drive circuits and more heat dissipation in the device.
LDs are also much more temperature sensitive than either SLEDs or ELEDs. Increases in the laser
temperature significantly reduce laser output power. Increases in laser temperature beyond certain limits
result in the loss of lasing. When lasers are used in many applications, the temperature of the laser must
be controlled. Typically, electronic coolers, called thermo-electric (TE) coolers, are used to cool LDs in
system applications.

[ Back ] [ Home ] [ Up ] [ Next ]

Leaky mode
From Wikipedia, the free encyclopedia
Jump to: navigation, search
A leaky mode or tunneling mode in an optical fiber or other waveguide is a mode having an electric
field that decays monotonically for a finite distance in the transverse direction but becomes oscillatory
everywhere beyond that finite distance. Such a mode gradually "leaks" out of the waveguide as it travels
down it, producing attenuation even if the waveguide is perfect in every respect. In order for a leaky
mode to be definable as a mode, the relative amplitude of the oscillatory part (the leakage rate) must be
sufficiently small that the mode substantially maintains its shape as it decays.
Leaky modes correspond to leaky rays in the terminology of geometric optics. The propagation of light
through optical fibre can take place via meridional rays or skew rays. These skew rays suffer only partial
reflection while meridional rays are completely guided. Thus the modes allowing propagation of skew
rays are called leaky modes. Some optical power is lost into clad due to these modes.
Coupling loss
From Wikipedia, the free encyclopedia
Jump to: navigation, search
This article does not cite any references or sources. Please help improve this article by adding
citations to reliable sources. Unsourced material may be challenged and removed. (December
2009)
Coupling loss also known as connection loss is the loss that occurs when energy is transferred from one
circuit, circuit element, or medium to another. Coupling loss is usually expressed in the same unitssuch
as watts or decibelsas in the originating circuit element or medium.
Coupling loss in fiber optics refers to the power loss that occurs when coupling light from one optical
device or medium to another. (See also Optical return loss.)
Coupling losses can result from a number of factors. In electronics (see Coupling (electronics)),
impedance mismatch between coupled components results in a reflection of a portion of the energy at the
interface. Likewise, in optical systems, where there is a change in index of refraction (most commonly at
a fiber/air interface), a portion of the energy is reflected back into the source component.
Another major source of optical coupling loss is geometrical. As an example, two fibers coupled end-toend may not be precisely aligned, with the result that the two cores overlap somewhat. Light exiting the
source fiber at a portion of its core that is not aligned with the core of the receiving fiber will not (in
general) be coupled into the second fiber. While some such light will be coupled into the second fiber, it
is not likely to be efficiently coupled, nor will it generally travel in an appropriate mode in the second
fiber.
Similarly, even for two perfectly aligned cores, where there is a gap of any significant distance between
the two fibers, there will be some geometric loss due to spread of the beam. Some percentage of the light
rays exiting the source fiber face will not intersect the second fiber within its entrance cone.
This electronics-related article is a stub. You can help Wikipedia by expanding it.
Retrieved from "http://en.wikipedia.org/wiki/Coupling_loss"
View page ratings
Rate this page
I am highly knowledgeable about this topic (optional)
Submit ratings
Saved successfully
Your ratings have not been submitted yet
Categories: Electronics terms | Fiber optics | Electronics stubs
Hidden categories: Articles lacking sources from December 2009 | All articles lacking sources
Fiber Optic Coupling
Contact Us Print View Feedback
The problem of coupling light into an optical fiber is really two separate problems. In one case, we have
the problem of coupling into multimode fibers, where the ray optics of the previous section can be used.

In the other case, coupling into single-mode fibers, we have a fundamentally different problem. In this
case, one must consider the problem of matching the mode of the incident laser light into the mode of the
fiber. This cannot be done using the ray optics approach, but must be done using the concepts of Gaussian
beam optics (see Gaussian Beam Optics).
Application 5: Coupling Laser Light into a Multimode Fiber
When we look at coupling light from a well-collimated laser beam into a multimode optical fiber, we
return to the situation that was illustrated in Figure 5. The radius of the fiber core will be our y2. We will
have to make sure that the lens focuses to a spot size less than this parameter. An even more important
restriction is that the angle from the lens to the fiber 2 must be less than the NA of the optical fiber.
Lets consider coupling the light from a Newport R-30990 HeNe laser into an F-MSD fiber. The laser has
a beam diameter of 0.81 mm and divergence 1.0 mrad. The fiber has a core diameter of 50 m and an NA
of 0.20. Lets look at the coupling from the beam into the fiber when a Newport M-20X objective lens is
used in an F-915 or F-915T fiber coupler.
The objective lens has an effective focal length of 9 mm. In this case, the focused beam will have a
diameter of 9 m and a maximal ray of angle 0.05, so both the spot diameter and angle are well within
the parameters of the fiber.
Generally, coupling light from a well-collimated laser source into a multimode fiber is not a difficult
problem. If the user assures that the maximal ray of the focused beam is well within the NA of the fiber,
then effective coupling will be accomplished.
Application 6: Coupling Light from a Diffuse Source into a Multimode Fiber
The previous application involved focusing a collimated source onto the end face of a fiber. This new
application involves imaging the light from an extended diffuse source onto the fibers end face. This is
similar to Application 4 where we focused the light from an extended source to a small spot. In the
present application, we use a fiber to collect light from a source. By diffuse, we mean that the source
emits light in all directions and that the intensity of light is independent of direction. An example of this
is an incandescent lamp or a fluorescing sample.

