Novel Optical Sensors for High Temperature Measurement

in Harsh Environments
Yibing Zhang
Dissertation submitted to the Faculty of the
Virginia Polytechnic Institute and State University
in partial fulfillment of the requirements for the degree of
Doctor of Philosophy
in
Electrical and Computer Engineering
Dr. Anbo Wang, Co-Chairman
Dr. Ahmad Safaai-Jazi, Co-Chairman
Dr. Gary R. Pickrell
Dr. Ira Jacobs
Dr. Roger H. Stolen
July 24, 2003
Blacksburg, Virginia
Keywords: Fiber optic sensors, Temperature, Birefringence, Polarimetry,
Low-coherence interferometry
Copyright 2003. Yibing Zhang
Novel Optical Sensors for High Temperature Measurement
in Harsh Environments
Yibing Zhang
(Abstract)
Accurate measurement of temperature is essential for the safe and efficient operation and
control of a vast range of industrial processes. Many of these processes involve harsh
environments, such as high temperature, high pressure, chemical corrosion, toxicity,
strong electromagnetic interference, and high-energy radiation exposure. These extreme
physical conditions often prevent conventional temperature sensors from being used or
make them difficult to use. Novel sensor systems should not only provide accurate and
reliable temperature measurements, but also survive the harsh environments through
proper fabrication material selections and mechanical structure designs.
This dissertation presents detailed research work on the design, modeling,
implementation, analysis, and performance evaluation of novel optical high temperature
sensors suitable for harsh environment applications. For the first time to our knowledge,
an optical temperature sensor based on the broadband polarimetric differential
interferometric (BPDI) technology is proposed and tested using single crystal sapphire
material. With a simple mechanically structured sensing probe, in conjunction with an
optical spectrum-coded interferometric signal processing technique, the proposed single
crystal sapphire optical sensor can measure high temperature up to 1600
o
C in the harsh
environments with high accuracy, corrosion resistance, and long-term measurement
stability. Based on the successfully demonstrated sensor prototype in the laboratory, we
are confident of the next research step on sensor optimization and scale-up for full field
implementations. The goal for this research has been to bring this temperature sensor to a
level where it will become commercially viable for harsh environment applications
associated with industries.
iii
Acknowledgements
I wish to express my deepest appreciation to Dr. Anbo Wang, and Dr. Ahmad Safaai-Jazi,
for serving as my advisors. Without their patient, guidance and constant supports, this
dissertation would not have been possible. As mentors and friends, they continually and
convincingly conveyed a spirit of adventure and an excitement in regard to research and
scholarship during the past years. With their dedication, confidence and incredible
achievements, they will continue to be my mentors shedding light on my future journey.
I also would like to sincerely thank Dr. Gary R. Pickrell, Dr. Guy J. Indebetouw, Dr. Ira
Jacobs, and Dr. Roger H. Stolen, for serving on my committee and for their
encouragements and valuable suggestions to improve the quality of the work presented
here. I hope Dr. Guy J. Indebetouw get well soon.
My gratitude also goes to all of my colleagues and friends at Center for Photonics (CPT),
which has been shared as a home. Among them, special thanks goes to former CPTers,
Dr. Russsll G. May, Dr Hai Xiao, Dr. Bing Qi, and Jiangdong Deng, who introduced me
to CPT three years ago, as well as current CPTers, Yan Zhang and Xiaopei Chen, for
their valuable suggestions and supports to this presented work. I am also grateful to
Debbie Collins, Kathy Acosta and Bill Cockey, they have made the CPT a place to work
with great pleasure.
Finally, words alone cannot express the thanks I own to my loving wife, Chimge, for her
caring, encouragement and sacrifice; to my parents, who raised me and trust me with
their endless love; and to my two young sisters, Shudan and Jialing, they remind me that
family is more important than any number of academic degrees.
iv
Table of Content
Abstract.............................................................................................................................. ii
Acknowledgements .......................................................................................................... iii
Table of Content............................................................................................................... iv
List of Figures.................................................................................................................. vii
List of Tables .................................................................................................................. xiii
Chapter 1. Introduction ................................................................................................... 1
1.1 Background information on the proposed research.................................................. 1
1.2 Review of non-optical techniques for temperature measurements........................... 2
1.2.1 High temperature thermocouples....................................................................... 2
1.2.2 Acoustic methods............................................................................................... 3
1.3 Review of optical techniques for temperature measurements .................................. 4
1.3.1 Remote pyrometers ............................................................................................ 4
1.3.2 Thermometers based on thermal expansion....................................................... 5
1.3.3 Fluorescence thermometers ............................................................................... 5
1.3.4 Thermometers based on optical scattering......................................................... 6
1.4 Industrial needs for novel high temperature sensors................................................. 6
1.5 Special requirements for temperature sensors in coal gasifier................................ 10
1.6 Scope of research.................................................................................................... 12
Chapter 2. Principles of operation of the designed optical sensors............................ 13
2.1 Optical sensing elements design............................................................................. 14
2.1.1 Review of EFPI sapphire fiber sensor ............................................................. 14
2.1.2 Broadband polarimetric sapphire sensor.......................................................... 17
2.2 Signal processing units ........................................................................................... 20
2.2.1 Review of signal processing methods for interferometric sensors .................. 20
2.2.2 SCIIB signal processor .................................................................................... 21
2.2.3 Spectral domain white light interferometric signal processor ......................... 22
2.3 Blackbody radiation subtraction............................................................................. 25
2.4 Configurations of designed temperature sensor systems ........................................ 27
2.5 Mathematical model for BPDI temperature sensor ................................................ 30
2.6 Major advantages of BPDI sensor system.............................................................. 34
Chapter 3. Design and implementation of temperature sensor systems.................... 36
3.1 Fabrication materials for sensing probes ................................................................ 36
v
3.1.1 Properties of single crystal sapphire ................................................................ 37
3.1.2 Properties of single crystal zirconia................................................................. 40
3.2 Implementation of sensing probes .......................................................................... 41
3.2.1 A sensing probe with a sapphire disk, zirconia prism (SDZP) structure......... 41
3.2.2 A sensing probe with a sapphire prism (SP) structure..................................... 44
3.2.3 Total internal reflection in the sensor probe .................................................... 48
3.3 Signal processor and software implementation ...................................................... 49
3.3.1 Signal processor implementation..................................................................... 49
3.3.2 Signal processing algorithm development ....................................................... 52
3.3.3 Software design and implementation............................................................... 55
3.4 Overview of BPDI temperature sensor system....................................................... 62
Chapter 4. System noise analysis and performance optimization.............................. 64
4.1 Optical spectrum induced noise.............................................................................. 64
4.1.1 Wavelength drift .............................................................................................. 65
4.1.2 Spectral bandwidth broadening/narrowing effects .......................................... 71
4.2 Degradation effect due to the visibility of the interference spectrum..................... 73
4.3 Optical birefringence effects................................................................................... 77
4.3.1 Rotation about the slow-axis (i.e. C-axis)........................................................ 80
4.3.2 Rotation about the fast-axis (i.e. A-axis) ......................................................... 83
4.3.3 Rotation about an axis 45º from the slow-axis/fast-axis.................................. 86
4.4 Opto-electronic noise in spectrum measurements .................................................. 88
4.5 Optical fiber induced noises.................................................................................... 91
4.6 Summary of system noises and their optimization ................................................. 92
4.7 Power budget .......................................................................................................... 98
Chapter 5. Performance evaluation of designed temperature sensors .................... 101
5.1 Definitions of performance characteristics ........................................................... 101
5.2 Characterization of white light signal processor................................................... 103
5.2.1 Capability of compensating optical source power fluctuations ..................... 103
5.2.2 Capability of compensating optical fiber transmission loss .......................... 106
5.2.3 Capability of compensating temperature fluctuations ................................... 107
5.3 Blackbody radiation subtraction........................................................................... 108
5.4 Calibration of BPDI sensing system..................................................................... 111
5.4.1 Construction of temperature calibration system............................................ 111
5.4.2 Temperature sensor calibration...................................................................... 112
5.5 Performance evaluations of BPDI sensing system............................................... 115
vi
5.5.1 Repeatability of the measurements ................................................................ 115
5.5.2 Evaluation of accuracy................................................................................... 117
5.5.3 Long-term stability tests ................................................................................ 118
5.5.4 Sensitivity (resolution) tests........................................................................... 120
5.5.5 Hysteresis in the temperature measurements................................................. 121
5.5.6 Frequency response........................................................................................ 123
Chapter 6 Expansion of the BPDI sensing technology for other measurements .... 127
6.1 Multi-parameter measurement toolbox based on optical birefringence ............... 127
6.2 Pressure sensor with temperature compensation capability.................................. 129
6.3 Rotary displacement sensor .................................................................................. 131
6.4 High electrical voltage sensor............................................................................... 132
Chapter 7 Conclusions and recommendations for future work............................... 134
7.1 Conclusions........................................................................................................... 134
7.2 Suggestions for future works ................................................................................ 137
Reference ....................................................................................................................... 140
Appendix A: Determination of the refraction angle and refractive index
corresponding to the extraordinary waves in the single crystal sapphire........... 146
Appendix B: Acronym.................................................................................................. 148
VITA .............................................................................................................................. 150
vii
List of Figures
Figure 2.1 Schematic design of the sapphire fiber based EFPI sensor……………….... 15
Figure 2.2 Interference in the FP cavity formed by multimode sapphire fibers……….. 16
Figure 2.3 Output spectrum from the EFPI sapphire fiber sensor with cavity
length=6.5µm…………….…………….…………….…………….………. 16
Figure 2.4 Output spectrum from the EFPI sapphire fiber sensor with cavity
length=21.8µm…………….…………….…………….…………….……... 16
Figure 2.5 Principle of an interference device using polarized light waves…………… 17
Figure 2.6 Conceptual schematic design of the sensing head: broadband polarimetric
differential interferometry (BPDI) ………………...………………………. 18
Figure.2.7 Measured output optical interference signal ) (λ I from the BPDI sensor … 19
Figure 2.8 Illustration of the principle of the SCIIB fiber optic signal processor……... 21
Figure 2.9 Illustration of a semi-linear operating range of the interference fringes…… 22
Figure 2.10 Basic structure of optical fiber sensor with white light interferometry……23
Figure 2.11 Normalized optical interference fringes for different OPD values
(with γ =1).…………………………………………………………………24
Figure 2.12 Theoretical intensity curves of the blackbody radiation above 1000
o
C
for the wavelength range: 800~900nm. …………………………………… 26
Figure 2.13 Relation between blackbody radiation and interference fringes at
different temperatures. …………………………………………...………... 26
Figure 2.14 Amplitude response of the IIR filter……………………………………… 27
Figure 2.15 Interference curve with blackbody radiation (at 1350
o
C) subtraction…… 27
Figure 2.16 Temperature measurement systems based on designed sensing
elements and detecting units..………………………………………………27
Figure. 2.17 Schematic design of the single-crystal sapphire based BPDI optical high
temperature sensor……………………………………………………..…... 29
Figure. 2.18. Interfering mechanism for polarimetric differential interferometer…….. 31
Figure 3.1 The approximate transmission bands of standard- and UV-grade sapphire
(window thickness of 0.039 inch). …………………………………………38
viii
Figure 3.2 Hexagonal inner structure of single-crystal sapphire material……………... 38
Figure 3.3 A sensing probe with Sapphire Disk Zirconia Prism (SDZP) structure……. 43
Figure 3.4 Optical propagation in the special crystallographic oriented single crystal
sapphire right angle prism. …………………………………………………46
Figure 3.5 Single crystal sapphire right angle prism…………………………………... 47
Figure 3.6 Sapphire Prism (SP) structure sensing probe………………………………. 47
Figure 3.7 Single channel spectrum domain white light system…………….…...……. 51
Figure 3.8 Block diagram of the implementation of the OPD calculation algorithm….. 56
Figure 3.9 Blackbody radiation subtraction……………………………………………. 59
Figure 3.10 Electrical square wave used to modulate the LED……………….……….. 59
Figure 3.11 Output optical signal from the modulated LED………………….……….. 60
Figure 3.12 The mass-centroid method for peaks locating in the spectrum…………… 61
Figure 3.13 BPDI system overview……………………………………………………. 62
Figure 4.1 Gaussian spectral intensity profile from a low coherence sourece (LED) … 65
Figure 4.2 Central wavelength of the LED dependence on the temperature…………... 66
Figure 4.3 Simulated results of the normalized spectrum with center wavelength
shift effect…………..……………….……………….…………………….. 68
Figure 4.4 Measurement results of the normalized spectrum at 45°C (the reference
is acquired at 20°C) ……………..……………….……………….……….. 68
Figure 4.5 OPD measurement uncertainties caused by center wavelength shift
(δλ) (γ=1, λ
p
= 857nm, L =30um,w=60nm)….……………..……………… 70
Figure 4.6 Temperature uncertainty caused by center wavelength shift (δλ)
(γ=1, λ
p
= 857nm, L =30um,w=60nm) …..……………..………………….. 71
Figure 4.7 OPD measurement uncertainties caused by the bandwidth changes
(δw) (γ=1, λ
p
= 857nm, L =30um,w=60nm) ....……………..………………73
Figure 4.8 Temperature uncertainty caused by the bandwidth changes (δw)
(γ=1, λ
p
= 857nm, L =30um,w=60nm) …..……………..……………..……73
Figure.4.9 Misalignment between the optical polarizer and the sensing element. ……. 74
Figure 4.10. Decomposition and interference of the linearly polarized input
light in the polarimeter. ....……………..……………..…………….………75
Figure 4.11 Visibilities dependence on the misalignment angle α between the light
ix
polarization direction and principal axes of the sensing element………….. 76
Figure 4.12 Experimental interferograms with different visibilities. …………….…… 77
Figure 4.13 Measured temperature deviation caused by the changes in the
visibility of the interferogram…………..……………..……………………77
Figure 4.14 Method of the index ellipsoid to determine the refractive indices
of light beams propagating along s direction. .…..……………..………….. 78
Figure 4.15 The refractive index ellipsoid for single crystal sapphire………………… 79
Figure4.16 Optical refractive index ellipse for the rotation about the s-axis
of the sensing element.………..……………..……………..……………….80
Figure 4.17 Refractive indices and birefringence vs. wavelength for the
sapphire sensing element.……..……………..……………..……………… 81
Figure 4.18 (a) Refraction angles of the ordinary wave and the extraordinary
waves (rotation about the s-axis).
(b) The refraction angle difference between the ordinary wave
and the extraordinary wave (rotation about the s-axis) ..…..……………. 82
Figure.4.19 Sensing element rotation effect on the optical path difference
(rotation about the s-axis). ….……..……………..……………..……. …82
Figure4.20. Sensing element rotation effects along s-axis on the OPD
measurements. …………..……………..…………………..……………. 82
Figure 4.21. Optical refractive index ellipse for the rotation about the f-axis
of the sensing element…………..……………..……………..………….. 83
Figure 4.22. Refractive index vs. light incident angle for the extraordinary waves.….. 84
Figure 4.23. (a) Refraction angles of the ordinary wave and the extraordinary
wave (rotation about the f-axis)
(b) The refraction angle difference between the ordinary wave
and the extraordinary wave (rotation about the f-axis) ……..……………85
Figure 4.24. Rotation effects on the optical path difference (rotation
about the f-axis) …………..……………..………………..……………... 85
Figure 4.25 Effect of rotation about the f-axis on the OPD measurements. ………….. 86
Figure 4.26 Rotation about the 45º axis relative to the s-axis of sensing element. …….87
Figure.4.27 Effect on the OPD measurements of rotation about the axis at 45º
x
relative to the f-axis of the sensing element. ……………..………………87
Figure 4.28 CCD array for the optical signal detection in the optical spectrometer…... 88
Figure 4.29 Opto-electronic noise effects on the valley point locations
determinations..……………..……………..……………..…………………90
Figure 4.30 Opto-electronic noise effects on the temperature
measurement uncertainties…………..……………..……………..……….. 91
Figure 4.31 The measured interference spectrum after dark current subtraction……… 94
Figure 4.32 Simulated result for spectrum curve smoothing with small
window width (w=5) ……………..……………..……………..………….. 95
Figure 4.33 Simulated result for spectrum smoothing with a large window
width (w=80) ……………..……………..……………..………………….. 95
Figure 4.34 Measured spectrum after curve-smoothing with a window
width (w=61) ……………..……………..……………..………………….. 96
Figure 4.35 Optical power losses in the BPDI system. ……………..…………………. 99
Figure 5.1. Output power levels of the LED with different driving currents………….. 104
Figure 5.2. Output spectra of the LED with different driving currents. ………………. 104
Figure 5.3 Normalized interference fringes for the LED with different driving
currents……………..……………..……………..……………..…………...105
Figure 5.4 Temperature deviations vs. optical source output powers…………………..105
Figure 5.5 Normalized interference fringes for transmission fiber with different
attenuations……………..……………..……………..…………………….. 106
Figure.5.6 Temperature deviation vs transmission fiber loss………………………….. 107
Figure.5.7 Experimental results of temperature compensation with the first order
approximation……………..……………..……………..………………….. 108
Figure 5.8 Normalization of the optical spectrum……………..………………………. 109
Figure 5.9 Blackbody radiation effect on the measurements of optical
temperature signatures………………..……………..……………………….. 109
Figure 5.10 Normalized optical spectra measured at different high temperature
levels……………..……………..……………..……………..………………. 110
Figure 5.11 Temperature acquisition subsystem for the calibration purpose………….. 111
xi
Figure 5.12 Real time temperature is related to an OPD value through a GUI
interface………………..……………..……………..……………..…….……112
Figure 5.13 Applied temperature during the sensor calibration process………………. 113
Figure 5.14 OPDs measured with the BPDI system during the sensor calibration
process……………..……………..……………..……………..…………….. 114
Figure 5.15 The BPDI sensor system calibration curve for a sapphire sensing
disk with thickness of 1.5mm…..……………..……………..………………. 114
Figure 5.16 The BPDI sensor system calibration curve for a sapphire sensing
prism (equivalent to a sensing disk with thickness of 8mm) ..…………... 115
Figure 5.17 Repeatability testing results of the temperature measurements…………... 116
Figure 5.18 Deviation of the measured temperatures with respect to the reference
data……………………..……………..……………..……………………116
Figure 5.19 Optical sensing system measurement results vs a B-type
thermocouple measured temperatures. ……………..…………………… 117
Figure 5.20 Deviation between the temperature measurement results from the
B type thermocouple and the optical sensing system….…..…………….. 117
Figure 5.21 Long-term stability testing results……………..………………………….. 120
Figure 5.22 Histogram of temperature measurement……………..……………….…... 121
Figure 5.23 Hysteresis of the BPDI temperature sensor……………..…………………122
Figure 5.24 Rise time characterization of the BPDI sensor system with boiling
water……………..……………..……………..……………..……………126
Figure 5.25 Rise time characterization of the BPDI sensor system with
high temperature furnace..………..……………..……………..…….…... 126
Figure 6.1 An one-end structure version BPDI sensor system ……………..…………. 128
Figure 6.2 A two-end structure version BPDI sensor system……………..……….…... 128
Figure 6.3 Schematic design of BPDI based optical single crystal sapphire
high-pressure sensing head……………..……………..…………………… 130
Figure 6.4 Applied pressure signals on the sapphire sensing element………………… 130
Figure 6.5 Pressure signals from the optical sensor……………..…………………….. 130
Figure 6.6 Pressure measurements with the calibrated sapphire pressure sensor….….. 130
Figure 6.7 Fiber Optic rotation sensor……………..……………..…………………… 131
xii
Figure 6.8 Rotation angle measurements with the calibrated fiber optic
rotation sensor. .…………..……………..………………………………… 132
Figure 6.9 Sensing head designs for electrical voltage measurement. ………………... 133
Figure 6.10 Electrical voltage sensor based on the BPDI technology. ……………….. 133
xiii
List of Tables
Table 3.1 The key features of the optical spectrum analyzer USB2000……………..50
Table 4.1 Noise in the BPDI sensor system and compensation methods..…………..93
1
Chapter 1. Introduction
1.1 Background information on the proposed research
As one of the seven basic quantities used in the SI (International System of Units)
system, temperature is probably the most measured physical parameter since virtually
every process in nature and in industry is temperature dependent. Its accurate
measurement is essential for the safe and efficient operation and control of a vast range of
industrial processes. Appropriate techniques and instrumentation are needed depending
on temperature measurement requirements in different industrial processes and working
environments.
The motivation of this research is to meet the recent increasing needs for temperature
sensors capable of operating accurately and reliably in harsh environments, such as coal-
based power generation and distribution industries, nuclear power industries, glass and
metal manufacturing and processing industries and other high-temperature chemically
corrosive environments. In general, optical sensors have many advantages over
conventional electronic sensors for applications in hash environments. These include:
small size, light weight, immunity to electromagnetic interference (EMI), resistance to
chemical corrosion, avoidance of ground loops, high sensitivity, large bandwidth,
capability of remote operation, and potential capability of operating at high temperatures
[1]. These advantages have promoted worldwide research activities in the area of optical
fiber sensor technologies for harsh environments.
The sensor prototype developed in this research is intended for non-intrusive, direct high
temperature measurement, in the primary and secondary stages of slagging gasifiers.
These gasifiers are used in the coal-based power generation industries, where a sensor is
required to withstand extremely harsh environments imposed by the high temperature,
high pressure and corrosive chemical materials. Sponsored by the National Energy
Technology Laboratory of the U.S. Department of Energy, the Center for Photonics
Technology (CPT) at the Bradley Department of Electrical and Computer Engineering
Yibing Zhang Chapter 1. Introduction 2
of Virginia Tech is currently leading the effort in developing this optical sensor
technology for real-time high temperature measurements, in collaboration with an
industrial partner, Global Energy Technology, Inc.
1.2 Review of non-optical techniques for temperature measurements
In science, temperature is defined in terms of the amount of heat transferred in a Carnot
cycle [2]. This is generally not the most practical way of measuring temperature, and in
practice many different techniques are used depending on the temperature measurement
requirements. In practice, every temperature measurement involves the use of certain
calibrated transducers to convert a measurable quantity into a temperature value. Those
transducers convert changes in the temperature into other measurable physical quantities,
such as volumetric expansion (liquid-filled thermometer), dimensional change (bimetallic
thermometer), electromotive force (thermocouple), resistance (resistance temperature
detector - RTD), radiated energy (radiation thermometer), or some other characteristics of
a material that varies reproducibly with temperature [3-4].
Thermocouple
For high temperature measurements over 1000
o
C, the existing non-optical measurement
techniques are very limited. Possible choices include high temperature thermocouples
and acoustic methods.
1.2.1 High temperature thermocouples
A thermocouple is an assembly of two wires of different metals joined at one end, the so-
called hot end that will physically locate at the measuring position, and at the other end,
the cold junction, that usually works as a reference at 0
o
C. The open circuit voltage from
these two wires, generated by an electromotive force (EMF) inside the wires, depends on
the difference in temperature between the hot and the cold junctions as well as the
Seebeck coefficient of the two wire metals. German physicist J.T. Seebeck discovered
this thermoelectric effect phenomenon in 1821, so the voltage is accordingly called
Yibing Zhang Chapter 1. Introduction 3
Seebeck voltage. Through the known cold junction temperature, the temperature of the
hot junction can be obtained from the measured Seeback voltage.
There are more than 300 types of thermocouple available, which are formed by different
metals. Eight types of them are standardized [4], including T, E, K, J, N B, R, S type
thermocouples. Different type thermocouples are characterized to measure temperature
up to certain levels with different resolutions. For high temperatures over 1000
o
C, Type
B, R, S thermocouples are commercially available. In chemically corrosive environments,
high temperature thermocouples that utilize precious metals are used and have a limited
life of only a few days because of their susceptibility to attack from corrosive chemicals.
They drift significantly under high temperature environments for a long-term operation.
1.2.2 Acoustic methods
It is well known that the speed of sound in a material depends on the temperature [5-6].
Temperature can thus be measured by detecting the speed of sound that propagates inside
a material. This technology is especially useful for measuring gas temperature in a
combustion chamber, where it is difficult to measure the temperature using inserted
probes due to low thermal mass and low conductivity of gases, and the strong radiation
coupling of the walls of the enclosure to the sensor at high temperatures. Using the gas
itself as the temperature sensor overcomes these problems.
The main difficulty with the technique is that the speed of sound is strongly dependent on
the composition of the gas along the path, which is generally not constant in the
combustion chamber. Furthermore, soot particles slow the acoustic wave significantly,
and will result in a large error. Another problem arises from refraction of the sound wave
front by the density and temperature gradients in the chamber. These are often turbulent
which distort the wave fronts, making the accurate determination of time of flight
difficult. Since the sound wave travels with the gas, the apparent speed of sound will also
be strongly affected by the flow velocity through the Doppler effect. Practical systems
usually compensate for this effect by performing measurements in both directions. The
Yibing Zhang Chapter 1. Introduction 4
temperature measurement uncertainty claimed for this method is typically 30
o
C at 1000
o
C [7].
1.3 Review of optical techniques for temperature measurements
Optical sensors are those instruments in which optical signals are changed in a
reproducible way by an external physical stimulus such as temperature, pressure, strain,
etc. An optical beam is characterized by several variables, such as intensity, spectrum,
phase and state of polarization. Many different physical phenomena related to these
characteristics are used to perform sensing functions.
Temperature sensors probably constitute the largest class of commercially available
optical sensors. Besides bulk optics based optical sensors, a wide variety of temperature
sensors using fiber optics have been developed [8]. They offer several significant
advantages over electric sensors, such as small size, light weight, immunity to
electromagnetic interference, etc. The main existing techniques for optical thermometry
are remote pyrometers (or radiation thermometers), thermal expansion thermometers,
fluorescence thermometers, and thermometers based on optical scatterings including
Raman scattering and Raleigh scattering.
1.3.1 Remote pyrometers
All materials with temperatures above absolute zero degree emit electromagnetic
radiation (thermal radiation) and the amount of thermal radiation emitted increases with
temperature. The measurement of the amount of thermal radiation emitted by a material
can therefore be used as an indicator of its temperature. The basic operating principle of
the radiation thermometers is to measure part of the thermal radiation emitted by an
object and relate it to the temperature of the object using a calibration curve that has been
determined either experimentally or theoretically (from Planck’s law) [9]. Typical
radiation thermometers measure temperature above 600
o
C, this type of instrument
dominates the temperature measurement instrument market for temperature over 2000
o
C.
Yibing Zhang Chapter 1. Introduction 5
Dils [10] proposed a fiber optic version of radiation thermometer. In this sensor the end
of the fiber is used as a small blackbody cavity, by shearing it at an angle and coating it
with a metal, such as platinum, for use as a radiation thermometer. Since the fiber
diameter can be very small (<1mm) and the sensor diameter is the same as that of the
fiber, such a fiber sensor can measure temperature from 600
o
C up to 2000
o
C with a fast
time response. Adam achieved kHz frequency response with this structure [11].
1.3.2 Thermometers based on thermal expansion
These are sensors that use the temperature dependence of the optical path length in a
small optical resonator cavity, i.e a Fabry-Pérot (FP) interferometer in an optical fiber
[12]. These temperature sensors measure the change in optical path length of a short
piece of material whose thermal expansion coefficient and refractive index as a function
of temperature are known. The temperature measurement range is dependent on the
fabrication materials. Silica fiber based FP temperature sensor has been demonstrated for
temperature measurement up to 800
o
C, while single crystal sapphire fiber based FP
temperature sensor has been demonstrated for temperature measurement up to 1500
o
C
[13].
1.3.3 Fluorescence thermometers
This type of sensor measures temperature by detecting the decay time or intensity of a
UV stimulated visible fluorescence pulse, which is temperature dependent. In optical
luminescence based fiber sensors, a UV light source is focused into an optical fiber, thus
illuminating a small sample of luminescent phosphor at the end of the fiber. The decay
time or intensity of the UV stimulated visible fluorescence pulse is guided along the fiber
and detected to measure temperatures. Some of the most commonly used phosphors are
made using a rare-earth element, such as Gadolinium or Europium, doped into a ceramic
crystal, such as yttria (Y
2
O
3
) [14-16]. The strong temperature dependence of the
luminescence allows temperature measurement with good accuracy [17]. Such sensors
have found applications in the measurement of temperatures within microwave ovens, or
in very high-magnetic field regions. Those sensors suffer from background noise and
Yibing Zhang Chapter 1. Introduction 6
require complex electrical detecting units for the weak intensity ultrashort optical pulse
signals.
1.3.4 Thermometers based on optical scattering
Another class of optical thermometers employs the temperature dependence of scattered
light [18]. Rayleigh scattering, resulting from scattering of light by particles smaller than
the wavelength of light, depends on both the size and number of scatters present and it is
these relationships that enable the effect to be used in a thermometer [19-20]. Raman
scattering [21], however, results from the scattering of light off phonons, or vibrational
modes of the crystal, and results in two wavelength-shifted scattered light signals. One is
called “Stokes scattering”, at a longer wavelength than the incident light, resulting from
phonon emission, and the other, at a shorter wavelength, resulting from phonon
absorption, is called “anti-Stokes scattering”. The intensity of the anti-Stokes scattering
depends strongly on the number of sufficiently energetic optical phonons in the crystal,
which is a strong function of temperature. The ratio of the anti-Stokes to Stokes
scattering is thus a sensitive indicator of temperature.
1.4 Industrial needs for novel high temperature sensors
The majority of commercially available temperature measurement instruments are made
using only a few basic types of instruments: liquid-in-glass thermometers, thermocouples,
resistance thermometers and radiation thermometers. These conventional and traditional
measurement instruments have been in use for several decades. The major sources of
their instability or drift, as well as possible systematic errors, are well understood. As
matured technologies, they are also generally unexpensive, and there are large ranges of
suppliers offering equivalent plug compatible equipment.
Although these conventional instruments are widely available for scientific and industrial
applications, innovation and research and development activities in temperature
measurements need to be pursued, because commonly used traditional techniques are not
suitable for specific measurement problems, or the performances attainable by these
traditional techniques concerning measurement sensitivity, accuracy and range are
Yibing Zhang Chapter 1. Introduction 7
limited. The specific measurement requirements become necessary in today’s modern
industrial environments, including low thermal mass sensors for fast time response, gas
temperature measurement, measurement of internal temperature profiles, measurements
in hostile environments, e.g., EMI (Electromagnetic interference), radiation, corrosion
and intrinsic safety, where the measurement environment should not be affected by the
measurement units.
One representative harsh environment is an entrained flow slagging gasifier, which is one
of the main units among coal gasification facilities. In the new emerging coal-fired power
plants for advanced power generation, the coal gasification technique [22] is developed to
generate extremely clean electricity and other high-value energy products. Rather than
burning coal directly, coal gasification reacts coal with steam using carefully controlled
amounts of air or oxygen under high temperatures and pressures. A gasification-based
power plant uses the hot, high-pressure coal gases exiting a gasifier to power a gas
turbine. Hot exhaust from the gas turbine is then fed into a conventional steam turbine,
producing a second source of power. This unique integrated gasification combined cycle
(IGCC) configuration of turbines offers major improvements in power plant efficiencies
compared with conventional coal combustion. To optimize performance for these IGCC
plants, certain important physical parameters should be monitored and controlled
precisely for coal gasification processes [23], such as real-time accurate and reliable
monitoring of temperatures at various locations in a coal gasifier, pressure distribution
monitoring in a gasifier, burning material flow patterns inside the gasifier control and
monitoring, air or oxygen monitoring, etc.
The gasifier must be operated at a temperature high enough for the ash in the fuel, such as
coal, to melt and become sufficiently fluid to flow out of the gasifier through the bottom
tap-hole. Load changes will also affect the temperature in the gasifier and downstream,
and would require adjusting the operating conditions. Operating at a too low temperature
would cause the molten slag to become viscous or freeze, plugging up the tap-hole and
preventing additional slag from draining out of the gasifier. Eventually the gasifier has to
be shut down, cooled off, and the slag, now in the form of a hard vitreous rock, has to be
Yibing Zhang Chapter 1. Introduction 8
manually chipped out to be removed from inside the gasifier. The shut down to clean up
the slag may take weeks, resulting in a lengthy loss of production. On the other hand,
operating at a too high temperature would significantly shorten the lifetime of the
refractory lining. In addition, more of the alkali species in the ash would be volatilized,
reacting with the ash particles entrained in the gas to form low temperature eutectics
which deposit in the cooler sections of the gasifier or on downstream equipment, such as
the boiler, causing plugging problems. Operating the gasifier at a too high temperature
will also reduce the conversion efficiency of the gasification process in the production of
the synthesis gas.
In order to realize the full economic potential of the gasification systems, there is an
increasing need to utilize a wide variety of feedstock in addition to coals, such as
biomass, refuse, wood wastes, etc., in the gasification plants. The ash properties of these
various feedstocks vary significantly, and would thus require operating the gasifier at
different temperatures to facilitate slag tapping. The flow condition inside the gasifier is
highly turbulent, with many pockets of gas recirculation zones. Temperature in various
regions of the gasifier could be widely different. The temperature at the exit of the
burners could be above 1927
o
C (3500F), whereas close to the wall temperature could
drop to less than 1316
o
C (2400F). Real-time accurate and reliable monitoring of
temperatures at various locations in a gasifier is thus highly desirable.
Various methods for measuring temperature in harsh environments have been
investigated in the past [24-28]. Among these are optical and acoustical pyrometers, and
high temperature thermocouples. However, due to the harsh environment involving
entrained molten slag, high temperatures and pressures, and corrosive gases, conventional
sensors and measurement devices are very difficult to apply [29-33].
In the non-contact optical pyrometers, an infrared transparent high temperature window
on the gasifier wall is necessary to maintain a large pressure differential, while allowing
transmission of the infrared radiation emitted by the product gases to the detector placed
outside the gasifier. Obstruction of the sight-path opening for the pyrometers in the
Yibing Zhang Chapter 1. Introduction 9
refractory wall by molten slag is a major problem. Shifting of the refractory lining, which
could have a thickness of 2 feet or more, during heat-up and operation processes could
also cause blockage of the sight path. In addition, the measurement is also subject to
interference from the radiation emitted by entrained particles or the relatively cold
refractory walls.
Acoustic pyrometers, which deduce gas temperature along a line of sight by measuring
the speed of sound along that line, have also been evaluated for use in gasifiers. Noise in
the plant or gasifier, such as from the high velocity burners, soot-blowers, etc., interferes
with acoustic temperature measurements.
Current high temperature thermocouples that utilize precious metals drift significantly
and have a limited life of only a few days. Slag buildup around the thermocouple is also a
problem, forming an insulating layer on the thermocouple.
Direct contact temperature measurement is preferred since it will give the measurement
for a specific location. Several of these devices installed at the critical locations inside
the gasifier could provide a temperature profile of the gasifier. Flow patterns inside the
gasifier could thus be deduced from the temperature profile, and performance of the
gasifier could be monitored and improved by making operating adjustments. However,
no direct contact-measuring device is available to date due to material issues. The highly
corrosive molten slag attacks both metals and ceramics. Ceramic materials are also
susceptible to attack from alkali vapors in the gas [34-38]. This situation suggests that
innovative techniques that can operate in the gasifier harsh environment for real-time,
reliable monitoring of temperature be developed.
In this research program, an optical broadband polarimetric differential interferometric
(BPDI) temperature sensor system was proposed and tested using single crystal sapphire
material, which possesses high melting temperature (over 2000°C), superior optical
transparency, and ability to resist chemical corrosion. With a simple mechanically-
structured sensing probe, and optical spectrum-coded interferometric signal processing
Yibing Zhang Chapter 1. Introduction 10
techniques, the single crystal sapphire optical sensor can measure high temperature in
harsh environments with great accuracy, corrosion resistance and long-term measurement
stability.
1.5 Special requirements for temperature sensors in coal gasifier
The optical temperature sensors that can be used in the coal gasifier have to satisfy
several special requirements as explained below.
1. High temperature capability
The operating temperature in the coal gasifier is in the range of 1200
o
C~1600
o
C,
depending on the physical locations in the chamber. The high temperature is the main
reason that renders most electronic sensors inapplicable. Some optical sensors cannot be
deployed because of the limitations of the thermal properties on the fabrication materials.
For example, silica optical fibers can only sustain temperatures up to 800°C before the
dopants start to thermally diffuse. To measure high temperatures accurately in a wide
measurement range with high resolution, proper fabrication materials are needed, as well
as a simple and stable mechanical structure of the sensing probe.
2. High pressure survivability
Pressures as high as 500 psi can be encountered in the coal gasifier chamber. In order to
be able to survive in such high pressure environments, the optical temperature sensing
probe must be designed and fabricated with enough mechanical strength and with its
optical paths entirely sealed to provide the necessary protection.
3. Good thermal stability
The optic temperature sensors designed for the harsh environmental sensing applications
must be thermally stable. Temperature-related degradation mechanisms, including
thermal shock, thermal cycling, thermal stress, thermal fatigue, and high heat flues, must
be considered in the elevated temperature sensor design for long-term stable
measurements. Solid mechanical structures and special fabrication materials are needed
to increase the thermal stability.
Yibing Zhang Chapter 1. Introduction 11
4. Chemical corrosion resistance
With temperatures exceeding 1200°C, pressure exceeding 500 psi, and chemically
corrosive agents such as alkalis, sulfur, transition metals and steam, it is hard to find a
material that is impervious to such an extensive corrosive attack. Conventionally,
commercially available temperature sensors exhibit greatly abbreviated lifetimes due to
the harsh environment. Proper fabrication materials are needed to implement the sensing
probe.
5. Absolute measurement and self-calibration capability
Optical temperature sensors with absolute readouts are much more attractive for
applications in harsh environments because of their no requiring of initialization and
recalibration when the power is switched on. In addition, the sensors are required to have
self-calibration capability so that the guiding fiber loss variations and the source power
fluctuations can be fully compensated, or absolute measurement becomes meaningless.
6. Cost-effectiveness
As the market for optical temperature sensors for harsh environment grows rapidly, the
cost of the sensors and instrumentation is becoming a concern of increasing importance.
In order to achieve successful commercialization, optical temperature sensor systems
must be robust as well as low cost. This requires that the complexity of the sensor system
is kept to the minimum and the technique and process of fabricating sensor probes have
the potential of allowing mass production.
7. Deployability
Optical temperature sensors designed for harsh environment applications must be capable
of remote operation and be flexible enough for easy deployment. Features of mechanical
vibration-proof, high mechanical strength, and remote monitoring and control capability
are thus necessary.
Yibing Zhang Chapter 1. Introduction 12
1.6 Scope of research
In this research, innovative techniques are presented for high temperature sensors capable
of operating at temperatures up to 1600°C. These sensors fulfill the need for real-time
monitoring and long-term direct measurement of high temperatures in harsh
environments.
The efforts devoted to innovative sensing techniques focus mainly on the following
issues.
1) The study of sensing schemes;
2) Cost-effective opto-electronic signal processor design and implementation;
3) Fabrication material selection and mechanical packaging structure design of the
sensing probe;
4) Sensor prototype implementation;
5) Optimization of the sensor prototype and performance evaluation;
6) Expansion of the developed technologies to sensing applications other than
temperature measurement.
The principle of sensing technology is presented in Chapter 2, including sensing schemes,
optoelectronic signal processors for temperature information extraction, as well as a
mathematical model for the temperature measurements. The work of implementing the
sensor prototypes is reported in Chapter 3. Four different optical temperature sensor
prototypes were proposed and tested. The optical broadband polarimetric differential
interferometric (BPDI) temperature sensor system is chosen as an optimal approach for
the high temperature measurement. Chapter 4 is dedicated to the BPDI sensor system
performance analysis and performance enhancement. The experiments and results of the
BPDI optical temperature sensor are presented in Chapter 5. Chapter 6 discusses the
possibility of expanding the application of BPDI technology to sensing applications other
than temperature measurements. Chapter 7 summarizes the research work and proposes
future research directions and improvements.
Chapter 2. Principles of operation of the designed optical
sensors
An optical wave is characterized by several variables such as intensity, spectrum, phase
and state of polarization. Many different physical phenomena related to these
characteristics are used to perform sensing functions. In the designed novel optical
temperature sensors, the temperature information is encoded in the phase, state of
polarization, as well as spectrum of the optical wave. This combined sensing scheme
guarantees the measurement of absolute high temperature with high accuracy, high
resolution, large dynamic measurement range and long-term measurement stability.
An optical measurement system is typically composed of two basic parts. One part is a
sensing element, also called transducer. It is a unit to convert measurand into measurable
physical signals through certain sensing schemes. The other part is a signal conditioning
and processing unit, including optical detecting units, signal amplifiers, supporting
electronic circuits, a signal A/D converter, a computer, computer software for signal
processing calculations and control, and data presentation elements. This chapter presents
the working principles of those separate elements in the optical sensors. Optical sensing
elements utilizing different sensing schemes are described in Section 2.1; the signal
conditioning and processing units are presented in Section 2.2; Section 2.3 describes the
digital signal processing techniques for blackbody radiation background subtraction; the
configurations of the potential temperature sensing systems are compared and the BPDI
sensor system is selected for further prototype development in Section 2.4; the
mathematical model of temperature sensor based on the BPDI technology will be given
in Section 2.5. Finally the advantages of the BPDI sensor system are listed in Section 2.6.
13
Yibing Zhang Chapter 2. Principles of operation of the designed optical sensors 14
2.1 Optical sensing elements design
Because of its high resolution and accuracy, optical interferomerty methods are widely
employed to build optical sensors by encoding measurand information in the phase of the
optical wave through the interference phenomena. There are several different types of
interferometric techniques which have been widely used to measure temperature and
other physical parameters. These are Mach-Zehnder[39], Michelson, Fabry-Perot, and
Sagnac interferometers. Optical fiber Fabry-Perot sensors are highly sensitive to
temperature, mechanical vibration, magnetic fields, and acoustic waves [40-41]. Many
techniques have been used to create intrinsic or extrinsic Fabry-Perot cavities, such as
Bragg gratings in or on the fiber, metal coatings on the end faces of the fiber, and the use
of air-glass interfaces at the fiber ends as the reflectors. Although intrinsic Fabry-Perot
sensors have been used in the past to measure several physical parameters including
temperature, extrinsic Fabry-Perot interferometers (EFPI) have shown their ability to be
insensitive to polarization and only sensitive to axial strain components, which gives
them an advantage over other electro-mechanical and intrinsic Fabry-Perot sensors.
2.1.1 Review of EFPI sapphire fiber sensor
Silica fiber based EFPI sensors have shown great promise for measuring temperatures
(and other physical parameters) below 800°C and have been designed and tested
extensively in the Center for Photonics Technology (CPT). At higher temperatures the
devices degrade rapidly. The degradation mechanisms fall into two general categories:
one is the crystallization of the silica cladding region which degrades the fiber
performance, the other is the diffusion of the germanium dopant (used to establish the
refractive index gradient between the core and the cladding) from the core region into the
cladding, which changes the waveguiding properties of the fiber. By replacing silica fiber
with sapphire optical fibers, the sensors can potentially operate at high temperature (up to
1500°C) environments [13].
As shown in Figure 2.1, a sapphire air gapped EFPI sensor head is formed by inserting
the lead-in single crystal sapphire fiber into the one end of the sapphire tube and inserting
Yibing Zhang Chapter 2. Principles of operation of the designed optical sensors 15
another short piece of sapphire fiber into the other end of the tube to act as a reflecting
surface (target fiber). Both the lead-in and the target sapphire fiber ends are polished to
optical quality. While one end of the target sapphire fiber is highly polished, the other
end is shattered to prevent any significant reflection that would interfere with the
measurements. The light propagating in the lead-in sapphire fiber is partly reflected (7%)
at the first sapphire-air boundary. The transmitted light travels through the air gap and is
also partially reflected (7%) at the end face of the target sapphire fiber. Reflectance R of
the well polished sapphire fiber end surface is approximately 7% as given by the
following equation:
2
]
) ' (
) ' (
[
n n
n n
R
+

(2.1)
for wavelength λ=580nm, n=1.768 is the refractive index of the sapphire fiber, and n’=1
is the refractive of the air.
Figure 2.1. Schematic design of the sapphire fiber based EFPI sensor
Commercially available sapphire fiber is actually a thin sapphire rod using air as a
cladding, which makes the sapphire fiber have a very large modal volume. With these
highly multimode fibers, interference signals are generated with poor visibilities in the FP
cavity formed by the ends of the two sapphire fibers. As shown in Figure 2.2, the
interference between beams from the same mode order (beam 1 and beam 2) are desired
to deduce the cavity length information, while interference between beams from inter-
modes (beam 1 and beam 3, or beam 2 and beam 3) will degrade the interference signals,
resulting in low interference fringe visibilities. When the low quality interference signals
propagate in the multimode fiber, the inter-modal dispersion in the sapphire fiber
degrades the interference fringes further, thus it is very difficult to fabricate high quality
Lead-in Sapphire
Fiber
Target Sapphire
Fiber
Sapphire tube
High temperature
Bonding
Shattered
end
R 1
R 2
Yibing Zhang Chapter 2. Principles of operation of the designed optical sensors 16
sapphire EFPI sensing elements for temperature sensing. In industrial environments
where mechanical vibration exists, all of the possible modes in the sapphire fiber can be
excited and get involved in the interfering process, the more modes in the sensing head,
the lower quality the interference signals.
B e a m 1
B e a m 2
B e a m 3
d
S a p p h i r e f i b e r S a p p h i r e f i b e r
Figure 2.2. Interference in the FP cavity formed by multimode sapphire fibers
The EFPI sensing element was fabricated by Xiao et al with two 100µm diameter sapphire
fibers [13], one single mode silica fiber guides light into the sensing head, and the
sensing F_P cavity is kept on an isolated optical table to avoid exciting higher order
modes in the sapphire fiber. Figure 2.3 shows the output interference spectrum of the
sapphire EFPI sensor with a F_P cavity length of 6.5µm. Figure 2.4 shows the output
spectrum with a cavity length of 21.8µm. As seen in these two figures, when the cavity
length increases, the visibility of the interference signals degrades significantly. Those
low quality interfering signals can not provide accurate and reliable high temperature
measurement in harsh environments.
750 800 850 900 950
-200
0
200
400
600
800
1000
1200
1400
I
n
t
e
n
s
i
t
y
,
a
r
b
i
t
r
a
r
y
u
n
i
t
s
Wavelength (nm)
750 800 850 900 950
-200
0
200
400
600
800
1000
1200
Wavelength ( nm)
I
n
t
e
n
s
i
t
y
,
a
r
b
i
t
r
a
r
y
u
n
i
t
s
Figure 2.3. Output spectrum from the
EFPI sapphire fiber sensor with cavity
length=6.5µm
Figure 2.4. Output spectrum from
the EFPI sapphire fiber sensor
with cavity length=21.8µm
Yibing Zhang Chapter 2. Principles of operation of the designed optical sensors 17
2.1.2 Broadband polarimetric sapphire sensor
Polarization is a fundamental property of light that can be used to measure physical
parameters in a sensing system. The determination of the state of polarization (SOP) in a
system can provide information about photochemical, photo-biological or physical
processes. Techniques employing polarized light are already widely adopted in the
spectroscopic techniques to provide different types of information about atoms or
molecules, such as the electronic structure of the molecules, or orientation of the
molecules [42-43].
An interference system utilizing polarized optical waves generally includes a polarizing
device that is located between an optical polarizer which sets the state of polarization at
the entrance and a polarization analyzer which makes the interfering states of polarization
identical. Figure 2.5 makes the general principle of the system clear. The polarizing
devices such as phase-shifters or retardators split the incident light wave into two
orthogonal linearly polarized waves. Since these two waves propagate at different
velocities in the polarizing device, their phases are shifted by a quantity Φ. After
analyzed by a polarization analyzer, the emerging optical intensity is a function of the
phase shift Φ, which can be used for temperature measurement.
Figure 2.5. Principle of operation of an interference device using polarized light waves
Due to the crystallographic arrangement of the atoms in the single crystal sapphire, the
material exhibits an inherent birefringence. For a wavelength λ=589nm, n
o
=1.768, and
n
e
=1.760, so the birefringence is 0.008. The single crystal sapphire disk can thus be used
as a sensing element in the polarimeter. Figure 2.6. illustrates the working principle of the
sapphire sensing head [44]. One single crystal sapphire disk with inherent birefringence
is sandwiched between a polarizer and an analyzer. The polarization directions of the
Polarizing
interferometric
device
Unpolarized light
Polarizer Analyzer Detector
Yibing Zhang Chapter 2. Principles of operation of the designed optical sensors 18
polarizer and the analyzer are parallel to each other and along the z-axis. The principal
axes (f-axis and s-axis) of the sapphire disk are oriented at 45
o
with respect to the z-axis.
At the exit of the disk, the phases of the two orthogonal states are shifted by a quantity Φ
and the emergent state of polarization is usually elliptical. In order to recover the phase
shift Φ introduced by the disk, the emerging light has to be analyzed through a
polarization analyzer. The phase shift is determined by the magnitude of the
birefringence and also by the length of the sapphire disk traveled by the two orthogonal
linearly polarized light beams. Since both the birefringence and thickness of the sapphire-
sensing element are functions of the temperature, the magnitude of the differential phase
shift will also be temperature dependent. Therefore, by sensing the magnitude of the
phase shift, the temperature can be determined.
O
u
tp
u
t
lig
h
t
y
d
f s
B
ro
a
d
b
a
n
d
in
p
u
t
lig
h
t
P1
P2
Sapphiredisk
x
z
The phase shift between the two orthogonal linearly polarized light waves is determined
by the optical path difference (OPD) between those two linearly polarized light beams,
given by:
d n n OPD
OPD
o e

× Φ λ π / 2
(2.2)
where d is the thickness of the sensing element, n
e
and n
o
represent the two refractive
indices corresponding to the two orthogonal linearly polarized light waves, and λ is the
Figure. 2.6. Conceptual schematic design of the sensing head: broadband
polarimetric differential interferometry (BPDI)
Yibing Zhang Chapter 2. Principles of operation of the designed optical sensors 19
light wavelength. Theoretically speaking, n
e
and n
o
are wavelength dependent, but their
differences are very small in the interested wavelength range from 800nm-900nm. They
are approximated as constant in this wavelength range, thus the OPDs for all the
wavelength are assumed to be equal. The same OPD will generate different phase shifts
at different wavelengths, thus light will experience constructive (corresponding to the
phase shift of 2mπ, m being an integer) or deconstructive (corresponding to the phase
shift of (2m+1)π, m being an integer) interferences depending on their wavelengths.
Utilizing the constructive and deconstructive interference fringes in the broadband
spectral domain, shown in Figure 2.7, the temperature dependent OPD can be obtained by
calculations, and thus temperature information can be obtained by calibration.
Figure. 2.7.Measured output optical interference signal ) (λ I from the BPDI sensor
By encoding and decoding temperature information in the phase of an optical signal, this
sensor design guarantees self-calibration capability, where the encoded temperature is not
corrupted by optical intensity fluctuations from long term operation point of view, hence
better measurement repeatability than intensity based optical sensors.
800 810 820 830 840 850 860 870 880 890 900
0
500
1000
1500
2000
2500
3000
3500
Wavelength (nm)
I
n
t
e
n
s
i
t
y
Yibing Zhang Chapter 2. Principles of operation of the designed optical sensors 20
2.2 Signal processing units
2.2.1 Review of signal processing methods for interferometric sensors
As one of the most sensitive measurement methods, optical interferometry has been used
in a variety of applications. In coherent mono-wavelength interferometric systems, the
most commonly used signal demodulation method is direct interference fringe counting,
i.e. using a power meter to detect the intensity of the interference signal generated by the
sensor head and an electric counter to count the total number of peaks from a reference
point. However, due to the nonlinear and periodic nature of the sinusoidal interference
fringes, the fringe counting method suffers from a number of problems such as sensitivity
reduction when the sensor reaches the peaks or valleys of the fringes, and fringe direction
ambiguity, which practically limited the accuracy of measurement. It was possible to
count fringes bi-directionally by using a quadrature-phase-shifted two-interferometer
structure [45]. However, it was also found difficult to maintain the exact phase difference
between these two interferometers in practical applications.
More recently, Jackson [46], Hogg [47], Steward [48], and Gangopadhyay [49] reported
the use of laser wavelength modulation based heterodyne interferometry to demodulate
the interference signal of EFPI sensors. However, due to the short initial cavity length of
the EFPI sensors, the resolution of the wavelength modulation based signal processing
method was limited.
For the polarization interferometry sensor, intensity based polarization quadrature
measurement can be carried out to detect the phase delay between the two orthogonally
polarized light waves. This method has limited measurement range and suffers from the
complex structure of the detection unit. It thus highly depends on the performance of
those separate components in the detection unit.
In the CPT, there are two signal processors developed to meet the efficient signal
processing needs for interferometric sensors, including polarimetric sensors. They are
self-calibrated interferometric/intensity-based (SCIIB) signal processor, and spectral
domain white light interferometry (also called low-coherence interferometry) signal
Yibing Zhang Chapter 2. Principles of operation of the designed optical sensors 21
processor. The signal processing units can reliably measure the optical path difference
(OPD) that exists between the two interfering optical waves in an interferometer, with
high resolution and high accuracy, as well as self-compensation capability for the optical
source power drifting and fiber loss variations, which can result in a large measurement
error when the signal processing methods based on intensity detection are directly
applied.
2.2.2 SCIIB signal processor
As a unique signal processing method, the self-calibrated interferometric/intensity-based
(SCIIB) [50] fiber optic sensor successfully combines the advantages of both the
interferometric and the intensity-based fiber sensors in a single system. Through a proper
design of the sensor head, the SCIIB technology can provide absolute measurement of
various parameters with the full self-compensation capability for the source power
fluctuation and the fiber loss variations.
Channel 1
Detector
Signal
processing
∆λ
splitter
Channel 2
Detector
λ
Optical signal from
sensing head
Figure 2.8. Illustration of the principle of operation of the SCIIB fiber optic signal
processor
As shown in Figure 2.8, the optical signal from the sensing head is split into two channels
with different optical properties through optical filtering. The light in Channel 1 remains
its original spectral width (broadband spectrum) while the light in Channel 2 has a
narrower spectrum by passing it through an optical bandpass filter. By taking the ratio of
these two channels outputs, the optical source intensity fluctuation and transmission fiber
loss can be fully compensated.
Yibing Zhang Chapter 2. Principles of operation of the designed optical sensors 22
As mentioned previously, regular interferometric sensors suffer from the disadvantages
of sensitivity reduction and the fringe direction ambiguity when the sensor output reaches
the peak or valley of an interference fringe. Sensitivity is reduced at the peak or valley of
the fringe since at that point the change in optical intensity is zero for a small change in
the sensor cavity length. Fringe direction ambiguity refers to the difficulty in determining
from the optical intensity whether the sensor OPD is increasing or decreasing. To avoid
these two problems, the sensor head has to be operated only over the semi-linear range of
a half fringe, as shown in Figure 2.9, so that a one-to-one quantitative relation between
the output intensity and the OPD in the sensor head is obtained. Because of this limitation
associated with the SCIIB signal processor, spectral domain white light interferometric
signal processor is adopted for the high temperature measurement application.
0 0.2 0.4 0.6 0.8 1
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
1.8
2
Change of OPD (um)
S
e
m
i
-
L
i
n
e
a
r
R
a
n
g
e
N
o
r
m
a
l
i
z
e
d
s
e
n
s
o
r
o
u
t
p
u
t
Figure 2.9. Illustration of a semi-linear operating range of the interference fringes.
2.2.3 Spectral domain white light interferometric signal processor
As shown in Figure 2.10, a white light or low coherence interfeometric technique uses a
broadband light source, such as light emitting diodes (LED), multimode laser diodes or
halogen lamps. Its operation principle was first adopted for single mode fiber sensing
systems [51] and then in multimode fiber sensing systems [52]. The white light
interferometric fiber optic sensor system inherits most of the advantages of the
conventional optical interferometers, such as immunity to the light source power drift and
variations of fiber transmission losses, and high resolution.
Yibing Zhang Chapter 2. Principles of operation of the designed optical sensors 23
Broad-band
source
Sensing
intereferometer
Optical processing unit
Electronic processing unit
Optical fiber circulator
Optical fiber
Figure2.10. Basic structure of optical fiber sensor based on white light interferometry
In the spectral domain white light interferometric signal processor, the optical path
difference (OPD) between the two interfering light waves is obtained through
interferometric fringe pattern analysis. For low frequency signal detection (typically less
than 100 Hz), the spectrum measurement method can provide absolute OPD
measurements with high resolution and large dynamic range. Since variations of
temperature in the coal gasifier environments are in the low frequency range, a compact
white light signal processor based on the spectrum measurement has been adopted in this
research.
In the spectrum measurement based white light interferometric system, an optical
spectrum analyzer (OSA) is used as the optical signal processing unit. The OSA is
composed of a grating and a CCD array. The intensities of the dispersed spectral
components of the optical signal are scanned electronically by the CCD array in the OSA.
The optical signal is thus converted into an electrical signal and processed in the
electronic processing unit.
If we assume that the broadband light source (light emitting diode, LED) in Figure 2.10
has a Gaussian spectral intensity distribution given by
( )
( )
( )
]
]
]


− −

2
2
0
0
exp
λ
λ λ
λ I I
s
(2.3)
where
0
λ is the central wavelength,
0
I is the peak intensity value, and λ ∆ is the source
spectral width. The two-beam interference signal is then given by:
Yibing Zhang Chapter 2. Principles of operation of the designed optical sensors 24
( )
]
]
]

