close

Вход

Забыли?

вход по аккаунту

?

Microwave nondestructive testing and evaluation of electricalproperties of lossy materials

код для вставкиСкачать
MICROWAVE NONDESTRUCTIVE TESTING AND EVALUATION OF
ELECTRICAL PROPERTIES OF LOSSY MATERIALS
BY
UGUR CEM HASAR
B.Sc., Cukurova University, Adana, Turkey, 2000
M.Sc., Cukurova University, Adana, Turkey, 2002
DISSERTATION
Submitted in partial fulfillment of the requirements for
the degree of Doctor of Philosophy in Electrical Engineering
in the Graduate School of
Binghamton University
State University of New York
2008
3324358
2008
3324358
© Copyright by Ugur Cem Hasar 2008
All Rights Reserved
Accepted in partial fulfillment of the requirements for
the degree of Doctor of Philosophy in Electrical Engineering
in the Graduate School of
Binghamton University
State University of New York
2008
Charles R. Westgate, Department of Electrical and Computer Engineering,
Binghamton University
Mehmet Ertugrul, Department of Electrical and Electronic Engineering,
Ataturk University, Turkey
James Constable, Department of Electrical and Computer Engineering,
Binghamton University
Vladimir Nikulin, Department of Electrical and Computer Engineering,
Binghamton University
iii
ABSTRACT
The measurements of reflection and transmission complex scattering (S−) parameters are
generally utilized for characterization of electrical properties of materials at microwave
frequencies. There are three disadvantages of these measurements: a) ambiguities in
phase; b) errors caused by shifts in calibration planes; and c) phase uncertainty of
reflection scattering parameters of low–loss materials. The connection between the
characterization of electrical properties of materials and the application of these
measurements in industrial−based applications is simply lacking. The first part of this
dissertation focuses on two new measurement methods to fill this gap by eliminating the
aforementioned disadvantages.
In the literature, microwave nonresonant transmission/reflection methods for
materials electrical characterization eliminate either the dependency of calibration−plane
on measurements or the dependency on thickness in the measurements. A recently
proposed method to simultaneously eliminate these dependencies is inadequate for both
complex permittivity and complex permeability measurements. In the second part of this
dissertation, we propose a unique technique for calibration−plane invariant and
thickness−independent complex permittivity and complex permeability determination of
granular and/or liquid materials.
Cement−based materials (cement paste, mortar, concrete, etc.) are the most
widely used construction materials in the world. Knowledge of mechanical, mixture
proportion, chemical properties as well as electrical properties of such materials is
iv
important for evaluation of their quality and integrity. The methods commonly used in
civil engineering for characterizing their mechanical properties are generally destructive.
The third part of this dissertation presents microwave nondestructive reflection and
transmission properties of young mortar specimens with different w/c ratios over their
early curing periods and hardened mortar and concrete specimens with different w/c
ratios over their late curing periods.
A direct application of available microwave methods in the literature to
industrial−based applications may not be appropriate since these methods generally
necessitate expensive instruments. Therefore, there is a need for a simple and relatively
inexpensive microwave sensor for electrical characterization of materials in
industrial−based applications. In the final part of the dissertation, we propose a
microcontroller−based microwave free−space sensor for measurements of electrical
properties of lossy materials.
v
DEDICATION
I would like to dedicate this dissertation to my parents, Harun Hasar and Ayse Hasar, my
brothers, Hasan Alper Hasar and Ahmet Alpay Hasar, and my fiancée, Hafize Yuzgulec.
I would not have been able to accomplish this goal without their invaluable and endless
supports, patience, and understanding.
Bu çalışmayı, aileme, kardeşlerime ve sözlüm Hafize Yüzgüleç’e ithaf ediyorum. Sizlere
ne kadar teşekkür etsem azdır.
vi
ACKNOWLEDGMENTS
I wish to express my gratitude to my academic advisors, Dr. Charles R. Westgate, of
Electrical and Computer Engineering Department, Binghamton University, and Dr.
Mehmet Ertugrul, of Electrical and Electronic Engineering Department, Atatürk
University, for all of their knowledge, guidance, patience, and encouragement throughout
the initiation, preparation, and completion of this dissertation. Without their great help
and supports, I would not have achieved this.
I would like to thank Dr. James Constable, Dr. Vladimir Nikulin, and Dr. John
Fillo of the Thomas J. Watson School of Engineering and Applied Science for
participating in and serving on my committee and for their practical insight regarding my
dissertation material.
I also would like to take this opportunity to thank my former academic advisors,
Dr. Sergey N. Kharkovsky, of Electrical and Computer Engineering Department,
Missouri University of Science and Technology (formerly University of Missouri-Rolla)
and Dr. Victor Pogrebnyak, of Electrical Engineering Department, University at Buffalo
for his inspiration and invaluable support.
With this opportunity, I also would like to thank all the faculty members in
Electrical and Electronics Engineering Department and my friend Cüneyt Gözü and
faculty members in Industrial Engineering Department at Atatürk University for their
help and endless support.
vii
Finally, I would like to thank the Higher Education Council of Turkey (YOK) for
providing me with a scholarship to pursue my Ph.D. degree in USA. Also, I would like to
express my thanks to the Scientific and Technological Research Council of Turkey
(TUBITAK) for supporting me two scholarships during my M.Sc. and earlier Ph.D.
studies.
viii
TABLE OF CONTENTS
LIST OF TABLES .......................................................................................................... xi
LIST OF FIGURES ....................................................................................................... xii
CHAPTER 1
INTRODUCTION ............................................................................... 1
1.1 Background .......................................................................................... 1
1.2 Overview of the Chapters .................................................................... 7
CHAPTER 2
2.1
2.2
2.3
2.4
CHAPTER 3
3.1
3.2
3.3
3.4
REVIEW OF PREVIOUS STUDIES ................................................ 10
Novel Methods for Lossy Materials Characterization ....................... 10
A New Sample Holder for Low-Loss and Lossy Materials
Characterization ................................................................................. 13
Electrical Characterization of Cement-Based Materials .................... 16
A Simple Microwave Sensor for Electrical Characterization of
Materials in Industrial-Based Applications ....................................... 21
TWO MICROWAVE METHODS FOR COMPLEX PERMITTIVITY
DETERMINATION OF LOSSY MATERIALS USING
AMPLITUDE-ONLY S-PARAMETER MEASUREMENTS ......... 23
Theoretical Analysis .......................................................................... 24
3.1.1 Complex Permittivity Extraction Using the Classical MultipleReflections Model .................................................................. 27
3.1.2 Validation of the Derived Forward Scattering Parameters and
Derivation of Reverse Scattering Parameters ........................ 29
An Amplitude-Only Method for Permittivity Determination of Lossy
Materials ............................................................................................ 31
3.2.1 Selection of a Suitable Approach for Permittivity Extraction 32
3.2.2 Single-Pass Technique ........................................................... 33
3.2.3 Derivation of a Simple Objective Function ........................... 35
3.2.4 Functional Analysis of the Derived Objective Function ........ 39
3.2.5 Refining the Computed Complex Permittivity ...................... 40
An Amplitude-Only Method for Thickness-Independent Complex
Permittivity Determination of Lossy Materials ................................. 41
3.3.1 Zero-Order (ZO) Approach for Complex Permittivity
Determination ........................................................................ 43
3.3.2 Higher-Order (HO) Approach for Complex Permittivity
Determination ........................................................................ 46
Measurements .................................................................................... 47
ix
CHAPTER 4
4.1
4.2
4.3
CHAPTER 5
5.1
5.2
5.3
5.4
5.5
5.6
CHAPTER 6
6.1
6.2
6.3
6.4
CHAPTER 7
7.1
7.2
A WAVEGUIDE SAMPLE HOLDER FOR AUTOMATED
THICKNESS-INDEPENDENT COMPLEX PERMITTIVITY AND
COMPLEX PERMEABILITY MEASUREMENTS OF GRANULAR
AND LIQUID MATERIALS ............................................................ 56
Theoretical Analysis .......................................................................... 58
The Method ........................................................................................ 60
Measurements .................................................................................... 62
A SIMPLE FREE-WAVE METHOD FOR ELECTRICAL
CHARACTERIZATION OF CEMENT-BASED MATERIALS …. 69
Theoretical Background ..................................................................... 71
The Method ........................................................................................ 72
The Measurement Set-up ................................................................... 73
Calibration Procedure ........................................................................ 75
Measurements .................................................................................... 77
5.5.1 Measurements of Young Mortar Specimens .......................... 77
5.5.2 Measurements of Hardened Mortar and Concrete Specimens 90
Measurement Uncertainties ............................................................. 101
A SIMPLE AND RELATIVELY INEXPENSIVE
MICROCONTROLLER-BASED MICROWAVE FREE-SPACE
MEASUREMENT SYSTEM FOR INDUSTRIAL-BASED
APPLICATIONS ............................................................................. 108
The Measurement System ................................................................ 109
Calibration Procedure ...................................................................... 112
Algorithm ......................................................................................... 116
Results .............................................................................................. 121
6.4.1 Validation of Programs and the Microwave Section ........... 121
6.4.2 Validation of Whole Measurement System ......................... 123
CONCLUSION AND FUTURE WORK ........................................ 127
Conclusion ………………............................................................... 127
Future Work ..................................................................................... 131
REFERENCES ........................................................................................................... 133
x
LIST OF TABLES
Table 4.1
Table 5.1
Table 5.2
Table 5.3
Table 5.4
Table 6.1
Table 6.2
The parameters ε ∞ , ε s , and τ for distilled water and methanol. The
data are estimated by interpolation from [104],[105] for room
temperature (20–25 Co) ............................................................................66
Mass percentages of raw materials of cement samples ........................... 78
The absolute temporal change of reflection coefficient of each young
mortar specimens ..................................................................................... 81
Mixture proportions of mortar and concrete specimens .......................... 90
Reflection properties from various sides of mortar and concrete
specimens with different w/c ratios after 3 months of curing .................. 96
The comparison of the permittivity results ............................................ 121
The comparison of the measured R and T from [66] and computed
Rc and Tc by all µC programs .............................................................. 122
xi
LIST OF FIGURES
Fig. 3.1
Fig. 3.2
Fig. 3.3
Fig. 3.4
Fig. 3.5
Fig. 3.6
Fig. 3.7
Fig. 3.8
Fig. 3.9
Fig. 4.1
Fig. 4.2
Fig. 4.3
Fig. 4.4
Fig. 4.5
Fig. 5.1
Fig. 5.2
The scattering parameter measurements of a sample sandwiched
between two dielectric plugs .................................................................... 25
The diagram for the analysis of the five–layer structure in Fig. 3.1 by
multiple reflections approach ................................................................... 27
Conformal transformation between ε 3 and ( χ , ξ ) pair ........................... 35
Dependency of f (ξ ) over ξ for different ε 3 ......................................... 40
The measurement set–up .......................................................................... 47
Comparison of the frequency response (R) with/out aperture smoothing
and the TRL calibrations for (a) waveguide short and (b) thru connections
................................................................................................................... 50
Measured ε of an antifreeze solution extracted by the plug–loaded two–
port transmission–line method (PLTL) [25] and the proposed method
(PM) in section 3.2 without aperture smoothing and with 20 % aperture
smoothing ................................................................................................. 52
Measured ε of a binary mixture of ethyl alcohol (75%) and water (25%)
by the plug–loaded two–port transmission–line method (PLTL) [25] and
the proposed method (PM) in section 3.2 without aperture smoothing and
with 20 % aperture smoothing ................................................................. 53
(a) Real part and (b) imaginary part of the measured ε of methanol by the
plug–loaded two–port transmission–line method (PLTL) [25] and by the
proposed method (PM) in section 3.3 without aperture smoothing and with
10 % and 20 % aperture smoothing ......................................................... 55
The complex S–parameter measurements of granular and liquid samples
sandwiched between stable and movable dielectric plugs ....................... 58
(a) Real part and (b) imaginary part of the measured (dashed line) and
theoretical (solid line) ε of distilled water ……...................................... 63
(a) Real part and (b) imaginary part of the measured (dashed line) and
theoretical (solid line) ε of methanol ….................................................. 64
(a) Real part and (b) imaginary part of the measured ε of an antifreeze
solution by the plug–loaded two–port transmission line method (PLTL)
[25] and by the proposed method ............................................................. 65
Dependency of the computed length of air–filled section, L0 , inside the
proposed sample holder from commercially available antifreeze solution
measurements over frequency .................................................................. 68
The configuration for the measurement of reflection and transmission
properties by using a free–space method ................................................. 71
The schematic diagram of the measurement set–up ................................ 74
xii
Fig. 5.3
Fig. 5.4
Fig. 5.5
Fig. 5.6
Fig. 5.7
Fig. 5.8
Fig. 5.9
Fig. 5.10
Fig. 5.11
Fig. 5.12
Fig. 6.1
Fig. 6.2
Fig. 6.3
Fig. 6.4
Fig. 6.5
Fig. 6.6
Fig. 6.7
Hourly reflection properties of young mortar samples with different w/c
ratios at 8.5 GHz (a) between 20–30 hours and (b) between 44–54 hours
.................................................................................................................. 80
Hourly transmission properties of young mortar samples with different w/c
ratios at 8.5 GHz (a) between 20–30 hours and (b) between 44–54 hours
.................................................................................................................. 84
(a) Real part and (b) imaginary part of the complex permittivities of young
mortar samples with different w/c ratios between 20–30 hours .............. 88
(a) Real part and (b) imaginary part of the complex permittivities of young
mortar samples with different w/c ratios between 44–54 hours .............. 89
Results of transmission properties from right sides of mortar and concrete
specimens with different w/c ratios .......................................................... 91
Application of the relative difference approach to (a) the same type
specimens and (b) different type specimens ............................................ 93
The real part of the ε of mortar and concrete specimens with different w/c
ratios and sides ......................................................................................... 99
The imaginary part of the ε of mortar and concrete specimens with
different w/c ratios and sides .................................................................. 100
Dependencies of reflection and transmission properties of w/c=0.40 mortar
sample with ε = 21 − j 0.69 on sample thickness, L ............................. 102
Dependency of Z 2 versus specimen thickness, L , for ε = 6.95 − j 0.22
and ε = 11.18 − j 0.60 .............................................................................. 105
The schematic diagram of the microcontroller–based microwave free–
space measurement system .................................................................... 110
The circuit board of the electronic section ............................................. 111
The schematic diagram of the calibration technique ............................. 113
The simplified algorithm for each program ........................................... 118
The memory footprint, total ROM space allocated, and some information
related to the program, which uses Newton’s method for ε , written in
BASCOM ............................................................................................... 120
Hourly measurement of r and t at 10 GHz for the cement paste sample
with w/c=0.375 ....................................................................................... 124
Determined ε of the cement paste sample using the measurements in
Fig. 6.6 ................................................................................................... 125
xiii
CHAPTER 1
INTRODUCTION
1.1 Background
Material characterization is an important issue in many material production processes and
management applications in agriculture, food engineering, medical treatments,
bioengineering, and the construction industry [1]. In addition, microwave engineering
requires precise knowledge of the electromagnetic properties of materials at microwave
frequencies since microwave communications are playing more and more important roles
in military, industrial, and civilian life [2]. For these reasons, various microwave
techniques, each with its unique advantageous and limitations [2], [3], are introduced to
characterize materials’ properties.
Materials and manufactured products are often tested before delivery to the user
to ensure they will meet expectations and remain reliable during a specified period of
service. It is essential that any test made on a product intended for future use in no way
impairs its properties and performance. Any technique used to test under these conditions
is called a nondestructive testing (NDT) method. The monitoring of any structural
changes and properties such as defect initiation and growth in materials and products
during the production and service lives by NDT methods is known as nondestructive
evaluation (NDE) [4].
1
Methods in NDT can be roughly divided into five categories. These are: 1)
radiological methods: X–rays, gamma rays, and neutron beams; 2) acoustical and
vibrational methods: ultrasound and mechanical impedance measurements; 3) electrical
and magnetic methods: eddy current, magnetic flux leakage, and microwave testing; 4)
visual and optical methods: interferometry, holography, and dye penetrants; and 5)
thermal methods: infrared radiation and thermal paints [4]. Two or more NDT methods
may often be suitable for NDE; but, this does not necessarily imply they may be regarded
as alternative techniques. It often happens that one of the methods is used to complement
another or perhaps to verify the findings of the other [4].
Microwave nondestructive testing (MNDT) methods have advantages over other
NDT methods (such as radiography, ultrasonics, and eddy current) in terms of low cost,
contactless feature of the microwave sensor (antenna), good resolution and superior
penetration in nonmetallic materials [5]. They are known to be excellent candidates for
aquametry (detection of moisture content or bound– and free–water) in environments in
which moisture content determination is sought [6],[7]. In addition, MNDT techniques
are insensitive to environmental conditions, such as water vapor and dust (compared to
infrared sensors) and high temperatures, safe (compared to radioactive radiation), and
fast. However, microwave sensors are expensive (the costs have generally fallen with
time), must be adapted to a specific application, sensitive to more than one variable, and
must be calibrated separately for different materials [8].
Microwave methods for materials characterization generally fall into resonant or
nonresonant methods [2]. Resonant techniques have much higher accuracies and
sensitivities than those of nonresonant techniques. A shift in the resonant frequency and a
2
decrease in quality factor after insertion of a material are, respectively, utilized for the
determination of the real part (dielectric constant) and the imaginary part (loss factor) of
the complex permittivity. However, materials resulting in a large frequency shift from the
resonance frequency or a considerable decrease in the quality factor are not accurately
measurable by resonator methods which use a perturbation technique [9]. As a result,
these methods are generally utilized for electrical characterization of medium– and low–
loss materials. In a recent study, it was shown that resonant techniques are also applicable
to high–loss materials provided that very small samples are prepared or higher volume
cavities are constructed [10]. Nonetheless, resonant methods require meticulous sample
preparation before measurement. In addition, for an analysis over a broad frequency
band, new measurement set–ups (or cavities) must be made, which are not feasible from a
practical point of view. Tunable resonators can be used for a wider frequency band
analysis; but, they are expensive and an increase in the frequency bandwidth
accompanies a decrease in the accuracy. Examples for some resonant instruments are
open (or Fabry–Perot) resonator [11]–[13], cavity resonators [14], whispering gallery
modes [15]–[17], etc.
Nonresonant methods have relatively higher accuracy over a broad frequency
band and necessitate less sample preparation compared to resonant methods.
Additionally, they allow frequency– or time–domain analysis, or both. These features
make these methods very attractive for industrial–based applications. In these methods,
electrical properties of materials are usually derived from amplitude and/or phase
measurements of reflection and/or transmission coefficients [2]. Reflection measurements
require that the reference plane be well defined, and the higher the frequency the more
3
they are affected by the surface characteristics. However, they are convenient in some
instances because the sensor can be placed on one side of the material (for instance,
measurements on walls and pavement, back–filled concrete wall, roadway, etc).
Additionally, transmission measurements have the advantage of providing more
information on the whole volume, because the wave propagates through all the material
in its path; however, they require two sensors for their measurements.
Because nonresonant methods are applicable and feasible for industrial–based
applications, we will restrict our study to focus solely on nonresonant methods. Outcomes
of this dissertation are fourfold. Firstly, we will propose two methods for characterization
of the electrical properties of lossy materials. In the literature, methods used for materials
characterization generally employ complex scattering (S–) parameter measurements. To
measure these parameters, one needs to utilize a vector network analyzer (VNA).
However, these instruments are expensive, complex in operation, and require expertise.
Therefore, there is a need in industrial–based applications for characterization of
materials by simple and relatively inexpensive measurement set–ups. Scalar network
analyzers (SNAs), for example, can be used as a relatively inexpensive solution for
materials property measurements. However, because they are limited to measurements of
the amplitudes of S–parameters, it is not obvious how to use amplitude–only
measurements to fully characterize materials. To fill this gap, we developed two
amplitude–only methods. The first method extracts complex permittivity of lossy
materials at one fixed frequency, and the second determines both the complex
permittivity and thickness of lossy materials using multiple frequencies.
4
Second, we will present a design of a highly sensitive and accurate waveguide
holder for characterization of liquid and granular materials. It is well known that
measurements of electrical properties of solid materials are easier than those of liquid and
granular materials provided that the solid materials are not very thin. This is because
granular and liquid materials produce a meniscus, which requires a different approach
than those used for solid materials. Another problem that an experimentalist encounters
while performing the measurements on granular and liquid materials is that in most
instances he cannot explicitly evaluate the thickness of these materials. The final
problem, which is related to the second one, is that one needs to implement the
transformation from the calibration plane (the measurement point the system is
calibrated) to the measurement plane (the measurement point where the sample is
located). In the literature, there is no unique technique which eliminates all these
drawbacks. In this dissertation, we propose a method which determines the thickness of
materials, calculates the calibration plane factors (used for transforming the calibration
plane to measurement plane), and characterizes the complex permittivity and complex
permeability of solid (thin), liquid, and granular materials.
Third, we will propose a method to measure and monitor the electrical properties
of cement–based materials (cement paste, mortar, concrete, etc.) over their long service
times using amplitude–only measurements. These materials are the most widely used
construction materials in the world. Knowledge of mechanical, mixture proportion, and
chemical properties (compressive and tensile strengths, water–to–cement ratio (w/c),
sand–to–cement ratio (s/c), coarse aggregate–to–cement ratio (ca/c), porosity, degree of
hydration, etc.) of such materials is important for
5
evaluation of their quality and
integrity. These mechanical and physical properties are correlated with the electrical
properties of materials. Therefore, the characterization of electrical properties of these
materials can be utilized to improve the physical and mechanical properties and to detect
any integrity problems such as air bubbles, cracks, etc.
The general methods used in civil engineering for characterizing the mechanical
properties of cement–based materials are either to prepare a sample with the same mix
proportions in the lab, or to extract a sample from the structure, and then test in the lab.
The sample preparation method does not reflect the actual measurement of interest since
the sample in the lab and the materials cast in the field are not cured in the same
environment [18]. The sample removal method is clearly harmful. As a result, the test
techniques used for both methods are destructive [19] because the sample is damaged and
cannot be used in future. In the literature, characterization of these materials has been
done extensively using reflection measurements. However, these measurements cannot
fully characterize these materials. It was shown that transmission properties are also
important for structural inspection of these materials [20]. For these reasons, in this
dissertation we present reflection and transmission measurements for inspection and full
characterization of these materials.
Finally, we will develop a simple microwave sensor which is constructed from
relatively inexpensive and readily available components for characterization of materials
for industrial–based applications. The sensor uses free–space (free–wave) measurements,
which are very suitable for noncontact, nondestructive, noninvasive, and in situ
characterization of materials. To reduce the cost of the system, we used a microcontroller
to acquire the measurement data, process these data, and calculate the complex
6
permittivity of materials. In addition, because free–space measurements suffer from
diffraction from sample edges, we also developed a calibration procedure for highly
accurate measurements by the set–up.
1.2 Overview of the Chapters
This dissertation is organized as follows: In chapter 2, we will present a review for each
of the four aforementioned research studies in this chapter. Then, in chapter 3, we will
derive forward and reverse reflection and transmission S–parameters for a five–layer
structure using a conventional multiple–reflections model. The derivations are
sufficiently general that they can be used in subsequent chapters. Furthermore, we derive
a well–defined objective function which depends on only one unknown. This is
significant since fast and accurate computations could be achieved by this function. We
introduced two amplitude–only methods (methods based on amplitudes of S–parameter
measurements) for complex permittivity determination of lossy materials. While the first
method extracts the complex permittivity of lossy materials sandwiched between two
low–loss dielectric plugs at one frequency, the second uses the complex permittivity of
lossy materials without any prior physical thickness measurement using amplitude–only
measurements at different frequencies. Both of these methods are validated by a general
purpose rectangular waveguide structure using amplitude–only measurements after a
simple calibration technique, which only corrects the frequency errors, and a full two–
port calibration technique are applied. We verified that the extracted complex permittivity
from amplitude–only measurements calibrated by a simple calibration technique is in
very good agreement with that achieved by the full two–port calibration. Therefore, the
7
proposed methods allow the measurements to be carried out by scalar network analyzers
which are relatively inexpensive and simple in structure compared to vector network
analyzers; and they are suitable candidates for industrial–based applications.
In chapter 4, a waveguide sample holder for automated thickness–independent
complex permittivity and complex permeability determination of low–, medium–, and
high–loss materials is introduced. There are three main advantages of the holder as a) it
eliminates the dependency of calibration plane on measurements; b) it takes into account
the effect of low–loss plugs, which are utilized to hold the liquid or granular samples in
place; and c) it determines the complex permittivity and complex permeability of
materials without any previous physical thickness measurement of samples. In addition,
the holder is very important for measurements of solid thin materials since these
materials can result in sagging over the plane that measurements are carried out, and
accurately determines their complex permittivity and complex permeability by avoiding
the sagging problem using low–loss dielectric plugs. For validation of the sample holder,
we conducted measurements of the methanol, distilled water, and a commercially
available antifreeze solution. It is observed that the extracted complex permittivities of
these samples are in good agreement with those obtained from a theoretical model
(Debye model) and with those determined from a robust technique reported in the
literature.
In chapter 5, we demonstrate the free–space microwave measurements of cement–
based materials (concrete, mortar, etc.) using a relatively inexpensive and simple set–up.
Because these measurements are noncontact, nondestructive, and noninvasive, the set–up
can be used for in situ measurements. We carried out two types of measurements as: a)
8
measurements of reflection and transmission properties of young mortar specimens with
different water–to–cement ratios over 20–54 hours after sample preparation, and b)
measurements of reflection and transmission properties of hardened mortar and concrete
specimens with different water–to–cement ratios over 3–36 months after sample
preparation. We measured these properties using relative measurements, and analyzed the
possible measurement uncertainties.
In chapter 6, we present a microcontroller–based microwave free–space sensor for
electrical characterization of lossy materials. We have tried to keep the overall cost of the
sensor minimum such that the sensor can easily be adapted for industrial–based
applications. We also developed a calibration technique, which is similar to the well–
known Thru–Reflect–Line technique, for the sensor to decrease and eliminate undesired
reflection from environments, the losses in the measurement set–up, and diffractions
from edges of the sample. This is imperative since free–space measurements are prone to
these unwanted reflections and diffractions. We refined the domain for computations by
using the objective function derived in chapter 3, and coded three different programs for
complex permittivity determination of lossy materials. We validated the programs using
the measurement data taken from the literature by comparing their results with those of
the techniques in the literature. For validation of the sensor, we carried out reflection and
transmission measurements of a lossy sample (fresh cement paste) immersed into a
wooded container and calculated its complex permittivity.
9
CHAPTER 2
REVIEW OF PREVIOUS STUDIES
Because the dissertation can be divided into four main sub–fields, in this chapter, we will
present a review of previous studies for each sub–field.
2.1 Novel Methods for Lossy Materials Characterization
Due to its relative simplicity, the transmission–reflection (TR) technique is a widely used
measurement technique [21] for broadband measurements of materials. The well–known
Nicolson–Ross–Weir (NRW) method [22], [23] is a type of TR technique, which
determines the complex permittivity, ε , and permeability, µ , of materials noniteratively.
Because initially the NRW method was considered as a three–layer structure, Williams et
al. modified it to suit for simultaneous ε and µ determination of lossy thin materials
backed by a low–loss material [24] and Bois et al. adapted it for simultaneous ε and µ
determination of granular and liquid materials plugged (or sandwiched) by two low–loss
materials [25].
The complex scattering (S–) parameters are generally measured to achieve this
goal by TR techniques [26]. However, there are three main problems associated with
these measurements. First, they often yield multiple solutions. The general approach to
determine one solution was to choose a thickness for the material less than one–quarter
wavelength [27]. However, thin materials not only result in sagging, which alters the
10
theoretical formulations [28], but also decrease the feasibility and repeatability of
measurements. Though other test materials can be attached or adhered to the material
under test, the inhomogeneity in these materials can notably reduce the accuracy of ε
and µ estimation. To circumvent the problems arising from using thin materials for
extracting one solution, Weir et al. proposed a technique which compares calculated and
measured group delays and determines an initial guess for ε and µ after this comparison
[21],[23]. The second problem with complex S–parameter measurements is that TR
techniques drastically suffer from the phase uncertainty of reflection measurements of
low–loss materials at frequencies which correspond to multiple–half wavelengths inside
these materials [21],[29],[30]. At these frequencies, amplitudes of reflection
measurements decrease significantly, and measured phases of reflection measurements
diverge from their actual values. This results in a sharp ripple in the extracted ε and µ
over frequencies. To overcome this problem, Baker–Jarvis et al. used an iterative
numerical technique [30], which removes the ripples at those frequencies. This technique
uses a parameter whose value depends on the length and electrical properties of the
material and can be set zero or infinity. Although the technique is effective, it yields
unambiguously extractions of ε only in low–loss materials since that parameter is set to
zero. This eliminates the reflection S–parameter measurements, and eventually forces one
to use transmission S–parameter measurements in the iteration process. In addition, the
technique requires a good initial guess for ε before the iteration. The authors proposed
that this initial guess can be determined from the solution of the NRW method at one
frequency. However, the extraction of ε for the entire frequency band may result in
