Dissertation presented to the Instituto Tecnológico de Aeronáutica, in

partial fulfillment of the requirements for the degree of Master of Science in

the Graduate Program of Science and Space Technologies, Field of Space
Sensors and Actuators.

Gabriel Fernandes de Souza Nunes

FIBER OPTIC ACCELEROMETER BASED ON AN

INTERFEROMETER WITH NONLINEAR CONTROL
FOR SPACE APPLICATIONS

Dissertation approved in its final version by signatories below:

Prof. Dr. João Marcos Salvi Sakamoto

Advisor

Prof. Dr. Pedro Teixeira Lacava

Pro-Rector of Graduate Courses

Campo Montenegro
São José dos Campos, SP - Brazil
2021
Cataloging-in Publication Data
Documentation and Information Division
Nunes, Gabriel Fernandes de Souza
Fiber optic accelerometer based on an interferometer with nonlinear control for space
applications / Gabriel Fernandes de Souza Nunes.
São José dos Campos, 2021.
135p.

Dissertation of Master of Science – Course of Science and Space Technologies. Area of Space
Sensors and Actuators – Instituto Tecnológico de Aeronáutica, 2021. Advisor: Prof. Dr. João
Marcos Salvi Sakamoto.

1. Acelerômetros. 2. Fibras ópticas. 3. Interferômetros. 4. Lasers. 5. Sensores. 6. Fı́sica.

I. Instituto Tecnológico de Aeronáutica. II. Fiber optic accelerometer based on an interferometer
with nonlinear control for space applications.

BIBLIOGRAPHIC REFERENCE
NUNES, Gabriel Fernandes de Souza. Fiber optic accelerometer based on an
interferometer with nonlinear control for space applications. 2021. 135p. Dissertation
of Master of Science – Instituto Tecnológico de Aeronáutica, São José dos Campos.

CESSION OF RIGHTS
AUTHOR’S NAME: Gabriel Fernandes de Souza Nunes
PUBLICATION TITLE: Fiber optic accelerometer based on an interferometer with
nonlinear control for space applications.
PUBLICATION KIND/YEAR: Dissertation / 2021

It is granted to Instituto Tecnológico de Aeronáutica permission to reproduce copies of

this dissertation and to only loan or to sell copies for academic and scientific purposes.
The author reserves other publication rights and no part of this dissertation can be
reproduced without the authorization of the author.

Gabriel Fernandes de Souza Nunes

Rua Teresina, 54, ap. 14
12.235-610 – São José dos Campos–SP
 FIBER OPTIC ACCELEROMETER BASED ON AN
INTERFEROMETER WITH NONLINEAR CONTROL
FOR SPACE APPLICATIONS

Gabriel Fernandes de Souza Nunes

Thesis Committee Composition:

Prof. Dr. Milton Sergio Fernandes de Lima Chairperson - ITA

Prof. Dr. João Marcos Salvi Sakamoto Advisor - ITA
Prof. Dr. Getúlio de Vasconcelos Member - ITA
Prof. Dr. Rodrigo Moreira Bacurau External member - UNICAMP

ITA
To my parents and to Gabriella
 Acknowledgments

The accomplishment of this dissertation wouldn’t be possible without the great support
and inspiration of the people around me. I am grateful to Dr. João Marcos Salvi Sakamoto
for the lessons, advisement, trust, and friendship. Also, for the encouragement when most
needed. To São Paulo Research Foundation (FAPESP) for grant #2019/04679-4.
To Dr. Kelly, for luring me into this wonderful world of discoveries, science, experi-
ments, and optics. To Professors Dr. Nicolau Rodrigues and Dr. Rudimar Riva, for the
inspirational classes. To Professor Dr. Cláudio Kitano, for his assistance in the paper
writing process.
To Dr. Roberta Martin, for showing me the best way to approach optical interferome-
try and lending me the tools. To my laboratory and work friends, Thiago Pentiado, Walter
Ferreira, Vinicius Palma, Dr. Luiz Felão, and Renato Giovanini, for the support, technical
conversations, and the much-needed quality procrastination between experiments. To the
IEAv staff and researchers, Erik, Gustavo, Keila, Lavras, Lira, Minelli, Sérgio, and Tales,
for the technical assistance.
To my mom, Edna, who always supported me. Who taught me since early days that
the only thing no one can take away from me is my knowledge. To my dad, Gustavo, and
my eight siblings, especially my closest sisters Natalia and Bia, for helping and influencing
me to be who I become. To all my nephews, nieces, aunts, uncles, and cousins, for being
part of this awesome family environment.
To my friends, Kiki, João, Abner, Carlos, and Vinicius, who helped this disserta-
tion indirectly by keeping me sane during all work obligations and deadlines while in a
pandemic.
To my dear friend Hermes, for the friendship, advisement, and all the music exploration
along with the chess games. Thanks for all the affection and trust placed in me.
Finally, thank you very much, Gabriella, for being this sweet person and for the
companionship during all these years. Your distributed daily dose of encouragement was
proven very much essential.
 “Having knowledge, to apply it;
not having knowledge, to confess your ignorance;
this is (real) knowledge.”
— Confucius
 Resumo

Acelerômetros a fibra óptica são sensores projetados para medir vibrações de veı́culos
ou a aceleração para fins de navegação, e que vem sendo estudados e empregados em
operações espaciais e aeronáuticas, devido à sua leveza, alta sensibilidade, imunidade a
interferências eletromagnéticas, capacidade de funcionar em altas temperaturas, resistên-
cia a vibrações extremas e nı́veis elevados de choque, além de apresentar capacidade de
multiplexação. Os acelerômetros de alto desempenho ainda são embargados pelos paı́ses
que detém tal tecnologia, o que motiva o desenvolvimento de acelerômetros de alto desem-
penho com tecnologia nacional. Além disso, os acelerômetros disponı́veis no mercado são
elétricos, exigindo blindagem em aplicações espaciais, podem necessitar de técnicas de mi-
cro usinagem e não fornecem alta precisão. Neste trabalho, é proposto um interferômetro
em fibra óptica com controle não linear para interrogar um acelerômetro opto-mecânico.
O controle não linear serve para estabilizar o ponto de operação do interferômetro em seu
ponto de maior sensibilidade (ponto de quadratura de fase) e suprimir o desvanecimento
do sinal. Os interferômetros de Michelson e de Mach-Zehnder foram estudados e mon-
tados com componentes de fibra óptica. O controle foi implementado digitalmente para
estabilizar os interferômetros tanto no ponto de quadratura como em um ponto genérico
de operação. Moduladores de fase óptica foram caracterizados para validar o correto fun-
cionamento do interferômetro controlado e a técnica de demodulação de sinais empregada.
O modelo mecânico de um acelerômetro massa-mola-amortecedor foi estudado e montado.
Analisou-se configurações de localização do elemento sensor, sendo no braço sensor do in-
terferômetro ou em uma cavidade de laser de fibra. Isso levou ao estudo e caracterização
de lasers de fibra óptica dopada com érbio. Assim, amplificadores a fibra óptica dopada
com érbio foram caracterizados e um laser multimodo foi montado e caracterizado para
atuar como um elemento sensor. Um acelerômetro opto-mecânico foi montado e posicio-
nado no braço sensor do interferômetro de Mach-Zehnder. Em seguida, realizou-se uma
caracterização manual e os seguintes valores foram estimados: frequência de ressonância
de 503,3 Hz, fator de amortecimento de 60,5 s−1 , sensibilidade de 4,432 rad/g em 100 Hz
e faixa dinâmica de 0,01 g a 0,187 g (25 dB). O laser de fibra resultante forneceu um sinal
interferométrico instável, não permitindo a utilização da cavidade do laser como elemento
sensor por enquanto. Por outro lado, apresentou potencial para trabalhos futuros, já que
sua sensibilidade estimada é 37 vezes maior que o elemento sensor no braço do interferô-
metro. Em conclusão, o acelerômetro opto-mecânico apresentou um valor de sensibilidade
estimada que concorda com a literatura e frequência de ressonância acima da região de
interesse para aplicações espaciais, demonstrando funcionamento adequado utilizando o
interferômetro com o controle não linear.
 Abstract

Fiber optic accelerometers are sensors designed to measure vehicle vibrations or accelera-
tion for navigation purposes, and which have been studied and used in space and aeronau-
tical operations, due to their lightweight, high sensitivity, immunity to electromagnetic
interference, ability to function in high temperatures, resistance to extreme vibrations
and high levels of shock, and multiplexing capability. High-performance accelerometers
are still embargoed by countries that possess such technology, which motivates the de-
velopment of high-performance accelerometers with national technology. Furthermore,
accelerometers available on the market are electrical, requiring shielding in space appli-
cations, may require micro-machining techniques, and do not provide high accuracy. In
this work, an optical fiber interferometer with nonlinear control is proposed to interrogate
an opto-mechanical accelerometer. The nonlinear control is used to stabilize the interfer-
ometer operating point at its most sensitive point (phase quadrature point) and suppress
signal fading. Michelson and Mach-Zehnder interferometers were studied and assembled
with fiber optic components. The control was digitally implemented to stabilize the inter-
ferometers both at the quadrature point and at a generic operating point. Optical phase
modulators were characterized to validate the correct operation of the controlled inter-
ferometer and the demodulation technique used to recover the signal. The mechanical
model of a mass-spring-damper accelerometer was studied and assembled. Sensor ele-
ment location configurations were analyzed, either in the interferometer sensor arm or in
a fiber laser cavity. This led to the study and characterization of erbium-doped fiber op-
tic lasers. Thus, erbium-doped fiber optic amplifiers were characterized and a multimode
laser was assembled and characterized to act as a sensing element. An opto-mechanical
accelerometer was mounted and positioned on the sensing arm of the Mach-Zehnder inter-
ferometer. Then, a manual characterization was performed and the following values were
estimated: resonant frequency of 503.3 Hz, the damping factor of 60.5 s−1 , the sensitivity
of 4.432 rad/g at 100 Hz and dynamic range from 0.01 g to 0.187 g (25 dB). The resulting
fiber laser provided an unstable interferometric signal, not allowing the laser cavity to be
used as a sensing element for the time being. On the other hand, it presented a potential
for future work, since its estimated sensitivity is 37 times greater than the sensor element
in the interferometer arm. In conclusion, the opto-mechanical accelerometer presented
an estimated sensitivity value that agrees with the literature and a resonance frequency
above the region of interest for spatial applications, demonstrating proper functioning
using the interferometer with nonlinear control.
 List of Figures

FIGURE 1.1 – Different optical fiber accelerometer sensing element configurations. 25

FIGURE 2.1 – Interference fringes. . . . . . . . . . . . . . . . . . . . . . . . . . . . 30

FIGURE 2.2 – Interferometric characteristic curve. . . . . . . . . . . . . . . . . . . 34
FIGURE 2.3 – Examples of interferometric responses. . . . . . . . . . . . . . . . . . 35
FIGURE 2.4 – Fundamental assemble of a fiber optic Michelson interferometer. . . 36
FIGURE 2.5 – Fundamental assemble of a fiber optic Mach-Zehnder interferometer. 37
FIGURE 2.6 – Polarization ellipse. . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
FIGURE 2.7 – Poincaré sphere. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
FIGURE 2.8 – Recombining electric fields state of polarization with and without
Faraday rotator mirror represented with polarization ellipses. . . . . 45
FIGURE 2.9 – Simulation of polarization states for Michelson interferometer with
and without Faraday rotator mirror represented on the Poincaré
sphere. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
FIGURE 2.10 –Block diagram of the controlled interferometer. . . . . . . . . . . . . 48
FIGURE 2.11 –Equilibrium points stability analysis in the curve A·V · cos(x1 ). . . . 50
FIGURE 2.12 –Phase plane of the system when the control condition is met (γ > |φ̇0 |). 52
FIGURE 2.13 –Phase plane of the system when the control condition is not met
(γ ≤ |φ̇0 |) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
FIGURE 2.14 –Phase plane of the system using sigmoid instead of signal function. . 53
FIGURE 2.15 –Control loop block diagram on Simulink. . . . . . . . . . . . . . . . 54
FIGURE 2.16 –Control actuating in simulation. . . . . . . . . . . . . . . . . . . . . 55
FIGURE 2.17 –Block diagram of the controlled interferometer with offset A removal
and random set-point sp. . . . . . . . . . . . . . . . . . . . . . . . . 56
FIGURE 2.18 –Phase plane of the system with setpoint equal to 0.8·A·V . . . . . . . 57
FIGURE 2.19 –Phase plane of the system with setpoint equal to −0.8·A·V . . . . . . 57
FIGURE 2.20 –Noises current simulation for a laser with supposed RIN -164 dB. . . 60
FIGURE 2.21 –Simulation of 1/f noise together with white noise. . . . . . . . . . . 61

FIGURE 3.1 – Mass body under influence of a force. . . . . . . . . . . . . . . . . . 64

FIGURE 3.2 – Free-body diagram of some possible setups for sensing elements. . . 65
FIGURE 3.3 – Scheme of the sensing element used in this work. . . . . . . . . . . . 66
FIGURE 3.4 – Mass-spring equivalent system for optical fiber accelerometer. . . . . 66
FIGURE 3.5 – Unforced mass-spring-damper response simulation. . . . . . . . . . . 68
FIGURE 3.6 – Simulation of the accelerometer using the transfer function. . . . . . 69
FIGURE 3.7 – Accelerometer frequency response simulation for a sinusoidal input. 70
FIGURE 3.8 – Michelson interferometer with nonlinear control. . . . . . . . . . . . 73
FIGURE 3.9 – Phase modulator based on temperature (FOM-MC). . . . . . . . . . 73
FIGURE 3.10 –Complete Mach-Zehnder interferometer setup. . . . . . . . . . . . . 75
FIGURE 3.11 –Erbium energy three-level diagram. . . . . . . . . . . . . . . . . . . 77
FIGURE 3.12 –Pump laser (laser diode) slope efficiency. . . . . . . . . . . . . . . . 81
FIGURE 3.13 –Setup for characterization of the erbium doped fiber amplifier. . . . 81
FIGURE 3.14 –FBG on-line characterization setup used at IEAv. . . . . . . . . . . 82
FIGURE 3.15 –Calibration setup of FBG on a single-mode photosensitive fiber. . . 83
FIGURE 3.16 –FBG detection setup by transmission. . . . . . . . . . . . . . . . . . 83
FIGURE 3.17 –Reference fluorescence spectrum by transmission. . . . . . . . . . . . 84
FIGURE 3.18 –FBG reflection spectrum and erbium fluorescence spectrum. . . . . 84
FIGURE 3.19 –Number of longitudinal modes varying with distance between FBGs. 85

FIGURE 4.1 – Nonlinear control actuating in the optical interferomter. . . . . . . . 87

FIGURE 4.2 – Different setpoint values in the control loop. . . . . . . . . . . . . . 88
FIGURE 4.3 – Controlled interferometer (γv = 1) without an input signal. . . . . . 89
FIGURE 4.4 – Chattering example with controlled interferometer (γv = 7) without
an input signal. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
FIGURE 4.5 – Chattering example with controlled interferometer (γv = 30) with-
out an input signal. . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
FIGURE 4.6 – Interferometer (γv = 1) with an sinusoidal input signal. . . . . . . . 91
FIGURE 4.7 – Chattering example with controlled interferometer (γv = 10) with
an sinusoidal input signal. . . . . . . . . . . . . . . . . . . . . . . . 92
FIGURE 4.8 – Picture of FOM1. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
FIGURE 4.9 – Frequency response of FOM1. . . . . . . . . . . . . . . . . . . . . . 93
FIGURE 4.10 –Linearity from FOM1 at 1 KHz. . . . . . . . . . . . . . . . . . . . . 94
FIGURE 4.11 –Picture of a 10 mm FOM-MC soldered to a circuit board with ther-
mal paste over it. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
FIGURE 4.12 –Frequency response of FOM-MC. . . . . . . . . . . . . . . . . . . . 95
FIGURE 4.13 –Linearity of FOM-MC at 30 Hz and 150 Hz. . . . . . . . . . . . . . 95
FIGURE 4.14 –Problems with fiber Michelson interferometer. . . . . . . . . . . . . 96
FIGURE 4.15 –Visibility of the Mach-Zehnder interferometer over time. . . . . . . . 97
FIGURE 4.16 –Characterization of the erbium doped optical fibers. . . . . . . . . . 99
FIGURE 4.17 –Comparison between optical amplifier model and experimental data. 100
FIGURE 4.18 –FBGs inscribed on standard photosensitive fiber. . . . . . . . . . . . 101
FIGURE 4.19 –Lasers optical power in function of pump current. . . . . . . . . . . 102
FIGURE 4.20 –Laser cavity pumped at a constant power. . . . . . . . . . . . . . . 103
FIGURE 4.21 –Straining FBG in a fiber laser. . . . . . . . . . . . . . . . . . . . . . 104
FIGURE 4.22 –Force versus strain curve to calculate the Young’s modulus of pho-
tosensitive optical fiber PS1250/1500. . . . . . . . . . . . . . . . . . 105
FIGURE 4.23 –Straining erbium-doped fiber in a fiber laser. . . . . . . . . . . . . . 105
FIGURE 4.24 –Laser spectrum of fiber laser made with 2 m of fiber ER30 when
changing pump current from 30 mA to 650 mA. . . . . . . . . . . . 106
FIGURE 4.25 –Fiber laser with 2 m of ER30 direct modulation. . . . . . . . . . . . 106
FIGURE 4.26 –Interferometer working with the fiber laser and an input signal of
5 Hz to reach maximum and minimum values. . . . . . . . . . . . . 107
FIGURE 4.27 –Accelerometer model. . . . . . . . . . . . . . . . . . . . . . . . . . . 108
FIGURE 4.28 –Picture of the accelerometer assembly. . . . . . . . . . . . . . . . . . 109
FIGURE 4.29 –Accelerometer (0.7 N, bolted to the optical breadboard) response to
impulse from metal stick at the right fixed support with exponential
fit. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109
FIGURE 4.30 –Accelerometer (2 N, bolted to the optical breadboard) response to
impulse from metal stick at the right fixed support with exponential
fit. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110
FIGURE 4.31 –Accelerometer (1 N, attached) response to impulse from cotton swab
at the right fixed support with exponential fit. . . . . . . . . . . . . 111
FIGURE 4.32 –Accelerometer (1 N, attached) response to impulse from cotton swab
at the left fixed support with exponential fit. . . . . . . . . . . . . . 111
FIGURE 4.33 –Accelerometer (1 N, attached) response to impulse from metal stick
at the breadboard on the vertical direction with exponential fit. . . 111
FIGURE 4.34 –Accelerometer (1 N, not attached) response to impulse from cotton
swab at the right fixed support with exponential fit. . . . . . . . . . 112
FIGURE 4.35 –Accelerometer (1 N, not attached) response to impulse from cotton
swab at the left fixed support with exponential fit. . . . . . . . . . . 112
FIGURE 4.36 –Accelerometer (1 N, not attached) response to impulse from metal
stick at the breadboard with exponential fit. . . . . . . . . . . . . . 112
FIGURE 4.37 –Accelerometer (1 N, not bolted to the optical breadboard) response
to impulse from a manual impulse at the breadboard on the vertical
direction with exponential fit. . . . . . . . . . . . . . . . . . . . . . 114
FIGURE 4.38 –Beginning of the accelerometer (1 N, not bolted to the optical bread-
board) response to impulse from a manual impulse at the breadboard
on the vertical direction with exponential fit. . . . . . . . . . . . . . 114
FIGURE 4.39 –Accelerometer (2 N, bolted to the optical breadboard) response to
impulse from a manual impulse at the breadboard on the vertical
direction with exponential fit. . . . . . . . . . . . . . . . . . . . . . 115
FIGURE 4.40 –Accelerometer (1 N, bolted to the optical breadboard) response to
impulse from a manual impulse at the breadboard on the vertical
direction with exponential fit. . . . . . . . . . . . . . . . . . . . . . 115
FIGURE 4.41 –Accelerometer (0.7 N, bolted to the optical breadboard) response to
an acoustic signal of 300 Hz. . . . . . . . . . . . . . . . . . . . . . . 116
FIGURE 4.42 –Accelerometer (2 N, bolted to the optical breadboard) response to
a handclap near the inertial mass. . . . . . . . . . . . . . . . . . . . 116
FIGURE 4.43 –Accelerometer (2 N, bolted to the optical breadboard) response to a
smartphone vibrating while standing at the base of the accelerometer.116
FIGURE 4.44 –Sensitivity estimation over frequency. . . . . . . . . . . . . . . . . . 117
 List of Tables

TABLE 2.1 – Jones matrices of optical elements necessary to simulate the Michel-
son interferometer. . . . . . . . . . . . . . . . . . . . . . . . . . . . 44

TABLE 3.1 – Constants used to simulate accelerometer. . . . . . . . . . . . . . . . 69

TABLE 4.1 – Linear fitting results from the erbium doped fibers characterization. 98
List of Abbreviations and Acronyms

FBG Fiber Bragg grating

FC Fiber directional coupler
FC/PC Optical fiber connector by physical contact
FOM Fiber optical modulator
FOM-MC Optical phase modulator based on temperature
FPGA Field programmable gate array
FRM Faraday rotator mirror
LPF Low-pass filter
IEAv Institute for Advanced Studies
MEMS Micro-electro-mechanical systems
OPD Optical path difference
OSA Optical spectrum analyzer
PD Photodetector
PID Proportional integral derivative
PZT Piezoelectric actuator
rin Relative intensity noise
SLD Superluminescent diode
SNR Signal-to-noise ratio
SOP State of polarization
THD Total harmonic distortion
VI Virtual instrument
VS/SM Variable structure and sliding modes
WDM Wavelength-division multiplexing
 List of Symbols

k Wavenumber
l Physical path length traveled by light
n Refractive index
φ Optical phase shift
φ0 Quasi-static optical phase shift
δ Phase shift
∆φ Optical phase shift signal of interest
λ Wavelength
E Electric field
I Optical Irradiance or Intensity of a light source
A Photodetector responsivity and gains
V Interferometer visibility
γ Total feedback gain from nonlinear control loop
γv Variable feedback gain inside digital platform
C Fiber laser conversion factor between wavelength and displacement
τ Time constant
τc Coherence time
Lc Coherence length
η Half angle between two states of polarization in the Poincaré sphere
χ Ellipticity
ψ Orientation angle of an ellipse
a Acceleration
g Standard gravity
x Displacement
m Mass
F Force
A0 Acceleration amplitude
Ac Optical fiber cladding cross-sectional area
K Spring constant
Y Young’s modulus
σ Mechanical stress
ε Mechanical strain
ω Angular frequency
ω0 Natural angular frequency
f0 Natural frequency
Γ Damping factor
b Damping coefficient
c The speed of light in vacuum
ν Frequency of light source
Er3+ Ions of the element Erbium
σe , σa Cross sections for the emission and absorption of optical signal transition
h The Planck constant
α Optical gain
α0 Low signal optical gain
IS Signal saturation intensity
N Population density of an energy level
P Power
W Energy level of laser transitions
 Contents

List of Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xi

List of Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xvi

List of Abbreviations and Acronyms . . . . . . . . . . . . . . . . . xvii

List of Symbols . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xviii

1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
1.1 Sensing element configuration . . . . . . . . . . . . . . . . . . . . . . . . 24
1.2 Objective . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
1.3 Outline . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28

2 Optical interferometry with nonlinear control . . . . 29

2.1 Optical interferometry theory . . . . . . . . . . . . . . . . . . . . . . . . 29
2.1.1 Michelson interferometer . . . . . . . . . . . . . . . . . . . . . . . . . 29
2.1.2 Mach-Zehnder interferometer . . . . . . . . . . . . . . . . . . . . . . . 37
2.1.3 Optical phase modulation . . . . . . . . . . . . . . . . . . . . . . . . 38
2.1.4 Polarization-induced fading . . . . . . . . . . . . . . . . . . . . . . . 41
2.1.5 Interferometer signal demodulation for low phase demodulation index 47
2.2 Nonlinear control for two beam interferometry . . . . . . . . . . . . . . . 48
2.2.1 State space representation . . . . . . . . . . . . . . . . . . . . . . . . 50
2.2.2 Phase plane . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
2.2.3 Simulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
2.2.4 Control setpoint out of zero . . . . . . . . . . . . . . . . . . . . . . . 55
2.3 Optical detection and noise analysis . . . . . . . . . . . . . . . . . . . . . 57
 2.3.1 Thermal noise . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
2.3.2 Shot noise . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
2.3.3 Relative intensity noise . . . . . . . . . . . . . . . . . . . . . . . . . . 58
2.3.4 1/f noise . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
2.3.5 Quantization error . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
2.3.6 Minimum detectable phase calculation . . . . . . . . . . . . . . . . . 62

3 Fiber optical accelerometer . . . . . . . . . . . . . . . . . . . . 64

3.1 Opto-mechanical accelerometer operation principle . . . . . . . . . . . . 64
3.1.1 Accelerometer interrogation . . . . . . . . . . . . . . . . . . . . . . . 71
3.2 Accelerometer based on an interferometer with nonlinear control exper-
imental setups . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
3.2.1 Michelson experimental setup . . . . . . . . . . . . . . . . . . . . . . 72
3.2.2 Mach-Zehnder experimental setup . . . . . . . . . . . . . . . . . . . . 74
3.2.3 Control Implementation . . . . . . . . . . . . . . . . . . . . . . . . . 75
3.3 Erbium-doped fiber laser . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
3.3.1 Optical fiber light amplifier . . . . . . . . . . . . . . . . . . . . . . . . 77
3.3.2 Erbium-doped fiber characterization . . . . . . . . . . . . . . . . . . . 80
3.3.3 Fiber Bragg grating inscription . . . . . . . . . . . . . . . . . . . . . 82
3.3.4 Longitudinal mode selection . . . . . . . . . . . . . . . . . . . . . . . 84

4 Results and discussions . . . . . . . . . . . . . . . . . . . . . . . . . 87

4.1 Nonlinear control . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
4.2 Michelson interferometer results . . . . . . . . . . . . . . . . . . . . . . . 93
4.2.1 Piezoeletric phase modulator characterization . . . . . . . . . . . . . 93
4.2.2 Metal on fiber phase modulator characterization . . . . . . . . . . . 94
4.2.3 Problems with the Michelson setup . . . . . . . . . . . . . . . . . . . 96
4.3 Mach-Zehnder interferometer results . . . . . . . . . . . . . . . . . . . . 96
4.4 Erbium-doped fiber laser results . . . . . . . . . . . . . . . . . . . . . . . 97
4.4.1 Optical fiber amplifier . . . . . . . . . . . . . . . . . . . . . . . . . . 97
4.4.2 Fiber Bragg gratings . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
 4.4.3 Erbium-doped fiber laser characterization . . . . . . . . . . . . . . . . 102
4.4.4 Erbium-doped fiber laser as a sensing element . . . . . . . . . . . . . 103
4.5 Opto-mechanical accelerometer . . . . . . . . . . . . . . . . . . . . . . . . 108
4.5.1 Sensitivity estimation . . . . . . . . . . . . . . . . . . . . . . . . . . . 117
4.5.2 Minimum detectable acceleration and dynamic range . . . . . . . . . 118

5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119
5.1 Future work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120

References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122

Appendix A – Digital control implementation . . . . . . . . 132

A.1 Real Time Processor VI . . . . . . . . . . . . . . . . . . . . . . . . . . . . 132
A.2 Calibration FPGA VI . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133
A.3 Control Loop FPGA VI . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134
A.4 Two Input Control Implementation FPGA . . . . . . . . . . . . . . . . . 135
1 Introduction

The accelerometer is a type of sensor able to measure the variation rate of the velocity
of a body or, in other words, its acceleration (DOEBELIN, 2007). An accelerometer can be
designed to measure the acceleration of a body for navigation purposes, working in this
case as an inertial sensor. On the other hand, an accelerometer can be used to measure
vibration for structural health monitoring or for predictive maintenance (LINESSIO, 2020)
or for measuring vibration during a vehicle launching (SHEA, 2006). Regarding space and
aeronautics applications, accelerometers are indispensable whether for measuring vibra-
tions or for navigation (LEFÈVRE, 2014), while the recent increase in space cargo demand
(REDDY, 2018) reassures this sensor importance. Spacecrafts user’s guide specifies the
acceleration envelope the payload should support, in terms of navigation acceleration and
sinusoidal vibrations (SPACEX, 2020), showing the relevance of this type of sensor on the
space sector.
There are accelerometers available commercially but with performance below the nec-
essary for proper space of aeronautics operation as, for example, navigation of an aircraft
or a spacecraft. Sensors for high-grade applications are still embargoed by countries that
possess these technologies. Therefore, the motivation for the accomplishment of this work
is to dominate and to develop a proprietary technology of high-performance accelerometers
which meet space application requirements.
Usually, accelerometers available in the market are of the electrical type, e.g., piezo-
electric, piezoresistive, and capacitive devices. These devices are susceptible to electro-
magnetic interference and often require shielding, which might increase the cost, size, and
weight of the final sensor (UDD; SPILLMAN, 2011). Also, the seemly preferred electrical ac-
celerometer is the micro-electro-mechanical systems (MEMS) which needs micromachining
techniques (JINDAL et al., 2020), and do not provide high precision and accuracy.
Optical fiber1 sensors, however, can be small-sized, lightweight, and immune to elec-
tromagnetic interference. They can also offer robustness factors, such as high-temperature
operation, and resistance to extreme vibration and shock levels. Two-wave optical fiber
interferometers (such as Michelson or Mach-Zehnder) are optical fiber sensors that present

1
The word fiber in this text refers to optical fiber made of silica unless explicitly told otherwise.
CHAPTER 1. INTRODUCTION 24

intrinsically high sensitivity, accuracy, and precision (UDD; SPILLMAN, 2011). Therefore,
this type of optical fiber sensor was chosen in this work to compose a high-performance
optical fiber accelerometer for space applications. One advantage of the optical fiber in-
terferometers is their high sensitivity, which, on the other hand, generates a drawback
of sensing even undesirable spurious external disturbances. As a result, this can lead to
signal fading and degradation of the measurement (MARTIN, 2018).
Some authors choose to isolate the interferometer from the environment by placing
it inside a container (VLASOV et al., 2020). In the literature, one can find several papers
on the stabilization of optical interferometers by linear control techniques, such as the
classical proportional-integral-derivative (PID) controller (JACKSON et al., 1980; FRITSCH;
ADAMOVSKY, 1981; XIE et al., 2009). On the other hand, our research group developed
nonlinear control systems for stabilization of optical interferometers based on variable
structure and sliding modes (VS/SM), which provides stabilization of the operation point
in its optimal point of sensitivity (in the phase quadrature point) and suppression of the
signal fading (MARTIN et al., 2017), while presenting high robustness, high operating range,
simplicity, and low-cost (MARTIN et al., 2017; MARTIN, 2018; FELÃO et al., 2019).
Therefore, in this work, it is proposed an all-fiber optic accelerometer presenting high
sensitivity, accuracy, precision, robustness, and low drift for space applications, compris-
ing an interrogator based on optical interferometry with nonlinear control. In order to
choose the sensing element configuration to compose the accelerometer, an study was ac-
complished on the existent configurations and their specificities, discussed in Section 1.1.