J
J
`
'
(
|
+
λ
π
γ λ
L
I I
s
2
cos 1 (2.4)
where the factor γ is the visibility of the interference fringes, and L represents the optical
path difference (OPD) between the two interference beams. The ideal value of γ will be
γ =1, and it may decrease because of the coherence length of the optical source and the
amplitude difference between the two interference optical waves.
After normalizing the interference spectrum given by Equation (2.4) with respect to the
source spectrum ) (λ
s
I , we have the normalized interference output expressed as
J
J
`
'
(
|
+
λ
π
γ
L
I
n
2
cos 1 (2.5)
It is shown in Equation (2.5) that the output spectrum of the sensor is modulated by a co-
sinusoidal function due to the interference. As shown in Figure 2.11, interference fringe
patterns are different for different OPD values.
750 800 850 900 950 1000 1050 1100 1150 1200 1250
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
1.8
2
Wavelength(nm)
N
o
r
m
a
l
i
z
e
d
i
n
t
e
n
s
i
t
y
L=10µm
L=30µm
Figure 2.11. Normalized optical interference fringes for different OPD values (with
γ =1).
Because the interference fringe patterns are a function of the optical path difference L, the
successful demodulation of this spectral signal can render an accurate and absolute
measurement of the L values, which may be related to the physical parameters, such as
temperature, pressure, and stain. The L measurement can also be related to biological or
Yibing Zhang Chapter 2. Principles of operation of the designed optical sensors 25
chemical parameters as well. It is these relations that make the spectral domain white
light interferomety a powerful tool in the field of optical sensing technologies.
2.3 Blackbody radiation subtraction
As well known, for a cavity with perfectly absorbing (i.e. black) walls at a fixed
temperature T, its interior will be filled with radiant energy of all wavelengths. This
radiation is called blackbody radiation. The intensity of the blackbody radiation becomes
stronger with increasing of the temperature. Since the designed temperature measurement
system will be applied to a temperature environment as high as 1600
o
C, blackbody
radiation will be a strong noise background built upon the optical signal emitted from the
LED, which is the temperature information carrier.
Based on the Plank’s blackbody radiation theory, the radiation intensity is specified by a
distribution function
λ
K , representing radiation emitted into a hemisphere, its unit is
W/cm
2
µm
-1
, such that the total radiation intensity is given by λ
λ
d K


0
and λ
λ
d K is the
intensity due to radiation with wavelength between λ and λ λ d + .
λ
K is a universal
function, depending only on the temperature and the wavelength, not on the size or shape
of the cavity or on the material of its wall.
1
2
5
1


T
C
e
C
K
λ
λ
λ
(2.6)
where C
1
=3.7405×10^(4), and C
2
=1.4388×10^(4), T is temperature in Kelvin [9]. High
temperature sensing systems employ this formula to detect temperature values are called
radiation pyrometers. Figure 2.12 shows the blackbody radiation theoretical intensity
distribution curves at different temperatures.
As shown in Figure 2.13, blackbody radiation background is superimposed to the optical
signals, which are temperature signatures generated by the light from the broadband
optical source, and detected by the spectrometer in the optical temperature measurement
system. These experimental measurement results show clearly that the blackbody
Yibing Zhang Chapter 2. Principles of operation of the designed optical sensors 26
radiation background increases dramatically with the temperature increase, thus the
signal-to-noise ratio (SNR) of the measured spectrum is becoming smaller.
800 820 840 860 880 900
10
8
10
9
10
10
10
11
Wavelength(nm)
N
o
r
m
a
l
i
z
e
d
i
n
t
e
n
s
i
t
y
1000 °C
1200 °C
1400 °C
1600 °C
Figure 2.12. Theoretical intensity curves of the blackbody radiation above 1000
o
C for the
wavelength range: 800-900nm.
0 500 1000 1500 2000 2500
0
500
1000
1500
2000
2500
3000
3500
pixel numbers
Intensity
Optical signal
at 79
o
C
Blackbody
radiation
Background
at 79
o
C
0 500 1000 1500 2000 2500
0
500
1000
1500
2000
2500
3000
3500
4000
Pixel numbers
Intensity
Optical signal at
1247
o
C
Blackbody radiation
background
at 1247
o
C
Blackbody radiation
background
at 1350
o
C
Optical signal
at 1350
o
C
(a) (b)
Figure 2.13, Relation between blackbody radiation and interference fringes at different
temperatures.
To increase the signal-to-noise ratio in the detected optical temperature signatures, the
LED output light is modulated with a certain frequency and converted into AC signals,
while the blackbody radiation can be treated as DC signals. Thus the DC background can
be subtracted with digital signal processing techniques in the computer. One method to
reduce the blackbody radiation is to utilize a band pass digital signal filter [53] (IIR filter,
Infinite-Impulse Response filter) filtering out the desired AC optical signal from the LED
Yibing Zhang Chapter 2. Principles of operation of the designed optical sensors 27
only. The background from blackbody radiation will be blocked as a DC component in
the detected signals. If the frequency used for the LED modulation is 3 Hz, the designed
filter is shown in Figure 2.14, whose pass band covers 2Hz to 5Hz. Using the digital
signal filter, the blackbody radiation background is reduced, as shown in Figure 2.15.
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1/2 frequency (Hz)
Aplitude Response
Figure 2.14. Amplitude response of a IIR Figure 2.15. Interference curve with
filter blackbody radiation (at 1350
o
C) subtraction
2.4 Configurations of designed temperature sensor systems
By employing the previously described optical sensing elements (EFPI sapphire fiber
sensor, and polarimetric sapphire sensor) and signal processors (SCIIB signal processor
EF P I
P olar im etric s tr ucture
(P S )
SCI IB
Lo w-co her ence
inter fer o metr y
( W hitelight s ys tem)
Sensi ng elements Signal pr oc e ss or s
T em pe r atur e
meas ur ement s yst em s
E FP I +SCII B
EF P I+W hitelight syst em
P S+SCII B
P S+whit el ight system
Figure 2.16. Temperature measurement systems based on designed sensing elements and
detecting units
800 810 820 830 840 850 860 870 880 890 900
-500
0
500
1000
1500
2000
2500
3000
3500
4000
Wavelength(nm)
I
n
t
e
n
s
i
t
y
Without filter
with filter
Yibing Zhang Chapter 2. Principles of operation of the designed optical sensors 28
and spectral domain whilte light interferometric signal processor), four temperature
measurement systems rooted in the optical interference metrology can be constructed as
shown in Figure 2.16.
As discussed in the optical sensing unit section, sapphire fiber based extrinsic Fabry-
Pérot interferometric (EFPI) sensing elements are difficult to fabricate and their
interfering signals are usually low quality because of the sapphire fiber’s multimode
nature. This type of optical sensor system has been tried in CPT and proved to be not
practical for temperature measurement up to 1600
o
C. The efforts toward this work is then
focused on the development of a polarimetric structure based sensing systems, which
only utilize bulk optics made of single crystal sapphire materials. Two systems:
PS+SCIIB and PS+white light signal processor are attractive candidates for further
development and evaluation.
The PS+SCIIB system is an optical intensity-measurement based polarimetric sapphire
sensor. Because of the periodic nature of the interferometric signal, the working point and
the working range of the sensor need to be selected properly for the wide dynamic range
high temperature measurement. Because the SCIIB signal processor can only operate
properly in a limited linear working range of the sinusoidal interferometric signals, the
thickness of the sapphire sensing element is one of the keys to control the initial working
point and working range of the measurement system. Birefringence of the sapphire
material, which is a function of temperature, is another factor to be considered for the
sensing element design. The measurement stability is also a concern related to the two-
channel structure in the SCIIB signal processor. The long-term stability at different
temperature environments of each of those two channels is hardly achievable. Although
this system can work through careful sensing element selection and signal processor
design, it is hard to maintain long-term accurate temperature measurements.
The PS+whitelight light signal processor, also named as a broadband polarimetric
differential interferometric (BPDI) sensor system, on the other hand, can solve the
problems associated with the PS+SCIIB system. It is chosen as an optimal approach for
Yibing Zhang Chapter 2. Principles of operation of the designed optical sensors 29
further prototype instrumentation development for high temperature measurement in
harsh environments. Based on its interferomatic sensing scheme, high resolution and
accurate temperature measurement in a very large dynamic range is achievable in this
system. The BPDI sensing system monitors temperature in real-time through measuring
optical spectrum instead of optical intensities. This assures its relative immunity to
optical source power fluctuations and transmission fiber losses, thus providing a high
degree of long-term temperature measurement stability.
Sensor
Head
Optical
fiber
Coupler
Optical
source
λ
∆λ
λ
o
Optical
power
Signal
Generator
Optical
Spectrum
Analyzer
Computer
toFiber
Fiber
Collimation lens
Polarizer
(Analyzer)
Extension
Tube
Sapphire
Tube
Sensing
Element
Right-
angle
Prism
0
1
t
SMAconnector
Figure. 2.17. Schematic design of the single crystal sapphire based BPDI optical high
temperature sensor.
In many industrial applications, it is not practical to use a transmission version of the
polarimetric sensing structure in the sensing probe, because of its two-end structure. A
reflection version polarimetric sensing structure setup, which is an one-end structure is
thus designed, as shown in Figure 2.17, where only one fiber collimator is used to
collimate input light and also re-collect light reflected back from the other end of the
sensing probe. The only optical polarizer is used both as a polarizer for the input light and
a polarization analyzer for the reflected light in the sensor probe, which not only reduces
the cost, but also simplifies the probe mechanical structure. The total sensing probe is
then integrated into one solid long tube, which makes it robust enough to survive the
harsh environments.
Yibing Zhang Chapter 2. Principles of operation of the designed optical sensors 30
In the sensing system, the broadband light from a high power light emitting diode (LED)
is modulated by the signal generator and converted into AC signals for the purpose of the
blackbody radiation background subtraction, then injected into a lead-in multimode
optical fiber and propagates through a fiber bundle to the sensor head. The light is first
collimated by a collimation lens, then converted into a linearly polarized optical beam
and travels across a free space enclosed by a high temperature ceramic tube and a single
crystal sapphire tube to a single crystal sapphire sensing element. The light is then
reflected by a right angle zirconia prism and passes through the sapphire-sensing element
again. When the light exits the polarimeter, the two linear polarized light components
with a differential phase delay are combined along the polarizer direction to interfere
with each other. The output light from the polarimeter is then re-collected by the same
input optical fiber bundle and travels back along the optical fiber to the optical detection
end, which is a fiber optic PC plug-in spectrometer, composed of a grating and a CCD
array. The intensities of the dispersed spectral components of the signal are scanned
electronically by the CCD array with a high spectrum measurement resolution.
Additional signal processing is performed in the computer to extract the temperature
information.
2.5 Mathematical model for BPDI temperature sensor
One way to describe the sensing principle of the optical sensor mathematically is using
the Jones matrix method. This method was invented in 1940 by R.C.Jones [55], and is a
powerful tool for analyzing the properties of transmitted light through a complex optical
system composed of many optical elements. In the Jones matrix method, the state of
polarization is represented by a two-component vector. Each optical element can be
represented by a 2*2 matrix. The overall transfer matrix for the whole optical system is
obtained by multiplying all of the individual element matrices, and the polarization state
of the transmitted light is computed by multiplying the vector representing the input
beam by the overall matrix. This method deals with mathematical formulas and is very
effective for a system with large numbers of optical elements. In this temperature sensor
system with a few optical elements, a more simple method is employed with emphasis on
the physical activities in the sensing probe.
Yibing Zhang Chapter 2. Principles of operation of the designed optical sensors 31
The optical interfering activity inside the polarimeter is shown in Figure 2.18, where
polarization directions of the polarizer and the analyzer are parallel to each other. The
input broadband light is polarized by a polarizer P1. The oscillation direction of the
electric field vector of the light wave is thus along P1 (z-axis). When this linear polarized
light propagates inside a sensing element with optical birefringence, the sensing element
behaves as if it split the incident polarization state into two orthogonal linearly polarized
states with the same amplitudes, oriented along the principal axes of the sensing element:
ordinary ray (along f-axis) and extraordinary ray (along s-axis), resulting from the 45
o
angle alignment between P1 and the principle axes of the sensing element. Assuming no
attenuation in the sensing element, these two rays will propagate independently without
changes in their amplitudes, while there will be relative phase delays between them. At
the end of the analyzer, only light polarized along the z-axis can pass, where the light is
decomposed according to their polarization direction again: the ordinary ray is
decomposed into ray1 and ray2, the extraordinary ray is decomposed into ray1' and ray2'.
Interfering results between ray2 and ray2' will pass the polarization analyzer, since the
polarization directions of the polarizer and the analyzer are parallel to each other.
P1 and P2
Ordinary ray Extrodinary ray
Ray 1
Ray 1'
Ray 2 Ray 2'
x
z
The interfering results can be expressed mathematically as following:
) (
2
1
' 2 ray ray2 out
E E E
r r r
+
(2.7)
where
' 2 ay2
E ,
ray r out
E and E
r r r
are all vectors. The output optical intensity is then:
Figure. 2.18. Interfering mechanism for polarimetric differential interferometer
Yibing Zhang Chapter 2. Principles of operation of the designed optical sensors 32
I=
2
out
E
r
(2.8)
It will not only depend on the magnitudes of the interfering electric fields, but also on the
phase shift Φ between the two electrical oscillation vectors:
)) 2 / ( cos
)) cos( 1 (
2
1
)) cos( 2 2 (
4
1
cos
2
1
)
2
1
( )
2
1
(
2 2
2
2
2
2
2
2 ' 2
2
' 2
2
2
Φ
Φ +
Φ +
Φ + +
ray
ray
ray
ray ray ray ray out
E
E
E
E E E E I
(2.9)
where λ π / 2 n d∆ Φ . Since both d and ∆n are functions of ambient temperatures, I
out
will be a temperature encoded signal. To decode the temperature information, a signal
processing model in the spectral domain for the white light interferometry is needed.
The interference fringes of a broadband light from the polarimeter form a co-sinusoidal
curve after it is normalized with respect to ( ) λ
2
2 ray
E according to Equation (2.9):
) ( cos ) (
2
λ
π
λ
n d
I
n

(2.10)
where the product of d and ∆n is a function of temperature, physically representing the
optical path difference (OPD) between the two orthogonal linearly polarized light waves.
) ( ) ( ) ( T n T d T f ∆ × (2.11)
where,
) ( ) ( ) (
) ( ) ( ) (
0 //
T n T n T n
T T T T d
e o
axis c
− ∆
− ×

α
(2.12)
where
axis c− //
α is the coefficient of thermal expansion along the C-axis of single crystal
sapphire. According to reference [54], for the temperature range 325-950K:
) ' ( , ) 273 ( 10 2578 . 0 ) 273 ( 10 4995 . 0 10 6582 . 0 ) (
2 11 8 5
//
K in unit s T T T T
axis c
− × × + − × × + ×
− − −

α
Yibing Zhang Chapter 2. Principles of operation of the designed optical sensors 33
and the refractive indices (n
o
and n
e
) also depend on the temperature, they are changed
along with temperature in the level of ] [ 10
1 6 − −
K .
From the detected optical spectrum, an internally developed algorithm [66] is employed
to measure the OPD ( ) f T values between the two orthogonal linearly polarized light
beams in the sensing element, which is uniquely related to the differential phase delay
between the two orthogonal linearly polarized light beams. The algorithm used to
calculate ( ) f T from the transmitted interferometric spectrum is described below.
According to Equation (2.10), the normalized interferometric spectrum consists of a
series of maxima or minima, shown as peaks and valleys in the spectrum at certain
wavelengths. These special points have a fixed phase relation. For example, the phase
difference between the two adjacent peaks (or valleys) is 2π. Therefore, by detecting the
spectral locations of the peaks or valleys in the interference spectrum, we can obtain OPD
values by applying Equation (2.10). Assuming the wavelengths of two consecutive peak
points are λ
1
and λ
2

1

2
) in the interference spectrum, their interference orders will be
different by 1 and represented by m and m+1. For λ
1
and λ
2
:
π
λ
π
2
) ( 2
1
m
T f

π
λ
π
2 ) 1 (
) ( 2
2
+ m
T f (2.13)
With these two equations, the interference order number m for the peak related to the
wavelength λ
1
can be determined, as well as the OPD values related to the interference
signals:
2 1
2 1
2 1
2
) ( ) ( ) (
λ λ
λ λ
λ λ
λ


T n t d T f
m
(2.14)
Yibing Zhang Chapter 2. Principles of operation of the designed optical sensors 34
By employing the white light interferometry method to measure the OPD values in the
sensing element, and relating them to the temperature by calibration, this sensor system
takes into account two factors at the same time: one is the dimensional changes of the
sensing element and the other is the refractive indices changes with the temperature.
These two factors completely characterize the sensing element behaviors at different
temperature levels. Thus this design provides a method for reliable absolute measurement
of temperature.
2.6 Major advantages of BPDI sensor system
In addition to the general advantages associated with optical sensors, such as lightweight,
remote operation, immunity to EMI, electrically non-conducting, and chemical inertness,
the BPDI sensor system incorporates a spectrometer with a polarimeter for the purpose of
measuring polarization properties as a function of wavelength, offering the following
major advantages over the other sensors designed for high temperature applications:
1. Absolute measurement in a wide dynamic temperature range.
BPDI technology extracts absolute temperature information by absolute measurement of
the phase delays, which are more attractive in applications for the harsh environments
because of its no requirement of initialization and/or calibration when power is switched
on. The one-to-one relation between the phase delays and the temperatures makes it
possible to measure temperature in a wide dynamic range without ambiguity, which is
usually associated with the periodic nature of the interference signals.
2. Minimized complexity in signal processor.
The BPDI sensor system uses a compact PC-plug in optical spectrum analyzer as an
optical signal processor, and a digital computer for signal decoding to extract the
temperature information. The complexity of the signal processor is then kept to its
minimum level; the stability of temperature measurements is thus increased.
Yibing Zhang Chapter 2. Principles of operation of the designed optical sensors 35
3. Self-compensation capability.
The BPDI sensor system monitors temperatures in real-time through measuring optical
spectrum instead of optical intensity. This guarantees its relative immunity to optical
source power fluctuations and transmission fiber losses, thus providing a high degree of
long-term measurement stability.
4. Deployment flexibility.
The BPDI optical temperature measurement system takes advantages of both optical
fibers and bulk optics to simplify the design of the sensing head. Bulk optics is
convenient for reducing the required tolerances on optical alignment and also for
reducing the sensitivity to mechanical vibrations. The optical fiber can transmit light
over a long distance with small attenuation and are easily implemented in industrial
environments because of their small size, lightweight and immunity to electromagnetic
interference (EMI).
5. Ultra-high sensitivity.
The BPDI sensor system offers high measurement resolution since the essence of the
sensing scheme is based on interferometry.
Chapter 3. Design and implementation of temperature sensor
systems
This chapter will describe the design and implementation of the temperature sensor
systems. As presented in Figure 2.16, four types of temperature measurement systems
rooted in the optical interference metrology can be constructed: EFPI+SCIIB,
EFPI+white light interferometric system, PS+SCIIB, and PS+white light interferometric
system. Based on the comparative evaluation and analysis of the experimental results on
these four different temperature sensing schemes, the PS+white light interferometric
system, also named as broadband polarimetric differential interferometric (BPDI) system,
is chosen for further prototype instrumentation development.
The implementation of the BPDI sensor system includes both hardware and software
design and implementation. Hardware implementation includes fabrication materials
selection for the sensing probe, sensing element structure design and fabrication, sensing
probe mechanical assembly, and signal processor implementation. Software design and
implementation includes algorithm development and implementation for temperature
measurements based on the white light interefrometry, and digital signal processing for
the blackbody radiation background subtraction in the computer. In this chapter, Section
3.1 introduces the properties of the materials for the sensing probe fabrication. Section
3.2 presents the sensing probe prototypes based on the different sensing element
structures. Section 3.3 concentrates on the signal processor implementation, algorithm
and digital signal processing software development issues. Finally Section 3.4 provides
an overview of the implemented BPDI sensor system.
3.1 Fabrication materials for sensing probes
Commercially available conventionally temperature sensors exhibit greatly abbreviated
lifetimes in harsh environments, where high temperature, high pressure, chemical
corrosion, strong electromagnetic interference, and high energy radiation exposure exist
simultaneously. In the primary and secondary stages of a two stage slagging
36
Yibing Zhang Chapter 3. Design and implementation of temperature sensor systems 37
gasifier, the temperature measurement devices are subjected to such extremely harsh
conditions. With temperatures exceeding 1400°C, pressure exceeding 400 psig, and
chemical corrosive agents such as alkalis, sulfur, transition metals and steam, it is hard to
find a temperature sensor that is impervious to such an extensive corrosive attack.
Very limited materials are available to fabricate an optical sensor to be physically present
in such harsh environments for direct temperature measurement. For example, no
polymer materials can survive temperatures up to 1600
o
C; only a few expensive metals,
such as platinum and tungsten, have higher melting points over 1600
o
C, but they can
hardly resist the chemical oxidization, sulfidation, carbonization, and nitridation agents.
Some special alumina ceramics can survive both high temperature and oxidization
environments, but not all of them are optically transparent. Bearing in mind the
requirements on the thermal properties, chemical stabilities, and optical properties of the
materials, single crystal sapphire and fully stabilized single crystal zirconia are very
attractive optical materials for high temperature measurement in harsh environments.
3.1.1 Properties of single crystal sapphire
Unlike other ceramic materials, such as polycrystalline alumina, single crystal sapphire
(Al
2
O
3
) is expected to show superior corrosion resistance because of the elimination of
polycrystalline grain boundaries, which have been proven to be vulnerable to corrosion
attack [38]. Single crystal sapphire is actually single crystal aluminum oxide (α-A1
2
0
3
).
Optical grade single crystal sapphire with very low scattering or distortion is
commercially available and widely employed for infrared optical windows and domes
that must survive in the most demanding environments, such as the head of the laser-
guide missiles in battle fields.
• Optical properties of single crystal sapphire
Light transmission in single crystal sapphire material is better than 80% from a
wavelength of 0.3 microns in the UV band, through the visible range of 0.4-0.7microns to
Yibing Zhang Chapter 3. Design and implementation of temperature sensor systems 38
5.5 microns in the IR band. In Figure 3.1, the transmission band for a sapphire window
with a thickness of 1mm (.039 in) provides a nearly uniform transmission of 85% [56].
The perfection and purity of the UV grade sapphire crystals offer a transmission
capability superior to that of the standard grade sapphire. Sapphire’s lower surface
scattering losses at the IR wavelengths also require less rigorous surface polishing.
Figure 3.1. The approximate transmission bands Figure. 3.2. Hexagonal inner
of standard- and UV-grade sapphire (window structure of single crystal sapphire
thickness of 0.039 inch). material.
Single crystal sapphire, processing hexagonal crystal structure as shown in Figure 3.2,
and has two principal optical axes, along which the refractive indices are different. The
unique crystallographic axis is the C-axis, also called optical axis. The refractive index
along this axis is n
o
. The refractive index along A-axis is n
e
. For single crystal sapphire
n
e
<n
o
, the values of n
e
and n
o
are temperature dependent. They can be obtained from the
following approximate formula at room-temperature [57], the unit of the wavelength λ is
micron (µm):
2 2
2
2 2
2
2 2
2
2
2 2
2
2 2
2
2 2
2
2
072248 . 20
5927379 . 6
1216529 . 0
55069141 . 0
0740288 . 0
5039759 . 1
1
028251 . 18
3414021 . 5
1193242 . 0
65054713 . 0
0726631 . 0
4313493 . 1
1

+

+



+

+


λ
λ
λ
λ
λ
λ
λ
λ
λ
λ
λ
λ
e
o
n
n
(3.1)
Yibing Zhang Chapter 3. Design and implementation of temperature sensor systems 39
• Thermo-optic properties of single crystal sapphire
Since the propagation speed of light in a particular direction in a crystal is a function of
the electron density distribution, and the electron probability density distribution is
related to the spacing between adjacently bonded atoms, thus the spacing between the
atoms affects the speed of light propagation directly. Temperature is one of the main
factors affecting the refractive index of solid materials. The thermal expansion
coefficients of sapphire are different along the A-axis and C-axis. The change in spacing
between the adjacent atoms in each of these directions will be different for a given
temperature change. This suggests that the magnitude of the birefringence of the single
crystal sapphire will be changing with temperature. With the temperature increase, the
atomic arrangements inside the sapphire material are becoming more homogenous, thus
the difference between the refractive indices along the A-axis and the C-axis are
becoming smaller, i.e. the birefringence is becoming smaller, which has been observed in
our experiments.
• Thermal stability of single crystal sapphire
Single crystal sapphire can operate at high temperatures and survive high stress induced
by rapid heating. However, the resistance to thermal stress is limited by a loss of
mechanical strength at elevated temperature [58-62]. At temperature above 400
o
C a
dramatic decrease in the sapphire’s compressive strength along the C-axis occurs due to
the activation of the deformation twinning on the rhombohedral planes.
Various ways to increase the strength of sapphire have been explored, including heat
treatments, and doping and improvement of fabrication techniques. Heat treatments at
1450
o
C for 48 hours in flowing oxygen increase the compressive strength by 60% and
biaxial flexure strength by 45% at 600
o
C. Doping with Mg achieves 160% strength
increase over the undoped and unheat-treated baseline strength [63-64].
Yibing Zhang Chapter 3. Design and implementation of temperature sensor systems 40
• Mechanical properties of single crystal sapphire
On the Mohs scale of hardness, which assigns a unit of 10 to diamond, sapphire is rated
at 9, quartz at 7, and glass at 4.5–6.5 [56]. Sapphire is an ideal material for the windows
that must withstand great pressure or vacuum. In contrast to other available optically
transparent materials, sapphire also offers maximum resistance to abrasion and scoring.
• Chemical inertness of single crystal sapphire
Sapphire’s chemical inertness in the presence of a wide variety of reagents at
temperatures greater than 1000ºC makes it ideal for chemical industrial applications. For
example, silica becomes soluble in hydrofluoric acid at room temperature, but sapphire
exhibits no solubility in alkalies or acids, including hydrofluoric acid. At elevated
temperatures, other acids (e.g., hydrochloric acid and nitric acid) attack silica, but not
sapphire [56].
3.1.2 Properties of single crystal zirconia
Pure Zirconia (Zirconium dioxide) has a high melting point (2700
o
C) and a low thermal
conductivity as well as chemical corrosive resistance. Its polymorphism, however,
restricts its widespread use in ceramic industry. During a heating process, zirconia will
undergo a phase transformation process. The change in volume associated with this
transformation makes the usage of pure zirconia in many applications impossible.
Addition of the appropriate amount of some oxides, such as CaO, MgO, or Y
2
O
3
, into the
zirconia structure results in a solid solution, which is a cubic form and has no phase
transformation during heating and cooling. This solid solution material is termed as
stabilized zirconia, which is a valuable refractory [56].
Because of its cubic inner structure, the refractive index of the single crystal zirconia is
constant, regardless of the crystal orientation, so there is no birefringence associated with
the single crystal zirconia. Without adding additional birefringence to the sapphire
sensing element, it is thus used as a right angle prism reflector behind the sapphire
Yibing Zhang Chapter 3. Design and implementation of temperature sensor systems 41
sensing element at the end of the sensor probe to reflect light back, forming an one-end
probe structure. Another advantage associated with this reflector is that no refraction or
reflection (i.e., no scatter) occurs, when light passes from grain to grain of the cubic
material.
3.2 Implementation of sensing probes
This sensor system is designed to measure high temperatures directly in harsh
environments. The sensing element should thus be physically located in the place where
temperature needs to be measured. Certain protection for the sensing element is necessary
in such aggressive environments, so that the sensor can monitor the temperature with
reliability and durability. Due to its high melting temperature (over 2000°C), superior
transparency, and ability to resist corrosion, single crystal sapphire material is chosen to
be used as both sensing elements and protection materials for the sensing probe
packaging in the designed temperature sensing system.
Following the working principle of the sensor described in Chapter 2, the sensing head
structure can be implemented as either a two-end structure (a transmission version of the
polarimeter) or a one-end structure (a reflection version of the polarimeter). In most
industrial applications, a sensing probe with a one-end structure is more convenient to be
deployed, since all of the necessary optical components can be integrated into one solid
tube, input light and output light are all at the same end of the sensing probe instead of
two separate ends. In this research, two one-end structured sensing probe prototypes were
developed with different sensing element structures. One is the sapphire disk, zirconia
prism (SDZP) structure, and the other is the sapphire prism (SP) structure.
3.2.1 A sensing probe with a sapphire disk, zirconia prism (SDZP) structure
This structure is a very straightforward implementation of the polarimetric sensor in a
reflection mode. In this reflective polarimeter, a light reflector with high temperature
survivability is necessary. In low temperature environments, a mirror is commonly used
Yibing Zhang Chapter 3. Design and implementation of temperature sensor systems 42
as a light reflector. The mirror is usually made by coating one surface of a glass plate
with different materials, typically dielectrics or metals. Such dielectric or metal coatings
can hardly survive the high temperature and chemically corrosive environments, either
because the coating materials degrade at high temperature in the chemically corrosive
environments, or because the coefficient of thermal expansion (CTE) of the materials do
not match well with those of the glass substrate, causing thermal stress on heating or
cooling, so that they cannot stick to the substrate and function as a mirror in the high
temperature/corrosive environments. Instead, a right-angle prism is an attractive design
since the total internal reflection is employed inside to change the light propagation
direction by 180
o
. In the prism, only one type of material is employed, thus the CTE
matching problem is avoided for the high temperature applications.
The SDZP structure of sensing probe is shown in Figure 3.3. In this structure, a sapphire
sensing disk and a zirconia right angle prism are used, and packaged into a solid sapphire
tube for protection. Both the thickness and the birefringence of the sapphire disk change
along with temperature changes, thus the temperature information can be obtained by
measuring the OPD between the two orthogonal linearly polarized light waves in the
sapphire sensing disk. The OPD is related to both the thickness d, and the birefringence
n ∆ of the sensing disk by:
nd d n n OPD
o e
∆ − (3.2)
The polarizer used in this system is an optical grade calcite Glan-Thompson polarizing
prism. The light is first collimated by a collimation lens then converted into a linearly
polarized optical beam that travels across the free space enclosed by a high temperature
ceramic tube and a single crystal sapphire tube to the single crystal sapphire disk (with
thickness of 1.5mm and diameter of 30mm). In the sensing probe, the sapphire disk is
arranged such that the linear polarization direction of the input light is at 45 degrees with
respect to its fast and slow axes. When passing through the sapphire disk, the two linearly
Yibing Zhang Chapter 3. Design and implementation of temperature sensor systems 43
Inner sapphire supporting tube
Sapphire sening disk
Zirconia prism reflector
Outer sapphire protection tube
Alumina Extension tube
packaged with polarizer, optical
fiber collimator
SDZP struture sensor probe
Figure 3.3 A sensing probe with a Sapphire Disk, Zirconia Prism (SDZP) structure
polarized light components along the fast and slow axes experience a differential phase
delay due to the sapphire material birefringence and its thickness. The light, containing
the two orthogonal linearly polarized light components, is then reflected by the right
angle zirconia prism and passes through the sapphire sensing element again, doubling the
differential phase retardation. Since the right-angle zirconia prism has no inner
birefringence because of its inner cubic crystallographic structure, no additional
differential phase delay will be added in the detected optical signals. When light exits the
polarimeter, the two linearly polarized light components with a differential phase delay
are combined along the polarization direction of the analyzer and interfere with each
Yibing Zhang Chapter 3. Design and implementation of temperature sensor systems 44
other. The output light from the polarimeter is then re-collected by the same input optical
fiber bundle and travels back along the optical fiber bundle to the optical detection end
which is a fiber optic PC plug-in spectrometer, composed of a grating and a CCD array.
Additional signal processing is performed in the computer for temperature information
extraction. With the protection of the single crystal sapphire outer tube, the end of the
sensor head is heated in a high temperature furnace up to 1600
o
C and temperature is
measured. The measured results are indicative of high accuracy and high resolution.
An extension tube for the sensor head is necessary to avoid thermal damage to the optical
fiber collimator and the optical polarizer. These components cannot survive high
temperatures over 200ºC. The total sensing tube will thus be about two meters long
together with the extension tube.
3.2.2 A sensing probe with a sapphire prism (SP) structure
The previously developed sensing probe prototype utilized single crystal sapphire disk as
a sensing element and a single crystal zirconia prism as the light reflector in the sensing
location. These materials do not have the same CTEs. This may generate instability for
the mechanical packaging in high temperature environments. This structure also suffers
additional optical reflection losses at the surfaces of the zirconia prism besides the
sapphire disk surface reflection losses. From a cost minimization standpoint, an
additional single crystal zirconia light reflector will cost more. In the new version of the
sensing probe, a novel single crystal sapphire right angle prism, functioning as both a
sensing element and a light reflector, is employed. This single crystal sapphire prism
possesses a special geometrical structure and crystallographic orientation for those dual
functions.
The inherent birefringence property of the single crystal sapphire is needed for the prism
to function as a sensing element. As shown in Figure 3.4 (a), when light passes through
the right angle prism, it will propagate in three different directions. Beam 1 propagates
along JC in the +X direction, along CD in the –Y direction, and along DK in the –X
Yibing Zhang Chapter 3. Design and implementation of temperature sensor systems 45
direction. Beam 2 propagates along LE in the +X direction, along EF in the –Y direction,
and along FM in the –X direction. Different incident light beams encounter different path
lengths in different directions. The crystallographic orientation of the prism is selected as
follows. The base of the sapphire prism is A-plane, while both the smaller sides of the
prism are C-plane, as shown in Figure 3.4(b). With this crystallographic orientation,
when light propagates in the x-direction, such as JC and LE, it experiences the
birefringence generated by both the A-axis and C-axis. Because of the symmetric inner
atomic structure of the single crystal sapphire, when light propagates in the y-direction,
such as CD and EF, it also experiences the birefringence generated by both the A-axis
and C-axis. Therefore the total birefringence will be equal regardless of the paths in
which optical beams propagate. According to the geometric relationships, all light beams
will experience the same length equal to AB when they propagate in the prism. Thus the
OPDs for all the input light waves in the prism are:
n AB n FM EF LE OPD
n AB n DK CD JC OPD
beam
beam
∆ ∆ + +
∆ ∆ + +
) (
) (
2
1
(3.3)
This prism thus functions equivalent to a sensing disk with thickness of the length AB, as
well as a light reflector. Figure 3.5 shows the designed geometrical dimensions of the
prism and three pieces of the novel single crystal sapphire prisms.
1
2
A
B
C
D
E
F
J
K
L
M
X
Y
O
(a) Optical propagation path in the prism
Yibing Zhang Chapter 3. Design and implementation of temperature sensor systems 46
C (Z)
A (Y)
A (X)
C-plane formed by A-A axes
A-plane formed by A-C axes
(b) Crystallographic orientation in the prism
K
M
C D
B
E F
Beam experiencing
birefringence formed by
A(Y) and C(Z)
Beam experiencing
birefringence formed by
A(X) and C(Z)
J
A
L
(c ) Optical path in the sapphire prism is equivalent to the optical path in the sapphire disk
Figure 3.4. Optical propagation in special crystallographic oriented single crystal
sapphire right angle prism.
45
o
45
a
-
a
x
i
s
a-axis 5
.
0
m
m
5
.
0
m
m
220mm
Side view
Bottom view
o
a
-
a
x
i
s
c-axis
(a) Geometrical sizes of the prism
Yibing Zhang Chapter 3. Design and implementation of temperature sensor systems 47
(b) Fabricated single crystal sapphire right angle prisms
Figure 3.5 Single crystal sapphire right angle prism
Inner sapphire supporting tube
Sapphire sening Prism
Outer sapphire protection tube
Alumina Extension tube
packaged with polarizer, optical
fiber collimator
Prism struture sensor probe
Figure 3.6 Sensing probe with a Sapphire Prism (SP) structure
Yibing Zhang Chapter 3. Design and implementation of temperature sensor systems 48
With a single right angle prism at the end of the sensing probe, the mechanical structure
of the sensing head is simplified, as shown in Figure 3.6. The sensing prism is fixed at the
end of the inner supporting tube, in which a slot of the correct size to fit the prism is
machined. The outer protective sapphire tube covers the inner tube and the sensing
prism, with a certain amount of free space between its end and the sensing prism so that
the prism can expand freely when the temperature increases. The total sensing tube is
about two meters long together with the extension tube, which is used to avoid thermal
damage to the optical fiber collimator and the optical polarizer at the other end of the
sensing probe.
3.2.3 Total internal reflection in the sensor probe
The right angle prism works as a light reflector in the sensing probe. We want to analyze
the optical loss at the reflection surface and the changes of the optical wave polarization
states when the light is reflected.
The total internal reflection angle, also called critical angle θc for the single crystal
zirconia –air interface is:
o
zirconia air zirconia c
n n 1 . 28 ) 12 . 2 / 1 ( sin ) / ( sin
1 1

− −

θ (3.4)
Total internal reflection angle for the single crystal sapphire-air interface is:
o
sapphire air sapphire c
n n 7 . 34 ) 76 . 1 / 1 ( sin ) / ( sin
1 1

− −

θ (3.5)
With a 45
o
reflection angle at the prism-air interface, which is larger than the critical
angle, light is totally reflected without loss in those right angle prisms. The light
propagation direction is then changed by 180
o
at the end of the sensing probe by two total
internal reflections in the right angle prisms.
In the prism reflector, light is reflected by the total internal reflection, the Jones matrix
method can be used to analyze the electric field polarization direction changes at the
reflection surfaces [66]. The electric field vector of the input light is related to the electric
field vector of the output light by:
Yibing Zhang Chapter 3. Design and implementation of temperature sensor systems 49
in out
E E
r r
Τ (3.6)
the Jones matrix T is given by
J
J
J
`
'
'
(
|

y
x
T
ρ
ρ
0
0
(3.7)
where,
2 2
0
2 2
0
2 2 2 1/ 2
0
1
1
( sin )
cos
+ α
ρ
− α
+ α
ρ
− α
ψ −
α
ψ
x
y
i
i
n in
n in
n n
n
(3.8)
n is the refractive index of the single crystal sapphire or the single crystal zirconia, n
0
=1
is the refractive index of air, and ψ is the incident angle of light in the prism, which is
45
o
, for light reflected at the critical angle,
) / ( sin
0
1
n n
c

ψ
1
y x
ρ ρ (3.9)
then the Jones matrix reduces to an identity matrix, which means the total internal
reflections in the right angle prism do not change the polarization directions of the
reflected light, thus no additional phase shift will be induced between the two orthogonal
linearly polarized light waves.
3.3 Signal processor and software implementation
3.3.1 Signal processor implementation
An important part of the spectral domain white light interferometry signal processor is
the spectrometer. The performance of the whole system, such as measurement resolution
and stability, are highly dependent on the performance of the spectrometer. In a
traditional spectrometer based system where a cumbersome and expensive spectrometer
is used, the strict requirements to environmental conditions exclude the possibility to use
this kind of instrument in real industrial fields. In the designed system, a compact fiber
Yibing Zhang Chapter 3. Design and implementation of temperature sensor systems 50
optic spectrometer (USB2000 manufactured by Ocean Optics, Inc,) is used. Its key
features are listed in Table 3.1.
Table 3.1. The Key Features of the optical spectrum analyzer USB2000:
Computer interface Universal Serial Bus (RS-232 available on
side connector)
Integration time 3 milliseconds-65 seconds
Data transfer rate Full scans (2048 wavelengths) into
memory every 13 milliseconds
Dimensions: (Length*Wideth*Hight) 3.5" x 2.5"x 1.31"
89 mm x 64 mm x 34 mm
Weight 0.45 lb. without cable
0.60 lb. with cable
Detector 2048-element linear silicon CCD array
Effective range 800-900 nm
Sensitivity (estimate) 86 photons/count;
Signal-to-noise 250:1 (at full signal)
Dark current noise 2.5-4.0 pA (RMS)
Slits optical fiber is entrance aperture
Focal length 42 mm (input); 68 mm (output)
Resolution ~ 0.3 nm-10.0 nm FWHM
Stray light < 0.05% at 600 nm; < 0.10% at 435 nm;
<0.10% at 250 nm
Fiber optic connector SMA 905 to single-strand optical fiber
(0.22 NA)
The input fiber is connected to the spectrometer directly through a SMA connector.
Inside the spectrometer, a 1200 line holographic grating is used to diffract the input light
on a 2048 pixels CCD array. The output electrical signals from each pixel in the CCD
array will be read out serially and converted into digital format by a built-in
analog/digital (A/D) card. The measured spectrum is then transferred into a computer for
further signal processing. Figure 3.7 shows the structure of a single channel spectral
Yibing Zhang Chapter 3. Design and implementation of temperature sensor systems 51
domain white light signal processor, including an optical fiber coupler, an optical
spectrometer, a LED light source and an electrical power supply.
APC
fiber
connector
USB 2000
spectrometer
LED driver
Power source underneath the box
LED
Fiber
coupler
USB interface
to computer
Figure 3.7. Single channel spectral domain white light signal processor
Yibing Zhang Chapter 3. Design and implementation of temperature sensor systems 52
3.3.2 Signal processing algorithm development
According to the working principle of the white light interferometry shown in Equation
(2.5), OPD between the two orthogonal linearly polarized light beams can be obtained
from the measured spectrum fringe patterns. The existing algorithms for the OPD
calculations either possess high resolution with limited measurement range (such as the
peak tracing method), or large dynamic measurement range with limited resolution (such
as the two-point method). A novel algorithm is then adopted for the BPDI sensor system
to achieve both high resolution and large dynamic range simultaneously.
Peak tracing method
Based on the measured spectrum fringes patterns, the value of the OPD can be obtained
through tracing a special point in the interference fringes (such as one peak point). In this
method, the wavelength λ
m
of a peak point in the interference spectrum satisfies the
relationship:
π
λ
π
λ
π
2
2 2
m
n d L
m m


(3.10)
where the spectral order m is a non-negative integer. Equation (3.10) can be rewritten
into the following form:
m
m
m
m
L λ
π
πλ

2
2
(3.11)
To calculate the OPD value L from a special peak wavelength λ
m
, the interference order
m must be known first. For the selected peak, the identification of the interference order
m is difficult so that the unambiguous operating range of the OPD is limited in only half
of the wavelength range. The resolution of this peak tracing method is mainly dependent
Yibing Zhang Chapter 3. Design and implementation of temperature sensor systems 53
on the resolution of the spectrometer as shown in Equation (3.11). The relative
uncertainly of the measurement can be described as
λ
λ ∆


L
L
(3.12)
where ∆λ is the measurement resolution of the wavelength, determined by the
measurement resolution of the spectrometer.
Two-point method
To increase the measurement range of the OPD values, two special points instead of one
point in the interference spectrum need to be used for the absolute measurement. Suppose
λ
1
and λ
2

1

2
) are the wavelengths of the two adjacent peak points in the interference
spectrum. Their interference orders are m and m+1. From Equation (2.10),
π
λ
π
2
2
1
m
L

π
λ
π
2 ) 1 (
2
2
+ m
L
(3.13)
The OPD can be determined,
1 2
2 1
λ λ
λ λ


L (3.14)
The wide dynamic measurement range can then be obtained by using two special
wavelengths, as long as the two such special wavelengths exist (two peaks or two
valleys) in the spectrum pattern, the OPD can then be calculated by Equation (3.14). In
this case, the relative uncertainty induced by the spectrometer is
1
1
1 2
2
λ
λ
λ λ
λ ∆




L
L
(3.15)
If the central wavelength of the light source (LED) is 850nm, for the operating range of
OPD (40-60 µm), this factor
1 2
2
λ λ
λ

is about 110—150. Thus, this method has lower
resolution in a large dynamic measurement range compared to the peak tracing method.
Yibing Zhang Chapter 3. Design and implementation of temperature sensor systems 54
New algorithm
Combining the advantages of the two methods described above, a novel data processing
algorithm has been developed in CPT [66], which can achieve both high resolution and
large dynamic range.
The basic idea of this method is to use two peak points in the interference spectrum to get
a rough OPD value first (large dynamic range is then achieved); this rough OPD is then
used to determine the rough order number m
r
of a special peak point in the interference
fringes. Then, an accurate m
a
value will be recovered from the rough m
r
. From Equation
(3.11), the accurate OPD can be calculated based on the accurate m
a
and the peak
wavelength location in the spectrum (high resolution is thus achieved).
The algorithm used to recover the accurate m
a
is the following:
In Equation (3.11), for a given peak, m is a constant value. For adjacent peaks, the
difference between the interference orders is 1. For example, if m for one peak is 12, then
for the adjacent peaks the interference order will be 13, 14, 15…and 11, 10, 9….By
calibration, the value of m for a special peak (m
o
) can be acquired accurately and stored
in computer. When the rough m
r
value for any peak has been acquired, the accurate m
a
will be obtained from m
o
by adding the integer part in the difference between the rough
m
r
and m
o
. The whole process of demodulating OPD from the interference spectrum can
then be separated into two sub-processes: the calibrating sub-process and the measuring
sub-process.
With the white light interferometer system in a stable condition, the OPD is set to a
known value L
0
. The m
o
of a peak in the interference spectrum can be calculated
accurately from the L
0
and the wavelength λ
0
. This is the calibrating sub-process.
With known m
o
, the measuring sub-process can be carried out as follows:
1. Use the two peak points in the interference spectrum to acquire a rough value of
L
r
using Equation (3.14);
Yibing Zhang Chapter 3. Design and implementation of temperature sensor systems 55
2. Select a peak point near the center of the interference spectrum, then use the
wavelength of this peak and the rough L
r
from previous step to calculate a rough
m
r
';
3. The accurate m
a
can be calculated from m
r
and the stored m
o
from:
) 5 . 0 int( + − +
o r o a
m m m m (3.16)
Where function int(…) means to take the integral part.
4. The accurate OPD can be calculated from the accurate m
a
and the wavelength of
the peak with Equation (3.11).
3.3.3 Software design and implementation
Advanced computer software has been developed to demodulate the OPD values from the
interference spectral patterns, based on the novel data processing algorithm presented
previously. The software program is implemented in the combination of Microsoft Visual
Basic and C++ languages so that both the graphic interfaces and high computational
speed are achieved and optimized. The block diagram of the program is shown in Figure
3.8.
The program is described in detail below:
Initialization
The program starts with the hardware initialization, which includes the initialization of
the spectrometer. Then most of the important parameters, which are stored in the
computer, will be restored in the memory of the computer. These data include the
parameters for setting up the spectrometer such as the integration time, scanning time
interval, the calibrated interference order m
o
value, the reference spectrum of the light
source, the dark current noise of the spectrometer, and the calibration curve of the
temperature sensor.
Yibing Zhang Chapter 3. Design and implementation of temperature sensor systems 56
Initialize white light system
Restore important parameters
Acquire interference spectrum
Data processing
Find coarse peak wavelengths
Use mass-center method to get the accurate
peak positions
Use two peaks to calculate the rough OPD value
Use the rough OPD value to calculate the rough
m
r
value
Use the m
o
stored in PC and the rough m
r
to
calculate accurate m
a
Use the accurate m
a
to calculate the accurate
OPD
Use the accurate OPD and calibrating curve to
demodulate temperature information
Figure 3.8. Block diagram for the implementation of the OPD calculation algorithm
Yibing Zhang Chapter 3. Design and implementation of temperature sensor systems 57
Spectrum acquisition
The interference spectrum from the sensor head will be sampled according to the given
time interval. Once the optical signal is coupled into the spectrometer through a SMA
connector from the optical fiber, the divergent light emerged from the optical fiber will
be collimated by a spherical mirror. Then, a plane grating is employed to diffract the
collimated light. A second spherical mirror then focuses the diffracted light and an image
of the spectrum is projected onto a 1-dimensional 2048 pixels linear CCD array. The
CCD array is actually serially connected reverse-biased photodiodes, they discharge a
capacitor at a rate proportional to the photon flux. When the integration period of the
detector is complete, a series of switches closes and transfers the charge of each pixel to a
shift register. After the transfer to the shift register is complete, the switches open and the
capacitors attached to the photodiodes are recharged and a new integration period begins.
At the same time as light energy is being integrated, the spectrum data is read out of the
shift register by a built-in A/D card. In the last step, the spectral data will be acquired by
a computer through a USB interface.
Data processing
To demodulate the temperature information from the interference spectrum accurately, an
advanced data processing algorithm has been developed and implemented in the
Microsoft Visual Basic environments. This part is the core of the whole program, and
will be discussed in several sub-steps, as data pre-processing for blackbody radiation
background subtraction, peaks and valleys localization and signal demodulation for the
temperature measurement.
Data pre-processing
The original spectrum data from the spectrometer is a 1x2048 array, which composes the
received light signals from the 2048 pixels of the CCD array. To recover the interference
spectrum, the calibrating curve of the CCD array is used to calculate the optical
wavelengths according to the serial number of array. To remove the dark current noise of
the CCD detector and other background noise, a dark spectrum stored in the computer is
Yibing Zhang Chapter 3. Design and implementation of temperature sensor systems 58
subtracted from the measured spectrum. The dark spectrum is pre-acquired when the light
source is not powered.
As discussed in Chapter 2, modulation of the optical signal is needed to convert the
temperature probing light into AC signals, so that the blackbody radiation background
can be subtracted as a DC signal. The straightforward way to implement this modulation
and demodulation scheme is to modulate the optical source with a square wave with
certain frequency, so that the probing light will be on and off following the square wave.
When the probing light is on, the detected signals will contain both the blackbody
radiation and optical signals with temperature information. The measured optical
spectrum is as shown in Figure 3.9 (a). When the probing light is off, only blackbody
radiation background is detected by the spectrometer as shown in Figure 3.9(b). Since the
blackbody radiation background is a slowly changing signal that follows the temperature
changes, it can be treated as constant during the time interval when the probing light is on
and off. Thus, the blackbody radiation can be subtracted, resulting in a spectrum as
shown in Figure 3.9 (c). The square wave used to modulate the LED is shown in Figure
3.10. The frequency of the square wave is 8Hz. The modulated output light from the LED
is shown in Figure 3.11. The light intensity curve is measured with a low quality optical
detector, so the signal noise ratio is poor, which does not represent the real optical signal
intensity levels.
(a) Measured spectrum in the presence of blackbody radiation at 1200
o
C
800 810 820 830 840 850 860 870 880 890 900
1000
1500
2000
2500
3000
3500
4000
Wavelength(nm)
I
n
t
e
n
s
i
t
y
Yibing Zhang Chapter 3. Design and implementation of temperature sensor systems 59
(b) Measured spectrum of blackbody radiation at 1200
o
C
(c) Spectrum after blackbody radiation subtraction at 1200
o
C
Figure 3.9 Illustrating a blackbody radiation subtraction
Figure 3.10 . Electrical square wave used to modulate the LED
800 810 820 830 840 850 860 870 880 890 900
1000
1500
2000
2500
3000
3500
4000
Wavelength(nm)
I
n
t
e
n
s
i
t
y
800 810 820 830 840 850 860 870 880 890 900
-500
0
500
1000
1500
2000
2500
Wavelength(nm)
I
n
t
e
n
s
i
t
y
Yibing Zhang Chapter 3. Design and implementation of temperature sensor systems 60
Figure 3.11. Modulated output optical signal from the LED
Peaks and valleys localization
Desired peak positions in the interference spectrum are selected to demodulate the OPDs
between the two orthogonal linearly polarized light waves. The valley positions in the
interference spectrum can also be located with a similar method. The technique for
locating peaks is described as follows:
The first step is to find the coarse locations of the peaks by a smart comparison
algorithm. The basic idea of this algorithm is to find all the local maximum points in a
special range in the interference spectrum. The size of the searching range, which is
defined as the window size of the peak searching, must be large enough to eliminate the
influence of the noise. To avoid missing some peaks, the window size must be smaller
than the space between adjacent peaks.
After the coarse positions of the peaks are found, a mass-centroid algorithm is applied to
find the accurate positions of the peaks. The basic idea of the mass-centroid algorithm is
shown in Figure 3.12. If the function Y (x) is symmetric around its peak position, then
the peak position coincides with the x coordinate of the centriod of Y (x).
Yibing Zhang Chapter 3. Design and implementation of temperature sensor systems 61
As shown in Figure 3.12, the peak position Xo can be calculated from:
0
( )
( )
xf x dx
X
f x dx
+∞
−∞
+∞
−∞



(3.17)
From Equation (2.5), around the peak point, the spectrum is a symmetrical function of the
wave number k=1/λ. So the spectrum y(λ) needs to be expressed in terms of the wave
number first (from {y
i
, λ
i
} to {y
i
, k
i
}).
Figure 3.12 Mass-centroid method for peaks locating in the spectrum
With the normalized spectrum {y
i
, k
i
}, the accurate peak positions can be calculated from
the coarse peak positions:
1
1
1
( )
( )
1
i
i
i
i i i i
p i
m
i i i
p
m p
m
k
y k k k
k
y k k
k
λ
λ
+∞

−∞



(3.18)
Demodulation of the temperature information
Once the peak positions are acquired, the temperature can be demodulated as described
below:
A coarse OPD is calculated using two peak positions, and then the rough m
r
value for a
selected peak near the center wavelength is determined. Using the calibrated m
o
value
Y=f(x)
X
X
0
Centroid
Yibing Zhang Chapter 3. Design and implementation of temperature sensor systems 62
stored in the computer, the accurate m
a
value is obtained and then is used to calculate the
accurate OPD values. Finally, using the calibration curve, the temperature is extracted
from the accurate OPDs.
3.4 Overview of BPDI temperature sensor system
Based on the BPDI technology, a portable temperature-sensing probe is formed, where
the sensing element is located at the end of the sensing probe. This structure of the
sensing head is simplified by eliminating one polarizer and one light collimator compared
with the two-end structure.
In Figure 3.13, the broadband light from a high power light emitting diode (LED, 480µW
output after coupling into a 200µm diameter pigtail fiber), has a center wavelength at
850nm and its spectrum covers a range from 800nm to 900nm. The modulated light
pulses (represented by yellow blocks) are injected into a 2-meter long lead-in multimode
optical fiber bundle and propagates to the sensor head. The polarizer used in this system
is an optical grade calcite Glan-Thompson polarizing prism. The light is first collimated
by a collimation lens, then converted into a linearly polarized optical beam and travels
across a free space enclosed by a high temperature ceramic tube and a single crystal
sapphire tube to a sensing element.
Figure 3.13. BPDI sensor system overview
Optical fiber coupler
Optica l spectrum analyzer
Optica l pola rizer
Yibing Zhang Chapter 3. Design and implementation of temperature sensor systems 63
The sapphire-sensing element is arranged such that the input linear polarization direction
of light is at 45 degrees with respect to its fast and slow axes. When light passing through
the sapphire sensing element, the two linearly polarized light components along the fast
and slow axes experience a differential phase delay due to the sapphire birefringence and
its thickness. The light, containing the two orthogonal linearly polarized light
components with relative phase delays, is then reflected by a reflector, the reflector can
be a single crystal zirconia right angle prism or a sapphire sensing right angle prism. The
output light from the sensing probe is then re-collected by the same input optical fiber
bundle and travels back along the optical fiber to the optical detection end, which is a
fiber optic PC plug-in spectrometer, composed of a grating and a CCD array. The
intensities of the dispersed spectral components of the signal are scanned electronically
by the CCD array, with a spectral resolution of 0.3nm. Additional signal processing is
performed in the computer. As a temperature reference, a type K thermocouple
(resolution 0.05
o
C) is used for temperatures lower than 200
o
C, and a type B high-
temperature thermocouple (resolution 0.2
o
C under 500
o
C and 0.1
o
C above 500
o
C) is
used for temperatures above 200
o
C. The sensor head is heated in a high temperature
furnace up to 1600
o
C for its performance testing.
The total length of the sensing probe is 2 meters long and fabricated with high
temperature ceramic material with a low thermal conductivity, which is sufficient to
isolate thermal damage of the optical elements (optical fiber, collimation lens and
polarizer) from sensing location (temperature over 1600
o
C). After passing through 4
meters long free space (round trip in the sensing probe), it is usually difficult to re-collect
optical energy into the optical fiber. To solve this problem, a fiber bundle with one fiber
at the center surrounded by six fibers is used. The diameter of each fiber is 200µm. Light
exits from the center fiber, the other six fibers re-collect light reflected back from the
sensing probe, through a diameter of 22mm collimation lens with a 35mm focal length.
About 10% of the input light can be recollected from the sensing probe with this
collimation lens and the fiber bundle, thus the recollected optical power is sufficient to be
detected by the CCD array in the spectrometer.
Chapter 4. System noise analysis and performance
optimization
As an optoelectronic measurement system, the performance of the BPDI sensor is limited
by the noise associated with individual electronic and optical components and their
combined effects. A detailed analysis of these noise effects on the system performance is
a very important design step since it provides a guideline to achieve an optimal system
performance.
As discussed in Chapter 3, the optical components in the BPDI sensor system include an
optical source (LED), an optical fiber, an optical collimator, a polarizer, and an optical
sensing element. In order to achieve the desirable performance of the sensor system, it is
necessary for these optical components to perform their functions accurately.
Unfortunately, environmental perturbations, such as temperature variations and
mechanical vibrations, can easily introduce noise to the system through the interaction
between the optical components and the host media. These noises may cause the optical
components to deviate from their desired functions and result in measurement errors.
Instead of analyzing the noise performance of each of the individual optical components
in the sensor system, we study the all-possible degradation factors on the temperature
signatures, i.e the normalized interference spectrum. This chapter is dedicated to discuss
these degradation factors, including the optical spectrum characteristic variations relative
to the pre-stored optical source spectrum used for the normalization; the visibility of the
normalized spectrum; the mechanical vibrations; transmission fiber generated spectrum
distortion and electrical noise inside the spectrometer. These degradation factors might be
due to noises generated by one single component or combined noises from several
different components. The advantage of this method is to get a systematic view and then
design compensation methods to optimize the system performance.
4.1 Optical spectrum induced noise
The BPDI sensor system measures temperatures through detecting the OPDs in the
64
Yibing Zhang Chapter 4. System noise analysis and performance optimization 65
sensing element sandwiched in a polarimeter in the optical spectral domain. The optical
power fluctuation of the source is no longer a concern in this system, while the spectrum
characteristic changes will affect the OPD measurement thus the temperature
measurement accuracy.
The output spectrum of the LED is pre-measured and pre-stored in a computer as a
reference spectrum for the normalization purposes. Suppose the central wavelength of the
reference spectrum is λ
0
and the spectral bandwidth of the light source is w, then the
spectral intensity distribution of the LED radiation can be approximated as a Gaussian
curve as shown in Figure 4.1:
2
0
0 2
( )
( ) exp( ) I I
w
λ λ
λ

