PFC AlbertoDiazSouto

Masters Thesis
Comparison of OFDM with Single

Carrier in high-data rate optical
communication systems
Alumne:
Alberto Diaz Souto
Barcelona, December 2011
Tutor/s:
Dr.Josep Prat i Dr.Antonio Napoli
Masters Thesis at
Enginyeria Tècnica Superior de Barcelona (ETSETB)
from Universitat Politcnica de Catalunya (UPC)
Titol : Comparison of OFDM with Single Carrier in high-data rate optical communication
systems
Author : Alberto Diaz Souto
Alberto Diaz Souto

Vilamur 4
08014 Barcelona
[email protected]
Contents
1. Introduction
11
1.1. Background and Motivation . . . . . . . . . . . . . . . . . . . . . . . . . .
11
1.2. Thesis Outline . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
14
2. Theory
15
2.1. Modulation Formats . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
15
2.1.1. Intensity Modulation Formats . . . . . . . . . . . . . . . . . . . . .
16
2.1.2. Phase Modulations . . . . . . . . . . . . . . . . . . . . . . . . . . .
17
2.1.3. Amplitude Modulations . . . . . . . . . . . . . . . . . . . . . . . .
19
2.2. Orthogonal Frequency Division Multiplexion vs. Single Carrier Transmission 20

2.2.1. Single Carrier . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
20
2.2.2. OFDM
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
20
2.2.3. Digital Signal Processing coherent receiver . . . . . . . . . . . . . .
22
2.2.4. Orthogonal Frequency Division Multiplexing and Single Carrier

comparison . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.3. Optical Link
24
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
26
2.3.1. Fiber Propagation . . . . . . . . . . . . . . . . . . . . . . . . . . .
28
2.3.2. Linear Impairments . . . . . . . . . . . . . . . . . . . . . . . . . .
28
2.3.3. Nonlinear Impairments . . . . . . . . . . . . . . . . . . . . . . . . .
31
3. Nonlinear mitigation of propagation effects
33
3.1. Radio-frequency Pilot . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
33
3.1.1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
33
3.1.2. Setup for Orthogonal Frequency Division Multiplexing . . . . . . .
35
3.1.3. Setup for Single Carrier . . . . . . . . . . . . . . . . . . . . . . . .
36
3.2. Back-propagation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
38
3.2.1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
38
3.2.2. Setup for Orthogonal Frequency Division Multiplexing and Single

Carrier
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
41
Contents
4. Dispersion-managed link
43
4.1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
43
4.2. Performance of Dispersion Management System . . . . . . . . . . . . . . .
43
4.3. Pulse Comparison of Single Carrier and OFDM in DM Systems . . . . . .
46
4.4. Nonlinear Mitigation for DM links . . . . . . . . . . . . . . . . . . . . . .
46
4.4.1. Back-propagation for DM systems . . . . . . . . . . . . . . . . . .
48
4.4.2. RFP for DM link . . . . . . . . . . . . . . . . . . . . . . . . . . . .
50
4.5. 9x224 Gb/s POLMUX-16QAM system with Nonlinear Mitigation . . . . .
51
5. Non-dispersion-managed link
5.1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
53
5.2. Perfomance of Non-Managed Dispersion Systems . . . . . . . . . . . . . .
53
5.3. Narrow Optical Filtering on 200-Gbit/s DP-16-QAM . . . . . . . . . . . .
54
5.4. Comparison performance of Single Carrier and OFDM in NDM systems .
55
5.5. Pulse shape comparison for SC with back-propagation . . . . . . . . . . .
56
5.6. Joint work of back-propagation and RF pilot-tone for SC and OFDM
59
6. Conclusion
A. List of Abbreviations
Bibliography
ii
53
. .
63
I
V
List of Figures
2.1. NRZ-OOK transmitter scheme . . . . . . . . . . . . . . . . . . . . . . . .
16
2.2. RZ-OOK transmitter scheme . . . . . . . . . . . . . . . . . . . . . . . . .
17
2.3. Spectrum comparison for 112-Gb/s PDM-QPSK between RZ50 and NRZ.
17
2.4. Setup implemented in a DBPSK transmitter . . . . . . . . . . . . . . . . .
18
2.5. Examples of QAM constellations . . . . . . . . . . . . . . . . . . . . . . .
19
2.6. OFDM time-domain transmission . . . . . . . . . . . . . . . . . . . . . . .
21
2.7. OFDM signal in time (left) and frequency (right) domain . . . . . . . . .
22
2.8. Differentially coherent phase detection of: (a) 2-DPSK and (b) M -DPSK
23
2.9. Example of a feedforward carrier-phase estimation . . . . . . . . . . . . .
24
2.10. Back to back performance for Single Carrier . . . . . . . . . . . . . . . . .
27
2.11. Back to back performance for OFDM . . . . . . . . . . . . . . . . . . . . .
27
2.12. Scheme of an Optical Communication System . . . . . . . . . . . . . . . .
28
2.13. Example of predicted information spectral density limits per polarisation

for non-linear transmission . . . . . . . . . . . . . . . . . . . . . . . . . . .
29
3.1. Spectrum of a transmitted and received OFDM signal with the RF-pilot
at: (a) Transmitter and (b) Zoom of (a) [35] . . . . . . . . . . . . . . . . .
35
3.2. OFDM system applying RFP technique . . . . . . . . . . . . . . . . . . .
36
3.3. PSR and RFP bandwidth optimization in a 16-QAM modulation system

when RFP and BP are implemented . . . . . . . . . . . . . . . . . . . . .
37
3.4. SC spectrum after RFP addition . . . . . . . . . . . . . . . . . . . . . . .
37
3.5. SC system applying RFP technique . . . . . . . . . . . . . . . . . . . . . .
38
3.6. Forward and Backward Propagation concept . . . . . . . . . . . . . . . . .
39
3.7. Block diagram of the BP algorithm . . . . . . . . . . . . . . . . . . . . . .
40
3.8. Comparison between different Step Modes . . . . . . . . . . . . . . . . . .
41
3.9. Receiver DSP block diagram for SC and OFDM case . . . . . . . . . . . .
42
4.1. Scheme of a Dispersion-Managed scenario . . . . . . . . . . . . . . . . . .
44
4.2. Different dispersion compensation schemes for optical communications systems [21] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
45
iii
List of Figures
4.3. Maximum reach distance versus launch power for 9 WDM channels through
a DM link [42] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
47
4.4. Full dispersion compensation map . . . . . . . . . . . . . . . . . . . . . .
48
4.5. BP computational complexity reduction demonstration. Comparison between applying BP with 15 and 1 steps
. . . . . . . . . . . . . . . . . . .
49
4.6. Dispersion compensation map deployed in our simulations . . . . . . . . .
50
4.7. Maximum reach distance versus launch power for 9 WDM channels transmission in a DM links for: (a) Single Carrier and (b) OFDM . . . . . . .
52
5.1. Maximum reach distance versus launch power for 9 WDM channels transmission for different pulse shapes . . . . . . . . . . . . . . . . . . . . . . .
56
5.2. Maximum reach distance versus launch power for 9 WDM channels transmission for a dispersion unmanaged link . . . . . . . . . . . . . . . . . . .
57
5.3. Maximum reach distance versus launch power for a NRZ and RZ50 Single
carrier transmission of 9 WDM channels . . . . . . . . . . . . . . . . . . .
58
5.4. Receiver DSP block diagram for nonlinear mitigation . . . . . . . . . . . .
59
5.5. Maximum reach distance versus launch power for 9 WDM channels POLMUX 16-QAM transmission with nonlinear mitigation in a NDM scenario
iv
61
List of Tables
3.1. Fiber Link paramerters [34] . . . . . . . . . . . . . . . . . . . . . . . . . .
42
4.1. Typical parameters of a Dispersion Compensation Fiber [43] . . . . . . . .
45
Acknowledgements
First and foremost, I would like to express my utmost gratitude to all those people who
have supported me along this nine months making this master thesis possible.
The first person I would like to mention and to be especially thankful is Dr.Antonio
Napoli. His support within and outside work time made me feel like home and letting me
give my best doing this work. Furthermore, his constant support and all the theoretical
support have been really remarkable. In addition, I would like to mention Dr.Maxim
Kuschnerov because without his professional support all the simulations done in this thesis have never could be afforded. I wouldnt like to forget the professor of my university,
Dr. Josep Prat, because without his help and motivation one year ago I wouldnt have
this great opportunity at Nokia Siemens Networks which has been one of the most important opportunities in my recent life. I am heartly thankful to all my colleagues at
Nokia Siemens. During all that time they have been more than only colleagues, they can
be considered like friends due to all the nice time we spent together.
Last but not least, I am forever indebted to all my family and friends for their constant
support throughout not only my thesis, throughout the entire duration of my degree.
Their words of encouragement never failed to keep me reaching my goal even through the
hardest of times and it is here that I express my sincerest gratitude to them.
Munich, 16 December 2011.
Alberto Diaz Souto
Collaborations
This master thesis can not be realized without the support of my university, Telecom
BCN and Nokia Siemens Networks, the company that have given me all the technical
support to do my investigations and simulation analyses.
Resum del projecte

En aquesta tesi, el rendiment de single carrier i OFDM sobre diferents situacions sha
investigat. Adicionalment, dos tècniques de compensacions de no-linearitats com backpropagation i RF pilot-tone shan implementat als nostres anàlisi per tal de veure el
diferent rendiment que aquests mètodes poden aportar al rendiment de la senyal. Es
pot veure que la mateixa millora sha assolit tant per SC com per OFDM, arribant a un
augment en la m`
axima distancia assolida daproximadament 18% i 13% pel cas de SC i
OFDM respectivament quan els mètodes de backpropagation i RF pilot-tone sapliquen
junts. Els resultats conclueixen que no hi ha diferències en la mitigacio de no-linearitats
en qualsevol tipus de format de modulacio.
Principalment shan considerat dos situacions en les nostres simulacions. Aquestes dos
situacions s
on: dispersion managed link i non-dispersion link. El primer es va considerar
degut al fet de que est`
a extensament implementat en les actuals comunicacions òptiques
i el segon sha considerat perque es supossa que serà el futur mes proper en els sistemes
de comunicacions `
optiques. Els resultats quan vem implementar el cas de non-dispersion
link van ser qualificat com els millors i mes interessants que es van trobar. La rao es
lalta toler`
ancia a les no-linearitats que es pot trobar en aquesta classe de situacions.
Despres de tots els diferents an`
alisis, els resultats han mostrat millor rendiment pel cas
de single carrier tot i incurrint en una pitjor eficiència espectral. Amb l
us conjunt de
backpropagation i RF pilot-tone, OFDM pot assolir el mateix rendiment que single carrier
sense cap tipus de compensaci
o de no-linearitats, el qual significa que la diferència no es
podria considerar com un problema difcil al qual poder enfrentar-shi.
Resumen del proyecto