erroneous results if the initial guess is evaluated at a frequency which corresponds to
11
minimum measured reflection S–parameters. More importantly, this technique does not
solve the multiple–solutions problem in ε determination since it uses, as an initial guess,
the extracted ε at one frequency from the NRW method, which inherently has multiple
solutions for ε and µ .
Later, Boughriet et al. proposed a noniterative and stable effective parameter
(NSEP) method to eliminate the ripples in the measured ε and µ [31]. The method is
superior to the iterative technique [30] in that it explicitly determines ε and µ as does
the NRW method. Even if the method effectively suppressed the ripples for ε
measurements by removing the ill–behaved term, 1 − Γ 1 + Γ , from the ε equation, it is
not suitable for broadband and simultaneous ε and µ measurements since that term still
appears in the equation of µ . Furthermore, the method also extracts multiple solutions
for ε and µ as do the NRW and iterative technique, and, in this sense, requires a prior
knowledge about or initial guess for the electrical properties of material under test.
Finally, for low– or high–loss materials, the total uncertainty in ε determined by the
NSEP method is greater than that by the iterative technique [30].
Another drawback in using complex S–parameter measurements is that measured
ε and µ are adversely affected from a small change in phase measurements that arise
from any shift in the calibration plane (the measurement point where the calibration is
performed) of materials. Finally, accurate complex S–parameter measurements require
using expensive VNAs, which makes microwave techniques unsuitable for industrial–
based applications [26],[32]. To eliminate all the aforementioned drawbacks of complex
S–parameter measurements, amplitude–only S–parameter measurements can be
employed. In this dissertation, we develop two methods which use amplitude–only S–
12
parameter measurements for ε determination of materials and also eliminate the well–
known multiple–solutions problem. The first method not only considers a three layer
structure (the material under test is located inside its holder) but also takes into account of
four– and five–layer structures which are needed for completeness of the analysis and for
application purposes (the liquid materials placed between two low–loss dielectric plugs).
We derived an equation, which depends solely on only one unknown quantity, for lossy
materials. This is very important for both practical reasons (fast and accurate ε
determination) and numerical stability (one solution for ε ). The second method uses
amplitude–only measurements at different frequencies for one–valued (correct and
unique) ε determination of lossy materials. This method is different than the first method
in that it also determines the thickness of these materials, which removes the necessity of
its measurement or priori knowledge. This is very important and provides immunity to
methods for industrial–based applications where thickness measurements are dynamically
difficult to achieve. For example, materials continuously moving on a conveyor belt.
2.2 A New Sample Holder for Low-Loss and Lossy Materials Characterization
When applying the TR techniques for measurements of granular and liquid materials,
some form of approximation in the physical nature of the measurement apparatus and/or
forward and inverse problem formulations are made [25], [33] or formulations require
that the dielectric specimen be low–loss [34]–[36]. To avoid any meniscus formation on
top of liquid samples, generally the measurement set–up is positioned vertically [37].
However, this positioning will not totally eliminate the meniscus. For the case of granular
specimens, planar measurement planes may not be easily achieved [25] and this can
13
degrade the measurement performance if the formulae characterizing the problem, which
in some instances cannot satisfy the requirement of the plane distortion as a result of its
unpredictable nature, are not provided. Another problem for measurements of these
materials which are retained between two low–loss dielectric plugs (or windows) is the
lack of appropriate calibration which takes into account the materials thickness. It is well
known that for high–loss liquid or granular materials, the amplitude level of transmission
measurements decreases considerably, and, in some instances, this circumstance
drastically affects the measurement accuracy if its level is comparable with the noise
floor or threshold of VNAs. Calibrating a sample fixture containing two low–loss
dielectric plugs where these plugs are separated from each other by a distance in which
the liquid or granular material will be retained may degrade the measurement accuracy if
the liquid or granular material attenuates the electromagnetic waves, while passing
through it, more than VNAs can accurately measure.
Coaxial sample holders can be utilized for broadband measurements [38];
however, they suffer from bad contacts between the sample and coaxial inner and outer
conductors [39], [40]. On the other hand, if a coaxial line sample holder is used and
dimensioned so that only single mode operation is possible, the dimensions become small
and meniscus and other dimensional uncertainties may adversely affect the measurement
[41]. Because waveguide measurements are highly sensitive, sample preparation for these
measurements is relatively easy, and mechanical connections of waveguide sections are
simple. Waveguide sample holders can be employed for precise dielectric
characterization of liquid samples [25], [33],[41].
14
Somlo has proposed a self–checking method for extracting ε and µ of liquid or
granular materials in waveguide measurements [41]. This method auto–monitors the
accuracy of measurements and also removes the calibration problem for these materials
by employing only one low–loss dielectric plug for holding the sample. However, the
measurement accuracy deteriorates because of the meniscus formation on top of liquid
materials where surfaces are in direct contact directly with the air section inside the
waveguide. To circumvent calibration plane unevenness, Bois et al. used a waveguide
sample holder for granular and liquid materials which are sandwiched between two
dielectric plugs [25]. This method not only solves the calibration problem by considering
the effects of the plugs on theoretical analysis but also eliminates the calibration plane
irregularities by employing two movable low–loss plugs. However, it may experience
liquid leakage passing through the movable plugs while compacting the sample, which
eventually can decrease the measurement repeatability. In addition, the method requires
precise knowledge of the measurement plane and the sample length. In practical
applications, a feasible method needs to be developed to eliminate the requirement for
precise knowledge of the thicknesses of samples. In this dissertation, we propose a
different waveguide sample holder which combines the advantages of these
aforementioned methods. It both eliminates the measurement plane dependency from
measurements and includes the effect of plugs on theoretical formulations. Other
advantages of the method are that the thickness of the sample does not need to be known
either precisely or approximately and that the holder could be horizontally or vertically
positioned since any meniscus that could form on top of the liquids is eliminated by
compacting samples using the plugs.
15
2.3 Electrical Characterization of Cement-Based Materials
Cement–based materials (cement paste, mortar, concrete, etc.) are the most widely used
construction materials in the world. Knowledge of mechanical, mixture proportion, and
chemical properties (compressive and tensile strengths, water–to–cement ratio (w/c),
sand–to–cement ratio (s/c), coarse aggregate–to–cement ratio (ca/c), porosity, degree of
hydration, etc.) of such materials is important for
evaluation of their quality and
integrity.
The general methods used in civil engineering for characterizing the mechanical
properties of cement–based materials are either to prepare a sample with the same mix
proportions in the lab, or to extract a sample from the structure, and then test in the lab.
The sample preparation method does not reflect the actual measure of interest since the
sample in the lab and the materials cast in the field are not cured in the same conditions
[42]. The sample removal method is clearly harmful. The test techniques used for both
methods are destructive [19] because the sample is damaged and cannot be used in future.
Therefore, several NDE techniques have been proposed and applied for quality
assessment, mixture content evaluation, and monitoring the internal integrity (voids,
cracks, degradation, etc.) of these materials [43]–[45]. Among the techniques, microwave
NDE (MNDE) techniques are not hazardous and low in cost compared to radioactive
methods, scatter little compared to acoustic waves, and have a good spatial resolution and
superior penetration in nonmetallic materials [5]. The spatial resolution of MNDE
techniques, varying roughly from 3 mm to 100 mm, depends upon the wavelength (or
frequency) and whether a near– or far–field approach is applied.
16
MNDE techniques used for cement–based materials’ investigation can be divided
into two groups: near–field and far–field. Assuming that microwave signal unboundedly
radiates into free–space, near–field techniques will have a higher resolution and accuracy
than far–field techniques since the footstep of the microwave signal for former techniques
is much smaller than that of latter ones [42], [46]–[54].
Dr. Zoughi and his colleagues at the Missouri University of Science and
Technology (formerly University of Missouri-Rolla) applied the near–field open–ended
waveguide contact with the sample under test (SUT) for cement–based materials’
investigation and demonstrated its potential for cure–state monitoring, mixture content
estimation, and early age strength prediction [42], [47]–[49]. However, contacting
techniques require close proximity probing with the SUT which limits their applicability
and efficacy since, if improper sensors are employed, undesired higher–order modes,
which drastically affect the measurements, can appear as a result of any minuscule air
gap between the sensor and the SUT [55]. Although the SUT can be placed at a distance
(the stand–off distance) from the antenna, open–ended waveguide measurements can be
drastically affected by any minute change (a few millimeters) of this distance during
measurements [50]–[52]. This can be a problem for investigation of real construction
structures since they have some surface roughness associated with them [56]. To
eliminate the adverse effects of this distance on measurements, a method which uses
orthogonally dual–polarized microwave signals can be employed [57]. However, this
method is applicable to only anisotropic materials such as carbon fiber reinforced
polymers. Next, it is known that electrical properties of the SUT in the near–field of an
antenna are known to influence the antenna beam [58]. Because cement–based materials
17
contain water and some varying mixture contents (coarse and fine aggregates, etc.), their
electrical properties can notably alter the antenna beam pattern at proximity. Finally, the
numerical modeling of wave–material interaction for near–field techniques is relatively
complex [59],[60].
Far–field MNDE techniques have also shown a great potential for the
investigation of cement–based materials [59]–[65]. These techniques, unlike near–field
techniques, are contact–free methods that deal only with the dominant excitation mode
and also allow plane wave assumption for modeling. Although their accuracy and
sensitivity is lower than near–field techniques, they enable large area coverage and are
very suitable for remote sensing [61].
Knowledge of electrical (dielectric, conductivity, or complex permittivity, etc.)
properties of cement–based materials is also important for propagation–related research;
for
example,
microwave
propagation
modeling
to
develop
indoor
wireless
communication systems [66]–[70]. This is because reflection and transmission
characteristics of buildings, walls, etc. are governed by these electrical properties [20],
[64]. In addition, for practical purposes, it is attractive to use the simplest measurement
system because using expensive microwave instruments, such as VNA, are not
convenient for industrial–based applications [32].
Recent investigations [42], [47] by Dr. Zoughi and his colleagues have
demonstrated the capability and potential of microwaves to detect the state and degree of
chemical reaction (hydration) in cement–based materials. In addition, a strong correlation
between the magnitude of reflection coefficient of microwave signals and the water–to–
cement ratio (w/c) of these materials, which eventually can be utilized for strength
18
estimation [54],[71], was shown. Although the results are promising, reflection–only
measurements do not disclose any new information about these materials after
approximately the 28th day of curing since reflection properties stay constant. Dr.
Kharkovksy, Dr. Sato, and their colleagues measured reflection and transmission
coefficients of these materials to gain more knowledge about their electrical properties.
Transmission coefficients are less dependent on surface properties of cement–based
materials and can evaluate the overall volume properties of these materials since the
wave passes through the whole volume. Periodic assessment of aging in civil structures a
over long period of time is important for structural integrity and public safety [59]. As a
result, transmission coefficients are also important for structural integrity of these
materials. In the literature, although transmission coefficients of these materials are
measured, minimal attention was given to what the effects of different surface properties
of these materials would be on transmission measurements and whether monitoring the
transmission measurements of these materials over long service times would disclose
new insight.
In this dissertation, we present the measurement results of free–space reflection
and transmission coefficients of some cement–based materials with varying mixture
contents over their long service lives. We measured early age reflection and transmission
properties of mortar specimens with different w/c ratios and introduced a new approach
(relative slope approach for reflection properties) to predict the history of the hydration
process of these specimens. It is shown that the difference between the relative slopes of
reflection properties of mortar samples with 0.40 and 0.45 w/c ratios is greater than that
of other mortar samples whose w/c ratio difference is 0.05. This result shows that a
19
mortar with a lower w/c ratio will gain hydration (and strength) faster than the one with a
higher w/c ratio (a non–linear relationship between the degree of hydration and w/c ratios
at early ages of curing). We demonstrated that monitoring early age transmission
properties can be very useful for the quality enhancement of these samples.
We also show free–space microwave transmission (and reflection) properties
measured from diverse sides of hardened mortar and concrete specimens with different
w/c ratios and the same sand–to–cement ratios during a 3–36 months period after
specimens’ preparation. The relative difference approach initially developed for
reflection measurements is applied to predict the state and degree of aging (curing) of
specimens using transmission properties obtained from various sides of specimens. It is
observed that while reflection properties of specimens are constant, their transmission
properties and aging changed, except for last few months, during 3–36 months. After
approximately 30 months, aging of all specimens became constant. It is shown that, for
measurements acquired from the same side of specimens, the mortar specimen with a
higher w / c ratio will lose free–water faster than the specimen with a lower w/c over
approximately 3–9 months (first stage) of curing. On the other hand, the mortar specimen
with a lower w/c ratio will drop free–water sooner than the mortar specimen with higher
w/c during approximately 9–30 months (second stage) of curing. Moreover, it is
illustrated that concrete specimens will complete the aging sooner than mortar specimens
with the same w/c ratio as a result of heavy aggregates. In addition, among the specimens,
while the concrete specimen with a higher w/c ratio will exhibit quicker aging during
approximately 3–6.1 months, the mortar specimen with a lower w/c ratio will display
faster aging during nearly 6.1–30 months.
20
2.4 A Simple Microwave Sensor for Electrical Characterization of Materials in
Industrial-Based Applications
Lossy dielectric materials contain high content of free–water. Microwaves are known as
excellent candidates for aquametry (detection of moisture content or bound– and free–
water) measurements [72], [73]. Among microwave techniques, nonresonant methods are
capable of characterizing low–loss and lossy materials, allow wide frequency coverage,
and are feasible for frequency– and time–domain analyses. These features make them
very attractive for industrial–based applications.
Among nonresonant methods, free–space methods have been used for
characterization of a wide variety of materials because they are nondestructive,
noncontact, noninvasive, require little, if any, sample preparation [73], and can be
effectively used at high temperatures or in chemically poisonous regions [2], [27], [28].
An amplitude–only microwave free–space technique was analyzed for dynamic ε
determination of materials immersed into a container [26]. Using a graphical method, it
was shown that if the material under test demonstrates at least 10 dB attenuation, its ε
can uniquely be determined by using amplitudes of reflection and transmission
coefficients. However, the graphical approach is unsuitable for industrial–based
applications and dynamic characterization of materials since it takes a long time to
determine ε [74] and requires a two–dimensional numerical technique [75]. An
optimization technique will also be helpful for this purpose. However, these techniques
are often time consuming due to slow convergence and the existence of spurious
solutions [76]. They also necessitate a well–defined domain for solutions because any
small deviation from a local optimum point can result in an erroneous solution [75].
21
In this dissertation, we will present a microcontroller–based microwave free–
space measurement system for ε determination of lossy materials since there is a need
for a simple and relatively inexpensive measurement method in industrial–based
applications. We will utilize the derived objective function (the function which is utilized
for ε determination) in chapter 3, which depends on only one unknown quantity, for ε
determination [20],[74],[77]–[79]. This function is very important since the
implementation of ε determination in microcontrollers necessitate a well–defined
domain and a simple objective function for computations. Since free–space
measurements are very prone to undesired effects of diffractions from sample edges, we
also show a useful calibration technique we recently developed to minimize these effects.
We describe the main properties of the measurement system and validate it using free–
space measurements.
22
CHAPTER 3
TWO MICROWAVE METHODS FOR COMPLEX PERMITTIVITY
DETERMINAITON OF LOSSY MATERIALS USING AMPLITUDE–ONLY S–
PARAMETER MEASUREMENTS
The complex scattering (S–) parameters are generally measured by nonresonant methods
to determine the relative complex permittivity, ε , and the relative complex permeability,
µ . However, there are three main problems associated with these measurements. First,
they often yield multiple solutions [23]. Second, the phase uncertainty of reflection S–
parameter measurements greatly increases when the thickness of materials is of integer
multiples of one–half guided wavelength [21],[29]. Finally, they are adversely affected
by any small shift from the calibration plane of materials which results in enormous
phase errors. To eliminate all these drawbacks simultaneously, amplitude–only S–
parameter measurements can be employed. In addition, the use of expensive vector
network analyzers (VNAs) for complex S–parameter measurements makes microwave
techniques inconvenient for industrial–based applications. Therefore, there is a need for a
simple and relatively inexpensive microwave method for dynamic measurements.
In this chapter, we will present two amplitude–only methods (methods based on
amplitudes of S–parameter measurements). In the first method, we will derive an
objective function which depends on only one real quantity for determination of the
dielectric
23
constants of materials. The objective function is quite general and allows the use of a
composite structure with at most five–layers. Second, we will present an amplitude–only
method for thickness–independent complex permittivity determination of lossy materials.
3.1
Theoretical Analysis
In this section, the problem at hand is first kept general so that the derivations in this
section could be used for the following chapters, and then it is restricted to dielectric
(nonmagnetic) materials with µ = 1 . The geometry of the problem is depicted in Fig. 3.1
in which a sample whose dielectric properties are unknown is sandwiched between two
dielectric plugs with different lengths and electrical properties. In the following
mathematical analysis, a harmonic time variation of the time ( e jωt ) is assumed. We also
consider that the waveguide walls behave like perfect conductors and the waveguide is
operating in its fundamental TE10 (transverse electric) mode.
For the following derivations, we assume that lateral dimensions of media 2, 3,
and 4 in Fig. 3.1 are flat and there is no air gap between external surfaces of media 2, 3,
and 4 and waveguide inner walls. If any of these assumptions is not satisfied, the electric
field vector which is transverse to the direction of electromagnetic propagation in an
empty waveguide section will deflect and eventually this produces TMmn (transverse
magnetic) waveguide propagation modes inside media 2, 3, and 4 [80],[81]. Here, m and
n are used to denote different modes of the TM waveguide propagation where
m = 1,2,3,... and n = 1, 2,3,... In addition, it is assumed that all media in Fig. 3.1 are
isotropic, homogenous, and reciprocal.
24
Fig. 3.1.
The scattering parameter measurements of a sample sandwiched between
two dielectric plugs.
For a homogenous medium, any solution for the time-harmonic electric and
magnetic fields must satisfy the following Maxwell’s equations [81]
∇ × E = − M − jωµ H , ∇ × H = J + jωε E ,
(3.1)
∇ i E = qev ε , ∇i H = qmv µ .
(3.2)
In (3.1) and (3.2), the superscript right arrow denotes the vector quantities; E and H
electric and magnetic field intensities; J and M are electric and magnetic current
densities; qev and qmv volume electric and magnetic charge densities; and ε and µ are
frequency-varying complex permittivity and permeability of the medium.
The forward and reverse S-parameters for the overall structure in Fig. 3.1 can be
obtained by using Maxwell’s equations in (3.1) and (3.2) and applying boundary
conditions for tangential electric and magnetic fields at different material boundaries in
Fig. 1 and at waveguide walls (tangential components of the electric field must vanish on
25
the walls of the waveguide). We can also electric and magnetic vector potentials, F and
A , for a representation of E and H to simplify the boundary solutions problem in Fig.
3.1. However, we prefer the network analysis to the problem in Fig. 3.1 than the field
analysis using E and H or F and A . This is because attempting to solve Maxwell’s
equations subject to boundary conditions for the entire structure in Fig. 3.1 is rather
complicated and takes considerable amount of time. In addition, Maxwell’s equations
result in expression for E and H (or F and A ) at all points in space, which is more
information than it is needed [14]. Instead, we apply microwave network analysis to our
problem.
There are basically two approaches to analyze this wave–material interaction
using microwave network analysis: transmission–line impedance and multiple
reflections. While the former is applicable to only a steady–state analysis, the latter one
can be applied to both transient– and steady–state analyses. In this dissertation, only the
steady–state analysis is considered since the time required the electromagnetic wave to
travel from the five–layer structure in Fig. 3.1 is so small. We will use the conventional
multi–reflections model and derive complex S–parameters for the structure in Fig. 3.1.
It should be stated that in a steady–state analysis, the transmission–line and
multiple–reflections approaches can be assumed to be equivalent and can equally be
applied to our problem [14]. However, to gain more knowledge about the wave–material
interaction, we will employ the multiple–reflections approach.
26
3.1.1
Complex Permittivity Extraction Using the Classical Multiple–Reflections
Model
The classical multiple–reflections model takes into account only the delays for multiple
reflected and transmitted signals in a medium [82]. To visualize how the complex S–
parameters are derived from this model for the five–layer structure in Fig. 3.1, a diagram,
which shows internal reflections and transmissions inside each medium, is illustrated in
Fig. 3.2.
Fig. 3.2.
The diagram for the analysis of the five–layer structure in Fig. 3.1 by
multiple reflections approach.
Using this diagram, the forward complex S–parameters ( S11 and S21 ) for the
structure in Fig. 3.1 are obtained as
S11 = rA + tA tB rBT32 + tA tB rB2 rC T34 + tA tB rB3rC2T36 + ...
(3.3)
S21 = tA tC T3 + tA tC rB rC T33 + tA tC rB2 rC2T35 + ...
(3.4)
where
27
(
r12 + r23T22
)
rA = r12 + t12 r23t21T22 1 + r23r21T22 + ... =
(
)
(
)
rB = r34 + t34 r45t43T42 1 + r45 r43T42 + ... =
,
(3.5)
,
(3.6)
1 + r12 r23T22
r34 + r45T42
1 + r34 r45T42
rC = r32 + t32 r21t23T22 1 + r21r23T22 + ... = −
r23 + r12T22
1 + r12 r23T22
(1 + r12 ) (1 + r23 ) T2
,
(3.7)
(
)
,
(3.8)
(
) (1 − r12 ) (1 − r232 )T2 ,
(3.9)
(
) (1 + r34 )(1 + r452 )T4 ,
(3.10)
tA = t12t23T2 1 + r23r21T22 + ... =
tB = t32t21T2 1 + r21r23T22 + ... =
tC = t34t45T4 1 + r45r43T42 + ... =
1 + r12 r23T22
1 + r12 r23T2
1 + r34 r45T4
and
Z − Zn
, n = 1,2,3,4; m = n + 1;
rnm = m
Zm + Zn
Zr =
jωµr
γr
, γr = j
2π
λ0
Tk = exp ( −γ k Lk ) , k = 2,3,4.
(3.11)
2
λ 
ε r µr −  0  ,
 λc 
r = 1,2,3,4,5.
(3.12)
Here, rnm and Tk are, respectively, the first reflection coefficients at media boundaries
and the first transmission coefficients inside the media; Z m , γ k , ε r , µr , and Lk
represent, correspondingly, the impedances, propagation constants, relative complex
permittivities, relative complex permeabilities, and lengths of media; and λ0 and λc are
the free–space and cut–off wavelengths. For free–space and/or coaxial–line
measurements, λc → ∞ .
28
Finally, by substituting (3.5)–(3.10) into (3.3) and (3.4), S11 and S21 for the five–
layer structure in Fig. 3.1 can be written in a compact form as
r12 + r23T22 )(1 + r34 r45T42 ) + ( r12 r23 + T22 )( r34 + r45T42 ) T32
(
,
S11 =
(1 + r34r45T42 )(1 + r12r23T22 ) + ( r34 + r45T42 )( r23 + r12T22 )T32
S21 =
3.1.2
(1 + r12 ) (1 + r23 )(1 + r34 )(1 + r45 )T2T3T4
(1 + r34r45T42 )(1 + r12r23T22 ) + ( r34 + r45T42 )( r23 + r12T22 )T32
.
(3.13)
(3.14)
Validation of the Derived Forward Scattering Parameters and Derivation of
Reverse Scattering Parameters
For validation of the derivations in (3.13) and (3.14), we consider two special
cases. In the first case, we assume that the lengths and electrical properties of medium 2
and 4 are the same and media 1 and 5 are air. In these circumstances, r45 = − r12 ,
r34 = − r23 , and T4 = T2 . Then, (3.13) and (3.14) reduce to
(1)
S11
r12 + r23T22 )(1 + r12 r23T22 ) − ( r12 r23 + T22 )( r23 + r12T22 ) T32
(
=
,
2 2
2 2 2
(1 + r12r23T2 ) − ( r23 + r12T2 ) T3
(3.15)
2
2
1 − r12
1 − r23
T22T3
(
)(
)
=
,
2 2
2 2 2
(1 + r12r23T2 ) − ( r23 + r12T2 ) T3
(3.16)
(1)
S21
where the number ‘1’ in parenthesis at the superscript of S11 and S21 in (3.15) and (3.16)
denotes the first special case. The complex S–parameters given in (3.15) and (3.16) are
the same of those in [25],[26].
29
In the second case, we assume that media 1, 2, 4, and 5 are air and the lengths of
media 2 and 4 are zero. Then, we can write r12 = 0 , r45 = 0 , r34 = − r23 , and T2 = T4 = 1 .
In this condition, complex S–parameters in (3.13) and (3.14) will reduce to
),
(3.17)
2
1 − r23
(
)T3 ,
=
(3.18)
(2)
S11
=
(2)
S 21
(
r23 1 − T32
2 2
1 − r23
T3
2 2
T3
1 − r23
where the number ‘2’ in parenthesis at the superscript of S11 and S21 in (3.17) and (3.18)
(2)
(2)
and S 21
are identical to those
signifies the second special case. It is obvious that S11
well–known S–parameters used in the literature [21]–[23],[29],[30].
In the manner we derived the expressions for the forward complex S–parameters,
we also derived the reverse complex S–parameters for the five–layer structure in Fig. 3.1
by systematically exchanging the individual reflection and transmission coefficients in
the media. To demonstrate this, we first exchanged the individual reflection coefficients
in the media as
r12 → r54 , r23 → r43 , r34 → r32 , r45 → r21.
(3.19)
Then, we replaced the individual transmission coefficients in media 2, 3, and 4 as
T2 → T4 , T4 → T2 .
(3.20)
Substituting the replacements in (3.19) and (3.20) into (3.13) and (3.14), we
derived the reverse complex S–parameters as
r54 + r43T42 )(1 + r32 r21T22 ) + ( r54 r43 + T42 )( r32 + r21T22 ) T32
(
S22 =
,
(1 + r32r21T22 )(1 + r54r43T42 ) + ( r32 + r21T22 )( r43 + r54T42 )T32
30
(3.21)
S12 =
(1 + r54 )(1 + r43 )(1 + r32 ) (1 + r21 ) T2T3T4
(1 + r32r21T22 )(1 + r54r43T42 ) + ( r32 + r21T22 )( r43 + r54T42 )T32
.
(3.22)
Finally, using r54 = − r45 , r43 = − r34 , r32 = − r23 , and r21 = −r12 , the expressions for
S 22 and S12 in (3.21) and (3.22) reduce to
r45 + r34T42 )(1 + r23r12T22 ) + ( r45r34 + T42 )( r23 + r12T22 ) T32
(
S22 = −
,
2
2
2
2
2
+
+
+
+
+
r
r
T
r
r
T
r
r
T
r
r
T
T
1
1
( 23 12 2 )( 45 34 4 ) ( 23 12 2 )( 34 45 4 ) 3
S12 =
(
(1 − r45 )(1 − r34 )(1 − r23 ) (1 − r12 ) T2T3T4
)(
) (
)(
)
1 + r23r12T22 1 + r45 r34T42 + r23 + r12T22 r34 + r45T42 T32
.
(3.23)
(3.24)
We assumed at the beginning of the derivation of the S-parameter expressions that
all media are reciprocal. At first moment, it is seen that the final expressions for S21 and
S12 in (3.14) and (3.24) are different from each other. However, a close analysis
demonstrates that they are identical. To this end, we use (3.11) and (3.12) and obtain
(1 + r12 ) (1 + r23 )(1 + r34 )(1 + r45 ) = (1 − r45 )(1 − r34 )(1 − r23 ) (1 − r12 ) .
(3.25)
On the other hand, it is obvious that the final expressions for S11 and S22 in
(3.13) and (3.23) are different from each other. This is because the five–layer structure in
Fig. 3.1 is asymmetric. However, if the lengths and electrical properties of medium 2 and
4 and those of media 1 and 5 are the same, the structure becomes symmetric and the
expression for S22 in (3.23) becomes equal to that for S11 in (3.13).
3.2
An Amplitude-Only Method for Permittivity Determination of Lossy Materials
In this section, we will derive an objective function, which depends on only one unknown
quantity, for accurate, fast, and unique complex permittivity of lossy materials using
31
amplitudes of S–parameters. This function is very important since speed is one of the key
factors in industrial–based applications.
In the analysis, we will use the expressions for complex S–parameters in (3.13)
and (3.14). We assume that media 1 and 5 are air and that complex permittivities and
lengths of media 2 and 4 are known. In addition, it is assumed that all layers in Fig. 3.1
are nonmagnetic; i.e., µr = 1 . Our goal is to determine ε 3 for a known sample length and
a given frequency, f , using the amplitudes of S11 and S21 in (3.13) and (3.14).
3.2.1
Selection of a Suitable Approach for Permittivity Extraction
There are two main techniques for ε 3 extraction of the sample using (3.13) and (3.14): a)
the direct method and b) the de–embed method. In the direct method, the theoretical S–
parameters of the structure in Fig. 3.1 are directly compared with the measured S–
parameters. In the de–embed method, known layers (media 2 and 4 in Fig. 3.1) are
removed (i.e., de–embedded) before comparing theoretical and measured S–parameters
of the sample [84]. In applying the direct method, firstly an initial guess for ε 3 is
provided and then theoretical ABCD (A–) parameters of each layer (media 2, 3, and 4 in
Fig. 3.1) are computed. Next, these parameters are used to obtain S–parameters of the
structure. Finally, theoretical and measured S–parameters of the structure are compared.
The value of ε 3 is iterated by a suitable numerical technique until the desired accuracy is
reached. In applying the de–embed method, firstly A–parameters of known layers are
computed. Then, measured S–parameters of the structure are converted into A–
parameters. Next, the sample is isolated from the structure by using the A–matrices of the
32
known layers. Finally, ε 3 can be extracted by the Nicolson–Ross–Weir technique
[22],[23] or any other method using S–parameters of the isolated sample [84].
It was shown that while the direct method detects possible sample
inhomogeneities, the de–embed method cannot because an averaging occurs while
converting S–parameters to A–parameters (e.g., A22 = ( S21S12 − S11S22 ) S21 ) [84]. This
detection is important in industrial–based applications since samples are not always made
so perfectly (e.g., formation of air inside samples) [84]. It can also be used for measuring
the integral ε 3 of quasi–heterogeneous materials (e.g., concrete specimens composed of
cement powder, water, and coarse and fine aggregates – all varying in content and size).
Therefore, we adapted the direct method as a solution to our problem. Before applying
this method to our problem, it is important to analyze S–parameter expressions in (3.13)
and (3.14) for lossy materials.
3.2.2
Single-Pass Technique
The single–pass technique in frequency domain measurements can be applied under two
conditions: a) the sample in Fig. 3.1 has a thickness that is large enough to allow
elimination of the signals due to multiple reflections in it, or b) the sample is a high–loss
sample. If the sample in Fig. 3.1 possesses at least 10 dB attenuation, the first reflected
power at interface 2-3 will be approximately 10 times greater than that at interface 3-4;
that is
10dB ≈ 10log
(1)
P23
(1)
(1)
⇒ P23
≈ 10 P34
,
(1)
P34
(3.26)
(1)
(1)
where P23
and P34
are, respectively, the powers of first reflected signals at interfaces 2-
33
3 and 3-4.
In the same manner, since the first reflected signal at interface 3-4 consumes the
same amount of power to return to the interface 2-3, using (3.26), we can write
10dB ≈ 10log
(1)
P34
⇒
(2)
P23
(1)
(2)
P23
≈ 100 P23
,
(3.27)
(2)
where P23
is the power of the second reflected signal at interface 2-3. As a result, total
power of all reflected signals at interface 2-3 will be
(1)
(2)
(3)
P23 = P23
+ P23
+ P23
+ ...,
(3.28)
(v)
where P23
for v = 1, 2,3,... is the power of v th reflected signal at interface 2-3. It is
obvious from (3.28) that for applications in engineering, we can approximate (3.28) to
(1)
P23 ≅ P23
,
(3.29)
because the contribution of each reflected signal for v > 1 at interface 2-3 is equal to or
(1)
less than one percent of P23
. Therefore, multiple reflections inside the sample can be
neglected.
Assuming that the sample in Fig. 3.1 results in at least 10 dB attenuation [26] and
conductor losses are much smaller than dielectric losses [81], provided that the sample
length is small, S11 and S21 in (3.13) and (3.14) reduce to
s
S11
=
s
=
S21
r12 + r23T22
1 + r12 r23T22
(3.30)
,
(1 + r12 ) (1 + r23 )(1 + r34 )(1 + r45 ) T2T4T3
(
)(
1 + r34 r45T42 1 + r12 r23T22
34
)
.
(3.31)
where the letter ‘s’ at the superscript of S11 and S21 in (3.30) and (3.31) denotes
simplified expressions. In the simplification, we made use of T32 << 1 . It is clear from
(3.30) that for lossy samples there will be no effect of the medium 4 on the S11 , and S21
includes only the first transmission coefficient of the sample while multiple reflections
inside media 2 and 4 are present.
3.2.3
Derivation of a Simple Objective Function
The direct method requires a numerical technique for ε 3 determination. In industrial–
based applications, it is always attractive to evaluate ε 3 fast and accurately [26]. In
addition, it is appropriate to have a well–defined domain for numerical computations
since any small deviation from a local optimum point can result in incorrect solutions
[75]. Because ε 3 is unknown, it is difficult to predict a domain for its computation.
Nonetheless, we succeeded in assigning a highly restrictive domain for ε 3 computations
using the conformal transformation, as shown in Fig. 3.3, between ε 3 and r23 since
r23 ≤ 1 .
Fig. 3.3.
Conformal transformation between ε 3 and ( χ , ξ ) pair.
35
To elaborate on this, we defined a new set of variables as
r23 = χ − jξ , T2 = Λ1 − j Λ 2 , r12 = Λ3 − j Λ 4 ,
(3.32)
r34 = k − jl , T4 = Λ5 − j Λ 6 , r45 = Λ 7 − j Λ8 .
The definition of new variables changed the inverse problem from the determination of
ε 3 into the determination of χ and ξ (or k and l ) because we can determine ε 3 from
(3.11) for a known ε 2 (or ε 4 ). Using (3.32), the amplitudes of simplified S–parameters
will be
)(
(
) (
) (
12
)
 2

