(MD Zahurul Haq) Applied Measurement System

APPLIED
MEASUREMENT SYSTEMS
Edited by
Md. Zahurul Haq
APPLIED
MEASUREMENT SYSTEMS
Edited by Md. Zahurul Haq
Applied Measurement Systems

Edited by Md. Zahurul Haq
Published by InTech
Janeza Trdine 9, 51000 Rijeka, Croatia
Copyright 2012 InTech
All chapters are Open Access distributed under the Creative Commons Attribution 3.0
license, which allows users to download, copy and build upon published articles even for
commercial purposes, as long as the author and publisher are properly credited, which
ensures maximum dissemination and a wider impact of our publications. After this work
has been published by InTech, authors have the right to republish it, in whole or part, in
any publication of which they are the author, and to make other personal use of the
work. Any republication, referencing or personal use of the work must explicitly identify
the original source.
As for readers, this license allows users to download, copy and build upon published
chapters even for commercial purposes, as long as the author and publisher are properly
credited, which ensures maximum dissemination and a wider impact of our publications.
Notice
Statements and opinions expressed in the chapters are these of the individual contributors
and not necessarily those of the editors or publisher. No responsibility is accepted for the
accuracy of information contained in the published chapters. The publisher assumes no
responsibility for any damage or injury to persons or property arising out of the use of any
materials, instructions, methods or ideas contained in the book.
Publishing Process Manager Mirna Cvijic
Technical Editor Teodora Smiljanic
Cover Designer InTech Design Team
First published February, 2012
Printed in Croatia
A free online edition of this book is available at www.intechopen.com
Additional hard copies can be obtained from [email protected]
Applied Measurement Systems, Edited by Md. Zahurul Haq

p. cm.
ISBN 978-953-51-0103-1
Contents
Preface IX
Chapter 1
Measurement: System, Uncertainty and Response 1

Md. Zahurul Haq
Chapter 2
Internal Combustion Engine Indicating Measurements

Andr V. Bueno, Jos A. Velsquez and Luiz F. Milanez
Chapter 3
Measurement Systems
for Electrical Machine Monitoring 45
Mario Vrazic, Ivan Gasparac and Marinko Kovacic
Chapter 4
Experimental System for Determining

the Magnetic Losses of Super Paramagnetic
Materials; Planning, Realization and Testing 63
Milo Bekovi and Anton Hamler
Chapter 5
Non Contact Measurement System with

Electromagnets for Vibration Tests on Bladed Disks
Christian Maria Firrone and Teresa Berruti
Chapter 6
Study on Wireless Torque

Measurement Using SAW Sensors 109
Chih-Jer Lin, Chii Ruey Lin, Shen-Kai Yu,
Guo-Xing Liu, Chih-Wei Hung and Hai-Pin Lin
Chapter 7
Shape Measurement by
Phase-Stepping Method Using Multi-Line LEDs
Yoshiharu Morimoto, Akihiro Masaya,
Motoharu Fujigaki and Daisuke Asai
Chapter 8
Electro-Luminescence Based
Pressure-Sensitive Paint System
and Its Application to Flow Field Measurement
Yoshimi Iijima and Hirotaka Sakaue
137
153
77
23
VI
Contents
Chapter 9
Computer-Based Measurement
System for Complex Investigation
of Shape Memory Alloy Actuators Behavior 169
Marek Kciuk and Grzegorz Kapyta
Chapter 10
Calibration of Measuring Systems

Based on Maximum Dynamic Error 189
Krzysztof Tomczyk
Chapter 11
Determining Exact Point Correspondences

in 3D Measurement Systems Using Fringe Projection
Concepts, Algorithms and Accuracy Determination 211
Christian Bruer-Burchardt, Max Mller, Christoph Munkelt,
Matthias Heinze, Peter Khmstedt and Gunther Notni
Chapter 12
Design Methodology to Construct

Information Measuring Systems Built on Piezoresonant
Mechanotrons with a Modulated Interelectrode Gap 229
Alla Taranchuk and Sergey Pidchenko
Chapter 13
Non Invasive Acoustic Measurements for

Faults Detecting in Building Materials and Structures 259
Barbara De Nicolo, Carlo Piga, Vlad Popescu and Giovanna Concu
Chapter 14
Approximation and Correction

of Measuring Transducers Characteristics
Janiczek Janusz
Chapter 15
Chapter 16
Chapter 17
293
Planar Microwave Sensors for Complex Permittivity

Characterization of Materials and Their Applications
Kashif Saeed, Muhammad F. Shafique,
Matthew B. Byrne and Ian C. Hunter
Basics on Radar Cross Section
Reduction Measurements of Simple
and Complex Targets Using Microwave Absorbers
Marcelo A. S. Miacci and Mirabel C. Rezende
An Intelligent System for Efficient
Rigid Film Anticounterfeiting Inspection 377
Michael Kohlert, Christian Kohlert and Andreas Knig
319
351
Preface
There is no such thing as an easy experiment, nor is there any substitute for careful
experimentation in many areas of basic research and applied product development.
From Experimental Methods for Engineers by J. P. Holman
Measurement is a multidisciplinary experimental science. Measurement systems
synergistically blend science, engineering and statistical methods to provide
fundamental data for research, design and development, control of processes and
operations, and facilitate safe and economic performance of systems. In recent years,
measuring techniques have expanded rapidly and gained maturity, through extensive
research activities and hardware advancements.
With individual chapters authored by eminent professionals in their respective topics,
Applied Measurement Systems attempt to provide a comprehensive presentation on
some of the key applied and advanced topics in measurements for scientists, engineers
and educators. These book illustrates the diversity of measurement systems, and
provide in-depth guidance for specific practical problems and applications.
I wish to express my gratitude to the authors of the chapters for their valuable and
highly professional contributions. I am very grateful to Ms. Gorana Scerbe and Ms.
Mirna Cvijic, publishing process managers of the present project and the editorial and
production staff at InTech.
Finally, I wish to acknowledge and appreciate the patience and understanding of my
family.
Prof. Md. Zahurul Haq, Ph.D.

Department of Mechanical Engineering,
Bangladesh University of Engineering and Technology, Dhaka,
Bangladesh
1
Measurement: System, Uncertainty
and Response
Md. Zahurul Haq*
Bangladesh University of Engineering & Technology (BUET)

Bangladesh
1. Introduction
A measuring instrument transforms a measurand, i.e., a physical variable to be measured, to

provide comprehensible output. An instrument can be aptly analysed and synthesized as a
system which is a set of interconnected components functioning as a unit, while response is a
measure of such units delity to its purpose. A real instruments response is neither perfect
nor identical even under static and replicate-conditions. So appropriate approach to quantify
the deviation from the true value is required. It is now widely recognized that measurement
results should be expressed in terms of estimated value and an associated uncertainty value
obtained by proper analysis. System response analysis is further complicated in dynamic
measurements as an instrument does not respond instantaneously to an input that varies in
time. This obviously creates a measurement problem, and if these effects are not accounted
for, dynamic errors are introduced. Therefore, the performance of a measuring instrument is
to be specied in terms of both static and dynamic performance parameters.
2. Measurement system
Measurement is the act of assigning a specic value to a measurand. Mass, distance,
time, temperature, force, and other physical quantities, as well as the properties of matters,
materials, and devices, must be measured and described in common terminology. Measuring
instruments are designed to generate a xed and reproducible magnitude of the measurand
which is expressed by a number (the magnitude ratio) followed by the matching unit, e.g.,
a length of 2.5 m. So measurement provides quantitative information on the actual state of
the measurand that otherwise could only be estimated. ISO/IEC Guide 98:2008 Uncertainty of
Measurement Guide to the Expression of Uncertainty in Measurement (GUM) (ISO, 2008) reports
the following scope of applications of measurements:
To maintain quality control and quality assurance in production;

To comply with and enforcing laws and regulations;
To conduct basic/applied research and development, in science and engineering;
To develop, maintain and compare international and national physical reference standards,
reference materials, and also to achieve traceability to national standards.
* http://teacher.buet.ac.bd/zahurul/
Applied Measurement
Systems
Will-be-set-by-IN-TECH
According to ISO/IEC Guide 99:2007 International Vocabulary of Metrology Basic and General
Concepts and Associated Terms (VIM) (ISO, 2007), measurement is dened as:
process of experimentally obtaining one or more quantity values that can reasonably be
attributed to a quantity.
Thus measurement is an experimental science and most experiments are classied into
following four categories (Dunn, 2010):
1. Variational experiments.
These are carried out with an objective to establish the
mathematical relations between the experiments variables.
2. Validation experiments. These are carried out to validate a specic hypothesis.
3. Pedagogical experiments. These are aimed to demonstrate something that is already known.
4. Exploration experiments. These are conducted to explore an idea or possible theory.
2.1 General measurement system
It is sensible to have a generalized description of both the operation and performance of

a measuring instrument without recourse to any specic physical hardware. A measuring
instrument can be described in terms of its functional elements (Fig. 1); and these elements
form the bridge between the input to the measurement system and the system output, a
quantity that is used to infer the value of the measurand. Most measurement systems fall
within the general framework consisting of three functional stages (Holman, 2001):
1. Sensor-transducer stage. Sensor is directly affected by the measurand, while transducer
transduces the sensed information to provide an output quantity having a specied
relation to the input quantity. Examples of sensors-transducer include thermocouple,
strain gauge, manometer, load-cell, etc. There are three basic phenomenon in effect in
any sensor operation:
i) The change (or the absolute value) in the measurand causes an equivalent change in
the sensor property, e.g., displacement, voltage, resistance, capacitance, inductance,
magnetic ux, etc.
ii) The change in the sensor property is converted into a more usable form, e.g.,
temperature change results in the change in generated voltage by a thermocouple.
iii) The exposure of the sensor to the effects of the measurement environment may lead
to some exchange of energy to cause loading effect; e.g., a thermometer when inserted
into a cup of tea takes some heat from it to cause a difference between the true value
and the indicated value.
2. Signal-conditioning stage. Transduced signal is modied by one or more basic operations,
such as amplication, ltering, differentiation, integration, averaging, etc. for further
processing, i.e., display, storage or use in feed-back control systems, etc.
3. Output stage. It provides the information sought in a form comprehensible to one of the
human senses or to a controller (Beckwith et al., 2007). Output may be analogue or digital, it
may be displayed (using LCD or seven-segment display) or saved (using data-loggers) or
may be transmitted to a computer or controller (using data-acquisition system) for further
use.
33
Measurement:
System,
and Response
andUncertainty
Response
Fig. 1. Functional elements of an instrument or a measurement system.

A glass-bulb mercury thermometer may be analysed to illustrate the above concepts. Here,
mercury acts as the sensor whose volume changes with change in its temperature. The
transducer is the thermometer-bulb where the change in mercury volume leads to mercury
displacement because of the bulbs xed volume. The stem of the thermometer is the
signal-conditioner which physically amplies the mercury displacement and the graduated
scale on the stem offers the required temperature indication. So a pre-established relationship
between the input (temperature) and output scale is utilized which is obtained by calibration.
2.2 Instruments performance characteristics and calibration
Measuring instrument transforms the measurand into suitable output that is functionally
related to the input and the relationship is established by calibration. Calibration is dened
in ISO (2007) as:
operation that, under specied conditions, in a rst step, establishes a relation between the
quantity values with measurement uncertainties provided by measurement standards and
corresponding indications with associated measurement uncertainties and, in a second step, uses
this information to establish a relation for obtaining a measurement result from an indication.
Calibration provides the opportunity to check the instrument against a known standard and
subsequently to reduce errors in measurements. There is a hierarchy of standards which
arranges themselves in order of decreasing accuracy with the primary standards being the most
accurate (Doebelin, 2004). Tables 1 and 2, respectively, list hierarchy of standards and errors
associated with various levels of temperature standards.
Primary standard
Transfer standard
Local standard
Working standard
Maintained as absolute unit standard

Used to calibrate local standards
Used to calibrate working standards
Used to calibrate local instruments
Table 1. Hierarchy of Standards (Figliola & Beasley, 2011).

Level
Primary
Transfer
Working
Local
Method
Fixed thermodynamic points
Platinum RTD
Platinum RTD
Thermocouple
Uncertainty [C]
0
0.005
0.05
0.5
Table 2. Examples of Temperature Standards (Figliola & Beasley, 2011).

Calibration is the process of comparison of the output of a measuring system to the values of a
range of known inputs and the results may be expressed by a statement, calibration function,
Applied Measurement
Systems
calibration diagram, calibration curve, or calibration table. In some cases, it may consist
of an additive or multiplicative correction of the indication with associated measurement
uncertainty. Over the time, it is possible for the indicated values to drift and it makes
recalibration necessary at regular intervals. In general, calibration of measuring instruments
needs to be traceable to national standardizing laboratory. Hence, traceability is dened by ISO
(2007) as:
property of a measurement result whereby the result can be related to a reference through a
documented unbroken chain of calibrations, each contributing to the measurement uncertainty.
When the measurand maintains a steady value or slowly varies with time, system
performance can be described in terms of static characteristics. Systems dealing with rapidly
varying measurand require additional performance parameters termed dynamic characteristics.
System performing satisfactorily during static calibration may not provide correct results
during dynamic conditions. So relationship between the dynamic input and output must be
examined and dynamic performance parameters are required to provide satisfactory results.
3. Static characteristics and specications

When reporting the measurement result of a measurand, it is necessary that some quantitative
indication of the quality of the result be given so that those who use it can assess its reliability
(ISO, 2008). Without such an indication, measurement results cannot be compared, either
among themselves or with reference values given in a specication or standard. ASME
Power Test Codes (PTC) 19.1: Test Uncertainty (ASME, 2005) cites the following objectives of
uncertainty analysis:
To facilitate communication regarding measurement and test results;
To foster an understanding of potential error sources in a measurement system and the
effects of those potential error sources on test results;
To guide the decision-making process for selecting appropriate and cost-effective
measurement systems and methodologies;
To reduce the risk of making erroneous decisions; and
To document uncertainty for assessing compliance with agreements.
3.1 Measurement errors and uncertainties
Every measurement has error, which results in a difference between the measured value and
the true value. The difference between the measured value and the true value is the total error.
Hence, accuracy is dened as the degree of conformity of an indicated value to a recognized
accepted standard, or ideal or true value (ISO, 2007). Since the true value is unknown, total
error cannot be known and therefore only its expected values can be estimated.
ASME (2005) quanties the following two components of total error:
1. Random error, , is the portion of the measurement error that varies randomly in repeated
measurements throughout the conduct of a test (ISO, 2007). Random errors may arise
from uncontrolled test conditions and nonrepeatabilities in the measurement system,
measurement methods, environmental conditions, data reduction techniques, etc.
2. Systematic error, , is the component of measurement error that in replicate measurements
remains constant or varies in a particular manner (ISO, 2007). These errors may arise from
imperfect calibration corrections, measurement methods, data reduction techniques, etc.
Measurement:
System,
and Response
andUncertainty
Response
55
Random errors are identied and quantied through repeated measurements while trying
to keep the conditions constant and by statistical analysis of the results. Systematic errors
can be revealed when the conditions are varied, whether deliberately or unintentionally.
Figliola & Beasley (2011) reported the following four methodologies to reduce systematic
errors:
1. Calibration, by checking the output(s) for known input(s).
2. Concomitant method, by using different methods of estimating the same thing and
comparing the results.
3. Inter-laboratory comparison, by comparing the results of similar measurements.
4. Experience.
Errors in the measurement process are classied into four groups (ASME, 2005;
Figliola & Beasley, 2011) and examples of the error sources are reported in Table 3.
Error group
Error source
1. Calibration error
Standard or reference value errors

Instrument or system errors
Calibration process errors
Calibration curve t
2. Loading error
Interaction between the instrument and test media

Interaction between test article and test facility
3. Data-acquisition error Measurement system operating conditions

Sensor-transducer stage (instrument error)
Signal conditioning stage (instrument error)
Output stage (instrument error)
Process operating conditions
Sensor installation effects
Environmental effects
Spatial variation error
Temporal variation error
4. Data-reduction error Calibration curve t
Truncation error
Table 3. Error classications and sources.
It is now widely recognized that, when all of the known or suspected components of error
have been evaluated and the appropriate corrections have been applied, there still remains
an uncertainty about the correctness of the stated result, i.e., a doubt about the quality of the
result of the measurement. The word uncertainty means doubt, and thus in its broadest sense
uncertainty of measurement means doubt about the validity of the result of a measurement.
According to ISO (2007), uncertainty is dened as:
non-negative parameter characterizing the dispersion of the quantity values being attributed to
a measurand, based on the information used.
There are two widely accepted professional documents on uncertainty analysis:
1. ASME Power Test Codes (PTC) 19.1: Test Uncertainty (ASME, 2005), and
Applied Measurement
Systems
2. ISO/IEC Guide 98:2008 Uncertainty of Measurement Guide to the Expression of Uncertainty

in Measurement (GUM) (ISO, 2008).
These two documents differ in some terminology and how errors are catalogued. For example,
ASME (2005) refers to random and systematic error terms to classify errors by how they
manifest themselves in the measurement; ISO (2008) refers to Type A and Type B errors.
Type A uncertainties have data with which standard deviation can be calculated, while Type B
uncertainties do not have data to calculate a standard deviation and must be estimated by
other means. These differences are real but the nal result of an uncertainty analysis by either
method will yield a similar value (Figliola & Beasley, 2011).
3.2 Analysis of measurement data
Any single measurement of a parameter, x, is inuenced by different elemental random error

sources. In successive measurements of the parameter, the values of these elemental random
error sources change resulting in the evident random scatter. Measurement scatters common
in science and engineering are, in general, described by Normal or Gaussian distribution
which predicts that the scatter in the measured data-set will be distributed symmetrically
about some central tendency (Fig. 2). If an innite number of measurements of a parameter
were to be taken following the dened test process, the resulting population of measurements
could be described statistically in terms of the population mean, , and the population standard
deviation, ; and the true value of x would be (Doebelin, 2004).
Fig. 2. Illustration of measurement errors.

In practice, only nite-sized data are available. So measured data can only provide an estimate
of the true value (Figliola & Beasley, 2011). If the measured variable is described by discrete
data of size N, the mean value, x, of the measured value, x j is given by:
x=
1 N
xj
N j
=1
(1)
77
Measurement:
System,
and Response
andUncertainty
Response
, and the standard deviation is given by

Sx =
N
1
( x x )2
N 1 j =1 j
(2)
In computing the sample average, x, all N data are independent. However, the standard
deviation uses the result of the previous calculation for x. Hence, number of degrees
of freedom is reduced by one and in calculating the standard deviation, ( N 1) is used
instead of N. It may also be noted that, for innite number of measurements (N ):
x and S x . For nite data sets, the standard deviation of the mean, S x , can be estimated
from a single nite data set as:
Sx
(3)
Sx =
N
For a normal distribution of x about some sample mean value, x, it can be stated statistically
x j = x u x = x k(, )S x = x t(, )S x
( P%)
(4)
where the u x is the uncertainty interval or precision index associated with the estimate of x. The
value of u x depends on a constant known as coverage factor that depends on the probability
distribution function, the condence level, (which is expressed by probability, P%) and the
amount of data, N. For nite data set, k(, ) = t(, ). The Students t value for a given
probability, P% and degrees of freedom in data, = N 1, can be obtained from Table 4,
which is a short tabulation of the Students t distribution. The estimate of the true mean value,
xt , based on the nite data-set can be estimated (Figliola & Beasley, 2011) using:
xt = x k( )S x
( P%)
t50
t90
t95
t99
1
2
5
10
20
30
1.000
0.816
0.727
0.700
0.687
0.683
0.674
6.314
2.920
2.015
1.812
1.725
1.697
1.645
12.706
4.303
2.571
2.228
2.086
2.042
1.960
63.657
9.925
4.032
3.169
2.845
2.750
2.576
(5)
Table 4. Students t distribution (Figliola & Beasley, 2011).

The uncertainty interval, u x , assumes a set of measured values with only random error
present. Furthermore, the set of measured values is assumed to have unbound signicant
digits and to have been obtained with a measuring system having innite resolution. When
nite resolution exits and truncation of digits occurs, the uncertainty interval may be larger
than that predicted by the consideration of the random errors only. The uncertainty interval
can never be less that the resolution limits or truncation limits of the measured values.
To illustrate the above concepts, consider the data of Table 5, where 20 values of an arbitrary
measurement is presented.
Applied Measurement
Systems
9.78 10.29 9.68 10.30 9.69 9.63 10.35 9.88 10.29 9.72
9.85 10.25 9.75 10.24 9.89 10.31 9.94 10.35 10.27 9.65
Table 5. Sample of random variable, x.
Sample mean, x =
N
j =1 x j = 10.0

2
Standard deviation, S x = N11 N
j =1 ( x j x ) = 0.28
1
N
If a 21st data points are to be taken:

x j=21 = 10.00 (1.729 0.28) = 10.00 0.48
= 10.00 (2.093 0.28) = 10.00 0.59
(90%)
(95%)
So, there is a 95% probability that the 21st value will be between 9.41 and 10.59.
The true value, xt , isestimated by the sample mean value, x and the standard deviation of
the mean, S x = S x / N = 0.0626:
xt = 10.00 0.11
= 10.00 0.13
(90%)
(95%)
3.3 Uncertainty analysis: error propagation
In calibration experiments, one measures the desired results directly. In nearly all other
measurements, results are obtained through functional relationship with measured values. So,
it is necessary to compute the uncertainty in the results from the estimates of the uncertainty
in the measurements. This computation process is called the propagation of uncertainty.
Consider a result, R, which is determined through some functional relationship between
independent variables, x1 , x2 , , x N , dened by
R = R ( x 1 , x 2 , , x N )
(6)
where N is the number of independent variables and each variable contains some measure of
uncertainty to affect the nal result. The best estimate of the true mean value, Rt , would be
stated as:
( P%)
(7)
Rt = R u R
where the sample mean, R is found from
R = R ( x 1 , x2 , , x N )
and the uncertainty in R is found from

2
2

2
R
R
R
uR =
u1 +
u2 + +
uN
x1
x2
x N
(8)
(9)
where, u1 , u2 , , u N are the uncertainties associated with x1 , x2 , , x N .

To illustrate the concepts presented in the present section, let us consider the calculation of
electrical power, P from
P = VI
99
Measurement:
System,
and Response
andUncertainty
Response
, where, voltage, V and current, I are measured as

V = 100 V 5 V
I = 10 A 0.1 A
Hence, the nominal value of power is 1000 W. By taking the worst possible variations in
voltage and current, we would calculate
Pmax = (100 + 5)(10 + 0.1) = 1060.5 W
Pmin = (100 5)(10 0.1) = 940.5 W
Thus, using the simple calculations, the uncertainty in the power is +6.05%, 5.95%.
However, it is quite unlikely that the power would be in error by these amount because the
variation of voltage reading would probably not be corresponding with the current reading.
Hence, it is quite unlikely these two independent parameters would achieve maximum or
minimum values simultaneously. Hence, using Eq. 9:
P
R
=
= I = 10 A
x1
V
uR =
u1 = uV = 5 V
R
P
=
= V = 100 V
x2
I
u2 = u I = 0.1 A
(10 5)2 + (100 0.1)2 =
2500 + 100 = 51.0 W = 5.1%
Hence, estimated power, P = 1000 W 5.1%.

If we ignore the uncertainty contribution due to the current reading (u I 0), the uncertainty
in power measurement would be 5%. Hence, very little is gained by trying to reduce the
small uncertainties. Any improvement in result should be achieved by improving the
instrumentation or technique connected with relatively large uncertainties.
4. Specication of dynamic systems and response

A measuring instrument requires a certain amount of time to achieve complete response. In
case of a dynamic measurement, actual response undergoes some attenuation, delay and
distortion. A measurement system should be capable of yielding the true value(s) from
the time varying input value(s). Study of dynamic characteristics of such system requires
considerable analyses and the responses are conveniently stated in terms of step response,
harmonic response and frequency response of the instrument.
4.1 Modelling of measuring instruments
The response of a measurement system, i.e., output, x (t), when subjected to an input forcing
function, f (t), may be expressed by a linear ordinary differential equation with constant
coefcients of the form (Doebelin, 2004):
0th order
dn x
d n 1 x
d2 x
dx

+ a0 x = f ( t )
a n n + a n 1 n 1 + + a2 2 + a1
dt
dt
dt
dt

1st order
2nd order
(10)
10
Applied Measurement
Systems
10
f (t) Input quantity imposed on the system,

where, x (t) Output or the response of the system,
as Physical system parameters, assumed constants.
Hence, the order of the system is designated by the order of the differential equation. Once the
governing equation of the instrument is established, its response can be obtained if the input
(measurand) is a known function of time. Dynamic response of a system can be estimated by
solving Eq. 10 with proper initial and boundary conditions. While the general model of Eq. 10
is adequate for handling any linear measurement system, certain special cases (e.g., zero-,
rst- and second-order systems) occur so frequently in practice that warrant special attention.
Furthermore, complex systems may protably be considered in terms of the combinations of
simple cases and the response can be inferred from the observations made from these cases.
4.1.1 Zero-order instrument
The simplest case of Eq. 10 occurs if all the values of as other than a0 are assumed to be zero,
and Eq. 10 is then transformed into a simple algebraic equation:
a0 x = f (t) : = x (t) = k f (t)
(11)
The constant, k 1/a0 is called the static sensitivity of the system, and it represents the scaling
between the input and the output. For any-order system, static sensitivity always has the
same physical interpretation, i.e., the amount of output per unit input when the input is static
and under such condition all the derivative terms of Eq. 10 are zero.
A practical example of a zero-order instrument is a displacement measuring potentiometer
where a strip of resistance material is excited with a voltage, Vs , and provided with a sliding
contact responding to displacement. If the resistance is linearly distributed along the length,
L, the output voltage, eo (t), may be written as a function of displacement, l (t), as:
eo (t) =
Vs
l (t) = k l (t)
L
(12)
where, k Vs /L (Volts per unit length).

4.1.2 First-order instrument
If in Eq. 10, all as other than a1 and ao are taken as zero, we get:
a1
where,
dx
dx
+ a0 x = f (t) :=
+ x = k f (t)
dt
dt
(13)
k 1/ao static sensitivity,

a1 /ao time-constant.
The time constant has the dimension of time, while the static sensitivity has the dimension of
output divided by input. When time constant of a system is very small, the effect of the
derivative term in Eq. 13 becomes negligible and the governing equation approaches to that
of a zero-order system. It is experimentally observed that an instrument with small value of
the time-constant is associated with good dynamic response.
11
11
Measurement:
System,
and Response
andUncertainty
Response
A practical example of a rst-order instrument is a mercury-in-glass thermometer. If a

thermometer, initially at room temperature, is inserted into a hot uid at temperature, T ,
then the convective heat gain from the hot uid will result in the increase of internal energy
of mercury at the same rate. The thermometer takes a while to give the correct reading, and
the temperature reading, T (t), can be obtained using the following energy-balance equation:
dT (t)
dT (t)
+ T (t) = T
Q in = hA [ T T (t)] = mC
: =
dt
dt
(14)
Hence, mC/hA and k = 1; and h is the convective heat transfer coefcient, A is the
surface area over which heat is transferred, m is the mass of mercury, and C is its specic heat.
4.1.3 Second-order instrument
A second-order instrument is one that follows the equation:

a2
d2 x
dx
1 d2 x
dx
+
a
+
a
x
=
f
(
t
)
:
=
+2
+ x = k f (t)
o
1
dt
n dt
dt2
n2 dt2
(15)
k 1/a
o static sensitivity,
a
where, n a2o undamped natural frequency,
2aa1 a dimensionless damping ratio.
o 2
A practical example of a second-order instrument is a weight measuring spring balance as

shown in Fig. 3. Hence, k is the spring constant of the ideal spring and b is the friction constant
where damper friction force is linearly proportional to the velocity of mass, x (t). Summing
the forces acting on the mass and utilizing the Newtons second law of motion yields,
f (t) kx b
dx
d2 x
dx
1 d2 x
= m 2 := 2 2 + 2
+ x = k f (t)
dt
n dt
dt
n dt
k spring constant
b damping constant
b
Spring
(16)
f (t) forcing function

m mass
bx(t)
kx(t)
Damper
x(t)
x(t)
x
(t)
+ ve
f (t)
f (t)
(a)
m
x(t)
(b)
Fig. 3. (a) Spring-mass-damper system, (b) Free-body diagram.

Final form of Eq. 15, could represent, among other things, an electrical
resistor-inductor-capacitor (RLC) circuit. The variables of the electric circuit behave exactly
as the analogous variables of the equivalent mechanical system. The spring-mass-damper
system and analogous RLC circuit are illustrated in Fig. 4. Both of these systems share the
same governing equation and therefore have analogous response when subjected to input
forcing.
12
Applied Measurement
Systems
12
v(t) applied voltage (V)

L inductance (H)
C capacitance (F)
R resistance ()
q charge (C)
i dq/dt current (A)
f (t) forcing function (N)

m mass (kg)
k spring constant (N/m)
b damping constant (N.s/m)
x displacement (m)
x dx/dt velocity (m/s)
R
Resistor
k
Spring
x(t)
x(t)
x
(t)
+ ve
v(t)
Damper
Inductor
C
Capacitor
m
2
f (t)
1
L ddt2q + R dq
dt + C q = v(t)
L m, R b, C1 k, v f
m ddtx2 + b dx
+ kx = f (t) =
dt
1 d2 x
n2 dt2
+ 2 n dx
dt + x = kf (t)
Fig. 4. Spring-mass-damper system and analogous RLC circuit.

4.2 Modelling of response of measuring instruments
Once the governing equation of a system is established, its dynamic response to forcing
element can be obtained if the forcing can be represented as a function of time. The
fundamental difculty in this approach lies in the fact that the quantities to be measured
usually do not follow some simple mathematical functions. However, it is fruitful to study
a systems response to common forcing inputs, e.g., step and harmonic inputs. These forcing
functions have been found to be very useful in the theoretical and experimental analyses of
measurement systems response. Mathematically, the step function of magnitude, A, can be
expressed as:

0
at t = 0
(17)
f (t) =
A
for t > 0
, and the harmonic function of circular frequency, , can be expressed as:

0
at t = 0
f (t) =
A sin t for t > 0
(18)
Graphical representation of these two functions are shown in Fig. 5.

4.2.1 Response of zero-order instruments
The response of a zero-order instrument is governed by a simple algebraic equation (Eq. 11).
So the output, x (t), follows the input, f (t), perfectly with no distortion or time lag. Hence,
zero-order instrument represents ideal dynamic performance, and is thus a standard against
which less perfect or dynamic instruments are compared. Typical responses of a zero-order
instrument are shown in Fig. 6.
13
13
Measurement:
System,
and Response
andUncertainty
Response
Fig. 5. Step and harmonic inputs.
Fig. 6. Zero-order instruments response for step and harmonic inputs (for k = 0.75).
Zero-order system concept is used to analyse real systems response in static calibration. When
dynamic signals are involved, higher order differential equations are required to estimate
the dynamic response as most of the real systems possess inertial/storage capability and are
subjected to some viscous resistance and dissipation (Figliola & Beasley, 2011).
4.2.2 Response of rst-order instruments
Response for Step Input: If the governing equation for rst-order system (Eq. 13) is solved
for a step input function and with initial condition x | t=0 = x0 , the solution is:
x (t) = ( xo Ak) exp( t/ ) +

transient response
A
k
(19)
steadystate response
For large values of time, x (t ) = Ak x . Hence, x steady-state response of the

system. Equation 19 can be rearranged to give non-dimensional response, M (t), as:
M (t) =
x (t) xo
= 1.0 exp( t/ )
x xo
(20)
, where, M (t) is the ratio between output and input amplitudes. Non-dimensional Eq. 20 is
valid for all rst-order systems response to step-input forcing; and plot of Eq. 20 is shown in
14
Applied Measurement
Systems
14
Fig. 7. It is observed that, the response approaches the steady-state value monotonically and
the response reaches 63.2% of its steady-state value when t = . When elapsed time is 2,
3, and 4, the response is 86.5%, 95% and 98%, respectively. It is also seen that, the slope of
the response curve (Fig. 7) at the origin is 1.0, i.e., if the initial response rate were maintained,
the response would be completed in one time constant. However, the nal value (steady-state
response) is achieved only after innite time (Fig. 7). In practice two response specications
are used to describe the rst-order system response:
1. Settling time, ts . It is dened as the time required for the system to reach and stay within
the 2% of the nal value. Hence, for rst-order system, ts 4.
2. Rise time, tr . It is dened as the time required for the response to go from a small percentage
(10%) to a large percentage (90%) of the step input. Hence, for rst-order system, tr 2.2.
Fig. 7. First-order system response to a unit step-input.

Response for Harmonic Input: If the governing equation for rst-order system (Eq. 13) is
solved for harmonic input and x | t=0 = 0, the solution is:
1
x (t)
=
exp( t/ ) +
sin(t + )
Ak
1 + ( )2
1 + ( )2

(21)
steadystate response
transient response
where, tan 1 ( ) phase lag. Hence, time delay, t, is related to phase lag as:
t =
(22)
For >> 1, response is attenuated and time/phase is lagged from input, and for
<< 1, the transient response becomes very small and response follows the input with
small attenuation and time/phase lag. Ideal response (without attenuation and phase lag)
is obtained when the system time constant is signicantly smaller than the forcing element
15
15
Measurement:
System,
and Response
andUncertainty
Response
period, T. As t , the rst term on the right side of Eq. 21 vanishes and leaves only the
steady-state solution:
Ak
Ak
sin(t + ) =
sin (t + t) = f (t) Ga
(23)
1 + ( )2
1 + ( )2

Hence, Ga k/ 1 + ( )2 steady-state gain. Equation 23 indicates that the attenuated
steady-state response is also a sine wave with a frequency equal to the input signal frequency,
, and it lags behind the input by phase angle, .
x (t)| s =
To illustrate the above concepts, the response of a thermocouple subjected to a harmonic

temperature variation is plotted in Figs. 8 and 9; and summary results are also reported in
Table. 6. Initial transient response is clearly visible and steady-state response is achieved only
after the expiration of 4 time. The response is of same frequency but attenuated in magnitude
and lags behinds the input signal. However, once the steady-state is achieved, knowing the
values of the gain, Ga , and phase lag, , it is possible to estimate the correct input value,
and an example is also shown in Fig. 8. The effect of time constants on dynamic response is
illustrated in Fig. 9 where response from three thermocouple systems with time constants of 1,
5 and 50 s are plotted. Figure 9 provides a meaningful insight into the effects of time constant
as it shows that systems with small time constants follow the input with less attenuation and
time delay, and vice versa. However, when average value of a measurand is desired (e.g., in
chemical reactant mixing chamber), system with higher time constants are suitable and near
average result is observed as shown in Fig. 9 for a thermocouple with = 50 s.
Unit
01
02
03
[s]
01
05
50
/T
0.04
0.2
2.0
[deg]
-14.0
-51.3
-85.4
t [s]
-0.98
-3.58
-5.96
Ga
0.97
0.62
0.08
Table 6. Thermocouples response to harmonic temperature variation.

4.2.3 Response of second-order instruments
Response for Step Input: For a second-order system subjected to a step input, the solution
(response) is readily available (Dunn, 2010) and may be written as:
1 cos( n t)
for = 0
1 1 exp( n t) sin[ n t 1 2 + cos1 ( )]
for 0 < < 1
1 2
x (t)
= 1 (1 + n t) exp( n t)
kA

for = 1

1 exp( n t) cosh( n t 2 1) + 2 sinh( n t 2 1) for > 1

1
(24)
The nature of the response of the second-order system for a step input depends on the value
of the damping ratio, , as depicted in Fig. 10, and four types of response may be identied:
1. Harmonic oscillation ( = 0). Response oscillates at the natural frequency, n , and the
oscillations are undamped.
16
16
Applied Measurement
Systems
Fig. 8. Response of thermocouple for harmonic temperature input.
Fig. 9. Effect of time constant on thermocouples response to harmonic temperature input.

2. Underdamped response (0 < < 1). Response overshoots the steady-state value initially,
and then eventually decays to the steady-state value. The smaller the value of , the larger
the overshoot.
3. Critically damped response ( = 1). An exponential rise in response occurs to approach the
steady-state value without any overshoot, and the response is as fast as possible without
overshoot.
4. Overdamped response ( > 1). Response approaches the steady-state value without
overshoot, but at a slower rate. Hence, excessive time is required to complete a
measurement and therefore frequency at which the measurement is possible is limited.
So, little attention is focused on the overdamped systems for dynamic measurements.
Measurement:
System,
and Response
andUncertainty
Response
17
17
Fig. 10. Transient response of a second-order system for a step-input at t = 0.

For underdamped system, as decreases, response becomes increasingly oscillatory. The
transient response oscillates about the steady-value and occurs with a period,Td, given by:

2
Td
:
d n 1 2
(25)
d
, where, d ringing frequency. Standard performance parameters in terms of step response
is shown in Fig. 11. It is observed that, the duration of the transient response is controlled
by n . In fact, its inuence is equivalent to that of a time constant in a rst-order system,
such that we could dene a second-order time constant as e 1/ n . The system settles to
steady state value, x = k A, more quickly when it is designed with large n , i.e., small e .
Nevertheless, for all systems with > 0, the response will eventually indicate the steady value
as t (Figliola & Beasley, 2011). Like a rst-order system, the swiftness of the response
may be described by settle time, ts and rise time, tr . For second-order system, ts 4e and
tr n
= 2.16 + 0.60 for 0.3 0.8 (Dorf & Bishop, 1998).
Equation 24 also reveals that for second-order system, whenever n appears, it appears as the
product of n t. This means that, if we say, double n , the same range of the response will
occur in exactly one half the time. Thus n is a direct and proportional indicator of response
speed. For a xed value of n , the speed of response is determined by (Doebelin, 1998).
4.2.4 Response in time-domain and postscript
The response of a zero-order system is an ideal one and response follows the input without
attenuation and phase lag. In practice, no measurement system is ideal and therefore does
not possess zero-order characteristics. First-order system follows the input signal with some
attenuation and phase lag, but no oscillations are present. Second-order system responses
are further complicated as in addition to output attenuation and phase lag, it may exhibit
oscillation. The total response of a non-zero-order dynamic system due to a harmonic input
is the sum of transient response, which is independent of the frequency of the input, and a
18
18
Applied Measurement
Systems
Fig. 11. Second-order underdamped response specications.

response which depends on the frequency of the input (Dorf & Bishop, 1998). For a stable
system, the transient response decays exponentially and eventually becomes insignicant
compared to the response that is specic to the input. When this occurs, the system has
reached the steady-state response (Kelly, 2003). Time-domain analysis is often complicated
because of the initial transient-response. However, in a measurement system, we are
interested in the steady-state response where the system responds to the input without being
affected by the transient effect of the system itself.
4.3 Frequency response of instruments
A very practical and convenient alternative to time-domain modelling to capture the dynamic
response of linear system is through its frequency response which is dened as the steady-state
response of the system to a sinusoidal input. The sinusoidal is a unique input signal, and
the resultant output signal for a linear system as well as signal throughout the system, is
sinusoidal in the steady-state; it differs from the input wave-form only in amplitude and
phase angle (Dorf & Bishop, 1998). Moreover, almost all types of functions can be described
through Fourier analysis in terms of the sums of sine and cosine functions. So, if a linear
systems response for sinusoidal input is determined, then its response to more complicated
input can be described by linearly superimposing the outputs determined for each of the
sinusoidal-input that were identied by Fourier analysis. In frequency response analysis,
sinusoidal input signal is varied over a range of frequencies, and for each frequency, there
is a gain and phase angle that give the characteristic response at that frequency. The results
are plotted in a pair of graphs known as Bode diagram which consists of two plots:
1. Logarithmic gain, L ( ) 20 log10 Ga ( ) vs. log10 ( ), and
2. Phase angle, ( ) vs. log10 ( )
The vertical scale of the amplitude Bode diagram is in decibels (dB), where a non-dimensional
frequency parameter such as frequency ratio, (/ n ), is often used on the horizontal axis.
19
19
Measurement:
System,
and Response
andUncertainty
Response
Laplace transformation (LT) is the key to frequency-domain modelling. It is a mathematical

tool to transform linear differential equations into an easier-to-manipulate algebraic form. In
frequency domain, the differential equations are easily solved and the solutions are converted
back into time-domain to provide system response (Fig. 12). Some of the common Laplace
transform (LT) pairs are reported in Table 7.
Frequency-domain
Time-domain
Dierential Equation
Input, f (t)
Output, x(t)
L{}
Algebraic Equation
Input, F (s)
Output, X(s)
L {}
Calculus
Algebra
Multiplication
Division
Exponentiation
Addition
Subtraction
Multiplication
Fig. 12. Laplace transform (LT) takes a function from time-domain to frequency-domain.
Inverse LT takes a function from frequency-domain to time-domain.
f (t)
F (s)
(t), unit impulse

Step function, A
At; Atn
Ae at
A sin(t)
cos(t)
f (t)
f (t)

n 2 en t sin n 1 2 t, < 1
1
A/s
A ; An
s2 s n +1
c1 f 1 ( t ) + c2 f 2 ( t )
c1 F1 (s) + c2 F2 (s)
A
s+a
A
s2 + 2
s
s2 + 2
sF (s) f (0)
s 2 F ( s ) s f (0) f (0)
2n
s2 +2 n s + 2n
Table 7. Laplace transform (LT) pairs of some common functions.

4.3.1 Laplace transform (LT) and Transfer function (TF) methods
Transfer function (TF) is widely used in frequency-domain analysis and it establishes the
size and timing relationship between the output and the input. Transfer function of a linear
system, G (s), is dened as the ratio of the Laplace transform (LT) of the output variable,
X (s) L{ x (t)}, to the LT of the input variable, F (s) L{ f (t)}, with all the initial conditions
are assumed to be zero. Hence,
X (s)
G (s)
(26)
F (s)
20
Applied Measurement
Systems
20
The Laplace operator, s + j, is a complex variable. Hence, j 1, and translates

into a sinusoid in time-domain, translates into an exponential term, exp(t). For steady-state
sinusoidal input, = 0, and system response can be evaluated by setting s = j. So,
F (s) G (s) X (s)
G ( j )
G ( )
Hence, a
( )
Ga ( )( )
G (s)| s = j
| G ( j )|
G ( j )
G a
: =
x ( t ) = f ( t ) G a
steady-state transfer function,

gain frequency response,
phase frequency response,
frequency response.
For example, transfer function of a rst-order system is evaluated below:

time-domain equation: dx
dt + x = k f ( t )
s-domain equation: sX (s) + X (s) = k F (s)
F (s) G (s) =
k
X (s)
1 + s

Ga = | G (s j )| = 1+kj = k 2
1+( )
= G ( j ) = tan1 ( )
For input, f (t) = A sin(t), steady-state response, x (t) = k A
1+( )2
sin(t + ).
Expressions of G (s), Ga and for zero-, rst- and second-order systems are reported in Table 8.
Differential Equation
Transfer function, G (s)
Gain, Ga
Phase angle,
Zero-order
First-order
x (t) = k f (t) dx
dt + x = k f ( t )
k
s +1
k
0
k
1+( )2
tan1 ( )
1 d2 x
2n dt2

Second-order
+ 2 n dx
dt + x = k f ( t )
k
1 2
s +2 n
2
n
1 ( n )
k

2 2
tan 1
s +1
+4 2 ( n )
2 ( n )
1 ( n )
Table 8. Transfer function, gain and phase angle of zero-, rst- and second-order systems.
4.3.2 Bode-diagram and selection of instrumentation parameters
Ideal system response of a zero-order system is shown in Fig. 13, where both the values of
L ( ) and ( ) are zero. In case of a rst-order system response, as shown in Fig. 13, when
the measurement system responds with values of L ( ) 0, system transfers all or nearly all
of the input signal amplitude to the output and with very little time delay. At larger values
of , measurement system will essentially lter out frequency information of the input by
responding with smaller amplitudes and larger time delays. So, any combination of will
produce the same results: if one wants to measure signals with high frequency content, then
a system with an appropriately small is required and vice versa (Figliola & Beasley, 2011).
Measurement:
System,
and Response
andUncertainty
Response
21
21
Fig. 13. Bode-diagram for zero- and rst-order systems.

In a second-order system response, as shown in Fig. 14, at low values of / n , values of
L ( ) and ( ) remain near to zero. This indicates that information concerning the input
signal of frequency, , will be passed through to the output with little attenuation and phase
lag. This region of frequency response curves is called the transmission band. The actual extent
of the frequency range for near unity gain depends on the system damping ratio, . The
transmission band of a system is typically dened as 3 dB L ( ) 3 dB. At large values
of / n , the system will attenuate the amplitude information of the input signal and a large
phase shift occurs (Figliola & Beasley, 2011). For a system with 0, L ( ) will be very
high and ( ) 180o in the vicinity of / n = 1. This behaviour is a characteristics of
Fig. 14. Bode-diagram for second-order systems.
22
22
Applied Measurement
Systems
system resonance. However, real systems possess some damping to limit the abruptness and
magnitude of resonance, but underdamped systems will experience some resonance.
For second-order systems, response shown in Fig. 14, following observations can be made:
System has a good linearity for low values of and up to a frequency ratio, / n , of about
0.3 as the amplitude gain is very nearly unity (Ga 1) with L ( ) 0.
For large values of , the amplitude is reduced substantially.
The phase shift characteristics are a strong function of / n for all frequencies.
As a general rule of thumb, the choice of = 0.707 is optimal since it results in the
best combination of amplitude linearity and phase linearity over the widest range of
frequencies (Holman, 2001).
The universal curves (Figs. 13 and 14) may be used as guidance in the selection of
measurement instruments and system components. As the values of L ( ) and ( ) deviate
from zero, these curves take on rather steep slopes, and in these regions, small errors in the
predictions of and the deviation of the real system from ideal one may lead to signicant
errors in measurements (Figliola & Beasley, 2011).
5. References
ASME (2005). ASME PTC 19.1-2005: Test Uncertainty. Amarical Society of Mechanical
Engineers (ASME), New York, USA.
Beckwith, T., Marangoni, R. & Lienhard, J. (2007). Mechanical Measurements, 6th edn, Pearson
Prentice Hall.
Doebelin, E. O. (1998). System Dynamics: Modeling, Analysis, Simulation, Design, Marcel Dekker,
Inc. NY, USA.
Doebelin, E. O. (2004). Measurement Systems: Applications and Design, 5th edn, McGraw-Hill.
Dorf, R. & Bishop, R. (1998). Modern Control Systems, 8th edn, Pearson Prentice Hall.
Dunn, P. F. (2010). Measurement and Data Analysis for Engineering and Science, 2nd edn, CRC
Press.
Figliola, R. & Beasley, D. (2011). Theory and Design for Mechanical Measurements, 5th edn, John
Wiley.
Holman, J. (2001). Experimental Methods for Engineers, 7th edn, McGraw-Hill.
ISO (2007). ISO/IEC Guide 99:2007 International Vocabulary of Metrology Basic and
General Concepts and Associated Terms (VIM). International Organization for
Standardization (ISO), Geneva, Switzerland.
ISO (2008). ISO/IEC Guide 98:2008 Uncertainty of Measurement Guide to the Expression of
Uncertainty in Measurement (GUM). International Organization for Standardization
(ISO), Geneva, Switzerland.
Kelly, S. G. (2003). System Dynamics, Cengage Learning.
2
Internal Combustion Engine
Indicating Measurements
Andr V. Bueno1, Jos A. Velsquez2 and Luiz F. Milanez3
2Federal
1Federal
University of Ceara, UFC,

University of Technology - Parana, UTFPR,
3University of Campinas, UNICAMP,
Brazil
1. Introduction
Engine indicating includes the measurement of instantaneous in-cylinder pressure, the
determination of the top dead centre (TDC) and the measurement of the instantaneous
crank angle (c.a.). These measurements are fundamental for engine combustion diagnosis
and for indicated work calculation.
In engine combustion diagnosis, the apparent heat release rate and the combustion reaction
extent are the most useful quantities obtainable from engine indicating data. The apparent
heat release rate is calculated by computing the amount of fuel chemical energy release
necessary to obtain the experimentally observed pressure, while the combustion reaction
extent is evaluated through the released fraction of the total fuel chemical energy. Heat
release analysis is often complemented using optical techniques and its utilization as a
diagnostic tool covers a wide range of objectives, including the development of new
combustion systems, the analysis of alternative fuel burning, the validation of mathematical
models for engine simulation, the investigation of combustion chamber insulation effects
and the study of new injection strategies.
Taking into account the importance of indicating measurements for engine research and
development, this chapter addresses the description of a modern engine indicating
measurement system and includes recommendations about good-practice procedures and
cares that a researcher must observe in order to perform accurate experiments and to obtain
valuable information from measured data.
1.1 A word on the evolution of engine indicating measurement systems
The measurement of the working fluid pressure of heat engines was a topic of interest for
engineers since the advent of the steam engine, for which the Watts indicator was
developed. When the internal combustion engine became the most widespread heat
machine, its analyses and improvements also demanded the measuring of in-cylinder
pressure data and early measurements were accomplished utilizing several configurations
of mechanical indicators (Amann, 1985). However, as the operating speeds increased with
24
the development of the engines, the frequency response of such mechanical indicators
became deficient, making them obsolete in the mid-sixties of the last century.
High-speed electronic transducers, capable of converting the deflection of a low inertia
diaphragm into an electrical signal were designed to meet the demand for pressure
measuring instruments with superior characteristics. Early versions of such devices already
had adequate frequency response to the phenomena occurring in the engine combustion
chamber, having been built using extensometers (Draper & Li, 1949) and piezoelectric
crystals (Kistler, 1956) as sensing elements. Nevertheless, these new electronic pressure
transducers were initially coupled to analog data acquisition systems made of a signal
conditioning amplifier, a cathode-ray oscilloscope and a photo camera that was used to
record the pressure signal from the oscilloscope screen (Brown, 1967). This data generating
procedure was cumbersome and had a number of uncertainties related to the film exposure
time as well as to the oscilloscope trace (Benson & Pick, 1974).
At the end of the 1960s, complex analog systems capable of carrying out a fully electronic
processing of the transducer signal became available. These devices were firstly used for
specific applications, such as determining the indicated power (Alyea, 1969), indicated mean
effective pressure (Brown, 1973), as well as studying knock and misfire in spark ignition
engines (Alwood et al., 1970). The user interface was through a voltmeter, which displayed a
voltage proportional to the indicated mean effective pressure or, alternatively, by means of
an electromechanical counter that showed the number of cycles in which knock or misfire
occurred.
In the mid-seventies, when analog-to-digital converters were included in the
instrumentation, multipurpose and of-less-complexity experimental sets became available
(Benson & Pick, 1974; Ficher & Macey, 1975; Marzouk & Watson, 1976). From this time, the
signal from instrumentation amplifiers has been digitalized and stored in a computer,
allowing further manipulation via software. Therefore, greater storage capacity and
flexibility in data analysis were assured, maintaining an acceptable level of accuracy.
Nowadays these are typical characteristics of a modern engine indicating system, as the one
schematically represented in Figure 1, which includes components suitable for
measurements in diesel engines. In this scheme, the fuel injection line pressure is measured
with a strain-gauge based sensor connected to an instrumentation amplifier, while the signal
generated by a piezoelectric pressure sensor can be conditioned following two measurement
procedures, which are addressed in detail in the present chapter. In the first procedure, the
in-cylinder pressure is obtained by employing a charge amplifier for conditioning the
transducer signal, while in the second procedure a current-to-voltage converter is used to
measure the in-cylinder pressure derivative.
Pressure data are indexed with the angle position of the crankshaft, with reference to the
compression TDC. Usually the crankshaft angle position is determined with an optical angle
encoder, which provides both one pulse per revolution in a channel used to establish the
TDC angle reference, and 720 pulses per revolution in a second channel to determinate the
instantaneous relative angle position. External pulse multipliers may also be available to
improve the relative angle position resolution up to 3600 pulses per revolution. Each angle
position pulse triggers a high speed data acquisition system, which should be able to
simultaneously acquire the signals provided by the conditioning amplifiers, collect the
acquired data among multiple cycles for cycle averaging, and save it in a storage computer.
25
Fig. 1. A typical engine indicating measurement system.
2. The transducer for in-cylinder pressure measurement

The analysis of the processes occurring in the cylinder of internal combustion engines
requires pressure transducers with high specifications regarding linearity, frequency
response and resistance to thermal solicitations. Studies comparing transducers available at
the end of 1960 decade (see for example Brown, 1967) found that those having piezoelectric
crystals as measuring elements exhibited better tolerance to thermal solicitations than those
based on strain gauges. By this reason, piezoelectric transducers eventually spread to
measure the in-cylinder pressure, while sensors based on strain gauges (metal or piezoresistive) were preferably used in measurements where the thermal solicitations are modest,
such as pressure measuring in the fuel injection line and in the intake manifold.
Piezoelectric transducers are capable of maintaining high characteristics in frequency
response and linearity over a wide range of pressures. On the other hand, the main reported
drawbacks to their use include instability of the baseline and low intensity of its output
signal. These drawbacks and their control will be discussed in Section 2.4, where the
techniques for conditioning the pressure transducer signal will be presented.
2.1 Operating principle of the piezoelectric pressure transducer
Figure 2 illustrates the principle of operation of a piezoelectric pressure transducer. The
pressure change rate (dP/dt) experienced by the transducer diaphragm is transmitted to a
piezoelectric crystal through intermediate elements, causing its deformation at a rate d/dt.
Due to the piezoelectric effect, this deformation polarizes charge q in the transducer
electrode originating an electric current i, which constitutes the transducer output signal:
dq
dP
Gs
dt
dt
where Gs is the transducer sensitivity (gain).
(1)
26
During the measurement of in-cylinder pressure, the transducer is exposed to a transient

heat flow that causes continuous changes in its temperature. These temperature changes
modify the sensitivity of the piezoelectric element and impose thermal stresses in the
diaphragm and in the sensor housing, generating spurious forces that act on the quartz
element and cause additional distortion of the signal provided by the transducer. The
inaccuracy due to these effects receives the name of temperature drift (Marzouk &
Watson, 1976).
It is usual to separate the temperature drift into two components. The first one corresponds
to changes in heat flow that occur within each cycle, and is named short-term drift or
thermal shock. The second component represents the slow change in the transducer
temperature resulting from variations in the engine operating conditions and is known as
load-change drift or long-term drift.
Normally, the long-term drift only promotes a slow instability in the baseline for transducer
signal, whose consequences and control depend on the circuitry chosen for its polarization.
Fig. 2. The piezoelectric pressure transducer.

The effects of the short-term drift, in turn, depend on the intensity with which this
phenomenon occurs. A severe short-term, such as that resulting from operating conditions
that can permanently damage the transducer, may produce pressure values below the
atmospheric pressure at the end of the expansion process. At more restrained levels,
however, the presence of short-term drift cannot be readily identified, causing the pressure
reading to be higher than the actual in-cylinder pressure during combustion, and lower
throughout the remainder of the expansion. Although modern piezoelectric sensors are
designed to minimize the effects of short-term drift, it must be taken into account that the
intensity of the latter strongly depends on the thermal load at the transducer location. This
thermal load is influenced by the intense flows during the gas exchange process, by the
approximation of the fuel jet in a diesel engine or by the flame front approximation in a
spark-ignition engine. Thus, checking the extent of the thermal shock at the transducer
location in each particular case is a good practice for an accurate measuring.
27
2.2 Choice of the transducer mounting location

The installation of the piezoelectric pressure transducer must be preceded by the calibration
of the complete measuring chain formed of the piezoelectric transducer, the signal
conditioning amplifier and the data acquisition system. It is possible to use a dead-weight
calibrator for this task in accordance with the recommendations made by Lancaster et al.
(1975). The choice of the location where the transducer will be mounted should prioritize
well-cooled regions of the head and avoid these where thermal stresses can induce
deformations in the transducer housing. The diaphragm of the sensor should be positioned
as recommended by the manufacturer (usually, with a gap of 1.5 to 3.0 mm from the inner
surface of the head). Water-cooled transducers feature superior gain (higher signal/noise
ratio), linearity and thermal resistance respective to the miniature uncooled transducers, and
should be the first choice for use in cases where there is enough space in the head. Channels
connecting the combustion chamber with the cavity where the sensor diaphragm is located
may behave as acoustic resonators, generating pressure oscillations and their consequent
measurement inaccuracies, which can invalidate the evaluation of both, the indicated
thermodynamic parameters and the combustion energy release. So, the use of these channels
(as typically occurs when the sensor is embedded in a spark plug) is recommended only for
the identification of abnormal combustion in spark ignition engines.
In-cylinder pressure measurements in direct injection (DI) diesel engines demand extra care
due to the high compression ratio and to the shape of the combustion chamber. In these
engines, when the piston is near the TDC, about 90% of the working fluid mass is inside the
piston bowl and in the region above this cavity. The pressure of this mass portion is
representative of the cylinder averaged pressure. The remaining mass occupies the crevice
regions between piston and head as well between piston and sleeve, and its pressure may
exhibit oscillations of up 10 bar amplitude, which are due to the turbulent in-cylinder flow
and to acoustic phenomena resulting from combustion. Thus, the transducer must be
located at a point from which the pressure of the mass above the piston bowl can be
accessed. Finally, it is important to point out that when choosing the transducer mounting
point the fuel jet impingement on the transducer diaphragm should also be avoided.
2.3 Validation of the transducer mounting location
In order to illustrate the procedure for validating the transducer mounting location we will
consider as example the case of a fast, direct injection diesel engine with three valves per
cylinder, in which indicating measurements were conducted using an uncooled minitransducer AVL GM 12 D, mounted above the piston bowl in the glow plug place, as shown
in Figure 3.
The method proposed by Randolph (1990) allows checking the occurrence of short-term
drift by comparing the cycle-to-cycle variability of the in-cylinder pressure readings at
specified instants along the working cycle. Certain variability is normal due to the random
nature of the combustion process, which causes one cycle to be slightly different from other
under unchanged operating conditions. This cycle-to-cycle variability leads to changes in
the thermal load acting on the transducer and, when short-term drift occurs, it leads also to
changes in transducer sensibility, thus enlarging the scattering of pressure readings.
28
In order to apply the Randolph test, consider two points along the cycle, designated as C1
and B2. The first point is the beginning of the exhaust process, thus characterizing a moment
when the transducer is under the influence of combustion thermal loads. The second point
is located in the compression stroke, thus soon after the gas exchange, during which the
transducer is cooled. So, if short-term drift was occurring, it should produce pressure
readings at C1 more scattered than those at B2 (Randolph, 1990; Roth el al., 2001). Figure 4
shows the pressure deviations from the sample mean-value, for 56 consecutive cycles.
According to the description given above, the short-term drift would cause the scattering of
points along the x-axis be greater than that along the y-axis. However, the points in this
figure are distributed equitably with respect to the axes, thus implying that short-term drift
is trivial in this case. The scattering of points with respect to the diagonal of the graph
allows judging about the repeatability of the experiment. The behaviour observed in Figure
4 attests for a good repeatability and is similar to that reported by Roth et al. (2001) for
transducers akin to the one used in this example.
Fig. 3. Pressure transducer mounting location.
0.50
0.25
0.00
-0.25
-0.50
-0.50
-0.25
0.00
0.25
0.50
Fig. 4. Deviation of pressure readings respective to sample mean-value. Point C1: 145 c.a.
degrees after compression TDC. Point B2: 80 c.a. degrees before compression TDC.
29
In cases of significant short-term drift, it is recommendable to mount the transducer via an

adapter, which avoids its direct contact with the cylinder gas, thus eliminating local heating
of the transducer components and, mainly, of its diaphragm. Another solution consists in
installing the transducer recessed by means of a measuring channel. However, such
mounting procedure may lead to inaccuracies caused by an oscillating flow in the channel.
The extent of such inaccuracies has been studied in detail by Hountalas & Anestis (1998).
2.4 Strategies for processing the piezoelectric transducer signal
The strong influence of the pressure derivative on the energy release rate was recognized by
several authors (Marzouk & Watson, 1976; Woschni, 1965; Krieger & Borman, 1966). Despite
this, indicating measurements usually aim obtaining the in-cylinder pressure, even when
combustion diagnosis is the major objective. As a rule, this task is carried out using a
piezoelectric transducer polarized by a charge amplifier (Brown, 1976; Marzouk & Watson,
1976; Benson & Pick, 1974; Lancaster et al., 1975; Lapuerta et al., 2000; Benson & Whitehouse,
1983). The preference for this practice can be attributed to the legacy of the mechanical
indicators, which were created especially for pressure measuring, as well as to the
importance of knowing the in-cylinder pressure for calculating the indicated work or for
characterizing the thermodynamic state of the working fluid. Thus, obtaining pressure
derivative data for heat release analysis was eventually relegated to the numerical
manipulation of available pressure data. As it will be discussed below, this procedure
amplifies the inaccuracies in the pressure data, originating pressure derivative curves with
oscillations that are eventually transferred to the results of energy release rate calculation.
The numerical instabilities inherent to the differentiation operation motivated the
development of experimental arrangements in which this operation was carried out by
means of electronic circuits. A reference to such an approach can be found in the work of
Marzouk & Watson (1976) who processed the analog signal from a charge amplifier by an
electronic differentiator. While avoiding the numerical differentiation of pressure data, these
authors did not obtain satisfactory results for the pressure variation rate, underestimating its
peak value during the premixed combustion. This was due to the deficient frequency
response of the adopted differentiation circuit, which severely damaged the performance
processing of those data that make up the high frequencies region of the spectral
distribution (see Section 4).
The drawbacks mentioned above can be overcome by employing an alternative technique of
polarization of the piezoelectric sensor the direct conversion of the current signal
supplied by the transducer into an analog voltage level proportional to the time rate of
change of the in-cylinder pressure. The circuitry presented in Section 2.4.2 is based on this
technique.
2.4.1 Signal processing with a charge amplifier
Figure 5 shows a simplified schematic of the experimental setup commonly used to obtain
in-cylinder pressure data. A shielded cable with high insulation resistance conducts the
charges polarized by the transducer to the inlet of the charge amplifier. This device is based
on an integrator circuit, which provides an output voltage proportional to the time
integration of the electric current applied at its input during a time interval t, taken from
30
the instant at which it was started (or reset) until the desired instant of measurement. Thus,
the pressure variation during the interval t is given by
P Pref vchG Ga
(2)
where P and Pref are values of the pressure acting on the transducer diaphragm at the end
and at the beginning of the time interval t, respectively; Ga is the gain of the charge
amplifier and vch is the charge amplifier output voltage.
Since the charge produced by the pressure transducer is pretty small, reaching just tens of
picocoulombs per bar, the association between a piezoelectric transducer and a charge
amplifier results extremely sensitive to non-idealities of electronic circuits, mainly to current
leakages occurring through the insulation resistance of the measurement system or even
within the integrator circuit. Such current leakages cause a slow and steady decline in the
level of the output voltage of the charge amplifier, thus yielding pressure values lesser than
the actual ones. In order to control this inaccuracy it is necessary to have high input
impedance in the charge amplifier, while taking especial care for maintaining the electrical
contacts clean and for operating the data conditioning system in a low-humidity
environment.
Fig. 5. Transducer signal conditioning through a charge amplifier.

Long-term drift also promotes an incessant and slow displacement of the data baseline as
the integrator circuit processes spurious current induced by changes in the transducer
temperature, which in its turn result from changes in engine operation conditions.
Therefore, even when properly using a good-quality measuring system, the combined action
of circuitry non-idealities and long-term drift leads to instability of the pressure baseline,
which is a drawback inherent to the use of a charge amplifier and is reported in the
literature as pressure baseline floating (Marzouk & Watson, 1976; Benson & Pick, 1974;
Lancaster et al., 1975; Benson & Whitehouse, 1983). This baseline floating can produce
pressure deviations of tens of bars during a long-duration measurement, making it
mandatory to reset periodically the amplifier in order to avoid its saturation.
The deviations mentioned above as well the fact that pressure data provided by a charge
amplifier are related to an initial measurement time make it necessary to correct the baseline
at each cycle, as well as the pressure readings. This task can be accomplished by
31
determining the pressure baseline that leads to a null value of the apparent heat release rate
over the first 80 c.a. degrees after the intake valve closing. This criterion for correcting the
pressure readings is based on the recommendations given by Lapuerta et al. (2000).
2.4.2 Signal processing with a current-to-voltage converter
The most evident experimental setup for converting the current polarized by a piezoelectric
transducer consists in conducting it to the ground through an electrical resistance, thus
originating a voltage drop equal to the product of the value of this resistance by the current.
Such procedure would result, however, in the formation of an RC circuit between the
transducer inherent capacitance and the measurement resistance, which, in turn, would
cause attenuation in the voltage drop and a phase delay with respect to the polarized
current. In order to avoid this problem Bueno et al. (2009, 2010, 2011) proposed to use a
circuit that converts current into voltage (Franco, 2001), consisting of an operational
amplifier and a negative feedback resistor, as shown in Figure 6. This circuit operates stably
in a range of gains typical for engine indicating measurements, and may undergo minor
changes related to balancing techniques, depending of which operational amplifier was
chosen for its construction. The virtually zero value of the input impedance of the currentto-voltage converter, given by the ratio of RA and the open-loop gain of the operational
amplifier A1, makes it immune to inaccuracies arising from the inherent capacitance of the
piezoelectric transducer. Besides that, in order to isolate the converter with respect to the
impedance of the instrument to which it is connected, a voltage follower amplifier was used.
Thus, the pressure rate of change dP/dt is given by:
Fig. 6. Transducer signal conditioning through a current-to-voltage converter.
vcv
dP
dt Gs RA
(3)
where vcv is the output voltage of the current-to-voltage converter and RA is the gain
adjusting resistance of the current-to-voltage converter. Taking the crankshaft angle speed
as a constant, the following expression for the pressure derivative with respect to crank
angle can be obtained:
vcv
dP
d 6 Gs RA RPM
in this equation RPM represents the engine operating speed, in revolutions per minute.
(4)
32
The in-cylinder pressure can be determined by numerically integrating the data obtained for
its derivative. A fourth order Runge-Kutta algorithm is recommended for this task
(Teukolsky, 1996) and the pressure baseline can be determined by applying the same
correction technique recommended in Section 2.4.1.
The use of a current-to-voltage converter eliminates the need of special care for insulation
resistance and leakage currents, as in this case the charges polarized by the transducer
unrestrictedly flow to the ground. In addition, the long-term drift makes the pressure
derivative baseline to deviate from its original position returning, however, after the engine
reaches a steady state. Therefore, this technique also eliminates the need for resetting the
transducer polarization circuit due to the large displacements of the data baseline. Further
details involving the conditioning of the pressure derivative signal using a current-tovoltage converter can be found in references Bueno et al. (2009) and Bueno et al. (2011).
3. Crank angle measurements and TDC position identification

In modern engine indicating systems the instantaneous position of the crankshaft is
determined with the aid of an optical angular encoder, whose operating principle is based
on photoelectric scanning of a sequence of thin opaque lines. These lines are etched on a
transparent disk that rotates together with the crankshaft and are arranged in the radial
direction at equal angular intervals, forming the so-named incremental track (see Figure 7).
Fig. 7. The optical angle encoder.

During operation of the optical angular encoder a light beam emitted by a LED falls
perpendicular to the disk plane on the incremental track. As the disk is rotating, this light
beam will be reflected if meets one of the etched lines, but it will pass across the disk while
falling on the gap between two lines. Then, a sequence of light pulses synchronized with the
crankshaft angular position will pass across the disk and will be conducted through an
optical fiber to a light-pulse converter, where photovoltaic cells transform this light signal
into an electrical one. Therefore, the encoder outputs a pulse string in response to the
amount of rotational displacement of the crankshaft. A separate counter counts the number
of output pulses to determine the crankshaft angle position.
33
In engine indicating measurements the angular position of the crankshaft must be

determined in relation to a reference, which is generally also used as reference to identify
the position of the TDC. Such reference is generated with the aid of a second incremental
track that has a single mark, which is used to reset the counting of the pulses.
Special attention should be given to the correct identification of the TDC position, as small
errors in this measurement lead to significant errors in the evaluation of indicated work as
well as combustion heat release rate (Krieger & Borman, 1966; Pischinger & Glaser, 1985;
Lapuerta et al., 2000). In order to achieve adequate precision in determining the TDC
position it is recommended to perform a dynamic measurement, running the engine
motored and unfired or, alternatively, preventing combustion only in the cylinder where the
measurement is going on, while the other cylinders function fired to keep the engine
running.
When performing dynamic measurement, the inaccuracies that would be generated by the
bearing clearances are eliminated. Such measurement is usually carried out with a
capacitive proximity sensor, or in the absence of such a sensor, the TDC position can be
inferred from the motored-engine pressure data.
The capacitive proximity sensor uses two conductive objects separated by a dielectric
material. A voltage difference applied to the conductive objects generates an imbalance of
electrical charges between them, originating an electric field in the dielectric material. When
this voltage is alternated the electrical charges move continuously, going from one of the
conductive objects to the other and generating an alternating electric current, which is the
output signal of the sensor. The amount of current flow is determined by the capacitance,
and the capacitance depends on the proximity of the conductive objects. Closer objects cause
greater current than more distant ones.
In the capacitive sensors used to determine the TDC position, one of the conductive objects
is the sensor probe itself, while the piston plays the role of the second conductive object
(Figure 8). The sensor is mounted in the head in such a way that when the engine is running
the piston will move closer or away from the sensor, but without actually touching it. Thus,
the sensor will produce a signal with amplitude, which is inversely proportional to the
distance between the TDC sensor tip and the piston top. The exact TDC position will
correspond to the maximum amplitude of the TDC sensor signal, which can be determined
with great accuracy because of the high degree of symmetry of the signal.
When using the motored-engine pressure data for identifying the TDC position, the main
arising difficulty is that the peak pressure precedes the actual TDC position, which
corresponds to the minimum volume. This occurs due to heat transfer and mass losses, and
the angle interval between these events is named loss angle (Figure 9). Several enough
accurate methods have been proposed for determining the loss angle (Pinchon, 1984; Stas,
1996), and usually manufacturers of indicating equipment include in their manual
recommendations to estimate loss angle values, which depend on the kind of the engine
(spark ignition or diesel) and compression ratio.
The advantage of the direct measurement of TDC position, compared with its determination
from the motored-engine pressure curve, is that there is no need for a correction involving
the loss angle, which always increases the uncertainty.
34
Fig. 8. The capacitive TDC sensor.
Fig. 9. Definition of loss angle.
4. Uncertainty sources and data treatment

Cylinder pressure data measured by a transducer can be understood as being the sum of
two components: (i) a smooth component, corresponding to the instantaneous average
pressure through the entire cylinder volume; and (ii) a spurious component, originated by
both the turbulent flow of gas inside the cylinder and the acoustic pulsations associated to
combustion. According to the hypothesis of the single-zone combustion model (Krieger &
Borman, 1966), the first component must be used to characterize the thermodynamic state of
the gases in the cylinder, constituting the information of interest for heat release analysis.
The second component is small when compared to the first one and its extent is influenced
by the transducer mounting location in the combustion chamber. In addition to these two
components of the measured pressure, the experimental data also include noise generated
by the measurement system during the conditioning of the transducer output signal. Thus,
for accurate heat release analysis it is necessary to isolate the volume-averaged pressure
component from both the flow and the combustion driven spurious component as well as
from the measurement noise.
35
Typically, the assessment of the smooth volume-averaged pressure component has been
accomplished by means of numerical treatment of experimental data, which is usually
carried out after data averaging over several consecutive cycles. A diversity of methods can
be used for this numerical treatment, including FFT filters, windowing filters, smoothing
splines and Wiebe function regression (Zhong et al., 2004; Payri et al., 2011; Ding et al.,
2011). However, the choice of a proper data treatment methodology requires, at least, a basic
knowledge about the behaviour of each one of the above mentioned components of the
experimental data. In order to assess this behaviour Bueno et al. (2009) proposed to estimate
the smooth component corresponding to the average pressure through the entire cylinder
volume and, then, to employ it to analyze the spurious component of the experimental data.
Sections 4.1 through 4.3 show how this task is accomplished.
4.1 The volume-averaged smooth data component
120
Cylinder Pressure [bar]
110
Cylinder Pressure Derivative [bar/deg]
Smooth component curves (referred here as reference curves) for both in-cylinder pressure
and in-cylinder pressure derivative data can be estimated by means of a computer routine
based on the single-zone combustion model. In this routine, the combustion heat release rate
is modelled through Wiebe functions (Wiebe, 1970) whose adjusting constants are obtained
by optimizing curve-fitting to heat release rate data calculated from measured values.
Figure 10 shows typical reference curves and experimental data obtained with each one of
the signal conditioning procedures described in Sections 2.4.1 and 2.4.2. The curves shown
in Figure 10 are related to a direct injection diesel engine and they show that the premixed
burning causes a steep slope in the pressure curve region subsequent to ignition (circled
area in Figure 10a), whilst in the pressure derivative curve it is reflected as a prominent peak
(circled area in Figure 10b). By comparing these two representations, it becomes evident that
the premixed combustion is more visible in the pressure derivative curve, a fact that led to
the choice of the pressure derivative as the basic representation for the analysis of the data
given by the indicator system.
Experimental Data: Charge Amplifier

Reference Data
100
90
80
70
60
50
40
30
20
10
0
-140 -120 -100 -80 -60 -40 -20
20 40
60
80 100 120
5
4
Experimental Data: Current-to-Voltage Converter

Reference Data
3
2
1
0
-1
-2
-3
-4
-140 -120 -100 -80 -60 -40 -20
20
40
Crank Angle [deg]
Crank Angle [deg]
(a)
(b)
60
80 100 120
Fig. 10. Typical reference curves for pressure and pressure derivative data (D.I. diesel engine
running at 2600 rpm and 80% of full load).
36
Frequency domain representations such as Lomb periodograms (Lomb, 1976) constitute a

suitable tool for the analysis of experimental data. In these periodograms the spectral power
is normalized with respect to the variance of the analyzed data, and its value represents the
degree of participation of the signal associated to a given frequency in the data composition.
In Figure 11, spectral distributions of two pressure derivative curves obtained through
numerical simulation are compared. The first curve, named Reference Data, was calculated
using two Wiebe functions, corresponding to premixed and diffusive stages of combustion,
respectively. The second curve, named Diffusive Combustion Data, was obtained suppressing
the premixed combustion and only one Wiebe function representing diffusive combustion
was used. As can be seen in Figure 11, these curves exhibit similar spectral power
distribution in the low-frequency region (up to 1000 Hz); however, in the region between
1000 and 5000 Hz the suppression of premixed combustion caused considerable attenuation
of the pressure derivative data. Thus, it may be concluded that the contribution of the
premixed combustion is located in this frequency range (from 1000 to 5000 Hz) and that the
utilization of low-pass filters or numerical treatment to smooth the transducer signal at
frequencies below 5000 Hz may cause distortion or even the elimination of the influence of
premixed combustion on experimental data. It may also be observed that the effects of both
the compression process and the diffusive combustion play a major role in the spectral
distribution of the data of interest, occupying the frequency range with highest spectral
power values (below 1000 Hz).
Normalized Spectral Power
100
Difusive Combustion Data

Reference Data
10
0.1
0.01
1E-3
1E-4
0
500
1000
1500
2000
2500
3000
3500
4000
4500
5000
5500
6000
Frequency [Hz]
Fig. 11. Spectral distribution for simulated in-cylinder pressure derivative data (D.I. diesel
engine running at 2600 rpm and 80% of full load).
4.2 Measurement system noise and the flow driven spurious component
Uncertainties associated to the data acquisition system are dominated by the truncation
error of the analog-to-digital converter, acting as a source of white noise in the acquired
experimental signal. Therefore, the measurement uncertainty due to this noise (MU) can be
estimated in terms of the analog-to-digital converter accuracy, which corresponds to the
least-significant bit (LSB) of the used measurement system.
When a transducer is polarized by a charge amplifier, the influence of this uncertainty on the
pressure derivative curve is indirect, because the measured parameter in this case is the incylinder pressure. Nevertheless, it is possible to estimate the measurement uncertainty
37
imposed by the quantization noise to the pressure derivative (MUdP) using finite differences
approximation of fourth order. Doing so and taking into account that data reported in Figures
10 and 11 refer to measurements made with a system whose LSB value is 0.131 bar, results
LSB 2 8LSB 2 8LSB 2 LSB 2
0.25 bar/ deg
MU dP
12 12 12 12
(5)
On the other hand, when the direct polarization of the transducer is used, the pressure
derivative is accessed directly and the measurement uncertainty imposed by the
quantization noise to the pressure derivative is given by the corresponding LSB value. In the
case of the data reported in Figures 10 and 11 this value is 54.694 bar/s, then
0,3
Charge Amplifier
Current-to-Voltage Converter
0,2
0,1
0,0
-0,1
-0,2
-0,3
-100
(6)
Charge Amplifier
Pressure Derivative Deviation [bar/deg]
MU dP 54.694 bar/s 3.5 10 3 bar /deg
7
6
5
4
3
2
1
0
-90
-80
-70
-60
Crank Angle [deg]
(a)
-50
-40
2000
4000
6000
8000
10000
12000
14000
Frequency [Hz]
(b)
Fig. 12. Deviations due to the measurement system noise and to the in-cylinder flow (D.I.
diesel engine running at 2600 rpm and 80% of full load).
It must be noticed that the utilization of a current-to-voltage converter allowed reducing the
measurement uncertainty imposed by the quantization noise to only 1.4% of the value
obtained using a charge amplifier. As it will be discussed later, this fact has important
consequences for obtaining smooth heat release rate diagrams. Additionally, when signal
conditioning is carried out using a charge amplifier, the refinement of the encoder resolution
increases the measurement uncertainty, whilst for the case of the current-to-voltage
converter the measurement uncertainty does not depend on the encoder resolution.
Figure 12a shows experimental data deviation from the reference pressure derivative curve
during the compression process, when combustion does not occur and, therefore, the
spurious components of the experimental data can be attributed only to the in-cylinder flow
and to the data acquisition system. It may be observed in this figure that the utilization of a
charge amplifier imposes oscillations, which are within the limits of the quantization noise
(0.25 bar/deg); whilst for the case of the current-to-voltage converter these oscillations
38
exceed the corresponding limits (3.510-3 bar/deg). Therefore, it is possible to conclude

that when a charge amplifier is used the effects of the noise generated by the data
acquisition system predominate over the spurious component generated by in-cylinder
flow, whilst in the case of the current-to-voltage converter this spurious component is the
major responsible for the oscillations observed along the compression process.
The spectral distributions corresponding to the deviation curves shown in Figure 12a can be
found in Figure 12b. These distributions characterize both the in-cylinder flow as well as the
quantization error, as sources of white noise (random noise), having a considerable part of
its spectral composition in the same frequency range occupied by the data of interest. Such
behaviour makes it difficult to use low-pass filters or numerical smoothing without loss of
useful signal, demanding especial care during the choice of the smoothing parameters,
which vary with engine running condition (Zhong et al., 2004). Due to the high quantization
noise in the data obtained with the charge amplifier, the remaining analysis of the
experimental data composition was based on the current-to-voltage converter approach.
4.3 The combustion-driven spurious component
The ignition in diesel engines exhibits an eminently random character, being strongly
influenced by the interaction of chemical and hydrodynamic phenomena occurring along
the ignition delay (Maunoury et al., 2002). Because of this, a diesel engine operating in
steady state presents cycle-to-cycle variations in the amount of fuel available for premixed
combustion as well as in the locations where ignition occurs by first. In addition, the natural
vibration modes of the gas contained in the cylinder are excited by the high burning rates
that occur in the beginning of the combustion (Schmillen & Schneider, 1985), generating
pressure oscillations of large amplitude that are rapidly dampened due to the high
area/volume ratio of the cylinder cavity. Due to cycle-to-cycle combustion variability,
different modes are excited in each cycle, causing the oscillations to vary in phase,
characteristic frequencies and amplitude from one cycle to other.
Figure 13a shows the typical behavior of experimental pressure derivative data, where the
cycle-to-cycle variability associated to the premixed combustion and to the following acoustic
oscillations contrasts with the repeatability inherent to both the compression process and the
diffusive combustion. In the periodograms presented in Figure 13b, the combustion-generated
variability appears in the frequencies above 1000 Hz. This result is in accordance with the
observations of Strahle (1977) and Strahle et al. (1977a, 1977b), who attributed the pressure
data spectrum above 1000 Hz to the random phenomena associated to combustion.
In Figure 14a the deviation of the experimental data respective to the reference pressure
derivative curve is shown along the combustion process for two cycles. As can be noticed,
after 25 c.a. degrees the amplitude of the oscillations is dampened to a constant level slightly
higher than that observed during compression, indicating that combustion amplifies the influence of the turbulence on experimental data. The spectral analysis of this deviation is
shown in Figure 14b, where a periodic signal whose frequencies should correspond to the
cylinder cavity characteristic frequencies can be seen (9.0 kHz for cycle 2; and 2.0, 6.0 and
11.0 kHz for cycle 13). Figure 14 confirms that the oscillations vary in phase, in characteristic
frequencies and in amplitude from one cycle to other. This behavior may be used to remove
the combustion-driven oscillations by data overlapping.
39
4
3
100
Cycle 02
Cycle 13
Cycle 40
Pressure Derivative [bar/deg]
2
1
0
-1
-2
Cycle 02
Cycle 13
Cycle 40
10
0.1
0.01
1E-3
-3
-4
-120 -100
1E-4
-80
-60
-40
-20
20
40
60
80
100
500 1000 1500 2000 2500 3000 3500 4000 4500 5000 5500 6000
Frequency [Hz]
Crank Angle [deg]
(a)
(b)
Fig. 13. Pressure derivative experimental data and its spectral composition (D.I. diesel
engine running at 2600 rpm and 80% of full load).
15
14
Cycle 02
Cycle 13
13
12
11
10
9
8
7
6
5
4
3
2
1
0
0
2000
4000
6000
8000
10000
12000
14000
Frequency [Hz]
(a)
(b)
Fig. 14. Combustion-driven spurious component of the experimental pressure derivative

data (D.I. diesel engine running at 2600 rpm and 80% of full load).
4.4 Cycle-averaging as a data smoothing method
Cycle-averaging over a significant number of successive cycles is an effective method of data

treatment, easing by overlapping the effects of cycle-to-cycle variations, combustion driven
oscillations and measurement noise. This can be seen in Figure 15, which shows in its upper
part the spectral composition of the spurious components of experimental data, whilst in its
lower part the spectral composition of single-cycle pressure derivative data is compared to
that of mean data (averaged over a sequence of 56 cycles). Data corresponding to cycles with
the lowest noise level (13th cycle) and with the highest noise level (2nd cycle) are shown in
the left and in the right parts of this figure, respectively.
40
14.00
Combustion Noise
ADC+Flow Noise
8.00
6.00
4.00
2.00
0.25
Cycle 13
Mean Data
0.20
Combustion Noise
ADC + Flow Noise
12.00
10.00
14.00
12.00
10.00
0.15
0.10
8.00
6.00
4.00
2.00
0.25
0.20
Cycle 02
Mean Data
0.15
0.10
0.05
0.05
0.00
0.00
2000
4000
6000
8000
10000
12000
2000
14000
4000
6000
8000
10000
12000
14000
Frequency [Hz]
Frequency [Hz]
(a)
(b)
Fig. 15. Noise in individual cycles and its reduction by cycle averaging (D.I. diesel engine
running at 2600 rpm and 80% of full load).
Figures 16 and 17 show data of cylinder pressure and its derivative for the same engine
operating conditions, where 95% confidence limits are displayed. Data of Figure 16 were
obtained using a charge amplifier while for data of Figure 17 a current-to-voltage converter
was used. In both cases data were averaged over a sequence of 56 cycles. Analyzing these
figures it can be concluded that both transducer signal-conditioning procedures resulted in
similar values for the cycle averaged cylinder pressure and its confidence limits, so that they
are equivalent when the cylinder pressure is the parameter of interest. However, the studied
procedures gave different results when evaluating pressure derivative, which is the data of
greater interest for heat release analysis. The pressure derivative confidence interval
obtained with the current-to-voltage converter was smaller than that obtained using the
charge amplifier and subsequent numerical derivation, the former reaching only about 1/50
of the latter during most of the compression process as well as during late combustion.
120
110
100
90
80
70
60
50
40
30
20
10
0
-140 -120 -100 -80 -60 -40 -20
20
40
Crank Angle [deg]
(a)
60
80 100 120
7
6
5
4
3
2
1
0
-1
-2
-3
-4
-5
-6
-140 -120 -100 -80 -60 -40 -20
20
40
60
80 100 120
Crank Angle [deg]
(b)
Fig. 16. Indicating data obtained using a charge amplifier (D.I. diesel engine running at 2900
rpm and full load).
41
120
110
100
90
80
70
60
50
40
30
20
10
0
-140 -120 -100 -80 -60 -40 -20
20
40
60
7
6
5
4
3
2
1
0
-1
-2
-3
-4
-5
-6
-140 -120 -100 -80 -60 -40 -20
80 100 120
20
40
Crank Angle [deg]
Crank Angle [deg]
(a)
(b)
60
80 100 120
Fig. 17. Indicating data obtained using a current-to-voltage converter (D.I. diesel engine
running at 2900 rpm and full load).
1.1
0.040
Charge Amplifier
0.035
1.0
0.9
Mass-Burned Fraction [-]
Normalized Rate of Heat Release [1/deg]
Considering the extension of the confidence intervals of pressure derivative data, it is

expected that heat release results calculated from the charge amplifier data exhibit a more
irregular and oscillatory pattern than those calculated from the current-to-voltage converter
data. It is worth mentioning that large oscillations in the calculated heat release rate curve
can result in negative values during the compression as well as during late combustion,
which makes no physical sense. These oscillations make it difficult to determine the
pressure data baseline, which in turn increases the uncertainty in the assessment of the
burned fraction of fuel mass. Data shown in Figure 18a confirm that heat release curves
0.030
0.025
0.020
0.015
0.010
0.005
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0.000
0.0
-20 -10
10 20 30 40 50 60 70 80 90 100 110 120
Crank Angle [deg]
(a)
-20 -10
10 20 30 40 50 60 70 80 90 100 110 120
Crank Angle [deg]
(b)
Fig. 18. Normalized heat release rate and fuel mass burned fraction (D.I. diesel engine
running at 2900 rpm and full load).
42
calculated from charge amplifier data exhibit a higher noise level than those calculated from
current-to-voltage converter data, while Figure 18b gives evidence of the differences in the
calculated burned fraction of fuel mass, which are caused by this higher noise level.
5. References
Alwood, H. I. S.; Harrow, G. A. & Rose, L. J. (1970) A Multichannel Electronic Gating and
Counting System for the Study of Cyclic Dispersion, Knock and Weak Mixture
Combustion in Spark Ignition Engines. SAE Paper 700063.
Alyea, J. W. (1969) The Development and Evaluation of an Electronic Indicated Horsepower
Meter. SAE Paper 690181.
Amann, A. C. (1985). Classical Combustion Diagnostics for Engine Research. SAE Paper
850395.
Benson, R. S. & Pick, R. (1974) Recent Advances in Internal Combustion Engine
Instrumentation with Particular Reference to High-Speed Data Acquisition and
Automated Test Bed. SAE Paper 740695.
Benson, R. S. & Whitehouse, N. D. (1983) Internal Combustion Engines, 1st Ed. Pergamon
Press.
Brown, W. L. (1967) Methods for Evaluating Requirements and Errors in Cylinder Pressure
Measurement. SAE Paper 670008.
Brown, W. L. (1973) The Caterpillar imep Meter and Engine Friction. SAE Paper 730150.
Bueno, A. V. ; Velsquez, J. A. & Milanez, L. F. (2009) A New Engine Indicating
Measurement Procedure for Combustion Heat Release Analysis. Applied Thermal
Engineering, Vol. 29, pp. 16571675.
Bueno, A. V. ; Velsquez, J. A. & Milanez, L. F. (2010) Heat Release and Engine Performance
Effects of Soybean Oil Ethyl Ester Blending into Diesel Fuel. Energy, Vol. 36, pp.
3907-3916.
Bueno, A. V., Velsquez, J. A. & Milanez, L. F. (2011) Notes on A Methodology for
Combustion Detection in Diesel Engines Through In-cylinder Pressure Derivative
Signal. Mechanical Systems and Signal Processing, Vol. 25, pp. 3209-3210.
Ding, Y. ; Stapersmal, D. ; Knoll, H. & Grimmelius, H. T. (2011) A New Method to Smooth
the In-cylinder Pressure Signal for Combustion Analysis in Diesel Engines.
Proceedings of the Institution of Mechanical Engineers, Part A: Journal of Power and
Energy, Vol. 225, pp. 309-318.
Draper, S. C. & Li, T. Y. (1949) New High-Speed Indicator of Strain-Gauge Type. Journal of
Aerospacial Science, Vol. 16, pp. 593-610.
Ficher, R. V. & Macey, J. P. (1975) Digital Data Acquisition with Emphasis on Measuring
Pressure Synchronously with Crank Angle. SAE Paper 750028.
Franco, S. (2001) Design with Operational Amplifiers and Analog Integrated Circuits, 3rd Ed.
McGraw-Hill.
Hountalas, D. T. & Anestis, A. (1998) Effect of Pressure Transducer Position on Measured
Cylinder Pressure Diagram of High Speed Diesel Engines. Energy Conversion and
Management, Vol 39, pp. 589-607.
Kistler (1956). SLM Pressure Indicator, DEMA Bull. 14.
43
Krieger, R. B. & Borman, G. L. (1966) The Computation of Apparent Heat Release for
Internal Combustion Engines, ASME paper 66-WA/DGP-4.
Lancaster, D. R. ; Krieger, R. B. & Lienesch, J. H. (1975) Measurement and Analysis of Engine
Pressure Data. SAE Paper 750026.
Lapuerta, M. ; Armas, O. & Bermdez, V. (2000) Sensitivity of Diesel Engine
Thermodynamic Cycle Calculation to Measurement Errors and Estimated
Parameters. Applied Thermal Engineering, Vol. 20, pp. 843-861.
Lomb, N. R. (1976) Least-Squares Frequency Analysis of Unequally Spaced Data.
Astrophysics and Space Science, Vol. 39, pp. 447-462.
Marzouk, M. & Watson, N. (1976) Some Problems in Diesel Engine Research with Special
Reference to Computer Control and Data Acquisition. Proceedings of the Institution of
Mechanical Engineers, Vol. 190, pp. 137-151.
Maunoury, B. ; Duverger, T. ; Mokaddem, K. & Lacas, F. (2002) Phenomenological Analysis
of Injection, Auto-ignition and Combustion in a Small D.I. Diesel Engine. SAE Paper
2002-01-1161.
Payri, F. ; Olmeda, P. ; Guardiola, C. & Martn, J. (2011) Adaptive Determination of Cut-off
Frequencies for Filtering the In-cylinder Pressure in Diesel Engines Combustion
Analysis. Applied Thermal Engineering, Vol. 31, pp. 2869-2876.
Pinchon, P. (1984) Calage Thermodynamique du Point Mort Haut des Moteurs Piston.
Revue de l'institut du Ptrole, Vol 39(1).
Pischinger, R. & Glaser, J. (1985) Problems of Pressure Indication in Internal Combustion
Engines. Proceedings of COMODIA 1985, Tokyo, Japan.
Randolph, A. L. (1990) Methods of Processing Cylinder-Pressure Transducer Signals to
Maximize Data Accuracy. SAE Paper 900170.
Roth, K. J. ; Sobiesiak, A. ; Robertson, L. & Yates, S. (2001) In-Cylinder Pressure
Measurements with Optical Fiber and Piezoelectric Transducers. SAE Paper 200201-0745.
Schmillen, K. & Schneider, M. (1985) Combustion Chamber Pressure Oscillations as a Source
of Diesel Engine Noise. Proceedings of COMODIA 1985, Tokyo, Japan.
Stas, M. J. (1996) Thermodynamic Determination of T.D.C. in Piston Combustion Engines.
SAE Paper 960610.
Strahle, W. C. (1977) Combustion Randomness and Diesel Engine Noise: Theory and Initial
Experiments. Combustion and Flame, Vol. 28, pp. 279-290.
Strahle, W. C. ; Handley, J. C. & Varma, M. S. (1977a) Cetane Rating and Load Effects on
Combustion Noise in Diesel Engines. Combustion Science and Technology, Vol. 17, pp.
51-61, 1977.
Strahle, W. C. ; Muthukrishnan, M. & Handley, J. C. (1977b) Turbulent Combustion and
Diesel Engine Noise. Proceedings of the Combustion Institute, Vol. 16, pp. 337-346.
Teukolsky, S. A. ; Vetterling, W. T. & Flannery, B. P. (1966) Numerical Recipes in C: The Art of
Scientific Computing. William H Press.
Wiebe, J. J. (1970) Brennverlauf und Kreisprozes von Verbrennungsmotoren, VEB-Verlag Technik
Berlin, Germany.
Woschni, G. (1965) Computer Programs to Determine the Relationship between Pressure,
Flow, Heat Release and Thermal Load in Diesel Engines. SAE Paper 650450.
44
Zhong, L. ; Henein, N. A. & Bryzik, W. (2004) Effect of Smoothing the Pressure Trace on the
Interpretation of Experimental data for Combustion in Diesel Engines. SAE Paper
2004-01-0931.
3
Measurement Systems for
Electrical Machine Monitoring
Mario Vrazic, Ivan Gasparac and Marinko Kovacic
University of Zagreb, Faculty of Electrical Engineering and Computing,
Croatia
1. Introduction
Monitoring systems on electrical machines, nowadays largely present on new electrical
rotating (synchronous generators) and non rotating machines (transformers) have to
endure harsh conditions such as vibrations and electromagnetic disturbances. Also,
especially for rotating machines, rotor physical quantities (temperature, strain, current),
have to be transferred to the standstill part of the measurement system without galvanic
contact.
Therefore, monitoring systems should be, and usually are, designed for a specific purpose
and for a specific machine. Some essential parts are the same but most of the monitoring
system should be custom made especially if one takes into consideration that there is a small
serial production of large electrical machines in the world and that most of the large units
are unique.
Modern monitoring of big transformers covers permanent on-line monitoring of electrical,
magnetic and mechanic quantities. It is specific when compared to rotational electrical
machines because it also usually includes monitoring of dissolved gasses in transformer oil.
This information, with proper interpretation, represents an important element for
transformer state determination. However, transformer monitoring systems are not the
focus of this chapter.
This chapter deals with all problems, obstacles and solutions applied to two measurement
systems installed on a 35 MVA hydro-generator and a 250 MVA turbo-generator.
Large electric machines justify investment in measurement systems for several reasons.
Usually, measurement systems for bearings are installed. In such systems vibrations and oil
parameters (pressure and temperature) are monitored to help maintenance. Measurements
of stator and less often rotor currents are also present as parts of different systems. For
example, a differential stator current measurement is present for protection purposes and
enables detection of current leakage (short circuit between phase windings). Monitoring
systems that incorporate all necessary measurements and, of course, interpretations of those
measurements are almost non existing and very rare. This chapter will concentrate on
synchronous generators.
46
2. The monitoring system

