1/269

Fiber-Optic Communication Systems

Govind P. Agrawal
Institute of Optics
University of Rochester
Rochester, NY 14627
email: gpa@optics.rochester.edu

JJ
II
J
I
c
2005
G. P. Agrawal

Back
Close

Course Outline
Introduction

2/269

Fiber Loss, Dispersion, and Nonlinearities

Receiver Noise and Bit Error Rate
System Design and Performance
Loss Management: Optical Amplifiers
Dispersion Management
Multichannel Lightwave Systems
Optical Solitons

JJ
II
J
I
Back
Close

Historical Perspective
Smoke signals;

<1500

Semaphore Devices;

3/269

1500-1800

Mechanical Coding (Chappe);

1792

JJ
II
J
I
Back
Close

Historical Perspective
Electrical Era

4/269

Telegraph;

1836

Telephone;

1876

Coaxial Cables;

1840

Microwaves;

1948

Optical Era
Optical Fibers;

1978

Optical Amplifiers;

1990

WDM Technology;

1996

Multiple bands;

2002

JJ
II
J
I
Back
Close

Communication Technologies
5/269

Performance criterion: B L

JJ
II
J
I
Back
Close

Information Revolution
Industrial revolution of 19th century gave way to
information revolution during the 1990s.

6/269

Internet made the rapid distribution of information possible.

Fiber-Optic Revolution is a natural consequence of the
Internet growth.

JJ
II
J
I
Back
Close

Five Generations
0.8-m systems (1980); Graded-index fibers
7/269

1.3-m systems (1985); Single-mode fibers

1.55-m systems (1990); Single-mode lasers
WDM systems (1996); Optical amplifiers
L and S bands (2001); Raman amplification
10000

Bit Rate (Gb/s)

1000

Research

100
10

Commercial

1
0.1
0.01
1980

1985

1990

1995
Year

2000

2005

JJ
II
J
I
Back
Close

Basic Concepts
Analog and Digital Signals
8/269

Lightwave Systems use digital format.

Optical signal is a stream of 0 and 1 bits.
Bit rate B determines the time slot TB = 1/B for each bit.

JJ
II
J
I
Back
Close

Analog to Digital Conversion

9/269

Sampling: fs 2f (sampling theorem).

Quantization: M > Amax/AN (error < noise).
Coding: M = 2m;

m bits/sample (Binary coding).

Bit rate: B = mfs (2f ) log2 M > (f /3)SNR.

JJ
II
J
I
Back
Close

Audio and Video Signals

Digital Audio Signal
10/269

f = 3.1 kHz (0.3 to 3.4 kHz); SNR = 30 dB.

Minimum B = (f /3)SNR = 31 kb/s.
In practice, B = 64 kb/s (fs = 8 kHz; 8 bits/sample).
Digital Video Signal
f = 4 MHz; SNR = 50 dB.
Minimum B = (f /3)SNR = 66 Mb/s.
In practice, B = 100 Mb/s (fs = 10 MHz; 10 bits/sample).

JJ
II
J
I
Back
Close

Channel Multiplexing
TDM: Time-division multiplexing
11/269

FDM: Frequency-division multiplexing

Optical FDM = WDM (Wavelength-Division Multiplexing)

JJ
II
J
I
Back
Close

Evolution of Standards
No standards until 1988.
12/269

US standard: synchronous optical network (SONET).

ITU standard: synchronous digital hierarchy (SDH).

SONET
OC-1
OC-3
OC-12
OC-48
OC-192
OC-768

SDH

B (Mb/s) Channels
51.84
672
STM-1
155.52
2,016
STM-4
622.08
8,064
STM-16
2,488.32
32,256
STM-64
9,953.28 129,024
STM-256 39,813.12 516,096

JJ
II
J
I
Back
Close

Modulation Formats
Optical Carrier
E(t) = eA cos(0t + )

13/269

Amplitude-shift keying (ASK): modulate A

Frequency-shift keying (FSK): modulate 0
Phase-shift keying (PSK): modulate
Polarization-shift keying (PoSK): modulate e
? Most lightwave systems employ ASK
? also called onof keying (OOK) or
? Intensity modulation with Direct Detection

JJ
II
J
I
Back
Close

Optical Bit Stream

Return-to-zero (RZ)
14/269

nonreturn-to-zero (NRZ)

JJ
II
J
I
Back
Close

Lightwave System Components

Generic System
15/269

Transmitter Design

Receiver Design
JJ
II
J
I
Back
Close

Optical Fibers
Most suitable as communication channel because of dielectric
waveguiding (acts like an optical wire).

16/269

Total internal reflection at the core-cladding interface.

Single-mode propagation for core size < 10 m.
What happens to Signal?
Fiber losses limit the transmission distance (minimum loss near
1.55 m).
Chromatic dispersion limits the bit rate through pulse broadening.
Nonlinear effects distort the signal and limit the system
performance.

JJ
II
J
I
Back
Close

Fiber Fundamentals
17/269

Step-index fibers: Refractive index constant inside the core.

Graded-index fibers: Refractive index varies inside the core.

JJ
II
J
I
Back
Close

Total internal reflection

Refraction at the airglass interface:

n0 sin i = n1 sin r

18/269

Total internal reflection at the core-cladding interface

if > c = sin1(n2/n1).

Numerical Aperture: Maximum angle of incidence

n0 sin imax

= n1 sin(/2 c) = n1 cos c =

n21 n22

JJ
II
J
I
Back
Close

Modal Dispersion
Multimode fibers suffer from modal dispersion.
19/269

Shortest path length Lmin = L (along the fiber axis).

Longest path length for the ray close to the critical angle
Lmax = L/ sin c = L(n1/n2).
Pulse broadening: T = (Lmax Lmin)(n1/c).
Modal dispersion: T /L = n21/(n2c).
Limitation on the bit rate
T < TB = 1/B;

BT < 1;

n2 c
BL < 2 .
n1

Single-mode fibers essential for high performance.

JJ
II
J
I
Back
Close

Graded-Index Fibers
20/269


Refractive index n() =

n1[1 (/a) ];
n1(1 ) = n2 ;

Ray path obtained by solving

d2
dz 2

< a,
a.

1 dn
n d .

For = 2, = 0 cos(pz) + (00/p) sin(pz).

All rays arrive simultaneously at periodic intervals.
Limitation on the Bit Rate: BL <

8c
.
n1 2

JJ
II
J
I
Back
Close

Fiber Modes
Shape-preserving solutions of Maxwells equations
21/269

E = B/t
H = D/t
D=0
B=0
Constitutive Relations
D = 0 E + P
B = 0 H + M
Linear Susceptibility
Z

P(r, t) = 0

(r, t t0)E(r, t0) dt0

JJ
II
J
I
Back
Close

Fiber Modes (cont.)

It is easier to work in the Fourier domain:
Z
) =
E(r, t) exp(it) dt
E(r,

22/269

Helmholtz Equation: Each frequency component satisfies

+ n2()k 2E
= 0.
2 E
0
k0 = /c = 2/,

n() = (1 + Re )
1/2.

n = n1 inside the core but changes to n2 in the cladding.

Frequency dependence of n responsible for dispersion.
Material absorption governed by Im .

JJ
II
J
I
Back
Close

Fiber Modes (cont.)

Any solution must satisfy boundary conditions at the
corecladding interface (x2 + y 2 = a2).

23/269

Useful to work with cylindrical coordinates , , z.

Only two components of E and H are independent.
Common to choose Ez and Hz as independent components.
Equation for Ez in cylindrical coordinates:
2Ez 1 Ez
1 2Ez 2Ez
+
+ 2
+
+ n2k02Ez = 0.
2
2
2

JJ
II
J
I
Back
Close

Fiber Modes (cont.)

Use the method of separation of variables:
24/269

Ez (, , z) = F ()()Z(z).
We then obtain three ODEs:
d2Z/dz 2 + 2Z = 0,
d2/d2 + m2 = 0,


2
m
d2F 1 dF
+
+ n2k02 2 2 F = 0.
2
d
d

and m are two constants (m must be an integer).

First two equations can be solved easily to obtain
Z = exp(iz),
F () satisfies the Bessel equation.

= exp(im).

JJ
II
J
I
Back
Close

Fiber Modes (cont.)

General solution for Ez and Hz :

AJm(p) exp(im) exp(iz) ;
Ez =
CKm(q) exp(im) exp(iz);

BJm(p) exp(im) exp(iz) ;
Hz =
DKm(q) exp(im) exp(iz);
where p2 = n21k02 2,

25/269

a,
> a.
a,
> a.

q 2 = 2 n22k02.

Other components of EandH can be written in terms of Ez and

Hz using Maxwells equations.
Boundary conditions: Ez , Hz , E, and H should be continuous
across the corecladding inteface ( = a).

JJ
II
J
I
Back
Close

Fiber Modes (cont.)

Eigenvalue Equation
 0
 0

0
0
Km
(qa)
n22 Km
Jm(pa)
Jm(pa)
(qa)
+
+
pJm(pa) qKm(qa) pJm(pa) n21 qKm(qa)



2
2
1
1
m
1
n2 1
= 2
+
+
a
p2 q 2
p2 n21 q 2

26/269

where p2 = n21k02 2 and q 2 = 2 n22k02.

This equation should be solved numerically.
Multiple solutions for for a given fiber.
Each solution represents an optical mode.
p
Number of modes governed by V = k0a n21 n22.

JJ
II
J
I
Back
Close

Fiber Modes (cont.)

27/269

= /k0 between n1 and n2.

Effective mode index n
Useful to introduce a normalized quantity
b = (
n n2)/(n1 n2),

(0 < b < 1).

Modes quantified through () or b(V ).

JJ
II
J
I
Back
Close

Single-Mode Condition
Modes denoted as HEmn or EHmn, where n = 1, 2, . . . in decreasing
order of the eigenvalues mn.

28/269

Also labelled as TE0n and TM0n for m = 0.

Also called LPmn for weakly guiding fibers.
Number of modes increases rapidly with V parameter.
A mode ceases to exist when q = 0 (no decay in the cladding).
TE01 and TM01 reach cutoff when J0(V ) = 0.
Single-mode fibers require V < 2.405.
Typically core radius a < 5 m and n1 n2 < 0.005.

JJ
II
J
I
Back
Close

Single-Mode Fibers
Fibers support only the HE11 mode when the core size is such that
V = k0a(n21 n22) < 2.405.

29/269

This mode is almost linearly polarized (|Ez |2  |ET |2).

Spatial mode distribution approximately Gaussian
 2

x + y2
Ex(x, y, z, ) = A0() exp
exp[i()z].
w2

JJ
II
J
I
Back
Close

Single-Mode Properties
Spot size: w/a 0.65 + 1.619V 3/2 + 2.879V 6.
30/269

Mode index:
= n2 + b(n1 n2) n2(1 + b),
n
b(V ) (1.1428 0.9960/V )2.
Confinement factor:
Ra


2
2
|E
|

d
Pcore
2a
x
=
= R 0
= 1 exp 2 .
2 d
Ptotal
w
|E
|
x
0
0.8 for V = 2 but drops to 0.2 for V = 1.

JJ
II
J
I
Back
Close

Fiber Birefringence
nx 6= n
y ).
Real fibers exhibit some birefringence (

31/269

y | 10 ).
Modal birefringence quite small (Bm = |
nx n
Beat length: LB = /Bm.
State of polarization evolves periodically.

Birefringence varies randomly along fiber length (PMD) because of

stress and core-size variations.

JJ
II
J
I
Back
Close

Fiber Dispersion
Origin: Frequency dependence of the mode index n():
32/269

()/c = 0 + 1( 0) + 2( 0)2 + ,
() = n
where 0 is the carrier frequency of the optical pulse.
Transit time for a fiber of length L : T = L/vg = 1L.
Different frequency components travel at different speeds and arrive
at different times at output end (pulse broadening).
JJ
II
J
I
Back
Close

Fiber Dispersion (continued)

Pulse broadening governed by group-velocity dispersion:
33/269

T =

dT
d L
d1
=
= L
= L2,
d
d vg
d

where is pulse bandwidth and L is fiber length.

 2 
d
GVD parameter: 2 = d
.
2
=
0
 
d
Alternate definition: D = d v1g = 2c
.
2 2
Limitation on the bit rate: T < TB = 1/B, or
B(T ) = BL2 BLD < 1.

JJ
II
J
I
Back
Close

Material Dispersion
Refractive index of of any material is frequency dependent
(chromatic dispersion).

34/269

Material dispersion for silica is governed by the Sellmeier equation

M
X
Bj j2
2
n () = 1 +
2 2 .

j=1 j

JJ
II
J
I
Back
Close

Waveguide Dispersion
() = n2() + nW ().
Mode index n

35/269

Material dispersion DM results from n2() (index of silica ).

Waveguide dispersion DW results from nW () and depends on
core size and dopant distribution.
Total dispersion D = DM + DW can be controlled.

JJ
II
J
I
Back
Close

Higher-Order Dispersion
Dispersive effects do not disappear at = ZD.

36/269

D cannot be made zero at all frequencies within the pulse spectrum.

Higher-order dispersive effects are governed by
the dispersion slope S = dD/d.

S can be related to third-order dispersion 3 as

S = (2c/2)23 + (4c/3)2.
At = ZD, 2 = 0, and S is proportional to 3.

JJ
II
J
I
Back
Close

Polarization-Mode Dispersion
nx 6= n
y ).
Real fibers exhibit some birefringence (
37/269

Orthogonally polarized components of a pulse travel at different

speeds. The relative delay is given by


L

L
 = L|1x 1y | = L(1).

T = 
vgx vgy 
Birefringence varies randomly along fiber length (PMD) because of
stress and core-size variations.
RMS Pulse broadening:

2lcL Dp L.

PMD parameter Dp 0.0110 ps/ km

T (1)

PMD can degrade the system performance considerably (especially

for old fibers).