− (4.1)
W avelength(nm)
I
0
I
0
/e
Bandwidth w
0
λ
O
p
t
i
c
a
l
I
n
t
e
n
s
i
t
y
Figure 4.1 Gaussian spectral intensity profile from a low coherence source (LED)
When the ambient conditions change, the wavelength of the LED will drift, i.e the center
wavelength λ
0
is changed, or the spectral bandwidth is broadened or narrowed. Most
likely, the changes of these two factors will occur simultaneously during the temperature
measurement process. To illustrate their effects on the temperature measurement, these
two cases will be analyzed separately.
4.1.1 Wavelength drift
Two components in the BPDI sensor system may induce the center wavelength drifts: the
LED source and the optical spectrometer. The measurements from the optical
Yibing Zhang Chapter 4. System noise analysis and performance optimization 66
spectrometer are temperature dependent, resulting from the temperature dependent
performances of its components (the volume grating and the CCD array). Also
determined by the characteristic of the fabrication materials of the semiconductor optical
sources, the spectrum of the LED is temperature dependent, and the central wavelength
of the LED will shift to longer wavelength as the temperature increases. When the
temperature changes from T
0
to T, the central wavelength of the LED will drift from λ
0
to
λ
T
. In the designed working temperature range from –50~80
o
C, which is normally
experienced by the electronics at the room environments, λ
T
can be expressed in term of
the temperature T:
T a T T a
T
∆ − − ) (
0 0
λ λ δ
λ
(4.2)
where a is a constant, and roughly equals to 0.2-0.3nm/°C.
The temperature dependence of the light source in the BPDI sensor system is measured
experimentally. The light source is kept in an electric oven; the spectrometer and the
sensing head were kept in the room environment. The temperature of the oven is
increased to 50°C, and kept at that temperature for about 2 hours. Then the temperature
was decreased to room temperature slowly (in about 3 hours). During the temperature
decreasing process, a thermocouple is used to monitor the temperature in the oven, and
the output spectrum of the LED is measured by the OceanOptics spectrometer. Figure 4.2
shows the test results. The temperature coefficient of the LED is estimated to be about
0.27nm/°C.
Figure 4.2 Temperature dependence of the central wavelength of the LED
25 30 35 40 45 50
852
853
854
855
856
857
858
859
Temperature (
o
C)
C
e
n
t
r
a
l
W
a
v
e
l
e
n
g
t
h
(
n
m
)
Yibing Zhang Chapter 4. System noise analysis and performance optimization 67
Without wavelength drifts, sinusoidal curves defined by Equation (2.5) containing
temperature information, is obtained after normalization of Equation (2.4) to Equation
(2.3). Suppose the wavelength of the peak point is λ
P
, from Equation (2.5), at the peak
point, we have:
2
cos( ) 1
2
2 , k is a positive integer
2
sin( ) 0
p
p
p
p
L
L
k L k
L
π
λ
π
π λ
π λ
λ
|

|
|
⇒ ⇒
|
|

|
|
(4.3)
With a known k value, we can calculate OPD from λ
P
, which in turn can be related to
temperature information by calibration.
Since no method can be used to acquire the spectrum of the LED in real time for the
interference spectrum normalization, the pre-recorded reference spectrum acquired at
temperature T
0
is used to normalize the interference spectrum acquired at temperature T.
The normalized spectrum N(λ) is then expressed as:
))
2
cos( 1 (
)
) (
exp(
)
) (
exp(
) (
2
2
0
2
2
λ
π
γ
λ λ
λ λ
λ
L
w
w
N
T
+




(4.4)
The normalized curve is a product of the sinusoidal curve, which contained the OPD
information for the temperature measurements, and the ratio between the two optical
spectrum curves corresponding to different temperatures. This ratio is not equal to unity
because of the center wavelength drift. One normalized curve is plotted in Figure4.3,
where the OPD is chosen to be 9.2µm, the spectral width w is 27nm, visibility of the
interference spectrum γ is 0.5, and the shift of the central wavelength is 5nm. Figure 4.4
shows the measured normalized spectrum when the temperature was increased to 45°C
(the reference spectrum was acquired at 25°C). Obviously, with the use of a reference
spectrum acquired at different temperatures, the normalized spectrums are deformed from
Yibing Zhang Chapter 4. System noise analysis and performance optimization 68
the sinusoidal curve and the peak/valley locations in the interference spectrum are
shifted.
Figure 4.3 Simulated results of the normalized spectrum with center wavelength shift
effect
Figure 4.4. Measurement results of the normalized spectrum at 45°C (the reference is
acquired at 20°C)
Let us analyze quantitatively the temperature uncertainty caused by the center
wavelength drift.
After the center wavelength of the spectrum drifts to
0
'
λ , the output spectrum from the
LED source becomes:
790 800 810 820 830 840 850 860 870 880 890 900
0
0.2
0.4
0.6
0.8
1
Wavelength(nm)
N
o
r
m
a
l
i
z
e
d
S
p
e
c
t
r
u
m
Yibing Zhang Chapter 4. System noise analysis and performance optimization 69
)
) (
exp( ) (
2
2
0
'
0
w
I I
λ λ
λ

− (4.5)
Because of the wavelength drift, the temperature measurement will be affected, and the
new normalized curve will be:
))
2
cos( 1 (
)
) (
exp(
)
) (
exp(
) (
) (
) (
2
2
0
'
2
2
0
'
λ
π
γ
λ λ
λ λ
λ
λ
λ
L
w
w
I
I
N +




(4.6)
Assuming that
0
0
'
λ λ δ
λ
− , and consider 1
0
<< λ δ
λ
, then using a first order
approximation yields,
)
) ( 2
1 (
)
) (
exp(
)
) (
exp(
2
0
2
2
0
2
2
0
'
λ
δ
λ λ
λ λ
λ λ
w
w
w

+ ≈




(4.7)
The normalized Equation (4.6) becomes:
))
2
cos( 1 ( )
) ( 2
1 (
) (
) (
) (
2
0
'
λ
π
γ δ
λ λ
λ
λ
λ
λ
L
w I
I
N + ×

+ (4.8)
The shifted peak position λ
P
'
can be acquired by differentiating Equation (4.8) and setting
it equal to zero,
)
2
cos(
2
)
2
sin(
2 2
' 2 ' 2
'
2
'
p p
p
L
w
L L
w
N
p
λ
π
γ
δ
λ
π
λ
γ π δ
λ
λ λ
λ λ
+ +

0 )
2
sin(
2
) ( 2
' 2
'
2
0
'
× ×

+
p
p
p
L L
w λ
π
λ
π
γ δ λ λ
λ
(4.9)
Because
p p p
λ λ λ λ << − ∆
'
, from Equation (4.3), we obtain,
2 2 2 '
'
2
)
2
2 sin( )
) ( 2
sin( )
2
sin(
1 )
2
cos(
p p p
p
p
p
L L
k
L
L
L
λ
λ π
λ
λ π
π
λ
λ λ π
λ
π
λ
π

− ≈


∆ −


(4.10)
Yibing Zhang Chapter 4. System noise analysis and performance optimization 70
Substituting these two equations into Equation (4.9) and assuming that 1 << w
λ
δ ,
1 << ∆
P
λ λ , a first order approximation yields:
P
p
p
w L
λ δ
γ π
λ γ
λ
λ
+
+

2 2 2
4
'
4
) 1 (
(4.11)
From Equation (3.11), the OPD demodulated from the shifted peak wavelength λ
P
'
is:
L m L
p
p
p
p
λ
λ
λ
λ
'
'
(4.12)
Using Equation (4.11) and Equation (4.12), we obtain:
λ λ
δ
γ π
λ γ
2 2
3
4
) 1 (
Lw
L L
p p
+
+ (4.13)
Equation (4.13) is used to evaluate the wavelength shift induced OPD measurement
uncertainties (δOPD) in Figure 4.5. The temperature measurement uncertainties can be
obtained from Figure 4.6 through the calibration relation between the OPD and the
temperature.
0 2 4 6 8 10
0
0.5
1
1.5
2
2.5
3
3.5
4
Center wavelength drift δ λ (nm)
δ
O
P
D
(
n
m
)
Figure 4.5. OPD measurement uncertainties caused by center wavelength shift
(δλ) (γ=1, λ
p
= 857nm, L =30um,w=60nm)
Yibing Zhang Chapter 4. System noise analysis and performance optimization 71
0 2 4 6 8 10
0
0.5
1
1.5
2
2.5
3
Center wavelength drift δ λ (nm)
T
e
m
p
e
r
a
t
u
r
e
u
n
c
e
r
t
a
i
n
t
y
(
°
C
)
Figure 4.6. Temperature uncertainty caused by center wavelength shift (δλ) (γ=1,
λ
p
= 857nm, L =30um,w=60nm)
4.1.2 Spectral bandwidth broadening/narrowing effects
According to the Gaussian curve approximation of the output spectrum of the LED,
shown in Equation (2.3), spectral bandwidth is another factor that can induce errors in the
temperature measurements. To see the effect of the spectral bandwidth changes on the
temperature measurements qualitatively, assuming that the center wavelength is fixed, a
first-order approximation is used to analyze the normalized spectral curves. The spectrum
curve with bandwidth change w δ relative to the original bandwidth w can be described
as:
)
) (
) (
exp( ) (
2
2
0
0
w w
I I
w
δ
λ λ
λ
+

− (4.14)
The normalized interference fringes will then be:
))
2
cos( 1 (
)
) (
exp(
)
) (
) (
exp(
) (
) (
) (
2
2
0
2
2
0
λ
π
γ
λ λ
δ
λ λ
λ
λ
λ
L
w
w w
I
I
N
w
w
+


+


(4.15)
Assuming that 1
0
<< λ δ
w
, yields a first order approximation:
)
) ( 2
1 (
)
) (
exp(
)
) (
) (
exp(
3
2
0
2
2
0
2
2
0
w
w
w
w w
δ
λ λ
λ λ
δ
λ λ

+ ≈


+


(4.16)
Yibing Zhang Chapter 4. System noise analysis and performance optimization 72
So the normalized Equation (4.15) becomes:
))
2
cos( 1 ( )
) ( 2
1 (
) (
) (
) (
3
2
0
λ
π
γ δ
λ λ
λ
λ
λ
L
w I
I
N
w
w
w
+ ×

+ (4.17)
The shifted peak position λ
P
'
can be acquired by differentiating Equation (4.17) and
setting it equal to zero. Doing so gives,
)
2
cos(
) ( 2
)
2
sin(
2 ) ( 4
' 3
0
' 2
'
3
0
'
p
w
p
p
w w
L
w
L L
w
N
p
λ
π
γ
δ λ λ
λ
π
λ
γ π δ λ λ
λ
λ λ

+ +


0 )
2
sin(
2 ) ( 4
' 2
'
3
0
× ×

+
p
p
w
L L
w λ
π
λ
π γ δ λ λ
(4.18)
Assuming that w
w
<< δ , and noting that in regions around the peaks in the interference
fringes:
2 2 2 '
'
2
)
2
2 sin( )
) ( 2
sin( )
2
sin(
1 )
2
cos(
p p p
p
p
p
L L
k
L
L
L
λ
λ π
λ
λ π
π
λ
λ λ π
λ
π
λ
π

− ≈


∆ −


(4.19)
then substituting these two equations back into (4.18), and further considering
that 1 << w
w
δ , and 1 << ∆
P
λ λ , yields the following first order approximation,
P w
p
w p
w L
λ δ
γ π
λ γ
λ +
+

− 3 2 2
4
'
2
) 1 (
(4.20)
Using Equation (3.11), the OPD demodulated from the shifted peak wavelength λ
P
'
will
be:
L m L
p
w p
w p
p
w
λ
λ
λ
λ
'
' −
− −
(4.21)
From Equation (4.20) and Equation (4.21), we obtain:
Yibing Zhang Chapter 4. System noise analysis and performance optimization 73
w
w L
L L
p p
w
δ
γ π
λ γ
λ 3 2
3
2
) 1 ( +
+

(4.22)
This equation can be used to estimate the OPD measurement errors caused by the
spectrum bandwidth changes as shown in Figure 4.7; Figure 4.8 shows the temperature
measurement uncertainty caused by the spectrum bandwidth changes.
0 2 4 6 8 10
0
0.01
0.02
0.03
0.04
0.05
0.06
0.07
Bandwidth changes δ w(nm)
δ
O
P
D
(
n
m
)
Figure 4.7. OPD measurement uncertainties caused by the bandwidth changes
(δw) (γ=1, λ
p
= 857nm, L =30um,w=60nm)
0 2 4 6 8 10
0
0.01
0.02
0.03
0.04
0.05
Bandwidth changes δ w(nm)
T
e
m
p
e
r
a
t
u
r
e
u
n
c
e
r
t
a
i
n
t
y
(
°
C
)
Figure 4.8. Temperature uncertainty caused by the bandwidth changes (δw) (γ=1,
λ
p
= 857nm, L =30um,w=60nm)
4.2 Degradation effect due to the visibility of the interference spectrum
The visibility of the interference spectrum may be influenced by the optical polarizer
extinction ratio, the optical alignment between the polarizer and the sensing element, and
the collimation quality of the optical collimator. Among these factors, for the given
Yibing Zhang Chapter 4. System noise analysis and performance optimization 74
qualities of the polarizer and the optical collimator, the influence from the optical
alignment between the optical polarizer and the sensing element is dominant. The optical
alignment changes are also unavoidable in industrial fields where the mechanical
vibration problem is severe. The following analysis is using the optical misalignment
caused visibility changes to evaluate the temperature measurement uncertainty. The
sensing element is intentionally rotated relative to the polarizer within a plane normal to
the light propagation direction, as shown in Figure 4.9, so that the visibilities are
changed.
The results shown in Equation (2.9) assume that the polarization direction P of the input
light beam is at 45º relative to the fast- and slow-axes of the sensing element in the plane
normal to the light propagation direction. In fact, it is very difficult to align all optical
components accurately in the sensor probe. Also, mechanical vibrations may cause
changes of the relative positions between the sensing element and the polarizer. Thus, it
is necessary to analyze the dependence of the temperature measurement uncertainties on
the angle α.
Optical
Polarizer
Rotation in
the plane
f-axis
s-axis
45
o
Optical wave
propogation
direction
f
s
P
P'
α
Figure.4.9. Misalignment between the optical polarizer and the sensing element.
When a linearly polarized light beam passes through the sapphire sensing element, the
light behaves as if it is divided along the fast- and slow-axis separately as shown in
Figure 4.10. The refractive index is n
e
along the f-axis, and n
o
along the s-axis. At the
input surface of the sapphire sensing element, the ordinary wave and the extraordinary
Yibing Zhang Chapter 4. System noise analysis and performance optimization 75
wave are still in phase, namely the phase difference is δ=0. After light passes through the
sensing element, the phase difference δ will be changed to
λ π δ / 2 n d∆ (4.23)
where d is the thickness of the sensing element, ∆n = n
o
-n
e
is the birefringence of the
single crystal sapphire material, and λ is the optical wavelength. Depending on the
magnitude of the δ value, as well as the alignment angle α between the linear polarization
direction of the input light and the principal axes in the sensing element, the output light
from the sensing element can be linearly polarized, circularly polarized or elliptically
polarized.
Polarizer/analyzer polarization
direction
f
s
α
E
Ecosα
Esin
2
α
Ecos
2
α
Esinα
Figure 4.10. Decomposition and interference of the linearly polarized input light in the
polarimeter.
If another polarization analyzer is placed behind the sensing element with its polarization
direction parallel to the linear polarization direction of the input light, the electric field of
the output optical wave can be expressed in a vector form as
α α
2
0
2
0
cos sin E E E
out
v v v
+ (4.24)
where
out
E
v
, α
2
0
sin E
v
and α
2
0
cos E
v
are all vectors. The intensity of the output light will
be:
2

r
out out
I E (4.25)
The interference results will contain the information about the phase delay δ between the
two linearly polarized beams in the sensing element, as well as the angle α:
Yibing Zhang Chapter 4. System noise analysis and performance optimization 76
)
2
sin 2 sin 1 (
)] 1 (cos 2 sin
2
1
1 [
cos )
2
2 sin
( 2 ) cos ( ) sin (
2 2
0
2 2
0
2 2
0
2 2
0
2 2
0
δ
α
δ α
δ
α
α α
× +
− × +
+ +
I
E
E E E I
out
(4.26)
From Equation (4.23), δ is a function of the product d∆n. When the temperature changes,
the thickness d of the sapphire sensing element will change, as well as the birefringence
∆n = n
o
-n
e
. Equation (4.26) is thus essentially a function of temperature.
The factor related to the angle α in Equation (4.26) will determine the visibility of the
interferogram of the two polarized beams. After normalization of the measured spectral
curve I(λ) with the intensity I
0
(λ) of the input broadband optical source, the visibility of
the interferogram will be
α γ 2 sin
2
. (4.27)
The dependence of the visibility on the angle α is plotted in Figure 4.11.
20 30 40 50 60 70
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Angle α (°)
V
i
s
i
b
i
l
i
t
y
Measurement results
Teoretical results
Figure 4.11. Visibilities dependence on the misalignment angle α between the light
polarization direction and principal axes of the sensing element.
By changing the angle α, the measured optical spectra with different visibilities are
plotted in Figure 4.12. The maximum and minimum intensities in these spectral curves,
Yibing Zhang Chapter 4. System noise analysis and performance optimization 77
corresponding to the peaks and valleys in the curves, are fixed. Since only these peaks
and valleys are used to extract the temperature information, accurate temperature
measurement results can thus still be obtained in a wide changing range of the visibilities.
Experimental results demonstrate that the temperature can be measured accurately with
the change of the angle α up to 10º, as shown in Figure 4.13. This tolerance of the angle α
can be easily met in the laboratory environment as well as in the industrial environments,
where mechanical vibration condition is usually severe.
800 820 840 860 880 900
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
Wavelength(nm)
N
o
r
m
a
l
i
z
e
d
I
n
t
e
n
s
i
t
y
α=30°
α=35°
α=40°
α=45°
Figure 4.12. Experimental interferograms with different visibilities.
45 50 55 60 65 70 75
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
Angle α (°)
T
e
m
p
e
r
a
t
u
r
e
d
e
v
i
a
t
i
o
n
(
°
C
)
Figure 4.13. Measured temperature deviation caused by the changes in the visibility of
the interferogram
4.3 Optical birefringence effects
As presented in Chapter 2, the BPDI technology measures the optical path difference
Yibing Zhang Chapter 4. System noise analysis and performance optimization 78
(OPD), or the phase retardation between the two orthogonal linearly polarized light
beams in a single crystal sapphire element, which is a function of both the temperature
dependent birefringence and the temperature dependent thickness. The retardation in the
sensing element is a refractive index effect. It is known that as light passes from the
vacuum into a material, the speed of light is reduced by a factor that is the reciprocal of
the refractive index. An anisotropic material is completely characterized by three
principal refractive indices, which can be represented by an index ellipsoid. In uniaxial
materials, one principal refractive index is different from the other two. The rays that are
used to define the unique (optic) axis are called ordinary rays and the refractive index is
n
o
. Those that are associated with the other axes are called extraordinary rays and the
refractive index is n
e
. Materials with different principal refractive indices are therefore
referred to as birefringent.
The magnitude of the birefringence in an anisotropic material depends on the propagation
direction of the optical wave in the material. As shown in Figure 4.14, the index ellipsoid
is used to determine the magnitude of the two indices of the refraction (n
o
and n
e
corresponding to D
1
and D
2
) for the two orthogonal linearly polarized waves in the
material. This is done by finding the intersection ellipse between a plane through the
center of the ellipsoid that is normal to the direction of light propagation. The axes of the
intersection ellipse are equal in length to 2n
o
and 2n
e
, where n
o
and n
e
are the indices of
refraction corresponding to the two orthogonal linearly polarized light waves. The two
waves are polarized in directions parallel to D
1
and D
2
.
s
D
1
D
2
x
z
Figure 4.14. Method of the index ellipsoid to determine the refractive indices of light
propagating along s direction.
Yibing Zhang Chapter 4. System noise analysis and performance optimization 79
Single crystal sapphire is a uniaxial crystal because of its hexagonal crystalline structure,
as shown in Figure 3.2. The index ellipsoid corresponding to the hexagonal inner atomic
structure is shown in Figure 4.15. The refractive index along the C-axis is n
o
, the
refractive index along the A-axis is n
e
. Single crystal sapphire is a negative uniaxial
crystal, meaning n
o
>n
e
; thus the C-axis is the slow axis and the A-axis is the fast axis. In
the single crystal sapphire, the refractive index is constant in the x-y plane, there is no
birefringence in this plane and is called the C-plane. In the y-z or x-z plane, the refractive
indices along different axes are different; birefringence ∆n = n
o
-n
e
exists, and called the
A-plane. With the propagation direction of the incident light wave perpendicular to the
sapphire sensing disk surface (A-plane), the phase delay between the two orthogonal
linearly polarized waves will be δ = 2πd∆n/λ, where d and ∆n are the only terms that are
functions of temperature. Therefore temperature can be measured directly by monitoring
the phase delay δ with the BPDI technology. For the case of non-normal incident light,
the birefringence ∆n = n
o
-n
e
will be incident angle dependent, as well as the optical path
length that the light passes through in the sapphire sensing element, thus a more general
formula for the phase delay calculation should be:
( )
( ) ( )
λ
θ θ π
θ δ
, , 2
,
T n T d
T

(4.28)
where T is the temperature, and θ is the light incident angle.
y(a-axis)
x(a-axis)
z(c-axis)
no
ne
ne
Figure 4.15. The refractive index ellipsoid for single crystal sapphire.
To analyze the temperature measurement uncertainty caused by the non-normal incident
light on the sapphire sensing element, the sensing element is rotated about its fast axis,
Yibing Zhang Chapter 4. System noise analysis and performance optimization 80
slow axis and an axis 45
o
from the slow/fast axis respectively to study their effects on the
OPD measurement results.
4.3.1 Rotation about the slow-axis (i.e. C-axis)
The rotation is shown in Figure 4.16 (a). In this case, the refractive indices for the
ordinary and extraordinary waves are all constant, shown in Figure 4.16(b), thus the
birefringence ∆n = n
o
-n
e
is independent of the light incident angle. For this situation, the
optical path L, along which light passes through the sapphire disk, is the only factor that
is incident angle dependent.
Optical Polarizer
Rotation about
s-axis
s-axis
Optical wave
propogation
direction
P
S
(a) Rotation about the s-axis of the sensing element
c-axis
e
o
(b)Wave vectors for the double refraction in the sapphire, with the optical axis parallel to the boundary
and perpendicular to the plane of the incidence.
Figure4.16. Optical refractive index ellipse for the rotation about the s-axis of the sensing
element
Yibing Zhang Chapter 4. System noise analysis and performance optimization 81
In the room temperature environment, the refractive indices of the single crystal sapphire
can be approximated by Equation (3.1), as plotted in Figure 4.17(a). The birefringence ∆n
= n
o
-n
e
is plotted in Figure 4.17(b) as a function of the optical wavelengths.
0.76 0.80 0.84 0.88 0.92
1.748
1.750
1.752
1.754
1.756
1.758
1.760
1.762
1.764
Wavelength (nm)
R
e
f
r
a
c
t
i
v
e
i
n
d
e
x
no
ne
0.76 0.8 0.84 0.88 0.92
7.89
7.9
7.91
7.92
7.93
7.94
7.95
7.96
7.97
x 10
-3
Wavelength (nm)
B
i
r
e
f
r
i
n
g
e
n
c
e
(
n
o
-
n
e
)
(a)Refractive indices Vs wavelengths (b) birefringence vs wavelengths
Figure 4.17. Refractive indices and birefringence vs. optical wavelength for sapphire.
Selecting the refractive indices at the wavelength of interest λ=850nm, n
e
=1.751 and
n
o
=1.7589, the refraction angles vs. incident angles for the ordinary and extraordinary
waves are plotted in Figure 4.18(a). The difference between the two angles is so small as
shown in Figure 4.18(b), that they can be treated as equal in analyzing the rotation effects
on the temperature measurements. As shown in Figure 4.19(a), the OPD change is only
related to the change in L as a result of the disk rotation. The OPD values are:
) angle an rotated disk sensing the ( cos / ) ( '
light) incident normal ( ; ) (
1 0


θ θ L n n OPD
L n n OPD
e o
e o
(4.29)
where θ
1
is the light refraction angle corresponding to the light with an incident angle θ
0
.
Changes of the OPDs vs. rotation angles (θ
0
) up to 50º have been measured, and the
results are compared with theoretical simulations as shown in Figure 4.20.
Yibing Zhang Chapter 4. System noise analysis and performance optimization 82
0 50 100 150 200
0
5
10
15
20
25
30
35
Incident angle(°)
R
e
f
r
a
c
t
i
o
n
a
n
g
e
l
(
°
)
Ordinary ray
Extrordinary ray
0 50 100 150 200
0
0.02
0.04
0.06
0.08
0.1
0.12
0.14
0.16
0.18
Incident angle(°)
R
e
f
r
a
c
t
i
o
n
a
n
g
e
l
d
i
f
f
e
r
e
n
c
e
(
°
)
(a) (b)
Figure 4.18. (a) Refraction angles of the ordinary waves and the extraordinary waves
(rotation about the s-axis).
(b) The refraction angle difference between the ordinary waves and the extraordinary
waves (rotation about the s-axis)
θ
L
θ
0
1
n
o
-n
e
Figure.4.19. Sensing element rotation effect on the optical path difference (OPD)
(rotation about the s-axis).
10 20 30 40 50 60 70 80
28
28.5
29
29.5
30
30.5
31
31.5
32
32.5
Rotation angle(°)
O
P
D
(
µ
m
)
Theoretical results
Measurements
Figure4.20. Sensing element rotation effects along s-axis on the OPD measurements.
Yibing Zhang Chapter 4. System noise analysis and performance optimization 83
In both the theoretical analysis and experimental measurements, it was shown that the
OPD values increase with the increasing of the rotation angles about the C-axis. In a
constant temperature environment, if the sensing element rotates, the OPD values change
rather than remain constant. This change in the OPD values will thus affect the accuracy
of the temperature measurements. Using the OPD vs. rotation angle curves shown in
Figure 4.20, the rotation effects on the temperature measurements can be estimated
according to the calibrated relationship between the OPDs and the temperatures. It is
shown that the temperature measurement uncertainty is less than 4ºC for rotation angles
up to 5º.
4.3.2 Rotation about the fast-axis (i.e. A-axis)
The rotation is shown in Figure 4.21 (a). In this case, the refractive index for the ordinary
light wave is independent of the incident angles, and it obeys Snell’s law:
1 0
sin sin θ θ
o i
n n (4.30)
Optical
Polarizer
Rotation
about
f-axis
f-axis
Optical wave
propogation
direction P
f
(a) Rotation about the s-axis of the sensing element
(b) Wave vectors for the double refraction in the sapphire, with the optic axis parallel to the boundary and
perpendicular to the plane of incidence.
Figure 4.21. Optical refractive index ellipse for the rotation about the f-axis of the sensing
element.
c-axis
o
e
Yibing Zhang Chapter 4. System noise analysis and performance optimization 84
where n
i
is the refractive index of the media from which the light is incident, and θ
0
and
θ
i
are the incident angle and refraction angle of the light, respectively.
For the extraordinary waves in the A-plane sapphire disk, the corresponding refractive
index depends on the direction of light propagation. With rigorous mathematical analysis
shown in Appendix I, the refraction angle θ
2
and refractive index n
e
can be expressed as,
2
2
2 2
2
2 2 2
0
2 2 2 2
0
2 2 2
1
2
) ( tan
)) ( tan 1 (
)
) sin (
sin
( tan
o eo
o eo
e
o eo
o
n n
n n
n
n n n
n n
+
+


θ
θ
θ
θ
θ
(4.31)
where n is the refractive index of the media from which the light is incident (n=1 for the
incident light from the air); n
eo
is a special n
e
value, which is the refractive index of the
extraordinary waves normal incident to the surface of the A-plane sapphire disk.
Refractive indices for the extraordinary waves increase as the refraction angles increase,
as shown in Figure 4.22.
0 5 10 15 20 25 30 35
1.751
1.7515
1.752
1.7525
1.753
1.7535
1.754
Refraction angle(°)
R
e
f
r
a
c
t
i
v
e
i
n
d
e
x
(
n
e
)
Figure 4.22. Refractive indices vs. light incident angles for the extraordinary waves.
Using the refractive index for the wavelength of interest, λ=850nm (n
eo
=1.751,
n
o
=1.7589), the refraction angles for the ordinary wave and the extraordinary wave are
plotted in Figure 4.23(a). The difference between the two angles is so small, as shown in
Figure 4.23(b), that the refraction angles in the sapphire disk for these two waves can be
approximated as equal for the analysis of the rotation effects on the temperature
Yibing Zhang Chapter 4. System noise analysis and performance optimization 85
measurements. The changes in the OPDs are related to the L changes as a result of the
disk rotation, as well as the changes in the refractive indices for the extraordinary wave,
as shown in Figure 4.24.
2 2
cos / ) )( ( '
; ) (
θ θ L n n d
L n n d
e o
e o


(4.32)
0 50 100 150 200
0
5
10
15
20
25
30
35
Incident angle(°)
R
e
f
r
a
c
t
i
o
n
a
n
g
e
l
(
°
)
Ordinary ray
Extrordinary ray
0 50 100 150 200
0
0.02
0.04
0.06
0.08
0.1
0.12
0.14
Incident angle(°)
R
e
f
r
a
c
t
i
o
n
a
n
g
e
l
d
i
f
f
e
r
e
n
c
e
(
°
)
(a) (b)
Figure 4.23. (a) Refraction angles of the ordinary waves and the extraordinary waves (rotation about the f -
axis)
(b) The refraction angle difference between the ordinary waves and the extraordinary waves (rotation
about the f-axis)
θ
L
θ
0
2
n
o
-n
e
(θ2)
Figure 4.24. Rotation effects on the optical path difference (rotation about the f-axis)
The changes in the OPDs vs. rotation angles up to 50º have been measured; the results are
compared with the theoretical simulations in Figure 4.25.
Yibing Zhang Chapter 4. System noise analysis and performance optimization 86
10 20 30 40 50 60
24
24.5
25
25.5
26
26.5
27
27.5
28
Rotation angle(°)
O
P
D
(
µ
m
)
Theoretical results
Measurements
Figure 4.25. Effect of rotation about the f-axis on the OPD measurements.
According to the theoretical analysis and experimental measurements, the OPD values
decrease when the sensing element rotates about its A-axis. By comparing curves shown
in Figure 4.20 and Figure 4.25, it is evident that the magnitude of the OPD changes are
smaller when the rotation is about the A-axis instead of the rotation about the C-axis for a
given rotation angle. These results can be explained as following: when the sensing
element rotates about the A-axis and the light incident angles increase, the L’ value
increases from the initial value L, the birefringence decreases from the initial value
∆n
o e
n n −
0
, and the combined effects cause the OPDs to decrease; when the rotation is
about the C-axis, OPDs increase because the L’ value increases from the initial value L,
and the birefringence is fixed constant at its initial value ∆n
o e
n n −
0
, the changing
magnitude is then larger than the OPD changes corresponding to the same angle rotation
about the A-axis.
With the relationship between the OPDs and rotation angles shown in Figure 4.25, we
can estimate that the temperature measurement uncertainty is less than 3ºC for rotation
angles up to 5º about the fast axis, based on the calibrated relationship between the OPDs
and the temperature values.
4.3.3 Rotation about an axis 45º from the slow-axis/fast-axis
According to the analytical and experimental results presented in the two previous
Yibing Zhang Chapter 4. System noise analysis and performance optimization 87
sections, it is expected that the dependence of OPD values on rotation angles will be
reduced if the rotation occurs about an axis between the slow axis and fast axis. The total
rotation effect on temperature measurements can thus be partially cancelled between the
increasing and decreasing values of OPDs. We tested the rotation of the sensing element
about the z-axis, shown in Figure 4.26. The direction of the z-axis is at 45º relative to
both the slow axis and the fast axis of the sensing element. This rotation can be treated as
partial rotation about the fast axis, which will cause the OPDs to decrease, and partial
rotation about the slow axis, which will cause the OPDs to increase. Experimental results
are shown in Figure 4.27. With a rotation angle less than 5º, the effect on the temperature
measurement uncertainty is less than 2ºC.
Optical
Polarizer
Rotation about the
axis between f-axis
and s-axis
f-axis s-axis
45
o
Optical wave
propogation
direction
P
f s
z
y
x
Figure 4.26. Rotation about an 45º from the s-axis/f-axis of the sensing element.
0 5 10 15 20 25 30
27.75
27.8
27.85
27.9
27.95
28
Rotation angle(°)
O
P
D
(
µ
m
)
Figure.4.27. Effect on the OPD measurements of rotation about an axis 45º from the f-
axis of the sensing element.
Yibing Zhang Chapter 4. System noise analysis and performance optimization 88
4.4 Opto-electronic noise in spectrummeasurements
The opto-electronic noise in this system is generated inside the optical spectrometer,
where dispersed light from the grating is scanned electronically by a 2048-element linear
silicon CCD array and converted into electric signals, then processed in the computer, as
shown in Figure 4.28. The CCD detecting system is a solid-state sensor consisting of a
wafer of silicon crystal, and has 2048 photodiode-capacitor pairs. Radiation coming from
the optical grating hits the photodiodes, causing the capacitors to discharge. These
capacitors contain charges proportional to the photon flux. At the end of the integration
period, a series of switches close and transfer the charge to a shift register. The charges
on the shift register are digitized and the values are amplified and sent to the computer,
which plots an optical spectrum. The noises associated with the CCD array are mainly
readout noise and dark current noise.
Computer
Dispersed light from grating
Photodiode
Capacitor
Shift register
2048 elements
Figure 4.28. CCD detecting array for the optical signal detection in the optical
spectrometer
Readout noise:
The readout noise of the CCD detector is a combination of thermal white noise, 1/f noise
from its electrical components, i.e. photodiodes, capacitors, shift registers and the on-chip
amplifier. The white noise and the 1/f noise depend on the on-chip amplifier size of the
CCD detector, and become lower when the amplifier sizes increase. But, as the amplifier
size grows, its input capacitance from the sensing pixels (each pixel is composed of one
photodiode, one capacitor and one shift register), also grows, lowering the sensitivities of
Yibing Zhang Chapter 4. System noise analysis and performance optimization 89
the sensing pixels and increasing the readout noise. With an optimum design, typical
readout noise is 4-5 electrons for the current commercial products.
Dark current noise:
The CCD array liberates electrons in the photodiode to generate electric signals. The
amount of electrons generated by each CCD pixel depends on the photo flux. This
process is extremely temperature dependent. Even without optical illumination, electrons
can also be generated by the finite temperature of the CCD. The thermally generated
electrons are called the dark current. The dark current noises can be minimized by
cooling the CCD detector.
The goal of the measurement of the optical spectrum electronically is the unambiguous
determination of the OPDs between the two orthogonal linearly polarized light waves,
which can be calculated by Equation (3.14) with the known spectral positions of the
peaks and valleys in the interference spectral curves. The opto-electronic noise in the
CCD detector affects the accurate determination of the peak and valley positions in the
interference spectral curves. Since the ideal measured signal intensities corresponding to
the valley positions are zero, the SNR at those points are thus very small and affected by
the noise more seriously than the peaks points. To be conservative, we analyze the opto-
electronic noise effects on temperature measurements by considering the valley points
with noises. By differentiating Equation (2.4), the noise amplitude I δ , or the signal
uncertainty, is related to the uncertainty of the valley positionsδλ :
, , 1 )
2
sin(
)
2
sin(
2
) (
2
valleys the for
L
L L
I I
s

λ
π
λ
π
λ
δλ π
γ λ δ
(4.33)
where L is the OPD value, λ is the optical wavelength, ) (λ
s
I is the optical signal, and γ is
the visibility of the interference spectral curves. This equation can be transformed into:
SNR L
I I SNR
I I L
s
s
πγ
λ
δλ
δ λ
δ λ πγ
λ
δλ
2
) / ) ( (
) / ) ( ( 2
2
2
2

|
|
|
|
|

(4.34)
Yibing Zhang Chapter 4. System noise analysis and performance optimization 90
Also according to Equation (2.14), the measurement uncertainty of the valley position
will affect the OPD calculation by the following relation:
2 1
2 1
) (
λ λ
λ λ

T f OPD (4.35_a)
2 / 1 2
2
2
2
1
1
] ) ( ) [( δλ
λ
δλ
λ
δ


+

f f
f (4.35_b)
The CCD array used in the optical spectrometer is Sony
®
ILX511 linear CCD array, with
a SNR of 250. Assuming γ=1, then the uncertainties of both OPD measurement and
temperature measurement are OPD value dependent. Combing Equation (4.34), Equation
(4.35_a) and Equation(4.35_b), where L=f(T), the OPD measurement uncertainties are
calculated and plotted in Figure 4.29. Based on the calibration curve between OPD and
temperature, Figure 4.30 shows the temperature measurement uncertainty related to the
results shown in Figure 4.29. We can see that the opto-electronic noise contributes
significantly to the measurement uncertainties of temperature. With the OPD increase,
the temperature measurement uncertainties increase. To control the opto-electronic noise
effect on temperature measurements, the OPD values should be selected properly for
temperature measurement over a wide dynamic range. To increase the SNR of the
spectrum measurement, there is another effective way to minimize the opto-electronic
noise, and will be discussed in section 4.6.
Figure 4.29. Opto-electronic noise effects on the valley point locations determinations
15 20 25
0.2
0.25
0.3
OPD (µm)
δ
λ
(
n
m
)
Yibing Zhang Chapter 4. System noise analysis and performance optimization 91
Figure 4.30. Opto-electronic noise effects on the temperature measurement uncertainties
4.5 Optical fiber induced noises
Optical fiber is used in the BPDI sensor system to guide light into the sensing probe, re-
collect light reflected from the end of the sensing probe, and then guide the optical
temperature signature into the spectrometer for signal processing. Because of its small
size, immunity to EMI and chemical corrosion resistance, optical fiber offers great
system deployability in harsh environments. At the same time the optical polarization
effects, optical dispersion, fiber bending-induced transmission spectrum drift in the
optical fiber might induce noises in the temperature measurement (signal distortion is
viewed as noise in this discussion).
Since the sensing scheme for the BPDI sensor system is based on polarimetric
interferometry, the state of polarization of the light is very important for the sensor to
perform accurate temperature measurements. The temperature information is encoded
into the polarization states and the spectrum of light inside the sensing element, the
output temperature signature from the sensing probe is a linearly polarized light. When
coupled back into the multimode optical fiber, its polarization state usually becomes
random, and the light waves with different polarization states still contain the same
spectrum information, thus the same temperature information. The polarization mode
15 20 25
0.7
0.8
0.9
1
1.1
1.2
1.3
1.4
1.5
OPD (µm)
T
e
m
p
e
r
a
t
u
r
e
d
e
v
i
a
t
i
o
n
(
C
)
°
Yibing Zhang Chapter 4. System noise analysis and performance optimization 92
dispersion (PMD) may be a concern in single mode fibers for long haul transmission.
However for multimode fiber, it is not a problem for a few kilometer transmission
distances. Muliti-mode optical fiber increases the energy coupling efficiency between the
LED source and the optical fiber; collimation of light is also easier for multimode fiber
with larger numerical aperture compared to single mode optical fibers.
The optical fiber dispersion in a multimode fiber includes inter-modal dispersion and
chromatic dispersion. Chromatic dispersion occurs because different colors of light travel
through the fiber at different speeds; inter-modal dispersion is also called multipath
dispersion, where different rays or modes in the multimode fiber travel along paths of
different lengths. The inter-modal dispersion is dominant in multimode fibers. Graded-
index multimode fibers used in the system have less inter-modal dispersion compared to
step-index multimode fibers with the same geometrical structure and numerical aperture.
A graded index multimode fiber with optimized index profile has a capacity of 100Mb/s
over a distance of up to 100km [68]. Experiments with a 3km long multimode fiber in the
BPDI sensor system revealed no obvious degradation effects on the optical pulse signals
with a repetition rate of 8Hz.
When light propagates inside the fiber, the spectrum may be distorted because of fiber
bending, so the center wavelength and the bandwidth of the transmission light may be
changed. Compared to the optical source spectrum distortion caused by the
environmental temperature fluctuations, this effect in the fiber is much small, and was not
observed in the BPDI sensor system.
4.6 Summary of system noises and their optimization
A summary is listed in Table4.1 for system noises and their degradation effects on
temperature measurements. The design of the system, including the software and the
hardware, is optimized based on systematic analyses of these noises.
Yibing Zhang Chapter 4. System noise analysis and performance optimization 93
Table 4.1. Noise in the BPDI sensor system and compensation methods
Noise type Noise source Temperature
measurement
uncertainty
Electrical
noise
1). Dark current
2).Thermal white noise
3). 1/f noise
CCD detector in the
spectrometer
2
o
C (0.063% full
measurement scale)
Optical
noise
1). Spectral center
wavelength drift
2).Spectral bandwidth
broadening/narrowing
3).Visibility changes of
the spectrum
Optical source: LED;
Optical spectrometer;
Polarizer (analyzer);
Transmission fiber;
1) 3
o
C (0.19% full
measurement scale)
2) 0.05
o
C (0.00312%
full measurement
scale)
3) 1
o
C (0.06% full
measurement scale)
Mechanical
noise
Mechanical vibration Polarizer and sensing
element relative
position changes
5
o
C (0.3% full
measurement scale) in
the range of 5
o
rotation
Opto-electronic noise reduction
In a relatively constant room temperature environment, to remove the dark current noise
of the CCD detector, a dark spectrum can be pre-stored in the computer and then
subtracted from each of the measured spectra. In Figure 4.31, the spectrum is measured
with dark current subtraction, readout noise can still be observed. To further improve the
signal-to-noise ratio (SNR) of the optical spectrum measurements for the accurate peak
and valley location determinations, a boxcar algorithm is developed to smooth the raw
data.
Yibing Zhang Chapter 4. System noise analysis and performance optimization 94
Figure 4.31. The measured interference spectrum after dark current subtraction
Since the noise from each of the CCD pixels is independent, noise in each individual
detector pixels can be minimized with information from the adjacent pixels, so that the
spectrum curves can be smoothed. Suppose the measured spectrum is {f (i), i=1,2..,
2048}, the detector pixel i is smoothed using the adjacent –w~+w pixels in the boxcar
smoothing process, the smoothed result ) (i f
s
will be:


+
+

w j
w j
s
j i f
w
i f ) (
1 2
1
) ( (4.36)
where 2w is defined as the smoothing window width.
One important parameter of the boxcar-smoothing algorithm is the window width, which
must be selected very carefully. Obviously, the larger the window width, the better the
performance can be achieved. On the other hand, a large window width will cause two
problems. First, the visibility of the spectrum will decrease when the window width is
increased, this is the character of the averaging algorithm. Second, because the
interference spectrum is not a symmetric function of the wavelength, the averaging
process will induce the deformation of the interference spectrum and shift the peak/valley
positions. These effects are shown in Figure 4.32, Figure 4.33 and Figure 4.34.
Yibing Zhang Chapter 4. System noise analysis and performance optimization 95
Figure 4.32 shows the simulated result for a small smoothing window width (w=5;
OPD=11µm). In this figure, no obvious deformation can be observed, and a high
visibility of the interferogram was achieved.
Figure 4.32 Simulated result for spectrum curve smoothing with small window width
(w=5)
Figure 4.33 shows the simulated result for spectrum curve smoothing with a large
window width (w=80). The visibility of the interference spectrum decreases dramatically
and obvious deformation can also be observed. Figure 4.34 shows the measured
normalized spectrum when a large window width was used (w=61), the spectrum
deformation is very similar to the simulated results shown in Figure 4.33.
Figure 4.33 Simulated result for spectrum curve smoothing with a large window width
(w=80)
800 810 820 830 840 850 860 870 880 890 900
0
0.2
0.4
0.6
0.8
1
Wavelength(nm)
N
o
r
m
a
l
i
z
e
d
S
p
e
c
t
r
u
m
800 810 820 830 840 850 860 870 880 890 900
0
0.2
0.4
0.6
0.8
1
Wavelength(nm)
N
o
r
m
a
l
i
z
e
d
S
p
e
c
t
r
u
m
Yibing Zhang Chapter 4. System noise analysis and performance optimization 96
Figure 4.34 Measured spectrum after curve-smoothing with a window width (w=61)
To obtain the best performance and prevent the deformation of the interference spectrum,
the window width needs to be optimized for different CCD detectors, the light source and
the OPD values of the sensing elements. The basic principles are: the more the CCD
pixels in a unit wavelength space, the larger window width can be used; the smaller the
OPD values of the sensor (which means the space between adjacent peaks/valleys is
large), the larger window width can be used.
Optical noise reduction
When the center wavelength of the optical source shifts, the positions of the valley points
in the interference spectrum are also affected. For the valley points in the interference
spectrum Equation (4.4) becomes:
π
λ
π
λ
π
λ
π
) 1 2 (
2
0 )
2
sin(
1 )
2
cos(
+


k
L
L
L
V
V
V
(4.37)
Yibing Zhang Chapter 4. System noise analysis and performance optimization 97
For the shift in the valley wavelength λ
V
'
, the demodulated OPD is:
λ λ
δ
γ π
λ γ
2 2
3
4
) 1 (
Lw
L L
V v
+
− (4.38)
Choosing the peak point and the valley point to be close to the central wavelength of the
light source, then
0
λ λ λ ≈ ≈
V p
, and Equation (4.13) and Equation (4.38) become
λ λ
δ
γ π
λ γ
2 2
3
0
4
) 1 (
Lw
L L
p
+
+ (4.39-a)
λ λ
δ
γ π
λ γ
2 2
3
0
4
) 1 (
Lw
L L
v
+
− (4.39-b)
With a first order approximation, Equation (4.39) shows that the location changes of the
peaks and the valleys caused by the center wavelength drifts affect the OPD measurement
in opposite manners, thus the wavelength drifts effect can be partially compensated by
combining Equation (4.39(a)) and Equation (4.39(b)), i.e. using both peaks and valleys to
calculate the OPD instead of using peaks or valleys alone,
) (
2
2
δλ
λ λ
O
L L
L
v p
+
+
(4.40)
In analogy to the compensation method used for the center wavelength shifts, the effect
from the bandwidth variations can also be compensated using both the valley points and
the peak points in the interference fringes. Applying Equation (4.37) on the valley points,
the calculated OPD based on the shift in the valley wavelength λ
V
'
is obtained as,
w
p v
w
w L
L L δ
γ π
λ γ
λ 3 2
3
2
) 1 ( +


(4.41)
Both the peak point and the valley point are chosen to be the ones closest to the central
wavelength of the light source, thus
0
λ λ λ ≈ ≈
V p
, and Equation (4.22) and Equation
(4.41) become,
Yibing Zhang Chapter 4. System noise analysis and performance optimization 98
w
p
w
w L
L L δ
γ π
λ γ
λ 3 2
3
0
2
) 1 ( +
+

(4.42-a)
w
p
w
w L
L L δ
γ π
λ γ
λ 3 2
3
0
2
) 1 ( +


(4.42-b)
With a first order approximation, Equation (4.42) shows that the spectral bandwidth
narrowing/broadening effects will also affect the OPD measurement in opposite manners.
Then, by combining equation (4.42-a) and Equation (4.42-b), i.e. using both peaks and
valleys to calculate the OPDs, the spectral bandwidth can also be partially compensated.
) (
2
2
w O
L L
L
v
w
p
w
δ
λ λ
+
+

− −
(4.43)
Mechanical noise reduction
Most likely, the rotation of the sensing element relative to the polarizer is caused by the
environmental mechanical vibrations. To prevent the optical misalignment caused by the
rotation of the sensing element in the sensing probe, a single crystal sapphire inner tube is
used to hold the sensing element tightly in a fixed position. The single crystal sapphire
holder has the same coefficient of thermal expansion (CTE) as that of the sensing
element, thus the holder is stable and no significant thermal stress is generated at
different temperature levels.
4.7 Power budget
The power attenuation factors in the BPDI sensor system include fiber transmission
losses; coupling losses between the optical fiber and the optical collimator; reflection
losses at the surface of the optical components; power losses at the polarizer/analyzer
because light with only one linear polarization direction can pass. These attenuation
factors are shown in Figure 4.35 and their magnitudes are estimated as follows:
Yibing Zhang Chapter 4. System noise analysis and performance optimization 99
LED
Fiber loss
Coupling loss
Polarization loss
Surface reflection loss
Collimation loss
OSA
Figure 4.35. Optical power losses in the BPDI sensor system.
Fiber coupling loss (L
coupling
)
The coupling efficiency between the edge-emitting LED (Honeywell HFE4854-014) and
the multimode optical fiber (200µm core diameter) with a ST connector is 80%, so the
loss is about 1dB.
Fiber loss (L
fiber
)
The attenuation in the multimode fiber is 2.2dB/km at optical wavelength of 850nm; the
total loss in the transmission fiber depends on the length of the fiber used for both
guiding light into the tube and transmitting light to the optical spectrometer.
Collimation loss (L
collimator
)
This loss includes the collimator insertion loss, which is a result of the surface reflection
from the collimation lens, and power loss because of the optical beam divergence (0.3
degree) relative to the ideal collimated propagation direction. The collimation loss in the
selected collimator is about 1dB.
Surface reflection loss of the optical components (L
reflection
)
For single crystal sapphire, each of the optical quality surface reflection loss is 7%; and
for single crystal zirconia, each of the optical quality surface reflection loss is 10%.
The output optical power from the sensing probe should be:
The sensing probe with a SP structure:
P
out
=P
in
(1-L
coupling
) (1-L
fiber
) (1-L
collimator
) (1-L
reflection-sapphire
)
2
(4.44)
Yibing Zhang Chapter 4. System noise analysis and performance optimization 100
The sensing probe with a SDZP structure:
P
out
=P
in
(1-L
coupling
) (1-L
fiber
) (1-L
collimator
) (1-L
reflection-sapphire
)
4
(1-L
reflection-zirconis
)
2
(4.45)
Experimentally, the input power to the SP-structured sensing probe from the 2-meter
multimode fiber is 480.4µW, the output power from the sensing probe after a 2-meter
long multimode fiber is 54.3µW, the power loss in the system is then about 89%, or 9.6
dB.
Chapter 5. Performance evaluation of designed temperature
sensors
With the design, implementation and optimization of the sensor system described in the
previous chapters, this chapter will summarize the experiments and results associated
with the single crystal sapphire based BPDI optical fiber sensor system. Some of the most
commonly used terminology for optical sensing system specifications and performance
characteristics are defined first. The evaluation of the system includes the
characterization of the spectral domain white light signal processing unit, sensor system
calibration, and overall performance evaluations.
5.1 Definitions of performance characteristics
A wide range of terms has been used to describe the essential performance characteristics
of measurement instruments and sensors. The most widely used terms include
repeatability, precision, accuracy, resolution, sensitivity, stability, hysteresis, and
frequency response [68]. There is always a trade off among these specifications that can
be achievable at a given cost. The intention of this section is to provide clear definitions
for the terms used to describe the performance of the BPDI temperature sensing system.
Repeatability
The repeatability of an instrument is an indication of its ability to give the same
measurement results on the same quantity with repeated measurements under the same
conditions. An instrument with good measurement repeatability needs a good design and
should be carefully manufactured, so it can provide the same readout. But an instrument
with a good measurement repeatability, does not necessarily mean that it has a good
accuracy, since it could give the same wrong answer all the time.
Accuracy and precision
The accuracy of an instrument indicates the deviation of the measurement results from
the true value of the measurand, so it is a measure of its ability to tell the truth. Accuracy
101
Yibing Zhang Chapter 5. Performance evaluation of designed temperature sensors 102
can be expressed as a percentage of the full scale readings, or as an absolute value over
all working ranges of the instrument. Accurate calibration on repeated measurement
results against a standard is necessary for a good accuracy for the instrument.
Resolution (Sensitivity)
Resolution or sensitivity of an instrument is defined as the minimum resolvable change in
the value of the measurand. The resolution of an instrument can be interpreted by the
statistical standard deviation of a series of measurements under stable circumstances. It is
common to use twice the standard deviation as the direct measurement of the resolution.
The resolution for an instrument usually has different values for different measurement
scales.
Stability (Drift)
It is defined as the capability of an instrument to maintain the same output within a
specific length of a time period. The stability of an instrument is usually measured by the
quantity of drift compared to a standard, which is a well-calibrated measurement
instrument.
Hysteresis
If there is a difference between outputs for a given value of the measured quality when
the value is approached from above or below, an instrument is said to exhibit hysteresis.
It can be significant when rapid level fluctuations are likely to occur, as the result of
mechanical friction, magnetic effect, elastic deformation or thermal effects.
Frequency response
This is a measure of the sensor’s capability to track dynamic changes of the temperature.
For the theromometer, it is charactrized by a rise time, which is defined as the time
required for the sensor to respond to an instantaneous step function, measured from the
10% to 90% points on the response waveforms [68].
Yibing Zhang Chapter 5. Performance evaluation of designed temperature sensors 103
5.2 Characterization of white light signal processor
By incorporating a spectrometer for the purpose of measuring polarization properties as a
function of the optical wavelength, the optical broadband polarimetric differential
interferometric (BPDI) temperature sensor system provides absolute measurement of the
optical path differences (OPD) between the two orthogonal linearly polarized light waves
in the single crystal sapphire sensing element. The BPDI sensing system extracts
temperature information by calibrated relationship between the OPDs and the
temperatures. The absolute temperature measurement is very attractive for harsh
environment applications because it requires no initialization and/or calibration when the
power is switched on. To make the absolute measurement meaningful, self-compensating
capability is desired so that the optical power fluctuations and fiber transmission loss
variations can be fully compensated.
5.2.1 Capability of compensating optical source power fluctuations
To evaluate the self-compensating capability of this optical temperature measurement
system, the output temperature variations were monitored when the optical source power
was altered by changing its driving current. Theoretically speaking, changing the driving
current of a semiconductor optical source, such as an LED, would also change the
spectrum in addition to the optical power levels. Figure 5.1 shows the measured output
powers under different driving currents. The power increases as the driving current
increases. The spectral characteristics profiles also change with different driving currents,
as shown in Figure 5.2. The distortion of the source spectrum would introduce errors into
the measurement results through the normalization process, since the normalization of the
interferometric signals is carried out with respect to the pre-stored LED output spectrum,
which is fixed. The measured spectral curves after normalization are shown in Figure 5.3
for different driving currents.
Yibing Zhang Chapter 5. Performance evaluation of designed temperature sensors 104
0 20 40 60 80 100
50
100
150
200
250
300
350
400
450
500
Driving current (mA)
P
o
w
e
r
(
µ
W
)
Figure 5.1. Output power levels of the LED with different driving currents
800 820 840 860 880 900
0
500
1000
1500
2000
2500
3000
3500
Wavelength(nm)
R
e
l
a
t
i
v
e
i
n
t
e
n
s
i
t
y
I=10mA
I=20mA
I=30mA
I=50mA
I=70mA
I=90mA
Figure 5.2. Output spectra of the LED with different driving currents.
Figure 5.4 shows the output temperature variations as a function of the normalized optical
power of the optical source (LED). As shown, the temperature variation range is limited
to ±0.45ºC for total optical power changes up to 90%. The measurement results shown in
Figure 5.4 also indicate the contribution from the source spectrum distortions.
Yibing Zhang Chapter 5. Performance evaluation of designed temperature sensors 105
Figure 5.3 Normalized interference fringes for the LED with different driving currents
0 20 40 60 80 100
-0.5
-0.4
-0.3
-0.2
-0.1
0
0.1
0.2
0.3
0.4
Normalized power (%)
T
e
m
p
e
r
a
t
u
r
e
d
e
v
i
a
t
i
o
n
(
°
C
)
Figure 5.4 Temperature deviations vs. optical source output powers
I=103.7 mA
I=98.1 mA
I=88.4 mA
I=58.3 mA
I=43.5 mA
I=28.5 mA
Yibing Zhang Chapter 5. Performance evaluation of designed temperature sensors 106
5.2.2 Capability of compensating optical fiber transmission loss
To evaluate the self-compensating capability of this optical temperature measurement
system, the output temperature variations were also monitored when the attenuation in
the multimode transmission fiber was changed. Keeping the spectral bandwidth and
center wavelength fixed with changes only to the power levels, the observed
interferometric signal after normalization is shown in Figure 5.5. Some of the detecting
pixels in the CCD array are saturated with 0dB and 1.8 dB attenuation, thus the curves
are deformed. Figure 5.6 shows that the temperature variation is in the range of ±0.4ºC
for the transmission fiber attenuation up to 9 dB.
Figure 5.5. Normalized interference fringes for transmission fiber with different
attenuations
0 dB
1.8 dB
2.9 dB
12.6 dB
16.5 dB
6.3 dB
Yibing Zhang Chapter 5. Performance evaluation of designed temperature sensors 107
2 4 6 8 10
-0.3
-0.2
-0.1
0
0.1
0.2
0.3
0.4
Fiber loss (d B)
T
e
m
p
e
r
a
t
u
r
e
d
e
v
i
a
t
i
o
n
(
°
C
)
Figure.5.6. Temperature deviation vs fiber transmission loss
5.2.3 Capability of compensating temperature fluctuations
To prove the effectiveness of the first order compensation method for the spectral center
wavelength drift and bandwidth broadening/narrowing, which mainly are caused by the
environment temperature fluctuations, the whole white light signal processor, including
the optical source and spectrometer, was placed inside a Testequity 1000 temperature
chamber. The interferometric sensing head structure was kept unchanged in a constant
temperature and pressure environment, so the OPD in the sensing element is fixed in
order to evaluate the performance of the white light signal processor alone. The
temperature in the chamber was set to increase by a step of 5°C in the range from 10°C to
45°C then decrease with the same steps. These temperature increase and decrease
processes were carried out twice. The white light signal processor was also tested at room
temperature for about 5 hours between the temperature increasing and decreasing
processes. The whole test lasted for about 90 hours and the results are shown in Figure
5.7.
The top curve in Figure.5.7 shows the temperatures inside the chamber during the test
process, the temperature values were sampled by a K-type thermocouple inside the white
light signal processor. The middle curve shows the OPD fluctuations relative to the true
Yibing Zhang Chapter 5. Performance evaluation of designed temperature sensors 108
OPD values. The measured OPD is obtained by using only the valley points for signal
processing. Its temperature dependence was predicted by Equation (4.38) and verified in
this test. In the temperature range of 10°C to 45°C, the temperature dependence of the
OPD fluctuations is in the range of about ±9nm before using compensation. The bottom
curve shows the OPD fluctuations relative to the true OPD values. The measured OPD is
obtained by using the white light signal processor with Equation (4.40) for the
compensation. Over the entire tested temperature range, the temperature dependent OPD
measurement by the white light signal processor is about ±1.5nm, which is improved
dramatically compared to the ±9nm deviation shown in the middle curve.
0 10 20 30 40 50 60 70 80 90 100
0
20
40
60
0 10 20 30 40 50 60 70 80 90 100
-10
0
10
0 10 20 30 40 50 60 70 80 90 100
-2
0
2
4
Time (hours)
Figure.5.7 Experimental results of temperature compensation with the first order
approximation
5.3 Blackbody radiation subtraction
As discussed previously, with a low coherence, broadband spectral light source
illuminating the polarimeter, the interference spectral signal contains the OPD
information. After normalization of the measured spectral curves ( ) λ I with respect
T
e
m
p
e
r
a
t
u
r
e
(
o
C
)
O
D
D
f
l
u
c
t
u
a
t
i
o
n
(
n
m
)
O
D
D
f
l
u
c
t
u
a
t
i
o
n
(
n
m
)
Yibing Zhang Chapter 5. Performance evaluation of designed temperature sensors 109
to the pre-stored spectrum ( ) λ
s
I of the broadband input light source, the interference
spectrum would form a sinusoidal curve if plotted in the wave numbers, as shown in
Figure 5.8. The normalized interferometric spectral curve consists of a series of maxima
and minima (peaks and valleys of the interference spectral curves) at certain wavelengths.
Based on these peaks and valleys, the OPD values can be calculated.
Figure 5.8. Normalization of the optical spectrum
As temperature increases, the blackbody radiation is superimposed on the optical
interference signals used for the temperature monitoring, and becomes increasingly
stronger with the increasing of temperature. The measured optical spectral curves will
then be elevated with the addition of the undesired background, as shown in Figure 5.9.
Theoretical simulations of the normalized intensity distribution of blackbody radiations at
different temperatures are plotted in Figure 2.12 for the wavelength range of interest,
which shows that the effect of the blackbody radiation becomes more apparent for
temperatures over 1000
o
C.
Figure.5.9. Blackbody radiation effect on the measurements of optical temperature
signatures
800 810 820 830 840 850 860 870 880 890 900
0
500
1000
1500
2000
2500
3000
3500
4000
1350
o
C
1247
o
C
79
o
C
Wavelength (nm)
I
n
t
e
n
s
i
t
y
800 810 820 830 840 850 860 870 880 890 900
0
500
1000
1500
2000
2500
3000
3500
Wavelength (nm)
In
te
n
s
ity
820 830 840 850 860 870 880
0
0.2
0.4
0.6
0.8
1
1.2
Wavelength (nm)
N
o
rm
a
liz
e
d
In
te
n
s
ity
800 810 820 830 840 850 860 870 880 890 900
0
500
1000
1500
2000
2500
3000
3500
Wavelength (nm)
In
te
n
s
ity

