Are you sure?
This action might not be possible to undo. Are you sure you want to continue?
DigitalCommons@McMaster
Open Access Dissertations and Theses Open Dissertations and Theses
1012010
AllOptical Mult~hop FreeSpace Optical
Communication Systems
Shabnam Kazemlou
This Thesis is brought to you for free and open access by the Open Dissertations and Theses at DigitalCommons@McMaster. It has been accepted for
inclusion in Open Access Dissertations and Theses by an authorized administrator of DigitalCommons@McMaster. For more information, please
contact scom@mcmaster.ca.
Recommended Citation
Kazemlou, Shabnam, "AllOptical Mult~hop FreeSpace Optical Communication Systems" (2010). Open Access Dissertations and
Theses. Paper 4185.
http://digitalcommons.mcmaster.ca/opendissertations/4185
AllOptical Multihop FreeSpace Optical
Communication Systems
ALLOPTICAL lVIULTIHOP FREESPACE OPTICAL
COMMUNICATION SYSTElVIS
BY
SHABNAM KAZEMLOD, B.Sc.
A THESIS
SUBMITTED TO THE DEPARTMENT OF ELECTRICAL & COMPUTER ENGINEERING
AND THE SCHOOL OF GRADUATE STUDIES
OF MCMASTER UNIVERSITY
IN PARTIAL FULFILMENT OF THE REQUIREMENTS
FOR THE DEGREE OF
MASTER OF ApPLIED SCIENCE
© Copyright by Shabnam Kazemlou, October 2010
All Rights Reserved
Master of Applied Science (2010)
(Electrical & Computer Engineering)
McMaster University
Hamilton, Ontario, Canada
TITLE:
AUTHOR:
SUPERVISOR:
AllOptical FreeSpace Optical Communication
Systems
Shabnam Kazemlou
B .Sc., (Electrical Engineering)
Sharif University of Technology, Tehran, Iran
Dr. Steve Hranilovic
Dr. Shiva Kumar
NUMBER OF PAGES: xvi, 147
11
To those WllO were always tllere wllen I needed tllem
Abstract
FreeSpace Optical (FSO) communication systems have recently attracted consider
able attention in lastmile applications. High bandwidth, unlicensed spectrum, ease
of installation, and high security have made them a good candidate for high data rate
transmissions. However, distancedependent atmospheric turbulence and channel loss
degrade the optical link reliability and confine FSO systems to shorthaul applica
tions. This thesis addresses innovative alloptical relaying techniques to mitigate the
degrading effects of atmospheric turbulenceinduced fading by relaying data from the
source to the destination using intermediate terminals. The proposed techniques, op
tical amplifyandforward (OAF) relaying and optical regenerateandforward (ORF)
relaying, are deployed in multihop FSO systems to extend the nlaximum accessible
communicating distance of high data rate wireless optical systems.
In alloptical relaying techniques, photo detection is performed once at the re
ceiver and intermediate terminals process optical field envelopes instead of optical
intensities. This major difference requires a new definition of channel model for prop
agation of optical waves through the atmosphere. By using the developed channel
model, bit error rate (BER) performance of multihop OAF FSO systems is analyzed
through MonteCarlo simulations. The simulation r e s u l ~ s indicate that OAF relaying
technique mitigates the channel impairments and enhances the BER performance. By
IV
employing more relays, longer distances become accessible, however distance improve
ment decreases due to accumulating background noise at relays. In order to remove
background noise effects, another optical relaying technique is developed. The ORF
relaying technique eliminates the received background noise at each relay and signifi
cantly outperforms OAF systems. For example at high bit rate BR= 10 Gbps, using
two equallyspaced OAF relays during a 3 km turbulencefree link increases the total
communicating distance by about 1.11 km. Replacing OAF relays by ORF relays
extends the total communicating distance to 4.48 km which is 1.66 km longer than
the similar OAF FSO system. By deploying more ORF relays, even longer distances
are achievable.
v
i
1
Acknow ledgements
I would like to express my heartily gratitude to my supervisor, Dr. Steve Hranilovic,
whose support, advice and encouragement have been key factors to accomplish this
work. It is an honor for me to study and research under his supervision.
My deepest thanks also go to my cosupervisor, Dr. Shiva Kumar, whose insightful
knowledge and helpful ideas involved me in an innovative and interesting topic of
research. This thesis would not have been possible without his efforts.
I also extend my thanks to my lab mates, Dr. Ahmeq. Farid and Mr. Kasra
Asadzadeh, whose endless helps have been a key factor for successfully accomplishing
this work.
Last but not least, I would like to send my special thanks to my family for their
unconditional love and giving me the opportunity to follow my dreams.
VI
List of Notations
Aeff
a,p
{3
{32
c
en
F
9
G
ha
hp
h
Hg
Hp
i(t)
I(t)
k
Lk
LT
M
n2
N(t, S)
P
b
p
LlVl
Fiber effective core area
Gaussian beam wave parameters
Wave number inside fiber
Fiber dispersion coefficient
Light velocity
Refractiveindex structure constant
Phase front radius of curvature
Channel loss
Amplifier gain
Atmospheric attenuation
Complex propagation loss
Complex channel gain
Conventional geometric power loss
Propagation power loss
Photodetector current
Optical intensity
Wave number in freespace
Length of kth hop
Total communicating distance
Number of relays
Fiber nonlinearindex coefficient
Gaussian noise distribution
Average background power
Optical peak power
Average transmit power
Photodetector responsivity
Symbol duration
Vll
n
Bit interval
T
FWHM
Full pulse width at half maximum
U
ASE
Spontaneous emission noise of amplifier
U(t) Optical field envelope
UF Fiber mode weight factor
j
H!
Gaussian beam radius
,
6wo Regenerator input pulsewidth
6wf
Regenerator output pulsewidth
6Wshift
Filter frequency offset
6WSPM
SPM spectral broadening
I
Fiber nonlinearity coefficient
).
Wavelength
/tx
Mean of intensity fluctuations
Wo
Carrier Frequency
wf
Optical filter center frequency
WSPM
SPMbroadened spectrum
c/J
Phase
c/JNdz, t)
SPMinduced phase shift
<pn(x) Kolmogorov spectrum
'ljJ(r,c/J) Transverse field profile in freespace
'l/JF(r, c/J)
Transverse field profile inside fiber
\[1(7,', c/J, z, t) Optical beam wave
0" Atmospheric attenuation coefficient
o"x
Variance of intensity fluctuations
O"s
Variance of phase fluctuations
O"F
Fiber attenuation coefficient
T Atmospheric turbulenceinduced fading
Vlll
Contents
Abstract
Acknowledgements
List of Notations
1 Introduction
1.1 FreeSpace Optical Communication Systems
1.2 FSO Systems and Channel Impairments ..
1.2.1 Atmospheric Channel Impairments: h(t)
1.2.2
1.2.3
1.2.4
1.2.5
1.2.6
Background Illumination Noise: Ub(t)" ..
Atmospheric Turbulence Mitigation Techniques
Forward Error Correction ........ .
MaximumLikelihood Sequence Detection
Spatial Diversity . . . .
1.3 Optical Multihop Transmission
1.4 Multihop Communication Systems
1.4.1 AmplifyandForward relaying technique
1.4.2 DecodeandForward relaying technique.
IX
iv
VI
Vll
1
1
5
7
8
9
9
10
11
12
14
14
16
1.4.3 Optical multihop relaying techniques proposed in the literature 17
1.5 Thesis Contributions
1.6 Thesis Structure ...
2 Channel Model
2.1 FreeSpace Optical Channel
2.1.1 Atmospheric Attenuation .
2.1.2 Atmospheric Turbulence
2.1.3 Propagation Loss ....
2.1.4 Gaussianbeam wave propagation
2.1.5 ABeD RayMatrix
2.2 The RelaytoRelay Channel Model
2.2.1 Signal projection onto SingleMode Fiber (SMF) .
2.3 Noise Projection onto SingleMode Fiber
2.4 Conclusion ................ .
3 Optical ArnplifyandForward Relaying Technique
3.1 OAF Relay Structure ..... .
3.2 Optimal Relaying Configuration
3.3 Numerical Results and System Performance
3.3.1 Fixed Total Communicating Distance
3.3.2 :t\/Iaximum Accessible Communicating Distance.
3.3.3 Fixed hop Lengths
3.4 Cond usion. . . . . . . . .
x
20
23
25
26
27
28
31
33
35
39
43
50
53
54
55
59
70
71
85
92
98
4 Optical RegenerateandForward Relaying Technique 100
4.1 ORF Relay Structure 101
4.2 Optical Regenerator 102
4.3 Numerical Results and System Performance 110
4.3.1 QFactor Estimation ....... 111
4.3.2 Optimal Relaying Configuration . 114
4.3.3 Fixed Total Communicating Distance . 115
4.3.4 Combination of OAF and ORF Relays 117
4.3.5 Maximum Accessible Communicating Distance . 119
4.3.6 Fixed hop Lengths 122
4.4 Conclusion. . . . . . . . . 124
5 Conclusion and Future Work 127
5.1 Conclusion .. 127
5.2 Future Work. 130
A Optimum relaying configuration 132
A.l Dualhop relaying s y s t e l l 1 ~ . 132
A.lol f{(L{) = 0 . 133
A.lo2 f{'(L{) > 0 133
A.lo3 Jr(x) > 0 for x E [O,L
T
] 135
A.2 (p + I)hop relaying system 136
A.2.1
f;+I(X) .
136
A.2.2 f;+1(x) . 137
Xl
List of Figures
1.1 An FSO Communication system . 5
1.2 Multihop Transmission 13
1.3 Cooperative Divetsity . 14
2.1 The RelaytoRelay Channel Model 33
2.2 Optical intensity of Gaussian beam waves at z = 0 and z = 1 km. 36
2.3 Subsequent optical elements . . . . . . . . . . . . . . . . . . . . . 38
2.4 Comparison between the conventional and new geometric power loss. 48
2.5 Estimation area for the conventional (Hg) and ·new propagation (Hp)
loss. . .......................... .
3.1 An AllOptical Multihop FSO Communication System
3.2 Structure of OAF Relays ................ .
3.3 Optical AmplifyandForeward Multihop FSO Systems
3.4 An OAF Ivlultihop FSO System with M= 1. . . . . . .
3.5 The received optical SNR of an OAF FSO system with lvI = 1 and
49
55
56
60
62
SNRo = 37 dB. . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
3.6 An OAF Multihop FSO System with M= 2 and SNRo = 37 dB. 65
3.7 Optical SNR of an OAF FSO system with M = 2 67
3.8 An OAF FSO system with lvI = p . . . . . . . . . 68
Xll
j
I
!
3.9 An OAF FSO system with M = p + 1. . . . . . . . . . . . . . . . .. 68
3.10
3.11
BER versus SNRo for a 3 km link and different number of relays NI at
BR= 1.25 Gbps, no fading effect is considered. (The plot descriptions
are given in Table 3.1) ............ .
BER versus SNRo for a 3 km link and different number of relays NI at
BR= 10 Gbps, no fading effect is considered. (The plot descriptions
are given in Table 3.1) ............ .
3.12 BER in low SNRo region for LT = 3 km and BR= 1.25 Gbps. (The
plot descriptions are given in Table 3.1) . . . . . . . . . . .
3.13 BER in low SNRo region for LT = 3 km and BR= 10 Gbps. (The plot
descriptions are given in Table 3.1) . . . . . . . . . . . . .
3.14 The signal and noise power variations during a 3 km link operating at
BR= 1.25Gbps with different number of relays M, P
t
= 10mW, and
No = 2 X 1O
15
W 1Hz.
3.15 The signal and noise power variations during a 3 km link operating at
BR= 1.25 Gbps with different number of relays M, P
t
= 40 m W, and
No = 2 X 10
15
W 1Hz. . . . .
3.16 BER versus SNR
o
, for LT = 3 km, BR= 1.25 Gbps, and different
number of relays NI, with fading effects. (The plot descriptions are
73
74
75
76
78
79
given in Table 3.1) ......................... . .. 80
3.17 BER versus SNR
o
, for LT = 3 km, BR= 10 Gbps, and different number
of relays JVI, with fading effects. (The plot descriptions are given in
Table 3.1) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 81
xiii
3.18 BER versus the total communicating distance, LT (km) , for different
number of relays ll.![, P t = 500 m \iV, No = 2 X 10
15
W 1Hz, BR= 1.25
Gbps, without f ~ d i n g effects. (The plot descriptions are given in Table
3.5) ............. .
3.19 BER versus the total communicating distance, LT (km) , for different
number of relays JIll, P
t
= 500 mW, No = 2 X 10
15
VV 1Hz, BR= 10
Gbps, without fading effects. (The plot descriptions are given in Table
3.5) ............. .
85
86
3.20 BER versus the total communicating distance, LT (km) , for different
number of relays JV[, P
t
= 500 mW, No = 2 X 10
15
W 1Hz, BR= 1.25
Gbps, with fading effects. (The plot descriptions are given in Table 3.5) 90
3.21 BER versus the total communicating distance, LT (km), for different
number of relays JIll, P
t
= 500 m W, No = 2 X 10
15
W 1Hz, BR= 10
Gbps, with fading effects. (The plot descriptions are given in Table 3.5) 91
3.22 BER versus SNRo for the constant hop distance Lk = 1 km, bit rate
BR= 1.25 Gbps, and different number of relays JIll without fading
effects. (The plot descriptions are given in Table 3.8) ........ , 93
3.23 BER versus SNRo for the constant hop distance Lk = 1 km, bit rate
BR= 10 Gbps, and different number of relays JIll without fading effects.
(The plot descriptions are given in Table 3.8)
3.24 BER versus SNRo for the constant hop distance Lk = 1 km, bit rate
BR= 1.25 Gbps, and different number of relays M considering fading
94
effects. (The plot descriptions are given in Table 3.8) ........ , 96
XIV
3.25 BER versus SNRo for the constant hop distance Lk = 1 km, bit rate
BR= 10 Gbps, and different number of relays lY[ considering fading
effects. (The plot descriptions are given in Table 3.8) 97
4.1 The typical structure of an ORF relay. . ...... . 102
4.2 Gaussian Pulses propagating through an ORF relay at different points. 103
4.3 The internal structure of a regenerator. .............. " 104
4.4 The Gaussian pulse spectrums for different propagated distances, z. 107
4.5 Typical transfer function of an ideal regenerator. . . . . . . . . . .. 110
4.6 BER of an ORF FSO system with },i[ = 1 and LT = 3 km, obtained
by Me and QF methods. . . . . . . . . ., . . . . . . . . . . . . . .. 113
4.7 BER of different ORF multihop FSO systems with M = 1, LT = 3
km, and BR= 10 Gbps. (The plot descriptions are provided in Table
4.1) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 115
4.8 BER of different ORF multihop FSO systems with 1\1£ = 1 and 1\1£ = 2
for a fixed total communicating distance of LT = 3 km and, BR= 10
Gbps. (The plot descriptions are provided in Table 4.2) . . . . . . .. 116
4.9 BER of different multihop FSO systems with M = 2 for a fixed total
communicating distance of LT = 3 km and, BR= 10 Gbps. (The plot
descriptions are provided in Table 4.3) .......... .
4.10 The maximum accessible communicating distance of different multihop
FSO systems with 1\1£ = 1 and M = 2 when P
t
= 500 m W, No = 2 X
10
15
W 1Hz, and BR= 10 Gbps. (The plot descriptions are provided
118
in Table 4.4) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 121
xv
4.11 BER of different ORF FSO systems with different nUIl1.ber ofrelays 1\([
for a fixed hop distance of Lk = 1 km and BR= 10 Gbps. (The plot
descriptions are provided in Table 4.5) .......... .
XVI
123
~
i
Chapter 1
Introduction
1.1 FreeSpace Optical Communication Systems
A tremendous evolution has recently emerged in telecommunication systems. Var
ious reliable and costeffective techniques and technologies have evolved to deliver
high data rate services to users. Optical fiber communications have the potential to
deliver high data rate applications for longhaul and network access. Beside opti
cal fiber communication, wireless optical communication has recently emerged as a
new technology to deliver various types of services such as indoor infrared wireless
communications [1, 2], visible light communications (VLC) [3], terrestrial links [4],
ground to air (UAV) [5, 6], satellite to ground [7], and intersatellite communications
[8,9].
Freespace optical (FSO) communication is one of the most prevalent applications
for wireless optical systems and refers to terrestrial lineofsight (LOS) optical trans
mission through the atnlOsphere [10]. Today's coml'nercial FSO systems provide bit
rates ranging from 100 Megabits/second (Mbps) to 10 Gigabits/second (Gbps) over
1
McMaster University  Electrical Engineering
Chapter 1. Introduction
distances on the order of a few kilometers, e.g. SONAbeam ™ 1250M by fSONA
[11] and TereScope@lOGE (TS lOGE) by MRV [12]. For a given system reliability,
by increasing data rate, link coverage decreases. For example SONAbeam™1250M
1 provides 1.25 Gbps transmission over distances up to 5400 n1., while MRV product TS
lOGE can deliver 10 Gbps data over only distances up to 350 m. These highspeed
systems are utilized in lastmile applications to connect buildings, companies, and
business offices to the fiber network.
Optical fiber systems also can be used in lastmile applications. Optical fiber is
one ofthe most reliable candidates for connecting the end user, e.g. buildings, offices,
etc, to high rate fiber networks without loosing data rate. However, its deployment
requires digging and a rightofway license which makes fiber deployment expensive
and even impractical in high density cities. Various communication systems exist to
connect high data rate fiber networks to the end users.
Microwave access via WiMAX technology has been introduced as a promising
technology deployed in pointtomultipoint broadband wireless applications. As an
example, the Proxim product Tsunami™ MP8100 exhibits speeds up to 300 Mbps
over a range of up to 8 kilometers (km) [13]. Although WiMAX technology provides
high data rates, it still does not reach the demanded Gbps data rates used in optical
networks. As an alternative, RF communication links operating at high frequencies
in the range 60  86 GHz can be deployed to connect local sites to the fiber network,
e.g. the commercial product GigaBeam  GiCORE G1.25 provides 1.25 Gbps link
over about 1.6 km [14]. Although these systems are capable of delivering Gbps data
rates, the cost and the high attenuation in rain are the main challenges for these
systems [15]. Moreover, they fail to match high speed,commercial optical networks
2
J
1
McMaster University  Electrical Engineering
Chapter 1. Introduction
working at data rates in excess of 1.25 Gbps.
Freespace optical (FSO) systems are licensefree with highbandwidth provid
ing a costeffective and easytoinstall alternative to fiber optics and RF systems.
They also have the features of simple deployment, diggingfree installation, and re
deployment felxibility [16]. They fmther provide inherent security due to the nature
of their directional and narrow beams that makes interception difficult [17]. IvIore
over, narrowbeam lineofsight FSO connections preserve the average transmit power
which is limited in wireless communication systems. Compared to stateoftheart RF
technologies such as GigaBeam  GiCORE G 1.25 accommodating 1.25 Gbps data
rate, commerciallyproduced FSO links provide higher data rates up to 10 Gbps. In
addition, FSO links suffer from smaller attenuation in rain compared to RF links
which operate in the 60  86 GHz band [18].
FSO links can be deployed as a highbandwidth bridge to connect local area net
works (LANs) to longhaul widearea networks (WANs) or metropolitan area networks
(IvIANs) [19]. They are also used as wireless backhaul for WiMAX or WiFi networks
[20]. Unlike \iViMAX which is mainly deployed in pointtomultipoint topologies, FSO
pointtopoint links provides higher data rate transmissions. In addition, FSO link
is an efficient alternative to a fiber system to establish high data rate transmission
in populated city areas where laying fiber is too expensive or impractical. These ad
vantages introduce FSO communications as a promising candidate for high data rate
transmissions.
Beside the advantages of FSO communication links, it is important to point out
the disadvantages and challenges in these systems. Since the laser beam is extrelnely
narrow, accurate alignment is required between transmitter and receiver. The weather
3
McMaster University  Electrical Engineering
Chapter 1. Introduction
dependence of the optical channel is another issue so that the system performance
highly depends on atmospheric conditions. In FSO systems, the propagating optical
beam is attenuated over the optical channel due to absorption and scattering [18].
The optical beam is absorbed mainly by water particles and carbon dioxide molecules
in the atmosphere (absorption). In addition, scattering deflects away a portion of the
optical beam from the intended direct path to the receiver and is mainly caused by
fog, haze, rain, snow, etc. Although rain is known as one of the most destructive
atmospheric condition for 60  86 GHz radio frequency (RF) links [18], FSO links
mostly suffer from fog and haze. Typical atmospheric attenuation of optical. beam at
clear air is 0.43 dB/km, at haze is 4.3 dB/km, and at fog is 43 dB/km [21]. That is,
at a foggy weather condition, FSO transmitter must send at least 43 dB more power
to overcome atmospheric attenuation during a 1 km link which is not feasible from
practical stand point. Other than atmospheric attenuation, optical links also suf
fer from atmospheric refractive index random fluctuations (atmospheric turbulence)
which induce intensity and phase fluctuations of optical beam at the receiver. This
fluctuations degrade the system performance ( atmospheric turbulence is described
in detail in Section 2.1.2.) Furthermore, due to eyesafety regulations governed by
International Electrotechnical Commission (IEC) [22], the average t r a n ~ m i t power is
limited in FSO systems. Ambient ilhunination is another challenge of FSO systems
that is a huge amount of background noise is collected along with the incident data
signal at the receive aperture [10]. In Chapter 3, it will be shown that these limita
tions confine FSO system transmissions to short range applications. In this thesis, the
degrading effects of randomly varying atmospheric conditions and background illumi
nation noise on the performance of FSO systems are investigated and new techniques
4
.MclVlaster University  Electrical Engineering
Chapter 1. Introduction
Input
data
Transmitter
electronic
unit I
La", diode r
w
I Receiver I
electronic
unit
Figure 1.1: An FSO ComlTlUnication system
Output
data
are proposed to increase the accessible communicating distance of FSO systems. The
next section presents a brief overview of FSO channels and summarizes the techniques
which have been proposed so far to combat the aforementioned challenges.
1.2 FSO Systems and Channel Impairments
A pointtopoint FSO communication link is shown in Fig. 1.1. In the transmitter
side, a laser diode modulates data onto an optical wave. The modulated optical wave
is shaped and directed through the atmosphere via an optical lens. For the sake of
simplicity, in practical direct FSO systems Intensity Modulation with Direct Detection
(IlVI/DD) is utilized in which data are modulated in the instantaneous intensity of
optical waves [23]. The intensity of optical waves corresponds to the squared absolute
value of optical field envelopes. The field envelope is the timevarying amplitude of
an optical wave which is modulated by data. Assume a laser oscillating at angular
frequency w which is placed at z = a in the transmitter side. A forward propagating
5
McMaster University  Electrical Engineering
Chapter 1. Introduction
wave along the positive zaxis can be expressed as
\fJ(T, if;, z, t) = Ut(t) 'ljJ(T, Z, if;)
'v'" ""v'
Field envelope n·ansverse field profile
exp[i(wt  kz)]
, j
v
Carrier
(1.1)
where T = J x
2
+ y2 is the radial distance from the beam center line, k is the prop
agation constant along the positive zaxis in free space, and if; is the phase. From
(1.1), the transmit instantaneous intensity I (t) is defined as
(1.2)
In such systems, at the receiver, a recelVlllg apertme followed by a lens collects
and focuses the incident beam onto a photo detector . The photodetector coverts the
collected optical power to an electrical cmrent. This cmrent is proportional to the
instantaneous intensity of the received signal at the receiver scaled by the detector
responsivity, R (AjW) ,
(1.3)
The electrical cmrent is sampled and detected inside the receiver electronic unit.
Finally, the detected data are decoded to extract the sent data stream.
In this thesis, data are modulated onto the instantaneous intensity of optical
fields and at the receiver direct detection technique is performed to detect the signal
(IMjDD). However, as will be discussed in Chapter 2, both intensity and phase
information of optical field envelopes are required for system analyses.
An additive channel nlOdel is considered for the direct FSG system and can be
6
McMaster University  Electrical Engineering
Chapter 1. Introduction
expressed as [10]
(1.4)
where Ut(t) and Ur(t) are respectively the transmitted and received optical field en
velopes at the transmitter and receiver (Fig. 1.1), h(t) is the temporal random fluctu
ations of optical channel, and. U
b
(t) is the temporal profile of background illumination
noise that are briefly overviewed in Section '1.2.1 and 1.2.2.
1.2.1 Atmospheric Channel Impairments: h(t)
The channel gain h(t) models the random fluctuations of the propagation path. The
FSO optical link is a timevarying channel whose random fluctuations arise due to
variation of atmospheric conditions. Atmospheric molecules cause absorption and·
scattering which attenuate the power of the light traveling through the atmosphere.
Furthermore, spatial and temporal variations of the air thermal inhomogeneities cause
random fluctuation of the refractive index (i.e., atmospheric turbulenceinduced fad
ing) which results in fluctuation of the optical intensity at the receiving lens plane ..
These intensity fluctuations degrade the performance of FSO communication systems.
The optical channel impairments will be discussed in more detail in Section 2.1.
The time scale at which the fading state remains approximately constant is de
noted as the coherence time. If the coherence time is much smaller than the symbol
duration T, many different fading states occur within the transmission, which is called
fastfading atmosphoric turbulent channel. However, if the coherence time is equal
or greater than T, the channel is a slowfading atmospheric turbulent channel [24].
Typical FSO fading coherence titne varies approximately between 10
3
and 10
1
sec
[23] and typical symbol duration T is on the order of 10
9
sec, therefore FSO channels
7
McMaster University  Electrical Engineering
Chapter 1. Introduction
can be modeled as slowfading channels. That is, the channel gain h(t) is assumed
constant over 10
6
 10
8
symbols. In what follows, h simply refers to the channel gain
and implicitly indicates that fading is constant over many transmitted symbols. The
1 constant channel gain h is a multiplicative factor [25] representing the amplitude and
phase fluctuations which arise due to atmospheric attenuation ha, geometric spread
of optical beam hp, and atmospheric turbulence T, and can be fonnulated as
(1.5)
The atmospheric attenuation ha depends on the weather condition and wavelength
and includes both absorption and scattering effects (Section 2.1.1). Since the weather
changes slowly on the order of minutes to hours, atmospheric attenuation can be
considered constant over a long time. Geometric spread of the optical beam hp is also
constant for a given optical beam and distance (Section 2.1.3). Despite ha and hp,
atmosphericinduced fading factor T is a random coefficient which causes the received
intensity fluctuates at the receive aperture (Section 2.1.2). Intensity fluctuations at
the receiver degrades the FSO systems performance.
1.2.2 Background Illumination Noise: Ub(t)
The additive noise in (1.4), arises due to ambient illuminations which is the dominant
source of noise in FSO communication systems. In addition to the desired signal,
strong undesirable background radiation is also collected at the receive aperture.
Background noise mainly originates from amhent light coming from optical sources
such as sun, sky, moon, incandescent bulbs, etc, and· is statistically modeled as an
additive white Gaussian noise in time and space with zero mean and variance CT; =
8
McMaster University  Electrical Engineering
Chapter 1. Introduction
N
o
/2 [26]. The collected background noise is processed along with the desired signal
and degrades the overall system performance.
The intensity and phase fluctuations at the receiver, that are induced by atmo
spheric turbulence random variations, degrade FSO system performance. Although,
atmospheric turbulence mitigation techniques have been extensively investigated in
literature, background illumination noise has not been separately considered as a lim
iting performance factor in FSO systems. Whereas, in Chapter 3, it will be shown
that background illumination noise can deteriorate the error performance and limit
the link distance coverage. The next section briefly denotes techniques and methods
which have been proposed so far to combat atmospheric turbulence degrading effects.
1.2.3 Atmospheric Turbulence Mitigation Techniques
Atmospheric turbulence fading is one of the major impairments over FSO channels
where the link range is longer than 1 km [27]. This factor limits the link rate, reliability
and distance [23]. Different techniques have been proposed to mitigate fading effects
such as error control coding [28, 29], maximumlikelihood sequence detection (MLSD)
[30], spatial diversity [31], cooperative diversity, and multihop transmission [32].
1.2.4 Forward Error Correction
Various forward error correcting (FEC) techniques have been deployed in the liter
ature to combat atmospheric turbulence degrading effects. Forward error correction
is aeeomplished by adding redundancy to the transmitted data through an error cor
rection coding algorithm such as turbo codes [33], block and convolutional codes [29],
ReedSolomon (RS) codes [34], and lowdensity paritycheck (LDPC) codes [35]. In
9
McMaster University  Electrical Engineering
Chapter 1. Introduction
[29], the performance of block codes and convolutional codes have been investigated
and compared with turbo codes in the presence of weak atmospheric turbulence con
dition. It has been numerically shown that turbo coding can achieve better BER
J
1 performance over atmospheric turbulence channel. However, turbo coding is not
I
suitable for highspeed optical transmissions because of its high complexity and long
encoding/decoding time that imposes delays on the system [36]. A promising alterna
tive for mitigating atmospheric fluctuation is lowdensity paritycheck (LDPC) coding
scheme which greatly improves BER performance of the system even under strong
atmospheric turbulence condition [35]. Although FEC is one of the best techniques to
combat atmospheric turbulenceinduced fading, use of them in high data rate optical
transmissions im.poses high delay and complexity to encoding/decoding units which
may not be practical.
1.2.5 MaximumLikelihood Sequence Detection
Based on the knowledge of the joint temporal statistics of turbulenceinduced inten
sity fluctuations, maximum likelihood sequence detection (MLSD) has been proposed
as another solution to mitigate atmospheric turbulence fading. The MLSD attempts
to track the instantaneous state of intensity fluctuations and adjusts the detection
threshold to the optimal value [30]. For transmitted sequences of length n, the MLSD
computes the likelihood ratio of each of 2
n
possible sequences which imposes complex
computational costs and long delays to decoder. Various suboptimal methods have
been proposed to reduce computational complexity in MLSD method, e.g. singlestep
Markov chain maximum likelihood detection (SMCIVIL) [30] and pilotsymbol as
sisted detection (PSAML) [30]. The suboptimal methods provide comparable error
10
McMaster University  Electrical Engineering
Chapter 1. Introduction
performance with respect to the MLSD method while they reduce the computational
complexity of MLSD detection method. All the proposed detection techniques pro
vide significant perfonnance improvenlents over symbolbysymbol detection method
at the cost of increased delays and computational complexity for simulating multidi
mensional integrals.
1.2.6 Spatial Diversity
Spatial diversity technique is another promising alternative for fading compensation
that imposes less latency to encoding/decoding units. This technique takes aavantage
of multiple transmit/receive apertures that are placed far apart from each other so
that each LOS path experiences a different fading condition [23]. Each receiver collects
the received optical beam(s) from different spatial angles, this way spatial diversity
.provides information redundancy at the receiver and improves system performance
[37]. In optical spatial diversity systems, the.total average transmit power is the sum
of maximum allowed transmit power of all transmit apertures, therefore, employing
multiple apertures at the transmitter increases the total average transmit power of the
link and allows the system to cover longer distances. Spatial diversity can significantly
reduce the outage probability [38] and improve the outage capacity [39] 0f a multiple
input multipleoutput (MIM 0) link.
To maximize spatial diversity performance, the receive apertures should be placed
as far apart as possible from each other, so that various receivers experience uncorre
lated turbulenceinduced fadings. In practice, it may not be always possible to place
the receivers sufficiently far apart [23]. On the other hand, FSO systems are based on
pointtopoint transmission that is the transmitted signal directs to a specific receiver.
11
McMaster University  Electrical Engineering
Chapter 1. Introduction
Spatial diversity requires a lineofsight communication link between the source and
destination terminals which may not be feasible in all situations. Furthermore, spa
tial diversity in FSO s y s t ~ m s must involve multiple transmit and receive apertures at
, each terminal to create a spatial diversity. Implementing multiple transm.it/receive
aperture scheme and designing encoding/decoding protocols increase the complexity
and implementation costs of the system. Multihop transmission is another alternative
to combat the atmospheric fading effects and increasing freespace link coverage. The
next section provides a brief overview of this technique.
1.3 Optical Multihop Transmission
Multihop transmission relays the data signal from the source node (transmitter) to
destination node (receiver) through intermediate terminals called relays or nodes and
has been introduced as a promising technique to improve FSO links covenige and
reliability through mitigating atmospheric fading impairment [40, 41]. Arranging re
lays in a serial configuration exempts FSO relays from being equipped by multiple
transmit/receive apertures. Despite spatial diversity which requires transmit/receive
apertures to be positioned in a LOS configuration, in multihop systems, source and
destination nodes are connected via intermediate relays and consequently LOS con
nectivity is not necessary. In other words, multihop techniques can support an optical
connection between two buildings which do not have a line of sight. Furthermore,
by subsequent deployment of slllaller multiple hops (hop is the distance between two
relays), more reliable FSO transmission over longer distances becomes achievable [32].
As shown in Fig. 1.3, by em.ploying multihop links in a parallel configuration, new
relaying scheme based on cooperative diversity is sleveloped [42, 43]. In this scheme
12
l'vIcl'vIaster University  Electrical Engineering
Chapter 1. Introduction
k=M+l
Figure 1.2: l'vIultihop Transmission
which is a special configuration of spatial diversity, intermediate relays involve only
one transmit/receive aperture and are placed far enough from each other so that each
receiver suffers from a different fading condition. The diversity of received signal
intensities at destination improves the link endtoend reliability [32]. Obviously, in
cooperative diversity systems, the total communicating distance is determined by the
length of the shortest multihop link. Therefore, to increase the total communicating
distance of the link, a huge number of relays must be placed between the cmimlUni
cating nodes, that in practice, may not be possible. Since the primary purpose of this
study is to propose methods of increasing the total communicating distance of FSO
systems, only multihop systems are of great interests.
In multihop transmission relays are employed in series. This scheme is typically
illustrated in Fig. 1.2. In serial relaying scheme, the source transmits an intensity
lTlOdulated signal to the first relay. The relay (depending on relaying technique)
performs optical operations on the signal and forwards it to the next relay. This
continues until the transmitted signal arrives at the destination. In this thesis in
novative techniques are developed for the multihop schemes to increase the total
comnlUnicating distance of FSO systems while guaranteeing a reliable high data rate.
The next section presents a brief overview of multihop FSO systems and summarizes
their achievements and drawbacks. Also techniques which have been proposed in the
literature to combat these challenges are described.
13
McIVIaster University  Electrical Engineering
Chapter 1. Introduction
1 ""___
,
'l: .
. ..
Figure 1.3: Cooperative Diversity
1.4 Multihop Communication SysteIIls
Multihop transmission has been introduced as a powerful technique to n'litigate the
fading and path loss effects in FSO systems. Two processing techniques are widely
used in literature to implement multihop FSO transmission; namely: decodeand
forward (DF) relaying and amplifyandforward (AF) relaying.
1.4.1 AmplifyandForward relaying
In the previously considered AF methods, the collected optical signal at the receiving
lens of each relay is converted to a photocurrent via a photo detector. The electrical
signal is amplified by an electrical amplifier with a gain specific to each relay. Then
the amplified signal is optically modulated and retransmitted through the next hop.
In all AF FSO systems proposed so far, the amplifier gain of each relay is determined
based on the knowledge of the channel state of the previous hop. The channel state
informationassisted (CSIassisted) gains have been widely utilized to analytically
14
McMaster University  Electrical Engineering
Chapter 1. Introduction
study endtoend performance of optical [27, 44, 45] and RF [11, 16] amplifyand
forward multihop systems. Relays using CSIassisted gains use instantaneous CSI of
previous hop to control the gain of the relay and as a result fix the instantaneous
1 output power of the relay [47, (4)]. In the most AF multihop FSO and RF systems
investigated in the literature, to make mathematical analyses tractable, an ideal model
for relay gains has been considered by which relay completely compensates the channel
fading effects of the last hop, regardless of the noise of that relay [47, (5)]. This
approach has been widely used in analyzing the performance of RF communication
systems [42, 41] and recently attracted attention in FSO system studies [27, 44, 45].
The CSIassisted relays continuously estimates the channel fading amplitude which
may not be practically feasible in all situations.
A practica.l a.lternative for CSIassisted gain is using fixedgain amplifiers in AF
relays [47, (13)]. In [47], it has been shown that RF systems with fixed gain relays have
comparable performance with systems equipped with CSIassisted relays. However
the fixedgain method low complexity and ease of deployment make it a promising
alternative for CSIassissted AF systems. Although average BER of CSIassisted and
fixedgain AF FSO systems over strong GammaGamma atmospheric turbulence have
been investigated in [45], no explicit comparison has been made between performance
of these to gaining techniques. In AF FSO systems, the noise added to the signal at
each relay propagates through transmission path. This accumulated noise is another
limiting factor in AF FSO systems as the number of relays increases.
15
McMaster University  Electrical Engineering
Chapter 1. Introduction
1.4.2 DecodeandForward relaying technique
Similar to AF systems, in DF FSO systenls, at each intermediate relay the received
optical signal is photo detected and converted to an electrical current. The electrical
signal is decoded and reencoded before retransmission. In DF technique, the noise
at each relay is eliminated and does not propagate through the channel. The recon
structed signal is ideally a noiseless signal with average power equal to the transmitted
power at the source. In [48] various DF techniques have been proposed for cooper
ative diversity systems. The primary concepts of these techniques can be applied to
DF multihop systems. In bit detectandforward (BDF) technique the bit sequence
is detected and the detected "0" or "1" bits are retransmitted through the next hop
without applying any error correction technique.
In the normal decodeandfqrward (DF) relaying technique which has been em
ployed in all multihop FSO systems proposed in the literature [27, 32, 49], the bit
stream is decoded. The incorrectly decoded bits are corrected before transmission via
an error correcting technique. At each relay a finite error occurs due to the decoding
process. This error accumulates over multiple hops for longlength hops. In [40], it is
shown that for hops longer than 500m, error grows rapidly as the number of relays.
In DF relaying, noise does not propagate through the channel and decoding error is
less than BDF technique, therefore DF method outperforms BDF technique at a cost
of higher complexity through decoding/encoding processes at each relay. Moreover,
decoding and reencoding processes at each relay inject a time delay into the systenl
which leads to an endtoend delay that increases by the number of relays. Although
DF techniques outperform AF methods, the simplicity of AF over DF, introduce AF
relaying technique a more desirable candidate for pract'ical purposes.
16
McMaster University  Electrical Engineering
Chapter 1. Introduction
1.4.3 Optical multihop relaying techniques proposed In the
literature
lVlultihop techniques have been extensively studied for years in RF communications,
however applying them to FSO systems emerged only recently. In [49], the capac
ity of multihop FSO systems is analyzed from a networking standpoint and channel
impairments such as atmospheric fading, attenuation, and geometric loss were not
considered in the analysis. The biterror rate (BER) of a DF FSO system was stud
ied in [40], in which atmospheric attenuation and geometric loss were the only effects
included in the channel model. In [40], it was demonstrated that the mean and
variance of the error rate is smaller in multihop systems rather than single hop com
munication systems for the same link range and'launched power. In contrast to [40],
a lossfree strong turbulence fading channel is considered in [27] to analyze the end
toend outage probability of an FSO link employing OOK modulation with IM/DD
scheme. Both DF and CSIassisted AF relaying techniques have been investigated.
The results indicate that the outage performance of AF multihop systems degrades
with increasing number of relays. Using the same gain model, outage probability
of AF multihop FSO systems has been analytically calculated for strong turbulence
fading channel modeled by Kchannel and Negative Exponential (NE) channel [44].
The results proposed in [44] justify that by increasing the number of relays while the
hop lengths are fixed, the outage probability increases. In [32], an aggregated optical
channel model including both pathloss and weak tmbulence fading effects was COl1
sidered and the endtoend outage probability of FSO systems utilizing DF relaying
technique was presented. A fixed spacing between the source and destination nodes
has been considered in [32] and the endtoend outage performance of the system for
17
j
I
McNlaster University  Electrical Engineering
Chapter 1. Introduction
different number of relays is simulated. It was shown that outage probability of a
fixedlength multihop system improves by increasing the number of relays because
hop lengths and consequently atmospheric turbulenceinduced fading decrease as the
number of relays increase.
In allmultihop systems proposed so far, each relay is equipped with a photodetec
tor which converts the incident optical power into a photocurrent. Then the electrical
signal is processed by electrical units performing either AF or DF relaying technique.
The processed electrical current is modulated through a high bandwidth laser and
retransmitted to the next relay. In fact, each relay is equipped with analogtodigital
(ADC) and digitaltoanalog (DAC) convertors operating at speeds up to a few Giga
samples per second (Gsps), e.g. ADX3500  3 Gbps XMC Digitizer produced by
PMCSierra® provides up to 3 Gsps data rate [50]. Even if it is assumed that ADC
(DAC) conserves the high data rate of the optical signal, relaying procedures, i.e.
amplification or detection (decoding) process, are performed by low speed electri
cal processors. In other words, instead of taking advantage of wide bandwidth (high
data rate) of optical systems to communicate a huge amount of information at a short
time, the overall system data rate is confined to the speed of electrical processors.
Although high speed electrical processors operating at Gbps rates are e ~ e r g i n g [50],
deploying them in all relays may be cost prohibitive. In addition, in DF systems, an
additional encoding/decoding delay is also added to the system and increases as the
number of relays. A solution for increasing the overall data rate of the multihop FSO
systems and accessing wider bandwidths in a costeffective way is utilizing alloptical
processors at all relays.
Although alloptical AF relaying technique has been widely used in fiber optic
18
McMaster University  Electrical Engineering
Chapter 1. Introduction
comn:lUnication systems [16], its application in FSO systems was tested for the first
time in [51]. A dualhop alloptical AF FSO system was implemented and its BER
was obtained at d i f f e r e n ~ bit rates BR= 2.5 Gbps and BR= 10 Gbps. In this ex
1 periment an alloptical automatic gain controller (OAGC) is utilized to adjust the
optical amplifier gain so that it compensates the power fluctuations of the received
signal due to turbulence effects. The experimental results indicate that at higher
bit rates the probability of error increases. In [51], dualhop relayassissted config
uration is examined for hop lengths of 75 m over which the effects of atmospheric
turbulence are negligible. Until now, no analytical investigation has been performed
on the performance analysis of alloptical multihop FSO systems impaired by atmo
spheric turbulence in longhaul applications. This approach is the primary focus of
this thesis.
19
McMaster University  Electrical Engineering
Chapter 1. Introduction
1.5 Thesis Contributions
This thesis proposes novel relaying techniques for optical multihop communication
~ systems in which all relaying processes, i.e. amplification and detection, are performed
in optical domain.
In this work, an aggregated channel model is considered which takes into account
atmospheric turbulenceinduced fading, atmospheric attenuation due to absorption
and scattering and optical beamspread propagation loss. Although data are mod
ulated in optical field intensity (INIjDD), instead of analyzing optical power, the
magnitude and phase variations of optical fields are considered. Because alloptical
components are employed in all relays and photo detection is only performed once
at the receiver. To analyze the optical field characteristics, a new channel gain is
proposed to completely characterize both magnitude and phase variations of optical
fields propagating through the channel. The main contribution of this work in defin
ing a new channel gain corresponds to deriving new complex propagation loss over
the optical channel including beam propagation through FSO channel, beam focusing
and its projection onto an optical fiber at the beginning of each relay. Although, it
is optical field intensity that is finally detected at the receiver and its phase does not
contribute in detection process, a complete knowledge of field distribution is required
at intermediate relays. The proposed method supports more accurate description of
propagation loss rather than conventional geometric loss (equation (2.16)), especially
at short distances.
After completely defining the channel model, the error performance of alloptical
amplifyandforward (OAF) multihop systems are investigated. In the OAF tech
nique, the received optical field is simply coupled into a fiber and amplified by an
20
lvIcMaster University  Electrical Engineering
Chapter 1. Introduction
optical amplifier. Then the amplified signal is retransmitted through the next hop. A
fixedgain protocol is deployed for mnplifier gain which only depends on the detennin
istic structure of the preceding hop and guarantees a constant eyesafe average power
at the output of each relay. An adjustable gain optical amplifier can also be used at
each relay to compensate the last hop fading effects and provides a fixed instantaneous
output power at each relay. This type of OAF systems are expected tv outperform
the fixedgain OAF systems, further investigations of these systems have been left
for future. Regardless of the amplification method, the collected background noise is
amplified along with the desired signal. The OAF technique considerably il11.proves
the BER of FSO systems over moderate distances (up to 5 km), but the amplified
background noise accumulates through the channel and deteriorates OAF FSO sys
tem performance. Through MonteCarlo simulations, it is nUInerically shown that
by increasing data rate and/or the number of relays, error performance improves
slowly because more background noise is injected into the system. This is the major
drawback of OAF FSO systems that limits the maximum accessible commlll1icating
distance to a few kilometers even if a large number of relays is employed. Despite
their simple structure and ease of implementation, OAF FSO systems are practi
cally suitable for moderaterange applications, i.e. on the order of a few kilometers.
To introduce FSO links as a promising candidate for longrange applications, new
techniques need to be developed to eliminate background noise effects at each relay.
Also known as regenerative technique in RF communications [46], a novel all
optical regenerateandforward (ORF) relaying technique is proposed to combat back
ground noise. In this technique, the optical Rignal is reconstructed so that the collected
background noise is eliminated at each relay. Atmospheric turbulence fading effects
21
McMaster University  Electrical Engineering
Chapter 1. Introduction
are not considered in this study and only the effectiveness of ORF technique ~ n re
moving background noise is investigated in terms of BER performance. The optimal
relaying configuration is determined based on numerical simulations. It is numer
I ically shown that ORF technique significantly outperforms OAF technique and by
increasing the number of relays its performance improves steadily. Despite OAF sys
tems whose maximum accessible distance is limited to a few kilometers, FSO systems
which utilize ORF relaying are able to access extensive communicating distances,
given atmospheric turbulence can be neglected. This outstanding accomplishment is
achieved for the highest available data rate 10 Gbps employed in commercial FSO
communication systems.
22
McMaster University  Electrical Engineering
Chapter 1. Introduction
1.6 Thesis Structure
This thesis is organized as follows: In Chapter 2, a detailed description of optical
wave propagation in free space is provided. An aggregated optical channel between
every two relays is defined which starts fro111. the fiber output at the transmit re
lay and ends at the inside of the fiber at the receive relay. It is assumed that the
output laser beam at the transmitter has a Gaussian transverse profile. The parax
ial approximation [52] is used to find the field distribution at the receive relay from
which propagation loss due to beam spreading through the channel is obtained: The
BeersLambert law [53] is modified to model atmospheric attenuation of optical fields
instead of optical intensities. Moreover, a weak lognormal atmospheric turbulence
regime is considered whose logamplitude and phase are normally distributed and
their statistical moments are mathematically described. Finally, the optical channel
has been modeled as a multiplicative factor which takes into account weak lognormal
atmospheric turbulence, propagation loss and atmospheric attenuation.
In Chapter 3, an optical amplifyandforward relaying technique is introduced
and applied to various FSO system configurations. The average BER is selected as
an evaluating metric for analyzing the error performance of different relaying schemes.
It is mathematically proved by induction that equallyspaced relaying configuration
provides the best average optical SNR at the receiver side of OAF systems and the
result is numerically illustrated through Matlab simulations for the systems with· one
and two relays. Hence, the equallyspaced relaying configuration is employed in all·of
the considered OAF systems. Employing this configuration, the maximum accessible
communicating distance for different number of relays at two bit rates BR= 1.25
Gbps and BR= 10 Gbps is obtained through NlonteCarlo simulations. The BER
23
l
,
McMaster University  Electrical Engineering
Chapter 1. Introduction
results are presented for no fading and weak fading atmospheric conditions and ana
lyzed to contrust the effects of atmospheric fading. Finally, the destructive effects of
background noise in degrading BER performance of OAF systems is investigated and
background noise is introduced as a major distancelimiting factor in OAF multihop
FSO systems.