The monitoring system is basically a data acquisition system with on line data processing
and possible data interpretation. A standard data acquisition system consists of a sensor,
signal transfer, A/D conversion, signal conditioning and data storage.
The sensor is usually analogue, like a Pt1000 probe for temperature measurement. The
temperature is measured as a resistance over two, three or four wires. If resistance is
measured over four wires, the current flows through two wires and the other two wires are
used for voltage measurement. This system configuration is very sensitive to
electromagnetic disturbances thus making measurement inherently inaccurate. In order to
avoid that and to make measurement more accurate signal amplifiers and converters are
designed. They enable steady current regardless temperature dependence of connecting
wires. But still, voltage measurement is subjected to electromagnetic disturbances since the
measured voltage has an mV range. Of course, all of these disturbances and most of their
influence can be avoided by proper wiring. For example, bifilar wire configuration and
shielded cables significantly reduce electromagnetic disturbances if the shield is properly
grounded.
So, such configuration, for example the one installed in Plomin thermo power plant, has a
sensor (a Pt1000 probe), a rather large amount of wire (15 m), a signal converter and
amplifier (placed in the measurement cabinet), a 10 VDC signal path (a 2 m cable), an
acquisition system, communication (90 m Ethernet FTP cable) and finally a measurement
computer.
The reasons for such configuration will be presented later in this chapter. Of course, if one
has the opportunity and technical possibility, one should install a more advanced
temperature measurement. For example, there are digital temperature probes on the market
that are basically small chips with temperature probes and all the necessary signal
conversion. All they need is power and communication. They obtain them both through a
three wire system that is completely independent of electromagnetic disturbances. Such a
system has a serial connection and can have up to 127 probes in one system.
This description shows one review of the acquisition system that, of course, can grow to an
enormous size. For example, in CERN in Switzerland, there is a measurement system with
over 175 000 channels.
The monitoring system incorporates data acquisition and enables instant data processing
and interpretation. For instance, before information enters the PC, the analogue signal goes
through the A/D conversion and scaling. With this, for example, a signal of 5.4 mV becomes
24.5 C. After that, the information about the exact temperature enters a PC application that
operates with it. The temperature, maybe, does not represent the physical quantity, so
vibrations will be used from now on.
Vibrations are measured with accelerometers. They measure vibration accelerations and
such information enters the PC. But for a reasonable vibration analysis different quantities
should be obtained from vibration acceleration such as vibration velocity and vibration
displacement. Furthermore, for all the three quantities (acceleration, velocity and
displacement) a base harmonic and some higher harmonics should be known. All of this can
be done within data acquisition unit but then this unit will be very expensive. So, the
Measurement Systems for Electrical Machine Monitoring
47
general idea is that such advance data processing is done within application in the
monitoring PC.
The monitoring system can now have access to all the necessary data and can conclude
upon the electrical machine condition. Of course, this conclusion can be based on simple or
complex algorithms. Simple algorithms will include allowed limits, alarm limits and
shutdown limits for some or most of the physical quantities. Complex algorithms include
monitoring of some physical quantity trend lines and from that info concludes on further
actions. Those conclusions can lead to a well-timed maintenance action that is more efficient
and cheaper than waiting for a fault to occur and lead to an unplanned stopping of electrical
machine and process for which this machine is necessary.
In the following text three measurement systems and one monitoring system will be
presented.
The general idea of building a measurement system, and further on, a monitoring system on
synchronous generators started with a desire to build a user P-Q diagram of a synchronous
aggregate. In order to adapt FEM and analytical models to a real generator a lot of
measurement of different physical quantities should be done. This demanded a
measurement system to be made. Such a measurement system should be able to measure
magnetic inductances B (T), temperatures T (C), acceleration of vibrations a (m/s2), stator
and rotor voltages U (V) and currents I (A).
So the whole idea was tested on synchronous hydro-generator and synchronous turbogenerator.
3. Measurements on the hydro-generator

When research for a user P-Q diagram started, the access was approved to Vinodol
hydropower plant where three generators (Fig. 1) are installed. Their rated data is: Sn=35
MVA, Un=10.5 kV, f=50 Hz, n=500 rpm, cos=0.9.
Fig. 1. Hydro-generators in hydro power plant Vinodol - 35MVA each

A larger measurement system was installed on the first generator. This enables a reduction
of measurement systems installed on the other two generators. This text will describe the
larger measurement system.
48
The research demanded a magnetic induction measurement, so 22 Hall probes were

installed in winding end part and in the air gap. Since calibrated Hall probes are very
expensive it was decided to buy only three calibrated Hall probes with their calibration
certificates. The rest of hall probes were cheap sensors (their cost was approx. $5 per probe).
Hall probes have 4 connections. Two of them need a steady current flow and on the other
two voltage can be measured. The voltage corresponds to the magnetic induction. Both
current (1 mA or 5 mA) and voltage (range of mV) have rather small values. To satisfy those
conditions current sources were designed. The current range can be selected with a switch
within the current source. The Hall probe voltage is amplified with a signal amplifier to a
10 VDC signal range and as such prepared for data acquisition. Of course, there are still
two problems: calibration and electromagnetic disturbances.
But before calibration, all Hall probes were placed on an electronic board (Fig. 2) that has
two possible connection places. If there is enough room for probe placement, then Fig. 2
option a can be used, but if there is not enough place (air gap, cooling channel) then a minor
board can be cut out from the bigger one (Fig. 2 option b) and wires can be connected to it.
(a)
(b)
Fig. 2. Hall probe board

The calibration problem was resolved when it was done in the laboratory. A three phase
choke was used (Fig. 3) for that purpose. The air gap on the choke can be changed from 0 to
32 mm.
The calibrated Hall probe (the expensive one) was used as a reference probe and all the
necessary calibrations (with AC and DC signals) for all the other Hall probes were made
with it: the calibration constant, the DC offset and the thermal dependence. Several Hall
probes were discarded because of their lack of linearity but most of the cheap hall probes
were successfully calibrated.
Still, there was a problem of electromagnetic disturbances since the voltage signal from Hall
probes is very small (range of mV). The distance from a probe to signal amplifiers is approx.
15 meters and that also contributes to the problem. This issue was minimized by using a 4
wire shielded cable. So there is one cable for each probe. Since those synchronous generators
are air cooled there was no problem with the cable entrance through generator housing.
Standard cable glands were used. All the cables were shielded on one side, near signal
amplifiers. Such configuration ensures electromagnetic disturbance discharge thus
minimizing their influence on the measured signal.

200 mm
air gap
0 40
49
Fig. 3. Choke for Hall probe calibration

Similar measurement was made for temperature. Pt1000 probes and 2 wire measurement
were used. Signal converters and amplifiers in one device were used to obtain a 10 VDC
signal for the data acquisition system. Two wire shielded cables were used for each signal
and were grounded as the ones from Hall probes.
But a different problem was identified here. The problem was heat transfer from the surface
to the probe. So before mounting Pt1000 probes onto the generator, the probes were
conditioned. They were aligned on a thin pertinax base to ease mounting on the generator
and as well as to assure the wanted relative position. Each probe was glued with super glue
to pertinax (Fig. 4). After that isolation foil was glued over the probe. This foil assures
electrical isolation but does not influence thermal transition. Cable wires were soldered on
the Pt probe and the cable was glued on pertinax. When mounting on the generator, thermal
paste was applied between the surface and the probe to ease thermal conductivity.
After that everything was suffused with thermal resistant glue.
In order to be able to obtain the exact wanted position of probes within the generator, Hall
(A1, B1-B4, C1-C4, D1-D4, E1-E4, F1-F3, G1-G2) and Pt1000 (GTS1-GTS14) probes were placed
on thin boards in the exact position (Fig. 5). The boards were then placed at the exact position
within the generator (Fig. 6, Fig. 7 and Fig. 8). All probe positions were carefully recorded and
later used when 2D and 3D FEM models were confirmed by measurements (Fig. 9).
One of the instruments developed during the measurement system design was an
instrument for measuring the load angle (). During the designing period several devices
were designed and commissioned. The method of load angle measurement is based on time
measurement between zero crossing of the generator voltage and the rotor position
represented by positive edge of proximity switch impulse. At first, this operation was
carried out with a laboratory device and then after the production of a more professional
version, the third, compact and universal model was produced.
50
Fig. 4. Mounting scheme of Pt1000 probes
(a)
Fig. 5. Prepared probes: a) Hall probes; b) Pt1000 probes
Fig. 6. Built-in Hall probes
Fig. 7. Placement of Hall probes in generator
(b)
51
Fig. 8. Placement of Hall and Pt1000 probes
Fig. 9. 2D and 3D FEM induction calculations for no load operation

Also, as previously mentioned, two vibration probes (accelerometers) were installed on the
stator package (Fig. 10).
Fig. 10. Vibration accelerometer installed on the stator package
52
After all of that, the measurement system needed to be designed and built. Since, at the time,
only 16 channel data acquisition systems were available to us, 4 different data acquisition
systems (Iotech) were used for complex measurements. All together there were a total of
51 signals (22 Hall probes, 14 Pt1000 probes, 3 armature currents, 3 armature voltages,
1 excitation current, 1 excitation voltage, 1 impulse per rotation, 1 load angle, 3 grid voltages,
2 vibration accelerations). All data acquisition systems had trigger possibility on
one channel so this was used for synchronous start of data acquisition. Since there were so
many signals there are lots of cables present. Most measurements were made with
a temporary measurement system (Fig. 11). In order to be sure that electromagnetic
Fig. 11. Temporary measurement system

0,0075
0,0000
-0,0075
0,003
0,000
-0,003
0,00020
0,00000
-0,00020
0,00100
0,00000
-0,00100
0,015
-0,020
0,020
0,000
-0,020
0,0125
0,0000
-0,0125
0,005
0,000
-0,005
25 50 75 100
B1_HS16 [T]
150
200
B2_HS15 [T]
B4_HS06 [T]
C1_HS07 [T]
C3_HS12 [T]
C4_HS13 [T]
250
300
350
B3_HS14 [T]
C2_HS11 [T]
Fig. 12. Hall probe measurements signals of different probes
400
450
ms
53
disturbances are not present, or at least are acceptable, each signal was checked several
times and compared to the theory. For example, each magnetic induction signal was
checked separately and together with the rest of them. If two signals are the same and their
probes are not in the position that can give the same signals (for example the same radial
position on the press plate) this is the indication that one signal is copied to the other
channel. Also, most of signals must have some angular difference and this is the other way
to check them (Fig. 12).
After several measurements with the temporary measurement system there was a 3 year
delay with this power plant due to circumstances beyond our influence. So, the
measurement and monitoring system development for this power plant was halted. But all
designs have been done and the equipment has been bought and installed in the electrical
cabinet (Fig. 13). This cabinet (measurement cabinet) will be installed in the power plant
during the first half of next year. Data acquisition will be realised with National instruments
Compact RIO product. The main interface will be touch panel. The plan is to place a PC near
this cabinet and connect the system to the Internet.
Fig. 13. Measurement cabinet for measurement systems installed on all three generators in
hydro power plant Vinodol
4. Measurements on the turbo-generator

A measurement system was also installed on one turbo-generator (Fig. 14). This generator is
installed in Plomin 2 thermo power plant. Its rated data are 250 MVA, 3000 rpm.
A similar set of probes was installed in this generator as it was in the generator in HPP
Vinodol. In total, 55 probes were installed in the generator: 30 Hall probes, 14 Pt1000 probes
and 11 accelerometers. All armature and excitation voltages and currents are measured
together with one grid voltage. All those currents and voltages are measured in such a way
that a malfunction in the measurement system does not influence systems in the power
plant. For instance, armature currents are measured with current transformers (5A/333mV)
54
Fig. 14. Turbo-generator 250 MVA in Plomin 2 thermo power plant

placed around the wires from the secondary windings of the major current measurement
transformers (2000A/5A) of the generator. Voltages are also measured on the secondary
windings of major voltage measurement transformers. This connection was made over
250mA fuses. Major voltage transformers have the ability to burn those fuses (in case of a
measurement system malfunction) without jeopardizing their primary function.
This generator is cooled with hydrogen. Since this is a highly explosive coolant all signals
were checked for energy level they import in the generator. The energy level must be lower
than the energy that is high enough for sparking and that may ignite hydrogen. This is the
first step to ensure that the measurement system can do no harm to the generator. The next
step was to shut down all the measurements and cut off the power in the measurement
cabinet in case of hydrogen leakage. So, if sensors detect hydrogen outside of generator, the
main contactor used for powering the measurement cabinet switches off. The third step was
installation of a hydrogen detection probe within the measurement cabinet since signal
cables can lead hydrogen from the generator to the measurement cabinet in case of cable
glands malfunction. The fourth step was to design special cable glands to ensure absence of
hydrogen leakage. Cable glands (Fig. 15) were designed, tested, installed and they have
been functioning flawlessly for two years.
Signals from the generator were conducted through a signal cable. This cable has 16 twisted
pair wires. Each twisted pair is shielded and the whole cable is shielded, too. This double
shielding ensures better protection from electromagnetic disturbances. The measurement
system is similar to the one in HPP Vinodol. The difference occurs in the data acquisition
system.
After conditioning all the signals go to one device (NIcDAQ9188) and not to four different
data acquisition systems. The conditions and customer demanded that the PC should be
positioned in the control room approximately 90m away from the measurement cabinet
(Fig. 16) so Ethernet cable FTP cat.6 was used for that connection.
55
Fig. 15. Signal cable glands
Fig. 16. Measurement cabinet with NIcDAQ9188 data acquisition system

Unlike the measurement system in HPP Vinodol, where measurement and monitoring
systems should be finished, everything is finished in TPP Plomin2. A lot of work was done
in order to produce the application in LabVIEW for measurement, signal processing and
data storage.
The front panel application was developed in LabVIEW and consists of several tabs. All of
the measurements are divided between multiple tabs. The first three tabs are used for
graphical view of the measured signals. It is possible to observe waveforms of all of the
acquired signals (magnetic field flux density, vibration accelerations and electrical signals
such as stator currents, voltages, field current, voltage etc.). The user is also able to read the
exact value of the waveform using a cursor tool and move over the waveform using zoom
in/out and drag tools (Fig. 17).
After successful acquisition of the all the signals the application computes all the necessary
quantities such as rms, peak-to-peak, base harmonics of the acceleration, velocity and
56
Fig. 17. Front panel of the application, showing real time waveforms of the 3-phase system
voltage
displacement for vibration signals; dc-offset, rms value, base harmonics for magnetic flux
density signals; mean value for temperature signals; full 3 phase power analysis for electrical
signals. The signals that can also be measured on the synchronous machine are field current,
voltage and load angle of the machine. The computed quantities are also divided in four
separate tabs (magnetic flux density, vibrations, temperature and electrical), (Fig. 18).
Fig. 18. Front panel of the application, showing real time electrical measurements and
capability diagram
After a successful calculation, all of the calculated quantities are logged into the historical
database (Fig. 19). A user can access the database from the application GUI, and observe
variations or trends of the selected quantities over a desired time period. Trend analysis is
especially valuable tool in vibration monitoring. Based on the trend of the calculated
vibrations user can predict a maintenance of the machine.
57
Fig. 19. Historical data view of computed variables (red-active power, green- reactive,
white- tangential end winding vibration, violet-axial end winding vibration)
The described method of monitoring incorporates measurement of the signal waveforms.
All of the data logged into the database are just physical quantities calculated from those
waveforms. There are cases when a user demands to log the waveform data for further,
more detailed, analysis. A continuous waveform logging requires a large storage capacity as
all of the waveforms are sampled at rather a high frequency (more than 10 kHz). To lower
the demands on storage capacity the application must be able to decide which of the
waveforms to store. The stationary waveforms are no of interest for most of the users so the
application must detect only a change in the signal and based on that change it has to trigger
storing of the waveform. When designing such an application one has to be careful that it
records a certain amount of signal before the actual trigger event. Furthermore, in this type
of measurement the acquisition system and the data processing have to be able to withstand
the continuous sampling of the signal to ensure that none of the samples are lost due to the
buffer overrun or the slow data processing. Fig. 20 shows an armature current transient
properly recorded with pre-buffer time set to 5 s and total transient time set to 20 s.
5. Measurements on the rotor

Modern synchronous generator monitoring and controlling systems require measurement of
various physical quantities. One of those is the temperature of the machine. It is relatively
easy to measure temperature of the stationary part of the machine, but the difficulties arise
with the measurement of the rotor temperature. Conventional temperature measurement
requires slip-rings for each sensor. The measurement system which uses digital wireless
data transfer and digital temperature sensors will be described in short.
The DS18B20 digital sensors were chosen for the temperature measurement. The accuracy of
the temperature sensors is 0.5 C. Sensors use a 1-wire protocol for communication with
the microcontroller. Bluetooth communication was selected because of the high operating
frequency 2.4 GHz, low power requirement and wide usage in various IT equipments. A
special embedded device was developed to convert data from temperature sensors. The
measurement system consists of several specific parts:
58
Bluetooth communication module

Microcontroller
Battery charge management
14-bit ADC
Li-ion battery
Fig. 20. TEP2 armature current transient caused by the short circuit on the power-line (5 s of
pre-buffer time and 20 s of total signal time)
PSU input
9-15 Vdc
BQ2057W
Li-ion
charger
250 mA
Li-ion
battery
7,4 V
2500 mAh
1-Wire
conncector
F2M03GLA
Bluetooth
module
3,3V DC/
DC
Bluetooth
connection active
TLC3578
14bit 8ch
ADC
+/-10Vin
5V DC/
DC
1-Wire
interface
REF3040
4,096V
PIC18F4680
10 MIPS
mcrocontroller
Status
LEDs
MMA3201
accelerom
eter X,Y
axis
+/-15V
DC/DC
Fig. 21. Block diagram and PCB of the wireless measurement system
The heart of the measurement system is a microcontroller which routes data from the sensors
to the Bluetooth interface. Each DS18B20 has a unique 64-bit serial code, which allows multiple
DS18B20 to function on the same 1-Wire bus. A parallel connection of the sensors simplifies
wire routing over the machine rotor. Routines for searching, addressing and reading 1-Wire
temperature sensors are integrated in the embedded code of the microcontroller.
59
The Bluetooth module is connected with the microcontroller via a USART interface so the
system acts as a virtual serial port on the host computer. It is advisable that the
microcontroller has a pre-programmed bootloader, a part of software which enables
updating of firmware in the rotation through the Bluetooth link.
Because the system operates on the rotating part of the machine it had to be battery
supplied. An absence of the alternating magnetic field on the rotor side of the synchronous
generator excluded the usage of coils for parasite battery charge. Instead, batteries can be
charged via a DC/DC converter connected to the synchronous machine excitation. This
procedure has to be performed with caution due to the high voltage spikes generated by the
excitation system.
To save the battery, only the Bluetooth module has to be powered while the connection is
inactive while all other circuits (microcontroller, sensor supply) should be powered after a
successful Bluetooth connection to the host system is established.
The measurement system has additional features intended for future use, such as a 2-axsis
accelerometer, an 8 channel 14-bit analogue to digital converter and a 15V sensor power
supply.
Data from all the sensors had to be monitored and logged on to the host computer. The GUI
application sends commands to the measurement system via the Bluetooth and waits for its
response. The time base of the measurement is generated by the host system. The
application searches for sensors on the 1-Wire bus and monitors the temperature of each
sensor. Temperatures of each sensor together with time base are logged in a textual file
which is used for further analysis. The GUI application can also read the battery voltage and
battery charge status via the analogue input of the microcontroller. After a successful sensor
search, the application enters in the temperature polling mode. A small database which joins
names to the unique 64-bit address of a specific sensor can be incorporated into the
application. This way, it is easier to monitor temperature of each sensor during an
experiment.
It can be useful if the application has an ability to render 3D temperature distribution over
the surface. The 3D representation greatly improves visualization of temperature
distribution, especially locations of the hot-spots. The 3D model of the salient, shown in Fig.
22 (right), has been drawn and imported in LABView using 3D Sensor Mapping Tool.
Sensors were positioned on the same coordinates as on the real salient pole. Temperature
distribution over the surface between sensors is calculated using interpolation.
The measurement system was mounted on the 400 kVA, 1000 rpm synchronous generator,
Fig. 23. Because of the Bluetooth transceiver it had to be placed in a non metal case. It was
decided that the measurement system would be mounted on a shaft using a custom made
holder. It was mounted in such a way that centrifugal force pushes the plastic case of the
measurement system towards the metal holder and provides reliable fixation. Balancing
weights were mounted on the opposite side of the shaft to reduce vibrations.
After a successful installation of the measurement system prototype on the laboratory
synchronous generator, various experiments have been conducted. Fig. 24 shows an
example of a measurement recorded by the measurement system. The excitation winding
was first heated in standstill and afterwards in the rotation.
60

LOG FILE
PATH
COMMUNICATION
INTERFACE NAME
ADDRESS
ARRAY
PROBE
NAME
ARRAY
TEMPERA
TURE
ARRAY
TEMPERATURE
OVER TIME
CHART
START
BUTTON
STOP
BUTTON
BATTERY
STATUS
Fig. 22. GUI application and 3D temperature field representation
Fig. 23. Mounting of the temperature probes (left); measurement system in rotation (right)
60
D1
D2
D3
D4
D5
J1
K1
L1
L2
L3
L4
L5
N1
N10
N2
N3
N4
N5
N6
N7
N8
N9
P1
P2
P3
S1
Rotation
stopped
Rotation
started
55
50
Temperature [C]
45
40
35
30
25
20
15
200
400
600
800
1000
Sample number
1200
1400
1600
1800
Fig. 24. Temperatures from 27 sensors placed on synchronous generator rotor (2 s sampling time)
61
6. Conclusion
Measurement systems described in this chapter show measurement of most of necessary
physical quantities needed for proper rotating electric machine monitoring. Of course,
measurement is not everything. A good monitoring system should be based on expert
knowledge so that the monitoring system can give advice for maintenance. For large units
such as synchronous generators monitoring systems are different for each machine. During
the creation of such monitoring systems electric machine manufacturer should at least be
present, because he did, during the design process, lots of modelling (analytically, with FEM
etc.) and knows most of limitations of the produced electric machine. It would be the best if
the manufacturer also produces such a monitoring system and has a permanent access to
measurements. In such a way the manufacturer can improve his product, correct some flaws
and optimize it.
7. References
Mohan, N. Electric Drives: An Integrative Approach, Mnpere, ISBN 978-0971529250,
Minneapolis, USA, 2003.
Tavner, P.; Ran, L.; Penman, J.; Sedding, H. Condition Monitoring of Rotating Electrical
Machines, The Institute of Engineering and Technology, ISBN 978-0-86341-739-9,
Stevenage Herts, United Kingdom, 2008.
Vas, P; Parameter estimation, condition monitoring, and diagnosis of electrical machines, Oxford
science publication, ISBN 0-19-859375-9, Oxford University Press Inc., New York,
USA, 2001
Babichev, S.A.; Zakharov, P.A.; Kryukov O.V. Automated monitoring system for drive
motors of gas-compressor units, Automation and remote control, Vol.72, No.1,
January 2011, pp. 175-180, ISSN: 0005-1179, DOI: 10.1134/S0005117911010176
Despalatovic M., Jadric M., Terzic B. Real-time power angle determination of salient-pole
synchronous machine based on air gap measurements, Electric power systems
research, Vol.78, No.11, pp. 1873-1880, November 2008, ISSN: 0378-7796, DOI:
10.1016/j.epsr.2008.03.013
Yang, W.; Tavner, P.J.; Wilkinson, M.R. Condition monitoring and fault diagnosis of a wind
turbine synchronous generator drive train, Renewable Power Generation, IET, Vol.3,
No.1, pp. 1-11, March 2009, ISSN: 1752-1416, DOI : 10.1049/iet-rpg:20080006
Yang W., Tavner P.J., Crabtree C.J., Wilkinson M. Cost-Effective Condition Monitoring for
Wind Turbines, IEEE Transactions on industrial electronics, Vol.57, No.1, pp. 263-271,
January 2010, ISSN: 0278-0046, DOI 10.1109/TIE.2009.2032202
Rangnekar, S.; Nema, R.K.; Raman, P.; PC based data acquisition and monitoring system for
synchronous machines, Proceedings of IEEE/IAS International Conference on Industrial
Automation and Control, 1995, pp. 195 197, DOI 10.1109/IACC.1995.465843
Rahimian, M.M.; Butler-Purry, K.; Modelling of synchronous machines with damper windings
for condition monitoring, Proceedings of Electric Machines and Drives Conference, 2009.,
pp. 577 584, ISBN 978 1 4244 4252 - 2, DOI 10. 1109 / IEMDC. 2009. 5075264
Batzel, T.D. Observer-Based Monitoring of Synchronous Generator Winding Health,
Proceedings of Power Systems Conference and Exposition, 2006., pp. 1150 1155,
DOI 10.1109/PSCE.2006.296470
62
Hanic, Z.; Kovacic, M.; Vrazic, M. Influence of Mounting Temperature Probes on the
Excitation Winding on its Temperature Field, Proceedings of the XIX International
Conference on Electrical Machines (ICEM 2010), ISBN 978-1-4244-4175-4, pp 1-6,
Roma, Italy, September 2010.,
Hanic, Z.; Vrazic, M.; Stipetic, S. Some problems related to surface temperature measurement
of synchronous generator excitation winding in rotation, Proceedings of 14th
International Power Electronics and Motion Control Conference (EPE-PEMC 2010), ISBN
978-1-4244-7854-5, pp T7-15 - T7-20, Skopje, Macedonia, September 2010.
Kovacic, M.; Vrazic, M.; Gasparac, I. Bluetooth wireless communication and 1-Wire digital
temperature sensors in synchronous machine rotor temperature measurement,
Proceedings of 14th International Power Electronics and Motion Control Conference (EPEPEMC 2010), ISBN 978-1-4244-7854-5, pp T7-25 - T7-28, Skopje, Macedonia,
September 2010.
Stipetic, S.; Hanic, Z.; Vrazic, M. Application of IR Thermography to Measurements on
Synchronous Hydro-generator in Rotation, Proceedings of the 10th edition of the
Quantitive Infrared Thermography (QIRT 10), ISBN 978-2-9809199-1-6, pp 1-8, Qubec
City, Canada, July 27-30, 2010
Vrazic, M.; Stipetic, S.; Kutija, M. Methodology of verifying IR temperature measurement on
synchronous generator in rotation, Proceedings of EDPE 2009 - 15th international
conference on electrical drives and power electronics, ISBN 978-953-6037-56-8, pp 1-5,
Dubrovnik, Croatia, October 12-14, 2009
Vrazic, M.; Gasparac, I.; Pavlica, M. Ensuring synchronous generator special standstill tests,
Proceedings of 2008 International Conference on Electrical Machines, pp. 1-5, Vilamoura,
Portugal, 6-9 September, 2008.
Vrazic, M.; Gasparac, I.; Pavlica, M. Some Problems of Synchronous Hydro-Generator
Temperature Measurement, Proceedings of 2008 International Conference on Electrical
Machines, pp. 1-4, Vilamoura, Portugal, 6-9 September, 2008.
Maljkovic, Z.; Gasparac, I.; Vrazic, M. Electromechanical oscillations in bulb type
hydropower unit, Proceedings of 13th International Conference on Electrical Drives and
Power Electronics, ISBN 953-6037-43-2, pp. 7-15, Dubrovnik, Croatia, 26-28.
September, 2005
Gasparac, I.; Maljkovic, Z.; Vrazic, M. Vibrations measurement during hydrogenerator
testing, Proceedings of XVII IMEKO WORLD CONGRESS - Metrology in the 3rd
Milenium, ISBM 953-7124-00-2, pp. 924-928, June 22-27, 2003, Dubrovnik, Croatia
4
Experimental System for Determining the
Magnetic Losses of Super Paramagnetic
Materials; Planning, Realization and Testing
Milo Bekovi and Anton Hamler
University of Maribor, Faculty of Electrical Engineering and Computer Science,

Slovenia
1. Introduction
In recent decades we have witnessed the development of nanotechnology throughout all
fields of science including the field of magnetic materials. When it comes to nanotechnology
is usual for at least one of the dimensions to be in nm scale and therefore the nano-materials
or nano-equipment exhibit certain unique characteristics. The area of magnetic materials
also followed the developments within this field, with the focus on magnetic nanoparticles,
which are typically super-paramagnetic. Within the context of nanotechnology, it is
necessary to observe particles in comparison with solid magnetic materials because the
unique characteristics enable some completely new applications.
Within this section, the magnetic fluid is exposed, as a representative of nano-materials,
because of its unique features of super-paramagnetic properties and liquid state. The merger
of these two properties is enabled because of the magnetic nanoparticles, as discussed in
details in the second chapter. At this point several technical and biomedical applications can
be listed and the benefits arising from the unique properties of magnetic fluids. Example
applications in engineering mechanics regarding rotary sealing have been described by the
authors (Jibin et al., 1992), whilst the biomedical use of these materials enabled new
methods in medical diagnostics and therapeutic methods (Kumar, 2009). In this context the
method of medical hyperthermia has been introduced, which is becoming a promising
alternative treatment for cancerous tissues in contrast to the use of conventional methods
(Barba et al., 2010; Pavel & Stancu, 2009; Pollert, et al., 2007). The aforementioned method
exploits the phenomenon of warming ferrofluids within alternating magnetic field with the
aim of the thermally destroying cancer cells. This paper presents a physical mechanism
responsible for the heating of magnetic fluid within a magnetic field, as well as the
realization of an experimental system heating evaluation.
If modern medicine actually succeeds its usage of magnetic fluid for hyperthermia
treatment then one of the important parameters is certainly its heating power. In reality, this
is about the magnetic losses of magnetic materials in alternating magnetic field but in the
context of a heating element for application within medical hyperthermia regarding heating
power. Therefore, the precise characterization of magnetic losses is crucial and due to the
specific properties of this material differing from conventional approaches, for determining
64
any losses. This requires a measurement system that can determine these losses, this being
the essence of this article. The following shows the design of such a system and its
realization, as well as the actual testing on a magnetic fluid sample.
2. Magnetic losses
Magnetic fluids consist of stable colloidal dispersions of magnetic nanoparticles within a
carrier liquid, where the Brownian motion keeps the particles permanently suspended,
whilst adsorbed long-chain species on the particles surfaces prevent their agglomeration
(Rosensweig, 1997). Their physical properties are governed by the usage of carrier liquid
whilst the magnetic properties are dependent on the types, sizes and concentrations of
magnetic nanoparticles.
If such a fluid is exposed to an alternating magnetic field, its magnetic particles tend to align
in the direction of the magnetic field, also known as magnetic relaxation. Each particle
alignment is accompanied by energy consumption from the magnetic field, which is
outwardly manifested as magnetic losses. Since these fluids usually contain a great number
of particles and the rotations are frequent, as in the case of high-frequency fields, then the
operation of the dissipation mechanisms elevates the temperature of the material. On the
other hand, the amount of heat released within one magnetization cycle is equal to the area
of the hysteresis loop (Carrey, et al., 2011, Bekovi, 2010).
These two properties of the fluid within the alternating magnetic field serve as the basis for
determining of its heating power. The calorimetric method, where the determination of
losses is based on the elevated temperature according to the equation below, is where the
specific absorption rate (SAR) is determined as
SAR
Cs T
,
mFe t t 0
(1)
where is the density of the fluid, Cs is the sample specific heat-capacity, mFe is the total
content of the maghemite (gFe/cm3sample), and T/t is the temperature change, respectively
the initial slope of the measured temperature rise curve. In order to determine any loss by
this method, in addition to the constant parameters of the liquid, measured temperature
response is required, which varies depending on the amplitude and frequency of applied
magnetic field.
The second method provides specific heating power (SHP) or magnetic losses Ph by
determining the surface of the BH hysteresis loop, as given by the equation
SHP
H (t )
0
dB(t )
dt ,
dt
(2)
where the H(t) is the magnetic field strength, B(t) is the measured magnetic induction and f
the frequency of of magnetic field. Both these constants can be used for calculating the
specific heating power p in W/mm3 according to (3), where is mass density of the
magnetic particles.
p SAR SHP
(3)
Experimental System for Determining the Magnetic Losses

of Super Paramagnetic Materials; Planning, Realization and Testing
65
Both presented approaches provide an equivalent loss, as demonstrated in (3), whilst

equations (1) and (2) serve as the base for designing a measurement system.
3. Experimental system
The basic idea for designing this experimental system was determining the magnetic losses
of the magnetic fluid exposed to the magnetic field. The primary need for the realization of
such a system is the generation of alternating magnetic fields with the possibility of
changing the amplitude and frequencies of the magnetic fields, and the container for the
measured sample. The specific needs of a measuring system for determining losses using
the chosen method where determined according to equations (1) and (2). In the case of
calorimetric methods, the temperature of the sample must be measured in order to
determine the derivative (T/t), whilst in case of the magnetic measurements method the
time-dependent magnetic field strength and magnetic induction must be evaluate in order
to determine the integral in (2). Other constants appearing in equations (1) and (2) are
within the domain of the fluids manufacturer of and are usually known.
3.1 The planning
The construction of the system, the design of which can be seen in the figure below (Fig. 1),
was compiled from the results carefully planning its electrical and thermal properties. The
figure only shows the basic elements of the system, but thereafter it is represented by the
result of analyse which has resulted in a decision in the design and selected materials.
Excitation coil (water cooled)

Heat insulation
Temperature probe
Sample of magnetic fluid and
pickup coils
Fig. 1. The scheme of the key elements of the measurement system for the characterization of
magnetic fluid heating power.
The measurement system was incorporated within a measurement scheme as seen in Fig. 2.
which shows it's crucial elements. It is evident that a function generator is used for power
amplifier control, by generating a sinus waveform of various amplitudes and frequencies.
The power amplifier (Amplifier research) supplies an LC circuit, where L is the inductance
of the excitation coil, and C serves to create resonance conditions at a selected frequency.
The temperature measurements are performed using two sensors, fibre optic and resistance.
The measurements are performed using an oscilloscope, whilst control and data acquisition
66
is done via the LabView software package; communication being performed via standard
GPIB and RS 232 busses.
Fig. 2. Measurement scheme.

3.1.1 Magnetic field properties
The authors of some articles on the topic of magnetic fluid-loss characterization use those
measurement systems where the measured sample is exposed to locally-different values of
the magnetic fields, which means that losses are defined as the average value of a given
field. In order to avoid this anomaly, the homogeneity of the field must be ensured; it can be
obtained via the relatively long axial-length of the excitation coil.
This study tested different lengths or coils by increasing the number of coil turns, as show
on the left side of Fig. 3. Turns of the coil were added, as seen, until the desired field
deviations of less than 0.5% for amplitude H had been achieved within the area housing the
sample magnetic fluid. The finite-element method (FEM) software Vector Fields OPERA was
used and a static FEM analysis was performed for each number of turns.
Magnetic field intensity Hp.u.
1,2
1,0
z - position of magnetic fluid sample
0,8
0,6
70 turns
48 turns
36 turns
30 turns
24 turns
16 turns
8 turns
0,4
0,2
0
-200 -150 -100
-50
0
50
z (mm)
100
150
200
Fig. 3. Left side shows the principle of adding image excitation coil turns, right side-the
calculated results of magnetic field strength along the z-axis for different numbers of turns.

67
Partial results for the FEM analyse for the seven cases of different excitation coil turns are
presented at the right side of Fig. 3, showing the amplitudes of the magnetic field strengths
along the z-axis. It can be seen that, in the case of the least coil turns, H has a normal
distribution. Thus it means that the sample was not exposed to the same magnetic field,
both in terms of the direction and the amplitude. In this case, the deviation from the
maximum value was up to 25 % less, so it was necessary to add the coil-turns in order to
achieve homogeneity of the magnetic field. The graph shows the result of increasing number
of turns where it can be seen that adding field-coil turns reduces the deviation of the
magnetic field within the position of the measured sample. A deviation of 0 %, would, in
theory be achieved with an endless coil, therefore, a tolerance less than 0.5 % in the zdirection was acceptable, which was achieved with 70 turns, whilst the discrepancies in the
x and y-directions were negligible.
Fig. 4 shows three examples of magnetic FEM analysis, which show the magnetic field
strengths at the zx-plate in the centre of the measurement system. These results show that, in
the case of a smaller number of turns, the magnetic field in the region of the sample is
clearly non-homogeneous; in addition to archive the same value of H higher values of
excitation current is required and it is limited by the amplifier.
(a)
(b)
(c)
Fig. 4. Results of FEM analysis; magnetic field intensity within the measurement system for
different numbers of excitation coil turns; a) 8 turns, b) 24 turns and c) 70 turns.
3.1.2 Thermal field properties
Besides magnetic calculation during the design process of the measurement system, a
thermal analysis was preformed to ensure heat insulation within the system and the
temperature stability of the sample in case of outside temperature disturbances. Several
papers on magnetic fluid heating power characterization have not focused on the thermal
isolation of their systems, which could result in improper temperature measurements;
essential in calorimetric method. The resulting temperature increase of a sample can also be
the consequence of heat transfer from the heated excitation coil. Therefore, sufficient
thermal insulation must be included within the system, in such manner that any generated
68
temperature increase is merely a result of the desired heating mechanism, which is the
power-loss of the magnetic fluid sample.
For this part of the paper, an excitation coil with 70 turns was selected and thermal analysis
for different examples performed. Fig. 5 displays part of an xz-cross-section of the
measurement system, and the used materials. It also displays the distances or thicknesses of
the materials. The first influence looked at was the impact of warming the excitation coil on
the temperature of the magnetic fluid sample. The heating of the excitation coil was taken
into consideration due to the Joule losses and magnetic fluid for heating caused by the
aforementioned heating mechanisms. The maximum coil-current was set as given by the
maximum output of power supply. By knowledge of the coil resistance, the maximum loss
of excitation coils was determined together with heating power of the measuring sample.
The second influence examined was the influence of the cooling water within the measuring
system, since the plan was to produce a water-cooled coil. Final influence observed was the
presence of a glass vacuum tube, as visible in Fig. 5 and marked within area-I and the
presence of polystyrene as an insulating material on the same figure marked within area-II.
Four different examples of the FEM analysis are collected on the Table. I. For this research,
none of the dimensions where changed, as marked in Fig . 5 but only the presence of the
materials and the water cooling of excitation coil.
Fig. 5. The realized measurement system on the left and 3D model for thermal FEM analysis
on the right.
Ex.
Heat: Mag. fluid

(W/mm3)
A
B
C
D
1e-4
1e-4
1e-4
1e-4
Heat: Excit.
coil
(W/mm3)
1e-7
1e-7
1e-7
1e-7
Cooling
Vacuum results:
results:
Water T Polystyrene
tube
Tmax (C) Tavg (C)
(C)
no
no
no
152
147
no
no
yes
148
139
no
yes
yes
129
121
yes
yes
yes
110
108
Table 1. Examples of thermal analyses.

The result for one thermal analysis for example C can be seen in Fig. 6, which shows a) the
temperature distribution in the measuring system on the xz-cross-section, b) the
temperature derivative and c) the heat flux.
The results for all four FEM examples where the calculated temperature field distributions
throughout the volume; in this case, the most interesting were displayed by the variables
along the x-axis. The straight line was thus at the height z = 0 and x between 60 and 60, as
demonstrated in Fig. 4.

(a)
69
(b)
(c)
Fig. 6. Results of FEM analysis; a) the temperature distribution in the system

b) the temperature derivative and c) the heat flux within the system; results obtained for ex- C.
Figure 7 shows the results for temperature evaluated along the aforementioned line, for all
four cases. It can be seen that the sample's temperature was generally higher; less of the
insulation was included in the calculation. Notably in the last case, when the cooling water
with the same temperature as the surroundings is included, the sample fluid is heated only
because of attributed heating power, hence the lowest temperature. The impact of the
ambient was minimized. When observing in the course of the heat flux (right picture) for all
four cases along the same line it appeared, that in the position 40 < z < 50, where the coil
was located, there was a source of heat-flow directed to the outside and inside of the system,
thus causing unnecessary heating of the sample. When a layer of insulation was added
between the coil and the sample, heating was reduced (ex. C).
Temperature T (C)
120
100
80
ex. B
ex. B
ex. C
ex. C
ex. D
Tex. D
60
40
20
-60
-40
-20
0
x (mm)
20
40
ex. A
ex. A
ex. A
140
1.0
Heat Flux Q ( mW/mm )
160
60
0.8
ex. B
ex. A
ex. C
0.6
ex. D
ex. B
0.4
ex. D
ex. C
0.2
0.0
-60
-40
-20
0
x (mm)
20
40
60
Fig. 7. Results of FEM analysis; temperature along the x-axis for examples A, B, C and D and
the right picture, the heat flux along the x-axis for the same examples.
70
3.2 The realization
The result of this planning was a realization of the measurement system shown at the left of
Fig. 8. The right picture is a 3D model and reveals more details, since it is partly cut out. This
model was also used for all the FEM calculations presented in the previous subchapter.
Vacuum tube
Excitation coil
Insulation - polystyrene
Pickup coils
Test tube
Electrical connection and

cooling water connection
Fig. 8. The realized measurement system left side and 3D model for thermal FEM analysis
right side.
3.3 Measurements of magnetic field parameters
So far the focus has been on the thermal insulation of the measurement system for better
implementation when characterizing fluids, according to calorimetric method (3), whilst
these sections have shown the principle of magnetic field variables measurement when
determining the losses based on the determination of the hysteresis loops surface (2). From
the equation, it is clear that it is necessary to measure the magnetic field strength H(t) and
magnetic flux density B(t). For this measurement a system of three measuring coils has been
created, which bond to each other, as shown in Fig. 9. The top view displays the magnetic
fluid sample, the middle shows around the sample containing a glass test tube for sample
storage, and further outwards the two pickup coils for measuring the induced voltages e1
and e2. These two voltages are used in determination of magnetic field strength and
magnetic flux density according to (5) and (6).
Fig. 9. Principle of measuring B(t) and H(t) by means of two pickup coils.

71
First let's look at the magnetic flux density B(t), as defined by equation (4), where N1 is the
number of inner pickup coil turns, As is the cross-section of the sample, and 1 the magnetic
flux in the sample.
Bt
1
N 1 As
(4)
The inner measuring tube coil tightly embraces the vial so that the magnetic flux can be
replaced by the integral of a measured induced voltage, as shown in (5).
Bt
1
e1 t dt
N 1 As
(5)
Measurement of the magnetic field strength is a greater challenge, since the magnetic flux
2 should be measured close to the sample, whilst excluding samples magnet flux. The
external pickup coil marked in Fig. 9 is used for this measurement and it is evidently wound
in such a manner that embraces the sample.
H t
0 N 2 A2
1
e2 t dt
0 N 2 A2
(6)
Equation (6) contains N2 and A2 which are constants and represent the number of coil turns
and its cross-section. It was determined experimentally in such a manner, that the magnetic
flux linkages where measured whilst exposing the coil to the known density of the magnetic
field.
4. Testing
This section verifies the operation of the measuring system on a sample of magnetic fluid.
Firstly, the calorimetric method is described and then those methods based on the magnetic
field quantities measurements.
The magnetic fluid sample used in this work is commercially available; therefore some tests
were performed to determine the basic structural properties of the fluid. Chemical analysis
revealed particles of maghemite -Fe2O3 dispersed in mineral oil. Transmission electron
microscopy revealed the mean diameter of particles to be 10.9 nm. Magnetic measurement
of its magnetization curve was performed and modified according to the Langevine theory
of paramagnetism (Pshenichnikov) the magnetization of particles, close to saturation, is
described by
kBT
1 ,
M Ms 1
3
M
/
6
d
0 s bulk
core
(7)
The volume fraction of the magnetic particles in the used sample of magnetic fluid was
determined from a ration of Ms/Ms bulk, where Ms bulk was 400 kA/m and it is 10.57 %.
4.1 Magnetic fluid characterization SAR
Calorimetric measurements were carried out using a calorimeter, which is an instrument

that allows the heat-effects within it to be determined by direct measurement of
72
temperature, as proposed in (1). The sample of magnetic fluid was exposed to an ac

magnetic field. The measurement principle for the calorimetric measurements required a
stable temperature of the sample, before the magnetic field could be switched on. For this
reason, the measurement could only begin when the temperature of the sample was
constant, usually at the value of cooling water. When this condition had been fulfilled, the
excitation coil was turned on and constant frequency and regulated current set up the
magnetic field at the desirable value. The measurement lasted until the temperature steady
state had been reached or if the other heat production equalled the heat dissipation.
120
100
80
60
40
20
0
0
50
100
150
Time t (s)
H = 9.5 kA/m
H = 7.4 kA/m
H = 6.5 kA/m
H = 4.0 kA/m
H = 2.6 kA/m
200
250
300
o
T/t ( C/s)
Temperature T (C)
Several tests, for an 8 ml sample of magnetic fluid, of different magnetic field strengths were
performed and the results for temperature rise are presented in Fig. 10. The measurements
were performed at a frequency of 400 kHz and constant field amplitudes, that were
regulated to achieve a discrepancy of 10 A/m around the selected values.
H = 9.5 kA/m
H = 7.4 kA/m
H = 6.5 kA/m
H = 4.0 kA/m
H = 2.6 kA/m
2
1
0
-1
0
50
100
150
200
Time t (s)
250
300
Fig. 10. Calorimetric measurements of temperature rise for different values of magnetic field
strengths at frequency of the supply field of 400 kHz, and the temperature derivatives.
This approach is not new regarding magnetic fluid determination, but it serves as a
verification method for determining losses through hysteresis loops or through complex
susceptibility, as presented below.
4.2 Magnetic fluid characterization SPL
This section explores the methods for determining the losses of magnetic fluids at constant
temperature. For this, equation (3) was used to calculate the losses from the area of the
hysteresis loop and, in the second case, out of a complex susceptibility, as explained later,
both based on the measurement of time courses H(t) from equation (6) and B(t) from (5). In
the analysis, magnetic polarization was used instead of magnetic flux density, where
B J 0 H .
(8)
Another approach for determining losses is based on (9) according to Carrey, where the
parameter of the complex susceptibility '' represents the essential component
P
0 H 2 '' f .
(9)

73
This section shows the equivalences of these methods for the sampling the magnetic fluid
and point out the advantages of these methods against the calorimetric method. In this case,
the characterization losses are executed under such conditions as would be applied during
actual treatment with hyperthermia.
The relationship between relative permeability and susceptibility was determined using this
expression
J 0 H
B
J
1 1.
0 H
0 H
0 H
(10)
The responses of the material to the applied magnetic field were studied using the complex
susceptibility approach. In this case the susceptibility was treated as a real ' component,
which is within the phase with the magnetic field and imaginary component '' that lags
behind the 90, the latter is also related to the energy losses or absorbed energy by the
sample from the ac field respectively. The terms for both were written (11), where was the
angle between the measured signals H(t) and J(t) or B(t)
J
0 H
'
J
J
cos , ''
sin ,
0 H
0 H
(11)
The characterization was carried out in such a manner that different values for power
amplifier currents were set, and thus the magnetic field strength whilst measuring the
voltages of the two pickup coils. Such an example occurs in the lower two graphs showing
the courses of the induced voltages e1 and e2 for the change of gain over seventeen steps.
3
Pickup voltage e2 (V)
Pickup voltage e1 (V)
-5
0
4
6
Time t (s)
10
2
1
0
-1
-2
-3
0
4
6
Time t (s)
10
Fig. 11. Measured induced voltages e1 and e2 for various examples of magnetic fields
between 5 and 15 kA/m.
Here it should be noted that the measurements must be carried out within a very short
period of time, since a long exposure of the sample to the high value of the field quickly
rises the temperature, as is clear from the heat curves in Fig. 10. Magnetic fluid at higher
temperatures behaves differently, since it's the closer to Curies temperature which also
decreases its heating power because of the narrower hysteresis loop. From the measured
induced voltage the desired variables H(t) and B(t) can be calculated, which are plotted in
Fig 12 for the same example.
74
30
Magnetic flux density B (mT)
Magnetic field strength H (kA/m)
20
10
-10
-20
0
4
6
Time t (s)
10
20
10
0
-10
-20
-30
0
4
6
Time t (s)
10
Fig. 12. The time courses of the calculated values H(t) and B(t) for the same example.
The measured signals provide the determination of losses using both methods, namely by
determining the hysteresis loop area using equation (2) or through the determination of the
complex susceptibility and losses calculated by equation (9). Fig. 13 below shows the two
approaches, the left images are the outlined JH hysteresis loops for determining the losses
by the first method. Determination of the surface hysteresis is no novelty, but an interesting
approach for the evaluation of the dynamic hysteresis loop areas of super paramagnetic
materials, and the determination of loss by this method.
The right side figure shows the magnetic filed 0H box and the response of materials to the
field of magnetic polarization J(t) (you could also use B(t )). The goal was to determine the
phase shifts of signals to be able to determine the '' according to (11).
0.03
Magnetic polarization J (mT)
Magnetic polarization J (mT)
30
20
10
0
-10
-20
-30
-20
-10
0
10
Magneti field strength H (kA/m)
20
t ~
0.02
0.01
0
-0.01
-0.02
-0.03
0
0H
2
4
6
Time t (s)
10
Fig. 13. Hysteresis loop and the time course of signals for determining magnetic losses.
If the plot of the phase shifts in delta is now done for all seventeen cases, it is found that
only the latter is not constant but decreases linearly, as shown in Fig. 14. According to the
formula for the losses determined from a complex susceptibility P (9) the desired
characteristic P = f (H) can be calculated, which is for example of the 100 kHz frequency
given in the same graph.

75
and H - J ( )
P (W/g)
H - J meas.
meas.
lin. aprox.
4
3
0
0
H - J aprox.
5
10
Magnetic field intensity H (kA/m)
15
Fig. 14. The result of the phase shift measurements for different magnetic field strengths
and associated loss of magnetic fluid sample at 100 kHz.
When the losses of Ph were plotted in the same graph as the loss of P it was evident, that
both analyses of the measured signals offered the same results. The results are summarized
in Fig. 15 for four different frequencies of the applied field and it was discovered that the
losses expectedly increased when increasing the frequencies.
25
Heating power P (W/g)
, f = 678 kHz
20
Ph ,
f = 678 kHz
, f = 403 kHz
15
Ph ,
f = 403 kHz
, f = 265 kHz
10
Ph ,
f = 65 kHz
, f = 100 kHz
5
0
0
Ph ,
f = 100 kHz
5
10
15
Magnetic field intensity H (kA/m)
20
Fig. 15. A comparison of heating powers P and Ph for four different frequencies, and the
amplitudes of magnetic field intensity.
5. Conclusion
In conclusion we can point out specific guidelines that apply to measuring systems for
determining the losses of magnetic fluids. In this context, the losses are a desired property,
because it means greater heat output and thus higher quality material for usage in medical
hyperthermia.
In order to build such a measurement system it is necessary firstly to provide a source
magnetic field that must be able to change both the amplitude and frequency of the magnetic
field. In achieving this objective, we have tried to approximate as closely as possible those
fields parameters that would actually be best suited for implementing this treatment.
76
Secondly, we need a measurement system which is sufficiently precise and will allow for a
clear characterization of magnetic fluid losses. This article presents three methods, where
each of them exposed their weakness, and the possibility of erroneous measurements.
In the calorimetric method, the emphasis is on adiabatic changes in temperature, at least at
the start of the measurement, which determines the maximum slope of warming. For FEM
analysis, we have demonstrated the importance of certain measures for improving the
methodology.
The method of magnetic measurement emphasizes the accuracies of the measuring coil
parameters, since only with precisely defined parameters can accurate measurements be
performed. We here shown the equivalence of both methods, where we see that the
determination of the hysteresis loop area and determination of the angle between the
polarization and the magnetic field brings almost the same result for determining the losses,
depending on the accuracies of certain constants.
6. References
Barba, P., Dughiero, F., & Seini, E. (2010). Magnetic field synthesis in the design of inductors
for magnetic fluid hyperthermia. IEEE Transaction on Magnetics, Vol. 46, No. 8, pp.
2931-2943, 0018-9464.
Bekovi, M., & Hamler, A. (2010). Determination of the heating effect of magnetic fluid in
alternating magnetic field. IEEE Transaction on Magnetics, Vol. 46, No.2, pp. 552-555,
0018-9464.
Carrey, J., Mehdaoui, B., & Respaud, M. (2011). Simple models for dynamic hysteresis loop
calculations of magnetic single-domain nanoparticles: Application to magnetic
hyperthermia optimization. Journal of Applied Physics 109, 083921-1-17, 0021-8979.
Jeong, J. R., Lee, S. J., Kim, J. D., & Shin, S. C. (2004), Magnetic properties of Fe3O4
nanoparticles encapsulated with poly (D,L Lactide-Co-Glycolide). IEEE Transaction
on Magnetics, Vol. 40, No.4, pp. 3015-3017, 0018-9464.
Jibin, Z., & Yongping L., (1992). Numerical calculations for ferrofluid seals. IEEE Transaction
on Magnetics vol. 28, No.6, pp. 3367 3371, 00189464.
Kim, D. H., Thai, Y. T., Nikels, D. E., & Brazel, C. S., (2009). Heating of aqueous dispersions
containing MnFe2O4 nanoparticles by radio-frequency magnetic field induction,
IEEE Transaction on Magnetics, Vol. 45, No.1, pp. 64-70, 0018-9464.
Kumar, C (Ed(s)). (2009). Magnetic Nanoparticles, nanomaterials for life science Volume 4,
Wiley-vch Verlag GmbH&Co.
Pavel, M., & Stancu, A. (2009). Ferromagnetic nanoparticles dose based on tumor size in
magnetic fluid hyperthermia cancer therapy. IEEE Transaction on Magnetics, Vol. 45,
No.11, pp. 5251-5254, 0018-9464.
Pshenichnikov A.F., Mekhonoshin V.V., & Lebedev A.V., Magneto-granulometric analysis
of concentrated ferrocolloids, J. Magn. Magn. Mater. 161, pp. 94-102, 1996.
Pollert, E., Kniek, K., Maryko, M., Kapar, P., Vasseur, S., & Duguet, E., (2007), New Tctuned magnetic nanoparticles for self-controlled hyperthermia. Journal of Magnetism
and Magnetic Materials, 316, pp. 122-125, 0304-8853.
Rosensweig, R.E., Ferrohydrodynamics, Dover publications, inc, Mineola New York 1997.
5
Non Contact Measurement System
with Electromagnets for Vibration Tests
on Bladed Disks
Christian Maria Firrone and Teresa Berruti
Department of Mechanics, Polytechnic University of Turin (POLITO), Turin,

Italy
1. Introduction
In vibration testing one of the main concerns in the measurement of the forced response of
vibrating structures is the coupling between the test article and the exciter. When the
structural behaviour of the structure is pretty linear a hammer test allows getting a first idea
of the resonant frequencies and mode shapes without perturbing the dynamic response.
Unfortunately, real applications of structures whose behaviour is linear are rare. Usually,
structures are made of components assembled together by means of joints whose behaviour
may result highly nonlinear. Depending on the amount of excitation, joints can dramatically
change the dynamic behaviour of the whole system and the modelling of this type of
constraint is therefore crucial for a correct design. For this reason, tests are performed to
characterize the contact parameters of joints focusing the attention in the material used in
the application or extending the study by considering the whole components coupled by a
particular geometry of joint. The characterization of joints in dynamics requires the
knowledge of the amplitude of the exciting force which is usually set as a harmonic
waveform whose frequency varies within a range of interest in the so called stepped-sine
test which overcomes two limits of the hammer test. In fact, the stepped-sine test controls
the force amplitude at each value of frequency in a given range by means of a feedback
control which activates the measurement only when the target force is reached. Moreover a
realistic value of force amplitude must be reached in order to explore the nonlinear
behaviour of the structure. For this reason electromagnetic shakers are used by connecting
the exciter to the test article by means of a stinger. Different strategies are used to detune the
dynamics of the test article itself from the dynamics of the exciter , they are based on the
design of the stinger or on the constraint of the shaker. Unfortunately, these strategies are
not useful in particular applications where a small perturbation may introduce large
changes in the dynamics of structures or in general it may interfere with the measurement of
the physical phenomena. In turbomachinery, this is the case when the nonlinear friction
damping of the blade root attachment to the disk is studied. The amount of damping
introduced by the contact is very small, nonetheless it is still of the same order of magnitude
of the structural damping of the blade and it can be increased by a proper design of the
blade root geometry (dovetail, fir-tree, flat, crowned). A contact excitation of the blade like
the one required by the electromagnetic shaker would perturb the measurement of the
78
friction damping, for this reason it is preferred to excite directly the base of the platform
where the blade is attached. Another example is the study of the forced response of rotors
excited by a multiple source of excitation, which is the application of the study presented in
this chapter. Two main features of the bladed disk design concerning the dynamic response
are veryimportant. 1) The rotor is subjected to an excitation pattern where each blade is loaded
with a series of pulses within a complete disk rotation. The number and the intensity of pulses
depend on the architecture of the engine preceding the rotor (number of combustion
chambers, stages, stator vanes) and the wide spectral content is usually characterized by
several harmonic components whose excitation frequencies are mainly a multiple of the rotor
angular speed. The harmonic index which defines the multiplicity is called Engine Order (EO).
2) A rotating disk is nominally characterized by cyclic symmetry properties (Thomas, 1979)
where one fundamental sector constituted by the blade and its associated disk sector is
repeated for a number of times around the rotation axis. The periodical repetition of geometric
and material features (tuned configuration) generates a particular dynamic behaviour of the
disk where natural frequencies, and corresponding modes, can be grouped into families
featuring the same deformed shape of the vibrating blades and disk (Castanier & Pierre, 2006).
The main objective at design stage is to avoid any matching of the rotor natural frequencies
with the excitation frequency of each harmonic component of the load spectrum which
shortens the life of the component due to HCF failure of blades. Unfortunately, bladed disk
dynamics may show unexpected response due to small differences between sectors
nominally identical which determine a mistuned configuration of the system (Castanier &
Pierre, 2006, Kenyon & Griffin, 2003). In worst cases, a distorted and localized forced
response occurs in particular for high modal density regions where one or more blades
vibration amplitude is higher than the tuned response amplitude. Causes of mistuning are
multiple: irregularities associated to the material properties constituting the blades,
asymmetry generated during assembly or during service due to wear assessment, different
contact condition at the blade root joints, snubbers, shrouds in case of non-integral bladed
disks, Foreign Object Damage (FOD) or modifications introduced during maintenance and
repair processes. Another source of mistuning comes from the so-called underplatform
dampers (UPDs) which are metal parts located within cavities between contiguous blades.
UPDs aim is to limit the vibration amplitude of blades in case of excitation at resonance by
means of friction damping at the contact surfaces. Mistuning is introduced at the contacts
which couple directly the blades subjected to wear and change of the contact position
during service. Interesting studies were carried out where UPDs are seen not only as a
further source of mistuning to take into account but also as a mean to mitigate negative
effects of inherent mistuning (Pierre et al., 2002, Petrov et al., 2010) even by distributing a
number of UPDs smaller than the number of the blades (Avalos & Mignolet, 2008) or by
systematically using UPDs having different mass (Gotting et al., 2004).
Fimite Element Modelling, mistuning identification and numerical results obtained by
taking into account mistuning in the design must be validated with tests on dummy or real
disks. Test campaigns are performed on test rigs which can be classified in two groups
having a basic dual property: the first group has rotating disks with fixed excitation source,
on the contrary the second group supports a static rotor excited by a travelling excitation.
Both test rigs have advantages and limitations which must be addressed with respect to the
objective of the test campaign.

with Electromagnets for Vibration Tests on Bladed Disks
79
Rigs with rotating test articles are generally more complex than static rigs, first of all due to
the design of the disk containment which must fulfil safety requirements in case of blade
release (usually they are built in a pit or within a protective bunker). Moreover, different
levels of complexity are found in the comprehension of the mistuning phenomenon if tests
are performed on real engines and cold flow rigs since fluid-structure interaction must be
considered. In detail, correlations between mistuning and aero-coupling, damping and
elastic effects of gas flow which decreases the magnification factor of the mistuned response
(Beirov et al., 2008) and the influence of mistuning on flutter instabilities onset can be
studied. On the contrary, fluid-dynamics effects can be avoided in vacuum chamber (spin
test rig). For both types of rotating rigs, however, gyroscopic and centrifugal force effects
(e.g., blades untwisting, dependence of UPDs contact pressure on disk velocity) should be
taken into account. The excitation provided by the gas flow must be substituted in vacuum
rigs with an alternative source. Broad applications of non-contact excitation provided by
magnetic fields are found in literature by means of permanent magnets or electromagnets
whose attractive force is exerted on ferritic bladed disks. Permanent magnets are compact,
not expensive exciters which can be easily bought at different material grade of
magnetization working in some cases at a temperature up to 200C. One blade is excited by
the magnets which are equally spaced around the circumferential coordinate and is loaded
during one disk rotation with a number of pulses equal to the number of magnets. The
excitation frequency is therefore strictly proportional to the rotation frequency by the
number of magnets which determines the engine order of the force pattern. Electromagnets
allows to release the excitation frequency from the rotation speed by using an AC feeding
(Prchlik et al., 2009) in spite of a more complexity of the excitation system: more room is
required to place the EMs, heat must be dissipated (which can be a critical issue in vacuum),
an amplifier system and a signal generator must be provided. Hybrid solution can be also
found in literature as in (Rice et al., 2007) where magnetic excitation is used as an extra
source of excitation in engine tests with gas flow in order to explore flutter onset at different
engine orders for the same rotation speed i.e., for the same aerodynamic load, air mass flow
and centrifugal effects. Moreover in (Szwedowicz et al., 2007) a really cheap exciter is
constituted by an air jet simply generated by the pressure drop between room condition and
the vacuum chamber. In this case only one exciter can be provided since it is not possible to
control a given time delay between more pipes which are necessary if an engine order type
excitation is required. On the contrary, a narrow, single pulse virtually contains all the
engine order indices which can be calculated by Fourier series decomposition of the force
profile along the spatial domain of the disk circumferential coordinate. A contacting
excitation system is used in (Szwedowicz et al., 2006, Gilbert et al., 2010) by means of
piezoelectric actuators applied to the root of the blades. Piezo-actuators need electric feeding
supplied by an end-shaft slip-ring connected to the shaft. Since they are contacting devices,
on one hand this technique may cause non-intentional mistuning which should be taken
into account. On the other hand, a precise force control is achievable through the calibration
of the piezo-devices. However, the dynamic perturbation induced by the exciters can be
limited if they are glued to the disk (disk mistuning) which usually does not generate critical
mistuning phenomena as perturbed blades do (blade mistuning).
The determination of the magnification factor requires the measurement of each blade
response typically obtained by means of strain gauges which are again contacting devices
80
which can cause a non-intentional blade mistuning (Beirov et al., 2009). Tip timing
techniques (Heath & Imregun, 1998, Heinz et al., 2010) proximity probes and rotating laser
Doppler vibrometer (LDV) purposely designed (Sever, 2004) or already available in
commerce are valid equipments which do not interfere with the original bladed disk
dynamics. The first two techniques measure the disk response with respect to a stationary
reference system while the third technique allows measuring the disk response on a
reference system rotating with the disk. Therefore, attention must be paid to correlate input
and output data of the rig with simulation results since the transformation of the coordinate
system from static to rotating and vice versa is required. A not negligible aspect involving
LDV measurement is the necessity of visual access to the rotating bladed disk which may be
in conflict with safety requirements.
The second group of test rigs carries a static rotor and a rotating excitation. The travelling
wave is in this case generated by contacting (piezo-actuators (Kruse & Pierre, part II, 1997),
shakers (Strehlau & Kuhhorn, 2010)) or non-contacting systems (electromagnets with AC
feeding (Kruse & Pierre, part I, 1997), speakers (Judge et al., 2003, Jones & Cross, 2010)).
Whatever the case, a control system must be purposely developed in order to activate the
exciters with a given phase shift in time to mimic the engine order force pattern. Since the
bladed disk does not rotate, the dynamic response is directly measured on the reference
system fixed to the rotor by using the same devices already described. In particular, the
rotating LDV is substituted with a scanning LDV which measures, in its basic configuration,
a set of points previously defined on the plane including the rotor (Strehlau & Kuhhorn,
2010, Berruti et al., 2010). Compared to the case of rotating rotors in this case aerodynamic
coupling effects are avoided as well as complications given by data transmission. Such test
rigs allow to easily collect measurements of the dynamics of dummy or real bladed disks
directly related to mistuning and are therefore suitable for validating mistuning
identification methods and reduction techniques with precision.
When non-linear phenomena are present as friction damping occurring at joints, the control
of the force input is of primary importance since the forced response is strongly influenced
by the amount of the local displacements of the structure at the contact surfaces. The force
pattern can be measured during tests or estimated by means of a calibration process. Forces
generated by air jets, for instance, can be measured with manometer probe measuring the
pressure field before testing (Chang & Wickert, 2002) or with probes embedded in the
surface of the blades (Petrov, 2010) or estimated as in (Szwedowicz et al., 2007) by
calculating analytically the peak force applied to the blade as the product of the velocity at
the end of the pipe and the mass flow hitting the blade. In this case the calculation relies on
some simplifications about the area of the impacting gas and the distribution of the pressure
on the blade airfoil. A similar approach can be used for oil jets whose higher density is
suitable to excite very large blades. Much lower force amplitudes can be reached by
speakers which can be calibrated by recording the spectra of a white noise by means of a
microphone (Judge et al., 2003, Jones & Cross, 2010). The dynamics of the speaker can be
recorded and used as a calibration curve in order to guarantee the same intensity at each
frequency, while the value can be estimated by measuring the pressure on the microphone
multiplied by the area of the speaker. Attention must be paid to the air gap between each
blade and speaker since a small difference can cause an unacceptable error on the time delay
between the harmonic force profiles which actually load the blades.

81
Force measurement using the magnetic field as a source of excitation can be performed by
placing a force transducer below the magnet. In the case of permanent magnets, since it is
usually a single piece, compact device, it can be considered as an additional mass which
must be taken into account in order to verify if the frequency upper limit of the force
transducer drops to values within the tested bandwidth. In the same way, force transducers
can be used with electromagnets. Unfortunately, electromagnets are more complex devices
than the permanent magnets, since components like conducting wires, packed plates, and
insulating materials are assembled together and may generate a non-linear dynamic
behaviour of the exciter itself. Similar to piezo-actuators (Kruse & Pierre, part II, 1997), they
must be supported by an adequate power amplifier system in order to generate the required
force amplitude in particular for high frequencies if non-linear dynamics characterization
induced by slipping phenomena is addressed.
In this chapter, the design, manufacture and calibration of a travelling wave excitation system
for bladed disks is described and tested. The typical application of the system is the Frequency
Response Function (FRF) measurement by means of the stepped-sine technique for the
characterization of the complex phenomenon of mistuning and non-linear dynamic response
in presence of friction contacts such as blade root joints, shrouds or underplatform dampers.
The measurements provide accurate results to be compared with numerical prediction.
The essential requirements of a travelling wave excitation system for such kind of tests in
presence of mistuning are:
-
the absence of contact between blade and exciter

the same excitation force amplitude on each blade modulated with a given time lag
which is determined by the engine order index.
If the test setup involves the characterization of nonlinearities due to friction the requisites
listed above are not enough, since the excitation system should be able to provide also:
-
a high excitation force amplitude in order to produce slipping between the friction
contacts,
an accurate measurement of the force amplitude of the rotating force pattern.
The excitation produced by an array of electromagnets (EMs) is non-contact and is then

suitable for this kind of measurements. Moreover, any travelling wave force can be
generated on the disk by changing the phase delay of the waveform of each current signal
flowing on the EMs. The system hardware and software are purposely designed to satisfy
all the requirements listed above. The system set up requires an accurate calibration process
to generate the travelling force applied to the structure by each EM. The calibration consists
of two steps: the first step is based on an instrumented EM which measures the force exerted
on the structure, the second step is a tuning process in order to regulate all the other EMs for
the generation of the same force amplitude on each blade.
In the first part of the chapter the hardware and software of the excitation system setup are
described, together with a summary of the theoretical background that is necessary for the
EM design. In the second part the system set up and calibration is presented by focusing on
the proposed tuning method that allows the calibration of the EMs on site i.e., under the
disk in their final position. In the last part of the chapter an example of application on a
dummy bladed disk is presented.
82
2. The excitation system

The excitation system sketched in Figure 1a) is based on an electromagnetic unit, that is the
electromagnet + the aluminium support (EM) repeated cyclically in order to excite a cyclic
structure like a blade array where one EM is placed under each blade (Berruti et al, 2010,
2011). Each EM is mounted with a certain gap with respect to the blade so that the blade is
not in contact with the exciter. The force is generated by the magnetic induction flowing
through the EM, the air gap and the ferromagnetic material of the blade. The EMs are
mounted on an aluminium circular plate (Figure 1-b). The aluminium is chosen in order to
not interfere with the magnetic flux flowing in the EMs. Each EM is fed with a voltage with
a given amplitude and phase. The signal generator system National Instruments cRIO 9074
is connected to the audio amplifiers Crown CDi 2000 (2 channels each, 800 W power supply
for each channel). Each EM is connected to one channel. A LabVIEW software is purposely
developed in order to generate a number of voltage harmonic signals equal to the number of
EMs. The voltage signals are amplified in order to feed the EMs with a given amplitude and
phase. The resulting excitation system has the following features:
-
by feeding each EM with a voltage with a different phase from the neighboring it is
possible to reproduce an engine order (EO) type excitation typical of the excitation of
turbine disks in the engine;
by controlling the force amplitude it is possible to keep it constant during the step-sine
test;
by tuning the EMs with respect to a reference one it is possible to reach the same
excitation force amplitude on each EM.
(a)
(b)
Fig. 1. a) The cyclical excitation system; b) The aluminium circular plate with two EMs.
2.1 The electromagnet unit design
The electromagnet design strictly depends on the type of application. The general design
criteria are hereinafter addressed. The electromagnet is designed in order to satisfy two
essential criteria related to i) the geometry of the cyclic structure (bladed disk) excited by the
EM and ii) the maximum force and maximum frequency of the planned travelling wave
excitation, in detail:
-
the EM size must be limited according to the room between two adjacent blades of the
disk under test. As an example, the bladed disk shown later in the application allows a

83
maximum square cross area for a commercial EM core S=16x16mm2. The U-shaped core
outer dimensions are therefore automatically determined: 64X48x16 mm.
the EM must produce a force amplitude up to 10N in a mechanical frequency range
from 100Hz to 600Hz given an average air gap lair=2mm minimum. The last parameter
is set in order to consider the variation of the gap during vibration negligible. For this
reason lair will be considered constant during the design process.
Figure 2 shows the designed EM unit which must face the blade with a gap that can be
adjusted according to the application. Each electromagnet is made of two coils (1) of wire
wrapped around a U-shaped core of ferromagnetic packed plates. Two prismatic
ferromagnetic extensions (2) are glued at the two ends of the U shaped core in order to face
the blade at the stagger angle . This solution was preferred instead of producing a single
piece EM core in order to change the prism in case of bladed disks with different stagger
angles. The prismatic extensions are made up of SMC, i.e. a sintered ferromagnetic material
which strongly limits the eddy currents.
(2)
(1)
Fig. 2. Detail of the EM unit.

2.2 The theoretical model of the magnetic force
The attractive force on the blade can be estimated by means of a simple model of the
magnetic circuit of the electromagnet facing the blade. Electromagnet and the equivalent
magnetic circuit are sketched in Figure 3 a) and b) respectively. The two coils are modelled
as two generators of magnetomotive force Ni, where N is the number of turns for each coil
and i is the current. The part of the circuit corresponding to the air and the part of the circuit
corresponding to the ferromagnetic materials are characterized by the values of the
reluctance :
l
S
(1)
where l and S are respectively the length and the cross area of the part of the circuit
characterized by , is the magnetic permeability. steel is the magnetic reluctance of the
steel core and of the steel blade facing the electromagnet while air is the magnetic
reluctance of the air gap. steel is much less than air since steel >> air . The circuit can then
be simplified with the circuit of Figure 3 c) where steel is neglected.
84
steel
Ni
Ni
N
steel
N
S
S
blade
lair
steel
air
2Ni
air
2 air
steel (blade)
(a)
(b)
(c)
Fig. 3. a) Sketch of the electromagnet facing the steel blade; b) Equivalent magnetic circuit;
c) Simplified equivalent magnetic circuit.
The following relationships can be written for the circuit of Figure 3 c):
2 Ni 2air 2 Ni 2
2 Ni 2
lair
air
BB
lair
airS '
Ni air
(2)
lair
where is the magnetic flux, lair is the air gap, S is the area of each coil of the EM facing
the blade (S=S/sin()), B is the magnetic induction. The attractive force f exerted from the
EM on the blade is the sum of two forces fcoil exerted by the two branches and is related to
the magnetic induction B by the following relationship (principle of conservation of energy):
B2 S '
f 2 f coil 2
2 air
(3)
Substituting Eq.(2) in Eq.(3), the following relationship between the force and the current
holds:
Ni 2 air S '
2
lair
(4)
When the electromagnet is fed by an alternating current i(t ) I sin t , with electrical
angular frequency =2fel (being fel the electrical frequency in Hz) and a given phase shift,
the force in Eq.(4) becomes:
N 2 air S ' 2 2
N 2 air S ' I 2
(
I
sen
t
1 cos 2t 2 fs f a
2
2
2
lair
lair
(5)
where fs and fa are the static and the alternate component of the attractive force. The
alternate component of the attractive force becomes:
fa
N I 2 air S ' cos 2t 2
2
2lair
FA cos 2 f el 2 t 2 FA cos f m 2 t 2
(6)

85
The force fa is the force that excites dynamically the blade at the mechanical frequency
fm = 2 fel. The blades are then excited with a mechanical frequency fm that is two times the
electrical frequency fel.
The force amplitude FA is determined by the product NI. The current amplitude should be
limited to a maximum value of 10A in order to avoid risk for the insulation due to the
generation of heat.
In the next paragraph the relationship between the exciting force amplitude FA and the
value of the voltage amplitude V generated by the amplifier is given, since it is of more
practical use to know the value of voltage that must be generated by the amplifier in order
to excite each blade at a desired force.
2.3 The electrical circuit model
In order to characterize the electrical circuit of the EM the amplitude value of the AC current
I is measured for different values of amplitude and frequency of the AC applied voltage V.
Figure 4 shows the value of the measured current amplitude I versus the input voltage
amplitude V for different frequency values. Note that the trends are linear and for a given
voltage value the current (and therefore the amplitude of the force FA, Eq.(6)) flowing in the
EM decreases as the frequency increases. The admittance quantity Y = I/V is diagrammed in
Figure 5 (experimental values are presented with black dots). Y is a useful quantity which
characterizes the EM in order to know what is the current (and then the actual force) that
can be expected given a value of voltage V (representing the generated signal) at a given
frequency. This information is of primary importance if controlled force, step-sine tests must
be performed. In this case in fact, the voltage must vary during the frequency sweep in
order to guarantee the same output current amplitude and therefore the same force
amplitude.
Assuming a simple RL model of the electrical circuit the value Y can be calculated as:
Y
1
2
R 2 L2
(7)
where R is the resistance, L the inductance and the angular frequency. The R value is
measured at the winding ends (R=0.3 ). The L value (5.5 mH, standard deviation 0.7 mH)
is estimated from Eq.(7) as a mean value of all the inductances calculated from each
measured Y values. The comparison of the calculated trend of Y (solid line) with the
measured Y in Figure 5 shows a good agreement. The simple RL model can be therefore
used to calculate the value of Y also outside the measured points. Substituting Y which
includes the information about the electrical frequency, I=Y()V , in Eq.(6):
fa
NV 2 air S '
2
2 R 2 2 L2 lair
FA cos 2t 2
cos 2t 2
(8)
86

9
60 Hz
Current, Ampl I [A]
8
7
75 Hz
87 Hz
100 Hz
125 Hz
4
3
150 Hz
175 Hz
200 Hz
1
0
0
12
15
18
21
Voltage, Ampl V [V]
Fig. 4. Measured output current amplitude vs. input voltage amplitude for different input
voltage frequency.
0,6
Experimental
Y average [A/V]
0,5
RL Model
0,4
0,3
0,2
0,1
0
50
100
freq [Hz]
150
200
Fig. 5. Comparison of admittance values Y derived from experiments and from RL model.
Equation (8) shows the relationship between the voltage amplitude V, the force amplitude
FA and the input electrical frequency ():
V2
2
2 Fa R 2 2 L2 lair
2
air N S '
(9)
Since R2<< 2L2 (R=0.3 , L = 5.5 mH, Berruti et al., 2010, 2011) the term R2 can be neglected.
Considering also that =2fel, Eq. (9) can be rewritten in the two forms:
4 2 F L2 l 2
A
air
f
V
el
air N 2 S '
4 2 f 2 L2 l 2
el
air
FA
V
2
N
S
'
air
(10)
The two relationships in Eq. (10) put in evidence that for a given value of the excitation force
amplitude FA, the corresponding value of the voltage V feeding the EM changes

87
proportionally with fel. In the same way, for a given frequency value fel, the voltage
amplitude value V changes proportionally with FA . These two linear trends of Equation
(10) (V vs. fel and V vs. FA ) will be confirmed by the experimental tests as shown in the
following sections. The formulas presented in this paragraph are useful to understand the
dependence of the output parameter of the exciter (force amplitude FA varying harmonically
with a given frequency fm) from the input parameter (the voltage amplitude V varying
harmonically with a given frequency fel) and to decide the dimension of the EM according to
the maximum force that is generated, but are not sufficient to determine the exact numerical
values since some parameters like the inductance L, the air gap lair or the active cross section
in front of the air gap S are not known with precision. For this reason a direct calibration of
the EMs is required as described in the following paragraph.
2.4 Calibration of the EMs
The EMs used in the multiple excitation system are nominally identical, however, in order
to check the EMs manufacture quality, each EM unit is calibrated on a calibration bench
shown in Figure 6 a). A force transducer (1), carrying a ferromagnetic anchor (2), is
connected to an inertial mass (3). The anchor has the same local shape and size of the disk
blade that is tested. The part of the inertial mass close to the anchor is made of aluminum (4)
in order to not interfere with the magnetic flux. The EM faces the anchor with a given gap (2
mm). A control software developed in Labview is here used to adjust the harmonic input
voltage of the EM amplifier in order to obtain the desired force amplitude (with a tolerance
of 1%) measured by the force transducer connected to the anchor.
(1)
(3)
Voltage, Ampl. V [V]
(2)
(4)
F A = 5N
1,8
1,6
1,4
1,2
1
0,8
0,6
60
70
80
90
100
110
120
130
140
f el [Hz]
(a)
(b)
Fig. 6. a) The EM on the calibration bench; b) The 24 EMs calibration curves.

