Chapter 1

Optical Fiber Transmission

1.1

Introduction

Until the mid-1970s, communication systems transmitted information over copper cables
or free space. In 1966, Charles Kao and George Hockham working at Standard Telecommunications in England proposed that an optical ber might be used as a means of
communication provided the signal loss could be much less than 20 dB/km [1]. They also
illustrated that the attenuation in bers available at that time was caused by impurities
which could be removed. At Corning Glass Works, Robert Maurer, Douald Keck and
Peter Schultz worked with fused silica, a material that can be made extremely pure. In
1970, they developed a single mode ber with attenuation below 20 dB/km [2]. In 1977,
the rst optical telecommunication system was installed about 1.5 miles under downtown
Chicago and each optical ber carried the equivalent of 672 voice channels. In 1979,
single-mode bers with a loss of only 0.2 dB/km at 1550 nm were fabricated [3]. The
availability of low loss bers combined with the advent of semiconductor lasers led to
a new era of optical ber communication. Today, more than 80% of the worlds long
distance trac is carried over optical ber cable and about 25 million kilometers of the
optical ber has been installed worldwide.
This chapter deals with the light propagation in optical bers. Multi-mode and single1

mode bers are discussed using ray optics description in section 2.3. A rigorous solution
of the wave equation is derived in section 2.4 and the wave-optics description of the singlemode and multi-mode bers are presented. The pulse propagation in single-mode bers
is discussed in section 2.5. Section 2.7 focuses on the design of single-mode bers.

1.2

Fiber Structure

An optical ber consists of a central core clad by a material of slightly lower refractive
index as shown in Fig. 1.1. If the refractive index of the core is constant, such a ber is

Figure 1.1. Refractive index prole and cross section of a step-index ber.
called step-index ber. Most of the bers are made from glass, i.e., silica. The refractive
index of the core is increased by doping the silica with GeO2 . The cladding is pure silica.
Polymer jacket is used to protect the ber from moisture and abrasion. For short distance
(< 1 Km) and low bit rate ( Mb/s) transmission systems, plastic bers can be used.
2

They are: (i) inexpensive, (ii) exible, and (iii) easy to install and connect. However,
they do not transmit light eciently because of high absorption. For long distance and
high bit rate systems, glass bers are typically used. Optical bers have the following
advantages over copper cable:
(i) Bandwidth: To transmit more number of bits of information in a given time period,
the transmission medium should have a high bandwidth. Typically, the bandwidth is of
the order of the carrier frequency. In the case of optical signals, the carrier frequency is
200 THz and the bandwidth of the ber is several THz whereas the bandwidth of the
copper cable is typically several GHz or MHz.
(ii) Attenuation: The loss of a silica optical ber is around 0.2 dB/km which is much
lower than that of copper cable. Because of the lower loss, optical signals can propagate
longer distance without requiring repeaters.
(iii) Electromagnetic interference (EMI): Optical bers are not aected by electromagnetic
interference. This is because optical bers are purely dielectric waveguides with no metal
parts. In the case of copper cables, electromagnetic noise elds set up conduction currents
which interfere with the signal transmission.

1.3

Ray Propagation in Fibers

Figure 1.2. Signal propagation in ber by frustrated total internal reection.

Consider a step-index ber with core index n1 greater than the cladding index n2 . Let
3

c be the critical angle. Consider a ray with an angle > c as shown in Fig. 1.2. This
ray undergoes total internal reection at B. The reected ray BC undergoes total internal
reection again at C. This process continues till the output end and is called frustrated
total internal reection. With this mechanism, light is successfully transmitted from the
input end to the output end of the ber.
The power reection coecient may be dened as
Rp =

Reected power
Incident power

(1.1)

In the case of total internal reection, Rp = 1 which means all the power of the incident
ray appears on the reected ray. That is why it is called total internal reection(TIR). On
the other hand, the normal reection is always accompanied with refraction and Rp < 1.
Suppose the core of a silica ber is surrounded by a reective coating instead of a dielectric
cladding as shown in Fig.1.3. In this case, the light is guided by the normal reections

Figure 1.3. Signal propagation in ber by repeated normal reections.

at the interface. By choosing a coating with high reectivity, power loss during each
reection can be minimized. However, signicant power is lost after multiple reections.
In contrast, TIR is a more ecient way to transmit an optical signal over a long distance.

1.3.1

Numerical Aperture

Consider a ray which is incident on the ber input making an angle i as shown in Fig.
1.4. Using the Snells law, we have
sin i = n1 sin = n1 cos ,
4

(1.2)

where we have assumed the refractive index of air to be unity. If this ray has to suer
total internal reection at the core-cladding interface, the angle should be larger than
critical angle c ,
> c ,
sin > sin c .

(1.3)

Using Eq.(??) , we obtain

n21 sin2 > n22 ,
or
n21 cos2 < n21 n22 .

(1.4)

Using Eqs. (1.2) and (1.4), it follows that, to have a total internal reection, we should
have the following condition,
sin i < (n21 n22 )1/2 .

(1.5)

If (n21 n22 )1/2 > 1, total internal reection occurs for any incidence angle i. But for most

Figure 1.4. Numerical aperture of the ber. If the incidence angle i is less than the acceptance
angle, it undergoes total internal reection.

of the practical ber designs, (n21 n22 )1/2 1. In this case, as the angle of incidence i
increases, decreases and the light ray could escape the core-cladding interface without
total internal reection. From Eq. (1.5), the maximum value of sin i for a ray to be guided
5

is given by
sin imax = (n21 n22 )1/2 .
Therefore, the numerical aperture (NA) of the ber is dened as
(
)1/2
NA = sin imax = n21 n22
,

(1.6)

(1.7)

and imax is called acceptance angle. Let us dene the relative index dierence as
=

n1 n2
n1

(1.8)

If the dierence between n1 and n2 is small, n1 + n2 2n1 and Eq. (1.7) can be approximated as
NA n1 (2)1/2

(1.9)

Let us construct a cone with the semi-angle being equal to imax , as shown in Fig. 1.5(a).
If the incident ray is within the cone (i < imax ), it will be guided through the ber.
Otherwise, it will escape to the cladding and then to the jacket as shown in Fig. 1.5(b).
From a practical standpoint, it is desirable to have most of the source power to be launched
to the ber which requires large NA.

Figure 1.5. (a) if i imax , light is guided (b) if i > imax , light escapes from the core.

Problem 2.1
Core and cladding refractive indices of a multi-mode ber are 1.47 and 1.45, respectively.
Find the (a) Numerical aperture (b) acceptance angle and (c) the relative index dierence
6

Solution:
(a) n1 = 1.47, n2 = 1.45. From Eq. (1.7), we nd
NA = (n21 n22 )1/2 = 0.2417.

(1.10)

(b) From Eq. (1.6), the acceptance angle is

imax = sin1 (NA) = 0.2441 rad.

(1.11)

(c) From Eq. (1.8), the refractive index dierence is

=

1.3.2

n1 n2
= 0.0136.
n1

(1.12)

Multi-mode and Single-mode Fibers

If the index dierence (n1 n2 ) is large or the core radius a is much larger than the
wavelength of light, an optical ber supports multiple guided modes. A guided mode can
be imagined as a ray that undergoes total internal reection. Mathematical description
of guided modes is provided in Section 1.4. From ray optics theory, it follows that total
internal reection occurs for any angle in the interval [c /2]. This implies innite number
of guided modes. However, from the wave optics theory, it follows that all the angles in
the interval [c /2] are not permitted. Light guidance occurs only at discrete angles
{1 , 2 , } in the interval [c /2] as shown in Fig.1.6. Each discrete angle in Fig.1.6
corresponds to a guided mode. Typically a multi-mode ber can support thousands of
guided modes. As the index dierence (n1 n2 ) becomes very large and/or the core
diameter becomes much larger than the wavelength of light, the ber supports a very
large number of modes N which approaches innity and total internal reection occurs
7

Figure 1.6. When the angle of incidence exceeds c , total internal reection occurs only for
certain discrete angles.

for nearly any angle in the interval [c /2]. In this case, the ray optics theory is valid.
As the index dierence (n1 n2 ) becomes smaller and/or the core diameter becomes
comparable to the wavelength of light, the number of guided modes decreases. In fact, by
the proper design, a ber could support only one guided mode (in ray optics language, one
ray with a specic angle). Such a ber is called single mode ber which is of signicant
importance for high speed optical communication.

1.3.3

Dispersion in Multi-mode Fibers

A light pulse launched into a ber broadens as it propagates down the ber because of the
dierent times taken by dierent rays or modes to propagate through the ber. This is
known as intermodal dispersion. In Fig. 1.7, the path length of ray 1 is longer than that
of ray 3 and therefore, the fraction of the incident pulse carried by ray 3 arrives earlier
than that by ray 1 leading to pulse broadening.

Figure 1.7. Pulse broadens because of the dierent times taken by dierent rays to pass
through the ber.

Suppose an impulse is launched to the ber. Let us now estimate the pulse width at
the output end. For a ray making an angle with the axis (see Fig. 1.8), the distance

Figure 1.8. A ray undergoing multiple total internal reections in a multi-mode ber.
AB is traversed in time
tAB =

AB
AC + CB
=
,
v1
v1 sin

(1.13)

where v1 = c/n1 is the speed of light in the core. ACB can be imagined as one unit cell.
Let the ber length L is composed of N such unit cells. The time taken by the ray to
traverse a ber length L is,
tL =

N (AB)n1
n1 L
=
.
c sin
c sin

(1.14)

For multi-mode bers, we assume that all the rays making angles in the interval [c /2]
are present. This is a good approximation if the multi-mode ber supports many modes.
The ray which makes an angle = /2 propagates almost along the axis and takes the
shortest time. From Eq. (1.14), the time taken by this ray is
tmin =

n1 L
n1 L
=
.
c sin /2
c

(1.15)

The ray which makes an angle = c takes the longest time. The time taken by this ray
is
tmax =

n1 L
n2 L
= 1 .
c sin c
cn2

(1.16)

The time taken by a ray with angle in the interval [c /2] is somewhere in between
tmin and tmax . If an impulse is incident at the input end, it would excite all the rays in
9

the interval [c /2] and the rays occupy a time interval at the output end of duration
T given by
T = tmax tmin

n21 L n1 L
n1 L
=

=
cn2
c
c

)
n1
n21 L
1 =
,
n2
cn2

(1.17)

where is dened in Eq. (1.8). Fig. 1.9(b) shows the power proles of individual pulses
in the absence of other pulses. The pulse width at the output end is T , as shown Fig.
1.9(b). If the bit rate is B, the interval between bits is given by

Figure 1.9. The pulse train at (a) Fiber input (b) Fiber output. The individual pulses shown
here are outputs in the absence of input pulses at the other bit slots.

TB =

1
B

(1.18)

To avoid inter-symbol interference, the pulse width T TB . Using Eqs.(1.17) and

(1.18), we have
BL

10

cn2
n21

(1.19)

Eq. (1.19) provides the maximum bit rate-distance product possible for multi-moded
bers. From Eq.(1.19), we see that the product BL can be maximized by decreasing ,
but from Eq.(1.9), we see that it leads to reduction in NA, which is undesirable since it
lowers the power launched to the ber. So, there is a trade-o between power coupling
eciency and the maximum achievable bit rate-distance product.
From a practical standpoint, it is desirable to reduce the delay T . From Eq. (1.17),
we see that the delay at T increases linearly with ber length L. The quantity T /L
is a measure of intermodal dispersion.

Example 2.1
Consider a multi-mode ber with n1 = 1.46, = 0.01 and ber length L = 1 km. From
Eq. (1.8),
n2 = n1 (1 ) = 1.4454,

(1.20)

and
T =

n21 L
50 ns.
cn2

(1.21)

This implies an impulse traversing through the ber becomes a pulse of duration of about
50 ns. If B = 2 Mb/s, TB = 500 ns and the pulses at the output end are quite resolvable
as shown in Fig. 1.10(a). However, if the bit interval is 10 ns (B = 100 Mb/s), the
pulses would be absolutely unresolvable at the output end, as shown in Fig.1.10(b). From
Eq.(1.19), the maximum bit rate distance product is
(BL)max =

cn2
= 20.3 Mb/s km
n21

(1.22)

This implies that the maximum achievable bit rate is 20.3 Mb/s for a system consisting
of a 1 km ber. Note that the power proles in Fig. 1.10 are that of individual pulses
in the absence of other pulses. To nd the actual power proles, the elds of individual
pulses should be added and then, the power of the combined eld should be calculated.

11

Figure 1.10. A pulse train at the ber output (a) bit rate B = 2 Mb/s (b) B = 100 Mb/s.
The individual pulses shown here are the outputs in the absence of input pulses at the other bit
slots.

Figure 1.11. (a) Parabolic index prole (b) Triangular index prole (c) Step-index ber.

12

1.3.4

Graded-index Multi-mode Fibers

In a step-index multi-mode ber, the pulse width at the output is given by

T =

n21 L
.
cn2

(1.23)

To minimize this delay, graded-index multi-mode ber are used. The refractive index
prole, in this case, is given by

n [1 (r/a) ] if r < a
1
n(r) =
n = n (1 ), otherwise
2
1

where a is the core radius, n2 is the cladding index and n1 is the core index at r = 0,
= (n1 n2 ) /n1 and determines the index prole. When = 2, such a prole is
called parabolic index prole as shown in Fig. 1.11(a). When = , we get a step index
prole. A rigorous mathematical calculation shows that, if = 2(1 )( 2), the pulse
width, T is minimum and it is given by [4], [5]
T =

n1 2 L
.
8c

(1.24)

From Eq.(1.23) and (1.24), we see that the pulse broadening is proportional to in step
index bers whereas it is proportional to 2 in graded index bers (with = 2(1 )).
Since 1, pulse broadening can be signicantly reduced using graded index bers.
The reduction in pulse delay can be understood from the following analogy: suppose
A takes a highway to go to the oce from his home which is faster, but longer distance,
as shown in Fig. 1.12. B takes a shorter route but there are many trac lights leading to
delays. It is possible that A and B arrive at the same time to the oce, which implies that
there is no delay between the arrival times of A and B. In the case of graded-index bers,
the axial ray is conned mostly to the core center (because it undergoes total internal
reection closer to the center of the core due to the higher refractive index) and travels
slowly because of higher refractive index. The o-axis ray travels faster because it passes
through the region of lower refractive index. But it has to travel a longer zig-zag path.
So, the arrival time dierence between these rays could be smaller. This explains why the
delay given by Eq.(1.24) is smaller than that given by Eq.(1.23).
13

Figure 1.12. Two paths to connect home and oce which could take roughly same time.

