Low-Cost Optical Reflection Techniques for the

Assessment of Intra-office Optical Channel Quality

João António Pires Mendes de Campos Pereira

Thesis to obtain the Master of Science Degree in

Electrical and Computer Engineering

Supervisors: Prof. Paulo Sérgio de Brito André

Dr. Miguel Ângelo Madureira

Examination Committee

Chairperson: Prof. José Eduardo Charters Ribeiro da Cunha Sanguino

Supervisor: Prof. Paulo Sérgio de Brito André

Member of the Committee: Prof. Mário José Neves de Lima

May 2015

ii

iii

To my family

iv

v

Acknowledgements

First of all, I would like to thank my supervisor, Prof. Paulo André, for his great availability to

supervise my work and for always answering my questions.

I would like to thank my co-supervisor, Dr. Miguel Madureira, for always being available to give

suggestions and for helping me with the experimental implementations.

I would like to thank Coriant for providing me access to their facilities in Alfragide (Lisbon), as

well as Instituto de Telecomunicações (IT) for the access to room LT3, where most of the work was

conducted.

Thanks to my family and friends for their support, which was essential for the development of

this work.

Special thanks to Francisco Rosário, João Lemos, Samuel Balula, Eva Campos Pereira and

Prof. João Pires for all their advices.

vi

vii

Resumo

O objetivo deste trabalho é propor e demonstrar experimentalmente um novo método de

localização de defeitos nos primeiros metros de uma fibra óptica. A zona da fibra que está mais próxima

do laser transmissor está exposta a um ambiente que potencia a existência de falhas devido ao elevado

número de juntas e conectores. A monitorização da fibra óptica é uma prática essencial para a

manutenção da qualidade de serviço, especialmente nessa secção da fibra. Nos dias de hoje existem

muitas técnicas de monitorização, no entanto a maioria destes métodos acarretam custos acrescidos e

não permitem uma avaliação contínua da fibra. Através da utilização do sinal de tráfego, o método

proposto permite uma monitorização constante e com menores custos.

O método proposto emprega a técnica da reflectometria e foi validado experimentalmente para

dois cenários distintos: i) com um sinal contínuo para aplicação em sistemas com amplificação de

Raman e ii) com sinais de tráfego para aplicação em redes ópticas em geral. Considerando um sinal

contínuo, demonstrou-se a possibilidade de monitorizar uma ligação óptica com vários defeitos. Usando

um sinal de tráfego, demonstrou-se a possibilidade de avaliar os primeiros 6500 metros de fibra.

Os resultados experimentais obtidos demonstram a viabilidade deste método para a

monitorização dos primeiros metros de fibra com custos reduzidos.

Palavras-chave: comunicações ópticas, reflectometria, técnica de monitorização de fibra óptica,

avaliação contínua de defeitos na fibra.

viii

ix

Abstract

The goal of this dissertation is to propose and demonstrate a novel and cost effective technique

aiming to monitor the integrity of the optical fiber within intra-office distances. Intra-office is the fiber zone

in the vicinity of the transmitter which is an environment that is subject to channel impairments due to

the high number of connectors, fiber splices and bends. Therefore, the monitoring of the optical fiber

infrastructure, especially within the intra-office range, is an essential practice for the maintenance of the

quality of service. Even though there are many fiber monitoring techniques, the majority of them carry

unbearable costs, and do not allow a constant fiber fault examination during the network operation. The

technique here proposed, with resort to the optical traffic signal, allows a continuous and low-cost

assessment of the fiber infrastructure.

The proposed monitoring structure employs reflectometry mixing and was experimentally

validated for two distinct situations: i) with continuous wave optical signal, aiming to the application in

networks employing distributed Raman amplification and ii) with SDH (synchronous digital hierarchy)

traffic signals, for general optical networks. For the situation with continuous wave signal, it was

demonstrated the feasibility to use this type of signal to monitor an optical link with several impairments.

Considering the broadcast of SDH traffic signals, it was demonstrated the possibility to monitor the initial

6500 meters of the fiber.

The obtained experimental results show the feasibility of the proposed technique to monitor

optical networks within intra-office distance at a lower cost than the usual employed techniques.

Keywords: optical communications, reflectometry, fiber monitoring technique, in-service fiber fault

assessment.

x

xi

Contents

List of figures xiii

Nomenclature xv

Glossary xvii

1. Introduction 1

Scope of the work ........................................................................................................................1

Motivation ....................................................................................................................................3

Objective and work structure .......................................................................................................3

Contribution of this dissertation ...................................................................................................4

Scientific dissemination ...............................................................................................................4

2. State of the art 5

Introduction ..................................................................................................................................5

Optical time-domain reflectometer ...............................................................................................7

Optical low-coherence reflectometer ...........................................................................................8

Optical frequency-domain reflectometer................................................................................... 10

Other monitoring techniques .................................................................................................... 13

Proposed monitoring structure ................................................................................................. 16

3. Online reflectometry mixing 17

Description ................................................................................................................................ 17

The theoretical model ............................................................................................................... 17

3.2.1 State of polarization scrambling .......................................................................................... 19

3.2.2 Multiple reflection scenario .................................................................................................. 21

Fault monitoring module ........................................................................................................... 22

Simulation model ...................................................................................................................... 26

4. Continuous signal operation 29

Single reflective event ............................................................................................................... 29

Multiple reflective events .......................................................................................................... 31

Experimental results for a continuous signal operation ............................................................ 33

5. Fault diagnosis with real traffic 37

Fault detection with random data correlation............................................................................ 37

Reflectometry mixing with random data ................................................................................... 38

Reflectometry mixing with real traffic ........................................................................................ 41

Technique limitations analysis .................................................................................................. 45

5.4.1 Multi-channel impact ............................................................................................................ 45

5.4.2 Spatial resolution ................................................................................................................. 46

5.4.3 Measuring range .................................................................................................................. 46

xii

6. Experimental results 47

Laboratory test scenario ........................................................................................................... 47

Results with random data ......................................................................................................... 48

Results with traffic data............................................................................................................. 49

7. Implementation scenario 51

Location in ROADM architecture .............................................................................................. 51

Monitoring module .................................................................................................................... 52

Optimum device characteristics ................................................................................................ 53

8. Conclusions and future work 55

Bibliography 57

xiii

List of figures

1.1 Total global IP traffic forecast by 2018 [5]. …………………..……………………………..….............2

2.1 Spectral dependence of a single-mode fiber (reproduced from [3]). ............................................... 5

2.2 Reflection and refraction of an electromagnetic wave in an interface (reproduced from [10])……..6

2.3 Typical OTDR reflectogram exposing several reflective events [12]. ............................................. 7

2.4 OTDR functional block diagram [11]. .............................................................................................. 8

2.5 OLCR functional block diagram [10]. ............................................................................................... 9

2.6 Typical OLCR reflectogram reporting two reflective events (reproduced from [16]). ...................... 9

2.7 NA-OFDR functional block diagram [13]. ...................................................................................... 10

2.8 I-FMCW functional block diagram [13]. ......................................................................................... 11

2.9 C-OFDR functional block diagram [13]. ......................................................................................... 11

2.10 Typical OFDR reflectogram exposing several reflective events [18]. ......................................... 12

2.11 Block diagram of an operating WDM-PON system with embedded OTDR features [28]……….13

2.12 Block diagram of a transceiver with embedded OTDR features [27]. ......................................... 14

2.13 Zhao et al. monitoring structure block diagram [26]. ................................................................... 14

2.14 a) Plot of the cross-correlation between the traffic reference signal and its back reflection

(reproduced from [26]). b) Zhao et al. monitoring technique reflectogram (reproduced from [26])....... 15

3.1 Online reflectometry mixing block diagram. .................................................................................. 17

3.2 Unidirectional block diagram of the online reflectometry mixing for a single reflective event, occuring

after a fiber with length 𝐿. ...................................................................................................................... 18

3.3 Polarization states displayed over the Poincaré sphere: a) Linear b) Circular c) Elliptical………..21

3.4 Unidirectional block diagram of the online reflectometry mixing for a multiple reflection scenario.22

3.5 a) Temporal evolution of the input signal. b) Auto-correlation plot of the input signal. ................. 23

3.6 a) Temporal evolution of the output signal for a reflective event at distance 𝐿. b) Auto-correlation

plot of the output signal. ........................................................................................................................ 24

3.7 Proposed monitoring technique output for a reflective event at distance 𝐿. ................................. 25

3.8 Schematic of the monitoring module algorithm. ............................................................................ 26

4.1 Photodiode electric output current histogram for a single reflection scenario obtained via Monte

Carlo simulation. The red arrows indicate the theoretical extreme current values. .............................. 30

4.2 Photodiode electric output current histogram for a double reflection scenario obtained via Monte

Carlo simulation. The red arrows indicate the theoretical extreme current values. .............................. 31

4.3 Photodiode electric output current histograms for different reflective event scenarios obtained with

Monte Carlo simulation. ......................................................................................................................... 32

4.4 Normalized FWHM of the Monte Carlo simulation output as a function of the number of reflective

events considering different values for factor θ. The independent blue asterisks represent the normalized

xiv

FWHM of histograms of experimental results. The Poincaré spheres represent the state of polarization

distribution when θ is 1, 0.1 and 0.01. ................................................................................................... 33

4.5 Experimental scenario for a single refletive event. ........................................................................ 34

4.6 Photodiode output current as function of the input power. ............................................................ 34

4.7 a) Plot of experimental photodiode’s output current over time for a double reflective event case. b)

The histogram of the current values from plot a)................................................................................... 35

5.1 Block diagram for correlation testing with PRBS data. .................................................................. 37

5.2 Plot of the correlation between the random data reference signal and its back reflection which was

imposed by an open-ended fiber with 15 meters length. ...................................................................... 38

5.3 Auto-correlation plot of the monitoring structure output signal for a fiber impaiment at 20 meters

from the optical coupler. ........................................................................................................................ 39

5.4 Proposed technique reflectogram for a fiber impaiment at 20 meters from the optical coupler using

PRBS data: a) without smoothing filter; b) with smoothing filter. .......................................................... 39

5.5 BNSR as a function of the coupler ratio coefficient. For each case, it was considered a reflective

event with a reflection ratio of 2% at 20 meters from the coupler. ........................................................ 40

5.6 Proposed technique reflectogram using PRBS data for reflective events at 12 and 139.6 meters

from the optical coupler. ........................................................................................................................ 41

5.7 a) STM-1 frame structure. b) Overhead section of the STM-1 frame [8]. ..................................... 41

5.8 Aproximatted STM-1 frame structure for simulation purposes. ..................................................... 42

5.9 Auto-correlation plot of the approximated STM-1 input signal. ..................................................... 43

5.10 Proposed technique outcome without plot window adjustment for a single reflective event at 302

meters from the optical coupler. ............................................................................................................ 43

5.11 Online reflectometry mixing reflectogram for a single reflective event at 302 meters from the optical

coupler. .................................................................................................................................................. 44

5.12 BNSR as a function of the reflective event location in the fiber. For each case, a reflection

coefficient of 1.5% and a coupling ratio of 90/10 were considered. ...................................................... 45

6.1 Experimental scenario for fiber fault detection using traffic signals. ............................................. 47

6.2 Proposed monitoring technique experimental reflectogram using input PRBS data for a reflective

event at 1 kilometer from the optical coupler......................................................................................... 48

6.3 Proposed monitoring technique experimental reflectogram using input data traffic for a reflective

point at: a) 1 kilometer; b) 2 kilometers. ................................................................................................ 49