Figure 9
Figure 9 shows the situation where we do 1:1 imaging from the source into the fiber. From the Gaussian
lens equation, we have s1 = s2 = 2f in this configuration. Since y2 is simply the radius of the receiving
fiber, the light is collected only from an area of the source of radius y1 = y2. The maximum angle coupled
into the fiber is restricted by 2 = NA. Since this is a 1:1 imaging problem, the maximum angle sampled
from the light source is then also 1 = NA. Thus, it is the fiber and not the lens that puts the restriction on
how much light is collected. As long as we choose a lens with sufficiently small f/#, that is f/# <1/(2NA),
then we will collect the maximum possible amount of light from the source into the fiber.
Note what happens if we remove the lens and move the fiber directly against the source. That is, we butt
couple the fiber to the source. In that case, the fiber collects light from an area of radius y2 and angle NA.
But, this is exactly what we accomplished in our 1:1 imaging. An imaging system collecting light from a
diffuse source into a fiber cannot collect more than could be collected by butt coupling.
In fact, this is a completely general result. Consider a configuration more like Figure 7, where we might
choose, for example, to condense light onto the fiber end face from an area of the source with y1 = 5y2.
We are now collecting light from an area that is 25x larger than the area of the fiber. However, the optical
invariant tells us that we must now have 1 = NA/5, so we are collecting light from a solid angle that is
reduced by the same factor of 25. Thus, the total light that is collected, irradiance x area x solid angle, is
constant. What a lens system can achieve is only to retrieve the efficiency of butt coupling when the fiber
must be placed at a distance from a diffuse source. Therefore, for maximum efficiency, choose a fiber
with the largest possible core diameter and the largest available numerical aperture.
Application 7: Coupling Laser Light into a Single-Mode Fiber
When we need to couple laser light into a single-mode fiber, we move from the ray optics picture in
which we have worked to this point to a Gaussian mode-matching problem. The subject of Gaussian laser
beam optics is reviewed beginning on Gaussian Beam Optics. This application is included here for
completeness in discussing coupling light into optical fibers.

In order to couple light of wavelength from a collimated laser beam of 1/e2 diameter D into a fiber of
mode field diameter , choose a lens with a focal length
f = D(/4)
Lets consider the situation where we use a Newport F-915 fiber coupler to couple light from a Newport
R-30992 laser ( = 633 nm, D = 1.2 mm) into a Newport F-SV fiber ( = 4.3 m). We find f = 6.4 mm.
For this application, use the Newport M-20X objective, f = 9 mm, as the closest fit to the correct focal
length.
The coupling efficiency depends upon the overlap integral of the Gaussian mode of the input laser beam
and the nearly Gaussian fundamental mode of the fiber. This overlap integral is the same whether the
input mode is the larger or the smaller of the two modes. The focal length of the M-20X is too large by a
factor of 1.4 while the focal length of the M-40X too short by a factor of 0.7, so the M-20X will be the
better fit for this application.
Fiber Optic Coupling
Contact Us Print View Feedback
The problem of coupling light into an optical fiber is really two separate problems. In one case, we have
the problem of coupling into multimode fibers, where the ray optics of the previous section can be used.
In the other case, coupling into single-mode fibers, we have a fundamentally different problem. In this
case, one must consider the problem of matching the mode of the incident laser light into the mode of the
fiber. This cannot be done using the ray optics approach, but must be done using the concepts of Gaussian
beam optics (see Gaussian Beam Optics).
Application 5: Coupling Laser Light into a Multimode Fiber
When we look at coupling light from a well-collimated laser beam into a multimode optical fiber, we
return to the situation that was illustrated in Figure 5. The radius of the fiber core will be our y2. We will
have to make sure that the lens focuses to a spot size less than this parameter. An even more important
restriction is that the angle from the lens to the fiber 2 must be less than the NA of the optical fiber.
Lets consider coupling the light from a Newport R-30990 HeNe laser into an F-MSD fiber. The laser has
a beam diameter of 0.81 mm and divergence 1.0 mrad. The fiber has a core diameter of 50 m and an NA
of 0.20. Lets look at the coupling from the beam into the fiber when a Newport M-20X objective lens is
used in an F-915 or F-915T fiber coupler.
The objective lens has an effective focal length of 9 mm. In this case, the focused beam will have a
diameter of 9 m and a maximal ray of angle 0.05, so both the spot diameter and angle are well within
the parameters of the fiber.
Generally, coupling light from a well-collimated laser source into a multimode fiber is not a difficult
problem. If the user assures that the maximal ray of the focused beam is well within the NA of the fiber,
then effective coupling will be accomplished.
Application 6: Coupling Light from a Diffuse Source into a Multimode Fiber
The previous application involved focusing a collimated source onto the end face of a fiber. This new
application involves imaging the light from an extended diffuse source onto the fibers end face. This is
similar to Application 4 where we focused the light from an extended source to a small spot. In the
present application, we use a fiber to collect light from a source. By diffuse, we mean that the source
emits light in all directions and that the intensity of light is independent of direction. An example of this
is an incandescent lamp or a fluorescing sample.

Figure 9
Figure 9 shows the situation where we do 1:1 imaging from the source into the fiber. From the Gaussian
lens equation, we have s1 = s2 = 2f in this configuration. Since y2 is simply the radius of the receiving
fiber, the light is collected only from an area of the source of radius y1 = y2. The maximum angle coupled
into the fiber is restricted by 2 = NA. Since this is a 1:1 imaging problem, the maximum angle sampled
from the light source is then also 1 = NA. Thus, it is the fiber and not the lens that puts the restriction on
how much light is collected. As long as we choose a lens with sufficiently small f/#, that is f/# <1/(2NA),
then we will collect the maximum possible amount of light from the source into the fiber.

Note what happens if we remove the lens and move the fiber directly against the source. That is, we butt
couple the fiber to the source. In that case, the fiber collects light from an area of radius y2 and angle NA.
But, this is exactly what we accomplished in our 1:1 imaging. An imaging system collecting light from a
diffuse source into a fiber cannot collect more than could be collected by butt coupling.
In fact, this is a completely general result. Consider a configuration more like Figure 7, where we might
choose, for example, to condense light onto the fiber end face from an area of the source with y1 = 5y2.
We are now collecting light from an area that is 25x larger than the area of the fiber. However, the optical
invariant tells us that we must now have 1 = NA/5, so we are collecting light from a solid angle that is
reduced by the same factor of 25. Thus, the total light that is collected, irradiance x area x solid angle, is
constant. What a lens system can achieve is only to retrieve the efficiency of butt coupling when the fiber
must be placed at a distance from a diffuse source. Therefore, for maximum efficiency, choose a fiber
with the largest possible core diameter and the largest available numerical aperture.
Application 7: Coupling Laser Light into a Single-Mode Fiber
When we need to couple laser light into a single-mode fiber, we move from the ray optics picture in
which we have worked to this point to a Gaussian mode-matching problem. The subject of Gaussian laser
beam optics is reviewed beginning on Gaussian Beam Optics. This application is included here for
completeness in discussing coupling light into optical fibers.
In order to couple light of wavelength from a collimated laser beam of 1/e2 diameter D into a fiber of
mode field diameter , choose a lens with a focal length
f = D(/4)
Lets consider the situation where we use a Newport F-915 fiber coupler to couple light from a Newport
R-30992 laser ( = 633 nm, D = 1.2 mm) into a Newport F-SV fiber ( = 4.3 m). We find f = 6.4 mm.
For this application, use the Newport M-20X objective, f = 9 mm, as the closest fit to the correct focal
length.
The coupling efficiency depends upon the overlap integral of the Gaussian mode of the input laser beam
and the nearly Gaussian fundamental mode of the fiber. This overlap integral is the same whether the
input mode is the larger or the smaller of the two modes. The focal length of the M-20X is too large by a
factor of 1.4 while the focal length of the M-40X too short by a factor of 0.7, so the M-20X will be the
better fit for this application.
Attenuation
Attenuation in an optical fiber is caused by absorption, scattering, and bending losses. Attenuation is the
loss of optical power as light travels along the fiber. Signal attenuation is defined as the ratio of optical
input power (Pi) to the optical output power (Po). Optical input power is the power injected into the fiber
from an optical source. Optical output power is the power received at the fiber end or optical detector.
The following equation defines signal attenuation as a unit of length:

