Hamouda Assia

Ultrasonic Flow Metering with Highly Accurate
Jitter and Offset Compensation
vorgelegt von
Dipl.-Ing.
Assia Hamouda
geb. in Batna, Algerien
von der Fakultät IV

- Elektrotechnik und Informatik –
der Technischen Universität Berlin
zur Erlangung des akademischen Grades
Doktor der Ingenieurwissenschaften

- Dr.-Ing. -
genehmigte Dissertation
Promotionsausschuss
Vorsitzender: Prof. Dr.-Ing. Friedel Gerfers

1. Gutachter: Prof. Dr. rer. nat. Otto Manck
2. Gutachter: Prof. Dr.-Ing. Roland Thewes
3. Gutachter: Prof. Dr.-Ing. Nour-Eddine Bouguechal
Tag der wissenschaftlichen Aussprache: 12. Mai 2016
Berlin 2017
Acknowledgements
First and foremost, all thanks and praises are to God the Almighty for his blessing which
made this work possible and allowed for it to be completed.
I would like to express my deepest and most sincere gratitude to my supervisor Prof. Dr.
rer. nat. Otto Manck for his unconditional support, valuable and continuous advice,
guidance and assistance throughout my Ph.D. program, and his very active part in
initiating this work. I am especially thankful for his help with reviewing and his
constructive comments and continuous support during the preparation of this thesis.
Without his strong support and endless patience I would have never been able to complete
this research work.
I would like to thank Prof. Nour-Eddine Bouguechal at the University of Batna, Algeria,
for his kind support, helpful suggestions, and encouragement.
I would like to thank Prof. Roland Thewes, Head of the Institute of Sensor and Actuator
Systems at TU-Berlin for reviewing my Ph.D. work. He has offered many helpful
suggestions and remarks.
I am indebted to Mr. Helmut Manck, the Senior Manager of eonas GmbH, for funding this
research and his commitment to support the development of this project.
I would like to acknowledge Dr. Ruediger Arnold, Boris Joesaar, Dr. Mohamed Lamine
Hafiane and Redouane Djeghader with whom I had interesting conversations.
I also wish to thank Hadjer, Khadra, Souad, and Latifa for their great friendship and for
helping me with various mathematical issues.
I would like to give special thanks to Dr. Peter Grametbauer and Dr. Sabine Seydel for
proofreading my thesis and helpful suggests and comments.
I would like to give my special thanks to my family, my lovely mother, my father, my
brother, my sisters, my mother in law, my brothers in law, and sisters in law. My warmest
thanks go to my dearest husband for his encouragement and patient love that inspired me
to complete this work and also to my little angel Zinedine.
Kurzfassung
Diese Dissertation schlägt eine neue Methode zur Messung des Wasserdurchflusses mit
einem Durchlaufzeit-Ultraschall-Durchflussmessgerät vor. Die entwickelte Methode
ermöglicht einem Ultraschall-Durchflussmesser eine genauere Erfassung von sehr
niedrigen Durchflussraten. Flüsse von weniger als zwei Liter pro Stunde (2 l/h) in einem
typische Haushaltswasserzähler sind möglich. Der Fluss einer gegebenen Flüssigkeit in
einem Rohr wird durch die Laufzeitdifferenzmessung von Ultraschallsignalen mit und
gegen den Fluss ermittelt. Je geringer der Durchfluss ist, desto kleiner ist die
Laufzeitdifferenz. Die Differenz liegt bei niedrigen Durchflussraten im Bereich von
wenigen Pikosekunden. Die vorgeschlagene Methode erlaubt das Messen der Differenz im
Bereich von wenigen Pikosekunden und überwindet technische Schwierigkeiten anderer
Messmethoden.
Die piezoelektrischen Wandler sind die kritischen Komponenten des Ultraschall-

Durchflussmessers. Sie können die Genauigkeit der Ultraschall-Durchflussmesser
erheblich beeinträchtigen. Die Wahl einer geeigneten analytischen Funktion für die
Beschreibung des Verhaltens des piezoelektrischen Wandlers (Transducers) ist notwendig
zur Ermittelung eines geeigneten Durchflussmessverfahrens, welches in der Lage ist,
genaue und robuste Messergebnisse zu liefern. Im Ersatzschaltbild wird ein Transducer
durch einen Oszillator mit parallel geschalteter Kapazität dargestellt. Die einfachste
analytische Lösung der entsprechenden Differentialgleichung ergibt sich, wenn der
Transducer mit einer Sinusfunktion angeregt wird. Im ersten Moment reagiert der
Transducer mit einer Schwingung bei seiner eigenen Resonanzfrequenz, die aber nach
einiger Zeit abklingt. Danach schwingt der Transducer nur noch mit der aufgezwungenen
Frequenz. Wartet man also lange genug, dann wird der transiente Teil abklingen und man
wird den stationären Zustand erreichen, wo der Transducer nur noch mit der
Zwangsfrequenz schwingt.
Das vorgeschlagene Verfahren beruht darauf, die Differenz der Laufzeiten indirekt zu
messen, indem die Phasendifferenz zwischen den stationären Teilen der empfangenen
Signale in der stromaufwärtigen und der stromabwärtigen Richtung berechnet wird und
indem eine Sinus-Fitting-Technik mit kleinstem quadratischen Fehler verwendet wird. Dies
verringert den Effekt des Jitters in der Laufzeit. Der Jitter begrenzt die Messgenauigkeit
bei sehr geringer Strömungsgeschwindigkeit.
Der letzte Teil der Arbeit untersucht das Offset-Verhalten. Der Offset ist die Abweichung
der Differenz der Laufzeiten von Null bei nicht-fließendem Wasser. Er ist u. a.
temperaturabhängig. Auch einige Parameter der Transducer sind temperaturabhängig, in
erster Linie die Resonanzfrequenz selbst. Bei einer Erwärmung von 20°C auf 80°C
verändert sich der Offset entsprechend und z.B. erreicht Werte um 300 ps, wenn er vorher
bei 20°C auf „Null“ abgeglichen wurde. Die Messanordnung erlaubt eine neue, bisher in
der Literatur nicht bekannte Art des Offset-Abgleichs durch Anpassung der
Zwangsfrequenzen bezogen auf die Temperatur des Mediums. Die Langzeitstabilität der
zur Einstellung der Offsetdrift verwendeten Zwangsfrequenz wurde bei verschiedenen
Temperaturen experimentell nachgewiesen.
Die erhaltenen Messergebnisse verdeutlichen die Genauigkeit und Robustheit des

vorgeschlagenen Verfahrens: die Differenz der Laufzeiten zeigt im Temperaturbereich von
20°C bis 80°C bei nicht-fließendem Wasser einem Peak-to-Peak-Jitter von nur 15 ps und
einen Offset von weniger als 5 ps. Dadurch kann man im Vergleich zu früheren Techniken
kleinere Durchflüsse messen.
Diese Arbeit liefert Ansätze für mögliche zukünftige Ultraschall-Durchflussmesser mit

hoher Genauigkeit. Um die Ansätze kommerziell gut in zukünftigen Durchflussmessern
nutzen können, bedarf es einer Integration der Messtechnik in einem integrierten
Schaltkreis.
Abstract
This thesis proposes a new method for measuring water flow with a transit time ultrasonic
flow meter device. The developed method allows the ultrasonic flow meter to reach a
better performance than currently available commercial flow meters by accurately
detecting very low flow rates of less than two liters per hour (2 l/h) in a typical household
water meter. In principle, the flow velocity of a given liquid is obtained by measuring the
transit times of an ultrasonic signal in the upstream and downstream directions. The
difference between the transit times is directly proportional to the flow velocity. However,
the fainter the flow is, the smaller the transit time difference (TTD) is. This difference can
be as low as a few picoseconds, which gives rise to many technical difficulties in
measuring such a small time difference with a given accuracy.
The piezoelectric transducers are critical components in ultrasonic flow meters since they
can significantly affect the accuracy of the ultrasonic flow meters. Choosing an appropriate
analytic function that describes the behavior of the basic part of a piezoelectric transducer
proved to be essential for defining a suitable flow measurement method that yields
accurate and robust measurement results. The electrical equivalent circuit of a transducer is
represented by an oscillator connected to a parallel capacitance. The simplest analytical
solution of the corresponding differential equation is obtained when the transducer is
excited by a sinus function. First, the transducer oscillates at its own resonant frequency,
and its oscillations die away after some time. If we wait long enough, the transient part
dies out and what is left is the steady-state part, where the transducer oscillates at the
forced frequency. The proposed method relies on measuring the TTD indirectly by
computing the phase difference between the steady-state parts of the received signals in the
upstream and downstream directions and by using the least squares sine-fitting technique.
This reduces the effect of the TTD-jitter of the measurement, which limits the
measurement accuracy at very low flow velocity.
The last part of the work addresses the issue of the TTD-offset, which refers to any
deviation of the TTD from zero at no-flow conditions. The behavior of the TTD-offset is
investigated over a temperature range. Some parameters of the transducer, such as
resonance frequency, are temperature-dependent. When the temperature of the medium
around the transducers increases by 80°C from ambient temperature, the TTD-offset
(adjusted to zero at ambient temperature) changes accordingly and reaches approximately
300 ps. The novel proposed approach allows the compensation of the TTD-offset by
adjusting the forced frequency with respect to the temperature of the medium. The long-
term stability of the driving frequency used to adjust the TTD-offset drift has been
experimentally proved at different temperatures. The obtained measurement results
illustrate the accuracy and robustness of the proposed method since the TTD is measured at
no-flow conditions, with a peak-to-peak TTD-jitter as low as 15 ps and the TTD-offset less
than 5 ps within a temperature range from ambient temperature to 80°C. This allows to
reach a smaller minimum detectable flow in comparison to previously developed
techniques.
This work offers some suggestions to design an ultrasonic flow meter with high accuracy
in the future. However, the commercial aspect of the future flow meter requires an
integration of the proposed measurement technique in an integrated circuit.
Abbreviations and Symbols
Abbreviations
ADC Analog-to-Digital Converter
CSV Comma-separated values
LSSF Least Squares Sine-Fitting
OPAMP Operational Amplifier
SNR Signal-to-Noise Ratio
STD Standard Deviation
TR Transducer
TTD-offset Transit Time Difference-offset
TTD-jitter Transit Time Difference-jitter
Symbols
A Pipe cross section [m2]
c Speed of sound [m s-1]
C Spring compliance [m N-1]
Cs Serial capacitance [F]
Cp Parallel capacitance [F]
d Viscous damping coefficient [N s m-1]
D Pipe inner diameter [m]
DR ADC dynamic range [dB]
fp Parallel frequency [Hz]
fs Serial frequency [Hz]
fr Resonance frequency [Hz]
fdr Driving frequency [Hz]
ff Forced frequency [Hz]
F0 Applied force [N]
I Current [A]
k Stiffness coefficient [N m-1]
K Water compressibility [m2 N-1]
L Transducer separation [m]
Ls Serial inductance [H]
m Mass [Kg]
N Sample size
Nbit ADC resolution [bits]
Q Flow rate [m3 s-1]
q Charge [C]
qh(t) Homogeneous solution [C]
qp(t) Particular solution [C]
Rs Serial resistance [Ω]
R Pipe Radius [m]
tup Upstream traveling time (against the flow) [s]
tdown Downstream traveling time (with the flow) [s]
T Disk thickness [m]
TTD Transit time difference [s]
Temp Temperature [°C]
v Flow velocity [m s-1]
vmin Minimum flow velocity [m s-1]
V Voltage [V]
VFRS Full scale voltage range [V]
VLSB Least significant bit [V]
VRD1 First direction receiver voltage [V]
VRD2 Second direction receiver voltage [V]
VTD1 First direction transmitter voltage [V]
VTD2 Second direction transmitter voltage [V]
x Displacement [m]
 Water density [Kg m-3]
up Upstream phase [rad]
down Downstream phase [rad]
 Standard deviation of the transit time difference [s]
Standard deviation error of the transit time difference [s]
 Damping coefficient
ωf Angular forced frequency [rad s-1]
ωr Angular resonance frequency [rad s-1]
Content
Chapter 1: Introduction ..................................................................................................1

1.1. Motivation .........................................................................................................1
1.2. State of the Art ..................................................................................................2
1.3. Main Work Objective ........................................................................................3
1.4. Thesis Outline ...................................................................................................3
Chapter 2: The Principles of Ultrasonic Flow Meter ...................................................5

2.1. Introduction .......................................................................................................5
2.2. Doppler Flow Meter ..........................................................................................5
2.3. Transit Time Flow Meter ...................................................................................6
2.3.1. Transit Time Flow Meter Configuration ...................................................6
2.3.2. The Principle of Transit Time Flow Meter ................................................8
Chapter 3: Ultrasonic Piezoelectric Transducer Theory and Modeling ...................10

3.1. Introduction .....................................................................................................10
3.2. Piezoelectric Effect .........................................................................................10
3.3. Ultrasonic Piezoelectric Transducer ................................................................ 11
3.3.1. Piezoelectric Disk ....................................................................................12
3.3.2. Piezoelectric Backing and Matching Layer.............................................13
3.4. Piezoelectric Transducer Modeling .................................................................13
3.5. Butterworth-Van Dyke Model .........................................................................18
3.6. Transmitted and Received Signals Modeling ..................................................24
3.6.1. Transmitter............................................................................................... 26
3.6.2. Receiver ...................................................................................................27
3.7. Quiet Region Aspect .......................................................................................29
3.8. Characteristic Frequencies of Piezoelectric Transducer .................................32
Chapter 4: Determination of the Transit Time Difference (TTD) ............................ 33
4.1. Introduction .....................................................................................................33
4.2. Experimental Setup for Measuring the TTD ...................................................34
4.3. Methodology for the Calculation of TTD ......................................................36
4.4. Least Squares Sine-Fitting .............................................................................37
4.5. Estimation of the Quiet Region ......................................................................38
4.6. Algorithm for Computing the TTD .................................................................40
Chapter 5: TTD-jitter Reduction .................................................................................42