In Figure 6 b) the calibration curves defined as the input voltage amplitude (V) vs. the input
electrical frequency (fel) are shown for the 24 electromagnets that are used later in the
application by keeping the force amplitude FA = 5N constant on the anchor. The
repeatability error on the measured voltage values obtained by repositioning the same EM
on the rig is 2% maximum. Despite the good calibration repeatability and the nominal equal
shape, the calibration curves of the 24 EMs are different from each other. The highest
88
voltage difference for the same electrical frequency is 13% with respect to the mean value .
The difference of the calibration curves is associated to the EM manufacture process. The
linear trend of the calibration curves in Figure 6 b) confirms the validity of the relationship
of Eq.(10) where the amplitude voltage V is proportional to the electrical frequency fel for a
given force amplitude FA.
2.5 The control of the excitation force amplitude
The control of the excitation force amplitude is obtained by means of a device, called Force
Measuring ElectroMagnet (FMEM) which allows the measurement of the force exerted by
one EM (Firrone & Berruti, 2011). The device is shown in Figure 7 a). The FMEM includes an
EM nominally equal to the other EMs which is mounted on a particular support carrying a
piezoelectric force transducer (1). The transducer measures the horizontal component (FM)
of the force (FA) produced by the EM and applied to the blade. The reaction of the vertical
component of FA as well as the mass of the EM is supported by the vertical leaf spring at the
base of the FMEM. By knowing the relationship occurring between FA and FM at different
frequencies, it is possible to control the force FA applied to the blade by controlling the force
FM measured by the force transducer.
FA
force transducer (FA)
anchor
(1)
FM
(a)
force transducer (FM)

(b)
Fig. 7. a) The force measuring electromagnet (FMEM); b) the FMEM on the calibration bench.
The FMEM needs a special calibration process in order to establish the relationship between
the force FA and the force FM measured by the force transducer (1) at different frequencies.
The FMEM is therefore calibrated on the same calibration bench used for the other EMs as
shown in the picture of Figure 7 b).
The FMEM faces the anchor with a given gap (2 mm). Once the harmonic force amplitude FA
is controlled by the software to a given value, the force amplitude FM is measured by the
FMEM transducer and recorded. The plot FM/FA vs. fel is the calibration curve shown in
Figure 8.

89
5,0
4,5
FM/FA [N/N]
4,0
3,5
3,0
2,5
2,0
1,5
1,0
0,5
0,0
50
60
70
80
90
100
110
120
130
140
150
fel [Hz]
Fig. 8. The FMEM calibration curve.

The curve of Figure 8 shows that the FMEM has a natural frequency within the range of
interest. A hammer test confirmed the presence of a natural frequency of the FMEM at 206
Hz (which corresponds to fel = 103 Hz). In order to avoid the resonance condition of the
FMEM that can lead to a poor measurement repeatability, the FMEM will be employed only
in the ranges of frequency outside the resonance (fel = 5075 Hz and fel = 140150 Hz).
2.5.1 Reliability of the FMEM on the calibration bench
In order to estimate the reliability of the force amplitude value imposed and controlled by
means of the FMEM, different calibration tests are performed on the calibration bench of
Figure 7 b). The tests are listed below.
1.
2.
Check of the force control. Calibration curves are measured for different nominal values
of the force on the anchor (FAnom =1, 2, 3 , 4, 5, 6 N) with the same air gap of 2 mm. As a
result, for each FAnom a calibration curve like that of Figure 8 is obtained. An inverse
calibration is then performed on the same bench in order to simulate the same control
force procedure that will be used on site under the bladed disk. In this case the FMEM
transducer that measures FM is controlled by the software. The set-point controlled by
the software at different frequencies is FM = (FM/FA)FA,nom where FA,nom is the nominal
force value and (FM/FA) is the calibration curve. When the FMEM transducer reaches
the target value FM, the actual force FA,act exerted on the anchor is measured. The
difference between the measured value FA,act and the nominal value FA,nom proved to be
always within 1% of FA,nom for all the tested frequencies confirming the quality of the
force control procedure.
Check of the FMEM repeatability. FA,nom is set to 5 N, and the air gap to 2 mm. Different
calibration curves FM/FA are obtained after repositioning the FMEM on the calibration
bench. The calibration curve FM/FA plotted in Figure 8 is the average result (FM/FA)AVG.
The inverse calibration is then performed by controlling the force value FM measured
by the transducer on the FMEM support. The set-point controlled by the software at
90
different frequencies is calculated as FM=(FM/FA)AVGFA,nom. When the FMEM transducer

reaches the target value FM, the actual force FA,act exerted on the anchor is measured.
The error of FA,act with respect to the nominal force value of 5N is 3% in the worst case.
It can then be concluded that after this calibration procedure it is possible to obtain the
desired value of force amplitude FA on the anchor with a precision inside 3% by controlling
the force FM on the FMEM support.
3. The test rig

The excitation system is designed to study the complex dynamics of bladed disks for
turbomachinery applications. In detail, the aim is to collect a database of measurements on
real or dummy bladed disks in order to highlight the effect of mistuning and friction
damping. The experimental results will be compared with results from numerical
simulation tools. The multiple excitation system described above is set up on the test rig
Octopus designed to test bladed disks carrying underplatform dampers (UPDs) between the
blade platforms (Berruti, 2010). The test rig shown in Figure 9 a) can be divided into two
parts, i.e., the central support carrying the real or dummy disk and an outer ring supporting
the arm structures used to couple the bladed disk with the UPDs. The required number of
arms depends on the number of blades of the specific disk. In this case the test rig is
prepared to test a dummy integral bladed disk (blisk) with 24 blades, therefore a number of
Nb=24 arms must be provided. Nevertheless, the outer ring can be used in part or can be
extended depending on the specific disk and its number of blades. Figure 9 a) shows the test
rig with the dummy blisk (1) already mounted on it.
(1)
(5)
(2)
(3)
(4)
(a)
Fig. 9. a) The test rig Octopus; b) detail of the UPD.
(b)

91
The blisk is constrained on the central fixture (2), which is basically a big inertial mass
(about 400kg) with two coaxial cylinders mounted on a circular plate laid on a rubber sheet.
The arm structures (3) support one pulley each and are equally spaced around the fixture.
Two wires attached to each UPD (Figure 9 b)) pass over the arm supporting the pulley and
are connected to a loading plate which carries dead weights simulating the centrifugal force
on the UPD. The pulley has low friction ball bearing and is mounted on a pin tightened on
the circular outer ring (4) (about 1.5m diameter) centered with respect to the disk fixture.
Each arm carrying the pulley can rotate around the pin in order to align with precision the
wires along the radial direction of the centrifugal force acting on the UPDs. The outer ring is
fixed to the floor and is not directly connected to the central fixture in order to minimize the
transmission of vibrations from the fixture to the ring. The dummy blisk is centered on the
fixture by means of a hub integral with the blisk which is tightened with seven screws using
a counter torque wrench in order to save the cyclic symmetry of the structure as much as
possible. The rig was designed to hold 30kg (294N) for each damper which is a realistic
centrifugal force on a real UPD having a mass of about 5g.
3.1 The dummy disk
The dummy disk was manufactured in order to fulfil two requirements: i) to test the
electromagnetic excitation system and the control software of the rotating force and ii) to
study the dynamic behaviour of a cyclic structure with underplatform dampers (UPDs)
under a specific engine order EO excitation in order to validate the numerical model of the
blisk with friction contacts between blade platforms and UPDs. The FE model of the blisk is
shown in Figure 10 a). The mode shapes of a bladed disk are characterized by a number of
Nodal Diameter (ND), i.e., lines passing for the center of the bladed disk where the modal
displacement is null. As an example, Figures 10 b)-d) show modes at ND=0,2,4. In particular,
only one mode is associated to ND=0 (the so called umbrella mode) and to ND=Nb/2 while
two modes share the same ND when 0<ND<Nb/2. In this case the natural frequencies of the
two modes are equal and the two modes are orthogonal (their scalar product is null). This
information can be resumed in a plot where the natural frequencies are plotted with respect to
ND (Figure 11). Lines connecting the natural frequencies define the modal families of the blisk
where modes share a type of motion of the sector. In this case, for instance, the first family
(lower frequencies) is characterized by the first bending mode (1F) of the blade, while the
second family (higher frequencies) is characterized by the second bending mode (2F,
edgewise) of the blade. Label free refers to the modal analysis of the blisk without damper
while stick refers to the modal analysis of the blisk coupled to the cylindrical underplatform
damper. The stick condition is simulated by constraining the UPDs to the blade platforms
through hinge constraints along the line contacts between the cylindrical surface of the UPD
and the flat surface of each platform. A cylindrical cross section of the damper is chosen
(Figure 11 b)) in order to guarantee the contact between UPDs and blade platforms the
simplest as possible. The design aimed at obtaining the first bending mode well isolate from
the second bending mode in order to obtain a simple kinematics of the blade platforms in
contact with the UPDs. Moreover, a 45 stagger angle is chosen for the blades since both 1F
and 2F modal families must have components along the out-of-plane direction since the
motion is detected by a Laser Doppler Vibrometer (LDV) which measures the orthogonal
component of the velocity with respect to the plane of the blisk.
92
(b) ND=0
(a)
(c) ND=2
(d) ND=4
Fig. 10. a) FE model of the blisk; b) umbrella mode (ND=0); c) ND=2 orthogonal modes;
d) ND=4 orthogonal modes.
1000
900
2F stick
800
2F free
freq [Hz]
700
600
500
1F stick
400
300
1F free
200
line
contact
100
0
0
10
11
12
Nodal Diameter ND
(a)
(b)
Fig. 11. a) First two bending families, without UPDs (free condition) and with UPDs constrained
between the platforms (stick condition); b) line contacts between UPD and platforms.
It can be also noticed that in fully stick condition the highest frequency of the 1F family is
under the threshold of 500 Hz. This requirement is determined by the EM currently used; it

93
is possible to see from Figure 5, in fact, that the value of admittance drops for values of
electrical frequency higher than fel=200Hz (i.e. 400Hz mechanical frequency). This means
that a too high voltages V may be required to excite the blisk for frequencies higher than
fm=400Hz if the required force amplitude FA is of the order of magnitude of 10N. For this
reason the study will be restricted to the first modal family (1F).
4. The excitation system under the disk

In the final configuration, each EM is positioned under the corresponding blade with an air
gap of 2mm. However, before placing all the EMs, different tests are performed by
mounting only the FMEM under the disk on a housing purposely machined in the
aluminum base (Figure 12). In this case the FMEM is facing the blade instead of the
ferromagnetic anchor of the calibration bench.
FA
FM
Fig. 12. The FMEM mounted under the disk.

Different tests are performed with the software control in order to obtain the desired FA,nom on
the blade. The control is performed by the measurement of FM (Figure 12). The software adjusts
the voltage value until the desired value of force amplitude FM = (FM/FA)AVGFA,nom is reached by
the transducer. This force control is repeated for different FA,nom and for different frequencies fel.
In each case the feeding input voltage amplitude V is recorded. In Figure 13 the voltage
amplitude V is plotted versus the input electrical frequency fel for different FA,nom values.
3.0
6N
Voltage amplitude V [V]
2.5
5N
4N
3N
2.0
1.5
2N
1N
1.0
0.5
0.0
50
60
70
80
90
100 110 120 130 140 150 160

f el [Hz]
Fig. 13. Feeding voltage amplitude V vs. electrical frequency fel for different excitation force
amplitude FA, nom.
94
It can be seen that the trend V vs. fel is still linear for a given value of the force amplitude
exciting the blade (compare with Figure 6 b) where the same measurement was performed
on the calibration bench at FA=5N). These tests confirm again the first theoretical
relationship of Eq.(10). Moreover, the plots V vs. FAnom (see Figure 14) are also linear for
different electrical frequency values fel confirming the second theoretical relationship of
Eq.(10).
3.0
150Hz
145Hz
140 Hz
2.5
2.0
1.5
75 Hz
70 Hz
1.0
50 Hz
0.5
0.0
1.00
1.25
1.50
1.75
2.00
2.25
2.50
2.75
Fig. 14. Feeding voltage amplitude V vs. the square root of the excitation force amplitude FA
for different electrical frequencies fel.
The linear trends V vs. fel and V vs. FAnom allow to calculate by interpolation the needed
input voltage also for force amplitudes or for frequency values different from those
employed during the calibration of the FMEM. It is interesting to compare the trends V vs. fel
obtained on the calibration bench (dashed lines) and under the disk (solid curves) for two
different values of FA,nom (1 N and 5 N) as shown in Figure 15.
2.5
5 N-disk
2.0
5 N- c. bench
1.5
1 N-disk
1.0
1 N- c. bench
0.5
0.0
50
60
70
80
90
100 110 120 130 140 150 160

f el [Hz]
Fig. 15. Comparison of the feeding voltage amplitude V vs. electrical frequency fel when the
FMEM is on the calibration bench (dashed line) and under the blisk (solid lines).

95
It can be noted that higher input voltages are requested by the EM under the disk with
respect to the case of the calibration bench in order to obtain the same value of FA,nom. The
difference is higher for higher frequencies. The difference of the input voltages is due
probably to the higher magnetic field loss in the blade connected to the disk compared to the
case of the single magnetic anchor on the calibration bench. This voltage difference between
the blisk and the calibration bench highlights the importance of having a device like the
FMEM which measures the force directly under the disk. The bench calibration of each EM
is not enough since the calibration curve does not correspond to the desired FA,nom under the
disk.
4.1 Verification of the blisk linearity
The main objective of the tests performed on the dummy blisk is the measurement of the
amount of nonlinear friction damping introduced by the UPD. For this reason it must be
verified that another source of nonlinear damping is not introduced by the test rig itself and
in particular by the contact interface between the blisk and the central hub. The verification
of the blisk linearity requires the measurement of the vibration response of the blades. The
measurement is performed by means of the scanning LDV which detects the blade velocity
along the disk axial direction. One measurement point is chosen for each blade in the same
position.
The FMEM is positioned under the bladed disk (as shown in Figure 12) and a harmonic
force is applied at different amplitude FA gradually increasing in order to verify that the
blade response measured by the LDV increases proportionally.
The FMEM is fed with an alternate voltage at an electrical frequency fel, the mechanical
excitation frequency value fm (that is two times fel) is chosen as far as possible from both the
disk and FMEM natural frequencies in order to avoid calibration complications due to disk
or FMEM dynamics. The disk and the FMEM Frequency Response Functions (FRF)
measured by impact testing are shown in Figure 16.
v/F [mm/s/N]
FMEM
nd 1
140
100
nd 3 disk
nd 2
120
140
150
160
300
170
180
200
220
240
260
280
300
f el [Hz]
Fig. 16. FRF for the FMEM and for the bladed disk (y-axis is not to scale).
96
The verification of the disk linearity is performed at two different mechanical excitation
frequencies (fm=150Hz and 300Hz that is, fel=75Hz and 150Hz) which are far from both the disk
and FMEM natural frequencies. Once the force control loop has reached the target value FA
at fm=2fel, a trigger signal activates the scanning LDV to measure the velocity amplitude of
the 24 blades. The blade velocity amplitudes vs. the blade number (velocity profile) can then
be plotted.
In Figure 17 and Figure 18 the velocity profiles obtained for different excitation force
FA,nom amplitude values are plotted in the case of fel=75Hz and in the case of fel=150Hz. It is
possible to see that the velocity profile (see red dots for clarity) increases linearly with the
force amplitude FA. The comparison of the ratios of the different force amplitudes applied to
the bladed disk and the corresponding average ratios of the velocity values marked with red
0.040
0.035
6N
vel [m/s]
0.030
5N
0.025
0.020
3N
0.015
2N
0.010
1N
0.005
0.000
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
Blade order
Fig. 17. Blade absolute velocity values for different excitation force amplitude, electrical
frequency fel=75 Hz.
0.09
0.08
6N
0.07
vel [m/s]
0.06
0.05
5N
0.04
0.03
3N
0.02
0.01
1N
0
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
Blade order
Fig. 18. Blade absolute velocity values for different excitation force amplitude, electrical
frequency fel=150 Hz.

97
dots shows a difference within 2% in the most cases, i.e., within the force value repeatability
allowed by the FMEM. As a consequence, the dynamic response of the bladed disk without
UPDs can then be considered linear.
4.2 The method for tuning the EMs
In order to obtain the same amplitude of the excitation force for each EMs a tuning process
of the magnetic forces is necessary in order to adjust the input voltage for each exciter. The
tuning involves an iterative process exciting directly one blade each time and measuring the
indirectly excited blade responses. Two system configurations are needed for this
calibration:
-
configuration 1: only the FMEM is needed under blade 1;

configuration 2: the 24 EM units are positioned under the bladed disk (after removal of
the FMEM).
In the configuration 1 the FMEM is fed with an alternate voltage at the electrical frequency
fel. The control software adjusts the voltage amplitude in order to reach the target force value
FA,nom. As in the case of the blisk linearity verification the scanning LDV measures the
velocity amplitude of the 24 blades as a reference velocity profile (blade velocity
amplitudes vs. the blade number) which is recorded together with the voltage amplitude V
and the electrical frequency fel. The mechanical excitation frequencies fm chosen for the
tuning process (140, 150, 170, 300Hz) are highlighted in Figure 16. The chosen mechanical
excitation frequency values are far from the disk and from the FMEM natural frequencies. In
order to excite the blisk with these excitation frequencies in configuration 1 the FMEM must
be supplied with an input voltage with half the value of the electrical frequency (fel = 70, 75,
85, 150Hz). The velocity profiles are plotted in Figure 19 and Figure 20 for the four values of
fel feeding the FMEM.
Fig. 19. Velocity amplitude profiles when the FMEM is switched on, force controlled to a
nominal force amplitude FA,nom=5N, electrical frequency 70Hz and 75Hz.
98
Fig. 20. Velocity amplitude profiles when the FMEM is switched on, force controlled to a
nominal force amplitude FA,nom=5N, electrical frequency 85Hz and 150Hz.
The velocity profiles of Figure 19 and Figure 20 are kept as reference profiles for the tuning
process since they are obtained with a known force amplitude of 5N (measured on site) on
the excited blade. Repeatability tests performed after repositioning the FMEM under the
disk proved that the repeatability error of the profiles of figure 19 and 20 is lower than 3%.
The system is then set up according to the configuration 2: the FMEM is removed and the 24
EMs are positioned under the blisk. Only one EM per time is switched on. Figure 21 shows the
24 velocity amplitude profiles at the excitation electrical frequency fel=75Hz when the
calibration curves of Figure 6 b) are used to set the initial voltage amplitude V for the 24 EMs.
Fig. 21. The 24 velocity profiles, each of them measured with one EM switched on, at
fel=75Hz and nominal force amplitude FA,nom =5N, before tuning.

99
Each velocity profile is measured when only one EM per time excites the structure. The
velocity amplitude of the directly excited blade is plotted always in the first position (Blade
Order 1). On the same figure the velocity reference profile (bold line) obtained with the
FMEM at fel=75Hz in the previous step (configuration 1) is plotted. The curves of Figure 21
point out two phenomena:
-
there are relevant differences among the velocity profiles when the working EM exciter
changes (the standard deviation of the force values compared to the mean value of the
force is about 16%),
the velocity profile obtained by FMEM is inside the profile dispersion of the other 24
EMs.
Considering the dynamics of the blisk linear as proved before and neglecting the presence of
mistuning phenomena since the excitation frequency is far from resonance conditions, the
differences among the profiles in Figure 22 must be referred to different excitation force
amplitudes. These differences are not related to the manufacture of the EM units since they
are taken into account in the calibration curves of Figure 6b). Other causes producing
different velocity profiles are probably due to i) different positions of the EMs under the
disk blades that cause different air gaps, ii) different shape and material properties of the
blade compared to the anchor used for calibration. These differences among the EMs
highlight the importance of having a FMEM measuring the real force amplitude on site (i.e.
under the disk). The 24 EMs can than be tuned to the reference FMEM. The tuning process
aims at adjusting the voltage on the 24 EMs until they produce the same velocity profiles of
the blisk as the one measured with the FMEM. A method is here proposed starting from the
velocity profiles shown in Figure 21. The voltages are adjusted in order to obtain 24 velocity
profiles overlapping the FMEM reference profile as shown in Figure 22.
Fig. 22. The 24 velocity profiles, each of them measured with one EM switched on, at
fel=75Hz and nominal force amplitude FA,nom=5N, after tuning.
In detail the method works as follows. Let X be and X bFMEM be the measured velocity
amplitudes of the blade corresponding to the blade order b when respectively the EM e and
100
the FMEM is switched on. Considering the system linear and assuming the FMEM as a
reference, the ratio X be / X bFMEM is proportional to the excitation force ratio FA,e /FA,FMEM
where FA,e is the actual excitation force amplitude on the blade facing the EM e and FA,FMEM is
the nominal value of the force (in this case 5N).
For each response of Figure 21 (corresponding to the e-th EM working) the correction factor
ke is calculated as:
X be
ke mean
, b 1,..., N b
XbFMEM
(11)
In order to get the same force amplitude for each EM unit the force is corrected as FA,e*=FA,e/ke
by adjusting the input voltage as Ve * Ve / ke since the voltage is proportional to the square
root of the force (Eq.10). The velocity profiles plotted in Figure 22 are measured after two
iterations of the tuning process at fel=75Hz. The dispersion of the force amplitude
corresponding to the different velocity profiles with respect to the nominal value of 5N are less
than 2%. It can be concluded that after this tuning process each of the 24 EMs produces on the
excited blade the same force amplitude of 5N with a dispersion of 2%. The same calibration
procedure is repeated for the other electrical frequency values (fel =70, 85, 150Hz) in order to
characterize the linear trend of the voltage as a function of frequency when the same force
FM=(FM/FA)AVGFA,nom is required, as it was done in the calibration bench (Figure 13) with the
FMEM. The voltage values at frequencies not directly tuned can be found by linear
interpolation since the experimental evidence verifies the linear trends of Eq.(10).
It must be noted that the tuning method here proposed is very efficient since it avoids the
calibration of one EM at a time on a separate bench which is time consuming. In fact, the
tuning process does not require a particular starting voltage values of the EMs, in particular
it is not necessary to start with voltage values coming from the EMs calibration (Figure 6b).
5. Example of test results

The test campaign is performed for the modes at ND=2, 4 without dampers (free, linear
configuration) and with UPDs (nonlinear configuration). Each mode is excited with a
travelling wave with an engine order EO=ND. First, two forced responses at two different
excitation frequencies are shown for the blisk without damper (free system), second the
FRFs are shown in terms of mobility, for the two ND at 7 different values of force
amplitude FAnom =0.1, 0.2, 0.3, 0.6, 1, 2.5, 5N when UPDs are mounted. Each FRF is the
envelope of the maximum value of the FRF of the 24 blades at each frequency. The test
results are selected since they show the capability of the rig to detect both the effect of the
friction damping on the blisk dynamics and the presence of mistuning.
5.1 Forced response of the blisk without UPDs under the travelling force
A rotating force is generated in order to excite the ND=2 rotating mode shape. For this reason the
voltage signal is generated for each EM in order to obtain a force pattern on the EMs equal to:
2
f a ,n FAnom cos f m 2 t n 1
EO
Nb
n 1,...., N b
f m 2 f el
(12)

101
where EO is chosen equal to 2 in order to excite the ND=2 rotating mode shape. The force
amplitude FAnom is in this case very low since the system is linear (FAnom=0.2N). The response is
measured by measuring one point per blade by means of the scanning LDV in order to
measure the rotating mode shape along the hoop direction of the blisk. The voltage signal of
the first EM is used as a trigger in order to sinchronize the 24 signals measured by the LDV in
24 different time instants. The blisk is excited at fm=134Hz which is far from the natural
frequency f1F=131Hz of the ND=2 rotating mode shape, first family. Figures 23 a)-d) show the
(a) t=0
(b) t=(1/3)T
(c) t=(2/3)T
(d) t=T
Fig. 23. Rotating Deformed Operative Shape, ND=2, EO=2, far from resonance.
102
Operative Deformed Shape (ODS) of the blisk for four equally spaced time istants within a
complete rotation of the travelling force T=EO(1/fm). It is possible to see how the two
orthogonal mode shapes as seen in the simulation of Figure 10 c) combines to generate a
rotating mode shape at ND=2 following the rotating force. Figure 23 d) shows all the steps
after one rotation is complete, the red line groups the maximum amplitude that each blade
reaches in order to prove that the difference in amplitude is acceptable. If the same test is
performed at a frequency which is closer to the natural frequency of the 1F mode, ND=2
(Figure 24 a)-d)), it is possible to see that the blades vibrate with different amplitudes
proving the existence of inherent mistuning due to the asymmetries produced by the
manufacture or the constraint. However, the mistuning can be considered small since it
only perturbs the ND=2 shape of the rotating mode which can be still identified and does
not localize the energy of the system to a small number of blades.
(a) t=0
(b) t=(1/3)T
(c) t=(2/3)T
(d)t=T
Fig. 24. Rotating Deformed Operative Shape, ND=2, EO=2, resonance condition.
5.2 The forced response with friction damping
The UPDs are positioned in their cavities and each of them is loaded with an equivalent
centrifugal load of about 50N (5kg of dead weights) (Firrone et al., 2011). In Figure 25 the free
(linear) and the nonlinear FRFs at ND=2 are plotted as the envelope of the maximum value of
the FRF of each single blades at each frequency. The FRFs show that the amount of damping
depends on the force amplitude value FAnom. When FAnom is low (FAnom=0.1N) the relative
displacement at the contacts is not large enough to provide a large amount of damping. At the
same time the UPDs act more as an additional constraint whose main effect is to stiffen the
structure as demonstrated by the increase of the frequency value of the peaks with respect to
the free response. As the excitation force amplitude increases (FAnom=0.2N, 0.3N), more

103
10
Free
mobility [m/s/N]
0.1 N
0.2 N
10
0.3 N
0.6 N
-1
10
2.5 N
1N
135
140
fm [Hz]
145
5N
-2
10
-3
10
120
125
130
150
155
Fig. 25. FRF of the bladed disk (free and with UPDs) for different amplitude values FA of the
excitation force. CF =50N, EO=ND=2.
damping is provided to the bladed disk, a general reduction of the vibration amplitudes of
the blades is observed and the distance of the peak frequency from the corresponding free
value decreases since large relative displacements at the contacts globally determine a less
stiff constraint for the blades. As the excitation becomes larger and larger (FA=1N, 2.5N, 5N)
the smooth peak is almost flattened while the frequency peak tends toward values lower
than that of the free response since the UPDs act as an additional mass of the blisk. These
results can be summed up as shown in Figure 26 in terms of the optimization curve where
the frequency and the amplitude values of the response peaks are plotted vs. the variable
parameter CF/FAnom. It is possible to see that a minimum response of the blisk amplitude
mobility exists corresponding to an optimum combination of damper mass (i.e., CF) and
amount of excitation (i.e., FAnom).
160
Opt. amplitude
Opt. frequency
Free response amplitude
140
50
100
150
200
250
CF/FA
Fig. 26. Optimization curves at ND=2.
300
350
400
450
120
500
Frequency [Hz]
mobility [m/s/N]
104
5.3 The forced response with mistuning introduced by the friction dampers
In order to show the effect of mistuning introduced by the UPD, the rotating mode shape
ND=4 is excited with a travelling force at EO=4. In Figure 27 two peaks are clearly visible
for low values of FAnom (FAnom= 0.1, 0.2, 0.3N) for fm=316Hz and 345Hz. The presence of two
peaks instead of one is an index of presence of mistuning since the lack of cyclic symmetry
causes the split of the natural frequency of the twin modes at ND=4. This can be verified by
plotting the Operative Deformed Shape (ODS) of each blade when the maximum absolute
value of the velocity during the blisk vibration is reached. The two stationary mode shapes
vibrate at resonant conditions for two different frequency values and the rotating response of
the blisk is no more obtained.
1
10
mobility [m/s/N]
Free
0.1 N
0.2 N
10
0.3 N
0.6 N
1N
-1
10
2.5 N
5N
-2
10
200
220
240
260
280 300
fm [Hz]
320
340
360
Fig. 27. FRF of the bladed disk (free and with UPDs) for different excitation force amplitude
values. CF =50N. EO=4.
0.4
0.3
imag(mobility)
0.2
0.1
0
-0.1
-0.2
-0.3
-0.4
10
15
blade number
20
Fig. 28. Maximum amplitude of the ODS (mobility) for the different blades at 316Hz (solid
line) and 345Hz (dotted line), FAnom =0.1N, EO=4.

105
In particular, the ODS for the two peak frequencies are plotted for the case FAnom= 0.1N in
Figure 28. It can be noted that each maximum amplitude of the ODS represents a mode
shape ND=4 (4 maxima are present). Moreover the two mode shapes of Figure 28 are
orthogonal. This consideration puts in evidence that the two peaks of Figure 27 are the
mistuned twin modes at ND=4.
Figure 27 suggests another important consideration. By increasing FAnom more damping is
provided in the system and the two mistuned peaks become flat and disappear confirming
that the right amount of damping can mitigate the mistuning phenomenon. The test rig
proved then to be able to put in evidence the interaction of friction damping and mistuning.
6. Conclusions
The design, calibration procedure and testing of a complete (hardware and software) noncontact excitation system is presented. The system aims at performing accurate
measurements of the dynamic behaviour of bladed disks in presence of mistuning and
nonlinearity due to friction contacts. The system yields accurate results and it is
characterized by three key features.
First, the purposely designed EMs generate force amplitudes that can be considered high for
a noncontact excitation system. Each EM can reach a force amplitude up to 10 N (up to
300Hz excitation frequency of the system) and 5 N (up to 600 Hz). The several features
affecting this force value are highlighted.
Second, thanks to a novel calibration method performed on site under the disk to iteratively
calibrate the force, differences of force amplitudes among the blades are highly reduced (2%
variation among force amplitudes). The method is efficient since it avoids the calibration of
each EM (one by one) separately which is time consuming and in any case it proved to be
inaccurate when the EMs are mounted under the blisk.
Third the value of the force amplitude exciting the blades is known with good accuracy (less
than 5% error with respect to the nominal value). This feature is essential for stepped sine,
force controlled tests to investigate the nonlinearity due to friction contacts. A device
carrying one of the EMs and instrumented with a force transducer is designed, constructed
and calibrated for this purpose. The device called FMEM (Force Measurement
ElectroMagnet) is able to measure the force directly on site under the disk and is used as a
reference for the other EMs during calibration.
The travelling excitation system is applied to the test rig Octopus where an integral
bladed disk carries underplatform dampers (UPDs) which introduce nonlinearities due to
friction.
The test campaign provided an example of results which proved that the system is capable
of an engine order type excitation at different controlled force amplitudes. The developed
noncontact travelling excitation system together with a noncontact laser Doppler vibrometer
measurement of the response highlight 1) the effectiveness of the UPDs in reducing the
blades vibration amplitude at different EOs, 2) the interaction of the damping provided by
the UPDs and the introduction of mistuning.
106
7. Acknowledgment
The work described in this paper has been developed within the PRIN and CORALE
projects (National Interest Research Projects).
8. References
Avalos, J. & Mignolet, M. P. (2008) On damping entire bladed disks through dampers on
only a few blades, GT-2008-51446, Proceedings of the of ASME Turbo Expo, June 9-13,
Berlin, Germany.
Beirov, B., Khhorn, A., Schrape, S. (2008). A discrete model to consider the influence of the
air flow on blade vibrations of an intergral blisk compressor rotor, Proceedings of the
of ASME Turbo Expo, June 9-13, Berlin, Germany.
Beirov, B., Kuhhorn, A., Nipkau, J. (2009). On the influence of strain gauge
instrumentation on blade vibrations of integral blisk compressor rotors applying
a discrete model, GT2009-59207, Proceedings of the of ASME Turbo, June 8-12,
Orlando, Florida.
Berruti, T. (2010). A test rig for the investigation of the dynamic response of a bladed disk
with underplatform dampers, Mechanics Research Communications, vol 37,
doi:10.1016/j.mechrescom.2010.07.008, pp 581-583.
Berruti, T., Firrone, C.M., Gola, M.M. (2010). A test rig for non-contact travelling wave
excitation of a bladed disk with underplatform dampers, GT2010-22879, Proceedings
of the of ASME Turbo, Glasgow, UK., pp.1-9.
Berruti, T., Firrone, C.M., Gola, M.M. (2011). A test rig for non-contact travelling wave
excitation of a bladed disk with underplatform dampers. Journal of Engineering for
Gas Turbines and Power, Vol 133, pp. 032502-1-8, ISSN 0742-4795, DOI:
10.1115/1.4002100.
Castanier, M. P. & Pierre, C. (2006). Modeling and analysis of mistuned bladed disk
vibration : Status and emerging directions. Journal of Propulsion and Power, Vol. 22,
Issue 2, pp.384-396, DOI: 10.2514/1.16345. ISSN 0748-4658.
Firrone, C.M., Berruti, T. (2011). An electromagnetic system for the non-contact excitation of
bladed disks, Experimental Mechanics, DOI: 10.1007/s11340-011-9504-1.
Firrone, C.M, Berruti, T., Gola, M.M. (2011). On force control of an engine order type
excitation applied to a bladed disk with underplatform dampers, IDETC201147255, Proceedings of the International Design Engineering Technical Conferences
(IDETC), Washington, pp. 1-10.
Chang, J. Y. & Wickert, J. A. (2002). Measurement and analysis of modulated doublet mode
response in mock bladed disks", Journal of Sound and Vibration, Vol.250, Issue 3, pp379-400, doi:10.1006/jsvi.2001.3942.
Gilbert, C., Kharyton, V., Thouverez, F., Jean, P. (2010). On forced response of a rotating
integrally bladed disk: predictions and experiments, GT2010-23610, Proceedings of
the of ASME Turbo Expo, June 14-18, Glasgow, UK.
Gotting, F., Sextro, W., Panning, L., Popp, K. (2004). Systematic mistuning of bladed disk
assemblies with friction contacts, GT2004-53310, Proceedings of the of ASME Turbo
Expo, June 14-17, Vienna, Austria.

107
Heath, S. & Imregun, M. (1998). A survey of Blade Tip-Timing Measurement Techniques for
turbomachinery Vibration", Journal of Engineering for Gas Turbines and Power,
Vol.120, Issue 4, pp.784-791, doi:10.1115/1.2818468.
Heinz, C., Schatz, M., Casey, M. V., Stuer, H. (2010). Experimental and analytical
investigations of a low pressure model turbine during forced response excitation,
GT2010-22146, Proceedings of the of ASME Turbo, June 14-18, Glasgow, UK.
Jones, K.W. & Cross, C.J. (2003). Travelling Wave Excitation System for Bladed Disks",
Journal of Propulsion and Power, Vol. 19, pp.135-141, ISSN 0748-4658.
Judge, J., Ceccio, S. L., Pierre, C. (2003). Traveling-wave excitation and optical measurement
techniques for non-contact investigation of bladed disk dynamics, The Shock and
vibration digest, Vol. 35, Issue 3, pp.183-910, ISSN 1741-3184.
Kenyon, J.A. & Griffin, J.H. (2003). Experimental Demonstration of Maximum Mistuned
Bladed Disk Forced Response. Journal of Turbomachinery, Vol. 125, Issue 4, pp. 673681. doi:10.1115/1.1624847.
Kruse, M. J. & Pierre, C. (1997). An Experimental Investigation of Vibration Localization in
Bladed Disks, Part I: Free Response", Proceedings of the 42nd ASME Gas Turbine &
Aeroengine Congress, User's Symposium & Exposition, Orlando, Florida.
Kruse, M. J. & Pierre, C. (1997). An Experimental Investigation of Vibration Localization in
Bladed Disks, Part II: Forced Response", Proceedings of the 42nd ASME Gas Turbine &
Aeroengine Congress, User's Symposium & Exposition, Orlando, Florida.
Petrov, E., Hennings, H., Di Mare, L., Elliott, R. (2010). Forced response of mistuned bladed
discs in gas flow: a comparative study of predictions and full-scale experimental
results. Journal of Engineering for Gas Turbines and Power, Vol. 132, Issue 5, 052504,
doi:10.1115/1.3205031. ISSN: 0742-4795.
Pierre, C., Judge, J., Ceccio, S. L., Castanier, M. P. (2002). Experimental Investigation of
the Effects of Random and Intentional Mistuning on the Vibration of Bladed
Disks, Proceedings of the 7th National High Cycle Fatigue Conference, Palm Beach,
Florida.
Prchlik, L., Misek, T., Kubin, Z., Duchek, K.. (2009). The measurement of dynamic vibration
modes and frequencies of a large LP bladed disc, GT2009-60002, Proceedings of the of
ASME Turbo Expo, June 812, Orlando, Florida, USA.
Rice, T., Bell, D., Singh, G. (2007). Identification of the stability margin between safe
operation and the onset of blade flutter, GT2007-27462, Proceedings of the of ASME
Turbo Expo, May 14-17, Montreal, Canada.
Sever, I., A. (2004). Experimental validation of turbomachinery blade vibration predictions,
PhD Thesis, Department of Mechanical Engineering, London.
Strehlau, U. & Kuhhorn, A. (2010). Experimental and numerical investigations of HPC blisks
with a focus on travelling waves, GT2010-22463, Proceedings of the of ASME Turbo,
June 14-18, Glasgow, UK.
Szwedowicz, J., Gilbert, C., Sommer, T. P., Kellerer, R. (2006). Numerical and experimental
damping assessment of a thinwalled friction damper in the rotating set-up with
high pressure turbine blades", Proceedings of the of ASME Turbo Expo, May 8-11,
Barcelona, Spain.
108
Szwedowicz, J., Secall-Wimmel, T., Dnck-Kerst,, P., Sonnenschein, A., Regnery, D.,
Westfahl, M. (2007). Scaling concept for axial turbine stages with loosely assembled
friction bolts: the linear dynamic assessment, part I, GT2007-27502, Proceedings of the
of ASME Turbo Expo, May 14-17, Montreal, Canada.
Thomas, D. L. (1979). Dynamics of rotationally periodic structures. International Journal for
numerical methods in engineering, Vol.14, pp. 81-102.
6
Study on Wireless Torque Measurement
Using SAW Sensors
Chih-Jer Lin*, Chii Ruey Lin, Shen-Kai Yu, Guo-Xing Liu,
Chih-Wei Hung and Hai-Pin Lin
National Taipei University of Technology,
Taiwan, R.O.C.
1. Introduction
An acoustic wave is a vibration in an elastic medium that propagates in space and time, thus
transferring the energy supplied by an excitation source along the medium in the form of
oscillation or vibration. Acoustic wave propagation entails elastic deformation of the
medium along the propagation axis or in other axes as well. In contrast to electromagnetic
waves, acoustic waves do require a medium to propagate, and their propagation speeds
depend on the mechanical properties of the wave-supporting material. Virtually any
material is capable of supporting acoustic wave propagation, including silicon.
Nevertheless, the piezoelectric properties of certain materials facilitate the wave
propagation. Thus, for improving the electromechanical energy conversion, piezoelectric
materials are usually chosen as the acoustic layer of many acoustic-wave resonators. Also
known as sound speed, the acoustic-wave phase velocities are several times slower than
those of the electromagnetic wave traveling in the same medium(Auld,B. A.,1990). There
exist two types of acoustic waves, surface acoustic waves (SAW) and bulk acoustic waves
(BAW). A surface acoustic wave is a type of mechanical wave motion which travels along
the surface of a solid material. As shown in Fig. 1, the surface particles of an isotropic solid
move in ellipses in planes normal to the surface and parallel to the direction of the wave
propagation. The particle displacement is significant at a depth of about one wavelength.
This motion decreases at the surface at thinner depths and increases at greater depths. As
the size of the ellipses is smaller, its eccentricity changes for particles in the deeper material.
Surface acoustic waves were discovered in 1885 by Lord Rayleigh (Rayleigh,1885); therefore,
they are often named after him: Rayleigh waves. Rayleigh showed that SAWs could explain
one component of the seismic signal due to an earthquake, a phenomenon not previously
understood. The velocity of acoustic waves is typically 3000 m/s, which is much lower than
the velocity of the electromagnetic waves.
On the other hand, bulk acoustic waves are longitudinal, shear-mode, or combination of
both. Longitudinal waves travel through the medium parallel to the same axis of the
oscillations or vibrations of the particles in the medium; that is, in the same or opposite
direction as the motion of the wave as shown in Fig. 2. Longitudinal mode waves are
confined in a resonant cavity, thus displaying a particular standing-wave pattern. All other
* Corresponding
Author
110
Fig. 1. Rayleigh wave propagation. (Auld ,1990).
Fig. 2. The longitudinal-mode waves (Auld ,1990).

wavelengths experience destructive interference and are suppressed. While longitudinal
modes have a pattern with their nodes located axially along the length of the cavity,
transverse modes, with nodes located perpendicular to the axis of the cavity, may also exist.
A transverse or shear-mode wave propagates and transfers its energy in the direction
perpendicular to the oscillations occurring in the medium.
Although the existence of the Surface Acoustic Wave (SAW) was first discussed in 1885 by
Lord Rayleigh, it did not receive engineering interest for a long time. The first SAW device
was actually made in 1965 by White and Voltmer, who found out how to launch a SAW in a
piezoelectric substrate by an electrical signal. White and Voltmer suggested that SAWs can
be excited and detected efficiently by using an interdigital transducer (IDT) placed on a
piezoelectric substrate(R.M. White and F.W. Voltmer,1965). Using IDT is a very convenient
way on a piezoelectric substrate for excitation and detection of an acoustic wave, because
very fine IDTs can be mass-produced by using photolithography, which has been well
developed for semiconductor device fabrication, and proper design of the IDT enables the
construction of transversal filters with outstanding performance. Starting around 1970, SAW
devices were developed for band-pass filters, pulse compression radar and oscillators for TV
sets and professional radio. Since 1980, the rise of mobile radio caused a dramatic increase in
demand for filters, particularly for todays cellular telephones.
All acoustic wave devices and sensors use a piezoelectric material, such as quartz crystal, to
generate the acoustic wave. Piezoelectricity, which was discovered by Pierre and Jacque
Curie in 1880, means pressure electricity due to the Greek word piezo for pressure. The
Curie brothers were able to demonstrate the generation of electric charge in response to
applied pressure or stress; this is so called the direct piezoelectric effect, as shown in Fig. 3.
Piezoelectricity is received its name in 1881 from Wilhelm Hankel, and remained largely a
curiosity until 1921, when Walter Cady discovered the quartz resonator for stabilizing
Study on Wireless Torque Measurement Using SAW Sensors
111
Fig. 3. The direct Piezoelectric Effect.

electronic oscillators. The piezoelectric effect has been understood as the linear
electromechanical interaction between the mechanical and the electrical state in crystalline
materials with no inversion symmetry (Gautschi, G., 2002). For precision positioning
purpose, the converse piezo effect is usually applied to produce precise displacements in
response to the applied electric field, as shown in Fig. 4. The phenomenon is reciprocal, but
not linear with hysteresis.
Fig. 4. The reverse Piezoelectric Effect.

Therefore, an oscillating electric field is used in piezoelectric acoustic wave sensors to create
a mechanical wave, which propagates through the substrate and is then converted back to
an electric field for measurement. Based on this principle, SAW devices have found diverse
applications for measuring physical quantities such as temperature, pressure, torque,
acceleration, tire-road friction, humidity, etc.. There are several methods to measure forces
and torques. The force to be measured is often converted into a strain on a flexible element.
The change of strain is subsequently measured by a sensor, for example, a piezoresistive, a
capacitive or a resonant sensor. Unfortunately, the metallic resistance strain gauge is
relatively insensitive such that the actual output voltage is usually only several mV of
analog voltage before amplification and the gauges must not be significantly overstrained.
The measurement range and overloading capabilities for strain gages are seriously
restricted. In general, measurement instrumentation needs smaller sensing devices of lower
power consumption and with large measurement range and overload capabilities. Digital
microelectronics with greater compatibility is highly desirable. Noncontact and wireless
operation is sometimes needed; in addition, batteryless devices are also necessary in some
application cases.
SAW devices became important for sensor purposes in recent years. SAW sensors are
applied as well as wired sensor elements in active circuits and as remote passive devices.
The SAW sensors are passive elements, which do not need power supply, and can be
accessed wirelessly, enabling remote monitoring in harsh environment. They can work in
the frequency range of 10 MHz to several GHz. In addition, SAW sensors have the
advantages such as compact structure, outstanding stability, high sensitivity, low cost, fast
112
real time response, extremely small size (lightweight), and their fabrication is compatible
with CMOS and micro-electro-mechanical (MEMS) integrated circuits technology. The SAW
sensors are used for identification of moving object and parts (so called ID tags) and
wireless measuring of temperature, pressure, stress, strain, torque, acceleration, tire-road
friction, and humidity. The SAW sensors are well suited for measuring pressure in car and
truck tires and tire-road friction monitoring. Their characteristics offer advantages over
technologies such as capacitive and piezoresistive sensors, which require operating power and
are not wireless. SAW has been used in electronics for many years, notably in quartz
resonators which provide high Q-value as a result of the low acoustic losses. Exploiting the
delay lines give a long delay in a small space with low acoustic velocities. SAW resonant
sensors offer many benefits, such as improved sensitivity and accuracy, and reduced power
consumption, among others. The resonance frequencies of the resonator-based sensors will be
changed due to the action of the external excitation; therefore, we can use this characteristic of
SAW device to detect the external force or torque via designing a interrogation (read-out)
electronic circuit. Therefore, in this study we focus on designing an embedded system to
measure the interrogation resonance frequency via a RF antenna set. In addition, this
researchs purpose is to establish a torque measurement based on a wireless SAW
measurement system. Finally, we illustrate how to measure the torque wirelessly using the
relation between the frequency shift and the torque, which is applied to a rotary rod.
2. Wirelessly torque measurement

2.1 Literature reviews for wireless torque measurement
A torque sensor or torquemeter is a device for measuring and recording the torque on a
rotating system, such as a bicycle crank, rotor, gearbox, crankshaft, transmission, or an
engine. The torque measurement can be separated into two categories, which are static
torque and dynamic torque measurements. For static torque measure, using strain gage
applied to a shaft or axle as torque sensors or torque transducers is the most common way.
Static torque measurement is relatively easy with respect to dynamic measurement, because
the measured shaft or axle is static. However, dynamic torque is not easy to measure
because the shaft is rotating. Therefore, for a dynamic torque measuring system, the torque
effect is generally transferred into some electric- or magnetic-type signals transmitting via
wireless technologies from the rotary shaft being measured to the static system. Dynamic
torquemeters is usually used to measure the torque being transmitted between the two
machines which are connected. There are several varieties of torquemeter couplings, which
are used for continuous on-line torque monitoring. Torquemeter designs are faced with the
task of detecting a physical change due to torsion in the coupling while the shaft is rotating.
Therefore, the dynamic torque measurement is necessary to perform through noncontacting means and get the torque information to a stationary output device via wireless
device. Over the years, many methods have been devised to measure the torsional effects
exhibited by the coupling; from the past literatures or present commercial products, we can
find the following types of non-contact torque measurement, which are strain gage type
torquemeters using twist angle or phase shift (torsional deflection) measurement,
magnetoelastic torque sensors, and SAW torque sensors. We will make a literature survey as
the following.
113
2.1.1 Strain gage integrated with a wireless transfer system

In the 1930s, the first in-line torque transducer was used on the liner Queen Mary, where
employed the phase displacement principle to measure torque. It needed a length of 4
meters to adequately measure the angle of twist with around 5% accuracy because of crude
electronics. The modern torquemeters use the twist between a pair of toothed flanges, which
is used to generate sinusoidal signals in magnetic pickups in the form of internally toothed
rings and circumferential coils, to measure phase displacement. Fig. 5 describes the modern
torque transducer using the phase displacement principle. Then, the resultant phase change
of the two signals is measured by digital electronics to compute the applied torque. Phase
displacement transducers are best suited to high speed and high temperature applications
and they can achieve the accuracies of 0.1%.
Fig. 5. The torquemeters using phase shift with 2 channels pick-up rings (Corcoran, joe and
DErcole, Steve, 2000).
Different from the phase shift measurement, another way to measure torque uses the angle
of twist resulting from the applied torque, which is measured using two angular position
sensors and a differential transformer circuit. The measurement system consists of a
differential transformer measuring system, two concentric cylinders fixed to a shaft either
side of a torsion section and two concentric coils attached to the housing, as shown in Fig. 6.
Both cylinders have circumferential slots, which rotate inside the coils. An alternating
current flows through a primary coil and when the slots start to overlap due to shaft twist, a
torqueproportional EMF is induced in a secondary coil. However, its performance is worse
than the phase displacement transducer, but its cost is generally lower than the phase
displacement transducer.
The strain gage is also called load cell, which is delivered by Load Delvin in 1856; the
principle of the strain gage is using the relation between the resistance and the strain for
metal materials. When the metal martial suffers pull or tension, its resistance will be
increased due to the strain. Otherwise, the resistance decreases when the martial suffers
pressure. According to the relation between the strain of the metal and the applid torque,
the strain gages can be used to detect torque. However, the strain gage needs the electrical
source to produce the variation of voltage for measurement so that the torque sensor using
strain gage needs an external power supply. On one hand, a means to power the strain
gauge bridge is necessary, as well as a means to receive the signal from the rotating shaft.
On the other hand, the non-contact measurement is accomplished using slip rings, wireless
telemetry, or rotary transformers. The first rotating strain gauge torque transducer
employed a system of slip rings to make the electrical connections from the casing to the
rotating shaft. For the slip ring torque transducers, there are two problems which should be
noted. The first one is that the slip rings are carrying only millivolt signals from the strain
114
Fig. 6. The torquemeters using angle of twist resulting from the applied torque (David
Schrand).
Fig. 7. The strain-gage type torquemeter with RTS (PCB load & torque).
gauges; therefore, the materials for both the slip rings and the brushes have to be carefully
selected. The second is that slip ring torque transducers should be applied for slower speed
and short-term test applications due to the lifetime of brushes. The rotary transformer
system (RTS), which uses induction principles to transfer signal or power to the
measurement system, became popular in the 1980s. RTS uses high-performance, hightemperature electronic components to be as a telemetry system but it only uses kHz carrier
frequencies, not the MHz of radio systems. Several companies use the rotary transformer
system with strain gauges, and with typical inaccuracies of 0.2%, and speeds up to
50,000rpm it can be a very cost-effective torque transducer.
For novel types of torque transducers, conditioning electronics and an A/D converter may
be integrated into a microchip which is fixed to the rotating shaft; then, stator electronics
read the digital signals and convert those signals to a high-level analog output signal. In this
case, the wirelessly strain-gage type torque sensor, which needs a a bettery to support the
power of the system, is not a passive device and it needs a wireless communication system
to transfer the measurement data via RF. Wireless communication technologies are convent
for remote control, such as ZigBee, Bluetooth or wirefire networks. ZigBee is a new
communication technology in the wireless field and ZigBee comes from the bee which
communicates with other bees about position of pollen via shape of dance. In the Fig.8
describes a measurement system using the strain gage integrated with Zigbee. The torque
sensor integrated with ZigBee wireless communication sensor can detect the torque of shaft
to prevent breaking due to the fact that the torque exceeds the material limit. The torque
sensor systems with ZigBee involves of two parts, measuring unit (strain gages) and
receiver unit (microprocessor). The measuring unit is used to measure torque from strain
gages and transmit digital signal to receiver unit. Then, the MCU is used to perform A/D
conversion and signal processing tasks. A simple MCU is enough perference for measuring
task and transmit signals. After that the Zigbee chip transfers the digital signal into RF
signal and transmits it to the receiver.
115
Fig. 8. The block diagram of ZigBee torque measuring system.
Fig. 9. The block diagram of wireless measuring unit (Binsfeld Engineering Inc.).
2.1.2 Magnetoelastic torque sensors
Besides of strain-gage type torquemeter, one way to achieve this non-contact measure uses the
magnetic characteristics of the shaft with a series of permanent magnetic devices. The
magnetic characteristics will vary according to the applied torque, and thus can be measured
using magnetoelastic torque sensors, which are generally used for in-vehicle applications on
racecars, automobiles, and aircraft. The United States Patent, the inventor Bunyer, Scott
L(Bunyer, Scott L, 2007) in his patent Magneto-Elastic Resonator Torque Sensor , the torque
sensor consists of a substrate and a magneto-elastic sensing component formed from or on the
substrate. The magneto-elastic sensing component and the substrate together form a magnetoelastic torque sensor, which when subject to a stress associated with a torque, shifts a
characteristic frequency thereof linearly in response to the torque, thereby inducing a pathway
by which magneto-elastic energy is coupled to excite vibrations in a basal plane of the
magneto-elastic sensor, thereby generating torque-based information based on resonator
frequency thereof. Figs. 10-11 illustrate the torque sensing system using magneto-elastic
resonator. A few years ago, a magneto-elastic device was developed in the USA, which uses a
magneto-elastic sleeve fixed to a stainless steel shaft. This permanent magnet generates a
magnetic field proportional to torque. The resultant magnetic field is less dense that the earths
field, hence internal shielding is needed for the hall-effect probes, which is fine for applications
whose accuracy is not critical. The ABB Company proposed the magneto-elastic torque sensor,
which is called Torductor-S (ABB company, 2007). Since the sensor is part of the loadcarrying shaft and Torductor-S gives a true non-contact and rugged torque sensor without
any moving parts. Therefore, the measured torque is the true transmitted torque, which
enables Torductor-S to combine high accuracy with high overload capacity and fast response
at all times. A high output signal ensures integrity against electrical or magnetic interference
from the surroundings. The magneto-strictive transducer, which is similar to the magnetoelastic torque sensor, is mechanically simpler than the magneto-elastic unit and relies on a premagnetized shaft proportionally changing its magnetic field when torque is applied. The
magneto-strictive transducer is a low-cost transducer; however, zero drift with time and
temperature and the effect of adjacent magnetic fields can be a problem for. These two devices
can provide an accuracy of around 1%.
116
Fig. 10. The torque sensing system using magneto-elastic resonator (Bunyer, Scott L, 2007).
Fig. 11. The torque sensing system using magneto-elastic resonator (Bunyer, Scott L, 2007).
2.1.3 SAW torque sensors
Although the existence of the Surface Acoustic Wave (SAW) was first discussed in 1885 by
Lord Rayleigh, it did not receive engineering interest for a long time. Until 1965, R.M. White
suggested that SAWs can be excited and detected efficiently by using an interdigital
transducer (IDT) placed on a piezoelectric substrate(R.M. White and F.W. Voltmer, 1965).
This is because very fine IDTs can be mass-produced by using photolithography, which has
been well developed for semiconductor device fabrication, and proper design of the IDT
enables the construction of transversal filters with outstanding performance. Acoustic
devices are robust in respect to temperature and mechanical stress discussed by Seifert et al.
(F. Seifert, A. Pohl, R. Steindl, L. Reindl, M.J. Vellekoop, and B. Jakoby, 1999). They are
reliable more than the lifetime of other electronic devices. Application of SAW devices to
non-contact torque measurement was first suggested and patented in 1991 by A. Lonsdale
and B. Lonsdale (A. Lonsdale, B. Lonsdale, 1991). Non-contact torque sensors based on SAW
reflective delay lines were also introduced in (U. Wolft et al.; R. Grossmann et al. 1996).
Since then some reasearchers fabricated and applied torque sensors based on SAW
resonators to a number of industrial customers (A. Lonsdale, 2001; P. Merewood, 2000). In
2002, non-contact torque sensors based on SAW resonators application was proposed by
Beckley et al. (J. Beckley, V. Kalinin, M .Lee, and K. Voliansky, 2002). Although the
operation of both types of sensor was successfully demonstrated in the abovementioned
publications, they did not cover such important aspects as limitations on the accuracy of the
sensors resulting from their non-contact interrogation and the temperature stabilization of
their sensitivity to torque. SAW resonators have been a commercial success on radio
frequency applications, especially for filter and oscillator implementations. Their impact has
made possible considerable reductions in size and power of the chipsets of mobile
devices(Fujitsu Media Devices Ltd., 2006; Epcos AG, 2008). More modest and important
applications of SAW resonators are the measurements in mass detector and pressure sensor
devices with application in bio-particle detection(Martin, F., 2004; Talbi, A., 2006).
117
In recent years some instrument manufacturers of force and torque measurement devices
have investigated instead of using resistance strain gauges. For example, one leading
manufacturer of weighing machines now uses metallic and quartz resonant tuning fork
technologies for industrial applications and the others have established some applications
using surface acoustic wave (SAW) technology, optical technology, and magneto-elastic
technology. Further commercial developments are taking place to enhance device
manufacturability and improve device sensitivity and robustness in operation.
Measurement on stiffer structures at much lower strain levels than before is possible now.
Reviewing the world patents, we can find many torque measure patents using Piezo or
SAW devices. In the United States Patent US 2003/0000309 A1, A. Lonsdale and B. Lonsdale
claimed the patent of Torque measurement, in which they proposed a method and
apparatus for measuring the torque transmitted by a member comprises a SAW device
secured to the member such that mechanical stress in the member due to torque transmitted
thereby induces bending of the SAW device (A. Lonsdale and B. Lonsdale, 2003), as shown
in Fig. 12. In another United States Patent, LEC, RYSZARD MARIAN, Magee et al. claimed
the Torque Measuring Piezoelectric Device and Method using non-contact measurement
of torque applied to a torque-bearing member such as a shaft. The method involves the use
of a piezoelectric transducer mechanically coupled to the shaft for rotation therewith, and
having electrical characteristics responsive to applied torque(Steven J. Magee, 2004).
Electrical signal characteristics are changed by the torque-dependent transducer
characteristics. In 2007, in another United States Patent, Steven J. Magee claimed that the
sensor systems and methods are disclosed herein, including a sensor chip, upon which at
least two surface acoustic wave (SAW) sensing elements are centrally located on a first side
of the sensor chip. The SAW sensing elements occupy a common area on the first side of the
sensor chip. An etched diaphragm is located centrally on the second side of the sensor chip
opposite the first side in association with the two SAW sensing elements in order to
concentrate the mechanical strain of the sensor system or sensor device in the etched
diagram, thereby providing high strength, high sensitivity and ease of manufacturing.
Low propagation velocity enabled the use of SAW devices for time delays and filtering in
radar systems and television but their application burgeoned within the mobile phone
market. Honeywell SAW sensors use small piezo-electric quartz die upon which two or
three single-port resonators, with natural resonant frequencies around 434 MHz, are
fabricated in aluminum using standard photo-lithographic techniques, as shown in Fig. 13.
(Magee et al., Honeywell, 2007). A SAW resonator is excited by a short radio-frequency (RF)
burst. The centrally placed interdigital transducer (IDT) converts the electrical input signal
to a mechanical wave through the piezo-electric effect. The waves propagate from IDT to
reflectors and back until a forced resonance exists as a standing wave. After the transmit
signal is switched off, the resonator continues to oscillate but at a frequency modified by any
applied mechanical and/or thermal strain. The decaying oscillation is converted back to an
electrical signal via the piezo-electric effect and retransmitted to the SAW interrogation
board where the frequencies are analyzed and converted to engineering parameters. A more
recent development is the use of SAW devices attached to the shaft and remotely
interrogated. The strain on these tiny devices as the shaft flexes can be read remotely and
output without the need for attached electronics on the shaft. The probable first use in
volume will be in the automotive field as, of May 2009, Schott announced it has a SAW
sensor package viable for in vehicle uses.
118
Fig. 12. The block diagram of a SAW torque sensor system. (A. Lonsdale and B. Lonsdale,
2003).
Fig. 13. The diagram of a SAW torque sensor system. (Magee et al., Honeywell, 2007).
A British company, Sensor Technology Ltd., has developed a Torqsense digital
RWT310/320 series transducers using surface acoustic wave (SAW) technology. Torqsense
is a registered trademark used for Electric or Electronic Transducers for providing electric or
electronic sensors for sensing torque and owned by Lonsdale, Bryan, Lonsdale, Anthony.
Based on a revolutionary new, patented, technology using SAW strain sensing elements,
Torqsense sensors offer a superior solution to non-contact rotary torque measurement in
industrial applications. The SAW torque devices are strain sensitive elements with very high
frequency modulating and it has the ability to convert an electrical signal into an acoustic
signal with the same frequency. Because of a reduction in propagation velocity of about five
orders of magnitude, the SAW device has a much smaller wavelength, which allows
manipulation of an RF signal in a very small package. Therefore, we can consider a SAW
device as a frequency dependent strain gauge, which can be used to measure the change in
resonant frequency caused by the strain in the shaft. For dynamic torque measurement, the
SAW device needs an RF couple to transmit the signal from the shaft to a fixed pick-up. It is
a relatively low-cost device offering a 0.25% accuracy giving a total system accuracy of
around 0.35% (patent Magee, 2007).
2.2 Types and applications of SAW devices
Wirelessly interrogated passive SAW sensors can employ one of the two basic types of SAW
devices, one-port resonators(F. Jerems et al, 2001; A. Lonsdale, B. Lonsdale, 1991; A.
Lonsdale, 2001; P. Merewood, 2000) and reflective delay lines (F. Schmidt, G. Scholl; A. Pohl,
F. Seifert, 1997; U. Wolft et al.). Obviously, ordinary delay lines and two-port resonators can
also be used as passive sensing elements, although they are less preferable because of the
119
larger number of RF coupling devices are required (F. Schmidt, G. Scholl). If active
components are allowed on the shaft then both the delay line and the resonator can be used
in a feedback loop of a SAW oscillator, wirelessly coupled to an interrogation unit and
wirelessly powered by an RF signal. This is an approach inherited from a traditional wired
SAW sensor design.
2.2.1 Reflective delay lines of SAW devices
Most SAW passive sensors are designed using a reflective delay line. The surface acoustic
wave in a reflective delay line propagates towards reflectors distributed in a characteristic
barcode-like pattern and is partially reflected at each reflector. Fig. 14 shows the SAW delay
line sensor and the principle of the sensor is drawn in Fig.15. Fig. 16 shows reflectors lined
of same track and different tracks.
Fig. 14. The SAW delay line sensor.
Fig. 15. The reflectors lined of same track (Weidong Cheng et al., 2001).
Fig. 16. The reflectors lined of different tracks(Weidong Cheng et al., 2001).
2.2.2 Resonators of SAW devices
SAW resonators exploit SAW propagation and electromechanical transduction to implement
electronic circuits as filters, oscillators, and sensors. A SAW resonator is basically a resonant
cavity in which a first transducer electrode converts the electric signal into a lateral
mechanical wave. The resulting SAW propagates on the piezoelectric to reach the second
electrode, where it is transduced back into the electrical domain. When arriving at the
second electrode, and typically aided by one or more reflector electrodes, the acoustic wave
bounces back in the direction of the first electrode, and the electromechanical conversion is
120
repeated indefinitely, as depicted in Fig. 17. Thus, the acoustic wave is trapped in the cavity
formed by the resonator electrodes.
Fig. 17. The surface acoustic wave (SAW) resonator (Humberto Campanella, 2010).
The available photolithography resolution limits the dimensions of the IDTs and the
piezoelectric layer determines the maximum operating frequency of the fundamental mode
of the resonator. However, as well as other electromechanical resonators, SAW devices can
be operated at over-tone modes to bypass near-to-fundamental bulk or other resonance
modes. Typical frequencies for SAW-resonator-based applications are in the UHF band
below 1 GHz, although high-performance commercial devices in the 1.5-GHz GPS and 1.9GHz PCS bands are available so far(Epcos AG, 2002; TriQuint Semiconductor, 2009). The
fundamental frequency of the SAW resonator mainly depends on the pitch of the IDTs,
which is chosen to be equal to the SAW wavelength , and the sound speed of the
piezoelectric layer v :
f0
(1)
SAW resonators are found in two types of port configurations one-port and two-port
resonators. One-port resonators are two-terminal devices, and they find application in
oscillator circuits like VCOs or Colpitts oscillators. Two-port resonators behave more like
narrow band-pass filters.
Fig. 18 shows a typical configuration One-port SAW resonators, The One-port SAW
resonators have a single IDT generating and receiving the SAW, and two grating reflectors,
which reflect the SAW and generate a standing wave between the two reflectors. The IDT
and reflectors are fabricated on quartz crystal substrate or another piezoelectric material and
patterned by photolithographic processes(Ken-Ya Hashimoto, 2000). As Fig. 19 shown, its
equivalent circuit near resonance, where CM and LM are the dyamical capacitance and
inductance, respectively, corresponding to the contributions of elasticity and inertia. On
other hand, CO is the static capacitance of the IDTs, and RM is the motional resistance
corresponding to the contribution of damping. On the other hand, two-port SAW resonators
exhibit two IDTs, one of them generating the SAW, and the second one picking it. As with
one-port devices, two grating reflectors aid the SAW to be reflected and confined between
the IDTs. The generic connection is connected to the first and second IDTs as shown in Fig.
20 for two-port SAW resonator and Fig. 21 shows the resonator equivalent circuit.
121
Fig. 18. The one-port SAW resonator (Humberto Campanella, 2010).
Fig. 19. The equivalent circuit for the one-port SAWR (Ken-Ya Hashimoto, 2000).
Fig. 20. The two-port SAW resonator (Humberto Campanella, 2010).
Fig. 21. The two-port SAW resonator equivalent circuit (Ken-Ya Hashimoto, 2000).
2.2.3 Characteristics of one-port SAW resonators
In this study, we using one-port SAW resonators device. Fig. 18 shows typical configuration
one-port SAW resonators. Fig. 22 shows frequency response for resonator and in this study
we just measure the resonance frequency r or antiresonance frequency a .
Fig. 22. The frequency response for a resonator (Ken-Ya Hashimoto, 2000).
122
From the equivalent circuit of one-port SAW device in Fig. 19, we have
r
a
1
,
LM CM
1
LM CM CO
CM CO
(2a)
(2b)
where CM and LM are the dyamical capacitance and inductance, respectivly. On the other
hand, CO is the static capacitance of the IDTs, and RM is the motional resistance
corresponding to the contribution of damping. Fig. 23 shows typical electrical resonance
characteristics of one-port SAW resonator. As shown in Fig. 22, the conductance G takes a
maximum at the frequency above the resonance frequency r ; in addition, there exists the
anti-resonance frequency a where the resistance G 1 takes a maximum.
Fig. 23. The electrical resonance characteristics of one-port SAW resonator (Ken-Ya
Hashimoto, 2000).
As mentioned in Ref. (Ken-Ya Hashimoto, 2000), CM and LM can be obtained from the
measured CO , r and a . RM can be determined by G 1 at r . The capacitance ratio
is frequently used as a measure of the resonator performance, and is for:
CO
1
,
CM (a / r )2 1
(3)
which corresponds to the inverse of the effective electromechanical coupling factor. The
quality factor Q at the resonance frequency is called the resonance quality factor, which is
denoted by Qr . This is also an important measure, which is shown by:
Qr
r LM
RM
r CM RM
(4)
123
Both resonance frequency r and antiresonance frequency a will be shift by temperature

change and effect of applied stress. Effect of temperature can have a considerable affect on
propagation velocity for surface acoustic wave device, since many of the material constants
involved are themselves temperature dependent. The temperature dependence of surface
wave attenuation for quartz in particular has been studied over various temperature and
frequency range(M. R. Daniel and J. De Klerk, 1970). Different piezoelectric materials,
propagation directions, and cuts all show different temperature dependencies over different
range. It is thus possible to engineer that a SAW device has specific temperature
characteristics, which has led to the development of several SAW based temperature
sensors(D. Hauden et al., 1981; J. Neumeister et al., 1990; M. Viens and J. D. N. Cheeke,
1990). In effect of applied stress also contributes to acoustic wave attenuation and velocity
change, and has been studied extensively by Slobodnik(A. J. Slobodnik, 1972). In this study,
the attenuation is due to the generation of surface acoustic waves in the gas in contact with
the SAW device. In other words, the shear vertical component of the wave causes periodic
compression and rarefaction of the gas, resulting in a coupling of acoustic energy from the
SAW device into the air.
2.2.4 Applications of torque sensors based on SAW resonators
Although the existence of the Surface Acoustic Wave (SAW) based on torque measurement
was first discussed in 1996 by A.Lonsdale (A. Lonsdale et al.). Torque (radial strain)
transducers are one of the more common devices used by development engineers.
Knowledge of torque and rotational speed can be used to indicate power, from which the
efficiency of gearboxes, transmissions, electrical machines and many other systems can
readily be assessed. To apply the SAW element principle, two devices are used in a half
bridge, analogous to the classic resistive strain gauging configuration; one positioned so as
to be sensitive to the principal compressive strain and the other positioned to observe the
principal tensile strain. In the absence of bending moments and axial forces, the principal
stress planes lie perpendicular to one another at 45 to the plane about which the tensional
moment is applied. This is illustrated in Fig. 24.
Fig. 24. Torque sensing element based on SAW resonators(Stephen Beedy et al., 2004).
In this SAW sensors applied rely on the fact that the torque M applied to the shaft creates
two principal components of strain, S XX SYY S . As a result, one of the SAW devices is
under tension and the other one is under compression, causing the opposite change of
resonant frequency in the devices. The resonators have the same or better performance for
the same size of substrate and are less demanding in terms of the receiver bandwidth and
124
sensitivity. Resonator Q factors are about 10,000. The torque sensor interrogation system can
employ continuous frequency tracking of reflected frequencies from the two SAW resonators.
This applied can achieve temperature compensation and eliminate sensitivity to shaft
bending. From the technique described it is apparent that the output signal will be in the
frequency domain as shown in Fig. 25.
Fig. 25. The schematic of frequency domain for SAW signals(A. Lonsdale, 1993).
3. Wirelessly measurement architecture based on SAW device

An acoustic wave is a disturbance in an elastic medium that propagates in space and time,
thus transferring the energy supplied by an excitation source along the medium in the form
of oscillation or vibration. There exist two types of acoustic waves: surface acoustic waves
(SAW) and bulk acoustic waves (BAW). In this study, two one-port SAWRs (manufactured
by f-tech Corporation) with the resonant frequency of 433.42 MHz and 433.92 MHz are used
to be torque sensing devices for this study. The specifications for the SAWR (433.92 MHz)
are descibled in Table 1. Fig. 26 shows the wafer of the SAW resonators using in the
experiments and Figs. 27 and 28 show the die size of the SAWR and its microimage.
Fig. 26. The Wafer of f-tech SAWRs (433.92MHz).
Fig. 27. The photo of the SAWRs die size.
125
IDT
Reflectors
Fig. 28. The microimage of the SAWRs die.
Table 1. 433.92 MHz SAWR specifications (www.f-tech.com.tw).

In this study, we used two one-port SAWRs to obtain the torque signal via measuring the
frequency shift of SAWs, f , which is relate to the strain and the fundamental requency
of the SAWR f 0 . For each SAWR, the relation between the frequency shift and the strain can
be described as follows.
f0
v0
4d
where f 0 is the fundamental frequency of the SAWR, which mainly depends on the pitch of
the IDT, d , and vO is the speed of SAW.
Obtaining the variation of the fundamental frequency according the the variation of the
speed and pitch for the SAW, we have:
f 0 f
v0 v
v v
0
4(d d ) 4d (1 )
(5)
where is the strain. Neglect the variation of the SAW speed, becasue it is very small. Then,
we have
f 0 f
4df 0
f
0
4d (1 ) 1
Using Taylar expansion to consider the approximation, we can obtain the following.
(6)
126
1
1
1
1 2 3 ... 1
1
2
3!
(7)
then the frequency shift can be approximated as follows
(1 1)
1
f f 0
1 f 0
f 0
1
1
(8)
Therefore, we can obtain the relation between the frequency shift and the strain as follows
f f 0
(9)
where f is the shift of the resonant frequency, is the strain and f 0 is the resonant
frequency. In the implementation, two one-port SAW resonators oriented at 45o to the
shaft axis, which are shown in Fig. 29. Fig. 30 shows the locations of the two SAWR on the
shaft; we also attach the strain gage with two directions of 45o on the opposit side of the
SAWR to obtain the actual strain as shown in Fig. 30. We use the network analyzer (Agilent
E5071A) to measure the S11 value of this two SAWRs for obtaining the fundamental
frequency of these two SAWRs as shown in Fig. 31. Then, we can find the relation between
the frequency shift and the applied torque as shown in Fig. 32. Fig. 32 shows that the
resonant frequency shifts for SAWR1 and SAWR2 on the location of 45o will move in the
opposite directions as soon as torque is applied to the shaft, because the stresss direction is
inverse due to the applied torque as x and y described in Fig. 29. Therefore, we can
obtain the value of torque measuring the difference between the two frequencies and this
arrangement can compensate as mestioned in the past literature (A. Lonsdale, 1993). From
the frequency shifts of the SAWR1 and SAWR2 measured in Fig. 32, we can obtain the actual
strain which is computed according to Eq. (9).
Fig. 29. The relation between the stress and the applied torque for the SAWRs.
Fig. 30. The setup of the SAWRs and the strain gage on the shaft.
127
Fig. 31. Torque measurement using the network analyzer (Agilent E5071A).
Fig. 32. The relation between the S11 value and the frequency of the two SAWRs.
Fig. 33 shows the relation between the applied torque and the strain, which is measured
using the SAWRs. The theoretical torque can be computed as follows.
max
max
Tc T r
J (r ) 4
2
2
max
Fig. 33. The relation between the applied torque and the strain using two SAWRs.
(10)
128
where E is Youngs modulus of the material,is the stress, is the strain, r is the radius of
the torsion bar and T is the applied torque. Although the SAWRs response only 40% strain
which is compared from the strain gage in Fig. 34, the relation between the torque and the
strain is linear; therefore, we can measure the torque applied on the shaft according to the
relation obtained in Fig. 34 for the SAWRs.
Fig. 34. The relation between the applied torque and the strain using the strain gage.
If the measurement of the resonant frequency in the torque sensor is performed by means of
continuous tracking of S11 minimum at the input of the rotary coupler then a remarkably
low standard deviation of the frequency measurement error of 0.3-0.4 ppm can be achieved
at less than 1 ms update period. However the continuous tracking requires one automatic
frequency control (AFC) loop per resonator. In this study, the RSSI value is used to obtain
the resonant frequency of SAWR instead of measuring S11 minimum. Fig. 35 shows the
fabricated torque sensor, which consists of the four microstrip lines and two SAWRs. The
connection between the microstrip and the SAWR is established by the IC wire bonding
machine. The parameters of the microstrip line are designed optimally by the software of
Advanced Design System 2006A. Moreover, the wireless signals are transmitted by four
coupling antennas, which are described in Fig. 36. The proposed wireless torque sensory
system is described in Fig. 37.
Fig. 35. The rotational shaft equipped with two SAWRs.
Fig. 36. The coupling antenna for the sensory system.
129
Fig. 37. The block diagram of the proposed wireless torque sensory method.
A wireless passive torque measurement system using SAW sensors is presented as shown in
Fig. 38. To measure the torque via wireless sensing, a dsPIC architecture based on two SAW
sensors via wireless RF signals is proposed in this paper. When the torque is applied to the
shaft, the two SAWRs will change their resonant frequencies in opposite directions in the
frequency domain. According to the past literature, measuring the difference between the
two frequencies one can obtain the value of torque and achieve partial temperature
compensation, which are performed by means of continuous tracking of minimumal S11 at
the input of the rotary coupler. Different from the past method using the frequency with the
minimal S11, this study tries to obtain the maximum of S21 to measure the value of torque
via continuous tracking of output signal. An dSPIC33F microchip is used to control the RF
chips (C1100) to achieve the transmitting and receiving task. The frequency with maximum
received signal strength indication (RSSI), where the value of S21 is maximal, will be
influenced by the applied torque. The measuring system mainly uses the RSSI signal with
respect to the frequency to find the SAW center frequency value, and then the torque values
can be obtain according to the frequency shifts. Therefore, the frequency, which has the
maximal RSSI value, can be used to obtain the applied torque of the rotation shaft according
to the frequency shift. In addition, to achieve the wireless transmission, two sets of couplingtype antenna are applied on the shaft. Finally, the torque signal is passed through CANBUS
to integrate with the other vehicles electron systems. In this study, a wireless torque sensing
system has been established to measure the torque of the rotaional shaft via the RF antenna.
To illustrate the process of the proposed wireless torque measuring system, a single
electromagnetic RF interrogation system is studied to produce RF interrogation signals to
the SAW device. According to the measurement of the frequency shift, the environmental
influence can be measured by the specified relation via experimental results. A dsPIC33 chip
is used to control the RF chip (C1100) to achieve the transmitting and receiving between
433.1MHz and 433.85 MHz, and the interval scanning frequency is 0.5 KHz. According to
the response of the RSSI value with repect to the scanning frequency, the torque value can
be obtained according the look-up table obtained by the experimental results in Fig. 34.
130
After all, the microcontroller (dsPIC33) itself is equipped with a controller area network
(CAN) interface, which is responsible for the calculation of the sensor data and providing
the communication with other electronic systems(dsPIC33FJ256GP710A user guid).
According to above system, the CC1101 chip is a low- power consumption RF chip
developed by Chipcon. The main operating frequency is designed at 315, 433, 868, and
915MHz. The receive chip of RF has an amplifier to enhance a signal, and then change the
signal to IF (Intermediate Frequency). When the CC1101 plays an IF medium frequency
receiver, the I/O signals are transferred to digital signals using A/D converter. On the other
hand, the transmit chip of RF includes a LC VCO (LC Voltage-controlled oscillator) and a
LO (Local Oscillator). Fig. 38 describes the block diagram of the proposed architecture using
CC1100 and dsPIC to measure the frequency shift of the SAW devices. In the measurement
system, the MCUs performance is the critical component, which dominates system
bandwidth, sensitivity, and flexibility. Therefore, we choose the dsPIC33F to improve our
measuring system. The dsPIC33F is a 16-bit DSC chip, which is developed by to integrate
16bits MCU and DSP with high speed computation. The dsPIC provides 84 commands, 16
registers and DMA as shown in Fig. 39. The wireless signals are transmitted by four
coupling antennas, which are described in Fig. 37. Fig. 38 shows the flow chart of the
proposed measurement system, which microchip controls two RF chips to perform
transmission and receiving. Then, the signals are transmitted through the SAW via coupling
antennas; at the same time, the dsPIC33 can obtain the RSSI with respect to frequency for
SAW devices. According to the property of the SAW, the frequency intensity of RSSI is
larger than others when the signals frequency approaches the resonant frequency. After
Fig. 38. The block diagram of the proposed measurement system.
Fig. 39. The architecture of dsPIC33.
131
dsPIC33 receives the RSSI analysis for the signal in the scanning frequency domain, we can
obtain a shift value of frequency with comparison from the initial resonant one for each
signal. To control the RF transmit and receive modules of CC1100, the kernel of
measurement system is programed in the microchip dsPIC33F which possesses UART and
CANBUS transferring functions. The dsPIC33 controller is connected with CC1101 RF
modules, which include transmitting and receiving task as shown in Fig. 38.
The dsPIC33 MCU uses four lines of SPI to communicate with CC1101 module such as
sweep scanning time. In addition, the RSSI value of CC1101 RF receiving module uses 8-bit
digital signal to communicate with the dsPIC33, which controls two RF chips for
transmitting and receiving via the SPI signals. For the RF chip which performs transmitting
task, dsPIC33 is only set the frequency for transmitting in the register of CC1100. When the
RF chip which performs receiving task, the dsPIC33 sets the frequency for receiving in the
register of CC1100 and receives the RSSI value in DMA at the same time via SPI signals.
Then, the RSSI value is transferred via UART to the host PC to describe the measurement
results. The flow-chart of the above process is describled in Fig. 40.
Fig. 40. The flow chart of the AFC Loop code in the dsPIC33F.
4. Experimental results and discussions

In this implementation, the SAW sensors with wireless RF coupling antennas as shown in
Fig. 41, are attached on a shaft; the dsPIC33 microchip is used to achieve the transmitting
Torque sensor
Input antenna
Output antenna
Fig. 41. The wireless torque sensory system.
132
and receiving RF signal tasks via the RF chips (CC1100). The program embedded in dsPIC33
microchip is coded using MPLAB software, which is the development environment based
on C++ language. The microchip dsPIC33 is used to set the transmit frequency of C1100
from 433.1MHz to 435MHz and the interval between the scanning frequency is from 0.5 to
10 KHz for the specified bandwidth. On the receiver side of CC1100, the dsPIC33 makes the
RF receiver chips frequency to match the frequency of transmitting synchronously and then
it read the DMA to obtain the intensity of the RSSI. Therefore, the frequency with the
maximal RSSI value is obtained and transmitted to PC using CANBUS network. The flow
chart of the above processes is described in Fig. 42. On the PC side, a Labview program is
developed to display the frequency response as shown in Fig. 43. In this case, we have
developed a simple virtual instrument whose purpose is similar to the network analyzer.
Fig. 42. The flow chart of the dSPIC 33F program.
Fig. 43. The RSSI response with respect to frequency in LabView interface.
133
Fig. 44 shows the relation between the torque and the frequency shift. We can establish a
linear mapping relation between the torque and the frequency shift via using the
experimental results in Fig. 44. Then, the torque can be obtained in real-time. For this
wireless torque sensing system, the performance index for the sensory system is important
for feasible applications. On one hand, we use dsPIC33F with a wide bandwidth to increase
torsion sensor performance. On the other hand, different frequency scanning interval
introduces the different measurement performance as shown in Fig. 45. The developed
sensory systems bandwidth is described in Table 2. From the experimental results, we can
summarize the following comments. (1) The smaller scanning interval will cause the lower
system bandwidth, but the measurement system is more stable. (2) The less number of
scanning points can improve the system performance with bad stability.
Fig. 44. The experimental results between the torque and frequency shift.
Fig. 45. Number of frequency scanning points v.s. the system bandwidth.
No. of
scanning
points
System
bandwidth
(Hz)
Scanning
interval
(KHz)
Perturbation
(KHz)
2000
1000
500
200
100
15
38.43
64.11
191.35
383.1
0.5
1
2
5
10
1
2
4
10
20
Table 2. The bandwidth of the developed torque sensory system.
134
5. Conclusion
In this paper, we study SAW devices to establish a wireless torque measurement system,
which uses the analysis of RSSI to obtain the frequency shift of SAW using embedded
microchip dsPIC33F. From the experimental results, a prototype of virtual instrument based
on microchip integrated with LabView is used to measure the frequency response of the
SAW devices; that is, we developed a simple PC-based Spectrum Analyzer. The
measurement through sweeping varies frequency to obtain intensity of RSSI signals can
establish the relation between the torque and the frequency shift via the proposed
measurement system using UART and CANBUS. The experimental results validated the
proposed measurement system.
6. Acknowledgment
This work was supported by the National Science Council under NSC grant NSC 100-2221E-027-031.
7. References
A. J. Budreau and P. H. Carr, Temperature Dependence of the Attenuation of Microwave
Frequency Elastic Surface Waves in Quartz, Appl. Phys. Lett., 1971, 18, 239.
A. J. Slobodnik, Jr., Attenuation of microwave acoustic surface waves due to gas loading,
Appl. Phys., 1972, 43, 2565.
A. Lonsdale, Dynamic rotary torque measurement using surface acoustic waves, 2001,
Sensors, vol. 18, pp. 5 1-55
A. Lonsdale, DYNAMIC ROTARY TORQUE MEASUREMENT USING SURFACE
ACOUSTIC WAVES, 2001, Technical Director Sensor Technology Ltd
A. Lonsdale, B. Lonsdale, Method and apparatus for measuring strain, Int. patent public.
No. WO 91/13832, 19 Sept. 1991, Int. Applic. No. PCT/GB91/00328, Int. filing date:
4 March 1991, Priority: 9004822.4, 3 March 1990, GB .
A. Pohl, F. Seifert, Wirelessly interrogable suacoustic wave sensors for vehicular
applications, 1997, IEEE Trans. Instrum. Meas., vol. 46, pp. 1031-1 038.
A.Lonsdale, Transence Technologies, and N.Schofield, MST Ltd. An Intergrated Low Cost
Sensor For The Direct Torque Control of Brushless DC MOTORS, 1996, Machines
for automotive applications, Vol. 4, pp. 6/1-7
Auld, B. A., 2nd, Acoustic Fields and Waves in Solids, 1990, Krieger Publishing Company,
Vols. I and II, Malabar.
Bunyer, Scott L et al., Magneto-Elastic Resonator Torque Sensor, Aug. 28, 2007, United
States Patent, Patent NO. US 7,261,005 B2.
Corcoran, Joe and DErcole, Steve, A New Development in Continuous Torque Monitoring
Couplings, 2000, Proceedings of ASME DETC 2000 Power Transmission and
Gearing Conference.
D. Hauden, G. Jaillet, and R. Coquerel, Temperature Sensor Using SAW Delay Line, IEEE
Ultrason. Symp. (Proc.), 1981, 148.
D. Hauden, M. Michel, G. Bardeche, and J. J. Gagnepain, Temperature Effects on Quartz
Crystal SurfaceWave Oscillators, Appl. Phys. Lett., 1977, 31, 315.
dsPIC33FJ256GP710A user guide, Microchip Inc.
135
Epcos AG, SAW Frontend Module GSM 850/EGSM/DCS/PCS/WCDMA 2100, 2008,

