You are on page 1of 136

University of Nevada, Reno

Carrier Wave Reutilization in Optical Fiber Communications

A thesis submitted in partial fulfillment of the


requirements for the degree of Master of Science in
Electrical Engineering

by

Mathew H. Henderson

Dr. Banmali Rawat/Thesis Advisor

December, 2016
Copyright by Mathew H. Henderson 2016
All Rights Reserved
THE GRADUATE SCHOOL

We recommend that the thesis


prepared under our supervision by

MATHEW H. HENDERSON

Entitled

Carrier Wave Reutilization In Optical Fiber Communications

be accepted in partial fulfillment of the


requirements for the degree of

MASTER OF SCIENCE

Banmali Rawat, Ph.D., Advisor

Indira Chatterjee, Ph.D., Committee Member

Hiroshi Sawada, Ph.D., Graduate School Representative

David W. Zeh, Ph.D., Dean, Graduate School

December, 2016
i

Abstract

Traditional communication links utilize a transmitter with a continuous wave laser

that creates a carrier wave. This carrier is then altered within the transmitter either by a

digital modulation technique or an analog modulation technique. These methods of carrier

wave modulation provide the means to transfer signal information to the receiving end of

the link. At the receiving end, the carrier wave is then filtered out to recover the modulating

signal.

This research presents a novel approach to improving power consumption in optical

fiber communication links. This improvement is accomplished by reutilizing the optical

carrier wave at the distant end of the link. This is done by use of an optical splitter, wherein

half of the carrier wave power is supplied to the detector and the other half is modulated

with new signal data. The new data is then transported back to the original system.

The aim of this thesis is to provide an overview of various carrier wave reutilization

techniques, their effects on the carrier wave and signal quality, and a comparative analysis of

the potential benefits of carrier wave reutilization as compared to traditional methodologies.


ii

Dedication

This thesis is dedicated to my amazing bride, Desiree. This entire endeavor has been

made possible by her constant encouragement, her devotion to our two wonderful children

and her patience. Thank you, my Love.


iii

Acknowledgments

I would like to express my sincere gratitude and appreciation to my thesis advisor,

Dr. Banmali Rawat. His passion for the field of optical fiber communication sparked my

interest in this area of research and his guidance was invaluable in ensuring the completion

of my thesis work.

I would also like to thank Dr. Indira Chatterjee (Department of Electrical and

Biomedical Engineering, University of Nevada, Reno) and Dr. Hiroshi Sawada (Department

of Physics, University of Nevada, Reno) for their willingness to serve on my advisory

committee and their dedication to advancing their respective fields. I truly enjoyed taking

your courses.
iv

Table of Contents

1 Introduction 1
1.1 Historical Trends in Communication 2
1.2 Power as a Limiting Factor 6
1.3 Carrier Wave Reutilization 8
1.4 Optiwave for Photonics Simulation 10
1.5 Thesis Overview 10
1.6 References 12

2 Carrier Wave Reutilization in a Digital Link 14


2.1 Digital Line Coding 15
2.2 Digital Modulation 18
2.3 Overview of a Digital Link 21
2.4 Proposed Implementation 23
2.5 Bit-Error Rate 25
2.6 Link Simulation 27
2.7 Numerical Results 34
2.8 References 41

3 Carrier Wave Reutilization in an Analog Link 42


3.1 Analog Modulation 43
3.2 Overview of an Analog Link 47
3.3 Proposed Implementation 48
3.4 Link Simulation 49
3.5 Numerical Results 60
3.6 References 69
v

4 Carrier Wave Attenuation and Dispersion 70


4.1 Attenuation 71
4.2 Dispersion 76
4.3 Attenuation and Dispersion Simulation – Link 1 82
4.4 Pulse Waveform at the Receiver 90
4.5 Attenuation and Dispersion Simulation – Link 2 92
4.6 References 96

5 Comparison of Modulation Techniques 98


5.1 Frequency and Phase Modulation 98
5.2 Digital DPSK-ASK CWR Link Model 102
5.3 Numerical Results 106
5.4 References 111

6 Carrier Wave Reutilization and a Standard Methodology 112


6.1 Power and Cost Comparison 112
6.2 Limitations of CWR 115
6.3 Numerical Results 116
6.4 References 122

7 Conclusion and Future Research Direction 123


vi

List of Figures

Figure 1-1: Optical Fiber Communication Link and Associated Waveforms 9


Figure 2-1: Non-return to zero line coding waveforms 16
Figure 2-2: Return to zero polar line coding with 50% duty cycle 17
Figure 2-3: NRZ ASK binary representation 19
Figure 2-4: Mach-Zehnder modulator diagram 20
Figure 2-5: Diagram of a simplex point-to-point link 22
Figure 2-6: OptiSystem model of a simplex point-to-point digital link 23
Figure 2-7: Proposed implementation of a digital link with CWR 24
Figure 2-8: Binary waveform showing variance at off and on levels 27
Figure 2-9: OptiSystem model of digital link with CWR 28
Figure 2-10: Eye diagrams of System 1 and System 2 35
Figure 2-11: Q Factor versus applied voltage during a zero-bit period 36
Figure 2-12: BER versus input optical power 36
Figure 2-13: OptiSystem model with optical power meters 38
Figure 2-14: Average power received versus output power, system 1 39
Figure 2-15: Average power received versus output power, system 2 39
Figure 3-1: D-IM of a 1550 nm optical carrier with a 10 MHz signal 44
Figure 3-2: Example subcarrier multiplexing system 45
Figure 3-3: OptiSystem model of a simplex point-to-point analog link 47
Figure 3-4: Proposed implementation of an analog SCM link with CWR 49
Figure 3-5: OptiSystem model of an analog SCM link with CWR, top half 50
Figure 3-6: OptiSystem model of an analog SCM link with CWR, bottom half 51
Figure 3-7: AM signals of system 2, 10 GHz carrier and 13 GHz carrier 54
Figure 3-8: Composite RF signal of system 2 55
Figure 3-9: SCM optical signal of system 2 56
Figure 3-10: AM signals of system 1, 17GHz carrier and 21 GHz carrier 58
Figure 3-11: Composite RF signal of system 1 59
Figure 3-12: SCM optical signal of system 1 60
vii

Figure 3-13: Eye diagrams for simulation 1 62


Figure 3-14: Eye diagrams for simulation 1 63
Figure 3-15: Variation in eye opening as power is increased 65
Figure 3-16: Minimum BER of all signals as power is increased 66
Figure 3-17: Minimum BER of each signal as power is increased 67
Figure 3-18: Minimum BER of each signal as bit rate is increased 68
Figure 4-1: Attenuation versus wavelength showing absorption effects 74
Figure 4-2: Dispersion and intersymbol interference 76
Figure 4-3: OptiSystem model of digital link with CWR 83
Figure 4-4: OptiSystem Gaussian pulse transmitter 84
Figure 4-5: Simulated Gaussian pulse waveforms 85
Figure 4-6: Simplified OptiSystem model 86
Figure 4-7: Attenuation of carrier pulse at 10 km 87
Figure 4-8: Intramodal dispersion at 45.09 km 89
Figure 4-9: OptiSystem model for receiver input waveform 91
Figure 4-10: Pulse response to dispersion and attenuation effects 91
Figure 4-11: OptiSystem model of Link 2 92
Figure 4-12: Attenuated signal of Link 2 93
Figure 4-13: Effect of intramodal dispersion in Link 2 94
Figure 4-14: Attenuation and dispersion in Link 2 95
Figure 5-1: Example waveforms for FM and PM 100
Figure 5-2: Example waveforms for FSK and PSK 101
Figure 5-3: DPSK receiver configuration 102
Figure 5-4: OptiSystem model of a DPSK-ASK CWR digital link 103
Figure 5-5: Output of DPSK transmitter 105
Figure 5-6: Eye diagram comparison of System 1 107
Figure 5-7: Eye diagram comparison of System 2 108
Figure 5-8: BER comparison of DPSK-ASK and ASK-ASK 110
Figure 6-1: Bidirectional ASK digital link 113
viii

Figure 6-2: ASK-ASK CWR digital link 113


Figure 6-3: ASK simplex digital link results 116
Figure 6-4: ASK-ASK CWR initial digital link results 117
Figure 6-5: ASK-ASK CWR return digital link results 118
Figure 6-6: BER of an ASK-ASK CWR digital link with increasing distances 119
Figure 6-7: BER of a simplex digital link with increasing distances 120
ix

List of Tables

Table 1-1: Power consumption of packet-switched routers 7


Table 2-1: Characteristics of point-to-point link components 22
Table 2-2: Typical operating parameters for pin photodiodes 32
Table 3-1: Electrical amplitude modulator parameter settings 52
Table 3-2: Simulation 1 eye pattern characteristics 64
Table 4-1: Absorption loss in silica glass due to common impurities 73
Table 5-1: Power comparison of ASK-ASK and DPSK-ASK CWR systems 108
Table 6-1: 1550nm Erbium CW laser input and output power 114
Table 6-2: Power Comparison of Traditional System and CWR System 115
1

CHAPTER 1

Introduction

All communication methodologies require three fundamental components: a

transmitter, a transmission medium and a receiver. This concept holds just as true for a

method as simple as speaking to a friend, as it does for the complex practice of sending a

coherent beam of light thousands of kilometers through an optical fiber medium. While the

methods of communication have changed over time, the fundamental components remain.

It is the communication method and its evolution that has created the capability to transfer

enormous amounts of data over long distances in a very short amount of time. However,

this increase in transmission capacity is not without bounds.

Historically, the primary limiting factor to transmission capacity has been the

bandwidth constraints imposed by physical components. However, with the enormous

growth of the Internet in recent years there has been a paradigm shift. For example, it is

asserted in [1] that the transmission capacity of the Internet may soon be constrained by its

power requirements rather than the bandwidth of its constituent components. It is also

shown in [2 - 3] that an important approach to limiting the energy consumption of optical

fiber communications within the Internet is to reduce inefficiencies in both the switching

infrastructures and transport technologies.

The aim of this thesis is to advance research in one-such transport technology. In

Section 1.1, the need for such a technology is expounded upon from the perspective of

historical trends. Furthermore, in Section 1.2, research is discussed regarding the concept of

power as a limiting factor. In Section 1.3, a conceptual overview is provided into the
2

proposed transport technology. In Section 1.4, the method by which simulations are

performed is discussed and Section 1.5 provides an overview into the organization of the

thesis.

1.1 Historical Trends in Communication

The history of communications is a story of technological advances followed by

exponential growth. There has always been a desire for people to communicate with each

other. This desire, coupled with ingenuity, has led to a myriad of communication systems

such as: signal fires, semaphore, the telegraph, radio, television and the Internet.

One of the oldest known methods of long-distance communication was used as early

as the eighth century B.C. and is attributed to the Greeks [4]. This method was done simply

by displaying various torches so as to be seen from a long distance. The number of torches

displayed represented a code. While this method was sufficient for simple messaging, a

number of factors limited its capability, including: requirement for line-of-site, ability of the

person sending the message to do so quickly and ability of the person receiving the message

to do so without errors.

In 1753, the first known method of electronic communication was documented. In

a letter to Scots magazine a writer known only by the initials C.M. delineated a method of

electro-static communication that was the predecessor of the telegraph. C.M.’s proposal

included a set of wires equivalent to the number of letters in the alphabet. A transmitter on

each line would then activate a mechanism on the receiving end to designate its associated

letter [5]. While C.M.’s concept was never implemented, there were a number of scientific

advances that propelled the design of a telegraph as it is now commonly known.


3

First, advances in materials were made that allowed for telegraph lines to be installed

either by stringing them up or by burying them. Second, the first electro-magnet was

constructed by William Sturgeon in 1825. This technological advance allowed for an electro-

magnetic means of communication. Finally, while working with Samuel Morse, Alfred Vail

devised the ‘Morse code’ in 1838 [5]. Despite all of these advances and numerous small-

scale demonstrations, the telegraph would not become part of the mainstream until 1843

when the United States congress signed the Telegraph Bill – the accomplishment of which

was largely due to the lobbying of Morse.

The success garnered by Morse resulted in a boom to the telegraph industry. What

began as a small conglomerate of competing firms coalesced to become a single company in

1866 – the Western Union Telegraph Company. Yearly messages sent over Western Union

lines increased dramatically. In 1867, the company sent 5.8 million messages and, in 1900, it

sent 63.2 million messages [6]. The industry also saw further improvements in its

technology, increasing its transmission rate from six words per minute to 60 words per

minute [6 - 7].

On March 10, 1876, the world of communications was changed once again. It was

on this date that the first words were spoken over a telephone line: “Mr. Watson, come here,

I want you,” [7]. By August of 1877 a total of 778 telephones were in use and, by 1878, the

Bell Telephone Company was organized with 12,000 telephones in service and $450,000 in

capital. The company floundered for a time as lawsuits were filed against it by the Western

Union Telegraph Company [7]. However, after striking a deal with Western Union in 1880,

the American Bell Telephone Company was created with 106,638 telephones in service. Just

ten years later 457,356 telephones were in service [8].


4

Following the advent of the telephone, the work of innovative pioneers such as

Michael Faraday, James Maxwell, and Heinrich Hertz eventually lead to a wireless telegraph

system which became known as the radio. In 1895, a young inventor by the name of

Gugleilmo Marconi had extended the wireless communication distance from 9 m to 2.4 km

and, in 1897, Marconi demonstrated his system to the British Post Office. Marconi had

envisioned that his communications system would eventually be used for seafaring

communications and, by 1899, a number of two-way exchanges from ship to shore were

operational [5].

Technological advances continued throughout the twentieth century. In the

telephone industry improvements were made to the circuit switching infrastructure as well as

the telephone transceiver. Wireless communication technologies improved as well. Wireless

transmitters were improved with the result being increased range. Concurrently, wireless

receivers were improved in both their sensitivity and selectivity. On October 4, 1957,

Sputnik 1 was launched by the Soviet Union and the first wireless satellite communications

was born [9]. At the same time, research into what would become the modern Internet was

underway.

In January of 1957, Leonard Kleinrock began his graduate studies at the

Massachusetts Institute of Technology [10]. Kleinrock believed that the current circuit

switching methodology was inadequate for use in large scale networks and, by 1962, had

written the first publication to introduce the concept of breaking messages into smaller

fixed-length pieces. These fixed-length pieces, or packets, improved the mean response time

by utilizing a packet switching method within a network. During the same time period,

President Eisenhower had created the Advanced Research Projects Agency (ARPA) which
5

funded technologies similar to Kleinrock’s research. In 1965, the ARPA director, Charlie

Herzfeld, managed the allocation of $1 million dollars to fund a project that would create a

network of the largest research computers in the United States. Thus, with the combination

of research into packet switching and funding from ARPA, the first wide area network was

launched. The first host-to-host message was sent at 10:30 p.m. on October 29, 1969 on the

network aptly named ARPANET [10].

The first four nodes of ARPANET were deployed later that same year and the

network had grown to fifteen nodes by 1971. Many advances followed rapidly after that.

For example, the AlohaNet was created. It was a 9600 bit per second radio network and the

research into its creation lead to a protocol known as carrier sense multiple access with

collision detection (CSMA/CD). The CSMA/CD then led to the advent of Ethernet

whereby CSMA/CD was implemented with coaxial cable as the medium. Then, both the

transmission control protocol (TCP) and the Internet protocol (IP) followed which allowed

other networks to be attached to ARPANET. This case of different networks

“internetworking” with one another became an important factor in the ARPANET

experiment. In fact, the term Internet was used due to the natural increase of

internetworking within ARPANET [10].

The growth of the Internet was exponential. By 1980, approximately two hundred

hosts were connected to ARPANET. Ten years later, the public Internet had over one

hundred thousand hosts interconnected. One of the greatest accomplishments during this

time period was the creation of the World Wide Web application. This invention, by Tim

Berners-Lee, created a method by which the Internet could be brought to the homes and

businesses of millions of people worldwide [11].


6

This exponential increase in the Internet led to greater demands for high-bandwidth

services. To date, this demand has been satisfied in large part because of the proliferation of

optical fiber communication links. In the late 1970’s the first optical fiber links were

installed. These links could transmit signals at a rate of six megabits per second at a distance

of approximately 10 kilometers. By the 1980’s, technological improvements had resulted in

fiber optic links that could carry aggregate data rates greater than terabits per second over

hundreds of kilometers. Furthermore, this could be done without the need to restore signal

fidelity along the path length [4].

Today, the Internet continues to increase in terms of both its availability and its link

capacity. Furthermore, with the addition of cellular technologies such as 4G LTE and a

large amount of research going into 5G, exponential growth is expected to continue.

Companies that create Internet services continue to provide bandwidth intensive

technologies such as high definition streaming video, telemedicine and video conferencing.

And, it is quite natural to infer that with new technologies such as 4K and Ultra HD that the

desire for more bandwidth will continue. However, these growth trends have resulted in an

infrastructure that is beginning to consume enough power so as to become a concern.

1.2 Power as a Limiting Factor

It has been shown in [12] that 0.4% of the total electricity consumption in

broadband-enabled countries is due to the Internet. Furthermore, based on the historical

growth of the Internet, this number is likely to increase in the future. For example, it is

shown in [13] that the power consumption of an electronic router, when scaled for greater

network speeds, will increase by a factor of 1,125. This factor is given under the assumption
7

that the power consumption of future high-capacity routers does not change. It can be seen

in Table 1-1, based on data from [13], that power consumption is improved in the case of

the optoelectronic router and greatly improved in the case of an all-optical router.