1.1 Sensing element configuration

There are four main configurations for fiber optic accelerometers in the literature, or
opto-mechanical accelerometers, shwon in Fig. 1.1. All of them consist of attaching the
optical fiber to a movable mechanical device to submit it to a strain due to a given force.
An inertial mass is used to generate this strain and the resulting displacement of the
inertial mass is measured by an interferometer, alternatively, other interrogation methods
may be used, as shown by CAZO et al. (2003), LINESSIO et al. (2019), and LIU et al. (2019).
The first configuration is an inertial mass supported by optical fibers, where the fibers
work like a spring. A simplified demonstration of this accelerometer is presented in
Fig. 1.1a, according to the authors (TVETEN et al., 1980). This configuration provided
a sensitivity of approximately 2.5 rad/g and a resonance frequency of 400 Hz. Then,
maintaining this principle, lateral move restraints were shown to improve the operation
bandwidth, reaching 500 Hz (CHEN et al., 1999). A similar configuration was developed
by members of our research group at the Institute for Advanced Studies (IEAv), although
CHAPTER 1. INTRODUCTION 25

FIGURE 1.1 – Different optical fiber accelerometer sensing element configurations. (a) Inertial mass
supported by optical fiber, (b) cantilever, (c) flexural disk, and (d) rubber mandrel.
(a) (b)

Optical fiber

Optical fiber
Inertial
mass

a a

Cantilever

Spring

Support Support

(c) (d)

Optical fiber

Flexural disk
Soft
rubber
mandrel
Optical a
fiber coil
a

Support
Support

Source: (a) adapted from TVETEN et al. 1980, (b) from ANTUNES et al. 2009, (c) from VOHRA et al. 1996,
and (d) from GARDNER; GARRETT 1988.

not using interferometric interrogation (CAZO et al., 2003; CARVALHO et al., 2013; FER-
REIRA, 2020), limiting the lateral movement by attaching two optical fibers in each of the
three orthogonal directions, which resulted in a six-degree freedom accelerometer with
cross-influence in the order of 10 parts per million (CAZO et al., 2008; CAZO et al., 2009).
The inertial mass can be held in place by the optical fibers together with some other
material, for example, an elastic cylinder (WANG et al., 2015). This group also reported
miniaturized elastic shell with an inertial mass to strain an fiber Bragg grating (FBG)
that was interrogated by an unbalanced interferometer, with a sensitivity of 54 pm/g
and resonant frequency of 480 Hz (WANG et al., 2016). Additionally, diaphragms inside a
rigid compartment can be used to reduce the lateral movement, ZHANG et al. (2016) used
an interferometer to interrogate a fiber laser, with resultant sensitivity of 50 pm/g and
CHAPTER 1. INTRODUCTION 26

flat frequency response up to 200 Hz. A recent work used a cylindrical metal housing
(with screw threads to facilitate the assembly) to increase sensitivity and reduce lateral
vibration, the resultant accelerometer has a working bandwidth up to 220 Hz (ZHOU et
al., 2021).

The second sensing element configuration uses a rigid cantilever to strain the optical
fiber, composing the mechanical dynamic response of the accelerometer together with
the optical fiber. Figure 1.1b is an example of a cantilever exerting force on an FBG
due to acceleration, with a resonance frequency of 39 Hz, damping coefficient of 1.5 s−1 ,
and sensitivity of 0.997 V/g (ANTUNES et al., 2009). The optical fiber can be attached
directly to the cantilever itself, then its deformation will strain the optical fiber. The
resonance frequency for this model is near 80 Hz (BASUMALLICK et al., 2016). As happened
with the first configuration, modifications are made to improve some characteristics but
maintaining its working principle. A double cantilever was used to strain two FBGs to
improve sensibility (406.7 pm/g), with a resonance frequency of 86 Hz (PARIDA et al.,
2019). A modified cantilever was designed and achieved a resonance frequency above
1200 Hz with 90 pm/g (LINESSIO et al., 2019). The cantilever can be used to compress
the lateral of a fiber laser cavity placed between the cantilever and a fixed base, which
resulted in a frequency resonance of 440 Hz (LIU et al., 2019).
The third configuration is the use of flexural disks to strain coils of optical fiber. An
example is shown in Fig. 1.1c, where the cross-section of the sensing element is represented
with two coils of optical fiber (up and down the disk). The sensitivity for this configura-
tion can be increased due to the possibility of using long optical fibers without increasing
its overall size, reaching 50 rad/g for 5 m of fiber and a resonance frequency of 2.45 kHz
(BROWN; GARRETT, 1991). Then, by improving the disk geometric parameters, the res-
onance frequency for an aluminum disk reached 9.5 kHz with a sensitivity of 20 rad/g
for 40 m of optical fiber (VOHRA et al., 1996). Sensitivity of 6448 rad/g was reached
by modeling the flexural disk configuration with finite elements and design optimization,
however, with a reduced resonance frequency of 180 Hz (PENG et al., 2017).
The fourth main configuration comprises the optical fiber wound around a soft cylinder,
where the inertial mass in the center will deform the cylinder, as shown in Fig. 1.1d. The
reported sensitivity for this model is 104 rad/g and the resonance frequency was 327 Hz
(GARDNER; GARRETT, 1988). The same principle can be applied if the inertial mass
deforms a linear compliant material, but using it to strain the optical fiber linearly. This
sensor provides a lower sensitivity, 85 mrad/g as reported by BERKOFF; KERSEY (1996).
This configuration was used in an accelerometer with reduced common-mode noise for
lower-frequencies (< 10 kHz), presenting a sensitivity of 80 rad/g, and a dynamic range
from 0 to 1 g (DUO et al., 2018). Also, a recent noise analysis was made and modeled to
improve this kind of accelerometer (YANG et al., 2019), which shows a relation between the
CHAPTER 1. INTRODUCTION 27

output intensity noise and the arm length difference in a Michelson interferometer. Then,
it presents an optimal value for the arm length difference to minimize the noise level of
the system.
There are other sensing element configurations that are not categorized in one of the
four configurations mentioned. For example, prototypes that don’t have a protection
for the optical sensing element, in which the optical fiber is bending freely without any
hard case (RONG et al., 2017). Mechanical diaphragms bending the optical fiber (modi-
fication of a cantilever), resulting in a resonance frequency above 1.6 kHz (ZHANG et al.,
2020). Prototype of supported inertial mass with bulk components with resonance fre-
quency near 50 Hz (PÉREZ-ALONZO; SANDOVAL-ROMERO, 2021). A micro-opto-electro-
mechanical that works similar to the first configuration, but with a micro-mechanical mass
and different interrogation, resonance frequency of 1382 Hz (TAGHAVI et al., 2021). Cylin-
ders compressing a coil of modified optical fiber, measuring from 0.03 to 40 g, resonance
frequency of 2300 Hz, and sensitivity of 44.3 pm/g (HTEIN et al., 2021).
In this work, it was chosen to study the first sensing element configuration which con-
sists of an inertial mass supported by optical fibers, since it presents sensitivity suitable to
space applications and resonance frequency above the signal of interest (reaching 500 Hz),
as shown in the literature, and due to the relative simplicity of the mechanical structure.
Finally, due to the similarity with previous work developed by our research group, the
expertise on this type of configuration is already settled and former results will benefit the
present work. Then, two different optical fiber accelerometer sensing element locations
will be studied whereby one counts with the sensing element located in the interferometer
arm, while in the other, the sensor element is the interferometer’s light source.

1.2 Objective

This work aims to develop a high performance optical fiber accelerometer based on
interferometry with nonlinear control with high sensitivity, accuracy, precision, and low-
drift, which is small-sized, lightweight, and robust to spurious external disturbances, to
be applied at space applications, for example, for measuring low-frequency vibration (up
to 100 Hz) during a vehicle launch and acoustic environment vibrations, from 30 Hz to
10 kHz (SHEA, 2006; SPACEX, 2020).
CHAPTER 1. INTRODUCTION 28

1.3 Outline

In order to reach the objective of this work, a theoretical analysis and experimental
assembly of the fiber optic interferometers (Michelson and Mach-Zehnder)2 was performed;
the study, simulation, and implementation of the nonlinear control system developed by
MARTIN et al. (2017), including implementation in the algorithm logic to allow it to work
with the two-output from the Mach-Zehnder interferometer, developed in the present
work; the test of the controlled interferometer by locking it in the most sensitive point
of operation and characterizing an optical phase modulator; the analysis of the noises
involved in the system to define the minimum phase shift measurement; the theoretical
analysis and experimental assembly of a mechanical mass-spring-damper accelerometer.
Besides that, the study on the use of a fiber laser cavity as the sensing element and the
interferometer as the interrogator will be performed and the two configurations developed
will be compared; to achieve that, the characterization of an erbium-doped optical fiber
will be performed, aiming to construct a single-mode fiber laser to work as this sensing
element.

2
Called from now on in this text simply as Michelson interferometer and Mach-Zehnder interferometer.
2 Optical interferometry with nonlinear
control

In this chapter it is presented the elements of the interrogator used for the accelerom-
eter developed in this work. This interrogator is based on an optical interferometer with
nonlinear control. Therefore, a brief theory on optical interferometry is outlined, followed
by the principles on the variable structure and sliding modes nonlinear control. At the
end, a noise analysis is performed to provide the minimal phase detectable estimation.

2.1 Optical interferometry theory

This section presents the Michelson interferometer and the Mach-Zehnder interferom-
eter theory and model. Practical differences between them are considered and shown to
the reader. Then, the influence of physical quantities on the interferometer optical phase
modulation is discussed to understand the principle of working of the accelerometer itself.
The polarization fading is also discussed to justify the use of Faraday rotator mirrors on
the Michelson interferometer. After that, the chosen method of signal demodulation is
addressed.

2.1.1 Michelson interferometer

The Michelson interferometer1 , as well as other two beams interferometers works based
on the principle of wave interference, using two or more light waves to provide an interfer-
ence pattern, called fringes (Fig. 2.1). The basic concept is to divide a laser beam into two
beams and then recombine them creating the interference pattern2 . Any variation in one
of the separated beams translates to changes in the fringes, moving the fringes radially
outward when the path difference between the beams increases (and vice-versa) (HECHT,
1
In this text, the Michelson interferometer will be used to describe the working principle of a two
beams optical interferometer, without loss of generality.
2
If both beams have the same frequency, this is known as a homodyne interferometer. Otherwise, if
one of the beams has a different frequency, it becomes a heterodyne interferometer.
CHAPTER 2. OPTICAL INTERFEROMETRY WITH NONLINEAR CONTROL 30

2017; UDD; SPILLMAN, 2011). Thus, an ideal setup would work with one beam completely
still (reference arm) and the other suffering only the interest physical variations (sensor
arm). However, even a nanometer-scale vibration or micro kelvin temperature variation
near an optical interferometer can cause a high variation in the output signal, due to its
high sensitivity.

FIGURE 2.1 – Interference fringes.

Source: Adapted from FELÃO (2019).

The fringe pattern in the output is very visual (for a light source in the visible range of
the spectrum) and makes easy to understand the sensor behavior, however, to achieve a
high precision measurement and to allow data acquisition, the optical signal is converted
to an electrical voltage by a photodetector with a transimpedance circuit.
The electrical fields of the reference [ER (t)] and the sensor [ES (t)] beams, in a homo-
dyne interferometer, can be expressed by:

ER (t) = ER0 ·e j·ω0 ·t , (2.1a)

ES (t) = ES0 ·e j[ω0 ·t+φ(t)] , (2.1b)

where ω0 is the optical angular frequency, ER0 is the reference beam electrical field am-
plitude, ES0 is the sensor beam electrical field amplitude, φ(t) is the optical phase shift
between the two beams, and t is the time variable. The total electrical field is given by
the sum of (2.1a) and (2.1b), as follows:

ET (t) = ER0 ·e j·ω0 ·t + ES0 ·e j[ω0 ·t+φ(t)] . (2.2)

Due to limitations in the photodetector to capture frequency information above mi-

crowaves, it is not possible to measure the variable electrical field that arrives as a signal
(usually in the order of 1015 rad/s, for an optical signal). So, the irradiance is measured
instead, which is the average energy per unit area per unit time (HECHT, 2017), this time
average is over a period T  2π/ω0 , it is also historically called intensity (I). Its value is
CHAPTER 2. OPTICAL INTERFEROMETRY WITH NONLINEAR CONTROL 31

proportional to the average of the Poynting vector (UDD; SPILLMAN, 2011):

|ET |2
I(t) ∝ . (2.3)
2

Nonetheless, it is expected that the photodetector is able to capture the variation of

the phase shift. This is one of the great features presented by an interferometer, since it
is capable of obtaining measurements in the order of a fraction of the wavelength without
measuring the variation of the electrical field itself.
Working with (2.3) and using Euler’s identity, one can find an equation that considers
the two electric field amplitudes and the phase shift between them:

ER2 0 ES20
I(t) = + + ER0 ·ES0 · cos(φ(t)). (2.4)
2 2

Applying (2.3) in (2.4):

p
I(t) = IR + IS + 2 IR ·IS · cos(φ(t)), (2.5)

where IR = ER2 0 /2 and IS = ES20 /2. Then, organizing (2.5):

√
2 IR ·IS
 
IR + IS
I(t) = 1+ · cos(φ(t)) . (2.6)
2 IR + IS

Finally, we have the intensity equation represented in terms of the optical source
intensity, the visibility, and the optical phase difference:

Io  
I(t) = 1 + V · cos(φ(t)) , (2.7)
2

where Io is the optical irradiance from the source and V is the visibility which is given by
√
2 IR ·IS /(IR + IS ).
The visibility can be defined as:

Imax − Imin
V = , (2.8)
Imax + Imin

where Imax and Imin are the maximum and minimum values of intensity in the fringes.
Therefore, it represents the contrast among fringes and have values from 0 to 1. Using
(2.8) is more convenient in practical terms because it uses the output signal I instead of
IR and IS , which in most cases we don’t have access to these individual beams inside the
interferometer.
In this mathematical model of the interferometer, the visibility reaches its maximum
CHAPTER 2. OPTICAL INTERFEROMETRY WITH NONLINEAR CONTROL 32

value (unitary) if the division of the initial beam is perfect (ER0 = ES0 ) and there are
no losses. However, in real applications and experiments, this value depends on the state
of polarization of the electrical fields (which will be seen in Section 2.1.4), the coherence
length of the light source, and the losses in each arm of the interferometer. All that can
result in electrical fields with different amplitudes interfering in the output. In the worst
case, the visibility can reach zero and no fringes of interference will exist in the output.
The total phase shift [φ(t)] in the interferometer signal comprises the quasi-static phase
shift [φ0 (t)], that is an inherited characteristic from the construction of the sensor (e.g.
caused by the difference of length between arms), and the signal of interest [∆φ(t)], which
is a phase shift produced by the physical quantity under measurement, as follows:

φ(t) = ∆φ(t) + φ0 (t). (2.9)

The optical phase shift can be expressed by (KIRKENDALL; DANDRIDGE, 2004)

φ(t) = n(t)·k(t)·`(t), (2.10)

where n is the refractive index of the fiber, k is equal to 2π/λ which is the angular
wavenumber, λ is the wavelength of the light source, and ` is the physical path length.
Since the application of the interferometers usually aims to measure dynamic signals,
(2.10) can be differentiated in time3 :

dφ d(n·k·`)
= . (2.11)
dt dt

Applying the chain rule:

dφ dn dk d`
= ·k·` + n· ·` + n·k· . (2.12)
dt dt dt dt

Then dividing both sides of (2.12) by (2.10), we have:

dφ dn dk d`
= + + . (2.13)
φ n k `

Now it is possible to use (2.13) to choose what physical quantity will modulate the
optical phase shift in the interferometer. The most common is to strain the optical fiber,
changing `, and maintain the other two variables (k and n) constant.
Considering the contribution of all electronic gains and the photodetector responsivity

3
To avoid repetition of the mathematical term that indicates the dependence of time, (t), it will be
suppressed in the next few equations.
CHAPTER 2. OPTICAL INTERFEROMETRY WITH NONLINEAR CONTROL 33

on a constant A, the intensity in (2.7) with the phase shift represented by (2.9), when
converted to a voltage signal will thus be given by:

v(t) = A [1 + V · cos(∆φ(t) + φ0 )] ,
(2.14)
v(t) = A + A·V · cos(∆φ(t) + φ0 ).

The A·V factor in (2.14) represents the product of visibility and the electronic gains of
the sensor. This factor is the amplitude of the interferometric response, it is an important
value that is used to demodulate the measured signals and to find the offset value, A, that
must be removed before sending the signal into the control loop (as will be explained in
Section. 2.2).
The visibility is also obtained with v(t), by replacing (2.14) in (2.8) instead of using
I(t). With v(t)max = A + A·V and v(t)min = A − A·V , the visibility is

v(t)max − v(t)min 2 A·V

= = V. (2.15)
v(t)max + v(t)min 2A

The interferometric characteristic curve represented by (2.14) is a nonlinear function

and it has a point of operation, called point of phase quadrature (Fig. 2.2), in which the
input phase signal produces the largest output signal for a given input (i.e., point of the
highest sensitivity). In this point of operation, the response for small amplitude input
signals is considered as linear. Hence, there is a desire to maintain an interferometer in
its quadrature point to have the highest sensitivity and a direct demodulation process for
small signal input.
To achieve that, linear control techniques can be applied to stabilize the system (JACK-
SON et al., 1980; FRITSCH; ADAMOVSKY, 1981; XIE et al., 2009). This type of technique
employs a linearization around the operation point. Yet, it is not a robust solution, be-
cause any slight variation takes the operation point outside the linearized zone affecting
the controlling capabilities of the control system. In this work, a nonlinear control tech-
nique (MARTIN et al., 2017) was applied to maintain the interferometer in quadrature.
This topic is addressed in more detail in Section 2.2.
Figure 2.3 shows simulations of interferometric output signals from distinct input sig-
nals in amplitude of modulation, being measured with the interferometer working in a few
operation points, from the most sensible region to the least one. It is shown in Fig. 2.3a,
the interferometer characteristic curve of (2.7) with φ0 = π/2 (in phase quadrature), as
the solid black line, and with ∆φ(t) as a sinusoidal signal of low amplitude, as the dash-
dotted blue line. The intensity in Fig. 2.3b shows that the output signal, in this case, is
linearly proportional to the input ∆φ(t), i.e., the operation under this situation allows the
correct signal recovering. The second case, in Fig. 2.3c, presents a low modulation depth
CHAPTER 2. OPTICAL INTERFEROMETRY WITH NONLINEAR CONTROL 34

FIGURE 2.2 – Interferometric characteristic curve. The vertical lines are the phase limits for the approx-
imation of low phase modulation index providing 3% of harmonic distortion.

Source: Author.

input signal, ∆φ(t) (shown as the dash-dotted blue line), with the interferometer out of
phase quadrature, i.e., φ0 6= π/2 (shown as the solid black line), which causes the output
signal to fade and distort, as shown in Fig. 2.3d. The third case, in Fig. 2.3e, shows the
input signal ∆φ(t) (shown as the dash-dotted blue line) with a high modulation depth,
and even for the interferometer being in the quadrature point, φ0 = π/2 (shown as the
solid black line), the output signal, in Fig. 2.3f, is not a linear reproduction of the input.
It can be observed instead, that this output signal presents a multi-fringe behavior, which
still can be recovered by some phase unwrapping method.
CHAPTER 2. OPTICAL INTERFEROMETRY WITH NONLINEAR CONTROL 35

FIGURE 2.3 – Examples of interferometric responses. The left column contains 3 examples, being the
blue dot-dashed line the input phase variations and the continuous black lines are the interferometric
curves. The right column shows the results. (a) and (c) have the same amplitude of phase variation
in the input, but their outputs have different peak-to-peak values and frequency [(b) and (d)] due to
different points of operation. (e) has a much greater phase amplitude in the input and the result (f)
shows a multi-fringe signal.

(a) (b)

1
0.7

I(t)
I

-0.7
-1

- π /2 π /2 3π /2 5π /2 7π /2 time [ s]
ϕ
(c) (d)

1
I(t)

-0.7
I

-1
-1

- π /2 π /2 3π /2 5π /2 7π /2 time [ s]
ϕ
(e) (f)

1 1
I(t)
I

-1 -1

- π /2 π /2 3π /2 5π /2 7π /2 time [ s]
ϕ

Source: Adapted from UDD; SPILLMAN (2011), p. 238.

Figure 2.4 shows the simplest and practical setup of an all-fiber Michelson interfer-
ometer. A 50/50 optical fiber directional coupler (FC) is used to divide the laser beam
into two equal parts, which reflect in the mirrors at the end of each arm and then are
recombined by FC to create an interference pattern in the output. If the laser is emitting
in the visible spectrum and a screen is placed in the output, the fringes can be visualized
as shown in Fig. 2.4, or a photodetector (PD) can be used to convert the optical signal
into an electric signal to be analyzed by an oscilloscope.
CHAPTER 2. OPTICAL INTERFEROMETRY WITH NONLINEAR CONTROL 36

FIGURE 2.4 – Fundamental assemble of a fiber optic Michelson interferometer.

Laser Isolator
FRM
Connector FC

PD FRM
Output

Fringes
Oscilloscope

Source: Author.

The mirrors reflecting the light at the end of the arms are Faraday Rotator Mirrors
(FRM). As mentioned earlier, in the interferometer model it is assumed that the electric
fields interfering have the same polarization and will always interfere with each other.
However, due to imperfections during the fiber fabrication or uneven tensions that are
being applied to the fiber during the experiment, the optical fiber has birefringent char-
acteristics and can change the state of polarization in the arms of the interferometer.
Using the FRM instead of traditional mirrors avoid this polarization fading by rotating
the polarization in 90° (KIRKENDALL; DANDRIDGE, 2004; GARTLAND, 2016), which will
be explained in more detail in Section 2.1.4.
Additional care must be taken in this setup, to ensure light does not return to the laser
cavity. According to MILES et al. (1980), 0.1% of power feedback in the laser diode cavity
can result in a linewidth broadening of 40 times. Such large broadening would reduce
drastically the coherence length (Lc ) of the laser and consequently the limit of the optical
path difference (OPD) between the arms of the interferometer. The temporal coherence
of an electromagnetic wave is a time constant (τc ) in which the phase difference is the
same between the wave at a reference time t and the wave at a time t + τc (SVELTO, 2010).
Therefore, the coherent length is the light speed times the coherence time (Lc = c · τc ).
The light source bandwidth is inversely proportional to the coherence time, thus, broader
bandwidth provides smaller coherence length. For the interferometer to work, the optical
path difference of the arms cannot be longer than the coherence length, otherwise the
beams become uncorrelated in phase. Therefore, an isolator of 29 dB4 is used to reduce
the feedback of light to the laser cavity, because in the all-fiber setup the interferometer
cannot be intentionally misaligned to avoid it.

4
This was the available isolator to use at the moment of the interferometer assembly. However, the
recommended isolation is at least 50 dB (UDD; SPILLMAN, 2011, p. 261).
CHAPTER 2. OPTICAL INTERFEROMETRY WITH NONLINEAR CONTROL 37

2.1.2 Mach-Zehnder interferometer

The Mach-Zehnder interferometer is very similar to the Michelson interferometer. In-

stead of reflecting the beams in the mirrors and then recombining in the same FC, the
Mach-Zehnder uses another FC to interfere them, as shown in Fig. 2.5. Note that now
there are two possible outputs in the system.

FIGURE 2.5 – Fundamental assemble of a fiber optic Mach-Zehnder interferometer.

Laser Diode Connector Sensor

Arm Output 1
FC FC

Output 2
Reference
Arm

Oscilloscope

Source: Author.

One output has exactly the same response as in the Michelson interferometer, shown
by (2.7), and the other is 180° degrees out of phase due to energy conservation. Since
cos(φ(t) + π) = − cos(φ(t)), the two outputs are:

Io
I1 (t) = 2
[1 + V · cos(φ(t))] , (2.16a)
Io
I2 (t) = 2
[1 − V · cos(φ(t))] . (2.16b)

Taking advantage of the two outputs in this configuration, it is possible to remove

the offset value of the optical input as well as the fading due to optical intensity noise
(CHANDRA et al., 2016). Using the following:

I1 − I2
= V · cos(φ(t)), (2.17)
I1 + I2

then the output depends only on the visibility and on the variation of phase generated by
the desired measurand.
One must be careful when analyzing this signal when converted into the electrical
domain. Because each photodetector will have its responsivity and gain (the A factor
from (2.14)), and they might be different, assuming A1 and A2 , where A1 6= A2 . So, it is
important to equalize these gains before applying the described method.
It can be noted that both fiber optical interferometers present the same principle of
CHAPTER 2. OPTICAL INTERFEROMETRY WITH NONLINEAR CONTROL 38

working and are ruled by the same equation. The Michelson interferometer multiplies
every phase variation (∆φ) by a factor of 2, because the phase is affected two times in
the interferometer arm (before and after the mirrors), i.e., it presents a higher sensitivity.
On the other hand, the Mach-Zehnder doesn’t have the laser feedback problem and has
two outputs to work with. These two types of interferometer were then employed in this
work in order to verify and compare their behavior experimentally.

2.1.3 Optical phase modulation

It is possible to use different types of physical quantities to modulate the phase shift
in the interferometer by using (2.13). Whether it is by varying the refractive index of the
medium (n), the traveled physical distance (`), or the light source wavelength (k). The
optical phase modulators work by changing one of these physical quantities and they are
generally controlled by an electrical signal (in volts), thus, an important information for
a phase modulator is its scale factor, i.e., how much phase modulation is generated by a
given electrical signal (represented in rad/V).
In this work, the effectiveness of each type of modulation will influence the decision of
where to place the sensing element that holds an inertial mass to construct the accelerom-
eter. This will be explained in greater detail in Chapter 3.1.

