ISEMA Proceedings 1999 PDF

COLLECTION
PRESENTED
TH I RD
OF
AT
PAPERS
THE
W 0 R K S ,H 0 P
ON
ELECTROMAGNETIC WAVE
INTERACTION WITH
WATER AND MOIST SUBSTANCES
,
APRIL
HOLI D AY
INN ,
SPONSORED BY THE
11 - 13 ,
ATHENS ,
1999
GEORGIA ,
USA
U .S . DEPARTMENT OF AGRICULTURE
AGRICULTURAL RESEARCH SERVICE
Collection of Papers for the
Third Workshop on
Electromagnetic Wave Interaction with
Water and Moist Substances
Holiday Inn
Athens, Georgia
April 11 - 13, 1999
Dr. Andrzej Kraszewski, Organizer

Dr. Kurt C. Lawrence, Organizer
Sponsored by U. S. Department of Agriculture,

Agricultural Research Service
Cover Design: Mr. Richard Fiala

Web Development: Mr. Jim Griffin
Collected papers for this Workshop were duplicated, for the benefit of the
participants, mainly as furnished by the authors. Their inclusion in this Collection
of Papers does not imply endorsement by the U. S. Department of Agriculture with
respect
to
technical
content,
equipment
used,
or
any
other
aspect.
Table of Contents
Welcome ......................................................................................................................... 1
Moisture Measurement I: Dr. Andrzej Kraszewski, Chair.. ......................................... 3
8:40
Fifteen years of research on moisture content determination in cereal grains. S.O.

Nelson, AW. Kraszewski, K.C. Lawrence, and S. Trabelsi, USDA-ARS, Athens,
GA ................................................................................................................................... 5
9:00
Methods and devices for density-independent moisture measurement. K. Kupfer,

Material Research Institute, Weimar University, Weimar, Germany . ............................. 11
9:20
Measurement of composition and prior treatment of foodstuffs using complex

dielectric spectra. M. Kent, K&S Associates, Biggar, Scotland; R. Knochel, F.
Daschner, Christian Albrechts University, Kiel, Germany; and U.-K. Berger, SIK,
Goteborg, Sweeden ........ ................................................................................................ 20
9:40
Three dielectric measurement techniques for liquids. J. Baker-Jarvis, and M.D.

Janezic, Nat!. Institute of Standards & Technology, Boulder, CO. ................................ 25
Moisture Measurement n: Dr. Stuart Nelson, Chair ........................... '" ................... 30
10:30 Moisture content measurements in solids - Limitations and improvements with

modern technology. R.S. Jachowicz, Institute of Electronic Systems, Warsaw
University of Technology, Warsaw, Poland (Invited paper) ......................................... 32
11:15 Microwave remote sensing of soil moisture. T.J. Schmugge, and T.J. Jackson,
USDA-ARS, Hydrology Laboratory, Beltsville, MD . ..................................................... 42
11:30 New moisture measurement method for timber using two microwave phase shifts.
Y. Zhang, and S. Okamura, Department of Electronic Engineering, Shizuoka
University, Hamamatsu, Shizuoka, Japan. .................................................................... 47
11:45 Microwave measurements for specifying the freshness of eggs. F. Volgyi,
Department ofMicrowave Telecommunication, University ofBudapest, Budapest,
Hungary. ...................................................................................................................... 52
Noon Moisture measurement in organic material with immersion microwave probes. J.
Landgraf, and A Goller, hf sensor GmbH, Leipzig, Germany. ...................................... 57
12:15 Microwave drying - A rapid method to determine the true water content of dairy
products. M. Walter, and H.-D. Isengard, Institute ofFood Technology, University
ofHohenheim, Stuttgart, Germany. .............................................................................. 62
Poster Session, Physics and Measurements of Water: Dr. Kurt Lawrence,
Chair ............................................................................................................................ 67
Multivariate calibration for optimization of a microwave moisture content method

for grain varying in temperature and bulk-density. D.D. Archibald, S. Trabelsi,
AW. Kraszewski, and S.O. Nelson, USDA-ARS, Athens, GA . ...................................... 69
iii
Determination of moisture content in wheat using an artificial neural network.

P.G. Bartley Jr., University of Georgia, Athens GA; S.O. Nelson, USDA-ARS,
Athens, GA; RW. McClendon, University of Georgia, Athens GA; and S.
Trabelsi, USDA-ARS, Athens, GA . ................................................................................ 74
Calibration of a microwave system for measuring grain moisture content. M. Ben

Slima, Warsaw University of Technology, Warsaw, Poland; RZ. Morawski,
University of Quebec at Trois-Rivieres, Quebec, Canada; AW. Kraszewski,
USDA-ARS, Athens, GA; A Barwicz, Warsaw University of Technology, Warsaw,
Poland; and S.O. Nelson, USDA-ARS, Athens, GA . ...................................................... 79
A new microwave method for moisture measurement in building materials. A.

Goller, A Handro, and J. Landgraf, hfsensor GmbH, Leipzig, Germany. ..................... 84
The application of electromagnetic field on water structure and biology. Q. Hou,

T.H. Qian, and Q.x. Song, Hankon Branch of Hua Zhong University, Wuhan Hu
Bei Province, P.R. of China. ......................................................................................... 89
Non-destructive microstrip resonator technique for moisture/permittivity

measurement. K.K. Joshi, Department of Electronics, S.P. College, Pune, India;
and RD. Pollard, Department of Electronics and Electrical Engineering,
University of Leeds, Leeds, u.K. ................................................................................... 91
Nonclassical effects of the microwave interaction with aqueous systems. AF.

Korolev, AS. Vshivtsev, Faculty of Physics, Moscow State University, Moscow,
Russia; I.v. Timoshkin, Imperial College of Science, Technology and Medicine,
London, UK; and A Pulino, Gamma Telecommunications, Cambridge, MA ................ 92.
Non-destructive and in-situ measurements of water content in plant tissue using

time-domain reflectometry. N.J. Livingston, S. Pepin, W.R Hook, D. Leight, and
C. Bassey, Department of Biology, University of Victoria, Victoria, British
Columbia, Canada. .................................................................................... ~ ................. 96
Dielectric study on dynamical structure of water in moist materials using

dielectric resonator. J.F. Rouleau, J. Goyette, T.K. Bose, Institute for Hydrogen
Research, University of Quebec, Trois-Rivieres, Quebec, Canada; S. Yagihara,
and N. Shinyashiki, Department of PhYSics, Tokai University, Hiratsuka,
Kanagawa, Japan. ........................................................................................................ 97
10
Quality forecast of particleboards using a microwave monitoring system. F.

Volgyi, Department of Microwave Telecommunications, Technical University of
Budapest, Budapest, Hungary. .................................................................................... 102
11
Observation of changes in dynamics of water molecules in various aqueous

systems using time domain reflectometry. S. Yagihara, Y. Hayashi, M. Miyairi,
M. Asano, N. Shinyashiki, Department of Physics, Tokai University, Hiratsuka,
Kanagawa, Japan; R Kita, T. Dobashi, Department Biological & Chemical
Engineering, Gunma University, Kiryu, Japan; and M. Shiotsubo, Measurement
Engineering Division, Komatsu Engineering Corp., Kawasaki-shi, Kanagawa,
Japan. ........................................................................................................................ 107
iv
12
Computer aided design for surface acoustic wave liquid sensors. M.S. Zaghloul,
Arab Academy of Science and Technology; T.E. Taha, Faculty of Electronic
Engineering, Menoufia University, Alexandria, Egypt; K. EI-Shennawy, Arab
Academy of Science and Technology; and AH. Moustafa, Faculty of Electronic
Engineering, Menoufia University, Alexandria, Egypt. ............................................... 112
Moisture Measuring Sensors: Dr. Ray King, Chair ................................................. 117
2:30
Dielectric grain moisture measurement - practical, technical, and regulatory

issues. D.B. Funk, 'Technical Services Division, Grain Inspection, Packers and
Stockyards Administration, USDA, Kansas City, MO. (Invited paper) ......................... 119
3:45
Swept-frequency grain moisture sensor. KC. Lawrence, USDA-ARS, Athens, GA;

D.B. Funk, USDA-GlPSA, Kansas City, MO; and W.R. Windham, USDA-ARS,
Athens, GA. ................................................................................................................ 128
4:00
Grain moisture measurements with time domain reflectometry. M. Stacheder, P.

Blume, R. Fundinger, and K Koehler, lMKO Micromodultechnik GmbH,
Ettlingen, Germany. ................................................................................................... 133
4: 15
A quadrupole to assess the spatial distribution of plant water content in crops. J.

Helbert, P. Boissard, INRA, Unite de Bioclimatologie, Thiverval Grignon,
France; and J.-P. Buis, ClMEL Electronique, Paris, France. ..................................... 138
4:30
A practical method of measuring moisture using an insulated monopole. A

Wisniewski, Telecommunication Research Institute, Gdansk, Poland; and P.S.
Debicki, Faculty of Electronics and Technology, University of Gdansk, Gdansk,
Poland ....................................................................................................................... 143
4:45
Moisture gradient studies in timber by the measurement of dielectric parameters

using a multiple electrode arrangement. S. Jazayeri, Eurotherm Controls, Ltd,
Worthing, UK; and K Ahmet, Faculty of Science, Technology and Design,
University of Luton, Bedfordshire, u.K. ...................................................................... 148
5 :00
Microwave converter for simultaneous measurement of the reflection and

transmission coefficients of a material. J. Kalinski, WILTECH Industrial
Metrology, Warszawa, Poland ................................................................................... 153
5:15
Application of open coaxial resonator as a sensor for microwave moisture meter.

B. Tsentsiper, B&G Microwave, Jerusalem, Israel. .................................................... 156
5:30
Development oftime domain reflectometry probes for combined use with a cone
penetrometer. C.M.P. Vaz, P.S.P. Herrmann, EMBRAPA Instrumentation Center,
Brazilian Enterprise for Agriculture Research, sao Carlos-SP, Brazil; and J.W.
Hopmans, Department of Land, Air and Water Resources, University California,
Davis, CA. . ................................................................................................................. 161
TUESDAY, APRIL 13,1999, Holiday Inn, Georgian East Room

Moisture Measurement ill: Prof Stan Stuchly, Chair ............................................. 166
8:30
Active microwave radiometry in determining sand moisture. G. Polivka, Spacek

Labs., Inc., SantaBarbara, CA . .................................................................................. 168
8:45
Resonator-based microwave moisture meter with optimized digital signal

processing. F. Daschner, R. Knochel, Christian Albrechts University, Kiel,
Germany; and M. Kent, K&S Associates, Biggar, Scotland ....................................... 173
9:00
A unified calibration method for moisture sensing in particulate materials. S.

Trabelsi, A.W. Kraszewski, and S.D. Nelson, USDA-ARS, Athens, GA . ...................... 178
9: 15
Free space moisture measurement of cellular concrete. T. Lasri, D. Glay, A.

Mamouni, and Y. Leroy, Northern Institute for Electronics and Microelectronics,
Villeneuve d'Ascq Cedex, France. .............................................................................. 184
9:30
Measurement of the moisture and salt content of building materials. W. Leschnik,

and U. Schlemm, Building Physics and Material Science Sec., Technology,
University Hamburg-Harburg, Hamburg, Germany. .................................................. 189
9:45
Dielectric properties measurement using a cavity perturbation technique. M.S.

Venkatesh, and G.S.V. Raghavan, Department Agriculture & Biosystems
Engineering, McGill University, Macdonald Campus, Ste. Anne de Bellevue,
Quebec, Canada. ........................................................................................................ 194
Physics of Porous Media: Prof Philip Rasmussen, Chair ......................................... 199
10:30 Dielectric relaxation in water saturated sedimentary rocks. R. Hilfer, J.

Widjajakusuma, and B. Biswal, Institute of Computer Science, University
Stuttgart, Stuttgart, Germany. (Invited paper) ............................................................ 201
11: 15 Dielectric relaxation of water in porous silica glasses. Y. Feldman, and A. Gutina,
Department ofApplied Physics, The Hebrew University of Jerusalem, Jerusalem,
Israel. ......................................................................................................................... 211
11 :30 Combined determination of the water content of capillary porous media and the
electrical conductivity of the pore solution by the TDR-SS-method. R. Plagge,
Institute Building Climatology, Technology, University of Dresden, Dresden,

Germany; M. Malicki, Institute of Agrophysics, Polish Academy Science, Lublin,
Poland; and P. Haupl, Institute Building Climatology, Technology, University of
Dresden, Dresden, Germany. ...................................................................................... 216
11 :45 Low frequency impedance behavior of colloidal suspensions. L.M. Dudley, and
D. Or, Department Plants, Soils and Biometerology, Utah St. University, Logan,
UT; S. Bialkowski, Department Chemistry and Biochemistry, Utah St. University,
Logan, UT; and C. Junkermeier, Department Physics, Utah St. University, Logan,
UT. ............................................................................................................................. 217
vi
Noon Thermodielectric response of variably saturated porous media: specific surface

area determination. J.M. Wraith, Department of Land Resources & Environmental
Science, Montana St. University, Bozeman, MT; and D. Or, Department of Plants,
Soils and Biometerology, Utah St. University, Logan, UT. .......................................... 222
12: 15 Water content and particle shape effects on the dielectric permittivity of
anisotropic porous media. S.B. Jones, and S.P. Friedman, S.P., Institute Soil,
Water and Environmental Science, The Volcani Center - ARO, Bet Dagan, Israel. ..... 227
Physics of Water and Mixtures: Dr. Michael Kent, Chair ....................................... 232
2:00
Dielectric properties of liquid water and saline solutions below 0 C as measured

by a resonant applicator compensation method. P.O. Risman, Microtrans AB,
Landvetter, Sweden. ................................................................................................... 234
2:15
Study of dielectric behaviour of human bile at microwave frequencies. S.B.

Kumar, U. Raveendranath, P. Augustine, Digestive Disease Center, P. V.S. Hosp.,
Cochin, Kerala, India; and K. T. Mathew, Department of Electronics, Cochin
University of Science and Technology, Cochin, Kerala, India. .................................... 235
2:30
Modeling the dielectric constant of rice grains. K.P. Thakur, Microwave

Engineering Group, Industrial Research Ltd, Auckland, New Zealand ...................... 239
2:45
Frequency and time domain methods for snow moisture and density
determination. A. Brandelik, and C. Huebner, Institute of Meteorology and
Climate Research, Karlsruhe Research Center, Karlsruhe, Germany. ........................ 244
3 :00
Dielectric properties of aqueous solutions using time domain reflectometry (TDR)

technique. S.M. Puranik, T.s. "Chanakya", Karave, Nerul, New Mumbai, India. ........ 249
3:15
Errors in TDR-determined water content in saline sand. P. Ferre, W.R. Hook, N.

Livingston, and C. Bassey, Department of Biology, University of Victoria,
Victoria, Brithish Columbia, Canada. ......................................................................... 253
3 :30
Effective frequency of TDR travel time-based measurement of bulk dielectric

permittivity. D. Or, and v.P. Rasmussen, Department Plants, Soils and
Biometerology, Utah St. University, Logan, UT. ......................................................... 257
3:45
Modeling TDR response in heterogeneous composite materials. 1. Woodhead, G.

Buchan, and D. Kulasiri, Lincoln University, Canterbury, New Zealand .................... 261
4:00
Summary and Closing Remarks
vii
viii
WELCOME TO THE WORKSHOP!

For the third time in six years I've had the pleasure and the privilege to welcome you
as attendees of the Workshop on Electromagnetic Wave Interaction with Water and
Moist Substances. It is a pleasure to meet old friends and colleagues and a privilege to
greet new ones. The birth of every tradition is painful and this is true also in relation
to the Workshop. It is encouraging that the idea conceived twenty years ago is still alive
and well. It is encouraging that the Workshop, in spite of its long scientific name, was
given nicknames, like the Workshop on Microwave Aquametry, or recently, simply the
Moisture Workshop. Nicknames mean that people like the kid and that it has a bright
future. It has been personally satisfying to take part in the development of this meeting
devoted to all aspects of microwave aquametry, from 4 hours duration in 1993 in
Atlanta, to 9 hours in San Francisco in 1996, to over 18 hours in Athens in 1999. But
duration of the meeting is only one indicator of the growth and well-being of the
Workshop. The number of active participants is still increasing, as is the number of
presented papers, and the technical level and significance of their contents. The
Collection of Papers presented at this year's Workshop contains 51 papers prepared by
120 authors from 16 countries from around the world. I believe we can call it an
international meeting !
The main topics of this Workshop are: metrological aspects of moisture content determination in solid materials, problems related to calibration and validation of the
measuring systems, general aspects of moisture measurements, technical as well as
regulatory issues, and the physics of porous media, which in practical use can be applied
in soil research and in the oil industry. Each topic will be discussed in an invited paper,
tutorial in character, and in a series of short contributions following that presentation,
which can be considered as voices in the discussion rather than the usual paper
presentation. The number of excellent papers submitted for the Workshop was amazing
- the best of them have been accepted for presentation, and to accommodate all of
them in the two-day program, the Organizers were forced to strictly limit the time
allocated for each presentation.
On a personal note, this Workshop is different from the previous ones, not only because
of its bigger size. Both previous meetings were organized in conjunction with the IEEE
MTT-S International Microwave Symposia, and the responsibility of the Organizer was
limited to soliciting and preparing for the technical program. This year everything from
the number of doughnuts for coffee breaks to the order of presentations and the audio
equipment in this room, relied on the efforts of the two - person Organizing Committee.
And let me express my sincere thanks to the Co-organizer, Dr. Kurt Lawrence for all
the hours he spent on this task, because without his unprecedented involvement, this
meeting here, today, would never have taken place. I hope that all our problems were
solved correctly and that you, as attendees, will notice nothing wrong and will consider
this as a customary well-prepared meeting.
In the course of preparation for this meeting over more than a year, at least one
thousand messages - letters, faxes, e-mails and telephone conversations, were
exchanged with hundreds of people around the world. Many of those persons are present
here today, but still a lot more were not able to attend for various reasons. However,
they have sent in different forms their best wishes to us gathered in this room. Among
many others, I should mention the following:
Prof. Dinesh Dube, Indian Institute of Technology, New Delhi, India; Prof. Lionel
Davis, UMIST, Manchester, England; Prof. Guido Gentili, University of Florence, Italy;
Dr. Kaida bin Khalid, Agriculture Academy, Kuala Lumpur, Malayasia; Prof. Bjarne
Nost, Oslo University, Norway; Prof. Ed Law, University of Georgia, Athens, USA,
Prof. Polykarpos Pissis, Technical University, Athens, Greece; Dr. Ronald Pethig,
University College of Wales, U.K.; Dr. Ethel San Roman, Bahia Blanca, Argentina;
Prof. Liang Shen, University of Houston, USA; Mr. Tohru Shimohara, Director of
Shizuoka Seiki Co., Japan; Prof. Nguyen Tran, Royal Institute of Technology,
Melbourne, Australia; and Prof. Fawwaz Ulaby, University of Michigan, USA.
I want to express my sincere thanks to all my colleagues in the Russell Research Center
who helped in various ways in preparation of the meeting and this Collection of Papers.
I thank the authors of the papers for their cooperation and effort, the Agency for
sponsoring the meeting and for financial support. I hope we will have a fruitful and
enjoyable meeting and pleasant stay in our little town of Athena.
Andrzej Kraszewski, D.Sc.

III Workshop Organizer.
Session on
MOISTURE MEASUREMENT I
Chair: Dr. Andrzej W. Kraszewski,

USDA - ARS - RRC, Athens, GA, U.S.A.
Monday, April 12, 1999.

8:40 - 10:00 AM
FIFTEEN YEARS OF RESEARCH ON MOISTURE CONTENT

DETERMINATION IN CEREAL GRAINS
Stuart O. Nelson, Andrzej W. Kraszewski, Kurt C. Lawrence and Samir Trabelsi
U. S. Department of Agriculture, Agricultural Research Service, Russell Research Center
Athens, GA 30604-5677, U.S.A. sonelson@qaru.ars.usda.gov
ABSTRA CT. Findings of research over the past fifteen years on sensing moisture content in
cereal grains by Radio-frequency (RF) and microwave measurements related to the dielectric
properties of the grain are summarized. Results are described that provide the basis for reliable
sensing of grain moisture content in single kernels of grain and in bulk grain samples, including
techniques explored for sensing moisture content independent of bulk density fluctuation and
which, therefore, have applicability to on-line moisture measurements necessary for moisture
monitoring in flowing grain.
Keywords: cereal grains, moisture content, radio-jrequency measurements, microwave

measurements
1.
INTRODUCTION
Because the moisture (water) content of cereal grains is the most important factor determining
safe storage potential, and because it is important in determining market value, suitability for
processing, and the quality of processed products, much effort has been devoted to development
of rapid methods for moisture content determination. The electrical properties of grain are highly
correlated with moisture content and have, therefore, provided the most practical means for
sensing moisture with electrical instruments in the grain trade for more than seventy years. Early
instruments were based on measurements of the dc conductivity of grain as it passed between
crushing rollers, but for at least the past 50 years, electronic instruments that sense the dielectric
properties of grain have been the dominant type of moisture meters used at the grain elevators and
grading stations in the United States and many other countries [1].
The first method for measuring the dielectric properties of grain and resulting data were published
in 1953 [2]. Since then, dielectric properties data on many kinds of grain have been measured for
different frequency ranges and collected and made available for use in grain moisture meter design
and other applications [3].
2.
DIELECTRIC PROPERTIES DEPENDENCE
The dielectric properties of usual interest are the dielectric constant E' and the dielectric loss
factor E", the real and imaginary parts, respectively, of the relative complex permittivity, E = E'
- jE" = lEI e- jO, where 0 is the loss angle of the dielectric. In this paper IIpermittivityll is
understood to represent the relative complex permittivity. Here, also, E" is interpreted to include
the energy losses in the dielectric due to all operating dielectric relaxation mechanisms and ionic
conduction. The dielectric constant is associated with electric energy storage capability of the
5
dielectric material, whereas the loss factor is associated with energy dissipation in the material in
the form of heat energy. These properties are also important in dielectric and microwave heating
applications, but for sensing moisture in materials, only very low signal levels are needed which
produce no detectable heating.
For moisture sensing in grains, it is important to understand the dependence of the dielectric
properties of the grain on moisture content and other factors as well [4], [5]. The dielectric
constant always increases with increasing moisture content, and the dielectric loss factor generally
does also. Knowledge of the variation of the dielectric properties with frequency is important in
selecting the frequency range to use for moisture sensing. Instrument designers have a wide range
of frequencies to consider in building grain moisture meters. Often the frequency selection is
based on tradition, familiarity with design in certain frequency ranges, and on component costs.
Dielectric properties also vary with the temperature of the grain [6]-[8]. Both the dielectric
constant and loss factor generally increase with increasing temperature.
In addition to frequency, moisture content, and temperature, the density of the grain is an
important factor influencing the dielectric properties. As the bulk density of a grain sample is
increased, additional mass is present to react with the electric fields, and the permittivity
(dielectric constant and loss factor) will increase. Thus, all of these factors must be taken into
consideration in designing and calibrating instruments for measuring grain moisture content.
3.
SINGLE KERNEL MOISTURE MEASUREMENTS
Single-Kernel moisture meters would be useful in detecting blended lots of high and low moisture
grain, which might be subject to spoilage under certain transport or storage conditions. They
could also be useful in checking the uniformity of drying in grain drier performance testing and in
other research applications. Three different techniques were studied to determine their suitability
for measuring the moisture content of individual kernels of com, Zea mays L. They included dc
conductance, dual-frequency parallel-plate RF impedance, and microwave resonant cavity
measurements.
A commercial instrument, operating on the crushing-roller conductance principle, which was
designed for small grains, was modified, calibrated, and evaluated for use with field com kernels
[9]. The manufacturer then developed a model suitable for use with com and soybeans. The
performance of this instrument was compared with a nondestructive technique developed in our
laboratory. This new technique involved measuring the impedance at two frequencies, 1 and 4.5
MHz, of a small parallel-plate capacitor that held the single kernel in contact between the two
small flat plates. By measuring the complex RF impedance at these two frequencies, moisture
contents within 1% were determined independent of kernel shape and size [10]. A relatively lowcost prototype instrument which automatically measured one kernel after another was built to
demonstrate the method, but it still remains to be commercialized. This method was also
extended for moisture measurements on small samples of about 20 to 30 kernels of popcorn and
wheat, Triticum aestivum L., with standard errors of less than 1% moisture content.
The third single-kernel moisture measurement technique involved the use of a microwave
resonant cavity into which the kernel was inserted. By measuring the changes produced by the
kernel in the resonant frequency of the cavity and in the cavity transmission factor, the kernel
moisture content could be determined [11]. A single simultaneous measurement of frequency
shift and change in the transmission factor was suitable for wheat kernels, which have reasonably
uniform shape. For com kernels of widely varying shapes and sizes, two sets of simultaneous
measurements, separated by a 90-degree rotation of the kernel, were required. The microwave
cavity technique required no contact with the kernel and could be developed for continuous
operation with kernels dropping through the cavity.
Performance comparisons between the dual-frequency impedance measurements and the dc
crushing-roller conductance measurements showed that both techniques performed about equally
well, with standard errors of performance of about 0.7% moisture content on com kernels ranging
from about 10% to 27% moisture. The resonant microwave cavity method provided similar
results with a standard error of 0.6% moisture for a narrower moisture range of 8% to 18%. The
resonant cavity method also measured moisture content in wheat kernels with a standard error of
0.8% for samples ranging from 10% to 21% moisture. It was also determined that kernel mass
could be obtained from the same microwave cavity measurements with accuracies in the range
from 3% to 5%.
4.
BULK GRAIN PERMITTIVITY MEASUREMENTS
Because moisture content measurement by dielectric methods depends on the dielectric properties
of grain and the dependence of those properties on moisture content and other factors, some
techniques of measurement were adapted for efficient and reliable measurements of the dielectric
properties. One method involved the use of a shielded open-circuit, coaxial sample holder for
grain, measurements with computer controlled impedance analyzers, and use of the invariance-ofthe-cross-ratio technique, to cover the frequency range from 100 kHz to 1 GHz [12]. This
system was used to obtain temperature-dependent permittivity data for wheat [6]. Another
method was developed with a coaxial sample holder designed for full two-port S-parameter
measurements and calibrated to provide permittivity measurements on grain samples from 1 to
350 MHz [13].
For determining the dielectric properties at microwave frequencies, free-space permittivity
measurement techniques have been used to obtain dielectric constants and loss factors of wheat
and com as functions of moisture content, temperature, and bulk density [7], [8]. Grain samples
were held in polyethylene or Styrofoam containers between a transmitting hom and a receiving
hom connected to a microwave network analyzer, and the dielectric properties were determined
from attenuation and phase shift measurements as the microwaves traversed the grain layer.
Homllens antennas, which provide a focused microwave beam, permit measurements on samples
of reasonable size without concern for diffraction at the edges of samples or the interference of
reflected energy from surroundings.
5.
DENSITY-INDEPENDENT GRAIN MOISTURE DETERMINATION
Because an increase in bulk density of a grain sample produces the same increase in permittivity as
an increase in moisture content of a certain amount, the influence of bulk density must be
accounted for if accurate moisture contents are to be determined by dielectric-type moisture
meters. In conventional moisture testers, which test static grain samples, the effects of bulk
density, or test weight, are taken into account in various ways. These include corrections in
calibrations determined empirically, special electrode configurations designed to compensate for
density variation due to packing of grain samples, and weighing of samples and provisions to
obtain uniform filling of sample holders. However, considerable effort has been devoted to
development of techniques that permit moisture determination from the dielectric properties alone
which are independent of bulk density variation in the grain. For measurements on flowing grain,
bulk density fluctuations are much greater than for static samples, and the need for densityindependent moisture sensing is even greater than it is for batch sample moisture testers.
Studies on multiple-frequency impedance measurements, in the range from 1 to 350 MHz, on
static samples in parallel-plate and coaxial-line sample holders designed for eventual moisture
sensing in flowing grain, revealed potential for density-independent moisture sensing in wheat
with standard errors of performance between 0.3 and 0.5% moisture content [14]-[16]. An
improved shielded parallel-plate sensor, tested with a wide range of com samples provided a
standard error of performance of 0.28% moisture content. The key to reliable broad-frequencyrange measurements was proper design of sample holders to limit impedance mismatches and data
analyses involving spectral data processing and principle component analysis. Densityindependent functions of the permittivity components were used in some of this research.
The density-independent functions had their origin in studies on moisture sensing with microwave
measurements. A number of these density-independent functions of the dielectric constant and
loss factor were developed and studied recently, applying them to data obtained on wheat and
com with free-space microwave measurements [17], [18]. These data were obtained over wide
ranges of bulk density, temperature and moisture content at frequencies between 11 and 18 GHz.
Those density-independent functions providing the best performance yielded standard errors of
about 0.2% moisture content.
In addition to providing density-independent moisture content with good accuracies, the
microwave measurements of attenuation and phase shift, at a single frequency, can simultaneously
provide a determination of the bulk density of the grain at the time of measurement [19], [20].
The accuracy of bulk density determination has been on the order of3% in studies with wheat.
The newest of the density-independent functions permits the expression of both moisture content
and density in terms of the permittivity components measured by any desired means [20, [21].
The expressions involve two constants which are dependent on the type of grain. One is the
dielectric constant divided by density, E' / p = k, for that grain at zero moisture content, or for
grain of any moisture content at very low temperature (Fig. 1). The other is a constant at a given
frequency, which represents the slope of the straight line resulting from the plot of E" / p vs. E'/ p in
the complex plane known as an Argand diagram. Other interesting features of such a plot are that
8
a change in frequency amounts to a rotation of the line about the point (k, 0), and that changes in
both moisture content or temperature amount to translations along the same straight line.
1.0
r-------,.------,-----...,.-.......-?--""'7'""1
0.8
~0
rri
0
.~
&0
11.3 GHz
~.
~rri.
S-<f
0.6
.. '
A,.0
a.
-(,I)
~0
0.4
rri
t!j
'1'
-J!i
8~
0.2
..
-1C
24C
t!J..
42C
'if
e
0.0
2.5
. 3.0
3.5
4.0
4.5
g'/p
Figure 1. Argand diagram of the permittivity components normalized to density for hard red
winter wheat at two indicated frequencies and three indicated temperatures.
6.
CONCLUSIONS
The findings of this research over the past 15 years have established the bases for several new
techniques that should eventually find practical use as new sensors are developed for application
in more sophisticated farming, grain marketing, and grain processing arenas. The need for reliable
on-line grain moisture sensors is already apparent in yield monitoring processes on harvesters
employing precision farming principles. Ability to more closely monitor grain moisture content
and bulk density will provide important tools for better management of grain handling and storage
operations and maintenance of higher quality. Similar instrumentation can provide energy savings
in crop drying and improved quality of processed products, all with important implications in
maintaining competitive market positions in the global economy.
REFERENCES
[1]
[2]
[3]
[4]
[5]
[6]
[7]
[8]
[9]
[10]
[11]
[12]
[13]
[14]
[15]
[16]
[17]
[18]
[19]
[20]
[21]
Nelson, S. 0., Use of electrical properties for grain-moisture measurement. J. Microwave Power, 1977,
12(1), 67-72.
Nelson, S. 0., Soderholm, L. R., and Yung, F. D., Determining the dielectric properties of grain. Agric.
Eng., 1953, 34(9), 608-610.
ASAE, ASAE D293.2 Dielectric properties of grain and seed. ASAE Standards 1998, S1. Joseph, MI,
Amer. Soc. Agric. Engrs., 1998, pp. 537-546.
Nelson, S. 0., Review of factors influencing the dielectric properties of cereal grains. Cereal Chern.,
1981,58(6),487-492.
Nelson, S. 0., Dielectric properties of agricultural products - Measurements and applications. CEIDP
Digest ofLiterature on Dielectrics, IEEE Trans. Elect. Insul., 1991,26(5),845-869.
Lawrence, K. C., Nelson, S. 0., and Kraszewski, A. W., Temperature dependence of the dielectric
properties of wheat. Trans. ASAE, 1990,33(2), 535-540.
Kraszewski, A. W., Trabelsi, S., and Nelson, S. 0., Wheat permittivity measurements in free space. J.
Microwave Power Electromagn. Energy, 1996, 31(3), 135-141.
Trabelsi, S., Kraszewski, A. W., and Nelson, S. 0., Microwave dielectric properties of shelled, yellowdent field com. J. Microwave Power Electromagn. Energy, 1997,32(3), 188-194.
Nelson, S. 0., and Lawrence, K. C., Evaluation of a crushing-roller conductance instrument for singlekernel com moisture measurement. Trans. ASAE, 1989,32(2), 737-743.
Kandala, C. V. K, Nelson, S. 0., and Lawrence, K. C., Non-destructive electrical measurement of
moisture content in single kernels of com. J. Agric. Eng. Res., 1989,44, 125-132.
Kraszewski, A. W., and Nelson, S. 0., Resonant cavity perturbation - Some new applications of an old
measuring technique. J. Microwave Power Electromagn. Energy, 1996, 31 (3), 178-187.
Lawrence, K. C., Nelson, S. 0., and Kraszewski, A. W., Automatic system for dielectric properties
measurements from 100 kHz to 1 GHz. Trans. ASAE, 1989,32(1),304-308.
Lawrence, K. C., Nelson, S. 0., and Bartley, P. G., Jr., Measuring dielectric properties of hard red winter
wheat from 1 to 350 MHz with a flow-through coaxial sample holder. Trans. ASAE, 1998,41(1), 143150.
Lawrence, K. C., and Nelson, S. 0., Radio-frequency density,-independent moisture determination in
wheat. Trans. ASAE, 1993, 36(2), 477-483.
Lawrence, K. C., Windham, W. R, and Nelson, S. 0., Wheat moisture determination by 1- to nO-MHz
swept-frequency admittance measurements. Trans. ASAE, 1998,41(1),135-142.
Lawrence, K. C., Windham, W. R, and Nelson, S. 0., Sensing wheat moisture content independent of
density. Trans. ASAE, 1998,41(3),693-699.
Trabelsi, S., and Nelson, S. 0., Density-independent functions for on-line microwave moisture meters: a
general discussion. Meas. Sci. Technol., 1998,9,570-578.
Kraszewski, A. W., Trabelsi, S., and Nelson, S. 0., Comparison of density-independent expressions for
moisture content determination in wheat at microwave frequencies. J. Agric. Eng. Res., 1998, 71, 227237.
Kraszewski, A. W., Trabelsi, S., and Nelson, S. 0., Moisture content determination in grain by measuring
microwave parameters. Meas. Sci. Technol., 1997,8,857-863. Addendum. Meas. Sci Technol., 1997,9,
543-544
Trabelsi, S., Kraszewski, A. W., and Nelson, S. 0., Simultaneous determination of density and water
content of particulate materials by microwave sensors. Electronics Letters, 1997,33(1),874-876.
Trabelsi, S., Kraszewski, A. W., and Nelson, S. 0., Nondestructive microwave characterization for
determining the bulk density and moisture content of shelled com. Meas. Sci. Technol., 1998, 9, 15481556.
Contact point: Stuart O. Nelson, U. S. Department of Agriculture, Agricultural Research Service, Richard B.
Russell Agricultural Research Center, P. O. Box 5677, Athens, GA 30604-5677, U.S.A. Phone: 706 5463101; Fax: 706-546-3607; E-mail: sonelson@qaru.ars.usda.gov
10
METHODS AND DEVICES FOR

DENSITY-INDEPENDENT MOISTURE MEASUREMENT
Klaus Kupfer
Material-Research and Testing Institute (MFPA) at the Bauhaus-University Weimar,

Amalienstr. 13, D-99423 Weimar, Germany ldaus.kupfer@mfpa.de
ABSTRACT. Bulk density variations cause measuring errors at nearly all indirect moisture measuring methods. In the microwave range different methods exist to increase the accuracy of moisture measurement in production processes and at the analysis of structural conditions. The most
promising methods to reduce the influence of the density use a two-parameter-measurement at
one or two frequencies in the range of radio frequencies or microwaves. This trend is a result of
developing opportunities in instrumentation design and production of numerous microwave components and fast computer equipment.
Keywords: moisture-measurement, density-independence, methods, devices
1. INTRODUCTION
Measuring errors caused by bulk density variations are in many cases bigger than those
caused by the influence of salt content at frequencies below 1 GHz. High-speed production
processes but also the quality certification on the basis of ISO 9000 require in many different
fields of industry the determination of moisture content or the mass of material with high
accuracy, and sensing mostly from one side. Applications of radiometric density sensors,
which radioactively irradiate and ionize the material are unsuitable or prohibited for foods or
agricultural products. Different investigations were carried out in order to reduce the influence
of density variations on the measuring results and to determine the density. The most successful method which is carried on development and applied all over the world is the twoparameter measurement with different equations and modifications of instruments. Two basic
electrical methods exist: simultaneous measurement of two electrical parameters of the substance
at one single frequency or the simultaneous measurement of one or two electrical parameters at
two different frequencies.
2. METHODS OF DENSITY ELIMINATION AT ONE-PARAMETER-METHODS
The first microwave instruments used microwave absorption or loss tangent as moisture dependent parameter. In laboratory set-ups the density has been kept constant by using a probe with a
certain mass in a constant volume container. It was calibrated for different materials, e.g. by
"Aquatrace" from the firm Rotronic Switzerland. In the flowing material stream on a conveyor
belt or at aggregate bins the density must be kept constant during the measurement with different
probe former systems.
In the concrete production microwave moisture sensors operate in most cases as open resonator systems using a ring antenna at a frequency of 433 MHz. The one-parameter sensors
11
have to be installed at locations with small density variations e.g. in the flowing material
stream of aggregates. By installation of a baffle plate or a chute below the discharge of an
aggregate bin, a homogenizing of the material stream but also the determination of mass
variations with a pressure gauge are possible. Both can reduce density dependent changes of
measuring values. Density variations are decreased too during the homogenizing of fresh concrete in the mixer. A microwave sensor at the bottom of a mixer can measure the moisture
content at different mixing phases and can detect the status of homogenizing of fresh concrete. By this the mixing time could be reduced up to 30 % [1].
3. METHODS FOR DENSITY COMPENSATION WITH GAMMA-RAYS
Moisture measurement by microwave absorption and density determination by gamma-rays was
firstly applied at a chute, which was installed in the product stream. The device "Accu-sense" was
developed from Kay-Ray at the beginning of the eighties [2]. The further development realised the
density independence by using the ratio of microwave phase-shift and density which was measured
simultaneously by gamma-rays. The phase shift is less influenced by ionic conductivity, temperature
changes and scattering as the attenuation [3]. The measurement set-up operating as a homodyne network-analyser at 3 GHz was produced and first successfully applied in the coal industry by the firm
EG&G Berthold (1989). The microwave measuring value is related to the mass per unit area and the
density-independent measuring value can be calculated with the following equation:
(/J
MCqJ = a---.lL+ C
mA
(1)
(mA - mass per unit area, (/Ju, AM - meas. values of phase shift and attenuation, a,b, c - coefficients}
For inhomogeneous materials a combination of attenuation and phase shift gives the following
result:
(2)
Meanwhile many different applications exist e.g. in lime-sand-brick-stone production, brick production, food industry, production of wood chipboards, etc.
4. MICROWAVE METHODS OF DENSITY COMPENSATION
Stuchly has introduced an attenuation measurement method at three different frequencies
which allows to determine the values of moisture content, density and temperature [4].
The first equation with a combination of microwave attenuation (A) and phase shift ((/J) was
published by Kraszewski 1976 [5]. The moisture content was calculated as follows:
(3)
12
The coefficients for sand were found by a regression analysis at a measurement frequency of
9.3 GHz: (4= 4,21; a2=O.014; al- a2=3.01; a3- (4=77.56. The correctness of equation (3) was proved
by microwave measurements ofWiltech at rye (0=0.2154% Me), barley (cr=0.2187% Me) and
wheat/rye (0=0.2489% Me) [6].
In the eighties Meyer and Schilz developed a density independent measurement system with different applicators for a variety of materials [7]. The following equation was tested with a heterodyne network analyser operating at 9 GHz which uses a transmission measurement unit [8; 9]:
(4)
Nelson showed the relations of permittivity dependent on density at different frequencies and
materials (coal, wheat) [10]. Kent and Kress-Rogers have investigated the density-independence
with a microwave bridge by using materials such as fish meal, wheat, coffee- / milk powder and
coal [11]. The density-dependence of Sr' and Sr" of investigated organic materials was described by
second or third-order polynomial relationships.
With resonant cavities the moisture measurement could be realised by using scalar measuring
values. The function is correlated with moisture content via the complex permittivity given by
the resonant :frequencies and quality factors respectively:
1
R{If/)=~=!...' Q;-Q;
8r -1
(5)
fa - f1
iJ
fa andjj are the resonant frequencies, Qoand Q1 are the Q-factors before and after insertion of the
sample [12]. The central part was a cylindrical resonant cavity. The sample was placed in the
centre of cavity and consisted e.g. of a bunch of cotton threads. Above and below there were
arranged two Gunn oscillators which were coupled to the cavity by means ofllcut-ofr' technique.
A resonator measurement system based on the equation (5) was developed by the firm TEWS [13].
With open and closed resonator systems it was applied successfully for different materials e.g. tobacco, foods, chemical and pharmaceutical products, paper, etc. It operates a) as a laboratory measurement system for bulk goods, b) as bypass-system in industrial online processes, c) as a process
measuring set-up for conveyor belts and with a special design c) for high-speed profile measurements up to 104 rnIsec.
For grain moisture determination an advanced density independent function was derived by
Powell et. al. [14]. He developed a system for predicting the moisture measurement of grain
by sensing dielectric properties at 10.53 GHz under flowing conditions. He found that the
relationship (4) of Meyer and Schilz was valid over a small range of densities. Based on data
published previously by Nelson [15], the following function was derived:
R{If/)
3r; -1
= \j6r
2e::V8
r
13
(6)
It was found, that this function was valid in the moisture range between 5 .. .25 % Me and had
density-independence over a wider range of densities and showed less deviations than the function
(4). This relation was applied from Lawrence and Nelson too, to determine the moisture content
of hard red winter wheat at a frequency of 1 MHz. Under static conditions the conductance and
susceptance of the shielded parallel plate capacitor with and without grain sample have been
measured. The moisture content has been ranging from approximately 11 to 22 % with bulk densities from 660 ... 855 kglm3 [16; 17]. Standard error of calibration SEe ~ 0.6 % and standard error
of prediction SEP ~ 0.5 % were obtained.
The two-frequency method have been applied at frequencies of 1 MHz and 10 MHz to measure the
moisture content of wheat by Lawrence [17]. For hard red winter wheat the following equation was
used to estimate moisture content in the called ranges of moisture and density:
,
MC=28.04-2.01clOMHz
"
+ 4.83 In C1MHz +4.410
-4
-1
'
[( J
C
-,,C
lMHz
(' J ]2
C -1
-,,-
(7)
lOMHz
Formula (7) is suited for high moisture content values Me > 14 %.

An analysis of density-independent equations for determination of moisture content of wheat
in the radiofrequency range was published by Berbert and Stenning [18]. By using non-linear
regression analysis on wheat varieties Mercia and Hereward, they reported the following density-independent relation for moisture content ranging from 11 to 22.1 % at 1 MHz.
MC = 1,541n(t; ) + 14,61
where
t; =
(
The equation can only be used in the range of 2.86
moisture content is reduced to 11 % ~ Me ~ 19 %.
13,44
cr
-:
cr
- 1
(8)
1J -2,86
lMHz
[(8'-1)/8"] < 16.29. The applicable range of
Kraszewski and Nelson have investigated the density-independent moisture measurement of soft red
winter wheat at a frequency of 9 GHz [19]. The measuring values of attenuation and phase-shift
have been normalised in the first case to the product of thickness by density. The calibration equation related to the moisture content may be obtained by solving the normalised values of attenuation
and phase shift (eq. 7a):
MC= 109.9
(~r
(7b)
(7a)
Another way of presenting the experimental results is to correlate directly the ratio <PIA with
the given moisture content for every density (eq. 7b). In the first case a standard error of prediction SEP ~ 0.26 % Me could be achieved, in the second case was SEP ~ 0.30 % Me.
14
By using the mass per unit area related values AlmA, /mA and the ratio of A and the density-independent measurement at 3 GHz and 9.3 GHz has been applied successfully at the
processing technology of calcium silicate brick production by Kupfer, too [20].
A further system was developed from the firm NUTech at a frequency of 2.45 GHz. The resonance curves of two dielectric resonators are tuned symmetrically to one operating frequency wop
[21]. The density independent measurement can be realised by application of differences of frequency shift and quality factors (eq. 9). This system bases on the operation mode of a discriminator. The need of the symmetric resonator pair and their required equal characteristics are disadvantages of the system. Furthermore moisture gradients between the different measuring spots
cause essential measuring errors. The following equations are valid in the neighbourhood of resonance frequencies Llw < 0)01; 0)02. The amount of the amplitudes Ui at the frequency OJ = wop can
be recorded by a peak detektor.
(9)
(0 01
(0
op
(0 02
Fig. 1 Two resonance curves symmetrically

to the operating frequency (Oop; the Influence
of /),:2.' causes frequency shifts ~(O; the influence of the dielectric losses ~g" reduces the
quality-factors.
For the accurate determination of the real and imaginary part of permittivity the following limitations of validity have to be taken into account:
Congruence of resonance curves UOI and U02

Application ofloose coupling between resonator and material under test
Other resonance modes of resonators should be in far distance
The TC of resonators have to be equal
These demands require the limitation of the measuring range and the measuring sensitivity
respectively.
On the basis of aggregates Kupfer developed a model for the density independent measurement
by using a microwave bridge with a transmission measurement set-up [22;23]. The model was
transformed onto an open resonator system operating at 150:MHz [24] Fig.2. The cot a =.M7~
shows the density independent moisture measuring value.
15
3,5
Me
3,0
dB W 2,5
1M Hz
I':::""~
3 7%
.",
~ //_0"'"
,s.>
-~
,/
,;..,....
1r::::: .' /,-:',> --.. .~.,,~1 .4
2,0
9 I CIT'
1 ,5
1 ,0
0,5
0,0
2,0
0,0
4,0
6,0
8,0
d f 1M Hz
10,0
12,0
Fig. 2 Model of density independent measurement by using a resonator system

Menke and Knochel have investigated the equation (4) for a mono-frequency density-independent
measurement for different materials and showed its limitations. A new density-independent
measurement could be realised by using the ratio N Irad
';
/'"
,.............0
10
30
20
40
50
'V 1%
Fig. 3a Attenuation versus phase shift from

tobacco; parameter: 'I' (Me)
Fig. 3b Gradient of A--sraight-lines

of tobacco versus moisture content
Fig. 3a shows that attenuation A and phase shift <P depend virtually linearly on frequency. The
slope of straight lines versus frequency increases with growing moisture content. The slope of the
calibration curve is smoothed and the saturation at high moisture values is missing Fig. 3b. The
measurement accuracy could be improved, especially at high moisture content values.
16
The model was demonstrated at different materials with a frequency-swept microwave transmission technique in X-band. Compared to eq. (4) the moisture range could be expanded. The
calibrations have been performed not only for one material but also for material groups. The
following formula uses the frequency dependence of the ratio A/(J) for evaluation. It gives the
extension of density-independent measurement in the range of high moisture contents which
is suitable for on-line calibration:
(11)
Using these measuring methods the phase measurement is greatly simplified, because of
phase tracking versus frequency. In contrast the integral phase through a material layer can
only be measured in the range 1t.
A new approach related to a plate capacitor was developed by Trinks and Kupfer. The model
bases on the density independent function of equation (5) but it does not include the inhomogeneities of the stray-field. It requires smaller dimensions of particles compared to the material wave
length (a < As). (A similiar mixing rule was proposed by Huebner and Brandelik too [26]). The
moisture and density dependence are calculated in a wide moisture range of dry material up to
material with high moisture content (Fig. 3). The evidence of the model was demonstrated by
measurements. V m Vs and Va are the volumetric fractions of the particles of water, solid and air
related to the total volume by using fill-factors Fs, Fw and Fa with their corresponding
dielectric coefficients &11'1 &s and &0. A solid-water-air-mixture consists ofn-volume particles which
can be arranged parallel and serial to the field vector:
parallel particles
& pn
=&s Fs +&w . Fw + &0 Fa
(12)
(13)
The fill-factors F s, F w,
Fa
and the density-factor
~ have
the following relation
(14)
The part of a parallel or serial particles to the total volume can be described by a material dependent structure factor z = Vp,/V. Thus the permittivity &mix of a 3-phase-water-solid-air mixture can approached as follows:
(15)
It is a function of density (;J, moisture content (lfIJ and material structure (z) with the loss factor:
tan8mix
z . & pn . tan 8 pn + (J - z) . & sn . tan 8 sn

~-~--~-------
z . & pn + (J - z) . & sn
17
(16)
The values Emix and tan8 could be detenmned by measurements and the following diagrams (Fig. 4)
have been obtained. Fig. 4b shows the density-independence at z = 0.5 with the moisture content 'V
as a parameter in a wide density range (substance loose layered - substance total compacted).
Q22
~-Q1'
-.
-c-0,2J
i0,18
~0,16
'r
,.
I1
~=Q9
7
I
~=Q5
1
~ 0,14
@.012
, I
D
-. 019
8'
~
'1'=0 o/c - w=2J %
... -. -=-- -::-
-~;.::
~0,16
O? 0,13
~---.
\jf=15%
S
10'10
aY/o
3Y/o
4)'/0
Wlo
0,07
0,0
.....
.. -_.- -~r
~~.-
'JF6% T
\jf=10%
'1'=0%
~0,10
0,10
CWo
0,22
-c-
./
0,2
0,4
0,6
0,8
1,0
Density-factor ~
Moisture content \V
Fig. 4aRatio of LillW/~fversus moisture
content; parameter: density-factor ~~
Fig. 4b Ratio of LillW/~f versus densityfactor ~~; parameter: moisture content 'V
RFERENCES:
[1] Kupfer, Ket. aI.: Genauigkeitsanforderungen an Feuchteme.6systeme bei der Betonherstellung.
Technisches Messen 1997 Vol. 64 (11), pp. 433-439
[2] Kay-Ray:" Accu-sense" online noncontacting moisture measurement system. Data sheet USA 1982
[3] Klein, A:Comparison of Attenuation and Phase Measurernent Jomn of Microwave Power 1981 Vol 16 (3), pp. 289-304
[4] Stuchly, S.; Hamid, M.: State of the art in microwave sensors for measuring non-electrical quantities. Int.
Journ. Electronics 1972 Vol. 1933 (6), pp. 617-633
[5] Kraszewski, A; Kulinski, S.: An improved Microwave Method of Moisture Content Measurement and
Control. IEEE Trans. on Ind. Electr. and Contr. Instr. IECI 1976 Vol. 23 (4), pp. 364-370
[6] Mlodzka-Stybel, A:Practica1 Verification of the Microwave Two-Parnmeter Method of Moisture Monitoring in
Grain in Harvest Time. Coni Proc. of the 201h Europ. Microwave Conference Budapest 1990, pp. 1679 - 1682
[7] Meyer, W.; Schilz, W.: A microwave method for density independent determination of the moisture
content of solids. J. Phys. D: Appl. Phys. 1980 Vo1.13, pp. 1823-1830
[8] Jacobsen, R.; Meyer, W.; Schrage, B.: Density independent moisture meter at X-band. Conf. Proc. of the
101h EuMc Warschau 1980, pp. 216-220
[9] Meyer, W.; Schilz, W.: High Frequency Dielectric Data on Selected Moist Materials. Joum. of Microwave
Power 1982 Vol. 17 (1), pp. 67-77
[10] Nelson, S.O. :Observations on the Density Dependence of Dielectric Properties of Particulate Materials.
Journ. of Microwave Power 1983 Vol. 18 (2), pp. 143-152
[11] Kent, M.; Kress-Rogers, E.: Microwave moisture and density measurements in particulate solids. Trans
Inst. MC 1986 Vol. 8 (3), pp. 161-168.
[12] Meyer, W.; Schilz, W.: Microwave measurement of moisture content in process materials. Philips techno
Review 1982 Vol. 40 (4), pp. 112-119
[13] Herrmann, R.: Mikrowellen-Feuchtemessung mit Resonatoren und ihre Anwendungen.
Materialfeuchtemessung, Renningen-Malmsheim expert-Verlag 1997, Editor K Kupfer
[14] Powell, S.D. et.al.:Use of a Density-Independent Function and Microwave Measurement System for Grain
Moisture Measurement. Trans. of the ASAE 1988 Vol. 31(6), pp.1875-1881
18
[15] Nelson, S.O.:Density dependence of the dielectric properties of wheat and whole-wheat flour. Journal of
Microwave Power 1984 19, pp. 55-64
[16] Lawrence, KC.; Nelson, S.O.:Radio-frequency density independent moisture determination in wheat.
Trans. of the ASAE 1993 Vol. 36 (2), pp. 477-483
[17] Lawrence, KC.:Density-independent multiple-frequency technique for measuring moisture content in grains with a
radio-frequency pennittivity sensor. Ph.D. Dissertation 1997, University of Georgia, Athens, Georgia
[18] Berbert, P.A; Stenning, B.C.:Analysis of Density-independent Equations for Determination of Moisture
Content of Wheat in the Radiofrequency Range. J. agric. Engng. Res. 1996, Vol. 65, pp. 275-286
[19] Kraszewski, A W.; Nelson, S.O.: Density-independent moisture determination in wheat by microwave
measurement. Trans. ASAE 1991 Vol. 34 (4), pp. 1776-1783
[20] Kupfer, K: Applikation von Mi1crowellenme6verfahren bei der Herstellung von Kalksandstein.
Materialfeuchtemessung Renningen-Malmsheim, expert-Verlag 1997 Editor, K. Kupfer
[21] Stang, G.: Verfahren und Vorrichtung zur Messung der Dielektrizitatskonstante von Probenmaterialien.
Patentschrift DE 43 42 505 Cl
[22] Kupfer, K :Feuchtemessung an Zuschlagstoffen fur die Betonherstellungunter Verwendung der MikrowelIenme6technik. Hochschule fur Architektur und Bauwesen Weimar I Baustoffverfahrenstechnik; 1990 Diss. A
[23] Kupfer, K: ''PoSSIbilities and Limitations ofDensity-lndependent Moisture Measurement with Microwaves"
Chapter 21 pp. 313-327. MicrowaveAquametry"; New York, IEEE Press Book-Series 19%, Editor A Kraszewski
[24] Kupfer, K: Dielectric Moisture Measurements - Applications in Civil Engineering. Symposium on
Humidity & Moisture London 1998 Vol. 2, pp. 242-249
[25] Menke, F.; Knochel, R.: New Density-IndependentMoisture Measurement Methods using Frequencyswept Microwave Transmission. IEEE MTT-S Digest 1996 Vol. 3, pp. 1415-1418
[26] Huebner, C.; Brandelik, A:Distinguished problems in soil and snow aquametry. In press
Contact point Dr.-lng. Klaus Kupfer, Materialforschungs- und -priifanstalt an der Bauhaus-Universitat Weimar,
Amalienstr. 13; D-99423 Weimar; Germany; Phone: +49 (3643) 564-361; Fax: +49 (3643) 564-202;
E-mail: klaus.kupfer@mfpa.de
19
MEASUREMENT OF COMPOSITION AND PRIOR TREATMENT OF

FOODSTUFFS USING COMPLEX DIELECTRIC SPECTRA.
M Kenf, R Knochel2, F Daschner and U-K Berger
1 K & S Associates, Biggar, Scotland
2 Christian Albrechts University, Kiel, Germany
3 SIK, Goteborg, Sweden
ABSTRACT. As part of an EU project (FAIR CT97 3020) microwave dielectric spectra of

processed prawns (pandalus borealis), and chicken have been measured and the data analysed
using the statistical method of Principal Component Analysis. The Principal Components were
regressed (principal Component Regression) against various compositional variables, notably
added water, protein, water and salt and the calibration validated using the method of internal
cross-validation. The Principal Components were also used as discriminators to construct
discriminant functions for the separation of the samples into groups, e.g. added water or no
added water, polyphosphate or no polyphosphate.
Keywords: Dielectrics, foods, composition, polyphosphate

1 INTRODUCTION
1.1 Added Water
Water is often added to foods during processing either inadvertently (e.g. during washing) or
deliberately (e.g. when adding polyphosphates for various reasons). EU and national legislations
exist to limit the amount of added water and other additives [1, 2] often by prescribed labelling
of product composition and additives. However, its measurement is very difficult to achieve with
any accuracy in the presence of the large amount of water that exists naturally in many
foodstuffs. If the total water content is measured, its normal range needs to be well known.
Studies of the seasonal variation in the composition of crustacea (Nephrops norvegicus) have
shown that this can vary within the range 78-82% [3]. (A fish with initially 78% water, on
gaining 20 % of its weight by water addition, then has a water content of 82%, which still lies
within its normal range). Thus legislation which specifies that the amount of added water should
not exceed a certain level may not be easily enforceable.
The complex dielectric properties, E*= E'- jE", offoods vary with frequency and temperature and the
variations depend, among other things, on the dipolar relaxation frequencies of the polar molecules
involved. The spectrum of these properties over a two to three decade range of frequencies can tell
us something unique about the material [4, 5]. Water added to foodstuffs, whether accidentally or
deliberately by the addition of polyphosphates, can have quite different effects on the shape of the
dielectric spectrum. Other constituents, notably ionic salts, are diluted and also diffuse into the
exterior water. This causes a fall in loss factor, E", at low frequencies because the dielectric losses
depend on both polar and ionic d.c. conduction effects, the latter increasing with decreasing
frequency. On the other hand polyphosphate or salt solutions add conducting ions as well as water
and the resultant spectrum is quite different showing a markedly increased loss factor at low
frequencies.
20
1.2 Principal Component Analysis

The problem is how to use such measurements when much of what is seen appears qualitative. A
common spectroscopy technique, which was applied in this work, is the multivariate approaches of
Principal Component Analysis (PCA) and regression (PCR). This technique applies a linear
transformation to a set of correlated variables, in this case, the complex spectral values, to produce a
new set of orthogonal uncorrelated variables called 'components' [6]. The coefficients (loadings) for
each spectral variable to achieve this are calculated in sequence by maximising the variance due to
each principal component. Often the first few principal components account for a large proportion
of the total variance of the original set of variables. If this is the case, they may be used to
summarise the original data and as an input to further analyses. They may be regarded as
summarising the shape of the spectrum, which is very dependent on treatment.
Principal Component Regression (PCR) uses the PCs in a multiple linear regression against, for
example, the water content or any other compositional variable, to describe these compositional
variables in terms of a linear equation involving a small number of PCs. The orthogonality of the
PCs makes this a relatively simple task since the addition of further components to the regression
does not alter the coefficients of the equation already found. After forming such a calibration any
further spectrum obtained from an unknown sample should yield the composition after calculating
the principal components from the known loading vectors.
2 MATERIALS AND :METHODS
2.1 Dielectric Measurement
Replicate (5) of the dielectric spectra of the samples were measured at 31 frequencies between 0.2
to 12.0 GHz using a network analyser and an open-ended coaxial probe and their means calculated.
The spectra were reduced with no significant loss of final accuracy to 5 harmonic frequencies with
400MHz as the fundamental. The measurements were made at 3C, 8C, 20C and 30C.
In general the standard deviation was small, between (1 and 4%) and this was used as a guide in
rejecting outliers. These can occur when an air bubble or a liquid filled cavity is close to the
dielectric probe. Large deviation from the mean value was usually obvious enough to reject the
measurement without further statistical tests.
2.2 Calibration and Validation
For calibration and validation Internal Cross-Validation (ICV) was used [7]. This procedure
uses the complete data set minus one or more samples for calibration and the excluded
samples for validation. This is repeated for all the samples to obtain the overall performance
for the unknowns.
2.3 Principal Component Analysis (PCA)
For each material the calibration set of spectral data was used to calculate the first 7 principal
components and their loadings using standard techniques. This number (7) was enough to
account for all the variance. Temperature was also included as a variable in the analysis or, as an
alternative, an attempt was made via a simple correction procedure to correct all the data to the
expected values for a particular temperature.
21
As shall be seen these measurements enable the determination of variables other than water
uptake. In fact they can be applied to discriminant analysis for the determination of more
qualitative parameters such as; water added or not, polyphosphate added or not, fresh or frozen,
raw or cooked.
2.4 Principal Component Regression (PCR)

Having obtained the PCs for a calibration set, the various variables of interest were then
regressed against them to obtain calibration equations.
The results of these regressions are presented in tables 1 and 2, being the Coefficient of
Determination (R2) and the standard error of calibration (SEC) for each calibration line.
Having acquired the calibration equation in terms of the PCs, the PCs for each validation
spectrum are obtained by multiplying the mean centred and standardised validation data by the
matrix of PC loadings found from the calibration set. This generates the PCs for the validation
data, which are then used with the calibration equation to generate predicted values of the
appropriate variables. The difference between these predicted values and the laboratory values
are the residuals, which are then used to obtain the standard error of prediction (SEP).
3 RESULTS
4.1 Composition Prediction
Some examples with calibration and validation data plotted are shown in figsl and 2.
statistical parameters measured are shown in Table l.
10
;R
~
....
Q)
.....ro
Q)
"0
Q)
:2
"0
~
0...
/xu
"0
"0
/
W)
;r
The
o Validation
X Calibration
10
-2
Added Water (%)
Fig 1 Prediction of added water content in chicken compared to measured value. The line is
merely the line of equality and not some regression line
For the chicken results the poor performance in the prediction of polyphosphate seems to be due
to the additional presence of added salt because both salt and polyphosphate contribute to the DC
conductivity and because the effects of salt are much more pronounced. When samples with only
water and polyphosphate added are considered then R2 for polyphosphate increases to 68.2% and
22
the SEC becomes 24.4. Note the large improvement in results when temperature is controlled
especially for the prawns
Table 1 Results for calibration and validation comparing the use of temperature as a variable in
the PCA with measurement at constant temperature
3.0
all
Temperature eC)
R
SEC
SEP
SEC
SEP
chicken
Uptake (%)
Protein (%)
Water (%)
Phosphate (mg per 100g)
Salt (%)
74.1
64.5
43.0
0.8
80.4
1.2
0.6
0.6
35.8
0.11
1.3
0.7
0.6
37.4
0.12
84.1
82.0
41.4
3.2
94.6
1.0
0.4
0.7
33.7
0.06
1.4
0.7
0.9
38.5
0.10
prawns
Uptake (%)
Protein (%)
Water (%)
Phosphate (mg per 100g)
Salt (%)
68.9
83.5
80.1
65.7
64.3
2.7
0.7
0.6
92.3
0.04
3.2
0.7
0.7
160
0.04
96.8
92.5
94.9
36.2
63.5
1.1
0.7
0.5
136
0.06
1.4
0.9
0.6
177
0.07
---e
a.
~
20
c:
(I)
c:
0
18
()
c:
'ii)
16
"0
14
(I)
X Calibration
o Validation
:.c
~
a. 12
12
14
16
18
20
Protein Content (%)
Fig 2 Prediction of protein content of prawns treated with polyphosphate solutions

3.2 Discrimination of treatment
The use of the Principal Components as variables in discriminant analysis has enabled the
separation of treated and untreated samples to a high degree of success as table 2 demonstrates.
23
Table 2 Discrimination of treatment using Principal Components as discriminators. The

numbers in brackets refer to the total of samples in that group
Salt
Uptake
Polyphosphate
Percent of untreated correctly allocated
86.8 (38)
87.5 (8)
66.7 (24)
Percent of treated correctly allocated
82.4 (34)
93.1 (64)
85.4 (48)
Percent of untreated correctly allocated
100 (32)
94.6 (56)
Percent of treated correctly allocated
100 (128)
100 (104)
Treatment
Sample
Chicken
Prawns
4 CONCLUSIONS
The use of PCA to reduce spectral data to a set of orthogonal variables has been shown to give
good results for the measurement of added water and other compositional variables. It has also
demonstrated its power in discriminant analysis enabling prior treatment of samples to be
determined.
REFERENCES
[1]
Anon, 'Council Directive 95/21EC, 'Food additives other than colours and sweeteners'
Official Journal of The European Communities, (1995), No L61, pI
[2]
Anon, 'Food labelling regulations', Statutory Instrument (England, Scotland and Wales),
(1996), SI 1996/1499
[3]
Kent, M and Stroud, G.D., 'Microwave attenuation of frozen Nephrops norvegicus', J Fd

Technol. (1981), 16,647-654
[4]
Kent, M. and Anderson, D., 'Dielectric studies of added water in poultry meat and
scallops', Journal of Food Engineering, (1996), 28, 239-259.
[5]
Hall, M., Zhuo, L., and Gabriel, c., 'Spectroscopic investigation of the dielectric
properties of some foods and food components' Microwave Science Series, (1994), no 8,
Ministry of Agriculture Fisheries and Food London, UK
[6]
Chatfield, C., and Collins, A.l, 'Introduction to Multivariate analysis', London,

Chapman and Hall, (1980), pp. 57-81
[7]
Martens Hand Nres T, 'Multivariate Calibration', Chichester, UK, John Wiley and Sons,
(1989), pp. 254-258
24
THREE DIELECTRIC MEASUREMENT

TECHNIQUES FOR LIQUIDS
James Baker-Jarvis and Michael D. Janezic

Radio-Frequency Technology Division
National Institute of Standards and Technology
M.S. 813.01 Boulder, CO 80303-3328 jjarvis@boulder.nist.gov
March 9, 1999
ABSTRACT Three methods for dielectric measurement of liquids are overviewed. These methods are the shielded open-circuited holder, open-ended coaxial probe, and a dielectric resonator
method. The shielded open-circuited holder has a frequency window of 1 kHz to 10 GHz. The
coaxial probe is useful from 200 MHz to 20 GHz. It is necessary to use more than one coaxialprobe fixture in order to obtain broad frequency coverage. A dielectric-resonator technique is
used for low to medium-loss liquids.
Keywords: frequency: liquids, measurements,permittivity
1. INTRODUCTION
There is a need for accurate dielectric measurements of liquids, powders, and semisolids from d.c.
to microwave frequencies. Applications include basic research and applications to the biotechnology, food, chemical, and petroleum industries. There have been many approaches used over the
years for liquid permittivity measurements. In this paper we review three methods for dielectric
measurement of liquids in the frequency range of low frequency to 20 GHz. These methods are
the shielded open-circuited holder, open-ended coaxial probe, and a dielectric-resonator method.
These three techniques allow measurement of low to high-loss liquids from low to high frequencies.
The shielded open-circuited holder consists of a coaxial line terminated in an open circuit and has
a frequency window of 10 kHz to 10 GHz. The coaxial probe is useful from 200 MHz to 20 GHz.
Dielectric-resonator techniques are accurate, but narrow banded.
Liquid measurements are complicated by high permittivities, high losses, strong temperature
dependence, and low-frequency electrode-polarization effects. The relative permittivity can range
from 2 to 5 for oils to well over 1000 for colloidal solutions. Loss can be low, for example, for
heptane or very high, for example, in saline solutions. Temperature changes as small as 10 C can
25
Liquid fill level
-----------b
++
Open circuit
S~Jnple
I:
;,,':
L-
>;
Connector
Figure 2.1. Coaxial sampleholder.

strongly influence measurements.
At low frequencies, that is below about 10 kHz, measurements on conducting liquids are
plagued by electrode polarization. Electrode polarization is due to free ions causing a double
layer near the electrode and producing an anomalously high permittivity.
2. OPEN-CIRCUITED SAMPLE HOLDERS

The shielded open-circuited coaxial line sample holder has been used for years for dielectric measurements of liquids and powders in the microwave band [1--4,7]. The shielded-open fixture is
composed of an outer conductor that extends beyond the end of the inner conductor as shown in
Figure (2.1). The sample holder operation is based on an accurate model of a coaxial line terminated in a shielded open circuit [5]. The advantage of an open-circuited holder is the ease of sample
installation, the broad frequency capability, and the strong electric field in the sample region. The
sample holder has been found useful in both ambient and high-temperature measurements. Von
Hippel [1,6,7] used an open-circuited sample holder for liquid measurements. Biological tissues
have been measured using the shielded open-circuited line, for example by Stuchly and Stuchly [1].
Bussey [2] extended the open-circuited holder technique to higher frequencies with an admittance
model. The model we use is based on a full-wave solution. The full-mode model is more accurate
than the commonly used capacitance expansion [2,5].
26
Figure 3.1. The coaxial probe
3. OPEN-ENDED COAXIAL PROBE

3.1. Theory of the Open-ended Probe
Over the years, the open-ended coaxial probe as shown in Figure (3.1) has been studied extensively both theoretically and experimentally. Open-ended coaxial probes are commonly used as
nondestructive testing tools and for liquid measurements over a frequency band of 200 MHz to
20 GHz.
The method, although nondestructive, does have limitations. For example, the fields at the
probe end contain both Ez and Ep components. If there is an air gap between sample and probe,
the discontinuity in the normal electric field causes a large error in the predicted permittivity. For
this reason the probe has been used primarily for liquid and semisolid measurements, where good
contact can be obtained. In liquid measurements, the coaxial probe is submersed in the sample
and the reflection coefficient is measured. The permittivity is then obtained by a Newton-Raphson
root finding technique. The size of the probe is related to the frequency band of interest. The
higher the frequency, the smaller the coaxial probe size. We use a full-mode model of the coaxial
probe [8-10].
4. TEO! RESONATOR FOR LIQUID MEASUREMENTS

Resonant techniques for high-loss liquids are hard to design. The primary use of a resonator
technique is to obtain reliable data at a single frequency. One resonator method that does work
is a TEol dielectric sleeve resonator as shown in Figure (4.1). The liquid sample is inserted at the
center of the cavity. Since the electric field is minimal at the center of the resonator the loss in
the sample does not damp out the Q.
27
Side View
Sample
Sleeve
Top View
Figure 4.1. Dielectric sleeve resonator.
5. References
[1] M. A. Stuchly and S. S. Stuchly, "Coaxial line reflection methods for measuring dielectric properties of biological substances at radio and microwave frequencies-A review," IEEE
Trans. Instrum. Meas., vol. IM-29, pp. 176-183, 1980.
[2] H. E. Bussey, "Dielectric measurements in a shielded open circuit coaxial line," IEEE Trans.
Instrum. Meas., vol. IM-29, pp. 120--124, June 1980.
[3] 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,
pp. 338-344, April 1998.
[4] S. O. Nelson and A. W. Kraszewski, "Grain moisture determination by microwave measurements," ASAE, vol. 33, p. 1303, 1990.
[5] P. 1. Somlo, "The discontinuity capacitance and the effective position of a shielded open
circuit in a coaxial line," Proc. Inst. Radio Elec. Eng. Australia, vol. 28, pp. 7-9, January
1967.
[6] V. Hippel, Dielectric Materials and Applications. Cambridge, MA: M.LT. Press, 1954.
28
[7] 1. Hartshorn and W. H. Ward, "The measurement of the permittivity and power factor of
dielectrics at frequencies 104 to 108 cycles per second," J. Inst. Elec. Eng., vol. 79, pp. 597609, 1936.
[8] S. Jenkins, T. E. Hodgetts, G. T. Symm, A. G. P. Warhamm, and R. N. Clarke, "Comparison

of three numerical treatments for the open-ended coaxial line sensor ," Electron. Lett., vol. 26,
pp. 234-235, 1990.
[9] L. L. Li, N. H. Ismail, 1. S. Taylor, and C. C. Davis, "Flanged coaxial microwave probes for
measuring thin moisture layers," IEEE Trans. Biomedical Eng., vol. 39, pp. 49-57, January
1992.
[10J 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., pp. 711718, October 1994.
29
Session on
MOISTURE MEASUREMENT II
Chair: Dr. Stuart O. Nelson,

USDA - ARS - RRC, Athens, GA, U.S.A.