Table 1-1: Power consumption of packet-switched routers

Data Rate Electronic Optoelectronic Optical


(Tbps) (kW) (kW) (kW)
100 900 68 2
1000 9000 784 21

It is evident from Table 1-1 that optical technologies will be a key component in

lowering the effect of power usage. However, despite the improvements of optical devices,

the energy consumption at the core of the Internet will continue to increase. In [1] it has

been estimated that as Internet access rates increase to 8 Mbps, the power consumption will

increase to 1% of the global energy supply. When the access rate is further increased to 100

Mbps, this consumption will continue to increase to 7% of the global energy supply. This

increase, from approximately 40 GW to 130 GW [3], is a considerable amount of power. In

a world where energy consumption has become a key environmental concern, this increase

in power consumption is not acceptable.

Numerous papers have been written [14 – 18] to address the issue of power

consumption within the Internet. In these papers, various approaches have been taken to

limit the amount of power consumed. These approaches include: improvements to

manufacturing processes, adaptive power control during low-load periods, improved

methods of packet labeling and improvements to the switching infrastructure. This thesis is
8

focused on a novel approach to improve the power consumption in the transport system as

discussed in Section 1.3.

1.3 Carrier Wave Reutilization

In a standard optical fiber communications link, light energy is emitted from a source

and travels through optical fiber to be detected by a photodiode at the distant end. The

source may be either a light emitting diode (LED) or an injection laser diode (ILD). This

light source can be modulated either by direct intensity modulation (IM) using electronic

drive circuitry or external modulation by a device such as a Mach-Zehnder modulator. In

either case, the light source acts as a carrier wave for the information bearing signal.

Fig. 1-1 shows an example of an optical fiber link where the carrier wave is externally

modulated using amplitude shift keying (ASK). In the figure, it can be seen that the

modulated carrier wave incident upon the photodiode is lost as the information bearing

signal is received in the baseband. This is due to the fact that the frequency of the carrier

wave is outside of the response capability of the detector. The overall effect is that the

power associated with the creation of the carrier wave signal is largely lost.

This thesis focuses on various methods to reutilize an optical carrier wave at the

receiving end of a transport system. This idea was originated with Prof. Banmali S. Rawat,

University of Nevada, Reno, U.S.A., and I had the first opportunity to research it as my M.S.

thesis. To the best of our knowledge this is the first time that this very important topic has

been investigated. These methods of carrier wave reutilization (CWR) will enable the

retransmission of data by the distant end of the system without the need for creating a new
9

Figure 1-1: Optical fiber communication link and associated waveforms.


10

carrier wave. This provides a means to save the energy associated with carrier wave creation

and reduce the costs associated with source components.

1.4 Optiwave for Photonics Simulation

The Optiwave suite of photonics design software is used extensively throughout this

thesis [19]. Optiwave provides design and simulation tools and has wide acceptance

throughout industry and academia. This thesis primarily utilizes the OptiSystem software

that is a part of the design suite. OptiSystem enables users to plan, test and simulate optical

links in the transmission layer of optical networks. This software allows for the simulation

of the following system components which are integral to the simulations found throughout

this thesis: lasers, transmitters, direct receivers, coherent receivers, and single-mode fibers.

Furthermore, the following modulation formats are supported natively: non-return to zero

(NRZ) and return to zero (RZ). Finally, all system performance can be visually analyzed

with the following graphs: eye diagrams, Q-factor graphs, optical signal to noise ratio

(OSNR) graphs, and constellation diagrams. OptiSystem also provides models for highly

parameterized optical fibers. These models are used to characterize the effects of single

mode signal propagation including linear dispersion and stochastic polarization mode

dispersion (PMD).

1.5 Thesis Overview

This thesis begins by progressing through a series of proposed CWR systems.

Chapter 2 provides an overview of a standard digital transmission link and goes on to show
11

an implementation of a digital CWR link. This implementation is simulated using

OptiSystem to determine performance metrics such-as bit-error rate (BER) and the optical

power requirements of the system. Chapter 3 follows a similar format, but the proposed

system investigates an analog CWR link. In Chapter 4, this thesis examines the limiting

factors of CWR to include the dispersion and attenuation effects on the carrier wave as it

progresses through the system. Chapter 5 provides a comparison of modulation techniques

used in conjunction with CWR. Finally, in Chapter 6, a comparison of the originally

proposed digital CWR system to a standard methodology is evaluated. The conclusion of

Chapter 7 provides an overview of the findings of this research as well as a summary of

future research possibilities.


12

1.6 References

[1] R. S. Tucker, R. Parthiban, J. Baliga, K. Hinton, R. W. A. Ayre and W. V. Sorin,


"Evolution of WDM Optical IP Networks: A Cost and Energy Perspective,"
in Journal of Lightwave Technology, vol. 27, no. 3, pp. 243-252, Feb.1, 2009.
[2] R. S. Tucker, "Green Optical Communications—Part I: Energy Limitations in
Transport," in IEEE Journal of Selected Topics in Quantum Electronics, vol. 17, no. 2, pp.
245-260, March-April 2011.
[3] R. S. Tucker, "Green Optical Communications—Part II: Energy Limitations in
Networks," in IEEE Journal of Selected Topics in Quantum Electronics, vol. 17, no. 2, pp.
261-274, March-April 2011.
[4] G. Keiser, “Optical Fiber Communications,” Fourth Edition, McGraw-Hill, New
York, 2011.
[5] R. W. Burns, “Communications: An International History of the Formative Years,”
The Institution of Engineering and Technology, London, 2004.
[6] T. Nonnenmacher, “History of the U.S. Telegraph Industry,” EH.Net Encyclopedia,
edited by Robert Whaples, August 14, 2001.
[7] H. N. Casson, “The History of the Telephone,” A. C. McClurg & Co., Chicago,
1910.
[8] A. R. Foote, “Statistics of Manufactures: 1890,” U.S. Department of the Interior,
Census Office, June 1, 1892.
[9] V. Labrador, “Satellite Communication,” in Encyclopædia Britannica. 2016. Retrieved
from https://www.britannica.com/technology/satellite-communication
[10] L. Kleinrock, "An Early History of the Internet [History of Communications],"
in IEEE Communications Magazine, vol. 48, no. 8, pp. 26-36, August 2010.
[11] J. F. Kurose and K. W. Ross, “Computer Networking: A Top-Down Approach,”
Fifth Edition, Addison-Wesley, New York, 2010.
[12] J. Baliga, R. Ayre, K. Hinton, W. V. Sorin and R. S. Tucker, "Energy Consumption in
Optical IP Networks," in Journal of Lightwave Technology, vol. 27, no. 13, pp. 2391-2403,
July 1, 2009.
13

[13] S. Aleksić, "Analysis of Power Consumption in Future High-Capacity Network


Nodes," in IEEE/OSA Journal of Optical Communications and Networking, vol. 1, no. 3,
pp. 245-258, August 2009.
[14] S. Kawaguchi and T. Yachi, "Adaptive Power Efficiency Control by Computer
Power Consumption Prediction Using Performance Counters," in IEEE Transactions
on Industry Applications, vol. 52, no. 1, pp. 407-413, Jan.-Feb. 2016.
[15] S. Ibrahim, T. Nakahara, H. Ishikawa, and R. Takahashi, "Burst-mode optical label
processor with ultralow power consumption," in Opt. Express, 24, 6985-6995, 2016.
[16] W. Hou, L. Guo, X. Wang, X. Wei, “Joint port-cost and power-consumption savings
in hybrid hierarchical optical networks,” in Optical Switching and Networking, Volume 8,
Issue 3, July 2011, Pages 214-224
[17] A. Coiro, M. Listanti, A. Valenti and F. Matera, "Reducing Power Consumption in
Wavelength Routed Networks by Selective Switch Off of Optical Links," in IEEE
Journal of Selected Topics in Quantum Electronics, vol. 17, no. 2, pp. 428-436, March-April
2011.
[18] S. Palit, J. Kirch, L. Mawst, T. Kuech and N. M. Jokerst, “Thermal Characteristics of
III/V Thin Film Edge Emitting Lasers on Silicon,” in International Journal of Microwave
and Optical Technology, vol. 5, no. 6, November 2010.
[19] OptiSystem [computer program], 14.0.0.483, Ontario, Canada, Optiwave Systems,
Inc., 2015.
14

CHAPTER 2

Carrier Wave Reutilization in a Digital Link

An optical fiber communication link is the method by which data is conveyed from a

transmitter to a receiver over an optical fiber medium. Oftentimes, the data originates in a

binary format in the electrical domain. From there, the data is converted into the optical

domain by modulation of an injection laser diode (ILD) or a light emitting diode (LED). In

this chapter, it is shown that various methods of line coding and optical modulation

techniques may be used. Some of these methods are employed to create an example of a

basic digital link upon which carrier wave reutilization (CWR) may be used. The research of

this chapter is focused on the simulation of a CWR digital link and the resultant factors that

are associated with the simulation. These factors include Q Factor, bit error rate (BER), and

optical power.

This chapter is broken into seven parts. The first part, Section 2.1, provides an

overview of line coding while Section 2.2 discusses optical modulation techniques. Special

attention is given to the advantages and disadvantages of each technique. In Section 2.3, a

basic configuration for a digital link used in optical fiber communication is shown. The link

is described in detail in order to provide a foundation upon which the next section is built.

In the next section, Section 2.4, it is shown that a CWR digital link can be implemented with

minimal changes to the basic configuration. Section 2.5 provides an in-depth discussion of

bit-error rate (BER). Section 2.6 delves into the various components that make up the

simulation and how each of them functions as a whole. Finally, the results of the simulation

are provided in Section 2.7.


15

2.1 Digital Line Coding

In a communication system, digital line coding refers to the means by which binary

digits are represented in an electrical waveform. The proper selection of a line code can

greatly aid in the transmission of digital data. For example, self-clocking can be employed at

a receiver when a line code that provides ample number of transitions between bits is used.

Another advantage of proper line code selection would be the elimination of DC bias.

There are three primary types of coding: unipolar, polar and bipolar [1].

An example of unipolar line coding can be seen in Fig. 2-1(a). This form is perhaps

the easiest to implement and is a common communication method among TTL and CMOS

devices. During data transmission, unipolar code provides a positive pulse of amplitude, 𝐴,

to represent a ‘1’ bit and does not transmit a pulse to represent a ‘0’ – this form is often

referred to as on-off signaling. Despite the simplicity of unipolar signaling, it does have

some disadvantages. One problem is that unipolar signaling produces a relatively large DC

bias at the receiver. This is due to the fact that its average value is never at 0. A large DC

bias causes heating which may result in baseline drift in the receiver thereby increasing the

signal to noise ratio (SNR) [2]. Another problem with unipolar signaling is that it does not

provide any clock information and will not allow for self-clocking.

Another form of line coding, shown in Fig. 2-1(b), is polar. In polar line coding, a ‘1’

bit is represented by a pulse of amplitude, 𝐴, while a ‘0’ bit is represented by the negative of

this amplitude, −𝐴. The primary advantage of this form of signaling is that it decreases the

DC bias. Furthermore, the rms value of polar signals is larger when compared to that of

unipolar signaling. This results in a comparatively large energy level and a decrease of the

SNR at the receiver [2].


16

0 1 1 1 0 1 0 1 0 0
A

Voltage

0 1 2 3 4 5 6 7 8 9 10
(a) Time

A
Voltage

0
Time

-A

(b)

A
Voltage

0
Time

-A

(c)

Figure 2-1: Non-return to zero line coding waveforms, (a) unipolar line coding, (b) polar line
coding and (c) bipolar line coding.

Non-return to zero (NRZ) polar signaling suffers from the same lack of clock

information as that of unipolar signaling. Unlike unipolar signaling, this issue can be

overcome by using the return to zero (RZ) pulse shaping in Fig. 2-2. However, the RZ form

of polar signaling increases the bandwidth requirement when compared to NRZ [1].
17

0 1 1 1 0 1 0 1 0 0
A

Voltage 0
Time

-A

Figure 2-2: Return to zero polar line coding with 50% duty cycle.

The third type of line coding is known as bipolar and is shown in Fig. 2-1(c). Bipolar

line coding is a three-level system in which a ‘0’ bit is represented by the absence of a signal

while a ‘1’ bit is alternated between a positive pulse, 𝐴, and a negative pulse, −𝐴. For this

reason, bipolar line coding is also referred to as alternate mark inversion (AMI). The

alternating bit in AMI practically eliminates all DC bias. The drawback of AMI is that a long

run of ‘0’ bits results in no transitions in the data stream thereby increasing the probability of

a loss of synchronization in a self-clocking communication system. This problem has been

overcome by newer encoding schemes based on AMI. Examples of these schemes include

Binary with 8 Zero Substitution (B8ZS) and High-Density Bipolar 3-Levels (HDB3) [2].

Electrical line coding is an important consideration for the transfer of digital

information. While each of the methods discussed provide both advantages and

disadvantages, unipolar line coding is sufficient for use in the laser drive circuitry shown in

this research. This is due to the fact that the relatively short distances between the electrical

transmitter and the modulator do not result in conditions producing a DC offset that would

be detrimental to the SNR.


18

2.2 Digital Modulation

In optical fiber communication, either an ILD or an LED produces a waveform that

acts as an optical carrier for a message signal. This carrier can have its amplitude, phase, or

frequency varied with a message signal [1]. The method of imposing a message signal on an

optical carrier is known as modulation. There are three primary methods of digital

modulation in optical fiber communications – Amplitude-Shift Keying (ASK), Phase-Shift

Keying (PSK) and Frequency Shift Keying (FSK). This chapter focuses on ASK as the

means of digital modulation.

An electric field of an optical carrier wave can be represented by a plane wave as,

𝐸𝑠 = 𝐴𝑠 𝑐𝑜𝑠(𝜔𝑠 𝑡 + 𝜃𝑠 𝑡), (2.1)

where 𝐴𝑠 is the amplitude of the carrier field, 𝜔𝑠 is the angular frequency and 𝜃𝑠 is phase

shift [3]. When the carrier field, 𝐴𝑠 , is varied in proportion to a message signal, 𝑚(𝑡), while

both the angular frequency and phase are held constant, the carrier wave is amplitude

modulated. In this case, 𝐴𝑠 can be represented as

𝐴𝑠 (𝑡) = 𝑘𝑎 𝑚(𝑡), (2.2)

where 𝑘𝑎 is the amplitude sensitivity [1]. The amplitude modulated carrier wave can then be

written as

𝐸𝑠 (𝑡) = 𝑘𝑎 𝑚(𝑡) cos(𝜔𝑠 𝑡 + 𝜃𝑠 ). (2.3)

If the message signal, 𝑚(𝑡), is a digital signal, then 𝑚(𝑡) will take one of two values

during each bit period. In this case, the carrier is said to be modulated by amplitude-shift

keying (ASK). Oftentimes, in ASK, transmission of the waveform represents a binary one

while non-transmission of the waveform represents a binary zero. For this reason, ASK is
19

also referred to as on-off keying (OOK) [4]. An example of this form of modulation is

shown in Fig. 2-3.

(a)

(b)

0 1 0 1 1 0 1 0

Figure 2-3: NRZ ASK binary representation, (a) input electrical signal, (b) transmitted
optical waveform showing digital values.

In an intensity modulated direct detection (IM/DD) system, the electrical signal

directly modulates the laser transmitter, thereby providing amplitude modulation of the

optical power level of the light source. In this case, the current at the receiver is directly

proportional to the intensity of the optical signal at the detector, 𝐼𝑑 . Therefore, based on

(2.1), this yields

𝐼𝑑 = 𝐸𝑠 𝐸𝑠∗ = 0.5𝐴2𝑠 [1 + cos(2𝜔𝑠 𝑡 + 2𝜃𝑠 )]. (2.4)

However, since the frequency, 𝜃𝑠 , is twice the optical carrier frequency, the term involving

cos(2𝜔𝑠 𝑡 + 2𝜃𝑠 ) gets eliminated from the receiver. This elimination occurs because the
20

carrier frequency is beyond the response capability of the detector. Furthermore, the

elimination of the term causes 𝐼𝑑 in (2.4) to be simplified to

𝐼𝑑 = 0.5𝐴2𝑠 . (2.5)

When used to produce an ASK waveform, this form of intensity modulation results in the

photodetector acting as a square-law device [3]. Direct intensity modulation of the

photodetector is easy to implement. However, a problem occurs because the semiconductor

laser lacks the ability to maintain a stable output frequency with a changing drive current.

This is known as chirping and the instability results in linewidth broadening [3].

Furthermore, effects such as laser chirping can also contribute to changes in the linewidth,

making directly modulated lasers undesirable for data rates greater than 2.5 Gb/s [4]. For

these reasons, external optical modulation is preferred over direct modulation.

One technique for performing external modulation is electro-optical modulation. It

is shown in [5] that this form of modulation yields the highest quality signal. Using a Mach-

Zehnder modulator provides a practical means for implementing electro-optical modulation.

v1(t)

so(t) s(t)

v2(t)

Figure 2-4: Mach-Zehnder modulator diagram.


21

As shown in Fig. 2-4, the optical carrier wave signal, 𝑠𝑜 (𝑡), is provided as an input. This

signal splits into two paths of equal length, each comprised of a Pockel’s medium. In this

medium an electrical field is applied which changes the refractive indices of the two paths

relative to each other. This change results in a retardance, or difference in phase delay, that

is given by [6],
3𝑟 𝑉
2π𝑛𝑜 63
Δφ = (2.6)
𝜆0

where 𝑛0 is the index of refraction, 𝑉 is the potential difference in volts, 𝜆0 is the vacuum

wavelength in meters and 𝑟63 is the electro-optic constant in meters per volt [6].

When voltages 𝑣1 (𝑡) or 𝑣2 (𝑡) are applied (Fig. 2-4), the index of refraction of each

branch is modified. This modification results in the effect of each signal undergoing a phase

shift. The recombined output signal, 𝑠(𝑡), is a superposition of the two branch signals. This

resultant signal experiences destructive interference which is dependent upon the phase

relationship of each of the individual branch signals. When destructive interference occurs,

the output signal had a decreased magnitude as compared to the input signal.