JJ
II
J
I
Back
Close

Commercial Fibers
Fiber Type and
Aeff
Trade Name
(m2)
Corning SMF-28
80
Lucent AllWave
80
Alcatel ColorLock 80
Corning Vascade
101
TrueWave-RS
50
Corning LEAF
72
TrueWave-XL
72
Alcatel TeraLight
65

ZD
D (C band)
Slope S
(nm)
ps/(km-nm) ps/(km-nm2)
13021322
16 to 19
0.090
13001322
17 to 20
0.088
13001320
16 to 19
0.090
13001310
18 to 20
0.060
14701490
2.6 to 6
0.050
14901500
2 to 6
0.060
15701580 1.4 to 4.6
0.112
14401450
5.5 to 10
0.058

38/269

JJ
II
J
I
Back
Close

Pulse Propagation Equation

Optical Field at frequency at z = 0:
39/269

) = x
) exp(iz).
F (x, y)B(0,
E(r,
Optical field at a distance z:
) = B(0,
) exp(iz).
B(z,
Expand () is a Taylor series around 0:
() = n
()

2
3
0 + 1() + ()2 + ()3.
c
2
6

Introduce Pulse envelope:

B(z, t) = A(z, t) exp[i(0z 0t)].

JJ
II
J
I
Back
Close

Pulse Propagation Equation

Pulse envelope is obtained using
1
A(z, t) =
2



i
i
2
3

d()A(0, ) exp i1 z + 2 z() + 3 z() i()t .

2
6

40/269

Calculate A/z and convert to time domain by replacing

with i(A/t).
Final equation:
A
A i2 2A 3 3A

= 0.
+ 1
+
z
t
2 t2
6 t3
With the transformation t0 = t 1z and z 0 = z, it reduces to
A i2 2A 3 3A
+

= 0.
z 0
2 t02
6 t03

JJ
II
J
I
Back
Close

Pulse Propagation Equation

If we neglect third-order dispersion, pulse evolution is governed by
41/269

A i2 2A
+
= 0.
z
2 t2
Compare with the paraxial equation governing diffraction:
A 2A
= 0.
2ik
+
z
x2
Slit-diffraction problem identical to pulse propagation problem.
The only difference is that 2 can be positive or negative.
Many results from diffraction theory can be used for pulses.
A Gaussian pulse should spread but remain Gaussian in shape.

JJ
II
J
I
Back
Close

Pulse Broadening
Chirped Gaussian Pulse
42/269



(1 + iC)t2
.
A(0, t) = A0 exp
2T02
Input pulse width TFWHM = 2(ln 2)1/2T0 1.665T0.
Input chirp: (t) =
t =

C
T02

t.

Pulse spectrum
) = A0
A(0,
Spectral width: 0 =

2T02

1/2

1 + iC

1 + C 2/T0.



2 2
T0
exp
2(1 + iC)

JJ
II
J
I
Back
Close

Pulse Broadening (cont.)

Optical Pulse after propagation (3 = 0):


A0
2z
(1 + iC)t2
A(z, t) = p
,
Q(z)
=
1
+
(C

i)
.
exp
2T02Q(z)
T02
Q(z)

43/269

Pulse maintains its Gaussian shape.

Its chirp changes to C1(z) = C + (1 + C 2)2z/T02.
Pulse width at a distance z is given by
"
2 
2#1/2
T1
C2z
2z
.
=
1+
+
T0
T02
T02
p
Unchirped pulse broadens by a factor 1 + (z/LD )2.
LD = T02/|2| is called the dispersion length.

JJ
II
J
I
Back
Close

Effect of Chirp
44/269

Broadening depends on the sign of 2C.

Pulse compresses initially if 2C < 0.
Pulse becomes shortest and unchirped at a distance


2
zmin = |C|/(1 + C ) LD ,
T1min = T0/(1 + C 2)1/2.

JJ
II
J
I
Back
Close

Third-Order Dispersion
When 2 = 0, pulse broadening depends on 3.
45/269

Analytic solution exists in terms of an Airy function.

Pulse does not maintain its Gaussian shape and develops considerable structure (ringing near an edge).
A proper measure of pulse width is the RMS width defined as

1/2
= < t2 > < t >2
,
R m
2
t |A(z, t)| dt
m
.
< t >= R
2 dt
|A(z,
t)|

Broadening factor:

2 
2

2
2

C2L
2L
3L
2 2

=
1
+
+
+
(1
+
C
)
.
3
02
202
202
4 20

JJ
II
J
I
Back
Close

Source Spectral Width

Source spectral width can exceed pulse spectral width when a lowcoherence source (LED) is used.

46/269

Introduce a parameter V = 2(s)0 for a source with Gaussian

spectrum of RMS width s.
RMS pulse width then increases as

2

2
2
C2L

L
2
= 1+
+ (1 + V2)
2
2
0
20
202

2
3L
+(1 + C 2 + V2)2 3 .
4 20
Quite useful for analyzing the GVD effects.

JJ
II
J
I
Back
Close

Dispersion Limitations
Large Source Spectral Width: V  1

47/269

Assume input pulse to be unchirped (C = 0).

Set 3 = 0 when 6= ZD:
2 = 02 + (2L )2 02 + (DL)2.
96% of pulse energy remains within the bit slot if 4 < TB = 1/B.
Using 4B 1, and  0, BL|D| 14 .
Set 2 = 0 when = ZD to obtain
2 = 02 + 21 (3L2 )2 02 + 12 (SL2 )2.
Dispersion limit:

BL|S|2

1/ 8.

JJ
II
J
I
Back
Close

Dispersion Limitations
Small Source Spectral Width: V  1

48/269

When 3 = 0 and C = 0,
2 = 02 + (2L/20)2.
One can minimize by adjusting input width 0.
Minimum occurs for 0 = (|2|L/2)1/2 and leads to = (|2|L)1/2.
Dispersion limit when 3 = 0:
p
B |2|L 14 .
Dispersion limit when 2 = 0:
B(|3|L)1/3 0.324.

JJ
II
J
I
Back
Close

Dispersion Limitations (cont.)

49/269

Even a 1-nm spectral width limits BL < 0.1 (Gb/s)-km.

DFB lasers essential for most lightwave systems.
For B > 2.5 Gb/s, dispersion management required.

JJ
II
J
I
Back
Close

Dispersion Limitations (cont.)

Effect of Frequency chirp
50/269

Numerical simulations necessary for more realistic pulses.


 2m
t
Super-Gaussian pulse: A(0, T ) = A0 exp 1+iC
.
2
T0
Chirp can affect the system performance drastically.

JJ
II
J
I
Back
Close

Fiber Losses
Definition: (dB/km) =

10
L

log10

Pout
Pin

4.343.
51/269

Material absorption (silica, impurities, dopants)

Rayleigh scattering (varies as 4)
Waveguide imperfections (macro and microbending)
Conventional Fiber
Dispersion

Dry Fiber

JJ
II
J
I
Back
Close

Major Nonlinear Effects

Stimulated Raman Scattering (SRS)
52/269

Stimulated Brillouin Scattering (SBS)

Self-Phase Modulation (SPM)
Cross-Phase Modulation (XPM)
Four-Wave Mixing (FWM)
Origin of Nonlinear Effects in Optical Fibers
Ultrafast third-order susceptibility (3).
Real part leads to SPM, XPM, and FWM.
Imaginary part leads to SBS and SRS.

JJ
II
J
I
Back
Close

Brillouin Scattering
Scattering of light from acoustic waves (electrostriction).
53/269

Energy and momentum conservation laws require

B = p s and kA = kp ks.
Brillouin shift: B = |kA|vA = 2vA|kp| sin(/2).
only possibility = for single-mode fibers
(backward propagating Stokes wave).
Using kp = 2
n/p, B = B /2 = 2
nvA/p.
With vA = 5.96 km/s and n
= 1.45, B 11 GHz near 1.55 m.
Stokes wave grows from noise.
Not a very efficient process at low pump powers.

JJ
II
J
I
Back
Close

Stimulated Brillouin Scattering

Becomes a stimulated process at high input power levels.
54/269

Governed by two coupled equations:

dIp
dIs
= gB IpIs pIp,
= +gB IpIs sIs.
dz
dz
Brillouin gain has a narrow Lorentzian spectrum
(TB 10 ns; 20 MHz)
gB () =

gB (B )
.
1 + ( B )2TB2
JJ
II
J
I
Back
Close

SBS Threshold
Threshold condition: gB PthLeff /Aeff 21.

55/269

Effective fiber length: Leff = [1 exp(L)]/.

Effective core area: Aeff 5080 m2.
Peak Brillouin gain: gB 5 1011 m/W.
Low threshold power for long fibers (5 mW).
Most of the power reflected backward after SBS threshold.
Threshold can be increased using
Phase modulation at frequencies >0.1 GHz.
Sinusoidal strain along the fiber.
Nonuniform core radius or dopant density.

JJ
II
J
I
Back
Close

Stimulated Raman Scattering

Scattering of light from vibrating molecules.
56/269

Scattered light shifted in frequency.

Raman gain spectrum extends over 40 THz.
(a)

(b)

Raman shift at Gain peak: R = p s 13 THz).

JJ
II
J
I
Back
Close

SRS Threshold
SRS governed by two coupled equations:
57/269

dIp
= gR IpIs pIp
dz
dIs
= gR IpIs sIs.
dz
Threshold condition: gR PthLeff /Aeff 16.
Peak Raman gain: gR 6 1014 m/W near 1.5 m.
Threshold power relatively large ( 0.6 W).
SRS is not of concern for single-channel systems.
Leads to interchannel crosstalk in WDM systems.

JJ
II
J
I
Back
Close

Self-Phase Modulation
Refractive index depends on intensity as
58/269

n(, I) = nL() + n2I(t).

n2 = 2.6 1020 m2/W for silica fibers.
2P/Aeff + P .
Propagation constant: 0 = + k0n
n2/(Aeff ).
Nonlinear parameter = 2
Leads to nonlinear Phase shift:
Z L
Z
NL =
( 0 ) dz =
0

P (z) dz = PinLeff .

Optical field modifies its own phase (SPM).

Phase shift varies with time for pulses (chirping).

JJ
II
J
I
Back
Close

SPM-Induced Chirp
59/269

Chirp depends on the pulse shape.

Gaussian pulses (m = 1): Nearly linear chirp across the pulse.
Super-Gaussian pulses (m = 1): Chirping only near pulse edges.
In all cases, SPM leads to spectral broadening.

JJ
II
J
I
Back
Close

Cross-Phase Modulation
Nonlinear refractive index seen by one wave depends on the
intensity of other copropagating channels.

60/269

Nonlinear index for two channels:

nNL = n2(|A1|2 + 2|A2|2).
Total nonlinear phase shift for multiple channels

X
NL

Pm .
j = Leff Pj + 2
m6=j

XPM induces a nonlinear coupling among channels.

JJ
II
J
I
Back
Close

Nonlinear Schr
odinger Equation
Nonlinear effects can be included by adding a nonlinear term to the
equation used earlier for dispersive effects.

61/269

This equation is known as the Nonlinear Schrodinger Equation:

A i2 2A
= i|A|2A.
+
2
z
2 t
Nonlinear parameter: = 2
n2/(Aeff ).
Fibers with large Aeff help through reduced .
Known as large effective-area fiber or LEAF.
Nonlinear effects leads to formation of optical solitons.

JJ
II
J
I
Back
Close

Four-Wave Mixing
FWM converts two photons from one or two pump beams into two
new frequency-shifted photons.

62/269

Energy conservation: 1 + 2 = 3 + 4.
Degenerate FWM: 21 = 3 + 4.
Momentum conservation or phase matching is required.
FWM efficiency governed by phase mismatch:
= (3) + (4) (1) (2).
In the degenerate case (1 = 2), 3 = 1 + , and 4 = 1 .
2

Expanding in a Taylor series, = 2 .

FWM becomes important for WDM systems designed with lowdispersion fibers.

JJ
II
J
I
Back
Close

FWM: Good or Bad?

FWM leads to interchannel crosstalk in WDM systems.
63/269

It can be avoided through dispersion management.

On the other hand . . .
FWM can be used beneficially for
Parametric amplification
Optical phase conjugation
Demultiplexing of OTDM channels
Wavelength conversion of WDM channels
Supercontinuum generation

JJ
II
J
I
Back
Close

Fiber Design
64/269

Core doped with GeO2; cladding with fluorine.

Index profile rectangular for standard fibers.
Triangular index profile for dispersion-shifted fibers.
Raised or depressed cladding for dispersion control.

JJ
II
J
I
Back
Close

Commercial Fibers
Corning SMF-28 ZD = 1.31 m, D (1.55 m) = 17 ps/(km-nm),
S = 0.09 ps/(km-nm2), =
0.21 dB/km,
Aeff = 60 m2, Dp > 0.2 ps/ km.

65/269

Lucent TrueWave RS ZD = 1.45 m, D (1.55 m)

2
= 4.4 ps/(km-nm), S = 0.04 ps/(km-nm
), = 0.205 dB/km,

Aeff = 55 m2, Dp < 0.05 ps/ km.

Corning LEAF ZD = 1.50 m, D (1.55 m) = 4.2 ps/(km-nm),
S = 0.085 ps/(km-nm2), =
0.21 dB/km,
2
Aeff = 72 m , Dp < 0.07 ps/ km.
Lucent AllWave ZD = 1.31 m, D (1.55 m) = 17 ps/(km-nm),
S = 0.09 ps/(km-nm2), = 0.19
dB/km,
2
Aeff = 80 m , Dp < 0.05 ps/ km.

JJ
II
J
I
Back
Close

Fabrication Methods
Two-Stage Fabrication
66/269

preform: Length 1 m, diameter 2 cm; correct index profile.