2.1.3.1 Refractive index variation

One possible way of modulate the refractive index of the optical fiber is by changing
the temperature. Differentiating (2.9) with respect to temperature (T):
 
dφ d(k·n·`) dn d`
= =k· ·` + n· . (2.18)
dT dT dT dT

Considering that the light source will not be affected by the temperature since it is only
affecting the interferometer arms, k doesn’t change. Then, for fused silica optical fiber,
the thermo-optic coefficient is dn/dT ≈ 1·10−5 K−1 and the thermal expansion coefficient
is d`/(`·dT ) ≈ 7·10−7 K−1 (GUO et al., 2015).
Therefore, (2.18) demonstrates the possibility of modulate the optical phase by vari-
ation of temperature. This effect will be used later in this work (Section 3.2.1), whereby
an optical phase modulator based on thermal effect was constructed adding molybdenum
to optical fiber by sputtering, then an electrical current is applied so the metal can heat
and modulate the optical phase.
CHAPTER 2. OPTICAL INTERFEROMETRY WITH NONLINEAR CONTROL 39

2.1.3.2 Physical length variation

If one chooses to modify only the physical length (`), with n and k unaltered, from
(2.13) we have:
dφ d`
= , (2.19)
φ `
where dφ is the phase variation measured by the interferometer, φ is the optical phase
shift from (2.10), ` is the initial physical length and d` is the variation of the physical
length. This effect can be achieved by stretching or compressing the optical fiber, changing
the distance traveled by the light in the interferometer arm. As an example, a type of
transducer is usually made by winding a stretch of optical fiber around a cylindrical
piezoelectric ceramic, which is located in one of the interferometer arms and can work as
a phase modulator.
However, stretching the optical fiber leads the refractive index of the core to change due
to the photoelastic effect (GIALLORENZI et al., 1982). So, a correction factor is necessary5
in (2.19), becoming:

dφ n2
= εz − · [(P11 + P12 )·εr + P12 ·εz ] , (2.20)
φ 2

where P11 and P12 are Pockels coefficients of the core, εz = d`/` is the axial strain, where
d` is the variation of axial length and ` is the initial axial length, and εr = dr/r is the
radial strain, where dr is the length variation in the radial direction and r is the initial
radius. Disregarding the radial strain, one obtains:

dφ n2
= εz − ·P12 ·εz . (2.21)
φ 2

Then making
1
ξ = 1 − ·n2 ·P12 , (2.22)
2
we arrive at a variation of (2.19) with the elasto-optic correction factor ξ:

dφ d`
= ξ· . (2.23)
φ `

Finally, replacing (2.10) in (2.23), one obtains an expression relating the optic phase shift
measured by the interferometer with the physical length variation:

2π
dφ = ·n·ξ·d`. (2.24)
λ

Using values of P12 = 0.255 (ROSELLÓ-MECHÓ et al., 2016) and n = 1.45, the correction
5
The correction factor is unnecessary for free-space bulk interferometers.
CHAPTER 2. OPTICAL INTERFEROMETRY WITH NONLINEAR CONTROL 40

factor is constant for the given optical fiber as ξ = 0.732.

2.1.3.3 Wavelength variation

The other type of physical quantity considered in this work is the one that works
by modulating the wavelength of the light source (λ), while holding ` and n unchanged.
This can be achieved by feeding the interferometer with a fiber laser and, then, stretching
or compressing the fiber Bragg grating in the fiber laser can tune the laser to different
wavelengths. Again, from (2.13):
dφ dk
= . (2.25)
φ k

First, converting the expression from dk to dλ. By differentiating k = 2π/λ with

respect to λ:
dk
= −2π·λ−2 , (2.26)
dλ
then replacing in (2.25)
2π
dφ = − 2 ·n·`·dλ, (2.27)
λ
where λ is the laser wavelength when the transducer is at rest, dλ is the resultant wave-
length variation, n is the optical fiber refractive index of the interferometer, and ` is,
especially to this configuration, the length difference between the interferometer arms
(`S − `R ). Note that now the transducer is not inside one of the arms of the interferom-
eter, it is outside the sensor causing a variation in the laser wavelength that feeds the
input of the interferometer. So, if the interferometer has both arms with exactly the same
length (` = 0), the variation of wavelength in the input would not change the phase shift
inside the interferometer, i.e., it would always stay in phase. Thus, to detect the phase
variation caused by the laser wavelength, the length should be different from zero and,
besides, the length may be increased to increase the sensitivity.
The product n·` is the OPD and it is directly proportional to the sensitivity of the
interferometer when using this sensor configuration. It must be smaller than the coherence
length of the laser. This setup would also work with a laser diode by changing the drive
current, but then the OPD could only be a few centimeters long (UDD; SPILLMAN, 2011).
Hence, the advantages of a fiber laser are that the OPD can be in the order of a hundred
meters if working with a single longitudinal mode, since the laser cavity is easily exposed to
attach a transducer and no electrically active component is required in the interferometer.
Another option is to use a broadband light reflecting in an FBG and then stretch it to
shift the spectrum (ZHANG et al., 2020).
In summary, in this section it was discussed that the refractive index of the optical fiber
(n), the physical length of the optical fiber (`) or the wavelength of the light source (λ) are
CHAPTER 2. OPTICAL INTERFEROMETRY WITH NONLINEAR CONTROL 41

physical quantities capable of modulate the interferometric signal. As an example, Fig. 2.4
shows a simplified interferometer configuration where n or ` can be used to modulate the
interferometer signal, thus turning it in a transducer. Through the thermo-optical effect,
if one heats the optical fiber, its refractive index changes and this configuration may
operate as a temperature sensor. Similarly, if the optical fiber is stretched or compressed
this setup may operate as a force sensor. Particularly this configuration is of interest in
this work since to measure acceleration, an indirect method should be used, for example
by measuring the force exerted on a mass body. Another interferometer configuration is
shown in Fig. 2.5 where the physical quantity actuates in the light source (a laser in this
case) on the input of the interferometer, using λ to modulate the signal. This configuration
is also of interest since the laser cavity length may be stretched or compressed by a mass
to change λ and measure acceleration. Alternatively, the laser cavity mirror (an FBG)
may be stretched or compressed to measure acceleration also. Therefore, in this work,
based on these physical principles, configurations based on Fig. 2.4 and Fig. 2.5 will be
evaluated to measure acceleration. In addition, a fiber optical phase modulator based on
thermo-optical effect is proposed and evaluated with a configuration based on Fig. 2.4.

2.1.4 Polarization-induced fading

For simplicity, Section 2.1.1 considered both interfering beams with the same state
of polarization. However, low-cost all-fiber interferometers can be built with standard
telecommunication single-mode optical fiber. This means that they are susceptible to
fabrication imperfections that result in uneven mechanical stresses distributed along the
optical fiber. The intrinsic mechanical stresses combined with random mechanical stresses6
make the optical fiber randomly birefringent (GARTLAND, 2016). Then, each interferom-
eter’s arm might alter the state of polarization to a different position and degrade the
visibility when the beams are recombined (KERSEY et al., 1988).
The first solution is to assemble the entire interferometer with polarization-maintaining
components. That would increase the cost and the complexity of construction (because of
the angle setting during the splices). Another solution, only for the Michelson interferom-
eter, is to use Faraday rotator mirrors as mentioned earlier (MARTINELLI, 1989; PISTONI;
MARTINELLI, 1991; KERSEY et al., 1991). For the Mach-Zehnder setup, a combination of
polarization controllers can be placed on the output signal to find the best signal to work
with (KIRKENDALL; DANDRIDGE, 2004).
To understand how the polarization is changed along the interferometers and how
the corrections really work, it is necessary to go back to (2.1) and consider each of the

6
Stresses from bending the optical fiber when constructing the interferometer or any other reason
that can cause mechanical stress.
CHAPTER 2. OPTICAL INTERFEROMETRY WITH NONLINEAR CONTROL 42

electrical fields from the interferometer arms as a superposition of electrical fields in x

and y, that are travelling in z, using Jones calculus:
" #
~ = x̂·Ex ·e j·δx + ŷ·Ey ·e j·δy Ex ·e j·δx
E = . (2.28)
Ey ·e j·δy

Ex and Ey are the electrical field amplitude vectors in x and y directions, δx and δy are
phase shift to some reference point in z. If δ = δy − δx , any state of polarization can be
represented by choosing Ex , Ey and δ (COLLET, 2005; UDD; SPILLMAN, 2011).
~ in
The state of polarization can be visualized by plotting the motion of the tip of E
the xy-plane for z = 0, resulting in the polarization ellipse as shown in Fig. 2.6.

FIGURE 2.6 – Polarization ellipse.

Source: Adapted from UDD; SPILLMAN, (2011).

From the polarization ellipse, θ = tan−1 (Ey /Ex ) indicates the angle between x-axis
and the electric field at z = 0, χ indicates the ellipticity ( − π/4 ≤ χ ≤ π/4), ψ indicates the
orientation angle of the ellipse (0 ≤ ψ ≤ π), and they can be calculated by the following
expressions:

2Ex ·Ey
tan(2ψ) = · cos(δ), (2.29a)
Ex2 − Ey2
2Ex ·Ey
tan(2χ) = 2 · sin(δ). (2.29b)
Ex + Ey2

The ellipse is very helpful to visualize the actual state of polarization, but if one
desires to observe the evolution among multiple states of polarization, then the Poincaré
sphere is often preferred because it makes it possible to create a trace on its surface
from the initial state to the final. It represents all the states of polarization possible
CHAPTER 2. OPTICAL INTERFEROMETRY WITH NONLINEAR CONTROL 43

on the surface of a sphere. The north pole is the right circular polarization and the
south pole is the left circular (IEEE STD, 1979, p. 62). The equator represents the linear
polarization, being the horizontal polarization when intersecting +x-axis and vertical at
−x-axis. Finally, the northern hemisphere represents the right elliptically polarized, and
the southern hemisphere the left elliptically polarized.
To define a polarization state location on the sphere it is used the Stokes vector:
   
s0 Ex2 + Ey2
s1   Ex2 − Ey2 
   
~s =   = 
   . (2.30)
s 2E
 2  x y ·E · cos(δ) 

s3 2Ex ·Ey · sin(δ)

It can be shown the independence of s1 , s2 and s3 . So, the geometric center of the sphere
is located in the origin of the coordinate system, s0 represents the total intensity of the
beam and is the radius of the sphere. The variables s1 , s2 and s3 are on the x, y and
y-axis, respectively (Fig. 2.7).

FIGURE 2.7 – Poincaré sphere.

s3 P

2χ s2 y
s1

2ψ
x

Source: Adapted from COLLET, (2005).

The visibility can be better defined as:

√
2 IR ·IS
V = · cos(η). (2.31)
IR + IS

where η is half angle between two states of polarization in the Poincaré sphere. Therefore,
if the two points are the same, V = 1. if they are in exact opposite sides of the sphere,
V = 0.
It is possible to simulate the Michelson interferometer by using Jones calculus (YARIV,
1987; BIRKS; MORKEL, 1988), all the electric fields are 2x1 matrices and the optical com-
CHAPTER 2. OPTICAL INTERFEROMETRY WITH NONLINEAR CONTROL 44

ponents are 2x2, as represented by Table 2.1. The laser source is represented by a generic
electric field (LP ), whatever the polarization of the laser is, if any, it won’t affect the result
of the simulation. In the end, what matters is, after all the possible random changes in
the state of polarization inside the interferometer arms, the returning fields present the
same state of polarization to fully interfere. The directional coupler does have a model
(BIRKS; MORKEL, 1988) and can be birefringent in the coupling region (which is very
short in length and close to each other), but in this case, it is disregard and approximated
to simply divide the amplitude by 2. Since the points on the sphere are going to be
normalized by s0 , it does not affect our simulation results. This means the same laser
polarization enters both arms. The sensor arm and the reference arm are going to have
different polarization changes to simulate the worst-case scenario, RSf and RR f 7
. Then,
◦
the fields travel through a Faraday rotator of 45 and reflect on usual mirrors. Returning
to the Faraday rotators, which now rotates -45◦ . Its rotation depends on the direction of
the magnetic field of the device, independent of the electric field direction. Finally, the
electric fields pass through each respective polarization change due to birefringence and
these last states of polarization are the ones to compare.

TABLE 2.1 – Jones matrices of optical elements necessary to simulate the Michelson interferometer.

Optical element Symbol Jones matrix

Ex ·e j·δx
 
Input electric field LP
Ey ·e j·δy
 
−1 0
Mirror M
0 1
 
Faraday rotator cos(θ) sin(θ)
Ff
forward − sin(θ) cos(θ) θ=45◦
 
Faraday rotator cos(θ) sin(θ)
Fb
backward − sin(θ) cos(θ) θ=−45◦

cos (θ)·e j·δ/2 + sin2 (θ)·e−j·δ/2

 2 
Generic retarder j· sin(2θ)· sin(δ/2)
Rf
forward j· sin(2θ)· sin(δ/2) cos2 (θ)·e−j·δ/2 + sin2 (θ)·e j·δ/2
cos (θ)·e j·δ/2 + sin2 (θ)·e−j·δ/2
 2 
Generic retarder −j· sin(2θ)· sin(δ/2)
Rb
backward −j· sin(2θ)· sin(δ/2) cos2 (θ)·e−j·δ/2 + sin2 (θ)·e j·δ/2

The system is equated starting from the last element to the first using Table. 2.1:

ES = RSb ·F b ·M ·F f ·RSf ·LP (2.32a)

b f
ER = RR ·F b ·M ·F f ·RR ·LP (2.32b)

The random changes in the states of polarization of the arms are θ = π/26 and δ = 1.3π
7
The f index indicates if the optical component is affecting the electric field going forward and b for
the backward.
CHAPTER 2. OPTICAL INTERFEROMETRY WITH NONLINEAR CONTROL 45

for the reference arm and θ = −π/3 and δ = 1.3π for the sensor arm. It is shown by
KERSEY et al. (1991) that the result does not depend on these parameters.

The recombining electric fields state of polarization is shown in Fig. 2.8 when the
interferometer is built using Faraday rotator mirrors and with standard mirrors.

FIGURE 2.8 – Recombining electric fields state of polarization with and without Faraday rotator mirror
represented with polarization ellipses.
(b) Reference arm with
(a) Sensor arm with FRM. FRM.

(c) Sensor arm with stan- (d) Reference arm with

dard mirror. standard mirror.

Source: Author.

Then, a more detailed analysis can be made by observing the polarization after each
optical component and plotting the points on the Poincaré sphere, Fig. 2.9. Therefore, it
can be concluded that using the FRMs instead of regular mirrors in a Michelson interfer-
ometer avoids the polarization fading.
The solution for polarization fading for the Mach-Zehnder interferometer was not
addressed in this work. Therefore, a visibility stability experiment in the results section
was accomplished to verify how much the visibility is affected due to unintentional changes
in the state of polarization.
CHAPTER 2. OPTICAL INTERFEROMETRY WITH NONLINEAR CONTROL 46

FIGURE 2.9 – Simulation of polarization states for Michelson interferometer with and without Faraday
rotator mirror represented on the Poincaré sphere. For (a) and (b), P0 → P1 is the generic retarder of
the optical fiber, P1 → P2 is the Faraday rotator, P2 → P3 is the mirror, P3 → P4 is the Faraday rotator
for the returning beam, P4 → P5 is the generic retarder for the returning beam. For (c) and (d), P0 → P1
is the generic retarder of the optical fiber, P1 → P2 is the mirror, P2 → P3 is the generic retarder for the
returning beam.
(a) Sensor arm with FRM. (b) Reference arm with FRM.

(c) Sensor arm with standard mirror. (d) Reference arm with standard mirror.

Source: Author.
CHAPTER 2. OPTICAL INTERFEROMETRY WITH NONLINEAR CONTROL 47

2.1.5 Interferometer signal demodulation for low phase demodulation

index

To avoid the need for phase demodulation techniques, the amplitude of the signal of
interest [∆φ(t)] in (2.14) should operate in low modulation index, i.e., should be small
enough to match the same condition as in Fig. 2.3a, approximating the cosine curve to a
straight line. The response of the interferometer locked in phase quadrature (Section 2.2)
will be  π
v(t) = A + A·V · cos ∆φ(t) + . (2.33)
2
Then, by trigonometry:
v(t) = A − A·V · sin(∆φ(t)). (2.34)

If ∆φ(t) is small enough, sin(∆φ(t)) ≈ ∆φ(t), which leads to

v(t) ≈ A − A·V ·∆φ(t). (2.35)

It can be noted that from (2.35) the signal of interest is directly retrievable from the
output of the interferometer without performing any demodulation, provided the A·V
parameter is known and removing the DC offset A.
Considering a dynamic signal

∆φ(t) = φs · sin(ω·t + ψ), (2.36)

where φs is the amplitude of the signal in radians (provided that it is inside the interfer-
ometric curve), ω is the angular frequency and ψ is the signal phase shift in relation to a
given reference, (2.35) becomes

v(t) ≈ A − A·V ·φs · sin(ω·t + ψ). (2.37)

The next step is how small must φs be to establish the small-signal condition. During
the experiments, ∆φ was always a sinusoidal signal, so before collecting the interferometer
output data, its total harmonic distortion (THD) could be measured as a verification
parameter. The THD value of 3% was chosen as the maximum value to consider small-
signal operation, otherwise, the amplitude φs should be reduced to keep the THD below
this value. Numerically simulating (2.33), it was found that 3% THD is equivalent to
a phase variation of 0.83 rad from the quadrature point. Corresponding to an error of
11.1% between the interferometer cosine function and the tangent line to the quadrature
point (Fig. 2.2).
CHAPTER 2. OPTICAL INTERFEROMETRY WITH NONLINEAR CONTROL 48

2.2 Nonlinear control for two beam interferometry

The nonlinear control theory used in this work is based on variable structure and
sliding modes, which has been developed, proved stable, and applied to two-beam inter-
ferometry by MARTIN et al. (2017). The block diagram of the controlled system comprises
the interferometer, the low pass filter, the sign function, and the integrator blocks, besides
summing operators all connected as shown in Fig 2.10. In this diagram, the interferome-
ter input is the signal x1 , providing the output signal v = A·V · cos(x1 ). The variable x1
is given by the sum of the control signal (x1 ) and the phase shift [φ(t)] which, in turn,
is given by the sum of the signal of interest (∆φ) and the quasi-static phase shift (φ0 ).
Note that the interferometer model (plant) is given by A·V · cos(−) is already without the
offset value (A) presented in (2.14). Thus, the removal of the offset value A should be
accomplished formerly to achieve the interferometer block.

FIGURE 2.10 – Block diagram of the controlled interferometer.

.
e x2 x1 .
x1 x1 x1 v
x2 AVcos(–)

w v

Source: Adapted from MARTIN et al. (2017).

Thus, the signal v is the output in volts of the interferometer, which is filtered by the
low pass filter, providing the signal w. This filtered signal contains only the disturbance
information, then it is compared to a null setpoint. The result of this comparison gives
the error value (e) which is the variable x2 . When the error is zero, the interferometer is
in phase quadrature because this means that the low frequency content of v is zero, which
implies that the corresponding phase argument on the interferometer should be equal to
π/2.
The error signal enters the sign function, which is mathematically represented by the
nonlinear sign function [sgn(−)] multiplied by the total feedback gain γ, resulting in
γ·sgn(−). Once the output of the sign block is the input of the integrator block, it is
represented as the derivative of x1 and thus ẋ1 = γ·sgn(x2 ). Considering the fact that
CHAPTER 2. OPTICAL INTERFEROMETRY WITH NONLINEAR CONTROL 49

γ > 0 and A·V > 0, the following system will help us in this analysis:



 1, if e > 0,

sgn(e) = 0, if e = 0, (2.38)


−1, if e < 0.


Finally, the last block in this system is the integrator that provides the control signal
x1 from the input ẋ1 .
Supposing the low pass filter is working properly and the only signal in the closed-loop
is the φ0 , we can imagine the system with a by-pass in the feedback part instead of the
filter (v = w) and the only input signal in the interferometer is going to be φ0 , ignoring
the variation ∆φ for now.
Next, the evaluation of a few examples for initial conditions are presented, which can
be visually expressed in Fig. 2.11:

1. For an initial condition as 0 6 φ0 6 π/2 the output is going to be v > 0, w > 0.

The error is e = w − 0, leading to e > 0. Using (2.38), ẋ1 > 0. This will result in
the increase of the integrator output, consequently increasing the value of x1 and
making it closer to π/2, until e = 0.

2. Similarly, if the initial condition is π/2 6 φ0 6 π, v < 0, w < 0, e < 0 and ẋ1 < 0,
this will lead the integrator output to decrease, carrying the value of x1 closer to
π/2, until e = 0.

3. If the initial condition is already the quadrature condition, φ0 = π/2, then v = 0,

w = 0, e = 0, ẋ1 = 0, and x1 = 0, maintaining the interferometer in quadrature.

4. Considering π 6 φ0 6 3π/2, then v < 0, e < 0, ẋ1 < 0, the system will decrease the
value of x1 leading to π. Since the system passes the value of π, one can apply the
same analysis of Item 2.

5. If the initial condition is 3π/2 6 φ0 6 2π, then v > 0, e > 0, ẋ1 > 0, x1 will increase
up to 2π. Once it reaches 2π, the analysis is the same as in Item 1.

Drawing some conclusions from these analyses, it is noticeable that the integer odd
numbers multiple of π/2 are equilibrium points in this system, as it will be shown next in
the space state representation of the system. Also, those equilibrium points with positive
derivatives are unstable and the ones with negative derivatives are stable.
The equilibrium point is the phase quadrature point, where it is possible to measure
the interest signal ∆φ directly on the interferometer output by keeping the operation in
CHAPTER 2. OPTICAL INTERFEROMETRY WITH NONLINEAR CONTROL 50

FIGURE 2.11 – Equilibrium points stability analysis in the curve A·V · cos(x1 ).

Source: Author.

low modulation index (Section 2.1.5). Therefore, this configuration was employed in this
work. Alternatively, in a controlled condition, the control signal is: x1 = π/2 − φ0 . This
allows one to measure φ0 by analyzing the control signal itself.

2.2.1 State space representation

The state variables chosen in this text are x1 and x2 (MARTIN, 2018). So, we need
to mathematically manipulate the system to find a system that represents this controlled
plant in terms of the state variables.
Departing from the sign function output:

ẋ1 (t) = γ · sgn(x2 (t)), (2.39)

and assuming the low pass filter is capturing only the disturbance into the closed loop,
x1 (t) = x1 (t) + φ0 (t), that when derived gives:

ẋ1 (t) = ẋ1 (t) + φ̇0 (t). (2.40)

Therefore, ẋ1 (t) comes from (2.39) and (2.40):

ẋ1 (t) = γ · sgn(x2 (t)) + φ̇0 (t). (2.41)

CHAPTER 2. OPTICAL INTERFEROMETRY WITH NONLINEAR CONTROL 51

The first order low pass filter has the following transfer function:

W (s) 1
= , (2.42)
V (s) 1 + τs

where τ = 1/(2π·fc ). Isolating V (s), making W (s) = X2 (s), and taking the inverse
Laplace transform, we have:

v(t) = x2 (t) + τ · ẋ2 (t). (2.43)

From the interferometer block:

v(t) = A·V · cos(x1 (t)). (2.44)

Then, one can find ẋ2 (t) from (2.43) and (2.44):

A·V x2 (t)
ẋ2 (t) = · cos(x1 (t)) − . (2.45)
τ τ

Finally, we can represent the closed loop by the following system using the state
variables (2.41) and (2.45):

ẋ1 (t) = γ·sgn(x2 (t)) + φ̇0 (t),


(2.46)
ẋ2 (t) = A·V · cos(x1 (t)) − x2 (t) .


τ τ

2.2.2 Phase plane

Using the MATLAB pplane8 code available in the open library of Mathworks, it is
possible to draw the phase plane (Fig. 2.12) using the state space equations (2.46). The
phase plane is a plot of ẋ1 × ẋ2 (blue lines) and the cosine curve of the interferometer
(black line) is represented together with the phase plane to better represent where the
stable and unstable points are located.
The phase plane of the system shows that the system is stable as long as the amplitude
of the sign function is greater than the rate of change of the perturbation, i.e., the control
condition is represented by γ > |φ̇0 |. The stability can be observed as the blue lines
converge to the stable point (π/2, 0). A saddle point (blue lines) can be observed in the
unstable point of equilibrium (3π/2, 0).
The instability can be checked numerically when the control condition is not obeyed,
i.e. γ ≤ |φ̇0 |, in Fig. 2.13 there is no convergence point.
CHAPTER 2. OPTICAL INTERFEROMETRY WITH NONLINEAR CONTROL 52

FIGURE 2.12 – Phase plane of the system when the control condition is met (γ > |φ̇0 |). The black curve
is the cosine function A·V · cos(φ(t)) and the blue lines are the phase plane trajectories.

Stable point

Unstable point

Source: Author.

FIGURE 2.13 – Phase plane of the system when the control condition is not met (γ ≤ |φ̇0 |). The black
curve is the cosine function A·V · cos(φ(t)) and the blue lines are the phase plane trajectories.

Source: Author.

As can be observed in Fig. 2.12, the blue lines around the stable point converge
abruptly to it, which is the cause of the chattering phenomenon. This can happen because
of non-idealities from the control switch. The chattering can be present on the output
signal, reducing the resolution of the system. Replacing the sign function in (2.46) by a
sigmoid, results in a less abrupt actuation to the system and it helps to reduce the effect
of the chattering phenomenon (MARTIN, 2018), as can be seen in Fig. 2.14. Note that, the
chattering can also appear due to overly high gain values (γ  |φ̇0 |). So, it is important
to experimentally calibrate the gain according to the level of the disturbances impacting
the interferometer in each application.
CHAPTER 2. OPTICAL INTERFEROMETRY WITH NONLINEAR CONTROL 53

FIGURE 2.14 – Phase plane of the system using sigmoid instead of signal function. The black curve is
the cosine function A·V · cos(φ(t)) and the blue lines are the phase plane trajectories.

Source: Author.

Thus, the controller used in this work employed the sigmoid function which presents
less chattering in practical applications.

2.2.3 Simulation

To analyze the response of the controlled plant to a sinusoidal signal of interest with
a spurious external disturbance, a simulation was executed using Simulink software and
the respective diagram is shown in Fig. 2.15. The signal of interest was synthesized by
a function generator providing a sine wave with 300 Hz of frequency and 1 V amplitude.
The spurious external disturbance was represented by a combination of a ramp with a
slope of 50 and a sine wave with 5 Hz of frequency and 2 V amplitude. The input signal
of the interferometer comprises these two signals (after passing through the piezoeletric
phase modulator “signal PZT”) summed with the control signal. A “delay to activate
control” block was inserted in the feedback branch to turn the control on after a given
time delay to show the difference in the output when the control is turned off and on. This
delay was set to activate after 100 ms. The scale factor values of the PZT actuators are
generic values of 0.18 rad/V for the feedback PZT and 0.408 rad/V for the signal PZT,
obtained from MARTIN (2018). Note that the values used as signals amplitudes and gains
are estimated, the sign function output voltage is 1 V followed by a gain γv = 15 V/V
(this will be the variable gain set inside the digital platform), the linear amplifier has gain
of 20 V/V, the simulated “feedback PZT” actuator has a scale factor of 0.18 rad/V. The
product of all these gains together with the integrator constant (1 divided by second)8 ,
8
A block diagram can represent unrelated systems, so, the output dimension of a transfer function
block is the input dimension multiplied by the transfer function dimension. For integrator blocks it is
used an integration constant divided by an integration time (OGATA, 2010, p. 18).
CHAPTER 2. OPTICAL INTERFEROMETRY WITH NONLINEAR CONTROL 54

provides a general feedback loop gain of 54 rad/s.

FIGURE 2.15 – Control loop block diagram on Simulink.

Signal of interest
Spurious external
disturbance
300 Hz

Input signal [rad]

Linear Input signal w/

amplifier disturbances [rad]
Control signal
x
Interferometer
output [V]

w v

Source: Author.