En este proyecto, el rendimiento de single carrier y OFDM sobre diferentes escenarios
se ha llevado a cabo. Adem
as, dos tecnicas de compensacion de no-linearidades como
backpropagation y RF pilot-tone se han implementado en nuestros analisis con el objetivo
de ver las diferentes mejoras que pueden aportar en el rendimiento de la se
nal. Se puede
ver que la misma mejora se alcanza tanto para SC y OFDM llegando a un aumento en la
maxima distancia alcanzada de aproximadamente 18% y 13% para el caso de SC y OFDM
respectivamente cuando backpropagation y RF pilot-tone son aplicadas conjuntamente.
Los resultados concluyen que no se puede ver una diferencia en la compensacion de nolinearidades para cualquier tipo de formato de modulacion.
Principalmente dos escenarios se han considerado en todas nuestras simulaciones. Estos
dos escenarios se componen de los llamados dispersion managed link y non-dispersion
link. El primero de ellos se consider
o debido a su extensa implementacion en las actuales
comunicaciones
opticas y el segundo es considerado porque se supone que sera el futuro
mas proximo de los sistemas de comunicaciones opticos. Los mejores y mas interesantes
resultados que se encontraron fueron cuando se considero el caso del non-dispersion link.
La razon reside en la alta tolerancia a las no-linearidades que se puede encontrar en este
tipo de escenarios.
Despues de todos nuestros analisis, los resultados han mostrado un mejor rendimiento
siempre y cuando single carrier se transmita incurriendo en una peor eficiencia espectral.
Con el uso conjunto de las tecnicas de backpropagation y RF pilot-tone OFDM puede
alcanzar el mismo rendimiento que single carrier sin cualquier tecnica de mitigacion de
no-linearidades, lo cual significa que la diferencia no es un aspecto inalcanzable con el
que poder lidiar.
Abstract
In this thesis work, the performance of single carrier and OFDM along different scenarios
is investigated. In addition, two nonlinear compensation techniques like backpropagation and RF pilot-tone were implemented in our analyses in order to see the different
improvements that these methods can bring to the signal performance. It is shown that
the same improvement can be achieved for SC and OFDM which can lead to an improvement at the maximum reach distance of approximatelly 18% and 13% for the SC and
OFDM case respectively when backpropagation and RF pilot-tone technique are applied
together. This results conclude that no difference in the nonlinear mitigation can be seen
between any modulation format.
Two main scenarios where considered in all our simulations. This two scenarios are dispersion managed link and non-dispersion link. The first one was considered due to the
fact that is widely deployed in actual optical communications and the second is considered because is supposed to be the near future in optical communication systems. Better
and interesting results were found to the case were non-dispersion link was considered.
The reason is the high nonlinear tolerance that can be found in this sort of scenarios.
After all the different simulation analyses, the results show better performance where
a single carrier is transmitted incurring in a worst spectral efficiency. With the joint
usage of backpropagation and RF pilot-tone technique OFDM can afford the same performance as single carrier without any nonlinear mitigation technique, which means that
the difference is not a difficult issue to deal with.
1. Introduction
1.1. Background and Motivation
In modern technological history, different facts have pushed the telecommunication researches toward a new technologies in order to face up with the recent exponential growth
in the required bandwidth, considering them to be the actual and future deployments in
different telecommunication scenarios.
During the last decade, the telecommunication bubble in conjunction with the modern culture have increased the number of mobile devices (i.e: mobile phones, laptops,
netbooks, tablet PC,etc) pointing Internet services as one of the main reason of this increasement. As it is already known, people have the necessity to be connected every
time and everywhere to different applications as Facebook [1], Twitter [2], Youtube [3],
and thus, a high-speed transmission systems are required if these services want to be covered. In this situation, optical fiber communications has emerged as a novel link to attain
high data rates reaching the best relation between quality and maximum reach distance
[4]. Actually, the most deployed links to provide all these services are the ones based on
copper, either co-axial or twisted pair. Moreover, the preferred transmission format to
increase bit rates over this infrastructure is digital subscriber line (DSL) technology. On
the other hand, as it was aforementioned, optical fibers has strongly emerged as an alternative to all the implemented copper networks due to its much higher bandwith-distance
product, enabling high-speed connections over longer distances.
The first generation of optical communications, considered at the beginning of this century, was the transmission systems where 10-Gb/s bit rates were used due to the low cost
basis. Since then, capacities have steadily increased and the cost per transmitted bit has
gradually decreased. Several solutions have been proposed in order to attain this high
capacity at a high transmission data rates. One remarkable solution, founded in 1990,
was the idea to transmit more than one channel over the transmission link, defined as
wavelength division multiplexing (WDM) architecture. This technology enables to raise
the amount of data traffic by multiplexing a number of lower capacity wavelength channels onto a single fiber. It is interesting to mention the importance that WDM systems
have over long-haul optical transmissions because a remarkable decrease in the cost of
11
1. Introduction
the transmitted bandwidth can be achieved.
Additional to this solution, other solutions have been proposed such as modulation formats and also at the receiver. At the beginning of this telecommunication bubble, coherent detection was introduced at the receiver in order to enhance the transmission distance
[5]. The problem of coherent detection was that during 20 years was kept in stand-by
because of the more efficient solution provided by direct-detection together with optical amplification at that time. After this period of time, coherent receiver has recently
re-gained the interest in actual researches due to the digital signal processing (DSP)
techniques. In a coherent receiver, not only the amplitude of the optical signal, but the
full baseband optical field is transferred to the electronic domain. DSP implementation
provides an improvement in singals robustness and increases the flexibility of the entire
optical signal. Furthermore, it enables the equalization of nearly high amounts of chromatic dispersion and PMD. The new interest in coherent receiver in conjunction with
DSP has pushed up the usage of advance modulation formats to realize high spectrally
efficient transmission [6].
Multi-level modulation formats have attracted the interest in the last years because of
the high spectral efficiency. The most important modulation format used in researches
today is the so called POLMUX-16-QAM. When POLMUX is mentioned, we refer to
polarization-multiplexed. This transmission format has become attractive as a result of
the independent data modulation transmitted into each of the two orthogonal polarizations of an optical fiber. The main advantage from implementing a POLMUX transmission is doubling the spectral efficiency and halving the symbol rate in comparison to
single-polarization modulation. Apart from the increasing interest in multi-level modulation formats, it is important to mention that there is raising interest in different nonlinear
mitigation techniques due to the feasibility brought by the DSP-based receiver. In the
last decade, several techniques have come out to deal with nonlinear impairments, such as
self-phase modulation (SPM) and cross-phase modulation (XPM), which are considered
the most limiting factors for a WDM long-haul optical system. Back-propagation (BP)
and RF-pilot-based tone (RFP) have been demonstrated to be specially effective for orthogonal frequency division multiplexing (OFDM) [17], and recently its effectiveness has
been studied for the case of single carrier.
As long as the spectral effiency becomes the most considered target in optical communication systems, OFDM has exponentially grown as a central point in actual and future
researches because of its better spectral efficiency. OFDM is also interesting because of
its extensive use in broadband wired and wireless communication systems. Thus, the
idea that OFDM could afford the same system performance in optical communications
systems is supposed but needs to be analyzed. Although OFDM has gained popularity
12
1.1. Background and Motivation

in the past decade and has widely been implemented in numerous communication standards, the problem that OFDM has to becompared with SC transmission (which is the
technique deployed in optical fiber systems) is still remaining. Few investigations have
been focused on the debate regarding wheter OFDM or single-carrier applying frequency
domain equalizer (SCFDE) is superior. After all those investigations it may be premature
to conclude which of both modulation formats can afford the best system performance,
because an extensive study between the advantages and disadvantages versus the scenario
where OFDM and SCFDE are performed is required.
The goal of the thesis presented here is to give an extense and comprehensive comparison between the two aforementioned modulation formats. No clear conclusions have
been reported in the last investigations of this topic in the last years not clarifying if the
advantages given by OFDM can overcome its own disadvantages. Moreover, considerable simulation analyses have been done in order to study all possible scenarios where
the comparison of SC and OFDM can be interesting. Due to the nonlinear influence in
long-haul optical communications, two novel nonlinear mitigation techniques such as BP
and RFP were applied in order to give a better performance and quantify the achievable
improvements of these compensation techniques.
In this work, it is shown that BP and RFP can give a significant improvement in signal
performance, well compensating for SPM and XPM respectively. Only simulated data
was considered in order to analyze the comparison of both modulation formats. Furthermore, narrow optical filtering, different pulse-shape transmission, and the different
effects by applying the two nonlinear compensation techniques alone were considered. As
an outcome of this work, several publications appeared in conference proceedings and
others have been submitted to leading conferences.
A.Lobato, M.Kuschnerov, A.Diaz, A.Napoli, Performance comparison of single

carrier and OFDM in coherent optical long-haul communication systems, in 2011
Asian Photonics Conference
A.Lobato, M.Kuschnerov, A.Diaz, A.Napoli, Impact of Narrow Optical Filtering
on 200Gb/s DP-16-QAM Long-Haul Optical Transmission, in 2011 IEEE Photonics Society Summer Topicals Meetings.
A.Diaz et al., Comparison of Single Carrier and OFDM with ack-propagation and
RF Pilot Tone for a 9x224 Gbs POLMUX-16QAM System, in IEEE Photonics,
2011.
13
1. Introduction
A.Diaz et al., Analysis of back-propagation and RF pilot-tone based nonlinearity
compensation for a 9x224Gb/s POLMUX- 16QAM system, in OFC, 2012.
1.2. Thesis Outline

In this section a general overview of the thesis structure is given. The first part of the
thesis consists of the Chapter 2 which review the basic concepts used in optical communications, taking a look at the transmitter, link, and receiver part of an optical system.
At the beginning of Chapter 3 the basics of nonlinear mitigation techniques BP and RFP
are explained. In this chapter the different simulation setups deployed in order to apply
the two techniques are displayed. All the simulations done during this comprehensive
comparison have been reported to the dispersion managed and non-dispersion managed
scenarios. In Chapter 4 and 5 all the analyses referred to the dispersion managed links
and non-dispersion managed links are respectively deployed. Finally, Chapter 6 summarizes the most important conclusions found in this thesis and some future outlines are
described if this work can be the basis of future investigations.
14
2. Theory
2.1. Modulation Formats
The most important aim of optical fiber communications is the interest to attain long
reach distances and high transport capacities without incurring higher implementation
costs per transmitted information bit while maintining a high spectral efficiency. Among
other enabling technologies, advanced optical modulation formats is playing a fundamental role in the development and design of modern wavelength division multiplexed
(WDM) fiber systems.
In a WDM scenario is important to achieve remarkable cost reduction. Sharing the optical components among the WDM channels and increasing the per-channel data rates are
two possible ways to accomplish this challenge. In addition to the low implementation
cost, high spectral efficiency and capacity are also required and can be applied by several
technologies [8].
Advance modulation formats1 are important because of their noise resilience, fiber propagation characteristics and resilience to narrowband optical filtering due to a casacde of
optical add/drop multiplexers (OADMs).
The different modulation formats are classified depending on how the information is
transported. Therefore, we distinguish between intensity, phase, and amplitude formats.
In this section, we limit our attention to only the different modulation formats used in
our simulations analysis.
Intensity
OOK Nonretun-to-Zero (NRZ)
OOK Return-to Zero (RZ)
Phase
Binary Phase-Shift Keying (BPSK) and Differential-BPSK (DPSK)
Quaternary Phase-Shift Keying (QPSK) and Differential-QPSK (DQPSK)
1
in optical communications, advanced are all the formats that go beyond on-off-keying (OOK)
15
2. Theory
Amplitude
Quaternary Amplitude Modulation (QAM)
2.1.1. Intensity Modulation Formats

OOK Nonreturn-to Zero
For quite a long time, OOK has been the most used modulation format for optical fiber
communication systems. NRZ gained a momentum based in the low electrical bandwidth
demanding for transmitters/receivers and, the resistance to LASER phase noise.
Fig. 2.1 shows the transmitter scheme implemented when NRZ-OOK is used. The
intensity modulator (in our system a Mach-Zender Modulator (MZM)) converts the OOK
electric signal into an optical one at the same input data rate. The main advantge that
NRZ provides is a narrower spectrum compared to other modulation formats such as
RZ. Furthermore, its performance was found to be degraded more than other modulation
formats in long-haul optical communications because of less resistance to fiber nonlinear
impairments is afford [25],[26]. As a result, RZ has been deeply investigated in order to
compare its advantages and performance against the NRZ format.
Figure 2.1.: NRZ-OOK transmitter scheme
Return-to Zero
RZ-OOK transmitters (see Fig. 2.2) can be easily implemented for high-data rates by
carving pulses out of an NRZ signal using an additional modulator, called pulse carver.
Pulse carvers can be deployed by electroabsorption modulators (EAMs) or MZMs. Due
to the variable absorption characteristics and residual chirp of EAMs, advanced RZ modulation formats are typically implemented using MZM-based pulse carvers. Depending
on where a MZM is sinusoidally driven, different RZ cases can be deployed. This different
16
Figure 2.2.: RZ-OOK transmitter scheme

RZ cases are: 50%, 33%, and 66%-duty-cycle [8].
RZ spectrum due to the narrow symbol duration is broader than NRZ. As a consequence,
RZ formats are more robust to nonlinear propagation distorsions but less spectrally efficient. In [9], an extensive comparison between NRZ and RZ50 was carried out, concluding that RZ50 is able to improve in WDM-QPSK transmission experiment at an optimal
launch power, 24% the NRZ performance. The main reason to explain this improvement
is the advantage that a narrow optical filtering at the transmitter gives to RZ modulation
format. After this narrower filtering, RZ pulse is converted into a high quality NRZ signal
with less intersymbol interference compared with a conventional NRZ (see Fig. 2.3).
Figure 2.3.: Spectrum comparison for 112-Gb/s PDM-QPSK between RZ50 and NRZ.
2.1.2. Phase Modulations

Differential Binary Phase-Shift Keying
Differential binary PSK encodes information on the binary phase change between adjacent bits: 1-bits are encoded onto a phase change, whereas 0-bits are represented
without changing the previous phase value.
Fig. 2.4 depicts the optical DPSK transmitter. The data signal is differentially encoded
at the transmitter because a differential encoding at the receiver can incurre in error
17
2. Theory
Figure 2.4.: Setup implemented in a DBPSK transmitter

propagation. In addition, the phase of the optical field of a narrow-linewidth laser source
is then flipped between 0 and using the precoded data sequence. Another important
element of the transmitter is the modulator, where usually a MZM is employed, since
using this sort of modulators allows an exact phase modulation.
The most restrictive aspect at the receiver is the interference between bits. For that
reason, a good avoidance to interferences is required because two optical fields can interfere destructively to each other whenever there is no phase change, and constructively
whenever a phase change between subsequent bits.
Differential Quaternary Phase-Shift Keying
When we refer to phase modulation formats, we also find multilevel formats. DQPSK
is the phase multilevel format that has attracted more attention in optical modulation
formats. It consists in the addition of two more phase shifts at the DBPSK ones, which
are /2 and /2.
At the transmitter, two MZMs are required to operate as phase modulators. A LASER,
before a splitter, feeds the two MZMs with equal intensity. The next stage is composed by
an optical /2-phase shifter that allow one of the paths to be shifted and then combined
with the other path to produce a single output signal. The main advantage of this
transmitter is that only binary electronic drive signals are required, giving an easier way
to generate high speeds rather than the multilevel drive waveforms.
Is also important to mention the shape of a DQPSK spectrum. In the optical domain,
the shape of the DQPSK is nearly the same as the one that DBPSK has, but the DQPSK
spectrum is compressed in frequency due to the halved symbol rate. In conclusion, this
compressed spectrum became a beneficial tool to achieve high spectral efficiencies in
18

WDM systems, increased tolerance to chromatic dispersion (CD), and a robustness to
polarization mode dispersion (PMD).
2.1.3. Amplitude Modulations

Nowadays amplitude modulations formats are mainly focused on quadrature amplitude
modulation (QAM) and multilevel modulation formats [27]. In the QAM case half the
transmitted symbols are modulated with one frequency and the other half symbols are
modulated with the same frequency but with a /2 phase change. At the end the sum
of both components will result in a QAM signal. As the definition noted, the modulated
symbols with the pilot without phase would be defined as the in-phase component and
the other phase-shifted component is called quadrature-phase component. Fig. 2.5 shows
the different QAM constellations mostly used in communications.
In our simulation analyses, the 16-QAM format will be the chosen QAM transmitted
(a) 4-QAM
(b) 16-QAM
(c) 64-QAM
Figure 2.5.: Examples of QAM constellations
modulation. The reason why we chose this QAM format is because M-ary QAM constellations enables data to reach higher data rates with a considerable spectral efficiency
[11]. For example, the error-rate performance of 8-QAM is closer to the 16-QAM (only
0.5 dB better), but the data-rate is only 3/4 that of 16-QAM. Morevover, the data signal
spectrum becomes much more narrower as the M is increased. On the other hand, symbols are subjected to more errors due to noise and interference. Thus, one transmitted
symbol is generated closer to neighbouring symbols which means that any interference
could induce more errors as soon as the number of transmitted symbols are increased.
Is important to note that the usage of multilevel modulation formats increases the complexity in hardware implementation
We carry out the analysis with 16-QAM because of the aforementioned advantages and
because of its emergence as a promising candidate to attain high-data rates for long-haul
networks [12].
19
2. Theory
2.2. Orthogonal Frequency Division Multiplexion vs.