5.1. Introduction ....................................................................................................42
5.2. The Effect of the Sampling Jitter on TTD.......................................................43
5.3. Sources of TTD-jitter ......................................................................................44
5.4. TTD-jitter Reduction Technique .....................................................................45
Chapter 6: TTD-Offset Cancellation ...........................................................................50

6.1. Introduction .....................................................................................................50
6.2. Sources of TTD-offset .....................................................................................50
6.3. The Impacts of Temperature on the Transducer Resonance Frequencies .......51
6.4. Zero Flow TTD-offset Correction ...................................................................53
6.5. Temperature Dependence of the TTD-offset Calibration................................ 57
Chapter 7: Results and Analyses ..................................................................................60

7.1. Introduction .....................................................................................................60
7.2. Hardware Design ............................................................................................. 61
7.3. Software Algorithm .........................................................................................62
7.3.1. Transmission of the Sinusoidal Burst ......................................................63
7.3.2. Digitizing of the Two Direction Signals ..................................................63
7.3.3.TTD-offset Cancellation Algorithm .........................................................64
7.4. TTD-offset Compensation Evaluation ............................................................ 66
7.5. The Effect of Temperature on TTD-offset Measurements .............................. 68
7.6. Discussion .......................................................................................................70
Chapter 8: Conclusion ..................................................................................................72

Chapter 1
Introduction
1.1. Motivation
Flow measurement is often critical in the household sector of the domestic economy [1].
Therefore, in order to make a profit rather than run at a loss, it is imperative to accurately
measure what flows through the measuring pipe under all circumstances.
Ultrasonic flow meters have been used successfully in industrial applications for several
decades [2-3]. They have gained approbation in wide metering applications such as dirty or
clean water, petrochemical products, natural gas, and so on. This is due to the significant
operational and economic advantages that the ultrasonic flow meters offer in contrast to
conventional meters. From the economic point of view, they are easy to install,
inexpensive, and require less maintenance than other meters such as mechanical flow
meters, which need to be checked periodically [2]. In terms of operation, ultrasonic flow
meters can be highly sensitive and accurate, and they typically have a broader flow rate
range, good measurement repeatability, and bi-directional flow capability.
As any other measurement technology, ultrasonic flow meters have their measuring
limitations, which are mostly caused by changes in the temperature of the medium and
several other factors, such as water viscosity and compressibility [4-7]. If there is no flow
in the measuring pipe (i. e., still water in the pipe), the upstream and downstream transit
times of such meters are equal and the difference between the two transit times should be
negligible. However, this may not always be the case since the transit time difference
(TTD) can be influenced by jitter and offset. Therefore, any delay offset between the
upstream and downstream transit times directly translates into a zero flow error. This zero
flow TTD-offset limits measurement accuracy at low flow velocities. The lower this error,
the higher the accuracy of the flow meter.
Chapter 1: Introduction 2
1.2. State of the Art
Following [7], the accuracy of the ultrasonic flow meter at no-flow conditions - for water
or gas applications - depends essentially on the reciprocity of the electro-acoustic
measurement system. This reciprocity can be achieved by perfect transducer symmetry or
perfect electrical symmetry in the ultrasonic flow meter system. This means that either the
impedances of the electric loads for both transmitting and receiving transducers are equal
or both transducers are identical. The electro-acoustical reciprocity principle as referenced
by Per Lunde’s paper [8-16] provides possibilities for reducing or even neglecting the need
for the zero flow calibration of ultrasonic flow meters for both liquid and gas applications.
In 2010, Borg Johan presented in his Ph.D. thesis "On electronics for measurement
systems" [17-18] a new methodology based on driving the transducer with a current source
rather than a voltage source in order to achieve good impedance matching between
transmitting and receiving circuits. This method achieved a significant improvement in
reducing the zero flow error compared to Lunde’s work.
What followed was the state-of-the-art publication of Yang Bo in March 2011 [19], when
he presented a new approach to improve the accuracy and the stability of ultrasonic
transducer flow metering over a wide temperature range under non-reciprocal operation
conditions. This approach is based on driving both transducers at a specific frequency
outside their resonance frequencies through a sinus burst in order to eliminate the effect of
temperature dependence of the resonance frequency and therefore reduce the long-term
drift of the transit time difference measurements, which is caused by temperature
variations. The drawbacks of this method are, on the one hand, the reduced signal-to-noise
ratio (SNR) caused by working outside the resonance frequency of the transducer and, on
the other hand, the fact that the excitation voltage level must be high even in the presence
of an amplifier. This results in a high power consumption of the whole measurement
system which presents a big disadvantage in battery applications. Nevertheless, this
approach achieved improved results compared with the two previously mentioned
methods.
1.3. Main Work Objectives
The main objective of this work is to develop a new methodology, which combines
hardware with appropriate digital evaluation software algorithms and which can overcome
the effect of the TTD-jitter noise and the TTD-offset drift. This methodology reduces the
false flow detection and improves the accuracy of the ultrasonic flow meter at low and
no-flow conditions. The first part of this research work focuses on the analytical modeling
or the mathematical presentation of the piezoelectric transducer, which is based on the
driven damped harmonic oscillator system. This model provides a better understanding of
the behavior of the piezoelectric transducer and provides the necessary concepts that allow
to measure the TTD very accurately.
The main contribution of this work is summarized by the two main steps that were
undertaken to minimize the effect of zero flow error and TTD-jitter on the measurement
results:
 Develop a software algorithm on the basis of the proposed approach, which would
be capable to effectively reduce the TTD-jitter.
 Develop another software algorithm on the basis of the proposed approach, which
would be capable to continuously correct the zero flow TTD-offset.
1.4. Thesis Outline
A significant amount of the current research focuses on ultrasonic flow meters, especially
in the field of measuring of flow rate in the transmission of gas through pipelines, in order
to develop more accurate measurement methodologies. We will elaborate and refer to the
appropriate theoretical background when describing our experiments. This dissertation is
organized as follows:
1. Chapter 1 outlines the motivation behind this work and positions our work amongst
the recent research efforts in ultrasonic flow meter measurement methodologies
currently available.
2. Chapter 2 describes the available ultrasonic flow meters used for measuring the
flow rate of a flowing fluid.
3. Chapter 3 presents the theoretical background of a piezoelectric transducer. We

address the relevant theories to derive an analytic model of the transducer that can
describe its behavior in both cases, transmitter and receiver.
4. Taking into account the analytic transducer model and all the aspects derived
through the mathematical representation of the transducer presented in Chapter 3,
Chapter 4 presents the analysis and evaluation of our proposed method can be used
to calculate the transit time difference. It also presents a detailed description of each
of the measurement setup system’s blocks.
5. Chapter 5 describes the applied TTD-jitter analysis and methodology used to

understand the causes of TTD-jitter. The proposed TTD-jitter reduction technique is
also described.
6. The zero flow TTD-offset correction technique is the main subject of Chapter 6,
starting with a study of the TTD-offset sources and ending with the proposed
approach to mitigate this TTD-offset.
7. In Chapter 7, the experimental results are presented and analyzed.
8. Finally, conclusions are drawn in Chapter 8.

Chapter 2
The Principles of Ultrasonic Flow Meter
2.1. Introduction
Most ultrasonic flow meters use one of the two main principles: Doppler effect or Transit
Time Difference. When a fluid is in motion with a certain velocity, the flow meter
measures the flow of this fluid either by calculating the difference between the two
traveling times of an ultrasonic signal propagating with and against the flow direction or by
measuring the frequency shift using the Doppler principle.
2.2. Doppler Flow Meter
This type of flow meter is based on the Doppler principle discovered in 1842. Typically,
one transducer is fitted in the pipe wall as shown in Figure 2.1. It continuously transmits an
ultrasonic signal at a constant frequency f1 into the flowing fluid. The particles inside the
fluid reflect the transmitted signal, and their movement shifts the frequency of the
ultrasonic signal to a frequency f2. The frequency shift is proportional to the speed v of the
particles and hence to the flow. It is given by the following equation [20-22]:
(2.1)
where ∆f is the frequency shift (the difference between transmitted and received
frequencies), θ is the angle of the transmitter and receiver crystal axis with respect to the
pipe axis, c is the sound velocity, and v is the flow velocity.
Chapter 2: The Principles of Ultrasonic Flow Meter 6
Fig. 2.1: Doppler flow meter
Due to many drawbacks and limitations of ultrasonic Doppler flow metering, this method
is now used only in a few specific applications such as for waste water that contains dirt
particles or gas bubbles. It has been replaced by the transit time ultrasonic flow metering
because in addition to flow rate measurement this method can also provide information on
the type of liquid and the working temperature on the basis of sound velocity measurement
[20].
2.3. Transit Time Flow Meter
2.3.1. Transit Time Flow Meter Configuration
The ultrasonic transit time flow meter consists of one pair of transducers facing each other,
which are separated by a known distance. The transducers are mounted according to
different geometries depending on the application. For instance, the in-line configurations
shown in Figures 2.2 and 2.3, as well as the configuration with reflectors shown in
Figure 2.4, are most often used in applications where the diameter of the pipe is less than
25 mm, whereas the diagonal configuration shown in Figure 2.5 is used in applications
where the diameter of the pipe is up to 10 meters [4].
Two flow meter pipes have been used in this work. The configuration of the first flow
meter pipe contains one pair of 4 MHz ultrasonic transducers (Figure 2.2). The distance
between the two transducers and the inner radius of the flow meter body are L = 42.2 mm
and R = 4 mm, respectively. The configuration of the second flow meter pipe contains one
pair of 1 MHz ultrasonic transducers with L = 49.5 mm (Figure 2.3). All experimental
work has been performed using these meters.
Fig. 2.2: Flow meter pipe with in-line transducers: (1) inlet, (2) cable connector,
(3) the upstream transducer, (4) flow meter body, (5) the downstream transducer,
(6) outlet
Fig. 2.3: Flow meter pipe with in-line transducers
Fig. 2.4: Flow meter pipe with reflectors
Fig. 2.5: Flow meter pipe with diagonal transducers

2.3.2. The Principle of Transit Time Flow Meter
The transit time flow meter is based on the transit time difference (TTD) principle and uses
two transducers. Each transducer can alternately transmit and receive an ultrasonic signal.
This signal is generated when a piezoelectric crystal is subjected to an alternating voltage.
Conversely, the piezoelectric crystal generates voltage when the ultrasonic signal impacts
the transducer. In the case of simultaneous excitation, the two transducers emit and receive
the ultrasonic signals at the same time. One ultrasonic signal travels through the pipe in the
direction of the flow (downstream direction) and the other against the flow (upstream
direction). Every signal needs a certain period of time (called transit time) passes before
the signal is received by the opposite transducer. This transit time depends on three
parameters: the speed of sound c, the ultrasonic path length L, and the flow velocity v as
illustrated by the following equations [23-24]:
(2.2)
(2.3)
where tup is the upstream transit time, and tdown is the downstream transit time.
If there is no flow, then:
. (2.4)
At no-flow conditions the transit times are equal. Once the fluid starts to flow, the sound
wave moving with the flow travels faster than the sound wave moving against the flow.
The difference between the two transit times is directly proportional to the flow velocity.
This can be mathematically expressed as follows:
(2.5)
Since flow velocity v is much smaller than the speed of sound c, it can be derived from
(2.5) as follows:
(2.6)
Since the internal cross-section of the pipe is known, the volume flow rate Q is determined
by the following formula:
, (2.7)
where A is the inner circular cross-section, and R is the inner radius of the flow meter pipe.
The ratio of t to the traveling time t0 measured at no-flow conditions is given by:
. (2.8)
Equation (2.8) can be used to calculate the needed measurement accuracy. For instance, the
minimum flow vmin calculated for a given flow rate of two liters per hour (2 l/h) and pipe
diameter of 0.8 cm is about vmin = 10 mm/s, or approximately 7 ppm compared to the
sound velocity (the speed of sound in pure water and room temperature is about
c = 1500 m/s). Therefore, according to Equation (2.5), this minimum flow variation of
10 mm/s produces a transit time difference value of about t = 380 ps, provided that the
travelling time of the used flow meter pipe is 28 µs (calculated for L = 42.2 mm using
Equation (2.4)). Therefore, to achieve an accuracy of 5 %, the desired ultrasonic flow
meter must be able to accurately measure the transit time difference of at least 20 ps (which
then would be its minimum measured value).
This thesis focuses on the transit time flow meter mostly because of its extensive industrial
usage [5]. Besides, it has the highest cost efficiency and, unlike the Doppler flow meter,
does not require the fluid to contain particles or air bubbles in order to reflect the ultrasonic
sound.
Chapter 3
Ultrasonic Piezoelectric Transducer Theory and
Modeling
3.1. Introduction
An accurate description of the behavior of ultrasonic transducer’s active element requires a

detailed investigation of the transmitted acoustic wave, which travels through the pipe, and
of the received acoustic wave, which is picked up by the opposite transducer after a
predetermined time.
This chapter starts with an overview of the piezoelectric effect, which is followed by a
description of the real transducer geometry and its different composite layers. Thereafter,
an analytical approach of the piezoelectric transducer developed through the solution of the
differential equation of driven damped harmonic oscillator, is analyzed in detail.
In the remainder of this chapter, the emulated transmitted and received waveforms, which
are generated using Matlab, are compared to the real signals obtained experimentally. We
show that a full agreement between the theoretical description and experimental signals
can be achieved with an appropriate choice of only three model parameters: resonance
frequency, damping factor, and excitation frequency.
3.2. Piezoelectric Effect
The word "piezo", of Greek origin, means "push". According to [25], the effect known as
piezoelectricity is a property exhibited by certain classes of crystalline materials that
consist of polarized molecules. It was discovered by the brothers Pierre and Jacques Curie
in 1880. When a piezoelectric material is subjected to a mechanical stress, it generates an
electrical charge, which is proportional to the applied stress. This behavior is called the
direct piezoelectric effect. Inversely, when the piezoelectric material is subjected to an
Chapter 3: Ultrasonic Piezoelectric Transducer Theory and Modeling 11
electric field, it changes dimension and becomes strained [25]. This strain is again
proportional to the applied field. hese two effects are shown in igure 3.1.
Besides quartz and Rochelle salts, many other piezoelectric materials are available
nowadays, such as Barium Titanate (BaTiO3), Lead Metaniobate (PbNb2O3) and Lead
Zirconate Titanate (PZT) [26-27].
Fig. 3.1: Piezoelectric effects: (a) direct, (b) inverse [25]
3.3. Ultrasonic Piezoelectric Transducer
The ultrasonic piezoelectric transducer is employed to convert electrical energy into

mechanical energy (sound wave) and vice versa. The piezoelectric material has the
following three properties [25]:
• the elasticity property which defines the mechanical aspect of the material
• the piezoelectric property which defines the electromechanical aspect of the
material
• the dielectric property which defines the electrical aspect of the material.
An ultrasonic transducer consists of three main parts as shown in Figure 3.2: piezoelectric
disk, the rear part or backing layer, and front or matching layer [28-29]. Many factors
determine the performance of ultrasonic transducers, the most important of them being the
material properties of the transducer components, including housing and connections, the
external mechanical, and electrical load conditions and damping.
Fig. 3.2: Typical construction of an ultrasonic transducer [29]:

(1) matching layer, (2) piezoelectric disk, (3) backing layer,
(4) cable connector, (5) transducer housing, (6) electrodes
3.3.1. Piezoelectric Disk
The piezoelectric disk is the main component for generating and receiving the ultrasonic
wave and the active element of the ultrasonic transducer. The angle at which the
piezoelectric crystal is cut in relation to its crystallographic axes defines its vibration mode
and the type of the generated ultrasonic wave, which can be a longitudinal or a shear wave
(Figure 3.3), depending on the application. For instance, in our application, the ultrasonic
wave propagates through the water by a longitudinal motion (compression/expansion) as
the fluid does not support the shear motion [30]. The surface of the disk moves up and
down, and the water has to follow this movement directly.
Fig. 3.3: Response of a piezoelectric element to an AC voltage:

(a) compression motion generating longitudinal waves,
(b) transverse motion generating shear waves
The active element of the transducer shows a typical resonance frequency whose value is
precisely related to the size and shape of the piezoelectric transducer according to
Equation (3.1).
(3.1)
where fr is the resonance frequency, T is the disk thickness (see Figure 3.4), and c
represents the wave velocity of the longitudinal vibration inside the disk and depends on
the acoustic properties of the disk.
Fig. 3.4: Piezoelectric disk with thickness T and diameter D [29]
3.3.2. Piezoelectric Backing and Matching Layer
The backing layer consists of a high-density material and is used to control the vibration by
absorbing the energy radiating from the back face of the active element. The piezoelectric
materials are characterized by high acoustic impedance in comparison to water and air.
Consequently, the bandwidth of the response function of the disk is low. The acoustic
impedance mismatch can be overcome by adding a matching layer to enhance the
bandwidth and source sensitivity. More reliable wide bandwidth transducers are obtained
by adding a quarter-wave matching layer [31-33].
3.4. Piezoelectric Transducer Modeling
The transducer may be driven by a voltage burst consisting of a finite number of sinus
cycles with a well-defined frequency, which is chosen near the resonance frequency of the
transducer. Vibrating at the same frequency as the applied voltage, the piezoelectric
material generates a wave, which propagates through the water. In order to understand
exactly how this waveform is generated, in particular the "transient behavior" and "steady-
state behavior", an analytical description is needed to emulate the excitation process.
The transducer, like many oscillators, can be modeled as a harmonic oscillator, which is
characterized by its "resonance frequency" and "damping coefficient". The excitation of the
harmonic oscillator by a voltage causes the oscillator to vibrate at the same frequency as
the applied voltage. It vibrates with greater amplitude if the frequency of the excited
voltage corresponds to the resonance frequency of the oscillator.
In the case of a harmonic excitation, three basic parameters dominate the behavior of the
oscillator’s response: angular resonance frequency ωr, damping coefficient ζ, and angular
forced frequency ωf. However, as the free vibration (transient part) dies out over time, only
the forced frequency dominates the remaining steady-state part. When there is no exciting
signal, the signal dies out gradually, which causes the decay of the signal to zero due to the
transfer of the energy to the water. This switching-off behavior is characterized only by ωr,
ζ.
The driven damped harmonic oscillator model attempts to develop a better understanding
of the ultrasonic transducer. The disk, which represents the transducer’s active element, can
be described by both models depicted in Figure 3.5. During resonance, the series resonant
circuit Rs, Ls, Cs shown in Figure 3.5.b (also called motional branch) forms a damped
harmonic oscillation and resonates in a similar way to the mechanical model shown in
Figure 3.5.a. Therefore, it is possible to compare the two models and use the electrical
parameters (inductance, capacitance, and resistance) to represent the mechanical
parameters (mass, stiffness, and damping) [25]. Hence, some mechanical and electrical
properties such as material density, elastic mechanical parameters, piezoelectricity, and
dielectricity define the values of the motional components. The impact of the parallel
capacitance shown in the electrical model is explained in more detail in the next section.
Fig. 3.5: Equivalent models of the piezoelectric element:

(a) mechanical model, (b) Butterworth-Van Dyke electrical model
For sinus excitation, the basic description of the damped externally driven oscillator is
provided by the following second order differential harmonic equation:
, (3.2)
where:
F = F0/m is the applied external force F0 divided by m,
m is the mass,
ωf is the angular forced frequency of the sinus burst,
d is the viscous damping coefficient,
k is the system stiffness coefficient,
is the damping coefficient, and
is the angular transducer resonance frequency.
The mechanical displacement is analogous to the electrical charge that can be used to
reformulate the previous differential Equation (3.2). The summary of the equivalent
quantities between the mechanical and electrical models is presented in Table 3.1 [34].
Electrical Mechanical
Charge Q Displacement x
Current I = dQ/dt Velocity v
Applied voltage V Applied force F0
Resistance Rs Friction c
Inductance Ls Mass m
Capacitance Cs Spring compliance C = 1/k
Table 3.1: Comparison of equivalent electrical and mechanical resonant circuits
According to Table 3.1, Equation (3.2) of the mechanical motion can be rewritten as:
(3.3)
where:
V is the amplitude of the applied voltage,
RS, LS, and CS are the parameters of the piezoelectric disk,
describes the damping coefficient, and

is the transducer serial resonance angular frequency.
Assuming < 1 (underdamped harmonic oscillator), the complete solution of Equation

(3.3) is given by Equation (3.4). Generally, this function can be divided into two parts: the
solution of the homogeneous differential equation with a zero right side giving the transient
response, and the particular solution of the non-homogeneous equation as the answer to the
externally applied voltage:
. (3.4)
The homogeneous solution of this equation is given by:
(3.5)
where is the damped natural angular frequency, equal to , and the

constants Ah and θ are dependent on the initial conditions.
The particular solution, the forced or stationary solution, is given by:
(3.6)
Ap and can be determined simply if we use the complex form of the differential
Equation (3.3), which can be written as:
(3.7)
Using Moivre’s formula, and Q(t)= , one can write:
. (3.8)
By dividing the previous equation into real and imaginary parts, we end up with the
following equations that correspond to the amplitude Ap and the phase , respectively:
(3.9)
(3.10)
The amplitude of the forced oscillations depends on the difference between the forced
frequency1 of the applied voltage and the resonance frequency of the piezoelectric disk.
The damping coefficient has a strong influence on the maximum amplitude reached when
both forced and resonance frequencies are similar. Figures 3.6 and 3.7 show the typical
behavior of the amplitude and the phase, respectively, when realistic parameters are
applied.
Regarding the amplitude response shown in Figure 3.6, a Gaussian type distribution around
the resonance point can be observed when realistic parameters are applied, especially for
the damping coefficient. Considering the main goal of this work, which is reducing the
TTD-jitter and TTD-offset, this illustration clearly demonstrates the strategy which has to
be applied: If both transducers are driven using a forced frequency, which is in the direct
vicinity of their resonance points, good SNR in the receiving part can be expected. In this
work, the recommended range is ±500 kHz or approximately 10% of the resonance
frequency.
Fig. 3.6: Amplitude of a driven damped harmonic oscillator at ζ =0.02 and

fr = 4.05 MHz
Regarding the phase response shown in Figure 3.7, it can be observed that the phase shift
between the excitation and the mechanical response of the transducer is constant and
depends only on the difference "ωr - ωf". It can be also observed in Figure 3.7 that the
phase tends either towards 0° or towards 180° at frequencies away from the resonance
frequency. This feature would allow to derive a reliable measurement method because in
1
Also called the driving frequency.
these two parts of the phase frequency response the phase shifts are quite steady and no
longer depend on the transducer parameters. From this observation it can be directly
deduced that the temperature dependence of the transducer parameters effects does not
influence the measurements (outside the well-known relationship of the speed of sound
versus temperature). Yang Bo [19], who applied this idea, reported an approximately
200 ps peak-to-peak TTD-jitter and −185.3 ps TTD-offset measured at no-flow conditions
due to drastically reduced SNR.
Fig. 3.7: Phase of a driven damped harmonic oscillator at ζ =0.02 and fr = 4.05 MHz
3.5. Butterworth-Van Dyke Model

The comparison of the mechanical and electrical models shown in Figure 3.5 shows that
the difference between the two models is given by the parallel capacitance Cp, which is
inherent in a transducer where metal electrodes are separated by a disk. With regards to the
experimental setup presented in Figure 3.13, the function generator which can be
represented by a voltage source with output resistance (≈ 50 Ω) is added to the total
electrical excitation circuitry (Figure 3.8). The transmitter transducer is connected to this
voltage source via an external resistance (R1) of a typical value of 200 Ω. In the case of
simultaneous excitation (Figure 4.1), two resistances R1 and R2 are used. The total
symmetry of the two electrical circuits can never be assumed. The values of R1 and R2 can
be different, and also the material parameters of the two disks are clearly not identical. This
difference in the two resistances does not cause a systematic TTD-offset error as long as
both directions are symmetric in terms of impedance.
The harmonic model, which is independent of either single or simultaneous excitation,

does not take into account the parallel capacitance Cp. Only the motional branch of the
transducer's electrical model (Figure 3.5.b) can be represented by the harmonic oscillator
model.
Fig. 3.8: Simulation circuit using the transducer electrical model
Regarding the "Vtrans" node, the AC analysis of the electrical circuit shown in Figure 3.8
results in the frequency response magnitude and phase plots shown in Figures 3.9 and 3.10,
respectively. Hence, these results show a different behavior when compared to the
harmonic model findings depicted in Figure 3.6 and 3.7. Therefore, due to the parallel
capacitance Cp, the AC simulation results obtained with the transducer's electrical model
(Figures 3.9 and 3.10) clearly demonstrate a restricted similarity compared to the findings
obtained with the harmonic oscillator model (Figures 3.6 and 3.7). However, the analytic
evaluation of the two parallel branches gives a rather complicated amplitude and phase
equations (Equations (3.11) and (3.12)) compared to those obtained previously during the
evaluation of the harmonic oscillator (Equations (3.9) and (3.10)). Thus, the transfer
function Vtrans/Vin of the electrical circuit shown in Figure 3.8 has a magnitude and phase
shift, which are respectively given by:
(3.11)
. (3.12)
Moreover, in the transducer electrical model the two parameters ωr and ζ, which
characterize the harmonic oscillator model, are extended to three parameters: ωr, ζ, and Cp.
The amplitude response over frequency (Figure 3.9) displays following qualitative
behavior: At low frequencies, the impedance of the motional branch is extremely high and
the voltage "Vtrans" rises with increasing frequency. As the frequency increases, the
capacitance Cs in the serial branch lowers the resulting impedance, and therefore the
voltage "Vtrans" decreases as well. At the series resonance point the minimum impedance
(Rs) is reached. At only slightly higher frequencies (at the parallel resonance point) the
inductance dominates the resulting impedance of the motional branch. It resonates with Cp,
causing the voltage "Vtrans" to increase. However, as the frequency gets higher, the
parallel capacitance prevails so that "Vtrans" tends towards 0 V.
Fig. 3.9: Amplitude response at "Vtrans" node at fs = 4.05 MHz and fp = 4.3 MHz
The phase behavior shown in Figure 3.10 can be easily explained as follows: At low
frequencies the resulting transducer impedance is given mainly by the capacitance Cs;
hence, when compared to the excitation phase shift, the phase tends towards -90°. At the
point where forced frequency increases and reaches the serial resonance frequency, the
effect of inductance and capacitance compensate each other, and the resulting impedance is
given only by the resistance Rs, which represents the acoustical losses of the transducer to
the surrounding environment and the energy transferred to the water [35]. In this case, the
phase tends to 0°, and the equivalent transducer electrical model is simplified to a parallel
connection of Rs and Cp. However, as the frequency gets high, the inductance dominates
the resulting impedance of the motional branch, and the phase tries to reach +90°. At the
operating point where the frequency becomes higher, the influence of Cp cannot be
avoided. At this point the phase makes a fast transition from +90° to -90°.
Fig. 3.10: Phase response at "Vtrans" node at fs = 4.05 MHz and fp = 4.3 MHz
Figures 3.9 and 3.10 show that the transducer exhibits two resonance frequencies at which
it appears resistive. The first frequency, at which the impedance of the transducer is the
smallest, is called series resonance frequency fs. This resonant first point is created when
Cs resonates with Ls. The second frequency, at which the impedance of the transducer is the
greatest, is called parallel resonance frequency fp. This resonant second point is created
when Ls and Cs resonate with the parallel capacitor Cp. These two frequencies can be
computed as follows:
, (3.13)
. (3.14)
With respect to the last results shown in Figures 3.9 and 3.10, the assumption that the
transducer can be modeled as a harmonic oscillator becomes valid only if the voltage over
Rs is analyzed instead of "Vtrans" since the acoustic energy transmitted by the transducer
to the water is proportional to the energy dissipated in the serial resistance Rs [36]. The
following two figures show the amplitude and the phase at node "Vr", respectively. There
is a direct correspondence between the harmonic oscillator responses shown in Figures 3.6
and 3.7 and "Vr" responses presented in Figures 3.11 and 3.12. To sum up, these curves
clearly demonstrate that the mathematic modeling based on the driven damped harmonic
oscillator can govern the transducer's behavior.
Fig. 3.11: Amplitude response at "Vr" node at fs = 4.05 MHz and fp = 4.3 MHz
Fig. 3.12: Phase response at "Vr" node at fs = 4.05 MHz and fp = 4.3 MHz
Even under the condition that the magnitude and phase responses at the "Vtrans" node do
not demonstrate the behavior of a driven damped harmonic oscillator, the "Vtrans" signal
can be separated into three different regions on the basis of the harmonic oscillator
modeling aspects and according to the measured examples depicted in Figures 3.16 and
3.17.
• The first region contains both the homogeneous and the particular solution because
the particular solution always represents a sine wave, and any deviations from sinus
oscillations are caused by the homogeneous solution.
• The next region is characterized by the pure sinus wave, i.e., the homogeneous
solution dies out because of the damping, and the remaining part is presented by the
particular solution.
• The last region can be observed if the excitation is switched off. Hence, the
homogeneous solution is active. This region is characterized by damped
oscillations.
In this work, the transit time difference (TTD) measurements are restricted only to the
second region (in further text quiet region or steady-state region). This region is
characterized by the pure sinus of the particular solution. In this region the influence of Cp
cannot be avoided. Hence, the current through Cp is given by:
(3.15)
For a pure sinus burst, this current can be represented by a sinus oscillation as follows:
(3.16)
The total current flowing into the two branches of the circuit shown in Figure 3.8 can be
calculated as the sum of the current flowing through Cp and the current flowing into the
motional branch:
(3.17)
where,
(3.18)
. (3.19)
The influence of the parallel capacitance on the second region, which is characterized by
the particular solution of forced oscillation, results in a simple phase shift. Hence, instead
of 0, a slightly changed value 2 can be detected. The amplitude undergoes a similar