Surface Acoustic Wave Components Division.
F. Jerems, M. Pro& D. Rachui and V. Kalinin, A new generation non-torsion bar torque
sensors for electromechanical power steereing applications, 2001, Proc. Int.
Congress on Electronic Systemsfor Vehicles, pp. 27-28 September, Baden Baden.
F. Schmidt, G. Scholl, Wireless SAW identification and sensor systems , Advances in Surface
Acoustic Wave Technology, Systems and Applications, vol. 2, C. C. W. Ruppel and T. A.
Fjeldly, Eds. Singapore, World Scientific.
F. Seifert, A. Pohl, R. Steindl, L. Reindl, M.J. Vellekoop, and B. Jakoby, Wirelessly
interrogable acoustic sensors. 1999 IEEE INTERNATIONAL FREQUENCY
CONTROL SYMPOSIUM., pp. 1013 1018.
Gautschi, G, Piezoelectric Sensorics, Force, Strain, Pressure, Acceleration and Acoustic
Emission Sensors, 2002, Materials and Amplifiers
http://www.fujitsu.com., Fujitsu Media Devices Ltd., Cellular SAW Duplexer, FAR-D5GA881M50-D1AA, 2006,
http://www.honeywell.com/sensing, Honeywell Inc.
http://www.pcb.com/Linked_Documents/ForceTorque/Catalog/Sections/FTQ200G_0107_Torque.pdf, Torque Sensor PCB
Piezotronics, Inc
http://www.sendev.com/catalog/pdf/torque-measurement.pdf, David Schrand, Sensor
Developments Inc
http://www.sensors.co.uk/torqsense/RWT310-230, sensor technology
http://www.triquint.com., 881.5/1960 MHz SAW Filter, 2009, TriQuint Semiconductor.
http://www05.abb.com/global/scot/scot264.nsf/veritydisplay/eef8846525950f25c1257380
003623c6/$File/3BSE022227R0101_-001.pdf, Torductor-S, 2007, ABB Company.
Humberto Campanella Acoustic Wave and Electromechanical Resonators2010 ARTECH
HOUSE
Humberto Campanella, Acoustic Wave and Electromechanical Resonators, 2010 ARTECH
HOUSE Inc..M. R. Daniel and J. de Klerk, TEMPERATUREDEPENDENT
ATTENUATION OF ULTRASONIC SURFACE WAVES IN QUARTZ, 1970,Appl.
Phys. Lett., Vol. 16, pp. 30.
J. Beckley, V. Kalinin, M. Lee, and K. Voliansky, Non-Contact Torque Sensors Based on
SAW Resonators, 2002 IEEE International Frequency Control Symposium and
PDA Exhibition.
J. Neumeister, R. Thum, and E. Luder, A SAW delay line oscillator as a high-resolution
temperature sensor, Sens. Actuators A, 1990, A21-23, 670.
Ken-Ya Hashimoto, Surface Acoustic Wave in Telecommunications Modeling and
Simulation 2000, Springer, Inc..
M. Viens and J. D. N. Cheeke, Highly Sensitive Temperature Sensor Using SAW
Oscillator, Sens. Actuators A, 1990, A24, 209.
Magee et al, Surface Acoustic Wave Sensor Methods and Systems, Jan. 23, 2007, United
States Patent, Patent NO. US 7,165,455 B2 .
Magee et al., Torque Measuring Piezoelectric Device and Method, Jan. 20, 2004, United
States Patent, Patent NO. US 6,679,123 B2 .
136
Martin, F., Pulse Mode Shear Horizontal-Surface Acoustic Wave (SH-SAW) System for
Liquid Based Sensing Applications, 2004, Biosensors and Bioelectr., Vol. 19, pp. 627
632.
P. Merewood, A novel ZnO/Si thin-film SAW torque measurement microsystem, Sept.,
2000, Int. Forum on Wave Electron and Its Applications, pp. 14-18, St.-Petersburg.
R. Grossmann, J. Michel, T. Sachs, E. Schrufer, Measurement of mechanical quantities using
quartz sensors, March 1996, Proc. Europ. Freq. Time Forum, pp. 376-381, 5-7.
R.M. White and F.W. Voltmer, Direct Piezoelectric Coupling to Surface Elastic Wave,
1965, Appl. Phy. Lett., Vol. 17 pp.314-316.
Rayleigh, J.W.S., On waves propagated along the plane surface of an elastic solid, 1885,
Proceedings of the London Mathematical Society, Vol. 17, No. 1, pp. 4-11.
Stephen Beeby, Graham Ensell, Michael Kraft, Neil White, MEMS Mechanical Sensors,
2004, Artech House, Inc.
Talbi, A., et al., ZnO/Quartz Structure Potentiality for Surface Acoustic Wave Pressure
Sensor, Sens. Actuators A: Phys., Vol. 128, 2006, pp. 7883.
U. Wolft; F. Schmidt, G. Scholl, V. Magory, Radio Accessible SAW sensors for non-contact
measurement of torque and temperature, 1996, Proc. IEEE Ultrason. Symp.
Weidong Cheng, Yonggui Dong, and Guanping Feng, A Multi - Resolution Wireless Force
Sensing System Based upon a Passive SAW Device, Sept. 2001, IEEE Trans.
Ultrason. Freq. Contr. Vol. 48, NO.5, pp. 1438 - 1441.
7
Shape Measurement by Phase-Stepping
Method Using Multi-Line LEDs
Yoshiharu Morimoto1, Akihiro Masaya1,
Motoharu Fujigaki2 and Daisuke Asai3
1Moire
Institute Inc.,
University,
3Hikari,
Japan
2Wakayama
1. Introduction
Three-dimensional shape measurement is used in various fields such as robotics and
inspection of industry production. It is requested to measure shape with high speed and
high accuracy. It is also requested to make the system compact and low-cost. It is, however,
difficult to satisfy these requests simultaneously for 3D shape measurement. In order to
measure shape in real-time, Kato et al. proposed a method using phase-shifting electronic
moir pattern (Kato et al., 1997). The authors developed integrated phase-shifting method
for real-time shape measurement by a grating projection method using phase-shifting
(Morimoto et al., 1999, Fujigaki & Morimoto, 2003). However, they showed only the phase
distribution of the contour lines.
In almost all conventional shape measurement methods, the optical systems are modelled
and the parameters of the model are obtained with a calibration process using calculation of
the geometrical parameters of optical devices such as positions of lens centres of a camera
and a projector. However, the model cannot contain all of the information of the optical
system such as lens distortion, intensity error of a projected grating, brightness linearity of a
projector and a camera, etc. The less information causes measurement errors. Furthermore,
it is time-consuming to calculate the spatial coordinates using the parameters.
In order to obtain a 3D-shape with grating projection method, the authors previously
proposed whole-space tabulation method (WSTM) (Fujigaki & Morimoto, 2008a, Fujigaki et
al., 2008b, 2009, Morimoto et al., 2009). The relationship between the coordinates and the
phase of the grating recorded at each pixel of a camera is obtained as calibration tables in a
three-dimensional space by experiment beforehand. Therefore the analysis is very fast
because of looking at the calibration tables from the phase information at the pixel without
any complex calculation. It provides fine resolution even when the phase distribution of the
grating is not linear.
The tabulation method is extended to a case of inaccurate phase-shifting of a grating by a
projector using five-line light emitted diode (LED) light sources (Morimoto et al., 2010a,
2010b). In this paper, a grating projector with nine-line LED light sources is newly developed.
The phase-shifting usually uses a grating projector with a high resolution stage or a liquid
138
crystal display (LCD) projector. It is expensive and limits the speed of phase-shifting. The
light-power efficiency of an LED is very high. The size of the light source is very small. It is
easy and very fast to control the power and the switch on/off timing. The nine-line LED light
sources are set in front of a grating. The switch for each LED line of the light sources is switch
on. The phase of the grating shadow is shifted by changing the switches for LEDs into on/off
in synchronization with the phase-shifting and the recording with the camera sequentially.
Using the LED light sources for a grating projector, it is possible to make phase-shifting
without any moving devices. The projector is low cost and high-speed. The authors call the
method phase-stepping method using multi-line LEDs. Even when the positions of the LED
light sources are not so accurate, the error is almost cancelled by using the calibration tables
obtained by the WSTM with the same experimental setup. This method will satisfy for shape
measurement system high-speed, high accuracy, compact and low-cost.
In this paper, the theory and the system of shape measurement using the WSTM and the
phase analysis methods using multi-line LED light-sources are explained. Furthermore,
some experimental results of shape measurements using the system are shown.
2. Phase-shifting method in grating projection method

2.1 Grating projection method
Figure 1 shows a schematic system for grating projection method. A grating with a
cosinusoidal brightness distribution is projected by a projector. The grating projected on an
object is deformed according to the shape of the object. The deformed grating is recorded by
Fig. 1. Schematic view of grating projection system
Shape Measurement by Phase-Stepping Method Using Multi-Line LEDs
139
a camera and analyzed the deformed grating to obtain the shape. As analysis method of the
deformed grating, phase shifting method which is the most popular accurate method, is
used as mentioned next section.
2.2 Phase-shifting method
In shape measurement using grating projection method, phase analysis of the projected
grating provides accurate results. In order to analyze the phase, there are many methods
(Takeda & Mutoh, 1983, Asundi & Zhou., 1999, Sitnik & Kujawinska, 2002, etc.). Phaseshifting method is the most popular method. Let us explain the phase-shifting method.
Figure 2(a) shows phase-shifted grating patterns projected onto an object. Figure 2(b) shows
the brightness (intensity) distributions along the horizontal center line of Fig. 2(a). The
intensity distribution is almost cosinusoidal.
(a)
(b)
Fig. 2. Phase analysis of projected grating using phase-shifting grating images; (a) Phaseshifted grating pattern projected onto object; (b) Brightness distribution along horizontal
center line of Fig. (a)
When a grating with a cosinusoidal brightness distribution is projected or displayed on a
reference plane or an object, and the phase of the grating is shifted N times for one cycle, the
k-th phase-shifted grating images can be expressed as follows:
2
] I b ( x, y )
N
(k 0,1,..., N 1)
I k ( x, y ) I a ( x, y ) cos[ ( x, y ) k
(1)
where Ib(x, y) represents the background brightness in the image, which is insensitive to the
change in phase. Ia(x, y) represents the amplitude of the grating brightness and (x, y) is the
140
initial phase value. The phase distribution of the grating pattern can be obtained as follows
(Srinivasan et al., 1984, 1985).
tan ( x, y )
2
)
N
N 1
2
I ( x, y )cos(k
)
k 0 k
N
N 1
k 0
I k ( x, y )sin (k
(2)
Usually N is selected as 3 or 4. If the number N of the phase-shifted moir patterns is larger,

the phase analysis becomes more accurate. In this study, N =4 and 6 are selected.
3. Whole-Space Tabulation Method (WSTM) (Fujigaki et al., 2008a, 2008b,

2009, Morimoto et al., 2009)
In grating projection method, the coordinates of a point on an object is usually calculated using
the geometric positions of the lenses of a camera and a projector. If the distortion of lenses is
considered, the calculation is complex and time-consuming. Therefore several methods using a
reference plane or reference planes were proposed for accurate shape measurement (Zhou &
Su, 1994, Asundi & Wensen, 1999, Su et al., 2004, Ha et al., 2004, Yen et al., 2006). For real-time
shape measurement, it is required to calculate spatial coordinates from the phase values of the
projected grating in short time. The authors previously proposed the calibration methods
which used two or multiple reference planes (Fujigaki & Morimoto, 1996, Shinke et al., 2001, Ri
et al., 2005). The calibration method was extended to a very large number of planes. All the
relationships between the phase of the projected grating and the spatial coordinates can be
obtained as calibration tables for each pixel of the camera. The method is called 'whole-space
tabulation method (WSTM). Let us explain the method.
Figure 1 also shows a schematic view of a shape measurement system by grating projection
method and explanation of the principle of the calibration method using multiple reference
planes. An LCD panel as a reference plane is set on a linear stage. The LCD panel is covered
with a scattered film. The scattered film functions as a screen when a grating pattern is
projected from the projector onto the reference plane. It is used to make calibration tables on
the relationship between the phase and the z coordinate. The scattered film also functions
as a backward screen when a grating patterns are displayed on the LCD panel to determine
the x and y coordinates.
The LCD reference plane is set vertically to the z-direction. It is translated in the z-direction
step by step. A camera and a projector are arranged and fixed in front of the reference plane.
The grating is projected on the reference plane at first, and it is also projected on an object to
be measured later. By translating the reference plane along the z-axis, a pixel of the camera
records the intensities as images at the points P0, P1, P2 . and PN, on the reference planes
R0, R1, R2. and RN, respectively. And also, by recording the phase-shifted grating images,
the phase distribution along the visual line L of a pixel of the camera is obtained.
In order to obtain the x and y coordinates on the reference plane, the phase-shifted gratings
displayed on the LCD panel are taken by the camera. From these phase-shifted images, the
calibration tables are formed to obtain the x, y and z coordinates from the phase at each
pixel as shown in Figs. 3 and 4.
141
Fig. 3. Schematic view of calibration tables to obtain x, y and z coordinates from phase in
whole space tabulation method
(a)
(b)
(c)
Fig. 4. Calibration tables to obtain x, y and z coordinates from phase at each pixel point of
camera; (a) Table of phase and x; (b) Table of phase and y; (c) Table of phase and z
142
In the measurement procedure, an object is placed between the reference planes R0 and RN.
Phase-shifted gratings are projected onto the object and the phase distributions of the
grating are analyzed from the phase-shifted grating images. The 3-D coordinates at each
pixel are obtained from the phase value in high-speed by referring the calibration tables for
each pixel.
This method is called the WSTM as mentioned already. It excludes the effect of lens
distortion and intensity error of the projected grating in measurement results theoretically.
Tabulation makes short-time measurement possible because the 3D coordinates are obtained
by looking at the calibration tables from the phase at each pixel point of the camera and it
does not require any time-consuming complex calculation.
4. Phase-stepping method by changing light sources

4.1 Geometrical relationship among phase, phase-stepping amount and intensity of
projected grating
Let us explain the principle of the phase-stepping method using multi-line LED light
sources. Figure 5 shows that a grating is projected from the light sources. This figure shows
in the case of four line LED light sources. If each line LED light source is switched on and off
one by one sequentially, the shadow of the grating is moved.
Fig. 5. Shadow of grating is moved when each line LED is switched on and off one by one
sequentially. Phase stepping amount is function of vertical distance from LED light source
in phase-stepping method using multi-line LED light sources
The amount of phase shifts changes with the z-directional distance from the light source to
each point in Fig. 5. The positions with the amounts of phase shift, /2 and are shown as
horizontal straight lines, respectively.
The equations of the geometric relationship among light sources, a grating and an object are
derived using Fig. 6. In this figure, five line LED light sources are shown as an example.
Now, the subscript number n is given to each line LED light source L, and the n-th line light
source is set to Ln (n=0, 1, 2, 3, 4). The light sources are mutually parallel and the five lines
have regular intervals with the pitch l. The plane including the five LED line light sources is
143
Fig. 6. Schematic of grating projection system and object in shape measurement system
using five line LED light sources
called an LED plane where z=0 with the field (x, y). The position of the first LED line L0 is
the origin O of the x, y, and z coordinates. In the LED plane, the x direction is normal to the
LED lines, and the y-direction is the parallel to the LED lines. The z-direction is normal to
the LED plane. The brightness distributions of the five line LEDs are assumed as uniform
and equal in the x- and y-directions in the observation region at fixed z.
A grating plane is parallel to the LED plane, and the distance between the LED plane and
the grating plane is d. The grating lines with pitch p drawn on the grating plane are parallel
to the LED lines. The intersection of the grating plane and the z-axis is set to C and the
distance between C and the origin E, i.e., the center E of the first grating line is set to e.
The transmissivity intensity distribution of the grating at z=d is denoted by the following
equation.
I g a g cos{ } b g a g cos{
2
( x g e )} bg
p
(3)
where is the phase, ag is the amplitude, bg is the intensity of background, xg is the xdirectional coordinate at the grating plane, that is, xg is the x-directional distance from the
point C.
When the light source L0 is switched on and the shadow of the grating is projected at the
position S(x, y, z), the intensity at the position S(x, y, z) is expressed by the following equation.
I 0 ag
d2
2 d
d2
cos{ ( x e )} b g 2
2
p z
z
z
(4)
where it is considered that the intensity at a point is inversely proportional to the square of
the distance z. As shown in Fig. 6, the point G of the grating is projected at the point S on the
144
object. Only the n-th line LED light source Ln is switched on one by one, and it is considered
that the shadow of the grating is projected on the object.
The intensity distribution of the shadow projected through the grating from the light source
Ln has the following distribution shown at z.
I n ag
d2
2 d
d
d2
cos[ {( x e ) nl(1 )}] b g 2 a cos{ n} b
2
p z
z
z
z
(5)
a ag
d2
z2
(6)
b bg
d2
z2
(7)
where
2 d
( x e)
p z
(8)
2l
d
(1 )
p
z
(9)
4.2 Relationship among phase, phase-shift amount and height

When the shift amount is obtained, the z-coordinate is determined using Eq. (9) as
follows.
2ld
2l p
(10)
When the phase is obtained, the z-coordinate is determined using Eq. (8) as follows.
2dx
p 2e
(11)
From Eq. (10) or (11), the z-coordinate is obtained.

4.3 Phase analysis by phase-stepping method by changing light sources
As shown in the case of and in Fig. 5, there are the positions where the phase
shift amounts are 2/N (N is an integer). At the positions, the phase analysis becomes
possible using Eq. (2). At the other positions, there are some errors in the phase value
obtained using Eq. (2). However, if the relationship between the calculated phase and z is
a single-valued function at some region, the phase determination is possible in the region by
combining the WSTM. The height is exactly obtained by looking up the tables of the WSTM
even if the phase shifting method has a little error.
145
5. Experiment
5.1 Experimental setup
In this study, some experiments are performed to confirm the performance of the phasestepping method using line LED light sources. Figure 7 shows the schematic optical system
of the experiment. Figure 8 shows the photograph of the LED board with 9 line LED light
sources we developed.
(a)
(b)
Fig. 7. Optical system used in experiment; (a) Schematic view; (b) Photograph
(a)
(b)
Fig. 8. Photographs of the developed line LED light sources; (a) Board with 9 LED lines and
slanting 6 lines; (b) Photo when switching on a line
146
The LED has nine lines perpendicular to an x-axis, and slanting six lines. Here, the
perpendicular lines are used. The pitch l of the LED lines is 0.5 mm. One line has 30 LED
tips of a 0.35-mm square. A Ronchi grating is used as a grating. The pitch p of the grating is
0.5 mm. The interval d between the LED plane and the grating plane is 30 mm. The
experiments are performed where the base level of z is 180 mm.
5.2 Experimental results
The experimental results were shown in Figs. 9 to 14.
Figure 9 shows the intensity distributions in the x direction in the center of the y direction of
the grating projected on the reference plane when the lighting line LED light sources were
changed. In this figure, only five distributions are shown. .Although there is some random
noise, the intensity distributions are almost cosinusoidal and the phase shift is also
performed at equal intervals. Although this random noise is decreased by time averaging,
any averaging is not used in this study to confirm the performance.
At first, only the first four lines of LED were used for the phase-stepping method and the
phases are calculated using N= 4 in Eq. (2). The phase distribution along the x-axis on the
reference plane at z=180mm is shown in Fig. 10(a). Although the phase distribution is
repeated from to several times, the relationship between z and phase at a point of the
reference plane for a 2 phase interval is shown in Fig. 10(b). The relationship is
considerably curved although the relationship is almost straight line in theoretical
treatment. However, by combining the WSTM, the errors will be cancelled by using the
same optical system when the tables are used. Then the measurement is performed during a
2 phase interval without unwrapping of phase.
In the same manner, the phase distribution in the case of N=6 using the first six lines of LED
is shown in Fig. 11. In both of the distributions in the x-direction and the z-direction, the
relationships are almost linear. Thus, the linearity is different according to the value N.
Fig. 9. Intensity distributions along x-axis of projected grating on reference plane when
changing light sources
(a)
(b)
Fig. 10. Phase distributions along x-direction and z-direction when phase calculation is
performed using N=4; (a) x-directional phase distribution; (b) z-directional phase
distribution
(a)
147
148
(b)
Fig. 11. Phase distributions along x-direction and z-direction when phase calculation is
performed using N=6; (a) x-directional phase distribution; (b) -directional phase distribution
Fig. 12. Height distributions when the height is changed from reference plane at z=180mm
(in case of N=4)
Fig. 13. Height distributions when the height is changed from reference plane at z=180mm
(in case of N=6)
149
(a)
(b)
Fig. 14. Measurement result of mold of face; (a) Height distribution along center line;
(b) Height distribution
Next, height measurement was performed. As the specimen, the reference plane was used.
The results are shown in Figs. 12 and 13 and Tables 1 and 2. The average errors are less than
5 m, and the standard deviations are 11~19 m. Although the standard deviation in the
case of N=6 is better than the one in the case of N=4, both have good accuracy.
As an example of shape measurement, the result of shape measurement of a face model cast
is shown in Fig. 14. Although the measurement value shows unusual values in the edge part
in small areas, almost all over the object is measured by the proposed method.
150
Height
1mm
2mm
3mm
4mm
5mm
6mm
7mm
Average
1.003
2.002
3.002
4.002
5.001
6.001
7.000
Error
0.003
0.002
0.002
0.002
0.001
0.001
0.000
Standard deviation
0.019
0.018
0.017
0.017
0.017
0.015
0.017
Table 1. Results of height measurement (in case of N=4)
Height
1mm
2mm
3mm
4mm
5mm
6mm
7mm
Average
1.005
2.003
3.004
4.002
5.001
6.002
7.000
Error
0.005
0.003
0.004
0.002
0.001
0.002
0.000
Standard deviation
0.013
0.013
0.012
0.012
0.012
0.012
0.011
Table 2. Results of height measurement (in case of N=6)
6. Conclusions
A new shape measurement method called phase-stepping method using multi-line LED
light sources is proposed. A special LED array with nine line LEDs was developed. By using
whole-space tabulation method (WSTM), a system using the multi-line LED light sources
was developed. It could shift the phase of the projected grating easily. The principle and a
system of the shape measurement using the phase stepping method using LEDs were
shown. It was applied to three-dimensional shape measurement, and height has been
measured with sufficient accuracy less than 20 m. It is possible to make a high speed, high
precision, low cost, small size and wide-dynamic range system. It also excludes lens
distortion and intensity error of the projected grating in measurement results theoretically.
7. Acknowledgments
This study was supported by Hyogo COE Program Promotion Project financially. We
appreciate the support.
8. References
Asundi, K. & Zhou, W. (1999a). Mapping Algorithm for 360-deg Profilometry with Time
Delayed Integration Imaging, Optical Engineering, 38, pp. 339-344.
Asundi, A. & Wensen, Z. (1999b). Unified Calibration Technique and its Applications in
Optical Triangular Profilometry, Applied Optics, 38-16, pp. 3556-3561.
Fujigaki M. & Morimoto, Y. (1996). Shape Measurement by Grid Projection Method without
Influence of Aberration of Lenses, Post Conference Proc. of the VIII International
Congress on Experimental Mechanics SEM, pp. 167-172.
151
Fujigaki M. & Morimoto, Y. (2003). Real-time Shape Measurement by Integrated PhaseShifting Method Using Ronchi Grating, Trans. Jpn. Soc. Mech. Eng., (in Japanese), 69685, C, pp. 2374-2380.
Fujigaki M. & Morimoto, Y. (2008a). Shape Measurement with Grating Projection Using
Whole-Space Tabulation Method, Journal of JSEM (in Japanese), 8-4, pp. 92-98.
Fujigaki, M., Takagishi, A., Matui T. & Morimoto, Y. (2008b). Development of Real-Time
Shape Measurement System Using Whole-Space Tabulation Method, SPIE
International Symposium, Proc. SPIE 7066, 706606.
Fujigaki M., Masaya, A., Murakami, R. & Morimoto, Y. (2009). Accuracy Improvement of
Shape Measurement Using Whole-Space Tabulation Method, Proc. of ICEM2009
held in Singapore.
Ha, T., Takaya, Y. & Miyoshi, T. (2004). High-precision On-machine 3D Shape Measurement
Using Hypersurface Calibration Method, Proc. SPIE, 5603, pp. 40-50.
Kato, J., Yamaguchi, I., T. Nakamura & Kuwashima, S. (1997). Video-rate Fringe Analyzer
Based on Phase-shifting Electronic Moire Patterns, Applied Optics, Vol. 36, No. 32,
pp. 8403-8412.
Morimoto, Y., Fujigaki M. & Toda, H. (1999). Real-Time Shape Measurement by Integrated
Phase-Shifting Method, Proc. of SPIE, Vol. 3744, pp. 118-125.
Morimoto, Y., Fujigaki M. & Masaya, A. (2009). Shape Measurement by Grating Projection
and Whole-space Tabulation Method, Proc. of ISOT2009 held in Istanbul
Morimoto, Y., M. Fujigaki, M. Masaya, A. & Amino, Y. (2010a). Shape Measurement by
Whole-space Tabulation Method Using Phase-shifting LED Projector, Proceedings of
International Conference on Advanced Phase Measurement Methods in Optics and
Imaging, Monte Verita, Locarno, Switzerland, 17 to 21 May.
Morimoto, Y., Fujigaki, M. Masaya, A. & OuraY. (2010b). Shape Measurement by Lightstepping Method Using LEDs, ISOT 2010 International Symposium on
Optomechatronic Technologies, 25-27 October 2010, Toronto, Canada
Ri, S., Fujigaki M. & Morimoto, Y. (2005). Phase Reliability Evaluation in Phase-shifting
Method Using Fourier Transform for Shape Measurement, Optical Engineering, 44-8,
083601-1-8.
Shinke, A., Morimoto, Y. & Fujigaki, M. (2001). Development of Accurate Shape
Measurement System Using Multiple Reference Planes, Proceedings of APCFS &
ATEM '01, pp. 811-814.
Sitnik, R. & Kujawinska, M. (2002). Digital Fringe Projection System for Large-volume 360deg Shape Measurement, Optical Engineering, 41-2, pp. 443-449.
Srinivasan, V., Liu, H. C. & Halioua, M. (1984). Automated phase-measuring profilometry of
3-D diffuse objects, Appl. .Opt. 23, pp. 3105-3108.
Srinivasan, V., Liu, H. C. & Halioua, M. (1985). Automated phase-measuring profilometry: a
phase mapping approach,, Appl. Opt. 24, pp. 185-188.
Su, X., Song, W, Cao, Y. & Xiang, I. (2004). Phase-height Mapping and Coordinate
Calibration Simultaneously in Phase-Measuring Profilometry, Optical Engineering,
43, pp. 708-712.
Takeda, M. & Mutoh, K. (1983). Fourier Transform Profilometry for the Automatic
Measurement of 3-D Object Shapes, Applied Optics, 22-24, pp. 3977-3982.
152
Yen, H. N., Tsai D. M. & Yang, J. Y. (2006). Full-field 3-D measurement of solder pastes
using LCD-based Phase Shifting Techniques, IEEE Transactions on Electronics
Packaging Manufacturing, 29-1, pp. 50-57.
Zhou, W-S. & Su, X.-Y. (1994). A Direct Mapping Algorithm for Phase-measuring
Profilometry," Journal of Modern Optics, 41-1, pp. 89-94.
8
Electro-Luminescence Based PressureSensitive Paint System and Its Application to
Flow Field Measurement
Yoshimi Iijima and Hirotaka Sakaue
Japan Aerospace Exploration Agency,

Japan
1. Introduction
Pressure-sensitive paint (PSP) relates a static or oxygen pressure in a testing fluid to a
luminescent signal. It uses a photophysical process of oxygen quenching [Lakowicz]. A PSP
measurement system requires an illumination source to excite the PSP that can be a xenon
lamp, an LED array, and a laser. These are a point illumination that creates a distribution of
the illumination on a PSP coated surface. This distribution results in a pressure-independent
luminescence from the PSP surface. By rationing with a reference image, the pressureindependent luminescence is cancelled [Liu and Sullivan]. The reference image is obtained
under a constant pressure. It is only related to the pressure-independent luminescence
mainly due to the distribution of the illumination and a movement and/or distortion in a
testing article. The rationing method is valid when there is no movement and/or distortion
in the article. If there is such a case, a misalignment of the images may occur that causes a
substantial error to a PSP measurement. In addition to the illumination problem, a PSP in
general has a temperature dependency [Liu and Sullivan]. This also creates an error to a PSP
measurement. For example, a platinum-porphyrin based PSP changes about 1% of the
luminescent signal by 1 C change in temperature. This is equivalent to the 1 kPa change in
the luminescence signal.
Electro-luminescence (EL) gives a surface illumination instead of a point illumination. This
may reduce the misalignment error. An inorganic EL has a potential to spray on a testing
article. This will open our application to provide a testing article with an illumination layer
mounted. Previous studies reported that an EL has a temperature dependency [Airaghi et al,
Schulze et al]. The illumination output of the EL increased with an increase of the
temperature. The temperature dependency is opposite to that of a conventional PSP. The
combined system of the PSP and EL may reduce the overall temperature dependency of the
PSP system.
In this chapter, we introduce an EL-based PSP system. A spectral characterization as well as
a spatial uniformity of an EL is described. The temperature and pressure calibrations of the
system are included. The resultant system is demonstrated to obtain a pressure distribution
created by an impinging jet.
154
2. Background
2.1 Inorganic electro-luminescence
An electro-luminescence (EL) uses a luminescent particle, which is electrically excited to
give an illumination output [Destriau]. An EL can be categorized as an inorganic and
organic EL based on a material of the particle and an illumination mechanism. An inorganic
EL, which is commercially available, uses a phosphor as a luminescent particle. It can be a
powder or a thin-film. A powder-type EL can be applicable to spray on a testing article. Fig.
1 shows a schematic of a powder-type inorganic EL. It consists of multi-layers in addition to
a phosphor layer.
Back electrode
Dielectric layer
Phosphor layer
Transparent electrode
Substrate
Fig. 1. Multi-layer construction of an electro-luminesence (EL)
2.2 Pressure-Sensitive Paint (PSP)
Based on the oxygen quenching, the luminescent signal, I, can be related to a static pressure
by using the Stern-Volmer equation [Liu and Sullivan].
I ref
I
AP BPP
(1)
Where AP and BP are calibration coefficients. As a pressure calibration, BP denotes the

pressure sensitivity, . A high gives more sensitive to the pressure.
d I ref I
dp
BP (%/kPa)
(2)
p pref
A temperature characterization of a PSP can be described as the second order polynomial in

Eq. (3):
I
I ref
cT 0 cT 1T cT 2T 2
(3)
Here, cT0, cT1, and cT2 are calibration constants at atmospheric conditions. The luminescent
signal is denoted as I, and T denotes the temperature. The subscript ref denotes the reference
conditions. The reference conditions were at 100 kPa and 23 C. Throughout the
Pressure-Sensitive Paint System and Its Application to Flow Field Measurement
155
temperature and pressure calibrations as well as the application to the flow measurement,
we used these reference conditions. We defined the temperature dependency, , which is a
slope of the temperature calibration at the reference conditions.
d I I ref
dT
cT 1 2cT 2T ref (%/C)
(4)
T T ref
If the absolute value of is large, it tells us that the change in I over a given temperature
change is also large. This is unfavorable condition as a pressure sensor of a PSP. On the
contrary, zero means the PSP is temperature independent, which is a favorable condition.
3. Development of Electro-Luminescence based Pressure-Sensitive Paint

system
3.1 Characterization setup
Fig. 2 shows the characterization setup for an EL, PSP, and EL-PSP system. We can control
the temperature and pressure inside a test chamber. An EL, PSP, and EL-PSP was placed in
the chamber under controlled temperature and pressure. The EL was excited by a sinusoidal
voltage (Yokogawa, FG300) that creates an AC input. The input was amplified (NF
Electronic Instrument, High Speed Power Amplifier 4055) to give the EL illumination. We
set the illumination frequency at 1 kHz throughout this chapter. A spectrometer
(Hamamatsu Photonics, Photonic Multichannel Analyzer) was used to obtain the
luminescent output related to the wavelength. When obtaining the PSP output, we placed a
high-pass filter of 600 nm in front of the spectrometer to exclude the illumination. The
characterization setup also used a 16-bit CCD camera (Hamamatsu Photonics) to acquire the
PSP intensity. Depending on the pressure dye used for a PSP, we placed a band-pass filter in
front of the camera to acquire the PSP image (section 3.3). As a comparison, a xenon lamp
with a band-pass filter of 400 50 nm was used as a conventional illumination source.
CCD camera
band-pass filter
spectrometer
high-pass filter
EL controller
sample
pressure
controller
test chamber
Fig. 2. Schematic of the characterization setup
temperature
controller
156
3.2 EL characterization
Fig. 3 shows the temperature spectra of the inorganic EL used (Nippon Membrane). A peak
exists at 500 nm with a broad spectrum from 400 to 650 nm. The output was normalized at
the peak illumination at the reference temperature. As we can see, the illumination output
increases with an increase of the temperature.
1.6
50
40
30
23
20
10
Normalize Intensity
1.4
1.2
1.0
0.8
0.6
0.4
0.2
0.0
350
400
450
500
550
Wavelength [nm]
600
650
Fig. 3. The temperature spectra of the inorganic EL used

We integrated the spectrum from 400 to 600 nm to determine the illumination intensity. Fig.
4 (a) and (b) show a temperature and pressure calibration. The temperature calibration was
obtained under atmospheric conditions. For the temperature calibration, Eq. (3) was fitted to
the calibration data. For the pressure calibration, Eq. (1) was fitted to the calibration data.
The values of and were 1.1 %/C and 0 %/kPa, derived from Eqs. (4) and (2),
respectively. As seen in the illumination spectra (Fig. 3), the intensity increased with an
increase of the temperature. As described in section 3.3, the temperature dependency of the
EL was opposite to that of the PSP; the EL increases its intensity with temperature but the
PSP decreases its intensity. On the other hand, the intensity was independent of the
pressure.
Fig. 5 (a) shows the illumination uniformity of the EL. As a comparison, an illumination
image was obtained from the xenon lamp (Fig. 5 (b)). The images were acquired by the CCD
camera with the fixed distance from the illuminated surface. The same area of 30-mm in
square was shown. This includes non-illuminated area, which is shown as dark area in Fig.
5. The EL has the size of 25-mm square. For the xenon lamp, it illuminated the EL surface,
while the EL was switched off. An illumination tip of a xenon lamp was adjusted to place at
the edge of the EL square. The dark area was used as the minimum illumination to
normalize the illumination output. The mean output at the center of the illuminated area
was used as the maximum illumination for the normalization. Here, d denotes the length of
the image area to extract the trend in the illumination in the horizontal axis. Because a xenon
157
lamp is a point illumination source, it showed a non-uniform illumination over the square.
On the other hand, the EL showed a uniform illumination over the square. To compare the
illumination uniformity, cross sectional distributions given in lines 1 through 4 were shown
below each illumination image. The difference from the area averaged value of illumination
was shown. As we can see, the EL showed fairly identical in the distributions, while the
xenon source showed variations in the illumination.
1.6
I/Iref
1.4
1.2
1.0
0.8
0.6
5
15
25
35
Temperature[C]
45
55
(a)
1.2
1.0
Iref/I
0.8
0.6
0.4
0.2
0.0
0
20
40
60
80
Pressure [kPa]
100
120
(b)
Fig. 4. (a). The temperature calibration of the EL used; (b). The pressure calibration of the EL
used
158
(a)
(b)
Fig. 5. (a). Illumination uniformity of the EL; (b). Illumination uniformity of a xenon lamp.
I denotes the difference in the illumination outputs
159
3.3 EL-PSP characterization

To tailor the EL as an illumination source for a PSP system, we need to exclude the overlaid
spectra between the EL and a PSP emission. We used platinum porphyrin (PtTFPP) and
bathophen ruthenium (Ru(dpp)) as a pressure dye in a PSP component, which is commonly
used for a conventional PSP system. We inserted a band-pass filter (Fujifilm, BPB50) to
exclude the spectral overlay. It is a sheet filter made of a tri acetyl cellulose with its thickness
of 90 m. Fig. 6 shows the transmittance of the band-pass filter. This filter cuts off from 580
to 680 nm, which corresponds to the emission of the PSPs for the EL-PSP system.
Transmissivity [%]
100
80
60
40
20
0
400
450
500
550
600
Wavelength [nm]
650
700
750
Fig. 6. Transmittance (%) of a sheet band-pass filter to prevent a spectral overlay between
the EL illumination of PSP emission.
The developed EL-PSP system consists of the EL, the band-pass filter, and a PtTFPP based
PSP. This PSP uses poly-IBM-co-TFEM as a polymer. Similar to PtTFPP, Ru(dpp) based ELPSP system consist of the EL, the same band-pass filter, and a Ru(dpp) based PSP. This PSP
uses RTV118 as a polymer. Overall thickness of the EL-PSP layer was 0.9 mm. In the present
case, an EL-PSP layer was provided. It can be applied on a simple curvature, such as a
cylinder. If a spraying process would be developed for an inorganic EL, it would be possible
to spray the EL-PSP on a complex geometry. Fig. 7 showed the spectral outputs of PtTFPP
based EL-PSP system and Ru(dpp) based EL-PSP system, respectively. The former emitted
at 650 nm peak, while the latter emitted a broad region with a peak at 600 nm. Note that,
due to the sheet filter, the EL illumination did not exist over 580 nm. Because the PSP
emission was on top of this filter, the emission was not influenced. Therefore, the PSP
emission of Ru(dpp) showed its emission below 580 nm.
Fig. 8 shows the temperature calibrations of the EL-PSP systems. We used a band-pass filter
of 650 20 nm in front of the CCD camera to determine the luminescent intensity for
PtTFPP based EL-PSP systems, while a band-pass filter of 600 750 nm for Ru(dpp) based
EL-PSP system. As a comparison, the results of a conventional PSP system were shown.
Here, a conventional PSP system uses the same luminophore and polymer as the EL-PSP
160
EL illumination
PtTFPP emission
Ru(dpp) emission
1.2
Normalize intensity
1.0
0.8
0.6
0.4
0.2
0.0
400
500
600
Wavelength [nm]
700
800
Fig. 7. Luminescent spectra of PtTFPP based EL-PSP system and Ru(dpp) based EL-PSP
system
1.6
EL-PSP PtTFPP
EL-PSP Ru(dpp)
Conventional PSP PtTFPP
Conventional PSP Ru(dpp)
I/Iref
1.4
1.2
1.0
0.8
0.6
5
15
25
35
Temparature[C]
45
55
Fig. 8. The temperature calibration of the EL-PSP systems compared to that of a

conventional PSP system
systems but uses a xenon lamp as an illumination. PtTFPP based EL-PSP system showed of 0.6 %/C, while the corresponding conventional PSP showed -1.3 %/C, respectively. This ELPSP system could reduce by 54%. On the other hand, Ru(dpp) based EL-PSP system showed
of -0.1 %/C, while the corresponding conventional PSP showed -1.2 %/C, respectively.
This EL-PSP system could reduce by 92%. The temperature calibrations tell that the
combination of an EL and PSP greatly influences the temperature dependency.
161
Fig. 9 shows the pressure calibrations of the EL-PSP systems. Similar to the temperature
calibration, the pressure calibrations of conventional PSPs were shown for comparisons. As
we can see from the calibration results, similar trends can be seen for all results. Based on
Eq. (2), of the EL-PSP systems and conventional PSPs were 0.8 %/kPa.
1.2
1.0
Iref/I
0.8
0.6
EL-PSP PtTFPP
EL-PSP Ru(dpp)
Conventional PSP PtTFPP
Conventional PSP Ru(dpp
0.4
0.2
0.0
0
20
40
60
80
Pressure [kPa]
100
120
Fig. 9. The pressure calibration of the EL-PSP systems compared to that of a conventional
PSP system
4. Demonstration
4.1 Demonstration setup
We applied the developed system to a sonic jet impingement for comparison. A schematic of
the experimental setup is shown in Fig. 10. A sonic jet from a 2-mm orifice was impinged on
the EL-PSP surface. The same camera and optical filter in calibrations (section 3.1) were used
to acquire the luminescent image from the system. The camera was placed above the system
as shown in Fig. 10. To discuss a temperature dependency of the measurement, we acquired
two reference images: one at the ambient temperature of 23 C, and the other at a higher
temperature of 30 C. These reference images were acquired when there was no flow that
provided the constant pressure at 100 kPa over the EL-PSP surface. As a comparison, a
conventional PSP system was used to acquire the same flow. In this system, a xenon lamp
through 400 50 nm instead of the EL illumination was used as an excitation. The EL was
switched off during this measurement.
4.2 Global pressure measurement and discussion
Fig. 11 (a) and (b) show pressure maps obtained from PtTFPP based and Ru(dpp) based ELPSP systems. A priori pressure calibration obtained in Fig. 9 was used to convert the
luminescent image to the pressure map. The ambient reference was used as a reference image.
It showed a clear diamond shock pattern created by the sonic jet impingement. We can notice
small scratch-like spots in the pressure maps. These were created during the preparation
process of the EL-PSP layer, which can be improved to layer the EL, the sheet filter, and PSP.
162
16bit CCD camera + *BPF

*BPF 630-670nm
(PtTFPP based system)
*BPF 600-750nm
(Ru(dpp) based system)
peltier device
sonic jet nozzle
(2-mm orifice)
[room temperature Iref ]
EL-PSP
2525mm
[heating Iref by 7 degrees ]
image area shown

18 18mm
EL controller
Fig. 10. Schematic description of the demonstration setup
P[kPa]
PtTFPP based
based EL-PSP
EL-PSP
PtTFPP
2mm
(a)
163
P[kPa]
Ru(dpp) based EL-PSP
2mm
(b)
Fig. 11. Pressure maps created by a sonic jet impingement obtained from (a) PtTFPP based
EL-PSP system and (b) Ru(dpp) based EL-PSP system
Fig. 12 (a) through (d) show cross-sectional pressure distributions along the centerline of the
jet impinged surface. The location of the centerline was shown as a solid line in Fig. 11. Here
L = 0 denotes the left edge of the solid line in Fig. 11. Fig. 12 (a) showed the distribution
from PtTFPP based EL-PSP system, and Fig. 12 (b) showed the results from the conventional
PSP system, respectively. Similarly, Fig. 12 (c) and (d) are a comparison between Ru(dpp)
based EL-PSP system and Ru(dpp) based conventional PSP. The scratch-like spots discussed
in Fig. 11 was obvious for the pressure distributions from Ru(dpp) based EL-PSP system.
For each figure, the pressure distributions obtained from the pressure map results under the
ambient and the high temperature reference were shown. As increasing the temperature, the
luminescent intensity was reduced (Fig. 8). The high temperature reference, therefore, gave
lower luminescent ratio in the left hand side of Eq. (1). This results in an underestimated
pressure. We can see that both systems showed the underestimation of the pressure
distributions. For the EL-PSP system, the pressure distribution showed smaller change in
the pressure measurement compared to that of the conventional PSP system. The difference
in the pressure measurement was an average of 3 kPa for PtTFPP based EL-PSP system,
while the corresponding conventional PSP was an average of 12 kPa. There was a 75%
improvement in the temperature dependency for the pressure measurement obtained from
PtTFPP based EL-PSP system. Because Ru(dpp) based EL-PSP system showed very small
temperature dependency (-0.1 %/C, section 3.3), pressure distributions obtained from the
different references were almost identical. The difference in the pressure measurement was
an average of 1 kPa. On the other hand, Ru(dpp) based conventional PSP showed the
difference of an average of 15 kPa. For Ru(dpp) based EL-PSP system, there was a 93%
improvement in the temperature dependency.
164
EL-PSP system
150
150
ref image of ambient temperature
125
125
P
[kPa]
P [kPa]
mean 3kPa error
100
100
75
75
50
50
ref image of high temperature
00
55
10
10
10
[mm]
LL [mm]
15
15
(a)
150
150
PtTFPP based Conventional PSP
125
125
P[kPa]
[kPa]
P
mean 12kPa error
100
100
75
75
50
50
10
10
[mm]
LL [mm]
15
15
(b)
EL-PSP system
150
150
125
125
P
[kPa]
P [kPa]
mean 1kPa error
100
100
75
75
50
50
00
55
10
10
10
L
[mm]
L [mm]
(c)
15
15
150
150
Ru(dpp) based Conventional PSP mean 15kPa error
125
125
P[kPa]
[kPa]
P
165
100
100
75
75
50
50
10
10
[mm]
LL [mm]
15
15
15
(d)
Fig. 12. Cross-sectional pressures obtained from (a) PtTFPP based EL-PSP system, (b)
PtTFPP based conventional PSP system, (c) Ru(dpp) based EL-PSP system, and (d) Ru(dpp)
based conventional PSP system. Pressures under ambient temperature and a high
temperature were shown for each figure
The reference image was shifted about 2 mm towards the downstream to provide a
misalignment of the image. This changed the distribution in the pressure independent
luminescence between the reference and jet impinged images. The shifted reference was
aligned by an image processing as commonly used for a PSP measurement. Fig. 13 (a) and (b)
show the pressure maps obtained from PtTFPP based EL-PSP system and conventional PSP
system with the image alignment process, respectively. Because the 2-mm width of the
upstream area was not acquired it was shown as a shaded area. The EL-PSP system can
provide a pressure distribution as seen in Fig. 11 (a). This tells us that a uniform illumination
of the EL can reduce the pressure-independent luminescence and extract the pressure
distribution. On the other hand, a point illumination of the conventional PSP system showed a
distribution which is not caused by the pressure (Fig. 11 (a)). The processed image could not
remove the pressure independent luminescence caused by a non-uniform illumination. This
was because the distribution in the pressure independent luminescence obtained by the point
illumination was different from the reference and jet impinged images. By using the EL as a
surface illumination, the EL-PSP system greatly reduced the misalignment error.
5. Conclusion
We introduced a pressure-sensitive paint (PSP) measurement system based on an electroluminescence (EL) as a surface illumination. This consisted of an inorganic EL as the
illumination, a band-pass filter, and a PSP. The band-pass filter, which passed below 580
nm, was used to separate an overlay of the EL illumination and the PSP emission. In this
champter, two types of PSPs were used to construct the EL-PSP system. One is platinum
porphyrin (PtTFPP) based PSP, which gave the emission at 650-nm peak. The other is
bathophen ruthenium (Ru(dpp) based PSP, which gave a braod emission with a peak at 600
nm. The EL showed an opposite temperature dependency to that of the PSPs; the
illumination intensity of the EL increased with increase temperature with the temperature
dependency of 1.1 %/C. The EL gave a uniform illumination compared to that of a point
illumination source such as a xenon lamp.
166
CCD camera
EL-PSP
P[kPa]
Nozzle
10
L [mm]
(a)
11
12
14
16
18
xenon lamp
CCD camera
PSP
P[kPa]
Nozzle
10
11
12
14
16
18
L [mm]
(b)
Fig. 13. Pressure maps processed from a misaligned reference image obtained from (a)
PtTFPP based EL-PSP system and (b) PtTFPP based conventional PSP system
167
168
A combination of EL and PSP greatly influenced the temperature dependency, while a

minimal influence could be seen for the pressure sensitivity. Under atmospheric conditions,
PtTFPP based EL-PSP system reduced the temperature dependency by 54% compared to
that of a conventional PSP system. For Ru(dpp) based EL-PSP system, the temperature
dependency was greatly reduced by 92% compared to that of a conventional PSP system.
Both EL-PSP systems showed the pressure sensitivity of 0.8 %/kPa, which was the same for
conventional PSP systems.
An application of the EL-PSP systems to a sonic jet impingement showed that the systems
demonstrated the reduction of the temperature dependency compared to that of the
conventional PSP system. The temperature dependency was reduced by 75% in a pressure
measurement for PtTFPP based EL-PSP system, while by a reduction of 93% was obtained
for Ru(dpp) based EL-PSP system. Because of a uniform illumination, the EL-PSP system
could reduce an image misalignment error compared to that of the conventional PSP
system.
6. References
B. Schulze, C. Klein, ICIASF05, Sendai, Japan, August 29-September 1, 274 282, (2005)
G. Destriau, J. Chem. Phys., 33, 587, 1936
J. R. Lakowicz, Principles of Fluorescence Spectroscopy, Kluwer Academic/Plenum
Publishers, New York, USA, 1999, Chapter 1
Liu, T. & Sullivan, J. P. (2004). Pressure and Temperature Sensitive Paints, Springer, ISBN 3-54022241-3, Heidelberg, Germany (Chapter 1 and Chapter 3)
S. Airaghi, T. Rsgen, M. Guille, Proceedings of 11th International Symposium on Flow
Visualization, University of Notre Dame, Notre Dame, Indiana, USA, paper id
number 198, Aug. 9 12, (2004)
9
Computer-Based Measurement System
for Complex Investigation of Shape Memory
Alloy Actuators Behavior
Marek Kciuk and Grzegorz Kapyta
Silesian University of Technology, Faculty of Electrical Engineering,

Department of Mechatronics,
Poland
1. Introduction
1.1 Basic information
Shape Memory Alloys belong to group of so called SMART materials, which became more
and more popular last years. The smartness of that kind of materials is defined as possibility
of its particular properties stimulation using physical or chemical way. Shape Memory
Alloy (SMA) is able to change its size (shape) when temperature is changing. Shape Memory
Effect (SME) is macroscopic effect of martensitic transition which overlap inside the
material. There are two characteristic states of SMA: low temperature martensitic state
and high temperature austenitic state. Material can be easily plastically deformed by
relatively small mechanical force in martensitic state. In opposite, material is strong and
very hard to deform in austenitic state - it is only elastic deformation viable. During
martensitic transition internal crystal order changes, which causes change of materials
properties. Material returning to its natural shape during reverse martensitic transition
(from austenitic to martensitic state) can put pressure or perform mechanical work.
Complex behavior of SMA materials can be illustrated by thermo-mechanical characteristic
(Fig.1).
Fig. 1. Thermo-mechanical characteristic of Shape Memory Alloy [1].
170
Martensitic transition is not isothermal transition. The characteristic is highly nonlinear and
exhibit permanent hysteresis loop. There are four important and specific temperature
points: MS and Mf are direct martensitic transition start and finish temperature, while AS and
Af are reverse martensitic transition start and finish temperature. The highest temperature
(Af) means that all the material is already in austenitic phase; the lowest temperature (Mf)
means that all the material is in martensitic phase. is measure of percentage content of
molecules in martensitic phase in the volume of the material. For all the temperatures
between Mf and Af the content martensitic fraction () of material is undefined.
Fig. 2. Shape of SMA material via bias force.

The bias force is important for shape and for martensitic transition parameters. If there is no
bias force the macroscopic shape in martensitic state is the same as in austenitic state. If
there is a bias force martensitic shape is deformed and is different than austenitic shape. It is
presented in Fig. 2.
Important temperatures of martensitic transitions strongly depend on mechanical bias force.
The higher bias force is, the higher important temperatures are (there is linear relationship
see Fig.3). Characteristic temperatures under no bias force are marked with 0 subscript.
Fig. 3. Characteristic temperatures of martensitic transition via bias force.

Therefore there are two possible ways of transition - one stimulated by temperature, second
- stimulated by bias force (respectively blue and red arrows in Fig.2). Temperature
stimulated transformation enables to achieve three types of shape: austenitic shape (natural
shape, ordered structure of material), martensitic twinned shape (natural shape, unordered

for Complex Investigation of Shape Memory Alloy Actuators Behavior
171
structure of material), martensitic deformed shape (deformed shape, unordered structure of

material. Bias force stimulated transformation is isothermal and it starts with low bias force
and temperature above A0f. Increased mechanical force causes that characteristic
temperatures shifts. It is so called superelasticity and enables to achieve only two types of
shape: austenitic shape and deformed martensitic shape. This phenomena is presented in
Fig.4.
Fig. 4. Superelasticity effect of SMA material.

1.2 Modeling of Shape Memory Alloys
SMAs are already well known and frequently used materials. Different types of
mathematical models describe SME behavior. Most popular is modeling based on
thermodynamical laws, using different types of free energy equations i.e. Helmholtzs,
Gibbs, or Landaua Devonshiers [3,6]. There are also mezoscopic models (describing
groups of molecules), micromechanical models (describing molecules behavior) and
macroscopic models, describing whole material. Ikutas model describes the state of material
by two elements: elastic element and suspension element -the factors are changing with
phase changes. There are several types of Ikutas models (i.e. two layers model, multilayer
model, running layers model etc.) [4,5].
Thermodynamical models are the most popular but they are very difficult to implement.
They need a lot of parameters and it is hard to use them in control systems.
1.3 Applications
There are two main application groups: SMA works as a sensor or as an actuator. There is
also third group of applications, which combine both previously mentioned together.
SMAs are most frequently used as temperature sensors in household, industry devices or air
conditioning. Natural hysteresis loop is useful advantage for these kinds of sensors.
Threshold temperature mainly depends on material. It is possible to adjust this temperature
in some range by mechanical bias force changing.
SMA are also used as active (controlled) or passive (not controlled) actuators. Passive
actuators exploit the phenomenon of superelasticity. They are used in civil engineering to
protect buildings bridges etc. (e.g. as seismic protection). Passive SMA actuators can
suppress vibrations. See Fig.5.
172
Fig. 5. Comparison of stress distribution of protected and not protected building [15].
Active SMA actuators are popular in mechatronic applications as unconventional drives.
They are light-weight actuators with high power to weight ratio. They are able to replace
sophisticated electro-mechanical devices in simple tasks. They generate linear displacement
in direct way. Sparkles work ensures the safety of work in flammable and wet areas. It is
also possible to build light-weight robots (or robotic tools) actuated by SMA.
Unconventional robotic hand driven by SMAs is described in [7]. Two types of
unconventional mobile robots (snake robot and net robot) are presented in [8, 9]. It is
possible to build micro robots made in MEMS technology, actuated by SMAs [10].
2. Measuring stand
Measuring stand for obtaining the complex SMA characteristics was designed and built in
Department of Mechatronics at Silesian University of Technology. The stand allows for
measurements of static as well as dynamic characteristics of SMA wires. Multiparameter
simultaneous measurement enables comparison of different characteristics and allows for
assessment of impact of particular parameters on the studied phenomena. Main
assumptions for measuring stand project were:
-
resistive heating of wire with current constraint,

possibility of mechanical load changing,
temperature measurement using infrared camera,
optical measurement of wire displacement,
simultaneous measurement of current, voltage, wire displacement and temperature,
maximum wire length 900mm,
minimal mechanical load 75g,
maximum wire displacement 80 mm,
automatic measurement process with data acquisition system.
Measurement system presented in this paper allows to investigate complex electro-thermomechanical characteristics. It is possible to measure electro-mechanical characteristics
(length of actuator vs. wire current L = f(I), wire resistance vs. length of actuator R = f(L),

173
etc.), thermo-mechanical characteristic (wire length vs. temperature of actuator L = f(T)),

electro-thermal characteristic (temperature of actuator vs. supplying current T = f(I)), etc.
The idea of measurement system is described in Fig.6.
Fig. 6. Idea of measurement system.

There are six main modules: mechanical construction (M) with bias force (F), supply
module, and measurement modules: current module (A), voltage module (V), temperature
module (T), length changing measurement module (L). Measurement stand is controlled
via PC computer in LabVIEW environment. Detailed construction of each module is
described below.
Mechanical construction: Design of a stand (Fig.7) allows to place SMA wire (1) between
stiff crosspiece (2) and moving piston (3). Gravitational mechanical load can be increased by
using additional masses and allows for constant load during measurement process. Base (4),
frame (5), and the crosspiece (2) are made of aluminum profiles. Infrared camera (6) is
mounted to the base at constant level but it can be shifted horizontally to allow picture
sharpness regulation. Electrical circuit is electrically separated from aluminum base by PCV
distances. To improve the quality of infrared measurement part of the stand was covered
with black cardboard sleeve. In window of Fig.7a. you can see also displacement sensor
placed in the bottom part of stand.
Heat supply is done in electrical way, hence SMA wire is a good conductor. According to
the Joule-Lenz Law the wire temperature rise is nearly proportional to the square of flowing
current. To set the constant temperature in SMA wire the supply system should allow for
current stabilization. The power supply PSH-3620A (36V, 20A) was chosen to feed the
system. It allows for current as well as for voltage stabilization. Stabilization is switched
automatically according to Ohms law. Measurement stand is closed inside black cartoon
shield which guarantees steady air conditions and is helpful for temperature measurements
using infrared camera.
Measurement of electrical parameters is done by three precise multimeters Rigol DM3052. It
allows for simultaneous measurement of current and voltage of SMA actuator and voltage
signal of displacement sensor. Data acquisition and control signals are realized by PC
computer. All elements of the measuring system are connected and synchronized by GPIB
network (IEEE 488.2 standard). Control program for static measurements was written in G
language (LabVIEW environment). For dynamic measurements data recording is performed
using digital oscilloscope Tektronix MSO 2024 equipped with four separated channels and
digital filter.
174
(a)
(b)
Fig. 7. Measuring stand: (a)real view; (b)3D model

175
Temperature measurement is realized using infrared camera Flir A325. It has resolution of
320x240 pixels and can record pictures up to 60Hz. For measuring temperature of very thin
wires (50-500m) special macroscopic lens (closeup 1x) were used. Its surface resolution is
25m. Camera signal is send to PC computer by Gigabit Ethernet connection (with
maximum speed of 1GB/s). IR camera software finds in the image point of the highest
temperature and acquires it as the actual temperature of the wire. Additionally, ambient
temperature is measured all the time by digital thermometer chip DS18B20, however its
result is only displayed but not recorded. Fig.8 presents infrared camera view for SMA wire
with 308m diameter.
Fig. 8. Infrared camera view for SMA wire with 308m diameter.
Displacement measurement module is working in two steps. First displacement-voltage
converter converts displacement into voltage signal, next voltage signal is measured by
multimeter. Module consists of optical sensor (converting piston movement into series of
pulses), pulses counter, PWM signal generator (PWM signal fulfillment is proportional to
number of pulses), lowpass filter and buffer for conditioning signal. Amplitude of
converters external voltage is measured by digital multimeter Rigol DM3052. Movement
detection is realized by integrated quadrature optical sensor H9720 and measuring tape
with 85 m resolution scale. Quadrature optical sensor H9720 detects direction of movement
based on two geometrically shifted signals (A and B). Pulse counting and PWM generating
is realized by 8-bit microcontroller AVR ATmega88. Microcontroller count pulses from one
input while second input is used only for direction detection. Pulse counting is realized by
detection of (both) signal ramps using external interrupt INT. When ramp is detected both
inputs are compared. If their signals have the same value - counter is decreased, if not - it is
increased. Block diagram presenting displacement-voltage converter is shown in Fig.9b
while idea of pulse counting (with direction detection) is shown in Fig.10.
176
Fig. 9(a). Block diagram of Q9720 sensor. [11]
Fig. 9(b). Block diagram of displacement-voltage converter.

177
In
Dir
Wy
t
Fig. 10. The idea of pulse counting.
Sensor (S) generate series of pulses (LS) which are counted by counter (C1) and compared in
comparator (K) with number of generator (G) pulses (LG) counted by counter (C2). Counter
(C2) works in one direction only the value is increased to overrun the counter. Time
needed to overflow this counter determines period of generated PWM signal. PWM signal
frequency is 3,5 kHz. Unless LG < LS in comparator its output is set to high in other case it
is low. Sensor and external microcontroller interrupt are working asynchronously. Output
of pulse counter (LS) is buffered to avoid errors caused by unpredictable changes of counted
value. Changing of number of pulses in buffer is possible only in moment of overflow of
counter L2 (Strobe is the signal controlling this operation). This solution is hardwareimplemented in timers built-in to AVR microcontrollers. The role of lowpass RC filter
connected to output of PWM generator is to convert digital PWM signal into analog
(average) signal. Cutoff frequency of the filter is CF=384Hz. Output buffer (operational
amplifier with k=1) is conditioning signal and assure high impedance of filter load.
Control software was written in G language (LabVIEW environment). It enables to control
wide variety of measurement instruments and protocols. The most popular available
protocols are RS-232, USB, FireWire, LAN, GPIB, etc.). All measurement instruments,
instead of IR camera which use Giga Ethernet, communicate each other by GPIB. One
communication protocol enables synchronization between instruments.
Measurement scheme is divided into two loops. First loop increase heating current from
minimum to maximum value with specified step; second loop decreases heating current
from maximum to minimum value with the same step (See Fig.11):

current increasing
(1)
current decreasing
where:
i(j) step current,
j number of step jN,
N number of steps,
I current step difference.
Values: N, j, I are set by the user. There are ten measurement points per each step. Time
between consecutive steps is set as step delay. This two direction measurement process
allows for obtaining characteristics of both - direct and reverse transitions.
178
Fig. 11. The idea of measurement process.

Measurement process is triggered simultaneously using send to group command.
Registered values of each multimeter are then acquired in serial way. Measurement
algorithm of parallel triggering and serial acquisition is presented in Fig.12. Controlling
algorithm is shown in Fig.13.
Fig. 12. Data acquisition subprogram with parallel triggering.

179
START
Set measurement parameters
Communication with IR camera
Instruments configuration
Image file create
Finish
sesion?
No
Yes
i(j) = I min + j I
Yes
All
measurement steps
done?
No
Paraller measurement trigger
Image aquisition
Read current value
Highest temp. finding
Read voltage value
Temperature calculation
Read displacement value

step delay
Sesion - current increasing
Save data file
i(j) = Imin + (N- j) I

Sesion - current decreasing
Save data file
IR camera close connection
END
Fig. 13. Measurement algorithm.
180
3. Measurements accuracy
Supply current and measurements taken in the stand are executed with a specified accuracy.
In the case of supply module, accuracy is related to heating current stability. Analysis of
measurement accuracy of the measuring system, built with manufactured components
(power supply, multimeters), is associated with the analysis of accuracy of each instrument.
Accuracy of all instruments are presented in Tab. 1.
Instrument
Power supply
Multimeter
Measurement range
20A
Resolution
200mV DCV
2V DCV
20V DCV
200mA DCI
1A DCI
Accuracy
0,2% + 10mA
10mA
0,003 + 0,003
0,002 + 0,0006
0,002 + 0,0004
0,02 + 0,002
0,02 + 0,016
Table 1. Accuracy of measuring instruments.

Heating-current stability is checked by the multimeter. Amendment for current stability is
done according to following equation:

(2)
where:
Iset current set by control unit of power supply.
Current measurement accuracy is 1mA.
Accuracy of SMA displacement consist of two parts. The first part is connected with
displacement-voltage converter. The second one is connected with multimeter accuracy.
Accuracy of displacement-voltage converter consist of accuracy of measurement tape
85m and accuracy of PWM generator. 8-bits PWM generator has 256 steps. One step is
equal to 6,3mV. Accuracy of multimeter (for this measurement range) is 1mV, so it is much
higher. Resulting accuracy of displacement measurement is 1 step of measuring tape, it is
equal to 0,17mm.
4. Results and analyses

Designed measurement system allows for wide research program. All measurement results
presented below concern one type of actuator. It is F2000, its nominal load is 2kg.
Some example results showing interesting characteristics and obtained relationships are
shown below. All characteristics were measured at constant conditions of heat dissipation.
Ambient temperature was 252 Celsius degrees. Cartoon shield avoided external air
movement around the tested element. One static measurements cycle took approximately
five hours (each step took nearby three minutes - depending on wires diameter). Each of
dynamic measurements was done in approximately ninety seconds. Measurements is
finished after reaching steady state, which is defined as no change in displacement over a
period of time longer than ten seconds.

181
4.1 Steady-state characteristics
Displacement dL, mm
Example result of steady-state electro-mechanical characteristic is presented in Fig.14.
L = f(I), F = 1,075 kg
20
18
16
14
12
10
8
6
4
2
0
0
0.2
0.4
0.6
0.8
meas 1,
activation
meas 1,
deactivation
meas 2,
activation
meas 2,
deactivation
meas 3,
activation
meas 3,
deactivation
Current I, A
Fig. 14. Steady-state electro-mechanical characteristic.
There are three measurements called meas 1, meas 2, meas 3, increasing current
procedure is called activation, decreasing current procedure is called deactivation.
Results of second and third measurement process are almost identical. Repeatability of
measurement results has been analyzed. Difference between first and other results is
connected with hysteresis phenomenon and history of actuator. Based on the test
measurements the complete measurement procedure was determined. Full measurement
procedure is presented below:
-
Set the SMA actuator in the measurement stand,

Load SMA actuator with proper mechanical bias,
Calibration of actuator,
Perform series of measurements.
Calibration is the process of rapid activation and deactivation of mechanically loaded

actuator performed three times. It causes history cleaning before proper measurement
process. Displacement measuring module is reset after calibration process. Example results
obtained using developed measurement procedure are shown in Fig.15.
During measurements the question raised what if to reverse temperature change before
the transition is complete. So incomplete hysteresis loop measurements were done as well.
Exemplary results are presented in Fig.16. There are results for different maximum heating
currents changing in range of 0,5A to 0,8A. Full hysteresis loop is achieved for current value
equal to 0,8A it is the reference measurement. Transition starts when heating current value
is about 0,4A. Each measurement was done three times. Time of martensitic transformation
is independent of the rate of change of the factors causing it [6]. It means that you cannot
accelerate the transition by increasing temperature (current). So process of activation
182
Displacement L, mm
performs identically each time, but halting the supply of heat energy needed to activate the
actuator, stops, and even change the direction of transition.
L = f(I), F=1,4kg
18
16
14
12
10
8
6
4
2
0
meas 1,
activation
meas 1,
deactivatio
n
meas 2,
activation
0
0.2
0.4
Current I, A
0.6
0.8
Fig. 15. Steady-state electro-mechanical characteristics acquired using developed

measurement procedure (with actuator calibration).
L = f(I), F=0,5kg
16
14
Displacement L, mm
12
10
8
6
4
2
0
0
0.2
0.4
Current I, A
0.6
0.8
Imax = 0,5A,
activation
Imax = 0,5A,
deactivation
Imax = 0,6A,
activation
Imax = 0,6A,
deactivation
Imax = 0,7A,
activation
Imax = 0,7A,
deactivation
Imax = 0,8A,
activation
Imax = 0,8A,
deactivation
Fig. 16. Incomplete hysteresis loop measurements.

The next important characteristic of SMA actuator is load characteristic. Exemplary load
characteristic is presented in Fig.17. According to the theory, increasing the mechanical load
causes a linear increase of characteristic temperatures of martensitic transition while
maintaining a constant width of the hysteresis loop.

L = f(I), F=const
20
Displacement L, mm
183
F=0,6kg,
activation
F=0,6kg,
deactivatio
n
F=1,2kg,
activation
15
10
F=1,2kg,
deactivatio
n
F=1,6kg,
activation
0
0
0.2
0.4
0.6
Current I, A
0.8
Fig. 17. Load influence on the process of martensitic transition.

It can be easily observed that higher mechanical load causes higher displacement. It is
natural phenomenon connected with plastic deformation in martensitic state. Width of
hysteresis loop changes according to mechanical load. For example width of hysteresis loop
for 0,6kg is 0,2A (0,52A-0,32A), while it is 0,12A (0,6A-0,48A) for bias force equal to 1,6kg
and 2,0 kg. It causes not linear change of important temperatures of martensitic transition.
Next interesting characteristic is electro-thermal characteristic showing relation between
current and wire temperature (for constant ambient temperature). It is shown in Fig.18.
T = f(I), F = const
Wire temperature T, OC
120
F = 2,0kg,
activation
100
80
F = 2,0kg,
deactivation
60
40
20
0
0.2
0.4
0.6
0.8
Current I, A
Fig. 18. Temperature vs. current characteristic, for constant ambient temperature.
184
Temperature characteristics of the martensitic state is a quadratic function of current. In the

case of austenite, the situation is similar. A small number of measurement points in
austenite state prevents the determination of the equation function. The most interesting
range is the martensitic transition phase. There are different cases of temperature behavior
in this state. Temperature is constant or even decreases. This is caused by consumption of
thermal energy by transition process. Similar results were mentioned in [13].
Very useful characteristic is resistance as a function of heating current. In some applications
the characteristics of resistance as a function of the length of the actuator are used to estimate
the actual length of the actuator [12]. There are two types of resistance characteristics. First of
them - resistance in function of heating current - is presented in Fig.19.
R = f(I), F = const
Resistance R,
3.7
3.5
F = 0,6kg,
activation
3.3
F = 0,6kg,
deactivation
3.1
F = 2,0kg,
activation
2.9
F = 2,0kg,
deactivation
2.7
2.5
0
0.2
0.4
Current I, A
0.6
0.8
Fig. 19. Resistance vs. Current.

It is easy to notice the difference of characteristics for small and big mechanical loads (close
to nominal bias force). The general rule is better noticeable for the deactivation process
(direct martensitic transition). Characteristics shift toward higher currents with increasing
mechanical load. This is a trend similar to the phenomenon of characteristic temperatures
rising with increasing mechanical load.
Another phenomenon that can be observed is linear change of resistance with extending the
wire in activation process (Fig.20). It is independent of mechanical load while for
deactivation the function is linear only for mechanical load close to the nominal load. This
phenomenon allows for quite accurate displacement approximation using measurements of
resistance.
Simultaneous multiparameter measurement enables to plot different characteristics and a
comprehensive assessment of the impact of occurring phenomena. Fig.21. shows the
characteristics of displacement and temperature as a function of heating current.
The most energy consuming part of the martensitic transition is finish of activation and start
of deactivation. It can happen (as can be seen in above picture) that the wire temperature
decreases in spite of constant delivering of energy.

185
R = f(L), F = const
Resistance R,
3.7
3.5
F = 0,6kg,
activation
3.3
F = 0,6kg,
deactivation
3.1
F = 2,0kg,
activation
2.9
F =2,0kg,
deactivation
2.7
360
365
370
375
380
385
Length of actuator L, mm
Fig. 20. Resistance vs. length of actuator.
D
i
s
p
l
a
c
e
m
e
n
L = f(I), T = f(I), F = 2,0kg
1.2
dL activation
t 1
dL deactivation
0.8
L 0.6
,
0.4
Temp activation
m0.2
m
0
0
0.2
0.4
Current I,A
0.6
140
T
120 e
100 m
p T
80
e ,
60 r
O
40 a
C
t
20 u
r
0
0.8
e
Fig. 21. Displacement and temperature characteristic vs. heating current.

4.2 Dynamic characteristics
Dynamic characteristics was acquired for two cases switch on and switch off the heating
current. In the first case dynamics of activating process is measured, in the second dynamics of deactivating. Different mechanical bias forces and different current values have
been considered. Example results are given in Fig. 22 25.
186
Displacement L, mm
L = f(t), F = const, IMAX = 0,8A

10
F = 0,5kg
F = 0,7kg
F = 1,3kg
0
0
10
15
Time t, s
20
25
30
Fig. 22. Activation process for different mechanical bias forces.
Displacement L, mm
L = f(t), F = const, IMAX = 0,8A

10
F = 0,5kg
5
F = 0,7kg
F = 1,3kg
0
0
20
40Time t, s 60
80
100
Displacement L, mm
Fig. 23. Deactivation process for different mechanical bias forces.
L = f(t), F = 1,3 kg
15
10
I = 0,5 A
5
I = 0,6 A
I = 0,8 A
0
0
10
20
30
Time t, s
40
Fig. 24. Activation process for different heating currents.
50
60

L = f(t), F = 1,3 kg
15
Diplacement L, mm
187
10
I = 0,5 A
5
I = 0,6 A
I = 0,8 A
0
0
10
20
30
Time t, s
40
50
60
Fig. 25. Deactivation process for different heating currents.

Deactivation characteristics in Fig.23. are the result of unconstrained SMA actuator cooling.
It is easy to notice the difference between time of activation and deactivation of the actuator.
The higher force is, the faster actuator recovers. There is small difference in activation time,
so mechanical load does not significantly affect the activation time but influence final
displacement value.
Heating current value significantly influences the speed of the activation process. Current
value 0,8A causes full activation, lower values does not. There is not big difference in final
displacement between heating current values 0,6A and 0,8A, but time is important factor.
The slope of the deactivation curve does not depend on the starting point but is strongly
connected with conditions of heat dissipation.
5. Conclusions
Designed measurement stand can be very helpful in research of Shape Memory Effect. It
allows for simultaneous multiparameter measurement which enables to draw various
characteristics of SMA wire actuators. Presented characteristics prove the utility of designed
measurement systems. It is possible to compare different characteristics of the SMA actuator
for one forced parameter (e.g. comparison of the displacement and wire temperature as a
function of heating current (Fig.21)). Using some dependences that occur among certain
physical quantities it is also possible to calculate some additional parameters and draw
additional characteristics (e.g. resistance as function of current or actuator length (Fig.18 and
Fig.19)).
Design of the stand allows for performance of reach research program. Measurement
process is performed in semiautomatic way. User has to set measurement parameters
described in (1) and start the process. All current changes and data acquisition is done
automatically which increase measurement accuracy. Automatic operation save a lot of time
and eliminates the distortions arising from the presence of people close to measurement
system (e.g. air movement around heated wire). Moreover process parameters can be
changed remotely (using Internet) which means that it is possible to obtain whole family of
characteristics without physical access to the stand. The use of virtual measuring
instruments and computer control makes it easy to develop the measuring stand. Some
planned developments of the measuring stand would give:
188
-
the ability to automatically adjust mechanical load,

the opportunity to ask about specific mechanical load parameters (time or
displacement),
possibility to adjust air movement,
automatic measurement of environmental conditions (temperature, pressure,
humidity),
possibility of automatic adjusting environmental conditions.
The stand was co-financed with governmental founds (project number N N510 353036).
6. References
[1] Kciuk M.: Computer based measurement system for shape memory alloy. IX International
Workshop for Candidates for Doctors Degree OWD2007, Wisa 20-23.10.2007, str.
177-180,
[2] Koji Ikuta, Masahiro Tsukumoto, Shigeo Hirose: Shape memory alloy servo actuator system
with electric resistance feedback and application for active endoscope. CH25551/88/0000/0427$01.00 0 1988 IEEE
[3] Krzysztof Biereg: Porwnanie wybranych rwna konstytutywnych stopw z pamici ksztatu.
Modelowanie inynierskie, Gliwice 2006.
[4] Koji Ikuta, Masahiro Tsukamoto, Shigeo Hirose: Mathematical model and experimental
verification of Shape Memory Alloy for designing micro actuator. IEEE 1991
[5] Koji Ikuta, Hidenori Shimizu: Two dimensional mathematical model of Shape Memory Alloy
and intelligent SMA-CAD. IEEE 1993.
[6] Zikowski A.: Pseudoelastyczno stopw z pamici ksztatu badania dowiadczalne i opis
teoretyczny. Praca habilitacyjna, IPPT PAN, Warszawa 2006.
[7] Kyu-Jin Cho, Harry H. Assada: Multi-axis SMA actuator array for driving anthropomorphic
robot hand. International Conference on Robotics & Automation, Spain, April, 2005.
[8] Liu C.Y., Liao W. H.: A snake robot using Shape Memory Alloys. International Conference
on Robotics and Biomimetics Shenyang, China, August 22-26, 2004, Proceedings of
the 2004 IEEE.
[9] Masaru Fujii, Hiroshi Yokoi: Modeling and movement control of mobile SMA-Net.
International Symposium on Computational Inteligence in Robotics and
Automation, Kobe, Japan, July, 16-20, 2003, Proceedings 2003 IEEE.
[10] Ikuta K.: Micro/miniature Shape Memory Alloy actuator. CH2876-1/90/0000/2156$01.00 0
1990 IEEE.
[11] Technical data for sensor H9720 http://www.avagotech.com/docs/AV02-0511EN
[12] Zurbitu J., Kustov S., Zabaleta A., Cesari E., Aurrekoetxea J.: Thermo-mechanical behavior
of NiTi at impact. Shape Memory Alloys. ISBN 978-953-307-106-0, publishing date:
October 2010
[13] Kciuk M., Kapyta G.: Koncepcja stanowiska pomiarowego do wyznaczania charakterystyk
elektro-termo-mechanicznych stopw z pamici ksztatu (SMA). Materiay XIII
Sympozjum Podstawowe Problemy Energoelektroniki, Elektromechaniki i
Mechatroniki PPEEm 2009, Wisa 14-17.12.2009, str. 242-247.
[14] Hui Li, Chen-Xi Mao Jin-Ping Ou: Experimental and theoretical study on two types of Shape
Memory Alloy devices. Earthquake Engineering And Structural Dynamics 19
September 2007 in Wiley InterScience (www.intersciecnce.wiley.com).
10
Calibration of Measuring Systems Based
on Maximum Dynamic Error
Krzysztof Tomczyk
Cracow University of Technology, Faculty of Electrical and Computer Engineering,

Poland
1. Introduction
This chapter presents methods and algorithms for determining the maximum errors which
can be generated by measuring systems in reference to their standards. The application of
such errors in the process of calibration of measuring systems intended for measurement of
dynamic signals is discussed in detail. The problem of maximum errors lies in the fact that it
is impossible to analyze the full range of all possible dynamic input signals. Therefore we
find out the one that represents all signals. It is the signal generating errors of maximum
value.
The existence and availability of signals maximizing both the integral-square error and the
absolute value of error are discussed, and relevant solutions are presented. Moreover, the
constraints imposed on the input signal are analyzed. These constraints refer to the
magnitude as well as maximum rate of signal change. The latter constraint is applied in
order to match the dynamic properties of the signal to the dynamic properties of the system
under test.
In the relevant literature it was proved that the signals maximizing the errors in question
always achieve one of the constraints imposed on it. If only the signal amplitude is
constrained, the maximizing signal is always of the bang-bang type, while in the case of
two constraints relating to the magnitude and the rate of change such a signal can be only of
triangular or trapezoid shape with slopes resulting from the values of constraints.
For the integral-square error, no analytic solution referring to the shape of the maximizing
signal has been found so far. This is why the solution of this problem presented in the
chapter is based on the application of the genetic algorithm method. This method
guarantees that the results are obtained in minimized calculation time which depends only
on the assumed number of population, the value of the stop condition and number of
assumed switches.
In the case of the absolute error, the analytical formulae and algorithm which give precise
results and can be realized in a very short calculation time are considered. The signal
maximizing this criterion is presented in the form of figures showing the successive steps of
proceeding as well as the error shape corresponding to this signal.
Fig. 1 presents a block diagram of our calibration process.
190
Fig. 1. Block diagram of calibration process of measuring systems based on maximum

dynamic error.
In this chapter respective procedures and essential mathematical operations, which allow
carrying out the calibration process according to the block diagram presented above, are
discussed in detail. In part two the input-output relations of measuring systems are
discussed, part three is devoted to the synthesis of the mathematical model of standard and,
in part four error functionals are presented and analyzed. The problem of existence and
attainability of maximizing signals is discussed in part five, while parts six and seven
present the procedures for determining signals maximizing the integral-square error and the
absolute value of error.
The optimization of the minimax type on the basis of maximizing signals is presented in part
eight. Such optimization is performed in two steps. In the first step the input signal
maximizing the assumed error functional is determined, while in the second step the model
parameters optimization based on the obtained signal is performed.
The identification problem of the mathematical model of measuring system is omitted in
this chapter as it does not concern the theory of calibration. It was assumed that this model
exist and is given in the form of transfer function.
The practical application of the presented theory and algorithms is illustrated on the
example of low-pass measuring systems.
191
Calibration of Measuring Systems Based on Maximum Dynamic Error
2. Input-output relations of measuring systems

Let output signal y(t ) of calibrated system be described by integral convolution
t
y(t ) x( )h(t )d , t [0, T ]
(1)
h(t ) hm (t ) hs (t )
where: x(t ) - input signal, hm (t ), hs (t ) - impulse responses of system and its standard.
The convolution (1) can be presented in a discrete form as a sum
n
y(n) x( k )h(n k ), n 0,1,..., N 1
(2)
k 0
sample interval of T
or in matrix form
y0
h0
y
h
1
1
y 2 h2
h
N 1
N 1
0
h0
0
0
h1
h0
hN 2
hN 3
0 x0
0 x1
0 x2

h0 xN 1
(3)
In many cases it is convenient to present the output y(t ) by means of the following formula
y( n)
2
mk
i
1 n n
x( k ) e n 1
n 1 m0 k 0
h( k ) e
2
mk
n1
k 0
2
mn
n1
(4)
or by the Fourier transform

Y ( i ) X( i )H (i )
Y (i ) F { y(n)}, X( i ) F { x(n)} , H (i ) F { h(n)}
(5)
where F presents the DFT or FFT (Tomczyk, 2011).

Multiplication of the Fourier transformations in (5) gives
Re Y (i ) Re X(i )Re H ( i ) Im X(i )Im H (i )
Im Y (i ) Im X(i )Re H (i ) Re X(i )Im H (i )
(6)
y(n) IF Re Y (i ) j Im Y (i )
(7)
From Eq. (6) we get
where IF denotes inverse DFT.
192
In state space we have

y(n 1) y(n) x(n)
y() state vector, y() p , p number of state variables
(8)
x() input control vector, x() ,q number of inputs

p
Presenting (8) in the matrix form, we have

y(n 1) 1,1

y p ( n 1)
p ,1
1,p y(n) 1

x(n)
p ,p y p (n) p
(9)
where
( p x p ) e A
A( p x p ) state matrix
(10)
and
( p x q ) e At dt B A 1 ( e A I )B
0
(11)
B( p x q ) input control matrix, I( p x p ) unit matrix
3. Synthesis of mathematical model of standard

The calibration procedures for measuring systems intended for measurement of static
quantities are commonly known and have been used for a long time in measuring practice.
They cover the hierarchy of standards and calibrations circuits.
For measuring systems intended for measurement of dynamic quantities neither legal
regulations concerning hierarchy of their accuracy nor specific calibrating procedures have
been worked out so far. It results from the fact that for such systems input signals are time
dependent, often undetermined, and whose shapes cannot be predicted in advance. The
second reason is that the problem of standards for such systems has not been solved till now,
because various measuring systems fulfill different objective functions (Tomczyk&Sieja, 2006).
For such a situation we expect that models of standard will satisfy mathematical notation of
these functions. In the dynamic measurement selection of standard parameters can be realized
by means of optimization methods, which assure the conditions of non-distortion
transformation, or by means of ideal filters of a band-pass which corresponds to the range of
the work of the calibrated system. For low-pass systems this standard is given by
st
c e 0 for 0 m
H s (i )
0 for m
Impulse response of (12) equals
(12)
hs (t )
mSa m t t0
193
(13)
where: c - amplification coefficient, t0 - filter delay, m - cut-off frequency.
4. Error functionals
Dynamic errors applied in our calibration are defined in different function spaces by means
of chosen functionals. If the values of these functionals are determined by means of
maximizing signals, it means that they will be valid for any dynamic signals which can
appear at the input of the calibrated system. In this way all the possible input signals are
included in this special, maximizing one.
The process of maximum error determination requires a special input signal to be used,
which warrants that the error determined with it will always be higher or at least equal to
the value generated by any other signal. The solution of this problem needs to prove that
there exists a signal maximizing the chosen error functional at all, and if so, to elaborate a
suitable procedure for its determination. The maximum values of error can constitute the
base to work out the hierarchy of accuracy for instruments for dynamic measurement.
The integral-square error
T
I ( x ) y 2 (t ) dt
(14)
D( x ) y(t )
(15)
and absolute value of error
will be discussed as an example.

The choice of the functionals (14) and (15) results from the fact that they are the most
common in many different engineering domains.
5. Existence and attainability of maximizing signals

In (Layer, 2002, Layer&Tomczyk, 2010) it was proved that signals maximizing I(x) and D(x)
functionals in [0,T] always achieve one of the constraints imposed on them.
The constraints in question concern the magnitude a and rate of change . For systems
described by means of linear differential equations, the constraint of magnitude x(t ) results
from the measuring range of this system. Imposition on the signal of only this one constraint
gives the solutions of bang-bang type. As a result unexpected great values of errors are
generated. It is affected by the dynamics of bang-bang signals which have infinitely high
derivative values in the instants of switching. Outside of these instants, they have constant
values. Because of this, the bang-bang signals are not matched to the dynamics of
physically existing systems, since they can only transmit signals of a limited value of the rate
of change. In order to match the dynamic behaviour of the input signals an additional
constraint resulting from the dynamic properties of the calibrated system should be imposed.
Proper matching is obtained by restricting the maximum rate of change of the signal.
194
5.1 Constraints of signal
In the case of two simultaneous constraints a and , the maximizing signal has a
triangular shape with the slopes inclination ,
tg
(16)
or a trapezoid shape with the slopes inclination and the height a.

We can analyze here two different premises referring to the rate of change value. The first
one refers to the time domain
max x (t ) max h(t )

t[0,T ]
t[0, ]
(17)
where h(t) denotes the impulse responses of the system.