2 2
χ 2 + ξ 2 + Λ32 + Λ 24 + 2 Λ12 − Λ 22 ( Λ 3 χ + Λ 4ξ ) 
 Λ1 + Λ 2


+4Λ1Λ 2 ( Λ 4 χ − Λ 3ξ )
s


S11 =
2


2
2
2
2
2
2
2
2
1 + Λ1 + Λ 2 Λ3 + Λ 4 χ + ξ + 2 Λ1 − Λ 2 ( Λ 3 χ − Λ 4ξ ) 


 −4Λ1Λ 2 ( Λ 4 χ + Λ3ξ )

)(
)(
(
) (
)(
(
,
)
)
(3.33)
12
 (1 + χ )2 + ξ 2 (1 + k )2 + l 2 


s
= Λ9 
S21
H1 ( χ , ξ ) H 2 ( k , l )
e
Im{− jγ 3}L3
,
)(
)
(3.34)
where
(
Λ9 =  (1 + Λ3 ) + Λ 42

2
(
) ((1 + Λ )
)(
7
2
)(
)(
12
+ Λ82 Λ12 + Λ 22 Λ 52 + Λ 62 

) (
(3.35)
12
)


2
2 2
1
Λ
+
+
Λ
Λ 32 + Λ 42 χ 2 + ξ 2 + 2 Λ12 − Λ 22 ( χΛ 3 − ξΛ 4 ) 
1
2

H1 ( χ , ξ ) =


 −4Λ1Λ 2 ( χΛ 4 + ξΛ 3 )

(
)(
)(
) (
)
(3.36)
12


2
2 2
1
Λ
+
+
Λ
Λ 72 + Λ82 k 2 + l 2 + 2 Λ 52 − Λ 62 ( k Λ 7 − l Λ8 ) 
5
6

H2 ( k,l ) =


 −4Λ5 Λ 6 ( k Λ8 + l Λ 7 )

(3.37)
and Im {i} in (3.34) means the imaginary part of the complex i and the vertical bar i
denotes the amplitude of the complex i .
36
We can readily determine χ from (3.33) in terms of ξ as
(
)

s 2
−C2 ∓ C22 − C1  C1ξ 2 + C3ξ + Λ 32 + Λ 42 − S11



χ=
,
C1
(3.38)
where C1 , C2 , and C3 are known quantities and are given as
(
C1 = Λ12 + Λ 22
(
)
(
2
)

s 2
1
S
−
Λ 32 + Λ 42  ,
11


(3.39)
)


s 2
s 2
C2 = Λ12 − Λ 22 Λ3 1 − S11
+
2
Λ
Λ
Λ
1
+
S
1
2
4
11


,




(
)


s 2
s 2
C3 = 2 Λ12 − Λ 22 Λ 4 1 + S11
 − 4Λ1Λ 2 Λ3 1 − S11  .




(3.40)
(3.41)
The correct root for χ in (3.38) is selected by imposing r23 ≤ 1 . Below, we will prove
that −1 < χ < 0 and −1 < ξ < 0 , which not only helps the selection of the correct root for
χ but also reduces the computation times considerably. The proof is as follows.
For a low–loss medium 2 and a lossy sample, we can assume that tan δ 3 > tan δ 2
where tan δ 3 and tan δ 2 are, respectively, the loss tangents of the sample and the medium
2 in Fig. 3.1. Next, using (3.32) we expressed χ and ξ as
n12 + n22 ) − ( a 2 + b 2 )
(
,
χ=
ξ=
2an1 ( n2 n1 − b a )
,
(3.42)
n1 = Re {− j ( λ0 2π ) γ 2 } , n2 = − Im {− j ( λ0 2π ) γ 2 } ,
(3.43)
( n1 + a )2 + ( n2 + b )2
( n1 + a )2 + ( n2 + b )2
where
a = Re {− j ( λ0 2π ) γ 3} , b = − Im {− j ( λ0 2π ) γ 3} ,
and Re{i} means the real part of the complex i . It is clear from (3.43) that a > 0 , b > 0 ,
n1 > 0 , and n2 > 0 for λ0 < λc . Since a and b , respectively, are mainly dependent on
37
Re {ε } and Im {ε } , then tan δ 3 will be proportional to b a . In the same manner, tan δ 2
will be proportional to n2 n1 . Then, it is obvious that ξ < 0 . Assuming, for the sake of
simplicity, a = n1 in (3.42) will prove that χ < 0 .
We can substitute χ in (3.38) into (3.34) to compute ξ . However, k , l , and
Im {− jγ 3} must be written in terms of χ and ξ for this computation. To this end, using
(3.11) and (3.32) we found k and l as
(1 − χ )
(
k=
l=
2
)(
) (
)(
+ ξ 2 n12 + n22 − (1 + χ ) + ξ 2 m12 + m22
2
Π1 − Π 2
(
)
2 χ 2 + ξ 2 − 1 ( m2 n1 − m1n2 ) − 4ξ ( m1n1 + m2 n2 )
Π1 − Π 2
(
2
)(
) (
),
(3.45)
,
)(
2
)
Π1 = (1 − χ ) + ξ 2 n12 + n22 + (1 + χ ) + ξ 2 m12 + m22 ,
(
(3.44)
)
(3.46)
Π 2 = 2 χ 2 + ξ 2 − 1 ( m1n1 + m2 n2 ) + 4ξ ( m1n2 − m2 n1 )
(3.47)
m1 = Re {− j ( λ0 2π ) γ 4 } , m2 = − Im {− j ( λ0 2π ) γ 4 }.
(3.48)
where
In a similar way, Im {− jγ 3} is expressed from (3.12) and (3.32) as
(
)
2
2


2π  2ξ n1 + χ + ξ − 1 n2 
Im {− jγ 3} =
.
2

λo 
1+ χ ) + ξ 2
(


(3.49)
Then, the simple objective function that depends on only ξ is derived as
s
m
f (ξ ) = S21
− S21
= 0.
(3.50)
where the superscript ‘m’ in (3.50) denotes the measured parameter. Because (3.50)
contains an exponential term, an explicit solution for ξ (and χ ) is not possible.
38
However, since we derived a simple objective function which depends on only one
unknown quantity, the solution can be cast by any simple one–dimensional numerical
technique [75]. For example, it is observed that the Newton’s method with ξ = 0 initial
guess converges to the actual solution in less than 15 iterations. After computing ξ from
(3.50), χ can be determined from (3.38). Finally, ε 3 will be
2
ε3 =
2
Λ10
2
λ
2 λ 
λ 
−  o  Im {− jγ 3} +  o  + j 2Λ10 o Im {− jγ 3} ,
2π
 2π 
 λc 
 
λ


Λ10 = ξ  n2 − o Im {− jγ 3}  + n1 (1 − χ ) 
2π

 

3.2.4
(1 + χ ) .
(3.51)
(3.52)
Functional Analysis of the Derived Objective Function
A function g : H → K (from mapping H to the mapping K ) is called “one–to–one” if
and only if its functional dependency is either increasing or decreasing for every possible
value of the independent quantity [85]. Because it is difficult to show the “one–to–one”
property of f (ξ ) mathematically, we choose a numerical technique instead. In the
analysis, we drew f (ξ ) over −1 < ξ < 0 for different ε 2 , ε 4 , ε 3 , λ0 , L2 , L4 , and L3 at
s
s
and S21
for given ε 2 , ε 4 , ε 3 , λ0 , L2 , L4 , and L3 .
X–band. Firstly, we computed S11
Then, we compared the computed ξ from (3.50) with the assumed one. For example,
Fig. 3.4 shows such a dependency of f (ξ ) for different ε 3 . It is assumed that L2 =10
mm, L4 =15 mm, L3 =25 mm, λ0 =29.979 mm, ε 2 = 2.5 − j 0.0184 , and ε 4 = 1.2 − j 0.012 .
It is seen in Fig. 3.4 that there is only one point which satisfies f (ξ ) ≅ 0 for different ε 3 .
This shows that f (ξ ) is a “one–to–one” function.
39
A function g : H → K is called “onto” if and only if for every possible variable in
H there is a correspondence in K [85]. It is clear from (3.38) that not every ξ will yield
s
since the value of the terms in the
a real χ for given different ε 2 , L2 , λ0 , and S11
square–root in (3.38) might be negative. However, by restricting the computations in
(3.38) for acceptable χ quantities will help us determine an acceptable ξ . It is because
of that the dependencies of f (ξ ) in Fig. 3.4 are given for −0.45 < ξ < 0 . It is noted that
restricting the computations not only shrinks the domain for computations but also
significantly decreases computation times.
3.2.5
Refining the Computed Complex Permittivity
In the derivation for ε 3 of lossy samples in (3.50), we used simplified expressions in
(3.33) and (3.34). Accordingly, the determined ε 3 will be fairly accurate. To compute its
precise value, the ε 3 determined from (3.50) can be used an initial guess for the
expressions in (3.13) and (3.14) in a search algorithm since we are searching for a
solution near to this fairly accurate value [75].
Fig. 3.4.
Dependency of f (ξ ) over ξ for different ε 3 .
40
3.3
An Amplitude-Only Method for Thickness-Independent Complex Permittivity
Determination of Lossy Materials
In the previous section, we proposed a method which uses only the amplitudes of S–
parameters and derived an objective function, which depends on only one unknown
quantity, for fast and accurate complex permittivity determination of lossy materials.
Although this method is very applicable for industrial–based applications, the
requirement of precise thickness information of these materials before permittivity
measurements can limit its potential. It is always attractive to eliminate this need not only
in the sense of improving the accuracy of measurements but also for saving time,
especially for measurements of liquid or granular materials, which are positioned
between two low–loss dielectric materials with known electrical properties [25].
In addition to the investigations of characterization of the behavior of materials
reported in the literature, frequency variation techniques have been used for several
purposes in nonresonant measurement methods. They have been applied as a decision
maker for eliminating the superfluous solutions of the extracted complex permittivity, ε ,
and complex permeability, µ , of materials by the well−known Nicolson−Ross−Weir
technique [22],[23]. They have also been employed for obtaining an initial guess for the
ε determination of materials using transmission scattering (S−) parameter measurements
[86]. They are also useful in assigning a correct propagation constant, from which the ε
of materials can be extracted, using uncalibrated S−parameter measurements [87] and for
determining the ε , µ , and/or thickness of materials [88]−[90]. An additional usage of
frequency variation techniques is for optimization of measurements of physical properties
(thickness, disbond, delamination, etc.) of materials [50]. The main process for the
41
aforementioned techniques is to obtain independent measurements at different
frequencies and then extract or optimize the desired information (permittivity,
permeability, thickness, etc.).
It is known that phase−measurements are not convenient for dynamic
characterization of materials since the measured phase may differ by an integral multiple
of 2π from its actual value [26]. To overcome this ambiguity, Weir proposed a method
involving a comparison between calculated and measured group delays of the material at
different frequencies [23],[29]. In a recently published paper, Varadan and Ro utilized the
well−known Kramers−Kronig relations to resolve the same ambiguity [91]. Although
these methods are very useful, the use of expensive vector network analyzers (VNAs) for
complex S−parameter measurements makes microwave techniques inconvenient for
industrial−based applications [32]. Therefore, there is a need for a simple and relatively
inexpensive microwave method for dynamic measurements. In this section, our purpose
is to elaborate on the usage of frequency variation for ε determination of lossy materials
with no a priori knowledge of the thicknesses of the samples using amplitudes of
S−parameter measurements at slightly separated frequencies.
In the analysis, we will use the expressions for complex S–parameters in (3.17)
and (3.18) and are repeated for convenience as
),
(3.53)
2
1 − r23
(
)T3 .
=
(3.54)
S11 =
S21
(
r23 1 − T32
2 2
1 − r23
T3
2 2
1 − r23
T3
42
Applying the single–pass technique described in the subsection 3.2.2 reduces
(3.53) and (3.54) to
s
S11
= r23 ,
(
)
s
2
S21
= 1 − r23
T3 ,
(3.55)
where we made use of T32 << 1 and the superscript ‘s’ at the expressions of S–parameters
in (3.55) denotes simplified expressions.
3.3.1
Zero-Order (ZO) Approach for Complex Permittivity Determination
Because ε changes over f , we have to consider or simulate that change in our ε
determination. However, for very small frequency shifts, we can approximate
ε ( f ) = ε ( f ∓ ∆f ) where ∆f
f [88]. We will name this approximation the “zero–order”
for the remaining of this section for the reasons which will be explained in the next
subsection. To demonstrate this in our problem, we define new variables to simplify the
derivations as
χ1 − jξ1 = ε − ( λ0 λc ) ,
2
κ1 = 1 − ( λ0 λc ) .
2
(3.56)
The definition of new variables changed the inverse problem from ε
determination into χ1 and ξ1 determination. Using (3.56), we obtain
s
S11 =
( χ1 − κ1 )2 + ξ12 ,
( χ1 + κ1 )2 + ξ12
s
S21 =
4κ1 χ12 + ξ12 e
−2πξ1L3 λ0
( χ1 + κ1 )2 + ξ12
.
(3.57)
Even if we assume that ε stays constant over small frequency shifts, χ1 , ξ1 , and κ1 in
(3.56) will all change with f as a result of fc . We denote their new values as χ 2 , ξ 2 ,
s
is independent of L3 , while
and κ 2 at f 2 = f ∓ ∆f . It is obvious from (3.57) that S11
43
s
S21
is exponentially decreasing with it. Consequently, L3 should be eliminated from
s
S21
in (3.57) for a thickness–independent ε determination. Using (3.57), we eliminate
s
L3 from S21
as
 ( χ + κ )2 + ξ 2
1
s
 1 1
S21
2
2
 4κ χ + ξ
1
1
1