In Chapter 4, an optical regenerateandforward (ORF) relaying technique is pro
posed to mitigate the degrading effects of background noise. Since lVlonteCarlo (M C)
simulations are timeconslUlling, an alternative method called Qfactor (QF) estima
tion is utilized for additive white Gaussian noise (AWGN) channels to accelerate the
simulation process. The perforHmnce of MC and QF methods and are compared for a
nonlinear dualhop ORF system., and it is shown .that QF provides a close approxima
tion of MC in such systems. The equallydistance relaying configuration is assumed
for ORF systems based on BER simulation of a dualhop ORF system at bit rate
BR= 10 Gbps. Using equallydistance scheme, the maximum accessible distance for
different number of relays is found. In this study, a very weak turbulence fading con
dition (C;, < 1 X 1O
17
m
2
/
3
[52]) has been considered under which the atmospheric
fading effects can be neglected over distances up to 1 km. The simulation results show
that ORF systems are able to remove the background noise completely at each relay
when atmospheric fading is neglected. This property introduces the ORF technique
as a superior method over the OAF in increasing the total communicating distance
of FSO systems.
Finally, Chapter 5 presents concluding remarks and future directions.
24
Chapter 2
Channel Model
In the previous chapter, a typical direct FSO communication link was illustrated
in Fig. 1.1 and a propagating optical beam wave was expressed by (1.1). Many
lasers emit beams with a Gaussian profile, that is the laser is operating in a lowest
order mode called TEMoo mode. At this mode of propaga;tion, the field profile is
independent of ¢ , therefore, 'ljJ(r, z, ¢) at the transmitting aperture (z = 0) can be
described as [52]
(2.1)
where eto is the complex parameter related to effective beam radius (spot size) Wo
and phase front radius of curvature Fo at the transmitter and is given by
2 . 1
eto = k liVe + ~ Fo
(2.2)
The factor Ao is set to normalize the transverse field to carry unit energy at the
transmitting aperture. From (2.1), it is readily shown that
25
McMaster University  Electrical Engineering
Chapter 2. Channel Model
1
1
IAol' J
o 0
(2.3)
For a bit stream of Gaussian pulses the transmitted optical field envelope is expressed
as
(2.4)
here an is the modulating coefficient taking the values 0 and 1 randomly when onoff
keying (OOK) modulation is employed, To is the half of pulse width at 1iein.tensity,
n is the bit interval, and PM is the peak: power of the optical signal. For Gaussian
pulses, PM is obtained as [16]
P _ Tb
M  t 1.
665T
o'
(2.5)
where P
t
is the average transmitted power at the transmitter.
2.1 FreeSpace Optical Channel
Lineofsight FSO communication systems generally use highpower lasers that op
erate in eye safety Class 1M [11], [12] band to achieve a good power budget. The
26
,
I
McMaster University  Electrical Engineering
Chapter 2. Channel Model
allowable safe laser power depends on the wavelength. According to the lEC stan
dards [22], the 1550 nnl band provides higher power budget, compared to the 850
nm band [15]. Consequently, the 1550 nm band is extensively used in long distance
commercial FSO systems, while inexpensive components operating at the 850 nm
band are utilized in short distance FSO systems. As mentioned, in the considered
optical system, data are modulated in the optical intensity of the light, therefore the
allowed power and its variations through the optical channel playa key role in overall
system performance. The considered FSO channel impairments originate from the
atmospheric attenuation, atmospheric turbulence, and propagation loss.
2.1.1 Atmospheric Attenuation
As the optical wave propagates through the atmosphere its power attenuates due to
absorption and scattering. Both absorption and scattering phenoll1.ena are depen
dent on weather and wavelength. Absorption comes from the interaction between
the photons and at0111S or molecules that leads to the extinction of the incident pho
ton, elevation of the temperature, and radiative emission. Scattering phenomenon
redirects the incident photons into a different direction with respect to the original
axis. The atmospheric optical power attenuation is determined by the exponential
BeersLambert law [53]. Using this definition, the atmospheric attenuation for optical
field, ha, can be expressed as
(2.6)
where CJ(ml) is the atmospheric attenuation coefficient. Each wavelength chosen as
a central wavelength has a specific attenuation coefficient, therefore it is preferred
to select the central wavelength corresponding to the least attenuation coefficient.
27
,
I
McMaster University  Electrical Engineering
Chapter 2. Channel Model
Table 2.1: Atmospheric attenuation coefficients for different weather conditions.
Weather Condition Attenuation (dB/km)
Clear 0.20.5
Haze 29
Fog 21272
Fortunately, at 850 nm, and 1550 nm the attenuation coefficient is 10V{, i.g. for the
clear weather condition (]" = 0.43 dBm/km at A = 1550 nm. The attenuation factor is
also dependent on weather condition. Table 2.1 provides the attenuation coefficients
for different weather conditiOlis at wavelength A = 1550nm. As seen from this table,
the fog weather condition has the highest attenuation that limits the FSO link range
to a few meters [54].
2.1.2 Atmospheric Turbulence
Optical beam traveling through the atmosphere experiences random phase and am
plitude fluctuations (scintillation) due to atmospheric turbulence. Turbulence is a
disordered state of the atmospheric flow which is caused by temperature variations·
in the atmosphere. An atmospheric turbulent media consists of many spherical re
gions or eddies with randomly varying diameters and different indices of refraction.
The propagating optical beam experiences random spatial and temporal fluctuations
in this randomly varying refractive. indexed medium with different scale sizes. Large
scale inhomogeneities produce refractive effects that steer the beam in a slightly differ
ent direction, therefore, large scale effects mostly distort the phase of the propagating
wave. Small scale inhomogeneities mostly produce diffractive effects and distort the
amplitude of the wave through beam spreading and amplitude fluctuations [52].
28
"
1
McMaster University  Electrical Engineering
Chapter 2. Channel Model
According to Rytov theory which is proposed for weak turbulence conditions, the
turbulent medium is assumed to consist of a series of thin slabs. Each slab HlOdulates
the optical field from the previous slab perturbation by some incremental values. The
received field can be expressed in terms of the transmitted field U
t
as
(2.7)
where T = e(i = e
x
+
jS
represents the effect of turbulenceinduced fading as a complex
multiplicative term. According to the central limit theorem 'l/J = 2:
i
'l/Ji approaches a
complex Gaussian random variable and therefore, the fading logamplitude (X) and
phase (8) are normally distributed [25].
(2.8)
As a result, the turbulenceinduced fading amplitude ( TA = ITI = eX ) is a lognormal
random variable with logamplitude mean flx and logamplitude variance ( J ~ .
(2.9)
Considering that turbulence does not absorb optical energy, only scatters it, the
energy is conserved for an infinite plane wave or a spherical wave [55]. Gaussian
beam waves behave like plane waves at long distances from the source [56], therefore
an energyconserving condition can be also applied to Gaussian beam waves. As
shown in [55] , to ensure that fading does not attenuate or amplify the average power
(E[ITkI2] = 1), the logamplitude mean !LX must be equal to the negative of the
variance of the logamplitude flx =  ( J ~ .
29
i
l
McMaster University  Electrical Engineering
Chapter 2. Channel Model
The variance of logamplitude fluctuation is a n1.easure of the strength of the
amplitude fluctuations. The principal contribution to the logamplitude fluctuations
is made by inhomogeneities whose sizes are close to Fresnel length J):L [52], where
L is the thickness of the turbulent medium. The Kolmogorov spectrum is considered
as a satisfactory model to calculate the logamplitude variance [52]:
(2.10)
where C ~ is the refractive index structure constant and index n indicates the refractive
index dependency of turbulence. Close to ground levels, for paths that are nearly
horizontal to the earth's surface, C ~ does not change considerably and is assumed to
be constant. In (2.10), x is the spatial wave number defined by x = 211/l, where l is
the scale size. The Kolmogorov' model is applicable throughout the inertial interval
which is defined as [57]
211 5.92
Lo = Xo « x « Xm = T'
(2.11)
where lo and Lo are the internal and external turbulence scale respectively. The
behavior of <pn(x) outside the inertial intervals is not essential. By estimating Gaus
sian beam waves with plane waves, the logamplitude variance within the range of
lo « J):L « Lo is calculated as [25]
(2.12)
where k = 211/ A is the wave number.
30
McIVlaster University  Electrical Engineering
Chapter 2. Channel Model
In [58], a zeromean normal distribution is assumed for the phase fluctuations. De
spite logamplitude fluctuations, the phase fluctuations are mainly determined by the
largescale inhomogenities (including the external scale Lo) [52]. Unfortunately the
1 phase refractiveindex power spectrum is not well defined for scale sizes equal or larger
than Lo. In order to estimate the phase spectrum, certain limiting assumptions need
to be imposed on the behavior of <pn(x) in the region of largescale inhomogeneities.
Different models for phase spectrum have been investigated [52] among which the
twoparameter spectrum model proposed by Kanmin is considered here
(2.13)
where Xo and Xrn "are given by (2.11). Under a geometrical optics approximation
(Lx2 / k < < 1), the phase variance for Gaussian beam waves is obtained from [52]
(2.14)
by substituting (2.13) in (2.14) and performing mathematical simplifications, the
phase fluctuation variance is expressed as
2 0 78c2k2L 5/3
O"s ~ . n Xo . (2.15)
2.1.3 Propagation Loss
The third factor is the geometric power loss which results from the diffractive proper
ties of optical wave propagating through optical media. The conventional model for
geometric loss which has been widely used in IM/DD FSO systems is approximated
31
NIcMaster University  Electrical Engineering
Chapter 2. Channel Model
for unbounded plane waves as [59]
;
(2.16)
1
where 'l/J3(r, ¢) is the transverse optical field profile at the receiving lens plane (point
®), L is the propagating distance, d
t
and d
r
are the transmitting and receiving lens
diameters in Fig. 2.1, and e(rad) is the transmit beam divergence. In this model it is
assumed that the field intensity is distributed uniformly over the receiving aperture
area which is the case for plane waves. Although this model is a satisfactory estimation
for modeling plane waves, in many applications the plane wave approximation is not
sufficient to characterize the properties of the wave. In addition, other than power
loss which is a metric for evaluating the field nmgnitude variations, the phase of the
optical field is also of interest in many applications. In this work, since alloptical
components in all relays are employed, photo detection is only performed once at the
receiver. Thus, instead of analyzing optical power loss, the magnitude and phase
variations of optical fields are considered. In this part, by assuming Gaussian field
distribution for the propagating wave a new model for geometric loss is introduced.
The optical channel between transmitter and receiver is shown in Fig.2.1. At the
transmitter, a signal is transmitted via a transmitting lens with focal leilgth ft. The
distance between the fiber output and lens at the transmitter, L
t
, is dependent upon
the required beam width at the receiving lens to guarantee a reliable alignment. The
receiver has a converging lens that focuses and couples the incident light into the
fiber. To gain a satisfactory coupling efficiency, the fiber is placed at the focal plane
of the receiving lens, i.e. Lr = in where Lr is the distance between the lens and fiber
32
lIIcllIaster University  Electrical Engineering
Chapter 2. Channel Model
TXLens
I
RXLens
TXFiber
I
I
I
I
I
I ...
I
I
I
I
I
)
00
I
CD
Figure 2.1: The RelaytoRelay Channel Model
at the receiving node and 17" is the focal length of the receiving lens.
RXFiber
I
I
I
I
I
8
The optical channel considered here starts from the fiber output at the transmitter
point CD) to the fiber input at the receiver (point ®). The wave coming out of
the fiber at point CD travels distance L
t
to reach to a thin lens at point @. After
propagating thorough a freespace path with length L, it is incident on an optical lens
that converges the received field and couples it onto a fiber at point ®. In order to
analyze the transmitted wave characteristics after undergoing such propagation, the
optical wave propagation thorough various optical media such as freespace, optical
lens, and optical fiber will be discussed first.
2.1.4 Gaussianbeam wave propagation
Over distances on the order of a few kilometers, it is reasonable to consider a Gaussian
model for optical wave profile propagating optical media. Given the field
33
McMaster University  Electrical Engineering
Chapter 2. Channel Model
profile at z = 0, the field distribution at distance z from the source must be found.
If diffraction effects on the optical wave change slowly with respect to propagation
distance z, the optical wave profile at a distance z from the source can be found by
solving the paraxial wave equation [52]
1 0 . o?jJ . o?jJ
(r) + 22k = 0
r or or oz
(2.17)
The paraxial approximation (02?jJ /OZ2 = 0) is valid when the separation distance
between optical elell1.ents is large compared with the transverse extent of the beam.
By solving (2.17) for ?jJ (r, z), the optical wave profile at distance z from the source
is obtained as [52]
?jJ(r, z) = . exp ikz  k . r
Ao [ 1 ( ao ) 2]
1 + UXoZ 2 1 + UXoZ
(2.18)
It is obvious that optical beam wave profile at distance z from the source has a
complex Gaussian distribution with new complex parameters a
z
and pz
(2.19)
where
1 + iaoz
(2.20)
z 2z
pz 1 + iaoz = 1  Fo +i k T ; f ! ~
'..' 'v'
80 Ao
In (2.20), 8
0
and Ao are input plane beam parameters and characterize, respectively,
34
McMaster University  Electrical Engineering
Chapter 2. Channel Model
the refractive(focusing) and diffractive c]langes in the onaxis (r = 0) amplitude of
the Gaussian beam [52]. The transverse field amplitude at distance z is expressed in
terms of 8
0
and Ao as
Ao
Az = r======
86 +
'v"
refraction
A6
'v"
diffraction
(2.21)
When a Gaussianbeam wave propagates thorough freespace, its beam waist (TV)
broadens due to diffractive pr.operties of Gaussian beam waves.To find a visual intu
ition about this concept, a numerical example is illustrated here. Let us consider a
freespace path with length L = 1 km. A unitamplitude Gaussianbeam with the
wavelength A = 1550 nm, the beam radius TVo = 4 cm, and the radius of curvature
Fo = 20 cm propagates thorough atmosphere. The intensity of the optical wave is
defined as
J(r, z) = I'l/J(r, Z)12 (2.22)
In Figure 2.2, optical beam intensity is plotted at z = 0 and z = llun. As shown in
this figure, the beam waist at z = 1 km is approximat"ely 10 times wider than TV
o
.
2.1.5 ABeD RayMatrix
In Fig. 2.1, the optical channel f{"om point CD to point ® is composed of various
optical elements placed at arbitrary positions along the propagation path i.g. free
space, optical lens, and optical fiber. In order to sin'lplify the process of finding the
Gaussianbeam wave profile after propagating thorough subsequent optical elements
through the channel, 2 x 2 matrices called ABeD ray matr'lces have been used. More
35
McMaster University  Electrical Engineering
Chapter 2. Channel Model
NS
0.8
.£0.6
'" c
2:l
oS 0.4
";j
(j
"R
0
.
2
o
z=O
5 4 3 2 1 0 1 2 3 4 5
rem)
4 3 2 1 o
rem)
z=lkm
2 3 4 5
Figure 2.2: Optical intensity of Gaussian beam waves at z = 0 and z = 1 km
details of ABCD ray matrix analysis can be found in [52]. Each optical .element can
be represented by a matrix whose elements construct the output beam characteristics
using input beam parameters. In Table 2.2, the matrix representations of a freespace
path with length L and a thin lens with focal length f are listed.
The field distribution at the output of an optical element can be simply found
by using its ABeD ray matrix. From (2.1), the Gaussian beam profile in the input
36
McMaster University  Electrical Engineering
Chapter 2. Channel Model
Table 2.2: Ray matrices for various optical elements
Matrix
structure
(
AC DB)
Lineofsight freespace path with length L
(
0
1
L1)
Thin lens with focal length f (\ 0)
1 1
plane (z = 0) of an optical element is described as
(2.23)
As expressed in (2.19), the Gaussianbeam wave profile in the output plane (z = L)
is defined exactly via two parameters aL and PL. These parameters can be chararc
terized in terms of ABCD matrix elements
aoD  iC 2 .1
=+2
A + iaoB kltVi FL
PL A+iaoB
(2.24)
where W
L
and FL are, respectively, the beam radius and phase front radius of curva
ture of the wave in the output plane. For example, for a freespace path with length
37
McIVIaster University  Electrical Engineering
Chapter 2. Channel Model
1 2 • • •
N
Figure 2.3: Subsequent optical elements
L, the beam characteristics at the end of the path are calculated as
PL
1 +iaoL
1 + iaoL
which are consistel)t with those derived at (2.20).
(2.25)
Moreover, the received Gaussianbeam wave parameters after passing through
a train of optical devices can be readily found by successively multiplying ABC D
matrix representations of all optical elements. In fact, an optical path including
several arbitrary optical structures can be modeled by a single ABCD ray lTmtrix.
Consider N arbitrary optical elements between transmitting aperture (z = 0) and
receiving aperture (z = L) as shown in Fig. 2.3. Each optical element can be
characterized by an ABCD ray matrix. The overall ABCD matrix at z = L can be
found by lTmltiplying all ABCD ray matrices in reverse order as
(2.26)
From (2.24), by using the overall ABCD matrix, the Gaussian beam characteristics
at the end of optical path can be found.
38
McMaster University  Electrical Engineering
Chapter 2. Channel NIodel
ABC D ray nlatrices invoke the paraxial approximation to simplify the description
of Gaussian wave propagation through optical media. Consequently it is valid when
the separation distance between optical elements is large compared with the transverse
extent of the beam. Under the assumption of lossless optical elements, ABCD matrix
transformation is power conservative and the power of the propagated Gaussian beam
is equal to the launch power at the transmitter.
2.2 The RelaytoRelay Channel Model
In Figure 2.1, the relaytorelay channel refers to the optical path which starts from
the fiber output at point <D. The beam is shaped and redirected to the receiving
end via an optical thin lens at point @. After propagating through a freespace path,
another thin lens (at point@) collects and focuses the optical beam ontothe fiber input
which is placed at point ®. The channel is composed of successive optical elements,
so ABC D ray matrices have been used to 'readily characterize the transverse field
distribution at each point in the channel. Using (2.1) the unitenergy transverse filed
distribution at point <D is expressed as
(2.27)
where (Xl is the Gaussian beam parameter at point <D and Al is the power normal
izing factor. Typically, the beam waist at point @ is considerably smaller than the
transmitting lens diameter d
t
, therefore ABCD matrix transformation corresponding
to the thin lens at the transmitter (point @) and free space conserve the beam energy.
Table 2.3 shows the corresponding ABCD matrices of the optical elements used in
39
McMaster University  Electrical Engineering
Chapter 2. Channel Model
the channel from point CD to point @. By multiplying ray matrices of the optical ele
Table 2.3: Ray matrices of optical elements in Figure 2.1
structure Matrix
!
Lineofsight freespace path with length L
t
(
10 L1t)
Thin lens with focal length it (\ 0)
 It 1
(0
1
L1)
Lineofsight freespace path with length L
Thin lens with focal length ir ( _ 1 ~ ~ )
I,.
(0
1
L1r) Lineofsight freespace path with length Lr
ments placed between point CD and point ® in a reverse order, the ABeD ray matrix
of the beam wave at point ® is obtained as
(2.28)
therefore from (2.19), the transverse field distribution at point ® is obtained as
0/' ( L) Al ikL (1 2)
'f/3 T, Z = 3 = e 3 exp kCt3T
P3 . 2
(2.29)
40
lVIcMaster University  Electrical Engineering
Chapter 2. Channel Model
where
L3 L
t
+L (2.30)
P3 A3 + iCiOB3 (2.31)
CioD3  iC
3
2 . 1
(2.32) Ci3 =+'/,
A3 + iCiOB3 k VVi F3
TY3 and F3 are the beam waist and phase front radius of curvature of the optical
beam wave at point @ respectively. By assuming a lossless optical lens and ignoring
atmospheric absorption loss and turbulence effects (these effects have been taken into
account separately), 'l/J3(r) retains a Gaussian profile with unitenergy, i.e.,
. (2.33)
TYPIcally, in order to have a reliable alignment between the source and destination in
FSO links, the optical beam waist at the receiving lens (point @) must be sufficiently
larger than the diameter of the receiving lens d
r
(TY
3
~ d
r
). The receiving lens is
only able to collect the portion of the received field which is incident on the lens plane
and the remaining parts of the propagating wave are discarded. This process clearly
imposes a considerable power loss to the signal in lineofsight FSO systems and is
conventionally called geometric loss and can be calculated via (2.16).
In the new model proposed here for propgation loss, two factors are considered as
the main contributors in power loss. One; 'rJz, is due to the finite size of the receiving
lens which is considerably smaller thaIl the received beam waist, so that a major part
of the signal power is failed to be collected by the lens and dissipates in space. The
other factor called coupling efficiency, 'Tlc, results from the misalignment between the
41
McMaster University  Electrical Engineering
Chapter 2. Channel Model
focused beam wave distribution at point ® and the field distribution inside the fiber.
Consequently the new model for propagation power loss is defined as
(2.34)
In order to find 'r/l and r]c, fitst the transverse field distribution at point ® must be
determinded. The effective cross section area of the fiber is very small (on the order
of micron), hence to gain a satisfactory coupling efficiency, the fiber is placed at the
focal plane of the receiving lens, i.e. Lr = ir, to guarantee the smallest spot size
( b e a l l 1 ~ waist) at point ®. Since there is no ABCD matrix which exactly models a
finite thin lens, to find the transverse field profile in the back of focal plane of the
lens, 'l/J4(r), the Fresnel diffraction integral is used [52],[60]
(2.35)
where rand s are the radial coordinates at the fiber input plane (point ®) and
receiving lens plane (point ®) respectively. By sub.stituting (2.29) in (2.35) and
performing mathematical simplifications, the transverse field distribution at point ®
is obtained as
d
ik Al ikr2 1 { 1 2 kr
'l/J4(r) = f  exp(
ikL
4) eXP(:f ) exp( 
ka
3s )Jo( f s)sds,
r P3 2 r 0 2 r
(2.36)
where L4 = L3 + Lr = L3 + ir. As shown in Fig. 2.2, in typical FSO systems,
optical beam waist is on the order of a few meters which is considerably larger than
the receiving lens diameter, on the order of a few centimeters. Hence, by reasonably
42
McMaster University  Electrical Engineering
Chapter 2. Channel Model
approximating the exponential term in (2.36), exp( 1 for small 8'S, the
beam profile at point ® is approximated as
J
(
kd )
id A ikT2 1 2/
r
'ljJ4(r) _T_l ex
p
(ikL
4
)ex
P
(j ) r,
2 P3' 2 T r
(2.37)
where J
1
(.) is the first order Bessel function of the fiTst kind. For d
T
7 00, 'l/J4 (r)
tends to a complete Gaussian beam profile whose characteristics can be simply found
via ABC D ray matrix. The thin lens transformation is power conservative that is
the average power of 'l/J4(r) is equal to the average collected power by the receiving
lens. The power loss imposed to the signal is due to the limited collecting area of the
receiving lens and is defined as
(2.38)
where () = II· rdrd¢ is the spatial average operator. From (2.33), the power loss 7]1
A
is obtained as
(2.39)
2.2.1 Signal projection onto SingleMode Fiber (SMF)
In a typical application, the received optical beam must first be coupled into a single
mode fiber (SMF) to be processed by optical elements [51]. Projection of the optical
field onto SNIF imposes additional loss to the system. Furthermore, atmospheric
turbulence degrades the spatial coherence of the propagating beam. and limits the
coupling efficiency [61]. However, the degrading effects of atmospheric turbulence
on fiber coupling is not taken into account and the coupling efficiency arises just
43
McMaster University  Electrical Engineering
Chapter 2. Channel Model
due to misalignment between the incident beam profile and SMF characteristic field
distribution. The characteristic field distribution inside a SMF is independent of
phase ¢ and is given by [16]
(2.40)
where Teo is the radius of the fiber core, C is a constant which can be determined
from the average power carried by the guided mode, and 1'1 and 1'2 are described as
(2.41)
here n1 and n2 are the core and cladding refractive indices respectively (n1 > n2),
and f3 is pmpagation constant inside the fiber which is also dependent on the mode
of propagation. In singlemode fibers, n1 and n2 are chosen such that there is only
one mode of propagation called fundamental mode of propagation or LP 01 inside the
fiber. In order to find the normalized field distribution corresponding to LP 01 mode
of propagation inside the SMF, the constant parameter C is determined such that
the average power carried by this mode is unity, i.e.,
44
McMaster University  Electrical Engineering
Chapter 2. Channel Model
1 J J 'l/JF(r, ¢)'l/J'F(r, ¢)rdrd¢
00
27r J 'l/JF(r)'l/J'F(r)rdr
o
21fC2ll Jg h T lrdr + 1 KJh,r lrdT ]
1
=? c=;=================
(2.42)
[
Teo () 2 00 1
27r J J,2(ry r)rdr + JO('YlTeo) J K2(ry r)rdr
o 11 ]{O('Y2Teo) 0 12
. 0
where AF is the fiber cross section area. The for.ward propagating optical field inside
the fiber is described as
\]! F(r, z, t) = uFUt(t) 'l/JF(r)
'v' 'v'
Field envelope Transverse field profile Carner
(2.43)
In this equation, UF is the mode weight factor which is determined by the projection
of received optical field onto the singlemode fiber
00
21f J'l/J4(r)'l/J'F(r)rdr
(2.44)
o
45
McMaster University  Electrical Engineering
Chapter 2. Channel Model
By substituting 1jJ4(r) from (2.36) in (2.44), the n:lOde weight is obtained as
00 d
i21rk Al J 1 :[ iks
2
1 krs
UF = j exp(ikL
4
) eXP(j )exp(ka
3
r
2
)J
o
(j )1jJ'}..(s)rsdrds
r P3 0 2 r' 2 r
o
(2.45)
Finally by approximating 1jJ4(r) with equation (2.37), the mode weight is approxi
mated as
(2.46)
The coupling efficiency is defined as the ratio of the average optical power coupled
into the fiber to the average collected power in the receiving lens plane [61]
=1
~
(luF1jJF(r) 12) (IUFI2) (I1jJF(r) 12) IUFI2
TJc = (11jJ4(r)1
2
) = (11jJ4(r) 12) = (11jJ4(r)12)
(2.47)
by substituting (2.39) and (2.47) in (2.34), the new propagation power loss is described
as
(2.48)
By comparing the field envelope of the transmitted Gaussian wave, Ut(t) in (1.1),
with the field envelope of the propagating wave inside the fiber, uFUt(t) in equation
(2.43), the field envelope of the launch optical beam wave is multiplied by a complex
constant UF while propagating through the fibertofiber optical channel. Therefore,
the complexvalued geometric field loss in a fibertofiber optical channel is introduced
as
(2.49)
46
McMaster University  Electrical Engineering
Chapter 2. Channel Model
Table 2.4: The considered FSO system characteristics.
Parameter
Transmitting lens d i ~ m e t e r d
t
Receiving lens diameter d
r
Transmitting lens focal length it
Receiving lens focal length ir
Distance between the fiber output plane and transmitting lens L
t
Beam divergence angle ()
Beam waist at laser output liV
I
Phase front radius of cmvatme at laser output PI
Wavelength A
Table 2.5: SingleMode Fiber Specifications.
parameter value
Core radius r co
Core refractive index nl
Cladding refractive index n2
5 J.Lm
1.5047
1.50
Value
5cm
24 cm
20 cm
58.2 cm
19.76 cm
2.7 mrad.
4 J.Lm
00
1550 J.Lm
The comparison between the conventional and new geometric loss is illustrated via
a numerical example. Table 2.4 shows typical parameters of a FSO communication
system which is considered through the thesis for simulation pmposes.
The distance L
t
is adjusted so that the desired beamwidth (TiV
3
~ 3 m) is achieved
at the receiving lens plane. Also, the receiving lens focal length ir is determined so
that the best fiber coupling efficiency (7Je ~ 82%) is obtained. Due to the diffractive
properties of Gaussian beam wave propagation in freespace, the beamwidth broadens
after propagating over long distances, hence, in order to adjust the desirable beam
radius at the receiving lens plane, a converging lens is used as the transmitter side,
47
McMaster University  Electrical Engineering
Chapter 2. Channel Ivlodel
4 5 .        ~        .         ,         .         .         .
~
H
g
+H
40 p
35
15
1000 2000 3000 4000 5000 6000
FSO link length(m)
Figure 2.4: Comparison between the conventional and new geometric power loss.
point @. Table 2.5 shows the singlemode fiber specifications which are commonly
used in fiber optic systems.
In Figure 2.4, the conventional model for geometric power loss in (2.16) is com
pared with the new model given by (2.48). To calculate the values of UF for different
lengths (L), the approximated formula given in (2.46) is used. In Fig. 2.4, the com
parison is made between these two models. For short distances where the beamwidth
is a few times bigger than the receiving lens diameter, approximation of Gaussian
48
Chapter 2. Channelldodel
lVIcMaster University  Electrical Engineering
Esthnation
area of Hp
Estimation
area of Hg
Figure 2.5: Estimation area for the conventional (Hg) and new propagation (Hp) loss.
beams by plane waves is not reliable and the beam power distribution is not uniforlTl'
over the estimation area shown in Fig. 2.5. In this case, the new model for prop a
gation loss provides more accurate approximation than the conventional model. As
shown in Fig. 2.5, the area over which the Gaussian beam profile is approximated by
a plane wave is much sn1.aller for the new model than the conventional model. For'
long distances where the Gaussian beam profile can be approximated bY' a uniformly
povirerdistributed plane wave, both the new and conventional models provide more
reasonable approximations, however, the new model is 1 ~ 1 . 0 r e accurate. Table 2.6 pro
vides more detailed specifications of a Gaussian wave propagating through an FSO
link at two different distances from the source, Ll = 800 m and L2 = 4000 m.
49
I
McMaster University  Electrical Engineering
Chapter 2. Channel Model
Table 2.6: Propagating Beam Wave Specifications for Different Link Length.
Parameter
Conventional geometric power loss Hg
New propagation power loss Hp
Propagation field loss hp
Beamwidth at point @ TiV2
Beamwidth at point @ H!3
Coupling efficiency 7Jc
Ll = 800 m
3.07 X 10
3
4.82 X 10
3
(6.82 + i1.30) x 10
2
2.44cm
1.lm
82.8%
L2 = 4000 m
1.23 X 10
4
1.99 X 10
4
(1.31 + iO.54) x 10
2
2.44 cm
5.45 m
82.5%
2.3 Noise Projection onto SingleMode Fiber
Background noise is the dominant source of noise in FSO communication systems
and is statistically modeled as an additive white Gaussian noise in time and space
with zero mean and variance 0; = N
o
/2 [26]. As expressed in (1.4), background noise
is added to the signal at the receiving lens, therefore it is projected onto the fiber
through the receiving lens. Let N
3
(t, S) denote the Gaussian noise distribution at
the receiving lens plane where s is the radial vector in the lens plane and t is the
time scale. Since background noise distributions in time and space are independent,
N
3
(t, s) can be written as
(2.50)
therefore by projection of N
3
(t, s) onto fiber, temporal statistics of the noise remains
unchanged. Based on the assumed statistical model for background light, the first
n1.oment of the noise spatial distribution at the lens plane (point @ in Fig. 2.1) is
defined as
(2.51)
50
McMaster University  Electrical Engineering
Chapter 2. Channel Model
where E [.J is the expectation operator. From definition, the correlation function can
be written as
(2.52)
which indicates spatially white noise distribution over the receiving lens plane. By
using the Fresnel diffraction integral, the picture of the noise in the back of focal
plane of the lens is obtained as
(2.53)
where sand f are the radial vectors in the lens plane and fiber input plane (point
®) respectively, and dot operator (.) denotes the vectors inner product. The first
moment of the noise picture N
4
(t, r') is calculated as
(2.54)
and spatial correlation is termed as
RN4 (tl' t 2, fl' f 2) = E [N4(t l , f l)N;(t2, r2)J = (2:fr) 2 exp ( ~ ; r (Jf112  Jf212))
(2.55)
51
McMaster University  Electrical Engineering
Chapter 2. Channel Model
x 11 11
Alens Alens
N
O
( k)2 (ik(I>1
2
1>12))IJ (ik> (> »)d>
2
? . exp  r1  1'2 exp 81· 1'1r2 81
27[" iT 2fT fT
Alens
Therefore, the background noise after being focused in the back of focal plane of
the lens has still zero mean and temporally and spatially white distribution, in other
words, the thin lens conserves statistics of the incident field. Now, it must be shown
that the projected onto the fiber has also the same statistics. From (2.44), the
noise weight factor inside the fiber is defined as
nF = 1 J N
4
(t, r)'ljJ}(r')d:,.z
(2.56)
Afiber
the mean and variance of n F are calculated as
. (2.57)
(2.58)
52
McMas'ter University  Electrical Engineering
Chapter 2. Channel Model
Equations (2.57) and (2.58) indicate that the statistics of the noise inside the single
mode fiber is the same as statistics of the background light incident on the receiving
lens plane. Therefore, in the considered FSO system, the additive noise is modeled
as a white Gaussian noise with zero mean and variance ( J ~ = N
o
/2.
2.4 Conclusion
In this chapter the optical channel has been modeled as an AWGN channel. Back
ground illumination is considered as the major source of noise and is modeled as a
zeromean white Gaussian noise. Atmospheric attenuation, atmospheric turbulence
inducedfading, and atmospheric propagation loss are the channel impairments con
sidered in the channel model. In this thesis, the field envelope of the optical signal
and noise are analyzed instead of optical power, therefore a complex model for atmo
spheric turbulence and propagation loss has been considereq. to completely describe
the amplitude and phase variations of the propagating beam wave. A weak atmo
spheric turbulence condition is assumed whose logamplitude and phase are normally
distributed. Also a new method for calculating propagation loss is developed and
numerically compared with the conventional model. Other than geometric spread of
Gaussian beams, the coupling loss induced by the field projection onto a singlemode
fiber has been considered in the new model. The proposed model provides more re
liable approximation of propagation loss rather than the conventional model. This
channel model is utilized to analyze performance of various FSO relaying techniques
presented in the following chapters.
53
Chapter 3
Optical AmplifyandForward
Relaying Technique
In this chapter, multihop FSO communications using alloptical components is stud
ied. The distance dependence of atmospheric turbulence and path loss lirriits the
total communicating distance in FSO systems. Using relaying techniques, "FSO trans
mission is possible over longer distances. Highbandwidth, shortdistance freespace
optical tranceivers, e.g 10 Gbps TereScope TS10GE, encourages replacement of elec
trical relaying processors by optical elements in all relays. In this chapter, it is shown
that by using alloptical relaying techniques, longer communicating distances can be
achieved in FSO systems while taking advantage of highrate optical transmissions.
In an alloptical multihop FSO communication system at each relay data are pro
cessed ill optical domain. An optical amplifyandforward (OAF) relaying technique is
developed and illustrated in Fig. 3.1. As shown in this figure, each relay is composed
of alloptical elements such as an optical lens, optical fiber, and optical amplifier.
At each relay, the background light is added to the received data field. Then the
54
NIcMaster University  Electrical Engineering
Chapter 3. Optical AmplifyandForward Relaying Technique
UO
f
Uk+l Uk+l U
M
+
1
...
Transmitter
k=O
Relay
r I
1 1
1 1
1 1
Optical AmplifyandForward Relay
Figure 3.1: An AllOpticalNIultihop FSO Communication System
Receiver
k=M+1
noisy data field is amplified by an optical amplifier and forwarded to the next relay
or receiver. The optical link between each two consecutive relays is modeled as a
fibertofiber channel which has been cOl11_pletely characterized in Chapter 2.
3.1 OAF Relay Structure
In OAF relaying, there is at least one optical amplifier which amplifies the received
optical field and retransmits it to the nex"t relay. The structure of a typical OAF relay
is simply shown in Fig. 3.2. As mentioned, there is a converging lens at the beginning
of each relay that collects and focuses the incident light onto the back focal plane of
the lens, a plane normal to the lens axis placed at distance itoea! behind the lens.
The complex amplitude distribution of the field in the focal plane of the lens is the
Fraunhofer diffraction pattern of the field incident on the lens. This field distribution
is projected onto a single mode fiber (SMF). The SNIF is connected to the optical
55
McMaster University  Electrical Engineering
Chapter 3. Optical AmplifyandForward Relaying Technique
Receiving
lens
Fiber Fiber
Optical
Figme 3.2: Structme of OAF Relays
amplifier that is mathem.atica,lly modeled as
Transmitting
lens
(3.1)
where U';(t) and Ul'(t) are the received and LramlIllitted signals at the kth relay
respectively, G
k
is the kth amplifier gain and V ~ S E ( t ) is the amplified spontaneous
emission (AS E) noise of the kthamplifier. The ASE noise is modeled as an additive
zeromean white Gaussian noise. The spectral density of ASE noise is given by [16]
(3.2)
where h is Planck's constant, f is frequency, and nsp is the amplifier spontaneous
emission parameter. For the sake ?f simplicity, in what follows every field envelope
U(t) is referred simply by U.
As noted before, optical multihop AF systems developed so far [44, 27, 45] de
ployed an adjustable gain which compensates the fading effects of the preceding hop,
regardless of the noise of that relay. Other than con1.plexity in implementing ad
justable gain unit at each relay, the relay output power does not satisfy eye safety
56
[
McMaster University  Electrical Engineering
Chapter 3. Optical AmplifyandForward Relaying Technique
limitations (IEC) on the maximum permissible average power. In this thesis a more
practical model is provided for amplifier gain at each relay. The proposed model
for the gain of kth relay, G
k
, is independent of random fading fluctuations and only
1
[ compensates the propagation loss and atmospheric attenuation induced by the last
hop (e
h
hop). Also it is chosen so that the average power of the transmitted signal
at the kth relay p
t
k
= E [I Uf n is constant and equal to the average launch power at
the source P
t
= E [IU?1
2
] (Uf is the transmitted field envelope at the source node),
I.e. ,
(3.3)
The transmitted signal at the kth relay, Uf, can be expressed in terms of the trans
mitted signal at the k  1 th relay, U
t
k
\ as
U
t
k
= ~ (hkU
t
k

1
+ uf:) + U ~ S E (3.4)
, 1
V
u}'
where hk is the complex gain of the channel connecting the k  1 th relay to the kth
relay, and Ut is the background noise collected at the receiving lens of the kth relay.
The data signal, atmospheric turbulence, background light and amplifier iloise are all
independent random processes, therefore, from (3.4)
As mentioned in Section 2.1.2, the lognormal fading is normalized so that the mean
intensity of the propagating wave is conserved (E[ITkI2] = 1). Consequently, the
57
!
!
McMaster University  Electrical Engineering
Chapter 3. Optical AmplifyandForward Relaying Technique
average of the squared magnitude of the complex channel gain is
(3.6)
where ha,k and hp,k are the atmospheric attenuation and complex propagation loss
of the kth link respectively. Here, gk is defined as the channel loss or path loss and
includes both the effects of atmospheric attenuation and propagation loss. Using (
3.3) and (3.6), Equation (3.5) can be sinlplified to
(3.7)
where H = E [lVbI
2
] = CT; is the average collected background light power at the
receiving lens plane and to be identical for all relays. = E
is the average ASE noise power of the kth amplifier and from (3.2) is expressed as:
(3.8)
where flJ is the bandwidth of the optical amplifier. Plugging (3.8) into (3.7) and
rearranging for G
k
gives
(3.9)
where P
A
= hJnspflJ. This way, it is guaranteed that the average output power of
each relay satisfies the eye safety constraints. Typically, at optical frequencies, P
A
·is
negligible with respect to P
t
and P
b
, so G
k
can be approximated as
(3.10)
58
McMaster University  Electrical Engineering
Chapter 3. Optical AmplifyandForward Relaying Technique
here, SNRo = Pt! P
b
is the average transmit signaltonoise ratio (SNR) at the output
of each relay. In the case of low background noise, P
b
~ 0, G
k
~ g;l ,that is , the
amplifier totally compensates the effects of the channel loss. In the presence of high
background noise, G
k
is set to a smaller value so as to keep the average output power
of each relay within the eyesafe region. Therefore, the channel power loss imposed on
the data field is not compensated completely via amplification process and leads to
additional degradation effects on the system performance. The amount of G
k
offset
from its noiseless value, g;l, is dependent on both SNRo and gk. From (3.6), it is clear
that gk is a function of the kth hop distance (Lk). In the next section, by optimizing
hop distances, an optimal relaying configuration is demonstrated for a given eyesafe
SNR
o
·
3.2 Optimal Relaying Configuration
In practice, relays are placed at fixed stations between the source and destination
nodes. It is important to arrange relays such that the best system performance is
achieved at the receiver. In this section, the performance of the system is analyzed
in terms of the average optical signaltonoise ratio (SNR) at the receiver .. In the
considered multihop system, relays are consecutively placed between the source (k =
0) and destination (k = Iv! + 1) nodes. Figure 3.3 shows an optical Amplifyand
Forward multihop system with M relays, where OAF relays are typically shown as
amplifiers.
The hop distance, L
k
, which is the length ofthe link connecting the (k1)th node
to the kth node varies for different relays. Let Up denotes the transmitted signal at
59
McIVlaster University  Electrical Engineering
Chapter 3. Optical AmplifyandForward Relaying Technique
Figure 3.3: Optical AmplifyandForeward rvlultihop FSO Systems
t.he source, t.he received field at. t.he lh relay, j = 1,2 ... 111 + 1, is expressed as
u! (fi + (ut + y'G.hk+JU
t
)
+ + JI, y'G.h',h
Pb
)
(3.11)
.Assuming that the signal, background noise, ASE noise and fading are all indepen
dent, the average received power at the receiver (j = 111[ + 1) is
pM+l
r
+
(3.12)
In order to analyze variations of the dat.a signal power and noise dming the chan
nel, the average optical SNR is defined as the ratio of the average data signal power
to the average total noise power. From. (3.12), the average optical SNR at the receiver
60
lVlcMaster University  Electrical Engineering
Chapt.er 3. Optical AmplifyandForward Relaying Technique
is obtained as
Typically amplifier spontaneous emission noise P
ASE
is negligible with respect to data
power P
t
and background noise power P
b
, therefore (3.13) can be approximated as
(3.14)
By substituting G
k
from (3.10) into (3.14), and performing some simplifications, the
average optical SNR is expressed as
(3.15)
where SNR
k
is the average receive SNR at the receiver of a direct FSO link ( where
there is no relay between transmitter and receiver) with length L
k
:
(3.16)
The path loss gk depends on the hop distance L
k
. Now, SNRII;J+l must be optimized
with respect to Lk'S. Consider the optimization problem
max SNRM+l
Lk
M+l
s.t L Lk = LT
k=l
61
(3.17)
1
1
!
McMaster University  Electrical Engineering
Chapter 3. Optical AmplifyandForward Relaying Technique
TX
RX
LI
L2
k=O
...
LT
•
k=2
Figure 3.4: An OAF lVlultihop FSO System with M= 1
The above optimization problem can be converted to a simpler format. Let h(x) =
ftx), where f (x) > 0 and f (x) and h( x) are continuous functions whose first (J' (x)
and h'(x)) and second (JI/(x) and hl/(x)) order derivatives are defined at Xo. In order
to show that h(x) is locally maximized at xo, it is enough to prove h'(xo) . 0 and
hl/ (xo) < 0 which is equivalent to
h'(x)
f'(x)
= 0 =? !'(xo) = 0

J2(x)
(3.18)
Xo
hl/(x)
fl/(x)
< 0 =? fl/(xo) > 0

J2(x)
(3.19)
Xo
therefore, (3.17) is simplified to
M+l
S.t L Lk = LT
(3.20)
k=l
Consider an FSO system with a single OAF relay placed between the source and
destination nodes as shown in Fig. 3.4.
62
McMaster University  Electrical Engineering
Chapter 3. Optical AmplifyandForward Relaying Technique
The optim.ization problem in (3.20) for this system is defined as
"'l;n [nU + gklSNIlo
1
)]
s.t L1 + L2 = LT (3.21 )
By changing the variables L1 = x and L2 = !:TX, and some algebra, t h ~ optimization
problem is simplified to
min·
x
(3.22)
where gl and g2 are the path loss of the first and second links. From·(3.6), the path
loss gk is dependent on the hop distance Lk and is defined as
Ih
h
1
2 uLkH
gk = a,k p,k = e p,k
(3.23)
where ha,k and hp,k have been replaced by their definitions given in (2.6) and (2.48)
from Chapter 2 and Hp,k = Ihp,kl
2
= IUFI2 is propagation power loss. From (2.46),
propagation power loss can be approximated as
(3.24)
where K, is a constant. By plugging (3.24) into (3.23), the path loss is approximated
as
(3.25)
63
McMaster University  Electrical Engineering
Chapter 3. Optical AmplifyandForward Relaying Technique
4.5,,.,,.,
4
~ 3.5
en
Cil
u
'.0
§<
0 3
(1)
:>
'Q:)
u
~
(1)
ES 2.5
2
500 1000 2000 2500 3000
Figure 3.5: The received optical SNR of an OAF FSO system with M
SNRo = 37 dB.
Using this approximation) (3.22) can be described in terms of L
min
x
It is shown in Appendix A that in typical FSO systems
f{(x)lx_!T 0
 2
f{' (x) L_LT > 0)
 2
64
1 and
(3.26)
(3.27)
i
1
McMaster University  Electrical Engineering
Chapter 3. Optical AmplifyandForward Relaying Technique
1   ;             T:
1
1 U
z
U
b
1
1)1 2 1
I I
r.I 1
1 1
TX
1
G,.,
RX
~ L,., :
1
L"
k=O
         1         ~
• ~ 4
k=3
Subsystem
Figure 3.6: An OAF Multihop FSO System with M= 2 and SNRo = 37 dB.
where ff (x) and ff' (x) are the first and second order derivatives of function II (x).
These conditions are sufficient to prove that II(x) has a local minimum at L = LT/2
over the desired range of LT. This range depends on 0:, K" and SNRo and is obtained
in Appendix A. In addition, it is also proved in Appendix A that ff'(x) > 0, for
all ° < x < L
T
, that indicates the function II (x) is convex for all 0 < x < LT and
L = LT/2 is a global optimum point. Therefore, the average optical SNR received
at the receiving lens of an OAF multihop system with )1.1 = 1, as shown in Fig. 3.4,
is maximized when the OAF relay is placed at the middle of the link connecting the
source and destination. Fig. 3.5 illustrates the optical SNR variations by the position
of the OAF relay.
In this plot, the total communicating distance is LT = 3 km which is inside the
acceptable region of LT for SNRo = 37 dB. Obviously, SNR is maximized when
L1 = L2 = LT/2 = 1.5 km which means the best performance of the system iR
achieved for equallyspaced relaying configuration.
Similar calculations can be considered for an OAF system with ]v! = 2 (Fig.3.6 ).
65
II/LcMaster University  Electrical Engineering
Chapter 3. Optical AmplifyandForward Relaying Technique
The optimi7.ation problem is defined as
(3.28)
Here, L
1
, L
2
, and L3 are the hop distances between the source and the first relay,
the first relay and the second relay, and the second relay and the destination node
respectively. By changing the variables Ll = X, L2 = y and L3 = LT  X  y, the
optimization problem is simplified to
mm h(x,y)
X,Y
[(1 + (n;SNRo)leCXXx2)(1 + (n;SNRo)leaYy2)
x (1 + (n;SNRo)lecx(LTXY)(L
T
 x  y)2)]
(3.29)
It can be easily shown that the Gradient vector of h(x, y) is zero at x = y = L
T
/3:
(
8
h
(X'Y») .