(a) Optical interference
signal ) (λ I
(b) Optical source
PSD ) (λ
s
I
(c ) Normalization
) ( / ) ( λ λ
s
I I
Yibing Zhang Chapter 5. Performance evaluation of designed temperature sensors 110
When those background-raised spectral curves are normalized with respect to the ( ) λ
s
I ,
deformed sinusoidal curves are obtained and the calculated OPD values are not accurate,
and therefore the temperature measurements would be inaccurate.
800 900
Wavelength (nm) (a) (b)
Figure. 5.10. Normalized optical spectra measured at different high temperature levels
In order to minimize the influence of the blackbody radiation background, especially for
the situation of temperature over 1000
o
C, a signal generator is used to modulate the LED
driving current with low frequency square-wave signals, thus the output light from the
LED will be modulated and converted into alternating component (AC) signals. The
background from the blackbody radiation is essentially a direct component (DC) signal.
The AC signal from the modulated optical source carries temperature information. Digital
signal filtering techniques, described in Chapter 4 are used to block the DC signal. The
AC signal will be filtered out for further signal processing for temperature measurement.
With the optical source modulation and the digital signal filtering techniques, the
blackbody background is subtracted sufficiently for the high temperature measurements,
as shown in Figure 5.10. The modulation frequency is selected to be 8Hz, such that the
CCD detector in the spectrometer can measure the spectrum without being affected by its
800
o
C
900
o
1000
o
C
1100
o
C
1300
o
C
1400
o
C
1500
o
C
820 830 840 850 860 870 880
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
1300
o
C
1400
o
C
1500
o
C
Wavelength (nm)
N
o
r
m
a
l
i
z
e
d
i
n
t
e
n
s
i
t
y
Yibing Zhang Chapter 5. Performance evaluation of designed temperature sensors 111
“deadtime” for electric charge transfer. The data transfer rate for the CCD array is 13ms
for the full scan into the computer memory, shown in Table 3.1.
5.4 Calibration of BPDI sensing system
In order to use the BPDI sensor system for temperature measurement, it must be
calibrated to relate the OPD values to the applied temperatures. The sensor calibration is
usually conducted by applying known temperatures within its operating range. The one-
to-one relation between the OPD and the applied temperatures forms the calibration curve
which can be stored in the host computer and later used to convert the OPD into the
temperature readings.
5.4.1 Construction of temperature calibration system
A temperature acquiring subsystem is constructed to obtain the real-time temperature for
the calibration purpose.In this subsystem, real time temperature values from a
thermocouple are displayed in the temperature monitor and then acquired by a computer
using a RS232 interface, as shown in Figure 5.11. In the computer, the temperature
values will then be one to one related to the OPD in the sapphire sensing element by
software communication between the temperature acquiring subsystem and the OPD
measurement and calculation subsystem.
Temperature
monitor
Computer
B-type thermocouple
Rs232 interfeace
Figure 5.11. Temperature acquisition subsystem for the calibration purpose
One DPi32-C24 temperature monitor from Omega
©
is used to acquire the temperature
values from the K-type and the B-type thermocouples. A type K thermocouple (with
resolution of 0.05
o
C) is used for temperatures below 200
o
C, and a type B high-
temperature thermocouple (with resolution of 0.2
o
C under 500
o
C and of 0.1
o
C above
500
o
C) is used for temperatures above 200
o
C. An RS-232 serial digital communication
Yibing Zhang Chapter 5. Performance evaluation of designed temperature sensors 112
port is used to communicate with the computer directly. By employing the Microsoft
©
Visual Basic software, a graphical user interface (GUI) is designed to display the real-
time temperature and the related OPDs simultaneously. Each of these temperature values
is uniquely related to one OPD value. As shown in Figure 5.12, the OPD value is
20.10391 µm at temperature 1181.0
o
C.
Figure 5.12. Real time temperature is related to an OPD value through a GUI interface
5.4.2 Temperature sensor calibration
During the calibration process, the sensing element was heated in a high temperature
furnace up to 1600
o
C. Each OPD value obtained from the optical spectral temperature
signature and each of the real-time temperatures acquired from the thermocouple are
related simultaneously and stored in the computer. To ensure the accuracy of the
calibration, thermal equilibrium between the sensing element and the environment is
necessary. The system is thus held at a stable temperature level for about half an hour
before moving to the next level. Figure 5.13 shows the typical applied temperature data,
and Figure 5.14 shows the OPDs recorded from the BPDI sensor system during the
sensor calibration process. The sensing element used in the tests was a single crystal
sapphire disk with a thickness of 1.5mm and an initial OPD value of 25.887µm at room
temperature environment. Figure 5.15 is a plot of the OPD versus the applied temperature
after averaging. By taking the average of the temperatures within the temperature-holding
period, the measurement error is minimized. The one-to-one relation of the applied
temperature and the OPD was then used to find the calibration equation through a
Yibing Zhang Chapter 5. Performance evaluation of designed temperature sensors 113
polynomial curve-fitting. With the minimum root mean square error between the
experiemental results and the fitting-curve, experimental results revealed that the optimal
order of the polynomial curve-fitting is 3:
) ( 7 . 1085 86 . 604 836 . 29 320 . 0
2
3
C d d d T
o
− + − (5.1)
With a single crystal sapphire prism as a sensing element, its initial OPD is 66.894µm
under room temperature, and the calibrated polynomial equation is:
) ( 7 . 2419 06 . 94 66 . 2 011 . 0
2
3
C d d d T
o
+ + − (5.2)
0 100 200 300 400 500 600
0
200
400
600
800
1000
1200
1400
1600
T
e
m
p
e
r
a
t
u
r
e
(
°
C
)
Sample index
Figure 5.13. Applied temperature during the sensor calibration process
Yibing Zhang Chapter 5. Performance evaluation of designed temperature sensors 114
0 100 200 300 400 500 600
20
21
22
23
24
25
26
Sample index
O
P
D
(
µ
m
)
Figure 5.14. OPDs measured with the BPDI sensor system during the sensor calibration
process
20 21 22 23 24 25 26
0
200
400
600
800
1000
1200
1400
1600
T
h
e
r
m
o
c
o
u
p
l
e
t
e
m
p
e
r
a
t
u
r
e
(
°
C
)
OPD values (µm)
data 1
cubic
Figure 5.15. The BPDI sensor system calibration curve for a sapphire sensing disk with
thickness of 1.5mm
Yibing Zhang Chapter 5. Performance evaluation of designed temperature sensors 115
50 55 60 65 70
-500
0
500
1000
1500
2000
T
e
m
p
e
r
a
t
u
r
e
(
°
C
)
OPD(µm)
data 1
cubic
Figure 5.16. The BPDI sensor system calibration curve for a sapphire sensing prism
(equivalent to a sensing disk with thickness of 8mm)
For one sensing element, the accurate calibration curve was usually obtained by taking
the average of several consecutive calibration data to further ensure the accuracy of the
calibration. With the calibrated sensor, the performance of the developed BPDI
temperature sensing system, such as repeatability, accuracy compared to the
thermocouple and temperature measurement resolution can then be evaluated in detail.
5.5 Performance evaluations of BPDI sensing system
5.5.1 Repeatability of the measurements
Repeatability of the sensor measurement can be evaluated by applying temperature to a
certain preset point repeatedly from one direction (increasing or decreasing). The largest
difference of the sensor output readings can be used to specify the repeatability of the
sensor. With a 1.5mm sapphire disk as the sensing element, four consecutive
measurements were performed for the temperature cooling down processes from 1600
o
C
to 280
o
C, the results are shown in Figure 5.17. The temperature values on the x-axis are
the thermocouple reference readouts.
Yibing Zhang Chapter 5. Performance evaluation of designed temperature sensors 116
As shown in Figure 5.18, the maximum deviation between the measured temperature and
the calibrated temperature was within the range of ±2.5
o
C. The normalized repeatability
of the sensor system with respect to its dynamic range is therefore ±0.14% of the full
measurement scale; this represents the measurement precision that this system can
achieve.
Figure 5.17. Repeatability testing results of the temperature measurements
Figure 5.18. Deviation of the measured temperatures with respect to the reference data
200 400 600 800 1000 1200 1400 1600
200
400
600
800
1000
1200
1400
1600
Temperature average from four measurements (
o
C)
M
e
a
s
u
r
e
d
t
e
m
p
e
r
a
t
u
r
e
(
o
C
)
measurement
1
measurement
2
measurement
3
measurement
4
200 400 600 800 1000 1200 1400 1600
-2.5
-2
-1.5
-1
-0.5
0
0.5
1
1.5
2
Temperature average from four measurements (
o
C)
°
T
e
m
p
e
r
a
t
u
r
e
d
e
v
i
a
t
i
o
n
f
r
o
m
t
h
e
a
v
e
r
a
g
e
(
o
C
)
measurement
1
measurement
2
measurement
3
measurement
4
Yibing Zhang Chapter 5. Performance evaluation of designed temperature sensors 117
5.5.2 Evaluation of accuracy
A K-type thermocouple is employed for the system calibration from room temperature up
to 200
o
C, and a B-type thermal couple is used from 200
o
C up to 1600
o
C. The type K
thermocouple has a resolution of 0.05
o
C, with an accuracy of 5 . 1 ±
o
C, type B high-
temperature thermocouple has a resolution of 0.2
o
C under 500
o
C and of 0.1
o
C above 500
o
C, with an accuracy of 2 ±
o
C. With a 3
rd
order polynomial as a calibration curve, Figure
5.19 shows four measurement results, and Figure 5.20 shows their deviations from the
thermocouple reference for each of the measurements, which gives an accuracy in the
range of 6 ±
o
C, and 0.43% for the full measurement range.
0 500 1000 1500 2000
0
200
400
600
800
1000
1200
1400
1600
Temperature from Thermocouple ( °C)
T
e
m
p
e
r
a
t
u
r
e
f
r
o
m
o
p
t
i
c
a
l
s
e
n
s
o
r
(
°
C
)
First trial
Second trial
Third trial
Fourth trial
Figure 5.19. Optical sensing system measurement results vs a B-type thermocouple
measured temperatures.
0 200 400 600 800 1000 1200 1400 1600
-6
-4
-2
0
2
4
6
Temperature from Thermocouple ( Degree C)
T
e
m
p
e
r
a
t
u
r
e
d
e
v
i
a
t
i
o
n
(
D
e
g
r
e
e
C
)
First trial
Second trial
Third trial
Fourth trial
Figure 5.20. Deviation between the temperature measurement results from the B type
thermocouple and the optical sensing system
Yibing Zhang Chapter 5. Performance evaluation of designed temperature sensors 118
5.5.3 Long-term stability tests
Most optical sensors require accurate optical alignments, and mechanical vibration-proof
working environments to maintain its long-term stable operation. The BPDI optical
thermometer takes advantages of both optical fibers and bulk optics to ensure its long-
term stability for industrial field applications. Bulk optics is convenient for alignment and
also relaxes requirements on vibration proof devices. And optical fibers can transmit light
for a long distance with small attenuation, as well as are implemented easily in the
industrial environments because of their small size, light weight and their immunity to the
electromagnetic interference (EMI). The spectral domain optical signal processing
technique also provides better stability than intensity based optical signal processing
techniques.
The long-term stability testing results are shown in Figure 5.21. These experiments were
carried out at different temperature levels: 26
o
C, 300
o
C, 600
o
C , 900
o
C and 1200
o
C. The
temperature readout differences between the thermocouple and the optical thermometer is
in the range of ±2
o
C for tests over 160 hours, which means the normalized maximum
variation is 0.13% of the full dynamic measurement range.
Figure 5.21 (e) also shows the long term performance degradation of the thermocouple
operated at 1200
o
C, while optical thermometer performed well.
(1) Long-term operation (2) Deviation between the thermocouple and the optical sensor
(a) At 26
o
C
0 40 80 120 160 200
21
22
23
24
25
26
27
28
29
30
31
Time (hours)
T
e
m
p
e
r
a
t
u
r
e
(
o
C
)
Thermal-couple output
Sapphire sensor output
0 40 80 120 160 200
-2
-1.5
-1
-0.5
0
0.5
1
1.5
2
2.5
Time (hours)
T
e
m
p
e
r
a
t
u
r
e
d
e
v
i
a
t
i
o
n
f
r
o
m
t
h
e
r
m
o
c
o
u
p
l
e
o
u
t
p
u
t
(
o
C
)
Yibing Zhang Chapter 5. Performance evaluation of designed temperature sensors 119
(1) Long-term operation (2) Deviation between the thermocouple and the optical sensor
(b) At 306
o
C
(1) Long-term operation (2) Deviation between the thermocouple and the optical sensor
(c ) At 619
o
C
(1) Long-term operation (2) Deviation between the thermocouple and the optical sensor
(d) At 915
o
C
0 20 40 60 80 100 120 140 160 180
303
304
305
306
307
308
309
310
Time (hours)
T
e
m
p
e
r
a
t
u
r
e
(
o
C
)
Thermal-couple output
Sapphire sensor output
0 20 40 60 80 100 120 140 160 180 200
616.5
617
617.5
618
618.5
619
619.5
620
620.5
621
621.5
Time (hours)
T
e
m
p
e
r
a
t
u
r
e
(
o
C
)
Thermal-couple output
Sapphire sensor output
0 30 60 90 120 150 180 210
913.5
914
914.5
915
915.5
916
916.5
917
Time (hours)
T
e
m
p
e
r
a
t
u
r
e
(
o
C
)
Thermal-couple output
Sapphire sensor output
0 20 40 60 80 100 120 140 160 180
-2.5
-2
-1.5
-1
-0.5
0
0.5
1
1.5
2
Time (hours)
T
e
m
p
e
r
a
t
u
r
e
d
e
v
i
a
t
i
o
n
f
r
o
m
t
h
e
r
m
o
c
o
u
p
l
e
o
u
t
p
u
t
(
o
C
)
0 20 40 60 80 100 120 140 160 180 200
-2
-1.5
-1
-0.5
0
0.5
1
1.5
2
2.5
Time (hours)
T
e
m
p
e
r
a
t
u
r
e
d
e
v
i
a
t
i
o
n
f
r
o
m
t
h
e
r
m
o
c
o
u
p
l
e
o
u
t
p
u
t
(
o
C
)
0 30 60 90 120 150 180 210
-1.5
-1
-0.5
0
0.5
1
1.5
2
Time (hours)
T
e
m
p
e
r
a
t
u
r
e
d
e
v
i
a
t
i
o
n
f
r
o
m
t
h
e
r
m
o
c
o
u
p
l
e
o
u
t
p
u
t
(
o
C
)
Yibing Zhang Chapter 5. Performance evaluation of designed temperature sensors 120
(1) Long-term operation (2) Deviation between the thermocouple and the optical sensor
(e) At1205
o
C
Figure 5.21. Long-term stability testing results
It must be noted that the complete evaluation of the system stability needs much longer
testing period (for example, one year). Therefore, the stability testing results of 160-hour
period are not quite conclusive.
5.5.4 Sensitivity (resolution) tests
Interferometric sensors have the reputation of being extremely high sensitive by detecting
the differential phase changes. The resolution of the sensor system can be interpreted by
its standard deviation of a series of temperature measurements at one constant
temperature value. It is common to use twice the standard deviation as the direct
measurement of the resolution. The evaluation of the sensor resolution was performed
using a calibrated sensor at room temperature and also at 618
o
C. In the tests, 1000
temperature values were acquired continuously and the resulting histogram is plotted in
Figure 5.22. The standard deviation of the temperature data was calculated to be
σ=0.0348
o
C and σ=0.0343
o
C. Therefore, the resolution of the sensor system was
estimated to be 2σ=0.0696
o
C and 2σ=0.0686
o
C. The normalized resolution with respect
to the dynamic range of the system was 0.005% of the full scale.
0 2 4 6 8 10 12 14 16 18
1185
1190
1195
1200
1205
1210
Time (hours)
T
e
m
p
e
r
a
t
u
r
e
(
o
C
)
Thermal-couple output
Sapphire sensor output
0 2 4 6 8 10 12 14 16 18
-5
0
5
10
15
20
Time (hours)
T
e
m
p
e
r
a
t
u
r
e
d
e
v
i
a
t
i
o
n
f
r
o
m
t
h
e
r
m
o
c
o
u
p
l
e
o
u
t
p
u
t
(
o
C
)
Yibing Zhang Chapter 5. Performance evaluation of designed temperature sensors 121
27.76 27.78 27.8 27.82 27.84 27.86 27.88 27.9 27.92 27.94 27.96
0
50
100
150
200
250
Temperature ( Degree C)
618.05 618.1 618.15 618.2 618.25 618.3
0
50
100
150
200
250
Temperature (Degree C)
(a) At room temperature (b) At 618.5
o
C
Figure 5.22: Histogram of temperature measurement
5.5.5 Hysteresis in the temperature measurements
Hysteresis of the temperature sensor can be measured by cycling the applied temperature
between the minimum and the maximum of the operating range in both increasing and
decreasing directions. The hysteresis can be calculated as the largest difference among
the output readings of the temperature cycles.
To accurately measure the temperature in a coal gasifier, it is important to reach the
temperature equilibrium between the host media and the sapphire sensing element, which
is enclosed by a sapphire tube as protection housing. The process to build the thermal
equilibrium is determined by the heat transfer between the sensing element and the host
media, which highly depends on the thermal properties of the sapphire material, such as
thermal conductivity, heat capacitance, and the dimensions of the sensing element.
protection housing as well as the temperature levels. Since all of these contribution
factors are temperature dependent, a well-designed thermal model is necessary to analyze
the hysteresis of the system. The theoretical descriptions and experimental data of the
thermal properties of the single crystal sapphire are reasonably well known at low
temperature levels, but its thermal properties at high temperature levels have not been
extensively studied so a pure theoretical description is not possible, the hysteresis has to
be derived experimentally.
Yibing Zhang Chapter 5. Performance evaluation of designed temperature sensors 122
21 22 23 24 25 26 27
200
400
600
800
1000
1200
1400
1600
OPD (µm)
T
e
m
p
e
r
a
t
u
r
e
(
°
C
)
Temperature
Increading
Temperature
Decreasing
Figure 5.23. Hysteresis of the BPDI temperature sensor
The experimental evaluation was conducted using the high temperature furnace after the
sensor system was calibrated. The applied temperature was first increased to the
temperature of 1580
o
C at a rate of 2
o
C/minute. The temperature was then decreased to
250
o
C at the same rate after it was maintained at 1580
o
C for half an hour. The
measurement results are shown in Figure 5.23. It is shown that the hysteresis can be as
little as 3-5
o
C for the operating temperature over 1200
o
C, which is in the range of the
measurement accuracy of the system, while for temperature below 1200
o
C, the
maximum hysteresis is about 50
o
C.
This experimental result is understandable since the thermal properties of the single
crystal sapphire material, and the dimensions of the sensing element and the protection
housing will not change much at different temperature levels. It is the temperature
dependent heat transfer processes that result in the larger hysterises at low and medium
temperature levels. Usually, three mechanisms of heat transfer—conduction, convection
and radiation—regulate the sapphire probe’s temperature in a coal gasifier. Gases have
relatively low thermal conductivities so conduction will not be important at all
temperature levels. The convection heat transfer will dominate the heat transfer to the
sensing element at low and medium temperatures (up to a few hundred degrees) because
of the gases circulation in the chamber. At high temperatures (above ~800ºC), radiation
becomes stronger for the heat transfer between hot walls and protruding probes in
Yibing Zhang Chapter 5. Performance evaluation of designed temperature sensors 123
addition to the convection. Radiation transfer is defined fundamentally by the Planck
function, and is proportional to the fourth power of the temperature. The single crystal
sapphire protection housing can absorb the radiation at its exterior surface and the
absorbed radiation thermal energy is then conducted to the sensing element to reach the
thermal equilibrium. It also transmits electromagnetic energy from a wavelength of 0.25
µ m in the UV range to the visible range of 6 µ m in the IR range with a transmission
larger than 80%, so the sensing element can also absorb the radiation energy directly.
Radiation plus convection can then transfer thermal energy to the sensing element more
efficiently than convection heat transfer alone, then the thermal equilibrium can be
reached fast enough at the high temperature level, and no large hysteresis can be noticed.
The goal for the designed sensor is to measure high temperature over 1200
o
C in coal
gasifiers. These testing results show that the hysteresis can be negligible at those
temperature levels.
5.5.6 Frequency response
The frequency response of a sensing system describes its dynamic response capability. In
certain application environments, such as in gas turbine engines, gas temperature can
reach over 1000
O
C in a few seconds; the dynamic response of the temperature
measurement equipment is critical in such applications. While for the temperature
measurement in coal gasifiers, the temperature variations are in the low frequency range,
and the frequency response is not so critical compared to the requirements on its long
term survivability and stability. In this section, the frequency response characteristic is
evaluated for the developed BPDI temperature sensing system.
The frequency response of the developed sensor system depends on three factors,
including the response time of the sensing element to the dynamic environmental
temperature changes; the heat transfer rates between the sensing element and the host
media; and the signal processing time needed for the temperature information extraction
from the optical signals.
Yibing Zhang Chapter 5. Performance evaluation of designed temperature sensors 124
The response time of the sensing element describes its thermal transient response
characteristic. Assuming that the temperature distribution in the sapphire-sensing element
is uniform at any instant of time, the temperature (T) of the sensing element is an
exponential function of time [68],
t mc hA
env
env
e
T T
T T
) / (
0



(5.5)
where T
env
is the environmental temperature, T
0
is the temperature of the sensing element
at time zero, h is the thermal conductivity of the sapphire material, A is the heat transfer
area of the sapphire material, m is the mass of the sensing element, and c is the heat
capacity of the sensing element. The thermal response rise time for the sensing element is
the time required to achieve a response from 10% up to 90% of the step input of the
temperature, represented by τ , and given by:
hA
mc
t t
hA
mc
t e
hA
mc
t e
t mc hA
t mc hA
198 . 2 1 2
303 . 2 1 . 0
105 . 0 9 . 0
2
) / (
1
) / (
2
1
− ⇒
|
|
|
|
|
|
|




τ (5.6)
From Equation (5.6), we can see that the rise time depends on the thermal properties of
the sensing element material, its dimensions and mass. These factors set the ultimate
limitation on the frequency response of the sensor.
Usually, the three types of heat transfer mechanisms, including conduction, convection
and radiation, transfer thermal energy to the sensing element with different rates. As
discussed in addressing the hysterises of the system, the heat transfer mechanisms
between the sensing element and the environment are mainly convection and radiation in
a coal gasifier. Convection heat transfer is dominant from room temperature up to several
hundred degrees Celsius, while the radiation and convection heat transfer will co-exist at
high temperature levels. Thus, the frequency response of the sensor will thus be faster at
high temperature levels than at low temperature levels.
The frequency response of the BPDI sensing system also depends on the signal
processing time on the optical temperature signatures. The signal processing time of the
Yibing Zhang Chapter 5. Performance evaluation of designed temperature sensors 125
signal processing unit, which is composed of an optical spectrometer and a computer in
the BPDI sensor system, is determined by the data sampling frequency, time delay
between each electrical scan, and the boxcar width used to smooth the spectral curves, as
well as the speed of digital signal processing in the computer.
It is difficult to predict the rise time of the sensor theoretically by taking all these three
factors into account. Instead, experimental tests are carried out to measure the rise time of
the sensor system. The probe and the sensing element are single crystal sapphire material,
and the thickness of the protection tube is 3.0mm with 25mm inner diameter. The sensing
element is a right-angle prism with a base size of ) ( 25 ) ( 8 length mm width mm × . In the
optical spectrometer, the data sampling frequency is 100Hz, and time delay between each
scanning is 3ms. The boxcar width is 5 pixels for the spectral curves smoothing. The
computer is a Pentinum II desktop with a CPU frequency of 300MHz.
The first test was performed in a boiling water bath. The sensor and the thermocouple for
temperature reference, are kept at room temperature (26.5
o
C), then inserted into the
boiling water (95.4
o
C) instantly while the water was kept boiling to maintain its
temperature, the temperature change can then be approximated as a step function from
26.5
o
C to 95.4
o
C, the temperature increasing process was monitored and plotted in
Figure 5.24. The rise time can then be determined from these curves. The rise time for the
thermocouple is 3.19 seconds, and is 153.96 seconds (2.6 minutes) for the single crystal
sapphire optical sensor. The heat transfer mechanisms in the boiling water are mainly
convection and conduction.
Yibing Zhang Chapter 5. Performance evaluation of designed temperature sensors 126
0 50 100 150 200 250 300 350
20
30
40
50
60
70
80
90
100
110
Time(second)
T
e
m
p
e
r
a
t
u
r
e
(
°
C
)
Thermocouple readouts
optical sensor readouts
Figure 5.24. Rise time characterization of the BPDI sensor system with boiling water
0 500 1000 1500 2000 2500 3000 3500
20
40
60
80
100
120
140
160
180
Time(second)
T
e
m
p
e
r
a
t
u
r
e
(
°
C
)
Thermocouple readouts
optical sensor readouts
Figure 5.25. Rise time characterization of the BPDI sensor system with high temperature
furnace
The second test was performed in a high temperature furnace, which is closer to the
environment in an actual coal gasifier. Here the dominant heat transfer mechanism is
convection at the tested temperature level of 174.4
o
C. The step change of the temperature
is from 26.5
o
C up to 174.4
o
C, as shown in Figure 5.27. The rise time for the reference K-
type thermocouple is 305.65seconds (about 5 minutes), and the rise time for the single
crystal sapphire optical sensor is 1133.01 seconds (about 19 minutes).
Chapter 6 Expansion of the BPDI sensing technology for other
measurements
This chapter presents several different sensor designs and testing results using the BPDI
technology, including a pressure sensor, an angular rotation sensor, and an electrical
voltage sensor.
6.1 Multi-parameter measurement toolbox based on optical
birefringence
In the broadband polarimetric differential interferometric (BPDI) sensing technology,
when a broad spectrum linearly polarized light passes through a polarimeter, the state of
polarization (SOP) of the light is modulated by a sensing element. The output light
becomes a wavelength-encoded signal, which can be detected directly by an optical
spectrometer. This wavelength-encoded signal is actually the signature of the optical path
difference (OPD) between the two orthogonal linearly polarized light beams propagating
inside the sensing element. By relating the OPD to the measurands of interest through
calibrations, the measurements of multi-parameters, such as pressure, rotary
displacement, electrical current/voltage, and materials characterization can be realized.
The sensing schemes are really rooted in the sensitivities of the birefringence and
dimensions of the sensing element to these parameters through corresponding physical,
chemical or biological phenomena.
With the other existing technologies, most of the physical, chemical and biological
parameter measurements are usually carried out in the separate systems, and each
measurement system has its own signal-processing unit. Measurement schemes are
usually quite different. By simply changing the sensing elements, the BPDI sensing
system, with its self-calibration capability associated with the optical polarimetric
characteristics and the spectral domain signal processing method, could be employed to
measure several parameters, with high resolution, great accuracy, and long-term stability.
127
Yibing Zhang Chapter 6 Expansion of the BPDI sensing technology for other measurements 128
The chief advantages of this measurement toolbox are its simplicity and compactness. By
changing different sensing elements, the same system setup (optical source, and signal
processing unit) for the BPDI temperature sensing system described in Chapter 3, can be
used for multi-parameters measurements, either in a one-end portable structure, where a
45-45-90 degree cubic prism is used to reflect light back from the end of the sensing
head, and only one optical polarizer is used to work both as polarizer and polarization
analyzer, as shown in Figure 6.1, or in a two end structure, where two optical polarizers
are used, and the sensing element is sandwiched between them, as shown in Figure 6.2.
Sensor Head
Optical fiber
Coupler Optical source
λ
∆λ
λ
o
Optical
power
Signal Generator
Optical Spectrum
Analyzer
Computer
to Fiber
Fiber
Collimation lens
Polarizer
(Analyzer)
Sensing Element
Right-angle
Prism
0
1
t
SMA connector
Sensing
element
holder
Figure 6.1 An one-end structure version BPDI sensor system
Sensor Head Optical fiber
Optical source
λ
∆λ
λ
o
Optical
power
Signal Generator
0
1
t
Optical fiber
Optical Spectrum
Analyzer
Computer
to Fiber
Fiber
Collimation lens
Polarizer
SMA connector
Sensing
element
and its holder
to Fiber
SMA connector
Analyzer
Figure 6.2. A two-end structure version BPDI sensor system
Yibing Zhang Chapter 6 Expansion of the BPDI sensing technology for other measurements 129
6.2 Pressure sensor with temperature compensation capability
Most conventional optical pressure sensors are based on movable diaphragms, small
Fabry-Pérot interferometers, or fiber micro-bendings. The problem with these pressure
sensors is often the high cross-sensitivity with temperature. The pressure measurement
based on photoelastic phenomena has been proposed and tested [70], in which the
pressure generates or changes the birefringence. At the same time, the dimensions of
the sensing element also varies if the surrounding pressure changes, thus the OPD is
uniquely related to the integrated effects from the changes of both the birefringence
and the dimensions of the sensing element, all related to the surrounding pressures.
The BPDI sensor system interrogates the optical signal by directly detecting the
spectrum instead of optical intensity, which provides immunity to the optical source
fluctuation and transmission losses, thus high stability can be achieved.
By redesigning the structure of a sapphire temperature sensor head, the pressure-
birefringent polarization-based pressure sensors with temperature compensation
capability can be obtained. This new design provides temperature compensation
capability by utilizing a birefringent sapphire prism as a sensing element. A birefringent
sapphire disk is used as a temperature compensator outside the pressure environments.
By positioning these two sapphire elements closely and aligning the fast axis of the
temperature compensator parallel to the slow axis of the pressure transducer, temperature
sensitivity of the system can be fully compensated by the temperature compensator with a
proper size.
Preliminary tests were carried out to test the pressure sensor. The schematic setup of the
sensing head is shown in Figure 6.3. One sapphire prism is put in a pressure chamber and
a white-light system is used to monitor the OPD changes caused by the pressure. As
shown in Figure 6.4 and Figure 6.5, there is a unique relation between the pressure and
the photoelasticity generated OPD in the single crystal sapphire sensing element. Figure
6.6 shows the calibrated pressure sensor measurement results.
Yibing Zhang Chapter 6 Expansion of the BPDI sensing technology for other measurements 130
a-axis
a-axis
a-axis
c-axis
Pressure chamber
Sensing element Temperature compensation
element
Window
Polarizer
Optical fiber collimator
Figure 6.3 Schematic design of BPDI based optical single crystal sapphire high-pressure
sensing head
Figure 6.6. Pressure measurements with the calibrated sapphire pressure sensor
40 60 80 100 120 140 160
40
60
80
100
120
140
160
Pressure from semiconductor sensor (psi)
S
a
p
p
h
i
r
e
o
p
t
i
c
a
l
s
e
n
s
o
r
m
e
a
s
u
r
e
d
p
r
e
s
s
u
r
e
(
p
s
i
)
Figure 6.4. Applied pressure
signals on the sapphire sensing
element
X-axis: sampling points, function
of time.
Y axis: pressure values (Unit: psi)
Figure 6.5. Pressure signals from
the optical sensor
X-axis: sampling points, function
of time.
Y axis: optical path difference
(Unit: micron)
Yibing Zhang Chapter 6 Expansion of the BPDI sensing technology for other measurements 131
6.3 Rotary displacement sensor
Spillman reported the rotary displacement sensor based on an optical retardation plate
[71]. When a retardation plate is sandwiched between two polarizers and rotates along
its fast axis, a “notch” intensity minimum is created in the broadband input optical
signal. The wavelength position of the notch in the detected spectrum envelope is
related to the rotation angle. This reported sensing scheme utilizes only one notch in
the spectral curve, which is determined by the intensity minimum. When the rotation
angle changes, the notch will move in the spectral curve. The measurement range of
the rotary displacement will be limited by the bandwidth of the optical source and the
optical spectrometer measurement range. The accuracy and resolution of the rotary
displacement measurement are limited by the optical source drift, optical fiber
modulation, etc.
O
u
tp
u
t
li
g
h
t
y
d
f
s
B
r
o
a
d
b
a
n
d
in
p
u
t
lig
h
t
P1
P2
Rotary
displacement
sensing
element
x
z
Figure 6.7. Fiber Optic rotation sensor
In the BPDI sensor system, the wavelength modulation produced by the rotated
retardation plate is identical to that produced by the temperature sensor. In the rotation
sensor, the rotation of a retardation plate about its optical axis (fast or slow axis) rather
than the thermal expansion, produce the same effects in the sensing element. With a
proper thickness of the retardation plate, several optical minima and maxima can be
generated simultaneously in the spectrum envelop of the broadband input optical
signal. The BPDI sensor utilizes these special points in the spectral curves instead of
one notch, so high resolution and accuracy can thus be achieved because of its self-
compensation capability to optical source center wavelength drift and bandwidth
Yibing Zhang Chapter 6 Expansion of the BPDI sensing technology for other measurements 132
changes, as well as fiber transmission losses. The rotation displace measurement range
is not limited by the measurement range of the optical spectrometer measurement
range. Based on a single crystal sapphire disk with a thickness of 1.5mm as a sensing
element, a preliminary sensor was constructed and tested. Figure 6.8 shows the
rotation angle measurement results.
0 10 20 30 40
0
5
10
15
20
25
30
35
40
O
p
t
i
c
a
l
s
e
n
s
o
r
m
e
a
s
u
r
e
d
r
o
t
a
t
i
o
n
a
n
g
l
e
(
°
)
Applied rotation angle(°)
Figure 6.8. Rotation angle measurements with the calibrated fiber optic rotation
sensor.
6.4 High electrical voltage sensor
When an electro-optic crystal is exposed to an electric field, polarization states of light
in the electrooptic crystal will be changed due to the Pockels effect or the Kerr effect.
The BPDI sensing technology can also be applied to the electrical voltage
measurement by selecting an appropriate sensing element. Preliminary tests were
carried out with a single crystal lithium niobate by utilizing its Pockels effect, which is
the linear electro-optic effect in the crystal.
As shown in Figure 6.9, in the presence of an electric field along the x-axis, the
refractive index ellipsoid of the lithium niobate becomes [44]:
1 2 2
61 51
2
2
2
2
2
2
+ + + + xy E r xz E r
n
z
n
y
n
x
x x
z x x
(6.1)
Yibing Zhang Chapter 6 Expansion of the BPDI sensing technology for other measurements 133
where 8 . 6 , 28
61 51
r r for low frequency electrical signals. The electric field
generates birefringence in the x-y plane. The magnitude of the induced birefringence
depends on the electric field according to:
x o
E r n n
61
3
∆ (6.2)
A piece of lithium niobate with dimensions of ) 25 ( ) 9 ( ) 9 ( mm z mm y mm x × × is used for
the electric voltage measurement. To minimize the temperature effect and pyroelectric
effect in the crystal, the electric field is applied in the x-axis and the light propagates
along the z-axis [72]. The electrical voltage was measured up to 10KV and results are
shown in Figure 6.10.
O
u
t
p
u
t
l
i
g
h
t
y
B
r
o
a
d
b
a
n
d
i
n
p
u
t
l
i
g
h
t
P1
P2
Electrical
voltage
x
z
Wavelplate
Figure 6.9. Sensing head designs for electrical voltage measurement.
0 2000 4000 6000 8000 10000 12000
0
2000
4000
6000
8000
10000
12000
Applied Voltage (V)
M
e
a
s
u
r
e
d
V
o
l
t
a
g
e
(
V
)
Figure 6.10. Electrical voltage sensor based on the BPDI technology.
Chapter 7 Conclusions and recommendations for future work
By incorporating a spectrometer with a polarimeter for the purpose of measuring
polarization properties as a function of optical wavelengths, an optical broadband
polarimetric differential interferometric (BPDI) sensing technology was developed in this
research. It has been applied for temperature measurement and proven to be able to
provide reliable self-calibrating measurement of temperatures up to 1600
o
C with
excellent repeatability, as well as high resolution. This optical sensing technology
possesses many advantages, such as wide dynamic measurement range, intrinsic
immunity to electromagnetic interference (EMI), high sensitivity, simplicity and
especially long-term stability in harsh environments.
This chapter summarizes the major conclusions obtained during the research. Future
research work to further improve the system performance are also outlined in this chapter.
7.1 Conclusions
In this research work, an optical broadband polarimetric differential interferometric
(BPDI) temperature sensor system was proposed and tested using single crystal sapphire
material, which possesses a high melting temperature (over 2000°C), superior optical
transparency, and ability to resist chemical corrosions. With a simple mechanically-
structured sensing probe, in conjunction with an optical spectrum-coded interferometric
signal processing technique, the single crystal sapphire optical sensor can potentially
measure high temperature in harsh environments with great accuracy, corrosion
resistance and long-term measurement stability. Contrary to intensity based optical
sensing schemes, spectrum measurement based BPDI sensor system encodes temperature
information in the wavelength of an optical signal, which is then decoded by an optical
spectrometer; this sensor guarantees that encoded spectrum data is not corrupted by
intensity fluctuations in the optical signals.
134
Yibing Zhang Chapter 7 Conclusions and recommendations for future work 135
Due to the crystallographic arrangement of atoms in single crystal sapphire, the material
exhibits an inherent birefringence: the propagation speeds of light in the
crystal are different depending on their polarization directions. The BPDI sensing system
provides absolute measurement of the optical path difference (OPD) between two
orthogonal linearly polarized light beams in the single crystal sapphire used as a sensing
element. The OPD values are determined by the birefringence and the dimensions of the
sapphire sensing element. Since both the birefringence and the dimensions of the sensing
element are temperature dependent, OPD values can be related to temperature by
calibration. The mathematical models of the sensor in response to the temperature have
been studied in order to provide guidelines for optimal design of the sensor.
Detailed implementation of the BPDI sensing system has been described, including signal
processing unit and sensing probe assembly. High performance sensor probes integrate a
sensing element and an optical polarizer/analyzer with a fiber collimator to form a
compact unit, thus the mechanical stability is improved. The sensing head can either be a
SP structure including one right-angle Sapphire Prism with a special crystallographic
orientation, acting both as a sensing element and as a light reflector, or a SDZP structure
including a single crystal Sapphire Disk (as a sensing element) and a right-angle single
crystal Zirconia Prism (as a light reflector). A single crystal sapphire tube is employed to
protect the sensing element and hence increase its survivability in harsh environments.
Testing results show that the sensor probe exhibits excellent reliability and stability
needed to survive the high temperature (up to 1600
0
C) environments. Spectral domain
white light interferometry was employed to interrogate optical temperature signature in
real-time. The integration of white light interfermetry signal processing scheme and the
specially designed single crystal sapphire sensor probe augment the sensor’s survivability
when operating in the harsh environments.
In order to achieve all the potential advantages of the BPDI technology, a detailed noise
analysis has been presented to gain a better understanding of the performance limitation
of the BPDI sensor system. The optical noise analysis of the system identifies the largest
error results from the rotation of the sensing element relative to the optical polarizer. The
Yibing Zhang Chapter 7 Conclusions and recommendations for future work 136
optical noise analysis also indicates that the change of the spectral characteristics of the
optical source, such as spectrum wavelength drift and bandwidth broadening/narrowing
contributes to the error of the system. The fiber bending induces spectrum wavelength
shift and bandwidth changes that can affect the system performance as well. The
optoelectronic noise analysis indicates that the 1/f noise and the thermal noise together
form the ultimate limit on the optical spectrometer performance.
Based on the system noise analysis results, optimization of measurement techniques is
suggested to improve the system performance. The algorithm used to perform the signal
demodulation and the OPD calculations are optimized to compensate the error due to
spectrum changes and the optoelectronic noises.
Comprehensive experiments have been performed to systematically evaluate the
performance of the optical signal processing unit and the sensor system, including
measurement repeatability and accuracy, long-term system stability, measurement
resolution, self-calibration capability, hysterisis, and frequency response. Measurement
repeatability of 5 . 2 ±
o
C was obtained. The measurement accuracy is about 6 ±
o
C. By
keeping the sensor head at certain temperature levels for a relatively longer time period,
the long-term drift, which is the deviation of the temperature readouts from the designed
optical temperature measurement system compared to the thermocouple readouts, were
evaluated to be within the range of 2 ±
o
C. Resolution of the system was estimated to be
about 0.07
o
C. The excellent self-calibration capability of the system was tested, which
proves that the system is relatively immune to transmission fiber losses, optical source
power fluctuations and spectrum distortions. For high temperatures over 1200
o
C the
hysterisis of the BPDI sensor system is negligible. The rise times of the system are also
evaluated on two different temperature levels.
In conclusion, the BPDI sensor system offers the following major advantages over other
sensors designed for the high temperature applications:
1. BPDI technology extracts absolute temperature information by absolute
measurement of phase delays between two orthogonal linearly polarized light
Yibing Zhang Chapter 7 Conclusions and recommendations for future work 137
waves, which is more attractive for applications in harsh environments, because
of no requirement of initialization and recalibration when power is switched on
and off.
2. The BPDI optical temperature measurement system takes advantages of both fiber
optics and bulk optics to simplify the design of the sensing head. Bulk optics is
convenient for reducing the required tolerances on optical alignment and also for
reducing the sensitivity to mechanical vibrations. The optical fibers can transmit
light over a long distance with small attenuation and can be easily implemented in
industrial environments because of their small size, light weight and immunity to
electromagnetic interference (EMI).
3. The BPDI sensor system offers high measurement resolution (better than 0.1
o
C)
because of its intrinsic interferometric characteristics.
4. The BPDI sensor system monitors temperatures in real-time through measuring
optical spectrum instead of optical intensity. This guarantees its relative immunity
to optical source power fluctuations and fiber transmission losses, thus providing
a high degree of long-term measurement stability.
5. The BPDI temperature sensor provides a wide dynamic measurement range. It
was tested from room temperature up to over 1600
O
C.
7.2 Suggestions for future works
A complete and systematic performance evaluation of this sensor provides feedback
information for further improving the system design. Based on the work done so far,
future works are suggested to further enhance the robustness and reliability of the BPDI
temperature sensor systems, to achieve the final goal of conducting accurate temperature
measurements in practical harsh environments, especially in the coal gasifier. An outline
of further investigation is given below.
Yibing Zhang Chapter 7 Conclusions and recommendations for future work 138
1. Sensor protection
Temperature, pressure, and harsh chemicals are commonly used in the industrial
processes; however, they degrade conventional sensor probes quickly. Temperature-
related degradation mechanisms include: thermal shock, thermal cycling, thermal stress,
thermal fatigue, and high heat fluxes. All of these factors must be considered in the
elevated temperature environments.
For the temperature measurement in the coal gasifier, although the corrosion resistance of
the single crystal sapphire material has been documented by laboratory testing, special
consideration should be given to the required mechanical protection of both the sensor
probe and the associated signal demodulation system, since the sensor probe is subjected
to the extreme mechanical stress when the gasifier is heated up from room temperature to
2800F. The same stresses will also be experienced when the gasifier is shutting down and
cooling down. If the sensor needs to be retrieved, it needs to be designed to handle the
shutdown as well as heatup processes. An ideal sensor probe should contain an
inexpensive and rugged shell, marginally larger than the probe. The shell can be made of
high temperature metals for the mechanical stress protection on the inner sensing element
and housing tube.
2. Sensor assembly
The sensor probe material not only has to survive the environment, but also it has to be
compatible with the thermal expansion of the inside sensing element and the joining or
sealing techniques. An optical probe must consist of a corrosion-resistant body
hermetically sealed to a sapphire sensor head, so that the sapphire sensor head can be
physically exposed inside the coal gasifier for temperature measurements. The whole
probe design and assembly must address the problem of the operating temperature and
heat flux to the sensing element. Other concerns are the operating temperature of the
probe, the internal diameter of the probe, and the outside diameter of the envelope shell.
External constraints consist of the port size in the refractory wall of the coal gasifier
Yibing Zhang Chapter 7 Conclusions and recommendations for future work 139
chamber, directness of the access, reactivity and temperature of the environments, period
of use, and other optical requirements.
In order to install the designed sensing system in the coal gasifier, slagging conditions at
the temperature measurement points, length into the hot zone, width of the refractory
wall, and location of electronics, need to be specified for the system integration with the
existing coal gasification facility.
3. Improvement of the signal processing in digital domain
The performance of the BPDI sensor system can be further improved with filtering
techniques in the digital domain by the host computer. Digital filters can be implemented
to reduce the blackbody radiation background noise. The electronic noise associated with
the spectrum measurements can also be minimized with digital filtering techniques, thus
the temperature measurement accuracy and resolution can be improved.
4. Field evaluation
Based on the requirements of industrial fields, a complete field test of the system is
necessary to comprehensively evaluate the overall performance of the BPDI sensor
system. The field test results should also serve well to further optimize the sensor and
system design for the successful commercialization of the BPDI sensing technology.
For the field tests, consideration of the required environmental performance in a slagging
gasifier was a prime concern in the design of the optical sensor components, which must
physically reside in the gasification unit. The sensor instrumentation should be made for
continuous operation in an actual coal gasification facility under high temperatures and
extremely corrosive conditions. The field test should run for multiple months in order to
be able to reliably demonstrate the capability of the operation in extremely harsh
gasification environments.
Reference
1. E. Udd, Fiber optic sensors: an introduction for engineers and scientists, John
Wiley & Sons, Inc., 1991.
2. R. Baierlein, Thermal physics, Cambridge University Press, 1999.
3. E. Magison, “Temperature measurement-physical principles underlie the four
common methods,” Industrial temperature measurement, November 2001, pp:
39-41
4. R. E. Bentley, “Temperature and humidity measurement,” Handbook of
temperature measurement, vol 1, Singapore; New York : Springer, 1998
5. L. C. Lynnworthand and E. H. Carnavale, “Ultrasonic thermometry using pulse
techniques,” In H. H. Plumb, editor, Temperature: its measurement and control in
science and industry, vol 4, Inst. Soc. America, Pittsburgh, 1972, pp: 715-732.
6. D. W. Varela, “Temperature measurement in industrial and laboratory furnace
using ultrasonic thermometry,” In J. F. Schooley, editor, Temperature: its
measurement and control in science and industry, vol 6, Amer. Inst. Phys. New
York, 1992, pp: 1027-1032.
7. S. F. Green, “Acoustic temperature and velocity measurement in combustion
gases,” 8
th
Int. Heat Transfer Conf., San Francisco, August 1986
8. S. M. Vaezi-Nejad, “Selected topics in advanced solid state and fiber optic
sensors,” Published by the Institute of Electrical Engineers, London, United
Kingdom. 2000, IEE circuits, devices and systems series 11.
9. D. P. Dewitt and G. D. Nutter, Theory and practice of radiation thermometry,
John Wiley&Sons, Inc. New York, 1988.
10. R. R. Dils, “High-temperature optical fiber thermometer,” J. Appl. Phys, 54(3),
March 1983, pp: 1198-1201.
11. B. E. Adams, “Optical fiber thermometry for use at high temperatures,” In J. F.
Schooley, editor, Temperature: its measurement and control in science and
industry, vol 6 Amer. Inst. Phys. New York, 1992, pp: 739-743.
140
Yibing Zhang Reference 141
12. R. O. Claus, M. F. Gunther, A. Wang, K. A. Murphy, “Exitrinsic Fabry-Perot
sensor for strain and crack opening displacement measurement from –200 to 900
o
C,” J. of Smart Materials and Structures, vol 1, no.3, 1992.
13. H. Xiao, J. Deng, R. May, and A. Wang, “Single crystal sapphire fiber-based
sensors for high-temperature applications,” Proc. SPIE 3201, Bonston, MA,
September 19, 1999.
14. M. Sun, “Fiberoptic thermometry based on photoluminescent decay times,” In J.
F. Schooley, editor. Temperature: its measurement and control in science and
industry, vol 6, Amer. Inst. Phys, New York, 1992, pp: 715-719.
15. Z. Zhang, K.T.V. Grattan, and A.W. Palmer, “Fiber-optic high temperature sensor
based on the fluorescence lifetime of alexandrite,” Rev. Sci. Instrum.63, 1992, pp:
3869-73.
16. B. W. Noel, W.D. Turley, W. Lewis, K. W. Tobin, and D. L. Beshears, “Phosphor
thermometry on turbine-engine blades and vanes,” In J. F. Schooley, editor.
Temperature: its measurement and control in science and industry, vol 6, Amer.
Inst. Phys, New York, 1992, pp: 1249-54.
17. K. A. Wickersheim, and M. H. Sun, “Fiberoptic thermometry and its
applications,” J. Microwave. Power. 22(2) 1987, pp: 85-94.
18. O. Iida, T. Iwamura, K. hashiba, and Y. Kurosawa, “A fiber optic distributed
temperature sensor for high-temperature measurements,” In J. F. Schooley, editor.
Temperature: its measurement and control in science and industry, vol 6, Amer.
Inst.Phys, New York, 1992, pp: 745-749.
19. A. B. Murphy and A. J. D. Farmer, “Temperature measurement in thermal plasma
by rayleigh scattering,” J. Phys. D: Appl. Phys, 25, 1992, pp: 634-43.
20. A. B. Murphy, “Laser-scattering temperature measurements of a free burning arc
in nitrogen,” J. Phys. D: Appl. Phys, 27, 1994, pp: 1492-1498.
21. J. P. Dakin, D. J. Pratt, G. W. Bibby, and J. N. Rosss, “Distributed optical fiber
Ramman temperature sensor using a semiconductor light source and detector,”
Electron. Lett. 21: 1985 pp: 569-570.
22. H. D. Schilling, B. Bonn, and U. Krauss, Coal gasification: existing processes
and new developments, [2nd rev. ed.], London, Graham & Trotman, 1981.
Yibing Zhang Reference 142
23. E. J. Hoffman, Coal gasifiers, Laramie, Wyo. Energon Co., 1981.
24. M. R. Werner, and W. R. Fahrner, “Review on materials, microsensors, systems,
and devices for high-temperature and harsh-environment applications,” IEEE
trans. Ind. Electron., vol. 48, no. 2, April 2001, pp: 249-257.
25. J. R. Leigh, Temperature measurement and control, Chapter 5, Perter Peregrinus
Ltd, London, United Kingdom, 1988
26. G. B. Lu, and H. C. Yan, “Concurrent measurements of temperature and soot
concentration of pulverized coal flames,” Instrumentation and Measurement
Technology Conference, 2001. IMTC 2001. Proc. the 18
th
IEEE, vol.2 pp: 1221 –
1226.
27. J. Castrellon, G. Paez, and M. Strojnik, “Radiometric analysis of a fiber optic
temperature sensor,” Opt. Eng. Vol. 41 No. 6, June 2002, pp: 1255-1261.
28. K. M. Koo, J. H. Kim, S. B. Kim, H. D. Kim, and D. Y. Ko, “Ultrasonis
thermometry system for measuring very high temperatures using high resolution
signal processing,” Proc. IECIC 2001, Aug 2001.pp: 229-232.
29. E. P. Hastie, and D. Bonnell, “Alkali Vapor Transport in Coal Conversion and
coal combustion systems,” ACS Symposium Series, vol. 179 Metal bonding and
interactions in high temperature systems with emphasis on alkali metals (1982)
30. T. Sun, “Alkali corrosion of ceramic materials”, Ceramic News, 2(4)1-2, 1991
31. J. Schlichting, “Oxidation and hot corrosion behavior of SiN and SiAlON,”
Nitrogen ceramics, Edited by F. L. Riley. Noordhoff, Leyden, 1977, pp: 627-632.
32. J. Smialik and N. Jacobson, “Mechanism of strength degradation for hot corrosion
of alpha-SiC,” J. Am. Ceram. Soc. 59(10), 1986, pp: 741-752.
33. M. Feber, J. Ogle, V. Tennery and T. Henson, “Characterization of corrosion
mechanisms occurring in a sintered SiC exposed to basic coal slags,” J. Am.
Ceram. Soc. 68(4), 1985, pp: 191-197.
34. M. Feber and V. Tennery, “Behavior of tubular ceramic heat exchanger materials
in acidic ash from coal-oil-mixture combustion,” Ceram. Bull, 62(2), 1983, pp:
236-243.
35. D. Mckee and D. Chatterji, “Corrosion of silicon carbide in gases and alkaline
melts,” J. Am, Ceram, Soc 59(9-10), 1976, pp: 441-444.
Yibing Zhang Reference 143
36. T. Sun, Alkali corrosion of coal gasifier linings, Master’s thesis, Virginia
Polytechnic Institute, 1986
37. M. V. Roode and J. Price, “Corrosion resistant coatings for ceramic heat
exchanger tubes operating in highly corrosive environments,” in high
performance ceramic coatings and films, published by Elsevier Science
Publishers, 1991, pp: 625-633.
38. G. R. Pickrell, High Temperature Alkali Corrosion of Low Expansion Ceramics,
Ph.D. Dissertation, Virginia Tech, 1994
39. A. Dandridge and A. D. Kersey, “Overview of Mach-Zehnder sensor technology
and applications,” Proc. SPIE 985, Fiber Optic and Laser Sensors VI, Boston,
Sept. 1988
40. A. Wang, S. Gollapudi, R. G. May, K. A. Murphy, and R. O. Claus, "Sapphire
optical fiber-based interferometer for high temperature environmental
applications," J. of Smart Materials and Structures, 4, 1995, pp: 147-151.
41. A. Wang, S. Gollapudi, K. A. Murphy, R. G. May, and R. O. Claus, “Sapphire-
fiber-based intrinsic Fabry-Perot interferometer” Opt. Lett. vol. 17, no. 14, 1992,
pp:1021-23.
42. J. Schellman and H. P. Jensen. “Optical spectroscopy of oriented molecules,”
Chem. Rev. 87, 1987, pp: 1359-1399.
43. D. S. Kliger, J.W. Lewis, and C.E. Randall. Polarized light in optics and
spectroscopy, Academic Press, Boston, 1990
44. A. Yariv, Optical electronics in modern communications, Chapter 1, Oxford,
1997
45. K. A. Murphy, M. F. Gunther, A. M. Vengsarkar, and Richard O. Claus
“Quadrature phase shifted extrinsic Fabry-Perot fiber optic sensors”, Opt. Lett.
vol. 16, 1991, pp: 273-275.
46. D. A. Jackson, et al, “Pseudo-heterodyne detection scheme for optical
interferometers”, Electron. Lett. Vol. 28, 1982, pp: 1081-1083.
47. D. Hogg, D. Janzen, T. Valis, and R. M. Measures, “Development of a fiber
Fabry-Perot (FFP) strain gauge,” Proc. SPIE vol. 1588, 1991, pp: 300-307.
Yibing Zhang Reference 144
48. G. Stewart,; A. Mencaglia, W. Philp, and W. Jin, “Interferometric signals in fiber
optic methane sensors with wavelength modulation of the DFB laser source,” J.
Lightwave. Technol. vol. 16, no.1, 1998, pp: 43-53.
49. T. K. Gangopadhyay and P. J. Henderson, “Vibration: history and measurement
with an extrinsic Fabry-Perot sensor with solid-state laser interferometry,” Appl.
Opt. vol. 38, No.12, Apr., 1999, pp: 2471-2477.
50. A. Wang, H. Xiao, J. Wang, Z. Wang, W. Zhao, and R. G. May, “Self-calibrated
interferometric/intensity-based optical fiber sensors,” J. Lightwave. Technol. vol.
19, no. 10, 2001, pp:1495-1501.
51. S. A. AI-Chalabi, B. Culshaw, and D. E. N. Davies, “Partially coherent sources in
interferometry,” Proc. 1
st
int conf optical fiber sensors, London, 1983, pp: 132-
135.
52. T. Bosselman, and R. Ulrich, “High accuracy position-sensing with fiber-coupled
white light interferometers,” Proc. 2
st
int conf optical fiber sensors, Stuttgart,
1984, pp: 361-365.
53. J. G. Proakis and D. G. Manolakis, Digital signal processing, third edition,
Prentice-Hall, 1996.
54. D. N. Nikogosyan, Properties of optical and laser-related materials: a handbook,
Chapter 1. Chichester, England; New York : J. Wiley, 1997.
55. R. C. Jones, “New calculus for the treatment of optical systems,” J. Opt. Soc. Am
31, 1941, pp: 488-495.
56. D. C. Harris, Materials for infrared windows and domes, SPIE Press, Bellingham
WA, 1999
57. M. Bass, Handbook of optics-fundamentals, techniques& design, second edition,
vol II. Sponsored by the OSA. McGraw-Hill, 1995.
58. J. B. Wachtman, Jr. and L.H. Maxwell, “Strength of synthetic single crystal
sapphire and ruby as a function of temperature and orientation,” J. Am. Ceram,
Soc. 42, 1959, pp: 432-433.
59. R. L. Gentilman, E. A. Maguire, H.S. Starrett, T.M. Hartnett and H.P. Kirchner,
“Strength and transmittance of sapphire and strengthened sapphire,” J. Am.
Ceram. Soc, 64, 1981, pp:116-117.
Yibing Zhang Reference 145
60. J. W. Fischer, W. R. Compton, N. A. Jaeger and D. C. Harris, “Strength of
sapphire as a function of temperature and crystal orientation.’ Proc. SPIE. 1326,
1990, pp:11-22.
61. D. C. Haris and L.F. Johnson, “Navy mechanical test results from the sapphire
statistical characterization and risk reduction program,” Proc, SPIE 3705, 1999,
pp: 44-50.
62. D. C. Harris, “Overview of progress in strengthening sapphire at elevated
temperature,” Proc, SPIE 3705, 1999, pp: 2-11.
63. F. Schmid, K. A. Schmid, C. P. Khattak, and P. Duggan, “Increase of the strength
of sapphire by heat treatments,” Proc, SPIE 3705, 1999, pp:36-43.
64. F. Schmid, C. P. Khattak, K. A. Schmid, and S. G. Ivanova, “Increase of high
temperature strength of sapphire by polishing, heat treatments and doping,” Proc,
SPIE 4102, 2000, pp:43-51.
65. S. P. Bush and D. A. Jackson, “Numerical investigation of the effects of
birefringence and total internal reflection on Farady effect current sensors”. Appl.
Opt. vol 31, No25. Spetember 1992, pp: 5366-5374.
66. B. Qi, G. Pickrell, et al. “High-resolution white light interferometer and its
application”. Submitted to Opt. Eng. SPIE.
67. J. P. Holman, Experimental methods for engineers, 6
th
edition, McGraw-Hill Inc.
1994
68. A. Tapia-Mercado, J. Khomenko, A.V. Garcia-Weidner, “Precision and
sensitivity optimization for white-light interferometric fiber-optic sensors,” J.
Lightwave. Technol. Vol: 19 no:1, 2001, pp: 70 -74.
69. A. Wang, S. He, X. Fang, X. Jin, J. Lin, “Optical fiber pressure sensor based on
photoeleasticity and its application.” J. Lightwave. Technol. Vol 10, no.10, 1992,
pp1466-1470.
70. W. B. Spillman, Jr. P. L. Fuhr, and P. J. Kajenski, “Fiber optic rotary
displacement sensor with wavelength encoding,” Appl. Opt. 27 (15) 1988, pp:
3081-3085.
71. S. C. Abrahams, Properties of lithium niobate, London; New York: INSPEC,
Institution of Electrical Engineers, 1989.
Appendix A: Determination of the refraction angle and
refractive index corresponding to the extraordinary waves in
the single crystal sapphire
a
-a
b
t
o
θ
0
θ
2
Air
Sapphire
O
K
Figure 1. Light wave refraction at the surface of the A-plane cut single crystal sapphire
When light propagates at the interface of air and a single crystal sapphire with inherent
birefringence (A-plane cut), the refractive index of the extraordinary wave inside the
single crystal sapphire will depend on the incident angles. Shown in Figure 1, for light
incident with an angle θ
0
, the refraction angle is θ
2
and the magnitude of refraction index
is OK , it follows the Snell’s law,
2
2
2
0
2
sin sin θ θ OK (1)
Within the half ellipse (arc –a-b-a), the length of line OK is:
t b t a OK
po K the for s coordiante
t b Y
t a X
K
K
2 2 2 2
2
sin cos
int
sin
cos
+
|
|
|

(2)
Where t is an intermediate variable with no physical meanings. According to the simple
trigonometric function, the relationship between t and θ
2
is:
2
2
tan
tan
tan sin
cos
tan
θ
θ
b
a
t
t b
a
t b
t a
Y
X
K
K
⇒ (3)
By putting Equation (2) and (3) into Equation (1), we have:
146
Yibing Zhang Appendix A 147
2 2 2 2 2 2
0 2
2 2 2
2
2 2
2 2 2 2
2 2
2 2 2 2
2 2
2 2 2
2
2 2 2
2
sin (a cos t b sin t) sin
a b tan t
sin
1 tan t
b a (1 tan ) tan
b tan a 1 tan
b a tan
b tan a
θ + θ
+
θ
+
+ θ θ

θ + + θ
θ

θ +
(4)
Where the following trigonometric functions are used:
2
2
2
2
2
2
tan 1
1
cos
tan 1
tan
sin
θ β
β
β
β
β
β
or t represents
|
|
|
|
|
|
|
+

+

(5)
Utilizing both Equation (4) and Equation (2), then:
2
2
2 2
2
2 2 2
0
2 2 2
0
2 2
1
2
) ( tan
)) ( tan 1 (
)
) sin (
sin
( tan
a b
a b
OK
a b
a
+
+


θ
θ
θ
θ
θ
(6)
Corresponding to the index ellipse of the single crystal sapphire, where
n
eo
is a special n
e
value, it is the n
e
for the normal incident light wave to the surface of the
A-plane sapphire disk. Then Equation (6) is then converted into:
2
2
2 2
2
2 2 2
0
2 2 2
0
2 2
1
2
) ( tan
)) ( tan 1 (
)
) sin (
sin
( tan
o eo
o eo
e
o eo
o
n n
n n
n
n n
n
+
+


θ
θ
θ
θ
θ
(7)
These refraction angles and refractive indices are corresponding to the extraordinary
waves with incident angle θ
0
in the single crystal sapphire.
e
eo
o
n OK
n b
n a

Appendix B: Acronym
A/D card Analog/Digital card
A/D converter Analog/Digital converter
AC signal Alternating Component signal
BPDI Broadband Polarimetric Differential Interefrometry
CCD Charge Coupled Device
CPT Center for Photonics Technology, Virginia Tech
CPU Central Processing Unit
CTE Coefficient of Thermal Expansion
DC Direct Component
SOP State of Polarization
GUI Graphic User Interface
LED Light Emitting Diode
OPD Optical Path Difference
PS Polarimetric Structure
EFPI Extrinsic Fabry-Perot Interferometry
EMF Electromotive Force
EMI Electromagnetic Interference
FFT Fast Fourier Transform
FP Fabry-Perot
FWHM Full Wavelength Half Maximum
IGCC Integrated Gasification Combined Cycle
IR Infrared Radiation
NA Numerical Aperture
PMD Polarization Mode Dispersion
RMS Root-Mean-Square
RTD Resistance Temperature Detector
SP Sapphire prism
SDZP Sapphire Disk Zirconia Prism
148
Yibing Zhang Appendix B 149
SNR Signal Noise Ratio
SCIIB Self-Calibrated Intensity/Interferometric Base
UV Ultraviolet
USB Universal Serial Bus
Yibing Zhang VITA 150
VITA
Yibing Zhang was born in China, 1973. He received his BS degree in physics in 1995 and
MS degree in optics in 1998 from Nankai University, China. From 1998 to 1999, he
studied Physics at Washington State University (WSU), and did research with Dr. Philip
L. Marston on light scatterings at the Physical Acoustics and Optics laboratory,
Department of Physics, WSU. He joined the Center for Photonics Technology at Virginia
Polytechnic Institute and State University (Virginia Tech) in 2000. He has been working
in the field of laser technologies for high-power laser and ultra-short laser pulse
generations, photonic components (FBG, EDFA, fiber coupler, modulator, etc) design
and implementation at the system level. Since 2000, he has worked on optical fiber
sensors, optical fiber communication devices, and 3-D optical imaging systems for
biomedical applications, and optical metrologies for environmental evaluation and
material measurements. He is one of the winners of the Paul E. Torgersen Graduate
Student Research Excellence Awards for year 2003, College of Engineering, Virginia
Tech.

