Post on 15-Apr-2018
High Frequency High Power Converters for
Industrial Applications
A Thesis by Publication submitted in
Partial Fulfilment of the Requirement for the
Degree of
Doctor of Philosophy
Pooya Davari
M.Sc, B.Eng (Electrical Engineering)
Science and Engineering Faculty
School of Electrical Engineering and Computer Science
Queensland University of Technology
Queensland, Australia
2013
II
III
Acknowledgement
My first and sincere appreciation goes to my principle supervisor A/Prof. Firuz Zare for the
continuous support of my PhD study and research, for his patience, motivation, enthusiasm,
and immense knowledge. His guidance helped me in all the time of research and writing of
this thesis. I could not have imagined having a better advisor and mentor for my PhD study.
He was more like a friend to me for being an open person to ideas, and for encouraging and
helping me to shape my interest and ideas.
I would like to express my deep gratitude and respect to Prof. Arindam Ghosh my associate
supervisor for his continuous help and support in all stages of my PhD study. I would also to
thank Prof. Peter O’Shea for all his support in signal processing aspect of my project.
My sincere appreciation goes to Prof. Gerard Ledwich for all I have learned from him
especially in the Power Engineering Group monthly meetings.
My PhD research was a multi-disciplinary project. Hence I was fortunate to be involved in
different projects and group activities. I do thank my colleagues and friends Mr. Ashkan
Heidarkhan Tehrani, Mrs. Negareh Ghasemi, and Mr. Meisam Babaie whom I had
contribution in part of their PhD projects. My sincere thanks also go to my colleagues and
friends at the Power Engineering Group, International Laboratory for Air Quality and Health,
and Biofuel Group for their encouragement, sharing knowledge and for providing a warm
research atmosphere.
I would like to convey my sincere thanks to Queensland University of Technology (QUT)
for financial support and providing me with a pleasant research area and laboratory facilities.
I gratefully acknowledge the Australian Research Council for financial assistance throughout
my research via the ARC Discovery Grant. The assistance of Laboratory Technicians and the
staff of the Research Portfolio were also greatly appreciated.
And most importantly, I am deeply and forever indebted to my parents for their love,
support and encouragement throughout my entire life. Without them I was totally lost and
never been succeeded.
Abstract
IV
Abstract
High power converters have found their way into many industries. Variety of industries
processes have increased their requirements in power level, economy of scale, efficiency,
new control method, new topologies and using new technologies (like new semiconductors).
The incentive to move to higher frequencies is not only given by the reduced sizes of the
passive components, magnetic core losses, but also the application demands. It is thus clear
that much can be gained from moving to a design that enables operation at very high
frequency. Hence in the recent years, high frequency high power converters have been one of
the most active areas in research.
In this PhD research project high power converters have been considered, regarding to their
power level, as two categories of high instantaneous and high average power converters.
High frequency high instantaneous power converters mostly known as pulsed power supplies
are studied as the main aspect of this research, and high frequency high average power
converters are investigated with the specific intention on high power piezoelectric transducer
applications as the second aspect.
The research was conducted at two levels; first the system level which mainly encapsulated
the circuit topology and control scheme and second the application level which not only
involves with real-world applications but also evaluate the proposed and designed power
supplies. In this regard, this PhD research was a multidisciplinary project and contribution
with different research groups was made as listed below:
Biofuel Research Group, Queensland University of Technology (QUT)
The International Laboratory for Air Quality and Health, QUT
MQ Photonics Research Centre, Macquarie University
Pulsed Power Laboratory, Kumamoto University
Ultrasound-System-Development, Fraunhofer Institute for Biomedical Eng
This thesis is divided into two main sections. The first section as the major goal of this
dissertation was to develop, design and practical implementation of effective pulsed power
supply topologies with respect to the application requirements considering the power
electronics techniques and current existing commercialized devices.
To gain insight into the principle and characteristics of pulsed power technology, variety of
topologies and applications were investigated. The main advantage of non-solid state
methods found to be utilizing gas-state and magnetic switches as they possess a very high
blocking voltage and fast rise time. However, as their drawbacks, they are bulky, unreliable,
Abstract
V
have short lifetime span and low repetition rate. Even with the magnetic switches, which have
a higher repetition rate, the problems remain. These conditions limit the mobility, efficiency
and increase the cost and the size of the pulsed power system. On the other hand, advances in
solid-state switches and exploiting power electronics techniques and topologies have led to
compact, efficient, high repetition rate and more reliable pulsed power systems. The main
drawbacks of solid-state switches are their limited power rating and operation speed.
Hence, the main objective was to remedy these two issues by exploiting power electronics
techniques and topologies. As the first step, different configuration of power converters found
beneficial both in improving power and operating speed limits. Using two 600V IGBTs
(Insulated Gate Bipolar Transistor) the experimental setup based on flyback converter was
evaluated up to 4kV. Simulation and practical implementation of the proposed method are
presented in Chapter 2 and Chapter 3.
To evaluate the performance of the proposed method the series configuration of the flyback
converter was extended up to 10 modules. The experimental setup utilizing just two 1400V
IGBTS was implemented with the capability of generating 40kV, however due to laboratory
health and safety issues it was tested up to 20kV. The results and analysis are addressed in
Chapter 3.
The proposed method depicted superior advantages both in improving the power rating and
rate of rise. However, it is load dependent and suitable for high impedance load. To remedy
this issue while preserving the obtained advantages a pulsed power topology based on
modularity concept was developed. Benefiting from cascaded switch-capacitor units and
effective control scheme a load independent pulsed power supply was proposed. Chapter 4
provides proposed method description and analysis.
To investigate real-world application characteristics and effect of pulsed power supply, two
different applications were considered. Exhaust gas treatment and analysing particle matter
(PM) mass and distribution is studied as the first application. In this regards, a DBD
(Dielectric Barrier Discharge) reactor with a multipoint to plate geometry electrodes was
gradually designed. A pulsed power setup based on push-pull converter with bipolar output
voltage was developed, which provides controllable parameters to excite the DBD load at
different operation point. The experiments were conducted on real diesel exhaust gas at
different voltage level from15kVpp to 19.44kVpp at 10kHz. An optimum operating point was
investigated with respect to PM distribution and highest PM removal efficiency. The analysis
and experimentations are described in Chapter 5.
Abstract
VI
As the second application, driving Xenon filled plasma lamp using pulsed power supply
was studied with the main intention of controlling the lamp intensity. Considering the
obtained results over wide range of voltage levels and operating frequencies, the plasma lamp
power consumption regarding to its intensity depicted nonlinear behaviour of plasma lamp
and the possibility of selecting an optimum operating point was found. Chapter 6 provided
the detailed analysis.
The second part of this project focused on developing high power converter for high power
ultrasound applications. In order to understand the specifications of required excitation signal
which needed to be generated using a power converter, characteristics of high power
piezoelectric transducer using different medium and operating conditions were studied. The
experimentation and frequency response analysis are presented in Chapter 7.
Based on the piezoelectric transducer characteristics outcomes, an efficient adaptive
algorithm along with a multi-level inverter was developed in order to increase the efficiency
of high power piezoelectric applications. The proposed algorithm is capable of detecting
piezoelectric resonance frequency variations and adapting the inverter operating frequency in
order to get high power conversion efficiency. Chapter 8 depicted the evaluation of the
proposed technique and effect of employing such algorithm on performance of an ultrasound
interface.
Finally, the conclusions and possible future works are drawn in Chapter 9. It is to be noted
that dealing with practical issues such as noise, EMI (Electro Magnetic Interference), high
voltage insulation, transformer saturation, spikes across the IGBTs, DSC (Digital Signal
Controller) programming and etc, caused vast studying in addition to the main objectives. In
addition to the gathered knowledge concerning the energy conversion of the power converter,
the experiments and investigated applications also yielded a large amount of valuable data
concerning converter development under different pulse conditions and load configurations.
This data is presented in a systematic manner as seven chapters (journal and conference
papers) so it can be of value for future related research.
VII
Keywords
Buck-boost converter
Corona
Current and voltage sources
Dielectric barrier discharge (DBD)
Digital signal controller (DSC)
Exhaust gas treatment
Flyback converter
Full bridge
Half bridge
H-bridge inverter
High power ultrasound
High voltage
Insulated-Gate Bipolar Transistor (IGBT)
Marx generator (MG)
Modular
Multi-level inverter
Non-Thermal Plasma (NTP)
Particulate Matter (PM)
Piezoelectric Transducer
Plasma
Power electronics
Pulse generator
Pulse width modulation (PWM)
Pulsed power
Push-pull converter
Rate of rise
Reactor
Resonance
Solid-state technology
Spark gap
Switch-capacitor
Unipolar and bipolar modulation
VIII
List of Abbreviation
AC: Alternating Current
ARFI: Acoustic Radiation Force Impulse
BERF: Biofuel Engine Research Facility
CCM: Continues Conduction Mode
CO2: Carbon Dioxide
CPC: Condensation Particle Counter
CRT: Cathode Ray Tube
CT: Current Transducer
DBD: Dielectric Barrier Discharge
DC: Direct Current
DCM: Discontinues Conduction Mode
DNA: Deoxyribonucleic Acid
DSC: Digital Signal Controller
EMF: Electromagnetic Metal Forming
EMI: Electro Magnetic Interference
FET: Field-Effect Transistor
FFT: Fast Fourier Transform
FWHM: Full Width at Half Maximum
GCT: Gate Commutated Thyristor
GTO: Gate Turn-off Thyristor
HIFU: High Intensity Focused Ultrasound
HV: High Voltage
IGBT: Insulated-Gate Bipolar Transistor
IGCT: Integrated Gate-Commutated Thyristor
MBL: Multistage Blumlein Lines
MG: Marx Generator
MOSFET: Metal-Oxide Semiconductor Field-Effect Transistor
MPC: Magnetic Pulse Compressors
NOX: Nitrogen Oxides
NTP: Non-Thermal Plasma
PFC: Power Factor Correction
PFN: Pulse Forming Network
IX
PM: Particulate Matter
PWM: Pulse Width Modulation
PZT: Piezoelectric Transducer
RC: Resistive-Capacitive
SCFM: Short Circuit Failure Mode
SHM: Structural Health Monitoring
SiC: Silicon Carbide
SMPS: Scanning Mobility Particle Sizer
SO2: Sulfur Dioxide
THD: Total Harmonics Distortion
UV: Ultraviolet Radiation
VOCs: Volatile Organic Compounds
VUV: Vacuum UV
X
Table of Contents
Acknowledgement ................................................................................................................. III
Abstract .................................................................................................................................. IV
Keywords .............................................................................................................................. VII
List of Abbreviation ........................................................................................................... VIII
Contributions.................................................................................................................... XXIV
List of Publications .......................................................................................................... XXVI
List of Chapters According to Publications and Contributions .............................. XXVIII
Scholarships and Awards ................................................................................................ XXIX
Statement of Original Authorship ................................................................................... XXX
Chapter 1 .................................................................................................................................. 1
1. Introduction ...................................................................................................................... 1
1.1. Definition of the Research Problem ........................................................................... 2
1.2. Literature Review....................................................................................................... 5
1.2.1. Introduction .......................................................................................................... 5
1.2.2. Pulsed Power (high frequency high instantaneous power converter) .................. 6
1.2.2.1. System Level ................................................................................................ 7
1.2.2.2. Application Level ....................................................................................... 24
1.2.3. High Frequency High Average Power Converter for Ultrasound Applications 33
1.2.3.1. Piezoelectric Transducer Characteristics .................................................... 34
1.2.3.2. Power Electronics Topologies for High Power Ultrasound Systems ......... 38
1.2.3.3. High Power Piezoelectric Transducer Applications ................................... 43
1.3. Account of Research Progress Linking the Research Papers .................................. 45
1.3.1. Introduction ........................................................................................................ 45
1.3.2. Pulsed Power Systems ....................................................................................... 45
1.3.2.1. Benefiting from Parallel and Series Configurations of Flyback Converter
for Pulsed Power Applications..................................................................................... 46
1.3.2.2. Flexible and modular pulsed power supply ................................................ 60
1.3.2.3. Pulsed power applications .......................................................................... 70
1.3.3. High frequency high power converters for ultrasound applications .................. 88
1.3.3.1. Power electronic converters for high power ultrasound transducers .......... 89
1.3.3.2. Improving the efficiency of high power piezoelectric transducer .............. 95
1.4. References ................................................................................................................... 106
XI
Chapter 2 .............................................................................................................................. 117
2 . Parallel and Series Configurations of Flyback Converter for Pulsed Power
Applications .......................................................................................................................... 117
2.1. Keywords ............................................................................................................... 118
2.2. Introduction ............................................................................................................ 118
2.3. Topology ................................................................................................................ 119
2.4. Proposed Method ................................................................................................... 122
2.4.1. Single Module .................................................................................................. 122
2.4.2. Parallel Modules .............................................................................................. 123
2.4.3. Series Modules ................................................................................................. 125
2.4.4. Operating Conditions ....................................................................................... 126
2.5. Simulation Results ................................................................................................. 127
2.5.1. Case1 ................................................................................................................ 128
2.5.2. Case2 ................................................................................................................ 129
2.5.3. Case3 ................................................................................................................ 130
2.5.4. Case4 ................................................................................................................ 130
2.6. Conclusion ............................................................................................................. 131
2.7. References .............................................................................................................. 131
Chapter 3 .............................................................................................................................. 134
3 . High Voltage Modular Power Supply Using Parallel and Series Configurations of
Flyback Converter for Pulsed Power Applications .......................................................... 134
3.1. Index Terms ........................................................................................................... 135
3.2. Introduction ............................................................................................................ 135
3.3. Topology ................................................................................................................ 137
3.4. Proposed Method ................................................................................................... 140
3.4.1. Single Module .................................................................................................. 141
3.4.2. Parallel Modules .............................................................................................. 142
3.4.3. Series Modules ................................................................................................. 142
3.4.4. Operating Conditions ....................................................................................... 144
3.5. Experimental Results and Discussion .................................................................... 146
3.5.1. Evaluating Parallel and Series Configurations ................................................ 146
3.5.1.1. Case1 ........................................................................................................ 147
3.5.1.2. Case2 ........................................................................................................ 150
3.5.2. High Voltage Modular Power Supply.............................................................. 151
3.6. Conclusion ............................................................................................................. 153
3.7. References .............................................................................................................. 154
XII
Chapter 4 .............................................................................................................................. 157
4 . A Flexible Solid-State Pulsed Power Topology ............................................................ 157
4.1. Keywords ............................................................................................................... 158
4.2. Introduction ............................................................................................................ 158
4.3. Topology ................................................................................................................ 160
4.4. Control Strategy and Operating Modes ................................................................. 161
4.4.1. Current & Voltage Control .............................................................................. 161
4.4.2. Charging & Load Control ................................................................................ 163
4.5. Simulation Results and Analysis ........................................................................... 167
4.5.1. Case1 ................................................................................................................ 167
4.5.2. Case2 ................................................................................................................ 168
4.5.3. Case3 ................................................................................................................ 168
4.6. Conclusion ............................................................................................................. 170
4.7. References .............................................................................................................. 170
Chapter 5 .............................................................................................................................. 173
5 . Effect of Pulsed Power on Particle Matter in Diesel Engine Exhaust Using a DBD
Plasma Reactor..................................................................................................................... 173
5.1. Index Terms ........................................................................................................... 174
5.2. Introduction ............................................................................................................ 174
5.3. Experimental method ............................................................................................. 176
5.3.1. Experimental Setup .......................................................................................... 176
5.3.2. DBD Reactor .................................................................................................... 178
5.3.3. Bipolar Pulsed Power Supply .......................................................................... 178
5.4. Results and Discussion .......................................................................................... 181
5.4.1. Plasma effect on PM Size Distribution ............................................................ 181
5.4.2. PM removal efficiency ..................................................................................... 187
5.4.3. Plasma Effect on PM Mass Reduction............................................................. 190
5.5. Conclusion ............................................................................................................. 191
5.6. References .............................................................................................................. 191
Chapter 6 .............................................................................................................................. 196
6 . Analysing DBD Plasma Lamp Intensity versus Power Consumption Using a Push-
Pull Pulsed Power Supply ................................................................................................... 196
6.1. Keywords ............................................................................................................... 197
6.2. Introduction ............................................................................................................ 197
6.3. Experimental Setup ................................................................................................ 198
XIII
6.3.1. DBD Lamp ....................................................................................................... 198
6.3.2. Pulsed Power Supply ....................................................................................... 199
6.3.3. Measurements .................................................................................................. 201
6.4. Results and Discussion .......................................................................................... 203
6.5. Conclusion ............................................................................................................. 205
6.6. Acknowledgement ................................................................................................. 197
6.7. References .............................................................................................................. 205
Chapter 7 .............................................................................................................................. 208
7 . Power Electronic Converters for High Power Ultrasound Transducers ................... 208
7.1. Keywords ............................................................................................................... 209
7.2. Introduction ............................................................................................................ 209
7.3. Experimental Procedure ......................................................................................... 211
7.4. Conclusion ............................................................................................................. 221
7.5. References .............................................................................................................. 222
Chapter 8 .............................................................................................................................. 224
8 . Improving the Efficiency of High Power Piezoelectric Transducers for Industrial
Applications .......................................................................................................................... 224
8.1. Introduction ............................................................................................................ 225
8.2. Methodology and Approach to Estimate Resonance Frequency ........................... 228
8.2.1. Excitation Signal .............................................................................................. 228
8.2.2. Estimating Impedance ...................................................................................... 229
8.2.3. Extracting Resonant Frequencies ..................................................................... 231
8.2.4. Noise Issues ..................................................................................................... 233
8.3. Simulation and Experiment.................................................................................... 233
8.3.1. Simulation Results ........................................................................................... 234
8.3.2. Experimental Results ....................................................................................... 236
8.3.3. Ultrasound Interface......................................................................................... 240
8.4. Conclusions ............................................................................................................ 241
8.5. References .............................................................................................................. 242
Chapter 9 .............................................................................................................................. 245
9 . Conclusions and Further Research ............................................................................... 245
9.1. Conclusions ............................................................................................................ 246
9.1.1. Improving solid-state pulsed power supplies regarding to the low power rating
and operating speed limits of the current switching devices ......................................... 246
XIV
9.1.2. Considering load characteristics and effect of the load on pulsed power supply
248
9.1.3. Increasing the efficiency of high power ultrasound applications .................... 249
9.1.4. Summary of advantages and drawbacks of proposed topologies and methods
250
9.2. Further Research .................................................................................................... 256
XV
List of Figures
Chapter 1
Fig. 1.1 General scheme of a pulsed power system. .................................................................. 7
Fig. 1.2 Basic Capacitive (a) and inductive (b) schemes for pulse generation. ......................... 8
Fig. 1.3 Pulse shape and its parameters. .................................................................................... 8
Fig. 1.4 Using multiple switches to overcome the switches power rating problem, (a) parallel
for high current and (b) series for high voltage applications. .................................................. 10
Fig. 1.5 Conventional Marx generator. .................................................................................... 11
Fig. 1.6 Conventional pulse forming network [45].................................................................. 12
Fig. 1.7 PFN with reversal voltage polarity [18]. .................................................................... 13
Fig. 1.8 MBL or adder topology [18]. ..................................................................................... 13
Fig. 1.9 Basic MPC: (a) circuit, (b) Voltage waveforms [47]. ................................................ 15
Fig. 1.10 Benefiting from diodes in MG [61] and PFN [18] topologies. ................................ 17
Fig. 1.11 Conventional Marx generator based on solid-state switches [60, 62]. ..................... 17
Fig. 1.12 Pulsed power generator based on magnetic pulse compression [10, 23]. ................ 18
Fig. 1.13 Circuit of a module of 6 MOS-FETs, connected in parallel [72]. ............................ 19
Fig. 1.14 Test circuit of MOS-FET switching unit [72]. ......................................................... 19
Fig. 1.15 High voltage switches (stack of switches) optimized for pulsed power applications:
(a) ABB [49], (b) BEHLKE [73]. ............................................................................................ 20
Fig. 1.16 Electrical circuit of the pulsed power generator using SI-thyristor [72]. ................. 21
Fig. 1.17 Waveforms of L2 current and the output voltage [72]. ............................................ 22
Fig. 1.18 (a) flyback, and (b) forward topologies [53]. ........................................................... 23
Fig. 1.19 (a) half bridge, and (b) full bridge [53]. ................................................................... 24
Fig. 1.20 Electrical equivalent circuit of a DBD load [55]. ..................................................... 25
Fig. 1.21 Photograph of a discharge induced in water [87]. .................................................... 30
Fig. 1.22 Photograph of effect of water discharges in algae treatment [27]. ........................... 31
Fig. 1.23 a) Biological cell structure, b) Double-shell model of a biological cell [30]. .......... 33
Fig. 1.24 Resonance and anti-resonance frequencies in piezoelectric impedance frequency
response.................................................................................................................................... 35
Fig. 1.25 Depiction of typical resonance frequency variations of piezoelectric devices due to
structural changes..................................................................................................................... 36
Fig. 1.26 The Van Dyke Model [107]...................................................................................... 36
XVI
Fig. 1.27 The Guan model-unloaded piezoelectric [107]. ....................................................... 37
Fig. 1.28 The extended Van Dyke model-loaded piezoelectric [107]. .................................... 37
Fig. 1.29 The complete Guan model-loaded piezoelectric [107]. ............................................ 38
Fig. 1.30 Two-stage resonant inverter based on flyback topology [126]. ............................... 39
Fig. 1.31 Two-stage resonant inverter based on H-bridge topology [5]. ................................. 40
Fig. 1.32 Effect of using multi-level waveform in harmonic elimination. .............................. 41
Fig. 1.33 Three-level PWM inverter combined with LLCC-filter for driving high power
piezoelectric transducer [102]. ................................................................................................. 42
Fig. 1.34 Typical output voltage and current of employed hybrid three-level inverter when
applying to a piezoelectric transducer [102]. ........................................................................... 42
Fig. 1.35 Basis of the ultrasonic atomizer [5]. ......................................................................... 44
Fig. 1.36 Flyback converter circuit with transformer equivalent circuit model. ..................... 47
Fig. 1.37 Flyback converter series and parallel connection (secondary side), a) Parallel, b)
Series. ....................................................................................................................................... 48
Fig. 1.38 Energy transmission in a single module flyback converter. ..................................... 48
Fig. 1.39 Effect of damping ratio on output voltage and rate of rise in a single flyback
module...................................................................................................................................... 54
Fig. 1.40 Output voltage waveform of single, parallel and series modules in Case1
(Co1=Co2=Co). .......................................................................................................................... 55
Fig. 1.41 Output voltage waveform of single, parallel and series modules in Case2
(Cop=Cos=Co). .......................................................................................................................... 56
Fig. 1.42 Output voltage waveform of single, parallel and series modules with 0.2ms delay
between modules gate signals. ................................................................................................. 56
Fig. 1.43 The effect of the load and the damping factor on the system performance. ............. 57
Fig. 1.44 Hardware setup for two flyback modules. ................................................................ 58
Fig. 1.45 (a) current sharing at the primary side for both parallel and series connections, (b)
output voltage waveform of single, parallel and series modules in Case1 (Co1=Co2=Co). ..... 59
Fig. 1.46 Block diagram of the implemented setup. ................................................................ 61
Fig. 1.47 Hardware setup for 10-series flyback modules. ....................................................... 62
Fig. 1.48 Measured output voltage of 10 series modules (VLoad) and a single module (Vo1). 62
Fig. 1.49 Proposed pulsed power topology for n level switch-capacitor units. ....................... 64
Fig. 1.50 The charging and supplying the load operating modes. ........................................... 66
Fig. 1.51 Charging and supplying the load control algorithm flowchart, a) based on individual
voltage feedback of each unit and b) based only one voltage feedback. ................................. 67
XVII
Fig. 1.52 The effect of the load impedance on the generated output voltage. ......................... 68
Fig. 1.53 Voltage waveforms with relative switching signal patterns. .................................... 69
Fig. 1.54 Voltage across the critical components. ................................................................... 69
Fig. 1.55 Schematic diagram of plasma treatment system developed at QUT engine lab. ..... 72
Fig. 1.56 DBD reactor in Solidworks: a) schematic, b) cross-section. .................................... 73
Fig. 1.57 Pulsed power supply circuit schematic diagram (push-pull inverter) ...................... 74
Fig. 1.58 A typical measured output voltage of the employed pulsed power supply. ............. 75
Fig. 1.59 Electrical Hardware setup with the DBD load. ........................................................ 75
Fig. 1.60 DBD image. .............................................................................................................. 76
Fig. 1.61 Particle Size Distribution (2000rpm, 25%Load, 19.44 kVpp). ................................ 77
Fig. 1.62 Particle Size Distribution (2000rpm, 25%Load, 17 kVpp). ..................................... 78
Fig. 1.63 Particle Size Distribution (2000rpm, 25%Load, 15 kVpp). ..................................... 79
Fig. 1.64 PM removal as a function of PM size (2000rpm, 25%Load, 19.44 kVpp) .............. 80
Fig. 1.65 PM removal as a function of PM size (2000rpm, 25%Load, 17 kVpp) ................... 80
Fig. 1.66 PM removal as a function of PM size (2000rpm, 25%Load, 15 kVpp) ................... 81
Fig. 1.67 Dielectric barrier discharge lamp ............................................................................. 84
Fig. 1.68 Pulsed power supply circuit schematic diagram (push-pull inverter) ...................... 85
Fig. 1.69 Electrical Hardware setup with the DBD load. ........................................................ 85
Fig. 1.70 V-Q cyclogram of the DBD lamp as a basis of power consumption calculation. .... 86
Fig. 1.71 Comparing different operation points regarding to the varied applied voltage levels
and repetition rates, a) applied voltage versus plasma lamp power consumption, b) applied
voltage versus plasma lamp intensity. ..................................................................................... 87
Fig. 1.72 block diagram of a lab prototype for test 1 and test 2 .............................................. 90
Fig. 1.73 (a) Input signals at 39 kHz (b) output signals at 39 kHz (c) Input signals at 61 kHz
(d) output signals at 61 kHz. .................................................................................................... 91
Fig. 1.74 (a) test 1: summation of two output signals (at 39 kHz & 61 kHz) for Vin=15V and
Vin=30V, (b) test 2: two output signals for Vin=15V and Vin=30V, (c) comparing the results
of test 1 and test 2 for Vin=15V, (d) comparing the results of test 1 and test 2 for Vin=30V. 92
Fig. 1.75 Schematic diagram of the setup using an inverter. ................................................... 92
Fig. 1.76 Hardware Setup. ....................................................................................................... 93
Fig. 1.77 Captured results of third experiment ........................................................................ 93
Fig. 1.78 Applying filter to the inverter output. ....................................................................... 94
Fig. 1.79 Captured results of forth experiment (applying a filter). .......................................... 94
Fig. 1.80 Captured results of the last experiment. ................................................................... 94
XVIII
Fig. 1.81 Exploiting online resonance frequency estimation in high power applications of
piezoelectric devices. ............................................................................................................... 97
Fig. 1.82 Frequency responses: a) ideal pulse frequency response, b) practical pulse
frequency response................................................................................................................... 98
Fig. 1.83 Extracting resonant frequencies: a) change of slope at relative minimum (resonant
frequency), b) flowchart of the proposed algorithm. ............................................................. 100
Fig. 1.84 Experimental setup for the proposed method. ........................................................ 102
Fig. 1.85 Experimental results obtained for the piezoelectric devices impedance response: 103
Fig. 1.86 Obtained results for the ultrasound interface: (a) applied uni-polar pulse at 39 kHz
in time domain, (b) frequency response of the applied uni-polar pulse (input signal), (c)
frequency response of the output signal when the input is a uni-polar pulse, (d) applied multi-
level pulse at 39 kHz in time domain, (e) frequency response of the applied multi-level pulse
(input signal), (f) frequency response of the output signal when the input is a multi-level
pulse. ...................................................................................................................................... 105
Chapter 2
Fig. 2.1 Flyback converter circuit with transformer equivalent circuit model. ..................... 121
Fig. 2.2 Operating modes in a flyback converter. .................................................................. 122
Fig. 2.3 Flyback converter series and parallel connection (secondary side), a) Parallel, b)
Series. ..................................................................................................................................... 124
Fig. 2.4 Energy transmission in a single module flyback converter. ...................................... 125
Fig. 2.5 Output voltage waveform of single, parallel and series modules in Case1
(Co1=Co2=Co). ......................................................................................................................... 129
Fig. 2.6 Output voltage waveform of single, parallel and series modules in Case2
(Cop=Cos=Co). ......................................................................................................................... 129
Fig. 2.7 Output voltage waveform of single, parallel and series modules with 0.2ms delay
between modules gate signals. ............................................................................................... 130
Fig. 2.8 The effect of the load and the damping factor on the system performance. .............. 131
Chapter 3
Fig. 3.1 Flyback converter circuit with transformer equivalent circuit model. ..................... 138
Fig. 3.2 Operating modes in a flyback converter. .................................................................. 139
Fig. 3.3 Flyback converter series and parallel connection (secondary side), a) Parallel, b)
Series. ..................................................................................................................................... 140
XIX
Fig. 3.4 Energy transmission in a single module flyback converter. ..................................... 141
Fig. 3.5 Effect of damping ratio on output voltage and rate of rise in a single flyback module.
................................................................................................................................................ 146
Fig. 3.6 Hardware setup for two flyback modules. ................................................................ 147
Fig. 3.7 (a) current sharing at the primary side for both parallel and series connections, (b)
output voltage waveform of single, parallel and series modules in Case1 (Co1=Co2=Co). ... 149
Fig. 3.8 Voltage across the switches in Case1. ...................................................................... 150
Fig. 3.9 Voltage across the diodes in Case1. ......................................................................... 150
Fig. 3.10 Output voltage waveform of single, parallel and series modules in Case2
(Cop=Cos=Co). ........................................................................................................................ 151
Fig. 3.11 Block diagram of the implemented setup. .............................................................. 152
Fig. 3.12 Hardware setup for 10-series flyback modules. ..................................................... 152
Fig. 3.13 Measured output voltage of 10 series modules (VLoad) and a single module (Vo1).
................................................................................................................................................ 153
Chapter 4
Fig. 4.1 Proposed pulsed power topology for n level switch-capacitor units. ....................... 160
Fig. 4.2 Operating modes for positive buck-boost converter. ................................................ 162
Fig. 4.3 The charging and supplying the load operating modes. .......................................... 164
Fig. 4.4 Charging and supplying the load control algorithm flowchart (based on individual
voltage feedback of each unit). .............................................................................................. 165
Fig. 4.5 Charging and supplying the load control algorithm flowchart (based on only one
voltage feedback). .................................................................................................................. 166
Fig. 4.6 The effect of the load impedance on the generated output voltage.. ........................ 168
Fig. 4.7 Voltage waveforms with relative switching signal patterns. .................................... 169
Fig. 4.8 Voltage across the critical components. ................................................................... 169
Chapter 5
Fig. 5.1 Schematic diagram of plasma treatment system developed at QUT engine lab. ..... 177
Fig. 5.2 DBD reactor in Solidworks: a) schematic, b) cross-section. .................................... 178
Fig. 5.3 Pulsed power supply circuit schematic diagram (push-pull inverter) ...................... 179
Fig. 5.4 A typical measured output voltage of the employed pulsed power supply. ............. 180
Fig. 5.5 Electrical Hardware setup with the DBD load. ........................................................ 180
XX
Fig. 5.6 The measured electrical parameters: a) output voltages across the DBD load, b)
output current, and c) voltage stress across the switch. ......................................................... 182
Fig. 5.7 DBD image. .............................................................................................................. 183
Fig. 5.8 V-Q cyclogram of the DBD load as a basis of power consumption calculation. ..... 183
Fig. 5.9 Particle Size Distribution (2000rpm, 25%Load, 19.44 kVpp). ................................ 185
Fig. 5.10 Particle Size Distribution (2000rpm, 25%Load, 17 kVpp). ................................... 186
Fig. 5.11 Particle Size Distribution (2000rpm, 25%Load, 15 kVpp). ................................... 187
Fig. 5.12 PM removal as a function of PM size (2000rpm, 25%Load, 19.44 kVpp) ............ 188
Fig. 5.13 PM removal as a function of PM size (2000rpm, 25%Load, 17 kVpp) ................. 188
Fig. 5.14 PM removal as a function of PM size (2000rpm, 25%Load, 15 kVpp) ................. 189
Chapter 6
Fig. 6.1 Dielectric barrier discharge lamp ............................................................................. 199
Fig. 6.2 Pulsed power supply circuit schematic diagram (push-pull inverter) ...................... 200
Fig. 6.3 Electrical Hardware setup with the DBD load. ........................................................ 200
Fig. 6.4 Output voltage of the employed pulsed power supply: a) at 10 kHz, b) typical
measured output voltages at 10 kHz and 5 kHz. .................................................................... 201
Fig. 6.5 V-Q cyclogram of the DBD lamp as a basis of power consumption calculation. .... 202
Fig. 6.6 Comparing different operation points regarding to the varied applied voltage levels
and repetition rates, a) applied voltage versus plasma lamp power consumption, b) applied
voltage versus plasma lamp intensity. ................................................................................... 204
Fig. 6.7 The DBD lamp light intensity measured for varied applied voltage levels.............. 204
Chapter 7
Fig. 7.1 (a) Van Dyke Model, (b) a power converter, (c) output voltage of power converter.
................................................................................................................................................ 211
Fig. 7.2 A block diagram of a lab prototype for test 1 and test 2. ......................................... 213
Fig. 7.3 (a) input signals at 39 kHz (b) output signals at 39 kHz (c) input signals at 61 kHz
(d) output signals at 61 kHz ................................................................................................... 214
Fig. 7.4 (a) Test 1: summation of two output signals (at 39 kHz & 61 kHz) for Vin=15V and
Vin=30 V (b) Test 2: two output signals for Vin=15V and Vin=30 V. (c) Comparing
the results of test 1 and test 2 for Vin=15 V. (d) Comparing the results of test 1 and test 2 for
Vin=30 V. .............................................................................................................................. 215
Fig. 7.5 The experimental setup. ............................................................................................ 216
XXI
Fig. 7.6 A block diagram of test 3. ........................................................................................ 216
Fig. 7.7 The results of test 3 (a) input signals and (b) output signals. ................................... 217
Fig. 7.8 A block diagram of test 4. ........................................................................................ 218
Fig. 7.9 The results of test 4 (a) input signals and (b) output signals. ................................... 219
Fig. 7.10 A block diagram of test 5. ...................................................................................... 219
Fig. 7.11 The results of test 5 (a) input signals and (b) output signals. ................................. 220
Fig. 7.12 A block diagram of test 6. ...................................................................................... 220
Fig. 7.13 The results of test 6 (a) input signals and (b) output signals. ................................. 221
Chapter 8
Fig. 8.1 Effect of using multi-level waveform in harmonic elimination. .............................. 227
Fig. 8.2 Exploiting online resonance frequency estimation in high power applications of
piezoelectric devices. ............................................................................................................. 228
Fig. 8.3 a) Ideal pulse frequency response, b) Captured voltage c) Captured response
(current), d) Practical pulse frequency response. ................................................................... 230
Fig. 8.4 Extracting resonant frequencies: a) change of slope at relative minimum (resonant
frequency), b) flowchart of the proposed algorithm. ............................................................. 232
Fig. 8.5 Simulation results for piezoelectric Type A: a) circuit model, b) impedance response.
................................................................................................................................................ 235
Fig. 8.6 Experimental setup for the proposed method. .......................................................... 237
Fig. 8.7 Experimental results obtained for the piezoelectric devices impedance response: (a)
Type A, (b) Type B. .............................................................................................................. 238
Fig. 8.8 Obtained results for the ultrasound interface: (a) applied uni-polar pulse at 39 kHz in
time domain, (b) frequency response of the applied uni-polar pulse (input signal), (c)
frequency response of the output signal when the input is a uni-polar pulse, (d) applied multi-
level pulse at 39 kHz in time domain, (e) frequency response of the applied multi-level pulse
(input signal), (f) ) frequency response of the output signal when the input is a multi-level
pulse. ...................................................................................................................................... 241
XXII
List of Tables
Chapter 1
Table 1.1 Various switch parameters ....................................................................................... 16
Table 1.2 Summary of typical pulsed power applications requirements [30]. ........................ 26
Table 1.3 Dielectric strength of common materials. ................................................................ 28
Table 1.4 Output voltage and rate of rise when Co1=Co2=… CoN=Co. ..................................... 51
Table 1.5 Output voltage and rate of rise when Cop=Cos=Co. ............................................... 51
Table 1.6 Simulation Parameters ............................................................................................. 55
Table 1.7 Simulation Parameters in Case 1 and 2 ................................................................... 55
Table 1.8 Simulation Parameters ............................................................................................. 67
Table 1.9 PM Mass Reductions at Different Voltage Levels .................................................. 82
Table 1.10 Considered voltages and frequencies for different experiments ............................ 87
Table 1.11 Test conditions and setups ..................................................................................... 90
Table 1.12 Estimated resonant frequencies of the Type A piezoelectric device ................... 103
Table 1.13 Estimated resonant frequencies of the Type B piezoelectric device ................... 104
Chapter 2
Table 2.1 Output voltage and rate of rise when Co1=Co2=… CoN=Co..................................... 126
Table 2.2 Output voltage and rate of rise when Cop=Cos=Co. .............................................. 126
Table 2.3 Simulation Parameters ............................................................................................ 128
Table 2.4 Simulation Parameters in Case 1 and 2 .................................................................. 128
Chapter 3
Table 3.1 Output voltage and rate of rise when Co1=Co2=… CoN=Co. ................................... 143
Table 3.2 Output voltage and rate of rise when Cop=Cos=Co. ................................................ 143
Chapter 4
Table 4.1 Simulation Parameters ........................................................................................... 168
Chapter 5
Table 5.1 PM Mass Reduction at Different Voltage Levels .................................................. 190
Chapter 6
Table 6.1 Considered voltages and frequencies for different experiments ............................ 203
XXIII
Chapter 7
Table 7.1 Test conditions and setups ..................................................................................... 213
Chapter 8
Table 8.1 Values of the Components in the Electrical Circuit Model ................................... 236
Table 8.2 Estimated Resonant Frequencies in Simulation..................................................... 236
Table 8.3 Estimated resonant frequencies of the Type A piezoelectric device ..................... 238
Table 8.4 Estimated resonant frequencies of the Type B piezoelectric device ..................... 239
Table 8.5 Survey on advantages and drawbacks of mentioned methods ............................... 239
XXIV
Contributions
Effective Flyback Converter Configurations for Pulsed Power
Applications
Improving the voltage level and rate of rise
Studying the effect of load impedance on system performance
Investigating the fault conditions
Developing a High Voltage Modular Pulsed Power Supply
Applying modularity concept on flyback topology
Studying the effect of load impedance on system performance
Investigating the performance of parallel and series
configurations
Reducing the number of switches to min imum
Improving the voltage level and rate of rise
Ability to produce up to 40kV pulses
Easy diagnostic due to modularity concept
Proposing a Flexible Pulsed Power Topology
Utilizing cascaded switch-capacitor topology
An efficient load independent topology
Studying the effect of load impedance on the system performance
Using modularity concept
Ability to operate at high repetition rate
Developing a smart control algorithm which can control the flow
of energy and prevent from any possible faults
Easy diagnostic
Effect of Pulsed Power on Particle Matter Distribution
Efficient method for exhaust gas treatment
Developing a DBD load
Studying the effect of pulsed power at different voltage levels
Optimizing the system operation
Studying the particle matt er distribution and mass variations
XXV
Developing High Frequency High Power Converter for Plasma
Lamp Applications
Studying the plasma lamp power consumption regarding to its
intensity
Investigating DBD plasma lamp characteristics at varied power
levels
Designing a power converter for plasma lamp applications with
variable voltage and frequency
Studying Desirable Characteristics of Power Converters for High
Power Ultrasound Applications
Investigating high power piezoelectric transducer requirement s
Studying linear or non-linear behaviour of high power
piezoelectric transducer
Investigating effect of different medium on ultrasound system
performance
Improving the efficiency of power converter based on high power
piezoelectric transducer behaviour variati ons
Applying an efficient frequency adaptation algorithm
Developing frequency detection algorithm
Studying the behaviour of piezoelectric transducer
Increasing the performance of high power ultrasound system
XXVI
List of Publications
The Queensland University of Technology (QUT) allows the presentation of a thesis for the
degree of Doctor of Philosophy in the format of published or submitted papers, where such
papers have been published, accepted or submitted during the period of candidature. This
thesis is composed of seven published/submitted papers, of which six have been published
and one is accepted. Note that each paper is selected for the thesis as one chapter.
Peer Reviewed Journal Articles:
1. P. Davari, F. Zare, A. Ghosh, and H. Akiyama, "High Voltage Modular Power
Supply Using Parallel and Series Configurations of Flyback Converter for Pulsed
Power Applications," Plasma Science, IEEE Transactions on, vol.40, no.10, pp.2578-
2587, Oct. 2012.
2. P. Davari, N. Ghasemi, F. Zare, P. O'Shea, and A. Ghosh, "Improving the efficiency
of high power piezoelectric transducers for industrial applications," Science,
Measurement & Technology, IET, vol. 6, pp. 213-221, 2012.
3. M. Babaie, P. Davari, F. Zare, Z. Ristovski, and R. Brown, "Effect of Pulsed Power
on Particle Matter in Diesel Engine Exhaust Using a DBD Plasma Reactor," Plasma
Science, IEEE Transactions on, vol.41, no.8, pp.2349-2358, Aug. 2013.
Peer Reviewed International Conference Articles:
4. P. Davari, F. Zare, and A. Ghosh, “Analyzing DBD Plasma Lamp Intensity versus
Power Consumption Using a Push-Pull Pulsed Power Supply", EPE, Lille, France,
2013 (accepted).
5. P. Davari, F. Zare, and A. Ghosh, “A Flexible Solid-State Pulsed Power Topology",
International Power Electronics and Motion Control (EPE-PEMC), 2012 the 15th
IEEE Conference on, 2012.
6. P. Davari, F. Zare, and A. Ghosh, “Parallel and Series Configurations of Flyback
Converter for Pulsed Power Applications", Industrial Electronics and Applications
(ICIEA), 2012 the 7th IEEE Conference on, 2012.
XXVII
7. N. Ghasemi, F. Zare, P. Davari, P. Weber, C. Langton, and A. Ghosh, “Power
Electronic Converters for High Power Ultrasound Transducers", Industrial Electronics
and Applications (ICIEA), 2012 the 7th IEEE Conference on, 2012.
XXVIII
List of Chapters According to Publications and Contributions
High Frequency High Power Converters for
Industrial Applications
Literature Review
"Stage two, confirmation and final thesis"
Research Problem #1
Power Electronics Converters for Pulsed
Power Topologies
Chapter 2:
Applying Flyback Converter for Pulsed Power
"Parallel and Series Configurations of Flyback Converter for Pulsed Power Applications"
The 7th IEEE Conference on Industrial Electronics and Applications ICIEA2012
Chapter3:
High Voltage Modulare Pulsed Power Supply
"High Voltage Modular Power Supply Using Parallel and Series Configurations of Flyback
Converter for Pulsed Power Applications"
IEEE Transactions on Plasma Science, vol.40, no.10, pp.2578-2587, 2012 Chapter4:
Flexible Pulsed Power Supply
"A Flexible Solid-State Pulsed Power Topology"
The 15th IEEE Conference on Power Elctroncis and Motion Control, EPE-PEMC 2012
Research Problem #2
Investigating Effect of Pulsed Power in Different
Applications
Chapter 5: Effect of Pulsed Power on Particle Matter
Distribution
"Effect of Pulsed Power on Particle Matter in Diesel Engine Exhaust Using a DBD Plasma
Reactor"
IEEE Transactions on Plasma Science, vol. 41, no.8, pp.2349-2358, 2013
Chapter 6:
High Frequency High Voltage Converter for Plasma Lamp Applications
"Analyzing Intensity Versus Power Consumption in DBD Plasma Lamp Using a High Frequency Push-
Pull Converter"
15th European Conference on Power Electronics and Applications, EPE'13 ECCE Europe
Research Problem #3
High Frequency High Power Converters for High Power Piezoelectric Transducers
Chapter 7:
Power Converter for High Power Ultrasound Applications
"Power Electronic Converters for High Power Ultrasound Transducers"
The 7th IEEE Conference on Industrial Electronics and Applications ICIEA2012
Chapter 8:
Desgining a Power Converter based on High Power Piezoelectric Charactrestics
"Improving the Efficiency of High Power Piezoelectric Transducers for Industrial
Applications"
IET Sci. Meas.Technol, vol.6, no.4, pp.213-221, 2012
XXIX
Scholarships and Awards
Queensland University of Technology Postgraduate Research Award
(QUTPRA) for PhD degree for three years (2010-2013).
Tuition Fee Waiver Scholarship by Queensland University of Technology
for PhD degree for three years (2010-2013).
Two Supervisor Top-Up Scholarships for PhD degree, one with the
period of 1.5 year (2011-2013) and one for the period of one year (2012-
2013).
Outstanding HDR Student of the Month, Science and Engineering
Faculty, Queensland University of Technology, June 2012.
Queensland Safe Work Awards (2011)
QUT was highly commended for best solution to an identified electrical safety issue.
I was being part of the team in designing a prototype acrylic LV testing enclosure to
house circuiting and prevent contact with exposed terminals. The LV test box is
a flexible solution that reduces the risk of potential electrical incidents.
QUT grant-in-aid for attendance in ICIEA conference in Singapore, 2012.
QUT grant-in-aid for attendance in EPE-PEMC conference in Serbia,
2012.
XXX
Statement of Original Authorship
I declare that the work contained in this thesis has not been previously submitted to meet
requirements for an award at this or any other higher education institution. To the best of my
knowledge and belied, the thesis contains no material previously published or written by
another person except where due reference is made.
Pooya Davari Date: 21 Feb 2013
QUT Verified Signature
1
Chapter
1. Introduction
1
Chapter 1
2
1.1. Definition of the Research Problem
High power converters have found their way into many industries. Variety of industries
processes have increased their requirements in power level, economy of scale, efficiency,
new control method, new topologies and using new technology (like new semiconductors).
The incentive to move to higher frequencies is not only given by the reduced sizes of the
passive components, magnetic core losses, but also the application requirements. It is thus
clear that much can be gained from moving to a design that enables operation at very high
frequency. Hence in the recent years, high frequency high power converters have been one of
the most active areas in research [1-7].
Generally, high power converters regarding to their power level can be divided into two
different categories of high instantaneous and high average power converters. High
instantaneous power converters are mostly known as pulsed power supplies. The need for
generating high instantaneous power pulses was started with defence-related applications.
But over the last two decades, non-military applications of pulsed power technology have
been gained lot of interest. More than one hundred possible applications can now be listed [2,
8-43].
Varieties of topologies have been introduced and used for pulsed power supplies such as
Marx generators (MG) [18, 44, 45], pulse forming network (PFN) [18, 45, 46], magnetic
pulse compressors (MPC) [18, 45, 47, 48], and multistage Blumlein lines (MBL) [18, 45].
The main advantage of these methods (non-solid state) is utilizing gas-state and magnetic
switches as they possess a very high blocking voltage and fast rise time [10, 30]. However,
Gas-state switches require special operating conditions such as high pressure, vacuum
equipments and gas supplies. In addition, they are bulky, unreliable, have short lifetime span
and low repetition rate. Even with the magnetic switches, which have a higher repetition rate,
the problems remain. These conditions limit the mobility, efficiency and increase the cost and
the size of the pulsed power system [10, 30].
There have been great improvements in the pulsed power area in the recent years. Advances
in solid-state switches and exploiting power electronics techniques and topologies have led to
compact, efficient, high repetition rate and more reliable pulsed power systems. The main
drawbacks of solid-state switches are their limited power rating and operation speed. The new
developed solid-state switches such as Insulated-Gate Bipolar Transistor (IGBT), Integrated
Gate-Commutated Thyristor (IGCT) and Silicon Carbide Devices (SiC) have high power
rating [30, 49-51], but their lower operation speed comparing with the gas-state switches and
Chapter 1
3
high cost still put limits in the pulsed power supplies. From these circumstances, Problem #1
arises:
1. Problem # 1: Solid-state pulsed power supplies drawbacks including low blocking
voltage of switching device, operating speed, and lack of flexibility
One way to increase the pulsed power supply performance and cover the switch
limits is to explore alternative circuit topologies. Since the above mentioned
topologies carries complexity, inflexibility and inefficiency as their main
drawbacks, applying power electronics topologies and techniques to remedy these
problems was one of the major objectives in this research project. Power electronic
topologies are considered not only as an alternative way to overcome the switch
limits but also in developing flexible and compact systems. Thus, several power
electronics topologies and control methods are studied and investigated in order to
optimize them for supplying pulsed power applications with maximum flexibility.
In addition to switching devices and circuit topology, power converter requirements are
dependent on the characteristics of the load as well. In other word this is the application
which defines the power supply specifications. Pulsed power applications present one of the
most varied ranges of loads in terms of load behavior and impedance. Hence, various load
conditions adversely affect the power supply flexibility which leads to the second research
problem:
Problem # 2: Load dependency problem
To remedy this problem, two different approaches can be studied. The first is to
design a pulsed power supply based on the application characteristics. This requires
clear study of the application or load behaviour, which can be results in a most
appropriate power supply for that specific application. The second method is to
design a flexible pulsed power supply. Here, flexibility stands for load
independency, which means to design a pulsed power supply which can cover wide
range of applications. Hence, to cover both methods varieties of pulsed power
applications were considered in order to study their behaviour and effects on the
pulsed power supply.