In the frequency domain constraint results from the maximum frequency of the bandpass of system m . In this case we have
max ( a sin mt ) a m .
t[0,T ]
(18)
where A denotes amplitude.
6. Procedure for determining signals maximizing the integral-square error

For the integral-square error (14) good results in the determining of signals maximizing this
functional are achieved by the application of genetic algorithm. Fig. 2 presents a flowchart of
this algorithm.
If I i x I h x , the switch s1 takes the position 1 and the value of I i x is assigned to
I h x . Then I h x is stored in the buffer memory. Parallel to this operation, the vector of
switching signal xi (t ) and the output y i (t ) are saved in the memory.
For i N the switch s 2 changes the position from 0 to 1 and the values xi (t ), y i (t ) and
I h ( x ) are assigned to x0 (t ), y(t ) and I max ( x ) , respectively.
The algorithm presented in Fig.3 includes three main operations: reproduction, crossing and
mutation.
In the first step, the initial population (c11 , c12 ,..., c1n ) including an even number of
chromosomes is selected at random. Each chromosome consists of detectors
( d11 , d12 ,..., d1m ),( d21 , d22 ,..., d2 m ),...,( dn 1 , dn 2 ,..., dnm ), which correspond to the switching times
of xi signal. For each chromosome the error (14) is determined. Then on the basis of the
obtained results, adaptation coefficient I N is calculated as
I N Ic1 Ic2 .. Icn

This coefficient presents the total error generated by all chromosomes (Tomczyk, 2006).
(19)
Fig. 2. Flowchart of calibration procedure
Fig. 3. Flowchart of genetic algorithm
195
196
In order to form the next populations, it is necessary to determine the percentage share Ic of
each chromosome in I N
Ic1
Ic1
IN
100%
(20)
Icn
I cn
IN
100%
Relations (19) and (20) allow for estimation of the usefulness of each chromosome in the
population. The higher Ic value of the given chromosome gives more probability to include
it in the next population.
When the difference between the obtained values of adaptation coefficients is too small, it is
necessary to carry out the scaling operation of the adaptation coefficient.
The next step is the operation of reproduction. According to the probability calculated by
(20), the chromosomes from the initial population are selected at random. Depending on the
value of Ic coefficient, chromosomes have a bigger or smaller chance to be included in the
next population. There are several ways of selection chromosomes for the new population.
The most popular way is represented by the roulette wheel method. The results of random
selection are rewritten to the new descendant population (c 21 , c 22 ,..., c 2 n ).
The next step is the crossing process. The chromosomes from the new population are joined
in pairs in a random way and for the given crossing probability Pc are crossed or not.
The crossing is carried out according to the following formulae (Tomczyk, 2006):
The first crossing of paired di 1 and d j 1 detectors gives
di 1 (1 ) di 1 d j 1
d j 1 di 1 1 d j 1
(21)
where: di 1 is a descendant detector of i chromosome, and d j 1 is a descendant detector of j

chromosome.
The coefficient in (21) is selected according to
i min
di 1
d j 1 di 1
d j 1
, i max
di 2 d i 1
d j 1 di 1
d j2 d j1
, j max
j min
di 1 d j 1
di 1 d j 1
(22)
where: i min and i max present the minimum and maximum limit of changeability for
the detector from i chromosome, while j min and j max present minimum and maximum
limit of changeability for the detector from the j chromosome.
197
The changeability range of is contained in the range between 0 and the third value
calculated by means of (22) minus this value multiplied by (sample interval of T ).
Then the value is selected at random from the above range, and is substituted into (21).
The second crossing of detectors di m from i chromosome and d j m from the j chromosome,
where k 2, 3,..., m , gives
di k (1 ) di k d j k
d j k di k (1 ) d j k
(23)
where di k is the k detector of i descendant chromosome and d j k is the k detector of j

descendant chromosome.
Now, the coefficient is selected according to
i min
j min
di k 1 di k
di k 1 di k
, i max
d j k di k
d j k di k
d j k 1 d j k
d j k 1 d j k
, j max
di k d j k
di k d j k
(24)
For the i chromosome, the mutation operation is described by
d i k ( d i ( k 1) d i ( k 1)) d i ( k 1)
[0, 1], k 1, 2,..., m
(25)
The time calculation of genetic algorithm can be reduced significantly if a stop condition is
applied. It stops the calculation, if the value of I h ( x ) stored in memory does not change
during a given number of iterations. (Tomczyk, 2006).
7. Algorithm for determining signals maximizing the absolute error

7.1 Signals constraint on magnitude
In the case of D( x ) error (15) the maximum y(t ) occurs for t T

T
max y(t ) y(T ) x( ) h(T ) d
(26)
if
x( ) x0 ( ) a sign[ h(T )]
(27)
where a is the magnitude of x( ).

Replacing by t in (26), we obtain
T
y(t ) y(T ) x(t ) h(T t ) dt

0
(28)
198
and x0 (t ) maximizing (28) has now the form

x0 (t ) a sign [ h(T t )]
(29)
Substituting (29) into (28) gives

max y(t ) y(T ) a
h(T t ) dt a h(t ) dt
(30)
which can be computed in the very simple way.

7.2 Signals constraint on magnitude and rate of change
Let
t
x(t ) ( )d
(31)
then
T
y(T ) h(T ) ( ) d dt
(32)
Constraints related to x(t ) and ( ) are as follows

t
| ( )d ||x(t )| a
(33)
| (t )||x (t )|
(34)
and
Changing the integration order in (32), we have

T
y(T ) ( ) h(T t ) dt d
(35)
and after replacing for t , we get finally

y(T ) (t ) h(T ) d dt
(36)
A formula (36) proves that (t ) maximizing y(T ) has the maximum magnitude (t ) by
virtue of the formula (34) if
T
(t ) sign h(T ) d
t
(37)
199
and (t ) 0 in such subintervals, for which (33) between the switching moments has the
form
t
| ( )d | a
(38)
Using the equations (32)-(38), we can determine signal x0 (t ) in the following cases
First case
t
When |0 ( )d a| for varying in the intervals [0, ] and [0, ], (Fig. 4a,b), where
0
0 (t ) for
h(T )d
and 0 (t ) for
h(T )d , than the signal
x0 (t ) is
determined in three steps, according to Eqs. (39)-(47)..
Fig. 4. Exemplary functions
h(T )d and 0 (t )
t
During the first step, the bang-bang functions 1 (t ) of the magnitude are determined
with switching moments resulting from (37) - Fig.5a
1 (t )
1 (t )
if (t ) 0
if (t ) 0
In the second step, we obtain the function 2 (t ) by integrating 1 (t ) - Fig. 5b.
(39)
200
Fig. 5. Exemplary functions
h(T )d , 1 (t ) and 2 (t ) 1 ( )d
Function 2 (t ) in particular switching intervals t1 , t2 ,..., tm of 1 (t ) is given by the following

relations
for t t1 , m 1
2 (t ) t
(40)
2 ( t ) t1 ( t t 1 )
(41)
for t1 t t2 , m 2
for ti t ti 1 , i 2, 3,..., m , tm 1 T , m - number of switchings
2 ( t ) t1
(1) j 1 (t j t j 1 ) ( 1)i (t ti )
(42)
j 2
In the last step, we determine the function 3 (t ) on the basis of 2 (t ) :
3 (t )
3 (t ) 0
if |2 (t )| a
if |2 (t )| a
(43)
Finally, through integration of 3 (t ), we obtain the signal x(t ) x0 (t ) and this is the aim.
This operation is shown in Fig. 6a.
During the intervals in which 3 (t ) , the signal shape is triangular, with the slope .
In the intervals in which 3 (t ) 0, the signal is a constant and has the magnitude a.
For m switching moments of 3 (t ) the value of error is described by the following
equations:
for m = 1
201
o1
t1
y(T )
t1
h(T ) d
oT o1
T t1
h(T ) ( t1 ) d
t1
o1 h(T ) d
(44)
t1
for m 2
y(T )
o1
t1
o om
T
T tm
t1
o o
ti
ti
ti 1
ti 1
i i 1 h(T )( ti 1 ) d oi 1 h(T ) d
h(T ) d
i 2 ti ti 1
tm
tm
(45)
h(T ) ( tm ) d om h(T ) d
where om x0 (tm ), oT x0 (T ).
Fig. 6b presents the signal x0 (t ) and the error y(t ) corresponding to it
Fig. 6. Function 3 (t ), signal x0 (t ) 3 ( )d and error y(t )

0
The equations (44-45) in the discrete form are defined by:

for n = 1
y( N )
N
o o1 n1
o1 n1
h( N k ) k N
h( N k )( k n1 ) o1 h( N k )
t1 k 0
N n1 k n1
k n1
(46)
for n 2
y( N )
n1
n1
m
o o
o1 n1
h( N k ) k i i 1 h( N k )( k ni 1 ) oi 1 h( N k )
n1 k 0
i 2
k ni 1
ni ni 1 k ni1
o om
N
N nm
k nm
k nm
h(N k )( k nm ) om h( N k )
(47)
202
Second case
If T a then the signal x0 (t ) is given directly by

t
x0 (t ) sign [ h(T )] d
(48)
and the error equals

t
y(t ) x0 ( ) h(t ) d
(49)
Equation (49) in the discrete form is defined by

N
y( N ) x( k ) h( n k ) , for n 0,1,..., N 1
(50)
k 0
Fig. 7 presents the signal x0 (t ) and error y(t ) obtained by (48)-(49)
Fig. 7. Signal x0 (t ) and error y(t )

Third case
If T a then the signal x0 (t ) is determined indirectly by means of the functions 4 (t ) 7 ( t )
4 (t )
h(T ) (t ) d
2a
(51)
5 (t ) a sign[ 4 (T t )]
6 (t ) 5 (t )
2a
6 (t ) 5 (t ) 5 t

if t
if
2a
(52)
2a
t T
(53)
203
7 ( t ) 6 ( t )
7 ( t ) 0
if t
if
7 ( t ) 6 t
2a
if
2a
2a
2a
(54)
t T
Equation (51) in the discrete form is given by
4 (n)
2a
h(n k ) n k , n 0,1,..., N 1
(55)
k 0
The functions 4 (t ) and 5 (t ) are shown in Fig. 8a, while 5 (t ) and 6 (t ) in Fig. 8b.
Fig. 8(a). Functions 4 (t ) and 5 (t ).
Fig. 8(b). Functions 5 (t ) and 6 (t )

t
Fig. 9a presents the function 7 (t ) and the signal x0 (t ) 7 ( ) d , while the signal x0 (t )
0
and the error y(t ) corresponding to it are shown in Fig. 9b (Layer&Tomczyk, 2009;
Layer&Tomczyk, 2010).
Fig. 9(a). Function 7 (t ) and signal x0 (t ).
Fig. 9(b). Signal x0 (t ) and error y(t )
204
8. Maximum errors in optimization of models of measuring systems

In the technical domain the application of higher order models usually gives a possibility to
better map the dynamic properties of real systems. On the other hand, the analysis of such
models is most often difficult and time-consuming. Hence a tendency has developed to
replace a higher order model by a simplified one. The class and order of such model are
adopted in an arbitrary way. Its parameters can be determined by means of methods that
minimize the mapping error, considering all possible input signals. Therefore it is suggested
to solve the problem by using one signal which maximizes the error. In this way, the
mapping error being determined is credible for any input. The application of the minimax
method by means of the Levenberg-Marquardt algorithm (Tomczyk, 2009) gives good
results.
This method includes two main numerical computation stages. At the first stage the
x0 (t ) signal maximizing the error (Fig.10) is determined.
Fig. 10. Block diagram for determining x0 (t ) signal

At the second stage the class and order of the simplified model are adopted. For this model
by means of x0 (t ) signal the parameters minimizing the error are determined.
The Levenberg-Marquardt algorithm is an iterative method, in which the vector of
unknown parameters is determined in k 1 step by the following formula
1
k 1 kT JT ( k , t )J( k , t ) k I JT ( k , t )y( k , t )
(56)
with the error

T
I y 2 ( k , t )dt
(57)
where
t
y( k , t ) h(t )x( ) d
0
and
(58)
y( k , t1 )

1
y( k , t2 )
J( k , t ) 1
y( k , tn )

1
y( k , t1 )
y( k , t1 )
2
m
y( k , t2 )
y( k , t2 )
2
m
y( k , tn )
y( k , tn )
2
m
205
(59)
The notations in (56)-(59) are as follows:

-
k 1, 2,..., p , p - number of iteration loops
J( n x m ) ( k , t ) - Jacobian matrix
I( m x m ) - unit matrix
k - scalar, its value changes during iteration

[ 1 , 2 ,..., m ] - model parameters searched for.
The Levenberg-Marquardt algorithm is used for computation in the following stages:

Stage 1, for k=1
1. Assume the initial values of the parameters of vector k
2.
Assume the initial value of the coefficient k (e.g. k = 0.1)
3.
4.
5.
Solve the matrix equation (59) and calculate (58)

Calculate the value of error (57)
Determine the parameters of vector k 1 , following (56).
Stage 2 and further steps, for k 2, 3,..., p

1.
Update the values of the parameters of vector k
2.
3.
Solve the matrix equations (59) , calculate (58) and (56)

Calculate the value of error (57)
Compare the values of error (57) for the step k and the step k 1. If the result is
y( k , t ) y( k 1 , t ), multiply k by the specified value (e.g. 10 ) and return to the
step 2 of stage 2. If the result is y( k , t ) y( k 1 , t ) divide k by the value and return to
the step 1 of stage 2.
Fig. 11. presents a block diagram of minimax optimization by means of the LevenbergMarquardt algorithm.
The iteration process can be finished if in the consecutive stage the decrease in the value of
error (57) is very small and insignificant. Then we fix k 0 and determine the final result
for the parameters of vector . If the coefficient k is high, it means that the solution is not
satisfactory, the values of the parameters of vector are not optimum and the value of error
(57) is not at minimum level. At this point we can assumed
JT ( k , t )J( k , t ) k I
(60)
206
Fig. 11. Block diagram of minimax optimization by means of the Levenberg-Marquardt algorithm
In such a case the formula (60) leads to the steepest descent method
k 1 kT
JT ( k , t )y( k , t )
(61)
If the value of coefficient k is small, it means that the values of the parameters of vector
are close to the optimum solution. Then
JT ( k , t )J( k , t ) k I
(62)
and the Levenberg-Marquardt algorithm is reduced to the Gauss-Newton method

1
k 1 kT JT ( k , t )J( k , t ) JT ( k , t )y( k , t )
(63)
The selection of coefficient values k and depends on the adopted programs and
selected software.
9. Example of application
As an example let us consider the fourth order Butterworth, Tchebyhev and Bessel analog
filters and integral-square error (14). These filters have the following transfer functions:
c
s 2
s 2
s
s
1.8478
1
0.7654
m
m
m
m
(64)
c
2
2
s
s
s
s
5.5339
1
2.1853
1
1.2039
0.1964
m
m
m
m
(65)
c
2
2
s
s
s
s
0.4889
1 0.3890
1
1.3397
0.7743
m
m
m
m
(66)
H Bt 4 (s )
HT 4 (s )
H Be 4 ( s )
207
As a reference let us adopt the mathematical model of ideal low-pass filter given by (12) and
(13) for c 1 and m 100 rad .
s
In order to determine the signals maximizing the error (14) let us adopt:
-
magnitude a 0.2V ,
rate of change , as a maximum absolute value of h d , where h(t ) is the
difference between the impulse responses of filter and its standard: Bt 4 1.11
V
V
and Be 4 1.01
s
s
respectively,
T 0 0.5s.,
discretization step 0.001s.
T 4 1.36
V
,
s
for Butterworth, Tchebyhev and Bessel filters
The obtained switching times of signals errors and their values for Butterworth, Tchebychev
and Bessel filters are presented in Tables 1-3.
Butterworth filter
a 0.2 V , 1.11
T[s]
V
s
x 0 (t )
I ( x0 )
0.05
[0.0, 0.045s.], [0.045, 0.05s.]
8.67 10 7
0.1
[0.0, 0.095s.], [0.095, 0.1s.]
4.57 10 5
0.15
[0.0, 0.13s.], [0.13, 0.15s.]
2.68 10 4
0.2
[0.0, 0.185s.], a [0.185, 0.2s.]
7.55 10 4
0.25
[0.0, 0.2s.], a [0.2, 0.25s.]
19 10 4
0.3
[0.0, 0.185s.], a [0.185, 0.275s.],

[0.275, 0.3s.]
25 10 4
0.35
[0.0, 0.05s.], [0.05, 0.285s.],

a [0.285, 0.35s.]
39 10 4
0.4
0.45
0.5
[0.0, 0.08s.], [0.08, 0.345s.],

a [0.345, 0.4s.]
[0.0, 0.11s.], [0.11, 0.405s.],
a [0.405, 0.45s.]
[0.0, 0.135s.], [0.135, 0.455s.],
a [0.455, 0.5s.]
57 10 4
79 10 4
113 10 4
Table 1. Time switchings and values of integral-square error for Butterworth filter
208
Tchebychev filter
a 0.2, 1.36
T[s]
V
s
x 0 (t )
I ( x0 )
0.05
[0.0, 0.045s.], [0.045, 0.05s.]
4.52 10 6
0.1
[0.0, 0.095s.], [0.095, 0.1s.]
4.61 10 5
0.15
[0.0, 0.12s.], [0.12, 0.15s.]
4.01 10 4
0.2
[0.0, 0.15s.], a [0.15, 0.2s.]
12 10 4
0.25
[0.0, 0.15s.], a [0.15, 0.25s.]
25 10 4
0.3
[0.0, 0.045s.], a [0.045, 0.24s.], [0.24, 0.3s.]
36 10 4
0.35
[0.0, 0.07s.], [0.07, 0.29s.], a [0.29, 0.35s.]
59 10 4
0.4
[0.0, 0.1s.], [0.1, 0.35s.], a [0.35, 0.4s.]
87 10 4
0.45
[0.0, 0.125s.], [0.125, 0.4s.], a [0.4, 0.45s.]
122 10 4
0.5
[0.0, 0.15s.], [0.15, 0.45s.], a [0.45, 0.5s.]
173 10 4
Table 2. Time switchings and values of integral-square error for Tchebychev filter
Bessel filter
a 0.2, 1.01
T[s]
V
s
x 0 (t )
I ( x0 )
0.05
[0.0, 0.045s.], [0.045, 0.05s.]
8.55 10 8
0.1
[0.0, 0.095s.], [0.095, 0.1s.]
5.66 10 5
0.15
[0.0, 0.12s.], [0.12, 0.15s.]
2.74 10 4
0.2
[0.0, 0.15s.], a [0.15, 0.2s.]
7.28 10 4
0.25
[0.0, 0.15s.], a [0.15, 0.25s.]
18 10 4
0.3
[0.0, 0.045s.], a [0.045, 0.24s.], [0.24, 0.3s.]
25 10 4
0.35
[0.0, 0.07s.], [0.07, 0.29s.], a [0.29, 0.35s.]
37 10 4
0.4
[0.0, 0.1s.], [0.1, 0.35s.], a [0.35, 0.4s.]
51 10 4
0.45
[0.0, 0.125s.], [0.125, 0.4s.], a [0.4, 0.45s.]
70 10 4
0.5
[0.0, 0.15s.], [0.15, 0.45s.], a [0.45, 0.5s.]
99 10 4
Table 3. Time switchings and values of integral-square error for Bessel filter
209
Fig. 12 presents signals x0 (t ) and errors y(t ) as well as Fig. 13 illustrates the integral-square
error as function of time duration T 0 0.5 s. for T 0.05 s.
Fig. 12. Signals x0 (t ) and error y(t ) for Butterworth a), Tchebyhev b) and Bessel c) filters.
Fig. 13. Integral-square error as function of time duration T 0 0.5 s. for Butterworth,
Tchebychev and Bessel filters.
210
From Fig. 13 it follows that for T 0.2 s. signal x0 (t ) generates similar values of error for all
filters. In the interval 0.2 s. T 0.35 s. the highest value of error generates the Tchebyshev
filter, however for the Butterworth and Bessel filters the values of error are similar. For
T 0.35 s. the higher error generates the Tchebyshev filter, whereas the Bessel filter
generates the lowest value.
10. Conclusion
The calibration of systems based on the theory of maximum dynamic errors, makes it
possible to create accuracy classes for such dynamic systems and the hierarchy of dynamic
accuracy resulting from them. These hierarchies are valid regardless of the shape of the
measured dynamic signal that can appear at the input of the investigated system.
The calibration methods presented in this chapter can be widely applied in many different
domains, eg. in electrical metrology (transducers, strain gauge amplifiers, filters, etc),
geophysics (accelerometer and vibrometer systems), medicine (electroencephalograph and
electrocardiograph systems), meteorology (autocomparators, autobridges), etc.
11. References
Layer, E. (2002). Modelling of Simplified Dynamical Systems, Springer-Verlag, ISBN 3-54043762-2, Berlin Heidelberg
Layer, E.; Tomczyk, K. (2009). Determination of non-standard input signal maximizing the
absolute error, Metrology and Measurement Systems. Vol. XVII, no. 2, pp. 199-208,
ISSN 0860-8229
Layer, E.; Tomczyk, K. (2010). Measurements, Modelling and Simulation of Dynamic Systems,
Springer-Verlag, ISBN 978-3-642-04587-5, Berlin Heidelberg
Tomczyk, K. (2006). Application of genetic algorithm to measurement system calibration intended
for dynamic measurement, Metrology and Measurement Systems. Vol. XIII, no. 1, pp.
93-103, ISSN 0860-8229
Tomczyk, K. (2009). Levenberg-Marquardt Algorithm for Optimization of Mathematical
Models according to Minimax Objective Function of Measurement Systems,
Metrology and Measurement Systems. Vol. XVI, no. 4, pp. 599-606, ISSN 0860-8229
Tomczyk, K. (2011). Procedure for correction of the ECG signal error introduced by skin-electrode
interface, Metrology and Measurement Systems. Vol. XVIII, no. 3, pp. 461-470, ISSN
0860-8229
Tomczyk, K.; Sieja, M. (2006). Acceleration transducers calibration based on maximum dynamic
error, Czasopismo Techniczne 3-E, Cracow University of Technology, pp. 37-49,
ISSN 0011-4561
11
Determining Exact Point Correspondences in
3D Measurement Systems Using Fringe
Projection Concepts, Algorithms and
Accuracy Determination
Christian Bruer-Burchardt, Max Mller, Christoph Munkelt,
Matthias Heinze, Peter Khmstedt and Gunther Notni
Fraunhofer IOF Jena, Jena,

Germany
1. Introduction
Fringe projection systems for contactless surface measurements are increasingly used in
applications of industrial inspection, quality control, rapid prototyping, archaeology, and in
medical applications. Such systems provide increasing accuracy together with shorter
measurement time and increasing data volume. The demands of increasing data volume,
accuracy, and measurement speed can be satisfied using a more powerful calculation
technique and better algorithms.
Calculation of 3D surface points by fringe projection systems is typically based on
triangulation known from photogrammetry. An extensive review gives e.g. Luhmann
(Luhmann et al., 2006). This involves the task of finding homologous points describing points
in different images depicting the same original measuring point. In fringe projection systems
the finding of homologous points typically uses phase information (Creath, 1986) of the points.
The phase value characterizes the origin of a projected ray in the projector image plane. The
projector is geometrically considered as an inverse camera (Schreiber & Notni, 2000).
When an optical contactless 3D measurement system has to be designed or selected among
different alternatives some different basics should be clear. It must be known which objects
should be measured, how fast the measurement should run, and how accurate the
measurement has to be. If these aspects are plain, the design or selection may start. As the
calculation of the resulting 3D points on the objects surface is based on triangulation, the
procedure of finding corresponding points is crucial. This procedure has mainly two
aspects: uniqueness and precision of the localized position of the corresponding points. In
this work the various methods to obtain unique point correspondences are not described in
detail, but rather the second aspect (precision) should be dealt with. Uniqueness of point
correspondences is obtained by phase unwrapping (Sansoni et al., 1999; Reich et al., 2000).
This can be realized e.g. by the use of multiple spatial frequencies (Li et al., 2005), temporal
phase unwrapping methods (Zhang et al., 1999), or use of Gray code sequences (Sansoni et
al., 1999). Due to its unambiguousness, the usage of Gray code leads to robust results.
212
An extensive survey over coded structured light techniques to solve the correspondence
problem which is the basis for 3D surface reconstruction is given by Battle (Battle et al.,
1998).
Zhang and Yau suggest a real-time coordinate measurement (Zhang & Yau, 2006) where
phase unwrapping is realized by determination and tracking of a marker. An interesting
method for phase unwrapping using at least two cameras is presented by Ishiyama
(Ishiyama et al., 2007a). There the number of possible correspondences is drastically reduced
by back-propagation of the correspondence candidates into the image of the second camera.
Ishiyama gives another suggestion (Ishiyama et al., 2007b) for 3D measurement using the
invariance of cross-ratio of perspective projection. Young (Young et al., 2007) suggests the
use of the limitation of the measuring volume in order to reduce the search area for
corresponding points on the epipolar line to segments achieving a reduction of the projected
binary code by careful placement of additional cameras (or additional measuring positions).
Li (Li et al., 2009) uses this approach in combination with the multi-frequency technique in
order to realize real-time 3D measurements.
We presented an algorithm (Bruer-Burchardt, 2011a) describing how one can obtain unique
point correspondences using geometric constraints. This algorithm works when certain
conditions as restricted measurement volume depth hold.
A number of works were published from our working group concerning accuracy
determination of phase values and phase value based measurements (Notni & Notni, 2003;
Bruer-Burchardt et al., 2010). These works show the importance of theoretical error
estimation as well as the performance of experimental error analysis of fringe projection
based 3D surface measurement systems.
Different arrangements of camera(s) and projector(s) in 3D measuring systems using fringe
projection have been recently proposed (Maas, 1992; Ishiyama et al., 2007a). Arising
methods are based on triangulation (Luhmann et al., 2006) between two or bundle
adjustment between more than two viewpoints (sensor positions), or between two camera
positions within one sensor position. The most common triangulation procedure to obtain
3D measurement data in fringe projection systems is between two cameras or one camera
and one projector (Schreiber & Notni, 2000; Reich, 2000).
We consider the case of having one or two cameras and one projector in our sensor head.
We distinguish three kinds of establishing correspondences using one sensor position. The
first one is the simple observation of a phase value (Creath, 1986; Schreiber & Notni, 2000) at
a certain camera coordinate. The corresponding point in the projector image is directly
yielded by the observed phase value. This method is called CP mode (one camera and one
projector are used for triangulation). Another method is to search for the observed phase
value in the second camera image denoted by CC (two cameras are used for triangulation).
The third one called VR mode (virtual raster) uses a virtual projection grid and finds the
phase values of the grid points in the two camera images. This method corresponds to the
technique described by Reich (Reich et al., 2000).
In this work these different methods are compared regarding measurement accuracy,
handling, and sensitivity against image disturbances. The measuring principles are briefly
presented and advantages and disadvantages are discussed.
Determining Exact Point Correspondences in 3D Measurement Systems

Using Fringe Projection Concepts, Algorithms and Accuracy Determination
213
A model describing the random error of the resulting 3D measurement data was developed
and confirmed by the results of simulation experiments and real data measurements
obtained by two different measuring devices. The results are analysed and discussed and an
outlook to future work is given.
2. Situation and measuring principles

Recently several fringe projection systems for 3D surface determination for different
measurement objects were developed at our institute (Khmstedt et al., 2005; Munkelt et al.,
2005; Khmstedt et al., 2007; Khmstedt et al. 2007b; Bruer-Burchardt et al., 2011b). Some of
them are shown in figs.1 and 2. They are based on the projection and observation of one or two
90 rotated fringe sequences consisting of sinusoidal fringe patterns (between 3 and 16 images)
and optionally of a Gray code sequence (usually 5 to 7 images). From these sequences phase
values are determined using a 3-, 4-, 6-, 8-, or 16-phase algorithm (Khmstedt et al., 2007).
Phase values are used together with the extrinsic and intrinsic parameters of the optical
components to determine the 3D point coordinates of the reconstructed object surface.
Fig. 1. Several fringe projection based measurement systems developed at Fraunhofer IOF:
kolibri flex mini, kolibri CORDLESS, kolibri step (from left to right)
The use of epipolar constraint in order to find point correspondences is a typical approach
in phototogrammetry (Luhmann et al., 2006). It reduces the task to a correspondence
problem between two one-dimensional vectors.
Let us consider using a sensor consisting of two cameras C1 and C2 and one projector P in a
fix geometric arrangement (see fig. 2 right).
In the following the basic principles of fringe projection based 3D surface measurement
should be briefly explained.
2.1 Phasogrammetry
Phasogrammetry is the connection of the mathematical principles of photogrammetry and
fringe projection. The classical approach of fringe projection is described e.g. by Schreiber
and Notni (Schreiber & Notni, 2000), which has been extended depending on several
applications (Reich et al., 2000; Chen & Brown, 2000). The principle should be briefly
explained as follows. A fringe projection unit projects one or two perpendicular, well
defined fringe sequences onto the object, which is observed by one or more cameras. These
214
sequences may consist of a binary code sequence as the Gray code (Sansoni et al., 1999;
Thesing, 2000) and a sequence of up to 16 sinusoidal fringe patterns. The so called rough
phase value (Schreiber & Notni, 2000), and in combination with the Gray code the unique
unwrapped phase value (Sansoni et al., 1999) is obtained using the sequence of sinusoidal
patterns. Unique phase values are used to realize point correspondences in order to obtain
measurement values by triangulation (Luhmann et al., 2006).
Fig. 2. High speed sensor HS (left), intraoral sensor DirectScan (middle), and typical sensor
arrangement (right)
The rough phase value is the phase position within one period and the unique phase value
is obtained by phase unwrapping leading to a monotone series of phase values (if error free)
in a certain direction.
Phase values are produced in order to identify projector image coordinates or to produce
virtual landmarks (Khmstedt et al., 2007). These markers may be used both for calibration
of the system and for the calculation of the 3D measurement data.
One sequence of fringe images is processed resulting in phase images i,x for each
measuring position and each camera Ci. After rotation of the fringe pattern by 90, the
sequence may be projected and observed again resulting in a second phase image i,y for
each camera. The phase values i,x and i,y (see fig. 3) correspond to image coordinates in
the projector plane. The resulting 3D points are obtained by triangulation between the
coordinates of the camera and the projector, or between corresponding points of two
cameras. This can be regarded as standard procedure in photogrammetry (Luhmann et al.,
2006).
2.2 Stereo vision and epipolar geometry
Using active stereo vision, images of the object are captured from two different
perspectives. Pairs of image coordinates resulting from the same object point (the
homologous points) have to be identified. The object can be reconstructed by
triangulation using these points. In the case of active stereo vision a single intensity
pattern or a sequence of patterns is projected onto the measurement object. There are
several techniques to identify the homologous points in both cameras as e.g. described by
Luhmann (Luhmann et al., 2006).

215
Fig. 3. Principle of phasogrammetry

Epipolar geometry is a well-known principle which is often used in computer vision when
stereo systems are present. It is characterized by an arrangement of two cameras observing
almost the same object scene. The projection centres O1 and O2 of the two cameras define
together with an object point M a plane E in the 3D space (see fig. 4). The images of E are
corresponding epipolar lines concerning M. When the image point m1 of M is selected in the
image I1 of camera C1, the corresponding point m2 in the image I2 of camera C2 must lie on
the corresponding epipolar line. This restricts the search area in the task of finding
corresponding points. In the following we assume a system consisting of two cameras C1
and C2 and one projector P in a fixed arrangement.
Epipolar constraint usage is advantageous compared to using second fringe projection
direction because only 50% of the recorded images are necessary. However, the calibration
should be accurate and stable because correspondence finding is performed only on a line.
Systematic errors of the sensor geometry lead to systematic errors of the position
determination of the corresponding point to be found (Bruer-Burchardt et al., 2011c).
Fig. 4. Epipolar geometry
216
2.3 Camera calibration

Camera calibration describes the process of the determination of the intrinsic and extrinsic
parameters (including lens distortion parameters) of an optical system. It has been
extensively described in the literature, e.g. in (Brown, 1971; Tsai, 1986; Chen and Brown,
2000; Luhmann et al., 2006). Different principles have been applied in order to conduct
camera calibration. The choice of the method depends on the kind of the optical system, the
exterior conditions, and the desired measurement quality. In case of the calibration of
photogrammetric stereo camera pairs, the intrinsic parameters (principal length, principal
point, and distortion description) of both cameras should be determined as well as the
relative orientation between the cameras.
The position of the camera in the 3D coordinate system is described by the position of the
projection centre O = (X, Y, Z) and a rotation matrix R obtained from the three orientation
angles , , and . Considering stereo camera systems, only the relative orientation between
the two cameras (Luhmann et al., 2006) is considered, because the absolute position of the
stereo sensor is usually out of interest.
Lens distortion may be considerable and should be corrected by a distortion correction
operator D. It can be described by distortion functions or by a field of selected distortion
vectors (distortion matrix). The determination of D is performed within the calibration
procedure (Tsai, 1986) or separately (Bruer-Burchardt, 2004).
3. Different concepts of accurate point correspondence determination

Correspondence determination in a stereo system means first the identification and, second,
the localization of the point coordinates in the two considered 2D images mapping the same
original 3D point.
The first task can be solved using a suitable method which depends on the actual conditions
of the measurement. As phase values are concerned as the basis of the point correspondence
finding we distinguish between methods producing unique phase values (e.g. using Gray
code sequences or multi frequency projection techniques) and periodically repeating rough
phase values.
In this work we assume to use a technique leading to unique phase values either in one or
two directions. How to obtain unique phase values can be read elsewhere e.g. in (Battle et
al., 1998; Zhang et al., 1999; Sansoni et al., 1999; Li et al., 2005; Bruer-Burchardt et al.,
2011a). Assume in the following that the uniqueness problem is solved.
In this section different concepts of correct point localization should be considered. These
concepts distinguish e.g. concerning the hardware effort (one or two cameras), the costs, and
the error sensitivity.
3.1 The CP method
The CP method is the simplest method to localize point correspondences and realize 3D
point calculation.
Finding point correspondences is performed as follows. The pixels of camera C1 are the
coordinates x1, y1 of a point p1 = (x1, y1) and the observed phase values ( and ) at p1

217
determine the coordinates of the corresponding point in the projector image plane. Together
with the intrinsic camera parameters of C1 and P and the relative orientation between C1 and
P, the 3D point is calculated by triangulation (see fig. 3).
For this method only one camera is used and no algorithmic effort for finding
corresponding points is necessary. Hence this is also the simplest method concerning the
computational effort. Additionally, because only two major hardware sensor components
(one camera and one projector) are necessary, such sensors become lightweight and cheap.
These aspects are advantageous whereas the dispensation with the second camera may be
unfavourable. Figure 5 (left) shows a sketch of the CP arrangement.
The random error of the 3D point measurement is determined by the uncertainty of the
phase determination and its scaling depends on the triangulation angle (see section 5).
3.1.1 CP algorithm
Precondition: Projection and observation (by one camera C1) of a fringe image series consisting of two 90 rotated sequences subsequently producing two phase images xand y (*)
Input: two phase images xand y (*) for the observation camera C1
Consider all pixels of the image I1 of camera C1. Let pi=(xi, yi) the coordinates of point pi.
Read the phase values i and i at pi: qi =(i, i)

Perform triangulation between pi and qi to obtain 3D point Pi
Output: point cloud {Pi}

(*) Remark: The CP algorithm works also using only one projected fringe direction. Then,
triangulation is performed between a ray and a plane in the 3D space.
3.2 The CPE method
The CPE method is the simple extension of the CP mode according to epipolar geometry.
The second projected fringe direction is omitted. Instead of it the geometric information
of the epipolar lines is used. The second coordinate of the observed phase value results
from the position of the epipolar line in the projector image plane as illustrated by fig. 5
(right).
The advantages of the CPE method are the same as those of CP together with a reduced
image sequence (half the number of images) meaning shorter projection and observation
time and subsequently leading to shorter measurement time. Additionally, calculation
time is reduced, too. The disadvantage is a reduced robustness of the coordinate
determination, mainly determined by the epipolar lines. This can be, however, prevented
by additional control measurements as e.g. suggested by the authors (Bruer-Burchardt et
al., 2011c).
3.2.1 CPE algorithm
Precondition: Projection of a fringe image series in one direction subsequently producing
one phase image x and observation (by one camera C1)
218
Fig. 5. Projector and camera arrangement for the CP (left) and the CPE (right) algorithms
Input: one phase image x for the observation camera C1
Calculate the epipolar line gi in the projector image plane corresponding to pi
Read the phase value i at pi:
Find the position qi =(i, i) of phase value i on gi

3.3 The CC method
The CC method (see Munkelt et al., 2005; Khmstedt et al., 2007) uses corresponding points
in the images of the two cameras C1 and C2 (see fig.1). For finding the point correspondence
camera C1 is disclaimed as primary camera. The camera pixels of C1 are the coordinates of
the point p1 = (x1, y1). The phase values = (, ) at p1 are searched for in the image of
camera C2. This search may use bilinear or bicubic interpolation and determines the exact
position p2 = (x2, y2) of (, ) with subpixel accuracy in the image of camera C2 (see fig. 6).
Together with the intrinsic camera parameters of C1 and C2 and the relative orientation
between C1 and C2 the 3D point is calculated by triangulation.
The same procedure can be performed using C2 as primary camera and determining the
position of the corresponding point in the C1 image with subpixel accuracy.
The main advantage of the CC technique over the CP method is the possibility to correct
errors influenced by disturbed illumination or characteristic line errors. If the illumination is
disturbed in both images in the same manner the error is similar and the correspondence
will be found correctly despite the disturbance. This effect was reported by Munkelt
(Munkelt et al., 2005).
The main disadvantage is the additional camera leading to higher weight and bigger
volume of the sensor, higher costs, and an increased calculation effort.
3.3.1 CC algorithm
Precondition: Projection and observation (by two cameras C1 and C2) of a fringe image series
consisting of two 90 rotated sequences subsequently producing two phase images xand
y for each camera

219
Fig. 6. Sensor arrangements for the CC (left) and the CCE (right) algorithms
Input: four phase images 1x1y, 2x and 2y for both observation cameras C1 and C2
Read the phase values i and i at pi: i =(i, i), i=1,2

Find coordinates qi=(ui, vi) in the image of C2 with phase values I = (i,i)

3.4 The CCE method
The CPE method is the extension of the CP mode according to epipolar geometry. Instead of
the second projected fringe direction the geometric information of the epipolar lines in the
image I2 of camera C2 concerning the considered image point coordinates p=(x, y) are used.
The advantage of the method is the connection of the advantages of the CC method together
with the epipolar constraint use. The disadvantages are analogous.
3.4.1 CCE algorithm
one phase image 1xand 2x for each of the two cameras C1 and C2
Input: phase images 1xand 2x for the observation cameras C1 and C2
Calculate the epipolar line gi in image I2 of camera C2 corresponding to pi

Read the phase value i at pi:
Find the position qi =(i, i) of phase value i on gi

3.5 The VR method
The VR method (VR means virtual raster) is derived from the CC method. Instead of taking
the measured phase values at the pixels of one of the two cameras, a virtual phase raster is
220
defined according to the desired 3D point density in the resulting point cloud. The position
of every phase raster point is determined in the image of the C1 camera as well as in the
image of the C2 camera (see fig.2, right). If both positions p1 = (x1, y1) and p2 = (x2, y2) are
detectable, the triangulation can be performed and the resulting 3D point can be calculated.
The advantage of the VR method is the possibility to choose an arbitrary spatial resolution
of the resulting 3D measurement data. However, it should be noticed that neighbouring
resulting points may be not independent. Here a particular careful error analysis based e.g.
on the work of Notni (Notni & Notni, 2003) should be performed.
3.5.1 VR algorithm
Precondition: Projection and observation (by two cameras C1 and C2) of a fringe image series
consisting of two orthogonal sequences subsequently producing two phase images xand
y for both cameras C1 and C2
Input: four phase images 1x1y, 2x and 2y for both observation cameras C1 and C2
Consider all phase values i =(i, i) of the selected raster.
Find coordinates pi=(xi, yi) with the phase values i =(i, i) in C1

Find coordinates qi=(ui, vi) with the phase values i =(i, i) in C2

3.6 The VRE method
The VRE method is obtained by extension of the VR method by epipolar constraint. It can be
achieved by fixing the raster of the searched phase values. However, the raster of the
selected epipolar lines must be fixed, too. Usually the raster step width must be different
because one direction is driven by the phase and the other by a metric default value.
Hence it is doubtful whether the usage of the VRE method would be meaningful. Too few
experiments have been performed yet. A really convincing contention could only be given,
if further analysis and experiments would have been successfully performed.
3.6.1 VRE algorithm
one phase image x for each camera C1 and C2
Input: two phase images 1x and 2x for the observation cameras C1 and C2
Predefine a bundle of corresponding epipolar lines gi,1 and gi,2 in both camera images
Consider all phase values i of the selected raster
Find coordinates pi=(xi, yi) with the phase value i on gi,1

Find coordinates qi=(ui, vi) with the phase value i on gi,2

221
Fig. 7. Sensor arrangements for the VR (left) and the VRE (right) algorithms
4. Random and systematic errors at point correspondence localization

In order to evaluate and to compare all the different correspondence finding methods, an
estimation of the error of the reconstructed 3D points was performed. The 3D error of a
reconstructed point depends on a number of influences including phase measurement error,
characteristic line error, calibration error, distortion, and others, having a systematic and a
random component. All error sources that are not depending on the correspondence
algorithm are assumed to have the same influence on the measurement accuracy of all
considered methods.
Assuming that the determined calibration is fixed and the error of the calibration occurs as a
systematic error, the phase noise has the main influence on the random error of the resulting
3D point. All other relevant error sources are expected to be systematic and are not considered
here. A detailed analysis of the phase error is described by Notni (Notni & Notni, 2003).
4.1 Error models
As suggested by Rivera-Rios (Rivera-Rios et al. 2000) we assume a Gaussian independent
random error p=(x,y) of the observed point coordinates in the camera images. This
leads to a certain random error P=(X,Y,Z) of the measured 3D coordinate of the
reconstructed point, additionally depending on the triangulation angle . This error is not
symmetric in X, Y, and Z. The smaller the triangulation angle the higher is the longitudinal
error compared to the lateral one. The amount of the error in the different coordinates
depends on the orientation of the cameras in the world coordinate system W, too.
Assume the orientation of the world coordinate system with Z-axis along the boresight of
the sensor obtained by a transform of W yielding W. Then the standard deviations in X-, Yand Z-direction of the measured point cloud characterize the lateral and longitudinal
aspects of the random error of the measurement.
However, in the case of our correspondence methods CP and CC the point clouds
characterizing the error of one measured point are deformed to a flat (situated nearly in a
plane) point cloud. This is because one coordinate (the phase value or one camera pixel) is
fixed. In the case of VR the shape of the error point cloud is the intersection of two cones
(see fig. 9).
222
The extensions to CPE, CCE, and VRE by epipolar constraint reduce the shape of the error
clouds again: to a line segment in the CPE and CCE case and to a flat ellipse in the VRE case
(see figs. 8 and 9). Initially, this means a reduction of the random error by use of epipolar
constraint. However, a systematic error may additionally occur in the case of a calibration
error. This error, however, may be reduced by some additional effort as shown by the
authors (Bruer-Burchardt et al., 2011c).
Fig. 8. Error model for CP/CC (left) and CPE/CCE (right)
Fig. 9. Error model for the VR (left) and the VRE (right) algorithms
In order to estimate the random error of a single point measurement the following
assumptions are made. Assume that the calibration is fix and error free, and a possible
distortion error is completely corrected. The error p of the camera coordinate of a certain
point representing a given phase value should be normally distributed with expectation
value E(p)=0 and a certain variance c(p) which should be equal in X- and Y- direction.
The amount of this error depends on the phase noise (depending on the measuring
conditions, number of images in the sequence (Notni & Notni, 2003), and the parameters of
the current filter operators (e.g. Gaussian).
Let c be the standard variation of the random error of the pixel coordinate after all filtering
operations, m the magnification of the mapping and M the inverse magnification M=1/m.
Let us consider first the case CP/CC and the transformed world coordinate system W.
Because of the fixed coordinate in the primary camera, the random error of the measured 3D
point depends only on the random error in one camera image and the triangulation angle .
We consider the standard deviations x (p), y (p), and z (p) of the calculated 3D point
coordinates. We obtain

1
2
x 0; y c M fac ; z
1
c M fac
sin
223
(1)
where fac is set to fac = 1 for both the CP and the CC mode. Rather it should be fac = 1 for CP
and fac 2 for the CC method because of the double uncertainty of the phase value.
However, a number of experiments showed that the random error of CC is equal to that of
CP mode using the same triangulation angle. We assume that usually some error sources
have the same influence to the phase error in both images of cameras and hence this causes
the reduction of fac 2 to fac = 1 also in the CC mode.
Considering CPE and CCE we obtain
x 0; y 0; z
1
c M fac
sin
(2)
for both the CPE and the CCE mode.

Considering VR and assuming an independent random error p with E(p)=0 and c(p)
of the same amount in the two camera images, we obtain
.x
1
1
2
c M; y
c M; z
c M
sin
2
2
(3)
and for VRE we get

. x 0; y 0; z
2
c M
sin
(4)
As we can see considering formulas (1) to (4) CPE and CCE provide the best estimation of
the random 3D error because the error only depends on the uncertainty of the phase
determination along the epipolar line. However, an additional systematic error can occur
due to the uncertainty of the epipolar line position. It may be erroneous due to calibration
errors caused by thermic or mechanic influences as shocks or vibrations. However, the
stability of the calibration can be checked, and hence this error can be estimated and
minimized as it was shown by the authors (Bruer-Burchardt et al., 2011c). CP/CC and
CPE/CCE errors are mainly influenced by the triangulation angle .
5. Experiments and results

5.1 Simulations
In order to estimate the random error of a 3D point measurement and to confirm the
theoretic assumptions according to formulas (1) and (2), the following simulation
experiment was performed. For a given geometric situation of the sensor (extrinsic and
intrinsic parameters of the camera(s) and the projector, triangulation angle between the two
relevant image normal vectors) a meaningful value of the random error p of the camera
coordinates was chosen. It was chosen a normally distributed random value p = (x, y)
with expectation value (0, 0) for the camera coordinate uncertainty. The variance was chosen
224
according to realistic conditions by experiment. The error E of the calculated 3D point

coordinate is given as (X, Y, Z) error vector in the virtual world coordinate system W.
The alignment of W is according to the orientation of the boresight of the contributing
cameras (C and P or C1 and C2, resp.).
Error estimation will be achieved by analysis of the covariance matrix of the coordinate
values of the point clouds. These point clouds are those of the measured or simulated points
in the world coordinate system W according to the calibration data. The variances of the
error vector components are obtained as the eigenvalues ev1, ev2, and ev3 of the covariance
matrix. Hence the random error of the 3D point calculation can be described by the standard
deviations in the virtual world coordinate system W corresponding to the square roots of
the eigenvalues:
lsx ev1 ; lsy ev2 ; lsz ev3
(5)
The simulation of a measurement with random error was performed 400 times. The
resulting 3D point cloud was analysed using covariance matrix leading to the 3D point error
characteristic (x, y, and z). For the VR case the image coordinate of the pixels of camera
C1 were disturbed by a 2D error value with the same statistical characteristics (normal
distribution, zero expectation value, same variance as for C2), too.
The standard deviations of the three coordinates in the transformed world coordinate
system W according to (1) and (2) and the simulation results using 400 measurements are
given in table 1 showing that it holds approximately x = lsx,y = lsy, and z = lsz.
exp. status
model
simulation
experiment
quantity \
mode
[]
x [m]
y [m]
z [m]
lsx [m]
lsy [m]
lsz [m]
lsx [m]
lsy [m]
lsz [m]
CP
CPE
CC
CCE
VR
VRE
11,2
0,0
1,2
11,9
0,0
1,2
13,0
0,003
1,1
11,0
11,2
0,0
0,0
11,9
0,0
0,0
12,8
0,003
0,003
11,0
24,0
0,0
1,2
5,7
0,0
1,3
6,0
0,003
1,0
6,0
24,0
0,0
0,0
5,7
0,0
0,0
5,9
0,003
0,003
6,0
24,0
1,6
1,6
8,0
1,7
1,7
8,0
1,0
1,2
9,4
24,0
1,6
0,0
8,0
1,6
0,0
8,1
1,0
0,005
9,4
Table 1. Model data, simulation data, and experimental data of measurements using
different methods
5.2 Experiments on real measurements
In order to evaluate the different correspondence finding methods several experiments

determining the accuracy of the measurements were performed. For the measurements the
kolibri flex mini system was used. This system (see fig.1) is a table top system with a
sensor head consisting of two cameras (1.4 Mpixels) and one projector. The working
distance is about 300 mm and the measuring volume is 90 mm (diameter) x 25 mm (height).

225
The measuring accuracy given by the standard deviation of the surface points on a plane is
about 5 m.
The first experiment was a measurement of a single point in repeated measurements in
order to confirm the simulation experiment described in the previous section. The number
of measurements was 400. The resulting ls-values of all measurements are listed in table 1
and confirm the theoretic assumptions.
The next experiment was performed in order to confirm the assumption about the
dependence of the random error on the triangulation angle. Different triangulation angles
were realized by a modification of the kolibri flex mini system. Each measurement with
one fix triangulation angle and using the VR mode was repeated at least 100 times. A range
between 5 and 40 was realized for the triangulation angle .
The results of the variation of the triangulation angle are shown by a plot of the accuracy
values in fig. 10.
Fig. 10. Plot of accuracy of VR mode depending on triangulation angle
6. Discussion
The results of the experiments and the simulations mainly confirm the theoretic analysis of the
measuring accuracy. Especially in the CP/CPE and CC/CCE modes the correlation between
the results is very high. Using the VR/VRE modes, too few experiments have been performed
in order to make significant statements. Here, additional effects may influence the random
error and lead to a difference to our model. This will be, however, analysed in our future work.
The completeness of a measurement is directly dependent on the triangulation angle. The
smaller the triangulation angle the bigger is the number of measured 3D points because
shadowing and occlusion effects are reduced. Otherwise, if triangulation angle becomes too
small, z error increases. However, a difference of the completeness between a CP/CPE and
226
a CC/CCE measurement with CC = 2CP also depends heavily on the shape of the
measuring object.
In general, CP/CPE mode and CC/CCE mode with the same triangulation angle CC = CP
provide the same results in the accuracy. However, having disturbances in the phase
generation which similarly occur in both camera images (e.g. by indirect illumination,
irregular reflexes, phase generation errors, characteristic line errors, or not corrected
distortion of the projector optics) the error using the CC/CCE mode may be considerable
smaller than using CP/CPE mode as described by Munkelt (Munkelt et al., 2005).
The results are helpful for the decision of the method to be used taking into account the
typical measuring objects, the demands on the measurement, and the costs already in the
phase of system design.
A combination of the methods may improve a measurement using only one method
concerning accuracy and/or completeness in certain cases depending on the measuring
object as suggested by the authors (Bruer-Burchardt et al., 2010). However, this requires
additional algorithmic effort in the generation of software tools which has not been realized
yet for our systems. It also extends the calculation time and will not be used in fast speed
applications. However, first tests of the combination of the CP and VR methods have been
successfully performed in order to obtain data of regions which were hidden or in the shade
using only CC mode.
Together with the results described by Khmstedt (Khmstedt et al., 2009), a comprehensive
prediction of the expected error of 3D measurement systems using fringe projection can be
made depending on the properties of the measuring device, the exterior conditions and the
measuring objects.
As already mentioned in section 3 the main advantage of the CP/CPE method is the minimal
effort in equipment and calculation in contrast to the usage of the CC/CCE mode. There,
however, illumination caused errors may be reduced by the principle (Munkelt et al., 2005).
The extension by use of the epipolar constraint is always meaningful because it saves half of
the recording time and reduces the random error in direction perpendicular to the epipolar
lines. However, it should be realized that the calibration is correct and stable which can be
checked by additional measurements.
In order to design a new sensor the choice of the selected point localization principle depends
on several aspects like requested measurement accuracy, field of application, measurement
conditions (e.g. illumination, sensor mounting, measuring objects), costs, sensor weight, and so
on. Hence, this work cannot give the optimal way to perfect measurement data but it may help
to understand the influences on the measurement accuracy.
7. Summary and outlook

In this work some different correspondence localization methods which are usually applied
in 3D measurement systems using fringe projection were compared concerning the
measuring accuracy. A theoretic model for the random error of the measured 3D point
coordinates was established. A simulation tool was developed and applied. The results
obtained by these estimations were confirmed by real measurement data.

227
Future work is addressed on several aspects. First, the accuracy determination of the
VR/VRE modes should be improved and the error model should be potentially extended.
Here a number of new experiments must be performed.
Second, a combination of some of the methods should be merged into one measuring
system in order to obtain more complete results in the case of difficult measuring object
geometries. Here, new algorithms which realize an automatic 3D data fusion, a smooth
passage between regions of different measuring uncertainties, and an optimization of the
connected data must be developed.
Third, the VR/VRE methods can be extended in order to obtain equidistant distributed 3D
points as result of a measurement. This involves the realization of an adaptive virtual phase
raster for the point corresponding algorithms. Here, information about the shape of the
measuring object is necessary that may be obtained in a preliminary rough measurement.
8. References
Battle J., Mouaddib, E. & Salvi, J. (1998). Recent progress in coded structured light as a
technique to solve the correspondence problem: a survey. PR, Vol. 31, No 7, 963-982
Bruer-Burchardt, C. (2004). A Simple New Method for Precise Lens Distortion Correction
of Low Cost Camera Systems. Pattern Recognition (Proc 26th DAGM-Symposium),
Springer, 570-77
Bruer-Burchardt, C., Mller, M., Munkelt, C., Khmstedt, P. & Notni, G. (2010).
Comparison and evaluation of correspondence finding methods in 3D
measurement systems using fringe projection. Proc. SPIE Vol. 7830, 783019-1 783019-9
Bruer-Burchardt, C., Munkelt, C., Heinze, M., Khmstedt, P. & Notni, G. (2011a). Using
Geometric Constraints to Solve the Point Correspondence Problem in Fringe
Projection Based 3D Measuring Systems. Proc. ICIAP 2011, Part II, Springer LNCS,
265-74
Bruer-Burchardt, C., Breitbarth, A., Khmstedt, P., Schmidt, I., Heinze, M. & Notni, G.
(2011b). Fringe Projection Based High Speed 3D Sensor for Real-Time
Measurements. Proc. SPIE, Vol. 8082, 808212-1 - 808212-8
Bruer-Burchardt, C., Khmstedt, P. & Notni, G. (2011c). Error compensation by sensor recalibration in fringe projection based optical 3D stereo scanners. Proc. ICIAP 2011,
Part II, Springer LNCS, 363 -73
Brown, D.C. (1971). Close-range camera calibration. Photogram. Eng. 37(8), 855-66
Chen, F. & Brown, G.M. (2000). Overview of three-dimensional shape measurement using
optical methods. Opt. Eng. 39, 10-22
Creath, K. (1986). Comparison of phase-measurement algorithms. Surface characterization
and testing. Proc. SPIE 680, 1928
Ishiyama, R., Sakamotom, S., Tajima, J., Okatani, T. & Deguchi, K. (2007a). Absolute phase
measurements using geometric constraints between multiple cameras and
projectors. Applied Optics, Vol. 46, Issue 17, 3528-3538
Ishiyama, R., Okatani, T. & Deguchi, K. (2007b). Precise 3-d measurement using
uncalibrated pattern projection. Proc. IEEE Int. Conf. on Image Proc., vol. 1, 225-228
228
Khmstedt, P., Heinze, M., Himmelreich, M., Bruer-Burchardt, C. & Notni, G. (2005).
Optical 3D sensor for large objects in industrial application. Proc. SPIE 5856, 118127
Khmstedt, P., Munkelt, C., Heinze, M., Himmelreich, M., Bruer-Burchardt, C. & Notni, G.
(2007). 3D shape measurement with phase correlation based fringe projection. Proc.
SPIE vol. 6616, 66160 B-1-B-9
Khmstedt, P., Bruer-Burchardt, C., Munkelt, C., Heinze, M., Schmidt, I., Hintersehr, J.,
Notni, G. (2007b). Intraoral 3D scanner. Proc. SPIE vol. 6762, 67620 E-1-E-9
Khmstedt, P., Bruer-Burchardt, C. & Notni, G. (2009). Measurement Accuracy of Fringe
Projection Depending on Surface Normal Direction. Proc. SPIE vol. 7432, 743203,1-9
Li, E.B., Peng, X., Xi, J., Chicaro, J.F., Yao, J.Q. & Zhang, D.W. (2005). Multi-frequency and
multiple phase-shift sinusoidal fringe projection for 3D profilometry. Optics
Express. 13, 1561-1569
Li, Z., Shi, Y. & Wang, C. (2009). Real-time complex object 3D measurement. Proc. ICCMS,
191-194
Luhmann, T., Robson, S., Kyle, S. & Harley, I. (2006). Close range photogrammetry. Wiley
Whittles Publishing
Maas, H.-G. (1992). Robust automatic surface reconstruction with structured light. IAPRS,
Vol. XXIX, Part B5, 709-713
Munkelt, C., Khmstedt, P., Heinze, M., Se, H. & Notni, G. (2005). How to detect objectcaused illumination effects in 3D fringe projection. Proc. SPIE vol. 5856, 632-639
Notni, G. H. & Notni, G. (2003). Digital fringe projection in 3D shape measurement an
error analysis. Proc SPIE 5144, 372-80
Reich, C., Ritter, R. & Thesing, J. (2000). 3-D shape measurement of complex objects by
combining photogrammetry and fringe projection. Opt. Eng. 39, 224231
Rivera-Rios, A.H. & Marefat, M. (2005). Stereo Camera Pose Determination with Error
Reduction and Tolerance Satisfaction for Dimensional Measurements. Proc. IEEE
Int Conf on Robotics and Automation, 423-428
Sansoni, G., Carocci, M. & Rodella, R. (1999). Three-dimensional vision based on a
combination of Gray-code and phase-shift light projection: Analysis and
compensation of the systematic errors. Applied Optics, Vol. 38, Issue 31, pp. 65656573
Schreiber, W. & Notni, G. (2000). Theory and arrangements of self-calibrating whole-body
three-dimensional measurement systems using fringe projection techniques. Opt.
Eng. 39, 159-169
Tsai, R. (1986). An efficient and accurate camera calibration technique for 3-D machine
vision. IEEE Proc CCVPR, 364-74
Young, M., Beeson, E., Davis, J., Rusinkiewicz, S. & Ramamoorthi, R. (2007) Viewpointcoded structured light. Proc. CVPR, 1-8
Zhang, H., Lalor, M.J. & Burton, D.R. (1999). Spatiotemporal phase unwrapping for the
measurement of discontinuous objects in dynamic fringe-projection phase-shifting
profilometry. Applied Optics. 38, 3534-3541
Zhang, S. & Yau, S.T. (2006). High-resolution, real-time 3D absolute coordinate
measurement based on a phase-shifting method. Optics Express, Vol. 14, Issue 7,
2644-2649
12
Design Methodology to Construct
Information Measuring Systems Built
on Piezoresonant Mechanotrons with a
Modulated Interelectrode Gap
Alla Taranchuk and Sergey Pidchenko
Khmelnitsky National University,

Ukraine
1. Introduction
One of the perspective tendencies of development of information measuring systems is
applying the sensors of physical magnitudes on basis of quartz resonators (QR) of special
type, the stimulation of piezoelement in which is accomplished in the vacuum or filled with
dielectric gap. The changes of geometry (modulation) of the gap between piezoelement and
electrode (electrodes) of quartz resonator leads to the shift of personal resonance frequencies
of QR, as well as the oscillatory system in general, the part of which it is. Despite some
substantial advantages of presented way of QR control, piezoresonance sensors (PRS) of
given type were not widely used, which, to the authors opinion, is caused by inappropriate
study of theoretical and practical material of construction of PRS of given type. The material
of present chapter fills in the gap in the study.
Present chapter covers the theoretical issues, projecting and applying in highly informative
measuring systems of piezoresonator with modeled under the influence of mechanical force
the interelectrode gap and mobile electrode in the form of membrane, which is named
piezoresonance mechanotron (PRMT) (Kolpakov et al., 2009). We present mechanic scheme
of PRMT, examine its basic constructive and electrical characteristics, describe the regimes
of movement of mobile electrode. We propose approximate method of linearization of
graduation characteristic characteristics of PRMT while its use in primary measuring
transducers nonelectric magnitudes. Mechanic scheme, mathematical model and the results
of numeric simulation of a PRMT membrane bending flexure using ANSYS - programme
and a thermal PRMT model as well as numeric MATHLAB/ FEMLAB analyses of
temperature frequency characteristics under the excitation power variation of a PRMT
piezoelement and outer temperature change are presented. We present real life
piezoresonance mechanotron designs to be used as high precision surplus air pressure
transducers and the results of the development research of their precision characteristics.
Examples of PRMT applications as well as methods of secondary transducing of its output
signals in real life information measuring systems to define human hemodynamic par
meters are presented.
230
2. Quartz resonator as an element of precision primary measuring

transducers of nonelectric magnitudes - Quartz resonator frequency control
Quartz resonator (QR) represents electro - mechanical oscillatory system, conventional
equivalent scheme of which is shown in Fig. 1. Dynamic elements L 1 - equivalent
inductivity, C 1 - equivalent capacity and R 1 - equivalent resistance of losses are caused by
presence of direct and reverse piezoeffects and resonance peculiarities of piezoelement in
quartz resonator. As the quality factor of quartz resonators is very high and comprises tens
and hundreds thousands for usual resonators and several millions for precision ones,
equivalent dynamic branch of quartz resonator has a sense only in narrow band of
frequency next to resonance, and in other area of frequency equivalent electric resistance of
QR is determined by static capacity C 0 (Cady, 1946).
C1
L1
R1
C0
Fig. 1. Equivalent scheme of quartz resonator
The basic parameters of quartz resonator are the frequencies of successive and parallel
resonance
f1 =
1
and fa = f1 1 + m ;
2 L 1C 1
(1)
the quality factor and the width of interresonance interval

Q
f f
2 f1L 1
1
and a 1 0.5m ,
R1
2 f1C 1
f1
(2)
where m C 1 C 0 - is capacity factor.

At the same time dynamic inductivity L 1 is connected with equivalent mass of QR, and the
dynamic capacity C 1 is connected with elastic peculiarities (equivalent compliance).
Control the frequency in using QR as the piezoresonant sensor of physical magnitudes, as it
can be seen from the analyses of equivalent resonator circuit, can be realized:
-
by the influence on equivalent compliance ( C 1 ) or on equivalent mass ( L 1 ) of

oscillatory system or on both mentioned parameters at the same time. In accordance
with (1) in this case both resonance frequencies are varied;
by the varieties of active losses ( R 1 ). Increasing the decrement of damping 1 Q , as
is known, decreases the frequency own oscillations in accordance with the correlation
w w 02 2 .
Design Methodology to Construct Information Measuring Systems

Built on Piezoresonant Mechanotrons with a Modulated Interelectrode Gap
231
by managing the meaning of value of interelectrode capacity C 0 or capacity

correlation m C 1 C 0 .
The first method, which is based on the modulation of equivalent compliance C 1 or the
mass L 1 of oscillatory system, is widely used in piezoresonant sensors (PRS) of pressure,
micromoving, microweighing, humidity, gas analysis and others (Malov, 1989; Kolpakov &
Pidchenko, 2011).
As for construction of PRS the high quality factor resonators are used, in which the
decrement of damping is so small that it does not practically influence the resonance
frequencies, the second method of managing on frequency did not find wide application.
The third method is interesting from the point of view of applying in resonators of special
type (piezoresonant mechanotron), in which between piezoelement and electrode
(electrodes) there is the gap, vacuum-processed or filled with material with dialectic
permeability . Changes of gap geometry (modulation), as well as varying its electric
characteristics, leads to the shift of resonance frequency of oscillatory system as a whole.
The given way, in comparison with the first, in some cases has certain advantages, for
example, at measurements of micromovings and physical magnitudes led by it. It is
connected with an exception of direct mechanical influences on a sensor piezoelement that
leads to change (deterioration) of its characteristics as highly quality and highly stable
element of oscillatory system of the quartz crystal oscillator.
3. Design methodology of piezorezonant mechanotrons built on quartz

resonators with a modulated interelectrode gap
Introduced PRMT is related to measuring converters with control of its resonance frequency
by changing magnitude of gap between launch electrodes of piezoelectric element (PE).
Change of parameters of oscillatory system under action physical magnitude (moving and
reducible magnitude) is used in PRMT. PRMT design map is shown in Fig. 2, (a), where x 0 initial gap; x m - magnitude stroke mobile electrode (membrane); x x0 xm - current value
of modulated interelectrode gap (Kolpakov F., Pidchenko S., Taranchuk A., et al, 2009).
Connection of capacitive parasitics with minor gap capacity enabled to reduce influence on
the resonance frequency of oscillatory system. Modulated gap application, i.e. measuring
capacitor connection inwards construction allows raising capacitive-ratio, and therefore
tuning range of oscillatory system.
Piezoresonant mehanotron with modulated interelectrode gap (PRMT with MIG) equivalent
electric circuit is shown in Fig. 2, (b). Following descriptions are made: L 1 ,R 1 ,C 1 - PE
dynamic equivalent parameters on the fundamental mode of the thickness-shear vibrations;
C
C PE - PE static capacity; 1
- capacitive ratio (for AT-cut crystal m 4...7 10 3 );
C PE
C 0 C m par - capacitive parasitics with mounting capacity.
Rather important is the fact, that parasitic capacitive is connected through small capacitive
gap and therefore it has little influence onto resonance frequency oscillatory system.
232
PRMT by its electric equivalent circuit is similar to any capacitor the operated quartz
resonator. However, the use of modulated gap, e.i. connection of measuring compensator
into the construction, dives the chance to increase considerably the capacitor relation, and
therefore, the band of oscillatory system restructure. For example, on frequency 10 MHz
under the gap modulation we can get the frequency deviation (2530) kHz.
Lets define PRMT characteristics. Circuit equivalent resistance is:
R 1k 1 jL 1k 2
where
k5
k4
(3)
k4
1
1 C PE
;
C 1 Cg Cg
C
k
jC 0 R 1 1 C PEk 3 jL 1 1 C PE k 3 j PE 5
C0C1
k1 1
C PE
;
Cg
k2 1
C PE
;
Cg
k3
1
1
;
Cg C 0
1
1
1
C
PE - coefficients.
C 0 Cg C 1 Cg C 1
In conditions of a resonance Im Z 0 , we obtain:
where n 4 L 1C 0 , n 2
n 4 4 n 2 2 n 0 0 ,
C PER 12 C 0 Cg
L 1 Cg C1
,n
(4)
C0
.
C 1 L 1C g
Solving (4) for can be written as follows:

2
r,a
2C 0L 1 C 1 Cg 2L 1C 0
C PER 12 C 0 Cg
C PER 12 C 0 Cg
L 1 Cg C 1
4C 02
.
C 1 Cg
(5)
For typical valuations of QCR parameters: R 1 10 hm, C 1 22 pF, L 1 11,5 H ,

C 0 (110) pF , C PE (2...3) pF, C b (0.1...1) pF. On basis of measurement results
characteristics of QR control are built. They are shown in Fig. 3.
f,Hz
58
f,10Hz4
43.5
.
1.510
1044
29
. 4
1.010
104
14.5
0.5.1044
10
2.25
2.5
2.75
0.88
0.7 5
0.63
0.5
PE, pF
(a)
0.38
0.25
0.13
0 g, pF
(b)
Fig. 2. (a) Dependence of absolute detuning on PE; (b) Dependence of absolute detuning on g

233
The analysis of received dependence shows, that the use of frequency control with the help
of interelectrode gap rather effective and under little changes of volume of clearance allows
to get information deviation of frequency of order of tens KHz.
(a)
(b)
Fig. 3. (a) PRMT with MIG design map; (b) Equivalent electric circuit of PRMT
For real PRMT with piezoresonator AT - cut (diameter of piezoelement dPE 18 mm,
electrode diameter is de 8 mm, nominal frequency of piezoelement fn 10.0 MHz,
F0 6500 Hz) experimentally was received modulation (graduation) characteristic, shown
in Fig. 4.
150
200
Fig. 4. Graduation characteristic of PRMT with MIG

It is obvious that it can be symbolically divided into two sectors (see Fig. 4). The first one,
OA - is the area with maximum deviation and nonlinearity. The second one, AB - is the
linear area of the curve, which is characterized by relatively small frequency changes. This
allows to talk about two working conditions of PMRT: nonlinear and linear. The work in
nonlinear conditions allows to achieve maximum resolution capability, in linear - maximum
linearity of modulation characteristic under the demodulation of measuring signal by the
linear frequency detector. Using PE in PMRT of pressure it is reasonable to use nonlinear
conditions, which provides maximum sensitivity of PMRT, but allows to linearity
graduation characteristic.
234
As opposed to PMRT it is possible to use external capacitive frequency control (ECFC),

under which in the capacity of sensor the capacitive sensitive element, playing a role of Cg ,
is used.
We propose the comparative characteristic of external capacitive frequency control and
control of modulation of interelectrode gap.
Equivalent electric circuit of QR with ECFC shown in Fig. 5. Standard piezoresonanator is
used and is connected sequentially with tuning inductance L t and capacitive sensor of
pressure t p (see Fig. 5).
Fig. 5. Equivalent electric circuit of QR with external capacity control

Comparing Fig. 1, (a) and Fig. 5, we can mark, that parasitic capacitance C 0 , including also
capacitance of assembling, on schemes it is included on a miscellaneous. Under external
capacitive control it is connected parallel to static capacitive of PE. As frequency deviation is
directly proportional to the capacitor relation
m
C1
C1
,
C PE C PE C 0
(6)
so controlling the oscillatory system with an increase C 0 considerably worsens.

As controlling the modulation of interelectrode gap C 0 (see Fig. 2) is connected through
small capacitance of gap and practically does not influence the size of static capacitance, i.e.
m
C1
.
C PE
(7)
From here a relative benefit on controllability of a variant with gap modulation

kc
C
m
1 0 .
m
C PE
(8)
This fact is confirmed by the calculation of deviation of frequency for two types of control
according to formulae (5) (see Fig. 6).
Comparing these two types of control (see Fig. 6) shows multiple benefit of PMRT on the
conversion conductance and the preference of its use. At this time in some devices of not
high accuracy, for their reduction in price, the ECFC can also be used (Pidchenko, Kolpakov,
Akulinichev, 2000).

235
4. Linearization of a PRMT modulation characteristic based on fractionallinear approximation of graduation characteristic

Having written down (5) in terms of graduation characteristics, well receive expression for
resonant frequency PRMT:
Fig. 6. Frequency mismatches comparison of ECFC and PRMT with MIG
f x f0
m
0.5m
,
f0 1
1
h
h
1 PE
1 PE
PE x
PE x
(9)
where f0 - frequency rating of oscillating system with x 0 , m - capacity ratio; h PE - PE

dimension which determine frequency (thickness); PE - dielectric conductivity of PE
material; x - current gap value.
Let us introduce the PE parameter a
h PE
PE
N
, where N - is frequency coefficient (for
f0 PE
AT cut N 1661 kHzmm). From (9) we get the relative frequency deviation of measuring
transducer (MT):
F x
f x f0
f0
0.5mx
.
xa
(10)
Taking into account that x x0 xi (10) it can be written in such a way
F xm
0.5m x0 xm
x0 a xm
Fractional - linear function (9) has such look:
(11)
236
F xm
a1 1 a 2 xm a 2 a0 a1xm
1 a2 xm
a0 a1xm
,
1 a 2 xm
(12)
where a0 ,a1 ,a 2 are the coefficients of approximation of normalized fractional characteristic:

a0
0.5mx0
0.5m
1
; a1
; a2
.
x0 a
x0 a
x0 a
(13)
We can show at PMRT with parameters, which are character to pressure transducer. Let
x 0 0.1 mm, m 6.29 10 3 , a 0.03691 mm, xi 0...0.06 mm. We substitute initial data
into (13), find coefficient of approximation of normalized fractional characteristic:
a0 2.297 10 3 , a1 2.297 10 2 1/mm, a 2 7.304 1/mm. Hence:
x m xm max 0.06 mm, F xm 1.6357 10 3 ;
xm
xm max
0.03 mm, F xm 2.059 10 3 ;
xm x m min 0 mm, F x m 22.97 10 3 .
Complete reorganization of output frequency of MT under given parameters of oscillatory

system is 22.97 kHz, and f 0 corresponds own frequency of quartz plate and
corresponds f0 10 MHz.
For PMRT of micromovings the variable x 0, x0
characteristic corresponds (10). In accordance with this
F x
and normalized graduation
a0 a1x ,
1 a2 x
(14)
m
1
, a2 .
a
a
Having differentiated (14), we get the slope S F of characteristic PMRT:
where a0 0, a1 0.5
SF
a1 a 2 a0
1 a 2 x 2
(15)
We write also the formula for measuring the equivalent dynamic resistance of quartz
oscillatory system R ekv under variation of interelectrode gap:
2
C
R 1 x PE
R 1x
R ekv R 0 1 0
0
0
C air
h PE
a
(16)
where R 0 is a dynamic resistance of under x 0 , C 0 is static size of QR, C air the size of air gap.
These parameters are necessary while projecting PMRT, and approximation is essential for
the synthesis of linear graduation characteristic (Kolpakov, Pidchenko, Hilchenko, 1999).

237
5. Accounting for a PRMT electrodes nonparallelity

All given above correlations are received in assumption that electrodes of PMRT are strictly
parallel to each other, however it's practically unrealistic. Firstly, it is extremely difficult to
provide the parallelism of electrodes while producing PMRT; secondly, the slightest
mechanical stress, which appears in the construction under the effect of outer factors can
lead to micron defects.
Taking into account everything said before, we take into account nonparallelism, defining
the condenser capacity with nonparallel flat electrodes (see Fig. 7) with the radius R of static
and R ' R of movable.
X3
R
h3/2
O
h3
R
dS
X2
B
O
X1
dx 2
Fig. 7. Record of nonparallelism of PE and flat electrode

The area of the element with the width dx 2 equals
dS=AB dx2 2 R 2 x 22 dx 2 .
(17)
The distance to appropriate element of upper cover equals

x 3 OO+
h 3 x2
h3 h3
h3
x2 .
2
R
2
2R
(18)
Elementary capacity between elementary areas
dC=
0 2 R 2 x22 dx 2
(h 3 h 3 / 2) x 2
h3
2R
(19)
whence full capacity of condenser with nonparallel relatively to the axis x 2 electrodes is
C
R 2 x 22 dx 2
.
h3
x h3
) 2
R (h 3
2
2R
dC 2 0
(20)
238
Implementing the transformation of under the integral formula (20), we move from the form
R
dC S
Z dx 2
,
p x2
(21)
h3
4 0 R
; S
, to the form
where Z a bx 22 ; a R 2 ; b 1 ; p
1 R ; h
h
h3
3
h
R
R x 2dx 2
dx
2
dC
S
b
p
Z
Z2 a bp
R
R
R
R
(x
R
dx 2
2 p) Z
(22)
Using the table meanings of integrals in (22) and performing some transformations, we get
C
0 R 2
h 3 np
(23)
where np 0.5 0.25 h3 0.5 1 h3 is the function of nonparallelism; h3
h3
h3
In the terms of graduating characteristic of QR with MIG the relation

C3
h PE
a
,
C 0 np PE x np x
(24)
whence normalized graduating characteristic of PMRT in the conditions of measuring the

micromovings
F (x)
0.5m np x 0.5mx
0.5m
.
a
np x a
x a
1
np x
(25)
The analysis for np 1 and h3 0,1 shows that the display of nonparallelism is
equivalent to increasing the gap between mobile and static electrodes in npi times for
h3 h3i .
In the condition of changing the pressure, the normalized graduating characteristic PMRT
under the record of nonparallelism of electrodes has the image
F xm
where a 0
0.5m np x0
np x0 a
, a 1
a 0 a 1xm
,
1 a 2 xm
(26)
0.5m
1
, a 2
.
np x0 a
np x 0 a
Hence, the increase of nonparallelism of electrodes (rotation of mobile round the axis X 1 ) is
equivalent to an increase of initial gap x 0 .

239
Breaking the flatness of two surfaces usually occurs by means of moving of one of it at the
same time relative to two axes X 1 X 2 . That is why in general case the function of
nonparallelism can be presented as following
np np1 np2 ,
(27)
Where np1 h31 and np2 h32 are partial functions, which allow for rotation of mobile
electrode round the axes X 1 X 2 accordingly. Independence of these rotations define the
type np1 and np2 according to (24).
Therefore, nonparallelism of electrodes of PRMT can be taken into consideration on a stage
of its projecting and does not influence linearity of graduating characteristic (Kolpakov,
Pidchenko, Taranchuk, et al., 2009).
6. Special features of piezoresonant mechanotrons built on a resonant

membrane
Building the measuring force transducers and adduced to it physical quantities (pressure,
motion, etc), which use the effect of tensosensitivity QR, is connected with solving
complicated problem of connecting force transducers element with quartz piezoelement.
Wide possibilities in solving of these tasks are opened while using non- contact frequency
control of QR (quartz resonator) by means of modulating the interelectrode gap (Kolpakov,
Akulinichev, 1999).
PRMT is shown in Fig. 8 combines the use of two types of control: tensosensitivity of quartz
resonator and its control over the modulation of interelectrode gap. Under the activity of
applied pressure in the plane of piezoelement, the mechanical strains , and as a result of
which the PE changes its properties, that is its resonance frequency changes. This is the
demonstration of the first control mechanism.
The second control mechanism consists in the following: the resonant membrane, which is
jammed on the contour, is under the effect of spread air pressure P . Under the effect of this
pressure there happens the PE bending flexure and, as a result, the extension of
interelectrode gap xg , which in its turn leads to the reduction of interelectrode gap
capacity Cg . Under the influence of this pressure there is deflection and, as
Fig. 8. PRMT structure: h - thickness RM; d - RM contour, free from a jamming
240
consequence, augmentation of an interelectrode gap that in turn leads to reduction of

capacity of an interelectrode gap.
Numerical modeling of tense - strained state RM of round (elliptic) shape, is performed on a
basis of the system of engineering analysis ANSYS, which uses the method of finite
elements.
Taking into account anisotropic properties of the model the following constants are set: the
model of elasticity Ex 78 GPa, E y 85.3 GP, Ez 92.6 GP, coefficient Poisson
0.077 and module of shift G xy 42 GP.
Taking into account the character of loading (the bending flexure is small in comparison
with the plate width) and condition d / h 10 we use the model for thin plate. For such
model normal to middle plane until the curve remains normal to this plane after curve as
well. That is why the deformation of shift is absent, i.e. xy yz 0 .
Taking into account geometry and force symmetry of calculated model, it is enough to look
at the sector of membrane, which is limited by main axis, with setting the corresponding
limited conditions: absence of moving and rotation in plane; plane of symmetry is xy
u z 0 ; plane of symmetry yz u x 0 ; xy yz 0 . Such assumptions allow to reduce
considerably the dimension of being solved task. Jamming the free edge of plate is modeled
by means of fixing the junctions, which are situated in the area of jamming, in the plane xy
(see Fig. 9).
The plate is loaded with excess pressure P 4 10 4 Pa (300 mmHg), which operating in a
plane of a cover.
While solving this task the geometrical nonlinearity - the changes of cylindrical rigidity of
the cover in the deformation process is taken into account. This is provided by means of
regenerating (recalculation) of matrix of rigidity after each iteration. Solving the system of
nonlinearity equations is done with the help of the method of Newton-Rawson with
automatic change of iteration step.
(a)
(b)
Fig. 9. Finite - element model of RM with visualization of limit conditions and the enclosed
pressure: a - round; b - elliptic

241
As a result of calculations for RM with round cover of jamming and the

parameters h 169 m and d 11 mm the values of jamming u z and normal tension x
are received, z (see Fig. 10). The calculation of parameters RM with elliptical cover of
jamming was done for a 5, 5 mm and b 4, 4 mm (see Fig. 11).
The analyses of received data in the process of numerous modeling shows, that the value of
motion RM for round jamming comprises approximately 20 m , elliptic - 13 m , and the
reserve of solidity on mechanical tensions, penetrating into plane of piezoelement,
comprises 5.5.
The results of modeling showed high effectiveness of proposed methods on the basis of
method of finite elements for performing high - precise calculations of complicated
constructive elements of piezoquartz measuring transducers on the stage of their projecting
and exploitation, making analytical decision for which is impossible.
(a)
(b)
Fig. 10. Calculations of parameters of round RM: (a) bending flexure; (b) tension x (1) and z (2)
7. Analyses of a piezoresonant mechanotron temperature characteristics

employing modulated interelectrode gap
In accordance with (12) and (14) temperature sensitivity of QR with IG is determined by
the variations x0 T , m T , PE T , h PE T . Let us look at factors more detailed, which
contribute into the changes of information frequency parameter.
Permittivity of quartz depends very little on orientation of plates, its capacity temperature
coefficient (T) under increasing the temperature to 100C is little and in general it
determines the character of temperature dependence C 0 T ( TC PE 5 10 5 / C 25% ).
Temperature coefficient C0:
TCC 0 C 0
20 1 5 10
0
20 .
(28)
242
(a)
(b)
(c)
(d)
Fig. 11. Calculated parameters of elliptic RM: (a), (b) - bending flexure of membrane on axis
a and b of ellipse accordingly; (c), (d) - tension x (1) and z (2) - on axis a and b of the
ellipse accordingly
Temperature dependence of dynamic capacity Cq T for flat PE - cut:
TCC q C q
20 1 2.5 10
20 2 10 7 t 20 .
(29)
For the wide temperature interval (-60+100)C quadratic terms can not be taken into
account. Temperature dependence h PE T is estimated by the value of temperature coefficient
of linear expansion (LE 10 5 cm / C ). Dependability x0 T is defined by LE of metal
envelope, glued material, difference of LE frame and regulated screw, termal deformations
of mobile electrode (membrane). In accordance with (12), taking into account said before,
1f T
a0 T a1 T T x x 0
1 a 2 T T x x0
(30)

where a0 T a0
1 m x0
1 x a
a 1 T a 0
1 m x0
1 x 1 x a
0
; a2 T
243
a 1 T
0.5m 1 m
It is known that detuning of relative frequency of successive resonance leads to a turn of the
temperature frequency characteristic (F) QR. Dependence of TF on detuning depends
on the change of capacitive relation of resonator in the temperature interval and is
calculated
f0 mcrp 0.5m m e y T T0 ,
(31)
where m is the temperature coefficient of capacitive relation (for resonators AT-cut

m 3 10 4 , e y - is relative detuning as to the frequency of successive resonance of QR
with IG when x 0 , T- is current value of temperature , T0 - temperature, relative to
which the change TF is defined. Therefore, the relative change of frequency
a 0 T a 1 T x 0
2f T
1 a 2 T x0
m T.
(32)
The influence of temperature warp of electrodes and changes x 0ecv are taken into account by
introducing the correction to the value of initial gap:
3f T
a 0 T a 1 T x0 x
1 a 2 T x0 x
(33)
It is necessary to take into account the changes of frequency, which are connected with
temperature instability of oscillator, piezoelement and occurred, because of temperature
changes of construction element size, strain load of PE. Therefore
4f T fget T fa T K F T.
(34)
Taking into account all said above, we write approximate model of termal- sensitivity of
measuring transducers with QR IG:
f T 1f T 2f T 3f T 4f T
a0 T a 1 T x0 x0
1 a 2 T x0 x0
a0 (T) a1 (T)( T x x0 ) a 0 (T) a 1 (T)x0
m T
1 a 2 (T)( T x x0 )
1 a 2 (T)x0
(35)
fget T fa T K F T.
Having grouped the components, we get
f T
a0 T a1 T x0 x T x0
1 a 2 T x0 x T x0
1 m T fget T fa T K F T.
(36)
Dependence of frequency of QR on the temperature is defined as physical properties of

crystal element, as well as the size and material of its electrodes, quartz holders, the
topology of fastening of a piezoplate. At fast changes of temperature (temperature blow) the
244
main reason of instability of frequency is the presence of temperature gradients in quartz PE

and mechanic tension, which appears in the space of piezocrystal, as well as in the place of
electrodes placement and fastening of quartz holders under not uniform resonator heating
(Kolpakov, Pidchenko, Taranchuk, et al., 2009).
Let's survey geometrical model of node of the quartz holder piezoresonator. Researched
construction (see Fig. 12) usually consists of two parts: quartz piezoplate (1) and quartz
holder (2), which is produced of duraluminium DT-16.
Fig. 12. The construction of a node of quartz holder of piezoresonant sensor with MIG
Solving the equation of term heat conductivity
k T Q ,
t
(37)
where is density, C is heat capacity, k is coefficient of heat conductivity, Q is the

meaning of heat source, by the method of finite elements in the system of mathematical
modeling MATLAB (FEMLAB) under abrupt changes of temperature from 0C to 50C
(limit conditions of third type) we get the distribution of temperature field of sensor PE in
character time moments (see Fig. 13) We use received information for calculating instability
of PRMT frequency.
For definition of the temperature induced shifts of frequency piezoelectric resonator we will
present the general shift of frequency in a kind
f
f0
f TR T0
f T
,
f0
f0 stat
dyn
3
f T
i
where
ai T T
f0 stat i 1
(38)
is static component of instability of PE, which is

T TPE
min
defined by its static temperature- frequency characters and minimal temperature TPE
min

245
f T0 TR
- is dynamic component of instability, which originates by means of
f0
dyn
distortion of temperature field of PE.
Fig. 13. Distribution of temperature field in character time moments

Static (quasi - static) thermal behavior of PE with enough accuracy is defined by the
polynomial of third degree. For defining the dynamic component of instability of frequency
we use the methods, which are described in (Taranchuk, Pidchenko, 2002 - 2005). The
relative shift of personal frequency of piezoplate fluctuations under the effect of external
force F applied onto the piezoelement ends of disc shape along its diameter, is
calculated as
f f0
K ( ) F( ),
2R n f
f0
(39)
where - is the azimuth of force applying F ; f0 - is the minimal fluctuations frequency;

n - is the number of mechanical harmonica; R is the radius of PE; K f ( ) is force - frequency
coefficient of Rataiskiy.
246
To find F we take into account the fact, that tensor of elastic deformations of
elementary volume of piezoplate is bound with changes of temperature T by a parity
rij l ij T , where l ij are the components of tensor of second rank of quartz thermal
expansion, and force tension, which originate in this are depicted by generalized law of
Gook t ij C ijkl rkl dij T , where C ijkl are the components of the tensor of the forth rank
of modules of frequency (Cady, 1946).
Realizing the transition d d 1 into the plane of piezoplates of the most prevailing onerotary yxl and two- rotary SC, FC yxbl - the cuts of piezoelements and
using with a glance of geometric peculiarities of PE the cylindrical coordinate system, we
get:
Fr ( ) drr ( ) S Tr ( ) ,
(40)
where Fr are changes of allocated forces, which act in the volume of piezoplate along
the radius r ; S R h is the area of an elementary platform of a face surface of PE, h
is the thickness of PE.
Taking into account (40) in parity (39) and carrying out integration on in the limits from
0 to and on r in the limits form 0 to R we come to integral from of writing
f f0
h K f ( ) drr ( ) TR T0 d ,
f0 2n 0
where T0
(41)
1
1
T0 h dh , TR ( ) TR (h, )dh - average in the volume temperature in the
h 0
h0
centre and on the edge of PE.