7.1 Implementation of the monitoring structure in a ROADM architecture. ......................................... 51

7.2 DSP sampling process applied to the monitoring structure a) input signal; b) output signal……..52

xv

Nomenclature

𝐴𝑡 Power attenuation coefficient

𝛼 Fiber attenuation coefficient

𝛽 Intensity amount of �̂� direction polarization

𝑐 Speed of light

𝜒 Ellipticity angle

𝛥𝑡 Back reflection time delay

𝜀 Coupler ratio coefficient

𝐸0 Amplitude of the electric field

𝜙 Phase imposed by fiber section

𝜑 Phase difference between orthogonal fields

𝜆 Wavelength

𝐿 Fiber section length

𝑀 Mirror reflectivity

𝑛 Refractive index

𝑤0 Angular frequency

𝜓 Azimuth angle

𝑅 Reflection coefficient

𝑅𝜆 Responsivity

𝑡 Time

�̂� Field along direction x

�̂� Field along direction y

xvi

xvii

Glossary

ADC Analog-to-Digital Converter

ADM Add-Drop Multiplexer

APC Angle-Polished Connector

APD Avalanche Photodiode

AWG Arrayed Waveguide Grating

BNSR Background Noise Suppression Ratio

C-OFDR Coherent Optical Frequency-Domain Reflectometer

DSP Digital Signal Processor

DWDM Dense Wavelength Division Multiplexing

EDFA Erbium Doped Fiber Amplifier

FDR Frequency-Domain Reflectometry

FTTH Fiber to the Home

FWHM Full Width at Half Maximum

GVD Group Velocity Dispersion

I-FMCW Incoherent Frequency-Modulated Continuous Wave

I-OFDR Incoherent Optical Frequency-Domain Reflectometer

IP Internet Protocol

NA-OFDR Network Analysis Optical Frequency-Domain Reflectometer

NMS Network Management System

NRZ Non-Return to Zero

OADM Optical Add-Drop Multiplexer

OCT Optical Coherent Tomography

OFDR Optical Frequency-Domain Reflectometer

OLCR Optical Low-Coherence Reflectometer

OTDR Optical Time-Domain Reflectometer

xviii

OTN Optical Transport Network

OTU-k Optical Channel Transport Unit-k

PC Physical Connector

PDH Plesiochronous Digital Hierarchy

PDL Polarization-Dependent Loss

PMD Polarization-Mode Dispersion

PON Passive Optical Network

PRBS Pseudo-Random Binary Sequence

ROADM Reconfigurable Optical Add-Drop Multiplexer

SDH Synchronous Digital Hierarchy

SPM Self-Phase Modulation

SMF Single Mode Fiber

SRS Stimulated Raman Scattering

STM-N Synchronous Transport Module Level-N

TDM Time Division Multiplexing

TDR Time-Domain Reflectometry

WDM Wavelength Division Multiplexing

WDM-PON Wavelength Division Multiplexing in Passive Optical Network

WSS Wavelength Selective Switch

XPM Cross-Phase Modulation

1

1. Introduction

In this chapter, the motivation, the objective and the contribution of this dissertation are exposed.

Scope of the work

Over the last decades, telecommunication systems have experienced a tremendous growth.

The continuous improvement of optical communications has been a constant challenge [1].

In the mid-60’s, of the last century, due to copper cables bandwidth limitations and to the

microwave overloaded connections, researchers have struggled to find new transmission solutions [1,

2]. In 1970, the first optical fibers, with an attenuation lower than 20 dB/km were introduced and rapidly

started to replace coaxial cables [2]. With the launch of the Internet in the decade of 1980s, optical fiber

became a crucial support to the communication system, thereby triggering the deployment of millions of

kilometers of optical cable into the network [3]. Currently, the access networks are an exception, where

cooper cables are still in use. Nevertheless, in the next years, due to the FTTH (fiber to the home)

technology, the copper cables will be completely replaced by optical fibers [4].

Since the development of the WDM (wavelength-division multiplexing) technology, in the late

80s, a single optical fiber can handle dozens of optical channels [3]. This approach became even more

popular with the incorporation of in-line optical amplification based on EDFA (Erbium doped fiber

amplifier) technology within the network, which can transparently amplify simultaneously several optical

channels in the C-band [2, 3]. The EDFA allow the compensation of the optical attenuation introduced

by optical fibers and components allowing longer propagation distances between regenerators. Over

the last two decades, internet data traffic has been growing exponentially and the network largely

expanded worldwide [5]. The DWDM (dense wavelength-division multiplexing), a WDM technique with

shorter channel spacing, was introduced in 1996, providing capacity enhancement [6, 7]. Furthermore,

the nodes on the optical network evolved from electric ADM (add/drop multiplexer) to OADM (optical

add/drop multiplexer) where optical to electric signal conversion is no longer needed [1]. Nowadays, an

operator, with the help of a network software management platform, can add or drop any set of optical

channels into the transport node, thus a quick reconfiguration of the network is possible [6]. This type of

reconfigurable OADM is now referred as ROADM (reconfigurable optical add/drop multiplexer) [6].

Currently, fiber optics communications are the leading communication technology, due to its

numerous advantages in comparison with other technologies, such as higher reliability, higher

bandwidth, lower cross-talk, low-cost and small size [2, 3].

The traffic signal in the optical network has also evolved [4]. The initial protocols were based on

PDH (plesiochronous digital hierarchy) which is an early core network technology based on TDM (time-

domain multiplexing) [4, 8]. It faces many challenges such as limited data rate, absence of normalized

2

optical interfaces and lacks management functionalities [4, 8]. Due to these limitations, other approaches

were conceived. The SDH is a core technology that answers to the PDH needs [4, 8]. With SDH, the

maximum bit rate is approximately 40 Gbit/s and the use of optical interfaces is possible [4, 8]. Also, due

to its complex overhead frame section, this technology has a powerful management system [4, 8].

Considering the evolution of the optical network nodes and due to the demand for higher bit rates, the

current traffic system is OTN (optical transport network), which permits normalized bit rates up to 112

Gbit/s and aggregates all the network traffic (PDH, SHD, Ethernet, etc.) [9]. Nowadays, the optical

network has a massive capacity and optimized management features [9].

By the end of 2016, as displayed in Figure 1.1, the annual global IP (internet protocol) traffic is

expected to overcome the zettabyte (1000 exabytes) threshold, predicting great demand for the

continuous optimization of the current optical systems [5].

Figure 1.1 - Total global IP traffic forecast by 2018 [5].

The high degree of complexity of the actual optical network imposes the physical optical channel

quality assessment to be of high importance, in order to mitigate signal degradation that may cause

system failure [10]. Fiber monitoring technologies, such as OTDR (optical time-domain reflectometer)

and OFDR (optical frequency-domain reflectometer) are currently used for this task [10]. However,

further improvements can still be done due to the high costs and challenges concerning each technique.

The aim of this work is to study a complementary low-cost monitoring technique to evaluate optical fiber

quality in the neighborhood of the transmitter (intra-office section).

3

Motivation

The scope of this dissertation is in the frame of the core network systems. Being the low-cost a

major driver of the proposed solution, a broader range of applications within metro and access systems

is also foreseen. Although monitoring techniques are already available, there is a lack of optimized

solutions in terms of efficiency and price to provide intra-office evaluation. Intra-office is an environment

that is subject to channel impairments (e.g. connector damaging/dirtiness, fiber patchcord

bending/breakage, ribbon-single fan-out damage, fiber splicing, etc.) which would equally benefit from

continuous short-reach monitoring. This work aims to address this target with the key differentiator of

using the optical payload signal itself as probe. The ultimate goal is to create a simple passive optical

add-on block to enable this functionality to all or the most relevant ports on the ROADM nodes.

Objective and work structure

The main objective of this work, is to develop a low-cost fiber fault monitoring technique that

employs the traffic signal to assess the intra-office fiber section. To fulfill the objective, this dissertation

is structured as follows.

In chapter 2, the main physical events behind fiber fault monitoring are explained and the state

of the art techniques are described.

In chapter 3, it is described the structure of the proposed monitoring system and the fault

detection technique is presented.

In chapter 4, the simulated monitoring structure is tested with a continuous optical signal and its

applicability to a Raman amplified network is proposed. The simulation results are compared with

theoretical and experimental ones.

In chapter 5, the proposed monitoring technique is implemented in a simulation environment

and tested with traffic signals. The results are analyzed and the challenges concerning the technique

are addressed.

In chapter 6, the setup of the monitoring structure in a laboratory test scenario is described. The

impairment detection technique that is explained in chapter 3, is validated with experimental data and

further computer processing. The experimental outcome is compared with simulation results.

In chapter 7, it is discussed a scenario for the application of the monitoring technique to

existing/future ROADM linecard designs.

In chapter 8, the final conclusions of this work and suggestions for future work on this topic are

presented.

4

Contribution of this dissertation

The main contributions of this work are:

Proposal of a novel monitoring technique that uses the network traffic signal to assess the

optical channel quality in an intra-office range based on reflectometry mixing.

Test and validation of the proposed technique applied to a Raman amplified network.

Test and validation of the proposed technique employing random and SDH data.

Scientific dissemination

The develop work resulted in the following scientific papers:

Pereira, J.A.; Madureira, M.; André, P.S., Monitoring of the fiber infrastructure with Raman

pump back reflection. Electronic Letters. Paper submitted.

Pereira, J.A.; Madureira, M.; André, P.S., In-service fiber fault location based on

reflectometry mixing, Journal of Optical Communications and Networking. In preparation.

5

2. State of the art

In this chapter, the physical events related to fiber monitoring systems are explained.

Furthermore, the operating principles of the main monitoring techniques are described.

Introduction

Nowadays, most of the technical solutions used to monitor the physical integrity of the optical

fiber networks are based on reflectometry. This technique enables the assessment of the characteristics

of a certain medium by the analysis of the back reflections that continuously occur [10]. The most

common reflectometry techniques, used to monitor optical fibers, are classified as time-domain

reflectometry (TDR) and frequency-domain reflectometry (FDR) [10]. As presented in Figure 2.1, the

attenuation in the optical cables are imposed by several phenomena such as scattering, absorption and

waveguide imperfections. The backscattering of an optical signal is due to microscopic variation in the

fiber material density resulting in light scattered back in the opposite direction of the pulse propagation,

known as Rayleigh backscattering [1]. This phenomenon is one of the main aspects that affect the signal

attenuation in a fiber and is wavelength dependent [3].

Figure 2.1 - Spectral dependence of a single-mode fiber (reproduced from [3]).

Moreover, as the optical signal propagates in the fiber, it may run into waveguide discontinuities,

such as breaks, splice points, connectors or the fiber end [11], resulting in back reflections. This is called

Fresnel reflection, which is used by reflectometers to determine the reflective points location [11]. The

6

Fresnel reflection for an optical signal, depends on polarization of the electric field of the optical signal.

Therefore, the perpendicular and parallel fields to the incident plane, 𝐸// and 𝐸⟘, respectively, must be

treated separately [10]. These components are illustrated in Figure 2.2 for the reflection and refraction

in an interface.

Figure 2.2 - Reflection and refraction of an electromagnetic wave in an interface (reproduced from [10]).

The expressions for the parallel and perpendicular amplitude reflectivity coefficients are

respectively given by [10]

𝜌// =−𝑛2

2 𝑐𝑜𝑠 𝜃1 + 𝑛1√(𝑛22 − 𝑛1

2(𝑠𝑖𝑛 𝜃1)2)

𝑛22 𝑐𝑜𝑠 𝜃1 + 𝑛1√(𝑛2

2 − 𝑛12(𝑠𝑖𝑛 𝜃1)2)

,

( 2.1 )

𝜌⟘ =𝑛1 𝑐𝑜𝑠 𝜃1 − √(𝑛2

2 − 𝑛12(𝑠𝑖𝑛 𝜃1)2)

𝑛1 𝑐𝑜𝑠 𝜃1 + √(𝑛22 − 𝑛1

2(𝑠𝑖𝑛 𝜃1)2) .

( 2.2 )

Considering perpendicular incidence, which is a good approximation for a monomode optical

fiber propagation [10]:

𝜌// = 𝜌⟘ =𝑛1 − 𝑛2

𝑛1 + 𝑛2

.