Novel Optical Sensors for High Temperature Measurement in Harsh Environments
Yibing Zhang

(Abstract)
Accurate measurement of temperature is essential for the safe and efficient operation and control of a vast range of industrial processes. Many of these processes involve harsh environments, such as high temperature, high pressure, chemical corrosion, toxicity, strong electromagnetic interference, and high-energy radiation exposure. These extreme physical conditions often prevent conventional temperature sensors from being used or make them difficult to use. Novel sensor systems should not only provide accurate and reliable temperature measurements, but also survive the harsh environments through proper fabrication material selections and mechanical structure designs.

This dissertation presents detailed research work on the design, modeling, implementation, analysis, and performance evaluation of novel optical high temperature sensors suitable for harsh environment applications. For the first time to our knowledge, an optical temperature sensor based on the broadband polarimetric differential interferometric (BPDI) technology is proposed and tested using single crystal sapphire material. With a simple mechanically structured sensing probe, in conjunction with an optical spectrum-coded interferometric signal processing technique, the proposed single crystal sapphire optical sensor can measure high temperature up to 1600 oC in the harsh environments with high accuracy, corrosion resistance, and long-term measurement stability. Based on the successfully demonstrated sensor prototype in the laboratory, we are confident of the next research step on sensor optimization and scale-up for full field implementations. The goal for this research has been to bring this temperature sensor to a level where it will become commercially viable for harsh environment applications associated with industries.

Acknowledgements
I wish to express my deepest appreciation to Dr. Anbo Wang, and Dr. Ahmad Safaai-Jazi, for serving as my advisors. Without their patient, guidance and constant supports, this dissertation would not have been possible. As mentors and friends, they continually and convincingly conveyed a spirit of adventure and an excitement in regard to research and scholarship during the past years. With their dedication, confidence and incredible achievements, they will continue to be my mentors shedding light on my future journey.

I also would like to sincerely thank Dr. Gary R. Pickrell, Dr. Guy J. Indebetouw, Dr. Ira Jacobs, and Dr. Roger H. Stolen, for serving on my committee and for their encouragements and valuable suggestions to improve the quality of the work presented here. I hope Dr. Guy J. Indebetouw get well soon.

My gratitude also goes to all of my colleagues and friends at Center for Photonics (CPT), which has been shared as a home. Among them, special thanks goes to former CPTers, Dr. Russsll G. May, Dr Hai Xiao, Dr. Bing Qi, and Jiangdong Deng, who introduced me to CPT three years ago, as well as current CPTers, Yan Zhang and Xiaopei Chen, for their valuable suggestions and supports to this presented work. I am also grateful to Debbie Collins, Kathy Acosta and Bill Cockey, they have made the CPT a place to work with great pleasure.

Finally, words alone cannot express the thanks I own to my loving wife, Chimge, for her caring, encouragement and sacrifice; to my parents, who raised me and trust me with their endless love; and to my two young sisters, Shudan and Jialing, they remind me that family is more important than any number of academic degrees.

iii

...................................... 12 Chapter 2...............2 Broadband polarimetric sapphire sensor........................................................................... 14 2.............................................................. Design and implementation of temperature sensor systems..... 14 2.................................................4 Configurations of designed temperature sensor systems ......................... 13 2...................2............................................ 21 2..........................1 Review of signal processing methods for interferometric sensors .....................................................2......................................... 4 1............................................................... 1 1......................2 Thermometers based on thermal expansion.......................2 Acoustic methods.. 36 iv ................2 Review of non-optical techniques for temperature measurements ...........6 Major advantages of BPDI sensor system ......................................... 4 1..4 Industrial needs for novel high temperature sensors.......................................3................................................ 30 2.......3 Fluorescence thermometers ....................................................................1 Fabrication materials for sensing probes ............................................1 High temperature thermocouples.6 Scope of research ................................................... vii List of Tables ................................... xiii Chapter 1...... 10 1.............................2 SCIIB signal processor ..............3............ Introduction ..1...................1 Background information on the proposed research ..................... 5 1................5 Special requirements for temperature sensors in coal gasifier.......... 20 2... 3 1....................................2..........................1 Review of EFPI sapphire fiber sensor ...................... 5 1....................................................1.......... iv List of Figures........................3 Spectral domain white light interferometric signal processor ..................................... iii Table of Content. ii Acknowledgements ......................1 Remote pyrometers .......................................... 36 3..................... 17 2....................................................................................................4 Thermometers based on optical scattering...... 27 2..................................................... 2 1..................................... 20 2.............................................................................................. 34 Chapter 3...............3 Blackbody radiation subtraction ........................................................................................... 1 1.........2.....2.......................................... 6 1.................................3 Review of optical techniques for temperature measurements ..3............3.............. Principles of operation of the designed optical sensors...................5 Mathematical model for BPDI temperature sensor .................... 25 2..................................................................................... 2 1....................................2 Signal processing units ....... 6 1.... 22 2...............................................................................1 Optical sensing elements design ..........................................................................Table of Content Abstract................

...................................................................................6 Summary of system noises and their optimization .........3 Total internal reflection in the sensor probe . zirconia prism (SDZP) structure..... 88 4...............3 Blackbody radiation subtraction ..................3 Optical birefringence effects.................................. 77 4...........3............ 49 3...... 107 5.2 Properties of single crystal zirconia............ 71 4...................................2..........4......................................... 44 3...3......5 Performance evaluations of BPDI sensing system .........................................................................4 Overview of BPDI temperature sensor system................... 111 5........... A-axis) ........................ 55 3.......2 Degradation effect due to the visibility of the interference spectrum. 41 3......1 Definitions of performance characteristics .... 52 3...........................2...............................................1.....................2........................................... 49 3................ 101 5.......................................................... 86 4........................................................1 A sensing probe with a sapphire disk........................................3 Capability of compensating temperature fluctuations . 112 5.............2...... 62 Chapter 4........................ 92 4.............. System noise analysis and performance optimization...............1....4. 37 3....3 Rotation about an axis 45º from the slow-axis/fast-axis...................................................................................3 Signal processor and software implementation ........ 64 4...................................... 65 4..................2 Temperature sensor calibration.......................2 Signal processing algorithm development ......... 83 4.........................1 Wavelength drift .............................2 Capability of compensating optical fiber transmission loss ................................................................... Performance evaluation of designed temperature sensors ........... 91 4......1 Capability of compensating optical source power fluctuations .......1..................e...............................2................... 98 Chapter 5................................2 Characterization of white light signal processor...................3 Software design and implementation....1 Construction of temperature calibration system ...........7 Power budget ....................3........... 64 4..................... 103 5..................4 Opto-electronic noise in spectrum measurements .......................................................... 111 5............................... 41 3...................2 Rotation about the fast-axis (i......... 103 5.............................e.......................................................2 Implementation of sensing probes ................................1 Optical spectrum induced noise ............. 106 5....................................... 108 5... C-axis).............2 Spectral bandwidth broadening/narrowing effects ........4 Calibration of BPDI sensing system ........3... 73 4...... 101 5..2 A sensing probe with a sapphire prism (SP) structure........1................................................3........ 48 3...............................3.........................1 Signal processor implementation..................... 40 3..2...........5 Optical fiber induced noises...... 115 v ....3................ 80 4..........1 Rotation about the slow-axis (i................1 Properties of single crystal sapphire ..........................................................

............ 134 7.............2 Suggestions for future works ................................................................... 127 6........4 High electrical voltage sensor................................................ 120 5......5.............5.............................................................................. 148 VITA ..............5 Hysteresis in the temperature measurements................................................... 129 6......................................................1 Multi-parameter measurement toolbox based on optical birefringence ...........................................................5....................................................................................... 123 Chapter 6 Expansion of the BPDI sensing technology for other measurements .....................2 Pressure sensor with temperature compensation capability........................................................................4 Sensitivity (resolution) tests....... 132 Chapter 7 Conclusions and recommendations for future work ...............................5............1 Conclusions.....................................5............................5...................................1 Repeatability of the measurements ........................ 127 6......................... 137 Reference .............. 134 7......6 Frequency response........................... 121 5..................................5............................2 Evaluation of accuracy........................ 131 6............. 146 Appendix B: Acronym........................................................... 118 5....... 117 5....................................................... 115 5...................3 Long-term stability tests .......................................................................... 150 vi .......3 Rotary displacement sensor ................................................................... 140 Appendix A: Determination of the refraction angle and refractive index corresponding to the extraordinary waves in the single crystal sapphire ........

15 Interference curve with blackbody radiation (at 1350 oC) subtraction…… 27 Figure 2...List of Figures Figure 2. 18 Figure.……………..………………………. …………………………………… 26 Figure 2.…….……………. 21 Figure 2.12 Theoretical intensity curves of the blackbody radiation above 1000oC for the wavelength range: 800~900nm.and UV-grade sapphire (window thickness of 0...1 The approximate transmission bands of standard.18.11 Normalized optical interference fringes for different OPD values (with γ =1). 16 Figure 2. 16 Figure 2.16 Temperature measurement systems based on designed sensing elements and detecting units.4 Output spectrum from the EFPI sapphire fiber sensor with cavity length=21. ………………………………………….9 Illustration of a semi-linear operating range of the interference fringes…… 22 Figure 2..7 Measured output optical interference signal I (λ ) from the BPDI sensor … 19 Figure 2..……………………………………………… 27 Figure... ………………………………………… 38 vii ..5 Principle of an interference device using polarized light waves…………… 17 Figure 2..……………....….10 Basic structure of optical fiber sensor with white light interferometry…… 23 Figure 2..………………………………………………………………… 24 Figure 2.……………. 16 Figure 2.……………..13 Relation between blackbody radiation and interference fringes at different temperatures.……….8 Illustration of the principle of the SCIIB fiber optic signal processor…….6 Conceptual schematic design of the sensing head: broadband polarimetric differential interferometry (BPDI) ………………. 26 Figure 2.1 Schematic design of the sapphire fiber based EFPI sensor………………. 2. Interfering mechanism for polarimetric differential interferometer…….17 Schematic design of the single-crystal sapphire based BPDI optical high temperature sensor……………………………………………………. 15 Figure 2.039 inch).3 Output spectrum from the EFPI sapphire fiber sensor with cavity length=6.5µm……………. 2..8µm……………..14 Amplitude response of the IIR filter……………………………………… 27 Figure 2.2. 29 Figure. 31 Figure 3..2 Interference in the FP cavity formed by multimode sapphire fibers……….…………….……….

...……………….…………………..….………75 Figure 4. . 68 Figure 4.8 Block diagram of the implementation of the OPD calculation algorithm….11 Visibilities dependence on the misalignment angle α between the light viii . L =30um.5 OPD measurement uncertainties caused by center wavelength shift (δλ) (γ=1.…………….………………. 66 Figure 4.8 Temperature uncertainty caused by the bandwidth changes (δw) (γ=1... 59 Figure 3.4 Optical propagation in the special crystallographic oriented single crystal sapphire right angle prism. L =30um. 51 Figure 3.……………. 59 Figure 3. L =30um. 43 Figure 3..……………..……….………………73 Figure 4.Figure 3.12 The mass-centroid method for peaks locating in the spectrum…………… 61 Figure 3.w=60nm) ... 71 Figure 4.……….…………….………………. …………………………………………………46 Figure 3.13 BPDI system overview…………………………………………………….6 Temperature uncertainty caused by center wavelength shift (δλ) (γ=1. 47 Figure 3..w=60nm) …...5 Single crystal sapphire right angle prism………………………………….. λp= 857nm.…….1 Gaussian spectral intensity profile from a low coherence sourece (LED) … 65 Figure 4.…………….……………..w=60nm) …... 38 Figure 3.3 A sensing probe with Sapphire Disk Zirconia Prism (SDZP) structure…….9 Misalignment between the optical polarizer and the sensing element. 68 Figure 4. 62 Figure 4. ……..………………...…… 73 Figure. L =30um.7 Single channel spectrum domain white light system……………. 56 Figure 3. λp= 857nm.3 Simulated results of the normalized spectrum with center wavelength shift effect…………. λp= 857nm.7 OPD measurement uncertainties caused by the bandwidth changes (δw) (γ=1.9 Blackbody radiation subtraction……………………………………………..…………………….…………….10.. Decomposition and interference of the linearly polarized input light in the polarimeter. 60 Figure 3.w=60nm)….2 Hexagonal inner structure of single-crystal sapphire material…………….……………… 70 Figure 4.……………...4 Measurement results of the normalized spectrum at 45°C (the reference is acquired at 20°C) ……………. 74 Figure 4.11 Output optical signal from the modulated LED…………………...………. λp= 857nm.2 Central wavelength of the LED dependence on the temperature…………..6 Sapphire Prism (SP) structure sensing probe……………………………….10 Electrical square wave used to modulate the LED……………….4.... 47 Figure 3..

Optical refractive index ellipse for the rotation about the f-axis of the sensing element………….4. 76 Figure 4.……………..12 Experimental interferograms with different visibilities..…………… 85 Figure 4.………………...…………….23.20.….…………….. 86 Figure 4.…… 77 Figure 4.………….19 Sensing element rotation effect on the optical path difference (rotation about the s-axis). ……... .24.. (b) The refraction angle difference between the ordinary wave and the extraordinary wave (rotation about the s-axis) .21.…………….…….……………. 83 Figure 4.. ………….……………… 81 Figure 4.. wavelength for the sapphire sensing element. (a) Refraction angles of the ordinary wave and the extraordinary wave (rotation about the f-axis) (b) The refraction angle difference between the ordinary wave and the extraordinary wave (rotation about the f-axis) ……...22.14 Method of the index ellipsoid to determine the refractive indices of light beams propagating along s direction.87 Figure.…….. Rotation effects on the optical path difference (rotation about the f-axis) ………….….polarization direction and principal axes of the sensing element………….…………….…………….. … 82 Figure4.…. 82 Figure.……………..…………………. 82 Figure 4.27 Effect on the OPD measurements of rotation about the axis at 45º ix .16 Optical refractive index ellipse for the rotation about the s-axis of the sensing element. 78 Figure 4..……………. …………. 84 Figure 4.26 Rotation about the 45º axis relative to the s-axis of sensing element...……………….……………... …………….……………..……………..18 (a) Refraction angles of the ordinary wave and the extraordinary waves (rotation about the s-axis)..………….……….13 Measured temperature deviation caused by the changes in the visibility of the interferogram………….……...4..15 The refractive index ellipsoid for single crystal sapphire………………… 79 Figure4.80 Figure 4... Sensing element rotation effects along s-axis on the OPD measurements.25 Effect of rotation about the f-axis on the OPD measurements.. 85 Figure 4. Refractive index vs.…………….……………. …..…………….…………………… 77 Figure 4. light incident angle for the extraordinary waves..17 Refractive indices and birefringence vs..

…………….. 109 Figure 5.. 104 Figure 5.…………………90 Figure 4.105 Figure 5.....1...99 Figure 5..2..………….…………….……………..105 Figure 5. ……………....…………………….……………..34 Measured spectrum after curve-smoothing with a window width (w=61) ……………..5. 111 x .5 Normalized interference fringes for transmission fiber with different attenuations……………..31 The measured interference spectrum after dark current subtraction……… 94 Figure 4.……………..…………….……………..29 Opto-electronic noise effects on the valley point locations determinations.4 Temperature deviations vs.………………………..……………..……………. 108 Figure 5...…………….6 Temperature deviation vs transmission fiber loss………………………….…………….………….. 96 Figure 4..………………87 Figure 4.. 88 Figure 4.....3 Normalized interference fringes for the LED with different driving currents……………...32 Simulated result for spectrum curve smoothing with small window width (w=5) ……………... 95 Figure 4.7 Experimental results of temperature compensation with the first order approximation…………….8 Normalization of the optical spectrum…………….…………….…………………. ………………..…………….10 Normalized optical spectra measured at different high temperature levels…………….……………. Output spectra of the LED with different driving currents.………………..…………….30 Opto-electronic noise effects on the temperature measurement uncertainties………….35 Optical power losses in the BPDI system..……………. 95 Figure 4..…………….11 Temperature acquisition subsystem for the calibration purpose………….………………………..28 CCD array for the optical signal detection in the optical spectrometer…. 104 Figure 5..…………………. optical source output powers…………………. Output power levels of the LED with different driving currents………….33 Simulated result for spectrum smoothing with a large window width (w=80) …………….. 106 Figure.... 110 Figure 5...……………..…………….…………….relative to the f-axis of the sensing element. ……………. 107 Figure.……….. 109 Figure 5.………………….………………….9 Blackbody radiation effect on the measurements of optical temperature signatures………………..……………. 91 Figure 4..…………….5.

.…………….……………...…………….5 Pressure signals from the optical sensor……………. 128 Figure 6.…………….……………126 Figure 5.12 Real time temperature is related to an OPD value through a GUI interface………………...6 Pressure measurements with the calibrated sapphire pressure sensor….……………..……………. 117 Figure 5..…..16 The BPDI sensor system calibration curve for a sapphire sensing prism (equivalent to a sensing disk with thickness of 8mm) .Figure 5.....…………….……112 Figure 5.………. 130 Figure 6. 120 Figure 5.………………………….. 116 Figure 5.3 Schematic design of BPDI based optical single crystal sapphire high-pressure sensing head……………..…………….……...21 Long-term stability testing results…………….. 113 Figure 5.5mm…...24 Rise time characterization of the BPDI sensor system with boiling water…………….17 Repeatability testing results of the temperature measurements………….. 115 Figure 5. 126 Figure 6.………………… 122 Figure 5.…………….……………..……..4 Applied pressure signals on the sapphire sensing element………………… 130 Figure 6..…………….……………….…………….…………... 114 Figure 5. …………….. 114 Figure 5..15 The BPDI sensor system calibration curve for a sapphire sensing disk with thickness of 1.14 OPDs measured with the BPDI system during the sensor calibration process…………….………………..25 Rise time characterization of the BPDI sensor system with high temperature furnace....…………….23 Hysteresis of the BPDI temperature sensor…………….…...……………...………….…………….……………...…………….…..13 Applied temperature during the sensor calibration process……………….. 130 Figure 6.18 Deviation of the measured temperatures with respect to the reference data…………………….7 Fiber Optic rotation sensor…………….22 Histogram of temperature measurement……………. 128 Figure 6..19 Optical sensing system measurement results vs a B-type thermocouple measured temperatures.……….…………………… 130 Figure 6..….……………………116 Figure 5.….……………..…………………… 117 Figure 5.…………………….......…………………… 131 xi . 121 Figure 5..1 An one-end structure version BPDI sensor system …………….……………..2 A two-end structure version BPDI sensor system…………….20 Deviation between the temperature measurement results from the B type thermocouple and the optical sensing system….

. 133 Figure 6..Figure 6.10 Electrical voltage sensor based on the BPDI technology.. ……………….………………………………… 132 Figure 6. 133 xii . ..……………..9 Sensing head designs for electrical voltage measurement.………….8 Rotation angle measurements with the calibrated fiber optic rotation sensor. ……………….

List of Tables Table 3.50 Table 4.………….1 The key features of the optical spectrum analyzer USB2000…………….93 xiii ....1 Noise in the BPDI sensor system and compensation methods.

Department of Energy. resistance to chemical corrosion. temperature is probably the most measured physical parameter since virtually every process in nature and in industry is temperature dependent. Introduction 1. In general. Its accurate measurement is essential for the safe and efficient operation and control of a vast range of industrial processes. optical sensors have many advantages over conventional electronic sensors for applications in hash environments.S.Chapter 1. avoidance of ground loops. light weight.1 Background information on the proposed research As one of the seven basic quantities used in the SI (International System of Units) system. These include: small size. high pressure and corrosive chemical materials. These advantages have promoted worldwide research activities in the area of optical fiber sensor technologies for harsh environments. Appropriate techniques and instrumentation are needed depending on temperature measurement requirements in different industrial processes and working environments. the Center for Photonics Technology (CPT) at the Bradley Department of Electrical and Computer Engineering 1 . These gasifiers are used in the coal-based power generation industries. Sponsored by the National Energy Technology Laboratory of the U. The motivation of this research is to meet the recent increasing needs for temperature sensors capable of operating accurately and reliably in harsh environments. and potential capability of operating at high temperatures [1]. where a sensor is required to withstand extremely harsh environments imposed by the high temperature. direct high temperature measurement. capability of remote operation. high sensitivity. nuclear power industries. large bandwidth. immunity to electromagnetic interference (EMI). in the primary and secondary stages of slagging gasifiers. glass and metal manufacturing and processing industries and other high-temperature chemically corrosive environments. The sensor prototype developed in this research is intended for non-intrusive. such as coalbased power generation and distribution industries.

the existing non-optical measurement techniques are very limited. Seebeck discovered this thermoelectric effect phenomenon in 1821. radiated energy (radiation thermometer). Global Energy Technology. 1. electromotive force (thermocouple). 1.2 Review of non-optical techniques for temperature measurements In science. so the voltage is accordingly called . such as volumetric expansion (liquid-filled thermometer). German physicist J. depends on the difference in temperature between the hot and the cold junctions as well as the Seebeck coefficient of the two wire metals.Yibing Zhang Chapter 1. Thermocouple For high temperature measurements over 1000 oC. every temperature measurement involves the use of certain calibrated transducers to convert a measurable quantity into a temperature value. Inc. or some other characteristics of a material that varies reproducibly with temperature [3-4]. In practice.2. generated by an electromotive force (EMF) inside the wires. the socalled hot end that will physically locate at the measuring position. temperature is defined in terms of the amount of heat transferred in a Carnot cycle [2]. resistance (resistance temperature detector .RTD). dimensional change (bimetallic thermometer). Introduction 2 of Virginia Tech is currently leading the effort in developing this optical sensor technology for real-time high temperature measurements. in collaboration with an industrial partner. and in practice many different techniques are used depending on the temperature measurement requirements. Those transducers convert changes in the temperature into other measurable physical quantities. the cold junction.T.1 High temperature thermocouples A thermocouple is an assembly of two wires of different metals joined at one end. and at the other end. The open circuit voltage from these two wires. Possible choices include high temperature thermocouples and acoustic methods. This is generally not the most practical way of measuring temperature. that usually works as a reference at 0 oC.

The . Introduction 3 Seebeck voltage. which are formed by different metals. Since the sound wave travels with the gas. Temperature can thus be measured by detecting the speed of sound that propagates inside a material. high temperature thermocouples that utilize precious metals are used and have a limited life of only a few days because of their susceptibility to attack from corrosive chemicals. These are often turbulent which distort the wave fronts. the apparent speed of sound will also be strongly affected by the flow velocity through the Doppler effect. For high temperatures over 1000 oC. making the accurate determination of time of flight difficult. the temperature of the hot junction can be obtained from the measured Seeback voltage. This technology is especially useful for measuring gas temperature in a combustion chamber.2. In chemically corrosive environments. N B. S type thermocouples. Using the gas itself as the temperature sensor overcomes these problems. soot particles slow the acoustic wave significantly. 1. K. S thermocouples are commercially available. The main difficulty with the technique is that the speed of sound is strongly dependent on the composition of the gas along the path. R. Through the known cold junction temperature.Yibing Zhang Chapter 1. Type B. Furthermore. where it is difficult to measure the temperature using inserted probes due to low thermal mass and low conductivity of gases. which is generally not constant in the combustion chamber. Practical systems usually compensate for this effect by performing measurements in both directions. They drift significantly under high temperature environments for a long-term operation. Another problem arises from refraction of the sound wave front by the density and temperature gradients in the chamber. and the strong radiation coupling of the walls of the enclosure to the sensor at high temperatures.2 Acoustic methods It is well known that the speed of sound in a material depends on the temperature [5-6]. including T. R. J. and will result in a large error. Eight types of them are standardized [4]. E. There are more than 300 types of thermocouple available. Different type thermocouples are characterized to measure temperature up to certain levels with different resolutions.

fluorescence thermometers.Yibing Zhang Chapter 1. such as small size. An optical beam is characterized by several variables. The main existing techniques for optical thermometry are remote pyrometers (or radiation thermometers). The measurement of the amount of thermal radiation emitted by a material can therefore be used as an indicator of its temperature. strain. and thermometers based on optical scatterings including Raman scattering and Raleigh scattering. phase and state of polarization.3. Many different physical phenomena related to these characteristics are used to perform sensing functions. immunity to electromagnetic interference. Introduction 4 temperature measurement uncertainty claimed for this method is typically 30oC at 1000 o C [7]. 1. The basic operating principle of the radiation thermometers is to measure part of the thermal radiation emitted by an object and relate it to the temperature of the object using a calibration curve that has been determined either experimentally or theoretically (from Planck’s law) [9]. Temperature sensors probably constitute the largest class of commercially available optical sensors. thermal expansion thermometers. pressure.3 Review of optical techniques for temperature measurements Optical sensors are those instruments in which optical signals are changed in a reproducible way by an external physical stimulus such as temperature. etc. etc.1 Remote pyrometers All materials with temperatures above absolute zero degree emit electromagnetic radiation (thermal radiation) and the amount of thermal radiation emitted increases with temperature. light weight. such as intensity. . this type of instrument dominates the temperature measurement instrument market for temperature over 2000oC. a wide variety of temperature sensors using fiber optics have been developed [8]. They offer several significant advantages over electric sensors. Typical radiation thermometers measure temperature above 600 oC. 1. spectrum. Besides bulk optics based optical sensors.

which is temperature dependent.e a Fabry-Pérot (FP) interferometer in an optical fiber [12].3. Such sensors have found applications in the measurement of temperatures within microwave ovens. such as yttria (Y2O3) [14-16]. Silica fiber based FP temperature sensor has been demonstrated for temperature measurement up to 800 oC.Yibing Zhang Chapter 1. Those sensors suffer from background noise and . The decay time or intensity of the UV stimulated visible fluorescence pulse is guided along the fiber and detected to measure temperatures. Since the fiber diameter can be very small (<1mm) and the sensor diameter is the same as that of the fiber. for use as a radiation thermometer. Adam achieved kHz frequency response with this structure [11]. Some of the most commonly used phosphors are made using a rare-earth element. Introduction 5 Dils [10] proposed a fiber optic version of radiation thermometer. 1.3 Fluorescence thermometers This type of sensor measures temperature by detecting the decay time or intensity of a UV stimulated visible fluorescence pulse. 1.2 Thermometers based on thermal expansion These are sensors that use the temperature dependence of the optical path length in a small optical resonator cavity. The temperature measurement range is dependent on the fabrication materials. thus illuminating a small sample of luminescent phosphor at the end of the fiber. i. such as Gadolinium or Europium.3. or in very high-magnetic field regions. while single crystal sapphire fiber based FP temperature sensor has been demonstrated for temperature measurement up to 1500 oC [13]. In optical luminescence based fiber sensors. doped into a ceramic crystal. such a fiber sensor can measure temperature from 600 oC up to 2000 oC with a fast time response. The strong temperature dependence of the luminescence allows temperature measurement with good accuracy [17]. a UV light source is focused into an optical fiber. by shearing it at an angle and coating it with a metal. These temperature sensors measure the change in optical path length of a short piece of material whose thermal expansion coefficient and refractive index as a function of temperature are known. In this sensor the end of the fiber is used as a small blackbody cavity. such as platinum.

resulting from scattering of light by particles smaller than the wavelength of light. accuracy and range are . These conventional and traditional measurement instruments have been in use for several decades. The ratio of the anti-Stokes to Stokes scattering is thus a sensitive indicator of temperature. results from the scattering of light off phonons. as well as possible systematic errors. at a longer wavelength than the incident light. are well understood. 1.Yibing Zhang Chapter 1. One is called “Stokes scattering”. The intensity of the anti-Stokes scattering depends strongly on the number of sufficiently energetic optical phonons in the crystal. and there are large ranges of suppliers offering equivalent plug compatible equipment. resulting from phonon emission. they are also generally unexpensive. because commonly used traditional techniques are not suitable for specific measurement problems. Introduction 6 require complex electrical detecting units for the weak intensity ultrashort optical pulse signals. which is a strong function of temperature. at a shorter wavelength. Although these conventional instruments are widely available for scientific and industrial applications. innovation and research and development activities in temperature measurements need to be pursued. or vibrational modes of the crystal. The major sources of their instability or drift. and the other. and results in two wavelength-shifted scattered light signals.4 Industrial needs for novel high temperature sensors The majority of commercially available temperature measurement instruments are made using only a few basic types of instruments: liquid-in-glass thermometers. however. thermocouples. 1. depends on both the size and number of scatters present and it is these relationships that enable the effect to be used in a thermometer [19-20]. resistance thermometers and radiation thermometers.4 Thermometers based on optical scattering Another class of optical thermometers employs the temperature dependence of scattered light [18]. or the performances attainable by these traditional techniques concerning measurement sensitivity. is called “anti-Stokes scattering”. Rayleigh scattering. As matured technologies. Raman scattering [21]. resulting from phonon absorption.3.

. plugging up the tap-hole and preventing additional slag from draining out of the gasifier. including low thermal mass sensors for fast time response. air or oxygen monitoring. Operating at a too low temperature would cause the molten slag to become viscous or freeze. certain important physical parameters should be monitored and controlled precisely for coal gasification processes [23]. To optimize performance for these IGCC plants. Hot exhaust from the gas turbine is then fed into a conventional steam turbine. measurement of internal temperature profiles. burning material flow patterns inside the gasifier control and monitoring. producing a second source of power. The specific measurement requirements become necessary in today’s modern industrial environments. high-pressure coal gases exiting a gasifier to power a gas turbine. One representative harsh environment is an entrained flow slagging gasifier. now in the form of a hard vitreous rock. In the new emerging coal-fired power plants for advanced power generation. coal gasification reacts coal with steam using carefully controlled amounts of air or oxygen under high temperatures and pressures. measurements in hostile environments.g. has to be . gas temperature measurement. to melt and become sufficiently fluid to flow out of the gasifier through the bottom tap-hole. EMI (Electromagnetic interference). The gasifier must be operated at a temperature high enough for the ash in the fuel. pressure distribution monitoring in a gasifier. corrosion and intrinsic safety. cooled off. etc. A gasification-based power plant uses the hot. such as real-time accurate and reliable monitoring of temperatures at various locations in a coal gasifier. This unique integrated gasification combined cycle (IGCC) configuration of turbines offers major improvements in power plant efficiencies compared with conventional coal combustion. e. and the slag. where the measurement environment should not be affected by the measurement units. Rather than burning coal directly.Yibing Zhang Chapter 1. and would require adjusting the operating conditions. the coal gasification technique [22] is developed to generate extremely clean electricity and other high-value energy products. which is one of the main units among coal gasification facilities. Eventually the gasifier has to be shut down. radiation. Load changes will also affect the temperature in the gasifier and downstream. such as coal. Introduction 7 limited.

in the gasification plants. Among these are optical and acoustical pyrometers. high temperatures and pressures. and would thus require operating the gasifier at different temperatures to facilitate slag tapping. On the other hand. an infrared transparent high temperature window on the gasifier wall is necessary to maintain a large pressure differential. resulting in a lengthy loss of production. Operating the gasifier at a too high temperature will also reduce the conversion efficiency of the gasification process in the production of the synthesis gas. refuse. The shut down to clean up the slag may take weeks. and corrosive gases. causing plugging problems. operating at a too high temperature would significantly shorten the lifetime of the refractory lining. However. conventional sensors and measurement devices are very difficult to apply [29-33]. such as the boiler. reacting with the ash particles entrained in the gas to form low temperature eutectics which deposit in the cooler sections of the gasifier or on downstream equipment. In addition. Various methods for measuring temperature in harsh environments have been investigated in the past [24-28]. Introduction 8 manually chipped out to be removed from inside the gasifier. Obstruction of the sight-path opening for the pyrometers in the . wood wastes. due to the harsh environment involving entrained molten slag. whereas close to the wall temperature could drop to less than 1316oC (2400F). with many pockets of gas recirculation zones. The ash properties of these various feedstocks vary significantly.Yibing Zhang Chapter 1. The flow condition inside the gasifier is highly turbulent. Temperature in various regions of the gasifier could be widely different. more of the alkali species in the ash would be volatilized. there is an increasing need to utilize a wide variety of feedstock in addition to coals. while allowing transmission of the infrared radiation emitted by the product gases to the detector placed outside the gasifier. and high temperature thermocouples. The temperature at the exit of the burners could be above 1927oC (3500F). In the non-contact optical pyrometers. In order to realize the full economic potential of the gasification systems. Real-time accurate and reliable monitoring of temperatures at various locations in a gasifier is thus highly desirable. such as biomass.. etc.

which deduce gas temperature along a line of sight by measuring the speed of sound along that line. etc.. which could have a thickness of 2 feet or more. the measurement is also subject to interference from the radiation emitted by entrained particles or the relatively cold refractory walls. In addition. forming an insulating layer on the thermocouple. Ceramic materials are also susceptible to attack from alkali vapors in the gas [34-38].Yibing Zhang Chapter 1. The highly corrosive molten slag attacks both metals and ceramics. reliable monitoring of temperature be developed. superior optical transparency. and optical spectrum-coded interferometric signal processing . With a simple mechanicallystructured sensing probe. interferes with acoustic temperature measurements. Direct contact temperature measurement is preferred since it will give the measurement for a specific location. Current high temperature thermocouples that utilize precious metals drift significantly and have a limited life of only a few days. This situation suggests that innovative techniques that can operate in the gasifier harsh environment for real-time. no direct contact-measuring device is available to date due to material issues. In this research program. Introduction 9 refractory wall by molten slag is a major problem. have also been evaluated for use in gasifiers. during heat-up and operation processes could also cause blockage of the sight path. Acoustic pyrometers. However. an optical broadband polarimetric differential interferometric (BPDI) temperature sensor system was proposed and tested using single crystal sapphire material. Shifting of the refractory lining. and ability to resist chemical corrosion. Flow patterns inside the gasifier could thus be deduced from the temperature profile. which possesses high melting temperature (over 2000°C). Noise in the plant or gasifier. soot-blowers. such as from the high velocity burners. and performance of the gasifier could be monitored and improved by making operating adjustments. Several of these devices installed at the critical locations inside the gasifier could provide a temperature profile of the gasifier. Slag buildup around the thermocouple is also a problem.

and high heat flues. Good thermal stability The optic temperature sensors designed for the harsh environmental sensing applications must be thermally stable. thermal fatigue. Temperature-related degradation mechanisms. the single crystal sapphire optical sensor can measure high temperature in harsh environments with great accuracy. Some optical sensors cannot be deployed because of the limitations of the thermal properties on the fabrication materials. To measure high temperatures accurately in a wide measurement range with high resolution. proper fabrication materials are needed. silica optical fibers can only sustain temperatures up to 800°C before the dopants start to thermally diffuse. 1. thermal cycling. the optical temperature sensing probe must be designed and fabricated with enough mechanical strength and with its optical paths entirely sealed to provide the necessary protection. must be considered in the elevated temperature sensor design for long-term stable measurements. depending on the physical locations in the chamber. Solid mechanical structures and special fabrication materials are needed to increase the thermal stability. For example. 3. . 2. including thermal shock. 1. The high temperature is the main reason that renders most electronic sensors inapplicable. corrosion resistance and long-term measurement stability.Yibing Zhang Chapter 1. thermal stress. High pressure survivability Pressures as high as 500 psi can be encountered in the coal gasifier chamber. as well as a simple and stable mechanical structure of the sensing probe.5 Special requirements for temperature sensors in coal gasifier The optical temperature sensors that can be used in the coal gasifier have to satisfy several special requirements as explained below. Introduction 10 techniques. High temperature capability The operating temperature in the coal gasifier is in the range of 1200oC~1600oC. In order to be able to survive in such high pressure environments.

7. Absolute measurement and self-calibration capability Optical temperature sensors with absolute readouts are much more attractive for applications in harsh environments because of their no requiring of initialization and recalibration when the power is switched on. high mechanical strength. In order to achieve successful commercialization. the cost of the sensors and instrumentation is becoming a concern of increasing importance. Introduction 11 4. Chemical corrosion resistance With temperatures exceeding 1200°C. Deployability Optical temperature sensors designed for harsh environment applications must be capable of remote operation and be flexible enough for easy deployment. pressure exceeding 500 psi. optical temperature sensor systems must be robust as well as low cost. Features of mechanical vibration-proof. and chemically corrosive agents such as alkalis. the sensors are required to have self-calibration capability so that the guiding fiber loss variations and the source power fluctuations can be fully compensated. . Cost-effectiveness As the market for optical temperature sensors for harsh environment grows rapidly. or absolute measurement becomes meaningless. 5. commercially available temperature sensors exhibit greatly abbreviated lifetimes due to the harsh environment. 6. it is hard to find a material that is impervious to such an extensive corrosive attack. sulfur. Proper fabrication materials are needed to implement the sensing probe. This requires that the complexity of the sensor system is kept to the minimum and the technique and process of fabricating sensor probes have the potential of allowing mass production. In addition. transition metals and steam.Yibing Zhang Chapter 1. Conventionally. and remote monitoring and control capability are thus necessary.

The experiments and results of the BPDI optical temperature sensor are presented in Chapter 5. Introduction 12 1. Chapter 7 summarizes the research work and proposes future research directions and improvements. as well as a mathematical model for the temperature measurements. innovative techniques are presented for high temperature sensors capable of operating at temperatures up to 1600°C.Yibing Zhang Chapter 1. 1) The study of sensing schemes. Chapter 4 is dedicated to the BPDI sensor system performance analysis and performance enhancement. The efforts devoted to innovative sensing techniques focus mainly on the following issues. 4) Sensor prototype implementation. 2) Cost-effective opto-electronic signal processor design and implementation. The work of implementing the sensor prototypes is reported in Chapter 3. These sensors fulfill the need for real-time monitoring and long-term direct measurement of high temperatures in harsh environments. optoelectronic signal processors for temperature information extraction. Four different optical temperature sensor prototypes were proposed and tested. 3) Fabrication material selection and mechanical packaging structure design of the sensing probe.6 Scope of research In this research. The optical broadband polarimetric differential interferometric (BPDI) temperature sensor system is chosen as an optimal approach for the high temperature measurement. 5) Optimization of the sensor prototype and performance evaluation. 6) Expansion of the developed technologies to sensing applications other than temperature measurement. including sensing schemes. . Chapter 6 discusses the possibility of expanding the application of BPDI technology to sensing applications other than temperature measurements. The principle of sensing technology is presented in Chapter 2.

phase and state of polarization. Section 2. computer software for signal processing calculations and control. high resolution.6.2. also called transducer. This chapter presents the working principles of those separate elements in the optical sensors. state of polarization.5. large dynamic measurement range and long-term measurement stability. the signal conditioning and processing units are presented in Section 2.3 describes the digital signal processing techniques for blackbody radiation background subtraction. One part is a sensing element. a computer. An optical measurement system is typically composed of two basic parts. Many different physical phenomena related to these characteristics are used to perform sensing functions. The other part is a signal conditioning and processing unit. Principles of operation of the designed optical sensors An optical wave is characterized by several variables such as intensity. signal amplifiers. Finally the advantages of the BPDI sensor system are listed in Section 2. the configurations of the potential temperature sensing systems are compared and the BPDI sensor system is selected for further prototype development in Section 2. and data presentation elements. the mathematical model of temperature sensor based on the BPDI technology will be given in Section 2.1. a signal A/D converter.4. supporting electronic circuits. as well as spectrum of the optical wave. This combined sensing scheme guarantees the measurement of absolute high temperature with high accuracy. the temperature information is encoded in the phase. Optical sensing elements utilizing different sensing schemes are described in Section 2. including optical detecting units. In the designed novel optical temperature sensors. It is a unit to convert measurand into measurable physical signals through certain sensing schemes. 13 .Chapter 2. spectrum.

the other is the diffusion of the germanium dopant (used to establish the refractive index gradient between the core and the cladding) from the core region into the cladding. These are Mach-Zehnder[39]. Michelson. Although intrinsic Fabry-Perot sensors have been used in the past to measure several physical parameters including temperature. such as Bragg gratings in or on the fiber. The degradation mechanisms fall into two general categories: one is the crystallization of the silica cladding region which degrades the fiber performance. 2. mechanical vibration. Many techniques have been used to create intrinsic or extrinsic Fabry-Perot cavities. a sapphire air gapped EFPI sensor head is formed by inserting the lead-in single crystal sapphire fiber into the one end of the sapphire tube and inserting .1 Optical sensing elements design Because of its high resolution and accuracy. metal coatings on the end faces of the fiber.1. optical interferomerty methods are widely employed to build optical sensors by encoding measurand information in the phase of the optical wave through the interference phenomena. Optical fiber Fabry-Perot sensors are highly sensitive to temperature. and Sagnac interferometers. By replacing silica fiber with sapphire optical fibers.1 Review of EFPI sapphire fiber sensor Silica fiber based EFPI sensors have shown great promise for measuring temperatures (and other physical parameters) below 800°C and have been designed and tested extensively in the Center for Photonics Technology (CPT). and acoustic waves [40-41]. extrinsic Fabry-Perot interferometers (EFPI) have shown their ability to be insensitive to polarization and only sensitive to axial strain components. At higher temperatures the devices degrade rapidly. There are several different types of interferometric techniques which have been widely used to measure temperature and other physical parameters. the sensors can potentially operate at high temperature (up to 1500°C) environments [13]. Fabry-Perot. which changes the waveguiding properties of the fiber. magnetic fields. and the use of air-glass interfaces at the fiber ends as the reflectors.Yibing Zhang Chapter 2.1. As shown in Figure 2. Principles of operation of the designed optical sensors 14 2. which gives them an advantage over other electro-mechanical and intrinsic Fabry-Perot sensors.

Sapphire tube Target Sapphire Fiber R 1R 2 Lead-in Sapphire Fiber Shattered end High temperature Bonding Figure 2. which makes the sapphire fiber have a very large modal volume. or beam 2 and beam 3) will degrade the interference signals. Schematic design of the sapphire fiber based EFPI sensor Commercially available sapphire fiber is actually a thin sapphire rod using air as a cladding. While one end of the target sapphire fiber is highly polished. Reflectance R of the well polished sapphire fiber end surface is approximately 7% as given by the following equation: (n − n' ) 2 ] (n + n' ) R =[ (2. and n’=1 is the refractive of the air. With these highly multimode fibers. the interference between beams from the same mode order (beam 1 and beam 2) are desired to deduce the cavity length information.1) for wavelength λ=580nm.1. Principles of operation of the designed optical sensors 15 another short piece of sapphire fiber into the other end of the tube to act as a reflecting surface (target fiber). while interference between beams from intermodes (beam 1 and beam 3. n=1. The transmitted light travels through the air gap and is also partially reflected (7%) at the end face of the target sapphire fiber. the inter-modal dispersion in the sapphire fiber degrades the interference fringes further. interference signals are generated with poor visibilities in the FP cavity formed by the ends of the two sapphire fibers. The light propagating in the lead-in sapphire fiber is partly reflected (7%) at the first sapphire-air boundary.Yibing Zhang Chapter 2. the other end is shattered to prevent any significant reflection that would interfere with the measurements. Both the lead-in and the target sapphire fiber ends are polished to optical quality. When the low quality interference signals propagate in the multimode fiber.2. resulting in low interference fringe visibilities. thus it is very difficult to fabricate high quality . As shown in Figure 2.768 is the refractive index of the sapphire fiber.

all of the possible modes in the sapphire fiber can be excited and get involved in the interfering process.8µm . arbitrary units 1000 800 600 400 200 Intensity.4 shows the output spectrum with a cavity length of 21. Principles of operation of the designed optical sensors 16 sapphire EFPI sensing elements for temperature sensing. the lower quality the interference signals. Interference in the FP cavity formed by multimode sapphire fibers The EFPI sensing element was fabricated by Xiao et al with two 100µm diameter sapphire fibers [13]. Those low quality interfering signals can not provide accurate and reliable high temperature measurement in harsh environments. the more modes in the sensing head.2. Figure 2.5µm Figure 2.Yibing Zhang Chapter 2. Figure 2. In industrial environments where mechanical vibration exists.3 shows the output interference spectrum of the sapphire EFPI sensor with a F_P cavity length of 6. Output spectrum from the EFPI sapphire fiber sensor with cavity length=6.5µm. when the cavity length increases. and the sensing F_P cavity is kept on an isolated optical table to avoid exciting higher order modes in the sapphire fiber. one single mode silica fiber guides light into the sensing head. 1400 1200 1200 Intensity. arbitrary units 1000 800 600 400 200 0 -200 750 0 800 Wavelength (nm) 850 900 950 -200 750 800 850 900 950 Wavelength ( nm) Figure 2. Output spectrum from the EFPI sapphire fiber sensor with cavity length=21. As seen in these two figures.3. the visibility of the interference signals degrades significantly.4. d Beam 3 Beam 2 Beam 1 S a p p h ir e fib e r S a p p h ir e fib e r Figure 2.8µm.

Techniques employing polarized light are already widely adopted in the spectroscopic techniques to provide different types of information about atoms or molecules. the material exhibits an inherent birefringence. Principles of operation of the designed optical sensors 17 2.5 makes the general principle of the system clear. One single crystal sapphire disk with inherent birefringence is sandwiched between a polarizer and an analyzer.5. Principle of operation of an interference device using polarized light waves Due to the crystallographic arrangement of the atoms in the single crystal sapphire. An interference system utilizing polarized optical waves generally includes a polarizing device that is located between an optical polarizer which sets the state of polarization at the entrance and a polarization analyzer which makes the interfering states of polarization identical. and ne=1.Yibing Zhang Chapter 2. photo-biological or physical processes.008. Figure 2. the emerging optical intensity is a function of the phase shift Φ. so the birefringence is 0. Since these two waves propagate at different velocities in the polarizing device. The polarizing devices such as phase-shifters or retardators split the incident light wave into two orthogonal linearly polarized waves.2 Broadband polarimetric sapphire sensor Polarization is a fundamental property of light that can be used to measure physical parameters in a sensing system. no=1. For a wavelength λ=589nm. which can be used for temperature measurement. illustrates the working principle of the sapphire sensing head [44].760.1. Polarizer Analyzer Detector Unpolarized light Polarizing interferometric device Figure 2. such as the electronic structure of the molecules. their phases are shifted by a quantity Φ. The single crystal sapphire disk can thus be used as a sensing element in the polarimeter.768.6. The polarization directions of the . The determination of the state of polarization (SOP) in a system can provide information about photochemical. Figure 2. After analyzed by a polarization analyzer. or orientation of the molecules [42-43].

ne and no represent the two refractive indices corresponding to the two orthogonal linearly polarized light waves. In order to recover the phase shift Φ introduced by the disk. the emerging light has to be analyzed through a polarization analyzer. 2. Since both the birefringence and thickness of the sapphiresensing element are functions of the temperature. and λ is the . At the exit of the disk. Principles of operation of the designed optical sensors 18 polarizer and the analyzer are parallel to each other and along the z-axis. by sensing the magnitude of the phase shift. The principal axes (f-axis and s-axis) of the sapphire disk are oriented at 45o with respect to the z-axis. the phases of the two orthogonal states are shifted by a quantity Φ and the emergent state of polarization is usually elliptical. the magnitude of the differential phase shift will also be temperature dependent. given by: Φ = 2π × OPD / λ OPD = ne − no d (2. t gh t li tpu Ou d z f s y x P2 Sapphire disk put d in banht ad g Bro li P1 Figure. Conceptual schematic design of the sensing head: broadband polarimetric differential interferometry (BPDI) The phase shift between the two orthogonal linearly polarized light waves is determined by the optical path difference (OPD) between those two linearly polarized light beams.6. the temperature can be determined.Yibing Zhang Chapter 2.2) where d is the thickness of the sensing element. Therefore. The phase shift is determined by the magnitude of the birefringence and also by the length of the sapphire disk traveled by the two orthogonal linearly polarized light beams.

and thus temperature information can be obtained by calibration. Theoretically speaking. m being an integer) interferences depending on their wavelengths. They are approximated as constant in this wavelength range.7. Utilizing the constructive and deconstructive interference fringes in the broadband spectral domain. the temperature dependent OPD can be obtained by calculations. this sensor design guarantees self-calibration capability. thus the OPDs for all the wavelength are assumed to be equal. hence better measurement repeatability than intensity based optical sensors.7. 3500 3000 Intensity 2500 2000 1500 1000 500 0 800 810 820 830 840 850 860 870 880 890 900 Wavelength (nm) Figure. thus light will experience constructive (corresponding to the phase shift of 2mπ. where the encoded temperature is not corrupted by optical intensity fluctuations from long term operation point of view. 2. . shown in Figure 2. m being an integer) or deconstructive (corresponding to the phase shift of (2m+1)π. The same OPD will generate different phase shifts at different wavelengths. but their differences are very small in the interested wavelength range from 800nm-900nm. ne and no are wavelength dependent.Yibing Zhang Chapter 2.Measured output optical interference signal I (λ ) from the BPDI sensor By encoding and decoding temperature information in the phase of an optical signal. Principles of operation of the designed optical sensors 19 light wavelength.

and Gangopadhyay [49] reported the use of laser wavelength modulation based heterodyne interferometry to demodulate the interference signal of EFPI sensors.Yibing Zhang Chapter 2. It was possible to count fringes bi-directionally by using a quadrature-phase-shifted two-interferometer structure [45]. Principles of operation of the designed optical sensors 20 2. which practically limited the accuracy of measurement. They are self-calibrated interferometric/intensity-based (SCIIB) signal processor. However. using a power meter to detect the intensity of the interference signal generated by the sensor head and an electric counter to count the total number of peaks from a reference point. due to the short initial cavity length of the EFPI sensors. optical interferometry has been used in a variety of applications. Jackson [46].2 Signal processing units 2. For the polarization interferometry sensor. This method has limited measurement range and suffers from the complex structure of the detection unit. intensity based polarization quadrature measurement can be carried out to detect the phase delay between the two orthogonally polarized light waves. the resolution of the wavelength modulation based signal processing method was limited. i. the fringe counting method suffers from a number of problems such as sensitivity reduction when the sensor reaches the peaks or valleys of the fringes. It thus highly depends on the performance of those separate components in the detection unit. However. Steward [48].1 Review of signal processing methods for interferometric sensors As one of the most sensitive measurement methods. More recently. However. and fringe direction ambiguity. In coherent mono-wavelength interferometric systems. the most commonly used signal demodulation method is direct interference fringe counting. including polarimetric sensors. In the CPT.e. there are two signal processors developed to meet the efficient signal processing needs for interferometric sensors. due to the nonlinear and periodic nature of the sinusoidal interference fringes. and spectral domain white light interferometry (also called low-coherence interferometry) signal . Hogg [47]. it was also found difficult to maintain the exact phase difference between these two interferometers in practical applications.2.