Chapter 1
4
The high frequency high average power converters (second category) in this project were
specifically considered for high power piezoelectric transducer applications. The critical
behaviour of a piezoelectric device is encapsulated in its resonance frequencies because of its
maximum transmission performance at these frequencies. An ideal scenario is to have a
sinusoidal excitation signal at the resonant frequency. But at high power and high frequency,
which is the focus area of this thesis, generating pure sinusoidal signals is not possible. The
most efficient way is to benefit from high power high frequency converters to excite the
piezoelectric device properly. This arises the last research problem:
Problem # 3: Increasing the efficiency of high power piezoelectric application
Piezoelectric devices typically have multiple resonant frequencies, but only the
major resonant frequency is generally targeted for excitation in practice. Structural
and environmental changes of a piezoelectric system can affect variations in the
resonant frequencies [5, 52]. Therefore it is important to estimate the main resonant
frequency in order to maintain efficient system operation. Regarding this issue,
piezoelectric behaviour and circuit model was fully investigated in order to optimize
the power supply. In addition, as the performance of the piezoelectric application
increases by adapting the power converter switching frequency with variations in
piezoelectric resonance frequency different frequency estimation methods were
studied.
Chapter 1
5
1.2. Literature Review
1.2.1. Introduction
High power converters have been employed in industries for many years. However, in recent
years, developing and investigating high power converters have gained substantial interest, as
the variety of industries processes have increased their requirements in power level, economy
of scale, efficiency, new control method, new topologies and using new technology. In
addition, interest in operating at higher frequencies has increased as well. The inducement in
moving to higher frequency levels is not only given by the reduced sizes of the passive
components, magnetic core losses, but also the application demands. It is thus clear that much
can be gained from moving to a design that enables operation at high frequency. Hence in the
last decade, high frequency high power converters have been one of the most active areas in
research [1-7].
Since the advantages of employing power electronics technologies have been exploited,
varied types of power converters are introduced and manufactured. The power converters
have been categorized in varied ways regarding to the application requirements and
specifications. One can classified the power converters based on the voltage level, frequency
level or power level. As the main goal of this PhD research was high power converters and
due to the presence of varied applications, in this research high power converters are divided
into two different categories:
High frequency high instantaneous power converters
High frequency high average power converters
High frequency high instantaneous power converters are mostly known as pulsed power
supplies. Pulsed power is rapid release of stored energy as electrical pulses into a load, which
can result in delivery of large amounts of instantaneous power over a short period of time.
The need for generating high instantaneous power pulses was started during the Second
World War for radar application. From that time on, the defence-related applications were
one of the key driving forces behind pulsed power technology, primarily in connection with
nuclear weapons simulation, applications of high-power microwave sources, high-power
laser sources, electromagnetic guns, etc. In addition, over the last two decades, non-military
applications of pulsed power technology have been studied. More than one hundred possible
applications can now be listed [2, 8-43].
High frequency high average power converters (second category) are considered
specifically for high power ultrasound application in this research. One of the major issues
with high power ultrasound systems is the short life-time span of the piezoelectric transducer
Chapter 1
6
which cannot withstand such high power signals for a long time. Recently applying power
electronics techniques for high power ultrasound applications has gained interest, as the life
time of the piezoelectric transducer increases due to the fact that the power supply is not
applying the high power signal continuously and it is in the pulse manner. But the big
challenge is how to design a power supply which can generate a proper excitation signal for
the piezoelectric device. The critical behavior of a piezoelectric device is encapsulated in its
resonance frequencies because of its maximum transmission performance at these
frequencies. The resonance frequencies fall within the range of ultrasound which means
above 20kHz. Hence high frequency high average power converters need to be considered
(second category).
1.2.2. Pulsed Power (high frequency high instantaneous power
converter)
The need for generating high instantaneous power pulses was started during the Second
World War for radar application. From that time on, the defense-related applications were
one of the key driving forces behind pulsed power technology, primarily in connection with
nuclear weapons simulation, applications of high-power microwave sources, high-power
laser sources, electromagnetic guns, etc. In addition, over the last two decades, non-military
applications of pulsed power technology have been studied. Pulsed power supplies are a key
component in systems that the load needs to be pulsed.
Pulsed power is rapid release of stored energy as electrical pulses into a load, which can
result in delivery of large amounts of instantaneous power over a short period of time [44,
45]. Interest in pulsed power technology has been growing extremely fast since 1923, the
year that Erwin Marx invented a very high voltage generator (Marx generator) [18]. Since
then, pulsed power technologies have found variety of applications; especially it is subjected
to the clean technology.
There have been great improvements in the pulsed power area in the recent years. The
highest energy and power that have been achieved in a single pulse are at present of the order
of 100MJ and a few hundred terawatts respectively. The highest voltage and current
amplitudes are obtained at 50MV and 10MA, respectively [44]. In addition to its power and
energy, the pulse rise time, fall time and repetition rate are also important, which are achieved
in a fraction of nanosecond and few MHz. Setting these features is completely depending on
the pulsed power system sections.
Chapter 1
7
Generally, as the Fig. 1.1 shows, a pulsed power system consists of three main sections of
energy storage, pulse generator and the load. Considering this, study and design of a pulsed
power system can be considered at two levels (Fig. 1.1):
System Level
Application level
Fig. 1.1 General scheme of a pulsed power system.
In order to have a high performance system, careful consideration of each of these levels is
needed. Therefore, each level is described in the following sections.
1.2.2.1. System Level
This level consists of two important parts of energy storage and pulse generator. At this
level the main concern is to accumulate and store the energy and finally release the stored
energy rapidly to get required pulses. The whole procedure is accomplished by benefiting
from the different circuit topologies, components and control algorithms.
1.2.2.1.1. Background
Energy storage can be capacitive, inductive [44]. Fig 1.2 illustrated simple schemes of
capacitive and inductive storage pulsed power supplies.
In Fig 1.2(a), first the switch S is open and in this interval energy should be stored from a
high voltage power supply in the capacitor. In the second step the switch turns on and energy
stored in capacitor transfers to the load. In the second scheme, Fig 1.2(b), first S1 is close and
S2 is open. In this stage the inductor is charging. In the second stage, S1 and S2 turn off and
on, respectively. Therefore, the stored energy in the inductor transfers to the load.
As mentioned before, despite the voltage level, there are other pulse characteristics which
are quite important. Fig 1.3, shows the typical pulse wave form that appears across the load.
In this figure three important features of rise time, pulse width and fall time are depicted. The
pulse rise time, which is defined as the time it takes the voltage to rise from 10% to 90%, is
the time that stored energy suddenly released to the load. This means that the voltage across
Chapter 1
8
the load rises from 0 to a certain level in a time interval. Hence, this time interval completely
depends on the switch characteristics. Some applications require a fast rise time (in a fraction
of nanosecond) otherwise the required phenomena never happens and the load may be
damage, such as applying pulsed power to a living cell in biomedical applications [10, 21, 26,
29, 30].
LoadC
S
Pulse Generator
Capacitive
Storage
HV
Power
Supply
(a)
S2
LoadS1
L
Inductive
Storage
Pulse Generator
(b)
Fig. 1.2 Basic Capacitive (a) and inductive (b) schemes for pulse generation.
Fig. 1.3 Pulse shape and its parameters.
Chapter 1
9
Pulse energy depends on both its duration, sometimes mentioned as FWHM (Full Width at
Half Maximum), and its amplitude. The pulse duration depends on the amount of the stored
energy and the load impedance. This means that if the stored energy is low regarding the load
impedance then the pulse amplitude should increased so that the required energy transferred
to the load, otherwise the expected phenomenon won’t happen. Therefore, the energy should
be tuned in a way that it satisfies the load demands.
Except for the pulsed power supplies which generate pulses based on the fall time, the load
defines the fall time. In some applications the load has resistive characteristics [18, 38, 53]
that a spark happens at the output while in some applications the capacitive behavior of the
load [35, 41, 53-56] causes the output voltage never decreases and an efficient way to
discharge the power supply is required.
The final feature of the generated pulse is the repetition rate. The repetition rate depends on
storage unit and switches. If the storage unit is not able to recharge in a required time then it
decreases the repetition rate of the system. Same thing happens when the switches are not fast
enough. Normally, when the voltage level increases the repetition rate decreases. That is why
the extra high voltage systems are working in the single shot mode [18].
Regarding the above mentioned features, the most prominent part of a pulsed power system
is the pulse generator, which is based on the utilized switch and topology. Hence, the pulse
generator is the connecting part between the storage and the load. This means that the switch
and employed topology power capability and operating speed define the whole system
characteristics.
Due to switches power rating limits, in some applications it is required to use multiple-
switch base circuit. A conventional way to utilize multiple switches is to directly stack them
in series or in parallel, as shown in Fig 1.4. The series and parallel connections can be used in
generation of high level of voltage and current, respectively [18, 57]. For high power
demands, multiple switches are stacked in series and parallel.
The implementation of series stack requires careful attention. Differences in switches
characteristics and their drive circuit should be considered in a way that they get
synchronized. Moreover, the system has to be able to handle the failure of the switches.
These problems decrease the interest of using this configuration.
Chapter 1
10
LoadC
S2
S3
S1
(a)
LoadC
S2 S3S1
(b)
Fig. 1.4 Using multiple switches to overcome the switches power rating problem, (a) parallel for high current
and (b) series for high voltage applications.
1.2.2.1.2. Non Solid-State Pulsed Power
The simple schemes mentioned above become ineffective if it is required to produce
voltage pulses of amplitude V, for the lack of capacitors and switches designed for
such high voltages. In this case, voltage multiplication schemes are applicable [45]. In this
section the most known non solid-state pulsed power topologies are reviewed.
Marx Generator (MG)
Marx generator is one of the most common topologies used in variety of applications [18,
44, 45] (Fig 1.5). In this circuit, N capacitors are connected in parallel and charged through
resistors to a voltage V0. If all switches close simultaneously, capacitors become connected in
series and a voltage pulse with the voltage level of close to NV0 appears across the load. The
total capacitance value versus the load will be C/N, and hence the pulse FWHM will be
⁄ .
Chapter 1
11
Load
C CS
R
R
CS
R
R
S
R
C
R0
Fig. 1.5 Conventional Marx generator.
This topology was proposed by Marx in 1923 for high-voltage generation based on the
spark gap switches. The spark gap switch can be categorized in gas-state or non solid-state
switches. The high voltage and power rating of this switch is its main advantage. The main
benefit of this generator is that the multiple switches will be synchronized automatically. The
closing of the first switch leads to an overvoltage across the other switches that are not yet
closed. Subsequently, this overvoltage forces them to close.
In the conventional MG the resistors must be high to prevent the discharging of the
capacitors. The high value of the resistance limits the pulse generation rate due to the long
charging time of the capacitors besides being a loss component [44, 45]. To overcome the
drawbacks of energy loss and long charging time, a modified version of the MG has been
introduced, which the resistors are replaced by inductors [45]. The only problem with this
topology is tuning the capacitors charging time due and resonance between the inductors and
the capacitors.
Regarding the above mentioned features, the MG is used extensively in pulsed power
application, and many different Marx based systems have been developed.
Pulse Forming Network (PFN)
A more efficient method of voltage multiplication was proposed by Mesyats [18, 45, 46].
This topology is based on L-C ladder network. This scheme is known as pulse forming
network and it consists of N stages, each containing an oscillatory LC circuit and a spark gap
switch (Fig 1.6). The component values should be as follows:
(1-1)
Chapter 1
12
C0 is the capacitance of the smoothing filter of the rectifier. Resistors are
connected in the circuit to ensure complete discharging of all LC-circuit capacitors. The
resistors are generally selected as:
√
⁄ (1-2)
where i is the LC-circuit number.
V0
C0
L1S1
C1 R1
L2S2
C2 R2
L3S3
C3 R3
LNSN
LoadCN RN
HV
Power
Supply
Fig. 1.6 Conventional pulse forming network [45].
As the switch S1 operates, the capacitor C0 charged to a voltage of V0, discharges into the
capacitor C1. Considering a lossless circuit the voltage across the C1:
[ (
√
)] (1-3)
Considering the above equation, at time √ the maximum voltage V1max is
equal to 2V0. If the switch S2 closes at the time t1, then regarding equation (1-1) C1
discharges into C2 much faster than into C0. In √ , the voltage across the C2
becomes . Therefore, if switch closes at proper time then eventually the
maximum voltage across CN will be:
(1-4)
Therefore, this topology is more efficient as it generates higher voltage level with same
number of switches comparing with MG.
Different combinations of the LC-circuit have been addressed in the literature in order to
increase the performance of the PFN. For example in 1964 just one year after the
conventional PFN has been introduced [18, 45, 58], the modified version of the conventional
PFN was introduced (see Fig. 1.7). The output voltage can reach to the same level of voltage
as MG but the number of the switches is reduced by a factor of 2.
Chapter 1
13
Fig. 1.7 PFN with reversal voltage polarity [18].
Multi-Stage Blumlein Lines (MBL)
The third alternative for voltage multiplication is inductive adder which is also known as
multi-stage Blumlein lines [18, 45]. As shown in Fig 1.8 several pulse transformers with the
secondary side connected in series are used. The advantage of this topology is benefiting
from transformer turns ratio in order to increase the voltage. Secondly, as each core is
independent the currents will naturally be shared between all switches. Therefore, low power
rate switches can be used in this topology. The output voltage will be the sum of all voltages
in the primary side multiply by the transformer turns ratio.
Fig. 1.8 MBL or adder topology [18].
Chapter 1
14
Magnetic Pulse Compressor (MPC)
Magnetic pulse compressor behaviour is similar to PFN because both operate based on the
resonant converter. The only difference is that the MPC is used to compress the generated
pulse to obtain fast rise and fall times. Therefore, it can be used as the final stage in a pulse
generator unit. This topology uses magnetic switches instead of gas-state switches. Figure 9
shows a typical MPC circuit [47]. Suppose that C1 = C2 = C3 = C and MSs L1 MSu. MSu
and MSs refer to the unsaturated inductance and the saturated inductance of the MS,
respectively. On switch closure with the above conditions, the charged energy in C1, which is
initially charged to a potential V0, is transferred to C2 by C−L−C resonance. As the potential
on C2 reaches a point at which MS will saturate, the energy transfer takes place from C2 to C3
once more. In the latter loop, L1 and the switch are replaced by the MS. The voltage V1 and
the current I1 are compressed to V2 and I2, respectively, by the function of MS during the
energy transfer, as shown in Fig. 1.9(b).
Considering the above mentioned topologies variety of pulse generators have been
introduced based on the different combinations of the mentioned topologies [18, 44, 45, 47,
48]. Some of them have been implemented in large scales for extra high power generation
such as Z-Machine, at Sandia National Laboratory, which stores 12MJ of electrical energy in
36 Marx generators and the intermediate stores reach a peak voltage of 5MV. This machine
has been developed now to shoot 27 million amperes (previously 18million) in 95
nanoseconds with output power of 350 terawatts. Improvements on increasing the ability of
machine in order to support 1 petawatts ( ) is going on [59].
The switches in the above mentioned pulse generators are gas-state or magnetic switches
(non-solid state). Gas-state and magnetic switches have been widely used in pulsed power
technology, as they possess a very high electric strength and fast rise time. Gas-state switches
require special operating conditions such as high pressure ( 1), vacuum equipments and gas
supplies. In addition, they are bulky, unreliable, have short lifetime span and low repetition
rate. Even with the magnetic switches, which have a higher repetition rate, the problems
remain. These conditions limit the mobility, efficiency and increase the cost and the size of
the pulsed power system [10, 30, 60].
Chapter 1
15
(a)
(b)
Fig. 1.9 Basic MPC: (a) circuit, (b) Voltage waveforms [47].
1.2.2.1.3. Solid-State Pulsed Power
Variety of topologies has been introduced for solid-state based or combination of solid-state
and non-solid state switches. In this section the well-known ones are mentioned.
The solid-state high power converters have found wide spread applications in industry. The
development of solid-state high power converters started in the mid-1980s when 4500-V gate
turn off (GTO) thyristors became commercially available. The GTO was the standard for the
medium-voltage drive until the advent of high-power insulated gate bipolar transistors
(IGBTs) and gate commutated thyristors (GCTs) in the late 1990s [6]. These switching
devices have rapidly progressed into the main areas of high-power electronics due to their
superior switching characteristics, reduced power losses, ease of gate control, and snubberless
operation.
Solid-state switches are compact, reliable, cost effective, and have a long lifetime and
repetition rate. Moreover, using solid-state switches has the advantage of benefiting from the
power electronic controlling techniques, which results in compact and more efficient pulsed
power supply [10, 25, 30, 40, 44, 53, 60].
Chapter 1
16
The main drawbacks of solid-state switches are their limited power rating and operation
speed. For the applications that needs high repetition rate, high stability, and a long life time
the only options is using the solid-state switches. Therefore, their power rating and operation
speed limits should be solved. Table 1.1 compares the solid-state and non-solid sate switches
together.
Table 1.1 Various switch parameters
Switch Max hold-off voltage (kV) Peak current (kA) Switching speed
Spark gap 100 10 to >1000 <5ns
Pseudo spark 35 5-100 >10ns
Thyratron 30 1-10 >10ns
Thyristor/GTO 1-10 1-80 >1
IGCT 6.5 0.5-3.8 >1
IGBT 6.5 0.1-2 >0.4
MOSFET 1.2 0.05-1 >2ns
Benefiting from Diode
At the beginning of benefiting from solid-state switches in pulsed power, most topologies
introduced for non-solid state pulsed power were used in combination with solid-state
switches. Fig 1.10 (a) shows the MG topology, which instead of using resistors or inductors,
diodes have been used [61]. Using diode resolve the problem of energy loss, resonance and
charging time in MG. Same thing has been done for PFN (LC generator) as depicted in Fig
1.10 (b). The diode here prevents from resonance or oscillations.
Chapter 1
17
(a)
(b)
Fig. 1.10 Benefiting from diodes in MG [61] and PFN [18] topologies.
Solid-State Marx Generator
After the developments in solid-state switches, as mentioned above, in order to benefit from
the long life and high repetition rate the spark gap switches have been replaced by the solid-
state ones. Fig 1.11 shows the Marx generator topology with IGBTs switches [60, 62].
Variety of topologies based on modified version of the MG has been introduced in literature
[48, 60-71].
Fig. 1.11 Conventional Marx generator based on solid-state switches [60, 62].
Chapter 1
18
High voltage Magnetic Compression Modulator
Magnetic compression topology has been used widely along with solid state switches. Fig
1.12 shows one of the effective introduced topologies [10, 23].
Fig. 1.12 Pulsed power generator based on magnetic pulse compression [10, 23].
The advantages of such a technique are voltage multiplication and pulse compression
without costly re-magnetization techniques. It operates as follows: When switch S1 closes,
capacitor C1 discharges and through transformer T1, charges capacitors C2 and C3. Upon the
saturation of the T1 core, capacitor C2 recharges to the opposite polarity and the voltage
across capacitors C2 and C3 doubles. At this moment, the core of the MS saturates, and the
voltage is applied to load Z. The charge path for capacitor C3 is provided by load Z and
freewheeling diode D. 60 kV peak voltage with 15-20ns rise time pulse have been reported
using the above circuit. The reported result shows the high performance of the introduced
pulsed power supply.
Using Stack of Switches for High Voltage Generation
In solid-state switches the operating speed of the switch has inverse relation with its power
ratings (hold-on voltage and current). Hence, to increase the hold on voltage while preserving
the operating speed the stack of switches can be used. As mentioned before, this solution has
problem in synchronization and switches failure.
Fig 1.13 shows one of the methods based on using stack of switches [72]. Here, a repetitive
pulsed power generator using MOSFET (Metal-Oxide Semiconductor Field-Effect
Transistor) has been developed for accelerator applications. In order to increase the voltage
and current capacity, MOSFETs have been stacked with 8 in series and 6 in parallel. Fig 1.13
shows the circuit of a module of 6 MOSFETs, connected in parallel. MOSFET times, high-
Chapter 1
19
speed ICs are used to control gate potential. In addition, the FETs (Field-Effect Transistor)
are carefully selected so that the response-time difference between them is reduced to
minimum. Finally, eight of such modules are connected in series (48 switches connected).
Fig 1.14 shows the test circuit. In this circuit a capacitor charged by 5kV (functions as the
energy storage), while the MOSFET unit turning on and off controlled by the trigger pulses.
A 70Ω resistor is used as the load. The pulses were generated at 5kV with rise time of 33ns,
fall time of 43ns and repetition rate of 2.1 MHz.
Fig. 1.13 Circuit of a module of 6 MOS-FETs, connected in parallel [72].
Fig. 1.14 Test circuit of MOS-FET switching unit [72].
Considering the synchronization issue, the manufacturing companies have optimized
semiconductor components for pulsed power applications. Hence, similar switches are
selected to resolve the synchronization problem and in the case of series configuration new
type switches which have SCFM (Short Circuit Failure Mode) are used. Fig 1.15(a) shows
ABB switch [49] for pulsed power applications which can handle 10Kv, 90kA with low
Chapter 1
20
repetition rate of 1Hz. But BEHLKE Company which is one of the companies who has
focused on extra high voltage switches has developed fast switches that can handle 90Kv,
100 A with repetition rate of 100 KHz [73] (Fig 1.15(b)). The only problem with these kinds
of switches their high cost and high rise time.
(a)
(b)
Fig. 1.15 High voltage switches (stack of switches) optimized for pulsed power applications: (a) ABB [49], (b)
BEHLKE [73].
1.2.2.1.4. Power Electronics Converters for Pulsed Power
Power electronic converters have been one of the fastest growing market sectors in the
electronics industry over the last 25 yrs. Power electronic devices are at the heart of many
modern industrial and consumer applications and account for $18 billion per year in direct
sales, with an estimated $570 billion through sales of other products that include power
Chapter 1
21
electronic modules [74]. Recently, applications of solid-state high power converters have
gained great interest.
The topologies that were not originally designed for the pulsed power have been used in
this area. The main reason is benefiting from power electronic techniques to control the
energy flow and increase the system efficiency [53]. In the following the most high
performance methods for pulsed power supplies are reviewed.
An ultra compact generator based on the boost topology using static-induction thyristor (SI-
Thyristor) is addressed in [72]. In this approach, a SI-thyristor is used to develop an ultra-
compact pulsed power generator that is aimed at automobile applications. The electrical
circuit is illustrated in Fig 1.16.
This topology is consists of two stages of inductive energy storage with opening switch.
The behaviour of stage 1 and 2 are controlled by FET1 and FET2, respectively. Although the
second stage is completely controlled by FET2, the SI-thyristor plays the role of opening
switch and holds the voltage during the output. The SI-thyristor used in this experiment is a
4cm device with nominal maximum voltage of 3 kV and current of 300 A. The typical
waveforms are shown in Fig 1.17. The capacitor C1 is charged by the DC voltage supply of
12 V. The capacitance of C1 is large enough so that its voltage keeps nearly constant during
operation. When FET1 is turned off, the inductive energy stored in L1 is transferred to C2,
resulting in the C2 charging voltage of 200 V. When FET2 is turned on, the current in L2 goes
up and reaches 150A. Finally, the opening of SI-thyristor causes a voltage across SI-thyristor
of 3kV, which is multiplied by the transformer to 12 kV at the output. The FWHM is 100ns
and the repetition rate is 2 KHz.
Fig. 1.16 Electrical circuit of the pulsed power generator using SI-thyristor [72].
Chapter 1
22
Fig. 1.17 Waveforms of L2 current and the output voltage [72].
The literature refers the use of flyback, forward, half-bridge and full-bridge converters for
applications such as rapid capacitor high-voltage charging, food processing, X-ray plasma
processing, and air and water pollution control [53, 75-78]. The advantage of all these
configurations is using the transformer which is advantageous in isolation and stepping up the
output voltage.
Flyback and forward converter are two well known topologies in power electronics [53, 76,
78-85]. Flyback topology has been used from long time ago in CRT (Cathode Ray Tube) TVs
for high voltage generation. Fig 1.18 shows these two topologies.
The forward and flyback topologies are suitable for the generation of unipolar pulses. In the
case of bipolar pulses, half-bridge or full-bridge configurations are required (Fig 1.19).
Comparing flyback with all other converters for the pulsed power applications shows that
flyback topology seems much more appropriate [53, 82, 83]. The major advantage of flyback
topology over other power electronic topologies is the way of using transformer. The
transformer in this topology stored energy, steps down reflected voltage across the switch and
isolation. The problem with this topology is high frequency issues due to using the
transformer [83, 86].
Chapter 1
23
(a)
(b)
Fig. 1.18 (a) flyback, and (b) forward topologies [53].
Most of the introduced method for solid-state based configurations didn’t apply to the real
applications the whole circuit has been just evaluated with a resistive load. In the real world
that the applications have different behaviour such as RC, C, or RL [53] these methods
become ineffective. As a conclusion, even with the recent developments on solid-state
switches, the solid-state topologies are still dealing with two important issues comparing with
the non-solid state ones in terms of voltage level and rise time.
Chapter 1
24
(a)
(b)
Fig. 1.19 (a) half bridge, and (b) full bridge [53].
1.2.2.2. Application Level
As mentioned earlier, the last section of a pulsed power system is the load. Pulsed power
applications present one of the most varied ranges of loads in terms of load behaviour and
requirements. Hence, various load conditions adversely affect the power supply performance.
Here, different load types associated with pulsed power systems are described.
1.2.2.2.1. Load Types
Resistive Loads
Pure resistive loads are not common in pulsed power applications; however high impedance
resistive loads are used frequently as a load reference for testing and evaluating different
pulsed power supplies. One of the most challenging load types in pulsed power area is low
impedance resistive loads, in applications such as water decontamination, liquid food
sterilization, and biomedical materials [12, 22, 26-28, 38, 47, 53, 87]. The problem with low
resistive loads is that the energy dissipated in the load before the output voltage reaches to
required voltage level. To overcome this problem designing a load independent pulsed power
supply is needed. Till now applying pulsed power supplies with rise time in the order of
fraction of nano-seconds have found to be a good solution [27, 53].
Chapter 1
25
Resistive Capacitive Loads
Most of the pulsed power applications have resistive-capacitive (RC) behaviour, such as
plasma and gas processing. In fact, one of the most challenging operating conditions is
related with operation with plasmas [39, 53, 54, 56].
A typical example is dielectric barrier discharge (DBD) loads. To understand the load
dynamic characteristics, it can be modelled with an RC electrical equivalent circuit as in Fig
1.20. Here Cg represents the gap capacitance and Cd is the dielectric capacitance. Once that
dielectric breakdown occurs, the displacement current flows from one electrode to the other,
through the tube walls, using the plasma as a conductor, and this current is limited by Cd in a
unique manner [55].There are two regions of operation for a DBD load. Prior to breakdown,
Cg is in series with Cd, and the discharge is off. During breakdown a nonlinear gaseous
discharge is in series with a capacitance Cd, which plays a major role in limiting the gap
current.
Fig. 1.20 Electrical equivalent circuit of a DBD load [55].
In high impedance-capacitive loads, it is extremely important that the power supply has the
capability to short circuit the load after applying high voltage pulse; otherwise the load can be
damaged as the load stays charged to almost full pulse voltage [53].
Inductive Loads
This type of the load presents when voltage pulses are applied to a coil. For instance, in
electromagnetic metal forming application (EMF) high voltage resonant power supply applies
high current pulses (kA) into very low-inductive system [53, 88]. With this type of loads, it is
important that the pulsed power supply clamped the load with an opposite polarity voltage
after high voltage pulse.
Chapter 1
26
1.2.2.2.2. Pulsed Power Applications
The load is actually the application which the pulsed power supply is going to apply to. It is
one of the important parts of the pulsed power system, as the whole system is designed based
on the load or application requirements. Each application is separated based on the generated
pulse characteristics such as pulse amplitude, rise time, full width at half maximum (FWHM),
fall time, etc. Table 1.2 shows the typical pulsed power applications requirements [30].
Table 1.2 Summary of typical pulsed power applications requirements [30].
Application Electrical Energy Peak Power/Pulse
High Energy Density Plasma Physics 20MJ >10’s TW
Intense Electron Beam Radiography 200 kJ <1 TW
High Power Microwave (Narrowband) 10 kJ 100 GW
High Power Microwave (Ultra-wideband) 10 J 10 GW
Ion Beam Modification of Materials <10 kJ 30 GW
Bioelectrics 0.1 mJ- few J 10kW – 100 MW
Generally these applications can be divided into three categories:
Military and Defence Applications
In fact, the development of equipment to drive magnetron oscillators for microwave
radar started the field of pulsed power. Since that time, applications have been
dominated by defence-related technologies [44, 89].
Industrial Applications
In recent years, several applications based on pulsed power technology have been
introduced, developed or commercialized. Pulsed power has covered a variety of
industrial applications such as, fusion systems, food pasteurization, water treatment,
ozone generation, etc [9-11, 13-15, 18, 19, 22-25, 27, 29-31, 33, 34, 36-43].
Biomedical Applications
Since demonstrated that lightly ionized air is extremely effective in killing harmful
bacteria and spores on contaminated surfaces, the pulsed power became an important
technology in the biomedical arenas. Pulsed power has had a great impact on
biomedical applications such as, cancer treatment, wound healing etc [2, 12, 17, 20, 21,
26, 28, 30, 32, 35].
Chapter 1
27
From another aspect the pulsed power applications can be divided into two different types
of pulse generation. One is generating electro-magnetic energy which is the main concern of
military applications and second is plasma pulsed generation. In the following sections the
major non-military applications of the pulsed power system, in order to show the importance
of this technology, are described.
Plasma
Plasma is defined as the fourth state of matter. It was first identified by Sir William Crookes
in 1879. Plasma is an ionized gas containing free moving charge carriers: electrons and ions.
Over 99% of the visible universe is made up of plasma. For example, the matter in stars or
nebulae is plasma. There are also man-made plasmas on our planet, daily used in industrial
and medical applications. One of the ways in generating plasma is using pulsed power
supplies [35, 54, 56, 90-96].
But why generating plasma is beneficial. Plasma is very effective in disinfection. This
makes plasma very useful for various medical and industrial applications. The factors that
make plasma advantageous are:
Ultraviolet radiation
Ultraviolet radiation (UV) causes DNA (Deoxyribonucleic Acid) damage. This DNA
damage inhibits the replication of bacteria. The wavelengths in the range 220 – 280 nm at the
irradiation intensities of several mW/cm2 are known to have the optimal inactivation effect.
However, studies showed that UV is not an important decontaminating factor in treatments
with low temperature atmospheric air plasmas [35].
Heat
Heat can inactivate bacteria. Conventional sterilization methods are based on heat.
Charged particles
Charged particles from the plasma cause cell membrane charging, which may rupture the
outer membranes of bacteria [11, 21, 35]. This is because the electrostatic force caused by the
build-up of charge on the membrane can overcome its tensile strength and cause rupture.
Reactive species
The reactive species in non-thermal atmospheric air plasmas are generated through electron
impact excitation and dissociation. In these air plasmas we can find nitrogen– and oxygen–
based species such as atomic oxygen, ozone, NOX (Nitric Oxides), and OH.
Chapter 1
28
When the electric field intensity increases, it affects the electrons in the atomic orbital and
may cause atoms or molecules to polarize or liberate electrons. The maximum electric field
that a dielectric material can withstand without conduction is known as the dielectric strength
of that material and is expressed in V/mm. Table 1.3 shows the dielectric strength of common
materials. When the electric field increases beyond the dielectric strength, the material
becomes conducting by a process called avalanche effect, whereby electrons collide with the
atomic or molecular structure, releasing more electrons which in turn lead to the further
breakdown of the material. Large currents are possible at breakdown. For example the
dielectric strength of air at normal temperature and pressure is 3 kV/mm. At this point air
ionizes rapidly and arcing occurs. This is what happens during a lightning strike. Therefore,
depending on the electrodes gap and material dielectric strength the required electric filed
intensity varies in different applications.
Table 1.3 Dielectric strength of common materials.
Substance Dielectric Strength (kV/mm)
Helium (relative to nitrogen) 0.15
Air (relative to nitrogen) 3
Window glass 9.8-13.8
Silicone oil, Mineral oil 10-15
Benzene 163
Polystyrene 19.7
Polyethylene 18.9-21.7
Neoprene rubber 15.7 - 26.7
Distilled Water 65-70
High Vacuum (field emission limited) 20 - 40 (depends on electrode shape)
Fused silica 470 - 670
Waxed paper 40 - 60
PTFE (Teflon, Extruded ) 19.7
PTFE (Teflon, Insulating Film) 60 - 173
Mica 118
Generally, plasma can be divided into three different types of:
1) Plasma Discharge
Plasma discharge, thermal plasma or hot spark happens when the electric field intensity
exceed from the dielectric strength of the matter. In this situation high current pass
Chapter 1
29
through the material and makes spark or arc. Due to this fact this type of plasma is not
preferable as the material, probe and pulsed power supply may get damage.
2) Cold Plasma
Cold plasma can happen in low pressure and it is also kwon as non-thermal plasma.
This situation is readily achieved under reduced pressures, in the range of 10 to 1000
Pa. But, these plasmas must be confined in massive vacuum reactors, their operation is
costly, and the access for observation or sample treatment is limited.
3) Atmospheric non-Thermal Plasma/Corona
In order to achieve the characteristics of non-thermal plasma but at atmospheric
pressure the current should be limited when the material starts to conduct. This can be
done by controlling electric field intensity in way that it never goes above the dielectric
strength. The second way is using a dielectric barrier discharge (DBD) which is a
special way of placing the electrodes [39-41, 53-56]. This kind of plasma in known as
corona discharge [35, 90, 91, 93, 96] and it is quite desirable due its non-destructive
features.
Interest in pulsed power technology in biomedical field is increasing since its disinfection
feature has been found. The biological effects of pulsed power can be categorized as water or
liquid discharge, gas discharge, and electromagnetic field [2, 17, 20, 21, 26, 28, 30, 32, 35].
Except for generating electromagnetic field which can be used in cancer treatment all other
applications benefit from the plasma inactivation features (described in the previous section).
On the other hand, with the pulsed power major features as efficiency, compactness, low
maintenance, reliability and clean technology, pulsed power industrial applications are
growing fast.
In the following two sections, the pulsed power applications have been considered in three
different categories with respect to biological and industrial applications. First one is
biological applications and the second one is industrial applications.
1.2.2.2.3. Water/Liquid Discharge
The most typical application of pulsed discharges in liquid is water treatment. The objects
of treatment are in various fields in both biomedical and industrial applications, such as
sewage, river, lake and etc. Liquid discharges are selected for sterilization or inactivation of
bacteria in liquid decontamination. A streamer discharge is usually used in liquid [9, 10, 12,
Chapter 1
30
22, 27, 38, 47, 87]. These streamer discharges in liquids are able to produce a high electric
field, high energy electrons, ozone, chemically activate species, ultraviolet rays, and shock
waves, which readily sterilize microorganisms and decompose molecules and materials. Fig
1.21 shows a discharge induced in water [87].
Fig. 1.21 Photograph of a discharge induced in water [87].
In addition to the mentioned features of the water discharges, generated ozone due to the
plasma phenomena can be used in ozone therapy. Medical ozone is used in three principal
fields including: treatment of circulatory disorders, disease produced by viruses such as liver
disease (hepatitis) and herpes and healing badly infected topical wounds and inflammatory
processes including open ulcers on the legs, inflammatory intestinal conditions, burns, scalds
and infected wounds as well as fungal infections.
This method is expected to have high efficiency for the treatment and decompose the
pollutants that cannot be easily decomposed by even ozone oxidation. Pulsed power systems
not only are more effective, compact and cost effective and reliable but also by using pulsed
power supplies there is no need to employ chemical approaches for example in
decontaminating water. For example, pulsed power is employed in cleaning of lake and dam
algae bloom by discharges in the water. Fig 1.22 shows the effect of using pulsed power in
algae treatment [27]. Usually algae exist on the surface of the water. As the figure shows after
applying pulsed power the algae were precipitated to the bottom.
Chapter 1
31
Fig. 1.22 Photograph of effect of water discharges in algae treatment [27].
However, liquid discharge also has some problems. One is the high breakdown voltage of
the liquid. Therefore, good insulation is needed inside the pulsed power supply. Secondly, it
is difficult to make a widespread discharge in water for getting a wide treatment area. Finally,
as the liquid, especially water, has small resistance therefore generated pulses should have
either high energy which increases the size of the system, or get connected to the liquid after
the voltage level has reached to the required amplitude.
1.2.2.2.4. Gas Discharge
Gas discharges are selected for sterilization of Escherichia coli and other bacteria for
decontamination of air. It has been used mostly in treatment of exhaust gases but is has also
applied for crop growth and water treatment as well. For crop growth, gas discharges were
used for cultivation of mushrooms [10], and for water treatment gas discharges with droplets
of water is suggested. But as mentioned its main application is on exhaust gas treatment.
Recently, non-thermal plasmas have been increasingly used to control harmful gases and to
generate ozone [13, 15, 19, 24, 33, 34, 36, 37, 39].
The environmental pollution caused by the consumption of fossil fuel energy is increased
rapidly; it became very important to protect the environment and to develop technologies
with less energy consumption and less pollutant exhaust. One the ways is using non-thermal
plasma in exhaust gas treatment. Non-thermal plasmas have many kinds of chemically
activate radicals, such as O, O3, N, N*, and OH, which are generated by the dissociation
Chapter 1
32
and ionization of the ambient gases caused by the impact of energetic electrons. Using pulsed
power technology, non-thermal plasmas have been generated by a pulsed electron beam or a
pulsed streamer discharge, and have been used to treat nitric oxides (NOX), sulfur dioxide
(SO2), carbon dioxide (CO2) and volatile organic compounds (VOCs), and to generate ozone.
Particularly, the treatment of exhaust gases (NOX and SO2) using a pulsed streamer discharge
has been studied for the past decade.
1.2.2.2.5. Material Processing
Plasma-based ion implantation and deposition are one of the well-known methods which
are used for surface treatment of complex shape materials by employing pulsed power
supplies. This method can now be considered a nature technology for surface modification
and thin film deposition. In addition to this method pulsed power can be used in other
applications of material processing, such as surface heating by laser, or microwave
irradiation, synthesis nano-composite powders, and joining of solid materials [10, 88].
1.2.2.2.6. Electromagnetic Field
Pulsed electromagnetic fields are selected for sterilization or inactivation of bacteria, but
they have gained more attention for cancer treatment. The electromagnetic field yields
electroporation of the cell membrane or influences the cell nuclei. Electroporation is usually
used to sterilize bacteria. This technique is commonly applied for sterilization in food
processing, such as inactivation of microorganisms naturally contaminated in orange juice
[97].
Recently, many researchers have interests in using shorter pulse durations with fast rise
time. The reason is that the electrical pulses should reach to the cytoplasm without affecting
the membrane. To make this happen the pulse duration and rise time should be faster than the
charging time of the membrane. To understand the behaviour of biological cell in dealing
with electrical pulses, electrical circuit of a cell is considered [30]. Fig 1.23 shows the
capacitive-resistive model of a biological cell. In addition to rise time, voltage level or the
pulse amplitude should be sufficiently high in order to affect the intracellular organelles [30].
The varied range of applications for pulsed power shows the importance of this technology.
The study and research on this technology is going with respect to designing more compact
and reliable system and studying biomedical features of the generated pulse.
Chapter 1
33
(a)
(b)
Fig. 1.23 a) Biological cell structure, b) Double-shell model of a biological cell [30].
1.2.3. High Frequency High Average Power Converter for
Ultrasound Applications
Wide spread research has been conducted on piezoelectric transducer applications since
1880, the year that Pierre and Jacque Curie discovered the phenomenon of piezoelectricity
[52, 98]. Piezoelectric transducers convert electric power to acoustic power and vice versa,
with most of the applications to date being low-power ones. In the last decade, however,
high-power ultrasound applications have gained significant importance [3, 99-104]. These
types of applications have great potential in chemical and bio-technology processing,
specifically for enhancing chemical reaction kinetics and new reaction pathways. Such
enhancements allow changing the production from batch processing to continuous flow
processing, thereby reducing investment and operational costs.
Improvement of ultrasound systems has very significant environmental implications,
particularly in the area of renewable energy (bio-mass and bio-fuel), waste-water treatment
and biomedical applications. High-power ultrasound technology is already being used to
address these areas of need, but to a very limited extent owing to the very inefficient nature of
the state of the art in energy conversion.
The big challenge is how to design a power supply which can generate a proper excitation
signal for the piezoelectric device. The critical behaviour of a piezoelectric device is
encapsulated in its resonance frequencies because of its maximum transmission performance
at these frequencies. The resonance frequencies fall within the range of ultrasound which
means above 20kHz.
Chapter 1
34
Recently, the most efficient way to generate power signals to excite the piezoelectric device
found to be high frequency high power converters. In order to design a suitable excitation
signal using a power supply, the behaviour of the piezoelectric transducer needs to be
investigated. Hence, designing stage of a power supply for piezoelectric applications falls
within two major steps:
Analyzing piezoelectric characteristics
Design and adapting a high frequency high power supply based on the piezoelectric
behavior
1.2.3.1. Piezoelectric Transducer Characteristics
Large amount of research have been conducted on developing methods to study
piezoelectric resonator. Among them the most effective way is evaluating its impedance
frequency response [105-117]. In the impedance methods, changes in the transducer
impedance due to mechanical resonance are detected. A minimum in the impedance response
corresponds to a resonant frequency, fr (see Fig 1.24). The impedance frequency response is
the ratio of the voltage spectrum to the current spectrum. To calculate the piezoelectric
impedance response, a voltage source needs to be applied to the device as an excitation signal
and current needs to be measured simultaneously. In order to obtain the response of the
device for a specific range of frequencies, the excitation signal should cover the entire
frequency range.
Piezoelectric devices typically have two kinds of electrical resonances (see Fig 1.24), the
first known as a resonance and the other one known as anti-resonance [52]. In the last few
years, varieties of approaches have been introduced for obtaining piezoelectric impedance
characteristics [105-117]. A knowledge of such impedance characteristics is important for,
structural health monitoring (SHM) [114, 118] and for electrical circuit modelling [106, 107,
113, 115] of the piezoelectric devices. In the case of high power applications, knowledge of
the impedance characteristics is important so that the maximum transmission performance
can be achieved [5, 115-119]. In the other words, at resonance frequencies, piezoelectric
elements convert the input electrical energy into mechanical energy most efficiently, and so it
is necessary to accurately know these frequencies and excite the system appropriately.
Chapter 1
35
Fig. 1.24 Resonance and anti-resonance frequencies in piezoelectric impedance frequency response.
Piezoelectric devices typically have multiple resonance frequencies (as illustrated in Fig
1.25) but only the major resonance frequency is generally targeted for excitation in practice.
Structural and environmental changes can affect the transducer elements and cause changes
in the values of the resonance frequencies [5, 52]. As such changes are occurring, it is
important to estimate the main resonance frequency if one is to maintain efficient system
operation (see Fig 1.25 for an illustration).
To measure piezoelectric transducer impedance, generally a network analyser is used.
However, due to the fact that the network analyser is expensive, bulky and it can’t operate at
high power other methods can be employed. In addition to network analyser, these methods
can be categorized as below:
Traditional method: In this method several single frequency sine wave signal
regarding to the range of frequency interest are applied to the transducer separately.
The impedance can be estimated after the responses are captured for each single
frequency. The accuracy of this method depends on the frequency resolution. This
method is not practical for studying the response of the transducer for wide range of
frequency as it is labour intensive.
The second method is based on exciting the piezoelectric transducer using a broad-
band signal which covers wide range of frequencies. The concept of this method is
based on the system identification which is well-known in the field of signal
processing [120, 121]. This method is based on the impulse response estimation and
can utilize signals such as chirp, white noise, step, and etc [120, 121]. The
advantage of this method is low computational complexity and high accuracy.
Chapter 1
36
It is to be noted that for high power transducer it is not possible to study the characteristics
of the transducer at high power using a network analyser, due to the fact that a network
analyser operates at low power. To remedy this issue the aforementioned methods should be
employed.
Fig. 1.25 Depiction of typical resonance frequency variations of piezoelectric devices due to structural changes.
In addition to impedance frequency response, a good understanding on the electrical
characteristics of piezoelectric transducer is essential in the design and analysis of sensory
systems relying on piezoelectricity, such as impedance-based SHM systems [107, 115].
Usually, an equivalent circuit model is used to describe the electrical characteristics of the
piezoelectric materials. Moreover, an equivalent circuit model makes piezoelectric-based
systems simulation possible.
The most basic equivalent electrical circuit model characterizing piezoelectric transducer is
the Van Dyke Model [107] (see Fig 1.26). The Van Dyke Model is a parallel connection of a
series RLC representing mechanical damping, mass, and elastic compliance and a capacitor
representing the electrostatic capacitance between the two parallel plates of the PZT
(Piezoelectric Transducer) patch [122].
Fig. 1.26 The Van Dyke Model [107].
As the Van Dyke model cannot accurately model the piezoelectric behaviour other models
have been proposed. The Guan Model, which is the most recently proposed, estimates values
Chapter 1
37
of the electrical components based on the electrical behaviour of the piezoelectric ceramics
[107, 115]. The Guan model, as shown in Fig. 1.27, adds a series resistor Rs and a parallel
resistor Rp to C0 of the Van Dyke Model to account for the energy dissipation. Determination
of the values of the electric components C0, C1, L1 and R1 relies on visual inspection on the
magnitude and phase of the impedance, and values of Rs and Rp are decided by the amount of
energy dissipation [107].
Fig. 1.27 The Guan model-unloaded piezoelectric [107].
When a piezoelectric ceramic is mounted to a mechanical structure, the mechanical
boundary conditions of the piezoelectric ceramic change [107, 115], and, accordingly, a
different circuit model is required for a loaded piezoelectric ceramic. Since a loaded
piezoelectric ceramic experiences multiple resonances, a circuit model for a wide frequency
range with multiple resonant frequencies can be employed to model the behaviour of a loaded
piezoelectric ceramic. As shown in Fig 1.28, additional R-C-L branches are added in parallel
to the R1-C1-L1 branch of the Van Dyke Model [107, 122]. The Guan Model is also extended
to accommodate a loaded piezoelectric ceramic based on the extended Van Dyke Model, as
shown in Fig. 1.29. Each series RLC branch physically stands for a mechanical resonant
mode (as shown in Fig 1.25).
Fig. 1.28 The extended Van Dyke model-loaded piezoelectric [107].
Chapter 1
38
Fig. 1.29 The complete Guan model-loaded piezoelectric [107].
Estimating an accurate model representing the piezoelectric transducer behaviour such as
linear and nonlinear characteristics and a systematic procedure to find an equivalent electrical
circuit model is an undergoing research due to the varied behaviour of the piezoelectric
device under different situation.
1.2.3.2. Power Electronics Topologies for High Power Ultrasound
Systems
As mentioned above, the critical behaviour of a piezoelectric device is encapsulated in its
resonance frequencies because of its maximum transmission performance at these frequencies
[52, 98, 114, 116, 119]. Hence, ideal scenario is to have a sinusoidal excitation signal at the
resonant frequency. But at high power and high voltage, which is the focus area of this paper,
generating pure sinusoidal signals is not possible.
The most efficient way to generate power signals is to use power electronics inverter
topologies. Specifically, switch mode inverters are used for piezoelectric high power
applications due to their high power density, efficiency, low cost and size compared to
conventional linear power supplies [5, 102, 123]. However, the harmonics present in the
output waveform produce undesired side bands which are not suitable in many applications.
Moreover, they also cause unnecessary power dissipation which reduces the efficiency of the
power converter [124, 125].
In the last decade several power electronics topologies have been introduced for high power
piezoelectric applications. To overcome the harmonic issue, generally either a filtering circuit
or a harmonic elimination technique is employed [125]. Here the most important topologies
recently introduced are briefly described.
Chapter 1
39
Two Level Inverters
Generally, with this type of inverters a resonant converter is employed. Resonant converter,
normally, is based on a typical power electronics topology and a resonant stage formed by
applying filter component and the piezoelectric transducer itself (as it has RLC
characteristics).
The first stage can be based on flyback, full-bridge, half-bridge, and etc. Fig 1.30 depicted a
flyback based resonant inverter. The flyback converter is supplied with a full-bridge rectifier.
This stage in addition to providing smoothed dc-link voltage performs a PFC (Power Factor
Correction) function by shaping the input current to be sinusoidal and in phase with the ac
input voltage [126]. The second stage is a high frequency resonant inverter.
Another alternative is depicted in Fig 1.31 using a full-bridge (H-bridge) inverter combined
with a resonant stage [5]. As can be seen different resonant configurations can be used which
totally depends on the piezoelectric transducer and the range of the operating frequency.
Fig. 1.30 Two-stage resonant inverter based on flyback topology [126].
Chapter 1
40
Fig. 1.31 Two-stage resonant inverter based on H-bridge topology [5].
Multilevel Inverter
Employing multilevel inverters is a well-known method in order to increase the generated
signal quality and reduce number of the harmonics close to the fundamental frequency.
Moreover, for ultrasound applications which operate at high fundamental frequency power
converters based on a PWM (Pulse Width Modulation) strategy with high switching
frequency index (fs/fo) cannot be a practical solution. In such applications, the maximum
possible fundamental frequency of a converter is restricted by the switching transients of each
power switching event and the number of switching events per cycle.
In this regard, to reduce the number of switching transients and eliminate the undesired
harmonics, the most effective way is to use multilevel inverters [127-130]. Through these
inverters, it is possible to produce quasi-sine waves with low total harmonic distortion at high
power. Multi-level inverters can increase the quality and efficiency of the high voltage supply
compared to the conventional 2-level inverters. This permits the semiconductor devices to
operate at lower switching frequencies with higher efficiency as well as lower voltage stress
across switches and loads which minimize electromagnetic and ultrasound noise emissions.
Chapter 1
41
Fig 1.32 depicts typical voltage waveforms and their harmonic spectra. It can be seen that the
harmonics are depressed for multilevel inverters such that they can be easily filtered out from
the frequency range of interest.
Fig. 1.32 Effect of using multi-level waveform in harmonic elimination.
Fig 1.33 illustrated one of the multilevel methods introduced for high power piezoelectric
excitation. The employed output filter is resonant LLCC type, which can reduce the total
harmonic distortion (THD). By combining a PWM controlled inverter and resonant filter, the
generated voltage can be varied in a suitable range of frequency. The inverter topology is a
hybrid three-level PWM inverter based on diode clamped topology [102]. Fig 1.34 shows a
typical generated output voltage and current and the output voltage across the piezoelectric.