The Figure 2.16 shows the response output of the simulation obtained from the block
diagram of Fig. 2.15. The first graph is the input signal, i.e., the signal of interest ∆φ.
The second graph is the signal of interest together with the spurious external disturbance
(∆φ + φ0 ). The third graph is the output signal (v) without control actuation before the
100 ms mark and controlled after that. The fourth graph is the response of sign function.
The fifth graph is the final feedback control signal composed of the integrated sign function
(x1 ). Note that when the interferometer output has an offset distant from zero, the sign
function stays unchanged for a longer period, for example, the first actuation at 0.1 s has
a duration of 14.8 ms, while the last actuation before the 0.3 s has a duration of 1.8 ms.
When the sign function stays unchanged for a longer period, the integrator output is
higher (e.g., 0.83 rad during 14.8 ms versus 0.10 rad during 1.8 ms). For the feedback
loop gain γ = 54 rad/s set in the simulation the control settling time is approximately
50 ms.
Therefore, in this simulation, it can be observed that the nonlinear control system
can work correctly by keeping the system in phase quadrature point and suppressing the
spurious external disturbance after turning on the control system and after the settling
time. This allows the system to directly recover the signal of interest under low modulation
index operation, as shown in the interferometer output signal in Fig. 2.16 after 150 ms.
CHAPTER 2. OPTICAL INTERFEROMETRY WITH NONLINEAR CONTROL 55

FIGURE 2.16 – Control actuating in simulation.

Source: Author.

2.2.4 Control setpoint out of zero

The nonlinear control plant stability was evaluated by choosing the setpoint out of zero.
This can be very useful in situations in which it is required to control the interferometer
in a point different than π/2. For example, in the work developed by ANDRADE (2019),
where two Michelson interferometers operate each one at one orthogonal polarization,
sharing most of the same optical path, forming a control interferometer and a sensor
interferometer. It is very unlikely that these two interferometers are assembled with
exactly the same phase shift. So, the control interferometer should operate at a point of
operation out of the quadrature point in order to place the sensor interferometer in the
CHAPTER 2. OPTICAL INTERFEROMETRY WITH NONLINEAR CONTROL 56

phase quadrature point of operation.

In this case, the control loop can be better represented by Fig. 2.17. The offset value
(A) is removed inside the digital platform myRIO by subtracting the measured value
before sending the signal to the control loop. Then, the subtraction of value A can be
placed before the low-pass filter. Another interesting fact about this controller is that the
setpoint can be changed in the same manner, hence the subtraction of the variable sp.

FIGURE 2.17 – Block diagram of the controlled interferometer with offset A removal and random set-
point sp.

.
e x2 x1 .
x1 x1 x1 v
x2 A+AVcos(–)

w v1

sp A

Source: Adapted from MARTIN et al. (2017).

Calculating the same steps from Section 2.2.1 for the updated block diagram in
Fig. 2.17, the state space variables are:

ẋ1 (t) = γ·sgn(x2 (t)) + φ̇0 (t),


(2.47)
ẋ2 (t) = A·V · cos(x1 (t)) − x2 (t) − sp .


τ τ τ

These variables are used to calculate the phase plane of the system for two values of
setpoints to show the system is still stable, 80% of A·V (Fig. 2.18) and minus 80% of A·V
(Fig. 2.19). From the point of view of the controller, the interferometric cosine curve is
shifted in the y-axis.
As can be seen in the phase planes of Fig. 2.18 and Fig. 2.19, the setpoint value at
80% and -80% of the A·V factor, respectivly, provide stable operation points. Regarding
the work of ANDRADE (2019), this means that the operation point of one interferometer
can be set out of quadrature to place the other interferometer in the phase quadrature
point.
CHAPTER 2. OPTICAL INTERFEROMETRY WITH NONLINEAR CONTROL 57

FIGURE 2.18 – Phase plane of the system with setpoint equal to 0.8·A·V .

AV{sp Stable point Unstable point

-AV{sp

Source: Author.

FIGURE 2.19 – Phase plane of the system with setpoint equal to −0.8·A·V .

AV{sp

Stable point

-AV{sp Unstable point

Source: Author.

2.3 Optical detection and noise analysis

To evaluate the minimum detectable phase shift of the interferometer, an analysis

on the acquisition of the optical signal, its conversion to electrical signal and the noises
present along this process (optical and electrical) should be performed. Typically, in an
optical link three noise sources are prominent, the thermal noise, the shot noise, and the
relative intensity noise. However, as the physical length of the optical system increases the
low-frequency 1/f noise becomes significant to the system as well. For example, as seen
in Section 2.1.3.3, the optical path difference can be increased to improve the sensitivity
of the interferometer which, in turn, may increase the 1/f noise.
Next, a brief analysis of noises affecting the sensors is performed to identify which
noise source may limit the system resolution.
CHAPTER 2. OPTICAL INTERFEROMETRY WITH NONLINEAR CONTROL 58

2.3.1 Thermal noise

This noise is introduced in the system every time there is a resistor in the circuit. The
thermal noise contributed by a resistance R is commonly represented by the mean square
noise current (CHANG et al., 2002):

4 kB ·T ·∆f
i2t = , (2.48)
R

kB is the Boltzmann constant, T is the room temperature in Kelvin, ∆f is the frequency

bandwidth of the system or in practical cases it can be the measurement bandwidth filter
of the spectrum analyzer. Note that the thermal noise does depend on the bandwidth,
but it does not depend on the central frequency. It is a white-noise source.
The electrical circuit model to calculate the power dissipated by the thermal noise is
a noiseless resistance in parallel with the noise current. Placing a matching impedance
load in the output, the power delivered to the load is going to be (i2t /4)R, resulting in

Pt = kB ·T ·∆f. (2.49)

Considering 1 Hz bandwidth and a standard condition of temperature and pressure,

T = 298.15 K (25◦ C), then Pt = −173.85 dBm.

2.3.2 Shot noise

The shot noise is generated in the photodetector due to the random charge carrier
transits in the semiconductor junction. The average current generated by the photodetec-
tor (Ipd ) is a result of a series of independent random events (it is a white noise), which
causes the shot noise, represented by the mean square current (CHANG et al., 2002):

i2sn = 2 q · Ipd · ∆f, (2.50)

where q is the elementary charge. Note that the shot noise is linearly proportional to Ipd ,
while the thermal noise is constant, considering the temperature is also usually constant.

2.3.3 Relative intensity noise

The relative intensity noise (rin) comes from the laser intensity fluctuations due to
random spontaneous emission. The maximum value is just above the laser threshold when
the laser cavity modes are still competing and decrease once the laser is more stable. It
CHAPTER 2. OPTICAL INTERFEROMETRY WITH NONLINEAR CONTROL 59

is defined as:
2
SδP (f ) · ∆f
rin(f ) = (2.51)
P2
where SδP (f ) is the spectral density of the optical power fluctuation and P 2 is the laser
average optical power squared. It is often referred in its logarithmic form:

RIN(f ) = 10 log10 [rin(f )]. (2.52)

Analyzing this noise in terms of current in the resistor after the optical detection, and
2 2
using same detector and circuit for both measurements, SδP /P 2 will be equal i2rin /Ipd .
The relative intensity noise mean square current is (CHANG et al., 2002):

2
i2rin = rin · Ipd · ∆f, (2.53)

where Ipd is the average current from photodetector used previously in the shot noise
calculation. Note that the rin noise also depends on the average current value generated
from the photodetector, however it is a quadratic function. Therefore, for high current
values, the rin noise is predominant over the shot noise.
The rin noise is different from the thermal noise and shot noise because it varies in
frequency, it is not white noise. However, typically for optical links with lasers in single-
mode operation, the rin is almost constant for low frequencies. Since this noise depends
on the spontaneous emission intensity, different constructions of lasers have different RIN
values, for example, solid-state lasers can have a RIN value around 30 dB lower than laser
diodes.
The figure 2.20 is a simulation of the last three noises considering a laser with illustra-
tive RIN value of -164 dB for different output optical power, for a bandwidth of 1 Hz. In
this simulation the thermal noise is the predominant noise source below 0.95 mW of op-
tical power, then the shot noise becomes predominant up to 7.78 mW, and finally the rin
noise is the predominant as the optical power from the laser increases (above 7.78 mW).

2.3.4 1/f noise

This noise is found in many areas of knowledge and yet its origin is not a consensus,
in optical fibers, it is believed to be related to temperature variations (DUAN, 2010) and
acoustic low-frequency vibrations (CONFORTI et al., 2020). It might be called flicker, pink,
excess, or contact noise, depending on the field of interest. But it is popular known as
1/f noise because the function of its spectral density is a power law (ALEXANDER, 1997;
CHAPTER 2. OPTICAL INTERFEROMETRY WITH NONLINEAR CONTROL 60

WARD; GREENWOOD, 2007),

1
S(f ) = α · , (2.54)
fβ
where, α and β are constants that fit to each system (usually 0.5< β <1.5).
Many authors identify the corner frequency (fc ) of the noise, which is quite noticeable
separating the 1/f noise from white noise sources when plotted in logarithmic scales, this
argument is valid for systems with only one source of 1/f noise (Fig. 2.21). If the system
has more than one 1/f noise source, each one may have different values for β and this
can make the division between noises less apparent.
That can be a tricky factor in the noise analysis and performance since our signal of
interest is in these low-frequencies as well. This high noise floor in the low-frequencies will
drastically increase the minimum detectable phase shift by the interferometer. The influ-
ence of 1/f noise can be minimized if the interferometer signal is modulated by a carrier
signal with high-frequency (e.g., reported from 8.8 kHz up to 500 kHz) (DANDRIDGE et
al., 1982; MICHAUD-BELLEAU et al., 2020).

2.3.5 Quantization error

The analog voltage signal usually is converted to digital signal to provide the data
sets that are the source to calculate everything else. In this work it was accomplished by
an oscilloscope (Agilent technologies MSO6034A). The user’s manual informs a resolution

FIGURE 2.20 – Noises current simulation for a laser with supposed RIN -164 dB.

×10−21
Thermal noise
4
Shot noise
rin noise
Mean square current [A2 ]

0 2 4 6 8 10
Optical Power [mW]

Source: Author.
CHAPTER 2. OPTICAL INTERFEROMETRY WITH NONLINEAR CONTROL 61

FIGURE 2.21 – Simulation of 1/f noise together with white noise.

1/f noise
White noise

Power Spectrum Density [dBm]

fc
Frequency [Hz]

Source: Author.

for the analog to digital converter of 8 bits and it increases to a maximum of 12 bits if
averaging is applied. There is also an option to limit the bandwidth to approximately
25 MHz, thus reducing the noise floor.
The root mean square voltage of the quantization error9 is represented by (PUPALAIKIS,
2017):
1 8 VDIV
VQ = √ · , (2.55)
2 3 2b
where VDIV is volts per division set in the oscilloscope vertical scale and b is the resolution
in number of bits. Considering the measurement is collected with the minimum average
count necessary to have a 12 bit resolution, the noise will depend only on the volts per
division value and a constant: VQ = 5.638 · 10−4 · VDIV .
Since the noise floor depends on the volts per division value, and this consequently
depends on the dynamic range of the interferometer output signal, this noise is usually
represented in terms of signal-to-noise ratio (SNR). The measurements in this work were
programmatically configured to adjust the signal to fit exactly 6 of the 8 voltage divisions
on the screen. A sine wave signal10 with 6/8 of the full scale has a root mean square value
equal to:
6 4 · VDIV
Vsignal = · √ . (2.56)
8 2
So, the signal to noise ratio in decibels does not depend on the volts per division, as

9
The term quantization noise is popular in the literature, but according to LEFÈVRE (2014) the term
error is more accurate because quantization is not random.
10
Because of the low-index modulation and the input signal is sinusoidal, the output interferometric
signal is also going to be a sine wave.
CHAPTER 2. OPTICAL INTERFEROMETRY WITH NONLINEAR CONTROL 62

follows:  
Vsignal
SNRoscilloscope = 20 log10 = 71.51 dB. (2.57)
VQ

In addition to this quantization error dependent on the dynamic range chosen, there
is another noise in the oscilloscope which is originated by the front-end attenuator and
amplifier, but it is negligible when the volts per division is set to the higher values (Keysight
Technologies, 2015). The volts per division is high because the A·V term from (2.14) is
usually around 2 to 6 V in the experiments, this noise from the attenuator and the amplifier
would dominate if the scale were set to around the units of millivolts.

2.3.5.1 Resolution for the control system

This subsection is an observation, not necessarily about the noise from the interfer-
ometer, but it is related to the present topic. The nonlinear control works with the same
analog voltage signal into a high impedance input in the myRIO 1900. The instrument
also has a 12 bit resolution, however, the full-scale range signal is always 20 V, which
results in a resolution of 4.883 mV. Thus, one should pay attention to this resolution if
the dynamic range of the interferometric signal is within the same magnitude. This would
lead to an incorrect measurement of the offset level, that must be subtracted before insert-
ing the signal into the closed-loop control operation. The possibility of picking a different
operation point (i.e., different than the ideal equilibrium point (2k − 1) · π/2, where k is
an integer) means that the linear relation established for the low-index modulation will
no longer be valid and the phase readings will be flawed.

2.3.6 Minimum detectable phase calculation

There are a few ways to measure the interferometer minimum detectable phase shift.
Having measured the noise signals mentioned above, one can calculate the minimum phase
by (KIRKENDALL; DANDRIDGE, 2004):
s
i2sn + i2t + i2rin
∆φmin = , (2.58)
(R · Po )2

where R is the photodetector responsivity, Po is the mean optical power and the numerator
is the sum of shot, thermal and rin noises.
Another way is to use a lock-in amplifier to measure the signal with a signal-to-noise
ratio equal to 1 (JACKSON et al., 1980). Similarly, and less rigorous, it can be measured by
the noise floor of the interferometer with an electric spectrum analyzer and then converting
this value to phase shift.
CHAPTER 2. OPTICAL INTERFEROMETRY WITH NONLINEAR CONTROL 63

In this work, the minimum detectable phase shift will be measured by taking into
consideration the noise coming from the nonlinear control loop that holds the interferom-
eter in quadrature (noise floor), since the chattering of the sliding mode can be higher
than the described noise levels. So, first, the interferometer is held in quadrature with a
chosen desired robustness to the system, i.e., with a control gain high enough to keep the
system in phase quadrature but low enough to avoid large chattering. Without any input
signal modulating the interferometric phase, the phase shift measured by the controlled
interferometer is considered the minimum detectable phase shift possible.
3 Fiber optical accelerometer

In this chapter it is presented the principle of operation of an opto-mechanical ac-

celerometer and the interrogation scheme using the optical interferometer. Following
that, the experimental setups of accelerometers based on Michelson and Mach-Zehnder
interferometers with nonlinear control are shown. At the end, the study on an erbium
doped fiber laser is reported since this type of laser will be evaluated as the sensing element
in one of the accelerometer configurations.

3.1 Opto-mechanical accelerometer operation principle

One possible way of measuring the acceleration of a inertial mass under the influence
of a resultant force (Fig. 3.1) is to use the Newton’s second law of motion (YOUNG et al.,
2008):
X
F~ = m · ~a, (3.1)
P~
where F is the vector sum of all the forces involved in the system, m is the body mass
and ~a is the acceleration vector.

FIGURE 3.1 – Mass body under influence of a force.

a
F

Source: Author.

Using the Young’s modulus of the optical fiber that is holding the inertial mass in
place (e.g., Fig 3.2), it is possible to represent the solid in the elastic regime of tension.

σ
Y = , (3.2)
ε

where Y is the Young’s modulus, σ is the stress (force by the cross-sectional area) and ε
CHAPTER 3. FIBER OPTICAL ACCELEROMETER 65

is the strain (relative deformation). Then,

F L
Y = · , (3.3)
Ac ∆L

where Ac is the optical fiber cladding cross-sectional area (in this work the optical fiber
is glued to the mass without the coating), L is the initial length of the optical fiber, ∆L
is the length variation of the optical fiber and F is the axial force.

FIGURE 3.2 – Free-body diagram of some possible setups for sensing elements. (a) Two optical fibers
holding the inertial mass on the same axis in one given direction, (b) two parallel optical fibers holding
the inertial mass in one given direction and (c) two parallel optical fibers holding the inertial mass in
each Cartesian direction.
(a) (b) (c)

F1 F2
a

F3 F4 a y
Optical Fiber
aa
z
x

Source: (a) Adapted from TVETEN et al. (1980), (b) and (c) adapted from CAZO (2011).

Isolating the force in (3.3), we have

Y ·Ac
F = ·∆L. (3.4)
L

Whatever is the sensing element design chosen (Fig. 3.2), every attachment of optical
fiber with the inertial mass will have this force actuating in the system. Then, one can
sum all these attachment points to obtain the resultant force and use (3.1) to calculate
the acceleration.
Using the Hooke’s law,
F = −K·∆L, (3.5)

where K is the spring constant1 (positive value) and the minus sign indicates the opposite
force direction exerted by the spring when compressed or stretched. Then, comparing
(3.5) with (3.4), the optical fiber spring constant is

Y ·Ac
K= . (3.6)
L

Depending on how many fibers there are glued to the inertial mass, the system will have
1
Do not confuse the variable K in this mechanical analysis with the angular wavenumber k from the
optical chapters.
CHAPTER 3. FIBER OPTICAL ACCELEROMETER 66

a resultant spring constant, KR .

The sensing element used in this work is represented by Fig. 3.3. The optical fibers
are exerting the forces F~ due to an acceleration ~a of the whole sensor. If the body is
initially at rest, then by inertia, the movement of the inertial mass relative to the walls
of the sensor, ~x, is contrary to acceleration.

FIGURE 3.3 – Scheme of the sensing element used in this work.

Optical fibers
F
a
F
m
x
y Metal rods
x
z

Source: Adapted from FERREIRA (2020).

Still studying the forces diagram in Fig. 3.3, we notice the forces are not applied to
the center of gravity of the rigid body. However, there are two metal rods passing through
the mass body to cancel the torques. One rod avoids rotation in y and z, while the two
rods combined avoid rotation in x. They also cancel out the forces in y and z directions.
So, since the system only allows forces in x direction, it can be reduced by using (3.5)
which provides the following spring-mass equivalent shown in Fig.3.4. The resultant spring
stiffness is KR = 4K.

FIGURE 3.4 – Mass-spring equivalent system for optical fiber accelerometer. (a) All four fibers converted
to springs. (b) Simplification of parallel springs. (c) Simplification of the two fixed springs in each side.
(d) Resultant mass-spring-damper mechanical model.
(a) (b) (c) (d)

K K 4K
2K 2K 4K
K m K m m m
b

Source: Author.

There must be a damper in the final model (Fig. 3.4d), because otherwise the system
would oscillate eternally with any input signal, or the position of the inertial mass would
go to infinity with an input at the resonance frequency. Neither of these scenarios is
consistent to any real system.
The analysis of the mechanical model begins with the consideration of a mass-spring
CHAPTER 3. FIBER OPTICAL ACCELEROMETER 67

system without a damper and no external force applied. Then, with (3.1):
X
F = m·a = Fspring . (3.7)

Considering that the velocity of the body is given by the derivation in time of the physical
2
displacement ( dx
dt
= ẋ) and the acceleration is the second derivative ( ddt2x = ẍ), then,
replacing (3.5) in (3.7) provides:

d2 x
m· = −K·x. (3.8)
dt2
The differential equation (3.8) describes the motion of an unforced mass-spring system
(FEYNMAN et al., 1963).
A proposed solution for (3.8) is the real part of x = x0 ·e j·ω0 ·t (ẋ = j·ω0 ·x0 ·e j·ω0 ·t and
ẍ = −ω0 2 ·x0 ·e j·ω0 ·t ). Where ω0 is the natural angular frequency of the system, because
there are no external forces, and x0 is the initial position of the inertial mass. Then,

m·ω0 2 = K. (3.9)

Therefore, the natural angular frequency of a mass-spring system is given by

r
K
ω0 = , (3.10)
m

where ω0 = 2π·f0 .
Now, let’s step up and analyze a mass-spring-damper system without external force.
X
F = m·a = Fspring + Fdamping , (3.11)
m·ẍ = −K·x − b·ẋ. (3.12)

where b is the damping coefficient.

Proposing the same kind of solution x = Re{x0 ·e j·ω·t }, where ω is not necessarily equal
to ω0 , (3.12) becomes:
ω 2 − j·Γ·ω − ω0 2 = 0, (3.13)

where Γ = b/m is the damping factor [1/s] and ω0 2 = K/m.

Solving (3.13) for ω,
Γ 1p 2
ω = j· ± −Γ + 4ω0 2 . (3.14)
2 2
CHAPTER 3. FIBER OPTICAL ACCELEROMETER 68

Considering that in most cases ω0  Γ and the frequency is always positive:

Γ
ω = j· + ω0 . (3.15)
2

We can replace (3.15) back in the proposed solution x, resulting in

Γ
x = x0 ·e − 2 ·t · e j·ω0 ·t . (3.16)

Equation (3.16) reveals that the displacement of the mass body (x) presents a sinusoidal
with exponential decay shape. Thus, with (3.16) it is possible to measure the damping
factor by observing the accelerometer movement induced by an input (e.g., step or impulse
input), resulting in a curve represented by Fig. 3.5.

FIGURE 3.5 – Unforced mass-spring-damper response simulation.

Source: Author

Finally, the analysis considering all the necessary components to compose the ac-
celerometer model was performed. A forced mass-spring-damper, where the force is a
consequence of the sensor’s acceleration.
X
F = m·a = Fspring + Fdamping + Fapplied , (3.17)
m·ẍ = −b·ẋ − K·x + Fapplied , (3.18)

then, knowing that b = m·Γ and K = m·ω0 2 , one obtains:

m·ẍ + m·Γ·ẋ + m·ω0 2 ·x = Fapplied , (3.19)

Fapplied
ẍ + Γ·ẋ + ω0 2 ·x = = aapplied . (3.20)
m

From (3.20), it is possible to recover the vehicle acceleration by measuring the dis-
placement x of the inertial mass. The natural frequency is calculated in terms of the
optical fiber dimension and the damping factor from observing the sensor’s response.
The Laplace transform can be applied in (3.20) to obtain a transfer function for the
accelerometer (OGATA, 2010). With that, the response of the accelerometer for a given
CHAPTER 3. FIBER OPTICAL ACCELEROMETER 69

input signal can be predicted.

s2 ·X(s) + Γ·s·X(s) + ω0 2 ·X(s) = A(s), (3.21)

where s is a complex frequency.

X(s) 1
= 2 . (3.22)
A(s) s + Γ·s + ω0 2

The simulation of (3.22) with an impulse acceleration as input (A(s) = 1) is shown by

Fig. 3.6a. The simulation for a step acceleration (A(s) = 1/s) is shown by Fig. 3.6b. The
natural frequency was calculated by taking the amplitude of the Fourier transform of the
resultant waveform and the damping factor by its envelope (ANDRADE, 2019). So, both
input signals provide the necessary dynamic to the system to allow its characterization.

FIGURE 3.6 – Simulation of the accelerometer using the transfer function. Real values used for kR and
m. Damping factor (Γ) is arbitrary.
(a) (b)

Source: Author.

All the simulations used the constants from Table. 3.1, which are real values from the
experimental setup.

TABLE 3.1 – Constants used to simulate accelerometer.

Description Symbol Value

Young’s modulus Y 7.54·1010 N/m2
Fiber cladding diameter D 125 µm
Mass m 22.23·10−3 kg
Fiber length L 4.5 cm

Back to (3.20), a sinusoidal acceleration signal (aapplied = A0 ·e j·ω·t ) can be supposed as

CHAPTER 3. FIBER OPTICAL ACCELEROMETER 70

an input to solve the equation and obtain a motion equation. Analogous to the previous
problems, the solution is the real part of x0 ·e j·ω·t . Then,

A0
x0 = 2
. (3.23)
ω0 − ω 2 + j·Γ·ω

Since x0 resulted in a complex number, from |x0 |2 = x0 ·x∗0 :

A0
|x0 | = p . (3.24)
(ω0 − ω 2 )2 + Γ2 ·ω 2
2

The numeric simulation of (3.24) is presented by Fig. 3.7. All the constants are arbitrary,
the objective is to visualize the frequency response shape of the accelerometer.

FIGURE 3.7 – Accelerometer frequency response simulation for a sinusoidal input. Arbitrary value for
Γ.

Source: Author

Figure 3.7 shows a typical shape of a mass-dumper-spring model, with a resonance

around 306 Hz, a nearly flat region for frequencies below the resonance, reaching 3 dB
difference from start frequency at around 168 Hz, and a decaying region (40 dB by decade)
for frequencies above the resonance. This result shows that the mechanical structure of
this accelerometer model should be operated for frequencies between DC and 168 Hz.
Regarding the interrogation used in this work, the lower frequency becomes equal to the
nonlinear control system low-pass filter cutoff frequency (in this case, 20 Hz). Thus, the
final frequency range would be from 20 Hz to 168 Hz. Also, by analyzing the phase from
(3.23) or making a bode plot using (3.22), at the resonance frequency the signal phase
shifts π rad.
CHAPTER 3. FIBER OPTICAL ACCELEROMETER 71

3.1.1 Accelerometer interrogation

To interrogate the accelerometer with the interferometer, the physical displacement x0

from the general motion equation (3.24) is replaced by d` measured by the arms of the
interferometer from (2.24) or from the fiber laser equivalent d` changing the wavelength
from (2.27).
The sensitivity of interferometric accelerometers are commonly represented by ∆φ/a
[rad/g]. For the first case, with the accelerometer inside the interferometer:

∆φ 2π 1
= ·n·ξ · p . (3.25)
a λ (ω0 2 − ω 2 )2 + Γ2 ·ω 2

The sensitivity is then calculated to an specific frequency by using (3.25), commercial

accelerometers usually inform the sensitivity value at 100 Hz, however, most authors
choose to indicate the sensitivity for a DC acceleration (ω = 0 rad/s) as a method to
obtain a figure of merit. Then, replacing w = 0 rad/s and w0 2 = KR /m into (3.25), one
obtains:
∆φ 2·n·ξ·L·m
= , (3.26)
a λ·Y ·D2
where D is the fiber cladding diameter.
For the second case, with the accelerometer in the fiber laser cavity that feeds the
interferometer, the sensitivity is given by:

∆φ −2π 1
= 2 ·n·`·C · p , (3.27)
a λ (ω0 − ω 2 )2 + Γ2 ·ω 2
2

where n·` is the OPD that can be increased to the limit of the laser coherence length
and C is a conversion factor for variation of laser wavelength due to accelerometer motion
(dλ/d`). Considering a constant acceleration, the sensitivity becomes

∆φ −2·n·`·C·L·m
= . (3.28)
a λ2 ·Y ·D2

The equations (3.26) and (3.28) are very useful to analyze and modify the project in
terms of the mechanical constants. On the other hand, (3.25) and (3.27), allow to estimate
the final accelerometer characterization in frequency.
Therefore, these equations provide information on how to set the accelerometer pa-
rameters to meet the requirements of a given application.
CHAPTER 3. FIBER OPTICAL ACCELEROMETER 72

3.2 Accelerometer based on an interferometer with nonlin-

ear control experimental setups

In this section, the experimental setup of a Michelson interferometer with nonlinear

control is reported. Next, a fiber optical phase modulator developed in this work is re-
ported since it was employed to verify the correct working of the Michelson interferometer.
Finally, the experimental setup of the Mach-Zehnder interferometer with nonlinear control
is also presented, which is used to analyze the erbium-doped fiber laser as light source to
the interferometer and to assemble the accelerometer sensing element in the sensor arm
of the interferometer.

3.2.1 Michelson experimental setup

A Michelson interferometer was assembled in the laboratory, as shown in the schematic

of Fig. 3.8. The setup comprised a laser diode (7 mW at 1536 nm), an isolator (29 dB), a
2×2 fiber coupler (FC)2 , two Faraday rotator mirrors (FRM) to avoid polarization fading,
three fiber optical phase modulators (FOM1, FOM2, and FOM3) based on piezoelectric
cylinders, one fiber optical phase modulator based on temperature (FOM-MC), and a
photodetector (PD). The nonlinear control system was implemented within the develop-
ment platform myRIO (National Instruments) and comprised a DC suppressor, a low pass
filter (LPF), a sign function multiplied by a virtual gain [γv ·sgn(−)], an integrator based
on the Simpson’s rule, after that, the signal was inverted and sent to the analog output.
Then, the myRIO output was connected to a signal conditioning circuit that fed a lin-
ear voltage amplifier, finally, the amplified control signal voltage was applied to FOM2,
closing the loop. Also, the setup comprised an oscilloscope and a function generator that
were connected to a personal computer by a GPIB bus.
The connector between the laser diode and the isolator is shown in the diagram because
it plays an important role in the noise introduced in the sensor. When it is fully connected
the interferometric signal is noisy, probably due to the phase noise from the laser diode
caused by the feedback of light into the cavity. Considering a good fiber alignment and
that the fibers do not touch each other in the connector, the Fresnel reflections will
feedback 7.68% of the laser power back into the laser cavity, which can broaden several
times the laser linewidth. So, the connector position should be set while monitoring the
interferometer output signal in the oscilloscope, aiming to minimize its noise.
The fiber optical phase modulators (FOMs) in this first setup are based on piezoelectric
cylinders that are wound with standard single-mode optical fiber. When electrical voltage
2
This setup works with only one FC. The second FC was used in this work to decrease the input
signal by -3 dB, to reduce the light feedback to the laser diode.
CHAPTER 3. FIBER OPTICAL ACCELEROMETER 73

FIGURE 3.8 – Michelson interferometer with nonlinear control.