V'h(x,y) = 8x = o.
8h(x,y) LT LT
8y X=T'Y=T
(3.30)
Proof of nonnegativity of the Hessian matrix of h(x, y) is laborious. However, the
numerical simulation in Fig. 3.7 which is plotted at SNRo = 37dB indicates that the
maximum SNR is achieved at Ll = L2 = L3 = 1 km.
( 8
2
h(x,y)
H(h) = ~ iJx2
8
2
h(x,y)
8y8x
> o. (3.31)
For higher number of relays (Ai> 1), the mathematical induction proof is utilized
66
~
I
i
lVIcMaster University  Electrical Engineering
Chapter 3. Optical AmplifyandForward Relaying Technique
500 1000
Lr and L2 do not have
value for this region
2000 2500 3000
5.5
5
4.5
4
3.5
3
2.5
Figure 3.7: Optical SNR of an OAF FSO system with 1111 = 2
to find the optimal relaying configuration. As the base case, it has been proved
that for the OAF system with M = 1 (dualhop system), equallyspaced relaying
configuration provides the best performance at the receiver. Now, assume that for
lVI = p, the equallyspaced relaying optimizes the problem (induction hypothesis).
Based on induction proof, it should be proved that equidistance relaying configuration
also maximizes the total average received SNR at the receiver when JY1 = p + 1. In
Fig. 3.8, p relays are equally spaced between the source and destination nodes with
total distance of x. From (3.20), the modified optimization problem for this system
is expressed as
67
McMaster University  Electrical Engineering
Chapter 3. Optical AmplifyandForward Relaying Technique
k=O
TX
10=0
x
k= p+l
                        ~ .
Figure 3.8: An OAF FSO system with lYI = p
RX
k=p+2
X           ~ . +   L p + 2 ~
LT ..
Figure 3.9: An OAF FSO system with lvI = p + 1
(3.32)
Based on the induction hypothesis, the function fp(Ll, L
2
, ... ,Lp+l) is minimized for
equal hop lengths, i.e. Ll = L2 = ... = L
p
+
1
= x/(p + 1). From (3.25) and (3.32),
the n'linimized f;in (x) is obtained as
(3.33)
In Fig. 3.9, one more relay is added to the link plotted in Fig. 3.8 and provides
a longer link with p + 1 relays and total communicating distance of LT = x + Lp 12.
The optimization problem for the extended link can be expressed as
68
McMaster University  Electrical Engineering
Chapter 3. Optical AmplifyandForward Relaying Technique
(3.34)
The function jp+1(L
1
, L
2
,' .. ,Lp+
2
) can be rewritten in terms of jp(L1' L
2
, ... ,Lp+
1
)
as follows
[g (1+ gklSNRnl)] (1
jp(L1' L2,'" ; Lp+l) x (1 + (3.35)
As shown in Fig. 3.9, the total communicating distance LT is divided into' x and
Lp+2' therefore Lp+2 = LT  x. Furthermore, in order to optimize the SNRJ\1+
1
at
the receiver of an OAF system with M = p + 1, the received SNR at the previous re
lay should maximize or equivalently jp(L1' L
2
, ... ,Lp+l) minimizes. By substituting
j;:in(x) from (3.33) and utilizing (3.25), the optimization problem defined in (3.34)
is simplified to
x
(3.36)
In Appendix A, it is shown that for the equallyspaced relaying configuration L1 =
69
McMaster University  Electrical Engineering
Chapter 3. Optical AmplifyandForward Relaying Technique
L2 = ... = Lp+1 = Lp+2 = ( P ~ l ) = (p
L
.J2) ' the function I p+1 (x) satisfies
1;+1 (x) 0
1;+1 (x) > O.
(3.37)
In other words, the equallyspaced relaying configuration provides t h ~ best perfor
mance at the receiver and all the previous relays for !vI = p + 1 and consequently
the induction proof is complete. Therefore for any arbitrary number of relays !vI,
equallyspaced relaying configuration, i.e.
LT
L1 = ... = LM+1 = 
!vI+1'
(3.38)
provides the maximum average SNR
lI
H1 at the receiver over the desired region of LT.
This configuration is utilized in the next section to simulate various OAF systems.
3.3 Numerical Results and System Performance
In this section, the performance of different OAF multihop FSO communication sys
terns with equallyspaced relaying configurations is analyzed. MonteCarlo simula
tions are used to calculate the BitError Rate (BER)' of various systems. System
performance is studied at two bitrates (BR) 1.25 Gbps and 10 Gbps. The degrading
effect of atmospheric turbulence on BER is analyzed at both bitrates and compared
with turbulencefree system performance.
The system under consideration operates at A = 1550 nm in clear atmospheric
conditions with attenuation coefficient of (J = 0.43 dB/km and weak turbulence with
70
\
1
McMaster University  Electrical Engineering
Chapter 3. Optical AmplifyandForward Relaying Technique
refractive index structme constant of C ~ = 1 x 1O
14
m
2
/
3
[52]. The background. noise
power spectral density is assumed to be No = 2 X 10
15
,iV/Hz [1, 10, 62, 40] and the
arnplifier spontaneous emission parameter is nsp = 5. The other characteristics of the
system are as given in Chapter 2 via Tables 2.4 and 2.5.
3.3.1 Fixed Total Communicating Distance
Consider an FSO system where the somce and destination nodes are placed at a
total communicating distance of LT = 3 km from each other. In this section, by
placing a different number of relays (different JlVJ) between the somce and destination
nodes, the significant role of relaying technique in improving the performance of the
system is justified. In order to simulate the FSO system in the slowlyvaraying optical
channel, 10
7
bits are transmitted per channel state. At both bit rates, 64 samples
per bit interval are provided. In the presence of atmospheric tmbulence, the BER
is averaged over NT = 1000 different fading conditions to reasonably simulate the
slowfading tmbulence channel.
The overall performance of the system for different number of relays, M, placed
between the somce and destination is analyzed by plotting the BER versus the trans
mit signaltonoise ratio SNRo = pt/ Pb' Fig. 3.10 and 3.11 correspond to the systems
working at bit rates BR= 1.25 Gbps andBR= 10 Gbps respectively. The tmbulence
fading effects are not considered in these plots. Table 3.3.1 summarizes the config
mation, i.e., number of relays and hop distance, of the systems considered in these
figmes.
From (3.15) and (3.16), the overall performance ?fthe system only depends on the
average transmit SNRo and the relaying configurations, gk. By comparing two figmes,
71
McMaster University  Electrical Engineering
Chapter 3. Optical AmplifyandForward Relaying Technique
Table 3.1: Different system configurations for LT = 3 km.
N1arker hop Distance . Number of Relays
Lk(kTn) ]1.1
e 3 0
~
1.5 1
B
1 2
e
0.75 3
+
0.6 4
~
0.5 5
it is justified that for a given SNRo the system performance for a specific configuration
is nearly the same at both bit rates. Since the amount of background power is relative
to the optical bandwidth or equivalently optical bit rate, the noise power collected at
each relay (or receiver) in 10 Gbps system is more than the noise power in 1.25 Gbps
system. Therefore, In order to gain relatively similar performance at both bit rates,
almost 9dB more power must be transmitted by the system operating at 10 Gbps.
The maximum average power transmitted by the stateoftheart FSO transceivers
is on the order of hundreds of miliwatts, SONAbeam™ 12501\11 transceiver sends
640m W power via focur transmitters each sending 160 m\i\T. By assuming P
t
= 640
m \i\T as the maximunl available transmit power, the maximum achievable SNRo at
BR= 1.25 Gps and BR= 10 Gbps is about 39 dB and 30 dB respectively.
In Fig. 3.11, the BER of the systems operating at BR= 10 Gbps with 11.11 = 0
(direct transmission) and J1;[ = 1 relay have been plotted in the region SNRo :> 30
dB vv'hich is called inaccessible power reg·ion. The best performance of a FSO direct
transmission inside the accessible power region is BER= 1.26 x 10
2
. By employing
one relay in the midway, the systenl achieves BER= 7 x 10
4
which is an order of
magnitude improvement. But, in order to reduce the biterror rate to BER= 10
5
or
72
McMaster University  Electrical Engineering
Chapter 3. Optical AmplifyandForward Relaying Technique
10
2
.::: ..
. ................................... .
............... .
"1'"
...............
I
..... :.,.
15 30 35
Figure 3.10: BER versus SNRo for a 3 km link and different number of relays M at
BR= 1.25 Gbps, no fading effect is considered. (The plot descriptions are given in
Table 3.1)
less, at least two relays must be placed between the transmitter and receiver,(M = 2
or more).
At both bit rates, for a given SNR
o
, by shortening the hop distances via inserting
more relays between communicating nodes (increasing M), BER at the destination
llode decreases .... lts 8110\;\l11 in Fig. 3.10 alld 3.11, inserting Olle relay at the Iniddle of a
3 km link gains 2.49 dB improvement at BER = 10
5
. Here the effects of atmospheric
fading are not considered. Thus this gain mainly comes from the reduction in path
loss achieved by shortening the hop distances.
73
McMaster University  Electrical Engineering
Chapter 3. Optical AmplifyandForward Relaying Technique
. . ... . . _.. : : :
...........•.•...... J .. ;.
Iil
 ..
10
3
:::::::::::::::::::::::::: I .
... ......... .
15
_ ..... _ ... ...... .
.. _ ..... ..... , ... .
....... i .... .
:::::::i
..........
.. ..
30 35
Figure 3.11: BER versus SNRo for a 3 kmo link and different number of relays 1\11
at BR= 10 Gbps, no fading effect is considered. (The plot descriptions are given in
Table 3.1) .
As the number of relays increases, the system performance improves slightly. Qual
itatively, by increasing the number of relays from 1\11 = 4 to JI/! = 5, at BER = 10
5
,
the system performance gains only 0.74 dB improvement which is considerably smaller
than 2.49 dB, the gain achieved by inserting just one relay at the middle of a 3km
link. This deficiency mainly occurs due to the presence of background light noise.
As mentioned before, ambient illumination is collected at each relay. On the other
hand, to keep the average output power of each relay inside the eyesafe region, the
amplifier boosts up the signal with a relatively smaller gain. Therefore, the power
74
McMaster University  Electrical Engineering
Chapter 3. Optical AmplifyandForward Relaying Technique
20 21
Low §NRoRegiolll
SNRo < 24.82 d ~
24
,
,
,
,
,
25
Figure 3.12: BER in low SNRo region for LT = 3 km and BR= 1.25 Gbps. (The plot
descriptions are given in Table 3.1)
of the attenuated signal over the previous hop can not be recovered completely via
amplification process. The other reason is the small changes in hop distances and
consequently smaller reductions in channel loss. As an example, by placing only one
relay in the midway (lY£ = 0 ~ 1M = 1) the hop distance reduces by 50%, however
by increasing M = 4 to 1\1£ = 5, hop distances change from 0.6 km to 0.5 km which
means only 16% reduction in L
k
. Therefore, the path loss and therefore the system
performance improvement is less than the former case.
In summary, by continuously increasing the number of relays, M, the system
performance always improves but after some points the gain achieved by inserting
one HlOre relay is not significant enough to justify the additional costs of inserting
the relay. Thus, depending on the required link coverage, BER, power budget, and
financial feasibility, the number of relays is determined.
75
McMaster University  Electrical Engineering
Chapter 3. Optical AmplifyandForward Relaying Technique
Low SNR Region
.. .... ···0 ... .,.
SNRo < 24.81 d ~
20 21 24 25
Figme 3.13: BER in low SNRo region for LT = 3 km and BR= 10 Gbps. (The plot
descriptions are given in Table 3.1)
For low SNRo the system performance is different. The low SNRo region corre
sponds to the region that for any SNRo bigger than this region, relaying technique
always improves the system performance. As shown in Figs. 3.12 and 3.13, incre.asing
the number of relays does not necessarily improve the system performance. For low
SNR
o
, the average transmit power is relatively small and background noise degrada
tion effects are dominant. For this region, BER of a direct transmission (DT) FSO
link, i.e. no relay is placed between the source and destination nodes 1\11 = 0, is less
than BER of a multihop system with one relay at the middle of the link (M = 1).
Although by dividing the link into two parts the channel loss reduces, the amount
of collected background noise becomes nearly double. So, at the very low SNR
o
,
the system can not compensate the effects of background noise. By inserting more
relays, hop distances decrease more and the channel loss reduces. Simultaneously,
76
McMaster University  Electrical Engineering
Chapter 3. Optical AmplifyandForward Relaying Technique
the amount of background noise increases. Com_promising a trade off between chan
nel loss reduction and background noise increase determines the minimum required
transmit power for which relaying technique outperforms the DT transmission. The
required transmit power decreases by the number of relays.
In order to understand the power variations in the low SNRo region, the received
signal power ~ ~ , the background noise power pI;, the ASE noise power PlsE and
SNR
k
variations during the above mentioned 3 km link are analyzed. From (3.12)
and (3.13), for equidistance relaying configuration, signal and noise powers received
at the kth relay are defined as
p,k _ (gG)k  1 R
b  gG 1 b
(gG)(kl)  1
P
k
ASE =
gG 1 gP
ASE
(3.39)
(3.40)
(3.41 )
(3.42)"
In Fig. 3.14, the signal and noise power variations during the 3 km link are
illustrated for BR= 1.25Gbps. Each square denotes a relay placed during the link in
an equallyspaced relaying configuration. The average transmit power is P
t
= 10 m W
(SNRo = 21 dB < 24.82 dB ) which is small enough to see the effects of background
noise. The SNR of a DT link is denoted by a blackfilled square. As expected, at
this power, performance of a DT link is better than multihop systems with 1\11 = 1,
2, and 3 relays. In other words, in the low SNRo region (SNRo < 24.81 dB), the
relaying technique not only does not improve the system performance but degrades
77
}'v1cMaster University  Electrical Engineering
Chapter 3. Optical AmplifyandForward Relaying Technique
a
j:Q
"0
(1)
10°
i> ...
10
2
,,, ........... , .. .
. . . . . . . ; . . .
__ __'_ __ __'
o 1000 2000 3000
L(m)
10
7
'_' __ __'_ __ '
o 1000 2000 3000
L(m)
(2)
10°
o 1000 2000 3000
L(m)
(4)
lOt
1O
2
L_' __ '__ '
o 1000 2000 3000
L(m)
Figure 3.14: The signal and noise power variations during a 3 km link operating
at BR= 1.25Gbps with different number of relays M, P
t
= 10mW, and No = 2 x
1O
15
W/Hz.
it. Whereas, for high SNRo (SNRo > 24.81 dB), the system performance always
improves by employing relaying techniques. Fig. 3.15 shows the power and noise
variations of the system for P
t
= 40 m W (SNRo = 27 dB> 24.81 dB) where the signal
power is strong enough to combat the background noise. Obviously, by inserting more
relays, SNR at the destination node increases gradually. From Figs. 3.12 and 3.13,
the desired BER region ( 10
6
 10
5
) is taken place at high SNR
o
. Therefore for
all desired SNR
o
, relaying technique improves the system performance. The same
78
I
\
McMaster University  Electrical Engineering
Chapter 3. Optical AmplifyandForward Relaying Technique
(1)
10
1
r================
(2)
10°
10
2
"
10
0 1000 2000 3000 0 1000 2000 3000
L(m) L(m)
104 ..........
(3)
10
2
(4)
•• _. < ••• _. _. ...........
•• _ ••• < ••• _. _ •••••••••
S
10
1
""
"0
10
5 p::
z
Vl <Zl
""p.. <t:
10°
....................... . .......
...... .. ....  . ..................
10
6
0 1000 2000 3000
10
1
0 1000 2000 3000
L(m) L(m)
Figure 3.15: The signal and noise power variations during a 3 km link operating at
BR= 1.25 Gbps with ·different number of relays NI, P
t
= 40 mW, and No = 2 X 10
15
W/Hz.
senario is applied to BR= 10 Gbps systems, because its performance in terms of
SNRo is similar to 1.25 Gbps systems.
Other than background noise, FSO systems strongly suffer from atmospherictur
bulence. In order to the degrading effects of atmospheric fading on FSO
systems performance, the BER of above mentioned systems, whose BER are plotted
in Figs. 3.10 and 3.11, has been simulated in the of lognormal fading and
shown in Figs. 3.16 and 3.17. The BER gets averaged over NT = 1000 different fading
79
McMaster University  Electrical Engineering
Chapter 3. Optical AmplifyandForward Relaying Technique
Ull'iadesslbie
TU
 ,
LowSNRRegion'
2
0_ ,
10
•• :::_ •• :.: ••
............. ,... • "1 .
. ................. "'y" .
,
,
........ +
15 20 35 40 45
Figure 3.16: BER versus SNR
o
, for LT = 3 km, BR= 1.25 Gbps, and different number
of relays M, with fading effects. (The plot descriptions are given in Table 3.1)
states and at each state 10
6
bits are sent.
Again, the system performance in terms of SNRo is similar at both bit rates. To
achieve the same BER at both bit rates for a given BER= 10
5
, the transmitter needs
to send about 9 dB more power when bit rate is 10 Gbps rather than 1.25 Gbps. By
comparing Figs. 3.16 and 3.17 with Figs. 3.10 and 3.11, it turns out that to mitigate
the effects of atmospheric. fading on a DT link, respectively 8.45 dB and 8.49 dB
more power must be transmitted at BR= 1.25 Gbps and BR= 10 Gbps. Table 3.2,
summatizes the difference between the average transmit power of different systems
80
McMaster University  Electrical Engineering
Chapter 3. Optical AmplifyandForward Relaying Technique
15 20
"" "Inaccessible 'Power""
RegioA"
.  ~ ....... """
.. . .. .. ....... ~ ........ .
35 40 45
Figure 3.17: BER versus SNR
Q
, for LT = 3 km, BR= 10 Gbps, and differentnumber
of relays lVI, with fading effects. (The plot descriptions are given in Table 3.1)
with and without fading effects when BER= 10
5
. According to Table 3.2, as 1\11
increases less average transmit power is required to mitigate the degrading effects of
abnospheric turbulence. Because by increasing the number of relays the hop distances
decrease and therefore the fading effects reduce. The atmospheric fading varies by
weather conditions and is not dependent on the system bit rate. Therefore; for a given
system configuration and weather condition, the BER performance of the systeln is
similar for both bit rates.
The huge required average transmit power confines FSO communication systems
81
j
,
MclVlaster University  Electrical Engineering
Chapter 3. Optical AmplifyandForward Relaying Technique
Table The difference between the transmit power of different systems with and
without. fading when BER= 10
5
.
Number of relays BER= 1.25 Gbps BER= 10 Gbps
M dB dB
0 8.45 8.49
1 4.39 4.41
2 2.25 2.33
3 2.06 2.00
4 1.56 1.55
5 1.33 1.33
to short distances even at low bit rates. Illustratively, at BR= 1.25 Gbps whose
corresponding plot is shown in Fig. 3.16, a direct FSO system requires 42.4 dB
transmit SNRo to reach the 3 km distance from the source. As mentioned before,
the maximum available transmit SNRo at BR= 1.25 Gbps is 39 dB, thus direct FSO
transmission is not possible over 3 km link at BR= 1.25 Gbps. By placing only one
relay in the midway, a 6.59 dB gain is achieved and the required transmit SNRo
decreases to 35.8 dB which is inside the accessible power region. In other words, in
order to communicate over a 3 km link at BR= 1.25 Gbps, at least one relay must
be placed between the source and destination node.
For higher bit rates the situation is even worse. As noted, at BR= 10 Gbps, the
maximum available SNRo is 30 dB. Clearly, without enlploying relaying techniques
to guarantee BER= 10
5
, the direct FSO system needs to provide at least 42.4dB
transmit SNRo which lays inside the inaccessible power region. The best performance
l' d' ':) 1 DOi\" 1 .L C11\TT) C)l"\ lrl' TYf:11\. 7 An t r\._'J T t' ....... ' ........
Ol a lIed v hill 1:' OV llnK aL onno = 0U UD IS .t10 X 1U . 1n sec IOn It
is shown that atmospheric turbulence directly depends on the propagation distance.
Therefore, the effects of atmospheric fading is considerably mitigated by reducing the
hop distances. By placing one relay at the middle of the 3 km link, BER reduces to
82
NlcMaster University  Electrical Engineering
Chapter 3. Optical AmplifyandForward Relaying Technique
Table 3.3: Power gains achieved by employing OAF relaying technique in a FSO
communication system operating at BR= 1.25 Gbps for a particular BER= 105.
Increasing the number of relays
M>lV[+1
Power gain without fading
dB
2.52
1.55
1.27
0.82
0.74
Power gain with fading
dB
6.59
3.68
1.46
1.32
0.97
Table 3.4: Power gains achieved by employing OAF relaying technique in a FSO
conllTlUnication system operating at BR= 10 Gbps for a particular BER= 105.
Increasing the number of relays
]I/[>M+l
Power gain without fading
dB
2.49
1.61
1.15
0.86
0.74
Power gain with fading
dB
6.63
3.63
1.47
1.31
0.96
1.39 X 10
2
. Although dividing the propagation path into two smaller parts reduces
the fading effects, to achieve the particular performance BER= 105 at the receiver,
the hop distances must be shortened more to overcome the degrading effects of channel
loss and high background noise. From Fig. 3.17, in order to communicate over a 3
k111 FSO link with BER= 105 at least four equidistance OAF relays must be placed
between the source and destination nodes.
Tables 3.3 and 3.4 respectively summarize the power gains achieved at a given
BER= 105 by increasing the number of relays for bit rates BR= 1.25 Gbps and
BR= 10 Gbps. By comparing the second and third columns of Table 3.3 or 3.4, the
83
;
McMaster University  Electrical Engineering
Chapter 3. Optical AmplifyandForward Relaying Technique
substantial role of relaying technique in mitigating the atmospheric fading effects is
realized. When no fading is considered in the channel n'lodel, the relaying technique
reduces the channel loss but increases the collected background noise. However,
.j when fading effects are taken into account, shortening the hop distances other than
the channel loss also decreases the degrading effects of fading. Therefore, in a slow
fading channel, relaying technique obtains a. considerable gain in the average transmit
power. Illustratively, at BER= 10
5
, by using only one relay in the midway of a 3
km link operating at BR=1.25 Gbps, 6.59 dB improvement is achieved. But when
fading is neglected, the system obtains only 2.52 dB gain, i.e. 4 dB less.
By increasing the number of relays, the difference between the power gains ob
tained in the absence and presence of atmospheric fading decreases. The reason can
be simply explained via the distance dependence of atmospheric fading. At short
distances (less than 1 km) fading effects are negligible so that by changing the hop
distances from 0.6 km to 0.5 km, they do not change by much. Consequently, at very
short hop distances relaying technique mainly decreases the channel loss to overcome
the effects of background noise and its contribution in mitigating fading effects is
negligible.
As shown in Figs. 3.16 and 3.17, at a given BER= 10
5
in order to communicate
over a 3 km FSO link some of the considered OAF FSO systems require to send a
huge amount of power which exceeds the maximum available average power. In other
words, for some configurations it is not possible to reach the 3 km distance from
the transmitter. In the following section, the maximum accessible distance from the
transmitter for various relaying configurations is presented.
84
McMaster University  Electrical Engineering
Chapter 3. Optical AmplifyandForward Relaying Technique
c.::
W 3
co 10:::
5 10 15 30 35 40
Figure 3.18: BER versus the total communicating distance, LT (km) , for different
number of relays lVI, P
t
= 500 m W, No = 2 X 10
15
W 1Hz, BR= 1.25 Gbps, without
fading effects. (The plot descriptions are given in Table 3.5)
3.3.2 Maximum Accessible Communicating Distance
As mentioned in section 3.3.1, by increasing the number of relays during a fixedlength
link, error performance of FSO systems improves and more power gain is achieved.
However, the gain improvement reduces as the number of relays increases because
the total accumulated background noise grows by the number of relays. For example,
for a 3 km link, by increasing the number of relarys from IV! = 4 to IV! = 5, at
BER= 10
5
, the system perform.ance gains only 0.74 dB improvement which from
85
McMaster University  Electrical Engineering
Chapter 3. Optical AmplifyandForward Relaying Technique
Figure 3.19: BER versus the total communicating distance, LT (km) , for different
number of relays iI/I, P
t
= 500 m\iV, No = 2 X 10
15
W/Hz, BR= 10 Gbps, without
fading effects. (The plot descriptions are given in Table 3.5)
commercial point of view is not satisfactory. Therefore, there exists a compromise
between the number of relays and average transmit power to achieve a particular
BER. Since average transmit power is restricted to the eyesafe region, the number
of relays determine the maximum total communicating distance at a given BER and
average transmit power. Figs. 3.18 and 3.19 show the BER of different relaying
systems with different number of relays, JI/I, which are arranged in an equallyspaced
relaying configuration. The average transmit power is assumed P
t
= 500 m W which
86
JVIcMaster University  Electrical Engineering
Chapter 3. Optical AmplifyandForward Relaying Technique
Table 3.5: .Maximum achievable communicating distance LT(km) without fading ef
fects
lIIarker Number of Relays BR= 1.25Gbps BR= 10Gbps
j\1
LT(km) LT(km)
e
0 4.46 1.71
B
1 5.95 2.28
e
3 8.34 3.09
+ 5 10.15 3.71
V
7 11.72 4.23
is inside the accessible power region. For each NI, BER is simulated for different hop
distances, i.e. for different total communicating lengths. The effects of atmospheric
turbulence are not considered in these plots.
Table 3.5 provides the maximum. achievable 'communicating distances at two bit
rates BR= 1.25 Gbps and BR= 10 Gbps for different number of relays .Ai so that
BER= 10
5
is obtained at the receiver. As indicated in this table, the maximum
accessible distance for a direct FSO system operating at BR= 1.25 Gbps is 4.46
km when the atmospheric fading is not considered. This distance for the system.
operating at BR= 10 Gbps is 1.71 km. The huge difference between these two systems
is referred to the wider bandwidth of optical elements operating at higher bit rates
and consequently collection of more additive. white noise at the receiving apertures.
By employing relaying technique (NI = 7) FSO systems can access to 11.72 km and
4.23 km distances at BR= 1.25 Gbps and BR= 10 Gbps respectively which is a
considerable accomplishment for shortrange highrate FSO communication systems.
By increasing the number of relays (more than 1\,f = 7) even longer distances are
accessible.
As demonstrated before, due to additive background noise, the performance of
87
MclVlaster University  Electrical Engineering
Chapter 3. Optical AmplifyandForward Relaying Technique
Table 3.6: Distance improvements (llLT) by increasing the number of relays.
Increasing the number of relays BR= 1.25 Gbps BR= 10 Gbps
M 7 lV[ + 2 llLT(km) llLT(km) llL!J.f(%)
1 7 3 2.39 40.01 0.81 35.53
3 7 5 1.81 21.71 0.62 20.06
5 7 7 1.57 15.46 0.52 14.01
multihop FSO systems improves nonlinearly by increasing the number of relays. Table
3.6 shows the distance improvements (llLT) by increasing the munber of relays at a
particular BER= 10
5
. llL!{! is the percentage change in LT and is defined as
M llLT
llLT = LM X 100,
T
where L¥J is the maxinlUIIl accessible communicating distance when ljl[ relays are
employed. llL!{! represents the relative improven'lent in LT and can be used as a
metric to compare the system performance at two different bit rates.
From Table 3.6, by increasing the number of relays, llLT decreases gradually.
Qualitatively at BR= 1.25 Gbps, by increasing 1\1£ = 1 to M = 3 the total communi
cating distance extends 2.39 km while by changing 1\1£ = 3 to 1\1£ = 5, it increases only
by 1.81 km. This deficiency is due to background noise. By increasing the number
of relays, more background noise is added to the signal. High noise power not only
confines the amplifier gain at each relay but also degrades the signaltonoise ratio
and system performance at the receiver. The degrading effects of background noise
are more dominant at higher bit rates. By comparing the relative distance improve
ments llL!{! at two bit rates, it turns out that OAF relaying technique has better
performance at lower bit rates.
88
McMaster University  Electrical Engineering
Chapter 3. Optical AmplifyandForward Relaying Technique
Table 3: 7: Maximum achievable communicating distance in the presence of fading at
BER= 10
5
, P
t
= 500 m W, and No = 2 X 10
15
\tV 1Hz.
N umber of Relays BR= 1.25 Gbps BR= 10 Gbps
NI '(km) Lf( km) (km) Lf (km)
0 2.32 2.32 47.98 1.33 1.33 22.22
1 3.48 1.74 41.51 1.87 0.94 17.98
3 5.12 1.28 38.61 2.59 0.65 16.18
5 6.32 1.05 37.78 3.16 0.53 14.82
7 7.29 0.91 37.73 3.63 0.45 14.18
To evaluates the effects of atmospheric turbulence on the maximum communicat
ing distance, Figs. 3.20 and 3.21 are provided. In these figures, the BER of the system
in the presence of lognormal fading versus the total communicating distance LT is
plotted. As mentioned before, atmospheric fading deteriorates system performance.
The lognormal atmospheric fading decreases the maximum communicating distance
of a direct FSO system operating at BR= 1.25 Gbps by 2.14 km when BER 10
5
and P
t
= 500 mVi,T. This value for the system working at BR= 10 Gbp's is 380 m.
Table 3.7 provides the maximum communicating distances for various OAF FSO sys
tems with different number of relays for both bit rates. Also the relative distance
reductions due to atmospheric fading, are presented for different systems.
t:,.L!;'. is defined as
I
L LFI
t:,.L
F
= T  T X 100
T LT '
where L!;'. and LT are respectively the maximum achievable distances over the FSO
channels with and without fading effects. The length Lf in Table 3.7 indicates the
hop distance when atmospheric fading is considered.
By comparing and LT given in Tables 3.5 and 3.7, it is obvious that atmospheric
89
MclVlaster University  Electrical Engineering
Chapter 3. Optical AmplifyandForward Relaying Technique
104 :::: ..
5 10 15 30 35 40
Figure 3.20: BER versus the total communicating distance, LT (km) , for different
number of relays NI, P
t
= 500 mW, No = 2 X 10
15
'iV/Hz, BR= 1.25 Gbps, with
fading effects. (The plot descriptions a n ~ given in Table 3.5)
fading decreases the maximum accessible distance. By employing relaying technique
the degrading effects of atmospheric fading are mitigated. In Table 3.7, the parameter
!::::.Lf expresses the relative decrement in the total communicating distance when
atmospheric fading is considered and is introduced as a metric to readily analyze the
fading effects on the BER of various systems with different number of relays and
bit rates. From Table 3.7, by increasing the number of relays, the relative distance
reduction !::::.Lf decreases which means fading effects are compensated by using more
90
McMaster University  Electrical Engineering
Chapter 3. Optical AmplifyandForward Relaying Technique
Figure 3.21: BER versus the total communicating distance, LT (km), for different
number of relays NI, P
t
= 500 mW, No = 2 X 10
1
.
5
W/Hz, BR= 10 Gbps, with
fading effects. (The plot descriptions are given in Table 3.5)
relays in multihop FSO systems. The reason is s i m ~ p l y realized by considering hop
distances (Ln. By increasing the number of relays,· BER= 10
5
is obtained at
shorter hop distances because the'system needs to compensate the effects of more
additive background noise by reducing the atmospheric fading and channel loss. At
short hop distances (shorter than 1km) fading effects are negligible and therefore the
relative distance reduction due to atmospheric fading decreases slightly. From Table
3.7, BER= 10
5
is achieved at shorter Lf in the systems working at BR= 10 Gbps
91
.J
NIcMaster University  Electrical Engineering
Chapter 3. Optical AmplifyandForward Relaying Technique
Table 3.8: Different system configurations for Figs.3.22 to 3.25 (Lk = 1 km)
Marker Number of Relays Total Communicating Distance
11/[ LT
~ 0 1km
~
1 2 km
B
2 3 km

3 4km
+ 4 5 km
V
5 6 km
compaxed with BR= 1.25 Gbps. Therefore, atmospheric fading has less contribution
in reducing the maximum accessible distance at higher bit rates and background noise
is more dominant.
Distancedependent atmospheric fading limits FSO systems to operate over short
distances. Employing relaying techniques makes FSO communication possible over
longer distances by shortening the hop distances and mitigating the effects of atmo
spheric fading. However, additive background noise still remains as a powerful factor
in degrading the performance of multihop FSO systems. In the next section, the
effects of background noise on BER of different systems are investigated at two bit
rates.
3.3.3 Fixed hop Lengths
Optical AF nmltihop FSO systems strongly suffer from background noise. As demon
strated in sections 3.3.1 and 3.3.2, by increasing the nUlnber of relays nlOre background
noise is received at. t.he receiver. On the other hand, to guarantee an eyesafe average
output power at each relay, the amplification process does not completely compensate
92
\
McMaster University  Electrical Engineering
Chapter 3. Optical AmplifyandForward Relaying Technique
Figure 3.22: BER versus SNRo for the constant hop distance Lk ~ 1 km, bit rate
BR= 1.25 Gbps, and different number of relays NI without fading effects. (The plot
descriptions are given in Table 3.8)
the attenuation of the previous channel. In Chapter 2, Equation (3.15) simply sum
marizes the amplification process and mathematically formulates the average optical
SNR at the receiver.
From (3.15), by increasing the number of relays while the hop distance is fixed,
i.e. constant gb the multiplica.tion term in the denominator increases and totally the
average received SNR decreases. By analyzing the system performance in terms of av
erage received SNR, it is readily realized that because of background noise, extending
the total communicating distance by consecutively adding more relays degrades the
93
j
McMaster University  Electrical Engineering
Chapter 3. Optical AmplifyandForward Relaying Technique
Figure 3.23: BER versus SNRo for the constant hop distance Lk = 1 km, bit rate
BR= 10 Gbps, and different number of relays 1M without fading effects. (The plot
descriptions are given in Table 3.8)
error performance. In order to analyze the degrading effects of background noise, an
OAF FSO system with a fixed hop distance of Lk = 1 km is considered. The "BER of
the system for different number of relays and therefore different total communicating
distances are plotted in Figs. 3.22 and 3.23. The effects of atmospheric turbulence are
not considered in these plots. Table 3.8 summarizes the configuration of the systems
considered in these figures.
At a given SNR
o
, both systems have similar performance at BER= 10
5
. A 2 km
link with an OAF relay in the midway consists of two 1 km DT links. From (3.12),
94
lVIcMaster University  Electrical Engineering
Chapter 3. Optical AmplifyandForward Relaying Technique
Table 3.9: Average transmit SNRo(dB) to obtain BER= 10
5
.
Number of relays BR= 1.25 Gbps BR= 10 Gbps
N! W /0 Fading With Fading W /0 Fading With Fading
o 14.68 16.47 23.67 25.48
1 18.64 20.84 27.75 29.90
2 20.85 23.11 29.85 32.17
3 22.21 25.32 31.29 34.36
4 23.27 26.86 32.32 35.91
5 24.12 27.96 33.23 37.01
the background power at the receiver of the 2 km link is nearly twice the background
power at the receiver of the 1 km DT link. At BER= 10
5
, the transmitter needs
to send 3.97 dB more power to guarantee the particular BER= 10
5
at the receiver.
By increasing the average transmit power, the FSO system overcomes the degrading
effects of accumulated background noise and provides longerrange FSO links. How
_ ever, the total communicating distance in OAF FSO systems is restricted because
the average transmit power is limited. From Fig. 3.23, the ma.ximum communicating
distance for the system operating at BR= 10 Gbps is 3 km when fading effects are not
considered. To reach longer distances, the system needs to send more power which
lays inside the inaccessible power region.
Figures 3.24 and 3.25 are provided to show the degrading effects of atmospheric
fading on the system perfornmnce. To mitigate the effects of atmospheric turbulence,
the transmitter must launch additional power. Table 3.9 summarizes the required
average transmit SNRo (dB) so as to obtain BER= 10
5
at the receiver.
In order to compare the atmospheric fading effects on the BER of various systems,
the an:lOunt of increase in the average transmit signaltonoise ratio, .6.SNR
o
, is defined
95
McMaster University  Electrical Engineering
Chapter 3. Optical AmplifyandForward Relaying Technique
Figure 3.24: BER versus SNRo for the constant hop distance Lk = 1 km, bit rate
BR= 1.25 Gbps, and different number of relays M considering fading effects. (The
plot descriptions are given in Table 3.8)
as
6SNR
o
= SNR[  SNR
o
,
where SNR[ and SNRo are respectively the average transmit signaltonoise ratio sent
over the channels with and without fading effects to hit the target BER= 10
5
. The
variable 6SNR
o
for two bit rates and different number of relays are given in Table
3.10. From this table, atmospheric fading imposes almost equal additional transmit
power to the system at both bit rates. By increasing the number of relays, the required
96
McMaster University  Electrical Engineering
Chapter 3. Optical AmplifyandForward Relaying Technique
10°
:::::::::::(:: ............. , ........ .
.. .. .
5 10 15 20 25
SNRo(dB)
... : "Inaccessible···
::::Rt:giQIl:::::
30 35
Figure 3.25: BER versus SNRo for the constant hop distance Lk = 1 km, bit rate
BR= 10 Gbps, and different number of relays J..if considering fading effects. (The plot
descriptions are given in Table 3.8)
additional transmit power increases because the overall channel fading is the result
of multiplication of J..if + 1 fading coefficients defined over 1\1£ + 1 different 1 km DT
links. Therefore the overall fading variance and consequently its degrading effects
increase. By increasing the average transmit power, the system compensates the
effects of background noise and atmospheric turbulence and extends the link distance
coverage. However the maximum average transmit power is restricted because of
eyesafty issues and can not increase arbitrarily. This restriction on the average
transmit power along with background noise and atmospheric effects confines OAF
97
,
McMaster University  Electrical Engineering
Chapter 3. Optical AmplifyandForward Relaying Technique
Table 3.10: The relative SNRo increase ,6.SNRo(dB)
Number of Relays(M) BR= 1.25 Gbps BR= 10 Gbps
o 1.79 1.81
1 2.20 2.15
2 2.26 2.32
3 3.11 3.07
4 3.59 3.59
5 3.84 3.78
FSO transmissions to short distances.
3.4 Conclusion
FSO communication systems suffer extensively from atmospheric turbulence and
background noise. A serial amplifyandforward relaying technique has been intro
duced as a powerful technique to mitigate the atmospheric turbulence effects at long
haul FSO systems. By deploying more relays and decreasing the hop distances, fad
ing effects are reduced and FSO systems access to long distances at a lower average
transmit SNR
o
. Illustratively, by inserting only one relay at the middle of an FSO
link with length LT = 3 km, about 6.6 dB improvement is achieved in the transmit
SNRo at BER= 10
5
. It has been numerically shown that for a given SNR
o
, perfor
mance of the system is similar at both bit rates. However, the maximum distance
coverage of the system operating at BR= 1.25 Gbps is more than the system work
ing at BR= 10 Gbps to obtain a similar BER. Qualitatively, for a given P
t
= 500
mW and BER= 10
5
, the 1.25 Gbps system with }I;I = 7 relays provides a 7.29 km
long link, however this value for the similar 10 Gbps system is 3.63 km. Because the
FSO system working at higher bit rates (wider bandwidths) collects more background
98
McMaster University  Electrical Engineering
Chapter 3. Optical AmplifyandForward Relaying Technique
noise at the receiving lens, therefore, for a given transmit power, the 10 Gbps system
affords less SNRo and hence shorter communicating distance.
Increasing the number of relays is accompanied by collecting more additive back
\ ground noise at relays that degrades the system performance. Therefore, to reach a
specific communicating distance at a given BER, a trade off is compromised between
the average transmit power and the number of relays. Since the average transmit
power is limited due to eyesafety regulations, the number of relays determines the
maximum communicating distance. By increasing the number of relays, besides the
total communicating distance, the background noise also increases so that the distance
improvement reduces. Although the OAF relaying technique reduces the effects of
atmospheric fading, background noise still remains as a limiting factor in FSO com
munication systems. In Chapter 4, an optical regenerative relaying method using
nonlinear optics is developed to reduce the background noise effects and increase the
communication distance coverage.
99
I
Chapter 4
Optical RegenerateandForward .
Relaying Technique
FSO communication systems are strongly' affected by the atmospheric turbulence
fading and background illumination. In Chapter 3, the OAF relaying technique has
been introduced as a powerful method to mitigate the degrading effects of atmospheric
turbulence. But growing background noise remains as the major drawback of free
space optical relaying systems. In electrical AmplifyandForward relaying technique
at each relay the optical signal along with the background noise is converted to an
electrical signal which is amplified by an electrical amplifier. After amplification, the
electrical signal is modulated by a photodiode and retransmitted to the next relay.
In an FSO system, the optical intensity the field distribution of the signal is an
alyzed. By elllploying an optical band pass filter (BPF) at the beginning of each
optical relay, the outofband frequency components of the accumulated background
noise are eliminated, however, the inband noise comp<,ments still remain in the sys
tem. In this chapter the Optical RegenerateandForward (ORF) relaying technique
100
1
McMaster University  Electrical Engineering
Chapter 4. Optical RegenerateandForward Relaying Technique
is proposed to eliminate the background noise at each relay. In ORF, the quality of
the received noisy signal is optically restored by a regenerator. In the next section,
the internal structure of an ORF relay is described in detail.
4.1 ORF Relay Structure
The internal structure of an ORF relay is shown in Fig. 4.1. When a regenerator is
used in an FSO system where broadband white noise is accumulated, an optical BPF
centered at the signal frequency Wo is required. This filter is placed at the input of the
regenerator so as to reject the noise outside the signal spectrum [63]. The bandwidth
of the filter is assumed to be equal to the bandwidth of the photodetector. The
photo detector bandwidth is dependent on the transmission bit rate and is found so
that the best BER is achieved at the receiver. Here it is assumed that the optical filter
does not attenuate the optical signal. The filtered signal is amplified with gain G
1
so
as to adjust the average power of the signal to a level suitable for the regeneration
process. The optimum gain G
1
depends on the average received SNR at each relay
and for a given average transmit power and background illumination irradiance is
constant. The regeneration process is performed by a regenerator. The rege.nerator
suppresses the noise in zeros and the amplitude fluctuations in ones of optical data
streams. The regeneration process will be analyzed in detail in Section 4.2. The
regenerated pulses are amplified by another amplifier with gain G
2
and transmitted
through the next relay. The gain G
2
is adjusted so that the average transmitted
power at the output of each relay is equal to the average transmitted power at the
source, Pt.
A sample of Gaussian pulses propagating through an ORF relay are shown in
101
McMaster University  Electrical Engineering
Chapter 4. Optical RegenerateandForward Relaying Technique
r1
I I
A
I B Ie
I Optical I
I ill I
~ R ~
I egenerator
: BPF
~                                                     
Figure 4.1: The typical structure of an ORF relay.
Fig. 4.2 shows at different points. As shown, the regenerator refines the signal by
removing the noise at zeros and amplitude fluctuations in ones. The regenerated
Gaussian pulses carry less noise and have a small displacement compared with the
original pulses. The regeneration process and internal structure of a regenerator are
described in the next section.
4.2 Optical Regenerator
To analyze the regeneration process for suppressing the signal background, th€ typical
structure of a regenerator is shown in Fig. 4.3. The regeneration process is performed
by utilizing the effects of selfphase modulation (SPM) on the signal in a nonlinear
(NL) medium followed by an optical filtering at a frequency of wf which is shifted
with respect to the carrier frequency of the input data Woo The optical fiber with high
nonlinearity coefficient 'Y is employed as the nonlinear medium used in the regenerator.
(4.1)
where n2 (m
2
/vV) is the nonlinearindex coefficient, c (m/s) is light velocity, and Aelf
(m
2
) is the effective core area.
102
McMaster University  Electrical Engineering
Chapter 4. Optical RegenerateandForward Relaying Technique
'"
:::>
4
3
2
1
..9
3
2 1
o
t(s)
1
(a) TI.'ansmitted Pulses
0.8
1\
0.6
0.4
0.2
a f\ ,r' f\ /\, fI.rv..
3 2 1 0 1 2
t(s) x 10
10
(c) The Regenerator Input (Point B)
0.01
,I
0.005
2 a
t(s)
(b) The ORF Relay Input (Point A)
8
6
4
2
..93 2 1 0 1
t(s)
(d) The ORF Relay Output (Point C)
Figure 4.2: Gaussian Pulses propagating through an ORF relay at different points.
103
lVIclVIaster University  Electrical Engineering
Chapter 4. Optical RegenerateandForward Relaying Technique
1,
1
1
1 J
1
t)
1
!\
1
J
1
1
NLMedium
1
1
1
m
1
1
{UO
f
1
1
1
1 Gaussian BPF
1
                               ~
Figme 4.3: The internal structme of a regenerator.
The dependence of the refractive index in nonlinear media, e.g optical fiber, on the
signal intensity causes selfphase modulation (SPlVI) which governs spectral broaden
ing of optical pulses. The propagation of optical pulses inside a singlemode fiber
in terms of normalized amplitude S(z, T) is quantified by the nonlinear Schrodinger
equa"tion as [64]
(4.2)
where z is the propagation distance, T = t/T
o
is normalized time scale and To is the
pulse width. The parameters /32 and (TF(m
1
) account for the fiber dispersion and
fiber loss respectively. The dispersion length LD and nonlinear length LNL are given
as
T,2
L D =  ~
/32
where PM is the peak power of optical signal given by (2.5).
(4.3)
To analyze the effects of fiber nonlinearity on the selfphase modulation, fiber
dispersion is ignored and the dispersion coefficient is set to zero. By substituting
104
McMaster University  Electrical Engineering
Chapter 4. Optical RegenerateandForward Relaying Technique
(32 = 0 in (4.2), the nonlinear Schrodinger equation can be solved as
( 4.4)
where Ut(O, t) is the field envelope at z = 0 and is defined from (2.4) for a single
Gaussian pulse as
and
with
1 exp (az)
Zeff = '..:...
a
(4.5)
(4.6)
(4.7).
As indicated in (4.4), the phase shift ¢NL (z, t) induced by SPM is intensitydependent
while the squared field envelope governed by IUt(z, t)12 remains unchanged ((32 = 0) .
From (4.6), the nonlinear phase shift ¢Ndz, t) increases with the propagated distance
z. The parameter Zeff is an effective distance that because of fiber loss is less than"
z (in the absence of fiber loss, a = 0, Zeff = z). Moreover, the time· dependence
of Ut(O, t) and consequently ¢Ndz, t) induces SPM spectral broadening. In general,
temporally varying phase shift induces temporally varying frequency shift, therefore
in NL medium the amount of spectral broadening differs across the pulse. By taking
the time derivative of phase shift, the instantaneous frequency shift with
respect to the central frequency Wo is obtained as
(t) = I a¢Nd
z
, t) I = (IU (0 t)12) I Zeff
WSF:M at at t, LNL
(4.8)
105
McMaster University  Electrical Engineering
Chapter 4. Optical RegenerateandForward Relaying Technique
This difference is induced by SPM and increases with the propagated distan<:;e, z.
In other words, new frequency components are continuously generated as the pulse
propagates through the fiber. These SPMinduced frequency components broaden
the pulse spectrum. Figure 4.4 shows the spectrums of the data stream presented in
Fig. 4.2 for different propagated distances z inside a NL optical fiber.