Chapter 1
42
Fig. 1.33 Three-level PWM inverter combined with LLCC-filter for driving high power piezoelectric transducer
[102].
Fig. 1.34 Typical output voltage and current of employed hybrid three-level inverter when applying to a
piezoelectric transducer [102].
As can be seen even by using a multilevel topology filtering is still employed. However, as
the advantage of multilevel inverter is to push the high frequency contents far away, when
using filters for attenuating the remaining harmonics the filter cost and size decreases [127,
130].
Although several improvements have been achieved in this area, but increasing the number
of the voltage levels with low circuit complexity, adapting with resonance frequency
Chapter 1
43
variations, and employing effective harmonic elimination techniques are still under
investigation.
1.2.3.3. High Power Piezoelectric Transducer Applications
To show the importance of high power piezoelectric transducer, in this section a brief review
on high power piezoelectric applications is provided. Applications of high-intensity
ultrasonics are those wherein the ultrasonic energy is used for producing some permanent
physical change in the treated medium [3]. These effects can be attributed to various
mechanisms, the most important of which are: radiation pressure, heat, streaming,
cavitations, agitation, interface instabilities and friction mechanical rupture. These
mechanisms are involved in a wide range of applications, such as machining, welding,
metal forming, powder densification, etc, in solids; cleaning, particle agglomeration and
flocculation, liquid atomization, drying and dewatering, degassing, etc., in fluids [3]. For
instance Fig 1.35 shows the application of high power ultrasonic atomizer in manufacturing
polymer powder [5].
Recently, high power piezoelectric applications have gained interests due to new
improvements in high power supplies. High-power piezoelectric devices, such as ultrasonic
motors and piezoelectric transformers, are being intensively investigated [3, 99-104, 131].
Because piezoelectric devices often have significant advantages over conventional
electromagnetic devices at smaller sizes, miniaturization of piezoelectric devices is attractive.
Multilayer structures offer better generative force, electromechanical coupling, and energy
density properties than other configurations such as bimorph [103, 132, 133] and moonie
structures [103, 134]; thus they are promising designs for compact devices [103, 135].
Applications including ultrasonic motor, transformers and medical ultrasonic, such as
acoustic radiation force impulse (ARFI) imaging and ultrasound-guided HIFU (High
Intensity Focused Ultrasound) therapy are now shifting to commercialization stage. Although
great improvements have been achieved in this area, but high power applications are still
limited. This limitation causes by restrictions on excitation signal (power level and quality)
and internal power dissipation which is related to the piezoelectric material.
Chapter 1
44
Fig. 1.35 Basis of the ultrasonic atomizer [5].
Chapter 1
45
1.3. Account of Research Progress Linking the Research Papers
1.3.1. Introduction
This research commenced with studying high instantaneous power supplies, generally known
as pulsed power. Varieties of previous introduced methods, which mainly include non-solid
state techniques, were investigated in order to understand their advantages and drawbacks.
Meanwhile, high average power supplies with specific intention on high power piezoelectric
device applications were also investigated. In both categories, the main aim was to employ
and adapt power electronics techniques to increase the system performance, compactness,
simplification, and flexibility based on the application characteristics.
The preliminary literature review emphasised on trends for utilizing power electronics
techniques and topologies in high frequency high power supplies. Hence, studying and
reviewing converters and topologies capable of providing high power at high frequency was
considered as the initial step. Following that, to design a suitable power supply, load and
application requirements and their electrical characteristics were investigated.
Simulation and practical implementation of varied power converters and exciting different
loads at varied operating points were conducted after the literature review stage. Dealing with
practical issues such as noise, EMI (Electro Magnetic Interference), high voltage insulation,
transformer saturation, spikes across the IGBTs, DSC (Digital Signal Controller)
programming and etc, caused vast studying in addition to the main objectives.
1.3.2. Pulsed Power Systems
This research started with an extensive study on non-solid state topologies, such as Marx
generators (MG) [18, 44, 45], pulse forming network (PFN) [18, 45, 46], magnetic pulse
compressors (MPC) [18, 45, 47, 48] and multistage Blumlein lines (MBL) [18, 45], with
respect to their advantages and drawbacks. These techniques possess a very high electric
strength and fast rise time. But they are bulky, unreliable, have short life time span, low
repetition rate, and etc. On the other hand, solid-state techniques are compact, reliable, cost
effective, and have a long lifetime and repetition rate. As the power electronics techniques
are solid-state based, applying power electronics techniques found to be suitable way to
remedy aforementioned issues existing with the non-solid state power supplies.
Recently, significant advances in solid-state switches (both in peak power and operation
speed) and exploiting power electronics techniques and topologies have led to compact and
Chapter 1
46
more reliable pulsed power systems. However, power electronics techniques still show two
main disadvantages of limited power ratings and limited operating speed comparing with the
non-solid state ones. Hence, the major concern was to improve these two important limits.
Moreover, as the power supply is designed based on the load requirements, the load and
application specifications needed to be taking into account along with the two
aforementioned drawbacks.
Regarding to the mentioned issues, the main objectives of studying, designing and practical
implementation of new alternative topologies and configurations of pulsed power supplies
were as followings:
1- Limited power ratings of the solid-state switches
2- Limited operating speed
3- Load/application characteristics
1.3.2.1. Benefiting from Parallel and Series Configurations of Flyback
Converter for Pulsed Power Applications
Interest in applying power electronics topologies and techniques to increase pulsed power
supply flexibility and reliability is growing fast. In the last decade, research and studies
designate the advantage of using power converters in pulsed power applications [10, 136,
137]. However, generation of extra high voltages is still problematic due to components hold-
on voltage limits (such as capacitors, solid-state switches and etc).
One solution is to combine pulsed power supplies. Variety of converter configurations have
been introduced for improving the power supply specifications such as output ripple, high
input voltage, output power ratings, etc. Parallel and series connections are one of the well
known combinations. Designing power supplies based on series or parallel connection are
widely employed for different applications.
Therefore, the main contribution of this part was to benefit from parallel and series
connections of flyback converter modules to develop power rating and rise time of the pulsed
power supply using conventional low voltage switches.
The proposed approach utilizes flyback converter (see Fig 1.36) as it is simple, has only one
switch and magnetic component, is able to generate high voltage, can provide multiple
outputs, isolation, etc [82, 138]. But when it comes to pulsed power applications, in addition
to the above-mentioned features it shows more advantages as follows:
Chapter 1
47
The transformer, in addition to electrical isolation and energy storage also steps down the
reflected voltage across the switch; therefore lower voltage rate switches are needed
unlike other topologies.
Fault tolerant, as the switch is in the off-state during the output pulse.
High voltage output with low input DC voltage.
As the pulsed power applications mostly have R-C characteristic, a current source
topology (such as flyback) is a suitable candidate.
Suitable for controlling the energy flow as it acts as both current source and voltage
source.
Lm
Vs
C0 Load
1 : n
Ll
Vo
+
-
Fig. 1.36 Flyback converter circuit with transformer equivalent circuit model.
1.3.2.1.1. Proposed Method
The proposed method which employs the advantage of parallel and series configuration of
the flyback converter as the pulsed power supply to increase the voltage level and the rise-
time is shown in Fig 1.37. As depicted the parallel and series connections are only considered
for the secondary side, while the primary side of both configurations are connected in
parallel.
Chapter 1
48
Ll
Lm
Vs
1 : n
Co2
Ll
Lm
n : 1
Co1Load
Vo
+
-
Converter (1)Converter (2)
(a)
Ll
Lm
Vs
1 : n
Co1
Load
Ll
Lm
1 : n
Co2
Vo
+
-
Converter (2)
Converter (1)
(b)
Fig. 1.37 Flyback converter series and parallel connection (secondary side), a) Parallel, b) Series.
CoIo Vo
+
-
Load
Fig. 1.38 Energy transmission in a single module flyback converter.
Chapter 1
49
To understand and compare the parallel and series configurations, first a single module
characteristic is described. Regarding to the fact that a flyback converter can operate as a
current source, circuit schematic shown in Fig 1.36 is simplified as in Fig 1.38. Comparing
these two figures, the estimation of the output voltage can be summarized as below:
First, the magnetizing inductance is charged by the DC power supply. This mode lasts
depending on the duty cycle of the PWM (Pulse Width Modulation) signal. The relationship
between Lm and Vs can be expressed as:
s
mm
s
mmsms
TD
IL
TD
ILV
t
iLV
.0.
0
(1-5)
where Vs is the dc supply voltage, Im is the maximum current, D is the duty cycle and Ts is
the switching period. The charged energy stored in the inductor Lm under ideal situation can
be expressed as:
2
2
1mmpri ILE (1-6)
Regarding (1-5) and (1-6) it is possible to control the energy flow in each pulse by limiting
the Im which can be done by varying the duty cycle or the input voltage. This stored energy
will be transferred to the output capacitor when the switch is in the off-state. The capacitive
stored energy is:
2sec
2
1ooVCE (1-7)
If a lossless system considered, then the stored energy in the primary side, Epri, is equal to
the stored energy in the secondary side, Esec. Therefore, the output voltage can be derived
based on equations (1-6) and (1-7) as:
o
mmo
C
LIV
(1-8)
The rate of rise for the generated voltage is defined as dv/dt. When the switch is turned off,
the magnetizing inductance current is at its peak value (Im). Since the current through a
capacitor is proportional to the time-rate of the stored voltage, the rate of rise is:
o
m
o
o
nC
I
C
I
dt
dv
(1-9)
Chapter 1
50
This equation is valid till the current through the capacitor is approximately constant;
otherwise the rate of rise completely depends on the resonant frequency.
The idea of parallel and series configurations of the pulsed power modules is inspired by
both equations (1-8) and (1-9), as these equations show the effect of stored current level,
magnetising inductance and the output capacitor on the generated output voltage and the rise-
time.
Considering Fig 1.37(a) and the fact that each transformer acts as a current source, the
stored energy in the primary side is doubled when two modules are connected in parallel.
This stored energy will be transferred to the output capacitors; hence considering (1-6) and
(1-7) the output voltage magnitude and its rate of rise are as follows:
op
mmparallelo
C
LIV 2
(1-10)
op
o
Parallel C
I
dt
dv 2
(1-11)
where Cop=Co1+Co2. Equations (1-10) and (1-11) indicate the importance of the output
capacitor. Even if the modules are paralleled, the output capacitance can affect the output
voltage level and rise-time and can keep the performance same as a single module. Therefore,
the parallel configuration is beneficial in pulsed power applications when Cop=Co. This
happens in the case of R-C load when the load capacitance is much bigger than the power
supply output capacitance.
Second alternative is series connection of the pulsed power modules, as illustrated in Fig
1.37(b). Here, the injected energy to the output is also doubled. But as described, the
generated voltage features also depend on the output capacitance. Taking into account the
equations (1-6) to (1-9), the output voltage magnitude and its rate of rise can be expressed as:
os
mmSerieso
C
LIV 2
(1-12)
os
o
Series C
I
dt
dv
(1-13)
Chapter 1
51
where Cos is o2o1
o2o1
CC
CC
. In series modules, same as in parallel modules, the output voltage
level and rise time improve as the output capacitance decreases. Here the series modules are
beneficial in both level of voltage and its rate of rise when Cos<Co.
In pulsed power applications two different cases may occur. One is when the load
capacitance is lower than the output capacitance of the pulsed power supply. In this situation
series modules exhibit better performance as illustrated in Table 1.4. The second case is when
the load has much higher capacitance, therefore all series, parallel and single modules have
same output capacitance. Under this circumstance the parallel connection has better
performance (see Table 1.5).
Hence, depending on the load capacitance one of the modules is beneficial to use.
Regarding the voltage level and rate of rise as presented in Table 1.4 and 1.5, it is possible to
generate wide range of voltage levels and improve dv/dt by connecting N modules in series or
parallel. This feature increases the flexibility of the pulsed power supply in varied
applications.
Table 1.4 Output voltage and rate of rise when Co1=Co2=… CoN=Co.
N Modules connected in oV
dt
dv
Parallel Singleo
o
mm V
C
LI
Singleo
o
dt
dv
C
I
Series Singleo
o
mm NV
C
LNI
Singleo
o
dt
dvN
C
IN
Table 1.5 Output voltage and rate of rise when Cop=Cos=Co.
N Modules
connected in
oV
dt
dv
Parallel Singleoo
mm VN
C
LIN
Singleo
o
dt
dvN
C
IN
Series Singleo
o
mm VN
C
LIN
Singleo
o
dt
dv
C
I
Chapter 1
52
In addition, regarding (1-5) and the calculated output voltages in Table 1.4 and 1.5, the
output voltage can be easily adjusted for the required level whether by Vin or D. For example
(1-14) shows the output voltage of series module based on the results presented in Table 1.5
as:
.TDCL
VNV
om
ino
(1-14)
Usually this adjustment is performed by limiting the current by selecting suitable duty cycle
(D). But Im should be calculated in a way that Im<Isat, where Isat is the current saturation level
of the transformer.
As Fig 1.37(b) shows, in the series connection each switch withstands its own module
reflected output voltage, while in the parallel connection; each switch should tolerate the
whole reflected output voltage. This fact makes the series connected modules more
appropriate for generating high voltage waveforms using low voltage switches. Therefore, the
reflected voltage across the switch for N-series module can be rewritten as below:
maxmax1
)( osswitch vNn
Vv (1-15)
By comparing Fig 1.37(a) and 1.37(b) it is obvious that the idea of parallel and series
connections of the flyback modules is applied to the secondary side, while the primary sides
for both cases are paralleled. The main advantage of paralleling the primary side of the pulsed
power modules are: current sharing so the proposed idea is applicable for high power
applications, reducing the number of the power supplies to one at the input side and ability to
employ low current rating switches.
1.3.2.1.2. Operating Conditions
The above mentioned calculations are correct if three important points are considered in the
designing procedure of the pulsed power modules. These are:
1) The output capacitor should be completely discharged during each period; otherwise
the whole stored energy in the magnetising inductor will not transfer to the capacitor.
2) Synchronization of each switch gate signal is required. Delays between each module
gate signal reduce the performance of the system. This is important when higher
number of modules is used.
3) The damping factor (ζ):
Chapter 1
53
As shown in Fig 1.38, the entire circuit acts as a parallel RLC circuit (the current source is
an inductor). Hence, the output voltage and the rise time are the results of the resonance
effect of this circuit. The output voltage can be expressed as:
tstso eAeAtV 21
21)( (1-16)
where s1 and s2 are given as:
20
22,1 S
(1-17)
here, RC2
1 is known as neper frequency and
LC
10 is the resonance frequency.
0)0( to
V and oItCI
)0( are the primary conditions. I0 is the stored current in the
magnetizing inductor (Im=nIo) transferred to the secondary side. The coefficients A1 and A2
are determined by the primary conditions as:
02
02
20
212
LIAA
(1-18)
Now depending on the damping factor coefficient can have two different values:
Underdamped ζ <1 or α <w0 , C
L
R2
1
C
LIA
2
01
(1-19)
Overdamped ζ >1 or α >w0
01 RIA
(1-20)
Equation (1-19) shows that if the circuit is underdamped, the generated voltage amplitude is
independent of the load impedance, which is the situation in which the proposed method is
valid. But under overdamped condition as in (1-20), the output voltage amplitude depends on
the load impedance and the initial current. Hence, in overdamped condition, the parallel
connection has better performance.
In Fig 1.39, the effect of damping factor on the output voltage and rise time is illustrated. A
decreasing damping factor results in a higher voltage level and faster rise time.
Chapter 1
54
Fig. 1.39 Effect of damping ratio on output voltage and rate of rise in a single flyback module.
Considering the above mentioned features, this effect should be considered especially in
low impedance applications such as the liquid discharge. Therefore, regarding the mentioned
characteristics, the proposed method is more beneficial for high impedance-capacitive load
such as DBD (Dielectric Barrier Discharge) loads.
1.3.2.1.3. Simulation Results
To verify the proposed method and the theoretical analysis, simulations under different
conditions were carried out. Here the performance of parallel and series flyback configurations
(Fig 1.37) is compared with a single flyback converter (Fig. 1.36). The software that was used
for simulation was Matlab 7.10. Table 1.6 shows the parameters value used in simulation
corresponding to Fig 1.36. In order to make the simulation results more close to reality, the
transformer core and copper losses (Rc and Rw) are also considered.
To evaluate the proposed method in developing the features of a single module pulsed
power supply, four distinct cases are considered. Case 1 and 2 have been conducted regarding
the conditions presented in Table 1.4 and 1.5, respectively. In these cases, to simulate a plasma
phenomenon, the load gets connected to the output after the output voltage reaches to a certain
voltage level (threshold) [137]. The different parameters in each case are indicated in Table
1.7.
Chapter 1
55
Table 1.6 Simulation Parameters
Vs
(v) Fs
(kHz) D
(%) Lm
(µH) Llp
(µH) Lls
(mH) Rc
(kΩ) Rw1,2
(Ω) n
10 1 8.9 160 2 4 1 0.1 10
Table 1.7 Simulation Parameters in Case 1 and 2
Parameters Value Case1 Case2
Single Parallel Series Single Parallel Series
Threshold 1000v 1000v 1978v 1000v 1410v 1410v
Co 4nF 8nF 2nF 4nF 4nF 4nF
Co1=Co2 4nF 4nF 2nF 8nF
Case1
In this case the performance of parallel and series connections of two identical power
supplies is compared with the single one. Therefore, in this case the output capacitor of each
module is equal (see Table 1.7). Fig 1.40 shows the obtained output voltage waveform. As can
be seen in the series connection, both voltage level and rate of rise are increased with the
factor of N, (N=2 is considered here).
Case2
If the load capacitance, which is paralleled with the power supply output, is much bigger
than the power supply output capacitance, then the entire single, parallel and series modules
will have the same equivalent output capacitance. In this case the output capacitance of all
modules is considered to be equal to 4nF (see Table 1.7). As depicted in Fig. 1.41, both
parallel and series modules have same voltage level but the parallel connection shows better
performance in the case of dv/dt.
Fig. 1.40 Output voltage waveform of single, parallel and series modules in Case1 (Co1=Co2=Co).
Chapter 1
56
Case3
In the case of time delay between gate drives, here the same situation as Case 2 is selected
but 0.2ms delay is considered between gate signals. Fig 1.42 shows the generated output
voltage for parallel and series modules. As can be seen the output amplitude never reached to
the threshold value. Therefore the load is not connected, so the capacitors are not completely
discharged. This creates another problem since the entire energy in the magnetizing inductor is
not fully transferred to the capacitor and the mentioned equations for estimating the output
voltage are not valid.
Case4
To consider the effect of low impedance load such as liquid discharge applications, in this
case series and parallel modules are considered, same as in Case 1, but RL=100Ω is directly
connected to the output of the modules. In such situations damping factor plays important role
in system performance. The selected resistive load puts all three power supplies in the
overdamped condition. As Fig 1.43 illustrates the parallel modules show better performance,
while series module had far better performance in Case1 when the system was underdamped.
The easiest way to overcome this problem, considering ζ in (1-19), is increasing the output
capacitance.
Fig. 1.41 Output voltage waveform of single, parallel and series modules in Case2 (Cop=Cos=Co).
Fig. 1.42 Output voltage waveform of single, parallel and series modules with 0.2ms delay between modules gate signals.
Chapter 1
57
Fig. 1.43 The effect of the load and the damping factor on the system performance.
Experimentation
Two laboratory prototypes for parallel and series connections, based on single module
flyback converter are implemented, to investigate performance of the proposed method
practically. Fig 1.44 shows the experimental hardware setup for two flyback modules.
Here 600V IGBT modules, SK25GB065, are used as power switches. Semikron Skyper 32-
pro gate drive modules are utilized to drive the IGBTs and provide the necessary isolation
between the switching-signal ground and the power ground. Four 1000 V diodes, STTH3010,
are connected in series for each module. A Texas Instrument TMSF28335 DSC (Digital
Signal Controller) is used for PWM signal generation. Two step-up transformer with an
UU100 core 3C90 grade material ferrite from Ferroxcube, are designed with N1 = 4 and N2 =
40. Magnetizing inductance and leakage inductance of each transformer is approximately
equal to 152µH and 1.6µH, respectively. Here a 5nF capacitor is placed across the switch to
damp the voltage spikes caused by leakage inductance. The circuit is implemented with Vs =
17V, fs = 1 kHz, 15% duty cycle and resistive load of 20kΩ.
Fig 1.45 depicted the measured results. Fig 1.45(a) shows the current sharing at the primary
sides. Due to parallel connection of the primary sides, in both parallel and series
configurations, the current is approximately equally shared between the two modules. As
mentioned before, parallel connection at the primary sides makes the proposed method
independent from using high current rate switches.
Chapter 1
58
The voltage pulses, in Fig 1.45(b), illustrate the better performance of series connection
over the other connections in both voltage level and rate of rise. In series modules the
maximum voltage level and rate of rise are 4.02 kV and 608 V/µs, respectively. While the
parallel and single modules achieved approximately same level of the voltage (2.37 kV) and
rate of rise (304 V/µs). As can be seen, the series modules performance is around 1.7 and 2
times better than the other modules in voltage level and rate of rise, respectively. From the
aforementioned theoretical analysis the performance of the series modules in voltage level
expected to be 2 times better, this difference is due to the resonance and dissipation
happening in the practical circuit. As the rate of charging is in terms of a time constant RC;
hence, as illustrated in the figure, series connection discharging time is shorter than the
parallel and single modules.
Fig. 1.44 Hardware setup for two flyback modules.
Chapter 1
59
Module1 primary current
Module2 primary current
(a)
(b)
Fig. 1.45 (a) current sharing at the primary side for both parallel and series connections, (b) output voltage
waveform of single, parallel and series modules in Case1 (Co1=Co2=Co).
Chapter 1
60
In this research the advantages of parallel and series configuration of flyback converters for
pulsed power applications are demonstrated. The proposed method aims at increasing the
voltage level and rise time of the generated pulses. Regarding to the above mentioned
evaluations by employing parallel and series connections it is possible to generate wide range
of voltage levels with improved rate of rise.
The extended analysis of the proposed method is presented in Chapter 2 and 3. The basic of
this work along with the simulation results were published and presented as a conference
paper entitled “Parallel and Series Configurations of Flyback Converter for Pulsed Power
Applications” at The 7th
IEEE Conference on Industrial Electronics and Applications in July
2012 in Singapore [139]. The hardware implementation and experimental results are
presented as a part of a journal paper entitled “High Voltage Modular Power Supply Using
Parallel and Series Configurations of Flyback Converter for Pulsed Power Applications”
published in IEEE Transaction on Plasma Science Journal, Oct 2012 [140].
1.3.2.2. Flexible and modular pulsed power supply
Pulsed power applications present one of the most varied ranges of loads in terms of load
behaviour and impedance. Hence, various load conditions adversely affect the power supply
flexibility. Regarding this, it is desirable to have flexible pulsed power supply that can cover
varied range of applications. This flexibility comes in terms of: adjustable output voltage,
controlling the flow of energy and repetition rate. Moreover, the generated voltage pulse
waveform must be load independent [53]. This is quite important especially in the low
impedance applications such as in water discharge.
1.3.2.2.1. High voltage modular f lyback topology
Considering the above study, combining the power converters found to be an efficient way to
improve the performance of pulsed power system based on generated voltage level and rate
of rise. Therefore, combining more than two power supplies sounds to be a good solution to
not only overcomes the power rating and operating speed problems but also increase the
system flexibility in generating different voltage levels. Moreover, due to modularity concept
easy diagnostics and ability to employ low voltage components are the other advantages.
Hence, as the first step to show the benefit of employing modular power supplies and
regarding the aforementioned advantages, series configuration is applied for 10 flyback
modules. Fig 1.46 illustrated the block diagram of the implemented hardware setup. To
decrease the number of switches, depending on the switch power ratings, a set of modules
Chapter 1
61
(here 5) is controlled with one switch. The transferred energy to the load can be controlled by
monitoring the primary stored energy. This is done through limiting the input current by
changing the duty cycle of the PWM signal in (1-5). As Fig 1.46 shows, a current sensor is
used for monitoring the input current. The experimental hardware setup for 10 series flyback
modules is depicted in Fig 1.47. Here 1700V IGBT modules (SKM200GB176D) are used as
power switches. Same gate drive modules and controller setup are used for this experiment as
the former one. Also same configurations for the transformers are used except here the
number of the cores per transformer is reduced to one. A HX10-NP is used as the current
transducer (CT).
Regarding the selected components each module is able to generate up to 4kV, which
makes the whole system capable of generating up to 40kV. The circuit is implemented with
fs=1kHz and VS=10V and 10% duty cycle. Each module has a 470pF capacitor (CO) and a
1MΩ high power resistor was connected as a load.
VSL
oa
d
Control
protocol of
switches
VLoad
+
-
Vo1
+
-
Vo2
+
-CO
CO
CO Vo3
+
-
CO Vo4
+
-
CO Vo5
+
-
Vo6
+
-
Vo7
+
-
Vo8
+
-
Vo9
+
-
Vo10
+
-
CO
CO
CO
CO
CO
Me
asu
rin
g th
e C
urr
en
t
Fig. 1.46 Block diagram of the implemented setup.
Chapter 1
62
Fig. 1.47 Hardware setup for 10-series flyback modules.
Fig. 1.48 Measured output voltage of 10 series modules (VLoad) and a single module (Vo1).
Fig 1.48 shows the measured practical results when the generated voltage is applied to a
resistive load. As can be seen from Fig 1.48 the voltage pulse of the series modules obtained
the peak amplitude of 20.8kV, while the single module has the peak voltage of 2.15kV. The
measured output voltage shows the rate of voltage rise of 8kV/µs in the series connection.
The peak voltage level and rate of voltage rise indicate approximately the 10 times (number
of series modules) better performance in comparison with a single module.
The proposed methods depicted advantage of generating high level of the voltages with
improved rise time, however, it is suitable for high impedance capacitive loads (such as DBD
loads) that means the load impedance can affect the system performance.
Chapter 1
63
Further analyses are presented in Chapter 3. The proposed method was published as a part
of a journal paper entitled “High Voltage Modular Power Supply Using Parallel and Series
Configurations of Flyback Converter for Pulsed Power Applications” published in IEEE
Transaction on Plasma Science Journal, Oct 2012 [140].
1.3.2.2.2. Modular switch-capacitor units
To design a flexible pulsed power supply which can be independent of load impedance but
still based on the modularity concept, here an efficient scheme that utilizes modular switch-
capacitor units in obtaining high voltage levels with fast rise time (dv/dt) using low voltage
solid-state switches is proposed. The proposed pulsed power supply has flexibility in terms of
controlling input energy and generating broad range of voltage levels. The energy flow can
be controlled by adjusting the utilized current source at the first stage of the system and
desirable voltage level can be obtained by connecting adequate number of switched-capacitor
units. The proposed topology is able to operate independently from the load while preserving
high repetition rate. Moreover, the control algorithm is designed in a way that it can prevent
from any possible faults and protect the system from the over-voltage.
Topology
Fig 1.49 illustrates the proposed method with a brief controlling algorithm scheme.
Generally, as can be seen the proposed topology consists of three different stages. The first
stage is a positive buck-boost converter as a current source and the energy storage section.
The next stage is the switch-capacitor units, which generates the required voltage level and
acts as a pulse generator, and the last stage is the load.
In order to have a flexible pulsed power supply, here the positive buck-boost converter
control the amount of stored energy by employing appropriate duty cycle for Si and SV. The
desirable voltage level can be obtained by connecting adequate number of switch-capacitor
units in series. This feature also makes the system capable of taking the advantage of
employing low voltage switches, as lower voltage rating switches have intrinsically better
behaviour [141].
In addition to the common aforementioned three sections, this topology utilizes a load
switch for low impedance applications such as water decontamination in which it disconnects
the load from the pulsed power supply before the required voltage level is obtained. This
makes this topology load independent, because the switch-capacitor units are not connected
Chapter 1
64
to the load while they are charging. In order to preserve a same break down voltage for the
load switch as the other switches, the switching signals need to be controlled precisely.
S11
S12
S21
S22
Sn1
Sn2
SL
Load
Si SVVS
L
D1
D2
C1
C2
Cn
Posi
tive
buck
-boost
conver
ter
Load Switch
Control
protocol of
switches
Sw
itch
-cap
acit
or
unit
s
Hysteresis
current control
Hysteresis
voltage control
Over
-rid
e si
gnal
Over
-rid
e si
gnal
Voltage
Feedback
Synchronizing
PWM Signal
Fig. 1.49 Proposed pulsed power topology for n level switch-capacitor units.
Control Strategy
The output current of the positive buck-boost converter need to be adjusted on a suitable
level. A hysteresis control is used for stabilizing this parameter. This control technique is
selected due to its simplicity, robust performance and stability [129, 142]. To charge switch-
capacitor units equally, the output voltage needs to be adjusted on an appropriate level. This
can be done same as the current control by utilizing the hysteresis method.
The charging process is synchronized with a PWM signal, which also determines the
repetition rate of the pulsed power supply. Here the procedure is explained for two switched-
capacitor units (Fig 1.50) which can be extended to n units. The procedure is controlled by
Sn1 and Sn2, which n is the capacitor unit number.
Chapter 1
65
When the PWM signal is high, the charging process begins with the first switch-capacitor
unit. At this mode (Fig 1.50a), S22 turns on and the stored energy form the current source
starts charging the capacitor of the first unit via anti-parallel body diode of S11. When the
capacitor voltage reaches to the required level the second mode (Fig 1.50b) starts by turning
S12 on and S22 off.
When the last unit charged up (here the second unit), SL and SV should be turned on prior to
applying all units voltages to the load (turning on the high-side switches). The reason is to
protect these switches from over voltage, as each switch blocking voltage is considered to be
equal to one capacitor unit voltage. As can be seen from Fig 1.50c in the third mode the load
is connected to the last unit and the current source is disconnected as SV is in the on-state.
Depending on the load impedance and its time constant in the worst scenario case a
low impedance load may fully discharge the last unit.
After SL and SV turned on, the forth mode occurred by turning on the high side switches of
the capacitor units. At this moment, as Fig 1.50d depicted, all of the capacitors connected in
series and a high voltage is applied to the load. Due to the last unit discharging in the
previous mode in the worst case the output voltage is:
Cout VnV )1(
(1-21)
where n is the number of switched-capacitor units and VC is the voltage of each unit.
Regarding (1-21) the voltage drop issue can be solved by increasing the number of the units.
Finally, when the PWM signal is low all of the high-side and low-side switches turn off and
on, respectively and the voltage across the load drops to zero (Fig 1.50e).
Regarding the mentioned charging and supplying the load process two different control
algorithms can be employed. The main difference between these two algorithms is the way
that the voltage feedback is provided.
One possible way to control the whole process is to provide individual voltage feedback
from each unit. Hence, the charging process of each unit can be started by monitoring the
previous stage charging level (Vch). Finally, when the unit’s capacitors charged to the
required level (VR) then it can be applied to the load. Fig 1.51a illustrated the algorithm
flowchart. The advantage of this algorithm is the ability of preventing from any possible
faults and indicating the failed unit as each unit’s voltage is measured. But the main
drawback is having multiple voltage feedback which increases the hardware implementation
complexity.
Chapter 1
66
The second potential way for the control algorithm is based on the measured voltage across
the output. This simplicity is the main advantage of this algorithm comparing with the
previous one, as it only employs one voltage feedback. Fig 1.51b shows the second algorithm
flowchart. As can be seen after charging each unit’s capacitor, low-side switches should turn
on in order to short circuit the output. Hence, at each charging stage the measured output
voltage corresponds to only one unit. By short circuiting the output after each charging stage
and using a flag (h in Fig 1.51b) it is possible to control the charging and supplying the load
process. The main advantages of the second algorithm are the ability to prevent from any
fault (as the output voltage is measured) and convenient practical implementation.
S11
S12
SL
LoadC1
S21
S22
C2
S11
S12
SL
LoadC1
S21
S22
C2
S11
S12
SL
LoadC1
S21
S22
C2
S11
S12
SL
LoadC1
S21
S22
C2
S11
S12
SL
LoadC1
S21
S22
C2
(a) (b)
(c) (d) (e)
Fig. 1.50 The charging and supplying the load operating modes.
Chapter 1
67
SK1: off
SK2: onK = 1,..., n
h =1
Start
If Sync
PWM=1
Yes
Yes
No
Sh1 & Sh2: off
Sg1: off
Sg2: on
g = 1,..., n – h
SL: off
If Vch=VR
NoYesIf VcK>VR
K =1,..., n
NoIf h = n
Yes
No
h = h + 1
Yes
SL & SV: on
Sn1: on
Sn2: off
SK1: off
SK2: on
K = 1,..., n-1 Delay
SK1: on
SK2: off
K = 1,..., n
If Sync
PWM=0
No
Charging unit h
SK1: off
SK2: on
Si: off
Fault detection
Supplying the Load
(a)
SK1: off
SK2: onK = 1,..., n
h =1
Start
If Sync
PWM=1
Yes
Yes
No
Sh1 & Sh2: off
Sg1: off
Sg2: on
g = 1,..., n – h
SL: off
If Vout=VR
NoYesIf Vout>VR
K =1,..., n
NoIf h = n
Yes
No
h = h + 1
Yes
SL & SV: on
Sn1: on
Sn2: off
SK1: off
SK2: on
K = 1,..., n-1 Delay
SK1: on
SK2: off
K = 1,..., n
If Sync
PWM=0
No
Charging unit h
SK1: off
SK2: on
Si: off
Fault detection
SK1: off
SK2: on
K = 1,..., n
Supplying the Load (b)
Fig. 1.51 Charging and supplying the load control algorithm flowchart, a) based on individual voltage feedback of each
unit and b) based only one voltage feedback.
Simulation Results and Analysis
In this section the proposed topology with three switched-capacitor units (n=3 in Fig 1.49)
is simulated in order to verify the performance of the proposed method. The software that
was used for simulation was MATLAB 7.10. Table 1.8 shows the parameters value used in
simulation corresponding to Fig 1.49. The upper and lower bands for adjusting the current are
selected as 20.5A and 19.5A, respectively. The charging level is selected as 1000V, and to
control the voltage of each unit on 1000V the bands are selected with 2V variations and fS is
the frequency of the synchronizing PWM signal. The delay time as mentioned in Fig. 1.51 is
selected as 1µs.
Table 1.8 Simulation Parameters
VS L C1,2,3 D fS iL VC1,2,3
100V 1mH 10nF %0.5 1kHz 20A 1000V
The effects of different loads impedance on the generated output voltage are depicted in Fig
1.52. For high impedance loads the considered delay for turning on the SL and SV doesn’t
Chapter 1
68
affect the output voltage level, but as the load impedance decreases the voltage drop across
the output voltage increases. This is due to the last unit capacitor discharging based on
the . For the low impedance applications this issue can be solved by increasing the
number of the switched-capacitor units.
Fig. 1.52 The effect of the load impedance on the generated output voltage.
The switching signals for the charging process and the generated output voltage are
demonstrated for 1kΩ load in Fig 1.53. As depicted during the charging process all of the
high-side switches are in the off-state. As can be seen, the last unit S31 turns on prior to the
other switches. Finally because the PWM signal becomes low, the output voltage drop to zero
before the load fully discharges the capacitors.
As mentioned before, in order to benefit from low voltage switches the break down voltage
of switches are selected to be equal to one capacitor unit voltage. Therefore, it is quite
important to turn on and off SL and SV effectively, otherwise they should tolerate the whole
output voltage. As Fig 1.54 shows, following the proposed switching strategy the voltage
across these two switches never exceeded from 1 kV. The only component in this topology
which must handle the whole output voltage is D2.
Chapter 1
69
Fig. 1.53 Voltage waveforms with relative switching signal patterns.
Fig. 1.54 Voltage across the critical components.
The proposed method is evaluated under different circumstances and obtained simulation
results indicate the effectiveness of the proposed approach. Various voltage levels can be
obtained by connecting different numbers of switched-capacitor units. The energy flow can
Chapter 1
70
be controlled via a positive buck-boost converter. The repetition rate can be adjusted using a
synchronizing PWM signal. Moreover, the proposed topology is load independent, as it is
disconnected from the load during the charging process. This feature makes the proposed
method suitable for supplying wide range of applications specially the one with low
impedance such as water discharge.
Chapter 4 presented extended analyses of this research. The proposed method was
published and presented entitled “A Flexible Solid-State Pulsed Power Topology” at the 15th
International Power Electronics and Motion Control Conference and Exposition, Novi Sad,
Serbia 2012 [143].
1.3.2.3. Pulsed power applications
After developing different pulsed power topologies studying and analysing different
application and load characteristics was considered at this stage. Applying pulsed power
supply at different voltage level and frequency and observing variety forms of plasma
generation were investigated. The designed pulsed power supplies under load condition not
only depicted a good performance but also the role of pulsed power for plasma generation in
industrial applications was also realized.
1.3.2.3.1. Effect of pulsed power on exhaust gas particle matter distribution
To evaluate pulsed power supply with a real world application, exhaust gas treatment was
selected as the first application. High voltage and high frequency pulses using a push-pull
based power electronics topology were applied to a DBD load and the effect of the generated
plasma was investigated.
Non-thermal plasma (NTP) treatment of exhaust gas is a promising technology for both
nitrogen oxides (NOX) and particulate matter (PM) reduction by introducing plasma into the
exhaust gases. This study considers the effect of NTP on PM mass reduction, PM size
distribution and PM removal efficiency. The experiments have been performed on real
exhaust gases from a diesel engine. The NTP is generated by applying high voltage pulses
using a pulsed power supply across a dielectric barrier discharge (DBD) reactor. The effects
of the applied high voltage pulses up to 19.44 kVpp with repetition rate of 10 kHz are
investigated. In this paper, it is shown that PM removal and PM size distribution need to be
considered both together, as it is possible to achieve high PM removal efficiency with
undesirable increase in the number of small particles. Regarding these two important factors
in this research, 17 kVpp voltage-level is determined to be an optimum point for the given
Chapter 1
71
configuration. Moreover, particles deposition on the surface of the DBD reactor was found to
be a significant phenomenon which should be considered in all plasma PM removal tests.
As mentioned, a pulsed power supply is employed to generate NTP using a dielectric
barrier discharge (DBD) reactor. Variety of DBD reactors existed, which essentially is a
multilayer capacitor and has been extensively used for various applications [40, 91, 144-149].
Also it can be a packed bed DBD reactor [36, 42, 150-155] or assisted by another catalyst or
adsorbent [24, 36, 151, 154, 156-160]. Here, a conventional DBD reactor with multipoint-to-
plane geometry is developed.
The particle size distribution of particulate matter from compression ignition engines has
become of increased concern. Particles differ in size, composition, solubility and therefore
also in their toxic properties. More than 90% of diesel exhaust-derived PM is smaller than 1
μm in diameter [161]. Most of the mass is in the 0.1–1.0 μm “accumulation” size fraction,
while most of the particle numbers are in the < 0.1 μm “nano-particle” fraction [162, 163].
These inhalable small particles are able to enter the bloodstream and even reach the brain
[164, 165]. It is believed that PM mass reduction and specially PM size distribution are not
fully considered in previous researches.
Regarding the aforementioned issues, the main concern of this study was to analyse the
effect of pulsed power on PM mass reduction and PM size distribution considering the pulsed
power effects on ultra-fine particles emitted from real diesel engine exhaust gas. In this
research, a pulsed power supply based on the push-pull inverter is developed to generate up
to 19.44 kVpp across the DBD load. The tests were conducted at different voltage levels with
fixed repetition rate of 10 kHz. PM mass reduction, PM removal efficiency and PM size
distribution are investigated by evaluating the results obtained.
Experimental Method
A. Experimental Setup
A schematic diagram of the experimental setup is shown in Fig 1.55. Experiments were
conducted on a modern turbo-charged 6-cylinder Cummins diesel engine (ISBe22031) at the
QUT Biofuel Engine Research Facility (BERF). The engine has a capacity of 5.9l, a bore of
102 mm, a stroke length of 120 mm, a compression ratio of 17.3:1 and maximum power of
162 kW at 2500 rpm. Particle number distributions are measured with a scanning mobility
particle sizer (SMPS) consisting of a TSI 3080 classifier, which pre-selects particles within a
narrow mobility (and hence size) range and a TSI 3025 condensation particle counter (CPC)
Chapter 1
72
which grows particles (via condensation) to optically detectable sizes. The SMPS software
increases the classifier voltage in a pre-determined manner so that particles within a 10-500
nm size range are pre-selected and subsequently counted using the CPC. The software also
integrates the particle number distribution to enable calculation of the total number of
particles emitted by the engine at each test mode. Gaseous emissions are measured with CAI
600 series gas analyses. CO2, NOX and CO concentrations can be measured by this gas
analyser, whereas particulate mass emissions are measured with a TSI 8530 Dust-Track II.
Fig. 1.55 Schematic diagram of plasma treatment system developed at QUT engine lab.
As depicted in Fig 1.55, three way valves to control the exhaust gas path ways are
employed. With this configuration, it is possible to measure both gaseous emissions and
particles before entering the reactor and after leaving it by changing the three way valve
directions. CO2 was used as a tracer gas in order to calculate the dilution ratio. CO2 was
measured either from the dilution tunnel with dilution ratios being calculated using the
following equation:
backgrounddiluted
backgroundexhaust
COCO
COCOratioDilution
,2,2
,2,2
(1-22)
Chapter 1
73
Laboratory background CO2 measurements were made before the commencement of each
test session. Every concentration measured after dilution should be modified by using a
dilution ratio.
B. DBD Reactor
A conventional dielectric barrier discharge reactor was designed for the experiments. Fig
1.56 shows a schematic of the reactor. As illustrated in Fig 1.56, it consists of two concentric
quartz tubes. Both tubes are 400 mm long and have a wall thickness of 1.5 mm. The outside
diameters of inner and outer quartz tubes are 20mm and 25mm, respectively. Exhaust gas
passes through the gap between these two quartz tubes. Based on pre-designed geometry, the
discharge gap is 1 mm. The DBD is connected to the pulsed power supply using internal and
external electrodes. The internal electrode is a copper cylinder and the external electrode is
made by a copper mesh that wraps the exterior part of the DBD. The electrodes are placed in
the middle of the DBD load with the length of 100 mm. Both tubes are fixed by two Teflon
caps at each end. Exhaust enters the reactor at the angle of 45 degree and flows throughout
the gap and leaves the reactor with the same angle.
(a)
(b)
Fig. 1.56 DBD reactor in Solidworks: a) schematic, b) cross-section.
C. Bipolar Pulsed Power Supply
Fig 1.57 shows a circuit schematic diagram of the pulsed power supply. As illustrated, it is
based on the push-pull inverter topology. The push-pull inverter contains two switches that
are driven with respect to ground. This is the main advantage of the inverter. This topology
Chapter 1
74
uses a centre-tapped transformer which is excited in both directions. A step up transformer is
used to boost the voltage and achieve galvanic isolation.
S2
S1
Vin
DB
D
NA
NB
NS
CS
CS
NA = NB << NS
Fig. 1.57 Pulsed power supply circuit schematic diagram (push-pull inverter)
The main reason to use a push-pull topology is to generate bipolar output voltage and
employing lower number of switches. The two switches S1 and S2 are switched alternately
with a controlled duty ratio to convert input DC voltage into high frequency AC voltage
suitable for exciting the DBD load. Hence, the generated output voltage is bipolar.
Adding a DBD load turns the push-pull inverter into a resonant stage with approximately
sinusoidal output. The frequency of the semi-sinusoidal shape signal is determined by an L-C
circuit comprising of the transformer inductance and capacitances of DBD and the
transformer. The repetition rate can be used to adjust the power and by optimizing the
resonance it is possible to obtain high frequency semi-sinusoidal waveform. A typical
measured output voltage of the employed pulsed power supply is depicted in Fig 1.58.
The first portion of the output voltage waveform is the resonant circuit dominated by the
magnetizing inductance of the transformer and the capacitances of the transformer and DBD.
The period of this signal is approximately 11.2 µs. The second one is the resonance
happening during the switches off-state between the leakage inductance and the capacitances
of the transformer and DBD. The period of this signal is equal to 34.8µs. As can be seen from
the figure the repetition rate is set to 10 kHz.
Fig 1.59 shows the experimental hardware setup for the pulsed power supply. Here 1200V
IGBT modules, SK75GB123, are used as power switches. Semikron Skyper 32-pro gate drive
modules are utilized to drive the IGBTs and provide the necessary isolation between the
Chapter 1
75
switching-signal ground and the power ground. A Texas Instrument TMS320F28335 DSC is
used for PWM signal generation. A centre-tapped step-up transformer with an UU100 core
3C90 grade material ferrite from Ferroxcube, are designed with NA = NB = 5 and NS = 293.
Here a 470pF capacitor (CS) is placed across each switch to protect them against the voltage
spikes. The output voltage is measured and captured using a Pintek DP-22Kpro differential
probe and RIGOL DS1204B oscilloscope, respectively.
Fig. 1.58 A typical measured output voltage of the employed pulsed power supply.
Fig. 1.59 Electrical Hardware setup with the DBD load.
Results and Discussion
A. Plasma Effect on PM Size Distribution
The effect of plasma on emission treatment is considered for varied experiments based on
the aforementioned Cummins diesel engine. In all experiments engine speed and load are
kept constant at 40 kW (25 %load) and 2000 rpm, respectively. A portion of raw exhaust gas
Chapter 1
76
directly from an iso-kinetic sampling port of the tailpipe was diluted with air and passed
through the reactor. Emissions concentration is measured before and after the applying pulse
to study the plasma effects. In addition, the median particle diameter which is another useful
parameter to study the effect of plasma technique is also measured. It is to be noted that all
illustrated results in each experiment have been obtained as an average over three consecutive
measurements.
The measured output voltages show the rate of voltage rise of 2840V/µs. In these
experiments, the output voltage amplitude is controlled by changing the input DC voltage
between 72.4 V, 84.4 V, and 94.8 V for generating output voltages of 15 kVpp, 17 kVpp, and
19.44 kVpp respectively. Under same situation, Fig 1.60 shows an image of DBD recorded at
19.44 kVpp. As can be seen, NTP is clearly occurring between the two electrodes.
Fig. 1.60 DBD image.
In the first experiment, the maximum voltage level (19.44 kVpp) is applied. To estimate
the deposition rate on the reactor surface, emissions in the reactor inlet (reactor inlet no pulse)
and reactor outlet (reactor outlet no pulse) without applying any pulse voltage are measured.
Finally, the pulsed power supply is applied across the DBD and the emissions in reactor
outlet are measured (reactor outlet with pulse). The same process is employed in all following
experiments.
Fig 1.61 illustrates the particle size distribution at 19.44 kVpp. There is a considerable
amount of PM deposition on reactor surface which is likely related to small gap between the
tubes (1mm). The median diameter in the reactor inlet is 70 nm, while in the reactor outlet is
decreased to 66 nm. This shows that larger particles deposited more on the reactor surface.
By applying pulsed power, as shown in this figure, the median diameter decreases
remarkably to 35 nm. This implies that lots of big particles have been oxidized or broken to
small particles by producing plasma inside the exhaust gasses at this voltage level.
Chapter 1
77
Fig. 1.61 Particle Size Distribution (2000rpm, 25%Load, 19.44 kVpp).
The peak value of particle number at the reactor inlet is around ⁄ .
This value declines to ⁄ at the reactor outlet due to deposition. By
applying pulsed power the graph peaks to 4.5 ⁄ which is approximately
four times bigger than the particle number at reactor outlet without any pulse. The number of
particles with diameter of less than 70 nm in the reactor outlet with applying pulse is higher
than the particle numbers at the reactor outlet without any pulse. This effect increases even
more at particle diameters less than 50 nm. At this level, the particle number at reactor outlet
surpasses the number of particles at reactor inlet. These findings imply that the 19.44 kVpp
pulse power at 10 kHz, increases the number of small particles considerably.
The effect of voltage level with 17 kVpp is considered in the second experiment. The
results obtained have been summarized in Fig 1.62. The figure shows the median diameter
changes from 70 nm at reactor inlet to 78 nm at reactor outlet without any pulse, and then
falls to 75 nm at reactor outlet with applying pulse. This shows that the larger particles are
deposited and also removed by plasma selectivity compared to smaller particles. As can be
seen, at this voltage level (17 kVpp) the number of small particles has not increased. This is
an important feature when compared with the previous experiment (19.44 kVpp).
Chapter 1
78
Fig. 1.62 Particle Size Distribution (2000rpm, 25%Load, 17 kVpp).
The last experiment is conducted by applying 15 kVpp pulses. The measured results are
depicted in Fig 1.63, which shows no growth in the number of small particles as well as
second experiment (17 kVpp). The maximum of PM concentration took place at around 71
nm. There is a small difference between PM concentrations with and without applying
plasma. Therefore, this level of voltage can remove particles at the same rate of deposition.
Comparing the values of median diameter in the reactor inlet and outlet shows that the
smaller particles are more likely to be deposited inside the reactor under the no pulse
condition. On the other hand, by applying the pulse the median diameter is in same range of
reactor inlet. There is no increase in the number of small particles at this voltage level. This
trend in median diameter variation is almost in complete agreement with the voltage of 17
kVpp.
Chapter 1
79
Fig. 1.63 Particle Size Distribution (2000rpm, 25%Load, 15 kVpp).
B. PM Removal Efficiency
PM removal can be calculated based on the following equation:
(1-23)
where and are the concentrations of PM (particle/cm3) at different PM
diameter. Fig 1.64 illustrates the PM removal efficiency at 19.44 kVpp. PM deposition on the
reactor surface increases with the PM size and gets to the maximum value of 70% removal at
around 80 nm. But after 80nm PM removal decreases again. For most of the particle sizes,
PM deposition on the reactor surface is more than 40%. When a 19.44 kVpp voltage is
applied to the reactor, PM removal efficiency reaches the value of 90% for larger particles.