FOM1
Laser Diode Isolator
FRM
FC FC
Connector FOM2 FOM3
PD I(t) FRM
vPD(t)
DC v(t) R1 Rf
Suppressor
LPF γv sgn(-) ʃ -1
R2 Linear
Nonlinear control
Amplifier
Function Generator
Signal conditioning circuit

Oscilloscope GPIB Bus

GPIB / USB
Computer

Source: Author.

is applied to the FOM, it changes its diameter causing a strain in the fiber and increases
the optical path length traveled by the light in the respective arm of the interferometer.
A fiber optical phase modulator based on temperature (FOM-MC) was used to ver-
ify the proper working of the Michelson interferometer. In addition, the FOM-MC was
studied and characterized using this experimental setup (NUNES et al., 2021). The FOM-
MC was made of a standard single-mode fiber coated with molybdenum. The modulator
schematic is presented in Fig. 3.9. The molybdenum is added to the fiber cladding by
sputtering. Wires are soldered to the FOM-MC and an electric current is applied to it.
Then, the current heats the optical fiber and induces an optical phase variation, which is
mainly due to changes of the refractive index (BELAL et al., 2010).

FIGURE 3.9 – Phase modulator based on temperature (FOM-MC).

Jacket 245 nm
10 mm

Core
Cladding
Metal

Source: NUNES et al. (2021).

CHAPTER 3. FIBER OPTICAL ACCELEROMETER 74

The control signal is feedback through the FOM2. Once the interferometer is controlled
in quadrature, fiber optical phase modulators (other than the FOM2) can be used to
verify the correct working of the setup. For example, the frequency response of the
FOM1 and FOM-MC can be acquired. To accomplish that, a computer program made
in LabVIEW communicates to the function generator and the oscilloscope via GPIB,
generating sinusoidal signals to insert in FOM1 and FOM-MC (just one at a time) and
receiving corresponding output signal data from the oscilloscope. The algorithm checks
the threshold of 3% THD and collects the necessary data for the frequency response.

3.2.2 Mach-Zehnder experimental setup

A Mach-Zehnder interferometer was used to assemble the accelerometer in two dif-

ferent configurations: in the first, the accelerometer sensing element was placed in the
interferometer sensing arm; in the second, the accelerometer sensing element was place in
the interferometer input, i.e., in the erbium-doped fiber laser.
The experimental setup comprises a commercial laser diode (7 mW at 1550 nm) or
a laboratory-made erbium-doped fiber laser (only one can be connected at a time, and
most experiments in this work used the laser diode), two 2×2 fiber couplers (FC1 and
FC2 ), both are necessary for the interferometer to work, and two photodetectors (PD1
and PD2 ), as shown in Fig. 3.10.
This interferometer is controlled using the two outputs as in (2.17), and the calibration
VI is still needed to find out the actual A·V factor to be able to demodulate the signal
in (2.14). However, the A·V factor not required to maintain the Mach-Zehnder interfer-
ometer in phase quadrature (see Section 3.2.3). The VI used to implement the nonlinear
control in the myRIO platform with the two output scheme is shown in Appendix A.4.
The control signal is inserted in FOM4, while FOM5 was used eventually to insert
signals with the function generator. The output signal was collected from PD1 with the
oscilloscope.
The sensing element that forms the accelerometer was placed in the sensing arm of
the interferometer. After that, in other configuration the erbium-doped fiber laser was
used to analyze the interferometric response.
CHAPTER 3. FIBER OPTICAL ACCELEROMETER 75

FIGURE 3.10 – Complete Mach-Zehnder interferometer setup.

Erbium-doped Fiber Laser

Er 3+
Pump Laser FBG FBG
WDM

Isolator Acceleromter
FC1 FC2
OR PD1 DC
Suppressor
Laser Diode

Control
FOM4 FOM5 PD2
Signal LPF
Mach-Zehnder Interferometer

γ sgn(-)
-1 ʃ
Linear
Amplifier

Nonlinear Control Loop

Source: Author.

3.2.3 Control Implementation

The implementation of the control loop was performed inside a Field Programmable
Gate Array (FPGA) as shown in Appendix A. The analog output from myRIO is limited
between -10 V and +10 V, which is the control signal range. The linear voltage amplifier
(model CTC 260-3 from Control Technics Corp.) input accepts positive signals only,
from 0 V to +10 V. Therefore, it is needed to condition this control signal to always be
positive, with an offset added to the setpoint. To achieve that, the control signal was
digitally inverted and then inserted into the summing op-amp circuit (FRANCO, 2002).
The conditioning circuit supply voltages are +15 V and -15 V, with resistances R1 =
100 kΩ, R2 = 150 kΩ, and Rf = 10 kΩ. The signal was then limited between 0 and 2 V.
The gain of the linear voltage amplifier was set to 50 and left in this value unless told
otherwise.
The signal conditioning was not done all digitally because of the dynamic configuration
of the FPGA. Sometimes in the transition and reorganization of the code, the control
signal spiked to the maximum value, risking damaging the piezoelectric actuator.
The low pass filter is usually calibrated to have a cut-off frequency of 20 Hz (fc ), but
it really depends on the application of the sensor. For example, if one is experimenting
CHAPTER 3. FIBER OPTICAL ACCELEROMETER 76

with laser ultrasound systems the fc can be raised (on the order of kHz), since the signal
to be analyzed has a much higher frequency (on the order of MHz) and the undesirable
signal is also higher (usually below 1 kHz). The 20 Hz cut-off frequency was empirically
found to be a good fit for suppressing spurious external disturbances originated from a
variation of temperature, airflow, and vibration. Thus, this is the cut-off frequency that
will be used in the present work.
To avoid possible misinterpretations, when said that the temperature can cause a
variation of a few hertz in the fringes of interference that means the temperature is
linearly increasing with a rate that causes the fringes of interference to run many fringes
per second, not necessarily meaning the temperature is changing at this same frequency.
In other words, if the external spurious disturbance is said to be, for example, 10 Hz, that
does not mean that the temperature is varying at 10 Hz.
A voltage ramp is applied to the FOM2 to force the interferometric signal [vPD (t) =
A + A·V · cos (φ(t))] to reach the maximum and minimum values (vPDmax and vPDmin ).3
The A·V factor is calculated with these values by making (vPDmax − vPDmin )/2 = A·V ,
and later on is used to demodulate the signal. Once the software has found the maxi-
mum and minimum values, it calculates the mean voltage (A) to remove the signal offset
(this step is not necessary if the interferometer is a Mach-Zehnder as explained in Sec-
tion 3.2.2). Then, without offset, the signal goes into the control loop Virtual Instrument
(VI) (Appendix A.3).

3.3 Erbium-doped fiber laser

As presented in Section 2.1.3.3, the interferometer phase can be modulated by the

laser wavelength. Thus, one of the accelerometer configurations evaluated in this work
counts with an erbium-doped fiber laser which comprises the sensing element. In this
section, the laser rate equations will be outlined and the cavity mirrors formed by fiber
Bragg gratings will be studied, aiming to achieve a single-mode laser.
The linear relationship between the optical path delay and the phase shift sensitivity
was shown by (2.27) in Section 2.1.3.3. A laser with a long coherence length is necessary
to explore this characteristic, allowing a long optical path difference in the interferometer
to increase the sensitivity, for example, from 10 m to 100 m of path difference between

3
When calibrating the interferometer with the voltage ramp, preferably the FOM under evaluation is
inserting a low-frequency signal (set in the function generator) or, ideally, is turned off. This is because
when there is a frequency applied to the phase modulator, the ramp applied by the real-time VI from
myRIO does not achieve the maximum and minimum values in the interferometric signal (A·V value).
This results in an erroneous DC offset value to pass to the DC suppressor and, more impactful yet,
passing wrong values of maximum and minimum voltages that are used in the demodulation technique
(Section 2.1.5).
CHAPTER 3. FIBER OPTICAL ACCELEROMETER 77

the arms (CRANCH et al., 2008), and still have an interference signal to work. The erbium-
doped fiber laser is a good fit for this purpose (AMES; MAGUIRE, 2007). As mentioned
before, the coherence length is inversely proportional to the laser spectral width. The
laser spectral width may be related to the number of modes oscillating in the cavity, a
single-mode would provide the narrowest spectral width. In turn, the number of modes is
related to the size of the laser cavity, the smaller is the cavity, fewer the number of modes.
Fundamentally, a laser is made of three parts, an active medium, a pump to excite
the medium and a resonant cavity (RODRIGUES, 2019). After the pump excites the ac-
tive medium and achieves population inversion it becomes a light amplifier by stimulated
emission. Then, the resonant cavity (usually made by a pair of mirrors) provides feedback
to the oscillator. In the first moment, the pumped active medium generate light by spon-
taneous emission, then theses photons reflect on mirrors and are amplified by stimulated
emission until saturation, the resultant light beam exits the cavity by one of the mirrors
which is semi-reflective. In this section it will be presented and explained the necessary
theories and technical practices to understand and obtain an erbium-doped fiber laser.

3.3.1 Optical fiber light amplifier

The gain in an erbium-doped fiber may be found by using the rate equations from the
simplified three-level energy diagram represented in Fig. 3.11 (BALL; GLENN, 1992), where
N1 , N2 and N3 are the ion population density from levels 1, 2, and 3. The spontaneous
emission lifetime from level 2 and 3 are τ2 and τ3 . The pump rate is given by Wp . In this
case, the erbium-doped fiber laser is optically pumped by a laser at 974.4 nm. The signal
absorption and emission rates are Wa and We , respectively. The stimulated emission from
the laser transition (1–2), or signal transition for amplifiers, are photons with energy h·ν,
where h is the Planck constant and ν is the optical frequency.

FIGURE 3.11 – Erbium energy three-level diagram.

Source: Adapted from (BALL; GLENN, 1992).

CHAPTER 3. FIBER OPTICAL ACCELEROMETER 78

The rate equations corresponding to these populations are:

dN3 N3
= Wp ·N1 − , (3.29)
dt τ3

dN2 N3 N2
= + Wa ·N1 − We ·N2 − , (3.30)
dt τ3 τ2

dN1 N2
= + We ·N2 − Wa ·N1 − Wp ·N1 . (3.31)
dt τ2

Considering a steady state regime (Ṅ1 = Ṅ2 = Ṅ3 = 0), we have:

N3
= Wp ·N1 , (3.32)
τ3

N2 N3
=− − Wa ·N1 + We ·N2 , (3.33)
τ2 τ3
N2
τ2
− We ·N2
N1 = . (3.34)
−(Wa + Wp )

Replacing (3.32) in (3.33):

N2
= −Wp ·N1 − Wa ·N1 + We ·N2 . (3.35)
τ2

The level 3 population is negligible (N3 ≈ 0), due to the spontaneous lifetime from
level 3 being much shorter than that of level 2 (τ3  τ2 ). So, the total population is:

NT ≈ N1 + N2 . (3.36)

Replacing (3.36) in (3.35):

N2
= −Wp (NT − N2 ) − Wa (NT − N2 ) + We N2 , (3.37a)
τ2
Wp + Wa
N2 = NT · 1 . (3.37b)
We + Wp + Wa + τ2
CHAPTER 3. FIBER OPTICAL ACCELEROMETER 79

Replacing (3.36) in (3.34):

(NT − N1 )( τ12 + We )
N1 = , (3.38a)
Wa + Wp
1
τ2
+ We
N1 = NT · 1 . (3.38b)
We + Wp + Wa + τ2

The gain is defined as (DESURVIRE et al., 1990; BALL; GLENN, 1992):

α = σe ·N2 − σa ·N1 , (3.39)

where σe and σa are the emission and absorption cross sections from the signal transition.
The transition rates are given by Wp = σp ·Ip /(h·νp ), We = σe ·I/(h·ν) and Wa =
σa ·I/(h·ν), where Ip is the pump intensity, σp is the pump absorption cross section, νp is
the pump frequency, I is the signal intensity. Therefore, replacing (3.37b) and (3.38b) in
(3.39), we have:

σa ·I 1 σe ·I
Wp + +
h·ν τ2 h·ν
α = σe · NT · − σa · NT · , (3.40a)
σe ·I σa ·I 1 σe ·I σa ·I 1
+ Wp + + + Wp + +
h·ν h·ν τ2 h·ν h·ν τ2
σe ·τ2 ·Wp − σa
1 + τ2 ·Wp
α = NT · σa + σe . (3.40b)
1+I ·
h·ν( τ12 + Wp )

Rewriting (3.40b) in terms of low signal gain (α0 ) and signal saturation intensity (IS ):

α0
α= , (3.41)
I
1+
IS

σe ·τ2 ·Wp − σa
α0 = NT · , (3.42)
1 + Wp ·τ2

h·ν( τ12 + Wp )
IS = . (3.43)
σa + σe

In (3.43) it is possible to notice the linear dependence of the saturation intensity (IS )
with the pumping rate (Wp ), which can be verified in the Results and discussion chapter
(Chapter 4).
CHAPTER 3. FIBER OPTICAL ACCELEROMETER 80

From the gain equation (3.41), one is able to find an amplifier general equation for a
steady state operation (RIGROD, 1965; SVELTO, 2010; RODRIGUES, 2019). For an amplifier
with longitudinal length in the z-axis, the amplified intensity can be described by:

1 dI α0
· = . (3.44)
I dz I
1+
IS

Solving (3.44) results in an expression providing the final intensity, I, from an initial
intensity, I0 , amplified after a length L along the z-axis

I − I0
 
I
ln + = α0 ·L. (3.45)
I0 IS

Changing the variables in (3.45) to x and y:

x = I − I0 , (3.46a)
 
I
y = ln , (3.46b)
I0

equation (3.45) becomes:

−1
y= · x + α0 ·L. (3.47)
IS
Thus, with (3.47), one can characterize the amplifier medium, calculating the low signal
gain and the signal saturation intensity, by measuring the input and the output intensity
in a certain length of amplifier. The result is a line with inclination −1/IS and crossing
y-axis at the point α0 ·L.

3.3.2 Erbium-doped fiber characterization

To characterize the amplifier medium, a laser diode (974.4 nm) was used to pump
(called in this text as pump laser) the active medium and its optical power in function of
the electrical current is shown in Fig. 3.12. The first derivative plot corresponds to the
slope of optical power, meaning that the highest change in slope (i.e., second derivative
maximum point) is when the lasing action begins. For this pump laser the threshold
is 32 mA. The inset of Fig 3.12 is a zoomed-in view of the threshold region. The first
derivative plot is also helpful to identify nonlinearities (kinks) in the pump laser (HERT-
SENS, 1989; ACHTENHAGEN et al., 2006), as well as providing a quick read of the direct
modulation value of the pump laser (inclination in mW/mA).
The active medium used in this experiment are optical fibers doped with ions of Er3+
(models M5 and M12 from Fibercore and ER30 from Liekki). When the active medium
CHAPTER 3. FIBER OPTICAL ACCELEROMETER 81

FIGURE 3.12 – Pump laser (laser diode) slope efficiency.

Source: Author.

is pumped by the pump laser a population inversion is achieved, meaning that there are
more ions in the excited level (level 2 in Fig. 3.11) than in the fundamental level (level
1 in Fig. 3.11). Once this condition is met, the medium becomes an amplifier of light.
Then, another laser diode with a wavelength near the amplifier transition, in this case
around 1550 nm (called as signal), is subjected to pass through the doped fiber making the
excited ions to decay with the same wavelength as the signal, i.e., by stimulated emission.
Thus, the active medium amplifies the input signal (DESURVIRE et al., 1990), working as
a light amplifier.
The setup used to measure the amplification characteristics of the different erbium
doped-fibers is represented in Fig. 3.13. The experiment consisted of varying the input
signal power and measuring the output.

FIGURE 3.13 – Setup for characterization of the erbium doped fiber amplifier.

Source: Author.
CHAPTER 3. FIBER OPTICAL ACCELEROMETER 82

3.3.3 Fiber Bragg grating inscription

After choosing what model of erbium-doped fiber to use, it is needed to make the
reflectors (i.e., the mirrors) of the laser cavity. In this work, the reflectors are fiber
Bragg gratings (FBGs) inscribed on optical fiber by using the transverse holographic
method with a phase mask based interferometer (MELTZ et al., 1989; DOCKNEY et al.,
1996; SAHOTA et al., 2020). The inscription in the optical fiber is made by a continuous
wave laser at 244 nm and 42 mW of power (model Innova 300c from Coherent), with
the online characterization setup usually employed at the Institute for Advanced Studies,
IEAv, (CAZO, 2001), as shown in Fig. 3.14.

FIGURE 3.14 – FBG on-line characterization setup used at IEAv. (a) setup to get reference signal and
(b) to measure FBG reflection.
(a) (b)
Reflection

Source: Author.

A superluminescent diode (SLD) is a broadband light source, its light goes through
the directional coupler and is reflected by the cleaved fiber at the end. In Figure 3.14a
is indicated 3.37% reflection, because the refractive index of the optical fiber is 1.45 and
the air is 1 and according to Fresnel equations (HECHT, 2017), the reflection for incidence
angle zero is calculated as 2
nfiber − nair

R= , (3.48)
nfiber + nair
resulting in the 3.37% reflection. Then, this reflected light is measured by the optical
spectrum analyzer (OSA), in dBm, and is used as reference (POSA1 ). By comparing the
3.37% reflection with the reflection from the FBG that is being inscribed in the optical
fiber (POSA2 ), Fig. 3.14b, one can calculate the FBG reflection by:

OSA2 −POSA1
P 
PFBG = 10 10
· R. (3.49)

Using this characterization method, the FBG spectrum and alignment of the system (nec-
essary to obtain a fast inscription) were adjusted using a photosensitive fiber (PS1250/1500
from Fibercore), providing a highly reflective grating as shown in Fig. 3.15.
The drawback in using FBGs from photosensitive fibers, that does not contain erbium
in its composition, is the requirement of having splices to join the active medium to the
CHAPTER 3. FIBER OPTICAL ACCELEROMETER 83

FIGURE 3.15 – Calibration setup of FBG on a single-mode photosensitive fiber. (a) contains the signals
captured by the OSA and (b) is the calculated reflectivity spectrum.
(a) (b)

Source: Author.

reflectors. The splices inside the resonant cavity introduce losses to the light round-trip
and the cavity length must be larger because the splicing machine needs a minimum
length of fiber to work. A solution to this problem is to inscribe FBGs directly on the
erbium-doped fiber, without splices in the cavity.
Then, a transmission interrogative system was assembled to analyze the erbium-doped
fibers (Fig. 3.16). Since the optical fiber in this experiment is not photosensitive anymore,
the FBG fabrication was going to be slower, the final reflection possibly lower, and the
erbium absorption could reduce the power from the fiber-air reflection used as reference
beyond the OSA sensitivity limit. Thus, a transmission interrogation was used to increase
the optical power reaching the OSA. The broadband light used as reference was a back-
ward spontaneous emission from 4.8 m of pumped erbium-doped fiber, model M12 from
Fibercore. Its spectrum is represented in Fig. 3.17 and its optical power is noticeably
higher than those in Fig. 3.15 (more than 40 dB). If the FBG is correctly inscribed in the
fiber, then a local valley would appear in the reference spectrum.

FIGURE 3.16 – FBG detection setup by transmission.

Source: Author.

This section showed the techniques used to inscribe FBGs on the optical fiber and
CHAPTER 3. FIBER OPTICAL ACCELEROMETER 84

FIGURE 3.17 – Reference fluorescence spectrum by transmission.

Source: Author.

how to measure their reflectivities. The FBG reflectors are necessary as the laser cavity
mirrors for an erbium-doped optical fiber laser, they need to be highly reflective and have
a narrow spectral bandwidth to select the longitudinal modes of the laser.

3.3.4 Longitudinal mode selection

Since erbium has a wide spontaneous emission spectrum (Fig. 3.18), the FBG band-
width is the one limiting the number of longitudinal modes in the cavity.

FIGURE 3.18 – FBG reflection spectrum and erbium fluorescence spectrum.

Source: Author.

So, to have a single longitudinal mode laser it is necessary to have narrow FBGs
bandwidth and a longer distance among longitudinal modes. In this way, only one mode
CHAPTER 3. FIBER OPTICAL ACCELEROMETER 85

oscillates in the cavity. The distance among modes in frequency is (SVELTO, 2010)

c
∆ν = , (3.50)
2 n·Lc

where n is the medium refractive index, Lc is the cavity length and c is the speed of light.
Derivating ν = c/λ in relation to λ, we get the conversion factor between wavelength
bandwidth and frequency bandwidth dν = −c·dλ/λ2 . The negative sign indicates they
are inversely related, but makes no physical sense to the resultant bandwidth value. Then,
the distance among modes in meters is:

λ2
∆λ = , (3.51)
2 n·Lc

where λ is the central wavelength. Finally, making the distance among modes equal or
greater than the FBG bandwidth, only one longitudinal mode prevails (MILES et al., 1980;
ZYSKIND et al., 1992; KERSEY et al., 1994; AMES; MAGUIRE, 2007).

A simulation of the number of modes inside the real spectrum of FBGs is presented
in Fig. 3.19, with a cavity of 5 mm and another of 2 mm. To splice a cavity of 2 mm of
length is not practical with the present laboratory equipment.

FIGURE 3.19 – Number of longitudinal modes varying with distance between FBGs. (a) 5 mm of optical
fiber length with 3 modes and (b) 2 mm of optical fiber length with 1 mode.
(a) (b)

Source: Author.

With (3.51) is possible to plan the laser cavity length based on the FBG spectral
bandwidth, since they are inversely related. Therefore, to select just one longitudinal
mode in the fiber laser it is necessary to obey this condition of the FBG bandwidth being
equal or narrower than the longitudinal modes distance. Concluding that, the broader is
the FBG the shorter must be the laser cavity length. Then, to facilitate its assembly and
increase the amplification power by the active medium, the FBG should be inscribed to
CHAPTER 3. FIBER OPTICAL ACCELEROMETER 86

be the narrowest possible. One way to achieve that is to inscribe a large length FBG,
around 10 mm, which results in a narrower spectrum.
4 Results and discussions

4.1 Nonlinear control

This first experiment was performed to verify the proper working of the nonlinear
control to stabilize the optical fiber interferometer. The experimental setup used for
achieving this result was shown in Fig. 3.8.
The control actuation result is shown in Fig. 4.1. From 0 s to 0.9 s the interferometer
is running in an open-loop, from 0.9 s to 1.4 s a ramp is applied to the feedback phase
modulator to provide the maximum and minimum values of the interferometer (VP D (t)
from Fig. 3.8). Then, myRIO dynamically changes the FPGA logic to the control opera-
tion that starts in 2.4 s. During all the time it was applied a sinusoidal signal in FOM1
(which is the FOM under evaluation).

FIGURE 4.1 – Nonlinear control actuating in the optical interferomter.

Source: Author.

The result of Fig. 4.1 shows the correct behavior of the interferometer with the nonlin-
ear control, with the interest signal being directly recovered after 2.5 s. In this case, the
interferometer is operating in the phase quadrature point (π/2) and the nonlinear control
is suppressing the signal fading.
CHAPTER 4. RESULTS AND DISCUSSIONS 88

Some applications require that the interferometer is controlled in an operation point

other than π/2 (ANDRADE, 2019). So, to show the stability and the correct operation
of the control system, the setpoint of the control loop was changed to arbitrary values.
For this experiment it was used the setup in Fig. 3.10, with control signal in FOM4 and
input signal in FOM5. The maximum and minimum values of the interferometric signal
were measured by Fig. 4.2b. The chosen setpoints are shown by the horizontal lines in the
model of the interferometric curve in Fig. 4.2a. Then, the simulations of the interferometer
working with a sinusoidal input signal in these different setpoints are presented in Fig. 4.2c
and the real experimental data with a sinusoidal input is in Fig. 4.2d.

FIGURE 4.2 – Different setpoint values in the control loop. (a) is the model of the interferometric
curve with real AV factor value and lines representing the different set-points. (b) is the open-loop
interferometer signal to acquire AV factor. (c) is the simulation for the different setpoints chosen,
with a signal amplitude of 0.3 rad at 500 Hz. (d) is the experimental result using the Mach-Zehnder
interferometer with an input signal amplitude of 100 mV at 500 Hz.
(a) (b)

(c) (d)

Source: Author.

The results in Fig. 4.2 shows that the nonlinear control system is able to keep the
interferometer in an arbitrary point of operation. In this case, note how the recovered
signals that are not in the quadrature phase point of operation (blue continuous line)
are distorted, with the extreme case being the purple dash-dotted line. In practice, the
interferometer when controlled in the extreme points (near 0 rad and near π rad) becomes
CHAPTER 4. RESULTS AND DISCUSSIONS 89

less robust. During experiments, the same feedback gain was used in the control loop, it
was noticed that these extreme points of operation kept losing control for small periods
of time followed by finding the operation point again, while the operation points near π/2
maintained stable during the experimental routine.
To show the chattering effect from the sliding mode control, the Mach-Zehnder inter-
ferometer was controlled without an input signal with different values of γ. The specific
gain that changes in value during the experiments is the programmable gain inside the
myRIO platform, γv [V/V] (from Section 2.2.3), the total feedback gain comprises the
digital variable γv that works as the sign function gain, the phase modulator scale factor
and the linear voltage amplifier gain. The composition of all these gains is the variable
that comprises the γ variable in the control condition (γ > |φ˙0 |). The next set of data were
collected on different days. Because of that, the A·V factor might be different resulting
in different DC voltage values.
Examples of the interferometer without input signal are shown in Fig. 4.3 with γv = 1,
Fig. 4.4 with γv = 7, and Fig. 4.5 with γv = 30. The chattering effect increases together
with the loop gain.

FIGURE 4.3 – Controlled interferometer (γv = 1) without an input signal. (a) is the interferometer
output signal at π/2, (b) is its frequency spectrum showing a negligible peak at 65 Hz, (c) is the control
signal before the voltage amplifier and (d) is the frequency spectrum of the control signal.
(a) (b)

(c) (d)

Source: Author.
CHAPTER 4. RESULTS AND DISCUSSIONS 90

FIGURE 4.4 – Chattering example with controlled interferometer (γv = 7) without an input signal. (a)
is the interferometer output signal at π/2, (b) is its frequency spectrum showing a peak at 80 Hz, (c) is
the control signal before the voltage amplifier and (d) is the frequency spectrum of the control signal.
(a) (b)

(c) (d)

Source: Author.

FIGURE 4.5 – Chattering example with controlled interferometer (γv = 30) without an input signal. (a)
is the interferometer output signal at π/2, (b) is its frequency spectrum showing peaks at 100 Hz and
300 Hz, (c) is the control signal before the voltage amplifier and (d) is the frequency spectrum of the
control signal.
(a) (b)

(c) (d)

Source: Author.
CHAPTER 4. RESULTS AND DISCUSSIONS 91

It can be observed that the chattering can be a detrimental effect in the output signal.
The controlled interferometer without input signal should present an output signal like in
Fig. 4.3, without any prominent spurious frequency added by the control chattering. When
γv was increased 7 times, the interferometric signal started to show a 80 Hz oscillation
due to chattering (Fig. 4.4), and when increased γv 30 times this frequency increased in
amplitude and shifted to 100 Hz with additional harmonics (Fig. 4.5).
To show how the controlled interferometer behaves when the input signal has a fre-
quency below 20 Hz, a signal was applied to FOM5 (Fig. 3.10) to simulate an spurious
signal that should be controlled (φ0 from Section 2.2). The Mach-Zehnder interferometer
with a 6 Hz sine wave of 250 mV amplitude applied to FOM5 is shown in Fig. 4.6 with
γv = 1, and then, with γv = 10 in Fig. 4.7.