Received on the basis (44), dynamic components of instability of frequency for two types of
quartz - holders are demonstrated on Fig. 14, (a), where curve 1- is the dynamic instability of
frequency of quartz holder with four points of fastening of PE (see Fig. 12), curve 2 - for
quartz holder with thermal contact on the contour of PE. It can be seed, that in second case
there are big deformations of temperature area of PE, which cause therefore much bigger
instability of frequency (approximately order).
On Fig. 14, (b) there are settlement (in accordance with (38)) and experimental curves of
general shift of frequency of sensor under the thermal effect (from 50C to 0C). Good
coincidence of curve data proves the adequacy of used mathematical model, which can be
used for further research.
The construction of quartz holder influences greatly the general shifts of frequency of PE
PRMT. Under this there appear considerable temperature gradients and mechanical tension
in piezoplate, which lead to substantial increase of dynamic component of instability. Its
maximum values can exceed the frequency shifts, which are caused by quazi-static
temperature frequency characteristics (TFC). That is why the construction of PRMT must be
optimized with a glance of thermal processes, which take place in quartz holder and
piezoelement (Taranchuk, Pidchenko, 2002 - 2005).

247
Fig. 14. Dynamic component (a) and frequency shift of piezoelectric resonator (b): curve 1 is the dynamic instability of frequency of quartz holder with four points of fastening of PE;
curve 2 - for quartz holder with thermal contact on the contour of PE; 3 - settlement curve;
4 - experimental curve
8. Implementation of piezoresonant mechanotrons employing modulated

interelectrode gap and development research
On the basis of research, the results of which are presented above, the authors worked out
the number of sensors of excessive pressure for medical appliances.
On Fig. 15 the PMRT of pressure is presented, in which for modulation of inter electrode
gap the metal membrane is used. The case 1 made of polystyrene is connected with metal
basis 2. Between top groove of base 2 and union node 11 there is seal ring 12. In a ground
part of the basis 2 the node quartz holder 6 which is rigidly bridged by means of a rivet 4 to
an elastic element 5 is established. On a working surface quartz holder 6 the flat disk
piezoelement 3 AT - cut with the round electrode bridged electrically with a potential
conclusion of the sensor 7 is established.
Elastic element 5 through the sealing ring 8, and also the membrane 10 are jammed on a
contour by means of a clamping node 11. All nodes of PMRT have hard fixing and do not
need applying welding operations for junctions. The size of initial gap between quartz
piezoelement 3 and membrane 10 is provided with changes of the flexure of elastic element
5 with the help of screw 9.
In the condition of absence, towards atmosphere, air pressure in the cuff, which is connected
by the tube to the nipple part of frame 1, the deformation of membrane 10 is absent, as inner
volume of basis 2 is not pressurized and the pressure onto the membrane from both sides is
identical. In such condition quartz resonator, which is formed by piezoelement 3, metal
membrane 10 and switched into the scheme auto generator PRMT is stimulated on the
frequency which corresponds zero excessive pressure. When pressure available exceeds the
atmosphere one, the membrane 10 is bended, the gap capacity between unused surface of
piezoelement 3 and the surface of central part of membrane goes down: it leads to
decreasing the resonance frequency of QR with variable gap (Kolpakov, Pidchenko,
Hilchenko, 1998).
248
Fig. 15. Construction of PMRT with metal membrane

The basic advantage of this PRMT is complete absence of hysteresis effects, which are
inherent in sensors with direct influence onto piezoelement, which substantially increases
the reproducibility of graduating characteristics and, as a result, accuracy of measurements.
Typical value of permitted ability of given PRMT with metal membrane comprises
0.050.08 mmHg, and main error does not exceed 0.10.15%. Highly effective appliance of
PMRT with capacity pneumatic control is its use in measuring transducer of pulse air
pressure of sphygmographic research (Taranchuk, Pidchenko, et al., 2010).
Pneumatic scheme of PMRT is show Fig. 16, (a) has two channels: K1, which has pneumatic
the filter of the bottom frequencies, which consists of pneumatic resistance of filter R f ,
pneumatic throttle R tr and under membrane capacity VK1 ; 2 - is direct channel of action
onto membrane. Resulting influence of two channels onto the movement of membrane is
equivalent to the operation on given pressure P P~ P0 to the pneumatic filter of the top
frequencies. Here P0 - in constant (slowly changeable) and P~ - variable (informative)
component of pressure.
The frequency of cut PRMT is defined by the formula
fcut
Where C FC
1
,
2 R C FC
(42)
V1
is pneumatic capacity of the filtrational chamber VK1 ; R R f R tr is
T
total pneumatic resistance; V1
2
D FC
h FC
, D FC and h FC - the volume, height and
diameter of filtrational chamber accordingly; 287
is gas constant; T - is the gas

temperature (air temperature) .

Resistance of cloth filter is defined as
RF
F
,
S
(43)

(a)
249
(b)
Fig. 16. (a) Pneumatic scheme of PRMT; (b) The amplitude- frequency characteristics of
sensor of pressure for values l tr 1 mm
where F is pressure drop on cloth filter; - speed of air flow through filter; S dF2 / 4 the area of cross- section of cloth filter opening, dF is diameter of opening of cloth filter.For
the material for cloth filter, type felt: PF 49 P, 0.02 m/min and for dF 1.5 mm R F 8.32 1010 Pa sec/ m 3 .
Resistance of throttle equals
R tr
128 l tr
,
d tr4
(44)
where l tr , d tr are the length and diameter of throttle; - the density of air under normal
conditions, 1.205kg / m 3 ; is dynamic coefficient of gases viscosity; , are
constant coefficients ( for air 37.4 10 7 P sec, 0.506 10 7 P sec ). In normal
conditions ( 293K ) the dynamic coefficient of air viscosity 0.186 10 4 P sec .
The cut frequency of PRMT s defined by the spectrum of sphygmograph signals
(0.03660) Hz and comprises approximately fcut 0.03 Hz. Having based on chosen cut
frequency, taking into account real size D FC 40 mm and received cloth filter
R F 8.32 1010 Pa sec/ m 3 , we let's define in conformity with (42) - (44) the constructional
parameters of throttle (diameter d tr and the length l tr ) and the height of filter chamber h FC
(see Table 1, Table 2 and Fig. 16). Analyses of received data shows that substantial influence
onto the cut frequency fcut makes the height of filter chamber h FC (see Table 1, Table 2).
For d tr 0.1 mm the resistance of throttle R tr considerably depends on l tr and should be
taken into account while calculating PRMT (Fig. 16, (b)), for d tr (0.2...1) mm and
l tr (1...10) mm the throttle resistance l tr is much smaller that the resistance of cloth filter
R f R tr and can not be taken into account while calculating total resistance R . In such
case parameters PRMT are defined as usual by the height of filter camera h FC (see Table 2).
250
fcut , Hz
h FC , mm
l tr , mm under d tr 0.1 mm
1
0,11
0.0595
0.0397
0.0298
0.0238
0.0119
1
2
3
4
5
10
2,5
0.107
0.0539
0.0359
0.0269
0.0215
0.0108
5
0.093
0.0465
0.031
0.0232
0.0186
0.0093
10
0.073
0.0365
0.0243
0.0182
0.0146
0.0073
Table 1. Choosing the geometry sizes of elements of PRMT

h FC , mm
10
fcut , Hz
0.128
0.064
0.0427
0.032
0.0256
0.0127
Table 2. Choosing the height of filter chamber while d tr (0.2...1) mm and l tr (1...10) mm
Thus, proposed construction PRMT with successive connection R f and R tr allows to vary
the parameters of throttle in accordance with technological possibilities of production and to
realize it without applying special capillary technology (Kolpakov, Dobrova, et al., 1999;
Pidchenko, Taranchuk, et al., 2002).
The adantage of given construction of PMRT is in following: the capacity of interelectrode
gap is defined only by the amplitude of pulse wave of blood pressure and technological
tolerance of membrane mounting, and additional errors of nonparallelism of membrane and
PE are excepted by means of removal of: quazi - static component of air pressure. This
allows to increase the allowing ability of measuring transformer for sphygmographic
research on basis of PMRT while measuring dynamic air pressure in more than three-five
times as compared with known devices.
On Fig. 17 it is presented the third type of frequency PMRT of pressure with the use of
resonant membrane (RM) , which is developed in paragraph 6 of given chapter.
PMRT with PM works as follows. Under the absence of excessive as to atmosphere pressure of
air in the chamber, which is limited by piezoelement 5 and inside surface of cover 1, the
deformation of piezoelemet 5, which plays the role of resonating membrane, does not happen.
Quartz resonator, connected in the scheme of oscillator PRMT is stimulated on the frequency,
which corresponds zero excessive pressure. From the exit of oscillator the information signal is
taken, the frequency of which f0 corresponds the beginning of modulation characteristic of
PMRT. Under the pressure in the pressure chamber, which exceeds the atmosphere pressure,
there happens the small bending flexure of piezoelement 5, as a result the size of a gap
between the free surface of piezoelement and the surface of cylindrical bump would increase,
which leads to increasing the frequency f P . In given PRMT two mechanisms of controlled
of frequency of the quartz resonator take place. Capacitor controlled, at the expense of
change of size of a gap between a free surface of a piezoelement and a surface of a cylindrical
ledge 6, and controlled on the basis of effect tensosensitivity:

p C p p ;
K
251
(45)
f
N
K f d ,
f0 P 2 0
(46)
where N is frequency constant; K f is the coefficient of Rataiskiy; is azimuth of load; f0 is

personal resonance frequency of piezoelement ( f0 10MHz ); P is applied spread pressure.
Fig. 17. Construction of pressure sensor with PM:

1- cover; 2 - regulating nut; 3 - annular groove; 4 ring; 5 - piezoelement; 6 - cylindrical
ledge; 7 - basis; 8 - compactor; 9 - potential lead; 10 - mobile electrode; 11- hermetic; 12
samples; 13 - basis ledge; 14 - groove; 15 - two pairs of diameter - opposite basis ledge
Experimantal researches were held for PMRT with PM with PE of diameter 14 mm and
electrode diameter 4 mm under P 0, 300 mmHg. Static modulation charactetistic of
PMRT with PM (see Fig. 18) includes demonstration of both mechanisms of changes the
Fig. 18. Summarized modulation characteristic for PMRT with PM
252
resonance frequency. For estimating the impact of the mechanism of tensosensitivity a

separate experiment was held, in which PM of the same size contained two electrodes. Thanks
to this the capacity mechanism of frequency was excluded. The analysis of tensosensitivity of
mechanisms (see Fig. 19) shows that impact of tensosensitivity of component comprises
(2.53.5)%. Therefore, the use of model, which takes into account only capacity mechanism of
control, for PMRT with PM brings the error, which does not exceed 3.5 %.
Fig. 19. Tensosensitivity modulation characteristic for PMRT with PM

The temperature error of proposed PMRT is substantially decreased thanks to two
peculiarities of its construction.
Firstly, surfaces of a piezoelement and the metal surfaces jamming it, have small
roughnesses, therefore the forces of a friction arising because of difference of temperature of
Rataiskiy equals zero (see Fig. 20).
This means that in ideal case this component of temperature error equals zero, and
practically it is very little.
Fig. 20. The construct of quartz holder

253
The increase of deviation of frequency MC in one and half times can be achieved by
increasing the diameter of PE. The results of experimental researches of PMRT with PM
18 mm is demonstrated in Fig. 21. As for PMRT with disc flat PE of AT-cut with the
diameter 18 mm, the diameter of electrode is 8 mm and thickness h PE 169 m which
corresponds nominal frequency f0 10 MHz, while applying the pressure from 0 to 300
mmHg, the maximal value of banding flexure of piezoelement in order less than its
thickness and comprises 16.23 m . Elements and construction of PMRT with PM are
demonstrated in Fig. 22.
Fig. 21. The relative characteristic of sensitivity
Fig. 22. Elements and construction of PRMT with PM

Typical technical characteristics of sensor of pressure on basis of PMRT with PM:
1.
Resolution in the range of applied pressure P 0, 300 :
2.
3.
4.
for small pressure, less that 10 mmHg.................................0.008 - 0.01 mmHg;

for maximum pressure not worse......................0.025 mmHg.
Full change-over of frequency, not less than ..............................9.5 kHz.
The value of hysteresis after maximum load, not more than ............................2 Hz;
Basic error of measurement the surplus pressure, not more than.............................0,15%.
Coefficients of approximation of characteristic of transformer of PMRT, calculated in

accordance with (13) form: a0 1938.87Hz, a1 63.95Hz / mmHg a1 0.002308 1 / mmHg
254
Advantages of this type of PMRT with resonant membrane are the exception of effects,
which are connected with nonparallelism of membrane and piezoelement, considerable
decrease of temperature errors, connected with construct of quartz - holder, simplification of
construct of sensor (Pidchenko, Taranchuk, et al., 2003).
9. Application of piezoresonant mechanotrons of surplus air pressure in

medical sphygmographic systems to register pulse variation of cardiovascular human system
On the basis piezoresonat mechanotrons of surplus air pressure the authors developed
medical automated multichannel sphygmographic system BIOTON. This system allows
to do computer diagnostics of parameters of human hemodynamics, the registration of local
and volume sphygmogrammes for solving the tasks of polycardiography and poly
sphygmography. System BIOTON differs from the existing POLISPECTR-PWV (Russia),
Arteriograph "TensioClinic" (Hungary), Complior (France) by its improving informativeness
and quality of reproduction of pulse fluctuations.
Sphygmographic system (see Fig. 23) consists of sensors of pulse oscillation on a basis of
PMRT (Sensor 1-n) and the block of microcontroller (MCU). After primary information
processing it is transferred into computer on standard interface USB 1.1/2.0.
Each sensor contains measuring transducer of pressure on a base of PMRT, connected into
oscillatory system of oscillator. Change of interelectrode gap 1...n under the influence
of pressure in a receiver funnel pulse oscillations leads to frequency change on an Oscillator
1...n exit. Further the signal arrives on Frequency multipliers 1...n for increase in information
deviation of frequency. On each channel it is realized the method of linearization of
graduating characteristics of sensors, for which Synthesizer of direct synthesis (DDS
AD 9959) is used, which allows to give the optimal frequency of heterodyning in each
channel with the accuracy of less than Hz. From the exit of Mixer 1...n signal of differential
frequency comes onto Microcontroller ADUC841, which is used in the regime of period
measuring.
Fig. 23. Structure of sphygmographic system BIOTON

255
Amplitude,
V
For an exception of additional errors the uniform quartz generator of clock frequency for DDS
and MCU is used. Received data in real time are given to standard IBM compatible personal
computer (Notebook), which is the part of the system. Further processing is done with the help
of worked out special software, which functions under control of OS Windows
2000/XP/Vista/7. The developed software allows to carry out the semi-automatic analysis
and data processing, recording of results and creation of databases for research of
hemodynamics of cardiovascular system of patients. Possibility contour-time (with allocation
of characteristic points) and the spectral analysis sphygmographic curves (see Fig. 24),
definitions of an index of augmentation AIx and speeds of distribution pulse waves (PWV)
(see Fig. 25), and also allocation of the most typical sites sphygmograms, comparisons two or
several pulse oscillation is provided. For elimination of influence of artifacts procedures of a
digital filtration and smoothing are used (Taranchuk, Pidchenko, 2008 - 2010).
Time,sec
Amplitude, V
Fig. 24. Contour - time analysis of sphygmogram
t
Time, sec
Fig. 25. Defining PWV by signals on carotid artery and femoral artery
System BIOTON is highly effective while definition variability of heart rhythm (VHR) on
basis of measured sphygmograph curves of carotid artery during certain time (not less than
five minutes). Developed software allows to do statistic, correlation, spectral analysis and
form the estimation of analysis results of VHR while doing functional tests (see Fig. 26).
System BIOTON is designed for applying in following medicine areas:
1. Cardiology (non- invasive monitoring of hemodynamics, scientific research).
2. Family medicine (primary examination, early diagnosis of atherosclerosis).
3. Nephrology (beginning and progressing of atherosclerosis because of problems with
kidneys).
4. Diabetes study (faster process of aging of vessels because of increasing sugar in blood).
256
5.
6.
Obstetrics and gynecology (preeclampsia - endothelial illness, menopause - main factor

of risk of atherosclerosis);
Children cardiology (often atherosclerosis appears in early age).
Constructive realization of the block of the microcontroller and sensor of surplus pressure
on the basis of PRMT are in show Fig. 27.
Fig. 26. Results of measurements variability of heart rhythm
Fig. 27. Constructive realization of the block of the microcontroller and sensor of surplus
pressure on the basis of PRMT

257
10. Conclusion
The result of scientific research presented in present chapter is the creation of new element,
named piezoresonance mechanotron, which can be used as basic for constructing wide
range of highly informative measuring and diagnostic devices and systems. In combination
with linearizer of graduating characteristics, which realize approximation approach,
presented constructions of PMRT provide precision measurement of dynamic as well as
static pressure on the level of the best world models. They can be used for modernization of
existing as well as in the designing the perspective biomedical systems. Wide functional
possibilities of PMRT allow, by means of slight changes of its construction, to measure wide
range of parameters, such as travel, pressure, acceleration, mass, temperature and others,
which lead to micromoving of physical magnitudes. The construct of PMRT permits its
realization in microminiature variant on basis of micromembrane element, formed by means
of MEMS technology.
11. References
Cady W. (1946). Piezoelectricity, New York and London: McGraw-Hill.
Kolpakov F., Pidchenko S., Hilchenko G. (1997). Pressure Sensor, Patent Russia
2098783, December 10, 1997.
Kolpakov F., Pidchenko S., Hilchenko G. (1998). Method for measuring blood pressure,
Patent Russia 2106796, March 20, 1998.
Kolpakov F., Akulinichev A. (1999). Piezoquartz pressure transmitter with an elliptical
resonating membrane, Aerospace technics and technology, Vol. 12, National Aerospace
University (KhAI), ISSN 1727-7337, Kharkov, Ukraine, pp. 103-105.
Kolpakov F., Dobrova V., et al. (1999). Estimation of time diagnostic parameters of blood
pressure variation in frequency domain, Proceeding of the 1999 Finish Signal
Processing Symposium, Oulu, Finland, pp. 189-193.
Kolpakov F., Pidchenko S., Hilchenko G. (1999). Method of linearization calibration
characteristic piezoresonance transmitter with variable interelectrode gap, Patent
Russia 2127496, March 10, 1999.
Kolpakov F., Slavinsky S. (2004). The sphygmography measuring converter pressure-frequency dynamic characteristics examination, Measurement Science Review, Vol. 4,
Sect. 2, pp. 52-58.
Kolpakov F., Pidchenko S., Taranchuk A., et al. (2009). Piezoresonance mehanotron in
measurements of parameters warmly vascular system of the human,
Radioelectronic and computer system, Vol. 2, National Aerospace University (KhAI),
ISSN 1814-4225, Kharkov, Ukraine, pp. 60-70.
Kolpakov F., Pidchenko S. (2011). Theory and fundamentals implementation of invariant
piezoresonance systems, National Aerospace University (KhAI), ISBN 978-966-662222-1, Kharkov, Ukraine.
Malov V. (1989). Piezoresonance sensors, Energoatomizdat, ISBN 5-283-01507-6, Moscow,
Russia.
Pidchenko S., Kolpakov F., Akulinichev A. (2000). Analysis of the characteristics of multifrequency controlled crystal oscillator, Measuring and Computing Devices in
Technological Processes, Vol. 3, Khmelnitsky national university, ISSN 2219-9365,
Khmelnitsky, Ukraine, pp. 70-75.
258
Pidchenko S., Taranchuk A., et al. (2002). Pressure Sensor, Patent Ukraine 44108 ,
January 15, 2002.
Pidchenko S., Taranchuk A., et al. (2003). Pressure Sensor, Patent Ukraine 59936 ,
September 15, 2003.
Pidchenko S., Taranchuk A. (2004). Identification of the thermal state of the crystal at the
stage of oscillation, Radioelectronic and computer system, Vol. 3, National Aerospace
University (KhAI), ISSN 1814-4225, Kharkov, Ukraine, pp. 36-42.
Pidchenko S., Taranchuk A., Opolska A. (2010). Utilization Features of the Mechanotron for
Information Measurement Systems, Modern Problems of Radio Engineering,
Telecommunication and Computer Science: Proceedings of the Xth International
Conference TCSET2010, ISBN 978-966-553-875-2, February 23-27, Lviv-Slavske,
Ukraine, P. 358.
Taranchuk A., Pidchenko S., et al. (2002). The calculation of the temperature field
piezoelectric pressure sensor with a modulated interelectrode gap, Metrology and
Computer Engineering: Proceedings of the III International Conference (Metrology
2002), Sect. 2, Kharkov, Ukraine, pp. 139-141.
Taranchuk A., Pidchenko S. (2005). Modelling of thermal processes in the piezoresonance
sensors with modulated interelectrode gap, Khmelnitsky State University's bulletin,
Vol. 1, Khmelnitsky national university, ISBN 978-966-330-114-3, Khmelnitsky,
Ukraine, pp. 218-222.
Taranchuk A., Pidchenko S., et al. (2008). Sphygmographic channel to study hemodynamic
parameters of the cardiovascular system of man, Khmelnitsky State University's
bulletin, Vol. 3, Khmelnitsky national university, ISBN 978-966-330-114-3,
Khmelnitsky, Ukraine, pp. 47-52.
Taranchuk A., Pidchenko S., Opolska A. (2010). Sphygmographic system of high accuracy
registration of pulse fluctuations, Medical and Biological Informatics and Cybernetics:
Proceedings of the First Ukrainian Congress, June 23-26, Kiev, Ukraine, P. 267.
Taranchuk A., Pidchenko S., et al. (2010). The transmitter with pressure-frequency sensor
for sphygmographic research, Patent Ukraine 53900, October 25, 2010.
13
Non Invasive Acoustic Measurements for Faults
Detecting in Building Materials and Structures
Barbara De Nicolo, Carlo Piga, Vlad Popescu and Giovanna Concu
University of Cagliari, Engineering Faculty,

Italy
1. Introduction
1.1 Backgrounds and generalities
Over the past years both large and small restoration and conservation works on
monuments, civil and industrial buildings have become of great interest. As indicators of
the historical period in which they were built, all construction works have both their
architectural style and the material used in their construction. Indeed, for thousands of years
humans built using the same materials (wood, stone, brick, mortar and gypsum) up to the
introduction of concrete at the beginning of the 19th century. Although concrete has replaced
the old materials used in historical buildings, there still remains the problem of forestalling
their deterioration and restoring them, also in the light of the importance of such works
from the historical, cultural and economic viewpoints. Problems connected with the
restoration of buildings, whether in reinforced concrete, masonry or wood, are quite
complex and are essentially linked to the reuse, and thus the redesign, of the existing
heritage of buildings. Indeed, cultural, social and economic reasons foster to the desire to
lengthen the life of this heritage beyond normal physiological limits and thus its fruition far
beyond its useful life. The problems to be addressed vary widely since there are noteworthy
differences from one job to another; it is sufficient to consider just the social value of a
building of great historical value, which is usually protected by severe restrictions aimed at
conserving its artistic and cultural features, or an industrial building the use of which must
be completely changed while at the same time maintaining its structural characteristics. It is
evident that there is not one single answer to such widely differing situations: each job must
be addressed from the cultural, technological and technical standpoints as a special case.
The proper management of the rehabilitation of a building implies a knowledge of its real
static conditions to be restored, the mechanical, physical and chemical characteristics of the
materials of which it is built and the presence and characteristics of defects, anomalies and
so on. Of fundamental importance is the diagnosis of materials and structures and many
researchers, as well as companies that produce restoration materials, have carried out
studies in this field. Methods for structural diagnosis and faults detecting are beginning to
appear, albeit in an extremely divergent way, in tenders, rules, guidelines and so on. To
exemplify, non-invasive diagnostic techniques are often used to determine whether or not
materials compatible with the original structure have been used in restoration works: if not,
260
the building may undergo greater damage than that caused by progressive deterioration.
All the factors illustrated above favour the development of new control and diagnostic
techniques that produce more and more information on the state of the structure and,
indirectly, represent a more precise instrument for use in the planning of restoration works.
In the field of structural faults detection particular importance is given to developments of
Non-Destructive Testing measurements techniques (NDT), including automated procedures
and information technology to support decision making and evaluation of data. NDT
appear to be of great usefulness since, compared to classic laboratory techniques, they are
non-invasive, faster and of a general rather than specific nature. These skills have led to the
creation, evolution and rapid diffusion of certain diagnostic non-destructive measurements
techniques. The main obstacle to the effective, systematic and economical use of NDT in
structural diagnostics lies in the gap that exists between the theoretical and interpretative
bases and codification of operative modalities, which is to say the almost total lack of
standardization procedures; moreover, practical experience has underscored the limits of
commercial devices, which have often proved to be incapable of adapting to specific
structural problems under investigation and have interfaces that do not allow users to check
the validity of data acquired and the accuracy of measurements, thus making the results
aleatory and difficult to interpret.
As a major NDT tool, acoustic techniques, based on measurements of the characteristics of
acoustic waves propagating through the material, are often used in quality control and
faults detection for engineering structures and infrastructures. The studies of acoustic
techniques have been focused on medical or materials engineering applications for
laboratory testing. The valuable handmade analysis has begun in the first of '90 and
nowadays it is performed using considerable approximations to the detriment of results
precision. The inaccurate results are also caused by the subjective methodology often used
for interpretations. This contributed to form the common opinion that the use of such
methodologies in structural diagnostics does not give reliable results. Conversely, the
aforesaid techniques could show notable diagnostic properties if used in an appropriate
way, that is providing them with innovative techniques, which employ and develop
advanced computational tools and testing devices. In the light of this, fuelled by the rapid
development of portable personal computers, high-performance computing algorithms and
electronic engineering technology, acoustic techniques have evolved dramatically during
recent years.
Acoustic material analysis is based on a simple principle of physics: the propagation of any
wave will be affected by the medium through which it travels. Thus, changes in measurable
parameters associated with the passage of a wave through a material can be correlated with
changes in physical properties of the material. Recently, thanks to continuing scientific and
technological progress, systematic studies and applications of such methods in many
different fields have been developed. The application of methods of automated graphic
representation to the so-called media transparent to vibrations such as even large metal pieces,
has now become routine. Outside the field of engineering, such methods have also been
applied to parts of the human body; it is in the techniques of medical analysis such as
ultrasonography and computerized axial tomography that some of the most important
methods of investigation of structures with acoustics NDT have been developed. The
Non Invasive Acoustic Measurements

for Faults Detecting in Building Materials and Structures
261
acoustic NDT of materials less transparent to vibrations, such as concrete, mortars, stones,
wood, masonry and so on, is still more challenging. For such substantially inhomogeneous
media, the development of acoustic testing techniques has been decidedly slower compared
to applications on transparent media. This depends mainly on the greater difficulties and
the important theoretical and interpretative challenges, not to mention the technological
ones represented by the reduced transparency of the materials to vibrations deriving from
their high degree of intrinsic lack of homogeneity.
1.2 Methodologies
1.2.1 Sonic and ultrasonic testing methods
Sonic and ultrasonic investigations refer to a complex method for the analysis of materials
and the structures of which they are made, based on the study of phenomena connected
with the propagation of elastic perturbations inside the materials under study. The signal
that penetrates into the material is generated artificially by an external source and acquired
by means of a receiver after passing through the medium following appropriate trajectories.
From analysis of the processes and parameters connected with the propagation of acoustic
perturbations inside the artefact it is possible to collect a great deal of information on the
material or structure under study (J. Krautkramer & H. Krautkramer, 1990). This
information includes:
the level of homogeneity of the material concerning several elements or a certain

number of samples or only one of these;
the quality and degree of deterioration of the material;
the estimate of certain elastomechanical characteristics;
the identification of possible faults in the material or structure, such as cavities,
inclusions and zones having different elastomechanical characteristics;
the trend in time of different phenomena and related to the stresses the materials
undergo.
The intrinsic characteristics of the medium under test intervene decidedly in the several
aspects connected with the use of acoustic methods (choice of the most suitable
instrumentation, application of the basic principles of the method, criteria for interpreting
the results), especially in the choice of the signal frequency to use in the investigation.
In the case of not large homogeneous media, for example, metal pieces, the characteristics of
homogeneity favour the use of ultrasonic frequencies above 500KHz. The absence of natural
non homogeneities allows the signal to propagate without appreciable reflection, refraction
or mode conversion diffusion phenomena, unless these are caused by the presence of
possible local anomalies. The analysis of such kinds of media is thus favoured by the
negligible attenuation of the signal. This requires the use of very little energy and makes it
possible to identify even very small defects by emitting very narrow irradiation bands
through strongly directional probes. Substantially, in these cases it is possible to plan the
analysis in the smallest details and accurately investigate by closely defined zones from the
geometric standpoint.
For investigations in media with a high degree of non homogeneity, for example the
stratigraphic analysis of terrains, due to the strong intrinsic heterogeneity of the medium
262
there is a noteworthy diffusion of the signal. This, together with the need to investigate large
areas, calls for the use of sources with high emission energy and therefore kept prevalent in
low frequencies (usually 5-150Hz), so as to compensate for the high absorption
characteristics of the medium and the total loss of signal direction.
When intrinsically inhomogeneous media of not very large dimensions such as those
represented by building structures are to be investigated, the problem of revealing
perturbations caused by anomalies of great interest arises; however, such anomalies differ
slightly in extension, geometry and dimensions from the natural non homogeneities of the
medium under study: an emblematic case is that of concrete in which the anomalies may
respond to the acoustic input in a way similar to the aggregate. For this category of media,
in which we find most of the materials and structures used in civil, industrial and
monumental engineering and architecture, it is of fundamental importance to choose
correctly the characteristics of the signal employed, attempting a mediation between the
characteristics of the ultrasonic signals (enhanced diagnostic precision at the price of greater
attenuation and thus a lesser penetrative capacity) and that of the sonic signals (high
penetrative capacity at the price of poor definition). This is the case of concrete elements,
masonry structures, wood elements and limited volumes of terrain.
Experimental results (Concu et al., 2003a) showed that both sonic and ultrasonic signals reveal
the presence of macroscopic anomalies inside a limestone masonry structure, but are
differently sensitive to intrinsic structural characteristics and environmental conditions. We
performed the experiment on a sample masonry wall of limestone blocks, inside which some
metal and wood elements, assumed as anomalies of the structure, were placed in a known
position. Moreover, in a central position of the masonry a cavity was excavated designed to
hold a rectangular section, but the irregularity of the external blocks and the intrusion of
mortar caused the section to become irregular in the executive stage. The propagation velocity
of high- and low-frequency signals through the wall thickness was measured: higher velocities
are generally associated with a better quality of the material. Velocity data were then
elaborated by interpolation and represented in the form of a map of velocity distribution in a
generic vertical section of the wall. Fig. 1 shows the velocity maps obtained with the sonic
(low-frequency signals) and ultrasonic (high-frequency signals) methods.
3000
2800
2600
2400
2200
2000
1800
1600
1400
1200
1000
800
600
3400
3200
3000
2800
2600
2400
2200
2000
1800
1600
1400
1200
1000
800
Fig. 1. Maps of velocity [m/s]. Sonic velocity (left) and ultrasonic velocity (right). Extracted
from Concu et al., 2003a

263
The two velocity distributions appear quite similar. Substantially, the two methods show the
same velocity range, with maxima and minima located in the same areas. Both maps reveal
a well-defined region of minimum velocity in correspondence to the cavity. These results
underscore that both methods are useful for revealing the presence of macroscopic
anomalies inside stone masonry. The test also made it possible to demonstrate that
environmental noises and vibrations impact more on sonic than on ultrasonic method; the
latter is instead less practical since it calls for the use of a transducer rather than a hammer
as the source, and this requires the precise positioning of source and receiver on the surface
of the structure before each signal acquisition.
Independently of the frequency characteristics of the signal employed in the investigation,
the parameters associated with the signal penetrating through the medium are the
following:
transit time (or travel time), that is, the time taken by the signal to cover the distance
from the source to the receiver inside the material under examination;
signal propagation velocity, in the sense of the ratio of the distance between source and
receiver to transit time;
signal attenuation characteristics in its passage through the material.
Traditional application of acoustic techniques is based on measurements of the velocity V of

acoustic waves propagating through the material. The velocity is obtained from the ratio
L/T, where T is the time wave needs to travel along the path of length L. The wave velocity
is directly related to structure's elastic parameters, e.g. elastic modulus E, Poisson's number
and density , thus its analysis provides information crucial for inspections of structures
inner conditions.
The propagation velocity, although a significant parameter, limits the investigation at the
analysis of propagation times, not taking into consideration important information
regarding the way the waves are propagating. For instance, when a wave passing through a
specific item encounters any discontinuity, the wave power is certainly attenuated because
of scattering phenomena, while the propagation time may not be moved if part of the signal
is already able to reach the receiver. Therefore, it would be reasonable to approach the
acoustic analysis also in terms of other waves features changes and not only in terms of
propagation times. The higher the intrinsic non-homogeneity level of structures, e.g.
masonries, the bigger the advisability of this integrated approach. In fact, as documented by
various studies predominantly performed in the geophysics and aeronautics environments,
other waves characteristics such as attenuation, scattering and frequency content, primarily
related to the elastic wave power, may allow one to get more and relevant information
about the material, because of the reliance of the propagation on the properties of the
medium through which waves travel. In fact, different materials absorb or attenuate the
wave power at different rates, depending on complex interactive effects of material
characteristics, such as density, viscosity, homogeneity. Additionally, waves are reflected by
boundaries between dissimilar materials, so that changes in materials structure, e.g.
presence of discontinuities or defects, can affect amplitude, direction, and frequency content
of scattered signals. Furthermore, all materials behave somehow as low pass filters,
attenuating or scattering the higher frequency components of a broadband wave more than
the lower. Thus, waves analysis in terms of multiple waves features can give information
on the combined effects of attenuation and scattering as previously described.
264
The comparative analysis of transit time and amplitude attenuation of acoustic signals
carried out on the stone wall previously described confirmed the opportuneness of the joint
use of both parameters (Concu et al., 2003b). In this study, we assumed the amplitude
attenuation as the ratio of received signal to transmitted signal maximum amplitude. Fig. 2
shows the map of transit time and amplitude attenuation in the generic vertical section of
the masonry.
520
480
440
400
360
320
280
240
200
160
120
8E-006
7E-006
6E-006
5E-006
4E-006
3E-006
2E-006
1E-006
0
Fig. 2. Map of transit time [s] (left) and amplitude attenuation (right). Extracted from
Concu et al.,2003b
By observing the map we can see that the distribution of transit time shows a clearly defined
section with maximum values in association with the cavity inside the wall. This section can be
illustrated as a geometric figure of size and position perfectly compatible with the design
geometry of the cavity before erection of the wall. Distribution of times in the rest of the map
appears on the whole uniform. From this we can deduce that ultrasonic signal transit time
(thus velocity) is a parameter capable of identifying macroscopic defects with a good degree of
approximation and immediacy, while it does not appear to be sensitive to minor anomalies
such as the absence of material between the stone blocks, the presence of mortar joints or small
elements of different material. Interpretation of distribution concerning amplitude attenuation
is less immediate: we can see a zone of maximum attenuation in correspondence to the cavity,
the borders of which are rather blurred. Attenuation values are on the whole rather dispersed.
This confirms that amplitude attenuation is a parameter extremely sensitive to all kinds of
discontinuities of the material that cause a loss of signal energy and it is therefore quite
suitable for use in tests at high definition in small areas. Transit time and amplitude
attenuation thus have different diagnostic capabilities and this emphasizes the usefulness of
different waves features integrated use in structural faults diagnoses.
The study of energy characteristics associated with the signal is today addressed also by
employing the spectral analysis. By spectral analysis is meant the method based on analysis
of the frequency content of the signal travelling through the material of interest. When a
signal goes through a medium, the frequency components associated with the input signal
are altered since the medium acts as a filter that transmits only a certain frequency band
with different degrees of attenuation and phase shifts. It is thus possible to study the effects
of the medium's properties on alteration of the input signal by performing a spectral

265
analysis on the signals passing through different materials or different portions of the same
material or the same material in different surrounding conditions: a given input signal will
emerge with different spectrum frequencies after passing through materials with different
characteristics (Priestley, 1981). Substantially, the spectral analysis provides a sort of
signature characteristic of the properties of the material travelled through. The material is
gone through by a signal of the impulsive kind expressed by means of a function of type
A(t), in which the oscillation amplitudes are given as a function of time. The spectrum of the
signal offers instead a representation of the same in the frequency domain, in which the
signal is expressed by means of a function of type a(f): the amplitudes of the elementary
oscillations that make up the impulse are given as a function of the respective frequencies.
1.2.2 Acoustic emission
The acoustic emission technique is based fundamentally on the study of the acoustic signal
emitted by the material during deformation, cracking, breaking, collapse and in general
during whatever phase causing a release of energy. The single event, that is, the single
emission, is thus an impulsive acoustic signal produced by a source within the material
following the triggering of any phenomenon capable of releasing energy. The main
parameters associated with acoustic emissions and used for the study and application of the
method in the field of structural monitoring and faults detection are:
number of events emitted during the phenomenon - deformation, loading and so on

under observation;
emission velocity, in the sense of the number of events in a given time interval;
maximum amplitude of the event;
duration of the event;
wave shape of the event;
frequency spectrum of the event;
arrival time of the signal at the transducers used for establishing it;
the so-called "bvalue", that is, the slope of the logarithmic curve representing the
maximum amplitude of the events as a function of the frequency with which they
appear.
These parameters are placed together with the factors that identify the specific issue being
tested, such as the stress-deformation load diagrams, microscopic observations and so on, so
as to find a relation that makes it possible to arrive at the state of the material starting from
observation of the magnitudes associated with its acoustic emission.
The potentialities of the method for the study and monitoring of the behavior of materials,
the prediction of their response to different kinds of stresses and the check of defects and
anomalies are many. They are based on special phenomena, such as the Kaiser effect, on the
application of instruments of mathematical analysis and numerical calculus, and on the
comparative study of acoustic parameters and elastomechanical characteristics (Enoki &
Kishi, 1991).
The Kaiser effect (Kaiser, 1950) is a phenomenon by which a material under stress emits
acoustic signals which are significant only when the level of stress to which the material was
previously submitted has been exceeded. In effect, there are emissions even below this level,
but the two kinds of events differ greatly: in rapid succession and high energy content the
266
former, associated with the triggering and propagation of new cracks; less frequent and with
much lesser amplitudes the latter, associated with the deformation and contact between the
surfaces of cracks opened in the previous load cycle. Substantially, the material has a
recollection of its history of load and deformation. The uses of this phenomenon are many. It
is used to arrive at the maximum stress that a material has undergone, for example the
original stress of a rock, estimated by submitting to loading and unloading test samples
taken from the rock mass. It is exploited in identifying the breakage surface associated with
the material in different load conditions. It can be an indicator of the state of deterioration of
the material, for example, of the damage caused to it by loads, since emissions of the second
type appear for lower load levels in the most damaged materials.
The possibility of characterizing the source of acoustic emissions arises on using a suitable
number of transducers (in any case more than three): it is in fact possible to return to the
position of the signal source by recording its arrival times at the different transducers and
knowing the value of the characteristic acoustic velocity of the material. By interpolation we
find the point inside the material which, for the assigned velocity, satisfies the values of all
the times recorded by the transducers. By means of a well-documented mathematical
treatment based on Green's functions, on the analysis of tensorial moments and on the
calculation of eigenvalue and eigenvector it is possible to identify not only the position but
also the volume and spatial orientation of the source inside the medium, thus obtaining its
complete geometric characterization (Ohtsu & Ono, 1986).
The comparative study of the acoustic emission parameters and factors of the
elastomechanical type makes it possible to obtain many supplementary data on the behavior
of the material. Acquisition of the acoustic emission parameters associated with knowledge
of the diagrams of load, stress and deformation, as well as the characteristics of the material
observed under the microscope before, during and after the phenomena under examination
leads to the definition of precise correspondences between acoustic parameter values and
certain fundamental data such as:
level of load, stress and deformation;

type of defect originating the emission (intra-, inter- and transgranular cracks);
way of propagation of breaks (tensile or shear) and thus the kind of breakage observed;
level of creep and the way in which it develops;
dimension of the damaged zone in the material and so on.
From this naturally derives also the possibility of foreseeing the responses of the material to
different kinds of stress, starting from the analysis of its acoustic emission. Substantially, the
study of acoustic emissions has great potentialities in the monitoring of complex civil,
industrial or natural structures, of single structural elements or laboratory test pieces made
up of many different materials such as metals, cement, rock, ceramics, synthetic materials
and so on. Interest concerning the use of the acoustic emission technique is proven also by
the presence on the market of many kinds of often quite complex instrumental sets capable
of acquiring and processing acoustic emission data in real time and proposing diagnostic
hypotheses on the materials or structure under investigation by means of comparisons with
databases on acoustic parameters. The application of this methodology is of great service in
real-time monitoring of the triggering and propagation of breaks in materials in use as well
as in the characterization of their behavior and the prediction of their responses to various
stresses.

267
2. Operative procedures
2.1 Direct and indirect transmission
The easier and faster way to get relevant information using sonic and ultrasonic methods is
the measurement of the waves propagation velocity V. In fact, the operative procedure to
acquire waves velocity and to process data in order to get immediate results is quite simple.
In addition, the skills of this parameter are very useful: in fact, from waves propagation
theory it is known that V is dependent on the following materials characteristics: dynamic
elastic modulus Ed, Poisson's number and density . For a homogeneous isotropic material
this function is:
V
2v
(1)
V is directly related to structures elastic parameters, so that it has been frequently applied
for evaluating structures integrity and restorations effectiveness. Thus the relation between
V and Ed, , can be exploited for achieving data regarding the structures health in terms of
elastomechanical conditions. In fact, the measurements of V along proper grids of paths
leads to the elaboration of maps of velocity, which allow one to define the level of
elastomechanical homogeneity of the investigated structure, emphasizing areas where
anomalies are located; moreover, the knowledge of V values distribution consents to express
a qualitative remark on the mechanical effectiveness of the structure, since it is empirically
known that the higher the strength the higher the velocity.
Waves velocity measurements are preferentially carried out applying the Direct
Transmission Technique (DTT), in which the wave is transmitted by a transducer (Emitter)
through the test object and received by a second transducer (Receiver) on the opposite side.
This allows measuring the time T that the wave needs to travel through the objects
thickness, from the emitter to the receiver, along a path of length L; the average velocity of
the wave is simply obtained from the ratio L/T. The DTT is very effective, since the broad
direction of wave propagation is perpendicular to the source surface and the signal travels
through the entire thickness of the item. Standards concerning the determination of waves
velocity in structures, e.g. Europeans EN 12504-4 (EN 12504-4, 2004) and EN 14579 (EN
14579, 2004), suggest, therefore, the application of this kind of signals transmission.
Nevertheless, there are many kinds of structures, e.g. slabs, retaining walls, piers, in which
the DTT cannot be performed, because only one side of the item is accessible. In these cases
the Indirect Transmission Technique (ITT), in which both the emitter and the receiver
transducers are placed on the same side of the investigated object, or the Semi-direct
Transmission Technique (STT), in which transducers are placed on adjacent faces, might be
used. Generally speaking, ITT and STT are less effective than the DTT because the
amplitude of the received signal is lower, and the pulse propagates in a concrete layer just
beneath the surface. These remarks have since now not allowed ITT and STT systematic
development, and the scientific literature concerning their use is still quite poor. Despite
that, ITT skills of ease to be performed, high potential to evaluate the quality and the
characteristics of concrete covering on site, immediacy and low cost, claim to thorough
examine its suitability in concrete diagnosis on site, and then to develop studies concerning
the standardization of its application. In Annex A of the EN 12504-4 the determination of the
pulse velocity via the ITT is illustrated. It is highlighted that there is some uncertainty
268
regarding the exact length of the transmission path, since the areas of contact between the
transducers and the item are of significant size. It is therefore suggested to make a series of
measurements with the transducers at different distances apart to eliminate this uncertainty.
The transmitting transducer shall be placed in contact with the item surface at a fixed point
x, and the receiving transducer shall be placed at fixed increments xn along a chosen line on
the surface. The signals transit times recorded should be plotted as points on a graph,
showing their relation to the distance separating the transducers. An example of such a plot
is shown in Fig. 3, extracted from Annex A. The slope of the best straight line drawn
through the points (tan ) shall be measured and recorded as the mean pulse velocity along
the chosen line on the concrete surface. Where the points measured and recorded in this
way indicate a discontinuity, it is likely that a surface crack or surface layer of inferior
quality is present and a velocity measured in such an instance is unreliable.
Fig. 3. Example of the determination of pulse velocity by ITT. Extracted from EN 12504-4,
2004 Annex A
One of the main skills of the ITT is the possibility of cracks depth estimation. An estimate of
the depth of a crack visible at the surface can be obtained by measuring the transit times across
the crack for two different arrangements of the transducers placed on the surface. One suitable
arrangement requires that the transmitting and receiving transducers are placed on opposite
sides of the crack and equidistant from it (BS 1881, 1986). Two values of this distance x are
chosen, one being twice that of the other, and the transit times corresponding to these are
measured. If the first value of x chosen is x1 and the second value x2 and the transit times
corresponding to these are T1 and T2 respectively, then the crack depth h is:
h
(2)
Equation (2) is derived by assuming that the plane of the crack is perpendicular to the item
surface and that the material in the surrounding area of the crack is of reasonably uniform
quality. A check may be made to assess whether the crack is lying in a plane perpendicular
to the surface by placing both transducers near to the crack and moving one of them far
from the crack. If the transit time decreases, this indicates that the crack slopes towards the
direction in which the transducer was moved. It is important that the distance x is measured
accurately and that a very good coupling is guaranteed between the transducers and the
concrete surface. The method is valid provided the crack is not filled with water.

269
Another method of test uses the indirect method where a discontinuity appears in the graph
drawn following the indication of the standards (EN 12504-4, 2004; EN 14579, 2004) as
previously explained. In this case, if L is the distance separating the transducers
corresponding to which the slope of the line distance-time changes, while T1 and T2 are the
transit times corresponding to this change, then the crack depth h is:
h
(3)
As previously stated, the direction in which the maximum energy is propagated is at right
angles to the face of the transmitting transducer, so that the DTT is the most effective
operative procedure. However, the DTT has some limits too. The major limit consists in
describing the waves characteristics field of the object using for each path only one value of
that characteristic, i.e., hypothesizing that the average value is homogeneous along each
wave path. This assumption prevents from pinpointing the position of the detected anomaly
inside the object. A promising way for overcoming this limit is the use of the tomographic
technique, which uses numerical analysis as a real measurement instrument, combining the
results of several DTT applications for a sharper and reliable investigation of the object.
2.2 Tomography
One emerging technique for advanced imaging of materials is Acoustic Tomography (AT).
AT uses technology invented for the biomedical field to display the interior of engineered
structures. The spatial distribution of acoustic velocity and attenuation are imaged and then
correlated with properties directly related to physical conditions (Belanger & Cawley, 2009;
Rhazi, 2006; Leonard Bond et al., 2000; Kepler et al.,2000; Meglis et al.,2005) . The velocity is
determined by the elastic properties and density, while the attenuation is determined by the
inelastic property of the medium.
2.2.1 Generalities
Travel time tomography, a type of AT, represents the natural evolution of the DTT: the
signals emitted by different sources are acquired by several receivers arranged so as to allow
the taking of a large number of measurements of transit time of signals travelling along
pathways at different inclinations which intersect each other on flat sections of the structure.
This makes it possible to apply an algebraic system whose unknowns are signal velocity at
the nodes of a network arranged on the flat section of the medium containing source and
receiver. Thus travel time tomography allows determination of velocity distribution on flat
sections of the item being investigated. The methods degree of resolution depends on the
distance between sources and receivers, on the measurement step, on angular coverage by
means of the trajectories of the studied section. Travel time tomography makes it possible to
overcome the major limit of the DTT, that is, the impossibility of discriminating an extended
alteration along the source-receiver pathway from an anomaly confined to a part of it only,
since for each trajectory joining source and receiver the behavior of the element inside the
medium is described by means of a single velocity value. In virtue of the thick net of
intersecting trajectories, travel time tomography returns velocity distribution on the section
under study with high definition, making it possible to localize precisely and in detail any
anomalies that may be present.
270
It has long been believed that attenuation is more suitable than velocity (or travel time) to
study the inner properties of materials, because an anomaly has a greater effect on the
attenuation of a signal than on the propagation time (Best et al., 1994; Hudson, 1981). In fact,
as previously mentioned, waves characteristics such as attenuation, scattering and
frequency content may allow one to get relevant information about the material, because of
the reliance of the propagation on the properties of the medium through which waves
travel. Different materials absorb or attenuate the wave power at different rates, depending
on complex interactive effects of material characteristics, such as density, viscosity,
homogeneity. Additionally, waves are reflected by boundaries between dissimilar materials,
so that changes in materials structure, e.g. presence of discontinuities or defects, can affect
amplitude, direction, and frequency content of scattered signals. In this context special
credit has to be given to spectral attenuation tomography, which returns the attenuation
coefficient distribution on flat sections of the item being investigated.
The main limitations on the AT widespread diffusion are the longer times of execution
compared to the traditional operative procedures, the higher cost of the instrumental sets
(arrays of sources and receivers, multichannel acquirers), the complexity of the
reconstruction of velocity or attenuation distribution starting from signals acquisition.
As previously stated, acoustic investigation methods exploit the transmission and reflection
characteristics of mechanical waves with appropriate frequencies passing through the
investigated item. Elastic waves propagate in different manner through different solid
materials and cavities, thus enabling fault detection. The waves are in most cases generated
by a piezoelectric transducer fed with a voltage pulse. The receivers are accelerometers,
appropriate positioned based on the measurement type. The tomography represents an
improvement to the classic techniques of wave direct transmission, being able to perform
tests also on non-perpendicular wave paths. It is so possible to reconstruct a 2D image of the
distribution of the wave propagation parameters (e.g. velocity, attenuation) within the
analyzed structure, or in one of its sections. These images allow the identification of
variations correlated with defects, malformations, cracks etc. Acoustic tomography implies
that a ill posed linear equations system has to be solved, in order to determine the
distribution of the chosen wave parameter (e.g. velocity, attenuation) in selected sections of
the tested structure, thus highlighting the presence of anomalies (Berryman, 1991). Different
inversion algorithms are available for determining this distribution starting from signal
transmission and acquisition.
2.2.2 Travel time tomography
An acoustic wave propagating through any object spends a definite time to travel from a
point to another of the object. The wave covers the path between the two points spending
the time t and propagating with a mean velocity V. When the distance l between the two
points reduces to zero a local velocity Vp, and a local slowness s = 1/Vp can be defined for
the point p. In mathematical terms this behavior can be expressed by the following equation:

dl
s dl
(4)
The acoustic behavior of a selected section of the object is then defined when the slowness
s(x) is known continuously in every point x of the investigated section. This function can be

271
approximated dividing the section into a grid of n rectangular cells (pixels) in which V is
supposed to be constant. The tomographic problem consists in obtaining the slowness of the
n pixels starting from the knowledge of m travel times t measured along a series of paths
joining couples of transducers located on opposite or adjacent sides of the section. The
waves paths depend on the velocity distribution, and their sharp definition is a not easy
problem to solve, especially when dealing with structures made of different materials, such
as stone masonry, or with a degree of intrinsic non-homogeneity, such as concrete; a valid
approximation may be the linear tomography, which considers the paths to be straight. In
order to obtain the values of the slowness in the grid, the following equations system, which
is the matrix form for equation (4), has to be solved:
PS=T
(5)
where :
T = [t1,t2,tm] is the vector of measured travel times;
S = [s1,s2,sn] is the vector of slowness;
P = [l11,l12,lnm] is the coefficients matrix, whose generic element lij is the length of the ith
path in the jth cell.
Thus, the tomographic solution consists in determining the vector S as:
S = Pg-1 T
(6)
To avoid instability in matrix inversion, the number n of cells must be smaller than the
number m of measured travel times. If the inverse of P exists it can be directly evaluated.
However, the inverse of P generally does not exist since P is not a square matrix, it is ill
conditioned, and it has not full rank. Thus, other methods, such as iterative ones, have to be
used to solve the problem.
2.2.3 Spectral attenuation tomography
As previously stated, the propagation velocity, although a significant parameter, limits the
investigation at the analysis of propagation times, not taking into consideration important
information regarding the way the waves are propagating. When a signal passes through a
specific item or structure, certain frequency components of this signal are altered, the item
behaving like a filter performing modifications on the frequency components magnitude and
phase. Therefore, it would be reasonable to approach the tomographic problem also in terms
of frequency spectrum changes and not only in terms of propagation times. This type of
approach is documented by various studies performed in the geophysics environment, and
furthermore in the approach of Quan and Harris (Quan & Harris, 1997), based on the
observation that the frequency attenuation increments with the frequency of the signal,
meaning that the higher frequencies of a signal are more rapidly attenuated than the lower
ones. The main advantage of the frequency analysis is the immunity against disturbing factors
such as spherical divergence, reflection and transmission effects and coupling of the receiver
with the transmitter, which can affect the correct interpretation of the received signals.
For the purpose of estimating attenuation, the process of waves propagation can be assumed
as described by linear system theory. If the amplitude spectrum of an incident wave is S(f)
and the instrument-medium response is G(f)H(f), then the received amplitude spectrum
R(f) may be, in general, expressed as :
272
R(f) = G(f)H(f)S(f)
(7)
where the factor G(f) includes geometrical spreading, instrument response, source-receiver
coupling, radiation patterns, and reflection-transmission coefficients, and the phase
accumulation caused by propagation, and H(f) describes the attenuation effect on the
amplitude. It can be assumed that the effects included in factor G(f) are not frequency
dependent, thus it can be simplified as G(f) = G. In structural diagnosis, the H(f) factor is of
greater interest. Experiments indicate that attenuation is usually proportional to frequency
(Johnston, 1981), that is, response H(f) may be expressed as:
H f
exp
dl
(8)
where the integral is taken along the supposed straight wavepath, and 0 can be regarded as
an intrinsic attenuation coefficient. The tomographys goal is to estimate the medium
response H(f), or more specifically, the attenuation coefficient 0, from knowledge of the
input spectrum S(f) and the output spectrum R(f). A direct approach is to solve equation (8)
by taking the logarithm and obtaining:

dl
ln G
(9)
Equation (9) may be used to estimate the integrated attenuation at each frequency and is
called the amplitude decay method. However, as described above, the factor G lumps many
complicated processes together, and is very difficult to be determined. Furthermore, the
calculation of attenuation based on individual frequencies is not robust because of poor
individual signal-to-noise.
To overcome these difficulties, Quan and Harris (Quan & Harris, 1997) developed a
statistically based method that estimates the attenuation coefficient 0 from the spectral
centroid downshift over a range of frequencies. An analog relationship to that between
signal velocity along the wavepath and travel time connects the attenuation to the difference
of the signals spectrum frequency centroid, the latter being a parameter indicating the
center of the signals distribution in frequency. As mentioned, during wave propagation the
higher frequencies are attenuated more rapidly than the lower frequency components,
downshifting the centroid towards the lower frequencies. The centroid frequency of the
input signal S(f) is defined as:
f
(10)
and the variance is:
(11)
Similarly, the centroid frequency of the received signal R(f) is:

f
and its variance is:
(12)

273
(13)
where R(f) is given by equation (7). If the factor G is independent on the frequency f, then fR
and R2 will be independent on G. This is a major advantage of using the spectral centroid
and variance rather than the actual amplitudes. For the special case where the incident
spectrum S(f) is Gaussian, assuming a linear-dependence model of attenuation to frequency,
the attenuation coefficient 0 can be estimated as follows:

dl
(14)
where fS and fR are the centroid frequency for the source and receiver, respectively, and S is
the variance, or bandwidth, of the source signal. The previous relationship states that the
attenuation is proportional to the centroid frequency difference which has downshifted
from the original source centroid fS, to the centroid of the received signal fR. The total
amount of centroid frequency downshift depends on the attenuation characteristics along
the acoustic path. The tomographic formula relating frequency shift with the attenuation
projection is exact only for Gaussian spectra. Yet, similar derivations can also be obtained
for other frequency compositions, which implies that the estimates of relative attenuation
are not sensitive to small changes in spectrum shapes, and points out the robustness of this
method.
It is worth noting that equation (14) is in the same form as (4), with the intrinsic attenuation
coefficient 0 in (14) corresponding to the slowness 1/V in (4). The expression of frequency
centroid difference in (14) corresponds to the travel time t in (4). This similarity makes the
attenuation tomographic inversion easy to conduct applying the same algorithms developed
for travel time tomography, simply replacing 1/V with 0 and t with (fS - fR)/ S2.
As stated above, equation (14) is the basic formula for spectral attenuation tomography. It
can be written also in a discrete form as:
(i= 1,..M)
(15)
where i represents the ith path, j the jth parameterized cell of the medium, and lij is the length
of the ith path in the jth cell (Fig. 4).
The previous equations system can be written in matrix form as:
LA=F
(16)
where :
F = [F1,F2,Fm] is the vector of calculated centroid frequency downshift, in which Fi = (fSi fRi)/ Si2 has been assumed;
A = [ 01,02,0n] is the vector of attenuation coefficients;
L = [l11,l12,lnm] is the coefficients matrix, whose generic element lij is the length of the ith
straight path in the jth cell.
The intrinsic attenuation coefficient 0 is in the unit of [dBsm-1].
Moreover, the attenuation coefficient 0 can be also expressed as:
274
Fig. 4. Example of spectral attenuation tomography system equation
(17)
with Q the quality factor of the material and V the propagation velocity. In wave
propagation problems the Q factor is useful for characterizing wave attenuation, being
defined as the ratio of the total kinetic energy and energy loss in one vibration cycle (Sheriff
& Geldart, 1995; Knopoff, 1964). Geophysicists and seismologists often use the Q factor to
study this attenuation in rocks. An infinite Q means that there is no attenuation. This factor
is a function of the mineral composition of rocks as well as of their mechanical performances
(Ilyas, 2010). Numerous field observations have demonstrated that the quality factor Q
appears to be a constant over a large frequency range in the signal bandwidth. This is
widely accepted in the geophysics community and is referred to as the constant Q model.
Hence, the tomographic inversion can be applied for the case of the spectral attenuation in a
similar way as for the arrival times and the propagation velocity. The model obtained for Q
is consistent with the one obtained for the velocity; therefore the information regarding the
velocity distribution can be used for calculating Q itself.
2.2.4 Resolution algorithms
A common method to obtain the solution of equations system (5) in the least square sense is
the Singular Value Decomposition (SVD) (Berryman, 1991; Herman, 1980; Ivansson, 1986).
The SVD can be used for computing the pseudo-inverse of the coefficient matrix P. Indeed,
this method produces a diagonal matrix D, of the same dimension of P and with nonnegative diagonal elements in decreasing order, and two unitary matrices U and V so that P
= UDVT. Then, the pseudo-inverse of the matrix P with singular value decomposition is P+ =
VD-1UT and the solution of equations system (5) can be written as S = P+T. The same result
can be assumed for the spectral attenuation tomography, simply considering system (16)
instead of system (5). This solution is the minimum norm solution and it is only

275
mathematically suitable. The inverse problem is ill posed and ill conditioned, making the
solution sensitive to measurement errors and noise. Regularization methods are needed to
treat this ill posedness. It can be shown that the small singular values mainly represents the
noise and can be discarded. Truncated SVD (SVDT) may be considered as having a filter,
and hence it is less sensitive to high frequency noise in the measurements.
Because of the non linear relationship velocity - travel time and attenuation coefficient centroid frequency downshift, it is almost impossible to find the solution for systems (5)
and (16) by a single step algorithm using a linear approximation. Thus, iterative methods
(Gilbert, 1972; Lo & Inderwiesen, 1994) can be used, such as Algebraic Reconstruction
Technique (ART) (Gordon, 1974; Gordon et al., 1970) and Simultaneous Iteration
Reconstruction Technique (SIRT) (Dines & Lytle, 1979; Lakshiminarayanan & Lent, 1979;
Jansen et al., 1991). Both methods need a starting value of velocity or attenuation, and then
they modify iteratively this value by minimizing the difference between the measured travel
time or centroid frequency downshift and the value calculated in the previous iteration.
While ART goes on wavepath after wavepath, SIRT takes into account the effect of all
wavepaths crossing each cell. In the n-dimensional space each equation in (5) and (16)
represents a hyperplane. When a unique solution exists, the intersection of all the
hyperplanes is a single point. A computational procedure to locate the solution consists in
starting with an initial solution, denoted by:
q
,q
,,q
(18)
where q signifies velocity or attenuation coefficient depending on whether travel time or

spectral attenuation tomography is being performed. This initial solution is projected on the
hyperplane represented by the first equation in (5) or (16) giving q(1). This value is then
projected on the hyperplane represented by the second equation in (5) or (16) to yield q(i-1)
and so on. When q(i-1) is projected on the hyperplane represented by the ith equation to yield
q(i), the process can be mathematically described by:
q
(19)
where li is the ith raw of the matrix L and bi represents ti or Fi depending on whether travel
time or spectral attenuation tomography is being performed. A single iteration of ART is
completed when each row of P has been cycled.
For an over determined problem, m > n, ART does not give a unique solution, but this
depends on the starting point. The tomographic system is normally over determined and
measurement noise is present. In this case a unique solution does not exist and the solution
found by ART will oscillate in the neighborhood of the intersections of the hyperplanes. The
SIRT algorithm uses the same equations as in the ART algorithm; the difference is that SIRT
modifies the attenuation model taking into account at each iteration the effect of all wave
paths crossing each cell. The new value of each cell is the average value of all the computed
values for each hyperplane:
q
(20)
276
Then, using the SIRT algorithm better solutions are usually obtained at the expense of
slower convergence.
2.3 Impact echo method
Another development of the classic operative procedures is the impact-echo model, which
represents in some way an evolution in the ITT. This method is based on the principle of
reflection: a signal propagating in a medium is reflected on encountering an anomaly of any
kind inside it (Sansalone & Streett, 1997; Carino, 2001). On striking the medium with a
succession of impulses, usually generated by mechanical impact with a hammer, these are
reflected from the interface between the medium and the air if the piece is homogeneous or
by the defect if the medium presents an interruption in continuity; in this case the reflection
caused by the limiting surface is addressed as base echo, while the reflection caused by any
imperfections there may be in the medium is addressed simply as echo. By arranging source
and receiver on the surface of the medium it is possible to visualize with suitable
instrumentation the echoes generated by the reflection of the signal in the medium; this
makes it possible to establish the kind of reflection it is, that is, if it is caused by limit
surfaces or something else, and to locate the obstacle inside the medium as a function of
amplitude and the reciprocal position of the signals corresponding to the echoes. The
extraction of information usually takes place in the frequency domain.
2.4 Measurements sets
The basic instrumental set for the performance of sonic and ultrasonic tests is composed of a
source, a receiver and a data acquisition and processing unit, often an oscilloscope and PC.
The fundamental difference lies in the characteristics of the source: the sonic method calls
for the use of signals containing high energy and thus characterized by low frequency for
which the excitation of the material is usually performed with the impact of an
instrumented hammer; the ultrasonic method requires the use of high-frequency signals
since no excessive dissipation of energy inside the material is foreseen, so the signal is
introduced by means of a transducer, usually piezoelectric. Fig. 5 shows the schematic of
instrumental sets commonly used in the two kinds of tests.
EMITTER TRANSDUCER
IMPACT HAMMER
PULSER
RECEIVER
CONTROL
UNIT
WALL
RECEIVER TRANSDUCER
WALL
RECEIVER TRANSDUCER
DIGITAL
OSCILLOSCOPE
Fig. 5. Instrumentation set up. Sonic Testing (left) and Ultrasonic Testing (right)
When applying the Acoustic Emission procedure, the signal associated to the acoustic
emission can be captured by a suitable instrumental setup, usually composed of highfrequency transducers, amplifiers, filters, systems of acquisition and storage such as digital
oscilloscopes or acquisition cards, data processing software and so on, as shown in Fig. 6.

277
Fig. 6. Schematic diagram of a basic four-channel Acoustic Emission testing system.

Extracted from NDT Education Resource Center
Impact-echo testing relies on three basic components, as shown in Fig. 7:
Fig. 7. Schematic diagram of the Impact-Echo test. Extracted from Sansalone & Streett, 1997
278
a mechanical impactor capable of producing short-duration impacts, the duration of

which can be varied;
a high-fidelity receiver to measure the surface response;
a data acquisition-signal analysis system to capture, process, and store the waveforms
of surface motion.
3. Case study
In order to deepen the reliability of acoustic methods in buildings faults detecting and
materials characterization, an experimental program has been started, experimenting both
DTT and AT approaches to a full scale masonry model.
3.1 Materials
The two operative procedures DTT and AT - have been carried out on a full scale real
stone masonry expressly made by the Lab of Structural Engineering Dept. The wall is 0.90 m
wide, 0.62 m high and 0.38 m thick, and it is made of Trachite blocks sized 0.20 m 0.38
m 0.12 m, settled as shown in Fig. 8 and jointed with cement lime mortar. The block
assigned to the central position of the wall was not settled, thus realising a macro-cavity
with the same size of the missing block, and assumed as a known anomaly. Mortar joints
have been assumed to be 1 cm thick, but since the wall was manually built by a builder,
actual dimensions are not so precise.
Fig. 8. The full scale real masonry. From left to right: front view, vertical section, horizontal
tomographic section
Trachite specimens have been prepared and then tested for the determination of
compressive strength and elastic modulus, following the Italian Standards UNI EN 1926,
2000 (UNI EN 1926, 2000) and UNI EN 14580, 2005 (UNI EN 14580, 2005) respectively.
Results are shown in Table 1.
Properties
Compressive strength
Static elastic modulus
Table 1. Materials properties
Value (MPa)
40.5
6100

279
3.2 Methods
3.2.1 Direct transmission technique
As previously stated, when using the DTT the acoustic signal is transmitted through the test
object and received by a second transducer on the opposite side of the structure. Changes in
received signal provide indications of variations in material continuity. In this case study
220 emitters and 220 receivers have been arranged in a grid of 1120 nodes in the opposite
surfaces of the wall (Fig. 9).
Fig. 9. Grid of test points

At the end of the experimental sessions, 220 signals have been obtained, one for each point
of the grid of receivers. From each node of the grid, two parameters velocity and signal
power - have been extracted, in order to investigate the presence of anomalies and obtain
some information on the material. The signal power is defined as:
P
(21)
where T is the time duration of the received signal x(t).

The excitation wave is an acceleration signal. It is a seven and half-cycle tone burst enclosed
in a Hanning window, expressed by the following equation (22) and shown in Fig. 10. This
waveform has been chosen to reduce the leakage phenomena. The parameter f = 60 kHz is
the characteristic frequency of the emitting transducer.
y t
sin 2ft 1
cos
2ft
(22)
The emitter and receiver are piezoelectric transducers with a frequency of 60 kHz. The
emitter is connected to a signal generator PCG10 Velleman Instruments. Both transducers
are connected to a digital oscilloscope interfaced with a laptop. Due to the strong
280
Fig. 10. Excitation signal used in the DTT

attenuation of the transmitted signal, the peak to peak voltage of the generator has been
amplified from 13V to 39V using a transformer between the signal generator and the
emitter. The transformer has a transformation ratio of 1/3 and a frequency band of (10-200)
kHz. Moreover, an amplifier has been connected between the receiver and the oscilloscope.
It has gain equal to 100 or 200, a frequency band of (10-200) kHz and input impedance of
1M. The experimental setup is shown in Fig. 11.
Fig. 11. Experimental setup for DTT application

281
3.2.2 Acoustic tomography

Aim of the experimental program was also to point out the reliability of both travel time and
spectral attenuation AT in detecting building structures faults. With this purpose, the
algebraic problem of the tomography inversion has been deepened, the SIRT algorithm has
been selected and then numerically developed. After that, the solving algorithm has been
implemented in an automated procedure that allows the user to easily obtain a map of the
distribution of acoustic parameters (velocity and attenuation) in the selected section of the
item. The AT has been applied to a horizontal plane section crossing the wall in order to
intercept the central void (Fig. 8). The investigated section was thus 0.90 m wide and 0.38 m
thick, and it has been divided in 40 cells 0.09 m 0.095 m. By using this measurements
configuration, the section is crossed by 138 paths, and its coverage by the wave paths is
excellent. For each path the quantity (fS - fR)/ S2 has been calculated. Thus, systems (5) and
(16) consist in 138 equations and 40 unknowns, so as the two equations systems are
satisfactorily over determined.
Three kinds of signal have been evaluated in the experimental setup: pulse, sweep and chirp
signal. The short voltage pulse p(t), defined as:
p t
A rect
(23)
where T is the pulse duration and A the amplitude, is generally preferred for estimating the
travel time of elastic waves because it involves high power; on the other hand, when using
this signal only a poor control on the signal spectrum is usually achievable, thus making the
spectral tomography hardly possible. Because of the huge importance of the signal to noise
ratio for FFT spectral analysis, a broadband sweep signal has been preferred as source
signal. This signal, expressed by the following:
2
(24)
where T is the pulse duration, f0 is the lower bound of the frequency bandwidth which
increases with k = 2f/T rate and A is the amplitude, shows a linear relationship between
time and frequency. The purpose of using this signal was to extend the frequencies involved in
measurements up to 300 kHz, aiming at increasing the analysis resolution and at estimating
the attenuation coefficient more accurately in a wider signal band. Indeed, the evaluation of
the spectral attenuation coefficient requires a spread spectrum of both transmitted and
received frequencies in order to accurately estimate both centroid frequency and variance
values. Finally, the feasibility of using a chirp signal, described by the equation:
s t
Acos 2f t
rect
(25)
has been evaluated. This kind of signal allowed the performing of received signal spectrum
analysis similar to that achievable using the sweep one; moreover, it allowed crosscorrelating the received signal with the source one for travel time estimating, so as to
perform an effective travel time tomography.
Based on the requirements posed by the mathematical background presented in the
previous sections, an innovative measurements system was set-up according to the diagram
presented in Fig. 12.
282
RX
PXI
TX
A
PC
A
Fig. 12. Set-up of the measurement system
In order to simplify the measurement process, a multi-receiver solution was adopted, using
eight accelerometers as receivers for the generated wave. More detailed, the instrumentation
setup is composed of the following elements:
broadband piezoelectric transducer GRW350-D50 Ultran used as an ultrasonic wave

generator. The transducer has diameter of 520 mm, a central frequency of 370 kHz and
is fed with a high voltage signal (200 Vpp) for overcoming the high impedance of the
analyzed materials;
eight VS-150-M Vallen piezoelectric sensors (accelerometers) with a good frequency
response in the band of interest (100 500 kHz). Each of these sensors is coupled with
its own preamplifier with a gain of 40 dB, thus ensuring an elevated sensibility;
data acquisition system National Instruments PXI DAQ with two PXIe-6124 boards
having a total of 8 analog inputs and 2 analog outputs. The inputs have a sampling rate
of 4 MS/s each, at 16-bit resolution. The PXI rack is connected through a PCI-Express
card to a laptop computer that commands the entire measuring process.
The interface between the PXI data acquisition rack and the computer is based on the virtual
instruments created in the LabView environment, used both for acquiring the output of
the eight reception sensors as for generating the sweep signal for the piezoelectric
transducer. This signal is generated on one of the analog outputs of the acquisition board
and is subsequently amplified using a custom-build amplifier.
The entire measuring process is encapsulated in a user-friendly application that guides the
user by the means of a step-by-step procedure. At the beginning of the procedure, the
physical parameters of the structure can be specified, allowing the application to calculate
the number and position of the measurement points on the surface of the structure at each

283
step of the measure. Then the application triggers the generation of the signal and acquires
the output of the accelerometers for each measurement step. At the end of the cycle, the
tomographic algorithm previously mentioned is applied on the entire set of acquired data in
order to produce the tomographic map of the analyzed section.
3.3 Results
Results of both DTT and AT operative procedures have been displayed in terms of
distribution maps, in order to facilitate data interpretation and anomalies identification. The
maps are represented by a 256-levels gray-scale diagram, where the lowest level
corresponds to the white color, and the highest level corresponds to the black color. The 256
levels are normalized with respect to the range of the parameter values measured in the
represented map.
The map of the distribution of both signal velocity (Fig. 13) and signal power (Fig. 14) has
been derived interpolating the data recorded in the grids nodes (Cannas et al., 2008). Each
maps emphasizes areas where anomalies are located, according to features specific
diagnostic skill. In the center of both the maps the presence of the cavity region can be
clearly seen, although it is smaller than expected. The distribution of propagation velocity in
the rest of the map seems to be quite uniform, thus confirming that this parameter can easily
detect macroscopic anomalies, while it seems to be less affected by minor materials
discontinuity such as the mortar joints. The distribution of the signal power, on the other
hand, allows to better distinguishing the cavity in terms of both position and extension, and
points out that the various interfaces inside the structure affect this feature much more than
the velocity.
3000m/s
2700m/s
2400m/s
2100m/s
1800m/s
1500m/s
1200m/s
900m/s
Fig. 13. Propagation velocity [m/s]: distribution map.
284
6,51E+01
1,00E+07
Fig. 14. Signal power [mV]: distribution map.
Thickness
Figures 15 and 16 show respectively the results of the performed Travel Time and Spectral
Attenuation Tomography measurements on the selected horizontal section of the full-scale
real masonry (Concu et al., 2010). Showed results of Spectral Attenuation AT have been
achieved by using the sweep signal as source.
Width
Fig. 15. Travel Time Tomography measurements

285
Thickness
50 dBsm1
Width
0.05 dBsm1
Fig. 16. Spectral Attenuation Tomography measurements

It can be noticed that both maps identify a central region corresponding to the position of
the real cavity. The velocity map presents a poor quality in terms of the position and also the
extension of the cavity, both parameters being significantly improved by using the
attenuation map. It can be also seen that the attenuation map presents a high degree of
scattering, confirming the hypothesis that this parameter is very sensible to all
discontinuities of a section such as joints, small irregularities, various interfaces. It is worth
noting that a tighter grid of measurements would enable a better definition of the cavitys
shape, implying longer measuring and computational times.
Results of DTT and AT thus emphasize the main role played by frequency dependent
features, such as signal power and spectral attenuation coefficient, when dealing with
materials and structures with a quite important degree of non homogeneity. It is worth
stressing that the maps derived from DTT do not allow the anomaly to be collocated at its
real and correct distance from the wall surface, while the AT maps give a faithful picture of
the selected section.
4. Comments on measurements problems

Acoustic measurements are still affected by a quite large amount of uncertainness in terms
of stability and reproducibility. This is primarily due to the large number of factors that
somehow exert influence on measurements:
materials (chemical-physical-mechanical conditions);

signals;
acoustic parameters;
instrumental sets;
data processing methods;
286
environmental conditions;
human factor.
Another big problem concerning acoustic measurements is the gap that exists between the
theoretical and interpretative bases and the codification of operative modalities, which is to
say the almost total lack of standardization procedures. Actually, only standards concerning
the determination of ultrasonic pulse velocity in concrete and natural stones are available
(EN 12504-4, 2004; EN 14579, 2004). These standards take into account the problem of
measurements stability, repeatability and reproducibility, and address the main factors
influencing pulse velocity measurements. It is stated that to obtain some measure of the
acoustic velocity that is reproducible and which is essentially a function of the properties of
the material (concrete and stone) submitted to testing, it is necessary to take into
consideration the different factors that exert an influence on the velocity. This is also
essential to establish the correlations existing with the different physical characteristics of
the material. The factors that should be carefully considered are the following:
a.
b.
c.
d.
e.
moisture content and temperature of the concrete

water content of the stone
path length
shape and size of the specimen
cracks, fissures and voids.
Moisture and temperature of the concrete and water content of the stone
Moisture content has both chemical and physical effects on ultrasonic pulse velocity.
These effects are important especially for establishing correlations with concrete
strength. As an example, acoustic velocity can vary significantly between a properly
cured standard specimen and a structural element made of the same concrete, because
of the influence of curing conditions on cement hydration and the presence of free
water in the voids. In the same way, water content has some effects on pulse velocity
propagation in stones. Stone humidity, that is to say the presence of water inside the
pores, can cause a variation of ultrasonic velocity value up to 50 % with respect to dry
specimens or structural components. Concrete temperature effects on ultrasonic pulse
velocity should be taken into account only outside the range 10C - 30C. Within this
range, no significant change in pulse velocity has been experimentally found if
corresponding changes in strength or elastic properties do not occur.
Path length, specimen shape and size
The path length over which the pulse velocity is measured should be long enough not
to be significantly influenced by the heterogeneous nature of the concrete or the stone.
For concrete specimen the standard recommends minimum values of the path length
depending on aggregates nominal maximum size. The velocity is not generally affected
by the variations in the path length, even though the electronic timing devices are likely
to give some indications that the velocity can slightly decrease if the length of the path
increases. This is because the higher frequency components of the pulse are attenuated
more than that lower frequency components and the shape of the onset of the pulse
becomes more rounded with increased distance travelled. Thus, the apparent reduction
of pulse velocity arises from the difficulty of defining exactly the onset of the pulse and
this depends on the particular method used for its definition. The apparent reduction in
velocity is normally slight and is within the accuracy of measurements commercial
apparatus. Nevertheless, particular care shall be taken when measurements are carried
out on important path lengths. The velocity of short vibratory impulses is independent

287
of the size and shape of specimen in which they travel, unless its least lateral dimension
is less than a minimum value. Below this value, the pulse velocity can be reduced
appreciably. The amount of this reduction depends primarily on the ratio ultrasonic
wavelength to smallest lateral dimension of the specimen, but is insignificant if the ratio
is less than unity. Standards give the relationships existing between pulse velocity,
transducers frequency and the smallest admissible lateral dimension of the test
specimen. If the minimum lateral dimension is less than the wavelength or if the
indirect transmission arrangement is used, propagation mode changes and, therefore,
measured velocity will be different. This is particularly important in cases where
elements of significantly different sizes are being compared.
Cracks, fissures and voids
When a wave passing through a specific item encounters any discontinuity, the wave
power is most likely attenuated because of scattering phenomena, so that any crack,
fissure or void inside the item might obstacle wave propagation. The higher the
projection of defects length with respect to transducers size and wavelength, the greater
the influence on wave propagation. In this case, the first impulse picked up by the
receiving transducer will undergo a diffraction at the edge of the anomaly and this
gives a longer path time compared to a transmission taking place in items with no
fissures or voids. If part of the signal is already able to reach the receiver e.g. in
cracked elements where the broken sides are kept firmly in contact by compression
then the propagation time may not significantly change with respect to paths with no
defects throughout. Transit time is not a decisive parameter also when the crack is filled
with a liquid able to transmit wave power. In all these cases, examination of signal
attenuation may also provide helpful information. Therefore, it would be reasonable to
approach the ultrasonic analysis not only in terms of propagation times, but also in
terms of other waves features changes. The higher the intrinsic non-homogeneity level
of structures, e.g. masonries, the greater the advisability of this integrated approach.
The different applications of sonic and ultrasonic investigations have brought to the fore
two fundamental issues in the performance of such tests and their repeatability: the
influence of pressure exerted on the transducers employed and the effect on acquired data
of the application of acoustic coupling agents inserted between the material under
examination and the transducers.
Experiments performed on granite specimens in different operative conditions, that is, in
presence or absence of the acoustic coupling agent and of pressure on the transducers,
allowed us to evaluate the impact of these factors on the characterization of the material and
the repeatability of measurements (Concu, 2002). It emerged that the best operating
conditions, which ensure excellent measurement repeatability and reliability, are those that
call for both the use of the coupling agent placed between the surface of the material and the
transducers used for measuring and the application of a constant pressure on the
transducers throughout the entire test period. The uncertainty introduced by the variability
in pressure and the conditions of the coupling agent (thickness, temperature, viscosity) is in
fact converted into a variation of the energy associated with the transmitted signal; this
impacts to a greater degree on the wave shape and the signal spectrum rather than on its
propagation velocity. From the experiments performed another interesting datum emerged,
that is, the influence on the measurement of time between application of the acoustic
coupling and data acquisition (Concu & Fais, 2003). Analysis of the values of maximum
wave amplitude, amplitude of the maximum spectral peak and velocity relating to
288
ultrasonic signals acquired repeatedly over time, starting from the instant of application of
the coupling agent, led to the observation of a behavior of the aforementioned characteristic
parameters of the material under examination: as can be seen in Figures 17-19 the three
parameters increase progressively for approximately the first sixty minutes from application
of the acoustic coupling agent and then become constant. Such behavior is to be taken into
account in defining for each tested material the optimal instant for data acquisition so as to
reduce the influence of the coupling agent on the test and its repeatability.
2,4
2
1,6
1,2
0,8
0,4
0
0
40
80
120
160
200
240
280
240
280
Fig. 17. Maximum wave amplitude [V] vs time [minutes]
12
10
8
6
4
2
0
0
40
80
120
160
200
Fig. 18. Amplitude of the maximum spectral peak vs time [minutes]
4200
4100
4000
3900
3800
3700
3600
0
40
80
120
160
Fig. 19. Propagation velocity [m/s] vs time [minutes]
200
240
280

289
5. Conclusions
The chapter presented an overview on non invasive acoustic measurements applied for
faults detecting in building materials and structures. The state of art has been described,
focusing on two main operative procedures: the Direct Transmission Technique and the
Acoustic Tomography. Special emphasis has been dedicate to both Travel Time and Spectral
Attenuation Tomography, by deepening the aspects of numerical modeling, resolution
algorithms, choice of source signal, measuring systems setup.
A case study, reporting the results of an experimental program started with the aim of
deepening the reliability of acoustic methods in buildings faults detecting and materials
characterization, has been described. Both DTT and AT approaches have been applied to a
full scale masonry model with known anomalies inside. Results showed that both DTT and
AT can successfully apply to stone masonry characterization. Moreover, results confirmed
the suitability of the Spectral Attenuation Tomography - developed according to the model
proposed by Quan and Harris, 1997, for seismic surveying in buildings materials and
structures faults detection.
Finally, an outline of the most common problems affecting the acoustic non destructive
testing has been illustrated, addressing the main factors influencing the acoustic
measurements in terms of stability, repeatability and reproducibility. It has been pointed out
that particular attention should be given to chemical-physical-mechanical conditions of the
material and to testing modalities such as the pressure exerted on the transducers employed
and the effect on acquired data of acoustic coupling agents.
Further researches should deepen various facets, including:
the potentiality of a novel approach involving the integrate analysis of different

features, associated with acoustic waves propagating through the material, acquired
both in time and frequency domain;
the implementation of different inversion algorithms, chosen among the most robust
and commonly used, for tomographic measurements, with the aim of highlighting the
most suitable one for the specific algebraic problem solution;
the development of proper measurements setup and automated measuring procedure
which allowed problems of measurements stability, repeatability and reproducibility to
be minimized.
6. References
Belanger, P.& Cawley, P. (2009). Feasibility of low frequency straight-ray guided wave
tomography. NDT&E International, Vol.42, No.2, March 2009, pp. 113-119
Berryman, J.G. (1991). Non linear inversion and tomography. Lecture notes, Earth Resource
Laboratory, MIT
Best, A.I., McCann, C. & Sothcott, J. (1994). The relationships between the velocities,
attenuations, and petrophysical properties of reservoir sedimentary rocks.
Geophysical Prospection, Vol.42, No.2, February 1994, pp. 151-178
290
BS 1881: Part 203, Recommendations for measurement of the velocity of ultrasonic pulses in
concrete, London, 1986
Cannas, B., Cau, F., Concu G. &Usai M. (2008). Features extraction techniques for sonic and
ultrasonic NDT on building materials. Proceedings of 1st International Symposium on
Life-Cycle Civil Engineering, ISBN 978-0-415-46857-2, Varenna, Lake Como, Italy,
June, 2008
Carino, N.J. (2001). The Impact-Echo Method: An Overview. Proceedings of the 2001
Structures Congress & Exposition, Washington, D.C., May 21-23, 2001
Concu, G. (2002). Il problema del controllo delle prestazioni degli elementi lapidei impiegati
nelle costruzioni, PhD. Thesis, University of Cagliari, Italy
Concu G., De Nicolo B. & Mistretta F. (2003a). Comparative evaluation of two nondestructive acoustic techniques applied in limestone masonry diagnosis,
Proceedings of 33rd International Conference and Exhibition Defectoscopy, ISBN 80-2142499-0, Ostrava, Czech Republic, November, 2003
Concu G., De Nicolo B., Mistretta F.& Valds M. (2003b). NDT ultrasonic method for ancient
stone masonry diagnosis in Cagliari (Italy) , Proceedings of Structural Faults and
Repair - 2003, ISBN 0-947644-52-0, London, UK, July, 2003
Concu, G. & Fais S. (2003). In time analysis of a viscous coupling agent effect in ultrasonic
measurements. Proceedings of 3rd International Conference on Non-Destructive Testing
of the Hellenic Society for NDT NDT in Antiquity and Nowadays Skills-ApplicationsInnovations, Chania, Crete, Greece, October, 2003
Concu, G., De Nicolo, B., Piga, C. & Popescu, V. (2010). Measurement system for nondestructive testing using ultrasonic tomography spectral attenuation. Proceedings of
12th International Conference on Power Electronics and Electrical Engineering, ISBN 9781-4244-7020-4, Brasov, Romania, May, 2010
Dines, K.A. & Lytle, R.J. (1979). Computerized geophysical tomography. Proceedings of IEEE
67(7), pp. 1065-1073
Enoki, M. & Kishi, T. (1991). Acoustic Emission. Current Practice and Future Directions, ASTM,
Philadelphia
EN 12504-4 (2004). Testing concrete- Part 4. Determination of ultrasonic pulse velocity
EN 14579. (2004). Natural stone test methods. Determination of sound speed propagation
Gilbert, P. (1972). Iterative methods for the reconstruction of three-dimensional objects
from projections. Journal of Theoretical Biology, Vol.36, No.1, July 1972, pp. 105117
Gordon, R., Bender, R. & Herman, G.T. (1970). Algebraic Reconstruction Technique (ART)
for three-dimensional electron microscopy and x-ray photography, Journal of
Theoretical Biology, Vol.29, No.3, December 1970, pp.471-81
Gordon, R. (1974), A tutorial on ART. Proceedings of IEEE Trans. Nuclear Science, NS-21, pp.
78-93
Herman, G.T. (1980). Image Reconstruction from Projection: the Fundamentals of Computerized
Tomography, Academic, New York
Hudson, J.A. (1981). Wave speeds and attenuation of elastic waves in material containing
cracks. Geophysical Journal International, Vol.64, no.1, January 1981, pp. 13350

291
Ilyas, A. (2010). Estimation of Q factor from reflection seismic data, Report GPM 6/09,
Department of Exploration Geophysics, Curtin University of Technology, p. 65
Ivansson, S. (1986). Seismic Borehole Tomography Theory and Computational Methods.
Proceedings of IEEE 74, pp. 328-338
Jansen, D.P., Hutchins, D.A., Ungar, P.J. & Young R.P. (1991). Acoustic tomography in solids
using a bent ray SIRT algorithm. Nondestructive Testing and Evaluation, Vol.6, No.3,
pp. 131-148
Johnston, D.H. (1981). Attenuation: A state-of-art summary. In Toksz and Johnston, Eds.,
Geophysics reprint series, Vol.2, Soc. of Expl. Geophys
Kaiser, J. (1950). An Investigation into the Occurrence of Noises in Tensile Tests or a Study
of Acoustic Phenomena in Tensile Tests. Ph.D. Thesis, Tech. Hosch. Munchen,
Munich, Germany
Kepler, W.F., Leonard Bond, J. & Frangopol, D.M. (2000). Improved assessment of mass
concrete dams using acoustic travel time tomography. Part II: application.
Construction and Building Materials, Vol.14, No.3, April 2000, pp. 147-156
Knopoff, L (1964). The Convection Current Hypothesis. Reviews of Geophysics, Vol.2, No.1,
pp. 89-122
Krautkramer, J.& Krautkramer, H. (1990). Ultrasonic testing of materials, Springer Verlag,
New York
Lakshiminarayanan, V.A. & Lent A. (1979). Methods of least squares and SIRT in
reconstruction. Journal of Theoretical Biology, Vol.76, No.3, February 1979, pp. 267295
Jansen Leonard Bond, J., Kepler, W.F. & Frangopol, D.M. (2000). Improved assessment of
mass concrete dams using acoustic travel time tomography. Part I: theory.
Construction and Building Material, Vol.14, No.3, April 2000, pp. 133-146
Lo, T.W., Inderwiesen, P. (1994). Fundamentals of Seismic Tomography. Soc. of Expl. Geoph.,
Tulsa
Meglis, I.L., Chow, T., Martin, C.D. & Young, R.P. (2005). Assessing in situ microcrack
damage using ultrasonic velocity tomography. International Journal of Rock
Mechanics & Mining Sciences, Vol.42, No.1, January 2005, pp. 25-34
NDT Education Resource Center. Larson, B. Editor. (2001-2011). The Collaboration for NDT
Education. Iowa State University. Available at:
<www.ndt-ed.org>
Ohtsu, M. & Ono K. (1986). The generalized theory and source representations of acoustic
emission. J. of Acoustic Emission, Vol. 5, No. 4, October-December 1986, pp. 124133
Priestley, M.B. (1981). Spectral Analysis and Time Series. vol. I, Academic Press, New York
Quan, Y. & Harris, J.M. (1997). Seismic attenuation tomography using the frequency shift
method. Geophysics, Vol.62, No.3, May-June 1997, pp. 895-905
Rhazi, J. (2006). Evaluation of concrete structures by the Acoustic Tomography Technique.
Structural Health Monitoring, Vol.5, no.4, December 2006, pp. 333-342
Sansalone, M. & Streett, W.B. (1997). Impact-Echo: Nondestructive Evaluation of Concrete and
Masonry. Ithaca, N.Y., Bullbrier Press
Sheriff, R.E. & Geldart L.P. (1995). Exploration Seismology. Cambridge Univ. Press, London
292
UNI EN 1926 (2000). Natural stone test methods. Determination of compressive strength
UNI EN 14580 (2005). Natural stone test methods. Determination of static elastic modulus
14
Approximation and Correction of Measuring
Transducers Characteristics
Janusz Janiczek
Wroclaw University of Technology,

Poland
1. Introduction
A very frequent case is that the characteristics of measuring transducers are non-linear and,
apart from that, the output signal value may depend on several input values or unwanted
influencing factors such as temperature, humidity etc. The need to adjust the form of
information about the measured value to human perception possibilities, or to simplify the
further processing of measurement information generally requires such adjusting of the
slotted line element characteristics so that the resultant characteristics of the slotted line is
linear, with assumed error, in the adopted coordinate system. (Bucci et al., 2000; Grzybowski
@ Wagner, 1971; Iglesias @ Iglesias, 1988; Patranabis et al., 1988; Patranabis @ Ghosh, 1989)
When the measuring transducers characteristics is very non-linear, for example in the case
of certain transducers for light intensity measurement, a problem with selecting a relevant
A/D transducer may occur. All methods of correcting measuring transducers
characteristics cannot eliminate the error related to insufficient A/D transducer resolution.
That issue is illustrated by figure 1.
N"2
Ux
U"x
U'x
N
1.
2.
3.
4.
N"3
0
N N'3
N'2
measuring transducers characteristics
A/C transducer characteristics of small resolution
A/C transducer characteristics of high resolution
non-linear A/C transducer characteristics
x
x
Fig. 1. Impact of A/D transducer characteristics on resolution error.