change. Since the transit time difference is computed as the phase difference between the
two received signals2, the changes of the amplitude do not affect the measurement
accuracy.
3.6. Transmitted and Received Signals Modeling

Numerical solutions of the motion equations are obtained using Matlab in order to emulate
the measured transmitted and received signals. These waveforms are subsequently
compared to the experimentally measured ones. The measurement principle used in the
experiments is shown in Figures 3.13 and 3.14. The transducer TR1 is excited through the
resistance R by a sinusoidal burst in order to generate an ultrasonic signal that can be
received by the second transducer TR2 after traveling through the water within the distance
that separates the two transducers.
Fig. 3.13: Experimental setup used to generate and detect the ultrasonic waves
Fig. 3.14: The experimental setup
2
For further details see Chapter 4.
Figure 3.15 shows a screenshot of Digilent Analog Discovery digital oscilloscope

displaying the transmitted and received signals measured with 1 MHz flow meter pipe.
Assuming that the sound velocity at ambient temperature is 1500 m/s, the sound wave
traveling time calculated for an ultrasonic path length between TR1 and TR2 of L 49.5 mm
is about 33 µs.
Fig. 3.15: A screenshot of the transmitted and received waveforms
The two channel measured waveforms demonstrate a loss in the amplitudes between
transmission and reception. The 1 MHz flow meter pipe shows that the ultrasonic received
signal is attenuated by a factor of 2.5 compared to the transmitted signal, whereas for the
4 MHz flow meter pipe the attenuation factor becomes nearly 10 (Figure 3.22).
3.6.1. Transmitter
From the modeling point of view, the transmitted signal can be divided into two parts. The
first part is characterized by the fact that the homogeneous and the particular solution are
both active. The second part is provided only by the homogeneous solution since the
excitation is switched off. Hence, the signal dies off gradually since there is no driving
force. According to our measurements, the decay of the amplitude lasts about 10 to 15
periods.
Regarding the particular and the homogeneous solution (Equations (3.5) and (3.6),
respectively), three parameters - transducer resonance frequency ωr, damping coefficient ζ,
and the angular driving frequency ωf - are used in order to emulate the transmitter signal.
Figure 3.16 shows the simulated as well as the measured transmitted waveforms for
27 periods of uniform sinusoidal applied voltage. The simulation is performed at a
damping coefficient ζ = 0.05, a resonance frequency fr = 1 MHz (fr = 2/ωr), and a forced
frequency ff = 1 MHz (ff = 2/ωf), whereas the measured signal is obtained at ff = 1 MHz.
The second example shown in Figure 3.17 is obtained by changing only the forced
frequency ff from 1 MHz to 1.1 MHz for the simulated and for the measured signals.
The differences caused by the parallel capacitance can be observed, especially at the
beginning when the excitation starts and in the moment when the excitation is switched off.
It can be seen that the blue and red curves look very similar outside the switching-on and
-off behavior. A 10% change in the driving frequency provides a remarkable difference in
the shape between the two waves depicted in Figures 3.16 and 3.17.
Fig. 3.16: Transmitter response for a 1 MHz flow meter pipe and L = 49.5 mm
at ff = 1 MHz: (a) emulated waveform, (b) measured waveform
Fig. 3.17: Transmitter response for a 1 MHz flow meter pipe and L = 49.5 mm
at ff = 1.1 MHz: (a) emulated waveform, (b) measured waveform
3.6.2. Receiver
Assuming that the transmitted signal, which is directed towards the second transducer, is
defined by the complete solution of Equation (3.4) (the homogeneous and the particular
solutions), this complete solution is used to replace the second part of the second order
differential harmonic equation (Equation (3.3)). The following equation can be used to
emulate the receiver waveforms:
(3.20)
According to the numerical solution of Equation (3.20) obtained using Matlab, five
parameters are now used in order to emulate the ultrasonic receiver signal: two parameters
of each of the two transducers ωr1,2, ζ1,2, and the angular forced frequency ωf. Hence, the
receiver waveform is characterized by two intervals. During the first time interval, the
external force is given by "Qh(t) + Qp(t)", while during the second interval the external
force is given by only Qh(t). Figure 3.18 shows the result of the simulated receiver results
achieved for the parameter values ζ1 = 0.01, ζ2 = 0.02, fr1 = 1.01 MHz, fr2 = 1.02 MHz. The
external force is activated at the same frequency of 1 MHz for 33 cycles for both simulated
and measured signals. The shapes of both waveforms look very similar. The measured
curve shows a difference at the end after roughly 103 µs, where the first reflected signal
already approaches the received signal, causing interference between both signals.
Fig. 3.18: Receiver response for a 1 MHz flow meter pipe and L = 49.5 mm
at ff = 1 MHz: (a) emulated waveform, (b) measured waveform
Fig. 3.19: Receiver response for a 1 MHz flow meter pipe and L = 49.5 mm
at ff = 1.02 MHz: (a) emulated waveform at ζ1 = 0.01, ζ2 =0.018, fr1 = 1 MHz and
fr2 = 1.02 MHz (b) measured waveform
By comparing the measured receiver and transmitter signals, it can be deduced that the
switch-on and the switch-off behavior of the excitation, which creates spikes in the
ultrasonic transmitted signals, is no longer seen in the received ones. Therefore, in contrast
to the transmitter, the receiver is activated and deactivated rather smoothly. The flowing of
the current over Cp does not produce any signal spikes since the ultrasonic transmitter
signal impacts the receiver transducers more or less continuously. The influence of Cp is
correspondingly small.
In summary, the model of the harmonic oscillator is found to be the most effective to
describe and analyze the measured curves. The different physical waveforms in Figures
3.16 to 3.19 can be easily emulated by an appropriate choice of three parameters (the
transmitter transducer parameters ωr, ζ, and the angular driving frequency ωf) in the case of
the transmitted signal, and five parameters (the two transducers parameters ωr1, 2, ζ1, 2, and
the angular forced frequency ωf) in the case of the received signal. Obviously, the
requirement specified previously - an analytical description of the resulting signals - is
fulfilled by modeling the piezoelectric transducer as a driven damped harmonic oscillator.
3.7. Quiet Region Aspect

After the decay of the transient part of the ultrasonic signal, the resulting signal is
characterized by a constant phase, uniform amplitude, and a frequency which matches the
forced frequency. Hence, the mathematic representation of this signal is given by the
particular solution equation3. As mentioned previously, the evaluation of the measured
curves has to be restricted to this "quiet" time interval. Therefore, it becomes necessary to
analyze quantitatively the influence of the homogeneous solution (transient response):
what is the minimum number of cycles required in a burst excitation to make sure the
transient region dies off completely and the steady-state is reached? This can be determined
by estimating the frequency variation through the transmitted or the received signals using
consecutive zero-crossings. The good agreement between the emulated and the measured
curves (Figures 3.16 - 3.19) allows to use of the former ones to perform the frequency
variation estimation. Figure 3.20 illustrates the frequency variation, which is performed on
the basis of the emulated transmitter signal example for ff = 1.02 MHz, ζ = 0.05, and fr =
1 MHz, and using the analysis of the zero-crossings.
3
See Equation (3.6).
Fig. 3.20: Frequency variation through the transmitted signal
The zero-crossings are calculated by the linear interpolation between every pair of
successive positive and negative signal values and their corresponding time on the x-axis.
As a result, the homogeneous solution or the resonance effect dies out after about
25 periods, which allows the transducer to reach the so-called quiet region or the steady-
state region [37]. In order to achieve high accuracy of less than 20 ps peak-to-peak TTD-
jitter and TTD-offset of zero, the TTD measurements are carried out only within this
steady-state region4.
It can be deduced from the results reported in Figure 3.20 that the transducer must be
excited by a burst of more than 25 cycles in order to generate enough measurement points
in the quiet region parts of the ultrasonic signals, i.e., typically 50 to 70 periods of
excitation. The number of the allowed excited cycles is strictly limited by wave reflections
generated by the other transducer on the opposite side of the pipe. This problem becomes
more serious when the excitation of the two transducers is done simultaneously and the
ultrasonic path length is between 5 and 8 cm. Then, the number of cycles is limited to only
about a half of the sound wave traveling time. However, when a single excitation is used to
excite the transducers, the number of cycles is limited to about a sound wave traveling time
since the transmitted and the received signals are acquired from different scope channels.
In order to reach a minimum TTD-jitter and TTD-offset in the receive mode, the transducer
has to recover from the transmit mode and achieve a state where it becomes completely
4
See Section 4.5 of Chapter 4 for further discussion.
calm after passing the transient part, where free vibration is still valid. Therefore, it is
necessary either to extend the pipe length or to use a pair of transducers with a sufficiently
high resonance frequency. For a 1 MHz flow meter pipe and within a typical distance of
about 5 cm, the number of periods is limited to roughly 20. However, a 4 MHz flow meter
pipe allows 4 times more periods. In this work, we chose to switch from 1 MHz to the
4 MHz flow meter pipes (Figure 2.2), where typically a sinus burst of 70 cycles is used for
excitation.
Fig. 3.21: Measured transmitted signal of the 4 MHz flow meter pipe
Fig. 3.22: Measured received signal of the 4 MHz flow meter pipe
3.8. Characteristic Frequencies of Piezoelectric Transducer
As mentioned in Section 3.5, there are two frequencies which characterize the piezoelectric
transducer impedance: the series and the parallel frequencies. In order to estimate the
frequencies which characterize the 4 MHz transducer pair located in the 4 MHz flow meter
pipe used in our experiments, a variation of transmitter voltages VTD1 and VTD2 in both
directions with respect to the driving frequency is obtained experimentally using the
measurement setup shown in Figure 3.13 and the 4 MHz flow meter pipe. The amplitude
voltages are estimated in the steady-state region of transmitted signals in both directions
using least square sine-fitting. This experiment result (Figure 3.23) shows the same general
characteristics when compared to the simulated magnitude response of the electrical model
of the transducer shown in Figure 3.9. The piezoelectric series resonance frequency
corresponds to the minimum impedance of the transducer (minimum amplitude in
Figure 3.23), and the parallel resonance frequency corresponds to the maximum impedance
of the transducer (maximum amplitude in Figure 3.23). The transducer is protected by its
high impedance in order to transmit the maximum energy to the water. The maximum
voltage results from the fact that the driver sends a minimum of energy to the transducer,
thus the current over Rs (Figure 3.8) becomes minimal.
fp1 fp2
fs1 fs2
Fig. 3.23: Measured transmitter voltages versus driving frequency, performed at

ambient temperature
Chapter 4
Determination of Transit Time Difference
4.1. Introduction
The transducer model based on the driven damped harmonic oscillator model shown in
Chapter 3 allows an accurate description of the behavior of a piezoelectric transducer. The
free oscillation of the transducer dies out when it is excited by a burst of a sufficient
number of cycles in order to allow the transducer to reach a steady-state. We take
advantage of this phenomenon - the dying out of the transient region - to measure
indirectly the transit time difference (TTD) by computing the phase difference between the
steady-state region of received signals in the upstream and downstream directions. Many
factors significantly influence the upstream and downstream transit times, such as water
viscosity, temperature-dependent transducer parameters, and the speed of sound c, which is
related to the compressibility K and the density  of water by the following equation:
(4.1)
All these factors impose great challenges to develop a new methodology for achieving high
measurement accuracy with reduced TTD-jitter and compensated TTD-offset at no-flow
conditions and within a temperature range from room temperature to 80°C.
The first part of this chapter introduces the experimental setup for measuring the TTD,
followed by the description of the transmission and reception electronics associated with
the transducers.
The second part of the chapter describes the methodology applied to compute the TTD.
Chapter 4: Determination of Transit Time Difference 34
4.2. Experimental Setup for Measuring the TTD