8.2. Principles of operation of the designed optical sensors 21 processor. the SCIIB technology can provide absolute measurement of various parameters with the full self-compensation capability for the source power fluctuation and the fiber loss variations.Yibing Zhang Chapter 2. . The light in Channel 1 remains its original spectral width (broadband spectrum) while the light in Channel 2 has a narrower spectrum by passing it through an optical bandpass filter.8. λ Channel 1 Detector splitter ∆λ Channel 2 Detector Signal processing Optical signal from sensing head Figure 2. which can result in a large measurement error when the signal processing methods based on intensity detection are directly applied. the optical signal from the sensing head is split into two channels with different optical properties through optical filtering. Through a proper design of the sensor head. with high resolution and high accuracy. By taking the ratio of these two channels outputs. the self-calibrated interferometric/intensity-based (SCIIB) [50] fiber optic sensor successfully combines the advantages of both the interferometric and the intensity-based fiber sensors in a single system. Illustration of the principle of operation of the SCIIB fiber optic signal processor As shown in Figure 2.2 SCIIB signal processor As a unique signal processing method. the optical source intensity fluctuation and transmission fiber loss can be fully compensated. The signal processing units can reliably measure the optical path difference (OPD) that exists between the two interfering optical waves in an interferometer. as well as self-compensation capability for the optical source power drifting and fiber loss variations. 2.

2 1 0.3 Spectral domain white light interferometric signal processor As shown in Figure 2. Fringe direction ambiguity refers to the difficulty in determining from the optical intensity whether the sensor OPD is increasing or decreasing. such as immunity to the light source power drift and variations of fiber transmission losses.8 1 Change of OPD (um) Figure 2. 2. spectral domain white light interferometric signal processor is adopted for the high temperature measurement application. and high resolution. the sensor head has to be operated only over the semi-linear range of a half fringe.10.2. Because of this limitation associated with the SCIIB signal processor.9. Sem i-Lin ea r Ra n ge . Principles of operation of the designed optical sensors 22 As mentioned previously. 2 1. as shown in Figure 2. such as light emitting diodes (LED).4 1. Its operation principle was first adopted for single mode fiber sensing systems [51] and then in multimode fiber sensing systems [52]. multimode laser diodes or halogen lamps.9.Yibing Zhang Chapter 2.8 Normalized sensor output 1. The white light interferometric fiber optic sensor system inherits most of the advantages of the conventional optical interferometers.4 0.6 1. a white light or low coherence interfeometric technique uses a broadband light source.4 0.2 0. so that a one-to-one quantitative relation between the output intensity and the OPD in the sensor head is obtained.6 0.6 0.2 0 0 0.8 0. To avoid these two problems. Sensitivity is reduced at the peak or valley of the fringe since at that point the change in optical intensity is zero for a small change in the sensor cavity length. regular interferometric sensors suffer from the disadvantages of sensitivity reduction and the fringe direction ambiguity when the sensor output reaches the peak or valley of an interference fringe. Illustration of a semi-linear operating range of the interference fringes.

the optical path difference (OPD) between the two interfering light waves is obtained through interferometric fringe pattern analysis. an optical spectrum analyzer (OSA) is used as the optical signal processing unit.3) where λ0 is the central wavelength. a compact white light signal processor based on the spectrum measurement has been adopted in this research. Principles of operation of the designed optical sensors Optical fiber circulator Optical fiber 23 Broad-band source Sensing intereferometer Optical processing unit Electronic processing unit Figure2. If we assume that the broadband light source (light emitting diode.10. The optical signal is thus converted into an electrical signal and processed in the electronic processing unit. I 0 is the peak intensity value. and ∆λ is the source spectral width. The intensities of the dispersed spectral components of the optical signal are scanned electronically by the CCD array in the OSA. The two-beam interference signal is then given by: . Basic structure of optical fiber sensor based on white light interferometry In the spectral domain white light interferometric signal processor.Yibing Zhang Chapter 2. In the spectrum measurement based white light interferometric system. the spectrum measurement method can provide absolute OPD measurements with high resolution and large dynamic range. For low frequency signal detection (typically less than 100 Hz). Since variations of temperature in the coal gasifier environments are in the low frequency range. LED) in Figure 2.10 has a Gaussian spectral intensity distribution given by  − (λ − λ0 )2  I s (λ ) = I 0 exp   2  (∆λ )  (2. The OSA is composed of a grating and a CCD array.

which may be related to the physical parameters. Normalized optical interference fringes for different OPD values (with γ =1). and stain.8 1. and L represents the optical path difference (OPD) between the two interference beams.5) It is shown in Equation (2.4 0. Principles of operation of the designed optical sensors 24   2πL  I = I s (λ )1 + γ cos   λ   (2.Yibing Zhang Chapter 2. 2 1. and it may decrease because of the coherence length of the optical source and the amplitude difference between the two interference optical waves. the successful demodulation of this spectral signal can render an accurate and absolute measurement of the L values. After normalizing the interference spectrum given by Equation (2.4) where the factor γ is the visibility of the interference fringes.11.2 1 0. The L measurement can also be related to biological or .4) with respect to the source spectrum I s (λ ) . interference fringe patterns are different for different OPD values.2 0 750 L=10µ m L=30µ m 800 850 900 950 1000 1050 Wavelength(nm) 1100 1150 1200 1250 Figure 2.11. pressure. we have the normalized interference output expressed as  2πL  I n = 1 + γ cos   λ  (2.6 1. such as temperature. The ideal value of γ will be γ =1.8 0. As shown in Figure 2.4 Normalized intensity 1.6 0. Because the interference fringe patterns are a function of the optical path difference L.5) that the output spectrum of the sensor is modulated by a cosinusoidal function due to the interference.

not on the size or shape of the cavity or on the material of its wall. As shown in Figure 2. representing radiation emitted into a hemisphere. black) walls at a fixed temperature T. It is these relations that make the spectral domain white light interferomety a powerful tool in the field of optical sensing technologies. which are temperature signatures generated by the light from the broadband optical source. depending only on the temperature and the wavelength. for a cavity with perfectly absorbing (i.Yibing Zhang Chapter 2. This radiation is called blackbody radiation.13. and detected by the spectrometer in the optical temperature measurement system. The intensity of the blackbody radiation becomes stronger with increasing of the temperature.3 Blackbody radiation subtraction As well known. These experimental measurement results show clearly that the blackbody .e. T is temperature in Kelvin [9]. and C2=1. which is the temperature information carrier. such that the total radiation intensity is given by ∫ ∞ 0 K λ dλ and K λ dλ is the intensity due to radiation with wavelength between λ and λ + dλ . K λ is a universal function.12 shows the blackbody radiation theoretical intensity distribution curves at different temperatures.6) λT −1 where C1=3. blackbody radiation will be a strong noise background built upon the optical signal emitted from the LED. Based on the Plank’s blackbody radiation theory. its interior will be filled with radiant energy of all wavelengths. the radiation intensity is specified by a distribution function K λ . Kλ = e C1λ−5 C2 (2. 2. Since the designed temperature measurement system will be applied to a temperature environment as high as 1600oC. blackbody radiation background is superimposed to the optical signals.4388×10^(4). its unit is W/cm2µm-1.7405×10^(4). Figure 2. Principles of operation of the designed optical sensors 25 chemical parameters as well. High temperature sensing systems employ this formula to detect temperature values are called radiation pyrometers.

13. To increase the signal-to-noise ratio in the detected optical temperature signatures. 10 11 Normalized intensity 10 10 10 9 10 8 1000 °C 1200 °C 1400 °C 1600 °C 800 820 840 860 Wavelength(nm) 880 900 Figure 2. Infinite-Impulse Response filter) filtering out the desired AC optical signal from the LED . Principles of operation of the designed optical sensors 26 radiation background increases dramatically with the temperature increase. Theoretical intensity curves of the blackbody radiation above 1000oC for the wavelength range: 800-900nm. Relation between blackbody radiation and interference fringes at different temperatures. Thus the DC background can be subtracted with digital signal processing techniques in the computer. thus the signal-to-noise ratio (SNR) of the measured spectrum is becoming smaller. the LED output light is modulated with a certain frequency and converted into AC signals.12. One method to reduce the blackbody radiation is to utilize a band pass digital signal filter [53] (IIR filter. while the blackbody radiation can be treated as DC signals. Intensity 3500 3000 2500 2000 1500 1000 500 0 0 Intensity 4000 3500 3000 2500 Optical signal o at 79 C Optical signal at 1350 o C Optical signal at 1247 o C Blackbody radiation background at 1350 oC Blackbody radiation background o at 1247 C Blackbody radiation Background at 79 o C 2000 1500 1000 500 0 500 1000 1500 2000 2500 0 500 1000 1500 2000 2500 pixel numbers Pixel numbers (a) (b) Figure 2.Yibing Zhang Chapter 2.

14.5 4 4. the blackbody radiation background is reduced.2 -500 Intensity 0 800 810 820 830 840 850 860 870 880 890 900 0 0.Yibing Zhang Chapter 2.8 0.5 2 2.16.4 3000 2500 2000 1500 1000 500 0 0. Temperature measurement systems based on designed sensing elements and detecting units . and polarimetric sapphire sensor) and signal processors (SCIIB signal processor S e n s i n g e le m e n ts S ig n a l p r o c e s s o r s T e m p e r a tu r e m easurem ent system s E F P I + S C II B EFPI S C I IB E F P I+ W h ite lig h t s y s te m P o la r im e tric s tr u c tu re (P S ) L o w -c o h e r e n c e in te r fe r o m e tr y ( W h ite lig h t s y s te m ) P S + S C II B P S + w h ite lig h t s y s te m Figure 2.6 0.5 3 3.5 1 1. Aplitude Response 1. whose pass band covers 2Hz to 5Hz.5 5 Wavelength(nm) 1/2 frequency (Hz) Figure 2.2 1 0.14.4 Configurations of designed temperature sensor systems By employing the previously described optical sensing elements (EFPI sapphire fiber sensor. Principles of operation of the designed optical sensors 27 only.15.4 3500 Without filter with filter 4000 1. The background from blackbody radiation will be blocked as a DC component in the detected signals. Interference curve with blackbody radiation (at 1350 oC) subtraction 2. Using the digital signal filter. Amplitude response of a IIR filter Figure 2. the designed filter is shown in Figure 2. as shown in Figure 2.15. If the frequency used for the LED modulation is 3 Hz.

Birefringence of the sapphire material. the working point and the working range of the sensor need to be selected properly for the wide dynamic range high temperature measurement. The long-term stability at different temperature environments of each of those two channels is hardly achievable.16. Although this system can work through careful sensing element selection and signal processor design. sapphire fiber based extrinsic FabryPérot interferometric (EFPI) sensing elements are difficult to fabricate and their interfering signals are usually low quality because of the sapphire fiber’s multimode nature. The measurement stability is also a concern related to the twochannel structure in the SCIIB signal processor. This type of optical sensor system has been tried in CPT and proved to be not practical for temperature measurement up to 1600oC. also named as a broadband polarimetric differential interferometric (BPDI) sensor system. which only utilize bulk optics made of single crystal sapphire materials. The PS+whitelight light signal processor. four temperature measurement systems rooted in the optical interference metrology can be constructed as shown in Figure 2. Because the SCIIB signal processor can only operate properly in a limited linear working range of the sinusoidal interferometric signals. It is chosen as an optimal approach for . As discussed in the optical sensing unit section.Yibing Zhang Chapter 2. the thickness of the sapphire sensing element is one of the keys to control the initial working point and working range of the measurement system. The PS+SCIIB system is an optical intensity-measurement based polarimetric sapphire sensor. Principles of operation of the designed optical sensors 28 and spectral domain whilte light interferometric signal processor). which is a function of temperature. Because of the periodic nature of the interferometric signal. is another factor to be considered for the sensing element design. Two systems: PS+SCIIB and PS+white light signal processor are attractive candidates for further development and evaluation. can solve the problems associated with the PS+SCIIB system. The efforts toward this work is then focused on the development of a polarimetric structure based sensing systems. it is hard to maintain long-term accurate temperature measurements. on the other hand.

A reflection version polarimetric sensing structure setup. which not only reduces the cost.Yibing Zhang Chapter 2. In many industrial applications. The total sensing probe is then integrated into one solid long tube. 2. but also simplifies the probe mechanical structure. The only optical polarizer is used both as a polarizer for the input light and a polarization analyzer for the reflected light in the sensor probe. it is not practical to use a transmission version of the polarimetric sensing structure in the sensing probe.17. high resolution and accurate temperature measurement in a very large dynamic range is achievable in this system. Schematic design of the single crystal sapphire based BPDI optical high temperature sensor. Optical power 1 0 t Signal Generator λo Optical source Optical Spectrum Analyzer SMA connector to Fiber Fiber Collimation lens Polarizer (Analyzer) Extension Tube Sapphire Tube ∆λ λ Coupler Optical fiber Sensor Head Computer Rightangle Prism Sensing Element Figure. The BPDI sensing system monitors temperature in real-time through measuring optical spectrum instead of optical intensities. Principles of operation of the designed optical sensors 29 further prototype instrumentation development for high temperature measurement in harsh environments. thus providing a high degree of long-term temperature measurement stability. which makes it robust enough to survive the harsh environments. because of its two-end structure. Based on its interferomatic sensing scheme. where only one fiber collimator is used to collimate input light and also re-collect light reflected back from the other end of the sensing probe. which is an one-end structure is thus designed. .17. This assures its relative immunity to optical source power fluctuations and transmission fiber losses. as shown in Figure 2.

Principles of operation of the designed optical sensors 30 In the sensing system. 2. This method deals with mathematical formulas and is very effective for a system with large numbers of optical elements. The intensities of the dispersed spectral components of the signal are scanned electronically by the CCD array with a high spectrum measurement resolution. Each optical element can be represented by a 2*2 matrix. This method was invented in 1940 by R.C. In the Jones matrix method.Jones [55]. . then injected into a lead-in multimode optical fiber and propagates through a fiber bundle to the sensor head. Additional signal processing is performed in the computer to extract the temperature information. The light is then reflected by a right angle zirconia prism and passes through the sapphire-sensing element again. and the polarization state of the transmitted light is computed by multiplying the vector representing the input beam by the overall matrix. then converted into a linearly polarized optical beam and travels across a free space enclosed by a high temperature ceramic tube and a single crystal sapphire tube to a single crystal sapphire sensing element. which is a fiber optic PC plug-in spectrometer. The output light from the polarimeter is then re-collected by the same input optical fiber bundle and travels back along the optical fiber to the optical detection end.Yibing Zhang Chapter 2. the broadband light from a high power light emitting diode (LED) is modulated by the signal generator and converted into AC signals for the purpose of the blackbody radiation background subtraction. a more simple method is employed with emphasis on the physical activities in the sensing probe. and is a powerful tool for analyzing the properties of transmitted light through a complex optical system composed of many optical elements.5 Mathematical model for BPDI temperature sensor One way to describe the sensing principle of the optical sensor mathematically is using the Jones matrix method. In this temperature sensor system with a few optical elements. the state of polarization is represented by a two-component vector. The light is first collimated by a collimation lens. the two linear polarized light components with a differential phase delay are combined along the polarizer direction to interfere with each other. composed of a grating and a CCD array. When the light exits the polarimeter. The overall transfer matrix for the whole optical system is obtained by multiplying all of the individual element matrices.

resulting from the 45o angle alignment between P1 and the principle axes of the sensing element. where the light is decomposed according to their polarization direction again: the ordinary ray is decomposed into ray1 and ray2.7) .Yibing Zhang Chapter 2. E ray2 and Eray 2' are all vectors. Assuming no attenuation in the sensing element. only light polarized along the z-axis can pass. Interfering results between ray2 and ray2' will pass the polarization analyzer. since the polarization directions of the polarizer and the analyzer are parallel to each other. the extraordinary ray is decomposed into ray1' and ray2'. these two rays will propagate independently without changes in their amplitudes. where polarization directions of the polarizer and the analyzer are parallel to each other. At the end of the analyzer. z P1 and P2 Ordinary ray Ray 2 Ray 2' Extrodinary ray Ray 1 x Ray 1' Figure. Interfering mechanism for polarimetric differential interferometer The interfering results can be expressed mathematically as following: r r r where Eout . oriented along the principal axes of the sensing element: ordinary ray (along f-axis) and extraordinary ray (along s-axis). The output optical intensity is then: r r 1 r E out = ( E ray2 + E ray 2 ' ) 2 (2.18. When this linear polarized light propagates inside a sensing element with optical birefringence. Principles of operation of the designed optical sensors 31 The optical interfering activity inside the polarimeter is shown in Figure 2. 2. The oscillation direction of the electric field vector of the light wave is thus along P1 (z-axis). The input broadband light is polarized by a polarizer P1. while there will be relative phase delays between them.18. the sensing element behaves as if it split the incident polarization state into two orthogonal linearly polarized states with the same amplitudes.

6582 × 10− 5 + 0.10) where the product of d and ∆n is a function of temperature. f (T ) = d (T ) × ∆n(T ) where. According to reference [54]. To decode the temperature information.4995 × 10− 8 × (T − 273) + 0.8) It will not only depend on the magnitudes of the interfering electric fields. Since both d and ∆n are functions of ambient temperatures.11) where α // c− axis is the coefficient of thermal expansion along the C-axis of single crystal sapphire.9): π d ∆n ) In (λ) = cos2 ( λ (2.12) (2. Iout will be a temperature encoded signal. physically representing the optical path difference (OPD) between the two orthogonal linearly polarized light waves. Principles of operation of the designed optical sensors 32 (2. but also on the phase shift Φ between the two electrical oscillation vectors: 1 1 1 I out = ( Eray 2 ) 2 + ( Eray 2' ) 2 + Eray 2' Eray 2 cos Φ 2 2 2 1 = Eray 2 2 (2 + 2 cos(Φ )) 4 1 = Eray 2 2 (1 + cos(Φ )) 2 = Eray 22 cos 2 (Φ / 2)) where Φ = 2πd∆n / λ .Yibing Zhang r 2 I= Eout Chapter 2. d (T ) = α // c − axis (T ) × (T − T0 ) ∆n(T ) = no (T ) − ne (T ) (2. a signal processing model in the spectral domain for the white light interferometry is needed.9) The interference fringes of a broadband light from the polarimeter form a co-sinusoidal curve after it is normalized with respect to Eray 2 2 (λ ) according to Equation (2.2578 × 10−11 × (T − 273) 2 . (2. for the temperature range 325-950K: α // c − axis (T ) = 0. (T ' s unit in K ) .

For example. which is uniquely related to the differential phase delay between the two orthogonal linearly polarized light beams. the normalized interferometric spectrum consists of a series of maxima or minima.13) With these two equations. the interference order number m for the peak related to the wavelength λ1 can be determined. by detecting the spectral locations of the peaks or valleys in the interference spectrum.10). shown as peaks and valleys in the spectrum at certain wavelengths. Assuming the wavelengths of two consecutive peak points are λ1 and λ2 (λ1>λ2) in the interference spectrum.Yibing Zhang Chapter 2. we can obtain OPD values by applying Equation (2. These special points have a fixed phase relation. For λ1 and λ2: 2 π f (T ) = m 2π λ 1 2 π f (T ) = ( m + 1) 2 π λ 2 (2. they are changed along with temperature in the level of 10 −6 [ K −1 ] . Therefore. The algorithm used to calculate f (T ) from the transmitted interferometric spectrum is described below. as well as the OPD values related to the interference signals: m= λ2 λ1 − λ2 λλ f (T ) = d (t )∆n(T ) = 1 2 λ1 − λ2 (2. their interference orders will be different by 1 and represented by m and m+1. Principles of operation of the designed optical sensors 33 and the refractive indices (no and ne) also depend on the temperature. an internally developed algorithm [66] is employed to measure the OPD f (T ) values between the two orthogonal linearly polarized light beams in the sensing element. According to Equation (2. the phase difference between the two adjacent peaks (or valleys) is 2π. From the detected optical spectrum.10).14) .

immunity to EMI. which is usually associated with the periodic nature of the interference signals. . The BPDI sensor system uses a compact PC-plug in optical spectrum analyzer as an optical signal processor. and chemical inertness. Minimized complexity in signal processor. 2. remote operation. and relating them to the temperature by calibration. the BPDI sensor system incorporates a spectrometer with a polarimeter for the purpose of measuring polarization properties as a function of wavelength. These two factors completely characterize the sensing element behaviors at different temperature levels. such as lightweight.6 Major advantages of BPDI sensor system In addition to the general advantages associated with optical sensors. Principles of operation of the designed optical sensors 34 By employing the white light interferometry method to measure the OPD values in the sensing element. the stability of temperature measurements is thus increased. electrically non-conducting. this sensor system takes into account two factors at the same time: one is the dimensional changes of the sensing element and the other is the refractive indices changes with the temperature. 2.Yibing Zhang Chapter 2. Thus this design provides a method for reliable absolute measurement of temperature. The one-to-one relation between the phase delays and the temperatures makes it possible to measure temperature in a wide dynamic range without ambiguity. BPDI technology extracts absolute temperature information by absolute measurement of the phase delays. offering the following major advantages over the other sensors designed for high temperature applications: 1. The complexity of the signal processor is then kept to its minimum level. Absolute measurement in a wide dynamic temperature range. and a digital computer for signal decoding to extract the temperature information. which are more attractive in applications for the harsh environments because of its no requirement of initialization and/or calibration when power is switched on.

The BPDI optical temperature measurement system takes advantages of both optical fibers and bulk optics to simplify the design of the sensing head. lightweight and immunity to electromagnetic interference (EMI). Self-compensation capability. . Bulk optics is convenient for reducing the required tolerances on optical alignment and also for reducing the sensitivity to mechanical vibrations. Principles of operation of the designed optical sensors 35 3. The BPDI sensor system offers high measurement resolution since the essence of the sensing scheme is based on interferometry. The optical fiber can transmit light over a long distance with small attenuation and are easily implemented in industrial environments because of their small size. The BPDI sensor system monitors temperatures in real-time through measuring optical spectrum instead of optical intensity. Ultra-high sensitivity.Yibing Zhang Chapter 2. 4. Deployment flexibility. 5. This guarantees its relative immunity to optical source power fluctuations and transmission fiber losses. thus providing a high degree of long-term measurement stability.

the PS+white light interferometric system. As presented in Figure 2. PS+SCIIB. strong electromagnetic interference.1 Fabrication materials for sensing probes Commercially available conventionally temperature sensors exhibit greatly abbreviated lifetimes in harsh environments. In the primary and secondary stages of a two stage slagging 36 . and signal processor implementation. Hardware implementation includes fabrication materials selection for the sensing probe. Based on the comparative evaluation and analysis of the experimental results on these four different temperature sensing schemes. Section 3. four types of temperature measurement systems rooted in the optical interference metrology can be constructed: EFPI+SCIIB. and high energy radiation exposure exist simultaneously. and digital signal processing for the blackbody radiation background subtraction in the computer. sensing probe mechanical assembly. The implementation of the BPDI sensor system includes both hardware and software design and implementation.3 concentrates on the signal processor implementation.Chapter 3. Section 3.1 introduces the properties of the materials for the sensing probe fabrication. Section 3. chemical corrosion. 3. algorithm and digital signal processing software development issues. Design and implementation of temperature sensor systems This chapter will describe the design and implementation of the temperature sensor systems. EFPI+white light interferometric system.16. sensing element structure design and fabrication. Finally Section 3.4 provides an overview of the implemented BPDI sensor system. is chosen for further prototype instrumentation development. where high temperature. Software design and implementation includes algorithm development and implementation for temperature measurements based on the white light interefrometry. In this chapter.2 presents the sensing probe prototypes based on the different sensing element structures. also named as broadband polarimetric differential interferometric (BPDI) system. high pressure. and PS+white light interferometric system.

sulfur. Design and implementation of temperature sensor systems 37 gasifier. pressure exceeding 400 psig. Optical grade single crystal sapphire with very low scattering or distortion is commercially available and widely employed for infrared optical windows and domes that must survive in the most demanding environments. Some special alumina ceramics can survive both high temperature and oxidization environments. no polymer materials can survive temperatures up to 1600 oC.3 microns in the UV band. have higher melting points over 1600 oC.1. such as platinum and tungsten. through the visible range of 0. such as polycrystalline alumina. With temperatures exceeding 1400°C.Yibing Zhang Chapter 3.1 Properties of single crystal sapphire Unlike other ceramic materials.7microns to . single crystal sapphire (Al2O3) is expected to show superior corrosion resistance because of the elimination of polycrystalline grain boundaries. Very limited materials are available to fabricate an optical sensor to be physically present in such harsh environments for direct temperature measurement.4-0. • Optical properties of single crystal sapphire Light transmission in single crystal sapphire material is better than 80% from a wavelength of 0. it is hard to find a temperature sensor that is impervious to such an extensive corrosive attack. and chemical corrosive agents such as alkalis. such as the head of the laserguide missiles in battle fields. and optical properties of the materials. 3. Single crystal sapphire is actually single crystal aluminum oxide (α-A1203). only a few expensive metals. single crystal sapphire and fully stabilized single crystal zirconia are very attractive optical materials for high temperature measurement in harsh environments. and nitridation agents. Bearing in mind the requirements on the thermal properties. which have been proven to be vulnerable to corrosion attack [38]. carbonization. sulfidation. but not all of them are optically transparent. the temperature measurement devices are subjected to such extremely harsh conditions. chemical stabilities. transition metals and steam. For example. but they can hardly resist the chemical oxidization.

0282512 1. Design and implementation of temperature sensor systems 38 5.1.Yibing Zhang Chapter 3.07266312 λ − 0.11932422 λ − 18.5 microns in the IR band.07402882 λ − 0. Figure 3. processing hexagonal crystal structure as shown in Figure 3. Hexagonal inner structure of single crystal sapphire material. They can be obtained from the following approximate formula at room-temperature [57]. Single crystal sapphire.0722482 no − 1 = 2 (3.039 in) provides a nearly uniform transmission of 85% [56]. The refractive index along this axis is no.3414021λ2 + 2 + 2 λ2 − 0. For single crystal sapphire ne<no. the values of ne and no are temperature dependent.65054713λ2 5. The unique crystallographic axis is the C-axis. the transmission band for a sapphire window with a thickness of 1mm (. The approximate transmission bands of standard. Figure. the unit of the wavelength λ is micron (µm): 1. The perfection and purity of the UV grade sapphire crystals offer a transmission capability superior to that of the standard grade sapphire.5039759λ2 0.1) .2. also called optical axis. In Figure 3. Sapphire’s lower surface scattering losses at the IR wavelengths also require less rigorous surface polishing.1.2.and UV-grade sapphire (window thickness of 0. and has two principal optical axes. 3. The refractive index along A-axis is ne.4313493λ2 0.5927379λ2 2 ne − 1 = 2 + 2 + 2 λ − 0.55069141λ2 6.12165292 λ − 20. along which the refractive indices are different.039 inch).

which has been observed in our experiments. This suggests that the magnitude of the birefringence of the single crystal sapphire will be changing with temperature. the resistance to thermal stress is limited by a loss of mechanical strength at elevated temperature [58-62]. • Thermal stability of single crystal sapphire Single crystal sapphire can operate at high temperatures and survive high stress induced by rapid heating. including heat treatments. the atomic arrangements inside the sapphire material are becoming more homogenous. Heat treatments at 1450oC for 48 hours in flowing oxygen increase the compressive strength by 60% and biaxial flexure strength by 45% at 600oC. At temperature above 400 oC a dramatic decrease in the sapphire’s compressive strength along the C-axis occurs due to the activation of the deformation twinning on the rhombohedral planes. and doping and improvement of fabrication techniques. thus the difference between the refractive indices along the A-axis and the C-axis are becoming smaller. With the temperature increase. thus the spacing between the atoms affects the speed of light propagation directly. However. Design and implementation of temperature sensor systems 39 Thermo-optic properties of single crystal sapphire Since the propagation speed of light in a particular direction in a crystal is a function of the electron density distribution. The thermal expansion coefficients of sapphire are different along the A-axis and C-axis.e. The change in spacing between the adjacent atoms in each of these directions will be different for a given temperature change. i.Yibing Zhang • Chapter 3. the birefringence is becoming smaller. Various ways to increase the strength of sapphire have been explored. Temperature is one of the main factors affecting the refractive index of solid materials. Doping with Mg achieves 160% strength increase over the undoped and unheat-treated baseline strength [63-64]. and the electron probability density distribution is related to the spacing between adjacently bonded atoms. .

Because of its cubic inner structure. which assigns a unit of 10 to diamond. Its polymorphism. For example. Without adding additional birefringence to the sapphire sensing element. sapphire also offers maximum resistance to abrasion and scoring. the refractive index of the single crystal zirconia is constant. it is thus used as a right angle prism reflector behind the sapphire . and glass at 4.5–6. into the zirconia structure results in a solid solution. such as CaO. Design and implementation of temperature sensor systems 40 Mechanical properties of single crystal sapphire On the Mohs scale of hardness. but sapphire exhibits no solubility in alkalies or acids. The change in volume associated with this transformation makes the usage of pure zirconia in many applications impossible. which is a valuable refractory [56]. other acids (e.5 [56]. • Chemical inertness of single crystal sapphire Sapphire’s chemical inertness in the presence of a wide variety of reagents at temperatures greater than 1000ºC makes it ideal for chemical industrial applications.Yibing Zhang • Chapter 3. Addition of the appropriate amount of some oxides. Sapphire is an ideal material for the windows that must withstand great pressure or vacuum.g. At elevated temperatures.. During a heating process. zirconia will undergo a phase transformation process.1. so there is no birefringence associated with the single crystal zirconia. including hydrofluoric acid. regardless of the crystal orientation. quartz at 7. hydrochloric acid and nitric acid) attack silica. silica becomes soluble in hydrofluoric acid at room temperature. restricts its widespread use in ceramic industry. or Y2O3. however. In contrast to other available optically transparent materials. MgO.2 Properties of single crystal zirconia Pure Zirconia (Zirconium dioxide) has a high melting point (2700oC) and a low thermal conductivity as well as chemical corrosive resistance. but not sapphire [56]. sapphire is rated at 9. This solid solution material is termed as stabilized zirconia. 3. which is a cubic form and has no phase transformation during heating and cooling.

Following the working principle of the sensor described in Chapter 2.1 A sensing probe with a sapphire disk.. a mirror is commonly used .Yibing Zhang Chapter 3. single crystal sapphire material is chosen to be used as both sensing elements and protection materials for the sensing probe packaging in the designed temperature sensing system. Design and implementation of temperature sensor systems 41 sensing element at the end of the sensor probe to reflect light back. input light and output light are all at the same end of the sensing probe instead of two separate ends. two one-end structured sensing probe prototypes were developed with different sensing element structures. Certain protection for the sensing element is necessary in such aggressive environments. 3.2 Implementation of sensing probes This sensor system is designed to measure high temperatures directly in harsh environments. Another advantage associated with this reflector is that no refraction or reflection (i.e. In low temperature environments. no scatter) occurs. In most industrial applications. In this reflective polarimeter. 3. zirconia prism (SDZP) structure. In this research. a light reflector with high temperature survivability is necessary. Due to its high melting temperature (over 2000°C). superior transparency. a sensing probe with a one-end structure is more convenient to be deployed. The sensing element should thus be physically located in the place where temperature needs to be measured. since all of the necessary optical components can be integrated into one solid tube. and ability to resist corrosion. the sensing head structure can be implemented as either a two-end structure (a transmission version of the polarimeter) or a one-end structure (a reflection version of the polarimeter). and the other is the sapphire prism (SP) structure. forming an one-end probe structure. when light passes from grain to grain of the cubic material. One is the sapphire disk. zirconia prism (SDZP) structure This structure is a very straightforward implementation of the polarimetric sensor in a reflection mode.2. so that the sensor can monitor the temperature with reliability and durability.

Yibing Zhang

Chapter 3. Design and implementation of temperature sensor systems 42

as a light reflector. The mirror is usually made by coating one surface of a glass plate with different materials, typically dielectrics or metals. Such dielectric or metal coatings can hardly survive the high temperature and chemically corrosive environments, either because the coating materials degrade at high temperature in the chemically corrosive environments, or because the coefficient of thermal expansion (CTE) of the materials do not match well with those of the glass substrate, causing thermal stress on heating or cooling, so that they cannot stick to the substrate and function as a mirror in the high temperature/corrosive environments. Instead, a right-angle prism is an attractive design since the total internal reflection is employed inside to change the light propagation direction by 180o. In the prism, only one type of material is employed, thus the CTE matching problem is avoided for the high temperature applications.

The SDZP structure of sensing probe is shown in Figure 3.3. In this structure, a sapphire sensing disk and a zirconia right angle prism are used, and packaged into a solid sapphire tube for protection. Both the thickness and the birefringence of the sapphire disk change along with temperature changes, thus the temperature information can be obtained by measuring the OPD between the two orthogonal linearly polarized light waves in the sapphire sensing disk. The OPD is related to both the thickness d, and the birefringence
∆n of the sensing disk by:

OPD = ne − no d = ∆nd

(3.2)

The polarizer used in this system is an optical grade calcite Glan-Thompson polarizing prism. The light is first collimated by a collimation lens then converted into a linearly polarized optical beam that travels across the free space enclosed by a high temperature ceramic tube and a single crystal sapphire tube to the single crystal sapphire disk (with thickness of 1.5mm and diameter of 30mm). In the sensing probe, the sapphire disk is arranged such that the linear polarization direction of the input light is at 45 degrees with respect to its fast and slow axes. When passing through the sapphire disk, the two linearly

Yibing Zhang

Chapter 3. Design and implementation of temperature sensor systems 43

Inner sapphire supporting tube

Sapphire sening disk

Zirconia prism reflector

Outer sapphire protection tube

Alumina Extension tube packaged with polarizer, optical fiber collimator

SDZP struture sensor probe

Figure 3.3 A sensing probe with a Sapphire Disk, Zirconia Prism (SDZP) structure polarized light components along the fast and slow axes experience a differential phase delay due to the sapphire material birefringence and its thickness. The light, containing the two orthogonal linearly polarized light components, is then reflected by the right angle zirconia prism and passes through the sapphire sensing element again, doubling the differential phase retardation. Since the right-angle zirconia prism has no inner birefringence because of its inner cubic crystallographic structure, no additional differential phase delay will be added in the detected optical signals. When light exits the polarimeter, the two linearly polarized light components with a differential phase delay are combined along the polarization direction of the analyzer and interfere with each

Yibing Zhang

Chapter 3. Design and implementation of temperature sensor systems 44

other. The output light from the polarimeter is then re-collected by the same input optical fiber bundle and travels back along the optical fiber bundle to the optical detection end which is a fiber optic PC plug-in spectrometer, composed of a grating and a CCD array. Additional signal processing is performed in the computer for temperature information extraction. With the protection of the single crystal sapphire outer tube, the end of the sensor head is heated in a high temperature furnace up to 1600oC and temperature is measured. The measured results are indicative of high accuracy and high resolution.

An extension tube for the sensor head is necessary to avoid thermal damage to the optical fiber collimator and the optical polarizer. These components cannot survive high temperatures over 200ºC. The total sensing tube will thus be about two meters long together with the extension tube. 3.2.2 A sensing probe with a sapphire prism (SP) structure The previously developed sensing probe prototype utilized single crystal sapphire disk as a sensing element and a single crystal zirconia prism as the light reflector in the sensing location. These materials do not have the same CTEs. This may generate instability for the mechanical packaging in high temperature environments. This structure also suffers additional optical reflection losses at the surfaces of the zirconia prism besides the sapphire disk surface reflection losses. From a cost minimization standpoint, an additional single crystal zirconia light reflector will cost more. In the new version of the sensing probe, a novel single crystal sapphire right angle prism, functioning as both a sensing element and a light reflector, is employed. This single crystal sapphire prism possesses a special geometrical structure and crystallographic orientation for those dual functions.

The inherent birefringence property of the single crystal sapphire is needed for the prism to function as a sensing element. As shown in Figure 3.4 (a), when light passes through the right angle prism, it will propagate in three different directions. Beam 1 propagates along JC in the +X direction, along CD in the –Y direction, and along DK in the –X

The base of the sapphire prism is A-plane. when light propagates in the y-direction. With this crystallographic orientation. such as CD and EF. along EF in the –Y direction. Different incident light beams encounter different path lengths in different directions. Thus the OPDs for all the input light waves in the prism are: OPDbeam1 = ( JC + CD + DK )∆n = AB ∆n OPDbeam 2 = ( LE + EF + FM )∆n = AB ∆n (3. while both the smaller sides of the prism are C-plane. it experiences the birefringence generated by both the A-axis and C-axis. According to the geometric relationships.3) This prism thus functions equivalent to a sensing disk with thickness of the length AB. as shown in Figure 3. The crystallographic orientation of the prism is selected as follows. it also experiences the birefringence generated by both the A-axis and C-axis. as well as a light reflector. Beam 2 propagates along LE in the +X direction. Figure 3. Design and implementation of temperature sensor systems 45 direction. and along FM in the –X direction. A 1 Y X O M K B (a) Optical propagation path in the prism C 2 J L E F D .5 shows the designed geometrical dimensions of the prism and three pieces of the novel single crystal sapphire prisms. such as JC and LE.4(b).Yibing Zhang Chapter 3. all light beams will experience the same length equal to AB when they propagate in the prism. Therefore the total birefringence will be equal regardless of the paths in which optical beams propagate. Because of the symmetric inner atomic structure of the single crystal sapphire. when light propagates in the x-direction.

Yibing Zhang Chapter 3. a-axis 5.0mm a-axis c-axis 220mm Bottom view o 45 Side view (a) Geometrical sizes of the prism . Design and implementation of temperature sensor systems 46 A (Y) A-plane form ed by A-C axes A (X) C-plane form ed by A-A axes C (Z) (b) Crystallographic orientation in the prism J C D K A L E F B Beam experiencing birefringence formed by A(Y) and C(Z) Beam experiencing birefringence formed by A(X) and C(Z) (c ) Optical path in the sapphire prism is equivalent to the optical path in the sapphire disk M Figure 3.4. Optical propagation in special crystallographic oriented single crystal sapphire right angle prism.0mm a-axis 45 o 5.

optical fiber collimator Prism struture sensor probe Figure 3. Design and implementation of temperature sensor systems 47 (b) Fabricated single crystal sapphire right angle prisms Figure 3.6 Sensing probe with a Sapphire Prism (SP) structure .5 Single crystal sapphire right angle prism Inner sapphire supporting tube Sapphire sening Prism Outer sapphire protection tube Alumina Extension tube packaged with polarizer.Yibing Zhang Chapter 3.

3. light is totally reflected without loss in those right angle prisms.3 Total internal reflection in the sensor probe The right angle prism works as a light reflector in the sensing probe. with a certain amount of free space between its end and the sensing prism so that the prism can expand freely when the temperature increases. In the prism reflector. The sensing prism is fixed at the end of the inner supporting tube. the Jones matrix method can be used to analyze the electric field polarization direction changes at the reflection surfaces [66]. The light propagation direction is then changed by 180o at the end of the sensing probe by two total internal reflections in the right angle prisms. Design and implementation of temperature sensor systems 48 With a single right angle prism at the end of the sensing probe. The electric field vector of the input light is related to the electric field vector of the output light by: .7 o (3.4) θ c − sapphire = sin −1 (nair / nsapphire ) = sin −1 (1 / 1.Yibing Zhang Chapter 3. The total internal reflection angle.12) = 28.76) = 34.1o Total internal reflection angle for the single crystal sapphire-air interface is: (3. in which a slot of the correct size to fit the prism is machined. The total sensing tube is about two meters long together with the extension tube. the mechanical structure of the sensing head is simplified. which is used to avoid thermal damage to the optical fiber collimator and the optical polarizer at the other end of the sensing probe.2. which is larger than the critical angle. as shown in Figure 3. We want to analyze the optical loss at the reflection surface and the changes of the optical wave polarization states when the light is reflected. The outer protective sapphire tube covers the inner tube and the sensing prism.6. also called critical angle θc for the single crystal zirconia –air interface is: θ c − zirconia = sin −1 (nair / n zirconia ) = sin −1 (1 / 2. light is reflected by the total internal reflection.5) With a 45o reflection angle at the prism-air interface.

n0=1 is the refractive index of air. such as measurement resolution and stability. the strict requirements to environmental conditions exclude the possibility to use this kind of instrument in real industrial fields. and ψ is the incident angle of light in the prism. The performance of the whole system. Design and implementation of temperature sensor systems 49 r r E out = ΤEin the Jones matrix T is given by  ρx T =  0  where. 3.3 Signal processor and software implementation 3. which is 45o. a compact fiber . In the designed system. which means the total internal reflections in the right angle prism do not change the polarization directions of the reflected light.8) n is the refractive index of the single crystal sapphire or the single crystal zirconia. ψ c = sin −1 (n0 / n) ρx = ρy = 1 (3. 1 + iα 1 − iα n 2 + in 2 α ρy = 0 2 n0 − in 2 α ρx = α= 2 (n 2 sin 2 ψ − n0 )1/ 2 n cos ψ (3. for light reflected at the critical angle.Yibing Zhang Chapter 3. In a traditional spectrometer based system where a cumbersome and expensive spectrometer is used. thus no additional phase shift will be induced between the two orthogonal linearly polarized light waves.9) then the Jones matrix reduces to an identity matrix.3.1 Signal processor implementation An important part of the spectral domain white light interferometry signal processor is the spectrometer. are highly dependent on the performance of the spectrometer.6) 0  ρy   (3.7) (3.

1. The measured spectrum is then transferred into a computer for further signal processing. <0.10% at 250 nm Fiber optic connector SMA 905 to single-strand optical fiber (0.) is used.3 nm-10.0 nm FWHM < 0.1.60 lb.5-4. < 0.5" x 2.10% at 435 nm. Design and implementation of temperature sensor systems 50 optic spectrometer (USB2000 manufactured by Ocean Optics. Inc.0 pA (RMS) optical fiber is entrance aperture 42 mm (input). The Key Features of the optical spectrum analyzer USB2000: Computer interface Universal Serial Bus (RS-232 available on side connector) Integration time Data transfer rate 3 milliseconds-65 seconds Full scans (2048 wavelengths) into memory every 13 milliseconds Dimensions: (Length*Wideth*Hight) 3.05% at 600 nm. 68 mm (output) ~ 0. Figure 3.Yibing Zhang Chapter 3. Table 3.45 lb.22 NA) The input fiber is connected to the spectrometer directly through a SMA connector. with cable Detector Effective range Sensitivity (estimate) Signal-to-noise Dark current noise Slits Focal length Resolution Stray light 2048-element linear silicon CCD array 800-900 nm 86 photons/count.7 shows the structure of a single channel spectral . The output electrical signals from each pixel in the CCD array will be read out serially and converted into digital format by a built-in analog/digital (A/D) card. 250:1 (at full signal) 2.31" 89 mm x 64 mm x 34 mm Weight 0. a 1200 line holographic grating is used to diffract the input light on a 2048 pixels CCD array. without cable 0. Inside the spectrometer.5"x 1. Its key features are listed in Table 3.

an optical spectrometer.7. Single channel spectral domain white light signal processor . Power source underneath the box USB interface to computer USB 2000 spectrometer LED driver LED APC fiber connector Fiber coupler Figure 3. a LED light source and an electrical power supply. Design and implementation of temperature sensor systems 51 domain white light signal processor. including an optical fiber coupler.Yibing Zhang Chapter 3.

For the selected peak.10) can be rewritten into the following form: m2πλm = mλm 2π L= (3. the interference order m must be known first. Design and implementation of temperature sensor systems 52 3. In this method. A novel algorithm is then adopted for the BPDI sensor system to achieve both high resolution and large dynamic range simultaneously. the wavelength λm of a peak point in the interference spectrum satisfies the relationship: 2πL 2πd∆n = = m2π λm λm (3. The resolution of this peak tracing method is mainly dependent .5).3. or large dynamic measurement range with limited resolution (such as the two-point method). Equation (3.11) To calculate the OPD value L from a special peak wavelength λm.2 Signal processing algorithm development According to the working principle of the white light interferometry shown in Equation (2. OPD between the two orthogonal linearly polarized light beams can be obtained from the measured spectrum fringe patterns. Peak tracing method Based on the measured spectrum fringes patterns. the identification of the interference order m is difficult so that the unambiguous operating range of the OPD is limited in only half of the wavelength range. the value of the OPD can be obtained through tracing a special point in the interference fringes (such as one peak point). The existing algorithms for the OPD calculations either possess high resolution with limited measurement range (such as the peak tracing method).Yibing Zhang Chapter 3.10) where the spectral order m is a non-negative integer.

Their interference orders are m and m+1. L= (3. this factor λ2 is about 110—150. two special points instead of one point in the interference spectrum need to be used for the absolute measurement.14) The wide dynamic measurement range can then be obtained by using two special wavelengths. Two-point method To increase the measurement range of the OPD values. the OPD can then be calculated by Equation (3. The relative uncertainly of the measurement can be described as ∆L ∆λ ≅ L λ where ∆λ is the measurement resolution of the wavelength. determined by the measurement resolution of the spectrometer. as long as the two such special wavelengths exist (two peaks or two valleys) in the spectrum pattern.Yibing Zhang Chapter 3. In this case. Suppose λ1 and λ2 (λ1>λ2) are the wavelengths of the two adjacent peak points in the interference spectrum. 2πL = m2π λ1 2πL = (m + 1)2π λ2 The OPD can be determined.12) λ1 ⋅ λ2 λ2 − λ1 (3. for the operating range of OPD (40-60 µm).10).15) If the central wavelength of the light source (LED) is 850nm. .11). this method has lower λ2 − λ1 resolution in a large dynamic measurement range compared to the peak tracing method.13) (3. Design and implementation of temperature sensor systems 53 on the resolution of the spectrometer as shown in Equation (3. From Equation (2. the relative uncertainty induced by the spectrometer is ∆L λ2 ∆λ ≅ • 1 L λ2 − λ1 λ1 (3. Thus.14).

then for the adjacent peaks the interference order will be 13. the OPD is set to a known value L0. the measuring sub-process can be carried out as follows: 1. the value of m for a special peak (mo) can be acquired accurately and stored in computer. This is the calibrating sub-process. The whole process of demodulating OPD from the interference spectrum can then be separated into two sub-processes: the calibrating sub-process and the measuring sub-process.11). the difference between the interference orders is 1. this rough OPD is then used to determine the rough order number mr of a special peak point in the interference fringes. m is a constant value.Yibing Zhang Chapter 3. . which can achieve both high resolution and large dynamic range. The algorithm used to recover the accurate ma is the following: In Equation (3. The basic idea of this method is to use two peak points in the interference spectrum to get a rough OPD value first (large dynamic range is then achieved). an accurate ma value will be recovered from the rough mr. Design and implementation of temperature sensor systems 54 New algorithm Combining the advantages of the two methods described above. 14.By calibration. the accurate OPD can be calculated based on the accurate ma and the peak wavelength location in the spectrum (high resolution is thus achieved). For adjacent peaks. With known mo. 9…. if m for one peak is 12. for a given peak. With the white light interferometer system in a stable condition. From Equation (3. 15…and 11.11). For example.14). 10. Use the two peak points in the interference spectrum to acquire a rough value of Lr using Equation (3. The mo of a peak in the interference spectrum can be calculated accurately from the L0 and the wavelength λ0. a novel data processing algorithm has been developed in CPT [66]. When the rough mr value for any peak has been acquired. Then. the accurate ma will be obtained from mo by adding the integer part in the difference between the rough mr and mo.

based on the novel data processing algorithm presented previously. 4.5) Where function int(…) means to take the integral part. Design and implementation of temperature sensor systems 55 2. These data include the parameters for setting up the spectrometer such as the integration time. Then most of the important parameters. The accurate OPD can be calculated from the accurate ma and the wavelength of the peak with Equation (3. The block diagram of the program is shown in Figure 3.8. scanning time interval. then use the wavelength of this peak and the rough Lr from previous step to calculate a rough mr '. the calibrated interference order mo value.Yibing Zhang Chapter 3. will be restored in the memory of the computer.3. the reference spectrum of the light source. The software program is implemented in the combination of Microsoft Visual Basic and C++ languages so that both the graphic interfaces and high computational speed are achieved and optimized. . which are stored in the computer. (3. and the calibration curve of the temperature sensor. The accurate ma can be calculated from mr and the stored mo from: ma = mo + int( mr − mo + 0. 3.16) 3. Select a peak point near the center of the interference spectrum.11).3 Software design and implementation Advanced computer software has been developed to demodulate the OPD values from the interference spectral patterns. which includes the initialization of the spectrometer. The program is described in detail below: Initialization The program starts with the hardware initialization. the dark current noise of the spectrometer.

Yibing Zhang Chapter 3. Design and implementation of temperature sensor systems 56 Initialize white light system Restore important parameters Acquire interference spectrum Data processing Find coarse peak wavelengths Use mass-center method to get the accurate peak positions Use two peaks to calculate the rough OPD value Use the rough OPD value to calculate the rough m r value Use the m o stored in PC and the rough m r to calculate accurate m a Use the accurate m a to calculate the accurate OPD Use the accurate OPD and calibrating curve to demodulate temperature information Figure 3. Block diagram for the implementation of the OPD calculation algorithm .8.

a plane grating is employed to diffract the collimated light. a dark spectrum stored in the computer is . Data processing To demodulate the temperature information from the interference spectrum accurately. they discharge a capacitor at a rate proportional to the photon flux. In the last step. To remove the dark current noise of the CCD detector and other background noise. At the same time as light energy is being integrated. the calibrating curve of the CCD array is used to calculate the optical wavelengths according to the serial number of array. The CCD array is actually serially connected reverse-biased photodiodes. a series of switches closes and transfers the charge of each pixel to a shift register. After the transfer to the shift register is complete. and will be discussed in several sub-steps. the switches open and the capacitors attached to the photodiodes are recharged and a new integration period begins. the spectrum data is read out of the shift register by a built-in A/D card. This part is the core of the whole program. peaks and valleys localization and signal demodulation for the temperature measurement. Data pre-processing The original spectrum data from the spectrometer is a 1x2048 array. Once the optical signal is coupled into the spectrometer through a SMA connector from the optical fiber.Yibing Zhang Chapter 3. Design and implementation of temperature sensor systems 57 Spectrum acquisition The interference spectrum from the sensor head will be sampled according to the given time interval. A second spherical mirror then focuses the diffracted light and an image of the spectrum is projected onto a 1-dimensional 2048 pixels linear CCD array. To recover the interference spectrum. an advanced data processing algorithm has been developed and implemented in the Microsoft Visual Basic environments. the spectral data will be acquired by a computer through a USB interface. When the integration period of the detector is complete. as data pre-processing for blackbody radiation background subtraction. which composes the received light signals from the 2048 pixels of the CCD array. Then. the divergent light emerged from the optical fiber will be collimated by a spherical mirror.

9 (c). modulation of the optical signal is needed to convert the temperature probing light into AC signals. The modulated output light from the LED is shown in Figure 3. so that the blackbody radiation background can be subtracted as a DC signal. so that the probing light will be on and off following the square wave. 4000 3500 3000 Intensity 2500 2000 1500 1000 800 810 820 830 840 850 860 870 880 890 900 Wavelength(nm) (a) Measured spectrum in the presence of blackbody radiation at 1200 oC . The straightforward way to implement this modulation and demodulation scheme is to modulate the optical source with a square wave with certain frequency. so the signal noise ratio is poor. Design and implementation of temperature sensor systems 58 subtracted from the measured spectrum.9 (a).9(b). the blackbody radiation can be subtracted. When the probing light is on.Yibing Zhang Chapter 3. The square wave used to modulate the LED is shown in Figure 3. The light intensity curve is measured with a low quality optical detector. When the probing light is off. Since the blackbody radiation background is a slowly changing signal that follows the temperature changes. Thus. resulting in a spectrum as shown in Figure 3. which does not represent the real optical signal intensity levels.11. the detected signals will contain both the blackbody radiation and optical signals with temperature information. only blackbody radiation background is detected by the spectrometer as shown in Figure 3. The measured optical spectrum is as shown in Figure 3. it can be treated as constant during the time interval when the probing light is on and off.10. The frequency of the square wave is 8Hz. The dark spectrum is pre-acquired when the light source is not powered. As discussed in Chapter 2.

9 Illustrating a blackbody radiation subtraction Figure 3. Design and implementation of temperature sensor systems 59 4000 3500 3000 Intensity 2500 2000 1500 1000 800 810 820 830 840 850 860 870 880 890 900 Wavelength(nm) (b) Measured spectrum of blackbody radiation at 1200 oC 2500 2000 1500 Intensity 1000 500 0 -500 800 810 820 830 840 850 860 870 880 890 900 Wavelength(nm) (c) Spectrum after blackbody radiation subtraction at 1200 oC Figure 3.Yibing Zhang Chapter 3.10 . Electrical square wave used to modulate the LED .

12. Modulated output optical signal from the LED Peaks and valleys localization Desired peak positions in the interference spectrum are selected to demodulate the OPDs between the two orthogonal linearly polarized light waves.Yibing Zhang Chapter 3. . The valley positions in the interference spectrum can also be located with a similar method. the window size must be smaller than the space between adjacent peaks. The technique for locating peaks is described as follows: The first step is to find the coarse locations of the peaks by a smart comparison algorithm. a mass-centroid algorithm is applied to find the accurate positions of the peaks. then the peak position coincides with the x coordinate of the centriod of Y (x). Design and implementation of temperature sensor systems 60 Figure 3. which is defined as the window size of the peak searching.11. The size of the searching range. To avoid missing some peaks. must be large enough to eliminate the influence of the noise. The basic idea of this algorithm is to find all the local maximum points in a special range in the interference spectrum. If the function Y (x) is symmetric around its peak position. The basic idea of the mass-centroid algorithm is shown in Figure 3. After the coarse positions of the peaks are found.

the spectrum is a symmetrical function of the wave number k=1/λ. ki}).18) p λm = 1 p km Demodulation of the temperature information Once the peak positions are acquired. the temperature can be demodulated as described below: A coarse OPD is calculated using two peak positions. and then the rough mr value for a selected peak near the center wavelength is determined. the accurate peak positions can be calculated from the coarse peak positions: ki = 1 λi i =+∞ i i i i =−∞ i −1 k p m ∑ y k (k − k ) = ∑ y (k − k ) i i i −1 (3. the peak position Xo can be calculated from: X0 ∫ = ∫ +∞ −∞ +∞ −∞ xf ( x) dx f ( x)dx (3. λi} to {yi.12 Mass-centroid method for peaks locating in the spectrum With the normalized spectrum {yi. Design and implementation of temperature sensor systems 61 As shown in Figure 3. Y=f(x) Centroid X X0 Figure 3.Yibing Zhang Chapter 3.17) From Equation (2. So the spectrum y(λ) needs to be expressed in terms of the wave number first (from {yi. Using the calibrated mo value .12.5). ki}. around the peak point.

3. where the sensing element is located at the end of the sensing probe. Finally. The modulated light pulses (represented by yellow blocks) are injected into a 2-meter long lead-in multimode optical fiber bundle and propagates to the sensor head. the accurate ma value is obtained and then is used to calculate the accurate OPD values. BPDI sensor system overview . The polarizer used in this system is an optical grade calcite Glan-Thompson polarizing prism. then converted into a linearly polarized optical beam and travels across a free space enclosed by a high temperature ceramic tube and a single crystal sapphire tube to a sensing element. O p tica l fib er co u p ler O p tica l sp ectru m a n a ly zer O p tica l p o la rizer Figure 3. has a center wavelength at 850nm and its spectrum covers a range from 800nm to 900nm. 480µW output after coupling into a 200µm diameter pigtail fiber). the broadband light from a high power light emitting diode (LED. the temperature is extracted from the accurate OPDs. The light is first collimated by a collimation lens. using the calibration curve.13.13.4 Overview of BPDI temperature sensor system Based on the BPDI technology. In Figure 3. a portable temperature-sensing probe is formed.Yibing Zhang Chapter 3. Design and implementation of temperature sensor systems 62 stored in the computer. This structure of the sensing head is simplified by eliminating one polarizer and one light collimator compared with the two-end structure.

The sensor head is heated in a high temperature furnace up to 1600oC for its performance testing. Design and implementation of temperature sensor systems 63 The sapphire-sensing element is arranged such that the input linear polarization direction of light is at 45 degrees with respect to its fast and slow axes. the other six fibers re-collect light reflected back from the sensing probe. The output light from the sensing probe is then re-collected by the same input optical fiber bundle and travels back along the optical fiber to the optical detection end. and a type B hightemperature thermocouple (resolution 0. Additional signal processing is performed in the computer. thus the recollected optical power is sufficient to be detected by the CCD array in the spectrometer. which is sufficient to isolate thermal damage of the optical elements (optical fiber. The diameter of each fiber is 200µm. The total length of the sensing probe is 2 meters long and fabricated with high temperature ceramic material with a low thermal conductivity. a fiber bundle with one fiber at the center surrounded by six fibers is used. containing the two orthogonal linearly polarized light components with relative phase delays.Yibing Zhang Chapter 3. a type K thermocouple (resolution 0.2 oC under 500 oC and 0. it is usually difficult to re-collect optical energy into the optical fiber. The light. .1oC above 500 oC) is used for temperatures above 200 oC. collimation lens and polarizer) from sensing location (temperature over 1600oC). When light passing through the sapphire sensing element. which is a fiber optic PC plug-in spectrometer. through a diameter of 22mm collimation lens with a 35mm focal length. the reflector can be a single crystal zirconia right angle prism or a sapphire sensing right angle prism.05 oC) is used for temperatures lower than 200 oC. After passing through 4 meters long free space (round trip in the sensing probe). The intensities of the dispersed spectral components of the signal are scanned electronically by the CCD array. the two linearly polarized light components along the fast and slow axes experience a differential phase delay due to the sapphire birefringence and its thickness. To solve this problem. As a temperature reference. composed of a grating and a CCD array.3nm. About 10% of the input light can be recollected from the sensing probe with this collimation lens and the fiber bundle. with a spectral resolution of 0. Light exits from the center fiber. is then reflected by a reflector.

the optical components in the BPDI sensor system include an optical source (LED). a polarizer.Chapter 4. environmental perturbations.1 Optical spectrum induced noise The BPDI sensor system measures temperatures through detecting the OPDs in the 64 . 4. transmission fiber generated spectrum distortion and electrical noise inside the spectrometer. In order to achieve the desirable performance of the sensor system. This chapter is dedicated to discuss these degradation factors. we study the all-possible degradation factors on the temperature signatures. Instead of analyzing the noise performance of each of the individual optical components in the sensor system. such as temperature variations and mechanical vibrations. As discussed in Chapter 3. the visibility of the normalized spectrum. the performance of the BPDI sensor is limited by the noise associated with individual electronic and optical components and their combined effects.e the normalized interference spectrum. including the optical spectrum characteristic variations relative to the pre-stored optical source spectrum used for the normalization. These degradation factors might be due to noises generated by one single component or combined noises from several different components. These noises may cause the optical components to deviate from their desired functions and result in measurement errors. and an optical sensing element. Unfortunately. A detailed analysis of these noise effects on the system performance is a very important design step since it provides a guideline to achieve an optimal system performance. the mechanical vibrations. The advantage of this method is to get a systematic view and then design compensation methods to optimize the system performance. it is necessary for these optical components to perform their functions accurately. an optical fiber. can easily introduce noise to the system through the interaction between the optical components and the host media. System noise analysis and performance optimization As an optoelectronic measurement system. i. an optical collimator.