( 2.3 )

The power reflectivity coefficient is given by

𝑅// = 𝑅⟘ = (𝑛1 − 𝑛2

𝑛1 + 𝑛2

)2

.

( 2.4 )

7

For an open fiber end, the Fresnel power reflection coefficient is approximately 4%, being this

case the worst case scenario [10].

Optical time-domain reflectometer

The OTDR is the best known technique for monitoring fibers and is classified as a TDR

application [10]. The OTDR is linked to one end of the fiber under test and analysis back reflected signals

[11]. Current state of the art OTDR equipment can assess fiber impairments with 5 cm resolution [11].

OTDRs are used in several applications such as measuring overall loss, fusion and mechanical splices

loss, reflectance of connectors and locating faults [11]. Figure 2.3 shows a typical reflectogram produced

by an OTDR.

Figure 2.3 - Typical OTDR reflectogram exposing several reflective events [12].

The basic OTDR structure, as presented in Figure 2.4, consists of a laser source, an optical

circulator (or optical power coupler) and a detector module [10]. The laser source produces optical

pulses, with a certain pulse width, which can be selected for different spatial resolution conditions [11].

The circulator connects the laser source with the fiber and directs the resulting back reflections into the

detector. The detector is a sensitive photodiode or an APD (avalanche photodiode) that measures the

intensity of the returned reflections [11].

8

Figure 2.4 – OTDR functional block diagram [11].

The OTDR technique faces several challenges. Namely, the sensor is designed to measure the

backscattering reflection and, in the presence of a high intensity reflection, it becomes saturated, hence

incapable of continuing the operation [11]. Therefore, if a reflective event comes right after another, the

OTDR will not be able the measure and present that second event. This chain of events leads to the

appearance of dead zones on the reflectogram. The dead zone period is the sum of the duration of the

reflective event and the time the sensor needs to readjust to its maximum sensibility [11]. The final

reflectogram always presents at least one dead zone located at the beginning of the fiber, which is

imposed by losses inherent to the connection of the OTDR with the fiber [11]. Moreover, the larger the

pulse width is, the larger the dead zones are [11]. Furthermore, if the fiber under test returns more than

one strong reflection, some additional mixing terms between different reflectors are produced [13].

These false terms, known as ghost reflections, appear as real reflection events which may impose

difficulties in the analysis of the reflectogram [11]. Additionally, it is not suitable to test short-length optical

fibers, since the reduction of the pulse width to impose resolution, results in a diminished backscattered

signal level [14].

Optical low-coherence reflectometer

The optical low-coherence reflectometer (OLCR) is based on the Michelson interferometer

configuration with a simplified block diagram shown in Figure 2.5 [10]. A low-coherence optical signal is

split (usually 50/50) between the fiber under test and a movable mirror (reference). The back reflected

signal from the fiber and the reference signal are then recombined in the coupler and the product is

received by a photodiode [10]. The reflectogram is obtained by changing the optical path difference

between the signal in both interferometer arms by moving this mirror [15]. If this path difference is shorter

than the light source coherent length, interference will occur [10]. Therefore, the final result is a

reflectogram whose spatial resolution is scaled by the optical signal coherence length [15].

9

Figure 2.5 - OLCR functional block diagram [10].

Figure 2.6 displays an OLCR reflectogram presenting two reflective events.

Figure 2.6 – Typical OLCR reflectogram reporting two reflective events (reproduced from [16]).

However, the OLCR technique has many challenges to overcome in order to be a relevant fiber

monitoring system. For instance, the maximum measurement range is typically lower than 1 m [17].

Also, for proper utilization, the state of polarization between the reflections coming from the two

interferometer arms must coincide [10]. Due to the random nature of the polarization-mode coupling in

the fiber, orthogonality between the test and the reference signal at the receiver may occur, imposing

temporary fading of the beating signal [10]. The OLCR, like the OTDR, has also a restriction either in

spatial resolution or in distance range [18]. Other limitation regarding the OLCR technique is the

10

requirement of a swept optical delay line. The sweeping speed is normally slow since it is accomplished

by mechanical means [10], which does not assure the long-term reliability of the equipment.

Due of its superior spatial resolution, OLCR has also been used in biomedical applications to

provide three dimensional images, with high resolution, in a technique known as OCT (optical coherent

tomography) [10].

Optical frequency-domain reflectometer

The OFDR technique can be categorized in: i) I-OFDR (incoherent OFDR) and ii) C-OFDR

(coherent OFDR) [13]. Like the OLCR, this technique is based on Michelson interferometer,

nevertheless a wavelength-swept laser is used and the final reflectogram is obtained through the Fourier

transform analysis [10].

The I-OFDR can be subdivided into two categories: NA-OFDR (Network analysis OFDR) and I-

FMCW (Incoherent Frequency-Modulated Continuous Wave) [13]. Regarding the NA-OFDR, an optical

signal is intensity modulated by a constant amplitude RF (radio frequency) signal. The resulting signal

is launched into the fiber under test and the reflected signal is detected as function of the modulation

frequency and then processed with a vector signal analyzer [13]. The measured fiber frequency

response is converted to the time domain by inverse Fourier transform [13]. The typical block diagram

of NA-OFDR technique is presented in Figure 2.7.

Figure 2.7 - NA-OFDR functional block diagram [13].

For the I-FMCW, the reflected signal is mixed with the modulated one. The outcome is analyzed

by a spectrum analyzer and then it is converted to time-domain [13]. Figure 2.8 represents the I-FMCW

scheme.

11

Figure 2.8 - I-FMCW functional block diagram [13].

In the C-OFDR the optical signal frequency is swept linearly in time [13]. Then, similarly to

OLCR, the test signal returning from the fiber and the reference signal are coupled and the interference

between them is observed [13]. This interference signal contains the beat frequencies which are

converted to the time-domain by inverse Fourier transform [13]. Figure 2.8 presents C-OFDR block

diagram.

Figure 2.9 - C-OFDR functional block diagram [13].

OFDR technique faces a unique challenge imposed by the lack of laser sources with high-speed

wavelength sweeping. Most of these sources do not sweep continuously across the entire spectral band,

resulting in the degradation of spatial resolution [10, 13]. For the C-OFDR, like in OLCR, the optical

signal polarization state mismatches can be a problem. Moreover, the possibility of ghost reflections

events cannot be ruled out [13]. Due to the great computational work required, an OFDR becomes a

very expensive technique [17]. Figure 2.10 presents a typical OFDR reflectogram exposing several fiber

impairments.

12

Figure 2.10 - Typical OFDR reflectogram exposing several reflective events [18].

Table 2.1 presents the main characteristics for the different optical reflectometry techniques.

Monitoring

technique

Maximum

measuring range

Maximum spatial

resolution Reference

OTDR 200 km 5 cm [19]

I-OFDR 151 km 2 mm [20, 21]

C-OFDR 170 km 50 μm [22, 23]

OLCR 810 m 1.5 μm [24, 25]

Table 2.1 – Most common reflectometers summary.

Coherent reflectometry techniques, such as OLCR and OFDR, have important advantages over

the OTDR technique: larger sensitivity, larger dynamic range and better resolution [17]. The OLCR, due

to its high spatial resolution, is widely used for small component characterization while OFDR is suited

for the measurement of optical modules with extended components [17]. However, the OTDR is the

most used technique for optical fiber assessment nowadays [10]. It can monitor long fiber spans and

the tradeoff between spatial resolution and measurement range is satisfactory [10]. In addition, the

OTDR is a much cheaper technique than both OFDR and OLCR [17].

13

Other monitoring techniques

As already referred, optical channel quality monitoring is a relevant concern in optical

telecommunication systems. Thus, several fiber assessment techniques have been proposed in order

to cut down the implementation and operation costs. The integration of the OTDR function into the

communication equipment, using the optical signal source of the transmitter as probe source has been

discussed [26, 27]. In 2005, Lim et al. proposed an embedded OTDR technique to localize fiber failures

in WDM-PON (wavelength multiplexing in passive optical network) systems [28]. The WDM-PON

technology provides an independent wavelength channel to each user in a bidirectional configuration

[28]. The technique functional diagram is presented in Figure 2.11.

Figure 2.11 – Block diagram of an operating WDM-PON system with embedded OTDR features [28].

Considering an operating PON (passive optical network) system, if the control unit detects that

the upstream optical power of a specific channel is substantially reduced, it automatically triggers an

electrical switch that connects the transmitter with the OTDR pulse generator [28]. Thus, there is no

need for concern about wavelength mismatch between the test light and the target channel of AWG

(arrayed waveguide grating) [28].

In 2006, Schmuck et al. developed a method through embedding a photodetector into the

transmitter module to receive the reflected probe signal [27]. The required functionalities proposed by

Schmuck et al. for a transceiver with embedded OTDR function are shown in Figure 2.12 [27].

14

Figure 2.12 – Block diagram of a transceiver with embedded OTDR features [27].

The OTDR functionality is currently embedded in most of the commercial transceiver equipment

[19]. However, the operation costs for this technique continue to be high due to the probe pulse

generator and the communication still needs to be halted for monitoring purposes [26].

In 2013, Zhao et al. presented a new method that reduces significantly the costs associated to

the fiber assessment [26]. This low-cost and innovative idea uses the traffic signal to monitor optical

fibers [26]. Hence, the interruption of the communication is not required to examine a fiber plant [26].

The block diagram of the monitoring structure is exposed in Figure 2.13.

Figure 2.13 – Zhao et al. monitoring structure block diagram [26].

A tap of the transmitted data traffic that is used as reference signal, and the back reflected signal

resulting from existing faults on the fiber are received by independent photodetectors [26]. In order to

approach their monitoring structure results to a real case scenario, Zhao et al. imposed the source to

transmit SDH data streams [26]. Figure 2.14 a) shows the result from the cross-correlation between the

reference and the back reflected signal while b) presents Zhao et al. monitoring technique reflectogram.

15

Figure 2.14 – a) Plot of the cross-correlation between the traffic reference signal and its back reflection

(reproduced from [26]). b) Zhao et al. monitoring technique reflectogram (reproduced from [26]).

The plot in Figure 2.14 a) presents several correlation peaks with the highest of them

corresponding to the fault position [26]. The authors explain that these regularly arranged peaks indicate

the existence of periodic sections in the traffic signal that are originated by the SDH overhead section

[26]. However, all these correlation peaks are undesired since it may cause misjudgment when

monitoring is in progress. Moreover, it is assumed that the information bits from the SDH frame payload

vary randomly [26]. According to the authors, if this random portion of the traffic signal is extracted, it

will eliminate the misjudgment of the fault caused by the periodic portion [26]. Thus, a method was

proposed to process the reference signal. Firstly, the auto-correlation trace of the reference signal is

achieved and the period 𝑇𝑧 between the correlation periodic peaks is retrieved [26]. Secondly, a section

of the reference signal with period 𝑇𝑧 is arbitrarily selected [26]. This selected portion is then correlated

with the entire reference signal. If this correlation trace presents a gap “window” between correlation

peaks it means that the extracted section is taken from a random section of the signal [26]. In case of

achieving a different correlation trace, another piece of the reference signal is selected and submitted

to the same process [26]. Afterwards, the selected section is extracted from the original reference signal

[26]. The final step for this fiber fault monitoring technique is to correlate the resulting reference signal

with the back reflection from the fiber under test [26]. Zhao et al. reports that a -8 dBm transmitted power

can achieve a detection range of over 23 km with an 8 cm spatial resolution [26].

However, this monitoring structure makes use of an optical circulator which is a relatively

expensive device. Also, Zhao et al. monitoring method is complex and computationally demanding.

16

Proposed monitoring structure

After a detailed summary of the well-known optical reflectometers and analyzing the advantages

and disadvantages of these techniques, a new monitoring structure is proposed. The purpose of this

structure is to address a specific challenge that is not optimized in today reflectometry equipment, which

is the assessment of optical network intra-office distances. Although Zhao et al. have already proposed

a monitoring structure with similar principles, operation expenditure can be further reduced. Since the

telecom market is a capital intensive sector, there is continuous pressure for low-cost equipment

solutions. The proposed structure aims to be an efficient and low-cost solution for the evaluation of the

intra-office fiber distance region.