The operational principle of the system used in the experiments is illustrated in Figure 4.1.
As mentioned previously in Chapter 3, in order to allow the receiving transducer to reach a
steady-state where both amplitude and frequency are settled, a sinusoidal burst of 70 cycles
generated over a time period of 100 ms by a function waveform generator (Hameg HMF
2550) is used to drive the two transducers (TR1 and TR2) simultaneously through the
resistances R1 and R2. These transducers convert electrical excitation into a mechanical
wave. The driving frequency of the burst is selected near the resonance frequencies of both
transducers in order to achieve the maximum output vibration with an enhanced SNR.
Fig. 4.1: Block diagram of the experimental setup for measuring the TTD
Figure 4.2 shows a picture that was taken while performing the measurements.
Fig. 4.2: Experimental setup for measuring the TTD

In order to eliminate the high starting jitter, which is caused by triggering and recording
ultrasonic signals from both directions separately (sequentially), we rely on the
simultaneous (synchronous) excitation approach. This means that the common
synchronized clock of the dual channel digital oscilloscope (Digilent Analog Discovery) is
used to trigger and acquire signals from both directions. Hence, we eliminate any
differential delay time between the two input channels.
As shown in Figure 4.3, the sound waves are received simultaneously by both transducers
after a traveling time that is proportional to the distance between the transducers
(L = 42.2 mm). The received signals are attenuated due to the loss or absorption of
acoustical energy by the medium [34]. Therefore, we use two low-noise operational
amplifiers (OPA2846ID), shown in Figure 4.1, in order to increase the signal levels. The
amplified signals are acquired by the digital oscilloscope, which has two 14-bit channels
with a sampling rate of 50  106 Samples/s per channel. The samples are saved in comma-
separated values (CSV) data format and imported to the computer (PC). An embedded
(PC) software algorithm written in Matlab handles the automatic TTD measurement
repetition, providing all necessary programs to control the function generator and digital
oscilloscope.
Fig. 4.3: Acquired ultrasonic signals in both directions
(measured using a 4 MHz driving frequency)

4.3. Methodology for the Calculation of TTD

The TTD can be deduced from the phase difference between the steady-state parts of the
received signals in the upstream and downstream directions by estimating the two phases
with the least squares sine-fitting algorithm of Matlab (see Figure 4.4) [19] [38]:
(4.2)
(4.3)
where and are the phase shifts between the received and transmitted signals
measured in the steady-state regions (see Figure 4.5) in the upstream and downstream
direction, respectively. In the quiet region, the received signal frequency f matches the
forced frequency of the exciting signal. Therefore, the TTD can be computed as follows:
(4.4)
Fig. 4.4: Phase difference between the steady-state region of received signals in both
directions
Fig. 4.5: Response of the 4 MHz transducer to the incident ultrasonic wave
Assuming that the phase shifts between the electric and acoustic signals on both
transducers for transmitted signals are t1, t2 and r1, r2 for received signals, the total
calculated phase difference ( T) between the signals recorded from both directions is given
by [19]:
. (4.5)
Since tup and tdown are the same at no-flow conditions according to [7] and [19] under
electro-acoustical reciprocal operation, in which t1 + r2 = t2 + r1, the phase difference
between the upstream and downstream directions can be compensated. Such operation
requires that the equivalent electrical impedances of the electronics and transducers are
equal in both directions. Therefore, under this condition the unmatched transducers are no
longer the only reason for the TTD-offset drift produced at no-flow conditions. The
dissymmetry between the two directions signal paths (in terms of the electrical impedances
of the electronics and transducers used in the meter), which mainly causes the TTD-offset.
4.4. Least Squares Sine-Fitting
The least squares sine-fitting (LSSF) is a very accurate method to estimate all three
parameters - amplitude, phase, and frequency - that characterize a digitized sinusoidal
signal sampled at a well-defined sampling rate [39]. The sine-fitting technique is used as a
filter and can significantly reduce noise such as the ADC quantization noise [40]. This
procedure is very often used to recover distorted and noisy signals in tests and
measurements [41-45]. A three-parameter sine-fit algorithm is used for a reasonable
estimation of amplitude, phase, and frequency of each recorded sample. Assuming that the
data of n samples has been saved, the fitted function is given by:
, (4.6)
where A and are the amplitude and the phase of the signal, respectively, f is the driving
frequency, and tn is the discrete-time vector.
The sum of squares of the error between these three extracted parameters and the measured
data is given by:
. (4.7)
The algorithm chooses successive values of A, f, and to minimize , taking into account
the signal-to-noise-and-distortion ratio [41] [45].
4.5. Estimation of the Quiet Region
As mentioned before, accurate phase estimation from the acquired recorded data is
performed in the quiet parts of received signals. As shown in Figure 4.6.b, relatively high
random fluctuations in voltage occur in the region between the transmitted and received
signals, which can influence measurement accuracy. Such voltage fluctuations arise due to
insufficient time, which leads to a complete decay of the transmitted signal to zero before
the arrival of the received signal due to simultaneous excitation. In the case of single
excitation, the transmitted and the received signals are acquired from different scope
channels, as shown in Figure 4.7. It demonstrates that only a few millivolts typical noise
floor of the digital oscilloscope is observed before the start of the transmitted signal and
the arrival of the received signal (Figure 4.7.a and 4.7.c). This also shows that in our
chosen technique the higher voltage noise arise due to the proximity of the transmitted and
received signals in the same channel.
Fig. 4.6: Voltage fluctuations in the region between the transmitted and received
signals measured at simultaneous excitation
Fig. 4.7: Noise of the oscilloscope measured at single excitation

In order to protect the phase measurement interval from these voltage fluctuations, the
transducer should be excited for a sufficiently long period of time. Figures 4.8 and 4.9
show the results of the cycle-by-cycle time delay of the transmitted and received signals,
respectively. These results are calculated through the estimation of phase shifts of the 70
periods between the electric and acoustic signals on both transmitting and receiving
transducers in both directions by using sine-fitting. These results show that after 50 cycles
of excitation the two transmitted and received signals look more steady and the transient
region (free oscillation of the transducer) has completely died out. Therefore, on the basis
of these results we decided to use the last 20 periods to extract sine-fitting parameters in all
our experiments.
Fig. 4.8: The cycle-by-cycle time delays of both transmitted signals in both directions
Fig. 4.9: The cycle-by-cycle time delays of both received signals in both directions
4.6. Algorithm for Computing the TTD
The main components of the algorithm used for estimating the phase and computing the
TTD are shown in Figure 4.10. A sine-fitting technique is used to reduce the error in the
phase estimation. This technique estimates the sine wave that best fits the recorded samples
of the last 20 cycles of signals received from both directions. The algorithm used for
computing the TTD value performs the following steps:
 Read the recorded data saved in CSV format.
 The DC offset correction is performed to remove it from the acquired signal. This
correction is done by computing the mean voltage value of all recorded voltage samples,
which is then subtracted from each data voltage sample.
 An estimation of both frequencies of sine waves from the recorded sampling data is
performed by using the least squares sine-fitting algorithm of Matlab. This estimation is
carried out in the steady-state parts of received signals in upstream and downstream
directions.
 An estimation of amplitude and phase, which characterize the digitized sinusoidal
signals, is performed by using the previously obtained mean value of the fitted
frequencies.
 Calculate the difference between the two fitted phases.
 Calculate the TTD.
In order to control the shape of the curve, the upper and lower bounds of the estimated
parameters (amplitude, phase, and frequency) that characterize the digitized sinusoidal is
introduced. For instance, the amplitude has bounds [0, 5], the phase has the bounds [- ].
With this construction in place, the fitted curve accurately extracts parameters from the
measured data, especially the two phases that are used to calculate the transit time
difference.
Read the CSV file

of both directions
DC offset correction
Estimate both directions

frequencies
Estimate A and  using the

frequencies mean value
estimated previously
Calculate the phase

difference
Calculate TTD
Fig. 4.10: Flowchart for computing the TTD
.
Chapter 5
TTD-jitter Reduction
5.1. Introduction
The purpose of sampling and digitizing the transmitted and received ultrasonic signals is to
perform advanced signal processing and accurately extract the transient time difference
(TTD). The continuous signal waveform is defined on the basis of a set of time-discrete
samples over a specified period of time. However, clock jitter causes uncertainty in the
sampling time during the data acquisition of ultrasonic waves. This leads to random
fluctuations in the corresponding amplitude. Therefore, the recorded data are corrupted by
noise and distortion. As the TTD is computed on the basis of a finite set of sampled data,
such uncertainties in the sampling time produce random fluctuations of TTD
measurements, known as the TTD-jitter.
The TTD-jitter in flow measurement has a huge impact on its accuracy. It can be
experimentally evaluated by repeatedly performing the same measurement and using the
same measurement setup under the same conditions.
This chapter investigates the TTD-jitter noise and its impact on the accuracy of flow meter
measurement results. The approaches used in order to mitigate the TTD-jitter are also
presented here.
Chapter 5: TTD-jitter Reduction 43
5.2. The Effect of Sampling Jitter on TTD

In the sampling-process, the sampling time uncertainty (also called the sampling time
jitter5) is the random fluctuations in time location of a given waveform sample.
Correspondingly, this uncertainty in sampling time introduces errors in the amplitude of
the ultrasonic signal. As illustrated in Figure 5.1, for a given sine wave of period T and
peak-to-peak amplitude Vpp the sampling time jitter Δt causes a change in the measured
value equal to ΔV = Vpp π Δt/T [46]. This contributes to the increase of the TTD-jitter. In
the present application, the TTD-jitter refers to the dispersion of the measured TTD around
the mean value (zero for no-flow). The latter is assumed to be a stationary Gaussian noise,
which means that the standard deviation does not change over time. The TTD-jitter is thus
defined by the mean value and the standard deviation.
Fig. 5.1: Effect of the sampling jitter on measured value
Typically, the transducers are excited by a sinus burst of 70 cycles at 4 MHz frequency.
The experimental results in Figure 5.2 show the last 19 periods of a transmitted signal
superimposed upon each other. Figure 5.2.b shows a magnification of Figure 5.2.a.
Interestedly, this figure shows that for consecutive periods of a measurement, the zero-
crossings jump from one period to the next period statistically around the zero point. The
Gaussian (or normal) probability distribution function can be used to statistically
characterize the maximum time deviation of the periods around the zero point caused by
the sampling time jitter. The maximum time deviation of the 19 consecutive periods around
the zero point is about 374 ps, which corresponds to 6 the standard deviations
( = 62.4 ps). Note that the standard deviation result is obtained by evaluating 19 time data
5
Not to be confused with the TTD-jitter.
samples around the zero point shown in Figure 5.2.b applying the probability density
function. The time deviation of the periods increases with increasing number of time data
samples.
Fig. 5.2: Time deviation of the periods around the zero point:
(a) the last 19 periods of the transmitted signal superimposed upon each other,
(b) a magnified around the zero point
5.3. Sources of TTD-jitter

The TTD-jitter noise arises from many sources which contribute to the measurement
uncertainty. These sources are uncorrelated and may originate from anywhere in the signal
path from transmitting to receiving waveforms. Some of these sources are dedicated to the
electronics associated with the transducers such as [39]:
 transmitter waveform phase noise or driving frequency instability of the
transmitted signal due the function generator
 jitter in the sampling instance that may occur due to imperfect sample-and-hold
circuit synchronization
 crosstalk between the cables
 irregularities in the transmission medium.
Other noise sources such as the effects of the side lobes also contribute to the TTD-jitter.
As shown in Figure 5.3, most of the acoustic energy is transmitted perpendicularly from
the piezoelectric disk. The main lobe is the lobe containing maximum sound pressure. This
lobe is calculated by finding the angle at which the sound pressure is halved [47]. The
acoustic wave directivity6 is temperature-dependent [48], especially at the side lobes level,
which increases from low temperature to high temperature. This can enhance the level of
reverberation in the measuring pipe, producing undesired echoes acting as noise. This
noise may corrupt the signal along the signal path, causing errors in the measured transit
time.
Fig. 5.3: An example of a sound directivity pattern [47]
5.4. TTD-jitter Reduction Technique
Since the recorded data of the received waveform are sampled with 50 MS/s, each period
in the steady-state region contains around 13 sample points. The sine-fitting applied
approach adjusts Equation (4.6) to the set of the recorded sampling data in order to extract
the characterized parameters of the trace, namely its amplitude, phase, and the frequency.
The following equation, obtained from Equation (4.4), is applied to compute the TTD
using the upstream and downstream extracted phases ( , :
(5.1)
The precision to which the TTD is determined can be influenced by two parameters: the
amplitude of the received signal and the number of the samples used in the fitting
algorithm. Concerning the first parameter, there is a clear correlation between the TTD-
6
The acoustic directivity describes how the pressure of a sound wave is transmitted from a transducer.
jitter level and the amplitude of the received signal in the quiet region. In other words, the
obtained TTD has a jitter that depends on the input dynamic range of the acquisition
system. This is optimized by adjusting the gain of the two amplifiers (Figure 4.1), aiming
to cover most of the input range of the ADC. The results shown in Figure 5.4 (carried out
at room temperature, no-flow condition, N = 250 fitted samples, and different receiver
voltages7 VRD1 and VRD2) suggest that the increase of the voltage headroom in both
directions reduces the TTD-jitter.
The used ADC has 14-bit resolution and an input full-scale range of ±2.5 V8. The least
significant bit (LSB) voltage can be expressed as:
(5.2)
where Nbit is the ADC’s resolution in bits and VFSR is the full-scale voltage range of the
ADC. According to Equation (5.2), the ADC has a voltage resolution VLSB of about
300 µV. The dynamic range of data acquisition system (DR), which is the ratio of the
maximum input voltage VFS9 (signal amplitude) to the minimum voltage VLSB, can be
expressed as:
(5.3)
Figure 5.5 illustrates the measurement precision (represented by the standard deviation)
versus the dynamic range of the acquisition system. Note that the standard deviations are
calculated from TTDs of 50 captured ultrasonic waveforms. It can be deduced on the basis
of these results that high dynamic range of the acquisition system can be achieved using a
high gain amplifier to cover the input full-scale range voltage (VFSR = 5V) of the ADC,
which results in very low TTD-jitter.
7
The amplitude voltages VRD1 and VRD2 are estimated in the quiet region of both directions of received
signals using sine-fitting
8
According to the datasheet of the used ADC (AD9648) [49], it achieves SINDR of 74.3 dB, SNR of
75.4 dB, and ENOB of 12-bit at 9.7 MHz.
9
VFS refers to the full-scale sine wave of the received signal in the quiet region.
Fig. 5.4: TTD measured at different receiver amplitudes
Fig. 5.5: The standard deviation of the transit time difference (TTD) versus the
dynamic range of the acquisition system (N = 250 fitted samples)
As mentioned previously, the TTD-jitter can also be reduced by increasing the number of
the signal samples that are used to adjust the sine-fitting parameters. Since the noise
present at each sampling sequence is uncorrelated, a given number of samples N reduces
the timing jitter values by a factor of (according to the averaging principle). This can
mathematically be expressed as:
(5.4)
where is the standard deviation error of the TTD (after averaging), σ is the standard
deviation of the TTD, and N is the size of the samples.
Figure 5.6 depicts a single TTD measurement carried out at room temperature and no-flow
conditions for different numbers of the fitted samples.
Fig. 5.6: TTD results measured at different numbers of fitted samples (N)
Figure 5.7 illustrates the measured TTD standard deviation (STD) versus the number of
fitted samples as well as the theoretical limits estimated using Equation (5.4). It can be
deduced on the basis of these results that the measured TTD-jitter exhibits the same
behavior as the calculated one. In other words, as the number of samples increases, the
standard deviation of the measured TTD decreases according to Equation (5.4).
Fig. 5.7: The standard deviation of TTD versus the number of fitted samples (N)
Chapter 6
TTD-offset Cancellation
6.1. Introduction
As described previously, in the ideal case when there is no flow in the measuring pipe the
upstream and downstream transit times should be identical. However, in reality this is not
the case as there are many factors that prevent the ultrasonic signals (see Section 4.1) from
reaching the receivers after exactly the same traveling time. Hence, any deviation of the
measured transit time difference (TTD) from zero at no-flow conditions is referred to as a
TTD-offset. This TTD-offset, which limits the minimum measured flow, presents a serious
drawback in high accuracy measurements. Therefore, it is worthwhile to develop a
theoretical analysis which provides a better understanding of the TTD-offset sources and
their contributions in order to develop a measurement strategy that allows effective TTD-
offset compensation.
6.2. Sources of TTD-Offset