Preform drawn into fiber using a draw tower.
Preform Fabrication Techniques
Modified chemical vapor deposition (MCVD).
Outside vapor deposition (OVD).
Vapor Axial deposition (VAD).
JJ
II
J
I
Back
Close

Fiber Draw Tower

67/269

JJ
II
J
I
Back
Close

Fiber Cables
Light-Duty Cables
68/269

Heavy-Duty Cables

JJ
II
J
I
Back
Close

Optical Transmitters
Use semiconductor lasers or light-emitting diodes (LED)
as an optical source.

69/269

Semiconductor lasers offer many advantages:

Compact size, high efficiency, good reliability.

Emissive area compatible with fibers.

Electrical pumping at modest current levels.
Output can be modulated directly at high frequencies.
Room-temperature operation first realized in 1970.

JJ
II
J
I
Back
Close

Laser or LED
70/269

Three fundamental processes: Absorption, spontaneous emission

and stimulated emission.
LEDs use spontaneous emission: 3040 nm wide spectrum, low
power, and low coherence.
Lasers use stimulated emission: narrow spectrum, high power, and
high coherence.

JJ
II
J
I
Back
Close

pn Junctions
71/269

(a)

(b)

A pn junction separates p-type and n-type semiconductors.

Diffusion of electrons and holes creates a built-in electric field in
the depletion region (tilted lines).
Forward biasing of pn junction reduces the built-in field and allows
electron-hole recombination in the depletion region.

JJ
II
J
I
Back
Close

Problems with pn Junctions

Early semiconductor lasers (before 1970) worked only at liquid-air
temperatures and required large current densities to reach threshold.

72/269

They suffered from the lack of carrier confinement.

Electrons and holes distributed over the entire depletion region.
At the same time, light was not confined to this region.
Both problems can be solved using the double-heterostructure design in which a thin layer of a different semiconductor is inserted
between the pn junction.

JJ
II
J
I
Back
Close

Double-Heterostructure Design
Proposed in 1963 (Noble Prize in 2000).
73/269

Active region sandwiched between two cladding layers.

Active layer has lower Eg and higher index n.

JJ
II
J
I
Back
Close

Semiconductor Materials
74/269

Need two semiconductors with different bandgaps but

same lattice constant.
GaAs lasers use ternary compound AlxGa1xAs
Eg (x) = 1.424 + 1.247x
(x < 0.45).
InP lasers use quaternary compound In1xGaxAsy P1y
Eg (y) = 1.35 0.72y + 1.247y 2
(x = 0.45y, y 1).

JJ
II
J
I
Back
Close

Semiconductor Lasers
Used routinely for almost all communication systems.
75/269

Can be modulated at bit rates >10 Gb/s.

Simplest structure: Broad-area laser.

Population inversion realized using a double-heterostructure design.

Stimulated emission within active layer produces coherent light.

JJ
II
J
I
Back
Close

Laser Structures
76/269

Light confined to the active layer by the waveguide.

No such confinement exists in the lateral direction.
Mode spreads over the entire width.
Highly elliptical spot size ( 100 m1 m.)
Poor beam quality without lateral confinement.

JJ
II
J
I
Back
Close

Lateral Mode Confinement

Gain-guided lasers
77/269

Index-guided lasers

JJ
II
J
I
Back
Close

Optical Spectrum
78/269

Gain spectrum relatively broad (3050 nm).

Mode spacing 100 GHz for a < 1-mm-long cavity.
Many modes experience gain simultaneously.
Typically, several side modes accompany main mode.
Spectral width 23 nm for FabryPerot lasers.
Such lasers cannot be used at bit rates > 1 Gb/s.

JJ
II
J
I
Back
Close

Single-Mode Lasers
Basic idea: Increase cavity losses for all modes except the one
designed to reach the threshold first.

79/269

Several techniques have been used (distributed feedback, external

grating, two coupled cavities, etc.)
Distributed feedback (DFB) technique is used commercially to make
DFB lasers.

JJ
II
J
I
Back
Close

DFB and DBR Lasers

80/269

Basic idea: Use an index grating for lowering losses at a specific wavelength set by the Bragg condition: B = 2
n.
Grating period sets the laser wavelength.
Typically 0.2 m for B = 1.55 m.
Grating made using a holographic technique.

JJ
II
J
I
Back
Close

Tunable Lasers
81/269

External grating provides tuning over >50 nm.

Multisection DBR lasers can provide tuning over 40 nm when a
superstructure grating is used.

JJ
II
J
I
Back
Close

VCSELs
82/269

Vertical cavity leads to surface emission.

Short cavity: mode spacing exceeds gain bandwidth.
Require high-reflectivity (>99.9%) DBR mirrors.
Low output power (1 mW) in a single mode.
Many lasers on the same chip (2D arrays)
Useful for datacom and LAN applications.

JJ
II
J
I
Back
Close

CW Laser Characteristics
83/269

Laser power increases linearly after threshold.

Threshold current 10 mA at room temperature but increases exponentially at high temperatures.
Thermo-electric cooler used for temperature control.
Output powers >100 mW possible even for DFB lasers.

JJ
II
J
I
Back
Close

Laser Rate Equations

Photon and electron populations (P and N ) satisfy:
84/269

dP
P
= GP + Rsp ,
dt
p
I N
dN
= GP,
dt
q c
G = vg gm = GN (N N0),

Rsp = nspG.

Spontaneous-emission factor nsp 2.

Photon lifetime related to the cavity loss as
p = 1/(vg cav ) 4 ps.
Carrier lifetime c = 23 ns.

JJ
II
J
I
Back
Close

Steady State Solution

Set time derivatives to zero (steady state). Neglect Rsp.

85/269

The photon equation becomes: (G 1/p)P = 0.

Below threshold G < 1/p. Thus, P = 0 and N = cI/q.
Threshold reached at a current for which Gp = 1.
N clamped to its threshold value Nth = N0 + (GN p)1.

Threshold current:

Ith =

Above threshold (I > Ith):

qNth
c

q
c

N0 + G

1
N p

P = (p/q)(I Ith).

Emitted power Pe = 21 (vg mir)~P .

JJ
II
J
I
Back
Close

Direct Modulation
Time-dependent current of the form I(t) = Ib + Imfp(t).

86/269

Laser output is also time-dependent.

Solutions of rate equations predicts P (t).
An additional rate equation needed for the optical phase:


d 1
1
= c GN (N N0)
dt
2
p
At high powers G = GN (N N0)(1 NLP ).
Small-signal modulation: Im  Ib.
Linearize the rate equations and solve the linear equations in the
frequency domain.

JJ
II
J
I
Back
Close

Large-Signal Modulation
87/269

Need to solve the rate equations numerically.

Time-dependent phase leads to frequency chirp


1 d
c
1
.
(t) =
=
GN (N N0)
2 dt
4
p

JJ
II
J
I
Back
Close

Transmitter Design
88/269

Important design issues:

Efficient laserfiber coupling necessary.
Control of optical feedback (angled facets).
Stability of output power with aging (backfacet monitoring).

JJ
II
J
I
Back
Close

External Modulators
89/269

External modulation necessary at B > 5 Gb/s.

LiNbO3 modulators use electro-optic effect.
Electroabsorption modulators use FranzKeldysh effect and can be
integrated with the laser on the same chip.
Both can operate at bit rates up to 40 Gb/s.

JJ
II
J
I
Back
Close

Optical Receivers
A photodiode converts optical signal into electrical domain.
90/269

Amplifiers and filters shape the electrical signal.

A decision circuit reconstructs the stream of 1 and 0 bits.

Electrical and optical noises corrupt the signal.

Performance measured through bit error rate (BER).
BER < 109 required for all lightwave systems.

Receiver sensitivity: Minimum amount of optical power required to

realize the desirable BER.

JJ
II
J
I
Back
Close

Photodetection
91/269

A semiconductor layer absorbs incident light if h > Eg .

Electronhole pairs collected by applying voltage.
Photocurrent I = RPin (R measures responsivity).
Quantum Efficiency: =
=

I/q
Pin /h

h
q R.

Pabs Pin Ptr

=
= 1 exp(W ).
Pin
Pin

JJ
II
J
I
Back
Close

Absorption in Semiconductors
92/269

Absorption occurs for < hc/Eg .

Large absorption coefficient for most semiconductors.
Layer thickness < 5 m sufficient in practice.

JJ
II
J
I
Back
Close

Photodiodes
93/269

A reverse-biased pn junction is used.

Electronhole pairs in the depletion region swept by the large
electric field across it (drift current).
Electron-hole pairs outside the depletion region produce
diffusion current.

JJ
II
J
I
Back
Close

Response Time of Photodiodes

94/269

Drift current is produced quickly (transit time <100 ps).

Diffusion is a relatively slow process (1 ns).
This mismatch leads to distortion.
Problem can be solved using pin photodiodes.

JJ
II
J
I
Back
Close

pin Photodiodes
95/269

Insert an intrinsic or i layer between the pn junction.

Reduce the thickness of p- and n-type layers.
Use a higher-bandgap material for p- and n-type layers.
No absorption in these layers if Eg > h.

JJ
II
J
I
Back
Close

Avalanche Photodiodes
96/269

APDs increase responsivity through internal gain.

A single photon produces many electron-hole pairs.
High electric field across the gain layer.
Accelerating electrons and holes generate new electronhole pairs
though impact ionization.

JJ
II
J
I
Back
Close

Receiver Design
97/269

Front end converts optical signal into electrical form.

Linear channel amplifies and filters the electrical signal.
Data recovery section creates electrical bit stream using clockrecovery and decision circuits.

JJ
II
J
I
Back
Close

Data-Recovery Section
98/269

A clock-recovery circuit isolates the frequency f = B from the

received signal.
The clock helps to synchronize the decision process.
Decision circuit compares the output to a threshold level at sampling
times set by the clock.
Eye diagram useful for system monitoring.
The best sampling time corresponds to maximum eye opening.

JJ
II
J
I
Back
Close

Receiver Noise
Shot noise: I(t) = Ip + is(t).

99/269

Average his(t)i = 0 but

Z

his(t)is(t + )i =

Ss(f ) exp(2if ) df.

White noise: Spectral density Ss(f ) constant.

R
2
Noise variance: s = Ss(f ) df = 2qIp f .
Effective noise bandwidth f related to detector bandwidth.
Adding the contribution of dark current Id,
s2 = 2q(Ip + Id)f.

JJ
II
J
I
Back
Close

Receiver Noise (continued)

Thermal noise: I(t) = Ip + is(t) + iT (t).

100/269

Spectral density ST (f ) = 2kB T /RL depends on temperature and

load resistor RL.
R
2
Noise variance: T = Ss(f ) df = (4kB T /RL)f .
Amplifier noise: All electrical amplifiers enhance thermal noise by
the amplifier noise figure Fn.
Total thermal noise: T2 = (4kB T /RL)Fnf .
Total Receiver Noise
2 = s2 + T2 = 2q(Ip + Id)f + (4kB T /RL)Fnf.

JJ
II
J
I
Back
Close

pin Receivers
Signal-to-Noise Ratio:
101/269

Ip2
R2Pin2
.
SNR = 2 =

2q(RPin + Id)f + 4(kB T /RL)Fnf

Thermal-Noise Limit (T  s):
RLR2Pin2
SNR =
.
4kB T Fnf
Noise-equivalent power defined as the minimum optical power per
unit bandwidth required to produce SNR = 1:

1/2

1/2
Pin
4kB T Fn
h 4kB T Fn
NEP =
=
=
.
RL R 2
q
RL
f

JJ
II
J
I
Back
Close

APD Receivers
Average current larger by the APD gain factor M :
102/269

Ip = M RPin = RAPDPin.
Thermal noise unchanged but shot noise enhanced by a factor FA
known as excess noise factor.
Shot-noise variance: s2 = 2qM 2FA(RPin + Id)f .
Signal-to-Noise Ratio:
Ip2
(M RPin)2
SNR = 2
=
.
s + T2
2qM 2FA(RPin + Id)f + 4(kB T /RL)Fnf

JJ
II
J
I
Back
Close

Bit Error Rate

103/269

BER = Error probability per bit

BER = p(1)P (0/1) + p(0)P (1/0) = 12 [P (0/1) + P (1/0)].
P (0/1) = conditional probability of deciding 0 when 1 is sent.
Since p(1) = p(0) = 1/2, BER = 21 [P (0/1) + P (1/0)].
Assume Gaussian statistics for the current.

JJ
II
J
I
Back
Close

Bit Error Rate (continued)

P (0/1) = Area below the decision level ID




Z ID
(I I1)2
1
I1 ID
1

exp
dI = erfc
P (0/1) =
.
212
2
1 2
1 2

104/269

P (1/0) = Area above the decision level ID





Z
1
(I I0)2
1
ID I0

P (1/0) =
exp
dI = erfc
.
2
2
2
0 2 ID
0 2
0
R
Complementary error function erfc(x) = 2 x exp(y 2) dy.
Final Answer
 



1
I1 ID
ID I0

BER =
erfc
+ erfc
.
4
1 2
0 2

JJ
II
J
I
Back
Close

Bit Error Rate (continued)

BER depends on the decision threshold ID .

105/269

Minimum BER occurs when ID is chosen such that

 
(ID I0)2 (I1 ID )2
1
=
+
ln
.
202
212
0
Last term negligible in most cases, and
(ID I0)/0 = (I1 ID )/1 Q.
ID =

0I1 + 1I0
,
0 + 1

Q=

I1 I0
.
1 + 0

Final Expression
 
1
Q
exp(Q2/2)

BER = erfc

.
2
2
Q 2

JJ
II
J
I
Back
Close

Q Parameter
106/269

Q=

I1 I0
1 +0

is a measure of SNR.

Q > 6 required for a BER of < 109.

Q = 7 provides a BER of < 1012.

JJ
II
J
I
Back
Close

Minimum Received Power

Receiver sensitivity = Minimum average received power needed to
keep the BER below a certain value (typically 109).