10:30 - 12:30 PM
30
31
MOISTURE CONTENT MEASUREMENTS IN SOLIDS - LIMITATIONS

AND IMPROVEMENTS WITH MODERN TECHNOLOGY
Ryszard S. Jachowicz
Institute of Electronic Systems, Warsaw University of Technology,

ul. Nowowiejska 15/19, 00-665 Warsaw, Poland, e-mail: rsjach@ise.pw.edu.pl
ABSTRACT. The paper reviews most of the important and specific aspects of moisture content
in solid materials measurement methods. An influence of unstable distributed resistances of moist
materials on all electrical measurement methods is discussed in the paper. The multiparameter
method, Artificial Neural Network and Fuzzy Logic approaches for disturbing parameters
elimination are evaluated in the report.
Keywords: moisture measurement methods, disturbing parameters elimination
1.
INTRODUCTION
Every year a growing number of users of instruments for moisture content measurements in solid
materials also have increasing requirements; they need universal instruments (i.e. operating for
many different materials), with higher measurement accuracy and they often need some additional
information, e.g. material density, water distribution in a measured sample, etc. In fact, at a time
of very advanced digital signal processing, with powerful and inexpensive
IJ.-processors built into instruments, it is easy to believe that all mentioned features of moisture
measurement instruments can be achieved very simply.
Unfortunately, the operating level of moisture content instruments is not so spectacular as
described above. So, what is the reason for that particular situation, and what are the limits of
measurement methods employed in the instruments?
To find answers for these questions, some aspects of the physical phenomena and the
measurement methods must be analyzed; specific physical parameters of water, behavior of moist
material in electric fields and the limits and abilities of electrical measurement methods operating
with presence of strong distributing parameters must be considereed.
2.
EVALUATION OF MOISTURE MEASUREMENT COMPLEXITY
2.1
Moist solid materials
Among the great variety of solid materials and because of some of their specific physical features
we can list different types: crystals, polycrystallines, granulates and grains, solid porous materials,
fibrous materials, and organic materials. Water molecules can act differently with solid material,
depending on the type of material. In our analysis we will focus our attention on those types in
which water molecules can be absorbed.
32
2.2
Physical parameters of water
H 20 - water is a polar dielectric in which

90
I
t
two hydrogen atoms are bound to an oxygen
tSOC
~
80
atom with angle 104,5. From the electrical
zs't'
~
~
70
point of view a water molecule is a dipole.
S5 't'
\ 1\ \
Complex electric permittivity for bulk water
60
a5t
\ \ \ 0"is well described by Debye's equations [1].
SO
~55"C~OC!
The real part of the water dielectric constant
"1
is relatively high - close to 80 (with slight
4J
\1St
i
dependence on temperature and almost no
30
~ ~ -20 t .
-Joe
\
\
dependence on frequency up to some 3 x 109
20 ~ \,\1
-sot
'-'
\fSt
Hz - Fig. 1). Specific electrical conductivity
--=--:..... \
8
3
-,60 't':,,~ ~,~~
varies from 10- to 10- [Ocm] with
J
f
- ..... - r - ,0dependence on level of water purity - even
fJ
109
small concentrations of additives can change Fig. 1 Dielectric constant of bulk water as a
the conductivity a few orders of magnitude. function of frequency for different temperatures [1]
Electrical parameters of water in a monolayer
or in a thin film depend strongly on binding forces to the material (different for every type of
material) and on the number of water molecule layers. Those parameter are very different from
the parameters of bulk water and they are very difficult to recognize.
\In
\ f'
\\.\
-I
2.3
Water in solid materials
The amount of water in a solid state material (in thermodynamic equilibrium with surrounding
gases) is described by sorption isotherms (BET theory [2]). For porous material, the amount of
water molecules sorbed into the material depends on the number, magnitude and shape of pores.
However, the material can contain much more water than determined by sorption isotherms, when
water molecules are introduced not from humid gas or when there is no thermodynamic
equilibrium between the gas and the moist material. So, water distribution in the material is a
function of temperature and humidity of the surrounding gas, time and way of water introduction
into the material. An electrical method employed for water content measurement must consider all
the mentioned aspects.
2.4
Electrical parameters of moist materials
Electrical parameters of water (specific conductivity, dielectric constant) are considerably

different from those parameters of most solid materials. Materials containing water should change
their electrical parameters with water concentration. It happens, indeed, but it is also necessary to
consider some fundamental effects which have a major influence on electrical methods for
moisture measurements.
33
Resistivity of moist materials

For the polycrystal materials, grains, granulates, etc., the final resistance of the material sample
between any electrode system depends mostly on resistance between grains (granulates). To
illustrate the problem. Let us consider two cases:
lOWe have two grains with known resistivity (for particular moisture content), in
contact with each other (Fig.2a). For simplicity we assume that electrodes are
located inside the grains. What is the resistance between them? It is obvious that for
the point contact (in mathematical sense) the resistance between the electrodes is
infinite, and in the real case it depends much more on the area of contact than on
resistivity of the material (i.e. moisture content).
20 We have a single grain where the surface layer (skin) has a considerably different
resistivity than the bulk grain (when surface is either wet or dry in reference to bulk
grain) - Fig.2b. It has been estimated that the resistance of the whole grain R has a
much bigger contribution to the resistivity of the surface layer than the resistivity of
bulk grain [3], [4]. According to calculations, the effect of "skin" on the resistance R
is an order stronger than the contribution of bulk grain even when the surface layer
is 1110 of the grain radius in thickness.
Generally, it should be noted that a)
the resistance of moist materials
depends very strongly on sample
density (i.e. surface of contacts
between grains and number of
contacts)
and
on
water
distribution in all grains. Because Fig. 2 llIustration of resistance, a)-between grains;
those parameters are not well
b)-in single grain
repeatable and stable in time, the
resistive moisture method is unstable and of poor accuracy.
Dielectric constant of moist materials
Wet solid material can be treated, in the first order of description, as a mixture of two different
dielectrics: dry material and water. If the electric field, to which a dielectric sample is exposed, is
not uniform, no analytical description of electric permittivity for the mixture is known (with
exception of a dielectric spheroid placed in an infinite electric field). However, many different
formulas are well known (e.g. Wagner's, Lorenz-Lorentz, Clausius-Mossotti, Landau and
Lifshitz, Looyenga), which only approximate a mean value of the dielectric constant 8 with an
accuracy dependent on grain shape, density, variation of grain size and uniform distribution in the
whole sample. That kind of approximation can be used for some kinds of moist materials, if no
changes of mean value of material 8 with frequency can be observed. In the other case, we deal
with a mixed resistive/capacitive effect which can be described with RC distributed networks.
34
Mixed RC effect in moist materials

In very many dry materials, nonsignificant changes in the dielectric constant s with frequency over
a relatively wide frequency band can be observed. The dielectric constant of pure bulk water itself
depends on frequency very slightly up to 3 X 109 Hz (see Fig.1). However, the same material
when moist very often has 0 strongly
dependent on frequency.
In fact, the sample of real moist material can
be represented by RC distributed networks
as shown in Fig.3. This is a large system of
one-ports with parallel connected RC
elements. Every one-port represents some
part of the material in micro scale. The
number ofRC one-port types depends on the
kind of material. For simplicity lets take
under consideration a network consisting of
two types of one-ports. One type represents
the pore filled with water (denoted A) and
the second represents the empty pore with
surrounding material (denoted B). It is
obvious that the equivalent for whole net
capacitance is a function of frequency with
strong dependence on the mutual relation Fig. 3 Representation of moist material with RC
between the real part and the imaginary part two types one-ports and with random two-points
of impedance of every one-port unit. In real probability distribution
material structure the number of one-ports, type A and type B, depend on the amount of water
and the material structure as well.
Now, from a mathematical point of view, it can be stated that one-port type A appears at a
particular point with probability "p" and type B with probability "l-p". This is a two-point
probability random-distributed net. The
C[pF)
equivalent capacitance for such a system as a
function of probability "p" is shown in Fig.4.
It is worth noting that the capacitance can
1kHz
grow considerably for low frequencies and
100
that, if critical point of probability is passed,
the equivalent capacitance can decrease even
10
with an increase of water content. It means
that the characteristic is not monotonic [5].
This analysis has been done for the simplest
case
(two-point
random
distribution
p
probability for flat square net). In most real
materials, the described problem is more
1.0
0.&
02
o.s
complex: three dimensional, often with more
advanced net shape and not necessarily with Fig. 4 Equivalent capacitance of the on-ports system
from Fig.3 as a function of probability p for different
only two types of one-ports in the net.
frequencies
Generally speaking if a dry material has a
35
dielectric constant e independent of frequency but its characteristic e = f (moisture content)

depends on frequency, it means there is not a pure dielectric effect but rather an RC mixed effect.
In that case, calculations of the equivalent dielectric constant on the basis of the volumetric
contribution of its two phases, solid matter and water, does not make too much sense. It also
means that the measurement characteristic (C = f (moisture content)) can be unstable in time
because of resistive instability.
3. THE PARAMETERS - DISTURBING MEASUREMENT OF MOISTURE CONTENT IN
SOLIDS
disturbing
parameters
disturbing
parameters
Xl n.
x.
X2
electrical
output
SENSORS
SYSTEM
SENSOR
electrical
output
variables
x.
moisture
content
moisture
content
a)
Y'
Y.
...... ...
Ym
b)
disturbing
parameters
x.
x.
...
moisture
content
random
disturbances
Fig.
electrical
output
variables
....
Yl
y.
~
...
Ym
c)
S Representation of moisture sensor

operating with presence of disturbing
parameters: a) with single electrical
output signal; b) with sensor system
for multiparameters measurement; c)
system
for
with
sensor
multiparameters measurement with
additional random disturbances
On the basis of analysis already presented in the paper, it is evident that the measurements of
moisture content are very difficult, mostly because of the strong influence ofmany disturbing
parameters, i.e., density, porosity, chemical composition, size and shape of grains (or granulates),
temperature and humidity of the surrounding gas, and even time. Let us denote disturbing
parameters with Xi; i = 2, ... , N and by Xl - volumetric fraction of water content. Then electrical
parameter y measured in the method is a function of multi parameters: y = f (Xl, ..., Xn) - Fig.Sa.
Some of the disturbing parameters such as temperature and time are simple to measure by
additional methods, but the rest of them are rather difficult. A multiparameter measurement
method lets us decrease or even sometimes eliminate the influence of disturbing parameters when
sensor system are employed (Fig.Sb).
36
3.1
Bases of multiparameter measurement method
Mathematical aspects of the multiparameter measurement method are based on theory and
conditions [6], [7], [8], which are employed for resolution of the system of equations given
below:
YI = fl (Xl, X2, ... , Xn)
Y2 = f2 (Xl, X2, ... , Xn)
(1)
where: (Xl, X2, ... , Xn) is a vector of input variables (i.e. moisture content and disturbing variables)
(yI, Y2, ... , Ym) is a vector of expected values of output variables (electrical parameter to be
measured)
The function fI, f2, ... , fm are continuous and have continuos partial derivatives of the first order.
To find moisture content Xl, a function gl must exist in a form:
V
(Xl, ... , Xn)
where:
Xi = gi (Yil, Yi2, ... , Yir)
(2)
E~
- is the range of input variables, and (Yl, ... , Yr) are values which fulfil system (1).
Function (2) must be at least locally single-valued and exists if functions fil, fi2, ... , fir,
chosen from system (1), fulfil conditions as below:
1 an inequality occurs:
(3)
r~n
2 rank of Jacoby's matrix f determined by:
Ofil ChI ... ChI

8x2
8x1
8xn
8hr Chr ... 8hr

8x2
8x1
8x n
is equal n for every point of the input variable range.
It means that r = nand f becomes a square matrix and then condition 2 is:
detf* 0
37
(4)
Moreover, when the above conditions are fulfilled, the function (2) exists, is continuous and has
continuous partial derivatives of the first order. The moisture content Xl can be measured even
with the presence of disturbing variables X2, ... , xn .
3.2
Advantages and limitations of multiparameter measurement method
To find function (2) on the basis of equations system (1) is rather easy to do from a mathematical
point of view. It is also easy to do in measurement system equipped with a Il-computer or
specialized Il-processor system. Unfortunately, a number oflimitations or difficulties still exist and
it is not simple to bypass them.
b)
a)
Yi
Yi
Xi min
c)
Xi max
d)
.....1
......... _.lJ-l YI
Yi
Xi max
Xi min
Yi
lt~*;>----'
Y2
1
I
1
I
r~~~~~~;:::;::=-~:~ :
I
It'~
'~...
I . . . ...-;:-.::.::. . . . . : : ": .:. ::.~h~:.:::.::.~

I ..' /
~............
I"
I
Xtmin
Xi max
Xi min
e)
........... -..-..
.. . . ~.~:~.:. . . . . . . . .!
"''''' ".
.......... __ 1 Y
. 1
Xi max
Fig. 6 Illustration of multiparameter method

with two disturbing parameters; a) true
characteristics of sensors (equation systems
(1))
- solid
line;
b)
approximated
characteristics on the basis of measurement
data - dashed line; c), d), e) combinations of
every two characteristics show nonresolution in
mathematical sense, but the triangle between
three dashed lines is the space of resolution in
the metrological sense.
Yi
Xi min
Xi max
38
The first major problem is the formulation of the system of equations (1). To do that, an
identification process must be done in the whole range of changes for every variable (moisture
and all disturbing variables).
Let us say that in the identification process, 10 values for every variable from the range of its
changes will be set up. Ifwe have n variables it means that we must generate IOn samples (e.g. for
the measured variable and 2 disturbing parameters it gives 1000 samples). Every sample is
prepared with some given accuracy OXi. On the basis of measurement data (from identification
process), an approximation with the analytical function should be done. An inaccuracy of
approximation is LlYi for every equation. In fact functions f 1, ... , f m achieved this way differ from
real functions fl, ... , fn (never known) and the system of equations built in the identification
process can have either slightly different resolution or can have no resolution at all. This problem
is illustrated in Fig.6. However to bypass the problem, the model form Fig.5b can be replaced by
the model as in Fig.5c where all inaccuracies from both calibration and approximation can be
treated as random disturbances. With the help of a covariance matrix, we can calculated the best
case in the sense of some quality measure. A more advanced description of the problem is given in
[6], [7], [8].
For the moisture content measurement, the calibration (identification) process is especially
difficult. It is not easy to prepare samples in the quantity of a few hundred with high accuracy
determination of all calibration values for all disturbing variables. In the age of computer aids, the
measurement characteristic approximation is a rather minor problem. So, in the end of the method
evaluation, this is very clear: the multiparameter method has the potential to decrease or even
eliminate the influence of all disturbing parameters on the measurement variable (moisture
content) for every measurement method (i.e. capacitive, microwave, etc.), but the identification
process must be done very accurately, and preparation of samples or variation with disturbing
parameters is difficult. The multiparameter method also has some additional advantage - all
disturbing parameters Xi can be calculated in the same way as the desired variable Xl (moisture
content). For that no additional calibration is necessary - the system of equations (1) is the basis
for the determination of every variable Xi Good example for simultaneous measurement of both
moisture content and density is given in [9].
4. ARTIFICIAL NEURAL NETWORKS FOR MOISTURE DETERMINATION IN SOLIDS
During the last 15 years massive application of Artificial Neural Networks (ANN) for modeling
measurement systems in particulates for recognition multiparameters measurement problems has
been observed. Emphasis of that method is based on the idea that the system is fully automated on
the stage of organization and learning. A designer of the system must only supply for the system
enough well recognized input data- the system itself will find the best topology of the network
(i.e. Hopfield's network, multilayered feed-forward, external feedback, self organizing maps,
etc.), the next number of network layers, the number of neurons per layer and the interconnection
weights will be defined. Usually great experience of the designer is necessary to determine full
topology, which is somewhat contrary to the initial idea that procedure of system organization is
"fully automated". Artificial Neural Network learning is the last step in the system preparation for
the work. During the learning procedure the weights of interconnections are established. For that,
a sufficient amount of data must be supplied to the system. The designer has to take care that the
data are covering the whole range of input variable or parameter changes [10], [11].
39
This requirement is often very difficult to fulfil. In practice, the data for learning are supplied in
much higher numbers than necessary, to be sure that variation of every Xi parameter has the full
range of that parameter. In a comparison of the ANN method with the multiparameter
measurement method, it is necessary to underline that the main advantage of the first method is no
obligation for the determination of every Xi disturbing parameters value during the learning
process. Then some disadvantage appear - the ANN method lets the influence of disturbing
parameters decrease during regular measurements, but determination of disturbing parameters
values are not possible (in contrast to multiparameter method).
But the greatest disadvantage is that the ANN during the learning process needs much more data
for training than the multiparameters method during the identification process. That feature is very
important in the case of moisture content measurement in solids, because preparation of material
samples is very difficult (consume a lot of work and time).
However, ANN signal processing was successfully applied to microwave moisture data for hard
winter wheat [12]. The system operated at 10.6% to 19.2% moisture content and the ANN was
designed and trained on 118 samples and tested on 61 measurements. The achieved mean absolute
error of measurement was 0.135%.
5
FUZZY LOGIC POTENTIAL IN MOISTURE MEASUREMENT SYSTEMS
Fuzzy Logic approaches to multidimensional technical problems are very successful in control of
dynamic processes. If the object is not well recognized or calculation of the optimal set up of
some regulator takes too much time to calculate exact values of controlled parameters, it is better
to "react" and to change values of parameter in the proper direction rather than waste time for
calculations. Fuzzy Logic is based on the idea that any information is not completely determined
and some rules of membership for each kind of information must be set up. The fuzzy rules allow
expressing the processing strategy compactly in the form of approximate reasoning. The use of a
fuzzy approach requires knowledge of some physical features of the system [11], [13].
An application of the Fuzzy Logic approach in moisture content measurement has the potential to
be successful in the case of moisture control in some particular processes. As an example, Fuzzy
Logic based procedures for control of a dew point hygrometer will be introduced during the paper
presentation.
CONCLUSIONS
For every one, who has been involved in measurement of moisture content in solids, it is obvious
that this kind of measurements are exceptionally difficult for two main reasons: difficulties with
moist material sample preparation and the great problem with variation in values of disturbing
parameters.
In modem moisture meters, advanced signal processing, based either on the multiparameter
method or on Artificial Neural Networks seems to be necessary to assure high measurement
accuracy.
ACKNOWLEDGEMENTS:
The author would like express his great appreciation to Miss A. Keczmerska, Mr Z Azgin and
Mr J. Weremczuk for their help in this manuscript preparation.
40
REFERENCES
[1]
Antoniewicz J. - "Properties of dielectrics", WNT, Warsaw, 1971 (in Polish)
[2]
Trapnell B.M.W - "Chemisorption", Butterworths Scientific Publication, London, 1955
[3]
Gierasimowicz H., Jankowski S. - "Direct current flow through symmetrically supplied

sphere", J. Rozprawy Elektrotechniczne, 1997, No. 23, pp. 981-989 (in Polish)
[4]
Gierasimowicz H - "Comparision of electric properties of grain junction models of granular

substances", 1. Rozprawy Elektrotechniczne, 1987, No.33, pp. 55-70 (in Polish)
[5]
Jachowicz RS., Dumania P. - "Evaluation of thin film humidity sensor type MCP-MOS",
Proc. of the 1985 International Symposium on Moisture and Humidity, Washington D.C.,
USA, April 15+18, 1985
[6]
Jachowicz R.S., Sochon J. - "Method of disturbing parameters elimination at moisture

content measurements in solids", Proc. of the 1985 International Symposium on
Moisture and Humidity, Washington D.C., USA, April 15-18, 1985
[7]
Jachowicz RS. - "Electrical methods for moisture measurements", Elektronika No. 54,
Warsaw Dniv. of Technology, Publishing Div., Warsaw 1981
[8]
Sochon J. - "Foundations of computer analysis of multiparameter of measuring

nonelectrical quantities", J. Rozprawy Elektr., No.28, 1982, pp. 3-18 (in Polish).
[9]
Trabelsi S., Kraszewski AW., Nelson S.O. - "A microwave method for on-line
determination of bulk density and moisture content of particulate materials", IEEE Trans.
on Instr. & Measurement, VoIA7, No.1, Feb. 1998
[10] Hertz 1., Krogh A, Palmer RG. - "Introduction to the theory of neutral computation",
Addison-Wesley, Pub. Co., 1991
[11] Alippi C., Ferrero A, Pinri V. - "Artificial intelligence for instruments and measurement
applications", IEEE Instr. & Meas. Magazine, Vol. 1, No.2, June 1998
[12] Bartley P.G.,Jr., Nelson S.O., McClendon RW., Trabelsi S. - "Determining Moisture
Content of Wheat with an Artificial Neural Network from Microwave Transmission
Measurements", IEEE Trans. on Instr. & Measurement, Vol. 47, No.1, Feb. 1998
[13] Yager RR, Filev D.P. - "Essentials of fuzzy modeling and control", 1.Wiley & Sons, Inc.,
1994
41
MICROWAVE REMOTE SENSING OF SOIL MOISTURE

T. J Schmugge and T. J Jackson
U.S. Department of Agriculture, Agricultural Research Service, Hydrology Laboratory
Beltsville, MD 20705-2350, USA schmugge@hydrolab.arsusda.gov
ABSTRACT. The large dielectric contrast between water (k = 80) and dry soil (k= 3 or 4) at the
lower microwave frequencies causes a large change in the emissivity of soils as they become wet,
from about 0.95 when dry to 0.7 or less when wet. This emissivity difference has been observed
with radiometers operating on field, aircraft and satellite platforms. It has been found that the longer
wavelengths work better, in particular, the 21-cm wavelength has been used most extensively. Atthis
wavelength the radiometers respond to moisture variations in a surface layer about 5 cm think with
for light to moderate amounts of vegetation cover. The background physics and results from large
scale field experiments with aircraft systems will be discussed.
Keywords: remote sensing, soil moisture, microwave radiometers
1. INTRODUCTION
At microwave frequencies the most striking feature of the thermal emission from the earth's surface
is the large contrast between water and land. This emissivity contrast is due to the large difference
between the dielectric constant of water (- 80) and that of dry soils (- 5). Thus a mixture of water
and dry soil has a dielectric constant between these two extremes, affording a mechanism for the
remote sensing of the moisture content of soils at microwave frequencies. This variation in the soil's
dielectric constant produces a range of emissivity from 0.95 for dry soils to less that 0.6 for smooth
wet soils which is easily observable with a microwave radiometer. Since the early 70's there has
been extensive research studying passive microwave approaches for the remote sensing of soil
moisture [1],[2]. The basic conclusion of this research is that it is indeed possible to determine the
moisture content of the surface layer of the soil a few tenths of a wavelength thick, i.e. about a 0 to
5-cm layer using observations at the 21-cm wavelength. In general it has been found that
observations at the longer wavelengths are better because of the increased sampling depth and
reduced effects of various noise factors. The factors which effect the accuracy of this determination
include: vegetation cover, soil properties, surface roughness, and the intervening atmosphere.
Vegetation is probably the most important because a thick enough layer can totally 9bscure the soil
surface from observation. It appears that a mature corn crop (- 6kg/m2 in plant water content) is the
limiting situation, reducing the sensitivity to about 25% of the bare soil case at the 21-cm
wavelength. Obviously, the effect of vegetation would be greater at shorter wavelengths. Thus
microwave approaches could not be used to readily determine soil moisture in forested regions. The
soil properties which effect the microwave response include density and texture, these factors will
change the sensitivity but do not reduce the range of the soil moisture effect. Surface roughness can
reduce the range of the microwave response by as much as half in extreme situations, but the more
common effect is perhaps a 10 or 20 % reduction in sensitivity. Also, surface roughness and soil
density are factors which will remain relatively constant due to infrequent working of agricultural
42
fields. These basic conclusions have been verified with measurements from field towers, aircraft
and to a limited extent satellite platforms. The underlying theory or physical principles and the this
experimental verification lead us to believe that a 21-cm radiometer with a suitable size antenna
( -10m) can provide repetitive information about surface soil moisture variations on a global scale
with a spatial resolution of about 10 km. In this paper we will briefly present the basis for this
conclusion along with some results from a large scale field experiment demonstrating the approach.
The 21-cm wavelength is appropriate for practical reasons also, it is in a radio astronomy restricted
band which minimizes interference from man-made sources.
2.0 MICROWAVE RADIOMETRIC SENSITIVITY TO SOIL MOISTURE
2.1 Dielectric Properties of Soils
The large contrast between the dielectric constant of water (- 80) and that of dry soil (- 3.5) arises
from the tendency of the electric dipole of the water molecule to align itself with the electric field
at microwave frequencies. For ice the dielectric constant is large at low frequencies but because of
the binding of the water molecule in the solid, the motion is inhibited at about 104 Hertz (Hz). For
liquid water this reduction in the molecule's ability to rotate does not occur until about 10 10 Hz or
10 GHz (1 GHz = 109Hz) in the microwave range. The dielectric constant is important because it
determines the propagation characteristics of an electromagnetic wave. These characteristics include
the velocity of propagation, the wavelength and absorption of energy in the medium. The square
root of the dielectric constant is the index of refraction (n) for the material. The contrast in n at the
boundary between two media determines the
reflection and transmission coefficients of an
electromagnetic
wave at that boundary and thus the
+
o-YUMA SAND, Wt=0.17
emissivity of the surface.
The reflectivity /
emissivity can be determined from the dielectric
constant using the Fresnel equations from
electromagnetic theory.
I-
The dielectric constant of soil increases from about

3 or 4 when dry to almost 30 when wet. While the
o
u
range of variation is about the same for most soils,
~
there are differences among the soils at the lower soil
~ul 15
moisture values. These arise from textural
i5
differences, i.e. the distribution of soil particle sizes.
10
Wang and Schmugge [3] observed that the initial
water
added to a soil is tightly bound to the particle
+
surfaces and thus has dielectric properties
5
somewhere between those of bound molecules in ice
I ..,Q~~~~::'---1=-_'--_-L-_J and those of freely rotating molecules in the liquid.
O~
I
o
0.1
0.2
0.3
0.4
0.5
0.6 On y after there are several layers of water on the
3
3
VOLUMETRIC WATER CONTENT. cm tcm
particle surfaces does the water begin to behave like
Figure 1 Real and imaginary parts of the
a liquid in its dielectric properties. Examples of this
dielectric constant for sand and clay soils.
behavior are shown in figure 1 where laboratory
~Z
20
43
measurements of the dielectric constant are plotted

MICROWAVE
along with the predicted values from the model by
RADIOMETER
Wang and Schmugge. Dobson et al. [4] found that
-1-----\\ el:
separation
of the soil water into bound and bulk
H
\ I
j
'J
components depends on soil texture or, more
c:::::::::!-!:=:::::::=:==.=-:===~:::::::,-= SURFACE
R( 9, A), TSURF
directly,
the specific surface area of the soil. Their
ATMOSPHERIC
THERMAL EMISSION
TRANSMISSIVITY
FROM SURFACE
model treats the soil as a host medium of dry soil
I
I
T8 =T(H,el[RTSKy + (I-R) TSURF J+TATM (H,e)
solids containing randomly distributed inclusions of
/
\
./
OBSERVED
REFLECTED
DIRECT ATMOSPHERIC
bound water, bulk water and air. With such models
BRIGHTNESS
SKY BRIGHTNESS
CONTRIBUTION
TEMPERATURE
it is possible to study the sensitivity of a soil's
Figure 2 Sources of radiation observed by a dielectric properties to such factors as density,
radiometer at height H and angle e.
texture, salinity, etc. (e.g. Jackson and O'Neill, [5]).
fSKY
2.1 Soil Emission and Field Verification

A microwave radiometer measures the thermal emission from the soil surface, which at these
wavelengths is proportional to product of the temperature of the soil and the surface emissivity. This
product is commonly called the brightness temperature (TB)' The brightness temperature observed
by a remote platform is represented schematically in Figure 2 for a radiometer at an altitude H. At
the 21-cm wavelength 1:, the atmospheric transmission, is close to one arid Tsky and Tatrn the down and
up welling atmospheric radiances are small and are generally neglected. The remaining term (1 R)-Tsurf= c'Tsurf is the surface emission which we are interested in. This term can be measured with
sensitive radiometers on towers, aircraft or satellites.
An example of the results from a field tower is given in Figure 3 which illustrates that the emissivity
ranges from 0;6 for the wet soil (- 30 % volumetric soil moisture) to 0.9 for the dry soil (- 8% )[5].
The soil in this case was a smooth loamy sand without any vegetative cover. As shown in this figure
the calculations from the two dielectric models agree with each other and with the data reasonably
well. Thus the basic sensitivity of microwave emissivity to soil moisture variations can be observed.
The complications arise when factors such as surface roughness and vegetative cover are added to
the problem, see [6] and [7] for details. It is also important to note in this figure the soil moisture
values are for the 0 to 2-cm layer of the soil which is indicative of the moisture sampling depth with
microwave systems. Recall that we are observing the variation of the surface emissivity with soil
moisture and that this is determined by the dielectric properties, i.e. soil moisture, within a few
tenths of the wavelength at the surface.
2.2 Large Scale Aircraft Experiments

In addition to field experiments there have numerous aircraft campaigns which have demonstrated
sensitivity of microwave radiometers to soil moisture from the arid conditions of southern Arizona
[8] and the Sahelian region of West Africa [9] to the more humid conditions of the US Great Plains
[10], [11]. We would like to describe the most recent of these experiments, The Southern Great
Plains Hydrology Experiment (SGP97) [12]. The experiment was conducted in June and July of
1997 over a 10,000 km2 area in north central Oklahoma. Microwave emission maps were acquired
on 18 days over a 30 period and used to map the surface soil moisture. The conversion from
44
1.0
.....
~------------
L BAND 20 DEG. H POL

0.9
I!.
300
1981 LOAMY SAND
41984 PLOT S1
280
.ae 260
Q)
~ 0.8
Q)
:>
Cl.
en
en
E 240
~
~
w
0.7
.E 220
Q)
.c:
'"
'1:
III
"-
MIXING MODELS
0.6
_
DOBSON ET AL. (1985)
""
1...-_ _- ' -_ _---'-_ _ _-'---......._
10
20
30
~.
"Z
200
SGP97
Washita
- Model
180
""
WANG AND SCHMUGGE (1980) " " "
0.5
r/)
r/)
'" "'~/.: ,."" ',.4
.,+
10
20
30
~
40
Volumetric Soil Moisture (%)
.....
40
VOLUMETRIC SOIL MOISTURE (%)
Figure 3 Observed and predicted emissivities

versus soil moisture in the 0 - 2 cm layer for a
bare smooth loamy sand soil.
Figure 4 Comparison of area averaged T B with

soil moisture for the Little Washita watershed
in Oklahoma. Data are from the SGP97 and
Washita 92 experiments.
microwave emission to soil moisture was performed using an algorithm developed for the Washita
92 experiment [11] and involves the estimation of the vegetation biomass from remotely sensed
vegetation indices, e.g. ndvi, using Landsat data. An example of the results is shown in Figure 4
the Little Washita in the southern part of the test area. This area was studied in the 1992 experiment
[11] and thus data from the two years are provided for comparison. The figure shows the average
value of TB over the area compared with average of a large number of ground measurements of soil
moisture. Each point is the area average over the Little Washita watershed (~ 750 km2) for a day.
The soil moisture values are for 16 samples from each of 9 fields distributed over the watershed.
The expected result for the major soil type of the region is also shown for comparison. These results
show that even with coarse resolution systems that would available in space the sensitivity to large
area soil moisture variations is evident. Indeed the only 21-cm wavelength radiometer to operate
in space was aboard Skylab in the 70's and had 100 km spatial resolution. Its limited observations
were shown to be sensitive to ground measures of soil moisture [13],[14].
Further details on theSGP97 and Washita 92 experiments can be obtained at the Hydrology lab web
site: http://hydrolab.arsusda.gov
3. DISCUSSION
The ability to observed the variations of the soil's microwave emissivity with moisture content has
been verified by radiometers operating on aircraft and satellite platforms. They can provide a very
useful tool for observing the spatial and temporal variations of the surface soil moisture which have
important hydrological, agricultural and meteorological applications. Proposals to fly such systems
in space have been made to both NASA and European Space Agency but without success.
45
REFERENCES
[1]
Jackson, T.J. and Schmugge, T.J., Passive microwave remote sensing of soil moisture.
Hydroscience, 1986, 14, 123-159.
[2]
Schmugge, T. and Jackson, T.J., Mapping Surface Soil Moisture with Microwave Radiometers, Meteorol.
Atmos. Phys., 1994, Vol 54, 213 - 223.
[3]
Wang, J.R. and Schmugge, T.J., An empirical model for the complex dielectric permittivity of soils as a
function of water content. IEEE Trans Geosci Remote Sensing, 1980, GE-18:288-295.
[4]
Dobson, M. C., Ulaby, F. T., Hallikainen, M. T., and Reyes, M., Microwave dielectric behavior of wet
soil-PartII: Dielectric mixing models. IEEE Trans Geosci Remote Sensing, 1985, GE-23:35-46.
[5]
Jackson, T.J. and O'Neill, P.E., Salinity Effects on the microwave emission of soil. IEEE Trans Geosci
Remote Sensing, 1987, GE-25:214-220.
[6]
Jackson, T. J. and Schmugge, T.J. , Vegetation effects on the microwave emission of soils.
Environment, 1991,36:203-212.
[7]
Choudhury, B. J., Schmugge, T. J., Chang, A. and Newton, R. W. Effect of surface roughness on the
microwave emission from soils. J. Geophys. Res., 1979,84:5699-5706.
[8]
Schmugge, T.J., Jackson, TJ., Kustas, W.P., Roberts, R., Parry, R., Goodrich, D., Amer, S. and Weltz, M.A.,
PBMR Observations of Surface Soil Moisture in Monsoon '90. Water Resources Research, 1994, 30, pp 13211327.
[9]
Chanzy, A, Schmugge, T.J., Calvet, J.-C., Kerr, Y., van Oevelen, P., Grosjean, 0., and Wang, J.R., Airborne
microwave radiometry on a semi-arid area during HAPEX-Sahel. Journal of Hydrology, 1997, 188-189,285309 ..
[10]
Wang, J.R., Some features observed by the L-band push broom microwave radiometer over the Konza Prairie
during 1985-1989. J Geophys. Res., 1995, 100, pp 25,469 - 25,479.
[11]
Jackson, T.J., LeVine, D.M., Swift, C.T., Schmugge, T., and Schiebe, F.R., Large Area Mapping of Soil
Moisture Using the ESTAR Passive Microwave Radiometer. Remote Sens. Environ., 1995, 54, pp 27-37.
[12]
Jackson, T.J., Le Vine, D.M., Hsu, AY., Oldak, A, Starks, P.J., Swift, C.T., Isham, J., and Haken, M., Soli
Moisture Mapping at Regional Scales Using Microwave Radiometry: The Southern Great Plains Hydrology
Experiment, 1999, IEEE Trans Geosci Remote Sensing, in press.
[13]
Eagleman, J.R. and Lin, W.C., Remote Sensing of Soil Moisture by a 21-cm Passive Radiometer. J. Geophys.
Res., 1976,81, pp 3660 - 3666.
[14]
Wang, J.R., The Effect of Vegetation on Soil Moisture Sensing from Orbiting Microwave Radiometers.
Remote Sens. Environ., 1987, 11, pp 510 - 516.
46
Advances in
Rem. Sens.
NEW MOISTURE MEASUREMENT METHOD FOR TIMBER

USING TWO MICROWAVE PHASE SHIFTS
Yangjun Zhang and Seichi Okamura
Dept. of Electronic Eng., Shizuoka University,
Hamamatsu, Shizuoka, 432-8561, Japan. E-mail: tesokam@eng.shizuoka.ac.jp
ABSTRACT. New method to measure moisture contents of timber is proposed in this paper. The
moisture content can be determined without knowledge of thickness and density of the timber by
measuring two phase shifts of transmitted microwaves at two different frequencies. The principle
was demonstrated experimentally in the moisture measurement of three species, Douglas Fir,
Spruce and Cedar. In moisture range from 2 to 30 % on dry basis, the mean value of
measurement error is 1.4%, and the maximum error is 4.4%.
Key words: Moisture content measurement, Microwave, Phase shift, Timber.

1. INTRODUCTION
Moisture content is an important parameter for timber in the storage and application. It was
usually determined by the oven-dry method, which is destructive and time-consuming. Recently
microwave techniques have been successfully used in moisture content determination for various
materials [1, 2]. Two methods with microwave techniques were also reported for determination
of moisture content of timber. One method was based on microwave free-space technique, and it
detennined the moisture content from the microwave attenuation [3]. In this method, an
additional y-ray density detector must be used to compensate the density effect. The other
method was based on microwave resonance technique, and the moisture content of timber was
obtained from measurement of resonance fr~quency and quality factor [4]. However, the
resonance frequency and the quality factor are difficulty to be measured.
In this paper, a new method is proposed for timber moisture content determination. The method
is based on microwave free-space technique, and the moisture content was obtained from two
phase shifts at two different frequencies. The method has two favorable features. One is. that
moisture determination depends neith~r on the thickness nor on the density. The other is that the
phase shifts can be measured by the same measurement setup. Both features are important to
simplifYing the measurement system and reducing the cost of the instrument in practical
application.
2. PRINCIPLE
The moisture content of timber, on dry basis, is usually defined as:
W
M=~xl00%
Wd
'
(1)
where Ww and Wd are the masses of water and dry material, respectively. When a microwave
47
signal propagates through a layer of material under test, the phase shift is related with the
moisture content, the thickness and the density of the material. For a timber with moisture
content M, the phase shift ..1.<1>1 of a transmitted microwave at frequency fl can be written as:
(2)
where p and d are density and thickness of timber, RJ (p) is the dependence of phase shift on the
density and gl(M) is a function of moisture content at frequency fl. The phase shift ..1.<1>2 at
another frequency f2 can be similarly written as:
(3)
where R2(p) is the dependence of phase shift on the density and g2(M) is a function of moisture
. content at frequency f2
By combining Eq. 2 and Eq. 3, a new moisfure content function geM) is obtained:
(4)
where R(p) is the ratio of RI (p) to R2(p). It has been shown that phase shift in linear with the
density of material is the case most often met in practice [5, 6]. Thus in practical application,
assuming R(p) as constant is reasonable for small change of sample density. From Eq. 4 the
moisture content can be expressed as:
(5)
Eq. 5 shows that the moisture content can be determined according to the ratio of two phase shift
at two different frequencies. It depends neither on the thickness nor on the density of the timber.
In our measurement system, the phase shifts were measured at 9 and 12 GHz. It was shown that
these two frequencies were suitable to obtain Eq. 5 for timber. Another reason to use these two
frequencies is that they are in the single frequency range of waveguide WRJ-10, thus two phase
shifts measurement can be performed with the same microwave elements of waveguide, adapters
and antennas.
3. EXPERIMENTS
The block diagram of the measurement setup is shown in Fig. 1. Two rectangular hom antennas
with 9.5 cm x 7.0 cm apertures were used as transmitting and receiving elements. The distance
between them was 30 cm. The phase shift was measured by a network analyzer (HP8722C). All
measurements were performed at room temperature of 24C 1 C and the samples were placed
with wood fiber direction perpendicular to the electric field. In most instances, the uncertainty in
the phase shift measurement by this setup is less than 2 degrees at both 9 and 12 GHz.
48
Vector
Network
Analyzer
Computer
--..
--..
~
Transmitting
Antenna
Sample
Receiving
Antenna
Fig. 1 Schematic diagram of measurement setup

The moisture content" of three wood species, Douglas Fir, Spruce and Cedar~ were measured by
this method. These wood species are widely used in Japan. The samples were cut from raw wood
with about 22 em x 24 cm cross section, and their thiclmess varied from 2.8 to 5.8 em. The
lowest density of all samples was 0.31 g/em3, and the highest was 0.49 g/cm3 (about 58%
variation). The samples were gradually dried at 75C 2 C for various period of time to
various moisture contents. At each moisture content level, the samples were stored in a
polyethylene bag for 24 hours to obtain a uniform moisture distribution, then the sample was
weighted and the phase shift was measured. The oven moisture content used as the standard
moisture content was calculated according to Eq. 1, where the mass of dry sample was obtained
by drying the sample for 20 hours at 110C 2 C after all phase shift measurements were
fmished.
4. RESULT
Six samples, which were selected from three wood species, were used for calibration
measurement. The phase shifts of the samples in moisture range from 2 to 30% on dry basis were
measured. The result at 9 GHz, normalized by the sample thickness, is shown in Fig. 2. The
species, thielmess and density of these samples were also given in the figure. The difference of
the normalized phase shift in the samples was about 20 degree/cm because the samples were
different in the species and the density. The similar result was obtained from the normalized
phase shift at 12 GHz.
The ratio of phase shift at 12 GHz to that at 9 GHz with respect to the moisture c'ontent of these
samples is shown in Fig. 3. It can be seen the ratio decreases monotonically with the moisture
content increasing, and the standard deviation is very small (0.6%). The relationship between the
ratio and the moisture content can be described by following calibration equation:
49
M = 2078.99 - 3005.97 . R + 1085.28 R2
(6)
where R is the ratio of phase shift at 12 GHz to that at 9 GHz.
W120
OJ
~
a -
a Spruce, d=3S.7mm, p=0.4Sglcm3
a-
100
.to
Spruce, d=3S.3mm, p=0.42g1cm3
:a: Douglas Fir, d=3S.Smm, p=0.42g/cm3
- Douglas Fir, d=44.5mm, p=O.38g/cm3

Cedar, d=4S.Smm, p=O.36g1cm3
+
10
20
Cedar, d=24.Smm, p=O.43g1cm3
40
30
Moisture content [%]
Fig. 2 Nonnalized phase shift of six samples with respected to moisture content at 9 GHz
1.34
1.32
a
Spruce, d=3S.7mm, p=0.4Sglcm3
.to
Spruce, d=3S.3mm, p=0.42g1cm3
1.30
.&
<l
-.t$
1.28
x Douglas Fir, d=3S.Smm, p=0.42g1cm
1.26
- Douglas Fir, d=44.5mm, p=O.38g/cm3
1.24
Cedar, d=4S.Smm, p=O.36g1cm3
<J
1.22
...
Cedar, d=24.5mm, p=0.43g1cm3
1.20
-5
10
15
20
25
30
35
40
Moisture content [%]
Fig. 3 The ratio of phase shift at 12 GHz (~<l>l) to that at 9 GHz (~<l>2) of six samples
The calibration curve was tested for other six samples. The samples also included three wood
species, and they were also different from each other in density and thickness. The comparison
of measured moisture constant with oven moisture constant is shown in Fig. 4. In the moisture
range from 2 to 30 % on dry basis, the mean value of error is 1.4%, and the maximum error is
4.4%.
50
30
,..-,
'$.
'='
O~
.....
25
Spruce, d=35.8mm, p=OA3g1cm3
20
x Douglas Fir, d=24.9mm, p=0.39g1cm3
~ 15
'0
10
<U
1
~
Spruce, d=24.7mm, p~A7g1cm3
Douglas Fir, d=44.6mm, p=OAOglcm3

Cedar, d=55.1mm, p=0.31g1cm3
3
+ Cedar, d=35.3mm, p=0.35g1cm
o ~--~----~--~--~----~--~
5
20
25
o
10
15
30
Oven moisture content [%]
Fig. 4 Moisture content by the new method compared to the oven moisture content
5. CONCLUSION
The new method for timber moisture content determination was proposed. The method was
based on the ratio of the phase shift at 12 GHz to that at 9 GHz. The measurement result of three
wood species, Douglas Fir, Spruce and Cedar, indicated that this method was effective. The
mean value of error in the moisture content range from 2 and 30% on dry basis is 1.4%.
REFERENCES
[1]
Kraszewski, A.W., Microwave Aquametry - A Review. J. Microwave Power & Electromagnetic Energy,
1980,15(4),209-220.
[2]
Okamura, S., Microwave System for Moisture Content Measurement of Green Tea: Leaves in Drying
Process, Proc. International Symposium on Humidity & Moisture, NPL, London, 1998,204-210.
[3]
Tiuri, M., Jokela, K. and Heikkila, S., Microwave Instrument for Accurate Moisture and Density
Measurement of Timber. J. Microwave Power & Electromagnetic Energy, 1980, 15(4),251-254.
[4]
Vainikainen, P.V., Nyfors, E.G. and Fischer, M.T., Radiowave Sensor for Measuring the Properties of
Dielectric Sheets: Application to Veneer Moisture Content and Mass per Unit Area Measurement. IEEE
Trans. on 1M, 1987,36(4),1036-1039.
[5]
Kraszewski, A.W., Kulinski, S. and Stosio, Z., A Preliminary Study on Microwave Monitoring of Moisture
Content in Wheat. J. Microwave Power & Electromagnetic Energy, 1977, 12(3),241-252.
[6]
Menke, F. and Knochel, R., New density-Independent Moisture Measurement Methods Using FrequencySwept Microwave Transmission. IEEE MTT-S digest, 1996, 1415-1417.
Content point: Seichi Okamura, Dept. of Electronic Eng., Shizuoka University. Hamamatsu, Shizuoka, 432-8561,
Japan, Phone & Fax: +81-53-478-1107; E-mail: tesokam@eng.shizuoka.ac.jp
51
MICROWAVE MEASUREMENT FOR SPECIFYING

THE FRESHNESS OF EGGS
Ferenc V61gyi
Technical University of Budapest, Dept. of Microwave Telecomm.
H - 1111 Budapest, Goldmann ter 3, Hungary. t-volgyi@nov.mht.bme.hu
ABSTRACT. This paper introduces a microwave measurement setup using three microstrip
antennas integrated with a microwave oscillator and Schottky-diode detectors used as bistatic
microwave sensors. We have measured the microwave attenuation (water content) and the
bistatic radar cross section (dimensions) of eggs having different time of storage.
The measured attenuation for fresh egg is 8-10 dB at the frequency of 14.2 GHz, the variation
with time is: dA [dB] = - 0.033 T (days), thus nearly -1 dB / 30 days. Our next program is to
create a microwave equipment for automatic selection of old eggs. The basic arrangement and the
results of our measurement are discussed.
Keywords: egg, microwave aquametry, microstrip antenna, bistatic sensor.

1. INTRODUCTION
In food industries and at production of cosmetics is important to know the freshness of eggs used
as basic components. The aim of the first part of this work is: finding out the characteristic
changes of eggs during the storage period, tracking the measurable chemical and physicochemical parameters along the ageing. The examined chemical parameters were: protein content,
free amino acid content, ammonia content, PH-value, while the physical parameters were: content
of dry substance, specific weight, mass and viscosity. It was found out that in the egg-white no
ammonia was measurable, while in the yolk an unambiguous increase is observable with the time
proceeding. The measured average ammonia content was 1.9 and 20.3 mg/lOOg, for eggs stored
10 and 30 days, respectively [1].
It is more important from microwave point of "view, the change of specific weight, change of
mass. With knowledge of these data the water loss along the ageing, the escape of gas produced
by enzymatic decomposing process, in other words the complex dehydration process, can be well
characterised. As the volume of egg remains unchanged, it is sufficient to know the function of
change of mass versus storage period. Because the egg-white contains 87.8% of water, the task:
"specifYing the freshness of eggs" is the problem of microwave aquametry [2]. Since the water
content of egg decreases with age (by nearly 20% in 30 days [3], see also Figure 1) detection of
water content of egg is a good measure of its freshness. The other information about egg is the
relative dimension (see Figure 2), which can be characterised by the radar cross section [4].
2. THEORY
The attenuation, expressed by markings from Figure 3, which shows the basic setup:
52
Table 1. Specifications of our microstrip antenna arrays used for egg-measurements, and the
typical measurement distances
Number of
elements
2x2
4x4
8x8
MSAType
D=J"2d
[cm]
2.5
5.6
12
Side
dim.
[cm]
3.54
7.9
17.0
Diagonal
dim.
Rm =2D
/'A.
[cm]
11.8
59.3
272.6
Far-field
region
!Measurement distance
[cm]
0.2Rm
0.07Rm
2.4
0.8
11.9
4.2
54.5
19.1
Near field
Two
max.
beams
BW
Gain
[deg]
32
16
8
G[dB]
13.0
19.0
24.8
-r------------.,..40
w = -O.25T + 39.3
35
M[g)
_7,
60
tJ . . . . ~--
55
M=-O.25T+60
--
30
-
__
75
I
I
w[g]
10
15
20
25
30
35
+1
I
I
70l
~+--+-+--+-+--r--r--r-~
't" M[gJ
I
T [days]
II
I
I
I
I
60t
65t
I
Figure 1: Mass and water content change versus time
D/'D2 [cm1
I
I
55.
100
Figure 2: Mass versus relative dimension

Microwave
Absorber
Measuring
\
\
\
\
\
Receiver 2
I
I
I
I
I
G2
..
I
I
I
I
I
I
I
110
\
\
\
\
\
Con veyer-belt
E
~EggS
I
:
J
Egg ,s container
Figure 3: Microwave setup for specifying the freshness of eggs

53
120
130
I
I
I
(1)
where
A: the attenuation due to egg

GT : gain of the transmitter antenna in dB-s
G 1 : gain of the first receiver antenna in dB-s
'A. : wavelength in cm-s
The radar cross section of the egg target:

1t
ReS = o{S) =-. D1(3) D2(3 )."

4
where
(2)
D 1 : the smaller diameter of egg, versus angle 3

D2 : the larger diameter of egg
" : efficiency, referred to a metallic backscatterer. For target dimensions large
compared to the wavelength after [4] p.65 :
,,=
1_,)2
( 1+'
== 0.552 + - 2.6dB
(3)
&'=46
For egg target the reflection attenuation measured in the receiver 2:
(4)
where G2 : gain of the receiver antenna
R2 : the distance marked in FigureI 3
The approximate mass of fresh egg, expressed by its dimensions, as average of numerous
measurements:
Mo [g] = 0.59 (D/D2) - 3.7
(5)
The so-called dehydration (or water-Iossing) function (for Mo = 60 g):

w [g] = 39.3 exp(-0.0076T) == - 0.25T + 39.3
(6)
where T: storage period (in days).

Herewith :finally the change of attenuation in dBs, for eggs examined currently:
(7)
M=-0.033T
which yields 0.23 dB/week or 1dB/30 days.
54
3 . MICROWAVE MEASUREMENT CONSIDERATIONS

Guidelines for choosing the frequency of the measurement:
1. It is practical to choose a high microwave frequency in order to get the attenuation caused by
the egg's water content into the well measurable range (6 - 10 dB).
2. High enough to get higher DtiA values than in the Rayleigh-region (Dl == 0.1 A) or resonance
region (Dl == A) according to the RCS (radar cross section) assuming ordinary egg sizes.
3. A relatively high frequency has to be chosen also to achieve an appropriate beam concentration
with reasonable antenna dimensions.
4. Important to mention, that increasing the frequency the costs of the equipment under
development increase significantly.
Considered all above, the chosen frequency band is at f = 14.2 GHz.
Guidelines for choosing the antennas:
We choose printed (microstrip) antennas for the reason of small dimensions and simple
manufacturing. One of our antennas is shown in a slightly reduced form in Figure 4. A fourth- or
sixteenth part of this linearly polarised antenna seems to be an optimal choice for our present
examinations. The main parameters are given in Table 1. Rut in the table means the begin of the far
field. In the case of rectangular aperture, the axial power density increases as far as 0.2 Rut (which
is advantageous in aspect of the egg's positioning), but there is a minimum at 0.07 Rut because the
antenna beam splits in two, which is to avoid. On the basis of these considerations the distance
between the antennas and the egg as radar target can be chosen. It is noteworthy, that in our
previous examinations (see [2], p. 236, Figure 11.15) a very advantageous power density
distribution has been achieved for circularly polarised microstrip antennas, even for small
measurement distances, which could be important in aspect of reducing the dimensions of the
equipment to be designed.
Further considerations for the geometry of the measurement setup:
The basic measurement setup shown in Figure 3 has been realised to minimize the coupling
between the transmitter-and the receiver antenna, which measures the reflection when the
reflecting egg is not present. It is at this point to mention that we got only a few dB-s smaller
reflection in the case of eggs compared to the diagram values concerning to metallic reflectors,
refraction index of water in egg is high. Eggs were placed into each second
since the (E'
position in the egg-holder, so we ensured that only the measured egg was present in the antenna
beam, moreover it filled out the -1 dB range.
t.5
4. EXPERIMENTAL RESULTS
Figure 5 shows the attenuation measurement versus frequency characteristic. Measured values at
14.2GHz: empty egg (shell):::: O.5dB, No.62 (smallest egg):7.0 dB, No.116 (heaviest egg): 8.6 dB
Figure 6 shows the measured attenuation versus frequency characteristics with zoomed dB-scale.
Measured values for different eggs at 14.2 GHz are: No. 62 (16 days old, broken, M = 41.5 g):
6.70 dB. Eggs of same size but different ages: No. 27 (27 days old, 60.1 g): 7.28 dB, No. 41 (16
days old, 64.4 g): 7.66 dB, No. 97 (2 days old, 68.4 g): 8.05 dB. Largest and heaviest egg:
No.116 (2 days old, 74.7 g): 8.34 dB.These results are self-explanatory and show the egg's
freshness.
55
lor ----.:em=p=ty~eg=g---------,~
,~
:k::
=I
-----= ,
No.62
" 'No.08
8~
,, ,1
f[GflZl,
,
14.1
14.2
I
, (dBJr
I
!
Figure 4: A 8x8 elements microstrip antenna array
70
1 .
t I,
7
versus
~c:
~
----
+!~~~_4--~~~--~~~--~
No.62 (l!roken)
\ 8.0
I,
14.3,
tI
1 .5
Figure 6: Measured attenuation

frequency with zoomed dB scale ~
,
,
No. 116
- - -_ _ _ _
,91
Figure 5: Measured attenuation versus

frequency
,I
14.1
14.2
14.3
'
....:I===============================;
I
Figure 7: Recorded attenuation curves ~:-____

Figure 7 shows the recorded I 8 dB
attenuation curves during continuous \
I
movement (by hand) of the egg-,
I
,
holder. It can be seen well that eggs' 9 t .
,
have an attenuation proportional to'
,
,
their weight (water content) when'!
No.62
they pass under the antenna. The', 10 ~I
(S2g),
broken
,
measured largest deviation in RCS I
I
No.72
was nearly IdB~ between eggs having I 11 ,
(61g)
)
the lowest and highest DI dimensions. !~+-----+s-----1+-0---~I-S----2+-0-----2-7S em
,I
I
I
ACKNOWLEDGEMENT
The author sincerely thanks his PhD students K. Kerekes and L. Nyul for their useful
contribution.
REFERENCES
Nagy J. et al.: "Procedure and Equipment for Specifying the Freshness of Eggs" (in Hungarian), BME-BET,
Budapest, November 1997, o.MFB-OS677960416.
[2] Volgyi F.: "Integrated Microwave Moisture Sensors for Automatic Process Control", Ch. IS in book:
Microwave Aquametry (edited by A Kraszewski), IEEE Press, New York, 1996, ISBN 0-7803-1146-9.
[3J Volgyi F.: "Microwave Measurements for Specifying the Freshness of Eggs", Report (in Hungarian), BMEMIlT, Budapest, November 1998.
[41 Ch.G. Bachman: "Radar Targets", Lexington Books 1982, ISBN 0-669-OS232-9.
[1 J
56
MOISTURE MEASUREMENT IN ORGANIC MATERIAL

WITH IMMERSION MICROWAVE PROBES
Dipl.-Phys. J. Landgraf, Dr.-Ing. A. Goller
hf sensor GmbH, PermoserstraBe 15, D-04318 Leipzig, Germany, www.hf-sensor.com
An essential problem in processing and storage of organic material is a quick and nondestructive determination of the interior moisture content. Specially designed immersion microwave
probes can overcome these difficulties in many cases. For this purpose a series of basically
known microwave antenna and resonator structures where evaluated, tested and optimized.
Results of laboratory measurements and practical applications are given.
Microwave Moisture Measurement, water content, Immersion Probe, Organic Material
1
INTRODUCTION
A quick and nondestructive determination of moisture content in the interior of organic material
is both essential for random testing and optimal process control in many application fields.
Conventional methods are often strongly affected by mechanical and chemical material inhomogenity, grain size and conductivity, are sensitive to corrosion and abrasion, need very smooth
contact areas or don't fit constructive size and shape requirements for use beneath the surface.
Microwave immersion probes have the potential to overcome these disadvantages. They can be
constructed very thin, compact and robust. The interaction volume between the electromagnetic
wave field and the wet material can be made large enough to balance inhomogenities in material
and contact area.
Thus, the goal of this work was to select and optimize microwave antenna and resonator structures best suited for the development of immers~on probes both to complete a pocket moisture
meter and as a monitoring device to control biotechnological processes.
2
CONSTRUCTIVE REQUIREMENTS
Premises for easy portable usage of the immersion sensor head are transportability, light weight,
operation by one (unskilled) person, independence of a clear defined sample volume and shape, a
thin figure suited for piercing into sample material, resistance to chemical and mechanical stress
and electrical stability. Moreover, by driving in the probe step by step a depth profiling of water
content should be possible.
The basic construction of such a sensor head is shown in Fig. 1. It consists of the grip case comprising most of the measuring electronics, a stainless steel tube with 20 mm diameter and the
actual microwave structure surrounded by a tip shaped dielectric cover made from a rigid chemically inert material.
57
stainless steel tube

500... 2000mm
cable with plug
microwave structure with cover
Fig. 1: Basic construction of the immersion microwave probe

Because samples may occur as shred, chips, fibre bundles, mats, bulk goods of various particle
size, pasteous substances etc. structural material inhomogenities are inevitable, and the contact
area between sensor cover and material is irregular too. To get a more integral measuring effect
the interaction volume of the electromagnetic wave and the contact area to the organic material
have to be as large as possible in comparison to the mean particle size. Moreover the measuring
sensitivity should be distributed nearly homogeneously around the tip.
3
CHOICE OF MEASURING PRINCIPLE AND MICROWAVB STRUCTURES
For measuring of water content with microwave methods the corresponding variation of dielectric material properties is utilized. This is done by analyzing the changes of electromagnetic
propagation properties along free transmission lines, TEM lines and wave guides or of parameter
shifts in resonator and antenna arrangements.
The use of free transmission lines is limited to stationary geometric arrangements. The stray field
of open aperiodic TEM line or stripline structures is restricted to a range of a few millimeters.
Therefore they are applicable only in very homogeneous and fine grainy substances.
In the preferred ISM microwave band at 2.45 GHz resonators reach stray field extensions of
some cm, the wave field of antenna is practically limited by the electromagnetic attenuation in
the lossy moist material. Thus in principle both resonators and antenna structures are qualified
for moisture measurement in coarser grainy or fibrous organic substances.
3.1 Stripline Resonator Immersion Probe
In a line resonator partially open to the sample the stray field can be roughly divided into two
parts: The field between the metallic electrodes extends only to regions comparable to the electrodes distance and is concentrated at their edges. Some millimeters away from the conductors
only the more homogeneous field between the charge centers or the voltage maxima of the resonator contributes to the interaction with sample material.
Therefore the strip line resonator immersion probe shown in Fig. 2 is enclosed by a dielectric
cover which produces a definite sample distance and reduces the field concentration at the electrodes edges. Additionally the interaction strength is tunable by cover thickness and dielectric
constant, and mode conversion by loadmg the structure with the dielectric medium is avoided.
58
coaxial cable
tube segment electrodes
substrate with stripline resonator
via
connection
dielectric cover
metallic support ring
Fig. 2: Stripline resonator immersion probe with cylinder shaped electrodes

The active microwave structure consists of two open ended A-stripline resonators connected to a
thin microwave coaxial cable at their middle points by vias and capacitive coupling gaps. Thus
the middle range of the resonator oscillates in antiphase to the ends of the stripes. For symmetry
both stripes are arranged opposite to each other with their hot electrodes directed outward.
To avoid different resonant frequencies by asymmetric loading the ends of the stripes are connected by metallic support rings. Metallic electrodes filling the gap between resonator structures
and dielectric cover ensure a sufficient interaction strength.
3.2 Dipole Antenna Immersion Probe