17

3. Online reflectometry mixing

In this chapter, it is theoretically proposed and analyzed a novel fiber monitoring structure and

the operation of a fiber impairment detection block is explained.

Description

Like the OLCR and OFDR methods, the proposed online reflectometry mixing monitoring

structure is based on the Michelson interferometer, as illustrated by the block diagram of Figure 3.1.

Figure 3.1 – Online reflectometry mixing block diagram.

The data modulated optical signal from the laser source, is injected into an optical coupler, which

splits the optical signal to the fiber under test and to a mirror. The reflected signal from the mirror, which

serves as reference, is mixed with the back reflections from the optical channel and the resulting signal

is received by the photodiode. The detected electric signal is then sampled and processed in order to

clearly distinguish existing faults in the fiber path.

The theoretical model

In order to study the proposed monitoring structure operation, the bidirectional block diagram

was transformed in a unidirectional model, which is exposed in Figure 3.2.

18

Figure 3.2 – Unidirectional block diagram of the online reflectometry mixing for a single reflective event, occuring

after a fiber with length 𝐿.

The optical signal at the laser output, 𝐸𝑜𝑝𝑡(𝑡), is described by

𝐸𝑜𝑝𝑡(𝑡) = 𝐸0(𝑡)𝑐𝑜𝑠 (𝑤0𝑡) ,

( 3.1 )

where 𝐸0(𝑡) is the amplitude of the electric field, 𝑤0 denotes the angular frequency and 𝑡 represents the

time.

The input/output relationship for a 2x2 fiber-based coupler can be represented as [10]

[𝑜1𝑜2

] = [√1 − 𝜀 𝑗√𝜀

𝑗√𝜀 √1 − 𝜀] [

𝑖1𝑖2

] ,

( 3.2 )

where 𝑖1 and 𝑖2 are the input optical signals, 𝑜1 and 𝑜1 are the output signals and 𝜀 defines the coupler

ratio coefficient.

Therefore, the optical signals at the optical coupler outputs, 𝐸𝑎(𝑡) and 𝐸𝑏(𝑡) are described by

𝐸𝑎(𝑡) = √1 − 𝜀 𝐸0(𝑡)𝑐𝑜𝑠 (𝑤0𝑡) ,

( 3.3 )

𝐸𝑏(𝑡) = √𝜀 𝐸0(𝑡)𝑐𝑜𝑠 (𝑤0𝑡 +𝜋

2) .

( 3.4 )

Since reflectance of the mirror, denoted as 𝑀, is considered to be almost 1, then 𝐸𝑏′(𝑡) ≈ 𝐸𝑏(𝑡).

On the other hand, 𝐸𝑎(𝑡) propagates through an optical fiber suffering a power attenuation

coefficient, 𝐴𝑡. From Figure 3.2 it is assumed that the fiber section is a small patchcord, with length 𝐿,

with an open end. In order to reproduce the back reflections imposed by this reflective event, the output

signal from the first fiber section is multiplied by a reflection coefficient 𝑅. Also, the back reflected signal

is affected with the phase introduced by a fiber section, 𝜙. Therefore, the electric field back reflected at

the coupler input, 𝐸𝑎′(𝑡), is given by

19

𝐸𝑎′(𝑡) = √𝑅√1 − 𝜀𝐴𝑡2𝐸0(𝑡)𝑐𝑜𝑠 (𝑤0𝑡 + 2𝜙) ,

( 3.5 )

with

𝐴𝑡 = √10−𝛼𝐿10 ,

( 3.6 )

where 𝛼 is the fiber attenuation coefficient (dB/km) and 𝐿 the fiber length in km. The phase difference 𝜙,

imposed by the fiber, is given by [29]

𝜙 =2𝜋

𝜆𝑛𝐿 ,

( 3.7 )

where 𝜆 is the optical signal wavelength and 𝑛 is the fiber refractive index, which is assumed to be ≈

1.45 [29].

However, the actual laser source coherence length reaches up to few dozens of meters [3].

Therefore, the interference imposed by the phase difference 𝜙 is only noticed when a back reflection

occurs at the first meters of the optical fiber.

Afterwards, signals 𝐸𝑎′(𝑡) and 𝐸𝑏′(𝑡) are mixed together when back reflected to the optical

coupler. Signals 𝐸1(𝑡) and 𝐸2(𝑡) are respectively given by

𝐸1(𝑡) = √1 − 𝜀√1 − 𝜀√𝑅𝐴𝑡2𝐸0(𝑡)𝑐𝑜𝑠 (𝑤0𝑡 + 2𝜙) + √𝜀√𝜀𝑀𝐸0(𝑡) 𝑐𝑜𝑠 (𝑤0𝑡 + 𝜋) ,

( 3.8 )

𝐸2(𝑡) = √𝜀√1 − 𝜀√𝑅𝐴𝑡2𝐸0(𝑡)𝑐𝑜𝑠 (𝑤0𝑡 + 2𝜙 +

𝜋

2) + √1 − 𝜀√𝜀𝑀𝐸0(𝑡)𝑐𝑜𝑠 (𝑤0𝑡 +

𝜋

2) .

( 3.9 )

Signal 𝐸2(𝑡) is received by a photodetector, producing an output electric current, 𝐼2(𝑡) [3]:

𝐼2(𝑡) = |𝐸2(𝑡)|2𝑅𝜆 ,

( 3.10 )

where 𝑅𝜆 is the photodiode responsivity.

3.2.1 State of polarization scrambling

When an optical signal propagates through an optical fiber, its state of polarization changes.

These changes in polarization depend on several aspects such as PMD (polarization mode dispersion)

or PDL (polarization-dependent loss) [1]. PMD is a form of modal dispersion where two orthogonal

polarization fields propagate at different speeds, due to physical imperfections in the fiber core, imposing

20

the spreading of the pulses. The PDL phenomena degrades the signal since the two polarizations suffer

different losses, due to the random nature of fiber imperfections [1].

Considering that the optical signal from the laser source is polarized along �̂�, the state of

polarization of the signals that pass through a fiber is randomly scrambled. Thus, the final expression

for signal 𝐸𝑎′(𝑡) is given by

𝐸𝑎′(𝑡) = (1 − 𝛽)√𝑅√1 − 𝜀𝐴𝑡2𝐸0(𝑡) 𝑐𝑜𝑠(𝑤0𝑡 + 2𝜙) �̂� + 𝛽√𝑅√1 − 𝜀𝐴𝑡

2𝐸0(𝑡) 𝑐𝑜𝑠(𝑤0𝑡 + 2𝜙 + 𝜑) �̂� ,

( 3.11 )

where �̂� and �̂� are the versors representing the polarization in two orthogonal directions, 𝛽 is the intensity

amount of �̂� direction polarization and 𝜑 is the phase difference between �̂� and �̂� orthogonal fields.

Although the polarization parameters are time dependent, they are assumed to be constant for simplicity

purposes.

In order to observe the polarization state of an optical signal, the Poincaré sphere is used. The

azimuth 𝜓, and ellipticity 𝜒, of the orthogonal optical field are needed to represent the signal polarization

state in the Poincaré sphere [10]:

𝑡𝑎𝑛 2𝜓 =2𝐸𝑜𝑥(𝑡)𝐸𝑜𝑦(𝑡)

𝐸𝑜𝑥2 (𝑡) − 𝐸𝑜𝑦

2 (𝑡)𝑐𝑜𝑠 𝜑 0 ≤ 𝜓 < 𝜋 ,

( 3.12 )

𝑠𝑖𝑛 2𝜒 =2𝐸𝑜𝑥(𝑡)𝐸𝑜𝑦(𝑡)

𝐸𝑜𝑥2 (𝑡) + 𝐸𝑜𝑦

2 (𝑡)𝑠𝑖𝑛 𝜑 −

𝜋

4< 𝜒 ≤

𝜋

4 ,

( 3.13 )

where 𝐸𝑜𝑥(𝑡) and 𝐸𝑜𝑦(𝑡), are the intensity values of fields �̂� and �̂� respectively, which are given for the

signal 𝐸𝑎′(𝑡) by

𝐸𝑜𝑥(𝑡) = (1 − 𝛽)√𝑅√1 − 𝜀𝐴𝑡2𝐸0(𝑡) ,

( 3.14 )

𝐸𝑜𝑦(𝑡) = 𝛽√𝑅√1 − 𝜀𝐴𝑡2𝐸0(𝑡) .

( 3.15 )

Depending on the point location in the sphere, the signal polarization can be denominated as

linear, elliptical or circular [10]. The exact point location in the sphere is achieved from the spherical

coordinates given by

𝑥 = 𝑐𝑜𝑠 2𝜒 𝑐𝑜𝑠 2𝜓 ,

𝑦 = 𝑐𝑜𝑠 2𝜒 𝑠𝑖𝑛 2𝜓 ,

𝑧 = 𝑠𝑖𝑛 2𝜒 .

( 3.16 )

21

Figure 3.3 represents three of the possible polarization states. The quantities 𝑠1, 𝑠2 and 𝑠3 are

the observables of the polarized field and are denominated as Stokes polarization parameters [30].

Figure 3.3 - Polarization states displayed over the Poincaré sphere: a) Linear b) Circular c) Elliptical.

Hence, the final expression that rules the optical signal 𝐸2(𝑡) is

𝐸2(𝑡) = [((1 − 𝛽)√𝑅𝐴𝑡2 𝑐𝑜𝑠(2𝜙) + √𝑀) �̂� + 𝛽√𝑅𝐴𝑡

2 𝑐𝑜𝑠 (2𝜙 +𝜋

2) �̂�] √𝜀(1 − 𝜀)𝐸0(𝑡) 𝑐𝑜𝑠(𝑤0𝑡 +

𝜋

2) .

( 3.17 )

3.2.2 Multiple reflection scenario

In section 3.2, it was presented the unidirectional monitoring model for a fiber section with one

reflective event. However, the fiber path that connects two network nodes may present several reflective

events. This method was expanded considering several concatenated fiber links with reflective events

between them. In order to represent multiple reflection scenarios, the unidirectional model must be

extended as shown in Figure 3.4.

22

Figure 3.4 – Unidirectional block diagram of the online reflectometry mixing for a multiple reflection scenario.

By adding several fiber systems in parallel to the original scheme from Figure 3.2, one can

represent the presence of several reflective events in a fiber path. For 𝑛 fiber impairments, the electric

field, 𝐸2(𝑡), that reaches the photodiode is given by

𝐸2(𝑡) = [(𝐹𝑟𝛽𝑒𝑗𝜑 + √𝑀)�̂� + 𝐹𝑟(1 − 𝛽)�̂�]√𝜀(1 − 𝜀)𝐸0(𝑡) 𝑐𝑜𝑠 (𝑤0𝑡 +𝜋

2) ,

( 3.18 )

where

𝐹𝑟 = ∑ √𝑅𝑘

𝑛

𝑘=1

∏ [(1 − 𝑅𝑚−1)(𝐴𝑡𝑚𝑒𝑗ɸ𝑚)

2]

𝑘

𝑚=1

; 𝑅0 = 0 .

( 3.19 )

Fault monitoring module

In order to be able to distinguish eventual back reflections from the fiber, the output detected

signal must be submitted to a digital signal processing step. This step is accomplished by a monitoring

module that employs the correlation technique. The auto-correlation function is used as a measure of

similarity between the signal’s time instants. The auto-correlation of a 𝑥 signal is given by [31]

𝜙𝑥𝑥(𝑡) = ∑ 𝑥∗(𝜏)𝑥(𝑡 + 𝜏)

∞

𝜏=−∞

.