In water flow meter applications, the environment temperature inside the measuring pipe
may vary from ambient temperature to 80°C. This leads to a variation of the transducer's
resonance frequency, which cannot withstand the change of the operating temperature [19]
[50-52]. However, according to the harmonic oscillation model10, if a pair of transducers
has different resonance frequencies, then the phase difference between the two steady-state
parts of the received signals is different from zero as well. Figure 6.1 shows an example of
two phase responses versus the driving frequency. This two plots of phase responses are
obtained from the phase equation of the particular solution of a damped driven harmonic
oscillator (Equation 3.10) for two transducers with different resonance frequencies (for this
example, the resonance frequencies are fr1 = 4 MHz and fr2 = 4.05 MHz for the first
10
The analytic modeling of the piezoelectric transducer disk described in Chapter 3.
Chapter 6: TTD-offset Cancellation 51
transducer and the second transducer, respectively). The difference between these two
phases is shown in Figure 6.2, where it can be observed that within the resonance
frequency range the phase difference is significantly different from zero. Thus, the zero
flow error can drift according to the temperature changes, which can cause significant
TTD-offset.
Fig. 6.1: Phase response versus driving frequency for two different transducers
Fig. 6.2: Phase difference response versus driving frequency
6.3. The Impact of Temperature on the Transducer Resonance

Frequencies
In order to predict what effect temperature changes have on the series and parallel
resonance frequencies of the 4 MHz transducer pair, the variation of transmitter voltages
(VTD1 and VTD2) with respect to the driving frequency is investigated experimentally (see
Section 3.8). The experiment allows to obtain an approximate determination of both
frequencies by estimating the maximum and the minimum transmitter amplitudes
measured at different temperatures. This experiment is carried out by exciting the

transducers with a sinus burst of 70 cycles at a specified frequency in the range between
3.8 MHz and 5 MHz. The results illustrated in Figures 3.23, 6.3, and 6.4 demonstrate the
variation of transmitter voltage amplitudes in both directions measured at ambient
temperature (25°C), 60°C, and 80°C, respectively, with respect to the driving frequency.
Note that the measuring pipe is heated using a thermostat. It can be deduced from the
obtained results that the frequency at which the amplitude becomes minimal (minimum
impedance) is close to series resonance frequency fs, whereas the frequency at which the
amplitude becomes maximal (maximum impedance) is close to the parallel resonance
frequency fp. However, because the parameters of the two transducers are not identical, the
variation of both amplitudes is also not identical. The temperature dependence of the
resonance frequencies of the pair of 4 MHz transducers is summarized in Table 6.1.
Fig. 6.3: Transmitter voltages versus driving frequency measured at 60°C
Fig. 6.4: Transmitter voltages versus driving frequency measured at 80°C

Resonance frequency [MHz]

Temperature [°C] Transducer 1 Transducer 2
fs fp fs fp
25°C 4.02 4.615 4.01 4.595
60°C 4 4.49 3.96 4.475
80°C 3.97 4.465 3.925 4.43
Table 6.1: The dependence of resonance frequencies on temperature
As shown in Figure 6.5, the series fs and the parallel fp resonance frequencies of the used
transducer pair decrease as the temperature goes up. Regardless of the mismatch between
transducers, the variation of the resonance frequency with the operating temperature range
can cause a dissymmetry between the two directions of the ultrasonic signal paths. Thus,
the temperature variation of the medium is the main reason for the zero flow TTD-offset
drift.
Fig. 6.5: The effect of temperature on the transducer resonance frequencies
6.4. Zero Flow TTD-offset Correction
As mentioned before, in order to eliminate the possibility that the meter detects a false flow
under no-flow conditions, the upstream and downstream transit times should ideally be the
same, although this may not be the case unless special precautions are taken. Due to the
fact that every flow direction exhibits a slightly different electrical impedance compared to
the other, there is a dissymmetry between upstream and downstream signal paths [7]. This
effect provides a different amount of currents flowing through R1 and R2 (see Figure 4.1),
and causes different amplitudes in the upstream and downstream transmitted signals.
Figure 6.6 shows the result of the measurement carried out at room temperature and
no-flow conditions. Both transducers are excited with the same sinus burst at 4 MHz forced
frequency. The plotted curves show that the measured difference between the two
amplitudes of the transmitted signals is about 250 mV.
Fig. 6.6: Acquired ultrasonic signals in the upstream and downstream

directions (measured at 4 MHz driving frequency)
By computing the phase shifts t1 and t2 between the excitation and the sound wave on
both transducers for transmitted signals, a delay time difference of -2.4 ns is obtained
between transmitted signals in both directions. This relatively high starting delay time
difference can be explained by upstream and downstream unmatched electrical impedance
values, which lead to the difference between the two transmitter amplitudes (of about
250 mV). The 250 mV transmitter amplitude difference shown in Figure 6.6 results in
about 150 ps zero flow TTD-offset.
In order to effectively eliminate TTD-offset at no-flow conditions, it is necessary to match

the upstream and downstream electrical impedance of both transducers and their associated
electronic to reach a highly symmetrical signal paths. According to the literature [25], the
electrical impedance of a transducer can be controlled by a driving frequency. We have
used this feature to eliminate the electrical impedance mismatch between both directions
because a well-matched upstream-downstream signal path reduces the transmitter
amplitude difference and results in a very small zero flow TTD-offset.
By changing the sinus burst frequency from 4 MHz to 4.19 MHz, the 150 ps zero flow
TTD-offset measured previously is substantially reduced to less than 5 ps (Figure 6.7).
Moreover, comparing the amplitudes of the transmitted signals depicted in Figures 6.6 and
6.7, it can be observed that the difference between the two transmitter amplitudes is
reduced from 250 mV measured at 4 MHz to less than 15 mV, which is achieved by
exciting the transducers at 4.19 MHz driving frequency under the previously mentioned
conditions. In addition, the delay time difference of -2.4 ns achieved between transmitted
signals in both directions at 4 MHz is reduced to -664 ps at 4.19 MHz driving frequency.
Therefore, choosing an appropriate driving frequency within the resonance frequency
range of the transducer is the key to cancel the long-term drift of the TTD caused by
temperature variations.
Fig. 6.7: Acquired ultrasonic signals in the upstream and downstream directions
(measured at 4.19 MHz driving frequency)
Figure 6.8 shows TTD measurement results obtained in a driving frequency range from
4 MHz to 5 MHz with 5 KHz step variation at room temperature and no-flow conditions.
The result shows that within the resonance frequency range of the two transducers TTD-
offset compensation can be achieved at 4.185 MHz driving frequency. Figures 6.9 and 6.10
represent the same measurements as before carried out at 60°C and 80°C temperature,
respectively, and they clearly show that these TTD-offsets compensation are achieved at
different driving frequencies (about 4.095 MHz for 60°C and about 4.075 MHz for 80°C)
due to the fact that the resonance frequency of the transducer is changing as a function of
temperature as well.
Fig. 6.8: TTD versus driving frequency measured at ambient temperature

Fig. 6.9: TTD versus driving frequency measured at 60°C
Fig. 6.10: TTD versus driving frequency measured at 80°C
To sum up, for a successful TTD-offset compensation it is necessary to adopt the two
following strategies:
• Offline TTD-offset calibration must be performed in the manufacturing process in

order to provide a table with driving frequencies and their corresponding
temperature within a temperature range from ambient temperature to 80°C.
• Online TTD-offset calibration must be performed with control algorithms that use
adequate driving frequency suitable to compensate the TTD-offset according to
the measured temperature inside the measuring pipe.
6.5. Temperature Dependence of the TTD-offset Calibration
As mentioned before, choosing an appropriate driving frequency can drastically reduce the
zero flow TTD-offset. To attain automatic compensation, one needs to accurately set a
suitable driving frequency to eliminate the TTD-offset caused by the temperature variation
inside the pipe. In order to ensure repeatability or precision of the driving frequency used
to compensate the TTD-offset, several experimental measurements are performed over
three months at four randomly chosen temperatures (35°C, 60°C, 70°C, and 80°C). These
experiments have lead to the results illustrated in Figures 6.11 and 6.12, where the heating
up and cooling down of the pipe is done by means of a thermostat. As it can be seen in
Figure 6.11, the first experiment starts by adjusting the temperature inside the pipe to 35°C;
the calibration of the TTD-offset to zero is reached at 4.153 MHz driving frequency. While
maintaining the same driving frequency, the heating up of the pipe to 80°C increases the
TTD-offset to about 270 ps. By changing the driving frequency from 4.153 MHz to
4.075 MHz, the TTD-offset is calibrated once again to nearly zero value. The cooling down
of the pipe from 80°C back to 35°C at the same driving frequency decreases the TTD-
offset to about -320.3 ps, while the TTD-offset reaches again zero at the same starting
excitation frequency (4.153 MHz) since in that moment the measured temperature inside
the pipe is again 35°C. The relatively high measured TTD-jitter that can be seen clearly in
the measured plots, especially at 35°C, can be explained by the fact that a full-scale voltage
range of the ADC is not reached completely. This happens because the received signal
attenuation is normally proportional to the viscosity of the water, which decreases with
increasing temperature. Therefore, the gain of the amplifier is controlled by four
resistances (see Figure 7.1) in order to bring the received signal to a level that is high
enough to resist the attenuation that it will be subjected to.
The same setup is used to perform the second experiment, apart from the fact that the
maximum reached temperature is 60°C instead of 80°C. It can be seen from Figure 6.12
that heating up and cooling down the pipe from 35°C to 60°C and from 60°C to 35°C,
respectively, provides TTD-offset values that vary between -220.3 ps and 338.8 ps.
Fig. 6.11: TTD measurements performed over a temperature range

from 35°C to 80°C
Fig. 6.12: TTD measurements performed over a temperature range

from 35°C to 60°C
The results showing the temperature dependence of the driving frequency are obtained in
five measurements carried out over a three-month period (see Table 6.2 and Figure 6.13).
Taking in consideration the suppression of the zero flow error, it can be noticed that the
driving frequency at which the two transducers are excited is strongly influenced by the
temperature. Therefore, we have shown that there is a clear, simple, stable, and reliable
relation between the required excitation forced frequency and temperature.
Note that the measured driving frequencies with respect to the temperature reported in
Table 6.2 change from one pair of the transducers to the other.
Driving frequency [MHz]