Another possibility to enlarge the interaction volume is given by the use of radiating antenna
structures, especially rod like dipole antenna. In addition to the near field interaction, which is
the same as for open resonators, these structures emit electromagnetic energy into the surrounding medium. For a dipole inside an unlimited sample volume the resulting feed point impedance is nearly inversely proportional to the square root of the samples effective dielectric constant [1]. This is the basic effect for use as moisture sensor.
However the feed point impedance is also influenced by the receiving of energy reflected back
from sample edges and reaching the antenna again. Therefore the sample volume and the water
content have to be large enough to ensure a sufficient attenuation of propagating waves, that
means a necessary sample volume in the order of 30 ... 100 liters. This requirement has to be kept
in mind for calibration of dipole probes for special materials.
A JJ2-dipole has an almost ideal field distribution, but the mechanical fixing and the feed at the
high impedance termination seemed to be problematic. A solution was found in the mounting of
the dipole rod in prolongation of the inner conductor of an open quarter wave coaxial resonator
serving as impedance transformer. The basic construction of such an antenna sometimes used by
radio enthusiasts is shown in Fig. 3.
The inner conductor of the coaxial cable is taken through the bottom of the resonator and connected to its inner rod. This arrangement allows a matching of the structures impedance by
shifting the feed point on the inner conductor between short and JJ4. For a good measuring sensitivity the diameter of the dipole enlarges conically up to the inner diameter of the dielectric
cover which stabilizes the whole construction mechanically.
59
coaxial cable
jeedpoint
halfwave dipol
quarter wave coaxial resonator
dielectric cover
Fig. 3: Dipole antenna immersion probe
TESTING AND APPLICATION
For evaluation the probes where immersed in moist foam clay ("Seramis") which is well suited
as a model substance because of its density and water storage capacity comparable to many biological materials. Fig. 4 and Fig. 5 show frequency sweeps of the scalar input reflection coefficient Sl1 of the stripline resonator and the dipole antenna probe inside the moist sample for water
contents in the range of 0 to 100% on wet base (dry to pure water).
511 stripline resonator immersion probe
511 dipole antenna inmerslon probe
1,0.,...------,------.------,-----,
1,0 r.==::::::::J=::=:::=::r:=::=::=r::.::::::;--:---""I
0,9
0,8
0,7
0,6
'.
~~~----r-~. .:';~~>/r
0,7
~",<
.... ;._" "'/~7"'-+-:- - - I
l
0,4 + - - - - - - l - - - - - t - I - - - - - t - - - - - - I
2,32
38%
- - - - - pure \Vater
----...1---------
' - - - - ' - - , - - - - - - , - - - - - - , = , . . J ___ ..;. ______ _
__ ~--- -
----- i
--- ---~
~:: +-------f-'i--.."""""'c---~------r-------~I~-:---0,2 -l---+----=li::::::==+==----L-------l
0,3 -.-------;==::===========;---j
--dry
-'-'-' 17%
0,2 +----;
-,,-,,-28%
38%
0,1
-----purewater
2,28
-'-'-' 17%
-,,-,,-28%
:::"==~~=,
-:-~.-.
0,5 +------l----'''''''-,----=1
--::~--+------I
0,0
2,24
--dry
0,1
+----t----t--1--t----+----
2,40
0,0 + - - - - I - - - - / - - - - I - - - - I - - - - j
2,2
2,3
2,4
2,5
2,6
2,7
frequency (GHz)
frequency (GHz)
2,36
Fig. 4: reflection coefficient of stripline resonator probe immersed in moist foam clay
Fig. 5: reflection coefficient of dipole antenna

probe immersed in moist foam clay
For both structures a significant shift of resonance frequency and minimum depth with water
content is clearly visible. Fig. 4 shows also a minimum broadening with increasing water portion. The minimum widths in Fig. 5 are already comparatively large for dry material (radiation
damping). Both structures are easy usable as moisture sensors even at a single frequency.
Very similar results are obtained for other biological substances like bio filter material, compost
or grain. The main difference between both applicators is, as would be expected from function
60
principle, a lower sensitivity to inhomogenities and grain structure effects for the dipole antenna
probe at the expense of a larger necessary sample volume.
Based on the dipole antenna a practically usable moisture probe shown in Fig. 6 was built to
complete a small pocket size moisture meter for spot checking. The same probes where also
utilized for a long tenn monitoring of the water content in a practical bio filter. The Measured
row data in Fig. 7 friendly given by [2] show the good reproducibility and the nearly linear
relationship to water content even for such coarse sample material.
sensor signal In arbitrary units
4000,0
rr=====::::;-r-----,---r---,
- 0 - - sensor
3500,0
--8-- sensor 2 H---+-L-..a-;--:",,-::;;JEJ="'-<>-"'sensor 3
3000,0
2500,0
+=======1===-=-:
!
"
~-I----+------,
, "
;;
;
.. ' ..
+---i--,~-+---+..,...a.:~-+----j
I "
2000,0
...
;:if'
.<> ..
(0"["
+-~,'-'-!'_ _:-..,+-1,,_'_~_._ _'--_ _

.,'
..
'
1500,0~~
1000,0 ' f - - - - - - ' I - - - - - + - - - ! - - - - ; - - - - 1

20,00
30,00
40,00
50,00
60,00
70.00
water content In % (on wet base)
Fig. 6: Test measurements with dipole antenna probe MOIST BIO together with moisture
meter MOIST 100
Fig. 7: working function for 3 different dipole

type immersion probes inside bio filter material
(long tenn experiment), results by [2]
CONCLUSIONS
Specially designed rod like microwave resonators and antenna arrangements are well suited to
build compact immersion moisture probes for pocket size meters and process monitoring.
Measurement accuracy is sufficient for repeated spot checks, depth profiling or online process
control even in grainy or other not to coarse structured organic material.
REFERENCES
[1]
Zinke, 0.; Brunswig, H.: Compendium of High Frequency Engineering, Lehrbuch der Hochfrequenztechnik, Berlin/Gottingen/HeidelberglNew York, Springer Verlag 1965, p. 199
[2]
Patzwahl, S,<I), Nacke, TP): Online Detennination of Water Content in Bio Filter Material, students work,
(1) Fachhochschule Harz - Wernigerode, Gennany, (2) Institute for Bioprocessing and Analytical Measurement Techniques, Heiligenstadt, Germany
Contact point: Jfugen Landgraf, hf sensor GmbH, PermoserstraBe 15, D-04318 Leipzig, Gennany,
Phone: +49-341-2354010, Fax: +49-341-2354154, E-mail: Juergen_Landgraf@hf-sensor.com
61
MICROWAVE DRYING - A RAPID METHOD TO DETERMINE THE

TRUE WATER CONTENT IN DAIRY PRODUCTS
Monilca Walter and Heinz-Dieter Isengard
University ofHohenheim, Institute of Food Technology, Stuttgart, Germany

ABSTRACT: Oven drying is the most frequently used moisture determination method. To obtain
results faster than with the convective heating principle of classical drying ovens, more rapid
heating techniques are used. One of these is the drying by microwaves. The result of all drying
methods is the mass loss under the applied conditions. Only in a few cases does it correspond
exactly to the water content. The mass loss is caused by the evaporation of all volatile matter
during the drying process, not only by water. Particularly foodstuffs risk decomposition reactions
. with formation of volatile substances at higher temperatures. The results of drying processes
depend on the drying parameters and. are variable, whereas the water content is a specific
property of the sample. If the true water content is to be determined by a drying process, the
parameters must be adjusted in a way that the results equal those obtained by a specific method
like the Karl Fischer titration. The dependence on the drying parameters varies for different
products. In man cases a set of parameters can be found to match the mass loss with the analysed
water content, in other situations a differing but reproducible result can be multiplied by a
constant factor to give the reference value. It could be shown for some dairy products that the
water content found by Karl Fischer titration can be reproduced by microwave drying.
Keywords: Microwave drying, water content, dairy products
INTRODUCTION
The water content is for a number of reasons one of the most important general properties of
foodstuffs. In many cases limits are fixed by law. To avoid out-of-specification products the
water, content must therefore be controlled. Even on-line methods exist for this purpose. They
comprise NMR and NIR spectroscopy as well as microwave techniques. These methods do,
however, not measure the water content specifically but a property of the product that depends on
the water content. The correlation between the measured entity and the water content is in most
of the real situations extremely complex. These methods need therefore an extremely productspecific calibration, which must be based on a reference method by convention or definition or
on a specific method [1].
Drying methods are still the most common way to determine the moisture of a product. The term
"water content" was deliberately avoided in this context. Drying techniques do in fact not
determine the water content but a mass loss under certain conditions. This mass loss is not only
caused by water, but by all substances volatile under the drying conditions, be they originally
contained in the sample or be they produced during and by the drying process itself. Strongly
bound water may on the other hand escape detection. As the drying conditions may be
principally freely chosen, the results for a given product are variable. The results of a drying
62
method should therefore not be called water content, even though in some cases (but rarely for
foodstuffs) the mass loss may correspond to the water content. As the convective heating
principle of common drying ovens leads to long determination times, attempts were made to
accelerate the heating process by other heating sources and heat transfer principles. One of these
techniques is the drying by microwaves.
The aim of this work was to investigate for some dairy products if parameters can be found in a
way that the mass loss by microwave drying corresponds to the water content found by Karl
Fischer titration. This would make the determinations of the water content very easy, and only
from time to time would it be necessary to check if the results still match those found by Karl
Fischer titration. A discrepancy can occur when the matrix of the product analysed differs from
the one of the "original" sample or when the technical status of the microwave dryer has
changed. A method to determine the true water content is the Karl Fischer titration [2, 3]. It
served therefore as reference for the measurements. Official drying methods were also carried out
for comparison of the results.
2
METHODS AND MATERIALS
2.1
Samples
The samples were commercially available butter, cream and yoghurt. Several batches of each
product with different water content were examined.
The experiments and results concerning yoghurt are described in detail as example.
2.2
Determination of the water content by Karl Fischer titration
The water content was determined by Karl Fischer titration using the KF-Titrino 701 from
Metrohm, Herisau, Switzerland. The main titration parameters were the following: polarizing
current 50 IlA, stop voltage 250 mY, maximal titration rate 3 mlfmin, minimal volume increment
0.5 Ill. The titrations were stopped using the drift-controlled end-point technique. The drift was
measured before each analysis and a value 1Ill/min above the measured drift was set as end-point
criterion. The consumption due to the drift, taking the titration time into account, was then
automatically deducted from the titration volume. The two-component volumetric technique was
used with Hydranal-Titrant 5 as titrating solution. In the case of yoghurt Hydranal-Solvent
served as working medium. For cream and butter a mixture of Hydranal-Solvent and chloroform
(2:1 by volume) was used as working medium. All chemicals were from Riedel-de Haen, Seelze,
Germany.
2.3
Drying oven techniques
Both German [4] and Swiss [5] official methods were applied. According to German law
yoghurt, cream and butter have to be dried at 102C for at least 2 h. Swiss law prescribes a
temperature of 87C for yoghurt and of 102C for cream and butter; yoghurt and cream have to
be mixed with sand. The drying time is at least 2 h. The drying has in all cases to be continued in
I-hour steps until the mass does not change by more than I mg. The methods for butter or not
identical, even though temperature, minimal drying time and massloss criterion are the same.
The sample sizes are different: approximately 5 g in Germany and about 3 g in Switzerland.
63
2.4
Microwave drying
The LabWave 9000 from CEM, Matthews, North Carolina, U.S.A., distributed by CEM, KampLintfort, Germany was used. The samples are administered on glass-fibre sample pads, if
necessary between two of them. The mass of the pads is registered by the built-in balance and
then automatically taken into account. The mass loss can continuously be followed. Drying
curves can be printed out. The fmal mass is automatically deducted from the initial mass to give
the result of the measurement.
As in household microwave ovens, the frequency of 2450 MHz is used. This is the only one
allowed by international convention. Several working modes and parameters are disposable. The
power is variable. When a lower value than the maximum of 630 W is chosen, the magnetron is
switched off periodically to obtain the desired amount. In the constant-time mode the power and a
fixed drying time are chosen. The determination ends after the chosen time. In the constantweight mode the power and a final mass gradient are set. For the latter, a mass difference between
0.1 and 0.9 mg and a time between 0 and 59 s can be chosen. When 0.1 mg and 10 s are the set
values, the analysis is stopped when the mass loss during the last lOs of measurement reaches a
value of or below 0.1 mg.
To imitate the drying oven methods with mixing the sample with sand, some nncrowave
experiments were also carried out with adding sand to the sample.
3
RESULTS AND DISCUSSION
3.1
Microwave working parameters
The constant-weight mode was preferred to the constant-time mode, because constant drying
times may yield different results when the sample sizes are varied. The gradient 0.1 mg/20 s was
fmally chosen. Less severe criteria led to slightly lower values, as the drying process is
interrupted earlier and when still a certain amount of water has not yet been evaporated. But the
differences are only small: Ten replicates each of the same sample batch yielded 86.070.08 %
with the gradient 0.1 mg/20 s and 85.950.07 % with 0.9 mg/IO s. This shows that the results are
very stable at the end of the drying process. The full microwave power of 630 W was used.
Lower power levels made the determination times longer and gave slightly lower results, because
the drying process is then so slow in the fmal phase that the set mass-loss gradient is reached
earlier. This effect also leads to higher standard deviations for very low power levels. Table 1
shows such results.
Table 1: Influence of the power level on the mass loss and average drying times
of yoghurt (same batch). Constant-weight mode: 0.1 mg/20 s, 10 replicates each
Power level
[%]
Mass loss [%]
srel [%rel]
Drying time
10
30
50
70
90
100
84.260.2
1
0.25
1049
84.580.0
9
0.10
667
84.710.0
5
0.06
456
84.870.0
5
0.06
366
84.970.0
6
0.07
312
85.120.0
3
0.04
301
64
The full microwave power could be used, as no visible product degradation was observed. Even
long exposition to the full power by setting certain durations in the time-constant mode did not
change the result considerably. A sample for which the mass loss was found to be 85.31 % by the
standard procedure (duration about 5 min) had lost 85.48 % of its mass after a set drying time of
60 min. The samples (sample sizes between 2 and 4 g) were administered on the pads with
disposable pipettes.
The samples proved to be stable concerning their mass-loss values. A sample with a mass loss of
85.100.06 % (10 replicates) was re-analysed after 24 h without and with cooling at 8C. The
values (10 replicates each) were 85.060.07 % and 85.150.03 % respectively.
For the determinations with sand glass-fibre crucibles were used. Two special circular glassfibre
pads of the same diameter were laid in the crucible and about the same amount of sand as the
later to be used sample was placed between the pads. A small glass rod was then put on top and
the whole dried using the drying programme. The sample was then added to the sand and stirred
with the glass rod. The mixture was covered with the second pad and the whole dried in the usual
way.
3.2
Comparison of the results of Karl Fischer titration, microwave and oven drying
15 different samples of yoghurt were examined with the all different methods. There were 10
replicates carried out for each sample and each method, except of the oven drying. The standard
deviations given for the oven-drying methods are not really a measure for repeatability, because
the two samples were determined together in the oven and not independently as was the case
with the 10 replicates each of Karl Fischer titration and microwave drying. The standard
deviations of the microwave technique are very small. The oven-dried samples had changed
considerably in colour and were more or less brown after the process. This was also the case,
though to a lesser extent, with the samples that were treated by microwaves with sand. The sand
particles seem to accumulate more energy and transfer it then to the sample which is,
consequently, exposed to higher temperatures.
As the Karl Fischer titration served as reference method, correction factors to obtain the Karl
Fischer values were calculated for every sample (average of the replicates). From all these
numbers the average factor for each method was calculated. These mean factors are given in
Table 2.
Table 2: Average correction factors to obtain the water content of yoghurt
determined by Karl Fischer titration (KFT) from microwave drying (MW)
and oven drying (OD) based on n analyses (g groups with e elements each)
KFT
MW 100 % power, constant weight mode 0.1 mg/20 s
MW as above, with sand
OD 102C (German law)
OD 87C (Swiss law)
n = 150 (g = 15, e = 10)

n = 150 (g = 15, e = 10)
n = 100 (g = 10, e = 10 )
n = 26 (g = 13, e= 2 )
n = 24 (g = 12 , e= 2)
1.0000
0.9992
0.9908
0.9929
0.9960
The microwave method proved to be highly compatible with the Karl Fischer titration and is
obviously capable to reproduce the water content of yoghurt very well. The results are practically
65
identical. The variation with sand is not recommended. Not only is the procedure more
complicated, but also are the results less satisfactory. In classical oven drying, however, the
addition of sand seems to be a good measure. If one of the oven methods should serve as
reference, the correction factors to obtain the "German" or the "Swiss" result from the microwave
value would be 0.9937 and 0.9968 respectively.
CONCLUSION
It could be shown that the mass loss of yoghurt by microwave drying is very close to the water
content found by Karl Fischer titration and also to the results of classical oven drying.
Corresponding results are also found for cream and butter. The microwave technique is very
rapid and. simple to carry out and can therefore be recommended as alternative method to
determine the water content or moisture of the products investigated.
REFERENCES
Isengard, H.-D., Rapid water detennination in foodstuffs. Trends Food Sci. Techno!., 1995,6, 155-162
2
Fischer, K., Neues Verfahren zur rnaBanalytischen Bestimmung des Wassergehaltes von FIUssigkeiten und
festen Korpern. Angew. Chern., 1935,48,394-396.
3
Scholz, E., Karl Fischer Titration. Berlin, Heidelberg, New York, Tokyo, Springer-Verlag, 1984.
4
Bundesgesundheitsamt (editor), Amtliche Sammlung von Untersuchungsverfahren nach 35 Lebensrnittelund Bedarfsgegenstandegesetz (LMBG), Berlin, Koln, Beuth Verlag
5
Schweizerisches Lebensrnittelbuch: Methoden fUr die Untersuchung und Beurteilung von Lebensrnitteln
und Gebrauchsgegenstanden, Bern, Eidgenossische Drucksachen- und Materialzentrale
Contact point:
Monika Walter, Schulstr. 4, D-72175 Dornban, TellFax: +-49 7455-2384, E-Mail: walteruw@fh-albsig.de
Prof Dr. Heinz-Dieter Isengard, University ofHohenheim, Institute of Food Technology
Garbenstr. 25, D-70593 Stuttgart, Gennany.
Tel. +-49 711-397 4670 or +49711-4593285, Fax +49 71l-397 4674 or +49711-4593443,
E-Mail: Isengard@uni-hohenheim.de
66
Poster Session on
PHYSICS AND MEASUREMENTS OF WATER
Chair: Dr. Kurt C. Lawrence,

USDA-ARS-RRC, Athens, GA, U.S.A..

1:30 - 2:30 PM
67
68
MULTIVARIATE CALIBRATION FOR OPTIMIZATION OF A

MICROWAVE MOISTURE CONTENT METHOD FOR GRAIN VARYING
IN TEMPERATURE AND BULK-DENSITY
Douglas D. Archibald, * Samir Trabelsi, Andrzej W. Kraszewski and Stuart 0. Nelson
Quality Assessment Research Unit, Agricultural Research Service, U.S. Department of

Agriculture, Russell Research Center,Athens, GA 30604-5677 U.S.A.
DArch@QARU.ARS.USDA.GOV
ABSTRACT. Multivariate calibration methods are adapted to multifrequency microwave data
for a set of wheat samples that were treated to introduce substantial variations in moisturecontent (M), temperature (K) and bulk-density (D). Three analysis issues are outlined: 1)
linearization of input variables, 2) scaling and weighting of variables, and 3) estimation of model
error. The results of the analysis make it possible to evaluate a wide variety of instrumental
configurations that might be proposed to suit particular engineering criteria such as measurement
accuracy, range of operating conditions, and hardware complexity. Moreover, additional insight
can be obtained by projecting frequency-dependent trends in model error beyond the
instrumental parameters employed in the original study.
Keywords: moisture, grain, microwaves, temperature-compensated, bulk-density independent
1. INTRODUCTION
When designing a microwave-based instrument for a particular noncontact moisture content (M)
sensing application, the engineer has many options to consider: how many and which
frequencies? use of attenuation and/or phase? compensation of temperature and/or bulk-density?
This manuscript describes the computational strategies that were employed to answer such
questions for an in-line sensor for wheat M under conditions where there is substantial variation
in the temperature (K) and bulk density (D) of the flowing grain[I]. Multifrequency microwave
attenuation and phase measurements were made in a laboratory by transmission through lO.4-cm
thick samples of wheat that had been treated to simulate the range of M, D and K that would be
found in the grain handling facility[2]. Subsequent to the microwave measurements of each
specimen, D was determined by weighing a fixed volume, and M was determined by the air-oven
method for whole grain[3]. The overall analysis approach was use of linear multivariate
calibration techniques [4].
2. ANALYSIS ISSUES AND THEIR SOLUTIONS
2.1. Linearization of Input Variables
For the microwave data set in this study, linearization is a critical aspect of the preparation of the
data prior to applying multivariate calibration methods, and this was accomplished by optimizing
a multivariate calibration model that uses all of the measured variables to predict M. The models
of this type are referred to as Al-SP1-sKD, because the measured variables include the following:
microwave attenuation (A) and phase (P) values at eight frequencies between 10.36 and
18.0 GHz, measured sample temperature in Kelvin (K), and measured sample D. Partial leastsquares regression (PLSR)[5], the chosen multivariate calibration method, involves compressing
the many predictor variables, commonly known as the X-data, into an optimally small number of
linear factors that are correlated with reference values, M, known as the y-data. The X-data is
organized into an array of m samples by n variables that are concatenated column vectors. For
69
example, if all the variables are used individually, X = {AI, A2 , .. , As, K, Ph P2 , , P S, D}, the
resulting X-matrix is 279 samples by 18 variables. Application of PLSR generates up to 18 new
variables called PLSR factors that are linear combinations of the original input variables and are
ranked according to how well they describe the variance in the y-vector. For a good choice of
data preprocessing, the first few PLSR factors should describe most of the variance due to
differences in M, while the other factors should describe noise. When such a scheme is applied
to the aforementioned X-matrix example, the results were rather poor as measured by the ability
to predict M and the only gradual lowering of error with increasing model dimensionality (i.e.,
the number of factors used). The reason is that the microwave attenuation (and phase) is highly
nonlinear with respect to moisture under the measured conditions. The type of nonlinearity is
illustrated in the attenuation and phase microwave spectra in Figure 1, where it is shown that
nearly identical spectra can arise from specimens with extremely different M. The nonlinear
response is due to K and D variation. The obvious question is why doesn't the PLSR adequately
-700 r - - - - - - - , - - - - - - - - - r - - - - - - - . , . . - - - - - - - ,
60r-------r-----~~----~------_,
(8)
(A)
-800
50
-900
-1000
m 40
'"<Il<Il
"0
C, -'100
<Il
"0
.,; -1200
'"
~'" -1300
-'400
---
-1500
10
-'600
~ '-0- - - - - - - ' ,'-2- - - - - -......
, 4---------'-'6--------"8
Frequency. GHz
-17001'="'0------7:'2:--------:-'-14:-------'~6-------:-"8
Frequency. GHz
Figure 1. Eight-point microwave attenuation (A) and phase (B) spectra for a random subset of wheat specimens
illustrating that due to K and D variation the offset and slope of the spectra can be similar for specimens with greatly
differing M. The spectral traces are coded according to M: low =dotted, medium =dashed, and high =solid.
compensate for K and D variance when these variables are input to the algorithm as noted above.
Most of the problem is due to variation in D, and at least to a first approximation the cause is
rather simple to understand; more densely packed grain will put relatively more moisture in the
path of the microwaves, resulting in greater attenuation for a fixed M. By the Beer-Lambert law
of absorption for scatter-free transmission across a fixed instrumental pathlength,
A = -log(T)oc C
(1)
where, T is the sample transmittance and C is the analyte concentration.

3
Because C is proportional to units [gmoisture!Cm ] and M is in units of [gmoisture!gwheat(wet)],
M oc
(AID)
(2)
where, D is the bulk-density of the wheat specimen. PLSR does a poor job compensating the
inverse scaling effect of D on measured A. Under the best of conditions each PLSR factor
provides a linear compensation (i.e., modeling of spectral offset and slope), and since M is a
70
nonlinear function of A and D, multiple factors are required to model the phenomenon, which
greatly reduces the ability of PLSR to filter signal from noise.
Fortunately, when D is known, the ratio AID does an excellent job of compensating for the
variations in D. Furthermore, it is also convenient that over the range simulated for the
application, K variation has a nearly linear affect on the microwave attenuation at each measured
frequency. Therefore, temperature compensation is readily achieved by simply concatenating K
as an independent input into the PLSR algorithm, provided some minor adjustments are made to
the mathematical form of the variable, as discussed below. Models that use K and D
compensation require the minimum number of linear PLSR factors and achieve the lowest errors
of M prediction even without including the measured phase values. Phase spectra alone were
poorer at M prediction than attenuation spectra. However, phase data were highly beneficial in
combination with attenuation data, and this had strong parallels with the way in which D data
works in combination with A values. Phase readily corrects for D variation using the ratio AlP
reported by Kraszewski and Kulinski[6]. The similarity between normalization by AID and AlP
was apparent in the magnitude of reduction in M prediction error, the pattern of model
convergence, and the structure of the regression vectors. Moreover, only one phase frequency is
required to normalize attenuation at any of the attenuation frequencies. Because both
normalization methods employ a ratio, there is a strong suggestion that the P values provide the
missing information about sample density. Even without a strict physical model of the processes
involved in microwave transmission through grain, it is easy to rationalize why this is so; the
phase delay in the waves will increase with either the volume of grain or the number of grain/air
interfaces within the transmission path.
2.2. Scaling and Transformation of Variable Weights:
The linearization formulations described above work quite well on their own, but small
adjustments are required to allow variables to be used in combinations that are input into the
multivariate calibration algorithm. There are two problems to be considered: relative scaling of
different types of variables, and relative weighting of variables across their range. Most of the
first problem is solved by autoscaling each variable. Autoscaling is simply adjusting each
variable so that they each have mean zero and standard deviation of one. The latter problem is
apparent by observing the uneven distribution of prediction y-residuals across the M range. The
underlying problem appears to be that the level of uncertainty is different for each datum. One
likely cause is the increase in the relative effect of instrumental noise for very dense or very high
M specimens in which the A is great. Without repeated measurements or a priori knowledge
about uncertainty, the analysis resorted to applying the strategy of transforming the variables
with inverses, logarithms and roots prior to performing autoscaling. These transformations
change the weighting across the variable ranges and the chosen transformation functions were
the ones that generated low and uniform residual prediction errors.
The final functional form of the A1-SP1-sKD data transformation was as follows:
{(-Atfpd 12 , (-A 21P2 )1I2, , (-AslPd 12 , (IlK), 10g(AtfD), log(A2ID), ... , 10g(AsID)}
Multivariate calibration reduces these 17 input variables to 5 relevant factors, two of which
explain most of the variance and were attributed to M and K offset of the microwave
measurements, based on the weights of the regression coefficients for each X variable. The latter
three factors were mainly due to downweighting of frequency channels that performed poorly.
The root-mean-square residual M prediction error is estimated to be between 0.19 and 0.20 %.
71
This is presumed to be close to the accuracy of the M reference method because though the
repeatability is substantially better, the accuracy of whole-grain wheat by the air-oven method
has been reported to be 0.19 % versus the Karl-Fischer method, the primary method for M[7J.
Most of the samples will fall within two standard deviations of their reference M value,
0.39 %. With the transformations optimized for AI-SP1-sKD as a starting point for the form of
the transformation, the study then ranked the performance of eleven other combinations of
variables (APK, APD, AP, AKD, PKD, AK, AD, PK, PD, A, and P). Two important benefits of
linear PLSR modeling are the speed of computation and the fact that the same input data always
yield the same numerical predictions. Consequently it was possible to perform a test of the
performance of all-possible combinations of frequencies in set sizes from one to eight
frequencies for each of the twelve model types. This approach yielded more definitive
conclusions about how many and which frequencies are best for a particular type of model.
Moreover, trends are apparent in the data, particularly for the unifrequency models. For example
the data suggest that the AK model performance may improve further by employing frequencies
higher than the 18.0 GHz maximum that was used in the study.
2.3. Estimation of Model Error
The PLSR analysis hinges on obtaining a good estimate of the ability of models to predict the M
reference data. The basic concept is quite straightforward; simply obtain the root of the mean
squared residual errors (RMSE) between actual (y) and predicted M (y) for a validation set
containing m objects:
(3)
where f; = Yi - y; .
Using the calibration samples to estimate the error of the calibration model is highly prone to
overfitting and generates overly optimistic estimates. The key to estimating the error is to obtain
an independent set that evenly spans the parameter space expected for future samples. Because
of practical considerations in carrying out the measurements, the data collection and reference
measurement sequence could only be partially randomized. Additionally, the data set is
unbalanced in that the extreme density treatments were only carried out on samples at 24C, and
temperature treatments were only made in discrete values (-1,9, 16,24,34, and 42C). Because
of this structure in the data, it was decided to prune and balance prediction sets by analysis of
normalized (M,K,D)-space, formed from the three most important physical parameters
influencing the microwave measurements. First, the 379 initial samples were pruned down to
310 by removing the samples that were closest to other samples. This reduces the chance that a
validation sample is essentially a replicate of a calibration sample. Second, care was taken to
balance the validation sets. The parameter space was divided into 64 volume elements by
dividing each parameter axis into four parts, and a prediction set of 31 elements was formed by
selecting one sample from each of the 31 elements containing more than 2 samples (Figure 2).
Optimization of models was done by using group cross-validation, in which a portion of the
samples was omitted from the model calculations and used as an independent prediction set, with
the process repeated to obtain independent prediction estimates for each sample. Forty of the 64
volume elements contained samples and were used as cross-validation sets. The prediction set
and cross-validation error estimates were very similar in most cases and it was judged that the
cross-validation estimates were better because the selection scheme forced a substantial degree
72
of interpolation and extrapolation and spanned a wider range of variation than the prediction set.
Finally, the validity of each model was also verified by examining the regression vectors to
ensure that the model made physical sense. As an example, some attenuation spectra had a
discontinuity at 12.3 GHz that was responsible for the improvement in the performance of
density- or temperature-independent models. It was presumed that the discontinuity was an
instrumental artifact, and so a poorer performing model with fewer factors was selected because
it lacked weight on the spectral artifact.
(B)
(A)
0
0
; ~~ i g~ ~o
l
~o O
"91
t~
~o Br ~! ~
E 1
.~
,0
::E 1
Ii>0
3J oj
'b9
20
50 0 .7
i<b
1
0
40
~1
'0
::El
<0
10
Density
10
20
30
40
500.7
Density
Temperature
Temperature
Figure 2. Visual representation the two ways that were used to select independent validation samples that span the
entire (M,K,D)-property space of treated wheat samples: (A) prediction set of 31 samples (*) among 279
calibration samples (0); and (B) samples forming one of 40 cross-validation segments (*) split from the 279
calibration samples (0). Both strategies balance the uneven distribution of samples. The cross-validation strategy
spans more property variation because samples with extreme properties are included even if the density of samples
is sparse within that region of the property space.
REFERENCES
[1]
[2]
[3]
[4]
[5]
[6]
[7]
Archibald, D. D., Trabelsi, S., Kraszewski, A W., and Nelson, S. 0., Regression analysis of microwave
spectra for temperature-compensated and density-independent determination of wheat moisture content,
Appl. Spectrosc., 1998,52(11), 1435-1446.
Kraszewski, A, Trabelsi, S. and Nelson, S. 0., Wheat permittivity measurements in free space, J.
Microwave Power and Electromagn. Energy, 1996,31, 135-141.
Am. Soc. Agric. Engrs., St. Joseph, MI, ASAE, 1993, "ASAE Standard: ASAE S352.2: Moisture
measurement - unground grain and seeds", in ASAE Standards, Engineering Practices, and Data, 449-449.
Martens, H. and Nres, T. Multivariate Calibration, New York, NY, John Wiley, 1989.
Wold,S., Ruhe,A, Wold,H., Dunn,W.J.,m, The collinearity problem in linear regression. The partial least
squares (PLS) approach to generalized inverses, SIAM Journal of Science and Statistical Computations,
1984,5,735-743.
Kraszewski, A and Kulinski, S., An improved microwave method of moisture content measurement and
control, IEEE Trans. Ind. Electron. Control Instrum. 1976,23,364-370.
Hart, J. R., Feinstein, L., and Golumbic, C. Oven methods for precise measurement of moisture in seeds.
USDA Marketing Research Report No. 304, Washington, D.C., U.S. Government Printing Office, 1959,
1-16.
73
DETERMINATION OF MOISTURE CONTENT IN WHEAT USING AN

ARTIFICIAL NEURAL NETWORK
Philip G. Bartley, Jr.*, Stuart O. Nelson**, Ronald W McClendon * and Samir Trabelsi**
*The University of Georgia, Dept. of Biological and Agricultural Engineering, Driftmier
Engineering Center, Athens, Georgia 30602-4435, U.S.A. pbartley@bae.uga.edu
**U.S. Dept. of Agriculture, Agricultural Research Center, Russell Research Center,
Athens, GA 30604-5677, U.S.A.
Abstract. An Artificial Neural Network (ANN) was used to determine the moisture content of
hard red winter wheat. The ANN was trained to recognize moisture content in the range from
10.6% to 19.2% (wet basis) from transmission coefficient measurements on samples of wheat.
The measurements were made at 8 microwave frequencies (10 to 18 GHz) on wheat samples of
varying bulk densities (0.72 to 0.88 g/cm3) at 24C. The trained network predicted moisture
content (%) with a mean absolute error of 0.135 (compared to oven-dried measurements).
Keywords: moisture content, artificial neural network, microwave measurements
1. INTRODUCTION
Moisture content is an important factor that affects the quality, price and storage of grain.
Moisture content, in percent, wet basis, is defined as follows:
(1)
where mt is the total mass of the grain sample and IIld is the mass of the dry material. The wet
and dry weights of grain samples are determined by weighing with an analytical balance and
standard oven drying procedures [1]. This technique is time consuming. For unground wheat,
the recommended drying time is 19 hours. Although oven moisture determinations are important
for reference data, the technique is destructive and has no potential for configuration into an online process.
Classical nondestructive techniques for determining moisture content of grain consist of
measuring the electrical properties of the grain in a sample holder and relating these properties to
the moisture content. These techniques have their roots in the early part of the century when
Briggs [2] studied the possibility of rapidly determining the moisture content of grain by
measuring the dc resistance of two brass rod electrodes inserted into the wheat sample. This
resistance was found to vary with moisture content. More recent works in this field are
summarized by Nelson [3]. Modern techniques typically involve placing the wheat in an
appropriate sample holder and determining the moisture content of the wheat from measurements
made on the sample holder. The major drawback to these classical approaches is the fact that
these parameters are not only a function of moisture but also of bulk density, frequency and
temperature [4-11]. Successful implementation of these techniques usually depends on making
independent measurements of the other quantities for accurate determination of moisture content.
74
Of these quantities, bulk density is the most difficult to determine when the grain is flowing.
This situation is of particular interest because of the desirability of measuring the moisture
content of grain when it is being transferred or stored. Several researchers have reported density
independent moisture measurement techniques [12]. Because of the complex nature of the
interaction between the applied electromagnetic field and the grain, air and moisture mixture,
empirical techniques are typically employed. This complexity makes the Artificial Neural
Network (ANN) an attractive candidate as a solution to this problem. In this study, an ANN was
trained to predict the moisture content of hard red winter wheat. The input to the ANN consisted
of free-space transmission coefficient measurements on a layer of wheat placed between two
antennas.
2. METHODS AND TECHNIQUES
2.1 Measurement Method
Figure 1 illustrates the measurement arrangement used to obtain the free-space transmission
coefficients. The measurement technique, the problems associated with such measurements, and
the description of the wheat samples have previously been reported in detail [5].
Nelwor\<
Analyzer
Cable
Transmitting
Antenna
Cable
Sample
ReceMng
Antenna
Figure 1 Transmission coefficient measurement configuration.

The bulk densities of the wheat samples varied from 0.72 to 0.88 g1cm3 (loosely packed to
compacted). Samples were measured at moisture contents ranging from 10.6% to 19.2%.
Transmission coefficient measurements were made at eight measurement frequencies ranging
from 10 to 18 GHz. All measurements were taken at 24C.
2.2. Artificial Neural Network Fundamentals
An ANN is a collection of simple interconnected analog signal processors. The purpose of the
ANN is to provide a mathematical structure that can be trained to map a set of inputs to a set of
outputs. Figure 2 illustrates an ANN similar to the one used in this study.
75
Input 1
Output
Input 2
Figure 2 Artificial Neural Network

This ANN has three layers; the input layer, one hidden layer and the output layer. Each layer
consists of nodes or neurons. Each node has a sigmoid activation function associated with it.
Each interconnection between the nodes has a weight associated with it. The nodes in the hidden
and output layers sum the weighted inputs from sending nodes and apply this net input to the
activation function. The output of the network is determined by applying the inputs and
computing the output from the various node activations and interconnection weights. The
differences between the ANN shown in Figure 2 and the network used are the number of inputs
and number of neurons in the hidden layer.
The feed-forward, backpropagation method was used to train the ANN [13]. This method
consists of initializing the network with random activation levels and weights. Training is
accomplished by adjusting the weights to minimize the error between the predicted ANN outputs
and the observed values. The activations and weights are adjusted by using a gradient-descent
technique. In this study, 179 sample measurements were available. These samples were divided
into a training set of 70 measurements, a test set of 48 measurements and a production set of 61
measurements. The training set was used to train the network. The test set was used in the feedforward mode only during training to determine the optimal point to stop training. This was
done to avoid overfitting the training set. The trained ANN was evaluated based on the
production set exclusively.
3. ANN DESIGN AND TRAINING
The number of inputs determined the number of neurons in the input layer. Each of the
transmission coefficients contains two parts, amplitude and phase. This results in 16 inputs. The
number of neurons in the output layer was one, the moisture content. The number of hidden
layers and the number of neurons associated with each hidden layer are somewhat arbitrary. One
hidden layer was selected. The number of neurons associated with this hidden layer was
determined by training the network for varying numbers of neurons. Of the configurations
considered, 15 hidden nodes delivered the best prediction.
4. RESULTS
The effectiveness of applying the trained network to the production set of measurements is
shown in Table I. As illustrated, there is a strong correlation between the measurements and
moisture content provided by the trained ANN (r2=0.99).
76
TABLE I
Statistics of production set measurements processed by trained neural network with indicated
inputs (errors in percent moisture content)
Statistics
r squared:
Mean squared error:
Mean absolute error:
Min. absolute error:
Max. absolute error:
Transffi. Coeff.
0.993
0.028
0.135
0.004
0.441
AffiQI. Transffi. Coeff.

0.982
0.073
0.219
0.005
0.634
Permittivitx
0.992
0.032
0.142
0.001
0.440
Figure 3 shows the results obtained when all the data (training, test and production sets) were
applied to the trained ANN. The statistics, to the precision shown in Table I, were identical.
of-
c:
~0
:::>
"t;
:'2:
U
(j)
of-
()
'5
(j)
a:
20
19
18
.....
;.,::>#V
17
16
15
14
13
12
11
10
.~+
+,'
.+
.,
..
..
'+
~;
,.";..
10 11 12 13 14 15 16 17 18 19 20
Reference Moisture Content
Figure 3 Moisture contents predicted by trained ANN compared to

oven-dried reference moisture determinations
Next, the ANN was trained using only the amplitude of the transmission coefficient
measurements as the inputs. The statistical results of processing the production set with the
trained ANN are also listed in Table I. Even though the mean absolute error (compared to the
oven-dried measurements) is larger than the one obtained when both the amplitude and phase of
the transmission coefficients were used as inputs to the ANN, this configuration is attractive.
Implementing such a system would require significantly less sophisticated hardware than one
requiring both amplitude and phase measurements. Simple diode detection could be used.
Finally, the procedure was then repeated with values of permittivity calculated from the
transmission coefficient [5] as inputs to the ANN. This has the appeal of making the
determination of moisture content independent of the sample holder and the measurement
technique used to determine permittivity. The statistical results obtained when the production set
of permittivities was processed by the trained ANN are shown in Table I.
77
5. CONCLUSIONS
An Artificial Neural Network can be used to make density-independent moisture content
measurements on hard red winter wheat from free-space transmission coefficient measurements.
The resultant mean absolute error was 0.135 (%) when compared to oven dried measurements.
The mean absolute error increased to 0.219 (%) when only the amplitude of the transmission
coefficient measurements were used as inputs to the ANN. Eliminating the need for the phase
measurements greatly reduces the complexity of the hardware required to make the
measurements. Furthermore, this study indicates that the same level of accuracy (0.142) can be
obtained with calculated values of permittivity. This permits the use of measurement techniques
and sample holders other than those reported. Even though the measurements reported were
made on static samples, the system has potential for on-line, nondestructive moisture content
measurement for flowing grain.
REFERENCES
[1]
ASAE Standards, 43rd ed., S352.2, Moisture measurement-Unground grain and seeds, ASAE, St.
Joseph,MI.1996,p.497.
[2]
Briggs, L. J., "An electrical resistance method for the rapid determination of the moisture content of grain,"
Bureau of Plant Industry Circular, No. 20, U.S. Dept. Agr., 1908.
[3]
Nelson, S. 0., "Dielectric properties of agricultural products-Measurements and applications," Digest of

Literature on Dielectrics, A. de Reggie, Ed., IEEE Trans. Electr. Insul. vol. 26, no. 5, 1991, pp. 845-869.
[4]
Nelson, S. 0. and Stetson, L. E., "Frequency and moisture dependence of the dielectric properties of hard red
winter wheat," J. Agric. Eng. Res., vol. 21,1976, pp. 181-192.
[5]
Kraszewski, A. W., Trabelsi , S. and. Nelson, S. 0, "Wheat permittivity measurements in free space," J.
Microwave Power Electromagn. Energy, vol. 31, no. 3,1996, pp. 135-141.
[6]
Kraszewski, A. W. and Nelson, S. 0, "Density-independent moisture determination in wheat by microwave

measurement," Trans. ASAE, vol. 34, no. 4,1991.
[7]
Kraszewski, A. W. and Nelson, S. 0., "Wheat moisture content and bulk density determination by microwave
parameters measurement," Canadian Agric. Engng., vol. 34, no. 4,1992, pp. 327-335.
[8]
Kraszewski, A. W., "Microwave monitoring of mositure content in grain - further considerations,

"J. Microwave Power and Electromagn. Energy, vol. 23, no. 4,1988, pp. 236-246.
[9]
Kraszewski, A. W., Kubinski, S. and Stosio, Z., "A preliminary study on microwave monitoring of moisture
content in wheat," J. Microwave Power, vol. 12, no. 3,1977, pp. 241-252.
[10] King, R. J., King, K. V. and Woo, K., "Microwave moisture measurement of grains," IEEE Trans. Instrum.
Meas., vol. 41, no. 1, 1992, pp 111-115.
[11] Nelson, S. 0., "Review of factors influencing the dielectric properties of cereal grains," Cereal Chem, vol. 58,
1981, pp. 487-492.
[12] Kraszewski, A. W., "Microwave aquametry-electromagnetic wave interaction with water-containing
materials," New York, IEEE Press, Piscataway, NJ, 1996.
[l3]
Skapura, D. M., Building Neural Networks. ACM Press, New York, New York, 1996.
78
CALIDRATION OF A MICROWAVE SYSTEM

FOR MEASURING GRAIN MOISTURE CONTENT
Mohamed Ben SlimaI, Roman Z. Morawski2, Andrzej Kraszewski3 ,
Andrzej Barwic;' and Stuart O. Nelson3
1
2
Universite du Quebec a Trois-Rivieres, Dept. d'ingenierie, Trois-Rivieres, Quebec, G9A 5H7 Canada
Warsaw University of Technology, Institute of Radioelectronics, 00-665 Warsaw, Poland
U.S. Dept. of Agriculture, R.B. Russell Agricultural Research Center, Athens, GA 30604-5677, USA
ABSTRACT. In this paper, a calibration of a microwave system for grain moisture content
determination based on simultaneous measurements of two parameters, the attenuation and
phase shift, of the electromagnetic wave transmitted through a layer of material is
discussed. The calibration of such a system consists of identifying a function modeling the
relationship between these parameters and the measured moisture content, taking into
account an influence quantity - the temperature of the material, T. Two algorithms for
the calibration are considered. The results of the calibration indicate that the moisture
content of hard red winter wheat can be determined with an uncertainty lower than 0.5%
moisture, at the 95% confidence level, in the moisture range for 10% to 19% at grain
temperatures between _1C and 42C, independent of the grain bulk density.
Keywords: moisture content, calibration, microwave measurements, wheat
I. INTRODUCTION
In agriculture and related industries, the moisture content is the key parameter for
determining grain quality during harvesting, storage, trading, transportation and
processing. Grains of high moisture content readily spoil, while those needlessly overdried
lose nutritional value and can incur physical damage during handling. Thus, the methods
and techniques designed for measuring this quantity are of vital importance for food and
other agriculture-related industries. The moisture content, m, in percent (wet basis) is
defined as:
m=m
mw
+m xlOO
d
(1)
where mw is the mass of water, and md is the mass of dry material. The most widely used
standards method for measuring moisture content in grain are based on oven drying
procedures. They are energy and time consuming, and not suitable for the real-time, on-line
monitoring and control needed in many modern industrial processes. These drawbacks are
relieved by using a free-space microwave techniques which provide a tool for continuous
moisture content monitoring, and have advantages of being contactless and nondestructive
[1, 2].
This paper discusses a technique of monitoring the moisture content m in grain using a
microwave system for moisture determination (MSM) schematically shown in Fig. 1. The
principle of its operation is based on the observation of the influence of water contained in
a sample of grain under study on the attenuation A and phase shift of an electromagnetic
plane wave transmitted through this sample. Unfortunately, those quantities also depend
on the frequency, f, of the wave, as well as on the bulk density p and on the temperature T
of the material. These relationships can be presented in a general form:
f, p)
 = FiJ>(m, T, f, p).
A = F A (m, T,
79
(2)
Static calibration of the MSM consists of identifying the functions F A and

their inverse with respect to m:
m
on the basis of the data
Deal
F~,
= F meA, <P, T, j, p)
= {m~al, A~al, <p~al, T~al, j~al, p~all n = 1, 2,
.. N}.
or rather
(3)
(4)
The main difficulty related to the calibration of the MSM results from the high uncertainty
in those data, and among others, from the uncertainty of the reference value of moisture
content, m~al. To circumvent this difficulty when comparing the methods for calibration of
MSM, a new methodology of algorithmic studies is proposed. It is based on the assumption
that the numerical behavior (properties) of the compared methods depends on:
the qualitative features of the Eqn. (1) rather than on their detailed form and
quantitative characteristics;
the level of uncertainty in the data used for calibration, rather than on the particular
probability distribution they follow.
These assumptions are derived from the previous experience of the authors related to
calibration of various measuring systems [3-6]. In this paper, all the methods compared are
based on the a priori choice of the form of the function F m with a vector of parameters to
be estimated during calibration by means of an algorithm of regularized least-squares.
*
*
II. PROPOSED METHOD OF CALIDRATION

Any method of calibration must involve the solution of two basic problems: choice of the
structure of the mathematical model of the physical relationship (3):
m= F" meA; <P; T;
(5)
j; Pm)
and estimation of the parameter Pm of this model. For the measuring system under
consideration, two models have been tested:
* polynomial model
~ = aoo(T) + alO(T)A + a2o (T)A 2 + aOl (T)<P + an (T)<p2 + a 2I (T)A<P
* rational model
"
aoo(T) + alO(T)A + a2o(T)<P

= 1 + aOl (T) + all (T)A + a 2I (T)<p .
(6)
(7)
Two versions of each model were considered: 1) a version with parameters aij(T) linearly
dependent on temperature (version L):
aij(T) = Pij,O+ Pij,IT,
and a version with parameters aiiT) quadratically dependent on temperature (version Q):
a'J.(T)
= pOJ,
.. 0 + pOJ,
.. IT + p'J,
.. 2T2.
The polynomial model presented by Eqn.(6) for set of data:

D
= {mi;
can be written in matrix form as
Ai; <Pi; T i ; j I i
= 1, 2,
... ,
N}
.;n = H Pm'
(8)
(9)
where the vector Pm = [poo,o I Poo,II IPn,11 Pn,2]T contains the parameters to be
estimates on the basis of the calibration data defined by Eqn.(4). Using the standard leastsquares (LS) method, one would obtain:
Pm = (HTH)-IHTm.
(10)
In order to improve the numerical conditioning of the estimation problem, the following
regularization technique is used:
Pm = (HTH + aI) - IHTm,
(11)
where I is the identity matrix, and a > 0 is the regularization parameter whose value can be
optimized during calibration so as to minimize the error of modeling. For the rational
80
Network Analyzer
Receiving
Antenna
Transmitting
Antenna
Fig. 1. Microwave system for measuring moisture content
(a)
(a)
~5
0
0
85
80
~
<c
....'"
T'
T
T-11
"0
T-2
T-3.
T"2
$~
0000000
75
()
.p
l1li
00
!P'
a Cl O
8'.... 'it:-....
..., *"',...
+.
\f'f"".
...
lIC
xx .+
.~
85
Ie
ao
a~
55
!C
so
<5
12
"
'0
.a
15
Moisture content m (%)
17
IS
...
x. w~~
xl
,.)11 J -:.'"
I
w"'~
'" o06r
12
11
13
14
15
1800
....'".
c
'60
1800
20
::s.
if 1200
Content m (%)
Moisture
IS
.8
17
Fig. 2. Synthetic static characteristics

of the measuring system:
(a) attenuation vs moisture
content, (b) phase shift versus
moisture content.
18
11 0
.0
000
#~
...
ot--
00 0
0'"
++'f'!i'c
..&+....
,.x +
'+ .....",....+~
+... v+ v~
~"'--:
13
+++
.
.
OOcv...
.+ -+-
".....
r.:.... ..,.
+000+0$0
.:+. . -ft:i;. ......':+.*.+

.tii~"'*~~'" +~ ....+
*_.~ ..
:;;~+ .:
12
00
0++
a d" ~r
.;+:,. ++~.~+:~:
lC
v'Q'+ Vv v~
11
0:
x-:~
'2 0
T-34
T-42
-.
1000
T-17
T-2"
o () .. ..I+O$+~
o ~ co;;
+ ...:
o ... ""
CI
.30
T-1
~t.oo
12
'8
(b)
.70Ie
11
11
16
Moisture Content m (%)
.80
vvvvVv
.. _
z.: v
(b)
2000
80.
,p
'"v
vv
VV
.0
""
0
00 "
...
~++
...
+++
flJ
J&
",,'b*i,;
oa~:::*"". .....
6'!i.j:i
,r.........: ;j
o~*
;t
~
70
0"
.
"'
1
+:
.
"';:1
o
00 0
i(
IS
16
Moisture Content m (%)
17
.... +
+
v
v v
18
'"
'"
Fig. 3. The real-world data: (a) attenuation, A, and (b) phase shift, <P,
as functions of moisture' content,
m, for various temperatures, T.
20
model - Eqn.(7), the vector of parameters Pm is estimated using nonlinear least squares
optimization procedure of MatLab called ((Zeastsq" [7].
m. METHODOLOGY AND THE RESULTS OF STUDY

The proposed methodology comprises the following steps:
a rich set of calibration data is fitted to a heuristically chosen functions potentially able
to well approximate the expression in Eqn.(2);
those functions are next used for generation of exact synthetic data in given ranges of m,
T, p and f, in this way several sets of exact data for calibration and for validation are
synthesized;
the exact data for calibration are then disturbed with pseudo-random numbers
simulating the measurement errors of all involved quantities - in this way several sets of
error-corrupted data for calibration are synthesized;
these sets are used for the calibration of the MSM by means of compared methods;
the results of calibration are compared using criteria related to the accuracy of
estimating the moisture content values contained in the sets designed for validation;
the method which performs the best when applied to the synthetic data, is supposed to
be the most appropriate when applied to real-world data.
The compared algorithms of calibration were studied using synthetic data generated by
means of the following model of the MSM:
A = .J0.032707m4 + 1.6578/ T
q,
vP + 0.359/3 + 1.3296 10 -3 m 2T2 + 0.9997
= 17.4628\h3.0203m2 + (1.0431 + 0.1181T)p + 0.11/3
[dB]
[deg]
(12)
(13)
for m E [10; 19.2] (%), p E [ 720; 880] (kg/m ), T E [-1; 42] CC) and / = 16.8 GHz. The
above-defined relationships between A and q, and moisture content m, for several values of
temperature T, are shown in Fig. 2. The studied algorithm was used for the calibration of a
MSM designed for determining the moisture content in hard red winter wheat. The
acquired measurement data are shown in Fig. 3.
The data were divided into two sets: Deal for calibration (odd numbers), and Dual for
validation (even numbers). The results of calibration were then compared with two other
calibrations using the same set of data and based on multiple regression, and designed as
MulReg1 [2] and MulReg2 [8]. The comparison of the algorithms for calibration was
assessed using the mean error of calibration (MEC) and the standard error of calibration
(SEC). The results of this comparison is listed in Table 1.
Table 1. Results of the comparison of the algorithms of calibration in % mc
Algorithm
Polynomial-L
Polynomial-Q
Rational-L
Rational-Q
MulReg1
MulReg2
J-L m
(MEC)
0.0054
-0.0196
-0.0159
-0.0106
-0.0060
0.0100
Urn
(SEC)
0.233
0.230
0.232
0.231
0.277
0.239
It may be pointed out, however, that approach presented in MulReg1 [2J, provides in the
same time values of the grain moisture content and its bulk density (dry mass of grain).
This additional information is obtained at the cost of decreased accuracy in moisture.
82
IV. DISCUSSION AND CONCLUSIONS
The method of moisture content determination in grain discussed in the paper is based on
simultaneous measurements of two parameters of the electromagnetic wave transmitted
through the layer of material, namely, its attenuation, A, and the phase shift (phase delay),
<P. Both these parameters are directly related to the amount of water in the material. Both
parameters, however, are also dependent upon the material temperature and density at the
time of measurements. It has been observed previously [1] that by measuring two
parameters of the electromagnetic wave transmitted through various moist materials, the
effect of the density variation can be minimized, leading to a density-independent
calibration equation. Temperature of the material is usually easy to measure, and it can be
introduced as another variable into the calibration equation, making the measuring system
both temperature- and density-independent. This technique can be applied in frequencies
much lower than reported in the paper, as long as the ionic conductivity of the material is
not a disturbing factor (usually below 1-2 GHz).
Using the calibration procedure described in the paper, the moisture content of hard red
winter wheat can be determined with uncertainty less than 0.5% moisture, at 95%
confidence level, in the moisture range from 10% to 19% at grain temperatures between
_1C and 42C, independent of the grain bulk density. It also should be kept in mind that
the uncertainty related with standard oven method of moisture content determination in
grain may be as high as 0.3% moisture [8]. The advantage of the measuring arrangement
shown in Fig. 1 is that it can be easily adapted for dynamic on-line measurement by
providing the flow of the material between the antennas. As contact between the measuring
equipment and the material is not required during the measurement of the microwave
parameters, this may be a convenient, fast, accurate and nondestructive method of
moisture content determination in flowing materiaL It may also be useful for many
materials other than grain, such as powdered, granulated, or semi-liquid materials.
REFERENCES
[1] Kraszewski, A.W. "Microwave Aquametry - Introduction". In: Microwave Aquametry,

A. Kraszewski (Ed.), IEEE Press, New York, 1996, Ch. 1, pp. 1-34.
[2] Kraszewski, A.W., Trabelsi, S. and Nelson, S.O. "Moisture content determination in
grain by measuring microwave parameters", Meas. Sci. & Technol., vol. 8 (8), pp. 857863, August 1997. Also Addendum: ibidem, vol. 9 (3), pp. 543-544, March 1998.
[3] Barwicz, A., Dion, J.-L. and Morawski, R.Z. "Calibration of an electronic measuring
system for ultrasonic analysis of solutions", IEEE Trans. Instrum. & Meas., voL 39 (6),
pp. 1030-1033, Dec. 1990.
[4] Morawski, R.Z., Lemire, L., Barwicz, A. and Bock, W.J. "Calibration of an electronic
multisensor system for measuring high pressures", ibidem, voL 41 (2), pp. 269-273, Apr.
1992.
[5] Huang, S., Morawski, R.Z. and Barwicz, A. "Static calibration of transducers using
Gauss-function-based approximation", ibidem, voL 45 (3), pp. 389-393, June 1996.
[6] Huang, S., Morawski, R.Z. and Barwicz, A. "Static calibration based on superposition of
splines in one variable", ibidem, vol. 46 (5), Oct. 1997.
[7] Grace, A. "Optimization toolbox for use with MATLAB", Math Works, Inc.
[8] Kraszewski, A.W., Trabelsi, S. and Nelson, S.O. "Simple grain moisture content
determination from microwave measurements", Trans. Amer. Soc. of Agriculture
Engineers, vol. 41 (1), pp. 129-134, Jan.-Feb., 1998.
Contact point: Prof. Andrzej Barwicz, Universite du Quebec a Trois-Rivieres, Dept. d'ingenierie,
C.P. 500 Trois-Rivieres, Quebec, G9A 5H7 Canada. Phone: (819) 376-5071; fax: (819) 376-5152
E-mail: Andrzej_Barwicz@uqtr.uquebec.ca
83
A NEW MICROWAVE METHOD FOR MOISTURE

MEASUREMENT IN BUILDING MATERIALS
Dr.-Ing. A. Goller, Dipl.-Ing. A. Handro, Dipl.-Phys. J. Landgraf
hfsensor GmbH, PennoserstraBe 15, D-04318 Leipzig, Gennany, www.hf-sensor.com
Moisture is one of the most controversial topics in the evaluation of the physical condition of
buildings. The expense of time and destroyment for commonly acknowledged drying and calciumcarbide methods limits their use to spot testing. Conductivity and low frequency capacitance
measurements suffer from salinity.
Microwave probes can overcome often these difficulties. Therefore microwave structures are
analyzed in respect to their usability for water content measurement. Furthermore a ready to use
pocket size moisture meter with interchangeable put on sensor heads is presented which is suited
for different recording depths. Finally some application examples are given.
Microwave Moisture Measurement, Water Content, Suiface Probe, Building Material
1
INTRODUCTION
Because commonly used direct methods for moisture measurement such as drying, calciumcarbide method or Carl-Fischer-titration need the removal of samples from the building, they are
destructive and time consuming. Therefore they are not suitable to get a more detailed view
about water distribution or to locate leakages by moisture mapping.
Additionally measurements often have to be done from one side of the Material as in basements
thus permitting transmission methods or definite sample sizes. Other established methods like
conductivity and low frequency capacitance measurement are strongly influenced by the salinity
due to long tenn water diffusion or additives in building material.
Microwaves with frequencies in the gigahertz range often can overcome this problem due to the
diminishing of salinity effects with the square root of frequency. Thus, the goal of this work was
to select and optimize microwave structures suitable for the development of a moisture meter for
portable use in diagnostics of buildings and construction material.
2
CONSTRUCTIVE REQUIREMENTS
Basic requirements for portable use of a moisture probe are quick and easy operation, usability
from one sample side by one person, light weight, independence of a fixed sample size, electrical
stability and stress resistance against working conditions in the building industry.
Moreover, for scanning of building parts or leakage locating the moisture meter should allow an
ordered storage of measured data according to their local position and give a possibility for data
transmission to a computer. Simple statistical methods like averaging should be available too.
84
Microwaves, as are all other methods, are affected by mechanical and chemical material inhomogenity due to layered wall structures, gaps, differing porosity and coarse surface quality. Thus
the interaction volume between the electromagnetic wave field and the wet material has to be
large enough to balance both short range material inhomogenities due to interior structure or
porosity and the effects of a coarse contact area. At the other hand the interaction volume should
be limited to a certain volume, area or depth to be investigated. Because these requirements
differ widely from case to case they can't be fit by one single microwave structure. Therefore a
measuring setup with interchangeable sensor heads according to Fig. 1 was chosen.
The setup consists of a pocket size meter controlling the connected sensor heads and displaying
the measured data and variable moisture probes connected with the meter by short cables. The
moisture probes contain disk shaped planar microwave structures as a closure of a grip like
metallic case comprising the high frequency detection electronics. The active microwave sensors
are slightly lowered in the front side of the cases to realize definite sample distances.
cable with.::p~lu_g_ _ _ _ _ _ _IIIilb......
DDD
D
resonator
structure
Fig. 1: Measuring setup with pocket meter and two interchangeable put on moisture probes
CHOICE OF MEASURING PRINCIPLE AND MICROWAVE STRUCTURES
Microwave moisture sensing methods utilize the variation of dielectric material properties with
water content by analyzing the changes of electromagnetic propagation properties along lines or
of parameter shifts in resonator and antenna arrangements.
For one side measurements at planar samples and a certain ability to balance short range inhomogenities in material composition and contact area only planar antenna and resonator structures
are usable. In the preferred ISM microwave band at 2.45 GHz specially designed resonators
reach stray field extensions of some cm and can integrate over inhomogenities in the millimeter
range. The wave field of antenna is limited by the electromagnetic attenuation length in the lossy
moist material, commonly some decimeters. Thus in principle both resonators and antenna
structures are suitable for moisture measurement in building constructions.
85
3.1 Stripeline Resonator Surface Probe
Resonators in stripeline technique are particularly suitable as moisture sensing devices due to
their inherent planar geometry, their robustness against deformations and their easy and reliable
production. The electric field between the front and backplane conductors is concentrated mostly
in the interior of the substrate. The middle range stray field in a distance of some millimeters,
however, is mostly due to oppositely charged line segments of the resonator. Therefore the extension of this field is comparable to the distance between oppositely oscillating charge centers
having a maximum distance of half the wavelength at the microwave frequency (substrate dependent, 3-5 cm at 2.45 GHz).
By an air gap due to a lowered mounting of the active microwave structure in the front plane of
the probe case as shown in Fig. 1 only this field part interacts with the sample. That results in a
balancing of small inhomogenities in the millimeter range. Additionally a mode conversion
caused by to much loading of the resonator is avoided. A further averaging effect is achieved by
electrically coupling a few of such oscillating line segments or rings spread over a larger region.
For a frequency of2.45 GHz a useful averaging area is in the range of30 to 60 mm in diameter,
small enough to resolve the individual moisture content of single bricks.
Fig. 2 shows the front part of such a stripline resonator probe. A disc made from commonly used
copper-plated substrate closes the grip like probe case. The substrate disc tears two symmetric
resonator rings in microstripeline technique. The resonator rings are feed from the circuitry
contained inside the case by a via and coupling gaps. The sensor head is connected with the
pocket meter by a short cable thus allowing its easy use at not so convenient available positions
near the floor or at the ceiling.
symmetric
copper-plated ring structure
substrate
copper-plated
substrate
patch
antenna
plate-shaped
Jrontend
via tofeed
point
Fig. 2: front end of strip line resonator probe
Fig. 3: front end of patch antenna probe
A highly symmetric resonator structure like that in Fig. 2 prevents sample thickness dependent
oscillations of sensor signal due to a reflection of radiated microwaves discussed in the following
antenna part. Without radiation effects resonator probes achieve a recording depth of 1-3 cm
depending on material and water content. Thus they are well suited for near surface moisture
86
measurement or for the inspection of layered structures like plastered walls. Due to the balancing
of moisture content by diffusion this also allows an indirect mapping of deeper lying inhomogenious building parts otherwise not accessible.
3.2 Patch Antenna Volume Probe
The use of radiating antenna structures gives larger recording depth in comparison to nonradiating resonators. This allows a measurement of the interior moisture content of thicker walls or
basements. Due to their planarity and simple construction a patch antenna in microstripeline
technique [1] shown in Fig. 3 is well suited for this purpose.
In addition to the near field interaction, which is the same as for resonators, an antenna emits
electromagnetic waves into the sample medium. For a patch antenna placed above the sample
surface with a definite air gap the resulting feed point impedance measurably depends on the
samples dielectric constant. This is the basic effect for use as moisture sensor. The width of the
air gap thereby can be used to tune the interaction strength between antenna and sample materiaL
However the feed point impedance of an antenna depends also on received energy reflected from
the sample's rear side. Therefore the samples have to be thick enough to ensure a sufficient
attenuation of back reflected waves. This requires a sample thickness of 20-50 cm, the lower
value for larger densities and a moisture content of at least a few percents, the higher thickness
for dry low density substances like heat insulation materiaL In addition the available sample area
should be the same size due to the divergence of the radiation beam.
The basic construction of the patch antenna probe in Fig. 3 is very similar to that of the resonator
probe. The only visible difference is the plate-shaped front end .of the case around the antenna
which improves the directional radiation pattern and indicates the minimal needed sample area.
Additionally it ensures a fixed antenna to sample distance.
4
TESTING AND APPLICATION
For both probes and a variety of building materials like bricks, porous concrete, sand stone or
wood the moisture dependence was investigated systematically. Therefore a network analyzer
HP8752A was used to take frequency sweep curves of complex reflection coefficient for all
substances and with different moisture contents. It was found that the shifts in the resonance
parameters are large enough for a moisture determination at one fixed frequency material specifically chosen.
The measured reflection coefficient then can be correlated to moisture content by interpolation
between experimentally determined data points calibrated by drying as reference method. Fig. 4
shows such a work function for the resonator type probe. The deviations of the single measurements are due to inhomogenities in material and surface contact.
The numerical interpolation algorithm and the calibration points are included in the microcontroller program code running in the pocket moist meter. Thus after menu driven choice of substance to be investigated the meter directly gives the moisture percentage inside the sample.
Additionally the meter allows an ordered storage of measured data in tables of variable size.
Because each measurement needs only some seconds a building part can be scanned in a short
87
time. An online averaging of up to 16 measurement shots improves the reliable at coarse surfaces. Later on the data arrays can be transmitted to a computer by a standard serial interface.
Fig. 5 shows the moisture distribution in a basement wall measured with the patch antenna volume probe in an array of 20 cm mesh width. The data are online averaged from three points
nearly at the same position. A leakage in the lower left comer and a general higher moisture near
floor are clearly visible.
sensor raw signal at brick material
3000.0 . , . . - - - r - - - - - - , . - - - - , - - - - . , . . . - - - - - ,
rmislure
(on wet
base)
o
2500,0
2000,0
+-----!i----+-Q--r-'-+~--+------i
1500.0
+--~f___,~--+---+---+-----I
10%
8%
6%
4%
1000.0
2%
0%
500,0 :;:----:i----+--'-+---+------i
hightabove
floor in m
0,0 +-----'1-----+---+---+-----1
6,0%
8,0%
10,0%
4,0%
0,0%
2,0%
distance from
0,9 0,1
comer in m
water content in % (on wet base)
Fig. 4: work function of resonator probe
Fig. 5: moisture distribution in a basement wall
A combined use of both probes in the same building area can help to localize the source of
moisture. By comparing the moisture content near surface and in the depth a leakage in insulation can be distinguished from condensed water caused by humidity.
5
CONCLUSIONS
Specially designed planar microwave resonators and antenna structures are well suited to build
compact probe heads for pocket size moisture meters. Both principles complement one another
very well in recording depth, averaging range and application fields.
Measurement accuracy is sufficient for repeated spot checks or leakage localization. The short
measuring time and an ergonomic operation with data arrays enable an investigation of water
distribution by moisture mapping of whole building parts.
REFERENCES
[1]
Meinke; Gundlach: Pocket-Book High Frequency Technique, Taschenbuch Hochfrequenztechnik, ed. by

Lange, K. and Locherer, K.-H., BerliniGottingen/HeidelbergINew York, Springer Verlag 1992, p. N24
Contact point: Dr. Arndt Goller, hf sensor GmbH, PermoserstraJ3e 15, D-04318 Leipzig, Germany,
Phone: +49-341-2354010, Fax: +49-341-2354154, E-mail: Arndt_Goeller@hf-sensor.com
88
THE APPLICATION OF ELECTROMAGNETIC

FIELD ON WATER STRUCTURE AND BIOLOGY
HouQun
Qian Tong Hui
Song Quan Xiang
Auto Control Dept. Hankon Branch ofHua Zhong University of Science and Technology
Wuhan Hu Bei Province (430012 P. R. Of China)
ABSTRACT. This paper presents the effects of the electromagnetic field on water structure, the
changes of structure inside water molecule and many biological effects. The effects are applied
in clinical practice, medical research, making wine, etc.
Keyword: Electromagnetic, Water structure, Biological effects
I.INTRODUCTION
The water molecule is a polar molecule, The electrons interact between water molecules, and
these molecules are connected by the hydrogen bond. The angle between the O-H bond is
10445'. The gravity centers of the positive charge and the negative charge does not coincide. So
the electron moment of water molecule is powerful. It forms the sequence structure of water
molecules. But the hydrogen bond between molecules is interrupted frequently by the hot
movement of molecules, and it destroys the sequent status to some degree. Eventually, the pure
water structure keeps balance between sequence and unsequence statistically.
Ordinary water is not pure as it contains dissolved salts and gases. Because the water molecule is
polar, it should make the ions insteady status. The reason is that the ions in water can form
hydrate, that is to say, each negative ion can draw the positive end of water molecule neighbored
and make the ions in good steady status.
Microwaves are electromagnetic waves, which combine the electric field energy and magnetic
field energy. But only the electric field energy effects water structure, because there is no
magnetic material in water. So, when discussing the effects of the electromagnetic field on water
structure, we just need to consider the influence of the electric field energy. From experiment, the
electric field would breaks the O-H bond of water molecule and proton results in excessive extraoxygen negative free radical.
2.THE EFFECT OF ELECTRIC FIELD AT WATER STRUCTURE.
Besides of IDO molecules, in water there are ID02 molecules, and 02 molecules, as well as
some gas molecules and other impurities. In water, hydrogen molecules appear as molecule or
ion, as it can be known from testing the conductivity of water. Because the amount of oxygen has
a critical effect on life appearance, the starting point of the study is the variety of the oxygen