Problem 2.2
Compare T for a step-index ber with that for a parabolic-index ber. Length= 1 km,
n1 = 1.47 and n2 = 1.45.

Solution:
n1 n2
= 0.0136
n1

(1.25)

For a step-index ber, from Eq. (1.23), we nd

T =

n21 L
= 67.58 ns
cn2

(1.26)

For a parabolic-index ber, from Eq. (1.24), we nd

T =

n21 2 L
= 0.1133 ns
8c

(1.27)

Thus, we see that intermodal dispersion can be signicantly reduced using parabolic-index
ber.

14

1.4

Modes of a Step-Index Optical Fiber

To understand the electromagnetic eld propagation in optical bers, one should solve
Maxwells equations with the condition that tangential components of electric and magnetic elds should be continuous at the interface between core and cladding [6]- [7]. When
the refraction index dierence between core and cladding is small, weakly guiding or scalar
wave approximation can be made [8]- [11] and in this approximation, electromagnetic eld
is assumed to be nearly transverse as in the case of free space propagation. Under this
approximation, one set of modes consists of Ex and Hy components (x-polarized) and
the other set of modes consists of Ey and Hx components (y-polarized). These two sets
of modes are independent and known as as linearly polarized (LP) modes. The x - or ycomponent of the electric eld intensity satises the scalar wave equation Eq. (??),
2

1 2
=0
2 (r) t2

(1.28)

c
n(r)

(1.29)

where (r) is the speed of light given by

(r) =
with

n for r < a
1
n(r) =
n for r a
2

(1.30)

where a = core radius. We assume that n1 > n2 as shown in Fig. 1.13. In cylindrical
coordinates, the Laplacian operator 2 can be written as
2 =

2 1
1 2 2
+
+
+ 2
r2
r r
r2 2
z

(1.31)

Suppose this ber is excited with a laser oscillating at angular frequency . In a linear
dielectric medium, the frequency of the electromagnetic eld should be same as that of
the source. Therefore, we look for the solution of Eq. (1.28) in the form
(r, , z, t) = f (r, , z)eit

15

(1.32)

Figure 1.13. Refractive index prole of a step-index ber

Substituting Eq. (1.32) in Eq. (1.28) and using Eqs. (1.29) and (1.31), we obtain
1 f
1 2f
2f
2f
+
+
+
+ k02 n2 (r)f = 0,
r2
r r r2 2 z 2

(1.33)

where k0 = /c = 2/0 is the free space wave number. The above equation is known as
the Helmholtz equation. We solve Eq. (1.33) by separation of variables:
f (r, , z) = R(r)()Z(z)

(1.34)

This technique may not work for all types of partial dierential equations. Especially, if
the partial dierential equation is nonlinear, the method of separation of variables fails.
Substituting Eq. (1.34) in Eq. (1.33), we obtain
( 2
)
d R 1 dR
1 d2
d2 Z
+
Z
+
RZ
+
R + k02 n2 (r)RZ = 0
dr2
r dr
r2 d2
dz 2
Dividing Eq. (1.35) by RZ, we obtain
(
)
1 d2 R 1 dR
d2 Z 1
1 d2
2 2
+
=

+
k
n
(r)
+
0
R dr2
r dr
r2 d2
dz 2 Z

(1.35)

(1.36)

In Eq. (1.34), we assumed that f can be decomposed into three parts R, and Z which
are functions of r, and z, respectively. Since the right hand side of Eq. (1.36) depends
only on z while the left hand side of Eq. (1.36) depends only on R and , they can be

16

equated only if each of them is a constant independent of r, and z. Let this constant
be 2
1 d2 Z

= 2
2
Z dz

(1.37)

Z(z) = A1 eiz + A2 eiz

(1.38)

Using Eq. (1.34) and substituting Eq. (1.38) in Eq. (1.32), we obtain
[
]
(r, , z, t) = R(r)() A1 ei(tz) + A2 ei(t+z)

(1.39)

The rst and second terms represent a forward and a backward propagating waves, respectively. In this section, let us consider only the forward propagating modes by setting
A2 = 0. For example, the laser output is launched to the ber from the left so that only
forward propagating modes are excited. If the ber medium has no defect, there would
be no reection occurring within the ber and assumption of forward propagating mode
is valid. From the left hand side of Eq. (1.36), we obtain
(
)
[ 2 2
]
1 d2
r2 d2 R 1 dR
2
2
+
+
r
k
n
(r)

0
R dr2
r dr
d2

(1.40)

The left hand side of Eq. (1.40) is a function of r only and the right hand side is a function
of only. As before, each of these terms should be a constant. Let this constant be m2

1 d2
= m2
d2

() = B1 eim + B2 eim

(1.41)
(1.42)

The rst and second terms represent the modes propagating in counter clockwise and
clockwise directions, respectively when m is positive. Let us consider only one set of
modes, say counter clockwise modes and set B2 = 0. If the initial conditions at the input
end of the ber is such that both types of modes are excited, we can not ignore the second
term in Eq. (1.42). In section 1.4.6, we will study how to combine various modes to satisfy
the given initial conditions. Using Eq. (1.41) in Eq. (1.40), we obtain
[
]
d2 R 1 dR
m2
2
2 2
+
+ k0 n (r) 2 R = 0
dr2
r dr
r
17

(1.43)

(b)

(a)
10

J0( x)

0.8

J1( x)
J2( x)

0.6
0.4

J3( x)

Y0( x)

10

0.2

Jm

Y1( x)

Ym 20

Y2( x)

30

Y ( x)

0.2
0.4

0.6
40

0.8
1

10

15

50

0.5

1.5

2.5

3.5

4.5

Figure 1.14. Bessel functions of the (a)rst kind (b)second kind.

Using Eq. (1.30) for n2 (r), we obtain
( 2 2
)
d2 R 1 dR m2
2
+

R
+
k
n

R = 0 r<a
0 1
dr2
r dr
r2
( 2 2
)
d2 R 1 dR m2
2
+

R
+
k
n

R = 0 ra
0
2
dr2
r dr
r2

(1.44)
(1.45)

Fiber modes can be classied into two types. (i) k02 n22 < 2 < k02 n21 . These modes are
called guided modes and (ii) 2 < k02 n22 . These modes are called radiation modes. It can
be shown that there exists no mode when 2 > k02 n21 .

1.4.1

Guided Modes

Since 2 < k02 n21 , the last term in Eq. (1.44) is positive and the solution of Eq. (1.44) for
this case is given by Bessel functions
R(r) = C1 Jm (1 r) + C2 Ym (1 r) ,
where 1 =

ra

(1.46)

k02 n21 2 , Jm (1 r) and Ym (1 r) are the Bessel functions of the rst kind

and second kind, respectively and are plotted in Fig. 1.14. The solution Ym (1 r) has to
18

(a)

(b)

4.5

4.5

K3( x)

3.5

3.5

K2( x)

Km 2.5

Im 2.5

K1( x)

K ( x)

1.5
1

0.5

0.5

0
0

I ( x)
0
I ( x)

1.5

I2( x)

I3( x)

Figure 1.15. Modied Bessel function of the (a)rst kind (b)second kind.
be rejected since it becomes as r 0. Therefore,
R(r) = C1 Jm (1 r) ,

ra

(1.47)

Since 2 > k02 n22 , the last term in Eq. (1.45) is negative. The solution of Eq. (1.45) is
given by modied Bessel function
R(r) = D1 Km (2 r) + D2 Im (2 r) ,
where 2 =

ra

(1.48)

2 k02 n22 . Im (2 r) and Km (2 r) are modied Bessel functions of the rst

kind and second kind, respectively and are plotted in Fig. 1.15. The solution Im (2 r)
has to be rejected since it becomes innite as r . Therefore,
R(r) = D1 Km (2 r) ,

ra

(1.49)

Now we make use of the fact that R and dR/dr should be continuous at the core-cladding
interface. If dR/dr is not continuous, d2 R/dr2 will be a Dirac delta function centered at
the interface and from Eq. (1.43), we nd that d2 R/dr2 could be a Dirac delta function
only if the refraction index at the interface is innity. Otherwise, Eq. (1.43) will not
19

be satised at the interface. Since the refractive index is nite, it follows that dR/dr
is continuous at the interface. Similarly, it can be proved that R(r) is continuous at
the interface. Continuity of R(r) and dR/dr at the core-cladding interface leads to the
following equations
C1 Jm (1 a) = D1 Km (2 a)

C1 1 Jm (1 a) = D1 2 Km (2 a)

(1.50)
(1.51)

where denotes dierentiation with respect to the argument. Dividing Eq. (1.51) by Eq.
(1.50), we obtain the following eigenvalue equation

Jm (1 a)
2 Km (2 a)
=
Jm (1 a)
1 Km (2 a)
where

1 = k02 n21 2

and

2 =

2 k02 n22 .

(1.52)

(1.53)

(1.54)

Note that in Eq. (1.52), the only unknown parameter is the propagation constant . It
is not possible to solve Eq. (1.52) analytically. Eq. (1.52) may be solved numerically to
obtain the possible values of . It would be easier to solve Eq. (1.52) numerically if we
avoid dierentiations in Eq. (1.52). This can be done using the following identities

1 aJm (1 a) = mJm (1 a) + 1 aJm1 (1 a)

2 aKm (2 a) = mKm (2 a) 2 aKm1 (2 a)

(1.55)
(1.56)

Using (1.55) and (1.56) in Eq. (1.52), we obtain

2 Km1 (2 a)
Jm1 (1 a)
=
Jm (1 a)
1 Km (2 a)

(1.57)

The propagation constants obtained after solving Eq. (1.57) lie in the interval [k0 n2 k0 n1 ].
It is convenient to dene the normalized propagation constant
b=

2 /k02 n22
n21 n22
20

(1.58)

so that when = k0 n2 , b = 0 and when = k0 n1 , b = 1. For any guided mode of a stepindex ber, we have 0 < b < 1. Eq. (1.57) can be solved for various design parameters such
as wavelength and core radius a and the numerically calculated propagation constant
can be plotted as a function of a specic design parameter. Instead, it is more convenient
to dene the normalized frequency
V

2
2
= a 1 + 2 = k0 a n21 n22
=

)1/2
2f a ( 2
n1 n22
,
c

(1.59)

where f is the mean frequency of the lightwave. Using Eq. (1.7), Eq. (1.59) may be
rewritten as
2a
NA.
(1.60)

The solutions of Eq. (1.57) for normalized propagation constant b as a function of norV =

malized frequency V give us universal curves shown in Fig. 1.16.

Figure 1.16. Plot of normalized propagation constant, b versus normalized frequency V .

To solve Eq. (1.57), we rst calculate V using Eq. (1.59). Let us rst set m = 0.
Since Eq. (1.57) is an implicit function of b, left hand side and right hand side of Eq.
21

(1.57) are plotted for various values of b in the interval [0, 1]. The point of intersection of
the curves corresponding to the left hand side and right hand side of Eq. (1.57) gives the
normalized propagation constant b of the guided mode supported by the ber. As can
be seen from Fig. 1.17, there can be many intersections which means there are several

Figure 1.17. Left hand side (LHS) and right hand side (RHS) of Eq. (1.57) as a function of
b. V = 8.

guided mode solutions. This procedure is repeated for m = 1, 2, . . . , M . For m = M + 1,

we nd that Eq. (1.57) admits no solution. When there is only one intersection (M = 1),
such a ber is called single-mode ber. The values of b corresponding to the intersections
for a particular value of V are shown in Fig. 1.16 by the broken lines. This process is
repeated for dierent value of V .
The advantage of the universal curve shown in Fig. 1.16 is that they can be used for a
step-index ber with arbitrary refractive indices and core radius. The dependence of the
propagation constant on the specic design parameter can be extracted from Fig. 1.16
using Eqs. (1.58) and (1.59).
For many applications, it is required to obtain the frequency dependence of propaga22

tion constant in a single-mode ber. This information can be obtained from Fig. 1.16.
For the given ber parameters and for the desired range of frequencies, V -parameters are
calculated using Eq. (1.59). Using Fig. 1.16, the corresponding normalized propagation
constants, b are calculated (corresponding to LP01 ) and with the help of Eq. (1.58), propagation constants for this range of frequencies can be calculated. For a specic value
of m, Eq. (1.57) has a nite number of solutions and n-th solution is known as LPmn
mode. LP stands for linearly polarized mode.Under the weakly guiding approximation,
the electromagnetic eld is nearly transverse and each LP mode corresponds to either
x-polarized or y-polarized mode. For an ideal cylindrical ber, the propagation constants
of x-polarized LPmn and y-polarized LPmn are identical. When the refractive index difference between the core and the cladding is large, weakly guiding approximation is not
valid. Under this condition, Ez and/or Hz could be non-zero [5]- [12].
Suppose one of the solutions of Eq. (1.57) be mn . Using Eqs. (1.47), (1.49), (1.42)
and (1.39), the electric eld distribution of this guided mode can be written as

C J ( r) ei(tmn zim) for r a

1 m
1
=
D K ( r) ei(tmn zim) for r a
1 m
2

(1.61)

From Eq. (1.50), we have

D1 =

C1 Jm (1 a)
Km (2 a)

(1.62)

Using Eq. (1.62) in Eq. (1.61), we see that the only unknown parameter in Eq. (1.61)
is C1 which can be determined from the average power carried by this guided mode. In
Section 1.7, the average power density carried by an electromagnetic wave in a homogenous
medium is found to be
Pzav =

||2
2

(1.63)

Under the weakly guiding approximation, the eld is nearly transverse and Eq. (1.63)
may be used to calculate the power. Total power carried by a mode is [8], [9]
}
{ 2 a 2
2
2
(1 a)
Jm (1 r)
Jm
2
2
K (2 r) rdrd
Ptot = |C1 |
rdrd +
2 ( a) m
21
22 Km
2
0
0
0
a
= Fmn |C1 |2 ,
(1.64)
23

where 1 and 2 are the characteristic impedances of core and cladding, respectively. Fmn
can be determined after performing the integrations in Eq. (1.64) numerically. Eq. (1.61)
can be normalized so that the power carried by this mode is unity,
Ptot = 1 or C1 =

1
Fmn

(1.65)

and
= Rmn (r)ei(tmn zim)

where
Rmn (r) =

(1.66)

Jm (1 r) / Fmn

for r a

[J ( a) /K ( a)] K ( r) / F
mn for r > a
m
1
m
2
m
2

(1.67)

Figs. 1.18, 1.19, 1.20 and 1.21 show the optical intensity as a function of radial distance
for various LPmn modes. Total number of guided modes M is given by an approximate
expression [4], [13],
V2
.
M
=
2

(1.68)

1
0.8
0.6
0.4
0.2
0
0.2
0.4
0.6
0.8
1
1

0.5

0.5

Figure 1.18. Optical eld and power distributions of LP01 mode. V =5 and b= 0.84.