( 3.20 )

In order to explain the result from the auto-correlation of the monitoring structure output, the

case in which the laser source transmits optical pulses with width 𝑊 and period 𝑇 is assumed. In this

explanation, it is considered that the reflection occurs in a distance longer than the optical signal

coherent length. Figure 3.5 a) presents the plot of the signal evolution over time while b) presents the

trace from its respective auto-correlation.

23

Figure 3.5 – a) Temporal evolution of the input signal. b) Auto-correlation plot of the input signal.

The optical signal in Figure 3.5 a) is described by

𝐻1(𝑡) = ∑ 𝐴𝑟𝑒𝑐𝑡 (𝑡 − 𝑘𝑇

𝑊) ,

∞

𝑘=1

( 3.21 )

where 𝐴 is the signal amplitude and 𝑟𝑒𝑐𝑡(𝑡) is the rectangular function.

The trace from the auto-correlation of this signal is a well-known result from the literature [32].

As already mentioned, when an impairment exists, the monitoring structure output signal is the

combination of the back reflection and the reference signal. When an optical signal propagates in a fiber,

it suffers a time delay, depending on the fiber length. Thus, the back reflection is delayed in comparison

with the reference signal. The back reflection time delay, 𝛥𝑡, introduced by an optical fiber is given by

[3]

𝛥𝑡 =2𝑛𝐿

𝑐 ,

( 3.22 )

where 𝐿 is the reflective event location relative to the optical coupler, 𝑛 is the fiber refractive index and

𝑐 is the speed of light.

Considering the optical signal 𝐻1(𝑡), its respective back reflection signal is described by

𝐻2(𝑡) = ∑ 𝐴𝑅(1 − 𝜀)𝑟𝑒𝑐𝑡 (𝑡 − 𝑘𝑇 − 𝛥𝑡

𝑊) ,

∞

𝑘=1

( 3.23 )

24

where 𝑅 is the reflection coefficient of the reflective event and 𝜀 corresponds to the coupler ratio

coefficient.

The temporal evolution of the output signal from the monitoring structure is described by the plot

in Figure 3.6 a), while b) presents the trace from its respective auto-correlation.

Figure 3.6 – a) Temporal evolution of the output signal for a reflective event at distance 𝐿. b) Auto-correlation plot

of the output signal.

The monitoring structure output signal that is presented in Figure 3.6 a) is described by

𝐻𝑡(𝑡) = (𝜀𝐻1(𝑡) + 𝐻2(𝑡))𝜀 .

( 3.24 )

The highest peaks in the auto-correlation trace of the output signal are due to the correlation

between the pulses of the reference signal. By observation of Figure 3.6 b), it is noticeable that there

will be an adjacent peak at each side of the highest peaks. The adjacent peaks are due to the correlation

between the pulses of the reference signal and the ones of the back reflection signal. As Figure 3.6 a)

suggests, the auto-correlation of the output signal, allows the retrieval of information regarding the back

reflected signal.

The highest peaks in the auto-correlation trace of the output signal add difficulty to the analysis

of the back reflection time delay. Therefore, in order to exclude these peaks from the trace, the designed

monitoring module computes the ratio between auto-correlation values of the monitoring structure output

signal with the auto-correlation values of the input signal. Since the auto-correlations have different

magnitude orders, before the ratio is computed, the values of both auto-correlations must be normalized.

Also, the normalized values from both auto-correlations may not match. Therefore, in order to mitigate

the undesired peaks and to accentuate the adjacent peaks, the normalized auto-correlation values of

the output signal are multiplied by a factor μ. This factor is the quotient between the mean value of an

25

interval of the normalized auto-correlation values of the output signal, and the mean value of the same

interval of the normalized auto-correlation values of the input signal. Figure 3.7 presents the outcome

of the designed technique.

Figure 3.7 – Proposed monitoring technique output for a reflective event at distance 𝐿.

As Figure 3.7 suggests, the back reflection time delay is retrieved by the time of the first adjacent

peak. All the remaining peaks are correlation mixing terms and are not considered in the analysis of the

trace. Therefore, the plot window dimension must be adjusted for a clear detection of fiber impairments.

This introduces a limitation to the measuring range, nevertheless the target intra-office fiber length range

is covered. Also, if 𝛥𝑡 >𝑇

2, then each back reflected pulse will have a significant time difference from its

original reference pulse. In fact, in this situation, the back reflected pulse will be closer to the following

reference pulse, resulting in an intersection between adjacent peaks in the final output trace. Thereby,

the first adjacent peak in the final trace will be an undesirable correlation mixing term that does not

represent the real time delay of the back reflection. Hence, the back reflection time delay must not

exceed 𝑇

2. The red dashed lines mark the adjustment window borders. Moreover, the time axis of the

plot is then converted to distance using equation (3.22).

In order to summarize the proposed monitoring technique, a schematic is presented in Figure

3.8.

26

Figure 3.8 – Schematic of the monitoring module algorithm.

Simulation model

Optilux toolbox [33], an open source optical system simulator containing several predefined

Matlab® functions, was chosen to implement the numerical simulation scenario. Since the simulator

does not support backscattering propagation, the unidirectional models, exposed in Figure 3.2 and

Figure 3.4, were used to simulate real scenarios. For all the simulation scenarios, there are constant

parameters concerning the transmitted optical signal from the laser source and the fiber under test.

These parameters are set to typical values. Table 3.1 presents the used transmission parameters.

Optical power 6 dBm

Wavelength 1550 nm

Linewidth Negligible

Peak cumulated non-linear phase 0.4π

Extinction ratio 13 dB

Table 3.1 - Transmission parameters for simulation purposes.

27

Additionally, a 10 dB attenuation is considered at the laser output in order to mitigate the effects

that the back reflected signals might impose on the laser source. Also, a 10 dB attenuation is included

at the photodiode input to avoid its saturation.

Table 3.2 presents the optical fiber parameters for simulation scenarios.

Fiber power attenuation 0.2 dB/km

Effective area 80 μm2

Non-linear index 2.7x10-20

Dispersion (at 1550 nm) 17 ps/nm/km

Slope (at 1550 nm) 0 ps/nm2/km

Table 3.2 – Optical fiber parameters for simulation purposes.

The simulated optical fiber takes into account the effects of the GVD (group velocity dispersion),

SPM (self-phase modulation) and XPM (cross-phase modulation).

The optical coupler loss insertion loss is negligible, therefore it is not considered in the simulator.

28

29

4. Continuous signal operation

In this chapter, the simulated monitoring structure is operated with a continuous optical signal

to test its applicability to a Raman amplified network. Raman amplification makes use of the SRS

(stimulated Raman scattering) occurring in silica fibers resulting in energy transference from the pump

to the traffic signal [3]. Considering the proposed monitoring structure, it is analyzed the possibility of

monitoring the fiber infrastructure employing the Raman pump back reflection. Moreover, the simulation

outcome is compared with theoretical and experimental results for validation purposes.

Single reflective event

The first tested scenario represents a case with one reflective event. It was considered a

continuous optical input signal and a coupling ratio of 80/20. The test optical fiber length was 1 meter,

followed by a reflective event with a reflection coefficient of 4%.

Although Raman amplified systems need a pump signal with high intensity power [3], the optical

input power was kept at 6 dBm for simplicity purposes.

The numerical analysis for this scenario was performed with a Monte Carlo based method. This

method is employed to predict the behavior of a system by randomly varying the value of the input

variables [34]. The model from Figure 3.2 was simulated repeatedly until a sufficient number of

combination of input variables was covered, being the output the sum of the results from every

simulation cycle. Each cycle produced a histogram, which presented the number of occurrences as a

function of the photodiode current values. In order to include the most possible combinations, several

system parameters were changed in each cycle: i) the reflection coefficient value, 𝑅, was changed from

0 to 4%, ii) the phase imposed by fiber sections, ɸ, was varied from 0 to 2π and iii) the signal polarization

at the receiver was randomly distributed all over the Poincaré sphere. The Monte Carlo simulation was

tested considering 10000 simulation cycles. This number of cycles was determined after repeating the

simulation and achieving consistent and almost identical results. The result from the Monte Carlo

simulation applied to the described scenario is exposed in Figure 4.1.

30

Figure 4.1 – Photodiode electric output current histogram for a single reflection scenario obtained via Monte Carlo

simulation. The red arrows indicate the theoretical extreme current values.

In order to validate the accuracy of the numerical simulation scenario, the analytical description

of the system was considered. The theoretical maximum and minimum photodiode current values are

obtained for the case when the reflected signal has the same polarization state as the reference signal.

For this case, signal 𝐸2(𝑡) is given by

𝐸2(𝑡) = (√𝑅𝐴𝑡2 𝑐𝑜𝑠(2𝜙) + √𝑀)√𝜀(1 − 𝜀)𝐸0(𝑡)𝑐𝑜𝑠 (𝑤0𝑡 +

𝜋

2)�̂� .

( 4.1 )

Thus, the photodiode current 𝐼2(𝑡) is given by

𝐼2(𝑡) =1

2(𝐸𝑥1

2 (𝑡) + 𝐸𝑥22 (𝑡) + 2𝐸𝑥1(𝑡)𝐸𝑥2(𝑡) 𝑐𝑜𝑠(2𝜙))𝑅𝜆 ,

( 4.2 )

where

𝐸𝑥1(𝑡) = √𝜀(1 − 𝜀)√𝑅𝐴𝑡2𝐸0(𝑡) ,

( 4.3 )

𝐸𝑥2(𝑡) = √𝜀(1 − 𝜀)√𝑀𝐸0(𝑡) .

( 4.4 )

From equation (4.2) it was retrieved that the minimum electric current value occurs for 𝜙 = ±𝜋

2,

and the maximum value occurs for 𝜙 = 0 𝖵 𝜙 = 𝜋 . Thereby, assuming that the input parameters are the

ones of the simulation scenario, the maximum and minimum photodiode current analytical values are

8.27 μA and 3.66 μA, respectively. These values are indicated in Figure 4.1 with the vertical arrows,

perfectly matching with the limits of the numerical analysis. Therefore, one can conclude that the

simulator is well designed for this specific case scenario.

31

Multiple reflective events

The same validation procedure, as in section 4.1, was applied to the case with two reflective

events. In order to achieve two back reflections, two fiber sections must be sequentially linked to the

monitoring structure. A 1 meter long fiber section was connected to a second fiber section with 1

kilometer extension, and at its end, a reflective event with a reflection coefficient of 4% was

implemented. An additional reflective point with a reflection ratio of 1% was considered between the

fiber sections.

The Monte Carlo simulation was applied to this scenario and its output results are presented in

Figure 4.2. Several input parameters were changed in each cycle: i) the reflection coefficient, 𝑅1, was

varied from 0 to 1% and 𝑅2 was varied from 0 to 4%, ii) the phase imposed by fiber sections was varied

from 0 to 2π and iii) the signal polarization was randomly distributed all over the Poincaré sphere.

Figure 4.2 – Photodiode electric output current histogram for a double reflection scenario obtained via Monte

Carlo simulation. The red arrows indicate the theoretical extreme current values.

In order to retrieve the photodiode maximum and minimum theoretical current values, the case

when the reflected signals have the same polarization state of the reference signal was considered. For

this situation, signal 𝐸2(𝑡) is given by

𝐸2(𝑡) = (√𝑅1𝐴𝑡12 𝑐𝑜𝑠(2𝜙1) + √𝑅2(1 − 𝑅1)𝐴𝑡1

2 𝐴𝑡22 𝑐𝑜𝑠(2𝜙1 + 2𝜙2) + √𝑀)√𝜀(1 − 𝜀)𝐸0(𝑡) 𝑐𝑜𝑠 (𝑤0𝑡 +

𝜋

2) �̂� ,

( 4.5 )

where 𝜙1 and 𝜙2 are the phases imposed by fiber sections 𝐿1 and 𝐿2, respectively, and 𝐴𝑡1 and 𝐴𝑡2 are

the power attenuation coefficients from fiber sections 𝐿1 and 𝐿2, respectively.