Signal attenuation is a log relationship. Length (L) is expressed in kilometers. Therefore, the unit of
attenuation is decibels/kilometer (dB/km). As previously stated, attenuation is caused by absorption,
scattering, and bending losses. Each mechanism of loss is influenced by fiber-material properties and
fiber structure. However, loss is also present at fiber connections. Fiber connector, splice, and coupler
losses are discussed in chapter 4. The present discussion remains relative to optical fiber attenuation
properties.
Q.38 Define attenuation.
ABSORPTION. - Absorption is a major cause of signal loss in an optical fiber. Absorption is defined as
the portion of attenuation resulting from the conversion of optical power into another energy form, such
as heat. Absorption in optical fibers is explained by three factors:
Imperfections in the atomic structure of the fiber material
The intrinsic or basic fiber-material properties
The extrinsic (presence of impurities) fiber-material properties
Imperfections in the atomic structure induce absorption by the presence of missing molecules or oxygen
defects. Absorption is also induced by the diffusion of hydrogen molecules into the glass fiber. Since
intrinsic and extrinsic material properties are the main cause of absorption, they are discussed further.

Intrinsic Absorption. - Intrinsic absorption is caused by basic fiber-material properties. If an optical

fiber were absolutely pure, with no imperfections or impurities, then all absorption would be intrinsic.
Intrinsic absorption sets the minimal level of absorption.
In fiber optics, silica (pure glass) fibers are used predominately. Silica fibers are used because of their low
intrinsic material absorption at the wavelengths of operation.
In silica glass, the wavelengths of operation range from 700 nanometers (nm) to 1600 nm. Figure 2-21
shows the level of attenuation at the wavelengths of operation. This wavelength of operation is between
two intrinsic absorption regions. The first region is the ultraviolet region (below 400-nm wavelength).
The second region is the infrared region (above 2000-nm wavelength).
Figure 2-21. - Fiber losses.
Intrinsic absorption in the ultraviolet region is caused by electronic absorption bands. Basically,
absorption occurs when a light particle (photon) interacts with an electron and excites it to a higher
energy level. The tail of the ultraviolet absorption band is shown in figure 2-21.
The main cause of intrinsic absorption in the infrared region is the characteristic vibration frequency of
atomic bonds. In silica glass, absorption is caused by the vibration of silicon-oxygen (Si-O) bonds. The
interaction between the vibrating bond and the electromagnetic field of the optical signal causes intrinsic
absorption. Light energy is transferred from the electromagnetic field to the bond. The tail of the infrared
absorption band is shown in figure 2-21.
Extrinsic Absorption. - Extrinsic absorption is caused by impurities introduced into the fiber material.
Trace metal impurities, such as iron, nickel, and chromium, are introduced into the fiber during
fabrication. Extrinsic absorption is caused by the electronic transition of these metal ions from one
energy level to another.
Extrinsic absorption also occurs when hydroxyl ions (OH-) are introduced into the fiber. Water in silica
glass forms a silicon-hydroxyl (Si-OH) bond. This bond has a fundamental absorption at 2700 nm.
However, the harmonics or overtones of the fundamental absorption occur in the region of operation.
These harmonics increase extrinsic absorption at 1383 nm, 1250 nm, and 950 nm. Figure 2-21 shows the
presence of the three OH- harmonics. The level of the OH- harmonic absorption is also indicated.
These absorption peaks define three regions or windows of preferred operation. The first window is
centered at 850 nm. The second window is centered at 1300 nm. The third window is centered at 1550
nm. Fiber optic systems operate at wavelengths defined by one of these windows.
The amount of water (OH-) impurities present in a fiber should be less than a few parts per billion. Fiber
attenuation caused by extrinsic absorption is affected by the level of impurities (OH-) present in the fiber.
If the amount of impurities in a fiber is reduced, then fiber attenuation is reduced.
Q.39 What are the main causes of absorption in optical fiber?
Q.40 Silica (pure glass) fibers are used because of their low intrinsic material absorption at the
wavelengths of operation. This wavelength of operation is between two intrinsic absorption regions.
What are these two regions called? What are the wavelengths of operation for these two regions?
Q.41 Extrinsic (OH-) absorption peaks define three regions or windows of preferred operation. List the
three windows of operation.
SCATTERING. - Basically, scattering losses are caused by the interaction of light with density
fluctuations within a fiber. Density changes are produced when optical fibers are manufactured.
During manufacturing, regions of higher and lower molecular density areas, relative to the average
density of the fiber, are created. Light traveling through the fiber interacts with the density areas as shown
in figure 2-22. Light is then partially scattered in all directions.
Figure 2-22. - Light scattering.
In commercial fibers operating between 700-nm and 1600-nm wavelength, the main source of loss is
called Rayleigh scattering. Rayleigh scattering is the main loss mechanism between the ultraviolet and
infrared regions as shown in figure 2-21. Rayleigh scattering occurs when the size of the density
fluctuation (fiber defect) is less than one-tenth of the operating wavelength of light. Loss caused by
Rayleigh scattering is proportional to the fourth power of the wavelength (1/&lambda;4). As the
wavelength increases, the loss caused by Rayleigh scattering decreases.