At large non-linearity of the measuring transducers characteristics (curve 1 on figure 3.5)
and small inclination of A/D converter characteristics (line 2), the A/D converter resolution
294
is sufficient to process the voltage signal Ux to an appropriate digital code N2 in the area
in which the inclination of measuring transducers characteristics is high. On the other hand,
in the area of small inclination of the measuring transducers characteristics, the changes in
output voltage Ux of the transducer are too small to be properly processed by the A/D
converter to digital code N2. That problem may be partially bypassed using an A/D
converter of high resolution (line 3). However, a better solution would be to adjust the
processing characteristics of an A/D converter to the measuring transducers characteristics
(curve 4). The A/D converter characteristics should be described by a function reverse to the
function describing the measuring transducers characteristics and then the following linear
relation would be obtained N=c x.
That chapter discusses issues related to the approximation of measuring transducers
characteristics and their correction in an analog to digital (A/D) converters The thesis will
consider only static characteristics of measuring transducers.
2. Transducers with successive compensation

The block diagram of a transducer with weight compensation is presented in figure 2.1.
Ux
S/H
UREF
+
-
Control block
Approximation register
MSB
LSB
Digital-to-analog
converter
Fig. 2.1. General scheme of transducers with successive compensation.

It consists of a sample-and-hold circuit, a comparator, a digital-to-analog converter, an
approximating register and a control circuit. The transducers operating principle can be
found in readily available literature.
In transducers with weight compensation, the only available element, through which the
shape of a transducer processing characteristics may be influenced is the reference voltage
UREF . For that reason, the change of processing characteristics can be achieved by generating
a non-linear reference voltage in relation to the digital output signal N:
UREF = ag(N)
(1)
where: g() - measuring transducers characteristics approximating function,

a - proportionality coefficient.
The circuit of adjusting transducer processing characteristics with weight compensation by
means of reference voltage generation method has been presented in figure 2.2.
Approximation and Correction of Measuring Transducers Characteristics
295
Ux
COD I
+
-
D/A I
control
circuit
UREF
COD II
D/A II
Register
B(0,..., i-1) B(i, n-1, n)
D/A III
Fig. 2.2. Circuit of adjusting transducer processing characteristics.

Converter circuit consists of a comparator, a control circuit, a digital register and a reference
voltage generator containing two coding circuits CODI and CODII, three digital-to-analog
converters D/AI, D/AII and D/AIII and an adder. D/A III converter is required to have
an input for external reference voltage U0III, thanks to which the multiplication function may
be performed:
UIII = U0IIIkIII N
(2)
where: kIII - processing coefficient of D/A converter III.

UIII - output voltage of C/A converter III
N" - digital signal value controlling D/A converter III, dependent on bits Bi-1, B0
That allows for an adequate change of the u value of voltage spike to one bit, and hence the
change of inclination of the generated voltage characteristics. The measuring transducers
characteristics should be approximated by a range-linear function with the assumed error,
figure 2.3 a), while the range points Ai, Aj, should be selected so that their values meet
directly, or with the proportionality coefficient, the values Ni, Nj, of the A/C transducer
output signal.
a)
b)
Ux
Uj
Ui
0
UREF
Uj
Ui
Ai
Aj
Ax
Ni
Nj
Fig. 2.3. a) Transducer characteristics and its approximation; b) waveform of reference voltage.
More significant bits (Bn, Bi) of the approximating register, through the coding circuit
KODI control the D/A converter I. The CODI circuit, which may be ROM memory, contains
a board converting the signal value from the register to such a signal value controlling the
D/A converter, as to ensure proper processing characteristics, figure 2.3 b).
The transducer converting process is conducted in the same way as for a linear transducer,
figure 2.4, i.e. bits are issued one after another, starting from the most significant. When
voltage from the generator exceeds the value of the voltage processed recently, the preset bit
is withdrawn and another bit is incorporated. Testing all the more significant bits Bn, Bi
means that a specified approximation range has been selected, in which the processed input
signal is located (status of bits Bn, Bi provides the values Ni, Nj ). Appropriate inclination of
the characteristics required for a given approximation range is received by changing the
reference voltage of D/A converter III. The value of that voltage is obtained from D/A
296

UI
Ux
tact
Bn Bi
Bi-1 B0
Fig. 2.4. Waveform of voltage generated in the generator circuit.

converter II controlled by bits Bn, Bi, through the CODII circuit. Hence, after setting a set of
bits Bn, Bi relevant to the input voltage by the approximating register, further converting
process takes place, during which the less significant bits (Bi-1, B0) control the D/A
converter III which already has properly selected inclination of the characteristics.
Beginning of the range voltage values are obtained from D/A converter I and the values of
voltages in particular ranges by adding voltage value from D/A converter I with voltage
from D/A converter III. Output voltages of converter s C/A I, C/A II and C/A III for i-th
range are respectively:
UIi = kII(N)
(3)
UIIi = kIIII(N)
(4)
UIIIi = kIII UII k (N)
(5)
where: I - coefficient value received from cell of memory COD I with address N'.
ki - coefficient values of transducer processing.
II - coefficient value received from cell of memory COD II of address N'.

N' - more significant part of number N of control circuit (bits Bn, Bi).
N" - less significant part of number N of control circuit (bits Bi-1, B0).
After summing up voltages from converters D/A in the adder, UERF voltage is obtained,
defined by the relation:
UREF = kII(N) + kIIII(N) k N
(6)
The manner of adjusting A/D converter processing characteristics with weight

compensation described above requires that the linearized measuring transducers
characteristics is approximated by a range-linear function. Therefore, designing the circuit
should be started by determining the approximating function with the assumed error.
However, in order not to complicate the shaping circuit, the lengths of approximation
ranges should be equal.
The additional errors, which may be introduced by the shaping circuit, are errors mainly
from the D/A converter I, especially the linearity error, and the adder. The errors of other
transducers have a smaller impact since the values of voltages generated by them are
smaller than the voltage from D/A converter I.
The above-mentioned shaping circuit, in a simplified configuration, was completed in a
universal temperature meter cooperating with transducers Pt-100, Pt-1000, Ni-100, PtRh-Pt,
NiCr-NiAl, Fe-CuNi. The simplified scheme of the meter which allows explanation of its
297
operating principle is presented in figure 2.5. Input circuits containing resistance to voltage
transducers and amplifiers have been omitted here.
V0
R1 R2 R3
R4 R5
R6 R7
MAX384
IA3
IA2
IA1
IA0
R8
R9
R10
R11
A
EN
RS
WR
IB3
IB2
A0
A1
IB1
Ux
VCC
Iwe
DAC08
+
K
-
control
circuit
approximation
register
B(0,..., 7) B(8, 9)
IB0
Fig. 2.5. Simplified scheme of a temperature meter.

Because the assumptions require that the absolute error of temperature measurement,
submitted by the meter, to be within 1 C for platinum transducers, 2 C for nickel
transducer and 5 C for thermocouples, the characteristics of those transducers could be
approximated by four sections. Thus, in the place of D/A converter I and the circuit COD I,
a voltage generator, containing dividers R1 R7 and half of analog selector MAX384, has
been used. The DAC08 circuit has been used as a C/A III converter. That is a digital-toanalog converter with current output whose inclination of characteristics can be selected by
changing the current input IIN. The current can be changed through selecting, in the second
part of the analog selector, one of resistors R8 R11. For each measuring transducer, a
separate module containing the selector along with dividers of voltages and current resistors
was intended. Module selection was done by specifying the signal for input EN.
Converter circuit operates like an ordinary converter with successive approximation, with
addition that in the first four timings (bits B9 and B8 ), the approximation range is
determined, where the Ux voltage is located. The voltage from the generator, supplied to K
comparator's reversing input, has the beginning of the approximation range value. The
value of DAC08 transducer input current is also determined for the selected approximation
range. In the other processing timings, the less significant bits are activated (B7 B0) and,
through voltage drop on resistor R caused by DAC08 transducer current output, the exact
definition of Ux voltage value takes place. The status of bits B9 B0 directly provides the
value of measured temperature.
The conducted meter measurements indicated that it meets the parameters adopted in the
assumptions concerning ambient temperatures 5 35 C. Processing time was 2.5 ms.
Short converter processing time with successive approximation was used in the meter to
measure carbon monoxide concentration. The TGS 2444 measuring transducer
manufactured by Figaro USA Inc was used in the meter. In that transducer, the value of
carbon monoxide concentration is processed into resistance. The measuring range is within
30 to 1000 ppm. Transducer characteristics has been presented in figure 2.6. It presents
relative changes of output resistance, related to nominal resistance. Nominal resistance is
given for concentration 100 ppm and ranges from 13.3 k to 133 k. In applied transducer
the nominal resistance was 15.2 k.
298
Fig. 2.6. Static characteristics of transducer TGS2442.

Transducer characteristics, presented in figure 2.6, shows a great change of inclination. For
30 ppm, the inclination is -1,52 k/ppm and for 1000 ppm - -1,26 /ppm (TGS 2611). With
such high non-linearity of the measuring transducers characteristics, using an A/C
transducer with influenced processing characteristics is beneficial.
Because the meter was expected to be a portable device with battery power supply, a
microcontroller MSP430F435 equipped with an LCD field direct control module was used in
it. There is an A/C transducer with weight compensation with resolution of 12 bits and
processing frequency of up to 200 khz within the internal structure of the micro-controller.
Calculations proved that the resolution of that transducer is too small to receive a proper
resolution of measurement for high gas concentrations. Therefore, a circuit capable of
adjusting transducer processing characteristics needed to be used.
The value of the digital A/D converter output signal is specified by the relation:
N = 4095
Ux
U REF
(7)
The reference voltage UREF is accessible from the outside, which enables manipulation of
transducer processing characteristics. The simplified scheme of the meter has been
presented in figure 2.7. The device consists of a microcontroller MSP430F435 manufactured
by Texas Instruments, an LCD field controlled directly from the microcontroller, a double
D/A converter, an RS measuring transducer, highly stable resistors R1, R2 and R3, a heating
circuit (elements T2 and R4), a circuit of switching on the measurement resistor (elements T1
and R5) and elements related to power supply and keyboard not shown in the figure.
The measuring transducer requires relevant control, which entail cyclic switching on the Rh
heater and in the appropriate phase, switching on the RS measurement resistor, figure 2.8.
Proper operation of a measuring transducer requires that the heating is switched on for 14
ms and, after 981 ms from turning the heating off, the measurement resistors power supply
299
VCC
Ain1
+
-
T1
R3 R2
R4
R5
Va
Vb
VCC
P3.1
P3.3
P3.0
P6.0
P6.1
T2
LCD
Ain0
UREF
R1
MSP430F435
CN6292
OUTB
Ux
OUTA
Rh RS
DIN
CLK
CS
UREF
MAX5522
Fig. 2.7. Scheme of A/C transducer characteristics matching circuit.
heating
Va
986 ms
14 ms
Vb
14 ms
measurement
5 ms
95 ms
5 ms
Fig. 2.8. Signals controlling measuring transducer.

is switched on. The measurement resistor activation time is to last 5 ms and the
measurement should be taken during that period.
The RS measuring transducer is connected with resistor R1 in a divider circuit, which
enables changing the measuring transducers resistance to a voltage corresponding to it.
Output voltage from the divider, after strengthening x6 in the amplifier, is supplied to A/D
converter input. VCC voltage value was adopted as 3.3V (voltage supplying the
microcontroller). The value of resistor R1 has been selected to obtain the broadest range of
voltage changes Ux . Calculations, which have been omitted here, provided the value of 10
k, which provides the range of voltage changes from 0.47 V to 2 V. The chart of relation
between gas concentration s and Ux voltage value has been presented in figure 2.9.
The function presented in figure 2.9 is approximated by a range-linear function. With
approximation error not exceeding 0.1%, 8 approximating sections were sufficient for
approximation. However, the approximating function does not need to be continuous on
range endpoints. The values of range points and coefficients ki of section inclination in
particular ranges have been presented in figure 2.10
Inclination coefficients ki are determined by the relation:
ki =
sPi+1 - sPi
U Pi+1 - U Pi
(8)
300
1000
900
800
s [ppm]
700
600
500
400
300
200
100
0
0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0 2.2 2.4
Ux [V]
2.6 2.8
Fig. 2.9. Relation between gas concentration s and Ux voltage value.

s[%]
k7
k6
sP7
k5
sP6
k4
sP5
k3
sP4
sP3
k2
k1
sP2
k0
sP1
sP0
UP0
UP1
UP2
UP3 UP4 UP5 UP6UP7
Ux[V]
Fig. 2.10. Approximation by range-linear function.

The measurement takes place in two phases. In the first phase, input An1 is active and the
initial voltage measurement Ux takes place on the divider output. In that phase, the
reference voltage for an A/C transducer is the micro-controller internal voltage. That
measurement provides information in which range voltage Ux is located. Then, for a given
range, voltage Uoi is generated from D/A converter output A for the amplifier. The value of
that voltage is selected so that for the value Ux = UPi, the voltage on amplifier output is equal
to zero. At the same time, from the transducer output B, a reference voltage UREF is
generated for A/D converter. The value of that voltage should allow to obtain a satisfactory
measurement resolution in a given range.
The voltage on output B of D/A converter is set proportionally to inclination coefficient
value ki. Such proceeding stems from the fact that the measurement value in i-th range is
specified by the relation:
s = sPi + k i (U x - U Pi )
(9)
301
The value of voltage difference Ux = UPi, after processing in A/D converter, can be specified
using the formulation (7):
U x - U pi =
N xi U REF
4095 k
(10)
where: k amplifier gain

Nxi - value of processing result in i-th range.
Thus:
s = spi +
k i N xi U REF
= spi + si
4095 k
(11)
If we assume that a linear relation has to be met:

si = c Nxi
(12)
then the reference voltage should meet the condition:

U REF =
4095 c k
ki
(13)
Meeting the aforementioned condition may be difficult owing to a limited range of changes
in reference voltage. In that particular case, in accordance with catalogue data, the minimum
value of that voltage cannot be smaller than 1.4 V, and the maximum may not be greater
than VCC 0.2 V. For that reason, it has been assumed that the reference voltage value
should ensure the required measurement resolution:
U REF
4095 c k
ki
(14)
The final measurement is calculated using the formulation (11). The shaping circuit
introduces two kinds of errors. The first is connected with the approximation of the
measuring transducers characteristics by a range-linear function. It stems from the possible
finite values of coefficients ki and threshold voltages UPi, what is caused by their values
being generated by converter D/A. The fact that the approximating sections do not have to
be continuous in range points, facilitates approximation.
The second type of errors stems from the accuracy of generating the UREF voltage and errors
of A/D converter. D/A converters are 10-bit converters with errors: linearity - 1 LSB, offset
1 mV, characteristics inclination 0.5 LSB. Those errors will have the greatest impact at
endpoints of the range, where there is greatest inclination of measuring transducers
characteristics. Taking into account the above values, it has been calculated from
formulation (11) that the maximum absolute error s should not exceed the value of 8 ppm.
The conducted simulation research, in which the measuring transducer was replaced with a
resistance decade, indicated that the relative measuring error d, (referred to the endpoint of
the measuring range) in the whole measuring range did not exceed the value of 0.7%, figure
2.11.
302
d [%]
0.8
0.6
0.4
0.2
0
-0.2
-0.4
-0.6
-0.8
100 200 300 400 500 600 700 800 1000

s [pm]
Fig. 2.11. Chart of relative error of carbon monoxide concentration meter.
3. Dual slope converters

Dual slope analog-to-digital converters with are very popular owing to their simple
structure, high reliability, high resolution, high processing accuracy and were one of the first
to be integrated. They are mainly intended for measuring constant voltages and most often
cooperate with different kinds of measuring transducers.
The block diagram of a double integral converter has been presented in figure 3.1 and the
waveforms in particular points of the transducer circuit in figure 3.2.
C
k1
k2
Ux
UREF+
UREF-
R
k3
UI
+
K
+
UK
US
T0 t+ t-
Fig. 3.1. Block diagram of double integral converter.

UI
Ux < 0
Ux > 0
UK
k1
k2
k3
T0
T0
Fig. 3.2. Waveforms in particular points of double integral converter.

The converting process takes place in two phases. In the first phase, which lasts a given time
T0, the processed voltage Ux is connected to the integrators input through key k1. Then, after
time T0, voltage Ux is disconnected and reference voltage with polarization opposite to the
303
polarization of voltage Ux is connected to the integrators input. The level of voltage UK at

the comparator output indicates the polarity of the processed voltage Ux. Reference voltage
integration lasts until the moment when voltage UI at the integrators output returns to the
initial state which will result in activation of comparator K. Time T of reference voltage
integration determines the result of processing.
Assuming the ideal integrator and comparator operation, voltage UI at the integrators
output is defined by the relation:
UI =
T0
U xdt 0
T0 +T
(15)
U REF (t)dt = 0
T0
where: = RC integrator time constant

If voltage Ux is constant over processing, then relation (15) gives the following:
T
Ux =
1
U REF (t)dt
T0 0
(16)
Voltage UREF also has a constant value in a transducer with linear processing and then:
U x = U REF
T
T0
(17)
Double integral converters typically have a resolution limited to about 13 14 bits. It stems
from the fact that, with an increasing transducer resolution, the comparator must
differentiate between increasingly smaller voltages at the integrators output. The
consequence is the extension of the comparator's response time and its unstable operation.
In addition, the presence of noise, interference, hysteresis of the comparator and its
insensibility threshold results in the fact that minimum voltage distinguished by the
comparator is limited. Increasing transducer resolution, at the same time of voltage
integration, would require increasing the clock generators signal frequency. The abovementioned factors would lead to transducers unstable operation.
The need to increase transducer resolution to 16 18 bits resulted in preparing a modified
double integral transducer circuit whose scheme has been presented in figure 3.3 and the
waveforms of signals in selected points of the circuit - in figure 3.4.
Ux
UREF+
UREF-
k1
k4
k2
k3
A1
+
R1
A2
+
R2
C
Cz
UI
-Up
Up
k5
k6
Fig. 3.3. Modified integrating transducer circuit.
K3
+
Generator fn
K2
+
UK2
+
K1
-
Uk1
Counter
US
Tz T0 t+ t-
304
UI
Up
t
fn
k4, k5, k6
k1
k2
UK1
Uk2
T0
t1
t2
Fig. 3.4. Waveforms of signals in selected points of transducer circuit.

In comparison to the basic circuit of double integral converter, the circuit presented in figure
3.3 has been expanded with two input buffers, two additional comparators and elements
allowing cyclical zeroing of the transducer circuit. The control circuit has also been changed.
The modified transducer circuit operates in the following manner: first, zeroing of the set
integrator comparator K1 takes place. That takes place when keys k4, k5 and k6 are closed.
Other keys are open. Over that period, the integrator and the comparator K1 are enclosed by
the feedback loop and the voltage on the integrators reversing input reaches the value of that
amplifiers offset voltage. Because the inputs of buffering amplifiers are also dense to the mass,
the difference of offset voltages of input amplifier and the integrator amplifier is depositing on
capacitor Cz. That voltage is remembered on the capacitor at all times throughout the process
of converting. As a result, the impact of all amplifiers offset voltage is disposed.
During integration of the processed input voltage Ux which lasts for T0, when voltage UI at
integrators output exceeds the value of voltage Up, which is the reference voltage for the
comparator. Then the comparator activates, which, in turn, causes such action of the control
circuit US that, synchronically with impulse fn of the clock generator switching on takes
place - respectively to polarization of the processed voltage - of key k2 or k3 supplying the
reference voltage with polarization opposite to the polarization of the processed voltage to
the integrators input. It is summed up together with the processed voltage. Under the
influence of the reference voltage, the voltage at the integrators output reduces its value. If
the value of that voltage falls below voltage Up of the comparator, a change of the
comparator's condition takes place and, as a consequence, the control circuit will turn off the
reference voltage key again synchronically with the clocking impulse. If the value of the
processed voltage is high, that process may be repeated several times. For that converter to
operate properly, the absolute value of reference voltage should be greater or at least equal
to the absolute value of processed voltage. After the period T0, the reference voltage is
switched on, as in a regular double integral converter.
Because the voltage at the integrators output returns to the initial state after the period of
processing, that proves that the charge supplied from the processed Ux voltage source and
the reference UREF voltage source are of equal value. Therefore, an appropriate equality can
be written:
305
1
R1C
T0
k ti
U x (t)dt = R2C
U REF (t)dt + UREF (t)dt
i=1
(18)
If we assume that R1=R2 and the processed voltage value is constant during integration, the
following is obtained:
T
U x T0 = U REF (t)dt
(19)
where: T total time of reference UREF voltage integration.

The processing method has changed in the aforementioned transducer, but function (19)
describing the relations occurring in it, is the same as function (16) for dual slope converter.
From the point of view of adjusting processing characteristics of A/D double integral
converter, the relation (16) is important, which indicates the process of reference UREF
voltage integration. That means that using a generator generating an appropriate waveform
of reference voltage will enable obtaining a desired A/D converter characteristics, (Janiczek,
1993).
In the transducer circuit presented in figure 3.5, the set of integrator, comparator, time
constant and RC keys remains unchanged. On the other hand, the reference voltage
generator correction voltage, consists of an adder containing an operational amplifier, a
D/A converter and EPROM memory. The meter block along with the control block US and
the generator block fn, form a circuit generating signals controlling keys k1 and k2. In
addition, more significant bits of the meter block are at the same time control signals for
address inputs of the EPROM memory. When the second phase of double integral converter
operation starts the phase of reference voltage integration Uo, through closed key k2,
reference voltage Uo is applied to the integrators input. The voltage is the sum of (with
appropriate coefficients specified by resistors R2, R3 and R4), constant voltage UREF and
voltage generated by C/A transducer, controlled by signals from the EPROM memory.
Signals from the meter, controlling the EPROM memory, in selected time moments t0, t1, ... tn
result in sending relevant codes controlling the C/A transducer from the memory, which
C
Ux
R1
k1
k2
Uo
R4
+
UI
+
K
+
R3
R2
Ug
UK
t US T0
UREF
D/A
EPROM
Counter
Fig. 3.5. Block diagram of transducer with adjusted characteristics.
fn
306
results in generating the assumed reference voltage. A sample waveform of reference

voltage Uo has been presented in figure 3.6.
Stepped reference voltage, integrated in the second phase of double integral converter
operation, gives, as a result, the voltage at the integrators output described by a rangelinear function. The voltage waveform at the integrators output during processing has been
presented in figure 3.7. Time T0 is the time of the processed voltage integration.
Uo
t0
t1
t2
tm
tn
Fig. 3.6. Sample waveform of reference voltage.

UI
t2
t1
T0
tm
tn
Fig. 3.7. Voltage waveform at integrators output of double integral converter.

Taking into account that UREF= Uo in the formulation (16), the following is obtained:
Ux =
1 m-1
T0 i=0
t i+1
ti
tm
U oidt + U omdt
(20)
where: Uoi - reference voltage in the ti ti+1 time interval

t - moment of completion of reference voltage integration phase.
After integration, the relation is obtained:
Ux =
1 m-1
U oi (t i+1 - t i ) + U om (t - t m )
T0 i=0
(21)
Since time slices ti ti+1 are the same throughout the entire reference voltage generating time
and amount to t, and the integration time t of reference voltage ends in moment tn, the
formulation (21) can be presented in the following form:
1 m-1
U oi t + U om t n
T0 i=0
Ux =
307
(22)
The formulation (22) is a record of a range-linear function in which t determines the range
length, Uoi inclination of function in i-th range. Because time slices T0 and t are
determined in the transducer by dividing frequency fn of the clock generator, then, taking
account that :
T0 =
k
fn
t =
p
fn
t=
N
fn
t = m t + t n
(23)
the formulation (22) can be presented in the following form:

m-1
U x k = p U oi + U om (N - m p)
(24)
i=0
where: N value of input signal processing result.

Taking into consideration that the reference voltage is equal:
Ug
U
U o = R4 REF +
R3
R2
(25)
the formulation (24) takes the form:

Ux k =
R4
R4 m-1
U REF N - p +
p U gi + U gm N - m p
R3
R2 i=0
(26)
Because, as it seems from the above-mentioned formulations, the function describing the
output signal on the integrator in the phase of reference voltage integration is a range-linear
function, the measuring transducers characteristics should also be approximated by a
range-linear function:
U(x) =
m -1
i (xi - xi+1 ) + m (x - xm )
(27)
i=0
where: i coefficients of section inclination in i-th range

xi range points
Formulations (24) and (27) lead to believe that to obtain a linear relation:
N = cx
(28)
pUoi = ckix
(29)
the condition must be fulfilled:
where: x range length.

Range-linear function inclination coefficients obtained from the generator of function
containing the C/A transducer, have finite values with discretization corresponding to D/A
308
converters output signal resolution. That fact should be considered when calculating the
function approximating the measuring transducers characteristics. In addition, the number
of values that the approximating function coefficients can have is equal to the number of
values generated by the D/A converter. In addition, due to simplicity of the shaping
circuits structure, the approximation ranges should have the same length, and, if possible,
the number of ranges should be the power of number 2.
The range of changes in coefficient values of approximating sections inclination is
determined by the following relation:
Ug
U
= R4 REF +
R2
R3
(30)
From formulations (17) and (23), it seems that accuracy of double integral converter depends
on the reference voltage stability. Therefore, also in the transducer with adjusted
characteristics, the processing accuracy will depend on voltage generation accuracy Uo, with
the assumption that the approximation error of the measuring transducers characteristics
by a range-linear function is negligibly small.
The formulation (30) shows that maintaining the assumed coefficient accuracy depends on
the stability of resistors R2, R3, R4 and voltage UREF as well as on the accuracy of generating
voltage Ug by D/A converter. The currently available parameters of resistors and sources of
reference voltages are such that their impact on the accuracy of coefficients may be
omitted. Thus, the main source of errors will be D/A converters errors.
The most important D/A converters errors which should be factored in when determining
the generated reference voltage accuracy include: characteristics linearity error,
characteristics inclination change error and zero drift. D/A converter characteristics zero
drift reduction as well as other analog blocks of slotted line can be performed by means of
cyclical zeroing of the slotted line.
The method of adjusting transducer processing characteristics described above is one of the
most effective, especially for converters of high non-linearity. An example may be the
methane concentration meter with a measuring transducer TGS 22611 E00 manufactured by
Figaro USA Inc. (TGS 2611). That is a resistance transducer and it measures methane
concentration ranging from 300 ppm to 10000 ppm. Its characteristics, determining the
relation of output resistance Rs to reference resistance Ro depending on methane
concentration has been presented in figure 3.8. Resistance Ro, defined at concentration 5000
ppm, depending on a specific transducer may vary from 0.68 k to 6.8 k. The transducer
used in the described meter, had 1.2 k. Thus, transducer output resistance changes were
ranging from 0.9 k to 3.6 k.
The currently available integrated A/D converters do not provide the feature of adjusting
their processing characteristics. That is why analog elements that can be found in microcontroller internal structures were used in the structure. The MSP430FG4618 microcontroller manufactured by Texas Instruments was used in the meter for measuring
methane concentration. Its structure contains three operational amplifiers, a 12-bit D/A
converter, resistors and a set of analog switches. The circuit of dual slope converter,
containing those elements, has been presented in figure 3.9.
309
Rs/Ro
Fig. 3.8. Characteristics of transducer for measuring methane concentration.

C
VCC
R1
R2
Vz
Rs
k1
R3
+
OA1
Cp
C/A
Vz
CO
Cz
OA0
k3
TA
+
OA2
k3
Fig. 3.9. Scheme of dual slope converter.

The integrator with amplifier OA0 and comparator OA2 is enclosed by the zeroing circuit
with a capacitor Cz storing offset voltage. Voltage Vz produces artificial mass, in relation to
which the input voltage and the reference voltage received from D/A transducer is
determined. The capacitor Cp provides an appropriate shift of the output signal level from
the D/A converter. Resistors inside the micro-controller have been used as resistor R of the
integrator. The measuring transducer is included in the divider circuit, which simplifies the
measuring circuit. Since voltage VCC, supplying the divider, is, at the same time, the
reference voltage for the D/A converter. Changes in that voltage do not affect the accuracy
of measurement. The transducer control circuit has been based on the structure of module
Timer A and an appropriate program.
The simplified scheme of the meter circuit has been presented in figure 3.10. It is clear, that
almost the entire transducer structure is located inside the micro-controller. Only capacitors
and divider resistors are located outside. The diagram has omitted the heater activation
circuit and the meters keyboard.
For selected concentration values, voltage values on bridge diagonal have been calculated
and for points obtained in such way, the second-degree approximating polynomial U(s) has
been calculated. Its waveform have been specified in figure 3.11.
310
10k
10k
VCC
Rs
MSP430FG4618
P6.4
1,2k
10
P6.0
P10.7
P5.0
P6.6
P5.1
LCD
P6.3
10nF
P6.1
P6.2
P6.5
Fig. 3.10. Simplified scheme of meter for measuring CO concentration.

1.0
0.9
0.8
0.7
U [V]
0.6
0.5
0.4
0.3
0.2
0.1
0
1000 2000 3000 4000 5000 6000 7000 8000 9000 10000
s [pm]
Fig. 3.11. Relation of voltage U on divider output to value of concentration s.

Then, the obtained function U(s) has been approximated by a range-linear function. 10
sections have been obtained. Inclination values of those sections were used to calculate the
values entered into the controlling register of the C/A transducer.
Testing of errors produced by the shaping circuit took place in such way that the transducer
output resistance, for selected concentration values, was simulated by a resistance decade
and the difference between the theoretical concentration value and the value indicated by
the meter was defined. The value of relative error d, calculated with respect to the
maximum range value, has been shown in figure 3.12.
1.0
d [%]
0.5
0
-.5
-1.0
1000 2000 3000 4000 5000 6000 7000 8000 9000 10000
s [pm]
Fig. 3.12. Chart of relative error of carbon monoxide concentration meter.
311
As it stems from the error chart, its maximum value does not exceed 0.6% and is present at
the beginning of the range.
4. Transducer with impulse feedback

Transducer with impulse feedback is also classified into integrating transducers. Its great
advantage is that it is a very simple circuit, but it processes signals of just one polarity,
although that limitation generally is not disturbing in the case of cooperation with
measuring transducers. Especially since it is very well suited for cooperation with
measuring resistance transducers. Due to that simplicity and easy control, they are used in
measuring devices cooperating with different measuring transducers. Block diagram of a
transducer with impulse feedback has been presented in figure 4.1, and voltage waveforms
in selected points of the transducer in figure 4.2.
C
Ux
UI
R1
R2
k
K
+
UK
Up
tp
CB
Counter
T0
Uo
Fig. 4.1. Block diagram of transducer with impulse feedback.

UI
tp
Up
U'
U"
t
Fig. 4.2. Voltage waveforms in selected transducer points.

The transducer circuit consists of an integrator with operational amplifier, capacitor C and
resistors R1, R2, comparator K, control block CB, key k and reference voltage source Uo. The
output signal of frequency f being the function of the processed input voltage Ux is obtained
at the comparators output. If numeric values are to be obtained from the transducer, then
the meter calculates output signal impulses for a given time T0.
The circuit operates in the following manner: processed voltage Ux is supplied to the
integrator, through resistor R1, at all times. As a result, voltage UI at integrators output
continues to increase until it reaches the reference level Up. At that time, the comparator is
activated and, as a consequence, control block CB is activated which, for time tp, closes key k
supplying the integrator with reference voltage with polarization opposite to the
polarization of input voltage Ux through resistor R2. That results in change of voltage
312
waveform at the integrators output. After time tp, key k is turned off and the processing
cycle starts over. In the first phase of transducer operation, voltage increase at the
integrators output can be described by the relation:
U' t =
1
R1 C
T -t p
U x t dt
(31)
In the second phase of transducer operation, in which reference voltage is added to the
integrator, voltage increase at the integrators output can be described by the following
relation:
U" t =
1
1
U x t dt - R 2C Uo t dt
R 1C T-t
T-t
p
(32)
In the steady state, voltage increases at the integrators output are equal:
U' + U" = 0
(33)
Thus, summing up the integrals from formulations (31) and (32) the following is obtained:
tp
1
1
U x t dt =
U o t dt
R 1 C 0
R 2 C 0
(34)
If voltages Ux and Uo are constant, then after transforming the relation (34), a formulation
determining the frequency of transducer output signal is obtained:
f=
Ux R2
1
=
T Uo R1 tp
(35)
and the number of impulses calculated during processing time T0 is:

N = T0 f =
U x R 2 T0
Uo R 1 tp
(36)
Formulation (34) shows that in a one processing cycle, the charge supplied to the integrator
from voltage Ux source is equal to the charge transferred from the integrator under the
influence of reference voltage Uo . Measurement time of processed voltage Ux, namely the
time of impulses counting in the meter is T0 and over the entire period there is balance
between the supplied and the transferred charge. Taking that into consideration, the
formulation (34) can be presented in the following form:
1
R1
T0
U x t dt =
0
1
R2
Nt p
U odt
(37)
Assuming that the processed voltage Ux and reference voltage Uo are constant over
processing period T0, then after transforming, formulation (37) takes the form:
313
Ux
N tp
T0
= Uo
R1
R2
(38)
On the other hand, when the reference voltage U0 is a variable voltage over the processing
period, but it maintains constant value over period tp, figure 4.3, then that condition can be
described by relation (39) arising from formulation (37)
Ux
UI
T0 N t p N-1
=
Uoi
R1
R 2 i=0
tp
(39)
tp
Up
Uoi
Uoi+1
ti
ti+1
Fig. 4.3. Waveforms at integrators output at variable voltage U0.

If voltage Uoi is constant over p long ranges, then for the m-th range the relation (39) may be
expressed in the form:
Ux =
R 1 t p m-1
p U oi + U om N - m p
R 2 T0 i=0
(40)
Formulation (40) corresponds to formulation (30) for double integral converter and, as a
result, shows that the characteristics of a transducer with impulse feedback can be adjusted
in a similar way as it has been presented for a double integral converter. Also, all presented
considerations are also valid here with one reservation concerning time tp.
In the transducer circuit presented in figure 4.1, time tp is generated by a monostable circuit
which does not provide too high accuracy. It may be remedied by introducing modifications
in the transducer circuit, presented in figure 4.4. (Janiczek, 1992)
C
Ux
UI
R1
R2
k
Uo
tp
K
+
UK
Up
f
CB
fn
T0
Counter
Fig. 4.4. Modified version of transducer circuit with impulse feedback.
314
The modification of transducer operation involves generation of time tp it begins not when
the comparator is activated, but synchronically with the first impulse of generator fn, which
comes when the comparator is activated. That is illustrated in figure 4.5.
UI
Up
U2
U1
fg
t
Fig. 4.5. Voltage waveforms in modified transducer circuit.

As a result, time slice tp may be generated by division of frequency fn in the divider. That
results in a certain deviation of transducer output signal frequency, but after averaging over
time interval t0, that does not affect the processing accuracy. Also, time interval T0 may be
obtained in the same manner. Thus, times tp and T0 can be identified as:
T0 =
k
fn
tp =
q
fn
(41)
where q, k frequency fn division coefficients.

Taking into consideration the relations (41) in formulation (40) the following is obtained:
Ux =
R 1 q m-1
p U oi + U om N - m p
R 2 k i=0
(42)
Thanks to such solution, the processing accuracy of a transducer depends only on the
constancy of relation R1/R2 and accuracy of generating voltage Uo. In addition, there is a
possibility to adjust transducer characteristics in the above-mentioned circuit by changing
duration time tp of the feedback impulse.
Assuming, in formulation (37), that both the values of processed voltage Ux and reference
voltage Uo are constant, the following is obtained:
Ux
T0 U o N-1
=
t pi
R 1 R 2 i=0
(43)
After considering the relations (43) and transformation, the following formulation is
obtained:
Ux =
R 1 U o N-1
qi
R 2 k i=0
(44)
from which it can be concluded that the characteristics of a transducer with impulse
feedback can be adjusted by changing the degree of meter division specifying impulse
duration tp.
315
Assuming, as previously, that values qi are constant over p long ranges, the following
relation is obtained:
Ux =
R 1 U o m-1
p qi + qm N - m p
R 2 k i=0
(45)
which shows that in such circuit of a transducer with impulse feedback, measuring
transducers characteristics approximated by a range-linear function may be adjusted in a
very wide range. Moreover, in that transducer, a very high processing resolution and
accuracy can be obtained, which practically depends on the offset voltage constancy of
integrators amplifier, the constancy of R1/R2 resistors relation and the constancy of
reference voltage source Uo.
The above-mentioned method of adjusting processing characteristics of transducer with
impulse feedback has been applied in the meter for measuring carbon monoxide
concentration. A transducer TGS 203 manufactured by Figaro Engineering Inc. (Japan)
whose simplified scheme has been presented in figure 4.6. was used in the meter. (Janiczek,
2009)
The A/D converter circuit consists of an integrator, comparator, keys k1 k5, voltage
divider R2, R3 and R4 and reference resistor R1. Rs is the measuring transducers resistance.
Particular phases of transducer operation are controlled by the micro-controller
manufactured by Texas Instruments MSP430F415. Additionally, to improve transducer
resistance to changes in temperature and supply voltage, the phase of zeroing the
integrators amplifiers and comparator has been introduced. The zeroing voltage is stored
on capacitor Cz . In order for the circuit to be supplied with one voltage, a divider of voltage
R2/R3, R4 providing the reference voltage Uz for the integrator has been introduced. That
voltage is equal approximately to half the supply voltage. As a result, voltages applied to
resistors R1 and Rs, through keys k1 and k2, have opposite polarizations in relation to
voltage Uz. The reference voltage Up for the comparator is obtained from divider R2, R3/R4.
The circuit is zeroed over the Tz time, figure 4.7. Keys k1 and k2, are open and keys k3, k4
and k5 are closed. The integrators and comparators amplifiers are enclosed by the feedback
loop, and capacitor Cz is charged up to the integrators amplifier offset voltage. After
zeroing the circuit, keys k4 are opened and key k1 is closed for the entire duration of
processing and the transducer operates in regular mode of transducer with impulse
feedback.
VCC
R2
Uz
R3
C1
k1
Rs
k3
k2
k4
Cz
R1
UI
A
+
Up
k5
+
K
-
Uk MSP430F415
P1.6
R4
T2
T1
Fig. 4.6. Scheme of carbon monoxide concentration meter.
P1.4
P1.1
P1.0
LCD
CN6292
316
UI
Up
Tz
T1
T2
Fig. 4.7. Voltage waveform at integrators output.

Comparators output impulses are counted in one of the meter's micro-controllers and a
different meter unit generates time slices T2 of feedback impulse duration. Using the basic
relation (38) describing transducer operation and taking account of relation (4.4) describing
the measuring transducers characteristics, the following is obtained:
N=
(VCC - U z ) R1 s0.95
T0
245.66 U z T2 (N)
(46)
where: N - number of impulses counted in the meter,

T0 - processing duration,
Uz - artificial mass voltage.
Figure 4.8 shows the required voltage-frequency transducer characteristics.
N
800
500
400
300
200
100
0.1 0.2 0.3
0.4 0.5 0.6 0.7 0.8 s[%]
Fig. 4.8. Characteristics of voltage-frequency transducer.

For proper charge balance over the T2 period, it has been assumed that resistor R1=150
and Uz=VCC/2, hence the formulation (45) takes the form:
N = 0.61
s0.95
T0
T2 (N)
(47)
In order to obtain a linear relation between the number of counted impulses N and
concentration s, the following condition must be fulfilled:
T2(N) = cs0.95
(48)
The measuring transducers characteristics has been approximated with ten sections, which
ensured approximation error not greater than 0.3%. Using relation (45), coefficient values
determining times T2 have been calculated.
317
Conducted tests have indicated that a stable resolution of so constructed A/D converter was
reached at the level of 14 bits. The impact of temperature changes in the range 0 30o C did
not exceed 0.02%. The main factor of that error was changes of R1 resistor values and change
of R2/R3, R4 dividers pitch. Change of supply voltage VCC within +/-10% was not
noticed. Those parameters significantly exceed the requirements necessary to measure
carbon monoxide concentration but they show the capabilities of the measuring circuit.
d [%]
Conducted meter characteristics tests, in which the measuring transducer was simulated
with a variable resistor, indicated that the relative measuring error d (with regard to the
range endpoint) did not exceed the value 0.3%, figure 4.9.
0.5
0.4
0.3
0.2
0.1
0
-0.1
-0.2
-0.3
-0.4
-0.5
0.1
0.2
0.3
0.4
s [%]
0.5
0.6
0.7
0.8
Fig. 4.9. Error chart of transducer for measuring carbon monoxide concentration.
5. Conclusion
This chapter presents a method for digital adjustment of analog-to-digital converter transfer
function. If the measuring transducers characteristics is non-linear, it is necessary to use
suitable correction methods. When the measuring transducers characteristics is very nonlinear, for example in the case of certain transducers for light intensity measurement, a
problem with selecting a relevant A/D transducer may occur.
A converter with digital adjustments is a useful tool for correcting non-linearity of
transducers, especially those with high linearity, in case of which the numerical correction
would lead to the lost of resolution. The conducted experiments proved the suitability of the
proposed new solution of converter. Such designed analog-to-digital converter achieves a
stable resolution and small error of the approximation function.
6. References
Bucci G., Marco Faccio M., Landi C. New ADC with Piecewise Linear Characteristic: Case
StudyImplementation of a Smart Humidity Sensor. IEEE Trans. Instr. Measur.,
Vol. 49, No. 6, Dec. 2000.
Grzybowski W. Wagner F. Nonlinear function from d-a converters. Electronic Engineering,
nr 512. 1971.
Iglesias G. E., Iglesias E. A., Linearization of Transducer Signals Using an Analog-to-Digital
Converter. IEEE Trans. Instr. Measur., Vol. 37, NO. 1, Mar. 1988.
318
Janiczek J: Le convertisseur analogique-numerique pour la correction des caracteristiques

statiques nonlineaires des capteurs. Measurement Science and Technology. 1992
vol. 3
Janiczek J. Analogue-to-digital converter with digitally controlled transfer function.
Measurement Science and Technology. 1993 vol. 4
Janiczek J. Digital adjustment of analog-to-digital converter transfer function. Metrology
and Measurement Systems, vol.16, no 3, 2009.
Patranabis D., Ghosh S., Bakshi C. Linearizing Transducer Characteristics. IEEE Trans. Instr.
Measur., Vol. 37, No. 1, Mar. 1988.
Patranabis D., Ghosh D. A Novel Software-Based Transducer Linearizer. IEEE Trans. Instr.
Measur., Vol. 38, No. 6, Oct. 1989.
Reis G. Nichtlinearen A/D-Umzetzer linearisiert Sensor-Kennlinien. Elektronik No 4, 1980.
TGS 2611 Product Information. Figaro USA Inc.
15
Planar Microwave Sensors for Complex
Permittivity Characterization of Materials
and Their Applications
2COMSATS
1University
Kashif Saeed1, Muhammad F. Shafique2,

Matthew B. Byrne1 and Ian C. Hunter1
of Leeds, School of Electronic and Electrical Engineering,

Institute of Information Technology, Department of Electrical Engineering,
1United Kingdom
2Pakistan
1. Introduction
Accurate measurement of material properties has gained considerable importance over the
last decade. The ability to non-destructively monitor specific properties of a material
undergoing physical or chemical changes has led to many applications in industry,
medicine and pharmaceuticals.
In the food industry, the interest in dielectric properties of agricultural and food materials
has been principally for predicting heating rates when the materials are subjected to high
frequency electric fields (Oliveira, 2002). Microwave volumetric heating for food
preservation and processing applications has been popular since the early 1970s and
products such as rubber, wood, paper and many other agricultural products have been
studied extensively (Nelson, 1991). An important application of microwaves in the food
industry has been the non-invasive determination of moisture content (Kraszewski, 1998)
and its effect on the dielectric properties of materials such as granular solids, meats,
vegetables and fruits (Mudgett, 1995). Microwave techniques for drying of food products
have also been very popular (Schubert, 2005).
The use of microwaves in therapeutic medicine has increased dramatically in the last few
years (Golio, 2003). The ability to discriminate between normal and malignant cancerous
tissues using microwaves is well reported and has led to the development of many noninvasive techniques for early detection of cancer and harmful tumors in the human body.
Moreover, the ability of microwaves for deep tissue heating has resulted in the development
of many novel microwave therapeutic treatments. An important and a growing application
is the treatment of cardiac arrhytmias using microwave ablation (Rosenbaum, 1993,
Greenspon, 2000). Furthermore, novel techniques such as microwave aidedliposuction
(Rosen, 2000) and the RF and microwave enhancement of drug absorption are currently
being investigated.
In the field of chemistry, dielectric measurements serve as a fundamental tool for the
characterisation and evaluation of solvent materials, and have proven to be very convenient
320
for investigating the molecular dynamics of materials. The dielectric analysis of

pharmaceutical materials (Benoit, 1996) has increased in importance in recent years and is
vital to the understanding and application of microwaves to synthetic organic chemistry
(Tierney, 2005). Microwave dielectric heating has been very popular since the early 1980s
due to the many benefits it offers over conventional heating (Galema, 1997, Gabriel, 1998).
Solvents with higher loss factors lead to higher heating rates than can be achieved
conventionally. Moreover, the interaction of microwaves with metal powders is used
extensively to create hot-spots which serve as catalysts to accelerate chemical reactions of
metals with gases, other inorganic solids, and organic substrates. The reduced reaction times
also assist in the synthesis of radiopharmaceuticals (Galema, 1997) and inorganic complexes,
and have had a significant impact in drug discovery (Larhed, 2001, Wathey, 2002).
Microwaves are also employed during many organic reactions which require dry reaction
conditions. In ceramic processing, microwave heating reduces cracking and thermal stress
(Michael, 1991), allows quick curing of polymers, and is more economical than conventional
heating methods.
In RF and microwave circuit design the dielectric permittivity of substrate plays an
important role and requires precise evaluation over a broad range of frequencies.
Knowledge of these properties plays a crucial role in the accurate design of multi-layered
circuits, Monolithic Microwave Integrated Circuits (MMICs) and Low Temperature Co-fired
Ceramic (LTCC) circuits (Edwards, 1991). The following sections aim to provide a general
discussion of some well established dielectric measurement techniques, primarily applicable
to moderate and high loss liquid samples.
2. Dielectric measurement techniques

The measurement methods relevant for any desired application depends on the nature of
the dielectric material to be measured, both physically and electrically, the frequency of
interest, and the degree of accuracy required. Being able to design an appropriate holder for
the sample and to obtain an adequate model of the circuit for reliable calculations of the
permittivity from electrical measurements often proves to be an important challenge.
At low and medium frequency ranges, bridge and resonant circuits have often been used for
characterising dielectric materials. At higher frequencies however, transmission line,
resonant cavity, and free-space methods are commonly used and have been illustrated in an
early review (Altschuler, 1973). In general, dielectric measurement techniques can be
categorised as reflection or transmission type, using resonant or non-resonant systems, with
open or closed structures.
2.1 Non-resonant methods
In non-resonant methods, the properties of the materials are fundamentally deduced from
their impedance and wave velocities therein. When an electromagnetic wave propagates
from one material to another, both the characteristic wave impedance and the wave velocity
change, resulting in a partial reflection of the wave from the interface between the two
materials. Measurements of the reflections from such an interface, and the transmission
through it, can provide information for the deduction of permittivity and permeability
relationships between the two materials. Non-resonant methods mainly include reflection

Permittivity Characterization of Materials and Their Applications
321
and reflection/transmission methods. In reflection methods, the properties of the material

are deduced from the magnitude and phase measurement of the reflected signals. In
reflection/transmission measurements the properties of the material are deduced from
measuring the magnitude and phase of both reflected and transmitted signals. To improve
confidence in findings it is generally preferred to use results from both reflection and
transmission measurements. Non-resonant methods require a means of directing the
electromagnetic energy towards a material, and then collecting what is reflected and
transmitted through it. For this purpose, any type of transmission line could be used; for
instance a coaxial line, a hollow metallic waveguide, a planar transmission line etc.
2.1.1 Waveguide and coaxial transmission line techniques
The use of waveguide and coaxial transmission line cells for complex permittivity
measurements was first reported by Nicholson-Ross (Nicolson, 1970) and then Weir (Wier,
1974) who analysed the structure in the time and frequency domains respectively. Following
this, Baker-Jarvis (Baker-Jarvis, 1990) produced iterative methods of solutions. The analysis
of Nicholson, Ross and Weir led to the development of explicit formulae for the calculation
of permittivity and permeability which is commonly referred to as the NRW algorithm. The
technique essentially involves filling a section of a transmission line of a certain length with
the sample to be measured (figure 1). The change in the propagation constant and the
characteristic impedance Zo leads to partial reflections of the wave at the interfaces. The
propagation constant is related to the attenuation coefficient and propagation coefficient
through the relation j . For a dielectric material the propagation coefficient of a
wave within a transmission line is related to the complex permittivity of the filling material
through the relation,
2 r r
c2
(1)
Where is the angular frequency, r is the permeability of the material which is equal to 1, c
is the speed of light and, c is the cutoff wavelength of the transmission line. In case of a
coaxial transmission line supporting TEM propagation the cutoff wavelength is taken to be
equal to infinity. r r' j r'' is the complex permittivity of the material filling the line
where, r' is the dielectric constant and r'' is the dielectric loss of the medium. From
equation 1 it becomes possible to extract the complex permittivity of the material filling the
transmission line.
2
2
c
2
2
r'

r''
c 2 2
(2)
(3)
The complex propagation coefficient within the section of the line filled with the material is
obtained from the S-parameter measurements with a network analyser. Reference planes are
322
Fig. 1. (a) Waveguide fixture and (b) Coaxial fixture for transmission measurements.
established using well known calibration methods such as the Thru-Reflect-Line (TRL) or
the Short-Open-Load-Thru (SOLT) techniques (Bryant, 1993). Owing to the simplicity of this
technique, it has been widely adopted and many improvements have been made over the
years which allow the permittivity of high and moderately lossy samples to be evaluated
with a good degree of accuracy.
This technique is applicable to almost any kind of transmission line but waveguides and
coaxial lines are generally preferred at frequencies below 30 GHz. Waveguide transmission
cells can be constructed from either rectangular or circular waveguides and the mode of
transmission depends on the structure. At frequencies below 2.45 GHz, the use of
waveguides is not well suited due to the large volume of the test sample required. Coaxial
transmission lines are preferable under these circumstances as they are comparatively small
in size and cover a broader bandwidth. It has to be ensured that higher order TE and TM
modes do not propagate as they lead to errors in the measured permittivity. For this
purpose, standard cell dimensions are used. These are 3.5 mm (0 - 34.5 GHz), 7 mm (0 - 18.2

323
GHz), and 14 mm (0 - 8.6 GHz). Transmission line cells also suffer from problems with air
gaps when the size of the cell becomes too small. The presence of air gaps in the cell leads to
significant measurement errors. Over the past 30 years there have been several
improvements and variations in the design of transmission line cells each being well suited
to a particular class of dielectrics. At frequencies above 30 GHz, the dimensions of cells
become too small and one has to resort to using free-space measurement techniques.
2.1.2 Free space transmission techniques
Free-space techniques are also grouped under non-destructive and contactless measuring
methods, and are generally employed at higher frequencies (above 10 GHz) (Varadan, 1991).
They do not require any special sample preparation, and are particularly suitable for
measuring materials at high temperatures and for inhomogeneous dielectrics. They have
also been implemented in many industrial applications for continuous monitoring and
control. In a typical free-space transmission measurement technique, a sample is placed
between a transmitting and a receiving antenna, and the attenuation and phase shift of the
signal are measured, from which the dielectric properties of the sample can be determined.
Accurate measurement of the permittivity over a wide range of frequencies can be achieved
by free-space techniques.
In most systems, the accuracy of the determined permittivity depends mainly on the
measuring system and the validity of the model used for calculations. The usual
assumptions made with this technique are that a uniform plane wave is normally incident
on the flat surface of a homogeneous material, and that the planar sample extends to infinity
laterally so that diffraction effects at the edges of the sample can be neglected. Figure 2
shows a typical arrangement of a free-space measurement setup. Multiple reflections,
mismatches, and diffraction effects at the edges of the sample are generally considered to be
the main sources of errors and have to be accounted for appropriately. To enhance the
measurement accuracy, special attention must be paid to the choice of the radiating
Fig. 2. Typical arrangement of a free space dielectric measurement setup.
324
elements, the design of the sample holder, and the sample geometry and location between
the radiating elements.
2.1.3 Open ended transmission line techniques
Open ended transmission line techniques provide a very convenient and a non-invasive
fixture for the evaluation of dielectric permittivity of liquids and semi-solids with no sample
preparation. This technique was pioneered by Stuchly and Stuchly (Athey, 1982), who used
it to measure the dielectric properties of biological materials.
Over the years, this technique has been improved. In this method, the material to be tested is
placed against the cut-off section of the transmission line, and the magnitude and phase of
the reflected signal are measured (figure 3). Over the past decade or so several models have
been used to relate the permittivity of the material to the reflection coefficient measured at
the aperture of the probe. Both waveguide and coaxial transmission lines have been used,
although coaxial lines are preferable since they are very broadband and have small
dimensions. This technique is well suited to measuring high dielectric constant and high
loss samples, and is ideal for characterising lossy solvent materials. It requires calibration
reference planes to be obtained at the probe aperture, which are challenging in the case of a
coaxial probe. Well known reference materials are often required for calibration.
Fig. 3. Open ended coaxial and waveguide transmission lines.

2.1.4 Dielectric waveguide techniques
Dielectric waveguide techniques (Abbas, 2001) have recently been proposed which allow the
determination of the dielectric permittivity of low loss planar sheet materials such as, teflon
or perspex. The sample to be measured is placed in direct contact between two dielectric
waveguides which can be of circular or rectangular geometry (figure 4). The sample can be

325
of arbitrary shape, but its transverse dimensions must be greater than, or in the limiting case
equal to, the transverse dimensions of the dielectric waveguide. The system is based on two
basic properties of dielectric waveguides. The first property is that the energy of the wave
propagating within the waveguide is entirely concentrated inside the waveguide. The other
property is that the phase velocity is equal to the phase velocity of a plane wave in an
c
. Suitable expressions which allow determination of
unbound dielectric medium
the complex permittivity from reflection and transmission measurements can be found in
(Abbas, 2001). In order to launch energy into a dielectric waveguide suitable transitions are
needed. These are generally made from metallic horn antennas which can be difficult to
design.
Fig. 4. Dielectric waveguide fixture for the complex permittivity measurement of low
dielectric constant materials.
2.1.5 Planar transmission line techniques
Planar transmission lines such as microstrip and coplanar waveguides have long been used as
microwave components. They allow ease of fabrication, low cost of manufacture, and
compactness, which makes them suitable for industrial applications which use dielectric
permittivity measurements. In planar transmission line methods, the material to be tested
usually serves as a superstrate or as a substrate, or as a part of either. In the case of solids, the
dielectric sample can serve both as the substrate or the superstrate. However, in the case of
liquids and semi-solids, it is easier to have the sample as a superstrate. There have been many
investigations into the use of planar circuits for complex permittivity measurements of liquids
(Stuchly, 1998, Raj, 2001, Facer, 2001, Queffelec, 1994, Hinojosa, 2001, Wadell, 1991, Chen, 2004)
Figures 5 and 6 show typical planar cells for dielectric permittivity measurements of liquid
materials. The liquid to be measured completely covers the planar circuit and is enclosed
inside a low loss container which is fixed firmly or epoxied on top of the board. The enclosure
introduces mismatches which can easily be calibrated. The introduction of the liquid dielectric
changes the effective permittivity eff and the characteristic impedance Zo of the line. The
dielectric properties are then extracted from the change in effective permittivity using suitable
expressions which can be found in literature (Wadell, 1991).
326
Fig. 5. Microstrip cell for complex permittivity measurement of liquids.
Fig. 6. Coplanar waveguide cell for complex permittivity measurement of liquids.

327
2.2 Resonant methods

Resonant methods offer the potential of characterising the properties of a material at a single
frequency or a discrete set of frequencies with high accuracy in comparison to broadband
methods. They can be classified into resonator methods and resonant perturbation methods.
Resonator methods are those in which the material to be measured serves as a resonator and are
only applicable to extremely low loss samples. The different types of resonators used for this
purpose can be further classified into dielectric, coaxial surface-wave, and split resonators.
Details on these methods can be found in (Chen, 2004). Resonant perturbation methods are those
in which the sample is introduced into a resonant structure causing a perturbation in the
response. The perturbation results in a shift of the resonant frequency and a decrease in the
unloaded quality factor of the resonator from which the dielectric properties can be evaluated.
The resonant perturbation technique is well suited for low and moderate loss samples. Both
reflection and transmission type resonators can be employed for this purpose.
2.2.1 Waveguide cavity resonators
Waveguide cavity resonators are frequently employed for resonant perturbation
measurements because of their high quality factors. The resonant cavities are designed in
the standard TM (transverse magnetic) or TE (transverse electric) modes of propagation of
the electromagnetic fields. The choice of cavity used depends on the particular field
distribution of interest. The material to be tested is inserted at a particular location within
the cavity where the electric field is at a maximum. Often a rod shaped sample is used, and
in the case of liquid and semi-solid materials an appropriate cylindrical enclosure is needed
to hold the sample. The insertion of the sample within the cavity causes perturbation to the
system which results in a shift in the resonant frequency and a decrease in the unloaded
quality factors. This perturbation in the response of the cavity is related to the material
properties through cavity perturbation theory which is the most common technique used
owing to its simplicity and accuracy. The cavity must be designed for a particular frequency
of interest and at lower frequencies the cavity resonators are often large.
Figure 7 shows a rectangular TE101 and a cylindrical waveguide TM010 cavity operating in
the two most widely used modes of propagation for permittivity measurements. Also
shown in figure 7 is the magnitude of the electric field distributions for each mode.
Fig. 7. Waveguide cavity resonators for complex permittivity measurements.
328
2.2.2 Coaxial cavity resonators

A coaxial cavity shown in figure 8 can also be used for complex permittivity measurements
of both moderate and high loss samples. This technique has been demonstrated by
Raveendranath and colleagues (Raveendranath, 2000), and can also be analysed using
perturbation theory. It offers high quality factors and thus can be used for measuring lossy
samples. The coaxial cavity resonator is essentially a straight piece of coaxial transmission
line which is coupled to another coaxial line from one end with the other end either shorted
or left open. The cavity exhibits resonance at the fundamental with only one electric field
maximum and multiple higher order resonances which could also be used for
measurements at discrete frequencies. A rod shaped material is inserted within the cavity
through a slot cut into the outer conductor, which can slide across the length of the cavity.
Since coaxial lines are broadband in nature, resonant measurements can be carried out over
a very broad bandwidth.
Fig. 8. Coaxial cavity resonators for complex permittivity measurements.

2.2.3 Open ended resonant lines
A very important but not so widely used resonant method of complex permittivity
measurement is the resonant open ended line method first demonstrated by Johnson and
colleagues (Johnson, 1992). The sensor consists of a long coaxial line with the input feed line
located a distance L1 = /4 away from the shorted end and L2 = /2 away from the open
end section (figure 9). At resonance, the fields are concentrated at the open end aperture of
the resonator where the material to be tested is brought into contact. This method is ideally
suited for measuring high loss samples and does not require any sample preparation. The
sensor is calibrated in air before bringing it in contact with the material under test. The effect
of the material in contact with the aperture of the sensor detunes the response.
The response is then tuned back to the original frequency by adjusting length L1. This
change in the length is directly proportional to the permittivity of the material. The
deterioration in the quality factor of the resonator after tuning provides information on the
loss of the sample.

329
Fig. 9. Open-ended resonant line for complex permittivity measurements.

2.2.4 Planar transmission line resonators
Almost any kind of planar topology can be used for permittivity measurements.
Unfortunately, all planar methods suffer from low quality factors (less than 500) and as a
result measurements on low to moderate lossy samples have been common. Just as in the
case of broadband planar transmission line methods in planar resonator perturbation
measurements, the sample under study forms either a substrate or as a superstrate or as a
part of either. The perturbation to the system is governed by the properties of the sample
and the extent of interaction with the electric fields. Figure 10 shows a planar straight ribbon
microstrip resonator fixture proposed by Abdulnour and colleagues (Abdulnour, 1995). This
structure is well suited for resonant perturbation measurements on liquid samples. The
sample is inserted in a small hole drilled inside the substrate. Because of the higher
concentration of the electric fields at the tips of the resonator the sample can be located near
the tips. The electric fields interact with the dielectric causing a capacitive perturbation in
the response. The advantage of using such a technique is that only a small amount of sample
(less than a few microlitres) is required for characterisation. This is highly beneficial, for
instance, in the pharmaceutical industry, where accurate characterisation of materials in
small quantities is vital. Another well known microstrip resonant measurement fixture
applicable to moderately lossy samples was proposed by Boosanovich [Bogosanovich, 2000].
The sensor is essentially a circular disk resonator which is coupled to a coaxial transmission
line that extends through the substrate from the bottom. The measurement fixture is shown
in figure 11. The sample to be measured serves as a superstrate and the dielectric properties
are measured from the shift in resonance and a deterioration in the quality factor of the
response. The sample size required for measurements is generally large at low frequencies
as the material has to completely cover the patch. Closed forms expression allow the
permittivity to be calculated without requiring any reference calibration samples.
330
Fig. 10. Microstrip fixture for resonant perturbation measurements on liquid samples.
Fig. 11. Patch resonator sensor for complex permittivity measurements.

331
3. Microstrip resonators
Planar transmission lines have been used extensively in the past to measure electromagnetic
properties of materials including thin films, sheet samples and substrate samples. They can
offer many attractive features which include compactness, ease of fabrication and, use in a
disposable manner. The three most common types of planar transmission lines used widely for
materials measurements are stripline, coplanar waveguide and microstrip. Similar to their
coaxial and waveguide counterparts, planar circuit methods can also be characterised into
broadband and resonant methods. Resonant methods offer higher sensitivity and accuracy
and therefore have been the focus of attention. Planar resonator methods are generally
applicable to low loss samples such as thin films and substrates. In resonant perturbation
methods part of the substrate is replaced by the sample under study. The interaction of the
electric field lines with the material under test leads to a perturbation of the system thereby
changing its resonant frequency and the unloaded quality factor. Since our aim is primarily to
characterise lossy solvent materials using planar technology we will therefore focus our
attention on resonant perturbation methods. The use of microstrip resonators for measuring
material properties has been investigated by many researchers (Bernard, 1991, Abdulnour,
1995, Bogosanovich, 2002, Fratticcioli, 2004). These methods usually include the material under
test (MUT) as a part of the substrate or as an overlay. The complex permittivity of the MUT is
related to the resonant frequency fr and the unloaded-Q factor Qu, through closed form
expressions which have limitations. The techniques can be destructive or non-destructive and
in the past have been limited to measuring materials with low or moderate permittivities. In
this section we demonstrate how a microstrip perturbation method could be used to
characterise common polar and non-polar lab solvents (Saeed, 2007).
3.1 Microstrip transmission line
A microstrip transmission line consists of a strip conductor and a ground plane separated by
a dielectric substrate as shown in figure 12(a). w and t are the width and the thickness of the
strip conductor respectively. h and d are the thickness and the dielectric constant of the
substrate. Since the dielectric constant of the substrate is much higher than that of air the
field is concentrated near the substrate. The field distribution on a microstrip line is shown
in figure 12(b). The effective permittivity eff and the characteristic impedance Zo of a
microstrip line can be determined using the well known closed form expressions found in
the literature (Wadell, 1991).
Microstrip resonators such as the ring resonator (figure 12(c)) and the open resonator also
sometimes referred to as the straight ribbon resonator (figure 12(d)) form an integral part of
many microwave circuits such as oscillators and filters. The ring resonator is a basic
resonant structure which offers high unloaded Q factors due to reduced radiative losses. For
rings constructed on substrates which have a much higher thickness than that of the strip
conductor, we have, at resonance,
ng
2
(4)
where n is an integer order of resonance and g is the wavelength of the guided structure. To
take into account the dispersive nature of a microstrip line the above equation can be
rewritten as,
332
nc
2 f eff
(5)
where c is the speed of light, f is the frequency and eff is the effective permittivity of the
structure. Similarly, the open loop resonator is resonant when its electrical length is half that
of the wavelength in the guided structure. The open resonator does however, suffer from
increased radiative losses due to the presence of the fringe fields at the ends which result in
the inaccurate determination of the resonant frequency. In order to take into account the
fringing fields, the length of the resonator is generally taken to be slightly longer.
Fig. 12. (a) Geometry of a microstrip; (b) Electric field distribution at the cross section of a
microstrip line; (c) Ring resonator; (d) Open loop resonator.

333
3.2 Microstrip ring resonator for complex permittivity measurements

The earliest use of ring resonators for dispersion measurements was reported by Troughton
(Troughton, 1969). Later on it was realised that this simple structure could also be used to
characterise the dielectric properties of thin sheet materials. Bernard and his colleagues
(Bernard, 1991) used a microstrip ring resonator in a multilayer substrate topology to
measure the dielectric constant of thin wafers. This technique was suitable for measuring
low dielectric constant and low loss wafers.
The ring resonator could also be used to measure the dielectric properties of solvents. When
the material under test is brought in close vicinity to the resonator. The fringing fields
interact with the dielectric and energy is coupled into it causing a shift in the resonant
frequency fo. This shift depends on the relative permittivity r of the solvent and is
independent of the dielectric loss r. The relationship between fo and r can be modelled
using a 2nd order polynomial fo = a+b1r +b2r2 which provides a suitable fit. The
coefficients a, b1 and b2 can be found by using standard solvents such as methanol, ethanol
and 1-propanol of which the dielectric properties are well known. Reducing the
perturbation even further gives a linear response and only two calibration media are now
required (air and water for which the dielectric permittivity is well known). For
demonstration purposes we will use a one port coupled ring resonator as shown in figure 13
(Saeed, 2007). The resonator is modelled as an equivalent RLC network where R is the
resistance, L is the inductance and C is the capacitance. Figure 14(a) is the simulated
response of a square ring resonator designed on an RT/Duroid 5880 substrate with a
dielectric constant of 2.2 and a loss tangent of 0.0009. The length of the resonator is 208 mm
with the fundamental resonance at 1 GHz. Higher order modes occur at integer multiples of
the fundamental frequency at 2 GHz, 3 GHz and so on. Enhanced coupling method with
extended input transmission line was used with a coupling gap of 0.4 mm. This is a critically
coupled device with a 3dB bandwidth of approximately 10 MHz and a return loss of up to
60 dB. The use of critical coupling allows more accurate measurement of the resonant
frequency. The response can be tuned by selectively adding/removing copper.
Fig. 13. Ring resonator electrically coupled to a microstrip transmission line through a gap.
334
Fig. 14. (a) Simulated return loss performance of ring resonator; (b) Current distribution on
ring resonator for first three resonant modes.
Also shown in figure 14(b) is the current distribution on the ring for the first three
resonances. It can be seen that the coupling is electric in nature and that several electric field
maximums are located along the length of the resonator. For lossy materials only a small
portion of the ring can be covered by the material under test. The size and the location of the
material governs the perturbation to the system. Liquid samples which can have high
dielectric constant and/or high loss can be contained inside a small low loss capillary made
of quartz or borosilicate glass and then introduced on the ring. Fixing the capillary on the
microstrip in this manner ensures the same amount of perturbation to the system each time
the measurement is made. A good location on the ring where the capillary can be located is
shown in figure 14(b). Here the electric field maximum is common in all modes.
3.2.1 Location of capillary containing material under test
The capillary containing the material under test can be located either on top of the copper
track or inside the substrate. The sensitivity depends on the extent of field penetration inside
the material. Figure 15 illustrates two possible locations of the capillary for material
measurements. It can be seen that the field lines are more concentrated within the substrate
then in air. Therefore it should be obvious that by locating the material inside the substrate
(below the copper track) results in higher sensitivity. To do this we can drill a small hole
inside the substrate and insert the capillary through it. However, this might not be easily
achievable in case of thin substrates. Figure 16 shows the change in the transmission
response of a microstrip line when a capillary of 0.25 mm inner radius and 0.3 mm outer
radius containing the material under test is placed on top of copper track and inside the
substrate. The difference in the transmission responses is shown in figure 17. We can clearly
see that higher sensitivity is achieved when the sample is located within the substrate.

335
(a)
(b)
Fig. 15. (a) Capillary located on top of a microstrip line and (b) inside the substrate.
(a)
(b)
Fig. 16. Transmission response of a microstrip line at 1 GHz with a bororsilicate glass
capillary containing material located (a) on top of microstrip line and (b) located inside the
substrate. Capillary made of bororsilicate glass with outer radius of 0.3 mm and inner radius
of 0.25 mm. RT/Duroid 5880 substrate with r = 2.2 and tan() = 0.0009 with thickness
h = 0.787 mm. Copper thickness t = 0.035 mm and width of microstrip line w = 2.5 mm.
336
Fig. 17. Difference in the transmission response of a microstrip line when the capillary is
located on top of copper track and inside substrate.
3.2.2 Measurements
The solvents used as standards and the unknowns are listed in Table 1 along with literature
(Gregory, 2001) and measured values of dielectric permittivity. All samples obtained were of
ANALAR standard and have purity greater than 99.5%. The measurements were carried out at
1 GHz, 2 GHz and at 3 GHz. All measurements were performed using an Agilent 8510C
network analyzer at room temperature. The analyzer was used to average the data 256 times
and the measurements repeated 4 times. Once the dependence of resonant frequency on
dielectric permittivity was modelled and fitted using a 2nd order polynomial, the permittivity
of the unknown solvents was determined. Four standards namely air, ethanol, methanol and
dimethylsulphoxide were used to characterize 1-propanol and ethanediol. Air, 1-propanol and
ethanol were then used as standards to characterize xylene, trichloroethylene, butyl acetate,
Table 1. Measured dielectric permittivity of solvents.

337
chlorobenzene and ethyl acetate. Figure 18 is the measured and plotted results at 1 GHz, 2
GHz and 3 GHz respectively. The graphs clearly indicate that the resonant frequency is a
good fit to the reciprocal of the square root of the dielectric permittivity.
(a)
(b)
(c)
(d)
(e)
(f)
Fig. 18. Resonant frequency fo versus dielectric permittivity r for measurement of 1propanol and ethanediol at (a) 997 MHz (c) 2 GHz (e) 3GHz. Resonant frequency fo versus
dielectric permittivity r for measurement of xylene, trichloroethylene, butyl acetate,
chlorobenzene and ethyl acetate at (b) 997 MHz (d) 2 GHz and (f) 3GHz.
338
3.3 Microstrip open loop resonator

To demonstrate the use of a microstrip resonator for solution sensing application we use an
open loop resonator to characterise glucose-water solutions (Saeed, 2007). The resonator was
designed at 1 GHz using RT/Duroid 5880. The length of the resonator is 111.98 mm with a
width of 2.4 mm. The resonator is magnetically coupled to a transmission line. The electric
field is concentrated at the two tips of the resonator as shown in Figure 19(a). Perturbation
can be introduced to the system by placing the capillary across the tips and can be
controlled by perturbing a single arm or both.
To demonstrate the high sensitivity of the resonators and the ability to measure small
quantities of dissolved impurity, glucose-water solutions have been detected. This could be
done either by noticing the change in the resonant frequency of the device or the unloadedQ or both. The concentration of glucose-water solutions was varied from 0% to 4% (w/v) i.e.
0 g/100 ml to 4 g/100 ml. Figure 19(b) shows the variation in the unloaded-Q and the
resonant frequency with varying glucose concentration. It can be seen that as the glucose
concentration increases the resonant frequency increases whereas the unloaded-Q decreases.
This indicates that the dielectric permittivity decreases and the dielectric loss increases with
the addition of glucose. This could be anticipated from dielectric mixing theory (Steeman,
1990). Another conclusion that can be drawn from the graph is that it is simple to relate the
concentration to the resonant frequencies and the unloaded-Q as the response is
approximately linear. The device shows a decrease of 0.348 in the unloaded-Q factor and an
increase of 20 kHz in the resonant frequency for every 1% increase in the concentration of
glucose; making this a very sensitive method of detecting impurity concentration.
Nevertheless, the sensitivity can be further increased by increasing the perturbation. From
the linear relationship the sensitivity f/(g/100ml) of the device is 25.313kHz/(g/100ml)
and in terms of the unloaded-Q Qu/(g/100ml) is 0.348/(g/100ml).
Fig. 19. (a) Current distribution on an open loop resonator at 1 GHz; (b) Resonant frequency
fo and unloaded-Q Qu for varying concentration levels of glucose solutions.
4. Substrate integrated waveguide cavity resonators

Traditionally, waveguide cavity resonators have been used to characterise materials through
resonant perturbation methods. They offer high quality factors, accuracy, and sensitivity

339
and the volume of the material required for characterisation can be as small as a few
microlitres. However, waveguide resonators can often be bulky, and in some cases difficult
to construct which greatly limits their use. In such circumstances, one has to resort to using
planar resonant devices which are generally compact, less expensive and easier to construct.
Substrate integrated waveguide resonators are a relatively new class of planar resonant
devices which offer higher quality factors in comparison to other planar devices. They are
compact and can be constructed easily on a substrate along with the transitions. The devices
can be made extremely compact and low loss by fabricating them on a high dielectric
constant and low loss substrate. This makes them a suitable candidate for many industrial
applications which require measurement of dielectric properties. The compactness of these
devices can allow measurements to be made on sample volumes as small as a few nanolitres
with better sensitivity. Their high quality factors can allow measurements to be made on a
broad range of materials.
The substrate-integrated-waveguide is an artificial waveguide structure that is fabricated on
a planar substrate with periodic metallised via-hole cylinders or metallised grooves which
serve as sidewalls and solid top and bottom walls of metallization (figure 20). Since their
first introduction (Hirokawa, 1998, Deslandes, 2001) a large number of microwave circuits
based on this technology have been constructed to demonstrate the validity of the concept in
easy integration with planar devices and its significant size reduction. The SIW resonant
cavity is an important component in microwave circuits, e.g. filters and oscillators etc., and
is constructed in the same way as a conventional waveguide cavity but with smaller
dimensions. Due to finite losses in the substrate the quality factor of the cavity is reduced
but is still significantly higher than its planar counterparts. If the integrated waveguide is to
Fig. 20. Construction of a substrate integrated waveguide from metallised grooves and vias
on a substrate.
340
be constructed from metallised grooves (figure 20(c)) the structure can be treated as a
rectangular waveguide filled with a dielectric and the dimensions can be calculated easily
using equations valid for a conventional rectangular waveguide. If however, the walls of the
integrated waveguide are constructed using metallised vias (figure 20(b)) the exact
calculations for the dimensions become rather more complex. The exact cut-off frequency of
a particular mode in the integrated waveguide then becomes dependent upon the diameter
of the vias and the spacing between them. In the past many authors have made use of
modelling techniques such as the finite element method and the method of moments to
compute the appropriate dimensions of the integrated waveguide and simulate the
structures (Deslandes, 2001, Hill, 2001). Cassivi and colleagues (Cassivi, 2002) report on the
dispersion characteristics of a substrate integrated rectangular waveguide for the first time
and provide approximate equations to estimate the cut-off frequency of the first and second
propagating modes. In this section we report on how the substrate integrated waveguide
cavity resonator could be used to characterize solvents in small volumes (Saeed, 2008). The
sensor is compact and highly sensitive which makes it an ideal candidate to be used for
dielectric measurements in the pharmaceutical industry.
4.1 Design of SIW cavity
The dimensions of the TE101 SIW resonant cavity constructed from complete metallised
2
. For an empty rectangular waveguide
walls/grooves can be estimated from o
r' a 2 l 2
cavity resonating at 8 GHz the typical dimensions for the width a, length c and height b
would be 23 mm, 32 mm and 10 mm respectively. However, the corresponding substrate
integrated waveguide cavity constructed using complete metalised walls and fabricated
using RT Duroid 5880 laminate with a thickness of 2.8 mm would have a width of 15 mm
and a length of 23 mm. This results in a great reduction in the overall volume of the cavity.
Moreover, the electric field is concentrated within a small portion of the cavity. The
unloaded Q factor for an empty waveguide cavity is approximately 3500 whereas for a
substrate integrated waveguide cavity it is reduced to approximately 700-800 owing
primarily due to the losses in the substrate. However, the use of low loss ceramics or
polymer substrates will significantly improve the Q. For demonstration purposes an X-band
SIW cavity resonator was designed to measure the complex permittivity of solvents. The
cavity was a one-port reflection type cavity and the energy was coupled into it by means of
a microstrip line [48]. The extent of coupling could be controlled through the use of an offset
which makes it possible to tune into the resistive losses of the cavity and achieve return
losses in excess of 50 dB. Figure 21(a) is a layout of the cavity resonator. The cavity was
formed by complete metallisation of the top, bottom and sidewalls of the substrate. At the
intersection of the microstrip and the cavity the metallisation extends towards the microstrip
so as to form a complete cavity.
4.2 Location of material under test and sensitivity analysis
As discussed in previous sections for the fundamental TE101 resonant mode the electric field
is dominant at the center of the cavity and a common practice is to locate the sample at this
position for maximum sensitivity. The sample can be inserted either through the broad
walls of the cavity or through the side walls (figure 21(b) positions A and B). In case of

(a)
341
(b)
Fig. 21. (a) Layout of substrate integrated waveguide cavity resonator with microstrip feed
section and hole within the cavity through which the material under test (MUT) is inserted;
(b) Location of capillary containing material under test inside the cavity.
liquid dielectrics the material is contained inside a capillary made of low loss material such
as quartz or borosilicate glass and then inserted into the cavity. We choose to insert the
capillary through the sidewalls of the cavity through a small hole drilled into the substrate.
This gives good mechanical stability and repeatability as the width a of the cavity is much
larger than the height b. Care must be taken to ensure that no air bubbles are created in the
portion of the capillary inside the cavity. The capillary does not need to be filled entirely as
only the portion of material inside the cavity interacts with the fields. Therefore, the volume
of sample required for measurements depends on the inner diameter of the glass capillary
and the width a of the cavity.
The perturbation to the system is governed primarily by the location and size of the
capillary. As an example we show simulation results for an X-band substrate integrated
cavity resonator (8 GHz) made from an RT/duroid substrate of dielectric constant of 2.2,
loss tangent of 0.0009 and a thickness of 2.8 mm with a width of 15 mm and a length of 23
mm. The simulation results were obtained using HFSS. Figures 22 and 23 show the
f f rs
dFs
and the sensitivity
to a lossless
fractional change in the resonant frequency Fs ro
f rs
d r'
dielectric material with varying dielectric constant located within a borosilicate glass
capillary (r=3.4, tan()=0.0015) of various inner diameters (outer diameter = 0.7 mm). fro is
the resonant frequency of the cavity without the sample and frs is the resonant frequency of
the cavity with the sample. From perturbation theory it is anticipated that the fractional
change in resonant frequency increases with increasing dielectric constant of material (figure
21). For capillaries with small inner diameters this increment is linear. As the inner diameter
of the capillary increases the linearity is no longer valid due to higher perturbation to the
system. For even larger inner diameters of the capillary the fractional change in resonant
frequency increases very significantly with increasing dielectric constant. As would be
expected the sensitivity is higher for low dielectric constant materials for any given diameter
of the capillary (figure 22).
342
Fig. 22. Fractional change in resonant frequency with varying dielectric constant of an ideal
lossless material for different inner diameters of capillary (Outer diameter of capillary = 0.7 mm).
Fig. 23. Sensitivity versus the dielectric constant of material under test with varying inner
diameter of capillary (Outer diameter of capillary = 0.7 mm).