FIGURE 4.6 – Interferometer (γv = 1) with an sinusoidal input signal. (a) is the interferometer output
signal, (b) is its frequency spectrum showing peaks at 6 Hz and 10 Hz, (c) is the control signal before the
voltage amplifier, (d) is the frequency spectrum of the control signal, (e) is the input signal with 250 mV
amplitude and (f) is its frequency spectrum with 6 Hz peak.
(a) (b)

(c) (d)

(e) (f)

Source: Author.
CHAPTER 4. RESULTS AND DISCUSSIONS 92

FIGURE 4.7 – Chattering example with controlled interferometer (γv = 10) with an sinusoidal input
signal. (a) is the interferometer output signal, (b) is its frequency spectrum showing peaks near 90 Hz,
(c) is the control signal before the voltage amplifier, (d) is the frequency spectrum of the control signal
with 6 Hz peak, (e) is the input signal with 250 mV amplitude and (f) is its frequency spectrum with
6 Hz peak.
(a) (b)

(c) (d)

(e) (f)

Source: Author.

It can be observed that for this disturbance signal (250 mV at 6 Hz) the control was
not able to cancel this low-frequency variation with γv = 1, the output interferometric
signal still shows the 6 Hz disturbance signal (Fig 4.6). In the second case with γv = 10
(Fig 4.7), the interferometer presents a triangular wave signal around the quadrature
point and does not reach the same amplitude as in Fig 4.6, it can be considered stable.
Note that the control signal in Fig 4.7 is exactly the input signal used to generate the
disturbance with an extra noise and inverted. Therefore, this is an example of high-gain
operation where the signal of interest is recovered from the control signal.
CHAPTER 4. RESULTS AND DISCUSSIONS 93

4.2 Michelson interferometer results

To validate the interrogation technique of low modulation index signal, the Michelson
interferometer was used to characterize some phase modulators. First, the popular cylin-
der of a piezoelectric material modulator and then a metal-coated optical fiber modulator.

4.2.1 Piezoeletric phase modulator characterization

Using the setup from Fig. 3.8, the frequency response of the fiber optical modulator
FOM1 (Fig. 4.8) was measured by feeding it with a known electric signal and measuring
the controlled interferometric output signal, sweeping from 50 Hz to 500 Hz with 10 Hz
step and from 500 Hz to 5 kHz with 25 Hz step.

FIGURE 4.8 – Picture of FOM1.

Source: Author.

The scale factor of the actuator is represented in Fig. 4.9a and the phase between the
input and output signals, in Fig. 4.9b. The grey area represents the standard deviation
among the 10 measurements acquired and the black line is the mean value.

FIGURE 4.9 – Frequency response of FOM1, (a) scale factor and (b) phase.
(a) (b)

Source: NUNES et al. (2021).

The linearity of the FOM1 was acquired at 1 kHz and is shown in Fig. 4.10.
CHAPTER 4. RESULTS AND DISCUSSIONS 94

FIGURE 4.10 – Linearity from FOM1 at 1 KHz.

Output [rad]

Input amplitude [V]

Source: NUNES et al. (2021).

The FOM1 frequency response indicates that higher amplitudes are achieved for lower
frequencies and no additional resonances were found in the frequency interval studied.
According to UDD; SPILLMAN (2011), the resonance frequency for a piezoelectric cylinder
with 25.25 mm diameter should be around 40 kHz. The knowledge of the frequency
response is important when using any FOM, e.g., feedback element in a closed-loop control
system since the scale factor should be used to properly calculate the feedback signal
amplitude in radians. The linearity was measured at 1 kHz and FOM1 shows a linear
behavior in the studied interval.

4.2.2 Metal on fiber phase modulator characterization

The Michelson interferometer was also used to characterize a phase modulator based
on thermal effect, called FOM-MC (Fig. 4.11). The frequency response of the FOM-MC
is shown in Fig. 4.12 and linearity in Fig. 4.13.

FIGURE 4.11 – Picture of a 10 mm FOM-MC soldered to a circuit board with thermal paste over it.

Source: Author.

It can be noticed in Fig. 4.12a that the overall frequency response shape is similar
to FOM1’s. The FOM-MC linearity was evaluated at 30 Hz and 150 Hz, presenting a
linear response. For a thermal effect modulator, the maximum frequency reached in this
experiment, 200 Hz, can be considered a high frequency.
CHAPTER 4. RESULTS AND DISCUSSIONS 95

FIGURE 4.12 – Frequency response of FOM-MC, (a) scale factor and (b) phase.
(a) (b)

Source: NUNES et al. (2021).

FIGURE 4.13 – Linearity of FOM-MC at 30 Hz and 150 Hz.

Source: NUNES et al. (2021).

The results regarding the characterization of the FOM1 and FOM-MC show the correct
behavior of the interferometer when measuring phase variation. The physical quantity
variation that resulted in the phase shift measured by the interferometer was ` by FOM1
and n by FOM-MC. The FOM-MC showed a good capability as a fiber optical phase
modulator, its normalized scale factor (radians by volts applied by meters of optical fiber)
is higher than FOM1 normalized scale factor, due to the fact that FOM-MC is 10 mm
long and FOM1 has 10.17 m of optical fiber in the PZT cylinder.
CHAPTER 4. RESULTS AND DISCUSSIONS 96

4.2.3 Problems with the Michelson setup

Although the Michelson interferometer was able to be used to characterize the phase
modulators, it presents a few problems in its current setup. The output optical fiber
connector by physical contact (FC/PC) from the laser diode is connected to an adapter
FC/PC to FC/PC, which avoids splices whenever the laser is used in a new setup. How-
ever, if the connector is fully connected, the interferometric signal loses quality, probably
due to back reflections on the laser cavity. Then, the connector should be loosen connected
to avoid light feedback to the laser, and should still allow enough intensity to provide a
high signal-to-noise ratio output signal. The problem in doing that is, the interferometer
becomes sensitive to spurious mechanical disturbances since any mechanical vibration mis-
aligns the input laser beam (in the connector) and changes the optical intensity entering
the sensor (see Fig. 4.14c).
Another problem is the laser diode itself, which probably due to mode competition
shows some phase jumps. These phase jumps can be very frequent when the laser is
still warming up (Fig. 4.14a), or less frequent and unpredictable after being on for a
few minutes (Fig. 4.14b). This can lead to wrong measurements interpretations in the
final sensor. So, the measurements had to be collected when these phase jumps were not
present.

FIGURE 4.14 – Problems with fiber Michelson interferometer. (a) Interferometer signal a few seconds
after laser is on, (b) after a few minutes and (c) presenting the sensitivity to mechanical taps on the
optical table near the input connector.
(a) (b) (c)

Source: Author.

4.3 Mach-Zehnder interferometer results

Since the polarization fading of this setup was not addressed by using multiple polar-
ization controllers on the output (KIRKENDALL; DANDRIDGE, 2004), the visibility stability
of the Mach-Zehnder interferometer was measured and analyzed. A low-frequency signal
with high amplitude was applied to a phase modulator to achieve the multi fringe regime,
CHAPTER 4. RESULTS AND DISCUSSIONS 97

allowing to measure the minimum and maximum values of the interferometric signal. The
first part in Fig. 4.15 shows the visibility when intentionally squeezing a coil of optical
fiber in one of the arms. The middle part, between the traced vertical lines, contains the
measurements without any induced variations to the interferometer. The blue horizontal
line is the mean value of visibility of 0.8986, with a standard deviation of 0.0060, in a time
interval of 20 minutes. The final part shows the maximum degradation of the visibility
achieved by simply squeezing the coil of optical fiber.

FIGURE 4.15 – Visibility of the Mach-Zehnder interferometer over time.

Source: Author.

Therefore, even without the polarization controllers, the visibility of the Mach-Zehnder
is very stable with low variation, as long as the optical fibers are kept in the same position.

4.4 Erbium-doped fiber laser results

To use the interrogation method in which the optical phase changes due to change of
the optical wavelength (Section 2.1.3.3), it was needed a longitudinal single-mode laser.
Next, the experimental results related to this quest for the single-mode laser, including
studies and procedures are presented.

4.4.1 Optical fiber amplifier

The characterization of the available optical fibers doped with erbium was made with
the setup from Fig. 3.13 by varying the input signal and measuring the amplified output
signal for different pumping optical power. Figure 4.16 shows these results in the right
column (Fig. 4.16b, Fig. 4.16d, and Fig. 4.16f) and the linear fit result from (3.47) in the
left column (Fig. 4.16a, Fig. 4.16c, and Fig. 4.16e) .
The values obtained for the small signal gain (α0 ) and the intensity of saturation (IS )
CHAPTER 4. RESULTS AND DISCUSSIONS 98

for each pump current are presented in Table. 4.1. A comparison between the experimental
values from Table. 4.1 and a simulation of the optical amplifier variables (α0 and IS ) due
to pumping rate (Wp ) is shown in Fig. 4.17, that is, arbitrary values are used in (3.42)
and (3.43) to verify the shape of the simulation data (Fig. 4.17e, and Fig. 4.17i) with the
experimental data (Fig. 4.17f, Fig. 4.17g, Fig. 4.17h, Fig. 4.17j, Fig. 4.17k, and Fig. 4.17l).
Also, the amplifier gain (G equals to output power divided by input power) simulation is
used to compare the curve shape when changing the input power (Fig. 4.17a) with the
experimental amplifier gains (Fig. 4.17b, Fig. 4.17c, and Fig. 4.17d). The fitted values
and the analytical simulation have a good match. The small-signal gain from M12 and
M5 (Fig. 4.17k and Fig. 4.17l) seems to be missing resolution, probably because they
have lower gains and more precise measurement is needed, i.e., an experiment with lower
values of input signal and lower values of output signal should be accomplished to better
characterize the lower power from Fig. 4.16d and Fig. 4.16f. However, this was not
detrimental on the determination of α0 and IS .

TABLE 4.1 – Linear fitting results from the erbium doped fibers characterization.

ER30 M12 M5
Pump [mA] −1 8 −1 8 −1
IS × 108
W  W  W 
α0 [m ] IS × 10 m2 α0 [m ] IS × 10 m2 α0 [m ] m2
100 4.08 0.77 2.74 0.69 0.64 1.06
200 4.35 2.30 2.64 2.25 0.66 2.94
300 4.44 3.85 2.64 3.74 0.66 4.67
400 4.47 5.39 2.67 4.66 0.64 6.92
500 4.48 6.97 2.68 5.92 0.67 7.63
600 4.50 8.56 2.69 7.07 0.68 8.36
650 4.52 9.14 2.65 7.79 0.68 8.21
CHAPTER 4. RESULTS AND DISCUSSIONS 99

FIGURE 4.16 – Characterization of the erbium doped optical fibers. The left column [(a), (c) and (e)]
is the linear fitting to find the small-signal gain and the saturation intensity, based on the line equation
from (3.47). The right column [(b), (d) and (f)] is the unprocessed collected data of input by output. All
the colors in these subfigures follow the same legend of subfigure (a).
(a) ER30 (b) ER30 98 cm

(c) M12 (d) M12 89 cm

(e) M5 (f) M5 86 cm

Source: Author.
CHAPTER 4. RESULTS AND DISCUSSIONS 100

FIGURE 4.17 – Comparison between optical amplifier model and experimental data from (3.41), (3.42)
and (3.43). The first row shows the variation of the amplifier net gain by input signal, the legends from
(c) and (d) are the same as in (b). The second row shows the linear dependence of the saturation intensity
to the pump. The third row shows the variation of the small-signal gain to the pump power.
(a) Model (b) ER30 (c) M12 (d) M5

(e) Model (f) ER30 (g) M12 (h) M5

(i) Model (j) ER30 (k) M12 (l) M5

Source: Author.
CHAPTER 4. RESULTS AND DISCUSSIONS 101

4.4.2 Fiber Bragg gratings

To construct the laser cavity mirrors, FBGs were inscribed on a photosensitive optical
fiber (Fibercore PS1250/1500). The reflectivities of the FBGs were measured and are
shown in Fig. 4.18. The FBGs were inscribed in different days with different setup config-
uration, one set with peaks around 1550 nm and other around 1559 nm. The laser cavity
can be then mounted by splicing the active medium (the erbium-doped fiber) between
two mirrors (the FBGs). These splices may prevent laser action since they may introduce
losses and the active medium presents low amplifier gain, because it should have a short
length to work as a single longitudinal mode.

FIGURE 4.18 – FBGs inscribed on standard photosensitive fiber, (a) with peaks around 1550 nm and
(b) around 1559 nm.
(a) (b)

Source: Author.

A solution to that would be to inscribe the FBGs directly on the erbium-doped fibers.
Thus, the two fibers with the highest gains, M12 and ER30, were submitted to the test.
The fiber M12 indicated a reflection valley of 0.6 dB after 40 minutes of exposure, de-
creasing to 0.3 dB after 1 hour and 30 minutes. The fiber ER30 did not indicate any
reflection after 1 hour of exposure. As a result, no significant reflection was obtained with
any of the erbium-doped fibers.
After that, a photosensitive erbium-doped optical fiber was purchased (Coherent/Nufern
PS-ESF-3/125) which was used by other research teams (SONG et al., 2014; MILLER, 2018).
But, due to the pandemic (COVID-19) and the period without activity in the laborato-
ries, the inscription laser is not in proper operation condition and is now with 10 mW of
optical power. This power is not suitable to inscribe the FBGs on the fiber and this was
not accomplished.
CHAPTER 4. RESULTS AND DISCUSSIONS 102

4.4.3 Erbium-doped fiber laser characterization

An erbium-doped fiber laser was then constructed based on the literature (ZYSKIND et
al., 1992; BALL; GLENN, 1992; DYNDGAARD, 2001; ZHANG et al., 2009; GUAN et al., 2011;
BERNHARDI, 2012; YU et al., 2014; LIU et al., 2019) with the FBG1 and FBG2 shown in
Fig. 4.18, and active medium length of 12 cm of the optical fiber ER30. This experiment
intended to splice the smallest fiber laser possible with the current equipment. The laser
cavity length is larger than the active medium itself because each FBG contains an extra
length of undoped optical fiber to allow multiple splicing fusions1 .
After verifying the laser action with 12 cm of active medium, the fiber laser was
disassembled and the active medium was reduced to 7 cm of ER30. Figure 4.19 shows
how much the laser threshold varies, considering different lengths of active medium and
slightly different fusion splices losses, the threshold was found using the method of the
derivative mentioned in Fig. 3.12. The threshold current increased almost 5 times when
the 12 cm of doped fiber was reduced to 7 cm.

FIGURE 4.19 – Lasers optical power in function of pump current for (a) 12 cm and (b) 7 cm of active
medium ER30.
(a) (b)

Source: Author.

These lasers are not single-mode yet, because the space between the FBGs is still
large. There are a few limitations in making lasers with fusion splicing with the current
equipment, the distance between the FBGs is limited by the physical length needed by
the splicing machine to work properly and the splice loss that is introduced in the cavity
may prevent the laser action.
The minimal cavity length achieved in this work of 7 cm is not nearly close to the
1
To fusion splice two optical fibers with minimum transmission loss, it is needed to cleave each fiber
to have a perfectly flat endface. Approximately 1 cm of fiber is lost in every cleavage with the current
equipment. So, the spare length must be considered when assembling experimental setups with optical
fiber.
CHAPTER 4. RESULTS AND DISCUSSIONS 103

desired 2 mm cavity length. Therefore, the assembling of a single longitudinal mode

erbium-doped fiber laser was not concluded in this work. Despite of that, a multimode
erbium-doped fiber laser will be used in order to characterize the laser as part of the
sensing element, to obtain the conversion factor between laser cavity strain and wavelength
necessary to use (2.27).

4.4.4 Erbium-doped fiber laser as a sensing element

To characterize the change of the laser wavelength due to a strain in the fiber laser
cavity, a laser was made with 20 cm of ER30. The FBG3 and FBG6 were chosen. With
this new pair of gratings FBG3 needed always to be stretched to match the wavelength
peaks and to obtain laser action. Figure 4.20 shows a rough experiment on 3 cases, with
FBG3 at rest (0 N), stretched enough to match peaks (0.20 N), and too stretched (0.45 N).
The FBG3 is the up side curve because it is the last reflector from Fig. 3.10). FBG6 is
the down side curve, is the FBG just before the laser output to the wavelength-division
multiplexing (WDM).

FIGURE 4.20 – Laser cavity pumped at a constant power (200 mA) with (a) FBG3 at rest position, (b)
stretched with 0.20 N and (c) stretched with 0.45 N.
(a) (b) (c)

Source: Author.

The characterization involved of the erbium-doped fiber laser as a sensing element was
accomplished by gluing (with cyanoacrylate ester) the laser cavity to aluminum supports
that were attached to a micrometer and a dynamometer. Thus, it was exerted tension
in the optical fiber laser by moving the micrometer and measuring the force with the
dynamometer. Two configurations were tested, one stretching just one of the FBGs in
the cavity and other stretching just the active medium (ER30). The experiment collected
simultaneously the position (micrometer), force (dynamometer), and the laser optical
spectrum.
CHAPTER 4. RESULTS AND DISCUSSIONS 104

The FBG spectrum shifts under strain and changes the output laser wavelength. In
order to evaluate that, it was used an optical fiber with initial length of 11.5 cm, the
relative position of the micrometer was varied from 0 µm to 73 µm and the corresponding
measured force was from 0.02 N to 0.40 N. Characterization of the fiber laser due to
strain in the FBG is presented in Fig. 4.21, showing a linear laser wavelength variation as
function of stretching the FBG, as

dλ nm m
= 4.3427 × 10−3 = 4.343 × 10−6 . (4.1)
d` µm m

Figure 4.21b shows the corresponding optical power variation as function of relative
position. In the range evaluated, although the wavelength response presented a linear
behavior, the output power presented a nonlinear one and, thus, this should be taken in
account on the interferometer since it affects the A·V factor. A solution could be restrict
the position range to a flat region or to use the Mach-Zehnder with two outputs and to
dismiss the laser intensity with (2.17). One can observe in Fig. 4.21a that even not being
a single longitudinal mode laser and having mode hop zones, it is still possible to linearly
tune the laser wavelength (HAPPACH et al., 2020).

FIGURE 4.21 – Straining FBG in a fiber laser to analyze (a) wavelength shift and (b) power shift.
(a) (b)

Displacement
Displacement

Source: Author.

Figure 4.22 is the plot of strain versus the force of the optical fiber containing the FBG,
model PS1250/1500 from Fibercore. Replacing the mechanical constant F/ε of the fiber
(provided by the linear fit) and cladding diameter (125 µm) into (3.3) gives the Young’s
modulus for this fiber of 52.68 GPa. The Young’s modulus measured for this fiber model
is 69.05 GPa (ANTUNES et al., 2008), whose value was measured by applying force until its
rupture (4.68 N). The fiber used in this work has an FBG written on it, has been through
hydrogenation (to increase the photosensitivity) and heat treatment afterward and thus,
it is considered that it was obtained a good agreement between those values.
The characterization of the fiber laser due to strain in the active medium, i.e., in
CHAPTER 4. RESULTS AND DISCUSSIONS 105

FIGURE 4.22 – Force versus strain curve to calculate the Young’s modulus of photosensitive optical fiber
PS1250/1500.

Source: Author.

the erbium-doped fiber (without straining the FBGs) is shown in Fig. 4.23. The initial
length of ER30 fiber before strain was also 11.5 cm, the micrometer was varied from 0 µm
to 827 µm and the corresponding measured force was from 0.02 N to 6 N. It does not
indicate any variation in the peak wavelength of the laser. The force by optical power
plot (Fig. 4.23b) shows a variation of power with no trend, which leads to conclude that
the variation in Fig. 4.21b can be due to instrumentation error or laser mode-hopping.

FIGURE 4.23 – Straining erbium-doped fiber in a fiber laser to analyze (a) wavelength shift and (b)
power shift.
(a) (b)

Source: Author.

The optical power from the fiber laser with 20 cm of the active medium was not
powerful enough for the photodetector to capture a strong signal after passing through
the interferometer. Therefore, the active medium (ER30) was increased to 2 m. This
laser spectra versus pumping current (from the pump laser) is in Fig. 4.24, without any
strain applied. It has a single wavelength of 1557.92 nm along all pumping power after
threshold and maximum optical power of 7.18 dBm.
CHAPTER 4. RESULTS AND DISCUSSIONS 106

FIGURE 4.24 – Laser spectrum of fiber laser made with 2 m of fiber ER30 when changing pump current
from 30 mA to 650 mA.

Source: Author.

Using the pump laser characterization from Fig. 3.12, we obtain the direct modulation
curve for the fiber laser (Fig. 4.25), resulting in an efficiency of 0.01713 mW/mW (or
0.009125 mW/mA).

FIGURE 4.25 – Fiber laser with 2 m of ER30 direct modulation.

Source: Author.

In order to measure the interferometer output due to the modulation of the laser
wavelength (Fig. 4.21a), the erbium-doped fiber laser with 3.35 mW of optical power
(20 cm of active medium) was used to feed the Mach-Zehnder interferometer. A sine wave
with 5 V amplitude and 5 Hz was inserted into FOM5, in order to generate a multi-fringe
interferometric response and calculate the A·V factor. Then, keeping the erbium-doped
fiber laser cavity fixed, the interference signal from FOM5 was measured by the Mach-
Zehnder working with the erbium-doped fiber laser (Fig. 4.26).
CHAPTER 4. RESULTS AND DISCUSSIONS 107

FIGURE 4.26 – Interferometer working with the fiber laser and an input signal of 5 Hz to reach maximum
and minimum values. (a) 50 s acquisition with a high amplitude signal of 5 Hz, (b) 5 s acquisition, and
(c) 1 s acquisition.
(a)

(b)

(c)

Source: Author.

In the first moment, Fig.4.26a is showing a data collection of a 50 s period, the offset
value was low and it indicates the photodetector was misaligned. In this acquisition,
however, is possible to notice the maximum and minimum value of the interferometric
signal changing, which means the A·V factor was changing. The visibility was calculated
according to (2.15) and is represented by the dashed line in the figures, the maximum
value observed was 0.28. In the second moment with the photodetector aligned the offset
value increased, Fig. 4.26b. In the third moment, Fig. 4.26c, an acquisition with 1 s period
allows the visualization of the interferometric response (similar to Fig. 2.3f). But even for
relatively fast acquisition, when compared with Fig.4.26a, shows the visibility changing.
With these results, it was possible to observe the variation in wavelength when strain-
ing the cavity mirror (FBG) from the erbium-doped fiber laser and obtain a conversion
factor from wavelength to displacement that can be used to the sensor estimate sensibil-
CHAPTER 4. RESULTS AND DISCUSSIONS 108

ity. However, the accelerometer sensing element with an inertial mass (Fig. 3.3) was not
assembled in the fiber laser yet, due to the changes in the interferometric visibility. The
visibility changing this way would affect the acceleration measurement, making a faulty
sensor. Therefore, the accelerometer final assembly will be limited, for the time being, to
use the laser diode and place the sensing element in the interferometer sensor arm.

4.5 Opto-mechanical accelerometer

The accelerometer was assembled with the support shown in Fig. 4.27 (FERREIRA,
2020). The sensor arm of the interferometer (shown in orange in Fig. 4.27) is one of the
four optical fibers that hold the inertial mass in place. The fixation points are made with a
cyanoacrylate ester adhesive (from Loctite) and a polypropylene adhesive tape (from 3M)
over top. The fibers were pre-strained with the assistance of a dynamometer (PCE-FM
50N).

FIGURE 4.27 – Accelerometer model.

Interferometer arm

Fixation points Inertial mass

4.5 cm 4.5 cm

Base
Optical fibers

Fixed supports

Source: Author.

To characterize the accelerometer manual impulse signals were applied on the fixed
supports, on the base, or directly on the inertial mass. These impulse-like signals were
created by quickly hitting a cotton swab, a metal stick, or a spring of a pen on the
accelerometer. During the experiments, it was noted that the accelerometer response to
impulse signals is influenced by it being attached to the optical breadboard or not, so,
this will be informed. Therefore, the input signals were not the same among all tests. The
goal was to excite the resonance frequency and observe the exponential signal decrease
due to damping, even without a standardized input signal.
Signals were collected for three different pre-strain forces of 0.7 N, 1 N, and 2 N (all
with 4.5 cm of length between the inertial mass and each of the fixed supports), i.e.,
CHAPTER 4. RESULTS AND DISCUSSIONS 109

FIGURE 4.28 – Picture of the accelerometer assembly, (a) top view and (b) front view with the Mach-
Zehnder interferometer in the back.
(a) (b)

Source: Author.

considering the force necessary to break the optical fiber is around 4 N (ANTUNES et al.,
2008), the pre-strain forces were kept below this value. To find the excited frequencies
it was used the fast Fourier transform algorithm. To remove the bias when analyzing
the frequency spectrum, a prominence algorithm from MATLAB was used to indicate
the frequency peaks with a minimum of 4% prominence. Then, an exponential curve was
fitted to the upper envelope of the accelerometer response, in order to obtain the damping
factor (Γ).
Thus, in this experiment, all the fibers were pre-strained with 0.7 N. Figure. 4.29 shows
the accelerometer response for an impulse input on the fixed support, with the base bolted
to the optical breadboard (Honeycomb from Newport).

FIGURE 4.29 – (a) Accelerometer (0.7 N, bolted to the optical breadboard) response to impulse from
metal stick at the right fixed support with exponential fit. (b) Frequency spectrum showing a peak at
494 Hz.
(a) (b)

Source: Author.
CHAPTER 4. RESULTS AND DISCUSSIONS 110

The accelerometer with 0.7 N pre-strain presented a damping factor Γ = 41.53 s−1
and resonance frequency f0 = 494 Hz (Fig. 4.29).
The accelerometer was reassembled with 2 N pre-strain force. While testing this new
configuration of sensor it was noticed that the responses to impulse signals were damping
too fast (Fig. 4.30) in relation to the 0.7 N setup, which led to difficulties in acquiring a
response with an exponential decay form to characterize the accelerometer.

FIGURE 4.30 – (a) Accelerometer (2 N, bolted to the optical breadboard) response to impulse from metal
stick at the right fixed support with exponential fit. (b) Frequency spectrum showing a peak at 5260 Hz.
(a) (b)

Source: Author.

Figure 4.30 shows the accelerometer (bolted to the optical breadboard and 2 N pre-
strain) response to an impulse applied with a metal stick at the fixed support, fitting
Γ = 474.63 s−1 and f0 = 5260 Hz. In spite of that, the accelerometer worked properly
responding to weak input signals, even a handclap and was able to measure acceleration
generated by smart phone vibrating on its base.
From the hypothesis that a small pre-strain provides a low damping factor and a large
pre-strain provides a high damping factor, the accelerometer was reassembled with 1 N
of pre-strain force. Then, impulses were applied to the accelerometer and to the optical
breadboard. When attached to the optical breadboard and when just supported on it
(indicated in the figure captions as attached or not attached).
Figure 4.31, Fig 4.32 and Fig 4.33 are signals with the accelerometer bolted to the
optical breadboard. Figure 4.34, Fig 4.35 and Fig 4.36 are signals with the accelerometer
not bolted to the optical breadboard. It is possible to notice that all these signals have the
same approximate value of damping factor, with a mean value of Γ = 60.5 s−1 ). Which
seems to corroborate with the hypothesis, Γ(0.7 N) < Γ(1 N) < Γ(2 N) . Yet, the data set is
not large enough to confirm this.
CHAPTER 4. RESULTS AND DISCUSSIONS 111

FIGURE 4.31 – (a) Accelerometer (1 N, attached) response to impulse from cotton swab at the right
fixed support with exponential fit. (b) Frequency spectrum with peaks at 380 Hz and 740 Hz.

(a) (b)

Source: Author.

FIGURE 4.32 – (a) Accelerometer (1 N, attached) response to impulse from cotton swab at the left fixed
support with exponential fit. (b) Frequency spectrum with peaks at 370 Hz and 740 Hz.

(a) (b)

Source: Author.

FIGURE 4.33 – (a) Accelerometer (1 N, attached) response to impulse from metal stick at the breadboard
on the vertical direction with exponential fit. (b) Frequency spectrum with peaks at 410 Hz and 740 Hz.