amount. Oxygen molecule has an absorbing peak near ultraviolet band at wavelength: 194 nm
and 210 nm. So we can get the variation of the oxygen amount in water by using ultraviolet
radiation meter. Moreover, we can determine the oxygen density in water pre- or post-processing
electric field by special oxygen amount and ph value testing.
89
In order to find the reason of oxygen amount variation, by using a numerical chemical method,
we can get the density ofID02 in water and the information on the reaction process.
The experiments have showed that:
(1) The amount of oxygen in water will increase with the intensity of the electric field and
reaction time.
(2) Under the electric field influence there are excessive ionic hydrate that are extra-oxygen
negative free radical and proton. They are stable for a long time.
(3) Except for the excessive ionic negative free radical, there are excessive ID02.
(4) Water molecule is polar molecule, it can get additional electrical energy to move repeatedly
at its balance position under the influence of the electric field.
(5) The appropriate amount of negative free radicals in creature has a function of metabolism
energy storage, transformation, excretion, protection and disinfection. But when it is excessive, it
will break nucleic acid and carbohydrate, which will damage the creature.
(6) The constant or variable electric field of the same power has the same result in treating
water.
3. THE APPLICATION OF ELECTROMAGNETIC FIELD
(1) The application of the electromagnetic fields for biological effects.
As we know, human body can be regarded as a physiological saline-filled big vessel which has
resistance and capacitance. Under the influence of the electromagnetic field, the electrons of
polar molecule like IDO in biological tissue polarized, and collide, rub each other, producing
heat. In the same time, ions in biological tissue move and produce heat and conductive current, it
heats biological tissues. On this basis the electromagnetic waves are applied in a clinical practice
and thermotherapy.
(2) The application of the electromagnetic field on alcohol.
Under the reaction of alternative electric field in microwave band, the polar molecules in alcohol
vibrate in two directions and accelerate the combining of molecule sets. After the polar
molecules absorbed energy of the electromagnetic field, a lot of free radical is produced and they
promote the chemical reaction in alcohol. The experiments have proved that the new liquor
which is affected for one second by the electric field is equal to a wine aged naturally for three or
five months.
4. SUMMARY
After study of the effect of electromagnetic field on water structure, many effects have been
found, and they have been used in agriculture and industry. The mechanism of electromagnetic
field on biological tissue is complicated. The study of it brings a new subject involving
electromagnetics, biology, and medicine - and its called: Bio-electromagnetics.
REFERENCES
[1]
[2]
Cheng Jia Shen. The Study On electrofield application in water. Physics, 1995, No.7, pp. 15-17.
Zhou Xing Wang. Bio-electromagnetics. Physics, 1995, No.1, pp. lO2-125.
90
NON-DESTRUCTIVE MICROSTRlP RESONATOR TECHNIQUE

FOR MOISTURE/PERMITTMTY MEASUREMENT
K.K. Joshi* and R.D. Pollard**
* Dept. of Electronics, Sir Parshurambhau College, Tilak Road, Pune-411030, India;
phone: +91-212-544392, Fax: +91-212-532479/353899; e-mail: kalpana@iucaa.ernet.in
** Dept. of Electronics and Electrical Engineering, The University of Leeds, Leeds LS2 9JT, UK;
phone: +44-113-233-2080; Fax: +44-113-233-2032.
Abstract
Moisture content of various agricultural and industrial materials like grain, seeds,
food products, rubbers, foams, soap cakes, pharmaceutical products, etc. is an important
factor for storage, shelf life and quality control. Microwave moisture meters with
stripline sensors measure moisture using microwave power loss in the material and they
are commercially available in the market. They are subject to errors due to the shift in
the solid state components. Waveguide resonant cavity techniques have been widely
used to perform nondestructive moisture measurements as well as complex permittivity
of materials. The waveguide cavity technique is associated with various difficulties like
loading and unloading of samples even though it is considered as an accurate technique.
Microwave attenuation based techniques in practice are being used as on-line sensors
using striplines, coaxial lines and free-space propagation. Microwave resonator technique
is a powerful tool for the measurement of moisture/dielectric constant of the material
under test as the frequency measurement is not subject to the device errors like drift
and repeatability with the availability of modern network analyzers.
The present paper reports on accurate and fast moisture/complex permittivity
measurement using a micros trip resonator (patent pending). The material under test is
kept as an overlay on the microstrip surface. The resonant frequency and the quality
factor of loaded microstrip resonator are measured. The change in the effective
permittivity of the sensor and the permittivity of the material under test is determined
with the help of electromagnetic numerical analysis.
The electromagnetic software may be used to predict the behavior of the sensor
in-situ moist conditions using the permittivity data of dry materials and dielectric
mixture equations. The measured frequency shift and quality factor data will be
numerically analyzed to give the unknown complex dielectric constant and hence
moisture content of the sample placed as an overlay on the microstrip resonator.
Alternatively, the simulated or postprocessed experimental calibration data can be
stored in the computer to give a directly correlated moisture percentage. Simulation of
calibration curves and sensitivity analysis may be realized with the present numerical
technique hence simplifying calibration process. The dielectric constant is measured
with the accuracy of 1.15% and 0.15%, if the frequency uncertainty of the measuring
device are 25MHz and 1 Hz, respectively.
91
NONCLASSICAL EFFECTS OF THE MICROWAVE INTERACTION WITH

AQUEOUS SYSTEMS
Anatoly F. Korolev, Alexander S. Vshivtsev,
Physical Department, Moscow State University, Moscow, 119899, Russia
Igor Timoshkin
Imerial College of Science, Technology and Medicine,
Huxley School of Environment, Prince Consort Road, SW7 2BP London,UK
i. timoshkin@ic.ac.uk
Andriano Pulino
Gamma Tel., Cambrige, MA, 02139
Abstract
Water represents the complex associated liquid structure, consisting of molecules combined in complex
dynamical network by hydrogen bonds. In the present paper the energetic levels of the proton in the Hbond and frequency of tunneling of proton have been calculated. Obtained frequencies are in the
millimeter wave range.
Introduction
A lot of papers describing the results of microwaves and millimeter wave radiation influence on biological
objects were published recently. Experiments performed with the different biological systems show that
this influence has resonant character. Moreover, medical investigation in different spectral bands carried
out show that, we can observe quite contrary effects - from therapeutical up to aggravating decease.
The thermal microwave effects are used in the applications for the on cological illness treatment, when
the given tissue area is heated by microwave up to the temperatures of malign formation destructure. The
effect of the nonthermal influence is not so common in medical practice at the first one, nevertheless, it is
used now. These applications are based on numerous facts of the biological effect of the exterior signals
of, the so called, informational (nonthermal) nature. We cannot point evident specific receptor systems
them.
The search of the primordial receptor for such signals is usually done in the membranous system of the
biological macromolecular complexes. Such effects arise during the influence of the low intensity
electromagnetic fields of various spectral bands: UV, visible light, infrared radiation. These effects are
very similar and, as a rule, are of common term "biostimulation". It points that, there must be a common
element in the response chains making biological system reaction identical on the heterogeneous signals.
Results
In the context of contemporary view on the biological effects of the low intensity millimeter wave
radiation are opposed to thermal. The creation of nonthermal influence is that, the elevation of
temperatures of the defined object is not more than 0.1 K. In that case the density of the power flow
usually does not exceed 10 mWt/cm.
92
We assume that a hydrogen bond between two water molecules under special condition can be the
receptor of electromagnetic radiation. Hydrogen bond owns strong polarisability and even a weak
influence can cause a system response [1]. The modification of the water structure in that case, can be
responsible for the changes of biological processes in it.
Hydrogen bond does not impose rigid limitations on system geometry that is why the potential ensuring it
.
can be of different form.
U(x)
~r
Figure 1. Double symmetric potential U(x) for proton n+ in hydrogen bridge between two water
molecules.
In cases of tunneling proton the ground proton energy level Eo splits into two sublevels E01 and 02 . The
magnitude of splitting depends on proton tunneling frequency. We have to check the hypothesis that,
splitting frequency according to our calculations can be in millimeter wavelength band. This situation can
be realized in symmetrical potential with two negative peaks, formed by two molecules connected by
hydrogen bond (Figure 1).
The fact of the matter this problem is that one finds the exact solution of non-linear Schrodinger equation
in this symmetric potential [2,3]. We shall consider this problem ofa following type
(1)
93
where A. is eigenvalue, parameter L1 r is determined by relative values: barrier height and distance between
weels. The potential in this Schrodinger equation has the form
with 80=1.
At the Figure 2 we present the spectrum lines of proton in double weel potential, for first 4 energy
levels, n=0,1,2,3. This spectrum lines have been calculated by Eq.(1). For obtaining the solution of initial
113
problem with the potential (2) with 80 *0, it is necessary to make the scale transformation En= 80 An.
Here En is the energy of proton at the energy level with the number n. The frequency COn , which
corresponds to the transition between the energy levels can be calculated be the simple relation con=(EnEro)/h. These frequencies depend on the height of potential barrier Vo and parameter ~ r, and lies in the
millimeter range.
The ground energy level of proton Eo splits into two closely lying levels due to tunneling of the proton
through the potential barrier Vo. The frequency of tunneling can be obtained by Eq. 3
il
OJtun
mil
= -;exp (- 4h Ltr)
(3)
where n is the frequency of proton vibration in the potential. Hence, the frequency of proton tunneling in
liquid water is approximatly 178 GHz. For ice D 20 this frequency is 20 GHz.
25
20
n=3
--
N"' 15
C\I
....
<l
n=2
10
n=1
n=O
0
0
(11 r /2)
Figure 2. Proton energetic levels Eo in the double symmetric potential.

94
So, the frequencies of proton in the hydrogen-bond have been calculated by exact solving of the of nonlinear Schrodinger equation with the symmetric potential. These frequencies strongly depend on the
parameter of the H-bond potential. Hydrogen bonds in liquid water have a wide range of variation of
these parameters, this fact provides the continuous spectrum of absorption of millimeter waves.
References
1. Bricman J. Proton motion in hydrogen bonds. North-Holland Publ. Co., Amsterdam, 1976.
2. Vshivtsev A. S., Norin N.V., Sorokin V.N., Izvestiya Vyzov, Physics, 1996, N5, pp.55-70
3. Vshivtsev A. S., Norin N.V., Sorokin V.N., Theoretical and mathematical physics, 1996, v.109,
pp.107-123.
95
NONDESTRUCTIVE AND IN-SITU MEASUREMENTS OF WATER CONTENT

IN PLANT TISSUE USING TIME - DOMAIN REFLECTOMETRY
N.J. Livingston} S. Pepin} W.R. Hook} D. Leight and C. Bassey
Department of Biology, University of Victoria, P.O. Box 3020,
Victoria, B.C. Canada
ABSTRACT. Determination of water content were made in variety of plant material,

including tree stems, pollen, embryogenic tissue cultures and broad leaves, using time
domain reflectometry (TDR) and an automatic network analyzer (ANA). In all cases
there was a linear relation between the travel time of the reflected pulse and water
content over a wide range of water content from 0.05 to 0.98. This relationship was not
significantly different to that predicted by a simple dielectric mixing model. TDR
measurements involved the use of remotely shorted diodes and wave-form subtraction
techniques that allowed unequivocal determination of pulse reflections.
Measurements of leaf water content were made with two-diode stripline probes
constructed from a single-sided printed circuit board. The spiral arrangement of the
copper traces provides a large "effective" transmission line length but minimizes the
surface area (less than 5 cm 2 ) of the probe. Spiral probes were clamped to the adaxial or
abaxial surface of leaves for period of up to 10 days. Changes in leaf water content of less
than 0.05 were detected. In separate experiments in which leaves were severed from
plants, placed on a balance and dehydrated in air, there was excellent agreement between
TDR determinations of water content and those determined gravimetrically.
Profiles of the radial distribution of water in tree stems were obtained using segmented
cylindrical probes (11 mm in diameter) that were inserted into the tree. Again, there was
very good agreement between gravimetric and TDR determinations of water content.
96
DIELECTRIC STUDY ON DYNAMICAL STRUCTURE OF WATER IN

MOIST MATERIALS USING DIELECTRIC RESONATOR
Jean F. Rouleau, Jacques Goyette, and Tapan K. Bose
Institut de Recherche sur l'Hydrogene, Universite du Quebec aTrois-Rivieres
c.P.500 Trois-Rivieres, Quebec, G9A 5H7 CANADA
Shin Yagihara* and Naoki Shinyashiki
Department of Physics, Tokai University, Hiratsuka, Kanagawa, 259-1292 JAPAN
Abstract. The dielectric resonator technique was used to study the dielectric behavior of aqueous
polymer systems. Shifts in the resonant frequency and the transmission coefficient suggest
enough sensitivity even when water content is very low. Comparisons of preliminary results with
relaxation measurements performed at higher water content imply free volume parameters related
to the relaxation time change for the low water content region. The chain dynamics restricted in
that range also confines free water molecules.
Keywords: dielectric resonator, dielectric relaxation, water, poly(vinylpyrrolidone).
1. INTRODUCTION
Microwave dielectric measurement techniques have been extensively used in the last decade for
studying various aqueous systems of associated liquids, polymers, and biological materials. The
dielectric behavior of water is affected by molecular interactions in those materials. For example,
the relaxation time of free water in polymer/water system is longer than that of bulk water,
reflecting a decrease of the free volume. 1,2 Furthermore, bound water3 of biomolecules has 100
times longer relaxation time, and strongly bound water in calcium silicate hydrate4 have been
observed with still larger relaxation time. Therefore, in order to detect water content in moist
materials, the distinction of these dynamics is required to develop a viable detection system.
The dielectric resonator technique has been develoEed recently to estimate water content of
various materials, like humidity detection in gases.,6 This technique is highly sensitive and
nondestructive for solid-like samples. 7,8 However, the dynamical behavior of water has not been
discussed enough in the resonance study so far. Poly(vinylpyrrolidone) (PVP)/water system is a
typical aqueous system without bound water. 1,2 Data extrapolation of the relaxation time of water
in this system suggests that the relaxation time for a dry sample is 300 times larger than pure
water. However, these behaviors have not been clearly understood yet because of lack of
measurements in the low water content region. Also, physical contact of an electrode to the
sample required in conventional dielectric measurements is not well suited for these solid-like
and fragile samples. Thus the dielectric resonator technique without physical contact should be
useful.
In this work, a comparison of the dielectric resonance measurement with the dielectric relaxation
measurement is performed for PVP/water system for low water content range in order to clarify
how the dynamical behavior of water molecules is reflected in the results of the resonance
measurement.
97
2. EXPERIMENTAL
PVP with a molecular weight of 40,000 has been purchased from Aldrich. The powder samples
were dissolved in pure water with concentrations of 25-33wt% without further purification. Film
samples were obtained by casting from solutions under ambient laboratory conditions. Each
sample was kept in a sealed container to control water content. Measurements of weight and
resonance were achieved within a few minutes to maintain moisture content. Water content was
determined from sample weight with a Sartorius balance ( 10-4 g).
A resonant mode dielectrometer manufactured by GDK Products (model RMD-C-080) was used
as a cylindrical resonator. 7 The inner diameter and the height of the resonator are 20.32mm and
16.36mm, respectively. A slab of alumina with a thickness of 2.00mm was inserted between the
upper and lower parts of the resonator in order to keep a high quality factor. Two PVP samples
with thicknesses of 1.32mm (sample a) and 0.61Omm (sample b) were carefully located, one by
one, at the center of the alumina disk, and measurements of the frequency shift (M) and the
transmission coefficient (S21) were perfonned at 25C using a network analyzer (HP851OB).
Time domain reflectometry (TDR) measurements 1,2 based on a reflected signal have already been
perfonned for PVP/water systems with various water contents above 15wt%.
3. RESULTS AND DISCUSSION
The resonant frequency and the transmission coefficient are shown at various weight fractions of
water w in Figures 1 and 2. Both quantities are decreasing with water content because of
increasing dielectric constant and loss. Figs. 1 and 2 also suggest enough sensitivity to measure
the permittivity of moist PVP, if relationships between these quantities and complex permittivity,
(e*), are obtained empirically. However, it is generally quite difficult to predict dielectric
behaviors at a single resonant frequency, since the dynamical behaviors of water make the
9,86
-1O
J:
"-
C}
--E>' 9,82
cQ)
::s
---c
'(3
-...
o sa.rT'4Jle a
(/)
...
Q)
.,
eo0
'w
(f)
'E -25
c
(f)
00
0
.sampleb
-20
eo 0
00
Q)
00
~ 9,78
o
0
o
-15
Q)
00 00
cQ)
IIJ
"'0
-...
<U
o sample a
sample b
-30
9,74
0,05
0,1
0,15
0,2
water content
Fig. 1. Resonant frequency is for moist PVP

at various water contents W'
0,05
0,1
0,15
0,2
water content
Fig. 2. Transmission coefficient S21s for

moist PVP at various water contents w.
98
estimation of e* complicated. Here we consider the former data from TDR which has been
already reported,I,2 and we compare them with the present results. TDR data shows the
concentration dependencies of relaxation parameters obtained from higher frequency relaxation
process due to free water in PVP (Mn :10,OOO, M n :40,OOO) / water systems. Relaxation parameters
show no molecular weight dependency for this relaxation process.
PVP is a typical aqueous polymer, for which bound water is not observed. The relaxation
strength of free water is related to the concentration of polymer, Cd, as,
(1)
where Pp = 1.245 g/ml is the density of pure PVP obtained from an extrapolation of the
concentration dependence of the density of PVP aqueous solutions. It was also reported 1 that the
relaxation time, r, is described by free volume theory as
1.83
logr = -12.18---Cd -1.66

The complex dielectric permittivity, e*, at the angular frequency
Havriliak-Negami equation
(2)
0)
is generally described by the
(3)
where ~c is the relaxation strength and Coo is the high frequency limit of the dielectric constant.
The exponents a and {3 are shape parameters used to determine the broadness of relaxation
curves. The high frequency relaxation process is represented by Cole-Cole equation (a = 1). The
density dependency of the shape parameter {3 follows the linear relation
f3 = 1 - O.330Cd
(4)
Furthermore, Coo is expressed as

(5)
where cp is the dielectric constant of pure PVP. The permittivity is obtained from equations (1-5)
at any water content. From perturbation theory, Kraszewski and Nelson8 related cavity
parameters to the permittivity of the sample,
(6)
and
99
(10' - 1) =4" K
where
k
Q.(::J
(7)
=0.05(S21O - S21s)
(8)
and
1
K=---1+ A(8'-1)
(9)
where fo and Is, are resonant frequencies without and with sample, respectively, Vo and Vs are
volumes of cavity and sample, S210 and S21s are transmission coefficients without and with
sample, and A is the depolarization factor. The value of Vs was calculated, assuming
(10)
where M is the mass of the sample, Pw =0.997 g/ml is the density of pure water. Employing Eqs.
(1-5) and the present data, the depolarization factor A is determined from Eqs. (6) or (7).
The slope shown in Fig. 3 represents the parameter A for each sample with Eqs. (6) or (7),
respectively. The slope of each line seems to be similar except for sample b at higher water
6.----------------,
5
00000
~ 3
,...
000 00 0
+ extrapolation
00
o sample a
sampleb
......
~samplea
A sample b
a (eq.7)
oL---~--~~~==~==~
2
2,2
2,4
A~
t+,tf
A sample b (eq.7)
1,8
A
"
sample b (eq.6)
1,6
~it~
d9
o
()(:iPcPe'
o sample a (eq.6)
~sample
*+* ~o
+ ++:F#
2
2
2,6
-1
Fig. 3. Plot of 11K against (8 - 1) from Eq. (6)

and (7) for each sample. The slope
represents the A parameter.
+*!!!~~~
0,05
0,1
0,15
0,2
water content
Fig. 4. Comparison of complex permittivity

with extrapolated values at the resonant
frequency for each measurement.
100
content because of shape modification. Since A values calculated from respective points are
dispersed, A was determined by fitting the actual data of the dielectric constant with the former
data for water content around 15wt%. We calculated the dielectric constant and loss from the
present data with the same A as shown in Fig. 4.
Figure 4 shows that the dielectric constant is smaller than the extrapolated values from the former
data on PVP solutions. This result implies that the relaxation time increases with decreasing
water content more than expected from Eq. (2). The free volume parameters in Eq. (2) are
determined from former relaxation measurements with more than 15wt% of water. Thus, it is
reasonable to consider the validity of Eq. (2), as smaller values of the relaxation time are
calculated in the lower content region where the relaxation time increases much with water
content. On the other hand, the dielectric loss is larger than the extrapolated value especially in
the higher water content region. This result is supposed to be affected by dc conductivity of the
sample. Moreover, for sample b, the change in the A factor brings a large error.
Dynamical behavior of water molecules in the glassy state have not been clarified yet. It is
concluded that the dielectric resonance method is expected to be the most effective technique for
dielectric measurements of materials with low water content, even when materials can be
considerably fragile in the low water content.
REFERENCES
[1] Miura, N., Shinyashiki, N., and Mashimo, S. "Dielectric relaxation of the Kohlrausch-type in aqueous polymer
solution", J. Chern. Phys. 1992,97(11),8722-8726.
[2] Shinyashiki, N., Asaka, N., Mashimo, S., and Yagihara, S. "Dielectric study on dynamics of water in polymer
matrix using a frequency range 106_10 1'1Iz", J. Chern. Phys. 1990,93(1),760-764.
[3] Umehara, T., Kuwabara, S., Mashimo, S., and Yagihara, S. "Dielectric study on hydration of B-, A-, and ZDNA", Biopolymers 1990,30,649-656.
[4] Miura, N., Shinyashiki, ", Yagihara, S., and Mashimo, S. "Microwave dielectric study of water structure in the
hydration process of cement paste", J. Am. Ceram. Soc. 1998,81(1),213-216.
[5] Rouleau, J. F., Goyette, J., Bose, T. K. and Frechette, M. F. "Detection of small quantities of contaminants in SF6
by a microwave method", Proc. 1996 IEEE Int. Symp. on Elec. Insul., Montreal, June 16-19, 1996 pp. 111-113.
[6) Le Noe, 0., Goyette, J., Bose, T. K. and Ingrain, D., "Determination of the compressibility factor and
measurement of traces of water in methane by a microwave method", Fluid Phase Equilibria 1996,114, 135-145.
[7] Kent, G. and Bell, S. M. "The gap correction for the resonant-mode dielectrometer", IEEE Trans. Instrum. Meas.,
1996,45(1),98-101; ibid., "Nondestructive permittivity measurement of substrates", 102-106.
[8) Kraszewski, A. W. and Nelson S. O. "Microwave resonator for sensing moisture content and mass of single
wheat kernels", Can. Agr. Eng. 1994,36(4),231-238.
Contact point: Shin Yagihara, Institut de Recherche sur I'Hydrogene, Universite du Quebec a Trois-Rivieres, C.P.500
Trois-Rivieres, Quebec, G9A 5H7 CANADA Phone: +1 (819) 376-5108; Fax: +1 (819) 376-5164; E-mail:
yagihara@ugtr.uguebec.ca, yagihara@keyaki.cc.u-tokai.ac.jp
* S. Yagihara spends a sabbatical year at IRH, UQTR.

101
QUALITY FORECAST OF PARTICLEBOARDS

USING A MICROWAVE MONITORING SYSTEM
Ferenc Volgyi
Technical University of Budapest, Dept. of Microwave Telecomm.
H - 1111 Budapest, Goldmann ter 3, Hungary. t-volgyi@nov.mht.bme.hu
ABSTRA CT. A non-destructive testing method of particleboards is introduced, which is based

on a microwave free-space/double-transmissionlreflection type two-parameter complex vector
measurement. Using this basic idea, a 5.8 GHz monitoring system was developed, which is used
for quality forecast of particleboards continuously, before mechanical testing would be
accomplished. This paper summarises the phases of this development, giving some useful ideas
about the instrument itself, and industrial experiences, too.
Keywords: particleboard, quality control, microwave aquametry, modulated backscatterer,

microwave sensor
1. INTRODUCTION
The new international standards (e.g. ISO-9002) require the continuous control of manufacturing
environment and quality of products. Until recently the mechanical properties (Modulus of
Elasticity: MOE, Modulus of Rupture: MOR) and physical ones (Moisture Content: MC,
swelling) of particleboards (PB, see Figure 1) were assessed by destructive standardised
experiments on specimens taken from the boards in every hour. There is no information about the
quality of the boards between two successive measurements. The continuous production line, in
spite of the severe control of the constituents and the process, might produce substantial amount
of lower grade particleboards within that hour. These boards are downgraded or rechipped again,
causing a significant deficit to the company annually. The solution is the continuous quality
monitoring of the boards by a non-destructive method. We have developed a microwave
monitoring system (see Figures 2-6) which is used for quality forecast of particleboards
continuously, before mechanical testings would be accomplished periodically.
Figure 2: PB and our backscatterer array
Figure 1: Particleboards in rotating equalizer

102
Fig.3: Mounting of the microwave monitoring

system: transceiver (up) and backscetterer
array (down, between roller conveyor and
milling machine)
Fig.4: Microwave transceiver using microstrip

antennas and PBs in rotating equalizer
Fig.5: Opened Transceiver Unit: IF-Unit,

digital parts and power supplies (up),
microwave unit (down)
Fig.6: PB and encapsulated backscatterer array
Particleboards are manufactured by heat pressing (~ 200 DC, ~ 20 kp/cm2, ~ 2 min) wood
particles of various kind and size, mixed with webbing resin glue (e.g. urea-formaldehyde). The
compressing due to pressure is approximately 20%, the moisture content after pressing is ~ 4%,
the content of air remaining in the board is ~ 5%. The most frequently occurring 19 mm thick
board contains ~ 9% glue. The tensile- and bending strength (MOE) decrease approximately with
the square of the increasing moisture content. In opposite to this, an increase of the resin glue by
5% (referred to the nominal value) yields a strength increase by 10-15%. The (MOR) increases
proportionally to glue content. The kind of wood is of decisive impact on the strength. With equal
board volume weight - at first hearing surprisingly - the 1:3 rated strength order is: beech, acacia,
birch, pine, poplar. By increasing the surface density (as well as the volume density) the strength
can be significantly increased. According to our measurements, the microwave parameters are in
relationship first of all with these densities.
103
2. PHASES OF TIllS DEVELOPMENT

Using our basic idea ([ 1], patent pending), first of all we have created a basic measurement setup
shown in Figure 7 for the purpose of laboratory experiments. More details are given in [2].
Because the measured attenuation (M) and phase shift (A<I mainly depends on the moisture
content ofPB, this setup is excellent for microwave aquametry [3].
In the next phase an industrial version of our equipment was realized and tested. Some cases high
power RF-dryers are used in PB-production,. the application of spread spectrum technique [4] was
profitable. Measurements were accomplished to select the appropriate BPSK-code for the
modulated backscatterers [5].
Using a simple dielectric mixing model some calculations and diagrams were presented [6] about
the effect of glue-overdose and quality degrading air-inclusions formed during pressing of the
board. The developed microwave monitoring system is also applicable for plastic-sheets and RFabsorbers [7].
Preliminary measurement show, that the correlation between measured MOE and the result
calculated from measured microwave parameters is acceptable (see Figure 9). For standard wood
particle mixtures (mentioned at the Introduction) the glue-overdose and air-inclusion plots are
shown in Figure 10 and 11, respectively. Industrial experiences show that the effect of the
measurement-site having a lot of reflective metals (see Figure 3) is negligible, because we have
used circular polarization at microstrip antennas. The effect of sawdust in Figure 6 is petty, too.
The distance between PDBs and MUT gives the heaviest effect among errors. The relative
amplitude error is significant (see Figure 8). This error was minimized, as a result of our further
enhancements using three frequencies (fl=5.78 GHz, fO=5.80GHz, and f2=5.82 GHz). Figure 12
shows the measured relative amplitude and phase variation of a particleboard.
Measuring Digital
Microwave (5.8GHz)
Transceiver
'7/
AA,--"-~_--+
flf-1
II
'lllI2ll~'''''''''''''''''
-
....
rR'B\
\5'
J\.1aterial Under
II
I
Test (MlXf)
Passive Detector
I
I
Back:scatterers
(PDB)
Activated PDB using

Biphase Code
ControlPC
and Display
I
I
I
Figure 7. Scheme of the basic measurement setup
104
MOEl a'
.8t
21
19
-1255
II 11
Figure 8: Graphical error analysis
12
13
14 15 16 17 18
No. of sample
19
I
I
.Jt
1235
Figure 9: Measured MOE, e' and calc. MOEI vs. No.
0.89
3
Moisture content [m1m%]
Figure 10:
(w -1) I';;:; versus MC
0.8 2
Moisture content [m1m%]
Figure 11:
(w -1) I';;:; versus MC
4. CONCLUSIONS
Complex permittivity monitoring of particleboards gives us information about abnormal density or
glue-overdose or presence of air inclusions in the board at the moment of its occurrence. Using
dielectric mixing models the effect of change of moisture content, fluctuations in amount of other
ingredients were predicted. Industrial tests show that the developed measurement system is
suitable for on-line, real-time quality monitoring of large boards (3x17 m2). In this phase the
system is installed in industrial environment to collect more measurement data. In addition we
make experiments to compare the real mechanical properties to the measured microwave
parameters. Then a computer aided correlation analysis is carried out to find empirical formulas
between measured values of attenuation, phase shift and the measured mechanical properties.
After all an appropriate software can also display graphically the surface distribution of these
parameters on the whole board so its quality could be judged simply at a glance.
ACKNOWLEDGEMENT
This work was supported by: Interspan Ltd (Hungary), Zoltan Bay Foundation, Hungarian
Ministry of Culture and Education (MKM 795/1996). The author wishes to thank Quini Ltd and
Microval Co. for their excellent cooperation. He also thanks his PhD students G. Gyimes~ P.
Olasz, K. Kerekes and L. Nyul for their useful contribution.
I would also like to thank Dr. Andrzej w. Kraszews~ of the United States Department of
Agriculture, for his encouragement to write this paper.
105
Figure . Measured amplitude and phase variation of a composite board having dimensions of
(0.017) x 2.07 x 16.5 m3 . Contrast ranges (from white to black) 2.48 1 dB and 132.2 30 are,
respectively. Horizontally compressed pictures, frequency is 5.8 GHz, time duration is 56 sec.
Backscatterers are marked using letters B,C .. .I.
REFERENCES
Volgyi F., Mojzes I., Kelemen M. , Racz I., Seller R., Sonkoly A , Divos F. and Zombori B.: "Method and
Equipment for SpecifYing the Parameters of Substances, Applied Mainly for Moving-, Multilayer Materials"
(In Hungarian: "Eljaras es berendezes anyagrendszerek parametereinek meresere, elonyosen mozgo,
tobbretegu anyagokon torteno alkalmazasara"), Announcement for Patent, MSzH - P 9601630, June 1996.
[2] Volgyi F. and Zombori B.: "A New Application of WLAN Concept: Complex Pennittivity Monitoring of
Large-Sized Composite Boards", IEEE MTTS-S, WMFB-Workshop, 17th June 1996, San Francisco, USA,
pp.119-122.
[3] Kraszewski AW. (editor): "Microwave Aquametry", IEEE Press, New York, !996, ISBN 0-7803-1146-9.
[4] Volgyi F., Mojzes I., Seller Rand Olasz P.: "Pennittivity Monitoring of Composite Boards Using Spread
Spectrum Techniques", URSI'98, Int. Symp. on Electromagnetic Theory, Thessaloniki, Greece, 25-28 May
1998, pp.97-99.
[5] Volgyi F., Olasz P. and Mojzes I.: "New Application ofWLAN Concept: Modulated Backscatter and Spread
Spectrum Techniques", PIERS'98, Progress in Electromagnetic Research Symposium, Nantes, France, 13-17
July 1998, p .565.
[6] Nyul L. and Volgyi F.: "Contact less Testing of Particleboards", APMC '98, Yokohama, Japan, 8-11 December
1998, pp.213 -216.
[7] Volgyi F., Nyul L. , Tatar S. and Mrovcza A:"A New Radio Frequency Absorber for GTEM Cell Applications"
tl1
EMC Zurich'99, 13 lnt. Zurich Symp. on Electromagnetic Compatibility, 16-18 February 1999, pp.675-678.
[1]
106
OBSERVATION OF CHANGES IN DYNAMICS OF WATER MOLECULES

IN VARIOUS AQUEOUS SYSTEMS USING TIME DOMAIN
REFLECTOMETRY
Shin Yagihara, * Yoshihito Hayashi, Hidenori Miyairi, Megumi Asano, and Naoki Shinyashiki
Department of Physics, Tokai University, Hiratsuka, Kanagawa, 259-1292 JAPAN
Rio Kita and Toshiaki Dobashi
Department of Biological and Chemical Engineering, Gunma University,
Kiryu, Gunma, 376-8515 Japan
Masami Shiotsubo
Measuring Engineering Division, Komatsu Engineering Corporation
3-20-1, Nakase, Kawasaki-ku, Kawasaki-shi, Kanagawa, 210-0818 JAPAN
ABSTRACT
Microwave dielectric measurements were performed in the frequency range from IMHz to 30GHz
using a time domain reflectometry(TDR) method for several aqueous systems such as emulsion,
double- and triple-helix DNA, and mortar. Relaxation processes due to free water, bound water, and
strongly bound water were observed. It is concluded that the TDR technique is available to
investigate confined structures of water in complex molecular systems.
Keywords: emulsion, BNA;-cement;', time domain rejlectometry ,
1.
INTRODUCTION
The dielectric relaxation studies using time domain reflectometry (TDR) in the frequency range of
IM-30GHz have been extensively performed for various materials including aqueous systems such
as water mixtures with associated liquids, aqueous solutions of polymers and biological molecules,
emulsions, cells, and biological tissues. Dielectric properties of water thus observed indicate a
variety of dynamical behaviors of the water molecule.
Dielectric relaxation processes arising from the behavior of water molecules are classified in three
categories. The highest frequency process reflecting the depolarization comes from
makinglbreaking processes of molecular clusters of water. This water molecule is called free water.
Free water has a relaxation time of 8ps in pure water at 25C. The relaxation time of free water
increases, if solute molecules reduce the free volume of free water. 1,2 The relaxation strength of free
water is usually treated to be proportional to the water content. However some examples indicate
that experimental values are less than expected values from the water content. 3 The relaxation
strength of free water has not been systematically discussed yet.
107
The middle frequency process generally observed around 100MHz is caused by an orientation of
water molecules in the molecular network formed by hydrogen bond with other water molecules on
the surface of hydrophilic molecules such as DNA46 and proteins. 79 The relaxation time of bound
water doesn't change more than one decade. The relaxation strength reflects the number of bound
water molecules. Then this number is closely related to the molecular structure of the solute
molecule.
It has been considered that bound water stabilizes the structure of double helix DNA and the helix
parameters are corresponding to the number of bound water. Recently triple to double helix
transformation of DNA has been of great interest as one of tools for mapping the gene and for
pharmaceutical use. 10,11 The third strand enters into the major groove of double helical chain, then
the structure of the helix including the pitch of the helix and the highly associated bound water along
the helical chain is modified as seen from NMR. 12 Then, it is quite interesting to inquire the change
in the number of the hydrated water on the surface of the DNA complex through the conformation
change from triple to double helix.
The lower frequency process recently found for water included in some inorganic materials suggests
an existence of strongly bound water. This water is closely related to the molecular structure and
physical properties of the aqueous system. The relaxation time of strongly bound water for cement
paste depends on curing time. 13 However details of the hydration process of cement and the role of
water molecules have not been understood yet.
In this work, we have performed TDR measurements on aqueous systems of emulsion
(oil/surfactant/water) for free water, triple-helix DNAibuffer for bound water, and mortar
(cement/sand/water) for bound water taken into the structure of calcium silicate hydrate. The
relaxation process for each water structure is discussed from the viewpoint of confined structures of
water.
2.
EXPERIMENTAL
Liposome of dipalmitoyl phosphatidilcholine (DPPC) (Sygena Ltd.)and O/W emulsion were

prepared by extruder(Lipex Biomembranes, Inc.). The oil and the surfactant employed in this work
are dl-a-tocopherol acetate(VE) (Nisshin Oil Mills, Ltd.), soy lecithin(SLE)[phospholipid content:
95%(PC70%, PE12-14%)](Nisshin Oil Mills.,Ltd.), decaglyceryl monomyristate (DMM) (Nikko
Chemicals Co., Ltd.), hexaglyceryl monomyristate (HMM) (Nikko Chemicals Co., Ltd.), and
pentaglyceryl monomyristate (PMM) (Taiyo Kagaku Co., Ltd.). These chemicals were used without
further purification. Distilled and three times deionized water was passed through the purifier
(Millipore Ltd.), and it was used to prepare the samples for the dielectric measurements.
For DNA samples, oligothymine (PT) and oligoadenine (PA) with the number of residue being 20
were purchased from Toa Chemical Synthesis. These oligonucreotides were mixed with the molar
ratio ofPA:PT=1:2 (System A) and PA:PT=l:l (System B) in a buffer solution at pH=7 and
108
[Na+]=0.3M. The concentration of the mixed oligonucreotides was 0.5wt%. The ultraviolet light
absorption study showed that the transformation temperature of this mixture from triple to double
helix was 51C. Thus in this solution the triple helix structure is the most stable at room
temperature: System B contains equal number of moles, N, of triple helix and single oligoadenine
and System A contains 4/3 N moles of triple helix only.
The portland cement was obtained from Chichibu-Onoda
Cement Co. The mortal was prepared as a mixture with the
ratio of 1: 2: 0.65 for cement, sand, and distilled water.
Sand prepared for botanical use was purchased.
Sample
Digitizing Oscilloscope
Mainframe
(HP54120B)
Coaxal
Cable
TDR measuring techniques have already been described in

details. I-3,s-1I Figure I shows the block diagram of the TDR
system. Flat-end sample cells with the outer conductor
diameters 5.0mm(electric length, yd=0.32) and 2.0mm
(yd=O .13) and coaxial cells with electric lengths 0.34.3mm were employed.
Ch.1
Digitizing
Oscilloscope
Four Channel Test Set
(HP54124A)
LCR meter (HP4284A) and a three-terminal parallel plate

cell with 2.87pF were employed for the measurements of
the cement paste in the frequency range of20Hz-IMHz.
inSid\DR
Step
Generator
Sampler
Figure 1. TDR system
Relaxation measurements were performed at 25C

except in the case with the description.
3.
RESULTS AND DISCUSSIONS
Figure 2 shows examples of dielectric relaxation

phenomena observed in the present work. A
relaxation process was observed around 20GHz for
all aqueous suspensions of liposome and emulsions.
The relaxation time for each sample is same with
pure water (8ps). This result clearly indicates
dielectric properties of free water.
30
20
w
10
DPPC Liposome(84%)
Emulsion (85%)
0 DMM Emulsion (75%)
\l HMM Emulsion (75%)
0 PMM Emulsion (75%)
C. SLE
Figure 3 shows water content dependence of the

9
relaxation strength, LlE. If LlE is proportional to the
log f(Hz)
number of water molecules per unit volume, the
Figure 2. Dielectric relaxation
plots are on the solid line. However, the plots appear
curves for liposome and various
below the solid line. The difference of LlE between the
emulsions.
DPPC liposome and the SLE emulsion cannot be
recognized, even if the former includes water inside and the latter does not.
109
10
80
The ~s values observed for emulsions ofDMM, HMM,

and PMM are different. This result suggests that the ~s
are detennined not only by the water concentration but
also by the size of the hydrophilic part of the surfactant
molecule. The surfactant with smaller hydrophilic part
tends to reduce ~s as shown in. the case of protein
solutions. 3
+ DPPC Liposome
w
<J
From the relaxation strength of the dielectric relaxation

curves at 100MHz it is indicated that there exists ca. 1.5
times of hydrated water in System A compared with that
in System B. If only structured water can contribute to
the peak at lOOMHz, this ratio is reasonable since the
peak should come from triple helix DNA only. This
assumption is consistent with the results for the helix-coil
transition of DNA. 7
We have not observed bound water in aqueous systems
of synthesized polymers, even if it is hydrophilic,
because of lack of the stable network structure of water
molecules surrounding the polymer molecule. However
chain motion is usually observed in the same frequency
region with bound water for these systems.
Emulsion
Emulsion
Emulsion
Emulsion
c,+
c,~
0
O~B
40
IS3
20
0~~~~~~~-7~~~~~
0.2
0.4
0.6
water content
0.8
(glcm 3)
Figure 3. The relaxation strength vs.

water content for various suspensions.
System A, 30'C
10 1
0 System B, 30'C
10
log f(Hz)
Figure 4. Dielectric relaxation curves

for DNA.
15
Recently we reported that the cement paste shows some

dielectric relaxation processes. \3 The dielectric relaxation
process due to the free water decreases with curing time,
since the free water is taken into calcium silicate hydrate.
A relaxation process due to the hydrated water occurs at the
lower frequency range, when hydration proceeds.
110
60
Figure 4 shows dielectric relaxation curves for aqueous

solutions System A and B. These relaxation curves
suggest some difference in the hydration processes of
the both system.
Figure 5 shows relaxation curves for mortar with various

curing time. A sample with curing time of33 days shows
a relaxation process around IMHz. This process is
supposed to reflect strongly bound water in the structure of
calcium silicate hydrate. Furthennore this process shift to
200Hz at curing time of 4 years. The relaxation time
SLE
DMM
\1 HMM
o PMM
C,
=",
3
2 4 years
~-~ ... 33 days
$'li"?
~~
,~.
5
10
log f(Hz)
Figure 5. Dielectric relaxation for

mortar with the various curing time.
increasing with the curing time and strength suggest curing and aging effect.
We reported that dynamics of water in aqueous systems drastically change, corresponding to the
molecular structure and property of the material. The TDR method is the moist available technique
to offer these indications in the wide frequency range.
REFERENCES
[1] Shinyashiki, N., Yagihara, S., Arita, I., and Mashimo, S. "Dynamics of water in a polymer matrix studied by a
microwave dielectric measurement", J. Chern. Phys., 1998,93(1),760-764.
[2] Treated in following paper of this workshop. Rouleau, J. F., Goyette, J., Bose, T. K., Yagihara, S., and Shinyashik, N.
"Dielectric study on dynamical structure of water in moist materials using dielectric resonator", The third workshop on
electromagnetic wave interaction with water and moist substances, Athens, Georgia, April, 1999.
[3] Naito, S., Hoshi, M., and Mashimo, S. "In vivo dielectric analysis of free water content ofbiomaterials by time
domain reflectometry", Anal. Biochem., 1997,251,163-172.
[4] Umehara, T., Kuwabara, S., Mashimo, S., and Yagihara, S. "Dielectric study on hydration ofB-, A-, and Z-DNA",
Biopolymers 1990,30,649-656.
[5] Mashimo, S., Umehara, T., Kuwabara, S., and Yagihara, S. "Dielectric study on Dynamics and structure of water
bound to DNA using a frequency range 107-10 IOHz", J. Phys. Chern., 1989,93(12),4963-4967.
[6] Kuwabara, Umehara, T., S., Mashimo, S., and Yagihara, S. "Dynamics and structure of water bound to DNA", 1.
Phys. Chern., 1988,92(17),4389-4391.
[7] Shinyashiki, N., Asaka, N., Mashimo, S., Yagihara, S., and Sasaki, N. "Dielectric study on hydration of moist
collagen", Biopolymers 1990,29, 1185-1191.
[8] Miura, N., Asaka, N., Shinyashiki, N., and Mashimo, S. "Microwave dielectric study on bound water of globule
proteins in aqueous solution", Biopolymers 1994,34,357-364.
[9] Miura, N., Hayashi, Y, and Mashimo, S. "Hinge-bending deformation of enzyme observed by microwave dielectric
measurement", Biopolymers 1996,39, 183-187.
[10] Felsenfeld, G., Davis, D, and Rich, A. "Formation of a three-stranded polynucleotide molecule", J.Am. Chern. Soc.
1957,79,2023-2024.
[11] Soyer, V.N. and Potaman, V.N. Triple-Helical Nucreic Acids, Springer-Verlag, NY, 1995.
[12] Radhakrishnan, L and Patel, J. D. "DNA triplexes: solution structures, hydration sites, energetics, interactions, and
function", Biochemistry, 1994,33,11405-11416.
[13] Miura, N., Shinyashiki, N., Yagihara, S., and Shiotsubo, M. "Microwave dielectric study of water structure in the
hydration process of cement paste", J. Am. Ceram. Soc. 1998,81(1),213-216.
Contact point: Shin Yagihara, Department of Physics, Tokai University, Hiratsuka, Kanagawa, 259-1292 JAPAN
Phone: +81 (463) 58-1211; Fax: +81 (463) 58-5023; E-mail: yagihara@keyaki.cc.u-tokaLacjp
111
COMPUTER AIDED DESIGN FOR SURFACE

ACOUSTIC WAVE LIQUID SENSORS
*MS.Zaghloul, **T.E. Taha, *K.EL-Shennawy, **Aly.H.Moustfa.
*Arab Academy for Science and Technology.
** Faculty of Electronic Eng. Minufiya University,

E-mail: ruasta@rusys.EG.net
ABSTRACT. The change of the physical properties of the Surface Acoustic Wave
(SA W) device surface or near surface region can be used as a physical sensor because
this device is highly sensitive of the change of the physical properties at its surface. In
SA W liquid sensors the interaction of the propagating waves on the piezoelectric
substrate with the mobile carrier of the adjacent loading liquid is used to measure the
parameters of the loading liquid properties. In this paper, a computer program is
developed, based on the perturbation theory for describing the performance of SA W
liquid sensors to arbitrary liquids, which can simulate the effect of various liquid
parameters on sensor response. The velocity and attenuation shifts due to perturbation
of piezoelectric materials are plotted against the liquid conductivity. The
electromechanical coupling coefficient incorporates the substrate permitivity 8s, and the
liquid permitivity~. To neglect the effect of the mechanical perturbation, a small air gap
between the substrate and the loading liquid is considered. The current research shows
that there is a good agreement between the theoretical and the experimental published
results. Moreover, the obtained results can be used as a useful tool for the development
of SA W liquid sensors and indicate the possibility of using these programs as
characterization technique for liquid properties.
I. INTRODUCTION
Recently, SA W sensors are considered as powerful application of SA W devices. Sensing
takes place by considering the mechanical and electrical perturbations of the surface
which affect the propagating acoustic waves. Many studies have appeared which
investigate various types of acoustic waves to improve sensitivity, resolution, and time
response. The loading of a piezoelectric surface with viscous conductive liquids has
been studied for acoustic wave liquid-phase-based sensor applications [1-4]. The
mechanisms of interaction between the liquid and the acoustic wave include both
mechanical and electrical loading. Experimental studies were the key to show which
type of acoustic wave is best for these applications. The SA W devices structure of high
interest because they have high sensitivity to the change in the physical properties .To
design such sensors a knowledge of the perturbation and the finite element techniques
are essential. In the present work, the considered technique is the perturbation
technique. The elastic surface wave, which propagates along the free surface of
piezoelectric crystal, will be affected by the electromagnetic properties of the medium
adjoined to the surface. This surface acoustic wave propagating on the piezoelectric
substrate can interact with the mobile carriers of the adjacent loading liquid, and hence
there will be an acoustoelectric effect that can be used for measuring the parameters and
the properties of the loading liquid. As a result, there will be a change in the physical
properties which can be the conductivity, permitivity, temperature, pressure, mass,
112
compliance, concentration, viscosity... etc. This will ~reate a perturbed wave velocity
and attenuation. The velocity and the attenuation shifts can be used as a measurant.
In section 2, the interaction analysis between liquid and SAW is presented. In section 3,
the analysis of different liquids is considered. Section 4 contains a comparison between
the theoretical obtained results and the experimental published results.
2. INTERACTION BETWEEN LIQUID AND SAW
The case under study is of a layered structure of an acoustic wave liquid phase sensor.
Thin-film layered SAW structure have been in production for many years [13]. For
some sensor devices a thin layer covers the surface or part of the surface. While in
others the surface is loaded with a dilute conductive liquid, in this case a cell can be
designed to confine the liquid to the propagation area, and to have a constant volume of
it on the surface of the substrate. In the shown Fig (I), (not drawn to scale) the loading
liquid of thickness d, air gap of thickness h separates the loading liquid and the
piezoelectric crystal, which supports the acoustic wave.
t - - Liquid
load
t-Airgap
I
~ Piezoelectric
material
--------------_.
Fig.(I): Layered structure of an acoustic wave liquid phase sensor
The selected type of the acoustic wave affects the designer choice to build the geometry
of the substrate of the sensing element. In the present case, the air gap height h must be
very small relative to the wavelength in order that an interaction between the SA Wand
the loading liquid takes place. Normally, to study the effect of one parameter, others are
tried to be avoided, that is why references are used [5,6]. In order to derive the
perturbation for both the velocity v and the attenuation cr, initially let us starting by
defining the unperturbed velocity uo as:
Vo
ro Ik
Where ro = angular velocity

k = the wave number.
As stated before, the velocity is a function of the conductivity "(j ", the permitivity "8",
temperature "T', pressure ''P'', mass ''m': compliance "c ': concentration "n': viscosity
"rf'...etc, so we can write U = th. (CF,S ,T, P, m, c, n, 1J).
Ingebristen found that when the permitivities of the piezoelectric and the adjoining
medium are large.1a small air gap h might drastically reduce the coupling.
113
The reduction of the coupling is small only if

(j)
Vs
81
tanh[( -)h] _0
Where
80
= air gap permitivty.
liquid permitivty.
Which means that when 81 is large the spacing h must be very small compared to the
wavelength.
81 =
Due to the previously stated assumption that the loading liquid is separated from the
piezoelectric surface by a small air gap with permitivity 8 0 , The capacitance per unit
length Cs is the sum of the permitivities of the substrate, air gap, and the liquid
Cs =8s +80
+81
As a result of applying the perturbation theory, two equations representing the velocity
shift and the increase in attenuation of acoustic wave were derived, these are:
A computer program was developed (based on the perturbation technique) and applied
to different cases of piezoelectric materials with characterizing parameters taken from
[3]. These characterizing parameters include both the coupling coefficient k2 and the
substrate permitivity Eg Also this program was used assuming the case when the liquid
is directly on the surface, i.e. was no air gap. Moreover, the program was applied for
different values for
Fig. (2) to Fig. (9).
representing different loading liquids. The results are shown in
3. SAW PERTURBATION VELOCITY AND ATTENUATION WITH

CONDUCTIVTY:
Although most of the reported sensors assume Raleigh wave, it is assumed here that the
used mode is the acoustic plate. The program was applied for LiNb03 substrate of
dielectric constant 8s =47.9Go. K2=O.027, velocity of 3488 mis, loading liquid of the
dielectric constant 79.2 80, and frequency of 13.6 :MHz. When there is no air gap
separation and taking irito consideration the effective length of transducer to be 32
times the wavelength. Some of the output charts of this program for both fractional
velocity change and attenuation coefficient are plotted versus (vs.) the liquid
conductivity in Fig. (2) and (3).
114
&;*10-6
v 4000
Q.l
0.2
0.4
0.3
0.5
0.1
crL (n-I.m-I)
0.2
0.3
0.4
0.5
crL(n-l.m-l )
Fzg (2): Fractional velocity change vs. conductivity

for 13.6MHz on LiNb0:3 with ~= 0.27and velocity
3488 mlsec when no air gap is present between the
Fig (3): Attenuation coefficient vs. Conductivity for

13.6 MHz onLiNb0:3 with~= 0.27 and velocity 3488
mlsec when no air gap is present between the loading
liquid of dielectric constant 79.2 So and LiNb0:3
substrate of dielectric constant 47.9 So
loading liquid of dielectric constant 79.280 and the

LiNb0:3 substrate of dielectric constant 47.98 0
Measured Curve
0.1
0.2
0.3
o~~--~--~--~--~
0.5
0.1
crL (n-I.m-l )
Fzg (4): Fractional velocity change vs. conductivity for
a 10-3
_*
0.4
0.5
15
----------
. ,'
0.3
Fig (5): Attenuation coefficient vs. Conductivity for

13.6 MHz on LiNb0.3 with ~ = 0.27 and velocity 3488
mlsec when air gap is present between the loading
liquid of dielectric constant 79.2 So and LiNb0:3
substrate of dielectric constant 47.9 So
13.6MHz on LiNb0:3 with ~ =0.27andvelocity 3488

mlsec when air gap is present between the loading liquid
of dielectric constant 79.2 So and the LiNb0:3 substrate
of dielectric constant 47.9 So
&J*I0-6
v
0.2
crL (n-l.m-I)
SL =
...............................
20.50
40.50
=60.50
SL= 80.50
SL = 100.50
SL =
SL
xxxxx
OL-~--~--~--~--~
0.1
0.2
0.3
0.4
0.5
crL (n-I.m-I)
Q.l
0.2
0.3
0.4
0.5
O'L (O-I .m-l)
Fig (6): Fraction velocity change vs. conductivity for

various values of liquid dielectric constant for the 13.6
MHz on LiNb0.3 with ~.27 and velocity 3488 mls
115
Fig (7): Attenuation Coefficient vs. conductivity for

various values of liquid dielectric constant for the 13.6
MHz on LiNb0:3 with ~0.27 and velocity 3488 mls
!w*I~
~r-------~----~
!:.*10-3
------------
eL = 20.80
rf' / ....... ~-=::.-~-
eL = 40.eo
eL = 60.80
eL = 80.80
I:~
-
XXXXX
eT.
= 100.1>0
O"L (O-l.m-l)
Fig (8): Fmction velocity change vs. conductivity for

various values of liquid dielectric constant for the
13.6 MHz on LiNb0:3 with kl=0.05 and velocity 3490
mls
Fig (9): Attenuation Coefficient vs. conductivity

for various values of liquid dielectric constant for
the 13.6 MHz on L~ with 1C=0.05 and
velocity 3490 mls
4. THEORETICAL AND EXPERIMENTAL RESULTS
The theoretical obtained results are compared with the experimental published result
[6], in Fig. (3) and there is a good agreement, this indicates that this program can be
used as good characterization technique for liquid construction.
5. CONCLUSION
The SAW sensor can be used as a physical sensor due to its high sensitivity. The
velocity and attenuation shift due to perturbation of different piezoelectric and liquid
materials via conductivity show that the simulated theoretical results are in good
agreement with the experimental published results, so it is best suitable for chemical
and biosensors applications.
REFRENCES
Martin. G. C. Frye, and A J. Ricco, Acoustoelectric Intemction of Plate

Modes with Solutions, J. Appl. Phys., vol.64, No.210.
2. J. Andle, J. Vetelino, R Lee, and D. Mcallister, An Acoustic Plate Mode Immunosensor. In Proc.
1989 IEEE Ultms. Symp.
3. F. Josse, D. T. Haworth, U. R Kelkar, and Z. A Shena, Li.Nb03 Acoustic Plate Mode Sensor For
Dilute Ionic Solutions, Electronic Letters, vol. 26, No. 13.
4. T. Morizzumi, Y. Unno, and S. Shiokawa, New Sensor in Liquid Using Leaky SAW, In Proc. 1987
IEEE Ultms. Symp.
5. A J. Ricco, S. J. Martin and T. E. Zipprian, Surface Acoustic Wave Gas Sensor Based on Film
Conductivity Changes, Sensor and Actuators International Journal Devoted Research and
Development of Solid State Tmnsducers, January 17,1986.
6. F. Josse, Z. A Shana, Electrical Surface Perturbation of A Piezoelectric Acoustic Plate Mode by A
Conductive Liquid Loading, IEEE Tmnsaction Ultmsonics, Ferroelectrics, and Frequency Control,
vol. 39, No.4, July 1992.
.
7. M. Feldmann and J. Henaif, Surface Acoustic Waves For Signal Processing English Tmnslation,
Artech House, 1989.
8.F. Joss, D. T. Hawrth, U.R KelkarandZ. A Shanna, Electronic Letters, vol. 26.1990.
9. K. A Ingebrigsten, Linear and Nonlinear Attenuation of Acoustic Surface Waves in A Piezoelectric
Coated with A Semiconducting Film, J. Appl. Phys., 44, 1970,454-459.
10. R Adler, Simple Theory of Acoustic Amplification, IEEE Trans. Sonics Ultrasonics, SU-18 .1971
115-118.
11. G. S. Kino and T. M Reeder, A Normal Mode Theory For the Reyleigh Wave Amplifier, IEEE
Tmns. Electron Devices, ED-18 ,1971, 909-920.
12. S. Datta, Surface Acoustic Wave Devices, Prentice-Hall, Englewood Cliffs, Nj, 1985.
13. C. Cambel, Smface Acoustic Wave Devices For Mobile and Wrreless Communications, Academic
Press Limited, CA 92101-4495,USA, 1998.
1. T. M. Niemczyk, S. J.
116
Session on
MOISTURE MEASURING SENSORS
Chair: Dr. Ray King,

KDC Technologies, Livermore, CA, U.S.A.

2:30 - 5:40 PM
117
118
DIELECTRIC GRAIN MOISTURE MEASUREMENTPRACTICAL, TECHNICAL, AND REGULATORY ISSUES

David B. Funk
United States Department of Agriculture

Grain Inspection, Packers and Stockyards Administration
Technical Services Division, 10383 North Executive Hills Boulevard,
Kansas City, MO 64153, USA dfunk@tsd.fgiskc.usda.gov
ABSTRACT. Dielectric-type moisture meters are the dominant method for determining the
moisture content of U.S. grain from production to processing. Moisture content is assessed
virtually every time grain is handled or traded because of its impact on grain value and stability.
The dielectric method, though conceptually simple, requires careful attention to many parameters
for successful application. Achieving consistency among production units of a model is
particularly important. The National Conference on Weights and Measures, in cooperation with
GIPSA, has implemented a National Type Evaluation Program to improve confidence in the
accuracy and consistency of grain moisture meters and to facilitate the commercialization of new
moisture measurement instrumentation.
Keywords: grain, calibration, standardization, regulation
1. INTRODUCTION
"Build a better mousetrap and the world will beat a path to your door." Life "should" work this
way.
Most scientists and engineers, however, have experienced the frustration and
disappointment of slow acceptance of a great technical idea or an elegant design. Scientists and
engineers who want their work to be accepted must understand user needs and expectations.
They must conduct research and design products that relate to those needs and expectations as
closely as possible.
Dielectric-type grain moisture meters share many technical characteristics with instruments used
to measure moisture in other materials, but only instruments designed specifically to meet the
needs of the grain industry have had significant impact on that market. This paper discusses many
of the factors that contribute to the commercial success and profitability of grain moisture meters.
These observations are based on the author's experiences with numerous grain moisture meter
models-some very successful, some not. They are shared in the hope they will contribute
pO$itively to future generations of grain moisture meters.
2. PRACTICAL CONSIDERATIONS FOR MARKET ACCEPTANCE
The overall markets for grain moisture meters are quite large because of the importance of
moisture content in producing, handling, and processing grain. In the United States alone, there
are about 12,000 facilities that buy and sell grain-all of which own at least one grain moisture
meter. Each time grain is sold, moisture content is measured, and the result is used, in part, to
119
determine the price. Millions of producers use moisture meters to determine whether grain is at
suitable moisture levels to harvest, store, and market.
Grain processing may involve moisture measurement at several points. The dry matter content in
the incoming grain determines the amount of output material. Optimum processing efficiency
often requires that grain be dried or wetted to a certain moisture level. Continuous-flow (on-line)
moisture measurement is highly desirable for feedback or feedforward moisture control in
processing.
Precision agriculture presents another opportunity for on-line moisture sensmg. Crop yield
mapping is an essential step in optimizing agronomic practices. Harvesting equipment that
collects instantaneous data on the mass of grain also must have reliable moisture sensing to
determine the dry matter yield.
2.1. Producer Expectations
Dielectric grain moisture meter models that are targeted for producer use must be inexpensive,
portable, easy to use, and rugged. These moisture meters are carried around in tractors, trucks,
and combines to be available when immediate moisture tests are needed. Producers base
harvesting and grain drying decisions on the moisture results. Typically, the producer adjusts his
moisture meter to make his meter's results agree with results on the moisture meter at the
elevator where his grain will be sold. The absolute accuracy expectations are not as great for
these relatively inexpensive models as for instruments that are used in commerce. Also, since
these instruments are not used in commerce, they are not subject to regulation. Producer-type
grain moisture meters generally provide direct readout with built-in density and temperature
corrections for one to several different grain types. The less expensive models may omit density
corrections.
For precision agriculture applications, producers expect continuous-flow moisture sensors that are
rugged and smoothly integrated in the harvesting equipment. Available continuous-flow moisture
sensors for precision agriculture generally do not include density corrections. They are not, of
course, subject to regulation. As precision agriculture becomes more "precise," users will expect
better accuracy from their on-board moisture sensors.
2.2. Processor Expectations
The push for ever greater efficiency in grain drying and processing is increasing the demand for
continuous-flow moisture systems. A review of recent patents related to grain moisture
measurement found a majority dealing with on-line systems.[I] On-line moisture measurement
systems must be extremely reliable because failures could result in large quantities of out-of-spec
product and plant down-time. Moisture accuracy must be consistent with what is needed for
efficient production processes, but on-line systems may be adjusted to be consistent with the
processor's in-house reference moisture methods. Density and temperature corrections are
particularly important for these on-line systems to meet expectations. On-line moisture systems
require intensive engineering support for system design, fabrication, installation, and maintenance
120
because most installed systems are unique. This drives up cost and restricts their widespread
implementation. In-house systems that are not used to determine prices are not subject to
government regulations.
2.3. Grain Handler Expectations
Demands are much more stringent for grain moisture meters that are used as the basis for
commerce. In addition to user expectations, regulatory requirements limit market access for these
types, as will be discussed later in this paper.
Reliability of moisture results and ease of use are of paramount importance to grain handlers.
Grain is received very rapidly during harvest (one truck every one to two minutes), and the
moisture measurement system must not slow the process. Grain moisture meter operators in
elevators may not be highly trained. Whoever is available does the moisture test. Moisture meter
manufacturers have responded to the need for fast, easy to use moisture meters. Highly
automated instruments accept an un-weighed sample and, in response to the press of a button,
perform a complete moisture test in less than a minute. Moisture results are available in electronic
form for printing or external storage.
Grain handlers operate on thin profit margins. Differences in moisture measurements can erode
those profit margins if the moisture measurement used in determining the purchase price is lower
than the moisture measurement subsequently used to determine the selling price. Conversely,
systematic moisture inconsistencies in the elevator's favor put the elevator at a competitive
disadvantage if their producer-customers are unwilling to sell grain to them based on the moisture
tests.
Grain handlers may receive many different types of grain over wide ranges of moisture and sample
temperature. They expect moisture meters to provide reliable moisture results for whatever grain
arrives in whatever condition it arrives. They don't hesitate to complain to the moisture meter
manufacturer if those expectations are unmet.
Continuous-flow moisture measurement would seem to be desirable for use by grain handlers,
since entire loads could be averaged. However, this has not been common (in the United States,
at least). Grain price is based on the quality of a statistically representative sample of the load
using bench-top instruments rather than continuous-flow or stopped-flow moisture sensors. The
requirement for a representative sample for grain quality tests other than moisture is one reason
that moisture tests are done on a representative sample. Another reason is that most continuousflow moisture sensors measure only a part of the grain stream-and it is very difficult to ensure
that the part tested in a flow-through system is truly representative.
3. MEETING REGULATORY EXPECTATIONS
Since moisture content is an important factor in the pricing of grain, many regional, national and
international organizations charged with ensuring fairness in commerce take an active role in
regulating grain moisture measurement. The International Organization of Legal Metrology has
121
adopted an International Recommendation for grain moisture meters. [2] In the United States, the
individual States have regulatory responsibility for grain moisture meters used in commerce.
Uniform regulations are developed and promulgated by the National Conference on Weights and
Measures (NCW1\1), a voluntary organization made up of State Weights and Measures officials
and industry representatives. The National Institute for Standards and Technology provides
administrative support and oversight for the NCWM.
3.1. Moisture Meter Specifications and Tolerances
The NCWM conducts a National Type Evaluation Program (NTEP) that deals specifically with
evaluating the suitability of instruments for use in commerce. A committee made up of state and
federal officials, grain industry personnel, and moisture meter manufacturers' representatives have
developed specifications, tolerances, and evaluation procedures for grain moisture meters. [3] The
NCWM grain moisture meter specifications and tolerances are based on and similar to the OIML
International Recommendation. Familiarity with these documents is essential for instrument
designers preparing grain moisture meters for the U.S. market. In those states that enforce the
NTEP requirements for grain moisture meters, only those instruments that are certified to comply
with the requirements can be marketed.
The specifications in the NCWM document are primarily design requirements that reduce the
possibility of accidental or intentional incorrect moisture measurement results. The design
requirements are extensive, dealing with issues as simple as the minimum size of digital displays
and as complex as the details of audit trails to track all calibration changes. Instruments that fail
to meet these design criteria are rejected summarily without performance testing.
The tolerances deal with accuracy, repeatability, reproducibility, and stability over the ranges of
moisture and environmental conditions anticipated in commercial service. A GIPSA laboratory in
Kansas City, Missouri conducts the actual instrument evaluations according to the procedures
adopted by the NCWM.[4] To date, seven grain moisture meter models have been evaluated and
certified to meet the NCWM-NTEP specifications and tolerances. Of these, two are near-infrared
transmission types. The other five are dielectric-type instruments operating in the low megahertz
region. No instruments based on microwave measurements have been submitted for evaluation.
3.2. Calibration Evaluation and Maintenance Program
Grain moisture meters that demonstrate compliance with NTEP specifications and tolerances are
automatically enrolled in an ongoing calibration evaluation and development program. The
GIPSA laboratory in Kansas City, Missouri annually collects over a thousand grain samples that
represent all growing regions and moisture levels for major U.S. grains. These samples are
analyzed by the GIPSA's standard oven method to determine the "true" moisture content. [5] The
same samples are tested using two of each of the NTEP-certified moisture meter models. These
data are used to annually evaluate instrument performance on a much more extensive sample
population than that used for the initial model evaluation. The data are provided to the moisture
meter manufacturers for use in improving calibration equations. This ongoing calibration
development program greatly benefits moisture meter manufacturers because it provides them
122
with the data needed to develop and maintain commercially acceptable calibrations for the most
significant U.S. grains.
3.3. Field Regulatory Enforcement
Several, but not all, State Weights and Measures organizations conduct field-testing to ensure that
each individual moisture meter is operating within NCWM specifications. Field inspectors carry
grain samples of known moisture contents to the elevators within their jurisdictions to check
moisture meter accuracy. Instruments that do not give accurate results must be repaired or
replaced. Users are not allowed to adjust instruments to meet tolerances on the inspector's test
samples.
4. TECHNICAL CONSIDERATIONS FOR MEETING EXPECTATIONS
Researchers and instrument designers are naturally sensitive to many of the factors required for a
successful grain moisture measurement system. Precision and sensitivity in the measured
parameters, temperature stability in the circuitry, high correlations between the predicted and
"actual" moisture content, and low direct production costs are sure to be considered. This section
discusses several other factors that largely determine the indirect support costs for grain moisture
meters. For grain moisture meters intended for use in commerce, attention to these factors can
mean the difference between success and failure.
4.1. Calibration Equations
Users expect grain moisture meters to provide reliable measurements for dozens of grain types
over changing crop conditions, grain varieties, cultivation practices, and drying and storage
practices. The cost of developing and maintaining calibrations may exceed the instrument
development cost by many times over the lifetime of the instrument. Moisture meter
manufacturers generally bear the cost of developing and maintaining the demanded calibrations.
They recover the cost indirectly through revenues from ongoing instrument sales. It is extremely
important that all aspects of the instrument's design contribute to effective and efficient
calibration development.
4.1.1. Instrument Data Outputs
Calibration equations will be evaluated and re-optimized repeatedly over the lifetime of the
instrument. Grain samples are inherently unstable and unsuitable for repeated testing over a
period of years. Therefore, the instrument must output all of the important measured parameters
for each test sample, not just the moisture content, in a form that is suitable for long-term external
storage. The measured parameters that are accessible in the output data stream must be adequate
to precisely re-predict moisture values using new calibrations. Minimizing the data available in
the output data stream may drastically limit capability to test and improve calibrations later.
123
4.1.2. Calibration Coefficient Resolution

Compact calibration coefficients are advantageous if the operator is expected to hand enter them,
but inadequate resolution in the coefficients can cause serious headaches. A change of one in the
least significant digit of any calibration coefficient should change the predicted moisture values by
much less than 0.1% moisture for all anticipated grain sample characteristics.
4.1.3. Calibration Equation Structure
The number of degrees of freedom in the prediction algorithm should be appropriate to fit the
dielectric response function to the moisture content. Too few degrees of freedom may limit the
instrument's ultimate accuracy. Too many degrees of freedom encourage data over-fitting.
Using linear interpolation between a number of calibration points would appear to be an extremely
flexible way of generating calibration curves. If the calibrations are to be produced by hand with a
pencil and graph paper, that is probably true, but calibrating with line segments is clumsy by
computer.
A small set of orthonormal basis functions that "naturally" fit the dielectric versus moisture
relationships work well in calibration equations. [6] Orthonormal basis functions are advantageous
because each term is independent of all others. An unfortunate choice of basis functions makes
calibrations unnecessarily complex and difficult to manipulate. For example, using Fourier
coefficients to model a nearly straight ramp function would require many more terms in the
calibration equation than using Legendre polynomials.
The interactions of all measured parameters should be considered when designing the prediction
algorithm. These interactions will affect the calibration development process. An algorithm that
requires iteration among several different steps in the calibration process may add a great deal of
cost over the life of the instrument. Ifpossible, eliminate "circular" interactions.
Physical insight, not just statistics, is important in developing an effective form for calibration
equations. A model that inherently fits many types of grain will save a great deal of effort and
money over an empirical fit that must be completely re-done for each new grain type. Dielectric
response is inherently continuous with moisture content. A model that allows step discontinuities
or other such unphysical curve shapes will lead to overfitting.
4.1.4. Secondary Correction Factors
Grain density and temperature affect dielectric characteristics profoundly and should be
considered very early in the model development. Poorly-conceived corrections significantly limit
moisture meter performance and complicate calibration development. Using separable linear
corrections for grain parameters such as temperature and density has some advantages over
combining them with dielectric parameters in a multi-term calibration equation because acquiring
data for samples that cover all possible combinations of temperature, density, and moisture
124
content is difficult. Separable correction equations facilitate development of corrections based on

a limited sample set. This may result in iterative calibration development processes, however.
4.2. Calibration Transfer
As discussed above, optimizing calibration coefficients for a grain is a costly and time consuming
process. These calibrations must be applied to commercial instruments to add value to grain
moisture measurement. This involves transferring calibration equations to different instruments
within a model and to different instrument models. Distributing the calibration coefficients is the
relatively easy part of the process. Ensuring that each instrument yields excellent accuracy with
the calibration coefficients is much more challenging.
4.2.1. Calibration Coefficient Distribution