Single Carrier Transmission
Optical modulation techniques can be divided in two main groups: single-carrier (SC)
modulation and multicarrier transmission. During the last years, optical frequency domain modulation (OFDM) modulation format has raised its importance in the optical
field due to its exceptional quality to send information at high data rates. For this reason, the idea of giving a brief theoretical introduction of SC and OFDM establishing their
advantages and disadavantages for a future comparison was supposed to be interesting.
2.2.1. Single Carrier

Single carrier (SC) technique has been widely employed in optical communication systems during the last three decades. During the 70s, many researches [10] focused their
investigations in the fact that frequency domain processing techniques could also be implemented as a simplified equalization for SC systems. Additionally, the SC implementation
implicity/performance was found to be nearly the same as the one used for OFDM systems (only when FDE is considered).
One main characteristic of this systems, that becomes an advantage for SC, is its low
measured peak-to-average ratio (PAPR) compared to the one presented in multicarrier
systems. As a result, the SC transmitted spectrum and its forward performance are not
severaly affected by the nonlinear effects. Furthermore, this kind of systems are also
important because of its robustness to frequency offset and phase noise.
Single carrier with frequency domain equalization (SC-FDE) has been chosen as the SC
format for our simulation analysis. The potential shown by SC is one of the points to
discuss in order to find if SC-FDE implementation is still being a real alternative to
OFDM.
2.2.2. OFDM
Nowadays OFDM is extensively used in broadband wired and wireless communication
systems because is an effective solution to intersymbol interference (ISI) caused by a dispersive channel.
Before OFDM, several techniques were utilized to transmit information over the frequency
domain. Frequency Division Multiplexing (FDM) was the technique where the main signal is transmitted over a set of independent signals in the frequency domain, which are
called subcarriers or tones [28]. It can be seen that if we use different subcarriers to
transmit the entire data, we should divide the original data into many parallel streams,
20
2.2. Orthogonal Frequency Division Multiplexion vs. Single Carrier Transmission

one for each subcarrier. After that, each subcarrier is modulated with a conventional
modulation scheme and combined together to form the FDM signal.
This can be understood as the theoretical beginning of the OFDM technique which will
be discused in this chapter. If the usage of large number of different subcarriers is desired,
the necessity of a wide spectral space in the frequency domain is required, so that means
being less spectral efficient. Thus OFDM appears as a solution to solve the mentioned
problem. The idea was mainly supported by the concept of overlapping the different
subcarriers with an orthogally condition between them, resulting in a more efficiency
implementation than the FDM one.
While many details of OFDM are very complex, the basic concept of OFDM as we pointed
before is quite simple. Data is transmitted in parallel on a number of different frequencies
as is shown in Fig. 2.6. The different subcarrier frequencies are chosen in order to induce
Figure 2.6.: OFDM time-domain transmission

a mathematically orthogonal property over one OFDM symbol period. Both modulation
and multiplexing functions are digitally achieved using an inverse fast fourier transform
(IFFT) and therefore, the orthogonal signals can be generated precisely and in a very
computational efficient way. Each subcarrier is centered at fk and separated by 1/Ts from
its neighbours and because of its orthogonal property, the spectra of individual subcarriers will idealy appear without interference and without the need for analog filtering to
separate the received subcarriers [29]. In Fig. 2.7, the spectrum of one OFDM subcarrier
21
2. Theory
Figure 2.7.: OFDM signal in time (left) and frequency (right) domain
has a sinc form to better achieve the orthogonality between channels. A consequence of
using this spectrum shapes is the sidelobes found along the total frequency range including many other subcarriers. This can be pointed as the main cause of one of the most
important OFDM disadvantages: the sensitivity to frequency offset and phase noise [16].
At the end, we can note two important advantages of OFDM: the complexity of OFDM
and of systems using serial modulation and frequency domain equalization, scale well as
data rates and dispersion increase. The other important advantage is that this technique
transfers the complexity of the transmitters and receivers from the analog to the digital
domain. These advantages and other OFDM disadvantages can be discussed over this
chapter.
2.2.3. Digital Signal Processing coherent receiver

Over the last years coherent optical receivers have gained considerable interest, becoming
nowadays the most promising solution for future high-speed optical networks in optical
communications. It is also important to mention the different past technologies implemented at the receiver to detect the signal. The most important considered technologies
were:
Noncoherent detection
Differentially coherent detection
Coherent detection
Noncoherent Detection
In noncoherent detection, the receiver computes decision variables based on a measurement of signal energy. This type of detection is usually deployed for on-off-keying (OFF),
22
Figure 2.8.: Differentially coherent phase detection of: (a) 2-DPSK and (b) M -DPSK
multi-level amplitude-shift keying (ASK) and frequency-shift keying (FSK) [13].
However, there are some important limitations: (a) detection based on energy measurement allows signals to encode only one degree of freedom (DOF) per polarization per
carrier, reducing spectral efficiency and power efficiency, and (b) the loss of phase information during detection.
Differentially Coherent Detection
The main characteristic that define this second technique is that the receiver computes
decision variables based on a mesurement of differential phase between the symbol of interest and one or more reference symbol(s). In Fig. 2.8 the diagram of the idea explained
in this paragraph for the DPSK case is shown. For example, if we compare binary differential phase-shift keying (DPSK) using differentially coherent detection and noncoherent
OOK, at a bit error rate (BER) of 102 , we will obtain a 2.8 dB of higher sensitivity in
the case of DPSK.
Coherent Detection
The last and the most advanced technique explained in this section is called coherent
detection. In this case, the receiver computes decision variables based on the recovery of
the full electric field (amplitude and phase information).
Recently, the exponential growth of the interest in advanced modulation formats has
increased the complexity requirement in coherent receivers. Traditionally, coherent receivers equipped with complicated PLLs (optical or electrical) fulfilled the purpose of
knowledging the carrier phase information in order to demodulate the received signal
[13]. But recent researches in analog-to-digital conversion (ADC) technology have made
possible to use digital signal processing (DSP) techniques to demodulate high-speed optically modulated signals.
The main challenge in coherent detection is locking the phase of the LO to the transmitter laser phase. This synchronization could be obtained using the different sort of PLL
23
2. Theory
mentioned before. That scenario had an important limitation: if a phase synchronization was desired, which was the high-speed nature of optically modulated signal, that
speed will restricts the PLL. Its inherent feedback must respond very fast to any signal
change, and for these reason, DSP has emerged as an interesting method to apply for
phase synchronization because of the only phase tracking required (i.e Fig. 2.9), which
can be implemented by a feedforward technique[15].
Figure 2.9.: Example of a feedforward carrier-phase estimation

Optical coherent receivers in conjunction with DSP has emerged as a promising technique
enabling to attain high-information-rate in optical transmission systems. DSP appeared
as the substitute to the PLL which was one of the barriers to use coherent receivers in
the optical domain. With this deployment, several optical impariment mitigations can
be addressed in the digital domain allowing a great flexibility in any optical network.
2.2.4. Orthogonal Frequency Division Multiplexing and Single Carrier

comparison
The discussion between OFDM and SC in optical communications has become an important challenge in research. In the last section we have introduced both modulation
formats in order to give a better idea of the most important theoretical concepts. After the introduction, a short comparison looking at the advantages and disadvantages is
given in this section in order to establish a theoretical background and a basis for our
numerical analysis.
Main advantages and disadvantages
After the last sections, both OFDM and SC advantages and disadvantages have been
pointed but not further discussed. In the case of OFDM [16], the most important disadvantages are the high PAPR and the sensitivity to phase noise and frequency offset.
Despite of OFDM disadvantages, also SC has its own limitations.
First of all, the implementation cost in the optical field appears as an important drawback
for OFDM systems. One way to define this cost implementation is calculating the cost
24

complexity that would have an OFDM system. In [17], the complexity can be calculated
as Eq. 2.1 where Cbit is the number of subcarriers in OFDM or the number of points
used in DFT-IDFT for each SCFDE block, D is the accumulated CD per channel, B is
the baud rate, and is the proportional constant. But this is not a complete disadvantatge due to the nearly complexity that SC-FDE contributes, making both techniques
interesting for systems with high-data rate.
Cbit log(Nsc ), Nsc = DB
(2.1)
Looking that the cost implementation, for both systems are nearly the same. Another
critical aspect that should be compared for optical communications is the effect of nonlinearities. In long-haul single mode optical fiber, the PAPR is a fundamental disadvantatge for OFDM, but this difference is only significant in systems where a dispersion
compensation map is applied. In any case, this comparison is highly dependent on the
studied scenario, for this reason our simulation analysis may result in a useful tool to
decide the influence of non-linearities in each system.
Finally, the last important characteristics to be mentioned are the spectral efficiency
in SC-FDE, the sensitivity to frequency offset and phase noise for OFDM. One main
advantage of OFDM is the spectral efficiency that can offer to any system. The spectrum
of SC-FDE is broader and the complexity in electrical and optical filters are increased as
a result of the larger spectrum required to enclose the entire signal. As it was explained
in the OFDM concepts, the main advantage of this modulation format is the ortogonality
between subcarriers. But then, a strong accuracy in the phase of the receiver local oscillator is required because any difference compared to the carrier of the received signal can
degrade system performance. Several researches have appeared with a novel technique to
reduce this effect [18].
In summary, every comparison could result in an equal comparison because there is
no strong feature that can definetely conclude which modulation is the best one. Due to
the abstractness of each comparison, an additional issue should be carry out, focusing on
numerical and analitical terms. Furthermore, the analyses are strongly dependent on the
scenario studied and thus, is important to note that any conclusion given in this thesis
is based on our researches.
25
2. Theory
Back-to-back simulations for both systems for different data-rate and
modulation formats
The aim of this thesis is the comparison between SC and OFDM in order to figure out
which of both modulation formats provide the best performance. One way to support our
conclusions is basing them on the simulations analyses that are going to be carried out
in the following chapters. Without exception, the theoretical performance, which means
the best reachable performance of each modulation format is the first step that will take
part in any future conclusion. This decission is going to be determined essentially taking
a look at the performances and checking how closer are to the ideal case. For this reason,
in Fig. 2.10 and in Fig. 2.11 the back-to-back simulations are plotted in order to set up
the best attainable performance for SC and OFDM. The back-to-back simulations have
been shown in terms of BER as a function of the optical signal-to-noise ratio (OSNR) for
different modulation formats that will be used in our analyses.
2.3. Optical Link

An optical communication system (OCS) consists, as any other communication system, of
three fundamental blocks: the transmitter, the communication channel and the receiver,
as displayed in Fig. 2.12. The first part of the OCS is the optical transmitter, whose
function is to convert an electrical signal into an optical one to be launched into the
optical fiber, being this the communication channel. LASERS are usually used as optical
sources because of their compact nature and the strong compatibility with optical fibers.
An important transmitter design parameter is the optical power launched into the optical
fiber. An optimal power launch power is required in order to make the signal possible to
be received at the transmitter with the minimum effect of channel impairments.
The second main component of an OCS is the optical channel. The role of every communication channel, independent of its nature, is to transport the signal from the transmitter
to the receiver. In our system, the channel is represented by the optical fiber. Optical
fiber has emerged as the optimal channel communication for OCS due to the low losses
that can affect the signal performance. Otherwise, optical fibers broaden signal because
of the chromatic dispersion effect. Another source of signal distortion that can appear
in optical fibers are the nonlinear effects. In the next sections, further details of all the
impairments induced in optical fibers are going to be explained.
The last component is the optical receiver. The purpose of the optical receiver is to
convert the optical signal into the electrical domain and to recover the transmitted data.
Recovering the trasmitted data means recover the distorsions suffered in the signal be-
26
2.3. Optical Link
Figure 2.10.: Back to back performance for Single Carrier
Figure 2.11.: Back to back performance for OFDM
27
2. Theory
cause of the tranmission through the fiber.
Figure 2.12.: Scheme of an Optical Communication System
2.3.1. Fiber Propagation

The propagation of a single polarization optical field E(z, t) through an optical fiber is
described by the Nonlinear Schr
odinger Equation (NLSE) as given by [14]:
E j2 2 E
+
= j|E|2 E E
2
z
2 t
2
(2.2)
where in [Np/km] is related to the attenuation that affects the signal power through
the fiber; 2 in [ps2 /(nm km)] is define as the dispersion parameter and in [(W km) 1]
is the nonlinear coefficient. Fiber propagation impairments are strongly dependent on
the physical fiber characteristics. On the other hand, from Eq. 2.2 can be clearly seen
that the signal power (|E 2 |) is directly related to the nonlinear effects, represented by
. This leads to the signal power dependence that is shown in Fig. 2.13 where nonlinearities increase the importance in signal performance as soon as the launch power
is raised up. Due to quantum-mechanical lower bounds on optical amplifier noise, the
received SNR can only be increased by launching higher signal power but is also effected
by nonlinearities. For this reason, both linear and nonlinear effects will be studied in
the next sections in order to understand the signal behaviour through the fiber and how
future techniques can mitigate nonlinearities.
2.3.2. Linear Impairments