(a) (b)

Source: Author.
CHAPTER 4. RESULTS AND DISCUSSIONS 112

FIGURE 4.34 – (a) Accelerometer (1 N, not attached) response to impulse from cotton swab at the right
fixed support with exponential fit. (b) Frequency spectrum with peaks at 520 Hz and 765 Hz.

(a) (b)

Source: Author.

FIGURE 4.35 – (a) Accelerometer (1 N, not attached) response to impulse from cotton swab at the left
fixed support with exponential fit. (b) Frequency spectrum with peak at 520 Hz.

(a) (b)

Source: Author.

FIGURE 4.36 – (a) Accelerometer (1 N, not attached) response to impulse from metal stick at the
breadboard with exponential fit. (b) Frequency spectrum with peaks at 470 Hz and 745 Hz.

(a) (b)

Source: Author.
CHAPTER 4. RESULTS AND DISCUSSIONS 113

The responses to impulse signals to the accelerometer with 1 N pre-strain and bolted
to the optical breadboard showed: Γ = 62.73 s−1 with two most prominent frequency
peaks of 380 Hz and 740 Hz, for an impulse applied with a cotton swab at the right fixed
support (Fig. 4.31); Γ = 51.05 s−1 with two most prominent frequency peaks of 370 Hz
and 740 Hz, for an impulse applied with a cotton swab at the left fixed support (Fig. 4.32);
Γ = 64.83 s−1 with two most prominent frequency peaks of 410 Hz and 740 Hz, for an
impulse applied with a metal stick at the optical breadboard (Fig. 4.33).
The responses to impulse signals to the accelerometer with 1 N pre-strain and not
bolted to the optical breadboard showed: Γ = 54.56 s−1 with two most prominent fre-
quency peaks of 520 Hz and 765 Hz, for an impulse applied with a cotton swab at the
right fixed support (Fig. 4.34); Γ = 75.18 s−1 with two most prominent frequency peaks
of 520 Hz and 765 Hz, for an impulse applied with a cotton swab at the left fixed support
(Fig. 4.35); Γ = 54.67 s−1 with two most prominent frequency peaks of 470 Hz and 745 Hz,
for an impulse applied with a metal stick at the optical breadboard (Fig. 4.36).
When the base of the accelerometer is bolted to the optical breadboard there is 740 Hz
peak frequency together with a peak around 380 Hz. When not bolted, the 740 Hz am-
plitude decreases, and the lower frequency shifts to around 520 Hz and increase its ampli-
tude. From (3.10), the predicted natural frequency for the proposed spring-mass system
is f0 = 306.2 Hz. This value can be used as an estimation for the accelerometer’s natural
frequency since it does not consider the damping effect and neither the other compo-
nents of the accelerometer (base, fixed supports, adhesives, etc.). Then, disregarding the
740 Hz frequency once the predicted frequency is 306.2 Hz, the mean frequency of the
measurements with the accelerometer bolted to the table is 386.6 Hz and 503.3 Hz when
free.
Next, there are a few signals of practical measurements with the accelerometer. Fig-
ure 4.37, Fig 4.38, Fig 4.39 and Fig 4.40 are measurements of the optical breadboard
dynamic behavior. The optical breadboard was punched strong enough to oscillate, and
thus oscillate the entire accelerometer body. A 17 Hz frequency was present in all mea-
surements but in Fig 4.38. Because of the small acquisition time in Fig 4.38 the low
frequency measured shifted to 15 Hz, on the other hand, made it possible to observe
a frequency peak at 460 Hz (frequency near the accelerometer impulses response). The
damping factor of the optical breadboard is consistent among the measurements (between
4.04 s−1 and 7.75 s−1 ), even for different setups (1 N and 2 N). Again, as we know, this
might be the damping factor of the entire group of equipment present on the laboratory
bench, not just the optical breadboard. This might also be the reason for difference in
frequencies when the accelerometer is bolted and not bolted to the breadboard, bolting
the accelerometer to the breadboard might alter the dynamic response because of better
energy transfer between the two.
CHAPTER 4. RESULTS AND DISCUSSIONS 114

FIGURE 4.37 – (a) Accelerometer (1 N, not bolted to the optical breadboard) response to impulse from a
manual impulse at the breadboard on the vertical direction with exponential fit. (b) Frequency spectrum
with peak at 17 Hz.

(a) (b)

Source: Author.

FIGURE 4.38 – (a) Beginning of the accelerometer (1 N, not bolted to the optical breadboard) response
to impulse from a manual impulse at the breadboard on the vertical direction with exponential fit. (b)
Frequency spectrum with peaks at 15 Hz and 460 Hz.

(a) (b)

Source: Author.
CHAPTER 4. RESULTS AND DISCUSSIONS 115

FIGURE 4.39 – (a) Accelerometer (2 N, bolted to the optical breadboard) response to impulse from a
manual impulse at the breadboard on the vertical direction with exponential fit. (b) Frequency spectrum
with peak at 17 Hz.

(a) (b)

Source: Author.

FIGURE 4.40 – (a) Accelerometer (1 N, bolted to the optical breadboard) response to impulse from a
manual impulse at the breadboard on the vertical direction with exponential fit. (b) Frequency spectrum
with peak at 17 Hz.

(a) (b)

Source: Author.

Figure 4.41 is the accelerometer response to a set of commercial computer speakers

standing on the table playing a pure tone signal of 300 Hz. Figure 4.42 is the response for
a handclap near the accelerometer and Fig. 4.43 is a smartphone vibrating on the base of
the accelerometer.
CHAPTER 4. RESULTS AND DISCUSSIONS 116

FIGURE 4.41 – (a) Accelerometer (0.7 N, bolted to the optical breadboard) response to an acoustic signal
of 300 Hz. (b) Frequency spectrum with peak at 300 Hz.

(a) (b)

Source: Author.

FIGURE 4.42 – (a) Accelerometer (2 N, bolted to the optical breadboard) response to a handclap near
the inertial mass. (b) Frequency spectrum.

(a) (b)

Source: Author.

FIGURE 4.43 – (a) Accelerometer (2 N, bolted to the optical breadboard) response to a smartphone
vibrating while standing at the base of the accelerometer. (b) Frequency spectrum with peaks at 40 Hz
and 200 Hz.

(a) (b)

Source: Author.
CHAPTER 4. RESULTS AND DISCUSSIONS 117

The accelerometer responded to manual impulse signals, to speaker signal, to a hand-

clap and a smartphone vibrating. With these results it is possible to confirm the proper
work accelerometer, even applying different pre-strains to hold the inertial mass in place.
The small amplitude frequency between 50 Hz and 100 Hz present in the accelerometer
signals is likely to be the controller chattering, as seen in Section 4.1.

4.5.1 Sensitivity estimation

The sensitivity of the accelerometer on the arms of a Mach-Zehnder interferometer can

be estimated by replacing the mean value measured resonance frequency of 503.3 Hz and
the damping factor of 60.5 s−1 on (3.25), resulting in a sensitivity of ∆φ
a
= 4.432 rad/g at
100 Hz.
Although the accelerometer was not assembled on the fiber laser cavity, it is possible
to use the current values to estimate what sensitivity it would have, considering the same
mechanical model and using (3.27). The Mach-Zehnder interferometer was constructed
to have a perfect match between the sensor and reference arms, to have the best visibility
value possible. But, this is difficult to obtain in the real world because one additional
cleavage means 1 cm less in one of the arms. So, let’s assume that the physical difference
between the interferometer arms is ` = 1 cm. Using the fiber laser wavelength by stretch
constant of 4.343·10−6 m/m, the sensitivity at 100 Hz is ∆φ a
= −166.423 rad/g.
The figure 4.44 is a representation of these sensitivities estimation over frequency
for each setup. The sensitivity for the fiber laser is represented in magnitude since the
negative value only means the opposite orientation of phase relative to acceleration.

FIGURE 4.44 – Sensitivity estimation over frequency.

X 100
Y 166.4

X 100
Y 4.433

Source: Author.
CHAPTER 4. RESULTS AND DISCUSSIONS 118

4.5.2 Minimum detectable acceleration and dynamic range

Analyzing the controlled Mach-Zehnder interferometer without any signal, the mini-
mum value of phase is 42.37 mrad2 . Which results in a minimum acceleration of 0.01 g
at 100 Hz.
The maximum phase is the limit to stay in the low modulation index regime, chosen to
have a maximum THD of 3%, equivalent to 0.83 rad. Therefore, the maximum acceleration
is of 0.187 g at 100 Hz. The dynamic range is 25 dB.

2
In some works, this value is acquired by an electric spectrum √
analyzer and the unit is presented in
the dimension of spectrum density, that for interferometers is rad/ Hz
5 Conclusion

A fiber optical accelerometer model based on inertial mass supported by optical fiber
was studied and constructed with three different pre-strain forces (0.7, 1, and 2 N). It was
interrogated by a Mach-Zehnder interferometer with nonlinear control. The estimated
sensitivity obtained was 4.432 rad/g at 100 Hz, with an estimated resonance frequency
around 503.3 Hz and damping factor of 60.5 s−1 . The dynamic range measured was from
0.01 g to 0.187 g, i.e., 25 dB. Therefore, this setup worked properly as an accelerometer
based on a Mach-Zehnder interferometer with nonlinear control, showing the capability to
work on space applications frequency range (between 20 Hz and 100 Hz). In addition, this
sensor achieved high performance with a high sensitivity, robustness, resonance frequency
above the region of interest, showing that most of the objectives of this work were reached
in this proof-of-concept. Once this was proved, an engineering design on the reduction of
the size and the weight of the setup can be accomplished in future work. This work also
provided a mathematical model and equations that allow one to design the accelerometer
parameters in order to meet a given application requirement.
Three different types of erbium-doped fiber were characterized and it was obtained a
working laser with 20 cm of active medium. The laser presented a linear relation between
the applied strain in one of the cavity FBG and the output laser wavelength, confirming
that it can be used as the sensing element of the accelerometer. With an imbalance of
just 1 cm in the interferometer, it is predicted an increase of 37 times in the accelerometer
sensitivity, when compared with the configuration comprising the sensing element in the
interferometer arm. The visibility variation issue creates the requirement to improve the
actual laser configuration, not necessarily to achieve a single longitudinal mode. These
results show the potential for this configuration to be used as an accelerometer, agreeing
with the literature.
The Michelson interferometer is polarization insensitive due to Faraday rotator mirrors,
but it reflects light to the laser cavity depending on the isolator, which can result in a
noisy signal. Whereas the Mach-Zehnder interferometer does have polarization fading
that must be addressed by using polarization controllers to achieve the best visibility,
it does not reflect light to the laser cavity. Moreover, the dual output availability in the
Mach-Zehnder allows the elimination of the laser intensity in the recovered interferometric
CHAPTER 5. CONCLUSION 120

signal. Then regarding the application as accelerometer it was concluded that the Mach-
Zehnder is the more suitable interferometer to work as the interrogator because of these
extra robustness characteristics related to signal demodulation.
Regarding the nonlinear control system, the state space equations were calculated
to control the interferometer in arbitrary points of operation, the phase plane for this
state space equations was calculated to assess the stability of the modified equilibrium
points. Then, the control logic was modified and this implementation was demonstrated
in practice. Also, for the Mach-Zehnder interferometer controlled at quadrature point
of operation, the control logic was modified to consider the two Mach-Zehnder outputs
and remove the offset voltage automatically (CHANDRA et al., 2016), similar to what
MARTIN (2018) did analogically, but with different mathematical operations.

In addition, during this work, it was used an algorithm in LabVIEW to automatically

collect the frequency response of optical phase modulators and guarantee the 3% THD
threshold for the low modulation index. Another algorithm was developed and used to
automatically collect optical power characterization of lasers, communicating via GPIB
with the laser diode controller and the OSA. These algorithms allow a fast data collection
of large data, and the researcher’s presence near the devices is not necessary, eliminating
this probable source of spurious signals. The noise study provided a direction to improve
the optical detection signal in the future. For the high optical power used in this work,
the limiting noise is the rin.
Part of the results obtained in this work was submitted to publication in Applied
Optics as a manuscript entitled “Characterization of a thin-film metal-coated fiber optical
phase modulator based on thermal effect with a nonlinear control interferometer”, which
was accepted for publication.

5.1 Future work

For future work, it is desired to inscribe FBGs directly on the erbium-doped fiber
to achieve the longitudinal single-mode laser. Having a single-mode laser interrogating
the accelerometer would increase the estimated sensitivity from 166 rad/g with 1 cm
imbalance to 1.331·106 rad/g with 80 m imbalance. Still regarding the fiber laser, it
would be interesting to study the reason behind the visibility instability and solve this
problem to make it work as a sensor even in a multimode regime.
Another future work may consider using another demodulation technique to increase
the dynamic range, or even consider using other nonlinear control system for interferom-
etry that has a greater dynamic range without demodulation (FELÃO et al., 2019). Also,
the analysis of passive interferometric signal recovery techniques with 3 by 3 coupler
CHAPTER 5. CONCLUSION 121

(DJINOVIC et al., 2018; DJINOVIC et al., 2020; SHI et al., 2021; WU et al., 2021) could be
performed, and then, compare it with the nonlinear active control method used here.
 References

ACHTENHAGEN, M.; HARDY, A. A.; HARDER, C. S. Coherent kinks in high-power

ridge waveguide laser diodes. Journal of lightwave technology, IEEE, v. 24, n. 5,
p. 2225, 2006. Available at: https://dx.doi.org/10.1109/JLT.2006.872313. Accessed on:
4 Aug. 2021.
ALEXANDER, S. B. Optical communication receiver design. Bellingham, Washington:
IET, 1997. ISBN 0819420239.
AMES, G. H.; MAGUIRE, J. M. Erbium fiber laser accelerometer. IEEE Sensors
Journal, IEEE, v. 7, n. 4, p. 557–561, 2007. Available at:
https://dx.doi.org/10.1109/JSEN.2007.891995. Accessed on: 4 Aug. 2021.
ANDRADE, A. W. F. Interferômetro óptico em malha fechada com duas polarizações
para aplicações espaciais. Dissertation (Master’s degree) — Aeronautics Institute of
Technology (ITA), São José dos Campos, 2019.
ANTUNES, P.; LIMA, H.; MONTEIRO, J.; ANDRÉ, P. Elastic constant measurement
for standard and photosensitive single mode optical fibres. Microwave and optical
technology letters, Wiley Online Library, v. 50, n. 9, p. 2467–2469, 2008. Available at:
https://dx.doi.org/10.1002/mop.23660. Accessed on: 4 Aug. 2021.
ANTUNES, P. F. da C.; LIMA, H. F.; ALBERTO, N. J.; RODRIGUES, H.; PINTO,
P. M.; PINTO, J. de L.; NOGUEIRA, R. N.; VARUM, H.; COSTA, A. G.; ANDRÉ,
P. S. de B. Optical fiber accelerometer system for structural dynamic monitoring. IEEE
Sensors Journal, IEEE, v. 9, n. 11, p. 1347–1354, 2009. Available at:
https://dx.doi.org/10.1109/JSEN.2009.2026548. Accessed on: 4 Aug. 2021.
BALL, G. A.; GLENN, W. Design of a single-mode linear-cavity erbium fiber laser
utilizing Bragg reflectors. Journal of Lightwave Technology, IEEE, v. 10, n. 10, p.
1338–1343, 1992. Available at: https://dx.doi.org/10.1109/50.166773. Accessed on: 4
Aug. 2021.
BASUMALLICK, N.; BISWAS, P.; CHAKRABORTY, R.; CHAKRABORTY, S.;
DASGUPTA, K.; BANDYOPADHYAY, S. Fibre Bragg grating based accelerometer
with extended bandwidth. Measurement Science and Technology, IOP Publishing,
v. 27, n. 3, p. 035008, 2016. Available at:
https://dx.doi.org/10.1088/0957-0233/27/3/035008. Accessed on: 4 Aug. 2021.
BELAL, M.; SONG, Z. qi; JUNG, Y.; BRAMBILLA, G.; NEWSON, T. An
interferometric current sensor based on optical fiber micro wires. Opt. Express, OSA,
REFERENCES 123

v. 18, n. 19, p. 19951–19956, 2010. Available at:

https://dx.doi.org/10.1364/OE.18.019951. Accessed on: 4 Aug. 2021.

BERKOFF, T.; KERSEY, A. Experimental demonstration of a fiber Bragg grating

accelerometer. IEEE Photonics Technology Letters, IEEE, v. 8, n. 12, p. 1677–1679,
1996. Available at: https://dx.doi.org/10.1109/68.544716. Accessed on: 4 Aug. 2021.

BERNHARDI, E. Bragg-grating-based rare-earth-ion-doped channel waveguide lasers

and their applications. Thesis (Doctorate) — University of Twente, 11 2012. Available
at: https://dx.doi.org/10.3990/1.9789036534505. Accessed on: 4 Aug. 2021.

BIRKS, T. A.; MORKEL, P. Jones calculus analysis of single-mode fiber sagnac

reflector. Appl. Opt., OSA, v. 27, n. 15, p. 3107–3113, Aug 1988. Available at:
https://dx.doi.org/10.1364/AO.27.003107. Accessed on: 4 Aug. 2021.

BROWN, D. A.; GARRETT, S. L. Interferometric fiber optic accelerometer. In: SPIE.

Fiber Optic and Laser Sensors VIII. Proceedings [...]. [S.l.: s.n.], 1991. v. 1367, p.
282–289. Available at: https://dx.doi.org/10.1117/12.24756. Accessed on: 4 Aug. 2021.

CARVALHO, M. M.; CAZO, R. M.; ALMEIDA, V. R.; BARBOSA, C. L. Fiber Bragg

grating based angular accelerometer: a first approach. In: JAROSZEWICZ, L. R. (Ed.).
Fifth European Workshop on Optical Fibre Sensors. Proceedings [...]. SPIE, 2013.
v. 8794, p. 37 – 41. Available at: https://dx.doi.org/10.1117/12.2026038. Accessed on: 4
Aug. 2021.

CAZO, R. M. Sistemas Interrogadores de Sensores Baseados em Grade de Bragg.

Dissertation (Master’s degree) — Aeronautics Institute of Technology (ITA), São José
dos Campos, 2001.

CAZO, R. M. Acelerômetro opto-mecânico baseado em grades de Bragg a fibra óptica.

111 p. Thesis (Doctorate) — Aeronautics Institute of Technology (ITA), São José dos
Campos, 2011.

CAZO, R. M.; BARBOSA, C. L.; RABELO, R. C.; SIQUEIRA, J. L. de; FERREIRA,

A. L. C. T.; HATTORI, H. T. Proposed interrogation system applied to a fbg
opto-mechanical accelerometer. Annals of Optics, ENFMC, vol, 2003. Available at:
http://www.sbf1.sbfisica.org.br/procs/2003/r estendido/c65.pdf. Accessed on: 4 Aug.
2021.

CAZO, R. M.; FERREIRA, J. L. de S.; CALDAS, T. de B.; ARRUDA, J. U. de;

ALLES, N.; LOUZADA, R. C. Improvement of fiber Bragg grating interrogator
sensitivity for triaxial accelerometer. In: IEEE. 2009 SBMO/IEEE MTT-S
International Microwave and Optoelectronics Conference (IMOC). Proceedings [...].
[S.l.: s.n.], 2009. p. 660–663.

CAZO, R. M.; RIBEIRO, E. dos R.; NUNES, M. B.; BARBOSA, C. L.; FERREIRA,
J. L. de S.; CALDAS, T. de B.; SANTOS, J. C. dos; ARRUDA, J. U. de. Six degree
freedom optical fiber accelerometer. In: AIP. AIP Conference Proceedings.
Proceedings [...]. [S.l.: s.n.], 2008. v. 1055, n. 1, p. 125–128. Available at:
https://dx.doi.org/10.1063/1.3002519. Accessed on: 4 Aug. 2021.
REFERENCES 124

CHANDRA, V.; TIWARI, U.; DAS, B. Elimination of light intensity noise using
dual-channel scheme for fiber MZI-based FBG sensor interrogation. IEEE Sensors
Journal, IEEE, v. 16, n. 8, p. 2431–2436, 2016. Available at:
https://dx.doi.org/10.1109/JSEN.2016.2515130. Accessed on: 4 Aug. 2021.

CHANG, W. S. et al. RF photonic technology in optical fiber links. Cambridge

University Press, 2002. Available at: https://dx.doi.org/10.1017/CBO9780511755729.
Accessed on: 4 Aug. 2021.

CHEN, C.; ZHANG, D.; DING, G.; CUI, Y. Broadband Michelson fiber-optic
accelerometer. Applied optics, Optical Society of America, v. 38, n. 4, p. 628–630, 1999.
Available at: https://dx.doi.org/10.1364/AO.38.000628. Accessed on: 4 Aug. 2021.

COLLET, E. Field guide to polarization. [S.l.]: SPIE, 2005. ISBN 0819458686.

CONFORTI, E.; RODIGHERI, M.; SUTILI, T.; GALDIERI, F. J. Acoustical and 1/ f

noises in narrow linewidth lasers. Optics Communications, Elsevier, v. 476, p. 126286,
2020. Available at: https://dx.doi.org/10.1016/j.optcom.2020.126286. Accessed on: 4
Aug. 2021.

CRANCH, G. A.; FLOCKHART, G. M. H.; KIRKENDALL, C. K. Distributed feedback

fiber laser strain sensors. IEEE Sensors Journal, v. 8, n. 7, p. 1161–1172, 2008.
Available at: https://dx.doi.org/10.1109/JSEN.2008.926876. Accessed on: 4 Aug. 2021.

DANDRIDGE, A.; TVETEN, A. B.; GIALLORENZI, T. G. Homodyne demodulation

scheme for fiber optic sensors using phase generated carrier. IEEE Transactions on
Microwave Theory and Techniques, v. 30, n. 10, p. 1635–1641, 1982. Available at:
https://dx.doi.org/10.1109/TMTT.1982.1131302. Accessed on: 4 Aug. 2021.

DESURVIRE, E.; ZYSKIND, J. L.; GILES, C. R. Design optimization for efficient

erbium-doped fiber amplifiers. Journal of lightwave technology, IEEE, v. 8, n. 11, p.
1730–1741, 1990. Available at: https://dx.doi.org/10.1109/50.60573. Accessed on: 4
Aug. 2021.

DJINOVIC, Z.; PAVELKA, R.; TOMIC, M.; SPRINZL, G.; PLENK, H.; LOSERT, U.;
BERGMEISTER, H.; PLASENZOTTI, R. In-vitro and in-vivo measurement of the
animal’s middle ear acoustical response by partially implantable fiber-optic sensing
system. Biosensors and Bioelectronics, v. 103, p. 176–181, 2018. ISSN 09565663.
Available at: https://dx.doi.org/10.1016/j.bios.2017.12.015. Accessed on: 4 Aug. 2021.

DJINOVIC, Z.; TOMIC, M.; PAVELKA, R.; SPRINZL, G.; TRAXLER, H.

Measurement of the human cadaver ossicle vibration amplitude by fiber-optic
interferometry. In: MIPRO 2020/Optoeletronics and Photonics. Proceedings [...].
Opatija, Croatia: [s.n.], 2020. p. 2232–2236. Available at:
http://docs.mipro-proceedings.com/opho/05 OPHO 6215.pdf. Accessed on: 4 Aug.
2021.

DOCKNEY, M. L.; JAMES, S. W.; TATAM, R. P. Fibre Bragg gratings fabricated

using a wavelength tuneable laser source and a phase mask based interferometer.
Measurement science and technology, IOP Publishing, v. 7, n. 4, p. 445, 1996. Available
at: https://dx.doi.org/10.1088/0957-0233/7/4/001. Accessed on: 4 Aug. 2021.
REFERENCES 125

DOEBELIN, E. O. Measurement systems application and design. 5th. ed. [S.l.]:

McGraw-Hill, 2007. ISBN 9780072438864.

DUAN, L. Z. Intrinsic thermal noise of optical fibres due to mechanical dissipation.

Electronics Letters, v. 46, n. 22, p. 1515–1516, 2010. Available at:
https://dx.doi.org/10.1049/el.2010.1653. Accessed on: 4 Aug. 2021.

DUO, Y.; XIANGGE, H.; FEI, L.; LIJUAN, G.; MIN, Z.; XIAOKANG, Q.; HAN, Y.
Self-suppression of common-mode noises of the different fiber optic interferometric
accelerometers. Optics express, Optical Society of America, v. 26, n. 12, p. 15384–15397,
2018. Available at: https://dx.doi.org/10.1364/OE.26.015384. Accessed on: 4 Aug. 2021.

DYNDGAARD, M. G. ytterbium-& erbium-doped silica for planar waveguide lasers &

amplifiers. Thesis (Doctorate) — Technical University of Denmark, 2001.

FELÃO, L. H. V. High gain approach and sliding mode control applied to quadrature
interferometer. Thesis (Doctorate) — Universidade Estadual Paulista (UNESP), 2019.
Available at: http://hdl.handle.net/11449/190782. Accessed on: 4 Aug. 2021.

FELÃO, L. H. V.; MARTIN, R. I.; SAKAMOTO, J. ao M. S.; TEIXEIRA, M. C. M.;

KITANO, C. Wide dynamic range quadrature interferometer with high-gain approach
and sliding mode control. Opt. Express, OSA, v. 27, n. 18, p. 25031–25045, Sep 2019.
Available at: https://dx.doi.org/10.1364/OE.27.025031. Accessed on: 4 Aug. 2021.

FERREIRA, W. S. J. Acelerômetro a fibra óptica baseado em interferômetro de

Mach-Zehnder com redes de Bragg de perı́odo longo. Dissertation (Master’s degree) —
Aeronautics Institute of Technology (ITA), São José dos Campos, 2020.

FEYNMAN, R. P.; LEIGHTON, R. B.; SANDS, M. The Feynman lectures on physics:

Vol. 1. New York, NY: Basic Books, 1963. Available at:
https://www.feynmanlectures.caltech.edu. Accessed on: 4 Aug. 2021.

FRANCO, S. Design with operational amplifiers and analog integrated circuits. 3rd.
ed. [S.l.]: McGraw-Hill, 2002. ISBN 0071121730.

FRITSCH, K.; ADAMOVSKY, G. Simple circuit for feedback stabilization of a

single-mode optical fiber interferometer. Review of Scientific Instruments, AIP, v. 52,
n. 7, p. 996–1000, 1981. Available at: https://dx.doi.org/10.1063/1.1136740. Accessed
on: 4 Aug. 2021.

GARDNER, D. L.; GARRETT, S. L. Fiber optic seismic sensor. In: DEPAULA, R. P.;
UDD, E. (Ed.). Fiber Optic and Laser Sensors V. Proceedings [...]. SPIE, 1988.
v. 0838, p. 271 – 278. Available at: https://dx.doi.org/10.1117/12.942518. Accessed on:
4 Aug. 2021.

GARTLAND, P. L. Fiber-Optic Michelson nterferometer with Faraday mirrors for

acoustic sensing using a 3 x 3 coupler and symmetric demodulation scheme.
Dissertation (Master’s degree) — Virginia Tech, 2016. Available at:
http://hdl.handle.net/10919/73370. Accessed on: 4 Aug. 2021.
REFERENCES 126

GIALLORENZI, T. G.; BUCARO, J. A.; DANDRIDGE, A.; SIGEL, G. H.; COLE,

J. H.; RASHLEIGH, S. C.; PRIEST, R. G. Optical fiber sensor technology. IEEE
Transactions on Microwave Theory and Techniques, v. 30, n. 4, p. 472–511, 1982.
Available at: https://dx.doi.org/10.1109/TMTT.1982.1131089. Accessed on: 4 Aug.
2021.

GUAN, B.-O.; JIN, L.; ZHANG, Y.; TAM, H.-Y. Polarimetric heterodyning fiber grating
laser sensors. Journal of lightwave technology, IEEE, v. 30, n. 8, p. 1097–1112, 2011.
Available at: https://dx.doi.org/10.1109/JLT.2011.2171916. Accessed on: 4 Aug. 2021.

GUO, Y.; WANG, Z.-Y.; QIU, Q.; SU, J.; WANG, Y.; SHI, S.; YU, Z. Theoretical and
experimental investigations on the temperature dependence of the refractive index of
amorphous silica. Journal of Non-Crystalline Solids, v. 429, p. 198–201, 2015. ISSN
0022-3093. Available at: https://dx.doi.org/10.1016/j.jnoncrysol.2015.09.008. Accessed
on: 4 Aug. 2021.