107/269

We need to relate the Q parameter to incident optical power.

Assume 0 bits carry no optical power so that P0 = I0 = 0.
I1 = M RP1 = 2M RPrec, where Prec = (P1 + P0)/2.
Including both shot and thermal noise,
1 = (s2 + T2 )1/2 and 0 = T ,
s2 = 2qM 2FAR(2Prec)f,
T2 = (4kB T /RL)Fnf.
Using these results,
I1
2M RPrec
Q=
=
.
1 + 0 (s2 + T2 )1/2 + T

JJ
II
J
I
Back
Close

Minimum Received Power (cont.)

Solving for received power, we obtain


Q

T
Prec =
qFAQf +
.
R
M

108/269

For a pin receiver, we set M = 1.

Since thermal noise dominates for such a receiver,
Prec QT /R.
Using R 1 A/W near 1.55 m, Prec = QT .
As an example, if we use R = 1 A/W, T = 100 nA, and Q = 6,
Prec = 0.6 W or 32.2 dBm.

JJ
II
J
I
Back
Close

Minimum Received Power (cont.)

Receiver sensitivity can be expressed in terms of number of photons
Np contained within a single 1 bit.

109/269

In the shot-noise limit, I0 = 0 and 0 = 0 when 0 bits carry no

power, and Q = I1/1 = (SNR)1/2.
SNR related to Np as SNR Np, or
p

1
Np/2 .
BER = 2 erfc
For = 1, BER = 1 109, Np = 36. Thus, 36 photons are
sufficient in the shot-noise limit.
In practice, most optical receivers require Np > 1000 because of
thermal noise.

JJ
II
J
I
Back
Close

Quantum Limit of Photodetection

The BER obtained in the shot-noise limit not totally accurate.
110/269

Its derivation based on the Gaussian approximation for noise.

Poisson statistics should be used for small number of photons.
For an ideal detector (no thermal noise, no dark current, and = 1),
0 bits produce no photons, and 0 = 0.
Error occurs only if 1 bit fails to produce even one eh pair.
Probability of generating m eh pairs: Pm = exp(Np)Npm/m!.
Since P (0/1) = exp(Np), BER = exp(Np)/2.
9

Np = 20 for BER = 1 10

(10 photons/bit on average).

Prec = NphB/2 = NphB = 13 nW or 48.9 dBm at B =

10 Gb/s.

JJ
II
J
I
Back
Close

Sensitivity Degradation
Real receivers need more power than Prec.

111/269

Increase in power is referred to as power penalty.

In decibel units, power penalty is defined as
Power penalty =

Increased Power
.
Original Power

Several mechanisms degrade the receiver sensitivity.

Three main sources of power penalty are:
? Finite Extinction ratio (P0 6= 0)
? Intensity Noise of the received optical signal.
? Timing Jitter of electronic circuits.

JJ
II
J
I
Back
Close

Extinction Ratio
extinction ratio is defined as rex = P0/P1.

112/269

Power penalty can be obtained by calculating Q parameter.

For a pin receiver I1 = RP1 and I0 = RP0.
using Prec = (P1 + P0)/2,


1 rex 2RPrec
Q=
.
1 + rex 1 + 0
In thermal noise limit, 1 0 T .
Received power for a finite extinction ratio


1
+
r
T Q
ex
Prec(rex) =
.
1 rex
R

JJ
II
J
I
Back
Close

Extinction Ratio (continued)

113/269

Power penalty is found to be





1 + rex
Prec(rex)
ex = 10 log10
= 10 log10
.
1 rex
Prec(0)
A 1-dB penalty occurs for rex = 0.12 and increases to 4.8 dB for
rex = 0.5.

JJ
II
J
I
Back
Close

Receiver Performance
114/269

Measured receiver sensitivities as a function of bit rate for pin

(circles) and APD (triangles) receivers.
Straight ines show the quantum limit.
Measured values 20-dB higher than quantum-noise limit (due to
thermal noise).

JJ
II
J
I
Back
Close

BER Measurements
115/269

Receiver sensitivity = Power required for BER < 109.

It degrades with increasing link length (power penalty).

JJ
II
J
I
Back
Close

Lightwave Systems
System Architectures
116/269

? Point-to-Point Links
? Distribution Networks
? Local-Area Networks
Design Guidelines
? Loss-Limited Lightwave Systems
? Dispersion-Limited Lightwave Systems
? Power Budget
Terrestrial Lightwave Systems
Undersea Lightwave Systems

JJ
II
J
I
Back
Close

Point-to-Point Links
117/269

Long-haul systems used for the network backbone.

Require periodic loss compensation.
Regenerators (receiver+transmitter) used until 1995.
Optical amplifiers keep the signal in the optical domain.
Dispersive and nonlinear degradations accumulate.
Necessary to use a regenerator after some amplifiers.

JJ
II
J
I
Back
Close

Distribution Networks
118/269

Examples include metropolitan-area networks and CATV networks.

Make use of hub or bus topology.
Distribution losses limit the optical power. For the bus topology
PN = PT C[(1 )(1 C)]N 1.

JJ
II
J
I
Back
Close

Local-Area Networks
119/269

Random bidirectional access (Ethernet).

Make use of ring or star topology.
For the star topology, PN = (PT /N )(1 )log2 N .

JJ
II
J
I
Back
Close

Design Guidelines
120/269

Step-index fiber

Graded-index Fiber

Solid lines: loss limit; Dashed lines: dispersion limit.

Transmitted Distance depends on the bit rate and wavelength chosen to design the system.

JJ
II
J
I
Back
Close

Power Budget
Power distribution along the link: Ptr = Prec + CL + Ms.

121/269

Channel loss CL = f L + con + splice

System Margin Ms = 2 to 6 dB.
Quantity
Symbol
Laser
LED
Transmitter power
Ptr
0 dBm 13 dBm
Receiver sensitivity
Prec
42 dBm 42 dBm
System margin
Ms
6 dB
6 dB
Available channel loss
CL
36 dB
23 dB
Connector loss
con
2 dB
2 dB
Fiber cable loss
f
3.5 dB/km 3.5 dB/km
Maximum fiber length
L
9.7 km
6 km

JJ
II
J
I
Back
Close

Performance-Limiting Factors
Fiber Dispersion: Impacts all long-haul systems.
122/269

Compensation of accumulated dispersion at the receiver or along

the link often employed.
Nonlinear Effects: Self-phase modulation (SPM) leads to spectral
broadening. Its effect studied by solving:
A i2 2A

2
=

+
A
+
i|A|
A.
z
2 t2
2
Amplifier noise: Spontaneous emission adds noise. It reduces SNR
and produces timing jitter.
PMD becomes a limiting factor for old fibers at high bit rates.
Other limiting factors for WDM systems
(XPM, FWM, Raman amplification, etc.).

JJ
II
J
I
Back
Close

Capacity Management
Increase bit rate of the data stream (up to 40 Gb/s possible).
123/269

Use multiple channels at different wavelengths (optical FDM or

WDM).
Use multiple channels interleaved in the time domain (optical TDM).

System capacities as large as 11 Tb/s have been demonstrated in

laboratory experiments.
Management of fiber nonlinearities and dispersion is essential.

JJ
II
J
I
Back
Close

Terrestrial Lightwave Systems

System
FT3
FT3C
FT3X
FTG
FTG-1.7
STM16
STM64
STM256

Year
1980
1983
1984
1985
1987
1991
1996
2002

B
L
Voice
(m) (Mb/s) (km) Channels
0.85
45 < 10
672
0.85
90 < 15
1,344
1.30
180 < 25
2,688
1.30
417 < 40
6,048
1.30
1,668 < 46
24,192
1.55
2,488 < 85
32,256
1.55
9,953 < 90 129,024
1.55 39,813 < 90 516,096

WDM systems commercialized after 1995.

160-channel system with 1.6 Tb/s became available by 2000.

124/269

JJ
II
J
I
Back
Close

Undersea Lightwave Systems

System
TAT8
TAT9
TAT10/11
TAT12/13
AC1
TAT14
AC2
360Atlantic
Tycom
FLAG

Year B (Gb/s) L(km) Comments

1988
0.28
70
1.3 m, multimode lasers
1991
0.56
80
1.55 m, DFB lasers
1993
0.56
80
1.55 m, DFB lasers
1996
5.00
50
1.55 m, optical amplifiers
1998
80.0
50
1.55 m, WDM, amplifiers
2001
1280
50
1.55 m, dense WDM
2001
1280
50
1.55 m, dense WDM
2001
1920
50
1.55 m, dense WDM
2002
2560
50
1.55 m, dense WDM
2002
4800
50
1.55 m, dense WDM

Optical amplifiers were first used in 1996.

Up to 4.8 Tb/s capacity using 4 fibers.

125/269

JJ
II
J
I
Back
Close

Computer-Aided Design
126/269

A software package simulates the performance of entire system.

Bit pattern generated by solving the rate equations and propagated
by solving the NLS equation.
Signal amplified periodically and ASE noise added.
Signal detected, filtered, and analyzed at the receiver.

JJ
II
J
I
Back
Close

Loss Management
127/269

Periodic regeneration of optical signal.

Regenerator = Receiver + Transmitter
Periodic amplification using optical amplifiers.
Amplifier spacing is an important design parameter.
It should be as large as possible for economical reasons.
LA = 80 km for terrestrial lightwave systems.

JJ
II
J
I
Back
Close

Optical Amplifiers
Used routinely for loss compensation since 1995.
128/269

Amplify the input signal but also add some noise.

Several kinds of amplifiers have been developed.
? Semiconductor optical amplifiers
? Raman-based fiber amplifiers
? Erbium-doped fiber amplifiers
EDFAs are used most commonly for lightwave systems.
Raman amplifiers work better for long-haul systems.

JJ
II
J
I
Back
Close

Basic Concepts
An amplifier is nothing but a laser without feedback.
129/269

Amplifier pumped optically or electrically to realize optical gain.

Optical gain has the general form
g(, P ) =

g0
.
1 + ( 0)2T22 + P/Ps

Gain reduced at high signal powers. Saturation power Ps depends

on material parameters.
Parameter T2 determines gain bandwidth (typically T2 <1 ps).
Gain maximum at = 0 where gain peak is located.

JJ
II
J
I
Back
Close

Amplifier Noise
Optical amplifiers introduce noise and degrade SNR.
130/269

Source of noise: Spontaneous emission

Noise spectral density
Ssp() = (G 1)nsph.
Population inversion factor nsp = N2/(N2 N1) > 1.
Amplifier noise figure Fn is defined as
Fn =

(SNR)in
.
(SNR)out

SNR refers to electrical signal-to-noise ratio.

JJ
II
J
I
Back
Close

Amplifier Noise (continued)

For any amplifier Pout = GPin.

131/269

SNR of the input signal in the shot-noise limit

hIi2
(RPin)2
Pin
(SNR)in = 2 =
=
.
s
2q(RPin)f
2hf
Amplified signal is accompanied with spontaneous emission.
Beating of signal and spontaneous emission produces at detector

I = R| GEin + Esp|2.
Since Ein and Esp have a random phase difference , Noise current
I = 2R(GPin)1/2|Esp| cos .

JJ
II
J
I
Back
Close

Amplifier Noise (continued)

Averaging over , and neglecting all other noise sources,
132/269

2 4(RGPin)(RSsp)f.
SNR of the amplified signal
(SNR)out

GPin
hIi2 (RGPin)2

= 2 =
.

2
4Sspf

Noise figure for the amplifier

Fn = 2nsp(G 1)/G 2nsp.
SNR degraded by 3 dB even for an ideal amplifier (nsp = 1).
In practice, Fn exceeds 3 dB and can be as large as 68 dB.

JJ
II
J
I
Back
Close

Amplifier Applications
133/269

JJ
II
J
I
Back
Close

Semiconductor Optical Amplifiers

A laser biased below threshold acts as a FP amplifier.
134/269

Feedback at facets provides most gain near FP resonances.

Amplification factor obtained considering multiple reflections:
GFP() =

(1 R1)(1 R2)G()

,
2
(1 G R1R2) + 4G R1R2 sin2[( m)/L]

G() is the single-pass amplification factor.

GFP reduces to G when R1 = R2 = 0.
Amplifier bandwidth set by sharpness of cavity resonances.
3-dB bandwidth is found to be



2L 1 1 G R1R2

A =
sin
.

(4G R1R2)1/2

JJ
II
J
I
Back
Close

Traveling-Wave Amplifiers
FP amplifiers not suitable because of their narrow bandwidth.
135/269

Feedback at the cleaved facets must be suppressed.

Common to coat facets with an antireflection coating.
AR coating should be good enough to suppress FP resonances.
Maximum and minimum values of GFP satisfy


2
max
1 + G R1 R2
G

G = FP
=
.
Gmin
1 G R1 R2
FP
If G exceeds 3 dB, amplifier bandwidth is set by cavity resonances
rather than by the gain spectrum.

To keep G < 2, facet reflectivities should satisfy G R1R2 <

0.17.

JJ
II
J
I
Back
Close

SOA Design
136/269

Difficult to produce AR coatings with Rm < 0.1%.

Tilted-stripe structure: Active stripe tilted.
Buried-facet structure: transparent regions near each facet.
Rm < 104 when used in combination with AR coatings.

JJ
II
J
I
Back
Close

Amplifier Characteristics
137/269

Wide gain spectrum with negligible ripples.

High gain (>25 dB) at low currents.
Output saturation power in the range 510 mW.
Noise figure Fn in the range 57 dB.

JJ
II
J
I
Back
Close

Problems with SOAs

138/269

SOA provide Polarization-dependent gain:

Several schemes can solve this problem partially.
Large coupling losses (68 dB or more).
Relatively large noise figure.

JJ
II
J
I
Back
Close

Raman Amplifiers
139/269

Pumped optically in the forward or backward direction.