Linear impairments can be defined as the basic but not the most important impairments
that degrade signal performance, hence full compensation is achieve by modern coherent
digital signal processing (DSP)-based receiver.
Fiber Attenuation
Once the signal is propagated into the fiber link, the power of the signal can suffer
changes in the average power value due to the fiber attenuation. As [19], this loses can
28
2.3. Optical Link
Figure 2.13.: Example of predicted information spectral density limits per polarisation
for non-linear transmission
be described by:
dP/dz = P
(2.3)
where the parameter is the attenuation coefficient. If the integral in Eq. 2.3 is calculated
and we assume that Pin is the input launch power to the fiber, the final equation that
would describe the output signal power at a certain point of the fiber is given by:
P (z) = P0 expz
(2.4)
where P0 is the input power and z is the length where the output power is calculated.
From Eq. 2.4 it can be seen that the optical output power decreases for longer distances.
For this reason, when a long-haul optical network is deployed, the necessity of periodic
optical amplifiers is strongly required in order to obtain an output signal power that
allows the receiver to recover the entire signal.
Fiber losses depend on the wavelength of transmitted light [19]. In our case, where a
wavelength near the 1.55 m is applied, the fiber exhibited a loss about 0.2 dB/km which
is the lowest value found in 1979 [20].
Chromatic Dispersion
The dispersion problem can be described as the pulse broadening result in the optical
signal inserted into a dispersive channel such as the optical fiber. Moreover, the chromatic dispersion effect is the cause of the resulting inter-symbol-interference (ISI) which
degrades the signal by corrupting the bits peak amplitude. When a 1-bit is transmitted,
if the amplitude is raised over the normal value, it can result in a optical interference
29
2. Theory
within the neighbouring 0-bits.
One of the most important constants that causes the dispersion effect is the mode propagation constant that can be described by:
() = n()
where n() is the wavelength dependent refractive index, the angular frequency and c
is the speed of light. The dependency of () can be seen as a polynomial using a Taylor
series expansion as is shown in [21].
1
1
() 0 + 1 ( 0 ) + 2 ( 0 )2 + 3 ( 0 )3 + ....
2
6
(2.5)
where 0 [1/km] denotes the constant phase shift, 1 [ps/km] corresponds to the speed
at which the envelope of the pulse propoagates, 2 [ps2 /km] is called group velocity
dispersion (GVD) and represents the change in group velocity with angular frequency
and finally the parameter 3 [ps3 /km] is referred to the change of GVD with angular
frequency. The dispersion parameter, which is the one that quantify the amount of
chromatic dispersion that affects an optical signal is given by D [ps/(km km)] and
sometimes refered as the second derivative, 2 . Its relation to the GVD mentioned before
is given by
D(z) =
2c
2 (z)
2
where is the signal wavelength. Finally, the coefficient 3 (z) of Eq. 2.5 is related to
the dispersion slope S(z) as the next equation shows
S(z)
4c
2c
2 (z) + ( 2 )2 3 (z)
3
By applying the same concept as the periodic amplifiers, periodic dispersion compensation
can be implemented. The resistance of a signal to CD helps to relax fiber specifications
and minimizes the need for installing dispersion compensating fibers (DCF). But, in some
scenarios the placement of periodic DCFs plays an important role. For this reason, further
studies over DCFs are going to be realised in Chapter 4.
Polarization-Mode Dispersion (PMD)
In an optical single-mode fiber, the two signal polarizations are simultanously propagated. Therefore, the output signal will be identically degenerated for both polarizations.
Practically, due to the manufacturing imperfections and other variations the two polarizations can not be equally degraded and are slightly nondegenerated. The resulting two
30
2.3. Optical Link

polarization-eigenstates exhibit different group velocities and hence, the different group
delay (DGD) is raised. This DGD effect is a dispersive pulse broadening that is commonly
called polarization-mode dispersion (PMD).
The main challenge of this linear impairment is its stochastic nature that makes PMD
prediction a really important issue. Nowadays, PMD compensation is not longer a considerable effect thanks to the equalization tecniques that can fully avoid this phenomenon
after the fiber propagation [22].
2.3.3. Nonlinear Impairments

The ultimate research in high capacity for optical communication systems has resulted in
the widely usage of WDM technology where the influence of nonlinearities have become
really important. Nonlinear impairments can be clasified into two major groups: (I) Kerr
nonlinearities and (II) Optical scattering.
Optical signals are strongly confined into the fiber core and thus, reaching high optical intensities. Due to that fact, the fibers index of refraction is affected and therefore
signal-induced refractive index changes translate to changes of the signals optical phases.
In this group of nonlinearities we can distingish 3 subgroups: (I) self-phase modulation
(SPM), (II) cross-phase modulation (XPM) and (III) four-wave mixing (FWM).
The last main group, optical scattering nonlinearities are caused by the interaction between the light and materials. Further information can be seen in [23] because in this
thesis only special attention to Kerr nonlinearities has been given.
Self-Phase Modulation
Power variation inside the channel causes changes in the refractive index of the fiber causing a nonlinear phase shift defined as self-phase modulation (SPM). To better understand
and quantify this effect, from Eq. 2.2 and without considering the chromatic dispersion
phenomenon, the resulting equation is
E
= j|E 2 |E ,
z
2
where the first group of terms is related to the nonlinear effects but if a solution of this
equation is found, the result can clearly show the nonlinear effects as is given by:
E(z, t) = E(0, t) exp(
1
) exp( jSP M (z, t)).
2
31
2. Theory
In this solution, the term that quantifies the SPM is define as
SP M (z, t) = |E(0, t)|2
1 exp(z)
.
It is clear that the phase shift induced to the signal performance is dependent on the power
of the signal. For this reason, the effect of SPM will be considerably more important at
the beginning of the fiber or at the beginning of each span (after the power amplifier),
when a high signal power value can be obtained.
Cross-Phase Modulation
SPM in conjunction with XPM limit the capacity of WDM transmission systems [24].
XPM effect is also caused by the intensity dependence of the refractive index but its nature
is related to the tranmission concept of a composite signal (WDM). The consequence of
transmiting a WDM signal is the effect that one pulse can force into another pulse with
different wavelength.
Once the transmission consists of two or more channels, the XPM that we would have to
consider in our system will be described by:
XP M,k(z) = (|Ek (0)|2 + c|E1k (0)|2 )
1 exp(z)
,
where k = 0,1 depending on the channel and c is a factor depending in the polarization
of both pulses.
Four-Wave Mixing
FWM as the case of XPM is a consequence of a composite optical transmission system.
The FWM is based in the interaction between four optical waves, that is the reason of
its own name. FWM can be reduced either by reducing the power per channel or by
preventing the perfect phase matching by increasing the CD or increasing the channel
spacing.
32
3. Nonlinear mitigation of propagation

effects
In the last chapter, a review of fiber propagation impairments hereafter considered has
been reported. The state of the art technology shows that linear impairments can be fully
compensated by using linear equalizer structures such as finite impulse response (FIR).
The key role that DSP has emerged with, makes possible to implement several techniques
that reduce and, in few occasions, fully compensate for nonlinearities. In this thesis, two
novel compensation techniques are presented: (I) Radio frequency pilot-tone (RFP) and
(II) backpropagation (BP).
3.1. Radio-frequency Pilot

3.1.1. Introduction
OFDM is a well-known modulation technique in the field of RF-communication and has
recently attracted researchers interest. For this reason, several investigations have been
carried out to addapt OFDM to optical systems and if needed to overcome its disadvantages where DSP plays an important role. One of the mentioned disadvantages in the
last chapter was the sensitivity of OFDM to phase noise.
The influence of phase noise on the OFDM system performance has been regarded, for
WIFI and optical systems, in [30]. In these publications, the influence of phase noise
was split into a multiplicative part, which is common to all the subcarriers and often
referred as Common Phase Error (CPE), and an additive part, which is usually named
as Inter-Carrier Interference (ICI). By comparing both effects, the main limiting factor
is indentified to be CPE in the case where coherent detection receiver is deployed. CPE
causes a constellation rotation affecting all transmitted symbols and, on the other hand,
ICI is a consequence of the orthogonality loss between subcarriers. Therefore, many investigations have came out with different techniques [32, 33]. But in our analysis, only
the RFP technique proposed in [18], has been implemented.
The main concept behind the RFP technique is the idea to revert the phase noise impairment by adding a RF-pilot tone in the middle of the OFDM spectrum at the transmitter.
33

Placing this tone in the middle of the spectrum has the main advantage that does not
affect the spectral efficiency thus no extra bandwidth is required. Furthermore, the pilot
is distorted by phase noise in the same way as the signal, experiencing an inverse phase
shift. The knowledge of pilot transmitted phase can be assumed and therefore simple
phase correction can be done, applying this information in a back phase-shift to the data
symbols.
The fundamental parameters that take part in this technique are: the pilot and in consequence its transmitted power value, the pilot frequency offset and the bandwidth of the
recovery pilot filter. In OFDM as it was mentioned before, no pilot frequency offset is
considered due to the placement of the pilot at the center of the OFDM spectrum. Moreover, it is important to note that the parameter related to the zero-forced subcarriers
appears and must be considered for the pilot insertion.
The first parameter that is considered at the transmitter is the pilot frequency offset,
which is a shift here referred to the center of the signal spectrum. In the case of OFDM,
no phase shift was applied but, in our SC simulation analysis, the pilot frequency offset
becomes an important factor.
Once the pilot is inserted, its transmitted power value is chosen through optimization.
The transmitted pilot power is associated to the power-to-signal ratio (PSR), defined as

PSR[dB] = 10 log10
PRF
POF DM

,
(3.1)
where PRF and POF DM represent the electrical power of the RF-pilot and the OFDM
baseband, respectively. In the transmission, the optimization of the PSR becomes a key
factor. If a low value of PSR is chosen, the RF-pilot is too weak and the amplified spontaneous emission (ASE) noise reduces the compensation quality at the receiver, whereas
for higher PSR values, the optical signal-to-noise ratio (OSNR) of the signal becomes too
low.
At the receiver, the pilot recovery is implemented at the DSP by downconverting and
then filtering the received pilot. Using the extracted phase information to recover the
signal rotation. As Fig. 3.1 depicts, the filter bandwidth is important depending on the
PSR because at the receiver, the tone can be broader after channel propagation and as
a consequence, an optimization in the filter bandwidth is required to perfectly filter the
pilot information.
34
Figure 3.1.: Spectrum of a transmitted and received OFDM signal with the RF-pilot at:
(a) Transmitter and (b) Zoom of (a) [35]
3.1.2. Setup for Orthogonal Frequency Division Multiplexing

In Fig. 3.2 the OFDM scheme used in our simulation analysis is depicted. This scheme
shows the three main parts of the system: the transmitter, the optical link and the
receiver part for the RFP method.
The WDM signal is composed by 9 channels which are generated at a net symbol rate
of 25 Gbaud. An overhead of 20% is added for forward error correction (FEC), 5% for
training symbol sequences and 7.8% for the cyclic prefix added in OFDM. After the
overhead addition, the gross symbol rate results being 34 Gbaud.
As depicted in Fig. 3.2, the next stage at the transmitter where the signal pass through,
is the fast fourier transform (FFT) where a standard size of 256 was chosen. Inside this
FFT size, 48 subcarriers were utilized for zero-padding, 206 were modulated and the 2
remaining corresponding to the Nyquist frequency. Afterwards, the pilot insertion in the
center of the spectrum forces a zero-gap of 10 subcarriers. An optimization of the PSR
should be carried out for every single launch power and in distance that one wants to
analyze. For example, Fig. 3.3 shows the optimized PSR values for the power values that
were chosen for our simulations. Then, the signal is fed to the digital-to-analog converter
(DAC) and into a low-pass filter (LPF) with joint 3-dB bandwidth of 15.9 GHz. The
system for OFDM employs a pre-emphasis stage for the compensation of the transmitter
low-pass filtering impairments.
As it was explained in the last chapter, after the transmitter, the next component of an
35
Figure 3.2.: OFDM system applying RFP technique
OCS is the optical link, where the signal is propagated until the optical receiver. In our
simulations the optical link consists of several spans of 95 km of standard single-mode
fiber (SSMF) with an Erbium Doped Fiber Amplifier (EDFA) with 6 dB of noise figure
(NF). Further details of the fiber loss and Kerr nonlinearities parameters can be found
in [34].
At the front-end, the signal is received by a coherent receiver and then fed into the
analog-to-digital converter (ADC) which is set to identical values as the DAC used at the
transmitter. Because of the pilot insertion at the transmitter, another stage to recover
the pilot-tone is placed at the receiver. As it was pointed before, the filter bandwidth
required to recover the pilot information is also optimized as the other RFP parameters
(see Fig. 3.3).
3.1.3. Setup for Single Carrier

Several researches have been carried out applying the RFP technique for the OFDM case
[35, 36, 38]. Recently, the idea of applying this technique to the case of SC has gained
more interest [37]. For this reason, in this thesis the implementation of the pilot tone for
the case of a SC has been considered in order to compare the benefits provided by RFP
technique in both cases.
The WDM signal is generated at a net symbol rate of 25 Gbaud. The same overhead as
in the OFDM case is added, but without the cyclic prefix, resulting in a symbol gross rate
of 31.5 GBaud. After signal generation, the same stage as for the OFDM is placed, where
the DAC in conjunction with a LPF with 3-dB bandwidth of 24 GHz are employed. In the
36
(a) PSR
(b) RFP bandwidth
Figure 3.3.: PSR and RFP bandwidth optimization in a 16-QAM modulation system
when RFP and BP are implemented
SC case, the RF pilot is implemented after the DAC and a required PSR was deployed. In
this thesis, no zero-padding for the SC case was considered because the pilot was applied
out of the signal spectrum, keeping in mind the idea of using more than one tone in
this technique for future investigations. This frequency offset was optimized, showing an
optimal value of 24 GHz as it can be seen in Fig. 3.4. At the end of the transmitter,
depending on the generated modulation used (RZ50 or NRZ) a pulse carver would be
utilized or not.
Figure 3.4.: SC spectrum after RFP addition