HAPPACH, M.; FELIPE, D. de; FRIEDHOFF, V. N.; IRMSCHER, G.; KRESSE, M.;
KLEINERT, M.; ZAWADZKI, C.; BRINKER, W.; MöHRLE, M.; KEIL, N.;
HOFMANN, W.; SCHELL, M. Effect of optical feedback on the wavelength tuning in
dbr lasers. Journal of Lightwave Technology, v. 38, n. 17, p. 4824–4833, 2020. Available
at: https://dx.doi.org/10.1109/JLT.2020.2996131. Accessed on: 4 Aug. 2021.

HECHT, E. Optics. [S.l.]: Pearson Education Inc, 2017. ISBN 9780133977226.

HERTSENS, T. Measuring diode laser characteristics: diode lasers approach ubiquity,

but they still can be frustrating to work with. Lasers & Optronics, Reed Business
Information, Inc.(US), v. 8, n. 2, p. 37–42, 1989.

HTEIN, L.; GUNAWARDENA, D. S.; CHUNG, W.-H.; AU, H.-Y.; TAM, H.-Y.
Accelerometer employing a side-hole fiber in a sagnac interferometer. Journal of
Lightwave Technology, v. 39, n. 10, p. 3303–3311, 2021. Available at:
https://dx.doi.org/10.1109/JLT.2021.3056474. Accessed on: 4 Aug. 2021.

JACKSON, D.; PRIEST, R.; DANDRIDGE, A.; TVETEN, A. B. Elimination of drift in

a single-mode optical fiber interferometer using a piezoelectrically streched coiled fiber.
Applied Optics, Optical Society of America, v. 19, n. 17, p. 2926–2929, 1980. Available
at: https://dx.doi.org/10.1364/AO.19.002926. Accessed on: 4 Aug. 2021.

JINDAL, S. K.; PRIYA, S.; PRASADH, S. K. Design guidelines for mems optical
accelerometer based on dependence of sensitivities on diaphragm dimensions. Journal of
Circuits, Systems and Computers, v. 29, n. 07, p. 2050107, 2020. Available at:
https://dx.doi.org/10.1142/S0218126620501078. Accessed on: 4 Aug. 2021.

KERSEY, A.; MARRONE, M.; DAVIS, M. Polarisation-insensitive fibre optic Michelson

interferometer. Electronics letters, IET, v. 27, n. 6, p. 518–520, 1991.

KERSEY, A. D.; DANDRIDGE, A.; TVETEN, A. B. Dependence of visibility on input

polarization in interferometric fiber-optic sensors. Opt. Lett., OSA, v. 13, n. 4, p.
288–290, Apr 1988. Available at: https://dx.doi.org/10.1364/OL.13.000288. Accessed
on: 4 Aug. 2021.
REFERENCES 127

KERSEY, A. D.; KOO, K. P.; DAVIS, M. A. Fiber optic Bragg grating laser sensors. In:
SPIE. Fiber Optic and Laser Sensors XII. Proceedings [...]. San Diego, 1994. v. 2292, p.
102–112. Available at: https://dx.doi.org/10.1117/12.191824. Accessed on: 4 Aug. 2021.

Keysight Technologies. Evaluating Oscilloscope Vertical Noise Characteristics. [S.l.],

2015. (Application notes).

KIRKENDALL, C. K.; DANDRIDGE, A. Overview of high performance fibre-optic

sensing. Journal of Physics D: Applied Physics, IOP Publishing, v. 37, n. 18, p. R197,
2004. Available at: https://dx.doi.org/10.1088/0022-3727/37/18/R01. Accessed on: 4
Aug. 2021.

LEFÈVRE, H. C. The fiber-optic gyroscope. [S.l.]: Artech house, 2014. ISBN

9781608076956.

LINESSIO, R. P. Estudo e desenvolvimento de sensores a fibra óptica para a análise e

controle de vibrações em máquinas elétricas e estruturas. Thesis (Doctorate) —
Universidade Tecnológica Federal do Paraná, 2020. Available at:
http://repositorio.utfpr.edu.br/jspui/handle/1/24495. Accessed on: 4 Aug. 2021.

LINESSIO, R. P.; TAVARES, L. H. dos S.; SILVA, T. d.; OLIVEIRA, V. de;

BAVASTRI, C. A.; ANTUNES, P. F. da C.; SILVA, J. C. Cardozo da. Biaxial optical
accelerometer based on ultra-high numerical aperture fiber. IEEE Sensors Journal,
v. 19, n. 10, p. 3690–3697, 2019. Available at:
https://dx.doi.org/10.1109/JSEN.2019.2894988. Accessed on: 4 Aug. 2021.

LIU, T.; CHENG, J.; LV, D.; LUO, Y.; YAN, Z.; WANG, K.; LI, C.; LIU, D.; SUN, Q.
DBR fiber laser based high-resolution accelerometer network. Journal of Lightwave
Technology, IEEE, v. 37, n. 13, p. 2946–2953, 2019. Available at:
https://dx.doi.org/10.1109/JLT.2019.2908005. Accessed on: 4 Aug. 2021.

MARTIN, R. I. Nonlinear control system based on variable structure and sliding

modes for two beam optical interferometry. Thesis (Doctorate) — Universidade
Estadual Paulista (UNESP), 2018. Available at: http://hdl.handle.net/11449/153459.
Accessed on: 4 Aug. 2021.

MARTIN, R. I.; SAKAMOTO, J. M.; TEIXEIRA, M. C.; MARTINEZ, G. A.;

PEREIRA, F. C.; KITANO, C. Nonlinear control system for optical interferometry
based on variable structure control and sliding modes. Optics express, Optical Society of
America, v. 25, n. 6, p. 6335–6348, 2017. Available at:
https://dx.doi.org/10.1364/OE.25.006335. Accessed on: 4 Aug. 2021.

MARTINELLI, M. A universal compensator for polarization changes induced by

birefringence on a retracing beam. Optics Communications, v. 72, n. 6, p. 341–344,
1989. ISSN 0030-4018. Available at: https://dx.doi.org/10.1016/0030-4018(89)90436-7.
Accessed on: 4 Aug. 2021.

MELTZ, G.; MOREY, W.; GLENN, W. Formation of Bragg gratings in optical fibers by
a transverse holographic method. Optics letters, Optical Society of America, v. 14,
n. 15, p. 823–825, 1989. Available at: https://dx.doi.org/10.1364/OL.14.000823.
Accessed on: 4 Aug. 2021.
REFERENCES 128

MICHAUD-BELLEAU, V.; CHARLET, M.; TOURIGNY-PLANTE, A.; DESCHÊNES,

J.-D.; GENEST, J. External serrodyne modulation for the suppression of low-frequency
noise in quadrature interferometry. Opt. Lett., OSA, v. 45, n. 3, p. 670–673, 2020.
Available at: https://dx.doi.org/10.1364/OL.376542. Accessed on: 4 Aug. 2021.

MILES, R.; DANDRIDGE, A.; TVETEN, A.; TAYLOR, H.; GIALLORENZI, T.

Feedback-induced line broadening in cw channel-substrate planar laser diodes. Applied
Physics Letter, AIP, v. 37, n. 11, p. 990–992, 1980. Available at:
https://dx.doi.org/10.1063/1.91744. Accessed on: 4 Aug. 2021.

MILLER, G. A. Decrease in photosensitivity of erbium-doped fiber under tension:

Implications for distributed feedback fiber lasers. Journal of Lightwave Technology,
IEEE, v. 36, n. 14, p. 3040–3045, 2018. Available at:
https://dx.doi.org/10.1109/JLT.2018.2830298. Accessed on: 4 Aug. 2021.

NUNES, G. F. S.; PALMA, V. B.; MARTIN, R. I.; FELÃO, L. H. V.; TEIXEIRA, M.

C. M.; KITANO, C.; CAZO, R. M.; SAKAMOTO, J. M. S. Characterization of a
thin-film metal-coated fiber optical phase modulator based on thermal effect with a
nonlinear control interferometer. Applied Optics, in press, 2021. Available at:
https://dx.doi.org/10.1364/AO.428703. Accessed on: 4 Aug. 2021.

OGATA, K. Modern control engineering. [S.l.]: Prentice hall, 2010. ISBN

9780136156734.

PARIDA, O. P.; THOMAS, J.; NAYAK, J.; ASOKAN, S. Double-l cantilever-based

fiber Bragg grating accelerometer. IEEE Sensors Journal, v. 19, n. 23, p. 11247–11254,
2019. Available at: https://dx.doi.org/10.1109/JSEN.2019.2936463. Accessed on: 4 Aug.
2021.

PENG, F.; LV, Y.; LI, H.; TIAN, S.; CHEN, W.; YANG, J. Sensitivity prediction of
multiturn fiber coil-based fiber-optic flexural disk seismometer via finite element method
analysis. Journal of Lightwave Technology, IEEE, v. 35, n. 18, p. 3870–3876, 2017.
Available at: https://dx.doi.org/10.1109/JLT.2017.2724603. Accessed on: 4 Aug. 2021.

PÉREZ-ALONZO, A.; SANDOVAL-ROMERO, G. Accelerometer prototype based on

enhanced fiber Bragg grating overlapping interrogation method. Optik, p. 167410, 2021.
ISSN 0030-4026. Available at: https://dx.doi.org/10.1016/j.ijleo.2021.167410. Accessed
on: 4 Aug. 2021.

PISTONI, N. C.; MARTINELLI, M. Polarization noise suppression in retracing optical

fiber circuits. Opt. Lett., OSA, v. 16, n. 10, p. 711–713, May 1991. Available at:
https://dx.doi.org/10.1364/OL.16.000711. Accessed on: 4 Aug. 2021.

PUPALAIKIS, P. J. Understanding vertical resolution in oscilloscopes. In: UBM

AMERICAS. DesignCon 2017. Proceedings [...]. Santa Clara, CA: DesignCon, 2017. p.
1–23. Available at: http://cdn.teledynelecroy.com/files/whitepapers/designcon-2017-
understanding-vertical-resolution.pdf. Accessed on: 4 Aug. 2021.

REDDY, V. S. The spacex effect. New Space, v. 6, n. 2, p. 125–134, 2018. Available at:
https://dx.doi.org/10.1089/space.2017.0032. Accessed on: 4 Aug. 2021.
REFERENCES 129

RIGROD, W. W. Saturation effects in high-gain lasers. Journal of Applied Physics,

American Institute of Physics, v. 36, n. 8, p. 2487–2490, 1965. Available at:
https://dx.doi.org/10.1063/1.1714517. Accessed on: 4 Aug. 2021.

RODRIGUES, N. Lasers - Notas De Aula. 1. ed. [S.l.]: Livraria Da Fisica, 2019. ISBN
9788578615840.

RONG, Q.; GUO, T.; BAO, W.; SHAO, Z.; PENG, G.-D.; QIAO, X. Highly sensitive
fiber-optic accelerometer by grating inscription in specific core dip fiber. Scientific
reports, Nature Publishing Group, v. 7, n. 1, p. 1–9, 2017. Available at:
https://dx.doi.org/10.1038/s41598-017-12322-6. Accessed on: 4 Aug. 2021.

ROSELLÓ-MECHÓ, X.; DELGADO-PINAR, M.; DÍEZ, A.; ANDRÉS, M. V.

Measurement of pockels’ coefficients and demonstration of the anisotropy of the
elasto-optic effect in optical fibers under axial strain. Opt. Lett., OSA, v. 41, n. 13, p.
2934–2937, Jul 2016. Available at: https://dx.doi.org/10.1364/OL.41.002934. Accessed
on: 4 Aug. 2021.

SAHOTA, J. K.; GUPTA, N.; DHAWAN, D. Fiber Bragg grating sensors for monitoring
of physical parameters: a comprehensive review. Optical Engineering, SPIE, v. 59, n. 6,
p. 060901, 2020. Available at: https://dx.doi.org/10.1117/1.OE.59.6.060901. Accessed
on: 4 Aug. 2021.

SHEA, H. R. Reliability of MEMS for space applications. In: TANNER, D. M.;

RAMESHAM, R. (Ed.). Reliability, Packaging, Testing, and Characterization of
MEMS/MOEMS V. Proceedings [...]. SPIE, 2006. v. 6111, p. 84 – 93. Available at:
https://dx.doi.org/10.1117/12.651008. Accessed on: 4 Aug. 2021.

SHI, J.; GUANG, D.; LI, S.; WU, X.; ZHANG, G.; ZUO, C.; ZHANG, G.; WANG, R.;
GE, Q.; YU, B. Phase-shifted demodulation technique with additional modulation based
on a 3x3 coupler and efa for the interrogation of fiber-optic interferometric sensors. Opt.
Lett., OSA, v. 46, n. 12, p. 2900–2903, Jun 2021. Available at:
https://dx.doi.org/10.1364/OL.420655. Accessed on: 4 Aug. 2021.

SONG, Z.; QI, H.; GUO, J.; WANG, C.; PENG, G. Characteristics research on
self-amplified distributed feedback fiber laser. Photonic Sensors, Springer, v. 4, n. 3, p.
265–268, 2014. Available at: https://dx.doi.org/10.1007/s13320-014-0170-7. Accessed
on: 4 Aug. 2021.

SPACE EXPLORATION TECHNOLOGIES CORP. FALCON user’s guide.

Hawthorne, CA, 2020. 72 p.

SVELTO, O. Principles of lasers. Springer, 2010. Available at:

https://dx.doi.org/10.1007/978-1-4419-1302-9. Accessed on: 4 Aug. 2021.

TAGHAVI, M.; LATIFI, H.; PARSANASAB, G. M.; ABEDI, A.; NIKBAKHT, H.;
POORGHADIRI, M. H. A dual-axis moems accelerometer. IEEE Sensors Journal,
v. 21, n. 12, p. 13156–13164, 2021. Available at:
https://dx.doi.org/10.1109/JSEN.2021.3072333. Accessed on: 4 Aug. 2021.
REFERENCES 130

THE INSTITUTE OF ELECTRICAL AND ELECTRONICS ENGINEERS.

ANSI/IEEE Std 149-1979: IEEE standard test procedures for antennas. New York, NY,
1979. 135 p. Available at: https://dx.doi.org/10.1109/IEEESTD.1979.120310. Accessed
on: 4 Aug. 2021.

TVETEN, A.; DANDRIDGE, A.; DAVIS, C. M.; GIALLORENZI, T. G. Fibre optic

accelerometer. Electronics Letters, IET, v. 16, n. 22, p. 854–856, 1980. Available at:
https://dx.doi.org//10.1049/el:19800607. Accessed on: 4 Aug. 2021.

UDD, E.; SPILLMAN, W. B. Fiber optic sensors: an introduction for engineers and
scientists. John Wiley & Sons, 2011. Available at:
https://dx.doi.org/10.1002/9781118014103. Accessed on: 4 Aug. 2021.

VLASOV, A. A.; PLOTNIKOV, M. Y.; VOLKOVSKY, S. A.; ALEINIK, A. S.;

MOTORIN, E. A.; SHARKOV, I. A.; MAKARENKO, A. A. Development of the
passive vibroacoustic isolation system for the path matched differential interferometry
based fiber-optic sensors. Optical Fiber Technology, Elsevier, v. 57, p. 102241, ago 2020.
Available at: https://dx.doi.org/10.1016/j.yofte.2020.102241. Accessed on: 4 Aug. 2021.

VOHRA, S. T.; DANVER, B.; TVETEN, A. B.; DANDRIDGE, A. High performance

fiber optic accelerometers. In: Optical Fiber Sensors. Proceedings [...]. Optical Society
of America, 1996. p. Th55. Available at: https://dx.doi.org/10.1364/OFS.1996.Th55.
Accessed on: 4 Aug. 2021.

WANG, J.; PENG, G.; HU, Z.; YANG, H.; HU, Y. Design and analysis of a high
sensitivity fbg accelerometer based on local strain amplification. IEEE Sensors Journal,
v. 15, n. 10, p. 5442–5449, 2015. Available at:
https://dx.doi.org/10.1109/JSEN.2014.2370640. Accessed on: 4 Aug. 2021.

WANG, J.; ZENG, Y.; LIN, C.; HU, Z.; PENG, G.; HU, Y. A miniaturized fbg
accelerometer based on a thin polyurethane shell. IEEE Sensors Journal, v. 16, n. 5, p.
1210–1216, 2016. Available at: https://dx.doi.org/10.1109/JSEN.2015.2501983.
Accessed on: 4 Aug. 2021.

WARD, L. M.; GREENWOOD, P. E. 1/f noise. Scholarpedia, v. 2, n. 12, p. 1537,

revision #137265, 2007. Available at: https://dx.doi.org/10.4249/scholarpedia.1537.
Accessed on: 4 Aug. 2021.

WU, Y.; TANG, H.; WANG, C. Study on the performance of passive digital phase
demodulation algorithm based on 3×3 coupler and the hardware implementation. In:
YAN, J. (Ed.). Optics Frontiers Online 2020: Distributed Optical Fiber Sensing
Technology and Applications. Proceedings [...]. SPIE, 2021. v. 11607, p. 49 – 58.
Available at: https://dx.doi.org/10.1117/12.2585238. Accessed on: 4 Aug. 2021.

XIE, F.; REN, J.; CHEN, Z.; FENG, Q. Vibration-displacement measurements with a
highly stabilised optical fiber Michelson interferometer system. Optics & Laser
Technology, Elsevier, 2009. Available at:
https://dx.doi.org/10.1016/j.optlastec.2009.06.010. Accessed on: 4 Aug. 2021.

YANG, Y.; WANG, Z.; CHANG, T.; CHENG, L.; YU, M.; CUI, H.-L. Performance
optimization of fiber optic interferometric accelerometer based on phase noise analysis.
REFERENCES 131

IEEE Sensors Journal, v. 19, n. 22, p. 10498–10505, 2019. Available at:

https://dx.doi.org/10.1109/JSEN.2019.2929763. Accessed on: 4 Aug. 2021.

YARIV, A. Operator algebra for propagation problems involving phase conjugation and
nonreciprocal elements. Appl. Opt., OSA, v. 26, n. 21, p. 4538–4540, Nov 1987.
Available at: https://dx.doi.org/10.1364/AO.26.004538. Accessed on: 4 Aug. 2021.

YOUNG, H. D.; FREEDMAN, R. A.; SEARS, F. W.; ZEMANSKY, M. Sears e

Zemansky fı́sica I: mecânica. [S.l.]: Pearson Addison Wesley, 2008. ISBN
9788588639300.

YU, K.; LAI, C.-C.; WU, C.; ZHAO, Y.; LU, C.; TAM, H.-Y. A high-frequency
accelerometer based on distributed Bragg reflector fiber laser. IEEE Photonics
Technology Letters, IEEE, v. 26, n. 14, p. 1418–1421, 2014. Available at:
https://dx.doi.org/10.1109/LPT.2014.2326558. Accessed on: 4 Aug. 2021.

ZHANG, F.; JIANG, S.; WANG, C.; NI, J.; ZHAO, Q. Broadband and high sensitivity
fbg accelerometer based on double diaphragms and h-shaped hinges. IEEE Sensors
Journal, p. 1–1, 2020. Available at: https://dx.doi.org/10.1109/JSEN.2020.3013611.
Accessed on: 4 Aug. 2021.

ZHANG, W.; WANG, Z.; HUANG, W.; LIU, W.; LI, L.; LI, F. Fiber laser sensors for
micro seismic monitoring. Measurement, v. 79, p. 203–210, 2016. ISSN 0263-2241.
Available at: https://dx.doi.org/10.1016/j.measurement.2015.09.046. Accessed on: 4
Aug. 2021.

ZHANG, Y.; GUAN, B.-O.; TAM, H.-Y. Ultra-short distributed Bragg reflector fiber
laser for sensing applications. Optics express, Optical Society of America, v. 17, n. 12,
p. 10050–10055, 2009. Available at: https://dx.doi.org/10.1364/OE.17.010050. Accessed
on: 4 Aug. 2021.

ZHOU, R.; CHEN, F.; LI, S.; WANG, R.; QIAO, X. Three-dimensional vector
accelerometer using a multicore fiber inscribed with three fbgs. J. Lightwave Technol.,
OSA, v. 39, n. 10, p. 3244–3250, 2021. Available at:
https://dx.doi.org/10.1109/JLT.2021.3058240. Accessed on: 4 Aug. 2021.

ZYSKIND, J.; MIZRAHI, V.; DIGIOVANNI, D.; SULHOFF, J. Short single frequency
erbium-doped fibre laser. Electronics Letters, IET, v. 28, n. 15, p. 1385–1387, 1992.
Available at: https://dx.doi.org/10.1049/el:19920881. Accessed on: 4 Aug. 2021.
Appendix A - Digital control
implementation

In this appendix the digital implementation of the control loop in LabVIEW is shown.

A.1
RT Main.vi
Real Time Processor VI

file path
/D/Users/Stinky/Documents/23.05.2019/Control/teste.tdms

Plot 0
Waveform to AV Max Value
Parameters FIFO
10 0
Min Value
8
Gain Epsilon (ex: 0.01) Amostragem temp(uSec) 0
1 0.01 100 6
Actual Depth
4 0

2 Elements Remaining

STOP 0
0
16bit integer DC Value
0 25 50 75 100 125 150 175 200
Time 0

Control Signal Phase-Plane Control x LPF

Interferometric Signal 1
0.8
LPF
Control Signal
0.6
10
0.4
7.5
0.2
5
0
2.5
-0.2
0
-0.4
-2.5
-0.6
-5
-0.8
-7.5
-1
-10 -10 -7.5 -5 -2.5 0 2.5 5 7.5 10
0 1023 x1bar
Time

0 Sinal de controle
Saida interferometro
file path Saida LPF

Amostragem temp(uSec)
create or replace
50

201 2 10
25000 3000
0
0 Gain
0 -1
Waveform to AV
Actual Epsilon (ex: 0.01)
20
False constant turns On the Control Logic. Control Signal
180
1000
Phase-Plane

0.05000000074506 5 V/100 Max Value Elements Remaining

stop

Min Value 2
0.05000000074506
0.0048828125
16bit integer DC Value
Measuring AV 20V/(2^12bits)
APPENDIX A. DIGITAL CONTROL IMPLEMENTATION 133

A.2 Calibration FPGA VI

calibracao.vi

rampa saida interferometro

0 0

Periodo de amostragem(mSec)
5

saida interferometro
Periodo de amostragem(mSec)
ConnectorC/AI0

rampa 0.0048828125

20V/(2^12bits)

ConnectorC/AO1
0.0048828125
APPENDIX A. DIGITAL CONTROL IMPLEMENTATION 134
FPGA.vi
A.3 Control Loop FPGA VI

Periodo de amostragem controle(uSec) saida LPF

10 0

Periodo amostragem temp(uSec) AI0

100 0
initial condition Sinal de controle
0.00 0

nivel DC
0

Ganho initialize
0 OFF

gamma
0

Low Index Modulation Nonlinear Control

Inverting to connect
False Fals Fals conditioning circuit
ConnectorC/AI0 after myRIO Sinal de controle
20/4096 = 20/(2^12) saida LPF
0.0048828125
10
-10
initialize
nivel DC
9.999999974752E-7 AI0 0.0048828125
gamma
4.0 20/4096 = 20V/(2^12bits)
Ganho 6.0

LPF
initial condition ConnectorC/AO1
----------*gain
|LPF|+Y

Periodo de amostragem(uSec) Simpson

Integrator

Communication with RT

Amostragem temp(uSec) Sinal de controle FIFO

Write
0.0048828125
Element
AI0 -1 Timeout
saida LPF Timed Out?

True F F

0.0 Fals Fals

Tru Tru
0.50
 APPENDIX A. DIGITAL CONTROL IMPLEMENTATION 135
FPGA.vi
A.4 Two Input Control Implementation FPGA
D:\Users\Stinky\Google Drive\ITA\Codes\LabVIEW\Low index modulation Control v3 - 2 inp
Last modified on 9/11/2019 at 8:09 AM
The present work made the implementation of the two input into the control VI as
Printed on 3/30/2020 at 7:26 PM
shown next, the rest of the VI logic is the same as in Appendix A.3.
Low Index Modulation Nonlinear Control

ConnectorC/AI0 20/4096 = 20/(2^12) False

ConnectorC/AI1 0.0048828125 saida LPF

2048

initialize

9.999999974752E-7
gamma
AI0 4.0

Ganho 6.0

Based on Dual channel MZI (Chandra,2016) LPF

initial condition
----------*gain
|LPF|+Y

Periodo de amostragem(uSec) Simpson

Integrator

Communication with RT

Amostragem temp(uSec) Sinal de controle FIFO

Write
0.0048828125
Element
AI0 -1 Timeout
saida LPF Timed Out?
 FOLHA DE REGISTRO DO DOCUMENTO

1. 2. 3. 4.
CLASSIFICAÇÃO/TIPO DATA DOCUMENTO Nº Nº DE PÁGINAS
DM 17 de agosto de 2021 DCTA/ITA/DM-069/2021 135
5.
TÍTULO E SUBTÍTULO:
Fiber optic accelerometer based on an interferometer with nonlinear control for space applications
6.
AUTOR(ES):
Gabriel Fernandes de Souza Nunes
7.
INSTITUIÇÃO(ÕES)/ÓRGÃO(S) INTERNO(S)/DIVISÃO(ÕES):
Instituto Tecnológico de Aeronáutica – ITA
8.
PALAVRAS-CHAVE SUGERIDAS PELO AUTOR:
Acelerômetro; Laser; Interferômetro; Fibra óptica; Sensores.
9.
PALAVRAS-CHAVE RESULTANTES DE INDEXAÇÃO:
Acelerômetros; Fibras ópticas; Interferômetros; Lasers; Sensores; Fı́sica.
10.
APRESENTAÇÃO: (X) Nacional ( ) Internacional
ITA, São José dos Campos. Curso de Mestrado. Programa de Pós-Graduação em Ciências e Técnologias
Espaciais. Área de Sensores e Atuadores Espaciais. Orientador: Prof. Dr. João Marcos Salvi Sakamoto. Defesa
em 30/07/2021. Publicada em 2021.
11.
RESUMO:
Fiber optic accelerometers are sensors designed to measure vehicle vibrations or acceleration for navigation
purposes, and which have been studied and used in space and aeronautical operations, due to their lightweight,
high sensitivity, immunity to electromagnetic interference, ability to function in high temperatures, resistance to
extreme vibrations and high levels of shock, and multiplexing capability. High-performance accelerometers are
still embargoed by countries that possess such technology, which motivates the development of high-performance
accelerometers with national technology. Furthermore, accelerometers available on the market are electrical,
requiring shielding in space applications, may require micro-machining techniques, and do not provide high
accuracy. In this work, an optical fiber interferometer with nonlinear control is proposed to interrogate an opto-
mechanical accelerometer. The nonlinear control is used to stabilize the interferometer operating point at its most
sensitive point (phase quadrature point) and suppress signal fading. Michelson and Mach-Zehnder interferometers
were studied and assembled with fiber optic components. The control was digitally implemented to stabilize the
interferometers both at the quadrature point and at a generic operating point. Optical phase modulators were
characterized to validate the correct operation of the controlled interferometer and the demodulation technique
used to recover the signal. The mechanical model of a mass-spring-damper accelerometer was studied and
assembled. Sensor element location configurations were analyzed, either in the interferometer sensor arm or
in a fiber laser cavity. This led to the study and characterization of erbium-doped fiber optic lasers. Thus,
erbium-doped fiber optic amplifiers were characterized and a multimode laser was assembled and characterized
to act as a sensing element. An opto-mechanical accelerometer was mounted and positioned on the sensing arm
of the Mach-Zehnder interferometer. Then, a manual characterization was performed and the following values
were estimated: resonant frequency of 503.3 Hz, the damping factor of 60.5 s−1 , the sensitivity of 4.432 rad/g
at 100 Hz and dynamic range from 0.01 g to 0.187 g (25 dB). The resulting fiber laser provided an unstable
interferometric signal, not allowing the laser cavity to be used as a sensing element for the time being. On
the other hand, it presented a potential for future work, since its estimated sensitivity is 37 times greater than
the sensor element in the interferometer arm. In conclusion, the opto-mechanical accelerometer presented an
estimated sensitivity value that agrees with the literature and a resonance frequency above the region of interest
for spatial applications, demonstrating proper functioning using the interferometer with nonlinear control.

12.
GRAU DE SIGILO:
(X) OSTENSIVO ( ) RESERVADO ( ) SECRETO