2πξ2
2πξ1
 λ02
 ( χ + κ )2 + ξ 2
 λ
2
2
s  0

S21
,
= 2
*

 4κ χ 2 + ξ 2

2
2
2



(3.58)
s
s
where λ02 = c f 2 and S21
is the measured amplitude of S21
at f 2 . We can easily
*
s
s
s
and S11
, which is the measured amplitude of S11
at
determine ε from (3.58) using S11
*
f 2 , by any two–dimensional root–searching algorithm [75]. However, for industrial–
based applications it is appropriate to have a well–defined domain for fast numerical
computations. In this respect, we decided to use one–dimensional root–finding
algorithms, which require a function with only one–variable. We obtained a function
which depends on only χ1 as follows.
s
Using S11
in (3.57), we expressed ξ1 in terms of χ1 and κ1
(
s 2
ξ1 = 2 χ1κ1Λ1 − ( χ12 + κ12 ) , Λ1 = 1 + S11
) (1 − S s ).
2
11
(3.59)
It is obvious from (3.56) that the equation for ξ 2 will be in similar from in (3.59) except
s
s
→ S11
. At this stage, we expressed ξ1 and ξ 2 in
for χ1 → χ 2 , κ1 → κ 2 , and S11
*
functions of χ1 and χ 2 , respectively. As a result, if we write χ 2 in terms of χ1 and ξ1 ,
we will obtain a function which depends on only χ1 . To this end, using (3.56) and (3.59),
we found
44
χ2 =
κ 2 Λ 2 ∓ κ 22 Λ 22 + 2 ( χ12 − ξ12 − κ12 )
2
, Λ2 =
s 2
1 + S11
*
s 2
.
(3.60)
1 − S11 *
Accordingly, we can then determine χ1 by substituting ξ1 , ξ 2 , and χ 2 into
(3.58) and using any suitable one–dimensional root–finding algorithm [75]. After
computing χ1 , we can analytically determine ξ1 , χ 2 , and ξ 2 from (3.59) and (3.60). As
a result, ε will be
2
2
ε = ε ' − jε " = χ m
− ξm
+ λ02m λc2 − 2 j χ mξ m ,
(3.61)
where χ m , ξ m , and λ0 m are, respectively, the mean values of χ1 and χ 2 , ξ1 and ξ2 ,
and λ0 and λ02 .
It is noted that the equation in (3.58) can still be employed for ε determination in
coaxial–line or free–space measurements where λc → ∞ . For these measurements,
χ 2 = χ1 = χ , ξ 2 = ξ1 = ξ , and κ 2 = κ1 = κ . However, since λ02 ≠ λ0 , we can still
determine ε from (3.58).
s
After computing ε from (3.61), L3 can be calculated from S21
in (3.57).
Because we use approximations to S11 and S21 , the computed ε and L3 will be fairly
accurate but must be refined for precise values. For this purpose, we used the least
squares minimization [75], which is given as
min
∑ ( Sα
α
− Pα
)2 + ∑ ( Sα * − Pα * )
α
2
for ε , L3 ,
(3.62)
where α = 11 or 21 , Sα and Pα are the measured and predicted amplitudes of S–
parameters at f , and the subscript '*' denotes the measured and predicted amplitudes of
S–parameters at f 2 .
45
3.3.2
Higher-Order (ZO) Approach for Complex Permittivity Determination
It is seen from (3.58) that for the ZO approximation to be applied to our problem one has
s
s
s
s
to measure different S11
, S11
, S21
, and S21
. If the sample does not yield different
*
*
amplitude–only measurements over small frequency shifts, we have to search for a new
frequency that yields different amplitude–only measurements. However, using
amplitude–only S–parameter measurements at two widely separated frequencies may not
be suitable for the ZO approximation and then the simplified expressions in (3.55) may
not be substituted for those in (3.53) and (3.54) since ε can change considerably. In this
circumstance, higher–order (HO) approximations for ε should be used. For these
approximations, Baker–Jarvis et al. proposed an extended Debye relaxation model [89],
which in a compact form is
ε( f )=
2
n
∑ An (1 + j 2π Bn f ) ,
(3.63)
n =0
where An and Bn are the unknown complex quantities. The equation in (3.63) can also
be written as
ε( f )=
m
∑ Cn f n = C0 + C1 f + C2 f 2 + C3 f 3 + ...,
(3.64)
n =0
where Cn ’s are the unknown complex quantities and m is the degree of approximation.
The equation in (3.64) will reduce to ZO approximation when m = 0 [88] and linear
interpolation when m = 1 [90].
It is clearly seen from (3.64) that HO approximations necessitate an initial guess
for ε extraction. Although they are more general than the ZO approximation, the latter
gives an insight on electrical behavior of the sample and presents a readily available
46
initial guess for HO approximations for ε . To determine ε using (3.64), we substitute an
initial guess for ε into (3.64) determined by the ZO approximation from measurements
at f and at an f 2 , which is the nearest frequency to f . Then, we refine this ε by
iterating it and L3 using (3.62).
3.4
Measurements
In this section, we present measurement results of some liquid materials (methanol, a
commercially available antifreeze solution, and a binary mixture solution of ethyl alcohol
and distilled water) for validation of the methods given in sections 3.2 and 3.3. We
constructed a general purpose waveguide measurement set–up for this purpose as shown
in Fig. 3.5. The set–up has a HP8720C VNA connected as a source and measurement
equipment, two dielectric plugs and a sample inside a waveguide holder, two coaxial–to–
waveguide adapters, and two extra waveguide sections more than 70 mm in length
between the calibration plane and adapters. It operates at 8.2–12.4 GHz (X–band).
Fig. 3.5.
The measurement set–up.
47
We assumed single–mode transmission (TE10) through the sample in this chapter.
This condition for empty and sample–filled sections of the waveguide will not be
consistent for a sample with ε ' > 4 [92]. In this case, higher–order modes may appear.
Using a sample thickness less than one–half guided wavelength of the fundamental mode
will suppress these modes [29]. However, this may decrease the measurement accuracy
as a result of sample thickness uncertainty. Another option could be using two extra
waveguide sections with lengths greater than 2λ0 between the sample and coaxial–to–
waveguide adapters [21],[29],[41]. This is because higher–order modes will die out
drastically in a short distance away from the sample, and actual measurements are
performed near the adapters. We used extra waveguide sections, which are greater than
2λ0 at X–band, to eliminate any higher–order mode in the waveguide set–up in Fig. 3.5.
Since our primary concern is to extract ε from amplitude–only S–parameter
measurements,
any
calibration
technique
which
uses
complex
S–parameter
measurements, such as the Thru–Reflect–Line (TRL) technique [93] and the Line–
Reflect–Line (LRL) technique [94], is not suitable for validation of the method.
Therefore, we calibrated the set–up by the frequency response calibration [95], which
only removes the frequency (tracking) error by normalizing the frequency responses of a
waveguide short and of a thru standard (extra waveguide sections are connected back–to–
back). To verify the measurements, we also calibrated the set–up by the TRL calibration.
We used a waveguide short and the shortest spacer in our lab (a 44.38 mm long
waveguide section) for reflect and line standards, respectively. The line has a ∓ 70o
maximum offset from 90o between 9.7 GHz and 11.7 GHz. Figs. 3.6(a) and 3.6(b),
respectively, show a comparison between the TRL calibration and the frequency response
48
calibration with/out aperture smoothing for measurements of waveguide short and thru
connections.
It is observed that calibration results in Figs. 3.6(a) and 3.6(b) are clearly in
agreement with those in the VNA operating manual [95]. We noted from the
measurements in Figs. 3.6(a) and 3.6(b) that the frequency response calibration results in
more fluctuations than the TRL calibration. This is because it corrects only frequency
errors while the TRL calibration performs a full two–port calibration [95]. However,
applying aperture smoothing to the measurements obtained from the frequency response
calibration allowed us to measure quite similar results for especially the thru connection
in Fig. 3.6(b). In addition, for measurements of lossy materials a maximum 0.5 dB
difference between the two calibrations for the waveguide short connection in Fig. 3.6(a)
will not much affect the ε determination.
It is appropriate to discuss measurement uncertainties in the system before
conducting measurements. Several factors contribute to the uncertainty in ε
determination in waveguide measurements [21], [30] namely: 1) the uncertainty in
measured S–parameters; 2) errors in the sample length (and the dielectric plugs); 3) the
uncertainty in reference plane positions; 4) guide losses and conductor mismatches; 5) air
gaps between the external surface of the sample (and dielectric plugs) and inner walls of
waveguides; and 6) higher order modes. All these uncertainties are well–treated in the
literature. The reader can refer to [21], [24], [25], [29]–[31], [96]–[99] for details.
Care can be taken to limit the uncertainties due to sample position, calibration,
and air gaps. Using calibration techniques, uncertainties in reference plane positions can
be minimized. Preparing the sample (and dielectric plugs) and using a sample holder with
49
no scratches, nicks, or cracks, and machining the sample (and dielectric plugs) to fit
precisely into the waveguide will increase the measurement accuracy and reduce the
effect of air gaps [21], [29].
Fig. 3.6.
Comparison of the frequency response (R) with/out aperture smoothing and
the TRL calibrations for (a) waveguide short and (b) thru connections.
50
We machined two identical polytetrafluoro–ethylene (PTFE) samples (10 mm) for
dielectric plugs. We utilized them for holding the liquids in place so that there will be no
meniscus formation on surfaces of the liquids. The permittivities of PTFE samples are
measured over 9.7–11.7 GHz using explicit formulae derived by Boughriet et al. [31]. Its
value is around 2.04 − j 0.0014 over the band. Then, we poured the ethyl alcohol–water
solution and antifreeze solutions into the measurement cell (waveguide section) and
pressed on the plugs to decrease any meniscus that can form on top of the liquids [25].
Next, we carried out complex S–parameter measurements and collected 801 data points
evenly spaced between 9.7 GHz and 11.7 GHz. We also extended the calibration plane to
the surface of the PTFE samples using the VNA port extension facility. Next, we
measured the ε of the liquids by the plug–loaded two–port transmission–line method
(PLTL) [25] using complex S–parameter measurements after the TRL calibration. We
also applied time–domain gating to the main transmission properties of measured
complex S–parameters to obtain smoother ε measurements.
Because frequency response calibration corrects only frequency errors, it
produced fluctuations on amplitude–only scattering parameter measurements throughout
the frequency band. The time–domain gating facility of the VNA cannot be applied to
decrease these fluctuations since it requires phase information. Instead, we applied 20 %
aperture smoothing to obtain an even dependency for amplitude–only S–parameter
measurements. Since the measured ε of the liquids from (3.50) and (3.51) by our method
are not precise, we refined it by using the Newton’s search algorithm [75]. Figs. 3.7 and
3.8 demonstrate the measured ε of the liquids by our method and the PLTL method [25].
It is seen from Figs. 3.7 and 3.8 that measured ε of the liquids from the
51
frequency response and TRL calibrations are completely consistent [33]. Small
differences between measured ε by the methods arise because of a maximum 0.5 dB
difference in measured S11 by different calibrations. We obtained highly accurate results
for both methods since the measured S21 was greater than –30 dB with respect to the
reference (thru) signal. As it is known, the main uncertainty in measurements of lossy
samples comes from S21 . If it is less than –40 dB (in fact, it is –50 dB for the HP8720C
VNA over 8–20 GHz), the level of measurements will be comparable with that of noise
present in the system, and thus the uncertainty will significantly increase [21], [25].
Fig. 3.7.
Measured ε of an antifreeze solution extracted by the plug–loaded two–
port transmission–line method (PLTL) [25] and the proposed method (PM)
in section 3.2 without aperture smoothing and with 20 % aperture
smoothing.
52
Fig. 3.8.
Measured ε of a binary mixture of ethyl alcohol (75%) and water (25%)
by the plug–loaded two–port transmission–line method (PLTL) [25] and
the proposed method (PM) in section 3.2 without aperture smoothing and
with 20 % aperture smoothing.
After validation of the proposed method in section 3.2, we also validated our
second method in section 3.3. Since the derivations given in section 3.3 are not applicable
for a dielectric–sample–dielectric structure, we removed the effect of PTFE samples from
measurements using the de–embedding technique [28]. After, we applied our second
method to measure the ε
of the methanol using amplitude–only S–parameter
measurements after applying 10 % and 20 % aperture smoothing to frequency response
calibration. In applying our method, firstly ε is computed using (3.58)–(3.62). This ε
gives its probable value over 9.7–11.7 GHz. Then, we refined it using (3.64) with m = 1
53
along with (3.62). We continued this refining with larger m values until the refined ε
with present and previous m values is approximately the same. Figs. 3.9(a) and 3.9(b)
illustrate, correspondingly, the measured real and imaginary parts of the permittivity of
the methanol. It is seen from Figs. 3.9(a) and 3.9(b) that the measured ε by the PLTL
method and by our method are in good agreement with those in [33]. Small differences
between measured ε by the methods arise because of a maximum 0.5 dB difference
between the applied calibrations. As discussed before, we obtained highly accurate
results for both methods since the measured S21 was greater than –30 dB with respect to
the reference (thru) signal.
54
Fig. 3.9.
(a) Real part and (b) imaginary part of the measured ε of methanol by the
plug–loaded two–port transmission–line method (PLTL) [25] and by the
proposed method (PM) in section 3.3 without aperture smoothing and with
10 % and 20 % aperture smoothing.
55
CHAPTER 4
A WAVEGUIDE SAMPLE HOLDER FOR AUTOMATED THICKNESS–
INDEPENDENT COMPLEX PERMITTIVITY AND COMPLEX
PERMITTIVITY MEASUREMENTS OF GRANULAR AND LIQUID
MATERIALS
In chapter 3, we introduced two methods for complex permittivity determination of lossy
materials using amplitudes of scattering (S–) parameter measurements. Although these
methods can easily be implemented by using a relatively inexpensive scalar network
analyzer (SNA), they are not sufficient for full electrical characterization (complex
permittivity, complex permeability, etc.) of low–to–high–loss materials. In addition, they
always require a numerical analysis for extracting the complex permittivity since real and
imaginary parts of the complex permittivity cannot be exclusively expressed in terms of
amplitudes of S–parameter measurements.
A problem encountered in practice in nonresonant methods is the transformation
of S– parameter measurements from the calibration plane to end surfaces of the sample
[30]. The port extension feature of the vector network analyzers (VNAs) can be used for
this transformation; however, they require precise lengths between the calibration plane
and the sample, which in some instances are unknown. In addition, thickness–
independent complex permittivity, ε , and/or complex permeability, µ , measurements of
56
materials are very attractive since the requirement of the physical thickness measurement
of the materials is eliminated. In the literature, the problems pertaining to either
calibration plane–invariant or sample length–independent ε and µ measurements of
materials have been addressed. For example, Altschuler proposed the determinant of a
network (admittance) presentation of materials under test to determine calibration plane–
invariant ε in a simple fashion [100]; and for the same purpose, Baek et al. proposed a
position insensitive method for ε and µ determination of materials [101]. On the other
hand, Sucher and Fox presented several methods for thickness–independent ε
determination of materials backed by a short circuit termination [102]. Because the short–
circuit termination at different locations away from the sample could result in similar S–
parameter measurements as a result of the same reactive termination, another method
which eliminates this problem using measurements of broadband and short–circuit
terminations was proposed in [103]. Baker–Jarvis et al. were the first researchers who
derived equations that are simultaneously invariant to the location of the calibration plane
and the sample length [30] for ε determination of materials. Although these equations
are useful, the relation between ε and the sample length is not explicit. Therefore, any
method using these equations requires a numerical technique for thickness–independent
ε determination. In addition, the derived equations are limited to ε extraction of
materials.
Owing to their relative simplicity, broad frequency coverage, feasibility for
analyzing the measurements either in time–domain or frequency–domain, and relatively
higher accuracy, nonresonant methods have been widely utilized [29]. When applying
these methods for measurements of granular and liquid materials, some form of
57
approximation in the physical nature of the measurement apparatus and/or forward and
inverse problem formulations are made [25],[33] or formulations require that the
dielectric specimen be low–loss [34]–[36]. To avoid any meniscus formation on the top
of liquid samples, generally the measurement set–up is positioned vertically [37]. For the
case of granular specimens, planar measurement planes may not be easily achieved [25]
and this can degrade the measurement performance.
In this chapter, we will propose a waveguide sample holder for calibration–plane
invariant thickness–independent complex permittivity, ε , and/or complex permeability,
µ , determination of low–, medium–, and high–loss granular and liquid materials. The
holder readily lends itself for automated measurements using a VNA.
4.1
Theoretical Analysis
The geometry for simultaneous ε and µ measurement of a liquid and/or granular sample
with length, Ls , in a waveguide is shown in Fig. 4.1.
Fig. 4.1.
The complex S–parameter measurements of granular and liquid samples
sandwiched between stable and movable dielectric plugs.
58
The sample is sandwiched between stable and movable plugs. It is assumed that
the sample and plugs are isotropic and homogenous. Also, we assume that the waveguide
operates at the dominant mode (TE10).
It is obvious that the structure we consider in Fig. 4.1 is similar to that in Fig. 3.1.
Therefore, the derived S–parameters for the five–layer structure in chapter 3 can be
utilized in this chapter with some modifications. Assuming that the measurement set–up
in Fig. 4.1 is calibrated to the calibration plane (it is assumed that the line standard is an
empty waveguide section), by making analogy to the five–layer structure in Fig. 3.1 we
can model, respectively, the stable plug, the sample, the movable plug, and air regions as
media 1 through 4. As a result, using the expressions in (3.13), (3.14), and (3.21), we
derived forward and reverse complex S–parameters for the structure in Fig. 4.1 as
Γ s − Γ p ) (1 − Γ p Γ sT p2 ) + ( Γ pT p2 − Γ s ) (1 − Γ p Γ s ) Ts2
(
S11 =
,
2
2
2
1
−
Γ
Γ
T
1
−
Γ
Γ
+
Γ
−
Γ
Γ
T
−
Γ
T
(
)
( p s p ) p s ( s p )( p p s ) s
)(
(
) (
)(
)(
)
(4.1)
 1− Γ Γ Γ T 2 − Γ + Γ −T 2 Γ − Γ T 2 
p s
s p
p
s
p
s
p s  2
=
T ,
2
2
2 0
 1− Γ Γ T 1− Γ Γ + Γ − Γ
T
T
Γ
−
Γ
p
s
s
s
p
s
s
p
p
p
p


(4.2)
1 − Γ 2s )(1 − Γ 2p ) TsT pT0
(
S21 = S12 =
,
(1 − Γ pΓsTp2 ) (1 − Γ pΓs ) + ( Γ s − Γ p ) ( Γ pTp2 − Γs )Ts2
(4.3)
S22
(
)(
) (
)
where
Γp =
(
γ p − γ 0µ p
γ p + γ 0µ p
, Γs =
)
γ p µ − µ pγ s
γ p µ + µ pγ s
,
T p = exp −γ p L p , Ts = exp ( −γ s Ls ) , T0 = exp ( −γ 0 L0 ) ,
59
(4.4)
(4.5)
γp = j
2π
λ0
2
2
2
2π
2π
 λ0 
λ 
λ 
, γs = j
1−  0  .
εµ −  0  , γ 0 = j

λ0
λ0
 λc 
 λc 
 λc 
ε pµ p − 
(4.6)
Here, Γ p , T p , Γ s , and Ts are, respectively, the first reflection and transmission
coefficients of the movable plug– and the sample–filled cells; T0 is the propagation factor
in the air–filled cell; γ 0 , γ p , and γ s represent, respectively, the propagation constants of
the air–, movable plug–, and the sample–filled cells; ε p , µ p , ε , and µ are,
correspondingly, the relative complex permittivities and permeabilities of the movable
plug and the sample; L p and L0 denote the lengths of the movable plug and the air
section inside the waveguide; and λ0 and λc correspond to the free–space and cut–off
wavelengths.
4.2
The Method
In the formulation, it is assumed that we know the electrical properties ( ε p and µ p ) and
the length ( L p ) of the movable plug and the length of the waveguide holder
Lg = Ls + L p + L0 . Our goal is to determine the constitutive parameters of the sample
while its length and the air–section length inside the holder are unknown. Because of the
asymmetry of the measurement cell (sample, movable plug, and air) inside the calibration
plane, we could not directly apply the Nicolson–Ross–Weir (NRW) technique [22],[23].
However, we obtained a different formulation as follows.
It is clear from (4.1)–(4.6) that S11 is independent of L0 while S 22 and S21 are
functions of L0 . We firstly eliminated L0 using (4.2) and (4.3) as
60
(
)
2
2
1 − Γ 2s (1 − Γ p ) Ts2T p2
S21
=
.
S 22
Λ AΛ B
)(
(
2
) (
(4.7)
)(
)
Λ A = 1 − Γ p Γ s Γ sT p2 − Γ p + Γ s − T p2 Γ s − Γ p Ts2 ,
(
)(
)(
) (
(4.8)
)
Λ B = 1 − Γ p Γ sT p2 1 − Γ p Γ s + Γ s − Γ p Γ pT p2 − Γ s Ts2 .
Then, because Γ p and T p are known, we expressed Ts2 in terms of Γ s using the S11 in
(4.1) as
1 − Γ p Γ sT p2 ) ( Γ s − Γ p ) − (1 − Γ p Γ s ) S11 
(

.
Ts =
2
( Γ pTp − Γs ) ( Γs − Γ p ) S11 − (1 − Γ pΓs )
2
(4.9)
Substituting Ts in (4.9) into (4.7), we find a function depending on Γ s , Γ p , and
T p as
(
)
F Γs , Γ p ,Tp =
(
Ω1 1 − Γ 2s
) ( Γ pΓsTp2 − 1)( Ω2Γ2s + Ω3Γs + Ω2 )
2
( Γs − Γ pTp ) ( S11 + Γ p ) Γs − S11Γ p − 1 ( Γ pΓs − 1)
2
2
2
= 0,
(4.10)
where
2
S21
2


Ω1 = Γ p − 1 , Ω 2 = U − S11 − Γ p S11 − 1 Γ p − S11 T p , U =
,


S22
(
{
2
)
((
2
(
)
)
(4.11)
}
)
Ω3 = ( ( S11 − U ) S11 + 1) Γ p + T p2 − 2 U + 4S11  Γ p + (1 + ( S11 − U ) S11 ) T p2 + U Γ p . (4.12)


After some functional analysis, it is observed that among the numerator terms,
only Ω 2 Γ 2s + Ω3Γ s + Ω 2 satisfies the equation in (4.10). Therefore, the explicit expression
for Γ s will be
Γ s(1,2 ) =
−Ω 4 ∓ Ω 42 − 4
2
61
, Ω4 =
Ω3
.
Ω2
(4.13)
The correct root of Γ s from (4.13) can be selected by imposing Γ s ≤ 1 . After
determining Γ s , Ts can be calculated by substituting that Γ s into (4.9). Then, we can
determine L0 using either (4.2) or (4.3). Finally, Ls , ε , and µ will be
Ls = Lg − L p − L0 ,
(4.14)
2
 j