Designers should plan how new and updated calibration coefficients will be distributed to
customers. If thousands of instruments of a model are produced, the update method needs to be
very efficient. Plug-in modules are secure and potentially provide a means of recovering the cost
of calibration development. However, the module approach makes rapid calibration updates
nearly impossible and creates documentation nightmares if customers demand different sets of
calibrations or special calibrations. Requiring that the instrument be returned to the factory or
service center for each calibration update is even more cumbersome.
Most grain moisture meters allow calibration coefficients to be typed in from a keyboard and
stored in non-volatile memory. The coefficient entry process should be very rapid and simple or it
will cause customer resistance to the model. The ubiquitous PC and the Internet offer effective
possibilities for rapid calibration distribution by downloading coefficients.
4.2.2. Calibration Transfer Across Models
Many have tried to transfer calibration equations developed for one instrument model to another
instrument model. Results are generally very disappointing unless the two instrument models are
virtually identical electrically and mechanically. Differences in measurement frequency, test cell
geometry, dielectric sensing circuitry, or secondary corrections are almost certain to cause
unacceptably large moisture measurement differences.
This difficulty in transferring calibration equations to different instrument models also applies to
grain moisture research. The relationships between moisture, density, temperature, and dielectric
characteristics so painstakingly derived on laboratory instruments have only qualitative
significance for commercial grain moisture meters with fundamentally different sensor
characteristics. Calibration equations could be more transferable from research to commercial use
and across moisture meter models if manufacturers designed moisture meters to measure standard
physical parameters. Measuring the real and imaginary parts of the dielectric constant plus sample
temperature and density would be consistent with much dielectric research. Even then, the
dielectric response and density would be dependent upon the specific test cell geometry.
Research is needed to determine how to minimize those residual errors.
125
4.2.3. Calibration Transfer Within Models-Instrument Standardization

Instrument standardization, for purposes of discussion in this paper, is defined as the process of
causing each instrument of a model to agree within acceptable limits with the physical or virtual
master instrument for the model. Ease and precision of standardization are probably the most
challenging aspects of developing and producing a successful grain moisture meter.
High expectations contribute to the difficulty of grain moisture meter standardization. Differences
in results (between instruments of the model) of over 0.15 percent moisture are completely
unacceptable for some grain testing situations. [7] Since moisture measurement differences have
direct impact on the value of grain in commerce, inconsistencies among instruments tend to result
in irate phone calls from customers who demand immediate improvements.
For moisture meters used in commerce, close agreement among units is required for all grain
types. No grain-specific standardization adjustments are allowed.[3] This requirement virtually
dictates that each measured parameter be individually standardized with sufficient precision that
their combined contributions fall within acceptable limits for all possible grain samples. Treating
the combined measurement systems as a single "black box" and expecting to "tweak" the final
result to get acceptable agreement among instruments is a fatal flaw in grain moisture meter
design.
Adequate standardization requires careful error analysis based on detailed mathematical modeling
of every subsystem. The error analysis will reveal which measured parameters are most critical to
the final results. In a typical elevator-type grain moisture meter, a 0.1 percent moisture difference
is induced by a 1 C temperature difference, a 1 gram sample weight difference, a 0.1 mm
electrode spacing difference, or a 0.1 pF capacitance difference.
Standardizing capacitance (or dielectric constant) measurements is quite challenging. An error
analysis of one of the most precise LCR meters available, the Hewlett-Packard-4285A\ provides
some appreciation for the problem. This instrument achieves an absolute accuracy at 2 MHz of
approximately 0.1 pF over the range of capacitance of interest for grain moisture meters.
During research or prototype development, it is all too easy to ignore the subtle effects of
mechanical, electrical, and instrument tolerances. Calibration equations developed using a single
instrument (or even several units built at the same time with components from the same
production lots) may show significant errors with instruments from the first production run of
instruments. During the first production run is NOT a good time to start to think about
instrument standardization.
Standardization difficulties can be reduced by design. The measurement algorithm should
minimize sensitivity to component differences and drift by using differences or ratios between
measurements. Built-in stable references for automatic re-standardization or performance testing
1 Mention of trade products or finn names does not imply endorsement or recommendation by the U.S. Department
of Agriculture over similar products or finns not mentioned.
126
are essential. Measurements that are designed to be inherently linear are easier to standardize
than nonlinear responses. Precise single-point standardization is much more efficient than
conducting multiple tests and creating an interpolation table to standardize the parameter. All the
potential sources of nonlinearity and waveform distortion hiding in the circuitry should be
examined and quantified. Eventually they will cause problems. Designing in more standardization
precision and flexibility than the minimum that seem to be required will be useful in handling
unanticipated standardization problems later.
An error analysis that considers standardization issues is a valuable addition to a research report.
How much of the scatter in the data could be due to the repeatability of the instrumentation?
How might the results change if the research were repeated with a different apparatus? Is there
potential for replicating the observed performance with a commercially viable instrument?
Research that addresses these questions is more likely to have positive impact.
5. SUMMARY
Moisture measurement instrumentation plays important roles throughout the grain industry and
presents significant opportunities for researchers and manufacturers. It is a difficult market,
however, with high expectations, established product lines, regulatory requirements, and technical
challenges. Careful consideration of the factors presented in this paper should aid those who
decide to face the challenges.
[1]
[2]
[3]
[4]
[5]
[6]
[7]
Lawrence, K.C. and Nelson, S.O. "Sensing Moisture Content of Cereal Grains with Radio-Frequency
Measurements--A Review," in Papers and Abstracts from the Third International Symposium on
Humidity and MOisture, London, England, April 6-8, 1998, edited by Stephanie Bell, Teddington,
Middlesex, UK, National Physical Laboratory, April 1998, pp. 195-203.
International Organization of Legal Metrology, "International Recommendation No. 59: Moisture Meters
for Cereal Grains and Oilseeds," Paris, France, International Bureau of Legal Metrology, 1984.
National Institute of Standards and Technology, "Sec. 5.56 Grain Moisture Meters," in NIST Handbook
44-Specifications, Tolerances, and Other Technical ReqUirements for Weighing and MeaSUring Devices,
edited by Tina Butcher, Washington, D.C., U.S. Government Printing Office, 1998, pp. 5.21-5.32.
National Institute of Standards and Technology, "Checklist for Grain Moisture Meters," in NCWM
Publication 14-National Conference on Weights and Measures National Type Evaluation Program-Administrative Procedures, Technical Policy Checklists, and Test Procedures, edited by Tina Butcher,
Washington, D.C., U.S. Government Printing Office, 1998, pp. 6.1-6.49.
Federal Grain Inspection Service, "Chapter 4: Air-Oven Methods," in Moisture Handbook, Washington,
D.C., United States Department of Agriculture, 1986, pp. 4.1-4.6.
Ariken, G. Mathematical Methods for PhYSiCists, San Diego, CA, Academic Press, Inc., 1985, pp. 516521.
Grain Inspection, Packers and Stockyards Administration, "Part 801.6-- Tolerances for moisture meters,"
in Code ofFederal Regulations, Washington, D.C., U.S. Government Printing Office, 1998.
Contact point: David B. Funk, U.S. Dept. of Agriculture, Grain Inspection, Packers and Stockyards
Administration, Technical Services Division, 10383 N. Executive Hills Boulevard, Kansas City, Missouri 64153,
USA Phone: + 1 (816) 891-0430; Fax: + 1 (816) 891-8070; E-mail: dfunk@tsd.fgiskc.usda.gov
127
SWEPT-FREQUENCY GRAIN MOISTURE SENSOR

Kurt C. Lawrence *, David B. Funkt, and William R. Windham *
*U. S. Department of Agriculture, Agricultural Research Service, Richard B. Russell
Agricultural Research Center, Athens, GA 30604-5677, USA klawrenc@ars.usda.gov
tInspection Systems Engineering, Technical Services Division, Grain Inspection, Packers and
Stockyards Administration, Kansas City, MO, USA
ABSTRACT. A shielded parallel-plate moisture sensor was designed, calibrated, and validated
to sense the moisture content (mc) of yellow-dent field com, Zea mays L., and soybeans, Glycine
max (L.) Merrill. The sensor uses complex reflection coefficient measurements and sample
weight to predict the moisture content of static samples at room temperature (-22C) with a
sample volume of 850 cm3. Multivariate analysis techniques were used to eliminate outliers and
to divide the population into calibration and validation sets. The sensor was calibrated with 29
com and 33 soybean samples of differing lots, grown throughout the U. S. in 1997, with mc
ranging from about 10% to 26% for com and about 8% to 16% for soybean, wet basis. The
standard error of calibration was 0.24% and 0.19% mc with a coefficient of determination of
0.997 and 0.995 for com and soybeans, respectively. The sensor was independently validated
with 77 com samples and 83 soybean samples. The validation bias was 0.04% and -0.02% mc
and the standard error was 0.28% and 0.18% mc for com and soybeans, respectively. The sensor
compares favorably with results obtained by commercial moisture meters.
Keywords: radio frequency, moisture content, corn, soybean
1.
INTRODUCTION
There has been a considerable amount of research on radio frequency (RF) sensing of moisture
content (mc) in agricultural products over the past decade [1]. Much of the research has centered
around on-line sensing of mc, where variations in sample bulk density affect the predicted mc.
To minimize this effect, density-independent equations and techniques are often utilized in
efforts to avoid the mc errors associated with flowing materials. These same techniques can also
be used to reduce measurement errors in static or batch systems, where measurements of a single
electrical parameter at a single frequency are typically used to sense mc. The current research
uses the complex electrical measurements at multiple frequencies to reduce the errors in sensing
the mc of static yellow-dent field com and soybeans.
2.
MATERIALS AND METHODS
Shelled yellow-dent field com, Zea mays L. and soybeans, Glycine max (L.) Merrill, were used
in this study. Both were a subset of samples used by the Grain Inspection, Packers and
Stockyards Administration (GIPSA) in their annual moisture calibration study for the 1997 crop
year. The com and soybeans were grown in 1997 at various locations throughout the U. S. All
samples were received in their natural condition; that is, water was neither artificially added nor
removed from the grain samples. Moisture contents ranged from about 10% to 26%, wet basis,
for the com and from about 8% to 16%, wet basis, for the soybeans. Of the samples received by
GIPSA, 113 com and 120 soybean samples were measured for this study. Oven moisture
128
measurements (wet basis) were obtained according to the USDA standard [2]. All measurements
were taken at room temperature (between 20.7 - 23.7 C).
A Hewlett Packard 1 8753C Network Analyzer was used to aid in the design and fabrication of
the sample holder. It was used to locate, identify, and isolate undesirable reflections along the
length of the sample holder, and to ensure that the sample holder was as closely matched to 50 0
as possible. The primary measuring instrument was a Hewlett-Packard 4291A RF Impedance
!Material Analyzer which measured the complex reflection coefficient (r = r x + jry) of the
sample and holder from 1 to 500 MHz in 2 MHz intervals. Measurements on both commodities
were taken with a shielded parallel-plate sample holder (figure 1) with a 50-0 characteristic
impedance. A detailed description of the sample holder and measurement procedures was
presented earlier [3].
The instrument was calibrated with an open, short, and matched load at the instrument test head.
A port extension was performed to extend the instrument reference plane to the type N connector
at the end of a coaxial cable. After instrument calibration, the sample holder was connected to
the instrument and a measurement sequence was initiated. To check for instrument drift or errors
in the system, an empty cell measurement and a check sample were measured each day. The
sample holder was uniformly filled, and excess material was removed prior to the 30-s
measurement sequence. After the electrical measurement, the sample was collected, weighed,
and the sample temperature was recorded.
3.
DATA ANALYSIS AND RESULTS
The data consisted of complex reflection coefficient measurements at 250 frequencies from 1 to
500 MHz, sample temperature, sample weight, test weight, and oven moisture determinations for
113 yellow-dent field com samples and 120 soybean samples. Most samples were of sound
grain. However, some com samples had mold on the broken surfaces and others had a "fruity"
smell. Some of the soybeans were shriveled. All samples were used in the initial analysis.
With measurements over a wide frequency range, it is difficult to distinguish sample outliers.
Therefore, multivariate analysis was used on the data sets (com and soybeans analyzed
separately) to check for population outliers and to divide the data sets into calibration and
validation data sets. The first task of multivariate analysis was to evaluate the data set for
spectral outliers [4]. Spectral outliers were samples that did not statistically belong to the sample
population. This analysis was performed with the r values only and did not include moisture
values. Sample loadings, eigenvectors for the first several principal components (those with the
largest eigenvalues) multiplied by the r values, were used to obtain a spectral distance, known as
a Mahalanobis distance [4]. Samples with large spectral distances (>3.0) were outliers and, for
calibration purposes, could not be considered as part of the population. Of the original 113 com
and 120 soybean samples, seven com and four soybean samples were found whose r values did
not fit the population. The com outliers were all measured on the same day. Two of the soybean
samples were measured consecutively and the other two samples were repeated measurements of
the same sample. Graphs of the com data indicated each had an unusual peak around 70 MHz
I Mention of company or trade names is for purpose of description only and does not imply endorsement by the U.S.
Department of Agriculture.
129
that was not evident on subsequent days. Similar unusual peaks were noted with the soybeans.
Thus, 106 and 116 samples were used for calibration and validation for the com and soybeans,
respectively.
After the outliers were removed from the sample population, the data were divided into
calibration and validation sets by comparing the Mahalanobis distance of one sample to that of
all other samples [4]. If the difference in the Mahalanobis distance between individual samples
was less than a specified cutoff value, those samples statistically had the same spectra and only
one sample from that "neighborhood" was needed for the calibration data set. This technique
assumes that only one sample is required to represent all samples in a neighborhood. It also
assures that a particular moisture level does not dominate the calibration simply because more
measurements were taken at a particular moisture level. With a Mahalanobis distance cutoff of
0.6, there were 29 samples selected for the com calibration set and 33 for the soybean calibration
set. The remaining 77 com and 83 soybean samples were placed in the validation data sets.
The data were then analyzed with partial least squares analysis, principal component analysis,
and with multiple linear regressions to minimize the number of measurement frequencies needed
for a moisture calibration. After numerous attempts, the multivariate techniques were not able to
identify frequency regions or bands that were prominently superior for moisture calibration.
Therefore, multiple linear regression was deemed sufficient to develop a calibration equation. r x
and ry at every twentieth frequency from 1 to 497 MHz (1,21,41,61, ... 481,497 MHz) were
used in the SAS RSQUARE mUltiple linear regression analysis [5]. SAS results showed that
many frequency combinations produced similar results, which was expected from the
inconclusive results obtained by the multivariate approaches. The mc calibration equation
selected for yellow-dent field com was
M =-32.200+209.855(rx )1 + 276.032(rY )41 +108.551(rY)141 +173451W
(1)
where M is mc in percent, W is sample weight in g, and the subscripts refer to the measurement
frequency, in MHz. The coefficient of determination r2 was 0.997 and the standard error of
calibration (SEC) was 0.24% mc for com. Besides the reflection coefficient terms, equation 1
contains a 11W term which can be considered as 1 I density (constant volume). For a given
material, increasing the sample weight has the same effect as an apparent increase in the mc
because of an increase in the measured dielectric constant of the material within the sample
holder. Thus, to correct for mc, sample weight should be inversely related to moisture.
Equation 1 was validated with 77 different com samples harvested throughout the U. S and
ranging from about 12% to 24% mc. The bias of the validation data set was 0.04% mc and the
standard error of prediction (SEP) was 0.28% mc. The graph of the calibration and validation
data for com is shown in figure 2.
Similar results were obtained for the soybeans. The calibration equation was
M=-25.615+37.368(rJ l -77.511(rY )41 +3s.167(rY \41 +12117IW
(2)
which produced an r2 of 0.995 and a SEC of 0.19% mc. The 83 validation samples had a bias of
-0.02% mc and a SEP of 0.18% mc. Figure 3 shows the results for soybeans.
l30
Since the data in this research were a subset of data used in the 1997 moisture calibration study,
some comparisons between equation 1 and commercial instruments are possible. However, one
should remember that only half the data from the full calibration study was used here, and that all
measurements were taken at room temperature (20.7 - 23.7 C). Also, recall that electrical
measurements in this research were taken with a sophisticated research instrument, not a
specialized measurement circuit suitable for a commercial grain moisture instrument. For the
1997 crop year, the moisture bias reported for the Dickey-john GAC 2100 was 0.08% mc and the
SEP was 0.33% mc. Thus, the parallel-plate sensor in this study had a moisture bias and SEP
slightly lower than the official moisture meter.
4.
CONCLUSIONS
A parallel-plate sample holder was designed to operate over the frequency range from 1 to 500
MHz. Complex reflection coefficient values were taken at room temperature with slightly more
than one hundred yellow-dent field com and soybean samples and collected from throughout the
U. S. in 1997. The final calibration, based on reflection coefficient measurements at about 1,40,
and 140 MHz, along with the sample weight, yielded a coefficient of determination r2 of 0.997
and 0.995, and a SEC of 0.24% and 0.19% mc for com and soybeans, respectively. Validation
resulted in a mean bias of 0.04% and -0.02% mc and a standard error of 0.28% and 0.18% mc for
com and soybeans, respectively. The system performance compared favorably with that of
commercially available instruments on a subset of the data used in a 1997 crop-year moisture
study performed by GIPSA. A simplified measurement circuit and processor would be required
for a prototype instrument.
REFERENCES
[1]
Lawrence, K. C., and Nelson, S. 0., Sensing moisture content of cereal grains with radiofrequency measurements - a review. Third Int. Symp. on Humidity and Moisture,
Teddington, UK, 1998,2, 195-203,.
[2]
USDA. Air-oven methods. Moisture Handbook, Ch. 4., Washington, DC., Federal Grain
Inspection Service, U.S. Dept. Agriculture, 1992,4.1-4.6.
[3]
Lawrence, K. C., Funk, D. B., and Windham, W. R., Parallel-plate moisture sensor for
yellow-dent field com. ASAE Paper No. 986022, 1998, St. Joseph, Michigan, ASAE.
[4]
Lawrence, K. C., Windham, W. R., and Nelson, S. 0., Wheat moisture determination by
1- to 11O-MHz swept-frequency admittance measurements. Trans. ASAE, 1998, 41(1),
135-142.
[5]
SAS, SAS/STAT Guide for Personal Computers, 6 Ed., Cary, NC, SAS Institute, Inc.,
1985, .269-336.
Contact Point: Kurt C. Lawrence, USDA, ARS, Russell Research Center, Athens, GA 30604-5677, USA. Phone:
(706) 546-3527, Fax: (706) 546-3633, email: klawrenc@ars.usda.gov
l31
............
./...........
'i
Figure 1. Parallel-plate sample holder for sensing moisture content of yellow-dent field com.
26r---~--
_ 24
c
CD
~ 22
Q.
'i
__
----~
______
__
--~
__
__
Calibration
Validation
Regression
20
(3 18
.E
.!!!
16
14
CD
>
12
10~--~--__----~--__- -__----~--__--~
10
16
22
14
18
20
Predicted Moisture Content, Percent
12
24
26
Figure 2 Calibration and validation me values predicted from equation 1 versus air-oven me
values for yellow-dent field com samples harvested in 1997 and measured at about 22C.
18r-----~--
1:
16
CD
____
______
~--
____
~----~
Calibration
Validation
Regression
Q.
~c
14
(.)
CD
E
., 12
"0
:E
~ 10
8~------~----~------
10
__
------~----~
16
12
14
Predicted Moisture Content, Percent
18
Figure 3 Calibration and validation me values predicted from equation 2 versus air-oven me
values for soybean samples harvested in 1997 and measured at about 22C.
132
GRAIN MOISTURE MEASUREMENTS WITH

TIME DOMAIN REFLECTOMETRY
Markus Stacheder, Peter Blume, Robin Fundinger, and Kurt Koehler
IMKO Micromodultechnik GmbH, Im Stock 2, D-76275 Ettlingen, Gennany
ABSTRACT. This paper reports on the development and demonstration of new sensors for the
determination of grain moisture content based on the TRIME-TDR teclmology.

Two types of sensors are presented. The first is an on-line sensor for automatic and continuous
moisture control of grain dryers (TRIME-GW). The system can be simply installed at the dryer
wall and measures directly in the grain flow. No influence of temperature up to 120C could be
found. The sensor was integrated into a SPS controlling system allowing easy monitoring and
processing of grain drying. The risk of over- and underdrying was significantly lower resulting in
energy savings and higher yield. The second sensor is a new mobile moisture meter (TRIMEGMC) for fast, precise, and reliable determination of moisture in whole grains and representative
sample volumes. Due to the results, the appropriatness for verification is granted.
Keywords: moisture, grain, TDR

1. INTRODUCTION
TDR and especially the TRIME-TDR technique is now a well established measurement method
for moisture determination of soils. But it is also applicable for numerous other porous materials
such as for the determination of grain moisture.
Two main applications can be distinguished here. To assess trading value and storage capability,
there is a need for precise and rapid mobile moisture meters. The second application focuses on
automatic moisture control for monitoring and processing the drying cycles in grain dryers in
order to minimize over- or underdrying and increase economic benefits. The incentives for
improved moisture control are considerable according to a study of [1]. Savings in energy of
0.5 .. 1 $/ton and additional sale margins of 1 .. 3 $/ton, depending on product, througput and
specific process, are realizable by using on-line measuring sensors.
Existing automatic control systems based on measuring temperature or the moisture of extracted
air are inherently imprecise, subject to the influence of interfering factors (air moisture,
temperature fluctuations) and partly require intensive maintenance. Consequently dryers are
nonnally controlled based on results of hourly manual samples analyzed for moisture in the lab.
A manually supervised and controlled drying process is labor-intensive and full of potential
errors because the teclmology utilized for measuring the grain moisture is antiquated or simply
has not been fully developed.
Manual moisture meters nonnally have too small sampling sizes, which are not very
representative, they mostly need grinding of the grain and have a limited moisture range. It is
well known that moisture meters based on resistance measurements are very sensitive against
changes in electrical conductivity of the grain, caused by site-specific mineral contents or varying
applications of fertilizers. Subject material that is still wann leads to even greater errors. All in
all a proper regulation of the dryer by manual control is an illusion.
133
Therefore we have been developing grain moisture sensors for on- and off-line use based on the
TRIME-TDR technology.
2. DESCRIPTION OF THE SENSORS
TRIME technology [2] was derived from the Time Domain Reflectometry principle, a
meanwhile well established electromagnetic measurement technique for determination of soil
moisture [3]. The transit time of a voltage pulse that propagates along metallic wave guides
yields the dielectric constant of the porous material, which is a function of the water content.
Although cereal grains such as com, wheat, rye, etc. are typical porous materials, there are some
differences to soil that had to be considered in the sensor design. The porosity of the bulk grain is
higher and the bulk densities are lower compared to soil. Also the gravimetric water content on a
wet mass basis is needed, and absolute accuracy of around 0.5 % by weight are indispensable.
2.1 On-line Sensor
The sensor for on-line measurements is a rod probe (Figure 1) with a high-temperature resistant
plastic probe head that can be easily installed at the dryer wall with an flange. Since moistures
occurring in grain products are
lower than in soil material, no
Sensor
PVC-coating of the two metallic
wave guides (0.15 m in length)
was necessary as it is normally
the case for the TRIME probes,
Device with connector
due to the new pulse evaluation Fixation nut
power-supply,
M70x2
IMP232-MICRONET
method [2]. The evaluation
or RS232N24,
electronics are housed in a small
analog output
aluminum die-cast box, a read
out unit with selection switch
150
allows selection of 15 different
calibration steps for grain
product and on-site calibration of
Display unit
the system for accurate gravi~40~ Rod diameter:
metric moisture determinations.
TRIME-GW's results can be Figure 1. Technical Data ofTRIME-GW on-line sensor for
displayed on LEDs for manual grain dryers (all dimensions in mm).
control, incorporated into a
monitoring system via an analogue input (0)4 - 20 rnA, or documented with a linear plotter or
PC. The modular concept of the system allows any enhancements.
2.2 Off-line Sensor

For off-line measurements we designed a mobile moisture meter with a cup-like probe
containing two plate-shaped wave guides at the inner wall of the cup and having a volume of
450 cm3 (Figure 2). The grains can be filled in the cup without preparation such as grinding. To
134
measure the moisture on a weight

basis, which is indispensable for
trading grain, we combined the
sensor with a weighing-cell. Wet
mass and volume of the sample yield
density, so that with the TRIME
measured volumetric moisture, the
gravimetric moisture can easily be
determined.
Canyingcas
3. RESULTS
3.1 On-line Sensor
The on-line-sensor was tested in pilot
experiments with different types of
dryers. Installation and operation
were easy and the sensor worked well Figure 2. TRIME-GMC mobile moisture meter (dimeneven under temperatures of up to sions: 220 x 280 x 290 mm)
120C. Figure 3 shows the results of
a TRIME-GW probe during a drying process of maize in a conventional rotary dryer. Reference
values were determined by manual sampling at the same spot and drying the samples in an oven.
1----------------------
10
20
30
40
50
60
70
80
Sarrple Nurri:>er
L _ __
Figure 3. TRIME-GW measurements and oven drying method for maize in a rotary dryer.
The correlation is very good, only at moistures higher than 20 % b.w. TRTh1E-GW Wlderestimates moisture. This is mainly due to the calibration function which was established with
rewetted grains and has to be adapted. In the beginning users also complained about fluctuating
135
values from 0,4 % .. I % caused by the strong movement of grain during outlet times. Elongation
of average time to 2 minutes stabilized the displayed value. Finally the accuracy was such that
the sensors could be integrated into a SPS processing and controlling system (Siemens!
SIMATIC S7) for automatic operation of the dryer [4].
The results of Figure 3 also impressively demonstrate the common habit of overdrying to add a
margin of safety, which results in a considerable financial loss due to a higher energy demand
and a smaller yield.
3.2 Off-line Sensor
The TRTh1E-GMC was calibrated with different grain products (maize, wheat, rye, barley) and a
universal calibration function was established. Then the mobile moisture meter was tested
intensively at the German Federal Institute of Physics and Techniques (PTB) with several
different grain types during harvest season 1998. Figure 4 shows the results for maize. Reference
method was the oven-drying method according to ISO 6540 [5].
10
15
20
25
30
35
40
Wet Mass Moisture [% b. weight]
Figure 4. Deviation ofTRIME-GMC moisture measurements from oven-drying method

(reference) for maize (verification limits of "class I (II)"-devices: < 10% moisture = 0.4% (0.5%)
absolute moisture, > 10% moisture = 4% (5%) from absolute moisture).
Even with the universal calibration function, established with rewetted grain products, and with
no grinding necessary, TRIME-GMC fulfills the stringent limits for 'class I'-device-rating for
maize of up to 40 % b.w. moisture. The appropriatness for verification was granted for 1999. As
can be deduced from the figure, even an improvement of the results is achievable by special grain
type calibrations. Also measurements in oil seeds like rape yield a standard deviation of 0.26 %
b.w. in the range of6 ., 14 % b.w. moisture.
I
Reference to specific products for the benefit of the reader only.
136
4. CONCLUSION
TRIME-TDR technology could successfully be adapted for wet mass water content determination
in grain products. Both a mobile moisture meter for the trading industry and an on-line sensor for
direct control of grain dryers was designed.
Advantages and accuracy of the new sensors are considerable. Representative measuring
volumes and whole grain measurements minimize errors due to sampling and milling and
enhance the ease of operation. Influences of bulk electrical conductivity (mineral content) or
grain temperature on the measurement are very low so that even with a universal calibration
function the accuracy necessary for the certification for verification could be fulfilled.
The on-line sensor offers the possibility of continuous measurements for on-line monitoring of
grain moisture in dryer and delivery areas with excellent spatial and temporal resolution. An
integration in processing and controlling programs is possible for reliable automatic operation of
dryers with reduced risks in over- and underdrying and higher economical benefits due to energy
savings and higher yields.
REFERENCES
[1]
Conrad, T., Possible savings in the grain drying process using precise and fast on-line measurement
technologies. Steinbeis Transfer Center for Agronomy, Environmental Technics and Power EconomiesInternal Report, University ofHohenheim,Germany, 1997,8 p.
[2]
Stacheder, M.; Fundinger, R & Koehler, K., A new time domain reflectometry system (TRIME) to measure
soil moisture and electrical conductivity. Proc. Symp. on TDR in environmental, infrastructure, and mining
applications, 1994, U.S. Bureau of Mines Spec. Pub!. SP 19-94: 56-65
[3]
Topp, G.C.; Davis, J.L. & Annan, A.P., Electromagnetic determination of soil water content: Measurements
in coaxial transmission lines. Water Resour. Res., 1980, 16: 574-582.
[4]
Blume, P., Fundinger, R., Koehler, K., Pfeiffer, B.-M., and Stacheder, M., Automation of grain drying plants
(Automatisierung von Trocknungsanlagen, in german). Siemens Report, 1999, in press
[5]
International Standard ISO 6540, Maize - Determination of moisture content (on milled grains and on whole
grains). International Organization for Standardization, 1980.
Contact point: Markus Stacheder, IMKO Micromodultechnik GmbH, 1m Stoeck 2, D-76275 Ettlingen, Germany.
Phone: + 49 (0) 72435921-17; Fax: + 49 (0) 724390856; E-mail: m.stacheder@imko.de
l37
A QUADRUPOLE TO ASSESS THE SPATIAL DISTRIBUTION OF

PLANT WATER CONTENT IN CROPS
J Helbert
INRA, Unite de BiocIimatologie, 78850 Thiverval Grignon, France.
helbert@bcgn.grignon.inra.fr
P. Boissard
INRA, Unite de BiocIimatologie, 78850 Thiverval Grignon, France.
boissard@bcgn.grignon.inra.fr
J-P' Buis.
CIMEL Electronique, 5 cite de Phalsbourg, 75011 Paris, France.
cimel@worldnet.fr
Abstract
An electromagnetic quadrupole allowing the estimation of wheat spike water content by a non
destructive in situ method is presented. Experimental results show a high correlation between
measurement and wheat spike water content. A simulation study is also presented. It computes
the measured signal from the dielectric characteristics of the crop, particularly the permittivity of
the spikes and the stems.
Keywords: quadrupole, dielectric permittivity, wheat, water content.
1. INTRODUCTION
Estimating the water content of plants, and particularly the wheat spike water content, is a very
important point since it can inform on the total biomass of the crop and the potential yield
[1,2,3]. The biomass amount is also a basic datum if one wants to control nitrogen supply. With
existing models, such as the dilution curve model [4], one can infer the nitrogen need from the
biomass amount. Knowing the grain water content is also important to decide the best time for
harvesting.
A new device allowing the estimation of water content of vegetation has then been developed by
CIMEL Electronique (Paris, France) in collaboration with INRA (Institut National de fa
Recherche Agronomique, France) and the Departement de Geophysique Appliquee (University of
Paris VI, France) [5]. It can assess both the horizontal and the vertical distributions of water in
crops by a non destructive in situ method. Until now, it had been used to follow stages of
development and vegetative growth, particularly the filling of wheat spikes. Using such a device,
it is thus possible to estimate under unstressed conditions the biomass of a crop, and by
extension, the yield. In the future, the purpose is also to detect rapid variations of crop water
content related to water stress and to decide irrigation with the aim of adjusting water supply to
the real needs of plants.
2. THE INSTRUMENT
l38
2.1. Presentation
The instrument is an electromagnetic quadrupole, two electrodes injecting an alternating current
and two others allowing the measurement of a voltage. It is derived from the quadrupoles used
for prospecting in geophysics [6].
The two pairs of electrodes are placed horizontally at the height of the spikes (figure I). The
dimensions of the quadrupole are quite small and make the instrument easily transportable in the
field: each electrode is cylindrical with a diameter of about 10 cm and placed on the vertices of a
40 cm x 40 cm square.
Pair of injection
electrodes
Pair of reception
electrodes
Figure 1. View of the quadrupole in situ.

The frequency of the injected current has been chosen so as to make the measurement
independent of the plant geometry. Therefore, the electomagnetic waves are supposed to be only
absorbed by the constitutive water content of the canopy volume. Resulting data represent
therefore a volumetric measurement.
2.2. Theoretical aspects
The instrument injects a current of intensity I, the resulting electromagnetic wave then interacts
with the plants, the air and the soil, and the device measures a voltage V between the reception
electrodes. The ratio VII is the impedance of the medium and is an inverse function of the
apparent dielectric permittivty c; *, which can be written in the following complex form:
(J
=8-1--
2nf
l39
(1)
where s is the real dielectric pennittivity, (j the electric conductivity and f the frequency of the
electromagnetic wave. Thus, the signal is composed of two voltages in quadrature. The
component in phase with the injected current corresponds to the imaginary part of the
pennittivity while the component in quadrature with the signal corresponds to the real part of the
pennittivity.
Furthennore, the crop is mainly composed of air and water, and the real pennittivity of water is
eighty times the pennittivity of air. Thus, concerning the real part of the pennittivity,
measurement is sensitive to the crop water content. And then, the higher measurement, the lower
crop water content.
3. FIRST MEASUREMENTS
First measurements using the quadrupole were made by Fechant on wheat crops [5]. These
experiments concerned the decrease of spike water content before maturity to point out the
evolution of the grain filling. The quadrupole was placed in a wheat crop and a measurement of
the component in quadrature was made. After the measurement, a sample of plants located
between the electrodes were taken from the crop to estimate by drying and weighting the spike
water content.
Figqre 2 presents the measurements as a function of the spike water content and figure 3 reports
the evolution along time of the measurements and the spike water content. It can be seen on the
two figures that there is a high correlation between water content and measurements and that the
quadrupole clearly detects the rapid water loss, corresponding to the end of the grain filling [3].
:>
5.,
.,
.,...
<:.>
~
:t:
:0
-'.,a
'"
p..
650
600
550
500
450
400
350
300
250
200
150
- 6(X)
- 5:Xl
>
.s
4>
g
...4>
4(X)
. 300 ~
"0
-;ro ~
- 100
.. a
0
200
400
600
800
1000
Spike water content (zlm2)
1200-
175
180
185
100
195
c2
o
0.
Z):)
03yofyear
~ Spike v.ater o:ntent (9m2) - - A:XentiaJ
difference (mV)
Figure 3. Detection of water loss with the

quadrupole (after Fechant [5]).
Figure 2. Measurements made with the

quadrupole (after Fechant [5]).
4. SIMULATION
A simulation study has also been performed in order to detennine the influence of the dielectric
pennittivities of the various media to the measured signal. Since the wavelength of the injected
140
current is very large in comparison to the dimensions of the spikes, the quasi-steady state
approximation can be used and the purpose is thus to solve the Laplace's equation: !1V = 0,
where V is the electrostatic potential.
The software used to compute the electrostatic potential is Modulef (INRJA, France), edited by
Simulog (France), and is based on the finite element method [7,8]. The domain of simulation
was meshed and splitted up into four layers representing the different media: the air, the spikes,
the stems, and the soil (figure 4). In a first step, we assume that the soil conductivity did not
contribute to the output signal, and therefore a real dielectric permittivity was attributed to each
layer. Then, the potential, fixed on the injection electrodes, was computed in the 3D space on
every knot of the mesh, and was forced to be constant on each reception electrode.
....~., .\~ . .....
~.
600
1.5
."'
550
:> 500
a
'-'
450
0
0
!:l
...00 400
350
1:::
:a 300
-;
.~ 250
0
200
0
~
150
100
---
IStems I
Gl ....
..~
2.0 B
'.
2.5 ",x,
2.7 -- ... --
:'~;::':;:k;;:. . .
2
3.0
"l<'
16
18
'- .
...
--0--
1.7 ..... + ....
..,
10
12
14
Spike dielectric permittivity
20
Figure 5. Potential difference in function

of spike dielectric permittivity for several
stem permittivities.
Figure 4.Vertical section of the mesh.

Detail showing the four layers and the
two electrodes (one used for injection
and the other for reception).
Figure 5 represents the potential difference between the reception electrodes as a function of the
permittivities of the spikes and the stems (the real dielectric permittivity of the soil did not
appear to influence the potential). It can be seen that the stem layer influences the potential, but
its contribution decreases when the spike permittivity becomes more important. So, to estimate
precisely the lower values of spike water content, it could be interesting to simultaneously know
the stem water content. This could be achieved by making multiple measurements at two heights
at least inside the canopy.
5. CONCLUSION
First, experiments made with the quadrupole showed that the potential difference between the
reception electrodes is highly correlated to the spike water content, and that the instrument can
be used to follow the evolution of spike water content during the grain filling period.
141
Secondly, the results of the simulation study confirm that measurements are correlated with the
spike permittivity. However, an uncertainty remains on the estimation of the lower values of the
spike permittivity due to the contribution of the stems.
In order to conclude on the feasibility of the in situ water content estimation by using the
quadrupole, we can see that the uncertainty on the measurement is about 15%, which
corresponds to an uncertainty of about 20% on the spike water content. The remaining problem
to be solved is then to reduce the measurement noise by improving the electronic device and also
by controlling the in situ conditions such as the water deposits on plant organs and the soil
moisture.
ACKNOLEDGEMENTS
The authors want to thank Prof. O. Picon (University of Marne-la-Vallee, France) for her
contribution to this work.
REFERENCES
[IJ Huet, Ph., Pons, Y., Saur, L. Detennination de la fin du remplissage du grain de ble a partir de I'humidite de I'epi.
Application a des experiences multi-sites. Colloque INRA-ITCF Eco-physiologie du ble, Versailles, France, 25th
and 26th April, 1989. 9p.
[2J Huet, Ph., Pons, Y, Saur, L. Remplissage de l'epi: elements de modelisation et comparaison entre sites. Colloque
INRA-ITCF Eco-physiologie du ble, Versailles, France. 25th and 26th April, 1989. 8p.
[3] Schnyder, H., Baum, U. Growth of the grain of wheat (TriticuIJJ aestivuJIl L.). The relationship between water
content and dry matter accumulation. Eur. J. Agron., 1992,1(2),51-57.
[4] Lemaire, G., Salette, J. Relationship between growth and nin'ogen uptake in a pure grass land. I. Environmental
effects. Relation entre dynamique de croissance et dynamique de prelevement d'azote pour un peuplement de
graminees fourrageres. I. Etude de l'effet du milieu. Agronomie, 1984,4(5),423-430.
[5J Fechant, C. Realisation d'un quadripole de mesure in situ de la perrnittivite dielectrique des vegetaux. Premiere
application ala determination du contenu en eau des epis de ble. PhD Thesis, Uni\'ersite de Paris VI. 1996. 190p.
[6J Tabbagh, A., Hesse, A., Grard, R. Determination of electrical properties of the ground at shallow depth with an
electrostatic quadrupole: Field trials on archeological sites. Geophys. Prospect., 1993,41(5),579-597.
[7J Du Toit, H. F., George, P. L., Lang, P., Pate, P., Steer, D., Vidrascu, M. An introduction to Modulef. Modulef
user guide #1. INRIA. 1995. 309p.
[8J Zienkiewicz, O. C..Thefinite element method, McGraw-Hill, Maidenhead, England, 1977, 787p.
Contact point: J. Helbert, INRA, Unite de Bioclimatologie, 78850 Thiverval Grignon. Phone: +33 (0) 1 30 81 55 01.
Fax: +33 (0)130815563. E-mail: helbert@bcgn.grignon.inra.fr
142
A PRACTICAL METHOD OF MEASURE MEASURING

USING INSULATED MONOPOL
A. WiSniewski
TeleconUnunications Research Institute, Gdansk Division,
al. Gen. 1. Hallera 13, 80-401 Gdansk, Poland. adaw@pit.gda.pl
P.S. Dfbicki
Technical University of Gdansk, Faculty of Electronics, Telecommunications and
Informatics, ul Narutowicza 11112, 80-952 Gdansk, Poland, pid@pg.gda.pl
ABSTRACT. Research has demonstrated that insulated monopol can be used for measuring
moisture content in loose substances. Insulated Antenna Theory by King (IAT) has been
employed in the analysis. The choice of parameters has been discussed' and illustrative
results for moist clay presented. Research was carried out at frequency f=2450 :MHz.
Keywords: moisture, measurements, insulated monopo/

1. INTRODUCTION
Water is the most popular matter on earth. As a component of various materials it
determines their electrical, chemical and mechanical properties. This is the reason why the
process of controlled water concentration- during dehydration or humidification- is
especially significant in many industrial applications (e. g. in agriculture or food industry). It
is required for two basic reasons:
a) in order to lower the cost of energy consumed during dehydration of products,
b) in order to achieve high quality products in the process of dehydration or humidification.
A microwave method of measuring moisture content in loose substances (such as clay, sand
or wheat), has been presented. The above substances are characterized by a low dielectric
constant 8'50 (Table 1), and hygroscopicity. Their frequency, temperature and density
also have an effect on their permittivity [1].
Table 1. Permittivity in selected loose substances [2,3].
e = e' - je"
Substance
Argillaceous soil 2.27 - 10.034
11.3 - j2.830
Clay
10.0 - i1.46
2.60
- jO.150
Sand
6.80 - i5.65
2.7 - jO.0022
Soil
20.0 - 10.22
Gypsum
3.5-iO.75
11.7 - j6.8
143
Comments
T=25, moisture 0%, f=10GHz
T=25, moisture 17%, f=10GHz
T=25, moisture 10%, f=lOGHz
T=20, moisture 0%, f= 1OGHz
T=25, moisture 9%,f=lOGHz
very dry
moisture
T=25, moisture 5%,f=lOGHz
T=25, moisture 15%,f=10GHz
2. MATERIALS AND METHODS

Insulated mono pol used for measuring moisture content in loose substances has the
following benefits:
a) can be applied in measuring various loose substances,
b) is easy to place in a given substance,
c) is sufficiently structurally resistant.
The primary instrument used for analysis of insulated monopol (Figure 1) is Insulated
Antenna Theory IAT radiating into loss matter, elaborated by King [4]. Despite certain
difficulties in applying this theory after some modifications [6] it is suitable to be employed
in this case.
2c
ftiMlIHh
o
FEEDING LINE
Figure 1. Structure of insulated mono pol.

In IAT formula the following are introduced: complex wave number in monopol kL and
complex characteristic impedance of the monopol Zc treated as a long loss line. Given that
k3 IkJI ~ 2lk2 1 expressions for kL and Zc are the following:
(1)
z = ~[lnYa +(k2)2 H~2)(k3b)

c
2trwC2
k3
k 3 bHf2)(k3 b)
1
(2)
where: k;- wave number ofi center,

b - insulation radius of the monopol,
a - inside radius of the monopol,
82 - complex permittivity of insulation of the monopol.
Entry impedance of the monopol radiating into loss matter, treated as a long line, can be
designated on the basis of the well-known correlation between comp!ex wave number kL
and complex characteristic impedance Zc :
144
- _.
wej -
Zc
J tg( k L h)
(3)
For illustration figures 2, 3 present theoretical figures of entry resistance and entry
reactance for monopol, size 2a=3mm, 2b=7mm, 1:;2=2, operating at frequency f=2.45GHz,
placed in moist clay. As shown in the figures significant range of variation of entry
impedance in function of moist clay is observed for monopol h=33-45mm long. It appears
from detailed analysis that the greatest variations in coefficient of reflection at entry of the
monopol are observed for h:=44mm, when the greatest relative changes of entry impedance
of the monopol are observed.
5
....::J
""-~-''''''-w=7%
-CIl
-
--w=O%
-oC
CIl III
.~ t)
iii'iii
o
z
~3
a
:5
- w=10%
10
15
20
25
30
35
40
45h[mm]
Figure 2. Resistance of the monopol placed in moisture clay in its length function.
2
h[mm]
45
w=10%
-6
Figure 3. Reactance of the monopol placed in moisture clay in its length function
In order to measure moisture accurately parameters of insulated monopol must be selected
in such a way that a minor change in moisture content should cause significant changes in
value of coefficient of reflection. Insulated monopol meets this condition when its length is
close to anti-resonance length, its insulation is thin and permittivity low.
Monopol of inside diameter 2a=3mm, 2b=5mm and 2b=7mm, h=44mm, Zc=50.o has been
analyzed in detail. Figure 4 presents the calculated properties of coefficient of reflection of
the above monopol placed in clay with moisture w=O%+ 10% (value of permittivity of clay
with various moisture content levels has been designated on the basis of modified Wiener's
method [6]).
As presented in figure 4, when insulation of monopol is thinner (2a12h=3/5), the range of
variation of coefficient of reflection becomes greater (both for magnitude Irl and phase <p).
Parameters of insulated mono pol can be further improved by modification of impedance of
feeding line. Table 2 presents the range of variation of coefficient of reflection in the
function of moisture of clay, given that line reference impedance is Zo=300, Zo=500,
Zo=80o. It is observed that when impedance of feeding line is greater than Zo=50 the
range of variation of phase of coefficient of reflection <p becomes greater.
145
Table 2. Range of variation of modulus and phase of coefficient of reflection of insulated

mono pol (h=34mm, 2b/2a=5mm/ 3mm) placed in clay with humidity 0%-10%.
Range of
Feeding line
imp_edance Zo
Range of
variation . q:>rol
rOl variation 111[dBl
30
4.15
23
50
5.2
47
80
2.36
72
3.RESULTS AND DISCUSSION

Insulated monopol eventually used for measuring moisture had the following parameters:
h=44mm, 2a=3mm, 2b=5mm, Zo=500. The following substances have been analyzed:
sand, clay, wheat and crop mixture. Figures 5, 6 presents results of the analyses. It is clear
that there are discrepancies between the calculated (Figure 4) and measured (Figure 5, 6)
properties, the reason being the applied model of description of moisture clay [6]. It can be
gathered on the basis of measured properties that most accurate information about moisture
is obtained on the basis of the value of coefficient of reflection <po A drawback of this
solution is the presence of ambiguities at very low values of moisture content.
The same monopol can be used for measuring moisture in other loose substances [6].
o
-8
-10
10
-70
>~
-80
-90
---
12
WlldB]
-0- 2a12b=315
-14
-2a12b=315
m(%]
-100
arg (I")
-110
-120
16
-130
-18
-140
-150
-20
-160
-22
-170
Figure 4. Calculated coefficient of reflection of insulated mono pol placed in clay.

4. CONCLUSIONS
The present world tendency to save energy and improve quality of products creates the
need to control water contents in products. Subject matter of this thesis is employing
insulated monopol to measure moisture content of loose substances such as clay, sand,
wheat and crop mixture. Monopol of this type is of simple construction and can be used for
penetration of deeper layers of substances.
Using microwaves to measure humidity seems an attractive method. The results of
measurements are immediate. An additional benefit of the method is that humidity in
substances at very high temperatures can be designated.
146
180
160
140
C
. 120
100
80
moisture fk]
60
2
10
12
14
Figure 5. Measured of coefficient of reflection (phase shift) of insulated monopol placed in

clay.
-2
-4
-6
-8
-10
-12
-14
+-----~----~------r-----~----_r----~----~
10
12
14
moisture r;..]
Figure 6. Measured of coefficient of reflection (module) of insulated monopol placed in

clay.
REFERENCES:
[1] A. Kraszewski, Microwave aquametry, New York, IEEE Press, 1996
[2] Journal ofMicrowave Power, Vol. 15, No 4, 1980.
[3] Journal ofMicrowave Power, Vol. 8, No 1,1973,s. 59-61
[4] R W.P. King, S. Smith, Antennas in Matters, MIT Press, 1981
[5] A.Wisniewski, Coaxial probe for moisture ofmaterials testing, Gdansk, 1998
[6] P.S. Debicki, M. Astrahan, Calculating input impedance of electrically small insulated antennas for
hyperthermia, IEEE Trans. Microwave Theory Tech. ,41,357-360
Contact point:
Adam Wisniewski, Telecommunications Research Institute, Gdansk Division, al. Gen. 1. Hallera 13,
80-401 Gdansk, Poland. Phone +48(58) 302-32-75, Fax: +48(58) 302-17-31; E-mail: adaw@pit.gda.pl
P .S. Debicki, Technical University of Gdansk, Faculty of Electronics, Telecommunications and Informatics,
ul. Narutowicza 11/12, 80-952 Gdansk, Poland. Phone +48(58) 347-16-59; E-mail: pid@pg.gda.pl
147
MOISTURE GRADIENT STUDIES IN TIMBER BY THE

MEASUREMENT OF DIELECTRIC PARAMETERS USING A
MULTIPLE ELECTRODE ARRANGEMENT
Sina Jazayeri
Eurothenn Controls Ltd, Faraday Close,

Worthing, BN13 3PL, UK, sina.jazayeri@controls.eurothenn.co.uk
KemalAhmet
Faculty of Science, Technology and Design, University of Luton, Park Square, Luton,
Bedfordshire, LUI 3JU, UK, kemal.ahmet@luton.ac.uk
ABSTRACT. This paper describes the second part of a study into the detection of transverse
moisture gradients in timber using a method of precision capacitance measurement.
Measurements were carried .out using a multiple electrode arrangement on stacked veneer
samples where systematic moisture gradients were introduced. Results from utilising various
configurations of live electrodes show that although the scanning depth is affected by the
moisture profile, a consistent increase is observed in the effective penetration depth of the
electromagnetic waves as electrode separation is increased, yielding accurate moisture profiles to
moderate depths.
Keywords: moisture gradient, moisture measurement, electrode
1 INTRODUCTION
Mismatches between initial timber moisture content and the value in the environment of
installation frequently causes distortion and warping and affects various properties.
Consequently, the ability to accurately quantify moisture content is important to the timber
industry. Measurement of moisture content in timber is associated with various complications,
many of which are due to the inhomogeneous structure of the material. Moisture gradients, too,
severely affect the accuracy. The process of commercial timber drying involves evaporating
moisture from the surface of the wood, often carried out by kiln-drying. The deep-seated
moisture gradually moves towards the surface, setting up a moisture gradient - drier towards the
outside where the water is being evaporated, wetter towards the centre; this is known as the
drying gradient.
Moisture gradients can also occur when low moisture content timber is
exposed to high relative humidity or dampness. In such cases the moisture gradient is known as
the regain gradient in which the timber surface absorbs moisture from the environment.
Although some manufacturers quote as great as 40 mm for the sensitive scanning depth of their
hand-held meters, various studies have shown that existence of transverse moisture gradients
severely affect the accuracy of dielectric moisture meters (see for example [1] and [2]). It was
previously reported that in presence of transverse moisture gradients scanning depths could be
reduced to only a few millimetre where conventional planar electrodes are used [3].
148
METHODOLOGY
A computer controlled resonance circuit (described in [1]) was used in conjunction with a multielectrode set-up to make measurements on samples with built-in systematic moisture gradients.
For these trials the electrode configuration consisted of six equidistant rectangular poles
(20 mm x 5 mm) with centre-to-centre separation of 6 mm. The aim was to investigate the effect
of electrode pole separation on the effective penetration of the electric field into the samples.
The electrode configuration is shown in Figure 1.
PCB
E-field
E-field
Figure 1: Electrode configuration used.

The electrode configuration was utilized in a procedure involving a sequence of measurement
cycles. During the first measurement cycle, pole 1 was used in conjunction with pole 2. Having
completed the measurements on all samples, pole 2 was disconnected and pole 3 connected up
for the next measurement cycle. Similarly, measurement cycles 3 to 5 utilized pole 1 with poles
four to six, respectively. Although all five combinations were used during the study, it was found
that data collected from pole 6 were not repeatable to within the required tolerance, 0.05 pF.
3
PRELIMINARY INVESTIGATIONS: UNIFORM M.e.
Initial investigations concentrated on samples with uniform moisture content; Figure 2 shows the
correlation between the oven dry moisture content and electrode-sample capacitance, for these
samples. Early tests confirmed that measurements on samples with uniform moisture content
yielded similar trends for all pole combinations; it was further found that the required accuracy
and repeatability could be obtained with maximum pole separation of about 24 mm (poles 1 - 5).
Figure 2 was obtained using poles 1 and 5. Each data point is the mean of five measurements
taken during the course of each measurement and was repeatable to better than 0.05pF.
A clear correlation exists enabling the relationship between the moisture content and capacitance
to be established. The equation for the line of best fit in the example shown is well described by
C=0.1927exp(6.83xlO-2MC), where C is the electrode-sample capacitance and MC is the percent
oven dry moisture content. These calibrations were used for the tests in (5) below.
149
u:
1.2
.e:
Q)
(.l
c
J:!l
0.8
'utV 0.6
Q.
tV
(.l
0.4
l!! 0.2
:I
1/1
tV
Q)
:e
10
15
20
25
30
Percent oven dry moisture content
Figure 2: Correlation between capacitance and moisture content where poles 1 - 5 were utilised.
4
PRODUCING SYSTEMATIC MOISTURE GRADIENTS
In the main experiments, tests were carried out on samples with systematic moisture gradients as
defined in Table 1 (under Gradients). The variation in the moisture profile in each sample was
achieved by positioning preconditioned veneers in a stack arrangement such that the change in
moisture content per unit length toward the central veneer determined the type of each gradient! .
Table 1: Details of the three gradients investigated; moisture contents for each veneer and
average moisture content ( MC) for specified number of layers in the stack have been specified.
Gradients
Veneer
N0 2
1
2
3
4
5
6
7
8
9
Regain
25.1
25.1
13.6
13.6
7.5
13.6
13.6
25.1
25.1
Steep Drying
(%MC)
7.5
18.4
18.4
25.1
25.1
25.1
18.4
18.4
7.5
Drying
Regain
MC(%MC)
Steep Drying
13.6
18.4
18.4
18.4
25.1
18.4
18.4
18.4
13.6
25.1
25.1
21.3
19.4
17.0
16.5
16.1
17.2
18.1
7.5
13.0
14.8
17.4
18.9
20.0
19.7
19.6
18.2
Drying
13.6
16.0
16.8
17.2
18.8
18.7
18.7
18.7
18.1
1 Details of material preparation for the experiment was fully discussed in [1].
2 Veneer numbers refer to the position of each veneer in the stack; veneer number 1 is the top veneer in contact with the
electrode and 9 is furthest away. See [1] for further information regarding the construction of the samples.
150
RESULTS IN THE PRESENCE OF MOISTURE GRADIENTS
These tested the effectiveness of the moisture gradient detection technique. Plots were made of
average moisture contents to various depths in the stack as tabulated in Table 1 (MC) above,
versus measured moisture content for four pole-pole combinations namely PI 2 PI 3 PI 4 PI 5'
These plots were used to check the effectiveness of this technique for m:ois~e gr~dien:t
detection. Table 2 summarises the findings from tests on the steep drying and regain gradients.
Table 2: Measured electrode-sample capacitance for the regain (Cr) and steep drying (CJ
gradients for four pole to pole separations.
Pole Separation
(mm)
6
12
18
24
Pole Combination
P I_2
Pu
P I_4
P I5
Cr (PF)
Cd (PF)
0.98
0.71
0.60
0.62
0.33
0.53
0.68
0.75
Figure 3 shows the superposition of measured values of average moisture content at various
depths, for the regain gradient. Each measured value (the cross symbols) has been plotted at the
nearest possible quantized depth with nearest matching moisture content. Capacitance data was
converted to moisture content using appropriate expressions; although the equation quoted above
is that for poles 1-5, preliminary experiments showed that all pole combinations produced near
identical results for the reference samples. The measured values reflect the true nature of the
moisture profile to depth around 10 mm. Figure 4 shows an analogous plot for the steep drying
gradient. Observations show that the moisture profile is reproducible to depth around 14 mm. In
this case the decrease in the electric field strength is considered to be partially compensated with
the increase in the average moisture content in the profile.
26.00 ,.--_ _ _ _ _ _ _ _ _- ,
CD
~ 24.00
i!!
22.00
.~
CD
20.00
'"
!!!
18.00
C
~ 16.00
~
14.00
o
x
0
0
+--~~~_~~
024
_
M
_,__-l
re re 20
Depth in the stackJnm
Figure 3 : Average measured moisture contents to defined, and measured values. Data presented
are for samples with the regain gradient. Crosses mark the data points obtained using pole 1 in
conjunction with poles 2, 3 and 4, from left to right respectively. The moisture profile
throughout the sample is shown at various depths.
151
21.00
c:
19.00
17.00
...
<>
<>
<>
<>
l!!
~ 15.00
'0
.,E
13.00
.,f!>
11.00
9.00
<>
Q)
OJ
.,
@
0..
7.00
10
12
14
16
18
20
Depth in the stack Inm
Figure 4: As Figure 3, but for the steep drying gradient.

These findings show that the multiple electrode arrangement can be used to detect gradients to
depths typically 14 mm. This is a realistic distance over which naturally occurring and process
induced moisture gradients often occur. Due to the symmetrical nature of the moisture gradients,
both in this study and generally in drying processes, the multiple electrode arrangement is
suitable for samples with overall thickness at least 20 mm. In addition, the arrangement used
relies on more than one pair of electrodes for evaluation of moisture constant and is therefore
more accurate than the typical single pair electrode arrangement used by dielectric moisture
meters.
6
CONCLUSION
The results from the study show that multiple electrodes can be used for detection of commonly
encountered moisture gradients, near surface regain and steep drying. This is promising for
applications where the detection and quantification of moisture gradients is necessary.
REFERENCES
[1] James W, 1996, The interaction of electrode design and moisture gradients in dielectric measurements on wood,
Wood and Fiber Science, 18(2),264 - 275
[2] Makay J, 1976, Effects of moisture gradients on the accuracy of a Power-Loss moisture meter, FPJ, 26(3), 49-53
[3] Jazayeri S and Ahmet K, Moisture Gradient Studies in Timber by Measurement of Dielectric Parameters, Third
International Symposium in Moisture And Humidity, National Physical Laboratory, ENGLAND, 6 - 8 April 1998,
editor: Stephanie Bell, 179 - 186
Contact points: Sina Jazayeri, Eurotherm Controls Ltd, Faraday Close, Worthing, BN13 3PL, UK
Tel: +44 (1903) 268500, Email: sina.jazayeri@controls.eurotherm.co.uk
Dr. Kemal Ahmet, The University of Luton, Park Square, Luton, Bedfordshire, LUI 3JU, UK
Tel: +44 (1582) 489370, Email: kemal.ahmet@luton.ac.uk
152
MICROWAVE CONVERTER FOR SIMULTANEOUS MEASUREMENT OF THE

REFLECTION AND TRANSMISSION COEFFICIENTS OF A MATERIAL
Jerzy Kali:6.ski
WILTECH Industrial Metrology, Warszawa, Poland.
ABSTRACT. A simple microwave converter was developed for moisture content

measurement in dielectric materials. Correlation between the microwave attenuation and
the moisture content has been utilized for simultaneous measurement of the reflection and
transmission coefficients of a materiaL
Keywords: moisture content, microwaves, attenuation, converswn
1. INTRODUCTION
Microwave radiation is commonly used for moisture content measurement and control of
various dielectric materials in industrial as well as laboratory situations. The simplest
design of a microwave system converting the information on moisture content in a
material into an electrical signal (called an M to e converter) is based on the dependence
of the microwave attenuation of the material upon its water content. Depending on the
layer thickness of the material under test and its moisture content, the reflection or
transmission coefficient of a microwave signal should be determined (both being ratios of
two microwave voltages or powers). However, each method is useful in certain, limited
ranges of moisture content.
A significant expansion of the moisture content measuring range and improvement of
measuring accuracy can be accomplished by a simultaneous determination of both
microwave signals: reflected from and transmitted through the layer of material under
test. The moisture content of the material is then computed from both these values. To
enable such measurements a microwave converter, Wiltech MIC, has been developed [1].
2. MICROWAVE INTEGRATED CIRCUIT

Before applying a microwave converter in an industrial measuring system, the following
parameters should be considered:
operational reliability;
- sensitivity to environmental disturbances (temperature, vibrations, dust, etc.);
- size and dimensions as compared with the whole measuring system;
- cost of the equipment, installation and maintenance.
Performance of a converter built from discrete microwave (waveguide or coaxial) devices
could not be satisfactory with respect to any of these factors. Thus, attention has been
directed to a stripline integrated design that is much more efficient and provides a light,
compact and reliable converter. The block diagram of the Wiltech MIC converter is shown
in Fig. 1.
153
8-12V
Er
STAB
1------1
PIN
Fig. 1. Block diagram of the MIC Converter.

The MIC converter consists of the following components:
Gas FE oscillator operating at 9.35 GHz with an output power of 30 m W,
Schottky barrier diode detector, p-i-n diode switch and
a ferrite isolator/circulator, all accommodated on a microstrip line.
The MIC converter provides, alternatively, reference and measuring output signals.
During the reference period, the output signal, E~, depends on instantaneous values of the
microwave oscillator output power and the detector sensitivity. During the measuring
period, the output signal, E~, depends, with the given value of E~, on the microwave signal
reflected from the surroundings. A pin-diode switch provides the period switching.
3. REFLECTION AND TRANSMISSION MEASUREMENT
When the MIC converter is connected with an impedance tuner and a radiating antenna,
it can be directly used for measurement of the reflection coefficient from the surface (and
layer near to the surface) of the material under test. In that case, the ratio of the output
signal E~ and the reference signal E~ is the measure of the reflection coefficient of the layer
of the material.
When a second antenna is located on the other side of the material layer, axially aligned
with the first antenna used for the reflection measurement, as shown in Fig. 2, a simple
system for the transmission coefficient is created. The antenna is connected to a detector
head (X113), thus in a reference period the output signal, E t, is proportional to the
amplitude of the microwave signal passing through the pin-diode switch in "off" position
and the layer of wet material. In normal conditions, this signal should be zero or very
close to zero. During the measuring period, the detector output E'( depends on the
attenuation of the material layer, and the ratio E'( / E t is the transmission coefficient of the
material layer .
154
8-12V
MIC
....
t---E r
011.
:JL
1(z% H 2 0 P/lllilfU/I/lllllWll
A
A
PIN
~--+-"i
X113
ne ative
Fig. 2. Block diagram of the converter

useful in moisture content determination in liquid and solid materials (granular or plate
formed) with constant layer thickness. The system is based exclusively on measurement of
the amplitudes of microwave signals and should provide a range of moisture content
measurement and measurement accuracy comparable with two-parameter measuring
systems [2].
REFERENCES
[1] Kalinski, J. "Microwave converter for simultaneous measurement of the reflection and transmission
coefficients of the tested objects", Polish Patent Application, 1999 (in process).
[2] Kraszewski, A. and Kulinski, S. "An improved microwave method of moisture content measurement
and control", IEEE Trans. IECI, vol. IECI-23, pp. 364-370, Nov. 1976.
Contact point: Dr. Jerzy Kalinski, President, WILTECH Industrial Metrology,

Niemcewicza 24/7, 02-306 Warszawa, Poland. Phone: 0-601-365-277.
155
APPLICATION OF OPEN COAXIAL RESONATOR AS SENSOR

FOR MICROWAVE MOISTURE METER
Boris Tsentsiper
B&G Microwave, St. Jacob De-Haas 423/11,97225, Jerusalem, Israel
ABSTRACT: This paper deals with the application of the resonator method for microwave
moisture metering in weakly-hydrated friable materials. The expressions which relate the
permittivity ofwet materials with the Q-factor and frequency shift ofthe resonator are presented
The test results of the microwave moisture meter equipped with the resonator sensor and
operating at the frequency of2.5 GHz are described
Keywords: moisture, quality, frequency, permittivity
1. INTRODUCTION
The use of traditional microwave absorption and phase shift methods for monitoring the
moisture content in weakly-hydrated friable materials comes up against certain difficulties. In
order to achieve the desired measuring accuracy it is necessary to expose a great bulk of the
tested material to microwave radiation and therefor considerably raise the operating frequency of
the moisture meter. This causes additional technical difficulties in the use of microwave
moisture meters and increases their cost a good deal. Also, in some cases it is technologically
impossible to place the sensors (the transmitting and receiving antennas) on two sides of the
material as the above said methods require.
Using an open coaxial resonator as sensor for a microwave moisture meter offers an alternative
technical solution [1]. On the one hand, such a sensor has all the advantages of other resonator
sensors: high sensitivity, the possibility of using a two-parameter method of analysis which is
invariant to bulk density, and operation in the low-frequency part of the microwave range. On
the other hand, it allows the employment of methods which require the placement of the sensor
on only one side of the tested material.
The aim of the present paper is to present some results of the theoretical and experimental
research on the use of an open quarter-wave resonator as a moisture meter sensor.
2. THEORETICAL BASIS
Figures la and 1b show a schematic drawing of the coaxial quarter-wave resonator and its circuit
design. Cm is the inside fringe capacitance whose value does not depend on the dielectric
permittivity of the tested material; Cfl is the outer fringe capacitance interacting with the wet
material; Grad is the conductivity produced by electromagnetic radiation from the open end of the
resonator; Gd is conductivity determined by dielectric losses in the tested material.
156
The two main parameters that defme the technical characteristics of a resonator are its
resonance frequency and quality factor. We will start the analysis with the resonance frequency
relation.
L
Grad
Figure 1
A schematic drawing of the coaxial quarter-wave resonator (a) and its circuit design (b).
The resonance frequency
OJo
is determined by the following transcendental equation
(1)
where CT is the resonator total fringing capacitance, v is the light velocity, L is the resonator
length, Yc is the resonator characteristic admittance.
The value of CT depends on the dielectric permittivity of the tested wet material. Inasmuch as
only materials with low moisture content are investigated, the value of CT is small and the
resonator length L is close to Ao/4, where Ao is the resonance wave length in air. In this case,
Cot(OJoL/v) can be expanded into a power series about nl2 accurate to within the second order in
(tz/2- aJoL/v), and the following expression for aJo is obtained:
01
1'Z".fo
= ----"'-2( CT + .foL / v)
(2)
Let us the dielectric permittivity of the tested material is &~ = &'u - j&~, the dielectric
permittivity of the protective layer is &1 and its thickness is t. The resonator will interact with the
157
wet material as it would with a medium having an effective dielectric pennittivity
&:.u., which is a
function of &],f and &~. Then (2) can be presented as:

(3)
The relation betweens:.u. and s], f, &~ in its general fonn is a complex expression. For practical
purposes, however the following empirical fonnula can be used [2]
(4)
where T is a parameter which has dimensions of length and depends on the coaxial size.
We can consider (4) as a variant of the dielectric mixing fonnulae. If we assume that the
volumetric concentration of the wet material is ql = exp(-
fIT)
we arrive at a more general
fonnula:
(5)
where the empirical exponent can take on different values depending on the model used. We
define moo as the sensor resonance frequency in the absence of wet material and mOl as the
resonance frequency when in contact with wet material. By combining (3) and (5) we get the
expression for the resonance frequency shift l::..mo = moo - mOl
(6)
Thus, in a first approximation l::..moo is proportional to (irM - 1). When using (7) we must
know the numerical values of the parameters it contains. The value of the fringe capacitance Cfl
can be obtained from the literature. The unknown parameters f and r can be found from the
results of experiments with materials of known dielectric permittivity.
To calculate the resonator quality factor Q we will take the general fonnula that is written as
OBI
mo - x Q=
2G 8m /V=/Vo
(7)
where B and G are the shunt susceptance and conductance of the resonator, respectively.
Taking into account (1), (2) and (7), we get Qo for the resonator unloaded resonator quality factor
158
(8)
The value of Grad for can be calculated when 21ib/ILo <0.5. It was shown that in the far region of
the medium with relative pennittivity 8 eff the magnetic field Hlp is related to the tangential
electric field in the aperture ErCO,rl) [3] of the fonn
(9)
where k m
= (lVo/v )~8eff
,R is the distance from the radiation point to the observation point, rl is
the distance from the center of the resonator radiating aperture (origin of coordinates) to an
arbitrary point in the aperture.
Adopting that the electric field Er(O,rl) on the aperture is the same function as inside the
resonator and expanding the exponent in (9) into a power series and taking into account only the
first two expansion tenns, we get after integration we get
(10)
Using (10) we can get total power PEemitted by the resonator and Grad
(11)
Relationship (11) shows that the resonator quality depends to a great extent on the real part of the
complex dielectric pennittivity of the tested medium. It should be noted that in the case of a very
small radiating aperture (a-.>b), Grad does not depend on the aperture area and is only detennined
by its radius
(12)
The value of Gd can be expressed by

(13)
For wet materials the relationship tgo < 0.5 always holds true in the microwave range. Therefore
(13) can be transfonned into
159
(14)
Thus, measuring quality factor Q and frequency shift of the sensor we can compute e' and e' of
the test wet material with the help of expressions (9), (11), and (14).
Knowing e' and e' we can calculate the A-parameter that is defined as [4]
A(m) =
e"/(e' -1)
(15)
The A-parameter is independent of the material density and is a function of only its moisture
content. Algorithm (15) was realized in the microwave moisture meter operating at a frequency
of 2.5 GHz. The quarter wave coaxial resonator is used as sensor for this moisture meter has the
following main parameters: length L = 27 mm, inside diameter of the coaxial cavity is 26 mm,
outside diameter of the radiating aperture b = 14 mm, inside conductor diameter a = 6 mm and a
flange diameter is 55mm. In the process of the experiments to determine the parameters r, Cfl,
and Tthe thickness of the protective layer varied from 0.8 mm to 3.15 mm and its e1 from 2.3 to
10. It has been established that the parameters y, Cfl and T are 1.5, 7.4 mm and 0.26 pF
respectively. In the fmal version, the protection layer had a thickness off 0.8 mm and e1 =2.5.
The cylindrical sample holder filled with the material to be tested was placed on the resonator's
flange. The walls of the sample holder are made of stainless steel and its bottom of the same
material as the protective layer also with a thickness of 0.8 . Thus the total thickness of the
protective layer is 1.6 mm. To prevent the undesirable microwave irradiation from the resonator
perpendicular to its axis, the sample holder is equipped with a metal flange. In order to improve
the contact between flanges, the sample holder flange has a quarter wave choke groove. The test
results of the microwave moisture meter equipped with the sensor described above showed the
following: if the moisture content is less than 4 percent the standard deviation does not exceed
0.07 percent, irrespective of the bulk density for such materials as quartz, sand, sand-clay casting
mixture, cement, white sugar.
REFERENCES
[1] King R.J., Basuel J.C, Werner M.J.,King K.V., Material Characterization Using Microwave Open Reflection
Resonator Sensors. Microwave Aquametry edited by A. Kraszewski, New York, IEEE Press, 1996, 153-175
[2] Kjetil F., Tjomsland T., Pennittivity measurement of thin liquid layers using open-ended coaxial
probes.Meas.Sci.Tecnol., 1996, Vol. 7,1164-1173.
[3] Blackham D.V, Pollard R.D., An Improved Technique for Permittivity Measurements Using a Coaxial Probe.
IEEE Trans. Instrumentation and Measurement, 1997, Vol.. 46, No.5, pp.l093-1099.
[4] Meyer W., Schilz W. A microwave method of dencity independent determination of moisture content in solids.