In Fig. 3.5 the same optical link as the one used in OFDM is reported, because it does not
37

depend on which modulation is transmitted. Finally, at the coherent receiver a bank of
4 ADC is applied to receive the transmitted signal. At the receiver the ADC parameters
can not be the same as the ones at the DAC due to the RFP insertion. For this reason
in our analyses, a larger ADC filter bandwidth of 23.1 GHz is required, incurring in a
less noise filtered efficiency as a consequence of ASE induced in this filtering. Then, the
RFP compensation is performed in order to recover the induced phase shift and apply it
to recover the trasmitted data symbols. The same RFP bandwidth optimization at the
receiver should be done in order to fully recover the pilot.
Figure 3.5.: SC system applying RFP technique
3.2. Back-propagation
3.2.1. Introduction
While the last section has reviewed one of the most important compensation techniques
to mitigate XPM, back-propagation method has gained more interest in the last years
as a novel technique for SPM compensation [39, 40, 41]. In Section 2.3.1 the NLSE has
been introduced, and from Eq. 2.2 it can be clearly seen that in the absence of noise,
the signal can be recovered by inverting the NLSE equation, defining this process as
backpropagating. We write the backpropagate equation as
E
N
)E,
= (D
z
is the linear and N
the nonlinear part. This operation is equivalent as inverting
where D
the propagation parameters that cause the nonlinearities which are , 2 and . Fig.
3.6 shows more clearly the concept behind BP. An exact solution to the last equation is
38
Figure 3.6.: Forward and Backward Propagation concept
given by [39] having as a result

+N
])E(z, t)
E(z + h, t) = exp(h[D
where t is the time, z is the exact position in each span and h the step size. The exact
solution to this equation is hardly affordable and therefore an approximation by the
Split-Step Fourier Method (SSFM) is used.
+N
])E(z, t) exp(hD)
exp(hN
)E(z, t)
exp(h[D
(3.2)
which is valid as long as the step size h is small enough. Therefore, every step is composed
of a linear and nonlinear step. In each step of every span the algorithm shown in Fig.
3.7 is applied. The linear step consists in the dispersion compensation applied into the
frequency domain by
E(z + h, t) = F 1 {F{E(z, t).H}},
where H = exp(jh2 (2f )2 /2). This step is similar as the disperson compensation employed by a standard linear frequency domain equalizer (FDE) where the CD has to be
estimated and then compensated. On the other hand, the power dependent nonlinear
mitigation can be given by this equation
E(z + h, t) = E(z, t). exp(jh|E(z, t)|2 )
Finally, the power normalization is the result of the amplifiers located after each span.
It is important to mention that BP requires an accurate knowledge of the fiber link
parameters in order to invert the value of these parameters in the BP implementation.
In addition, the span length, the fiber parameters, the dispersion map and the launch
power have to be known by the receiver.
39
Figure 3.7.: Block diagram of the BP algorithm

Nonlinear Channel Models
As the power is not constant through the span length due to the fiber losses, large steps
cannot be used with the last equation mentioned in the previous section. However, the
solution in Eq. 3.2 can be refined by evaluating the nonlinear part with a constant
envelope profile and varying the intensity. Once this modification is applied, larger steps
sizes can be used because the power does not have to be constant throughout the step as
it was in the case before. The approximated solution becomes
)E(z, t),
exp(hN N
E(z + h, t) exp(hD)
(3.3)
where hN is the nonlinear step size. In [39], three nonlinear models where presented depending on the position of the linear and nonlinear parts in the solution of the NLSE. The
Wiener model performs first the linear before the nonlinear one, having as a consequence
the chromatic dispersion compensation at the beginning of this model. The opposite to
the last model is done by the Hammerstein model where the nonlinear part precedes the
linear one. Finally, the third model called Wiener-Hammerstein is a cascade of the two
previous models and therefore is composed by three blocks. At the beginning, a linear
step is performed compensating for CD. Then, the nonlinear compensation is performed.
At the end, the remaining CD is compensated in an additional step.
Step Modes
The step mode can be defined as a key parameter inside BP algorithm and thus, an
optimization for the nonlinear and linear step value is required. The main challenge in the
step value is the fact that an increment in the number of steps per spans results in better
accuracy and performance of the BP algorithm, but incurring in a higher computational
complexity not implementable in real-time. Therefore an optimal value, between the
improved performance and the reduced complexity requirements, must be defined. The
first trivial step mode derivates from Eq. 3.2, where the nonlinear length is equal to the
linear part. In this mode, the span is divided into N equal parts. Hence, hN = h =
40
L/N, where L is the total span length and N the number of steps. This method hardly
estimates the signal power where small number of steps per span are utilized. As soon
as the N value is increased, better power estimation is approached, reaching a perfect
power approximation for N . As mentioned previously in this thesis, the nonlinear
effects are power dependent. The power behaviour decreases exponentially, neglecting
the nonlinear impact when low power values are assumed. Therefore, it is easy to find
the span length where the signal power plays a non negligible role. This length is called
effective length and can be calculated as follows
Z
Lef f =
Z
P (z)dz =
P0 exp(z)dz =
0
1 exp(z)
(3.4)
The second approach is to divide the span into equal steps and then calculate the effective
length in each step by using the equation mentioned before. In this case, a perfect power
estimation is found due to the fact that the surface under the power curve is equal to the
power in the effective length. An illustrative comparison is shown in Fig. 3.8.
(a) Step Mode 1
(b) Step Mode 2
Figure 3.8.: Comparison between different Step Modes
3.2.2. Setup for Orthogonal Frequency Division Multiplexing and

Single Carrier
In Section 3.1 different setups for OFDM and for the SC case have been deployed due
to the implementation differences in both cases. The transmitter scheme is the same as
in any other usual transmitters without BP implementation, due to the nature that the
BP is a compensation technique which is placed at the receiver [42]. Only the knowledge
of fiber link parameters are an important factor, and for this reason the parameters of
Table 3.1 should be transmitted and utilized at the BP stage in the DSP.
In Fig. 3.9 the DSP configuration at the receiver is shown. This can be described as
41
Attenuation [dB/km]
Dispersion Coefficient at 1550 nm [ps/nm/km]
Dispersion Slope [ps/nm2 /km]
Effective Core Area [m2 ]
Nonlinear Coefficient [1/(W.km)]
SSMF
0.19
16.8
0.057
80
1.14
Table 3.1.: Fiber Link paramerters [34]

the most important part in the BP technique, because is where the BP algorithm and
therefore the SPM compensation are applied. When the signal is coherently received at
the receiver, each channel is filtered out and synchronized in order to fed it into the BP
stage where the algorithm is compensating for SPM in each single channel.
Figure 3.9.: Receiver DSP block diagram for SC and OFDM case
42
4.1. Introduction
Lin et al. were the first to propose the usage of the dispersion compensating fiber (DCF)
in 1980 [44]. The first demonstration of the use of DCF was in 1992 [45]. Since then,
dispersion management (DM) techniques have been studied in particular because of their
effectiveness in mitigating effects such as CD. During the beginning of the last decade,
several DM techniques have been considered as an alternative to DCF. However, conventional DCF have been worldwide deployed in the order of more than hundreds of
thousands modules becoming then, as the most widespread technique for CD compensation.
As explained in Section 2.2.3, with the emerging of DSP-based technology applied to
coherent receiver, the compensation of CD using DM schemes is an option that will be no
longer considered for future optical systems. At the moment, a major issue concerns the
existing long-haul optical transmission links (usually employed at line rates of 10-Gb/s),
where dispersion compensation components (mainly DCF) have been employed. Since
it would be not cost-efficient to replace actual optical communication systems, in this
work we will investigate the performance of a POLMUX signal received by a DSP-based
coherent receiver for the case of DM. The other option called non-dispersion managed
link (NDM) will be ananalyzed in Chapter 5.
4.2. Performance of Dispersion Management System

As we previously pointed out, data-rate of 10-Gb/s or 40-Gb/s are currently implemented
into long-haul networks employing DM solution. In order to compensate for the overall
accumulated CD, DCF method was developed as a considerable effective option to solve
this problem.
The idea behind DCF method is simple and it is illustrated in Fig. 4.1. Each span
basically consists of two kinds of fibers; one is deployed as a SSMF which is the transmission fiber, and a second DCF that has positive second and third-order dispersion values
(DCF), act as CD canceler in the optical domain. This implementation is the chosen one
43
Figure 4.1.: Scheme of a Dispersion-Managed scenario
for our following simulations. Another example of DM implementation scheme are the
ones used in transoceanic communications which are composed by Dispersion-Managed
Cables (DMCs) [43]. As Fig. 4.1 shows, the fiber length of the DCF is not specified as
it depends on the scheme we want to use for the dispersion map, which is the evolution
of the dispersion compensation along the link. The design of a dispersion map is strictly
dependent on the nonlinear effects we want to compensate for. Two important parameters in a dispersion maps needs to be considered: residual and inline compensation. No
pre-compensation was considered in our analyses.
After each span, residual dispersion can be expressed as
Dres () = DTF ()LTF + DDCF ()LDCF ,
(4.1)
where DTF () and DDCF () are the dispersion coefficients of the transmission fiber and
DCF as function of the wavelength , while LTF and LDCF are the lengths of the transmission fiber and DCF, respectively. Fig. 4.2 shows two kinds of different dispersion maps
design. In our case, we considered the compensation for the 90% of the accumulated CD
per span and therefore, a 10% of residual dispersion remain after each span. Once the
desired amount of residual dispersion is decided, the resulting DCF length can be calculated and for our specific link, it corresponds to 8.45 km. Typically, in a DM scheme
the length of the SSMF is around 100km while for a DCF length must be considerable
shorter because of tis physical parameters of the DCF.
As explained in [21], DCF has a smaller core effective area and different core index profile
in comparison to SSMF. In addition, the triple cladding index profile forces a larger part
of the optical field to propagate in the cladding and thus, an increasement in the amount
of waveguide dispersion is found resulting than in large dispersion values. Another reason
why short lengths are chosen for DCF is the high attenuation value (see Table 4.1) and
44
4.2. Performance of Dispersion Management System
Figure 4.2.: Different dispersion compensation schemes for optical communications

systems [21]
the relatively high nonlinearity.
Fiber Longitude [km]

Insertion Loss [dB/km]
Dispersion Coefficient at 1550 nm [ps/km/nm]
Nonlinear Coefficient [1/(W.km)]
DCF
8.45
0.5
-170
5.1
Table 4.1.: Typical parameters of a Dispersion Compensation Fiber [43]

Besides the importance of choosing the right dispersion map in our DM scheme, there are
considerable effects such as the loss, PMD effects, and nonlinear impairment tolerance
which should be take into account. In this section, special attention only to the nonlinear
tolerance is paid, the loss is no longer considered a problem since low values of DCF
lengths are choosen [46].
In Chapter 3, an extensive review of nonlinear effects has been reported. Nonlinear impairments are among the strongest limiting factors of current optical networks. Therefore,
an analysis considering the nonlinear perfomance in DCF is important to be considered
[47],[48]. K.Forozesh et al. reported that the nonlinear tolerance with respect to both
SPM and XPM is severely degraded when inline dispersion compensation is employed
[49]. For these reasons, the removal of the DCF modules presents two main advantages:
(I) it allows maximum reachable distance as CD also reduces the nonlinear inter-channel
interactions (i.e. effects such as XPM become less important) and (II) it rises the flexibility of the link. Moreover the system design cost is reduced due to the DCF removal.
In a near future, optical links will be deployed without DCF modules. This new concept
is commonly identified by the name non-dispersion managed (NDM) link and it has been
shown in [50] that it provides a better performance compared to DM. In Chapter 5 a
better understanding of NDM systems will be given.
45
4.3. Pulse Comparison of Single Carrier and OFDM in DM

Systems
In this section, we analyze the performance of RZ50, NRZ, root raised-cosine filter (RRC)
[51] pulse shapes and OFDM, for an optical system where a DM compensation scheme is
employed. This analysis is carried out for different modulation formats on a SSMF transmission link. The performance of 50-Gb/s DP-BPSK, 100-Gb/s DP-QPSK and 200-Gb/s
DP-16-QAM were evaluated and compared for single carrier and OFDM transmission.
The simulation setup used for this simulation analysis is shown in [42], having the same
symbol rate for all the modulation formats mentioned before.
The study of 9 equally modulated and co-propagationg channels within a 50-GHz grid
is focused on the nonlinear tolerance and the spectral efficiency properties when inline
dispersion compensation is employed. Concerning the pulse-shaping obtained by using
several roll-off values, only the roll-off values = 0.3, 0.5 and 0.7 are evaluated. The
reasons of this range of values for is the following: if we had employed lower values,
the system performance would have been limited by the timing jitter. Moreover, values
for higher than 0.7 would have been meaningless, since the selected optical filter at
transmitter would have filtered out the excess spectrum generated by higher values.
At the receiver, a target BER of 102 was assumed, which corresponds to the current
FEC limit for 20% overhead and soft decoding. Fig. 4.3 shows the maximum reach in
a DM link as function of the launch power for the different modulation formats aforementioned. It can be clearly seen that the best performance are obtained by RZ50 pulse
shape, which reaches the longest distance for all the modulation formats, while the shortest corresponds to the OFDM case. The increased distance was calculated between the
lowest and the maximum distance reach resulting in 9%, 6% and 5% for DP-BPSK,
DP-QPSK and DP-16QAM, respectively. For these calculations, OFDM was no longer
considered due to the low performance compared to the other pulse shapes. The poor
results obtained by OFDM can be explained as in [52]. In fact, the combination of high
PAPR in OFDM systems in conjuntion with DM links leads to a coherently addition of
nonlinear distorsions from span to span giving a worse perfomance.
4.4. Nonlinear Mitigation for DM links

In last section, we considered the effects of using different pulse shapes on the nonlinear
performance of several modulation format over DM links. We already mentioned that
BP and RFP have been identified among the most promising techniques for nonlinear
46
(a) DP-BPSK
(b) DP-QPSK
(c) DP-16QAM
Figure 4.3.: Maximum reach distance versus launch power for 9 WDM channels through
a DM link [42]
47
Figure 4.4.: Full dispersion compensation map
compensation. At the time we are writing, no comprehensive investigations on the mitigation of nonlinear effects over DM scenarios has been reported. Regardless the poor
tolerance against nonlinear impairments, DM link are still the most deployed scheme in
actual optical communications transmissions and therefore an extensive investigation of
nonlinear mitigation is required.
4.4.1. Back-propagation for DM systems