=
ln (Ts )  ,
2
Λ
 2π Ls

1
µ = µp
ε=
4.3
(1 + Γ s )
(1 − Γ s )
(4.15)
1
Λ
ε pµ p
λ02
−
1
,
(4.16)
λc2
2
2
1  λ0   λ0  
  +    .
µ  Λ   λc  
(4.17)
Measurements
We utilized the general purpose waveguide set–up introduced in chapter 3 for validation
of the proposed method in this chapter. Because the proposed method in this chapter
requires one of the plugs (left one) in Fig. 3.6 be fixed at the left terminal of the
calibration plane, we shifted that plug and changed the configuration of the set–up to be
similar to that in Fig. 4.1. We utilized two 10 mm polytetrafluoro−ethylene (PTFE)
samples as plugs for holding the liquid samples in place so that there will be no meniscus
formation on surfaces of liquid samples. We measured the ε p of the PTFE samples by
the plug-loaded two-port transmission line (PLTL) method [25] while only the PTFE
sample is present inside the holder in Fig. 4.1, and found ε p ≅ 2.03–j0.0014 over 9.7–
11.7 GHz. We calibrated the set–up by the Thru–Reflect–Line (TRL) technique [93]
62
since it does not necessitate precise or well–defined calibration standards such as the load
standard. We used methanol, a commercially available antifreeze solution, and distilled
water as test samples to validate the method. Figs. 4.2–4.4 demonstrate the measured ε
of these liquids over 9.7–11.7 GHz.
Fig. 4.2.
(a) Real part and (b) imaginary part of the measured (dashed line) and
theoretical (solid line) ε of distilled water.
63
Fig. 4.3.
(a) Real part and (b) imaginary part of the measured (dashed line) and
theoretical (solid line) ε of methanol.
64
Fig. 4.4.
(a) Real part and (b) imaginary part of the measured ε of an antifreeze
solution by the plug–loaded two–port transmission line method (PLTL)
[25] and by the proposed method.
65
It is seen from Figs. 4.2 and 4.3 that both the real and imaginary parts of the
measured ε of the distilled water and methanol are in good agreement with those
computed from Debye model [104], [105]. The theoretical curve is obtained using the
Debye equation
ε ' = ε∞ +
ε" =
ε s − ε∞
,
1 + ω 2τ 2
(4.18)
(ε s − ε ∞ ) ωτ ,
(4.19)
1 + ω 2τ 2
where ω is the angular frequency; ε ∞ is infinite frequency dielectric constant (or the
dielectric constant when ω goes to infinite); ε s is static dielectric constant (or the
dielectric constant when ω goes to zero); τ is the relaxation time. In the literature, one
frequently finds a ‘relaxation wavelength’ or ‘Sprungwellenlange’ λs , defined as
λs = 2π cτ ,
(4.20)
where c is the speed of light.
We obtained ε ∞ , ε s , and τ values for the distilled water and methanol from
[104] and [105] for room temperature (20–25 Co) and are given in Table 4.1.
Table 4.1. The parameters ε ∞ , ε s , and τ for distilled water and methanol. The data
are estimated by interpolation from [104],[105] for room temperature (20–25 Co).
εs
ε∞
τ (ps)
Distilled water
78.5
5.2
8.3
Methanol
32.6
5.6
48
Parameters
66
Fig. 4.4 illustrates the measured ε of an antifreeze solution over 9.7–11.7 GHz by
our method and by the plug–loaded two–port transmission line method (PLTL) [25]. It is
seen from Fig. 4.4 that the real and imaginary parts of the measured ε of the antifreeze
solution by both methods are very good agreement with each other. To apply the PLTL
method, we used the set–up in Fig. 3.6. We also kept the distances between calibration
plane and end surfaces of the plugs minimum so that we could easily measure them. In
our proposed method, we eliminated this necessity using (4.7) and (4.9).
The proposed method assumes that the sample is homogenous and isotropic. In
practice, it may not be homogenous. Naturally–occurring materials (e.g., timber) may
have density and grain orientation variations while manmade materials may have density
and constituent variations [41]. To evaluate the correctness of extracted constitutive
parameters of and/or the homogeneity of samples, as a self–checking feature, the
dependency of computed L0 can be drawn over the frequency band. For example, Fig.
4.5 shows this dependency for measurements of commercially available antifreeze
solution over 9.7–11.7 GHz.
It is observed that standard deviation of the computed L0 is little higher than that
of the air section length of another sample holder operating at 3 GHz [41]. The reason for
this can be the higher frequencies (approximately 10.7 GHz) used for testing our
proposed holder.
67
Fig. 4.5.
Dependency of the computed length of air–filled section, L0 , inside the
proposed sample holder from commercially available antifreeze solution
measurements over frequency.
68
CHAPTER 5
A SIMPLE FREE–WAVE METHOD FOR ELECTRICAL
CHARACTERIZATION OF CEMENT–BASED MATERIALS
In this chapter, we will present a free–wave method for complex permittivity, ε ,
measurements of cement–based materials (mortar, concrete, etc.) with different internal
contents. These materials are widely used in the construction industry. Mortar contains
cement powder, fine aggregate (sand), and water while concrete is composed of mortar
and coarse aggregate (gravel); i.e., cement powder, fine aggregate (sand), water, and
coarse aggregate (gravel). Measurement and monitoring of performance and properties
(compressive and tensile strengths, porosity, permeability, hydration, mixture content,
etc.) of these materials are important for evaluation of their quality and integrity.
As it was discussed in chapter 2, generally reflection properties of these materials
are measured to characterize them. Reflection measurements are very convenient for
applications where the sensor (antenna) can only be placed on one side of the material
(especially for measurements on pavement, back–filled concrete wall, roadway, etc.).
However, any surface roughness of materials can significantly affect reflection
measurements more than transmission measurements at high frequencies [20], [65]. In
addition, transmission measurements provide longitudinal averaging of variations in
material properties, which is particularly important for relatively lossy heterogeneous
materials such as cement–based materials and moist coal [86], [106]; they are also more
69
sensitive to the dielectric properties of high–loss samples [86] and offer a wide dynamic
range for measurements [86].
Although the results obtained from reflection measurements are very promising
[42], [46]–[49], [53], [54], [59], [60], reflection–only measurements do not disclose any
new information about cement–based materials, such as periodic assessment of aging of
civil structures for structural integrity and public safety [59], after approximately 28th day
of curing since reflection properties stay constant. Therefore, transmission measurements
can be utilized to gather more information about these materials [20], [64], [66].
Among nonresonant methods, the free–space or free–wave technique appears to
be more applicable for non–contact, noninvasive, and in situ applications [65]. In the
literature, methods employing free–space measurements use vector network analyzers
(VNAs) for highly accurate measurements and their time–gating facilities to reduce the
multiple reflections inside the sample and between antennas and the sample
[27],[28],[56],[59],[60],[65],[66],[107]–[111]. However, for practical purposes, it is
attractive to use the simplest measurement system because using expensive VNAs is not
convenient for industrial–based applications [32].
In this dissertation, we present microwave reflection and transmission properties
of young/hardened mortar and concrete specimens at X–band (8–12 GHz) by a simple
and relatively inexpensive set–up. Firstly, we will illustrate hourly measurements of
young mortar specimens with different water–to–cement ratios (w/c) changing between
0.40 and 0.60 with a 0.05 of increment during 20–30 and 44.54 hours. Second, we
demonstrate microwave reflection and transmission properties measured from different
sides of hardened mortar and concrete specimens with different w/c ratios at X–band.
70
5.1
Theoretical Background
Fig. 5.1 demonstrates a general configuration for measuring reflection and transmission
properties of cement–materials at far–field. In the theoretical analysis, it is assumed that
the sample is isotropic, non–magnetic, and extends to infinity in length in transverse
dimensions.
Fig. 5.1.
The configuration for the measurement of reflection and transmission
properties by using a free–space method.
Using the S–parameters derived in chapter 3 in (3.17) and (3.18), we can write
(
)
(
)
2
Er Γ 1 − Z
S11 =
=
,
Ei 1 − Γ 2 Z 2
2
Et Z 1 − Γ
S21 =
=
,
Ei 1 − Γ 2 Z 2
where
71
(5.1)
(5.2)
(
)
Z = exp − jk0 L ε ,
Γ=
1− ε
,
1+ ε
ε = ε ' − jε " ,
k0 =
2π f
.
c
(5.3)
Here, Ei , Er , Et , ε , L , k0 , f , c , Γ , and Z are, respectively, the electric field
intensity of incident, reflected, and transmitted waves, the complex relative permittivity
and thickness of the sample, the free–space propagation (or phase) constant (free–space
wave number), the operating frequency, the velocity of light, the reflection coefficient
when the sample is infinite in length, and the propagation factor of the sample.
5.2
The Method
Our goal is to determine ε from amplitudes of S–parameters in (5.1) and (5.2).
Assuming the sample in Fig. 5.1 shows at least 10 dB attenuation and then applying the
single–pass technique, the expressions in (5.1) and (5.2) can be approximated to
s
S11
= Γ,
(
)
s
S21
= Z 1 − Γ2 .
(5.4)
where the letter ‘s’ at the superscript of S11 and S21 in (5.4) denotes simplified
expressions. In the simplification, we made use of Z 2 <<1 where i denotes the
s
s
magnitude of the expression 'i' . In (5.4), S11
is independent of L while S21
is a function
of it.
s
s
It is obvious that the expressions of S11
and S21
in (5.4) and those in (3.33) and
(3.34) are the same for L2 = L4 = 0 and ε 2 = ε 4 = 1 . Therefore, the derived objective
function in chapter 3 can be employed for ε determination using L2 = L4 = 0 and
ε 2 = ε 4 = 1 . Substituting them into (3.36)–(3.50), we find an objective function for this
problem:
72
f 3l (ξ ) =
(1 − χ 2 + ξ 2 )
2


2k0ξ L
s
 − S21
+ 4 χ 2ξ 2 exp 
=0
2
2
 (1 + χ ) + ξ 


(5.5)
where k0 = 2π λ0 , λ0 is free–space wavelength, and
χ=
2
s
S11
−ξ2.
(5.6)
In the derivation of (5.5) and (5.6) from (3.36)–(3.50), we made use of
χ − jξ = Γ, H1 ( χ , ξ ) = 1, H 2 ( k , l ) = 1, k = − χ , l = −ξ .
(5.7)
Because the ‘one–to–one’ property of the objective function derived for the five–
layer structure in Fig. 3.1 is independent on the electrical properties of the dielectric
plugs, the objective function in (5.5) will accurately determine a unique ε . This is
because we already satisfy the condition that the loss tangent of the sample be greater
than that of the medium which is next (left) to the sample (in our problem, it is air).
5.3
The Measurement Set–up
The schematic diagram of the free–space measurement set–up used in measurements is
shown in Fig. 5.2. The set–up measures amplitudes of the reflection and transmission
coefficients (or S11 and S21 ) from electric field strengths of incident, reflected, and
transmitted waves. It is composed of rectangular waveguide sections, a microwave source
(OSC), a rotary attenuator (ATT), a frequency meter (FM), two directional couplers, two
antennas, and three diode detectors. The OSC operates at X–band. Its output power is
approximately 10 mW at X–band and is suitable and applicable for most microwave
nondestructive applications [112]. The signal from the source is modulated by a 1 kHz
73
signal to measure amplitude–modulated reflected and transmitted signals by a simple
reflectometer and an attenuation–meter as shown in Fig. 5.2.
Fig. 5.2.
The schematic diagram of the measurement set–up.
The simple reflectometer is constructed to measure S11 and consists of two
square–law detectors and the couplers. While Coupler I measures incident waves,
Coupler II measures reflection properties. The attenuation–meter (T–meter) measures
S21 and consists of two square–law detectors, Coupler I, a precision attenuator, an
amplifier, and an indicator.
The ATT is a rotary type attenuator, which is used to change the level of the
incident signals and to decrease the level of reflected signals at the source (an isolator is
much preferred for preventing the source from any harmful reflections in the
measurement system or from the measurement cell; however, in our lab there is no
available isolator). It is a FLANN 16110 precision attenuator with 0 to 60 dB attenuation
range and 0.1 dB reading accuracy. The FM (or wavemeter) is utilized for highly
sensitive frequency measurements. It is a FLANN 16072 tunable cavity resonator with
74
0.10 % accuracy and high Q (5500) when loaded. We used FLANN 16131 multihole
directional couplers. They operate at 8.2–12.5 GHz and have a minimum 40 dB
directivity and a 20 dB coupling. Standard gain horn antennas are used for matching the
impedances of air and waveguide apertures. These antennas have an aperture of 60.5 mm
x 45 mm and a gain of nearly 15 dB at X–band.
It is noted that the measurement set–up in Fig. 5.2 was constructed from
microwave components available in the Electromagnetic Fields and Microwave
Techniques laboratory in Electrical and Electronic Engineering Department at Cukurova
University, Adana, 01330, Turkey.
5.4
Calibration Procedure
The transmit and receive antennas in Fig. 5.2 are separated according to: a) the distance
between them fulfills the plane wave condition; b) the center of the samples is exactly
matching the center of horn antennas; c) the aperture of the antennas is positioned to be
parallel to the transverse dimension of the samples; d) the minimum amount of wave is
scattered by the edges of the sample; and e) maximum incident signal is received when
there is no sample between antennas to increase the sensitivity of the measurement set–
up.
The plane wave assumption is satisfied practically by LU >> 2 Dr2 λ where Dr ,
λ ( = c f ), and LU are the maximum lateral dimension of the radiator (diagonal length
for rectangular and square antennas and diameter for cylindrical antennas), the
wavelength, and the distance between the radiator (antenna) and each specimen [113].
According to the cross section of used horn antennas, 60.5 mm x 45 mm, the value of Dr
75
is determined approximately 75 mm and maximum LU is calculated as 380 mm at X–
band. The transmit and receive antennas are separated from each other approximately one
meter. In a recent study [59], it was shown that if the distance between the transmit
antenna and the specimen is more than 8 times the maximum wavelength ( LU > 8λmax ) at
which measurements are carried out, the plane wave assumption can be applied.
According to the minimum frequency at X−band (8.2 GHz) used in our measurement set–
up, 8λmax ≅ 292 mm which is much less than the antenna separation in our set–up
(samples are located at equal distance from both antennas to reduce multiple reflection
between antennas and specimens).
Previous investigations show that the edge diffraction might be practically
omitted by selecting Ds >> 5λ where Ds is the maximum lateral dimension of the
samples (diagonal length for rectangular and square samples and diameter for cylindrical
samples) [113]. The cross dimension of prepared samples, which will be discussed in the
next chapters, is 150 mm x 150 mm. It is readily noted from this information that the
maximum transverse dimension of the prepared samples, Ds =212.1 mm, is greater than
5λmax ≅ 183 mm. The horn antennas with narrow 3−dB beam widths at X−band are used
for further decreasing the diffraction [27].
The maximum amount of the received signal is obtained by arranging various
components of measurement apparatus; the center and parallel alignment and middle
distance of samples according to aperture of the antennas are provided by placing the
samples and antennas appropriately.
We performed relative measurements. The reference level for reflection
measurements is obtained by putting a metal plate (copper) exactly the position at which
76
front face of samples is located, and measuring the reflected signal from this metal plate.
Likewise, reference level for transmission measurements is set by measuring the
transmitted signal when there is no sample present between horn antennas.
Because the output signal of the square-law detector is proportional to S11
2
(or
r ) (assuming higher order terms are suppressed by detector [114]), the reflection
coefficient of samples is obtained as
r =
Vsr
Vm
(5.8)
where Vsr and Vm are the dc voltage values of reflected signal from the sample and metal
plate (reference), respectively. Likewise, transmission coefficient is measured as
t =
Vst
Vf
(5.9)
where Vst and V f are the dc voltage values of transmitted signal through the sample and
in free-space (reference), respectively.
5.5
Measurements
We carried out two different measurements for characterization of cement–based
materials. The discussion of the results for each measurement is presented in two
subsections.
5.5.1
Measurements of Young Mortar Specimens
We present the measurement results of free–space reflection and transmission
coefficients of five mortar specimens with varying mixture contents. We introduced a
77
new approach (relative slope approach for reflection properties) to predict the history of
the hydration (a chemical process that water molecules bond with cement molecules)
process of these specimens. Five mortar specimens are prepared with different w/c ratios
(w/c=0.40, 0.45, 0.50, 0.55, and 0.60). The mix proportions are given in Table 5.1. The
cement used in the experiment is the ASTM Type I normal Portland Cement, 100 % of
the sand mass consists of particles less than 4 mm in diameter. Well–graded, quartzite,
natural sand is used. The sand complies with the requirements of ASTM C–33 [115]. The
absorption capacity is 0.1 % and its specific gravity (relative density) at saturated surface
dry condition is 2.72.
Table 5.1. Mass percentages of raw materials of cement samples.
Mortar
Sand(s) Cement(c) Water (w)
s/c
w/c
I
58.80
29.40
11.80
2
0.40
II
57.96
28.98
13.06
2
0.45
III
57.14
28.57
14.29
2
0.50
IV
56.34
28.17
15.49
2
0.55
V
55.56
27.78
16.66
2
0.60
The thickness of the samples is chosen in such a way that transmission
measurements of young mortar samples can be detected. According to the previous study
[63], the optimum thickness of samples is selected as 100 mm after initial transmission
measurements. The lateral dimensions of the samples are 150 mm x 150 mm.
Temperature and relative humidity for curing conditions were kept between 28–32 Co and
50–60 %, respectively, at ordinary laboratory conditions.
78
The hourly measurement of reflection and transmission properties of five young
mortar specimens with raw material properties in Table 5.1 has been conducted at
X−band by the nondestructive and non–contact free–space measurement set–up in Fig.
5.2. For example, Figs. 5.3(a) and 5.3(b) show the hourly measurements of amplitude of
the reflection coefficient, S11 (or r ), at 8.5 GHz of mortar specimens in Table 5.1.
It can be seen from Figs. 5.3(a) and 5.3(b) that dependency of reflection
properties of each sample is monotonic (decreasing with time), and the higher amplitude
of reflection coefficient corresponds to the mortar sample with higher w/c ratio at early
ages. This is the result of the effect of both evaporation of free–water from the surface
and hydration process [20],[42],[63]. It means that the reflection properties depend on
w/c ratio. These measurement results compare well with the previous results of 2–3 day
cured mortar and concrete samples at room temperature [20],[42],[63].
Since sensing of the material using microwave energy interrogates the dielectric
properties of the material on a molecular level, it can be assumed that the curing process
is complete once the dielectric properties of these chemically active materials become
constant. Therefore, by periodic monitoring of the reflection properties of the specimens
during their curing process, an estimation of the cure–state of the specimens can be
ascertained [42]. In this regard, the slope or time rate change of the reflection properties
can be taken as a metric of the state and speed (or degree) of hydration process [42]. This
is also meaningful if we consider different curing (wet or dry) and environmental
conditions (temperature, humidity, etc.) for the measurements of properties of cement–
samples because the amplitude of reflection properties changes with them.
79
(a)
(b)
Fig. 5.3.
Hourly reflection properties of young mortar samples with different w/c
ratios at 8.5 GHz (a) between 20–30 hours and (b) between 44–54 hours.
80
In this study, we proposed a relative slope approach for reflection properties of
young mortar samples to predict the history of the hydration process of these samples
with equally increased w/c ratios; i.e., ∆w ∆c = 0.05 , and for this purpose monitored the
amplitude of the reflection properties. The proposed approach is reasonable since mortar
can be assumed as a homogenous dielectric mixture at relatively low microwave
frequencies (even at 10 GHz) [42]. The wavelength corresponding to the frequency used
in our measurements (8.5 GHz) is 35.3 mm, which is much greater than the diameter of
the maximum sand particles (less than 4 mm).
It is demonstrated in Table 5.2 that, considering only the samples whose w/c ratio
difference is 0.05, the average absolute temporal amplitude difference in reflection
coefficients for mortar samples I and II is much greater than that of other mortar samples.
This difference can also be seen from Figs. 5.3(a) and 5.3(b). This means a non–linear
relationship between the degree of hydration and w/c ratios for early ages of curing.
Table 5.2. The absolute temporal change of reflection coefficient
of each young mortar specimens.
Sample Number
Hours
20–30
44–54
20–54
I – w c =0.40
0.0300
0.0155
0.0543
II – w c =0.45
0.0296
0.0170
0.0570
III – w c =0.50
0.0325
0.0170
0.0578
IV – w c =0.55
0.0271
0.0152
0.0584
V – w c =0.60
0.0271
0.0135
0.0595
81
From the physical point of view, during curing, evaporation of free–water and
hydration inside samples occur simultaneously. Because the free–water very near to the
cement molecules will yield immediate reaction and hydration, the binding inside each
sample will be different depending upon the amount of free–water. All factors being the
same, the strength of mortar decreases as the w/c ratio increases. A mortar sample with a
lower w/c ratio will gain strength quicker than the one with a higher w/c ratio [19],[42].
From the electrical point of view, it is known that measured reflection properties
for a mixture consisting of a relatively low permittivity of dielectric constituent (cement)
and a high permittivity of dielectric constituent (water) is expected to increase as a
function of increasing water content [116]. In addition, while dielectric properties of
free–water are much higher than those of cement powder, dielectric properties of bound–
water are similar to those of cement powder [117]. Therefore, a considerably lower
amplitude of reflection coefficient of mortar sample I means that a fast chemical reaction
or hydration (an indication of chemical binding) occurred inside it. Therefore, we can
deduce the history (i.e., quickness) of hydration process (and strength) of mortar samples
from our proposed relative slope approach.
While reflection measurements show surface characteristics of samples, the
transmission measurements show the internal structure (central) and provide longitudinal
averaging of variations in sample properties [20]. For example, Figs. 5.4(a) and 5.4(b)
show the result of the hourly measurements of amplitude of transmission coefficient,
S21 (or t ), at 8.5 GHz of mortar specimens in Table 5.1.
It can be seen from Figs. 5.4(a) and 5.4(b) that a higher transmission coefficient is
for the mortar sample with a lower w/c ratio (meaning that transmission properties are
82
w/c ratio dependent) and that there is no intersection point observed between these
dependencies.
In a fresh state of mixture during hardening (setting) time, the heavy ingredients
of the mixture tend to move downward (aggregates) and light ingredients of the mixture
move upward (water), resulting in bleeding (a special form of segregation) and porosity
formation [20],[118]. Pore structure inside cement samples can be thought as an air
medium with a relative permittivity of almost one. This means that microwave signals
will penetrate much easily through cement samples with higher porosity than the ones
with lower porosity.
It is well known that hardened cement sample with a bigger w/c ratio will have a
higher porosity than the one with a smaller w/c ratio [20],[118]–[121]. The reason is the
evaporation of free–water from the surface of samples. On the other hand, since the
transmission properties of the mortar sample with lower w/c ratio are higher than those
with higher w/c ratio at early ages of curing as shown in Figs. 5.4(a) and 5.4(b), we can
conclude that the potential porosity inside the young mortar sample with higher w/c ratio
should be less than the one with lower w/c ratio. This is because free water cannot
immediately escape from the cement mixture at early stages of curing. As a result,
monitoring of transmission properties along with reflection properties can be used for the
quality enhancement of cement samples since early age detection of properties is
important for quality enhancement of these samples [42].
83
(a)
(b)
Fig. 5.4.
Hourly transmission properties of young mortar samples with different w/c
ratios at 8.5 GHz (a) between 20–30 hours and (b) between 44–54 hours.
84
Because measurements were conducted during only day and evening times, there
is no measurement data between 30 and 44 hours (night times). We expect that there are
no significant changes of the dependencies of reflection and transmission properties (the
higher reflection is for the sample with higher w/c ratio, and the higher transmission is for
the sample with lower w/c ratio) since we kept the environmental conditions constant
throughout the measurements. However, as it can be seen from Figs. 5.3(a) and 5.3(b)
that we predict the average relative distance between the slopes of the reflection
properties of mortar samples I and II during 30–44 hours will be greater than the average
distance during 20–30 hours and less than the average distance during 44–54 hours.
It can be seen from Figs. 5.3 and 5.4 that we observed some variations in
reflection and transmission properties. This is because samples are switched for each
measurement step and reflection and transmission properties are surface and integral
content dependent, respectively. It is noticed that the variations seen in reflection
measurements (Fig. 5.3) are greater than those of transmission measurements (Fig. 5.4).
It is known that any misalignment of the aperture of sample (angle deviation) yields
different reflection and transmission properties, and reflection properties are affected
more [20]. It should be noted that the effect of the environment (temperature, humidity,
etc.) is minimized throughout measurements by keeping the environmental conditions
constant. We also assumed that the characteristics of the measurement setup are stable
during measurement.
We used two methods for the ε determination of mortar samples. These methods
are the graphical and Newton’s methods. The graphical method is used because it shows
the general pattern of the dependency of reflection and transmission properties on real
85
and imaginary parts of the ε [20],[26],[74]. For given measured amplitudes of the
reflection and transmission coefficients, we can obtain the constant value lines of these
coefficients, CR and CT [20],[26],[74]. The necessary and sufficient condition for
determining the ε from the measured amplitudes is that there is just one cross point
between the lines CR and CT for high–loss samples. Although the graphical method is
very clear and useful for primary evaluation of the dielectric properties, the speed of this
method is very slow.
We utilized the Newton’s method for ε determination of mortar samples using
the objective function in (5.5) along with (5.6). We set ξ = 0 as initial guess and searched
for the minimum of the objective function. Figs. 5.5 and 5.6 show the ε of young mortar
samples in Table 5.1 during early ages of curing (20–30 and 44–54 hours). It is seen that
the values of real and imaginary parts of the complex permittivity, ε ' and ε "
respectively, of mortar samples are in good agreement with known complex permittivity
values of hardened cement samples obtained from reflection measurements by vector
network analyzer (VNA) [122] and amplitudes–only reflection and transmission
measurements with a simple set–up [20].
It is demonstrated in these figures that a higher w/c ratio corresponds to a higher
imaginary (real) part of the permittivity, ε " ( ε ' ). From the physical point of view, since
at early ages of curing the free water content will be greater than the bound water content
inside mortar samples, a higher content of water will result in bigger losses (or
attenuations) and bigger reflections. Additionally, because ε " represents the loss of
electric field in the sample and ε ' represents the ability of the sample to store energy
86
[123], one can conclude that transmission (reflection) measurements will be mainly
affected by ε " ( ε ' ).
From the mathematical point of view, assuming a high–loss case, it is seen from
(5.5) that the effect of the imaginary part of the complex permittivity on transmission
coefficient is almost directly present in exponential term since the imaginary part of the
square root of a complex number is much more dependent on the imaginary part than real
part of this complex number. This also validates the strong relation between transmission
(reflection) measurements and ε " ( ε ' ) at early ages of curing.
From the empirical point of view, the same conclusions can be made. To this end,
we apply a newly developed relative slope approach for reflection properties to
transmission properties. It was mentioned that the relative difference between reflection
properties of mortar samples I and II are much bigger than any other nearly the same
relative differences in Fig. 5.3. However, it seen from Fig. 5.4 that all relative differences
between transmission properties of mortar samples are almost the same. It is noted that
only the samples whose w/c ratio difference is 0.05 are considered for comparisons. In
addition, it is observed that whereas reflection properties change from one measurement
step to another in Fig. 5.3(a), transmission properties show a monotonic behavior in Fig.
5.4(a). The reason of this change might be the misalignment of the position of surface of
the samples with aperture of antennas (angle deviation). Comparing these measurement
results with calculated ε ' and ε " values in Figs. 5.5 and 5.6, it is realized that ε ' and ε "
follow the change patterns of reflection and transmission properties, respectively.
Therefore, we can conclude that ε ' ( ε " ) is mainly affected by reflection (transmission)
properties at early ages of curing.
87
(a)
(b)
Fig. 5.5.
(a) Real part and (b) imaginary part of the complex permittivities of young
mortar samples with different w/c ratios between 20–30 hours.
88
(a)
(b)
Fig. 5.6.
(a) Real part and (b) imaginary part of the complex permittivities of young
mortar samples with different w/c ratios between 44–54 hours.
89
5.5.2
Measurements of Hardened Mortar and Concrete Specimens
Measurements of reflection and transmission properties from various sides of hardened
mortar and concrete specimens with different w/c ratios are carried out at X–band over 3–
36 months by the free–space set–up in Fig. 5.2. Several cubic mortar specimens with
different w/c ratios and dimensions of 150 mm
× 150
mm
×
150 mm are prepared. The
raw materials of the mortar and concrete specimens are shown in Table 5.3.
Concrete
Mortar
Table 5.3. Mixture proportions of mortar and concrete specimens.
Coarse aggregate
Water/cement
Sand/cement
I
0.4
1.5
/cement
––
II
0.7
1.5
––
I
0.4
1.5
2.0
II
0.7
1.5
2.0
The cement used in the experiment is the ASTM Type I normal Portland Cement,
100 % of the sand mass consists of particles less than 4 mm in diameter. Well–graded,
quartzite, natural sand is used. The sand complies with the requirements of ASTM C–33
[115]. The absorption capacity is 0.1% and its specific gravity (relative density) at
saturated surface dry condition is 2.72. Coarse aggregates have a maximum size of 16
mm and are natural round shaped obtained from the river.
We prepared all specimens with a surface protuberance less than 5 mm by
vibrating each specimen inside its metal mould. We removed all molds after 1 day. Then,
the specimens were left to cure at ordinary room conditions (a temperature and a relative
90
humidity between 28 and 32 oC and 50 and 60 %, respectively). For comparing the side
effects of specimens on reflection and transmission properties, we, respectively, labeled
the bottom, right, top, and left sides of specimens with numbers from 1 to 4.
At early times of curing, the transmission properties are too low to measure
accurately. Therefore, we present transmission properties of mortar and concrete
specimens after 3 months. For example, Fig. 5.7 shows the results of monthly
measurements of S21 from right sides of mortar and concrete specimens with different
w/c ratios at 10 GHz. For easy reference in Figs. 5.7–5.10, whereas ‘M’ and ‘C’ in the
figure legend refer to mortar and concrete specimens, the numbers from 1 to 4 denote
sides of the specimens, and ‘I’ and ‘II’ mean w/c=0.4 and 0.7, respectively.
Fig. 5.7.
Results of transmission properties from right sides of mortar and concrete
specimens with different w/c ratios.
It is seen from Fig. 5.7 that the higher w/c ratio results in bigger transmission
properties of mortar (concrete) specimens. The reason is the evaporation of free–water
91
from the surface of samples and formation of pore structure inside samples. It is well
known that a hardened cement sample with a bigger w/c ratio will have a higher porosity
than the one with a smaller w/c ratio [20],[63],[118]–[121].
A crossover point was observed between transmission properties of the same type
specimens with different w/c ratios at 5th days of curing [20]. We did not observe any
crossover point between transmission properties of the specimens with different w/c=0.4
and 0.7 ratios during 3–36 months of curing.
To monitor the state and degree of aging (curing) inside hardened specimens, we
developed a relative difference approach. It uses temporal differences between
transmission properties of the specimens and can be applied to: a) the same type
specimens with different w/c ratios and b) different type specimens with the same w/c
ratio. While a temporal difference demonstrates the state of aging, its tangent relates the
degree of aging of specimens. The application of the technique to the same type and
different type specimens using the measurements in Fig. 5.7 is shown in Figs. 5.8(a) and
5.8(b), respectively.
92
(a)
(b)
Fig. 5.8.
Application of the relative difference approach to (a) the same type
specimens and (b) different type specimens.
The temporal dependencies in Fig. 5.8 are investigated as three separate regions.
These regions, respectively from first to third, are labeled with plot symbols ‘○’, ‘□’, and
‘∆’. The main conclusions drawn from the measurements in Fig. 5.8 are the following:
93
1) Regions–I and II correspond to periods of aging for specimens whereas Region–III
refers to the end of aging since temporal differences change throughout Regions–I and II
and are stable and monotonic (constant) during Region–III. It is seen from Fig. 5.8(a) that
the aging periods for mortar and concrete specimens are during 3–30 and 3–24 months,
respectively. The difference between aging periods comes from the fact that gravel
particles are heavier in mass than sand particles and force free–water to escape from the
open surfaces of the specimens. As a result, concrete specimens will complete or finalize
the aging process sooner than mortar specimens with the same w/c ratio.
2) For Region–I in Fig. 5.8(a), both dependencies significantly increase for mortar
specimens from 0.076 to 0.138 (an increment of 0.062) and for concrete specimens from
0.038 to 0.121 (an increment of 0.083) during 3–9 months. This means that tangents of
the dependencies are both positive for Region–I. The increments in the dependencies are
because the evaporation or desiccation of free–water inside the mortar (concrete)
specimen with a higher w/c ratio is faster than that inside the same specimen with a lower
w/c ratio. As opposed to this, during Region–I in Fig. 5.8(b), both dependencies decrease
throughout 3–6.1 months. This time, the effect is that the evaporation of free–water inside
a concrete specimen is faster than that inside a mortar specimen with the same w/c ratio.
Accordingly, a concrete specimen with a higher w/c ratio will lose free–water much faster
than other specimens with different w/c ratios in Region–I. The reason for the increments
and decrements in the dependencies is, as discussed in 1), the difference in mass
proportions of gravel and sand particles and the w/c ratio.
3) During Region–II, we observed smaller decreases for mortar specimens from 0.138 to
0.131 (0.007 difference) and for concrete specimens from 0.121 to 0.118 (0.003
94
difference) during 9–30 months in Fig. 5.8(a). In contrast, a small increase in the
dependencies is recorded for the ‘M.II.2–C.II.2’ dependency from 0.024 to 0.034 (0.01
difference) and for the ‘M.I.2–C.I.2’ dependency from 0.006 to 0.021 (0.015 difference)
during 6.1–30 months. Therefore, a mortar specimen with lower w/c ratio will drop free–
water sooner than other specimens with different w/c ratios in Region–II. This can be
because of the continuing evaporation of free–water inside mortar specimens as contrary
to decreasing evaporation of concrete specimens.
4) We note that the conclusions made for the temporal transmission properties for the
right and left sides of mortar and concrete specimens are nearly the same. Only a small
difference (<3%) between the dependencies is examined. This demonstrates the integrity
and homogeneity of prepared specimens.
5) The temporal transmission properties for the top and bottom sides of mortar and
concrete specimens are the same. This clearly proves that the specimens under
investigation are inactive materials [14].
6) The magnitude of transmission properties for the top (or bottom) is less than those for
the right (or left) side for each measurement step. This is due to the effect of gravel and
sand present inside the specimens after hardening (setting) [63],[118]. Because
microwave signals will scatter from the aggregates (scatterer) which moved to the bottom
of the specimens during setting, lower transmission properties were measured for
specimens with top side is located toward to the microwave signal.
7) While the patterns of the dependencies of transmission properties for the same type
specimens (different w/c ratios) stay the same for the top and right sides, the difference
between the dependencies of these properties for the top side is more pronounced than
95
those for the right side. This is because of greater scattering effect of gravel particles than
that of sand particles.
8) We did not record any significant change between the dependencies for different type
specimens (the same w/c ratios) for top and right sides. The reason for this is
approximately the same mixture proportions of gravel and sand particles inside mortar
(and concrete) specimens with different w/c ratios. As shown in Table 5.3, because w/c
ratios of the specimens are much lower than sand–to–cement (s/c) and coarse aggregate–
to–cement (ca/c) ratios, the same type specimens will approximately have the same
gravel and sand particles.
Whereas transmission properties of mortar and concrete specimens change over
long service lives and become constant after approximately 30 months, reflection
properties of these specimens vary during 1 month after preparation and stay constant
since then. Reflection properties of these specimens were analyzed in the literature
[20],[42],[47],[63]. Table 5.4 demonstrates measured reflection properties of prepared
samples after 3 months of curing.
Table 5.4. Reflection properties from various sides of mortar and
Concrete
Mortar
concrete specimens with different w/c ratios after 3 months of curing.
Right
Left
Top
Bottom
I
0.51
0.50
0.49
0.52
II
0.46
0.45
0.44
0.49
I
0.48
0.47
0.46
0.54
II
0.45
0.44
0.43
0.52
96
The main findings from reflection measurements in Table 5.4 are as follows:
1) The difference between reflection properties from the right (or left) side of mortar
specimens with different w/c ratios is greater than that of the concrete specimens. This is
because of the fact that cement proportions inside mortar specimens are much greater
than those inside concrete specimens because of gravel and the same s/c ratio. While
hydration occurs, cement and water molecules chemically combine with each other. This
process greatly changes the permittivity of specimens [42]. Because reflection properties
are affected by the permittivity changes, we can conclude that the higher cement content
inside mortar and concrete specimens the bigger reflections from the surface of these
specimens while other ingredients are kept approximately the same.
2) Reflection properties from the bottom of mortar and concrete specimens are greater
than those from other sides of the same type specimens for each w/c ratio. This is because
of the scattering effect of aggregates. While hardening (or setting), heavy ingredients
inside specimens move downward and stays there. Because reflection properties are
measured after hardening, not only hydration but also scatterers (aggregates) affect
microwave reflection properties.
3) Reflection properties from the top of mortar and concrete specimens are less than those
from other sides of the same type specimen for each w/c ratio. This can be explained
using the relative permittivity values of pore and aggregates. While hardening (or
setting), heavy aggregates force lighter ingredients (free–water molecules) to the top.
This force creates pore structures inside mortar and concrete specimens after water
evaporates from the sides of the specimens. Because pore structures inside cement
97
samples can be thought as air medium with a relative permittivity of one, microwave
signals will penetrate inside samples easier from the top.
We utilized the Newton’s method for ε determination of mortar and concrete
specimens using the objective function in (5.5) along with (5.6). We set ξ = 0 as initial
guess and searched for the minimum of the objective function. Figs. 5.9 and 5.10
demonstrate, respectively, the real and imaginary parts of the ε of mortar and concrete
specimens with different w/c ratios and sides using the measurements in Fig. 5.7 and
Table 5.4.
It is seen from Figs. 5.9 and 5.10 that whereas ε ' stays approximately constant,
ε " decreases over 3–36 months for all specimens. Because reflection measurements were
constant and transmission properties increased over 3–36 months and ε ' represents the
ability of the substance to store energy [123], the results in Figs. 5.9 and 5.10 clearly
demonstrate a strong relation between the pattern of reflection properties and ε ' of mortar
and concrete specimens [20]. In other words, the higher the reflection properties the
bigger ε ' . In addition, since ε " (loss factor) represents the loss of electric field energy in
the substance [123], the higher transmission properties correspond to a lower loss factor.
Therefore, we also observed a correlation between the change patterns of transmission
properties and ε " of specimens [20].
98
(a)
(b)
Fig. 5.9.
The real part of the ε of mortar and concrete specimens with different w/c
ratios and sides.
99
(a)
(b)
Fig. 5.10. The imaginary part of the ε of mortar and concrete specimens with
different w/c ratios and sides.
In the light of the conclusions made in this subsection, we can extract some useful
information about the reflection and transmission properties of mortar and concrete
specimens using the information present in Figs. 5.9 and 5.10. For example, the highest
and the lowest ε ' of mortar and concrete specimens correspond to the reflection
100
properties from top and bottom, respectively, for each w/c ratios. As another example,
because it is well known that the specimens with higher w/c ratio contain pore structures
much more than those with a lower w/c ratio and these structures can be thought as air
medium inside specimens, the smallest ε " corresponds to mortar or concrete specimens
with the highest w/c ratio for a given side (top, bottom, or etc.).
It can be seen from Figs. 5.3, 5.4, 5.7, and 5.8 that we observed some variations in
reflection and transmission properties. This is because samples are switched for each
measurement step and reflection and transmission properties are surface and integral
content dependent, respectively. It is noticed that the variations seen in reflection
measurements (Fig. 5.3) are greater than those of transmission measurements (Figs. 5.4,
5.7, and 5.8). A detailed discussion about these variations can be found in subsection
5.5.1.
5.6
Measurement Uncertainties
The reliability of reflection and transmission measurements depends on measurement
uncertainties. Although we developed a calibration technique to eliminate some of
systematic errors, it does not consider MR between the measurement cell and antennas,
between antennas, and inside the sample, deviations from the characteristics of
directional and square–law detectors, and surface roughness.
It is assumed in section 5.2 that the sample has enough attenuation so that MR
between the two surfaces of the sample can be neglected. To validate this assumption for
measurements of young mortar samples (Figs. 5.3 and 5.4), the dependencies of
reflection and transmission (attenuation) properties for lossless sample case presented by
101
(5.1) and (5.2) versus sample thickness, L , are drawn. For example, Fig. 5.11 illustrates
such kind of dependencies for the w/c=0.40 mortar sample with ε = 21 − j 0.69 at 54
hours of curing in Fig. 5.6.
Fig. 5.11. Dependencies of reflection and transmission properties of w/c=0.40 mortar
sample with ε = 21 − j 0.69 on sample thickness, L .
The dB values for reflection and attenuation properties in the curves in Fig. 5.11
are obtained
R ( dB ) = −20log10 S21 ,
(5.10)
T ( dB ) = −20log10 S21 ,
(5.11)
where R and T , respectively, denote the reflection and attenuation properties in dB.
Because L in (5.1) and (5.2) is expressed in complex exponential terms, we
expect an oscillatory behavior of reflection and transmission properties with thickness for
102
lossless samples. The oscillations in these dependencies demonstrate MR inside the
sample because a small change in thickness results in a large alteration in the
s
dependencies. It was stated in section 5.2 that for high–loss samples (with no MR) S11
is
independent of L . Therefore, a small change in thickness will not result in an oscillatory
behavior for reflection properties for high–loss samples. The same situation is true for
s
S21
since it is dependent upon Γ , which is a function of only ε , and a linear term