Removal efficiency for particulate matter less than 80 nm is less than PM removal without
applying any pulse. This indicates undesirable operation of the plasma within this particle
size range. Also there is no removal for particles smaller than 50 nm. We suggest that this
incense in particle numbers can be related to the following two factors: Firstly, fragmentation
of larger particles by electron impact reactions or incomplete oxidation and secondly,
oxidation of exhaust gaseous emissions to particle by plasma ozone generated.
Chapter 1
80
Fig. 1.64 PM removal as a function of PM size (2000rpm, 25%Load, 19.44 kVpp)
Fig. 1.65 PM removal as a function of PM size (2000rpm, 25%Load, 17 kVpp)
Fig 1.65 Shows the PM removal at 17 kVpp which shows that PM removal by plasma is
more effective than deposition removal. This means that at this level of voltage, all deposited
particles and also some large particles inside the flow can be removed or oxidized. For
particles smaller than 35nm, PM removal percentage by deposition is higher than PM
removal by plasma. Apparently, smaller particles cannot be removed by plasma within this
Chapter 1
81
Fig. 1.66 PM removal as a function of PM size (2000rpm, 25%Load, 15 kVpp)
range (<35nm). However another possibility can be dissociation of larger particles to smaller
ones by electron impact reactions.
PM removal efficiency at 15 kVpp is illustrated in Fig 1.66. As can be seen the PM removal
for both graphs increases to an optimum value then decreases. The maximum PM removals in
reactor outlet with and without the pulse are 61% and 69% respectively. For particles larger
than the 60 nm diameter, PM removal when applying pulse voltage is slightly higher than PM
removal without any pulse. However, for particles with smaller diameters, these values are in
the same ranges.
Regarding the results obtained, it can be concluded that the voltage level has an important
role in the size dependent removal efficiency. At 15 kVpp the particle size distribution has
been affected slightly. The PM removal without producing small particles can be improved
as the voltage increases to 17 kVpp. The increase in the number of small particles has been
noticed when the voltage level goes up to 19.44 kVpp, while larger particles have been
reduced considerably. By taking into account all the above mentioned features, the 17 kVpp
experiment shows better efficiency in terms of particle size distribution for the given
configuration.
C. Plasma Effect on PM Mass Reduction
After studying the effect of different voltages on particle size distribution, in this section,
the effect of plasma on particle mass reduction is considered. In a similar way to the previous
Chapter 1
82
sections three different voltage levels have been applied to the DBD load. All results obtained
have been summarized in Table I. PM mass concentrations in the reactor inlet were 4.56 (
),
4.26 (
) and 5.14 (
) respecting to the variation of engine operating conditions for the
three tests conducted. These values decreased to 2.48 (
), 3.68 (
) and 4.37 (
) at reactor
outlet in the no pulse condition, respectively. This shows a significant particle deposition
inside the reactor. When the pulsed power is applied, plasma PM removal occurred. This
causes the PM mass concentration reduction of 43.9 %, 38.6% and 27.1% at 19.44 kVpp, 17
kVpp and 15 kVpp respectively.
The maximum PM mass reduction has been obtained when the voltage level is 19.44 kVpp.
However, according to the particle size distribution measurements, this voltage level
increases the number of small particles, which is not a desirable feature. The 17 kVpp applied
voltage shows a more suitable performance with good mass reduction of around 40% without
any increase in ultra-fine particle numbers. The 15 kVpp voltage level is found to be almost
the threshold breakdown voltage for the given configuration, below which no significant PM
mass reduction, PM removal, and PM size distribution which have not been affected too
much.
In this research the industrial application of pulsed power along with new findings, such as
PM distribution, PM mass reduction, and optimum operating point have been investigated. In
addition, the capability of employing power electronics topology to generate suitable voltage
levels with controllable parameters for a pulsed power application is demonstrated.
Table 1.9 PM Mass Reductions at Different Voltage Levels
Applied Voltage
Measurement
19.44 kVpp 17 kVpp 15 kVpp
Reactor Inlet PM
Concentration (3m
mg)
4.56 4.26 5.14
Reactor Outlet No Pulse PM
Concentration(3m
mg)
2.84 3.68 4.37
Reactor Outlet By-Pulse PM
Concentration (3m
mg)
2.56 2.62 3.74
Plasma PM Removal
Efficiency (%)
43.9 38.6 27.1
Chapter 1
83
Chapter 5 presented extended description about this work. This study entitled “Effect of
Pulsed Power on Particle matter in Diesel Engine Exhaust Using a DBD Plasma Reactor” is
published in IEEE Transactions on Plasma Science Journal on Aug 2013.
1.3.2.3.2. Analyzing plasma lamp intensity versus power consumption using
pulsed power supply
To study another type of DBD load with multi-point to multi-point geometry, a plasma lamp
is considered as the second application.
Recently, DBD lamps have gained much attention as they are mercury free, easily scalable
and simple to construct [166]. DBD lamps are mostly filled with rare gases and rare gas-
halides, which can provide efficient scheme for generating incoherent UV and VUV
(Vacuum UV) radiation [166]. The range of applications is broad owing to the variability of
output wavelength (88–310 nm) with the choice of gas fill.
DBD Lamps can be operated with continuous excitation or with pulsed excitation. Because
a pulsed discharge can operate at much higher peak voltages and peak currents for the same
average power as in a dc glow discharge, higher instantaneous sputtering, ionization and
excitation can be expected and hence better efficiencies [166]. Therefore, a solid-state pulsed
power supply can be a suitable choice for exciting a DBD lamp as it provides controllable
variables such as duty cycle, frequency, generated voltage, and etc.
One of the important features of a DBD lamp is the light intensity. The light intensity is quit
important in sensitive applications such as in therapy of skin diseases and for disinfecting the
skin surface. Hence, increasing or decreasing generated UV intensity can harm or not
efficiently affect the under treatment area respectively. The light intensity can be controlled
by factors such as frequency (pulse repetition rate) and applied voltage. However, another
important issue which needs to take into account is the lamp power consumption. Increment
in power consumption not only reduce the system performance but also can lead to short
operational lifetime of the lamp. Thus, finding the relation between the lamp intensity and
power consumption in order to select a proper operating point is important.
To study the behaviour of a DBD lamp, light intensity versus power consumption is
analysed at different voltage level and repetition rates. The DBD lamp is excited using a
push-pull based pulsed power supply. The light intensity is measured using a UV detector
and the power consumption is measured based on Lissajous V-Q diagram.
Chapter 1
84
Experimental Setup
A. DBD Lamp
Here a DBD lamp, as shown in Fig 1.67, is selected which is due to their particular interest
as they possess high efficiency [166]. The lamp tube has a coaxial geometry with a double
dielectric barrier, and is fabricated from UV grade fused silica tubing. The external electrodes
are wire/wire mesh yielding an active region of ~60mm length with a discharge gap of ~9mm
between the inner and outer dielectrics. The gas fill (sealed off) is a Xenon buffer at ~100mb
pressure, seeded with a low partial pressure (~1torr) of halogen (Cl) to yield XeCl excimers
upon discharge excitation which emit in the UV at 308m. Typical output irradiances from the
lamp are 1-10mW per cm2.
Fig. 1.67 Dielectric barrier discharge lamp
B. Pulsed Power Supply
Fig 1.68 shows a circuit schematic diagram of the developed pulsed power supply. As
illustrated, it is based on the push-pull inverter topology. The push-pull inverter contains two
switches that are driven with respect to ground. This is the main advantage of the inverter.
This topology uses a centre-tapped transformer which is excited in both directions. A step up
transformer is used to boost the voltage and achieve galvanic isolation.
The main reason to use a push-pull topology is to generate bipolar output voltage and
employing lower number of switches. The two switches S1 and S2 are switched alternately
with a controlled duty ratio to convert input DC voltage into high frequency AC voltage
suitable for exciting the DBD load. Hence, the generated output voltage is bipolar.
Adding a DBD load turns the push-pull inverter into a resonant stage with approximately
sinusoidal output. The frequency of the semi-sinusoidal shape signal is determined by an L-C
Chapter 1
85
circuit comprising of the transformer inductance and capacitances of DBD and the
transformer. The repetition rate can be used to adjust the power and by optimizing the
resonance it is possible to obtain high frequency semi-sinusoidal waveform.
Fig 1.69 shows the experimental hardware setup for the pulsed power supply. Here 1200V
IGBT modules, SK75GB123, are used as power switches. Semikron Skyper 32-pro gate drive
modules are utilized to drive the IGBTs and provide the necessary isolation between the
switching-signal ground and the power ground. A Texas Instrument TMS320F28335 DSC is
used for PWM signal generation. A centre-tapped step-up transformer with an UU100 core
3C90 grade material ferrite from Ferroxcube, are designed with NA = NB = 5 and NS = 293.
Here a 470pF capacitor (CS) is placed across each switch to protect them against the voltage
spikes. To calculate the power consumption a 4.2nF capacitor (Cm) is connected in series
with the DBD lamp. Cm is selected large enough to not to affect the DBD lamp capacitance.
The output voltage is measured and captured using a Pintek DP-22Kpro differential probe
and RIGOL DS1204B oscilloscope, respectively.
S2
S1
Vin DB
D L
am
p
NA
NB
NS
CS
CS
NA = NB << NS
Cm
Fig. 1.68 Pulsed power supply circuit schematic diagram (push-pull inverter)
Fig. 1.69 Electrical Hardware setup with the DBD load.
Chapter 1
86
C. Measurements
The lamp intensity was measured using a UV-photodetector (see Fig 1.69). The photo-
detector is equipped with JIC157 which has a spectral range of 210...390 nm. The averaged
intensity was considered for evaluation regarding to the area of the captured intensity
waveform and employed frequency.
Fig. 1.70 V-Q cyclogram of the DBD lamp as a basis of power consumption calculation.
To measure the power consumption of the DBD lamp, the energy transferred to the DBD
lamp has been calculated by applying the Lissajous (V −Q) diagram [167]. To measure Q one
capacitor (Cm) is placed in series with the DBD lamp, as depicted in Fig 1.68. Thus, by
measuring the voltage across Cm and multiplying it by Cm value it is possible to calculate Q.
The energy consumed by the plasma for one cycle is calculated from the area of V−Q curve
for different experiments (see Fig 1.70). Hence, by considering the employed repetition rate
(frequency) it is possible to calculate the average consumed power by the DBD lamp. Due to
presence of noise during measurement the captured data is filtered before power consumption
estimation stage (see Fig 1.70).
Results and Discussion
To analysis the lamp characteristics the lamp intensity versus the lamp power consumption
is considered at different voltage level and operating frequency. Input voltage (Vin) and
repetition rate (fr) were assigned as controlling parameters to determine the plasma lamp
characteristics. The high voltage pulses were applied to the plasma lamp at 7 different voltage
levels and 4 different repetition frequencies of 2kHz, 5kHz, 10kHz, and 15kHz (see Table
1.10). The measured results are illustrated as averaged intensity versus averaged power
Chapter 1
87
consumption (see Fig 1.71). It is to be noted that the results are normalized for a better
comparison.
Table 1.10 Considered voltages and frequencies for different experiments
Frequency
Output Voltage Level
2kHz
5kHz
10kHz
15kHz
Vin= 36V D11 D21 D31 D41
Vin= 40V D12 D22 D32 D42
Vin= 44V D13 D23 D33 D43
Vin= 48V D14 D24 D34 D44
Vin= 52V D15 D25 D35 D45
Vin= 56V D16 D26 D36 D46
Vin= 60V D17 D27 D37 D47
(a)
(b)
Fig. 1.71 Comparing different operation points regarding to the varied applied voltage levels and repetition
rates, a) applied voltage versus plasma lamp power consumption, b) applied voltage versus plasma lamp
intensity.
Chapter 1
88
The obtained results clearly indicate that the DBD lamp power consumption
correspondingly increases with the applied voltage level and the repetition rate. Despite the
power consumption which is always desirable to be as low as possible, it is the lamp intensity
which is the main goal. Obtaining high intensity at low power consumption in plasma lamp
applications is one of the major concerns. The obtained results regarding to the applied
voltage versus the lamp intensity are illustrated in Fig. 1.71. The measured results indicate
that not necessarily the higher intensity the high power consumption is. In the other word, the
intensity doesn’t have a linear relation with the power consumption. The measured results in
Fig. 1.71 depicted the possibility of achieving same or even higher intensity at lower power
consumption by increasing the repetition rate. For instance, comparing D44 with D36 shows
that, D44 has higher intensity than D36 while it has same power consumption. Such
behaviour can be realized by comparing D32 with D27 as well. This can be due to the
different slope of change regarding to the applied voltage and consumed power.
Regarding to the aforementioned facts, it is feasible to find an optimum operation point
where the required intensity can be achieved at lower power consumption. This can be
obtained by operating at lower voltage levels but with the cost of higher repetition rate.
Reducing the generated voltage level while obtaining the required performance is always
desirable due to high voltage insulation and power switches limited breakdown voltage
issues.
More analyses are provided in Chapter 6. This work entitled “Analysing DBD Plasma
Lamp Intensity versus Power Consumption Using a Push-Pull Pulsed Power Supply” is
accepted in 15th European Conference on Power Electronics and Applications to be held in
Lille, France, 2013. We thank Dr. Robert Carman of the Department of Physics and
Astronomy at Macquarie University, Sydney, Australia, for the helpful advice, and loan of
the XeCl lamp which was originally supplied by Prof. G. Zissis, LAPLACE Laboratory,
Universite Paul Sabatier, Toulouse, France.
1.3.3. High frequency high power converters for ultrasound
applications
The second aspect of this research project was improving the efficiency of high power
piezoelectric applications by applying high frequency high power converters. Extensive study
was conducted on high power piezoelectric device characteristics in order to generate suitable
excitation signal. Variety of power electronics topologies and control techniques were
Chapter 1
89
investigated in order to excite the high power piezoelectric device at the desirable range of
frequency. Two major steps were considered in this study:
Investigating the characteristics of a high power piezoelectric transducer
Improving the efficiency high power ultrasound system
1.3.3.1. Power electronic converters for high power ultrasound
transducers
Ultrasound power transducer performance is strongly related to the applied electrical
excitation [52, 98]. To have a suitable excitation for maximum energy conversion [5, 115-
119], it is required to analyse the effects of input signal waveform, medium and input signal
distortion on the characteristics of a high power ultrasound system (including ultrasound
transducer). In this research to investigate the features of a suitable excitation signal, varied
condition is considered. Different input voltage signals are generated using a single-phase
power inverter and a linear power amplifier to excite a high power ultrasound transducer in
different mediums (water and oil) in order to study the characteristics of the system.
Moreover, the effect of power converter output voltage distortions on the performance of the
high power ultrasound transducer using a passive filter is also analysed.
To study the performance of the piezoelectric transducer under the applied situation, an
ultrasound system was considered. The frequency responses of the applied voltage and the
measured voltage are considered as it is one of the most efficient ways to study the behaviour
of a system. In addition to that, it is possible to understand the linear or non-linear behaviour
of the system and the transducer (main goal) based on the applied signal and measured
response by employing the superposition method.
Experiments and Results
Table 1.11 shows a summary of all test conditions which have been carried out at different
mediums and input voltages. As can be seen, for low voltage evaluations sinusoidal
waveforms are generated by a linear power amplifier (OPA459) and a step-up transformer
(see Fig 1.72). But due to existing restrictions for the higher voltage levels, a single phase
inverter is used which results in a non-sinusoidal waveform.
The employed piezoelectric transducer has some resonant frequencies around 39 kHz. To
study the piezoelectric characteristics using superposition method and as the highest energy
Chapter 1
90
conversion happens at resonant frequencies one of the signals is selected at 39 kHz the other
one is selected at 61 kHz which is a non-resonant frequency.
Table 1.11 Test conditions and setups
Medium Excitation System
Input Voltage Magnitude
(peak)
Frequency of Input
Signal (s)
Test 1 water
Signal Generator, a power
amplifier and a high
frequency transformer
15 V and 30 V 39 kHz and 61 kHz
Test 2 water
Signal Generator, a power
amplifier and a high
frequency transformer
15 V + 30 V (time
domain) 39 kHz and 61 kHz
Test 3 oil three-level inverter 50 V, 100 V, 200 V and
300 V 39 kHz
Test 4 oil three-level inverter and a
filter
50 V, 100 V, 200 V and
300 V 39 kHz
Test 5 water three-level inverter 50 V, 100 V, 200 V and
300 V 39 kHz
Fig. 1.72 block diagram of a lab prototype for test 1 and test 2
In the first experiment, the first transducer was excited by applying a sinusoidal waveform
with two voltage levels (15V and 30V) at two different frequencies (39 kHz and 61 kHz).
The responses were captured across the second transducer in time domain. The obtained
results don’t show a proportional relation between the input and output voltages (see Fig
1.73). This can be due to the non-linear behaviour of the transducer. But to make sure the
system should be analysed under superposition principle. Based on this principle, if the
ultrasound system has linear characteristics, the responses of the system with two sinusoidal
signals (as follow) should be same:
a) Separately excited by two signals at 39 kHz and 61 kHz and adding the output signals
b) Simultaneously excited by two signals at 39 kHz and 61 kHz
Chapter 1
91
Therefore, in the second experiment, the same system is excited by two signals at 39 kHz
and 61 kHz which are added together at the input side of the power amplifier. It is to be
expected that the output voltages of the ultrasound system in two different tests should be
exactly same as each other if the system has a linear characteristics. To evaluate the linearity,
the previous obtained results are also added up together (based on superposition principle).
The measured result illustrated in Fig 1.74 shows that the output voltages (separately and
simultaneously excited) are not same, which implies the nonlinear characteristics of the
ultrasound system under the applied situation.
To study the non-linearity of ultrasound system at higher power ratings and different
mediums, experimentation were carried out using a single phase inverter. A coupling box of
laboratory setup is filled of oil or water for further experiments. A block diagram and the
experimental setup are shown in Fig 1.75 and Fig 1.76 respectively. The generated voltage
was a square wave uni-polar voltage waveform.
(a)
(b)
(c)
(d)
Fig. 1.73 (a) Input signals at 39 kHz (b) output signals at 39 kHz (c) Input signals at 61 kHz (d) output signals
at 61 kHz.
Chapter 1
92
(a)
(b)
(c)
(d)
Fig. 1.74 (a) test 1: summation of two output signals (at 39 kHz & 61 kHz) for Vin=15V and Vin=30V, (b) test
2: two output signals for Vin=15V and Vin=30V, (c) comparing the results of test 1 and test 2 for Vin=15V, (d)
comparing the results of test 1 and test 2 for Vin=30V.
Fig. 1.75 Schematic diagram of the setup using an inverter.
Chapter 1
93
Fig. 1.76 Hardware Setup.
(a)
(b)
Fig. 1.77 Captured results of third experiment
The third experiment was conducted at different voltage levels (50V, 100V, 200V and
300V) and at 39 kHz. Due to uni-polar modulation the generated signal has some harmonics.
To have a fare comparison all the input voltages are normalized so they have same amplitude.
Based on the same factor the captured output voltages are normalized as well (see Fig 1.77).
The results show that the ultrasound system has nonlinear characteristics at different voltage
levels and the output voltages are not same when they are normalized.
The output voltage of the power inverter generates voltage stress (dv/dt) across the
piezoelectric transducer. This voltage with the capacitive characteristics of the piezoelectric
transducer can generate significant current spikes which increases losses and high frequency
noise. The input voltage distortion can tend to deteriorate the output voltage quality and can
increase the power dissipation. Moreover, these distortions can cause the nonlinear behaviour
of the ultrasound system. Hence, a 990 μH inductor as a filter is placed between the power
converter and the transducer to reduce the input voltage distortion across the first transducer.
Chapter 1
94
Fig 1.78 shows a block diagram of the setup. The captured results after applying filter implies
that the ultrasound system still shows nonlinear characteristics (see Fig 1.79).
The final experiment was performed using water as medium same as third experiment but
this time by applying an inverter. Similar to the previous test results, when the input and
output voltages are normalized at 50 V, there are still significant differences between the
output voltages at different frequencies shown in Fig 1.80. The nonlinear behaviour of a high
power ultrasound system is obvious in this figure due to mismatch of the responses of the
ultrasound system to the normalized input voltages.
Fig. 1.78 Applying filter to the inverter output.
(a)
(b)
Fig. 1.79 Captured results of forth experiment (applying a filter).
(a)
(b)
Fig. 1.80 Captured results of the last experiment.
Chapter 1
95
Regarding to the mentioned experiments and evaluations, in this research, the nonlinearity
of the ultrasound system has been shown under varied situation. The obtained results indicate
the fact that for the high power ultrasound applications not necessarily the output voltage
increases with the increment in the input voltage. In addition, an inverter can be a suitable
choice for generating the required voltage even without filtering out the generated harmonics.
The extended analysis of this research is presented in chapter 7. This work was published
and presented entitled “Power Electronic Converters for High Power Ultrasound
Transducers” at The 7th
IEEE Conference on Industrial Electronics and Applications in July
2012 in Singapore [168].
1.3.3.2. Improving the efficiency of high power piezoelectric transducer
Regarding to the aforementioned outcomes the broad spectral content of the 2-level pulse
results in undesired harmonics which can decrease the performance of the system
significantly. On the other hand, it is crucial to excite the piezoelectric devices at their main
resonant frequency in order to have maximum energy conversion. Therefore a high quality,
low distorted power signal is needed to excite the high power piezoelectric transducer at its
resonant frequency. This paper proposes an efficient approach to develop the performance of
high power ultrasonic applications using multilevel inverters along with a frequency
estimation algorithm. In this method, the resonant frequencies are estimated based on relative
minimums of the piezoelectric impedance frequency response. The algorithm follows the
resonant frequency variation and adapts the multilevel inverter reference frequency to drive
an ultrasound transducer at high power. Extensive simulation and experimental results
indicate the effectiveness of the proposed approach.
The most efficient way to generate power signals is to use a 2-level power inverter.
Specifically, switch mode inverters are used for piezoelectric high power applications due to
their high power density, efficiency, low cost and size compared to conventional linear power
supplies [5, 102, 123]. However, the harmonics present in the output waveform produce
undesired side bands which are not suitable in many applications. Moreover, they also cause
unnecessary power dissipation which reduces the efficiency of the power converter [124,
125].
On the other hand, for ultrasound applications which operate at high fundamental frequency
power converters based on a PWM strategy with high switching frequency index (fs/fo)
cannot be a practical solution. In such applications, the maximum possible fundamental
Chapter 1
96
frequency of a converter is restricted by the switching transients of each power switching
event and the number of switching events per cycle.
In this regard, to reduce the number of switching transients and eliminate the undesired
harmonics, the most effective way is to use multilevel inverters [127-130]. Through these
inverters, it is possible to produce quasi-sine waves with low total harmonic distortion at high
power. Multi-level inverters can increase the quality and efficiency of the high voltage supply
compared to the conventional 2-level inverters. This permits the semiconductor devices to
operate at lower switching frequencies with higher efficiency as well as lower voltage stress
across switches and loads which minimize electromagnetic and ultrasound noise emissions.
Fig 1.32 depicted typical voltage waveforms and their harmonic spectra. It can be seen that
the harmonics are depressed for multilevel inverters such that they can be easily filtered out
from the frequency range of interest.
To improve the performance of the piezoelectric transducer for high power applications, in
addition to the multilevel converter, the device needs to be excited at its resonant frequency.
Piezoelectric devices typically have multiple resonant frequencies, but only the major
resonant frequency is generally targeted for excitation in practice. Structural and
environmental changes of a piezoelectric system can affect variations in the resonant
frequencies [5, 52]. Therefore, it is important to estimate the main resonant frequency in
order to maintain efficient system operation. Therefore, the multilevel converter needs to be
adapted with a suitable frequency estimation algorithm.
The most effective way to find the resonant frequencies of a piezoelectric transducer is by
evaluating its impedance frequency response [105, 108-113, 115, 117, 118]. A minimum in
the impedance response corresponds to a resonant frequency, fr. The impedance frequency
response is the ratio of the voltage spectrum to the current spectrum. To calculate the
piezoelectric impedance response, a voltage source needs to be applied to the device as an
excitation signal and current needs to be measured simultaneously. In order to obtain the
response of the device for a specific range of frequencies, the excitation signal should cover
the entire frequency range. The idea of calculating the piezoelectric impedance is inspired by
the general concept of performing system identification (i.e. finding the system transfer
function [121]). It is therefore possible to benefit from the existing knowledge base of
excitation signals for system identification. Considering the above mentioned features, a
broad-band excitation signal is the most appropriate candidate [120, 121].
This study advocates the advantage of using multilevel inverters along with an efficient
frequency estimation algorithm (as depicted in Fig 1.81) to develop the performance of the
Chapter 1
97
high power ultrasonic applications. In the proposed algorithm, a 1 kHz rectangular pulse with
10% duty cycle is applied to the device and the current is measured as the response. Then the
captured data is cropped to retain only one cycle of the applied input voltage and
corresponding output current. In the next step, FFT (Fast Fourier Transform) is applied to
both the signals and the impedance response is found as a ratio of the voltage to current
transforms. Finally, the relative minimum values are estimated and sorted according to both
impedance derivative and impedance magnitude. The FFT is used because it can be computed
relatively efficiently (with order NlogN operations), thus enabling real-time operations.
Finally, the multilevel inverter reference frequency is updated with the estimated resonance
frequency. The proposed method has been evaluated in simulations (using an electrical circuit
model) and experimentally (using two piezoelectric devices). The results obtained indicate
the efficiency and high performance of the proposed method.
DC/AC
Inverter
AC
Current Sensor
Piezoelectric
Frequency
Estimation
Fig. 1.81 Exploiting online resonance frequency estimation in high power applications of piezoelectric devices.
Methodology and Approach to Estimate Resonance Frequency
A. Excitation Signal
Piezoelectric devices have different resonance frequencies which are sometimes described
as vibration modes. To identify these frequencies, the excitation signal needs to be wide-
band. In the proposed approach, a sequence of rectangular pulses is applied to the device. The
pulse widths are 0.1ms and the pulses are reapplied every 1ms. It is quite easy to generate the
selected pulse stream and the pulses occur fast enough to enable the system to create regular
updates of the resonance frequency online.
Fig 1.82(a) shows the frequency response of a typical rectangular pulse. As can be seen the
energy in the spectrum is well spread, but varies considerably in intensity as a function of
frequency. This imperfection actually proves to an advantage because the practical spectrum
Chapter 1
98
does not suffer from the problem of having zero energy at some frequency components (see
Fig 1.82 (b)).
(a)
(b)
Fig. 1.82 Frequency responses: a) ideal pulse frequency response, b) practical pulse frequency response.
B. Estimating Impedance
The response due to the excitation signal (known as the ‘residual vibrations’) is given by:
n
k
kkk tatfSinAty1
)exp()2()( (1-24)
where n is the number of resonance frequencies, Ak is the amplitude, fk is the frequency and
ak is the damping coefficient of the thk resonance frequency.
A current sensor is used to capture the response (residual vibrations) of the device. The
current and voltage across the piezoelectric are captured simultaneously. To get the best
results, one cycle of the excitation pulse, along with the corresponding current response,
needs to be extracted from the captured signals. The starting point of the voltage waveform is
specified on the leading edge of the pulse. Then, from a knowledge of the sampling rate and
the length of the signal (which is 1ms) the end point is determined. Based on these starting
and end points the voltage and current waveforms are cropped from the captured signals.
After cropping, the FFT of the both signals are calculated as follows:
Chapter 1
99
1,...,1,0,)()(
1
1
2
NkenxkX
N
n
N
nki
(1-24)
where x(n) and X(k) are the discrete inputs and outputs respectively. To find the power
spectrum of the voltage and current signals the FFT outputs are multiplied by their conjugates
as per equation (1-25) and finally the impedance is calculated based on equation (1-26).
N
XXPx
.
(1-25)
NN
FffP
fPfP s
i
vz /)
2:0(,
)(
)()( (1-26)
where Fs denotes the sampling frequency and N is the number of FFT points.
C. Extracting Resonant Frequencies
The resonant frequencies correspond to the local (or relative) minimums of the piezoelectric
impedance. The relative minimums of a function are the points where that the slope of the
tangent changes from – to +. Fig 1.83(a) shows a power spectrum, of impedance and
the change of slope in the resonant frequency. At point A, is equal to zero. Moreover,
immediately to the left of this point at point B the slope is negative while at point C the slope
is positive.
Motivated by the above, the derivative of is calculated and all the points where
changes sign from – to + are extracted. As the final step, at the extracted frequencies, a
sorting is performed based on the magnitude of . The main resonant frequency is the
one with the lowest magnitude.
The procedure described above is depicted as an algorithm flowchart in Fig 1.83(b). Within
that flowchart R is a constant defining the number of resonant frequencies needing to be
extracted. The algorithm can be used both for offline and online systems. It is important to
note that, repeating the proposed algorithm, while applying a repetitive pulse and averaging,
results in an increase of the estimation accuracy.
Chapter 1
100
(a)
Input
captured
signals
Finding
starting and
end points
Cropping the
signals
Saving as
relative
minimum
Sorting using
amplitude
Yes
No
Finding the
minimum
amplitude
Number of extracted
frequencies = R
No
Yes
(b)
Fig. 1.83 Extracting resonant frequencies: a) change of slope at relative minimum (resonant frequency), b)
flowchart of the proposed algorithm.
Experimentation and Discussion
Chapter 1
101
The proposed algorithm is evaluated via both simulations and experimentation. The
comparisons were made based on three key types of alternative impedance analysing
methods. These alternative methods are:
a) The traditional method: In this method several single frequency sine waves from 30
kHz to 80 kHz have been applied separately to the piezoelectric. The frequency step
size was set to 1 kHz.
b) The Network Analyser method: An R&S ZVL3 vector network analyser has been used
to obtain the impedance frequency response of a piezoelectric device in the frequency
range of 30 kHz to 80 kHz.
c) The unit step and white noise excitation method: In order to compare the proposed
method with different wide-band excitation signals, the impedance frequency behaviour
has been obtained based on step response and white noise.
Step Pulse: From system identification theory, it is known that the Fourier transform
of the impulse response (h(t)) of a system gives the system transfer function. Since
generating an impulse ( ) at high power is highly impractical a step (u(t)) is
preferable. Here a 1ms step is applied to the device and as discussed in the previous
section the impedance response can be obtained by dividing the Fourier transform of
the output into the Fourier transform of the input.
White Noise: One of the most popular excitation signals used in system identification is
white noise [120, 121]. As white noise theoretically has a constant amount of energy per
frequency band, it is possible to simply look at the captured current to find the resonant
frequencies. To account for the fact that white noise spectra are not always perfectly flat
in practice, the impedance response is calculated in the same way as the other two
methods – by dividing the Fourier transform of the output into Fourier transform of the
input.
Two different piezoelectric devices Type A and Type B were considered for experimental
evaluation. As was the case with the simulation testing, the impedance of both devices was
obtained in the frequency range of 30 kHz to 80 kHz. For the proposed method a power
converter was used for generating pulses. For the traditional and broadband excitation
methods a G5100A function waveform generator was used as the signal generator source and
the signals were amplified using an OPA549. Fig 1.84 shows the experimental setup for the
proposed method. All signals were captured using a RIGOL DS1204B oscilloscope. As the
frequency was between 30 kHz and 80 kHz, the cut-off frequency of the filter was set to 100
Chapter 1
102
kHz. Hence capacitor and resistor values of Cf =1nF and Rf =1.59KΩ respectively were
selected for the low pass filter. The measured impedance frequency behaviour of the Type A
piezoelectric device for all methods is shown in Fig 1.85(a). Table 1.12 also shows the
extracted frequencies for the three strongest resonant frequencies.
Fig. 1.84 Experimental setup for the proposed method.
Fig. 1.85(b) shows the impedance frequency response of the second piezoelectric device.
The estimated frequencies for the first three resonant frequencies are shown in Table 1.12. As
can be seen from the measured results, the white noise method did not result in accurate
estimation of Fr2 and Fr3 compared to the other methods. The main reason is that in each
experiment the level of power in white noise changes randomly over time, and may even go
to zero at particular frequencies. Therefore for resonances which are not especially strong (as
was the case for Fr2 and Fr3 in the Type B device) the results are unreliable.
It should be noted that the electrical circuit model of the piezoelectric device is a simple
model which is not able to perfectly model the piezoelectric device nonlinearity, time
variations and high frequency behaviour. That is why there are small differences between the
simulation and test results. The advantages and drawbacks of the various methods are
summarized in Table 1.13. The proposed method offers simplicity and high performance in
addition to its ability to be used for online systems.
The performance of the proposed frequency estimation algorithm was evaluated in the
above sections, but as already mentioned, exciting the device at its resonant frequency is not
enough to achieve maximum power conversion. The excitation signal harmonics also need to
be considered. To do that an ultrasound interface is setup in the next section.
Chapter 1
103
(a)
(b)
Fig. 1.85 Experimental results obtained for the piezoelectric devices impedance response:
(a) Type A, (b) Type B.
Table 1.12 Estimated resonant frequencies of the Type A piezoelectric device
Method
Resonant Frequency(kHz) Fr1 Fr2 Fr3
Network Analyser 70.18 48.62 38.82
Impulse Response (Step Excitation) 70.31 48.83 39.06
White Noise 70.31 48.83 39.06
Traditional 70 49 39
Proposed 70.31 48.83 39.06
Chapter 1
104
Table 1.13 Estimated resonant frequencies of the Type B piezoelectric device
Method
Resonant Frequency(kHz) Fr1 Fr2 Fr3
Network Analyser 78.08 47.48 66.94
Impulse Response (Step Excitation) 78.13 46.88 66.41
White Noise 78.13 44.92 68.36
Traditional 78 47 67
Proposed 78.13 46.88 66.41
In this section the advantage of exciting a piezoelectric transducer using a multi-level
waveform at the resonant frequency compared with a uni-polar waveform is illustrated. For
the comparison, one multi-level waveform and one uni-polar waveform were generated with
peak to peak voltage of 120V at 39 kHz (Type A device resonant frequency).
To perform the evaluation, one pair of the Type A piezoelectric transducers was placed face
to face as sender and receiver. For the first experiment, a unipolar pulse was applied to one of
the piezoelectric devices and the voltage across the other one was captured. Fig 1.86(a) shows
the applied voltages. To show their influence on piezoelectric devices the frequency
responses of applied and captured voltages are illustrated in Fig 1.86 (b). As can be seen from
Fig 1.86(b) the captured response contains several harmonics, due to the excitation signal
having much energy away from the fundamental frequency.
For the next test, a multi-level waveform was applied to the piezoelectric device. As can be
seen from Fig 1.86(c), harmonic levels are attenuated significantly. This was due to the use of
a multi-level waveform which damped the harmonics in the vast area around the fundamental
frequency.
Comparing the Fig 1.86(b) with Fig 1.86(c), illustrates that the maximum energy is
achieved at the resonant frequency. Higher efficiency is obtained for multi-level signal. In
particular the method is quite effective for reducing the harmonic content. In addition, the
presence of harmonics not only adversely affects frequency sensitive applications, but also
causes an increase in temperature and an increase in power loss. Moreover, when using filters
for attenuating remaining harmonics the filter cost and size decreases when multi-level
topology is employed.
Chapter 1
105
(a)
(d)
Input voltage (unipolar)
(b)
Input voltage (multilevel)
(e)
Output voltage (unipolar)
(c)
Output voltage (multilevel)
(f)
Fig. 1.86 Obtained results for the ultrasound interface: (a) applied uni-polar pulse at 39 kHz in time domain, (b)
frequency response of the applied uni-polar pulse (input signal), (c) frequency response of the output signal when
the input is a uni-polar pulse, (d) applied multi-level pulse at 39 kHz in time domain, (e) frequency response of the
applied multi-level pulse (input signal), (f) frequency response of the output signal when the input is a multi-level
pulse.
Further description of this study is provided in Chapter 8. This work entitled “Improving
the Efficiency of High Power Piezoelectric Transducers for Industrial Applications” was
published in IET Science, Measurement and Technology Journal, in Feb 2012 [169].
Chapter 1
106
1.4. References
[1] S. Bernet, "Recent developments of high power converters for industry and traction
applications," Power Electronics, IEEE Transactions on, vol. 15, pp. 1102-1117, 2000.
[2] C. Carastro, C. Castellazzi, C. Clare, and W. Wheeler, "High-Efficiency High-Reliability
Pulsed Power Converters for Industrial Processes," Power Electronics, IEEE Transactions on,
vol. 27, pp. 37-45, 2012.
[3] J. A. Gallego-Juarez, "New technologies in high-power ultrasonic industrial applications," in
Ultrasonics Symposium, 1994. Proceedings., 1994 IEEE, 1994, pp. 1343-1352 vol.3.
[4] B. Hua, Z. Zhengming, and C. Mi, "Framework and Research Methodology of Short-
Timescale Pulsed Power Phenomena in High-Voltage and High-Power Converters,"
Industrial Electronics, IEEE Transactions on, vol. 56, pp. 805-816, 2009.
[5] C. Kauczor and N. Frohleke, "Inverter topologies for ultrasonic piezoelectric transducers with
high mechanical Q-factor," in Power Electronics Specialists Conference, 2004. PESC 04.
2004 IEEE 35th Annual, 2004, pp. 2736-2741 Vol.4.
[6] J. Rodriguez, S. Bernet, W. Bin, J. O. Pontt, and S. Kouro, "Multilevel Voltage-Source-
Converter Topologies for Industrial Medium-Voltage Drives," Industrial Electronics, IEEE
Transactions on, vol. 54, pp. 2930-2945, 2007.
[7] K. Shenai, P. G. Neudeck, and G. Schwarze, "Design and technology of compact high-power
converters," Aerospace and Electronic Systems Magazine, IEEE, vol. 16, pp. 27-31, 2001.
[8] H. Akiyama, "Pulsed power in Japan," in Pulsed Power Conference, 1995. Digest of
Technical Papers., Tenth IEEE International, 1995, pp. 13-16 vol.1.
[9] H. Akiyama, S. Sakai, T. Sakugawa, and T. Namihira, "Invited Paper - Environmental
Applications of Repetitive Pulsed Power," Dielectrics and Electrical Insulation, IEEE
Transactions on, vol. 14, pp. 825-833, 2007.
[10] H. Akiyama, T. Sakugawa, T. Namihira, K. Takaki, Y. Minamitani, and N. Shimomura,
"Industrial Applications of Pulsed Power Technology," Dielectrics and Electrical Insulation,
IEEE Transactions on, vol. 14, pp. 1051-1064, 2007.
[11] R. Brandenburg, H. Lange, T. von Woedtke, M. Stieber, E. Kindel, J. Ehlbeck, et al.,
"Antimicrobial Effects of UV and VUV Radiation of Nonthermal Plasma Jets," Plasma
Science, IEEE Transactions on, vol. 37, pp. 877-883, 2009.
[12] J. Foster, B. S. Sommers, S. N. Gucker, I. M. Blankson, and G. Adamovsky, "Perspectives on
the Interaction of Plasmas With Liquid Water for Water Purification," Plasma Science, IEEE
Transactions on, vol. 40, pp. 1311-1323, 2012.
[13] F. Fukawa, N. Shimomura, T. Yano, S. Yamanaka, K. Teranishi, and H. Akiyama,
"Application of Nanosecond Pulsed Power to Ozone Production by Streamer Corona,"
Plasma Science, IEEE Transactions on, vol. 36, pp. 2592-2597, 2008.
[14] A. Golberg, J. Kandel, M. Belkin, and B. Rubinsky, "Intermittently Delivered Pulsed Electric
Fields for Sterile Storage of Turbid Media," Plasma Science, IEEE Transactions on, vol. 38,
pp. 3211-3218, 2010.
[15] W. Hartmann, T. Hammer, T. Kishimoto, M. Romheld, and A. Safitri, "Ozone Generation in
a Wire-Plate Pulsed Corona Plasma Reactor," in Pulsed Power Conference, 2005 IEEE, 2005,
pp. 856-859.
[16] T. Heeren, J. F. Kolb, S. Xiao, K. H. Schoenbach, and H. Akiyama, "Pulsed Power
Generators and Delivery Devices for Bioelectrical Applications," in Power Modulator
Symposium, 2006. Conference Record of the 2006 Twenty-Seventh International, 2006, pp.
486-489.
Chapter 1
107
[17] Y. A. Kotov and S. Y. Sokovnin, "Overview of the application of nanosecond electron beams
for radiochemical sterilization," Plasma Science, IEEE Transactions on, vol. 28, pp. 133-136,
2000.
[18] Z. Liu, "Multiple-switch pulsed power generation based on a transmission line transformer,"
Doctoral degree Department of Electrical Engineering;, Eindhoven : Technische Universiteit
Eindhoven, 2008.
[19] T. Matsumoto, D. Wang, T. Namihira, and H. Akiyama, "Energy Efficiency Improvement of
Nitric Oxide Treatment Using Nanosecond Pulsed Discharge," Plasma Science, IEEE
Transactions on, vol. 38, pp. 2639-2643, 2010.
[20] Y. Matsumoto, N. Shioji, T. Satake, and A. Sakuma, "Inactivation of microorganisms by
pulsed high voltage application," in Industry Applications Society Annual Meeting, 1991.,
Conference Record of the 1991 IEEE, 1991, pp. 652-659 vol.1.
[21] A. Mizuno and Y. Hori, "Destruction of living cells by pulsed high-voltage application,"
Industry Applications, IEEE Transactions on, vol. 24, pp. 387-394, 1988.
[22] R. Narsetti, R. D. Curry, K. F. McDonald, L. M. Nichols, and T. Clevenger, "Application of
Pulsed Electric Fields and Magnetic Pulse Compressor Technology for Water Sterilization,"
in Pulsed Power Conference, 2005 IEEE, 2005, pp. 1282-1285.
[23] A. Pokryvailo, M. Wolf, Y. Yankelevich, S. Wald, L. R. Grabowski, E. M. van Veldhuizen, et
al., "High-Power Pulsed Corona for Treatment of Pollutants in Heterogeneous Media,"
Plasma Science, IEEE Transactions on, vol. 34, pp. 1731-1743, 2006.
[24] B. S. Rajanikanth and A. D. Srinivasan, "Pulsed plasma promoted adsorption/catalysis for
NOx removal from stationary diesel engine exhaust," Dielectrics and Electrical Insulation,
IEEE Transactions on, vol. 14, pp. 302-311, 2007.
[25] J. O. Rossi, E. Schamiloglu, and M. Ueda, "Advances in High-Voltage Modulators for
Applications in Pulsed Power and Plasma-Based Ion Implantation," Plasma Science, IEEE
Transactions on, vol. 39, pp. 3033-3044, 2011.
[26] Z. Ruobing, W. Liming, W. Yan, G. Zhicheng, and J. Zhidong, "Bacterial Decontamination
of Water by Bipolar Pulsed Discharge in a Gas-Liquid-Solid Three-Phase Discharge
Reactor," Plasma Science, IEEE Transactions on, vol. 34, pp. 1370-1374, 2006.
[27] T. Sakugawa, T. Yamaguchi, K. Yamamoto, T. Kiyan, T. Namihira, S. Katsuki, et al., "All
Solid State Pulsed Power System for Water Discharge," in Pulsed Power Conference, 2005
IEEE, 2005, pp. 1057-1060.
[28] M. Sato, N. M. Ishida, A. T. Sugiarto, T. Ohshima, and H. Taniguchi, "High-efficiency
sterilizer by high-voltage pulse using concentrated-field electrode system," Industry
Applications, IEEE Transactions on, vol. 37, pp. 1646-1650, 2001.
[29] T. Sato, O. Furuya, and T. Nakatani, "Characteristics of Nonequilibrium Plasma Flow and Its
Sterilization Efficacy in a Tube at Atmospheric Pressure," Industry Applications, IEEE
Transactions on, vol. 45, pp. 44-49, 2009.
[30] E. Schamiloglu, R. J. Barker, M. Gundersen, and A. A. Neuber, "Modern Pulsed Power:
Charlie Martin and Beyond," Proceedings of the IEEE, vol. 92, pp. 1014-1020, 2004.
[31] T. Sheng-Yu, W. Tsai-Fu, and C. Yaow-Ming, "Wide Pulse Combined With Narrow-Pulse
Generator for Food Sterilization," Industrial Electronics, IEEE Transactions on, vol. 55, pp.
741-748, 2008.
[32] K. Shimizu, M. Blajan, and S. Tatematsu, "Basic Study of Remote Disinfection and
Sterilization Effect by Using Atmospheric Microplasma," Industry Applications, IEEE
Transactions on, vol. 48, pp. 1182-1188, 2012.
Chapter 1
108
[33] N. Shimomura, K. Nakano, H. Nakajima, T. Kageyama, K. Teranishi, and H. Akiyama,
"Consideration of reactors configuration for NOx treatment by nanosecond pulsed power," in
Power Modulator and High Voltage Conference (IPMHVC), 2010 IEEE International, 2010,
pp. 693-696.
[34] N. Shimomura, K. Nakano, H. Nakajima, T. Kageyama, K. Teranishi, F. Fukawa, et al.,
"Nanosecond pulsed power application to nitrogen oxides treatment with coaxial reactors,"
Dielectrics and Electrical Insulation, IEEE Transactions on, vol. 18, pp. 1274-1280, 2011.
[35] R. E. J. Sladek, "Plasma needle: non-thermal atmospheric plasmas in dentistry," PhD thesis,
Eindhoven University of Technology,, The Netherlands, 2006.
[36] A. D. Srinivasan and B. S. Rajanikanth, "Nonthermal-Plasma-Promoted Catalysis for the
Removal of NO<sub><span class="roman">x</span></sub> From a Stationary Diesel-
Engine Exhaust," Industry Applications, IEEE Transactions on, vol. 43, pp. 1507-1514, 2007.
[37] A. D. Srinivasan and B. S. Rajanikanth, "Pulsed Plasma Treatment for NOx Reduction from
Filtered/Unfiltered Stationary Diesel Engine Exhaust," in Industry Applications Conference,
2007. 42nd IAS Annual Meeting. Conference Record of the 2007 IEEE, 2007, pp. 1893-1900.
[38] T. Suzuki, Y. Minamitani, and T. Nose, "Investigation of a pulse circuit design and pulse
condition for the high energy efficiency on water treatment using pulsed power discharge in a
water droplet spray," Dielectrics and Electrical Insulation, IEEE Transactions on, vol. 18, pp.
1281-1286, 2011.
[39] K. Takaki, M. A. Jani, and T. Fujiwara, "Removal of nitric oxide in flue gases by multi-point
to plane dielectric barrier discharge," Plasma Science, IEEE Transactions on, vol. 27, pp.
1137-1145, 1999.
[40] S. Tao, Z. Dongdong, Y. Yang, Z. Cheng, W. Jue, Y. Ping, et al., "A Compact Repetitive
Unipolar Nanosecond-Pulse Generator for Dielectric Barrier Discharge Application," Plasma
Science, IEEE Transactions on, vol. 38, pp. 1651-1655, 2010.
[41] S. Tao, Y. Yang, Z. Cheng, N. Zheng, Z. Dongdong, W. Jue, et al., "Nanosecond-pulse
dielectric barrier discharge using magnetic compression solid-state pulsed power," in Power
Modulator and High Voltage Conference (IPMHVC), 2010 IEEE International, 2010, pp.
481-484.
[42] T. Yamamoto, B. S. Rajanikanth, M. Okubo, T. Kuroki, and M. Nishino, "Performance
evaluation of nonthermal plasma reactors for NO oxidation in diesel engine exhaust gas
treatment," Industry Applications, IEEE Transactions on, vol. 39, pp. 1608-1613, 2003.
[43] K. Yatsui, "PROGRESS OF PULSED POWER COMMERCIAL APPLICATIONS IN
JAPAN."
[44] H. Bluhm, Pulsed Power System: Principle and Applications. Berlin Heidelberg: Springer-
Verlag, 2006.
[45] G. A. Mesyats, Pulsed Power. New York: Kluwer Academic, 2005.
[46] G. A. Mesyats, "Methods of Pulsed Voltage Multiplication," Prib, Tekh. Exp, vol. 6, pp. 95-
97, 1963.
[47] C. Jaegu, T. Yamaguchi, K. Yamamoto, T. Namihira, T. Sakugawa, S. Katsuki, et al.,
"Feasibility Studies of EMTP Simulation for the Design of the Pulsed-Power Generator Using
MPC and BPFN for Water Treatments," Plasma Science, IEEE Transactions on, vol. 34, pp.
1744-1750, 2006.
[48] D. D. P. Kumar, K. S. S. Mitra, A. Sharma, K. V. Nagesh, S. K. Singh, A. Roy, et al.,
"Characterization and analysis of a pulse power system based on Marx generator and
Blumlein," Review of Scientific Instruments, vol. 78, 2007.
[49] ABB, "High Power Semiconductors Short Form Catalogue," 2011.
Chapter 1
109
[50] ABB, "High Power Semiconductors Short Form Catalogue," 2003.
[51] ABB, "Asymmetric Integrated Gate Commutated Thyristor," ed, Jun. 2010.
[52] C. Steinem and A. Janshoff, Piezoelectric Sensors: Springer, 2007.
[53] L. Redondo and J. F. Silva, "Solid State Pulsed Power Electronics," in POWER
ELECTRONICS HANDBOOK DEVICES, CIRCUITS,ANDAPPLICATIONS, M. H.Rashid,
Ed., Third Edition ed: Elsevier, 2011.
[54] R. Diez, H. Piquet, M. Cousineau, and S. Bhosle, "Current-Mode Power Converter for
Radiation Control in DBD Excimer Lamps," Industrial Electronics, IEEE Transactions on,
vol. 59, pp. 1912-1919, 2012.
[55] V. H. Olivares, M. Ponce, J. A. Gutierrez, A. Tellez, and D. Mora, "Short Pulsed Voltage
Dielectric Barrier Discharge on Tubular Fluorescent Lamps," in Electronics, Robotics and
Automotive Mechanics Conference, 2009. CERMA '09., 2009, pp. 417-422.