Using external modulation provides greater signal fidelity and improves detection by

removing the instability associated with direct intensity modulation [3]. However, it is

important to note that the power of the applied optical carrier wave is being modulated and

that the destructive interference that occurs results in relatively large power losses. Despite

this limitation external modulation is used for all simulations conducted in this thesis.

2.3 Overview of a Digital Link

A simple link can be modeled by creating a point-to-point line with a transmitter on

one side and a receiver on the other as shown in Fig. 2-5. A link of this type places the least
22

demand on the optical fiber technology, creating a foundation by which more complicated

systems can be evaluated [4].

Optical Optical Fiber


Signal Source Optical Receiver User
Transmitter

Figure 2-5: Diagram of a simplex point-to-point link.

A number of technology decisions must be considered when designing a link due to

the interdependency of the fiber, the transmitter and the receiver. Some of these factors

include: single-mode versus multimode fiber, LED versus ILD source, and pin versus

avalanche photodiode (APD). Furthermore, while selecting each of these technologies, the

designer must consider the interrelated characteristics of each device as seen in Table 2-1.

Table 2-1: Characteristics of point-to-point link components [4]

Component Characteristics
 Core Size
 Refractive Index
Optical Fiber  Bandwidth/Dispersion
 Attenuation
 Numerical Aperture/Mode-Field Diameter
 Emission Wavelength
 Spectral Line Width
 Output Power
Optical Transmitter
 Effective Radiating Area
 Emission Pattern
 Number of Emitting Modes
 Responsivity
 Operating Wavelength
Optical Receiver
 Speed
 Sensitivity
23

A simplex point-to-point digital link modeled in OptiSystem is shown in Fig. 2-6. A

continuous wave (CW) ILD produces a constant optical signal as an input to the Lithium

Niobate (LiNbO3) Mach-Zehnder modulator. The modulator is controlled by two voltage

inputs from an NRZ Pulse Generator. The optical signal power that is output from the

Mach-Zehnder modulator is launched into 20 km of single-mode fiber before being coupled

to the pin photodetector. The photodetector produces a current that is proportional to the

optical input power, recovering the signal directly in the baseband. For the purposes of

simulation, this signal is then filtered and sent to a bit-error rate (BER) analyzer.

Figure 2-6: OptiSystem model of a simplex point-to-point digital link.

The digital link modeled in Fig. 2-6 serves as the foundation upon which system

simulations throughout this chapter are built. The first of these systems is discussed in the

next section.

2.4 Proposed Implementation

The first implementation of CWR is shown in Fig. 2-7. A series of digital ASK

pulses are produced by the transmitter of System 2 and output from the amplitude

modulator. However, prior to reaching the detector of System 1, the optical power is split
24

between two paths. The first path reaches the detector while the second path is used to

overlay a second signal utilizing the same carrier wave that was produced by System 2. This

reutilization is accomplished by sending the original signal through an amplitude modulator

in System 1. The signal then continues back to the detector in System 2.

SYSTEM 2 SYSTEM 1
Continuous Amplitude Detector Amplifier
Wave Laser Modulator
Laser Controller Decision

DATA IN
Drive Electronics

Amplifier Detector Amplitude


Modulator DATA IN
Decision Drive Electronics

Figure 2-7: Proposed implementation of a digital link with CWR.

In order to make this system feasible, the ASK modulation initially performed in

System 2 cannot be true OOK. Instead, the ASK signal is modulated so that a zero bit

period is not truly off, but rather, a percentage of the power used during a one bit period.

The decision circuitry of System 1 can then be set to identify the difference between both a

zero and a one bit. This proposed methodology has two advantages. First, the power of the

applied optical carrier wave in System 2 is not completely lost in the amplitude modulator

during the zero bit periods; such is the case if OOK were employed. Second, the optical

power of the carrier is used by System 1 for return signaling. As previously discussed, this

power is lost in a traditional digital link and System 1 would be required to generate its own

CW. This requirement would necessitate the use of another CW laser and control
25

electronics in System 1, thereby increasing the cost and the overall power required by this

system.

One potential drawback in this system is an increase in the bit-error rate (BER). The

BER increase as seen at System 1 would be a result of having a zero bit period that is

transmitted with a modulated pulse that is not truly at a power level of zero. The BER

increase as seen at System 2 is due to several factors. First, System 1 superimposes a signal

on a carrier that has already been modulated. Second, the power level from System 1 to

System 2 may be lower than in the traditional case wherein a new CW would have been

produced at System 1. Finally, the effects of signal dispersion would be increased as

compared to a traditional case. While these issues are examined throughout this thesis, the

use of a digital CWR link and its associated BER are shown in the next sections.

2.5 Bit-Error Rate

The rate of error occurrences in a digital stream can be measured by several different

means [4]. One of the simplest approaches is to divide the number of errors, 𝑁𝑒 , that occur

over a given time interval, 𝑡, by the number of pulses, 𝑁𝑡 , transmitted during the interval.

Thus, the BER is defined as [4]


𝑁𝑒 𝑁 1
𝐵𝐸𝑅 = = 𝐵𝑡𝑒 , 𝐵 = 𝑇 , (2.7)
𝑁𝑡 𝑏

where 𝐵 is the bit rate and 𝑇𝑏 is the bit transmission time or bit period. The BER is

dependent upon a number of factors such as signal-to-noise ratio (SNR) at the receiver,

interference from adjacent pulses and, the light source not fully turning off during a zero bit

period [4].
26

While (2.7) provides a simple definition of BER, the probability distribution of the

signal at the equalizer of the receiver must be known in order to compute the BER at the

receiver. For example, the probability that a one bit is identified at the receiver when a zero

bit was actually sent is given as [4]


2
∞ 1 ∞ (𝑣−𝑏𝑜𝑓𝑓 )
𝑃0 (𝑣𝑡ℎ ) = ∫𝑣 𝑝(𝑦|0)𝑑𝑦 = 2𝜋𝜎 ∫𝑣 exp [− 2𝜎2 ] 𝑑𝑣, (2.8)
𝑡ℎ 𝑜𝑓𝑓 𝑡ℎ 𝑜𝑓𝑓

where 𝑣𝑡ℎ is the threshold voltage, 𝑣 is the equalizer output voltage, 𝑝(𝑦|0) is the

probability that the output voltage is 𝑦 given that the input voltage is 0, 𝜎𝑜𝑓𝑓 is the standard

deviation of the noise during a zero pulse and, 𝑏𝑜𝑓𝑓 is the mean noise value during this same

period as shown in Fig. 2-8. Furthermore, the probability that a zero bit is identified at the

receiver as a one is also given [4]

𝑣
𝑡ℎ 𝑡ℎ 1 𝑣 (𝑏𝑜𝑛 −𝑣)2
𝑃1 (𝑣𝑡ℎ ) = ∫−∞ 𝑝(𝑦|1)𝑑𝑦 = 2𝜋𝜎 ∫−∞ exp [− 2 ] 𝑑𝑣, (2.9)
𝑜𝑛 2𝜎𝑜𝑛

where 𝑏𝑜𝑛 is the mean noise value during a one pulse and 𝜎𝑜𝑛 is the standard deviation of

the noise during this same period. Assuming that the probabilities of zero bits and one bits

are equally likely, then


1
𝑃0 (𝑣𝑡ℎ ) = 𝑃1 (𝑣𝑡ℎ ) = 2 𝑃𝑒 (2.10)

where

𝑃𝑒 = 𝑎𝑃1 (𝑣𝑡ℎ ) + 𝑏𝑃0 (𝑣𝑡ℎ ). (2.11)

The weighting factors 𝑎 and 𝑏 are determined by a priori distribution of the data and, in this

case, 𝑎 = 𝑏 = 0.5. Therefore, combining equations (2.8) and (2.11) yields the BER [4]

1 ∞
𝐵𝐸𝑅 = 𝑃𝑒 (𝑄) = ∫𝑄 exp[−𝑥 2 ] 𝑑𝑥
√𝜋 √2
1 𝑄
= 2 [1 − erf ( )] (2.12)
√2
27

where erf(𝑥) is the error function and 𝑄 is defined as

𝑣𝑡ℎ −𝑏𝑜𝑓𝑓 𝑏𝑜𝑛 −𝑣𝑡ℎ 𝑏𝑜𝑛 −𝑏𝑜𝑓𝑓


𝑄= = =𝜎 . (2.13)
𝜎𝑜𝑓𝑓 𝜎𝑜𝑛 𝑜𝑛 −𝜎𝑜𝑓𝑓

σ 2on
bon
Signal Voltage

νth
Time

boff
σ 2off

Figure 2-8: Binary waveform showing variance at off and on levels.

The value of 𝑄 is often used as a metric to specify receiver performance because it shows

the relationship between BER and the SNR. For example, 𝑄 may be used to determine the

SNR required to achieve a specific BER [4]. Values of BER and Q are referenced

throughout the link simulations in this thesis.

2.6 Link Simulation

In this section, the system defined in Section 2.4 is modeled and simulated. The

purpose of these simulations is to examine the effect that varying parameters, such as the

ASK level during a zero bit period or the power level of the optical carrier wave, will have

on the BER. The model components are discussed, including their transfer functions. The

system is modeled as shown in Fig. 2-9.


28

Figure 2-9: OptiSystem model of digital link with CWR.


29

The pseudo-random bit sequence generator shown in Fig. 2-9 generates a pseudo-

random binary sequence (PRBS) of 𝑁 bits where, [7]

𝑁 = 𝑇𝑤 𝐵𝑟 . (2.14)

The time window is set to, 𝑇𝑤 = 204.8 𝑛𝑠 to ensure that 2048 bits are simulated when the

bit rate is set to, 𝐵𝑟 = 10 𝐺𝑏/𝑠. Furthermore, a random number generator is used where a

Mark Probability parameter specifies the probability of ones in the sequence and has been

set to 0.5. The output of the bit sequence generator is fed to the NRZ pulse generator

block. The pulse generator is initialized to create pulses with an exponential edge shape

where, [7]
𝑡
−( )
1−𝑒 , 0 ≤ 𝑡 < 𝑡1
𝑐𝑟

𝐸(𝑡) = 1, 𝑡1 ≤ 𝑡 < 𝑡2 (2.15)


𝑡
−( )
𝑐𝑓
{ 𝑒 , 𝑡2 ≤ 𝑡 < 𝑇

𝑐𝑟 and 𝑐𝑓 are the rise time and fall time coefficients, 𝑡1 and 𝑡2 are aligned with the rise time

and fall time coefficients to generate pulses that align with 𝑇, the bit period. Furthermore,

the first simulation is run in sweep iterations of 100, where the amplitude of the pulse during

a zero bit period is swept in increments of 0.02 from 0 to 1.98. Further simulations are run

in sweep iterations of 50.

The output of the NRZ pulse generator is duplicated by the Fork 1x2 component,

where one copy is fed to the Dual Port Mach-Zehnder Modulator (MZM). The second copy

is inverted before being applied to the second port of the MZM. The modulator simulates a

Lithium Niobate (LiNbO3) MZM that consists of an optical input branch fed by a CW laser

operating at a wavelength of 1550 nm. The average output power of the laser is set to 0

dBm and the laser phase noise is modeled using the probability density function [7]
30

(Δ𝜑)2
1 −
𝑓(Δ𝜑) = ( )𝑒 4𝜋Δ𝑓𝑑𝑡 (2.16)
2𝜋√Δ𝑓𝑑𝑡

where Δ𝜑 is the phase difference between two successive time instants and 𝑑𝑡 is the time

discretization. A Gaussian random variable is assumed for the phase difference with zero

mean and a variance of 2𝜋√Δ𝑓 where Δ𝑓 is the laser line-width at full width half maximum

(FWHM). The output of the laser is multiplied with a complex vector considering the state

of polarization where [7]

𝐸𝑥 (𝑡) −𝑘
( ) = (√1 ) · √𝑃(𝑡) (2.17)
𝐸𝑦 (𝑡) √𝑘𝑒 𝑗𝜃

In this case, the power splitting, 𝑘, and the phase difference, 𝜃, are related to the parameters

of Azimuth, 𝛼, and Ellipticity, 𝜀, by the equations [7]

(2√𝑘(1−𝑘) cos(𝜃))
tan(2𝛼) = (2.18)
1−2𝑘

and

sin(2𝜀) = 2√𝑘(1 − 𝑘) sin(𝜃). (2.19)

Returning to the MZM, the NRZ pulse generator signals are applied to the electrical

input of the optical arms thereby controlling the degree of interference at the output of the

MZM. This optical field is determined by [7] as,

𝑗𝜋𝑣2 (𝑡) 𝑗𝜋𝑣


𝐸𝑖𝑛 (𝑡) ( )+( 𝑏𝑖𝑎𝑠2 )
𝐸𝑜 (𝑡) = 𝐼𝐿 [𝛾𝑒 𝑉𝜋𝑅𝐹 𝑉𝜋𝐷𝐶
( )
10 20

𝑗𝜋𝑣1 (𝑡) 𝑗𝜋𝑣


( )+( 𝑏𝑖𝑎𝑠1 )
+(1 − 𝛾)𝑒 𝑉𝜋𝑅𝐹 𝑉𝜋𝐷𝐶
], (2.20)

where 𝐸𝑖𝑛 (𝑡) is the optical input signal, 𝐼𝐿 is the insertion loss, 𝑣1 (𝑡) and 𝑣2 (𝑡) are the input

electrical voltages for the upper and lower arms respectively, 𝑣𝑏𝑖𝑎𝑠1 and 𝑣𝑏𝑖𝑎𝑠2 are the bias

voltages for the upper and lower arms, 𝑣𝜋𝑅𝐹 is the switching modulation voltage, 𝑣𝜋𝐷𝐶 is the
31

switching bias voltage and 𝛾 denotes the splitting ratio of the branches. These branches are

assumed to be symmetrical and the splitting ratio is given by [7]


1
1−( ) 𝜌𝑒
√𝜀𝑟
𝛾= , 𝜀𝑟 = 10 10 , (2.21)
2

where the Extinction Ratio, 𝜌𝑒 = 20 𝑑𝐵. In this simulation, the MZM is operated at a null

bias setting where (𝑣1 − 𝑣2 ) = 𝑣𝜋𝐷𝐶 = 𝑣𝜋𝑅𝐹 = 4𝑉.

The modulated optical signal produced by the MZM is then launched into a 10 km

length of single-mode fiber. This model takes into account the following types of

dispersion: group velocity dispersion (known also as chromatic or intramodal dispersion),

third-order dispersion and polarization mode dispersion. Furthermore, both material

dispersion and waveguide dispersion are subsets of group velocity dispersion. The types and

effects of dispersion are examined in more detail in Chapter 4. For now, it is sufficient to

say that the optical fiber is simulated with the default OptiSystem values associated with a

1550 nm optical wavelength.

The signal is then coupled to a 3 dB optical power splitter. The signal output for

each port is attenuated using [7]


𝛼

𝐸𝑖𝑛 (𝑡)10 20
𝐸𝑜𝑢𝑡 (𝑡) = , (2.22)
√𝑁

where 𝛼 is the power attenuation and 𝑁 is the number of output ports which, in this case, is

equal to 2. For the purpose of this simulation, the splitter is assumed to be ideal and

𝛼 = 0 𝑑𝐵. At this point, our signal has traveled from System 2 to System 1 (Fig. 2-7).

In System 1, one branch of the power splitter provides the input to an Indium

Gallium Arsenide (InGaAs) pin photodiode. An InGaAs pin was chosen as it provides the

highest responsivity for a wavelength of 1550 nm as compared to Silicon (Si) or Germanium


32

(Ge) [4]. Furthermore, the selection of a pin over an avalanche photodiode (APD) was

made in order to provide a worst-case scenario. An APD would provide better results as it

has a multiplicative factor that produces an increased sensitivity as compared to a pin.

Typical values for the responsivity of pin photodiodes can be seen in Table 2-2.

Table 2-2: Typical operating parameters for pin photodiodes [4]

Parameter Symbol Unit Si Ge InGaAs


Wavelength 𝜆 nm 400 - 1100 800 - 1650 1100 - 1700
Responsivity 𝓡 A/W 0.4 - 0.6 0.4 – 0.5 0.75 – 0.95
Dark Current 𝑖𝐷 nA 1 – 10 50 – 500 0.5 – 2.0
Modulation Bandwidth 𝐵𝑚 GHz 0.3 – 0.7 0.5 – 3 1-2

The pin photodiode converts the input optical signal into an electrical current based

upon its responsivity. The model also includes noise source modeling for dark current,

thermal noise and shot noise. The model has been configured to convert optical noise to

Gaussian noise inside of the signal bandwidth. The combined optical field is then converted

to optical power. The thermal noise calculation is based on the thermal power density as

follows [7]

√< 𝑖 𝑇2 >= √𝑃𝑤𝑟𝐷𝑒𝑛𝑠𝑖𝑡𝑦 × 𝑇ℎ𝑒𝑟𝑚𝑎𝑙𝑁𝑜𝑖𝑠𝑒𝐵𝑊, (2.23)

and the 𝑇ℎ𝑒𝑟𝑚𝑎𝑙𝑁𝑜𝑖𝑠𝑒𝐵𝑊 is set to the model sample rate of 320 GHz.

The shot noise is modeled with a Gaussian distribution where the optical power is

converted to electrical current by [7]

𝑖(𝑡) = 𝑖𝑠 (𝑡) + 𝑖𝐷 + 𝑖𝑠ℎ𝑜𝑡 (𝑡), (2.24)

where 𝑖𝑠 (𝑡) is the optical signal calculated from the responsivity, ℛ, as [7]

𝑖𝑠 (𝑡) = ℛ𝑃𝑠 (𝑡) (2.25)


33

and 𝑖𝐷 is the dark current, which is assumed to be 1.25 nA. Furthermore, the mean square

of the shot noise is obtained as [7]


2
< 𝑖𝐷+𝑠ℎ𝑜𝑡 >= 2𝑞(𝑖𝑠 + 𝑖𝐷 )B, (2.26)

where 𝑞 = 1.602 × 10−19 𝐶 and 𝐵 is the shot bandwidth which is set to the bit rate of 10

Gbps.