24

1
0.8
0.6
0.4
0.2
0
0.2
0.4
0.6
0.8
1
1

0.5

0.5

Figure 1.19. Optical eld and power distributions of LP02 mode. V = 5.3 and b= 0.278.

1
0.8
0.6
0.4
0.2
0
0.2
0.4
0.6
0.8
1
1

0.5

0.5

Figure 1.20. Optical eld and power distributions of LP11 mode. V = 4 and b= 0.44.

25

1
0.8
0.6
0.4
0.2
0
0.2
0.4
0.6
0.8
1
1

0.5

0.5

Figure 1.21. Optical eld and power distributions of LP22 mode. V = 10.3 and b= 0.456.

1.4.2

Mode Cuto

Fig. 1.16 shows the plot of normalized propagation constant b as a function of normalized
frequency V obtained by solving Eq. (1.57). From Fig. 1.16, we see that when V = 10,
there are 6 possible values of b which means there are 6 guided modes. From Eq. (1.59),
it follows that V is large if the ratio of core radius and wavelength is large or the index
dierence is large. When V = 3, the ber supports two modes, LP01 and LP11 . From Fig.
1.16, we see that when V < 2.4048, the ber supports only one mode. This can be proved
as follows. Any LPmn mode ceases to exist when b < 0 because it then corresponds to a
radiation mode since < k0 n2 (Eq. (1.58)). For LP11 mode, when = k0 n2 , using Eq.
(1.54) and Eq. (1.58), we obtain
2 = 0

and b = 0

(1.69)

In this case, using Eq. (1.53) in Eq. (1.59), we nd V = 1 a and with m = 1, Eq. (1.57)
becomes
J0 (V ) = 0

(1.70)

The rst zero of the zeroth order Bessel function (Fig. 1.14) occurs at V = 2.4048.
Therefore, LP11 ceases to exist if V < 2.4048. Thus, when 0 < V < 2.4048, the ber
26

supports only LP01 mode. In other words, to ensure that the ber is single-moded, V
should be smaller than 2.4048.

1.4.3

Eective Index

From Eq. (1.39), we nd that the forward propagating mode can be written as
(r, , z, t) = A1 R(r)() exp [i (t z)]

(1.71)

In analogy with one-dimensional plan wave (Section. ??), the phase speed of this mode
can be written as

v=

(1.72)

The phase speed of electromagnetic wave propagating in a uniform medium of refractive

index ne is
c
ne

(1.73)

ne = k0 ne
c

(1.74)

v=
Combining Eqs. (1.72) and (1.73), we nd
=

Thus, ne can be interpreted as the eective index seen by the mode with propagating
constant . Guided modes occur if k0 n2 < < k0 n1 or n2 < ne < n1 .

1.4.4

2-Dimensional Planar Waveguide Analogy

Consider a 2-D planar waveguide shown in Fig 1.22. Let

n for |x| < d/2

1
n(x) =
n
otherwise
2
and n1 > n2 . For simplicity, let us ignore the variations with respect to y coordinate. The
ray AB corresponds to a plane wave
Eincident = Aeit+i(x+iz) .
27

(1.75)

Figure 1.22. A two dimensional planar waveguide

There is a total internal reection at the interface. The reected ray BC corresponds to
Eref = Aeitix+iz

(1.76)

Note that the z-component of the wave vector does not change after the reection, but
the x-component reverses its sign. The total eld in the waveguide can be written as
E = Eincident + Eref = 2A cos(x)ei(tz)

(1.77)

Thus, incident and reected plane waves set up a standing wave in the x-direction. The
rigorous solution to the planar waveguide problem by solving the Maxwells equation
shows that can take discrete values n and 2n2 /0 < n < 2n1 /0 . In the case of an
optical ber, cos(x) is replaced by the Bessel function and the rest is nearly the same.
In section 1.4, we have found that for single-mode bers, there is only one mode with eld
distribution given by
= A1 R01 (r)eit+i01 z

(1.78)

where 01 is the propagation constant obtained by solving Eq. (1.57). Therefore, a guided
mode of an optical ber can be imagined as a standing wave in transverse directions and
a propagating wave in z-direction resulting from the superposition of the ray AB and the

28

reected ray BC. The propagation constant 01 and angle are related by
01 = k0 n1 sin

(1.79)

Discrete value of the propagation constant 01 implies that can not take arbitrary
values in the interval [c , /2], but only a discrete value as determined from Eq. (1.57)
and (1.79).

1.4.5

Radiation Modes

For radiation modes, 2 < k02 n22 . In this case, the last terms on the left hand sides of Eqs.
(1.44) and (1.45) are both positive and their solutions are given by Bessel functions,

C1 Jm (1 r)
ra
R(r) =
E J ( r) + E Y ( r) r > a
1 m
2
2 m
2
Continuity of the R(r) and dR/dr at the core-cladding interface leads to two equations
as before. But now we have four unknowns C1 , E1 , E2 and . C1 can be determined
from the power carried by the mode and this leaves us with three unknowns and two
equations of continuity. Therefore, we can not write an eigenvalue equation for as was
done in section 1.4.1 for guided modes. In fact, can take arbitrary values in the range
0 < < k0 n2 . Connection with ray optics can be made by dening
= k1 sin i = k0 n1 sin i

(1.80)

where i is the angle of incidence, as shown in Fig. 1.23. The ray undergoes refraction as
it goes from core to cladding if the angel of incidence i < c . When i = c , we have
sin c =

n2
n1

(1.81)

and from Eq. (1.80), we obtain

= c = k0 n 2

(1.82)

Therefore, the condition that < k0 n2 for radiation modes corresponds to i < c and
the ray undergoes refraction, as shown in Fig. 1.23(a). Similarly, the condition that
29

Figure 1.23. (a) Radiation modes correspond to refraction with angle of incidence i < c
(b) Guided modes correspond to total internal reection with i c .

k0 n2 < < k0 n1 for guided modes correspond to c < < /2 and rays undergo
total internal reection for this range of angle of incidence as shown in Fig. 1.23(b). The
dierence between guided modes and radiation modes is that the propagation constants of
guided modes form a discrete set while those of radiation modes are continuous. Radiation
modes do not propagate a longer distance since they are absorbed by the polymer jacket.

1.4.6

Excitation of Guided Modes

Total eld in optical ber can be expressed as a superposition of elds due to guided
modes and radiation modes. Radiation modes are attenuated strongly due to absorption
by the polymer jacket. Therefore, total eld can be expressed as the superposition of
elds due to guided modes given by Eq. (1.66)
(r, , z, t) =

M
Nm

Amn Rmn (r)ei(tmn zim)

(1.83)

m=M n=1

where Nm is the number of solutions of the eigenvalue equation (1.57) for the given m,
Amn is the mode weight factor which is to be determined from launch conditions and
Rmn (r)eim is the transverse eld distribution given by Eq. (1.67). For convenience, Eq.
30

(1.83) may be rewritten as

J

(x, y, z, t) =

Aj j (x, y)ei(tj z)

(1.84)

j=1

where
Aj Amn
j (x, y) Rmn (r)eim
j mn

(1.85)

and J is the total number of modes. Suppose the output of a laser is monochromatic and
it is used as ber input. Fiber input eld can be written as
(x, y, z = 0, t) = f (x, y)eit

(1.86)

Using Eq. (1.86) in Eq. (1.84), we obtain

f (x, y) =

Aj j (x, y)

(1.87)

j=1

To determine Aj , multiply Eq. (1.87) by k (x, y) and integrate over the cross-section to
obtain

f (x, y)k (x, y)dxdy

Aj

j=1

Using the orthogonality relation,

+ +

j (x, y)k (x, y)dxdy

j (x, y)k (x, y)dxdy = jk

where jk is the Kronecker delta function dened as

1, if j = k
jk =
0, otherwise
Eq. (1.88) reduces to

Ak =

f (x, y)k (x, y)dxdy

31

(1.88)

(1.89)

(1.90)

(1.91)

Thus, for the given input eld distribution f (x, y), we can nd the mode weight factors
Ak using Eq. (1.91) and the total eld distribution at any distance z is given by Eq.
(1.84). Suppose the output of the laser has exactly the same transverse distribution as
that of the fundamental mode of the step-index ber, i.e., if f (x, y) = 1 (x, y) = R01 (r),
from Eq. (1.91), we nd A1 = 1 and Am = 0 for m > 1. Therefore, from Eq. (1.84), the
eld distribution at z is
(x, y, z, t) = 1 (x, y)ei(t1 z)

(1.92)

Ideally speaking, fundamental mode LP01 can be launched to the ber which propagates
down the ber without any change in shape and thereby, intermodal dispersion can be
avoided. However, in practice, the ber imperfections and refractive index uctuations
due to temperature and stress can easily transfer power from LP01 mode to higher order
modes. Therefore, the safest way to avoid intermodal dispersion is by ensuring that the
ber is single-moded at the operating wavelength.

Figure 1.24. Mode weight factor versus mode index p. Core radius = 31.25 microns, =
0.01 and R0 = 15 microns

32

Suppose a multimode ber is excited with the Gaussian input

( 2
)
x + y2
f (x, y) = exp
2R02
The mode weight factors Ap can be calculated using Eq. (1.91) and are shown in Fig.
1.24. As can be seen, in this example, most of the power is carried by LP01 mode (p = 1).

1.5

Pulse Propagation in Single-mode Fibers

In the case of multi-mode bers (MMF), pulse broadening occurs because of the dierent
times taken by the dierent rays (or modes) to propagate through the ber. This broadening can be avoided by using single-mode bers (SMF). One may think that in the case
of SMF, there is only one path and hence, pulses should not broaden, but it is not true.
If a monochromatic light wave of innite duration is launched to a SMF, it corresponds
to a single ray path. However, such an optical signal does not convey any information.
To transmit data over a ber, the optical carrier has to be modulated. As a result, the
optical signal propagating in the ber consists of a range of frequency components. Since
the propagation constant is frequency dependent (see Fig.1.16), dierent frequency components undergo dierent amounts of delays (or phase-shifts) and arrive at dierent times
at the receiver, leading to pulse broadening even in a SMF. This is known as intra-modal
dispersion. The degree of pulse broadening in a SMF is much smaller than that in MMF,
but for high rate transmission systems (> 2.5 Gb/s), even the pulse spreading in a SMF
could limit the maximum error-free transmission distance.
When the output of a CW (continuous wave) laser operating at the frequency is
incident on a single-mode ber, the optical eld distribution can be written as (Eq. (1.84)
with j = 1)
(x, y, z, t) = (x, y, )A()ei[t()z]

(1.93)

Mode weight factor A and the transverse eld distribution could vary with frequency .
So far we have assumed that the ber is lossless. In the presence of ber loss, the refractive
33

index appearing in Eq. (1.29) should be complex and as a result, the propagation constant
becomes complex,
() = r () + i()/2

(1.94)

where r () = Re [()] and () = 2Im [()]. Using Eq. (1.94) in Eq. (1.93), we
obtain
(x, y, z, t) = (x, y, )A()e()z/2 ei[tr ()z]

(1.95)

If an optical ber is excited with multiple frequency components, the total eld distribution is the superposition of the elds due to each frequency component,
(x, y, z, t) = (x, y)e

z/2

A(n )ein t+ir (n )z

(1.96)

n=1

In Eq. (1.96), we have ignored the frequency dependency of the transverse eld distribution and also that of the loss coecient . This is valid if the frequency spread
= |N 1 | is much smaller than the mean frequency of incident eld. If the incident
eld envelope is a pulse, its frequency components are closely spaced and we can replace
the summation in Eq.(1.96) with an integral
(x, y, z, t) = (x, y)F (t, z)
where

ez/2
F (t, z) =
2

i[tr ()z]
e
A()e
d

(1.97)

(1.98)

A(n )
e
A()
= 2 lim
n 0 n

(1.99)