J.Phys D, 1980,13, pp.874-876.
Contact point: Boris Tsentsiper, B&G Microwave, StJacob De-Haas 423/11, 97225, Jerusalem, Israel. Phone: +2
(972) 5714804; Fax: +2 (972) 5714804; E-mail: bang7888@netvision.net.il
160
DEVELOPMENT OF TIME DOMAIN REFLECTOMETRY PROBES FOR

COMBINED USE WITH A CONE PENETROMETER
Carlos M P. Vaz(l,2); Paulo S. P. Herrmann(l); Jan W. Hopmans(2)
(1) EMBRAPA Instrumentation Center, Brazilian Enterprise for Agricultural Research, Sao
Carlos-SP, 13560-970, Brazil (2) Hydrology, Dept. of Land Air and Water Resources, University
of California, Davis, CA, 95616, USA. cmvaz@ucdavis.edu
ABSTRACT.
Measurements of soil water content using TDR is widely used to monitor water content in the
soil profile for irrigation, crop management and hydrologic studies. The objective of this work
was to design and test short TDR probes (5 cm long) for measuring in-situ soil water content
when combined with a cone penetrometer. The basic configuration of the short TDR probe
design consists of2 parallel brass rods guided along the outside of a PVC rod and installed above
the penetrometer cone. Samples of 3 different soils with known water content values were used
to obtain calibration curves relating the water content and dielectric constant. Results show that
although loosing sensitivity compared with conventional probes, the penetrometer probe can be
used to perform soil water content measurements combined with the cone penetrometer, with
relatively good sensitivity for laboratory and field applications.
Keywords: TDR, soil water content, cone penetrometer

1. INTRODUCTION
Soil is a three phase system consisting of solid material (minerals and organic matter) and the
voids that can be either water or gas filled. Determination of the soil water content (volume of
water divided by the total volume) is an important parameter required in many agronomic studies
like irrigation and crop management.
Measurement of soil water content (9) with time domain reflectometry (TDR) takes advantage of
the high dielectric constant (e) of water (ew=80) compared with air (ea=l) or dry soil (e=3 to 7)
[1]. Applications of TDR in soil science for the determination of soil water content were first
introduced by Clark Topp and collaborators in 1980 [2], and it is now a widely used technique.
Its main advantages compared with other techniques such as neutron moderation, electrical
resistance blocks, tensiometers, heat dissipation and gravimetric methods are: (1) possibility of
performing both laboratory and field measurements; (2) the low influence of dissolved solute; (3)
the use of non-ionizing electromagnetic radiation; (4) the low influence of bulk density
variations; (5) potential of automatic data acquisition and multiplexing and (6) the use of
different probe sizes and geometry. The objective of this study was to design and test short TDR
probes for measuring soil water content in the laboratory and field when combined with a cone
penetrometer.
Soil cone penetrometers measure the soil penetrability that is a measure of the ease with which
an object can be pushed or driven into the soil profile [3]. The reading is called penetration
161
resistance (PR) and depends on mainly soil properties such as bulk density, water content and
potential, texture, aggregation and mineralogy. Hence, cone penetrometers have been used
extensively in soil science studies to asses the effect of different tillage practices [4], to identify
natural and induced compacted layers [5] and to predict related soil properties [6]. However, soil
water content changes spatially and temporally in field soils, limiting the application of cone
penetrometers as an indicator for compaction or tillage practices. An instrument that measures
simultaneously the penetration resistance and the water content at the same point will improve
and bring new applications to penetrometers in soil science and civil engineering.
Most commercial TDR-equipment use standard waveguides (probes) provided by the
manufacturer, generally longer than 15 cm. The soil water content value is obtained from
calibrations using the travel time of an EM wave along the waveguide to determine the bulk
dielectric constant of the soil. When using commercial cable testers the lower limit of probe
length is around 10 cm if an accuracy better than 0.02 cm3/cm 3 is required [7]. Few studies have
used TDR probes shorter than 10 cm. Petersen et al. (1995) [8] examined the importance of
probe-rod length and diameter, distance between rods and horizontal installation depth. They
obtained excellent traces using a 5 cm probe length in a coarse sandy soil with a water content of
3
3
3
3
0.21 cm /cm . Kelly et al. (1995) [9] obtained accuracy of 0.035 cm /cm using TDR probes as
short as 2.5 cm and a high-bandwidth TDR system (20 MHz). Sri Ranjan and Domytrak (1997)
[10] described the successful use ofTDR miniprobes (5 cm long) for a clay loam soil. Selker et
al. (1993) [11] introduced an serpentine type surface probe (10 cm x 10 cm) that allows
miniaturization of TDR probes with great spatial resolution. Nissen et al. (1998) [7] using a
similar approach developed a coil type probe allowing a reduction in probe length of a factor of
five without loss in the probe sensitivity.
2. MATERIALS AND :METHODS
The basic configuration of the short TDR probe designed here consists of 2 parallel brass rods (2
mm diameter) guided along the outside of a PVC rod and soldered at its upper end with two
poles of a 50 .Q coaxial cable. The probe was installed just above the penetrometer cone at the
end of the penetrometer probing rod. The coaxial cable is guided inside the probing rod and
connected at its end to a cable tester (Tektronix 1502C). Figure 1 shows the TDR probe
combined with the cone penetrometer and the system being used for the laboratory tests. The
wave form (trace) is visualized on the cable tester display. The data are transferred to a PC
microcomputer through the serial RS232 interface. Cone and probing rod sizes are in accordance
to the American Society of Agricultural Engineers Standards [12]. As the dielectric constant
measured by the penetrometer probe is an average dielectric constant of the soil and the PVCsteel between the TDR rods, a conventional 2 parallel rods TDR probe (length = 5 cm) was also
used to estimate the true bulk soil dielectric constant of the soil samples.
Three different materials were used to test the probe. A yolo silt clay soil, a columbia sandy loam
mixed with sand and a washed sand (SRI supreme sand-30). Samples with different water
content were obtained adding known amount of water to the dry soil samples, which after
through mixing was packed in cores (diameter = 8.25 cm; height = 8 cm). Subsequently, samples
stead for 2 days to reach equilibrium. Number of samples prepared were 16, 13 and 12 for yolo,
columbia and sand, respectively, in the range of water content between air dry and saturation.
162
Figure 1. Probe combined with the cone penetrometer (a) and system used in laboratory (b).
TDR traces were obtained at three different locations in each soil sample with both the
penetrometer probe (P) and the conventional probe (C). After TDR readings, soil samples were
weighed, oven dried (100 degrees centigrade for 24 hours) and weighed again for volumetric
water content and bulk density determinations. The average of the 3 replicates TDR
measurements was plotted versus water content values for calibration. Dielectric constant values
were determined from the trace by the following equation: e = 1.0 1(La / Li [1], where La (cm) is
the apparent length of the probe determined from the trace and L (cm) is the probe length (5 cm).
RESULTS
Typical traces of the yolo soil obtained with the (P) and the (C) probes are presented in Figure 2.
Traces related to 2 different soil water content are shown. Apparent length (La) is determined as
shown, identifying the first reflection at the beginning and the second at the end of the TDR rods.
140.-----------------~~~----~~==3
6 = 29.2%
--6=15.6%
120
Q)
()
c:
1\1
"0
80
Q)
c..
Q)
60
.~
15
(jj
....
. "' ............ .
~
.. '
100
-->,,,
,
,:,
. . .. ':1........
I:
~
c~
.. ,
!==========!~~L ' ....... (:
40
,
:
L _________________________ ~
20
'
L.
0
0
10
15
20
25
30
35
40
45
50
distance (cm)
Figure 2. Traces obtained for yolo soil samples of different water contents in both penetrometer
(P) and conventional probes (C).
163
Figure 3 presents the dielectric constant as a function of the volumetric water content. Soil
dielectric constant values were similar in the low water content range, but differences between
soils increased as water content increased. This behavior is a consequence of the presence of
more bound water in the yolo soil than the sand and columbia soil [13]. Bulk density of the soil
samples ranged from 1.34 to 1.65 g/cm3 for sand, 1.02 to 1.49 g/cm3 for columbia and 1.00 to
1.26 g/cm3 for yolo soil.
20
35
penetrometer probe
18
conventional probe
30
16
14
lIE
columbia-sand
sand
yolo
25
c:
....co
c:
10
0
0
.;:: 15
ti
Q)
Qi
Q)
'6
colum bia-sand
sand
yolo
0
.;::
+'
'*
ti 20
c:
0
co
12
I/)
....c:
+'
'6
10
4
5
2
0
0
0
10
20
30
40
50
e (%)
10
20
30
40
e (%)
Figure 3. Variation of dielectric constant with the water content (8) for three different soils on
the penetrometer and conventional probe.
All experimental data obtained for the three soils combined are presented in Figure 4 and
compared with the experimental equation introuced by Topp et al. (1980) [2]. Soil dielectric
constants obtained with the 5 cm conventional probe are quite similar to those obtained by Topp
equation, but higher than those obtained by the penetrometer probe. This is caused by the larger
effective volume of soil interacting with the microwaves in the conventional probe compared
with the penetrometer one. Consequently, penetrometer probe presents a lower sensitivity for soil
water content determinations compared with the conventional probe. However, the calibration
curves obtained here for the penetrometer probe allow us to conclude that the designed probe can
be used for soil water content determinations combined with a cone penetrometer. Further
studies should include a field evaluation of the probe and tests with other probes geometry.
164
40~------------------------~
35
o penetrometer probe
conventional probe
----- Topp equation
,
,,
1: 25
.. ."
,
,,
30
./
/
I
,.'
~
t::
20
()
~ 15
~
"C
<cO
10
10
20
30
40
50
e (%)
Figure 4. Comparison among results obtained with the penetrometer and conventional 5 cm long
probes for the three soils and Topp equation [2].
ACKNOWLEDGEMENT. Funding support received (98/04740-7) by C.M.P. Vaz from
F APESP (Sao Paulo State Funding Agency, Brazil) is acknowledged. The authors wish to thank
technical assistance of David Page on the penetrometer probe construction.
REFERENCES.
[1] Cassel, D.R.~ Kachanoski, R.G.~ Topp, G.C., Practical considerations for using a TOR cable tester. Soil Techno!.,
1994, 7, 113-126.
[2] Topp, G.C.~ Davis, J.L.~ Annan, AP., Electromagnetic detennination of soil water content: measurements in
coaxial transmission lines. Water Resour. Res., 1980, 16, 574-582.
[3] Bradford, J. L., Penetrability. m: Methods of Soil Analysis, Madison, ASA-SSSA mc. Publisher, 1986,463-477.
[4] Stelluti, M.~ Maiorana, M.; Giorgio, D., Multivariate approach to evaluate the penetrometer resistance in
different tillage systems. Soil Till. Res., 1998,46, 145-151.
[5] Henderson, C.W.L., Using a penetrometer to predict the effects of soil compaction on the growth and yeld of
wheat on uniform, sandy soils. Aust J. Agric. Res., 1989, 40 (3), 498-508.
[6] Ayers, P.D.~ Bowen, H.D., Predicting soil density using cone penetration resistance and moisture profile. ASAE
Trans., 1987,30, 1331-1336.
[7] Nissen, H.H.; Moldrup, P.~ Henriksen, K, High-resolution time domain reflectometry coil probe for measuring
soil water content. Soil Sci. Soc. Am. 1., 1998,62, 1203-121l.
[8] Petersen, L.W.; Tomsen, P.; Moldrup, P.~ Jacobsen, O.H.; Rolston, D.E. High resolution time domain
reflectometry: sensitivity dependency on probe design. Soil Sci. 1995, 159, 149-154.
[9] Kelly, S.F.~ Selker, 1.S.~ Green, 1.L. Using short soil moisture probes with high-bandwidth time domain
reflectometry instruments. Soil Sci. Soc. Am. 1., 1995,59,97-102.
[10] Sty Ranjan, R~ Domytrak, C.J. Effective volume measured by TOR miniprobes. Transactions of ASAE, 1997,
40 (4), 1059-1066.
[11] Selker, J.S.~ Graff, L.~ Steenhuis, T.,Noninvasive time domain reflectometry moisture measurement probe. Soil
Sci. Soc. Am. J., 1993,57,934-936.
[12] ASAE Standards engineering practices data (Soil cone penetrometer, S313.2), St Joseph, ASAE, 1994,687.
[13] Dasberg, S. Hopmans, J.W. Time domain reflectometry calibration for uniformly wetted sandy and clayey soils.
Soil Sci. Soc. Am. J., 49, 293-297.
165
Session on
MOISTURE MEASUREMENT III
Chair: Prof. Stan Stuchly,

Dept. Electrical & Computer Engineering.
University of Victoria, Victoria, B.c. Canada
Tuesday, April 13, 1999.

8:30 - 10:00 AM
166
167
ACTIVE MICROWAVE RADIOMETRY IN DETERMINING SAND MOISTURE

George POLIVKA, SPACEK Labs., Inc., 212 E.Gutierrez St.,
Santa Barbara, 93101 CA, U.S.A.
Polivka 0 webtv.net
ABSTRACT
There are numerous methods to determine sand moisture
mainly by microwave absorption but almost all methods
require laboratory work and almost none is known to operate
reliably in construction site environment.
The application of the Active Radiometry brings a novelty
discovered first in experiments with noise fields:
as
microwave noise is composed of wideband random fluctuations,
all spatial and temporal interference is random and the
instrument response is smoothed by an integrator in the
detecting radiometer. The resulting effect suppresses all
local interference we normally observe in the near field:
this means that object loss can be measured when the object
is moving and the measuring system is also insensitive to
object shape.
1. INTRODUCTION
The presented method of Active Radiometry consists in
using a microwave noise source emitting noise toward another
radiator connected to a microwave radiometer. The noise
radiator utilizes an avalanche diode as noise generator : as
a radiator, one can use a standard waveguide horn or another
suitable antenna that can concentrate the generated noise
field to a suitable test spot.
The radiometer is a sensitive microwave receiver capable
to evaluate the low power levels of noise generated in the
noise radiator and attenuated by the tested object. Fig.l
shows the diagram of the test system and Fig.2 gives an
estimation of a determinable attenuation of the object
located between the radiators. With modern noise sources and
moderately sensitive radiometers now available, the loss
limit of
tested objects is
around 50 dB
including
propagation loss.
168
2. LOSS DUE TO SAND MOISTURE

Water is a good absorber at microwaves. Depending on
frequency, we expected loss up to 10 dB/cm for wet sand at
12 GHz. Certain variations could also be expected due to
impurities in water like salts that
increase conductivity
and loss by up to 100 times, also depending upon frequency.
3.COMPARISON OF CW AND NOISE METHODS
While with a CW sounding signal, the main problem is always
the intense interference in the near-field region. With
noise there is no interference and the tested object can
even be moving without any effect upon the measured loss.
4. EXPERIMENTS
With the setup shown in Fig.l, a practical loss limit was
40 dB approx. With 1 kHz modulation, this limit was
extended up to more than 50 dB. Sand samples with a known
moisture were evaluated welded in polyetylene bags. When CW
signal was used, sample position had to be kept within one
millimeter - with noise, no problem was observed when sample
moved by more than several centimeters. Sand density was
important: it was kept constant by light shaking the bags
with sand samples. The results are shown in Fig. 4.
5. INDUSTRIAL SENSOR DEVELOPMENT
For industrial use, the following principles were applied:
A. Microwave noise is better than a CW signal. Noise caused
no interference in "near zone" and is safe for humans as
instead of 10 -100 mW, only picowatts of power are radiated.
The low-noise receiver is free from interference as only
weak satellite signals are allowed in 10.95-12.8 GHz band.
Fig.3 shows the block diagram of our system, and Fig.5 the
noise source and receiver with horn antennas for field use.
B. Measured loss is proportional to sample thickness. From
Fig.4 it can be seen that a maximum loss can be expected of
10 dB/cm approx. Sand layer forming remains to be tested
extensively in order to keep measurement error small.
C. Salt contamination remains the major source of error.
Thus, when sand source is changed, it is recommended to
test run samples checked by standard laboratory test.
169
6. CONCLUSION.
The Active Radiometry has demonstrated its advantage in a
demanding industrial application. Sample structure, form and
position may vary
without substantially affecting the
reading. All possible arrangements were tested including
polarization and reflectivity: the normal transmittance has
produced the best reproducible results.
The author acknowledges Mr.Zeman of CSE Company for his
challenge to develop the described sensor. Mr. Burda and
Zatocil of Autocont Co. are acknowledged for their effort in
developing the industrial version of the described sensor.
The above work was sponsored by POL-CONSULT of Prague, Czech
Republic.
REFERENCES:
1. Polivka, J.: "Active"
And Applications, paper
Oct.98
2.
Kraszewski, A.:
80, p.48-58
Microwave Radiometry - Experiments

presented to EuMC 98, Amsterdam,
Microwave Aquametry,1980,
Proc.of EuMC
3. Polivka, J.: Measuring Concrete Permittivity at MM Waves,

1996, Internat. Journal of IR and MM Waves, 17, Oct.1996,
No.IO,p. 1672-1683
4. Polivka, J.: Czechoslovak Patent No. AO 260 205, 1987,
Contactless System to Determine Loss and Reflectivity of
Objects by Microwave Noise
Fig.l
Basic Active Radiometry System to determine object loss

NG - Noise Generator, RM - Radiometer
170
ENR,dB
30
~------T-------~------~------~------~
AT=10K
20
10
o
o
Fig.2
10
20
40
30
50 dB, loss
Estimation of Measurable Loss in a System of Fig.l

ENR: Excess Noise Ratio of Noise Generator,
e.g. 30 dB for avalanche noise diodes,
15 dB approx. for noise discharge tubes_M
L{T : Radiometer resolution,
AT = 'rs (B t ) 2.
IF amp.
det.
audio amp.1 kHz
indicator
RS-232
Fig.3
System to Measure Moving Object Loss by Noise

NG - Noise Generator; 1 kHz Modulated
LNC - Low-Noise Converter, in our case a 11 GHz LNC
IF - IF Amplifier and Detector
MVM- Selective 1 kHz Audio Level Meter
AID - A-D Converter
1 kHz- Square-Wave Generator, 1 kHz
171
Loss
dB/em
8
/
'/
~
6
Y'
~
~
NOISE
9'
CW SIGNAL
o
o
Fig.
4 Measured
Fig.S
10
12
\vater
14% Content
Loss of Sand Layer at 2.5 cm Wavelength
Noise Generator and Noise Receiver for Sand Moisture

Determination - Industrial Sensor Version
172
RESONATOR-BASED MICROWAVE MOISTURE METER WITH

OPTIMIZED DIGITAL SIGNAL PROCESSING
F. Daschne/, R. Knochel1, M. Ken?
1 Christian-Albrechts-University, Kiel, Germany

2 K&S Associates, Biggar, Scotland
Abstract.
If moisture content is derived from resonance frequency or quality factor of a
resonator detuned by a moist dielectric, problems arise if digital techniques are used. The reason
is, that digital data are only available at discrete frequencies, i.e. a resonant frequency does not
normally coincide with a frequency point. If rapid measurements are required, digital algorithms
for determining moisture content are critical. This paper discusses various digital signal processing algorithms with respect to measurement speed, accuracy, speed of data processing, numerical stability and hardware effort.
Keywords: Moisture measurement, resonator, digital algorithm
1. INTRODUCTION
Microwave moisture meters are often based on resonators, where the fields are loaded by the
moist material. This solution offers high sensitivity. Only scalar measurements are needed for
determining resonant frequency and quality factor of the resonator loop. If volume, shape and
orientation of the moist material are kept constant, the cited quantities are able to deliver independent measurements for real and imaginary parts of the complex dielectric constant E*=E'-jE",
and then permit for example the determination of moisture content 'I' or mass (density because of
known volume) or both. If such instruments are applied in process control, often fast moving
mass streams on conveyor belts or in transportation tubes are involved, which show fluctuating
homogeneity, thus demanding for continuous averaging in order to reach the required accuracy.
High-speed measurement and data
antenna
processing of the resonance curve
is unavoidable. Often measurement
speed
and
accuracy
are
contradictory features, hence a
compromise has to be found.
Most modem moisture meters use
digitally controlled instrumentation
and digital signal processing afterwards. Collecting data for a resonance curve by sweeping frequency delivers amplitude values
at discrete frequencies, typically
spaced by constant frequency increments, and not necessarily including the resonance frequency or
trigger
grip
Fig. 1 Fringe field capacitive sensor
173
the 3 dB-bandwidth. Those have to be determined from the measurements by applying digital
algorithms. This paper compares various measurement procedures and their subsequent digital
signal processing with respect to accuracy, measurement speed, speed of data processing, numerical stability and required hardware effort. Hitherto most frequently used analog evaluation procedures could not be transposed to digital techniques directly because of the discrete sampling of
the resonance curve.
2.1 Microwave moisture meter
The microwave moisture meter used for these measurements is a portable self-contained fringefield capacitive sensor (Fig. 1), based on an open ended quarter wave coaxial line. The microwave generator is a PLL-synthesizer operating between 2,4 and 2,5 GHz, having a settling time
of:;:; 2 ms after switching the frequency. The measurement procedure is controlled by a digital
signal processor (DSP). Data are transferred to a digital computer and evaluated by applying
various digital algorithms.
2.2 Measurements
Measurements were made on moist terracotta particles (typically used for gardening purposes).
Density was virtually constant. Moisture content varied between '11=0 .. .33%, graduated in 13
different moisture values. Data were collected at 50 discrete frequencies between 2,4 and 2,5
GHz, repeating each measurement ten times at different places of the same sample for averaging.
2.3 Evaluation procedures
All investigated algorithms work with the same set of data enabling comparison. Amplitude values near the resonant frequency are used. The following algorithms were tested:
a. Nonlinear fitting to a Lorenzian curve at 9 frequencies.
b. Solving a system of simultaneous linear equations at 9 frequencies.
c. Principle component analysis and multiple linear regression using 9 frequencies.
d. Same as c. but applying nonlinear regression.
e. Measurement of transmitted power at a fixed frequency.
f. Measurement of transmitted power at two fixed frequencies.
Due to the relatively small number of prepared measurement samples, internal cross validation
(ICV) [1] was used for greatly improving data evaluation. Calculations were made with
MATLAB.
3. RESULTS
3.1 Nonlinear fitting to a Lorenzian curve (Method a)
Using a lumped element model of a resonator, the amplitude behaviour of the resonance curve
can be described as a Lorenzian curve. If U is the normalized power transmitted through the
174
2.46
O.
2.46
2.46
2.46
N
2.4
J:
(!)
"1
-~
2.4
0
2.461
2.4~--'---'----'----'---'---'---.I
10
15
20
moisture content
25
35
30
Fig. 2 Resonance frequency vs. moisture content

(Methods a or b)
10
15
20
moisture contenl
25
30
35
Fig 3 lIQ versus moisture content

(Methods a or b)
resonator and f is the nonnalized frequency, the resonance characteristic is given as

a
u= b+ (c-f )2'
(1)
where a, band c are fitting parameters. The resonance frequency and the quality factor are
Ir =c
and
Q=
(2)
2.Jb
For every moisture value, fr and Q were determined from the 9 points in the vicinity of the resonance and calibration curves were generated by linear regression. Moisture content 'II linearly
depends on fr or lIQ, as are shown in fig. 2 and fig. 3. Nonlinear fitting to a Lorenzian curve can
be classified as being very accurate, however, requiring high computational effort, if floating
point operations (Flops) are taken as a measure. Also, care has to be taken when choosing the
frequency increment for collecting the amplitudes: If the resolution is too coarse, and some of
the amplitudes are lying in the noisy region outside the resonance, numerical instability can occur.
3.2 Solving a system of simultaneous linear equations for fr and Q (Method b)
Numerical effort is reduced by calculating the inverse of the Lorenzian characteristic, i.e.
y=J...= b+(c-f)2 =kof2+klf+k2'
(3)
Taking measurements at m=9 frequencies into account the parabola is overdetermined and the
system of simultaneous linear equations of eq. 4 results:
m
1 2-1/ 2-1/
2i=m1
i=m1 i i=m1
2-1/ 2-1/
2-1/
i=m1
i=m1
m
2-1i4
2-1/
2-1/
;=1
;=1
i=1
j=l
I_Ili
i=1 U
I_IU; 1/
I_IU; 1/ )
o
{t]=
175
;=1
;=1
(4)
The equation is solved by applying the Cholesky-approach[2], because the matrix is symmetric
and positive definite. The algorithm is optimized by using the symmetry and by minimizing
floating point operations, delivering leo, k}, and k2 or
k
k
fr =- 2~ andQ=
2'
(5)
o
4 k 02
k -~
RI
4
Moisture content is again expressed as a linear function of fr or lIQ. Compared with method a,
fitted data have approximately the same quality, but the computational effort is less by the factor
100. The applied measurement values have to be near the resonance, because by forming the
inverse value noise is amplified, and numerical instability can occur.
3.3 Principle component analysis and multiple linear regression (Method c)
Again starting from 9 measurement points at each grade of moisture, the principle components
(PCs) of the data sets are calculated [1]. The various PCs deliver orthogonal statistical information about the measurements. They can not directly be related to resonance frequency and quality
factor, however, they reflect the dependence of the resonance characteristics on moisture content.
In the present case, moisture content is constructed as a function of the first two PCs by linear
regression. This yields the best results, requiring low computational effort and being numerically
stable, once the calibration is carried out.
3.4. Principle component analysis and nonlinear regression (Method d)
After determining the PCs, the moisture content 'If is constructed as a third order polynomial
using PCI only and nonlinear regression (Fig. 4). PCI contains the largest variance. Compared
to method c, only a minor decrease in computational effort results.
35
30
c:
c:., 2
c:
25
J!l
c
0
.,2
:s
20
'"
u;
]i
'0
E 15
E 1
10
-2.5
-2
-1:5
-1
-<l.5
PC1
0.5
1.5'
Fig. 4 Moisture content as third order polynomial

of principle component 1 (Method d)
176
50
100
150
200
250
300
350
measured value at 2.4755GHz
400
Fig. 5 Moisture content as third order polynomial

of the measurement value at 2.48 GHz (Method e)
Procedure
a
b
c
d
e
f
SEP
1,25
1,28
1,50
1,67
1,13
1,40
SEC
1,02
1,02
0,97
0,87
0,87
1,37
Flops
34844
366
69
62
8
36
No. of Meas.
9. -- -9
9
9
1
2
Meas. Time [ms]

18
18
18
18
ADC sampling-time
4
Hardware-effort
high
high
high
high
low
medium
Table 1 Comparison of the results
3.5 Measurement of transmitted power at one or two fixed frequencies (Methods e and f)
For method e, an amplitude at a fixed frequency slightly above the resonance is chosen, which is
influenced by both the shift of fr to lower frequencies and the decrease of the Q-factor when
loaded with a moist dielectric. A measurement set-up operating this way is very simple, although
it is only suitable if density changes in the sample are negligible. Hence the approach has only
limited applicability and only serves here for comparison purposes. The change of amplitude
with moisture content is fitted by a third order polynomial (Fig. 5). Uniqueness of the measurements and sufficient signal amplitudes are also required. This limits the applicable range of
moisture content. Computational effort is low and rapid measurements are possible.
The idea of method f is to improve the quality of the results of method e by an additional measurement frequency below resonance. Even when loaded with a sample having the highest possible moisture content, the resonance frequency must always stay above the second test frequency.
It is found, that the additional effort is not worthwhile, if only moisture content is being determined and density changes do not occur.
4. COMPARISON AND CONCLUSIONS
Table 1 shows a comparison of the results. If standard error of prediction (SEP) is the main criterion, methods a and b achieve approximately the same accuracy. Hence method b has to be preferred because of 100 times less computational effort (Flops). Accepting a slightly deteriorated
accuracy, method c allows a further reduction of Flops by a factor of -5. Measurement time is
dictated by tuning speed (2 ms/point, 18 ms/9 points). Reducing the number of points leads to a
deterioration of accuracy with 3 points- being the lower limit, while increasing the number of
points does not necessarily increase accuracy. Hardware effort can be minimized by using a
fixed frequency oscillator (method e). This also decreases measurement time to ADC-sampling
time. SEP is also lowest for this approach. However, one has to keep in mind, that method e is
only applicable as long as density changes are negligible. If a density-independent measurement
is required, algorithms b and c have to be preferred.
5. REFERENCES
[l]Martens H and Naes T, 'Multivariate Calibration', John Wiley and Sons, (1989), Chichester,
UK.
[2]Golub G and van Loan C, Matrix computations, The Johns Hopkins University Press(1996)
A UNIFIED CALIBRATION METHOD FOR MOISTURE SENSING IN

PARTICULATE MATERIALS
Samir Trabelsi, Andrzej W. Kraszewski, and Stuart 0. Nelson
Richard B. Russell Agricultural Research Center, Agricultural Research Service,

U. S. Department of Agriculture, Athens, GA 30604-5677. E-mail: strabelsi@qaru.ars.usda.gov
ABSTRACT. A unified calibration method for nondestructive determination of moisture content in
granular materials by microwave sensors is presented. Results are shown for wheat and com with a
permittivity-based calibration function that is bulk-density and material independent.
Keywords: complex permittivity, grain, moisture content, calibration.
1. INTRODUCTION
Moisture content of natural or manufactured granular materials is an important parameter often

used in quality control and in defining optimum processing conditions in many industries
including mining, construction, pharmaceutical, food and agricultural. Conventional methods for
moisture determination rely on off-line oven drying of a few samples at a specific temperature
for several hours or on batch type electrical moisture meters calibrated to oven standards. These
methods may be destructive and time consuming. Moreover, they do not meet the needs of
highly automated industrial processes for which moisture content has to be determined in real
time for large quantities of material. In the last two decades several microwave moisture sensors
have been successfully used for nondestructive testing of a wide variety of materials in different
industries [1]. They are based on indirect determination of moisture content from measurement
at microwave frequencies of entities which characterize the wave-material interaction. However,
although water content is expected to dominate the wave-material interaction at these
frequencies, an individual calibration is required for each particular material and instrument [2][6]. This is mainly related to effects of the bulk density, geometry and dimensions of the
particles, and, perhaps to a lesser degree, because of differences in composition. In most
applications this will complicate the calibration procedure and to some extent the design of the
sensor and consequently its cost. For instance, many existing microwave moisture meters for
particulate materials are composed of two subsystems, i. e., a microwave subsystem for moisture
measurement and a second subsystem for bulk-density determination. In this way the bulk
density effect is compensated for empirically in the reading output of the moisture meter. It is
also possible to eliminate the bulk density effect by using density-independent calibration
functions [2]-[8]. However, the particle size and geometry have more subtle effects, and though
they are known to affect the scattering and absorption properties of the material, it is difficult to
quantify their respective effects by measurements or theoretical means. In this paper, a unified
calibration method is presented for moisture content determination in granular materials
independent of bulk density and independent of the material under test. Calibration equations are
given for wheat and com at two particular frequencies at room temperature.
178
2. PERMITTIVITY-BASED FUNCTIONS FOR BULK DENSITY AND MOISTURE

CONTENT DETERMINATION IN PARTICULATE MATERIALS
Particulate moist materials are heterogenous mixtures of components exhibiting different
dielectric behaviors. Therefore, a rigorous theoretical analysis of the wave-material interaction
leading to explicit relations between the effective complex permittivity (E = E' - jE", where E' is
the dielectric constant and E" is the loss factor) and the physical properties of the mixture is
difficult to comprehend. However, because of its intrinsic character, the complex permittivity
remains the most meaningful entity that can be used to extract other physical properties. Thus, at
a given frequency (j), the moisture content (M), bulk density (p), and temperature (T) can be
determined from measurement of the complex permittivity. Another major advantage, if
appropriate correlations between E, or expressions that are a function of E, and M, p, and T are
identified, they can be used regardless of the measurement technique, which gives more
flexibility in the design of the best sensing device for a given application. In this paper, some
relations are identified and given for wheat and com in the frequency range between 11.3 GHz
and 18.0 GHz. The complex permittivities were measured in free space for wheat and com of
different moisture contents, b~lk densities and temperatures [9], [10]. Table 1 summarizes the
ranges of the different variables under consideration,/, p, M, and T.
To accommodate variations of the complex permittivity with all variables in a single diagram, a
complex-plane plot is used with E'IP as the x-axis and E"Ip as the y-axis. Fig. 1 and Fig. 2 show
such a diagram for wheat and com, respectively. At each frequency, data corresponding to
different moisture contents and different temperatures fall along the same straight line of slope
af' This particular behavior reveals that M and T affect the complex permittivity in ways that are
interchangeable. Hence, at a given frequency, the normalized dielectric properties for a given
moisture content at high temperatures are the same as those for a higher moisture content at
lower temperatures. The x-axis intercept k for each material is common to all frequencies and
represents the normalized dielectric properties of the air and dry matter mixture or that
corresponding to the air-kernel mixture at a given moisture content at a very low temperature.
The constant k is characteristic of each material and can be determined either by extrapolation or
experimentally [7], [11]. For com, the value of k was determined as 2.95 by extrapolation and
2.92 by measurement. For wheat, the value of k was determined as 2.74 by extrapolation and
2.76 by measurement. The frequency effect translates into a rotation of the straight line about (k,
0) in the complex plane.
For each material, a linear regression at a given frequency provides the bulk density as follows
[12]:
(1)
where af is a function of the frequency alone [7], [11]. As shown in Eq. (1), neither the moisture
content nor the temperature of the material are needed to determine p. In general, for any
material of known (af' k), measurement of the dielectric properties will provide the bulk density
179
according to Eq. (1) without prior knowledge of the moisture content or the grain temperature.
Eq. (1) provided p with a standard error of performance (SEP) of 7.8 kg/m3 on average for wheat
and 12.9 kg/m3 on average for corn.
For moisture determination, the concept of density-independence has been widely used [8], [13].
Many density-independent calibration functions, fully defined in term of the complex
permittivity, have been reported [3], [7]. Historically, the first one was based on empirical
observations [2], [3] and is expressed analytically as:
E"
'l/Jl =
E -
(2)
Recently, a new density-independent calibration function was introduced. It is based on

empirical observations and the balance of energy distribution between stored energy and
dissipated energy in dielectrics [7]. It is expressed analytically as:
(3)
In the next section variations of 'l/Jl and 'l/J2 with moisture content are shown and calibration
equations are given for wheat and corn at two particular frequencies, 11.3 GHz and 18.0 GHz, at
room temperature.
3. UNIFIED CALIBRATION METHOD
Calibration is probably one of the most critical phases in the development of a sensor. It reflects
the validity of a given technique and the repeatability of the measurements within a given
confidence interval. In this respect, accurate determination of the different constants and
correction or compensation factors involved in the calibration equation is essential. Most
existing moisture meters require an individual calibration equation for each material which
complicates the calibration procedure and to some extent the design of the meter itself. As an
example, Fig. 3 and Fig. 4 show variations of 'l/Jl with moisture content in wheat and corn at 11.3
GHz and 18.0 GHz at room temperature. At each frequency, a linear relationship between 'l/Jl
and moisture content of each material can be identified:
'l/JI,wheat = aI,wheat
+ bI,wheat
(4)
'l/JI,corn = aI,corn
+ b1,corn
(5)
Therefore, two different calibration equations have to be used to predict moisture content from
measurements of 'l/Jl.
Fig. 5 and Fig. 6 show variations of 'l/J2 with moisture content in wheat and corn. Although corn
and wheat have pronounced differences in kernel dimensions and shape, bulk density, and
180
composition, data points for a given frequency are nearly superimposed. Consequently, data can
be fitted by the same equation:
(6)
resulting in a single calibration equation for both materials:
M=
'z - bz
az
This simplifies considerably the calibration procedure and the design of a moisture meter
dedicated to moisture determination independent of density and independent of the material.
Table 2 gives the calibration constants for both calibration functions '1 and 'z. Eq. (7) provided
M with a SEP of 0.25% on average for wheat and 0.45% on average for com.
4. PERSPECTIVE
Standardization of microwave methods for nondestructive moisture determination relies on the
versatility and accuracy of the meters designed for this purpose. In this paper, a unified
calibration method is proposed which eliminates effects of bulk density and structure. This is a
first step toward the design of a universal cost-effective moisture meter for natural and
manufactured particulate materials. Further investigations are needed to verify the behavior of
'z for other particulate materials.
REFERENCES
[1] Nyfors, E. and Vainikainen, P., Industrial Microwave Sensors, Artech House, 1989.
[2] Kraszewski, A. W., Kulinski, S., and Stosio, Z., A preliminary study on microwave monitoring of moisture content
in wheat. J. Microwave Power, 1977, Vol. 12, No.3, pp. 241-252.
[3] Meyer, W. and Schilz, W. M., Feasibility study of density-independent moisture measurement with microwaves.
IEEE Trans. Microwave Theory Techn., 1981, Vol. MTT-29, No.7, pp. 732-739.
[4] Kent, M. and Kress-Rogers, E., Microwave moisture and density measurements in particulate solids. Trans. Inst. M
C, 1986, Vol. 8, pp. 161-168.
[5] Kupfer, K., Possibilities and limitations of density-independent moisture measurement with microwaves. Microwave
Aquametry, 1996, Chap. 21, pp. 313-327, New York, IEEE Press Book Series, Ed. A. W. Kraszewski.
[6] Menke, F. and Knochel, R., New density-independent moisture measurement methods using frequency-swept
microwave transmission. IEEE MTT-S International Microwave Symp. Digest, 1996, Vol. 3, pp. 1415-1418.
[7] Trabelsi, S., Kraszewski, A. W. and Nelson, S. 0., New density-independent calibration function for moisture
sensing in particulate materials. IEEE Trans. Instrum. Meas., 1998, Vol. 47, pp. 613-622.
[8] Trabelsi, S. and Nelson, S. 0., Density-independent functions for on-line microwave moisture meters: a general
discussion. Measurement Science and Technology, 1998, Vol. 9, No.4, pp. 570-578.
[9] Kraszewski, A. W., Trabelsi, S. and Nelson, S. 0., Wheat permittivity measurements in free space. J. Microwave
Power Electromagn. Energy, 1996, Vol. 31, No.3, pp. 135-141.
[10] Trabelsi, S., Kraszewski, A. W. and Nelson, S. 0., Microwave dielectric properties of shelled, yellow-dent
field com. J. Microwave Power Electromagn. Energy, 1997, Vol. 32, pp. 188-194.
[11] Trabelsi, S., Kraszewski, A. W. and Nelson, S. 0., Nondestructive microwave characterization for bulk density
and moisture content determination in shelled com. Meas. Sci. Technol., 1998., Vol. 9, pp. 1548-1556.
[12] Trabelsi, S., Kraszewski, A. W. and Nelson, S. 0., Simultaneous determination of density and water content of
particulate materials by microwave sensors. Electronics Letters, 1997, Vol. 33, pp. 874-876.
181
[13] Kraszewski, A. W., Trabelsi, S. and Nelson, S. 0., comparison of density-independent expressions for moisture
content determination in wheat at microwave frequencies. J. agric. Engng Res., 1998, Vol. 71, pp. 227-237.
Contact point: Samir Trabelsi, Richard B. RusselI Agricultural Research Center, Agricultural Research Service, U.
S. Department of Agriculture, Athens, GA 30604-5677. U.S.A. Phone: 1 (706) 546 3527; Fax: I (706) 546 3607;
E-mail: strabelsi@qaru.ars.usda.gov
Table 1. Ranges of frequency, bulk density, moisture content and temperature.

Material
Frequency, GHz
Bulk density, kg/m3
Moisture content, %
Temperatures,OC
Wheat
11.3 to 18.0
720 to 880
10.6 to 19.2
-1,9,17,24,34,42
Com
11.3 to 18.0
695 to 830
9.0 to 19.2
14,24,34
Table 2. Linear regression coefficients for wheat and com at two frequencies 11.3 GHz and 18.0
GHzat24C.
Frequency
11.3 GHz
Function
18.0 GHz
7/Jl
'l/J2
Material
wheat
com
Slope, a
0.0193
0.0247
Intercept, b
-0.0832
r2
0.9876
'l/Jl
wheat and com
'l/J2
wheat
com
wheat and com
0.0183
0.0207
0.0304
0.0160
-0.1089
0.0392
-0.0896
-0.0823
0.0588
0.9600
0.9619
0.9794
0.9534
0.9559
2.0 ,.------,.----,..----r----r--.,..---,
1.0
0.8
1.5
0.6
..: 10.
,t
0.4
'l"
~
'
i~
0.2
_1C
240C
0.5
642C
.#
14C
24C
34C
If!
.i}
.~
$
0.0
2.5
3.0
3.5
4.0
0.0
4.5
t'/p
Fig. 1 Argand diagram of the normalized effective

complex permittivity for wheat at indicated
frequencies and temperatures.
k 3
&'/p
Fig. 2 Argand diagram of the normalized effective

complex permittivity for com at indicated
frequencies and temperatures.
182
0040
..
0.6,--,.---,---,.----,---.,-----.
I--,.---r--..,---r-----r--~
'
Com
Wheat
0.35
Com
Wheat
0.5
0.30
0.4
~
0.25
0.3
0.20
0.15
0.1 ::-_~--..L----l.--...1---'----.J
8
10
12
14
16
18
20
0.10 '--_-'-_ _.l...-_---L_ _....L.._ _'--_--.I

18
10
14
16
20
8
12
MOISTURE CONTENT, %
MOISTURE CONTENT,%
Fig. 4 Dependence of 'I on moisture content in

wheat and shelled com at 18.0 GHz and 24C.
Fig. 3 Dependence of 'I on moisture content in

0.45
r----r---r--..,---r-----r--~
0.2
0.40
r---,---,----,---,.---.---.--,
Com
Wheat
Com
Wheat
0040
0.35
0.35
0.30
0.30
0.25
0.25
0.20 '--_-.l.._ _...l.-_--l_ _-L_ _.l.-_-.J

14
16
18
20
8
10
12
0.20 ~----''-----'----.L----1.----1.--_
8
10
12
14
16
18
20
MOISTURE CONTENT, %
MOISTURE CONTENT, %
Fig. 6 Dependence of '2 on moisture content in

Fig. 5 Dependence of '2 on moisture content in

183
FREE SPACE MOISTURE MEASUREMENT

OF CELLULAR CONCRETE
T. Lasri, D. Glay, A. Mamouni, Y. Leroy
Institut d'Electronique et de Micro-electronique du Nord, UMR CNRS 9929
IElVlN-DHS, Avenue Poincare - B.P. 69 59652 Villeneuve d'Ascq Cedex FRANCE
Tel: 33-3 20 197938, Fax: 33-3 20 197896, E-mail Tuami.Lasri@IElVlN.Univ-Lille1.fr
Abstract
The need of non-destructive moisture determination is crucial in civil engineering. Thus,

the purpose of this paper is to present a contact-less method of moisture measurement in
building materials. The measurements are achieved by means of a system (S-Parameters
Measurement System) conceived and built in our laboratory, to provide the S-parameters of the
material under test. The modeling of the reflection coefficient of the metal-backed material
associated to the theory of dielectric mixtures allows, by using an inversion algorithm, the
determination of the moisture content of the sample under test.
Key-words: Moisture, Microwave system, Free space, Building materials.
INTRODUCTION
Moisture measurement systems are needed to take up the scientific, industrial and
economic challenge of the water content determination in solid and liquid materials.
For a long time the use of a network analyzer makes the microwave techniques unsuitable for
industrial purposes. So, many efforts have been undertaken in order to conceive and develop
devices necessary to carry out measurements out of laboratories. Thus, a few indirect moisture
measurement systems based on various physical theories are now rising [1,2]. We have developed
such systems, based on the use of IQ demodulators, which are able to provide the
S parameters of the material under test. According to the application, these systems have been
used to perform wave-guide or free space measurements in a transmission or reflection
configuration. In this study we present a moisture measurement device that makes use of the
reflection coefficient of a metal-backed material. The materials tested in this investigation, sand
and concrete, concern the building industry. For conditioning reasons the feasibility study for
concrete has been led on cellular samples.
In keeping within the objective of the materials characterization in an on-line industrial setting, we
have opted for the free space technique.
MEASUREMENT SYSTEM
There is a considerable interest in methods and devices for the determination of the
moisture content in many industrial branches [3] and in particular in building industry [1]. One of
the more attractive mean for online moisture content monitoring is indisputably the free space
technique. Such a technique associated to a system operating at 10 GHz has already been used to
characterize materials in terms of permittivity and thickness [4]. In the present study, a second
S-Parameters Measurement System (SPMS), operating at 2.45 GHz, has been built according to
the same principle. This device, based on the use of an IQ demodulator, is able to measure the
scattering parameters with an accuracy of 0.2 dB in magnitude and 2 in phase in the range
[0, -30 dB]. This system is associated to a hom antenna to perform reflection coefficient
measurements according to the schematic diagram of the free space measurement set-up given in
Figure 1.
184
SPMS
F=2.45 GHz
S11m
V
fl
horn antenna
TI
L
sample
e- if!o{L'-h)
L'
f2
e- if!o{L'-h)
SII
8 21
"
T2
S!J
S21
-1
metal plate
Figure 1 : Schematic diagram and flow graph of the experimental set-up.

In this configuration a single transmit/receive hom antenna is used to collect the signal that is back
scattered from the metal. We have shown [4] that the reflection coefficient of the metal-backed
sample, SlIm' deduced from the flow graph system (Fig. 1) is given by the following equation:
S
11m
7lTzS~ls
= f I + ---,-"::-":;;"""::"::;'--2jPo{i-h)
,
e
(1)
- f 2 S 11s
where the reflection coefficient S~lS is given by :
,s
ils
= Sus - 1 + S
Sl1s
(2)
lIs
In this expression Sl1s and S21s are the reflection and transmission coefficients that depend
on the permittivity of the material under test (thickness h). The coefficients [1, [2, and T1, T2 are
respectively the reflection and transmission coefficients which take into account the antenna and
the space contained between the antenna and the distance dref (dref is chosen so that the far field
conditions are fulfilled). These coefficients (fl, [2, and T1, T2) are determined when there is no
sample on the metal plate.
Thus, inversion methods allow, by means of this model and the reflection coefficient
measurement, the retrieval of the material permittivity for a known thickness h.
METHOD AND RESULTS
This system has been used to perform measurements on sand and cellular concrete. First,
we give the results obtained when sand was tested. The experimental conditions that have been
chosen consist in moistening a sample of sand and to lay it on a metal plate. Then, versus time and
at room temperature, we have determined the absolute moisture content Ha by weighing. In
addition, for each value of Ha we have collected, by means of the system previously presented, the
magnitude and the phase shift of the reflection coefficient of the metal-backed layer of sand. We
give in figure 2 the results of the measurements of Ha versus time.
185
Ha(%)
25~~------------------------~
20
15
10
5
12
16
20
24
28
drying time (days)
Figure 2 : Drying of a sample of sand initially moistened versus time at room temperature.
As one can see the drying is relatively slow. On this graph is also given the curve resulting
from the fitting of the set of data points by the following polynomial equation :
Ha (%) = -1.76310-4 d 4 +1.192 10-2 d 3 -2.31310-1 d 2 +1.05810-1 d +24.85
(3)
The regression coefficient r2 is equal to 0.9994.
Concerning the reflection coefficient measurement, the results are given in figure 3.
dB (Sllm)
deg(Sllm)
45
Or-------------------------~
T=15C
22.5
0
-22.5
-45
-67.5
-25 +------+---+------+-----1,....-----1
15
20
25
o
5
10
-90
0
~~
10
15
25
20
~~
Figure 3 : Reflection coefficient ( SlIm) versus the moisture content ( Ha ) of sand :

8;1)' =3.2 - O.2j and Pd =1.6 ( - - model, _ measurement)
In order to have a representative model of this configuration measurement, we have used
of dielectric mixing formulae. Several empiric, semi-empiric or simplified theoretical mixing rules
can be applied regarding to the application [5]. In the mixing law approach, the sand and the
cellular concrete are considered as a mixture of three phases. So, the effective permittivity 8~ is
described by the following expression:
K=v1R" +v2R +v3 .g-
(4)
s;
In this relation VI, V2 and V3 are the fraction volume of each constituent and 8t, 8; and
their
permittivity [6]. From this equation for three phases such as dry material, water and air, we have
derived another expression of the permittivity :
=~8;1)' +PdHa(~8:ater
186
-1)
(5)
where &~ is the permittivity of dry material,Pd its dry density and Ha
= mwater
the absolute
mdry
moisture content.
Thus, from equation (5), we obtain the expression of the moisture content Ha :
-~e;ry 1
H=---::==---'a (~e:ater -1) Pd
(6)
The combination of this relation, the expression of the reflection coefficient, established for such
experimental conditions, and the measurements performed leads to the moisture content by using
an inversion process. The results of this computation are also reported on the figure 3.
Due to the fact that the drying process lasted 28 days (Fig. 2) and that the room temperature was
not controlled, relatively big temperature variations between the days and the nights can have
occurred. So, to appreciate the impact of this parameter on the model, we have made use of the
inversion algorithm for two temperatures (Fig. 3). In this way, we have imposed &:ater = 77 -12}
for T=25 C and e:ater =79-16} for T=15 C. The results of these computations show, first that
this model gives a good prediction of the reflection coefficient and then that the influence of the
temperature is not too strong.
The same proceeding has been led on concrete and has permitted to collect the absolute
moisture content and the reflection coefficient of a metal-backed sample of concrete. We give in
figure 4 the comparison between the experimental and the theoretical values for the magnitude
and the phase shift of the reflection coefficient.
dB (SUm)
deg(Sllm)
-8
-9
-10
-10
-11
-20
-12
-30
-13
-40
-14
-15
-50
0
10
15
20
25
30
Ha(%)
10
15
20
25
30
Ha(%)
Figure 4 : Reflection coefficient (SlIm) versus the cellular concrete moisture content (Ha) :
e:ater = 77 -12} (T=25C), e;ry = 2.05- 0.005} and Pd = 0.5 ( - - model, measurement)
F or this second example we also note a good agreement between the measurement and the
prediction of the model. The advantage of the method presented in this paper is that there is no
need of a calibration curve for the moisture determination. In fact, the knowledge of the
parameters that characterize the dry material (e~ and Pd) and the permittivity e~, calculated by
the inversion of the equation (1), are sufficient to determine the absolute moisture content. We
give in figure 5a the comparison between the moisture content obtained by weighing and the one
given by the model.
187
Ha (%) by model
25~----------------~
relative error (% )
30r---------------------------~
20
20
10
15
10
-10
-20
~O~----~--~----_+--
5
10 15 20
Ha (%) by weighing
__
15
10
~----~
20
25
Ha (%) by weighing
25
(b)
(a)
Figure 5 : Validity of the model for the cellular concrete moisture content retrieval
computation for T=25 C (~ ) and T=15 C (I> )
These results exhibit a good agreement between computed and experimental results.
A meaningful illustration of the model precision is given in figure 5b. On this graph is represented
the relative error observed between the moisture content obtained by weighing and the one
determined by the model. The relative error is defined as :
.
0
_
Ha (by model) - Ha (by weighing)
(7)
relatIve error (Yo) -100 x
H (b
'ghi)
a
ywel ng
Finally, the agreement observed between the measurement values and those predicted for both
materials selected suggests that the approach chosen is realistic.
CONCLUSION
This study shows that microwave techniques offer a useful method for measuring the
moisture content of building materials. We have demonstrated that a low cost system, associated
to a relatively simple modeling of the reflection coefficient and a mixing theory, allows a good
prediction of the moisture content in sand and cellular concrete.
Furthermore, the free space technique seems to be the method of choice for moisture content
determination in an industrial context.
REFERENCES
[1] : K. Kupfer (ed.) "9th Meeting on Moisture Measurement (in German, 9 Feuchtetag 1997)". Weimar, Germany,
MFPA University of Weimar, September 1997.
[2] : T. Hauschild, F. Menke "Moisture measurement in masonry walls using a non-invasive reflectometer".
Electronics Letters, Vol. 34, n025, pp 2413-2414, 10 December 1998.
[3] : AW. Kraszewski "Microwave Aquametry - Needs and Perspectives ". IEEE Transactions on microwave
theory and techniques, Vol. 39, n05, pp 828-835, May 1991.
[4] : T. Lasri, D. Glay, A Mamouni, Y. Leroy "Non-destructive testing of materials by microwave systems".
Electronics Letters, Vol. 34, n05, pp 470-471, 5 March 1998.
[5] : D.S. McLachlan, A Priou, I. Chenerie, E. Issac, "Modeling the permittivity of composite materials with a
general effective medium equation". Journal of Electromagnetic Waves and Applications, Vol. 6, N.9,
pp 1099-1131, 1992.
[6]: K.B. Khalid, Z.B. Abbas "Development of microstrip sensor for oil palm fruits". Workshop on
Electromagnetic Wave Interaction with Water and Moist Substances, Atlanta, pp 45-50, June 1993.
188
Measurement of the Moisture and Salt Content

of Building Materials
Werner Leschnik, Udo Schlemm
Building Physics and Materials Science Section, Technical University ofHamburg-Harburg
D-21071 Hamburg, Germany E-Mail: leschnik@tu-harburg.de
ABSTRACT. For measuring the moisture and salt content at building materials and components a
reliable, non- or less-destructive moisture meter for repeated measuring is demanded. To develop
a new electric meter, an investigation into the dielectric properties of building materials was made.
This investigation proved, that at microwave frequencies there is a good correlation between the
real part of the complex relative permittivity and the moisture content. The imaginary part of the
complex relative permittivity depends both on the moisture and salt content of the material. Based
on these results a microwave method and a prototype of a measuring instrument for measuring the
moisture and salt content of building materials were developed and tested in laboratory and in situ
as well.
1 INTRODUCTION
Many damages in buildings are caused by moisture and salt. To eliminate these defects, the measurement of the moisture and salt content of building materials and components is necessary. For
precise measuring of the moisture content of components only classic destructive methods are
disposable, f e. the oven-drying method. A continuous observation of the moisture content is not
possible. Also for measuring the salt content of components only destructive methods for analysis
in laboratory are applied.
It was the aim of several research projects of the authors, the collaborators and the partners in cooperation 1 2 3, to develop a mostly ideal moisture meter for less destructive measuring of the
moisture content of building materials and lateron to expand the method for in-situ and lessdestructive measuring of the salt content. This paper reports on the applied measuring method and
the results which were obtained.
2 MEASURING METHOD
To develop a moisture meter with the demands of Table 1 different methods are applicable. Comparing the advantages and disadvantages of the different techniques it results, that an electrical
method seems to be most suitable for developing a moisture meter meeting the requirements
Research project of the Deutsche Forschungsgemeinschaft, No. LE 765/1 and LE 765/2 Diagnostik und Sanierung, partly in cooporation with Prof Dr Venzmer, Technical University Wismar, final report 1995 (in German)
Research project of the Stiftung Volkswagenwerk, No. II67 193 Kontinuierliche und zerstorungsfreie Feuchteme6verfahren mit Mikrowellen, in cooperation with Prof Dr Knochel, Prof Dr Kammeyer and 3) Prof Dr Eggers, Technical University ofHamburg-Harburg, final report 1996 (in Gennan)
Research project of the Deutsche Forschungsgemeinschaft, No. LE 765/3-1 and LE 765/3-2 Entwicldung eines
Feuchte- und Salzmessers, final report in preparation (in German)
189
shown in Table 1 [1]. Therefore, further investigations were carried out to develop a moisture
meter measuring and evaluating electric parameter.
Table 1: Properties of an ideal moisture meter
Measurement
non-destructive or less-destructive
continuous and repeated at the same
position
independent of structural parameters of
the sample
determining the moisture profile
independent of solved salts of the sample
save in view of injuring radiation
3 DIELECTRIC PROPERTIES OF BUILDING MATERIALS

As the real and the imaginary part as well of the complex relative permittivity E are much more
sensitive to water than to dry building materials, both parameters, the dielectric constant E' and
the loss factor E" can be used to measure the moisture content of building materials. To get more
information on the dependence of the electromagnetic field parameter on frequency and on the
structural, mechanical and chemical properties of the building materials, a detailed investigation
into the dielectric properties of building materials in the frequency range from 50 Hz to 4.000
MHz was made [2]. Samples were made of different sorts of brick, sandstone, sand-lime brick,
lime mortar and cement mortar. The moisture and salt content of the samples were systematically
varied and determined absolutely by classic methods. For salt contamination a mixture of 34%
NaCl, 63% Na2S04 and 14% NaN03 was used, which is relevant for salt damages at components
[3]. By means ofa network analyser the electric signals were measured and the complex dielectric
constant was calculated. The main results of this investigation are shown in Table 2.
Table 2: Frequency ranges and dielectric properties of building materials
Frequency
range
Properties of building materials
100Hz
to
100 MHz
A precise reproducible measuring of the moisture content is impossible. Contact

and dielectric transition problems are of priority, and in general a separation of
the influence of water and salt on the dielectric parameter is impossible.
100 MHz
to
1000 MHz
The dielectric constant and the loss factor depend in different way on the moisture and the salt content of the probe. By measuring both parameter, in general
a calculation of moisture and salt content is possible.
1000 MHz
to
4000 MHz
The dielectric constant becomes more and more independent of the salt content
of the probe. At frequencies above 2.000 MHz the real part E' of the complex
relative permittivity E is correlated with the moisture content and to a large independent of the salt content and independent of the building material as well.
The imaginary part of the complex relative permittivity depends strictly on the
moisture and salt content of the material. Thus the moisture and salt content of
building material can be determined by measuring both dielectric parameter at
Microwave frequencies (see also [2][4]).
190
4 THE MICROWAVE SENSOR SYSTEM FOR MOISTURE AND SALT MEASUREMENT

The Microwave Sensor System was initially developed for moisture measurements at building
materials and lateron extended for salt measurements as well. To determine the moisture profile of
a component, a sensor (aerial) is necessary penetrating into the building material. The displacement of the aerial(s) enables the detection of the moisture profile of the specimen. In general, it is
possible to look from outside into the building component by applying the Time-DomainReflectometry. However, usual inhomegeneities inside the component cause reflection and diffraction, so a distinguished analysis of the reflected signal is nearly impossible.
Suitable is a measuring system that requires two parallel bore holes, each for the transmission and
the receiving aerial. A laboratory measuring system based on a two-borehole-method was f. e.
developed by RUDOLF and WIGGENHAUSER [5]. More comfortable in application is a system,
which requires only one bore hole. A measuring system using this method was developed by the
co-operation partners and was already presented on various occasions, f. e. [6][7][8]. A plane
microwave field (frequency about 2.450 MHz) is radiated by a patch antenna into the component.
The electric field is detected by a dipole antenna within a bore-hole of 16 mm diameter and the
detected signal is analysed by a network analyser. All electric and mechanic functions are computer controlled. By simulating the dielectric field, f. e. by numerical methods [9], both parts of
the dielectric constant can be calculated of the measured electric data.
The measuring system was tested in laboratory and in situ as well and improved step by step. It
resulted, that the developed measuring system is suitable for determining the moisture content and
moisture profiles of building materials and structural elements.
The investigation into the dielectric properties of building materials had proved, that at a frequency of about 2.450 MHz (i) the real part of the complex relative permittivity, the dielectric
constant, is correlated with the moisture content of the specimen and is fairly independent of the
type and the salt content of the material, and (ii) that there is a relation between the imaginary part
of the complex relative permittivity, the dielectric loss, and the moisture and salt content of the
material. With this, it is possible to determine the moisture and the salt content of the specimen as
well by calculating the real and the imaginary part of the permittivity. Further investigations show,
that in fact a separation of different salt contents is possible by analysing both parts of the complex permittivity.
In Fig. 1 and Fig. 2 the results of dielectric measurements at lime-bricks are shown. The moisture
and salt content of the probes was systematically changed. The real part of the complex permittivity, the dielectric constant, is nearly independent of the salt content and there is a good correlation between the moisture content and the measured value of the dielectric constant. A determination of the moisture content is possible. The imaginary part if the complex permittivity, the loss
factor, depends obviously both on the moisture and salt content.
Thus a determination of the salt content is possible. Whilst the correlation between moisture content and the dielectric constant had already bee calculated, the derivation of the functional connection between salt content and the dielectric loss is still in process. [10][11]
191
salt content
-~-0,8
-0-0,7
-0-0,41
- ....-0,32
-.-0,16
12
11
10
-.-0
M.%
M.%
M.%
M.%
M.%
8
-0)
7
6
5
.. :..... f
4
3
468
10
12
moisture [M .%]
Figure 1
Dielectric measurements at a lime-brick wall at 2.450 MHz, Dielectric Constant s' depending on
moisture and salt content
salt content
-0-0,8 M.%
-0-0,7 M.%
. .
:. : ~:~! ~:~ : '-. :': :::::::::::::tzS~::::::::

-':.: ..... ~:::.:: ..
. -.-0
M.%
............................ , ........... '

........................... ~ . . .
.
........................~................
:to)
:i~.];:
>ii~j:::- ......
o~~--~~--~~--~~--~~--~~--~~
10
12
moisture [M. %]
Figure 2
Dielectric measurements at a lime-brick wall at 2.450 MHz, Loss Factor S" depending on moisture
and salt content
Of course, there are still some problems to be solved. For instance: if there is a high moisture
content, a high attenuation of the microwave signal will result and the measurement of the moisture content inside a wall may be difficult at a depth of more than about 25 em, Or: a reflection of
microwaves f e. at holes inside the component may disturb the measurements. Therefore, a new
measuring apparatus is developed with a rotating antenna to avoid interference zones.
More detailed information on the applied method, the measuring system and measuring results
concerning the measurements of the moisture and salt content at building materials are published
by one of the authors in [12].
192
5 SUMMARY
A new method for measuring the moisture and salt content in building material and structural elements by means of Microwaves has been developed. Tests in laboratory and in situ as well confirm that this system really is suitable for this application.
The advantages of the developed system for the application in civil engineering are:
The moisture content of building materials can be determined independently of the type of
material and independently of the salt content by detecting the dielectric constant only
The salt content of building materials and components can be measured by detecting both the
dielectric constant and the loss factor
The measurement is safely (no radiation, very low microwave power), and quickly
Moisture and salt profiles can be measured, a continuous and repeated measuring is possible
Only one bore-hole of 16 mm diameter is necessary
The application of microwaves at the public frequency of2.450 MHz requires no permission or
special admission
By applying small and cheap microwave components of the satellite communication technique
it will be possible to develop a small and comparably cheap moisture and salt meter
6 ACKNOWLEDGEMENT
The authors wants to give thanks to the Deutsche Forschungsgemeinschaft and the Stiftung
Volkswagenwerk, who supported this research by grants.
7 REFERENCES
[1]
[2]
[3]
[4]
[5]
[6]
[7]
[8]
[9]
[10
[11]
[12]
Leschnik W: Zur Messung der Feuchte an Bauteilen. 4. Tagung Feuchtegeschadigte Wande, Technische
Hochschule, Wismar, 1989, proceedings.
Hauenschild C and Leschnik W: Untersuchungen zum dielektrischen Verhalten von Baustoffen - Grundlage
fur die Entwicldung von Feuchtesensoren. 9. Bauklimatisches Symposium, (1994), Dresden, proceedings.
WTA-Recommendation 2-2-91: Sanierputzsysteme (protective Rendering Systems). Published in: Bautenschutz+Bausanierung 15 (1992), S. 59-63.
Hauenschild C and Leschnik W: Dielektrisches Verhalten von mineralischen Baustoffen. In: Materialfeuchtemessung. Hrsg.: K. Kupfer. Expert-Verlag, Renningen-Malmsheim (1997).
Rudolf M and Wiggenhauser H: Feuchteprofilmessungen mit Mikrowellen im Mauenverk. Feuchtetag,
BAM, Berlin, Sept. 1993, proceedings.
Leschnik W, Hauenschild C, Boltze T, Menke F. and Knochel R: Tiefenaufgeloste Feuchtemessung an
Feststoffen
Menke F, Knochel R, Boltze T, Hauenschild C and Leschnik W: Moisture measurement in walls using
microwaves. MTT, Orlando (USA), Mai 1995, proceedings.
Menke F, Knochel R: Portable Moisture Measurement System. IMS Workshop, Athens, GA (USA), June
1996, proceedings.
Boltze T, Hauenschild C, Kammeyer K D and Leschnik W: Robuste Schatzverfahren fur tiefenaufgeloste
Feuchtemessung mit Mikrowellen. Feuchtetag '95, Berlin, September 1995, proceedings.
Leschnik W and Schlemm U: Entwicldung eines Verfahrens zur Feuchte- und Salzmessung an Bauteilen.
Feuchtetag, Wismar (1997), proceedings.
Leschnik W, Schlemm U and Pinsler F: Entwickung eines Verfahrens zur Feuchte- und Salzmessung an
Bauteilen. DFG-Kolloquium, Kassel (1997), proceedings in preparation.
Leschnik: Measurement of the Moisture Content of Building Materials. Conference on New Requirements
for Materials and Structures, Czech Technical University, Prague (1998), Proceedings.
193
DIELECTRIC PROPERTIES MEASUREMENT

USING A CAVITY PERTURBATION TECHNIQUE
Venkatesh, M.S. and G.S.V. Raghavan
Dept of Agricultural and Biosytems Engineering, Macdonald Campus, McGill University
21,111 Lakeshore, Ste.Anne.de.Bellevue, Quebec H9X3V9, Canada
ABSTRACT. In recent years, dielectric properties have gained more importance because of their
capabilities in monitoring specific properties of materials undergoing thermo-physical, chemical &
biological changes. A quick procedure for determining dielectric constant & loss factor at microwave
frequencies (915, 2450 MHz) and certain temperature range is presented. It exploits an extension of
Cavity Perturbation Technique (CPT). The core of this measuring procedure consists in the analysis of
the power at resonance condition by means of quick and narrow frequency sweeps. The concept of an
integrated dielectric analyzer has been introduced and discussed.
Keywords: dielectric constant, loss/actor, permittivity, CPT
1. INTRODUCTION
Microwave (MW) methods of moisture content measurements received the greatest development and
practical application due to the high sensitivity of microwave radiation to water, to the possibility of
achieving high accuracy in non-contact measurements. Dielectric properties, usually expressed by the
relative complex permittivity, E = E' - jE", where E' is the dielectric constant and E" is the dielectric loss
factor, are often used to study the material's characteristics.
Several methods may be used to measure dielectric properties of materials and include transmission,
reflection and resonance techniques. The development of automatic network analyzer and sweep
frequency measurement techniques has lead to faster measurement techniques. Most transmission
techniques utilize either a waveguide or a free-space system Reflection techniques utilize slotted line
or coaxial .probe systems. Phase measurements are important when used for the dielectric
characterization of materials used in material monitoring (Trabelsi et al, 1996). Broadhurst (1970)
utilized a slotted line system in which he measured dielectric constant by relating it to the admittance of
a coaxial transmission line. Use of open-ended coaxial lines have been reported. Resonant cavities are
also used to measure dielectric properties.
2. IMPORTANCE OF DIELECTRIC PROPERTIES
Measurement of dielectric properties is important from two points of view. First, since dielectric
properties are transient during processing, functional relationships describing these changes are needed
for process control and optimization. Second, since dielectric properties are intrinsically related to
specific and bulk characteristics of materials, the development of rapid measurement techniques can
potentially lead to rapid indirect determination of a multitude of material properties. The best known
instance of this use is the dielectric determination of moisture content in materials requiring this
parameter for quality control (Nelson et al, 1990). In recent years, dielectric measurements have gained
more importance because of their capabilities in monitoring specific properties of materials. At the MW
194
frequencies, in particular, the variation of the e can monitor selectively the concentration change of
groups with high mobility thus allowing the kinetics of polymerization reactions as well as vaporization
phenomena can be monitored (Bhutta et al., 1995). For this kind of application, single frequency (915
or 2450 MHz) measurement techniques are more accurate, faster and less expensive; in particular, those
based on the use of a cavity resonator reach acceptable sensitivity and accuracy. Although a great deal
of progress has been achieved in the development of instrumentation and in the understanding of the
functional relationships between dielectric properties and other parameters, the growing trend towards
MW processing has created a need for more flexible equipment to quantify the dielectric properties of
many agri-food products under a wider variety of conditions. Existing measurement technology is very
expensive and the applicability of the various measurement techniques to certain types of material has
been questioned. The study presented here is based on the assumption that it is possible to develop a
system to measure the dielectric properties of agri-food products which is less expensive and more
flexible than existing equipment.
2.1. Objectives
The main objective is to further our understanding of the dielectric properties, measuring technique
(CPT) and behavior of agri-food materials subjected to electromagnetic radiation. The specific objectives
are:
To develop a permittivity analyzer to operate at 2 major ISM frequencies (915, 2450 MHz) and
various temperatures using the concepts of cavity perturbation technique (CPT);
To estimate the dielectric constant and loss factor of few standard reference materials.
3. PERTURBATION THEORY
It is well known that the insertion of a dielectric material different from air inside the cavity affects its
resonant frequency, fo and quality factor, Q. If the sample is small enough compared to the cavity
dimensions, perturbation theory may be used to link frequency shift ilf to the dielectric constant, e' and
the quality factor shift ilQ to the loss factor e". The survey and the experimental procedures of CPT,
commonly employed for measuring the dielectric properties of materials have been reported (Altschuler,
1963; Gallone et aI, 1996). This method allows for simple and precise cavity measurements and also
improves the rapidity.
3.1. CPT
The Concepts of cavity perturbation technique have been developed and widely applied. Presently, many
passive and active measuring techniques are available and much work has been carried out by researchers
using this method, thus proving it to be a very powerful tool of investigation. The experimental
requirement of CPT consists in the determination of the characteristic parameters of a resonant cavity,
namely the resonance frequency and the merit factor Q (Q-factor) of the cavity. The parameters have
to be measured with the empty cavity and with a small dielectric sample, properly inserted. Then the real
and imaginary parts of the complex permittivity can be calculated. All known techniques feature the
resonant frequency measurement by taking the minimum (maximum) of the power reflected (transmitted)
by the cavity to the external network during an appropriate frequency sweep. The Q-factor is usually
found by measuring the positions of the -3 dB points in the resonance curve with appropriate coupling
conditions, although it is possible to measure the Q-factor directly from the decay time of the output
195
signal of a cavity excited by microwave pulses (Montgomery, 1947) or through the phase shift with a
frequency tracking system (Akyel et al., 1978).
4.1 Resonant Cavities (915,2450 MHz)
The permittivity measurement is based on the perturbation technique and uses a cavity operating in
standard TMolO mode (ASTM, 1986). The material used for cavities are copper (915 MHz) and brass
(2450 MHz) of high thermal conductivities and are recommended for such measuring systems (Gauthier,
1997). These circular cavities were connected to an external constant, recirculating temperature control
bath (Fisher Scientific ISOTEMP 1006S) permitting the dielectric properties measurement at various
temperatures. TMolO mode was chosen for its simplicity in mathematics and symmetry. The design and
fabrication of the resonant cavities have been reported elsewhere. The excitation of the microwave signal
and the coupling is usually done using standard antenna loop and with proper tuning, one can aim at the
best peak conditions for the desired resonant frequency of the cavity. The main perturbation equations
followed by the perturbation theory are :
'=l+(Besselxv,x(f';,f,))+v,
//=
Bessel
2
x(_l___l_)x~
Qs
(1)
(2)
Qv
Vs
where :Bessel = Bessel function for circular (TMolO ) cavity = 0.539, Qs = Q of cavity with sample, Qv
=Q of empty cavity, Vs = volume of sample (mml, Vc = volume of cavity (mm)3,fv= resonant frequency
of empty cavity (GHz),/s = resonant frequency when sample is in place (GHz). To perform measurement
with this technique, the user has to know the following main parameters regarding the cavity; (i) resonant
frequency of empty cavity and when loaded with sample (ii)Q of cavity with and without sample (iii)
volume of the cavity (iv) volume of the sample (all SI units). All samples were filled, carefully sealed
using micropipets (1 0-100JIL) made of borosilicate glass and were measured in two replicates. The
following section briefly describe the electronics and essentials of the dielectric analyzer. The present