The output current of the pin photodiode is provided as an input to a low-pass

Bessel filter with the transfer function [7]

𝑑0 (2𝑁)!
𝐻(𝑠) = 𝛼 (𝐵 ), 𝑑0 = (2.27)
𝑁 (𝑠) 2𝑁 𝑁!

where 𝛼 is the insertion loss which is assumed to be ideal (0 dB). The value 𝑁 is the order,

which is configured to be a fourth order Bessel polynomial of the form

(2𝑁−𝑘)!
𝐵𝑁 (𝑠) = ∑𝑁 𝑘
𝑘=0 𝑑𝑘 𝑠 , 𝑑𝑘 = 2𝑁−𝑘 𝑘!(𝑁−𝑘)! (2.28)

and

𝑓𝑤𝑏
𝑠 = 𝑗( ) (2.29)
𝑓𝑐

where 𝑓𝑐 is the filter cutoff frequency defined as 𝑓𝑐 = 0.75 × 10 𝐺𝐻𝑧 and 𝑤𝑏 denotes the

normalized 3 dB bandwidth which is obtained via a table of values internal to OptiSystem

[7]. The final waveform calculated via the Bessel transfer function is fed as an input to a

BER analyzer to provide values of 𝑄 and BER.

Finally, the second branch of the power splitter is fed to the input of another MZM

block. This MZM is also ASK modulated by a second NRZ pulse generator. However, the

modulation performed here is a true OOK which results in no optical power being sent

during zero-bit periods. All other components are modeled with the same specifications as

were previously discussed.


34

2.7 Numerical Results

The digital link using CWR is simulated using the parameters specified in Section 2.6.

A 10 Gbps link is modeled and the amplitude of the zero-bit period, 𝑉𝑇𝑧𝑏 , of the NRZ Pulse

Generator in System 1 (Fig. 2-7) is swept from 0 to 1.98 V in increments of 0.02. It is

shown in Fig. 2-10(a), that the eye opening at System 1 is clearly visible at 𝑉𝑇𝑧𝑏 = 0 while

the eye opening at System 2 is not visible at all. In fact, the BER of System 2 is equal to 1 at

this point – one error for every bit transmitted. However, as the value of 𝑉𝑇𝑧𝑏 is increased,

the eye opening of System 2 increases, while that of System 1 decreases. Eventually, the eye

opening of System 1 closes completely as shown in Fig. 2-10(d) which results in that system

having a BER equal to 1. These results imply that there is a point of equilibrium wherein the

BER of both System 1 and System 2 can be minimized or, conversely, the Q factor

maximized.

In order to explore this concept, the maximum 𝑄 factor for each system is examined

as 𝑉𝑇𝑧𝑏 is varied, as shown in Fig. 2-11. It is shown when 𝑉𝑇𝑧𝑏 = 1.44𝑉 that the Q factor

for each system is maximized. At this voltage, the 𝑄 factor is approximately 9.5 for both

systems and, from (2.12), results in an extremely low BER of 10-21. This is an acceptable

value, as even high-speed SONET links operating at 622 Mbps only require BERs of 10-12

[4]. However, it is worth noting that a BER of this magnitude, which is theoretical

mathematically, may not be achievable in a practical system.

In finding that the optimal value for 𝑉𝑇𝑧𝑏 is 1.44 V, a second round of simulations is

shown in Fig. 2-12. In this simulation, the amplitude of the zero-bit period is maintained at

its optimal value while the power output from the CW laser is varied from -6 dBm to 6 dBm.
35

(a)

(b)

(c)

(d)

Figure 2-10: Eye diagrams of System 1 (left) and System 2 (right). The value of 𝑽𝑻𝒛𝒃 is
(a) 0V, (b) 1V, (c) 1.4V and, (d) 1.98V.
36

45 System 1

40 System 2

35

30

Q Factor 25

20

15

10 1.44 V

0
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
Voltage Applied to MZM of System 1 During a Zero-Bit Period
(V)

Figure 2-11: Q Factor versus applied voltage during a zero-bit period.

It is shown that, as the optical output power is increased, the BER is improved. However,

the BER of System 1 improves dramatically while that of System 2 only improves slightly.

Figure 2-12: BER versus input optical power.


37

This variation in the BER of System 1 can be attributed to the smaller pulses as compared to

System 2. At lower power levels, these pulses are diminished by the noise effects of the

system. However, at greater power levels, these pulses are sufficiently large so as to

overcome these effects. That, coupled with the fact that the carrier wave is not greatly

diminished, results in an improvement of the BER of System 1 past that of System 2,

wherein reutilization has been performed. This shows that the optimal biasing of the initial

transmitter must be modified for differing power levels.

Finally, the average optical power coupled to the input of the photodetector of both

System 1 and System 2 is examined. The Optical Power Meter block is added to the output

branch of the Power Splitter and to the output of the Optical Fiber block shown in Fig. 2-

13. A simulation is then run while sweeping the optical output power of the CW laser. The

same parameters that were used in the previous simulation are maintained for this one.

The results of the power sweep iterations are shown in Figs. 2-14 and 2-15. Both

graphs increase linearly as the output power of the CW laser is increased from -6 dBm to 6

dBm. This is expected as the distribution of our binary signal provides for an equal

probability of zero bits to one bits. There are, however, two important points to be noted.

First, average optical signal power received at System 1 is 3 dB lower than that of a

traditional system. This is evident due to the usage of the Power Splitter prior to the input

of the photodetector. Second, the average optical signal power received at System 2 can be

considered a decrease in wasted power as compared to a traditional system. This is true

because that power was not needed to successfully transmit the information to System 1 as

evidenced by the Q Factor in Fig. 2-12 and, it has been used to successfully transmit a

second message signal to System 2.


38

Figure 2-13: OptiSystem model with optical power meters.


39

Figure 2-14: Average power received versus output power, system 1.

Figure 2-15: Average power received versus output power, system 2.


40

These series of simulations prove that using the CWR methodology on a digital link

is a feasible means to attaining bidirectional communications. The BER is well within an

acceptable level and can be maintained with varying levels of optical input power.

Furthermore, optical power that would not have provided an added benefit using a

traditional digital link is utilized to create a communications link on the return path. This

has the added value of removing the need for circuitry and components associated with

having a CW laser in System 1.


41

2.8 References

[1] S. Kumar and M. J. Deen, “Fiber Optic Communications: Fundamentals and


Applications,” John Wiley & Sons, Ltd., West Sussex, United Kingdom, 2014.
[2] B. P. Lathi and Z. Ding, “Modern Digital and Analog Communications Systems,”
Fourth Edition, Oxford University Press, 2009.
[3] J. Senior, “Optical Fiber Communications: Principles and Practice,” Third Edition,
Pearson Education, Harlow, England, 2009.
[4] G. Keiser, “Optical Fiber Communications,” Fourth Edition, McGraw-Hill, New
York, 2011.
[5] J. C. Mauro and S. Raghavan, “Advanced modulation formats for fiber optic
communication systems,” Journal of Scientific Modeling and Simulation, vol.15, no.1,
pp.283-312, 2008.
[6] E. Hecht, “Optics,” Fourth Edition, Pearson Education Inc., San Francisco, CA,
2002.
[7] OptiSystem [computer program], 14.0.0.483, Ontario, Canada, Optiwave Systems,
Inc., 2015.
42

CHAPTER 3

Carrier Wave Reutilization in an Analog Link

A majority of the optical fiber communication systems are designed with digital

communications in mind [1]. Despite this fact, there are a number of systems where analog

optical fiber communications have an advantage. Some examples of these systems include:

transmission of microwave-multiplexed signals, subscriber services using hybrid fiber-coax

(HFC), video distribution, radio-over-fiber (RoF) and RADAR signal processing [2]. In

these systems, the use of analog transmission avoids both the complexity and the cost of

digital terminal equipment [1]. In addition, an analog link also avoids signal degradation due

to quantization noise – the error introduced during analog to digital conversion (ADC).

Intensity modulation of an optical carrier wave with an analog signal is simple to implement.

However, there are inherent challenges to the implementation of the analog link itself. This

chapter provides a discussion of these challenges and introduces a novel approach to

implementing CWR on an analog link.

This chapter is broken into five sections. In Section 3.1 an overview of various

analog modulation methods is provided while focusing on the advantages and disadvantages

of each. In Section 3.2, a basic analog link used in optical fiber communication is discussed.

The link is described in detail in order to provide a foundation upon which the next section

is built. In Section 3.3, it is shown that a CWR analog link can be implemented with minimal

changes to the basic configuration. Section 3.4 introduces new components that are used in

the analog link and how they interact with one another. Finally, the results of the simulation

are provided in Section 3.5.


43

3.1 Analog Modulation

As in digital modulation, optical fiber communication utilizes either an ILD or an

LED to create an optical carrier wave (CW) for a message signal. This CW may have its

amplitude, phase or frequency modulated with a message signal. Again, similarly to digital

modulation, the optical source may be directly modulated or externally modulated. The

benefits and drawbacks of each are discussed in this section.

Direct intensity modulation (D-IM) is inexpensive and easy to implement as no

electrical modulation or demodulation is required [1]. In a D-IM system, the waveform of

the optical power may be represented as [1]

𝑃𝑜𝑝𝑡 (𝑡) = 𝑃𝑖 (1 + 𝑚(𝑡)), (3.1)

where 𝑃𝑖 is the power of the unmodulated carrier and 𝑚(𝑡) is the intensity modulating signal

and is given as [1]

𝑚(𝑡) = 𝑚𝑎 𝑐𝑜𝑠𝜔𝑚 𝑡, (3.2)

where 𝜔𝑚 is the angular frequency of the modulating signal and 𝑚𝑎 is the modulation index

– the ratio of the peak power to the average power and [1]
𝑃𝑚𝑎𝑥 −𝑃𝑖
𝑚𝑎 = . (3.3)
𝑃𝑖

Combining (3.1) and (3.2) results in

𝑃𝑜𝑝𝑡 (𝑡) = 𝑃𝑖 (1 + 𝑚𝑎 𝑐𝑜𝑠𝜔𝑚 𝑡), (3.4)

which was used to produce the waveform shown in Fig. 3-1.


44

1
𝑃𝑜𝑝𝑡 (𝑡)
0.9

0.8

0.7
Optical Power (mW)

0.6

0.5

0.4

0.3

0.2

0.1

0
0 10 20 30 40 50 60 70 80 90 100
Time (ns)

Figure 3-1: D-IM of a 1550 nm optical carrier with a 10 MHz signal.

The SNR for a D-IM signal is given as [1]

𝑆 (𝑅𝑃0 )2 𝑃𝑀
(𝑁) = , (3.5)
𝑟𝑚𝑠,𝐷−𝐼𝑀 2𝐵𝑀 𝑁0

where 𝑅 is the responsivity of the photodiode, 𝑃0 is the average received optical power, 𝑃𝑀

is the total power of 𝑚(𝑡) from (3.2), 𝐵𝑀 is the one-sided bandwidth and 𝑁0 is defined as

[1]
4𝐾𝑇𝐵𝐹𝑛
𝑁0 = 𝑞(𝑅𝑃0 + 𝐼𝑑 )𝐹(𝑀) + , (3.6)
𝑀 2 𝑅𝐿

where the electric charge, 𝑞 = 1.602 × 10−19 𝐶, 𝐼𝑑 is the dark current, 𝐹(𝑀) is the excess

avalanche noise factor and 𝑀 is the multiplication factor - when an avalanche photodiode

(APD) is used. Finally, 𝐾 is Boltzmann’s constant, 𝑇 is the temperature in degrees Kelvin, 𝐵


45

is the post-detection bandwidth, 𝐹𝑛 is the amplifier noise figure and 𝑅𝐿 is the amplifier load

resistance. As previously stated, D-IM is inexpensive and easy to implement. However, in

order to fully realize the wideband capability of the optical fiber medium, a form of

multiplexing numerous channels in the electrical domain must be used [1].

In order to transmit multiple signals over the same channel, a subcarrier multiplexing

(SCM) technique is used. SCM provides the capability of multiplexing both multichannel

analog and digital signals within the same system [2]. To accomplish SCM, the baseband

channels are translated onto electrical carriers of differing frequencies prior to being

simultaneously transmitted as a frequency division multiplexed (FDM) signal. The baseband

channels may be used to modulate the electrical carrier by amplitude modulation (AM),

frequency modulation (FM) or phase modulation (PM) [1].

f1
MODULATED
m1(t) CARRIERS
f2

m2(t) COMPOSITE
RF
RF Power SIGNAL Laser Diode
Laser Driver
Combiner

fN

mN(t)

Figure 3-2: Example subcarrier multiplexing system.

An example of an SCM system is shown in Fig. 3-2. 𝑁 independent baseband

signals are provided as an input to the transmitter. The signals may be either analog or

digital. Each signal is mixed with a local oscillator (LO) using a frequency, 𝑓𝑁 . The LO
46

frequencies employed are generally within the range of 2 to 8 GHz and are called subcarriers

[2]. However, simulations have been performed that utilize carriers up to 25 GHz [3 – 4].

The combination of the resultant signals at the radio frequency (RF) power combiner

provides an FDM signal which is used to drive the laser.

It is important to note that a major advantage of SCM is the improvement in SNR

that may be obtained during subcarrier demodulation. It is shown in [1] that the use of

either AM-IM or double-sideband intensity modulation (DSB-IM) results in a 3 dB

degradation in SNR as compared to D-IM. This can be seen by comparing (3.5) with the

equation for DSB-IM which is given as [1]

𝑆 (𝑅𝑃0 )2 𝑃𝑀
(𝑁 ) = . (3.7)
𝑟𝑚𝑠,𝐷𝑆𝐵−𝐼𝑀 4𝐵𝑀 𝑁0

For this reason, these modulation formats are generally not used. However, by modulating

the subcarrier with FM-IM there is a significant improvement in the post-detection SNR as

compared to D-IM. This improvement is derived as follows.

The SNR for an FM-IM carrier as given as [1]

𝑆 3𝐷𝑓2 𝑃𝑀 (𝑅𝑃0 )2
(𝑁) = . (3.8)
𝑟𝑚𝑠,𝐹𝑀−𝐼𝑀 4𝐵𝑀 𝑁0

where 𝐷𝑓 is the peak frequency deviation ratio and is expressed as [1]

𝑝𝑒𝑎𝑘 𝑓𝑟𝑒𝑞𝑢𝑒𝑛𝑐𝑦 𝑑𝑒𝑣𝑖𝑎𝑡𝑖𝑜𝑛 𝑓


𝐷𝑓 = = 𝐵𝑑 . (3.9)
𝑏𝑎𝑛𝑑𝑤𝑖𝑑𝑡ℎ 𝑜𝑓 𝑚(𝑡) 𝑀

The improvement of SNR of FM-IM as compared to D-IM can be expressed by dividing

(3.5) by (3.8) which results in

𝑆 (3𝐷𝑓2 𝑃𝑀 (𝑅𝑃0 )2 )/(4𝐵𝑀 𝑁0 )


(𝑁) =
𝑖𝑚𝑝𝑟𝑜𝑣𝑒𝑚𝑒𝑛𝑡 (𝑃𝑀 (𝑅𝑃0 )2 )/(2𝐵𝑀 𝑁0 )

3𝐷𝑓2
= . (3.10)
2
47

Similarly, PM-IM provides an improvement over D-IM. However, it is not as great an

improvement as FM-IM [1]. This chapter focuses on the simulation of a DSB subcarrier

modulation in an analog link, thus providing a basis for the study of FM subcarrier

modulation in chapter 5.

3.2 Overview of an Analog Link

The OptiSystem model of a simplex point-to-point analog link is shown in Fig. 3-3.

For the sake of simplicity, a single LO of 10 GHz is amplitude modulated by a digital signal

in the electrical domain. The resultant signal is an RF modulated carrier which is used to

Figure 3-3: OptiSystem model of a simplex point-to-point analog link.

modulate a CW laser via the MZM. In a standard SCM system, several RF carriers would be

power combined prior to modulation of the optical CW. This is shown in the simulations of
48

a CWR analog link in Section 3.4. The composite optical signal is then launched into 5 km

of single-mode optical fiber where it is then detected by a pin photodiode at the distant end.

The original signal is then demodulated from the carrier at the distant end in the electrical

domain. The signal is then fed through a low-pass Bessel filter to remove high frequency

noise. The filtered signal is finally sent to a BER analyzer for comparative analysis.

3.3 Proposed Implementation

Fig. 3-4 provides a block diagram of an SCM analog link with CWR. In this

implementation, two message signals, 𝑚1 (𝑡) and 𝑚2 (𝑡) are each modulated with a different

subcarrier frequency, 𝑓1 and 𝑓2 respectively. An RF power combiner creates a composite

frequency division multiplexed (FDM) signal which is then used to drive the input of an

MZM. The output of the MZM is a modulated optical carrier wave which was provided as

an input by the CW laser. This optical carrier is then sent from System 2 to System 1.

At System 1, the optical power is split using a beamsplitter. The first path reaches

the detector where it is subsequently power split in the electrical domain and demodulated

providing the original signals. The second path provides CWR by sending the optical signal

through the MZM of System 1, where it is then modulated with a new FDM signal. This

signal is comprised of two new LO frequencies, 𝑓3 and 𝑓4 that are modulated with message

signals 𝑚3 (𝑡) and 𝑚4 (𝑡).


49

SYSTEM 2 SYSTEM 1
f1 f1

Detector
m1(t) m1(t)

f2 RF Power Optical f2
Drive
Combiner Power Splitter
Electronics Detector
m2(t) m2(t)

f3

Continuous Wave m3(t)


Laser MZM MZM
RF Power f4
Combiner
m4(t)

f3

m3(t) Detector

f4
RF Power
Splitter
m4(t) Detector

Figure 3-4: Proposed implementation of an analog SCM link with CWR.