Use stimulated Raman scattering for signal amplification.
Can work at any wavelength with suitable pumping.
Can provide large bandwidth using multiple pumps.
Fiber used for data transmission can itself be used.
Gain can be made polarization independent using orthogonally polarized pumps at each wavelength.

JJ
II
J
I
Back
Close

Raman Gain and Bandwidth

140/269

Main gain peak shifted by 13.2 THz (about 100 nm) from
pump wavelength.
Raman gain spectrum extends over >20 THz.
Gain coefficient g() = gR ()(Pp/ap) depends on
mode area ap of the pump.
Can differ by a factor of 5 or more for SMF, DSF, and DCF.

JJ
II
J
I
Back
Close

Simple Theory
dPs/dz = sPs + (gR /ap)PpPs,

141/269

dPp/dz = pPp (p/s)(gR /ap)PsPp.

Ignoring pump depletion, Pp(z) = Pp(0) exp(pz).
Signal power for a Raman amplifier of length L:
Ps(L) = Ps(0) exp(gR P0Leff /ap sL),
where P0 = Pp(0) and Leff = [1 exp(pL)]/p.
Amplifier gain is given by
Ps(L)
= exp(g0L),
Ps(0) exp(sL)
 

P0
Leff
gR P0
g0 = gR

.
ap
L
appL

GA =

JJ
II
J
I
Back
Close

Raman Gain Saturation

142/269

30 dB gain at a pump power 1 W for L 1 km.

No gain saturation up to 0.5 W (Ps 1 W).

JJ
II
J
I
Back
Close

Raman Amplifier Design

143/269

Multiple pumps can provide wide and flat gain spectrum.

Noise is smaller for Raman amplifiers (nsp 1).
Suffer from double Rayleigh scattering and pump-noise transfer.
Backward pumping provides a better design.

JJ
II
J
I
Back
Close

Erbium-Doped Fiber Amplifiers

Developed after 1987 and commercialized during the 1990s.
144/269

Fiber core doped with erbium (length 20-200 m).

Pumped using diode lasers operating at 980 or 1480 nm.
Provide 2030 dB gain at pump powers below 50 mW.
Gain bandwidth up to 40 nm possible.
Noise figure in the range 45 dB.
Provide polarization-independent gain.
Gain pattern independent (Response time 10 ms).
Can be designed to work in both the C and L bands.

JJ
II
J
I
Back
Close

Pump Characteristics
145/269

Amorphous nature of silica broadens energy levels into bands.

Semiconductor lasers at 980 and 1480 nm used most commonly.
30-dB gain obtained with only 1015 mW of absorbed pump power.
Pumping efficiency up to 11 dB/mW possible at 980 nm.
Pumping at 1480 nm requires longer fibers and higher powers.

JJ
II
J
I
Back
Close

Amplifier Characteristics
146/269

L = 1015 m enough to provide 2030 dB gain.

For a given pump power, G becomes maximum at an optimum
value of L and drops sharply when L exceeds this optimum value.
It is necessary to choose both L and Pp appropriately.

JJ
II
J
I
Back
Close

Amplifier Noise
147/269

Fn = 2nsp but nsp = N2/(N2 N1).

For a three-level pumping scheme, N1 6= 0 and nsp > 1.
Both N1 and N2 vary along fiber length, and nsp should be averaged
over length.
Noise figure depends both on length L and pump power Pp.

JJ
II
J
I
Back
Close

Amplifier Design
C-Band EDFAs
148/269

L-Band EDFAs

JJ
II
J
I
Back
Close

Preamplification at Receiver
Sensitivity of optical receivers can be improved by preamplifying the
signal before it falls on the photodetector.

149/269

Preamplification makes the signal strong enough that thermal noise

becomes negligible compared with shot noise.
Amplifier noise can be reduced by using an optical filter after the
amplifier.
Filter works because spontaneous emission is spread over the entire
amplifier bandwidth (40 nm or so).
Filter bandwidth is chosen to be <1 nm.
Receiver sensitivity can be improved by 10 dB or more using an
EDFA as a preamplifier.

JJ
II
J
I
Back
Close

Preamplification at Receiver
150/269

Sensitivity in terms of number of photons/bit:

Np = 1 Fn[Q2 + Q(2opt/B)1/2].
2

Best receiver sensitivity Np = 45 photons/bit.

Np exceeds 1000 without an optical amplifier.

JJ
II
J
I
Back
Close

Long-Haul Systems
151/269

Multiple amplifiers cascaded to overcome fiber losses.

Buildup of amplifier noise becomes the most critical factor.
ASE accumulates and degrades SNR as NA increases.
As the level of ASE grows, it saturates the amplifier and reduces its
gain.

JJ
II
J
I
Back
Close

Signal-to-Noise Ratio
Amplifiers are spaced apart by a constant distance LA.

152/269

Amplifier gain G exp(LA) just large enough to compensate for

fiber losses over a length LA.
Total ASE power for a chain of NA amplifiers:
Psp = 2NASspopt = 2nsph0NA(G 1)opt.
Factor of 2 accounts for the unpolarized nature of ASE.
Optical SNR: SNRopt = Pin/Psp.
Electrical SNR dominated by the signal-spontaneous beat noise:
R2Pin2
opt
SNRopt.
SNRel =
=
2
NAsigsp
2f

JJ
II
J
I
Back
Close

Optimum Amplifier Spacing

15000

153/269

System Length (km)

2 mW
1.0

10000

0.5
5000

0.2

20

40

60
Amplifier Spacing (km)

80

100

Large LA requires less amplifiers but more gain [G = exp(LA)].

Noting NA = LT /LA = LT / ln G, Psp scales with G as
(G 1)/ ln G.
SNR can be increased by lowering G.
Figure shows LT as a function of LA to maintain SNR = 20 dB.

JJ
II
J
I
Back
Close

ASE-Induced Timing Jitter

Amplifiers induce timing jitter by shifting pulses from their original
time slot in a random fashion.

154/269

This effect was first studied in 1986 and is known as

the GordonHaus jitter.
Spontaneous emission affects the phase and changes signal frequency by a random amount.
Group velocity depends on frequency because of dispersion.
Speed at which pulse propagates through the fiber is affected by
each amplifier in a random fashion.
Such random speed changes produce random shifts in the pulse
position at the receiver and leads to timing jitter.

JJ
II
J
I
Back
Close

Timing Jitter
10

Timing Jitter (ps)

0.3

155/269
0.2
0.1

0.01 ps /km

2000

4000
6000
Distance (km)

8000

10000

Timing jitter can be calculated using the moment method:



2
2
2
t = (Ssp/E0)Tm NA + NA(NAd + C0 + df ) ,
where d = 2LA/T02 = LA/LD .
T0, E0, and C0 are width, energy, and chirp of input pulses.
Large jitter for large average dispersion.
Can be reduced using the postcompensation technique.

JJ
II
J
I
Back
Close

Dispersive and Nonlinear Effects

156/269

Numerical simulations essential to account for dispersive and nonlinear effects.

Q factor depends on input power for long-haul systems (9000 km).
(a) 5-Gb/s system over 9000 km (D = 1ps/nm/km, LA = 40 km).
(b) Improvement realized with 150-GHz optical filters.
(c) 6-Gb/s operation near zero-dispersion wavelength.
(d) 10-Gb/s system with dispersion management.

JJ
II
J
I
Back
Close

Dispersion Management
Standard fibers have large dispersion near 1.55 m.
157/269

Operation near zero-dispersion wavelength not realistic

for WDM systems.
Transmission distance limited to
L < (16|2|B 2)1
even when DFB lasers are used.
L < 35 km at B =10 Gb/s for standard fibers with |2| 21 ps2/km).
Dispersion must be compensated or managed using a suitable technique.

JJ
II
J
I
Back
Close

Basic Idea
Pulse propagation in the linear case governed by
158/269

A i2 A
+
= 0.
z
2 t2
Using the Fourier-transform method, the solution is


Z
i
1
) exp
A(z, t) =
A(0,
2z 2 it d.
2
2
Phase factor exp(i2z 2/2) is the source of degradation.
A dispersion-management scheme cancels this phase factor.
Actual implementation can be carried out at the transmitter, at the
receiver, or along the fiber link.
Such a scheme works only if nonlinear effects are negligible.

JJ
II
J
I
Back
Close

Dispersion-Compensating Fibers
Fibers with opposite dispersion characteristics used.
159/269

Two-section map: D1L1 + D2L2 = 0.

Special dispersion-compensating fibers (DCFs) developed with D2
100 ps/(nm-km).
Required length L2 = D1L1/D2 (typically 5-10 km).
DCF modules inserted periodically along the link.
Each module introduces 56 dB losses that require compensation
and increase the noise level.
DCFs have a relatively small core diameter
(nonlinear effects enhanced).

JJ
II
J
I
Back
Close

Two-Mode DCFs
160/269

(a)

(b)

A new type of DCF uses a two-mode fiber (V > 2.405).

Long-period fiber gratings transfer power from one mode to another.
Dispersion for the higher-order mode can be as large as
500 ps/(km-nm).
Low losses and low nonlinearities for such DCFs.

JJ
II
J
I
Back
Close

Optical Filters
161/269

Optical field after the filter:



Z
i
1
)H() exp
A(0,
A(L, t) =
2L 2 it d,
2
2
Expanding the phase of H() in a Taylor series:
H() = |H()| exp[i()] |H()| exp[i(0 + 1 + 21 2 2)],
GVD totally compensated if 2 = 2L.
Requirements: |H()| = 1 and negligible higher-order terms.

JJ
II
J
I
Back
Close

Optical Filters (continued)

Simple example: FabryPerot interferometer.
162/269

GiresTournois interferometer: A FabryPerot whose back

mirror is 100% reflective. Its transfer function has the form
HFP() = H0

1 + r exp(iT )
,
1 + r exp(iT )

T =

2L
.
vg

In the vicinity of each peak, phase variation nearly quadratic

with 2 = 2T 2r(1 r)/(1 + r)3.
For a 2-cm-long FP cavity with r = 0.8, 2 2200 ps2.
Such a filter can compensate GVD acquired over 110 km.
Optical circulator required to separate the filtered signal.

JJ
II
J
I
Back
Close

MachZehnder Interferometer
163/269

A MZ interferometer constructed by connecting two

3-dB directional couplers in series.
First coupler splits input signal into two equal parts.
Different phase shifts acquired in the MZ arms.
Two fields interfere at the second coupler and may exit from either
or both output ports.
Transfer function for the bar port
HMZ() = 21 [1 + exp(i )].

JJ
II
J
I
Back
Close

MachZehnder Interferometer Chain

164/269

A cascaded chain of several MZ interferometers used in practice.

Fabricated in the form of a planar lightwave circuit using silica-onsilicone technology.
A chromium heater provides thermo-optic control of phase shift.

JJ
II
J
I
Back
Close

Fiber Bragg Gratings

Bragg gratings act as optical filters because of a stop band.
165/269

Light reflected back if its wavelengths falls within stop band.

Stop band centered at the Bragg wavelength: B = 2
n.
Grating period 0.5 m near 1.55 m.
A holographic technique is used for making Bragg gratings.
Use of gratings for dispersion compensation proposed in the 1980s.
Their use became practical after 1990.
Fiber gratings are available commercially and used routinely.

JJ
II
J
I
Back
Close

Coupled-Mode Equations
Refractive index varies along the length periodically as
n(z) = n
+ ng cos(2z/).

166/269

Index modulation depth ng 104.

Bragg gratings analyzed using coupled-mode equations
dAf /dz = +iAf + iAb,
dAb/dz = iAb iAf ,
where detuning =

2
0

2 and coupling coefficient =

B

ng
B .

Transfer function is found to be

i sin(qLg )
Ab(0)
H() = r() =
=
,
Af (0) q cos(qLg ) i sin(qLg )
where q 2 = 2 2 and Lg is the grating length.

JJ
II
J
I
Back
Close

Reflectivity Spectrum
1.0

167/269
0

0.6

Phase

Reflectivity

0.8

0.4

-10

0.2
0.0

-5

-10

-5

0
Detuning
(a)

10

-15

-10

-5

0
Detuning
(b)

10

Reflectivity |H()|2 and phase plotted for Lg = 2 and 3.

Grating-induced dispersion exists near stop-band edges.
Such gratings useful only in transmission.
Input signal tuned close to but outside the stop band.

JJ
II
J
I
Back
Close

Grating Dispersion
600

168/269
400

10 cm1

2g (ps2/cm)

200
0
-200
-400
-600
-20

-10

10

20

(cm1)

Using = B q and expanding in a Taylor series

2g

sgn()2/vg2
= 2
,
( 2)3/2

3g

3||2/vg3
= 2
.
( 2)5/2

JJ
II
J
I
Back
Close

Chirped Fiber Gratings

169/269

Period n
of a chirped grating not uniform over its length.
Bragg wavelength B = 2
n also varies along the grating.
Different spectral components of optical pulse reflected at different
points depending on where Bragg condition is satisfied locally.
The resulting stop band can be quite wide.

JJ
II
J
I
Back
Close

Normalized Reflectivity (dB)

Chirped Fiber Gratings

Time Delay (ps)

170/269

Wavelength (nm)

Group delay (d/d) for a 10-cm-long grating with

= 0.12 bandwidth.
Slope of group delay (5000 ps/nm) relateed to dispersion provided
the grating [Dg 5 107 ps/(km-nm)].
Grating period changes only 0.008% over the length.

JJ
II
J
I
Back
Close

Optical Phase Conjugation

171/269

Four-wave mixing used to generate phase-conjugated field

in the middle of fiber link.
2 reversed for the phase-conjugated field:
A i2 2A
+
=0
z
2 t2

A i2 2A

= 0.
z
2 t2

Pulse shape restored at the fiber end.

Basic idea patented in 1979.
First experimental demonstration in 1993.