If the size of the defect is greater than one-tenth of the wavelength of light, the scattering mechanism is
called Mie scattering. Mie scattering, caused by these large defects in the fiber core, scatters light out of
the fiber core. However, in commercial fibers, the effects of Mie scattering are insignificant. Optical
fibers are manufactured with very few large defects.
Q.42 What is the main loss mechanism between the ultraviolet and infrared absorption regions?
Q.43 Scattering losses are caused by the interaction of light with density fluctuations within a fiber. What
are the two scattering mechanisms called when the size of the density fluctuations is (a) greater than and
(b) less than one-tenth of the operating wavelength?
BENDING LOSS. - Bending the fiber also causes attenuation. Bending loss is classified according to the
bend radius of curvature: microbend loss or macrobend loss.
Microbends are small microscopic bends of the fiber axis that occur mainly when a fiber is cabled.
Macrobends are bends having a large radius of curvature relative to the fiber diameter. Microbend and
macrobend losses are very important loss mechanisms. Fiber loss caused by microbending can still occur
even if the fiber is cabled correctly. During installation, if fibers are bent too sharply, macrobend losses
will occur.
Microbend losses are caused by small discontinuities or imperfections in the fiber. Uneven coating
applications and improper cabling procedures increase microbend loss. External forces are also a source
of microbends. An external force deforms the cabled jacket surrounding the fiber but causes only a small
bend in the fiber. Microbends change the path that propagating modes take, as shown in figure 2-23.
Microbend loss increases attenuation because low-order modes become coupled with high-order modes
that are naturally lossy.
Figure 2-23. - Microbend loss.
Macrobend losses are observed when a fiber bend's radius of curvature is large compared to the fiber
diameter.
These bends become a great source of loss when the radius of curvature is less than several centimeters.
Light propagating at the inner side of the bend travels a shorter distance than that on the outer side. To
maintain the phase of the light wave, the mode phase velocity must increase. When the fiber bend is less
than some critical radius, the mode phase velocity must increase to a speed greater than the speed of light.
However, it is impossible to exceed the speed of light. This condition causes some of the light within the
fiber to be converted to high-order modes. These high-order modes are then lost or radiated out of the
fiber.
Fiber sensitivity to bending losses can be reduced. If the refractive index of the core is increased, then
fiber sensitivity decreases. Sensitivity also decreases as the diameter of the overall fiber increases.
However, increases in the fiber core diameter increase fiber sensitivity. Fibers with larger core size
propagate more modes. These additional modes tend to be more lossy.
Q.44 Microbend loss is caused by microscopic bends of the fiber axis. List three sources of microbend
loss.
Q.45 How is fiber sensitivity to bending losses reduced?
[ Back ] [ Home ] [ Up ] [ Next ]
DISPERSION
There are two different types of dispersion in optical fibers.
The types are intramodal and intermodal dispersion. Intramodal, or chromatic, dispersion occurs in all
types of fibers. Intermodal, or modal, dispersion occurs only in multimode fibers. Each type of dispersion
mechanism leads to pulse spreading. As a pulse spreads, energy is overlapped. This condition is shown in
figure 2-24. The spreading of the optical pulse as it travels along the fiber limits the information capacity
of the fiber.
Figure 2-24. - Pulse overlap.
Intramodal Dispersion
Intramodal, or chromatic, dispersion depends primarily on fiber materials. There are two types of
intramodal dispersion. The first type is material dispersion. The second type is waveguide dispersion.
Intramodal dispersion occurs because different colors of light travel through different materials and
different waveguide structures at different speeds.

Material dispersion occurs because the spreading of a light pulse is dependent on the wavelengths'
interaction with the refractive index of the fiber core. Different wavelengths travel at different speeds in
the fiber material. Different wavelengths of a light pulse that enter a fiber at one time exit the fiber at
different times. Material dispersion is a function of the source spectral width. The spectral width specifies
the range of wavelengths that can propagate in the fiber. Material dispersion is less at longer wavelengths.
Waveguide dispersion occurs because the mode propagation constant (&beta;) is a function of the size
of the fiber's core relative to the wavelength of operation. Waveguide dispersion also occurs because light
propagates differently in the core than in the cladding.
In multimode fibers, waveguide dispersion and material dispersion are basically separate properties.
Multimode waveguide dispersion is generally small compared to material dispersion. Waveguide
dispersion is usually neglected.
However, in single mode fibers, material and waveguide dispersion are interrelated.
The total dispersion present in single mode fibers may be minimized by trading material and waveguide
properties depending on the wavelength of operation.
Q.46 Name the two types of intramodal, or chromatic, dispersion.
Q.47 Which dispersion mechanism (material or waveguide) is a function of the size of the fiber's core
relative to the wavelength of operation?
Intermodal Dispersion
Intermodal or modal dispersion causes the input light pulse to spread. The input light pulse is made up of
a group of modes. As the modes propagate along the fiber, light energy distributed among the modes is
delayed by different amounts. The pulse spreads because each mode propagates along the fiber at
different speeds. Since modes travel in different directions, some modes travel longer distances. Modal
dispersion occurs because each mode travels a different distance over the same time span, as shown in
figure 2-25. The modes of a light pulse that enter the fiber at one time exit the fiber a different times. This
condition causes the light pulse to spread. As the length of the fiber increases, modal dispersion increases.
Figure 2-25. - Distance traveled by each mode over the same time span.
Modal dispersion is the dominant source of dispersion in multimode fibers. Modal dispersion does not
exist in single mode fibers. Single mode fibers propagate only the fundamental mode. Therefore, single
mode fibers exhibit the lowest amount of total dispersion. Single mode fibers also exhibit the highest
possible bandwidth.
Q.48 Modes of a light pulse that enter the fiber at one time exit the fiber at different times. This condition
causes the light pulse to spread. What is this condition called?
Vibrational absorption
When a chemical bond is dipolar (one atom more electronegative than the other) its vibration is an
oscillating dipole
If signal at telecom wavelength is close enough in frequency to that of the vibration, the
oscillating electric field goes into resonance with the vibration and loses energy to it
Vibrational energies are typically measured in cm-1 (inverse of wavelength). 1550 nm = 6500 cm
Scattering loss: from index discontinuity
Scatterers are much smaller than the wavelength: Rayleigh and Raman scattering
Scatterers are much bigger than the wavelength: geometric ray optics
Scatterers are about the same size as the wavelength: Mie scattering
Scatterers are sound waves: Brillouin scattering