The output of the MZM in System 1 is then sent back to System 2 where it is detected and

demodulated in the same manner as was previously described for System 1.

3.4 Link Simulation

The system defined in Section 3.3 is modeled and simulated in this section. The

purpose of these simulations is to examine the ability to use CWR on an analog link. While

the model is an SCM analog link, digital signals are utilized to modulate the RF carriers. This

is done in order to provide simple analysis of the effects of CWR on the link. This section

also provides a description of the model components shown in Fig. 3-5 and a discussion of

their transfer functions.


50

Figure 3-5: OptiSystem model of an analog SCM link with CWR, top half.
51

Figure 3-6: OptiSystem model of an analog SCM link with CWR, bottom half.
52

SCM is initially achieved in system 2 (Fig. 3-4) by the seven components shown in

the top-left of Fig. 3-5. The two PRBS generators each produce a pseudo-random number

of binary bits as defined by (2.14). The generators are set to a Mark Probability mode with

the probability of ones in the sequence set to 0.5. In this simulation, the time window is

defined as 𝑇𝑤 = 8.192 𝜇𝑠 and the bit rate is set to, 𝐵𝑟 = 250 𝑀𝑏/𝑠. The number of bits

produced by each generator is

𝑁 = 𝑇𝑤 𝐵𝑟 = 2048 𝑏𝑖𝑡𝑠. (3.11)

The output of each generator is then fed to the NRZ pulse generators which are used to

create pulses with exponential edges as defined by (2.15). At this point, a new component is

introduced, the electrical amplitude modulator. This component simulates an analog

amplitude modulator. The output signal is modulated according to [5]

𝑣𝑜𝑢𝑡 (𝑡) = 𝐺𝑣𝑖𝑛 (𝑡) cos(2𝜋𝑓𝑐 𝑡 + 𝜙𝑐 ) + 𝑏, (3.12)

where 𝑣𝑖𝑛 is the input electrical signal, 𝐺 is the parameter gain, 𝑏 is the bias, 𝑓𝑐 is the carrier

frequency and 𝜙𝑐 is the phase of the carrier. The parameters associated with this simulation

for each of the electrical amplitude modulators are shown in Table 3-1 and the resultant

signals are shown in Fig. 3-7.

Table 3-1: Electrical amplitude modulator parameter settings.

Electrical
Gain, 𝑮 Bias, 𝒃 Phase, 𝝓𝒄 Frequency, 𝒇𝒄
Amplitude
(V) (°) (GHz)
Modulator
0 1 0 0 10
1 1 0 0 13
2 1 0 0 17
3 1 0 0 21
53

The resultant signals are provided as an input to the 2x1 power combiner block. The signals

are combined evenly into a single output port using the s-parameters [5]

𝑠01 1𝑗 = [−3𝑑𝐵 − 𝛼]∠0°, (3.13)

where 𝛼 is the insertion loss, 𝑁 is the number of input ports and 𝑗 is the input port index.

The combiner is assumed to have an ideal insertion loss of 0 dB. As previously discussed,

the combination of the two AM signals results in a composite RF signal, shown in Fig. 3-8,

which is used to modulate the optical carrier wave.


54

(a)

(b)

Figure 3-7: AM signals of system 2, (a) 10 GHz carrier and (b) 13 GHz carrier.

Modulation of the optical carrier wave is performed in a similar manner as described

in Section 2.6. The 1x2 fork, electrical gain, and CW laser blocks used in this simulation

have the same characteristics as previously described. The MZM, however, is operated at a
55

negative quadrature operating point as opposed to the null operating point used in

simulations of a digital link. As recalled from (2.20), in order to operate in negative

Figure 3-8: Composite RF signal of system 2

𝑣𝜋𝐷𝐶 𝑣𝜋𝑅𝐹
quadrature with 𝑣𝜋𝐷𝐶 set to 4V, (𝑣1 − 𝑣2 ) = = = 2𝑉 [5]. Therefore, the values
2 2

used for this simulation are 𝑣1 = 1 and 𝑣2 = −1. The output signal from the MZM block

is the SCM optical carrier shown in Fig. 3-9. It can be seen in the figure that this signal

consists of the carrier at 193.1 THz (1550 nm) as well as the two AM carriers spaced 10

GHz and 13 GHz from the optical carrier.


56

Figure 3-9: SCM optical signal of system 2

The SCM signal is then launched down 1 km of single-mode optical fiber before

reaching the power splitter (Fig. 3-5). The power splitter serves a dual purpose. First, it

splits power between the two optical receivers of system 1 (Fig. 3-4). Second, it provides

power to the MZM of system 1 for the purpose of CWR. When received at system 1, the

two optical signals are detected by pin photodiodes. The photodiode blocks have the same

parameters and transfer functions as described in Section 2.6. The electrical signal created is

then provided as an input to the electrical amplitude demodulator blocks.

The two demodulator blocks function as coherent amplitude demodulators and the

RF signal is demodulated according to [5]


57

𝑣𝑜𝑢𝑡 (𝑡) = [𝐺𝑣𝑖𝑛 (𝑡) cos(2𝜋𝑓𝑐 𝑡 + 𝜙𝑐 )] × ℎ𝑙𝑜𝑤 (𝑡), (3.14)

where ℎ𝑙𝑜𝑤 is the time response of the low pass filter. The low pass filter associated with

the demodulator is defined to be a cosine roll off filter with the transfer function [5]

𝛼, (|𝑓| < 𝑓1 )
|𝑓𝑐 |−𝑓1
𝐻(𝑓) = √0.5𝛼 2 [1 + cos ( × 𝜋)] , (𝑓1 ≤ |𝑓| < 𝑓2 ) (3.15)
𝑟 𝑝 𝛥𝑓𝐹𝑊𝐻𝑀

{ 0, (𝑓2 ≤ |𝑓|)

where 𝛼 is the insertion loss, 𝑓𝑐 is the filter cutoff frequency and 𝑟𝑝 is the parameter ‘roll off

factor’. The parameters 𝑓1 and 𝑓2 are defined as [5]

𝑓1 = 1 − 𝑟𝑝 𝑓𝑐 , (0 ≤ 𝑟𝑝 < 1)
. (3.16)
𝑓2 = 1 + 𝑟𝑝 𝑓𝑐 , (0 ≤ 𝑟𝑝 < 1)

Finally, the parameters for simulation of the amplitude demodulator are set to match the

values of the transmitters shown in Table 3-1. Additional parameters for the demodulator’s

low pass filter are 𝛼 = 0 𝑑𝐵, 𝑟𝑝 = 0.5 and 𝑓𝑐 = 0.75 × 250 𝑀𝑏/𝑠. The final signals are

sent to their respective BER analyzers via low pass Bessel filters in the same manner as was

shown in Section 2.6.

Returning to the MZM of system 1 labeled Dual Port MZ Modulator Measured_,1

shown in Fig. 3-6, the third branch of the power splitter (Fig. 3-5) provides the input to this

block so that CWR can be performed. The seven components in the bottom left of the

figure provide the same functionality for system 1 as those that were previously discussed for

system 2. In this case, the electrical amplitude modulators create two new carriers at 17

GHz and 21 GHz shown in Fig. 3-10. These signals are then combined to produce the

composite RF signal of system 1 that is used to modulate the reutilized carrier wave as seen

in Fig. 3-11.
58

(a)

(b)

Figure 3-10: AM signals of system 1, (a) 17 GHz carrier and (b) 21 GHz carrier.
59

The modulation of the optical carrier is performed using the same method and system

blocks that were previously used for system 2. This produces a CWR signal composed of

Fig 3-11: Composite RF signal of system 1.

the originally modulated RF carriers produced by system 2 and the new RF carriers

produced by system 1, Fig. 3-12. The newly created SCM optical signal is then launched

down 1 km of single-mode optical fiber back to system 2. At system 2, the 1x2 power

splitter, described by (2.12), equally distributes the power to two pin photodetectors. As

with the simulation in chapter 2, the power attenuation, 𝛼, has been set to 0.
60

The two signals produced by the previous operation are detected at system 2 and

demodulated using the same method as described for system 1.

Figure 3-12: SCM optical signal of system 1.

3.5 Numerical Results

Two simulations of the aforementioned system have been performed. In the first

simulation, the power output of the CW laser is varied from -3 dBm to 6 dBm. The effect

of increasing the power is analyzed by examining the eye diagram of each of the digital

signals at the distant end. (As a reminder, analog modulation was used on the optical carrier

wave while performing SCM. However, the original information bearing signals used to
61

modulate the RF carriers were digital.) The improvement of the eye opening and respective

BERs can then be reviewed. The bit rate for each of the digital signals is held at 250 Mbps.

In Figs. 3-13 and 3-14, the eye diagrams for each of our signals are shown. Each

column of the figure represents a different RF modulated carrier while each row represents

an increase in the optical carrier power output. It is evident from the figures that the signals

received at system 1 (Fig. 3-13) provide a much cleaner signal than those that are from the

remodulated carrier wave (the right two columns). It is shown that, from system 1 to system

2, the noise margin is decreased, the sensitivity is decreased and the timing jitter is increased.

The noise margin is calculated using [1]


𝑣
𝑁𝑜𝑖𝑠𝑒 𝑚𝑎𝑟𝑔𝑖𝑛 = 𝑣1 × 100, (3.17)
2

where 𝑣1 is the difference between the threshold and the lowest part of the upper eye and 𝑣2

is the difference between the threshold and the maximum signal voltage. The timing jitter is

calculated using [1]

Δ
𝑇𝑖𝑚𝑖𝑛𝑔 𝑗𝑖𝑡𝑡𝑒𝑟 = 𝑇 𝑡 × 100, (3.18)
𝑏

where Δ𝑡 is the length of time of the distortion at the zero crossing. The rise time is defined

as the difference between the points where the rising edge of the signal reaches 10 percent of

its final amplitude to 90 percent of its final amplitude. The results of these eye pattern

parameters have been tabulated in Table 3-2. It is noted that, despite the decreased

performance along the return path, the eye opening is still well-defined and the original

signal can still be recovered.


62

(a)

(b)

(c)

(d)

Figure 3-13: Eye diagrams for simulation 1. The subcarriers used for each column from left
to right are 10 GHz and 13GHz. The optical carrier power for each row is (a) -3 dBm,
(b) 0dBm, (c) 3 dBm and (d) 6 dBm.
63

(a)

(b)

(c)

(d)

Figure 3-14: Eye diagrams for simulation 1. The subcarriers used for each column from left
to right are 17GHz and 21 GHz. The optical carrier power for each row is (a) -3 dBm,
(b) 0dBm, (c) 3 dBm and (d) 6 dBm.
64

Table 3-2: Simulation 1 eye pattern characteristics

CW Subcarrier at 10 GHz Subcarrier at 13 GHz Subcarrier at 17 GHz Subcarrier at 21 GHz


Noise Timing Rise Noise Timing Rise Noise Timing Rise Noise Timing Rise
Power Margin Jitter Time Margin Jitter Time Margin Jitter Time Margin Jitter Time
(dBm) (%) (%) (ns) (%) (%) (ns) (%) (%) (ns) (%) (%) (ns)
-3 71.6 10.0 2.0 81.8 12.0 2.0 15.6 35.0 1.7 3.75 25.0 1.9
0 66.7 9.0 2.0 58.3 12.0 1.9 18.2 37.5 1.8 39.8 23.0 2.0
3 71.1 8.0 2.0 65.2 12.0 2.0 23.1 28.0 1.8 42.1 24.0 1.7
6 71.1 8.0 2.0 63.2 12.0 2.0 18.4 32.0 1.7 44.4 24.0 1.7

The maximum eye opening factor is another parameter that is used as an observation of

signal quality. It is shown in Fig. 3-15 that, while the eye opening varies slightly as the power

of the carrier wave is increased, the two factors that play the most important role in eye

height are the frequency of the subcarrier and the application of CWR.

This fact is further emphasized by examining the BER of each of the signals as the

power of the optical carrier is increased as shown in Fig. 3-16. There is a dramatic difference

between the BERs of the 10 GHz and 13 GHz subcarriers despite the fact that they are used

to modulate the initial, undisturbed, optical carrier wave. This may be due to undesirable

intermodulation products which occur when multiple carrier frequencies pass through a

nonlinear device, such as a laser diode [1]. While a number of papers [6 - 8] have examined

these effects, the analyzation of these effects is outside of the scope of this thesis. The other

difference that is observed between the subcarrier BERs is that of the signals that were used

in CWR compared to those that were not (i.e. – the 10 and 13 GHz subcarriers versus the 17

and 21 GHz subcarriers). Furthermore, it is shown in Fig. 3-17 that when examining each

of the signals individually, the BER improves as the power is increased.


65

10 GHz Subcarrier

13 GHz Subcarrier

21 GHz Subcarrier

17 GHz Subcarrier

Figure 3-15: Variation in eye opening as power is increased.


66

17 GHz Subcarrier

21 GHz Subcarrier

13 GHz Subcarrier

10 GHz Subcarrier

Figure 3-16: Minimum BER of all signals as power is increased.


67

Figure 3-17: Minimum BER of each signal as power is increased; the subcarrier used for
each signal is (top left) 10 GHz, (top right) 13 GHz, (bottom left) 17 GHz and,
(bottom right) 21 GHz.

For the second simulation, the optical carrier wave power is maintained at a constant

6 dBm while the bit rate is varied from 100 Mb/s to 1 Gb/s. It is shown in Fig. 3-18 that,

until approximately 700 Mb/s, the BER for each of the signals remains relatively constant.

For the CWR signals, the BER begins to increase sooner than it does for the non-CWR

signals.
68

Figure 3-18: Minimum BER of each signal as bit rate is increased; the subcarrier used for
each signal is (top left) 10 GHz, (top right) 13 GHz, (bottom left) 17 GHz and,
(bottom right) 21 GHz.

These simulations show that using CWR on an analog link creates a viable alternative

to the traditional means of optical fiber communication over an analog link. As with the

digital link, the BERs achieved are within an acceptable level and can be maintained both

with varying optical power levels and with varying bit rates.
69

3.6 References

[1] J. Senior, “Optical Fiber Communications: Principles and Practice,” Third Edition,
Pearson Education, Harlow, England, 2009.
[2] G. Keiser, “Optical Fiber Communications,” Fourth Edition, McGraw-Hill, New
York, 2011.
[3] R. Aparna and S. Chandran, “Investigations on Spectral Efficiency of Optical
Communication System Using Wavelength Division and Sub Carrier Multiplexing,”
International Journal of Science and Research (IJSR), www.ijsr.net, vol. 4, iss. 2, pp. 166-169
February 2015.
[4] J. Johny and S. Shashidharan, “Design and simulation of a Radio Over Fiber system
and its performance analysis,” 2012 4th International Congress on Ultra Modern
Telecommunications and Control Systems and Workshops (ICUMT), St. Petersburg, pp. 636-
639, 2012.
[5] OptiSystem [computer program], 14.0.0.483, Ontario, Canada, Optiwave Systems,
Inc., 2015.
[6] Y. Cui, Y. Dai, F. Yin, J. Dai, K. Xu, J. Li and J. Lin, “Intermodulation distortion
suppression for intensity-modulated analog fiber-optic link incorporating optical
carrier band processing,” Optics Express, pp. 23433-23440, 2013.
[7] T. Cho and I. Kwon, “Performance optimization of radio-on-fiber systems
employing erbium doped fiber amplifier for optical single sideband signals
considering intermodulation distortion,” Photonic Network Communications, pp. 73-78,
2013.
[8] M. Madero-Ayora, M. Allegue-Martinez, C. Crespo-Cadenas, J. Reina-Tosina and J.
Navarro-Lazaro, “Experimental study of two-tone intermodulation products in a
communications modulator,” Microwave and Optical Technology Letters, pp. 58-61, 2011.
70

CHAPTER 4

Carrier Wave Attenuation and Dispersion

This chapter examines the performance of the optical fibers as a transmission

medium and the Mach-Zehnder Modulator (MZM) as the primary component used to re-

modulate a carrier wave. The primary focus is on the transmission characteristics of

attenuation and dispersion of the carrier wave. Signal attenuation in an optical fiber link is a

vital characteristic, in that its value largely dictates the maximum achievable distance from

transmitter to receiver without the use of a repeater [1]. Furthermore, the potential

bandwidth that is espoused as one of the greatest virtues of optical fiber communication, is

limited by the signal dispersion within the fiber [2].

Throughout this chapter the effects of both attenuation and dispersion on the optical

carrier wave are examined. The characteristic of attenuation and its basic mechanisms in a

fiber are discussed in Section 4.1. Section 4.2 provides an overview of signal dispersion

along with its constituent contributing factors. In Section 4.3, attenuation and dispersion are

simulated in the optical fiber from the generation of the carrier wave to the initial receiver.

Section 4.4 examines the combined effects of attenuation and dispersion and how it relates

to the signal as seen at the initial receiver. Finally, in Section 4.5, attenuation and dispersion

of the reutilized carrier wave are simulated.


71

4.1 Attenuation

Attenuation, also known as transmission loss, fiber loss or signal loss [1 – 2], was a

gating factor in bringing about wide-scale acceptance of optical fiber in telecommunications

[2]. Prior to the 1970’s, the prospect of utilizing fiber optics for communications was not

feasible due to the fact that a dielectric waveguide made of glass could only support optical

transmission for a few meters [1]. However, in 1970 a breakthrough occurred which allowed

for the propagation of light waves with an attenuation of 20 dB km-1 [3]. It was not long

until silica-based glass fibers with signal losses of less than 0.2 dB km-1 were reported [4],

thereby bringing the signal attenuation to far lower levels than that of their competing

metallic conductors which were operating at less than 5 dB km-1 [1].