From Eq.(1.98), we have

1
F (t, 0) =
2

it
e
A()e
d

(1.100)

e
Eq. (1.100) represents the inverse Fourier transform of A().
Therefore, the Fourier
e
transform A()
of the incident pulse F (t, 0) is
+
e
A()
=
F (t, 0)eit dt

34

(1.101)

e
Thus, for the given incident pulse shape, we can calculate A()
using Eq. (1.101) and
the optical eld distribution at any z can be calculated using Eqs. (1.97) and (1.98). The
impact of the ber is characterized by (). However, in practice, the dependence of the
propagation constant on the frequency for the commercially available bers are not known.
Besides, from the ber-optic system design point of view, it is desirable to characterize
the ber using a few parameters. Therefore, we do the following approximation. The
propagation constant at any frequency can be written in terms of the propagation
constant and its derivative at some reference frequency (typically the carrier frequency)
0 using Taylor series,
1
r () = 0 + 1 ( 0 ) + 2 ( 0 )2 +
2

(1.102)

where
0 = r (0 )

dr 
1
1 =
=

d =0
g

d2 r 
2 =
d 2 

(1.103)
(1.104)
(1.105)

=0

1 is the inverse group velocity and 2 is the second order dispersion coecient (see
section ??). If the signal bandwidth is much smaller than the carrier frequency 0 , we
can truncate the Taylor series after the second term on the right hand side. To simplify
Eq.(1.98), let us choose the variable = 0 . Using Eq. (1.102) in Eq. (1.98), we
obtain
+
[
]
1
e
B()
exp z/2 i (0 t 0 z) + i1 z + i2 2 z/2 exp (it) d
F (t, z) =
2

(
)
exp [z/2 i (0 t 0 z)] + e
=
B() exp i1 z + i2 2 z/2 it d
2

exp [i (0 t 0 z)] + e
=
B()Hf (, z) exp (it) d
(1.106)
2

where

(
)
Hf (, z) = exp z/2 + i1 z + i2 2 z/2
35

(1.107)

is called the ber transfer function and

0 + )
B()
A(

(1.108)

The linear phase shift 1 z corresponds to a delay in time domain. To see that, set 2 = 0
in Eq.(1.107) and the ber output z = L,

exp [z/2 i (0 t 0 L)] + e

F (t, L) =
B() exp [i (t 1 L)] d
2

= exp [z/2 i (0 t 0 L)] B (t 1 L)

(1.109)
In a dispersion-free ber (2 = 0), the pulse is simply delayed by 1 L at the ber output
without any change in pulse shape, as in the free space propagation. Using Eqs. (1.98)
and (1.106), the optical eld distribution can be written as
(x, y)
| {z }

exp [i (0 t 0 z)]
|
{z
}

s(t, z)
| {z }

transverse eld

carrier

eld envelope

(x, y, z, t) =
where
1
s(t, z) =
2

and
e
B()
=

e
B()H
f (, z) exp (it) d

(1.110)

(1.111)

s(t, 0) exp (it) dt

(1.112)

Eqs. (1.111) and (1.112) can be rewritten as

[
]
1 e
s(t, z) = F
B()Hf (, z) ,

(1.113)

e
B()
= F [s(t, 0)] ,

(1.114)

e
B()

s(t, 0),

(1.115)

where, F and F 1 denote Fourier and inverse Fourier transforms, respectively, and

indicates that they are Fourier transform pairs. In this section, we mainly focus on the
eld envelope s(t, z). Let us assume that the transverse eld distribution of the laser
output is the same as that of the ber, and therefore, there is no change in the transverse
eld distribution along the ber. Let the eld envelope of the laser output be si (t),
si (t) = s(t, 0)
36

(1.116)

and
e
B()
= F [si (t)] = sei ()

(1.117)

The ber can be imagined as a linear system with the transfer function Hf (, z). The
impact of the ber nonlinearity is discussed in Chapter 10. Let the eld envelope of the
ber output s(t, L) be so (t),
s(t, L) = so (t)
F [so (t)] = seo () = Hf (, L)e
si ().

(1.118)
(1.119)

The optical signal propagation in a single-mode ber can be summarized as follows.

Figure 1.25. Optical ber as a linear time-invariant system.

Step 1: Input eld envelope si (t) is known. Take its Fourier transform to obtain sei ().
Step 2: Multiply sei () by Hf (, L) to get the output spectrum, seo ().
Step 3: Take the inverse Fourier transform of seo () to obtain the output eld envelope
so (t).
Step 4: Total eld distribution at the output is obtained by
(x, y, L, t) = (x, y) exp [i (0 t 0 L)] so (t)

(1.120)

The advantage of this approach over that uses Eq. (1.98) is that the ber is characterized
by three parameters 0 , 1 and 2 instead of (). As the spectral width of the signal
transmitted over the ber increases, it may be necessary to include the higher order
37

dispersion coecients such as 3 and 4 . 1 and 2 can be measured experimentally even

if the ber index prole is unknown. For example, by transmitting the output of a cw
laser of angular frequency 0 over a ber of length L, the time of ight T0 to traverse
the distance L can be measured and 1 (0 ) is T0 /L. Repeating the same experiment at
0 + , 1 (0 + ) can be calculated. 2 can be estimated as
1 (0 + ) 1 (0 )
2
=

1.5.1

(1.121)

Power and dBm Unit

The average power density of a plane wave is given by (cite rst chapter)
 2
e 
Ex 
Pzav =
2

(1.122)

ex is the peak amplitude of the electric eld intensity and is the intrinsic
where E
impedance of the dielectric medium. A plane wave has innite spatial extension in x
and y directions and therefore, the power carried by a plane wave is innite. Under the
LP mode approximation, a ber mode can be interpreted as a plane wave with nite
spatial extension in x and y directions and therefore, power carried by a ber mode can
be obtained by integrating the absolute square of electric eld intensity as done in Eq.
(1.64),

 2 1
|s(t, z)|2 |(x, y)|2
e 
dxdy =
dxdy
Ex 
2
2

P =

= K |s(t, z)|2

(1.123)

Thus, we see that the power is proportional to the absolute square of the eld envelope
s(t, z). Throughout this book, unless otherwise specied, we set K = 1 so that absolute
square of the electric eld envelope is equal to the power.
Often it is convenient to use the logarithmic unit for power. The optical power in
dBm unit is expressed as

Power (dBm) = 10 log10

38

]
Power (mW)
.
1 mW

(1.124)

In Eq. (1.124), 1 mW is chosen as a reference power and the letter m in dBm is a

reminder of the 1 mW reference. For example, 1 mW of transmitter power corresponds
to 0 dBm. If the transmitter power is increased to 2 mW, a factor of 2 in linear scale
corresponds to 3 dB and therefore, in this case, the transmitter power is 3 dBm. Note
that the optical power expressed in dBm units is not really a unit of power such as mW,
but the ratio of the power in mW and 1 mW expressed in dB unit. Typically, the loss
and gain in ber-optic system are expressed in dB units. The advantage of using dBm
unit is that multiplications and divisions involving power and loss factors can be replaced
by additions and subtractions as illustrated in Problems 2.7 and 2.8.
Inverting Eq. (1.124), we nd
Power (mW) = 10Power (dBm)/10 mW.

(1.125)

Problem 2.3
The power transmitted in a ber-optic system is 0.012 W. (a)Convert this into dBm units.
(b)The received power is 5 dBm. Convert this into mW units.

Solution:
(a) From (1.124), the transmitted power in dBm unit is
[
]
12 mW
Ptr (dBm) = 10 log10
= 10. 79 dBm.
1 mW

(1.126)

(b) The received power is

Prec (dBm) = 5dBm,

(1.127)

Prec (mW ) = 105/10 mW = 0.3162 mW.

(1.128)

Using Eq.(1.125), we nd

39

Problem 2.4: Rectangular Pulse

Figure 1.26. Pulse propagation in free space and optical ber.

The laser shown in Fig. 1.26 operates at 375 THz. It is turned on for 50 ps and then
turned o. Sketch the eld envelope at the screen if the medium is (a) free space (b) ber
with 2 = 0 (c) ber with 2 = 21 ps2 /km. Ignore ber loss.

Solution:
(a) Under the steady state conditions, the electric eld intensity of a cw laser (ignoring
the transverse eld distribution) may be written as
F (t, 0) = f (t) = A exp[i2f0 t],

(1.129)

where f0 = 375 THz. When the laser is turned on for 50 ps and then turned o, a
rectangular pulse is generated and in this case, the electric eld intensity is
F (t, 0) = f (t) = si (t) exp[i2f0 t],
40

(1.130)

where

(
si (t) = Arect

t
T0

)
,

(1.131)

and T0 = 50 ps.
(a) In Section 1.6, the electric eld intensity at the screen (z = L) is found to be
F (t, L) = f (t T1 ) = so (t T1 ) cos [2f0 (t T1 )] ,
where

(
so (t) = rect

t T1
T0

(1.132)

)
,

(1.133)

and T1 = L/c, and c is the velocity of light in the free space. The eld is delayed by
T1 = L/c which is the propagation delay as shown in Fig. 1.27.

Figure 1.27. In free space, the pulse shape does not change.
(b) In the case of an optical ber, let us rst consider the case 2 = 0,
Step 1:

(
si (t) = Arect

sei (f ) = A

t
T0

T0 /2

exp (i2f t) dt =
T0 /2

A sin (f T0 )
f

(1.134)

Step 2: The transfer function of a loss-free ber in the absence of 2 is

Hf (f, L) = exp (i2f 1 L)
41

(1.135)

seo (f ) = sei (f )Hf (f, L) =

A sin (f T0 )
exp (i2f 1 L)
f

(1.136)

Step 3: The delay in time domain corresponds to a constant phase shift in frequency
domain,
g(t T0 )
ge(f ) exp (i2f T0 )
Using Eq.(1.134) and (1.137), the output eld envelope may be written as,
(
)
t 1 L
so (t) = si (t 1 L) = rect
T0

(1.137)

(1.138)

Fig. 1.28 shows the eld envelope. As can be seen, there is no change in the pulse shape
at z = L. It is simply delayed by 1 L similar to the case of free-space propagation.

Figure 1.28. The eld envelopes at the laser and at the screen. In optical bers with 2 = 0,
the pulse shape does not change.

(c) When 2 = 0, Eq. (1.137) may be written as

seo (f ) =

[
]
A sin (f T0 )
exp i2f 1 L + i(2f )2 2 L/2
f
so (t) = F 1 [e
so (f )]

(1.139)
(1.140)

It is not possible to do the inverse Fourier transform analytically. Fig. 1.29 shows the
output eld envelope so (t) obtained using numerical techniques when 2 = 21 ps2 /km
and L = 80 km. As can be seen, there is a signicant pulse broadening after ber
42

Figure 1.29. Optical eld envelopes when 2 = 0. L = 80 km, 2 = 21 ps2 /km.

propagation.
Step 4: Total eld distribution at the ber output is
(x, y, L, t) = (x, y) exp [i (0 t 0 z)] s0 (t).

(1.141)

Fig. 1.30 shows the total eld distribution at the ber input and output (transverse eld

Figure 1.30. Total eld distribution at the laser and at the screen when 2 = 0.
distribution is not shown).

43

Problem 2.5: Gaussian Pulse

The input eld envelope is

(
)
si (t) = A exp t2 /2T02

(1.142)

where T0 represents the half-width at 1/e-intensity point and A is the peak amplitude.
Find the output eld envelope in a dispersive ber. Ignore ber loss and constant delay
due to 1 .

Solution:
in
To relate T0 to full width at half-maximum (FWHM) TFWHM
, let us rst write an equation

for power
P (t) = |si (t)|2 = A2 et

2 /T 2
0

Pmax = P (0) = A2 ,

(1.143)
(1.144)

Let th be the time at which the power is half of the peak power as shown in Fig. 1.31.
Since FWHM means the full width at half-power point, we have

Figure 1.31. A Gaussian pulse

)
(
P (th ) = A2 /2 = A2 exp t2h /T02 .
44

(1.145)

Taking logarithms on both sides, we obtain

th = T0 (ln 2)1/2 .

(1.146)

in
TFWHM
= 2th = 2 (ln 2)1/2 T0 1.665T0 .

(1.147)

and

The transfer function of an optical ber in the absence of ber loss is given by Eq. (1.107)
as

[
]
Hf (f, L) = exp i1 (2f ) L + i2 (2f )2 L/2

(1.148)

As mentioned before, the rst term on the right hand side introduces a constant delay
and hence, it can be ignored for the purpose of evaluating the output pulse shape. Using
the following identity

(
)
(
)
exp t2
exp f 2 ,

(1.149)

where
indicates that they are Fourier transform pairs and using the scaling property
1
g(at)
ge (f /a) ,
a

Re(a) > 0,

(1.150)

Fourier transform of si (t) can be calculated. Taking

1
a=
,
2T0

(1.151)

[
]
[
]
A
si (t) = A exp (at)2
exp (f /a)2 = sei (f ).
a

(1.152)

Therefore, we have
seo (f ) = sei (f )Hf (f, L)
[
]
A
f 2
2
=
exp 2 + i2 (2f ) L/2
a
a
(
)
A
=
exp f 2 /b2 ,
a

(1.153)

where
1
1
= 2 i22 L
2
b
a
45

(1.154)

Using Eqs. (1.149 and (1.150), the inverse Fourier transform of seo (f ) is
so (t) =

[
]
Ab
exp (tb)2 .
a

(1.155)

Using Eqs.(1.151) and (1.154), we have

1/2

a
[T02 i2 L]
=
b
T0
2
t
b2 t2 =
2 (T02 i2 L)

(1.156)
(1.157)

Therefore, the output eld envelope is

so (t) =

AT0
1/2

(T02 i2 L)

t2
exp
2 (T02 i2 L)

]
(1.158)

To nd the pulse width at the output, let us rst calculate the output power,

]2
[
2 2


A
T
t2
0
2
 .