study is aimed at measuring the dielectric properties of an array of agri-food materials at least 2 ISM
approved frequencies and various temperature ranges.
5. CONCEPT OF AN INTEGRATED DIELECTRIC ANALYZER (components and functions)
Venkatesh et al (1997) reported the development of a new system that determines the dielectric
properties ofMUT (material under test). It consists of a resonant cavity calibrated to 915 MHz and 2450
MHz, and apermittivity (dielectric) analyzer which consists oftwojrequency synthesizers (915, 2450
MHz), a micro-controller, a detector, an analog-to-digital converter (ADC) and a common power

supply. These synthesizers are in fact the programmable microwave generators operating in a frequency
ranges :600 to 1200 MHz and 2.2 to 2.5 GHz. It is programmable by incremental steps of 20 kHz
frequency, thus giving the possibility of 50,000 and 12,000 points in the wavelength of interest,
respectively. A micro-controller is included in the analyzer for the control of the synthesizer and the
196
ADC. In order to reduce the reaction period of the software during acquisition, the software starts with
a wide scanning in the selected band, and once the resonant frequency has been tracked, the scanning
converges to this area. The power is measured using a zero-bias diode. The voltage of the diode is a
function of the power of the received signal. A calibration curve is stored by the software. The voltage
measurement is made by a 16-bit ADC. For each voltage measurement, signal is converted and returned
to the computer through the serial port. Complete permittivity analysis was carried within 10 seconds
or even less. This equipment was connected to an IBM compatible PC 486 through a RS232 cable to
a serial port. With the help of a computer pro gram, the value of the maximum resonant frequency can
be found and from this information, E' can be computed. At the same time, with a calibrated output
detector is calibrated, we can find the Q of the cavity Q = (f Ill) at -3 dB, where I1f is the shift in
resonant frequency. Once Q is known, E"is computed.

Table 1. shows the measured values (avg. of 3 readings I replicate) of E' and E" of certain materials
measured at 915, 2450 MHz and different temperatures using CPT. The reported values are in
agreement with the values illustrated elsewhere. However the percentage variation caused by other
measuring methods, instrumental errors and transient nature of materials, cannot be discounted. The
integrated permittivity analyzer is simple, easy and flexible to operate. It is currently under
developmental stage and has a promising potential to serve as an excellent investigative tool in the
microwave measurement aspects. It is less expensive and portable, when compared to the commercially
available network analyzers. Cavity perturbation technique (CPT) is found to reliable measuring system
in terms of accuracy, high temperature capability and low loss measurements. The disadvantages of this
system include; difficulty in measuring bulk and high lossy samples, limited frequency sweeps, limited
modes of cavity configuration (higher TM I TE modes), etc. The method presented here may find useful
for fast and accurate measurements in both static &dynarnic applications and MW-material interactions.
Acknowledgement: The authors thank the financial support of NSERC and FCAR.
REFERENCES
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
(9)
(10)
Akyel, C., Bosisio, R., and April, G.E. 1978. An active frequency technique for precise measurements on dynamic
microwave cavity perturbations. IEEE Transaction on 1M 27:363-368.
Altschuler, H.M. 1963. Dielectric constant. Handbook ofMicrowave Measurements (M. Sucher & J. Fox, eds.).
Brooklyn Polytechnic Press, New York, NY. vol. 2:530-536.
ASTM (American Society for Testing and Materials).1986. Standard test methods for complex permittivity
(dielectric constant). Designation D 2520-86, Method B (resonant cavities).
Bhutta, E., Livi, A, Levita, G., and Rolla, P.A 1995. Dielectric analysis of an epoxy resin during cross-linking.
J. Polym. Sci. PartB: Polym. Phys. 33:2253-2261.
Broadhurst, M.G. 1970. Complex dielectric constant and dissipation factor offoliage, NBS special report 9592,
October 1970.
Gallone, G., Lucardesi, P., Martinelli, M., and Rolla, P.A. 1996. A fast and precise method for the measurement
of dielectric permittivity at microwave frequencies. Journal of Microwave Power and Electromagnetic Energy,
vo1.31, No.3: ppI58-164.
Gauthier, J. Personal communications. 1997-98. GauTel1M, Montreal, Canada.
Montgomery, C.G. 1947. Technique of Microwave measurements. McGraw-Hill, New York, NY, 340.
Nelson, S.O., and Kraszewski, AW. 1990. Grain moisture content determination by microwave measurements.
Transactions of ASAE; 33(4):1303-7.
Trabelsi, S., Kraszewski, AW., and Nelson, S.O. 1996. Use of measurements at two frequencies for phase shift
determination. Proceedings; IMS Workshop "WMFB", June 17-21, San Francisco, USA
197
(11)
Venkatesh M.S., Raghavan, G.S.V. andSotocinal, S.A.1997. DevelopmentofaPermittivity Analyzer to operate

at 915, 2450:MHz frequency using Cavity Perturbation Technique. Paper# 315; CSAE /SCGR Annual
Conference, July 5-8, Vancouver, B.C.
Table 1. Dielectric properties of some standard materials and agri-foods

Sample
f(MHz)
T (1.5C)
e'
e"
Teflon
2450
22
2.74
0.002
Borosilicate glass
2450
22
4.65
0.008
Clay soil (12% m.c.)
2450
22
2.44
0.14
Loamy soil (7 % m.c.)
2450
22
7.01
0.05
Grape seed extract (dry)
2450
22
3.57
0.26
Tap water
2450
915
22
22
78.45
77.69
12.57
9.82
Methanol
2450
22
23.13
13.92
Soya oil
2450 (915)
0
20
40
60
80
100
11.94 (1.99)
15.22 (8.39)
12.06 (4.05)
17.12 (10.46)
12.3 (16.82)
13.89 (22.6)
0.30 (0.04)
0.34 (0.27)
0.35 (0.29)
0.36 (0.42)
0.37 (0.38)
0.37 (0.66)
Canola oil
2450 (915)
0
20
40
60
80
100
11.85 (2.74)
14.98 (8.96)
13.37 (3.73)
12.06 (10.33)
12.3 (16.45)
13.65 (22.19
0.25 (0.16)
0.32 (0.36)
0.37 (0.25
0.34 (0.38
0.36 (038)
0.37 (0.58)
Sunflower oil
2450 (915)
0
20
40
60
80
100
11.98 (2.74)
14.98 (8.86)
13.25 (3.49)
16.74 (10.46)
12.07(17.20)
13.65 (22.69)
0.32 (0.17)
0.33 (0.33)
0.34 (0.29)
0.32 (0.37)
0.36 (0.38)
0.38 (0.61)
Maple syrup
2450
-14.0
-1.5
23.2
39.84
61.49
50.41
13.54
13.77
6.43
Contact point: Dr. G.S.V. Raghavan, McGill University, Canada. Email: raghavan@agreng.lan.mcgill.ca
198
Session on
PHYSICS OF POROUS MEDIA
Chair: Prof. Philip Rasmussen,

Dept. Plants, Soils and Biometeorology,
Utah State University, Logan, ur, U.S.A.

10:30 - 12:30 PM
199
200
DIELECTRIC RELAXATION IN WATER SATURATED

SEDIMENTARY ROCKS
R. Hilje-?,2, J. Widjajakusuma 1 and B. Biswal1,3
1ICA-I,
Universitat Stuttgart, 70569 Stuttgart, Gennany

2Institut fUr Physik, Universitat Mainz, 55099 Mainz, Gennany
3Department of Physics & Electronics, Sri Venkateswara College,
University of Delhi, New Delhi - 110 021, India
ABSTRACT. An approximate method to calculate dielectric response and relaxation functions

for water saturated sedimentary rocks is tested for realistic threedimensional pore space images.
The test is perfonned by comparing the pre~iction from the approximate method against the
exact solution. The approximate method is based on image analysis and local porosity theory.
An empirical rule for the specification of the length scale in local porosity theory is advanced.
The results from the exact solution are compared to those obtained using local porosity theory
and various other approximate mixing laws. The calculation based on local porosity theory is
found to yield improved quantitative agreement with the exact result.
Keywords: local porosity theory, dielectric relaxation, water satured porous media
1. INTRODUCTION
Applying a small electric field to a heterogeneous mixture of homogeneous and isotropic dielectrics can give rise to an effective dielectric behaviour that may differ substantially from
that of the constituents. The ~ffective dielectric constant of the mixture d~pends not only on
the dielectric constants of the constiqIent materials, but also on the geometrical microstructure of the mixtUre [1, 2]. It is tempting to utilize the dependence on the microstructure for
deducing microstructural infonnation from dielectric measurements. This motivates (at least
partially) the continued interest in the dielectric response of waterfilled sedimentary rocks or
soils [3, 4, 5, 6, 7, 8, 9]. Such studies are important for the interpretation of geophysical or
petrophysical borehole measurements [10, 11]. Before microstructural infonnation can be deduced from dielectric measurements a reliable theory is needed that links the observed dielectric
response with the desired microstructural infonnation. Our objective in this paper is to discuss
local porosity theory [12] as an approximate relation between microstructural infonnation and
dielectric response. We compare its prediction against classical mixing laws and against the
exact value for the effective dielectric constant. Our discussion starts with reminding the reader
how the microstructure enters via the microscopic equations. Next we review briefly the most
popular mixing laws. In Section 4 we recall the basics of local porosity theory, and advance a
new length scale, called the percolation length. In Section 5 we present results for four different
201
threedimensional sandstone samples. We give only a short summary of selected results. A more
detailed account is in preparation [13].
2. MICROSCOPIC EQUATIONS
Microscopically the electric fields and potentials are governed by Maxwells equations in the
quasistatic approximation. To be more specific let us consider from now on a two phase mixture of water and rock. The water fills the pore space of the rock. The pore space will be denoted
as lP, while the rock matrix will be denoted as M. The sets lP, M E ~3 are subsets of threedimensional space, and their union = lP u M represents the sample. Within the quasi static
approximation the electrical potential U(r) obeys the equation
\7 . D(r)
D(r) -
0,
-e(r)\7U(r),
rE,r~8M
(1)
rE,r~8M
(2)
where D(r) is the electric displacement vector,

(3)
is the inhomogeneous dielectric function, and 8M = 8lP denotes the internal boundary between
the two phases. Here
I : for rEG
(4)
XG (r) = { 0 : for r ~ G
denotes the indicator function of a set G, and ep, eM are the dielectric constants (possibly frequency dependent) of the constituents (water and rock). For real rocks with typical pore sizes of
order 100,um the quasistatic approximation remains valid for frequencies up to severallOOGHz.
Differentiation of the dielectric function c(r) is not allowed at the internal boundary 8M =
8lP, and hence the equations must be supplemented with boundary conditions on the internal
boundary. Assuming that there is no surface charge density, and denoting the unit normal to the
surface by n we have
lim n . D (r + 'Tln) -
17-. 0
limn x V'U(r + 'Tln) -
17-.0
lim n D(r - 'Tln),
rE8M
(5)
lim n x V'U(r - 'Tln),
rE8M
(6)
17~0
17~O
where the second condition expresses continuity of the electric field component tangential to the
surface. These equations must be supplemented with further boundary conditions representing
the applied external potential at the sample boundaries 8.
Given a solution U(r; 8M) of the above equations all macroscopic properties of interest can
in principle be calculated from it. An example would be the effective macroscopic dielectric
constant . It can be defined by averaging the solution U(r; 8M) of the microscopic problem
D(r) = - V'U(r)
202
(7)
where A formally represents a suitably defined ensemble or spatial average of a quantity A over
microstructures. The corresponding measure on the space of microstructures generally depends
on parameters 7ri characterizing the statistical properties of the microstructure. In practice the
microstructure is usually not known in sufficient detail. As a consequence it is impossible to
calculate U(r; aM), and one has to resort to approximate theories based on incomplete knowledge. This gives rise to the so called mixing laws. Mixing laws provide a relationship between
the dielectric constants Ci of the constituent phases i = 1,2, ... and the effective macroscopic
dielectric constant of the mixture E. They have the general form
(8)
where 7ri are parameters reflecting the dependence on the microstructure. These parameters
determine which information about the microstructure can be obtained from measurements of
the effective dielectric constant.
3. MIXING LAWS AND RIGOROUS BOUNDS

Let us illustrate the approximations involved in deriving mixing laws with a simple example. If
we define a spatial average over the whole sample as
(9)
where
(10)
denotes the volume of a set G E :lR3 , then we obain by inserting eq. (3) into eq. (2) and applying
eq. (9)
(11)
If we assume X]p> \/U = X]p> \/U in the spirit of mean field theories, then we obtain by comparison with eq. (7) the mixing law of the arithmetic average
(12)
where
IIP I
</>= fT
(13)
denotes the porosity, i.e. the volume fraction of pore space.

Other popular mixing laws for homogeneous and isotropic systems, similar to eq. (12), include
the harmonic averages
E=
(cjp1 + (1 - )c;;,/) -1 ,
(14)
the Clausius-Mossotti approximation with IP as background phase

(15)
203
the Clausius-Mossotti approximation with M as background phase

(16)
and the self-consistent effective medium approximation
(17)
which leads to a quadratic equation for"t. In all of these mixing laws the porosity (j) is all that
is left to characterize the microstructure. Measurement of"t combined with the knowledge of
eMb eIP allows to deduce the porosity from such formulae. It is clear, however, that only the
value for if> given in eq. (13) is correct, and this raises the issue of how accurate the approximate
mixing laws actually are.
In such a situation it is useful to have rigorous upper and lower bounds for"t. Such bounds can
be derived by variational methods under various assumptions about the stochastic nature of the
microstructure [14, 15, 16]. If the microstructure is known to be homogeneous and isotropic
with bulk porosity (j), and if eIP > CM, then
(18)
holds, where the upper and the lower bound are given by the Clausius-Mossotti formulae, eqs.
(16) and (15). For CIP < eM the bounds are reversed. It turns out that the simplest mixing laws,
eq. (12) and (14), violate these bounds. They will not be considered further.
4. LOCAL POROSITY THEORY

A fundamental drawback of the classical bounds and mixing laws is that they depend only on
porosity if> as a single geometric parameter characterising the complex microstructures. Recently a new mixing law was developed which circumvents this restriction by incorporating
fluctuations in porosity and connectivity [12, 17, 18, 19,20,21]. The basic idea of the new approach, called local porosity theory, is to measure fundamental geometric observables (such as
volume fraction, surface density, mean curvature density, Euler-characteristic or connectivity)
within a bounded (compact) subset of the porous medium and to collect these measurements
into various histograms. These histograms are then used in a generalization of the effective
medium approximation to predict effective transport properties.
Let lK(r, L) E denote a cube of sidelength L centered at r. The set lK(r, L) defines a measurement cell (window) inside of which local geometric properties such as porosity or specific
internal surface are measured [12]. The local porosity in this measurement celllK( r, L) is defined as
"'( L)=llPnlK(r,L)1
(19)
'fJ r,
IlK (r, L)I
204
The local porosity distribution p,( >, L) is defined as
p,(>,L)
= m I>5(> -
>(r,L))
(20)
where 6 (x) denotes Diracs 6-distribution, and the summation runs over placements of the measurement cell. The integer
(21)
i=l
is the number of placements of JK( r, L) for a discretized sample (assumed to be a parallelepiped)

with sidelengths Mi' Ideally all measurement cells should be disjoint [12], but in practice this
would give very poor statistics. The support of p,( >, L) is the unit interval 0 :::; > :::; 1 for all L.
The second geometrical ingredient for local porosity theory characterizes the connectivity of
each measurement cell. We define
Ax(r,L)
if
aLx (r, L) "'* alI4x (r, L) in JK (r, L) n lP
otherwise
(22)
where "'* indicates that there exists a path inside the pore space of the measurement cell which
connects the left boundary a Lx (r, L) of JK(r, L) perpendicular to the x-direction to its right
boundary lI4x (r, L). Similarly we define Ay(r, L) for percolation in the y-direction, and
Az(r, L) for the z-direction. It is possible to relate these quantities to the Euler-characteristic of
lP [22].Given eq. (22) we define the local percolation probability
)..(>, L) =
I:r Ax(r, L)Ay(r, L)Az(r, L)6rj)(/>(r,L)
Lr 6r/>r/>(r,L}
(23)
which gives the probability that a cell with local porosity > percolates in all three directions.
With these preparations the mixing law of local porosity theory reads [12]
t cd
cc( - l
11 CB( - l
.
+ 2l)..( >, L )p,( >, L )d> + cB( + 2l(1 - )..( >, L) )p,(>, L )d> = 0
}0
(24)
where CB and cc are given in eqs. (16) and (15), p, in eq. (20), and)" in eq. (23).
The mixing law (24) is a generalization of the effective medium approximation. In fact, it
reduces to eq. (17) in the limit L -+ O. In the limit L -+ 00 it also reduces to eq. (17)
albeit with in eq. (17) replaced with )..(). In both limits the basic assumptions underlying all
effective medium approaches become invalid. For small L the local geometries become strongly
correlated, and this is at variance with the basic assumption of weak or no correlations. For
large L on the other hand the assumption that the local geometry is sufficiently simple becomes
invalid [12]. Hence we expect that formula (24) will yield good results only for intermediate L.
The question which L to choose has been discussed in the literature [23,24,21,25]. Here we
advance a new proposal. We suggest to use a length scale L p , called the percolation length. It
205
is defined using the function
p(L) =
11
f.J,(,L))..(;,L) d
(25)
which gives the total fraction of percolating cells at length L. Experience shows that p( L) frequently has a sigmoidal shape, and this has led us to define Lp as the length scale corresponding
to the inflection point of p(L). Hence we define Lp through the condition
d?pl
dL2
L=Lp
=0
(26)
assuming that it is unique. The idea behind this definition is that at the inflection point the
function p(L) changes most rapidly from its trivial value p(O) = at small L to its equally
trivial value p( (0) = 1 at large L (assuming that the pore space percolates). We have observed
that the length Lp is typically much larger than the correlation length [21,26].
We remark that there are two other important length scales associated with p( L ). The first of
these is the threshold length Lc defined by
(27)
where Pc can be taken as the percolation threshold for the underlying lattice (Pc ~ 0.248812
for the simple cubic lattice [27]) or as Pc = 1/3 for the effective medium approximation. This
length scale is particularly important for network models which attempt to replace the complex
microstructure by an effective lattice with similar statistical properties as the real sample. Of
course Lc may not exist when = p(O) > Pc.
The second length scale La is the length at which p( L) approaches its asymptotic value p( 00 )
to a given degree of accuracy. For percolating samples p( (0) = 1. We define La through
Ip(oo) - p(La)1 < 0
(28)
for small 0 > O. The length La may be equated with the size of the so called "representative
elementary volume" (REV)[28] required for representativity with respect to connectivity. La
represents the scale of the averaging (smoothing) region that is needed to ensure that the fluctuating microscopic connectivity can be replaced with an averaged connectivity field defined on
the continuum. The small parameter 0 controls the degree of smoothness. Naturally we expect
Lp < La for small enough o.
5. RESULTS
We have analyzed four samples of sedimentary rocks whose pore spaces were obtained by
computer assisted microtomography. Images of two of the samples are given in Figures 1 and 3
of [21]. Each data set consists of a threedimensional array of O's and 1's indicating pore space
1P' or matrix M. The dimensions of the array are M 1 , M2 and M 3 Table I gives a synopsis of the
characteristics of the four samples.
206
TABLE I: Overview over properties of the data sets for four reservoir sandstones.
Sample
A
B
C
D
Description
Berea
coarse Sst20d
fine Sst6d
Fontainebleau
a
10jtm
30jtm
10jtm
7.5jtm
Ml x M2 X M3
128 x 128 x 128
73 x 128 x 128
95 x 128 x 128
300 x 300 x 299
Lp
180jtm
420jtm
100jtm
225jtm
J
0.1775
0.2470
0.3200
0.1355
e
0.112
0.149
0.193
0.098
Here a is the resolution, and Mi are the dimensionless sidelengths of the sample in units of a.
The bulk porosity was defined in eq. (13) and the length Lp in eq. (26).
We have solved the microscopic equations (2)-(6) numerically using the values e]!> ~ 87.74 and
= 4.7 (in units of co = 8.854 1O- 12 F/m) for the dielectric constants, and calculated from
equation (7) using the averaging procedure defined in eq. (9). The results are shown in the last
column in Table 1.
eM
0.007
Sample
Sample
Sample
Sample
0.006
0.005
0.004
....J
"0
""'-
c..
A
8
C
0
-----)(----e
-.-.-~.-.-.
0.003
"0
0.002
0.001
0
-0.001
100 200 300 400 500 600 700 800
L
Figure 1: Derivative of p(L) (in jtm- 1) defined in eq. (25). The abscissa is the size of
measurement cells in jtm. The percolation length Lp corresponds to the position of the
mru(lmum.
Next we measured jt( J, L) and ).( J, L) as a function of J and L. For the determination of ).
we used a Hoshen-Kopelman algorithm [29]. Integrating the product according to eq. (25) and
207
differentiating the result with respect to L we find the curves shown in Figure 1. The locations
of the maxima give the values of Lp tabulated in Table I.
Finally we solve equation (24) iteratively to find the value of LPT predicted by local porosity theory. We have plotted these values together with the predictions from the other mixing
formulae in Figure 2. We emphasize that, contrary to spectral theories or network models, neither the mixing laws nor the local porosity theory contains any free fitting parameters. While
the Clausius-Mossotti predictions (upper and lower bounds) do not give good estimates the results from the effective medium approximation and the local porosity theory are in much better
agreement with the exact result. Note however that the effective medium values approach zero
for infinite contrast, i.e. EM -+ 0 for crr>/cM -+ 00. The values of LPT on the other hand
remain finite and are in similarly good agreement also in that limit [13, 25].
40
~----~----~------~----~----~
CMl
CM2
32
EMA
exact
LPT
24
II.L>
I .......'
........
."
.'
.'.'
....
..
.' .'
'
.' .'
....
.' .'
.'
.'.'
0
0
.... 8 0 ....--,-8 L~~

.. ...~.~---~~-~--~-~-~--------.'
I ................
16
.. .' .'
'
....
........
.,."
.,.,....
...."
."".;
.",..".",..,-
'
o ~----~----~------~----~----~
0.0
0.08
0.16
0.24
0.32
0.4
Figure 2: Comparison of approximate calculations for evaluated from eq. (15) (upward
triangles), eq.(16) (downward triangles), eq.(17) (squares), and eq.(24 (diamonds) with the
exact result (circles).
Of course the comparison with at zero frequency is not sufficient to judge the quality of the
approximations. Solutions of the frequency dependent complex dielectric function are in preparation [13], and the simultaneous comparison ofreal and imaginary parts is expected to provide
further insight [19]. Specifically, the solution of the frequency dependent inverse problem for
local porosity theory is expected to yield information about porosity and connectivity fluctuations. We emphasize, however, that first the use of Lp must be better established. Currently we
208
view it as a successful empirical rule based on the available data. Further tests are necessary to
corroborate or reject it.
ACKNOWLEDGEMENT: We are grateful to Dr. P.E. 0ren and Dr.S. Bakke for discussion, and
for providing us with the experimental data sets. We thank the Deutsche Forschungsgemeinschaft and the GKKS at the Universitat Stuttgart for financial support.
References
[1]
c. Bottcher, Theory of Electric Polarization, vol. I.
Amsterdam: Elsevier Scientific Pub-
lishing Co., 1973.
[2] C. Bottcher and P. Bordewijk, Theory of Electric Polarization, vol. ll. Amsterdam: Elsevier Scientific Publishing Co., 1978.
[3] J. Korringa, "The influence of pore geometry on the dielectric properties of clean sandstones," Geophysics, vol. 49, p. 1760, 1984.
[4] W. Kenyon, "Texture effects on megahertz dielectric properties of calcite rock samples,"
1. Appl. Phys., vol. 55, p. 3153, 1984.
[5] R. Knight and A. Nur, ''The dielectric constant of sandstones, 60 kHz to 4 MHz," Geophysics, vol. 52, p. 644, 1987.
[6] I. Holwech and B. Njijst, "Dielectric dispersion measurements of salt-water saturated
porous glass," Phys.Rev.B, vol. 38, p. 12845, 1989.
[7] M. Taherian, W. Kenyon, and K. Safinya, "Measurement of dielectric response of water
saturated rocks," Geophysics, vol. 55, p. 1530, 1990.
[8] E. Haslund, "Dielectric dispersion of salt water saturated porous glass containing thin
glass plates," Geophysics, vol. 61, p. 722, 1996.
[9] S. Friedman, "A saturation-dependent composite sphere model for describing the effective
dielectric constant of unsaturated porous media," Water. Resources Res., vol. 34, p. 2949,
1998.
[10] J. Poley, J. Noteboom, and P. de Waal, "Use of VHF dielectric measurements for borehole
formation analysis," The Log Analyst, vol. 19, p. 8, 1978.
[11] J. Hearst and P. Nelson, Well Logging for Physical Properties. New York: McGraw-Hill,
1985.
[12] R. Rilfer, "Geometric and dielectric characterization of porous media," Phys. Rev. B,
vol. 44, p. 60, 1991.
[13] J. Widjajakusuma, B. Biswal, and R. Hilfer. to be published.
209
[14] Z. Hashin and S. Shtrikman, "A variational approach to the theory of effective magnetic
permeability of multiphase materials;' J. Appl. Phys., vol. 33, p. 3125, 1962.
[15] D. Bergman, "Rigorous bounds for the complex dielectric constant of a two-component
composite," Ann. Phys., vol. 138, p. 78, 1982.
[16] S. Torquato, "Random heterogeneous media: Microstructure and improved bounds on
effective properties," Applied mechanics reviews, vol. 44, p. 37, 1991.
[17] R. Hilfer, "Local porosity theory forftow in porous media," Phys. Rev. B, vol. 45, p. 7115,
1992.
[18] R. Hilfer, B.Nst, E.Haslund, Th.Kautzsch, B.Virgin, and B.D.Hansen, "Local porosity theory for the frequency dependent dielectric function of porous rocks and polymer
blends," PhysicaA, vol. 207, p. 19, 1994.
[19] E. Haslund, B. Hansen, R. Hilfer, and B. Nst, "Measurement of local porosities and
dielectric dispersion for a water saturated porous medium," J. Appl. Phys., vol. 76, p. 5473,
1994.
[20] R. Hilfer, "Transport and relaxation phenomena in porous media," Advances in Chemical
Physics, vol. xcn, p. 299, 1996.
[21] B. Biswal, C. Manwart, and R. Hilfer, "Threedimensionallocal porosity analysis of porous
media," Physica A, vol. 255, p. 221, 1998.
[22] R. Hilfer. to be published.
[23] F. Boger, J. Feder, R. Hilfer, and T. Jssang, "Microstructural sensitivity of local porosity
distributions," Physica A, vol. 187, p. 55, 1992.
[24] C. Andraud, A. Beghdadi, E. Haslund, R. Hilfer, J. Lafait, and B. Virgin, "Local entropy
characterization of correlated random microstructures," Physica A, vol. 235, p. 307, 1997.
[25] J. Widjajakusuma, B. Biswal, and R. Hilfer, "Quantitative prediction of effective material
properties of heterogeneous media," Compo Mat. Sci., p. in print, 1999.
[26] B. Biswal, C. Manwart, R. Hilfer, S. Bakke, and P.E.0ren, "Quantitative analysis of experimental and synthetic microstructures for sedimentary rock." to be published.
[27] B. Hughes, Random Walks and Random Environments, vol. 2. Oxford: Clarendon Press,
1996.
[28] J. Bear, Dynamics of Fluids in Porous Media. New York: Elsevier Publ. Co., 1972.
[29] D. Stauffer and A. Aharony, Introduction to Percolation Theory. London: Taylor and
Francis, 1992.
Contact: R. Hilfer, lCA-l, Pfaffenwaldring 27, Universitat Stuttgart, 70569 Stuttgart, Germany
210
DIELECTRIC RELAXATION OF WATER IN POROUS SILICA GLASSES

Yuri Feldman and Anna Gutina
Department of Applied Physics, The Hebrew University of Jerusalem,
91904, Jerusalem, Israel yurif@vms.huji.ac.il
ABSTRACT. The dielectric relaxation properties of porous glasses obtained from sodium
borosilicate glass were studied by Dielectric Spectroscopy in the wide frequency (20 Hz + 1
MHz) and temperature (-100 0c + +300 0C) ranges. The dielectric behavior reflecting the
geometrical disorder was characterized by using models describing the non-Debye slow decay
dynamics. The dielectric response is found to be very sensitive to the geometrical micro- and
mesostructural features of the porous matrix and the properties of the material filling the pores. It
provides information on the hindered dynamics of water molecules, located within the pores and
affected by the surfaces. A simple model has been developed in order to explain the relaxation
process which is well marked in the temperature range: -50C to + 150 c and has a specific
saddle-like shape.
Keywords:
Dielectric relaxation, porous glasses, adsorbed water.
1. INTRODUCTION
Recently much attention has been paid to the properties of water adsorbed on surfaces of porous
silica glasses [1,2]. These glasses can be defined as a bicontinous random structure of two
interpenetrating percolating phases, the solid and the pore networks. The pores in the glasses are
connected to each other and a pore size distribution is narrow. The characteristic pore spacing
depends on the method of preparation, and can be between 2 and 500 nm [3].
It has been shown [4] that silica glasses contain three types of adsorbed water on the pore
surface. The first type forms a physisorbed layer which is desorbed by heating at 90C. The
second and third types form a chemisorbed layer associated with spatially distributed individual
molecules (or groups of several molecules) on the surface. The second type of water comes off
above 190C, while the third type is a more strongly chemisorbed layer, consisting of water
clusters of 40-60 molecules [4], connected to each other by hydrogen bonds and forming an
ordered array of water molecules.
The dynamic properties of adsorbed water obtained from NMR and dielectric spectroscopy are
different from those of the bulk water due to geometrical confinement of the liquid in pores and
interaction of the molecules with active surface sites [5,6].
In our prior paper [7] we investigated the dielectric properties of silica glasses in broad
frequency and temperature regions with the purpose of investigating the dynamics and inferring
the morphological properties of the porous materials. It was shown that the complex dielectric
behavior could be described in terms of the four distributed relaxation processes. The typical
spectrum of the dielectric losses associated with a relaxation of water molecules of the
adsorptive layer for the studied porous glasses versus frequency and temperature is displayed in
Figure 1.
211
FIGURE 1. Three dimensional plot of the frequency and temperature dependence of the
dielectric losses for the porous glass (Reproduced from reference [7]. Copyright 1998, Elsevier
Science).
The first relaxation process which is observed in the low temperature region -100C to + 10C
is due to the reorientation of water molecules in ice-like structures of water clusters. The second
relaxation process has a specific saddle-like shape and is well marked in the temperature range 50 c to +150 DC. This relaxation process is thought to be a kinetic transition due to the water
molecule reorientation in the vicinity of a defect. The third process is located in the lowfrequency region and the temperature interval 50C + 80 DC. The amplitude of this process
essentially decreases when the frequency increases and the maximum of dielectric losses has
almost no temperature dependence. This relaxation process is related to the percolation of dipole
moment excitation within the developed fractal structure of connected pores due to the selfdiffusion of the charge carriers filling the porous space. In the high temperature region, above
150C, the glasses become electrically conductive and show increase in dielectric constant and
dielectric losses in the low frequency limit. This relaxation process is thought to be related to the
Maxwell-Wagner-Sillars polarization process as a result of trapping free charge carriers at the
interface, thus causing build-up of macroscopic charge separation, or space charge with a
relatively long relaxation time.
2. EXPERIMENTAL SECTION
2.1. Sample preparation
We studied three different porous glass samples A, B and C, obtained from the same sodium
borosilicate glass. Glass A was heated at 490C for 165 hours and glass C was heated at 650C
for 100 hours. All the samples were then etched in hydrochloric acid and rinsed in deionized
water. Glass B was obtained from glass A by additional immersion in KOH. This sample was
also rinsed in deionized water. The glasses, which were obtained, indicated different structures.
212
We can assume that glass A contains silica gel in the pore's volume, albeit the silica gel is almost
not present in glasses B and C.
2.2. Experimental technique
Dielectric measurements in the frequency range 20 Hz -;- 1 MHz were performed by using a
Broad Band Dielectric Spectrometer BDS 4284 (NOVOCONTROL) with automatic temperature
control by QUATRO Cryosystem. The accuracy of the measured dielectric permittivity and
losses is estimated to be better than 3%. The measurements were carried out by cooling the
samples from 20 DC down to -100 DC. The samples were then measured at intervals of SDC upon
heating them from -100 DC to 300C.
3.
RESULTS AND DISCUSSION
The temperature dependencies of the relaxation times of the first process for samples A and C
(Fig. 2) demonstrate the Arrhenius behavior with the energies of activation 42 kJ/mol and 80
kJ/mol, respectively, and correlate with the activation energy of bulk ice [8]. However, unlike
the Debye relaxation of bulk ice, the relaxation in samples A and C markedly demonstrates the
non-Debye behavior.
IOOOIT, (K-l)
3.4
3.6
3.8
4.0
4.2
4.4
4.6
10-5 T.=,=.,.,.,.,.,..~=~="..,.,.,.,.~m,.,.=~==~:;),-,.,.,.=.,.,.,j
p'"
4.6
4.8
5.0
5.2
5.4
IOOOIT, (K-l)
5.6
5.8
FIGURE 2. Temperature dependence of the characteristic relaxation time of the first process for
samples A (0) and C (D), and relaxation time of bulk ice (Ll). The top axis corresponds to the
temperature behavior of bulk ice. The calculated activation energies are - 40, - 80 and - 70
kJ/mol, correspondingly.
The fitting non- Debye parameters show different temperature dependencies for sample A and C,
that can be connected to variance of the morphology of pores and the interaction of water
molecules with the pores surface. A similar dielectric relaxation behavior in silica-water systems
at low water content has been observed recently [6,9] and ascribed to reorientation of water
213
molecules in ice-like structures. The formation of the structure of this ice-like water is strongly
dependent on the number of water monolayers covering the pore surface.
The relaxation process manifested in the temperature range -50C to +150 c has a specific
saddle-like shape (see Fig.I). This process is not likely to be due to the physisorbed water, which
mostly participates in the first relaxation process. However, it is apt to be due to the dynamics of
chemisorbed water molecules forming clusters near hydration centers distributed at the interface.
Indeed, one can see that the amplitude of this process remains unchanged in the temperature
range for which this process occurs. This means that despite the heating of the sample, the
number of the kinetic units participating in the process almost does not change.
For the description of the second relaxation process we developed a model, based on the idea of
Macedo and Litovitz [10]. This model considers the relaxation act as an occurrence of two
simultaneous events: a) the molecule must attain sufficient energy to break away from neighbors
and change its orientation, and b) there must be a site with a defect in the vicinity of the moving
molecule with sufficient local free volume for a reorientation to occur. The probability of a
transition from one orientation of the molecule to another is given by
(1)
where Pais the probability of attaining sufficient energy to brake bonds and Pd is the probability
that there is a defect, i.e. sufficient local empty space near the molecule for its reorientation to
occur.
Assuming that the relaxation time is inversely proportional to the transition probability we obtain
the equation
(2)
that describes the temperature dependence of the relaxation time corresponding to the second
process of the relaxation in silica glasses. Here Ha is the height of the potential barrier between
equilibrium positions, T is the temperature and k is the Boltzmann constant, Hd is the energy of
the defect formation. The parameter C has a meaning of an order parameter. It can be defmed by
simple formula as follows: C= V*No IV, where No is the maximal number of defects, V* is close
to the value of the water molecule and the volume of the water clusters V varies with temperature
as a linear function. The pre-exponential constant 'fo can vary with temperature but far less than
the exponential term. The detailed description of the model is presented elsewhere [11].
The experimental data of the relaxation time were fitted to Eq. (2) by means of a least-squares
procedure with the purpose of determining the numerical values of the parameters Ha, Hd, C, and
'ro. The model can be applied for the interpretation of the dynamics of water molecules
chemisorbed by the active hydrate centers and forming water clusters distributed on the pore
surface. According to the mechanism suggested in the model, a reorientation of water molecules
in such clusters can occur near the orientation defect. One has to bear in mind that the nature of
orientation defects in water and ice is very complex. In our case, for instance, the Bjerrum model
of the orientation defects in ice can be applied [12].
214
In the framework of the model suggested above, the observed temperature dependence of the
relaxation time can be described in the following way. The increase of the temperature leads, on
the one hand, to the increase of the probability to attain a sufficient energy to brake bonds
between the reorienting molecule and its neighbors. On the other hand, it leads to the decrease of
the probability of finding a defect in the vicinity of the moving molecule. The interplay of these
two terms in Eq. (2) yields the appearance of the minimum observed in the temperature
dependence of the relaxation time.
The fitting value of the pre-exponential constant 7:0 is _10- 12 + 10-13 s for different glasses. It is in
a good agreement with the characteristic time of water molecule oscillation near its equilibrium
position. The values of activation energies Ha and Hd obtained from the fitting of experimental
data to Eq. (2) are -40 kJ/mol and -30 kJ/mol, respectively. These values are in a satisfactory
agreement with the energies of molecular orientation and energy of defect formation in ice-like
water matter [13].
REFERENCES
[1] Gallegos D. P., Smith D. M., and Brinker C. 1., "An NMR Technique of the Analysis of Pore Structure:
Application to Mesopores and Micropores". J. Colloid Interface Sci., 1988, vol. 124, pp. 186-198.
[2] Allen S. G., Stephenson P. C. L., and Strange J. H., "Morphology of Porous Media Studied by Nuclear
Magnetic Resonance". J. Chem Phys., 1997, vol. 106 (18), pp. 7802-7809.
[3] Rysiakiewicz-Pasek E., and Marczuk K., "The Properties of Porous Glass after Impregnation with Carbon". J.
ofPorous Materials, 1996 vol. 3, pp. 17-22.
[4] Agamalian M., Drake J. M., Sinha S. K., and Axe J. D., "Neutron Diffraction Study of the Pore Surface Layer
ofVycor Glass". Phys. Rev. E, 1997, vol. 55, pp. 3021-3027.
[5] Hirama Y., Takahashi T., Hino M., and Sato T., "Studies of Water Adsorbed in Porous Vycor Glass". J. Colloid
Interface Sci., 1996, vol. 184, pp. 349-359.
[6] Pissis P., Laudat J., Daoukaki D., and Kyritsis A., "Dynamic Properties of Water in Porous Vycor Glass
Studied by Dielectric Techniques". J. Non-Cryst. Solids, 1994, vol. 171, pp. 201-207.
[7] Gutina A., Axelrod E., Puzenko A., Rysiakiewicz-Pasek E., Kozlovich N., and Feldman Yu., "Dielectric
Relaxation of Porous Glasses". J. Non-Cryst. Solids, 1998, vol. 235/237, pp. 302-307.
[8] Chan R. K., Davidson D. W., and Whalley E., "Effect of Pressure on the Dielectric Properties of Ice I". J.
Chem. Phys., 1965, vol. 43, pp. 2376-2383.
[9] Piss is P., Anagnostopoulou-Konsta A., Apekis L., Daoukaki-Diamanti D., and Christodoulides C., "Dielectric
Effects of Water in Water-Containing Systems". J. Non-Cryst. Solids, 1991, vol. 131-133, pp. 1174-1181.
[10] Macedo P. B., and Litovitz T. A., "On the Relative Roles of Free Volume and Activation Energy in the
Viscosity of Liquids". J. Chem. Phys., 1965, vol. 42 (1), pp. 245-256.
[11] Feldman Yu., Ryabov Ya., Puzenko A., Gutina A., and Kozlovich N., "Dielectric Relaxation of Water
Adsorbed on the Pore Surface of Silica Glasses". In: 8'h Tohwa University International Symposium on Slow
Dynamics in Complex Systems, Conference Proceeding, Fukuoka, Japan, November 9-14, 1998, "edited by
Michio Tokuyama and Irwin Oppenheim" ,(in press).
[12] Eisenberg D., and Kauzmann W., "The Structure and Properties of Water". Oxford: At the Clarendon Press,
1969.
[13] Stillinger F. H., "Low Frequency Dielectric Properties ofLiquid and Solid Water II The Liquid State ofMatter:
Fluids, Simple and Complex". Amsterdam: North Holland, 1982, pp. 341-431.
Contact point: Yuri Feldman, Department of Applied Physics, The Hebrew University of Jerusalem, 91904,
Jerusalem, Israel. Phone: +972-2-6586187; Fax: +972-2-5663878; E-mail: yurif@vms.huji.ac.il
215
COMBINED DETERMINATION OF THE WATER CONTENT

OF CAPILLARY POROUS MEDIA AND THE ELECTRICAL CONDUCTIVITY
OF THE PORE SOLUTION BY THE TDR-SS-METHOD
R. Plagge!, M. Malicki2 and P. Haupll
ABSTRACT. Using the information from the travel time measurements of electromagnetic
waves along two metal rods embedded in a capillary porous material, the apparent
dielectric constant can be determined using the time-domain reflectometry technique
(TDR). Because free water has a dielectric constant of fw:::: 81 which is remarkably higher
than that of mineral materials, fmin:::: 4 - 6 or air, fair:::: 1, the volumetric water content of
the porous media can be calculated using a calibration function. The performance of
different conversion functions are discussed by Plagge et al (1996).
To determine the bulk electrical conductivity oh the medium, the resistance of the TDR
probes is measured directly by the steady state method described by Oliver and Cage
(1977). Additionally to the TDR pulse, another step pulse having a rise time <10- 9
seconds is sent to the cable-sensor system, where the specific resistance of the probe can be
measured after fade away of the multireflections. If the relationship between the specific
resistance of the probe and the electrical conductivity is known, the bulk soil electrical
conductivity ca be determined. To estimate the electrical conductivity of the porous
solution, a mixed constituent approach is applied, identifying the influence of water content
and solid matrix.
REFERENCES
Plagge, R., C.H. Roth and M. Renger (1996). Dielectric soil water content determination
using time-domain reflectometry (TDR). In: A. Kraszewski (ed), Second Workshop on
Electromagnetic Wave Interaction with Water and Moist Substances, at the 1996 IEEE
Microwave Theory and Techniques Society International Microwave Symposium, San
Francisco, CA, 17 June 1996, pp. 59-62.
Oliver, B.M. and J.M. Cage (1977). Square-wave and Pulse Testing of Linear Systems.
Electronic Measurements and Instrumentation. McGraw-Hill, New York, pp. 61-64.
Contact points:
1 Institute of Building Climatology, University of Technology Dresden (TU-Dresden), Zellescher Weg 17,
01062 Dresden, Germany. Phone: +49 351-463-3756; fax: +49 351-463-2627;
e-mail: plagge@abkfs1.arch.tu-dresden.de
2 Institute of Agrophysics, Polish Academy of Sciences (PAN), ul. Doswiadczalna 4, 20-236 Lublin, Poland
216
LOW FREQUENCY IMPEDANCE BERAVIOR

OF COLLOIDAL SUSPENSIONS
Lynn Dudley, Dani Or, Stephen Bialkowski, Chad lunkermeier
Dept. of Plants, Soils and Biomet., Dept. of Plants, Soils, and Biomet., Dept. of Chemistry
and Biochemisty, Dept. of Physics, respectively, Utah State University, Logan UT
ABSTRACT. Colloid suspensions exhibit large changes in conductivity and dielectric constants
at low frequencies of kHz to MHz. These responses have been used in geophysics to study
electrochemical properties of variably saturated rocks. We have used an HP 4194A impedance
analyzer for complex impedance measurements of clay suspensions. A method for obtaining
relaxation time constants from impedance spectra is described. Results show that the impedance
measurements reveal information about particle geometry and mobility. The temperature
dependance of the relaxation time constants indicate that polarization of the clay counter ions
takes place in mobile water outside the Stem layer. In this presentation, we show the need to
modify existing physically-based impedance models for variably saturated porous media.
Keywords: diffusion, montmorillonite, permittivity, relaxation

1. INTRODUCTION
Impedance spectroscopy provides the basis for many nondestructive measurement techniques in
different fields including, food sciences - determining electrical properties of foods, fruits and
vegetables, grain moisture, the curing of concrete, and intact porous media leading [1]. Recently,
impedance tomography has been used for 3-D imaging of soil properties during solute transport
experiments [2] and attempts have been made to extend this application to field scale using cross
borehole tomography to delineate spatial variations in soil properties [3]. We are attempting to
introduce these potentially useful techniques to investigations of soil geometrical and
electrochemical properties. The prospects for using a small soil sample in suspension to
determine properties such as surface charge density, surface diffusion coefficients and particle
size distribution are appealing considering presently-available alternatives. Ability to measure
(potentially in-situ) surface electrochemical and geometrical soil properties should enhance
understanding and predictability of transport processes. The success of these and many other
new impedance and dielectric-based applications hinges on the development and availability of
sound models for impedance (and dielectric) behavior of variably-saturated soils. Toward
developing a model for impedance behavior of variably-saturated soils we conducted impedance
experiments on clay suspensions. Our objectives were, first, to develop a conceptual
understanding of the impedance spectra of clay suspensions and second to determine the efficacy
of recently proposed mathematical models of the complex dielectric and conductivity of clay
suspensions.
We used an HP model 4194A impedance analyzer with an HP cell that had two-electrode
configuration. The electrode spacing was 0.3 mm and an electrical field of 160 Vim was applied
across the sample. The cell was suspended in a water bath and temperatures were maintained at
5, 25 or 55C. Electrolyte solution capacitance was measured at various spacing and was a linear
217
function of the spacing, thus extraneous capacitance due to electrode polarization was negligible
in the system. The instrument required two scans to cover the frequency range of interest, 0.1
kHz to 10 MHz and the data was spliced together (Fig. I). An equivalent circuit composed of a
capacitor and resistor in parallel was used to compute the complex impedance, Z*.
Clay was prepared for the experiments by three washings of SWy-1 montmorillonite with 1 M
CaCI2. Clay and electrolyte were placed on a reciprocal shaker for 15 minutes and the clay and
electrolyte solution were subsequently separated by centrifugation. The Ca-saturated clay was
transferred to a dialysis bag and washed salt free. The clay was then separated into three particle
size classes, 1-2, 0.2-1 and <0.2 !lm (hereafter referred to as large, medium and small) by
centrifugation. Time and g force were computed from Stoke's Law for disk-shaped particles.
The particle sizes were verified by scanning electron microscopy. The micrographs showed good
agreement between the desired and actual particle sizes for the largest and smallest fractions, but
the medium fraction contained particles larger and smaller than the target. The surface charge
density was taken as I !lmoljm2 [4]. The clay concentrations in the suspensions were 15 gIL.
Electrolyte conductivity values of 7 to 11 mS/m were measured after the clay settled from
suspension samples. For most measurements the electrodes were positioned so that the electrical
field was parallel to the long axis of the settling clay tactoids. The cell was agitated before each
experiment and allowed to stand for 10 to 20 seconds before taking measurements.
A plot of the real y imaginary components of the impedance (Fig. 1) shows a linear section at
low frequency (frequency increases from right to left in the diagram) and a nearly semi-circular
section that were characteristic of the data sets. The prime and double prime superscripts are
used throughout to denote real and imaginary components, respectively. The small offset in the
data at about 70 Q probably results from slightly different cell conditions and calibration for the
two measurement ranges used for the two scans. The symmetry of the main arch (centered at
120
--a
90
~ 0.8
laW III
<u 0.6
c-- H---I' --+-1"
c::
60
E 0.4
30
E 0.2
II
III
'"
..,
i/~
,~
0
0
30
60
Z' (Q)
90
1.E+02
120
1.E+04
1.E+06
1.E+08
ro (rad/s)
Figure 2. Real and imaginar components

of the immitance as a function of angular
frequency, w.
Figure 1. Impedance plane plot for the small

particle size Ca-saturated monmorillonite.
Z'-25 Q) indicates an impedance element with well-behaved dielectric properties and the slight
218
deviation from a semi-circle indicates a distributed impedance element [5]. The linear section
with a slope approximately W l/2 (where w is the angular frequency =2nf) is usually attributed to
a diffusion process that may be occurring at the electrode suspension interface. The distributed
impedance element is most likely a result of diffusion along particles that vary in size [5].
The real and imaginary immittance, Iz defined as: Iz =Z*/Ro where Ro is the low frequency
resistance [5], show a relaxation in the MradJs range (Fig. 2). The imaginary component was
used for deconvolution (or inversion) of the relaxation time distribution using an expectationmaximization (EM) algorithm [6] because of the well-defined maximum. A model for the
distribution of relaxation time-constant was fitted to the immittance data using Debye-type
behavior for each fraction of the relaxation process [5]. The real part of the complex immittance,
I' was modeled by fitting the imaginary, I", component of the measurement data set and faithful
reproduction of the I' trends from the aCt) for I", even at frequencies not included in the
measurement data set, resulted. The resulting density function of relaxation times, denoted by
aCt), is depicted in Figs. 3 and 4 as a relaxation-time spectra. The spectra exhibit a "shoulder" at
relaxation times> 10-3 s, attributed to a diffusion-limited processes (possibly at electrodesuspension interface); and to motions of colloidal particles [7]. Moreover, the amplitude of the
spectra in the low frequency range increases with particle size (Figs. 3 and 4) strongly suggesting
that this region contains a geometric signature.
0.3 ..,---.,--..---,---,--.----.--...----.--..---,
r'=(2hl)ra:.('t)CJlu[I+(ro'tfl
't max=251 ns
0't112-61 ns
Go)
"C
:::I
~
't max=57 ns
0't112-10 ns
(\)
't:I
0.2 +--+--I---+---1I---1---!-........+--+--I---I
~ 0.2 +--+---!--+--+--+--1-+--+---!----t
c..
Co
c:{
,...... 0.1 +--+-+---I-----1I---+--+-.+--+-+--I
81
0.0
0.0
-1
-2
-3
-4
-5
-6
-7
-8
-9 -10
-1
-2
-3
-4
.s
-6
-7
-9 -10
Log Relaxation Time (5)
Log Relaxation Time (s)
Figure 3. Relaxation time distribution for the

small particle size Ca-saturated clay.
Figure 4. Relaxation time distribution for

the large particle size Ca-saturated clay.
The spectra also exhibit a well-defined peak between 10-6 and 1O-7 s attributed to diffusion of ions
in the DDL [8]. The relaxation time at the maximum (1:m) and the relaxation time peak half width
are given in the spectra. The mean relaxation time constant is related to the self-diffusion
constant, D, by the equation D =(a + KI)hm [8], where a is the particle diameter and K- I is the
Debye screening length. Diffusion coefficients computed from the relaxation times were 7.1 x
10-5 , 1.5 X 10-5 and 4.8 x 10-7 m 2/s, respectively. These values are 10 to 1000 times greater than
self-diffusion coefficients in water. The increase in mobility is explained, in part, by elimination
of random walk in the vertical direction within the DDL where cations movement is essentially
constrained to a plane (i.e. 3-D to 2-D). The reason for an inverse relationship between the
relaxation time constant and the particle size is not clear.
219
The 1'm decrease with increasing temperature. An "Arrhenius plot" (Fig. 5) was used to determine
the activation energy for the rate process where the effective first-order rate constant, k, is simply
the inverse of the mean relaxation time, 1'm. The slope of the line (Fig. 5) corresponds to an
activation energy barrier (to relaxation of the polarized state) 21.45 kJ/mole or 5.13 kcal/mole,
about activation energy is about that of water [9]. The intercept of the line (Fig. 5) is e23 .2 , or 1.2
x 1010 Sl is the high temperature limited relaxation rate constant. It represents a relaxation time of
84 psec which is about the same as the dielectric relaxation time of room temperature liquid
water, -25 psec, estimated by [10]. The activation energy and high temperature limiting rate
both support the notion that dielectric relaxation of water controls the rate of the processes
limiting conduction. If the 1'm correspond to relaxation of a polarized DDL and the energy barrier
to rearrangement of bulk water surrounding the ions, the polarization-relaxation process takes
place in Gouy-Chapman portion of the DDL and not within the structured water near the clay
surface, i.e. the Stern layer. Further evidence for relaxation of the DDL was obtained by rotation
of the axis of the electrical field. When the electrical field was oriented such that the polarization
occurred along the clay surface (perpendicular) the real portion of the dielectric constant was
greater than when the polarization was induced in along the z axis of the DDL (parallel) (Fig. 6).
Note that the maximum corresponds approximately to 1'max.
1.6
~ 1.5
16.5
In(k)= -2.5BE+03(1fT)
232E+01
16.0
'k 15.5 ~
:i'
:E 15.0
14.5
JJJllllt.-JjjJl . . . . .
()
:sC 1A .._.
." .....
/
~1.3
~ 1.2 =.
~ 1.1 -=-
14.0
0.0028
II.
ii
'"......
... 1 _
&. 0.9 ...'"'
---I-+-WIU!-~~~.w
1E2
0.0034
Figure 5. The relaxation rate constant as a

function of the inverse temperature (K).
1E3
1E4
U)
1E5
1E6
1E7
1EB
(rad/s)
Figure 6. Effect of electrode orientation on

the real and imaginary dielectric constant.
Estimating soil properties such as particle size, particle arrangement, water-film thickness and
surface charge density is predicated on a physically based model of impedance behavior of
variably saturated soils. We selected a physically based model [11], considered among the most
advanced, (not presented here due to space limitations) for interpretation of the frequencydependant impedance behavior. Our goal was to use the model to predict a geometric property,
particle size, and an electrochemical property, surface charge density by fitting the experimental
data. Model parameters were particle size, low frequency permittivity and conductivity, surface
charge density, suspension clay density and mean relaxation frequency that were all
independently determined. The particle size and surface charge density were adjusted by trial and
error while attempting to keep them as close as possible to the known values. The fit best was
obtained with a particle size an order of magnitude greater than the determined value and a
surface charge density six orders of magnitude greater than the true value ..
220
4. SUMMARY AND CONCLUSIONS

Preliminary results from impedance measurements of clay-water suspensions of different particle
size, temperature and measurement orientation reveal the potential of these measurements to
characterize geometrical and electrochemical properties of these surfaces. Much of the
interpretation hinges on proper understanding and quantifying of the physical processes.
Estimation of distribution of relaxation times offers insights on diffusion processes and
geometrical constraints/features. The use of controlled temperature variations added another
dimension to the information and was instrumental in resolving the activation energy which, in
turn, implicated ion motions at distances greater than 3-4 molecular layers of water from the
surface. Shortcomings in the physically based model [11] illustrate the importance of expanding
our understanding in this area. Future studies will focus on varying particle sizes and ionic
compositions of suspensions. We plan use more realistic systems resembling natural porous
media by studying clay-sand mixtures and varying, among other things, the degree of saturation.
Prototype measurement cell for unsaturated mixtures has been developed and is being tested.
REFERENCES
[1]
Olhoeft, G.R 1985. Low-frequency electrical properties. Geophys. 50(12):2492-2503.
[2]
Huang, S. M., A B. Plaskowski, C. G. Xie, and M. S. Beck. 1989. Tomographic imaging of twocomponent flow using capacitance sensors. J. Phys. E. Sci. Instrum. 22:173-177.
[3]
Roberts, J. J., and W. Lin. 1997. Electrical properties of partially saturated Topopah Spring tuff: Water
distribution as a function of saturation. Water Resour. Res. 3(4):577-587.
[4]
Sposito,G. 1981. Suifacechemistryofsoils. Oxford University Press. New York. p.37.
[5J
Mcdonald, J. R, and W. B. Johnson. 1987. Fundamentals of impedance spectroscopy. Impedance

spectroscopy. "edited by J. R Mcdonald", John Wiley and Sons, New York., pp.1-26.
[6J
Bialkowski, S. E. 1991 Expectation-Maximization Algorithm for Regression, Deconvolution, and

Smoothing of Shot-Noise-Limit Data, J. Chemomet. 5:211-225
[7]
Loewenberg, M. and R W. O'Brian, 1992. The dynamic mobility of nonsperical particles. J. Colloid
Interface Sci. 150(1):158-168.
[8]
Garrouch AA, and M.M. Sharma. 1994. The influence of clay content, salinity, stress, and wettability on
the dielectric properties of brine-saturated rocks: 10 kHz to 10 MHZ. Geophys. 59(6):909-917.
[9J
Bockris, J. O'M, E. Gileadi, and K. Muller. 1966. Dielectric relaxation in the electrical double layer. J.
Chern. Phys. 44(4):1445-1456.
[1 OJ
Debye, P.1929. Polar Molecules. Dover Publ., Mineola, New York pp. 77-108.
[11]
de Lima, O. A., and M. M. Sharma. 1992. A generalized Maxwell-Wagner theory for membrane
polarization in shaly sands. Geophys. 57(3):431-440.
Contact point: Lynn Dudley, Department of Plants, Soils, and Biometeorology, 4820 Old Main Hill, Logan, UT
84322-4820, U.S.A Phone: +1 (435) 797-2184; Fax +1 (435) 797-3376; E-mail: ldud@rbt.agsci.usu.edu
221
THERMODIELECTRIC RESPONSE OF VARIABLY SATURATED

POROUS MEDIA: SPECIFIC SURFACE AREA DETERMINATION
Jon M Wraith and Dani Or
Land Resources and Environmental Sciences Dept., Montana State University, P.O. Box
173120, Bozeman, MT 59717-3120, U.S.A. jwraith@montana.edu.
Dept. of Plants, Soils, and Biometeorology, Utah State University, Logan, UT 84322-4820,
U.S.A. dani@mendel.usu.edu.
ABSTRACT. We developed a physical model for the temperature (T) dependency of TDRmeasured bulk dielectric permittivity (Eb) of soils and other porous media. The model considers
modified properties of water near solid surfaces to define a layer of rotationally-hindered water
having temperature dependent thickness. The thermo dielectric sensitivity ofTDR-measured Eb to
bound water reflects interactions with the medium surface area, and permits model inversion to infer
specific surface area (As) from measured response to thermal perturbation. Mixtures of sand and
clays having a range in As and wetness were sealed in coaxial TDR cells, and subjected to discrete
temperature steps. Measured Eb(T) using TDR along with independently measured mass water
contents and bulk densities were used in the model to calculate As. Results were in agreement with
As ofthe mixtures measured using ethylene glycol mono ethyl ether (EGME) monolayer adsorption.
Keywords: temperature, bound water
1. INTRODUCTION
Time domain reflectometry (TDR) is a widely used method to infer water content and bulk electrical
conductivity of soils and other porous media. Inference of soil water content from measured Eb using
TDR travel time is based on the dominance ofthe high dielectric permittivity ofliquid water relative
to that of the solid constituents and air. Recent experimental results [1] showed that Eb of soils
measured using TDR exhibits temperature dependence proportional to soil specific surface area and
wetness. This led to a hypothesis that TDR-measured Eb ofwetted porous media is determined by two
competing phenomena: (i) a reduction of dielectric permittivity of bulk water with increased T; and
(ii) an increase in measured Eb with increased T resulting from release ofbound water. We developed
a physical model [2] which considers this interplay of competing dielectric phenomena.
1.1 A physical model for thermo dielectric behavior of wetted porous media
Development of the physical model is detailed elsewhere [2], so only the main points will be
summarized here. Vicinal water, water very near solid surfaces, has different properties than for bulk
water due to an array of short- and long-range forces acting on it. Viscosity profiles as a function of
distance from solid surfaces and T may be calculated using relationships developed by Low [3,4].
The Debye [5] model for dipole polarization may then be used to predict relaxation time (or
frequency) of vicinal soil water using these viscosity profiles. Debye's model predicts a strong T222
dependence of relaxation time (frequency), and may be rearranged to provide calculated relaxation
frequencies as a function ofT and distance from solids. The thermo dielectric model [2] assumes that
rotationally hindered water layers near solid surfaces experience dielectric relaxation at frequencies
within the TDR bandwidth, thus becoming partially 'invisible' to TDR travel time analysis of Eb That
is, a given amount ofbound water has much lower effect on measured travel time, hence Eb, than does
the same amount of bulk water. Molecular water layers having dielectric relaxation at frequencies
below a certain cutoff frequency are considered as bound, with lower E than for free water. Though
we are incorporating differential E for specific vicinal molecular water layers into the model at
present, the version used here assumed a single value of E for all bound water. Bulk water E decreases
with increasing T [6], while bound water becomes increasingly rotational or 'visible' to TDR with
increased T, and thereby tends to increase measured Eb The relative effects of the competing bound
vs. bulk water responses on measured Eb (i.e., the direction and magnitude of the T influence) is
therefore proportional to wetness and surface area. The bound water content (8 bw, kg/kg) is calculated
as the product ofthe thickness of the 'relaxed' (within the TDR bandwidth) water layer surrounding
solids - a function of temperature only (Fig. 1, [2]) - and As. The dielectric permittivity of bound
water (Ebw) is calculated using the relationships ofBockris et al. [7]. Therefore, by independently
measuring As, bulk density and total 8 we may
calculate the 8 bw and its Ebw as a function ofT only. A
10 .,.----r--,--.,.----,---,--..---.--,
dielectric mixing model is then used to incorporate $
1*=1.0 GHz
the effects ofthe bound water 8 bw(T) into measured E b B 8 +-4+--+--+---+--!--+--+--I
Several assumptions in this model, based on lack of S 6 -I-::-=-~-"o:-f--+--+--I--+---+--i
better information at present, require clarification.
B
4~~~~~~--+_-!-_+--+_-I
In the present study we used the measured ~ 2 L---l:::j::::;~!

thermo dielectric response ofwetted porous media to ~
estimate their specific surface areas, by inversion of
270
290
310
330
350
the physical model described in the previous
Temperature (K)
paragraph. This serves both as a verification for the
model itself, and as an evaluation of a potential new
method for As determination. The specific surface Fig. I. Dielectric relaxation distance from
area of soils and other porous media has critical solid surfaces as function of temperature.
importance to a number of physical and chemical
processes, yet is seldom measured. Provision of a new measurement technique might serve as
impetus to alleviate this situation.
2. METHODS
Mixtures of sand with montmorillonite or kaolinite clays, having variable known As and 8, were
packed in sealed TDR coaxial cells and subjected to step temperature changes in a circulating water
bath. A coaxial cell filled with distilled and deionized CDI) water was included in each replicate
batch to serve as a control. The temperature perturbation ranged between 5 and 55C. Computercontrolled, multiplexed TDR systems (Tektronix 1502B or 1502C) were used to measure Eb.
Measured Eb responses to changing T were then used in the model to predict As of media within the
coaxial cells. The mixture specific surface areas were independently measured using EGME
adsorption.
223

Measured Eb(T) for the sand-clay mixtures were consistent with our previous findings [1], and again
documented the phenomenon of increased dielectric permittivity with increased T due to the bound
water response (e.g., Fig 2). Increased signal attenuation at high T for some high 8 or high As mixtures
led to uncertainty in TDR waveform travel time analysis, due to 'flattening' ofthe second reflection
signifying the end ofthe probe. Mixtures having high total 8 and/or low As showed dominance ofthe
temperature dependent dielectric constant ofwater, while those having low 8 and/or high As followed
the bound water response of increasing measured dielectric permittivity with increasing T (Fig. 3).
Optimization of As to measured Eb(T) .
using the physical model revealed an c 13 . - - - - - - , I - r f - - - - - - - - - - - - ,
apparent lag in the T response. The .;
model was modified to reflect an
12+--+---~---r-----~~--_r----1
'effective' (smaller than measured) o
I.)
';:
temperature, which produced better u
CI)
agreement between measured and a; 11 +---1k::dI~----_r--jr---=---_I_------_I
C
estimated Eb (Fig. 3). We believe this ::!!:
lag in the T response may represent ::::I
m 10+-~~__--~~-~~~~--~~~~_I
an activation energy barrier for 'tI~
rotation of bound water molecules. II)::::I
We are currently evaluating m 9~-~~~~~~~~~~~--~~==
5 20 35 50 30 10 40 45 15 5
modification of the thermo dielectric
Temperature (C)
model to incorporate consideration of
the first few molecular layers ofwater Fig. 2. Measured time series of Eb in response to step changes
(currently treated in aggregate) as in T. Thick line with symbols is sand-clay mixture
discrete layers whose activation response, thin line is DI water.
energy for rotation will be calculated
rigorously using Morse functions [8]. This modeling approach will specify investment of thermal
energy to raise the energy level of each discrete layer of water molecules above the 'rotational
threshold'. In addition, other aspects of water-clay surface interactions such as those resulting from
counter-ions in the diffuse double layer will be considered. Thermodielectric model estimates of As
based on optimization to measured Eb(T) were in agreement with those independently measured using
standard laboratory EGME methodology (Fig. 4).
4. SUMMARY AND CONCLUSIONS
The method described here uses the behavior ofwater to estimate wettable As of porous media based
on the measured thermodielectric response using TDR. Good agreement ofTDR thermo dielectric
As estimates with standard EGME As determination are highly encouraging. Increasing T degrades
the' quality' ofTDR waveforms for travel time analysis, due to increased attenuation. This may limit
the ability to make inferences for high As materials at high T. We found an apparent threshold
temperature effect in measured Eb. The physical model was revised to account for this temperature lag
by using an 'effective temperature', and the underlying mechanism is under consideration. We
anticipate incorporating the 'molecular energy well' concept [8] to describe this phenomenon.
224
Fine sand, 15% mass water
-.
~ 8.8
c
"u 8.6
.;::
-~
.!!!8.4
GI
is
=s
Measured
..
f! 80
<C
G>
u 60
_120
< 100
..
Original
Revised
8.2
SandMontmorilionite Mixtures
140
:::I
CO
(J)
...
0:
Q
I-
...
+---t---:::!:------::;,.jL---t--Ic--T-'bI---I 15% we
III
40 +---t---..!~-+---t-i---1,r-----I 23%
we
20 -r---111~
0
270
280
290
300
310
320
330
!
I
-==--T----+---1I--l-- 32% we
20
40
60
80
100
120
140
EGME Surface Area (m"21g)
Temperature (OK)
SandKaolinite Mixtures
9% Montmorillonite, 15% Mass Water
15 , . - - - - - - - - , - - - - - , - - - - - - - : 7 1
10
1:1
"E
2 9.8
...
we
I
I
l----.---:;;,-<:-------i-------
U)
8% we
o
U9.6
o
c
t>
"*9.4
is
/'
--......
:9.2
- - ... ~
.............
"'"
II
280
290
300
310
320
!
I
0~--_ _~1--__--_+1--__--~
co
9
270
15%
r---..
330
Temperature ("K)
10
15
EGME Surface Area (m A 2Jg)
Fig. 4. Specific surface areas of sand-clay

mixtures measured using EGME and from TDR
13"--~~--r--.---~_--r-~
by inversion of the thermo dielectric model.
e
Calculations indicate that apparent TDR under~c 12+--~r-----t---+--~~-,.r-~
estimate for 2 circled points in upper graph is a
o
".g 11 +----t----;/ttI-'----+-:'""--t---t-----; result of insufficient water to cover all surfaces
with 3 molecular water layers as assumed in the
U
GI
a;
model.
18% montmorillonite, 15% mass water
is 10 +----t~--+-..".-+-----t---+----l
..lo:
:;
CO
9+-__- - t -__-+-__-+-__---t-__-+-__--l
However, the model provided similar As

330 estimates using eitherT or 'effective' T. Further
Temperature (K)
investigations are underway, and we anticipate
Fig. 3. Mean measured Eb(T) (symbols) for three that the thermodielectric measurement approach
sand-clay mixtures. Dominance of bulk water
may provide an effective alternative approach
(top) for low surface areas, and bound water
for As estimation of soils and other wetted (or
(bottom) for high surface areas are illustrated.
wettable) porous media. Finally, these results
Also shown is effective neutralization of the
combined with others not included in this paper
competing responses (middle) for moderate
strongly indicate that the thermodie1ectric
surface areas. Model predictions use actual T
phenomenon may be important to remotely(thin lines) and 'effective' T (thick lines).
sensed land surface wetness measurements. In
addition to providing required thermal
270
280
290
300
310
320
225
corrections to remotely sensed data, this may lead to an ability to map surface soil textures (strongly
related to As) from existing or new satellite platforms.
5. ACKNOWLEDGMENTS
Partial funding for this work was provided by NRI Competitive Grants ProgramJUSDA (Agreement
no. 95-37102-2175), and the Israel-US Binational Agricultural Research and Development (BARD)
Fund (Project IS-2839-97).
6. REFERENCES
[1]
Wraith, lM., and Or, D., Temperature effects on soil bulk dielectric permittivity measured by time domain
reflectometry: Experimental evidence and hypothesis development. Water Resour. Res. 1999.35,361-369.
[2]
Or, D., and Wraith, lM., Temperature effects on soil bulk dielectric permittivity measured by time domain
reflectometry: A physical model. Water Resour. Res. 1999.35,371-383.
[3]
Low, P.F., Viscosity ofinterlayer water in montmorillonite. Soil Sci. Soc. Am. J. 1976,40,500-505.
[4]
Low, P.F., Nature and properties of water in montmorillonite-water systems. Soil Sci. Soc. Am. J. 1979,43,
651-658.
[5]
Debye, P., Polar Molecules, Mineola, New York, Dover Pub., 1929,77-108.
[6]
Weast, R.C., CRC Handbook of Chemistry and PhysiCS, 67th ed., Boca Raton, FL, CRC Press, 1986.
[7]
Bockris, J.O'M., Devanathan, M.A.V., and Muller, K., On the structure of charged interfaces. Proc. Roy.
Soc. London, Ser. A, 1963,274,55-79.
[8]
Bockris J. O'M., Gileadi, E., and Muller, K., Dielectric relaxation in the electric double layer. l Chern. Phys.
1966,44,1445-1456.
Contact point: Jon M. Wraith, Land Resources and Environmental Sciences Dept., Montana State University, P.O.
Box 173120, Bozeman, MT 59717-3120, U.S.A. Phone: +1 (406) 994-1997; Fax +1 (406) 994-3933; E-mail:
jwraith@montana.edu
226
WATER CONTENT AND PARTICLE SHAPE EFFECTS ON THE

DIELECTRIC PERMITTIVITY OF ANISOTROPIC POROUS MEDIA
Scott B. Jones and Shmulik P. Friedman
Inst. Soil, Water, and Environmental Sciences, The Volcani Center - ARO,
P.O. Box 6, Bet Dagan 50-250, ISRAEL. vwsfried@volcani.agri.gov.il
ABSTRA CT. The anisotropy of porous media affecting its directional permittivity is caused by
extreme aspect ratio particles (e.g., disks, needles) aligned with respect to a bedding plane in rock
or soil. Our objectives were to determine the extent of particle shape and layering effects on the
dielectric permittivity of porous media. Permittivity measurements in a packing of disk-shaped
particles (mica) using time domain reflectometry (TDR) showed alteration ofthe permittivity due to
the particle shape effects in accordance with predictions of a Maxwell-Garnett based mixing model.
Keywords - particle-shape, anisotropy. three-phase mixture

1. INTRODUCTION
Electromagnetic determination ofvolumetric water content of porous materials such as soil depends
upon its constituent components consisting of solids of various size, shape and mineralogy and of
pores filled by air, water, or NAPL. Many natural soils contain some degree of anisotropy which is
due both to soil structure and texture. Particle shape effects have been observed using
electromagnetic measurements techniques within various materials. Kraszewski and Nelsonl
observed shape effects of peanuts on microwave measurements in microwave resonant cavities. The
"platy" particle effect due to oblate shaped clayey particles has been addressed by many researchers
measuring the dielectric permittivity of saturated rock2,3. Although shape effects have been observed
and mathematical theories exist for modeling their effects little work has been done to verify the
orientational and particle shape effects on permittivity measurements in soils. Furthermore,
confounding effects of bound water, and interfacial polarization phenomena have not been separated
from shape effects, which would be of great benefit in understanding the role each factor plays.
2. DIELECTRIC MIXING MODELS
The particle shape effect on the permittivity of two-phase mixtures can be viewed by considering
ellipsoidal inclusions of permittivity El lying in a background of permittivity Eo. We also consider a
three-phase confocal ellipsoid - background mixture with a 2-phase inclusion composed of
permittivity El and E2 again lying in a background of Eo. For the purpose of simplification, all
inclusions are ellipsoids of revolution having a unique axial dimension "a" with two common axial
dimensions band c (i.e., b = c). Therefore we refer to the effective permittivity with respect to the
electrical field alignment parallel to the "a" axis as Eef/ and correspondingly, Eeffb refers to an
electrical field parallel to either b or c. The largest dimension, a or b should be smaller than the freespace wavelength to avoid scattering effects. The depolarization factor as a function of aspect ratio,
a:b, for the axis of rotation is given by, N 3 = [1+1.6(a:b)+0.4(a:b)2]"1 (r = 0.999) from which N b=
0.5(1-N3). A family of mixing models describe the effective permittivity ofa two-phase anisotropic
227
mixture of ellipsoidal inclusions of volume fraction, 4> l' The resulting effective permittivity of the
mixture in the ith direction is strongly dependent on the particle's aspect ratio and alignment with
respect to the electric field. It can be written in a general form as4
i
&eff
1{&1 - &0)[&0 + V{&eff - &0)]

i
&0 ) + N [ (&1 - &0)(1- 1) ]
= &0 + &0 + V (&eff -
(1)
where v is a parameter which gives the MG5 mixing rule when equal to 0, the PVS 6, or effective
medium approximation mixing formula when v= 1 - N, and the CP ("coherent potential") mixing
rule when v = 1. These three theories represent different possible approaches for answering the
question of how to represent the background permittivity as "felt" by the inclusions within the mixing
scheme. The MG formula adopted here refers to a dilute inclusion concentration, but has also been
successfully applied in the past to mixtures having a relatively high inclusion concentration' .
Modeling the dielectric permittivity ofthree-phase systems is more difficult because ofthe complexity
of geometry, phase configuration and distinguishing between host and inclusion. One approach is
to allow for multiple independent inclusion geometries, each of different permittivities, contained in
a background8 , Another approach uses concentric spheres or confocal ellipses as inclusions, which
represent the different medium phases-t9.10,ll. For the three-phase mixtures, we have selected a
mixing scheme based on the MG theory for two confocal inclusion phases (El and E2) contained in
a background representing the third phase (Eo). The effective permittivity of a three-phase confocal
system of ellipsoids aligned parallel to the ith axis (i.e., Ec~) is presented as 12
(2)
(3)
and 4>2 are the volumetric fractions and N 1i and N 2i are the depolarization factors of the outer and
inner ellipsoids, respectively.
4>1
Among the six potential phase configurations, Friedman 11 found SWA (solid-water confocal
ellipsoids lying in a background air phase) and ASW to more closely describe the permittivity of
natural unsaturated soils, where SWA tends to underestimate and ASW gives an overestimate, but
228
both maintain the solid-water interface characteristic of water-wetted minerals. Among various
methods of mixing the dual SWA-ASW system for best describing a moist soil, Friedmanll used the