343
For a capillary with an inner diameter of 0.05 mm the volume of material required is 29 nl.
The shift in resonant frequency is 0.86 MHz for a dielectric constant change of 60 which is
significantly high for samples in nanolitre volumes. This change can be measured easily
with a network analyzer. In our case the volume required was less than 0.47 l as we chose
to use a capillary which had an inner diameter of 0.2 mm. The graph given in figure 22 also
provides good insight into the range of solvents that can be characterized using a specific
size of capillary. It can be concluded that for a capillary of an inner diameter of 0.07 mm this
technique is most suitable for measuring solvents with a moderate dielectric constant i.e. 1
r 20 as the sensitivity is high. This technique is also suitable for measuring small dielectric
differences in a range of solvents which can be of high dielectric constant. This is achieved
by careful adjustment of coupling. The cavity containing the reference material is tuned by
varying the coupling offset until the load is critically coupled to the input feed section. The
solutions are then compared to the reference material by studying the relative change in the
response.
4.3 Perturbation of higher order modes
In order to study the dispersive dielectric properties of solvents over a desired bandwidth it
is also possible to observe the perturbation of higher order modes in the resonant cavity.
However, this requires the ability to vary the location of the material which can be achieved
by drilling holes at various locations along the length of the cavity. To illustrate this we
show simulation results on an integrated transmission type cavity formed using the same
method discussed in the previous sections but with a longer length of cavity (l = 95 mm).
The cavity is loosely coupled to the input and output microstrip feed sections. Figure 24(a)
and 24(b) are the reflection and transmission responses of the cavity. The graphs show the
perturbation of five resonant modes observed when a cylindrical sample of a dielectric
constant r = 20 is located at the centre of the cavity inserted through the sidewalls. Also
shown is the magnitude of the electric field distribution at each resonance. It can be clearly
seen that several electric field maximum occur along the length of the cavity. The extent of
perturbation of each resonance can be controlled by varying the location of the capillary.
4.4 Fabricated resonant cavity
To study the dielectric properties of solvents an X-band substrate integrated waveguide cavity
resonator operating at 8 GHz has been fabricated with complete metallised walls on an
RT/Duroid 5880 substrate with a thickness of 2.8 mm. Metallisation of the sidewalls of the
cavity was achieved by uniformly depositing a thin layer of silver epoxy. At the intersection of
the microstrip and the cavity the metallization extends towards the microstrip by 6.5 mm from
either side so as to form a complete cavity. The width and the length of the cavity were
optimized using HFSS. The optimized width of the cavity a was 15 mm and the length l was 23
mm. A microstrip to substrate integrated waveguide transition was used and the reflection
type cavity was coupled to a microstrip line having a width of 1 mm through an offset length
of 3.1 mm. The offset length for critical coupling was optimized to a borosilicate glass capillary
located within the substrate through the side walls of cavity (Figure 21(b) position B with
capillary along the x-axis). The capillary had an outer diameter of 0.5 mm and an inner
diameter of 0.2 mm (0.47 l volume of material) respectively. The optimised response showed
a resonance close to 8 GHz and the electric field distribution in the cavity at resonance is
shown in figure 25(a). The fabricated resonator is shown in figure 25(b) and 25(c).
344
Fig. 24. Simulated (a) Return loss performance and (b) transmission response of integrated
waveguide resonant cavity of 95 mm length with sample of dielectric constant r= 20
located at the center of the cavity.
Fig. 25. (a) Magnitude of electric field distribution on the substrate integrated waveguide
resonator for the fundamental TE101 resonant mode; (b) Top view of fabricated resonator; (c)
Side view of fabricated resonator; (d) Measured return loss performance of various solvents.

345
4.5 Measurements
To illustrate the ability of the compact resonator to discriminate between solvents the
capillary was filled with various solvents and their response was measured using an Agilent
PNA E8361A at room temperature. The data presented was the result of averaging 256 times
and the measurements were repeated 4 times. Figure 25(d) is the measured response for
various materials clearly indicating its ability to discriminate among solvents. The resonant
frequency shifts and the Q factor changes as various materials are inserted in the capillary.
The measured unloaded Q for the empty capillary was approximately 700 and the resonant
frequency was close to 8 GHz. As stated in previous discussion the resonator is highly
sensitive to moderate dielectric constant materials. Therefore, this technique can be used to
characterize and study many alcohol solutions. For demonstration purposes we choose to
measure the dielectric properties of isobutanol-isopropanol mixtures. Mixtures with varying
volume fractions of dissolved isopropanol were carefully prepared and their response was
measured using an Agilent PNA E8361A at room temperature. The shift in the resonant
frequency was linearly related to the change in the dielectric constant (figure 26(a)) whereas
the change in the 3 dB bandwidth was linearly related to the change in the loss tangent of
the mixture (figure 26(b)). The measured return loss performance of the mixtures is shown
in figure 26(c). The dielectric properties of the mixtures were measured and compared to
those obtained using a TE101 transmission type rectangular waveguide cavity and an open
ended coaxial probe. A comparison is presented in table 2. From figure 26(a) it can be seen
that the resonant frequency decreases as the volume fraction of dissolved isopropanol
increases. This is expected as the dielectric constant of the mixture increases with increasing
concentration of isopropanol since isopropanol has a higher dielectric constant than that of
isobutanol. There is also an increase in the 3 dB bandwidth with increasing volume fraction
of dissolved isopropanol due to an increase in loss tangent (figure 26(b)).
Fig. 26. (a) Measured resonant frequency versus dielectric constant of mixture for increasing
volume fraction of dissolved isopropanol (%). (b) Measured 3-dB bandwidth versus loss
tangent of mixture for increasing volume fraction of dissolved isopropanol (%). (c) Returnloss performance of isobutanol and isopropanol mixture.
The dielectric data measured using the integrated waveguide technique is in excellent
agreement to that obtained using a rectangular waveguide resonant cavity method. For a
change in dielectric constant of 0.5 there is a shift of almost 4.4 MHz in the resonant
frequency indicating a high sensitivity. Moreover, a change of 2.9 MHz in the 3 dB
bandwidth is observed for a change in loss tangent of 0.1. The error in the measured results
for both resonant cavity and the integrated waveguide cavity is within 0.5 %.
346
Table 2. Comparison of measured complex permittivity of isobutanol-isopropanol mixtures

using a substrate integrated waveguide cavity resonator method, a rectangular waveguide
cavity resonator method and an open ended coaxial probe method.
5. Conclusion
In conclusion, we have presented a thorough introduction into the various resonant and
non-resonant methods for the characterization of complex permittivity using a range of
various transmission line topologies. We have highlighted the advantages and
disadvantages of the various techniques and have presented a thorough comparison.
Discussion has been focused primarily on the use of planar resonant perturbation methods
for dielectric characterization. Two topologies namely the microstrip resonator technique
and the substrate integrated waveguide technique have been highlighted as potential
candidates for use as compact sensitive sensors for use within the pharmaceutical industry.
The discussion has been supported through careful design of sensors and measurements on
polar and non-polar common lab solvents for each of the sensors. It has been concluded that
the substrate integrated waveguide sensors are highly compact offering high unloaded-Q
factors and are capable of performing accurate and sensitive measurements on solvents in
volumes of less than a few nano-litres making them ideally for use within the
pharmaceutical industry.
6. References
Abbas, Z., Pollard, R., & Kelsall, R. (2001). Complex permittivity measurements at ka band
using rectangular dielectric waveguide. IEEE Transactions on Instrumentation and
Measurement, Vol. 50, No. 5, (October 2001), pp. 1334-1342, ISSN 0018-9456
Abdulnour, J., Akyel, C., & Wu, K. (1985). A generic approach for permittivity measurement
of dielectric materials using discontinuity in a rectangular waveguide or a
microstrip line. IEEE Transactions on Microwave Theory and Techniques, Vol. 43, No. 5,
(May 1985), pp. 1060-1066, ISSN 0018-9480

347
Altschuler, H. (1963). Dielectric constant, In: Handbook of Microwave Measurements, Vol. 2, M.

Sucher & J. Fox, (Eds.), pp. 530-536, Polytechnic Press, New York (NY), ISBN B0011007T-C
Athey, T., Stuchly, M., & Stuchly, S. (1982). Measurement of radio frequency permittivity of
biological tissues with an open ended coaxial line: Part I. IEEE Transactions on
Microwave Theory and Techniques, Vol. 30, No. 1, (January 1982), pp. 82-86, ISSN
0018-9480
Baker-Jarvis, J., Vanzura, E., & Kissick, W. (1990). Improved technique for determining
complex permittivity with the transmission/reflection method. IEEE Transactions
on Microwave Theory and Techniques, Vol. 38, No. 8, (August 1990) pp. 1096-1103,
ISSN 0018-9480
Benoit, E., Prot, O., Maincent, P., & Bessiere, J. (1996). Applicability of dielectric
measurements in the field of pharmaceutical formulation. Bioelectrochemistry and
Bioenergetics, Vol. 40, No. 2, (August 1996), pp. 175-179, ISBN 0302-4598
Bernard, P., & Gautray, J. (1991). Measurement of dielectric constant using a microstrip ring
resonator. IEEE Transactions on Microwave Theory and Techniques, Vol. 39, No. 3,
(March 1991), pp. 592-595, ISSN 0018-9480
Bogosanovich, M. (2000). Microstrip patch sensor for measurement of the permittivity of
homogeneous dielectric materials. IEEE Transactions on Instrumentation and
Measurement, Vol. 49, No. 5, (October 2000), pp. 1144-1148, ISSN 0018-9456
Bryant, G. (1993). Principles of microwave measurements, Peregrinus on behalf of Institution of
Electrical Engineers, ISBN 0-83641296-3, London (UK)
Cassivi, Y., Perregrini, L., Arcioni, P., Bressan, M., Wu, K., & Conciauro, G. (2002).
Dispersion characteristics of substrate integrated rectangular waveguide. IEEE
Microwave and Wireless Components Letters, Vol. 12, No. 9, (September 2002), pp. 333335, ISSN 1531-1309
Chen, L., Ong, C., Neo, C., Varadan V. V, & Varadan, V. K. (2004). Microwave electronics:
Measurement and materials characterization, John Wiley and Sons, ISBN 0-47084492-2,
Chichester (UK)
Deslandes, D., & Wu, K. (2001). Integrated microstrip and rectangular waveguide in planar
form. IEEE Microwave and Wireless Components Letters, Vol. 11, No. 2, (February
2001), pp. 68-70, ISSN 1531-1309
Deslandes, D., & Wu, K. (2001). Integrated transition of coplanar to rectangular waveguides,
Proceedings of IEEE MTT-S International Microwave Symposium (Vol. 2), ISBN 078036538-0, Phoenix (AZ), May 2001
Edwards, T., & Steer, M. (1991). Foundations of interconnect and microstrip circuit design (3rd
ed.), John Wiley and Sons, ISBN 0-47160701-0, New York (NY)
Facer, G., Notterman, D., & Sohn, L. (2001). Dielectric spectroscopy for bioanalysis: From 40
Hz to 26.5 GHz in a microfabricated waveguide. Applied Physics Letters, Vol. 78, No.
7, (February 2001), pp. 996-998, ISSN 1077-3118
Fratticcioli, E., Dionigi, M., & Sorrentino, R. (2004). A simple low cost measurement system
for the complex permittivity characterisation of materials. IEEE Transactions on
Instrumentation and Measurement, Vol. 53, No. 4, (August 2004), pp. 1071-1077, ISSN
0018-9456
348
Gabriel, C., Gabriel, S., Grant, E., Halstead, B., & Mingos, D. (1998). Dielectric parameters
relevant to microwave dielectric heating. Chemical Society Reviews, Vol. 27, No. 3,
(May 1998), pp. 213-223, ISSN 0306-0012
Galema, S. (1997). Microwave chemistry. Chemical Society Reviews, Vol. 26, No. 3, (June 1997),
pp. 233-238, ISSN 0306-0012
Golio, M. (2003). Microwave and RF product applications, CRC Press, ISBN 0-84931732-0,
London (UK)
Greenspon, A. (2000). Advances in Catheter ablation for the treatment of cardiac arrhytmias,
Proceedings of IEEE MTT-S International Microwave Symposium, ISBN 0-7803-6435-X,
Boston (MA), June 2000
Hill, M., Ziolkowski, R., & Papapolymerou, J. (2001). A high-Q reconfigurable planar EBG
cavity resonator. IEEE Microwave and Wireless Components Letters, Vol. 11, No. 6,
(June 2001), pp. 255-257, ISSN 1531-1309
Hinojosa, J. (2001). S parameter broadband measurements on microstrip and fast extraction
of the substrate intrinsic properties. IEEE Microwave and Wireless Component Letters,
Vol. 11, No. 7, (July 2001), pp. 305-307, ISSN 1531-1309
Hirokawa, J., & Ando, M. (1998). Single-layer feed waveguide consisting of posts for plane
TEM wave excitation in parallel plates. IEEE Transactions on Antennas and
Propagation, Vol. 46, No. 5, (May 1998), pp. 625-630, ISSN 0018-926X
Johnson, R., Green, J., Robinson, M., Preece, A., & Clarke, R. (1992). Resonant open ended
coaxial line sensor for measuring complex permittivity. IEE Proceedings-A, Vol. 139,
No. 5, (September 1992), pp. 261-264, ISSN 0960-7641
Kraszewski, A., Trabelsi, S., & Nelson, S. (1997). Moisture content determination in grain by
measuring microwave parameters. Measurements in Science & Technology, Vol. 8, No.
8, (August 1997), pp. 857-863, ISSN 0957-0233
Larhed, M., & Hallberg, A. (2001). Microwave-assisted high-speed chemistry: a new
technique in drug discovery. Drug Discovery Today, Vol. 6, No. 8, (April 2001), pp.
406-416, ISSN 1359-6446
Michael, D., Mingos, P., & Baghurst, D. (1991). Applications of microwave dielectric heating
effects to synthetic problems in chemistry. Chemical Society Reviews, Vol. 20, No. 1,
(March 1991), pp. 1-47, ISSN 0306-0012
Mudgett, R. (1995). Electrical properties of foods, In: Engineering properties of foods, (2nd ed.),
M. A. Rao & S. S. H. Rizvi, (Eds.), pp. 389-455, Marcel Dekker, New York (NY),
ISBN 0-82475328-3
Nelson, S. (1991). Dielectric properties of agricultural products: measurements and
applications. IEEE Transactions of Electrical Insulation, Vol. 26, No. 5, (October 1991),
pp. 845-869, ISSN 0018-9367
Nicolson, A., & Ross, G. (1970). Measurement of the intrinsic properties of materials by time
domain techniques. IEEE Transactions on Instrumentation and Measurement, Vol. 19,
No. 4, (November 1970), pp. 377-382, ISSN 0018-9456
Oliveira, M., & Franca, A. (2002). Microwave heating of foodstuffs. Journal of Food
engineering, Vol. 53, No. 4, (August 2002), pp. 347-359, ISSN 0260-8774
Queffelec, P., Gelin, P., Gieraltowski, J., & Loaec, J. (1994). A microstrip device for the
broadband simultaneous measurement of complex permeability and permittivity.

349
IEEE Transactions on Magnetics, Vol. 30, No. 2, (March 1994), pp. 224-231, ISSN 00189464
Raj, A., Holmes, W., & Judah, S. (2001). Wide bandwidth measurement of complex
permittivity of liquids using coplanar lines. IEEE Transactions on Instrumentation and
Measurement, Vol. 50, No. 4, (August 2001), pp. 905-909, ISSN 0018-9456
Raveendranath, U., Bijukumar, S., & Matthew, K. (2000). Broadband coaxial cavity resonator
for complex permittivity measurements of liquids. IEEE Transactions on
Instrumentation and Measurement, Vol. 49, No. 6, (December 2000), pp. 1305-1312,
ISSN 0018-9456
Rosen, A., Rosen, D., Tuma, G., & Bucky, L. (2000). RF/Microwave aided tumescent
liposuction. IEEE Transactions on Microwave Theory and Techniques, Vol. 48, No. 11,
(November 2000), pp. 1879-1884, ISSN 0018-9480
Rosenbaum, R., Greenspon, A., Hsu, S., Walinsky, P., & Rosen, A. (1993). RF and
microwave ablation for the treatment of ventricular tachycardia, Proceedings of
IEEE MTT-S International Microwave Symposium, ISBN 0-7803-1352-6, Atlanta
(GA), June 1993
Saeed, K., Guyette, A., Pollard, R., & Hunter, I. (2007). Microstrip resonator technique for
measuring dielectric permittivity of liquid solvents and for solution sensing,
Proceedings of IEEE MTT-S International Microwave Symposium, ISBN 1-42440688-9,
Honolulu (HI), June 2007
Saeed, K., Pollard, R. D., Hunter, I. C. (2008). Substrate integrated waveguide cavity
resonators for the complex permittivity characterisation of material, IEEE
Transactions on Microwave Theory and Techniques, Vol. 56, No. 10, (October 2008), pp.
2340-2347, ISSN 0018-9480
Schubert, H., & Regier, M. (2005). The microwave processing of foods, Woodhead, ISBN 185573964-X, Cambridge (UK)
Stuchly, S. & Bassey, C. (1998). Microwave coplanar sensors for dielectric measurements.
Measurement Science & Technology, Vol. 9, No. 8, (August 1998), pp. 1324-1329, ISSN
0957-0233
Tierney, J., & Lidstrom, P. (2005). Microwave assisted organic synthesis, Blackwell, ISBN 140511560-2, Oxford (UK)
Troughton, P. (1969). Measurement techniques in microstrip. Electronics Letters, Vol. 5, No. 2,
(January 1969), pp. 25-26, ISSN 0013-5194
Varadan, V. V., Hollinger, R., Ghodgaonkar, D., & Varadan, V. K. (1991). Free-space,
broadband measurements of high-temperature, complex dielectric properties at
microwave frequencies. IEEE Transactions on Instrumentation and Measurement, Vol.
40, No. 5, (October 1991), pp. 842-846, ISSN 0018-9456
Wadell, B. (1991). Microstrip line structures, In: Transmission line handbook, B. C. Wadell
(Ed.), pp. 93-110, Artech House, ISBN 0-89006436-9, London (UK)
Wadell, B. (1991). Transmission line design handbook, Artech House, ISBN 0-9006436-9, London
(UK)
Wathey, B., Tierney, J., Lidstrom, P., & Westman, J. (2002). The impact of microwaveassisted organic chemistry on drug discovery. Drug Discovery Today, Vol. 7, No. 6,
(March 2002), pp. 373-380, ISSN 1359-6446
350
Weir, W. (1974). Automatic measurement of complex dielectric constant and permeability at

microwave frequencies. Proceedings of the IEEE, Vol. 62, No. 1, pp. 33-36, (January
1974), ISSN 0018-9219
16
Basics on Radar Cross Section Reduction
Measurements of Simple and Complex Targets
Using Microwave Absorbers
Marcelo A. S. Miacci1 and Mirabel C. Rezende2
2Institute
1National Institute for Space Research (INPE),

of Aeronautics and Space, Department of Aerospace Science and Technology,
Brazil
1. Introduction
The research of Radar Cross Section (RCS) of simple and complex objects is decisively
important to identify targets such as aircraft, missiles, rockets, ships and other objects, with
the purpose of improving or rendering difficult their radar visibility in various frequency
ranges. The use of RCS measurements of targets have expanded to more than solely military
applications in the identification and control processes of defense systems (Burgess, 1988).
Higher the RCS value is, easier it becomes to detect and identify an object. However, when
these targets present different geometrical forms and different types of electromagnetic
radiation absorber materials (ERAM) on their surfaces, they can become stealthy and
practically invisible to radars at determined frequency ranges.
In order to make target identification more precise, it is indispensable to analyze and
understand the RCS patterns generated by the targets. These patterns represent the
reflection mechanisms in the interaction process of the wave with the target, i.e., the
interaction of the wave with the aspects of the targets geometry and the materials physicalchemical characteristics of its surface (Dybdal, 1987).
RCS measurements aim to determine the equivalent effective area of the target when it is
impinged by a radar wave. In other words, it is the ratio between the electromagnetic
energy irradiated by radar over a target and the energy scattered by it. The scattering
measurements can be performed in monostatic condition, whereby the electromagnetic
waves reflected by the target are measured in the same direction as the emitting source
(radar), or in bistatic condition, when the reflected waves are detected in other directions.
The present work describes experimental studies of RCS measurements of targets with
simple and complex shapes and the RCS reduction by using ERAM in the microwave
frequency range of 5 12 GHz (C and X bands). This chapter intends also to give some
highlights about the theory involved and some measurement topics concerning target
reflectivity, calibration techniques, enhancement methods and some experimental results
achieved in this research performed at Materials Division of the Institute of Aeronautics and
Space from Brazil.
352
2. Main definitions
When an electromagnetic wave focuses on an object the energy is spread in all directions.
The spatial distribution of energy depends on the target geometry, material composition, the
operating frequency and polarization of the incident wave. This distribution of energy is
called scattering and the material is often called target (Blake, 1986).
Based on this principle, we can define the radar cross section, or RCS, as a measure of the power
that returns to source or reflects in a given direction, normalized in relation to an incident wave
power density. The purpose of normalization is to remove the effect of distance and facilitate
the description of the cross section independent of the distance between the target and radar
(R). The RCS is defined as shown in equation 1 (Bhattacharyya & Sengupta, 1991).
4 lim
4 lim
(1)
where:
: radar cross section of the target (m2);
: reflected or scattered electric field (V/m);
: reflected or scattered magnetic field (A/m);
: incident electric field (V/m);
: incident magnetic field (A/m).
The scattered electric and magnetic fields are due to the presence of a target, so, the total
field is the sum of the incident and the scattered fields (equation 2):
(2)
The RCS unit is usually given in square meters, or expressed in dB, relative to one square
meter (dBm2 or dBsm) as in equation 3 (Currie, 1989).
10 log
(3)
These concepts can be applied in the radar equation (Skolnik, 1990), correlating the received
power in terms of transmitted power, scattering, distance and antennas gain (equation 4).
(4)
where:
: radar received power (W);
: radar transmitted power (W);
R: distance between radar and target (m);
: radar cross section of the target (m2);
G: radar antenna gain (dimensionless);
: wavelength (m).
The received power by the antenna from the transmitted radar pulse is directly related with
the physical characteristics of the target through the backscattering coefficient. The value of
this backscattering coefficient is basically dependent on the following factors (Jenn, 1995):
Basics on Radar Cross Section Reduction Measurements

of Simple and Complex Targets Using Microwave Absorbers
353
the target geometry and surface roughness;

the target material composition (electrical and/or magnetic properties);
the wavelength and polarization of the radar; and
the aspect angle related with the incident wave.
The conceptual definition of RCS includes the fact only one part of the radiated energy
reaches the target. The RCS of a target () can be most easily visualized as the product of
three factors:
= Projected cross section x reflectivity x directivity
The reflectivity is defined as the intercepted radiated (scattered) power by the target. The
directivity is given by the ratio of the backscattered power into the radars direction to the
power that would have been backscattered, considering an uniform scattering in all
directions (isotropic).
2.1 Scattering matrix
In order to determine the dependence of the RCS in relation to the polarization of the wave,
we must consider the relationship between the transmitted (t) and the received (r) fields, in
terms of the horizontal and vertical components of linear polarization of the wave, EH and
EV, respectively, where the index H denotes horizontal polarization and V the vertical
polarization. Then, EH and EV can be expressed in terms of the reflectivity of a target
illuminated by both polarizations through the proportionality constant aij, where i denotes
the polarization of the transmitter and j the polarization of the receiver (equation 5) (Crispin,
1968) or in matrix notation (equation 6):
(5)
(6)
The constants aij shall be considered independent of the distance, but in complex notation
due to its phase relationship between the electric field components. For a monostatic radar
configuration, aHV and aVH are equal. Therefore, the scattering matrix defines the relationship
between amplitude and phase components of the electric fields transmitted and received.
A similar matrix to that one mentioned for linear polarization of the wave can be obtained
for elliptical and circular polarizations, using the constants of proportionality describing the
left or right hand polarization (Crispin, 1968).
2.2 Frequency regions
When a target has very small physical dimensions compared to the wavelength of the
incident wave, it is considered that your analysis is being done in the so called Rayleigh
region (Blake, 1986). In this region, the shape of the object does not influence in determining
their RCS, and for some types of objects the RCS is determined from the volume, instead of
considering the dimensions and physical forms. For targets that are comparable in size to
354
the wavelength of incident wave, the RCS varies depending on the frequency and is called
the resonant region or Mie region.
When the dimensions of the target are large compared with the wavelength of incident
wave, the RCS can be determined using the methodology of geometrical optics or by the
method of optical physics, and this region is called the optical region. In the following
sections, the analysis of some RCS targets of simple and complex geometry is shown in the
optical region. Figure 1 is a RCS curve of a sphere as a function of the ratio of the target
radius (a)/ wavelength (), normalized in the optical region. The analysis of this figure
shows that there are the distinct regions discussed above. Although it is a perfectly
conducting sphere, this behavior is observed for all types of targets (Currie, 1989;
Bhattacharyya & Sengupta, 1991).
Fig. 1. Normalized RCS of a sphere as a function of ka (where k=2/).
3. Radar cross section reduction measurements

RCS measurements are performed for many reasons; the main one is to verify target
detection limits by a radar system to which it is subjected, verifying conformity of the
practical models, designed by theoretical equations (Kouyoumjian, 1965).
The measurement as a whole is a complex task due to the many factors that may affect these
measurements. Instrumental errors, spurious interferences and reflections are some of
contributions to degrade the quality of the data. These problems are compounded when one
is interested in reducing the RCS of an object, as in this case, the magnitude of these effects
can overshadow the true RCS values.
Many methods have been proposed and used for the measurement of several types of
targets. Depending on the size of the targets and radar frequencies used, RCS measurements
can be performed on outdoor or indoor ranges, in the last condition, inside anechoic
chambers. In measurements, it is important that the radar is illuminated by an
electromagnetic wave which is uniform in phase and amplitude. For practical purposes, the
maximum tolerance for amplitude variation over a target is 0.5 dB, and the phase should not
deviate more than 22.5o. These conditions denote the far-field condition (Balanis, 1982; Jasik
& Johnson, 1984), given by:
2
(7)

355
where:
R is the distance between radar and target;
d is the largest dimension of the target; and
is the wavelength of the radar.
This condition ensures good measurements. Errors produced by the instrumentation should
neither exceed 0.5 dB nor vary in time in order to avoid instability in the measurement of the
RCS patterns. Thus, it is decisive the careful selection of the experimental parameters. In this
case, the dynamic range of the system should be at least 40 dB when measuring targets with
small RCS. Dynamic range values in the order of 60 dB or higher are preferable when RCS
reduction studies are conducted or when absorber materials are used. Another important
factor to be taken into account is the target support structure that should not interfere with
the incident wave, however in practice such condition isnt always possible.
Analysis of the electromagnetic energy scattering by metallic objects is very important in the
understanding of RCS of targets, in the same way that reflections of dielectric and magnetic
surfaces are important when studying RCS reduction. Within this context, RCS analysis of
targets with simple shapes is fundamental to support the understanding of RCS patterns of
targets with complex shapes.
To evaluate the electromagnetic behavior of targets, many methods have been proposed and
experimentally used for many decades with well accepted results (Birtcher & Balanis, 1994)
and the RCS measurement errors depend on the nature of the target under test, the distance
at which this target is being measured and the place of measurements.
Systems are projected in such way that respect the parameters described above. However,
they present technical challenges, making adequate the use of experimental procedures that
involve new techniques that can minimize errors. When an anechoic chamber is used, the
emissions and reflections of spurious radiation are controlled and minimized by using
commercial off the shelf ERAM inside the chamber, becoming the background noise levels
of this radiation almost null. On the other hand, by using the outdoor range, the
measurements are affected by environment variations, therefore needing greater control of
the parameters.
Even in measurements performed in indoor range, the RCS patterns of the targets may
become impaired due to the occurrence of noise in the system or on account of low
backward radiation contribution of the targets with lower RCS.
3.1 Basic instrumentation
The basic instrumentation required for RCS measurements consists of four subsystems that
can be controlled by a central station; these subsystems are: positioners and drivers,
receiver, transmitter and data acquisition system.
Using a simple setup and placing the transmitter (TX) and receiver (RX) antennas in the
scheme showed in Figure 2, it is possible to measure the RCS of targets in different frequency
ranges. The distance between both antennas needs to be tested to eliminate the radiation
coupling between them.
In order to have good precision, the transmitter must have means to provide enough power
to allow a good signal/noise ratio for the measurement system and it is necessary to take
356
Fig. 2. RCS measurement setup (R is the distance between the antennas and the target).
care with the alignment between the transmission/reception antennas and the target. The
use of a laser beam helps in the alignment of the system improving the precision of the
measurements. The support column for the target, called as pylon, needs to be recovered
with ERAM in order to avoid any possible contribution of reflected waves that prejudices
the target characterization.
Figure 3 shows a system used in this study that basically consists of two antennas, one
transmitter and one receiver mounted on a tower keeping a certain distance from each other.
The target is mounted on a dielectric support and this one is mounted on a positioner which
will allow recording the RCS patterns. The transmission is made through a microwave
generator that feeds, by a low loss coaxial cable, the antenna input terminals. A spectrum
analyzer can be used as a receiver that collects the signals from the receiving antenna. The
system has a positioner controller (driver) that controls the rotational speed, azimuth angle
and limits. Some equipment is in a control room, where the user monitors the tests through
a computer with a GPIB interface.
Figure 3 shows the system composed by:
a.
b.
c.
d.
e.
Target under test (square flat plate with 0.2 m side);

Pyramidal microwave absorbers;
Horn antennas to the 8.2 12.4 GHz frequency range;
Low loss coaxial cables; and
Microwave generator.
Fig. 3. Anechoic chamber view assembled with the RCS measurement system (courtesy of
IFI/DCTA, Brazil).

357
Figure 4 shows the internal view of the control room (shelter), which is integrated with the
instrumentation located inside the anechoic chamber. This figure shows: (a) antenna
positioner controllers; (b) spectrum analyzer (receiver) and (c) computer.
Fig. 4. Internal view of the control room (courtesy of IFI/DCTA, Brazil).

3.2 Amplitude calibration methods
There are two main methods for calibration of systems for RCS measurements: direct and
indirect calibrations. The direct calibration involves the use of reference targets with well
known values of RCS, such as spheres, flat plates, dihedrals and dielectric lens. The RCS of
these targets can be accurately inferred by its physical sizes, to provide a radar return signal
that can be compared to the signal of a target under test with unknown value of RCS. In
some cases, the reference target may be in a different position relative to the target under
test and a modified form of the radar equation is used to estimate the difference, as seen in
equation 8:
(8)
where:
: RCS of the target under test (m2);
Pr: received power ratio between the reference target and target under test (dimensionless);
Ru: distance between target under test and radar (m);
Rr: distance between reference target and radar (m);
Fu: propagation factor of the target under test (dimensionless);
Fr: propagation factor of the reference target (dimensionless); and
: RCS of the reference target (m2).
From equation 8, the propagation factors Fr and Fu contain information about the conditions
of propagation between the radar and the two targets, including any atmospheric
attenuation, scattering or even any undesired multiple reflection, as ground reflection that
may be present. If the two factors are the same during measurements, equation 8 is reduced
to equation 9:
(9)
358
The accuracy of this calibration method depends on the accuracy of the RCS reference target
and the assumption that there are no other targets or significant sources of reflection on the
measuring environment and the propagation factor is well known.
The second method of amplitude calibration is known as the indirect calibration method or
closure. This method involves precise measurements of the characteristics of the radar
system or the transmission and reception system by measuring the received power of a
target under test with unknown RCS, and the calculation of RCS is done by using the radar
range equation (equation 10) below:
4
(10)
where:
: RCS of the unknown target (m2);
PR: received power from the unknown target (W);
LT: transmitter losses (dimensionless);
LR: receiver losses (dimensionless);
Ru: distance between radar to unknown target (m);
PT: transmitted power (W);
G: radar antennas gain (dimensionless);
: wavelength (m), and
Fu: propagation factor (dimensionless).
For this method is considered accurate, the parameters of the radar should be carefully
measured, the received power must be precisely determined and the propagation factor
should be equal to unity. The errors associated with this technique can occur in the
measurements of radar parameters such as received power and also in conditions where the
propagation factor is not equal to unity. One advantage of the indirect technique is that the
reference target is not required.
3.3 Enhancement measurement techniques (active cancellation)
In order to reduce some error sources present in RCS measurements such as coupling
between antennas, reflections in the anechoic chamber, reflection of target support and poor
dynamic range, it can be used an active noise cancellation circuit which employs a phase
cancelling technique.
A result of works performed at Materials Division/IAE/DCTA in Brazil shows an
experimental set-up developed to reduce the error sources presented in indoor RCS
measurements in C-band (5.8 to 6.4 GHz). By means of this research it was possible to
perform comparisons, with and without the developed system, and observe its influence in
measurements of simple objects.
By proposing the implementation of the system, it is expected to allow detection of small
RCS targets by increasing sensitivity and dynamic range, so the effective study of RCS
reduction can be established. The complete measurement system has a simplified block
diagram shown in Figure 5.
The operation principle of the device is based on the detection of spurious signals and the
application of the principle of phase cancelling, by generating a signal of equal amplitude

359
Fig. 5. Block diagram of the measurement system with the active noise cancellation circuit.
and opposite phase. Thus, after operation of the system only the signal of the target under
test is expected to be detected. In this case, errors inherent to the RCS measurement systems
in an indoor or outdoor range are reduced.
In accordance with Figure 5, operation of the developed device consists of a sample of the
signal transmitted by the radar through the directional coupler 1. This signal goes through
an isolator, followed by the modulator that performs phase and amplitude modulation
(with fine tuning provided by the variable attenuator and phase shifter) and are combined
in the directional coupler 2 with the interfering signals. Figure 6 shows the components
used, such as: isolator, attenuator, coupler, filter and etc.
Fig. 6. View of the designed circuit assembled for C-band.

For the evaluation of the system, this circuit was tested by using an anechoic chamber with
dimensions of 9.5 m x 4.5 m x 4.5 m available at IFI/DCTA. For this, two corrugated horn
antennas were installed, with a transmitter and a receiver and a flat plate measuring 0.3 m x
0.3 m was used as a target under test, at frequencies of 5.9, 6.0, 6.2, and 6.4 GHz. Firstly, the
noise level in the chamber was measured with and without the proposed circuit by rotating
the azimuth positioner with no targets mounted on that. Some results are shown in Figure 7
and Table 1 presents the results obtained with the circuit for the frequency range.
360
Fig. 7. Noise level of the anechoic chamber, with and without the circuit, at 5.9 GHz.
Frequency
(GHz)
5.9
6.0
6.2
6.4
Noise level without the

circuit (dBm)
-59
-64
-64
-60
Noise level with the

circuit (dBm)
-85
-84
-85
-82
Reduction (dB)
26
20
21
22
Table 1. Anechoic chamber noise level reduction in C-band (deviation: 0.7 dB).
It is observed that due to the reflection from the anechoic chamber, added to the coupling
between the antennas, the noise level is relatively high (approximately -59 dBm at 5.9 GHz).
Afterwards, a RCS pattern of the flat plate was measured without the circuit (Figure 8 - left)
and it was observed a considerable deformation of the flat plate RCS pattern that can be
assessed by foreseen results in literature (Ross, 1966) and experimental measurements
(Miacci, 2002). After the circuits operation, where the interfering signals were reduced, a
new RCS pattern of the flat plate was measured and it can be observed as being closer than
expected, as seen in Figure 8 (right).
Fig. 8. RCS measurements of the flat plate (0.3 m x 0.3 m) without the circuit (left) and with
the circuit (right) at 6.0 GHz.

361
This last RCS pattern evidences well defined null points and symmetry with a minimal
contamination from external signals. It is also observed that the circuit does not influence
the maximum peak, hence not interfering in the maximum RCS of the target under test.
Table 2 shows the minimum detectable RCS values and it is shown the increase of the
probability of detection of smaller targets, associated to a higher dynamic range of the
system, consequently, allowing the study of RCS reduction to be more effective when
dealing with targets covered by high attenuation ERAM.
Frequency (GHz)
5.9
6.0
6.2
6.4
Detectable RCS without circuit

(dBsm)
-09
-16
-15
-11
Detectable RCS with

circuit (dBsm)
-35
-36
-36
-33
Table 2. Minimum detectable RCS values with and without the circuit (deviation: 0.7 dB).
The results denote that the RCS measurements performed with the developed and built
circuit improved the RCS patterns with a remarkable reduction of interfering signals in the
frequency range. This consideration takes into account the testing environment being almost
completely free of spurious signals that reduce precision of measurement systems and can
lead to false conclusions in analyses of RCS patterns. These preliminary measurements
support other applications in bigger anechoic chambers or even in outdoor measurements.
4. Reflectivity characteristics of simple and complex shapes targets and RCS

reduction by using ERAM
In this section it will be treated the reflectivity characteristics of simple and complex targets
and discussed the experimental results of the RCS reduction research by using ERAM
developed in Brazil.
4.1 Simple shapes targets
The electromagnetic theory allows the RCS calculation of targets that can be mathematically
well defined, considering its size and physical forms. In practice this statement refers to
objects with simple geometries. The solutions for simple objects are very important for two
reasons, firstly, these objects can be considered as reference targets and they can be
constructed and used as calibration references to measure complex geometry targets;
second, approximations of equations used to determine the RCS of complex geometry
objects can be made by detailed knowledge of the behavior of objects of simple geometries
and their combinations (Miacci, 2002).
Table 3 shows the equations used for the RCS theoretical calculation for some simple
geometry shapes in the optical region (Ruck, 1970; Knott et. al., 1993).
In the next sections is presented some experimental results achieved by the research group
in RCS measurements of targets with simple and complex geometries and its RCS reduction
when applying ERAM.
362
Geometry
RCS (m2)
Flat Plate
4
Sphere
= a2
Diedral
8
Cylinder
4
Table 3. RCS of some simple geometries in optical region.

4.1.1 Dielectric lenses (Luneberg reflectors)
In order to validate a RCS measurement setup, firstly, it can be used a certified reference
target called Luneberg reflector. A Luneberg reflector (Figure 9 - left) is a sphere constituted
of dielectric massive shells. Figure 9 (right) depicts a measured RCS pattern of this target
using the setup described in section 3.1, in function of the aspect angles between 0 and 180o
in the frequency of 9.375 GHz.
The maximum power observed in this pattern corresponds to the RCS value of 45 m2,
according to the certified value given by the Luneberg reflectors manufacturer (Thomson
CSF International Inc.). Using the relationship given by equation 9, it is possible to
determine the RCS values of single targets in the same frequency (9.375 GHz).
The almost constant intensity level of the signal, between the aspect angles of 65 and +65o,
corresponds to the region with the same scattering level of the reflector, being a
characteristic RCS pattern for this target. The obtained RCS pattern shows a good agreement
with that one provided by the manufacturer.
Using these data obtained with the Luneberg reflector it is possible to determine the RCS
value for other single targets based on equation 9.

363
Fig. 9. Luneberg reflector positioned in an anechoic chamber (left) and a measured RCS
pattern at 9.375 GHz (right).
4.1.2 Flat plates RCS reduction measurements
The RCS values of a perfect flat rectangular reflector can have its theoretical RCS value
calculated as a function of the incident radiation frequency, according to equation depicted
in Table 3. The RCS reduction method to evaluate a flat plate requires a double face panel,
where one side is used as reflector material (reference) and the other is coated with ERAM.
The panel is fixed on a rotating support, which is positioned in front of the receiving and
transmitting horns according to the setup described in 3.1. The advantage of this
methodology is that it allows the evaluation of the reference and ERAM, by rotating the
device from 0 to 360o, evaluating both sides of the panel, one after the other. Figure 10
shows a simplified scheme of the device used in RCS method. With this method, it is not
necessary to make two separate measurements, because the RCS pattern of the ERAM is
made by rotating the device from 0 to 180o and the reference (metal plate-reflector) is made
from 180 to 360o. Thus, it is a self-calibrating measurement (Knott et al., 1993).
Fig. 10. Device scheme used for the RCS measurements.

The tested ERAM coating was prepared at Materials Division/IAE/DCTA, from Brazil. The
ERAM preparation involved the mixture of 60 % (w/w) of a commercial polyurethane
matrix loaded with 40 % (w/w) of fillers, being carbonyl iron (20 %) and ferrites of MnZn
(10 %), NiZn (5 %) and MgZn (5 %). Physico-chemical characteristics of the fillers and the
polyurethane resin as well as the coating preparation procedures were previously described
364
(Martin, 2002). The ERAM was applied on the aluminum flat panel of 300 mm x 200 mm
surface by brushing with thickness of 0.7 to 2.0 mm.
RCS measurements were carried out at the frequency of 8 GHz and the panel (300 mm x 200
mm) was fixed on the rotating support (Figure 11) and rotated from 0 to 360o, at a scanning
rate of 0.080 rad/s, characterizing both sides of the panel, i.e., the reference side and the
ERAM painted side. From 0 to -1800 it is scanned the reference side and from 0 to + 1800 the
ERAM coated side.
Fig. 11. Flat plate assembled in an anechoic chamber: metal face (left) and ERAM face (right).
Figure 12 shows the RCS pattern of the reference aluminum plate, obtained at 8 GHz, with a
rotation of 180o. It is observed a peak at 0o corresponding to values of -25.3 dBm (that results
in a RCS value about 32.1 m2 using an appropriate reference target), due to the normal
incidence of the electromagnetic waves on the reference plate. The position of the plate is a
critical point for the success of the measurements, where changes of 8o, for example, can
change the signal intensity for nearly 35 dBm. This abrupt dropping of the signal for angles
different from 0o is due to the flat geometry of the target, scattering the electromagnetic
wave impinged on it in different directions of the receiving antenna.
Fig. 12. RCS pattern of an aluminum plate (300 mm x 200 mm), at 8 GHz.
Afterwards, RCS measurements were carried out with the panel having one side coated by
ERAM. Figure 13 shows the RCS pattern expressed in dBm and in square meters, both

365
obtained at 8 GHz. In the range of 0 to -180o it is observed a peak at 900, corresponding to

the normal incidence of the radiation on the reference side of the panel.
(a)
(b)
Fig. 13. RCS measurements of the panel (300 mm x 200 mm), at 8 GHz: (a) in dBm and (b) in
m2. Reference side (-180o to 0o) and ERAM coated side (00 to +180o).
The attenuation obtained at the frequency of 8 GHz, presented above, at the incidence angle
of 90o is about 10 dB. The measurements showed clearly that the RCS values decrease when
the ERAM is applied on the flat plate. The processed ERAM coating is able to attenuate
nearly 94 % of the electromagnetic wave, when impinged at 90o, at 8 GHz.
4.1.3 Dihedral and trihedral corner reflectors
Corner reflectors consist in geometries of special interest in electromagnetic scattering
problems because they provide a large bistatic or monostatic radar cross section over a
broad range of observation and aspect angles. The large echoes from these targets arise from
the multiple reflections between two or three mutually orthogonal flat surfaces forming the
reflectors. Dihedral corners have been used by many works in RCS reduction studies and
also as RCS calibration target (reference target). Trihedral corners yield large backscattering
RCS over a wide azimuth and elevation angular ranges and are widely used for external
radar calibration.
Both dihedral and trihedral corners are usually present in mechanical structures of ships,
aircraft and vehicles, contributing as efficient scattering centers of such targets.
Thus RCS reduction of such geometric corner structures is of great importance in the design
and construction of ships and aircraft which will be under surveillance of radar systems. It
has been reported that RCS reduction of dihedral corners can be achieved by altering the
mutual orthogonality of the flat surfaces. This technique involves changes in original
engineering design of the target. Meanwhile, the use of ERAM can also overcome this
problem.
This section shows the results involving the RCS reduction of a square dihedral and
trihedral corner reflectors, coated with ERAM developed at Materials Division/IAE/DCTA.
The characterization was performed using an anechoic chamber in the frequency range of 8
12 GHz.
366
RCS measurements were carried out at the frequencies of 8, 10 e 12 GHz. In each frequency,
the dihedral and the trihedral were fixed on the rotating support and rotated from 90 to
+90, at a scanning rate of 0.15 rad/s. In this process two pieces were utilized, one as
reference and the other coated with the ERAM (previously described). Figure 14 shows the
assembly of a dihedral inside the anechoic chamber used in this work.
Fig. 14. Assembly of a dihedral on the azimuth positioner.

Figure 15 shows the results at 10 GHz of the reference dihedral (-180o to 0o) and this target
coated with ERAM (0o to +180o). The dihedral RCS pattern is characterized by multiple
reflections of the wave between the orthogonal faces, with two peaks in the aspect angles of
135o and 45o. These peaks are attributed to the wave normal incidence on the flat side of
the dihedral. It can be observed at 10 GHz (Figure 15) an attenuation of nearly 13.6 dB and
10 dB at -135o and 45o, respectively in the RCS pattern of the ERAM coated dihedral, in
comparison to the reference one. Between +45o and +135o angles it is verified an attenuation
of 20 - 24 dB, attributed to the multiple reflections of the wave between the orthogonal faces
of the dihedral.
Fig. 15. RCS pattern of a dihedral (120 mm x 120 mm), reference side (-180 to 0), ERAM
coated side (0 to +180).
Figure 16 depicts the results at 8 GHz of both, the reference trihedral (-180o to 0o) and the
ERAM coated one (0o to +180o). The trihedral RCS pattern is characterized by multiple
reflections of the wave among the three faces. In a similar way, the resulted RCS pattern is
similar to that one verified for dihedrals. At 8 GHz, the RCS pattern shows a reduction of 17
dB for the ERAM coated trihedral in comparison to the reference one, between the aspect
angles of +45o and +135o. However, in the angle range of +135o to +180o and 0o to 45o is
observed a signal increase related to the reference trihedral.

367
Fig. 16. RCS pattern of a trihedral (120 mm x 120 mm), reference side (-180 to 0), ERAM
coated side (0 to +180).
4.1.4 Cylinders
The RCS of a metal cylinder has well-known theoretical equations and making use of some
considerations, the RCS of a cylinder can be calculated. However, unlike the perfectly
conducting sphere, the cylinder is sensitive to wave polarization.
The calculations involving cylinders generally assume that the axial length of the cylinder is
large compared to the wavelength. The RCS pattern of a cylinder can be represented by an
almost constant value over the entire range of the aspect angle that should, ideally, result in
an omnidirectional pattern.
The advantage of using a metal cylinder with a low roughness, as support for testing the
reflectivity of ERAM, lies on RCS values nearly constant with the angle of the incident wave.
Therefore, the ERAM coated cylinder characterization is more influenced by possible
variations of ERAM texture, homogeneity of the absorbing additives distribution and also
the absorber thickness. Possible variations of these parameters will result in change of RCS
pattern as a function of aspect angle of the incident wave. Figure 17 shows the cylinder
mounted on the positioner, located inside an anechoic chamber, uncoated (Figure 17a) and
ERAM coated (Figure 17b).
(a)
(b)
Fig. 17. Assembly of an uncoated (a) and a coated (b) cylinder inside an anechoic chamber
for RCS reduction measurements.
368
Figure 18 depicts the plotted RCS pattern of a metallic cylinder of 32 cm of length and 15 cm
of diameter, rotating the cylinder from +90o to 90o on its axes and keeping the TX and RX
antennas in the same position, at 9.375 GHz. The determined RCS value is constant and
omnidirectional, equal to 1.61 m2. The cylinder RCS measurement needs a tight adjusts on
the dielectric support to avoid contributions of the reflected waves from this apparatus.
Fig. 18. RCS pattern of a cylinder at 9.375 GHz.

The received power level is almost constant at 9.375 GHz (-49.7 1.0 dBm) rotating the
support in 360o around its vertical axe. This curve is typical for this kind of target and it is
attributed to the cylinder shape, which contributes only as a line when the wave impinges
on it. Afterwards, using the same device and the same cylinder it was obtained the RCS
pattern of this target coated with a processed ERAM (Figure 19).
Fig. 19. RCS patterns of cylinder without ERAM (black curve) and coated with ERAM (red
curve) at 10 GHz (left) and 12 GHz (right).
Figure 19 shows the RCS pattern of the cylinder at 10 GHz and 12 GHz coated with the
ERAM presenting 1.2 mm of thickness loaded with NiZn ferrite and carbon black. In this
case, the pattern shows the reflectivity variation in function of the aspect angles,
characterizing a RCS reduction of the target. This variation is attributed to the heterogeneity
of the processed ERAM, related to both the ferrite and carbon black particles distribution
and the absorber thickness.

369
4.2 Complex shapes targets

The total RCS of a target of complex geometry can be determined by various techniques of
different levels of complexity. One approach involves dividing the object into several parts,
and these should approach targets of simple geometry. The determination of the RCS of
each of these parts, the sum of their individual contributions to the total reflected field,
taking into account their respective phase angles and reflectivity, provide the final value of
the target RCS of complex geometry.
A complex target has multiple scatters interacting with each other, changing the RCS as a
function of aspect angle. For those targets, the peaks of RCS pattern may be many wavelengths
apart, but they can add up constructively or destructively by a fraction of degree.
The RCS pattern of a complex target is a function of many variables in the radar system and
environmental measures. In addition to the change in aspect angle and shape of the target,
there are multipath and atmospheric effects. The main reflection mechanisms of a complex
object are: specular reflection, surface waves, edges diffraction and cavities.
In the next section will be treated the RCS studies of complex targets (mainly missiles) done
by the Materials Division/IAE/DCTA from Brazil.
4.2.1 Missiles
Missiles possess one of the most fundamental complex geometries to RCS prediction and
electromagnetic characterization. These ones present a relatively simple geometry in
comparison with more elaborate targets, such as aircraft and ships. However, these shapes
allow detailed studies of the scattering phenomenon of the simple geometries that constitute
them (Ruck, 1970). In this section, it is shown some RCS results achieved in the
characterization of a hypothetical missile and a real missile section, in the microwave range
of 5 to 7 GHz. ERAM application on the studied targets is also evaluated.
4.2.1.1 Characterization of a hypothetical missile in indoor measurements in C-band
A prototype was projected and constructed and called hypothetical missile. The purpose is
to understand the behavior of an actual complex target, from the well known physical and
mechanical characteristics of a prototype. The dimensions of the hypothetical missile
correspond to those of a cylinder measuring 0.32 m x 0.15 m and four square flat plates
measuring 0.15 m sideways.
The study presented in this section intends to compare the results obtained experimentally
with data obtained theoretically, by predictions based on literature (Knott, 1993). An
algorithm to simulate the radar signature of complex geometry was developed with the
purpose of both estimating the contributions of each simple component of the target and
assisting the analyses of the experimental measurements (Miacci, 2006).
The foreseen aspect, by simulation, of the RCS pattern for this type of target is shown in
Figure 20. This figure demonstrates the appearance of the peaks relative to the frontal, back
and sides contributions of the target. Figure 20 shows also a view inside the anechoic
chamber, where the system was assembled for the RCS experimental measurements for the
frequency range of 5.8 to 6.4 GHz.
370
Fig. 20. Simulated RCS pattern for the Hypothetic Missile at 6.0 GHz and the Hypothetic
Missile placed in the anechoic chamber.
The same missile was placed in the anechoic chamber and characterized in accordance with
the procedures and the RCS patterns obtained at frequencies 6.0 GHz and 6.4 GHz are
shown in Figure 21. Comparing Figures 20 and 21 good agreements of the patterns is
verified. However, the figures relative to the experimental measurements show less
detailing of the secondary lobes. From these patterns, the RCS values can be calculated for
the hypothetical missile, as shown in Table 4.
Fig. 21. RCS pattern measured from the Hypothetic Missile at 6.0 GHz (left) and 6.4 GHz (right).
Frequency (GHz)
5.9
6.0
6.2
6.4
RCS maximum (dBsm)

09.2
11.5
11.9
13.0
Table 4. Experimental RCS Values of the Hypothetical Missile (deviation: 0.7 dB).
The analysis of Table 4 shows the maximum RCS variation of the hypothetical missile (9.2
13.0 dBsm) as a function of frequency, hence evidencing the sensibility of this measure with
small variations on the wavelength of incident radiation. From such measurements it can be
affirmed that the system is reliable to accomplish RCS measurements of complex geometry
targets.
4.2.1.2 RCS reduction of the hypothetical missile
Reflectivity measurements of the Hypothetical Missile coated with ERAM were performed
(Figure 22) and the results were compared with those presented in preceding section. In

371
order to improve the evaluation of the ERAM effect on the Hypothetical Missile, this
material was applied in only a portion of this target under test, i.e., a face of the missile was
not covered with the absorber. Thus, we applied the ERAM on the front and over one side,
while the rear and the other side did not have their surfaces coated (Figure 23).
Fig. 22. View of the Hypothetical Missile with applying of ERAM.
Fig. 23. Reflectivity curve of the Hypothetical Missile partially covered with ERAM, at 6.2
GHz (left) and 6.4 GHz (right).
Table 5 shows the results and compares the RCS reductions achieved.
Frequency (GHz)
5.9
6.0
6.2
6.4
RCS Reduction (dB)

16
19
19
21
Measured RCS (dBsm)

-6.8
-7.5
-7.1
-8.0
Table 5. RCS reduction with the ERAM on the Hypothetical Missile (deviation: 0.7 dB).
The analysis of Table 5 shows that the RCS reduction values (16 - 21 dB) are in accordance
with the attenuation values of the used ERAM (~ 20 dB attenuation). This behavior shows
that the proposed experimental method is effective for the study of RCS reduction of
complex geometries. The main contributions are highlighted in the figures.
372
4.2.1.3 Characterization of a missiles section in indoor range in C-band

After verifying the reliability of the parameters adopted in the employed methodology, the
characterization of the actual complex target was performed. For this, the selected section of
the actual missile was placed in the anechoic chamber (Figure 24). The results obtained in
the reflectivity measurements are illustrated in Figure 25.
Fig. 24. View of the missiles section positioned in the anechoic chamber.
Fig. 25. RCS patterns of the missiles section at 6.0 GHz (left) and 6.4 GHz (right).
Figure 25 represent the obtained curves that are typical of the reflectivity patterns of the
missiles section characterized in the anechoic chamber at a frequency range of 5.9 to 6.4
GHz. The comparison of these figures shows that the frequency increase is accompanied by
greater detailing of the patterns. This behavior shows a larger contribution of the secondary
lobes. These figures reveal also the real complexity of the RCS patterns of complex targets.
The results of these measurements suggest the presence of phenomena commonly seen in
experiments performed with this type of geometry. As the missiles section presents similar
geometry to that one observed in the hypothetical missile, it is expected that its pattern
adjust adequately to those measured for the hypothetical missile.
However, the presences of triangular-shaped flat plates (the missiles warp) and the
irregular aspect of the surface contribute with new peaks, increasing the complexity of the
pattern. Besides the front part of the missile is not a perfect flat plate, since before this area
has the boundary with the tracking compartment, modifying the outline of this targets
reflectivity pattern, contributing also to the complexity of the RCS pattern.

373
Another phenomenon observed is with regard to the non-symmetry of the pattern between
the angles of 180 to 0 and 0 to +180, which reveal peaks of different shapes and
amplitudes. This is due to the contributions of physical details of the missile, such as
compartments of electronic circuits, conduits where the cables pass through and connectors
integrated to the cylindrical body of the target.
Such devices positioned on the missiles body, which at first seem to be of little influence,
cause wave scattering on the targets surface, which alter the reflection as a whole, such as
diffraction edges and traveling waves. These constructive attributes of the missile are
displayed in the measurements and results in alterations in the pattern peaks. These results
justify the complexity of the reduction of the targets RCS of complex geometry, as foreseen
in literature (Ufimtsev, 1996).
The measured RCS values for these cases can be seen in Table 6.
Frequency (GHz)
5.9
6.0
6.2
6.4
Maximum RCS (dBsm)

7.3
7.9
10.2
10.5
Table 6. Experimental RCS values for the missiles section (deviation: 0.7 dB).
4.2.1.4 RCS reduction of a missiles section
ERAM was used covering selected faces of the missile, i.e., the faces covered and uncovered
resulted in a pattern for direct comparison of RCS reduction. Analogously to make with the
hypothetical missile, ERAM was applied on the front part and on one side, while the rear
and the other side remained uncoated (Figure 26). Figure 27 show the RCS reduction
patterns obtained at 6.0 and 6.4 GHz.
Similarly to that observed in the Hypothetical Missile measurements, it is verified the RCS
reduction of the faces with ERAM. In a similar way, the application of ERAM promotes
attenuation values of the targets RCS close to those determined for the ERAM (~ 20 dB).
Table 7 shows the reduction obtained in this case.
Fig. 26. View of the missiles section covered with ERAM.
374
Fig. 27. Reflectivity curve of the missiles section partially covered with ERAM at 6.0 GHz
(left) and 6.4 GHz (right).
Frequency
(GHz)
5.9
6.0
6.2
6.4
RCS Reduction
(dB)
11.5
14.0
18.5
18.0
Measured RCS
(dBsm)
-4.2
-6.1
-8.3
-7.5
Table 7. RCS reduction of the missiles section coated with ERAM (deviation: 0.7 dB).
5. Conclusions
This work shows a reliable methodology for RCS measurements of simple and complex
targets. Indoor RCS measurements of simple and complex targets in C and X-band are
explored and the results show good agreement with the literature. Reflectivity measurements
of the studied targets coated with electromagnetic radiation absorbing materials demonstrate
that the proposed measurement system is satisfactory for RCS reduction studies.
This work presents also, results of an active cancellation circuit as an enhancement
technique for the RCS measurements. The proposed circuit increases the detection
probability of smaller targets, associated with a greater dynamic range of the measurement
system. Therefore, the proposed circuit makes the study of RCS reduction more effective
concerning small targets covered with high attenuation ERAM. An increase of dynamic
range of up to 26 dB at 5.9 GHz frequency is observed. Consequently, the minimum
detectable RCS values using the proposed circuit increased significantly, as for instance -9.0
dBsm (without the circuit) up to -35 dBsm (with the circuit) at 5.9 GHz.
The recovering of any targets with a determined ERAM developed at Materials
Division/IAE/DCTA in Brazil it was possible to compare and to discuss the influence of the
ERAM on the observed RCS reduction. The used methodology showed useful for attending
aeronautical and telecommunication applications.
For the first time in Brazil, the RCS of missiles sections (hypothetical missile and an actual
missile section) was experimentally determined, in C-band in an anechoic chamber.

375
The evaluated RCS measurement system proved to be effective, with measured values
within the expected error range (lower than 1 dB). The obtained results enable to conclude
that this methodology is applicable in the characterization of various types of simple and
complex geometry targets in indoor and outdoor ranges, attending the far-field condition. In
function of the results obtained, employing the presented system, further in-depth studies of
RCS reduction is feasible for other military platforms, given its functionality.
6. Acknowledgment
The authors acknowledge FAPESP (So Paulo Research Foundation) (Process 03/00716-4
and 11/03093-4), CNPq (National Council for Scientific and Technological Development)
(Process N. 305478/2009-5) for the financial support and also to Dr. Carlos Frederico de
Angelis from INPE/Brazil and the Ministry of Defense / Aeronautics Command, for their
collaboration and support in this project.
7. References
Balanis, C. A. (1982) Antenna theory, analysis and design, Harper & Row, ISBN 0-47160-352-X
New York, USA
Bhattacaryya, A. K.; Sengupta, D. L. (1991), Radar cross section analysis and control, Artech
House, ISBN 0-89006-371-0, Norwood, USA
Birtcher, C. R.; Balanis, C. A. (1994), RCS measurements, transformations and comparisons
under cylindrical and plane wave illumination. IEEE Transactions on Antennas and
Propagation, v. 42, n. 3, p.329-333.
Blake, L. V. (1986), Radar range performance analysis, Artech House, ISBN 0-89006-224-2,
Norwood, USA
Burgess, L. R., Berlekamp, J. (1988) Understanding Radar Cross-Section Measurements.
Microwaves Systems News and Communications Technology, p. 54-61.
Crispin, J. W.; Siegel, K. M. (1968), Methods of radar cross section analysis, Academic Press,
New York, N.Y., USA
Currie, N. C. (1989), Radar reflectivity measurement: techniques and applications, Artech House,
ISBN 0-89006-345-1, Norwood, USA
Dybdal, R. B. (1987) Radar cross section measurements. Proceedings of the IEEE, v. 75, n. 4,
p.498-516.
Jasik, H.; Johnson, R.C. (1984) Antenna engineering handbook, 2 ed., Mcgraw Hill, New York,
USA
Jenn, D. C. (1995), Radar and laser cross section engineering, AIAA Education Series,
Washington, DC, USA
Knott, E. F.; Schaeffer, J. F.; Tuley, M. T. (1993) Radar cross section (2nd ed.), Artech House,
ISBN 1-891121-25-1, Norwood, USA.
Kouyoumjian, R. G.; Peters JR. (1965), Range requirements in radar cross section
measurements. Proceedings of the IEEE, v. 53, p. 920-928.
Martin, I. M.; Dias, J. C.; Rezende, M. C. (2002), Reflectivity in the microwave range of
polyurethane coatings loaded with NiZn ferrites. Journal of Materials Research, v.1,
n.1, p. 1-10
Miacci, M. A. S. (2002), Experimental Measurements of the Microwave Backscattering of Single
Shapes Targets, (In Portuguese), Dissertation (Master of Science) ITA, Brazil
376
Miacci, M. A. S. (2006) Improvement of Missile RCS Measurements by using of Active Cancellation

Technique, (In Portuguese), Thesis (PhD) ITA, Brazil
Ross, R. A. (1966), Radar cross section of rectangular flat plates as function of aspect angle.
IEEE Transactions on Antennas and Propagation, v. AP-14, p. 329-335.
Ruck, G. T. (1970), Radar cross section handbook, Plenum Press, New York, USA.
Skolnik, M. I. (1990), Radar handbook (2nd ed.), Mc Graw Hill, ISBN 0-07-057913-X, New York,
USA.
Thomson Inc. (1988), Luneberg Reflectors and Lenses, Technical Bulletin, September, 8 p,
France.
Ufimtsev, P.Y. (1996), Comments on: Diffraction principles and limitations for RCS
reduction techniques. Proceedings of the IEEE, v. 84, n.12, p.1828-1851.
17
An Intelligent System for Efficient Rigid Film
Anticounterfeiting Inspection
Michael Kohlert1, Christian Kohlert2 and Andreas Knig1
Institute of Integrated Sensor Systems, Department of Electrical and Computer

Engineering, University of Kaiserslautern, Kaiserslautern,
2Department of Research and Development, Klckner Pentaplast
GmbH & Co. KG, Heiligenroth,
Germany
1. Introduction
The danger that results from taking ineffective pharmaceutical products or consuming food
that does not comply with the requirements, leads to significant health, and life hazard for
people. Anticounterfeiting of, e.g., pharmaceutical products, guarantees customers the
delivery of original products. The rigid film industry is developing automated systems to
detect product piracy. For this aim, an approach to an automated inspection system for
rapid and reliable product verification based on optimized insertion of infrared (IR) and
ultraviolet (UV) m-pigments in rigid films has been developed in this work and
statistically meaningful samplesets have been extracted. The industrial manufacturing
process has been enhanced for optimized insertion of pigments in rigid films with regard to
size, type, density, and process step. The pigments are activated with IR or UV light in an
encapsulated laboratory system specially developed here. Filter on illumination sources and
colour cameras limit the activation and the emission range. Due to optimized film
manufacturing and measurement system, the evaluation of m-pigments can be achieved by
a two-stage process of state of the art supervised colour segmentation, followed by a blob
analysis. The recognition results of the conceived intelligent engineering system fully meet
the industrial specification requirements. The project has advanced from a laboratory, small
volume production to a large volume test customer production and evaluation.
2. An intelligent inspection system

Product piracy causes a loss of 600 bn USD world wide every year (Barbieri, 2004). A
product copy due to plagiarism is difficult to differentiate from originals. The danger of
unknowingly consuming vitally important pharmaceutical products that are either not
effective or even poisonous is a stringent societal problem. The financial situation of product
markets shows the problem of capital investment in new products when no profit is realized
due to illegal copies. To disclose illegal copying of original products the rigid film industry
developes reliable application and authentification systems to make product packages and
their content safer. As one part of the first stage of the supply chain the rigid film industry is
able to achieve this goal for the whole supply chain.
378
The rigid film industry wants to secure the path from the producer to the customer to avoid
replication at low additional costs. The package appearance should not change and the
whole system has to be integrated easily in the packaging-process.
Klckner Pentaplast, Montabaur, Germany, has been looking into ways of marking
packaging film in such a way that it can be subsequently recognized by the film
manufacturer and can also be used for product protection by the converter.
This work`s aim is to develop measuring systems with costs lower than 2000 for visual
detection of pigments and authentification of the product package. The variable cost of the
pigment has to be lower than 0.1 Cent/kg of foil. The reproducibility of equal distribution
has to be guaranteed during the production process.
The specifications of the research and development department of Klckner Pentaplast
request the ability to measure the UV- and IR-pigment density in films by using light
sources, filters and cameras, and achieve a correct separation of actual pigments and fibres
due to contamination.
In this work, the process of calendering is described in section 2.1 (Fig. 1), focusing on the
insertion of light emitting pigments for anticounterfeiting.
Fig. 1. Insertion points for pigments in the process of rigid film production: 1. mixing
machine, 2. kneader, 3. Calendar
A short description on the kinds of pigments, which are used, their activation under IR and
UV light sources and related data acquisition will be shown in section 2.2. A hierarchical,
two-step classification- approach, that serves to segment the acquired images of pigmented
films explained in section 2.3. An overview of the whole anticounterfeiting process is
presented in section 2.4, before concluding.
An Intelligent System for Efficient Rigid Film Anticounterfeiting Inspection
379
2.1 Basic process and application specific adaptations

The standard production of marked transparent and coloured films can be described in a
complete process diagram (Fig. 1). Additional pigments can be inserted in a master batch,
which is mixed in two mixing machines with high and low temperature, and the kneader,
which plastifies the batch. The insertion of pigments in the calendar where the film becomes
formed is inefficient. The produced foil is slittered at the end of the calendaring process for
the transport stage. Further changes on the film will be made in a printing firm or the other
industrial plants after transportation.
The pigments dedicated to product identification were inserted in the recipe of the mixing
machines. Anorganic and organic types have been investigated. Anorganic pigments are
FDA approved and do not cause any problems to industrial usage. Thus, polymeric films
can be doped with fluorescent pigments (FDA approved, anorganic) for protection of
polymeric packages. Standard pigment sizes from manufacturers are 3m, 5m, or 10m.
An inserted pigment density of 0.0001% up to 1% is used in this work. A higher density of
more than 0.1% leads to an unintentional coloration of the polymeric film, which was a
motivation for an upper and lower boundary and stepwise increment in variation.
The intensity and distribution of the pigments is depending on the thickness of the film. The
following permutations have been explored by preliminary work with 300 tests on pigment
size, type, density, and film thickness empirically. Starting with the smallest size, smallest
density, random type, and smallest film thickness, about 1500 pigmented films were
produced with 20 different pigments of three different sizes, 3m, 5m, and 10m.
Successful laboratory tests with films of 60m up to 800m thickness allow the use of
product identification even for multilayered films.
Dedicated measurement laboratory equipment is developed for the detection of the
pigments in the films. UV fluorescent tubes (UV-A, UV-B), and light emitting diodes (UV-A,
UV-B) activate pigments, and other materials. IR light emitting diodes (880nm 1050nm),
and a laser-diode (980nm) activate the IR pigments used in the film.
The range of settings (Tab. 1) will be completed by achieving best recognition results with
measuring devices in chapter 2.2.
Types : = 20
Size : = 3
Density : = 5
Film Thickness : = 5
Organic (natural)
3 m
0,0001%
200 m
Anorganic (synthetic)
5 m
0,001%
300 m
Ultraviolet (300-370nm)
10 m
0,01%
400 m
0,1%
600 m
1%
800 m
Infrared (940-980nm)
Table 1. Ranges of preliminary tested parameters for 1500 films with 20 different pigments
The surface analysis and classification methods show advances for anorganic IR pigments
with 5m, 0.001% density, and 300m film thickness. The whole tested setting are shown in
figure 2.
380
Fig. 2. Tested pigment parameters for 20 types, 3 sizes, 5 densities, and 5 film thicknesses
Four measurement units where deployed. The measurement setup and data acquisition will
be explained in chapter 2.2.
2.2 Measurement setup and data acquisition
For the required measurements, four different measuring devices (UV lamp system, visual
UV-IR LED system, non visual UV-IR LED system, and laser diode system) were developed
in this work.
The settings from chapter 2 have been investigated during the measurement setups and data
acquisition with these four units.
The UV illumination is generated by fluorescent lamps and light emitting diodes at
activation wavelengths of 300nm to 370nm. Depending on the pigment emission a
bandpass- filter is used to detect the pigment wavelength with a colour camera. It is very
efficient to use filters for camera and light source to minimize reflections and other
influences.
Pigments are activated in UV or IR light and emit in lower or higher wavelengths (Fig. 3
shows an example). These pigments are called up- and down- converters. This feature is
used for the following measuring devices.
Blisters are prototypes of pharmaceutical packages that show the produced film in the way
they are used later for products. The activation light source for the film is placed in 45
above the object. A camera captures the image in 90. The coloured image acquisition allows
a colour-spaced segmentation as a next step. In this work about 100 blisters were captured.
381
Fig. 3. Measuring device (left) for UV pigments in a blister (right)

The second system is an IR recognition system, which detects pigments with an activation
wavelength from 900nm to 1000nm. (Fig. 4)
Fig. 4. Image acquisition on 10cm with a 11-mega-pixel-camera (b/w), image section of

2cm with a fluorescent IR pigment density of 0,001% and 5m
A laser diode with an optical spreading device is able to illuminate an area of 10 cm in an
appropriate normalised distribution for image acquisition. The IR lighted area and the
selected image section with fluorescent IR pigments are shown in Fig. 4. The illuminated
pigments are captured by the black/white camera. For this specific measurement system,
100 testing films have been captured for ensuing analysis.
In Fig. 5 a measurement unit (Visual UV-IR LED System) is shown, which is in development
for the recognition of UV, and IR pigments. The hand-held unit allows a fast analysis of
rigid film surface at all places in the plant.
382
Fig. 5. Hand-held unit for internal use for fast recognition of UV, and IR pigments
Different measurement units were tested as described in this chapter. A compact installed
laboratory or inline- system is expensive. (Tab. 2)
UV lamp system
Laser-diode System
3000
365nm/ 300nm
2 lamps
15.000
975nm
1 laser-diode
Visual UV-IR LED

System
1500
320nm/940nm
8 diodes/ 8 diodes
Non-Visual UV-IR
LED System
1000
365nm/940nm
12 diodes/ 12 diodes
Table 2. Price range for the different illumination systems

As an alternative the hand-held systems can be used after the calendaring process for fast
identification in the plant, at the pharmacy or integrated in a cash terminal. In the textile
industry this system can be used for identification of jackets after the production process.
2.3 Two-stage recognition system and results
Based on the optimized rigid films and measurement setups the ensuing evaluation is
already feasible by state of the art techniques applied in a two stage approach. The first
stage of classification separates the light emitting pigments from the background and the
reflections caused by the light source of the measuring device by supervised segmentation.
This stage is called here microclassification. To separate the pigments from contaminations
(fibres) a second classification is carried out on the segmented images, denoted here as
macroclassification. The origin image is shown in figure 6.
Supervised segmentation of images generates binary images with candidate pigment pixels
highlighted. Expert knowledge is integrated during data selection and labelling in the
design process. For manually selected relevant sample pixels from 30 specially fabricated
films (type: anorganic, size: 5m, density: 0,001%, not thermoformed), class affiliations were
interactively generated. (Fig.7). From the RGB images LAB color-spaced images of the same
size were computed, and from the selected pixel coordinates 1200 labeled three dimensional
383
vectors have been acquired, balanced for both classes and randomly split into training and
sample sets of 600 vectors each.
Fig. 6. 640x480 pixel RGB image of a rigid film under UV (302nm)
Fig. 7. Manual selection of example pixels representing the two classes reflections/background
(left, middle), and pigments (right) for segmentation classifier training
The selected pixels (Fig. 7) define vectors of LAB-colour space for each class to be presented.
In a pre-examination the LAB-colour space generated best segmentation results, but other
colour spaces can also be used (Hunter Labs [HL], 1996).
The classifier achieves supervised segmentation by assigning a class to each pixel of the RGB
image.
For the underlying problem complexity, a non-parametric classifier should be chosen for
optimum classification results.
The training vectors of the defined classes separate all samples into different classes by
using the L, a, or b data of a sample.
The L, and a data of each vector achieved better results for the Support Vector Machine
(Matlab-Toolbox) with 99.83%. Pigment and Background vectors can be distinguished more
efficiently. (Fig. 8)
384
The CorrectRate defines the possibility for becoming classified as a pigment. For efficient
use the whole Anticounterfeiting system has to detect 99% correct. Errors in the
microclassification will be dominantly compensated in the second stage. The size
parameters prevent a wrong assignment (Christianini & Shawe-Taylor, 2000).
a
Fig. 8. Scatterplot of 600 training- and 600 testing vectors of L, and a data from L,a,b colour
space (SVM Matlab Toolbox) CorrectRate = 99.83 %
The a, and b data of each vector showed similar recognition results in comparison with the
L, and a data correlation.
Thus, the L, and b data correlates with 99.5%, the L, and a data will be preferred for further
classification (99.83%).
The classification results of the training data of 600 vectors (features: L,a,b of colour space
L,a,b) and the testing data of 600 vectors are shown in Tab. 3. Four classifiers (kNN, PNN,
RNN, and SVM) are known as state of the art methods. These types use different
mathematical algorithms for the separation, and classification of sample sets.
kNN
Classification rate k=1 99.5 %
for test set using k=2 99.5 %
hold-out approach k=3 99.5 %
k=5 99.5 %
k=6 99.5 %
k=8 99%
k=10 99 %
Table 3. Microclassification results
PNN
: 0.1
99%
: 0.03 99%
: 0.001 60%
RNN
SVM
99%
99.83 %
quadratic
Kernel
C = 0.1 100000
Gamma = 0.001 - 10
385
Thus, the k-neareast-neighbor (kNN), the reduced-nearest-neighbor (RNN), the probabilistic

neural network (PNN), and the support vector machine (SVM) classifier have been applied,
and found already to be satisfactorily working (Duda et al., 2001).
The segmented images from the 99.83% correct (SVM) microclassification are used for blob
analysis, which has the potential to eliminate residual erroneous pixel classification by the
context of pixel cluster shapes. The macroclassification, based on geometrical features of
blob analysis, separates light emitting pigments from fibres. (Fig. 9)
Fig. 9. Separation of fibres from true pigments after segmentation

Fibres originate from contamination or clustered particles (Fig. 10). The geometrical
structure of fibres is different from the structure of the pigments.
Fig. 10. Subregions containing fibres and pigments for training and test data
386
From eight blob features, area, perimeter, balance point, compactness, outer circle, inner
circle, symmetry, limitative rectangle, the most significant ones (area, rectangle) are
automatically chosen by feature selection to distinguish between both classes.
Out of pre-selected 300 vectors for pigments, the training set of 100 feature vectors of
pigments, 100 feature vectors of fibres, and the test data consisting of 50 feature vectors of
pigments and 50 feature vectors of fibres from the same images were employed. (Tab. 4)
Classification
rate for test set
using hold-out
approach
kNN
k=1
k=2
k=3
k=5
k=6
k=8
k=10
100 %
100 %
100 %
100 %
100 %
100 %
100 %
PNN
: 0.1
100 %
: 0.03 100 %
: 0.02 72.5 %
: 0.001 52.5 %
RNN
100 %
SVM
100 %
quadratic
Kernel
C = 0.1 100000
Gamma = 0.001 - 10
Table 4. Classification results for the blob analysis

The result of the k-nearest neighbor classification is an image only consisting of pigments.
Classes like background, reflections or fibres are not included any more (Gonzales &
Woods, 2008; Knig et al., 1999; Knig, 2001).
After detection and elimination of the fibres, the located pigments can serve for a higherlevel authentification process in a following step (Fig. 11).
Fig. 11. Selection process (Blue identification pigments in software analysis)

Using the described state-of-the-art method for plane films or blisters as shown in the
image, a selection of partial areas allows a defined analysis of completed packages at the
387
end of the production process. This has been experimentally verified on the detection results
for three randomly chosen rigid films. The coding process separates package/production id,
date, corners and angles to define a unique digital signature for each package. The selected
pigments have a geometrical correlation (angles) which is saved in the digital signature of
the package code. The pigments are selected randomly, but their coordinates can be
recalculated in the verification process (Fig. 12).
Fig. 12. Coding process

After the polymeric coding a database server saves the scanned angles for further
comparison of the packages at the client. The small code allows a fast, brand recognition
(Fig. 13).
Database with
Coding Information
Fig. 13. Polymeric Package Coding

An inline-inspection system (see Fig. 14) is actually in progress for first detection during the
calendaring process. The patented technology is available to all users of rigid film from
Klockner Pentaplast. The inspection system is not implemented yet. Figure 14 shows an
impression of the final system at the end of the development process.
Authentic products can be reliably protected with a fingerprint which is obtained through
the incorporation of randomly-distributed pigments in the packaging, which are then made
visible (C. Kohlert et al., 2010).
388
Fig. 14. Inline-Inspection system at the end of the calendaring process shortly after the
thermoforming-/ or the slitting process
2.4 Complete anticounterfeiting system
Using UV or IR pigments for anticounterfeiting is a novel approach in the rigid film industry.
The insertion in polymeric recipes allows high level security for the whole supply chain. The
complete process of anticounterfeiting rigid film is shown in the next figure (Fig. 15).
Fig. 15. Processing diagram for pigment insertion, image acquisition, two-stage classification,
and coding stage
389
The insertion of pigments can be done at the mixing machines, the kneader, or the calender.
The mixing machines, as the best insertion point, consisting of a mixture of particles (batch)
allow a fast disposal of the pigments within the batch.
The choice of the pigment type depends on the measuring device. For optimal results IR
pigments are better to distinguish from other batch particles. Most batch particles emit in
UV ranges, so the UV emission can be ten times higher than the defined insertion.
A two-stage classification separates pigments from other materials in an image with a
correct rate of 100%.
For anticounterfeiting of rigid films a specially developed software authentificates a
polymeric package via recording, and reading of random pigment structures correctly. It
calculates the geometrical positions of the pigments with 99% accuracy, and checks the
apparent authenticity of a film, or a blister within 6 seconds.
3. Conclusion
In this work, a viable and economic implementation of a product authentification system
was developed, that is able to recognize, and locate IR and UV pigments in specially
fabricated polymeric films. A special feature of this approach is, that film manufacturing
and measurement were optimized with regard to recognition accuracy, so that rather basic
methods could be employed in the back-end for successful identification, which is also
important for aspired low-cost detection devices.
The modification of the polymeric standard process of film production by insertion of
pigments was implemented in two of three possible ways. The mixing machine showed best
results for optimized distribution. An insertion during the calendaring process causes no
optimal results. The kneader is another possibility to insert pigments as fluid, and will be
tested in further tasks.
For optimal detection, and separation from fibres, anorganic IR pigments with a density of
0,001% and a size of 5m showed best results in comparison with UV pigments.
Four different measurement systems, an UV lamp system, a visual UV-IR LED system, a
non visual UV-IR LED system, and a laser-diode system were established. The IR LED
systems showed best results for film recognition.
A hierarchical two-stage recognition system with state of the art methods for pigment and
localization of 100% accuracy has been achieved.
Classification techniques, e.g. Support Vector Machine have been investigated with existing
and new data sets to improve system reliability for viable inline-process measuring, and
portable inspection systems.
The industrial specification of the company with regard to a homogenous insertion of IR
and UV pigments in the calendaring process, and a separation of fibres is fully achieved. In
future work, more extensive datasets will be generated, in particular for blob analysis, to
ensure wider generality.
The real time behaviour and resource demands of the proposed system have to be regarded
carefully, and potentially optimized for hand-held inspection devices (M. Kohlert et al., 2010).
390
4. References
Barbieri, A. (Sept. 29, 2004). Getting real about fake designer goods, In: Bankrate.com, Feb. 26,
2011, Available from
http://www.bankrate.com/brm/news/advice/scams/20040929a1.asp
Christianini, N., & Shawe-Taylor, J. (2000). An Introduction to Support Vector Machines and
Other Kernel-based Learning Methods (1st edition), Cambridge University Press, 9780521780193, Cambridge, UK
Duda, R.O., Hart, P.E., & Stork, D.G. (2000). Pattern Classification (2nd edtion), WileyInterscience, 978-0471056690
Gonzales, R. C., & Woods, R. E. (2008). Digital Image Processing (3rd edition), Prentice Hall,
Upper Saddle River, NJ
Hunter Labs (Aug. 1-15, 1996). Hunter Lab Color Scale. Insight on Color 8 9, Feb. 26, 2011,
Available from: http://www.hunterlab.com/appnotes/an08_96a2.pdf
Knig, A., Eberhardt, M., & Wenzel, R. (1999). Image Processing Europe, QuickCog SelfLearning Recognition System Exploiting machine learning techniques for transparent
and fast industrial recognition system design, (Sept./Oct. 1999), pp. 1019
Knig, A. (2001). Pattern Recognition in Soft Computing Paradigm, Dimensionality Reduction
Techniques for Interactive Visualisation, Exploratory Data Analysis and Classification,
Vol. 2, (2001), pp. 137, ISBN 981-02-4491-6
Kohlert, M., Kohlert, C., & Knig, A. (2010). Lecture Notes in Computer Science, Automated
Anticounterfeiting Inspection Methods for Rigid Films Based on Infrared and Ultraviolet
Pigments and Supervised Image Segmentation and Classification, Volume 6276,
(2010),pp. 321-330
Kohlert, C., Schmidt, B., Egenolf, & W., Zistjakova, T. (2010). Verpackungsfolie fr
Produktauthentifizierung, Authentifizierungsverfahren und system. DE 10 2008 032 781
A1

(MD Zahurul Haq) Applied Measurement System

Uploaded by

Copyright:

Available Formats

(MD Zahurul Haq) Applied Measurement System

Uploaded by

Document Information

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

(MD Zahurul Haq) Applied Measurement System

Uploaded by

Copyright:

Available Formats

APPLIED

Md. Zahurul Haq

Applied Measurement Systems

Applied Measurement Systems, Edited by Md. Zahurul Haq

Measurement: System, Uncertainty and Response 1

Internal Combustion Engine Indicating Measurements

Experimental System for Determining

Non Contact Measurement System with

Study on Wireless Torque

Calibration of Measuring Systems

Determining Exact Point Correspondences

Design Methodology to Construct

Non Invasive Acoustic Measurements for

Approximation and Correction

Planar Microwave Sensors for Complex Permittivity

Prof. Md. Zahurul Haq, Ph.D.

Bangladesh University of Engineering & Technology (BUET)

A measuring instrument transforms a measurand, i.e., a physical variable to be measured, to

To maintain quality control and quality assurance in production;

It is sensible to have a generalized description of both the operation and performance of

Fig. 1. Functional elements of an instrument or a measurement system.

Maintained as absolute unit standard

Table 1. Hierarchy of Standards (Figliola & Beasley, 2011).

Table 2. Examples of Temperature Standards (Figliola & Beasley, 2011).

3. Static characteristics and specications

Standard or reference value errors

Interaction between the instrument and test media

3. Data-acquisition error Measurement system operating conditions

2. ISO/IEC Guide 98:2008 Uncertainty of Measurement Guide to the Expression of Uncertainty

Any single measurement of a parameter, x, is inuenced by different elemental random error

Fig. 2. Illustration of measurement errors.

, and the standard deviation is given by

Table 4. Students t distribution (Figliola & Beasley, 2011).

If a 21st data points are to be taken:

= 10.00 (2.093 0.28) = 10.00 0.59

3.3 Uncertainty analysis: error propagation

where, u1 , u2 , , u N are the uncertainties associated with x1 , x2 , , x N .

, where, voltage, V and current, I are measured as

(10 5)2 + (100 0.1)2 =

2500 + 100 = 51.0 W = 5.1%

Hence, estimated power, P = 1000 W 5.1%.

4. Specication of dynamic systems and response

f (t) Input quantity imposed on the system,

where, k Vs /L (Volts per unit length).

k 1/ao static sensitivity,

A practical example of a rst-order instrument is a mercury-in-glass thermometer. If a

A second-order instrument is one that follows the equation:

A practical example of a second-order instrument is a weight measuring spring balance as

f (t) forcing function

Fig. 3. (a) Spring-mass-damper system, (b) Free-body diagram.

v(t) applied voltage (V)

f (t) forcing function (N)

Fig. 4. Spring-mass-damper system and analogous RLC circuit.

Graphical representation of these two functions are shown in Fig. 5.

Fig. 5. Step and harmonic inputs.

For large values of time, x (t ) = Ak x . Hence, x steady-state response of the

Fig. 7. First-order system response to a unit step-input.

To illustrate the above concepts, the response of a thermocouple subjected to a harmonic

Table 6. Thermocouples response to harmonic temperature variation.