Thus, the photodiode current 𝐼2(𝑡) is given by

32

𝐼2(𝑡) =1

2[𝐸𝑥11

2 (𝑡) + 𝐸𝑥122 (𝑡) + 2𝐸𝑥11(𝑡)𝐸𝑥12(𝑡) 𝑐𝑜𝑠(2𝜙2) + 𝐸𝑥2

2 (𝑡) +

2𝐸𝑥2(𝑡)√𝐸𝑥112 (𝑡) + 𝐸𝑥12

2 (𝑡) + 2𝐸𝑥11(𝑡)𝐸𝑥12(𝑡) 𝑐𝑜𝑠(2𝜙2) 𝑐𝑜𝑠 (𝑡𝑎𝑛−1(𝐸𝑥11(𝑡) 𝑠𝑖𝑛(2𝜙1+

𝜋

2)+𝐸𝑥12(𝑡) 𝑠𝑖𝑛(2𝜙1+2𝜙2+

𝜋

2)

𝐸𝑥11 (𝑡)𝑐𝑜𝑠(2𝜙1+𝜋

2)+𝐸𝑥12 (𝑡)𝑐𝑜𝑠(2𝜙1+2𝜙2+

𝜋

2)

−𝜋

2))] 𝑅𝜆 ,

( 4.6 )

where

𝐸𝑥11(𝑡) = √𝜀(1 − 𝜀)√𝑅1𝐴𝑡12 𝐸0(𝑡) ,

( 4.7 )

𝐸𝑥12(𝑡) = √𝜀(1 − 𝜀)𝐴𝑡12 𝐴𝑡2

2 √𝑅2(1 − 𝑅1)𝐸0(𝑡) ,

( 4.8 )

𝐸𝑥2(𝑡) = √𝜀(1 − 𝜀)√𝑀𝐸0(𝑡) .

( 4.9 )

The maximum and minimum values for 𝐼2(𝑡) are retrieved from equation (4.6). Assuming the

input parameters used in the simulator, the maximum and minimum current values are 9.55 μA and 2.45

μA, respectively. These values are indicated in Figure 4.2 with the red arrows which match with the

limits of the numerical analysis.

With the purpose of representing cases when there is up to 4 reflections in an optical link, 4

fibers sections of 1 kilometer were sequentially connected. At the end of each fiber section, a reflective

event with a reflection coefficient of 1% was implemented. Furthermore, the Monte Carlo simulation was

applied to each case with the same varying input parameters. Figure 4.3 presents the Monte Carlo

method output for situations with up to 4 reflective events.

Figure 4.3 – Photodiode electric output current histograms for different reflective event scenarios obtained with

Monte Carlo simulation.

33

As expected, the Monte Carlo simulation output histogram is narrower for cases with less

reflective events. With the increasing number of reflective events, the photodiode outputs current values

with higher intensity.

In Figure 4.4 the connected black squares represent, for different number of reflective events,

the values of the FWHM (full width at half maximum) of the Monte Carlo simulation output divided by

the output histogram central current value. The remaining connected symbols represent, for different

number of reflective events, the normalized FWHM from the Monte Carlo simulation output when its

varying inputs are multiplied by a factor θ. Moreover, the Poincaré spheres represent the state of

polarization distribution when factor θ is 1, 0.1 and 0.01. Also, Figure 4.4 shows experimental results

that were obtained at Coriant facilities in Alfragide (Lisbon). The individual blue asterisks in Figure 4.4

represent the normalized FWHM of the histograms of the experimental monitoring structure output

signal, for different number of reflective events.

Figure 4.4 – Normalized FWHM of the Monte Carlo simulation output as a function of the number of reflective

events considering different values for factor θ. The independent blue asterisks represent the normalized FWHM

of histograms of experimental results. The Poincaré spheres represent the state of polarization distribution when

θ is 1, 0.1 and 0.01.

Experimental results for a continuous signal operation

Figure 4.5 shows a snapshot of the experimental test scenario of the monitoring structure with

a continuous optical signal.

34

Figure 4.5 - Experimental scenario for a single refletive event.

The optical source was the continuous laser Agilent 81980A which was linked to the monitoring

structure input. The used optical coupler has a specific ratio of 80/20, however, after an experimental

characterization, it was calculated that the coupler presents an insertion loss of 0.8 dB and the real

coupling ratio was approximately 81.5/18.5. The used mirror was connected to a SMF (single mode

fiber) pigtail and has a specific reflectance of 99%. Also, SMF optical fibers with 1 km extension were

sequentially linked to the coupler output and terminated with PC (physical connector) connectors to

achieve several reflective events. With PC terminated fibers, back reflections with a reflection coefficient

of 4% were accomplished [35].

The monitoring structure output optical signal was detected with a photodiode. The photodiode

output electrical current as a function of the input optical power, for a wavelength of 1550 nm, is

displayed in Figure 4.6.

Figure 4.6 – Photodiode output current as function of the input power.

35

The data points in Figure 4.6 were linear fitted, which allowed the estimation of the photodiode

responsivity to be 0.952 A/W.

The photodiode’s output signals were acquired by the Tektronix DPO4104 oscilloscope which

has a bandwidth of 1 GHz and a sampling rate of 5 GS/s. The acquired signals were loaded to a

computer for the histogram’s achievement.

The analyzed scenarios were demonstrated experimentally. For these test scenarios, the

transmission parameters were set to the same values as in the simulator. Figure 4.7 a) presents the

photodiode´s output current evolution over time for the case with two reflective events and b) shows the

respective histogram.

Figure 4.7 - a) Plot of experimental photodiode’s output current over time for a double reflective event case. b)

The histogram of the current values from plot a).

Considering the experimental outcome that is marked in Figure 4.4 with blue asterisks, it is

demonstrated that through the Monte Carlo simulation, for θ between 0.01 and 0.05, one can reproduce

the experimental results. From Figure 4.4 it is demonstrated the feasibility to use the Raman pump signal

to monitor an optical link with several impairments.

36

37

5. Fault diagnosis with real traffic

In this chapter, the simulated monitoring structure is operated with random and SDH data

signals and the fiber impairment detection method is applied. The results are analyzed and the

challenges concerning the proposed technique are addressed.

Fault detection with random data correlation

In order to demonstrate how fiber fault detection can be achieved with correlation technique, the

test scenario described in Figure 5.1 was simulated [36].

Figure 5.1 – Block diagram for correlation testing with PRBS data.

The transmitted optical signal was amplitude modulated with a NRZ (non-return to zero) format

by PRBS (pseudo-random binary sequence) data, and a bit rate of 1 Gbit/s was considered. The

coupling ratio for this scenario was 50/50.

Due to a open fiber end, a back reflection with a reflection coefficient of 4% is achieved. Figure

5.2 presents the result of the correlation between the reference and the back reflected signal for a fiber

with 15 meters length. For the illustrated example, the correlation plot exposes a peak at 150 ns, which

corresponds to the reflective event time delay. Considering equation (3.22), the time delay of this event

can be converted to the distance relative to the optical coupler. This result shows that it is possible to

retrieve the fiber impairment location with random data as traffic signal.

38

Figure 5.2 – Plot of the correlation between the random data reference signal and its back reflection which was

imposed by an open-ended fiber with 15 meters length.

The results from this approach have been extensively analyzed [36-39]. Takushima et al. used

a similar structure to retrieve fiber fault location and developed an algorithm to improve the dynamic

range of the output reflectogram.

Besides the high cost of the optical circulator, it is assumed that the traffic signal varies in

random form, which does not correspond to a real environment situation [26].

Reflectometry mixing with random data

Considering the proposed monitoring structure that is described in chapter 3, it is relevant to

realize which information can be retrieved from auto-correlation trace of the output signal. To test this

approach, an optical signal was transmitted with the characteristics described in section 5.1. Moreover,

the coupling ratio for this experiment was 80/20. A reflective event with a reflection ratio of 2% of the

incident optical signal was implemented at 20 meters from the optical coupler. Figure 5.3 presents the

auto-correlation of the photodiode electric output current.

39

Figure 5.3 – Auto-correlation plot of the monitoring structure output signal for a fiber impaiment at 20 meters from

the optical coupler.

Apparently, one cannot detect the fiber impairment from the plot in Figure 5.3. Therefore, using

the proposed monitoring structure, a simple auto-correlation trace of the output signal is not suitable to

detect reflective events.

In order to accentuate the reflective event peak, it is necessary to compute the auto-correlation

from the input signal and to apply the designed method that is described in section 3.3. Figure 5.4 a)

represents the result from the designed method while b) shows the outcome after using a smoothing

filter. This filter employs a local regression using weighted linear least squares and a first degree

polynomial model. As Figure 5.4 presents, the BNSR (background noise suppression ratio) value

indicates the magnitude of the detected peak, which is retrieved considering an imaginary horizontal

line at the mean noise level.

Figure 5.4 – Proposed technique reflectogram for a fiber impaiment at 20 meters from the optical coupler using

PRBS data: a) without smoothing filter; b) with smoothing filter.

40

Figure 5.4 demonstrates the operation of the proposed technique. This approach is able to

distinguish reflective events with lower intensities. Back reflections with 0.5% of the intensity of the

incident optical signal can easily be detected by this method.

Considering the previous case, Figure 5.5 presents how the BNSR varies with the increasing

coupler ratio coefficient, 𝜀.

Figure 5.5 - BNSR as a function of the coupler ratio coefficient. For each case, it was considered a reflective

event with a reflection ratio of 2% at 20 meters from the coupler.

From Figure 5.5, one can conclude that the lower the coupler ratio coefficient, the more

accentuated the detected peak is. For lower coupling coefficients, the back reflections will have stronger

intensities.

The proposed technique was applied to cases with multiple reflective events. Figure 5.6

represents the outcome for the case when there are two reflective events in a fiber link. For this scenario,

one reflective point was located at 12 meters while the other was located at 139.6 meters from the

optical coupler. Moreover, the first reflective event had a reflection coefficient of 0.5% and the second

had a reflection coefficient of 1%. The optical signal was transmitted with the characteristics described

in section 5.1, and due to the results exposed in Figure 5.5, a coupling ratio of 90/10 was considered in

this experiment.

41

Figure 5.6 – Proposed technique reflectogram using PRBS data for reflective events at 12 and 139.6 meters from

the optical coupler.

Reflectometry mixing with real traffic

For a real operating network, the data traffic does not have a random form [4]. For instance, an

STM-1 (synchronous transport module level-1) frame has overhead information whose purpose is to

implement communication protocol features [4, 8]. Figure 5.7 a) represents the STM-1 frame structure,

which is the SDH basic frame. The SDH frame is transmitted row by row. Figure 5.7 b) presents the

content of the overhead section from the STM-1 frame.

Figure 5.7 – a) STM-1 frame structure. b) Overhead section of the STM-1 frame [8].

The first 6 bytes from the overhead section of the STM-1 frame (represented in Figure 5.7 by

A1 and A2) are denominated by frame alignment bytes [4, 8]. These bytes have a constant bit pattern

42

and are used to indicate the beginning of the frame [4, 8]. All the other overhead and payload bytes vary

in each transmitted frame [4]. Therefore, an STM-1 frame can be approximated as shown in Figure 5.8.

Figure 5.8 – Aproximatted STM-1 frame structure for simulation purposes.

In the approximated STM-1 frame structure, the first 6 bytes represent the alignment bytes while

the remaining bytes are pseudo-random bit sequence data.

The STM-N frame is compiled through the assembling of N STM-1 frames in a process called

octet multiplexing [4]. In this arrangement, each byte from each STM-1 frame is sequentially inserted in

the final frame [4]. Thus, considering the STM-4 frame, the beginning of the frame starts with 4x6=24

alignment bytes.