Interference and Diffraction

14.1 Superposition of Waves
Consider a region in space where two or more waves pass through at the same time. According to the
superposition principle, the net displacement is simply given by the vector or the algebraic sum of the
individual displacements. Interference is the combination of two or more waves to form a composite
wave, based on such principle. The idea of the superposition principle is illustrated in Figure 14.1.1. (a)
(b) (c) (d)
Figure 14.1.1 Superposition of waves. (b) Constructive
interference, and (c) destructive interference. Suppose we
are given two waves, (,) = sin( kx t+ ), (,) =
sin( kx t+ )
xt xt (14.1.1)
1 10 111 2 20 222
the resulting wave is simply (,) = sin(kx t+ ) + sin(
xt kx t+ ) (14.1.2)
10 11120 2 22
The interference is constructive if the amplitude of (,)
xt is greater than the individual ones (Figure
14.1.1b), and destructive if smaller (Figure 14.1.1c).
As an example, consider the superposition of the following two waves at t= 0:

1( ) = sin x, 2( ) = 2sin x
xx + (14.1.3) 4
The resultant wave is given by
14-2 () =()x +2 x =sin x +2sin x + 1
x () =+ 2 sin x +2 cos x (14.1.4) 4 where we have used sin( +)
=sin cos +cos sin 1 ( )
(14.1.5) and sin(/4) =cos( /4) =2 / 2 . Further use of the identity
a sin x +bcos x =a2 +b2 2 a 2 sin x + 2 b 2 cos x

a +ba +b =a2 +b2 [cossin x +sin cos x] (14.1.6) =a2 +b2 sin( x
+) with
then leads to () =5 +2 2 sin( x +) (14.1.8)
x
where =tan 1( 2/(1 +2)) =30.4=0.53 rad. The superposition of the waves is depicted in Figure 14.1.2.
Figure 14.1.2 Superposition of two sinusoidal waves.
We see that the wave has a maximum amplitude when sin(x +) =1, or x =/2 . The interference there
is constructive. On the other hand, destructive interference occurs at x == , wheresin( ).
2.61 rad =0
14-3 In order to form an interference pattern, the incident light must satisfy two conditions:
(i) The light sources must be coherent. This means that the plane waves from the sources must maintain a
constant phase relation. For example, if two waves are completely out of phase with = , this phase
difference must not change with time.
(ii) The light must be monochromatic. This means that the light consists of just one wavelength = 2/ k
.
Light emitted from an incandescent lightbulb is incoherent because the light consists o waves of different
wavelengths and they do not maintain a constant phase relationship. Thus, no interference pattern is
observed.
Figure 14.1.3 Incoherent light source
14.2 Youngs Double-Slit Experiment
In 1801 Thomas Young carried out an experiment in which the wave nature of light was demonstrated.
The schematic diagram of the double-slit experiment is shown in Figure
14.2.1.

Figure 14.2.1 Youngs double-slit experiment.

A monochromatic light source is incident on the first screen which contains a slit S. The
0emerging light then
arrives at the second screen which has two parallel slits S1 and S2. which serve as the sources of coherent
light. The light waves emerging from the two slits then interfere and form an interference pattern on the
viewing screen. The bright bands (fringes) correspond to interference maxima, and the dark band
interference minima.
14-4 Figure 14.2.2 shows the ways in which the waves could combine to interfere constructively or
destructively.
Figure 14.2.2 Constructive interference (a) at P, and (b) at P1. (c) Destructive interference at P2.
The geometry of the double-slit interference is shown in the Figure 14.2.3.
Figure 14.2.3 Double-slit experiment
Consider light that falls on the screen at a point P a distance y from the point O that lies on the screen a
perpendicular distance L from the double-slit system. The two slits are separated by a distance d. The
light from slit 2 will travel an extra distance rr1
=
2 to the point P than
the light from slit 1. This extra distance is called the path difference. From Figure 14.2.3, we have, using
the law of cosines, In the limit Ld, i.e., the distance to the screen is much greater than the distance
between the slits, the sum of r1 and r2 may be approximated by rr2+r , and the path
12
difference becomes
=rrd sin (14.2.4)
21
In this limit, the two rays r1and r2 are essentially treated as being parallel (see Figure 14.2.4).
Figure 14.2.4 Path difference between the two rays, assuming Ld.
Whether the two waves are in phase or out of phase is determined by the value of . Constructive
interference occurs when is zero or an integer multiple of the wavelength
:
=d sin=m , m 0, 1, 2, 3, ... (constructive interference)
=
(14.2.5)
where m is called the order number. The zeroth-order (m = 0) maximum corresponds to the central bright
fringe at =0 , and the first-order maxima ( m =1 ) are the bright fringes on either side of the central
fringe.
On the other hand, when is equal to an odd integer multiple of /2 , the waves will be 180out of phase
at P, resulting in destructive interference with a dark fringe on the screen. The condition for destructive
interference is given by
=d sin =m +1 , m 0, 1, 2, 3, ... (destructive interference)
=
(14.2.6) 2
In Figure 14.2.5, we show how a path difference of /2 ( m =0 ) results in a
= destructive interference and =( m =1)
leads to a constructive interference.
14-6
Figure 14.2.5 (a) Destructive interference. (b) Constructive interference.
To locate the positions of the fringes as measured vertically from the central point O, in addition to L d ,
we shall also assume that the distance between the slits is much greater than the wavelength of the
monochromatic light, d . The conditions imply that the angle is very small, so that
sin tan =y (14.2.7)
L

Substituting the above expression into the constructive and destructive interference conditions given in
Eqs. (14.2.5) and (14.2.6), the positions of the bright and dark fringes are, respectively,
L
yb =m (14.2.8)
d
and 1 L
yd = m + (14.2.9) 2 d
Example 14.1: Double-Slit Experiment
Suppose in the double-slit arrangement, d = 0.150mm, L = 120cm, =833nm, and y = 2.00cm .
(a) What is the path difference for the rays from the two slits arriving at point P?
(b) Express this path difference in terms of .
(c) Does point P correspond to a maximum, a minimum, or an intermediate condition?
Solutions: 14-7 (a) The path difference is given by =d sin. When L y , is small and we can make
the approximationsin tan =y / L . Thus,
y 4 2.00102 6
m
d =(1.5010 m)=2.50 10 m
L 1.20m

(b) From the answer in part (a), we have

2.50 10 6m
=7 3.00
8.33 10 m or =3.00.
(c) Since the path difference is an integer multiple of the wavelength, the intensity at point P is a
maximum.
14.3 Intensity Distribution
Consider the double-slit experiment shown in Figure 14.3.1.
Figure 14.3.1 Double-slit interference

The total instantaneous electric field E at the point P on the screen is equal to the vector

sum of the two sources: EE1 +E2 . On the other hand, the Poynting flux S is
=
proportional to the square of the total field:

2 222
S E =(E +E ) =E ++ 2 E (14.3.1)
EE
12 1212
Taking the time average of S, the intensity I of the light at P may be obtained as:

E12

represents the correlation between the two light waves. For incoherent light sources, since there is no
definite phase relation between E1 and E2 , the cross term vanishes, and the intensity due to the incoherent
source is simply the sum of the two individual intensities:
Iinc I1 I2 (14.3.3)
=+
For coherent sources, the cross term is non-zero. In fact, for constructive interference, and the resulting
intensity is I =4I1 (14.3.4)
which is four times greater than the intensity due to a single source. On the other hand,
I1, and the total when destructive interference takes place, E1 =E2 , and
12
intensity becomes
II12I1 I1 0 (14.3.5)
= +=
as expected.
Suppose that the waves emerged from the slits are coherent sinusoidal plane waves. Let the electric field
components of the wave from slits 1 and 2 at P be given by
E1 =E0 sint (14.3.6)
and E2 = 0
E sin( t+) (14.3.7)
respectively, where the waves from both slits are assumed have the same amplitude E0. For simplicity, we
have chosen the point P to be the origin, so that the kx dependence in the wave function is eliminated.
Since the wave from slit 2 has traveled an extra distance to P , E2 has an extra phase shift relative to E1
from slit 1.
For constructive interference, a path difference of =would correspond to a phase shift of =2. This
then implies

= (14.3.8)
2

or
14-9 2 2
= =dsin (14.3.9)

Assuming that both fields point in the same direction, the total electric field may be obtained by using the
superposition principle discussed in Section 13.4.1:
EEE E sin (14.3.10)
=+ =[t+sin( t+)]=2Ecos sin t
120 0 +
22 where we have used
the trigonometric identity
+

sin+sin =2sin cos (14.3.11)

2 2 The intensity Iis proportional to the time average of
the square of the total electric field:

22
22
2
IE

=4E0 cos
sin 2 t+
=2E0 cos (14.3.12)
or

II0 cos (14.3.13)

where I0 is the maximum intensity on the screen. Upon substituting Eq. (14.3.4), the above expression
becomes
II0 cos2 dsin (14.3.14)
=
Figure 14.3.2 Intensity as a function
of dsin/ For small angle , using Eq. (14.2.5) the
intensity can be rewritten as d
I=I0 cos 2 y (14.3.15)
L
14-10 Example 14.2: Intensity of Three-Slit Interference
Suppose a monochromatic coherent source of light passes through three parallel slits, each separated by a
distance d from its neighbor, as shown in Figure 14.3.3.
Figure 14.3.3 Three-slit interference.
The waves have the same amplitude E0and angular frequency , but a constant phase difference =2 d sin /
.
(a) Show that the intensity is
I 2d sin 2
I =0 12cos
9 + (14.3.16)
where I0 is the maximum intensity associated with the primary maxima.
(b) What is the ratio of the intensities of the primary and secondary maxima?
Solutions:
(a) Let the three waves emerging from the slits be E sin t, E =E sin (t +), E =E sin t +2 ) (14.3.17)
E = (
1020 30
Using the trigonometric identity
+
sin+sin =2cos sin (14.3.18)

2 2 the sum of E1 and E3 is
E1 +E3 =E0 sin t +sin (t +2)=2E0 cos sin( t +) (14.3.19)

The total electric field at the point P on the screen is
14-11 EEE E = 1 + 2 + 30 + + 0
=2Ecos sin( t ) Esin( t+)
(14.3.20)
=E0
(1+2cos )sin( t+) where =2 dsin /
. The intensity is proportional to
E2: E02 2
IE 02 (12cos )2
+
sin 2 (t+)

(12cos ) (14.3.21)
=+
2 where we have
used
sin2
=1/ 2 . The maximum intensity I0 is attained when
(t+)
cos=1. Thus, +I =(12cos )2 (14.3.22)
I 0
2
I 0
I09 which implies I= 9 (12cos ) = 9 1+2cos 2d sin 2 (14.3.23)

+
(b) The interference pattern is shown in Figure 14.3.4.
From the figure, we see that the minimum intensity is zero, and occurs when cos=1/2 . The condition
for primary maxima is cos =+1, which gives /0 =. In
II 1 addition, there are also
secondary maxima which are located at cos=1. The condition implies =(2m+, or dsin
=(m+1/2),m=0, 1,
1) / 2,... The intensity ratio is II/0 =1/9 . 14-12 14.4 Diffraction
In addition to interference, waves also exhibit another property diffraction, which is the bending of
waves as they pass by some objects or through an aperture. The phenomenon of diffraction can be
understood using Huygenss principle which states that
Every unobstructed point on a wavefront will act a source of secondary spherical waves. The new
wavefront is the surface tangent to all the secondary spherical waves.
Figure 14.4.1 illustrates the propagation of the wave based on Huygenss principle.
Figure 14.4.1 Propagation of wave based on Huygenss principle.
According to Huygenss principle, light waves incident on two slits will spread out and exhibit an
interference pattern in the region beyond (Figure 14.4.2a). The pattern is called a diffraction pattern. On
the other hand, if no bending occurs and the light wave continue to travel in straight lines, then no
diffraction pattern would be observed (Figure 14.4.2b).
Figure 14.4.2 (a) Spreading of light leading to a diffraction pattern. (b) Absence of diffraction pattern if
the paths of the light wave are straight lines.
We shall restrict ourselves to a special case of diffraction called the Fraunhofer diffraction. In this case,
all light rays that emerge from the slit are approximately parallel to each other. For a diffraction pattern to
appear on the screen, a convex lens is placed between the slit and screen to provide convergence of the
light rays.
14.5 Single-Slit Diffraction
In our consideration of the Youngs double-slit experiments, we have assumed the width of the slits to be
so small that each slit is a point source. In this section we shall take the width of slit to be finite and see
how Fraunhofer diffraction arises.
14-13 Let a source of monochromatic light be incident on a slit of finite width a, as shown in Figure
14.5.1.
Figure 14.5.1 Diffraction of light by a slit of width a.
In diffraction of Fraunhofer type, all rays passing through the slit are
approximately parallel. In addition, each portion of the slit will act as a source
of light waves according to Huygenss principle. For simplicity we divide the
slit into two halves. At the first minimum, each ray from the upper half will be
exactly 180out of phase with a corresponding ray form the lower half. For
example, suppose there are 100 point sources, with the first 50 in the lower
half, and 51 to 100 in the upper half. Source 1 and source 51 are separated by a
distance a /2 and are out of phase with a path difference =/2. Similar
observation applies to source 2 and source 52, as well as any pair that are a
distance
sin 2 a = 2