Po (t) = |so (t)| = 
2 exp
 2
2 (T02 i2 L) 
1/2 
(T0 i2 L) 

(1.159)

Since

we obtain

t2
t2 (T02 + i2 L)
=
,
2 (T02 i2 L)
2 (T04 + 22 L2 )

(1.160)

( 2)
T0 A2
t
Po (t) =
exp 2
T1
T1

(1.161)

where T12 = (T04 + 22 L2 ) /T02 . The FWHM at the output is given by

out
TFWHM

= 2 (ln 2)

1/2

1/2

T1 = 2 (ln 2)

(T04 + 22 L2 )
T0

1/2

(1.162)

To determine the amount of pulse broadening, the ratio of output and input pulse widths
is calculated as

1/2

T out
(T04 + 22 L2 )
= FWHM
=
in
TFWHM
T02
When
|2 L| =

46

3T02 ,

(1.163)

(1.164)

we nd = 2 which means that the output pulse width is twice the input pulse width.
Note that the amount of pulse broadening is independent of the sign of the dispersion
coecient 2 .
The frequency chirp or instantaneous frequency deviation is dened as
(t) =

d
dt

(1.165)

where is the instantaneous phase of the eld envelope and (t) is the instantaneous
frequency deviation from the carrier frequency. Note that the optical carrier is of the form
exp(i0 t). A negative sign is introduced above so that a positive value of implies
that the frequency is upshifted. At the ber output, from Eq. (1.160), the instantaneous
phase is
(t) =

t2 2 L
+ const
2T12 T02

Substituting Eq. (1.166) in Eq. (1.165), we nd

(
)
2 L
(t) =
t
T12 T02

(1.166)

(1.167)

Fig. 1.32 illustrates the evolution of power and chirp along the ber length. At
the ber output, the pulse becomes chirped and the sign of chirp depends on the sign of
dispersion coecient 2 . In Figs. 1.32b and c, we see that the trailing edge is down-shifted
in frequency (or red shifted) and the leading edge is up-shifted (or blue shifted). This
can be explained as follows. When the dispersion is anomalous (2 < 0), high frequency
components of the pulse travel faster than the low frequency components (See Section ??).
Since these components arrive at dierent times, this leads to pulse broadening. Since
the low frequency components of the pulse (components whose frequency is lower than
the carrier frequency) travel slower, they arrive late and therefore, they are present near
the trailing edge which is another way of saying that the trailing edge is down-shifted in
frequency.

47

Figure 1.32. Evolution of unchirped Gaussian pulse in optical ber. 2 = 21 ps2 /km,
T0 = 30 ps.

1.6

Comparison between multi-mode and single-mode

bers

Multi-mode bers (MMFs) have several advantages over single-mode bers (SMFs). The
core radius of multi-mode ber (25 35 m) is much larger that that of single-mode ber
(4 9 m). Therefore, it is easier to launch optical power into a MMF and also to splice
two MMFs. Large core of MMF facilitates simple ber-to-ber or ber-to-transceiver
alignement and hence, is best suited to local area network (LAN) applications [14]. The
relative index dierence of a MMF is larger than that of a SMF. Therefore, the numerical
aperture of a MMF is large which implies more light can be launched to ber from
an inexpensive optical source that has a large angular spread, such as a light emitting
diode(LED). To have a reasonable power coupling eciency, SMFs are excited with laser
diodes. Inexpensive short haul ber-optic links can be designed using LED and multimode bers. However, multi-mode bers are not used for long haul and/or high bit rate
48

applications because of intermodal dispersion. Although the dispersion can be reduced to

some extent using graded-index multi-mode ber, the pulse broadening increases linearly
with distance (Eq. (1.24) and it becomes unacceptably large for a ber-optic link that
is hundreds of kilometers long. Typically, the transmission reach of a MMF ber-optic
link at a bit rate of 1 Gb/s is limited to a few kilometers. Intermodal dispersion would
be absent if there is only one mode. Therefore, single-mode bers are used for long haul
(1000km-30000 km) and high bit rate (10 Gb/s-100 Gb/s) applications.
From the information theory point of view, the channel capacity of a multi-mode ber
is larger than that of a single mode ber. This is because, in principle, each mode of
a MMF can carry information as much as that by a SMF. When dierent modes of a
MMF carry independent sets of data, it is known as mode division multiplexing which
has drawn signicant attention recently [15]- [19]. In an ideal MMF with M guided
modes, there is no power-coupling among modes and the channel capacity can be enhanced
by the factor M. However, due to refractive index uctuations along the ber, there is
exchange of power among modes leading to crosstalk between channels of a mode division
multiplexed system. This crosstalk can be compensated using the digital signal processing
techniques [15].

1.7

Single-Mode Fiber Design Considerations

The parameters that are important for the design of a single-mode ber are (i) cuto wavelength (ii) ber loss (iii) dispersion (iv) dispersion slope,(v) polarization mode dispersion,
and (vi) spot size. Using a step-index optical ber, it is not possible to optimize all these
parameters. Therefore, refractive index prole n(r) is so chosen that the design parameters listed above are optimum for a specic application. For the given refractive index
prole n(r), the Helmholtz equation (1.28) is solved to obtain the propagation constant
() and the mode distribution function (x, y). From this data, design parameters can
be calculated. As an inverse problem, refractive index prole n(r) can be constructed to

49

meet the given specications on the design parameters. However, in some cases, solution
to the inverse problem does not exist. For example, it is desirable to have a large spot size
(to reduce nonlinear eects) as well as low dispersion slope to improve the performance
of a WDM system. But it turns out that as the spot size increases the dispersion slope
also increases. In the following subsections, important design parameters of a single-mode
ber and their interrelationships are discussed.

1.7.1

Cut-O Wavelength

For high bit-rate and long-haul applications, it is essential that the ber is single-moded.
The single mode condition for a step-index ber is given by Eqs. (1.70) and (1.59),
V =

)1/2
2a ( 2
n1 n22
2.4048

(1.168)

For example, if = 1.55 m, a = 4 m and (n21 n22 )1/2 = 0.1, V = 1.62. Therefore,
this ber is single-moded at this wavelength. However, if = 0.7 m corresponding to
the optical communication in the visible spectrum, V becomes 3.59 and the ber is not
single-moded at this wavelength. For the given ber parameters, the cut-o wavelength
is dened as
2a(n21 n22 )1/2
2.4048
2a N A
=
2.4048

c =

(1.169)

If the operating wavelength is less than c , ber will not be single-moded. For bers
with arbitrary index proles, the Helmholtz equation (1.28) should be solved numerically
to nd the propagation constants n as a function of frequency from which the conditions
for the cuto of higher order modes can be established.

Problem 2.6
The cuto wavelength for a step-index ber is 1.1 m. Core index n1 = 1.45 and =
0.005. Find the core radius. Is this ber single-mode at 1.55 m?
50

Solution:
From Eq. (1.8), we nd
n2 = n1 (1 ) = 1.4428.

(1.170)

Using Eq. (1.169), we nd

a=

2.4048c
= 2.907 m.
2(n21 n22 )1/2

(1.171)

Since the operating wavelength = 1.55 m > c , it is single-moded at this wavelength.

1.7.2

Fiber Loss

Before the advent of optical ampliers, the maximum transmission distance of a ber-optic
system was determined by the ber loss as the optical receivers need a certain amount
of optical power to detect the transmitted signal reliably. Now the optical ampliers are
widely used and yet the maximum reach is aected by the ber loss. This is because the
optical ampliers add noise whose power spectral density is proportional to amplier gain,
which in turn is proportional to ber loss (See Chapter 6). In other words, the amount of
noise in a long haul communication system is directly related to ber loss. In addition,
if the ber loss is small, the amplier spacing can be increased which reduces the system
cost. So, it is important to design a ber with the lowest possible loss.
Let us consider a cw input to the ber. the optical eld distribution is given by Eq.
(1.95)
(x, y, z, t) =

A() exp (z/2) exp [i (t r ()z)]

|
{z
}|
{z
}

(x, y, )
| {z }
transverse distribution

eld envelope

(1.172)

optical carrier

The optical power is given by Eq. (1.123),

P (z) = |A() exp (z/2)|2

(1.173)

Pin = P (0) = |A()|2

(1.174)

At the ber input,

51

At the ber output z = L,

Pout = P (L) = |A()|2 exp (L) = Pin (t) exp (L)

(1.175)

The optical power loss in dB unit due to propagation in a ber of length L is dened as
by
loss(dB) = 10 log10

Pout
= 10(L) log10 e = 4.343L
Pin

(1.176)

Here, is the attenuation coecient in the unit of km-1 . The loss per unit length is
(dB/km) = 4.343.

(1.177)

Next, let us consider the origin of ber loss. The lightwave is attenuated as it propagates in
ber mainly due to (i) Rayleigh scattering, and (ii) material absorption. In the following
subsections, we discuss these mechanisms in detail.
1.7.2.1

Rayleigh Scattering

Consider a perfect crystal with uniformly spaced atoms or molecules. When a lightwave
is incident on this crystal, electrons in atoms oscillate and emit lightwaves of the same
frequency as the incident lightwave under the linear approximation (See Chapter 10).
In other words, each atom acts as a tiny receiving and transmitting antenna. The light
emitted by an atom could be in all directions. However, for a perfect crystal with uniformly
spaced atoms or molecules, it can be shown that the emitted lightwaves add up coherently
in the direction of incident lightwave; in any other direction, we get no light as they add
up destructively [20]. In other words, in a perfect crystal, there is no scattering of incident
light. Next consider a crystal with defects such as atoms missing or they are irregularly
placed in the lattice structure. In this case, lightwaves emitted by atoms may not add up
destructively over a range of directions which leads to scattering.
Rayleigh scattering is the scattering of light by atoms or molecules of size much smaller
than the wavelength of the light. It is an important mechanism arising from local microscopic uctuations in density and compositional variations. The uctuations in density
52

corresponds to irregularly spaced atoms or molecules in a lattice structure and as a result, incident light is scattered over a range of angles as shown in Fig.1.33. If the angle of
scattering is less than the critical angle, it will escape to the cladding and then will be
absorbed at the polymer jacket. A part of the optical eld is back-reected as well due
to Rayleigh scattering which propagates as a backward propagating mode. These eects
lead to loss of power in the forward propagating direction. The loss coecient due to
Rayleigh scattering can be written as

Figure 1.33. Rayleigh scattering in optical bers

1
,
4

(1.178)

Because of the strong wavelength dependence of Rayleigh scattering, short wavelengths

(blue) are scattered more than long wavelengths (red). The scattering at 400 nm is 9.2
times as great as that at 700 nm for equal incident intensity. Rayleigh scattering of sun
light in clear atmosphere is the reason why the sky is blue. This also explains why the
sun looks red in the morning/evening. The distance between the sun and an observer
is large in the morning/evening and the light has to go through a thicker atmosphere
causing the lower wavelengths (violet, blue) to suer higher losses (see Eq.(1.178)) and
their intensities would be too low to detect.
The dominant contributions to the ber loss comes from Rayleigh scattering in the
wavelength range of practical interest, 1550 nm-1620 nm. One of the reasons why the

53

ber-optic communication systems operate in infra-red region instead of visible region

(400 nm-700 nm) is that the loss due to Rayleigh scattering is much smaller in the former
region because of its 4 dependence. Fig. 1.34 shows the measured loss spectrum of a
single-mode ber with 9.4 micron core diameter and = 0.0019 [3]. As can be seen, the
lowest ber loss occurs at 1.55 m wavelength. For the silica ber, at = 1.55 m, the
loss due to Rayleigh scattering alone is R = 0.1559 dB/km. Thus, for the ber shown in
Fig. 1.34, 77% of the total loss at 1.55 m comes from Rayleigh scattering.
Conventional optical bers are fabricated by doping the silica with GeO2 . The addition
of small amount of GeO2 increases the refractive index and therefore, the enhancement of
core refractive index relative to cladding index is achieved. However, the addition of GeO2
increases the Rayleigh scattering. Therefore, eorts have been made to fabricate pure
silica core bers (PSCF) in which the core is pure silica [21] - [22]. The refractive index
of cladding is reduced relative to the core index by adding a small amount of Flourine.
Since most of the light is conned to core, the pure silica core ber has lower Rayleigh
scattering coecient than the conventional silica-GeO2 core ber. The attenuation of
1570 nm for PSCF is 0.154 dB/km [21] which is the lowest attenuation reported whereas
the lowest attenuation for silica-GeO2 core ber is about 0.19 dB/km .
1.7.2.2

Material Absorption

Material absorption can be divided into two types. (a) Intrinsic absorption, and (b)
extrinsic absorption.
Intrinsic Absorption This loss is caused by the interaction of light with pure-silica.
An electron in the silica molecule absorbs light and it makes a transition from one electron
state to another. This kind of resonances occur in the ultraviolet region ( < 0.4 m)
for silica and the tail of the absorption band extends throughout the visible spectrum.
A photon could interact with a molecule causing a change in its vibrational state. This
also leads to photon absorption or optical power loss. This kind of vibrational resonances

54

Figure 1.34. The measured loss spectrum of a single-mode ber. (After Ref. [3].(c) IET.
Reprinted with permission.