JJ
II
J
I
Back
Close

Thory Behind Phase Conjugation

Consider a phase conjugator located at z = L/2.

) exp(i 22L/4).
Pulse spectrum just before: A(L/2,
) = A(0,

172/269

Pulse spectrum just after phase conjugation:

A(L/2, ) = A(0, ) exp(i 22L/4).
Spectrum inverted because c = 2p .
Optical field at the end of fiber link:




Z
i
1
L
A(L, t) =
A
, exp 2L 2 it d.
2
2
4

It is easy to see that A(L, t) = A (0, t).

Pulse shape restored to its input form irrespective of
how much pulse broadened in the first section.

JJ
II
J
I
Back
Close

Experimental Results
173/269

A long fiber used for OPC in a 1993 experiment.

Pump wavelength coincided with zero-dispersion wavelength.
Practical issues: Wavelength shift of OPC signal, polarization sensitivity, insertion losses, higher-order dispersion, etc.

JJ
II
J
I
Back
Close

Long-Haul Lightwave Systems

Without nonlinear effects, total GVD can be compensated once at
the receiver end.

174/269

This is not the case when nonlinear effects cannot be neglected.

Simple solution: periodic dispersion management.
Fibers with positive and negative GVD used in a periodic fashion
such that
= (D1L1 + D2L2)/Lm 0.
D
Dispersion map period Lm = L1 + L2 equal to amplifier spacing
LA in practice.
Lm = LA 80 km for terrestrial systems but reduced to about
50 km for submarine systems.

JJ
II
J
I
Back
Close

Recirculating Fiber Loop

175/269

A recirculating fiber loop used for long-haul experiments.

Loop length typically 300500 km. An optical bandpass filter inserted used to reduce amplifier noise.
10-Gb/s signal transmitted over 10,000 km of standard fiber.

JJ
II
J
I
Back
Close

Theory Behind Dispersion Maps

Propagation governed by the NLS equation:
176/269

A 2 A
i
2

+
|A|
A
=

A.
z
2 t2
2

2, , and are periodic functions of z.

Useful to eliminate the last term with the transformation


Z
1 z
A(z, t) = B(z, t) exp
(z) dz .
2 0
NLS Equation then takes the form
B 2(z) 2B
2
i

(z)|B|
B = 0,
z
2 t2
Rz
where (z) = exp[ 0 (z) dz].

JJ
II
J
I
Back
Close

Variational Method
Gaussian pulses maintain their functional form when = 0.
177/269

Since nonlinear effects are relatively weak locally, pulse shape is

likely to retain its Gaussian shape even when 6= 0.
Assume that each pulse evolves along the fiber as
B(z, t) = a exp[(1 + iC)t2/2T 2 + i],
All four pulse parameters (a, C, T, ) vary with z.
Variational method useful to find their z dependence.
NLS equation can be derived from the EulerLagrange equation
using the Lagrangian density:
"

2 #




i
B
1
B 
B
4

Lden =
B
B
+
(z)|B| 2(z) 
.
2
z
z
2
t 

JJ
II
J
I
Back
Close

Variational Method (continued)

Find reduced Lagrangian density by integrating over time.
178/269

Then, use the EulerLagrange equation to find ODEs

for a, T, C, and .
Amplitude equation eliminated using conservation of pulse energy.
Need to solve only the following two ODEs:
dT
2C
=
,
dz
T

E0
dC
2
=
+ (1 + C 2) 2 .
dz
T
2 T

Solution in the linear case ( = 0):

Z z
2Z z
1
+
C
0
T 2(z) = T02+2
2(z) dz.
2(z)C(z) dz, C(z) = C0+
2
T0
0
0
Numerical solution necessary in the nonlinear case.

JJ
II
J
I
Back
Close

Quasi-linear Propagation
A nonlinear system performs best when GVD compensation is only
9095% (some residual dispersion left after each map period).

179/269

CRZ format works best for dispersion-managed links.

Dispersion map made such that the pulse broadensin the first section
and compresses in the second section.
Pulse peak power reduced considerably in the first section because
of rapid broadening of chirped pulses.
Peak power lower in the second section because of accumulated
fiber losses.
Impact of nonlinear effects reduced significantly.
Such dispersion-managed links are called quasi-linear.

JJ
II
J
I
Back
Close

Broadband Dispersion Compensation

WDM systems require simultaneous compensation for all channels:
D1(n)L1 + D2(n)L2 = 0.

180/269

Dispersion slope of the DCF should satisfy

S2 = S1(L1/L2) = S1(D2/D1).
Ratio S/D should match for both sections.
Multiple fiber gratings can also be used.
RIANT et al.: CHIRPED FBGS FOR WDM CHROMATIC DISPERSION COMPENSATION

1317

JJ
II
J
I
Back
Fig. 12.

Experimental setup for system characterization of CDCs.

Close

Tunable Dispersion Compensation

Not all WDM channels can be compensated perfectly
by a single DCF.

181/269

Residual dispersion for each channel needs compensation at the

receiver (called postcompensation).
Precise amount of residual dispersion not known in practice
(dispersion variations along fiber length).
Dynamic variations can occur because of temperature
fluctuations.
Solution: Tunable dispersion compensation

JJ
II
J
I
Back
Close

Tunable Compensation (continued)

182/269

(a)

(b)

Fiber grating designed with a nonlinear chirp.

Fiber grating with a temperature gradient along its length;
segmented thin-film heater used in practice.

Z Lg
dg 2 d
n
(z) dz
=
Dg () =
d
c d
0

JJ
II
J
I
Back
Close

Tunable Compensation (continued)

183/269

A 160-Gb/s transmission experiment with a chirped fiber Bragg

grating performed in 2000.
Large power penally for a residual dispersion of >8 ps/nm.
Eliminated using tunable dispersion provided by the grating.

JJ
II
J
I
Back
Close

Third-Order Dispersion Management

184/269

Both second- and third-order dispersion require compensation when

short pulses are used at high bit rates:
21L1 + 22L2 = 0

and

31L1 + 32L2 = 0.

Design DCFs with a negative dispersion slope such that

32 = (22/21)31 = (L1/L2)31.

JJ
II
J
I
Back
Close

PMD Compensation
PMD-induced average pulse broadening:

185/269

T < (T )2 >1/2= Dp L
PMD control requires an outage probability < 105 (5 min/year)
or T < 0.1/B.

Dp L < 2.5 ps required at 40 Gb/s.

For L = 2500 km, Dp < 0.05 ps/ km.

PMD compensation necessary for such systems.

(a)

(b)

JJ
II
J
I
Back
Close

Tunable PMD Compensation

186/269

(a)

(b)

Chirped grating made in a polarization-maintaining fiber.

Stop bands shifted because of birefringence.
Differential delay used for PMD compensation.
Tunable over 5 nm by stretching the grating.

JJ
II
J
I
Back
Close

Multichannel Systems
187/269

Optical fibers offer a huge bandwidth (100 THz).

Single-channel bit rate limited to 40 Gb/s by electronics.
Solution: Wavelength-division multiplexing (WDM).
Many 10 or 40-Gb/s channels sent over the same fiber.

JJ
II
J
I
Back
Close

Point-to-Point WDM Links

Tx

188/269
Tx

Tx

Data from several transmitters multiplexed together.

A demultiplexer separates channels and feeds them into individual
receivers.
Channel spacing in the range 25100 GHz.
ITU grid specifies source wavelengths from 1530 to 1610 nm.

JJ
II
J
I
Back
Close

High-capacity Experiments
Channels Bit Rate
Capacity Distance N BL Product
N
B (Gb/s) N B (Tb/s) L (km) [(Pb/s)-km]
120
20
2.40
6200
14.88
132
20
2.64
120
0.317
160
20
3.20
1500
4.80
82
40
3.28
300
0.984
256
40
10.24
100
1.024
273
40
10.92
117
1.278
Capacity increased using C and L bands simultaneously.
C band = 15251565 nm; L band = 15701610 nm.
Other bands defined to cover 1.31.6 m range.
Total fiber capacity exceeds 30 Tb/s.

189/269

JJ
II
J
I
Back
Close

Wide-Area Networks
190/269

WANs cover a large geographical area.

Multiple interconnected rings provide fault tolerance.
Each node on the ring is a Metro network.

JJ
II
J
I
Back
Close

Metro-Area Networks
191/269

MANs cover a large metropolitan area.

A central ring connects multiple LANs at access nodes.
Connected to a WAN at egress nodes.

JJ
II
J
I
Back
Close

Local-Area Networks
Lambdanet Architecture
192/269

Passive Photonic Loop

JJ
II
J
I
Back
Close

WDM Components
Tunable Optical filters
193/269

Multiplexers and demultiplexers

Channel adddrop filters
Star couplers
WDM routers
Optical cross-connects
Wavelength converters
WDM transmitters
WDM receivers

JJ
II
J
I
Back
Close

Optical Filters
194/269

JJ
II
J
I
Back
Close

FabryPerot Filters
195/269

Periodic transmission peaks separated by L = c/(2ng L).

Can be tuned using a piezoelectric transducer.
Two conditions for single channel selection:
sig = N ch = N B/s < L,
s = spectral efficiency;

FP = L/F > B.

F = Finesse.

Number of channels N < s(L/FP) = sF .

JJ
II
J
I
Back
Close

MachZehnder Filters
MZ interferometer with extra delay in one of its arms.
196/269

Transmissivity T () = |H()|2 = cos2( ).

Such a filter is not sharp enough for channel selection.
Transmittivity of a chain of M such interferometers:
T () =

M
Y

cos2(m).

m=1

When m = (2mch)1, each MZ stage blocks alternate channels.

A 10-stage MZ chain is as good as a FP filter with F = 1600.
Silica-on-silicon technology used to make such filters.

JJ
II
J
I
Back
Close

Fiber Bragg Gratings

A fiber grating acts as a reflection filter and will reflect any channel
that falls inside its stop band.

197/269

Bragg wavelength can be tuned thermally or by stretching

the grating.
Fiber gratings require the use of an optical circulator.
A phase shift in the middle of grating can convert it into
a transmission filter.
Gratings can be used in combination with FP, MZ, and
Michelson interferometers.
Gratings can also be made using InGaAsP/InP technology.
Tunable filters can be designed using a DBR structure.

JJ
II
J
I
Back
Close

Acouto-Optic Filters
Grating is formed dynamically using acoustic waves.
198/269

Such filters exhibit a wide tuning range (> 100 nm).

Physical mechanism: TE-to-TM mode conversion by the
index grating created by the acoustic wave.
Tuning realized by changing the acoustic-wave frequency.
LiNbO3 waveguide technology used to produce compact,
polarization-independent acousto-optic filters.
WDM signal first split into its orthogonally polarized components.
Channel whose wavelength satisfies the Bragg condition
directed to a different output port.

JJ
II
J
I
Back
Close

Multiplexers/Demultiplexers
199/269

JJ
II
J
I
Back
Close

Waveguide-Grating Demultiplexers
200/269

A phased array of optical waveguides acts as a grating.

Such arrayed waveguide gratings can be fabricated using silicon,
InP, or LiNbO3 technology.
Waveguides lengths change by a constant amount.
Different channels focus to different output waveguides.

JJ
II
J
I
Back
Close

AddDrop Multiplexers
201/269

Generic design provides flexibility but is expensive.

All-optical solution: MachZehnder with Bragg gratings.

JJ
II
J
I
Back
Close

AddDrop Multiplexers
Optical circulators can be used in combination with fiber gratings.
202/269

Such a device can be made by connecting a fiber grating

to two 3-port optical circulators.
Channel reflected by the grating is dropped at the unused port of
the input-end circulator.
IEEE PHOTONICS TECHNOLOGY LETTERS, VOL. 13, NO. 6, JUNE 2001

AddDrop Multiplexers with Low Crosstalk

Device can also be made using only one circulator provided it has
more than three ports.

ent Member, IEEE, Wen De Zhong, Member, IEEE, Rodney C. Tucker, Fellow, IEEE, and
Richard Lauder, Member, IEEE

e and demonstrate three new optical

OADM) architectures with simple and
nd excellent crosstalk performance. The
iport optical circulator and either one or
ragg gratings, depending on the crosstalk

tor, crosstalk, fiber grating, optical

avelength-division multiplexing.

INTRODUCTION

ddrop multiplexer (OADM) is a key

elength-division multiplexing (WDM).
ssue for OADM design is the reduction

JJ
II
J
I
Back
Close

Star Couplers
203/269

JJ
II
J
I
Back
Close

WDM routers
204/269

Waveguide-grating router:
An array of optical waveguides between two star couplers

JJ
II
J
I
Back
Close

Optical Cross-Connects
205/269

Optical analog of electronic crossbar switches.

OXCs demultiplex all WDM channels and use optical switches to
make connections.

JJ
II
J
I
Back
Close

Optical Switching
206/269

Several technologies can be used for making optical switches.

Micro-electro-mechanical-system (MEMS) mirrors
Liquid crystals (spatial light modulators, flat-panel displays)
Thermo-optic MachZehnder switches (silica-on-silicon)

JJ
II
J
I
Back
Close

Optical Switching
207/269

(a)

(b)

Y-junction switches with semiconductor optical amplifiers

Directional-couplers switches (LiNbO3 waveguides)
Total internal reflection at air bubbles (bubble-jet technology).
Electroholographic switches (LiNbO3 crystal)

JJ
II
J
I
Back
Close

Wavelength Converters
208/269

JJ
II
J
I
Back
Close

Wavelength Converters
Optoelectronic regenerators: Covert data into electric domain and use
a transmitter at the desired wavelength.

209/269

All-Optical Techniques
Cross-gain saturation in a semiconductor optical amplifier (SOA).
Cross-absorption saturation (Electro-absorption modulators).
SOA-induced phase shift in a Mach-Zehnder interferometer.
Four-wave mixing inside a semiconductor optical amplifier.
Four-wave mixing inside an optical fiber.