a
/2

apart. Thus, the

condition for
the first
minimum is

(14.5.1)

or
sin=

(14.5.2)

a
ATTENUATION : Attenuation in an optical fiber is caused by absorption, scattering, and bending
losses. Attenuation is the loss of optical power as light travels along the fiber. Signal attenuation is
defined as the ratio of optical input power (Pi) to the optical output power (Po). Optical input power is the

power injected into the fiber from an optical source. Optical output power is the power received at the
fiber end or optical detector. The following equation defines signal attenuation as a unit of length: Signal
attenuation is a log relationship. Length (L) is expressed in kilometers. Therefore, the unit of attenuation
is decibels/kilometer (dB/km). As previously stated, attenuation is caused by absorption, scattering, and
bending losses. Each mechanism of loss is influenced by fiber-material properties and fiber structure.
However, loss is also present at fiber connections. Fiber connector, splice, and coupler losses are
discussed in chapter 4. The present discussion remains relative to optical fiber attenuation properties.
2.2 ABSORPTION. - Absorption is a major cause of signal loss in an optical fiber. Absorption is
defined as the portion of attenuation resulting from the conversion of optical power into another energy
form, such as heat. Absorption in optical fibers is explained by three factors:
Imperfections in the atomic structure of the fiber material
The intrinsic or basic fiber-material properties
The extrinsic (presence of impurities) fiber-material properties
Imperfections in the atomic structure induce absorption by the presence of missing molecules or oxygen
defects. Absorption is also induced by the diffusion of hydrogen molecules into the glass fiber. Since
intrinsic and extrinsic material properties are the main cause of absorption, they are discussed further.
Intrinsic Absorption. - Intrinsic absorption is caused by basic fiber-material properties. If an optical
fiber were absolutely pure, with no imperfections or impurities, then all absorption would be intrinsic.
Intrinsic absorption sets the minimal level of absorption. In fiber optics, silica (pure glass) fibers are used
predominately. Silica fibers are used because of their low intrinsic material absorption at the wavelengths
of operation. Sub Code: EC1402 Optical Communication Page 41 of 98 Einstein College of
Engineering
In silica glass, the wavelengths of operation range from 700 nanometers (nm) to 1600 nm. Figure 2-21
shows the level of attenuation at the wavelengths of operation. This wavelength of operation is between
two intrinsic absorption regions. The first region is the ultraviolet region (below 400-nm wavelength).
The second region is the infrared region (above 2000-nm wavelength).
Intrinsic absorption in the ultraviolet region is caused by electronic absorption bands. Basically,
absorption occurs when a light particle (photon) interacts with an electron and excites it to a higher
energy level. The tail of the ultraviolet absorption band is shown in figure 2-21. The main cause of
intrinsic absorption in the infrared region is the characteristic vibration frequency of atomic bonds. In
silica glass, absorption is caused by the vibration of silicon-oxygen (Si-O) bonds. The interaction
between the vibrating bond and the electromagnetic field of the optical signal causes intrinsic absorption.
Light energy is transferred from the electromagnetic field to the bond. The tail of the infrared absorption
band is shown in figure 2-21.
Extrinsic Absorption. - Extrinsic absorption is caused by impurities introduced into the fiber material.
Trace metal impurities, such as iron, nickel, and chromium, are introduced Sub Code: EC1402 Optical
Communication Page 42 of 98 Einstein College of Engineering into the fiber during fabrication.
Extrinsic absorption is caused by the electronic transition of these metal ions from one energy level to
another. Extrinsic absorption also occurs when hydroxyl ions (OH-) are introduced into the fiber. Water in
silica glass forms a silicon-hydroxyl (Si-OH) bond. This bond has a fundamental absorption at 2700 nm.
However, the harmonics or overtones of the fundamental absorption occur in the region of operation.
These harmonics increase extrinsic absorption at 1383 nm, 1250 nm, and 950 nm. Figure 2-21 shows the
presence of the three OH- harmonics. The level of the OH- harmonic absorption is also indicated. These
absorption peaks define three regions or windows of preferred operation. The first window is centered at
850 nm. The second window is centered at 1300 nm. The third window is centered at 1550 nm. Fiber
optic systems operate at wavelengths defined by one of these windows. The amount of water (OH-)
impurities present in a fiber should be less than a few parts per billion. Fiber attenuation caused by
extrinsic absorption is affected by the level of impurities (OH-) present in the fiber. If the amount of
impurities in a fiber is reduced, then fiber attenuation is reduced.
2.3 SCATTERING. - Basically, scattering losses are caused by the interaction of light with density
fluctuations within a fiber. Density changes are produced when optical fibers are manufactured. During
manufacturing, regions of higher and lower molecular density areas, relative to the average density of the
fiber, are created. Light traveling through the fiber interacts with the density areas as shown in
In commercial fibers operating between 700-nm and 1600-nm wavelength, the main source of loss is
called Rayleigh scattering. Rayleigh scattering is the main loss mechanism between the ultraviolet and