To illustrate their effects on the temperature measurement.1 Wavelength drift Two components in the BPDI sensor system may induce the center wavelength drifts: the LED source and the optical spectrometer. 4. then the spectral intensity distribution of the LED radiation can be approximated as a Gaussian curve as shown in Figure 4. Suppose the central wavelength of the reference spectrum is λ0 and the spectral bandwidth of the light source is w.e the center wavelength λ0 is changed.Yibing Zhang Chapter 4. the wavelength of the LED will drift. the changes of these two factors will occur simultaneously during the temperature measurement process. these two cases will be analyzed separately. The output spectrum of the LED is pre-measured and pre-stored in a computer as a reference spectrum for the normalization purposes.1. Most likely. The measurements from the optical . System noise analysis and performance optimization 65 sensing element sandwiched in a polarimeter in the optical spectral domain.1 Gaussian spectral intensity profile from a low coherence source (LED) When the ambient conditions change.1: (λ − λ0 ) 2 I (λ ) = I 0 exp(− ) w2 (4. i.1) I 0 /e Optical Intensity I0 λ0 B andw idth w W avelength(nm ) Figure 4. The optical power fluctuation of the source is no longer a concern in this system. while the spectrum characteristic changes will affect the OPD measurement thus the temperature measurement accuracy. or the spectral bandwidth is broadened or narrowed.

In the designed working temperature range from –50~80 oC. System noise analysis and performance optimization 66 spectrometer are temperature dependent. the central wavelength of the LED will drift from λ0 to λT. The light source is kept in an electric oven. which is normally experienced by the electronics at the room environments. resulting from the temperature dependent performances of its components (the volume grating and the CCD array). Then the temperature was decreased to room temperature slowly (in about 3 hours). and kept at that temperature for about 2 hours. The temperature coefficient of the LED is estimated to be about 0.2 shows the test results.27nm/°C. Also determined by the characteristic of the fabrication materials of the semiconductor optical sources. a thermocouple is used to monitor the temperature in the oven. During the temperature decreasing process. and the central wavelength of the LED will shift to longer wavelength as the temperature increases. the spectrum of the LED is temperature dependent.2 Temperature dependence of the central wavelength of the LED . Figure 4. the spectrometer and the sensing head were kept in the room environment.2) The temperature dependence of the light source in the BPDI sensor system is measured experimentally. The temperature of the oven is increased to 50°C. and the output spectrum of the LED is measured by the OceanOptics spectrometer.3nm/°C. λT can be expressed in term of the temperature T: δ λ = λT − λ0 = a(T − T0 ) = a∆T where a is a constant. and roughly equals to 0.Yibing Zhang Chapter 4. 859 858 Central Wavelength (nm) 857 856 855 854 853 852 25 30 Temperature (o C) 35 40 45 50 Figure 4. When the temperature changes from T0 to T.2-0. (4.

the spectral width w is 27nm. One normalized curve is plotted in Figure4. we have: 2π L  ) = 1 λp  2π L = 2kπ ⇒ L = k λ p . with the use of a reference spectrum acquired at different temperatures. This ratio is not equal to unity because of the center wavelength drift.3).4) The normalized curve is a product of the sinusoidal curve. and the shift of the central wavelength is 5nm. Obviously. we can calculate OPD from λP.2µm. visibility of the interference spectrum γ is 0. is obtained after normalization of Equation (2. k is a positive integer ⇒ 2π L λp  sin( )=0  λp  cos( (4. and the ratio between the two optical spectrum curves corresponding to different temperatures.5. Since no method can be used to acquire the spectrum of the LED in real time for the interference spectrum normalization. which contained the OPD information for the temperature measurements. the normalized spectrums are deformed from . System noise analysis and performance optimization 67 Without wavelength drifts.Yibing Zhang Chapter 4. at the peak point.3.4 shows the measured normalized spectrum when the temperature was increased to 45°C (the reference spectrum was acquired at 25°C). the pre-recorded reference spectrum acquired at temperature T0 is used to normalize the interference spectrum acquired at temperature T. Figure 4. which in turn can be related to temperature information by calibration. where the OPD is chosen to be 9.5). from Equation (2.5) containing temperature information.4) to Equation (2. sinusoidal curves defined by Equation (2.3) With a known k value. The normalized spectrum N(λ) is then expressed as: (λ − λT ) 2 ) 2πL w2 N (λ ) = (1 + γ cos( )) 2 (λ − λ0 ) λ exp(− ) w2 exp(− (4. Suppose the wavelength of the peak point is λP.

6 0. the output spectrum from the LED source becomes: .4 0. After the center wavelength of the spectrum drifts to λ'0 .Yibing Zhang Chapter 4.8 0.4. Measurement results of the normalized spectrum at 45°C (the reference is acquired at 20°C) Let us analyze quantitatively the temperature uncertainty caused by the center wavelength drift. System noise analysis and performance optimization 68 the sinusoidal curve and the peak/valley locations in the interference spectrum are shifted.3 Simulated results of the normalized spectrum with center wavelength shift effect Figure 4. 1 Normalized Spectrum 0.2 0 790 800 810 820 830 840 850 860 870 880 890 900 Wavelength(nm) Figure 4.

6) Assuming that δ λ = λ'0 − λ0 .8) The shifted peak position λP' can be acquired by differentiating Equation (4.9) λ = λ'p 2(λ'p − λ0 )δ λγ w 2 λ ' 2 p Because ∆λ = λ'p − λ p << λ p .10) .7) The normalized Equation (4. System noise analysis and performance optimization (λ − λ'0 ) 2 w2 69 I (λ ) = I 0 exp(− ) (4. we obtain.6) becomes: N (λ ) = I ' (λ ) 2(λ − λ0 ) 2πL = (1 + δ λ ) × (1 + γ cos( )) 2 I (λ ) w λ (4. from Equation (4.5) Because of the wavelength drift. then using a first order approximation yields.8) and setting it equal to zero. the temperature measurement will be affected.3).Yibing Zhang Chapter 4. exp(− ) 2(λ − λ0 ) w2 δλ ) ≈ (1 + (λ − λ0 ) 2 w2 exp(− ) w2 (λ − λ'0 )2 (4. cos( 2πL ) ≈1 λ'p 2πL(λ p − ∆λ ) 2πL 2πL∆λ 2πL∆λ sin( ' ) ≈ sin( ) = sin( 2kπ − )≈− 2 2 λp λp λp λ2p (4. and consider δ λ λ0 << 1 . and the new normalized curve will be: (λ − λ0 ) 2 ) I (λ ) 2πL w2 N (λ ) = = (1 + γ cos( )) ' 2 (λ − λ 0 ) I (λ ) λ exp(− ) w2 ' exp(− (4. ∂N ∂λ + = 2δ λ 2πLγ 2πL 2δ 2πL + ' 2 sin( ' ) + 2λ γ cos( ' ) 2 w w λp λp λp × 2πL × sin( 2πL )=0 λ'p (4.

the OPD demodulated from the shifted peak wavelength λP' is: λ'p L = mλ = L λp p λ ' p (4.11) and Equation (4. λp= 857nm. ∆λ λ P << 1 . we obtain: (1 + γ )λ3p L =L+ 2 δλ 4π γLw2 p λ (4. 4 3. a first order approximation yields: λ'p = (1 + γ )λ4 p δ λ + λP 4π 2γL2 w2 (4. The temperature measurement uncertainties can be obtained from Figure 4. OPD measurement uncertainties caused by center wavelength shift (δλ) (γ=1.11) From Equation (3. System noise analysis and performance optimization 70 Substituting these two equations into Equation (4.w=60nm) .5. L =30um.12).Yibing Zhang Chapter 4.5 0 0 2 4 6 8 Center wavelength drift δ λ (nm) 10 Figure 4.13) is used to evaluate the wavelength shift induced OPD measurement uncertainties (δOPD) in Figure 4.5 2 1.13) Equation (4.11).5.5 3 δ OPD(nm) 2.6 through the calibration relation between the OPD and the temperature.12) Using Equation (4.5 1 0.9) and assuming that δ λ w << 1 .

1.14) (4. Temperature uncertainty caused by center wavelength shift (δλ) (γ=1.5 0 0 Temperature uncertainty (°C) 2 4 6 8 Center wavelength drift δ λ (nm) 10 Figure 4. L =30um.Yibing Zhang Chapter 4. yields a first order approximation: exp(− (λ − λ0 )2 ( w + δw) 2 ) 2(λ − λ0 ) 2 w3 I w (λ ) = I 0 exp(− (λ − λ 0 ) 2 (4. assuming that the center wavelength is fixed. To see the effect of the spectral bandwidth changes on the temperature measurements qualitatively. λp= 857nm.5 1 0.w=60nm) 4. a first-order approximation is used to analyze the normalized spectral curves. shown in Equation (2. System noise analysis and performance optimization 71 3 2.3).6.16) . spectral bandwidth is another factor that can induce errors in the temperature measurements. The spectrum curve with bandwidth change δw relative to the original bandwidth w can be described as: ) ( w + δw) 2 The normalized interference fringes will then be: (λ − λ0 ) 2 exp(− ) I w (λ ) 2πL ( w + δw) 2 N w (λ ) = = (1 + γ cos( )) 2 (λ − λ 0 ) I (λ ) λ exp(− ) w2 Assuming that δ w λ0 << 1 .5 2 1.2 Spectral bandwidth broadening/narrowing effects According to the Gaussian curve approximation of the output spectrum of the LED.15) ≈ (1 + (λ − λ 0 ) 2 exp(− ) w2 δw) (4.

Yibing Zhang Chapter 4.19) 2πL(λ p − ∆λ ) 2πL 2πL∆λ 2πL∆λ ) ≈ sin( ) = sin( 2kπ − )≈− ' 2 2 λp λp λp λ2p then substituting these two equations back into (4.21) From Equation (4. Doing so gives.20) Using Equation (3. and noting that in regions around the peaks in the interference fringes: cos( sin( 2πL ) ≈1 λ'p (4. yields the following first order approximation. and further considering that δ w w << 1 .18) λ = λ 'p Assuming that δ w << w . λ'p − w = (1 + γ )λ4 p δ + λP 2 2 3 w 2π γL w (4.20) and Equation (4. we obtain: .17) and setting it equal to zero. and ∆λ λ P << 1 .21). ∂N w ∂λ = 4(λ − λ0 )δ w 2πLγ 2πL 2(λ − λ0 )δ w 2πL γ cos( ' ) + ' 2 sin( ' ) + 3 3 w λp w λp λp + 4(λ − λ0 )δ wγ 2πL 2πL × ' 2 × sin( ' ) = 0 3 w λp λp (4.18). the OPD demodulated from the shifted peak wavelength λP' will be: L p λ−w = mλ ' p−w λ'p − w = L λp (4.17) The shifted peak position λP' can be acquired by differentiating Equation (4. System noise analysis and performance optimization 72 So the normalized Equation (4.15) becomes: 2(λ − λ0 ) 2 2πL I w (λ ) = (1 + )) δ w ) × (1 + γ cos( N w (λ ) = 3 I (λ ) w λ (4.11).

w=60nm) 0. OPD measurement uncertainties caused by the bandwidth changes (δw) (γ=1.01 0 0 2 4 6 8 Bandwidth changes δ w(nm) 10 Figure 4.04 0.Yibing Zhang Chapter 4. Figure 4.7.02 0. L =30um.07 0.03 0. λp= 857nm.8. for the given .05 Temperature uncertainty (°C) 0. λp= 857nm.02 0. Temperature uncertainty caused by the bandwidth changes (δw) (γ=1. L =30um.w=60nm) 4.06 0.03 0. 0.04 0.01 0 0 2 4 6 8 Bandwidth changes δ w(nm) 10 Figure 4. System noise analysis and performance optimization (1 + γ )λ3p δw =L+ 2 2π γL w3 73 L p λ−w (4.8 shows the temperature measurement uncertainty caused by the spectrum bandwidth changes.05 δ OPD(nm) 0. Among these factors. and the collimation quality of the optical collimator. the optical alignment between the polarizer and the sensing element.2 Degradation effect due to the visibility of the interference spectrum The visibility of the interference spectrum may be influenced by the optical polarizer extinction ratio.22) This equation can be used to estimate the OPD measurement errors caused by the spectrum bandwidth changes as shown in Figure 4.7.

System noise analysis and performance optimization 74 qualities of the polarizer and the optical collimator.9. The refractive index is ne along the f-axis. The sensing element is intentionally rotated relative to the polarizer within a plane normal to the light propagation direction.Yibing Zhang Chapter 4. Misalignment between the optical polarizer and the sensing element.10. The optical alignment changes are also unavoidable in industrial fields where the mechanical vibration problem is severe. Optical wave propogation direction s-axis 45o f-axis P s α P' f Rotation in the plane Optical Polarizer Figure.and slow-axis separately as shown in Figure 4. the ordinary wave and the extraordinary . it is very difficult to align all optical components accurately in the sensor probe. Thus. The following analysis is using the optical misalignment caused visibility changes to evaluate the temperature measurement uncertainty. In fact.9) assume that the polarization direction P of the input light beam is at 45º relative to the fast. When a linearly polarized light beam passes through the sapphire sensing element.4. mechanical vibrations may cause changes of the relative positions between the sensing element and the polarizer.9.and slow-axes of the sensing element in the plane normal to the light propagation direction. Also. so that the visibilities are changed. it is necessary to analyze the dependence of the temperature measurement uncertainties on the angle α. the influence from the optical alignment between the optical polarizer and the sensing element is dominant. The results shown in Equation (2. and no along the s-axis. At the input surface of the sapphire sensing element. as shown in Figure 4. the light behaves as if it is divided along the fast.

the phase difference δ will be changed to δ = 2πd∆n / λ (4. as well as the alignment angle α between the linear polarization direction of the input light and the principal axes in the sensing element. Polarizer/analyzer polarization direction f E Ecos α s Esinα Esin2α Ecos α 2 α Figure 4.23) where d is the thickness of the sensing element.24) v v v where E out . circularly polarized or elliptically polarized. the output light from the sensing element can be linearly polarized.25) The interference results will contain the information about the phase delay δ between the two linearly polarized beams in the sensing element.Yibing Zhang Chapter 4.10. the electric field of the output optical wave can be expressed in a vector form as v v v Eout = E0 sin 2 α + E0 cos 2 α (4. namely the phase difference is δ=0. E0 sin 2 α and E0 cos 2 α are all vectors. The intensity of the output light will be: r I out = Eout 2 (4. as well as the angle α: . After light passes through the sensing element. ∆n = no-ne is the birefringence of the single crystal sapphire material. Decomposition and interference of the linearly polarized input light in the polarimeter. If another polarization analyzer is placed behind the sensing element with its polarization direction parallel to the linear polarization direction of the input light. Depending on the magnitude of the δ value. System noise analysis and performance optimization 75 wave are still in phase. and λ is the optical wavelength.

5 0. as well as the birefringence ∆n = no-ne.6 0.26) is thus essentially a function of temperature. . The factor related to the angle α in Equation (4. When the temperature changes.3 0.11. the measured optical spectra with different visibilities are plotted in Figure 4. the thickness d of the sapphire sensing element will change.7 Visibility 0. The maximum and minimum intensities in these spectral curves.11.26) 76 I out = ( E0 sin 2 α ) 2 + ( E0 cos 2 α ) 2 + 2 E02 ( 1 = E02 [1 + sin 2 2α × (cos δ − 1)] 2 δ = I 0 (1 + sin 2 2α × sin 2 ) 2 From Equation (4.12. After normalization of the measured spectral curve I(λ) with the intensity I0(λ) of the input broadband optical source. Visibilities dependence on the misalignment angle α between the light polarization direction and principal axes of the sensing element.23). the visibility of the interferogram will be γ = sin 2 2α .Yibing Zhang Chapter 4. The dependence of the visibility on the angle α is plotted in Figure 4. System noise analysis and performance optimization sin 2α 2 ) cos δ 2 (4.26) will determine the visibility of the interferogram of the two polarized beams.9 0. δ is a function of the product d∆n. 1 0. By changing the angle α.27) Figure 4.2 20 30 Measurement results Teoretical results 40 50 Angle α (°) 60 70 (4. Equation (4.8 0.4 0.

13.6 0.7 0.3 0. the BPDI technology measures the optical path difference . Measured temperature deviation caused by the changes in the visibility of the interferogram 4.2 0.1 0 45 50 55 60 Angle α (°) 65 70 75 Figure 4.4 0.12.13. as shown in Figure 4.8 0.1 0 800 820 840 860 Wavelength(nm) 880 α =30° α =35° α =40° α =45° 900 Figure 4. where mechanical vibration condition is usually severe. accurate temperature measurement results can thus still be obtained in a wide changing range of the visibilities. 0.7 Normalized Intensity 0.Yibing Zhang Chapter 4. are fixed. Since only these peaks and valleys are used to extract the temperature information.3 0.5 0.9 0.2 0. This tolerance of the angle α can be easily met in the laboratory environment as well as in the industrial environments. Experimental results demonstrate that the temperature can be measured accurately with the change of the angle α up to 10º.6 Temperature deviation (° C) 0.3 Optical birefringence effects As presented in Chapter 2. 0. System noise analysis and performance optimization 77 corresponding to the peaks and valleys in the curves. Experimental interferograms with different visibilities.4 0.5 0.

System noise analysis and performance optimization 78 (OPD). The rays that are used to define the unique (optic) axis are called ordinary rays and the refractive index is no.14. The retardation in the sensing element is a refractive index effect. the speed of light is reduced by a factor that is the reciprocal of the refractive index. An anisotropic material is completely characterized by three principal refractive indices. where no and ne are the indices of refraction corresponding to the two orthogonal linearly polarized light waves. The axes of the intersection ellipse are equal in length to 2no and 2ne.14. This is done by finding the intersection ellipse between a plane through the center of the ellipsoid that is normal to the direction of light propagation. which can be represented by an index ellipsoid. the index ellipsoid is used to determine the magnitude of the two indices of the refraction (no and ne corresponding to D1 and D2) for the two orthogonal linearly polarized waves in the material.Yibing Zhang Chapter 4. Those that are associated with the other axes are called extraordinary rays and the refractive index is ne. Method of the index ellipsoid to determine the refractive indices of light propagating along s direction. The magnitude of the birefringence in an anisotropic material depends on the propagation direction of the optical wave in the material. one principal refractive index is different from the other two. or the phase retardation between the two orthogonal linearly polarized light beams in a single crystal sapphire element. Materials with different principal refractive indices are therefore referred to as birefringent. In uniaxial materials. As shown in Figure 4. The two waves are polarized in directions parallel to D1 and D2. It is known that as light passes from the vacuum into a material. which is a function of both the temperature dependent birefringence and the temperature dependent thickness. z D1 D2 x s Figure 4. .

In the y-z or x-z plane. as shown in Figure 3. θ ) λ (4. Single crystal sapphire is a negative uniaxial crystal.2. the refractive index along the A-axis is ne.28) where T is the temperature.Yibing Zhang Chapter 4. the refractive indices along different axes are different. The refractive index along the C-axis is no.15.θ ) = 2πd (T . the phase delay between the two orthogonal linearly polarized waves will be δ = 2πd∆n/λ. thus a more general formula for the phase delay calculation should be: δ (T . The refractive index ellipsoid for single crystal sapphire. In the single crystal sapphire. where d and ∆n are the only terms that are functions of temperature. System noise analysis and performance optimization 79 Single crystal sapphire is a uniaxial crystal because of its hexagonal crystalline structure. and called the A-plane. θ )∆n(T . ne y(a-axis) To analyze the temperature measurement uncertainty caused by the non-normal incident light on the sapphire sensing element. meaning no>ne. The index ellipsoid corresponding to the hexagonal inner atomic structure is shown in Figure 4. For the case of non-normal incident light. With the propagation direction of the incident light wave perpendicular to the sapphire sensing disk surface (A-plane). the refractive index is constant in the x-y plane. there is no birefringence in this plane and is called the C-plane. and θ is the light incident angle. as well as the optical path length that the light passes through in the sapphire sensing element. z(c-axis) no ne x(a-axis) Figure 4. birefringence ∆n = no-ne exists. the sensing element is rotated about its fast axis. the birefringence ∆n = no-ne will be incident angle dependent. thus the C-axis is the slow axis and the A-axis is the fast axis.15. Therefore temperature can be measured directly by monitoring the phase delay δ with the BPDI technology. .

For this situation.16(b). Optical refractive index ellipse for the rotation about the s-axis of the sensing element .Yibing Zhang Chapter 4.3. thus the birefringence ∆n = no-ne is independent of the light incident angle. with the optical axis parallel to the boundary and perpendicular to the plane of the incidence.16. Figure4. s-axis Optical wave propogation direction P S Optical Polarizer Rotation about s-axis (a) Rotation about the s-axis of the sensing element c-axis e o (b)Wave vectors for the double refraction in the sapphire. C-axis) The rotation is shown in Figure 4.16 (a). 4. along which light passes through the sapphire disk. System noise analysis and performance optimization 80 slow axis and an axis 45o from the slow/fast axis respectively to study their effects on the OPD measurement results. shown in Figure 4. is the only factor that is incident angle dependent. the optical path L. the refractive indices for the ordinary and extraordinary waves are all constant.1 Rotation about the slow-axis (i. In this case.e.

19(a). System noise analysis and performance optimization 81 In the room temperature environment.92 7.1).76 0.18(b). as plotted in Figure 4.29) where θ1 is the light refraction angle corresponding to the light with an incident angle θ0.751 and no=1.97 7.88 Wavelength (nm) 0. Selecting the refractive indices at the wavelength of interest λ=850nm.Yibing Zhang Chapter 4.92 7.96 7.94 7.20.93 7. As shown in Figure 4. incident angles for the ordinary and extraordinary waves are plotted in Figure 4.764 1.18(a). . Refractive indices and birefringence vs. ne=1.758 1.80 0. that they can be treated as equal in analyzing the rotation effects on the temperature measurements.752 1. 1. The birefringence ∆n = no-ne is plotted in Figure 4.754 1.84 Wavelength (nm) 0. the OPD change is only related to the change in L as a result of the disk rotation.92 Birefringence (no-ne) no ne x 10 -3 7.7589.748 0.91 (a)Refractive indices Vs wavelengths (b) birefringence vs wavelengths Figure 4.84 0.750 7. The OPD values are: OPD = (no − ne ) L.89 0.17(b) as a function of the optical wavelengths. OPD' = (no − ne ) L / cosθ1 (normal incident light) ( the sensing disk rotated an angle θ 0 ) (4. the refraction angles vs. the refractive indices of the single crystal sapphire can be approximated by Equation (3.76 0.95 7.88 0.8 0.17.756 1.17(a). optical wavelength for sapphire. and the results are compared with theoretical simulations as shown in Figure 4.760 Refractive index 1.762 1. rotation angles (θ0) up to 50º have been measured.9 1. Changes of the OPDs vs. The difference between the two angles is so small as shown in Figure 4.

5 30 29.20. .19. Sensing element rotation effects along s-axis on the OPD measurements. (b) The refraction angle difference between the ordinary waves and the extraordinary waves (rotation about the s-axis) θ0 L θ1 no-ne Figure.5 31 OPD(µm) 30.04 0.5 29 28.4.5 32 31.16 0.1 0. 32. Sensing element rotation effect on the optical path difference (OPD) (rotation about the s-axis).12 0.5 28 10 20 30 40 50 Rotation angle(°) 60 70 80 Theoretical results Measurements Figure4.18 0. System noise analysis and performance optimization 82 35 Refraction angel difference (°) 0.06 0.18. (a) Refraction angles of the ordinary waves and the extraordinary waves (rotation about the s-axis).Yibing Zhang Chapter 4.14 0.02 0 0 50 100 Incident angle(°) 150 200 30 Refraction angel(°) 25 20 15 10 5 0 0 Ordinary ray Extrordinary ray 50 100 Incident angle(°) 150 200 (a) (b) Figure 4.08 0.

the rotation effects on the temperature measurements can be estimated according to the calibrated relationship between the OPDs and the temperatures.21. if the sensing element rotates. and it obeys Snell’s law: ni sin θ0 = no sin θ1 (4. with the optic axis parallel to the boundary and perpendicular to the plane of incidence. A-axis) The rotation is shown in Figure 4. In this case. the refractive index for the ordinary light wave is independent of the incident angles. Optical refractive index ellipse for the rotation about the f-axis of the sensing element. In a constant temperature environment.e. rotation angle curves shown in Figure 4. the OPD values change rather than remain constant. System noise analysis and performance optimization 83 In both the theoretical analysis and experimental measurements.2 Rotation about the fast-axis (i. It is shown that the temperature measurement uncertainty is less than 4ºC for rotation angles up to 5º. Figure 4. . Using the OPD vs.30) f-axis Optical wave propogation direction P f Optical Polarizer Rotation about f-axis (a) Rotation about the s-axis of the sensing element c-axis o e (b) Wave vectors for the double refraction in the sapphire.Yibing Zhang Chapter 4.20. it was shown that the OPD values increase with the increasing of the rotation angles about the C-axis.21 (a).3. 4. This change in the OPD values will thus affect the accuracy of the temperature measurements.

respectively.754 1.752 1. For the extraordinary waves in the A-plane sapphire disk. With rigorous mathematical analysis shown in Appendix I.Yibing Zhang Chapter 4. Refractive indices vs. the corresponding refractive index depends on the direction of light propagation. which is the refractive index of the extraordinary waves normal incident to the surface of the A-plane sapphire disk.7589). 1. the refraction angle θ2 and refractive index ne can be expressed as.22.23(b). no=1. the refraction angles for the ordinary wave and the extraordinary wave are plotted in Figure 4.753 1.7515 1. as shown in Figure 4. neo is a special ne value.23(a).31) n n (1 + tan (θ 2 )) ne = eo 2 o neo tan 2 (θ 2 ) + no 2 where n is the refractive index of the media from which the light is incident (n=1 for the incident light from the air).7525 1. Refractive indices for the extraordinary waves increase as the refraction angles increase. θ 2 = tan −1 ( 2 2 no 2 n 2 sin 2 θ 0 neo 2 (no 2 − n 2 sin 2 θ0 ) 2 ) (4. Using the refractive index for the wavelength of interest.22. light incident angles for the extraordinary waves. λ=850nm (neo=1. as shown in Figure 4. The difference between the two angles is so small.7535 Refractive index(ne) 1. System noise analysis and performance optimization 84 where ni is the refractive index of the media from which the light is incident.751. that the refraction angles in the sapphire disk for these two waves can be approximated as equal for the analysis of the rotation effects on the temperature . and θ0 and θi are the incident angle and refraction angle of the light.751 0 5 10 15 20 25 Refraction angle(°) 30 35 Figure 4.

as well as the changes in the refractive indices for the extraordinary wave.Yibing Zhang Chapter 4. System noise analysis and performance optimization 85 measurements.14 Refraction angel difference (°) 150 200 0.24.1 0.23.08 0. the results are compared with the theoretical simulations in Figure 4. d = (no − ne ) L. d ' = (no − ne )(θ 2 ) L / cosθ 2 (4. The changes in the OPDs are related to the L changes as a result of the disk rotation. (a) Refraction angles of the ordinary waves and the extraordinary waves (rotation about the f axis) (b) The refraction angle difference between the ordinary waves and the extraordinary waves (rotation about the f-axis) θ0 L θ2 no-ne(θ2 ) Figure 4.02 0 0 50 100 Incident angle(°) 50 100 Incident angle(°) 150 200 (a) (b) Figure 4.12 0. Rotation effects on the optical path difference (rotation about the f-axis) The changes in the OPDs vs.06 0.04 0.24.25. . as shown in Figure 4. rotation angles up to 50º have been measured.32) 35 30 Refraction angel(°) 25 20 15 10 5 0 0 Ordinary ray Extrordinary ray 0.

By comparing curves shown in Figure 4. the L’ value increases from the initial value L.25.5 25 24.25.3 Rotation about an axis 45º from the slow-axis/fast-axis According to the analytical and experimental results presented in the two previous .5 26 25. when the rotation is about the C-axis. and the birefringence is fixed constant at its initial value ∆n = ne 0 − no . With the relationship between the OPDs and rotation angles shown in Figure 4. the birefringence decreases from the initial value ∆n = ne 0 − no . Effect of rotation about the f-axis on the OPD measurements. based on the calibrated relationship between the OPDs and the temperature values. it is evident that the magnitude of the OPD changes are smaller when the rotation is about the A-axis instead of the rotation about the C-axis for a given rotation angle.5 27 OPD(µm) 26.25. System noise analysis and performance optimization 86 28 27. 4.5 24 10 20 Theoretical results Measurements 30 40 Rotation angle(°) 50 60 Figure 4. we can estimate that the temperature measurement uncertainty is less than 3ºC for rotation angles up to 5º about the fast axis. These results can be explained as following: when the sensing element rotates about the A-axis and the light incident angles increase. the changing magnitude is then larger than the OPD changes corresponding to the same angle rotation about the A-axis.20 and Figure 4.Yibing Zhang Chapter 4. According to the theoretical analysis and experimental measurements. the OPD values decrease when the sensing element rotates about its A-axis.3. and the combined effects cause the OPDs to decrease. OPDs increase because the L’ value increases from the initial value L.

4. it is expected that the dependence of OPD values on rotation angles will be reduced if the rotation occurs about an axis between the slow axis and fast axis.Yibing Zhang Chapter 4. which will cause the OPDs to increase.75 0 5 10 15 20 Rotation angle(°) 25 30 Figure.95 OPD(µm) 27. Experimental results are shown in Figure 4. 28 27. The total rotation effect on temperature measurements can thus be partially cancelled between the increasing and decreasing values of OPDs. . s-axis Optical wave propogation direction 45o f-axis P f s z Rotation about the axis between f-axis and s-axis Optical Polarizer y x Figure 4.8 27. This rotation can be treated as partial rotation about the fast axis.26. We tested the rotation of the sensing element about the z-axis.26.27. The direction of the z-axis is at 45º relative to both the slow axis and the fast axis of the sensing element. and partial rotation about the slow axis. Effect on the OPD measurements of rotation about an axis 45º from the faxis of the sensing element.85 27. System noise analysis and performance optimization 87 sections. With a rotation angle less than 5º.9 27. shown in Figure 4. Rotation about an 45º from the s-axis/f-axis of the sensing element.27. which will cause the OPDs to decrease. the effect on the temperature measurement uncertainty is less than 2ºC.

Yibing Zhang

Chapter 4. System noise analysis and performance optimization

88

4.4 Opto-electronic noise in spectrum measurements
The opto-electronic noise in this system is generated inside the optical spectrometer, where dispersed light from the grating is scanned electronically by a 2048-element linear silicon CCD array and converted into electric signals, then processed in the computer, as shown in Figure 4.28. The CCD detecting system is a solid-state sensor consisting of a wafer of silicon crystal, and has 2048 photodiode-capacitor pairs. Radiation coming from the optical grating hits the photodiodes, causing the capacitors to discharge. These capacitors contain charges proportional to the photon flux. At the end of the integration period, a series of switches close and transfer the charge to a shift register. The charges on the shift register are digitized and the values are amplified and sent to the computer, which plots an optical spectrum. The noises associated with the CCD array are mainly readout noise and dark current noise.

Dispersed light from grating

Photodiode

Capacitor

Shift register

2048 elements

Computer

Figure 4.28. CCD detecting array for the optical signal detection in the optical spectrometer Readout noise: The readout noise of the CCD detector is a combination of thermal white noise, 1/f noise from its electrical components, i.e. photodiodes, capacitors, shift registers and the on-chip amplifier. The white noise and the 1/f noise depend on the on-chip amplifier size of the CCD detector, and become lower when the amplifier sizes increase. But, as the amplifier size grows, its input capacitance from the sensing pixels (each pixel is composed of one photodiode, one capacitor and one shift register), also grows, lowering the sensitivities of

Yibing Zhang

Chapter 4. System noise analysis and performance optimization

89

the sensing pixels and increasing the readout noise. With an optimum design, typical readout noise is 4-5 electrons for the current commercial products. Dark current noise: The CCD array liberates electrons in the photodiode to generate electric signals. The amount of electrons generated by each CCD pixel depends on the photo flux. This process is extremely temperature dependent. Even without optical illumination, electrons can also be generated by the finite temperature of the CCD. The thermally generated electrons are called the dark current. The dark current noises can be minimized by cooling the CCD detector. The goal of the measurement of the optical spectrum electronically is the unambiguous determination of the OPDs between the two orthogonal linearly polarized light waves, which can be calculated by Equation (3.14) with the known spectral positions of the peaks and valleys in the interference spectral curves. The opto-electronic noise in the CCD detector affects the accurate determination of the peak and valley positions in the interference spectral curves. Since the ideal measured signal intensities corresponding to the valley positions are zero, the SNR at those points are thus very small and affected by the noise more seriously than the peaks points. To be conservative, we analyze the optoelectronic noise effects on temperature measurements by considering the valley points with noises. By differentiating Equation (2.4), the noise amplitude δI , or the signal uncertainty, is related to the uncertainty of the valley positions δλ :

δI = I s (λ )γ

2πLδλ 2πL sin( ) 2 λ λ

2πL sin( ) = 1, for the valleys, λ

(4.33)

where L is the OPD value, λ is the optical wavelength, I s (λ ) is the optical signal, and γ is the visibility of the interference spectral curves. This equation can be transformed into:

δλ =

 λ2 λ2  2πγL( I s(λ ) / δI )  ⇒ δλ = 2πγL SNR SNR = ( I s(λ ) / δI ) 2  

(4.34)

Yibing Zhang

Chapter 4. System noise analysis and performance optimization

90

Also according to Equation (2.14), the measurement uncertainty of the valley position will affect the OPD calculation by the following relation: OPD = f (T ) =

λ1λ2 λ1 − λ2

(4.35_a)

δf = [(

∂f ∂f δλ1 ) 2 + ( δλ2 ) 2 ]1 / 2 ∂λ1 ∂λ2

(4.35_b)

The CCD array used in the optical spectrometer is Sony® ILX511 linear CCD array, with a SNR of 250. Assuming γ=1, then the uncertainties of both OPD measurement and temperature measurement are OPD value dependent. Combing Equation (4.34), Equation (4.35_a) and Equation(4.35_b), where L=f(T), the OPD measurement uncertainties are calculated and plotted in Figure 4.29. Based on the calibration curve between OPD and temperature, Figure 4.30 shows the temperature measurement uncertainty related to the results shown in Figure 4.29. We can see that the opto-electronic noise contributes significantly to the measurement uncertainties of temperature. With the OPD increase, the temperature measurement uncertainties increase. To control the opto-electronic noise effect on temperature measurements, the OPD values should be selected properly for temperature measurement over a wide dynamic range. To increase the SNR of the spectrum measurement, there is another effective way to minimize the opto-electronic noise, and will be discussed in section 4.6.
0.3

δλ (nm)

0.25

0.2

15

20

25

OPD (µm)

Figure 4.29. Opto-electronic noise effects on the valley point locations determinations

immunity to EMI and chemical corrosion resistance. recollect light reflected from the end of the sensing probe.8 0. fiber bending-induced transmission spectrum drift in the optical fiber might induce noises in the temperature measurement (signal distortion is viewed as noise in this discussion). optical dispersion.5 1.30.2 1. and the light waves with different polarization states still contain the same spectrum information. the output temperature signature from the sensing probe is a linearly polarized light.7 15 20 OPD (µm) 25 Figure 4.4 ° Temperature deviation ( C) 1. and then guide the optical temperature signature into the spectrometer for signal processing. System noise analysis and performance optimization 91 1. optical fiber offers great system deployability in harsh environments.1 1 0.5 Optical fiber induced noises Optical fiber is used in the BPDI sensor system to guide light into the sensing probe. Since the sensing scheme for the BPDI sensor system is based on polarimetric interferometry. the state of polarization of the light is very important for the sensor to perform accurate temperature measurements. Because of its small size.9 0. The polarization mode . At the same time the optical polarization effects. When coupled back into the multimode optical fiber. its polarization state usually becomes random. Opto-electronic noise effects on the temperature measurement uncertainties 4.3 1. thus the same temperature information. The temperature information is encoded into the polarization states and the spectrum of light inside the sensing element.Yibing Zhang Chapter 4.

1 for system noises and their degradation effects on temperature measurements. Gradedindex multimode fibers used in the system have less inter-modal dispersion compared to step-index multimode fibers with the same geometrical structure and numerical aperture. and was not observed in the BPDI sensor system. this effect in the fiber is much small. Chromatic dispersion occurs because different colors of light travel through the fiber at different speeds. When light propagates inside the fiber. 4. the spectrum may be distorted because of fiber bending. . inter-modal dispersion is also called multipath dispersion. However for multimode fiber. The optical fiber dispersion in a multimode fiber includes inter-modal dispersion and chromatic dispersion. so the center wavelength and the bandwidth of the transmission light may be changed. The design of the system. collimation of light is also easier for multimode fiber with larger numerical aperture compared to single mode optical fibers. including the software and the hardware. Muliti-mode optical fiber increases the energy coupling efficiency between the LED source and the optical fiber.6 Summary of system noises and their optimization A summary is listed in Table4. System noise analysis and performance optimization 92 dispersion (PMD) may be a concern in single mode fibers for long haul transmission. is optimized based on systematic analyses of these noises. it is not a problem for a few kilometer transmission distances. where different rays or modes in the multimode fiber travel along paths of different lengths. A graded index multimode fiber with optimized index profile has a capacity of 100Mb/s over a distance of up to 100km [68]. The inter-modal dispersion is dominant in multimode fibers.Yibing Zhang Chapter 4. Compared to the optical source spectrum distortion caused by the environmental temperature fluctuations. Experiments with a 3km long multimode fiber in the BPDI sensor system revealed no obvious degradation effects on the optical pulse signals with a repetition rate of 8Hz.

06% full measurement scale) Mechanical noise Mechanical vibration (0. the spectrum is measured with dark current subtraction. Noise in the BPDI sensor system and compensation methods Noise type Noise source Temperature measurement uncertainty Electrical noise 1).00312% measurement Polarizer (analyzer).Yibing Zhang Chapter 4.Thermal white noise 3).31. to remove the dark current noise of the CCD detector.1.19% full CCD detector in the 2oC spectrometer (0. a dark spectrum can be pre-stored in the computer and then subtracted from each of the measured spectra. System noise analysis and performance optimization 93 Table 4. Spectral center wavelength drift 2).063% full measurement scale) measurement scale) (0.3% full relative measurement scale) in the range of 5o rotation position changes Opto-electronic noise reduction In a relatively constant room temperature environment.05oC Transmission fiber. To further improve the signal-to-noise ratio (SNR) of the optical spectrum measurements for the accurate peak and valley location determinations. Dark current 2).Visibility changes of the spectrum Optical source: LED. full scale) 3) 1oC Polarizer and sensing 5oC element (0. . 1/f noise Optical noise 1).Spectral bandwidth broadening/narrowing 3). Optical spectrometer. 1) 3oC (0. 2) 0. a boxcar algorithm is developed to smooth the raw data. readout noise can still be observed. In Figure 4.

a large window width will cause two problems. Second. the larger the window width. noise in each individual detector pixels can be minimized with information from the adjacent pixels. i=1. Obviously. because the interference spectrum is not a symmetric function of the wavelength. These effects are shown in Figure 4. One important parameter of the boxcar-smoothing algorithm is the window width.31.. Figure 4. . the smoothed result f s (i ) will be: f s (i ) = 1 j =w ∑ f (i + j) 2w + 1 j = − w (4. First. the averaging process will induce the deformation of the interference spectrum and shift the peak/valley positions. Suppose the measured spectrum is {f (i).. The measured interference spectrum after dark current subtraction Since the noise from each of the CCD pixels is independent. 2048}. On the other hand. the detector pixel i is smoothed using the adjacent –w~+w pixels in the boxcar smoothing process.34.36) where 2w is defined as the smoothing window width.Yibing Zhang Chapter 4.33 and Figure 4. the visibility of the spectrum will decrease when the window width is increased. the better the performance can be achieved.32. this is the character of the averaging algorithm.2. System noise analysis and performance optimization 94 Figure 4. which must be selected very carefully. so that the spectrum curves can be smoothed.

In this figure.8 0.33 shows the simulated result for spectrum curve smoothing with a large window width (w=80).8 0. the spectrum deformation is very similar to the simulated results shown in Figure 4.4 0.4 0. no obvious deformation can be observed.34 shows the measured normalized spectrum when a large window width was used (w=61). System noise analysis and performance optimization 95 Figure 4.2 0 800 810 820 830 840 850 860 870 880 890 900 Wavelength(nm) Figure 4. 1 Normalized Spectrum 0. Figure 4.33 Simulated result for spectrum curve smoothing with a large window width (w=80) . and a high visibility of the interferogram was achieved.32 shows the simulated result for a small smoothing window width (w=5. The visibility of the interference spectrum decreases dramatically and obvious deformation can also be observed.6 0. OPD=11µm).Yibing Zhang Chapter 4. 1 Normalized Spectrum 0.33.6 0.32 Simulated result for spectrum curve smoothing with small window width (w=5) Figure 4.2 0 800 810 820 830 840 850 860 870 880 890 900 Wavelength(nm) Figure 4.

Optical noise reduction When the center wavelength of the optical source shifts.Yibing Zhang Chapter 4. The basic principles are: the more the CCD pixels in a unit wavelength space. the smaller the OPD values of the sensor (which means the space between adjacent peaks/valleys is large). the positions of the valley points in the interference spectrum are also affected.37) 2πL )=0 λV 2πL = (2k + 1)π λV . the window width needs to be optimized for different CCD detectors. System noise analysis and performance optimization 96 Figure 4. the light source and the OPD values of the sensing elements. the larger window width can be used. the larger window width can be used. For the valley points in the interference spectrum Equation (4.4) becomes: cos( sin( 2πL ) = −1 λV (4.34 Measured spectrum after curve-smoothing with a window width (w=61) To obtain the best performance and prevent the deformation of the interference spectrum.

39-a) (1 + γ )λ3 0 L =L− 2 δλ 4π γLw2 v λ (4.39-b) With a first order approximation.e. the demodulated OPD is: Lv = L − λ 3 (1 + γ )λV δλ 4π 2γLw2 (4. the effect from the bandwidth variations can also be compensated using both the valley points and the peak points in the interference fringes.41) Both the peak point and the valley point are chosen to be the ones closest to the central wavelength of the light source. and Equation (4.39) shows that the location changes of the peaks and the valleys caused by the center wavelength drifts affect the OPD measurement in opposite manners.39(a)) and Equation (4.38) become p Lλ = L + (1 + γ )λ3 0 δλ 2 4π γLw2 (4.13) and Equation (4. using both peaks and valleys to calculate the OPD instead of using peaks or valleys alone.22) and Equation (4. i.38) Choosing the peak point and the valley point to be close to the central wavelength of the light source. the calculated OPD based on the shift in the valley wavelength λV' is obtained as. Lv − w = L − λ (1 + γ )λ3p δw 2π 2γL w3 (4.41) become. Applying Equation (4. System noise analysis and performance optimization 97 For the shift in the valley wavelength λV'. L= p Lλ + Lv λ + O(δλ2 ) 2 (4. and Equation (4. then λ p ≈ λV ≈ λ 0 .39(b)). Equation (4.Yibing Zhang Chapter 4. thus the wavelength drifts effect can be partially compensated by combining Equation (4. .37) on the valley points. thus λ p ≈ λV ≈ λ 0 .40) In analogy to the compensation method used for the center wavelength shifts.

The single crystal sapphire holder has the same coefficient of thermal expansion (CTE) as that of the sensing element.43) Mechanical noise reduction Most likely. reflection losses at the surface of the optical components.42-a) p Lλ − w = L − (4. the rotation of the sensing element relative to the polarizer is caused by the environmental mechanical vibrations.42-a) and Equation (4.42-b). System noise analysis and performance optimization 98 L p λ−w (1 + γ )λ3 0 δw =L+ 2 2π γL w3 (1 + γ )λ3 0 δw 2π 2γL w3 (4. a single crystal sapphire inner tube is used to hold the sensing element tightly in a fixed position.e. power losses at the polarizer/analyzer because light with only one linear polarization direction can pass. L= p Lλ − w + Lv − w λ + O(δw2 ) 2 (4.7 Power budget The power attenuation factors in the BPDI sensor system include fiber transmission losses.Yibing Zhang Chapter 4. 4. These attenuation factors are shown in Figure 4. Equation (4. i. using both peaks and valleys to calculate the OPDs.42-b) With a first order approximation.35 and their magnitudes are estimated as follows: . by combining equation (4. Then. coupling losses between the optical fiber and the optical collimator.42) shows that the spectral bandwidth narrowing/broadening effects will also affect the OPD measurement in opposite manners. thus the holder is stable and no significant thermal stress is generated at different temperature levels. the spectral bandwidth can also be partially compensated. To prevent the optical misalignment caused by the rotation of the sensing element in the sensing probe.

each of the optical quality surface reflection loss is 10%. which is a result of the surface reflection from the collimation lens.3 degree) relative to the ideal collimated propagation direction. Collimation loss (Lcollimator ) This loss includes the collimator insertion loss.Yibing Zhang Chapter 4.2dB/km at optical wavelength of 850nm.35. The output optical power from the sensing probe should be: The sensing probe with a SP structure: Pout=Pin(1-Lcoupling ) (1-Lfiber) (1-Lcollimator) (1-Lreflection-sapphire)2 (4. the total loss in the transmission fiber depends on the length of the fiber used for both guiding light into the tube and transmitting light to the optical spectrometer. so the loss is about 1dB. each of the optical quality surface reflection loss is 7%. Surface reflection loss of the optical components (Lreflection) For single crystal sapphire. Fiber loss (Lfiber) The attenuation in the multimode fiber is 2. The collimation loss in the selected collimator is about 1dB. Fiber coupling loss (Lcoupling) The coupling efficiency between the edge-emitting LED (Honeywell HFE4854-014) and the multimode optical fiber (200µm core diameter) with a ST connector is 80%. and power loss because of the optical beam divergence (0.44) . System noise analysis and performance optimization 99 Coupling loss Collimation loss LED Fiber loss Surface reflection loss Polarization loss OSA Figure 4. and for single crystal zirconia. Optical power losses in the BPDI sensor system.

Yibing Zhang Chapter 4.45) Experimentally.3µW.4µW. the input power to the SP-structured sensing probe from the 2-meter multimode fiber is 480. the output power from the sensing probe after a 2-meter long multimode fiber is 54. or 9. System noise analysis and performance optimization 100 The sensing probe with a SDZP structure: Pout=Pin(1-Lcoupling ) (1-Lfiber) (1-Lcollimator) (1-Lreflection-sapphire)4(1-Lreflection-zirconis)2 (4. .6 dB. the power loss in the system is then about 89%.

and overall performance evaluations. precision. Accuracy and precision The accuracy of an instrument indicates the deviation of the measurement results from the true value of the measurand. An instrument with good measurement repeatability needs a good design and should be carefully manufactured.1 Definitions of performance characteristics A wide range of terms has been used to describe the essential performance characteristics of measurement instruments and sensors. so it can provide the same readout. 5. resolution. But an instrument with a good measurement repeatability. does not necessarily mean that it has a good accuracy. since it could give the same wrong answer all the time. The most widely used terms include repeatability. Repeatability The repeatability of an instrument is an indication of its ability to give the same measurement results on the same quantity with repeated measurements under the same conditions. and frequency response [68]. There is always a trade off among these specifications that can be achievable at a given cost. Some of the most commonly used terminology for optical sensing system specifications and performance characteristics are defined first. Performance evaluation of designed temperature sensors With the design. accuracy.Chapter 5. implementation and optimization of the sensor system described in the previous chapters. Accuracy 101 . sensor system calibration. sensitivity. hysteresis. this chapter will summarize the experiments and results associated with the single crystal sapphire based BPDI optical fiber sensor system. The evaluation of the system includes the characterization of the spectral domain white light signal processing unit. The intention of this section is to provide clear definitions for the terms used to describe the performance of the BPDI temperature sensing system. stability. so it is a measure of its ability to tell the truth.

or as an absolute value over all working ranges of the instrument. it is charactrized by a rise time. which is a well-calibrated measurement instrument. Performance evaluation of designed temperature sensors 102 can be expressed as a percentage of the full scale readings. Stability (Drift) It is defined as the capability of an instrument to maintain the same output within a specific length of a time period. an instrument is said to exhibit hysteresis. which is defined as the time required for the sensor to respond to an instantaneous step function. as the result of mechanical friction. It is common to use twice the standard deviation as the direct measurement of the resolution. Resolution (Sensitivity) Resolution or sensitivity of an instrument is defined as the minimum resolvable change in the value of the measurand. Hysteresis If there is a difference between outputs for a given value of the measured quality when the value is approached from above or below. . magnetic effect. Accurate calibration on repeated measurement results against a standard is necessary for a good accuracy for the instrument.Yibing Zhang Chapter 5. It can be significant when rapid level fluctuations are likely to occur. elastic deformation or thermal effects. The resolution for an instrument usually has different values for different measurement scales. measured from the 10% to 90% points on the response waveforms [68]. Frequency response This is a measure of the sensor’s capability to track dynamic changes of the temperature. For the theromometer. The resolution of an instrument can be interpreted by the statistical standard deviation of a series of measurements under stable circumstances. The stability of an instrument is usually measured by the quantity of drift compared to a standard.

Performance evaluation of designed temperature sensors 103 5. changing the driving current of a semiconductor optical source. 5. .2. To make the absolute measurement meaningful. The absolute temperature measurement is very attractive for harsh environment applications because it requires no initialization and/or calibration when the power is switched on. the optical broadband polarimetric differential interferometric (BPDI) temperature sensor system provides absolute measurement of the optical path differences (OPD) between the two orthogonal linearly polarized light waves in the single crystal sapphire sensing element. The spectral characteristics profiles also change with different driving currents. The BPDI sensing system extracts temperature information by calibrated relationship between the OPDs and the temperatures.1 shows the measured output powers under different driving currents. self-compensating capability is desired so that the optical power fluctuations and fiber transmission loss variations can be fully compensated. such as an LED. The power increases as the driving current increases. as shown in Figure 5.Yibing Zhang Chapter 5.2.3 for different driving currents.2 Characterization of white light signal processor By incorporating a spectrometer for the purpose of measuring polarization properties as a function of the optical wavelength.1 Capability of compensating optical source power fluctuations To evaluate the self-compensating capability of this optical temperature measurement system. since the normalization of the interferometric signals is carried out with respect to the pre-stored LED output spectrum. Theoretically speaking. The measured spectral curves after normalization are shown in Figure 5. Figure 5. The distortion of the source spectrum would introduce errors into the measurement results through the normalization process. the output temperature variations were monitored when the optical source power was altered by changing its driving current. would also change the spectrum in addition to the optical power levels. which is fixed.

45ºC for total optical power changes up to 90%. Figure 5. .2. As shown. Output power levels of the LED with different driving currents 3500 3000 2500 2000 1500 1000 500 0 800 I=10mA I=20mA I=30mA I=50mA I=70mA I=90mA Relative intensity 820 840 860 Wavelength(nm) 880 900 Figure 5.1.4 shows the output temperature variations as a function of the normalized optical power of the optical source (LED).Yibing Zhang Chapter 5. the temperature variation range is limited to ±0.4 also indicate the contribution from the source spectrum distortions. Output spectra of the LED with different driving currents. The measurement results shown in Figure 5. Performance evaluation of designed temperature sensors 104 500 450 400 350 Power(µW) 300 250 200 150 100 50 0 20 40 60 Driving current (mA) 80 100 Figure 5.

1 mA I=88.4 -0.5 0 20 40 60 Normalized power (%) 80 100 Figure 5. Performance evaluation of designed temperature sensors 105 I=103.4 Temperature deviations vs.2 0.Yibing Zhang Chapter 5.3 mA I=43.1 0 -0.3 Temperature deviation( °C) 0.2 -0.5 mA Figure 5.1 -0.7 mA I=98. optical source output powers .4 mA I=58.3 -0.5 mA I=28.4 0.3 Normalized interference fringes for the LED with different driving currents 0.

Normalized interference fringes for transmission fiber with different attenuations . Some of the detecting pixels in the CCD array are saturated with 0dB and 1.2 Capability of compensating optical fiber transmission loss To evaluate the self-compensating capability of this optical temperature measurement system. thus the curves are deformed. the output temperature variations were also monitored when the attenuation in the multimode transmission fiber was changed.8 dB attenuation. 16.5 dB 12.4ºC for the transmission fiber attenuation up to 9 dB.5. the observed interferometric signal after normalization is shown in Figure 5.6 dB 6.9 dB 1.Yibing Zhang Chapter 5.2.3 dB 2.8 dB 0 dB Figure 5. Performance evaluation of designed temperature sensors 106 5. Keeping the spectral bandwidth and center wavelength fixed with changes only to the power levels. Figure 5.6 shows that the temperature variation is in the range of ±0.5.

The white light signal processor was also tested at room temperature for about 5 hours between the temperature increasing and decreasing processes. including the optical source and spectrometer.3 0.5. the whole white light signal processor.2 -0.3 Capability of compensating temperature fluctuations To prove the effectiveness of the first order compensation method for the spectral center wavelength drift and bandwidth broadening/narrowing. Temperature deviation vs fiber transmission loss 5. The temperature in the chamber was set to increase by a step of 5°C in the range from 10°C to 45°C then decrease with the same steps.2.7 shows the temperatures inside the chamber during the test process. so the OPD in the sensing element is fixed in order to evaluate the performance of the white light signal processor alone.4 Temperature deviation (°C) 0.6.Yibing Zhang Chapter 5. which mainly are caused by the environment temperature fluctuations. the temperature values were sampled by a K-type thermocouple inside the white light signal processor. The whole test lasted for about 90 hours and the results are shown in Figure 5.3 2 4 6 Fiber loss (d B) 8 10 Figure.1 -0. These temperature increase and decrease processes were carried out twice.2 0. was placed inside a Testequity 1000 temperature chamber.1 0 -0. The top curve in Figure. The middle curve shows the OPD fluctuations relative to the true .5. Performance evaluation of designed temperature sensors 107 0. The interferometric sensing head structure was kept unchanged in a constant temperature and pressure environment.7.

The bottom curve shows the OPD fluctuations relative to the true OPD values. broadband spectral light source illuminating the polarimeter.3 Blackbody radiation subtraction As discussed previously. 60 Temperature( C) 40 20 0 0 10 10 20 30 40 50 60 70 80 90 100 ODDfluctuation (nm) o 0 ODDfluctuation -10 0 4 2 0 -2 0 10 20 30 40 50 60 70 80 90 100 (nm) 10 20 30 40 50 60 Time (hours) 70 80 90 100 Figure.7 Experimental results of temperature compensation with the first order approximation 5.38) and verified in this test. which is improved dramatically compared to the ±9nm deviation shown in the middle curve. the temperature dependence of the OPD fluctuations is in the range of about ±9nm before using compensation. Over the entire tested temperature range.Yibing Zhang Chapter 5. In the temperature range of 10°C to 45°C. with a low coherence.5. After normalization of the measured spectral curves I (λ ) with respect .5nm. The measured OPD is obtained by using only the valley points for signal processing. Performance evaluation of designed temperature sensors 108 OPD values. the interference spectral signal contains the OPD information. the temperature dependent OPD measurement by the white light signal processor is about ±1. The measured OPD is obtained by using the white light signal processor with Equation (4.40) for the compensation. Its temperature dependence was predicted by Equation (4.

The normalized interferometric spectral curve consists of a series of maxima and minima (peaks and valleys of the interference spectral curves) at certain wavelengths.2 0 800 0 800 0 820 830 840 850 860 870 880 Wavelength (nm) (a) Optical interference signal I (λ ) (b) Optical source PSD I s (λ ) (c ) Normalization I (λ ) / I s ( λ ) Figure 5. the blackbody radiation is superimposed on the optical interference signals used for the temperature monitoring.4 500 500 0.9.9. as shown in Figure 5. Based on these peaks and valleys.6 1000 1000 0.12 for the wavelength range of interest.8.8. which shows that the effect of the blackbody radiation becomes more apparent for temperatures over 1000 oC. 3500 3500 2500 2000 1500 ÷ 810 820 830 840 850 860 870 880 890 900 Wavelength (nm) 2500 2000 1500 = 810 820 830 840 850 860 870 880 890 900 Wavelength (nm) Normalized Intensity 3000 Intensity Intensity 3000 1. 4000 o Intensity 3500 3000 2500 2000 1500 1000 500 0 800 810 1350 o C 1247 C 79 oC 820 830 840 850 860 870 880 890 900 Wavelength (nm) Figure. and becomes increasingly stronger with the increasing of temperature.2 1 0.8 0. the interference spectrum would form a sinusoidal curve if plotted in the wave numbers. the OPD values can be calculated. The measured optical spectral curves will then be elevated with the addition of the undesired background.5. Blackbody radiation effect on the measurements of optical temperature signatures . Performance evaluation of designed temperature sensors 109 to the pre-stored spectrum I s (λ ) of the broadband input light source.Yibing Zhang Chapter 5. as shown in Figure 5. Theoretical simulations of the normalized intensity distribution of blackbody radiations at different temperatures are plotted in Figure 2. Normalization of the optical spectrum As temperature increases.

thus the output light from the LED will be modulated and converted into alternating component (AC) signals. 1500oC 1400oC Normalized intensity 1. 5.6 1. and therefore the temperature measurements would be inaccurate.2 1 0. such that the CCD detector in the spectrometer can measure the spectrum without being affected by its . Digital signal filtering techniques. especially for the situation of temperature over 1000 oC.10.2 1300 o C o 1400 C 1500 o C 1300 C 1100 C 1000oC 900 o o o 0 820 830 840 850 860 870 880 Wavelength (nm) 800 oC 800 Wavelength (nm) 900 (a) (b) Figure. With the optical source modulation and the digital signal filtering techniques. as shown in Figure 5. The modulation frequency is selected to be 8Hz. The background from the blackbody radiation is essentially a direct component (DC) signal. The AC signal from the modulated optical source carries temperature information.6 0. Normalized optical spectra measured at different high temperature levels In order to minimize the influence of the blackbody radiation background. deformed sinusoidal curves are obtained and the calculated OPD values are not accurate.10. a signal generator is used to modulate the LED driving current with low frequency square-wave signals.8 0.4 1. described in Chapter 4 are used to block the DC signal. the blackbody background is subtracted sufficiently for the high temperature measurements.4 0. The AC signal will be filtered out for further signal processing for temperature measurement.Yibing Zhang Chapter 5. Performance evaluation of designed temperature sensors 110 When those background-raised spectral curves are normalized with respect to the I s (λ ) .

shown in Table 3.1 Construction of temperature calibration system A temperature acquiring subsystem is constructed to obtain the real-time temperature for the calibration purpose.1. Performance evaluation of designed temperature sensors 111 “deadtime” for electric charge transfer. the temperature values will then be one to one related to the OPD in the sapphire sensing element by software communication between the temperature acquiring subsystem and the OPD measurement and calculation subsystem.In this subsystem.05 oC) is used for temperatures below 200 oC. The sensor calibration is usually conducted by applying known temperatures within its operating range. An RS-232 serial digital communication .1oC above 500 oC) is used for temperatures above 200 oC. In the computer.11.2 oC under 500 oC and of 0.11.Yibing Zhang Chapter 5. as shown in Figure 5.4. 5. 5.4 Calibration of BPDI sensing system In order to use the BPDI sensor system for temperature measurement. it must be calibrated to relate the OPD values to the applied temperatures. Temperature monitor B-type thermocouple Rs232 interfeace Computer Figure 5. A type K thermocouple (with resolution of 0. real time temperature values from a thermocouple are displayed in the temperature monitor and then acquired by a computer using a RS232 interface. The data transfer rate for the CCD array is 13ms for the full scan into the computer memory. and a type B hightemperature thermocouple (with resolution of 0. Temperature acquisition subsystem for the calibration purpose One DPi32-C24 temperature monitor from Omega is used to acquire the temperature values from the K-type and the B-type thermocouples. The oneto-one relation between the OPD and the applied temperatures forms the calibration curve which can be stored in the host computer and later used to convert the OPD into the temperature readings.