Signal attenuation is generally expressed in the logarithmic unit of the decibel – a

comparison of two power levels. For a given wavelength, this comparison is defined as the

ratio between input optical power, 𝑃𝑖 , to the optical power received, 𝑃𝑜 , as shown in (4.1):

𝑃
𝛼 = 10𝑙𝑜𝑔10 𝑃 𝑖 . (4.1)
𝑜

Furthermore, as it pertains to an optical fiber optic medium, this ratio is often defined in

terms of decibels per unit length, following the equation


10 𝑃
𝛼𝑑𝐵 = 𝑙𝑜𝑔10 𝑃 𝑖 , (4.2)
𝐿 𝑜

where 𝛼𝑑𝐵 is the attenuation per unit length in decibels and 𝐿 is the fiber length in km [1].

Note that this parameter is generally referred to as the fiber loss and it is a function of the

wavelength. Moreover, the attenuation mechanisms associated with optical fiber

communications are due to the absorption, scattering and radiative losses of optical energy

within the optical fiber medium [2].


72

4.1.1 Absorption

One of the attenuation mechanisms, absorption, is caused by three different sub-

effects. First, there is absorption by atomic defects in the glass composition. Second, there

is extrinsic absorption caused by impurity atoms in the glass material. Third, intrinsic

absorption is caused by the basic constituent atoms of the fiber material. Each of these loss

mechanisms are discussed in detail to gain an understanding of their limiting factors.

Atomic defects occur because of the imperfections in the atomic structure of the

fiber material. These imperfections include: missing molecules, high-density clusters of atom

groups, and oxygen defects [2]. While absorption losses arising from this particular effect

are generally negligible, these losses can be significant if the fiber is exposed to ionizing

radiation [5 – 8]. Exposing fiber to this type of radiation may happen in a nuclear reactor,

medical radiation therapy, space missions or accelerator instrumentation [2]. When optical

fiber is exposed to any of the previously described types of ionizing radiation, the associated

attenuation will increase. This is due to the fact that the number of atomic defects increases

and each of these defects contribute to the absorption of optical energy.

Impurity atoms in the fiber material are the largest contributing factor to absorption

and are known as extrinsic absorption [1 – 2]. These impurities arise because of the

conventional melting techniques which leave behind various transition metals as well as

water ions (OH-), which are bonded into the glass structure. Many of these impurities are

shown in Table 4-1. Extrinsic absorption can occur either because of the electron

transitions between energy levels within the impurities or because of the charge transitions

between the ions [2].


73

Table 4-1: Absorption loss in silica glass due to common impurities.

Impurity Absorption Peak (nm) Loss from 1 ppb (dB km-1)


Cr3+ 625 1.6
Cr2+ 625 1.6
C2+ 685 0.1
Cu2+ 850 1.1
Fe2+ 1100 0.68
Fe3+ 400 0.15
Mn3+ 460 0.2
Ni2+ 650 0.1
V4+ 725 2.7
OH- 950 1.0
OH- 1240 2.0
OH- 1380 4.0

It is also interesting to note that, while OH- was historically a large contributor around 1350

nm, modern fabrication techniques have lessened the number of these impurities found in

silica fiber [2]. This is evidenced by the measured fiber-optics installation graphs in Fig. 4-1.

It is shown that the attenuation due to OH- impurities results in a peak at approximately

1380 nm and, that in newer installations, fiber manufacturers have learned to reduce these

levels of impurities [9].

The final mechanism of absorption is known as intrinsic absorption. Intrinsic

absorption is due to the basic fiber material, i.e., pure silicate class. Furthermore, this type of

absorption sets the lower bounds on absorption effects. The mechanism results from

electronic absorption bands in the ultraviolet region and from atomic vibration bands in the

near-infrared region [2]. An empirical formula for the effect of ultraviolet absorption has

been given as [2]


4.63
1.542𝑥
𝛼𝑢𝑣 = 46.6𝑥+60 𝑒 𝜆 , (4.3)
74

where 𝑥 is the mole fraction of Germanium dioxide (GeO2) and 𝜆 is the wavelength in µm.

This function has been plotted as UV Absorption in Fig. 4-1 under the assumption that a

fiber has been doped with 6% GeO2. An empirical formula for infrared absorption has also

been given as [2]


−48.48
𝛼𝑖𝑟 = 7.81 × 1011 𝑒 𝜆 . (4.4)

The attenuation due to infrared absorption has also been plotted in Fig. 4-1 for comparison.

1.0

0.9
Fiber Installed pre 1990
0.8

0.7
Attenuation (dB/km)

0.6

0.5
Fiber Installed 2003
0.4

0.3

0.2 Rayleigh Scattering

0.1 UV Absorption
IR Absorption

0.0
1200 1250 1300 1350 1400 1450 1500 1550 1600 1650
Wavelength (nm)

Figure 4-1: Attenuation versus wavelength showing absorption effects. Note that
experimental data graphs from fiber installations are based on measurements from [9].

In this figure, it is shown that the effects of intrinsic absorption are magnitudes lower than

the actual measured values over the given range of wavelengths. However, as the graph
75

approaches 1650 nm, it can be seen that the effect of infrared absorption trends in such a

way that it becomes a mitigating factor at longer wavelengths.

4.1.2 Scattering

Attenuation due to scattering arises from several factors, such as: microscopic

variations in material density, compositional fluctuations and structural inhomogeneities [2].

If a structural inhomogeneity was to occur that was comparable in size with the guided

wavelength, then Mie scattering may occur. Mie scattering can result in significant losses.

However, the fabrication process has improved so that structural inhomogeneities have been

controlled [1].

The dominant loss mechanism for optical fiber is due to Rayleigh scattering [1]. The

natural changes in molecular density and compositional fluctuations due to the fact that glass

is made up of several oxides (SiO2, GeO2, P2O5), results in refractive index variations [2]. In

single-component glass, these fluctuations can be approximated by [1]

8𝜋 3
𝛾𝑅 = 𝑛8 𝑝2 𝛽𝑐 𝑘𝐵 𝑇𝑓 , (4.5)
3𝜆4

where 𝛾𝑅 is the Rayleigh scattering coefficient, 𝜆 is the wavelength, 𝑛 is the index of

refraction, 𝑝 is the average photoelastic coefficient, 𝛽𝑐 is the isothermal compressibility at

the fictive temperature 𝑇𝑓 , and 𝑘𝐵 is Boltzmann’s constant. This function has been graphed

in Fig. 4-1 using the following equation to convert to the attenuation in decibels

1
𝛼𝑑𝐵 = 10 log ( 3 ). (4.6)
𝑒 −𝛾𝑅 ×10

In the graph, it was assumed that 𝑛 = 1.46, 𝑝 = 0.286, 𝛽𝑐 = 7 × 10−11 𝑚2 𝑛−1 and

𝑇𝑓 = 1400𝐾 [10]. It is shown in Fig. 4-1 that, for lower wavelengths, the dominant
76

absorption characteristic is Rayleigh scattering and, as the wavelength increases, the

dominant factor becomes intrinsic absorption.

4.2 Dispersion

While attenuation describes the weakening of an optical signal due to various

mechanisms, dispersion describes the distortion of an optical signal. This distortion causes

pulses to broaden as they travel down a length of optical fiber until they begin to overlap as

seen in Fig. 4-2. Overlapping of optical pulses is known as intersymbol interference (ISI)

and it results in an increasing number of errors at the receiving end of a system. For this

reason, signal dispersion exerts a limit on the maximum possible bandwidth attainable [1].

Figure 4-2: Dispersion and intersymbol interference, (a) input signal, (b) signal begins to
overlap with distance, (c) as distance increases overlap becomes problematic.
77

For example, in order to refrain from having pulses overlap, the digital bit rate, 𝑇𝑏 , must be

less than the reciprocal of the broadened pulse duration [1]


1
𝑇𝑏 ≤ 2𝜏, (4.7)

where 𝜏 is the input pulse duration. Therefore, (4.7) provides a conservative estimate of the

maximum achievable bit rate. However, in order to obtain a more accurate estimate, the

pulses must be considered as a Gaussian shape with an rms width of 𝜎 [1]. In this case, the

maximum achievable bit rate is derived in [1] as


0.2
𝑇𝑏 ≤ . (4.8)
𝜎

As discussed in Chapter 2, the conversion of bit rate to bandwidth is dependent on the

digital coding format used. For example, if NRZ is used, the binary one period is held high

for the entirety of 𝜏. This means that there are two bit periods for every one wavelength or

two bits per second per Hertz [1]. So, in the case of NRZ, the maximum bandwidth is
𝑇𝑏
𝐵= . (4.9)
2

Dispersion is the result of several factors, including: intermodal delay, intramodal

dispersion, polarization-mode dispersion and higher-order dispersion. Intermodal

dispersion occurs only in multimode fibers and is the result of each mode within the fiber

having a different group velocity at a given frequency. Since this thesis focuses on the

effects within single-mode fiber, intermodal delay is not covered in detail.


78

4.2.1 Intramodal Dispersion

Intramodal dispersion refers to the pulse spreading that takes place within a single

mode. Intramodal dispersion is also referred to as chromatic dispersion or group velocity

dispersion (GVD). This type of dispersion can occur in all categories of optical fiber and it

results from the finite spectral linewidth of the optical source [1]. An optical source does

not emit a single frequency but rather a band of frequencies. Due to this fact, there is a

propagation delay difference between the wavelengths over which the source emits. The

two main causes of intramodal dispersion are material dispersion and waveguide dispersion.

Material dispersion describes the broadening associated with variations in the

refractive index of the core material as a function of wavelength [2]. This is the same effect

by which a prism spreads out a spectrum and gives rise to the term chromatic dispersion. As

a signal travels through an optical waveguide, it is assumed that each spectral component

travels independently and a group delay per unit length is given by [2]

𝜏𝑔 1 1 𝑑𝛽 𝜆2 𝑑𝛽
= 𝑉 = 𝑐 𝑑𝑘 = − 2𝜋𝑐 𝑑𝜆 , (4.10)
𝐿 𝑔

where 𝐿 is the distance traveled by the pulse, 𝛽 is the propagation constant, 𝑘 = 2𝜋/𝜆 and

the group velocity is the velocity at which energy in a pulse travels along a fiber given by

𝑑𝛽 −1
𝑉𝑔 = 𝑐 (𝑑𝑘 ) . (4.11)

Furthermore, the pulse spreading associated with a spectral width characterized by its rms

value, 𝜎𝜆 , is approximated by [2]

𝑑𝜏𝑔
𝜎𝑔 = | 𝑑𝜆 | 𝜎𝜆 , (4.12)

where 𝜏𝑔 is the group delay.


79

To calculate material dispersion, a plane-wave is considered in which it is

propagating in an infinite dielectric medium with a refractive index of 𝑛(𝜆). In this case, the

propagation constant is given by [2]

2𝜋𝑛(𝜆)
𝛽= . (4.13)
𝜆

Substituting (4.13) into (4.10) yields the following equation for group delay resulting from

material dispersion

𝐿 𝑑𝑛
𝜏𝑚𝑎𝑡 = (𝑛 − 𝜆 ). (4.14)
𝑐 𝑑𝜆

Finally, by differentiating the group delay of (4.14) with respect to wavelength and

substituting it into (4.12) the following expression is obtained,

𝑑𝜏𝑚𝑎𝑡 𝜎𝜆 𝐿 𝑑2 𝑛
𝜎𝑚𝑎𝑡 = | | 𝜎𝜆 = |𝜆 𝑑𝜆2 | = 𝜎𝜆 𝐿|𝐷𝑚𝑎𝑡 (𝜆)|, (4.15)
𝑑𝜆 𝑐

where 𝐷𝑚𝑎𝑡 (𝜆) is the material dispersion [2].

The second cause of intramodal dispersion, waveguide dispersion, is a result of a

portion of the optical power propagating along the cladding rather than the core of the fiber.

This occurs because shorter wavelengths are confined to the core, while longer wavelengths

can propagate in the cladding [2]. While waveguide dispersion is minimal in multimode fiber

and can generally be ignored, its effect is significant in single-mode fibers. Waveguide

dispersion can be approximated based on the assumption that the refractive index of the

material is independent of the wavelength. In order to determine the group delay, the results

must be independent of the fiber configuration and should be expressed in terms of a

normalized propagation constant defined as [2]


𝛽
−𝑛2
𝑏 ≈ 𝑛𝑘 −𝑛 , (4.16)
1 2
80

where it is assumed that there are small values of the index difference. Solving (4.16) for 𝛽

and substituting into (4.10) gives the group delay that arises from waveguide dispersion as

𝐿 𝑑𝛽
𝜏𝑤𝑔 = 𝑐 𝑑𝑘 . (4.17)

The modal propagation constant 𝛽 can be expressed in terms of the normalized frequency,

𝜈𝑛 , as [2]
1
𝜈𝑛 = 𝑘𝑎(𝑛12 − 𝑛22 )2 ≈ 𝑘𝑎𝑛1 √2Δ , (4.18)

which is valid for small changes in the refractive index of a material where
𝑛1 −𝑛2
Δ= . (4.19)
𝑛1

Therefore, writing (4.17) in terms of 𝜈 results in

𝐿 𝑑(𝜈𝑛 𝑏)
𝜏𝑤𝑔 = 𝑐 [𝑛2 + 𝑛2 Δ ], (4.20)
𝑑𝜈𝑛

The first term in (4.20) is constant and the second term is the group delay from waveguide

dispersion. This factor can be expressed as [2]

𝑑(𝜈𝑛 𝑏) 2𝐽𝜈2 (𝑢𝑎)


= 𝑏 [1 − 𝐽 ], (4.21)
𝑑𝜈𝑛 𝜈+1 (𝑢𝑎)𝐽𝜈−1 (𝑢𝑎)

where 𝑎 is the fiber radius, 𝐽𝜈 is a Bessel function and 𝑢 is defined as [2]

2𝜋𝑛1 2
𝑢=( ) − 𝛽2 . (4.22)
𝜆

As previously mentioned, in single-mode fibers waveguide dispersion is important as

it can be of the same order of magnitude as that of material dispersion. By taking the

derivative of the group delay, 𝜏𝑤𝑔 , with respect to wavelength, the pulse spread over a

distribution of wavelengths, 𝜎𝜆 is obtained [2]:


81

𝑑𝜏𝑤𝑔 𝜈𝑛 𝑑𝜏𝑤𝑔
𝜎𝑤𝑔 = | | 𝜎𝜆 = 𝐿|𝐷𝑤𝑔 (𝜆)|𝜎𝜆 = | | 𝜎𝜆
𝑑𝜆 𝜆 𝑑𝜈𝑛

𝑛2 𝐿Δ𝜎𝜆 𝑑2 (𝜈𝑛 𝑏)
= 𝜈𝑛 . (4.23)
𝑐𝜆 𝑑𝜈𝑛2

Therefore, solving for the waveguide dispersion results in

𝑛2 Δ 𝑑2 (𝜈𝑛 𝑏)
|𝐷𝑤𝑔 (𝜆)| = − [𝜈𝑛 ]. (4.24)
𝑐𝜆 𝑑𝜈𝑛2

Finally, it has been shown in [11] that a good estimate of the total intramodal dispersion can

be calculated by summing the waveguide dispersion and the material dispersion.

4.2.2 Polarization-Mode Dispersion

Polarization-mode dispersion (PMD) is caused by the fact that optical energy at a

particular wavelength occupies two orthogonal polarization states (modes) in a single-mode

fiber. Each of these modes will encounter a slightly different refractive index as they

propagate through the fiber due to imperfections – this is known as birefringence. The two

different propagation times between the modes causes pulse spreading. Given two group

velocities between the modes, 𝑉𝑔𝑥 and 𝑉𝑔𝑦 , the differential time delay of a distance, 𝐿, is

given as [2]

𝐿 𝐿
Δ𝜏𝑃𝑀𝐷 = |𝑉 − 𝑉 |. (4.25)
𝑔𝑥 𝑔𝑦

It is important to note that PMD varies randomly along a fiber due to the birefringence

effects that cause variations with applied stress and temperature. As such, statistical

estimations have been used to derive the following equation for PMD:

√𝐿
𝐷𝑃𝑀𝐷 ≈ Δ𝜏 , (4.26)
𝑃𝑀𝐷
82

where 𝐷𝑃𝑀𝐷 is measured in 𝑝𝑠/√𝑘𝑚. In order to limit the possibility of errors due to

PMD, a standard limit on the maximum allowable value of Δ𝜏𝑃𝑀𝐷 ranges between 10 and 20

percent of a bit duration [2].

4.3 Attenuation and Dispersion Simulation – Link 1

The system that was simulated in Chapter 2 is considered in reference to the

attenuation and dispersion simulations in this section. This original model can be seen in

Fig. 2-9 and is shown here for convenience. The model produces a pseudo-random binary

sequence (PRBS) with a bit rate of 10 Gb/s. In order to simulate the effects of attenuation

and dispersion within this system, a single pulse is considered in this section. The effects of

attenuation, GVD and PMD are simulated individually from the original transmission of the

CW laser up to the power splitter of System 1. Then, all effects are simulated simultaneously

in order to compare the individual effects to the final result.

Since many optical fibers exhibit pulse outputs with a temporal variation that is

closely approximated by a Gaussian distribution [1], i.e.

𝑡2
1 −( )
𝑃𝑜 (𝑡) = 𝑒 2𝜎2 , (4.27)
√2𝜋

the model of Fig. 4-3 is modified to produce Gaussian pulse shapes by adding a low pass

Gaussian filter to the transmitter as shown in Fig. 4-4. Furthermore, the PRBS block is

removed and a user-defined bit sequence is added to produce three pulses.


83

Figure 4-3: OptiSystem model of digital link with CWR.