2k0ξ L
.
exp 
 (1 + χ )2 + ξ 2 


(5.12)
This linear term will yield a constant value for a given L (not an oscillation). Therefore,
the oscillatory behavior of the dependencies of S11 and S21 over L for lossless
samples will turn to a monotonic behavior for high–loss samples.
It is seen from Fig. 5.11 that there are large oscillations in both transmission
(attenuation) and reflection properties in the 0–8 cm thickness range. Beyond this
thickness, these oscillations decrease with increasing thickness, and their behaviors start
to be monotonic. Therefore, whereas 0–8 cm thickness range corresponds to a lossless (or
low–loss) behavior, thicknesses greater than 8 cm result in a high–loss behavior. In
practice, MR inside samples can be neglected if the sample shows at least 10 dB
attenuation [20],[26],[123]. It is seen in Fig. 5.11 that the selected thickness ( L =100 mm)
demonstrates approximately 15 dB attenuation. The same monotonic behavior of
reflection and transmission dependencies of w/c=0.4 young mortar sample at 54 hours of
curing is observed for the same sample at other measurement steps and the other mortar
samples at all measurement steps. As a result, we conclude that the selected thickness of
mortar samples shows at least 10 dB attenuation and assures the correctness of complex
103
permittivity computations using (5.5). It is noted that determined ε can be refined, if
needed, using that ε as an initial guess for the amplitudes of (5.1) and (5.2) in a two–
dimensional search algorithm [75].
We can monitor the dependencies of R and T in (5.10) and (5.11) for validation
of the determined ε of hardened mortar and concrete specimens in Figs. 5.9 and 5.10.
Instead, we apply another technique to demonstrate the validation of this assumption that
s
s
S11
and S21
can, respectively, be approximated to S11 and S21 (or MR inside the
specimens can be neglected for our measurements). In the section 5.2, we utilized
Z 2 <<1 for simplifying the expressions in (5.1) and (5.2). This condition can be written
in an explicit from using (5.3) and (5.7) as


4k0ξ L
 << 1.
exp 
 (1 + χ )2 + ξ 2 


(5.13)
The dependency of (5.13) versus L is used for correctness of the determined ε in
Fig. 5.9 and 5.10. It is clearly seen from (5.13) that for a known f both χ and ξ are in
effect. It is noted from (5.7) that ξ is highly dominated by ε " . Because determined ε '
stays approximately constant in Fig. 5.9 and ε " changes with time in Fig. 5.10,
monitoring the effect of ξ on (5.13) will help us validate the foregoing assumption. It
was proven in chapter 3 that χ < 0 and ξ < 0 . Since it is easy to satisfy (5.13) when ξ is
large for a constant χ , we have to take into consideration the worst case (the smallest
ξ ) in order to prove the foregoing assumption for all ξ values.
From Figs. 5.9(a) and (b), the lowest and the highest ξ values are, respectively,
obtained from the mortar specimen with w/c=0.7 and from the concrete specimen with
104
w/c=0.4 after 36 months of curing because ξ is proportional to ε " . For example, Fig.
5.12 shows the dependency of Z 2 over L for ε = 6.95 − j 0.22 and ε = 11.18 − j 0.60 .
Fig. 5.12. Dependency of Z 2 versus specimen thickness, L , for ε = 6.95 − j 0.22
and ε = 11.18 − j 0.60 .
It is clearly seen from Fig. 5.12 that the dependencies of Z 2 decrease with L . At
L =150 mm, Z 2 =0.075 for ε = 6.95 − j 0.22 and Z 2 =0.0034 for ε = 11.18 − j 0.60 .
Because Fig. 5.12 shows the dependencies for the lowest and the highest ξ , the
dependencies of Z 2 for other hardened mortar and concrete specimens will lie between
the curves in Fig. 5.12. As a result, we can conclude that MR inside the specimens can be
neglected for all prepared hardened specimens.
Since MR between antennas depend upon MR between sample surfaces
(assuming this effect is already omitted), its effect can also be neglected. Because relative
105
measurements are performed, the effects of MR between antenna and sample and the
apparatus uncertainties are already small and/or eliminated.
Another effect might be the finite directivity of the directional coupler used in the
measurement set–up in Fig. 5.2. Finite directivity reflects a possibility of a linkage
between the forward and reverse ways of a directional coupler [14]. The oscillator signal
coupled into the detector output for reflection measurements (because of the finite
directivity of coupler more than 40 dB) shows that its value is much small compared with
the reflected signal value and can be neglected [14].
On the other hand, the accuracy of S11 and/or S21 measurements might be
reduced by the deviations from the square–law characteristics of the crystal diode
detectors used in our set–up. It is well known that the square–law characteristic may be
altered if the level of measured voltage value is too high. Considering this, we selected
the thickness of the young mortar samples and hardened mortar and concrete samples in
such a way that both the level of transmission measurements is in the range of crystal
diode characteristics and the levels are high enough to be distinguishable for different w/c
ratio samples.
Finally, S11 and S21 in (5.1) and (5.2) account only for specular reflection, which
occurs for smooth surfaces. When the surface is rough, the impinging energy will be
scattered in angels other than the specular angle of reflection, thereby reducing energy in
the specularly reflected and transmitted components. As a result, the surface protuberance
might change reflection and transmission measurements, and its effect must be analyzed.
The modified S11 , which takes into account the surface roughness, can be written as [67]
r
S11
= ρ s S11 ,
106
(5.14)
where ρ s is the correlation factor and the letter ‘r’ at the superscript of S11 in (5.14)
denotes the expression for surface roughness. The expression of ρ s is given as [67]

 πσ h cosθi 

λ0


ρs = exp  −8 

2
,


(5.15)
where σ h is the standard deviation of the surface height in the first Fresnel zone of the
illuminating antenna and θi is the angle of incidence.
The Rayleigh criterion is commonly used as a test for surface roughness (critical
height) of materials [67] as
hc =
λ0
8cosθi
.
(5.16)
The height, h , of a given rough surface is defined as the minimum to maximum surface
protuberance. The surface is considered smooth if h < hc and rough if h > hc . It can be
seen from equation (5.16) that hc reaches minimum value for a given wavelength λ0 at
θi = 0 .
The minimum value of hc in (5.16) is between 4–5mm at X–band. In practice,
cement–based samples have surface roughness with heights less than 5 mm, and thus can
be assumed that they are smooth [20]. In addition, most cement–based structures have
surfaces relatively having larger curvatures so that at X–band they are locally quite
smooth. As a result, for elaborately prepared samples the surface protuberance effect can
be omitted.
107
CHAPTER 6
A SIMPLE AND RELATIVELY INEXPENSIVE MICROCONTROLLER–BASED
MICROWAVE FREE–SPACE MEASUREMENT SYSTEM FOR INDUSTRIAL–
BASED APPLICATIONS
Microwave measurements lend themselves to automation because many readings and
considerable computation are often needed to achieve reasonable accuracies. Therefore,
microwave measurements, especially broadband microwave measurements, can be made
considerably faster and more accurately when they are automated [124]. Automated
systems have been demonstrated to provide some principal advantages over manual ones
in microwave measurements:
a) Increased speed
b) Computational enhancement
c) Repeatability
d) More complex measurements
Any measurement system that would benefit from any of the above advantages is
a likely candidate for automation [124]. The introduction of inexpensive microprocessors,
the availability of programmable sources and measurement instruments have made it
possible to build custom automated network analyzers automated network analyzers at a
substantially lower cost than commercial system purchase price [125].
108
In this chapter, we will present a microcontroller based microwave free–space
(free–wave) measurement system for electrical characterization (complex permittivity, ε ,
determination) of lossy materials. This system is very suitable for industrial–based
applications since the total cost to construct a microwave measurement system for
electrical characterization of lossy materials is kept at a minimum. It is composed of a
microwave section and an electronic section. While the microwave section measures the
amplitudes of reflection and transmission measurements of samples, the electronic
section calculates the ε using those measurement data. In this chapter, we also present a
calibration procedure, which not only takes into account of diffraction effects from the
sample edges but also undesired reflections from the surrounding environment, multiple–
path reflections from the ground, and the losses in the measurement system. We validate
the whole measurement system by amplitude only measurements of a fresh sample,
which is poured into a wooden container, at 10 GHz.
6.1
The Measurement System
The proposed microcontroller–based microwave free–space measurement system is
shown in Fig. 6.1. It measures the amplitudes of reflection and transmission coefficients
( r ( S11 ) and t ( S21 )) at discrete frequencies and computes the complex permittivity,
ε , of lossy sample (or a measurement cell which contains or holds the sample under test)
using a general purpose microcontroller (µC). It can be separated into two main sections:
a) microwave section and b) electronic section. Since the microwave section is already
discussed in chapter 5, we will only describe the electronic section.
109
Fig. 6.1.
The schematic diagram of the microcontroller–based microwave free–space
measurement system.
The electronic section gathers and processes measurement data to determine ε
and outputs the result. It consists of a µC, an analog–to–digital converter (ADC), an
electrically erasable programmable read–only memory (EEPROM), a liquid crystal
display (LCD), and a printer (Fig. 6.1). The circuit board of the electronic section is
shown in Fig. 6.2.
The µC commands every operation in this section. We employed a multi–purpose
8 bit AT89S8252–24PI µC from ATMEL for our project. It has a 8K Bytes flash
memory, a 2K Bytes EEPROM, and a 24 MHz internal crystal frequency. It is a powerful
110
µC, which provides a highly–flexible and cost–effective solution to many control
applications. An 18.432 MHz external crystal frequency is used to operate this µC.
Fig. 6.2.
The circuit board of the electronic section.
Since the microwave detector output is analog, it can readily be used as an input
signal to any ADC. In our preliminary studies, we employed an 8 bit ADC (ADC 0808)
for converting the analog signal from the microwave section into a digital signal for
processing in the µC [79]. Then, we added amplifiers between the microwave section and
the µC as a buffer to increase the sensitivity [78]. However, since the resolution or the
accuracy of an 8 bit ADC is not suitable for precise ε determination, in this dissertation,
we employ a 12 bit LTC1298 ADC to process the data. Its operating range is adjusted
between 0 and 1000 mV. This sets a resolution or a quantile level less than 250 µV.
111
We utilized an external 24LC512 EEPROM (512 Kbit) to save and transport the
measurement data for post–processing. Chip select and clock digital signals are used to
access this EEPROM by the µC, and a serial transfer is utilized to acquire/write the data
(Fig. 6.1). The LCD from Hantronix has a 16x2 character display (two lines where each
line can show at most 16 characters). An enable and a read/write digital signals are used
to control it by the µC (Fig. 6.1).
A general purpose HP 6 MP Laserjet printer is connected for printing r , t , and
ε . A common interface standard (centronics) is utilized to communicate with the printer
and to print the results. We employed this standard because the majority of printers use
this basic handshaking procedure. The computed ε can be saved into the internal 2KB
EEPROM of the µC, stored into the external EEPROM, demonstrated on the LCD
display, or printed using the printer.
6.2
Calibration Procedure
The microwave source of the measurement set–up in Fig. 6.1 was operated at least two
hours before measurements to stabilize source power. Care was taken to minimize the
interference to measurements. For the set–up in Fig. 6.1, we can write
Y (ω ) = U (ω ) H (ω ) .
(6.1)
Here, U (ω ) , H (ω ) , and Y (ω ) correspond to the expected or theoretical signal in
the frequency domain, the system transfer function, and the measured signal in the
frequency domain. We applied a calibration technique similar to the Thru–Reflect–Line
(TRL) technique [93]. Its schematic diagram is shown in Fig. 6.3.
112
Fig. 6.3.
The schematic diagram of the calibration technique.
Error terms denoted by Ed (ω ) , E pf (ω ) , E pr (ω ) , EDf (ω ) , EDr (ω ) , and
E f (ω ) correspond to errors of directivity, forward and reverse port match, forward and
reverse diffraction, and frequency response (tracking), respectively. The r and t are
the actual amplitudes of reflection and transmission properties of specimens. The error
terms EDf (ω ) and EDr (ω ) include not only diffraction effects from sample edges but
also undesired reflections from the surrounding environment, multiple–path reflections
from the ground, and the loss in the measurement system. In the analysis, firstly it is
assumed that E f (ω ) = 1 , E pf (ω ) = 0 , and E pr (ω ) = 0 throughout the frequency domain.
Before finding the error terms, the transmit and receive antennas are separated as
follows: a) the distance between them fulfills the plane wave condition, b) the center of
the samples is exactly matching the center of horn antennas, and c) maximum incident
signal is received when there is no sample between antennas to increase the sensitivity of
the measurement set–up. All these criteria were discussed in detail in Section 5.
113
Next, a highly reflective metal plate (copper) having the same cross section and
surface properties as of prepared specimens is placed at the middle distance of the
antennas, and amplitudes of incident– I m (ω ) , reflected– R m (ω ) , and transmitted– T m (ω )
signals of the metal plate are measured. Using the diagram in Fig. 6.3 and (6.1), we can
write
Γ m (ω ) H r (ω ) = R m (ω ) − EDf (ω ) EDr (ω ) − Ed (ω ) ,
(6.2)
Z m (ω ) H t (ω ) = T m (ω ) − EDf (ω ) ,
(6.3)
where Γ m (ω ) , Z m (ω ) , H r (ω ) , and H t (ω ) are the theoretical amplitudes of reflection
and transmission coefficients for the metal plate, and transfer functions for reflected and
transmitted signals of the system, respectively. For each frequency of interest, we set
Γ m (ω ) = 1 and Z m (ω ) = 0 . Accordingly, we obtain
H r (ω ) = R m (ω ) − EDf (ω ) EDr (ω ) − Ed (ω ) ,
(6.4)
T m (ω ) = EDf (ω ) .
(6.5)
Then, we removed the metal plate between antennas and measured amplitude of
the incident signal in free–space, I f (ω ) . Since the level of the incident signal changed
because of the mismatch produced by removing the plate, its level is firstly returned to
I m (ω ) using a rotary attenuator. Afterwards, we measured reflected– R f (ω ) and
transmitted– T f (ω ) signals in free–space. Using the diagram in Fig. 6.3 and (6.1), we
can find
Γ f (ω ) H r (ω ) = R f (ω ) − EDr (ω ) T f (ω ) − Ed (ω ) ,
(6.6)
Z f (ω ) H t (ω ) = T f (ω ) − EDf (ω ) ,
(6.7)
114
where Γ f (ω ) and Z f (ω ) are the theoretical amplitudes of reflection and transmission
coefficients for free–space, respectively. For each frequency of interest, we set
Γ f (ω ) = 0 and Z f (ω ) = 1 . Consequently, we find
EDr (ω ) =
R f (ω ) − Ed (ω )
,
(6.8)
H t (ω ) = T f (ω ) − EDf (ω ) .
(6.9)
T f (ω )
To obtain the calibrated r and t , we have to determine the transfer functions of
the system, H r (ω ) and H t (ω ) . We already know H t (ω ) from (6.5) and (6.9). However,
the determination of H r (ω ) requires another step. To this end, first, we determined
Ed (ω ) by measuring the amplitude of reflected signals when a matched waveguide load
is connected to the terminal of the waveguide section where the transmit antenna is
connected in Fig. 6.3. Next, we calculated EDr (ω ) from (6.8). Then, we determine
H r (ω ) from (6.4) using EDf (ω ) , EDr (ω ) , and Ed (ω ) .
Finally, we placed measurement cells and measured amplitudes of the incident–
I s (ω ) , reflected– R s (ω ) , and transmitted– T s (ω ) signals of them. We also set I s (ω )
equal to I m (ω ) by using the attenuator. Since measured signals from a square–law
detector are proportional to square of the reflection (transmission) coefficient, α r 2
( α 2 t 2 ) where α ( α 2 ) is a constant of proportionally, the calibrated r and t are
obtained by substituting H r (ω ) and H t (ω ) as
115
12
 R s (ω ) −  EDf (ω ) EDr (ω ) + Ed (ω )  


r (ω ) =  m
 R (ω ) −  EDf (ω ) EDr (ω ) + Ed (ω )  