As shown, the signal power is distributed over more frequency components as the
pulses propagate down the fiber. Clearly, the length of the fiber is a main factor in
determining the desired spectrum bandwidth at the output of the NL medium.
By substituting (4.5) into (4.8), the spectrum broadening of a Gaussian pulse
propagating inside a lossless optical fiber (Zeff = z) is obtained as
(4.9)
By plugging the value of LNL from (4.3) into (4.9) and taking the time average over one
bit interval T
b
, the average spectrum broadening for one Gaussian pulse is calculated
as [63]
( 4.10)
where Ilwo/27r = l/T
FwHM
is the 3 dB spectral bandwidth (TFWHM is the full width
at half maximum) and Ip = Pt/A
eff
is the average pulse intensity which varies for
different pulses. It is clear that the spectrum broadening increases by the initial
bandwidth Ilwo, the nonlinear refractive index n2, the average pulse intensity Ip and
the propagated distance z.
After the nonlinear medium, the pulses pass through a Gaussian optical filter [65]
whose center frequency wf is shifted with respect to the input signal carrier frequency
106
1
McMaster University  Electrical Engineering
Chapter 4. Optical RegenerateandForward Relaying Technique
7 0 0 0 1 .         ~   ~   ~  
6000
5000
4000
N
~
3000
2000
1000
.94
2 o
f(Hz)
(a) Transmitted Pulses (z = 0)
(c)z=4km
2
2
o
f(Hz)
(b) z = 2 km
o
f(Hz)
(d) z = 6 km
2 4
X 1011 .
Figure 4.4: The Gaussian pulse spectrums for different propagated distances, z.
107
McMaster University  Electrical Engineering
Chapter 4. Optical RegenerateandForward Relaying Technique
Wo by a certain value .6.w
s
hift:
W f = Wo + .6.wshift
(4.11)
As mentioned before, there exists a displacement between the regenerator input pulses
and the filtered ones. This misalignment originates from the shifted center frequency
of the Gaussian filter.
When the pulse spectral broadening is small so that the spectral bandwidth of the
selfphase modulated signal at the output of the NL medium, WSPM = .6.wo + .6.wSPM,
satisfies the following equation
WSPM .
2 < .6.wshift,
(4.12)
the pulse is rejected by the filter. In the regenerator, this occurs when the pulse
intensity is small, e.g. noise in zeros. If the pulse intensity is high, e.g. at ones, so
that the new spectral bandwidth satisfies
( 4.13)
depending on the amount of pulse broadening; the filter center frequency, and the filter
bandwidth .6.wf, a part of the SPMbroadened spectrum passes through the filter and
the rest of them are rejected. The spectral width of the filtered pulse is determined
by the filter spectral bandwidth. By changing the filter spectral bandwidth, the
bandwidth of the SPMbroadened spectrum changes. If .6.wf = .6.wo, the output
pulsewidth is the same as the input pulsewidth.
108
1
McMaster University  Electrical Engineering
Chapter 4. Optical RegenerateandForward Relaying Technique
In the case where the input pulse intensity Ip is very high, Ip > > I
er
, so that the
pulse spectrum broadens extensively, WSPM > > .6w
s
hift, the intensity of the output
pulse is independent of the input pulse intensity. Therefore, a pulse transfer function
for regenerator in terms of output pulse intensity versus input pulse intensity can be
established as
lout = { 0
Ie
if Ip < Ier
(4.14)
if Ip > Ier
where Ie is the constant output pulse intensity and Ier is the critical pulse intensity
which is adjusted to a level so that compromise a trade off between removing .the noise
at zeros and suppressing the amplitude fluctuations at ones. If leT' is chosen to be
very small, the regenerator cannot completely remove the noise at zeros. Otherwise,
if the critical intensity is selected too big, the pulse spectrum broadens excessively so
that the signal is distorted by generation of new outofband frequency components.
On the other hand, by extensively broadening the pulse spectrum, the more parts
of the signal are rejected by the filter that imposes more power loss to the system.
The transfer function expressed by (4.14) is an ideal transfer function. Fig. 4.5
shows a typical transfer function of an ideal regenerator and its perfect performance
in removing noise at zeros and suppressing the amplitude fluctuations at ones.
As expressed earlier, the regeneration process imposes a considerable power loss
to the signal by rejecting a major portion of the pulse spectrum. Therefore, an
amplifier is required after the regenerator to boost up the signal and retrieve its
average power to the average transmitted power at the source, Pt. As shown in Table
4.2, the ORF relay receives the noisy signal and refines it via regeneration process,
then amplifies and retransmits it through the next channel. Section 4.3 provides the
109
lVIcNIaster University  Electrical Engineering
Chapter 4. Optical RegenerateandForward Relaying Technique
lout lout
i"
rrr,  r    ,...       
\
Output Signal
Figme 4.5: Typical transfer function of an ideal regenerator.
..
r
t
numerical simulations on ORF relaying technique and compares its performance with
OAF multihop systems.
4.3 Numeric;:}l Results and System Performance
In this section, the perforIn.anc.e of various ORF n:lLLltihop FSO communication sys
tems is investigated. Because of the long simulation time of MATLAB , a Qfactor
110
McMaster University  Electrical Engineering
Chapter 4. Optical RegenerateandForward Relaying Technique
estimation method is utilized to calculate the BER of FSO systems. When the ad
ditive noise has a Gaussian distribution, the Qfactor approximation is widely used
instead of MonteCarlo method for BER simulations. As demonstrated in Chapter
3, wideband FSO systems suffer from background noise and their maximum accessi
ble distances are limited due to accumulated background noise. On the other hand,
electrical processing limits the bandwidth of high rate optical systems, therefore all
optical relaying techniques are attractive. In this thesis, ORF relaying technique is
proposed as a powerful method for optically suppressing the effects of background
noise at high bitrates. Here, the performance of ORF multihop systems are inves
tigated at 10 Gbps and compared with the performance of OAF multihop systems.
The degrading effects of atmospheric turbulence are not considered in ORF systen'ls.
The considered FSO system has the same specifications as the OAF FSO systeln
that .are previously introduced in Chapter 3. The NL medium specifications are chosen
as what Mamyshev considered in [63] for experimental simulations. At BR= 10Gbps,
a fiber of length z = 8 km with effective core area Aeff = 45 JLm
2
at .A = 155"0
nm is used. The Gaussian filter inside the regenerator has the 3 dB bandwidth of
f1Wj /27r = 29 GHz, and the filter frequency offset with respect to the input signal
carrier frequency is f1wshift/27r = 100 GHz. The gain of the first amplifier G
1
is
adjusted for each launch power and hop distance so that the best performance at the
receiver is obtained.
4.3.1 Q.Nlctor Est"inlation
Simulating the BER of an ORF ll1ultihop system takes an extremely long time. Q
factor is frequently utilized to evaluate the performance of optically amplified systems
111
J
McMaster University  Electrical Engineering
Chapter 4. Optical RegenerateandForward Relaying Technique
[66]. In ORF relaying systems due to fiber nonlinearity, the accumulated noi::;e at
the receiver is not exactly Gaussiandistributed. However, it can be shown that a
Gaussian approximation is a reasonable estimation for the noise distribution. The
BER of an OOK modulated signal at the receiver is related to the Q factor as [66]
(4.15)
where the complementary error function erfc(x) is defined as
00
erfc(x) = In J e
t2
dt.
(4.16)
x
The Q factor is also expressed as
( 4.17)
where ~ l O ( ~ 1 ) and aD (a1) are the means and the standard deviations of zeros (ones)
respectively. In numerical simulations, the mean ~ O (hl1) of a pulse stream is simply
found by getting average over the values obtained by sampling pulses at the middle
of the bit intervals, t = kT
b
/2, k = 0,1,2, .... By squaring the sampled values and
taking average, the variance aD (a1) of the pulse stream" is also obtained.
Fig. 4.6 shows the BER of an ORF multihop system with the total communicating
distance of 3 km where one ORF relay is placed at the middle of the link. BER of
the system is simulated using both MonteCarlo (MC) and Qfactor estimation (QF)
methods. The total number of transmitted bits is N = 214. In the all MC analyses
considered in this work, at least 100 errors in the received bit streams are required
112
McMaster University  Electrical Engineering
Chapter 4. Optical RegenerateandForward Relaying Technique
Figure 4.6: BER of an ORF FSO system with M = 1 and LT = 3 km, obtained by
MC and QF methods.
to be able to rely on the received BER. By sending N = 214 bits and taking average
on about 20 different iterations, the smallest reliable BER obtained by MC method
will be around BER=1O
3
. However, simulation results indicate that 214 sent bits are
enough to rely on BER as low as 10
5
when QF method is employed. As seen from
the figure, the performance of the system driven by Qfactor estimation is very close
to the MonteCarlo simulation result (less than 1 dB offset). Therefore, in the rest of
this section, QF is considered as a reliable method used to calculate BER of different
113
MclVlaster University  Electrical Engineering
Chapter 4. Optical RegenerateandForward Relaying Technique
Table 4.1: Different system. configurations for Fig.4.7 (LT = 3 km)
lVlaTker Ll (kIll") L2 (km)
B 1 2
~ 1.5 1.5
2 1
ORF FSO systems.
4.3.2 Optimal Relaying Configuration
As proved in Chapter 3, the equallyspaced relaying configuration provides the best
performance at the receiver of OAF systems. For ORF systems, there exists no
explicit mathematical relations between the hop lengths and the average received
SNR at the receiver. Therefore) numerical simulations is utilized to found the best
configurations. Fig. 4.7 corresponds to the BER of three different ORF systems. In
all systems, the total communicating distance is LT = 3 km and one relay is placed
between the source and destination nodes. The distance bet\;veen the source and ORF
relay is denoted by Ll and the distance between the 0 RF relay and the destination is
L
2
. Table 4.1 summarizes the configurations of the systems considered in this figure.
From Fig. 4.7, the lowest BER corresponds to the system where the ORF relay
is placed at the middle of the link (Ll = L2 = l.5 km). Hence, it can be assumed
that equallyspace relaying may also provide one of the optimal configuration1') for
ORF multihop systems, although enough evidences do not exist for its proof. In
the follmving simulations, ORF relays are arranged in serial equallyspace relaying
configurations.
114
,
!
MclVlaster University  Electrical Engineering
Chapter 4. Optical RegenerateandForward Relaying Technique
Figure 4.7: BER of different ORF multihop FSO systems with M = 1, LT = 3 km,
and BR= 10 Gbps. (The plot descriptions are provided in Table 4.1)
4.3.3 Fixed Total Communicating Distance
In Chapter 3, it was shown via numerical sill1Ulations that increasing the number of
relays between the source and destination nodes while keeping the total communicat
ing distance fixed, improves the system performance (Fig. 3.13). Fig. 4.8 compares
the performance of ORF multihop systems with OAF systems. In this plot, the BER
of ORF FSO systems for NI = 1 and N! = 2 relays are illustrated while the total com
municating distance is fixed (LT = 3 km). Table 4.2 summarizes the configurations
115
McMaster University  Electrical Engineering
Chapter 4. Optical RegenerateandForward Relaying Technique
5 10 15 20
25 30 35
Figure 4.8: BER of different ORF multihop FSO systems with All = 1 and 1\11 = 2 for
a fixed total communicating distance of LT = 3 km and, BR= 10 Gbps. (The plot
descriptions are provided in Table 4.2)
of the systems considered in this figure.
From Table 4.2, at a given BER= 10
5
, by replacing OAF relays with ORF relays,
the system gains 2.93 dB (All = 1) and 3.51 dB (All = 2) improvement in SNRo with
respect to OAF relaying technique. It is expected that by increasing the number of
relays, ORF systems outperform OAF systems more rapidly. As mentioned in Chapter
3, in OAF systems, utilizing more relays injects more accumulated background noise
to the system. Thus by increasing the number of relays, BER improvement decreases
116
JVlcJVlaster University  Electrical Engineering
Chapter 4. Optical RegenerateandForward Relaying Technique
Table 4.2: Different system configurations for Fig.4.8 (LT = 3 km)
lVlarker ]1/[ Lk (km) Relaying Technique SNRo (dB)
& 0 3 DT (no relay) 33.93
x  l' 1.5 OAF 31.46
 EJ
2
1
2
1
1.5
1
OAF
ORF
ORF
29.84
28.53
26.33
(Table 3.4). In ORF systems, background noise is ideally eliminated at each relay,
hence background noise does not propagates through the channel and its average
power is almost constant at all relays. On the other hand, increasing the number
of relays is equivalent to decreasing hop distances and channel loss and therefore
the average received SNR increases by the number of relays. In sunlnmry, in ORF
systems, it is expected that in a fixedlength link, BER improves by increasing the
number of relays and its improvement increases rapidly.
4.3.4 Combination of OAF and ORF Relays
In this section, the performance of various multihop systems composed of both OAF
and ORF relays is investigated. Consider a 3 km FSO link with two equallyspaced
relays placed between the source and destination nodes (the hop distances are 1
km). Fig. 4.9 provides the BER of various relaying schemes whose configurations are
summarized in Table 4.3.
As illustrated in Fig. 4.9, by replacing one OAF relay by an ORF relay, i.e.
System 2 + System 3 (4) in Table 4.3, 1.01 dB (1.33 dB) gain is achieved in SNRo
at BER= 10
5
. Surprisingly, if the signal is first amplified by the first relay and then
117
1
,
McMaster University  Electrical Engineering
Chapter 4. Optical RegenerateandForward Relaying Technique
25 30 35
Figure 4.9: BER of different nmltihop FSO systems with M = 2 for a fixed total
communicating distance of LT = 3 km and, BR= 10 Gbps. (The plot descriptions
are provided in Table 4.3)
regenerated at the second one (System 3) or it is first regenerated and then amplified
(System 4), the performance of both systems is nearly the same. At high SNRos,
System 4 has slightly better performance. Because the received SNR is higher and
the regenerator suppresses t.he amplitude fluctuations at ones more evenly, in other
words, regenerator has better performance at high SNR
o
.
By replacing both OAF relays by ORF relays, i.e. System 2 (3) + System 5,
the system performance improves by an additional 2.50 dB (2.18 dB). This value is
118
~
MclVIaster University  Electrical Engineering
Chapter 4. Optical RegenerateandForward Relaying Technique
Table 4.3: Different system configurations for Fig.4.9 (LT = 3 km, NI = 2)
System No. lVIarker First Relay Second relay SNRo (dB)
1
e
no relay no relay 33.93
2
 EJ
OAF OAF 29.84
3 +
OAF ORF 28.83
4
~
ORF OAF 28.51
5
B
ORF ORF 26.33
almost twice the gain achieved by replacing only one OAF relay with an ORF relay
(System 2 7 System 3 or 4).' The reason can be simply explained: the regenerator
eliminates the background noise at each relay therefore by using all ORF relays. a
negligible amount of background noise remains in the system. In general, regenerators
have better performance at higher SNRs. In System 3, the signal is first aInplified by'
an amplifier in the first relay, and then the noisy signal is regenerated at the second
relay. Obviously, System 5 benefits from the regeneration process at the second relay
more than System 3. In System 4, the noiseless signal regenerated at the first relay is
amplified along 'with the collected background light at the second relay. This amplified'
background noise considerably deteriorates the perform.ance of System 4. System 5
achieves totally 3.57 dB gain with respect to System 2 and 8.18 dB gain with respect
to the direct transmission (System 0.)
4.3.5 Maximum Accessibie Communicating Distance
The OAF relaying technique has been introduced as a powerful method in mitigating
the effect of atlnospheric fading and consequently increasing the maximum achievable
distance in FSO con1Hmnication systems. Although atmospheric fading is suppressed
119
I
j
lVIclVlaster University  Electrical Engineering
Chapter 4. Optical RegenerateandForward Relaying Technique
Table 4.4: Different system configmations for Fig.4.10 (P
t
= 500 mvV, No = 2 X 10
15
\iV/Hz, and BER= 10
5
)
System
. :Marker N umber of Relays Relaying
LT
Relaying Distance Lk
No. I'll Technique (km) (km)
0
e
0 DT 1.70 1.70
1
x
1 OAF 2.20 1.10
2
 EJ
2 OAF+OAF 2.62 0.87
3 * 1 ORF 3.08 1.5"4
4
B
2 ORF+ORF 4.48 1.50
5
v
2 ORF+OAF 3.40 1.13
by OAF relaying technique, amplifyandforward FSO systems encounter an essential
difficulty originating from accumulating background noise at each relay. The ORF
relaying technique is proposed as an effective technique to reduce the background noise
and increase the maximum. accessible distance in FSO communication systems. Fig.
4.10 provides BER of various OAF, ORF and hybrid OAF /ORF multihop systems.
The average transmit power is P
t
= 500 m VV for all schemes and atnlOspheric fading
effects are neglected in this study. The number of relays at each system is either one
or two and relays are arranged in an equallyspaced relaying configmation. Table 4.4
summarizes the configmations of the systems considered in this figme.
By comparing ORF and OAF relaying systems in Fig. 4.10, it is realized that ORF
technique improves the maximum achievable distance several times more than OAF
technique. By replacing OAF relays with ORF relays in System 1 and 2, i.e. System.
1 + System 3 and System 2 + System 4 in Table 4.4, the ma.'{imum accessible
distance increases respectively by 0.88 km and 1.86 km. Despite OAF technique,
in ORF systems as the number of relays increases, maximum accessible distance
120
\
McMaster University  Electrical Engineering
Chapter 4. Optical RegenerateandForward Relaying Technique
Figure 4.10: The maximum accessible communicating distance of different multihop
FSO systems with lv'I = 1 and lv'I = 2 when P
t
= 500 m VV, No = 2 X 10
15
W 1Hz,
and BR= 10 Gbps. (The plot descriptions are provided in Table 4.4)
improves more rapidly, e.g. by utilizing only one ORF relay (System 3), LT increases
by 1.38 km, by using another ORF relay (System 4), additional 1.4 km increment
is obtained in LT' As mentioned in Chapter 3, by increasing the number of relays,
maximum LT occurs at shorter hop distances so that the system overcomes the effects
of accumulated background noise. By deploying ORF relaying technique, background
noise is mostly eliminated at each relay and shortening hop distances no longer is
required, e.g. from Table 4.4, the hop distances in the' considered ORF systems are
121
McMaster University  Electrical Engineering
Chapter 4. Optical RegenerateandForward Relaying Technique
almost 1.5 km for both System" 3 and System 4. Under this circumstance, in ORF
systems, the maximum accessible distance extends linearly by the number of relays.
Ideally, if atmospheric turbulence effects are neglected, ORF relaying technique can
greatly extend the total communicating distance LT by desirably increasing the number
of relays.
The relaying system which consists of two equallyspaced ORF and OAF relays
(System 5) has slightly better performance than the system with just one ORF relay
(System 3). This small improvement corresponds to the shorter hop distances to
compensate the effects of background noise. For long hop distances where the received
SNR at relays is very small, employing OAF relays not only does not improve the
system performance but it also deteriorates it (Fig. 3.13). As shown in Fig. 4.10,
at very long hop distances (Lk ~ 2.2 km), the performance of System 3 is better
than performance of System 5, because at very low received SNRs, using OAF relays
is equivalent to injecting background noise to the system. Although System 5 takes
advantage of one more OAF relay, totally, performance of System 3 and System 5
are very similar even at short hop distances. This similarity accentuates the superior
performance of ORF technique to OAF method and its dominant role in hybrid
systems for improving the system performance. Next section is spent to compare
resistivity of both techniques in the presence" of background noise.
4.3.6 Fixed hop Lengths
In order to compare ORF technique resistivity to background noise with OAF tech
nique, Fig. 4.11 is provided. In the systems under consideration, hop lengths are fixed
Lk = 1 km. The total communicating distance increases by subsequently adding a
122
McMaster University  Electrical Engineering
Chapter 4. Optical RegenerateandForward Relaying Technique
100 .. .. ... .. ..
....  ..... ". _..... _........ .  ....... _ ... _... . .... _.  .. . ...... ... .
.......;
\i
5 10
............ .. .....
.. ···········c: \ \
1 ' 0 .
::::X:: ..
..... ..... ....... ... :
....... , .. ,
25
...... L ..... \.: ....
.. ;q ..
,
, .
. ___, ...... 1 ..... .
.. \ .... :.\ ...
\ : \
...  .. \. .: 1:
:::::',:i::Q
 .. .\...
· ...... or ... \
........ ( I
·······t· "'T
, ! Q)
, 1 __ ... l ...
::: :.Jf ....,:.:.:.:.:.:.:.: .•..,'.:. ::: :1:::':
::::c:::
30 
Figure 4.11: BER of different ORF FSO systems with different number of relays M
for a fixed hop distance of Lk = 1 km and BR= 10 Gbps. (The plot descriptions are
provided in Table 4.5)
new relay to the previous system. The amount of accumulated background noise in
creases by the number of relays. In the OAF relaying technique, it is shown that this
increment in background power deteriorates the system BER and limits its maximum
accessible distance. However Fig. 4.11 indicates that increasing the number of relays
in ORF systems does not influence the system performance because the accumulated
background noise is eliminated at each relay. Table 4.5 summarizes the configurations
of the systems considered in this figure.
From Fig. 4.11, by replacing ORF relays with OAF relays, at a given BER= 10
5
,
123
~
McNIaster University  Electrical Engineering
Chapter 4. Optical RegenerateandForward Relaying Technique
. Table 4.5: Different system configurations for Fig.4.11 (Lk = 1 km)
Marker
1\,£
LT(km) Relaying Technique SNRo (dB)
0
3 4 OAF 31.29
 8
2 3 OAF 29.84
x
I 2 OAF 27.75
e
3 4 ORF 25.96
B
2 3 ORF 26.31
~
1 2 ORF 26.33
the average transmit SNRo improves by 1.44 dB, 2.92 dB and 5.36 dB when the num
ber of relays are 1\,£ = 1, 1\,£ = 2, and 1\.1 = 3 respectively. In OAF systems, by
increasing the number of relays (increasing the total communicating distance) more
average transmit power is required to guarantee a specific BER at the receiver. How
ever when ORF technique is used, by subsequently adding more relays, the system
BER relTlains nearly unchanged. This feature makes the ORF relaying technique
a distinguished method for removing the background noise in multihop' FSO com
munication systems and leading to significantly extending the FSO communication
distance coverage.
4.4 Conclusion
In Chapter 3, the OAF relaying technique has been developed to reduce the degrading
effects of atmospheric turbulence. It has been s11m'ln that by Initigating atmospheric:
fading effects via reducing hop distances, the link coverage can be extended. But the
magnitude of improvement is limited due to growing accumulated background noise
during the channel. In this chapter, ORF relaying technique has been proposed as
124
McMaster University  Electrical Engineering
Chapter 4. Optical RegenerateandForward Relaying Technique
a powerful method to optically remove background noise at each relay. A very weak
turbulence fading condition (C;, < 1 X 10
17
m
2
/
3
[52]) has been considered under
which the atmospheric fading effects can be neglected over distances up to 1 km.
Assuming this condition, BER performance of different ORF systems operating at
BR= 10 Gbps are investigated and compared with the performance of OAF systems.
The results indicate that the distance coverage of ORF FSO links can be significantly
extended by increasing the number of relays because collected background light at
each relay is mostly eliminated by a regenerator. Qualitatively, by employing only
one ORF relay, the total communicating distance increases respectively by 1.38 km
and 0.88 km with respect to direct transmission ruid a similar OAF multihop system.
In ORF systems, background noise is mainly eliminated by the relay and does
not propagate through the channel, this feature which is also involved in the decode
andforward relaying technique causes ORF systems to significantly outperform OAF
systems, especially for large number of relays and longer distances. Although the
ORF technique provides superior performance to OAF technique, it requires more
sophisticated equipment such as an automatic gain controller, optical Gaussian fil
ters, adjustable nonlinear optical medium, two amplifier at each relay, etc, which
increase the complexity and implementation costs of the system. The regeneration
process is highly sensitive to the incident signal SNR, the preamplifier adjusted gain,
and the Gaussian filter frequency offset and bandwidth. Also as mentioned, the re
generator output signal has a small displacement with respect to the original pulses
that in some applications this delay might be undesired [63]. The error propagation
is another performance limiting factor in ORF systems. In general, the amplitude
of the regenerator output pulses are not completely equal and they have negligible
125
J
i
McMaster University  Electrical Engineering
Chapter 4. Optical RegenerateandForward Relaying Technique
fluctuations. If the regenerator is not well adjusted and does not work at its opti
mal operating condition, it can even add extra amplitude fluctuations to the output
pulses. These fluctuations, e.g. at bit" 1", propagate through the channel and by
subsequent regeneration in the following relays lead to wrong detection of the pulse
(completely rejecting the pUlse). These challenges in the proposed ORF system pro
vide motivation to finding methods to the system stability al'ld reduce the
implementation complexity.
126
Chapter 5
Conclusion and Future Work
5.1 Conclusion
This thesis presents new optical relaying techniques to mitigate atmospheric turbulence
induced fading effects and eliminate background noise in free space optical (FSO) com
munication systems. The main contributions of the thesis are proposing alloptical
amplifyandforward (OAF) and regenerateandforward (ORF) relaying techniques
and applying them to relay assisted FSO systems.
In order to define an all optical relayassisted FSO system, a new channel model
is developed which characterizes the variations of intensity and phase of the optical
signal during wave propagation. An additive AWGN channel is assumed in which
background illumination is the dominant source of noise. Three primary factors have
been considered to model the free space channel effects: 1) atmospheric attenuation
which includes both absorption and scattering contributions; 2) lognormal fading
under weak atmospheric turbulence conditions; and 3) propagation loss due to optical
beam spreading through optical channel. The BeersLambert law is modified to find
127
McMaster University  Electrical Engineering
Chapter 5. Conclusion and Future Work
the atmospheric attenuation factor applied to the optical field envelope. Since the
optical field envelope is analyzed, both amplitude and phase of atmosphericinduced
fading are statistically modeled and new definition for propagation loss is defined.
The numerical results show that the new model for propagation loss is close to the
conventional model (geometric loss), however, the proposed model provides more
accurate estimation of beam propagation loss especially over short ranges (a few
hundred meters). The atmospheric channel effects are assumed as multiplicative
complex terms which are multiplied by the field envelope and induce intensity and
phase fluctuations at the receive aperture. These fluctuations degrade the overall
system performance.
The OAF relaying technique is proposed as a powerful technique for mitigating
the atmnspheric turbulenceinduced fading while all relaying processes, e.g. amplifi
cation, filtering (as needed), etc, are in optical domain. The alloptical AF
relaying technique allows users to take advantage of high data rate (wide bandwidth)
optical transmissions over longer distances. It also provides less delay and complexity
in implementation rather than the previously considered AF communication systems
which need OE and EO conversions at each relay. Performance of OAF systems has
been considered for two bit rates BR= 1.25 Gbps and BR= 10 Gbps under no fad
ing and weak fading effects. It is numerically shown that by increasing the number
of relays between source and destination, hop distances decrease and consequently
distancedependent atmosphericinduced fading is mitigated. In fact by employing
more OAF relays, longer communicating distances are accessible for a given aver
age transmit power. However, distance improvements slow down by increasing the
number of relays because the collected background light at each relay is accumulated
128
\
I
NIcMaster University  Electrical Engineering
Chapter 5. Conclusion and Future Work
during propagation and totally more noise is added to the signal. The accumulated
background noise is the major limiting factor in improving the maximum accessible
distance of OAF multihop systems.
The decodeandforward (DF) relaying method has been widely used in RF sys
tems to remove the effects of background noise at each relay. Due to its superior
performance to AF relaying technique, DF method is recently applied to FSO sys
ten'lS as well. The considerable loss in data rate, complexity of encoding/decoding
processes, and delay induced by electrical processors at each relay accentuate the
need for an innovative technique to remove background noise completely in optical
domain. In this thesis, the ORF relaying method is developed as a promising tech
nique for removing background noise at each re.lay. The ORF systems operating at
high bit rate BR= 10 Gbps are considered under a very weak atmospheric turbulence
condition where atmospheric fading effects can be neglected over short distances (less
than 1 km). Under this assumption, BER performance of ORF systems are analyzed.
The simulation results indicate that by employing ORF technique, the background
noise can be eliminated at each relay that results in accessing greatly longer commu
nicating distances. Despite the OAF method, in ORF systems the communicating
distance improves steadily as the number of relays. In other words, the ORF tech
nique used in multihop FSO systems makes high data rate communications possible
over a greatly extended communicating distance by employing reasonable number of
relays. Although the ORF technique distinguishably outperforms the OAF method in
terms of BER performance, it requires complex relay structures and adjustable gain
amplifiers that imposes n:lOre complexity and implementation costs to the system.
129
McMaster University  Electrical Engineering
Chapter 5. Conclusion and Future Work
5.2 Future Work
In this thesis, alloptical multihop FSO communication systems are studied and its
~ BER performance is analyzed for the first time. Therefore many future directions
still remain to be investigated in the future, some of which are described as follows .
• Although most FSO manufactures utilize the bit error rate (BER) as a stan
dard performance metric to characterize their products, the BER is not itself a
comprehensive metric to evaluate the performance of FSO systems. In multi
hop systenls, the endtoend outage occurs when a local outage happens during
one of intermediate hops. The intensity fluctuations induced by atmospheric
turbulence, blockage of the lineofsight link due to temporary obstructions, e.g.
birds, temporary disalignmcnt between two consequent relays, etc, lead to an
endtoend outage in the whole system. Therefore, analyzing outage probability
factor would be a critical step in performance analyses of alloptical multihop
FSO systems .
• In Chapter 4, the error performance of ORF FSO systems has been investigated
when the atmospheric turbulenceinduced fading effects are neglected. This
assumption holds for short hopdistances (less than 1 km) and under weak
atmospheric turbulence conditions. In order to utilize ORF technique to support
longhaul applications, atmospheric turbulence effects must been taken into
account. The ORF FSO system performance has been analyzed in the presence
of background noise and it has been shown that in the absence of fading, ORF
technique greatly extends the total communicating distance. The next work
can be devoted on investigating the limitations that atmospheric turbulence
130
:McMaster University  Electrical Engineering
Chapter 5. Conclusion and Future Work
imposes to the maximum achievable distance of ORF FSO systems .
• The performance of multihop FSO systems has been analyzed in the absence/presence
of weak atmospheriC turbulenceinduced fading. The performance of FSO links
is very dependent upon weather conditions therefore investigating the perfor
mance of multihop FSO systems under different turbulence conditions is another
interesting approach for the future work. On the other hand, the system per
formance can be analyzed for different amounts of collected background power
at the receive apertures when the ambient illumination varies during the day
and for different locations.
• The cooperative diversity scheme is another future approach for increasing both
the reliability and distance coverage of FSO systems. The information redun
dancy at the receiver of such systems can improve the error performance. On
the other hand, since each path experiences a different atmospheric condition,
it is expected that the endtoend outage probability decreases.
131
I
1
1
Appendix A
Optimum relaying configuration
A.I Dualhop relaying system
Recall the expression for JI(x) given in Eq. (3.26):
It will be shown that
f{ (x) I X= Li = 0
f{'(x) > 0, for 0 < x < LT. (A.2)
132
McMaster University  Electrical Engineering
Appendix A. Optimum relaying configuration
A.I.1 J{( L!) = 0
f{ (x)
(A.3)
=? f{(x)
(AA)
From (AA), it is clear that at x = L:J the first derivative of fl (x) is zero: f ~ ( L:J) = o.
A.I.2 J{'(L{) > 0
First it will be shown that the second derivative is nonnegative at x = L:J. From
(A A)
f{' (x)
(A.5)
133
McMaster University  Electrical Engineering
Appendix A. Optimum relaying configuration
By substituting x = L{ in (A.5)
f{' 20: [L; eD:LT/2(L; + 2)] + 4 + 1)] 2(K:SNRo)l eD:LT
e
oVr
/
2
 e
O
£T/2), Li. HaLT H] . (A.6)
In the FSO systems which are considered in this work, the atmospheric atten
uation coefficient is 0: = 0.43 dB/Inn which in the linear system is equivalent to
0: = 1O
4
(m1). In typical FSO systems, K: 10
5
and the average transmit SNRo
varies between 0 dB and 40 dB in our simulations or equivalently 1 < SNRo < 10
4
in the linear system. The term c in (A.6) is the coefficient of the dominant factor
and is described as c = (0:/2  (K:SNRo)leD:LT/2). Since the two other terms in
(A.6) are positive, if c > 0, therefore f{'(LT/2) > 0 and the proof is complete. For
o :s; LT :s; 40 km, the exponential term changes between 1 :s; eD:LT/2 :s; 7.39, in other
words eD:LT/2 < 10 for a long range of LT. On the other hand, for the above mentioned
range of SNR
o
, 10
5
:s; (K:SNR
o
)l :s; 10
9
. At the worst case where SNRo = 0 dB
((K:SNR
o
)l = 10
5
), the maximum LT over which fr(L
T
/2) is nonnegative can be
obtained as
0: SNRo=O 10
4
5 1O4xLT
C = (  (/l:SNRo)leaLT/2) > 0 >   10 e 2 > 0 =? LT < 35 km.
2 2
(A.7)
For higher SNR
o
, f{'(LT/2) is nonnegative over longer distances. Since, in this work,
shorter distances and higher SNRo are considered in the simulations, it can be assumed
that fr(L
T
/2) is nonnegative in the desired range of LT'
134
McMaster University  Electrical Engineering
Appendix A. Optimum relaying configuration
A.1.3 ff'(x) > 0 for x E [0, L
T
]
N ow it will be shown that (x) > 0 is also nonnegative over the range of typical LT
values where is nonnegative. In (A.5), the first two terms are nonnegative
for every arbitrary value of 0 < x < LT. However the third term depends on the
value of a, L
T
, 1'1" and SNR
o
. In A.1.2, the range of LT over which f{'(LT/2) is
nonnegative has been discussed. In this section, it will be shown that f{'(LT/2) is
less than f{'(x) for every x =I L{ and 0 < x < L
T
, and therefore, is nonnegative
over the region where f{'(L
T
/2) > O. Consider the third term in (A.5).
g(x) = [(LT  2x?  2x(LT  x)] , (A.8)
the first and second order derivatives of g( x) are easily obtained as
g'(x)
I
 = 4(LT  2x)  2(LT  2x) L = 0
ax
2
g"(x) = 8 + 8 = 16 > O. (A.9)
From (A.9), g(x) is convex over all possible values of x and x = L2T is a global
minimum point for g(x). That is, g(x) > g(LT/2) or equivalently f{'(x) > f{'(LT/2)
for every x inside the region (0 < x < LT)' Since, f{'(L
T
/2) is nonnegative for the
desired range of L
T
, therefore is also nonnegative for every 0 < x < L
T
, and
x = L2T is a global minimum point for h (x) over the desired range of LT.
135
l\/IcMaster University  Electrical Engineering
Appendix A. Optimum relaying configuration
A.2 (p + 1 )hop relaying system
From (3.36)
ajp+1(x) = (p+ 1) [1 + (_X_)2]P
ax p+ 1
x. + _2_(_X_))]
p+l p+1 p+1 p+1
x [1 + (A;SNRo)lect(LTX)(L
T
 X)2]
[1+ (p: S]'*'
x [(A;SNRo)lect(LTx) (a(LT  X)2 + "2(L
T
 x))] . (A.ll)
Define XM = LT/p + 2. For x = (p + l)XMl the first order derivative is zero
136
.j
I.
McMaster University  Electrical Engineering
Appendix A. Optimum relaying configuration
From (A.11), the second order derivative is derived as
= p [1 + (_X_)2jPl
fu p+1
x + _2_(_X_))]
p+1 p+1 p+1 p+1
x [1 + (Il:SNRo)lect(LTX)(L
T
 x?]
[
ct( x ) (0: X 2 2 x) ]
x e pH p + 1 (p + 1) + p + 1 (p + 1 )
[1+ ("SNIlo)l (p: sr
x [(Il:SNRo)lect(LTX) (o:(L
T
 x? + 2(LT  x))]
x e p+l ( + ()
[
ct( x ) (0: X)2 2 x)]
p+1 p+1 p+1 p+1
+ [1+ ("SNlloj1c"(, ;,) (p: l)'r
x [1 + (Il:SNRo)lect(LTX)(L
T
 x?]
x L:1 (0:(P:1)2+
2
(P:1)) + (p2:
1
(p:1)+ P!l)]]
(p+ 1) [1+ ("SNllo) C: l)'r
x + _2_(_X_))]
p+1 p+1 p+1 p+1
x [ect(LTX) (o:(L
T
 x? + 2(LT  x))]
+ [1 + 1 ) 21
P
+l
\Jl T J.. / J
x [ect(LTX) [0: (o:(L
T
 X)2 + 2(LT  x)) + (20:(LT  x) + 2)]] . (A.13)
137
.j
McMaster University  Electrical Engineering
Appendix A. Optimum relaying configuration
By substituting (p + l)xM into (A.13), a similar equation to (A.6) is found
For p = 0, (A.14) results in (A.6). Here, XM depends on the number of relays
1\1£ = p + 1
i
ll (LT )
p+l p+2
By comparing (A.6) and (A.15), it is evident that by increasing the number of relays
at a given SNR
o
, a new acceptable region for LT is obtained which is larger than what
obtained in (A. 7). In this region, equally spaced relaying provides maximum received
SNR at point XM. Using the same approach expressed in Section A.1.3, it is possible
to show that jp+ 1 (x) is also convex over all 0 < x < LT for typical values of system
parameters and XM is the global optimum point over the acceptable region of LT'
138
I
i
Bibliography
[1] J. R. Barry, Wireless Infrared Communications. Kluwer Academic Publisher,
1994.
[2] M. R. Pakravan and M. Kavehrad, "Indoor wireless infrared channel character
ization by measurements," IEEE Transaction on Vehicular Technology, vol. 50,
2001.
[3] T. Komine and M. Nakagawa, "Fundamental analysis for visiblelight communi
cation systems using LED lights," IEEE Transaction on Consumer Electronics,
vol. 50, pp. 100107, February 2004.
[4] S. S. Muhammad, P. kohldorfer, and E. Leitgeb, "Channel modeling for terres
trial free space optical channels," in Proceedings of the International Conference
on Transparent Optical Networks, vol. 1, pp. 407410, july 2005.
[5] A. Harris, J. J. Sluss, H. H. Refai, and P. G. LoPresti, "Alignment and tracking
of a free space optical.communications link to a DAV," in Proceeding of the 24th
Digital Avionics Systems Conference (DASCj, pp. l.C.21 to l.G29, 2005.
[6] E. Leitgeb, K. Zettl, S. Muhammad, N. Schmitt, and W. Rehm, "Investiga
tion in free space optical communication links between unmanned aerial vehicles
139
~
I
M.A.Sc. Thesis  your name
Bibliography
lVlcMaster  Electrical Engineering
(UAVs)," in Proceeding of the 9th International Conference on Transparent Op
tical Networks (152155, ed.), July 2007.
[7] D. Giggenbach, J. Horwath, and B. Epple, "Optical satellite downlinks to opti
cal ground stations and highaltitude platforms," in The 16th 1ST JvIobile and
Wireless Comm1mications Summit, pp. 14, 2007.
[8] C. G. C.C. Chen, "Impact of random pointing and tracking errors on the design
of coherent and incoherent optical intersatellite communication links," IEEE
Transactions on Commu·nications, vol. 37, pp. 252260, March 1989.
[9] A. Polishuk and S. Arnon, "Optimization of laser satellite communication sys
tem with an optical preamplifier," Journal of Optical Society America, vol. 21,
pp. 13071315, July 2004.
[10] R. M. Gagliardi and S. Karp, Optical Communications. John Wiley and Sons
Inc., second ed., 1995.
[11] fSONA Optical \i\Tireless. [Online]. Available: http://www.fsona.com.
[12] MRV Optical Communication Systems™. [Online]. Available:
http://www.mrv.com.
[13] Proxim Wireless. [Online]. Av.ailable: http://www.proxim.com.
[14] GigaBeam Corporation. [Online]. Available: http://www.gigabeam.com.
[15] A. A. Youssef, Capacity and Signaling for FreeSpace Optical Channels. PhD
thesis, MclVlaster University, October 2009.
140
M.A.Sc. Thesis  your name
Bibliography
McMaster  Electrical Engineering
[16] G. P. Agrawal, Fiber Optic Communication Systems. John Wiley and Sons Inc.,
third ed., 2002.
[17] D. Kedar and S. Arnon, "Urban optical wireless communication networks:
the main challenges and possible solutions," IEEE Comm'llnications lvlagazine,
vol. 42, pp. 82S7, May 2004.
[18] F. Nadeem, V. Kvicera, M. Awan, E. Leitgeb, S. Muhammad, and G. Kandus,
"Weather effects on hybrid F80/RF communication link," IEEE JO'llrnal on
Selected Areas in Communications, vol. 27, pp. 16871697, December 2009.
[19] D. A. ·Rockwell and G. S. Mecherle, "Optical wireless: lowcost, broadband,
optical access," in fSONA white paper, 2001.
[20] LightPointe™, "Lightpointe unveils performance enhanced Flightlite™G FSO
wireless bridging solution.," in http://www.lightpointe.com. April 2009.
[21] 1. 1. Kim, B. McArthur, and E. J. Korevaar, "Comparison of laser beam propaga
tion at 785 nm and 1550 nm in fog and haze for optical wireless communications,"
Proceedings SPIE, vol. 4214, pp. 2637, February 2001.
[22] International Electrotechnical Commission. (2010). [Online]. Available:
http://www.iec.ch.
[23] X. Zhu, "Freespace optical communication through atmospheric turbulence
channels," in IEEE Transactions on Communications, vol. 50, August 2002.
[24] E. Biglieri, J. Proakis, and 8. Shamai, "Fading channels: informationtheoretic
and communications aspects," IEEE Transaction on Information Theory, vol. 44,
pp. 26192692, 1998.
141
M.A.Sc. Thesis  your name
Bibliography
McMaster  Electrical Engineering
[25] G. R. Osche, Optical Detection Theory for Laser Applications. New York, NY,
USA: John Wiley and Sons, 2002.
[26] J. W. Goodman, Statistical Optics. A WileyInterscience Publication, 2000.
[27] T. Tsiftsis, H. Sandalidis, G. Karagiannidis, and N. Sagias, "Multihop free space
optical communications over strong turbulence channels," in IEEE International
Conference on Comm'LLnications ICC, pp. June 2006.
[28] M. Uysal, S. Navidpour, and L. Jing, "Error rate performance of coded freespace
optical links over strong turbulence channels," IEEE Communications Letters,
vol. 8, pp. October 2004.
[29] X. Zhu and J. Kahn, "Performance bounds for coded freespace optical com
munications through atmospheric turbulence channels," IEEE Transactions on
Communications, vol. 51, pp. 1239, August 2003.
[30] X. Zhu and J. Kahn, "Markov chain model in maximumlikelihood sequence
detection for freespace optical communication through atmospheric turbulence
channels," IEEE Transactions on Communications, vol. 51, pp. March
2003.
[31] E. Lee and V. Chan, "Part 1: Optical communication over the clear turbulent
atmospheric channel using diversity," IEEE Journal on selected areas in commu
nications, vol. 22, pp. November 2004.
[32] M. Safari and M. Uysal, "Relayassisted freespace optical communication," in
IEEE Transactions on Wireless Communications, yol. 7, pp. 122008.
142
M.A.Sc. Thesis  your name
Bibliography
McMaster  Electrical Engineering
[33] K. Kiasaleh, "Turbocoded optical PPM COl11.HlUnication systen1.s," Journal of
Lightwave Technology, vol. 16, pp. 1826, January 1998.
I [34] E. Forestieri, Optical communication theory and techniques. Springer, 2004.
[35] 1. Djordjevic, B. Vasic, and M. Neifeld, "Power efficient LDPCcoded modulation
for freespace optical communication over the atmospheric channel,"
in Conference on Optical Fiber Communication and the National Fiber Optic
Engineers Conference (OFCjNFOEC), pp. 13, Nlarch 2007.
[36] B. Vasic, 1. Djordjevic, and R. Kostuk, "Lowdensity parity check codes and iter
ative decoding for longhaul optical communication systems," Journal of Light
wave Technology, vol. 21, pp. 438446, February 2003.
[37] A. Paulraj, R. Nabar, and D. Gore, Introduction to SpaceTime Wireless Com
munications. Cambridge University Press, 2003.
[38] E. Shin and V. Chan, "Optical communication over the turbulent atmospheric
channel using spatial diversity," in IEEE Global Telecommunications Conference
(GLOBECOM), vol. 3, pp. 20552060, March 2002.
[39] S. Hass and J. Shapiro, "Capacity of \Yireless optical communications," IEEE
Journal on selected areas in communications, vol. 21, pp. 13461357, October
2003.
[40] J. Akella, M. Yuksel, and S. Kalyanaraman, "Error analysis of multihop free
space optical communication," in IEEE International Conference on Communi
cations ICC, vol. 3, pp. 17771781, May 2005.
143
M.A.Sc. Thesis  your name
Bibliography
McMaster  Electrical Engineering
[41] M. Hasna and M.S. Alouini, "Outage probability of multihop transmission over
Nakagami fading channels," IEEE Communications Letters, vol. 7, pp. 216218,
May 2003.
[42] J. Laneman and G. "Vornell, "Energyefficient antenna sharing and relaying for
wireless networks," in IEEE Wireless Communications and Networking Confer
ence (WCNC), vol. 1, pp. 712, 2000.
[43] J. Laneman, D. Tse, and G. Wornell, "Cooperative diversity in wireless networks:
Efficient protocols and outage behavior," IEEE Transactions on Information
Theory, vol. 50, pp. 30623080, December 2004.
[44] G. Karagiannidis, T. Tsiftsis, and H. Sandalidis, "Outage probability of relayed
free space optical communication systems," Electronics Letters, vol. 42, pp. 994
995, August 2006.
[45] C. Datsikas, K. Peppas, N. Sagias, and G. Tombras, "Serial free space optical
relaying communications over gammagamma atmospheric turbulence cliannels,"
IEEEjOSA Journal of Optical Communications and Networking, vol. 2, pp. 576
586, August 2010.
[46] G. Karagiannidis, A. Tsiftsis, and R. Mallik, "Bounds for multihop relayed com
munications in nakagamim fading," IEEE Transactions on Communications,
vol. 54, pp. 1822, January 2006.
[47] M. Hasna and M.S. Alouini, "A performance study of dualhop transmissions
with fixed gain relays," IEEE Transactions on Wireless Communications, vol. 3,
pp. 19631968, November 2004.
144
M.A.Sc. Thesis  your name
Bibliography
McMaster  Electrical Engineering
[48] lV1. Karimi and M. NasiriKenari, "BER analysis of cooperative systems in frec
space optical networks," Journal of Lightwave Technology, vol. 27, pp. 56395647,
December 2009.
[49] A. Acampora and S. Krishnamurthy, "A broadband wireless access network
based on meshconnected freespace optical links," IEEE Personal Communi
cations, vol. 6, pp. 10709916, October 1999.
[50] PMCSierra. [Online]. Available: http://www.pmcsierra.com.
[51] U. Adar, T. Kuzniz, 1. Haber, and M. Tur, "An alloptical multihop (cascaded)
high bit rate wireless communication field trial," in Optical Fiber Communication
Conference and Exhibit (OFC), pp. 446448, March 2002.
[52] L. C. Andrews and R. L. Philips, Laser Beam Propagation through Random
Media. SPIEThe International Society for Optical Engineering, second ed.,
2005.
[53] 1. Kim, B. McArthur, and E. Korevaar, "Comparison of laser beam propagation
at 785 nm and 1550 nm in fog and haze for optical wireless communications," in
Proceeding SPIE, vol. 4214, pp. 2627, February 2001.