84

These pulses are fed to the NRZ pulse generator which is set to a minimum of 1.44V to

mimic the results of Chapter 2 (see Fig. 2-11). Finally, to examine only the attenuation and

dispersion effects produced by the optical fiber, the insertion loss of the Mach-Zehnder

modulator (MZM) is set to 0 dB and the extinction ration is set to 200 dB. Setting the

extinction ratio this high minimizes any chirp caused by asymmetric Y-branch waveguides so

as to become negligible [12]. The noise induced by the system up to this point has also been

turned off in the simulation.

Figure 4-4: OptiSystem Gaussian pulse transmitter.

The pulses created have a Gaussian temporal variation that varies in accordance with (4.26)

and are shown in Fig. 4-5. As with the original system of Chapter 2, the MZM is set to

operate at the null bias setting.


85

Figure 4-5: Simulated Gaussian pulse waveforms.

Having demonstrated the ability to produce Gaussian pulses from the MZM

transmitter, a simplified model is constructed. This model is used to examine Link 1, the

transmission of a carrier wave pulse from System 2 to System 1 (Fig. 2-7). The simplified

model is shown in Fig. 4-6. The user defined bit sequence generator creates a 16 bit

sequence of zeros and ones with the following pattern: 0001000100010000. This sequence

runs at 10 Gb/s and is used to produce three pulses from the optical Gaussian pulse

generator. The generator creates pulses according to (4.26) with the addition of a bias

parameter.
86

Figure 4-6: Simplified OptiSystem model.

The bias parameter is set to -0.8187 dBm as obtained from the simulation that produced Fig.

4-4. The optical fiber block is then set so-as to only model attenuation. The attenuation

factor of 0.21 dB km-1 is used as obtained from [9] for a 1550 nm wavelength. The initial

optical waveform and the waveform at a length of 10 km are shown in Fig. 4-7.

It is shown in the figure that the carrier wave is attenuated approximately 2.148 dB

which is in keeping with what would be expected by an attenuation factor of 0.21 dB km -1.

While the effect of attenuation is rather straightforward, the effect of GVD is more

complex.
87

(a)

(b)

Figure 4-7: Attenuation of carrier pulse at 10 km (a) pulse train and (b) individual pulse.
88

The effect of GVD is not profound on the simulated link at only 10 km and, in fact,

is hardly noticeable in graphs. In order to more closely examine this type of dispersion, the

length of the optical fiber is increased based upon the following equation [13]

𝑇2
𝐿𝐷 = |𝛽𝑜 |, (4.28)
2

where 𝛽2 is the GVD parameter and is set to -20 ps2 km-1 , and

𝑇𝐹𝑊𝐻𝑀
𝑇𝑜 = , (4.29)
1.665

Solving for 𝐿𝐷 results in the distance at which the pulse broadens by a factor of √2 [13]. In

order to obtain the full-width at half maximum (FWHM) parameter the following is

considered. With the bit rate is set to 10 Gb/s, it is known that the corresponding bit

duration is 100 ps. Since the default width value of the optical Gaussian pulse generator is

set to 0.5, the FWHM is 50 ps. Solving for 𝑇𝑜 then gives us

50×10−12 [𝑠]
𝑇𝑜 = = 30.03 𝑝𝑠. (4.30)
1.665

Substituting (4.30) into (4.28) results in the following distance calculation for the optical fiber

length:

30.032 [𝑝𝑠]
𝐿𝐷 = 20[𝑝𝑠]/[𝑘𝑚] = 45.09 𝑘𝑚. (4.31)

The parameters of the optical fiber model are then changed to remove the attenuation

effects and add the GVD effect. The length is increased to 45.09 km, the GVD is enabled

with the frequency domain parameter for GVD, 𝛽2, set to -20 ps2 km-1. The pulse

broadening due to GVD is shown in Fig. 4-8. It is worth noting that, the above equations

are for a full pulse (i.e. – from 0 to a maximum amplitude). The pulses used in this

simulation actually begin at approximately 81% of the maximum amplitude and the FWHM

is actually wider than that calculated. Despite this, these calculations still serve to examine
89

TRANSMITTER OUTPUT

FIBER OUTPUT

(a)

TRANSMITTER OUTPUT

FIBER OUTPUT

(b)

Figure 4-8: Intramodal dispersion at 45.09 km with (a) pulse train and (b) individual pulse.
The brown signal is the output from the transmitter while the blue signal is measured at the
end of the fiber.
90

the effect of intramodal dispersion on the pulses.

While polarization-mode dispersion (PMD) has been simulated to determine the

effects in Link 1, the results obtained showed a negligible difference between the input pulse

to the output pulse – even at a distance of 40 km.

4.4 Pulse Waveform at the Receiver

This section is concerned with the condition of the optical pulse at the input to the

receiver’s pin photodiode. While this simulation is similar to that of the previous section, a

new component is introduced, the optical power splitter. As seen in Fig. 4-3, the optical

power splitter provides the input to the receiver as well as the input to the MZM. The

optical power splitter model produces two outputs following the equation
𝛼

𝐸𝑖𝑛 (𝑡)10 20
𝐸𝑜𝑢𝑡 (𝑡) = , (4.32)
√𝑁

where 𝛼 is the power attenuation and 𝑁 is the number of output ports which, in this case, is

equal to 2. While the power splitter was considered ideal for the simulations of Chapter 2,

this simulation follows the recommendation of the Fiber Optics Association in that the

average excess power attenuation of a 3 dB power splitter is 1 dB [14]. Therefore, in this

model, 𝛼 = 1. Furthermore, all of the effects of attenuation and dispersion are simulated

simultaneously to obtain the pulse as seen at the receiver. As PMD is simulated as well, the

birefringence type is simulated using a stochastic method with the differential group delay,

𝜏𝑃𝑀𝐷 , set to 0.5 ps (km0.5)-1 [15]. This updated model is shown in Fig. 4-9.
91

Figure 4-9: OptiSystem model for receiver input waveform.

It is shown in Fig. 4-10 that, from the pulses input into the optical fiber to the pulses as seen

at the receiver, there has been an overall signal attenuation of 6.01 dB. However, the effects

of pulse spreading at a distance of 10 km are negligible.

TRANSMITTER OUTPUT

RECEIVER INPUT

Figure 4-10: Pulse response to dispersion and attenuation effects.


92

4.5 Attenuation and Dispersion Simulation – Link 2

Link 2 is simulated by adding an MZM to the second output of the power splitter

shown in Fig. 4-9. This input optical signal is then modulated by the same means as the

Gaussian pulse transmitter in Fig. 4-4. In this case, the bit sequence generator is set to send

the following series of pulses: 0001010001000010. This sequence allows for the examination

of pulses that align with the previously generated pulses of the optical Gaussian pulse

generator as well as those that do not. For this simulation, the signal provided as an input to

the MZM is the same as the simulated signal of Section 4.4 – i.e., the pulses as seen by the

receiver in Fig. 4-10. Furthermore, it is assumed that there is no insertion loss at the input

to the MZM. The system model is shown in Fig. 4-11.

Figure 4-11: OptiSystem model of Link 2.

For the first simulation, the system is modeled to observe attenuation on the second

length of optical fiber. The parameters for the fiber model are set as described in Section 4.3

to observe only the effect of attenuation. The results are shown in Fig. 4-12.
93

Figure 4-12: Attenuated signal of Link 2.

It is shown that, for the pulse that aligns with the original pulse, some of the waveform

shape is truncated. This is due to the fact that the original pulse begins to decline in power

prior to the new pulse’s creation. This timing misalignment occurs because of the delays

inherent to the electrical models prior to the creation of the pulse in the MZM. The shape is

simply a function of the available power. It can also be seen that when the available power is

at a lower level, that the maximum output power of the pulse is also lower – this is to be
94

expected. Finally, the attenuation associated with the newly created pulses as seen at the

output of the fiber is 2.141 dB which is in keeping with the 2.1 dB km-1 parameter entered

into the optical fiber model.

The second simulation for Link 2 is to show the effects of intramodal dispersion or

GVD. Again, the parameters of the second optical fiber are set as described in Section 4.3

to display only the effect of GVD on the output pulses. A minimal amount of GVD can be

seen on the pulses shown in Fig. 4-13.

MZM OUTPUT
FIBER OUTPUT

Figure 4-13: Effect of intramodal dispersion in Link 2.


95

Once again, the effects of PMD were also simulated in this link in a similar fashion

described for Link 1. The effects were, once again, negligible. Finally, all effects were

simulated simultaneously and can be seen in Fig. 4-14.

Figure 4-14: Attenuation and dispersion in Link 2.

These simulations have shown the effects of attenuation and dispersion on the ASK CWR

system that was modeled in Chapter 2. It can be seen in Fig. 4-14 that the effects of ISI due

to attenuation and dispersion are negligible and that the system maintains excellent pulse

shaping capabilities even after the carrier wave has been reutilized.
96

4.6 References

[1] J. Senior, “Optical Fiber Communications: Principles and Practice,” Third Edition,
Pearson Education, Harlow, England, 2009.
[2] G. Keiser, “Optical Fiber Communications,” Fourth Edition, McGraw-Hill, New
York, 2011.
[3] F. Kapron, D. Keck, R. Maurer, “Radiation losses in optical waveguides,” Applied
Physics Letters, pp. 423-423, 1970.
[4] S. Nagel, “Optical fiber – the expanding medium,” IEEE Circuits and Devices
Magazine, pp. 36-45, 1989.
[5] A. Lino and J. Tamura, “Radiation resistivity in silica optical fibers,” Journal of
Lightwave Technology, vol. 6, pp. 145-149, 1988.
[6] R. West, H. Buker, E. Friebele, H. Henschel and P. Lyons, “The use of optical time-
domain reflectometers to measure radiation-induced losses in optical fibers,” Journal
of Lightwave Technology, vol. 12, pp.614-620, 1994.
[7] H. Henschel and O. Kohn, “Regeneration of irradiated optical fibers by
photobleaching,” IEEE Transactions on Nuclear Science, pp. 699-704, 2000.
[8] S. Girard, J. Baggio and J. Leray, “Radiation-induced effects in a new class of optical
waveguides: The air-guiding photonic crystal fibers,” IEEE Transactions on Nuclear
Science, pp. 2683-2688, 2005.
[9] ITU-T G-series Recommendations – Supplement 39, “Optical system design and
engineering considerations,” International Telecommunication Union, 2016.
[10] J. Schroeder, R. Mohr, P. Macedo and C. Montrose, “Rayleigh and Brillouin
scattering in K2O-SiO2 glasses”, Journal of the American Ceramic Society, vol. 56,
pp. 510-514, 1973.
[11] D. Marcuse, “Interdependence of waveguide and material dispersion,” Applied
Optics, vol. 18, pp. 2930-2932, 1979.
[12] J. Cartledge, “Performance of 10 Gb/s lightwave systems based on lithium niobate
Mach-Zehnder modulators with asymmetric Y-branch waveguides,” IEEE Photonics
Technology Letters, vol. 7, iss. 9, pp. 1090-1092, 1995.
[13] OptiSystem [computer program], 14.0.0.483, Ontario, Canada, Optiwave Systems,
Inc., 2015.
97

[14] “Testing fiber optic couplers, splitters or other passive devices,” www.thefoa.org,
The Fiber Optic Association, Inc., Web, 23 October 2016.
[15] ITU-T G.652, “Characteristics of a single-mode optical fibre and cable,”
International Telecommunication Union, 2009.
98

CHAPTER 5

Comparison of Modulation Techniques

In Chapter 2, the concepts of amplitude modulation (AM) and amplitude shift keying

(ASK) were introduced. Throughout this thesis, AM has been the primary method by which

the carrier waves have been modulated. This chapter examines alternative modulation

techniques and the means by which they may be employed in a carrier wave reutilized

system. These alternative system configurations are simulated and compared to earlier

results.

In Section 5.1, the concepts of frequency modulation (FM) and phase modulation

(PM) are introduced with a discussion regarding their digital forms – frequency shift keying

(FSK) and phase shift keying (PSK). Furthermore, the concept of differential PSK (DPSK)

is discussed along with a means of asynchronous detection. In Section 5.2, a new digital link

is introduced and compared to the digital link of Chapter 2. The link model is discussed in

detail. In Section 5.3, the model is simulated and the numerical results are discussed. These

results are compared to the results in Chapter 2.

5.1 Frequency and Phase Modulation

Both FM and PM belong to a technique of modulation where the angle of the carrier

is modulated with an information-bearing signal, 𝑚(𝑡). This technique is known as angle

modulation or exponential modulation [1]. Consider the electric field of an optical carrier

wave that is represented by

𝐸𝑠 (𝑡) = 𝐴𝑠 𝑐𝑜𝑠(𝜃𝑠 (𝑡)), (5.1)


99

where 𝜃𝑠 (𝑡) is the generalized angle. Now, assuming the case where, at some instant, the

generalized angle is tangential to the angle 𝜙 = (𝜔𝑠 𝑡 + 𝜃0 ) then, at a small interval as

Δ𝑡 → 0 the signal 𝐸𝑠 and the sinusoid associated with 𝜙 are identical, i.e., [1]

𝜃𝑠 (𝑡) = 𝐴𝑠 𝑐𝑜𝑠(𝜙) = 𝐴𝑠 cos(𝜔𝑠 𝑡 + 𝜃0 ). (5.2)

This concept can be generalized at every instant, thereby defining the instantaneous

frequency, 𝜔𝑖 , as the slope of 𝜃𝑠 (𝑡) and any instant, 𝑡. Therefore, the instantaneous angular

frequency and the generalized angle can be related by [1]

𝑑𝜃𝑠
𝜔𝑖 (𝑡) = (5.3)
𝑑𝑡

and
𝑡
𝜃𝑠 (𝑡) = ∫−∞ 𝜔𝑖 (𝛼)𝑑𝛼. (5.4)

In PM, the angle 𝜃𝑠 (𝑡) of the carrier is varied linearly with the message signal, 𝑚(𝑡). And,

assuming that 𝜃0 is 0, this variation is given as

𝜃𝑠 (𝑡) = 𝜔𝑐 𝑡 + 𝑘𝑝 𝑚(𝑡), (5.5)

where 𝑘𝑝 is an arbitrary constant and 𝜔𝑐 is the carrier frequency. The resulting PM wave is

then derived by substituting (5.5) into (5.1):

𝐸𝑠(𝑝𝑚) (𝑡) = 𝐴𝑠 cos[𝜔𝑐 𝑡 + 𝑘𝑝 𝑚(𝑡)]. (5.6)

Furthermore, the instantaneous frequency of a PM wave is given by [1]

𝑑𝜃𝑠
𝜔𝑖 (𝑡) = = (𝜔𝑐 + 𝑘𝑝 𝑚̇(𝑡)). (5.7)
𝑑𝑡

It can be seen from (5.7) that the instantaneous angular frequency varies linearly with the

derivative of the modulating signal. This is not the case, however, with FM.

The instantaneous angular frequency of an FM wave is given by [1]


100

𝜔𝑖 (𝑡) = (𝜔𝑐 + 𝑘𝑓 𝑚(𝑡)). (5.8)

Therefore, the instantaneous angular frequency of an FM signal varies linearly with the

modulating signal – rather than the derivative of the modulating signal. The angle, 𝜃𝑠 (𝑡), is

now
𝑡
𝜃𝑠 (𝑡) = ∫−∞[𝜔𝑐 + 𝑘𝑓 𝑚(𝛼)]𝑑𝛼 (5.9)

again, assuming that the 𝜃0 constant is equal to zero. The FM wave is then shown to be

𝑡
𝐸𝑠(𝑓𝑚) (𝑡) = 𝐴𝑠 cos [𝜔𝑐 𝑡 + 𝑘𝑓 ∫−∞ 𝑚(𝛼)𝑑𝛼 ]. (5.10)

It can be seen from (5.6) and (5.10), in both PM and FM, that the angle of a carrier is varied

in proportion to 𝑚(𝑡). The primary difference being that, in PM, it is varied directly

proportional to the signal, whereas in FM, the variation is proportional to the integral of the

signal [1]. It is important to note that, unlike AM, the instantaneous frequency and phase is

modulated but the amplitude, 𝐴𝑠 , remains constant. Example waveforms for FM and PM

are shown in Fig. 5-1.

(a) (b)

Figure 5-1: Example waveforms for (a) FM and (b) PM. The modulating signal is shown in
red.
101

In the case of digital modulating waveforms, during FM, the wave frequency

switches between two distinct frequencies. This is known as frequency shift keying (FSK)

because information bits are transmitted by keying different frequencies. In PM, however,

the derivative of the modulating signal in (5.7) results in zero, except at points of

discontinuity where impulses is present [1]. This results in the frequency of the PM signal

remaining constant, except at the points of discontinuity (i.e. – when the waveform

transitions from a 0 bit to a 1 bit or vice versa). The resultant waveform is known as phase

shift keying (PSK) because the information bits are transmitted by shifting the carrier phase.

Examples of these waveforms are shown in Fig. 5-2.

(a) (b)

Figure 5-2: Example waveforms for (a) FSK and (b) PSK.

The digital simulation that is shown in Section 5.2 utilizes a simple form of

asynchronous heterodyne detection known as DPSK. As shown in Fig. 5-3, the

asynchronous detection utilizes a one-bit delay line, rather than having to utilize a phase-

locked loop (PLL). When the information is encoded with a PSK method there are changes

in the optical phase. This allows the mixer to produce a positive or negative output
102

depending on whether the phase of the received signal has changed from the previous bit

[2].

Detector

1-bit Delay
Interferometer
Optical Signal - LP FILTER Digital Signal

Detector

Figure 5-3: DPSK receiver configuration.

In other words, the concept of differential detection is for the receiver to detect the relative

phase change between successive modulated phases (𝜃𝑘 − 𝜃𝑘−1 ) [1]. In PSK, the phase

value is finite and in binary PSK (BPSK), the finite values are 0 and 𝜋. Therefore, a phase

difference of 0 would represent a binary 0 and a phase difference of 𝜋 would represent a

binary 1.