Temperature
1st 2nd 3rd 4th 5th
[°C]
measurement measurement measurement measurement measurement
35°C 4.152 4.153 4.151 4.154 4.152
60°C 4.098 4.098 4.097 4.096 4.097
70°C 4.088 4.089 4.088 4.087 4.085
80°C 4.076 4.073 4.073 4.072 4.071
Table 6.2: Temperature dependence of the driving frequency
Fig. 6.13: Temperature dependence of the driving frequency used for the zero flow
TTD-offset compensation
Chapter 7
Results and Analysis
7.1. Introduction
In this chapter, we demonstrate a practical application of the developed measurement
methodology described in the previous chapter. Hence, the electronic system design
introduced in Chapter 4 and the software algorithm used for signal processing are
described in more detail. Afterwards, the experimental results are presented and analyzed.
This analysis allows to highlight the robustness of TTD-jitter reduction and TTD-offset
correction methodologies, developed in this thesis. Note that all measurement results
reported in this chapter are performed at no-flow conditions within a specified temperature
range from ambient temperature to 80°C
Figure 7.1 illustrates the electronic instrumentation system utilized to perform the
ultrasonic measurement with significantly improved accuracy. The designed electronics
and the software algorithm written in Matlab are capable of handling automatic
measurement repetition, providing all control sub-programs, namely, the control of the
function generator and analog discovery device.
Chapter 7: Results and Analysis 61
Fig. 7.1: Block diagram of the experimental setup
7.2. Hardware Design
The hardware elements of the experimental setup presented in Figure 7.1 are:
• An arbitrary waveform generator, which generates a sinusoidal burst of a finite
number of cycles in order to excite the transducers.
• A pair of 4 MHz transducers located in the flow meter pipe, which are capable of
transmitting and receiving an ultrasonic signal.
• DC source to power the low-noise amplifiers, which are used to amplify the
voltage levels of the two received signals. Amplifiers gain is controlled by four
resistors R3, R4, R5, and R6 in order to reach 5 V full-scale voltage of the ADC.
• Digital oscilloscope, which has dual-channel, 14-bit ADCs for digital signal
acquisition.
• An embedded (PC) software algorithm written in Matlab handles automatic TTD
measurement repetition.
• Additionally, the flow meter pipe is put in a water bath with a thermostat to
regulate the temperature of the transducers at given levels, assuming that the
temperature of the transducer can be considered to be the same as the temperature
of still water filling in the pipe.
7.3. Software Algorithm
Completely automatic measurement process is achieved through the developed software

algorithm, which is used to automate manual tasks such as the configuration of the function
generator and the digital oscilloscope to output the desired signal and save numerical data,
respectively. The algorithm is also used to compute TTD and compensate the TTD-offset
error if necessary. This automatic procedure is summarized in Figure 7.2, and it mainly
consists of the following four sub-algorithms:
Transmit the sinusoidal

burst
Digitize the two

direction signals
Determine the shift phases of

both received signals using the
LSSF, and calculate the TTD
through the phase difference
between the received signals
Offset adjustment
Fig. 7.2: Flowchart of the automatic measurement repetition
• A sub-algorithm controls the arbitrary waveform generator in order to generate

sinusoidal burst.
• A sub-algorithm controls the two 14-bit ADCs in order to digitize the upstream
and downstream signals.
• The sub-algorithm described in Section 4.6 is used to compute the TTD. The
proposed method described in Chapter 5 is applied to attain reduced TTD-jitter.
• The last sub-algorithm is used to check whether the TTD-offset tends towards
zero value or not.
o If the TTD-offset is not approximately equal to zero, the adjustment of the

TTD-offset to zero is performed by choosing an appropriate excitation
frequency (see Section 6.5), which is capable of avoiding the TTD-offset
error.
7.3.1. Transmission of the Sinusoidal Burst
A sinusoidal burst of 70 cycles simultaneously excites the two transducers over a time
period of 100 ms at a driving frequency of 4 MHz (17.5 µs burst length) after the power up
of the system. The automatic configuration of the arbitrary waveform generator is
performed using the algorithm shown in Figure 7.3. The transmission procedure starts by
opening the serial port so that the communication between the PC and the function
generator can be enabled. All burst signal parameters (function, amplitude, frequency, the
number of cycles, and period) must be specified in the program in order for them to be sent
to the function generator through the serial port.
Open the serial port
Initialize the function

generator
Burst function: SIN

Burst frequency: 4 MHz
Number of burst cycles: 70
Burst amplitude: 20 V
Burst period: 100 ms
Fig. 7.3: Configuration of the function generator
7.3.2. Digitizing the Two Direction Signals
The two 14-bit ADCs of the digital oscilloscope operate up to 100 MHz maximum
sampling rate. The trigger controls enable us to capture a stable waveform to be digitized
and saved into a file containing numerical data regarding time and amplitude. One of the
channels is used as a trigger source, where the trigger level and slope controls provide the
basic trigger point definition. Therefore, the slope control, which determines whether the
trigger point is on the rising or the falling edge of the indicated signal, must be indicated in
the control software algorithm depicted in Figure 7.4 in such a way that the system can
deliver direct digital signals via the two integrated USB powered ADCs. The capture of a
single-shot waveform is controlled by triggering; thereafter, the performed acquisitions are
transferred to the PC in CSV format via USB with the maximum sample storage length of
8192 samples in order to be processed in Matlab.
Power up
Trigger configuration
Start the acquisition
Triggering No
activated ?
Yes
Saving and exporting

frequencies
data
Fig. 7.4: The acquisition of both direction signals
7.3.3. TTD-Offset Cancellation Algorithm
The TTD-offset compensation strategy presented in Chapter 6 can be accomplished

automatically by using an algorithm. This compensation is done with respect to the
temperature of the medium around the transducers (see Section 6.5) in order to extract the
required driving frequency that corresponds to the temperature of the medium (Table 6.2)
so that the TTD-offset error can be compensated. Choosing of the correct driving
frequency can be done accurately by first determining the temperature of the water inside
the measuring pipe. The measurement software system includes a conceptual TTD-offset
compensation algorithm depicted in Figure 7.5, which allows continuous detection and
compensation of the zero flow TTD-offset error.
Start the TTD-offset

adjustment
Yes Return
TTD-offset  0
?
No
Measure the temperature

inside the pipe
Determine the driving frequency

according to the measured temperature
from the Table 6.5
Excite the transducers using

the new driving frequency
Save status
Return
Fig. 7.5: Compensation of the TTD-offset
The TTD-offset compensation algorithm is executed immediately after computing the

TTD. This algorithm starts by comparing the obtained TTD value to zero. If the TTD-
offset is almost equal to zero, the new status information (driving frequency) is stored in
the PC buffer to be used to excite the transducers again (i.e., it is the return to the
transmission of the sinusoidal burst sub-algorithm). However, if the TTD-offset is different
from zero, the adjustment of the TTD-offset must be performed by choosing the correct
driving frequency that corresponds to the measured water temperature around the
transducers. This driving frequency, which can be used as new transmission frequency, is
obtained from Table 6.2. Thereafter, the new driving frequency is stored and used to excite
the transducers (i.e., it is the return to the transmission of the sinusoidal burst sub-
algorithm).
7.4. TTD-offset Compensation Evaluation
In order to evaluate the offset compensation described in Chapter 6, the measurements are
carried out using a 4 MHz flow meter pipe at no-flow conditions. As mentioned above, the
two received signals are sampled at 50 MS/s, around 13 samples per period. Since the
TTD-jitter can be reduced by a factor of (see Section 5.4), the TTD value is evaluated
on the basis of 250 samples taken from the last 20 cycles in the steady-state region for both
received signals. For each measurement setup, the TTD is computed several times with a
repetition rate of about 1.6 s. Figure 7.6 shows that at 35°C when the transducers are
excited at 4 MHz driving frequency, the average value of TTD-offset is approximately
−370 ps and the value of peak-to-peak TTD-jitter is 30 ps. The TTD-jitter measurement
precision, which refers to the standard deviation, is about 6 ps. The standard deviation is
calculated from TTDs of 250 captured ultrasonic waveforms. In order to check the validity
of our TTD-offset cancellation method, the first chosen 4 MHz frequency is changed to
4.153 MHz according to Table 6.2, considering that the temperature inside the measuring
pipe is 35°C. Therefore, the measured -370 ps TTD-offset value is reduced to almost zero
as shown in Figure 7.7.
Since the dynamic range of the acquisition system considerably influences TTD-jitter
performance, it is necessary to use controllable gain amplifiers in order to reach the
maximum amplitude of the received signal and therefore to cover input full-scale range
voltage (VFSR = 5V) of the ADC. However, as expected, the two voltage amplitudes
measured in the quiet region of both received signals increase along with the temperature
due to the fact that the viscosity of water decreases as the temperature goes up. This can be
deduced from the results shown in Figure 7.8. This test is performed by changing the
temperature of the thermostat from 80°C to 35°C.
Fig. 7.6: Uncompensated zero flow TTD-offset measured at 35°C
Fig. 7.7: Compensated zero flow TTD-offset measured at 35°C

Fig. 7.8: Two received amplitudes measured over a temperature range

from 80°C to 35°C
7.5. The Effect of Temperature on TTD-offset Measurements
Since the temperature is the main cause of the TTD-offset, additional TTD measurements
are carried out at different temperatures (ambient temperature, 60°C, and 80°C), always
using the 4 MHz flow meter pipe. The first result is illustrated in Figure 7.9, where the
TTD measurements are performed continuously over 6 hours at ambient temperature, and
the zero flow TTD-offset is adjusted only at the beginning by applying a driving frequency
of about 4.180 MHz. In this result, the deviation from zero of the TTD-offset value is less
than 50 ps over the whole measurement time. This is due to about 3 to 6°C variation of the
ambient temperature during the measurement period.
Under the same conditions as shown in Figure 7.9, experimental measurements are
performed at differently-regulated temperatures. Figures 7.10 and 7.11 depict the TTD
measurements performed at 60°C and 80°C, respectively. According to Table 6.2, the TTD-
offset cancellation is reached by setting the driving frequencies to 4.098 MHz at 60°C and
4.075 MHz at 80°C. As mentioned previously (see Section 7.4), the relatively high peak-to-
peak TTD-jitter of 30 ps measured at 35°C can be explained by the fact that the received
signals are not amplified enough to cover the dynamic range of the of the acquisition
system. However, the peak-to-peak TTD-jitter is reduced to 20 ps and 15 ps when
measured at 60°C and 80°C, respectively. This is due to the fact that the received signal
attenuation is proportional to the viscosity of the water, which decreases with increasing
temperature as shown in Figure 7.8. The standard deviation (calculated from TTDs of 250
captured ultrasonic waveforms) of the TTD-jitter decreases from 6 ps at 35°C to 4 ps at
80°C.
Fig. 7.9: TTD measurements performed over six hours at ambient temperature
Fig. 7.10: TTD measurements performed at 60°C

Fig. 7.11: TTD measurements performed at 80°C
7.6. Discussion
The transducers suffer from variations in their characteristic parameters such as static
capacitance, electrical impedance, and resonance frequency that is temperature dependent,
as demonstrated by the experimental measurements shown in Section 6.3. Furthermore, the
exact value of the transducer working resonance frequency depends on the input
impedance when the transducer acts as a transmitter and on the output impedance when the
transducer acts as a receiver. On the other hand, under the same input or output impedance
conditions the variation of the temperature changes the transducer resonance frequency as
well. The validity of the proposed methodology is checked by evaluating the variation of
the transducer’s resonance frequencies and the driving frequency, which is used to adjust
the TTD-offset with respect to the temperature changes (Figures 6.5 and 6.13). These
results show that an increase in the temperature leads to a decrease in the resonance
frequency of the transducer and the driving frequency at which the TTD-offset can be
compensated. Moreover, the long-term stability of the driving frequency used to adjust the
TTD-offset error has been experimentally proved at four different temperatures. As shown
in Table 6.2, the variation of the driving frequency, which is used to compensate the TTD-
offset, is less than 4 Hz at the same temperature of the medium around the transducers.
Chapter 8
Conclusion
The main goal of this thesis is to improve the performance accuracy of the ultrasonic
transit time flow meters . This work also provides a detailed description of the theoretical
development and the experimental validation of the ultrasonic system and develops a new
water flow measurement approach that solves the TTD-jitter and TTD-offset problems
raised by the ultrasonic flow meters. The thesis starts with a proposal for the analytic
modeling of the piezoelectric transducer disk. This proposal was borrowed from damped
driven harmonic oscillator theory. By investigating the amplitude and phase responses of
the voltage across the resistive element of the transducer's electrical model, which
represents the energy transferred to the water, it is found that the harmonic oscillator model
is highly suitable to effectively capture the electrical behavior of the piezoelectric
transducer disk, the most critical part of the ultrasonic water flow meter.
The numerical solution of the second order differential equation of damped harmonic
forced system includes the transient or the homogeneous solution and the steady-state or
the particular solution. This steady-state can be reached only by exciting the transducer for
a sufficiently long time period. We have deduced through these solutions that the accuracy
of the flow meter can be further improved if we rely on the particular solution instead of
the homogeneous one. This steady-state part of the signal is characterized by a fixed
amplitude, a phase shift, and a frequency that matches the driving frequency. Therefore, the
TTD is measured as the phase difference between the received signals in both directions.
Furthermore, the phase shifts on both transducer for the received signals are estimated in
the steady-state parts of the signals using the least square sine-fitting, which filters out all
the noise related to the signal digitization carried out by the two 14-bit ADCs.
Chapter 8: Conclusion 72
The second development established through this work is the TTD-jitter reduction
approach, which considerably enhances TTD-jitter performance. As it can be inferred from
the experimental results, the standard deviation of the TTD-jitter is reduced drastically by
using the sine-fitting technique. Furthermore, the system accuracy is carried out one step
further by exploring two techniques, namely, employing low-noise amplifiers to improve
the acquisition system dynamic range and using an adequate number of samples in a
specific time interval (steady-state region).
Another achievement of this study is the extension of the TTD-jitter reduction technique to
TTD-offset cancellation, which also degrades the ultrasonic flow meters. A novel approach
to the problem has been developed to cancel the long-term zero flow TTD-offset drift
caused by temperature variations. This approach is based on the principle that continuously
adjusting the driving frequency nearby the resonance working area of the transducers
reduces the zero-flow TTD-offset to zero. The TTD-offset dependence of the forced
frequency finds a simple explanation in the theory of the oscillator: if the forced frequency
is in the range between the respective resonance frequencies, a slight shift of the frequency
produces an increase or a decrease in the amplitudes of the transmitted acoustic waves.
Choosing a correct forced frequency according to the temperature of the medium can
adjust the amplitude of the transmitted signals and compensate the TTD-offset. The
robustness of the zero flow TTD-offset cancellation methodology is experimentally
validated in harsh conditions, such as high temperatures(up to 80°C).
.
Bibliography
[01] M. Sc. Yaoying Lin, “Signal processing and experimental technology in ultrasonic
flow measurement”, Ph.D. dissertation, Duisburg Univ., Essen, 2004.
[02] J. Yoder, “Ultrasonic flow meters in the energy measurement spotlight”, Pipeline and
Gas J., vol. 236, no. 7, pp. 20-28, Jul. 2009.
[03] M. Kupnik, A. Schröder, and P. O’Leary, “Adaptive pulse repetition frequency
technique for an ultrasonic transit-time gas flow meter for hot pulsating gases”,
IEEE. Sensor J., vol. 6, no. 4, pp. 906-615, Aug. 2006.
[04] J. Berribi, “Self-diagnosis techniques and their applications to error reduction for
ultrasonic flow measurements”, Ph.D. dissertation, Luleå Univ., Sweden, 2004.
[05] J. Reyes, and A. Acevedo, “Simulation and experimental validation of a transit time
in an ultrasonic gas flow meter using air”, in Proc. conf. IEEE. ANDESCON, Bogota,
Colombia, Sep. 2010, pp. 1-6.
[06] P. Lunde, M. Vestrheim, B. Reidar, S. Smorgrav, and A. K. Abrahamsen,
“Reciprocity and its utilization in ultrasonic flow meters”, in Proc. 23rd Int. North
Sea Flow Measurement Workshop, Tønsberg, Norway, Oct. 2005, pp. 18–21.
[07] P. Lunde, M. Vestrheim, B. Reidar, S. Smorgrav, and A. K. Abrahamsen, “Reciprocal
operation of ultrasonic flow meters: criteria and applications”, in Proc. Conf. IEEE
Ultrasonics Symp., New York, USA, Oct. 2007, pp. 381–386.
[08] S. Ballantine, “Reciprocity in electromagnetic, mechanical, acoustical, and
interconnected systems”, in Proc. Inst. Radio Engrs., New Jersey, USA, vol. 17, no.
6, Jun. 1929, pp. 929-951.
[09] W. R. MacLean, “Absolute measurement of sound without a primary standard”,
Acoust. Soc. Am. J., vol. 12, Jul. 1940, pp. 140-146.
[10] L. L. Foldy, and H. Primakoff, “A general theory of passive linear electroacoustic
transducers and the electroacoustic reciprocity theorem. I”, Acoust. Soc. Am. J., vol.
17, Oct. 1945, pp. 109-120.
Bibliography 74
[11] E. M. McMillan, “Violation of the reciprocity theorem in linear passive