In Section 4.2 the concept of DM compensation was introduced. Depending on the quantity of CD compensated, the dispersion map varies from one system to another. For
example, as illustrated in Fig. 4.4, the whole communication link results in a symmetric span, and CD and power value are equal after each span. If this dispersion map is
implemented, the BP would be equivalent to several nonlinear steps without linear steps
between them. Therefore, it can be clearly seen that the nonlinear steps can be merged
into one single step, and as a consequence, only one single BP for the whole link is required.
Fig. 4.4 shows the dispersion map of a fully compensated link composed of 8 spans
each of 95 km SSMF and 8.45 km of DCF. As it is depicts in Fig. 4.4 all the spans
have similiar CD behaviour. The 8 spans of the link can be merged into a single span
and the BP can be performed in one single step. Zied et al. showed a computational
complexity reduction under the assumption that the nonlinearity of the DCF is negligible
[53]. In order to verify that this complexity reduction does not lower the performance,
a simulation of 2 WDM POLMUX 16-QAM channels at a 224-Gb/s over a fully inline
compensated link was carried out. Fig. 4.5 displays the post-processing simulated data
for the case mentioned before. From our simulations we can conclude that the results
applying the BP algorithm with 15 steps are comparable to the ones where one single
48

step per link is applied.
Figure 4.5.: BP computational complexity reduction demonstration. Comparison

between applying BP with 15 and 1 steps
Nowadays fully compensation maps are not considered suitable for long-haul optical communications because by fully compensating for the CD we reduce the overall system performance making it less robust against nonlinearities. Since now, Fig. 4.6 is going to be
considered for the rest of this chapter as the referring dispersion map, where a 90% inline
compensation is employed. Starting from the point on the right (indicated with the letter
A) and tracing an horizontal line towards left, a series of points over the link with the
same amount of CD can be identified. Therefore, the fact that several points have the
same CD value can be exploited in order to reduce the BP operation complexity over the
transmission link. All the points where an equal CD value can be identified in Fig. 4.6
are not located at the beginning of each span. With respect to the back-propagation,
this value corresponds to the CD at a certain point of the span, being located within
the nonlinear length1 . As reported in [54], the BP performance (usually measured in
terms of improved nonlinear tolerance) is at least similar to the performance when the
Wiener model is deployed, if the nonlinear step is performed within the nonlinear length.
According to these results, the last 4 spans of the link in Fig. 4.6 can be merged together
and the nonlinear compensation can be performed within a unique step over them. Furthermore, the value of CD at point A can correspond to a CD value at an optimal point
of another span, so that the nonlinear compensation is optimized for this span.
We also point out that if the under-compensation fraction is smaller, more steps can
be packed together, reducing the complexity of the BP algorithm. As it was mentioned
1
Nonlinear Length depends on the fiber loss parameter and the span length [19]
49
Figure 4.6.: Dispersion compensation map deployed in our simulations
before, a residual dispersion has to be compensated at the end of the hole optical link
because not full compensation was done. This residual CD value can be easily compensated by a Time Domain Equalizer (TDE) due to the small reamining CD values.
For the dispersion compensation map displayed in Fig. 4.6, BP performs the nonlinear
compensation as follows. Firstly, the accumulated CD of the last span till the CD value
indicated by A is compensated in one single linear BP step. Afterwards, the nonlinear
compensation through the BP algorithm over the last and the following 3 neighbours
spans is performed. Another compensation over the following 3 neighbours spans should
be done. Finally, the residual CD is compensated as we mentioned before by applying a
TDE with an increased number of taps.
4.4.2. RFP for DM link

Alternative to BP, in Section 3.1 the RFP technique has been pointed as another important technique for nonlinear impairments mitigation. As it was explained in the chapter
before, RFP has been proposed for XPM mitigation [18]. Furthermore DM was found
to increase the importance of nonlinearities, especially of XPM [56]. For this reason, the
interest of RFP implementation in a DM scheme became to be significant.
As soon as OFDM emerged as a promising technique for long-haul communication systems, the amount of researches in this topic rose drammatically. Since DM have been
deployed worldwide, it becomes meaningfull to focus on the performance of such a modulation formats over this kind of link. So far, no comprehensive researches for single
carrier in DM systems applying RFP technique were carried out. In [42] an extensive
study on the XPM compensation in a CO-OFDM system was reported, investigating
50
4.5. 9x224 Gb/s POLMUX-16QAM system with Nonlinear Mitigation

different fibers and dispersion management options. It was observed that in systems with
in-line dispersion compensation, the nonlinear distorsions are correlated resulting in a
less nonlinear efficiency and thus, the RFP-based compensation is imperceptibly effective
for any evaluated constellation sizes for a OFDM system.
On the other hand, besides the application of RFP to multi-carrier transmission, we
considered important the effect of this compensation technique applied to single carrier
transmission, where so far only few investigations have been carried out. Afterwards, a
comprehensive comparison between the effects of RFP compensation with SC and OFDM
systems over DM maps is hereafter presented.
4.5. 9x224 Gb/s POLMUX-16QAM system with Nonlinear

Mitigation
In order to further investigate the effects of nonlinear mitigation over a dispersion compensation scheme, a 9 WDM POLMUX-16QAM at 200-Gb/s channels were transmitted
over a DM link employing SSMF. The performance of a 200-Gb/s POLMUX-16QAM
OFDM system is evaluated over the same system scheme applied in [61]. In this section,
different configurations are presented one after an other applying: (I) only BP, (II) only
RFP and (III) both techniques.
For the SC, only the NRZ digital format is considered in the comparison analysis. From
Fig. 4.3 when 16-QAM is performed, we can clearly see that NRZ and RZ50 have nearly
the same performance and therefore the investigation over both modulation formats would
result in a redundant information. All simulations are presented in terms of maximum
reachable distance as a function of launch power where the target bit error rate assumed
in our system was 102 .
Fig. 4.7(a) depicts the implementation of the above described configurations to analyze
the reliance of these techniques against SPM and XPM compensation for a SC system.
It has been mentioned in this chapter that XPM effect plays a significant limiting role
in DM links, for this reason RFP gives us a better performance compensating most of
the XPM effects. In the case when BP is applied, the performance is nearly the same as
NRZ case below a launch power value of 1 dBm. For higher power values, the mitigation
of SPM gives better improvement due to the raising importance of nonlinearities for high
launch power values. We can conclude that no XPM compensation is done applying only
BP for the DM link case. Finally we reported the same study for the OFDM case as
is displayed in Fig. 4.7(b). No improvement can be seen for launch power values under
1 dBm applying either BP or RFP. For higher power values the BP technique brings
better performance due to SPM compensation. On the other hand, RFP can not afford
51
a strong XPM compensation because high values of XPM present within DM link. For
this reason, the conclusion drawn for the SC case is valid also for OFDM, where actually
the improvement is even smaller.
(a)
(b)
Figure 4.7.: Maximum reach distance versus launch power for 9 WDM channels
transmission in a DM links for: (a) Single Carrier and (b) OFDM
52
5.1. Introduction
In last chapter an explanation of the basic principles of DM links was reported. As we
already mentioned, in a near future, all DM links will be substituted with NDM ones,
which are characterized by a significant simplification along the optimal link and by an
increased complexity of the receiver by means of the DSP. This flexibility relies on the
potencial that DSP can provide to optical communications. Implementing a DSP-based
coherent receiver shows an improvement in the system efficiency because the complexity
is moved to the receiver and thus NDMs links will become attractive for future long-haul
communications.
In this chapter we firstly introduce the main aspects dealing with this kind of systems and
secondly we show the performance assessment by means of post-processing the simulated
data. Finally, a comparison of NDM against DM will be carried out highlighting the
advantages and disadvantages of both systems.
5.2. Perfomance of Non-Managed Dispersion Systems

Non-dispersion managed systems have been found to be more interesting than DM systems because the electronic dispersion compensation (EDC) approach a better CD compensation rather than dispersion compensation schemes in presence of substantial nonlinearities (mostly XPM) [57].
Recently, several researches have been focused in WDM schemes in order to reach high
capacity values for long-haul optical communications. When a WDM transmission signal
is fed into the fiber, the XPM impairments raised exponentially their effects over the
transmitted signal. Therefore, apart from SPM, another impairment would degrade the
signal performance, which is the case of XPM. In last chapter the interaction between CD
and nonlinearities was considered. As soon as no inline CD compensation is performed,
a better nonlinear tolerance can be found in this kind of systems. K.Forozesh et al. investigated the differences between implementing DM and NDM links, showing that when
53
DCF are removed from the transmission line, all impairments are completely uncorrelated
and average out [49]. The averaging in this perturbations can lead to the high nonlinear
tolerance that was mentioned before. In this case, the simulations were considered for
OFDM signals transmitted over a SSMF fiber, which resulted in a degraded performance
when DM is employed.
It is clear that in order to further improve the signal maximum reach distance the advantages that a DSP-based receiver can bring into our system should be considered as the
best ones to obtain a better performance. As it was done in the last Chapter 4, RFP and
BP techniques will be considered in our following analyses. The effects of RFP and BP
mitigation techniques can be better considered for the case of NDM transmission links
due to the fact that SPM compensation for BP has a higher efficiency and RFP proved
to be better for NDM system since the effect of XPM is reduced compared to the DM
case [56]. Looking at the advantages of both mitigation techniques, in our simulations
we examinated the improvement applying RFP and BP together in different scenarios in
order to verify if the joint usage of RFP and BP can mitigate for most of the nonlinear
impairments (SPM and XPM).
5.3. Narrow Optical Filtering on 200-Gbit/s DP-16-QAM

It is well known that techniques like BP and RFP (see Chapter 2) are considered as novel
methods in the nonlinear compensation of an optical field. Additionally, there has been
significant interest in reducing the effect of nonlinearities by optimizing pulse shaping
[58, 51, 59]. Despite all these researches, the aim of the simulations carried out in this
section is to test the susceptibility of the pulse shaping to narrow optical filtering coming
from a cascade of reconfigurable optical add/drop multiplexers (ROADMs). The concatenation of ROADMs was emulated by a single band-pass filter (BPF) at the transmitter
which is expected to be the worst case scenario in terms of performance when distributed
filtering over the link is employed. In these simulations a dual polarization 16-QAM 200Gb/s was chosen for NRZ, RZ50 and RRC modulation formats. Moreover two different
types of fiber link were considered: SSMF and large effective area fiber (LEAF).
The performance of 200-Gbit/s DP-16-QAM with different pulse shapes and for different
transmitter optical filter bandwidth at the transmitter is evaluated. The same transmitter setup as the one in Section 4.3 was used. The optical link consists of a variable
number of spans of 95 km of SSMF or LEAF, each with an EDFA of 6 dB of NF. In order
to focus on the nonlinear effects, neither the transmitter LASER nor the local oscillator
are assumed to generate LASER phase noise.
54
5.4. Comparison performance of Single Carrier and OFDM in NDM systems

The nonlinear tolerance of 220-Gb/s DP-16-QAM systems with 9 WDM channels is evaluated when 34 GHz, 38 GHz and 42 GHz transmitter optical BPF are used, corresponding
to a 3 and 8 cascaded filters (for 38 GHz and 42 GHz), respectively. For all the simulations, only the roll-off factor of 0.3 is assumed for RRC case. In Section 4.3 it was shown
that the relation of efficiency versus best maximum reach distance is the best when rolloff factor equal to 0.3 is transmitted and for this reason in these simulations is going to
be considered the only case for RRC. Although a lower roll-off decreases the bandwidth
occupation, it introduces penalties due to its low timing jitter tolerance.
Fig. 5.1 shows the maximum transmission reached distance as a function of the launch
power for SSMF and LEAF fibers. In both cases, the maximum reach distance was calculated assuming a FEC limit of 102 . Similar to the results presented in Section 4.3
when SSMF is the optical link studied, the RZ50 presents the highest perfomance due
to the best nonlinear tolerance. However, this nonlinear tolerance property has a strong
dependence on the filter bandwidth applied at the transmitter. The difference between
the maximum reach for 34 GHz and 42 GHz is about 140 km, corresponding to a relative decrease in distance of 8.1% when the filter bandwith becomes narrower. For the
NRZ case this decrease corresponds to 4%, and for the RRC pulse shape, the effect of a
narrowing filter bandwidth is negligible. On the other hand, when LEAF is implemented
instead of SSMF the results show a lower reach distance for all pulse shapes. The lower
dispersion parameter of the LEAF causes a higher coherence between the nonlinear regions, which can be compared to the case when DM links are considered. For LEAF, the
difference between the considered filter bandwidths is not apparently clear as for SSMF.
That suggests a less benefit of pulse shaping for LEAF than for SSMF.
5.4. Comparison performance of Single Carrier and OFDM

in NDM systems
In the last chapter, the same analysis comparing SC and OFDM but with the difference
of implementing DM systems was carried out. The performance of 50-Gbit/s DP-BPSK,
100-Gb/s DP-QPSK and 200-Gb/s DP-16-QAM is evaluated for single carrier and OFDM
transmission. The same setup was employed having as a result a gross symbol rate of 31.5
Gbaud and 34 Gbaud for SC and OFDM respectively. Nonlinear mitigation techniques
were applied because the interest of these simulations is to investigate the different pulse
shapes and the advantages that can bring to long-haul optical communications.
Fig. 5.2 shows the maximum reach distance in a dispersion unmanaged link as a function
of launch power for three studied modulation formats. A target BER of 102 was assumed, which is below the FEC limit of 20% overhead and soft decoding. Similar to the
55
transmission for different pulse shapes
results presented in Section 4.3, the pulse shape RZ50 reaches the longest distance for all
modulation formats considered and the shortest corresponds to RRC with 0.3 of roll-off
factor. The difference between the lowest and the highest values of distance is 1510 km
for DP-BPSK, 707 km for DP-QPSK, and 233 km for DP-16-QAM, corresponding to an
increase of the reach of 11.1%, 11.3% and 15.5% respectively. It can also be observed
that the RRC and NRZ plots present similar results, meaning that adding a digital RRC
does not have an advantage regarding the nonlinear tolerance, but as shown in [42] it is
beneficial for increasing the spectral efficiency.
The results have shown that the maximum reach distance can be achieve by transmitting a RZ50 signal. For non-dispersion managed links the benefit using RZ50 was more
evident reaching an improvement of the transmission distance of approximately 11% with
respect to OFDM and the other pulse shapes. Regardless the lowest spectral efficiency
of RZ50, it obtains the highest nonlinear tolerance.
5.5. Pulse shape comparison for SC with back-propagation