[54] D. Bushuev and S. Arnon, "Analysis of the performance of a wireless optical
multiinput to multioutput communication system," Journal of Optical Society
of America A, vol. 23, pp. 17221730, July 2006.
[55] D. L. Fried, "Optical heterodyne detection of an atmospherically distorted signal
wave front," Proceeding of IEEE, vol. 55, January 1967.
145
lVLA.Sc. Thesis  your name
Bibliography
McMaster  Electrical Engineering
[56] B. E. A. Saleh and M. C. Teich, Fundamentals of Photonics. John Wiley and
Sons Inc., 1991.
J [57] V. P. Lukin and V. V. Pokasov, "Phase fluctuations of an optical wave
;
which propagates in a turbulent atmosphere," Institute of Atmospheric Optics,
Translated from Izvestiya Vysshikh Uchebnykh Zavedenii, Radiojizika, vol. 16,
pp. 17261729, November 1973.
[58] V. 1. Tatraski, Wave Propagation in a Turbulent Medium. The McGrawHill
Book Company, Inc., 1961.
[59] H. Willebrand and B. S. Ghuman, FreeSpace Optics: Enabling Optical Connec
tivity in Today's Networks. Sams Publishing, 2002.
[60] J. W. Goodman, Introduction to Fourier Optics. The McGrawHill Compa
nies,INC, second ed., 1988.
[61] Y. Dikmelik and F. M. Davidson, "Fibercoupling efficiency for freespace op
tical communication through atmosphere turbulence," Applied Optics, vol. 44,
pp. 49464952, August 2005.
[62] S. B. Alexander, Optical Communication Receiver Design. SPIEThe Interna
tional Society for Optical Engineering, 1997.
[63] P. Mamyshev, "Alloptical data regeneration based on selfphase modulation
effect," in 24th European Conference on Optical Comm'llnication, vol. 1, pp. 475
476, 2024 September 1998.
[64] G. P. Agrawal, Nonlinear Fiber Optics. Academic Press, Inc., second ed., 1995.
146
j
M.A.Sc. Thesis  your name
Bibliography
McMaster  Electrical Engineering
[65] J. Berger, F. Ilkov, D. King, and A. Tselikov, "Widely tunable, naiTow optical
bandpass Gaussian filter using a silicon microactuator," in Optical Fiber Com
munication Conference, vol. 1, pp. 252253, March 2003.
[66] S. Norimatsu and M. Maruoka, "Accurate Qfactor estimation of optically am
plified systems in the presence of waveform distortion," Journal of Lightwave
Technology, vol. 20, pp. 1927, January 2002.
147
AllOptical Multihop FreeSpace Optical Communication Systems
ALLOPTICAL lVIULTIHOP FREESPACE OPTICAL COMMUNICATION SYSTElVIS
BY
SHABNAM KAZEMLOD, B.Sc.
A THESIS SUBMITTED TO THE DEPARTMENT OF ELECTRICAL
& COMPUTER ENGINEERING
AND THE SCHOOL OF GRADUATE STUDIES OF MCMASTER UNIVERSITY IN PARTIAL FULFILMENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF ApPLIED SCIENCE
©
Copyright by Shabnam Kazemlou, October 2010 All Rights Reserved
Master of Applied Science (2010) (Electrical & Computer Engineering) McMaster University Hamilton. Canada TITLE: AllOptical Mult~hop FreeSpace Optical Communication Systems AUTHOR: Shabnam Kazemlou B .. Steve Hranilovic Dr.Sc. 147 11 . (Electrical Engineering) Sharif University of Technology. Tehran. Iran SUPERVISOR: Dr. Ontario. Shiva Kumar NUMBER OF PAGES: xvi.
To those WllO were always tllere wllen I needed tllem .
The simulation resul~s indicate that OAF relaying technique mitigates the channel impairments and enhances the BER performance. High bandwidth. This thesis addresses innovative alloptical relaying techniques to mitigate the degrading effects of atmospheric turbulenceinduced fading by relaying data from the source to the destination using intermediate terminals. bit error rate (BER) performance of multihop OAF FSO systems is analyzed through MonteCarlo simulations. unlicensed spectrum. In alloptical relaying techniques. are deployed in multihop FSO systems to extend the nlaximum accessible communicating distance of high data rate wireless optical systems. By IV .Abstract FreeSpace Optical (FSO) communication systems have recently attracted considerable attention in lastmile applications. photo detection is performed once at the re ceiver and intermediate terminals process optical field envelopes instead of optical intensities. optical amplifyandforward (OAF) relaying and optical regenerateandforward (ORF) relaying. and high security have made them a good candidate for high data rate transmissions. The proposed techniques. ease of installation. By using the developed channel model. distancedependent atmospheric turbulence and channel loss degrade the optical link reliability and confine FSO systems to shorthaul applications. However. This major difference requires a new definition of channel model for propagation of optical waves through the atmosphere.
Replacing OAF relays by ORF relays extends the total communicating distance to 4. longer distances become accessible.48 km which is 1. however distance improvement decreases due to accumulating background noise at relays.66 km longer than the similar OAF FSO system.11 km. even longer distances are achievable. v . For example at high bit rate BR= 10 Gbps. The ORF relaying technique eliminates the received background noise at each relay and significantly outperforms OAF systems. In order to remove background noise effects. using two equallyspaced OAF relays during a 3 km turbulencefree link increases the total communicating distance by about 1.employing more relays. another optical relaying technique is developed. By deploying more ORF relays.
Kasra Asadzadeh. Shiva Kumar. Farid and Mr. Dr. whose endless helps have been a key factor for successfully accomplishing this work. This thesis would not have been possible without his efforts. My deepest thanks also go to my cosupervisor. I would like to send my special thanks to my family for their unconditional love and giving me the opportunity to follow my dreams. advice and encouragement have been key factors to accomplish this work. It is an honor for me to study and research under his supervision. Steve Hranilovic.I also extend my thanks to my lab mates. whose insightful knowledge and helpful ideas involved me in an innovative and interesting topic of research. Dr. whose support.i 1 Acknow ledgements I would like to express my heartily gratitude to my supervisor. Dr. Last but not least. VI . Ahmeq.
List of Notations Fiber effective core area Gaussian beam wave parameters a.p {3 Wave number inside fiber Fiber dispersion coefficient {32 Light velocity c Refractiveindex structure constant Phase front radius of curvature F Channel loss 9 Amplifier gain G Atmospheric attenuation ha Complex propagation loss hp Complex channel gain h Conventional geometric power loss Hg Propagation power loss Hp i(t) Photodetector current Optical intensity I(t) Wave number in freespace k Length of kth hop Lk Total communicating distance LT M Number of relays Fiber nonlinearindex coefficient n2 N(t. S) Gaussian noise distribution Pb Average background power pOptical peak power LlVl Average transmit power Photodetector responsivity Symbol duration Aeff en Vll .
TFWHM n UASE U(t) UF j. t) <pn(x) 'ljJ(r. z. H! 6wo 6wf 6Ws hift 6WSPM I ). c/J. /tx Wo wf WSPM c/J c/JNdz. t) 0" o"x O"s O"F T Bit interval Full pulse width at half maximum Spontaneous emission noise of amplifier Optical field envelope Fiber mode weight factor Gaussian beam radius Regenerator input pulsewidth Regenerator output pulsewidth Filter frequency offset SPM spectral broadening Fiber nonlinearity coefficient Wavelength Mean of intensity fluctuations Carrier Frequency Optical filter center frequency SPMbroadened spectrum Phase SPMinduced phase shift Kolmogorov spectrum Transverse field profile in freespace Transverse field profile inside fiber Optical beam wave Atmospheric attenuation coefficient Variance of intensity fluctuations Variance of phase fluctuations Fiber attenuation coefficient Atmospheric turbulenceinduced fading Vlll . c/J) \[1(7.'.c/J) 'l/JF(r.
..2.. .4.1 1..5 1.2. 10 11 Optical Multihop Transmission Multihop Communication Systems 1.3 1.4 1.2.2.. IX 12 14 14 16 . Atmospheric Turbulence Mitigation Techniques Forward Error Correction .1 1...2.4.4 .2 AmplifyandForward relaying technique DecodeandForward relaying technique.. .1 1.2.6 1.2 1. ..Contents Abstract Acknowledgements List of Notations 1 Introduction 1..3 1. MaximumLikelihood Sequence Detection Spatial Diversity . iv VI Vll 1 1 5 7 8 9 9 Atmospheric Channel Impairments: h(t) Background Illumination Noise: Ub(t)" .2 FreeSpace Optical Communication Systems FSO Systems and Channel Impairments 1.
1. 2. .3 1.1.4 Fixed Total Communicating Distance :t\/Iaximum Accessible Communicating Distance. .2 3. .2 2.3. . . . . . . .5 2. .1 3.6 2 Optical multihop relaying techniques proposed in the literature 17 Thesis Contributions Thesis Structure . . . . Fixed hop Lengths 55 59 70 71 85 92 98 Cond usion. . .1 Signal projection onto SingleMode Fiber (SMF) . . Atmospheric Turbulence Propagation Loss . 53 54 3 Optical ArnplifyandForward Relaying Technique 3. x . . Gaussianbeam wave propagation ABeD RayMatrix 26 27 28 31 33 35 39 43 50 The RelaytoRelay Channel Model 2. . Optimal Relaying Configuration Numerical Results and System Performance 3. .1 3.3.4.2. .2 Atmospheric Attenuation . . . . .3 2.1 2.1. . 20 23 25 Channel Model 2.5 1.1.4 2. . . .1. .3. . . .3 OAF Relay Structure .1.2 3.3 3. .4 Noise Projection onto SingleMode Fiber Conclusion .1 FreeSpace Optical Channel 2.3 2.
. A. Combination of OAF and ORF Relays Maximum Accessible Communicating Distance ..2. .2 (p systell1~ .4 5 Conclusion. .2 Conclusion . . .4 4.3. 127 130 132 A Optimum relaying configuration A.5 4. 132 133 133 f{(L{) = 0 .3.L T] 135 136 136 137 + I)hop relaying system f. .2 4.4 Optical RegenerateandForward Relaying Technique 100 4. . .1 5.3 ORF Relay Structure Optical Regenerator Numerical Results and System Performance 4. . Conclusion and Future Work 5. Fixed Total Communicating Distance .+I(X) .1 A.3 4.l Dualhop relaying A.6 101 102 110 111 114 115 117 119 122 124 127 Q. .lo3 A. Future Work. f.2 4. . . Fixed hop Lengths 4. .3.lol A.3.2 Xl . Optimal Relaying Configuration .1 4.Factor Estimation .lo2 A.1 4.2.3.+1(x) .3. f{'(L{) > 0 Jr(x) > 0 for x E [O. .
.. .8 An OAF Multihop FSO System with M= 2 and SNRo = 37 dB. . . . . .. . . . . . .. . . . . .6 3. .. .. ... .5 49 55 3... . . . .3 3.4 3. . .. . . .2 1.. . . Subsequent optical elements . . Xll . .4 An FSO Communication system .. Comparison between the conventional and new geometric power loss. 5 13 14 33 36 38 48 2... . . Estimation area for the conventional (Hg) and ·new propagation (Hp) loss. . The received optical SNR of an OAF FSO system with lvI = 1 and SNRo = 56 60 62 37 dB. Optical SNR of an OAF FSO system with M = 2 An OAF FSO system with lvI = p ... . . .7 3. . . . .. Multihop Transmission Cooperative Divetsity . . . .. .2 3. The RelaytoRelay Channel Model Optical intensity of Gaussian beam waves at z = 0 and z = 1 km.. . . . .. Optical AmplifyandForeward Multihop FSO Systems An OAF Ivlultihop FSO System with M= 1..List of Figures 1. . ..3 2. . . . .1 2. .1 3. . . .2 2. . .5 An AllOptical Multihop FSO Communication System Structure of OAF Relays . . . . . .. .. . . . . .1 1.3 2.. . . . 64 65 67 68 3. .
. and No = 2 X 1O15 W 1Hz. .14 The signal and noise power variations during a 3 km link operating at 76 75 74 BR= 1. . . . . . . . .25 Gbps with different number of relays M. . . . . .13 BER in low SNRo region for LT = 3 km and BR= 10 Gbps. (The plot descriptions are given in Table 3. (The plot descriptions I j ! are given in Table 3. . . . . . . and No = 2 X 1015 W 1Hz.25Gbps with different number of relays M.1) . . . . for LT = 3 km. . (The plot descriptions are given in Table 3. .1) . (The plot descriptions are given in Table 3. . . . . Pt = 40 m W. and different number of relays NI. . . . 78 3.. . . no fading effect is considered.1) .25 Gbps. .17 BER versus SNRo. . .15 The signal and noise power variations during a 3 km link operating at BR= 1. with fading effects. . . . . . . and different number of relays JVI. . . . . 68 3. . . . (The plot descriptions are given in Table 3. . .9 An OAF FSO system with M = p + 1. .1) . . . . 81 80 xiii . . . . . .16 BER versus SNRo. . . . . . 3. 79 3. . . . .11 BER versus SNRo for a 3 km link and different number of relays NI at 73 BR= 10 Gbps. 3.1) . . . . . .25 Gbps. 3. . . . . . . . .1) . Pt = 10mW.10 BER versus SNRo for a 3 km link and different number of relays NI at BR= 1. . . . . for LT = 3 km. BR= 1. . . . . . .. BR= 10 Gbps.. . 3. . . . . . . . . . 3. . . . .3. . . . . .25 Gbps. with fading effects. . . . . no fading effect is considered. . . . . .12 BER in low SNRo region for LT = 3 km and BR= 1. . (The plot descriptions are given in Table 3.
21 BER versus the total communicating distance. BR= 1. . .19 BER versus the total communicating distance. . . . . .25 Gbps.5) . for different number of relays ll.24 BER versus SNRo for the constant hop distance Lk = 1 km. (The plot descriptions are given in Table 3.3.25 effects. BR= 10 Gbps. . (The plot descriptions are given in Table 3. . . . and different number of relays M considering fading effects. . BR= 10 Gbps.20 BER versus the total communicating distance.22 BER versus SNRo for the constant hop distance Lk = 1 km. . P t = 500 m \iV. (The plot descriptions are given in Table 85 3. . Pt = 500 mW. and different number of relays JIll without fading effects. for different number of relays JIll. with fading effects. . (The plot descriptions are given in Table 3. . 3. . . for different number of relays JIll.25 Gbps. (The plot descriptions are given in Table 3.5) 91 3.15 VV 1Hz. 3.15 W 1Hz. LT (km) . . . . . No = 2 Gbps. . . for different number of relays JV[. No = 2 X 10. . BR= 1.8) . . and different number of relays JIll without fading effects.![. LT (km) . bit rate BR= 1. 3. with fading effects. Pt = 500 m W.23 BER versus SNRo for the constant hop distance Lk = 1 km.5) . . LT (km) . Pt = 500 mW. . . without f~ding X 10. No = 2 X 10. .25 Gbps. bit rate BR= 1. . . (The plot descriptions are given in Table 3. .15 W 1Hz. (The plot descriptions are given in Table 3.8) . . . 96 94 93 XIV . . bit rate BR= 10 Gbps. .8) 3.18 BER versus the total communicating distance.5) 90 3. No = 2 X 86 10. LT (km). . without fading effects. . . . .15 W 1Hz.
. . . . . . . . .. . LT = 3 km. . . . . . . . . . . . " The Gaussian pulse spectrums for different propagated distances. . . . . . .. BR= 10 Gbps. .2 4.1) . . (The plot descriptions are given in Table 3. . . . . . .3) . . (The plot descriptions are provided in Table 4. . . . 121 xv . . z. . ... (The plot descriptions are provided in Table 4. BR= 10 Gbps. .. Gaussian Pulses propagating through an ORF relay at different points. . 4. . . . Typical transfer function of an ideal regenerator. (The plot descriptions are provided in Table 4. . .1 4. .10 The maximum accessible communicating distance of different multihop FSO systems with 1\1£ = 1 and M = 2 when Pt = 500 m W.. bit rate BR= 10 Gbps.i[ = 1 and LT = 3 km. . (The plot descriptions are provided in Table 4. .. No = 2 X 10. . . . . . . . .2) . and BR= 10 Gbps.5 4..9 BER of different multihop FSO systems with M = 2 for a fixed total communicating distance of LT = 3 km and. . . .8 BER of different ORF multihop FSO systems with 1\1£ = 1 and 1\1£ = 2 for a fixed total communicating distance of LT = 3 km and. . . .8) 97 102 4.4 4. . . . . . 118 4. . . and BR= 10 Gbps. . . . . . .. . . . . . . .4) .15 W 1Hz. . . .. . .. . ..6 The typical structure of an ORF relay. . obtained 113 by Me and QF methods. . 104 107 110 BER of an ORF FSO system with }. . .3. .25 BER versus SNRo for the constant hop distance Lk = 1 km. . . . 115 116 4. . . .. . . . .. . 103 The internal structure of a regenerator. . and different number of relays lY[ considering fading effects. .3 4. . . . . 4. .. . . . . . . . .7 BER of different ORF multihop FSO systems with M = 1. . . .
.. (The plot descriptions are provided in Table 4.ber ofrelays 1\([ for a fixed hop distance of Lk = 1 km and BR= 10 Gbps..4... 123 XVI .....11 BER of different ORF FSO systems with different nUIl1.5) . .
Optical fiber communications have the potential to deliver high data rate applications for longhaul and network access.1 FreeSpace Optical Communication Systems A tremendous evolution has recently emerged in telecommunication systems. Freespace optical (FSO) communication is one of the most prevalent applications for wireless optical systems and refers to terrestrial lineofsight (LOS) optical transmission through the atnlOsphere [10]. ground to air (UAV) [5. Various reliable and costeffective techniques and technologies have evolved to deliver high data rate services to users. satellite to ground [7]. and intersatellite communications [8. visible light communications (VLC) [3]. 2].9]. Today's coml'nercial FSO systems provide bit rates ranging from 100 Megabits/second (Mbps) to 10 Gigabits/second (Gbps) over 1 . terrestrial links [4]. Beside optical fiber communication. wireless optical communication has recently emerged as a new technology to deliver various types of services such as indoor infrared wireless communications [1.~ i Chapter 1 Introduction 1. 6].
while MRV product TS lOGE can deliver 10 Gbps data over only distances up to 350 m. it still does not reach the demanded Gbps data rates used in optical networks. However. e. Microwave access via WiMAX technology has been introduced as a promising technology deployed in pointtomultipoint broadband wireless applications. etc.Electrical Engineering Chapter 1. RF communication links operating at high frequencies in the range 60 . its deployment requires digging and a rightofway license which makes fiber deployment expensive and even impractical in high density cities.g.25 Gbps transmission over distances up to 5400 n1. Optical fiber is one ofthe most reliable candidates for connecting the end user. buildings. e. the cost and the high attenuation in rain are the main challenges for these systems [15]. Introduction distances on the order of a few kilometers. offices. These highspeed systems are utilized in lastmile applications to connect buildings. Moreover. Although these systems are capable of delivering Gbps data rates. Although WiMAX technology provides high data rates. by increasing data rate.McMaster University . For example SONAbeam™1250M provides 1. e.86 GHz can be deployed to connect local sites to the fiber network. to high rate fiber networks without loosing data rate. link coverage decreases. As an alternative.. companies. they fail to match high speed.6 km [14]. Various communication systems exist to connect high data rate fiber networks to the end users. and business offices to the fiber network. SONAbeam™ 1250M by fSONA [11] and TereScope@lOGE (TS lOGE) by MRV [12]. Optical fiber systems also can be used in lastmile applications.25 provides 1. For a given system reliability.g. As an example.25 Gbps link over about 1. the Proxim product Tsunami™ MP8100 exhibits speeds up to 300 Mbps over a range of up to 8 kilometers (km) [13].commercial optical networks 1 2 .g.GiCORE G1. the commercial product GigaBeam .
Introduction working at data rates in excess of 1.25 Gbps data rate. accurate alignment is required between transmitter and receiver. and redeployment felxibility [16]. 1 J They also have the features of simple deployment. They are also used as wireless backhaul for WiMAX or WiFi networks [20]. The weather 3 . FSO link is an efficient alternative to a fiber system to establish high data rate transmission in populated city areas where laying fiber is too expensive or impractical.86 GHz band [18]. In addition. narrowbeam lineofsight FSO connections preserve the average transmit power which is limited in wireless communication systems.Electrical Engineering Chapter 1. Unlike \iViMAX which is mainly deployed in pointtomultipoint topologies. Compared to stateoftheart RF technologies such as GigaBeam . FSO links suffer from smaller attenuation in rain compared to RF links which operate in the 60 . In addition. These advantages introduce FSO communications as a promising candidate for high data rate transmissions. FSO links can be deployed as a highbandwidth bridge to connect local area networks (LANs) to longhaul widearea networks (WANs) or metropolitan area networks (IvIANs) [19].25 Gbps. Freespace optical (FSO) systems are licensefree with highbandwidth providing a costeffective and easytoinstall alternative to fiber optics and RF systems. IvIoreover. Beside the advantages of FSO communication links.McMaster University . FSO pointtopoint links provides higher data rate transmissions. commerciallyproduced FSO links provide higher data rates up to 10 Gbps.25 accommodating 1. it is important to point out the disadvantages and challenges in these systems.GiCORE G 1. diggingfree installation. Since the laser beam is extrelnely narrow. They fmther provide inherent security due to the nature of their directional and narrow beams that makes interception difficult [17].
In addition. at haze is 4. The optical beam is absorbed mainly by water particles and carbon dioxide molecules in the atmosphere (absorption). at a foggy weather condition. the degrading effects of randomly varying atmospheric conditions and background illumination noise on the performance of FSO systems are investigated and new techniques 4 .43 dB/km. etc. In this thesis. due to eyesafety regulations governed by International Electrotechnical Commission (IEC) [22]. FSO links mostly suffer from fog and haze.2.) Furthermore. In FSO systems. Ambient ilhunination is another challenge of FSO systems that is a huge amount of background noise is collected along with the incident data signal at the receive aperture [10]. and at fog is 43 dB/km [21].86 GHz radio frequency (RF) links [18]. That is. scattering deflects away a portion of the optical beam from the intended direct path to the receiver and is mainly caused by fog. haze. the propagating optical beam is attenuated over the optical channel due to absorption and scattering [18]. optical links also suffer from atmospheric refractive index random fluctuations (atmospheric turbulence) which induce intensity and phase fluctuations of optical beam at the receiver. Other than atmospheric attenuation. snow. rain. FSO transmitter must send at least 43 dB more power to overcome atmospheric attenuation during a 1 km link which is not feasible from practical stand point.McMaster University . the average tran~mit power is limited in FSO systems.3 dB/km. Introduction dependence of the optical channel is another issue so that the system performance highly depends on atmospheric conditions. This fluctuations degrade the system performance ( atmospheric turbulence is described in detail in Section 2.1. beam at clear air is 0.Electrical Engineering Chapter 1. In Chapter 3. Typical atmospheric attenuation of optical. it will be shown that these limitations confine FSO system transmissions to short range applications. Although rain is known as one of the most destructive atmospheric condition for 60 .
in practical direct FSO systems Intensity Modulation with Direct Detection (IlVI/DD) is utilized in which data are modulated in the instantaneous intensity of optical waves [23].MclVlaster University . The modulated optical wave is shaped and directed through the atmosphere via an optical lens. 1.. In the transmitter side. A forward propagating 5 . Introduction Input data Transmitter electronic unit I La". The intensity of optical waves corresponds to the squared absolute value of optical field envelopes.2 FSO Systems and Channel Impairments A pointtopoint FSO communication link is shown in Fig. 1.1: An FSO ComlTlUnication system are proposed to increase the accessible communicating distance of FSO systems. a laser diode modulates data onto an optical wave. For the sake of simplicity. Assume a laser oscillating at angular frequency w which is placed at z = a in the transmitter side. The field envelope is the timevarying amplitude of an optical wave which is modulated by data.1.Electrical Engineering Chapter 1. The next section presents a brief overview of FSO channels and summarizes the techniques which have been proposed so far to combat the aforementioned challenges. diode r I Receiver electronic unit I Output data w Figure 1.
z.Electrical Engineering Chapter 1.3) The electrical cmrent is sampled and detected inside the receiver electronic unit. the detected data are decoded to extract the sent data stream. both intensity and phase information of optical field envelopes are required for system analyses. (1. k is the prop agation constant along the positive zaxis in free space. t) = Ut(t) 'v'" 'ljJ(T.2) In such systems. An additive channel nlOdel is considered for the direct FSG system and can be 6 . In this thesis.1). Introduction wave along the positive zaxis can be expressed as \fJ(T. R (AjW) . and if. the transmit instantaneous intensity I (t) is defined as (1. a recelVlllg apertme followed by a lens collects and focuses the incident beam onto a photo detector .) ""v' exp[i(wt . if.McMaster University . However. From (1. (1. The photodetector coverts the collected optical power to an electrical cmrent. Z.. if.kz)] . data are modulated onto the instantaneous intensity of optical fields and at the receiver direct detection technique is performed to detect the signal (IMjDD). Finally. as will be discussed in Chapter 2. This cmrent is proportional to the instantaneous intensity of the received signal at the receiver scaled by the detector responsivity.1) v j Field envelope n·ansverse field profile Carrier where T = Jx 2 + y2 is the radial distance from the beam center line. at the receiver. is the phase.
1 Atmospheric Channel Impairments: h(t) The channel gain h(t) models the random fluctuations of the propagation path.2.. If the coherence time is much smaller than the symbol duration T.Electrical Engineering Chapter 1. Furthermore.4) where Ut(t) and Ur(t) are respectively the transmitted and received optical field envelopes at the transmitter and receiver (Fig. which is called fastfading atmosphoric turbulent channel. many different fading states occur within the transmission. Typical FSO fading coherence titne varies approximately between 103 and 10.1 and 1. 1. if the coherence time is equal or greater than T.2. The optical channel impairments will be discussed in more detail in Section 2. Ub (t) is the temporal profile of background illumination noise that are briefly overviewed in Section '1.. However. 1.1).2. h(t) is the temporal random fluctuations of optical channel.2. therefore FSO channels 7 . atmospheric turbulenceinduced fading) which results in fluctuation of the optical intensity at the receiving lens plane . Introduction expressed as [10] (1.McMaster University . the channel is a slowfading atmospheric turbulent channel [24]. The FSO optical link is a timevarying channel whose random fluctuations arise due to variation of atmospheric conditions. These intensity fluctuations degrade the performance of FSO communication systems. The time scale at which the fading state remains approximately constant is denoted as the coherence time.1. spatial and temporal variations of the air thermal inhomogeneities cause random fluctuation of the refractive index (i. Atmospheric molecules cause absorption and· scattering which attenuate the power of the light traveling through the atmosphere.e.1 sec [23] and typical symbol duration T is on the order of 109 sec. and.
1.5) The atmospheric attenuation ha depends on the weather condition and wavelength and includes both absorption and scattering effects (Section 2. = 8 .2 Background Illumination Noise: Ub(t) The additive noise in (1. atmospheric attenuation can be considered constant over a long time. sky.1. Since the weather changes slowly on the order of minutes to hours.McMaster University .3). arises due to ambient illuminations which is the dominant source of noise in FSO communication systems. moon. strong undesirable background radiation is also collected at the receive aperture.4).2).Electrical Engineering Chapter 1. Geometric spread of the optical beam hp is also constant for a given optical beam and distance (Section 2.1). incandescent bulbs. and atmospheric turbulence T. and· is statistically modeled as an additive white Gaussian noise in time and space with zero mean and variance CT.2.1.108 symbols. 1 and can be fonnulated as (1. h simply refers to the channel gain and implicitly indicates that fading is constant over many transmitted symbols. In addition to the desired signal. Despite ha and hp. geometric spread of optical beam hp. That is. atmosphericinduced fading factor T is a random coefficient which causes the received intensity fluctuates at the receive aperture (Section 2. Background noise mainly originates from amhent light coming from optical sources such as sun. the channel gain h(t) is assumed constant over 10 6 . In what follows. Intensity fluctuations at the receiver degrades the FSO systems performance. The constant channel gain h is a multiplicative factor [25] representing the amplitude and phase fluctuations which arise due to atmospheric attenuation ha. 1. etc. Introduction can be modeled as slowfading channels.
spatial diversity [31]. Introduction N o/2 [26]. Different techniques have been proposed to mitigate fading effects such as error control coding [28. it will be shown that background illumination noise can deteriorate the error performance and limit the link distance coverage. The intensity and phase fluctuations at the receiver.McMaster University . and multihop transmission [32]. In 9 . The collected background noise is processed along with the desired signal and degrades the overall system performance. 1. maximumlikelihood sequence detection (MLSD) [30].4 Forward Error Correction Various forward error correcting (FEC) techniques have been deployed in the literature to combat atmospheric turbulence degrading effects. reliability and distance [23]. The next section briefly denotes techniques and methods which have been proposed so far to combat atmospheric turbulence degrading effects. background illumination noise has not been separately considered as a limiting performance factor in FSO systems. Whereas. Forward error correction is aeeomplished by adding redundancy to the transmitted data through an error correction coding algorithm such as turbo codes [33].2.2. ReedSolomon (RS) codes [34]. that are induced by atmospheric turbulence random variations. block and convolutional codes [29]. Although. degrade FSO system performance.3 Atmospheric Turbulence Mitigation Techniques Atmospheric turbulence fading is one of the major impairments over FSO channels where the link range is longer than 1 km [27]. and lowdensity paritycheck (LDPC) codes [35]. in Chapter 3. This factor limits the link rate.Electrical Engineering Chapter 1. atmospheric turbulence mitigation techniques have been extensively investigated in literature. 29]. cooperative diversity. 1.
e.Electrical Engineering Chapter 1.5 MaximumLikelihood Sequence Detection Based on the knowledge of the joint temporal statistics of turbulenceinduced intensity fluctuations. maximum likelihood sequence detection (MLSD) has been proposed as another solution to mitigate atmospheric turbulence fading. Various suboptimal methods have been proposed to reduce computational complexity in MLSD method. The MLSD attempts to track the instantaneous state of intensity fluctuations and adjusts the detection threshold to the optimal value [30]. It has been numerically shown that turbo coding can achieve better BER I J 1 performance over atmospheric turbulence channel. 1. use of them in high data rate optical transmissions im. turbo coding is not suitable for highspeed optical transmissions because of its high complexity and long encoding/decoding time that imposes delays on the system [36]. singlestep Markov chain maximum likelihood detection (SMCIVIL) [30] and pilotsymbol assisted detection (PSAML) [30]. the MLSD computes the likelihood ratio of each of 2n possible sequences which imposes complex computational costs and long delays to decoder.poses high delay and complexity to encoding/decoding units which may not be practical.2. However.g. For transmitted sequences of length n. the performance of block codes and convolutional codes have been investigated and compared with turbo codes in the presence of weak atmospheric turbulence condition. Introduction [29].McMaster University . Although FEC is one of the best techniques to combat atmospheric turbulenceinduced fading. The suboptimal methods provide comparable error 10 . A promising alternative for mitigating atmospheric fluctuation is lowdensity paritycheck (LDPC) coding scheme which greatly improves BER performance of the system even under strong atmospheric turbulence condition [35].
employing multiple apertures at the transmitter increases the total average transmit power of the link and allows the system to cover longer distances. 11 . In practice. it may not be always possible to place the receivers sufficiently far apart [23]. All the proposed detection techniques provide significant perfonnance improvenlents over symbolbysymbol detection method at the cost of increased delays and computational complexity for simulating multidimensional integrals. This technique takes aavantage of multiple transmit/receive apertures that are placed far apart from each other so that each LOS path experiences a different fading condition [23].2. therefore. so that various receivers experience uncorrelated turbulenceinduced fadings. On the other hand.provides information redundancy at the receiver and improves system performance [37].Electrical Engineering Chapter 1. Introduction performance with respect to the MLSD method while they reduce the computational complexity of MLSD detection method. Each receiver collects the received optical beam(s) from different spatial angles.McMaster University . 1. Spatial diversity can significantly reduce the outage probability [38] and improve the outage capacity [39] 0f a multipleinput multipleoutput (MIM 0) link. FSO systems are based on pointtopoint transmission that is the transmitted signal directs to a specific receiver. the.6 Spatial Diversity Spatial diversity technique is another promising alternative for fading compensation that imposes less latency to encoding/decoding units.total average transmit power is the sum of maximum allowed transmit power of all transmit apertures. this way spatial diversity . To maximize spatial diversity performance. In optical spatial diversity systems. the receive apertures should be placed as far apart as possible from each other.
41].it/receive aperture scheme and designing encoding/decoding protocols increase the complexity and implementation costs of the system. Introduction Spatial diversity requires a lineofsight communication link between the source and destination terminals which may not be feasible in all situations. each terminal to create a spatial diversity. The next section provides a brief overview of this technique. spatial diversity in FSO syst~ms must involve multiple transmit and receive apertures at . Arranging relays in a serial configuration exempts FSO relays from being equipped by multiple transmit/receive apertures.Electrical Engineering Chapter 1.ploying multihop links in a parallel configuration. 1. As shown in Fig. by em. Multihop transmission is another alternative to combat the atmospheric fading effects and increasing freespace link coverage. Furthermore. Furthermore. In other words. multihop techniques can support an optical connection between two buildings which do not have a line of sight.3 Optical Multihop Transmission Multihop transmission relays the data signal from the source node (transmitter) to destination node (receiver) through intermediate terminals called relays or nodes and has been introduced as a promising technique to improve FSO links covenige and reliability through mitigating atmospheric fading impairment [40.3. Implementing multiple transm. in multihop systems. 1. more reliable FSO transmission over longer distances becomes achievable [32]. 43]. Despite spatial diversity which requires transmit/receive apertures to be positioned in a LOS configuration. new relaying scheme based on cooperative diversity is sleveloped [42. source and destination nodes are connected via intermediate relays and consequently LOS connectivity is not necessary. by subsequent deployment of slllaller multiple hops (hop is the distance between two relays).McMaster University . In this scheme 12 .
l'vIcl'vIaster University . a huge number of relays must be placed between the cmimlUnicating nodes. Also techniques which have been proposed in the literature to combat these challenges are described. 1. The relay (depending on relaying technique) performs optical operations on the signal and forwards it to the next relay.2. This scheme is typically illustrated in Fig. Introduction k=M+l Figure 1. Therefore. The next section presents a brief overview of multihop FSO systems and summarizes their achievements and drawbacks. only multihop systems are of great interests. to increase the total communicating distance of the link. the total communicating distance is determined by the length of the shortest multihop link. In this thesis innovative techniques are developed for the multihop schemes to increase the total comnlUnicating distance of FSO systems while guaranteeing a reliable high data rate. the source transmits an intensity lTlOdulated signal to the first relay. in cooperative diversity systems. The diversity of received signal intensities at destination improves the link endtoend reliability [32]. Obviously. that in practice. This continues until the transmitted signal arrives at the destination.Electrical Engineering Chapter 1. In serial relaying scheme. In multihop transmission relays are employed in series. intermediate relays involve only one transmit/receive aperture and are placed far enough from each other so that each receiver suffers from a different fading condition. may not be possible.2: l'vIultihop Transmission which is a special configuration of spatial diversity. Since the primary purpose of this study is to propose methods of increasing the total communicating distance of FSO systems. 13 .
the collected optical signal at the receiving lens of each relay is converted to a photocurrent via a photo detector... . namely: decodeandforward (DF) relaying and amplifyandforward (AF) relaying.Electrical Engineering Chapter 1. Figure 1. Then the amplified signal is optically modulated and retransmitted through the next hop.McIVIaster University .4 Multihop Communication SysteIIls Multihop transmission has been introduced as a powerful technique to n'litigate the fading and path loss effects in FSO systems.3: Cooperative Diversity 1. The channel state informationassisted (CSIassisted) gains have been widely utilized to analytically 14 .'i¥~~.":.S~~litter ""___~ ". the amplifier gain of each relay is determined based on the knowledge of the channel state of the previous hop.~.1 AmplifyandForward relaying t~chnique In the previously considered AF methods. In all AF FSO systems proposed so far. Introduction 1 :. 1.. . The electrical signal is amplified by an electrical amplifier with a gain specific to each relay.4. Two processing techniques are widely used in literature to implement multihop FSO transmission. ~ . 'l: .
(4)].lternative for CSIassisted gain is using fixedgain amplifiers in AF relays [47. (13)]. This approach has been widely used in analyzing the performance of RF communication systems [42. In [47]. The CSIassisted relays continuously estimates the channel fading amplitude which may not be practically feasible in all situations. Introduction study endtoend performance of optical [27. In the most AF multihop FSO and RF systems investigated in the literature. A practica. Relays using CSIassisted gains use instantaneous CSI of previous hop to control the gain of the relay and as a result fix the instantaneous output power of the relay [47.McMaster University .Electrical Engineering Chapter 1. 1 15 . 45]. 16] amplifyandforward multihop systems. the noise added to the signal at each relay propagates through transmission path. an ideal model for relay gains has been considered by which relay completely compensates the channel fading effects of the last hop. 45] and RF [11. However the fixedgain method low complexity and ease of deployment make it a promising alternative for CSIassissted AF systems. 44. 41] and recently attracted attention in FSO system studies [27. to make mathematical analyses tractable.l a. Although average BER of CSIassisted and fixedgain AF FSO systems over strong GammaGamma atmospheric turbulence have been investigated in [45]. 44. (5)]. regardless of the noise of that relay [47. This accumulated noise is another limiting factor in AF FSO systems as the number of relays increases. no explicit comparison has been made between performance of these to gaining techniques. it has been shown that RF systems with fixed gain relays have comparable performance with systems equipped with CSIassisted relays. In AF FSO systems.
therefore DF method outperforms BDF technique at a cost of higher complexity through decoding/encoding processes at each relay. The reconstructed signal is ideally a noiseless signal with average power equal to the transmitted power at the source. the simplicity of AF over DF. error grows rapidly as the number of relays. In DF technique. the noise at each relay is eliminated and does not propagate through the channel. in DF FSO systenls. In bit detectandforward (BDF) technique the bit sequence is detected and the detected "0" or "1" bits are retransmitted through the next hop without applying any error correction technique. This error accumulates over multiple hops for longlength hops.4. 16 . The primary concepts of these techniques can be applied to DF multihop systems. In [40]. In DF relaying. it is shown that for hops longer than 500m.2 DecodeandForward relaying technique Similar to AF systems. Moreover. The electrical signal is decoded and reencoded before retransmission. 49]. decoding and reencoding processes at each relay inject a time delay into the systenl which leads to an endtoend delay that increases by the number of relays. the bit stream is decoded. Although DF techniques outperform AF methods. 32. The incorrectly decoded bits are corrected before transmission via an error correcting technique. introduce AF relaying technique a more desirable candidate for pract'ical purposes. In the normal decodeandfqrward (DF) relaying technique which has been employed in all multihop FSO systems proposed in the literature [27. Introduction 1. At each relay a finite error occurs due to the decoding process.Electrical Engineering Chapter 1. at each intermediate relay the received optical signal is photo detected and converted to an electrical current. noise does not propagate through the channel and decoding error is less than BDF technique.McMaster University . In [48] various DF techniques have been proposed for cooperative diversity systems.
the outage probability increases. attenuation.McMaster University . however applying them to FSO systems emerged only recently. a lossfree strong turbulence fading channel is considered in [27] to analyze the endtoend outage probability of an FSO link employing OOK modulation with IM/DD scheme. In [32]. The results proposed in [44] justify that by increasing the number of relays while the hop lengths are fixed. an aggregated optical channel model including both pathloss and weak tmbulence fading effects was COl1 sidered and the endtoend outage probability of FSO systems utilizing DF relaying technique was presented. and geometric loss were not considered in the analysis. In [49]. the capacity of multihop FSO systems is analyzed from a networking standpoint and channel impairments such as atmospheric fading. it was demonstrated that the mean and variance of the error rate is smaller in multihop systems rather than single hop communication systems for the same link range and'launched power. The biterror rate (BER) of a DF FSO system was studied in [40].4. In contrast to [40]. outage probability of AF multihop FSO systems has been analytically calculated for strong turbulence fading channel modeled by Kchannel and Negative Exponential (NE) channel [44]. The results indicate that the outage performance of AF multihop systems degrades with increasing number of relays. in which atmospheric attenuation and geometric loss were the only effects included in the channel model.Electrical Engineering Chapter 1. Both DF and CSIassisted AF relaying techniques have been investigated. Using the same gain model.3 Optical multihop relaying techniques proposed In the literature lVlultihop techniques have been extensively studied for years in RF communications. A fixed spacing between the source and destination nodes has been considered in [32] and the endtoend outage performance of the system for 17 . In [40]. Introduction 1.
instead of taking advantage of wide bandwidth (high data rate) of optical systems to communicate a huge amount of information at a short time. Then the electrical signal is processed by electrical units performing either AF or DF relaying technique.Electrical Engineering Chapter 1. deploying them in all relays may be cost prohibitive. amplification or detection (decoding) process. e.g. Introduction different number of relays is simulated. i. the overall system data rate is confined to the speed of electrical processors. The processed electrical current is modulated through a high bandwidth laser and retransmitted to the next relay. A solution for increasing the overall data rate of the multihop FSO systems and accessing wider bandwidths in a costeffective way is utilizing alloptical processors at all relays.McNlaster University . an additional encoding/decoding delay is also added to the system and increases as the number of relays. In fact. each relay is equipped with analogtodigital (ADC) and digitaltoanalog (DAC) convertors operating at speeds up to a few Giga samples per second (Gsps). Although high speed electrical processors operating at Gbps rates are e~erging [50]. In allmultihop systems proposed so far.e.3 Gbps XMC Digitizer produced by PMCSierra® provides up to 3 Gsps data rate [50]. In addition. each relay is equipped with a photodetector which converts the incident optical power into a photocurrent. are performed by low speed electrical processors. ADX3500 . relaying procedures. In other words. It was shown that outage probability of a fixedlength multihop system improves by increasing the number of relays because hop lengths and consequently atmospheric turbulenceinduced fading decrease as the I j number of relays increase. in DF systems. Although alloptical AF relaying technique has been widely used in fiber optic 18 . Even if it is assumed that ADC (DAC) conserves the high data rate of the optical signal.
Introduction comn:lUnication systems [16]. In [51]. 19 . In this ex 1 periment an alloptical automatic gain controller (OAGC) is utilized to adjust the optical amplifier gain so that it compensates the power fluctuations of the received signal due to turbulence effects.McMaster University .Electrical Engineering Chapter 1. no analytical investigation has been performed on the performance analysis of alloptical multihop FSO systems impaired by atmospheric turbulence in longhaul applications. dualhop relayassissted configuration is examined for hop lengths of 75 m over which the effects of atmospheric turbulence are negligible. This approach is the primary focus of this thesis. A dualhop alloptical AF FSO system was implemented and its BER was obtained at differen~ bit rates BR= 2. Until now. The experimental results indicate that at higher bit rates the probability of error increases. its application in FSO systems was tested for the first time in [51].5 Gbps and BR= 10 Gbps.
Because alloptical components are employed in all relays and photo detection is only performed once at the receiver.e. The proposed method supports more accurate description of propagation loss rather than conventional geometric loss (equation (2.16)). After completely defining the channel model. Introduction 1. especially at short distances.Electrical Engineering Chapter 1. beam focusing and its projection onto an optical fiber at the beginning of each relay. Although. instead of analyzing optical power. atmospheric attenuation due to absorption and scattering and optical beamspread propagation loss. the magnitude and phase variations of optical fields are considered. a complete knowledge of field distribution is required at intermediate relays.McMaster University . The main contribution of this work in defining a new channel gain corresponds to deriving new complex propagation loss over the optical channel including beam propagation through FSO channel. i. Although data are modulated in optical field intensity (INIjDD). amplification and detection. the received optical field is simply coupled into a fiber and amplified by an ~ 20 .5 Thesis Contributions This thesis proposes novel relaying techniques for optical multihop communication systems in which all relaying processes. the error performance of alloptical amplifyandforward (OAF) multihop systems are investigated. a new channel gain is proposed to completely characterize both magnitude and phase variations of optical fields propagating through the channel. an aggregated channel model is considered which takes into account atmospheric turbulenceinduced fading. it is optical field intensity that is finally detected at the receiver and its phase does not contribute in detection process. are performed in optical domain. To analyze the optical field characteristics. In this work. In the OAF technique.
lvIcMaster University .e. but the amplified background noise accumulates through the channel and deteriorates OAF FSO system performance. A fixedgain protocol is deployed for mnplifier gain which only depends on the detenninistic structure of the preceding hop and guarantees a constant eyesafe average power at the output of each relay. An adjustable gain optical amplifier can also be used at each relay to compensate the last hop fading effects and provides a fixed instantaneous output power at each relay. new techniques need to be developed to eliminate background noise effects at each relay. the optical Rignal is reconstructed so that the collected background noise is eliminated at each relay. The OAF technique considerably il11. This is the major drawback of OAF FSO systems that limits the maximum accessible commlll1icating distance to a few kilometers even if a large number of relays is employed. it is nUInerically shown that by increasing data rate and/or the number of relays. Introduction optical amplifier. on the order of a few kilometers. a novel alloptical regenerateandforward (ORF) relaying technique is proposed to combat background noise. error performance improves slowly because more background noise is injected into the system. Atmospheric turbulence fading effects 21 . This type of OAF systems are expected tv outperform the fixedgain OAF systems. further investigations of these systems have been left for future. Despite their simple structure and ease of implementation. To introduce FSO links as a promising candidate for longrange applications. Also known as regenerative technique in RF communications [46]. OAF FSO systems are practically suitable for moderaterange applications. i. the collected background noise is amplified along with the desired signal. Regardless of the amplification method.proves the BER of FSO systems over moderate distances (up to 5 km). Then the amplified signal is retransmitted through the next hop. Through MonteCarlo simulations. In this technique.Electrical Engineering Chapter 1.
The optimal relaying configuration is determined based on numerical simulations. I 22 . Despite OAF systems whose maximum accessible distance is limited to a few kilometers.McMaster University . given atmospheric turbulence can be neglected. It is numerically shown that ORF technique significantly outperforms OAF technique and by increasing the number of relays its performance improves steadily.Electrical Engineering Chapter 1. This outstanding accomplishment is achieved for the highest available data rate 10 Gbps employed in commercial FSO communication systems. Introduction are not considered in this study and only the effectiveness of ORF technique ~n re moving background noise is investigated in terms of BER performance. FSO systems which utilize ORF relaying are able to access extensive communicating distances.
Electrical Engineering Chapter 1. the optical channel has been modeled as a multiplicative factor which takes into account weak lognormal atmospheric turbulence. Hence. The BER 23 . a weak lognormal atmospheric turbulence regime is considered whose logamplitude and phase are normally distributed and their statistical moments are mathematically described. propagation loss and atmospheric attenuation. the fiber output at the transmit relay and ends at the inside of the fiber at the receive relay. the maximum accessible communicating distance for different number of relays at two bit rates BR= 1. Introduction 1. It is assumed that the output laser beam at the transmitter has a Gaussian transverse profile. An aggregated optical channel between every two relays is defined which starts fro111. The average BER is selected as an evaluating metric for analyzing the error performance of different relaying schemes. Moreover. Finally. the equallyspaced relaying configuration is employed in all·of the considered OAF systems. a detailed description of optical wave propagation in free space is provided.6 Thesis Structure This thesis is organized as follows: In Chapter 2. Employing this configuration. It is mathematically proved by induction that equallyspaced relaying configuration provides the best average optical SNR at the receiver side of OAF systems and the result is numerically illustrated through Matlab simulations for the systems with· one and two relays. an optical amplifyandforward relaying technique is introduced and applied to various FSO system configurations. In Chapter 3.McMaster University .25 Gbps and BR= 10 Gbps is obtained through NlonteCarlo simulations. The paraxial approximation [52] is used to find the field distribution at the receive relay from which propagation loss due to beam spreading through the channel is obtained: The BeersLambert law [53] is modified to model atmospheric attenuation of optical fields instead of optical intensities.