14 shows the OPDs recorded from the BPDI sensor system during the sensor calibration process. and Figure 5.2 Temperature sensor calibration During the calibration process. the measurement error is minimized.10391 µm at temperature 1181.12.0 oC. Each of these temperature values is uniquely related to one OPD value. Figure 5. The system is thus held at a stable temperature level for about half an hour before moving to the next level. By employing the Microsoft Visual Basic software. The one-to-one relation of the applied temperature and the OPD was then used to find the calibration equation through a .887µm at room temperature environment. Figure 5. the sensing element was heated in a high temperature furnace up to 1600oC. thermal equilibrium between the sensing element and the environment is necessary.5mm and an initial OPD value of 25.Yibing Zhang Chapter 5.12. Figure 5. Real time temperature is related to an OPD value through a GUI interface 5. The sensing element used in the tests was a single crystal sapphire disk with a thickness of 1.4. Each OPD value obtained from the optical spectral temperature signature and each of the real-time temperatures acquired from the thermocouple are related simultaneously and stored in the computer. To ensure the accuracy of the calibration.13 shows the typical applied temperature data. Performance evaluation of designed temperature sensors 112 port is used to communicate with the computer directly. the OPD value is 20.15 is a plot of the OPD versus the applied temperature after averaging. As shown in Figure 5. a graphical user interface (GUI) is designed to display the realtime temperature and the related OPDs simultaneously. By taking the average of the temperatures within the temperature-holding period.

320d 3 − 29.894µm under room temperature.06d + 2419.86d − 1085. its initial OPD is 66.7 (o C ) 2 (5.836d + 604. Performance evaluation of designed temperature sensors 113 polynomial curve-fitting.011d 3 − 2.Yibing Zhang Chapter 5. Applied temperature during the sensor calibration process . and the calibrated polynomial equation is: T = 0. With the minimum root mean square error between the experiemental results and the fitting-curve.13.66d + 94. experimental results revealed that the optimal order of the polynomial curve-fitting is 3: T = 0.2) 1600 1400 1200 1000 800 600 400 200 0 Temperature(°C) 0 100 200 300 Sample index 400 500 600 Figure 5.7 (o C ) 2 (5.1) With a single crystal sapphire prism as a sensing element.

5mm .15. Performance evaluation of designed temperature sensors 114 26 25 24 OPD(µm) 23 22 21 20 0 100 200 300 Sample index 400 500 600 Figure 5. OPDs measured with the BPDI sensor system during the sensor calibration process 1600 Thermocouple temperature( °C) 1400 1200 1000 800 600 400 200 0 20 21 22 23 24 OPD values (µm) 25 26 data 1 cubic Figure 5.Yibing Zhang Chapter 5. The BPDI sensor system calibration curve for a sapphire sensing disk with thickness of 1.14.

With the calibrated sensor.1 Repeatability of the measurements Repeatability of the sensor measurement can be evaluated by applying temperature to a certain preset point repeatedly from one direction (increasing or decreasing).5.5 Performance evaluations of BPDI sensing system 5.16.17. The BPDI sensor system calibration curve for a sapphire sensing prism (equivalent to a sensing disk with thickness of 8mm) For one sensing element. accuracy compared to the thermocouple and temperature measurement resolution can then be evaluated in detail.5mm sapphire disk as the sensing element. With a 1. the accurate calibration curve was usually obtained by taking the average of several consecutive calibration data to further ensure the accuracy of the calibration. 5. the results are shown in Figure 5. such as repeatability. The temperature values on the x-axis are the thermocouple reference readouts.Yibing Zhang Chapter 5. Performance evaluation of designed temperature sensors 115 2000 data 1 cubic 1500 Temperature( °C) 1000 500 0 -500 50 55 60 OPD(µm) 65 70 Figure 5. . the performance of the developed BPDI temperature sensing system. four consecutive measurements were performed for the temperature cooling down processes from 1600 oC to 280 oC. The largest difference of the sensor output readings can be used to specify the repeatability of the sensor.

5 200 measurement measurement measurement measurement 400 600 800 1000 1200 1400 ° measurements ( o 1 2 3 4 1600 Temperature average from four C) Figure 5.17. Repeatability testing results of the temperature measurements Temperature deviation from the average( o C) 2 1. The normalized repeatability of the sensor system with respect to its dynamic range is therefore ±0.5 oC. the maximum deviation between the measured temperature and the calibrated temperature was within the range of ±2. Performance evaluation of designed temperature sensors 116 As shown in Figure 5.18.14% of the full measurement scale. Deviation of the measured temperatures with respect to the reference data .5 -2 -2.5 -1 -1. this represents the measurement precision that this system can achieve.5 0 -0. 1600 1400 Measured temperature( o C) 1200 1000 800 600 measurement 1 400 measurement 2 measurement 3 measurement 4 200 200 400 600 800 1000 1200 1400 1600 Temperature average from four measurements (oC) Figure 5.Yibing Zhang Chapter 5.5 1 0.18.

05 oC.19 shows four measurement results.2 oC under 500 oC and of 0. With a 3rd order polynomial as a calibration curve.43% for the full measurement range. The type K thermocouple has a resolution of 0.5. 1600 Temperature from optical sensor ( °C) 1400 1200 1000 800 600 400 200 0 0 First trial Second trial Third trial Fourth trial 500 1000 1500 Temperature from Thermocouple ( °C) 2000 Figure 5.5 oC. 6 First trial Second trial Third trial Fourth trial 4 Temperature deviation( Degree C) 2 0 -2 -4 -6 0 200 400 600 800 1000 1200 Temperature from Thermocouple ( Degree C) 1400 1600 Figure 5. Optical sensing system measurement results vs a B-type thermocouple measured temperatures. and a B-type thermal couple is used from 200oC up to 1600 oC.20 shows their deviations from the thermocouple reference for each of the measurements. Performance evaluation of designed temperature sensors 117 5.2 Evaluation of accuracy A K-type thermocouple is employed for the system calibration from room temperature up to 200 oC. and Figure 5. Figure 5. Deviation between the temperature measurement results from the B type thermocouple and the optical sensing system .Yibing Zhang Chapter 5. with an accuracy of ± 2 oC. with an accuracy of ± 1. which gives an accuracy in the range of ± 6 oC.19. type B hightemperature thermocouple has a resolution of 0. and 0.1oC above 500 o C.20.

Performance evaluation of designed temperature sensors 118 5.Yibing Zhang Chapter 5. Bulk optics is convenient for alignment and also relaxes requirements on vibration proof devices. The long-term stability testing results are shown in Figure 5. These experiments were carried out at different temperature levels: 26oC. The spectral domain optical signal processing technique also provides better stability than intensity based optical signal processing techniques.5 2 1. Figure 5.5. 300oC. The temperature readout differences between the thermocouple and the optical thermometer is in the range of ±2 oC for tests over 160 hours.3 Long-term stability tests Most optical sensors require accurate optical alignments.21. The BPDI optical thermometer takes advantages of both optical fibers and bulk optics to ensure its longterm stability for industrial field applications. light weight and their immunity to the electromagnetic interference (EMI). while optical thermometer performed well. which means the normalized maximum variation is 0. 31 30 29 Thermal-couple output Sapphire sensor output Temperature deviation from thermocouple output ( oC) 2. and mechanical vibration-proof working environments to maintain its long-term stable operation.5 -1 -1. 600 oC . And optical fibers can transmit light for a long distance with small attenuation.5 0 -0.13% of the full dynamic measurement range.5 -2 0 Temperature ( oC) 28 27 26 25 24 23 22 21 0 40 80 120 160 200 Time (hours) 40 80 120 160 200 Time (hours) (1) Long-term operation (a) At 26 oC (2) Deviation between the thermocouple and the optical sensor .5 1 0. 900 oC and 1200 oC.21 (e) also shows the long term performance degradation of the thermocouple operated at 1200 oC. as well as are implemented easily in the industrial environments because of their small size.

5 o 1.5 617 616.5 -1 -1.5 -1 -1.5 914 -1 913.5 915 0 914.5 621 620.5 -2 -2.5 -2 0 Temperature ( oC) 620 619.5 (2) Deviation between the thermocouple and the optical sensor Temperature deviation from thermocouple output ( C) Thermal-couple output Sapphire sensor output o 2.5 -0.Yibing Zhang Chapter 5.5 0 30 60 90 120 150 180 210 Time (hours) Time (hours) (1) Long-term operation (d) At 915 oC (2) Deviation between the thermocouple and the optical sensor .5 0 309 Temperature ( oC) 308 307 306 305 304 Thermal-couple output Sapphire sensor output 20 40 60 80 100 120 303 0 20 40 60 80 100 120 140 160 180 Time (hours) Time (hours) (1) Long-term operation (b) At 306oC 621.5 618 617.5 0 -0. Performance evaluation of designed temperature sensors 119 Temperature deviation from thermocouple output ( oC) 140 160 180 310 2 1.5 1 0.5 0 -0.5 0 20 40 60 80 100 120 140 160 180 200 20 40 60 80 100 120 140 160 180 200 Time (hours) Time (hours) (1) Long-term operation (c ) At 619 oC 917 (2) Deviation between the thermocouple and the optical sensor Temperature deviation from thermocouple output ( C) 2 Thermal-couple output Sapphire sensor output 916.5 0.5 0 30 60 90 120 150 180 210 -1.5 619 618.5 2 1.5 1 0.5 916 1 Temperature ( oC) 915.

the stability testing results of 160-hour period are not quite conclusive. Long-term stability testing results It must be noted that the complete evaluation of the system stability needs much longer testing period (for example.5. . Therefore.0686oC. The normalized resolution with respect to the dynamic range of the system was 0. one year).0696oC and 2σ=0. 1000 temperature values were acquired continuously and the resulting histogram is plotted in Figure 5.0343oC. 5. In the tests.0348oC and σ=0. Performance evaluation of designed temperature sensors 120 Temperature deviation from thermocouple output ( oC) 12 14 16 18 1210 20 1205 15 Temperature ( oC) 1200 10 1195 Thermal-couple output Sapphire sensor output 5 1190 0 1185 0 2 4 6 8 10 -5 0 2 4 6 8 10 12 14 16 18 Time (hours) Time (hours) (1) Long-term operation (e) At1205 oC (2) Deviation between the thermocouple and the optical sensor Figure 5. Therefore.005% of the full scale.22.21. the resolution of the sensor system was estimated to be 2σ=0. The standard deviation of the temperature data was calculated to be σ=0. The evaluation of the sensor resolution was performed using a calibrated sensor at room temperature and also at 618oC. It is common to use twice the standard deviation as the direct measurement of the resolution.4 Sensitivity (resolution) tests Interferometric sensors have the reputation of being extremely high sensitive by detecting the differential phase changes.Yibing Zhang Chapter 5. The resolution of the sensor system can be interpreted by its standard deviation of a series of temperature measurements at one constant temperature value.

Yibing Zhang Chapter 5. Performance evaluation of designed temperature sensors 121

250
250

200
200

150

150

100

100

50

50

0 27.76 27.78

27.8

27.82 27.84 27.86 27.88

27.9 27.92 27.94 27.96 Temperature ( Degree C)

0 618.05

618.1

618.15

618.2

618.25 618.3 Temperature (Degree C)

(a) At room temperature

(b) At 618.5oC

Figure 5.22: Histogram of temperature measurement 5.5.5 Hysteresis in the temperature measurements Hysteresis of the temperature sensor can be measured by cycling the applied temperature between the minimum and the maximum of the operating range in both increasing and decreasing directions. The hysteresis can be calculated as the largest difference among the output readings of the temperature cycles.

To accurately measure the temperature in a coal gasifier, it is important to reach the temperature equilibrium between the host media and the sapphire sensing element, which is enclosed by a sapphire tube as protection housing. The process to build the thermal equilibrium is determined by the heat transfer between the sensing element and the host media, which highly depends on the thermal properties of the sapphire material, such as thermal conductivity, heat capacitance, and the dimensions of the sensing element. protection housing as well as the temperature levels. Since all of these contribution factors are temperature dependent, a well-designed thermal model is necessary to analyze the hysteresis of the system. The theoretical descriptions and experimental data of the thermal properties of the single crystal sapphire are reasonably well known at low temperature levels, but its thermal properties at high temperature levels have not been extensively studied so a pure theoretical description is not possible, the hysteresis has to be derived experimentally.

Yibing Zhang Chapter 5. Performance evaluation of designed temperature sensors 122

1600

1400

1200 Temperature (° C)

1000

800 Temperature Decreasing Temperature Increading 400

600

200 21

22

23

24 OPD (µm)

25

26

27

Figure 5.23. Hysteresis of the BPDI temperature sensor The experimental evaluation was conducted using the high temperature furnace after the sensor system was calibrated. The applied temperature was first increased to the temperature of 1580 oC at a rate of 2 oC/minute. The temperature was then decreased to 250 oC at the same rate after it was maintained at 1580 oC for half an hour. The measurement results are shown in Figure 5.23. It is shown that the hysteresis can be as little as 3-5 oC for the operating temperature over 1200 oC, which is in the range of the measurement accuracy of the system, while for temperature below 1200 oC, the maximum hysteresis is about 50 oC. This experimental result is understandable since the thermal properties of the single crystal sapphire material, and the dimensions of the sensing element and the protection housing will not change much at different temperature levels. It is the temperature dependent heat transfer processes that result in the larger hysterises at low and medium temperature levels. Usually, three mechanisms of heat transfer—conduction, convection and radiation—regulate the sapphire probe’s temperature in a coal gasifier. Gases have relatively low thermal conductivities so conduction will not be important at all temperature levels. The convection heat transfer will dominate the heat transfer to the sensing element at low and medium temperatures (up to a few hundred degrees) because of the gases circulation in the chamber. At high temperatures (above ~800ºC), radiation becomes stronger for the heat transfer between hot walls and protruding probes in

Yibing Zhang Chapter 5. Performance evaluation of designed temperature sensors 123 addition to the convection. Radiation transfer is defined fundamentally by the Planck function, and is proportional to the fourth power of the temperature. The single crystal sapphire protection housing can absorb the radiation at its exterior surface and the absorbed radiation thermal energy is then conducted to the sensing element to reach the thermal equilibrium. It also transmits electromagnetic energy from a wavelength of 0.25

µ m in the UV range to the visible range of 6 µ m in the IR range with a transmission
larger than 80%, so the sensing element can also absorb the radiation energy directly. Radiation plus convection can then transfer thermal energy to the sensing element more efficiently than convection heat transfer alone, then the thermal equilibrium can be reached fast enough at the high temperature level, and no large hysteresis can be noticed. The goal for the designed sensor is to measure high temperature over 1200 oC in coal gasifiers. These testing results show that the hysteresis can be negligible at those temperature levels. 5.5.6 Frequency response The frequency response of a sensing system describes its dynamic response capability. In certain application environments, such as in gas turbine engines, gas temperature can reach over 1000OC in a few seconds; the dynamic response of the temperature measurement equipment is critical in such applications. While for the temperature measurement in coal gasifiers, the temperature variations are in the low frequency range, and the frequency response is not so critical compared to the requirements on its long term survivability and stability. In this section, the frequency response characteristic is evaluated for the developed BPDI temperature sensing system.

The frequency response of the developed sensor system depends on three factors, including the response time of the sensing element to the dynamic environmental temperature changes; the heat transfer rates between the sensing element and the host media; and the signal processing time needed for the temperature information extraction from the optical signals.

Assuming that the temperature distribution in the sapphire-sensing element is uniform at any instant of time. T0 is the temperature of the sensing element at time zero. transfer thermal energy to the sensing element with different rates.Yibing Zhang Chapter 5. the heat transfer mechanisms between the sensing element and the environment are mainly convection and radiation in a coal gasifier.5) where Tenv is the environmental temperature. Usually.6) From Equation (5. As discussed in addressing the hysterises of the system. The frequency response of the BPDI sensing system also depends on the signal processing time on the optical temperature signatures. T − Tenv = e( − hA / mc )t T0 − Tenv (5. and c is the heat capacity of the sensing element. including conduction.198 mc hA = 0. the three types of heat transfer mechanisms. represented by τ . These factors set the ultimate limitation on the frequency response of the sensor. Thus.105 e ( − hA / mc ) t 2 mc  mc  hA   ⇒ τ = t 2 − t1 = 2. A is the heat transfer area of the sapphire material. h is the thermal conductivity of the sapphire material. while the radiation and convection heat transfer will co-exist at high temperature levels. its dimensions and mass. The thermal response rise time for the sensing element is the time required to achieve a response from 10% up to 90% of the step input of the temperature. m is the mass of the sensing element. Performance evaluation of designed temperature sensors 124 The response time of the sensing element describes its thermal transient response characteristic.303   hA  (5.1 ⇒ t2 = 2. The signal processing time of the .6).9 ⇒ t1 = 0. Convection heat transfer is dominant from room temperature up to several hundred degrees Celsius. and given by: e( − hA / mc )t1 = 0. the temperature (T) of the sensing element is an exponential function of time [68]. convection and radiation. the frequency response of the sensor will thus be faster at high temperature levels than at low temperature levels. we can see that the rise time depends on the thermal properties of the sensing element material.

In the optical spectrometer. are kept at room temperature (26. The boxcar width is 5 pixels for the spectral curves smoothing. The probe and the sensing element are single crystal sapphire material. time delay between each electrical scan. which is composed of an optical spectrometer and a computer in the BPDI sensor system. and the boxcar width used to smooth the spectral curves. and is 153. . experimental tests are carried out to measure the rise time of the sensor system. The rise time can then be determined from these curves. Performance evaluation of designed temperature sensors 125 signal processing unit. the temperature increasing process was monitored and plotted in Figure 5. as well as the speed of digital signal processing in the computer.0mm with 25mm inner diameter. The computer is a Pentinum II desktop with a CPU frequency of 300MHz.4oC) instantly while the water was kept boiling to maintain its temperature.24. The sensor and the thermocouple for temperature reference.96 seconds (2.4 oC.5 oC to 95. and time delay between each scanning is 3ms. Instead. and the thickness of the protection tube is 3. The heat transfer mechanisms in the boiling water are mainly convection and conduction. the temperature change can then be approximated as a step function from 26.6 minutes) for the single crystal sapphire optical sensor. then inserted into the boiling water (95.19 seconds.Yibing Zhang Chapter 5. is determined by the data sampling frequency. the data sampling frequency is 100Hz. The sensing element is a right-angle prism with a base size of 8mm( width) × 25mm(length) . It is difficult to predict the rise time of the sensor theoretically by taking all these three factors into account. The first test was performed in a boiling water bath. The rise time for the thermocouple is 3.5oC).

Here the dominant heat transfer mechanism is convection at the tested temperature level of 174. as shown in Figure 5.27.4 oC.Yibing Zhang Chapter 5.01 seconds (about 19 minutes). which is closer to the environment in an actual coal gasifier. The step change of the temperature is from 26. Rise time characterization of the BPDI sensor system with high temperature furnace The second test was performed in a high temperature furnace. and the rise time for the single crystal sapphire optical sensor is 1133. Rise time characterization of the BPDI sensor system with boiling water 180 160 140 Temperature (° C) 120 100 80 60 40 20 0 500 1000 1500 2000 Time(second) 2500 3000 3500 Thermocouple readouts optical sensor readouts Figure 5.24.4 oC.25. .5 oC up to 174. The rise time for the reference Ktype thermocouple is 305. Performance evaluation of designed temperature sensors 126 110 100 90 Temperature (° C) 80 70 60 50 40 30 20 0 50 100 150 200 Time(second) 250 300 350 Thermocouple readouts optical sensor readouts Figure 5.65seconds (about 5 minutes).

chemical or biological phenomena. The sensing schemes are really rooted in the sensitivities of the birefringence and dimensions of the sensing element to these parameters through corresponding physical. 6. with its self-calibration capability associated with the optical polarimetric characteristics and the spectral domain signal processing method. rotary displacement. With the other existing technologies. The output light becomes a wavelength-encoded signal. the BPDI sensing system. the state of polarization (SOP) of the light is modulated by a sensing element. such as pressure. and an electrical voltage sensor. chemical and biological parameter measurements are usually carried out in the separate systems. most of the physical. could be employed to measure several parameters. including a pressure sensor. an angular rotation sensor. By simply changing the sensing elements.1 Multi-parameter measurement toolbox based on optical birefringence In the broadband polarimetric differential interferometric (BPDI) sensing technology. when a broad spectrum linearly polarized light passes through a polarimeter. This wavelength-encoded signal is actually the signature of the optical path difference (OPD) between the two orthogonal linearly polarized light beams propagating inside the sensing element. and long-term stability. which can be detected directly by an optical spectrometer. electrical current/voltage. with high resolution. 127 . and each measurement system has its own signal-processing unit. By relating the OPD to the measurands of interest through calibrations. and materials characterization can be realized.Chapter 6 Expansion of the BPDI sensing technology for other measurements This chapter presents several different sensor designs and testing results using the BPDI technology. great accuracy. Measurement schemes are usually quite different. the measurements of multi-parameters.

and the sensing element is sandwiched between them. the same system setup (optical source.2. By changing different sensing elements. where two optical polarizers are used.Yibing Zhang Chapter 6 Expansion of the BPDI sensing technology for other measurements 128 The chief advantages of this measurement toolbox are its simplicity and compactness. where a 45-45-90 degree cubic prism is used to reflect light back from the end of the sensing head. A two-end structure version BPDI sensor system .1. or in a two end structure. can be used for multi-parameters measurements. either in a one-end portable structure. as shown in Figure 6. Optical power 1 0 t λo Optical source ∆λ λ Coupler Optical fiber Sensor Head Signal Generator Computer Optical Spectrum Analyzer SMA connector to Fiber Fiber Collimation lens Polarizer (Analyzer) Sensing element holder Right-angle Prism Sensing Element Figure 6. as shown in Figure 6.2.1 An one-end structure version BPDI sensor system Optical power ∆λ λo Optical source λ Optical fiber Sensor Head Optical fiber Optical Spectrum Analyzer 1 0 t Signal Generator Computer SMA connector to Fiber Fiber Collimation lens Analyzer Polarizer Sensing element and its holder SMA connector to Fiber Figure 6. and only one optical polarizer is used to work both as polarizer and polarization analyzer. and signal processing unit) for the BPDI temperature sensing system described in Chapter 3.

By positioning these two sapphire elements closely and aligning the fast axis of the temperature compensator parallel to the slow axis of the pressure transducer. thus high stability can be achieved. .3. or fiber micro-bendings.Yibing Zhang Chapter 6 Expansion of the BPDI sensing technology for other measurements 129 6.4 and Figure 6. The schematic setup of the sensing head is shown in Figure 6. The pressure measurement based on photoelastic phenomena has been proposed and tested [70]. A birefringent sapphire disk is used as a temperature compensator outside the pressure environments. The problem with these pressure sensors is often the high cross-sensitivity with temperature. there is a unique relation between the pressure and the photoelasticity generated OPD in the single crystal sapphire sensing element.2 Pressure sensor with temperature compensation capability Most conventional optical pressure sensors are based on movable diaphragms. At the same time. thus the OPD is uniquely related to the integrated effects from the changes of both the birefringence and the dimensions of the sensing element. in which the pressure generates or changes the birefringence.5. One sapphire prism is put in a pressure chamber and a white-light system is used to monitor the OPD changes caused by the pressure. The BPDI sensor system interrogates the optical signal by directly detecting the spectrum instead of optical intensity. which provides immunity to the optical source fluctuation and transmission losses. all related to the surrounding pressures. the pressurebirefringent polarization-based pressure sensors with temperature compensation capability can be obtained. Preliminary tests were carried out to test the pressure sensor.6 shows the calibrated pressure sensor measurement results. the dimensions of the sensing element also varies if the surrounding pressure changes. Figure 6. small Fabry-Pérot interferometers. temperature sensitivity of the system can be fully compensated by the temperature compensator with a proper size. As shown in Figure 6. By redesigning the structure of a sapphire temperature sensor head. This new design provides temperature compensation capability by utilizing a birefringent sapphire prism as a sensing element.

Applied pressure signals on the sapphire sensing element X-axis: sampling points.5. Pressure measurements with the calibrated sapphire pressure sensor .Yibing Zhang Chapter 6 Expansion of the BPDI sensing technology for other measurements 130 Temperature compensation element Sensing element c-axis a-axis a-axis a-axis Optical fiber collimator Polarizer Window Pressure chamber Figure 6.4. Y axis: optical path difference (Unit: micron) 160 Sapphire optical sensor measured pressure (psi) 140 120 100 80 60 40 40 60 80 100 120 140 160 Pressure from semiconductor sensor (psi) Figure 6. Y axis: pressure values (Unit: psi) Figure 6. function of time.6. Pressure signals from the optical sensor X-axis: sampling points. function of time.3 Schematic design of BPDI based optical single crystal sapphire high-pressure sensing head Figure 6.

the wavelength modulation produced by the rotated retardation plate is identical to that produced by the temperature sensor. When the rotation angle changes. so high resolution and accuracy can thus be achieved because of its selfcompensation capability to optical source center wavelength drift and bandwidth . This reported sensing scheme utilizes only one notch in the spectral curve. the notch will move in the spectral curve. When a retardation plate is sandwiched between two polarizers and rotates along its fast axis. Fiber Optic rotation sensor In the BPDI sensor system. a “notch” intensity minimum is created in the broadband input optical signal. which is determined by the intensity minimum.7.3 Rotary displacement sensor Spillman reported the rotary displacement sensor based on an optical retardation plate [71]. the rotation of a retardation plate about its optical axis (fast or slow axis) rather than the thermal expansion. With a proper thickness of the retardation plate. produce the same effects in the sensing element. The measurement range of the rotary displacement will be limited by the bandwidth of the optical source and the optical spectrometer measurement range. The BPDI sensor utilizes these special points in the spectral curves instead of one notch. In the rotation sensor. optical fiber modulation. several optical minima and maxima can be generated simultaneously in the spectrum envelop of the broadband input optical signal. The accuracy and resolution of the rotary displacement measurement are limited by the optical source drift. etc. The wavelength position of the notch in the detected spectrum envelope is related to the rotation angle.Yibing Zhang Chapter 6 Expansion of the BPDI sensing technology for other measurements 131 6. d f t t pu Ou h t lig z y x s P2 nd db a r o a lig h t B ut inp P1 Rotary displacement sensing element Figure 6.

the refractive index ellipsoid of the lithium niobate becomes [44]: y2 z 2 + 2 + 2 + 2r51E x xz + 2r61Ex xy = 1 2 n x nx nz x2 (6. polarization states of light in the electrooptic crystal will be changed due to the Pockels effect or the Kerr effect. Figure 6. in the presence of an electric field along the x-axis. Optical sensor measured rotation angle(°) 40 35 30 25 20 15 10 5 0 0 10 20 30 Applied rotation angle(°) 40 Figure 6. 6.9. Based on a single crystal sapphire disk with a thickness of 1.8. which is the linear electro-optic effect in the crystal. As shown in Figure 6.1) .5mm as a sensing element.4 High electrical voltage sensor When an electro-optic crystal is exposed to an electric field.Yibing Zhang Chapter 6 Expansion of the BPDI sensing technology for other measurements 132 changes. The rotation displace measurement range is not limited by the measurement range of the optical spectrometer measurement range.8 shows the rotation angle measurement results. a preliminary sensor was constructed and tested. The BPDI sensing technology can also be applied to the electrical voltage measurement by selecting an appropriate sensing element. Preliminary tests were carried out with a single crystal lithium niobate by utilizing its Pockels effect. Rotation angle measurements with the calibrated fiber optic rotation sensor. as well as fiber transmission losses.

Yibing Zhang Chapter 6 Expansion of the BPDI sensing technology for other measurements 133 where r51 = 28, r61 = 6.8 for low frequency electrical signals. The electric field generates birefringence in the x-y plane. The magnitude of the induced birefringence depends on the electric field according to:
3 ∆n = no r61Ex

(6.2)

A piece of lithium niobate with dimensions of x(9mm) × y (9mm) × z (25mm) is used for the electric voltage measurement. To minimize the temperature effect and pyroelectric effect in the crystal, the electric field is applied in the x-axis and the light propagates along the z-axis [72]. The electrical voltage was measured up to 10KV and results are shown in Figure 6.10.

Electrical voltage

t tpu Ou ght li

P2 x y
d ban oad light Br ut inp
Wavelplate

z

P1

Figure 6.9. Sensing head designs for electrical voltage measurement.
12000 10000 Measured Voltage(V) 8000 6000 4000 2000 0 0

2000

4000 6000 8000 Applied Voltage (V)

10000

12000

Figure 6.10. Electrical voltage sensor based on the BPDI technology.

Chapter 7 Conclusions and recommendations for future work
By incorporating a spectrometer with a polarimeter for the purpose of measuring polarization properties as a function of optical wavelengths, an optical broadband polarimetric differential interferometric (BPDI) sensing technology was developed in this research. It has been applied for temperature measurement and proven to be able to provide reliable self-calibrating measurement of temperatures up to 1600oC with excellent repeatability, as well as high resolution. This optical sensing technology possesses many advantages, such as wide dynamic measurement range, intrinsic immunity to electromagnetic interference (EMI), high sensitivity, simplicity and especially long-term stability in harsh environments. This chapter summarizes the major conclusions obtained during the research. Future research work to further improve the system performance are also outlined in this chapter.

7.1 Conclusions
In this research work, an optical broadband polarimetric differential interferometric (BPDI) temperature sensor system was proposed and tested using single crystal sapphire material, which possesses a high melting temperature (over 2000°C), superior optical transparency, and ability to resist chemical corrosions. With a simple mechanicallystructured sensing probe, in conjunction with an optical spectrum-coded interferometric signal processing technique, the single crystal sapphire optical sensor can potentially measure high temperature in harsh environments with great accuracy, corrosion resistance and long-term measurement stability. Contrary to intensity based optical sensing schemes, spectrum measurement based BPDI sensor system encodes temperature information in the wavelength of an optical signal, which is then decoded by an optical spectrometer; this sensor guarantees that encoded spectrum data is not corrupted by intensity fluctuations in the optical signals.

134

Yibing Zhang

Chapter 7 Conclusions and recommendations for future work

135

Due to the crystallographic arrangement of atoms in single crystal sapphire, the material exhibits an inherent birefringence: the propagation speeds of light in the crystal are different depending on their polarization directions. The BPDI sensing system provides absolute measurement of the optical path difference (OPD) between two orthogonal linearly polarized light beams in the single crystal sapphire used as a sensing element. The OPD values are determined by the birefringence and the dimensions of the sapphire sensing element. Since both the birefringence and the dimensions of the sensing element are temperature dependent, OPD values can be related to temperature by calibration. The mathematical models of the sensor in response to the temperature have been studied in order to provide guidelines for optimal design of the sensor.

Detailed implementation of the BPDI sensing system has been described, including signal processing unit and sensing probe assembly. High performance sensor probes integrate a sensing element and an optical polarizer/analyzer with a fiber collimator to form a compact unit, thus the mechanical stability is improved. The sensing head can either be a SP structure including one right-angle Sapphire Prism with a special crystallographic orientation, acting both as a sensing element and as a light reflector, or a SDZP structure including a single crystal Sapphire Disk (as a sensing element) and a right-angle single crystal Zirconia Prism (as a light reflector). A single crystal sapphire tube is employed to protect the sensing element and hence increase its survivability in harsh environments. Testing results show that the sensor probe exhibits excellent reliability and stability needed to survive the high temperature (up to 16000C) environments. Spectral domain white light interferometry was employed to interrogate optical temperature signature in real-time. The integration of white light interfermetry signal processing scheme and the specially designed single crystal sapphire sensor probe augment the sensor’s survivability when operating in the harsh environments. In order to achieve all the potential advantages of the BPDI technology, a detailed noise analysis has been presented to gain a better understanding of the performance limitation of the BPDI sensor system. The optical noise analysis of the system identifies the largest error results from the rotation of the sensing element relative to the optical polarizer. The

In conclusion. which is the deviation of the temperature readouts from the designed optical temperature measurement system compared to the thermocouple readouts. The excellent self-calibration capability of the system was tested. Resolution of the system was estimated to be about 0. long-term system stability. optical source power fluctuations and spectrum distortions.Yibing Zhang Chapter 7 Conclusions and recommendations for future work 136 optical noise analysis also indicates that the change of the spectral characteristics of the optical source. For high temperatures over 1200 oC the hysterisis of the BPDI sensor system is negligible.07oC. measurement resolution. Comprehensive experiments have been performed to systematically evaluate the performance of the optical signal processing unit and the sensor system. hysterisis. self-calibration capability. The algorithm used to perform the signal demodulation and the OPD calculations are optimized to compensate the error due to spectrum changes and the optoelectronic noises. the BPDI sensor system offers the following major advantages over other sensors designed for the high temperature applications: 1.5 oC was obtained. The rise times of the system are also evaluated on two different temperature levels. By keeping the sensor head at certain temperature levels for a relatively longer time period. The optoelectronic noise analysis indicates that the 1/f noise and the thermal noise together form the ultimate limit on the optical spectrometer performance. such as spectrum wavelength drift and bandwidth broadening/narrowing contributes to the error of the system. the long-term drift. BPDI technology extracts absolute temperature information by absolute measurement of phase delays between two orthogonal linearly polarized light . Based on the system noise analysis results. were evaluated to be within the range of ± 2 oC. The fiber bending induces spectrum wavelength shift and bandwidth changes that can affect the system performance as well. including measurement repeatability and accuracy. Measurement repeatability of ± 2. and frequency response. The measurement accuracy is about ± 6 oC. optimization of measurement techniques is suggested to improve the system performance. which proves that the system is relatively immune to transmission fiber losses.

Yibing Zhang Chapter 7 Conclusions and recommendations for future work 137 waves. 7. because of no requirement of initialization and recalibration when power is switched on and off. The optical fibers can transmit light over a long distance with small attenuation and can be easily implemented in industrial environments because of their small size. 5. The BPDI temperature sensor provides a wide dynamic measurement range. An outline of further investigation is given below. thus providing a high degree of long-term measurement stability. 3. The BPDI optical temperature measurement system takes advantages of both fiber optics and bulk optics to simplify the design of the sensing head. to achieve the final goal of conducting accurate temperature measurements in practical harsh environments. Bulk optics is convenient for reducing the required tolerances on optical alignment and also for reducing the sensitivity to mechanical vibrations. especially in the coal gasifier. The BPDI sensor system offers high measurement resolution (better than 0. This guarantees its relative immunity to optical source power fluctuations and fiber transmission losses. 4.1oC) because of its intrinsic interferometric characteristics. 2. . future works are suggested to further enhance the robustness and reliability of the BPDI temperature sensor systems. Based on the work done so far.2 Suggestions for future works A complete and systematic performance evaluation of this sensor provides feedback information for further improving the system design. light weight and immunity to electromagnetic interference (EMI). It was tested from room temperature up to over 1600 OC. The BPDI sensor system monitors temperatures in real-time through measuring optical spectrum instead of optical intensity. which is more attractive for applications in harsh environments.

pressure. however. special consideration should be given to the required mechanical protection of both the sensor probe and the associated signal demodulation system. and the outside diameter of the envelope shell. they degrade conventional sensor probes quickly. thermal fatigue. An ideal sensor probe should contain an inexpensive and rugged shell.Yibing Zhang Chapter 7 Conclusions and recommendations for future work 138 1. Sensor protection Temperature. Sensor assembly The sensor probe material not only has to survive the environment. For the temperature measurement in the coal gasifier. The same stresses will also be experienced when the gasifier is shutting down and cooling down. and harsh chemicals are commonly used in the industrial processes. If the sensor needs to be retrieved. marginally larger than the probe. thermal stress. An optical probe must consist of a corrosion-resistant body hermetically sealed to a sapphire sensor head. it needs to be designed to handle the shutdown as well as heatup processes. since the sensor probe is subjected to the extreme mechanical stress when the gasifier is heated up from room temperature to 2800F. thermal cycling. External constraints consist of the port size in the refractory wall of the coal gasifier . but also it has to be compatible with the thermal expansion of the inside sensing element and the joining or sealing techniques. and high heat fluxes. The whole probe design and assembly must address the problem of the operating temperature and heat flux to the sensing element. although the corrosion resistance of the single crystal sapphire material has been documented by laboratory testing. 2. The shell can be made of high temperature metals for the mechanical stress protection on the inner sensing element and housing tube. the internal diameter of the probe. so that the sapphire sensor head can be physically exposed inside the coal gasifier for temperature measurements. All of these factors must be considered in the elevated temperature environments. Other concerns are the operating temperature of the probe. Temperaturerelated degradation mechanisms include: thermal shock.

Yibing Zhang Chapter 7 Conclusions and recommendations for future work 139 chamber. . For the field tests. reactivity and temperature of the environments. In order to install the designed sensing system in the coal gasifier. a complete field test of the system is necessary to comprehensively evaluate the overall performance of the BPDI sensor system. The field test should run for multiple months in order to be able to reliably demonstrate the capability of the operation in extremely harsh gasification environments. period of use. and other optical requirements. length into the hot zone. Improvement of the signal processing in digital domain The performance of the BPDI sensor system can be further improved with filtering techniques in the digital domain by the host computer. Digital filters can be implemented to reduce the blackbody radiation background noise. directness of the access. slagging conditions at the temperature measurement points. thus the temperature measurement accuracy and resolution can be improved. The field test results should also serve well to further optimize the sensor and system design for the successful commercialization of the BPDI sensing technology. Field evaluation Based on the requirements of industrial fields. 3. need to be specified for the system integration with the existing coal gasification facility. The sensor instrumentation should be made for continuous operation in an actual coal gasification facility under high temperatures and extremely corrosive conditions. which must physically reside in the gasification unit. consideration of the required environmental performance in a slagging gasifier was a prime concern in the design of the optical sensor components. and location of electronics. The electronic noise associated with the spectrum measurements can also be minimized with digital filtering techniques. 4. width of the refractory wall.

Adams.Reference 1. John Wiley & Sons. 10. Dewitt and G. San Francisco. New York : Springer. Temperature: its measurement and control in science and industry. E. Carnavale. 1988. L.. vol 6 Amer. E. March 1983. 140 . Fiber optic sensors: an introduction for engineers and scientists. R.” J. 7. Thermal physics. B. S. “Temperature and humidity measurement. 1999. Theory and practice of radiation thermometry. 6. America. Green. editor. F.” In H. 2. W. United Kingdom. D. New York. 1992.” In J. vol 6. C. New York. 2000. editor. “High-temperature optical fiber thermometer. H. pp: 39-41 4. Soc. Phys. Udd. 9. “Optical fiber thermometry for use at high temperatures. H. 11.” Industrial temperature measurement. Dils. pp: 1027-1032. Schooley. Appl. “Ultrasonic thermometry using pulse techniques. Magison. Bentley. Vaezi-Nejad. vol 4. Baierlein. Inst. 1992. Inst. P. 54(3). November 2001. pp: 739-743.” Published by the Institute of Electrical Engineers. 3. pp: 715-732. F. M. D. New York. Temperature: its measurement and control in science and industry. Pittsburgh. D. “Temperature measurement-physical principles underlie the four common methods. London. Phys. “Acoustic temperature and velocity measurement in combustion gases. devices and systems series 11. R. R. Singapore. Varela. John Wiley&Sons. 1998 5. pp: 1198-1201. Plumb. E. Inc. Temperature: its measurement and control in science and industry. “Selected topics in advanced solid state and fiber optic sensors. F. E.” Handbook of temperature measurement. editor. Cambridge University Press.” 8th Int. S. Phys. 1972. IEE circuits. Lynnworthand and E. Inst. “Temperature measurement in industrial and laboratory furnace using ultrasonic thermometry. 1991. Schooley.” In J. vol 1. August 1986 8. Nutter. Heat Transfer Conf.. R. Amer. Inc.

“Laser-scattering temperature measurements of a free burning arc in nitrogen. Phys. vol 1. ed. Temperature: its measurement and control in science and industry. Instrum. Sun. Phys. P.” J. “Exitrinsic Fabry-Perot sensor for strain and crack opening displacement measurement from –200 to 900 o C.T. Iwamura. F. Power. Temperature: its measurement and control in science and industry. W.” J. 1992. Microwave. SPIE 3201. K. Murphy and A. G. 1992.63. K. Inst. Amer. vol 6. B. New York. 14. O. Deng. and J. September 19. K. New York. F.]. editor. 1992. W. Pratt. Murphy.” Proc. and U. Wang. Noel. Claus. . Inst. D. Gunther. Graham & Trotman. Wang. A. K. London. 27. J. 22(2) 1987. of Smart Materials and Structures. 1992. Grattan. “Fiber-optic high temperature sensor based on the fluorescence lifetime of alexandrite. H. A. editor. J. Bonston. Iida. Sci. M. Phys. 21: 1985 pp: 569-570.3. editor. Wickersheim. D: Appl. F. pp: 3869-73. Amer.” In J. Dakin. Murphy. and A. [2nd rev. Bibby. D. R. Farmer. pp: 1249-54. B. 16. “A fiber optic distributed temperature sensor for high-temperature measurements. B. T. Temperature: its measurement and control in science and industry. New York. and A. A. vol 6. 17. Palmer. “Single crystal sapphire fiber-based sensors for high-temperature applications. Lewis. Schooley. Tobin. “Temperature measurement in thermal plasma by rayleigh scattering. A. D: Appl. 13. J. 1992. vol 6. F. H. pp: 715-719. Schilling. Beshears. Kurosawa. 22. W. “Fiberoptic thermometry based on photoluminescent decay times.W.” In J. N. Sun. hashiba. W. pp: 1492-1498. Lett. pp: 634-43. 1994. 21. Inst. pp: 745-749. 15. 25. and Y. L. Phys. M. D. B.” J. O. Bonn. 19. Krauss.D. Schooley. Zhang. and D. 20. “Fiberoptic thermometry and its applications. Coal gasification: existing processes and new developments. MA. “Phosphor thermometry on turbine-engine blades and vanes.V. Z. May. Amer. Phys. Schooley. K.” Electron. 18. 1992. and M.Phys.” J. A. Xiao. W. J. R. “Distributed optical fiber Ramman temperature sensor using a semiconductor light source and detector.Yibing Zhang Reference 141 12. 1981. no. Rosss. 1999. Turley. H.” Rev. pp: 85-94. Phys.” In J.

Paez. April 2001. Feber and V.” J. 1981. Am. and M. “Radiometric analysis of a fiber optic temperature sensor. Temperature measurement and control. H. and D. London. 59(10). J. Ogle. Ceramic News.” Instrumentation and Measurement Technology Conference. June 2002. Riley.. Werner. Leyden. 1983. Aug 2001. 32. Leigh.” J. Smialik and N. 1991 31. L. “Behavior of tubular ceramic heat exchanger materials in acidic ash from coal-oil-mixture combustion. Feber. and H. Wyo. . 179 Metal bonding and interactions in high temperature systems with emphasis on alkali metals (1982) 30. 1988 26. 1986. microsensors.” Proc. M. Bonnell. Chatterji. “Concurrent measurements of temperature and soot concentration of pulverized coal flames. systems. pp: 441-444. V. vol. D. Mckee and D. 2. Bull. B. vol. Laramie.2 pp: 1221 – 1226. 27. Am. and D. and W. 48. pp: 236-243. 41 No.” Nitrogen ceramics. the 18th IEEE. E. J. J.Yibing Zhang Reference 142 23. 2001.” J. Soc. Castrellon. Edited by F. pp: 741-752. M. Coal gasifiers. Energon Co. “Oxidation and hot corrosion behavior of SiN and SiAlON. Tennery. J. pp: 627-632. Strojnik. Soc 59(9-10). Hoffman. Ceram. “Alkali Vapor Transport in Coal Conversion and coal combustion systems. K. Vol. Noordhoff. “Mechanism of strength degradation for hot corrosion of alpha-SiC. 33. Fahrner. Am. Jacobson. “Review on materials. R. G.pp: 229-232. Soc. Koo. Schlichting. and devices for high-temperature and harsh-environment applications. H. pp: 249-257. Ind. J.” Ceram.. Chapter 5. “Alkali corrosion of ceramic materials”. Ceram. Y. 34. P. Ko. pp: 191-197. E.” IEEE trans. J. 68(4). Tennery and T. 62(2). “Ultrasonis thermometry system for measuring very high temperatures using high resolution signal processing. 24. Ceram. R. S. “Characterization of corrosion mechanisms occurring in a sintered SiC exposed to basic coal slags. United Kingdom. “Corrosion of silicon carbide in gases and alkaline melts. G. Lu. 6. J. IECIC 2001. M. vol. no. Kim. 1976. 1985. 2(4)1-2. Yan. Perter Peregrinus Ltd.” ACS Symposium Series.” Opt. T. M. Kim. Hastie. Eng. Henson. Kim. C. 28. R. Electron. 25. IMTC 2001. B. 35. Proc. D. Sun. 1977. 29. pp: 1255-1261.

14. R. 1995. Murphy. Gollapudi. pp: 147-151. P. D. Polarized light in optics and spectroscopy. Measures. Lett. R. Alkali corrosion of coal gasifier linings. 1991. 1992. Boston. 41. A. 1982. SPIE 985. pp:1021-23. Sept. G. Optical electronics in modern communications. Opt. 46. 38. “Pseudo-heterodyne detection scheme for optical interferometers”. Pickrell. Price. Wang. Janzen. Hogg. 1994 39. Kersey. A. Gollapudi. A. Randall. G. et al. M. S. 1588. M. Electron. 1986 37. D. pp: 625-633. 1991. High Temperature Alkali Corrosion of Low Expansion Ceramics. Dissertation." J. pp: 273-275. 1987. Boston. Chapter 1. M. pp: 1081-1083. A. Virginia Tech. R. Claus “Quadrature phase shifted extrinsic Fabry-Perot fiber optic sensors”. Oxford. of Smart Materials and Structures. Lewis. Roode and J. and C.E. S. Wang. Gunther. . T. "Sapphire optical fiber-based interferometer for high temperature environmental applications. A. D. O. Fiber Optic and Laser Sensors VI. Valis. “Overview of Mach-Zehnder sensor technology and applications. K. and Richard O. pp: 300-307. “Sapphirefiber-based intrinsic Fabry-Perot interferometer” Opt. 17. Lett.” Proc. D. and R. Kliger.” Proc. 4. Jensen. J. K. J. F. Murphy. Claus. Sun. 87. “Optical spectroscopy of oriented molecules. S. vol.Yibing Zhang Reference 143 36. A. Jackson. Vol. 28. Lett. O. 1988 40. 1990 44. “Development of a fiber Fabry-Perot (FFP) strain gauge.” Chem. pp: 1359-1399. Academic Press. K. A. Yariv. M. May. Virginia Polytechnic Institute. Dandridge and A. G. T. published by Elsevier Science Publishers. 1991. V. 16. Ph. May.D. 1997 45. 42. no. Vengsarkar. 47.W.” in high performance ceramic coatings and films. vol. 43. Rev. SPIE vol. A. Schellman and H. and R. “Corrosion resistant coatings for ceramic heat exchanger tubes operating in highly corrosive environments. D. Master’s thesis. and R. Claus. A. Murphy.

“Interferometric signals in fiber optic methane sensors with wavelength modulation of the DFB laser source. G. Stuttgart. 16.” J. 38.Yibing Zhang Reference 144 48. “Strength and transmittance of sapphire and strengthened sapphire. Bellingham WA. Lightwave. London. G. Handbook of optics-fundamentals. J. pp: 361-365.P. Ceram. Kirchner. 2st int conf optical fiber sensors. G. M.. Manolakis. A. pp:1495-1501. Wang. R. Soc. techniques& design. Sponsored by the OSA. New York : J. Chichester. 55. 1999. May. AI-Chalabi. and D. Zhao. No. 58. 59. Technol.. third edition. 1996. Technol. Opt. Jr. Nikogosyan.” J. 2001.1. B. Philp. 1981.12. vol. J. Z.” Appl. T. pp:116-117. Bosselman. Soc. E. B. vol II. SPIE Press. J. Wang. Lightwave. 1959. H. Davies. Mencaglia. Properties of optical and laser-related materials: a handbook. “New calculus for the treatment of optical systems. Wang. R. pp: 2471-2477. Culshaw. Hartnett and H. C. Am. Stewart. Maxwell. 49. A. A. and R. 1984. “Strength of synthetic single crystal sapphire and ruby as a function of temperature and orientation. “High accuracy position-sensing with fiber-coupled white light interferometers.” J. McGraw-Hill. second edition. vol. 1983.M. vol. Henderson. Wachtman. England. and W. W. 42. 50.H. W. 19. Harris. A. Jones. . Chapter 1. no. E.” Proc. no. 1st int conf optical fiber sensors. Gangopadhyay and P. “Partially coherent sources in interferometry. 1995. C. and R. 52. Materials for infrared windows and domes. Am. Prentice-Hall. Ceram. 64. Gentilman.” J. T. Am 31. 1941. Starrett. pp: 43-53. 51. Jin. S. Opt. T. Wiley. Ulrich. 53. “Self-calibrated interferometric/intensity-based optical fiber sensors. K. 1998. Proakis and D. pp: 488-495. 54. Soc. D.S. “Vibration: history and measurement with an extrinsic Fabry-Perot sensor with solid-state laser interferometry. 10. G. L. H. N. Xiao. 56. Maguire. and L. 1997. D.” J. pp: 432-433. 1999 57. Apr. J. Bass. N.” Proc. pp: 132135. Digital signal processing.

X. C. C. Properties of lithium niobate. “Precision and sensitivity optimization for white-light interferometric fiber-optic sensors. D. Haris and L. X. P. pp: 70 -74.” J.F. pp:11-22. 62. K. Khattak. R. L. J. and S. “Strength of sapphire as a function of temperature and crystal orientation. McGraw-Hill Inc. pp:43-51. 1990.’ Proc. P. Harris. SPIE 3705. A. 71. Wang. 1999. heat treatments and doping. P. pp:36-43. A.” Proc. Johnson. SPIE. W. 66. Institution of Electrical Engineers. He. 70. 27 (15) 1988. Lightwave. pp: 3081-3085. B. 1999. 65. “Overview of progress in strengthening sapphire at elevated temperature. Harris. Jackson. “High-resolution white light interferometer and its application”. 64. W. 2000. Duggan. Lightwave. Schmid. Opt. and P. F. SPIE. D. Schmid. Holman. C. “Numerical investigation of the effects of birefringence and total internal reflection on Farady effect current sensors”. Lin. pp: 5366-5374. C. 1999. Abrahams.” Appl. J. Compton. No25. 1994 68. 1992. “Optical fiber pressure sensor based on photoeleasticity and its application. G. “Increase of high temperature strength of sapphire by polishing. “Fiber optic rotary displacement sensor with wavelength encoding. K. Eng. 69. SPIE 3705. 61. Fuhr. A. New York: INSPEC. Experimental methods for engineers. 2001. Vol 10. pp1466-1470. 1326. pp: 2-11. S. Kajenski.” Proc. Bush and D. “Navy mechanical test results from the sapphire statistical characterization and risk reduction program.V. Khomenko. Garcia-Weidner. pp: 44-50.Yibing Zhang Reference 145 60. . no. Schmid. 63. Tapia-Mercado. Schmid. J. 67. J. and P. SPIE 4102. Pickrell. A. Jin. Appl. A.” J. SPIE 3705. G. Fang. et al. W.” Proc. Qi. Khattak. B. S. A. C. S. 6th edition. F. J. Technol. A. London. Technol. P. N. Jaeger and D. Vol: 19 no:1. Spillman.” Proc. “Increase of the strength of sapphire by heat treatments.10. Jr. Fischer. Submitted to Opt. C. vol 31. 1989. Ivanova. Opt. P. Spetember 1992.

Appendix A: Determination of the refraction angle and refractive index corresponding to the extraordinary waves in the single crystal sapphire b θ2 -a θ0 oO t Sapphire K a Air Figure 1. Shown in Figure 1. it follows the Snell’s law. According to the simple trigonometric function. sin 2 θ 0 = OK sin 2 θ 2 Within the half ellipse (arc –a-b-a). the refractive index of the extraordinary wave inside the single crystal sapphire will depend on the incident angles. Light wave refraction at the surface of the A-plane cut single crystal sapphire When light propagates at the interface of air and a single crystal sapphire with inherent birefringence (A-plane cut). the relationship between t and θ2 is: tan θ 2 = X K a cos t a a = = ⇒ tan t = YK b sin t b tan t b tan θ 2 (3) 2 2 (1) (2) By putting Equation (2) and (3) into Equation (1). for light incident with an angle θ0. the length of line OK is: X K = a cos t  coordiantes for the K po int YK = b sin t  OK = a 2 cos 2 t + b 2 sin 2 t Where t is an intermediate variable with no physical meanings. the refraction angle is θ2 and the magnitude of refraction index is OK . we have: 146 .

. Then Equation (6) is then converted into: θ 2 = tan −1( 2 2 no 2 sin 2 θ0 neo 2 (no 2 − sin 2 θ0 ) 2 ) (7) n n (1 + tan (θ 2 )) ne = eo 2 o 2 neo tan (θ 2 ) + no 2 These refraction angles and refractive indices are corresponding to the extraordinary waves with incident angle θ0 in the single crystal sapphire. where a = no b = neo OK = ne neo is a special ne value. then: θ 2 = tan −1 ( OK = 2 2 a 2 sin 2 θ0 b2 (a 2 − sin 2 θ0 ) 2 ) (6) b a (1 + tan (θ 2 )) b 2 tan 2 (θ 2 ) + a 2 Corresponding to the index ellipse of the single crystal sapphire.Yibing Zhang Appendix A 147 sin 2 θ0 = (a 2 cos2 t + b2 sin 2 t) sin 2 θ2 a 2 + b 2 tan 2 t 2 = sin θ2 1 + tan 2 t b2 a 2 (1 + tan 2 θ2 ) tan 2 θ2 = 2 b tan 2 θ2 + a 2 1 + tan 2 θ2 = b2 a 2 tan 2 θ2 b2 tan 2 θ2 + a 2 (4) Where the following trigonometric functions are used: sin 2 β = tan 2 β  1 + tan 2 β    β represents t or θ 2 1 2  cos β = 1 + tan 2 β   (5) Utilizing both Equation (4) and Equation (2). it is the ne for the normal incident light wave to the surface of the A-plane sapphire disk.

Virginia Tech Central Processing Unit Coefficient of Thermal Expansion Direct Component State of Polarization Graphic User Interface Light Emitting Diode Optical Path Difference Polarimetric Structure Extrinsic Fabry-Perot Interferometry Electromotive Force Electromagnetic Interference Fast Fourier Transform Fabry-Perot Full Wavelength Half Maximum Integrated Gasification Combined Cycle Infrared Radiation Numerical Aperture Polarization Mode Dispersion Root-Mean-Square Resistance Temperature Detector Sapphire prism Sapphire Disk Zirconia Prism 148 .Appendix B: Acronym A/D card A/D converter AC signal BPDI CCD CPT CPU CTE DC SOP GUI LED OPD PS EFPI EMF EMI FFT FP FWHM IGCC IR NA PMD RMS RTD SP SDZP Analog/Digital card Analog/Digital converter Alternating Component signal Broadband Polarimetric Differential Interefrometry Charge Coupled Device Center for Photonics Technology.

Yibing Zhang SNR SCIIB UV USB Appendix B Signal Noise Ratio Self-Calibrated Intensity/Interferometric Base Ultraviolet Universal Serial Bus 149 .

optical fiber communication devices. fiber coupler. EDFA. he studied Physics at Washington State University (WSU). Philip L. he has worked on optical fiber sensors. and 3-D optical imaging systems for biomedical applications. He has been working in the field of laser technologies for high-power laser and ultra-short laser pulse generations. Department of Physics. WSU. photonic components (FBG. Marston on light scatterings at the Physical Acoustics and Optics laboratory. modulator. etc) design and implementation at the system level. . 1973. China. and optical metrologies for environmental evaluation and material measurements.Yibing Zhang VITA VITA 150 Yibing Zhang was born in China. College of Engineering. He received his BS degree in physics in 1995 and MS degree in optics in 1998 from Nankai University. and did research with Dr. Torgersen Graduate Student Research Excellence Awards for year 2003. He joined the Center for Photonics Technology at Virginia Polytechnic Institute and State University (Virginia Tech) in 2000. He is one of the winners of the Paul E. Virginia Tech. From 1998 to 1999. Since 2000.