symmetric effective medium approximation (EMA). Each configuration permittivity is weighted
according to the saturation degree, S, a function of moisture content, 8, and porosity, n, as
fW SWA
= 1- s
fW ASW
s=-Bn
=S
(4)
where at low 8, SWAis more heavily weighted and at high 8, ASW has more weight. The resulting
quadratic expression is given asH
_(C16 +
8SWA8ASW)"2 _
8 eff -
C = fwSWA{8ASW
- 28SWA
C
4
)+ f/sW{8sWA -
(5)
28ASW )
We will refer here only to packings of perfectly aligned ellipsoidal particles, forming a uniaxial
anisotropic medium (effb = effC), and will calculate only ef/ and effb using eqs. (2) and (3).
3. EXPERIMENTAL RESULTS AND DISCUSSION
Because high-aspect ratio disk shaped particles produce the most significant alteration ofpermittivity
compared to that of a sphere, mica particles of relatively narrow particle size distribution and high
aspect ratio (about 0.1) were selected for this study. Phlogopite mica (Suzorite Mica Products,
Boucherville, Quebec, Canada), previously manufacturer-sieved, was rinsed in deionized water and
oven dried. Mono-sized glass bead diameters varied from 0.025 to 0.5 mm and binary mixtures were
also used to obtain lower porosities. The three different sizes were loosely deposited into separate
5L beakers filled with saline solutions of 5.2
(0.25 mm), 2.1 (0.5 mm) and 5.1 (1 mm) dS
80~==~~--------------------~
m- 1. Three-wire waveguides made of 4 mm
1 mm
diameter rods of length 100 mm and spaced
:t!' 60
,.
at 50 mm were prepositioned, two in each 's
0.5
E
beaker, one with rods parallel to the expected 'E
horizontal bedding plane of the mica and the ~ 40 0.25

other with rods perpendicular to it. The mica ~
0.50.025
was gradually desaturated by applying ~
suction using a water-filled cylindrical lD 20
ceramic cup located in the bottom of the
container. Daily permittivity measurements
were recorded using a Tektronix 1502C
o
0.2
0.8
0.4
0.6
cable tester and WINTDR98 waveform
Porosity
analysis software13 .
Figure 1. Dielectric permittivity measured in waterThe measured eff of the different glass bead saturated mica and glass beads ~? modeled using a
packings are in agreement with the predicted the 2-p~ase ~axwell-Garne~t mIX1~g t~eory for eff
permittivity of the MG model for an aspect and eff' PartIcle aspect ratIO of mIca IS 0.085 and
as shown for model predictions.
229
ratio of 1 as expected (see Figure 1). More interesting are the mica permittivity data which refer to
different porosities resulting from a suction-induced shrinkage process (mostly vertical and prior to
desaturation). The measurements are generally correlated to the MG theory for E effa and Eeffb with
an exception of the 0.5 mm measurements of Eeffb which lie significantly below the MG predictions.
We expect partial disorder of the particle alignment to shift measured values to something between
model predictions for Eeffa and Eeffb. Some particle rearrangement is likely to occur near rod surfaces,
depending on the amount of movement of the mica particles relative to the probes. The differences
between modeled permittivities for E effa and Eeffb are greatest between porosities of 0.5 to 0.7 which
happens to be in the range of porosities of clayey soils. Salinity of the water phase in each medium
may also have influenced permittivity readings. However, interfacial polarization which occurs at
low frequencies is not a likely factor owing to the relatively high effective frequency of the TDR
device (1 GHz14). Maxwell-Wagner effects, which occur at frequencies between 0.1 and 500 MHzlS
have been shown to increase permittivity readings, especially in the presence of platy particles2,16.
Here we do not see evidence of increased permittivities and therefore feel Maxwell-Wagner effects
have a minimal influence on permittivity readings compared to the particle shape effects which are
predicted by the dielectric mixing theory.
For testing the experimental setup, we also measured the Eeff in glass beads, where E effa and Eeffb are
expected to be the same for the entire range of volumetric water content(Figure 2, inset). Measured
permittivities in three different size fractions
of mica particles are generally correlated to
70 "'I''''-r:-=-=-=-=-_-_-_-_-_-_-_-_---,---------.
a
b
electrical field alignment (i.e., E eff and E eff )
60" ~ont'l
as determined by probe positioning (Figure
20
~ertiC.l
2). Substantial deviations in permittivity data
50 15 ?
are attributable to the hydrostatic distribution ' 40
~~;::
oflower moisture contents in the upper part ~
of the probe's volume of influence which is ~ 30
highly correlated to particle size (i.e., most ~ 20
significant for Eeff3 of 1 mm particles). Model w
predictions are computed using a mica
10
permittivity of 6.35, water permittivity of
0 ~~~~~~:::::~~:=~~-+-U
78.5 (25C), and 1 for the permittivity of
0
air. The mica was found to have an average
0.1 0.2vo~~!e~~4wa~~; co~!ntO.7 0.8 0.9
aspect ratio of 0.085 and porosities of 0.75 .
.. . .
.
to 0.85 were used in calculations. FIgure 2. EffectIve permittivIty of three dIfferent
Permittivity data generally lie between the mica flake dia.mete~s plotted as a function of water
E a and E b model predictions with E a content. PartIcle SIzes of 0.25, 0.5, and 1 mm are
eff
eff
eff
db'
I
. I
d
data better described by the EMA modeling. represe~te y t:Iang es, Circ es an squares,
Some mis-alignment of the mica particles :es?ectively. ~Igment dependent meas~rements are
with respect to the bedding plane caused mdicated by Eeff (filled symb.ol~) and Eeff (empty
some gravitation of permittivity readings symb~ls). EMA. model predIctI?nS for both
towards a central value between the electncal field alIgnment scenanos are shown for
predicted permittivities for the two principal porosities of 0.75 (~olid ~ine) and 0.8.5 ~~ashed line).
directions. Data overall lie well below Effects of probe onentatlOn on permIttIVIty
Topp's17 curve except for E b values near measurements is demonstrated in glass beads using
eff
one horizontal and one vertical probe (inset figure).
10
10
230
saturation. Assumptions and limitations of the proposed mixing models merit further discussion.
The use of a single aspect ratio ellipsoidal geometry to model soil particles is oflimited practicality.
Additionally, the elliptical water shells are not necessarily consistent with the water-air interface
behavior. Modeling the phases of solid, water, and air using a confocal arrangement while
mathematically efficient, does not ideally mimic the complex and variable configuration of partially
saturated soils. Some of these discrepancies are partially corrected by the EMA mixing of the SWA
and ASW phase configurations.
REFERENCES
[1]
[2]
[3]
[4]
[5]
[6]
[7]
[8]
[9]
[10]
[11]
[12]
[13]
[14]
[15]
[16]
[17]
Kraszewski, AW. and S.O. Nelson. 1996. Moisture content detennination in single kernels and seeds with
microwave resonant sensors. p 177-203. In: A Kraszewski (ed.). Microwave Aquametry, IEEE Press, NY.
Sen, P.N. 1981. Relation of certain geometrical features to the dielectric anomaly of rocks. Geophysics.
46(12): 1714-1720.
Mendelson, K.S. and M.H. Cohen. 1982. The effect of grain anisotropy on the electrical properties of
sedimentary rocks. Geophysics. 47(2):257-263.
Sihvola, A and I.V. Lindell. 1989. Polarizability and effective pennittivity of layered and continuously
inhomogeneous dielectric spheres. 1 Electro. Waves Applic. 3(1):37-60.
Ma"\.'Well-Garnett, J.C. 1904. Colours in metal glasses and in metallic films. Trans. Royal Soc. (London).
CCIII:385-420.
.
Polder, D. and J.H. vanSanten. 19~6. The effective penneability of mixtures of solids. Physica.
XII(5):257-271.
Sihvola, A.H. and J.A Kong. 1988. Effective pennittivity of dielectric mixtures. IEEE Trans. Geosci. Remote
Sens.24(4):420-429.
deLoor, G.P. 1953. Method of obtaining infonnation on the internal dielectric constant of mi:\wres. Appl.
Sci. Res. 3(B):479-482.
Tinga, W.R, WAG. Voss, and D.F. Blossey. 1973. Generalized approach to multiphase dielectric mixture
theory. 1 Appl. Phys. 44(9):3897-3902.
Sihvola, A and I.v. Lindell. 1990. Po!arizability and effective pennittivity of layered and continuously
inhomogeneous dielectric ellipsoids. 1 Electro. Waves Applic. 4(1):1-26.
Friedman, S.P. 1998. A saturation degree-dependent composite spheres model for describing the effective
dielectric constant of unsaturated porous media. Water Resour. Res. 34(11):2949-2961.
Sihvola, A 1996. Dielectric mixture theories in pennittivity prediction: Effects of water on macroscopic
parameters. p 111-122. In: A Kraszewski (ed.). Microwave Aquametry, IEEE Press,NY.
Or, D., B. Fisher, R.A Hubscher, and I.M. Wraith. 1998. WinTDR98 V4.0 - Users Guide (Windows-Based
Time Domain Reflectometry Program for Measurement ofSoi! Water Content and Electrical Conductivity).
Utah Agric. E:\-p. Sta. Res. Rpt.
Heimovaara, T.l., E.J. G. deWinter, W.K.P. vanLoon, and D. C. Esveld. 1996. Frequency-dependent dielectric
pennittivity from 0 to 1 GHz: Time domain reflectometry measurements compared with frequency domain
network analyzer measurements. Water Resour. Res. 32(12):3603-3610.
Hilhorst, M.A 1998. Dielectric Characterization of Soil. Doctoral Thesis submitted in partial fulfillment of
the requirements for the degree of Ph.D .. Wageningen Agricultural University, Wageningen.
Hilfer, R. 1991. Geometric and dielectric characterization of porous media. Physical Review B. 44( 1) :60-75.
Topp, G.C., lL. Davis, and AP. Annan. 1982. Electromagnetic detennination of soil water content using
TDR: II. Evaluation of installation and configuration of parallel transmission lines. Soil Sci. Soc. Am. J.
46:678-684.
Acknowledgments: This research was supported by Postdoctoral Award No. FU-267 -97 from Bard, The United States
- Israel Binational Agricultural Research and Development Fund.
Contact Point: Scott B. Jones, Dept. Plants, Soils. and Biometeorology; Utah State University; Logan, UT 84322-4820;
USA Phone: (435) 797-2175; Fa"{:(435) 797-2117; Email: sjones@.mendel.usu.edu.
231
Session on
PHYSICS OF WATER AND MIXTURES
Chair: Dr. Michael Kent,

K&S Associates, Beggar, Scotland.

2:00 - 4:00 PM
232
233
DIELECTRIC PROPERTIES OF LIQUID WATER AND SALINE SOLUTIONS

BELOW OC AS MEASURED BY
A RESONANT APPLICATOR COMPENSATION METHOD
Per O. Risman, Microtrans AB, Sweden

ABSTRACT. Purified water in sealed glass capillary tube can rather easily be supercooled
down to about - 20C and then measured by the common TMolO transmission resonant
cavity technique. However, accurate measurements of the absolute kind (where the
dielectric data are obtained by only geometric and microwave resonance data) require very
well-controlled sample and tube dimensions and complicated correction for sample
insertion holes, coupling port properties, and losses in these and the whole cavity.
Instead, a compensation method based on the known variation of the dielectric
properties of water above DoC and up to about 60C - taken from the literature - has
been used. The resonant frequencies, Q-factor values and transmission at resonance at a
number of temperatures above DoC are then used to quantify the imperfections of the
measuring system, and to calibrate it. A "virtual" sample diameter is also obtained using
the known data for water and no-load measurements of the cavity with and without the
empty capillary tubes before these are filled with water and sealed. These no-load
measurements provide also the temperature dependence calibration of the cavity data.
Similar measurements are also made below DoC, to provide compensation data for the
actual sample measurements.
The reason why a resonant cavity method is used is its very good resolution and a
possibility of achieving a high accuracy. However, the data are obtained at only one
frequency. Therefore, two different cavities with resonant frequencies about 2800 and 950
MHz were used. Quantification of the validity of the simple Debye relaxation concept with
the dispersion factor (l' could therefore be made in this frequency range used in practice
for heating applications. Also NaCl solutions in three different concentrations were
measured down to - 20C. One of theses was the eutectic solution, which has a normal
freezing point below that temperature.
Literature data on the microwave permittivity of supercooled water is scare. The
obtained results are quite plausible, in view of the rather poorly specified accuracies of the
literature data. The relative permittivity and relaxation frequency of pure supercooled
water was found to fit the following functions very well:
f D = 594exp[ -455/(T -165)]

where t is the temperature in C and T in K, and f D is the relaxation frequency in MHz.
Interestingly, the conductivity contribution by the dissolved NaCl to the effective
f" seems to vanish just at the eutectic temperature about - 21C.
The temperature at
which fD according to the above function vanishes is about -108C; this may have some
interesting implications. No dispersion of f D has been found within the accuracy of
measurement. The accuracies will be discussed in some detaiL
fs
= 88.1- OA03t + 0.0009t2
foo
= 4.5
Contact point: Dr. Per O. Risman, Microtrans AB, P.O. Box 7, S-43821 Landvetter, Sweden.
Phone: +46-301-31100; Fax: +46-301-30102; E-mail: por@por.se
234
STUDY OF DIELECTRIC BEHAVIOUR OF HUMAN BILE AT

MICROWAVE FREQUENCIES
S.Biju Kumar, URaveendranath, Philip Augustine * and K.TMathew
Department of Electronics
Cochin University of Science and Technology
Cochin-682 022, Kerala, India
mathew@doe.cusat.edu
*Director
Digestive Disease Centre
P.V.S.Hospital
Cochin-682 017, Kerala, India
ABSTRACT
Dielectric behaviour of human bile under normal and infected conditions was studied at
Microwave frequencies. The dielectric parameters have been measured using the rectangular
cavity perturbation technique. It is observed that the complex permittivity of bile differs by a
good margin under normal and infected conditions.
Key words: Cavity Perturbation, Complex Permittivity, biological samples
1. INTRODUCTION
Microwave is an efficient tool for material characterisation. This capability of microwave is
utilised for the detection of disease. All biological tissues and fluids are dielectrics and their
properties change when they get infected. Investigations on the dielectric properties of certain
biological materials were reported by Cook [1] and Schwan [2].
In our investigation complex permittivity of three samples of bile was measured. The donor of
normal sample was ,a healthy person. The other two samples were collected from patients
suffering from Primary Sclerosing Cholangitis (PSC) (inflammatory bowel diseases by which
bile ducts get narrowed and irregular) and Biliary Obstruction (BO) (abdominal pain and
jaundice due to bile duct stone) at the time ofERCP surgery. The dielectric properties of these
samples were studied and the complex permittivity and conductivity were measured at S and C
bands. All the measurements were made at 27 C.
2. EXPERIMENTAL SET UP AND MEASUREMENT
The method of measurement is the same as given in [3]. The measurement is done with HP
8510C Vector Network Analyser, using cavity perturbation technique. The sample is taken in a
capillary tube of small dimension and both ends are sealed. The tube is placed at the position of
maximum electric field of the cavity resonator. In each case resonance frequency and Q-factor
of the cavity resonator with and without sample are measured. Since the biological samples are
high loss materials, volume of the sample taken in the capillary tube should be as small as
235
possible. Otherwise resonant peak becomes very flat and it affects the accuracy of the
measurement. Transmission type rectangular waveguide cavities are used for the study. The
cavity is excited in TE IOn mode. Block diagram of the experimental set up is shown in Fig. I.
HP 83651 B
SWEEP OSCILLATOR
HP 85lO C
NETWORK ANALYSER
HP 8514 B
S-PARAMETER TEST SET
HP 9000/300 SERIES
INSTRUMENTATION
COMPUTER
J
RECTANGULAR CAVITY
RESONATOR
TEMPERATURE
CONTROLLER
Fig.I. Block diagram of the Experimental Set up

3. THEORY
If ft and fs are the resonance frequencies and Qt and Qs the corresponding Quality factors of the
cavity resonator without and with the sample in the tube respectively the real and imaginary parts
of the complex permittivity, and conductivity are given by [3] as
'Cr/ -1 =
Ell
r
O'e
(f -f) V
t
s c
2f
s Vs
= ~[_1
4 Vs Q
s
= 00/1 =
__1]
Qt
21tfE o/1
r
where Vcand Vs are the volumes of the cavity resonator and sample respectively.
236
(1)
(2)
(3)
4. RESULTS
The complex permittivity and conductivity of human bile under normal and diseased conditions
are measured. Normal bile has low dielectric constant and conductivity. When bile gets affected
by diseases dielectric constant becomes high, showing an increase in water content. Variation of
real and imaginary parts of complex permittivity and conductivity with frequency for different
samples are shown in Fig. 2(a), 2(b) and 3 respectively.
75 ~~~~~~~~~~~~~======~~~~
73
Normal
71 _
PSC
69
BO
'\
.?!-67
.s;:
::::65
--
'63
e::o 61
CD
'.- .. ~ . '. .......... -........
16 59
...... -........ ---_ ... --_ .
0-
co 57
CD
a: 55
53
51 2
Frequency (GHz)
Fig.2(a) Variation of Real Part of Complex Permittivity of bile with frequency

31
29
27
Normal
PSC
eo
.... ,1"""---._---- -.-
8
Frequency (GHz)
Fig.2(b) Variation of Imaginary Part of Complex Permittivity of bile with frequency
237
High dielectric loss indicates that bile contains more water showing mal functioning of liver.
Large variation in the complex permittivity is observed for infected bile. For normal bile
frequency dependency is small compared to diseased bile. An irregular variation in the real part
of complex permittivity at low frequency region is due to the fact that the resonance peaks are
13
,_..
12
__
......._._._-_.-._---_..
:~c
10
--e-n
:::s
c::
::-
os;
:;:::;
-0
(.)
./
Normal
11
_-_...__._..,
I
./
./
./
./'
./
.~--~~.------j
./
./
3 _....
. .
2
2
Frequency (GHz)
Fig. 3. Variation of conductivity of bile with frequency under normal and infected conditions
very flat at these frequencies and m~asurements may not be very accurate. However the range of
fluctuation is not very large.
REFERENCES
[1]
H.F.Cook, "The dielectric behaviour of some types of human tissues at microwave frequencies",
Brit.J.Phys., 1951, vo1.2, pp.295-304.
[2]
H.P.Schwan, "Electrical Properties of tissue and cell suspensions", Advances Biological and Medical
Physics, New York, Academic, 1957, vo1.5.
[3]
K.T.Mathew and U.Raveendranath, "Wave guide Cavity Perturbation Method for measuring complex
Permittivity of Water", Microwave and Optical Technology Letters, 1993, vo1.6, pp.104-106.
Contact Point: Dr.K.T.Mathew, Professor, Department of Electronics, Cochin University of Science and
Technology, Cochin - 682 022, Kerala, India. Phone: +91 484 532161; Fax: +91 0484 532800; E-mail:
mathew@doe.cusat.edu
238
MODELLING THE DIELECTRIC CONSTANT OF RICE GRAINS

Kailash Prasad Thakur
Microwave Engineering Group, Imaging and Sensing Team, Industrial Research Limited,
24 Balfour Road, PO Box 2225, Parnell, Auckland, New Zealand, k.thakur@irl.cri.nz
ABSTRACT. A new technique for the measurement of dielectric constant of materials, using a
partially filled waveguide cell and a new mixture equation, has been developed which is first
tested for teflon and then used to measure the dielectric constant of rice and air mixture in the
frequency range of 2.6GHz to 3.9GHz. The values of dielectric constant of rice grain obtained
using the mixture model varied between 4.5 and 7.0. The dielectric constant of rice grains
generated by the new model compares well with the present experimental data and published
data of You and Nelson. The moisture content of rice was also measured by slow drying of
three different samples of rice in an oven which was found to be 11.6%. The packing fraction of
rice grains in the waveguide cell is 0.641, which was further increased to 0.81 by using a mixture
of rice grains (80% by weight) and rice powder (20% by weight).
Keywords: rice, moisture, permittivity, model
1. INTRODUCTION
Methods for measuring the permittivity of material using a waveguide cell is well known and
described extensively[I-2].This paper presents a new technique for the measurement of dielectric
constant of material using a partially filled waveguide cell and a new mixture equation [3].
2. MATERIALS AND METHOD
When the waveguide cell is partly full of a rectangular sample of teflon such that the
crossectional area of the waveguide is completely occupied by the sample and an air gap exists
along the direction of propagation as shown in Figure 1, the measured value of the dielectric
constant is lower than that of teflon and greater than that of air. This situation can be considered
as the mixture of two dielectrics (teflon and air) and the dielectric mixture model can be applied
to estimate the dielectric constant of the sample. We have applied a linear mixture model
(1)
where Sa = 1, the dielectric constant of air, VI (= tid) is the fraction of material (v/v) in the
waveguide cell, t is the thickness of the sample and d is the length of the waveguide cell. The
waveguide cell having dimensions; a (width), b (height) and d (length) is assumed to carry TEIO
mode when completely or partially filled with the material under test which has a complex
relative permittivity Sm = s'm - j slim. Considering the material to be non-magnetic and neglecting
the ohmic losses in the waveguide walls, the propagation constant in the guide is given by [4]
239
(2)
where
OJ
is the angular frequency in rad/s.

t
Figure 1: Partly filled waveguide cell is considered as a mixture of two dielectrics in the cell
The characteristic impedance of the waveguide is
(3)
where flo is the permeability of free space.

p=
The voltage reflection coefficient p is defined as

ZTE(Sm) - ZTE(sa)
ZTE (s m) + ZTE ( sa)
(4)
where ZTE(8J is impedance of the waveguide filled with air. The transmission coefficient is
given by
(5)
From equations (4) and (5), the following expressions for S-parameters can be derived
(6)
and
(7)
From the measured values of S-parameters, the values of permittivity of the sample can be
obtained numerically using equations (3) to (7). The computed values of 8 'm for t = 1.5 em and t
= 2.1 em were obtained. The results are shown in Figure 2 together with the experimental values
for three samples of teflon of different thickness. It is interesting to note from figure 2 that the
results obtained from the mixture model for t = 1.5 em and t = 2.1 em compare well with the
240
measured values of the dielectric constant. The exact match between the computed and
experimental values is obtained at frequency 2.9 GHz. The discrepancy between the computed
and experimental results could be attributed to the mixture model (1) used in this computation.
If we use an improved dielectric mixture model, it will be possible to remove the discrepancy at
all frequencies. A generalised mixture model has been proposed recently [3] which is given by
(8)
The mixture model (8) together with equations (2) to (7) generates an exact match between
experimental and theoretical values of the dielectric constant of teflon that removes the
discrepancies shown in Figure 2. This has further established the applicability of the mixture
model (8) to solid teflon and gave us encouragement to use the technique for the measurement of
the dielectric constant of rice in the same waveguide cell.
2.20 ~~_________________________ b
2.00
1.80
1.60
.....
~~----~=:========~====::::
1.40
1.20
+-___________________________
1.00
O.BO
2.6
2.B
3~
3.2
3A
3~
3~
4~
Frequency. GHz
Figure 2: Dielectric constant of Teflon of different thickness as measured in waveguide cell (Sband) , a, air in the cell; b, teflon (29.8mm) completely filled in the waveguide cell; c,
teflon (21 mm) partly filled in the waveguide cell; d, teflon (15 mm) partly filled in the
waveguide cell; e, model value for teflon (21 mm) partly filled in the waveguide cell;
f, model value for teflon (15 mm) partly filled in the waveguide cell.
2.1. Long Grain Rice

A sample of long grain rice was used for this experiment. Rice has an ellipsoidal shape.
Average dimension and other physical properties of the rice is shown in Table 1. The moisture
content in rice was measured by drying three samples of 100 gm each of rice at 100C over a long
period and periodically measuring the weight loss until the weight loss became negligibly small.
The length of the waveguide cell was kept vertical and various quantities of rice grains were
loaded into the waveguide and the value of S-parameters were recorded in the frequency range
2.6GHz to 3.96GHz. The values of the dielectric constant of the mixture (air and rice) was
obtained from the S-parameters using the algorithms developed by Nicolson and Ross [5] and
Weir [6]. The results are presented in Figure 3.
241
Table1: Physical property of the long grain rice

Property
length
thickness
thickness
volume
mass
density
packing fraction
moisture content
value
0.652
0.213
0.188
0.0132
0.0178
1.347
64.1%
11.6%
unit
cm
cm
cm
cc
gm
gmlcc
It is interesting to note in Figure 3 that the value of dielectric constant of rice when the cell is
completely full of rice is between 3.33 at 3.95GHz and 3.48 at 2.6GHz. However, it must be
remembered that the volume occupied by rice at this stage is 64.1 % (equal to the packing
fraction of rice). The rest of the volume (= 35.9 %) is still occupied by air.
___ 3.50
................ __ ........... .
3.00
~2'50
2.00
1.50
1.00
,,,
,
0.50
__ 1
Frequency. GHz
Figure 3: Dielectric constant of rice and air mixture as a function of volume fraction of rice (VI)
at different frequencies.
8.50
8.00
7.50
7.00
~
6.50
.",
6.00
5.50
5.00
4.50
4.00
2.6
2.8
3.0
3.2
3.4
3.6
3.8
Frequency, GHz
Figure 4: Variation of dielectric constant for rice obtained from the dielectric mixture model (8).
242
o 3.84 GHz, Experiment

-Model
o 11GHz, You & Nelson (1988)
-w
H
H
1D-""=--f-----+---+---+---1
o
0.2
0.4
0.6
0.8
Figure 5: Dielectric constant for rice and air mixture as a function of concentration of rice (v/v)
In order to get the dielectric constant of rice we need to extrapolate the surface in figure 1 to VI =
1. This extrapolation can be carried out either statistically or by using the dielectric mixture
equation. We have carried out the extrapolation of the surface in Figure 1 to VI = 1 using the
dielectric mixture model (8). The result of this extrapolation is shown in Figure 4. For air and
rice mixture at 3.84 GHz, we have a = -0.38005, s/ = 7.867685 and sa' = 1.0. The variation
of dielectric constant of rice and air mixture as a function of concentration of rice (v/v) is shown
in Figure 5 for f = 3.84Ghz. Figure 5 also show the results of You and Nelson [7] measured at
llGHz. There is a good agreement between the present results and that of Nelson et. al. [7-9]
showing the validity of the experimental technique and model used in the present investigation.
To conclude the paper what we have got here is a new technique of using a waveguide cell for
the measurement of dielectric constant of materials which have packing fraction less than one.
Acknowlwdgment:
The author thanks the Foundation for Research Science and Technology, New Zealand for
funding this research through their contract #C08606.
References:
[1]
[2]
[3]
[4]
[5]
[6]
[7]
[8]
[9]
Kraszewski, A. W. and Nelson, S. O. Moisture content detennination in single kernels and seeds with
microwave resonance sensors. In Microwave Aquametry - Electromagnetic Wave Interaction with WaterContaining Materials, "edited by A. W. Kraszewski", New York, The Intsitute of Electrical and Electronic
Engineers, 1996, 177-203.
Green, A. Measurements of the dielectric properties of cheddar cheese. J. Microwave Power and
Electromagnetic Energy, 1997, 32 (1), 16-27.
Thakur, K. P., Cresswell, K., Bogosanovich, M and Holmes, W. Modelling the pennittivity of liquid mixtures.
J. Microwave Power and Electromagnetic Energy, 1999. in press.
Baker-Jarvis, J., Vanzura, E. J. and Kissick, W. A. 1990. Improved Technique for Detennining Complex
Pennittivity with TransmissionlReflection Method, IEEE Trans. Microwave Theory & Tech. 38: 1096-1103.
Nicolson, A. M. and Ross, G. F. Measurement of the intrinsic propereties of metrials by time domain
techniques, IEEE Trans. Instrurn. Meas. 1968. 17, 395-402.
Weir, W. B. Automatic measurements of complex dielectric constant and penneability at microwave
frequencies, Proc. IEEE, 1974.62,33-36.
You, T. S. and Nelson, S. O. Microwave dielectric properties of rice kernels. 1. of Microwave Power, 1988,
23 (3), 150-59.
Noh, S. H. and Nelson, S. O. Dielectric Properties of Rice at frequencies from 50 Hz to 12 GHz. Transection
of the ASAE, 1989, 32 (3), 991-998.
Nelson, S. O. and Noh, S. H. Mathematical models for the Dielectric constant of Rice. Transection of the
ASAE, 1992, 35 (5), 1533-36.
243
FREQUENCY AND TIME DOMAIN METHODS FOR

SNOW MOISTURE AND DENSITY DETERMINATION
Alexander Brandelik and ChristolHuebner
Karlsruhe Research Center / University of Karlsruhe

Institute for Meteorology and Climate Research
Postbox 3640, D-76021 Karlsruhe, Germany
alexander.brandelik@imk.fzk.de
ABSTRACT. Water content and density of a natural snow covers are measured with a new
sensor based on a three wire flat band cable. The cable is enclosed by snow fall and changes its
electromagnetic characteristics depending on the dielectric properties of the surrounding snow.
In the dry snow phase the pulse travel time along the cable is a function of the density. In the wet
snow phase high and low frequency measurements are required to determine water content and
density at the same time. A new algorithm detects inevitable air gaps around the sensor cable and
corrects the measurement results. Continuous monitoring of the snow properties has been carried
out on a glacier in Switzerland. With time domain reflectometry it is possible to investigate
natural snow inhomogenities like ice lenses and percolation zones and observe their temporal
behavior.
Keywords: snow moisture and density measurement, air gap correction, snow inhomogenities
1. INTRODUCTION
The liquid water content in snow covers on glaciers is a key parameter for global warming
investigations. Avalanche and flood warning authorities are interested in snow property data as
well. Furthermore hydro-power stations need snow parameters of their catchment in order to
predict the filling stage of their water storage reservoirs. Radar remote sensing promises to
provide this information by the help of accurate ground reference measurements. The authors
have developed a new sensor system for snow moisture and. density determination and examined
it during several field campaigns. The system consists of a net of radio frequency transmission
lines up to about 100 m long. The dielectric coefficient of the material surrounding the lines is
measured with a time domain reflectometer (TDR) at high frequencies and with a low frequency
impedance analyzer as well. The preinstalled lines are enclosed by snow fall and change their
propagation characteristics in respect to variations of the dielectric coefficient of the surrounding
snow.
2. TRANSMISSION LINE DESIGN
The sensor consists of a three wire flat band cable with insulation as shown in Fig. 1. The total
capacitance C of the equivalent circuit is a nonlinear function of the dielectric properties em of
the surrounding material [1]:
(1)
244
White polyethylene insulation
Copper conductors
Fig. 1 Flat band cable sensor cross section and equivalent circuit
The parameters C], C2 and Cs are measured (travel time and wave impedance) and calculated
with electromagnetic field simulations. The values for symmetric excitement are compared in
table 1 together with the transmission line inductance.
Table 1 Comparison of measurement and electromagnetic field calculation
Circuit element
C]
C2
C3
L
Measurement
Field calculation
3.4 pF/m
323 pF/m
14.8 pF/m
756nH1m
3.4 pF/m
276pF/m
15.3 pF/m
741 nHlm
There is a good agreement between the measurement and the calculation. The deviation for C2 is
due to the manufacturing process with its large uncertainty of insulation thickness.
3. AIR GAP CORRECTION
One of the distinguished problems for continuous measurements in snow is the disturbing
interaction of the sensor with the surrounding material. There is always an air gap around the
sensor due to mechanical influences or partial melting. In order to detect and correct the effect of
an air gap the following procedure is applied. The three wire sensor cable is measured in two
modes, with small and with large spacing as shown in Fig. 2.
Small spacing
Large spacing
Fig.2 Electric Field extension for small (1 +,2-,3 open) and large (1 +,2 open, 3-) spacing
245
In the case of large spacing, the air gap has less influence on the measurement of the dielectric
coefficient than in the small spacing mode. This influence can be calculated and a correction
algorithm can be derived [2]. The method was applied successfully to data gathered during an
extensive field campaign on a glacier in Switzerland
3. MEASUREMENT RESULTS
A complete measurement system was installed on the glacier, consisting of horizontal snow
wetness measurement cables connected to a time domain refiectometer, a weather station, snow
height and temperature sensors and a portable computer for the control and data acquisition. The
horizontal cables were arranged in different heights as seen in Fig. 3. They were about 40 m long
which is essential for providing representative data from inhomogeneous snow covers.
Exemplary measurement results of a horizontal cable are shown in Fig. 4.
Fig. 3 Test site with horizontal cables
w 2.5
-f-;-----i--;---i---i--f'-'-ff-'--t---wet
1.5-'-r---;---r---i---;---t---;--.---T--'
75
90
105
120
135
150
165
180
195
Julian day of the year
Fig. 4 Dielectric coefficient of a snow cover versus time
246
It can be distinguished easily between the dry snow and the wet snow phase. In the dry snow
phase there is a constant increase of the dielectric coefficient which is due to the compaction of
the snow for higher density. The relation between dielectric coefficient Cd and density p for the
dry phase is given by the recommendation ofTiuri et al. [3]:
(2)
They also recommend a formula for the additional increase of the dielectric coefficient Ltc under
wet snow conditions with the volumetric water content Vw :
I1c = 0.089 Vw + 0.0072 V;
(3)
This can be seen as a sharp rise in the dielectric coefficient around the 135th day at the beginning
of the melting season. Furthermore, there are strong daily variations which could be seen at finer
time resolutions and related to the actual weather conditions. Another important difference
between the dry snow and the wet snow periods can be seen in the voltage-versus-time curves
measured with the time domain refiectometer (Fig. 5).
1.8
...,-;--------~---,---....,_--.,...,...--,
1.7
Wetsnow
--'--'-,
bl)
.s
(5
>
0
2:
1.6
{;j
~
~
---,. ....--.-...-.-~~.
.i
:
--1---...,.
..... Dry snow
.~ ......... -..... ~ .. -..... .
;.
.. .............:
~ /#
! !
;
1.5
1.4
---!-~I--
. - . . .--. - ' . . . . .,f-....""
....l-+-'----+---+----i-----+----I
50
100
150
200
Relative travel time in ns
Fig. 5 TDR diagram for wet and dry snow periods

Dry (high voltage) and wet (low voltage) sections along the cable can be distinguished.
Percolation funnels can be localized, too. During the dry snow phase which lasts till beginning of
May the snow cover is relatively homogeneous. The melting conditions at the end of May cause
pronounced variations in the voltage pattern. They appear symmetric to the middle of the curve
due to a loop arrangement of the cables. The voltage bumps can be identified with the
percolation zones on the glacier surface.
The aforementioned method is suitable if the density development can be extrapolated from a
long enough dry snow period into the wet snow period. Otherwise two independent
measurements are required to determine the two unknown parameters water content and density.
This can be achieved with a high frequency and a low frequency measurement. The method is
247
based on the fact, that the dielectric coefficient of ice is nearly the same as water (:::::88) at low
frequencies whereas at high frequencies it is only 3.2. The measurement results of another
campaign in the Black Forest show a good agreement between the gravimetric sampling and the
dielectric density determination.
T-;::::=============:;-;----i--r 3.0
400
350 - -
gravimetric sampling
dielectric measurement
300 --- - - high frequency g
..,
r-'-+-;
-_.!-
2.5
2.0
'"
1-...:------..-..-1- 1.5
i __ .
50 --------,
;;
(
I
15
20
25
\t;
\.
30
35
40
45
1.0
50
Julian day of the year
Fig. 6 Temporal development of density and high frequency dielectric coefficient

4. CONCLUSIONS
A new sensor for measuring water content and density of natural snow covers has been
presented. It consists of a long flat band cable enclosed by snow fall. This way the snow
properties of large areas can be monitored continuously. The measurement results show the
temporal development of the dielectric coefficient of a snow cover, wet phase inhomogenities
due to percolation zones and the possibility of determining water content and density at the same
time with a high and low frequency measurement.
REFERENCES
[1]
Huebner, C, Brandelik, A, Doepke, G., Wunderle, S., Rau, F., Braun, M. and F. Weber, A New Method
for Snow Moisture Sensing, Proceedings of the EARSeL Workshop Remote Sensing of Land Ice and
Snow, Freiburg, Germany, 17-18 April 1997, edited by S. Wunderie, 103-109.
[2]
Huebner, C., Brandelik, A, Wunderie, S. and D. Beppler, Ground truth for soil and snow moisture sensing,
Proceedings of the 7th International Symposium on Physical Measurements and Signatures in Remote
Sensing, Courchevel, France, 7-11 April 1997, edited by G. Guyot and T. Phulpin, 281-286.
[3]
Tiuri, M. E., Sihvola, A H., Nyfors, E. G. and M. T. HaUikainen, The complex dielectric constant of snow
at microwave frequencies, IEEE Journal of Oceanic Engineering, 9, 377-382.
[4]
Brandelik, A., Huebner, C., Doepke, G. and S. Wunderie, Advanced ground truth for snow and glacier
sensing, Proceedings of IEEE International Geoscience and Remote Sensing Symposium, Seattle,
Washington, 6-10 July 1998.
Contact point: Alexander Brandelik, Forschungszentrum Karlsruhe, IMK, Postfach 3640, D-76021 Karlsruhe,
Germany, Phone: + 49 (0) 7247-823913, Fax: + 49 (0) 7247-824742, E-mail: alexander.brandelik@imk.fzk.de
248
DIELECTRIC PROPERTIES OF AQUEOUS SOLUTIONS USING T.D.R TECHNIQUE
Sanjay M Puranik
T.S."CHANAKYA", Karave, Nerol, New Mumbai 400706. (India).
ABSTRACT
The dielectric properties of polar-polar liquid mixtures gives valuable information about
hydrogen bonding in the mixtures. Since alcohols are hydrogen bonded liquids, like water, there
is a great interest in dielectric relaxation and process of molecular inter-action in alcohol-water
mixture. Further it is assumed that the primary alcohol-water mixture are the systems with
cooperative process involving large number of molecules with hydrogen bond linkages. They are
microscopically homogeneous, whereas the mixture of higher alcohol-water are microscopically
heterogeneous.
Keywords: Polar-polar liquids, hydrogen bond.
1.) INTRODUCTION
In the present paper the results of dielectric properties (fs;t) and molecular interactions using
effective correlation factors (gr' goff) are discussed for mixture of water & primary (methanol),
secondary (2-propanal), tertiary (tert.Butyl alcohol) and polyhydric alcohols (ethylene Glycol).
The Bruggeman mixture formula for static permittivity has been used with some modification to
gain better understanding of the dipole interaction on the mixture of two liquids. The Kirkwood
equation is also used with some modification to calculate correlation factors in the mixture.
2.) EXPERIMENTAL
The liquids, methanol, propanol, tert Butyl alCohol & ethylene glycol were obtained with 99%
purity & used without any purification. Water used in the preparation of solutions was distilled &
deionised before use. All samples were prepared by the different volume percentage of water.
Dielectric relaxation measurements were carried out by time domain reflectometry (T.D.R). The
detail of the experimental system and method of data analysis have already been described earlier
[1]. The setup consist of a Tektronix 7854 sampling oscilloscope with 7S12 TDR unit. An
oscilloscope is interfaced with PCIXT which performs necessary numerical calculations. The
resulting values of static permittivity and relaxation time are listed in Table 1 at room
temperature.
The values of effective correlation factors (~, g<H) are also calculated in order to understand the
molecular interactions in the liquids by modifying slightly the Kirkwoods equation [2,3] as
~~:
fl<= \OM-CO:>r>I)(2tOM"Tfa:",,)=
QI"
t () I"f' C~Q:! "'" -+ 2
L.
4TIN
CJ K r
249
r.A;~!1-- + .ub~l? c1-';ql
MQ
M '0
geff
(1)
and
f=
Koh
(~e", -;- ~a:n'" ) (2
0'(>0
<l'M (~Cbrt\ -T 2 ) ...
4rr~ [;.1;\":,9. 1; -t ..q~~b~b C;-)()]
-+ E:ol'>, )
=
Met
<j1<T
Mb
'f
(~)
The values of the correlation factors are listed in Table 2.
Table 1: Dielectric relaxation parameters.

% water
in alcohol
MeOH-HOH
~s
-z:
0
10
20
30
40
50
60
70
80
90
100
33.7
40.3
45.9
51.2
56.1
60.8
65.6
68.3
72.1
76.4
79.5
51.4
50.2
35.1
29.4
25.0
17.0
16.8
16.1
13.9
13.1
8.8
2-PrOH-HOH
fs
18.4 245.8
21.3 107.4
31.5
62.1
33.1
61.4
42.6
30.3
49.7
24.2
57.2
22.0
63.3
16.2
12.0
68.9
70.2
11.0
79.5
8.8
tBA-HOH
Es
11.5
15.2
20.0
26.5
34.5
50.1
55.0
55.4
65.1
73.8
79.5
EG-HOH
r:s
276.9
132.3
78.8
65.6
38.4
26.5
21.0
20.0
13.4
13.0
8.8
40.8
46.4
52.0
56.0
59.0
64.3
64.6
69.7
73.0
75.6
79.5
104.6
67.4
57.8
44.8
33.7
26.8
25.9
19.5
13.7
10.8
8.8
The Kirkwood equation gives the significance of the association effect in the liquids due to
dipole correlation and correlation factor may provide some information about the orientation of
the dipoles in the mixture of two liquids.
Table 2: Kirkwoods correlation factors.
% water
in alcohol
0
10
20
30
40
50
60
70
80
90
100
MeOH-HOH
9\>ff 9f
3.19
3.28
3.26
3.21
3.15
3.08
3.03
2.89
2.81
2.76
2.67
1.00
1.06
1.08
1.10
1.10
1.09
1.09
1.05
1.03
1.02
1.00
2-PrOH-HOH
9~~F
3.25
2.58
3.00
2.56
2.79
2.82
2.86
2.82
2.77
2.57
2.67
9f
1.00
0.86
1.04
0.90
1.00
1.02
1.04
1.04
1.02
0.95
1.00
250
tBA-HOH
9i'ff 9f
2.28
2.01
2.00
2.13
2.32
2.91
2.79
2.49
2.63
2.71
2.67
1.00
0.82
0.79
0.83
0.89
1.11
1.05
0.94
0.99
1.01
1.00
EG-HOH
gef?
2.61
2.71
2.81
2.81
2.76
2.82
2.66
2.72
2.70
2.66
2.67
9.f
1.00
1.03
1.06
1.06
1.04
1.06
1.00
1.02
1.01
0.99
1.00
/
/
/
/
/
I
I
I
/
/
/
I
/
lO
ci
~
0
I
I
to
....'"
3
X.
r-
cr
.:r
,/
2..~
0-
,
I
-J
I:
"1
0
c:i
I.
.--,---
..:!>
c::>
Co)
tl"?
...:::
251
.-,
0
'"
c..;)
..n
d
'3
L
r..Il
v
.......
CS
><:
.J::.
;...
,/
......
<D
The Bruggeman mixture formula [4,5] of static permittivity has been modified for binary
mixture ofpolar liquids which may provide information about interactions in the mixtures by the
follOwing equation:
(3 )
where f B, the Bruggeman factor; v, the volume fraction of solvent; tSf\." ESV' and tss are static
permittivity, respectively, for
mixture, water and solvent and "a/tis the parameter gives the information regarding the change in
orientation of solvent molecule in the mixture.
3.) RESULTS AND DISCUSSION.

As can be seen from Table 1 the values of and at room temperature agree well with the values
reported earlier by different techniques and T.D.R.
For alcohol-water system the values of goff and ~ are given in Table 2. Note that effective
correlation factor g<ff decreases from pure alcohol to pure water in general. Some times it shows
higher values indicating structure making effect. The value of gf' on the other hand, remains close
to one generally. Deviation from one indicates high correlated dipole interaction in the mixture.
The graphical representation of Bruggeman factor, fa is plotted for all the samples of alcoholwater in Fig. I. For tBA-water and EG-water systems the correlation factors are smaller than that
of water indicating stronger hydrogen bonded linkages than lower alcohols. According to
Bruggeman a linear relationship is expected. But for the systems studied here we found that our
experimental data for alcohol-water system do not agree with theoretical prediction of
Bruggeman. In Eqn. (3) if a= 1 then equation leads to original Bruggeman model and indicates no
additional interaction between two systems. Deviation of 'a' from 1 implies stronger molecular
interaction in the mixture. It should be noted that the original equation is proposed for the mixture
of polar-nonpolar liquids in which molecular interaction is essentially one. The Original
Bruggeman equation [5] fails for mixture of 2 polar liquids. Therefore one has to use the
modified Bruggeman equation as above.
4.) CONCLUSION
The present study with slight modification in Kirkwood and Bruggeman equation may provides
useful information about hydrogen bonding linkages for primary to polyhydric alcohols.
REFERENCES:
[1] Kumbharkhane et al.,J.Chem.Soc.Farad.Trans,(1991),87/1O,1569.
[2] Kirkwood,J.chem.phys . (1939),7,911
[3] Kumbhaekhane et al . IMol.liqs . (1992).51,307.
[4] D.A.G. Bruggeman. Ann.Physik.,(1935),24,636.
[5] Puranik et al., J.Mol.liqs . (l994).59.173.
252
ERRORS IN TDR-DETERMINED WATER CONTENT

IN A SALINE SAND
Paul Ferre, William Hook, Nigel Livingston and Christopher Bassey
Dept. of Biology, P.O. Box 3020, University of Victoria, Victoria, B.C., Canada.
tyferre@uvic.ca
ABSTRACT.
Measurements with an Automated Network Analyzer (ANA) in
transmission show that there are increases in the travel time and rise-time of a step pulse
transmitted through sands saturated with KCI solutions. The rise time is shown to be a
useful diagnostic of the presence of errors in water content measurements using time
domain reflectometry systems.
Keywords: time domain rejlectometry, automated network analyzer, lossy media.
1. INTRODUCTION
Time domain reflectometry (TDR) has been shown to be a very effective, rapid,
nondestructive method of volumetric water content measurement in a wide variety of
porous media. The goal of this investigation was to determine if there are errors in the
TDR-measured water content associated with high pore water salinity in sand and, if so, if
the errors are correctable.
An ANA was used to allow for simultaneous frequency domain and time domain analyses,
in addition to providing more accurate measurements than can be achieved with standard
TDR instrumentation. Operation in the transmission mode removes any uncertainties
associated with end effects.
2. INSTRUMENTATION AND :METHODS

The ANA used in this study is a Hewlett-Packard HP 8752C Network Analyzer, suitable
for measurements in the frequency range from 300kHz to 3 GHz. Using the built-in Fast
Fourier Transform (FFT) capabilities of the instrument, the time-domain response to a
step pulse was simulated for comparison with conventional TDR instruments.
Following Zegelin et al. (1989) the probes used in this study were comprised of three
stainless steel rods, 32 cm long, 0.3 cm in diameter, separated by 0.85 cm gaps. The
three rods were aligned in a horizontal plane. The rods extended through two opposite
faces of a 28 by 17 by 19 cm Plexiglas box, allowing for connection to the ANA in a
transmission configuration. Measurements were made in sand saturated with KCI
solutions with electrical conductivities ranging from 0 to 40 dS/m. The probes were
connected to the ANA through 50-ohm coaxial cables. Measurements were made at 401
points in the frequency range of 0.3 to 1500 MHz. The phase and amplitude of the
transmitted signal were measured in the selected frequency range, first with the line
253
empty and then with the line loaded with the test material. Then the time delay for the
empty and sample filled cases (tf and t e) and the rise time of the signal were computed
from the resulting FFT step-pulse using standard straight-line curve-fitting techniques.
The water content was determined from the time delay measurements using a general
relationship derived from that presented by Hook and Livingston (1995) for nonclay
soils,
e = 0.1256[tf ~te +1-1.756J
(1)
tmr
Four solutions of KCI were prepared in distilled water to electrical conductivities of 0,

10, 25, and 40 dS/m. Clean sand was added to each solution. Each saturated sand
mixture was packed into the testing box for measurement. The sand was packed to a
height of at least 5 cm above the horizontal plane in which the three rods were located to
ensure that the sample area of the probes did not extend beyond the soil surface. Solution
was ponded on the soil surface to assure full saturation. Saturated sand was used rather
than attempting to pack to a specified level of partial saturation to ensure repeatability of
packing and spatial uniformity of the water content in the sampling volume. In addition,
full saturation represents the condition of greatest electrical loss for a given solution EC.
Finally, for these preliminary experiments, clean sand was chosen as the medium of study
to minimize surface conductivity and the uncertainties associated with water content
measurement in clay-rich soils.
3. RESULTS
The frequency domain results (not shown) demonstrated that the magnitudes of the high
frequency components dropped sharply to very low values for solutions with electrical
conductivities greater than 10 dS/m, leaving only low frequency components.
Furthermore, the phase, and hence the time delay, at a given low frequency increased
with the salinity of the pore water.
Both the time delay, t, and the rise time, tr, are necessary to explain errors in TDRdetermined water contents in saline soils. The rise time is defined here as the time for the
transmitted pulse to rise to one half of the eventual maximum amplitude. Figure 1 shows
the highly linear (R2 = 0.9976) relationship between tr and the pore water EC for one of
our experimental trials. Figure 2 shows the water content, determined with Equation 1,
as a function of solution Ee. A second order polynomial is fit to the data. Using 1/10th
of the rise time as an estimate of the time measurement error (Hook and Livingston,
1995), an envelope is drawn about this best fit polynomial. This represents overestimates
and underestimates of 1/1Oth of the rise time in the measured travel time, showing the
expected upper and lower limits of water content errors as a function of pore water EC.
Data from multiple trials was largely contained within this envelope.
254
12
10
!S
r~
....e
E-i
8
6
.~
4
2
0
0
10
20
30
40
EC, dS/m
Figure 1: Rise time vs. soil water electrical conductivity for four KCI solutions.
0.95
--==
e
~
;..
0.85
0.75
0.65
0.55
0.45
0.35
0
10
20
30
40
EC, dS/m
Figure 2: Water content vs. soil water BC for four KCl solutions.
3. CONCLUSIONS
Our results demonstrate that there is an increase in the travel time as a function of pore
water EC. This increase is due to the progressive loss of high frequency components with
increasing salinity coupled with the Maxwell-Wagner effect causing an increase in the
time delay for the remaining lower frequency components. Further calculations have
shown that corrections for the contributions of the loss tangent are insufficient to explain
the observed time delays and that dispersion plays only a minor role in this system. The
255
time domain data also demonstrate an increase in the measured rise time with increasing
salinity, caused by the loss of high frequency components as seen in the frequency
domain. The increased attenuation of higher frequencies may be caused by the highly
concentrated ions inhibiting the water molecules from responding freely to the alternating
field, resulting in increased dissipation of energy with increasing frequency (Conway
1981)' The rise time can predict the average water content error as a function of solution
EC, serving as a potential diagnostic measure of data quality. However, the uncertainty
in the time delay measurements, which is as large as the average error, cannot be
eliminated, suggesting an inherent limitation to water content measurement with
conventional TDR in highly saline soils.
REFERENCES
[1]
Zegelin, S.J., I. White, and D.R. Jenkins, Improved field probes for soil water content and electrical
conductivity measurements using time domain reflectometIy. Water Resour. Res .. 1989, 25, 23672376.
[2]
Hook W.R. and N.J. Livingston, Errors in converting Time Domain ReflectometIy measurements of
propagation velocity to estimates of soil water content, Soil Sci. Soc. Am. I, 1995, 59, 35-41.
[3]
Conway B.E, Ionic Hydration in ChemistIy and Biophysics, Amsterdam, Elsevier Scientific
Publishing Company, 1981.
Contact point: Paul Ferre, Dept. of Biology, P.O. Box 3020, University of Victoria,
Victoria, B.C., V8W 3N5, Canada. Phone: (250) 721-7121; Fax: (250) 7217120; E-main: tyferre@uvic.ca
256
EFFECTIVE FREQUENCY OF TDR TRAVEL TIME-BASED

MEASUREMENT OF BULK DIELECTRIC PERMITTIVITY
Dani Or and V Philip Rasmussen*
Department of Plants, Soils and Biometeorology, Utah State University
Logan, UT 84322-4820, U.S.A. dani@mendel.usu.edu
ABSTRACT. The correspondence between the complete spectrum of dielectric permittivity (E vs.
frequency) and the single value provided by TDR travel time analysis of waveforms (E b) was studied
by identifying an effective frequency (f') for which E' (f')=E b. Dielectric spectra of water and ethanol
mixtures and soils at different wetness levels were measured with an HP-8752C network analyzer
(equipped with 85070B dielectric probe) in the frequency range of 0-6 GHz. The mixtures were also
measured with a Tektronix 1502B TDR using 0.2 m long stainless steel coaxial cells. Results show
the correspondence between Eband E'(f) varied with the mixture dielectric constant and occurred at
different frequencies. The effective frequency f' for the water-ethanol mixtures was nonlinearly
dependent on the mixture's Eb value according to:/=fJl-e -aEb), withfoo=2.34 GHz and a= 0.0216.
Keywords: Time Domain Reflectometry (TDR), dielectric spectrum, waveforms

1. INTRODUCTION
Time Domain Reflectometry (TDR) measurement of soil bulk dielectric permittivity (Eb)
provides the basis for accurate determination of soil water content (8) used in many hydrological
settings. In contrast to frequency domain techniques and classical TDR spectroscopy methods [1,
2, 3] that provide information on the dielectric permittivity as a function of frequency, present
TDR applications yield a single value of bulk dielectric constant (E b) deduced from travel time
analysis reflection of waveform. Measurements in heterogeneous and often lossy Earth materials
and practical constraints necessitating the use of nonideal waveguide designs for insertion into
porous media [4] raise questions regarding the relationships between the singular permittivity
value obtained by TDR and the detailed dielectric spectrum (for a similar frequency range).
Improved understanding of these relationships is important for evaluation of temperature and soil
texture effects on TDR measurements. Of particular interest is the effective frequency (f') of
TDR-determined Eb, or the frequency at which E'(f)=Eb.
2. EXPERIMENTAL PROCEDURES
Dielectric spectra of deionized water and ethanol mixtures (with 100,33,50,66, and 0% ethanol
in water) were measured with an HP-8752C network analyzer (equipped with 85070B dielectric
probe) for the frequency range of 0-6 GHz. The mixtures were also measured with a Tektronix
1502B TDR using 0.2 m long stainless steel coaxial cells (open ended). All measurements were
made at room temperature (25C 2C ). Additional measurements of soil samples at different
water contents and clay-sand mixtures were also made. The HP 85070B dielectric probe kit and
software provide directly real and imaginary dielectric permittivity values as functions of
frequency. TDR waveforms were analyzed for Eb using methods described.in [5, 6].
257
3. RESULTS
Bulk dielectric permittivity values for different water-ethanol mixtures obtained from TDR
waveform analysis (travel time analysis) and respective values from the network analyzer
dielectric spectra are depicted in Fig. 1. Results show the correspondence between Eb determined
from TDR travel time analyses and network analyzer dielectric spectra varied with water-ethanol
ratio and occurred at different frequencies.
25
~
\
20
15
10
5
It:tnanol -1 Uuv/o
I
0.6
g'
~
e"......
0.2
e1T19.1
-0.2
Il
-0.6
'/
-1
0.5
1.5
2.5
50
~40
>
E
:::
1:
30
!ECI)
'
'(3
CI)
0.2
eb=45.7
Il..
-'*
.5:! 20
o
~
c -0.2
i5
0.6
CI)
.s!
....
10
II'\....
f*
80
-1
0.6
,Ivvaler It;'(bL;) 1uuv1
0.2
60
40
-0.2
20
-- --gil
f.--'
o
o
i - -~
0.5
1.5
2.5
g'
fl
1\
-0.6
-0.6
--
e~78.3
~l\
~/
-1
Frequency (GHz)
0.5
1
1.5
Distance (m)
2.5
FIGURE 1: Dielectric spectra and TDR waveforms (dashed lines mark locations of
fIrst and fmal reflections) for ethanol, water, and their mixture. The concept of
effective frequency is illustrated for the mixture.
The resulting effective frequencies f' for the different mixtures are illustrated in Fig. 2
(symbols). The dependency of f" on the mixture bulk dielectric value (E b) is expressed as:
j* =fJl-e -aEb), withf",=2.34 GHz and a= 0.0216 (denoted by a line in Fig.2).
258
'N2.0
J:
~-----~-----_ _ _--,
C)
~1.5
CI)
:::s
g"1.0
i-
LL.
CI)
~ 0.5
(.)
~
W
0.0 +----1----+----+----+---1
35
5
20
50
65
80
Relative Dielectric Permittivity
FIGURE 2: Effective TDR frequency as a function
of pennittivity of water-ethanol mixtures.
The reliability of estimation of r from dielectric spectra hinges on the slope of E' (f), and to a
certain extent, also on interpretation errors associated with Eb. Unlike the large changes observed
for water-ethanol dielectric spectra (induced by relaxation of ethanol occurring within the
measurement bandwidth ::::800 MHz), the spectra for partially-wet soils were practically flat (i.e.,
aE'laf::::O) in the frequency range of interest (Fig. 3). For the results depicted in Fig. 3, the
estimated r values ranged from 1.0 to 3.6 GHz (varying with values Ebfor replicate soil
samples).
40
'7 0.2
'>
~ 30
c'
0.0
-0.2
CI)
20
<3
to)
:sg 10
-0.4
g -0.6
Qj
1:
CI)
o
o
"
cb=23.8
:0;
E'
(a)
234
Millville silt loam

20%W.C.
Frequency (GHz)
;;::
-0.8
CI)
0:: -1.0
0.0
'"' --"
0.5
1/
1.0
(b)
1.5
2.0
Distance (m)
FIGURE 3: (a) Dielectric pennittivity spectrum and (b) TDR wavefonn for Millville silt loam soil
with 20% mass water content (dashed lines mark 1'1 and 2 nd reflection used for travel time analyses).
4. DISCUSSION AND CONCLUSIONS
The high sensitivity of estimation of in soils to slight inaccuracies in Eb, and possible contact
problems between the HP 85070B dielectric probe and unsaturated soil surfaces, prohibited
identification of clear trends similar to that depicted in Fig. 2 for water-ethanol mixtures.
Nevertheless, if the results in Fig. 2 are representative (at least for low loss materials), we can
make the following inferences: The lower the bulk dielectric constant, the lower TDR effective
259
frequency. This may be interpreted as follows: for mixtures with a low bulk dielectric constant,
lower frequency components in a TDR signal are sufficient for resolving Eb through travel time
analysis of the corresponding wavefonns. With the increase in dielectric constant, higher
frequency components are needed to obtain E b. A part of this dependency is attributed to the
reliance of TDR wavefonn analysis on rapid changes in the wavefonn for identifying or marking
the first and second reflections (e.g., Fig. 3b). In turn, these features are defined by the highest
frequency components present in a TDR signal [7], and as these attenuate or are filtered out (i.e.,
reflections become more "rounded") the locations of these reflections on a wavefonn become
uncertain and affect the accuracy of travel time analysis. Further loss of frequency components
ultimately causes wavefonn analysis to fail.
These results may also provide some practical constraints for low-cost TDR units and similar
sensors (e.g., reflectometers), regarding essential frequency components (or rise times) required
for measurements in materials with low dielectric constants. The results for a lossy medium (wet
soil) depicted in Fig. 3 illustrate the important effect of ohmic losses on TDR signal quality and
on accuracy ofwavefonn analysis [5,6,7]. The relatively "flat" dielectric spectrum precludes
definite conclusions, yet it is likely that the effective frequency in this system would be higher
than for a water-ethanol mixture with a comparable dielectric constant (i.e., 80% ethanol).
In summary, the frequency at which TDR-measured Eb matches E'(f) was nonlinearly dependent
on the value of E b , due to interactions between features in TDR wavefonn travel time analyses
and loss of signal frequency components as the signal traverses the sample. Future studies will
focus on establishing a theoretical basis for the dependency of r on E b, and on the role of ohmic
losses on r using mixtures of polar liquids and suspensions to reduce contact issues.
REFERENCES
[1]
Clarkston, T. S., Glasser, L., Tuxworth, R. W., and Williams, G., An appreciation of experimental factors
in time-domain spectroscopy. Adv. Molec. Relax. Proc., 1977, 10,173-202.
[2]
Cole, R.H., Time domain reflectometry. Ann. Rev. Phys. Chern., 1977,28,283-300.
[3]
Rau, R. N., and Wharton, R. P., Measurement of core electrical parameters at ultrahigh and microwave
frequencies. J. Petrol. Techn., 1982, November, 2689-2700.
[4]
Zegelin, S. J, White, I., and Jenkins, D. R., Improved field probes for soil-water content and electrical
conductivity measurements using time domain reflectometry: Water Resour. Res., 1990,25,2367-2376.
[5]
Or, D., and Wraith, J.M., Temperature effects on soil bulk dielectric permittivity measured by time domain
reflectometry: A physical model. Water Resour. Res., 1999,35:371-383.
[6]
Wraith, J.M., and Or, D., Temperature effects on soil bulk dielectric permittivity measured by time domain
reflectometry: Experimental evidence and hypothesis development. Water Resour. Res., 1999,35:361-369.
[7]
Friel, R., and Or, D., Frequency analysis oftime domain reflectometry (TDR) with applications to
dielectric spectroscopy of soil constituents. Geophysics, 1999,64:1-12.
Contact point: Dani Or, Dept. Plant, Soils, and Biometeorology, Utah State University, Logan, Utah 84322-4820,
U.S.A, Phone: +1 (435) 797-2637; Fax +1 (435) 797-2117; E-Mail: dani@mendel.usu.edu
260
MODELLING TDR RESPONSE IN HETEROGENEOUS COMPOSITE

MATERIALS
Ian Woodhead, Graeme Buchan, Don Kulasiri
Lincoln University, Canterbury, New Zealand. Woodhead@lincoln.ac.nz.
ABSTRACT. The use of time domain reflectometry (TDR) techniques for measuring the moisture
content of composite materials is a mature art but usually makes assumptions about the
homogeneity of the material, since the response of a parallel transmission line to transverse
dielectric variation is non-linear. Further, because the electric field distribution depends on the
moisture distribution within the material, an analytic solution to the sensitivity distribution in the
transverse plane is impracticable, so numerical solutions are required. We describe an integral
equation approach to model the response of the TDR system to a heterogeneous dielectric body.
Then, in conjunction with a suitable dielectric model of the composite material, the TDR response
to moisture content distribution may be quantified.
Keywords: TDR, heterogeneous, dielectric, spatial.
1. INTRODUCTION
TDR (time domain reflectometry) is used extensively for measurement ofe, the volumetric
moisture content in soil, and applicable but less widely used in other materials such as grains,
powders, and minerals. For measurement ofe, a short open-ended transmission line that is
typically 300 mm long, is buried in the material under test. The travel time of a pulse with very
short risetime (typically < 300 ps) is measured and provides the mean propagation velocity v, on
the line of known length. Since most biological and composite materials make no contribution to
the permeability of the region, v indicates the mean relative permittivity Br of the material
surrounding the transmission line. When the loss tangent is small, and the relative permeability is
one, Br is:
(1)
where c is the velocity oflight. Since Br for most biological and composite materials is typically in
the range of three to five whereas that of water is typically 80, Br ofa material forms a useful
surrogate for its moisture content. Frequently, empirical calibration techniques are used since
practical dielectric models are usually unable to account for the subtle interactions between water
molecules and the material, and which affect the water's polarisability. For example Topp et al
[1] developed a polynomial relating the measured Br to the moisture content of soil. This
calibration is applicable to quite a wide range of soil types (and hence orders of magnitude
variation in particle size with their attendant variable interactions with water molecules) and
typically has an accuracy of better than 2% in e over the range 5 to 50%.
261
It has been shown [2] that longitudinal variation in moisture content is fairly accurately integrated
by TDR systems. However the lateral sensitivity is not a linear function of distance from the
transmission line, so lateral variations are not integrated linearly. Knight [3] derived analytically,
an approximate lateral sensitivity weighting function, useful if the actual permittivity distribution
was nearly uniform, but for an exact solution a numerical approach is required. Knight et al [4]
used a numerical method that agreed well with previous analytical calculations of Annan [5] and
Knight [2]. The method used a finite element algorithm adapted to the electrostatic case since
this approach adequately represents a TEM transmission on parallel lines provided the line is
essentially lossless.
Here an IE (integral equation) method is chosen as the forward model for use with a tomographic
inversion algorithm which we are currently developing. This application exploits a key
characteristic of IE methods, namely that where the region of anomalous permittivity is
surrounded by free space, only the anomalous region need be calculated. The IE approach also
has advantages in tomography, since matrix recalculation is not necessary for changes in the
impressed electric field. However a disadvantage is that when applied to arbitrary permittivity
distributions, IE methods require volume integration even where one dimension is invarian~.
Under these circumstances as in the present case, quasi 3-D or 2.5-D variants reduce the
computational burden of the IE method to compare more favourably with the DE approach. In
our application, we use the output from the model to directly calculate the velocity of propagation
on the transmission line
2 INTEGRAL EQUATION AND DISCRETISATION
The polarisation of a discretised zone or cell within a dielectric material may be represented by a
dipole at its geometric centre. In most dielectric materials, there is no net polarisation until
generated by an external or impressed field. When applied to this quasi-static electric field
problem, the method of moments may be considered as the summation in each cell, of the electric
field contributions due to the polarisation in all other cells. The potential p at point p{x,y,z)
generated by a dipole with dipole moment or polarisation P, is:
_"
P.r
41r8
(2)
or2
1\
where ris a unit vector pointing from the centre of the dipole to p [6]. In Cartesian 3-space:
= P.(x+y+z)
p
41r8 r3
o
(3)
where x,y and z are the rectangular components of r. The potential arising from many such
dipoles in a region is:
(4)
262
/\
where dv is the differential volume over which each P. r applies. Reverting to the single dipole
case, its electric field is the space rate of change of potential ( -grad p) so that:
E px
= -p s [/\x(r2 -
/\
/\
3X2) - y(3xy) - z(3xz)
(5)
41r&or
and with corresponding equations for Epy and Epz, may be expressed as a dyadic equation
1\ 1\
I\. 1\
1\ A
1\ A
/\ 1\
1\ 1\
X x(3x 2 - r2) + x y(3xy) + xz(3xz)

Ep =
P
41r&o r
s yy(3y2 _r2)+ yx(3xy) + yz(3xz)

1\ 1\
1\./\
(6)
1\ 1\
zz(3z2 - r2) + zx(3zx) +zy(3zy)
We may combine the above in an integral equation describing the electric field Ep at a point p:
/\
Ep(x,y,z)
=-V(fff
P.r 2 dv)
(7)
41r&or
The polarisation region may now be discretised, and following the method of moments [7], we
calculate the matrix of polarisation vectors p(x,y,z) using:
L(P) =-EJ (x,y,z)
P(x,y,z)
=E p ( x,y,z) --..:........:..:'--'-':.....
(8)
&oZ(x,y,z)
where L is a linear operator, Ei the external impressed field and X(x,y,z) the electric susceptibility
(er(X,y,Z) - 1). Equation 8 is converted to matrix form and solved for the vector of polarisations
P, and the electric field strength in each cell is recovered from the polarisation:
E( x,y,z)
= P(x,y,z)
(9)
&oZ(x,y,z)
The inputs required for the method are: a vector comprising sets of three elements describing the
impressed field, a matrix describing the permittivity within each cell, and the dimensionality of the
problem. While the above method applies to any impressed field distribution, in this case E; is the
vector of impressed field components due to a parallel transmission line. To obtain the potential
difference between the two lines and hence determine line capacitance, the field described by the
matrix E is integrated along a path connecting the two lines (along the x-axis for example). Then
to obtain the velocity of an electrical edge on the transmission line (assumed lossless), the
standard transmission line formula is used:
263
V=
1
.JLC =
fE(x,y,z).dl
QPCOSh-{!)
7r
(10)
Here dl is the length element of the numerical integration (the cell length in this discretised case),
q the same initial line charge density that defined the impressed field, P the total permeability, b
the transmission line rod spacing, and a the rod diameter.
3 VERIFICATION
We chose to verify the above calculation by comparison with propagation times from a Tektronix
1502C connected with 0.8 m ofURM 43 coaxial cable to a 1:4 balun. Balun construction
followed [8], but omitted the initial 1: 1 transformer, and used a single, grade S3 ferrite toroid. A
relay (similar to Teledyne 172) switched the balanced line to either a reference transmission line or
the measuring line. The 6 mm diameter stainless steel rods were spaced 60 mm apart, with the
measuring rods 300 mm longer than the reference rods. At the end of the transmission lines 6 by
Imm steel shorting straps provided sharper, better defined reflections than unterminated lines.
Waveform data retrieved from the 1502 were smoothed and differentiated using 25 point routines
[9]. The intersection between the tangents to the maximum negative slope and the immediately
preceding stationary point defined the edge of the pulse. Finally, the reading from the reference
line was subtracted from that of the measuring line to obtain the actual travel time of the edge.
A rectangular thin walled plastic container 150 by 500 by 80 mm filled with water, formed the
phantom dielectric body. The transmission line was positioned near the phantom and used
computer readable position sensing with 1 mm precision to record relative positions.
4 RESULTS AND DISCUSSION
Table 1. Comparison of measured and calculated data
Position (mm) Measured (ns) Model (ns)

-30,8
-30, 13
-30,23
-30,33
45,8
45, 13
45,23
45,33
Distant
1.071
1.016
1.001
0.992
1.172
1.047
1.009
0.998
0.987
1.076
1.039
1.017
1.010
1.204
1.103
1.035
1.019
1.008
Discrepancy (ps)
Normalised (ps)
5
23
18
18
32
56
26
21
21
-16
2
-3
-3
11
35
5
0
0
In Table 1, the position is defined as the (x, y) distances (mm) between top edge of the container
and the geometric centre of transmission line. The 'distant' position provided a reference reading
to normalise the data. Model predictions were calculated using 5 mm cubic cells and a quasi 3-D
264
approach that includes the influence of the neighbouring cells in the z direction within the 2-D (xy)
matrix. The constant portion of the discrepancy between the model prediction and the
measurements is attributed to the imperfect measuring system. The value of .s;.for water that was
used in the model, took account of the temperature of the water in the container.
5 CONCLUSIONS
We describe an integral equation method for determining the electric field distribution in a low
loss, inhomogenous dielectric material given a pre-determined impressed field, E i . In the case of a
parallel transmission line generating Ei , the procedure has been extended to calculate line
parameters and hence the propagation velocity of a pulse on the line. Thus the procedure enables
prediction of the impact of arbitrary dielectric (or moisture content) distributions on a TDR
measurement system, and quantification of its sensitivity distribution. Experimental verification
using water as a dielectric body, provided good agreement with the model predictions.
REFERENCES
[1]
G. C. Topp, J. L. Davis, and A. P. Annan, Electromagnetic detennination of soil water content:

measurements in coaxial transmission lines, Water Resources Research, vol. 16, pp. 574-582, 1980.
[2] P. A. Ferre, D. L. Rudolph, and R G. Kachanoski, Spatial averaging of water content by time domain
reflectometry: Implications for twin rod probes with and without dielectric coatings, Water Resources
Research, vol. 32, pp. 271-279,1996.
[3] J. H. Knight, Sensitivity of time domain reflectometry measurements to lateral variations in soil water
content, Water Resources Research, vol. 28, pp. 2345-2352, 1992.
[4] J. H. Knight, P. A. Ferre, D. L. Rudolph, and R G. Kachanowski, A numerical analysis of the effects of
coatings and gaps upon relative dielectric pennittivity measurement with time domain reflectometry, Water
Resources Research, vol. 33, pp. 1455-1460, 1997.
[5] A. P. Annan, Time-domain reflectometry - air gap problem for parallel wire transmission lines, Geol. Surv.
Can, Report of activities 77 -lB, 1977.
[6] A. F. Kip, Fundamentals of Electricity and Magnetism, 2 ed: McGraw-Hill, 1962.
[7] R F. Harrington, Field Computation by Moment Methods: R E Kreiger, 1968.
[8] E. J. A. Spaans and J. M. Baker, Simple Baluns in Parallel Probes for Time Domain Reflectometry, Soil Sci
Soc Am J, vol. 57, pp. 668-673,1993.
[9] A. Savitsky and M. J. E. Golay, Smoothing and Differentiation of Data by Simplified Least Squares
Procedures, Analytical Chemistry, vol. 36, pp. 1627-1639, 1964.
Contact point: Ian Woodhead, Lincoln Ventures Ltd, Lincoln University, Canterbury, New Zealand. Ph +6433253721; Fax +64 3325-3725; E-mail: woodhead@lincoln.ac.nz
265

ISEMA Proceedings 1999 PDF

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

ISEMA Proceedings 1999 PDF

Uploaded by

Copyright:

Available Formats

COLLECTION

AGRICULTURAL RESEARCH SERVICE

Collection of Papers for the

Dr. Andrzej Kraszewski, Organizer

Sponsored by U. S. Department of Agriculture,

Cover Design: Mr. Richard Fiala

Fifteen years of research on moisture content determination in cereal grains. S.O.

Methods and devices for density-independent moisture measurement. K. Kupfer,

Measurement of composition and prior treatment of foodstuffs using complex

Three dielectric measurement techniques for liquids. J. Baker-Jarvis, and M.D.

10:30 Moisture content measurements in solids - Limitations and improvements with

Multivariate calibration for optimization of a microwave moisture content method

Determination of moisture content in wheat using an artificial neural network.

Calibration of a microwave system for measuring grain moisture content. M. Ben

A new microwave method for moisture measurement in building materials. A.

The application of electromagnetic field on water structure and biology. Q. Hou,

Non-destructive microstrip resonator technique for moisture/permittivity

Nonclassical effects of the microwave interaction with aqueous systems. AF.

Non-destructive and in-situ measurements of water content in plant tissue using

Dielectric study on dynamical structure of water in moist materials using

Quality forecast of particleboards using a microwave monitoring system. F.

Observation of changes in dynamics of water molecules in various aqueous

Dielectric grain moisture measurement - practical, technical, and regulatory

Swept-frequency grain moisture sensor. KC. Lawrence, USDA-ARS, Athens, GA;

Grain moisture measurements with time domain reflectometry. M. Stacheder, P.

A quadrupole to assess the spatial distribution of plant water content in crops. J.

A practical method of measuring moisture using an insulated monopole. A

Moisture gradient studies in timber by the measurement of dielectric parameters

Microwave converter for simultaneous measurement of the reflection and

Application of open coaxial resonator as a sensor for microwave moisture meter.

TUESDAY, APRIL 13,1999, Holiday Inn, Georgian East Room

Active microwave radiometry in determining sand moisture. G. Polivka, Spacek

Resonator-based microwave moisture meter with optimized digital signal

A unified calibration method for moisture sensing in particulate materials. S.

Free space moisture measurement of cellular concrete. T. Lasri, D. Glay, A.

Measurement of the moisture and salt content of building materials. W. Leschnik,

Dielectric properties measurement using a cavity perturbation technique. M.S.

10:30 Dielectric relaxation in water saturated sedimentary rocks. R. Hilfer, J.

Institute Building Climatology, Technology, University of Dresden, Dresden,

UT. ............................................................................................................................. 217

Noon Thermodielectric response of variably saturated porous media: specific surface

Dielectric properties of liquid water and saline solutions below 0 C as measured

Study of dielectric behaviour of human bile at microwave frequencies. S.B.

Modeling the dielectric constant of rice grains. K.P. Thakur, Microwave

Dielectric properties of aqueous solutions using time domain reflectometry (TDR)

Errors in TDR-determined water content in saline sand. P. Ferre, W.R. Hook, N.

Effective frequency of TDR travel time-based measurement of bulk dielectric

Modeling TDR response in heterogeneous composite materials. 1. Woodhead, G.

Summary and Closing Remarks

WELCOME TO THE WORKSHOP!

Andrzej Kraszewski, D.Sc.

Chair: Dr. Andrzej W. Kraszewski,

Monday, April 12, 1999.

FIFTEEN YEARS OF RESEARCH ON MOISTURE CONTENT

Keywords: cereal grains, moisture content, radio-jrequency measurements, microwave

DIELECTRIC PROPERTIES DEPENDENCE

SINGLE KERNEL MOISTURE MEASUREMENTS

BULK GRAIN PERMITTIVITY MEASUREMENTS

DENSITY-INDEPENDENT GRAIN MOISTURE DETERMINATION

METHODS AND DEVICES FOR

Material-Research and Testing Institute (MFPA) at the Bauhaus-University Weimar,

+ 4.83 In C1MHz +4.410