In order to simulate the broadcast of STM-1 signals, the frame exposed in Figure 5.8 was

repeatedly transmitted at the laser source of the monitoring structure. The transmitted optical signal was

amplitude modulated with an NRZ format by the STM-1 simulated data, and a bit rate of 155.52 Mbit/s

(STM-1 bit rate [4]) was considered. The coupling ratio for this scenario was 90/10. A reflective event

with 1.5% of the intensity of the incident optical signal was implemented at 302 meters from the optical

coupler.

The auto-correlation trace from the STM-1 input traffic signal is a relevant matter of discussion.

Regarding the example from section 3.3, one can consider that the period 𝑇 corresponds to the STM-1

frame time while the pulse width 𝑊 corresponds to the STM-1 start of the frame section. Figure 5.9

presents the auto-correlation trace of this input signal normalized to its maximum value. This result is

analogous to the auto-correlation trace of the input signal that is described in section 3.3.

43

Figure 5.9 – Auto-correlation plot of the approximated STM-1 input signal.

Since the time interval of an SDH signal frame is 125 μs [4], the auto-correlation trace of the

input signal presents peaks at multiples of 125 μs.

The proposed technique, which is described in section 3.3, was applied to this case. The plot of

the proposed technique outcome without plot window adjustment is presented in Figure 5.10.

Figure 5.10 – Proposed technique outcome without plot window adjustment for a single reflective event at 302

meters from the optical coupler.

As expected, the first detected peak appears at 3020 ns, which corresponds to the time delay

of the back reflection imposed by the implemented reflective event.

44

Taking into account the example from section 3.3, the appearance of the consecutive peaks in

the trace from Figure 5.10 is comprehensible. Considering the SDH frame period, the time delay of any

reflective event must not exceed 62.5 μs. The red dashed lines in Figure 5.10 mark the adjustment

window borders.

Besides the peaks originated by the start of the frame section, other peaks from eventual

correlations may appear in the technique output trace. However, considering that the remaining bytes

in the SDH frame do not have a constant pattern, it is unlikely that any other peak has a BNSR with

such high order of magnitude. In fact, all the other peaks appear in the final plot mixed in the background

noise.

Figure 5.11 presents the online reflectometry mixing reflectogram for this case scenario.

Figure 5.11 – Online reflectometry mixing reflectogram for a single reflective event at 302 meters from the optical

coupler.

Therefore, a technique for fiber impairment detection, using the traffic signal as probe, is

demonstrated.

Figure 5.12 shows how the BNSR value varies for reflective events with a reflection coefficient

of 1.5% at different locations in the fiber. For these experiments, the coupling ratio was considered to

be 90/10.

45

Figure 5.12 – BNSR as a function of the reflective event location in the fiber. For each case, a reflection

coefficient of 1.5% and a coupling ratio of 90/10 were considered.

From this plot, it is concluded that, the BNSR value decreases for fiber impairments at longer

distances. This result might be related with the power attenuation imposed by the optical fiber.

Technique limitations analysis

To analyze the monitoring technique limitations, it was prepared a simulation scenario where

the transmitted optical signal was amplitude modulated with NRZ format by STM-1 data. Moreover, the

coupling ratio was considered to be 90/10.

5.4.1 Multi-channel impact

The previous results were achieved considering the transmission of a single channel. However,

transport optical transmission systems make use of DWDM technology with up to 96 channels in the C-

band [7]. Thus, it is relevant to realize the impact of the multi-channel on the monitoring technique. It

was considered a 10 channel system with a channel spacing of 50 GHz and a reflective event was

implemented at 25 meters from the optical coupler. The goal is to realize if the BNSR depends on the

number of channels. The obtained results indicate that, as the number of channels increases, a slight

decrease on BNSR is noticed. The BNSR of the technique output for a 10 channel situation is

approximately 11% lower than the BNSR for a single channel situation. Moreover, these experiments

were repeated for different channel spacing conditions (100 GHz) and the results were the same.

46

5.4.2 Spatial resolution

A determinant parameter of any monitoring technique is its level of accuracy. The resolution of

a monitoring technique defines how close the individual data points that make up a reflectogram are

spaced in time (and corresponding distance) [11]. It is also relevant to determine the parameters that

directly influence the fault location precision. Thereby, the bit rate was varied between 0.1 and 90 Gbit/s

considering a sampling rate of 10 samples/bit. Also, the sampling rate was varied from 10 to 100

samples/bit considering a bit rate of 155.52 Mbit/s. In order to test the technique detection accuracy, a

reflective event was implemented at 45.6738 meters from the optical coupler. It was concluded that the

variation of the bit rate has no influence on the results accuracy, considering a constant number of

samples per bit. However, with the increasing number of samples per bit, the fault location precision

becomes more accurate. For an optical system working at 100 GS/s, with a bit rate of 10 Gbit/s and 10

samples per bit, one can achieve a precision of centimeters.

5.4.3 Measuring range

The maximum range for this monitoring technique is obtained with the following expression:

𝑅𝑎𝑛𝑔𝑒 =𝑐𝑁𝑏

4𝑛𝑅𝑏

,

( 5.1 )

where 𝑁𝑏 is the number of bits of the traffic signal frame and 𝑅𝑏 is the bit rate.

Since the normalized SDH frame time is 125 μs [4], for higher level SDH frames, the bit rate

must be higher. From equation (5.1) it was concluded that, for SDH signals, the maximum monitoring

range is 6465.5 meters. Thereby, with SDH traffic signals the intra-office propagation range is covered.

However, as mentioned before, due to its limitations, the SDH technology was replaced by the

modern OTN technology [9]. Since the beginning of the OTU-k (optical channel transport unit-k) frame

presents also alignment bytes [9], the proposed monitoring technique can easily be employed using the

current core traffic system.

The measuring range of this monitoring technique depends on the used traffic protocol. The

presented work was developed considering the implementation of the proposed monitoring system in

the core network nodes. However, since the majority of the traffic signals contain alignment bytes [4], it

can also be used in access solutions. In fact, due to the short fiber length between the central office and

the subscribers, this technique would be also useful in PON systems.

47

6. Experimental results

In this chapter, the proposed technique is tested in an experimental scenario. Furthermore, the

experimental outcome is compared with the simulation results.

Laboratory test scenario

In order to demonstrate the operation of the proposed monitoring technique, the monitoring

structure was assembled in a laboratory environment. The experiments were carried out at Coriant

facilities in Alfragide (Lisbon). Figure 6.1 shows a snapshot of the experimental scenario for the

monitoring structure operation with traffic signals.

Figure 6.1 – Experimental scenario for fiber fault detection using traffic signals.

In this setup, the Agilent N4901B Serial BERT generated a traffic signal which was converted

by the SFP 2.5 Gb/s transceiver module to an optical signal with a wavelength of 1550 nm. Several SMF

fibers, with 1 kilometer extension, were sequentially connected to the optical coupler and the reflective

events were achieved from the open end. By changing the type of connectors at the fiber end, back

reflections with different intensities were achieved [35]. For an open termination, due to differences in

the connector’s ferrule terminated angle, PC or APC (angle polished connector) lead to back reflections

with reflection coefficients of approximately 4% and 1%, respectively [35]. The used optical coupler and

mirror are described in section 4.3. The photodiode was a Nortel PP-10G which has a responsivity of

0.88 A/W. This photodiode was used to detect the monitoring structure input and output signals, which

were then acquired by the Tektronix DPO4104 oscilloscope. Finally, the acquired signals were loaded

to a computer, and the reflective event location algorithm was applied.

48

Results with random data

Initially, the experiment was carried out using PRBS data. The optical signal was transmitted

with a bit rate of 1 Gbit/s, and amplitude modulated with NRZ format. The transmitted optical power was

set to 6 dBm, followed by an attenuation of 10 dB at the laser output, which mitigated the effects at the

laser source of an eventual back reflection. Also, an additional attenuation of 6 dB was included at the

photodiode input to avoid its saturation. This scenario was tested for reflective events at 1 and 4

kilometers with different connectors. It was retrieved a 400 μs sample of monitoring structure input and

output signals and the process described in section 3.3 was applied. The sample length was chosen

considering the tradeoff between the oscilloscope sampling rate and the size of the output data file. The

technique was able to identify a reflective point at 1 kilometer for the case with a PC termination. Figure

6.2 presents the technique outcome for this case.

Figure 6.2 – Proposed monitoring technique experimental reflectogram using input PRBS data for a reflective

event at 1 kilometer from the optical coupler.

Although there is high background noise, it is possible to distinguish a peak at approximately

1000 meters. The reflectogram from Figure 6.2 presents a precision error of 2.5%. The fiber refractive

index was not characterized which imposed error in fault detection precision. The spatial axis of the final

reflectogram was defined by equation (3.22), and a slight variation in the real fiber refractive index

introduces a measurement error of several meters. Moreover, the tested fiber was rolled in a box, thus

its refractive index might have been altered [40].

The comparison with the simulation outcome from Figure 5.4 is inevitable. Besides the

difference in the location of the reflective points, the background noise is higher in experimental results

since the data load processed in laboratory was significantly larger than in that particular simulation.

49

Therefore, the BNSR for the experimental result is lower. In fact, when applying the smoothing filter

(described in section 5.2) to this case, the detected peak gets mixed with the background noise. Also,

there is a high peak at the center of the experimental reflectogram from Figure 6.2. In this case, the

monitoring technique fails to mitigate the auto-correlation peaks from the final reflectogram. The central

peak appears due to the lack of similarity between the auto-correlations of the structure input and output

signals. Nevertheless, this peak has no influence on the fault location. The optimization of the factor μ,

which is described in section 3.3, is an important matter not only for the clearance of undesirable peaks,

but also for the improvement of the BNSR.

Further, due to the high background noise values, one cannot detect reflective events in the

case of APC terminations. The reflective event at 4 kilometers terminated with a PC connector might

not be detected due to the power attenuation imposed by the fiber.

Results with traffic data

In order to generate a traffic signal, the Agilent N4901B Serial BERT was set to transmit a STM-

4 frame that was compiled taking into account the octet multiplexing specifications. The optical signal

was transmitted with a bit rate of 622.080 Mbit/s (STM-4 bit rate [4]) and was amplitude modulated with

NRZ format. The transmitted optical power was set to 2.3 dBm and suffered an attenuation of 6 dB at

the laser output. The scenario was tested for 1 and 2 kilometers terminated with different connectors.

Once again, for the cases with APC termination, one cannot detect fiber impairments as these are

expected to be immersed in the background noise. Nevertheless, when fibers are ended with PC

connectors, perfectly discernible peaks can be identified in the final trace. Figure 6.3 presents the

proposed technique reflectogram for reflective points at 1 kilometer and 2 kilometers.

Figure 6.3 – Proposed monitoring technique experimental reflectogram using input data traffic for a reflective point

at: a) 1 kilometer; b) 2 kilometers.

50

The reflectogram in Figure 6.3 a) presents a precision error of 2.5%, while the reflectogram

precision error in b) is 2.1%. The lack of accuracy can be explained by the same reasons described in

section 6.2.

Using traffic signals as probe, the BNSR of the experimentally detected peak has a similar

magnitude order as the one from simulation results. Considering the results in Figure 6.3, the BNSR at

1 kilometer and 2 kilometers is 0.026 dB and 0.0195 dB, respectively. As expected, the BNSR value

decreases with the increasing distance.

Although there is some uncertainty on the detected peaks as to the exact position of the fault,

for a first approach these results are satisfactory. In future experiments, all the aspects that impose

measurement error have to be taken in consideration to carefully determine the proposed technique

accuracy.

51

7. Implementation scenario

In this chapter, it is presented a possible field scenario for the application of the monitoring

structure to ROADM nodes. Moreover, it is explained an adaptation of the monitoring module in order

to detect real fiber impairments.

Location in ROADM architecture

The location of the monitoring structure in the ROADM nodes, considering a WDM system, is

presented in Figure 7.1.

Figure 7.1 – Implementation of the monitoring structure in a ROADM architecture.