Keeping on with the investigations where the comparison between different pulse shapes
for SC transmission systems were analyzed, in this section the nonlinear mitigation of
SPM is hereafter considered. In this section, the comparison between two different pulse
shapes (NRZ and RZ50) for single carrier, at a net data rate of 200-Gb/s employing a
56
5.5. Pulse shape comparison for SC with back-propagation
transmission for a dispersion unmanaged link
57
Figure 5.3.: Maximum reach distance versus launch power for a NRZ and RZ50 Single
carrier transmission of 9 WDM channels
POLMUX-16QAM is reported. The signals are coherently detected with the nonlinear
compensation scheme illustrated in Section 3.2.
Both signals are generated at the transmitter at a net symbol rate of 25 Gbaud, an overhead of 20% is added for FEC and 5% for training symbol sequences, having as a result a
gross symbol rate of 31.5 Gbaud for both cases. The scenario where the simulations are
going to take place are several spans composed of a dispersion unmanaged optical link
of SSMF fiber and EDFA with 6 dB of noise figure. As it is known, in a WDM system
SPM is not the only effect which degrades the transmitted signal; the mitigation of interchannel nonlinearities, particularly XPM represents a challenge for BP, as it requires the
information of the co-propagating channels. After the analysis of the BP implementation,
further simulations applying RFP in conjunction with BP will be reported within this
chapter in order to show the efficiency of RFP in mitigating XPM.
Fig. 5.3 shows the comparison between the results with and without BP in terms of maximum reached distance as a function of launch power for NRZ and RZ50 pulse shaping.
A target BER of 102 was assumed. For launch powers below -1 dBm, no improvement
by employing BP can be seen since this is still within the linear region. A significant improvement is obtained when the optical launched power is higher than 0 dBm. A 11.9%
of better performance is obtained at the optimal launch power, for the NRZ case. Similar
to the NRZ results, RZ50 gains about of 15.8%. Besides increased maximum distance,
both transmission pulse shapes reduce the SPM influence, shifting the optimal launch
power from 0 dBm to 1 dBm as a result of the nonlinear mitigation scheme employed.
58
Figure 5.4.: Receiver DSP block diagram for nonlinear mitigation
5.6. Joint work of back-propagation and RF pilot-tone for

SC and OFDM
Since nonlinearities are becoming one of the limiting factor for long-haul transmission, in
this section the joint usage of RFP and BP is prosposed to mitigate the nonlinearities of
a 200-Gb/s POLMUX 16-QAM SC and for OFDM system. The idea of the joint work of
BP and RFP have not attracted the attention for SC systems so far, only few researches
have been carried out using this idea. For this reason, we considered that would be interesting to analyze the performance of BP and RFP together in a SC system, in order
to later compare it to the OFDM case [62, 61].
As it was explained in Section 3.1 the insertion of the pilot in our simulations is placed at
the transmitter, with a specific offset respect to the central channel. The simulation setup
implemented in this simulation is the same as in Fig. 5.4. The performance of a 200-Gb/s
POLMUX-16QAM is evaluated for both modulation formats. At the receiver, the BP
and RFP stages are applied in order to compensate for SPM and XPM respectively. It is
important where and whether SPM and XPM compensation are located or used because
system performance in the DSP could be degraded. In the case when nonlinear compensation is performed first with RFP and then BP, the XPM and partially SPM would
be compensated by RFP and subsequently the remaining SPM should be compensated
by BP. The result of this implementation would be a redundant SPM compensation as
explained in [36]. For that reason and as shown in Fig. 5.4, BP is performed in our
simulation analysis before the RFP technique, thus BP removes SPM from the signal and
the pilot tone, and finally RFP compensates for the remaining XPM.
The obtained results for the aforementioned 200-Gb/s POLMUX-16QAM system are
reported presenting one after the other, the performance of this system with and with-
59
out BP applying RFP and finally for the case where both techniques are simultaneously
implemented. In the SC case only the NRZ pulse shape was considered. The comparison
between four different mitigation schemes are presented in terms of maximum reachable
distance as a function of launch power for NRZ pulse shaping where the target BER
assumed in our system is still 102 . As reported in the results of last section, no significant improvement can be seen for power values below -1 dBm since the signal is only
ASE-noise limited. For launched power values higher than 0 dBm the signal experiences
the effects of nonlinearities and the three compensation schemes mentioned before show
different tolerance against them. Compared to the case where no compensation is applied,
all techniques provide a clear benefit. Directly from Fig. 5.5(a) where NRZ results are
displayed, we can draw the following conclusions: the compensation provided by single
RFP is less performing than the one obtained by using a standard BP module. Moreover,
the joint use of BP and RFP mitigation provide the absolute best performance, and the
maximum improvement achieved is of about 18% compared to the NRZ case where nonlinear compensation was not implemented. On the other hand, in Fig. 5.5(b) the results
where OFDM is trasmitted are shown. The results with BP show an improvement of
2%, similar than the only usage of RFP which is 1%. In both cases the improvement
is less than for the case of SC. However, the combination of BP and RFP provides an
improvement, at the optimal launch power of 13%, more than applying only BP or RFP
separately, which means that XPM is efficiently compensated and both techniques must
be implemented together in order to push the OFDM performance to a value comparable
to the NRZ where no nonlinear mitigation techniques are applied.
60
(a) Single Carrier
(b) OFDM
POLMUX 16-QAM transmission with nonlinear mitigation in a NDM scenario
61
6. Conclusion
In this thesis, the comparison between the SC and OFDM modulation formats have been
investigated. Several scenarios and simulation setups have helped us to support all the
analysis shown in this work. The nonlinear mitigation technique has also been considered
as an important issue because better signal performance can be achieved, reducing the
effects of nonlinearities. With or without nonlinear mitigation single carrier has always
reported the best result in all the considered scenarios compared to the OFDM results.
The comparison between SC and OFDM was at the beginning primarily focused on the
scenarios where nonlinear effects were not compensated for. First of all the comparisons
of different pulse shapes and OFDM with different modulation formats were carried out.
The results reported that in the case where a DM is employed, OFDM performed the
worst in every situation. The best performance was obtained for SC when RZ50 pulse
shaping was transmitted. In fact, as we explained in this thesis that employing a RZ50
gives you the advantage of having a high quality NRZ pulse due to the optical filtering
at the transmitter. On the other hand, when comparing both modulations formats in a
NDM scenario OFDM reached the same performance as the worst RRC roll-off factor,
giving a difference between the OFDM and the RZ50 performance of 15.5% when a DP16-QAM was transmitted.
After looking at different systems performance where the most limiting factor was defined as the nonlinear effects, the two most popular nonlinear mitigation techniques were
employed, which are: BP and RFP. Both techniques show more efficient mitigation when
NDM scenario was considered due to the fact that in DM nonlinear effects dramatically
increase their importance as it increases XPM making it more challenging for the overall
compensation. After demonstrating that BP and RFP can really compensate for SPM
and XPM effects respectively, we proposed the idea of the joint usage of both techniques
in order to compare the performance of SC and OFDM when the nonlinear effects could
be considered a limiting factor any more. In that case, where BP and RFP where applied together in a NDM scenario we showed an improvement of 18% and 13% to SC and
OFDM respectively. These improvements where referred to the cases where no nonlinear mitigation was applied to the signal. However when we analyzed the DM links the
results showed how complicate is to compensate for the different nonlinear impairments
63
6. Conclusion
and thus, lower improvement was obtained.
After analyzing all the results obtained in all the investigations presented in this thesis certain results can summarized. SC has always reported a better performance than
OFDM concluding that SC is better than OFDM for the cases considered in this work.
Furthermore, the usage of nonlinear mitigation techniques concluded that the joint work
of BP and RFP could lead into an important advantage for long-haul optical communication systems.
64
ADC
Analog-to-Digital Converter
ASE
Amplified Spontaneous Emission
BER
Bit Error Rate
BP
Back-Propagation
BPSK
Binary Phase-Shift Keying
CD
Chromatic Dispersion
CO-OFDM
Coherent Detection Optical Orthogonal Frequency Division Multiplexing
CPE
Common Phase Error
DAC
Digital-to-analog Converter
DBPSK
Differential Binary Phase Shift Keying
DCF
Dispersion Compensating Fibers
DM
Dispersion Management
DOF
Degree of Freedom
DPSK
Differential Phase Shift Keying
DQPSK
Differential Quadrature Phase Shift Keying
DSP
Digital Signal Processing
EAM
Electroabsorption Modulators
EDC
Electronic Dispersion Compensation
EDFA
Erbium Doped Fiber Amplifier
FDE
Frequency Domain Equalizer
FDM
Forward Domain Multiplexing
FEC
Forward Error Correction
FFT
Fast Fourier Transform
FSK
Frequency-shift Keying
FWM
Four Wave Mixing
GVD
Group-Velocity Dispersion
ICI
Inter-Carrier Interference
IFFT
Inverted Fast Fourier Transform
ISI
Inter-Symbol-Interference
LEAF
Large Effective Area Fiber
LO
Local Oscillator
II
MIMO
Multiple-Input Multiple Output
MZM
Mach-Zehnder-Modulator
NF
Noise Figure
NLSE
Nonlinear Schr
odinger Equation
NRZ
Nonreturn-to-zero
OADM
Optical add/drop multiplexers
OCS
Optical Communication System
OFDM
Orthogonal Frequency Domain Modulation
OOK
On-Off Keying
OSNR
Optical Signal-to-Noise Ratio
QPSK
Quadrature Phase-Shift Keying
PAPR
Peak-to-Average Power Ratio
PLL
Phase-locked Loop
PMD
Polarization Mode Dispersion
POLMUX
Polarization-Multiplexed
PSR
Pilot-to-Signal Ratio
RFP
Radio Frequency Pilot
ROADM
Reconfigurable Optical add/drop Multiplexers
RZ
Return-to-Zero
SC
Single Carrier
SC-FDE
Single Carrier with Frequency Domain Equalization
SPM
Self-Phase Modulation
SSFM
Split Step Fourier Method
SSMF
Standard Single Mode Fiber
TDE
Time Domain Equalizer
WDM
Wavelength Division Mutliplexing
XPM
Cross Phase Modulation
III
Bibliography
[1] www.facebook.com
[2] www.twitter.com
[3] www.youtube.com
[4] I.P.Kaminow, T. Li, A.E.Willner,Optical Fiber telecommunications, vol. V-A, ELSEVIER in 2008.
[5] Kazovsky L., Multichannel coherent optical communications systems, J Lightwave
Technol, vol.5, pp.1095102 in 1987.
[6] M.Seimetz,Performance of coherent optical square-16-QAM-systems based on IQtransmitters and homodyne receivers with digital phase estimation, OFC/NFOEC,
NWA4 in 2006.
[7] W.Shieh, et al., Opt. Express in 2007.
[8] P.J.Winzer and R.J.Essiambre,Advanced Optical modulation formats, Proceedings
of IEEE, vol.94, no.5, pp 952-986 in 2006.
[9] E.Torrengo,
S. Makovjs,
D.S.Millar,
I.Fatadin,
R.I.Killey,
S.J.Savory,
and
P.Baybel,Influence of Pulse Shape in 112-Gb/s WDM PDM-QPSK Tranmission,

IEEE Photonics Technology Letters, vol. 22, no. 23 in 2010.
[10] T.Walzman and M.Schwartz,Automatic equalization using the discrete frequency
domain, IEEE Trans. Inf. Theory, vol.19, pp.59-68, Jan.1973.
[11] P.J.Winzer, A.H.Gnauck, C.R.Doerr, M.Magarini, and L.L.Buhl, Spectrally Efficient Long-Haul Optical networking Using 112-Gb/s Polarization-Multiplexed
16QAM, vol.28, no.4, pp.547-556 in Journal of Lightwave Technology, 2010.
[12] X.Zhou and J.Yu, Multi-Level, multi-dimensional coding for high-speed and highspectral-efficiency optical transmission, vol.27, no 16 in Journal Lightwave Technology in 2009.
Bibliography
[13] E.Ip, A.P.T.Lau, D.J.F.Barros, J.M.Kahn,Coherent detection in optical fiber systems, vol.16, no.2, in Optics Express in 2008.
[14] G.P.AGrawal, Lightwave technology :
Telecommunciations systems, Wiley-
Interscience in 2005.
[15] T.Pfau, S.Hoffmann,R.Peveling,S.Bhandare,S.K.Ibrahim,O.Adamczyk,M.Porrmann,R.No,
and Y.Achiam,Real-time synchronous QPSK Transmission with standard DFB
lasers and digital I&Q receiver, Tecnical Digest (CD), CThC5 in 2006.
[16] J.Armstrong,OFDM for Optical Communications, Journal of Lightwave Technology, vol.27, no 3 in 2009.
[17] W.Sieh, X.Yi, Y.Ma, and Q.Yang, Coherent optical OFDM: has its time come?,
vol 7, no 3, pp. 234-255 in Journal of Optical Networking, 2008.
[18] S.L.Jansen, I.Morita, T.C.W.Schenk, N.Takeda, and H.Tanaka, Coherent optical
25.8-Gb/s OFDM transmission over 4160-km SSMF, J.Lightw.Technol., vol.26, pp.615 in 2008.
[19] C.P.Agrawal, Fiber-Optic Communications Systems, Third Edition, in 2002.
[20] T. Miya, Y. Terunuma, T. Hosaka, and T. Miyoshita, Electron. Lett. vol.15, no.106
in 1979.
[21] Dirk van Den Borne, Robust Optical Transmission, in 2008.
[22] H.Y.Pua, K. Peddanarappagari, B.Zhu, C.Allen, K.Demarest, and R.Hui,An Adaptive first-Order Polarization-Mode Compensation system aided by polarization scrambling: Theory and Demonstration, J.Lightw.Techonol. in 2000.
[23] W. Sieh and I.Djordjevic,OFDM for Optical communications, in 2010.
[24] D.Marcuse, A.R.Chraplyvy and R.W.Tkach, Effect of fiber nonlinearity on longdistance transmission, J.Lightwave Technol. 9, 121-129 in 1991.
[25] C.Xie, G.Raybon and S.Chandrasekhar,Comparison of RZ and NRZ formats in
112-Gb/s PDM-QPSK Long Haul coherent Transmission Systems, OFC in 2011.
[26] B.Bakhshi,
M.Vaa,
E.A.Golovchenko,
W.W.Patterson,
R.L.Maybach,
and
N.S.Bergano,Comparison of CRZ,RZ and NRZ modulation formats in a 64x12.3

Gb/s WDM transmission experiment over 9000 km, in OFC.
VI
Bibliography
[27] X.Zhou and J.Yu,Spectrally-Efficient high-speed optical transmission technologies,
OECC in 2009.
[28] K.Nosu,
H.Toba,
and
K.Iwashita,Optical
FDM
transmission
technique,
J.Lightwave Technol. vol LT-5, no 9 in 1987.