infrared regions as shown in figure 2-21. Rayleigh scattering occurs when the size of the density
fluctuation (fiber defect) is less than one-tenth of the operating wavelength of light. Loss caused by
Rayleigh scattering is proportional to the fourth power of the wavelength (1/&lambda;4). As the
wavelength increases, the loss caused by Rayleigh scattering decreases. If the size of the defect is greater
than one-tenth of the wavelength of light, the scattering mechanism is called Mie scattering. Mie
scattering, caused by these large defects in the fiber core, scatters light out of the fiber core. However, in
commercial fibers, the effects of Mie scattering are insignificant. Optical fibers are manufactured with
very few large defects.
2.4 BENDING LOSS. - Bending the fiber also causes attenuation. Bending loss is classified according to
the bend radius of curvature: microbend loss or macrobend loss. Microbends are small microscopic
bends of the fiber axis that occur mainly when a fiber is cabled. Macrobends are bends having a large
radius of curvature relative to the fiber diameter. Microbend and macrobend losses are very important
loss mechanisms. Fiber loss caused by microbending can still occur even if the fiber is cabled correctly.
During installation, if fibers are bent too sharply, macrobend losses will occur. Microbend losses are
caused by small discontinuities or imperfections in the fiber. Uneven coating applications and improper
cabling procedures increase microbend loss. External forces are also a source of microbends. An external
force deforms the cabled jacket surrounding the fiber but causes only a small bend in the fiber.
Microbends change the path that propagating modes take, as shown in figure 2-23. Microbend loss
increases attenuation because low-order modes become coupled with high-order modes that are naturally
lossy. Macrobend losses are observed when a fiber bend's radius of curvature is large compared to the
fiber diameter. These bends become a great source of loss when the radius of curvature is less than
several centimeters. Light propagating at the inner side of the bend travels a shorter distance than that on
the outer side. To maintain the phase of the light wave, the mode phase velocity must increase. When the
fiber bend is less than some critical radius, the mode phase velocity must increase to a speed greater than
the speed of light. However, it is impossible to exceed the speed of light. This condition causes some of
the light within the fiber to be converted to high-order modes. These high-order modes are then lost or
radiated out of the fiber. Sub Code: EC1402 Optical Communication Page 44 of 98 Einstein College of
Engineering
Fiber sensitivity to bending losses can be reduced. If the refractive index of the core is increased, then
fiber sensitivity decreases. Sensitivity also decreases as the diameter of the overall fiber increases.
However, increases in the fiber core diameter increase fiber sensitivity. Fibers with larger core size
propagate more modes. These additional modes tend to be more lossy.
2.5 SIGNAL DISTORTION IN OPTICAL WAVE GUIDES:
2.5.1 DISPERSION There are two different types of dispersion in optical fibers. The types are
intramodal and intermodal dispersion. Intramodal, or chromatic, dispersion occurs in all types of fibers.
Intermodal, or modal, dispersion occurs only in multimode fibers. Each type of dispersion mechanism
leads to pulse spreading. As a pulse spreads, energy is overlapped. This condition is shown in figure 2-24.
The spreading of the optical pulse as it travels along the fiber limits the information capacity of the fiber.
Intramodal Dispersion Intramodal, or chromatic, dispersion depends primarily on fiber materials. There
are two types of intramodal dispersion. The first type is material dispersion. The second type is
waveguide dispersion. Sub Code: EC1402 Optical Communication Page 45 of 98 Einstein College of
Engineering
Intramodal dispersion occurs because different colors of light travel through different materials and
different waveguide structures at different speeds.
Material dispersion occurs because the spreading of a light pulse is dependent on the wavelengths'
interaction with the refractive index of the fiber core. Different wavelengths travel at different speeds in
the fiber material. Different wavelengths of a light pulse that enter a fiber at one time exit the fiber at
different times. Material dispersion is a function of the source spectral width. The spectral width specifies
the range of wavelengths that can propagate in the fiber. Material dispersion is less at longer wavelengths.
Waveguide dispersion occurs because the mode propagation constant (&beta;) is a function of the size
of the fiber's core relative to the wavelength of operation. Waveguide dispersion also occurs because light
propagates differently in the core than in the cladding. In multimode fibers, waveguide dispersion and
material dispersion are basically separate properties. Multimode waveguide dispersion is generally small
compared to material dispersion. Waveguide dispersion is usually neglected. However, in single mode
fibers, material and waveguide dispersion are interrelated. The total dispersion present in single mode

fibers may be minimized by trading material and waveguide properties depending on the wavelength of
operation. Sub Code: EC1402 Optical Communication Page 46 of 98 Einstein College of Engineering
Intermodal Dispersion Intermodal or modal dispersion causes the input light pulse to spread. The input
light pulse is made up of a group of modes. As the modes propagate along the fiber, light energy
distributed among the modes is delayed by different amounts. The pulse spreads because each mode
propagates along the fiber at different speeds. Since modes travel in different directions, some modes
travel longer distances. Modal dispersion occurs because each mode travels a different distance over the
same time span, as shown in figure 2-25. The modes of a light pulse that enter the fiber at one time exit
the fiber a different times. This condition causes the light pulse to spread. As the length of the fiber
increases, modal dispersion increases. Modal dispersion is the dominant source of dispersion in
multimode fibers. Modal dispersion does not exist in single mode fibers. Single mode fibers propagate
only the fundamental mode. Therefore, single mode fibers exhibit the lowest amount of total dispersion.
Single mode fibers also exhibit the highest possible bandwidth. Figure 2-25. - Distance traveled by each
mode over the same time span.
2.5.2 Information Capacity Determination: A light pulse will broaden as it travels along the fiber. The
pulse broadening will eventually cause neighboring pulses to overlap. After a certain amount of overlap,
neighboring pulses will not be distinguishable. Thus, this dispersive mechanism limits the information
capacity of a fiber. One measure of the information capacity of an optical waveguide is called the
bandwidth-distance product. This value is usually given as x Hz km. For example, you are given a fiber
with an bandwidth-distance product of 250 MHz km and you need to use a five kilometer length. The
useable bandwidth over this link is then (ignoring other issues) Sub
2.5.3 Group Delay Let us presume a linear system where an optical source launches light power into a
fiber where all the modes carry equal power. The spectral components will be assumed to travel
independently. Each component will undergo a time delay called the group delay per unit length in the
direction of travel.
There are two main contributions to chromatic dispersion:
Material dispersion. The refractive index of any medium is a function of wavelength, and hence
different wavelengths that see different refractive indices will propagate with different velocities,
resulting in intramodal dispersion.
Waveguide dispersion. Even if the refractive index is constant, and material dispersion eliminated, the
propagation constant would vary with wavelength for any waveguide structure, resulting in intramodal
dispersion.
Total chromatic dispersion D is the sum of the material dispersion DM and the waveguide dispersion Dw.
There is zero dispersion close to 1.31 m (ZD) in glass optical fiber.
Figure 3.11 The variation in chromatic dispersion with wavelength.
Intramodal (Chromatic) Dispersion
in Dispersion-modified Single-mode Optical Fiber
The contribution of waveguide dispersion is dependent on fiber parameters such as refractive indices and
core diameter.
Dispersion shifted fibers: shift ZD typically to 1.55 m, where the optical loss of optical fiber is a
minimum.
Dispersion flattened fibers: the total chromatic dispersion is relatively small over the wavelength range
1.3-1.6 m.
Figure 3.12 Dispersion-shifted and dispersion-flattened fibers.
For dispersion flattened single-mode optical fiber,
operated near ZD , e.g. D = 1 ps/(nmkm)
500 Gbit/skm
Comparison
SI MM fiber BT 10 Mbit/skm
GI MM fiber BT 8 Gbit/skm
SI SM fiber BT 500 Gbit/skm