,
(6.10)
12
 T s (ω ) − EDf (ω ) 

t (ω ) =  f
 T (ω ) − EDf (ω ) 
.
(6.11)
Because we divide amplitudes of measured signals in (6.10) and (6.11), the
constant proportionalities will cancel each other. This means that we do not need to
determine them. Since we leveled the source to produce constant incident signals for
reflection and transmission measurements, we can assume that E pf (ω ) = 0 and
E pr (ω ) = 0 in the signal block diagram in Fig. 6.3. In addition, we also utilized a
matched waveguide load connected to the waveguide section to which the receive
antenna is connected to realize E pr (ω ) = 0 . Finally, we can presume that E f (ω ) = 1 by
using constant incident power for each frequency of interest.
6.3
Algorithm
The complex scattering (S–) parameters derived for the five–layer structure in chapter 3
will be utilized in this chapter for ε determination of lossy materials by the µC based
free–space measurement system. To apply the derived objective function, f (ξ ) , in (3.50)
to the our problem in this chapter, we let λc → ∞ .
Our goal is to code a program into the µC which determines ε by using r and t
at one frequency. While writing the program into the µC, we paid attention on three
important facts. These are: 1) the implementation of the square–root and exponential
functions in (3.34) and (3.38) for a wide range of values, 2) increasing the speed of
116
iterations (code optimization), and 3) organizing the variables to occupy less space
(memory optimization).
It is proven in chapter 3 that for a lossy medium 3, −1 < χ ≤ 0 and −1 < ξ ≤ 0 .
However, in some cases, while performing the computations for ε , the square–root in
(3.38) may be negative which results in an imaginary χ . This situation adversely affects
computations in µCs since they do computations using real numbers. As a result, to
resolve this problem, we propose two solutions depending on the applied numerical
method. As a first solution, for numerical methods which require two initial guesses for
ξ such as bisection and secant methods [75], ξ1 = 0 and ξ 2 = ξ min can be used where
ξ min is obtained by iterating ξ 2 = −1 until the first real χ is achieved. As a second
solution, for numerical methods which only necessitate one initial guess such as
Newton’s method [75], it is observed that ξ3 = 0 initial guess accurately determines the
correct ξ .
We coded three programs using different one–dimensional root–finding numerical
methods (bisection, secant, and Newton’s methods) based on the objective function in
(3.50). The simplified algorithm for each program is shown in Fig. 6.4.
The programs firstly acquire measurement data from the microwave section, and
then assign a suitable domain for the solution using ξ1 = 0 and ξ 2 = ξ min (for bisection
and secant methods) or using ξ3 = 0 (for Newton’s method). Next, the tolerance for
computations, TOL , and the total number of iterations, N p , for evaluation of ε are set.
Each program halts whenever a) N p is reached or b) f (ξ ) ≤ TOL . Finally, ε is computed
using (3.51) and (3.52) and the result is shown.
117
Fig. 6.4.
The simplified algorithm for each program.
To implement the exponentials and square roots in (3.34) and (3.38), Taylor series
expansions to the degree N for the exponential and to the second degree for the square
root are used. They are
N Xn
eX = ∑
,
n = 0 n!
M 
1  s2 + X
X = ∑  sn =  n−1
2  sn−1
n = 0 