JJ
II
J
I
Back
Close

WDM Transmitters and Receivers

210/269

Waveguide grating integrated with a wideband amplifier.

Spectral slicing of a wideband source.
Supercontinuum generation using nonlinear effects.
DFB-laser arrays integrated on the same chip.

JJ
II
J
I
Back
Close

Crosstalk in WDM Systems

System performance degrades whenever power from one channel
leaks into another.

211/269

Such a power transfer can occur because of the nonlinear effects in

optical fibers (nonlinear crosstalk).
Crosstalk occurs even in a perfectly linear channel because of imperfections in WDM components.
Linear crosstalk can be classified into two categories.
Heterowavelength or Out-of-band crosstalk: Leaked power is at a
different wavelength from the channel wavelength.
Homowavelength or In-band crosstalk: Leaked power is at the same
wavelength as the channel wavelength.

JJ
II
J
I
Back
Close

Out-of-Band Linear Crosstalk

Introduced by an optical filter if its passband leaks other channels.
P
Power after filter: P = Pm + N
n6=m Tmn Pn .
P
Photocurrent at receiver: I = RmPm + N
n6=m Rn Tmn Pn Ich +
IX .


PN
R
T
P
n mn n
Crosstalk penalty: X = 10 log10 1 + n6=mRmPm
.

212/269

Assuming equal powers and same responsivities,

X 10 log10(1 + X).
PN

X = n6=m Tmn represents the fraction of total power leaked into

a specific channel from all other channels.

JJ
II
J
I
Back
Close

Crosstalk for FP Filters

213/269

Penalty depends on FP finesse

F and the number N of channels.
To keep penalty below 0.2 dB,
N/F < 0.33.
Channel spacing > 3B at
bit rate B.
JJ
II
J
I
Back
Close

In-band Linear Crosstalk

Occurs for WDM components used for routing and switching.
214/269

In a waveguide-grating router, N 2 1 interfering signals can accompany the desired signal.

Among these N 1 have the same carrier wavelength m and can
induce in-band crosstalk.
Because of coherent nature of interfering
signals, one must add


PN
fields: Em(t) = Em + n6=m En exp(imt).
Receiver current:
I(t) RPm(t) + 2R

N p
X
n6=m

Pm(t)Pn(t) cos[m(t) n(t)].

JJ
II
J
I
Back
Close

In-band Crosstalk (continued)

Since phases are likely to fluctuate randomly, I(t) = R(Pm + P ),
where P is a random quantity.

215/269

Treating the crosstalk as intensity noise, crosstalk penalty

2
X = 10 log10(1 rX
Q2),
2
where rX
= h(P )2i/Pm2 = X(N 1),

X = Pn/Pm is the fraction of power leaking through the WGR

assuming equal channel powers.
To keep power penalty below 2 dB, rX < 0.07 is required.
This condition limits X(N 1) to below 23 dB.
crosstalk level X must be below 38 dB for N = 16 and below
43 dB for N = 100 (quite stringent requirement).

JJ
II
J
I
Back
Close

Nonlinear Raman Crosstalk

SRS not of concern for single-channel systems because of its
high threshold (about 500 mW).

216/269

In the case of WDM systems, fiber acts as a Raman amplifier.

Long-wavelength channels amplified by short-wavelength channels.
Power transfer depends on the bit pattern: amplification occurs
only when 1 bits are present in both channels simultaneously.
SRS induces power fluctuations (noise) in all channels.
Shortest-wavelength channel most depleted.
One can estimate Raman crosstalk from noise level of this channel.

JJ
II
J
I
Back
Close

Nonlinear Raman Crosstalk

5.0

Power Penalty

217/269

10 mW

4.0

5
3.0

2
2.0

1.0
0.0

20

40
60
Number of Channels

80

100

Power penalty R = 10 log(1 DR ), where

DR =

M
X

gR (m)PchLeff /Aeff ,

CR = SR ch/(2Aeff ).

m=2

Slope of Raman gain SR = 4.9 1018 m/(W-GHz).

Pch should be below 1 mW when M > 70.

JJ
II
J
I
Back
Close

Stimulated Brillouin Scattering

218/269

SBS does not induce interchannel crosstalk when all channels propagate in the forward direction.
SBS limits the launched power of each channel.
SBS threshold condition gB PthLeff /Aeff 21 implies Pch should be
below 5 mW or so.

JJ
II
J
I
Back
Close

Cross-Phase Modulation
219/269

XPM-induced phase shift depends on bit pattern of channels.

Dispersion converts pattern-dependent phase shifts into power fluctuations (noise). Probe power fluctuates by as much as 6% after
320 km.
Level of fluctuations depends on channel powers.

JJ
II
J
I
Back
Close

Four-Wave Mixing
FWM generates new waves at frequencies ijk = i + j k .

220/269

In the case of equally spaced channels, new frequencies coincide

with the existing frequencies and produce in-band crosstalk.
Coherent crosstalk is unacceptable for WDM systems.
In the case of nonuniform channel spacing, most FWM components
fall in between the channels and produce out-of-band crosstalk.
Nonuniform channel spacing not practical because many WDM
components require equal channel spacings.
A practical solution offered by the periodic dispersion management
technique.
GVD high locally but its average value is quite low.

JJ
II
J
I
Back
Close

Time-Division Multiplexing
221/269

Multiple channels interleaved in time domain.

Optical TDM (OTDM) requires RZ format.
Optical pulses much shorter than bit slot.
OTDM transmitter:
A mode-locked pulse train at the single-channel bit rate is split into
many branches and recombined after suitable delays.

JJ
II
J
I
Back
Close

OTDM Demultiplexers
222/269

JJ
II
J
I
Back
Close

Electro-Optic Technique
Electro-optic technique uses multiple LiNbO3 modulators in series.

223/269

Each modulator halves the bit rate by rejecting alternate bits in the
incoming signal.
An 8-channel OTDM system requires three modulators, each
driven by the same electrical clock but different voltages.
Require voltages are 4V0, 2V0, and V0, where V0 produces
phase shift in one arm of MZ interferometer.
Different channels can be selected by changing the phase of clock
signal.
Limited by the speed of modulators, Requires expensive components
and high drive voltage.

JJ
II
J
I
Back
Close

Nonlinear Optical Loop Mirror

Sagnac interferometer constructed using a fiber loop and
a 3-dB fiber coupler.

224/269

NOLM reflects its input entirely when counterpropagating waves

experience the same phase shift over one round trip.
With a relative phase shift of , signal is fully transmitted.
Clock signal propagates only in the clockwise direction and
introduces phase shift on pulses belonging to a specific channel.
This single channel is demultiplexed by the NOLM.
Multiple channels can be demultiplexed simultaneously using
several NOLMs in parallel.
By 1998, NOLM was used to demultiplex a 640-Gb/s OTDM signal.

JJ
II
J
I
Back
Close

Four-Wave Mixing
Uses FWM in a nonlinear medium (similar to
wavelength conversion).

225/269

Clock signal plays the role of the pump for the FWM process.
In time slots in which a clock pulse overlaps with 1 bits of the
selected channel, FWM produces a pulse at the idler wavelength.
Pulse train at the idler wavelength is an exact replica of the
selected channel.
An optical filter separates the demultiplexed channel.
A polarization-preserving fiber used as the nonlinear medium.
By 1996, error-free demultiplexing of 10-Gb/s channels from a 500Gb/s OTDM signal was demonstrated.

JJ
II
J
I
Back
Close

Soliton Systems
Use a nonlinear effect (SPM) to eliminate dispersion-induced pulse
broadening.

226/269

Solitons maintain their shape by balancing the dispersive and nonlinear effects.
Ultimate solution of the dispersion problem.

Soliton systems require RZ format and mode-locked lasers.

Their launched power depends on the pulse width.
New concepts of dispersion-managed solitons quite useful for system
design.

JJ
II
J
I
Back
Close

Fiber Solitons
Combination of SPM and anomalous GVD required.
227/269

GVD broadens optical pulses except when the pulse is initially chirped
such that 2 < 0.
SPM imposes a chirp on the optical pulse such that C > 0.
Soliton formation possible only when 2 < 0.
SPM-induced chirp is power dependent.
SPM and GVD can cooperate when input power is adjusted such
that SPM-induced chirp just cancels GVD-induced broadening.
Nonlinear Schrodinger Equation
A 2 2A
2
i

+
|A|
A = 0.
z
2 t2

JJ
II
J
I
Back
Close

Bright Solitons

Normalized variables = z/Ld, = t/T0, and U = A/ P0 lead

to
U 1 2U

i
+ N 2|U |2U = 0.
2

228/269

Its solution depends on a single parameter N defined as

N 2 = LD = LD /LNL = P0T02/|2|.
Dispersive and nonlinear lengths LD =

T02
|2 | ,

LNL =

1
P0 .

The two are balanced when LNL = LD or N = 1.

NLS equation can be solved exactly with the inverse scattering
method.

JJ
II
J
I
Back
Close

Soliton Shape and Chirp

229/269

Temporal and chirp profiles for N = 1 and N = 3.

Fundamental shape-preserving solitons for N = 1.
Periodic evolution for higher-order solitons (N > 1).

JJ
II
J
I
Back
Close

Fundamental Soliton Solution

For fundamental solitons, NLS equation becomes
230/269

u 1 u
+
+ |u|2u = 0.
2
2

Assuming u(, ) = V ( ) exp[i()], V satisfies

V 2).

d2 V
d 2

= 2V (K

Multiplying by 2 (dV /d ) and integrating over ,

(dV /d )2 = 2KV 2 V 4 + C.
C = 0 from the boundary condition V 0 as | | .
Constant K =

1
2

using V = 1 and dV /d = 0 at = 0.

Final Solution: u(, ) = sech( ) exp(i/2).

JJ
II
J
I
Back
Close

Stability of Fundamental Solitons

Very stable; can be excited using any pulse shape.
231/269

Evolution of a Gaussian pulse with N = 1:

Nonlinear index n = n2I(t) larger near the pulse center.

Temporal mode of a SPM-induced waveguide.

JJ
II
J
I
Back
Close

Stability of Solitons
Peak Power and pulse width require matching for N = 1.
232/269

Evolution of a sech pulse with N = 1.2:

Pulse changes its width to make N = 1.

Temporal mode of a SPM-induced waveguide.

JJ
II
J
I
Back
Close

Dark Solitons
In the case of normal dispersion, NLS equation becomes:
233/269

u 1 2u
2
+
|u|
u = 0.
i
2 2
This equation does not have pulse-like soliton solutions.
Shape-maintaining solitons exist but appear as a dip in a constant
background; called dark solitons.
General solution: ud(, ) = (tanh i) exp(iu20), where
= ( ),

= u0 cos ,

= u0 sin .

u0 is amplitude of CW background and is any angle in the range

0 to /2.

JJ
II
J
I
Back
Close

Dark Solitons (continued)

234/269

For = 0, ud(, ) = u0 tanh(u0 ) exp(iu20).

black soliton: Peak power drops to zero at the center.
For 6= 0, soliton is referred to as a gray soliton.
Phase of a dark soliton changes across its width (chirping).

JJ
II
J
I
Back
Close

Information Transmission with Solitons

235/269

Basic Idea: Use bright solitons as 1 bits.

Requires RZ format and a short-pulse optical source.
Solitons should be well separated (T0  TB ).
Using 2q0 = TB /T0, bit rate B is related to T0 as
1
1
B=
=
.
TB 2q0T0
For N = 1, pulse width and peak power are related as
P0 = |2|/( T02).

JJ
II
J
I
Back
Close

Input Signal properties

Pulse shape: P (t) = P0 sech2(t/T0).
T0 is related to FWHM as Ts = 2T0 ln(1 +
R
Pulse energy Es = P (t) dt = 2P0T0.

236/269

2) ' 1.763T0.

Assuming 1 and 0 bits are equally likely to occur, average power of

the RZ signal is Ps = Es(B/2) = P0/2q0.
T0 = 10 ps for a 10-Gb/s soliton system if we choose q0 = 5.
Peak power of input pulse is 5 mW using 2 = 1 ps2/km and
= 2 W1/km.
Pulse energy = 0.1 pJ and average power of the bit stream is only
0.5 mW.

JJ
II
J
I
Back
Close

Soliton Interaction
237/269

Two neighboring solitons interact through XPM.

Interaction depends strongly on their spacing.
It also depends on amplitude ratio r and relative phase .

JJ
II
J
I
Back
Close

Loss-Managed Solitons
Fiber losses destroy the balance needed for solitons.
238/269

Soliton energy and peak power decrease along the fiber.

Nonlinear effects become weaker and cannot balance dispersion
completely.
Pulse width begins to increase.
Solution: Compensate losses periodically using amplifiers.
Solitons sustained through periodic amplification are called
loss-managed solitons.
They are launched with a higher energy.

JJ
II
J
I
Back
Close

Loss-Induced Soliton Broadening

NLS equation with loss:
239/269

u 1 u
i
2
+
+
|u|
u
=

u,
2 2
2

If = LD  1, last term treated as a small perturbation.

Variational or perturbation solution:
u(, ) e sech( e ) exp[i(1 e2 )/4].
Soliton width increases exponentially as
T1() = T0 exp() = T0 exp(z).
An exponential increase cannot be expected to continue.
Numerical solutions show slower increase for  1.

JJ
II
J
I
Back
Close

Loss-Induced Soliton Broadening

240/269

Perturbative result reasonably accurate up to = 1.

soliton broadening much less compared with the linear case.
Nonlinear effects beneficial even when solitons cannot be maintained
perfectly because of fiber losses.

JJ
II
J
I
Back
Close

Soliton Amplification
241/269

Optical amplification necessary for long-haul systems.

Lumped amplifiers placed periodically along the link.
Distributed Raman amplification is a better alternative.