[56] U. N. Pal, P. Gulati, N. Kumar, M. Kumar, M. S. Tyagi, B. L. Meena, et al., "Analysis of
Discharge Parameters in Xenon-Filled Coaxial DBD Tube," Plasma Science, IEEE
Transactions on, vol. 39, pp. 1475-1481, 2011.
[57] B. Ju Won, Y. Dong-Wook, and K. Heung-Geun, "High-voltage switch using series-
connected IGBTs with simple auxiliary circuit," Industry Applications, IEEE Transactions
on, vol. 37, pp. 1832-1839, 2001.
[58] R. A. Fitch and V. T. S. Howell, "Novel principle of transient high-voltage generation,"
Proceedings of the Institution of Electrical Engineers, Science and General, vol. 111, pp.
849-855, 1964.
[59] D. E. BLISS and B. ESPINOZA. (2010) Automated alignment keeps Z machine on target.
Laser Focus World.
[60] S. Zabihi, F. Zare, G. Ledwich, A. Ghosh, and H. Akiyama, "A new family of marx
generators based on commutation circuits," Dielectrics and Electrical Insulation, IEEE
Transactions on, vol. 18, pp. 1181-1188, 2011.
[61] H. Kirbie and G. E. Dale, "Diode-directed solid-state marx generator " US Patent, 2006.
[62] H. Canacsinh, L. M. Redondo, and J. F. Silva, "New solid-state Marx topology for bipolar
repetitive high-voltage pulses," in Power Electronics Specialists Conference, 2008. PESC
2008. IEEE, 2008, pp. 791-795.
[63] T. Sakamoto, A. Nami, M. Akiyama, and H. Akiyama, "A Repetitive Solid State Marx-Type
Pulsed Power Generator Using Multistage Switch-Capacitor Cells," Plasma Science, IEEE
Transactions on, vol. 40, pp. 2316-2321, 2012.
[64] L. M. Redondo and H. Canacsinh, "Bipolar solid state arbitrary-waveform Marx generator for
capacitive loads," in Pulsed Power Conference (PPC), 2011 IEEE, 2011, pp. 598-601.
[65] M. Lehmann, "High Energy Output Marx Generator Design," in Power Modulator and High
Voltage Conference (IPMHVC), 2010 IEEE International, 2010, pp. 576-578.
[66] G. Jingming, L. Yonggui, L. Jinliang, Y. Jianhua, and Z. Jiande, "Development of a
Repetitive Wave Erection Marx Generator," Plasma Science, IEEE Transactions on, vol. 37,
pp. 1936-1942, 2009.
[67] L. Hongtao, R. Hong-Je, K. Jong-Soo, R. Geun-Hie, K. Young-Bae, and D. Jianjun,
"Development of Rectangle-Pulse Marx Generator Based on PFN," Plasma Science, IEEE
Transactions on, vol. 37, pp. 190-194, 2009.
[68] H. Heo, S. S. Park, S. C. Kim, J. H. Seo, S. H. Kim, O. R. Choi, et al., "Performance of the
Marx generator for repetitive applications," in IEEE International Power Modulators and
High Voltage Conference, Proceedings of the 2008, 2008, pp. 526-528.
Chapter 1
110
[69] W. Yifan, L. Kefu, Q. Jian, L. Xiaoxu, and X. Houxiu, "Repetitive and High Voltage Marx
Generator Using Solid-state Devices," Dielectrics and Electrical Insulation, IEEE
Transactions on, vol. 14, pp. 937-940, 2007.
[70] S. C. Glidden and H. D. Sanders, "Solid State Marx Generator," in Power Modulator
Symposium, 2006. Conference Record of the 2006 Twenty-Seventh International, 2006, pp.
314-317.
[71] J. W. Baek, D. W. Yoo, G. H. Rim, and J. S. Lai, "Solid State Marx Generator Using Series-
Connected IGBTs," Plasma Science, IEEE Transactions on, vol. 33, pp. 1198-1204, 2005.
[72] W. Jiang, K. Yatsui, N. Shimizu, K. Iida, and A. Tokuchi, "Compact pulsed power generators
for industrial applications," in Pulsed Power Conference, 2003. Digest of Technical Papers.
PPC-2003. 14th IEEE International, 2003, pp. 261-264 Vol.1.
[73] "Fast high voltage transistor switches," Behlke Electronic GmbH, Ed., ed. Kronberg,
Germany, 901-10-lc2.
[74] D. O. Neacsu, Power Switching Converters: Medium and High Power. Boca Raton, FL:
Tayler and Francis Group, CRC Press, 2006.
[75] M. Giesselmann and T. Heeren, "Rapid capacitor charger," in Power Modulator Symposium,
2002 and 2002 High-Voltage Workshop. Conference Record of the Twenty-Fifth
International, 2002, pp. 146-149.
[76] L. M. Redondo, E. Margato, and J. Fernando Silva, "A new method to build a high-voltage
pulse supply using only semiconductor switches for plasma-immersion ion implantation,"
Surface and Coatings Technology, vol. 136, pp. 51-54, 2/2/ 2001.
[77] X. Tian, X. Wang, B. Tang, P. K. Chu, P. K. Ko, and Y.-C. Cheng, "Special modulator for
high frequency, low-voltage plasma immersion ion implantation," Review of Scientific
Instruments, vol. 70, pp. 1824-1828, 1999.
[78] L. Ching-Shan, "A forward configuration for high-voltage applications: multiple-switch
integrated filter forward converter (MSIFFC)," in Applied Power Electronics Conference and
Exposition, 1998. APEC '98. Conference Proceedings 1998., Thirteenth Annual, 1998, pp.
609-613 vol.2.
[79] Prudi, x, M. k, and P. Vorel, "Advantages of using two-switch forward converter for high-
voltage applications," in Power Electronics, Electrical Drives, Automation and Motion
(SPEEDAM), 2012 International Symposium on, 2012, pp. 326-330.
[80] G. Spiazzi, P. Mattavelli, and A. Costabeber, "High Step-Up Ratio Flyback Converter With
Active Clamp and Voltage Multiplier," Power Electronics, IEEE Transactions on, vol. 26, pp.
3205-3214, 2011.
[81] T. Sheng-Yu, H. Guan-Wei, J. Yi-Ren, F. Shu-Yuan, and C. Geeng-Kwei, "High voltage
generator using rapid response boost / flyback converters for stun gun applications," in
Industrial Electronics and Applications (ICIEA), 2011 6th IEEE Conference on, 2011, pp.
1780-1785.
[82] L. M. Redondo and J. F. Silva, "Flyback Versus Forward Switching Power Supply
Topologies For Unipolar Pulsed-Power Applications," Plasma Science, IEEE Transactions
on, vol. 37, pp. 171-178, 2009.
[83] T. E. Salem, C. W. Tipton, and D. Porschet, "Fabrication and Practical Considerations of a
Flyback Transformer for Use in High Pulsed-Power Applications," in System Theory, 2006.
SSST '06. Proceeding of the Thirty-Eighth Southeastern Symposium on, 2006, pp. 406-409.
[84] F. Wang, A. Kuthi, C. Jiang, Q. Zhou, and M. Gundersen, "Flyback resonant charger for high
repetition rate pseudospark pulse generator," in Power Modulator Symposium, 2004 and 2004
High-Voltage Workshop. Conference Record of the Twenty-Sixth International, 2004, pp. 85-
88.
Chapter 1
111
[85] A. M. Rahimi, F. Rahimi, and I. Hassanzadeh, "Analysis of high-voltage flyback converter in
color TVs, and its regulation," in Power Electronics and Drive Systems, 2003. PEDS 2003.
The Fifth International Conference on, 2003, pp. 353-358 Vol.1.
[86] S. K. Chung, "Transient characteristics of high-voltage flyback transformer operating in
discontinuous conduction mode," Electric Power Applications, IEE Proceedings -, vol. 151,
pp. 628-634, 2004.
[87] T. Namihira, T. Yamaguchi, K. Yamamoto, J. Choi, T. Kiyan, T. Sakugawa, et al.,
"Characteristics of Pulsed Discharge Plasma in Water," in Pulsed Power Conference, 2005
IEEE, 2005, pp. 1013-1016.
[88] G. S. Daehn, Metalworking: Sheet Forming vol. 14B. USA: ASM International, 2006.
[89] D. E. Anderson, "Recent Developments in Pulsed High Power Systems," in Proceedings of
LINAC 2006 conf., Knoxville, USA, 2006.
[90] D. Weidong, R. Hang, Z. Qiaogen, and Y. Lanjun, "Repetitive Frequency Marx Generator
Based on Magnetic Switches and Its Application in Dielectric Barrier Discharge," Plasma
Science, IEEE Transactions on, vol. 40, pp. 2373-2378, 2012.
[91] H. Ghomi, N. N. Safa, and S. Ghasemi, "Investigation on a DBD Plasma Reactor," Plasma
Science, IEEE Transactions on, vol. 39, pp. 2104-2105, 2011.
[92] Q. G. Zhang, F. B. Tao, Z. Li, W. D. Ding, and A. C. Qiu, "Effect of Pulse Rise Time on the
Glow Discharge in Nonuniform Electric Field," Plasma Science, IEEE Transactions on, vol.
36, pp. 1008-1009, 2008.
[93] Z. Machala, I. Jedlovsky, and V. Martisovits, "DC Discharges in Atmospheric Air and Their
Transitions," Plasma Science, IEEE Transactions on, vol. 36, pp. 918-919, 2008.
[94] W. Douyan, M. Jikuya, S. Yoshida, T. Namihira, S. Katsuki, and H. Akiyama, "Positive- and
Negative-Pulsed Streamer Discharges Generated by a 100-ns Pulsed-Power in Atmospheric
Air," Plasma Science, IEEE Transactions on, vol. 35, pp. 1098-1103, 2007.
[95] P. Lukes, A. T. Appleton, and B. R. Locke, "Hydrogen peroxide and ozone formation in
hybrid gas-liquid electrical discharge Reactors," Industry Applications, IEEE Transactions
on, vol. 40, pp. 60-67, 2004.
[96] J. S. Chang, P. A. Lawless, and T. Yamamoto, "Corona discharge processes," Plasma
Science, IEEE Transactions on, vol. 19, pp. 1152-1166, 1991.
[97] A. H. El-Hag, S. H. Jayaram, and M. W. Griffiths, "Inactivation of Naturally Grown
Microorganisms in Orange Juice Using Pulsed Electric Fields," Plasma Science, IEEE
Transactions on, vol. 34, pp. 1412-1415, 2006.
[98] W. Heywang, K. Lubitz, and W. Wersing, Piezoelectricity. New York: Springer, 2008.
[99] L. Rongyuan, N. Frohleke, and J. Bocker, "LLCC-PWM inverter for driving high-power
piezoelectric actuators," in Power Electronics and Motion Control Conference, 2008. EPE-
PEMC 2008. 13th, 2008, pp. 159-164.
[100] S. Priya, "High power universal piezoelectric transformer," Ultrasonics, Ferroelectrics and
Frequency Control, IEEE Transactions on, vol. 53, pp. 23-29, 2006.
[101] S. Zhang, R. Xia, L. Lebrun, D. Anderson, and T. R. Shrout, "Piezoelectric materials for high
power, high temperature applications," Materials Letters, vol. 59, pp. 3471-3475, 11// 2005.
[102] R. Li, N. Frohleke, and J. Bocker, "DESIGN AND IMPLEMENTATION OF A POWER
INVERTER FOR A HIGH POOWER PEIZOELECTRIC BRAKE ACTUATOR IN
AIRCRAFTS," presented at the 9th Brazilian Power Electronics Conference, 2003.
Chapter 1
112
[103] Y. Kui, K. Uchino, X. Yuan, D. Shuxiang, and L. Leong Chew, "Compact piezoelectric
stacked actuators for high power applications," Ultrasonics, Ferroelectrics and Frequency
Control, IEEE Transactions on, vol. 47, pp. 819-825, 2000.
[104] A. Shoh, "Industrial Applications of Ultrasound - A Review I. High-Power Ultrasound,"
Sonics and Ultrasonics, IEEE Transactions on, vol. 22, pp. 60-70, 1975.
[105] T. S. O. Märtens, M. Min, R. Land, M. Reidla, "Fast Impedance Spectroscopy of
Piezosensors for Structural Health Monitoring," ELECTRONICS AND ELECTRICAL
ENGINEERING, vol. 7, pp. 31-34, 2010.
[106] T. Li, X. Zhang, C. Jiang, and L. Hou, "Analysis of the characteristics of piezoelectric sensor
and research of its application," in Applications of Ferroelectrics, 2009. ISAF 2009. 18th
IEEE International Symposium on the, 2009, pp. 1-4.
[107] K. Jina, B. L. Grisso, J. K. Kim, H. Dong Sam, and D. J. Inman, "Electrical modeling of
Piezoelectric ceramics for analysis and evaluation of sensory systems," in Sensors
Applications Symposium, 2008. SAS 2008. IEEE, 2008, pp. 122-127.
[108] J. George K Lewis, George K Lewis, Sr, and William Olbricht, "Cost-effective broad-band
electrical impedance spectroscopy measurement circuit and signal analysis for piezo-materials
and ultrasound transducers," MEASUREMENT SCIENCE AND TECHNOLOGY, vol. 19, pp.
1-13, 2008.
[109] L. W. Schmerr, A. Lopez-Sanchez, and R. Huang, "Complete ultrasonic transducer
characterization and its use for models and measurements," Ultrasonics, vol. 44, pp. e753-
e757, 2006.
[110] V. Loyau and G. Feuillard, "Relationship between electrical impedance of a transducer and its
electroacoustic behavior: Measurement without primary source," Journal of Applied Physics,
vol. 100, pp. 034909-034909-7, 2006.
[111] A. L. Lopez-Sanchez and L. W. Schmerr, "Determination of an ultrasonic transducer's
sensitivity and impedance in a pulse-echo setup," Ultrasonics, Ferroelectrics and Frequency
Control, IEEE Transactions on, vol. 53, pp. 2101-2112, 2006.
[112] K. Kin Wing, H. L. W. Chan, and C. L. Choy, "Evaluation of the material parameters of
piezoelectric materials by various methods," Ultrasonics, Ferroelectrics and Frequency
Control, IEEE Transactions on, vol. 44, pp. 733-742, 1997.
[113] J. S. Kim, K. Choi, and I. Yu, "A new method of determining the equivalent circuit
parameters of piezoelectric resonators and analysis of the piezoelectric loading effect,"
Ultrasonics, Ferroelectrics and Frequency Control, IEEE Transactions on, vol. 40, pp. 424-
426, 1993.
[114] T. Saar, O. Martens, M. Reidla, and A. Ronk, "Chirp-based impedance spectroscopy of piezo-
sensors," presented at the Electronics Conference (BEC), 2010.
[115] G. Mingjie and L. Wei-Hsin, "Studies on the circuit models of piezoelectric ceramics,"
presented at the Proc. Int. Conf. Information Acquisition, 2004.
[116] D. Huijuan, W. Jian, Z. Hui, and Z. Guangyu, "Measurement of a piezoelectric transducer's
mechanical resonant frequency based on residual vibration signals," presented at the
Information and Automation (ICIA), 2010 IEEE International Conference on, 2010.
[117] I. Getman and S. Lopatin, "Matching of series and parallel resonance frequencies for
ultrasonic piezoelectric transducers," presented at the Proc. Int. Symp. Applications of
Ferroelectrics.
[118] Y. Y. Lim, S. Bhalla, and C. K. Soh, "Structural identification and damage diagnosis using
self-sensing piezo-impedance transducers," Smart Materials and Structures, vol. 15, pp. 987-
995, 2006.
Chapter 1
113
[119] E. Dallago and A. Danioni, "Resonance frequency tracking control for piezoelectric
transformer," Electronics Letters, vol. 37, pp. 1317-1318, 2001.
[120] P. Davari and H. Hassanpour, "Designing a new robust on-line secondary path modeling
technique for feedforward active noise control systems," Signal Processing, vol. 89, pp. 1195-
1204, 2009.
[121] R. Pintelon and J. Schoukens, "Basic Choices in System Identification," in System
Identification: A Frequency Domain Approach, ed: IEEE Press and John Wiley, 2001, pp.
351-375.
[122] "IEEE Standard on Piezoelectricity," ANSI/IEEE Std 176-1987, p. 0_1, 1988.
[123] K. Agbossou, J. L. Dion, S. Carignan, M. Abdelkrim, and A. Cheriti, "Class D amplifier for a
power piezoelectric load," Ultrasonics, Ferroelectrics and Frequency Control, IEEE
Transactions on, vol. 47, pp. 1036-1041, 2000.
[124] C. Volosencu, "Methods for Parameter Estimation and Frequency Control of Piezoelectric
Transducers, Automation Control - Theory and Practice," in Automation Control - Theory and
Practice, A. D. Rodić, Ed., ed: InTech, 2009, pp. 115-136.
[125] T. Sai Chun and G. T. Clement, "A harmonic cancellation technique for an ultrasound
transducer excited by a switched-mode power converter," Ultrasonics, Ferroelectrics and
Frequency Control, IEEE Transactions on, vol. 55, pp. 359-367, 2008.
[126] C. Hung Liang, C. Chun An, F. Chun Chieh, and Y. Hau Chen, "Single-switch high power
factor inverter for driving piezoelectric ceramic transducer," in Power Electronics and Drive
Systems, 2009. PEDS 2009. International Conference on, 2009, pp. 1571-1576.
[127] A. Nami and F. Zare, "Multilevel Converters in Renewable Energy Systems," in Renewable
Energy, T. J. Hammons, Ed., ed: InTech, 2009, pp. 271-296.
[128] F. Zare and G. Ledwich, "A New Predictive Current Control Technique for Multilevel
Converters," presented at the TENCON IEEE Region 10 Conference, 2006.
[129] F. Zare and G. Ledwich, "A hysteresis current control for single-phase multilevel voltage
source inverters: PLD implementation," Power Electronics, IEEE Transactions on, vol. 17,
pp. 731-738, 2002.
[130] J. Rodriguez, L. Jih-Sheng, and P. Fang Zheng, "Multilevel inverters: a survey of topologies,
controls, and applications," Industrial Electronics, IEEE Transactions on, vol. 49, pp. 724-
738, 2002.
[131] K. Uchino, Piezoelectric Actuators/Ultrasonic Motors. Boston: Kluwer, 1996.
[132] J. G. Smits and S. I. Dalke, "The constituent equations of piezoelectric bimorphs," in
Ultrasonics Symposium, 1989. Proceedings., IEEE 1989, 1989, pp. 781-784 vol.2.
[133] E. G. Thurston, "The Theoretical Sensitivity of Three Types of Rectangular Bimorph
Transducers," The Journal of the Acoustical Society of America, vol. 25, pp. 870-872, 09/00/
1953.
[134] A. Dogan, K. Uchino, and R. E. Newnham, "Composite piezoelectric transducer with
truncated conical endcaps "cymbal"," Ultrasonics, Ferroelectrics and Frequency Control,
IEEE Transactions on, vol. 44, pp. 597-605, 1997.
[135] Y. Kui, Z. Weiguang, E. Uchino, Z. Zhe, and L. Leong Chew, "Design and fabrication of a
high performance multilayer piezoelectric actuator with bending deformation," Ultrasonics,
Ferroelectrics and Frequency Control, IEEE Transactions on, vol. 46, pp. 1020-1027, 1999.
[136] S. Zabihi, F. Zare, G. Ledwich, A. Ghosh, and H. Akiyama, "A Novel High-Voltage Pulsed-
Power Supply Based on Low-Voltage Switch-Capacitor Units," Plasma Science, IEEE
Transactions on, vol. 38, pp. 2877-2887, 2010.
Chapter 1
114
[137] S. Zabihi, F. Zare, G. Ledwich, A. Ghosh, and H. Akiyama, "A new pulsed power supply
topology based on positive buck-boost converters concept," Dielectrics and Electrical
Insulation, IEEE Transactions on, vol. 17, pp. 1901-1911, 2010.
[138] N. Coruh, S. Urgun, and T. Erfidan, "Design and implementation of flyback converters," in
Industrial Electronics and Applications (ICIEA), 2010 the 5th IEEE Conference on, 2010, pp.
1189-1193.
[139] P. Davari, F. Zare, and A. Ghosh, "Parallel and series configurations of flyback converter for
pulsed power applications," in Industrial Electronics and Applications (ICIEA), 2012 7th
IEEE Conference on, 2012, pp. 1517-1522.
[140] P. Davari, F. Zare, A. Ghosh, and H. Akiyama, "High-Voltage Modular Power Supply Using
Parallel and Series Configurations of Flyback Converter for Pulsed Power Applications,"
Plasma Science, IEEE Transactions on, vol. 40, pp. 2578-2587, 2012.
[141] S. Honggang, S. Wei, W. Hongfang, F. Dianbo, P. Yunqing, Y. Xu, et al., "Design and
Implementation of a High Power Density Three-Level Parallel Resonant Converter for
Capacitor Charging Pulsed-Power Supply," Plasma Science, IEEE Transactions on, vol. 39,
pp. 1131-1140, 2011.
[142] A. A. Boora, F. Zare, G. Ledwich, and A. Ghosh, "A general approach to control a Positive
Buck-Boost converter to achieve robustness against input voltage fluctuations and load
changes," in Power Electronics Specialists Conference, 2008. PESC 2008. IEEE, 2008, pp.
2011-2017.
[143] P. Davari, F. Zare, and A. Ghosh, "Flexible Solid-State Pulsed Power Topology," presented at
the 15th International Power Electronics and Motion Control Conference, EPE-PEMC 2012
ECCE Europe, Novi Sad, Serbia, Sep 2012.
[144] N. Osawa and Y. Yoshioka, "Generation of low-frequency homogeneous dielectric barrier
discharge at atmospheric pressure," Plasma Science, IEEE Transactions on, vol. 40, pp. 2-8,
2012.
[145] H. Piquet, S. Bhosle, R. Diez, and M. V. Erofeev, "Pulsed Current-Mode Supply of Dielectric
Barrier Discharge Excilamps for the Control of the Radiated Ultraviolet Power," Plasma
Science, IEEE Transactions on, vol. 38, pp. 2531-2538, 2010.
[146] A. A. El-Deib, F. Dawson, S. Bhosle, and G. Zissis, "Circuit-Based Model for a Dielectric
Barrier Discharge Lamp Using the Finite Volume Method," Plasma Science, IEEE
Transactions on, vol. 38, pp. 2260-2273, 2010.
[147] B. Rahmani, S. Bhosle, and G. Zissis, "Dielectric-Barrier-Discharge Excilamp in Mixtures of
Krypton and Molecular Chlorine," Plasma Science, IEEE Transactions on, vol. 37, pp. 546-
550, 2009.
[148] H. Ayan, G. Fridman, A. F. Gutsol, V. N. Vasilets, A. Fridman, and G. Friedman,
"Nanosecond-Pulsed Uniform Dielectric-Barrier Discharge," Plasma Science, IEEE
Transactions on, vol. 36, pp. 504-508, 2008.
[149] K. Takaki, M. Shimizu, S. Mukaigawa, and T. Fujiwara, "Effect of electrode shape in
dielectric barrier discharge plasma reactor for NOx removal," Plasma Science, IEEE
Transactions on, vol. 32, pp. 32-38, 2004.
[150] B. S. Rajanikanth and B. R. Sushma, "Injection of N-radicals into diesel engine exhaust
treated by plasma," Plasma Science and Technology, vol. 8, pp. 202-206, 2006.
[151] V. Ravi, Y. S. Mok, B. S. Rajanikanth, and H. C. Kang, "Studies on nitrogen oxides removal
using plasma assisted catalytic reactor," Plasma Science and Technology, vol. 5, pp. 2057-
2062, 2003.
Chapter 1
115
[152] V. Ravi, Y. S. Mok, B. S. Rajanikanth, and H. C. Kang, "Temperature effect on hydrocarbon-
enhanced nitric oxide conversion using a dielectric barrier discharge reactor," Fuel
Processing Technology, vol. 81, pp. 187-199, 2003.
[153] B. S. Rajanikanth and V. Ravi, "Removal of nitrogen oxides in diesel engine exhaust by
plasma assisted molecular sieves," Plasma Science and Technology, vol. 4, pp. 1399-1406,
2002.
[154] T. Yamamoto, M. Okubo, K. Hayakawa, and K. Kitaura, "Towards ideal NO x control
technology using plasma-chemical hybrid process," Phoenix, AZ, USA, 1999, pp. 1495-1502.
[155] A. D. Srinivasan, B. S. Rajanikanth, and S. Mahapatro, "CORONA TREATMENT FOR
NOX REDUCTION FROM STATIONARY DIESEL ENGINE EXHAUST IMPACT OF
NATURE OF ENERGIZATION AND EXHAUST COMPOSITION," 2009.
[156] B. S. Rajanikanth and V. Ravi, "DeNOx study in diesel engine exhaust using barrier
discharge corona assisted by v2O5/TiO2 catalyst," Plasma Science and Technology, vol. 6,
pp. 2411-2415, 2004.
[157] B. S. Rajanikanth, S. Das, and A. D. Srinivasan, "Unfiltered Diesel Engine Exhaust
Treatment by Discharge Plasma: Effect of Soot Oxidation," Plasma Science & Technology,
vol. 6, pp. 2475-2480, 2004.
[158] B. S. Rajanikanth, A. D. Srinivasan, and B. A. Nandiny, "A cascaded discharge plasma-
adsorbent technique for engine exhaust treatment," Plasma Science and Technology, vol. 5,
pp. 1825-1833, 2003.
[159] B. S. Rajanikanth, D. Sinha, and P. Emmanuel, "Discharge plasma assisted adsorbents for
exhaust treatment: A comparative analysis on enhancing NOx removal," Plasma Science and
Technology, vol. 10, pp. 307-312, 2008.
[160] B. S. Rajanikanth, S. Mohapatro, and L. Umanand, "Solar powered high voltage energization
for vehicular exhaust cleaning: A step towards possible retrofitting in vehicles," Fuel
Processing Technology, vol. 90, pp. 343-352, 2009.
[161] R. O. McClellan, "Health Effects of Exposure to Diesel Exhaust Particles," Ann. Rev.
Pharmacol. Toxicol, vol. 27, pp. 279-300, 1989.
[162] J. H. Seinfeld, Air pollution: Physical and chemical fundamentals. New York: McGraw-Hill,
Inc., 1975.
[163] D. B. Kittelson, "Engines and nanoparticles: A review," J. Aerosol Sci., vol. 29, pp. 575-588,
1998.
[164] L. Zhua, J. Yu, and X. Wang, "Oxidation treatment of diesel soot particulate on CexZr1−xO2,
," J. Hazard. Mater, vol. 140, pp. 205-210, 2007.
[165] Oberdörster G, Sharp Z, Atudorei V, Elder A, Gelein R, Kreyling W, et al., "Translocation of
inhaled ultrafine particles to the brain," Inhalation Toxicology, vol. 16, pp. 437–45, 2004.
[166] R. P. Mildren and R. J. Carman, "Enhanced performance of a dielectric barrier discharge
lamp using shortpulsed excitation," J. Phys. D.: Appl. Phys., vol. 34, pp. L1-L6, Jan. 2001.
[167] J. Kriegseis, S. Grundmann, and C. Tropea, "Power consumption, discharge capacitance and
light emission as measures for thrust production of dielectric barrier discharge plasma
actuators," Journal of Applied Physics, vol. 110, pp. 013305-9, 07/01/ 2011.
[168] N. Ghasemi, F. Zare, P. Davari, P. Weber, C. Langton, and A. Ghosh, "Power electronic
converters for high power ultrasound transducers," in Industrial Electronics and Applications
(ICIEA), 2012 7th IEEE Conference on, 2012, pp. 647-652.
[169] P. Davari, N. Ghasemi, F. Zare, P. O'Shea, and A. Ghosh, "Improving the efficiency of high
power piezoelectric transducers for industrial applications," Science, Measurement &
Technology, IET, vol. 6, pp. 213-221, 2012.
116
Statement of Contribution of Co-Authors
The authors listed below have certified that:
1- They meet the criteria for authorship in that they have participated in the conception, execution, or
interpretation, of at least that part of the publication in their field of expertise;
2- They take public responsibility for their part of the publication, except for the responsible author who
accepts overall responsibility for the publications;
3- There are no other authors of the publication according to these criteria;
4- Potential conflicts of interest have been disclosed to (a) granting bodies, (b) the editor or publisher of
journals or other publications, and (c) the head of the responsible academic unit;
5- They agree to the use of the publication in the student’s thesis and its publication on the Australasian
Digital Thesis database consistent with any limitations set by publisher requirements.
In the case of this chapter:
Parallel and Series Configurations of Flyback Converter for Pulsed Power Applications
Published in proceedings of: The 7th IEEE Conference on Industrial Electronics and Applications, ICIEA
2012, Singapore
Contributor Statement of contribution
Pooya Davari Proposed the initial design and conducted mathematical proof, simulation studies,
data analysis, and wrote the manuscript.
22 Feb 2013
Firuz Zare Proposed the initial idea and supervised the validity studies including: conducting
the simulations and writing the manuscript.
Arindam Ghosh Aided data analysis and writing the paper.
Supervisor Confirmation
I have sighted email or other correspondence from all Co-authors confirming their certifying authorship.
Prof. Arindam Ghosh
Name Signature Date: 22 Feb 2013
117
Chapter
2 . P a r a l l e l a n d S e r i e s C o n f i g u r a t i o n s o f
F l y b a c k C o n v e r t e r f o r P u l s e d P o w e r
A p p l i c a t i o n s
* School of Electrical Engineering and Computer Science, Queensland University of Technology,
GPO BOX 2434, Brisbane, Australia
Presented and Published at: The 7th
IEEE Conference on Industrial Electronics and
Applications, ICIEA 2012, Singapore, July 2012.
2
Chapter 2
118
Abstract— Advances in solid-state switches and power electronics techniques have led to
the development of compact, efficient and more reliable pulsed power systems. Although, the
power rating and operation speed of the new solid-state switches are considerably increased,
their low blocking voltage level puts a limits in the pulsed power operation. This paper
proposes the advantage of parallel and series configurations of pulsed power modules in
obtaining high voltage levels with fast rise time (dv/dt) using only conventional switches. The
proposed configuration is based on two flyback modules. The effectiveness of the proposed
approach is verified by numerical simulations, and the advantages of each configuration are
indicated in comparison with a single module.
2.1. Keywords
Pulsed power, Flyback converter, Parallel and series connection, High voltage pulse
2.2. Introduction
Rapid release of stored energy as electrical pulses into a load can result in delivery of large
amounts of instantaneous power over a short period. This strategy is called pulsed power [1].
Generally, a pulsed power system consists of three main parts: energy storage, pulse generator
and the load. The most prominent part of a pulsed power system is the pulse generator, which
is based on utilized switch and topology. Hence, the switch is the connecting device between
the storage and the load. This means that the whole system characteristics such as rise time,
repetition rate, voltage rating, efficiency, cost, life time, etc, depend on the employed switch
specifications.
Gas-state and magnetic switches have been widely used in pulsed power technology, as
they possess a very high electric strength and fast rise time [2]. Gas-state switches require
special operating conditions such as high pressure, vacuum equipments and gas supplies. In
addition, they are bulky, unreliable, have short lifetime span and low repetition rate. Even with
the magnetic switches, which have a higher repetition rate, the problems remain. These
conditions limit the mobility, efficiency and increase the cost and the size of the pulsed power
system [2, 3]. On the other hand, solid-state switches are compact, reliable, cost effective, and
have a long lifetime and repetition rate. The main drawbacks of solid-state switches are their
limited power rating and operation speed [2, 3].
Recently, significant advances in solid-state switches (both in peak power and operation
speed) and exploiting power electronics techniques and topologies have led to compact,
Chapter 2
119
efficient and more reliable pulsed power systems. The new developed solid-state switches
such as IGBT (Insulated-Gate Bipolar Transistor) and Integrated Gate-Commutated Thyristor
(IGCT) have high power rating [4], but their lower operation speed comparing with the gas-
state switches and high cost still put limits in the pulsed power supplies.
One way to increase the pulsed power supply performance and cover the switch limits is to
explore alternative circuit topologies. Varieties of circuit topologies such as Marx generators
(MG) [5], pulse forming network (PFN) [6], magnetic pulse compressors (MPC) [7] and
multistage Blumlein lines (MBL) [8] have been introduced. These topologies have been
widely used in pulsed power supplies, but complexity, inflexibility and inefficiency are their
main drawbacks [2, 9]. Interest in applying power electronics topologies and techniques to
increase power supply efficiency and reliability is growing fast. In the last decade, research
and studies designate the advantage of using DC-DC converters in pulsed power applications
[3, 9, 10]. However, if one wants to generate extra high voltages, due to components hold-on
voltage limits (such as capacitors, solid-state switches and etc), these schemes become
ineffective.
In addition to topology, another way is to combine power supplies. Variety of converter
configurations have been introduced for improving the power supply specifications such as
output ripple, high input voltage, output power ratings, etc [11, 12]. Parallel and series
connections are one of the well known combinations. Designing power supplies based on
series or parallel connection are widely employed for different applications [11, 12].
This paper proposes parallel and series connections of flyback converter modules to
develop power rating and rise time of the pulsed power supply using conventional low voltage
switches. Flyback converter is selected due to its unique properties in pulsed power as it is
discussed in the next section. The proposed approach is evaluated by numerical simulations,
and the advantages of each configuration are indicated in comparison with a single module.
The evaluated results indicate the effectiveness and efficiency of the proposed method.
2.3. Topology
The proposed approach utilized flyback topology, which is one of the well known
topologies in the power electronics [13]. Conventional flyback converter is usually preferred
as it is simple, has only one switch and magnetic component, is able to generate high voltage,
can provide multiple outputs, isolation, etc [13, 14]. But when it comes to pulsed power
Chapter 2
120
applications, in addition to the mentioned features, it has some more advantages which make it
more suitable. Main features of a flyback converter for pulsed power applications are:
The transformer, in addition to electrical isolation and energy storage also steps down
the reflected voltage across the switch; therefore lower voltage rate switches are needed
comparing with other topologies.
Fault tolerant, as the switch is in the off-state during the output pulse.
High voltage output with low input DC voltage.
As the pulsed power applications mostly have R-C characteristics [9], a current source
topology (such as flyback) is a suitable candidate.
Considering above mentioned features the current source topology is selected for the
proposed method. Here basic principle and operation modes of this converter are briefly
described.
The behaviour of a flyback converter can be realized by modelling the transformer with a
simple equivalent circuit consisting of an ideal transformer, magnetizing inductance (Lm) and
leakage inductance (Ll). Fig. 2.1 shows a flyback converter including the simple model of the
transformer connected to a load. In this figure CO is the converter equivalent output
capacitance. This capacitance can get affected in the case of an R-C load. It is to be mentioned
that, to keep the analysis simple, following assumptions are made:
1) The switch and diode are ideal (voltage drop across them is zero).
2) The switch and diode output capacitances are zero.
3) Conduction and switching losses are negligible.
4) The transformer stray capacitances are negligible.
Basically a flyback converter transfers energy from a source into the transformer
magnetizing inductance when the switch is on, and then transfers the stored energy to the load
while the switch is off. The proposed approach operates in the DCM (Discontinues
Conduction Mode). The operating modes are briefly summarized for 4 different modes
(illustrated in Fig. 2.2) below:
Mode 1: In this mode the switch is in the on-state. As Fig. 2.2 depicts, the current flows
through Lm and, during this stage, the energy is stored in the inductor. This mode lasts
Chapter 2
121
depending on the duty cycle of the PWM signal. The relationship between Lm and Vs can be
expressed as:
s
mm
s
mmsms
TD
IL
TD
ILV
t
iLV
.0.
0
(2-1)
where Vs is the dc supply voltage, Im is the maximum current, D is the duty cycle and Ts is the
switching period.
Mode 2: When the switch is turned off the magnetizing current circulates through the
primary side of the transformer and the diode in the secondary side is turned on. As the switch
is turned off the current flowing through Ll is decreased to zero, which induces voltage spike
across the switch according todt
diLv ll . This voltage may damage the switch if it exceeds
the switch break down voltage level. A snubber can provide a path for this current and damps
the spike to protect the switch [15].
Mode 3: The diode is turned on and the current stored in the magnetising inductance flows
to the secondary side. The maximum voltage across the switch at this stage is:
maxmax1
)( osswitch vn
Vv (2-2)
where, Vo is the maximum output voltage level and n is the transformer turns ratio.
Mode 4: As the converter operates in DCM in this mode, all the stored energy in Lm is
completely transferred to CO, causing the diode to get turned off.
Lm
Vs
C0 Load
1 : n
Ll
Vo
+
-
Fig. 2.1 Flyback converter circuit with transformer equivalent circuit model.
Chapter 2
122
Ll
Lm
VS
1 : n
LoadCo
VS VS
Mode 1
Mode 3 Mode 4
1 : n
Co
Mode 2
Ll
Lm
VS
Load
Lm Lm
Ll Ll
1 : n 1 : n
Co CoLoad Load
Vo
+
-
Vo
+
-
Vo
+
-
Vo
+
-
Fig. 2.2 Operating modes in a flyback converter.
2.4. Proposed Method
High voltage pulses with fast rise-time are the two main important features of a pulsed
power supply [1-3]. Fig. 2.3 illustrates the proposed method, which employs the advantage of
parallel and series configuration of the pulsed power modules to increase the voltage level and
the rise-time, while using the low-voltage switches. As the figure shows, the parallel and
series connections are only considered for the secondary side of the converter.
2.4.1. Single Module
To understand and compare the parallel and series configuration, first a single module
characteristic is described. As mentioned, a flyback converter can operate as a current source.
Therefore, circuit schematic shown in Fig. 2.1 is simplified as in Fig. 2.4 Comparing these two
figures, the estimation of the output voltage can be summarized as below:
First, as described earlier, the magnetizing inductance is charged by the DC power supply.
The charged energy stored in the inductor Lm under ideal situation can be expressed as:
2
2
1mmpri ILE
(2-3)
This stored energy will be transferred to the output capacitor when the switch is in the off-
state. The capacitive stored energy is:
Chapter 2
123
2sec
2
1ooVCE
(2-4)
If a lossless system considered, then the stored energy in the primary side, Epri, is equal to
the stored energy in the secondary side, Esec. Therefore, the output voltage can be derived
based on equations (2-3) and (2-4) as:
o
mmo
C
LIV
(2-5)
The rise time for the generated voltage is defined as dv/dt. When the switch is turned off,
the magnetizing inductance current is at its peak value (Im). Since the current through a
capacitor is proportional to the time-rate of the stored voltage, the rate of rise is:
o
m
o
o
nC
I
C
I
dt
dv
(2-6)
This equation is valid till the current through the capacitor is approximately constant;
otherwise the rate of rise completely depends on the resonant frequency.
The above equations show the effect of stored current level, magnetising inductance and
the output capacitor on the generated output voltage and the rise-time. The idea of parallel and
series configurations of the pulsed power modules is inspired by both equations (2-5) and (2-
6).
2.4.2. Parallel Modules
Considering Fig. 2.3.(a) and the fact that each transformer is acting as a current source, the
stored energy in the primary side is doubled when two modules are connected in parallel. This
stored energy will be transferred to the output capacitors; hence considering (2-3) and (2-4)
the output voltage magnitude and its rate of rise are as follows:
op
mmparallelo
C
LIV 2
(2-7)
op
o
Parallel C
I
dt
dv 2
(2-8)
Chapter 2
124
where Cop=Co1+Co2. Equations (2-7) and (2-8) indicate the importance of the output capacitor.
Even if the modules are paralleled the output capacitance can affect the output voltage level
and rise-time and keep them same as the single module. Therefore, the parallel configuration
is beneficial in pulsed power applications when Cop=Co. This happens in the case of R-C load
when the load capacitance is much bigger than the power supply output capacitance.
Ll
Lm
Vs
1 : n
Co2
Ll
Lm
Vs
n : 1
Co1Load
Vo
+
-
(a)
Ll
Lm
Vs
1 : n
Co1
Load
Ll
Lm
1 : n
Co2
Vo
+
-
(b)
Fig. 2.3 Flyback converter series and parallel connection (secondary side), a) Parallel, b) Series.
Chapter 2
125
CoIo Vo
+
-
Load
Fig. 2.4 Energy transmission in a single module flyback converter.
2.4.3. Series Modules
Second alternative is series connection of the pulsed power modules, as illustrated in Fig.
2.3(b). Here, the injected energy to the output is also doubled. But as described, the generated
voltage features also depend on the output capacitance. Taking into account the equations (2-
3) to (2-6), the output voltage magnitude and its rate of rise can be expressed as:
os
mmSerieso
C
LIV 2
(2-9)
os
o
Series C
I
dt
dv
(2-10)
where Cos is Co1Co2/(Co1+Co2). Same as in parallel modules, in series modules, the output
voltage level and rise time improve as the output capacitance decreases. Here the series
modules are beneficial in both level of voltage and its rate of rise when Cos<Co. The influence
of N modules and equivalent output capacitor (for two different cases) on the output voltage
and rise time are summarized in Tables 2.1 and 2.2.
As Fig. 2.3(b) shows, in the series connection each switch withstands its own module
reflected output voltage, while in the parallel connection, each switch should tolerate the
whole reflected output voltage.
Chapter 2
126
Table 2.1 Output voltage and rate of rise when Co1=Co2=… CoN=Co.
N Modules connected in oV
dt
dv
Parallel Singleo
o
mm V
C
LI
Singleo
o
dt
dv
C
I
Series Singleo
o
mm NV
C
LNI
Singleo
o
dt
dvN
C
IN
Table 2.2 Output voltage and rate of rise when Cop=Cos=Co.
N Modules connected in oV
dt
dv
Parallel Singleo
o
mm VN
C
LIN
Singleo
o
dt
dvN
C
IN
Series Singleo
o
mm VN
C
LIN
Singleo
o
dt
dv
C
I
This fact makes the series connected modules more appropriate for generating high voltage
waveforms using low voltage switches. Therefore, (2-2) can be rewritten as below for N-series
module:
maxmax1
)( osswitch vNn
Vv
(2-11)
2.4.4. Operating Conditions
The above mentioned calculations are correct if three important points are considered in the
designing procedure of pulsed power modules. These are:
1) The output capacitor should be completely discharged during each period; otherwise the
whole stored energy in the magnetising inductor will not transfer to the capacitor.
2) Synchronization of each switch gate signal is required. Delays between each module
gate signal reduce the performance of the system.
3) It is also important to consider the damping factor (ζ). As shown in Fig. 2.4, the entire
circuit acts as a parallel RLC circuit (the current source is an inductor). Hence, the
output voltage and the rise time are the results of the resonance effect of this circuit.
The output voltage can be expressed as:
Chapter 2
127
tstso eAeAtV 21
21)( (2-12)
where s1 and s2 are given as:
20
22,1 S
(2-13)
here, RC2
1 is known as neper frequency and LC
1
0 is the resonance frequency.
0)0(
toV and oItCI
)0( are the primary conditions. Io is the stored current in the
magnetizing inductor (Im=nIo) transferred to the secondary side. The coefficients A1 and A2 are
determined by the primary conditions as:
02
02
20
212
LIAA
(2-14)
Now depending on the damping factor coefficient can have two different values:
Underdamped ζ <1 or α < 0 , C
L
R2
1
C
LIA
2
01
(2-15)
Overdamped ζ >1 or α > 0
01 RIA
(2-16)
Equation (2-15) shows that if the circuit is in underdamped situation, then the generated
voltage amplitude is independent of the load impedance, which is the situation where the
proposed method is valid. But under overdamped condition as in (2-16), the output voltage
amplitude depends on the load impedance and the initial current. Hence, in overdamped
condition, the parallel connection has better performance.
2.5. Simulation Results
To verify the proposed method and the theoretical analysis, simulations under different
conditions were carried out. Here the performance of parallel and series flyback configurations
Chapter 2
128
(Fig. 2.3) is compared with a single flyback converter (Fig. 2.1). The software that was used
for simulation was Matlab 7.10. Table 2.3 shows the parameters value used in simulation
corresponding to Fig. 2.1. In order to make the simulation results more close to reality, the
transformer core and copper losses (Rc and Rw) are also considered.
To evaluate the proposed method in developing the features of a single module pulsed
power supply, four distinct cases are considered. Case 1 and 2 have been conducted regarding
the conditions presented in Table 2.1 and 2.2, respectively. In these cases, to simulate a plasma
phenomenon, the load gets connected to the output after the output voltage reaches to a certain
voltage level (threshold) [9]. The different parameters in each case are indicated in Table 2.4.
Like all other methods, the proposed method has its own drawbacks under certain
conditions. These conditions, mentioned as operating conditions in the previous section, are
considered in Case 3 and 4. In Case 3 the consequence of unsynchronized gate signals on the
system performance is described. Finally, in Case 4 the effect of damping factor on the
generated pulses is examined. Except in Case 4 for all the other cases a 1kΩ resistor is
selected as the load.
Table 2.3 Simulation Parameters
Vs
(v) Fs
(kHz) D
(%) Lm
(µH) Llp
(µH) Lls
(mH) Rc
(kΩ) Rw1,2
(Ω) n
10 1 8.9 160 2 4 1 0.1 10
Table 2.4 Simulation Parameters in Case 1 and 2
Parameters Value Case1 Case2
Single Parallel Series Single Parallel Series
Threshold 1000v 1000v 1978v 1000v 1410v 1410v
Co 4nF 8nF 2nF 4nF 4nF 4nF
Co1=Co2 4nF 4nF 2nF 8nF
2.5.1. Case1
In this case the performance of parallel and series connections of two identical power
supplies is compared with the single one. Therefore, in this case the output capacitor of each
module is equal.
Regarding the capacitor value mentioned in Table 2.4, 4nF for each module, the equivalent
capacitance of the series and parallel modules are equal to 2nF and 8nF, respectively. Fig. 2.5
shows the obtained output voltage waveform. As can be seen in the series connection, both
Chapter 2
129
voltage level and rate of rise are increased with the factor of N, (N=2 is considered here). This
has been shown in the analysis of section 2.3 and Table 2.1.
Regarding (2-2) and (2-11) the reflected voltage across the switches is about 110V for both
parallel and series modules. This shows ability of the proposed method in generating high
voltages with low voltage switches. So, it is possible to benefit from ultrafast switches, as
switch operation speed increases when its hold-on voltage decreases.
As the rate of charging is in terms of a time constant RC; hence, as illustrated in the figure,
series connection discharging time is shorter than the parallel and single modules.
2.5.2. Case2
If the load capacitance, which is paralleled with the power supply output, is much bigger
than the power supply output capacitance, then the entire single, parallel and series modules
will have the same equivalent output capacitance. In this case the output capacitance of all
modules is considered to be equal to 4nF (see Table 2.4). As depicted in Fig. 2.6, both parallel
and series modules have same voltage level but the parallel connection shows better
performance in the case of dv/dt.
Fig. 2.5 Output voltage waveform of single, parallel and series modules in Case1 (Co1=Co2=Co).
Fig. 2.6 Output voltage waveform of single, parallel and series modules in Case2 (Cop=Cos=Co).
Chapter 2
130
2.5.3. Case3
Due to differences in semiconductor characteristics and the gate drive circuits, it is quite
important to select identical devices and synchronize the gate signals. Here the same situation
in Case 2 is selected, but 0.2ms delay between gate signals is considered to show the effect.
Fig. 2.7 shows the generated output voltage for parallel and series modules. As can be seen the
output amplitude never reached to the threshold value. Therefore the load is not connected, so
the capacitors are not completely discharged. This creates another problem since the entire
energy in the magnetizing inductor is not fully transferred to the capacitor and the mentioned
equations for estimating the output voltage are not valid.
2.5.4. Case4
In some applications, the load has low impedance like discharge plasmas in water [3]. In
these cases damping factor plays important role in system performance as it discussed in part
D of the previous section.
Here the series and parallel modules are considered, same as in Case 1, but RL = 100Ω is
directly connected. This value puts the all three power supplies in the overdamped condition.
As Fig. 2.8 illustrates the parallel modules show better performance, while series module had
far better performance in Case1 when the system was underdamped. This means that all the
mentioned features about parallel and series connections are valid when the system is
underdamped (see (2-15)). The overdamping effect of the load can be seen as the output
voltage levels are dropped tremendously compared to those in Case 1. Because when the load
resistance decreases, the time constant, τ=L/R, increases, and hence the inductor energy is not
completely transferred to the output capacitor. The easiest way to overcome this problem,
considering ζ in (2-15), is increasing the output capacitance.
Fig. 2.7 Output voltage waveform of single, parallel and series modules with 0.2ms delay between modules gate
signals.
Chapter 2
131
Fig. 2.8 The effect of the load and the damping factor on the system performance.
2.6. Conclusion
This paper demonstrates the advantages of parallel and series configuration of flyback
converters for pulsed power applications. The proposed method aims at increasing the voltage
level and rise time of the generated pulses using conventional low voltage switches. In this
method the flyback converter topology is employed as it shows beneficial characteristics
especially for pulsed power applications. The proposed method is evaluated under different
circumstances and obtained simulation results indicate the effectiveness of the approach.
2.7. References
[1] H. Bluhm, Pulsed Power Systems: Principle and Applications. Berlin Heidelberg: Springer-
Verlag, 2006.
[2] E. Schamiloglu, R. J. Barker, M. Gundersen, and A. A. Neuber, "Modern Pulsed Power:
Charlie Martin and Beyond," Proceedings of the IEEE, vol. 92, pp. 1014-1020, 2004.
[3] H. Akiyama, T. Sakugawa, T. Namihira, K. Takaki, Y. Minamitani, and N. Shimomura,
"Industrial Applications of Pulsed Power Technology," Dielectrics and Electrical Insulation,
IEEE Transactions on, vol. 14, pp. 1051-1064, 2007.
[4] ABB, "High power semiconductors," in Short form catalogue, ed, 2003.
[5] T. Heeren, T. Ueno, D. Wang, T. Namihira, S. Katsuki, and H. Akiyama, "Novel Dual Marx
Generator for Microplasma Applications," Plasma Science, IEEE Transactions on, vol. 33,
pp. 1205-1209, 2005.