 ,


(6.12)
where X , n , s−1 , N , and M are the value of which the exponential or square root is
evaluated, index number, initial guess for the square–root, and total iteration numbers,
respectively. Because N , M , N p , and TOL set the resolution or accuracy of
computations, we firstly computed the exponential and square–root using (6.12) for
118
different values of N and M . It is observed that N =50 and M =50 granted highly
accurate results. Next, for precise computations we set TOL = 10−4 . Then, we assigned
N p =50 after some computations using different frequency, f , and sample thickness, L ,
( N =50 and M =50).
Each program is coded using a private basic compiler (BASCOM) from MCS
Electronics. This compiler can perform calculations using a wide range of variable set.
The set can change from an 8 bit unsigned variable to a 32 bit signed variable. We
obtained HEX files after compiling the programs by this compiler. Then, these files are
transferred into the µC using a serial programming mode. It is clear from (3.34)–(3.38)
that the determination of ε in the µC is a difficult task. AT89S8252–24PI µC has only
8K Bytes Flash Memory, and we have to allocate this limited memory for variables and
calculations. To evaluate the written program in BASCOM, we show the compiler results
(the memory footprint of the variables, total ROM memory space allocated, and other
specific information such as crystal frequency, ROM start, and RAM start) of the written
program which uses Newton’s method for the ε determination in Fig. 6.5. It is assumed
that the thicknesses and complex permittivities of left and right dielectric plugs are equal,
and λc → ∞ .
119
Fig. 6.5.
The memory footprint, total ROM space allocated, and some information
related to the program, which uses Newton’s method for ε , written in
BASCOM.
It is clearly seen from Fig. 6.5 that the total allocated memory space for
computations and variables is approximately 8K Bytes, which is the maximum available
space in the µC Flash Memory. This clearly shows the difficulty of coding a program into
the µC. We observed that the total allocated space for the programs, which use bisection
and secant methods, is much less than that for the program which utilizes the Newton’s
method. This is because the Newton’s method requires the derivative of the objective
function, f (ξ ) , for computations of ε while bisection and secant methods do not [75].
120
6.4
Results
6.4.1
Validation of Programs and the Microwave Section
Before carrying out measurements, we performed some computations for validation of
programs coded into the µC. The permittivity results obtained from the programs for a
three–layer structure (air–sample–air) are compared with those of the papers [66],[79]
using the measured R and T at f =57.5 GHz of a concrete sample with L =30 mm for
different times of its aging in [66]. We included the determined ε from [66] as a
reference data for comparison. Here, R and T are the dB values of r and t and are
expressed as T = −20log t and R = −20log r . Since we use a three–layer structure for
validation of the programs, we employ the objective function in (5.5) and (5.6). The
results of the comparison are given in Table 6.1.
Table 6.1. The comparison of the permittivity results.
Measurements
Complex permittivity, ε
T
R
[66]
[79]
µC
73.6
6.44
7.73–j1.28
7.76–j1.27
7.76–j1.27
68.4
6.83
6.98–j1.11
6.97–j1.12
6.97–j1.12
46.5
7.15
6.49–j0.71
6.51–j0.72
6.51–j0.72
28.1
7.19
6.49–j0.42
6.49–j0.42
6.49–j0.42
It is seen from Table 6.1 that results of the µC programs are in good agreement
with those in [66],[79]. Because all µC programs compute approximately the same ε 3 ,
we demonstrated only one result in Table 6.1 as a representative for all programs. To
validate computed ε by all µC programs, we firstly substituted them into (5.1) and (5.2)
121
and then computed R and T . Next, we compared the measured R and T in Table 6.1
and computed Rc and Tc by all µC programs. The result is shown in Table 6.2.
Table 6.2. The comparison of the measured R and T from [66] and
computed Rc and Tc by all µC programs.
Measurements
Computed Results
Comparison (%)
T
R
Tc
Rc
∆T
∆R
73.6
6.44
73.547
6.449
0.072
0.149
68.4
6.83
68.382
6.844
0.027
0.207
46.5
7.15
46.086
7.152
0.892
0.030
28.1
7.19
27.710
7.183
1.387
0.101
It is seen from Table 6.2 that the measured R and T are computed Rc and Tc are
very good agreement with each other. There are two important results we can conclude
from the results in Table 6.2. First, the simplified expressions of S11 (or r ) and S21 (or
t ) in (5.4) can be approximated to those in (5.1) and (5.2) provided that the sample under
test indicates at least 10 dB attenuation ( T >10 dB). It is seen from Table 6.1 that the
sample shows at least 10 dB attenuation during all its aging period (from 1 week to 14
months). Second, the programs written in µC are very accurate (less than 2 % difference
between measured and computed values in Table 6.2) and thus can readily be used in
industrial−based applications whenever the accuracy and low–cost are key concerns.
We discussed the accuracy and drawbacks of the microwave section of the set−up
in Fig. 6.1 in chapter 5. We also validated that section in chapter 5 by amplitude–only
measurements of reflection and transmission properties of young mortar and hardened
mortar and concrete specimens with different mixture contents.
122
6.4.2
Validation of Whole Measurement System
We tested the whole measurement system using a fresh cement paste sample with a 0.375
water–to–cement ratio (w/c). Cement paste is comprised of the mixture of the cement
powder and water. High–density particle boards are used as container for holding the
fresh cement paste sample. The boards have a cross section of 150 mm x 150 mm and a
thickness of L2 = L4 ≅ 11 mm. The complex permittivity of the boards is assumed
2.80 − j 0.20 at 10 GHz [126]. The distance between two inner container walls is L3 ≅ 30
mm. The cement paste mixture with different w/c=0.375 is prepared as follows. Firstly, in
order to eliminate any possible effect between cement mixtures and wooden containers, a
nylon thin film is placed into the container. Then, the prepared cement mixture is vibrated
to achieve homogenous internal structure. Next, it is poured into the container, and the
container is vibrated so that there will be no gap between the container walls and the
mixture. This process is continued until the mixture covers the top of the container.
We calibrated the whole set−up in Fig. 6.1 using the calibration procedure in
section 6.4 before measurements. Then, amplitude–only measurements of the container
with the cement paste sample are conducted at X–band. For example, Fig. 6.6 shows
hourly measurement of these amplitudes at 10 GHz during 2–26 hours of curing.
123
Fig. 6.6.
Hourly measurement of r and t at 10 GHz for the cement paste sample
with w/c=0.375.
It is seen from Fig. 6.6 that r ( t ) decreases (increases) with time. In addition, t
is much lower than r . It is an expected result since over time free–water molecules
inside the sample evaporates and this circumstance allows more microwave signals to
pass through the sample. We removed the measurement cell between antennas after each
measurement time and than located it back at the middle distance of antennas. It is
obvious from Fig. 6.6 that reflection properties have more oscillatory behavior over 2–26
hours of curing than transmission properties. This is because the measurement cell for
each measurement may not be located at the same place (theoretically, at the middle
distance of antennas). This result is in agreement with the predictions discussed in
chapter 5. We observed that the measurements are in good agreement with those in the
literature [42],[47],[122].
124
We determined the ε of the cement paste sample from the µC program, which
uses Newton’s method, using the measured r and t in Fig. 6.6. The result of the ε
determination is shown in Fig. 6.7.
Fig. 6.7.
Determined ε of the cement paste sample using the measurements in Fig.
6.6.
It is seen from Fig. 6.7 that both real and imaginary parts ( ε ' and ε " ) of the
complex permittivity of the cement sample decrease with time. This is completely in
agreement with the conclusions made in chapter 5 that ε ' ( ε " ) is dominated by reflection
(transmission) measurements. Therefore, it is no surprise that both ε ' and ε " decrease
with time since, in Fig. 6.6, reflection properties decrease while transmission properties
increase with time.
The numerical method used for ε determination of the cement sample was
Newton’s method. To compare its speed, we also computed ε by bisection and secant
methods. It is noted that Newton’s method computes ε much faster than do secant and
125
bisection methods, since it performs fewer iterations and its convergence to the solution
is quadratically [75]. As a measure, programs using Newton’s and secant methods needed
approximately 10 and 32 iterations, respectively, while the program using bisection
method did not result in a solution before maximum iteration number, N p = 50 . In
addition, while simulation times of computations were firstly around 4 seconds for
Newton’s method [78], in this study, it is noted that these times are around 1 second. This
is because we defined a highly restrictive and well–behaved domain for computations.
126
CHAPTER 7
CONCLUSION AND FUTURE WORK
7.1
Conclusion
In this dissertation, we investigated the application of microwave techniques for electrical
characterization of materials. In the first part of this dissertation, we derived forward and
reverse complex scattering (S−) parameters for a five−layer structure using conventional
multiple−reflections models. The derived S−parameter expressions can also represent a
four− or three−layer structure if appropriate substitutions for the electrical and thickness
of discarded layers are provided. This is important for materials characterization regime
since each structure (three−, four− and five−layer) finds applications in different
circumstances. Then, we proposed two methods for complex permittivity determination
of lossy materials. While the first method utilizes amplitude only reflection and
transmission S−parameter measurements at one fixed frequency for complex permittivity
determination of lossy materials, the second one uses the same measurements at different
frequencies for thickness−independent complex permittivity extraction of lossy materials.
For the first method, we derived an objective function, which depends on only one
variable, for permittivity determination of lossy materials sandwiched between two
low−loss dielectric plugs. We defined the domain for computations for this function and
showed that a unique solution exits for given physical, dimensional, and measurement
parameters in the domain. Therefore, this function readily permits fast and dynamic
127
computations as well as unique complex permittivity extraction of lossy materials. For
the second method, we simulated the change in the complex permittivity of lossy
materials for a change in frequency using two approximations (zero−order and
higher−order). We derived an objective function, which depends on only one unknown
quantity, for complex permittivity computations using the zero−order approximation.
This function also provides an initial guess for the higher−order approximations. We
expanded the power series of the extended Debye relaxation model to simulate the
change in permittivity over frequency for higher−order approximations. For the
validation of both methods, we constructed a general purpose rectangular waveguide test
cell, and calibrated the cell by two different calibration techniques. While the first
technique corrects only the frequency errors (frequency response calibration), the latter
performs a full two−port calibration (the Thru−Reflect−Line calibration). It is shown that
the extracted complex permittivity of three test liquid samples (methanol, binary mixture
of ethyl alcohol and distilled water, and a commercially antifreeze solution) by the
proposed method, which uses frequency response calibration measurements, is in good
agreement with that determined from the full two−port calibrated measurements. As a
result, the proposed methods are good candidates for industrial−based applications since
both of them use one−variable objective functions for fast and dynamic computations,
and a very simple calibration which can easily be implemented by inexpensive
microwave instruments such as a scalar network analyzer.
In the second part of the dissertation, a waveguide sample holder for automated
complex permittivity and complex permeability determination of granular and liquid
materials, which are positioned into two low−loss dielectric plugs, is presented. There are
128
three key features of the holder: a) it takes into account the effect of low−loss dielectric
plugs on measurements; b) it eliminates the dependency of the calibration plane on
measurements so that there is no need to fully fill the calibration plane with the unknown
sample; and c) it automatically measures the thickness of liquid samples so that a prior
physical measurement or knowledge of the sample thickness is not necessary. For
validation of the sample holder, we measured the complex permittivity of three test
liquids (methanol, distilled water, and a commercially available antifreeze solution) and
observed that the proposed sample holder is superior to the available sample holders
reported in the literature.
In the third section of the dissertation, reflection and transmission properties of
cement−based materials measured at X−band (8.2−12.4 GHz) by a free−space
microwave set−up are illustrated. Firstly, a simple and relatively inexpensive free−space
set−up is described, and a simple calibration procedure before measurements is
presented. Then, we demonstrated the reflection and transmission properties of young
mortar specimens with different water−to−cement (w/c) ratios over 20−54 hours of
curing after sample preparation. We introduced a new approach (relative slope approach
for reflection properties) to predict the history of the hydration (a chemical reaction
between water molecules and cement powder) process of these specimens. The main
results concluded from the measurements are as follows: a) the mortar with a lower w/c
ratio will gain hydration (and strength) quicker than the one with a higher w/c ratio (a
non–linear relationship exits between the degree of hydration and w/c ratios at early ages
of curing), and b) early age transmission measurements can be used to monitor the
porosity level inside cement samples, and this monitoring can be very useful for the
129
quality enhancement of these samples. Next, we showed the reflection and transmission
properties of hardened mortar and concrete specimens with different w/c ratios over 3−36
months of curing after sample preparation. We introduced a new approach (relative
difference approach for transmission measurements) to monitor the aging inside these
specimens. The main results of the measurements are: a) concrete specimens will
complete the aging process sooner than mortar specimens with the same w/c ratio as a
result of heavy aggregates; b) the concrete specimen with a higher w/c ratio will exhibit
faster aging during approximately 3−6.1 months; and c) mortar specimen with a lower
w/c ratio will display faster ageing during approximately 6.1−30 months. Measurement
uncertainties of the simple measurement set−up are also discussed.
The final portion of the dissertation presents a simple and relatively inexpensive
microwave sensor for complex permittivity determination of lossy materials in
industrial−based applications. We firstly presented the electronic and microwave sections
of the sensor. Then, we introduced a newly developed calibration procedure, which is
similar to the well−known Thru−Reflect−Line calibration technique, to improve the
accuracy of measurements. This calibration corrects the diffraction effect from sample
edges, which is very important for free−space measurements, as well as undesired
reflections from the surrounding environment, multi−path reflections from the ground,
and the loss in the measurement system. Next, we coded three programs into a general
purpose microcontroller for complex permittivity determination and validated them by
the measurement data taken from the literature. Finally, we measured the reflection and
transmission properties of a fresh cement paste sample, which is poured into a wooden
container, at X−band for validation of the sensor. Because the sensor is noncontacting,
130
nondestructive, noninvasive, and uses amplitudes of reflection and transmission
properties, it will find many applications for electrical and physical characterization of
materials in industry.
7.2
Future Work
The waveguide sample holder discussed in chapter 4 is based on reflection and
transmission complex S−parameter measurements. In some applications such as the
measurements of low−loss granular and liquid materials, reflection measurements cannot
accurately measure the electrical properties of specimens. In this circumstance, a sample
holder based on solely transmission complex S−parameter measurements can be
developed. In addition, another remedy for the same problem is to develop a waveguide
sample holder for measurements of low-loss specimens backed by a metal conductor.
Both of these sample holders are currently under investigation.
The prediction and estimation of compressive strength of cement−based materials
are the key research fields in construction industry. It was shown that amplitudes of the
microwave reflection properties over time can be correlated to the compressive strength
of these materials. Although this correlation is verified by measurements, amplitudes of
reflection properties of two cement−based materials with close w/c ratios will result in
nearly the sample reflection properties. This is because the resolution of reflection
measurements from highly reflective materials such as fresh cement−based materials
decreases. As a result, the performance of estimating the accurate correlation decreases.
Therefore, there is a need for accurate compressive strength estimation of these materials
in the construction industry. In chapter 5, we generally focused on reflection and
131
transmission properties of young mortar specimens over their early curing periods and
hardened mortar and concrete specimens during their late curing periods. It was observed
that the temporal dependencies of reflection and transmission properties of young
specimens intersect each other at approximately the 5th day of curing. This intersection
can be used as a metric for compressive strength development in cement−based materials.
In this way, the accuracy of measurements will increase provided that the sample under
investigation is moderately lossy, since we increase the resolution by using transmission
properties. There is much more that can be studied from the results of this dissertation,
and just a few possibilities are discussed above.
132
REFERENCES
[1]
[2]
[3]
[4]
[5]
[6]
[7]
[8]
[9]
[10]
[11]
[12]
[13]
[14]
[15]
R. Zoughi, Microwave Non–destructive Testing and Evaluation. Dordrecht, The
Netherlands: Kluwer Academic Publishers, 2000.
L. F. Chen, C. K. Ong, C. P. Neo, V. V. Varadan, and V. K. Varadan, Microwave
Electronics: Measurement and Materials Characterization. West Sussex,
England: John Wiley & Sons, 2004.
M. N. Afsar, J. R. Birch, and R. N. Clarke, “The measurement of the properties of
materials,” Proc. IEEE, vol. 74, pp.183–199, 1986.
J. Blitz, Electrical and Magnetic Methods of Non–destructive Testing. London,
UK: Chapman & Hall, 1997.
G. W. Carriveau and R. Zoughi, “Nondestructive evaluation and characterization
of complex composite structures”, in Proc. 11th Int. Sym. on Nondestructive
Characterization of Materials, Berlin, Germany, 2002, pp.273–280.
A. W. Kraszewski, Microwave Aquametry: Electromagnetic Wave Interaction
with Water–Containing Materials. A. W. Kraszewski, Ed. Piscataway, NJ: IEEE,
1996, pp 3–34.
S. O. Nelson and S. Trabelsi, Electromagnetic Aquametry. K. Kupfer, Ed. Berlin,
Germany: Springer–Verlag, 2005, pp 419–442.
E. Nyfors, “Industrial Microwave Sensors–A Review,” Subsurface Sensing Tech.
and App., vol. 1, no. 1, pp. 23–43, 2000.
A. Enders, “An accurate measurement technique for line properties, junction
effects, and dielectric and magnetic material parameters,” IEEE Trans. Microw.
Theory. Tech., vol. 37, no. 3, pp. 598–605, March 1989.
C. P. L. Rubinger and L. C. Costa, “Building a resonant cavity for the
measurement of microwave dielectric permittivity of high loss materials,”
Microwave Opt. Tech. Lett., vol. 49, no.7, pp. 1687–1690, 2007.
M. H. Afsar, X. Li, and H. Chi, “An automated 60 GHz open resonator system for
precision dielectric measurement,” IEEE Trans. Instrum. Meas., vol. 38, no. 12,
pp.1845–1853, 1990.
P. Coquet, T. Matsui, and M. Kiyokawa, “Measurements of low loss dielectric
materials in the 60 GHz band using a high–Q beam open resonator,” Proc.
Instrum. Meas. Technol. Conf., pp. 1265–1268, May 10–12, Hamamatsu, 1994.
T. M. Hirvonen, P. Vainikainen, A. Lozowski, and A. V. Raisanen,
“Measurement of dielectrics at 100 GHz with an open resonator connected to a
network analyzer,” IEEE Trans. Instrum. Meas., vol. 45, no. 4, 1996.
D. M. Pozar, Microwave Engineering, New York: John Wiley & Sons, Inc., 2005
J. Krupka, K. Derzakowski, A. Abramowicz, M. Tobar, and R. G. Geyer,
“Measurements of the complex permittivity of extremely low loss dielectric
materials using whispering galley modes,” IEEE MTT–Symp. Dig., Denver, June
8–14, pp. 1347–1350, 1997.
133
[16]
[17]
[18]
[19]
[20]
[21]
[22]
[23]
[24]
[25]
[26]
[27]
[28]
[29]
M. E. Tobar, J. Krupka, E. N. Ivanov, and R. A. Woode, “Measurement of the
complex permittivity of rutile between 10 to 300 Kelvin using whispering gallery
modes,” J. Appl. Phys., vol. 83, no. 3, pp.1604–1609, 1998.
J. Krupka, K. Derzakowski, A. Abrampwicz, M. Tobar, and R. G. Geyer, “Use of
whispering gallery modes for complex permittivity determinations of ultra low
loss dielectric materials”, IEEE Trans. Microw. Theory Tech., vol. 47, no. 6, pp.
752–759, June 1999.
K. J. Bois, “Near–field microwave inspection and characterization of cement
based materials,” Ph.D. dissertation, Colorado State Univ., Fort Collins, Jan.
1999.
A. M. Neville, Properties of Concrete. London: Longman Groups, 1996.
S. N. Kharkovsky, M. F. Akay, U. C. Hasar, and C. D. Atis, “Measurement and
monitoring of microwave reflection and transmission properties of cement–based
specimens,” IEEE Trans. Instrum. Meas., vol. 51, no. 6, pp. 1210–1218, Dec.
2002.
J. Baker–Jarvis, “Transmission/reflection and short–circuit line permittivity
measurements,” Natl. Inst. Stand. Technol., Boulder, CO, Tech. Note 1341, July
1990.
A. M. Nicolson and G. Ross, “Measurement of the intrinsic properties of
materials by time–domain techniques,” IEEE Trans. Instrum. Meas., vol. 19, no.
4, pp. 377–382, Nov. 1970.
W. B. Weir, “Automatic measurement of complex dielectric constant and
permeability at microwave frequencies,” Proc. IEEE, vol.62, no.1, pp. 33–36,
Jan. 1974.
T. C. Williams, M. A. Stuchly, P. Saville, “Modified transmission–reflection
method for measuring constitutive parameters of thin flexible high–loss
materials,” IEEE. Trans. Microw. Theory Tech., vol. 51, no.5, pp. 1560–1566,
May 2003.
K. J. Bois, L. F. Handjojo, A. D. Benally, K. Mubarak, and R. Zoughi, “Dielectric
plug–loaded two–port transmission line measurement technique for dielectric
property characterization of granular and liquid materials,” IEEE Trans. Instrum.
Meas., vol. 48, no. 6, pp. 1141–1148, Dec. 1999.
Z. Ma and S. Okamura, “Permittivity determination using amplitudes of
transmission and reflection coefficients at microwave frequency,” IEEE Trans.
Microw. Theory Tech., vol. 47, no. 5, pp. 546–550, May 1999.
D. K. Ghodgaonkar, V. V. Varadan, and V. K. Varadan, “A free–space method
for measurement of dielectric constants and loss tangents at microwave
frequencies,” IEEE Trans. Instrum. Meas., vol. 38, no. 3, pp. 789–793, June
1989.
D. K. Ghodgaonkar, V. V. Varadan, and V. K. Varadan, “Free–space
measurement of complex permittivity and complex permeability of magnetic
materials at microwave frequencies,” IEEE Trans. Instrum. Meas., vol. 39, no. 2,
pp. 387–394, Apr. 1990.
J. Baker–Jarvis, M. D. Janezic, J. H. Grosvenor, Jr., and R. G. Geyer,
“Transmission/reflection and short–circuit line methods for measuring
134
[30]
[31]
[32]
[33]
[34]
[35]
[36]
[37]
[38]
[39]
[40]
[41]
[42]
[43]
permittivity and permeability,” Natl. Inst. Stand. Technol., Boulder, CO, Tech.
Note 1355, May 1992.
J. Baker–Jarvis, E. J. Vanzura, and W. A. Kissick, “Improved technique for
determining complex permittivity with the Transmission/Reflection method,”
IEEE Trans. Microw. Theory Tech., vol. 38, no. 8, pp. 1096–1103, Aug. 1990.
A. H. Boughriet, C. Legrand, and A. Chapoton, “A noniterative stable
transmission/reflection method for low–loss material complex permittivity
determination,” IEEE. Trans. Microw. Theory Tech., vol. 45, no. 1, pp. 52–57,
Jan. 1997.
T. Lasri, D. Glay, L. Achrait, A. Mamoni, and Y. Leriy, “Microwave methods
and systems for nondestructive control,” Subsurface Sensing Technol. Appl., vol.
1, pp. 141–160, Jan. 2000.
Y. Wang and M. N. Afsar, “Measurement of complex permittivity of liquids
using waveguide techniques,” Progress in Electromagnetics Research, vol. PIER
42, pp. 131–142, 2003.
K. Folgero, “Broad–band dielectric spectroscopy of low–permittivity fluids using
one measurement cell,” IEEE Trans. Instrum. Meas., vol. 47, no. 4, pp. 881–885,
Aug. 1998.
K. Folgero, T. Friiso, J. Hilland, and T. Tjomsland, “A broad–band and high–
sensitivity dielectric spectroscopy measurement system for quality determination
of low–permittivity fluids,” Meas. Sci. Technol., vol. 6, pp. 995–1008, 1995.
K. Folgero, “Bilinear calibration of coaxial transmission/reflection cells for
permittivity measurement of low–loss liquids,” Meas. Sci. Technol., vol. 7, pp.
1260–1269, 1996.
N. Qaddoumi, S. Ganchev, and R. Zoughi, “Microwave diagnosis of low density
glass fibers with resin binder,” Res. Nondestruc. Eval., vol. 8, no.3, pp.177–188,
1996.
J. Baker–Jarvis, M. D. Janezic, and C. A. Jones, “Shielded open–circuited sample
holder for dielectric measurements of solids and liquids,” IEEE Trans. Instrum.
Meas., vol. 47, no.2, pp. 338–344, Apr. 1998.
E. J. Vanzura, J. Baker–Jarvis, J. H. Grosvenor, and M. Janezic, “Intercomparison
of permittivity measurements using the transmission/reflection method in 7–mm
coaxial transmission lines,” IEEE Trans. Microw. Theory Tech., vol. 42, no. 11,
pp. 2063–2070, 1994.
K. E. Mattar, D. G. Watters, and M. E. Brodwin, “Influence of wall contacts on
measured complex permittivity spectra at coaxial line frequencies,” IEEE Trans.
Microw. Theory Tech., vol. 39, no. 3, pp. 532–537, 1991.
P. I. Somlo, “A convenient self–checking method for the automated microwave
measurement of µ and ε ,” IEEE Trans. Instrum. Meas., vol. 42, no. 2, pp. 213–
216, Apr. 1993.
K. J. Bois, A. D. Benally, P. S. Nowak, and R. Zoughi, “Cure–state monitoring
and water–to–cement ratio determination of fresh Portland cement–based
materials using near–field microwave techniques,” IEEE Trans. Instrum. Meas.,
vol. 47, no.3, pp. 628–637, June 1998.
J. S. Popovics, “Non–destructive evaluation for civil engineering structures and
materials,” Review of Progress in QNDE, vol. 700, pp. 32–42, 2004.
135
[44]
[45]
[46]
[47]
[48]
[49]
[50]
[51]
[52]
[53]
[54]
[55]
[56]
[57]
[58]
V. M. Malhotra, N. J. Carino, Handbook on nondestructive testing of concrete.
Boca Raton, F.L: CRC Press, 2004.
D. M. McCan, and M. C. Forde, “Review of NDT methods in the assessment of
concrete and masonry structures,” NDT & E Int., vol. 34, no. 2, pp. 71–84, 2001.
K. Mubarak, K. J. Bois, and R. Zoughi, “A simple, robust, and on–site microwave
technique for determining water–to–cement ratio (w/c) of fresh Portland cement–
based materials,” IEEE Trans. Instrum. Meas., vol. 50, no. 5, pp. 1255–1263, Oct.
2001.
K. J. Bois, A. D. Benally, and R. Zoughi, “Microwave near–field reflection
property analysis of concrete for material content determination,” IEEE Trans.
Instrum. Meas., vol. 49, no. 1, pp. 49–55, Feb. 2000.
R. Zoughi, G. L. Cone, and P. S. Nowak, “Microwave nondestructive detection of
rebars in concrete slabs,” Mater. Eval., vol. 19, pp. 1385–1386, 1991.
Y. J. Kim, L. Jofre, F. De Flaviis, and M. Q. Feng, “Microwave reflection
tomographic array for damage detection of civil structures,” IEEE Trans. Ant.
Prop., vol. 51, no. 11, pp. 3022–3032, 2003.
M. Abou–Khousa and R. Zoughi, “Disbond thickness evaluation employing
multiple–frequency near–field microwave measurements,” IEEE Trans. Instrum.
Meas., vol. 56, no. 4, pp. 1107–1113, Aug. 2007.
S. Bakhtiari, N. Qaddoumi, S. I. Ganchev, and R. Zoughi, “Microwave
noncontact examination of disbond and thickness variation in stratified composite
media,” IEEE Trans. Microw. Theory Tech., vol. 42, no. 3, pp. 389–395, Mar.
1994.
S. Bakhtiari, S. I. Ganchev, and R. Zoughi, “Open–ended rectangular waveguide
for nondestructive thickness measurement and variation detection of lossy
dielectric slabs backed by a conducting plate.” IEEE Trans. Instrum. Meas., vol.
42, no. 1, pp. 19–24, Feb. 1993.
K. J. Bois, A. D. Benally, P. S. Nowak, and R. Zoughi, “Microwave non–
destructive determination of sand to cement (s/c) ratio in mortar,” Res.
Nondestruc. Eval., vol. 9, no.4, pp.227–238, 1997.
R. Zoughi, S. D. Glay, and P. S. Nowak, “Microwave nondestructive estimation
of cement paste compressive strength,” ACI Mater. J., vol. 92, no. 1, 1995.
J. Baker–Jarvis, M. D. Janezic, P. D. Domich, and R. G. Geyer, “Analysis of an
open–ended coaxial probe with lift–off for nondestructive testing,” IEEE Trans.
Instrum. Meas., vol. 43, no. 5, pp. 711–718, Oct. 1994.
M. T. Ghasr and N. Qaddoumi, “Microwave far–field nondestructive detection
and characterization of disbands in concrete structures,” Proc., 11th Int. Symp. on
Nondestructive Characterization of Materials, Springer, Berlin, pp. 365–372,
2002.
S. Kharkovsky, C. Ryley, V. Stephen, and R. Zoughi, “Dual–polarized near–field
microwave reflectometer for noninvasive inspection of carbon fiber reinforced
polymer–strengthened structures,” IEEE Trans. Instrum. Meas., vol. 57, no.1, pp.
168–175, Jan. 2008.
I. E. Pakulis, Moisture detector apparatus and method. Advanced Moisture
Technology, Inc., Wauconda, Ill, Patent No: 4716360, 1987.
136
[59]
[60]
[61]
[62]
[63]
[64]
[65]
[66]
[67]
[68]
[69]
[70]
[71]
[72]
[73]
K. Arunachalam, V. R. Melapudi, L. Udpa, and S. S. Udpa, “Microwave NDT of
cement–based materials using far–field reflection coefficients,” NDT&E
International, vol. 39, no. 7, pp. 585–593, 2006.
K. Arunachalam, V. R. Melapudi, E. J. Rothwell, L. Udpa, and S. S. Udpa,
“Microwave NDE for reinforced concrete,” Review of Quantitative
Nondestructive Eval., vol. 25, pp. 455–460, 2006.
M. Pieraccini, G. Luzi, D. Mecatti, L. Noferini, and C. Atzeni, “A microwave
radar technique for dynamic testing of large structures,” IEEE Trans. Microw.
Theory. Tech., vol. 51, no. 5, pp. 1603–1609, May 2003.
I. L. Al–Qadi, D. K. Ghodgaonkar, V. K. Varadan, and V. V. Varadan, “Effect of
moisture on asphaltic concrete at microwave frequencies,” IEEE Tran. Geosci.
Rem. Sensing, vol. 29, no. 5, pp. 710–717, Sept. 1991.
S. N. Kharkovsky and C. D. Atis, “Nondestructive testing of mortar specimen by
using the microwave free–space method,” J. Mater. Civ. Engn., vol. 15, no.2, pp.
200–204, 2003.
U. C. Hasar, “Free–space nondestructive characterization of young mortar
samples,” J. Mater. Civ. Engn., vol. 19, no. 8, pp. 674–682, 2007.
O. Büyüköztürk, T.–Y. Yu, J. A. Ortega, “A methodology for determining
complex permittivity of construction materials based on transmission–only
coherent, wide–bandwidth free–space measurements,” Cem. Concr. Compos.,
vol. 28, no. 4, pp.349–359, 2006.
K. Sato, T. Manabe, J. Povilka, T. Ihara, Y. Kasashima, and K. Yamaki,
“Measurement of the complex refractive index of concrete at 57.5 GHz,” IEEE
Trans. Antennas Propagat., vol. 44, no. 1, pp. 35–40, 1996.
O. Landron, M. J. Feuerstein, and T. S. Rappaport, “A comparison of theoretical
and empirical reflection coefficients for typical exterior wall surfaces in a mobile
radio environment,” IEEE Trans. Antennas Propagat., vol. 44, no. 3, pp. 341–
351, Mar. 1996.
R. J. Luebbers, “A ray–tracing propagation model for digital broadcast system in
urban areas,” IEEE Trans. on Broadcasting, vol. 39, no. 3, pp. 309–317, Sept.
1993.
I. Cuinas and M. G. Sanchez, “Building material characterization from complex
transmissivity measurements at 5.8 GHz,” IEEE Trans. Antennas Propag., vol.
48, no. 8, pp. 1269–1271, Aug. 2000.
I. Cuinas and M. G. Sanchez, “Measuring, modeling, and characterizing of indoor
radio channel at 5.8 GHz,” IEEE Trans. Veh. Technol., vol. 50, no. 2, pp. 526–
535, Mar. 2001.
W. Shalaby and R. Zoughi, “Analysis of monopole sensors for cement paste
compressive strength estimation,” Res. Nondestruc. Eval., vol. 7, no. 2/3, pp.
101–105, 1995.
A. W. Kraszewski and S. O. Nelson, “Moisture content determination in single
kernels and seeds with microwave resonant sensors,” in Microwave Aquametry:
Electromagnetic Wave Interaction with Water–Containing Materials, A. W.
Kraszewski, Ed. Piscataway, NJ: IEEE Press, 1996, pp. 177–203.
S. O. Nelson and S. Trabelsi, “Permittivity measurements and agricultural
applications,” in Electromagnetic Aquametry: Electromagnetic Wave Interaction
137
[74]
[75]
[76]
[77]
[78]
[79]
[80]
[81]
[82]
[83]
[84]
[85]
[86]
[87]
with Water and Moist Substances, K. Kupfer, Ed. Berlin, Germany: Springer–
Verlag, 2005, pp. 419–442.
U. C. Hasar, M. F. Akay, and S. N. Kharkovsky, “Determination of complex
dielectric permittivity of loss materials at microwave frequencies,” J.
Mathematical and Computational Appl., vol. 8, pp. 319–26, 2003.
W. H. Press, S. A. Teukolsky, W. T. Vetterling, and B. P. Flannery, Numerical
Recipes in C: The Art of Scientific Computing. New York, NY: Cambridge
University Press, 1992.
A. H. Muqaibel and A. Safaai–Jazi, “A new formulation for characterization of
materials based on measured insertion transfer function,” IEEE Trans. Microw.
Theory Tech., vol. 51, no. 8, pp. 1946–1951, Aug. 2003.
U. C. Hasar, “A microcontroller–based microwave measurement system for
permittivity determination of fresh cement–based materials,” in Proc. IEEE
Instrumentation and Measurement Technology Conf. (IMTC’07), Warsaw,
Poland, 2007, pp. 1–6.
U. C. Hasar, “Far–field microcontroller based automated microwave
measurement system for characterization of dielectric permittivity of high–loss
materials,” in Proc. IEEE Instrumentation and Measurement Technology Conf.
(IMTC’06), Sorrento, Italy, 2006, pp. 942–947.
M. F. Akay, S. N. Kharkovsky, and U. C. Hasar, “An automated amplitudes–only
measurement system for permittivity determination using free–space method,” in
Proc. IEEE Instrumentation and Measurement Technology Conf. (IMTC’01),
Budapest, Hungary, 2001, pp. 503–506.
R. F. Harrington, Time–Harmonic Electromagnetic Fields. New York, NY: IEEE
Press and Wiley–Interscience, 2001, pp. 1–197.
C. A. Balanis, Advanced Engineering Electromagnetics. New York, NY: J. Wiley
& Sons, Inc., 1989.
W. D. Burnside, and K. W. Burgener, “High frequency scattering by a thin
lossless dielectric slab,” IEEE Trans. Antennas Propagat., vol. AP–31, pp. 104–
110, Jan. 1983.
I. Cuinas, and M. G. Sanchez, “Building material characterization from complex
transmissivity measurements at 5.8 GHz,” IEEE Trans. Antennas Propagat., vol.
48, pp. 1269–1271, Aug. 2000.
M. J. Havrilla and D. P. Nyquist, “Electromagnetic characterization of layered
materials via direct and de–embed methods,” IEEE Instrum. Meas., vol. 55, no. 1,
pp. 158–163, Feb. 2006.
K. W. Anderson and D. W. Hall, Elementary Real Analysis. New York, NY:
McGraw–Hill, 1979.
J. Ness, “Broad–band permittivity measurements using the semi–automatic
network analyzer,” IEEE Trans. Microw. Theory Tech., vol. 33, no.11, pp.1222–
1226, Nov. 1985.
J. A. Reynoso–Hernandez, C. F. Estrada–Maldonado, T. Parra, K. Grenier, and J.
Graffeuil, “An improved method for the wave propagation constant γ estimation
in broadband uniform millimeter–wave transmission line,” Microwave Opt.
Technol. Lett., vol. 22, pp. 268–271, Aug. 1999.
138
[88]
Y. Huang and M. Nakhkash, “Characterization of layered dielectric medium
using reflection coefficient,” Electron. Lett., vol. 34, pp.1207–1208, June 1998.
[89] J. Baker–Jarvis, R. G. Geyer, and P. D. Domich, “A nonlinear least–squares
solution with causality constraints applied to transmission line permittivity and
permeability determination,” IEEE Trans. Instrum. Meas., vol. 41, no. 5, pp. 646–
652, Oct. 1992.
[90] S. Wang, M. Niu, and D. Xu, “A frequency–varying method for simultaneous
measurement of complex permittivity and permeability with an open–ended
coaxial probe,” IEEE Trans. Microw. Theory Tech., vol. 46, no. 12, pp. 2145–
2147, Dec. 1998.
[91] V. V. Varadan and R. Ro, “Unique retrieval of complex permittivity and
permeability of dispersive materials from reflection and transmitted fields by
enforcing causality,” IEEE Trans. Microw. Theory Tech., vol. 55, no. 10, pp.
2224–2230, Oct. 2007
[92] A. Nishikata, “A swept–frequency measurement of complex permittivity and
complex permeability of a columnar specimen inserted in a rectangular
waveguide,” IEEE Trans. Microw. Theory Tech., vol. 55, no.7, pp. 1554–1567,
July 2007.
[93] G. F. Engen and C. A. Hoer, “‘Thru–reflect–line’: An improved technique for
calibrating the dual six–port automatic network analyzer,” IEEE Microw. Theory
and Tech., vol. MTT–27, no. 12, pp. 987–993, Dec. 1979.
[94] C. A. Hoer and G. F. Engen, “Calibrating a dual six–port of four–port for
measuring two–ports with any connectors,” in IEEE MTT–S Int. Microwave
Symp. Dig., Baltimore, MD, pp. 665–668, June 1986.
[95] Hewlett–Packard. HP 8719C, HP 8720C, HP 8722A Network Analyzer operating
manual, pp. 6.6 – 6.7, Oct. 1992.
[96] F. C. Smith, “Uncertainty in the reflection/transmission method for measuring
permeability and permittivity and its effects on radar absorber design,” in Proc.
7th British Electromagnetic Measurement Conf., Malvern, UK, 1995, pp. 1–4.
[97] E. Ni, “An uncertainty analysis for the measurement of intrinsic properties of
material by the combined transmission–reflection method,” IEEE Trans. Instrum.
Meas., vol. 41, no. 4, pp. 495–499, Aug. 1992.
[98] Z. A. Delecki and S. S. Stuchly, “Uncertainties in the Determination of Dielectric
Properties in the Infinite Sample Mehtod,” IEEE Trans. Instrum. Meas., vol. 38,
no. 4, pp. 902–906, Aug. 1989.
[99] L. P. Ligthart, “A fast computational technique for accurate permittivity
determination using transmission line method,” IEEE Trans. Microw. Theory
Tech., vol. 31, no. 3, pp. 249–254, Mar. 1983.
[100] H. M. Altschuler, “Dielectric constant,” in Handbook of Microwave
Measurements, vol. 2, Sucher and Fox, Eds. Brooklyn, NY: Polytechnic Press,
1963, ch. 9, p.502.
[101] K. H. Baek, J. C. Chun, and W. S. Park, “A position–insensitive measurement of
the permittivity and permeability in coaxial airline,” in 43rd Automatic RF Tech.
Group. Conf. Dig., pp. 112–116, May 1997.
[102] M. Sucher and J. Fox, Handbook of Microwave Measurements. New York, NY:
Polytechnic Press, vol.2, 1963.
139
[103] R. Olmi, M. Tedesco, C. Riminesi, and A. Ignesti, “Thickness–independent
measurement of the permittivity of thin samples in the X band,” Meas. Sci.
Technol., vol. 13, pp. 503–509, Feb. 2002.
[104] J. B. Hasted, Aqueous Dielectrics. London, UK: Chapman and Hall, 1973.
[105] F. Buckley and A. A. Maryott, Tables of Dielectric Dispersion Data for Pure
Liquids and Dilute Solutions, Natl. Bur. Stand. (U.S.) (U.S. GPO, Washington,
D.C., 1958), Circ. No. 589.
[106] J. A. R. Ball and B. Horsfield, “Resolving ambiguity in broadband waveguide
permittivity measurements on moist materials,” IEEE Trans. Instrum. Meas., vol.
47, no. 2, pp. 390–392, Apr. 1998.
[107] J. Biju Kumar, U. Raveendranath, P. Mohanan, K. T. Mathew, M. Hajian, and L.
P. Ligthart, “A simple free–space method for measuring the complex permittivity
of single and compound dielectric materials,” Microw. Opt. Technol. Lett., vol.
26, no. 2, pp. 100–105, July 2000.
[108] V. V. Varadan, R. D. Hollinger, D. K. Ghodgaonkar, and V. K. Varadan, “Free–
space, broadband measurements of high–temperature, complex dielectric
properties at microwave frequencies,” IEEE Tans. Instrum. Meas., vol. 40, no. 5,
pp. 842–846, Oct. 1991.
[109] R. D. Hollinger, K. A. Jose, A. Tellakula, V. V. Varadan, and V. K. Varadan,
“Microwave characterization of dielectric materials from 8 to 110 GHz using a
free–space setup,” Microw. Opt. Technol. Lett., vol. 26, no. 2, pp. 100–105, July
2000.
[110] N. Gagnon, J. Shaker, L. Roy, A. Petosa, and P. Berini, “Low-cost free-space
measurement of dielectric constant at Ka band,” IEE Proc.-Microw. Antennas
Propag., vol. 151, no. 3, pp. 271–276, June 2004.
[111] V. V. Varadan, K. A. Jose, and V. K. Varadan, “In situ microwave
characterization of nonplanar dielectric objects,” IEEE Trans. Microw. Theory
Tech., vol. 48, no. 3, pp. 388–394, Mar. 2000.
[112] N. Ida, Microwave NDT. The Netherlands: Kluwer Academic Publishers, vol. 10,
1992.
[113] U. S. Inan and A. S. Inan, Electromagnetic Waves. Englewood Cliffs, NJ:
Prentice Hall, 1999.
[114] P. I. Somlo and J. D. Hunter, Microwave Impedance Measurement. London, UK:
Peter Peregrines Ltd., 1985.
[115] ASTM C-33, “Standart specification for concrete aggregates,” Annual book of
ASTM standarts, C-33, West Conshohocken, Pa, 1990.
[116] A. H. Shivola, “Self consistency aspects of dielectric mixing theories,” IEEE
Trans. Geosci. Remote Sensing, vol. 27, no. 4, pp. 403–415, July 1989.
[117] F. T. Ulaby, R. K. Moore, and A. K. Fung, Microwave Remote Sensing: Active
and Passive. Dedham, MA: Artech House, pp. 2017–2025, vol. 3, 1986.
[118] P. K. Mehta, Concrete: Structure, Properties, and Materials. Englewood Cliffs,
NJ: Prentice-Hall, 1986.
[119] B. Postacioglu, Concrete, Istanbul, Turkey: Press Technicians, 1986 (in Turkish).
[120] S. Popovics, Concrete Materials: Properties, Specifications, and Testing. NJ:
Noyes Pubs, 1992.
140
[121] A. M. Neville, and J. J. Brooks, Concrete Technology. London, UK: Longman
Group, 1993.
[122] H. C. Rhim and O. Büyüköztürk, “Electromagnetic properties of concrete at
microwave frequencies range,” ACI Mater. J., vol. 95, no. 3, pp. 262–271, 1998.
[123] S. Trabelsi and S. O. Nelson, “Free-space measurement of dielectric properties of
cereal grain and oilseed at microwave frequencies,” Meas. Sci. Tecnol., vol. 14,
pp. 589–600, 2003.
[124] A.van Roggen, “An overview of dielectric measurements,” IEEE Trans. Elec.
Insulation, vol. 25, no. 1, pp.95-106, Feb. 1990.
[125] V. G. Gelnovatch, “A computer program for the direct calibration of two-port
reflectometers for automated microwave measurements,” IEEE Trans. Microw.
Theory Tech., vol. 24, no. 1, pp. 45–47, Jan. 1976.
[126] G. I. Torgovnikov, Dielectric Properties of Wood and Wood-Based Materials.
Berlin, Germany: Springer-Verlag, 1993.
141
Документ
Категория
Без категории
Просмотров
0
Размер файла
2 263 Кб
Теги
sdewsdweddes
1/--страниц
Пожаловаться на содержимое документа