55

occur in the far infrared region > 7 m and tail of the vibrational resonances can be
seen in Fig. 1.34 for > 1.6 m.
Extrinsic Absorption This loss is caused by the interaction of light with impurities in
silica. Metal impurities such as Cu, Fe, Cr, Ni and V leads to a strong signal attenuation.
These impurities can be reduced to less than one part in 1010 by glass rening techniques
such as vapour-phase oxidation [23]. One of the major source of extrinsic absorption is
the water vapors present in silica bers. The OH ion of the water vapor is bonded into
the glass structure and has a fundamental vibrational resonance at 2.73 m. Its overtones
and combination tones with the fundamental silica vibrational resonances lead to strong
absorption at 1.38, 1.24, 0.95 and 0.88 m wavelengths. As shown in Fig. 1.34, the
absorption at 1.31 m is the strongest and its tail at 1.3 m was the main hurdle for the
development of ber-optic communication systems at 1.3 m. Eorts have been made to
reduce the absorption at 1.31 m to less than about 0.35 dB/km by reducing the water
content in the glass [24]- [25]. Majority of the ber-optic systems operate around the
wavelength window centered around 1.3 m and 1.55 m. This is because the window
centered around 1.3 m has the lowest dispersion for the standard SMF and the window
around 1.55 m has the lowest loss.

Problem 2.7
A ber of length 80 km has a loss coecient of 0.046 km1 . Find the total loss. If the
power launched to this ber is 3 dBm, nd the output power in mW and dBm units.

56

Solution:
The loss per unit length (dB/km) = 4.3430.046 = 0.2 dB/km. Total loss = 0.280 = 16
dB. From Eq. (1.124), we have
P (dBm) = 10 log10

P (mW)
.
1 mW

(1.179)

From Eq. (1.175), we nd

Pout (mW) = Pin (mW) exp(L).
Dividing Eq. (1.180) by 1 mW and taking logarithms, we nd
{
}
Pin (mW) exp(L)
Pout (dBm) = 10 log10
1 mW
= Pin (dBm) + 10 log10 exp(L)
Pout
.
= Pin (dBm) + 10 log10
Pin

(1.180)

(1.181)

Using Eq. (1.176), we nd

Pout (dBm) = Pin (dBm) loss(dB)
= 3 dBm 16 dB
= 13 dBm.

(1.182)

Note that using dBm unit, multiplication (Eq. (1.180)) is replaced by subtraction (Eq.
(1.182)). Using Eq. (1.125), we nd
Pout (mW) = 1013/10 mW = 0.05mW.

(1.183)

Problem 2.8
Consider a ber-optic system consisting of a ber with loss F followed by an amplier of
gain G. The launch power is Pin . Calculate the output power of the amplier in dBm
unit.
57

Figure 1.35. A ber-optic link consisting of a ber and an amplier

Solution:
Fiber-out
,
Pin
Fiber-out
Fiber loss in linear units, F =
.
Pin

Fiber loss (dB) = F (dB) = 10 log10

Fiber-out = F Pin .

(1.184)
(1.185)
(1.186)

Similarly, the amplier output power is

Pout = GF Pin .

(1.187)

From Eq. (1.124), we nd

P (dBm) = 10 log10

P (mW)
.
1 mW

(1.188)

Suppose that Pout and Pin in Eq. (1.187) are given in units of mW. Divide Eq. (1.187)
by 1 mW and take logarithms both sides to obtain
10 log10

[
]
GF Pin
Pout
= 10 log10
,
1 mW
1 mW

(1.189)

Pout = 10 log10 G + 10 log10 F + 10 log10 Pin

= G(dB) F (dB) + Pin (dBm).

(1.190)

Note that any loss (such as ber attenuations) in the system is subtracted from the input
power in dBm units and any gain (such as amplier gain) is added.

58

1.7.3

Fiber Dispersion

A medium is said to be dispersive if the group speed of light changes with frequency of the
optical wave. For example, a pulse p1 with carrier frequency f1 = 193 THz and inverse
group speed 1 (f1 ) = 50 s/km is delayed by
T1 = 1 (f1 )L = 50 s

(1.191)

after propagating through a 10 km ber. Consider another pulse p2 with a dierent carrier
frequency f2 = f1 + f . If the ber is not dispersive (2 = 0), the inverse group speed
2 (f2 ) is same as 1 (f1 ) and therefore, the pulse p2 is delayed by the same amount
T2 = 1 (f2 )L = 50 s

(1.192)

In this case, the dierential delay between the pulses is zero. Next consider a dispersive
ber with 2 = 100 ps2 /km at 193 THz. By denition
d2 0
d1
1 d1
=
=
2
d
d
2 df
1 1 (f1 + f ) 1 (f )

2
f

2 () =

(1.193)

or
1 (f2 ) = 1 (f1 ) + 2f 2

(1.194)

Let f = 1 THz. The pulse p2 is delayed by

T2 = 1 (f2 )L = 1 (f1 )L + 2f 2 L
= 50.00628 s.

(1.195)
(1.196)

The dierential delay between the pulses is given by

T = T2 T1 = 2f 2 L
= 6.28 ns.

(1.197)
(1.198)

In other words, the pulse p2 arrives at the ber output later than pulse p1 by 6.28 ns.
Instead of nding the derivative of 1 with respect to frequency as in Eq (1.193), we could
59

dene its derivative with respect to wavelength

D=

d1
d

(1.199)

where D is called the Dispersion Parameter. Since

c

c
df = 2 d

f =

(1.200)
(1.201)

Using Eq. (1.201) in Eq. (1.199) and making use of Eq. (1.193), we obtain
D=

2c
2
2

(1.202)

Substituting Eq. (1.202) in Eq. (1.197), the dierential delay can be rewritten as
T = DL

(1.203)

where = f /f12 . The above relation can be understood from the fact that 1 is
the delay per unit length and D is the delay per unit length per unit wavelength (Eq.
(1.199)).
Fiber dispersion can be divided into two parts (i) material dispersion (ii) waveguide
dispersion. Material dispersion is due to the frequency dependence of the refractive index
of glass. Just like a prism spreads the white light into rain bow of colors (See section
??), dierent frequency components travel at dierent speeds in glass leading to pulse
spreading. The second contribution to the ber dispersion comes from the waveguide eect
and it is known as waveguide dispersion. The dependence of the propagation constant
on frequency can be varied by changing the refractive index prole. For example, if
we change the refractive index prole from step-index to parabolic index, the dispersion
coecient 2 could vary signicantly. In a hypothetical case in which the refractive index
of core/clading does not change with frequency, ber dispersion coecient 2 could be
non-zero because of waveguide dispersion. The product of the dispersion parameters D
and ber length is called accumulated dispersion.
60

A curve tting to experimentally measured dispersion parameter of a standard singlemode ber (SSMF) is given by

[
]
S0
40
3 ps/ (nm km)
D() =
4

(1.204)

where 0 = 1317 nm and S0 = 0.088 ps/ (nm2 km). From Eq. (1.204), at = 0 ,
D(0 ) = 0 and therefore, 0 is called the zero dispersion wavelength and

dD 
= S0
d =0

(1.205)

S0 is called the dispersion slope at 0 . Fiber loss is the lowest at 1550 nm and therefore,
most of the optical communication systems operate in the wavelength range 1520 nm 1620 nm. In this wavelength range, the dispersion parameter of a standard SMF is quite
high which leads to strong inter-symbol interference (ISI) at the receiver. To avoid this
problem, dispersion-shifted (DS) bers were invented in 1980s and 1990s [26]- [27] which
have 0 = 1550 nm. In the absence of ber nonlinearity, ideal characteristics of a ber are
low dispersion parameter |D| and low loss which can be achieved using DS bers. However,
it was soon realized that DS bers are not suitable for wavelength-division multiplexed
(WDM) systems since nonlinear interactions between channels such as four wave mixing
(FWM) and cross-phase modulation (XPM) are enhanced because of low dispersion (see
Chapter 10). In mid 1990s, non-zero dispersion-shifted bers (NZ-DSF) were invented
for which 0 is chosen to be out of the wavelength region 1530 nm-1565 nm [28]. For
NZ-DSF, dispersion near 1550 nm is large enough to suppress the nonlinear eects and
yet low enough to avoid strong ISI due to dispersion. Alternatively, the dispersion of the
transmission ber can be compensated by using dispersion compensating bers (DCF)
(see section 1.8) or using equalizer in electrical domain. This topic is discussed in detail
in Chapter 11.

1.7.4

Dispersion Slope

Dispersion parameter depends on the wavelength. For a single channel low bit rate optical
communication systems, the spectral width f (and therefore, wavelength spread, || )
61

is quite small and dispersion parameter D() can be considered as a constant over a small
. However, for a high bit rate wide band communication system, the dependence of
D on the wavelength can not be ignored. For example, the spectral width of an optical
non-return-to-zero (NRZ) signal at B = 160 Gb/s is roughly (see Chapter 4),
f 2B = 320 GHz

(1.206)

f
=

(1.207)

Since c = f ,

At the carrier wavelength c = 1550 nm, = 2.56 nm. The change in dispersion over
such a large wavelength spread around 1550 nm can not be ignored. It is useful to dene
the dispersion slope as
S=

dD
d

Using Eq. (1.204), dispersion slope for a standard SMF can be calculated as
[
]
S0
340
S=
1+ 4
4

(1.208)

(1.209)

If the the dispersion parameter Dc and the dispersion slope Sc at the carrier wavelength
c are known, the dispersion parameter in the vicinity of the carrier wavelength can be
obtained by the linear approximation [29]
D() = Dc + Sc ( 0 )

(1.210)

In Eq. (1.102), we have retained the Taylor expansion terms up to 2 2 /2. Under this
approximation, 2 or D is constant over the spectral width of the signal. To include the
impact of dispersion slope, we need to include a higher order term in the Taylor expansion
of Eq. (1.102),

where

() = 0 + 1 + 2 2 /2 + 3 3 /6

(1.211)



d2 
d3 
=
3 =
d 3 =0
d =0

(1.212)

62

Substituting Eq. (1.211) in Eq. (1.98), we obtain

e
F (t, z) = exp [i (0 t 0 z)]
B()H
f (, z) exp (it) d,

(1.213)

where the ber transfer function is modied as

[
]
Hf (, z) = exp z + i1 z + i2 2 z/2 + i3 3 z/6

(1.214)

The eld envelope at the output is given by Eq. (1.113) as before

so (t) = F 1 [e
so ()] = F 1 [Hf (, z)e
si ()]
Dispersion slope, S and 3 are related using Eqs. (1.208) (1.202),
( 2 )2
d2

3 D
d (D2 /2c) d
3 =
+ 2 2
=
=S
d
d
d
2c
2 c

(1.215)

(1.216)

Fig. 1.36(a) shows the eld envelope at the output of a ber in the presence of 3 when

Figure 1.36. Impact of third order dispersion on an ultra-short Gaussian pulse. FWHM=
1.56 ps, L= 4.7 km.

1 = 2 = 0. The eect of 3 is to cause a pulse broadening asymmetrically. When

2 = 0, pulses broaden symmetrically due to 2 as well as asymmetrically due to 3
(Fig. 1.36(b)). If the sign of 3 is changed, the pulse distortion occurs at the other edge
whereas if the sign of 2 is changed, the amount of pulse broadening is not aected for
an unchirped pulse (Eq. (1.162)).
63

Problem 2.9
For a ber-optic system, the channel wavelengths are in the range of 1530-1560 nm. Design
a single-mode ber such that the absolute accumulated dispersion should not exceed 1100
ps/nm over the wavelength range of interest. Assume that the dispersion varies linearly
with wavelength. Fiber length= 80 km.

Solution:
The dependence of dispersion on wavelength is given by
D() = ( 0 )S.

(1.217)

If we choose the zero-dispersion wavelength 0 in the center of the band, maximum allowable dispersion slope could be large. Therefore, we choose 0 = 1545 nm. Using Eq.
(1.217), the absolute dispersion at the right edge of the band is
|D(1560 nm)| = |S|(1560 1545) ps/nm/km,
or
|S| =

|D(1560 nm)|
ps/nm2 /km.
15

(1.218)

(1.219)

Since
|D(1560 nm)L| 1100 ps/nm,

(1.220)

With L = 80 km, and using Eqs. (1.219) and (1.220), we obtain

|S| 0.917 ps/nm2 /km,

1.7.5

(1.221)

Polarization Mode Dispersion

Weakly guiding approximation implies that the electromagnetic eld propagates almost
in z-direction and therefore, electric and magnetic eld components are nearly transverse
(See section 2.4). The scalar eld of section 2.4 could either represent the electric
64

eld intensity Ex or Ey . Therefore, single-mode ber supports two independent propagation modes [30]- [32]. For one of the modes, the electric eld component is along x
direction, the magnetic led component is along y direction and the propagation constant is x (x-polarization). For the other mode, the electric eld component is along
y-direction, magnetic eld component is along x-direction and the propagation constant
is y (y-polarization). If the ber cross-section is perfectly circular, these two modes are
degenerate, i.e. x = y . However, it is hard to fabricate a ber whose cross-section is
perfectly circular. Because of the asymmetry introduced during the ber manufacturing
process and external factors such as bending or twisting, the propagation constants x
and y dier. The inverse group speeds 1x and 1y corresponding to x- and y- polarization components are also dierent. As a result, x- and y- polarization components of the
input signal arrive at the ber output at dierent times leading to pulse broadening if
direct detection receiver is used. This phenomenon is known as polarization mode dispersion (PMD) [33]- [37]. Due to random uctuations in the ber refractive index along the
ber axis, there is exchange of power between these polarization components that occurs
randomly along the ber. Therefore, the pulse broadening due to PMD is stochastic in
nature.
Using Eq. (1.110), x- and y- components of electric eld intensity can be written as
Ex (x, y, z, t) = sx (t, z)(x, y) exp [i (t x z)] ,
Ey (x, y, z, t) = sy (t, z)(x, y) exp [i (t y z)] ,

(1.222)

where sx and sy are electrical eld envelopes, x and y are propagation constants for xand y- polarization components, respectively. The transverse eld distributions are nearly
the same for x- and y- polarization. As in section 2.5, input and output electrical eld
envelopes are related by
sex,out () = sex,in ()Hx (, L),
sey,out () = sey,in ()Hy (, L),

65

(1.223)

where

)]
(
2 2 z
, a = x, y.
(1.224)
Ha (, L) = exp z + i 1a z +
2
Eqs. (1.223) and (1.223) can be written in a matrix form using Jones vector notation (See
section ??)

esout () =

sex,out ()
sey,out ()

e
sin () =

sex,in ()
sey,in ()

esout () = H(, L)esin (),

Hx (, L)
0
,
H(, L) =
0
Hy (, L)

(1.225)
(1.226)
(1.227)

In the case of multi-mode bers, modes propagate at dierent speeds and arrive at dierent
times. Similarly, in a single-mode ber, x- (y-) polarization component propagates at the
speed of 1/1x (1/1y ) and therefore, the the time delay T between two polarization
components at the output of the ber of length L is
T = L |1x 1y | .