24 . Since lVlonteCarlo (M C) simulations are timeconslUlling. In this study. Using equallydistance scheme. a very weak turbulence fading condition (C. In Chapter 4.17 m. the maximum accessible distance for different number of relays is found..that QF provides a close approximation of MC in such systems. the destructive effects of background noise in degrading BER performance of OAF systems is investigated and background noise is introduced as a major distancelimiting factor in OAF multihop FSO systems. an optical regenerateandforward (ORF) relaying technique is proposed to mitigate the degrading effects of background noise.2/ 3 [52]) has been considered under which the atmospheric fading effects can be neglected over distances up to 1 km. The simulation results show that ORF systems are able to remove the background noise completely at each relay when atmospheric fading is neglected. This property introduces the ORF technique as a superior method over the OAF in increasing the total communicating distance of FSO systems. an alternative method called Qfactor (QF) estimation is utilized for additive white Gaussian noise (AWGN) channels to accelerate the simulation process. Finally. The perforHmnce of MC and QF methods and are compared for a nonlinear dualhop ORF system. Introduction results are presented for no fading and weak fading atmospheric conditions and analyzed to contrust the effects of atmospheric fading.Electrical Engineering Chapter 1.. The equallydistance relaying configuration is assumed for ORF systems based on BER simulation of a dualhop ORF system at bit rate BR= 10 Gbps. 1O. < 1 X l. and it is shown . Finally.McMaster University . Chapter 5 presents concluding remarks and future directions.
Many lasers emit beams with a Gaussian profile.2) The factor Ao is set to normalize the transverse field to carry unit energy at the transmitting aperture. that is the laser is operating in a lowestorder mode called TEMoo mode. At this mode of propaga. 'ljJ(r. 1. the field profile is independent of ¢ . ¢) at the transmitting aperture (z described as [52] = 0) can be (2.tion. a typical direct FSO communication link was illustrated in Fig.1).1) where eto is the complex parameter related to effective beam radius (spot size) Wo and phase front radius of curvature Fo at the transmitter and is given by 2 eto = k liVe . it is readily shown that 25 . From (2. z.Chapter 2 Channel Model In the previous chapter.1 + ~ Fo (2. therefore.1 and a propagating optical beam wave was expressed by (1.1).
5) where Pt is the average transmitted power at the transmitter.Electrical Engineering Chapter 2.4) here an is the modulating coefficient taking the values 0 and 1 randomly when onoffkeying (OOK) modulation is employed. PM is obtained as [16] For Gaussian P _ 2~ t M  1. and PM is the peak: power of the optical signal.~~v~~' o 0 ~7rwJ (2.)rdrd¢ . The 26 .3) For a bit stream of Gaussian pulses the transmitted optical field envelope is expressed as (2. 2. pulses.tensity. Channel Model 1 1 IAol' JrXp(~. [12] band to achieve a good power budget. 665To' Tb (2. To is the half of pulse width at 1iein.McMaster University . n is the bit interval.1 FreeSpace Optical Channel Lineofsight FSO communication systems generally use highpower lasers that operate in eye safety Class 1M [11].
elevation of the temperature. Channel Model allowable safe laser power depends on the wavelength. According to the lEC standards [22]. The considered FSO channel impairments originate from the atmospheric attenuation.ena are dependent on weather and wavelength. Scattering phenomenon redirects the incident photons into a different direction with respect to the original axis. compared to the 850 . I 2. the 1550 nm band is extensively used in long distance commercial FSO systems.1. Consequently. Each wavelength chosen as a central wavelength has a specific attenuation coefficient. ha. Absorption comes from the interaction between the photons and at0111S or molecules that leads to the extinction of the incident photon. atmospheric turbulence. data are modulated in the optical intensity of the light. 27 .Electrical Engineering Chapter 2. can be expressed as (2.6) where CJ(ml) is the atmospheric attenuation coefficient. As mentioned. the 1550 nnl band provides higher power budget.1 Atmospheric Attenuation As the optical wave propagates through the atmosphere its power attenuates due to absorption and scattering. Both absorption and scattering phenoll1. while inexpensive components operating at the 850 nm band are utilized in short distance FSO systems. therefore it is preferred to select the central wavelength corresponding to the least attenuation coefficient. and propagation loss. nm band [15]. and radiative emission. Using this definition. The atmospheric optical power attenuation is determined by the exponential BeersLambert law [53]. in the considered optical system. therefore the allowed power and its variations through the optical channel playa key role in overall system performance.McMaster University . the atmospheric attenuation for optical field.
the fog weather condition has the highest attenuation that limits the FSO link range to a few meters [54].5 Haze 29 Fog 21272 Fortunately. The attenuation factor is also dependent on weather condition. and 1550 nm the attenuation coefficient is clear weather condition (]" = 10V{. indexed medium with different scale sizes. Turbulence is a disordered state of the atmospheric flow which is caused by temperature variations· in the atmosphere. 28 .1. As seen from this table. for the 0. Small scale inhomogeneities mostly produce diffractive effects and distort the amplitude of the wave through beam spreading and amplitude fluctuations [52]. Table 2. 2. Large scale inhomogeneities produce refractive effects that steer the beam in a slightly different direction. An atmospheric turbulent media consists of many spherical regions or eddies with randomly varying diameters and different indices of refraction.1 provides the attenuation coefficients for different weather conditiOlis at wavelength A = 1550nm.McMaster University .g. Weather Condition Attenuation (dB/km) Clear 0. at 850 nm. large scale effects mostly distort the phase of the propagating wave.Electrical Engineering Chapter 2.1: Atmospheric attenuation coefficients for different weather conditions.43 dBm/km at A = 1550 nm.2 Atmospheric Turbulence Optical beam traveling through the atmosphere experiences random phase and amplitude fluctuations (scintillation) due to atmospheric turbulence.20. Channel Model . I Table 2. therefore. The propagating optical beam experiences random spatial and temporal fluctuations in this randomly varying refractive. i.
The received field can be expressed in terms of the transmitted field Ut as 1 (2. (2. According to the central limit theorem 'l/J = 2:i 'l/Ji approaches a complex Gaussian random variable and therefore.9) Considering that turbulence does not absorb optical energy. to ensure that fading does not attenuate or amplify the average power (E[ITkI2] = 1). the turbulenceinduced fading amplitude ( TA = ITI = eX ) is a lognormal random variable with logamplitude mean flx and logamplitude variance (J~. therefore an energyconserving condition can be also applied to Gaussian beam waves. As shown in [55] . (2. the logamplitude mean !LX must be equal to the negative of the variance of the logamplitude flx = (J~.7) where T = e(i = ex +jS represents the effect of turbulenceinduced fading as a complex multiplicative term. Each slab HlOdulates " the optical field from the previous slab perturbation by some incremental values. the energy is conserved for an infinite plane wave or a spherical wave [55]. only scatters it.Electrical Engineering Chapter 2.8) As a result.McMaster University . 29 . the fading logamplitude (X) and phase (8) are normally distributed [25]. the turbulent medium is assumed to consist of a series of thin slabs. Gaussian beam waves behave like plane waves at long distances from the source [56]. Channel Model According to Rytov theory which is proposed for weak turbulence conditions.
30 . C~ does not change considerably and is assumed to be constant.12) where k = 211/ A is the wave number. where L is the thickness of the turbulent medium. Channel Model The variance of logamplitude fluctuation is a n1. The Kolmogorov spectrum is considered as a satisfactory model to calculate the logamplitude variance [52]: l (2. By estimating Gaussian beam waves with plane waves. for paths that are nearly horizontal to the earth's surface.10).easure of the strength of the amplitude fluctuations. the logamplitude variance within the range of lo « J):L « Lo is calculated as [25] (2. The principal contribution to the logamplitude fluctuations i is made by inhomogeneities whose sizes are close to Fresnel length J):L [52]. The Kolmogorov' model is applicable throughout the inertial interval which is defined as [57] 211 Lo = Xo « x « Xm = T' 5. Close to ground levels.McMaster University .92 (2. where l is the scale size.11) where lo and Lo are the internal and external turbulence scale respectively.Electrical Engineering Chapter 2. The behavior of <pn(x) outside the inertial intervals is not essential. x is the spatial wave number defined by x = 211/l.10) where C~ is the refractive index structure constant and index n indicates the refractive index dependency of turbulence. In (2.
14) by substituting (2. the phase variance for Gaussian beam waves is obtained from [52] (2. The conventional model for geometric loss which has been widely used in IM/DD FSO systems is approximated 31 .11). Channel Model In [58]. De spite logamplitude fluctuations. the phase fluctuation variance is expressed as O"s ~. Unfortunately the phase refractiveindex power spectrum is not well defined for scale sizes equal or larger than Lo.14) and performing mathematical simplifications.15) 2. a zeromean normal distribution is assumed for the phase fluctuations.3 Propagation Loss The third factor is the geometric power loss which results from the diffractive properties of optical wave propagating through optical media. In order to estimate the phase spectrum.13) in (2. Different models for phase spectrum have been investigated [52] among which the twoparameter spectrum model proposed by Kanmin is considered here 1 (2.1. 2 0 78c2k2L Xo 5/3 n .Electrical Engineering Chapter 2. Under a geometrical optics approximation (Lx2 / k < < 1). (2. certain limiting assumptions need to be imposed on the behavior of <pn(x) in the region of largescale inhomogeneities. the phase fluctuations are mainly determined by the largescale inhomogenities (including the external scale Lo) [52].13) where Xo and Xrn "are given by (2.McIVlaster University .
i.1.Electrical Engineering Chapter 2. Lr = in where Lr is the distance between the lens and fiber 32 . L t . a signal is transmitted via a transmitting lens with focal leilgth ft.2. other than power loss which is a metric for evaluating the field nmgnitude variations. by assuming Gaussian field distribution for the propagating wave a new model for geometric loss is introduced. in many applications the plane wave approximation is not sufficient to characterize the properties of the wave. photo detection is only performed once at the receiver. instead of analyzing optical power loss. To gain a satisfactory coupling efficiency. Channel Model for unbounded plane waves as [59] . the fiber is placed at the focal plane of the receiving lens.e. is dependent upon the required beam width at the receiving lens to guarantee a reliable alignment. L is the propagating distance.1.16) 1 where 'l/J3(r. Although this model is a satisfactory estimation for modeling plane waves. and e(rad) is the transmit beam divergence. Thus. In this part. the phase of the optical field is also of interest in many applications. The distance between the fiber output and lens at the transmitter. 2. (2. In addition.NIcMaster University . In this model it is assumed that the field intensity is distributed uniformly over the receiving aperture area which is the case for plane waves. since alloptical components in all relays are employed. In this work. The optical channel between transmitter and receiver is shown in Fig. the magnitude and phase variations of optical fields are considered. The receiver has a converging lens that focuses and couples the incident light into the fiber. dt and dr are the transmitting and receiving lens diameters in Fig. ¢) is the transverse optical field profile at the receiving lens plane (point ®). At the transmitter.
1. it is reasonable to consider a Gaussian model for optical wave profile propagating thorou~h optical media. the optical wave propagation thorough various optical media such as freespace. and optical fiber will be discussed first.1: The RelaytoRelay Channel Model at the receiving node and 17" is the focal length of the receiving lens.. Given the field 33 . The wave coming out of the fiber at point CD travels distance L t to reach to a thin lens at point @. optical lens. 2.Electrical Engineering Chapter 2. Channel Model TXLens TXFiber RXLens RXFiber I I I ~==?11 I I I I I I I I I ) I I I.4 Gaussianbeam wave propagation Over distances on the order of a few kilometers. it is incident on an optical lens that converges the received field and couples it onto a fiber at point ®.lIIcllIaster University . After propagating thorough a freespace path with length L. In order to analyze the transmitted wave characteristics after undergoing such propagation.. 00 I CD 8 I I Figure 2. The optical channel considered here starts from the fiber output at the transmitter point CD) to the fiber input at the receiver (point ®).
8 0 and Ao are input plane beam parameters and characterize. o?jJ (r) + 22kr or or .20). 2 1 + UXoZ 2] (2. the optical wave profile at a distance z from the source can be found by solving the paraxial wave equation [52] 1 0 .19) where 1 + iaoz (2.f!~ 'v' Ao 2z In (2. o?jJ oz = 0 (2.k ( .' 80 +i kT. z) = 1 + UXoZ Ao . Channel Model profile at z = 0.Electrical Engineering Chapter 2.Fo '. 1 ao ) r exp [ikz . the field distribution at distance z from the source must be found. If diffraction effects on the optical wave change slowly with respect to propagation distance z.. By solving (2.20) pz z 1 + iaoz = 1 .ents is large compared with the transverse extent of the beam. 34 .17) The paraxial approximation (02?jJ /OZ2 = 0) is valid when the separation distance between optical elell1.18) It is obvious that optical beam wave profile at distance z from the source has a complex Gaussian distribution with new complex parameters a z and pz (2.17) for ?jJ (r. respectively. z).McMaster University . the optical wave profile at distance z from the source is obtained as [52] ?jJ(r.
optical lens.1. A unitamplitude Gaussianbeam with the wavelength A = 1550 nm. optical beam intensity is plotted at z = 0 and z = llun. Channel Model the refractive(focusing) and diffractive c]langes in the onaxis (r = 0) amplitude of the Gaussian beam [52].To find a visual intuition about this concept. and optical fiber. 2. and the radius of curvature Fo = 20 cm propagates thorough atmosphere.g.McMaster University .Electrical Engineering Chapter 2. 2 x 2 matrices called ABeD ray matr'lces have been used.2. Z)12 (2. the beam radius TVo = 4 cm. More 35 . its beam waist (TV) broadens due to diffractive pr. As shown in this figure.21) refraction diffraction When a Gaussianbeam wave propagates thorough freespace. In order to sin'lplify the process of finding the Gaussianbeam wave profile after propagating thorough subsequent optical elements through the channel. a numerical example is illustrated here. the optical channel f{"om point CD to point ® is composed of various optical elements placed at arbitrary positions along the propagation path i.1. The transverse field amplitude at distance z is expressed in terms of 8 0 and Ao as Az = r====== Ao A6 86 + 'v" 'v" (2.5 ABeD RayMatrix In Fig. The intensity of the optical wave is defined as J(r. freespace.operties of Gaussian beam waves. 2. the beam waist at z = 1 km is approximat"ely 10 times wider than TVo. z) = I'l/J(r.22) In Figure 2. Let us consider a freespace path with length L = 1 km.
Channel Model 1.2...£0.1).Electrical Engineering Chapter 2. the matrix representations of a freespace path with length L and a thin lens with focal length f are listed.8 .~. From (2.element can be represented by a matrix whose elements construct the output beam characteristics using input beam parameters.2 o OL~J~~~~~~~~~ 5 4 3 2 1 0 rem) 1 2 3 4 5 z=lkm 4 3 2 1 rem) o 2 3 4 5 Figure 2.~.4 (j ". Each optical .2: Optical intensity of Gaussian beam waves at z = 0 and z = 1 km details of ABCD ray matrix analysis can be found in [52]... NS z=O ~ 0. In Table 2.6 c '" 2:l oS 0.McMaster University .. The field distribution at the output of an optical element can be simply found by using its ABeD ray matrix. the Gaussian beam profile in the input 36 .j "R 0 ..
McMaster University .1 FL (2.23) As expressed in (2.2: Ray matrices for various optical elements Matrix structure ( AC DB) Lineofsight freespace path with length L 1 ( 0 L1) Thin lens with focal length f (\ 1 0) 1 plane (z = 0) of an optical element is described as (2. the beam radius and phase front radius of curvature of the wave in the output plane. Channel Model Table 2. for a freespace path with length 37 . the Gaussianbeam wave profile in the output plane (z = L) is defined exactly via two parameters aL and PL.19).Electrical Engineering Chapter 2. respectively. These parameters can be chararcterized in terms of ABCD matrix elements =+2PL aoD .24) A+iaoB where W L and FL are.iC A + iaoB 2 kltVi . For example.
by using the overall ABCD matrix.20).Electrical Engineering Chapter 2.3. 2. the beam characteristics at the end of the path are calculated as 1 +iaoL (2. Consider N arbitrary optical elements between transmitting aperture (z = 0) and receiving aperture (z = L) as shown in Fig. an optical path including several arbitrary optical structures can be modeled by a single ABCD ray lTmtrix. the Gaussian beam characteristics at the end of optical path can be found.26) From (2. the received Gaussianbeam wave parameters after passing through a train of optical devices can be readily found by successively multiplying ABC D matrix representations of all optical elements. Channel Model 1 2 • • • N Figure 2.25) PL 1 + iaoL which are consistel)t with those derived at (2. Each optical element can be characterized by an ABCD ray matrix. In fact.3: Subsequent optical elements L.McIVIaster University . The overall ABCD matrix at z = L can be found by lTmltiplying all ABCD ray matrices in reverse order as (2. 38 .24). Moreover.
Typically. 2. Channel NIodel ABC D ray nlatrices invoke the paraxial approximation to simplify the description of Gaussian wave propagation through optical media. the beam waist at point is considerably smaller than the transmitting lens diameter dt . After propagating through a freespace path. the relaytorelay channel refers to the optical path which starts from the fiber output at point <D. The channel is composed of successive optical elements.1) the unitenergy transverse filed distribution at point <D is expressed as (2. Table 2. ABCD matrix transformation is power conservative and the power of the propagated Gaussian beam is equal to the launch power at the transmitter.2 The RelaytoRelay Channel Model In Figure 2. The beam is shaped and redirected to the receiving end via an optical thin lens at point @. Under the assumption of lossless optical elements.Electrical Engineering Chapter 2.27) where (Xl is the Gaussian beam parameter at point <D and @ Al is the power normal izing factor. so ABC D ray matrices have been used to 'readily characterize the transverse field distribution at each point in the channel.1. another thin lens (at point@) collects and focuses the optical beam ontothe fiber input which is placed at point ®.3 shows the corresponding ABCD matrices of the optical elements used in 39 . Consequently it is valid when the separation distance between optical elements is large compared with the transverse extent of the beam. Using (2.McMaster University . therefore ABCD matrix transformation corresponding to the thin lens at the transmitter (point @) and free space conserve the beam energy.
McMaster University .1 structure Matrix Lineofsight freespace path with length L t Thin lens with focal length ! (10 L1t) (\ it Lineofsight freespace path with length L Thin lens with focal length 1 L1) (0 (_1~ ~) . 1 L1r) Lineofsight freespace path with length Lr (0 ments placed between point CD and point ® in a reverse order.19). Channel Model the channel from point CD to point @. 2 (2. By multiplying ray matrices of the optical ele Table 2..29) 40 . the transverse field distribution at point ® is obtained as 0/' ( T.28) therefore from (2.3: Ray matrices of optical elements in Figure 2.eikL3 exp (1 Al 3 kCt3T 2) P3 .It 0) 1 ir I. the ABeD ray matrix of the beam wave at point ® is obtained as (2.Electrical Engineering Chapter 2. Z 'f/3 = L) = .
is due to the finite size of the receiving lens which is considerably smaller thaIl the received beam waist. i. (2. The other factor called coupling efficiency. .iC3 2 .Electrical Engineering Chapter 2. 'Tlc.1 =+'/.33) TYPIcally.e.A3 + iCiOB3 kVVi F3 TY3 and F3 are the beam waist and phase front radius of curvature of the optical beam wave at point @ respectively. results from the misalignment between the 41 .32) + iCiOB3 CioD3 .lVIcMaster University .30) (2. in order to have a reliable alignment between the source and destination in FSO links.31) (2. the optical beam waist at the receiving lens (point @) must be sufficiently larger than the diameter of the receiving lens dr (TY3 ~ dr ). Channel Model where L3 P3 Ci3 L t +L A3 (2. By assuming a lossless optical lens and ignoring atmospheric absorption loss and turbulence effects (these effects have been taken into account separately). One. two factors are considered as the main contributors in power loss. This process clearly imposes a considerable power loss to the signal in lineofsight FSO systems and is conventionally called geometric loss and can be calculated via (2. 'l/J3(r) retains a Gaussian profile with unitenergy. The receiving lens is only able to collect the portion of the received field which is incident on the lens plane and the remaining parts of the propagating wave are discarded.16). In the new model proposed here for propgation loss. 'rJz.. so that a major part of the signal power is failed to be collected by the lens and dissipates in space.
the fiber is placed at the focal plane of the receiving lens. i.34) In order to find 'r/l and r]c. the transverse field distribution at point ® is obtained as 'l/J4(r) = ik Al ikr2 . 'l/J4(r).McMaster University . on the order of a few centimeters. 2. 2 r (2. hence to gain a satisfactory coupling efficiency. Channel Model focused beam wave distribution at point ® and the field distribution inside the fiber. in typical FSO systems. fitst the transverse field distribution at point ® must be determinded.ka 3s )Jo( f s)sds.35) and performing mathematical simplifications. Hence. to guarantee the smallest spot size (beall1~ waist) at point ®.stituting (2.35) where rand s are the radial coordinates at the fiber input plane (point ®) and receiving lens plane (point ®) respectively.29) in (2.[60] (2. Consequently the new model for propagation power loss is defined as (2.exp( ikL4) eXP(:f ) r P3 2 r 1{ d 0 1 2 kr exp( . Lr = ir. the Fresnel diffraction integral is used [52].2. to find the transverse field profile in the back of focal plane of the lens. By sub. by reasonably 42 . As shown in Fig.36) where L4 = L3 + Lr = L3 + ir.Electrical Engineering Chapter 2. The effective cross section area of the fiber is very small (on the order of micron).e. optical beam waist is on the order of a few meters which is considerably larger than the receiving lens diameter.f . Since there is no ABCD matrix which exactly models a finite thin lens.
McMaster University  Electrical Engineering Chapter 2. Channel Model approximating the exponential term in (2.36), exp( ~ka3s2) ~ 1 for small beam profile at point ® is approximated as
kd ) id A ikT2 J 1 ( 2/ r 'ljJ4(r) ~ _T_l exp (ikL 4)exP(  j ) r, 2 P3' 2 T r
8'S,
the
(2.37)
where J 1 (.) is the first order Bessel function of the fiTst kind. For dT
7
00, 'l/J4 (r)
tends to a complete Gaussian beam profile whose characteristics can be simply found via ABC D ray matrix. The thin lens transformation is power conservative that is the average power of 'l/J4(r) is equal to the average collected power by the receiving lens. The power loss imposed to the signal is due to the limited collecting area of the receiving lens and is defined as
(2.38)
where () =
II· rdrd¢ is the spatial average operator.
A
From (2.33), the power loss
7]1
is obtained as (2.39)
2.2.1
Signal projection onto SingleMode Fiber (SMF)
In a typical application, the received optical beam must first be coupled into a singlemode fiber (SMF) to be processed by optical elements [51]. Projection of the optical field onto SNIF imposes additional loss to the system. Furthermore, atmospheric turbulence degrades the spatial coherence of the propagating beam. and limits the coupling efficiency [61]. However, the degrading effects of atmospheric turbulence on fiber coupling is not taken into account and the coupling efficiency arises just
43
McMaster University  Electrical Engineering Chapter 2. Channel Model due to misalignment between the incident beam profile and SMF characteristic field distribution. The characteristic field distribution inside a SMF is independent of phase ¢ and is given by [16]
(2.40)
where
Teo
is the radius of the fiber core, C is a constant which can be determined
from the average power carried by the guided mode, and 1'1 and 1'2 are described as
(2.41)
here n1 and n2 are the core and cladding refractive indices respectively (n1 > n2), and
f3 is pmpagation constant inside the fiber which is also dependent on the mode
of propagation. In singlemode fibers, n1 and n2 are chosen such that there is only one mode of propagation called fundamental mode of propagation or LP 01 inside the fiber. In order to find the normalized field distribution corresponding to LP 01 mode of propagation inside the SMF, the constant parameter C is determined such that the average power carried by this mode is unity, i.e.,
44
McMaster University  Electrical Engineering Chapter 2. Channel Model
1
J J 'l/JF(r, ¢)'l/J'F(r, ¢)rdrd¢
00
27r J 'l/JF(r)'l/J'F(r)rdr
o
21fC2llJg h Tlrdr + (~i~2:::J
1
=?
.
[
1
00
KJh,r lrdT]
(2.42)
c=;================= 27r Teo J,2(ry r)rdr + ( ) 2 J K2(ry r)rdr J o 11 JO('YlTeo) ]{O('Y2Teo) 0 12
0
~
1
where AF is the fiber cross section area. The for.ward propagating optical field inside the fiber is described as
\]! F(r,
z, t) = uFUt(t)
'v'
'l/JF(r)
'v'
Field envelope Transverse field profile
~
Carner
e~i(wt~{3z)
(2.43)
In this equation,
UF
is the mode weight factor which is determined by the projection
of received optical field onto the singlemode fiber
00
21f J'l/J4(r)'l/J'F(r)rdr
(2.44)
o
45
the complexvalued geometric field loss in a fibertofiber optical channel is introduced as (2.46) The coupling efficiency is defined as the ratio of the average optical power coupled into the fiber to the average collected power in the receiving lens plane [61] =1 ~ ) TJc = (11jJ4(r)1 2 = (luF1jJF(r) 12) (IUFI2) (I1jJF(r) 12) (11jJ4(r) 12) IUFI2 = (11jJ4(r)12) (2. the n:lOde weight is obtained as i21rk Al UF = j exp(ikL4) r P3 J1:[ 00 d o 0 2 iks 1 krs 2 eXP(j )exp(ka 3 r )Jo(j )1jJ'}.(s)rsdrds 2 r' 2 r (2.48) By comparing the field envelope of the transmitted Gaussian wave. the mode weight is approximated as (2. uFUt(t) in equation (2.45) Finally by approximating 1jJ4(r) with equation (2.49) 46 . the new propagation power loss is described as (2.Electrical Engineering Chapter 2.43).44). Therefore. the field envelope of the launch optical beam wave is multiplied by a complex constant UF while propagating through the fibertofiber optical channel.36) in (2. Channel Model By substituting 1jJ4(r) from (2..34). Ut(t) in (1.37).McMaster University .1). with the field envelope of the propagating wave inside the fiber.39) and (2.47) in (2.47) by substituting (2.
76 cm 2. 47 . 4 J.2 cm 19.Lm 00 1550 J. a converging lens is used as the transmitter side. parameter value Core radius r co Core refractive index nl Cladding refractive index n2 5 J.4: The considered FSO system characteristics. hence.Electrical Engineering Chapter 2.Lm 1. in order to adjust the desirable beam radius at the receiving lens plane. Due to the diffractive properties of Gaussian beam wave propagation in freespace.7 mrad. Channel Model Table 2. the receiving lens focal length that the best fiber coupling efficiency (7Je ~ ~ 3 m) is achieved is determined so ir 82%) is obtained.4 shows typical parameters of a FSO communication system which is considered through the thesis for simulation pmposes. The distance L t is adjusted so that the desired beamwidth (TiV3 at the receiving lens plane.McMaster University .5: SingleMode Fiber Specifications.50 The comparison between the conventional and new geometric loss is illustrated via a numerical example. Parameter Transmitting lens di~meter dt Receiving lens diameter dr Transmitting lens focal length it Receiving lens focal length ir Distance between the fiber output plane and transmitting lens L t Beam divergence angle () Beam waist at laser output liVI Phase front radius of cmvatme at laser output PI Wavelength A Value 5cm 24 cm 20 cm 58. the beamwidth broadens after propagating over long distances.Lm Table 2. Also.5047 1. Table 2.
16) is compared with the new model given by (2. 2.. Table 2. the conventional model for geometric power loss in (2...Electrical Engineering Chapter 2. Channel Ivlodel 45..4.~. For short distances where the beamwidth is a few times bigger than the receiving lens diameter.46) is used. To calculate the values of UF for different lengths (L).5 shows the singlemode fiber specifications which are commonly used in fiber optic systems.McMaster University . 40 ~ H g +H p 35 15 1000 2000 3000 4000 FSO link length(m) 5000 6000 Figure 2. the comparison is made between these two models.48). the approximated formula given in (2. In Figure 2.4.4: Comparison between the conventional and new geometric power loss. point @. approximation of Gaussian 48 . In Fig.
As shown in Fig. For' long distances where the Gaussian beam profile can be approximated bY' a uniformly povirerdistributed plane wave. the area over which the Gaussian beam profile is approximated by a plane wave is much sn1. 49 .lVIcMaster University . the new model for prop agation loss provides more accurate approximation than the conventional model. however. Channelldodel Esthnation area of Hp Estimation area of Hg Figure 2.6 pro vides more detailed specifications of a Gaussian wave propagating through an FSO link at two different distances from the source.aller for the new model than the conventional model.5.0re accurate. 2. both the new and conventional models provide more reasonable approximations.5: Estimation area for the conventional (Hg) and new propagation (Hp) loss. 2.Electrical Engineering Chapter 2. In this case. Ll = 800 m and L2 = 4000 m. Table 2. the new model is 1~1. beams by plane waves is not reliable and the beam power distribution is not uniforlTl' over the estimation area shown in Fig.5.
Let the receiving lens plane where N 3 (t.44cm 1.31 + iO. s) can be written as (2.Electrical Engineering Chapter 2.oment of the noise spatial distribution at the lens plane (point @ in Fig.51) 50 . the first n1. S) denote the Gaussian noise distribution at is the s is the radial vector in the lens plane and t time scale.2 2.8% L2 = 4000 m 1.44 cm 5.3 (6.07 X 10. 2. As expressed in (1.McMaster University .54) x 10.45 m 82.50) therefore by projection of N 3 (t.3 Noise Projection onto SingleMode Fiber Background noise is the dominant source of noise in FSO communication systems and is statistically modeled as an additive white Gaussian noise in time and space with zero mean and variance 0.lm 82. Since background noise distributions in time and space are independent.4). Parameter Conventional geometric power loss Hg New propagation power loss Hp Propagation field loss hp Beamwidth at point @ TiV2 Beamwidth at point @ H!3 Coupling efficiency 7Jc Ll = 800 m 3.6: Propagating Beam Wave Specifications for Different Link Length.23 X 104 1.99 X 104 (1.3 4. therefore it is projected onto the fiber through the receiving lens.1) is defined as (2. N 3 (t. s) onto fiber. Based on the assumed statistical model for background light. temporal statistics of the noise remains unchanged.82 + i1.30) x 10.5% I 2. = N o/2 [26]. Channel Model Table 2.82 X 10.2 2. background noise is added to the signal at the receiving lens.
Electrical Engineering Chapter 2.r (Jf112  Jf212)) (2. From definition.McMaster University . Channel Model where E [. f l)N. r') is calculated as (2.55) 51 .54) and spatial correlation is termed as RN4 (tl' t 2. The first moment of the noise picture N 4 (t. the picture of the noise in the back of focal plane of the lens is obtained as (2.52) which indicates spatially white noise distribution over the receiving lens plane.(t2. By using the Fresnel diffraction integral. fl' f 2) = E 4(t l . the correlation function can be written as (2. and dot operator (. r2)J = (2:fr) 2 [N exp (~.) denotes the vectors inner product.53) where sand f are the radial vectors in the lens plane and fiber input plane (point ®) respectively.J is the expectation operator.
56) the mean and variance of n F are calculated as . it must be shown that the projected ~loise onto the fiber has also the same statistics.McMaster University . the thin lens conserves statistics of the incident field. exp (ik(I>1 2 ..z (2. (2. From (2.r1$2.57) (2.12d. the noise weight factor inside the fiber is defined as nF = 1 J Afiber N 4 (t.s~N.(t2.r2))d.44).1>12))IJ exp r1 1'2 ~ 27[" iT 2fT (ik> · (> »)d> 2 81 1'1r2 81 fT Alens Therefore. the background noise after being focused in the back of focal plane of the lens has still zero mean and temporally and spatially white distribution.Electrical Engineering Chapter 2. Channel Model x 11 11 ![N3(t1. Now. r)'ljJ}(r')d:.58) 52 .s2nexp(~(Sl. in other words.22 Alens Alens ~Do(t2h)o(S2S1) N k)2 ? O ( .
Therefore. to completely describe the amplitude and phase variations of the propagating beam wave. in the considered FSO system. The proposed model provides more reliable approximation of propagation loss rather than the conventional model. This channel model is utilized to analyze performance of various FSO relaying techniques presented in the following chapters. the additive noise is modeled as a white Gaussian noise with zero mean and variance (J~ = N o/2. In this thesis. Atmospheric attenuation.McMas'ter University . therefore a complex model for atmospheric turbulence and propagation loss has been considereq. the coupling loss induced by the field projection onto a singlemode fiber has been considered in the new model. atmospheric turbulence inducedfading. A weak atmospheric turbulence condition is assumed whose logamplitude and phase are normally distributed. Other than geometric spread of Gaussian beams.4 Conclusion In this chapter the optical channel has been modeled as an AWGN channel. 2.57) and (2. 53 . Channel Model Equations (2.58) indicate that the statistics of the noise inside the singlemode fiber is the same as statistics of the background light incident on the receiving lens plane. Background illumination is considered as the major source of noise and is modeled as a zeromean white Gaussian noise. and atmospheric propagation loss are the channel impairments considered in the channel model.Electrical Engineering Chapter 2. Also a new method for calculating propagation loss is developed and numerically compared with the conventional model. the field envelope of the optical signal and noise are analyzed instead of optical power.
In an alloptical multihop FSO communication system at each relay data are processed ill optical domain. In this chapter.Chapter 3 Optical AmplifyandForward Relaying Technique In this chapter. Then the 54 . and optical amplifier.1.g 10 Gbps TereScope TS10GE. e. encourages replacement of electrical relaying processors by optical elements in all relays. longer communicating distances can be achieved in FSO systems while taking advantage of highrate optical transmissions. Highbandwidth. An optical amplifyandforward (OAF) relaying technique is developed and illustrated in Fig. shortdistance freespace optical tranceivers. the background light is added to the received data field. each relay is composed of alloptical elements such as an optical lens. At each relay. it is shown that by using alloptical relaying techniques. "FSO transmission is possible over longer distances. As shown in this figure. multihop FSO communications using alloptical components is studied. The distance dependence of atmospheric turbulence and path loss lirriits the total communicating distance in FSO systems. 3. Using relaying techniques. optical fiber.
3. a plane normal to the lens axis placed at distance itoea! behind the lens. Optical AmplifyandForward Relaying Technique UO f Uk+l Uk+l U M +1 I'T~!~ Transmitter k=O 1 1 1 ::E~ .. The structure of a typical OAF relay is simply shown in Fig.1 OAF Relay Structure In OAF relaying.. This field distribution is projected onto a single mode fiber (SMF).. The SNIF is connected to the optical 55 . ~I>{RX! Relay r.. there is at least one optical amplifier which amplifies the received optical field and retransmits it to the nex"t relay. As mentioned.1: An AllOpticalNIultihop FSO Communication System noisy data field is amplified by an optical amplifier and forwarded to the next relay or receiver.NIcMaster University . 3. there is a converging lens at the beginning of each relay that collects and focuses the incident light onto the back focal plane of the lens.Electrical Engineering Chapter 3. The complex amplitude distribution of the field in the focal plane of the lens is the Fraunhofer diffraction pattern of the field incident on the lens. The optical link between each two consecutive relays is modeled as a fibertofiber channel which has been cOl11_pletely characterized in Chapter 2..I 1 1 1 Receiver k=M+1 Optical AmplifyandForward Relay Figure 3..2.
regardless of the noise of that relay. Other than con1. f is frequency. and nsp is the amplifier spontaneous emission parameter.lly modeled as (3. optical multihop AF systems developed so far [44. in what follows every field envelope U(t) is referred simply by U. 45] deployed an adjustable gain which compensates the fading effects of the preceding hop. 27. The spectral density of ASE noise is given by [16] (3.2: Structme of OAF Relays amplifier that is mathem.Electrical Engineering Chapter 3. As noted before.plexity in implementing ad justable gain unit at each relay.McMaster University .(t) and Ul'(t) are the received and LramlIllitted signals at the respectively. The ASE noise is modeled as an additive zeromean white Gaussian noise. G k is the kth kth relay amplifier gain and V~SE(t) is the amplified spontaneous emission (AS E) noise of the kthamplifier. Optical AmplifyandForward Relaying Technique Fiber Receiving lens Optical Fiber Transmitting lens Figme 3.atica.2) where h is Planck's constant. the relay output power does not satisfy eye safety 56 .1) where U'. For the sake ?f simplicity.
1 .1. G k . E [IU?1 2 ] (Uf is the transmitted field envelope at the source node). can be expressed in terms of the trans 1th relay. therefore. The proposed model for the gain of [ kth relay. (3.\ as Utk = ~ (hkUtk. Uf. Also it is chosen so that the average power of the transmitted signal at the kth relay p tk = = E [I Uf n is constant and equal to the average launch power at the source Pt I.McMaster University .4) As mentioned in Section 2. Utk .Electrical Engineering Chapter 3. is independent of random fading fluctuations and only 1[ compensates the propagation loss and atmospheric attenuation induced by the last hop (e h hop). atmospheric turbulence. the lognormal fading is normalized so that the mean intensity of the propagating wave is conserved (E[ITkI2] = 1). the 57 . Consequently. Optical AmplifyandForward Relaying Technique limitations (IEC) on the maximum permissible average power.2. and k  1th relay to the kth Ut is the background noise collected at the receiving lens of the kth relay. + uf:) +U~SE 1 (3. In this thesis a more practical model is provided for amplifier gain at each relay. . from (3. The data signal.e.4) V u}' where hk is the complex gain of the channel connecting the relay.3) The transmitted signal at the mitted signal at the k kth relay. background light and amplifier iloise are all independent random processes.
it is guaranteed that the average output power of each relay satisfies the eye safety constraints.k and hp.3) and (3.k are the atmospheric attenuation and complex propagation loss of the kth link respectively.McMaster University .10) 58 . Plugging (3. is the average collected background light power at the = receiving lens plane and assume~ to be identical for all relays. P~SE kth E [IU~SEI2] is the average ASE noise power of the amplifier and from (3.8) into (3.Electrical Engineering Chapter 3. gk is defined as the channel loss or path loss and includes both the effects of atmospheric attenuation and propagation loss. PA ·is negligible with respect to Pt and Pb . Here. Using ( 3. Optical AmplifyandForward Relaying Technique average of the squared magnitude of the complex channel gain is (3.7) where H= E [lVbI 2] = CT. so G k can be approximated as (3. at optical frequencies.5) can be sinlplified to (3. Equation (3.8) where flJ is the bandwidth of the optical amplifier. This way.9) where PA = hJnspflJ. Typically.6) ~! ! where ha.6).7) and rearranging for G k gives (3.2) is expressed as: (3.
In the next section.2 Optimal Relaying Configuration In practice. Therefore. where OAF relays are typically shown as amplifiers. It is important to arrange relays such that the best system performance is achieved at the receiver. L k . SNRo Pt! Pb is the average transmit signaltonoise ratio (SNR) at the output = of each relay.6). Figure 3.. In the considered multihop system. g. Let Up denotes the transmitted signal at 59 . The hop distance. relays are placed at fixed stations between the source and destination nodes.that is . In this section. is dependent on both SNRo and gk. In the presence of high background noise.l.Electrical Engineering Chapter 3. G k ~ g. G k is set to a smaller value so as to keep the average output power of each relay within the eyesafe region. In the case of low background noise. it is clear that gk is a function of the kth hop distance (Lk). the amplifier totally compensates the effects of the channel loss. relays are consecutively placed between the source (k = 0) and destination (k = Iv! + 1) nodes.McMaster University . an optimal relaying configuration is demonstrated for a given eyesafe SNR o· 3. by optimizing hop distances.3 shows an optical Amplifyand Forward multihop system with M relays. the performance of the system is analyzed in terms of the average optical signaltonoise ratio (SNR) at the receiver . From (3. Optical AmplifyandForward Relaying Technique here. which is the length ofthe link connecting the (k1)th node to the kth node varies for different relays. Pb ~ 0. The amount of G k offset from its noiseless value.l . the channel power loss imposed on the data field is not compensated completely via amplification process and leads to additional degradation effects on the system performance.
hPb ) (3.U!JS~ + ~ JI.. Optical AmplifyandForward Relaying Technique Figure 3. t. y'G.11) .12). From.McIVlaster University .3: Optical AmplifyandForeward rvlultihop FSO Systems t. the average received power at the receiver (j = 111[ + 1) is pM+l r + (3.he source. ASE noise and fading are all independent.Assuming that the signal. the average optical SNR at the receiver 60 . t.h'.hk+JUt) (h.he received field at..a signal power and noise dming the channel.12) In order to analyze variations of the dat.2 . the average optical SNR is defined as the ratio of the average data signal power to the average total noise power.Electrical Engineering Chapter 3. background noise. is expressed as u! + (fi JG~hk) hP~ + (ut + ~g y'G. (3. 111 + 1. j = 1.he lh relay.
16) The path loss gk depends on the hop distance L k .14) By substituting Gk from (3.15) where SNRk is the average receive SNR at the receiver of a direct FSO link ( where there is no relay between transmitter and receiver) with length L k : (3. the average optical SNR is expressed as (3. SNRII. and performing some simplifications.10) into (3.t L (3.J+l must be optimized with respect to Lk'S.Electrical Engineering Chapt.13) can be approximated as (3. Consider the optimization problem max Lk SNRM+l Lk = LT M+l s. Optical AmplifyandForward Relaying Technique is obtained as Typically amplifier spontaneous emission noise PASE is negligible with respect to data power Pt and background noise power Pb .lVlcMaster University .14). therefore (3.er 3.17) k=l 61 . Now.
Let h(x) = ftx). (3.17) is simplified to M+l S.4: An OAF lVlultihop FSO System with M= 1 The above optimization problem can be converted to a simpler format. 3.t L Lk = LT (3. where f (x) > 0 and f (x) and h( x) are continuous functions whose first (J' (x) and h'(x)) and second (JI/(x) and hl/(x)) order derivatives are defined at Xo.fl/(xo) > 0 Xo (3.4.19) therefore.!'(xo) = 0 < 0 =?.McMaster University .18)  fl/(x) J2(x) (3. In order to show that h(x) is locally maximized at xo. 0 and hl/ (xo) < 0 which is equivalent to f'(x) J2(x) h'(x) hl/(x) Xo = 0 =?. Optical AmplifyandForward Relaying Technique 11 ! TX RX LI L2 LT k=O . 62 .Electrical Engineering Chapter 3. • k=2 Figure 3. it is enough to prove h'(xo) ..20) k=l Consider an FSO system with a single OAF relay placed between the source and destination nodes as shown in Fig..
and some algebra.Electrical Engineering Chapter 3.McMaster University .ization problem in (3. the path loss is approximated (3.6) and (2.k 2 = 1 e uLkHp. the path loss gk is dependent on the hop distance Lk and is defined as gk = Iha.kl = IUFI2 is propagation power loss.n [n U+ s.k and hp.24) where as K.k have been replaced by their definitions given in (2. By plugging (3.21 ) By changing the variables L1 = x and L2 = !:TX.46). From (2.23). From·(3.48) 2 from Chapter 2 and Hp.t gklSNIlo 1 )] L1 + L2 = LT (3.22) where gl and g2 are the path loss of the first and second links. is a constant. Optical AmplifyandForward Relaying Technique The optim.k = Ihp.k (3.20) for this system is defined as "'l.6). propagation power loss can be approximated as (3.25) 63 . problem is simplified to th~ optimization min· x (3.k hp.24) into (3.23) where ha.
5.22) can be described in terms of L 1 and min x (3.. 4 ~ 3.5 en u '.....26) It is shown in Appendix A that in typical FSO systems f{(x)lx_!T 2 2 0 f{' (x) L_LT > 0) (3.McMaster University .Electrical Engineering Chapter 3..0 0 (1) Cil §< 3 :> 'Q:) u ~ (1) ES 2.5 2 500 1000 2000 2500 3000 Figure 3. Using this approximation) (3.27) 64 . Optical AmplifyandForward Relaying Technique 4.5: The received optical SNR of an OAF FSO system with M SNRo = 37 dB.
Therefore. K" and SNRo and is obtained in Appendix A.1 = 1.. . the average optical SNR received at the receiving lens of an OAF multihop system with )1. These conditions are sufficient to prove that II(x) has a local minimum at L = LT/2 over the desired range of LT. In this plot. the total communicating distance is LT = 3 km which is inside the acceptable region of LT for SNRo = 37 dB. RX L. Uz 1)1 1 i U b2 T:.Electrical Engineering Chapter 3...6 )..5 illustrates the optical SNR variations by the position of the OAF relay.. where ff (x) and ff' (x) are the first and second order derivatives of function II (x).. Optical AmplifyandForward Relaying Technique 1.. is maximized when the OAF relay is placed at the middle of the link connecting the source and destination.I 1 1 TX k=O 1 ~ G.. : 1 L" k=3 1~ • ~ 4 Subsystem Figure 3. This range depends on 0:. In addition... Obviously. 3.McMaster University . that indicates the function II (x) is convex for all 0 < x < LT and L = LT/2 is a global optimum point. for all °< x < L T .. Similar calculations can be considered for an OAF system with 65 ]v! = 2 (Fig.. it is also proved in Appendix A that ff'(x) > 0...3.. SNR is maximized when L1 = L2 = LT/2 = 1. .6: An OAF Multihop FSO System with M= 2 and SNRo = 37 dB.5 km which means the best performance of the system iR achieved for equallyspaced relaying configuration.. as shown in Fig.1 1 I 1 1 1 I r. 3.4. Fig..