5.2 Digital DPSK-ASK CWR Link Model

In this section, a complimentary DPSK-ASK CWR digital link is modeled and

simulated. The purpose of these simulations is to examine the performance associated with

utilizing DPSK on the initial link, and ASK on the return link. These simulations are then

compared to the digital link simulation of Chapter 2. The model components are discussed,

including their related functions. The system is modeled as shown in Fig. 5-4.
103

Figure 5-4: OptiSystem model of a DPSK-ASK CWR digital link.


104

For comparison purposes, the PRBS is set to the same parameters as the proposed

system in Section 2.6. The output is provided as an input to the DPSK transmitter model

and it is also provided as an input to the BER analyzer. The DPSK transmitter model is

complex and requires further examination.

The transmitter simulates a single channel optical transmitter with DPSK modulation

[3]. Internally, it is modeled using two MZMs, wherein one MZM is used for phase

modulation of the data and the other is used for amplitude modulation of the clock for RZ

pulse carving [3]. Therefore, this component operates as a serial DPSK encoder and

transmitter [4]. The transmitter can simulate four different types of DPSK signals: RZ with

a 33% duty cycle, RZ with a 50% duty cycle, RZ with a 66% duty cycle and NRZ. For this

simulation, an RZ signal with 66% duty cycle is used. It provides a higher power level for

the ASK return link as well as the benefits associated with RZ coding. The RF spectral

content and the time domain signals produced by the transmitter are shown in Fig. 5-5.
105

(a)

(b)

Figure 5-5: Output of DPSK transmitter showing (a) the frequency domain and (b) the time
domain.
106

The signal produced by the DPSK transmitter is then launched down 10 km of

single-mode fiber with the default OptiSystem parameters associated with a 1550nm optical

wavelength. The signal is then coupled to a 3 dB optical power splitter that is assumed to be

ideal. Note that if the splitter were not ideal, an additional attenuation factor would be

introduced as discussed in Chapter 4. One branch of the signal is then fed to the input of a

1-bit delay interferometer which provides the input to the two pin photodiodes and an

electrical subtractor. This configuration creates the DPSK receiver as shown in Fig. 5-3.

The output current of the electrical subtractor model is filtered with a low-pass Bessel filter

and is provided to the BER analyzer for signal evaluation.

The second branch of the signal is provided as an input to the MZM. The MZM

acts as a transmitter using the same configuration that was used in System 1 of Chapter 2

(Fig. 2-7). This transmitter modulates the signal with ASK and launches it back to the initial

system via 10 km of optical fiber. At this point, all component parameters are equivalent to

the return path of the ASK-ASK digital link described in Chapter 2.

5.3 Numerical Results

In the first simulation, the BER of the first digital receiver is compared to that of the

ASK-ASK model in Chapter 2. These receivers are part of System 1 (Fig. 2-7). It is shown

in Fig. 5-6 that the eye opening of the DPSK receiver is much larger than that of the ASK

receiver: 547 × 10−6 amplitude units as compared to 14 × 10−6 . Furthermore, the timing

jitter in the ASK-ASK model’s eye diagram is much greater than that of the DPSK-ASK.
107

(a) (b)

Figure 5-6: Eye diagram comparison of System 1 with (a) DPSK-ASK system and (b) ASK-
ASK system.

This is evidenced by the amount of distortion at the zero crossing. These results are

expected, as DPSK has been experimentally shown to drive 40 Gb/s signals at 2.2 km at a

BER of 10-9 [5] and PSK is known to have a much lower BER in general when compared to

ASK [2]. Furthermore, the ASK signal is degraded due to the smaller pulses associated with

biasing the ASK transmission at 1.44V as described in Chapter 2.

Next, it is shown in Fig. 5-7 that the eye diagram of the DPSK-ASK receiver of

System 2 is much larger than that of that used in ASK-ASK: 39.8 × 10−6 amplitude units

as opposed to 12.7 × 10−6 . However, the timing jitter of the DPSK-ASK system is larger.
108

(a) (b)

Figure 5-7: Eye diagram comparison of System 2 with (a) ASK of current simulation and (b)
ASK of Chapter 2 simulation.

The comparative improvement in the eye opening of the DPSK-ASK model can be partially

attributed to the larger transfer of power throughout the system. For example, Table 5-1

shows that when transmitting a 0 dBm optical carrier, the DPSK-ASK power levels are

larger at the input of each receiver when compared to the ASK-ASK model.

Table 5-1: Power comparison of ASK-ASK and DPSK-ASK CWR systems

Optical Power ASK-ASK System DPSK-ASK System


Transmitted Receiver 1 Receiver 2 Receiver 1 Receiver 2
(dBm) (dBm) (dBm) (dBm) (dBm)
-3 -11.8 -22.2 -9.8 -20.3
0 -10.0 -20.4 -6.8 -17.1
3 -7.8 -18.1 -3.8 -14.0

Finally, when comparing the BERs of each system, with the transmitter at 0 dBm, it

is obvious that the DPSK link’s performance exceeds that of the ASK model. For example,

at the first receiver of the DPSK-ASK model, the Q-factor is 31.1 with a theoretical BER of
109

3.03 × 10−213 , while that of the ASK-ASK model is 9.5 with a BER of 10 × 10−22. Note

that the theoretical BER obtained for the DPSK-ASK model is extremely high and may not

be a practical value in an experimental system. At the second receiver, the DPSK-ASK

model has a BER of 14.6 × 10−22 as compared to the 10 × 10−22 BER of the ASK-ASK

model. This increased BER can be attributed to the increase in timing jitter. However, as

Table 5-1 shows, as the transmitted optical power increases, the power at the ASK receiver

begins to greatly improve. This is shown in the next simulation.

In the second simulation, the power is increased from -6 dBm to 6 dBm and is

compared to the proposed system of Chapter 2. It is shown in Fig. 5-8 that the BER of

System 2 improves almost linearly as the transmitted optical power increases. There is,

therefore, a point at which the BER of System 2 in the DPSK-ASK model has a better

performance than that of the ASK-ASK model. In this specific simulation, the point is at

approximately 0 dBm. The results of System 1 for the DPSK-ASK model are not shown

because the BER remains in the range of 10−200 to the point where it becomes so small that

the model shows that it is zero. While these are not practical values, they do provide a

method of comparison whilst reviewing the BERs. For example, one might question at what

point the BER of System 1 would become a mitigating factor when examining greater

transmission distances. Furthermore, it is interesting to note that this extremely low BER

provides an opportunity for system improvement by directing less power to the receiver of

System 1 and redirecting that power to System 2. This could be accomplished by use of a

variable beam splitter. The simulations of this chapter show that the DPSK-ASK proposed

system has the capability of greatly outperforming the ASK-ASK system of Chapter 2.
110

(a)

(b)

Figure 5-8: BER comparison of (a) DPSK-ASK and (b) ASK-ASK. Note that the BER of the
DPSK signal is not shown in (a) because it is negligible at these distances.
111

5.4 References

[1] B. P. Lathi and Z. Ding, “Modern Digital and Analog Communications Systems,”
Fourth Edition, Oxford University Press, 2009.
[2] G. Keiser, “Optical Fiber Communications,” Fourth Edition, McGraw-Hill, New
York, 2011.
[3] OptiSystem [computer program], 14.0.0.483, Ontario, Canada, Optiwave Systems,
Inc., 2015.
[4] M. Seimetz, “High-Order Modulation for Optical Fiber Transmission,” Springer,
2009.
[5] A. Martinez, V. Polo, J. L. Corral and J. Marti, “Experimental demonstration of
dispersion-tolerant 155-Mb/s BPSK data transmission at 40 GHz using an optical
coherent harmonic generation technique,” IEEE Photonics Technology Letters, vol. 15,
no. 5, 2003.
112

CHAPTER 6

Carrier Wave Reutilization and a Standard Methodology

Throughout this thesis, techniques and theoretical methodologies of implementing

carrier wave reutilization (CWR) have been elucidated. However, just because something

can be done, does not mean that it should be done. In this chapter, the benefits and

drawbacks of a CWR implementation are examined in comparison to a standard

methodology. This is done in order to provide a determination as to whether or not CWR is

worth pursuing.

In Section 6.1, the power utilized in a CWR link is compared to that used in a

comparable digital link. Additional analysis of the cost of operating each link is performed.

The comparative limitation of a CWR link is examined in Section 6.2, with a simulation to

determine the drawback of using such a link.

6.1 Power and Cost Comparison

The primary aim of this thesis is to provide an effective transport technology that

also lowers the power consumption in an optical fiber link. It has been shown throughout

that CWR provides an effective method of transporting data streams in two directions. In

this section, both power usage and cost are considered in comparison to a standard digital

link.

Fig. 6-1 shows a bidirectional digital ASK link while Fig. 6-2 shows the CWR ASK

link. The most obvious cost and power savings occur because there is no need for a laser at

System 1 when using CWR. The only additional component required by the CWR link is
113

that of a 3 dB power splitter which is far more cost effective and requires no additional

power.

System 2 System 1
Amplitude
Continuous Modulator Detector Amplifier
Laser Wave Laser
Controller
Decision

Drive
DATA IN
Electronics

Amplitude
Amplifier Detector Modulator Continuous
Wave Laser Laser
Decision Controller

Drive
DATA IN
Electronics

Figure 6-1: Bidirectional ASK digital link.

SYSTEM 2 SYSTEM 1
Continuous Wave Amplitude Detector Amplifier
Laser Modulator
Laser Controller Decision

DATA IN
Drive Electronics

Amplifier Detector Amplitude


Modulator DATA IN
Decision Drive Electronics

Figure 6-2: ASK-ASK CWR digital link.


114

The fact that an additional laser is not required does not just yield a small

improvement. On the contrary, the cost of a continuous wave laser can range from

hundreds of dollars to tens of thousands of dollars depending on the optical power required.

Furthermore, the output power of these lasers requires a much larger investment of input

power as the output power requirement increases. While the amount of input power varies

among manufacturers and models, an example of this increase can be seen in Table 6-1.

Table 6-1: 1550nm Erbium CW Laser Input and Output Power [1]

Input Power Output Power


Laser Model
(W) (W)
ELM-1 30 1
ELM-5 40 5
ELR-1 50 1
ELM-10 80 10
ELR-5 100 5
ELM-20 145 20
ELR-10 180 10
ELM-30 200 30
ELM-50 300 50
ELR-30 320 30
ELR-50 400 50

Furthermore, as a system is scaled to provide more links, both the cost and the power

requirements increase as well. For these reasons, CWR can be considered a viable alternative

to the current method of bidirectional communications.

In order to provide a power savings comparison, the systems of Figs. 6-1 and 6-2 are

simulated. The optical power is sampled at four points: the output of the amplitude

modulator of System 2, the input to the receiver at System 1, the output of the amplitude

modulator of System 1 and the input to the receiver at System 2. The simulation for the

CWR System is identical to the ASK-ASK CWR system shown in Chapter 2. The systems
115

are simulated at 10 Gbps with the CWR Laser outputting 0 dBm. These results are shown in

Table 6-2.

Table 6-2: Power Comparison of Traditional System and CWR System

Traditional System CWR System


Power Sampling Point Power Level Power Level
(dBm) (dBm)
Amplitude Modulator
-11.159 -5.422
System 2
Receiver System 1 -15.159 -10.432
Amplitude Modulator
-26.159 -18.649
System 1
Receiver System 2 -30.159 -20.649

It is shown in Table 6-2 that the total optical power savings of a digital ASK-ASK CWR

system as compared to a traditional bidirectional system is 9.51 dB. Note that the power

values obtained for the traditional system are based on the assumption that the carrier wave

must be provided by a CW Laser in System 1 as shown in Fig. 6-1. Therefore, there is a

marked improvement in power usage in the CWR system.

6.2 Limitations of CWR

The primary limitations of CWR when compared to a standard methodology are an

increase in BER and an increased attenuation of the return path. This is obvious from the

standpoint that, in a standard link, a new carrier wave is directly produced for the return

path, thereby providing source that is free of the effects of dispersion and attenuation from

the initial path as shown in Figs. 6-1 and 6-2.

In order to determine how these limitations affect a system, two models are

simulated. The first model is the digital link of Chapter 2 (Fig. 2-9) while the second link is
116

the simplex digital link shown in Fig. 2-6. The simplex link may be used as it is the same

system for both the initial and return paths in a standard methodology. The first simulation

provides a comparison of BERs with each laser producing 0 dBm power output. The

second simulation is used to determine the distance limitation by sweeping the length

parameter of the optical fiber model while maintaining a constant laser power of 0 dBm.

6.3 Numerical Results

In the first simulation, the eye diagram and its associated parameters of Q-factor,

BER and eye opening amplitude are examined. Fig. 6-3 shows the results of a standard

implementation at a distance of 10 km with a 0 dBm power output. The bit rate is set to 10

Gb/s. The CWR implementation results are shown in Figs. 6-4 and 6-5. They show the

initial link and return link respectively. It is evident from the extremely high Q factor of the

simplex digital link that the BER is greatly improved as compared to both links of the CWR

model. Furthermore, the eye opening is very large with minimal timing jitter and distortion.

The noise margin is very high in the simplex digital link. All of these factors contribute to

the greatly improved results found in the standard methodology.


117

(a) (b)

(c) (d)

Figure 6-3: ASK simplex digital link results, (a) Q-Factor, (b) minimum BER,
(c) eye height and (d) eye diagram.
118

(a) (b)

(c) (d)

Figure 6-4: ASK-ASK CWR initial digital link results, (a) Q-Factor, (b) minimum BER,
(c) eye height and (d) eye diagram.
119

(a) (b)

(c) (d)

Figure 6-5: ASK-ASK CWR return digital link results, (a) Q-Factor, (b) minimum BER,
(c) eye height and (d) eye diagram.
120

The aforementioned results lend themselves to the information provided by the

second simulation. In this simulation, the parameters of both the models are maintained,

except for the distance. The length of the optical fiber is swept from 11 km to 60 km and

the BER of each model is examined - for the purpose of comparison, a BER of 10−9 is

considered.

It is shown in Fig. 6-6 that, in the case of the ASK-ASK CWR model, the BER

increases dramatically as the length of the fiber is increased. In fact, the BER of the initial

link surpasses 10−9 somewhere between 18 and 19 km. The return link, however, maintains

a low BER until approximately 35 km – where it surpasses the 10−9 point. It is interesting

to note that, at approximately 30 km, the BER of the initial link becomes unreliable. This is

due to the fact that, at distances greater than 30km, the BER is primarily a function of the bit

sequence and the simulation occasionally results in a BER of 1 (one error for every bit

transmitted).

Figure 6-6: BER of an ASK-ASK CWR digital link with increasing distances.
121

The results shown in Fig. 6-7 show the advantage associated with a traditional

methodology. The BER is greatly improved as compared to the CWR results and a BER of

greater than 10−9 does not occur until approximately 56 km. The other benefit is that these

results are similar for both the initial and return paths. Obviously, this methodology would

be better suited for long-reach applications than the CWR method. However, for short-

reach and intermediate-reach applications (up to approximately 15 km [2]), the CWR method

is sufficient.

Figure 6-7: BER of a simplex digital link with increasing distances.


122

6.4 References

[1] IPG Photonics Corp., “ELM and ELR, 1-50W Erbium Single-mode Lasers,” [web],
http://www.ipgphotonics.com/laser/view/6/Lasers/Low_Power_CW_Fiber_Laser
s/1_53___1_65__m/ELM_and_ELR__1_50_W#[Overview], Accessed 29OCT16.
[2] G. Keiser, “Optical Fiber Communications,” Fourth Edition, McGraw-Hill, New
York, 2011.
123

CHAPTER 7

Conclusion and Future Research Direction

A novel approach to a power saving transport technology has been presented in this

research in the form of carrier wave reutilization (CWR). It has been shown that there is a

need for such technologies because the core Internet infrastructure is becoming limited by

its power consumption. Furthermore, it has been shown that CWR is a viable method of

data transport in both digital and analog forms of modulation. Considerations such as

attenuation and dispersion of the optical carrier have been examined and it is shown that

CWR provides a useful means for sending data on a previously modulated carrier wave.

Another method of digital CWR was then investigated using a complimentary DPSK-ASK

system. It is shown that this implementation could provide further performance

improvements when compared to an ASK-ASK system. Finally, it has been shown that the

cost and power consumption of both short and intermediate reach optical fiber

communication systems are reduced with CWR as compared to a traditional implementation.

The applications of CWR are just beginning to be realized. While both an ASK-ASK

and DPSK-ASK CWR link have been explained, other modulation formats can, and should

be considered. Furthermore, while an SCM Analog CWR link is demonstrated using ASK,

the other sub-modulation formats could be implemented as well. The following related

topics require further investigation for future research:

(i) Design and simulation of an FSK-ASK digital link

(ii) Minimization of intramodulation distortion through proper biasing of MZM on

return link.
124

(iii) Optimized channel spacing for an SCM CWR system

(iv) Effects signal filtering prior to the use of an RF combiner in an SCM CWR

system.

(v) Improved detection of SCM CWR signals through optical and RF filtering.

(vi) Performance comparison of FSK, PSK, ASK, FM, PM and AM RF modulation in

an SCM CWR system.

(vii) Design and simulation of a differential quadrature PSK-ASK CWR system.

(viii) Improved BER performance of return path in a DPSK-ASK CWR system

through the use of a variable beam splitter.

(ix) Wave division multiplexing and CWR.

(x) Examination of multiple path links to determine the maximum number of times

CWR can be performed.

(xi) Experimental verification of the simulation results which could not be performed

due to unavailability of proper instrumentation in our laboratory.

The implications of this research are numerous. For example, new concepts have

been introduced that, with minimal changes to infrastructure, lower both the operating and

equipment costs for the telecommunications industry. Furthermore, these concepts could

be modified to operate with different modulation formats and, with adjustments to power

distribution, could produce even greater results than those elucidated in this thesis.