(1.228)

The above equation is valid when there is no coupling between x- and y- polarization
components. However, for standard telecommunication bers, there is a random coupling
between these components due to the perturbations such as stress and micro bending.
The ber vector transfer function given by Eq. (1.227) does not take into account the
random coupling between x- and y- polarization components. In general, ber vector
transfer function can be written as [38]

Hxx (, L) Hxy (, L)
.
H(, L) =
Hyx (, L) Hyy (, L)

(1.229)

The transfer functions Hxy (, L) and Hyx (, L) represent the random coupling between xand y- polarization components. Because of the random nature of the coupling, it is hard
to characterize these functions. Nevertheless, these functions change over a time scale that
is longer than the symbol period and therefore, it is possible to estimate H(, L) and compensate for it using digital signal processing (See chapter 11) in coherent communication
systems.
66

1.7.6

Spot Size

The transverse extent of the eld distribution of the fundamental mode plays an important role in determining splice loss between bers, bending loss, ber dispersion and the
threshold power required to have signicant nonlinear eects (discussed in Chapter 10).
The root mean square (r.m.s.) spot size or petermann-1 spot size is dened as [39]- [40]
wp1

[ 2
]1/2
2 0 (r)r3 dr
= 2
(r)rdr
0

(1.230)

where (r) is the transverse eld distribution of the fundamental mode which is radially
symmetric. When a ber mode has a large transverse extent, the spot size is large leading
to enhancement of bending losses. On the other hand, large spot size is desirable to reduce
the eect of ber nonlinearity on optical pulses and thereby, transmission performance
can be improved. This is because for the given launch power, the power per unit crosssectional area (= optical intensity) is larger when the spot size is smaller and the nonlinear
change in refractive index is directly proportional to the optical intensity. Typically, as
the spot size increases, the dispersion slope increases, too. Therefore, refractive index
prole n(r) of a ber should be optimized so that (i) it has a single mode and has low
loss at the desired wavelength range and (ii) the spot size should be suciently large so
that the transmission performance is not impaired by nonlinear eects and yet, it should
be quite small so that the dispersion slope and bending losses are not enhanced.

1.8

Dispersion Compensating Fibers

For long haul and/or high bit rate optical communication system, the pulse broadening
due to intra-modal dispersion leads to inter-symbol interference which degrades transmission performance. The pulse broadening can be compensated using dispersion compensating ber (DCF), as shown in Fig. 1.37. Using Eq. (1.107), the transfer functions of

67

Figure 1.37. Fiber-optic system consisting of TF and DCF. TF= transmission ber, DCF=
dispersion compensating ber, TX= transmitter.

the transmission ber and DCF after ignoring 1 and ber loss can be written as
[
]
i(2f )2 2TF LTF
,
HTF (f ) = exp
2
[
]
i(2f )2 2DCF LDCF
HDCF (f ) = exp
,
(1.231)
2
where the superscripts TF and DCF stand for transmission ber and dispersion compensating ber, respectively. Total transfer function is
Htot (f ) = HTF (f )HDCF (f ).

(1.232)

To keep the output pulse width equal to input pulse width, we require
Htot (f ) = 1,

(1.233)

2TF LTF = 2DCF LDCF .

(1.234)

or

When the loss is not ignored, the output of the DCF is attenuated by a factor exp [TF LTF
DCF LDCF ] without aecting the pulse broadening. When 1 is included, the pulse is
simply delayed by the factor 1TF LTF + 1DCF LDCF . Typically, the standard single-mode
bers (SSMF) or non-zero dispersion shifted bers (NZDSF) are used as the transmission
bers which have anomalous dispersion. Therefore, DCF should have normal dispersion.
As mentioned before, ber dispersion coecient can be altered by changing the amount
of waveguide dispersion. The material dispersion of glass at 1550 nm is anomalous and
the waveguide dispersion of the standard SMF is a small fraction of the total dispersion
68

(at 1550 nm). If the sign of the waveguide dispersion is made to be opposite of that of
the material dispersion by the proper choice of refractive index prole, total ber dispersion becomes normal. This is the underlying principle behind the design of DCF.
To design a DCF, the refractive index prole of a ber is divided into several segments.
Each segment is characterized by two or more parameters. For example, a segment could
have a parabolic index prole or it could have a triangular index prole. By optimizing
the parameters of these segments, the desired dispersion coecient 2 can be obtained.
For WDM applications (See Chapter 9), it is desirable to compensate dispersion over a
wide band. With proper design, dispersion slope of the transmission ber can also be
compensated [41]. Such a ber is called dispersion slope compensating ber.

Problem 2.10
A transmission ber of length 80 km has a dispersion of 21 ps2 /km. The transmission
ber is followed by a dispersion compensating ber (DCF) of dispersion 130 ps2 /km. (a)
Find the length of the DCF such that the pulse width at the input of the transmission
ber is the same as that at the output of the DCF. (b) Suppose the power launched into
the transmission ber is 2 mW, losses of the transmission ber and DCF are 0.2 dB/km
and 0.5 dB/km, respectively. Calculate the power at the output of DCF. Assume a splice
loss of 0.5 dB between transmission ber and DCF. (c) Find the gain of the amplier such
that the signal power at the output of the amplier is same as that at the input.

Solution:
(a) From Eq. (1.234), we have
2TF LTF = 2DCF LDCF ,

(1.235)

2TF = 21 ps2 /km LTF = 80 km,

(1.236)

2DCF = 130 ps2 /km,

(1.237)

69

LDCF =

2TF LTF
= 12.9 km.
2DCF

(1.238)

(b) The launch power in dBm unit is given by Eq. (1.124) as

Pin (dBm) = 10 log10

2 mW
= 3 dBm.
1 mW

(1.239)

Loss budget:
Loss in transmission ber = 0.2 80 dB = 16 dB,
Loss in DCF = 0.5 12.9 dB = 6.45 dB,
Splice loss = 0.5 dB,
Total loss = 16 + 6.45 + 0.5 dB = 22.95 dB.
The power at the output of DCF is
Pout,DCF = 3dBm 22.95dB = 19.95dBm.

(1.240)

(c) To keep the signal power at the output of the amplier same as the input, the
amplier gain should be equal to the total loss in the system, i.e.,
Amplier gain = 22.95 dBm.

(1.241)

Additional Problems
Problem 2.11
The numerical aperture of a multi-mode ber is 0.2. Find the delay between the shortest
and longest path. Fiber length= 2 km and core index= 1.45. Assume that the dierence
between the core index and cladding index is small.

70

Solution:
The NA is given by Eq. (1.9) as

NA = n1 2,
(NA)2
=
.
2n21

(1.242)
(1.243)

NA = 0.2, n1 = 1.45,
From Eq. (1.243), we nd
= 0.0095.
When the dierence between the core index and cladding index is small, n1 n2 and Eq.
(1.17) can be approximated as
T =

n1 L
,
c

(1.244)

n1 = 1.45, L = 2 km, = 0.0095, and c = 3 108 m/s,

Substituting these values in Eq. (1.244), the delay between the shortest and longest path
is
T = 91.95 ns.

Problem 2.12
The propagation constant at the wavelength 0 = 1550 nm is 6 106 rad/m. Calculate
the propagation constant at 1 = 1551 nm. Assume 1 = 0.5 108 s/m and 2 = 10
ps2 /km. Ignore n , n > 2.

Solution:
From Eq. (1.102), we nd
(1 ) = 0 + 1 (1 0 ) + 2 (1 0 )2 /2.

71

(1.245)

Using c = f , we have
0 = 2f0 =

2c
= 1.2161 1015 rad/s,
0

1 = 2f1 =

2c
= 1.2153 1015 rad/s,
1

1 0 = 8.168 1011 rad/s,

0 = 6 1016 rad/m,
Substituting these values in Eq. (1.245), we nd
(1 ) = 5.9959 106 rad/s.
Note that the change in propagation constant is very small.

Problem 2.13
Consider a ber-optic system shown in Fig. 1.38. Fiber loss= 0.2 dB/km, length= 80 km,
loss in optical lter= 0.5 dB and amplier gain= 15 dB. If the minimum power required
at the receiver to have a good signal-to-noise ratio is 3 dBm, calculate the lower limit
on the transmitter power in dBm and mW units.

Figure 1.38. A ber-optic system with loss and gain.

72

Solution:

Fiber loss, F1 (dB) = 0.2 80 = 16 dB,

Filter loss, F2 (dB) = 0.5 dB,
Amplier gain, G(dB) = 15 dB.
The minimum power required at the receiver is
Pout (dBm) = 3 dBm,
Pout (dBm) = Pin (dBm) + F1 (dB) + F2 (dB) + G(dB).
Therefore, the lower limit on the transmitter power is
Pin (dBm) = 3 + 16 + 0.5 15dBm = 1.5 dBm.
Using Eq. (1.125), transmitter power in mW is
Pin = 100.1Pin (dBm) = 0.7079 mW.

Problem 2.14
Electric eld envelope at the ber input is
sin (t) = A cos(2fm t).

(1.246)

Show that the electric eld envelope at the ber output is

[
]
sout (t) = A cos(2fm t) exp i(2fm )2 2 L/2 .
Ignore ber loss and 1 .

73

(1.247)

Solution:
Taking the Fourier transform of Eq. (1.246), we nd
sein (f ) = F [sin (t)] =

A
[(f fm ) + (f + fm )] ,
2

(1.248)

where is the Dirac delta function. From Eq. (1.107), we have the ber transfer function
(after ignoreing 1 and loss)
[
]
e ) = exp i2 (2f )2 L/2 .
H(f

(1.249)

The output spectrum is

seout (f ) = sein (f )H(f )
[
]
A
=
[(f fm ) + (f + fm )] exp i2 (2f )2 L/2 .
2
Taking the inverse Fourier transform of Eq. (1.250), we obtain
[
]

A
2 2 L
so (t) =
[(f fm ) + (f + fm )] exp i2f t + i(2f )
df.
2
2

(1.250)

(1.251)

Using the following relation

(f fm )X(f )df = X(fm ),

(1.252)

Eq. (1.251) is simplied as

[
]
A
[exp(i2fm t) + exp(i2fm t)] exp i(2fm )2 2 L/2
2
[
]
= A cos(2fm t) exp i(2fm )2 2 L/2 .

so (t) =

(1.253)

Comparing Eqs. (1.246) and (1.253), we nd that if the eld envelope is a sinusoid, it
acquires only a phase shift.

Problem 2.15
Consider a ber-optic system shown in Fig. 1.39. A Gaussian pulse is launched into
the transmission ber. Find the length of DCF so that the pulse width (FWHM) at the
74

Figure 1.39. Fiber optic system consisting of TF and DCF. TF= transmission ber, DCF=
dispersion compensating ber, Tx= transmitter and Rx= receiver.

output of DCF is twice the pulse width at the input of TF. Assume 2TF = 21 ps2 /km,
2DCF = 130 ps2 /km, LTF = 80 km. FWHM at the input of TF= 12.5 ps. Ignore loss and
1 .

Solution:
The eective transfer function is given by Eq. (1.232) as
{ [
]}
2
( TF TF
)
DCF DCF (2f )
He (f ) = exp i 2 L + 2 L
2
[ e e
]
2
= exp i2 L (2f ) /2 ,

(1.254)

where
2e Le = 2TF LTF + 2DCF LDCF .
(1.255)

In Problem 2.5, we have found that when |2 L| = 3T02 , the output pulse width is twice
the input pulse width. Replacing 2 L by 2e Le , we nd
 TF TF

2 L + 2DCF LDCF  = 3T02 .
2TF = 21 ps2 /km,
2DCF = 130 ps2 /km,

LTF = 80 km,
TFWHM = 12.5 ps,

T0 = TFWHM /1.665 = 7.507 ps.

Eq. (1.256) may be written as

21 80 + 130LDCF = 3 7.5072
75

(1.256)

Therefore, LDCF = 12.17km or 13.67 km. As the pulse propagates in the DCF, the pulse
undergoes compression. At LDCF = 12.17 km, the pulse width is twice the initial pulse
width. If 2TF LTF = 2DCF LDCF , the pulse width becomes equal to the initial pulse
width. This corresponds to a propagation distance of 12.92 km in DCF. After this, the
pulse broadening takes place and when LDCF = 13.67 km, the output pulse width is the
twice the initial pulse width, again.

Problem 2.16
Zero dispersion wavelength of a transmission ber (TF) is chosen as 1490 nm so that the
local dispersion in the desired wavelength range 1530 nm-1565nm is not zero (so as to avoid
the enhancement of nonlinear eects). Find the accumulated dispersion of the dispersion
compensating ber (DCF) so that the net accumulated dispersion does not exceed 1100
ps/nm. Assume that the dispersion slope of the TF and DCF are 0.08 ps/nm2 /km and 0
ps/nm2 /km, respectively. Other parameters are same as that in Problem 2.9.

Solution:
In the absence of DCF, the dispersion at 1560 nm is
DTF (1560 nm) = 0.08(1560 1490) ps/nm/km = 5.6 ps/nm/km.