[6] T. G. Engel and W. C. Nunnally, "Design and operation of a sequentially-fired pulse forming
network for non-linear loads," Plasma Science, IEEE Transactions on, vol. 33, pp. 2060-
2065, 2005.
[7] C. Jaegu, T. Yamaguchi, K. Yamamoto, T. Namihira, T. Sakugawa, S. Katsuki, et al.,
"Feasibility Studies of EMTP Simulation for the Design of the Pulsed-Power Generator Using
MPC and BPFN for Water Treatments," Plasma Science, IEEE Transactions on, vol. 34, pp.
1744-1750, 2006.
Chapter 2
132
[8] D. Durga Praveen Kumar, S. Mitra, K. Senthil, A. Sharma, K. V. Nagesh, S. K. Singh, et al.,
"Characterization and analysis of a pulse power system based on Marx generator and
Blumlein," Review of Scientific Instruments, vol. 78, pp. 115107-115107-4, 2007.
[9] S. Zabihi, F. Zare, G. Ledwich, A. Ghosh, and H. Akiyama, "A new pulsed power supply
topology based on positive buck-boost converters concept," Dielectrics and Electrical
Insulation, IEEE Transactions on, vol. 17, pp. 1901-1911, 2010.
[10] S. Zabihi, F. Zare, G. Ledwich, and A. Ghosh, "A novel high voltage pulsed power supply
based on low voltage switch-capacitor units," in Pulsed Power Conference, 2009 IET
European, 2009, pp. 1-4.
[11] A. Cid-Pastor, L. Martinez-Salamero, C. Alonso, R. Leyva, and S. Singer, "Paralleling DC-
DC Switching Converters by Means of Power Gyrators," Power Electronics, IEEE
Transactions on, vol. 22, pp. 2444-2453, 2007.
[12] B. R. Lin, H. K. Chiang, C. C. Chen, C. S. Lin, and A. Chiang, "Analysis and implementation
of soft switching converter with series-connected transformers," Electric Power Applications,
IET, vol. 1, pp. 82-92, 2007.
[13] L. M. Redondo and J. F. Silva, "Flyback Versus Forward Switching Power Supply
Topologies For Unipolar Pulsed-Power Applications," Plasma Science, IEEE Transactions
on, vol. 37, pp. 171-178, 2009.
[14] N. Coruh, S. Urgun, and T. Erfidan, "Design and implementation of flyback converters," in
Industrial Electronics and Applications (ICIEA), 2010 the 5th IEEE Conference on, 2010, pp.
1189-1193.
[15] M. Peipei, W. Xinke, Y. Jianyou, C. Henglin, and Q. Zhaoming, "Analysis and design
considerations for EMI and losses of RCD snubber in flyback converter," in Applied Power
Electronics Conference and Exposition (APEC), 2010 Twenty-Fifth Annual IEEE, 2010, pp.
642-647.
133
Statement of Contribution of Co-Authors
The authors listed below have certified that:
1- They meet the criteria for authorship in that they have participated in the conception, execution, or
interpretation, of at least that part of the publication in their field of expertise;
2- They take public responsibility for their part of the publication, except for the responsible author who
accepts overall responsibility for the publications;
3- There are no other authors of the publication according to these criteria;
4- Potential conflicts of interest have been disclosed to (a) granting bodies, (b) the editor or publisher of
journals or other publications, and (c) the head of the responsible academic unit;
5- They agree to the use of the publication in the student’s thesis and its publication on the Australasian
Digital Thesis database consistent with any limitations set by publisher requirements.
In the case of this chapter:
High Voltage Modular Power Supply Using Parallel and Series Configurations of Flyback Converter
for Pulsed Power Applications
Published in: IEEE Transactions on Plasma Science, Vol.40, No.10, pp. 2578-2587, Oct. 2012.
Contributor Statement of contribution
Pooya Davari Proposed the initial design and conducted simulation and data analysis.
Implemented the hardware setup, designed the control strategy, conducted
experimental verifications and wrote the manuscript
22 Feb 2013
Firuz Zare Proposed the initial idea and supervised implementation process, experimentation,
and writing the manuscript
Arindam Ghosh Aided data analysis and writing the paper
Hidenori Akiyama Provided general information about the pulsed power applications demands
Supervisor Confirmation
I have sighted email or other correspondence from all Co-authors confirming their certifying authorship.
Prof. Arindam Ghosh
Name Signature Date: 22 Feb 2013
134
Chapter
3 . H i g h V o l t a g e M o d u l a r P o w e r S u p p l y
U s i n g P a r a l l e l a n d S e r i e s C o n f i g u r a t i o n s
o f F l y b a c k C o n v e r t e r f o r P u l s e d P o w e r
A p p l i c a t i o n s
* School of Electrical Engineering and Computer Science, Queensland University of Technology,
GPO BOX 2434, Brisbane, Australia
† Kumamoto University, Japan
Published in: IEEE Transactions on Plasma Science, Vol.40, No.10, pp. 2578-2587, Oct.
2012.
3
Chapter 3
135
Abstract— To cover wide range of pulsed power applications, this paper proposes a
modularity concept to improve the performance and flexibility of the pulsed power supply.
The proposed scheme utilizes the advantage of parallel and series configurations of flyback
modules in obtaining high voltage levels with fast rise time (dv/dt). Prototypes were
implemented using 600V IGBTs (Insulated-Gate Bipolar Transistor) switches to generate up
to 4 kV output pulses with 1 kHz repetition rate for experimentation. To assess the proposed
modular approach for higher number of the modules, prototypes were implemented using
1700V IGBTs switches, based on 10 series modules, and tested up to 20kV. Conducted
experimental results verified the effectiveness of the proposed method.
3.1. Index Terms
Pulsed power, Flyback converter, Parallel and series connection, High voltage pulse
3.2. Introduction
Rapid release of stored energy as electrical pulses into a load can result in delivery of large
amounts of instantaneous power over a short period. This strategy is called pulsed power [1].
To generate such pulses variety of research and studies have been conducted in pulsed power
area. The most prominent part of a pulsed power system is the utilized switch and topology.
This means that the whole system characteristics such as rise time, repetition rate, voltage
rating, cost, life time, etc, depend on the designed topology and employed switches
specifications.
Gas-state and magnetic switches have been widely used in pulsed power technology, as
they possess a very high electric strength and fast rise time [2]. Gas-state switches require
special operating conditions such as high pressure, vacuum equipments and gas supplies. In
addition, they are bulky, unreliable, have short lifetime span and low repetition rate. Even with
the magnetic switches, which have a higher repetition rate, the problems remain. These
conditions limit the mobility, efficiency and increase the cost and the size of the pulsed power
system [2, 3]. On the other hand, solid-state switches are compact, reliable, cost effective, and
have a long lifetime and repetition rate. The main drawbacks of solid-state switches are their
limited power rating and operation speed [2, 3].
Recently, significant advances in solid-state switches (both in peak power and operation
speed) and exploiting power electronics techniques and topologies have led to compact and
Chapter 3
136
more reliable pulsed power systems. The new developed solid-state switches such as IGBT
and Integrated Gate-Commutated Thyristor (IGCT) have high power rating, but their lower
operation speed comparing with the gas-state switches and high cost still put limits in the
pulsed power supplies [2, 4-6]. For example 5SHY42L6500 which is one of the recent
developed IGCT switches can handle voltage up to 6.5kV and its rise time is 1µs [6], while
GP-81B (triggered spark gap) break down voltage is 120kV with 200ns rise time [7].
Topologies are considered not only as an alternative way to overcome the switch limits but
also in developing flexible and compact systems. Varieties of circuit topologies such as Marx
generators (MG) [8], pulse forming network (PFN) [9], magnetic pulse compressors (MPC)
[10] and multistage Blumlein lines (MBL) [11] have been introduced. These topologies have
been widely used in pulsed power supplies, but complexity, inflexibility and inefficiency are
their main drawbacks [2, 12]. Interest in applying power electronics topologies and techniques
to increase power supply flexibility and reliability is growing fast. In the last decade, research
and studies designate the advantage of using DC-DC converters in pulsed power applications
[12, 13]. However, generation of extra high voltages is still problematic due to components
hold-on voltage limits (such as capacitors, solid-state switches and etc).
Another solution is to combine power supplies. Variety of converter configurations have
been introduced for improving the power supply specifications such as output ripple, high
input voltage, output power ratings, etc [14, 15]. Parallel and series connections are one of the
well known combinations. Designing power supplies based on series or parallel connection are
widely employed for different applications [14, 15].
This paper proposes parallel and series connections of flyback converter modules to
develop power rating and rise time of the pulsed power supply using conventional low voltage
switches. It is to be noted that the proposed idea is applied to the secondary sides of the pulsed
power modules, while the primary sides are connected in parallel due to the current sharing
purpose. The proposed scheme is flexible in increasing the output power and voltage due to its
modular design. Flyback converter is selected due to its unique properties in pulsed power as
is discussed in the next section. Taking the advantage of the parallel connection of the primary
side, it is possible to employ low current rate switches.
The proposed approach is evaluated based on two separate experimentations. In the first
experiment the efficiency of parallel and series configurations theory are evaluated through
two experimental setups. To insure applicability of the proposed modular approach, in terms
of performance and higher number of the modules, generating 20kV output voltage using 10
Chapter 3
137
series flyback modules is presented. The evaluated results indicate the effectiveness of the
proposed method.
3.3. Topology
The proposed approach utilizes flyback converter, which is one of the well known
topologies in the power electronics [16]. Conventional flyback converter is usually preferred
as it is simple, has only one switch and magnetic component, is able to generate high voltage,
can provide multiple outputs, isolation, etc [16, 17]. But when it comes to pulsed power
applications, in addition to the above-mentioned features, it has some more advantages which
make it more suitable. Main features of a flyback converter for pulsed power applications are:
The transformer, in addition to electrical isolation and energy storage also steps down the
reflected voltage across the switch; therefore lower voltage rate switches are needed
unlike other topologies.
Fault tolerant, as the switch is in the off-state during the output pulse.
High voltage output with low input DC voltage.
As the pulsed power applications mostly have R-C characteristics [12], a current source
topology (such as flyback) is a suitable candidate.
Suitable for controlling the energy flow as it acts as both current source and voltage
source.
Below basic principle and operation modes of this converter are briefly described.
The behaviour of a flyback converter can be realized by modelling the transformer with a
simple equivalent circuit consisting of an ideal transformer, magnetizing inductance (Lm) and
leakage inductance (Ll). Fig. 3.1 shows a flyback converter including the simple model of the
transformer connected to a load. In this figure CO is the converter equivalent output
capacitance. This capacitance can get affected in the case of an R-C load. It is to be mentioned
that, to keep the analysis simple, following assumptions are made:
1. The switch and diode are ideal (voltage drop across them is zero).
2. The switch and diode output capacitances are zero.
3. Conduction and switching losses are negligible.
4. The transformer stray capacitances are negligible.
Chapter 3
138
Lm
Vs
C0 Load
1 : n
Ll
Vo
+
-
Fig. 3.1 Flyback converter circuit with transformer equivalent circuit model.
Basically a flyback converter transfers energy from a source into the transformer
magnetizing inductance when the switch is on, and then transfers the stored energy to the load
while the switch is off. The proposed approach operates in the DCM (Discontinues
Conduction Mode). The operating modes are briefly summarized for 4 different modes
(illustrated in Fig. 3.2) below:
Mode 1: In this mode the switch is in the on-state. As Fig. 3.2 depicts, the current flows
through Lm and, during this stage, the energy is stored in the inductor. This mode lasts
depending on the duty cycle of the PWM (Pulse Width Modulation) signal. The
relationship between Lm and Vs can be expressed as:
s
mm
s
mmsms
TD
IL
TD
ILV
t
iLV
.0.
0
(3-1)
where Vs is the dc supply voltage, Im is the maximum current, D is the duty cycle and Ts
is the switching period.
Mode 2: When the switch is turned off the magnetizing current circulates through the
primary side of the transformer and the diode in the secondary side is turned on. As the
switch is turned off the current flowing through Ll is decreased to zero, which induces
voltage spike across the switch according todt
diLv ll . This voltage may damage the
switch if it exceeds the switch break down voltage level. To overcome this problem a
snubber can provide a path for this current and damps the spike to protect the switch
[18] or Ll can be reduced by employing optimum transformer design. This transient
Chapter 3
139
state is considered as one of the operating modes due to its importance (protecting the
switch) and understanding the procedure of the next operating mode.
Mode 3: The stored current in the magnetising inductance flows fully to the secondary side
of the transformer and charges the output capacitor. At this stage the converter is acting
as a current source. The maximum voltage across the switch at this stage is:
maxmax1
)( osswitch vn
Vv (3-2)
where, vomax is the maximum output voltage level and n is the transformer turns ratio.
Mode 4: As the converter operates in DCM, in this mode all the stored energy in Lm is
completely discharged to the capacitor, causing the diode to get turned off. Here the
converter is a voltage source and a high level of the voltage is applied to the load.
Ll
Lm
Vs
1 : n
LoadC0
Vs Vs
Mode 1
Mode 3 Mode 4
1 : n
C0
Mode 2
Ll
Lm
Vs
Load
Lm Lm
Ll Ll
1 : n 1 : n
C0 C0Load Load
Vo
+
-
Vo
+
-
Vo
+
-
Vo
+
-
Fig. 3.2 Operating modes in a flyback converter.
Chapter 3
140
3.4. Proposed Method
High voltage pulses with fast rise-time are the two main important features of a pulsed
power supply [1-3]. Fig. 3.3 illustrates the proposed method, which employs the advantage of
parallel and series configuration of the pulsed power modules to increase the voltage level and
the rise-time, while using the low-voltage switches. As the figure shows, the parallel and
series connections are only considered for the secondary side of the converter.
Ll
Lm
Vs
1 : n
Co2
Ll
Lm
n : 1
Co1Load
Vo
+
-
Converter (1)Converter (2)
(a)
Ll
Lm
Vs
1 : n
Co1
Load
Ll
Lm
1 : n
Co2
Vo
+
-
Converter (2)
Converter (1)
(b)
Fig. 3.3 Flyback converter series and parallel connection (secondary side), a) Parallel, b) Series.
Chapter 3
141
3.4.1. Single Module
To understand and compare the parallel and series configuration, first a single module
characteristic is described. As mentioned, a flyback converter can operate as a current source.
CoIo Vo
+
-
Load
Fig. 3.4 Energy transmission in a single module flyback converter.
Therefore, circuit schematic shown in Fig. 3.1 is simplified as in Fig. 3.4. Comparing these
two figures, the estimation of the output voltage can be summarized as below:
First, as described earlier, the magnetizing inductance is charged by the DC power supply.
The charged energy stored in the inductor Lm under ideal situation can be expressed as:
2
2
1mmpri ILE (3-3)
Regarding (3-1) and (3-3) it is possible to control the energy flow in each pulse by limiting
the Im which can be done by varying the duty cycle or the input voltage. This stored energy
will be transferred to the output capacitor when the switch is in the off-state. The capacitive
stored energy is:
2sec
2
1ooVCE (3-4)
If a lossless system considered, then the stored energy in the primary side, Epri, is equal to
the stored energy in the secondary side, Esec. Therefore, the output voltage can be derived
based on equations (3-3) and (3-4) as:
o
mmo
C
LIV (3-5)
The rate of rise for the generated voltage is defined as dv/dt. When the switch is turned off,
the magnetizing inductance current is at its peak value (Im). Since the current through a
capacitor is proportional to the time-rate of the stored voltage, the rate of rise is:
Chapter 3
142
o
m
o
o
nC
I
C
I
dt
dv
(3-6)
This equation is valid till the current through the capacitor is approximately constant;
otherwise the rate of rise completely depends on the resonant frequency.
The idea of parallel and series configurations of the pulsed power modules is inspired by
both equations (3-5) and (3-6), as these equations show the effect of stored current level,
magnetising inductance and the output capacitor on the generated output voltage and the rise-
time.
3.4.2. Parallel Modules
Considering Fig. 3.3(a) and the fact that each transformer acts as a current source, the
stored energy in the primary side is doubled when two modules are connected in parallel. This
stored energy will be transferred to the output capacitors; hence considering (3-3) and (3-4)
the output voltage magnitude and its rate of rise are as follows:
op
mmparallelo
C
LIV 2
(3-7)
op
o
Parallel C
I
dt
dv 2
(3-8)
where Cop=Co1+Co2. Equations (3-7) and (3-8) indicate the importance of the output capacitor.
Even if the modules are paralleled, the output capacitance can affect the output voltage level
and rise-time and can keep the performance same as a single module. Therefore, the parallel
configuration is beneficial in pulsed power applications when Cop=Co. This happens in the
case of R-C load when the load capacitance is much bigger than the power supply output
capacitance.
3.4.3. Series Modules
Second alternative is series connection of the pulsed power modules, as illustrated in Fig.
3.3(b). Here, the injected energy to the output is also doubled. But as described, the generated
voltage features also depend on the output capacitance. Taking into account the equations (3-
3) to (3-6), the output voltage magnitude and its rate of rise can be expressed as:
Chapter 3
143
os
mmSerieso
C
LIV 2 (3-9)
os
o
Series C
I
dt
dv
(3-10)
where Cos is o2o1
o2o1
CC
CC
. In series modules, same as in parallel modules, the output voltage
level and rise time improve as the output capacitance decreases. Here the series modules are
beneficial in both level of voltage and its rate of rise when Cos<Co.
In pulsed power applications two different cases may occur. One is when the load
capacitance is lower than the output capacitance of the pulsed power supply. In this situation
series modules exhibit better performance as illustrated in Table 3.1. The second case is when
the load has much higher capacitance, therefore all series, parallel and single modules have
same output capacitance. Under this circumstance the parallel connection has better
performance (see Table 3.2).
Hence, depending on the load capacitance one of the modules is beneficial to use.
Regarding the voltage level and rate of rise as presented in Table 3.1 and 3.2, it is possible to
generate wide range of voltage levels and improve dv/dt by connecting N modules in series or
parallel. This feature increases the flexibility of the pulsed power supply in varied
applications.
Table 3.1 Output voltage and rate of rise when Co1=Co2=… CoN=Co.
N Modules connected in oV
dt
dv
Parallel Singleo
o
mm V
C
LI
Singleo
o
dt
dv
C
I
Series Singleo
o
mm NV
C
LNI
Singleo
o
dt
dvN
C
IN
Table 3.2 Output voltage and rate of rise when Cop=Cos=Co.
N Modules connected in oV
dt
dv
Parallel Singleoo
mm VN
C
LIN
Singleo
o
dt
dvN
C
IN
Series Singleo
o
mm VN
C
LIN
Singleo
o
dt
dv
C
I
Chapter 3
144
In addition, regarding (3-1) and the calculated output voltages in Table 3.1 and 3.2, the output
voltage can be easily adjusted for the required level whether by Vin or D. For example (3-11)
shows the output voltage of series module based on the results presented in Table II as:
.TDCL
VNV
om
ino (3-11)
Usually this adjustment is performed by limiting the current by selecting suitable duty cycle
(D). But Im should be calculated in a way that Im<Isat, where Isat is the current saturation level
of the transformer.
As Fig. 3.3(b) shows, in the series connection each switch withstands its own module
reflected output voltage, while in the parallel connection; each switch should tolerate the
whole reflected output voltage. This fact makes the series connected modules more
appropriate for generating high voltage waveforms using low voltage switches. Therefore, (3-
2) can be rewritten as below for N-series module:
maxmax1
)( osswitch vNn
Vv (3-12)
By comparing Fig. 3.3(a) and (b) it is obvious that the idea of parallel and series
connections of the flyback modules is applied to the secondary side, while the primary sides
for both cases are paralleled. The main advantage of paralleling the primary side of the pulsed
power modules are: current sharing so the proposed idea is applicable for high power
applications, reducing the number of the power supplies to one at the input side and ability to
employ low current rating switches.
3.4.4. Operating Conditions
The above mentioned calculations are correct if three important points are considered in the
designing procedure of the pulsed power modules. These are:
1) The output capacitor should be completely discharged during each period; otherwise the
whole stored energy in the magnetising inductor will not transfer to the capacitor.
2) Synchronization of each switch gate signal is required. Delays between each module gate
signal reduce the performance of the system. This is important when higher number of
modules is used.
3) The damping factor (ζ):
Chapter 3
145
As shown in Fig. 3.4, the entire circuit acts as a parallel RLC circuit (the current source is
an inductor). Hence, the output voltage and the rise time are the results of the resonance effect
of this circuit. The output voltage can be expressed as:
tstso eAeAtV 21
21)(
(3-13)
where s1 and s2 are given as:
20
22,1 S
(3-14)
here, RC2
1 is known as neper frequency and
LC
1
0 is the resonance frequency.
0)0(
toV and oItCI
)0( are the primary conditions. I0 is the stored current in the
magnetizing inductor (Im=nI0) transferred to the secondary side. The coefficients A1 and A2
are determined by the primary conditions as:
02
02
20
212
LIAA
(3-15)
Now depending on the damping factor coefficient can have two different values:
Underdamped ζ <1 or α < 0 , C
L
R2
1
C
LIA
2
01
(2-16)
Overdamped ζ >1 or α > 0
01 RIA
(2-17)
Equation (3-16) shows that if the circuit is underdamped, the generated voltage amplitude
is independent of the load impedance, which is the situation in which the proposed method is
valid. But under overdamped condition as in (3-17), the output voltage amplitude depends on
the load impedance and the initial current. Hence, in overdamped condition, the parallel
connection has better performance.
In Fig. 3.5, the effect of damping factor on the output voltage and rise time is illustrated. A
decreasing damping factor results in a higher voltage level and faster rise time.
Chapter 3
146
Fig. 3.5 Effect of damping ratio on output voltage and rate of rise in a single flyback module.
Considering the above mentioned features, this effect should be considered especially in
low impedance applications such as the liquid discharge. Therefore, regarding the mentioned
characteristics, the proposed method is more beneficial for high impedance-capacitive load
such as DBD (Dielectric Barrier Discharge) loads.
3.5. Experimental Results and Discussion
In this section the proposed idea has been evaluated based on practical experimentations. In
the first set of experimental results the idea of parallel and series configurations is evaluated
based on two flyback modules. In the second part, in order to assess the proposed method for
higher number of the modules, the obtained results of a high voltage prototype based on 10
series flyback modules is presented.
3.5.1. Evaluating Parallel and Series Configurations
Two laboratory prototypes for parallel and series connections, based on single module
flyback converter are implemented, to investigate performance of the proposed method
practically. Fig. 3.6 shows the experimental hardware setup for two flyback modules. Here
600V IGBT modules, SK25GB065, are used as power switches. Semikron Skyper 32-pro gate
drive modules are utilized to drive the IGBTs and provide the necessary isolation between the
switching-signal ground and the power ground. Four 1000 V diodes, STTH3010, are
Chapter 3
147
connected in series for each module. A Texas Instrument TMSF28335 DSC (Digital Signal
Controller) is used for PWM signal generation. Two step-up transformer with an UU100 core
3C90 grade material ferrite from Ferroxcube, are designed with N1 = 4 and N2 = 40.
Magnetizing inductance and leakage inductance of each transformer is approximately equal to
152µH and 1.6µH, respectively. Here a 5nF capacitor is placed across the switch to damp the
voltage spikes caused by leakage inductance. The circuit is implemented with Vs = 17V, fs = 1
kHz, 15% duty cycle and resistive load of 20kΩ.
To verify the proposed method and the theoretical analysis, experimental evaluations under
two different conditions (Case 1 and 2) are carried out. Both cases have been conducted
regarding the conditions presented in Table 3.1 and 3.2, respectively.
Fig. 3.6 Hardware setup for two flyback modules.
3.5.1.1. Case1
In this case the performance of parallel and series connections of two identical power
supplies is compared with the single one. Therefore, the output capacitor of each module is
equal (2.35nF for each module). So the equivalent capacitance of the series and parallel
modules are equal to 1.175nF and 4.7nF, respectively.
Chapter 3
148
Fig. 3.7 depicted the measured results. Fig. 3.7(a) shows the current sharing at the primary
sides. Due to parallel connection of the primary sides, in both parallel and series
configurations, the current is approximately equally shared between the two modules. As
mentioned before, parallel connection at the primary sides makes the proposed method
independent from using high current rate switches.
The voltage pulses, in Fig. 3.7(b), illustrate the better performance of series connection
over the other connections in both voltage level and rate of rise. In series modules the
maximum voltage level and rate of rise are 4.02 kV and 608 V/µs, respectively. While the
parallel and single modules achieved approximately same level of the voltage (2.37 kV) and
rate of rise (304 V/µs). As can be seen, the series modules performance is around 1.7 and 2
times better than the other modules in voltage level and rate of rise, respectively. From the
aforementioned theoretical analysis the performance of the series modules in voltage level
expected to be 2 times better, this difference is due to the resonance and dissipation happening
in the practical circuit. As the rate of charging is in terms of a time constant RC; hence, as
illustrated in the figure, series connection discharging time is shorter than the parallel and
single modules.
The voltage across the switch, shown in Fig. 3.8, in series connection is same as the other
modules, while series module generates higher level of voltage. This shows ability of the
series modules in generating high voltages with low voltage switches (see (3-12)). However,
as the figure shows the resonance appeared across the switch should be considered too. This
issue which mainly causes by stray capacitances and inductances of the transformer can be
reduced by proper transformer design and fabrication.
Chapter 3
149
Module1 primary current
Module2 primary current
(a)
(b)
Fig. 3.7 (a) current sharing at the primary side for both parallel and series connections, (b) output voltage
waveform of single, parallel and series modules in Case1 (Co1=Co2=Co).
Chapter 3
150
Fig. 3.8 Voltage across the switches in Case1.
Fig. 3.9 depicts voltage stress across the output diodes. As can be seen the reverse blocking
voltage of the diodes in series connection modules are less than the other ones while it is
operating at higher voltage level. This is due to the fact that each module in series connection
sees the output voltage divided by the number of the modules. This is another advantage of
series connection, because in the parallel connection module’s diode should withstand the
whole generated output voltage. It is also important to consider the current rating of the diodes
for high power applications.
Fig. 3.9 Voltage across the diodes in Case1.
3.5.1.2. Case2
If the load capacitance is much higher than the power supply output capacitance, then all of
the single, parallel and series modules will have the same equivalent output capacitance. In
this case the output capacitance of all modules is considered to be equal to 4.7nF. As depicted
Chapter 3
151
in Fig. 3.10, single module maximum voltage level is 1.72 kV and both parallel and series
modules obtained same voltage levels (2.37kV). But the parallel connection has better rate of
rise (304V/µs) compared to the two other modules (162 V/µs).
Fig. 3.10 Output voltage waveform of single, parallel and series modules in Case2 (Cop=Cos=Co).
Although in this case parallel configuration shows better performance, but as its circuit
components should with-stand higher level of voltages compared to the series connection, the
use of the series modules is more convenient.
3.5.2. High Voltage Modular Power Supply
To insure applicability of the proposed approach for a modular power supply, series
configuration is applied for 10 flyback modules. Fig. 3.11 illustrated the block diagram of the
implemented hardware setup. To decrease the number of switches, depending on the switch
power ratings, a set of modules (here 5) is controlled with one switch. The transferred energy
to the load can be controlled by monitoring the primary stored energy. This is done through
limiting the input current by changing the duty cycle of the PWM signal in (3-1). As Fig. 3.11
shows, a current sensor is used for monitoring the input current. The experimental hardware
setup for 10 series flyback modules is depicted in Fig. 3.12. 1700V IGBT modules
(SKM200GB176D) are used as power switches. Same gate drive modules and controller setup
are used for this experiment as the former one. Also same configurations for the transformers
are used except here the number of the cores per transformer is reduced to one. A HX10-NP is
used as the current transducer (CT).
Regarding the selected components each module is able to generate up to 4kV, which
makes the whole system capable of generating up to 40kV. The circuit is implemented with
Chapter 3
152
fs=1kHz and VS=10V and 10% duty cycle. Each module has a 470pF capacitor (CO) and a
1MΩ high power resistor was connected as a load.
VS
Lo
ad
Control
protocol of
switches
VLoad
+
-
Vo1
+
-
Vo2
+
-CO
CO
CO Vo3
+
-
CO Vo4
+
-
CO Vo5
+
-
Vo6
+
-
Vo7
+
-
Vo8
+
-
Vo9
+
-
Vo10
+
-
CO
CO
CO
CO
CO
Me
asu
rin
g th
e C
urr
en
t
Fig. 3.11 Block diagram of the implemented setup.
Fig. 3.12 Hardware setup for 10-series flyback modules.
Chapter 3
153
Fig. 3.13 Measured output voltage of 10 series modules (VLoad) and a single module (Vo1).
Fig. 3.13 shows the measured practical results when the generated voltage is applied to a
resistive load. As can be seen from Fig. 3.13 the voltage pulse of the series modules obtained
the peak amplitude of 20.8kV, while the single module has the peak voltage of 2.15kV. The
measured output voltage shows the rate of voltage rise of 8kV/µs in the series connection. The
peak voltage level and rate of voltage rise indicate approximately the 10 times (number of
series modules) better performance in comparison with a single module.
3.6. Conclusion
This paper demonstrates the advantages of parallel and series configuration of flyback
converters for pulsed power applications. The proposed method aims at increasing the voltage
level and rise time of the generated pulses while emphasizing on the modularity concept. By
employing parallel and series connections it is possible to generate wide range of voltage
levels with improved rate of rise.
In this method the flyback converter topology is employed as it shows beneficial
characteristics especially for pulsed power applications. The proposed method was evaluated
through two different hardware setups. Two prototypes capable of generating up to 4 kV to
prove the proposed parallel and series configurations theory were used as the first experiment.
In the second experiment the idea of modularity and employing higher number of the modules
were evaluated, by implementing a series connection of 10 flyback modules and tested up to
20kV. The experimentations and analysis demonstrated that the proposed scheme is beneficial
Chapter 3
154
for high impedance-capacitive loads. Generally series modules show better performance, as
the parallel modules has more constraints due to its circuit components blocking voltage level.
3.7. References
[1] H. Bluhm, Pulsed Power Systems: Principle and Applications. Berlin Heidelberg: Springer-
Verlag, 2006.
[2] E. Schamiloglu, R. J. Barker, M. Gundersen, and A. A. Neuber, "Modern Pulsed Power:
Charlie Martin and Beyond," Proceedings of the IEEE, vol. 92, pp. 1014-1020, 2004.
[3] H. Akiyama, T. Sakugawa, T. Namihira, K. Takaki, Y. Minamitani, and N. Shimomura,
"Industrial Applications of Pulsed Power Technology," Dielectrics and Electrical Insulation,
IEEE Transactions on, vol. 14, pp. 1051-1064, 2007.
[4] ABB, "High Power Semiconductors," in Short Form Catalogue, ed, 2003.
[5] ABB, "High Power Semiconductors," in Short From Catalogue, ed, 2011.
[6] ABB, "Asymmetric Integrated Gate Commutated Thyristor," ed, Jun. 2010.
[7] P. Elmer, "Triggered Spark Gaps, Ceramic-Metal ", ed, 2001.
[8] T. Heeren, T. Ueno, D. Wang, T. Namihira, S. Katsuki, and H. Akiyama, "Novel Dual Marx
Generator for Microplasma Applications," Plasma Science, IEEE Transactions on, vol. 33,
pp. 1205-1209, 2005.
[9] T. G. Engel and W. C. Nunnally, "Design and operation of a sequentially-fired pulse forming
network for non-linear loads," Plasma Science, IEEE Transactions on, vol. 33, pp. 2060-
2065, 2005.
[10] C. Jaegu, T. Yamaguchi, K. Yamamoto, T. Namihira, T. Sakugawa, S. Katsuki, et al.,
"Feasibility Studies of EMTP Simulation for the Design of the Pulsed-Power Generator Using
MPC and BPFN for Water Treatments," Plasma Science, IEEE Transactions on, vol. 34, pp.
1744-1750, 2006.
[11] D. Durga Praveen Kumar, S. Mitra, K. Senthil, A. Sharma, K. V. Nagesh, S. K. Singh, et al.,
"Characterization and analysis of a pulse power system based on Marx generator and
Blumlein," Review of Scientific Instruments, vol. 78, pp. 115107-115107-4, 2007.
[12] S. Zabihi, F. Zare, G. Ledwich, A. Ghosh, and H. Akiyama, "A new pulsed power supply
topology based on positive buck-boost converters concept," Dielectrics and Electrical
Insulation, IEEE Transactions on, vol. 17, pp. 1901-1911, 2010.
[13] S. Zabihi, F. Zare, G. Ledwich, and A. Ghosh, "A novel high voltage pulsed power supply
based on low voltage switch-capacitor units," in Pulsed Power Conference, 2009 IET
European, 2009, pp. 1-4.
[14] A. Cid-Pastor, L. Martinez-Salamero, C. Alonso, R. Leyva, and S. Singer, "Paralleling DC-
DC Switching Converters by Means of Power Gyrators," Power Electronics, IEEE
Transactions on, vol. 22, pp. 2444-2453, 2007.
[15] B. R. Lin, H. K. Chiang, C. C. Chen, C. S. Lin, and A. Chiang, "Analysis and implementation
of soft switching converter with series-connected transformers," Electric Power Applications,
IET, vol. 1, pp. 82-92, 2007.
[16] L. M. Redondo and J. F. Silva, "Flyback Versus Forward Switching Power Supply
Topologies For Unipolar Pulsed-Power Applications," Plasma Science, IEEE Transactions
on, vol. 37, pp. 171-178, 2009.
Chapter 3
155
[17] N. Coruh, S. Urgun, and T. Erfidan, "Design and implementation of flyback converters," in
Industrial Electronics and Applications (ICIEA), 2010 the 5th IEEE Conference on, 2010, pp.
1189-1193.
[18] M. Peipei, W. Xinke, Y. Jianyou, C. Henglin, and Q. Zhaoming, "Analysis and design
considerations for EMI and losses of RCD snubber in flyback converter," in Applied Power
Electronics Conference and Exposition (APEC), 2010 Twenty-Fifth Annual IEEE, 2010, pp.
642-647.
156
Statement of Contribution of Co-Authors
The authors listed below have certified that:
1- They meet the criteria for authorship in that they have participated in the conception, execution, or
interpretation, of at least that part of the publication in their field of expertise;
2- They take public responsibility for their part of the publication, except for the responsible author who
accepts overall responsibility for the publications;
3- There are no other authors of the publication according to these criteria;
4- Potential conflicts of interest have been disclosed to (a) granting bodies, (b) the editor or publisher of
journals or other publications, and (c) the head of the responsible academic unit;
5- They agree to the use of the publication in the student’s thesis and its publication on the Australasian
Digital Thesis database consistent with any limitations set by publisher requirements.
In the case of this chapter:
A Flexible Solid-State Pulsed Power Topology
Published in proceedings of: 15th International Power Electronics and Motion Control Conference, EPE-
PEMC 2012 ECCE Europe, Novi Sad, Serbia.
Contributor Statement of contribution
Pooya Davari Proposed the initial design and control strategy. Develop the control strategy
regarding the simulation and data analysis. Wrote the manuscript.
22 Feb 2013
Firuz Zare Proposed the initial idea and supervised the simulation process and writing the
manuscript.
Arindam Ghosh Aided data analysis and writing the paper
Supervisor Confirmation
I have sighted email or other correspondence from all Co-authors confirming their certifying authorship.
Prof. Arindam Ghosh
Name Signature Date: 22 Feb 2013
157
Chapter
4 . A F l e x i b l e S o l i d - S t a t e P u l s e d P o w e r
T o p o l o g y
* School of Electrical Engineering and Computer Science, Queensland University of Technology,
GPO BOX 2434, Brisbane, Australia
Presented and Published at: 15th International Power Electronics and Motion Control
Conference, EPE-PEMC 2012 ECCE Europe, Novi Sad, Serbia, Sep. 2012.
4
Chapter 4
158
Abstract — Advances in solid-state switches and power electronics techniques have led to the
development of compact, efficient and more reliable pulsed power systems. This paper
proposes an efficient scheme that utilizes modular switch-capacitor units in obtaining high
voltage levels with fast rise time (dv/dt) using low voltage solid-state switches. The proposed
pulsed power supply has flexibility in terms of controlling energy and generating broad range
of voltage levels. The energy flow can be controlled as the stored energy can be adjusted by a
current source utilized at the first stage of the system. Desirable voltage level can be obtained
by connecting adequate number of switch-capacitor units. Moreover, the proposed topology is
load independent. Therefore it can easily supply wide range of applications especially the low
impedance ones. The effectiveness of the proposed approach is verified by simulations.
4.1. Keywords
Pulsed power supply, Current source, High voltage generator, Power Converter control
4.2. Introduction
Rapid release of stored energy as electrical pulses into a load can result in delivery of large
amounts of instantaneous power over a short period. This strategy is called pulsed power [1].
Generally, a pulsed power system consists of three main parts: energy storage, pulse generator
and load. The most prominent part of a pulsed power system is the pulse generator, which is
based on utilized switch and topology. Hence, the switch is the connecting device between the
storage and the load. This means that the whole system characteristics such as rise time,
repetition rate, voltage rating, efficiency, cost, life time, etc, depend on the employed switch
specifications.
Gas-state and magnetic switches have been widely used in pulsed power technology, as
they possess a very high electric strength and fast rise time [2]. Gas-state switches require
special operating conditions such as high pressure, vacuum equipments and gas supplies. In
addition, they are bulky, unreliable, have short lifetime span and low repetition rate. Even with
the magnetic switches, which have a higher repetition rate, the problems remain. These
conditions limit the mobility, efficiency and increase the cost and the size of the pulsed power
system [2], [3]. On the other hand, solid-state switches are compact, reliable, cost effective,
and have a long lifetime and repetition rate. The main drawbacks of solid-state switches are
their limited power rating and operation speed [2], [3].
Chapter 4
159
Recently, significant advances in solid-state switches (both in peak power and operation
speed) and exploiting power electronics techniques and topologies have led to compact,
efficient and more reliable pulsed power systems. The new developed solid-state switches
such as Insulated-Gate Bipolar Transistor (IGBT) and Integrated Gate-Commutated Thyristor
(IGCT) have high power rating [4], but their lower operation speed comparing with the gas-
state switches and high cost still put limits in the pulsed power supplies.
One way to increase the pulsed power supply performance and cover the switch limits is to
explore alternative circuit topologies. Varieties of circuit topologies such as Marx generators
(MG) [5], pulse forming network (PFN) [6], magnetic pulse compressors (MPC) [7] and
multistage Blumlein lines (MBL) [8] have been introduced. These topologies have been
widely used in pulsed power supplies, but complexity, inflexibility and inefficiency are their
main drawbacks [2], [9]. Interest in applying power electronics topologies and techniques to
increase power supply efficiency and reliability is growing fast. In the last decade, research
and studies designate the advantage of using converters in pulsed power applications [3], [9-
11].
Pulsed power applications present one of the most varied ranges of loads in terms of load
behavior and impedance. Hence, various load conditions adversely affect the power supply
flexibility. Regarding this, it is desirable to have flexible pulsed power supply that can cover
varied range of applications. This flexibility comes in terms of: adjustable output voltage,
controlling the flow of energy and repetition rate. Moreover, the generated voltage pulse
waveform must be load independent [11]. This is quite important especially in the low
impedance applications such as in water discharge.
To achieve all the mentioned features, this paper proposes an efficient scheme that utilizes
modular switch-capacitor units in obtaining high voltage levels with fast rise time (dv/dt)
using low voltage solid-state switches. The proposed pulsed power supply has flexibility in
terms of controlling input energy and generating broad range of voltage levels. The energy
flow can be controlled by adjusting the utilized current source at the first stage of the system
and desirable voltage level can be obtained by connecting adequate number of switched-
capacitor units. The proposed topology is able to operate independently from the load while
preserving high repetition rate. Moreover, the control algorithm is designed in a way that it
can prevent from any possible faults and protect the system from the over-voltage.
Chapter 4
160
S11
S12
S21
S22
Sn1
Sn2
SL
Load
Si SVVS
L
D1
D2
C1
C2
Cn
Posi
tive
buck
-boost
conver
ter
Load Switch
Control
protocol of
switches
Sw
itch
-cap
acit
or
unit
s
Hysteresis
current control
Hysteresis
voltage control
Ov
er-r
ide
sig
nal
Ov
er-r
ide
sig
nal
Voltage
Feedback
Synchronizing
PWM Signal
Fig. 4.1 Proposed pulsed power topology for n level switch-capacitor units.
4.3. Topology
Fig. 4.1 illustrates the proposed method with a brief controlling algorithm scheme. Generally,
as can be seen the proposed topology consists of three different stages. The first stage is a
positive buck-boost converter as a current source and the energy storage section. The next
stage is the switch-capacitor units, which generates the required voltage level and acts as a
pulse generator, and the last stage is the load.
In order to have a flexible pulsed power supply, here the positive buck-boost converter
control the amount of stored energy by employing appropriate duty cycle for Si and SV. The
desirable voltage level can be obtained by connecting adequate number of switch-capacitor
units in series. This feature also makes the system capable of taking the advantage of
employing low voltage switches, as lower voltage rating switches have intrinsically better
behavior [12].
Chapter 4
161
In addition to the common aforementioned three sections, this topology utilizes a load
switch for low impedance applications such as water decontamination in which it disconnects
the load from the pulsed power supply before the required voltage level is obtained. This
makes this topology load independent, because while the switch-capacitor units are charging it
is not connected to the load. In order to preserve a same break down voltage for the load
switch as the other switches, the switching signals need to be controlled precisely.
4.4. Control Strategy and Operating Modes
Effective algorithms are considered to define the turn on and off procedures of the power
switches. Fig. 4.2 and Fig. 4.3 illustrate the operating modes of the proposed topology
regarding the designed control algorithm.
4.4.1. Current & Voltage Control
The output current of the positive buck-boost converter need to be adjusted on a suitable
level. A hysteresis control is used for stabilizing this parameter. This control technique is
selected due to its simplicity, robust performance and stability [13, 14].
The hysteresis control determines the duty cycle of Si by comparing the inductor current
with the predefined reference current. The hysteresis control performs based on the upper and
the lower bands. When the inductor current iL reaches to the upper-band Si turns off and when
it reduces down to the lower-band it turns on. Here the hysteresis bands are selected in such
way that the converter operates in CCM (Continues Conduction Mode). Therefore this
converter operates as a current source for the switch-capacitor unit.
To charge switch-capacitor units equally, the output voltage needs to be adjusted on an
appropriate level. This can be done same as the current control by utilizing the hysteresis
method. Here the Vref is selected as the controllable voltage. When Vref increased to the upper-
band then the SV turns on and when it reduces down to the lower-band it turns off.
Considering the mentioned control procedure the operating modes can be depicted as in
Fig. 4.2. In the first mode, Fig. 4.2a, both Si and SV are in the on-state. During this mode the
inductor L starts to store energy. The required charging time can be calculated as:
S
LLSL
V
iLt
t
iLVV
(4-1)
At this stage the required energy based on the pulsed power application can be estimate as:
Chapter 4
162
2
2
1LLiE (4-2)
Therefore, it is possible to control the energy flow by selecting proper values for iL and L.
Si SVVS D1
C R
VrefL
iL
D2
(a)
Si SVVS
L
D1
D2
C R
Vref
iL
(b)
Si SVVS
L
D1
D2
C R
Vref
iL
(c)
Fig. 4.2 Operating modes for positive buck-boost converter.
In the voltage control algorithm SV turns on when iL drops below the desired level, this
override signal speed up the inductor charging process. This is why in the first mode SV is in
the on-state. This mode last when iL reaches to the upper-band of the current hysteresis
control.
In the second mode, Fig. 4.2b, the inductor is charged to the required level. Hence, Si and
SV will turn off and both diodes D1 and D2 turn on in order to transfer the stored energy to the
output which is the capacitor units.
In the last mode Vref reaches to its upper limit when the capacitor unit is charged up to the
required level. Here, SV turns on as shown in Fig. 4.2c. As the proposed topology is designed
in a way that each switches breakdown voltage should be equal to one switched-capacitor unit,
therefore in order to protect this switch, SV also turns on prior to connecting all switch-
capacitor units in series and supplying the load.
Chapter 4
163
4.4.2. Charging & Load Control
The charging process is synchronized with a PWM (Pulse Width Modulation) signal, which
also determines the repetition rate of the pulsed power supply. Here the procedure is explained
for two switched-capacitor units (Fig. 4.3) which can be extended to n units. The procedure is
controlled by Sn1 and Sn2, which n is the capacitor unit number.
When the PWM signal is high, the charging process begins with the first switch-capacitor
unit. At this mode (Fig. 4.3a), S22 turns on and the stored energy form the current source starts
charging the capacitor of the first unit via anti-parallel body diode of S11. When the capacitor
voltage reaches to the required level the second mode (Fig. 4.3b) starts by turning S12 on and
S22 off.
When the last unit charged up (here the second unit), SL and SV should be turned on prior to
applying all units voltages to the load (turning on the high-side switches). The reason is to
protect these switches from over voltage, as each switch blocking voltage is considered to be
equal to one capacitor unit voltage. As can be seen from Fig. 4.3c in the third mode the load is
connected to the last unit and the current source is disconnected as SV is in the on-state.
Depending on the load impedance and its time constant, , in the worst scenario case a
low impedance load may fully discharge the last unit.
After SL and SV turned on, the forth mode occurred by turning on the high side switches of
the capacitor units. At this moment, as Fig. 4.3d depicted, all of the capacitors connected in
series and a high voltage is applied to the load. Due to the last unit discharging in the previous
mode in the worst case the output voltage is:
Cout VnV )1( (4-3)
where n is the number of switched-capacitor units and VC is the voltage of each unit.
Regarding (4-3) the voltage drop issue can be solved by increasing the number of the units.
Finally, when the PWM signal is low all of the high-side and low-side switches turn off and
on, respectively and the voltage across the load drops to zero (Fig. 4.3e).
The reason that the last unit should be connected to the load prior to the rest of the units is
due to the presence of SL. To protect SL this switch needs to turn on before supplying the high
voltage to the load. Before supplying the load the low-side switches are in on-state (the output
is short circuit). Practically a switch can turns on when it has been biased enough. Hence, the
last unit gets connected to provide a voltage across SL, so it can turn on.
Chapter 4
164
The load switch (SL) is considered in the proposed topology to make load independent. In
the other word, SL causes high impedance across the capacitor units when the load has low
impedance so the load impedance (especially in low impedance applications) doesn’t affect
the system. Hence, for high impedance loads there is no need to place SL and therefore there is
no need to connect last unit prior to connection of the other units.
The break-down voltage of the employed components is equal to one capacitor unit except
D2. Therefore, the control algorithm should design in a way to prevent from any overvoltage
across the components. In practical implementation, for generating high level of voltages, D2
can be several series connected diodes.
S11
S12
SL
LoadC1
S21
S22
C2
S11
S12
SL
LoadC1
S21
S22
C2
S11
S12
SL
LoadC1
S21
S22
C2
S11
S12
SL
LoadC1
S21
S22
C2
S11
S12
SL
LoadC1
S21
S22
C2
(a) (b)
(c) (d) (e)
Fig. 4.3 The charging and supplying the load operating modes.
Regarding the mentioned charging and supplying the load process two different control
algorithms can be employed. The main difference between these two algorithms is the way
that the voltage feedback is provided.
One possible way to control the whole process is to provide individual voltage feedback
from each capacitor unit. Hence, the charging process of each unit can be started by
monitoring the previous stage charging level (Vch). Finally, when the unit’s capacitors charged
to the required level (VR) then it can be applied to the load. Fig. 4.4 illustrated the algorithm
flowchart. The advantage of this algorithm is the ability of preventing from any possible faults
Chapter 4
165
and indicating the failed unit as each unit’s voltage is measured. But the main drawback is
having multiple voltage feedback which increases the hardware implementation complexity.
The second potential way for the control algorithm is based on the measured voltage across
the output. The simplicity is the main advantage of this algorithm comparing with the previous
one, as it only employs one voltage feedback. Fig. 4.5 shows the second algorithm flowchart.
As can be seen after charging each unit’s capacitor, low-side switches should turn on in order
to short circuit the output. Hence, at each charging stage the measured output voltage
corresponds to only one unit. By short circuiting the output after each charging stage and using
a flag (h in Fig. 4.5) it is possible to control the charging and supplying the load process.
SK1: off
SK2: onK = 1,..., n
h =1
Start
If Sync
PWM=1
Yes
Yes
No
Sh1 & Sh2: off
Sg1: off
Sg2: on
g = 1,..., n – h
SL: off
If Vch=VR
NoYesIf VcK>VR
K =1,..., n
NoIf h = n
Yes
No
h = h + 1
Yes
SL & SV: on
Sn1: on
Sn2: off
SK1: off
SK2: on
K = 1,..., n-1 Delay
SK1: on
SK2: off
K = 1,..., n
If Sync
PWM=0
No
Charging unit h
SK1: off
SK2: on
Si: off
Fault detection
Supplying the Load
Fig. 4.4 Charging and supplying the load control algorithm flowchart (based on individual voltage feedback of
each unit).
Chapter 4
166
SK1: off
SK2: onK = 1,..., n
h =1
Start
If Sync
PWM=1
Yes
Yes
No
Sh1 & Sh2: off
Sg1: off
Sg2: on
g = 1,..., n – h
SL: off
If Vout=VR
NoYesIf Vout>VR
K =1,..., n
NoIf h = n
Yes
No
h = h + 1
Yes
SL & SV: on
Sn1: on
Sn2: off
SK1: off
SK2: on
K = 1,..., n-1 Delay
SK1: on
SK2: off
K = 1,..., n
If Sync
PWM=0
No
Charging unit h
SK1: off
SK2: on
Si: off
Fault detection
SK1: off
SK2: on
K = 1,..., n
Supplying the Load
Fig. 4.5 Charging and supplying the load control algorithm flowchart (based on only one voltage feedback).
The main advantages of the second algorithm are the ability to prevent from any fault (as the
output voltage is measured) and convenient practical implementation.