electromechanical systems”, Acoust. Soc. Am. J., vol. 18, Oct. 1946.pp. 344-347.
[12] L. L. Foldy, and H. Primakoff, “A general theory of passive linear electroacoustic
transducers and the electroacoustic reciprocity theorem. II”, Acoust. Soc. Am. J.,
vol. 19, Jan. 1947, pp. 50-58.
[13] R. J. Bobber, “General reciprocity parameter”, Acoust. Soc. Am. J., vol. 39, no.4,
Apr. 1966, pp. 680-687.
[14] F. V. Hunt, “ he analysis of transduction and its historical background”, 2nd ed.,
Harvard Univ. Press, Acoust. Soc. Am., New York, 1982.
[15] J. Hemp, “Improvements in or relating to ultrasonic flow meters”, Patent
UK2017914 A, 1979.
[16] M. L. Sanderson, and J. Hemp, “Ultrasonic flow meters – a review of the state of
the art”, in Proc. Conf. on advances in flow measurement techniques, Warwick
Univ., UK, Sep. 1981, pp. 157-178.
[17] J. Borg, “On electronics for measurement systems”, Ph.D. dissertation, Dept. Elect.
Eng., Luleå Univ., Sweden, 2010.
[18] J. Borg, J. Johansson, J. V. Deventer, and J. Delsing, “Reciprocal operation of
ultrasonic transducers: experimental results”, in Proc. Conf. IEEE Ultrasonics Symp.,
New Jersey, USA, Oct. 2006, pp. 1013-1016.
[19] Y. Bo, C. Li, and L. Yupin, “ orced oscillation to reduce zero flow error and thermal
drift for non-reciprocal operating liquid ultrasonic flow meters”, Flow Meas. Instr. J.,
vol. 22, Aug. 2011, pp. 257–264.
[20] F. M. Birleanu, “Design of RPA representation spaces for the analysis of transient
signals”, Ph.D. dissertation, Grenoble Univ., France, 2006.
[21] M. L. Sanderson, and H. Yeung, “Guidelines for the use of ultrasonic non-invasive
metering techniques”, Flow Meas. Instr. J., vol. 13, Aug. 2002, pp. 125–142.
[22] W. D. Barber, J. W. Eberhard, and S. G. Karr, “A New Time Domain Technique for
Velocity Measurements Using Doppler Ultrasound”, IEEE. Trans. Biomedical Eng.,
vol. BME-32, no. 3, pp. 213-229, Mar. 1985.
[23] Y. Bo, and C. Li, “High-speed and precise measurement for ultrasonic liquid flow
metering based on a single PGA”, in Proc. Conf. IEEE Instr. Meas. Techno.,
Singapore, May 2009. pp. 309–312.
Bibliography 75
[24] N. C. Temperley, “Optimisation of an Ultrasonic Flow Meter Based on Experimental

and Numerical Investigation of Flow and Ultrasound Propagation”, Ph.D.
dissertation, New South Wales Univ., Australia, 2002.
[25] H. D. Al-Budairi, “Design and analysis of ultrasonic horns operating in longitudinal
and torsional vibration”, Ph.D. dissertation, Glasgow Univ., Scotland, UK, 2012.
[26] C. Rosen, B. Hiremath, and R. Newnham, “Piezoelectricity”, American Institute of
Physics, New York, USA, 1992.
[27] ”IEEE Standard on Piezoelectricity”, The Institute of Electrical and Electronics
Engineers, Inc., New York, USA, 1988.
[28] P. Acevedo, and I. S. Domínguez, “Simulation of an Ultrasonic Transducer for
Medical Applications Using the Finite Element Method”, Mater. Sci. Eng. J., vol. 7,
Aug. 2015, pp. 293-297.
[29] J. Kocbach, “ inite Element Modeling of Ultrasonic Piezoelectric ransducers”,
Ph.D. dissertation, Bergan Univ., Norway, 2000.
[30] D. R. Franca, C. K. Jen, and Y. Ono, “Contrapropagating Ultrasonic Flow meter
Using Clad Buffer Rods for High Temperature measurement”, Dynamic System
Measurement and Control. J., vol. 133, no. 1, Jan. 2011, pp. 7-13.
[31] G. Kossoff, “Effects of Backing and Matching on the Performance of Piezoelectric
Ceramic ransducers”, IEEE Trans. Sonics and Ultrasonic, vol. 13, no. 1, Mar. 1966,
pp. 20-30.
[32] M. Goldfarb, and N. Celanovic, “Modeling Piezoelectric Stack Actuators for Control
of Micromanipulation”, IEEE Control Systems J., vol. 17, no. 3, Jun. 1997,
pp. 69-79.
[33] F. Wein, “ opology Optimization of Smart Piezoelectric ransducers”, Ph.D.
dissertation, Erlangen-Nürnberg Univ., Germany, 2011.
[34] J. David, and N. Cheeke, “ undamentals and Applications of Ultrasonic Waves”,
CRC Press LCC, Montreal, Qc, Canada, 2002, ch. 2, pp. 14-38.
[35] J. M. Friedt, and É. Carry, “Introduction to the quartz tuning fork”, Am. Phys. J.,
vol. 75, No. 5, May 2007, pp. 415–422.
[36] J. Ramon, G. Hernandez, and J. Bleakey, “Low-Cost, Wideband Ultrasonic
Transmitter and Receiver for Array Signal Processing Applications”, IEEE Sensors
J., vol. 11, no. 5, May 2011, pp. 1284-1292.
Bibliography 76
[37] E. Storheim, P. Lunde, and M. Vertrheim, “Diffraction Correction in Ultrasonic

Fields for Measurements of Sound Velocity in Gas. Conventional and Alternative
Methods”, in Proc. 34th Scandinavian Symposium on Physical Acoustic, Jan. 30-
Feb. 2, 2011, pp. 1–27.
[38] W. C. Zarnstorff, C. A. Castillo, and C. W. Crumpton, “A phase-shift ultrasonic flow
meter”, IRE Trans. Bio-Medical Electronics, vol. 9, no. 2, Jan. 1962, pp. 199–203.
[39] F. C. Alegria”, Bias of amplitude estimation using three-parameter sine fitting in the
presence of additive noise”, Meas. J., vol. 42, Jun. 2009, pp. 748-756.
[40] J. Šaligaa, I. Kollár, L. Michaeli, J. Bušaa, J. Liptáka, and T. Virosztek, “A
comparison of least squares and maximum likelihood methods using sine fitting in
ADC testing”, Meas. J., vol. 46, Dec. 2011, pp. 4362–4368.
[41] V. Dumbrava, and L. Svilainis, “ he Automated Complex Impedance Measurement
System”, Electron. Elec. Eng. J., vol. 76, no. 4, Mar. 2007, pp. 59-62.
[42] F. d. Silva, P. M. Ramos, and A. C. Serra, “A new four parameter sine fitting
technique”, Meas. J., vol. 35, Mar. 2004, pp. 131–137.
[43] P. M. Ramos, F. D. Silva, and A. C. Serra, “Improving sine-fitting algorithms for
amplitude and phase measurements”, in Proc. XVII IMEKO World Congress,
Dubrovnik, Croatia, Jun. 22–27, 2003, pp. 614–619.
[44] P. M. Ramos, F. d. Silva, and A. Cruz Serra, “Low frequency impedance
measurement using sine-fitting”, Meas. J., vol. 35, Jan. 2004, pp. 89–96.
[45] P. M. Ramos, and A. C. Serra, "A new sine-fitting algorithm for accurate amplitude
and phase measurements in two channel acquisition systems”, Meas. J., vol. 41, Feb.
2008, pp. 135–143.
[46] P. A. Picot, “Blood Flow Visualization and Flow Rate Estimation with Colour
Doppler Ultrasound”, Ph.D. dissertation, Western Ontario Univ., Canada, 1997.
[47] H. Øberg, “Ultrasound sensor for biomedical applications”, M.S. thesis, Tromsø
Univ., Norway, 2011.
[48] P. Lunde, K. E. Frøysa, and M. Vertrheim, “GERG Project on Ultrasonic Gas Flow
Meters, Phase II”, GERG Technical Monograph, Norway, 2000.
[49] Analog Devices Data sheet, Analog Devices, Inc., USA, 2011-2015.
[50] V. Petkus, A. Ragauskas, and P. Borodicas, “ emperature dependence of a
piezoceramic transducer electric impedance”, ULTRAGARSAS J., vol. 56, no. 3,
2005, pp. 22–25.
Bibliography 77
[51] P. Fletcher-Haynes, “Method and apparatus for correcting temperature variations in

ultrasonic flow meters”, Patent US5831175, 1998.
[52] R. J. Hazelden, and K. P. F. Smith, “Improvements in fluid monitoring”, Patent
WO2004070358, 2004.

Hamouda Assia

Uploaded by

Copyright:

Available Formats

Hamouda Assia

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Hamouda Assia

Uploaded by

Copyright:

Available Formats

Ultrasonic Flow Metering with Highly Accurate

Jitter and Offset Compensation

von der Fakultät IV

Doktor der Ingenieurwissenschaften

Vorsitzender: Prof. Dr.-Ing. Friedel Gerfers

Tag der wissenschaftlichen Aussprache: 12. Mai 2016

Die piezoelektrischen Wandler sind die kritischen Komponenten des Ultraschall-

Die erhaltenen Messergebnisse verdeutlichen die Genauigkeit und Robustheit des

Diese Arbeit liefert Ansätze für mögliche zukünftige Ultraschall-Durchflussmesser mit

Chapter 1: Introduction ..................................................................................................1

Chapter 2: The Principles of Ultrasonic Flow Meter ...................................................5

Chapter 3: Ultrasonic Piezoelectric Transducer Theory and Modeling ...................10

Chapter 5: TTD-jitter Reduction .................................................................................42

Chapter 6: TTD-Offset Cancellation ...........................................................................50

Chapter 7: Results and Analyses ..................................................................................60

Chapter 8: Conclusion ..................................................................................................72

1.2. State of the Art

1.3. Main Work Objectives

1.4. Thesis Outline

3. Chapter 3 presents the theoretical background of a piezoelectric transducer. We

5. Chapter 5 describes the applied TTD-jitter analysis and methodology used to

7. In Chapter 7, the experimental results are presented and analyzed.

8. Finally, conclusions are drawn in Chapter 8.

2.2. Doppler Flow Meter

Fig. 2.1: Doppler flow meter

2.3. Transit Time Flow Meter

2.3.1. Transit Time Flow Meter Configuration

Fig. 2.3: Flow meter pipe with in-line transducers

Fig. 2.4: Flow meter pipe with reflectors

Fig. 2.5: Flow meter pipe with diagonal transducers

2.3.2. The Principle of Transit Time Flow Meter

An accurate description of the behavior of ultrasonic transducer’s active element requires a

3.2. Piezoelectric Effect

Fig. 3.1: Piezoelectric effects: (a) direct, (b) inverse [25]

3.3. Ultrasonic Piezoelectric Transducer

The ultrasonic piezoelectric transducer is employed to convert electrical energy into

Fig. 3.2: Typical construction of an ultrasonic transducer [29]:

3.3.1. Piezoelectric Disk

Fig. 3.3: Response of a piezoelectric element to an AC voltage:

Fig. 3.4: Piezoelectric disk with thickness T and diameter D [29]

3.3.2. Piezoelectric Backing and Matching Layer

3.4. Piezoelectric Transducer Modeling

Fig. 3.5: Equivalent models of the piezoelectric element:

is the damping coefficient, and

is the angular transducer resonance frequency.

Table 3.1: Comparison of equivalent electrical and mechanical resonant circuits

describes the damping coefficient, and

is the transducer serial resonance angular frequency.

Assuming < 1 (underdamped harmonic oscillator), the complete solution of Equation

The homogeneous solution of this equation is given by:

where is the damped natural angular frequency, equal to , and the

Using Moivre’s formula, and Q(t)= , one can write:

Fig. 3.6: Amplitude of a driven damped harmonic oscillator at ζ =0.02 and

3.5. Butterworth-Van Dyke Model

The harmonic model, which is independent of either single or simultaneous excitation,

Fig. 3.8: Simulation circuit using the transducer electrical model

of 0, a slightly changed value 2 can be detected. The amplitude undergoes a similar

3.6. Transmitted and Received Signals Modeling