[29] W.Shieh, H.Bao, and Y.Tang, Coherent optical OFDM: theory and design, vol.16,
no.2, Optics Exrpesss in 2008.
[30] E.Costa and s.Pupolin, M-QAM-OFDM System performance in the presence of a
nonlinear amplifier and phase noise, IEEE Trans. Commun., vol.50, no.3, pp.462-472
in 2002.
[31] A.G.Armada, Understanding the effects of phase noise in orthogonal frequency
division multiplexing (OFDM), IEEE Trans. on Broadc., vol.47, no. 2, pp.153-159,
in 2001.
[32] Q.Sou, A.Tarighat, and A.H.Sayed, Compensation of phase noise in OFDM wireless
systems, IEEE Trans. on Signal Processing in 2007.
[33] M.E.Mousa-Pasandi, and D.V.Plant,Noniterative interpolation-based partial phase
noise ICI mitigation for CO-OFDM transport systems, IEEE Photonics Techn. Letters, vol.23, no. 21 in 2011.
[34] S.Adhikari, On the nonlinear tolerance of 42.8-Gb/s DPSK with co-propagating
OFDM neighbours, IEEE LEOS Annu.Meeting, pp.40-41 in 2009.
[35] B.Inan, S.Randel, S.L.Jansen, A.Lobato, S.Adhikari, N.Hanik,Pilot-Tone-based
Nonlinearity compensation for optical OFDM systems, EUREL, paper Tu.4.A.16
in 2010.
[36] L.B.Du and A.J.Lowery, Pilot-based cross-phase modulation compensation for coherent optical orthogonal frequency division multiplexing long-haul optical communications systems, Opticas Letters, vol.36, no.9, in 2011.
[37] M.H.Morsy-Osman, L.R.Chen and D.V.Plant,Joint mitigation of laser phase noise
and fiber nonlinearity using pilot-aided transmission for single-carrier systems,
ECOC 2011.
[38] S.Randel, S.Adhikari, and S.L.Jansen,Analysis of RF-Pilot-Based phase noise compensation for coherent optical OFDM systems, IEEE Photonics Tech. Letters, vol.22,
no.17 in 2010.
VII
Bibliography
[39] D.S.Millar, S.Makovejs, C.Behrens, S.Hellerbrand, R.I.Killey, P.Bayvel, and
S.J.Savory, Mitigation of fiber nonlinearity using a digital coherent receiver, IEEE
Journal of selected topics in Quantum Electronics in 2010.
[40] F.Yaman and G.Li,Nonlinear impairment compensation for polarization-division
multiplexed WDM transmission using digital backward propagation, IEEE Photonics
Journal, vol. 1, no 2, pp.144-152 in 2009.
[41] E.Ip and J.M.Kahn,Compensation of dispersion and nonlinear impairments using
digital backpropagation,Journal of Tightwave Technology, vol 26, no.20, pp.34163425 in 2008.
[42] A.Lobato, M.Kuschnerov, A.Diaz, A.Napoli, Performance comparison of single carrier and OFDM in coherent optical long-haul communication systems, Asian Photonics Conference in 2011.
[43] L.Grner-Nielsen,
M.Wandel,
P.Kristensen,
C.Jorgensen,
L.V.Jorgensen,
B.Edvold, b.Plsdttir, and D.Jakobsen, Dispersion-Compensating Fibers, in

J.Light.Techonlogy, vol.23, no.11 in 2005.
[44] C.Lin, H.Kogelnik, adn L.G.Cohem,Optical-pulse equalization of low-dispersion
transmission in single-mode fibers in the 1.3-1.7 um spectral region, in Opt.Letters,
vol.5, no.11,pp.476-478, in 1980.
[45] J.M.Dugan, A.J.Price, M.Ramadan, D.L.Wolf, E.F.Murphy, A.J.Antos, D.K.Smith,
and D.W.Hall, All-optical, fiber-based 1550 nm dispersion compensation in a 10
Gb/s, 150 km transmission experiment over 1310 nm optimized fiber, OFC, paper
PD14 in 1992.
[46] J.Chesnoy, P.Nouchi, J.Y.Boniort, C.Brehm, C. Le Sergent, and P.Sansonetti, Optimum dispersion compensating fiber: birefringence and PMD simulation and fiber
process improvement, in Proc.Int.Wire and CAble Symp, pp.343-346 in 1995.
[47] F.Forghieri, R.W.Tkach, and A.R.Chraplyvy, Dispersion compensatin g fiber: Is
there merit in the figure of merit? IEEE Phot.Technol. Letters, vol.9, no.7, pp.970972 in 1997.
[48] P.Sillard, B.Dany, A.Bertaina, L.Curinckx, C.Bastide, O.Courtois, J.-C.Antona, and
S.Bigo,Simple criterion of quality to evaluate DCM impact on WDM system performance, OFC, paper FA3 in 2004.
VIII
Bibliography
[49] K.Forozesh, S.L.Jansen, S.Randel, I.Morita, and H.Tanaka,The inlfuence of the dispersion map in coherent optical OFDM transmission systems, IEEE LEOS Summer
Topical Meetings, paper WC2.4 in 2008.
[50] M.S.Alfiad, D.van den Borne, S.L.Jansen, T.Wuth, M.Kuschnerov, G.Grosso,
A.Napoli, H.de Waardt, A comparison of electrical and optical dispersion compensation for 111-Gb/s POLMUX-RZ-DQPSK, J.Light.Technology, vol. 27, no.16 in
2009.
[51] S.Makovejs et al.,Comparison of Pulse Shapes in a 224 Gb/s PDM-QAM16 Long
Haul Transmission Experiment, OFC in 2011.
[52] A.Lobato et al.,On the efficiency of RF-Pilot-based nonlinearity compensation for
CO-OFDM, OFC in 2011.
[53] Z.Maalej, A.Napoli, V.Sleiffer, M.Kuschnerov, B.Spinnler, and N.Hanik, Reduced
complexity back-propagation for optical communication systems, to be accepted for
OFC in 2012.
[54] Z.Maalej et al.,Reduced complexity for back-propagation method algorithm, IEEE
Phot. Conf., to be published (2011).
[55] A.Lobato, B.Inan, S.Adhikari, and S.L.Jansen,On the efficiency of RF-Pilot-based
nonlinearity compensation for CO-OFDM, OFC, OThF2 in 2011.
[56] A.Bononi et al.,Which is the dominant nonlinearity in long-haul PDM-QPSK coherent transmissions?, ECOC in 2010.
[57] V.Curri, P.Poggiolini, A.Carena, and F.Forghieri,Dispersion compensation and
mitigation of nonlinear effects in 111-Gb/s WDM coherent PM-QPSK systems,
IEEE.Photonics.Techonology Letters, vol.20, no.17 in 2008.
[58] B. Chtelain et al., SPM-Tolerant Pulse Shaping for 40- and 100-Gb/s DualPolarization QPSK Systems, in IEEE Photonics Technology Letters, vol.22, no.22
in 2010.
[59] G. Bosco et al., On the performance of Nyquist-WDM terabit superchannels based
on PM-BPSK, PM-QPSK, PM-8QAM or PM-16QAM Subcarriers, Journal of Lightwave Technology, vol.29, no1, pp 53-61 in 2011.
[60] D.S.Millar et al.,Mitigation of fiber nonlinearity using a digital coherent receiver,
JoST QE, vol.16, no.5, pp.1217-1226 in 2010.
IX
Bibliography
[61] A.Diaz,
A.Napoli,
S.Adhikari,
Z.Maalej,
A.Lobato,
M.Kuschnerov
and
J.Prat,Analysis of back-propagation and RF pilot-tone based nonlinearity compensation for a 9x224 Gb/s POLMUX-16QAM system, to be accepted for OFC, in
2012.
[62] TKobayashi et al.,45.2 Tb/s C-band WDM transmission over 240km using 538
Gb/s PDM-64QAM single carrier FDM signal with digital pilot tone, in ECOC in
2011.

PFC AlbertoDiazSouto

Uploaded by

Copyright:

Available Formats

PFC AlbertoDiazSouto

Uploaded by

Document Information

Original Description:

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

PFC AlbertoDiazSouto

Uploaded by

Copyright:

Available Formats

Masters Thesis

Comparison of OFDM with Single

Alberto Diaz Souto

1.1. Background and Motivation . . . . . . . . . . . . . . . . . . . . . . . . . .

1.2. Thesis Outline . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

2.1. Modulation Formats . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

2.1.1. Intensity Modulation Formats . . . . . . . . . . . . . . . . . . . . .

2.1.2. Phase Modulations . . . . . . . . . . . . . . . . . . . . . . . . . . .

2.1.3. Amplitude Modulations . . . . . . . . . . . . . . . . . . . . . . . .

2.2. Orthogonal Frequency Division Multiplexion vs. Single Carrier Transmission 20

2.2.3. Digital Signal Processing coherent receiver . . . . . . . . . . . . . .

2.2.4. Orthogonal Frequency Division Multiplexing and Single Carrier

2.3.1. Fiber Propagation . . . . . . . . . . . . . . . . . . . . . . . . . . .

2.3.2. Linear Impairments . . . . . . . . . . . . . . . . . . . . . . . . . .

2.3.3. Nonlinear Impairments . . . . . . . . . . . . . . . . . . . . . . . . .

3. Nonlinear mitigation of propagation effects

3.1. Radio-frequency Pilot . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

3.1.2. Setup for Orthogonal Frequency Division Multiplexing . . . . . . .

3.1.3. Setup for Single Carrier . . . . . . . . . . . . . . . . . . . . . . . .

3.2.2. Setup for Orthogonal Frequency Division Multiplexing and Single

4.2. Performance of Dispersion Management System . . . . . . . . . . . . . . .

4.3. Pulse Comparison of Single Carrier and OFDM in DM Systems . . . . . .

4.4. Nonlinear Mitigation for DM links . . . . . . . . . . . . . . . . . . . . . .

4.4.1. Back-propagation for DM systems . . . . . . . . . . . . . . . . . .

4.4.2. RFP for DM link . . . . . . . . . . . . . . . . . . . . . . . . . . . .

4.5. 9x224 Gb/s POLMUX-16QAM system with Nonlinear Mitigation . . . . .

5.2. Perfomance of Non-Managed Dispersion Systems . . . . . . . . . . . . . .

5.3. Narrow Optical Filtering on 200-Gbit/s DP-16-QAM . . . . . . . . . . . .

5.4. Comparison performance of Single Carrier and OFDM in NDM systems .

5.5. Pulse shape comparison for SC with back-propagation . . . . . . . . . . .

5.6. Joint work of back-propagation and RF pilot-tone for SC and OFDM

2.2. RZ-OOK transmitter scheme . . . . . . . . . . . . . . . . . . . . . . . . .

2.4. Setup implemented in a DBPSK transmitter . . . . . . . . . . . . . . . . .

2.5. Examples of QAM constellations . . . . . . . . . . . . . . . . . . . . . . .

2.6. OFDM time-domain transmission . . . . . . . . . . . . . . . . . . . . . . .

2.7. OFDM signal in time (left) and frequency (right) domain . . . . . . . . .

2.9. Example of a feedforward carrier-phase estimation . . . . . . . . . . . . .

2.10. Back to back performance for Single Carrier . . . . . . . . . . . . . . . . .

2.11. Back to back performance for OFDM . . . . . . . . . . . . . . . . . . . . .

2.12. Scheme of an Optical Communication System . . . . . . . . . . . . . . . .

2.13. Example of predicted information spectral density limits per polarisation

3.2. OFDM system applying RFP technique . . . . . . . . . . . . . . . . . . .

3.3. PSR and RFP bandwidth optimization in a 16-QAM modulation system

3.4. SC spectrum after RFP addition . . . . . . . . . . . . . . . . . . . . . . .

3.5. SC system applying RFP technique . . . . . . . . . . . . . . . . . . . . . .

3.6. Forward and Backward Propagation concept . . . . . . . . . . . . . . . . .

3.7. Block diagram of the BP algorithm . . . . . . . . . . . . . . . . . . . . . .

3.8. Comparison between different Step Modes . . . . . . . . . . . . . . . . . .

3.9. Receiver DSP block diagram for SC and OFDM case . . . . . . . . . . . .

4.1. Scheme of a Dispersion-Managed scenario . . . . . . . . . . . . . . . . . .

4.4. Full dispersion compensation map . . . . . . . . . . . . . . . . . . . . . .

4.6. Dispersion compensation map deployed in our simulations . . . . . . . . .

5.4. Receiver DSP block diagram for nonlinear mitigation . . . . . . . . . . . .

4.1. Typical parameters of a Dispersion Compensation Fiber [43] . . . . . . . .

Resum del projecte

Resumen del proyecto

1.1. Background and Motivation

A.Lobato, M.Kuschnerov, A.Diaz, A.Napoli, Performance comparison of single

1.2. Thesis Outline