By changing the variables Ll = optimization problem is simplified to X.SNRo)leCXXx2)(1 + (n. and the second relay and the destination node respectively.y) = 8h (X'Y») ( 8x 8h(x.Y h(x. y) is laborious.7 which is plotted at SNRo maximum SNR is achieved at Ll = L2 = L3 = 1 km.y) = 37dB indicates that the H(h) = ~ 8 2 h(x. (3.SNRo)leaYy2) (3. However. the mathematical induction proof is utilized 66 . L 2. L 1 .ation problem is defined as (3.30) X=T'Y=T Proof of nonnegativity of the Hessian matrix of h(x.29) (1 + (n. L2 = y and L3 = LT  X  y.31) For higher number of relays (Ai> 1).SNRo)lecx(LTXY)(LT . the mm X. LT = o. the first relay and the second relay. and L3 are the hop distances between the source and the first relay.o. the numerical simulation in Fig. 3.x .Electrical Engineering Chapter 3. y) is zero at x = y = L T /3: V'h(x. Optical AmplifyandForward Relaying Technique The optimi7.II/LcMaster University . (3.y)2)] It can be easily shown that the Gradient vector of h(x.28) Here.y) 8y LT .y) x [(1 + (n. ( 8 2 h(x.y) 8y8x iJx2 >.
lVIcMaster University . As the base case. p relays are equally spaced between the source and destination nodes with total distance of x. assume that for lVI = p. In Fig. 3.5 4 I ~ i 3. Now.Electrical Engineering Chapter 3. it has been proved that for the OAF system with M = 1 (dualhop system). From (3.8.5 5 Lr and L2 do not have value for this region 4.5 3 2. the equallyspaced relaying optimizes the problem (induction hypothesis). the modified optimization problem for this system is expressed as 67 .5 500 1000 2000 2500 3000 Figure 3. equallyspaced relaying configuration provides the best performance at the receiver.7: Optical SNR of an OAF FSO system with 1111 = 2 to find the optimal relaying configuration. Based on induction proof. Optical AmplifyandForward Relaying Technique 5.20). it should be proved that equidistance relaying configuration also maximizes the total average received SNR at the receiver when JY1 = p + 1.
f.9. .25) and (3..32).e.8: An OAF FSO system with lYI = p TX 10=0 RX k=p+2 X ~.Lp+l) is minimized for equal hop lengths. From (3.McMaster University .. 3. 3.8 and provides a longer link with p + 1 relays and total communicating distance of LT = x + Lp 12..in (x) is obtained as (3. . Figure 3. one more relay is added to the link plotted in Fig. = L p+1 = x/(p + 1). i.9: An OAF FSO system with lvI = p +1 (3. L 2. Optical AmplifyandForward Relaying Technique k=O x k= p+l ~. Ll = L2 the n'linimized = .32) Based on the induction hypothesis.Electrical Engineering Chapter 3. the function fp(Ll. Figure 3.33) In Fig..+Lp+2~ LT .. The optimization problem for the extended link can be expressed as 68 .
the total communicating distance LT is divided into' x and LT . in order to optimize the SNRJ\1+ 1 at = p the receiver of an OAF system with M + 1. .:in(x) from (3.. By substituting j. Lp+l) x (1 As shown in Fig.. Furthermore.'" .25).34) The function jp+1(L 1.'~2SNRn') + g...Lp+1) as follows [g Lp+2' therefore Lp+2 = (1+ gklSNRnl)] (1 +g.Electrical Engineering Chapter 3.' .~2SNR(1) (3. ..Lp+l) minimizes.36) In Appendix A.McMaster University . . 3. L 2.33) and utilizing (3.35) jp(L1' L 2.Lp+2) can be rewritten in terms of jp(L1' L 2. .x. Optical AmplifyandForward Relaying Technique (3.9. the optimization problem defined in (3.34) is simplified to x (3. the received SNR at the previous re lay should maximize or equivalently jp(L1' L 2. it is shown that for the equallyspaced relaying configuration L1 = 69 . .
38) provides the maximum average SNRlIH1 at the receiver over the desired region of LT.e.+1 (x) 0 > O..J2) ' the function I p+1 (x) satisfies L 1.McMaster University . The system under consideration operates at A = 1550 nm in clear atmospheric conditions with attenuation coefficient of (J = 0. System performance is studied at two bitrates (BR) 1. Optical AmplifyandForward Relaying Technique L2 = ... i.Electrical Engineering Chapter 3.43 dB/km and weak turbulence with 70 . the equallyspaced relaying configuration provides th~ best performance at the receiver and all the previous relays for !vI = p +1 and consequently the induction proof is complete. Therefore for any arbitrary number of relays !vI.+1 (x) 1. the performance of different OAF multihop FSO communication systerns with equallyspaced relaying configurations is analyzed.25 Gbps and 10 Gbps. MonteCarlo simulations are used to calculate the BitError Rate (BER)' of various systems. This configuration is utilized in the next section to simulate various OAF systems.3 Numerical Results and System Performance In this section.. = LM+1 = . = Lp+1 = Lp+2 = (P~l) = (p . equallyspaced relaying configuration.. 3. (3. The degrading effect of atmospheric turbulence on BER is analyzed at both bitrates and compared with turbulencefree system performance. LT L1 = .37) In other words. !vI+1' (3.
64 samples per bit interval are provided. The background. 10. In this section. number of relays and hop distance. noise power spectral density is assumed to be No = 2 arnplifier spontaneous emission parameter is \ X 10. 62. placed between the somce and destination is analyzed by plotting the BER versus the transmit signaltonoise ratio SNRo = pt/ Pb' Fig. In the presence of atmospheric tmbulence.McMaster University . gk. The tmbulence fading effects are not considered in these plots. By comparing two figmes. 40] and the nsp = 5.3. The other characteristics of the 1 system are as given in Chapter 2 via Tables 2. The overall performance of the system for different number of relays.1 Fixed Total Communicating Distance Consider an FSO system where the somce and destination nodes are placed at a total communicating distance of LT = 3 km from each other. Optical AmplifyandForward Relaying Technique refractive index structme constant of C~ = 1 x 1O. by placing a different number of relays (different JlVJ) between the somce and destination nodes.15 .14 m. the overall performance ?fthe system only depends on the average transmit SNRo and the relaying configurations.5.10 and 3.2 / 3 [52].iV/Hz [1. From (3.. of the systems considered in these figmes.25 Gbps andBR= 10 Gbps respectively. the significant role of relaying technique in improving the performance of the system is justified.Electrical Engineering Chapter 3.e. 3. 10 7 bits are transmitted per channel state. 71 .3. Table 3. the BER is averaged over NT = 1000 different fading conditions to reasonably simulate the slowfading tmbulence channel. In order to simulate the FSO system in the slowlyvaraying optical channel. i.15) and (3.4 and 2. 3.1 summarizes the configmation.11 correspond to the systems working at bit rates BR= 1. At both bit rates. M.16).
By assuming Pt = 640 m\i\T as the maximunl available transmit power.6 0.11 = 0 (direct transmission) and J1.25 Gps and BR= 10 Gbps is about 39 dB and 30 dB respectively. N1arker hop Distance . the BER of the systems operating at BR= 10 Gbps with 11.5 0 1 2 3 4 5 B e +~ it is justified that for a given SNRo the system performance for a specific configuration is nearly the same at both bit rates.5 1 0. the noise power collected at each relay (or receiver) in 10 Gbps system is more than the noise power in 1. in order to reduce the biterror rate to BER= 10. The best performance of a FSO direct transmission inside the accessible power region is BER= 1.26 x 10. By employing one relay in the midway. Optical AmplifyandForward Relaying Technique Table 3.25 Gbps system.1 e~ 3 1. almost 9dB more power must be transmitted by the system operating at 10 Gbps.1: Different system configurations for LT = 3 km. SONAbeam™ 12501\11 transceiver sends 640mW power via focur transmitters each sending 160 m\i\T. 3.2 . In Fig.5 or 72 . The maximum average power transmitted by the stateoftheart FSO transceivers is on the order of hundreds of miliwatts.75 0. Number of Relays Lk(kTn) ]1.11. Therefore.Electrical Engineering Chapter 3.[ = 1 relay have been plotted in the region SNRo :> 30 dB vv'hich is called inaccessible power reg·ion.McMaster University . the maximum achievable SNRo at BR= 1. In order to gain relatively similar performance at both bit rates. Since the amount of background power is relative to the optical bandwidth or equivalently optical bit rate. But. the systenl achieves BER= 7 x 104 which is an order of magnitude improvement.
......... for a given SNRo....10 alld 3..::: .11.McMaster University .:... I ::::. Thus this gain mainly comes from the reduction in path loss achieved by shortening the hop distances.25 Gbps..·~O:~~gi()l"lT' "1'" .. BER at the destination llode decreases .. . . ..Electrical Engineering Chapter 3. :::::·:~?\\r:~.lts 8110\... At both bit rates....\l11 in Fig...~N~(j:s:~1~:~:~::~~ ........ Optical AmplifyandForward Relaying Technique 102 . ~..5 . Here the effects of atmospheric fading are not considered.. 15 30 35 Figure 3. no fading effect is considered........ 3. (The plot descriptions are given in Table 3... ...49 dB improvement at BER = 10...... .(M = 2 or more)........ inserting Olle relay at the Iniddle of a 3 km link gains 2.. by shortening the hop distances via inserting more relays between communicating nodes (increasing M).1) less....10: BER versus SNRo for a 3 km link and different number of relays M at BR= 1...... at least two relays must be placed between the transmitter and receiver... 73 ......
. no fading effect is considered.... _ ..49 dB... .. the system performance gains only 0...McMaster University ... Therefore...... ... Qualitatively.. .... ....11: BER versus SNRo for a 3 kmo link and different number of relays 1\11 at BR= 10 Gbps. As the number of relays increases. Optical AmplifyandForward Relaying Technique .. J...... .. by increasing the number of relays from 1\11 = 4 to JI/! = 5.. the gain achieved by inserting just one relay at the middle of a 3km link.. As mentioned before.1) .. the amplifier boosts up the signal with a relatively smaller gain. ·+~·Power ... .......74 dB improvement which is considerably smaller than 2.... ~.. i ... _..... at BER = 105 . _..... :::::~~gjpn . ambient illumination is collected at each relay.:..... (The plot descriptions are given in Table 3.. _.. Iil ~ 103 :::::::::::::::::::::::::: ..•... ..•.. :::::::i :::::jiiiiGG'~sslble . . the power 74 . . On the other hand... 15 30 35 Figure 3.Electrical Engineering Chapter 3.. ::::::::::::::::::::~:: ~ ::: . This deficiency mainly occurs due to the presence of background light noise. .... . .. ::::::::~~o:St4:~+:dI3:: I .. . to keep the average output power of each relay inside the eyesafe region.. . the system performance improves slightly..
In summary. by continuously increasing the number of relays. 75 . . Optical AmplifyandForward Relaying Technique Low §NRoRegiolll SNRo < 24. The other reason is the small changes in hop distances and consequently smaller reductions in channel loss. .6 km to 0. (The plot descriptions are given in Table 3. however by increasing M = 4 to 1\1£ = 5. by placing only one relay in the midway (lY£ = 0 ~ 1M = 1) the hop distance reduces by 50%. .1) of the attenuated signal over the previous hop can not be recovered completely via amplification process.Electrical Engineering Chapter 3. power budget. Therefore. the system performance always improves but after some points the gain achieved by inserting one HlOre relay is not significant enough to justify the additional costs of inserting the relay. Thus. depending on the required link coverage. the path loss and therefore the system performance improvement is less than the former case. M.82 d~ . .25 Gbps. 20 21 24 25 Figure 3.12: BER in low SNRo region for LT = 3 km and BR= 1.5 km which means only 16% reduction in L k . and financial feasibility. BER. As an example. hop distances change from 0. the number of relays is determined.McMaster University .
81 d~ 20 21 24 25 Figme 3.13. So. hop distances decrease more and the channel loss reduces. Although by dividing the link into two parts the channel loss reduces. BER of a direct transmission (DT) FSO link. the amount of collected background noise becomes nearly double. For low SNRo. is less than BER of a multihop system with one relay at the middle of the link (M = 1). . The low SNRo region corresponds to the region that for any SNRo bigger than this region.Electrical Engineering Chapter 3.. 76 .1) For low SNRo the system performance is different. i.. By inserting more relays... relaying technique always improves the system performance. no relay is placed between the source and destination nodes 1\11 = 0.. the system can not compensate the effects of background noise. . ···0 . at the very low SNRo.asing the number of relays does not necessarily improve the system performance. the average transmit power is relatively small and background noise degradation effects are dominant.. SNRo < 24. As shown in Figs. 3.McMaster University .. Simultaneously.. (The plot descriptions are given in Table 3.12 and 3. For this region.13: BER in low SNRo region for LT = 3 km and BR= 10 Gbps.e. Optical AmplifyandForward Relaying Technique Low SNR Region . incre.
signal and noise powers received at the kth relay are defined as (3.1 gG 1 gPASE (3.12) and (3.McMaster University .82 dB ) which is small enough to see the effects of background noise.1 R b gG 1 b k PASE = (3.25Gbps. the signal and noise power variations during the 3 km link are illustrated for BR= 1. Each square denotes a relay placed during the link in an equallyspaced relaying configuration. The SNR of a DT link is denoted by a blackfilled square. Optical AmplifyandForward Relaying Technique the amount of background noise increases.k _ (gG)k .39) p.42)" In Fig. In other words. The average transmit power is Pt = 10 m W (SNRo = 21 dB < 24.81 dB). and 3 relays.13). Com_promising a trade off between channel loss reduction and background noise increase determines the minimum required transmit power for which relaying technique outperforms the DT transmission. for equidistance relaying configuration. In order to understand the power variations in the low SNRo region.Electrical Engineering Chapter 3. the received signal power ~~. As expected. From (3.41 ) (3.. at this power. The required transmit power decreases by the number of relays. the background noise power pI.40) (gG)(kl) . the relaying technique not only does not improve the system performance but degrades 77 . the ASE noise power PlsE and SNRk variations during the above mentioned 3 km link are analyzed.14. performance of a DT link is better than multihop systems with 1\11 = 1. 2. 3. in the low SNRo region (SNRo < 24.
. .14: The signal and noise power variations during a 3 km link operating at BR= 1... Obviously.. relaying technique improves the system performance.81 dB) where the signal power is strong enough to combat the background noise..15 W/Hz. .'_ ___'__ _' o 1000 2000 3000 L(m) 1O2 L_'_ _'_ _' o 1000 2000 3000 L(m) Figure 3. .5 ) is taken place at high SNRo._ . .. . SNR at the destination node increases gradually. . Pt = 10mW..}'v1cMaster University . .81 dB).12 and 3. Fig. From Figs.. by inserting more relays. Whereas. Therefore for all desired SNRo.. .13. the desired BER region ( 10.. the system performance always improves by employing relaying techniques. . . .Electrical Engineering Chapter 3. . 1O3'_~ o _ ___'__ ___' 1000 2000 3000 L(m) 102'~'' o 1000 2000 3000 L(m) lOt ~=cc~~~=~~= (4) 107 ' .15 shows the power and noise variations of the system for Pt = 40 m W (SNRo = 27 dB> 24.6  10... Optical AmplifyandForward Relaying Technique 10° ~~=~~~~~""'" (1) 10° r:::cc~==:c_~=:c~:::_:c_:~ (2) i> j:Q "0 a 102 .. .25Gbps with different number of relays M.... . 3.. .. for high SNRo (SNRo > 24. The same 78 .. and No = 2 x 1O. 3. it.
. ..25 Gbps with ·different number of relays NI...17. Pt = 40 mW.. .. < ••• _. because its performance in terms of SNRo is similar to 1............25 Gbps systems........... ...........<t: . The BER gets averaged over NT = 1000 different fading 79 ....16 and 3..10 and 3. Optical AmplifyandForward Relaying Technique 10 1 (1) r================ 10° r=....11. 10 2 (4) •• _ ••• < ••• _.. ......15: The signal and noise power variations during a 3 km link operating at BR= 1.... FSO systems strongly suffer from atmosphericturbulence.. ..McMaster University ...... 1 ""p:: z <Zl ""p... 3.. 3......15 W/Hz... senario is applied to BR= 10 Gbps systems... 106 0 1000 L(m) 2000 3000 101 0 1000 L(m) 2000 3000 Figure 3. In order to reali~e the degrading effects of atmospheric fading on FSO systems performance. . 10° . has been simulated in the pres~nce of lognormal fading and shown in Figs. _.Electrical Engineering Chapter 3. •• _. and No = 2 X 10. _ • • • • • • • • • S ~ "0 ~~ Vl 10 105 ...:=::c=~:c:::::" (2) \ I 102 0 1000 L(m) 2000 (3) 3000 " 100 1000 L(m) 2000 3000 104 . whose BER are plotted in Figs. the BER of above mentioned systems. Other than background noise.........
Optical AmplifyandForward Relaying Technique Ull'iadesslbie ::::::::::::::::...::R~~@i~·::· TU 102 LowSNRRegion'... 3...1) states and at each state 106 bits are sent...25 Gbps and BR= 10 Gbps. To achieve the same BER at both bit rates for a given BER= 10...~: .16 and 3... 0_ . .. the system performance in terms of SNRo is similar at both bit rates. and different number of relays M.. it turns out that to mitigate the effects of atmospheric....~R:05?~. summatizes the difference between the average transmit power of different systems 80 ..:::::::::::::::::l:::B()~~I~:::: ····..: •• .Electrical Engineering Chapter 3..  .. respectively 8.49 dB more power must be transmitted at BR= 1. with fading effects. for LT = 3 km.t5. •• :::_ •• :.·.11. BR= 1.. "'y" .McMaster University . ..25 Gbps... Table 3.~l3.. Again. fading on a DT link.10 and 3. • "1 .. By comparing Figs.25 Gbps. ... (The plot descriptions are given in Table 3. 3...17 with Figs..5 . .s... ..2.16: BER versus SNRo.45 dB and 8.. + 15 20 35 40 45 Figure 3.. the transmitter needs to send about 9 dB more power when bit rate is 10 Gbps rather than 1..
McMaster University  Electrical Engineering Chapter 3. Optical AmplifyandForward Relaying Technique
"" "Inaccessible 'Power""
RegioA"
~
~
..
. ........ . ..  ..  ....... ....... .
"""
15
20
35
40
45
Figure 3.17: BER versus SNRQ, for LT = 3 km, BR= 10 Gbps, and differentnumber of relays lVI, with fading effects. (The plot descriptions are given in Table 3.1) with and without fading effects when BER= 10 5 . According to Table 3.2, as 1\11 increases less average transmit power is required to mitigate the degrading effects of abnospheric turbulence. Because by increasing the number of relays the hop distances decrease and therefore the fading effects reduce. The atmospheric fading varies by weather conditions and is not dependent on the system bit rate. Therefore; for a given system configuration and weather condition, the BER performance of the systeln is similar for both bit rates. The huge required average transmit power confines FSO communication systems
81
MclVlaster University  Electrical Engineering Chapter 3. Optical AmplifyandForward Relaying Technique
j
,
Table ~.2: The difference between the transmit power of different systems with and without. fading when BER= 10 5 . Number of relays BER= 1.25 Gbps BER= 10 Gbps M dB dB 8.49 8.45 0 4.41 1 4.39 2 2.25 2.33 2.00 2.06 3 1.55 4 1.56 1.33 1.33 5
to short distances even at low bit rates. Illustratively, at BR= 1.25 Gbps whose corresponding plot is shown in Fig. 3.16, a direct FSO system requires 42.4 dB
transmit SNRo to reach the 3 km distance from the source. As mentioned before, the maximum available transmit SNRo at BR= 1.25 Gbps is 39 dB, thus direct FSO transmission is not possible over 3 km link at BR= 1.25 Gbps. By placing only one relay in the midway, a 6.59 dB gain is achieved and the required transmit SNRo decreases to 35.8 dB which is inside the accessible power region. In other words, in order to communicate over a 3 km link at BR= 1.25 Gbps, at least one relay must be placed between the source and destination node. For higher bit rates the situation is even worse. As noted, at BR= 10 Gbps, the maximum available SNRo is 30 dB. Clearly, without enlploying relaying techniques to guarantee BER= 10 5 , the direct FSO system needs to provide at least 42.4dB transmit SNRo which lays inside the inaccessible power region. The best performance
Ol
l'
': a d' lIed v)
1
hill 1:' OV llnK
DOi\"
1
C11\TT) aL onno =
.L
0U
C)l"\
lr UDl ' IS
TYf:11\.
D~.tl=
7 .t10
An
X
r\._'J . . . . ' ........ 1U . 1n sect'IOn ~.l.~, It
t
T
is shown that atmospheric turbulence directly depends on the propagation distance. Therefore, the effects of atmospheric fading is considerably mitigated by reducing the hop distances. By placing one relay at the middle of the 3 km link, BER reduces to 82
NlcMaster University  Electrical Engineering Chapter 3. Optical AmplifyandForward Relaying Technique
Table 3.3: Power gains achieved by employing OAF relaying technique in a FSO communication system operating at BR= 1.25 Gbps for a particular BER= 105. Increasing the number of relays M>lV[+1 Power gain without fading dB 2.52 1.55 1.27 0.82 0.74 Power gain with fading dB 6.59 3.68 1.46 1.32 0.97
Table 3.4: Power gains achieved by employing OAF relaying technique in a FSO conllTlUnication system operating at BR= 10 Gbps for a particular BER= 105. Increasing the number of relays ]I/[>M+l Power gain without fading dB 2.49 1.61 1.15 0.86 0.74 Power gain with fading dB 6.633.63 1.47 1.31 0.96
1.39
X
10 2 . Although dividing the propagation path into two smaller parts reduces
the fading effects, to achieve the particular performance BER= 105 at the receiver, the hop distances must be shortened more to overcome the degrading effects of channel loss and high background noise. From Fig. 3.17, in order to communicate over a 3 k111 FSO link with BER= 105 at least four equidistance OAF relays must be placed between the source and destination nodes. Tables 3.3 and 3.4 respectively summarize the power gains achieved at a given BER= 105 by increasing the number of relays for bit rates BR= 1.25 Gbps and BR= 10 Gbps. By comparing the second and third columns of Table 3.3 or 3.4, the
83
25 Gbps. As shown in Figs. at BER= 10.e. Illustratively. . Optical AmplifyandForward Relaying Technique substantial role of relaying technique in mitigating the atmospheric fading effects is realized. In the following section. in a slow fading channel. the difference between the power gains obtained in the absence and presence of atmospheric fading decreases. 84 .5 . shortening the hop distances other than the channel loss also decreases the degrading effects of fading. 6.52 dB gain. Therefore.Electrical Engineering Chapter 3.McMaster University . at very short hop distances relaying technique mainly decreases the channel loss to overcome the effects of background noise and its contribution in mitigating fading effects is negligible. 3. 4 dB less. i. the system obtains only 2.5 in order to communicate over a 3 km FSO link some of the considered OAF FSO systems require to send a huge amount of power which exceeds the maximum available average power. for some configurations it is not possible to reach the 3 km distance from the transmitter.59 dB improvement is achieved. the relaying technique reduces the channel loss but increases the collected background noise.16 and 3. But when fading is neglected. However.17.5 km. at a given BER= 10.6 km to 0. considerable gain in the average transmit power. by using only one relay in the midway of a 3 km link operating at BR=1. When no fading is considered in the channel n'lodel. Consequently. . they do not change by much. The reason can be simply explained via the distance dependence of atmospheric fading. By increasing the number of relays.j when fading effects are taken into account. At short distances (less than 1 km) fading effects are negligible so that by changing the hop distances from 0. relaying technique obtains a. In other words. the maximum accessible distance from the transmitter for various relaying configurations is presented.
5) 3. BR= 1. by increasing the number of relarys from IV! = 4 to IV! = 5. without fading effects.Electrical Engineering Chapter 3.1. For example.3. However. by increasing the number of relays during a fixedlength link. at BER= 105 . P t = 500 m W. for a 3 km link. Optical AmplifyandForward Relaying Technique c.2 Maximum Accessible Communicating Distance As mentioned in section 3. for different number of relays lVI. the gain improvement reduces as the number of relays increases because the total accumulated background noise grows by the number of relays.74 dB improvement which from 85 . (The plot descriptions are given in Table 3.ance gains only 0.:: co 10::: W 3 5 10 15 30 35 40 Figure 3.3.McMaster University .18: BER versus the total communicating distance.15 W 1Hz.25 Gbps. the system perform. No = 2 X 10. LT (km) . error performance of FSO systems improves and more power gain is achieved.
there exists a compromise between the number of relays and average transmit power to achieve a particular BER. Optical AmplifyandForward Relaying Technique Figure 3. LT (km) .19: BER versus the total communicating distance. 3.Electrical Engineering Chapter 3. which are arranged in an equallyspaced relaying configuration. for different number of relays iI/I. without fading effects.18 and 3. No = 2 X 10.McMaster University .19 show the BER of different relaying systems with different number of relays.5) commercial point of view is not satisfactory.15 W/Hz. JI/I. Since average transmit power is restricted to the eyesafe region. BR= 10 Gbps. The average transmit power is assumed Pt = 500 m W which 86 . (The plot descriptions are given in Table 3. Therefore. the number of relays determine the maximum total communicating distance at a given BER and average transmit power. Figs. P t = 500 m\iV.
Ai so that BER= 10. As demonstrated before.71 0 B1 2. due to additive background noise.23 is inside the accessible power region. This distance for the system.5 provides the maximum. The effects of atmospheric turbulence are not considered in these plots.Maximum achievable communicating distance LT(km) without fading effects lIIarker Number of Relays BR= 1. for different total communicating lengths.95 e3 8.15 3.e.25Gbps BR= 10Gbps j\1 LT(km) LT(km) e4. the performance of 87 .25 Gbps and BR= 10 Gbps for different number of relays . Optical AmplifyandForward Relaying Technique Table 3.46 km when the atmospheric fading is not considered.JVIcMaster University . For each NI. i. achievable 'communicating distances at two bit rates BR= 1.5 is obtained at the receiver.72 4.5: . white noise at the receiving apertures. operating at BR= 10 Gbps is 1.f = 7) even longer distances are accessible.34 3.25 Gbps and BR= 10 Gbps respectively which is a considerable accomplishment for shortrange highrate FSO communication systems.72 km and 4.25 Gbps is 4.71 + V 7 11.09 5 10.28 5.46 1. BER is simulated for different hop distances. The huge difference between these two systems is referred to the wider bandwidth of optical elements operating at higher bit rates and consequently collection of more additive. Table 3. By increasing the number of relays (more than 1\. the maximum accessible distance for a direct FSO system operating at BR= 1. By employing relaying technique (NI = 7) FSO systems can access to 11. As indicated in this table.23 km distances at BR= 1.71 km.Electrical Engineering Chapter 3.
39 km while by changing 1\1£ = 3 to 1\1£ = 5.62 20.71 0. By increasing the number of relays. llL!{! is the percentage change in LT and is defined as M llLT llLT = LM T X 100. it turns out that OAF relaying technique has better performance at lower bit rates. Table 3.6: Distance improvements (llLT) by increasing the number of relays.06 5 7 7 1.52 14.25 Gbps BR= 10 Gbps M 7 lV[ + 2 llLT(km) llL~i[(%) llLT(km) llL!J.53 3 7 5 1. Qualitatively at BR= 1.01 multihop FSO systems improves nonlinearly by increasing the number of relays.46 0.39 40.MclVlaster University . 88 .81 km. By comparing the relative distance improvements llL!{! at two bit rates.81 21.Electrical Engineering Chapter 3.6 shows the distance improvements (llLT) by increasing the munber of relays at a particular BER= 10. High noise power not only confines the amplifier gain at each relay but also degrades the signaltonoise ratio and system performance at the receiver.25 Gbps. Increasing the number of relays BR= 1. more background noise is added to the signal.f(%) 1 7 3 2. by increasing the number of relays.01 0. where L¥J is the maxinlUIIl accessible communicating distance when ljl[ relays are employed. llLT decreases gradually.81 35. llL!{! represents the relative improven'lent in LT and can be used as a metric to compare the system performance at two different bit rates. by increasing 1\1£ = 1 to M = 3 the total communicating distance extends 2.57 15.5 . This deficiency is due to background noise. From Table 3. Optical AmplifyandForward Relaying Technique Table 3. it increases only by 1. The degrading effects of background noise are more dominant at higher bit rates.6.
.65 16.'. is defined as t:.7 provides the maximum communicating distances for various OAF FSO systems with different number of relays for both bit rates. Table 3.5 .18 3.32 7.25 Gbps by 2.98 2.Electrical Engineering Chapter 3. The lognormal atmospheric fading decreases the maximum communicating distance of a direct FSO system operating at BR= 1.7 indicates the 89 .21 are provided. In these figures. and LT are respectively the maximum achievable distances over the FSO channels with and without fading effects.33 1.33 1.L!.L~(%).LFI T LT X 100 ' where L!. 3.73 0.59 0..32 3.82 3.61 1.12 6. the BER of the system in the presence of lognormal fading versus the total communicating distance LT is plotted..T. By comparing L~ and LT given in Tables 3.29 BR= 1. Also the relative distance reductions due to atmospheric fading. N umber of Relays NI 0 1 3 5 7 L~ '(km) 2.5 and 3. and No = 2 X 10.LF T = IL T . it is obvious that atmospheric Lf in Table 3.22 1.87 0.53 14.32 47.74 38.51 1.48 5.45 To evaluates the effects of atmospheric turbulence on the maximum communicating distance.. Figs.91 BR= 10 Gbps L~ (km) Lf (km) t:.78 1.18 0.25 Gbps Lf( km) t:.10.L~(%) 22.16 0. Optical AmplifyandForward Relaying Technique Table 3: 7: Maximum achievable communicating distance in the presence of fading at BER= 10.'.98 41. t:. Pt = 500 m W.94 17.15 \tV1Hz. This value for the system working at BR= 10 Gbp's is 380 m. As mentioned before.7. t:. The length hop distance when atmospheric fading is considered. atmospheric fading deteriorates system performance.McMaster University .14 km when BER.63 14.L~(%) 2.5 and Pt = 500 mVi..20 and 3.05 37. are presented for different systems.28 37.
MclVlaster University  Electrical Engineering Chapter 3. Optical AmplifyandForward Relaying Technique
104 :::: ..
5
10
15
30
35
40
Figure 3.20: BER versus the total communicating distance, LT (km) , for different number of relays NI, P t = 500 mW, No = 2 X 10 15 'iV/Hz, BR= 1.25 Gbps, with fading effects. (The plot descriptions an~ given in Table 3.5) fading decreases the maximum accessible distance. By employing relaying technique the degrading effects of atmospheric fading are mitigated. In Table 3.7, the parameter
!::::.Lf expresses the relative decrement in the total communicating distance when
atmospheric fading is considered and is introduced as a metric to readily analyze the fading effects on the BER of various systems with different number of relays and bit rates. From Table 3.7, by increasing the number of relays, the relative distance reduction !::::.Lf decreases which means fading effects are compensated by using more
90
McMaster University  Electrical Engineering Chapter 3. Optical AmplifyandForward Relaying Technique
Figure 3.21: BER versus the total communicating distance, LT (km), for different number of relays NI, P t = 500 mW, No = 2 X 10 1.5 W/Hz, BR= 10 Gbps, with fading effects. (The plot descriptions are given in Table 3.5) relays in multihop FSO systems. The reason is distances (Ln.
sim~ply
realized by considering hop
By increasing the number of relays,· BER= 10 5 is obtained at
shorter hop distances because the'system needs to compensate the effects of more additive background noise by reducing the atmospheric fading and channel loss. At short hop distances (shorter than 1km) fading effects are negligible and therefore the relative distance reduction due to atmospheric fading decreases slightly. From Table 3.7, BER= 10 5 is achieved at shorter Lf in the systems working at BR= 10 Gbps
91
NIcMaster University  Electrical Engineering Chapter 3. Optical AmplifyandForward Relaying Technique
Table 3.8: Different system configurations for Figs.3.22 to 3.25 (Lk = 1 km) Marker Number of Relays Total Communicating Distance 11/[ LT
~
~
0 1 2 3 4 5
1km 2 km 3 km 4km 5 km 6 km
J
B
+V
compaxed with BR= 1.25 Gbps. Therefore, atmospheric fading has less contribution in reducing the maximum accessible distance at higher bit rates and background noise is more dominant. Distancedependent atmospheric fading limits FSO systems to operate over short distances. Employing relaying techniques makes FSO communication possible over longer distances by shortening the hop distances and mitigating the effects of atmospheric fading. However, additive background noise still remains as a powerful factor in degrading the performance of multihop FSO systems. In the next section, the effects of background noise on BER of different systems are investigated at two bit rates.
3.3.3
Fixed hop Lengths
Optical AF nmltihop FSO systems strongly suffer from background noise. As demonstrated in sections 3.3.1 and 3.3.2, by increasing the nUlnber of relays nlOre background noise is received at. t.he receiver. On the other hand, to guarantee an eyesafe average output power at each relay, the amplification process does not completely compensate
92
Equation (3.e. By analyzing the system performance in terms of average received SNR. and different number of relays NI without fading effects.McMaster University .22: BER versus SNRo for the constant hop distance Lk ~ 1 km.25 Gbps.Electrical Engineering Chapter 3. i.8) the attenuation of the previous channel. (The plot descriptions are given in Table 3. it is readily realized that because of background noise. Optical AmplifyandForward Relaying Technique \ Figure 3.15) simply summarizes the amplification process and mathematically formulates the average optical SNR at the receiver. extending the total communicating distance by consecutively adding more relays degrades the 93 .tion term in the denominator increases and totally the average received SNR decreases. bit rate BR= 1. constant gb the multiplica. by increasing the number of relays while the hop distance is fixed.15). From (3. In Chapter 2.
(The plot descriptions are given in Table 3. bit rate BR= 10 Gbps.23: BER versus SNRo for the constant hop distance Lk = 1 km.McMaster University . At a given SNRo. Optical AmplifyandForward Relaying Technique j Figure 3. Table 3.8 summarizes the configuration of the systems considered in these figures. both systems have similar performance at BER= 10. In order to analyze the degrading effects of background noise. From (3. an OAF FSO system with a fixed hop distance of Lk = 1 km is considered. 94 .5 . 3.8) error performance.22 and 3.23. The effects of atmospheric turbulence are not considered in these plots. and different number of relays 1M without fading effects.Electrical Engineering Chapter 3. The "BER of the system for different number of relays and therefore different total communicating distances are plotted in Figs.12). A 2 km link with an OAF relay in the midway consists of two 1 km DT links.
23 37.96 33.6.47 23. the system needs to send more power which lays inside the inaccessible power region. Optical AmplifyandForward Relaying Technique Table 3. From Fig.lVIcMaster University . How_ever.5 .12 27.Electrical Engineering Chapter 3. the ma.25 are provided to show the degrading effects of atmospheric fading on the system perfornmnce.90 2 20. the transmitter must launch additional power.84 27.9 summarizes the required average transmit SNRo (dB) so as to obtain BER= 10.5 .85 32.17 3 22. In order to compare the atmospheric fading effects on the BER of various systems. 3.5 at the receiver.5 at the receiver.25 Gbps BR= 10 Gbps N! W /0 Fading With Fading W /0 Fading With Fading o 14.01 the background power at the receiver of the 2 km link is nearly twice the background power at the receiver of the 1 km DT link.SNRo.64 20. To reach longer distances. Table 3. To mitigate the effects of atmospheric turbulence. is defined 95 .68 16.11 29.ximum communicating distance for the system operating at BR= 10 Gbps is 3 km when fading effects are not considered.75 29. Figures 3.67 25.9: Average transmit SNRo(dB) to obtain BER= 10.27 26.32 35. At BER= 10.86 32. the transmitter needs to send 3.36 4 23.48 1 18.32 31. . the FSO system overcomes the degrading effects of accumulated background noise and provides longerrange FSO links.97 dB more power to guarantee the particular BER= 10.21 25.85 23.91 5 24.24 and 3. By increasing the average transmit power. the total communicating distance in OAF FSO systems is restricted because the average transmit power is limited. Number of relays BR= 1.23. the an:lOunt of increase in the average transmit signaltonoise ratio.29 34.
McMaster University . where SNR[ and SNRo are respectively the average transmit signaltonoise ratio sent over the channels with and without fading effects to hit the target BER= 10.Electrical Engineering Chapter 3. bit rate BR= 1.24: BER versus SNRo for the constant hop distance Lk = 1 km.8) as 6SNRo = SNR[ . the required 96 .5 .25 Gbps. atmospheric fading imposes almost equal additional transmit power to the system at both bit rates. and different number of relays M considering fading effects. The variable 6SNRo for two bit rates and different number of relays are given in Table 3. Optical AmplifyandForward Relaying Technique Figure 3.SNRo. By increasing the number of relays. From this table.10. (The plot descriptions are given in Table 3.
::~ . Therefore the overall fading variance and consequently its degrading effects increase.Electrical Engineering Chapter 3.. and different number of relays J.. .. However the maximum average transmit power is restricted because of eyesafty issues and can not increase arbitrarily.. ..if + 1 fading coefficients defined over 1\1£ + 1 different 1 km DT links. By increasing the average transmit power.. :.if considering fading effects.25: BER versus SNRo for the constant hop distance Lk = 1 km..::::. Optical AmplifyandForward Relaying Technique 10° ~~r=~~~~==~r===~~~~~~~~~=~~~ :::::::::::(:: .8) additional transmit power increases because the overall channel fading is the result of multiplication of J. ... ."Inaccessible··· ·····Pow~r ::::Rt:giQIl::::: 5 10 15 SNRo(dB) 20 25 30 35 Figure 3. .. bit rate BR= 10 Gbps... ~~~~~~~~~~~....... This restriction on the average transmit power along with background noise and atmospheric effects confines OAF 97 ..::::.McMaster University .. ... the system compensates the effects of background noise and atmospheric turbulence and extends the link distance coverage.~~~~~::. (The plot descriptions are given in Table 3....
Because the FSO system working at higher bit rates (wider bandwidths) collects more background 98 .79 1.SNRo(dB) Number of Relays(M) BR= 1.07 3.81 2. Illustratively. the maximum distance coverage of the system operating at BR= 1.4 Conclusion FSO communication systems suffer extensively from atmospheric turbulence and background noise.29 km long link. Table 3.32 3. about 6. for a given P t = 500 mW and BER= 10. however this value for the similar 10 Gbps system is 3.25 Gbps is more than the system working at BR= 10 Gbps to obtain a similar BER. A serial amplifyandforward relaying technique has been introduced as a powerful technique to mitigate the atmospheric turbulence effects at long haul FSO systems.15 2 2. Qualitatively.I = 7 relays provides a 7.5 .59 3.6. Optical AmplifyandForward Relaying Technique .10: The relative SNRo increase . fading effects are reduced and FSO systems access to long distances at a lower average transmit SNRo.84 5 3.5 .59 4 3.6 dB improvement is achieved in the transmit SNRo at BER= 10.78 FSO transmissions to short distances. By deploying more relays and decreasing the hop distances. by inserting only one relay at the middle of an FSO link with length LT = 3 km.25 Gbps system with }I.20 1 2.McMaster University .11 3 3.25 Gbps BR= 10 Gbps o 1.63 km.Electrical Engineering Chapter 3. However. the 1. 3.26 2. performance of the system is similar at both bit rates. It has been numerically shown that for a given SNRo.
\ 99 .McMaster University . for a given transmit power. Since the average transmit power is limited due to eyesafety regulations. an optical regenerative relaying method using nonlinear optics is developed to reduce the background noise effects and increase the communication distance coverage. the background noise also increases so that the distance improvement reduces. Therefore. By increasing the number of relays. a trade off is compromised between the average transmit power and the number of relays. the 10 Gbps system affords less SNRo and hence shorter communicating distance. besides the total communicating distance.Electrical Engineering Chapter 3. to reach a specific communicating distance at a given BER. Optical AmplifyandForward Relaying Technique noise at the receiving lens. Increasing the number of relays is accompanied by collecting more additive background noise at relays that degrades the system performance. therefore. background noise still remains as a limiting factor in FSO communication systems. the number of relays determines the maximum communicating distance. In Chapter 4. Although the OAF relaying technique reduces the effects of atmospheric fading.
I Chapter 4 Optical RegenerateandForward . In this chapter the Optical RegenerateandForward (ORF) relaying technique 100 . however. In electrical AmplifyandForward relaying technique at each relay the optical signal along with the background noise is converted to an electrical signal which is amplified by an electrical amplifier. But growing background noise remains as the major drawback of freespace optical relaying systems.ments still remain in the system. By elllploying an optical band pass filter (BPF) at the beginning of each optical relay. the outofband frequency components of the accumulated background noise are eliminated. the inband noise comp<. After amplification. Relaying Technique FSO communication systems are strongly' affected by the atmospheric turbulence fading and background illumination. the optical intensity the field distribution of the signal is analyzed. the OAF relaying technique has been introduced as a powerful method to mitigate the degrading effects of atmospheric turbulence. In an FSO system. the electrical signal is modulated by a photodiode and retransmitted to the next relay. In Chapter 3.
The rege. an optical BPF centered at the signal frequency Wo is required.2. Optical RegenerateandForward Relaying Technique is proposed to eliminate the background noise at each relay. the internal structure of an ORF relay is described in detail.nerator suppresses the noise in zeros and the amplitude fluctuations in ones of optical data streams. The optimum gain G 1 depends on the average received SNR at each relay and for a given average transmit power and background illumination irradiance is constant. The regenerated pulses are amplified by another amplifier with gain G 2 and transmitted through the next relay. A sample of Gaussian pulses propagating through an ORF relay are shown in 101 . 1 4. The bandwidth of the filter is assumed to be equal to the bandwidth of the photodetector.Electrical Engineering Chapter 4. In ORF. When a regenerator is used in an FSO system where broadband white noise is accumulated. The regeneration process is performed by a regenerator. Pt.1 ORF Relay Structure The internal structure of an ORF relay is shown in Fig. Here it is assumed that the optical filter does not attenuate the optical signal. the quality of the received noisy signal is optically restored by a regenerator. The filtered signal is amplified with gain G 1 so as to adjust the average power of the signal to a level suitable for the regeneration process. The photo detector bandwidth is dependent on the transmission bit rate and is found so that the best BER is achieved at the receiver. The regeneration process will be analyzed in detail in Section 4.McMaster University . This filter is placed at the input of the regenerator so as to reject the noise outside the signal spectrum [63]. In the next section.1. The gain G 2 is adjusted so that the average transmitted power at the output of each relay is equal to the average transmitted power at the source. 4.
The regeneration process is performed by utilizing the effects of selfphase modulation (SPM) on the signal in a nonlinear (NL) medium followed by an optical filtering at a frequency of wf which is shifted with respect to the carrier frequency of the input data Woo The optical fiber with high nonlinearity coefficient 'Y is employed as the nonlinear medium used in the regenerator. 4. and (m 2 ) is the effective core area.3. 4.McMaster University .1: The typical structure of an ORF relay. c (m/s) is light velocity.2 Optical Regenerator To analyze the regeneration process for suppressing the signal background. (4. 4. th€ typical structure of a regenerator is shown in Fig. Aelf 102 . the regenerator refines the signal by removing the noise at zeros and amplitude fluctuations in ones. The regenerated Gaussian pulses carry less noise and have a small displacement compared with the original pulses.2 shows at different points. As shown. Optical RegenerateandForward Relaying Technique r1 I I A I I I I ill BPF B ~ Regenerator Optical Ie I ~ I : ~ Figure 4. Fig.1) where n2 (m 2 /vV) is the nonlinearindex coefficient.Electrical Engineering Chapter 4. The regeneration process and internal structure of a regenerator are described in the next section.
93 2 1 t(s) 0 1 (c) The Regenerator Input (Point B) (d) The ORF Relay Output (Point C) Figure 4.93 2 1 t(s) o 1 2 t(s) a (a) TI.rv.4 4 0. fI.8 1\ 8 0. 2 1 0 t(s) 1 2 x 1010 ..'ansmitted Pulses (b) The ORF Relay Input (Point A) 0.r' a 3 ~ f\ /\.005 1 O~~~~=CL''"~=""".I 2 0.2: Gaussian Pulses propagating through an ORF relay at different points.McMaster University . 103 ..6 6 '":::> 0.'' .Electrical Engineering Chapter 4.. Optical RegenerateandForward Relaying Technique 4 0.2 2 f\.01 3 .
2 /32 (4. on the signal intensity causes selfphase modulation (SPlVI) which governs spectral broadening of optical pulses. T) is quantified by the nonlinear Schrodinger equa"tion as [64] (4. The propagation of optical pulses inside a singlemode fiber in terms of normalized amplitude S(z. The dispersion length LD and nonlinear length LNL are given as LD=~ T.1) account for the fiber dispersion and fiber loss respectively.Electrical Engineering Chapter 4. e.g optical fiber.5). The dependence of the refractive index in nonlinear media.3: The internal structme of a regenerator.lVIclVIaster University .3) where PM is the peak power of optical signal given by (2. By substituting 104 .2) where z is the propagation distance. To analyze the effects of fiber nonlinearity on the selfphase modulation. fiber dispersion is ignored and the dispersion coefficient is set to zero. The parameters T = t/To is normalized time scale and To is the /32 and (TF(m. pulse width. Optical RegenerateandForward Relaying Technique 1. 1 1 1 1 1 1 1 1 1 J t) NLMedium {UO !\ m f 1 1 1 1 1 1 1 1 J 1 Gaussian BPF ~ Figme 4.
5) and (4. t) increases with the propagated distance z.4) where Ut(O. t) induces SPM spectral broadening. t) and consequently ¢Ndz. Moreover. temporally varying phase shift induces temporally varying frequency shift. The parameter Zeff is an effective distance that because of fiber loss is less than" = z (in the absence of fiber loss. Optical RegenerateandForward Relaying Technique (32 = 0 in (4. t)12 remains unchanged ((32 = 0) .8) 105 . the phase shift ¢NL (z.exp (az) = '. t) is the field envelope at z = 0 and is defined from (2. a 0. As indicated in (4. From (4. In general. t)12) (0 at at z I I Zeff LNL (4. Zeff = z).6).McMaster University . the instantaneous frequency shift respect to the central frequency Wo ~WSPM(t) with is obtained as ~WSF:M (t) = I a¢Nd .. t) induced by SPM is intensitydependent while the squared field envelope governed by IUt(z.7).Electrical Engineering Chapter 4. t) = I~ (IUt .4) for a single Gaussian pulse as (4. the time· dependence of Ut(O.6) with Zeff 1. a (4.:. therefore in NL medium the amount of spectral broadening differs across the pulse. the nonlinear Schrodinger equation can be solved as (4.4).2). the nonlinear phase shift ¢Ndz... By taking the time derivative of phase shift.
the spectrum broadening of a Gaussian pulse propagating inside a lossless optical fiber (Zeff = z) is obtained as (4. z. the length of the fiber is a main factor in determining the desired spectrum bandwidth at the output of the NL medium.e. the pulses pass through a Gaussian optical filter [65] whose center frequency wf is shifted with respect to the input signal carrier frequency 106 . By substituting (4. the signal power is distributed over more frequency components as the pulses propagate down the fiber. These SPMinduced frequency components broaden the pulse spectrum. 4. In other words. new frequency components are continuously generated as the pulse propagates through the fiber.4 shows the spectrums of the data stream presented in Fig. It is clear that the spectrum broadening increases by the initial bandwidth Ilwo.8). the average pulse intensity Ip and the propagated distance z. Optical RegenerateandForward Relaying Technique This difference is induced by SPM and increases with the propagated distan<:. Clearly.9) By plugging the value of LNL from (4. the average spectrum broadening for one Gaussian pulse is calculated as [63] (4. As shown.McMaster University .5) into (4.3) into (4.10) where Ilwo/27r = l/TFwHM is the 3 dB spectral bandwidth (TFWHM is the full width at half maximum) and Ip = Pt/A eff is the average pulse intensity which varies for different pulses. After the nonlinear medium. the nonlinear refractive index n2.Electrical Engineering Chapter 4.2 for different propagated distances z inside a NL optical fiber. Figure 4.9) and taking the time average over one bit interval T b .
McMaster University . 107 .Electrical Engineering Chapter 4. Optical RegenerateandForward Relaying Technique 1 70001.4: The Gaussian pulse spectrums for different propagated distances. z.94 2 f(Hz) = o 2 f(Hz) o (a) Transmitted Pulses (z 0) (b) z = 2 km 2 f(Hz) o 2 4 X 1011 .~~~ 6000 5000 N 4000 ~ 3000 2000 1000 . (c)z=4km (d) z = 6 km Figure 4.