One of the important issues is to protect the circuit against the failures. These failures
happen if switch/switches stop operating, which can leads to over-voltage. As both described
algorithms monitors output voltage so they have the ability to prevent from any faults by
stopping the charging process. As an example, Fig. 4.4 and 4.5 illustrate one way of detecting
system failures.
Another issue needs to be pointed out is the duty cycle of the PWM signal, which should
adjusted in a way that it covers the whole charging process. To calculate the required pulse
width, let’s consider an ideal system. In an ideal situation the current source is charging each
capacitor unit with a constant current. Therefore without considering any losses:
Chapter 4
167
LC it
vC
dt
dvCi
(4-4)
here ∆v is equal to the voltage level that each switched-capacitor units need to charge up.
Considering the time require for charging the capacitor, ∆t, and the number of the units the
minimum pulse width can be expressed as:
100
T
tD
i
vCnt
L
(4-5)
where n is the number of the units, D is the duty cycle and T is the PWM signal period.
4.5. Simulation Results and Analysis
In this section the proposed topology with three switched-capacitor units (n=3 in Fig. 4.1)
is simulated in order to verify the performance of the proposed method. The software that was
used for simulation was MATLAB 7.10. Table 4.1 shows the parameters value used in
simulation corresponding to Fig. 4.1. The upper and lower bands for adjusting the current are
selected as 20.5A and 19.5A, respectively. The charging level is selected as 1000V, and to
control the voltage of each unit on 1000V the bands are selected with 2V variations and fS is
the frequency of the synchronizing PWM signal. The delay time as mentioned in Fig. 4.4 and
4.5 is selected as 1µs.
The simulation results and analysis are presented for three distinguished cases. In Case 1
the output voltage for three different loads are depicted. The switching signals and switched-
capacitor unit voltages are considered for one cycle in Case 2. Finally, the voltage stresses
across the critical components are depicted in Case 3.
4.5.1. Case1
Here the effects of different loads impedance on the generated output voltage are depicted
in Fig. 4.6. For high impedance loads the considered delay for turning on the SL and SV
doesn’t affect the output voltage level, but as the load impedance decreases the voltage drop
across the output voltage increases. This is due to the last unit capacitor discharging based on
the For the low impedance applications this issue can be solved by increasing the
number of the switched-capacitor units.
Chapter 4
168
4.5.2. Case2
In this case the switching signals for the charging process and the generated output voltage
are demonstrated for 1kΩ load. Fig. 4.7 completely demonstrates the charging and supplying
processes. As depicted during the charging process all of the high-side switches are in the off-
state. As can be seen, the last unit S31 turns on prior to the other switches. Finally because the
PWM signal becomes low, the output voltage drop to zero before the load fully discharges the
capacitors.
Table 4.1 Simulation Parameters
VS L C1,2,3 D fS iL VC1,2,3
100V 1mH 10nF %0.5 1kHz 20A 1000V
Fig. 4.6 The effect of the load impedance on the generated output voltage.
4.5.3. Case3
As mentioned before, in order to benefit from low voltage switches the break down voltage
of switches are selected to be equal to one capacitor unit voltage. Therefore, it is quite
important to turn on and off SL and SV effectively, otherwise they should tolerate the whole
output voltage. As Fig. 4.8 shows, following the proposed switching strategy the voltage
Chapter 4
169
across these two switches never exceeded from 1 kV. The only component in this topology
which must handle the whole output voltage is D2.
Fig. 4.7 Voltage waveforms with relative switching signal patterns.
Fig. 4.8 Voltage across the critical components.
Chapter 4
170
4.6. Conclusion
This paper demonstrates a flexible high voltage pulsed power topology. Various voltage
levels can be obtained by connecting different numbers of switched-capacitor units. The
energy flow can be controlled via a positive buck-boost converter. The repetition rate can be
adjusted using a synchronizing PWM signal. Moreover, the proposed topology is load
independent, as it is disconnected from the load during the charging process. This feature
makes the proposed method suitable for supplying wide range of applications specially the one
with low impedance such as water discharge.
The proposed method is evaluated under different circumstances and obtained simulation
results indicate the effectiveness of the proposed approach.
4.7. References
[1] Hansjoachim Bluhm, Pulsed Power System: Principle and Applications, Berlin Heidelberg,
Springer-Verlag, 2006.
[2] E. Schamiloglu, R.J. Barker, M. Gundersen, and A.A. Neuber, "Modern Pulsed Power: Charlie
Martin and Beyond," Proceedings of the IEEE , vol.92, no.7, pp. 1014- 1020, July 2004.
[3] H. Akiyama, T. Sakugawa, T. Namihira, K. Takaki, Y. Minamitani, and N. Shimomura,
"Industrial Applications of Pulsed Power Technology," Dielectrics and Electrical Insulation,
IEEE Transactions on, vol.14, no.5, pp.1051-1064, October 2007.
[4] ABB, High power semiconductors, Short form catalogue, 2003.
[5] T. Heeren, T. Ueno, D. Wang, T. Namihira, S. Katsuki, and H. Akiyama, "Novel Dual Marx
Generator for Microplasma Applications," Plasma Science, IEEE Transactions on , vol.33, no.4,
pp.1205- 1209, Aug. 2005.
[6] T.G. Engel and W.C. Nunnally, "Design and operation of a sequentially-fired pulse forming
network for non-linear loads," Plasma Science, IEEE Transactions on, vol.33, no.6, pp. 2060-
2065, Dec. 2005.
[7] Choi. Jaegu, T. Yamaguchi, K. Yamamoto, T. Namihira, T. Sakugawa, S. Katsuki, and H.
Akiyama, "Feasibility Studies of EMTP Simulation for the Design of the Pulsed-P ower
Generator Using MPC and BPFN for Water Treatments," Plasma Science, IEEE Transactions on,
vol.34, no.5, pp.1744-1750, Oct. 2006.
[8] D. Durga Praveen Kumar, S. Mitra, K. Senthil, A. Sharma, K. V. Nagesh, S. K. Singh, J. Mondal,
A. Roy, and D. P. Chakravarthy, "Characterization and analysis of a pulse power system based on
Marx generator and Blumlein," Review of Scientific Instruments, vol.78, no.11, pp.115107-
115107-4, Nov 2007.
[9] S. Zabihi, F. Zare, G. Ledwich, A. Ghosh, and H. Akiyama, "A new pulsed power supply
topology based on positive buck-boost converters concept," Dielectrics and Electrical Insulation,
IEEE Transactions on, vol.17, no.6, pp.1901-1911, Dec. 2010.
[10] P. Davari, F. Zare, A. Ghosh, and H. Akiyama, "High-Voltage Modular Power Supply Using
Parallel and Series Configurations of Flyback Converters for Pulsed Power Applications," Plasma
Science, IEEE Transactions on, 2012.
Chapter 4
171
[11] L. Redondo and J. F. Silva, "26 - Solid State Pulsed Power Electronics," in Power Electronics
Handbook (Third Edition), ed Boston: Butterworth-Heinemann, 2011, pp. 669-707.
[12] Honggang Sheng; Wei Shen; Hongfang Wang; Dianbo Fu; Yunqing Pei; Xu Yang; Fei Wang;
Boroyevich, D.; Lee, F.C.; Tipton, C.W.; , "Design and Implementation of a High Power Density
Three-Level Parallel Resonant Converter for Capacitor Charging Pulsed-Power Supply," Plasma
Science, IEEE Transactions on , vol.39, no.4, pp.1131-1140, April 2011.
[13] F. Zare, G. Ledwich, "A hysteresis current control for single-phase multilevel voltage source
inverters: PLD implementation," Power Electronics, IEEE Transactions on, vol.17, no.5, pp. 731-
738, Sep 2002.
[14] A.A. Boora, F. Zare, G. Ledwich, A. Ghosh, " A general approach to control a Positive Buck- Boost converter to achieve robustness against input voltage fluctuations and load changes,"
Power Electronics Specialists Conference, PESC 2008, pp.2011-2017, 15-19 June 2008.
172
Statement of Contribution of Co-Authors
The authors listed below have certified that:
1- They meet the criteria for authorship in that they have participated in the conception, execution, or
interpretation, of at least that part of the publication in their field of expertise;
2- They take public responsibility for their part of the publication, except for the responsible author who
accepts overall responsibility for the publications;
3- There are no other authors of the publication according to these criteria;
4- Potential conflicts of interest have been disclosed to (a) granting bodies, (b) the editor or publisher of
journals or other publications, and (c) the head of the responsible academic unit;
5- They agree to the use of the publication in the student’s thesis and its publication on the Australasian
Digital Thesis database consistent with any limitations set by publisher requirements.
In the case of this chapter:
Effect of Pulsed Power on Particle Matter in Diesel Engine Exhaust Using a DBD Plasma Reactor
Submitted to: IEEE Transactions of Plasma Science
Contributor Statement of contribution
Meisam Babaie Designed and developed the mechanical setup. Planned and conducted
experimental verification and wrote the paper.
Pooya Davari Proposed the initial idea and designed and implemented the pulsed power supply.
Planned and conducted experimental verification and wrote the paper.
22 Feb 2013
Firuz Zare
Proposed the initial idea and evaluation process. Supervised the data analysis,
experimentation and writing the paper.
MD Mostafizur Rahman Aided experimentation process
Hassan Rahimzadeh Aided experimentation process
Zoran Ristovski Aided data analysis, supervised experimentation process and writing the paper
Richard Brown Aided data analysis, supervised experimentation process and writing the paper
Supervisor Confirmation
I have sighted email or other correspondence from all Co-authors confirming their certifying authorship.
Prof. Arindam Ghosh
Name Signature Date: 22 Feb 2013
173
Chapter
5 . E f f e c t o f P u l s e d P o w e r o n P a r t i c l e M a t t e r
i n D i e s e l E n g i n e E x h a u s t U s i n g a D B D
P l a s m a R e a c t o r
School of Chemistry, Physics and Mechanical Engineering, Queensland University of Technology,
GPO BOX 2434, Brisbane, Australia
* School of Electrical Engineering and Computer Science, Queensland University of Technology,
GPO BOX 2434, Brisbane, Australia
+ Danfoss Power Electronics, Graasten DK-6300, Denmark
Amirkabir University of Technology, Iran
Published in: IEEE Transactions on Plasma Science, Vol.41, No.8, pp. 2349-2358, Aug.
2013.
5
Chapter 5
174
Abstract— Non-thermal plasma (NTP) treatment of exhaust gas is a promising technology
for both nitrogen oxides (NOX) and particulate matter (PM) reduction by introducing plasma
into the exhaust gases. This study considers the effect of NTP on PM mass reduction, PM size
distribution and PM removal efficiency. The experiments have been performed on real exhaust
gases from a diesel engine. The NTP is generated by applying high voltage pulses using a
pulsed power supply across a dielectric barrier discharge (DBD) reactor. The effects of the
applied high voltage pulses up to 19.44 kVpp with repetition rate of 10 kHz are investigated.
In this paper, it is shown that PM removal and PM size distribution need to be considered both
together, as it is possible to achieve high PM removal efficiency with undesirable increase in
the number of small particles. Regarding these two important factors, in this research, 17
kVpp voltage level is determined to be an optimum point for the given configuration.
Moreover, particles deposition on the surface of the DBD reactor was found to be a significant
phenomenon which should be considered in all plasma PM removal tests.
5.1. Index Terms
Particle size distribution, Particle mass reduction, Diesel exhaust gas, Dielectric barrier
discharge (DBD), Pulsed power, Push-pull converter
5.2. Introduction
There is a continuous increase in the number of diesel engines in both stationary and
mobile application due to the lower operating cost, higher thermal efficiency, longer durability
as well as lower hydrocarbons (HC) and carbon monoxide (CO) emissions [1]. However NOX
and particulate matter (PM) emissions still remain the two main environmental concerns in
diesel engine applications. Studies focused on risk assessment have showed that high outdoor
NOX concentration observed in residential areas contributes to increased respiratory and
cardiovascular diseases and mortality [2]. Moreover, the health effects of diesel particulate
matter have been an area of concern for many years, due to both the chemical composition and
the particle size distribution [3]. The small particles are inhalable and penetrate deep into the
lungs where they are able to enter the bloodstream and even reach the brain [4, 5].
Up to now, several technologies have been applied for NOX and particulate treatment of
diesel engines. In recent years, application of non-thermal plasma (NTP) in exhaust gas
treatment has gained lot of interests [6-9]. NTP treatment of exhaust gas is a promising
technology for both NOX and PM reduction by introducing plasma inside the exhaust gases.
Chapter 5
175
In the non-thermal plasma, electrons have a kinetic energy higher than the energy
corresponding to the random motion of the background gas molecules. The intent of using
non-thermal plasma is to selectively transfer the input electrical energy to the electrons which
would generate free radicals through collisions and promote the desired chemical changes in
the exhaust gas. These reactions can be accomplished at a fraction of the energy which is
required in a thermal plasma system [10]. NOX, unburned hydrocarbons, CO, and PM will be
oxidized due to oxidation processes which happen by introducing plasma in the exhaust gas.
Applying pulsed power is one of the efficient ways to generate NTP. Pulsed power is the
rapid release of stored energy in the form of electrical pulses into a load, which can result in
delivery of large amounts of instantaneous power over a short period of time. Recently, solid-
state pulsed power has gained more interest as it is compact, reliable, has a long lifetime and
high repetition rate. In the last decade, research and studies established the advantage of using
power electronics topologies in pulsed power applications [11-13]. In this research a bipolar
pulsed power supply based on push-pull topology is implemented.
Diesel particulate matter (DPM) consist mostly of carbonaceous soot with minor
components of volatile organic fraction (VOF) from unburned fuel, lubricating oil, inorganic
compounds such as ash and sulphur compounds and metals including zinc from lubricating
oil [38]. DPM are the cause of a series of adverse effects on environment [39] and human
health [40-42]. Particulate formation begins with nucleation in the engine cylinder and
dilution tunnel, and is followed thereafter by agglomeration [43]. Most of the diesel
particulate matter mass is in the accumulation mode, whereas in terms of particle number,
most particles are found in the nucleation mode. More than 90% of diesel exhaust-derived
PM is smaller than 1 μm in diameter [44]. Most of the mass is in the 0.1–1.0 μm
“accumulation” size fraction, while most of the particle numbers are in the <0.1 μm “nano-
particle” fraction [45, 46]. Ultrafine particles have an aerodynamic diameter less than 100 nm
and are emitted in high number by compression ignition engines. Whilst ultrafine particles
do not contribute much to the total mass of particulate matter emitted from an engine, they
contribute greatly to the total number of particles. The particle size distribution of particulate
matter from compression ignition engines has become of increased concern since a study by
the Health Effects Institute demonstrated an increased number of nanoparticles emitted from
a 1991 Cummins engine, despite a reduction in overall particle mass, relative to an older
1988 Cummins engine [47].
Ultrafine particles can penetrate deep into the lungs where they are able to enter the
bloodstream and even reach the brain [4]. The respiratory health effects (in particular asthma)
Chapter 5
176
from particle emissions correlate strongly with particle number, rather than particle mass
emissions [48] . In 2014 the Euro VI regulation will be implemented and the number of
particles emitted by compression ignition engines (in addition to a new particle mass limit)
will be regulated. This demonstrates that particle number emissions are becoming a very
prominent issue in engine design and research [49, 50].
The main concern of this paper is to analyse the effect of pulsed power on PM mass
reduction and PM size distribution considering the pulsed power effects on ultra-fine particles
emitted from real diesel engine exhaust gas. In this research, a pulsed power supply based on
the push-pull inverter is developed to generate up to 19.44 kVpp across the DBD load. The
experiments were conducted at different voltage levels with fixed repetition rate of 10 kHz.
PM mass reduction, PM removal efficiency and PM size distribution are investigated by
evaluating the results obtained.
5.3. Experimental method
5.3.1. Experimental Setup
A schematic diagram of the experimental setup is shown in Fig. 5.1. Experiments were
conducted on a modern turbo-charged 6-cylinder Cummins diesel engine (ISBe22031) at the
Queensland University of Technology (QUT) Biofuel Engine Research Facility (BERF). The
engine has a capacity of 5.9l, a bore of 102 mm, a stroke length of 120 mm, a compression
ratio of 17.3:1 and maximum power of 162 kW at 2500 rpm. Particle number distributions
are measured with a scanning mobility particle sizer (SMPS) consisting of a TSI 3080
classifier, which pre-selects particles within a narrow mobility (and hence size) range and a
TSI 3025 condensation particle counter (CPC) which grows particles (via condensation) to
optically detectable sizes. The SMPS software increases the classifier voltage in a pre-
determined manner so that particles within a 10-500 nm size range are pre-selected and
subsequently counted using the CPC. The software also integrates the particle number
distribution to enable calculation of the total number of particles emitted by the engine at
each test mode. Gaseous emissions are measured with CAI 600 series gas analyses. CO2,
NOX and CO concentrations can be measured by this gas analyser, whereas particulate mass
emissions are measured with a TSI 8530 Dust-Track II.
Chapter 5
177
As depicted in Fig. 5.1, three way valves to control the exhaust gas path ways are
employed. With this configuration, it is possible to measure both gaseous emissions and
particles before entering the reactor and after leaving it by changing the three way valve
directions. CO2 was used as a tracer gas in order to calculate the dilution ratio. CO2 was
measured from the dilution tunnel with dilution ratios being calculated using the following
equation:
(5-1)
Laboratory background CO2 measurements were made before the commencement of each
test session. Every concentration measured after dilution should be modified by using a
dilution ratio.
Fig. 5.1 Schematic diagram of plasma treatment system developed at QUT engine lab.
Chapter 5
178
5.3.2. DBD Reactor
A conventional dielectric barrier discharge reactor was designed for the experiments. Fig.
5.2 shows a schematic of the reactor. As illustrated in Fig. 5.2, it consists of two concentric
quartz tubes. Both tubes are 400 mm long and have a wall thickness of 1.5 mm. The outside
diameters of inner and outer quartz tubes are 20 mm and 25 mm, respectively. Exhaust gas
passes through the gap between these two quartz tubes. Based on pre-designed geometry, the
discharge gap is 1 mm. The DBD is connected to the pulsed power supply using internal and
external electrodes. The internal electrode is a copper cylinder and the external electrode is
made by a copper mesh that wraps the exterior part of the DBD. The electrodes are placed in
the middle of the DBD load with the length of 100 mm. Both tubes are fixed by two Teflon
caps at each end. Exhaust enters the reactor at the angle of 45 degree and flows throughout the
gap and leaves the reactor with the same angle.
(a)
(b)
Fig. 5.2 DBD reactor in Solidworks: a) Schematic view, b) Cross-sectional view.
5.3.3. Bipolar Pulsed Power Supply
Fig. 5.3 shows a circuit schematic diagram of the pulsed power supply. As illustrated, it is
based on the push-pull inverter topology. The push-pull inverter contains two switches that are
driven with respect to ground. This is the main advantage of the inverter. This topology uses a
centre-tapped transformer which is excited in both directions. A step up transformer is used to
boost the voltage and achieve galvanic isolation.
Chapter 5
179
The main reason to use a push-pull topology is to generate bipolar output voltage.
Applying voltage periodically build-up charges across the electrodes, which can results in
arcing. In order to sustained non-thermal plasma and prevent from arcing, bipolar pulse
generation can employed for clearing charges [50]. Employing lower number of switches is
another advantage of push-pull inverter. The two switches S1 and S2 are switched alternately
with a controlled duty ratio to convert input DC voltage into high frequency AC voltage
suitable for exciting the DBD load. Hence, the generated output voltage is bipolar.
Adding a DBD load turns the push-pull inverter into a resonant stage with approximately
sinusoidal output. The frequency of the semi-sinusoidal shape signal is determined by an L-C
circuit comprising of the transformer inductance and capacitances of DBD and the
transformer. The repetition rate can be used to adjust the power and by optimizing the
resonance it is possible to obtain high frequency semi-sinusoidal waveform. A typical
measured output voltage of the employed pulsed power supply is depicted in Fig. 5.4.
S2
S1
Vin
DB
DNA
NB
NS
CS
CS
NA = NB << NS
Fig. 5.3 Pulsed power supply circuit schematic diagram (push-pull inverter)
The first portion of the output voltage waveform is the resonant circuit dominated by the
magnetizing inductance of the transformer and the capacitances of the transformer and DBD.
The period of this signal is approximately 11.2 µs. The second one is the resonance
happening during the switches off-state between the leakage inductance and the capacitances
of the transformer and DBD. The period of this signal is equal to 88.8µs. As can be seen from
Chapter 5
180
the figure the repetition rate is set to 10 kHz. It is quite important to generate symmetrical
waveform regarding to clearing charge purpose and avoiding transformer saturation.
Fig. 5.5 shows the experimental hardware setup for the pulsed power supply. Here 1200 V
IGBT modules, SK75GB123, are used as power switches. Semikron Skyper 32-pro gate drive
modules are utilized to drive the IGBTs and provide the necessary isolation between the
switching-signal ground and the power ground. A Texas Instrument TMS320F28335 DSC
(Digital Signal Controller) is used for PWM signal generation. A centre-tapped step-up
transformer with an UU100 core 3C90 grade material ferrite from Ferroxcube, are designed
with NA = NB = 5 and NS = 293. Here a 470 pF capacitor (CS) is placed across each switch to
protect them against the voltage spikes. The output voltage is measured and captured using a
Pintek DP-22Kpro differential probe and RIGOL DS1204B oscilloscope, respectively.
Fig. 5.4 Typical measured output voltage of the employed pulsed power supply.
Fig. 5.5 Electrical hardware setup with the DBD load.
Chapter 5
181
5.4. Results and Discussion
5.4.1. Plasma effect on PM Size Distribution
The effect of plasma on emission treatment is considered for varied experiments based on
the aforementioned Cummins diesel engine. In all experiments engine speed and load are kept
constant at 40 kW (25 %load) and 2000 rpm, respectively. A portion of raw exhaust gas
directly from an iso-kinetic sampling port of the tailpipe was diluted with air and passed
through the reactor. Emissions concentration is measured before and after the applying pulse
to study the plasma effects. In addition, the median particle diameter which is another useful
parameter to study the effect of plasma technique is also measured. It is to be noted that all
illustrated results in each experiment have been obtained as an average over three consecutive
measurements.
The experiments were made by applying output voltage from 10 kVpp up to 19.44 kVpp.
However, the first effect of plasma was appeared at 15 kVpp following with optimum
operation at 17kVpp and finally high amount of small particle generation at 19.44 kVpp.
Hence, the measurements are reported in this section regarding to the mentioned three applied
voltage levels. Fig. 5.6 illustrates the measured results. The applied output voltages across the
DBD load for three different voltage levels of 15 kVpp, 17 kVpp, and 19.44 kVpp are
depicted in Fig. 5.6a. The measured output voltages show the rate of voltage rise of
2840V/µs.
The measured load currents are illustrated in Fig. 5.6b, which shows many narrow pulsed
current spikes occurring in each half-cycle of the applied voltage. The measured current at
19.44 kVpp comparing with the other applied voltages shows higher number of the micro-
discharges in the gap with much higher amplitude. This is due to the fact that the applied
voltage has reached to the value of the breakdown voltage, which depends on the gap
distance, dielectric material, repetition rate, and etc. Controlling the amplitude of current
discharges is quite important as it can directly affect the plasma reaction which is discussed
further. Fig. 5.6c shows the voltage stress across the switch during the switching transition.
As can be seen, due to employing a centre-tapped transformer, the peak-to-peak voltage
stress across the switch in a push-pull inverter is approximately two times of the input
voltage.
Chapter 5
182
(a)
(b)
(c)
Fig. 5.6 Measured electrical parameters: a) output voltages across the DBD load, b) output current, and c)
voltage stress across the switch.
Chapter 5
183
Fig. 5.7 DBD image.
Fig. 5.8 V-Q cyclogram of the DBD load as a basis of power consumption calculation.
In these experiments, the output voltage amplitude is controlled by changing the input DC
voltage between 72.4 V, 84.4 V, and 94.8 V. Under same situation, Fig. 5.7 shows an image
of DBD recorded at 19.44 kVpp. As can be seen, NTP is clearly occurring between the two
electrodes.
To measure the power consumption of the DBD load, the energy transferred to the DBD
load has been calculated by employing the Lissajous (V−Q) diagram [51, 52].To measure Q a
4nF capacitor is placed in series with the DBD reactor. Thus, by measuring the voltage across
the capacitor and multiplying it by its capacitance value it is possible to calculate Q. The
energy consumed by the DBD reactor for one cycle is calculated from the area of V−Q curve
for different experiments, where V is the measured voltage across the DBD reactor (see Fig.
5.8). Hence, by considering the employed repetition rate (frequency) it is possible to calculate
the average consumed power by the DBD load. It is to be noted that the series connected
capacitor is selected large enough to not to affect the DBD reactor capacitance. The relevant
equations are:
Chapter 5
184
dvtQW )( (5-2)
)(
)()(
tdV
tdQtC (5-3)
By substituting (3) in (2):
2
2
1)(CVdQ
C
tQW (5-4)
Therefore, considering (5-2) to (5-4) the averaged consumed power can be calculated as
below:
2
2
1)( CVfdvtQfWfPA (5-5)
The illustrated data in Fig. 5.8 clearly indicate that the DBD reactor power consumption
correspondingly increases with the applied voltage level. This can be also realized from (5-5),
which shows the relation between the power and the applied voltage. The calculated averaged
power consumption (PA) for the applied voltages of 15 kVpp, 17 kVpp, and 19.44 kVpp are
27.37 W, 36.54 W, and 55.17 W respectively. The higher averaged power at 19.44 kVpp can
be realized through the measured discharged current as it has occurred at higher amplitude
and higher number of the micro-discharges.
In the first experiment, the maximum voltage level (19.44 kVpp) is applied. To estimate
the deposition rate on the reactor surface, emissions in the reactor inlet (reactor inlet no pulse)
and reactor outlet (reactor outlet no pulse) without applying any pulse voltage are measured.
Finally, the pulsed power supply is applied across the DBD and the emissions in reactor
outlet are measured (reactor outlet with pulse). The same process is employed in all following
experiments.
Fig. 5.9 illustrates the particle size distribution at 19.44 kVpp. There is a considerable
amount of PM deposition on reactor surface which is likely related to small gap between the
tubes (1 mm). The median diameter in the reactor inlet is 70 nm, while in the reactor outlet is
decreased to 66 nm. This shows that larger particles deposited more on the reactor surface.
By applying pulsed power, as shown in this figure, the median diameter decreases
remarkably to 35 nm. This implies that lots of big particles have been oxidized or broken to
small particles by producing plasma inside the exhaust gasses at this voltage level.
Chapter 5
185
Fig. 5.9 Particle Size Distribution (2000rpm, 25%Load, 19.44 kVpp).
The peak value of particle number at the reactor inlet is around ⁄ .
This value declines to ⁄ at the reactor outlet due to deposition. By
applying pulsed power the graph peaks to 4.5 ⁄ which is approximately
four times bigger than the particle number at reactor outlet without any pulse. The number of
particles with diameter of less than 70 nm in the reactor outlet with applying pulse is higher
than the particle numbers at the reactor outlet without any pulse. This effect increases even
more at particle diameters less than 50 nm. At this level, the particle number at reactor outlet
surpasses the number of particles at reactor inlet. These findings imply that the 19.44 kVpp
pulse power at 10 kHz, increases the number of small particles considerably. The origin and
nature of these particles is still not clear and will be of interests in future investigations.
However, the effect of 19.44 kVpp on particle size can be understood regarding to the high
level of micro-discharges in the discharged current as depicted in Fig. 5.6b.
The effect of voltage level with 17 kVpp is considered in the second experiment. The
results obtained have been summarized in Fig. 5.10. The figure shows the median diameter
changes from 70 nm at reactor inlet to 78 nm at reactor outlet without any pulse, and then falls
to 75 nm at reactor outlet with applying pulse. This shows that the larger particles are
deposited and also removed by plasma selectivity compared to smaller particles. As can be
Chapter 5
186
seen, at this voltage level (17 kVpp) the number of small particles has not increased. This is an
important feature when compared with the previous experiment (19.44 kVpp).
The last experiment is conducted by applying 15 kVpp pulses. The measured results are
depicted in Fig. 5.11, which shows no growth in the number of small particles as well as
second experiment (17 kVpp). The maximum of PM concentration took place at around 71
nm. There is a small difference between PM concentrations with and without applying plasma.
Therefore, this level of voltage can remove particles at the same rate of deposition.
Comparing the values of median diameter in the reactor inlet and outlet shows that the
smaller particles are more likely to be deposited inside the reactor under the no pulse
condition. On the other hand, by applying the pulse the median diameter is in same range of
reactor inlet. There is no increase in the number of small particles at this voltage level. This
trend in median diameter variation is almost in complete agreement with the voltage of 17
kVpp.
Careful comparison of the reactor outlet particle size distribution, when there is no pulsed
power, indicates that the distribution in Fig. 5.9 shows a reduction in particle median
diameter, whereas Fig. 5.10 shows a slight increase in particle median diameter. Experiments
for Fig. 5.10 were conducted approximately 30 min after that of Fig. 5.9. Therefore, there is a
possibility that wall initial deposition occurred with larger particles and the later experiments
for Fig. 5.10 favoured slightly smaller particles due to the larger surface area of the wall and
the attraction of particles to deposited particles rather than the quartz wall alone.
Fig. 5.10 Particle Size Distribution (2000rpm, 25%Load, 17 kVpp).
Chapter 5
187
Fig. 5.11 Particle Size Distribution (2000rpm, 25%Load, 15 kVpp).
5.4.2. PM removal efficiency
PM removal can be calculated based on the following equation:
100)(
ionconcentratPMinlet
pulsewithionconcentratPMoutletionconcentratPMinletremovalPM (5-6)
Where the PM concentration unit is (particle/cm3) and PM removal is calculated for all
PM diameters.
Fig. 5.12 illustrates the PM removal at 19.44 kVpp. PM deposition on the reactor surface
increases with the PM size and gets to the maximum value of 70% removal at around 80 nm.
But below 80nm PM removal decreases again. For most of the particle sizes, PM deposition
on the reactor surface is more than 40%. When a 19.44 kVpp voltage is applied to the
reactor, PM removal efficiency reaches the value of 90% for larger particles. Removal
efficiency for particulate matter less than 80 nm is less than PM removal without applying
any pulse. This indicates undesirable operation of the plasma within this particle size range.
Also there is no removal for particles smaller than 50 nm. There is a high possibility that this
incense in particle numbers can be related to the following two factors: firstly, fragmentation
of larger particles by electron impact reactions or incomplete oxidation and secondly,
oxidation of gaseous exhaust emissions to particles by plasma generated ozone.
Chapter 5
188
Fig. 5.12 PM removal as a function of PM size (2000rpm, 25%Load, 19.44 kVpp)
Fig. 5.13 PM removal as a function of PM size (2000rpm, 25%Load, 17 kVpp)
Fig. 5.13 shows the PM removal at 17 kVpp which shows that PM removal by plasma is
more effective than deposition removal. This means that at this level of voltage, all deposited
particles and also some large particles inside the flow can be removed or oxidized. For
particles smaller than 35 nm, PM removal percentage by deposition is higher than PM
Chapter 5
189
removal by plasma. Apparently, smaller particles cannot be removed by plasma within this
range (<35nm). However another possibility can be dissociation of larger particles to smaller
ones by electron impact reactions.
PM removal efficiency at 15 kVpp is illustrated in Fig. 5.14. As can be seen the PM
removal for both graphs increases to an optimum value then decreases. The maximum PM
removals in reactor outlet with and without the pulse are 61% and 69% respectively. For
particles larger than the 60 nm diameter, PM removal when applying pulsed power is slightly
higher than PM removal without any pulse. However, for particles with smaller diameters,
these values are in the same ranges.
Regarding the results obtained, it can be concluded that the voltage level has an important
role in the size dependent removal efficiency. At 15 kVpp the particle size distribution has
been affected slightly. The PM removal without producing small particles can be improved
as the voltage increases to 17 kVpp. The increase in the number of small particles has been
noticed when the voltage level goes up to 19.44 kVpp, while larger particles have been
reduced considerably. By taking into account all the above mentioned features, the
experiment with 17 kVpp voltage shows better efficiency in terms of particle size distribution
for the given configuration.
Fig. 5.14 PM removal as a function of PM size (2000rpm, 25%Load, 15 kVpp)
Chapter 5
190
5.4.3. Plasma Effect on PM Mass Reduction
After studying the effect of different voltages on particle size distribution, in this section, the
effect of plasma on particle mass reduction is considered. In a similar way to the previous
sections three different voltage levels have been applied to the DBD load. All results obtained
have been summarized in Table 5.1. PM mass concentrations in the reactor inlet were 4.56
(mg/m3), 4.26 (mg/m
3) and 5.14 (mg/m
3) respecting to the variation of engine operating
conditions for the three tests conducted. These values decreased to 2.84(mg/m3), 3.68 (mg/m
3)
and 4.37 (mg/m3) at reactor outlet in the no pulse condition, respectively. This shows a
significant particle deposition inside the reactor. When the pulsed power is applied, plasma
PM removal occurred. This causes the PM mass concentration reduction of 43.9 %, 38.6% and
27.1% at 19.44 kVpp, 17 kVpp and 15 kVpp respectively.
The maximum PM mass reduction has been obtained when the voltage level is 19.44 kVpp.
However, according to the particle size distribution measurements, this voltage level increases
the number of small particles, which is not a desirable feature. The 17 kVpp applied voltage
shows a more suitable performance with good mass reduction of around 40% without any
increase in ultra-fine particle numbers. The 15 kVpp voltage level is found to be almost the
threshold breakdown voltage for the given configuration, below which no significant PM mass
reduction, PM removal, and PM size distribution which have not been affected too much.
Table 5.1 PM Mass Reduction at Different Voltage Levels
Applied Voltage
Measurement
19.44 kVpp 17 kVpp 15 kVpp
Reactor Inlet PM
Concentration (
)
4.56 4.26 5.14
Reactor Outlet No Pulse PM
Concentration(
)
2.84 3.68 4.37
Reactor Outlet By-Pulse PM
Concentration (
)
2.56 2.62 3.74
Plasma PM Removal
Efficiency (%)
43.9 38.6 27.1
Chapter 5
191
5.5. Conclusion
In this study the effect of non-thermal plasma obtained by applying high voltage pulses on
PM size distribution and PM mass reduction were investigated. It was found that NTP plasma
not only affects the PM mass concentration, but also changes the PM size distribution. At
very high voltage levels (here 19.44 kVpp), NTP was very effective for PM mass reduction.
However, PM mass reduction is not the only concern. It became clear that at high voltage
levels the number of ultra-fine particles increases significantly. Regarding the negative health
effects of tiny particles, the performance of plasma at such a high voltage levels is not
desirable. Considering the PM mass reduction and PM size distribution simultaneously, an
optimum voltage level of 17 kVpp at 10 kHz was found for the given configuration and
operating condition. Moreover the wall attachments of particulates are another important
parameter which should be considered in all experiments. It was found that wall attachments
are variable even without introducing any plasma. Therefore, particle deposition insides the
reactor and its effect on plasma PM removal should be considered with more details in future.
5.6. References
[1] C.-L. Song, F. Bin, Z.-M. Tao, F.-C. Li, and Q.-F. Huang, "Simultaneous removals of NOx,
HC and PM from diesel exhaust emissions by dielectric barrier discharges," Journal of
Hazardous Materials, vol. 166, pp. 523-530, 2009.
[2] A. Chaloulakou, M. I, and G. I, "Compliance with the annual NO2 air quality standard in
Athens. Required NOx levels an expected health implications," Atmos Environ, vol. 42, pp.
454–465, 2008.
[3] A. Mayer, H. Egli, J. Burtscher, T. Czerwinski, and D. Gehrig, "Particle size distribution
downstream traps of different design," SAE vol. Paper No. 950373 1995.
[4] L. Zhua, J. Yu, and X. Wang, "Oxidation treatment of diesel soot particulate on CexZr1−xO2,
," J. Hazard. Mater, vol. 140, pp. 205-210, 2007.
[5] Oberdörster G, Sharp Z, Atudorei V, Elder A, Gelein R, Kreyling W, et al., "Translocation of
inhaled ultrafine particles to the brain," Inhalation Toxicology, vol. 16, pp. 437–45, 2004.
[6] T. Matsumoto, D. Wang, T. Namihira, and H. Akiyama, "Energy Efficiency Improvement of
Nitric Oxide Treatment Using Nanosecond Pulsed Discharge," Plasma Science, IEEE
Transactions on, vol. 38, pp. 2639-2643, 2010.
[7] K. Takaki, M. A. Jani, and T. Fujiwara, "Removal of nitric oxide in flue gases by multi-point
to plane dielectric barrier discharge," Plasma Science, IEEE Transactions on, vol. 27, pp.
1137-1145, 1999.
[8] A. Mizuno, "Industrial applications of atmospheric non-thermal plasma in environmental
remediation," Plasma Physics and Controlled Fusion, vol. 49, p. A1, 2007.
[9] M. Saito, H. Hoshino, T. Furuhata, and M. Arai, "Continuous regeneration of an electrically
heated diesel particulate trap: Mechanism of particulate matter trapping and improvement of
Chapter 5
192
trapping efficiency," International Journal of Engine Research, vol. 11, pp. 127-136, April 1,
2010 2010.
[10] Plasma Exhaust Treatment [Online]. Available: http://www.dieselnet.com/tech/plasma.html
[11] P. Davari, F. Zare, and A. Ghosh, "A flexible solid-state pulsed power topology," in Power
Electronics and Motion Control Conference (EPE/PEMC), 2012 15th International, 2012, pp.
DS2b.12-1-DS2b.12-6.
[12] P. Davari, F. Zare, A. Ghosh, and H. Akiyama, "High-Voltage Modular Power Supply Using
Parallel and Series Configurations of Flyback Converter for Pulsed Power Applications,"
Plasma Science, IEEE Transactions on, vol. 40, pp. 2578-2587, 2012.
[13] S. Zabihi, F. Zare, G. Ledwich, A. Ghosh, and H. Akiyama, "A Novel High-Voltage Pulsed-
Power Supply Based on Low-Voltage Switch-Capacitor Units," Plasma Science, IEEE
Transactions on, vol. 38, pp. 2877-2887, 2010.
[14] A. A. El-Deib, F. Dawson, S. Bhosle, and G. Zissis, "Circuit-Based Model for a Dielectric
Barrier Discharge Lamp Using the Finite Volume Method," Plasma Science, IEEE
Transactions on, vol. 38, pp. 2260-2273, 2010.
[15] S. Tao, Z. Dongdong, Y. Yang, Z. Cheng, W. Jue, Y. Ping, et al., "A Compact Repetitive
Unipolar Nanosecond-Pulse Generator for Dielectric Barrier Discharge Application," Plasma
Science, IEEE Transactions on, vol. 38, pp. 1651-1655, 2010.
[16] B. Rahmani, S. Bhosle, and G. Zissis, "Dielectric-Barrier-Discharge Excilamp in Mixtures of
Krypton and Molecular Chlorine," Plasma Science, IEEE Transactions on, vol. 37, pp. 546-
550, 2009.
[17] K. Takaki, M. Shimizu, S. Mukaigawa, and T. Fujiwara, "Effect of electrode shape in
dielectric barrier discharge plasma reactor for NOx removal," Plasma Science, IEEE
Transactions on, vol. 32, pp. 32-38, 2004.
[18] N. Osawa and Y. Yoshioka, "Generation of low-frequency homogeneous dielectric barrier
discharge at atmospheric pressure," Plasma Science, IEEE Transactions on, vol. 40, pp. 2-8,
2012.
[19] H. Ghomi, N. N. Safa, and S. Ghasemi, "Investigation on a DBD Plasma Reactor," Plasma
Science, IEEE Transactions on, vol. 39, pp. 2104-2105, 2011.
[20] H. Ayan, G. Fridman, A. F. Gutsol, V. N. Vasilets, A. Fridman, and G. Friedman,
"Nanosecond-Pulsed Uniform Dielectric-Barrier Discharge," Plasma Science, IEEE
Transactions on, vol. 36, pp. 504-508, 2008.
[21] H. Piquet, S. Bhosle, R. Diez, and M. V. Erofeev, "Pulsed Current-Mode Supply of Dielectric
Barrier Discharge Excilamps for the Control of the Radiated Ultraviolet Power," Plasma
Science, IEEE Transactions on, vol. 38, pp. 2531-2538, 2010.
[22] T. Yamamoto, M. Okubo, K. Hayakawa, and K. Kitaura, "Towards ideal NO x control
technology using plasma-chemical hybrid process," Phoenix, AZ, USA, 1999, pp. 1495-1502.
[23] B. S. Rajanikanth and S. Rout, "Studies on nitric oxide removal in simulated gas
compositions under plasma-dielectric/catalytic discharges," Fuel Processing Technology, vol.
74, pp. 177-195, 2001.
[24] V. Ravi, Y. S. Mok, B. S. Rajanikanth, and H. C. Kang, "Temperature effect on hydrocarbon-
enhanced nitric oxide conversion using a dielectric barrier discharge reactor," Fuel
Processing Technology, vol. 81, pp. 187-199, 2003.
[25] B. S. Rajanikanth and V. Ravi, "Removal of nitrogen oxides in diesel engine exhaust by
plasma assisted molecular sieves," Plasma Science and Technology, vol. 4, pp. 1399-1406,
2002.
Chapter 5
193
[26] T. Yamamoto, B. S. Rajanikanth, M. Okubo, T. Kuroki, and M. Nishino, "Performance
evaluation of nonthermal plasma reactors for NO oxidation in diesel engine exhaust gas
treatment," IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, vol. 39, pp. 1608-1613,
2003.
[27] B. S. Rajanikanth and B. R. Sushma, "Injection of N-radicals into diesel engine exhaust
treated by plasma," Plasma Science and Technology, vol. 8, pp. 202-206, 2006.
[28] A. D. Srinivasan and B. S. Rajanikanth, "Nonthermal-plasma-promoted catalysis for the
removal of NOX from a stationary diesel-engine exhaust," IEEE TRANSACTIONS ON
INDUSTRY APPLICATIONS, vol. 43, pp. 1507-1514, 2007.
[29] A. D. Srinivasan, B. S. Rajanikanth, and S. Mahapatro, "CORONA TREATMENT FOR
NOX REDUCTION FROM STATIONARY DIESEL ENGINE EXHAUST IMPACT OF
NATURE OF ENERGIZATION AND EXHAUST COMPOSITION," 2009.
[30] V. Ravi, Y. S. Mok, B. S. Rajanikanth, and H. C. Kang, "Studies on nitrogen oxides removal
using plasma assisted catalytic reactor," Plasma Science and Technology, vol. 5, pp. 2057-
2062, 2003.
[31] B. S. Rajanikanth, A. D. Srinivasan, and B. A. Nandiny, "A cascaded discharge plasma-
adsorbent technique for engine exhaust treatment," Plasma Science and Technology, vol. 5,
pp. 1825-1833, 2003.
[32] B. S. Rajanikanth, S. Das, and A. D. Srinivasan, "Unfiltered Diesel Engine Exhaust
Treatment by Discharge Plasma: Effect of Soot Oxidation," Plasma Science & Technology,
vol. 6, pp. 2475-2480, 2004.
[33] B. S. Rajanikanth and V. Ravi, "DeNOx study in diesel engine exhaust using barrier
discharge corona assisted by v2O5/TiO2 catalyst," Plasma Science and Technology, vol. 6,
pp. 2411-2415, 2004.
[34] B. S. Rajanikanth and A. D. Srinivasan, "Pulsed plasma promoted adsorption/catalysis for
NOx removal from stationary diesel engine exhaust," IEEE Transactions on Dielectrics and
Electrical Insulation, vol. 14, pp. 302-311, 2007.
[35] A. D. Srinivasan and B. S. Rajanikanth, "Pulsed plasma treatment for NOx reduction from
filtered/unfiltered stationary diesel engine exhaust," New Orleans, LA, 2007, pp. 1893-1900.
[36] B. S. Rajanikanth, D. Sinha, and P. Emmanuel, "Discharge plasma assisted adsorbents for
exhaust treatment: A comparative analysis on enhancing NOx removal," Plasma Science and
Technology, vol. 10, pp. 307-312, 2008.
[37] B. S. Rajanikanth, S. Mohapatro, and L. Umanand, "Solar powered high voltage energization
for vehicular exhaust cleaning: A step towards possible retrofitting in vehicles," Fuel
Processing Technology, vol. 90, pp. 343-352, 2009.
[38] G. A. Stratakis, "Experimental investigation of catalytic soot oxidation and pressure drop
characteristics in wall flow diesel particulate filters,," Ph.D. Thesis, University of Thessaly,
Greece., 2004.
[39] V. Ramanathan, "Global dimming by air pollution and global warming by greenhouse gases,"
Nucleation and Atmospheric Aerosols vol. 6, pp. 473-483, 2007
[40] A. Seaton, W. MacNee, D. K, and D. Godden, " Particulate air pollution and acute health
effects," Lancet, vol. 345, pp. 176-178., 1995.
[41] A. Sydbom, A. Blomberg, S. Parnia, N. Stenfors, T. Sandstrom, and S. E. Dahlen, "Health
effects of diesel exhaust emissions," Eur respir J vol. 17, pp. 733-746, 2001.
Chapter 5
194
[42] C. M. omers, B. E. McCarry, F. Malek, and J. S. Quinn, " Reduction of particulate air
pollution lowers the risk of heritable mutations in mice," Science vol. 304, pp. 1008–1010,
2004.
[43] W. A. Majewski. Diesel exhaust particle size [Online]. Available:
http://www.dieselnet.com/tech/dpm_size.html
[44] R. O. McClellan, "Health Effects of Exposure to Diesel Exhaust Particles," Ann. Rev.
Pharmacol. Toxicol, vol. 27, pp. 279-300, 1989.
[45] J. H. Seinfeld, Air pollution: Physical and chemical fundamentals. New York: McGraw-Hill,
Inc., 1975.
[46] D. B. Kittelson, "Engines and nanoparticles: A review," J. Aerosol Sci., vol. 29, pp. 575-588,
1998.
[47] W. A. Majewski. Diesel particulate matter [Online]. Available:
http://www.dieselnet.com/tech/dpm.html
[48] A. Peters, H. E. Wichmann, T. Tuch, J. Heinrich, and J. Heyder, "Respiratory effects are
associated with the number of ultrafine particles," American journal of respiratory and
critical care medicine, vol. 155, pp. 1376-1383, 1997.
[49] N. C. Surawski, Z. D. Ristovski, R. J. Brown, and R. Situ, "Gaseous and particle emissions
from an ethanol fumigated compression ignition engine," Energy Conversion and
Management, vol. 54, pp. 145-151, 2012.
[50] Z. D. Ristovski, B. Miljevic, N. C. Surawski, L. Morawska, K. M. Fong, F. Goh, et al.,
"Respiratory health effects of diesel particulate matter," Respirology, vol. 17, pp. 201-212,
2012.
[51] J. Kriegseis, S. Grundmann, and C. Tropea, "Power consumption, discharge capacitance and
light emission as measures for thrust production of dielectric barrier discharge plasma
actuators," Journal of Applied Physics, vol. 110, pp. 013305-9, 07/01/ 2011.
[52] R. P. Mildren and R. J. Carman, "Enhanced performance of a dielectric barrier discharge
lamp using shortpulsed excitation," J. Phys. D.: Appl. Phys., vol. 34, pp. L1-L6, Jan. 2001.
195
Statement of Contribution of Co-Authors
The authors listed below have certified that:
1- They meet the criteria for authorship in that they have participated in the conception, execution, or
interpretation, of at least that part of the publication in their field of expertise;
2- They take public responsibility for their part of the publication, except for the responsible author who
accepts overall responsibility for the publications;
3- There are no other authors of the publication according to these criteria;
4- Potential conflicts of interest have been disclosed to (a) granting bodies, (b) the editor or publisher of
journals or other publications, and (c) the head of the responsible academic unit;
5- They agree to the use of the publication in the student’s thesis and its publication on the Australasian
Digital Thesis database consistent with any limitations set by publisher requirements.
In the case of this chapter:
Analysing DBD Plasma Lamp Intensity versus Power Consumption Using a Push-Pull Pulsed Power
Supply
Submitted to: 15th
European Conference on Power Electronics and Applications, France, Sep 2013.
Contributor Statement of contribution
Pooya Davari Proposed the initial idea. Designed and conducted simulation and data analysis.
Implemented the hardware setup, designed the control strategy, conducted
experimental verifications and wrote the manuscript
22 Feb 2013
Firuz Zare Proposed the initial idea and supervised implementation process, experimentation,
and writing the manuscript
Arindam Ghosh Aided data analysis and writing the manuscript
Supervisor Confirmation
I have sighted email or other correspondence from all Co-authors confirming their certifying authorship.
Prof. Arindam Ghosh
Name Signature Date: 22 Feb 2013
196
Chapter
6 . A n a l y s i n g D B D P l a s m a L a m p I n t e n s i t y
v e r s u s P o w e r C o n s u m p t i o n U s i n g a P u s h -
P u l l P u l s e d P o w e r S u p p l y
* School of Electrical Engineering and Computer Science, Queensland University of Technology,
GPO BOX 2434, Brisbane, Australia
† Danfoss Power Electronics, Graasten DK-6300, Denmark
Accepted in: 15th
European Conference on Power Electronics and Applications, France, Sep
2013.
6
Chapter 6
197
6.1. Acknowledgement
We thank Dr. Robert Carman of the Department of Physics and Astronomy at Macquarie
University, Sydney, Australia, for the helpful advice, and loan of the XeCl lamp which was
originally supplied by Prof. G. Zissis, LAPLACE Laboratory, Universite Paul Sabatier,
Toulouse, France.
Abstract- In this paper characteristic of a DBD (Dielectric Barrier Discharge) plasma lamp is
investigated based on the lamp intensity and power consumption. A pulsed power supply
with controllable parameters based on a push-pull converter is developed for lamp excitation
at different voltage levels and repetition rate. The experimentations were conducted for 28
different operating points with the frequency range of 2 kHz to 15 kHz at output voltage
levels of between 7.4 kVpp up to 13 kVpp. The obtained results show the feasibility of finding
an optimum operation point due to nonlinear behaviour of the DBD lamp.