(1.257)

The accumulated dispersion at 1560 nm is

DTF LTF = 448 ps/nm.

(1.258)

The net accumulated dispersion is

|DTF LTF + DDCF LDCF | 1100 ps/nm.

(1.259)

Therefore, the accumulated dispersion of DCF should be

DDCF LDCF > 652 ps/nm.
76

(1.260)

Problem 2.17
Let the input eld envelope be
]
t2 (1 + iC)
si (t) = A exp
,
2T02
[

(1.261)

where C is the chirp parameter. Show that the eld envelope at the ber output is
[
]
AT0
(1 + iC) t2
exp
,
so (t) =
T1
2T12
where
T1 =

[( 2
)
]1/2
T0 + 2 LC i2 L
.

Plot the power and frequency deviation at dierent ber lengths for 2 C > 0 and
2 C < 0. Ignore ber loss and 1 .

Solution:
Using Eqs. (1.150) and (1.152), we nd
sei (f ) =

[
]
A
exp (f /a)2 ,
a

(1.262)

where
a2 =

1 + iC
.
2T02

(1.263)

The output eld envelope is given by Eq. (1.155) with b2 = a2 /(1 i22 La2 ) as
[
]
AT0
(1 + iC) t2
so (t) =
exp
,
(1.264)
T1
2T12
where
T1 =

)
]1/2
[( 2
.
T0 + 2 LC i2 L

(1.265)

As in Problem 2.5, the output pulse width can be calculated as

out
TFWHM

[(
)2 (
)2 ]1/2
C
L

L
2
2
in
= TFWHM
1+
+
.
2
T0
T02
77

(1.266)

The eld envelope may be written as

s(t, z) = A(t, z) exp[i(t, z)]

(1.267)

The instantaneous frequency deviation from the carrier frequency is given by Eq. (1.165
as
d
dt

(1.268)

t2 C
.
2T02

(1.269)

(t, z) =
At the ber input, we have
(t, 0) =

So, the instantaneous frequency deviation from the carrier frequency is

(t, 0) =

Ct
T02

(1.270)

Figure 1.40. Output pulse width of a chirped Gaussian pulse. 2 = 21 ps2 /km.
Fig. 1.40 shows the output pulse width as a function of propagation distance L. As
can be seen from Eq. (1.266), the pulse broadening depends on the sign of 2 C. When
2 C 0, pulse width increases with distance monotonically. When 2 C < 0, the rst
term within the square bracket of Eq. (1.266) becomes less than unity and therefore, the
output pulse width can be less than the input pulse width for certain distances. Fig. 1.40
shows that the pulse undergoes compression initially for C = 4 and 2 < 0. The physical
78

Figure 1.41. Power and frequency deviations of a chirped Gaussian pulse. 2 = 21 ps2 /km,
C = +3.

explanation for pulse compression is as follows. When C > 0, from Eq. (1.270), we see
that the leading edge is downshifted in frequency and the trailing edge is upshifted at the
ber input. In anomalous dispersion ber (2 < 0), low frequency (red) components travel
slower than high frequency (blue) components and therefore, the frequency components
at the leading edge travel slowly. In other words, they are delayed and move to the right
(later time) as shown by the arrow in Fig. 1.41(a)) and the frequency components at
the leading edge move to the left (earlier time) leading to pulse compression as shown in
Fig. 1.41(b). Since the frequency chirp imposed on the pulse at the input is of opposite
sign from the frequency chirp developed due to pulse propagation in anomalous dispersion
ber,these two frequency chirps cancel at L = 12.5 km and the pulse becomes unchirped
(see the bottom of Fig. 1.41(b)). At this distance, the pulse width is the shortest.
Thereafter, pulse propagation is the same as that discussed in Problem 2.5 leading to
pulse broadening.

79

Figure 1.42. Power and frequency deviations of a chirped Gaussian pulse. 2 = 21 ps2 /km,
C= -3.

If C < 0, at the ber input, the leading edge is upshifted in frequency whereas the
trailing edge is downshifted. The frequency components corresponding to the leading
edge travel faster than that corresponding to the trailing edge in an anomalous dispersion
ber. Therefore, as shown by the arrows in Fig. 1.42a, the leading edge moves to the left
(earlier time) and the trailing edge moves to the right which leads to pulse broadening.
The frequency chirp at the ber input in this case has the same sign as that due to
dispersion as given by Eq. (1.167) and therefore, these two chirps add up leading to
enhanced broadening seen in Fig. 1.40 (C = -4) as compared to the case of unchirped
pulse.

80

Exercises
1. A step-index ber has a cuto wavelength = 900 nm, and NA = 0.22 (a) Calculate
core radius (b) what could be the maximum allowable core radius to make this ber
single-moded at 500 nm? (Ans: (a) 3.44 microns (b) 2.29 microns)

Figure 1.43. An innitesimal ber section

2. Consider a small ber section of length L as shown in Fig. 1.43. Let F (L) =
P (L)/Pin . Next, consider a cascade of identical ber sections shown in Fig. 1.44.
Let M be the total number of ber sections. When M (or L 0), show

Figure 1.44. A ber of length L with M sections of length L

that
Ftot =

Pout
= exp (L) ,
Pin

(1.271)

where = dF/d(L).
Hint: Ftot = [F (L)]L/L , expand F (L) in a Taylor series with F (0) = 1 and use
e = LimM (1 + 1/M )M

81

(1.272)

3. A step-index multi-mode ber has to be designed to support bit rates up to 10 Mb/s.

The link length is 1.5 km. Calculate the upper limit on the relative index dierence,
. Assume core refractive index is approximately equal to cladding refractive index.
(Ans: 0.02.)
4. In a step-index multi-mode ber, the critical angle for the core-cladding interface
is 85 . The core refractive index is 1.46 and core diameter is 100 m. Find (a)
V -number, and (b) the approximate number of guided modes M at the wavelength
of 1.3 m.
(Ans: V = 25.79, M = 332.)
5. The maximum achievable bit rate distance product in a step-index multi-mode ber
is 16 Mb/skm. The core refractive index n1 ( cladding index n2 )= 1.45. Calculate
(a) numerical aperture (b) critical angle for the core-cladding interface.
(Ans: (a) 0.2795 (b) 1.3768 rad.)
6. A step-index multi-mode ber has an acceptable angle of 0.2077 rad and the critical
angle for the core-cladding interface is 1.4266 rad. Calculate the speed of light in
the ber core. It may be assumed that ray optics theory is valid.
(Ans: 2.076 108 m/s.)
7. A single-mode ber has to be designed to operate at 1.55 m with a cut-o wavelength less than 1.5 m. The core and cladding refractive indices are 1.45 and 1.445,
respectively. Calculate the maximum allowable core radius.
(Ans: 4.77 m.)
8. The power launched into a ber is 5 mW and the power at the ber output is 0.3
mW. Calculate the ber loss in dBm units.
(Ans: 12.22 dB.)
9. A 40 km single-mode ber has a dispersion parameter D = 10 ps/nmkm. An
optical signal of bandwidth 10 GHz is launched into ber. Find the delay between
82

highest and lowest frequency components. The carrier wavelength= 1.55 m.

(Ans: 32.033 ps.)
10. A single-mode ber has a zero-dispersion wavelength at 1550 nm. The dispersion
slope= 0.06 ps/nm2 /km. Find the absolute dispersion |D at 1600 nm. Assume that
the dispersion varies linearly with wavelength.
(Ans: 3 ps/nm/km.)
11. A Gaussian pulse is transmitted in a long single-mode ber with dispersion coecient 2 = 10 ps2 /km. The input and output pulse widths (FWHM) are 8 ps and
32 ps, respectively. Calculate the ber length.
(Ans: 8.9 km.)
12. The outputs of cw lasers of frequencies 193.5 THz and 193.6 THz are transmitted
over a ber of length 2 km. It is found that the lower frequency optical wave
arrived later than the higher frequency component by 6 ps. Calculate the dispersion
coecient 2 .
(Ans: 4.77 ps2 /km.)
13. A ber-optic link consists of a transmission ber (TF) followed by a dispersion
compensating ber (DCF) of length 5 km. The transmission ber has a dispersion
parameter D = 10 ps/nm/km, loss= 0.25 dB/km, and length= 50 km. (a) Find
the dispersion parameter of DCF such that the pulse width (FWHM) at the output
of the DCF is the same as the pulse width at the input of the TF (b) The power
launched into the TF is 2 mW and the power at the output of DCF is -12 dBm.
Find the loss coecient of DCF in dB/km.
(Ans: (a) -100 ps/nm/km (b) 0.5 dB/km

83

Further Reading
K. Okamoto, Fundamentals of optical waveguide, Academic, New York, Second Edition,
2006.
T. Okoshi, Optical Fibers, Academic Press, San Diego, CA, 1982.
L.B. Jeunhomme, Single-mode Fiber Optics, Marcel Dekker, New York, 2nd edition, 1990.
E.G. Neumann, Single-mode Fibers, Springer-Verlag, New York, 1988.
A.W. Snyder and J. D. Love, Optical Waveguide Theory Chapman Hall, London, 1983.
J. M. Senior, Optical Fiber Communications, Second edition, Prentice Hall Europe, 1992.
H. Kolimbris, Fiber Optic Communications, Prentice Hall, New Jersey, 2004.
R. P. Khare, Fiber Optics and Optoelectronics, Oxford University Press, New York, 2004.

84

References
[1] C. K. Kao and G. A. Hockham, Proc. IEE, Vol. 113, p. 1151, 1966.
[2] R.D. Maurer, D.B. Keck and P. Schultz, Optical waveguide bers, US patent #
3, 711, 262, 1973.
[3] T. Miya, Y. Terunuma, T. Husaka and T. Miyoshita, Electron. Lett., Vol. 15, p. 106,
1979.
[4] J. Gower, Optical Communication systems, 2nd ed., Prentice Hall, London, 1993.
[5] G. P. Agrawal, Fiber-Optic Communication Systems, Fourth edition, John Wiley and
Sons, Inc, New York, 2010.
[6] D. Marcuse, Theory of dielectric optical waveguides, Academic, New York, Second
Edition, 1991.
[7] J.A. Buck, Fundamentals of optical bers, Wiley, Hoboken, New Jersey, Second Edition, 2004.
[8] A. Ghatak and K. Thyagarajan, Optical Electronics, Cambridge University Press,
Cambridge, 1991.
[9] D. Gloge, Weakly guiding bers, Appl. Opt., Vol. 10, pp. 2252, 1971.
[10] D. Marcuse, D. Gloge and E.A.J. Marcatili, Guiding properties of bers in S.E.
Miller and A.G. Chynoweth eds., Optical Fiber Telecommunications Academic, New
York, 1979.
85

[11] A. W. Snyder, IEEE Trans. Microwave Theory Tech., vol. MTT-17, p. 1130, 1969.
[12] G. Keiser, Optical Fiber Communications, McGraw Hill, New York, Fourth Edition,2011.
[13] R.M. Gagliardi and S. Karp, Optical Communications, Wiley, New York, Second
Edition, 1995.
[14] Communication News, Mar. 2006, Nelson Publishing Inc, www.comnews.com.
[15] R. Ryf et al, J. Lightwave Technol., vol. 30, p. 521, 2012
[16] X. Chen et al, Opt. Exp., vol. 20, p. 14302, 2012
[17] B. Neng et al, Opt. Exp., vol. 20, p. 2668, 2012
[18] N. Hanzawa et al, Optical Fiber Conference, paper OWA4, 2012
[19] H.S. Chen, H.P.A. van den Boom and A.M.J. Koonen, IEEE Photon. Tech. Lett.,
vol. 23, 2011
[20] R.P.Feynmann, R.B. Leighton and M. Sands, The Feyman Lectures on physicsvolume 1, Chapter 32, Addison-Wesley Publishing Company, 1963.
[21] H. Kanamori et al J. Lightwave Technol., vol. LT-4, p. 1144, 1986.
[22] Y. Chigusa et al J. Lightwave Technol., vol. 23, p. 3541, 2005.
[23] Osanai et al Electron. Lett., vol. 12, p. 549, 1976.
[24] Berkey et al , US patent EP1181254 B1, 2006.
[25] F.W. Dabby, US patent 20130025326 A1, 2013.
[26] Ohashi et al, US patent no. 4,755,022, 1988.
[27] Kato it etal, US patent no. 5,721,800, 1998.
86

[28] Y. Liu and M. A. Newhouse, US patent EP1158323 A1, 1996.

[29] Corining product sheet information - www.corning.com/opticalber.
[30] I.P. Kaminov, IEEE J. Quantum Electron., vol. 17, p. 15, 1981.
[31] S.C. Rashleigh, J. Lightwave Technol., vol.LT-1, p. 312, 1983.
[32] X.-H. Zheng, W.M. Henry and A.W. Snyder, J. Lightwave Technol., vol. 6, p. 312,
1988.
[33] C.D. Poole, J. H. Winters and J. A. Nagel, Opt. Lett.,vol. 16, p. 372, 1991.
[34] G.J. Foschini and C.D. Poole, J. Lightwave Technol., vol. 9, p. 1439, 1991.
[35] J. Zhou and M.J.O Mahony, IEEE Photon. Tech. Lett., vol. 6, p. 1265, 1994.
[36] P.K.A. Wai and C.R. Menyuk, J. Lightwave Technol., vol. 14, p. 148, 1996.
[37] A. Galtarossa and C.R. Menyuk (Eds.),Polarization Mode Dispersion Springer, New
York, 2005.
[38] C.D. Poole and R.E. Wagner, Electron. Lett. , vol. 22, p. 1029, 1986.
[39] K. Petermann, Electron. Lett., vol. 19, p. 712, 1983.
[40] C. Pask, Electron. Lett., vol. 20, p. 144, 1984.
[41] S. Bickham et al, US Patent no. 06671445, 2003.

87