11) As mentioned before.6. Optical RegenerateandForward Relaying Technique by a certain value Wo .wshift (4. the output 108 . = . noise in zeros. so that the new spectral bandwidth satisfies (4.12) the pulse is rejected by the filter. WSPM = . The spectral width of the filtered pulse is determined by the filter spectral bandwidth. e.6.McMaster University .13) depending on the amount of pulse broadening. and the filter bandwidth .6.< (4. e. By changing the filter spectral bandwidth.wf.6.wo. there exists a displacement between the regenerator input pulses and the filtered ones. .g.6. the bandwidth of the SPMbroadened spectrum changes. If the pulse intensity is high.g.Electrical Engineering Chapter 4. at ones.ws hift: Wf = Wo + . When the pulse spectral broadening is small so that the spectral bandwidth of the selfphase modulated signal at the output of the NL medium. If .wo + . a part of the SPMbroadened spectrum passes through the filter and the rest of them are rejected. this occurs when the pulse intensity is small.wf pulsewidth is the same as the input pulsewidth.wSPM. the filter center frequency. 2. satisfies the following equation WSPM .6.6. This misalignment originates from the shifted center frequency of the Gaussian filter.6.ws hift. In the regenerator.
Section 4. an amplifier is required after the regenerator to boost up the signal and retrieve its average power to the average transmitted power at the source.14) Ie if Ip > Ier where Ie is the constant output pulse intensity and Ier is the critical pulse intensity which is adjusted to a level so that compromise a trade off between removing . the ORF relay receives the noisy signal and refines it via regeneration process.3 provides the 109 . If leT' is chosen to be very small. if the critical intensity is selected too big. the regenerator cannot completely remove the noise at zeros.2. the regeneration process imposes a considerable power loss to the signal by rejecting a major portion of the pulse spectrum.5 shows a typical transfer function of an ideal regenerator and its perfect performance in removing noise at zeros and suppressing the amplitude fluctuations at ones. Therefore. the pulse spectrum broadens excessively so that the signal is distorted by generation of new outofband frequency components. Therefore. the intensity of the output pulse is independent of the input pulse intensity. a pulse transfer function for regenerator in terms of output pulse intensity versus input pulse intensity can be established as if Ip < Ier lout = { 1 0 (4. 4. On the other hand. WSPM > > . then amplifies and retransmits it through the next channel. Optical RegenerateandForward Relaying Technique In the case where the input pulse intensity Ip is very high. so that the pulse spectrum broadens extensively.14) is an ideal transfer function. by extensively broadening the pulse spectrum.the noise at zeros and suppressing the amplitude fluctuations at ones. the more parts of the signal are rejected by the filter that imposes more power loss to the system. Pt. The transfer function expressed by (4. Fig.McMaster University . As expressed earlier. As shown in Table 4.Electrical Engineering Chapter 4.6ws hift. Ip > > I er . Otherwise.
.anc. Optical RegenerateandForward Relaying Technique lout lout i" rrr. r    .3 Numeric.lVIcNIaster University . 4.e of various ORF n:lLLltihop FSO communication systems is investigated.      . t r Output Signal Figme 4.. Because of the long simulation time of MATLAB ... numerical simulations on ORF relaying technique and compares its performance with OAF multihop systems.Electrical Engineering Chapter 4. \ . a Qfactor 110 .5: Typical transfer function of an ideal regenerator.:}l Results and System Performance In this section. the perforIn.
are previously introduced in Chapter 3.Electrical Engineering Chapter 4.Nlctor Est"inlation Simulating the BER of an ORF ll1ultihop system takes an extremely long time. ORF relaying technique is proposed as a powerful method for optically suppressing the effects of background noise at high bitrates. The gain of the first amplifier G 1 is adjusted for each launch power and hop distance so that the best performance at the receiver is obtained. The Gaussian filter inside the regenerator has the 3 dB bandwidth of f1Wj /27r = 29 GHz. a fiber of length z = 8 km with effective core area Aeff = 45 JLm 2 at . The considered FSO system has the same specifications as the OAF FSO systeln that . The degrading effects of atmospheric turbulence are not considered in ORF systen'ls. In this thesis. the performance of ORF multihop systems are investigated at 10 Gbps and compared with the performance of OAF multihop systems. At BR= 10Gbps.3. Q factor is frequently utilized to evaluate the performance of optically amplified systems 111 . the Qfactor approximation is widely used instead of MonteCarlo method for BER simulations. therefore alloptical relaying techniques are attractive. 4. Optical RegenerateandForward Relaying Technique estimation method is utilized to calculate the BER of FSO systems. and the filter frequency offset with respect to the input signal carrier frequency is f1wshift/27r = 100 GHz.1 Q. As demonstrated in Chapter 3. The NL medium specifications are chosen as what Mamyshev considered in [63] for experimental simulations.McMaster University . On the other hand. wideband FSO systems suffer from background noise and their maximum accessible distances are limited due to accumulated background noise. electrical processing limits the bandwidth of high rate optical systems.A = 155"0 nm is used. When the additive noise has a Gaussian distribution. Here.
In the all MC analyses considered in this work. The BER of an OOK modulated signal at the receiver is related to the Q factor as [66] J (4. The total number of transmitted bits is N = 214.15) where the complementary error function erfc(x) is defined as erfc(x) = In J 00 e. t2 (4.6 shows the BER of an ORF multihop system with the total communicating distance of 3 km where one ORF relay is placed at the middle of the link. BER of the system is simulated using both MonteCarlo (MC) and Qfactor estimation (QF) methods. In ORF relaying systems due to fiber nonlinearity.2.Electrical Engineering Chapter 4. In numerical simulations. By squaring the sampled values and taking average. However. t = kTb /2.e at the receiver is not exactly Gaussiandistributed. it can be shown that a Gaussian approximation is a reasonable estimation for the noise distribution. the variance aD (a1) of the pulse stream" is also obtained. the mean ~O (hl1) of a pulse stream is simply found by getting average over the values obtained by sampling pulses at the middle of the bit intervals.dt... .McMaster University .1. the accumulated noi::. 4.. k = 0.16) x The Q factor is also expressed as (4. at least 100 errors in the received bit streams are required 112 . Optical RegenerateandForward Relaying Technique [66].17) where ~lO (~1) and aD (a1) are the means and the standard deviations of zeros (ones) respectively. Fig.
Electrical Engineering Chapter 4.5 when QF method is employed.3 . the smallest reliable BER obtained by MC method will be around BER=1O. to be able to rely on the received BER.6: BER of an ORF FSO system with M = 1 and LT = 3 km. simulation results indicate that 214 sent bits are enough to rely on BER as low as 10. the performance of the system driven by Qfactor estimation is very close to the MonteCarlo simulation result (less than 1 dB offset). obtained by MC and QF methods. As seen from the figure. in the rest of this section. Optical RegenerateandForward Relaying Technique Figure 4. QF is considered as a reliable method used to calculate BER of different 113 . Therefore.McMaster University . By sending N = 214 bits and taking average on about 20 different iterations. However.
In all systems.5 km).5 2 1. The distance bet\. From Fig. configurations for Fig. 4. 114 . In the follmving simulations.3.7 corresponds to the BER of three different ORF systems. it can be assumed that equallyspace relaying may also provide one of the optimal configuration1') for ORF multihop systems. 4.1 summarizes the configurations of the systems considered in this figure.7 (LT = 3 km) lVlaTker Ll (kIll") L2 (km) B1 2 ~ 1. although enough evidences do not exist for its proof.7. Table 4. ORF relays are arranged in serial equallyspace relaying configurations. Optical RegenerateandForward Relaying Technique Table 4. there exists no explicit mathematical relations between the hop lengths and the average received SNR at the receiver.MclVlaster University .4. the lowest BER corresponds to the system where the ORF relay is placed at the middle of the link (Ll = L2 = l. Hence. For ORF systems.veen the source and ORF relay is denoted by Ll and the distance between the 0 RF relay and the destination is L 2 . the equallyspaced relaying configuration provides the best performance at the receiver of OAF systems.1: Different system. Therefore) numerical simulations is utilized to found the best configurations. 4.5 1 ORF FSO systems.2 Optimal Relaying Configuration As proved in Chapter 3. the total communicating distance is LT = 3 km and one relay is placed between the source and destination nodes.Electrical Engineering Chapter 4. Fig.
13). (The plot descriptions are provided in Table 4.8 compares the performance of ORF multihop systems with OAF systems. the BER of ORF FSO systems for NI = 1 and N! = 2 relays are illustrated while the total communicating distance is fixed (LT = 3 km). Fig.MclVlaster University . ! Figure 4. Table 4. In this plot.3.3 Fixed Total Communicating Distance In Chapter 3. 3. it was shown via numerical sill1Ulations that increasing the number of relays between the source and destination nodes while keeping the total communicating distance fixed.1) 4. improves the system performance (Fig.2 summarizes the configurations 115 . LT = 3 km.Electrical Engineering Chapter 4. 4. and BR= 10 Gbps. Optical RegenerateandForward Relaying Technique .7: BER of different ORF multihop FSO systems with M = 1.
Electrical Engineering Chapter 4. It is expected that by increasing the number of relays.51 dB (All = 2) improvement in SNRo with respect to OAF relaying technique. in OAF systems.McMaster University . Thus by increasing the number of relays.8: BER of different ORF multihop FSO systems with All = 1 and 1\11 = 2 for a fixed total communicating distance of LT = 3 km and.2) of the systems considered in this figure.5 . As mentioned in Chapter 3. at a given BER= 10. BER improvement decreases 116 . (The plot descriptions are provided in Table 4. utilizing more relays injects more accumulated background noise to the system. ORF systems outperform OAF systems more rapidly. by replacing OAF relays with ORF relays. Optical RegenerateandForward Relaying Technique 5 10 15 SNRo~dB) 20 25 30 35 Figure 4. the system gains 2. From Table 4. BR= 10 Gbps.2.93 dB (All = 1) and 3.
in ORF systems.8 (LT = 3 km) lVlarker ]1/[ Lk (km) Relaying Technique SNRo (dB) & x.4 Combination of OAF and ORF Relays In this section. by replacing one OAF relay by an ORF relay. System 2 + System 3 (4) in Table 4.e.JVlcJVlaster University . 4. the performance of various multihop systems composed of both OAF and ORF relays is investigated.5 1 (Table 3.53 26.Electrical Engineering Chapter 4. Fig. increasing the number of relays is equivalent to decreasing hop distances and channel loss and therefore the average received SNR increases by the number of relays. it is expected that in a fixedlength link.. Optical RegenerateandForward Relaying Technique Table 4. background noise is ideally eliminated at each relay.4. Consider a 3 km FSO link with two equallyspaced relays placed between the source and destination nodes (the hop distances are 1 km).2: Different system configurations for Fig.33 dB) gain is achieved in SNRo at BER= 10.4).3.93 31. Surprisingly.01 dB (1. if the signal is first amplified by the first relay and then 117 . As illustrated in Fig.9. In sunlnmry.84 28.EJ 0 l' 2 3 1.3. i.46 29. 1.3. 4.5 . 4. hence background noise does not propagates through the channel and its average power is almost constant at all relays. On the other hand.33 1 2 1. BER improves by increasing the number of relays and its improvement increases rapidly. In ORF systems.9 provides the BER of various relaying schemes whose configurations are summarized in Table 4.5 1 DT (no relay) OAF OAF ORF ORF 33.
By replacing both OAF relays by ORF relays. in other words. 25 30 35 Figure 4. At high SNRos.3) regenerated at the second one (System 3) or it is first regenerated and then amplified (System 4). System 4 has slightly better performance. BR= 10 Gbps.50 dB (2. System 2 (3) + System 5. Optical RegenerateandForward Relaying Technique 1. regenerator has better performance at high SNRo.McMaster University . (The plot descriptions are provided in Table 4. i. the performance of both systems is nearly the same. Because the received SNR is higher and the regenerator suppresses t. This value is 118 . the system performance improves by an additional 2.he amplitude fluctuations at ones more evenly.18 dB).9: BER of different nmltihop FSO systems with M = 2 for a fixed total communicating distance of LT = 3 km and.e.Electrical Engineering Chapter 4.
' The reason can be simply explained: the regenerator eliminates the background noise at each relay therefore by using all ORF relays. System 5 benefits from the regeneration process at the second relay more than System 3. NI System No. Although atmospheric fading is suppressed 119 . and then the noisy signal is regenerated at the second relay.18 dB gain with respect to the direct transmission (System 0.3. Obviously. the signal is first aInplified by' an amplifier in the first relay.51 26. In general.33 ~ +~ B almost twice the gain achieved by replacing only one OAF relay with an ORF relay (System 2 7 System 3 or 4).9 (LT = 3 km.MclVIaster University .EJ no relay OAF OAF ORF ORF no relay OAF ORF OAF ORF 33. System 5 achieves totally 3. regenerators have better performance at higher SNRs.83 28.5 Maximum Accessibie Communicating Distance The OAF relaying technique has been introduced as a powerful method in mitigating the effect of atlnospheric fading and consequently increasing the maximum achievable distance in FSO con1Hmnication systems. a negligible amount of background noise remains in the system. the noiseless signal regenerated at the first relay is amplified along 'with the collected background light at the second relay.Electrical Engineering Chapter 4.84 28. In System 4.) 4.57 dB gain with respect to System 2 and 8.93 29.3: Different system configurations for Fig. Optical RegenerateandForward Relaying Technique Table 4.ance of System 4. lVIarker First Relay Second relay SNRo (dB) 1 2 3 4 5 = 2) e. In System 3.4. This amplified' background noise considerably deteriorates the perform.
70 2.'{imum accessible distance increases respectively by 0. The number of relays at each system is either one or two and relays are arranged in an equallyspaced relaying configmation.10 provides BER of various OAF.10 0. the ma.lVIclVlaster University .70 1.87 1. ORF and hybrid OAF /ORF multihop systems.15 \iV/Hz.5 ) Relaying Relaying Distance Lk System . 0 1 2 3 4 5 Ij e.40 1.50 1.Electrical Engineering Chapter 4.10 (Pt = 500 mvV.62 3.4: Different system configmations for Fig. Optical RegenerateandForward Relaying Technique Table 4. :Marker Number of Relays LT I'll Technique (km) (km) No. The ORF relaying technique is proposed as an effective technique to reduce the background noise and increase the maximum. maximum accessible distance 120 . By replacing OAF relays with ORF relays in System 1 and 2. Table 4. in ORF systems as the number of relays increases. i. No = 2 X 10. 4.4. 4.86 km.10. and BER= 10. 1 + System 3 and System 2 + System 4 in Table 4.EJ 0 1 2 1 2 2 DT OAF OAF+OAF ORF ORF+ORF ORF+OAF 1. By comparing ORF and OAF relaying systems in Fig.e. amplifyandforward FSO systems encounter an essential difficulty originating from accumulating background noise at each relay. The average transmit power is Pt = 500 m VV for all schemes and atnlOspheric fading effects are neglected in this study.48 3. accessible distance in FSO communication systems.08 4. it is realized that ORF technique improves the maximum achievable distance several times more than OAF technique.4 summarizes the configmations of the systems considered in this figme. Despite OAF technique.13 x*B v by OAF relaying technique.5"4 1.20 2. Fig. System.4.88 km and 1.
g. additional 1. Optical RegenerateandForward Relaying Technique \ Figure 4. by using another ORF relay (System 4). (The plot descriptions are provided in Table 4.10: The maximum accessible communicating distance of different multihop FSO systems with lv'I = 1 and lv'I = 2 when Pt = 500 m VV. e. No = 2 X 10. background noise is mostly eliminated at each relay and shortening hop distances no longer is required. the hop distances in the' considered ORF systems are 121 .McMaster University .15 W 1Hz.4) improves more rapidly. from Table 4.4 km increment is obtained in LT' As mentioned in Chapter 3. and BR= 10 Gbps.38 km. By deploying ORF relaying technique.g. LT increases by 1. e.Electrical Engineering Chapter 4.4. maximum LT occurs at shorter hop distances so that the system overcomes the effects of accumulated background noise. by utilizing only one ORF relay (System 3). by increasing the number of relays.
using OAF relays is equivalent to injecting background noise to the system. if atmospheric turbulence effects are neglected. For long hop distances where the received SNR at relays is very small. This similarity accentuates the superior performance of ORF technique to OAF method and its dominant role in hybrid systems for improving the system performance. The total communicating distance increases by subsequently adding a 122 .McMaster University .13).3. 4. Although System 5 takes advantage of one more OAF relay. The relaying system which consists of two equallyspaced ORF and OAF relays (System 5) has slightly better performance than the system with just one ORF relay (System 3).10. employing OAF relays not only does not improve the system performance but it also deteriorates it (Fig. Fig. This small improvement corresponds to the shorter hop distances to compensate the effects of background noise. the maximum accessible distance extends linearly by the number of relays. the performance of System 3 is better than performance of System 5. in ORF systems. 3.6 Fixed hop Lengths In order to compare ORF technique resistivity to background noise with OAF technique. Under this circumstance.11 is provided. 4. ORF relaying technique can greatly extend the total communicating distance LT by desirably increasing the number of relays. at very long hop distances (Lk ~ 2. 4. Optical RegenerateandForward Relaying Technique almost 1.5 km for both System" 3 and System 4. because at very low received SNRs. Next section is spent to compare resistivity of both techniques in the presence" of background noise. performance of System 3 and System 5 are very similar even at short hop distances. In the systems under consideration. As shown in Fig.Electrical Engineering Chapter 4. hop lengths are fixed Lk = 1 km.2 km). Ideally. totally.
. 1 ' 0 . ..Jf. :.e~~~~~~~x:~::~~<.. ... . \ t::\:~ : \ .. ··i·········~~\"1>\2~{·········· .... L .McMaster University . .. .:.. ..:.. _... ..... \. ___... it is shown that this increment in background power deteriorates the system BER and limits its maximum accessible distance... _ ... ( I i~' ·······t· "'T ... ... .'. ..: . ::: :. .:.. ···········c: \ \ ... .. \.. .5 .. _..q . .. ~ .~~..5) new relay to the previous system.:.. (The plot descriptions are provided in Table 4...... Table 4. . .... ... In the OAF relaying technique....:.. ... .. _ ~ ... 4.\... . .or . .Electrical Engineering Chapter 4..~.......11.. ::: :1:::': 5 10 25 30  Figure 4... .:. ::::X:: :::~~\~~~~::\~~~~:~: .... . ..11 indicates that increasing the number of relays in ORF systems does not influence the system performance because the accumulated background noise is eliminated at each relay. . \. :8~> ......  ". ::::c::: .: 1: :::::'.~~~~~~~~~~~~~~<=~= ..•. The amount of accumulated background noise increases by the number of relays. From Fig...... ..:. 1 ......11: BER of different ORF FSO systems with different number of relays M for a fixed hop distance of Lk = 1 km and BR= 10 Gbps..... . Optical RegenerateandForward Relaying Technique 100 ~~~"~"~'~"'~:~:~~~. .. .... .\ .. ... .. .. .... ... .. \i ~ .. . : .:i::Q ·...5 summarizes the configurations of the systems considered in this figure.l .....:.. . ..:::::~:::~:::~.... . However Fig...\.. ! Q) 1 __ . by replacing ORF relays with OAF relays. . 123 ....... .. at a given BER= 10.:. 4..:.. _.. .
It has been s11m'ln that by Initigating atmospheric: fading effects via reducing hop distances.84 27.11 (Lk = 1 km) Marker 1\.75 25. 2.33 ~ x eB~ 3 2 1 4 3 2 the average transmit SNRo improves by 1.4. 1\. In this chapter. 4. the link coverage can be extended. and 1\. by increasing the number of relays (increasing the total communicating distance) more average transmit power is required to guarantee a specific BER at the receiver. the OAF relaying technique has been developed to reduce the degrading effects of atmospheric turbulence. This feature makes the ORF relaying technique a distinguished method for removing the background noise in multihop' FSO communication systems and leading to significantly extending the FSO communication distance coverage.36 dB when the number of relays are 1\.5: Different system configurations for Fig.8 3 2 I 4 3 2 OAF OAF OAF ORF ORF ORF 31.£ = 2. the system BER relTlains nearly unchanged. by subsequently adding more relays. But the magnitude of improvement is limited due to growing accumulated background noise during the channel.£ = 1.29 29.92 dB and 5. However when ORF technique is used. In OAF systems.4 Conclusion In Chapter 3.£ LT(km) Relaying Technique SNRo (dB) 0.McNIaster University .96 26.1 = 3 respectively. Table 4.31 26. ORF relaying technique has been proposed as 124 . Optical RegenerateandForward Relaying Technique .44 dB.Electrical Engineering Chapter 4.
< 1 X 10. the amplitude of the regenerator output pulses are not completely equal and they have negligible 125 . it requires more sophisticated equipment such as an automatic gain controller.88 km with respect to direct transmission ruid a similar OAF multihop system.. BER performance of different ORF systems operating at BR= 10 Gbps are investigated and compared with the performance of OAF systems. Optical RegenerateandForward Relaying Technique a powerful method to optically remove background noise at each relay.2/ 3 [52]) has been considered under which the atmospheric fading effects can be neglected over distances up to 1 km. The regeneration process is highly sensitive to the incident signal SNR. etc. especially for large number of relays and longer distances. In ORF systems. by employing only one ORF relay. optical Gaussian filters. the preamplifier adjusted gain.17 m. background noise is mainly eliminated by the relay and does not propagate through the channel. Also as mentioned.38 km and 0. which increase the complexity and implementation costs of the system. The results indicate that the distance coverage of ORF FSO links can be significantly extended by increasing the number of relays because collected background light at each relay is mostly eliminated by a regenerator.Electrical Engineering Chapter 4. The error propagation is another performance limiting factor in ORF systems. two amplifier at each relay. adjustable nonlinear optical medium. In general. the total communicating distance increases respectively by 1. Assuming this condition. and the Gaussian filter frequency offset and bandwidth. this feature which is also involved in the decodeandforward relaying technique causes ORF systems to significantly outperform OAF systems.McMaster University . A very weak turbulence fading condition (C. Qualitatively. the regenerator output signal has a small displacement with respect to the original pulses that in some applications this delay might be undesired [63]. Although the ORF technique provides superior performance to OAF technique.
g. e. Optical RegenerateandForward Relaying Technique fluctuations. it can even add extra amplitude fluctuations to the output pulses. These fluctuations.Electrical Engineering Chapter 4. at bit" 1". incr~ase J i the system stability al'ld reduce the 126 . These challenges in the proposed ORF system provide motivation to finding methods to implementation complexity.McMaster University . If the regenerator is not well adjusted and does not work at its optimal operating condition. propagate through the channel and by subsequent regeneration in the following relays lead to wrong detection of the pulse (completely rejecting the pUlse).
1 Conclusion This thesis presents new optical relaying techniques to mitigate atmospheric turbulenceinduced fading effects and eliminate background noise in free space optical (FSO) communication systems. The BeersLambert law is modified to find 127 . a new channel model is developed which characterizes the variations of intensity and phase of the optical signal during wave propagation. An additive AWGN channel is assumed in which background illumination is the dominant source of noise. 2) lognormal fading under weak atmospheric turbulence conditions. Three primary factors have been considered to model the free space channel effects: 1) atmospheric attenuation which includes both absorption and scattering contributions. and 3) propagation loss due to optical beam spreading through optical channel. In order to define an all optical relayassisted FSO system.Chapter 5 Conclusion and Future Work 5. The main contributions of the thesis are proposing alloptical amplifyandforward (OAF) and regenerateandforward (ORF) relaying techniques and applying them to relay assisted FSO systems.
filtering (as needed). amplification. e. Performance of OAF systems has been considered for two bit rates BR= 1. etc.g. Since the optical field envelope is analyzed. It is numerically shown that by increasing the number of relays between source and destination.25 Gbps and BR= 10 Gbps under no fading and weak fading effects. It also provides less delay and complexity in implementation rather than the previously considered AF communication systems which need OE and EO conversions at each relay. hop distances decrease and consequently distancedependent atmosphericinduced fading is mitigated. The OAF relaying technique is proposed as a powerful technique for mitigating the atmnspheric turbulenceinduced fading while all relaying processes. However. distance improvements slow down by increasing the number of relays because the collected background light at each relay is accumulated 128 . the proposed model provides more accurate estimation of beam propagation loss especially over short ranges (a few hundred meters).McMaster University . are perfon~ed in optical domain. The numerical results show that the new model for propagation loss is close to the conventional model (geometric loss). longer communicating distances are accessible for a given average transmit power. both amplitude and phase of atmosphericinduced fading are statistically modeled and new definition for propagation loss is defined. These fluctuations degrade the overall system performance. however. In fact by employing more OAF relays. The atmospheric channel effects are assumed as multiplicative complex terms which are multiplied by the field envelope and induce intensity and phase fluctuations at the receive aperture. The alloptical AF relaying technique allows users to take advantage of high data rate (wide bandwidth) optical transmissions over longer distances. Conclusion and Future Work the atmospheric attenuation factor applied to the optical field envelope.Electrical Engineering Chapter 5.
Electrical Engineering Chapter 5. Due to its superior performance to AF relaying technique. complexity of encoding/decoding processes. Although the ORF technique distinguishably outperforms the OAF method in terms of BER performance. DF method is recently applied to FSO systen'lS as well. The accumulated background noise is the major limiting factor in improving the maximum accessible distance of OAF multihop systems. The simulation results indicate that by employing ORF technique. the background noise can be eliminated at each relay that results in accessing greatly longer communicating distances. Conclusion and Future Work during propagation and totally more noise is added to the signal. 129 . Under this assumption. In this thesis. Despite the OAF method. In other words. the ORF technique used in multihop FSO systems makes high data rate communications possible over a greatly extended communicating distance by employing reasonable number of relays.lay. I \ The decodeandforward (DF) relaying method has been widely used in RF systems to remove the effects of background noise at each relay. The ORF systems operating at high bit rate BR= 10 Gbps are considered under a very weak atmospheric turbulence condition where atmospheric fading effects can be neglected over short distances (less than 1 km). The considerable loss in data rate. the ORF relaying method is developed as a promising technique for removing background noise at each re. and delay induced by electrical processors at each relay accentuate the need for an innovative technique to remove background noise completely in optical domain.NIcMaster University . in ORF systems the communicating distance improves steadily as the number of relays. BER performance of ORF systems are analyzed. it requires complex relay structures and adjustable gain amplifiers that imposes n:lOre complexity and implementation costs to the system.
2 Future Work In this thesis. Conclusion and Future Work 5. some of which are described as follows .Electrical Engineering Chapter 5.McMaster University . • Although most FSO manufactures utilize the bit error rate (BER) as a standard performance metric to characterize their products. The ORF FSO system performance has been analyzed in the presence of background noise and it has been shown that in the absence of fading. Therefore. In order to utilize ORF technique to support longhaul applications. Therefore many future directions still remain to be investigated in the future.g. The intensity fluctuations induced by atmospheric turbulence. e. analyzing outage probability factor would be a critical step in performance analyses of alloptical multihop FSO systems . etc. lead to an endtoend outage in the whole system. blockage of the lineofsight link due to temporary obstructions. the endtoend outage occurs when a local outage happens during one of intermediate hops. the error performance of ORF FSO systems has been investigated when the atmospheric turbulenceinduced fading effects are neglected. This assumption holds for short hopdistances (less than 1 km) and under weak atmospheric turbulence conditions. In multihop systenls. ORF technique greatly extends the total communicating distance. the BER is not itself a comprehensive metric to evaluate the performance of FSO systems. temporary disalignmcnt between two consequent relays. • In Chapter 4. birds. The next work can be devoted on investigating the limitations that atmospheric turbulence ~ 130 . atmospheric turbulence effects must been taken into account. alloptical multihop FSO communication systems are studied and its BER performance is analyzed for the first time.
On the other hand. Conclusion and Future Work imposes to the maximum achievable distance of ORF FSO systems . 131 .Electrical Engineering Chapter 5. On the other hand.:McMaster University . • The performance of multihop FSO systems has been analyzed in the absence/presence of weak atmospheriC turbulenceinduced fading. since each path experiences a different atmospheric condition. the system performance can be analyzed for different amounts of collected background power at the receive apertures when the ambient illumination varies during the day and for different locations. it is expected that the endtoend outage probability decreases. The performance of FSO links is very dependent upon weather conditions therefore investigating the performance of multihop FSO systems under different turbulence conditions is another interesting approach for the future work. The information redundancy at the receiver of such systems can improve the error performance. • The cooperative diversity scheme is another future approach for increasing both the reliability and distance coverage of FSO systems.
26): It will be shown that f{ (x) IX= Li = 0 (A. 132 .2) f{'(x) > 0. (3.I Dualhop relaying system Recall the expression for JI(x) given in Eq. for 0 < x < LT.I 1 1 Appendix A Optimum relaying configuration A.
1 J{( L!) = 0 f{ (x) (A.Electrical Engineering Appendix A. Optimum relaying configuration A.McMaster University . it is clear that at x = L:J the first derivative of fl (x) is zero: f~ ( L:J) = o.I.3) =? f{(x) (AA) From (AA). A. From (A A) f{' (x) (A.2 J{'(L{) > 0 First it will be shown that the second derivative is nonnegative at x = L:J.5) 133 .I.
Optimum relaying configuration By substituting x = L{ in (A.McMaster University ..7) For higher SNRo.Electrical Engineering Appendix A.39.. f{'(LT/2) is nonnegative over longer distances. eD:LT/2 :s. The term c in (A.6) is the coefficient of the dominant factor (L~) and is described as c = (0:/2 . (A. On the other hand. HaLT H] . the maximum LT over which fr(L T/2) is nonnegative can be obtained as C = (. 7.43 dB/Inn which in the linear system is equivalent to 0: = 1O4 (m1). in other words eD:LT/2 < 10 for a long range of LT. eD:LT/2(L. K: ~ 10 5 and the average transmit SNRo varies between 0 dB and 40 dB in our simulations or equivalently 1 < SNRo < 104 in the linear system. 40 km.6) In the FSO systems which are considered in this work.5) f{' (~T) 20: [L. the exponential term changes between 1 :s. shorter distances and higher SNRo are considered in the simulations. therefore f{'(LT/2) > 0 and the proof is complete.6) are positive. (K:SNRo)l :s. it can be assumed that fr(L T /2) is nonnegative in the desired range of LT' 134 . in this work. Li. For o :s. 10. if c > 0.5 :s. Since. for the above mentioned range of SNRo. Since the two other terms in (A. the atmospheric attenuation coefficient is 0: = 0. At the worst case where SNRo = 0 dB ((K:SNR o)l = 105). 0: 2 (/l:SNRo)leaLT/2) > 0 SNRo=O > .(~ (~SN~o)l e  O £T/2).(K:SNRo)leD:LT/2). 109 . (A. In typical FSO systems.10 e 2 104 5 1O4xLT 2 > 0 =? LT < 35 km. LT :s.2(K:SNRo)leD:LT L2~ ~ eoVr / 2 [. + 2)] + 4 [eaLT/2(0:~T + 1)].
L T ] N ow it will be shown that f~' (x) > 0 is also nonnegative over the range of typical LT values where f~'(LT/2) is nonnegative. 1'1" and SNRo.9) From (A. LT .1.2(LT . the first two terms are nonnegative for every arbitrary value of 0 < x < LT. In (A.Electrical Engineering Appendix A.9).8) the first and second order derivatives of g( x) are easily obtained as g'(x) g"(x) ~~) .x)] . (A. g(x) > g(LT/2) or equivalently f{'(x) > f{'(LT/2) for every x inside the region (0 < x < LT)' Since. g(x) is convex over all possible values of x and x = L2T is a global minimum point for g(x). the range of LT over which f{'(LT/2) is nonnegative has been discussed.McMaster University . Consider the third term in (A. f{'(L T/2) is nonnegative for the desired range of LT . Optimum relaying configuration A. it will be shown that f{'(LT/2) is less than f{'(x) for every x =I L{ and 0 < x < LT . x=~ 2 L = 0 a~~x) = (A.1.3 ff'(x) > 0 for x E [0.5).2x) I 8 + 8 = 16 > O. In this section. That is.2x) . g(x) = [(LT . However the third term depends on the value of a.2.= ax 4(LT . therefore f~'(x) is also nonnegative for every 0 < x < LT . f~(x) is nonnegative over the region where f{'(L T /2) > O.2x(LT . In A. 135 . and therefore. and x= L2T is a global minimum point for h (x) over the desired range of LT.2x? .5).
Optimum relaying configuration A. For x = (p + l)XMl the first order derivative is zero 136 .Electrical Engineering Appendix A.) (p: S]'*' X)2 + "2(LT .SNRo)lect(LTX)(L T .SNRo)lect(LTx) (a(LT  Define XM = LT/p + 2..~.X)2] [1+ (KSNn.ll) x [(A.l\/IcMaster University . [(A.SNRo)lect(p~l) (~(_X_? + x _2_(_X_))] [1 + (A.x))] .2 (p + 1)hop relaying system From (3.36) ajp+1(x) ax = (p+ 1) [1 + (A.)'ca(.SNRo)lect(p~l) (_X_)2]P p+ 1 p+l p+1 p+1 p+1 x. (A.
(A.x? + 2(LT ..Electrical Engineering Appendix A. / [ect(LTX) [0: (o:(LT  J +l X)2 + 2(LT .x?] x [ect(p~l) (p+ 1) L:1 (0:(P:1)2+ 2(P:1)) o) + (p2: 1(p:1)+ P!l)]] [1+ ("SNll 'e"(.13) 137 .McMaster University .x)) + (20:(LT ..x?] ct( x ) [e pH [(Il:SNRo)lect(p~l) (~(_X_)2 + _2_(_X_))] x (0: 1 (p + 1) p+ 2 + p + 1 (p + 1 ) 2 [1+ ("SNIlo)l e"('~') (p: x x sr x) ] [(Il:SNRo)lect(LTX) (o:(L T .~') C: l)'r P x x p+1 p+1 p+1 p+1 [ect(LTX) (o:(L T .x))] ct( x ) [e p+l (0: p+1 ( X)2 p+1 2 x) + () ] p+1 p+1 + x [1+ ("SNlloj1c"(. the second order derivative is derived as j I. Optimum relaying configuration From (A..x))] [(Il:SNRo)lect(p~l) (~(_X_? + _2_(_X_))] + x [1 + (Il:SNRo)lect(p~l) (~ ~ \Jl T 1 ) 21 J.x) + 2)]] .x? + 2(LT .) (p: l)'r [1 + (Il:SNRo)lect(LTX)(L T . af~+l(X) fu = p [1 + (Il:SNRo)lect(p~l) (_X_)2jPl p+1 p+1 p+1 X x x p+1 p+1 [1 + (Il:SNRo)lect(LTX)(LT .11).
3. a new acceptable region for LT is obtained which is larger than what obtained in (A. a similar equation to (A. Optimum relaying configuration By substituting (p + l)xM into (A. it is evident that by increasing the number of relays at a given SNRo. 7).1. 1\1£ = p Here.Electrical Engineering Appendix A.McMaster University . Using the same approach expressed in Section A. equally spaced relaying provides maximum received SNR at point XM.6) is found j For p = 0. (A.6) and (A.6).14) results in (A. XM depends on the number of relays +1 ll p+2 i p+l (LT ) By comparing (A.15). it is possible to show that jp+ 1 (x) is also convex over all 0 < x < LT for typical values of system parameters and XM is the global optimum point over the acceptable region of LT' 138 . In this region.13).
G. K.communications link to a DAV. "Investigation in free space optical communication links between unmanned aerial vehicles 139 . [6] E. Wireless Infrared Communications. vol. Komine and M. LoPresti. kohldorfer. Leitgeb. Refai. 1. 50. pp. "Fundamental analysis for visiblelight communication systems using LED lights. and W.C." in Proceeding of the 24th Digital Avionics Systems Conference (DASCj. Muhammad. and P. [2] M. R. J. Nakagawa. 407410.G29.21 to l. l. 100107. Barry. S. Schmitt. [3] T. P. Zettl. H. "Channel modeling for terrestrial free space optical channels. 2001. Rehm. 1994. Muhammad. "Indoor wireless infrared channel characterization by measurements.I i Bibliography [1] J. Kavehrad. and E. Pakravan and M." IEEE Transaction on Vehicular Technology. Sluss. [4] S. 2005. pp. Harris." in Proceedings of the International Conference on Transparent Optical Networks. N. vol. vol. july 2005. "Alignment and tracking of a free space optical. R. Leitgeb. pp. February 2004. 50." IEEE Transaction on Consumer Electronics. H. J. [5] A. S. Kluwer Academic Publisher.
Arnon. July 2004. March 1989. J. A. [11] fSONA Optical \i\Tireless. PhD thesis. second ed. [9] A.com.). Giggenbach. M. 14. Available: http://www.proxim.M. Gagliardi and S. Thesis . 1995. [7] D.fsona. [15] A. Epple.A. Horwath. G." IEEE Transactions on Commu·nications." in The 16th 1ST JvIobile and ~ I Wireless Comm1mications Summit. Optical Communications. 13071315. Youssef. [10] R. 21. pp.. John Wiley and Sons Inc. [8] C.Sc.com.your name Bibliography lVlcMaster ." Journal of Optical Society America. 2007. [13] Proxim Wireless. vol. [Online].gigabeam. Capacity and Signaling for FreeSpace Optical Channels.C. Chen.com.com. "Impact of random pointing and tracking errors on the design of coherent and incoherent optical intersatellite communication links. pp. Available: http://www. July 2007.Electrical Engineering (UAVs). Available: http://www.mrv. 252260. C. 140 .ailable: http://www. and B. pp. [Online]. "Optical satellite downlinks to optical ground stations and highaltitude platforms. [12] MRV Optical Communication Systems™. [Online]. MclVlaster University. Polishuk and S. October 2009.. [Online]." in Proceeding of the 9th International Conference on Transparent Op tical Networks (152155. 37. Karp. [14] GigaBeam Corporation. vol. "Optimization of laser satellite communication system with an optical preamplifier. ed. Av.
S. Awan. Korevaar. third ed. vol.iec.Sc. M. 27. Available: http://www. May 2004. Nadeem. "Urban optical wireless communication networks: the main challenges and possible solutions. McArthur. 26192692. vol. Proakis. and E. 141 . 2001. "Weather effects on hybrid F80/RF communication link.com. Muhammad. [20] LightPointe™. [21] 1. 4214. 1. Kandus.ch. A. 44. [24] E. Kvicera. S..lightpointe. broadband. P. Agrawal. Fiber Optic Communication Systems.A.Electrical Engineering [16] G. vol. 16871697. April 2009. [22] International Electrotechnical Commission. 2002. "Lightpointe unveils performance enhanced Flightlite™G FSO wireless bridging solution. "Fading channels: informationtheoretic and communications aspects." in fSONA white paper. Biglieri. (2010)." Proceedings SPIE. J. J." IEEE JO'llrnal on Selected Areas in Communications. vol. 42. B. 2637. vol." IEEE Transaction on Information Theory. [Online]. 50. August 2002. V. Leitgeb. and 8. pp.M. "Comparison of laser beam propagation at 785 nm and 1550 nm in fog and haze for optical wireless communications. December 2009. pp.. E. Thesis . "Optical wireless: lowcost. [23] X. Arnon. 82S7. Kim. Zhu. ·Rockwell and G. Mecherle.your name Bibliography McMaster . February 2001." in http://www.. John Wiley and Sons Inc. [19] D. "Freespace optical communication through atmospheric turbulence channels. and G. pp. [18] F. Kedar and S. optical access." in IEEE Transactions on Communications. [17] D. pp. 1998. Shamai." IEEE Comm'llnications lvlagazine.
R. New York. 51. Thesis . [26] J. pp. 635~637. 2000. 2003. Uysal. 7.your name Bibliography McMaster . Chan. Karagiannidis. W. 1896~1906. August 2003. 2755~2759. [32] M." in IEEE International Conference on Comm'LLnications ICC. 22. [29] X. A WileyInterscience Publication." IEEE Journal on selected areas in commu509~516. [31] E. Goodman. 1233~ 1239." in IEEE Transactions on Wireless Communications. pp. November 2004. USA: John Wiley and Sons. Uysal. Lee and V. 8.M. vol.A. Jing. Zhu and J. 142 . "Multihop free space optical communications over strong turbulence channels. [28] M. Sandalidis. Zhu and J. "Part 1: Optical communication over the clear turbulent atmospheric channel using diversity. Statistical Optics. Osche. pp. pp. Optical Detection Theory for Laser Applications. H.Electrical Engineering [25] G. NY. vol. 5441~5449." IEEE Transactions on Communications. March nications. Kahn. S. "Markov chain model in maximumlikelihood sequence detection for freespace optical communication through atmospheric turbulence channels. and L. 51. yol. "Performance bounds for coded freespace optical communications through atmospheric turbulence channels. "Error rate performance of coded freespace optical links over strong turbulence channels. Sagias. June 2006. October 2004. Tsiftsis." IEEE Communications Letters. 122008. and N. "Relayassisted freespace optical communication. Navidpour. [27] T. vol. pp. 2002.Sc." IEEE Transactions on Communications. vol. Kahn. pp. Safari and M. [30] X. G.
M.A.Sc. Thesis  your name Bibliography
McMaster  Electrical Engineering
[33] K. Kiasaleh, "Turbocoded optical PPM COl11.HlUnication systen1.s," Journal of
Lightwave Technology, vol. 16, pp. 1826, January 1998.
[34] E. Forestieri, Optical communication theory and techniques. Springer, 2004. [35] 1. Djordjevic, B. Vasic, and M. Neifeld, "Power efficient LDPCcoded modulation
I
for freespace optical communication over the atmospheric
turbulenc~
channel,"
in Conference on Optical Fiber Communication and the National Fiber Optic
Engineers Conference (OFCjNFOEC), pp. 13, Nlarch 2007.
[36] B. Vasic, 1. Djordjevic, and R. Kostuk, "Lowdensity parity check codes and iter
ative decoding for longhaul optical communication systems," Journal of Light
wave Technology, vol. 21, pp. 438446, February 2003.
[37] A. Paulraj, R. Nabar, and D. Gore, Introduction to SpaceTime Wireless Com
munications. Cambridge University Press, 2003.
[38] E. Shin and V. Chan, "Optical communication over the turbulent atmospheric
channel using spatial diversity," in IEEE Global Telecommunications Conference
(GLOBECOM), vol. 3, pp. 20552060, March 2002.
[39] S. Hass and J. Shapiro, "Capacity of \Yireless optical communications," IEEE
Journal on selected areas in communications, vol. 21, pp. 13461357, October
2003. [40] J. Akella, M. Yuksel, and S. Kalyanaraman, "Error analysis of multihop free
space optical communication," in IEEE International Conference on Communi
cations ICC, vol. 3, pp. 17771781, May 2005.
143
M.A.Sc. Thesis  your name Bibliography
McMaster  Electrical Engineering
[41] M. Hasna and M.S. Alouini, "Outage probability of multihop transmission over Nakagami fading channels," IEEE Communications Letters, vol. 7, pp. 216218, May 2003. [42] J. Laneman and G. "Vornell, "Energyefficient antenna sharing and relaying for wireless networks," in IEEE Wireless Communications and Networking Confer
ence (WCNC), vol. 1, pp. 712, 2000.
[43] J. Laneman, D. Tse, and G. Wornell, "Cooperative diversity in wireless networks: Efficient protocols and outage behavior," IEEE Transactions on Information
Theory, vol. 50, pp. 30623080, December 2004.
[44] G. Karagiannidis, T. Tsiftsis, and H. Sandalidis, "Outage probability of relayed free space optical communication systems," Electronics Letters, vol. 42, pp. 994995, August 2006. [45] C. Datsikas, K. Peppas, N. Sagias, and G. Tombras, "Serial free space optical relaying communications over gammagamma atmospheric turbulence cliannels,"
IEEEjOSA Journal of Optical Communications and Networking, vol. 2, pp. 576586, August 2010. [46] G. Karagiannidis, A. Tsiftsis, and R. Mallik, "Bounds for multihop relayed communications in nakagamim fading," IEEE Transactions on Communications, vol. 54, pp. 1822, January 2006. [47] M. Hasna and M.S. Alouini, "A performance study of dualhop transmissions with fixed gain relays," IEEE Transactions on Wireless Communications, vol. 3, pp. 19631968, November 2004.
144
M.A.Sc. Thesis  your name Bibliography
McMaster  Electrical Engineering
[48] lV1. Karimi and M. NasiriKenari, "BER analysis of cooperative systems in frec
space optical networks," Journal of Lightwave Technology, vol. 27, pp. 56395647, December 2009.
[49] A. Acampora and S. Krishnamurthy, "A broadband wireless access network
based on meshconnected freespace optical links," IEEE Personal Communi
cations, vol. 6, pp. 10709916, October 1999.
[50] PMCSierra. [Online]. Available: http://www.pmcsierra.com. [51] U. Adar, T. Kuzniz, 1. Haber, and M. Tur, "An alloptical multihop (cascaded)
high bit rate wireless communication field trial," in Optical Fiber Communication
Conference and Exhibit (OFC), pp. 446448, March 2002.
[52] L. C. Andrews and R. L. Philips, Laser Beam Propagation through Random
Media.
2005.
SPIEThe International Society for Optical Engineering, second ed.,
[53] 1. Kim, B. McArthur, and E. Korevaar, "Comparison of laser beam propagation
at 785 nm and 1550 nm in fog and haze for optical wireless communications," in
Proceeding SPIE, vol. 4214, pp. 2627, February 2001.
[54] D. Bushuev and S. Arnon, "Analysis of the performance of a wireless optical
multiinput to multioutput communication system," Journal of Optical Society
of America A, vol. 23, pp. 17221730, July 2006.
[55] D. L. Fried, "Optical heterodyne detection of an atmospherically distorted signal
wave front," Proceeding of IEEE, vol. 55, January 1967.
145
lVLA. 146 .. The McGrawHill Compa nies. C. [61] Y. pp. Fundamentals of Photonics. vol.Sc. Mamyshev. Introduction to Fourier Optics. [58] V.your name Bibliography McMaster . Wave Propagation in a Turbulent Medium. M. Teich. Dikmelik and F. [59] H. Radiojizika. Inc." Institute of Atmospheric Optics. Agrawal. The McGrawHill Book Company. Optical Communication Receiver Design. Lukin and V. Pokasov. FreeSpace Optics: Enabling Optical Connec tivity in Today's Networks. Alexander. pp. 44. "Alloptical data regeneration based on selfphase modulation effect. "Phase fluctuations of an optical wave J . Thesis . August 2005. 2024 September 1998.. [63] P. Davidson. 1997. November 1973. vol. W. E. [64] G. Sams Publishing. Ghuman.. [57] V. which propagates in a turbulent atmosphere. 49464952. 1995. 475476. pp. [62] S. 1988. 17261729. 2002. P.." Applied Optics. vol. Nonlinear Fiber Optics. B. P. Goodman. "Fibercoupling efficiency for freespace op tical communication through atmosphere turbulence. Inc. 16. 1961. Tatraski.Electrical Engineering [56] B. S. second ed. [60] J. A." in 24th European Conference on Optical Comm'llnication. Academic Press.INC. Saleh and M. 1. John Wiley and Sons Inc. SPIEThe Interna tional Society for Optical Engineering. V. 1. second ed. 1991. Willebrand and B.. Translated from Izvestiya Vysshikh Uchebnykh Zavedenii.
Berger." in Optical Fiber Com munication Conference. "Accurate Qfactor estimation of optically amplified systems in the presence of waveform distortion. Thesis .your name Bibliography McMaster . pp. 20.Electrical Engineering [65] J. [66] S. vol. 1927. 147 . "Widely tunable. Tselikov. March 2003." Journal of Lightwave Technology. 1. 252253. F. Ilkov.Sc.M. and A. January 2002. Maruoka. pp. vol. Norimatsu and M. naiTow optical bandpass Gaussian filter using a silicon microactuator. King.A. D.
This action might not be possible to undo. Are you sure you want to continue?
Use one of your book credits to continue reading from where you left off, or restart the preview.