In WDM systems, different wavelengths share the same propagation fibers [7]. In Figure 7.1

each colored line represents a single wavelength and the NMS (network management system) controls

the network configuration. The WSS (wavelength selective switch) multiplexes/demultiplexes tributary

52

wavelengths into/from a single ROADM port and the monitoring structure receives all the traffic signal

from that port [6]. In order to receive both the reference and the back reflected signals from a single

wavelength, an optical filter must be placed at the input of the photodiodes. If a tunable filter is chosen,

it will increase the proposed structure cost. The use of a passive filter will reduce costs, however when

the filtered wavelength is not in use, the monitoring process is halted. Alternatively, auxiliary channels

at specific service wavelengths that commonly are used in real systems for the transmission of

supervisory and service data, could be used for the implementation of the proposed monitoring structure.

Following the detection, the ADC (analog-to-digital converter) converts the detected analog signals to

digital and the monitoring module processes its inputs by means of a DSP (digital signal processor).

Monitoring module

In order to detect real fiber impairments, the monitoring algorithm that is explained in section

3.3 must suffer a slight adaptation. In the adapted process, the DSP selects a sample of the monitoring

structure input and output signals. In order to explain the sampling process, and based on the example

from section 3.3, Figure 7.2 is presented.

Figure 7.2 – DSP sampling process applied to the monitoring structure a) input signal; b) output signal.

For this explanation, it is assumed that the period 𝑊 corresponds to the traffic signal start of the

frame period, while period 𝑇 corresponds to the frame time. The DSP is set to capture half of the traffic

signal frame time, starting from the alignment bytes. It does not matter which part of the input and output

signals is extracted, nevertheless the capture must begin with these overhead bytes. Thus, like Figure

7.2 suggests, if there is a back reflection with a time delay that exceeds half of the traffic signal frame

time, it will not be visible in the final reflectogram.

Once the samples are extracted, the algorithm to obtain the final reflectogram is quite similar to

the computational model that is described in section 3.3. The auto-correlation from both samples is

53

computed and normalized to the maximum value. Then, the normalized auto-correlation values of the

output signal sample are multiplied by the factor μ. Finally, it is computed the ratio between the values

of both auto-correlations and the outcome passes through a smoothing filter in order to reduce the

background noise. This filter must be optimized in order to achieve the best performance possible. Due

to the previously explained sampling process, the plot window adjustment is not required.

In section 2.1, it is explained that the fiber under test will always return back reflections, mainly

due to Rayleigh backscattering. However, these back reflected signals have such low intensity that no

peak can be observed in this monitoring technique output. If there are no reflective points in the fiber

under test, the reflectogram presents only background noise.

Optimum device characteristics

In order to achieve the best performance from the proposed monitoring technique, the devices

must be carefully chosen. As Figure 5.5 presents, regarding the optical coupler, the lower the coupling

ratio, the more accentuated the peaks in the final reflectogram will be. From the practical ROADM

application point of view, this is highly desirable since the monitoring block contributes with a low

insertion loss penalty to the WDM signal path, therefore it does not affect significantly the penalty margin

of the system. Moreover, the photodiodes must have a bandwidth high enough to avoid the degradation

of the detected signal. The ADCs should have a high sampling rate for the technique to achieve the best

resolution. DSP blocks are commonly present in ROADM architectures [41], therefore the proposed

algorithm could be implemented considering these resources.

54

55

8. Conclusions and future work

In this work, a novel technique for optical fiber assessment based on reflectometry mixing was

proposed and studied. This method makes use of the optically transmitted traffic data to monitor the

intra-office fiber section, thus allowing a continuous assessment of the quality. Due to its simplicity, this

is a low-intrusive and low-cost solution that can upgrade the current ROADM architecture with

monitoring capability. Moreover, it was concluded that the applicability of this new method can be

extended to metro and access networks.

The developed monitoring structure was completely characterized as well as a signal processing

block that employs the correlation technique. It was demonstrated the possibility of monitoring optical

fiber employing the Raman pump back reflection. Considering the transmission of SDH signals, it was

concluded that this technique achieves a measuring range around 6500 meters.

The developed technique was tested in laboratory for impairments in the initial 2000 meters of

fiber. The experimental results demonstrated the feasibility of this technique and indicated a precision

error of less than 2.5%.

In addition to the theoretical proposal and experimental validation of the described monitoring

technique, a possible field scenario application to ROADM architectures was addressed.

Consequent to the developed work, several suggestions for future work are presented:

Study of optimization processes to improve the dynamic range of the proposed technique

reflectogram.

Study of the proposed technique operation for different modulation conditions.

Implementation of the required optical components in an integrated circuit board to be

included in ROADM architectures.

56

57

Bibliography

[1] Ramaswamin, R., K.N. Sivarajan, and G.H. Sasaki, Optical Networks - A Practical Perspective. 3rd ed. 2010: Elsevier.

[2] Corning and D. Keck. Optical fiber spans 30 years. 2000 October 15, 2014]; Available from:

http://www.corning.com/docs/opticalfiber/r3461.pdf. [3] Agrawal, G.P., Fiber-optic Communication Systems. 4th ed. 2010: Wiley. [4] Vladar, A., Understanding Telecommunications Networks. 1st ed. 2006: Institution of

Engineering and Technology. [5] Cisco. The Zettabyte Era: Trends and Analysis. 2014 October 10, 2014]; Available from:

http://www.cisco.com/c/en/us/solutions/collateral/service-provider/visual-networking-index-vni/VNI_Hyperconnectivity_WP.pdf.

[6] Gringeri, S., et al., Flexible architectures for optical transport nodes and networks.

Communications Magazine, IEEE, 2010. 48(7): p. 40-50. [7] Cisco. Introduction to DWDM technology. 2001 [cited March 2, 2015; Available from:

http://www.cisco.com/application/pdf/en/us/guest/products/ps2011/c2001/ccmigration_09186a00802342cf.pdf.

[8] Pires, J., Sistemas e Redes de Telecomunicações. 1st ed. 2006: IST. [9] ITU and T. Walker. Optical Transport Network (OTN) Tutorial. 2014 March 9, 2015]; Available

from: https://www.itu.int/ITU-T/studygroups/com15/otn/OTNtutorial.pdf. [10] Hui, R. and M. O'Sullivan, Fiber Optic Measurement Techniques. 1st ed. 2009: Elsevier. [11] Anritsu. Understanding OTDRs. 2011 October 4, 2014]; Available from:

http://www.anritsu.com/en-GB/Media-Room/Newsletters/files/Anritsu-understanding-otdrs.pdf. [12] OSP Magazine, T. Irujo, and T. West. Coming to Grips With Loss. 2008 September 23, 2014];

Available from: http://www.ospmag.com/issue/article/052008-grip. [13] Yuksel, K., et al. Optical frequency domain reflectometry: A review. in Transparent Optical

Networks, 2009. ICTON '09. 11th International Conference on. 2009. [14] Rishvanth, K.P., et al. Design and simulation of Optical Frequency Domain reflectometer for

short distance fault detection in optical fibers and integrated optical devices using ptolemy-II. in Ultra Modern Telecommunications & Workshops, 2009. ICUMT '09. International Conference on. 2009.

[15] Obaton, A.F., et al., Characterization of Fiber Bragg Gratings by Phase-Sensitive Optical Low-

Coherence Reflectometry. Instrumentation and Measurement, IEEE Transactions on, 2006. 55(5): p. 1696-1703.

[16] Takada, K., M. Yamada, and S. Mitachi, Optical low coherence reflectometer with a 47-dB

dynamic range achieved by using a fluoride-based erbium-doped fiber amplifier. Lightwave Technology, Journal of, 1998. 16(4): p. 593-597.

[17] Wegmuller, M., et al., Overview of coherent reflectometry techniques: characterization of

components and small systems. Technical Digest Symposium on Optical Fiber Measurements, 2000.

58

[18] Gifford, D.K., et al. Millimeter Resolution Otpical Reflectometry Over Up to Two Kilometers of Fiber Length. in Avionics, Fiber-Optics and Photonics Technology Conference, 2007 IEEE. 2007.

[19] Champavere, A. New OTDR measurement and monitoring techniques. in Optical Fiber

Communications Conference and Exhibition (OFC), 2014. 2014. [20] Baker, C., et al., Incoherent optical frequency domain reflectometry based on a Kerr phase-

interrogator. Optics Express, 2014. 22(13): p. 15370-15375. [21] Ryu, G. and M. Dagenais. Incoherent optical frequency domain reflectometry for health

monitoring of avionics fiber optics networks. in Avionics, Fiber-Optics and Photonics Technology Conference, 2008 IEEE. 2008.

[22] Zhenyang, D., et al., Long Measurement Range OFDR Beyond Laser Coherence Length.

Photonics Technology Letters, IEEE, 2013. 25(2): p. 202-205. [23] Ndiaye, C., T. Hara, and H. Ito. Profilometry using a frequency-shifted feedback laser. in Lasers

and Electro-Optics, 2005. (CLEO). Conference on. 2005. [24] Li, S., et al., Rangeability extension of fiber-optic low-coherence measurement based on

cascaded multistage fiber delay line. Applied Optics, 2012. 51(6): p. 771-775. [25] Clivaz, X., F.D. Marquis-Weible, and R.-P. Salathe. 1.5-μm resolution optical low-coherence

reflectometry in biological tissues. 1994. [26] Zhao, T., et al., Fiber fault location utilizing traffic signal in optical network. Optics Express,

2013. 21(20): p. 23978-23984. [27] Schmuck, H., et al. Embedded OTDR techniques for cost-efficient fibre monitoring in optical

access networks. in Optical Communications, 2006. ECOC 2006. European Conference on. 2006.

[28] Lim, K.W., et al., Fault localization in WDM passive optical network by reusing downstream light

sources. Photonics Technology Letters, IEEE, 2005. 17(12): p. 2691-2693. [29] Hecht, E., Optics. 2002: Addison-Wesley. [30] Collet, E., Field Guide to Polarization. 1st ed. 2005: SPIE press. [31] Diniz, P., E. Silva, and S. Netto, Digital Signal Processing - System Analysis and Design. 2nd

ed. 2010: Cambridge University Press. [32] Haykin, S., Communication Systems. 4th ed. 2001: John Wiley & Sons, Inc. [33] Serena, P., M. Bertolini, and A. Vannuci, Optilux Toolbox. 2009. [34] Hammersley, J.M. and D.C. Handscomb, Monte Carlo Methods. 1st ed. 1964: Metheun & Co

Ltd. [35] ADC. Choosing the right connector. 2008 March 20th, 2015]; Available from:

http://www.te.com/content/dam/te/global/english/industries/telecom-networks/knowledge-center/documents/telecom-white-paper-choosing-the-right-connector-105662ae.pdf.

[36] Takushima, Y. and Y.C. Chung, Optical reflectometry based on correlation detection and its

application to the in-service monitoring of WDM passive optical network. Optics Express, 2007. 15(9): p. 5318-5326.

[37] Yuncai, W., W. Bingjie, and W. Anbang, Chaotic Correlation Optical Time Domain Reflectometer

Utilizing Laser Diode. Photonics Technology Letters, IEEE, 2008. 20(19): p. 1636-1638.

59

[38] Dawson, J. and T.C. Tozer, New technique for nonintrusive OTDR based on traffic data correlation. Electronics Letters, 1994. 30(17): p. 1443-1444.

[39] Zoboli, M. and P. Bassi, High spatial resolution OTDR attenuation measurements by a

correlation technique. Applied Optics, 1983. 22(23): p. 3680-3681. [40] Liu, G., et al., Bent optical fiber taper for refractive index detection with a high sensitivity.

Sensors and Actuators A: Physical, 2013. 201(0): p. 352-356. [41] Tang, J., et al. New ROADMs with DSP-Enabled Dynamic and Flexible Operations for Elastic

Optical Networks. in Optical Fiber Communication Conference. 2015. Los Angeles, California: Optical Society of America.

60