JJ
II
J
I
Back
Close

Lumped Amplification
Loss Management: Match the average value of peak power to the
fundamental soliton power P0:

242/269

Z LA
1
PineLA dz = P0.
LA 0
Loss Enhancement Factor
Pin
LA
G ln G
=
=
,
P0
1 eLA
G1
where G = exp(LA) is the amplifier gain.
fLM =

Launch power larger by a factor of fLM.

Typically G = 10 for LA = 40 km, and fLM = 2.56.

JJ
II
J
I
Back
Close

Loss-Managed Solitons
Peak power drops by a factor eLA for an amplifier spacing of LA.

243/269

For LA = 80 km, power drops by a factor of about 100.

soliton width will increase considerably unless LD  LA.
LA should be a small fraction of LD for lumped amplification to
work.

(a)

(b)

JJ
II
J
I
Back
Close

Design Constraints
Soliton interaction should be avoided by ensuring that tails of two
neighboring solitons do not overlap.

244/269

Soliton width should be a small fraction of the bit slot.

Amplifier spacing should be a small fraction of dispersion length.
p
2
LA  LD = T0 /|2|
=
T0  |2|LA.
For 2 = 1 ps2/km and LA = 80 km, T0  9 ps.
Limits the bit rate to below 20 Gb/s for LA > 40 km.
Smaller amplifier spacing impractical for economical reasons.

JJ
II
J
I
Back
Close

Distributed Amplification
1.2

245/269

Normalized Energy

1.0
0.8
0.6
0.4
0.2
0.0

10

20
30
Distance (km)

40

50

Local gain g(z) 6= because of pump losses.

Peak power evolves as

dp
dz

= [g(z) ]p
R LA
distributed gain must satisfy 0 g(z) dz = LA.
Soliton power variations smallest for bidirectional pumping.

JJ
II
J
I
Back
Close

Dispersion Management
Local GVD should be high to avoid FWM.
246/269

Dispersion-managed links employ two types of fibers with opposite

GVD (normal and anomalous).
Local dispersion high but average GVD is kept low.
2ala + 2nln
2 =
la + ln
Such periodic dispersion maps support new type of solitons
called dispersion-managed solitons.
DM solitons change their width within each map period Lmap =
la + ln.
They evolve in a periodic fashion with the period Lmap.

JJ
II
J
I
Back
Close

Dispersion-Managed solitons
Nonlinear Schrodinger Equation
247/269

B 2(z) 2B
2

+
p(z)|B|
B = 0.
i
z
2 t2
2(z) is a periodic function with period Lmap.
p(z) accounts for loss-induced power variations; it varies periodically
with the period LA.
LA = mLmap, where m is an integer.
Often LA = LD (m = 1) in practice.
Dense dispersion management needed at high bit rates (m  1).
DM solitons are solutions of the modified NLS equation.

JJ
II
J
I
Back
Close

Variational Equations
Gaussian Ansatz: B(z, t) = a exp[(1 + iC)t2/2T 2 + i]
248/269

Pulse parameters a, C, T , and evolve with z periodically.

Pulse energy E0 = a2(z)T (z) does not change.

Pulse width T and chirp C satisfy
dT
C
= 2(z) ,
dz
T
dC E0p(z) 2
=
+ 2 (1 + C 2).
dz
T
2T
Solve with periodic boundary conditions
T (0) = T (LA),

C(0) = C(LA).

JJ
II
J
I
Back
Close

DM Solitons
20

Input chirp

Pulse width (ps)

15

249/269

2
1
0
0.0

10

0.5

1.0

Input energy (pJ)

T0
5

Tm
0
0.001

0.010

0.100

1.000

Input energy (pJ)

(a)

(b)

Dispersion Map: 2 = 4 ps2/km; la ln = 5 km; 2 = 0.01 ps2/km;

LA = 80 km; = 0.2 dB/km.
For a given dispersion map, input pulse width T0 and chirp fixed but
vary with pulse energy; Tm = minimum pulse width.

JJ
II
J
I
Back
Close

-1

-1

-1

-1

-2
10

-2

-2

Distance (km)
(a)

250/269

Chirp

Pulse width (ps)

Chirp

Pulse width (ps)

DM Solitons (continued)

-2
10

Distance (km)
(b)

Pulse width and chirp of DM solitons for two pulse energies.

Pulse width minimum where chirp vanishes.
Shortest pulse occurs in the middle of anomalous-GVD section.
DM soliton does not maintain its chirp, width, or peak power.

JJ
II
J
I
Back
Close

Approximate Solution
Find T0 and Tm solving variational equations approximately.

251/269

Width equation can be integrated to relate T and C as

Z z
T 2(z) = T02 + 2
2(z)C(z) dz.
0

Chirp equation cannot be integrated but numerical solutions show

that C(z) varies almost linearly from C0 to C0.
Ratio (1+C 2)/T 2 related to pulse spectral width that changes little
over one map period.
Averaging it over one map period:
s

1/2
2
1 + C0
|2n2alnla|
T0 = Tmap
,
Tmap =
.
|C0|
2nln 2ala

JJ
II
J
I
Back
Close

Design Issues
Tmap is a useful design parameter with dimensions of time involving
only four map parameters.

252/269

It provides a time scale associated with an arbitrary dispersion map.

Minimum
value of T0 occurs for |C0| = 1 and is given by

T0min = 2Tmap.
Shortest pulse occurs at the chirpfree point:
p
2 1/2
Tm = T0/(1 + C0 ) = Tmap/ |C0| Tmap.

Optimum value of pulse stretching factor is 2.

The choice T0 = 2Tmap and C0 = 1 is optimum from a system

standpoint.

JJ
II
J
I
Back
Close

Benefits of DM Solitons
Energy enhancement factor for DM solitons: fDM = E0DM/E0av .

253/269

Its value can exceed 10 depending on the system design.

Higher soliton energy improves SNR and reduces timing jitter.
A large number of experiments have shown benefits of DM solitons.
A 10-Gb/s experiment transmitted signals over 28,000 km using a
dispersion map consisting of 100 km of normal-GVD fiber and 8-km
of anomalous-GVD fiber with a negative value of 2av .
In a 1998 experiment, the loop was modified to make average-GVD
zero or slightly positive.
Experimental results in agreement with numerical simulations.

JJ
II
J
I
Back
Close

Impact of Amplifier Noise

Amplifiers add spontaneous-emission noise.
254/269

It affects amplitude, frequency, and phase of solitons.

Amplitude noise degrades the SNR.
Frequency fluctuations change soliton speed.
Random changes in speed produce jitter in arrival time.
Both SNR degradation and timing jitter can be calculated using the
moment method:
Z
Z
1
E=
t|B|2dt,
|B|2dt,
q=
E


Z 

i
B
B
=
B
B
dt.
2E
t
t

JJ
II
J
I
Back
Close

Timing Jitter
Timing jitter (ps)

255/269

Standard Soliton

6
4
2
0
0

DM Soliton

2000

4000
Distance (km)

6000

8000

Jitter Variance for DM Solitons:

t2

SspTm2
=
[NA(1+C02)+NA(NA 1)C0d+ 61 NA(NA 1)(2NA 1)d2]
E0
Z LA
1
2LA
d= 2
2(z) dz =
,
Ssp = nsph0(G 1).
Tm 0
Tm2

JJ
II
J
I
Back
Close

Timing Jitter (continued)

Timing jitter limits the system performance.
256/269

It should be a small fraction of the bit slot.

If bj is the tolerable fraction of the bit slot, Bit RateDistance
Limitation has the form
!1/3
2
9bj fLMLA
.
BLT <
Sspq0 2
For a standard soliton system with parameters q0 = 5, = 0.2 dB/km,
= 2 W1/km, 2 = 1 ps/(km-nm), nsp = 1.5, LA = 80 km,
and bj = 0.1, BLT must be below 132 (Tb/s)-km.
For a 40-Gb/s system, LT is limited to below 3300 km.
This value can be increased to 10,000 km for DM solitons.

JJ
II
J
I
Back
Close

Control of Timing Jitter

257/269

Optical filters (FabryPerot or other kinds)

Better control with sliding-frequency filters.
Synchronous modulation (LiNbO3 modulators)
Fast saturable absorbers, nonlinear optical-loop mirrors
Other nonlinear effects (XPM, FWM, SRS)

JJ
II
J
I
Back
Close

High-Speed Soliton Systems

Main design parameters are: GVD of each fiber section, map period
Lmap, and amplifier spacing LA.

258/269

Parameter Tmap sets the scale for the shortest pulse that can be
propagated in a periodic fashion.
For terrestrial systems, map consists of 6070 km of standard fiber
and 1015 km of dispersion-compensating fiber.
Tmap exceeds 25 ps for such maps.
Difficult to realize a bit rate of 40 Gb/s for such systems.
For B 10 Gb/s systems, Tmap should be reduced to below 10 ps.
Map period should be less than LA (dense DM).

JJ
II
J
I
Back
Close

Dense Dispersion Management

259/269

Normalized Power

1
0.8
0.6
0.4
0.2
0

60

40

20

0
20
Time (ps)

40

60

8000
6000 )
km
4000
e(
nc
2000
a
t
Dis

Evolution of a DM soliton over 8000 km using dense DM

(Lmap = LA/8).
Dispersion map: 2 = 4 ps2/km, la ln = 5 km, 2av = 0.01
ps2/km, Lmap = 10 km, LA = 80 km, Tmap = 3.2 ps, T0 = 5 ps.

JJ
II
J
I
Back
Close

Soliton Interaction Problem

Distance (km)

8000
6000

260/269

4000
2000
0
-200

-150

-100

-50

50

100

150

200

-200

-150

-100

-50

50

100

150

200

Distance (km)

8000
6000
4000
2000
0

Time (ps)

Left: Pulse energy 100 fJ; Right: Optimum value (10 fJ)
Interaction can be minimized by optimizing dispersion map.
Intrachannel polarization multiplexing is also helpful.

JJ
II
J
I
Back
Close

Polarization Multiplexing
261/269
llel

Neighboring bits are orthogonally polarized.

Improves system performance by reducing soliton interaction.
Nonlinear interaction weaker for orthogonally polarized solitons.
Helps to increase bit rate as solitons can be packed more tightly.

JJ
II
J
I
Back
Close

Higher-Order Effects
Third-order dispersion and Intrapulse Raman scattering
B 2(z) 2B
i3 3B
|B|2
2
i

.
+ p(z)|B| B =
+ TR p(z)B
z
2 t2
6 t3
t

262/269

Raman parameter TR 3 fs, 3 0.1 ps3/km.

Raman term leads to spectral shift (soliton self-frequency shift).
Solitons travel slower because of spectral red shift.

JJ
II
J
I
Back
Close

Raman-induced Timing Jitter

Timing jitter (ps)

263/269

4
Raman

Total

GordonHaus

1
0
0

1000

2000

Distance (Km)

3000

4000

Raman-induced frequency shift depends on pulse energy.

Amplifier noise induced pulse-energy fluctuations.
Raman effects converts them into frequency fluctuations.
GVD translates frequency fluctuations into timing jitter.
Results for a 160-Gb/s system designed using 1.25-ps pulses.

JJ
II
J
I
Back
Close

Acoustic Timing Jitter

264/269

(a)

(b)

Intensity gradient of fiber mode creates acoustic waves traveling in

the radial direction through electrostriction.
Acoustic waves produce time-dependent index changes.
Index changes produced by one pulse affect the speed of many following pulses (number set by phonon decay time).
Random nature of bit pattern produces timing jitter.

JJ
II
J
I
Back
Close

Control of Timing Jitter

Optical phase conjugation is quite effective in reducing timing jitter.
265/269

Rquires use of parametric amplifiers in place of EDFAs.

For the idler wave, A changes to A (phase conjugation).
Equivalent to inverting pulse spectrum.
In the presence of parametric amplifiers, GordonHaus contribution
scales as NA rather than NA3 .
At the same time, Raman-induced jitter scales as NA rather than
NA5 .
OPC does not compensate for third-order dispersion.

JJ
II
J
I
Back
Close

Timing Jitter with TOD

266/269

For 3 = 0.05 ps3/km (typical value), distance is limited by TOD

to below 1500 km.
Transmission over 7500 km is possible for 3 = 0.
Distance can be increased further for slightly negative values of 3.
Compensation of both 2 and 3 necessary at high bit rates.

JJ
II
J
I
Back
Close

WDM Soliton Systems

Design similar to regular WDM systems.
267/269

Dispersion map used for suppressing FWM.

New feature: XPM interaction among solitons belonging
to different channels.
Solitons in different WDM channels travel at different speeds.
When pulses overlap temporally (collision), they interact
nonlinearly through XPM.
Soliton frequency shifts during a collision.
No net frequency shift for a symmetric collision.

JJ
II
J
I
Back
Close

Interchannel Collisions
268/269

(a)

(b)

Fiber losses destroys symmetric nature of soliton collisions.

Residual frequency shift 50 MHz depending on the collision length
Lcoll = (V )(2Ts), where V = (|2|ch/T0)1.
Residual frequency shift negligible when Lcoll > 2LA.

JJ
II
J
I
Back
Close

Experimental Progress
A recirculating loop is often used in practice.
269/269

Periodic amplification of solitons using amplifiers.

Dispersion management essential for WDM systems.
By 2000, eight 20-Gb/s channels over 10,000 km using optical filters
and synchronous modulators inside a 250-km loop.
640-Gb/s Experiment: 16 channels at 40 Gb/s transmitted over
1000 km in a 250-km loop (Suzuki et al., EL, 2000).
2.56-Tb/s Experiment: 32 channels at 80 Gb/s transmitted over
120 km (Lee et al., EL, 2001).
2002 Field Trial: 63 channels at 40 Gb/s transmitted over 320 km
of installed fiber (Otani et al., OFC, 2002).

JJ
II
J
I
Back
Close