6.2. Keywords
Pulsed power supply, Plasma, DBD Lamp, High voltage pulse.
6.3. Introduction
Dielectric barrier discharge (DBD) is a promising method in producing non-thermal
plasma, which is widely used in a variety of industrial applications. Recently, DBD lamps
have gained much attention as they are mercury free, easily scalable and simple to construct
[1-4]. DBD lamps are mostly filled with rare gases and rare gas-halides, which can provide
efficient scheme for generating incoherent UV (Ultra Violet) and VUV (Vacuum UV)
radiation [3]. The range of applications is broad owing to the variability of output wavelength
(88–310 nm) with the choice of gas fill.
DBD Lamps can be operated with continuous excitation or with pulsed excitation. Because
a pulsed discharge can operate at much higher peak voltages and peak currents for the same
average power as in a dc glow discharge, higher instantaneous sputtering, ionization and
excitation can be expected and hence better efficiencies [5, 6]. Recently, solid-state pulsed
power has gained more interest as it is compact, reliable, has a long lifetime and high
repetition rate. In the last decade, research and studies established the advantage of using
power electronics topologies in developing pulsed power supplies for variety of applications
Chapter 6
198
[6-10]. Therefore, a solid-state pulsed power supply can be a suitable choice for exciting a
DBD lamp as it provides controllable variables such as duty cycle, frequency, generated
voltage, and etc.
One of the important features of a DBD lamp is the light intensity. The light intensity is
quite important in sensitive applications such as in therapy of skin diseases and for
disinfecting the skin surface. Hence, increasing or decreasing generated UV intensity can
harm or not efficiently affect the under treatment area respectively. The light intensity can be
controlled by factors such as frequency (pulse repetition rate) and applied voltage. However,
another important issue which needs to take into account is the lamp power consumption [2].
Increment in power consumption not only reduces the system efficiency but also can lead to
short operational lifetime of the lamp. Thus, finding the relation between the lamp intensity
and power consumption in order to select a proper operating point is vital.
In this paper light intensity versus power consumption of a Xenon-filled coaxial DBD lamp
is analysed at different voltage level and repetition rates. The DBD lamp is excited using a
push-pull based pulsed power supply with the frequency range of 2 kHz to15 kHz and at
output voltage starting from 7.4kVpp up to 13kVpp. The light intensity is measured using a
UV detector and the power consumption is measured based on Lissajous V-Q diagram [2, 3]
at 28 different operating points. The obtained results not only depicted the performance of the
implemented pulsed power supply in driving the DBD lamp, but also the DBD lamp
illustrated a non-linear characteristic which can be advantageous in selecting an optimum
operating point.
6.4. Experimental Setup
6.4.1. DBD Lamp
Here a DBD lamp, as shown in Fig 6.1, is selected which is due to their particular interest as
they possess high efficiency [3]. The lamp tube has a coaxial geometry with a double
dielectric barrier, and is fabricated from UV grade fused silica tubing. The external electrodes
are wire/wire mesh yielding an active region of ~60mm length with a discharge gap of ~9mm
between the inner and outer dielectrics. The gas fill (sealed off) is a Xenon buffer at ~100mb
pressure, seeded with a low partial pressure (~1torr) of halogen (Cl) to yield XeCl excimers
upon discharge excitation which emit in the UV at 308m. Typical output irradiances from the
lamp are 1-10mW per cm2.
Chapter 6
199
6.4.2. Pulsed Power Supply
Fig 6.2 depicted a circuit schematic diagram of the developed pulsed power supply. As
illustrated, it is based on a push-pull inverter topology. The push-pull inverter contains two
switches that are driven with respect to ground. This is the main advantage of the inverter.
This topology uses a centre-tapped transformer which is excited in both directions. A step up
transformer is used to boost the voltage and achieve galvanic isolation.
Fig. 6.1 Dielectric barrier discharge lamp
In order to sustained NTP and prevent from arcing, bipolar pulse generation is employed
for clearing charges [11].The two switches S1 and S2 are switched alternately with a
controlled duty ratio to convert input DC voltage into high frequency AC voltage suitable for
exciting the DBD load. Hence, the generated output voltage is bipolar.
Adding a DBD load turns the push-pull inverter into a resonant stage with an
approximately sinusoidal output. The frequency of the semi-sinusoidal shape signal is
determined by an L-C circuit comprising of the transformer inductance and capacitances of
DBD and the transformer. The repetition rate can be used to adjust the power and by
optimizing the resonance it is possible to obtain high frequency semi-sinusoidal waveform.
The pulsed power supply was developed to provide complete control over output voltage
and repetition rate by means of regulating the input voltage and the duty cycle of power
switches. Fig. 6.3 shows the experimental hardware setup for the pulsed power supply. Here
1200V IGBT modules, SK75GB123, are used as power switches. Semikron Skyper 32-pro
gate drive modules are utilized to drive the IGBTs and provide the necessary isolation
between the switching-signal ground and the power ground. A Texas Instrument
TMS320F28335 DSC (Digital Signal Controller) is used for PWM signal generation. A
Chapter 6
200
centre-tapped step-up transformer with an UU100 core 3C90 grade material ferrite from
Ferroxcube, are designed with NA = NB = 5 and NS = 293. Here a 470 pF capacitor (CS) is
placed across each switch to protect them against the voltage spikes. To calculate the power
consumption a 4nF capacitor (Cm) is connected in series with the DBD lamp. Cm is selected
large enough to not to affect the DBD lamp capacitance. The output voltage is measured and
captured using a Pintek DP-22Kpro differential probe and RIGOL DS1204B oscilloscope,
respectively.
S2
S1
Vin DB
D L
am
p
NA
NB
NS
CS
CS
NA = NB << NS
Cm
Fig. 6.2 Pulsed power supply circuit schematic diagram (push-pull inverter)
Fig. 6.3 Electrical Hardware setup with the DBD load.
An example of a measured output voltage of the employed pulsed power supply at the output
voltage of 13kVpp is depicted in Fig. 6.4. The depicted bipolar voltage waveform is generated
by switching S1 and S2 alternately (see Fig. 6.4a). This part (the first portion) of the output
voltage waveform is the resonant circuit dominated by the magnetizing inductance of the
transformer and the capacitances of the transformer and DBD. The period of this signal is
approximately 11.5 µs. The second part is the resonance happening during the switches off-
state between the leakage inductance and the capacitances of the transformer and DBD.
Chapter 6
201
(a)
(b)
Fig. 6.4 Output voltage of the employed pulsed power supply: a) at 10 kHz, b) typical measured output voltages
at 10 kHz and 5 kHz.
As the depicted signal (Fig. 6.4a) was generated at 10 kHz (repetition rate), therefore the
period of the second portion is equal to 88.5µs. It is to be noted that for all the applied
voltages and repetition rates the bipolar pulses were generated with a fixed duration of 11.5
µs (see Fig. 6.4b). This implies the importance of repetition rate which differs in four types of
applied pulses and can results in delivering varied power level to the DBD lamp.
6.4.3. Measurements
The lamp intensity was measured using a UV-photodetector (see Fig 6.3). The photo-detector
is equipped with JIC157 which has a spectral range of 210...390 nm. The averaged intensity
was considered for evaluation regarding to the area of the captured intensity waveform and
employed frequency.
Chapter 6
202
Fig. 6.5 V-Q cyclogram of the DBD lamp as a basis of power consumption calculation.
To measure the power consumption of the DBD lamp, the energy transferred to the DBD
lamp has been calculated by applying the Lissajous (V −Q) diagram [2, 3]. To measure Q one
capacitor (Cm) is placed in series with the DBD lamp, as depicted in Fig 6.2. Thus, by
measuring the voltage across Cm and multiplying it by Cm value it is possible to calculate Q.
The energy consumed by the plasma for one cycle is calculated from the area of V−Q curve
for different experiments (see Fig 6.5). Hence, by considering the employed repetition rate
(frequency) it is possible to calculate the average consumed power by the DBD lamp. Due to
presence of noise during measurement the captured data is filtered before power consumption
estimation stage (see Fig 6.5). The relevant equations are:
∮ (6-1)
(6-2)
By substituting (6-2) in (6-1):
∮
(6-3)
Therefore, considering (6-1) to (6-3) the averaged consumed power can be calculated as
below:
∮
(6-4)
Chapter 6
203
6.5. Results and Discussion
To analysis the lamp characteristics the lamp intensity versus the lamp power consumption
is considered at different voltage level and operating frequency. Input voltage (Vin) and
repetition rate (fr) were assigned as controlling parameters to determine the plasma lamp
characteristics. The high voltage pulses were applied to the plasma lamp at seven different
voltage levels and four different repetition rates of 2 kHz, 5 kHz, 10 kHz, and 15 kHz (see
Table 6.1). The 28 operating points are selected based on combination of varied operating
frequencies and voltages to evaluate the behaviour of the DBD lamp at variety of power
levels and intensities. The starting point was selected as Vin = 36 V as the first sustainable
plasma was achieved at this point. The measured results are illustrated as averaged intensity
versus averaged power consumption (see Fig. 6.6). It is to be noted that the results are
normalized for a better comparison.
Fig. 6.6a depicted the applied voltage versus power consumption. The illustrated data
clearly indicate that the DBD lamp power consumption correspondingly increases with the
applied voltage level and the repetition rate. This can be also realized from (6-4), which
shows the relation between the power, repetition rate and the applied voltage. However, as
the power is proportional to the applied voltage squared, the applied voltage is much more
effective than the repetition rate on the power level.
Table 6.1 Considered voltages and frequencies for different experiments
Frequency
Output Voltage Level
2kHz
5kHz
10kHz
15kHz
Vin= 36V D11 D21 D31 D41
Vin= 40V D12 D22 D32 D42
Vin= 44V D13 D23 D33 D43
Vin= 48V D14 D24 D34 D44
Vin= 52V D15 D25 D35 D45
Vin= 56V D16 D26 D36 D46
Vin= 60V D17 D27 D37 D47
Chapter 6
204
(a)
(b)
Fig. 6.6 Comparing different operation points regarding to the varied applied voltage levels and repetition rates,
a) applied voltage versus plasma lamp power consumption, b) applied voltage versus plasma lamp intensity.
Fig. 6.7 The DBD lamp light intensity measured for varied applied voltage levels.
Despite the power consumption which is always desirable to be as low as possible, it is the
lamp intensity which is the main goal. Obtaining high intensity at low power consumption in
plasma lamp applications is one of the major concerns. The obtained results regarding to the
applied voltage versus the lamp intensity are illustrated in Fig. 6.6b. It is obvious that
Chapter 6
205
increasing the applied voltage results higher intensity (see Figs 6.6b and 6.7). However, the
obtained results indicate that not necessarily the higher intensity the high power consumption
is. In the other word, the intensity doesn’t have a linear relation with the power consumption.
The measured results in Fig. 6.6 depicted the possibility of achieving same or even higher
intensity at lower power consumption by increasing the repetition rate. For instance,
comparing D44 with D36 shows that, D44 has higher intensity than D36 while it has same
power consumption. Such behaviour can be realized by comparing D32 with D27 as well.
This can be due to the different slope of change regarding to the applied voltage and
consumed power.
Regarding to the aforementioned facts, it is feasible to find an optimum operation point
where the required intensity can be achieved at lower power consumption. This can be
obtained by operating at lower voltage levels but with the cost of higher repetition rate.
Reducing the generated voltage level while obtaining the required performance is always
desirable due to high voltage insulation and power switches limited breakdown voltage
issues.
6.6. Conclusion
In this paper, the behaviour of a DBD plasma lamp was analysed with respect to its
intensity and power consumption. A pulsed power supply based on the push-pull topology
was developed with controllable parameters to trigger the DBD lamp over a wide range of
operating conditions. The feasibility of optimum operation point due to non-linear
characteristics of the plasma lamp was concluded based on the results obtained. To ensure
that the following conclusions apply for different power levels of the DBD lamp, all
experiments have been conducted on 28 different operating points by combining different
repetition frequencies and applied voltages.
6.7. References
[1] H. Ghomi, N. N. Safa, and S. Ghasemi, "Investigation on a DBD Plasma Reactor," Plasma
Science, IEEE Transactions on, vol. 39, pp. 2104-2105, 2011.
[2] J. Kriegseis, S. Grundmann, and C. Tropea, "Power consumption, discharge capacitance and
light emission as measures for thrust production of dielectric barrier discharge plasma
actuators," Journal of Applied Physics, vol. 110, pp. 013305-9, 07/01/ 2011.
Chapter 6
206
[3] R. P. Mildren and R. J. Carman, "Enhanced performance of a dielectric barrier discharge
lamp using shortpulsed excitation," J. Phys. D.: Appl. Phys., vol. 34, pp. L1-L6, Jan. 2001.
[4] U. N. Pal, P. Gulati, N. Kumar, M. Kumar, M. S. Tyagi, B. L. Meena, et al., "Analysis of
Discharge Parameters in Xenon-Filled Coaxial DBD Tube," Plasma Science, IEEE
Transactions on, vol. 39, pp. 1475-1481, 2011.
[5] H. Ayan, G. Fridman, A. F. Gutsol, V. N. Vasilets, A. Fridman, and G. Friedman,
"Nanosecond-Pulsed Uniform Dielectric-Barrier Discharge," Plasma Science, IEEE
Transactions on, vol. 36, pp. 504-508, 2008.
[6] L. Redondo and J. F. Silva, "Solid State Pulsed Power Electronics," in POWER
ELECTRONICS HANDBOOK DEVICES, CIRCUITS,ANDAPPLICATIONS, M. H.Rashid,
Ed., Third Edition ed: Elsevier, 2011.
[7] P. Davari, F. Zare, and A. Ghosh, "Parallel and series configurations of flyback converter for
pulsed power applications," in Industrial Electronics and Applications (ICIEA), 2012 7th
IEEE Conference on, 2012, pp. 1517-1522.
[8] P. Davari, F. Zare, and A. Ghosh, "Flexible Solid-State Pulsed Power Topology," presented at
the 15th International Power Electronics and Motion Control Conference, EPE-PEMC 2012
ECCE Europe, Novi Sad, Serbia, Sep 2012.
[9] P. Davari, F. Zare, A. Ghosh, and H. Akiyama, "High-Voltage Modular Power Supply Using
Parallel and Series Configurations of Flyback Converter for Pulsed Power Applications,"
Plasma Science, IEEE Transactions on, vol. 40, pp. 2578-2587, 2012.
[10] L. M. Redondo and J. F. Silva, "Flyback Versus Forward Switching Power Supply
Topologies For Unipolar Pulsed-Power Applications," Plasma Science, IEEE Transactions
on, vol. 37, pp. 171-178, 2009.
[11] R. A. Scholl, "Power Supplies for Pulsed Plasma Technologies: State-Of-The-Art And
Outlook,"" Advanced Energy Industries, Inc, pp. 1-8, 1999.
207
Statement of Contribution of Co-Authors
The authors listed below have certified that:
1- They meet the criteria for authorship in that they have participated in the conception, execution, or
interpretation, of at least that part of the publication in their field of expertise;
2- They take public responsibility for their part of the publication, except for the responsible author who
accepts overall responsibility for the publications;
3- There are no other authors of the publication according to these criteria;
4- Potential conflicts of interest have been disclosed to (a) granting bodies, (b) the editor or publisher of
journals or other publications, and (c) the head of the responsible academic unit;
5- They agree to the use of the publication in the student’s thesis and its publication on the Australasian
Digital Thesis database consistent with any limitations set by publisher requirements.
In the case of this chapter:
Power Electronic Converters for High Power Ultrasound Transducers
Published in proceedings of: The 7th IEEE Conference on Industrial Electronics and Applications, ICIEA
2012, Singapore.
Contributor Statement of contribution
Negareh Ghasemi Proposed the initial idea and designed and implemented the pulsed power supply.
Planned and conducted experimental verification and wrote the manuscript
Firuz Zare Proposed the initial idea and evaluation process. Supervised the data analysis,
experimentation and writing the manuscript
Pooya Davari Aided data analysis, experimental verification, and writing the manuscript
22 Feb 2013
Christian Langton Aided data analysis and writing the manuscript
Peter Weber Provided the piezoelectric transducer and general information about the ultrasound
applications and writing the manuscript
Arindam Ghosh Aided data analysis and writing the manuscript
Supervisor Confirmation
I have sighted email or other correspondence from all Co-authors confirming their certifying authorship.
Prof. Arindam Ghosh
Name Signature Date: 22 Feb 2013
208
Chapter
7 . P o w e r E l e c t r o n i c C o n v e r t e r s f o r H i g h
P o w e r U l t r a s o u n d T r a n s d u c e r s
* School of Electrical Engineering and Computer Science, Queensland University of Technology,
GPO BOX 2434, Brisbane, Australia
Science and Engineering Faculty, Queensland University of Technology, GPO BOX 2434,
Brisbane, Australia
Ultrasound-System-Development, Fraunhofer Institute for Biomedical Engineering, Germany
Presented and Published at: The 7th
IEEE Conference on Industrial Electronics and
Applications, ICIEA 2012, Singapore, July 2012.
7
Chapter 7
209
Abstract- Piezoelectric transducers convert electrical energy to mechanical energy and play a
great role in ultrasound systems. Ultrasound power transducer performance is strongly related
to the applied electrical excitation. To have a suitable excitation for maximum energy
conversion, it is required to analyze the effects of input signal waveform, medium and input
signal distortion on the characteristics of a high power ultrasound system (including
ultrasound transducer). In this research, different input voltage signals are generated using a
single-phase power inverter and a linear power amplifier to excite a high power ultrasound
transducer in different mediums (water and oil) in order to study the characteristics of the
system. We have also considered and analyzed the effect of power converter output voltage
distortions on the performance of the high power ultrasound transducer using a passive filter.
7.1. Keywords
Component; Power Converter, Piezoelectric Transducer Excitation
7.2. Introduction
Much research has been conducted on piezoelectric behavior since 1880, the year that Pierre
and Jacque Curie discovered the phenomenon of piezoelectricity. The critical behavior of a
piezoelectric device is encapsulated in its resonant frequencies and the most efficient way to
find the critical piezoelectric specifications is to analyze its impedance frequency response [1].
IEEE Standard on Piezoelectricity introduced the basic equivalent circuit model characterizing
a piezoelectric ceramic near the resonant frequency which is known as Van Dyke Model. This
model is often adapted to model electromechanical resonance characteristics of crystal
oscillators.
The Van Dyke Model is a parallel connection of a series RLC representing mechanical
damping, mass, and elastic compliance and a capacitor representing the electrostatic
capacitance between the two parallel ceramic plates [2]. When a piezoelectric ceramic is
mounted to a mechanical structure, a loaded piezoelectric ceramic experiences multiple
resonances, a circuit model (Fig. 7.1 (a)) for a wide frequency range with multiple resonant
frequencies can be employed to model the behavior of a loaded piezoelectric ceramic.
Ultrasound systems are used in different industrial and medical applications. According to
various applications, an ultrasound system can be used in low (1-100 W) and high (0.1- 50
kW) power and frequency ranges. For instance, in biomedical applications a high frequency
Chapter 7
210
ultrasound system is used for diagnosis (low power) or therapeutic application (high power).
In order to generate ultrasound wave, piezoelectric transducer is a key part of the ultrasound
system which converts electrical to mechanical energy. A most important issue in exciting a
high power ultrasound transducer is quality and shape of the electrical signal which drives the
power transducer [3-9]. It is important to generate a high quality power signal at its resonant
frequency with low distortion to attain the highest energy conversion. Different methods are
introduced to generate a suitable signal to drive a power transducer such as radio-frequency
linear amplifiers and switched mode power converters [6].
A main advantage of switched mode power converters compare to power amplifiers is its
high efficiency at high power operation. Fig. 7.1 (b) and Fig. 7.1 (c) show a power converter
connected to a piezoelectric transducer as a load and the output voltage levels of a power
converter respectively. Multilevel converters are suitable power converters to drive power
transducers due to their attractive ability to generate a high quality power waveform with low
harmonic distortion and voltage stress [10-12].
In order to drive a power transducer with an appropriate signal, it is essential to study and
analyze the impedance of a piezoelectric transducer at different frequency and power ranges.
Usually, a network analyzer is used to measure piezoelectric transducer impedance and its
resonant frequencies in frequency domain. It is not possible to analyze and study the
characteristics of a high power transducer using a network analyzer due to the fact that a
network analyzer operates at low power[5]. Therefore, some tests have been carried out in this
research to study the performance and behavior (linear or nonlinear) of ultrasound system at
high power range which are presented in the next sections.
Chapter 7
211
(a)
(b)
(c)
Fig. 7.1 (a) Van Dyke Model, (b) a power converter, (c) output voltage of power converter.
7.3. Experimental Procedure
In this research work we have used a power ultrasound transducer which has some resonant
frequencies below 100 kHz and high energy conversion happens at those resonant frequencies.
Table 7.1 shows a summary of all test conditions which have been carried out at different
mediums and input voltages. In the first two tests, sinusoidal voltage waveforms are generated
by a linear power amplifier to drive the piezoelectric transducers. In these two tests, nonlinear
characteristics of the high power ultrasound system is analyzed using superposition method.
Chapter 7
212
In the other tests, a high power single phase inverter is used to generate high voltage and
high frequency signals (square wave) to analyze behavior of ultrasound system when it is
driven by non-sinusoidal signals.
In test 1 and test 2, a power amplifier (OPA 459) is connected to a step-up transformer to
generate a high voltage signal to excite a high power transducer as shown in Fig. 7.2. Two
transducers (same type) are placed in a container (exactly opposite to each other) and one
transducer is excited by the electrical signal as a transmitter and the other one converts the
generated ultrasound signal to electrical energy at the other side of the container as a receiver.
Since the performance of the piezoelectric transducer is highly dependent on its excitation,
generating a proper sinusoidal signal with low order harmonics is required. If the frequency of
the excitation signal is same as the resonant frequency of the transducer, the highest energy
conversion will be attained. A signal generator and a high power amplifier (OPA 459) are used
to generate a high power signal (up to 8 A at 30 Volts) but a high frequency step-up
transformer is used to increase the output voltage level. The piezoelectric has some resonant
frequencies around 39 kHz. We have generated two sinusoidal signals at 39 kHz and 61 kHz
in two different tests. In test 1, the magnitudes of the sinusoidal signals are adjusted at 15 V
and 30 V and we have excited the first transducer at 39 kHz and 61 kHz separately and have
measured the output voltage of the second transducer in time domain. In order to check the
quality of the input and output voltages at 15 V and 30 V, the signals are shown in frequency
domain (Fig. 7.3).
The test results show that the output voltage measured by the second transducer is not
proportional to the input voltage. This test verifies that the ultrasound system has nonlinear
behavior at its resonant frequencies. In order to study the performance of the ultrasound
system more, a superposition law is used as a key factor. Based on this principle, if the
ultrasound system has a linear characteristics, the responses of the system with two sinusoidal
signals (as follow) should be same:
a) Separately excited by two signals at 39 kHz and 61 kHz and adding the output signals
b) Simultaneously excited by two signals at 39 kHz and 61 kHz
Chapter 7
213
Table 7.1 Test conditions and setups
Test # Medium Excitation System Input Voltage
Magnitude (peak)
Frequency of Input
Signal (s)
Test 1 water
Signal Generator, a
power amplifier and a
high frequency
transformer
15 V and 30 V 39 kHz and 61 kHz
Test 2 water
Signal Generator, a
power amplifier and a
high frequency
transformer
15 V + 30 V
(time domain) 39 kHz and 61 kHz
Test 3 oil three-level inverter
50 V, 100 V, 200 V
and
300 V
39 kHz
Test 4 oil
three-level inverter and
a tube between two
transducers
50 V, 100 V, 200 V
and
300 V
39 kHz
Test 5 oil three-level inverter and
a filter
50 V, 100 V, 200 V
and
300 V
39 kHz
Test 6 water three-level inverter
50 V, 100 V, 200 V
and
300 V
39 kHz
Fig. 7.2 A block diagram of a lab prototype for test 1 and test 2.
Chapter 7
214
(a)
(b)
(c)
(d)
Fig. 7.3 (a) input signals at 39 kHz (b) output signals at 39 kHz (c) input signals at 61 kHz (d) output signals at
61 kHz
Therefore, in test 2, the same transducer is excited by two signals at 39 kHz and 61 kHz
which are added together at the input side of the power amplifier. We expect the output
voltages of the ultrasound system in two different tests should be exactly same as each other if
the system has a linear characteristics. The test result for each voltage level is shown in Fig.
7.4 (a). In order to compare this test results with the previous one, we have added the output
voltage results of test 1 at 39 kHz and 61 kHz in time domain and the results are shown in Fig.
7.4(b). The output voltage of each test at 15 V and 30 V are shown in Fig. 7.4 (c) and Fig. 7.4
(d) and it is clear that the output voltages (separately and simultaneously excited) are not
same. A difference between these test results shows the ultrasound system has a nonlinear
characteristics when the input voltage and power are increased.
Chapter 7
215
(a)
(b)
(c)
(d)
Fig. 7.4 (a) Test 1: summation of two output signals (at 39 kHz & 61 kHz) for Vin=15V and Vin=30 V
(b) Test 2: two output signals for Vin=15V and Vin=30 V. (c) Comparing the results of test 1 and test 2 for
Vin=15 V. (d) Comparing the results of test 1 and test 2 for Vin=30 V.
The results of Fig. 7.4 show that the ultrasound system does not obey the superposition
principle and it has nonlinear behavior as the output voltages of the two tests at 30 V are not
same. In order to study the nonlinearity of the ultrasound system at higher power and different
mediums, different tests have been carried out using a single phase inverter, generating a
square wave uni-polar voltage waveform. A laboratory setup of this configuration is shown in
Fig. 7.5.
A coupling box of laboratory setup is filled of oil or water for different tests in this research.
A block diagram of the setup of test 3 is shown in Fig. 7.6. In this test, a power converter
generates a square wave signal (uni-polar modulation) at different voltage levels (50V, 100V,
200V and 300V) and at 39 kHz including some harmonics. In order to compare the quality of
all input voltages, we have normalized all input voltage at 50 Volts (dividing all input voltages
by factors of 1, 2, 4 and 6, respectively) and the input voltage waveforms are shown in
frequency domain (Fig. 7.7 (a)). Then, we have measured the output voltage of the second
transducer at different excitation voltages and the results are shown in Fig. 7.7 (b). Similar to
Chapter 7
216
the input voltages, we have divided the output voltages by the same factors (1, 2, 4 and 6,
respectively) in order to compare the output voltages. The results show that the ultrasound
system has nonlinear characteristics at different voltage levels and the output voltages are not
same when they are normalized.
Fig. 7.5 The experimental setup.
Fig. 7.6 A block diagram of test 3.
Chapter 7
217
(a)
(b)
Fig. 7.7 The results of test 3 (a) input signals and (b) output signals.
According to the applied input voltages, an ultrasound wave is generated and transferred to
the second transducer. But some of the generated wave will be attenuated due to interaction
with inhomogeneous material. The velocity and attenuation of generated wave are highly
dependent on the mechanical and structural properties of the medium [13, 14]. To study the
effect of attenuation of generated ultrasound wave, a tube is placed between the two
transducers in which ultrasound wave is guided from the first transducer to the second
transducer. It is expected that the tube reduces the propagation of generated wave and thus
increases the intensity of the wave at the second transducer. A block diagram of this
configuration is illustrated in Fig. 7.8.
Similar to the previous test, we have normalized all input and output voltages in order to
check the quality of the input voltages and compare the output voltages in frequency domain.
As is shown in Fig. 7.9 the output voltage magnitudes are increased compare to the previous
Chapter 7
218
test result (Fig. 7.7 (b)) but there are still differences in the output voltage magnitudes due to
the nonlinear behaviour of the system at those frequencies.
The output voltage of the power inverter generates voltage stress (dv/dt) across the
piezoelectric transducer. This voltage with the capacitive characteristics of the piezoelectric
transducer can generate significant current spikes which increases losses and high frequency
noise. Since the input voltage distortion can tend to deteriorate the output voltage quality and
can increase the power dissipation, a 990 μH inductor as a filter is placed between the power
converter and the transducer to reduce the input voltage distortion across the first transducer.
Fig. 7.10 shows a block diagram of the setup.
The added filter reduces the amplitude of the output voltage in frequency domain (Fig. 7.11)
compared to the test results shown in Fig. 7.7 but this filter has not a significant effect on the
characteristics of the ultrasound system and it has still a nonlinear characteristics at those
frequencies which is shown in Fig. 7.11 (b).
In order to study the effects of medium on the ultrasound system characteristics, we have
performed a new test similar to test 3 but with water (instead of oil). A block diagram of the
setup is shown in Fig. 7.12.
Fig. 7.8 A block diagram of test 4.
Chapter 7
219
(a)
(b)
Fig. 7.9 The results of test 4 (a) input signals and (b) output signals.
Fig. 7.10 A block diagram of test 5.
Similar to the previous test results, when the input and output voltages are normalized at 50
V, there are still significant differences between the output voltages at different frequencies
shown in Fig. 7.13. The nonlinear behavior of a high power ultrasound system is obvious in
this figure due to mismatch of the responses of the ultrasound system to the normalized input
voltages. The nonlinearity of the ultrasound system has been shown in different cases where a
high power ultrasound transducer was excited by a sinusoidal and a pulse voltage waveform.
Chapter 7
220
(a)
(b)
Fig. 7.11 The results of test 5 (a) input signals and (b) output signals.
Fig. 7.12 A block diagram of test 6.
Chapter 7
221
(a)
(b)
Fig. 7.13 The results of test 6 (a) input signals and (b) output signals.
7.4. Conclusion
In this research a high power transducer is driven by different input signals and the
performance of ultrasound system is studied and analyzed in several frequencies. According to
the first and the second test results, it is obvious that the ultrasound system has nonlinear
behavior at high voltage and high power. In order to analyze the effects of a) input signal
waveform, b) medium and c) input signal distortion on characteristics of an ultrasound system,
several tests with different setups are carried out. According to the test results of this study and
research work, it is concluded that a high power ultrasound system has nonlinear
characteristics with respect to the input voltage magnitude. The results verify that the
nonlinear characteristics of a high power ultrasound system exists in different medium. It
means that a high power ultrasound system may be saturated when the input voltage
magnitude is increased and the piezoelectric ceramic plates cannot vibrate proportional to the
input voltage.
Chapter 7
222
7.5. References
[1] H. Walter, L. Karl, W. Wolfram, Piezoelectricity. New York: Springer, 2008.
[2] N. E. R. Centre, "Charactrestic of Piezoeletric Transducer," 2001-2011.
[3] S. Ozeri, D. Shmilovitz, "High frequency resonant inverter for excitation of piezoelectric
devices," in IEEE Power Electronics Specialists Conf., Jun 2008.
[4] S. B. Yaakov, N. Krihely "Modeling and driving piezoelectric resonant blade elements," in in
Nineteenth Annual IEEEApplied Power Electronics Conf. , 2004, pp. 1733-1739.
[5] A. A. Vives. (2008). Piezoelectric Transducers and Applications (2 ed.). Available:
http://QUT.eblib.com.au/patron/FullRecord.aspx?p=372502
[6] T. S. Chu, G. T. Clement, "A Harmonic Cancellation Technique for an Ultrasound
Transducer Excited by a Switched-Mode Power Converter," IEEE TRANS. ON
ULTRASONICS, FEROELECTRICS, AND FREQUENCY CONTROL, vol. 55, Feb 2008.
[7] T. Suzuki, H. Ikeda, H. Yoshida, S. Shinohara, "Megasonic transducer drive utilizing
MOSFET DC-to-RF inverter with output power of 600 W at 1 MHz," IEEE Trans. on
Industrial Electronics, vol. 46, pp. 1159-1173, 1999.
[8] K. Agbossou, J. L. Dion, S. Carignan, M. Abdelkrim, A. Cheriti, "Class D amplifier for a
power piezoelectric load," IEEE Trans. on Ultrasonics, Ferroelectrics and Frequency
Control, vol. 47, pp. 1036-1041, 2000.
[9] P. Fabijanski, R. Lagoda, "Series resonant converter with sandwich-type piezoelectric
ceramic transducers," in Proc. IEEE International Industrial TechnologyConf., 1996, pp.
252-256.
[10] A. Nami, F. Zare, A. Ghosh, "Asymmetrical DC link voltage configuration for a diode-
clamped inverter," IEEJ Transactions on Industry Applications (Denki Gakkai Ronbunshi. D,
Sangyo Oyo Bumonshi), vol. 130, pp. 195-206, 2010.
[11] A. Nami, F. Zare, A. Ghosh, F. Blaabjerg, "A Hybrid Cascade Converter Topology With
Series-Connected Symmetrical and Asymmetrical Diode-Clamped H-Bridge Cells," IEEE
Transactions on Power Electronics, vol. 26, pp. 51-65, 2011.
[12] A. A. Boora, A. Nami, F. Zare, A. Ghosh, F. Blaabjerg, "Voltage-Sharing Converter to
Supply Single-Phase Asymmetrical Four-Level Diode-Clamped Inverter With High Power
Factor Loads," IEEE Transactions on Power Electronics, , vol. 25, pp. 2507-2520, October
2010.
[13] M. Willatzen, "Ultrasound transducer modeling-received voltage signals and the use of half-
wavelength window layers with acoustic coupling layers," IEEE Trans. on Ultrasonics,
Ferroelectrics and Frequency Control, vol. 46, pp. 1164-1174, 1999.
[14] A. Lanata, E. P. Scilingo, R. Francesconi, D. De Rossi, , "Performance Analysis and Early
Validation of a Bi-modal Ultrasound Transducer," in in 28th Annual International IEEE
Medicine and Biology Society Engineering Conf., 2006, pp. 1858-1861.
223
Statement of Contribution of Co-Authors
The authors listed below have certified that:
1- They meet the criteria for authorship in that they have participated in the conception, execution, or
interpretation, of at least that part of the publication in their field of expertise;
2- They take public responsibility for their part of the publication, except for the responsible author who
accepts overall responsibility for the publications;
3- There are no other authors of the publication according to these criteria;
4- Potential conflicts of interest have been disclosed to (a) granting bodies, (b) the editor or publisher of
journals or other publications, and (c) the head of the responsible academic unit;
5- They agree to the use of the publication in the student’s thesis and its publication on the Australasian
Digital Thesis database consistent with any limitations set by publisher requirements.
In the case of this chapter:
Improving the Efficiency of High Power Piezoelectric Transducers for Industrial Applications
Published in: IET Science, Measurement, and Technology, vol.6, no.4, pp.213-221, July 2012.
Contributor Statement of contribution
Pooya Davari Proposed the initial idea. Designed and conducted simulation and data analysis.
Implemented the hardware setup, designed the control strategy, conducted
experimental verifications and wrote the manuscript
22 Feb 2013
Negareh Ghasemi Aided data analysis, experimental verifications and writing the manuscript
Firuz Zare Proposed the initial idea and supervised implementation process, experimentation,
and writing the manuscript
Peter O’Shea Aided data analysis, designing the algorithm and writing the manuscript
Arindam Ghosh Proposed the initial idea. Designed and conducted simulation and data analysis.
Implemented the hardware setup, designed the control strategy, conducted
experimental verifications and wrote the manuscript
Supervisor Confirmation
I have sighted email or other correspondence from all Co-authors confirming their certifying authorship.
Prof. Arindam Ghosh
Name Signature Date: 22 Feb 2013
224
Chapter
8 . I m p r o v i n g t h e E f f i c i e n c y o f H i g h P o w e r
P i e z o e l e c t r i c T r a n s d u c e r s f o r I n d u s t r i a l
A p p l i c a t i o n s
* School of Electrical Engineering and Computer Science, Queensland University of Technology,
GPO BOX 2434, Brisbane, Australia
Published in: IET Science, Measurement, and Technology, vol.6, no.4, pp.213-221, July
2012.
8
Chapter 8
225
Abstract: Most high power ultrasound applications are driven by 2-level inverters. However,
the broad spectral content of the 2-level pulse results in undesired harmonics which can
decrease the performance of the system significantly. On the other hand, it is crucial to excite
the piezoelectric devices at their main resonant frequency in order to have maximum energy
conversion. Therefore a high quality, low distorted power signal is needed to excite the high
power piezoelectric transducer at its resonant frequency. This paper proposes an efficient
approach to develop the performance of high power ultrasonic applications using multilevel
inverters along with a frequency estimation algorithm. In this method, the resonant frequencies
are estimated based on relative minimums of the piezoelectric impedance frequency response.
The algorithm follows the resonant frequency variation and adapts the multilevel inverter
reference frequency to drive an ultrasound transducer at high power. Extensive simulation and
experimental results indicate the effectiveness of the proposed approach.
8.1. Introduction
Wide spread research has been conducted on piezoelectric transducer applications since
1880, the year that Pierre and Jacque Curie discovered the phenomenon of piezoelectricity [1,
2]. Piezoelectric transducers convert electric power to acoustic power and vice versa, with
most of the applications to date being low power ones. In the last decade, however, high
power ultrasound applications have gained significant importance [3-8]. These types of
applications have great potential in chemical and bio-technology processing, specifically for
enhancing chemical reaction kinetics and new reaction pathways. Such enhancements allow
changing the production from batch processing to continuous flow processing, thereby
reducing investment and operational costs.
Improvement of ultrasound systems has very significant environmental implications,
particularly in the area of renewable energy (bio-mass and bio-fuel), waste-water treatment
and biomedical applications. High power ultrasound technology is already being used to
address these areas of need, but to a very limited extent due to the very inefficient nature of the
state of the art in energy conversion.
The critical behavior of a piezoelectric device is encapsulated in its resonance frequencies
due to its maximum transmission performance at these frequencies [1, 2, 9-11]. Hence, ideal
scenario is to have a sinusoidal excitation signal at the resonant frequency. But at high power
and high voltage, which is the focus area of this paper, generating pure sinusoidal signals is
not possible.
Chapter 8
226
The most efficient way to generate power signals is to use a 2-level power inverter.
Specifically, switch mode inverters are used for piezoelectric high power applications due to
their high power density, efficiency, low cost and size compared to conventional linear power
supplies [5, 12, 13]. However, the harmonics present in the output waveform produce
undesired side bands which are not suitable in many applications. Moreover, they also cause
unnecessary power dissipation which reduces the efficiency of the power converter [8, 14].
On the other hand, for ultrasound applications which operate at high fundamental frequency
power converters based on a PWM strategy with high switching frequency index (fs/fo) cannot
be a practical solution. In such applications, the maximum possible fundamental frequency of
a converter is restricted by the switching transients of each power switching event and the
number of switching events per cycle.
In this regard, to reduce the number of switching transients and eliminate the undesired
harmonics, the most effective way is to use multilevel inverters [15-18]. Through these
inverters, it is possible to produce quasi-sine waves with low total harmonic distortion at high
power. Multi-level inverters can increase the quality and efficiency of the high voltage supply
compared to the conventional 2-level inverters. This permits the semiconductor devices to
operate at lower switching frequencies with higher efficiency as well as lower voltage stress
across switches and loads which minimize electromagnetic and ultrasound noise emissions.
Fig. 8.1 depicts typical voltage waveforms and their harmonic spectra. It can be seen that the
harmonics are depressed for multilevel inverters such that they can be easily filtered out from
the frequency range of interest.
To improve the performance of the piezoelectric transducer for high power applications, in
addition to the multilevel converter, the device needs to be excited at its resonant frequency.
Piezoelectric devices typically have multiple resonant frequencies, but only the major resonant
frequency is generally targeted for excitation in practice. Structural and environmental
changes of a piezoelectric system can affect variations in the resonant frequencies [2, 13].
Therefore, it is important to estimate the main resonant frequency in order to maintain
efficient system operation. Therefore, the multilevel converter needs to be adapted with a
suitable frequency estimation algorithm.
Chapter 8
227
Fig. 8.1 Effect of using multi-level waveform in harmonic elimination.
The most effective way to find the resonant frequencies of a piezoelectric transducer is by
evaluating its impedance frequency response [19-28]. A minimum in the impedance response
corresponds to a resonant frequency, fr. The impedance frequency response is the ratio of the
voltage spectrum to the current spectrum. To calculate the piezoelectric impedance response, a
voltage source needs to be applied to the device as an excitation signal and current needs to be
measured simultaneously. In order to obtain the response of the device for a specific range of
frequencies, the excitation signal should cover the entire frequency range. The idea of
calculating the piezoelectric impedance is inspired by the general concept of performing
system identification (i.e. finding the system transfer function [29]). It is therefore possible to
benefit from the existing knowledge base of excitation signals for system identification.
Considering the above mentioned features, a broad-band excitation signal is the most
appropriate candidate [29, 30].
Chapter 8
228
DC/AC
Inverter
AC
Current Sensor
Piezoelectric
Frequency
Estimation
Fig. 8.2 Exploiting online resonance frequency estimation in high power applications of piezoelectric devices.
This paper advocates the advantage of using multilevel inverters along with an efficient
frequency estimation algorithm (as depicted in Fig. 8.2) to develop the performance of the
high power ultrasonic applications. In the proposed algorithm, a 1 kHz rectangular pulse with
10% duty cycle is applied to the device and the current is measured as the response. Then the
captured data is cropped to retain only one cycle of the applied input voltage and
corresponding output current. In the next step, FFT (Fast Fourier Transform) is applied to both
the signals and the impedance response is found as a ratio of the voltage to current transforms.
Finally, the relative minimum values are estimated and sorted according to both impedance
derivative and impedance magnitude. The FFT is used because it can be computed relatively
efficiently (with order NlogN operations), thus enabling real-time operations. Finally, the
multilevel inverter reference frequency is updated with the estimated resonance frequency.
The proposed method has been evaluated in simulations (using an electrical circuit model) and
experimentally (using two piezoelectric devices). The results obtained indicate the efficiency
and high performance of the proposed method.
8.2. Methodology and Approach to Estimate Resonance
Frequency
8.2.1. Excitation Signal
Piezoelectric devices have different resonance frequencies which are sometimes described as
vibration modes. To identify these frequencies, the excitation signal needs to be wide-band. In
the proposed approach, a sequence of rectangular pulses is applied to the device. The pulse
widths are 0.1ms and the pulses are reapplied every 1ms. It is quite easy to generate the
Chapter 8
229
selected pulse stream and the pulses occur fast enough to enable the system to create regular
updates of the resonance frequency online.
Fig. 8.3 (a) shows the frequency response of a typical rectangular pulse. As can be seen the
energy in the spectrum is well spread, but varies considerably in intensity as a function of
frequency. The frequency response of the output current will be moderated by this input
spectrum since it is not possible to study the piezoelectric characteristic by just looking at the
response to the rectangular pulse. One needs to compute the impedance response by forming
the quotient of the input and output responses.
Fig. 8.3(b) and (c) show a sample of the captured voltage and current after applying the
proposed excitation signal. As can be seen, an imperfect pulse is obtained in a practical
situation. This imperfection actually proves to an advantage because the practical spectrum
does not suffer from the problem of having zero energy at some frequency components (see
Fig. 8.3 (d)).
8.2.2. Estimating Impedance
The response due to the excitation signal (also known as the ‘residual vibrations’) is given
by:
n
k
kkk tatfSinAty1
)exp()2()( (8-1)
where, n is the number of resonance frequencies, Ak is the amplitude, fk is the frequency and
ak is the damping coefficient of the thk resonance frequency.
A current sensor is used to capture the response (residual vibrations) of the device. The
current and voltage across the piezoelectric are captured simultaneously. To get the best
results, one cycle of the excitation pulse, along with the corresponding current response, needs
to be extracted from the captured signals. The starting point of the voltage waveform is
specified on the leading edge of the pulse. Then, from a knowledge of the sampling rate and
the length of the signal (which is 1ms) the end point is determined. Based on these starting and
end points the voltage and current waveforms are cropped from the captured signals (see Fig.
8.3 (b) and (c) for sample captured signals).
Chapter 8
230
(a)
(b)
(c)
(d)
Fig. 8.3 a) Ideal pulse frequency response, b) Captured voltage c) Captured response (current), d) Practical pulse
frequency response.
Chapter 8
231
After cropping, the FFT of the both signals are calculated as follows:
1,...,1,0,)()(
1
1
2
NkenxkX
N
n
N
nki
(8-2)
where x(n) and X(k) are the discrete inputs and outputs respectively. To find the power
spectrum of the voltage and current signals the FFT outputs are multiplied by their conjugates
as per equation (8-3) and finally the impedance is calculated based on equation (8-4).
N
XXPx
.
(8-3)
NN
FffP
fPfP s
i
vz /)
2:0(,
)(
)()( (8-4)
where Fs denotes the sampling frequency and N is the number of FFT points.
8.2.3. Extracting Resonant Frequencies
The resonant frequencies correspond to the local (or relative) minimums of the piezoelectric
impedance. The relative minimums of a function are the points where that the slope of the
tangent changes from – to +. Fig. 8. 4(a) shows a power spectrum, )( fPzof an impedance and
the change of slope in the resonant frequency. At point A, )( fPz is equal to zero. Moreover,
immediately to the left of this point at point B the slope is negative while at point C the slope
is positive.
Motivated by the above, the derivative of )( fPz is calculated and all the points where )( fPz
changes sign from – to + are extracted. As the final step, at the extracted frequencies, a sorting
is performed based on the magnitude of )( fPz. The main resonant frequency is the one with the
lowest magnitude.
The procedure described above is depicted as an algorithm flowchart in Fig. 8.4(b). Within
that flowchart R is a constant defining the number of resonant frequencies needing to be
extracted. The algorithm can be used both for offline and online systems. It is important to
note that, repeating the proposed algorithm, while applying a repetitive pulse and averaging,
results in an increase of the estimation accuracy.
Chapter 8
232
(a)
Input
captured
signals
Finding
starting and
end points
Cropping the
signals
Saving as
relative
minimum
Sorting using
amplitude
Yes
No
Finding the
minimum
amplitude
Number of extracted
frequencies = R
No
Yes
(b)
Fig. 8.4 Extracting resonant frequencies: a) change of slope at relative minimum (resonant frequency), b)
flowchart of the proposed algorithm.
Chapter 8
233
8.2.4. Noise Issues
In practical systems the presence of noise is inevitable. Fortunately, at resonant frequencies,
the impedance of the device is minimum and current increases significantly. The signal-to-
noise ratio is therefore relatively high. Noise is nonetheless an issue, and noise removal is
therefore required.
Several approaches have been introduced for noise reduction. One of the most common
ways to reduce the noise is to use an anti-aliasing filter [31]. This type of filter is used before
sampling to limit the bandwidth of a signal. A simple anti-aliasing filtering is proposed for use
on the analog signal before sampling. This is in keeping with the goal to devise a simple
method which is easy to implement and fast enough to follow the resonant frequency
variations. In addition to a low-pass filter, it is also important to choose the sampling
frequency in accordance with the Nyquist theorem [30] and to ensure that there is a good
coupling at sampling points.
It is recommended that a simple RC low-pass filter be used between the current sensor and
the capturing device. The cut-off frequency of the filter is selected according to:
ff
cCR
f2
1 (8-5)
8.3. Simulation and Experiment
In this section, in order to show the performance of the proposed method in frequency
estimation, the proposed algorithm is evaluated via both simulations and experimentation. The
effect of using a multilevel inverter is shown in the last part as a practical evaluation using
ultrasound interface. The proposed frequency estimation algorithm is compared with the three
key types of alternative impedance analysing methods. These alternative methods are:
a) The traditional method: In this method several single frequency sine waves from 30 kHz
to 80 kHz have been applied separately to the piezoelectric. The frequency step size was
set to 1 kHz.
b) The Network Analyser method: An R&S ZVL3 vector network analyser has been used
to obtain the impedance frequency response of a piezoelectric device in the frequency
range of 30 kHz to 80 kHz.
Chapter 8
234
c) The unit step and white noise excitation method: In order to compare the proposed
method with different wide-band excitation signals, the impedance frequency behaviour
has been obtained based on step response and white noise.
Step Pulse: From system identification theory, it is known that the Fourier transform
of the impulse response (h(t)) of a system gives the system transfer function. Since
generating an impulse ( )(t ) at high power is highly impractical a step (u(t)) is
preferable. Here a 1ms step is applied to the device and as discussed in the previous
section the impedance response can be obtained by dividing the Fourier transform of
the output into the Fourier transform of the input.
White Noise: One of the most popular excitation signals used in system identification
is white noise [29, 30]. As white noise theoretically has a constant amount of energy
per frequency band, it is possible to simply look at the captured current to find the
resonant frequencies. To account for the fact that white noise spectra are not always
perfectly flat in practice, the impedance response is calculated in the same way as the
other two methods – by dividing the Fourier transform of the output into Fourier
transform of the input.
The software that was used for simulation and analysis of experimental results was Matlab
7.10. The sampling frequency in both simulation and experimental results was 2 MHz and the
number of FFT points, N, was set to 1024. Here two different kinds of piezoelectric devices
were used (Type A and Type B). Type A has three dominant resonant frequencies while Type
B has just one. For the simulation and experimental evaluations the piezoelectric devices were
immersed in housing containing water.
8.3.1. Simulation Results
In order to simulate and compare the proposed method with other methods, a circuit model
of the piezoelectric device was needed. The electrical circuit model [26] of a piezoelectric
transducer is shown in Fig. 8.5(a). The parameters in the electrical circuit model were
determined for the experimental piezoelectric device by performing measurements on a
network analyser. These values are presented in Table 8.1.
Chapter 8
235
Rp
Rs
R1
C0
C1
L1
R2
C2
L2
R3
C3
L3
(a)
(b)
Fig. 8.5 Simulation results for piezoelectric Type A: a) circuit model, b) impedance response.
The accuracy of the circuit model was further evaluated by comparing simulated and
measured impedance responses from a network analyser. Fig. 8.5(b) shows the impedance
responses of the piezoelectric device obtained from real measurements and from simulation.
As can be seen from the figure only three resonant frequencies have been modelled. With the
model established the next step was to compare the proposed method with the alternatives via
simulations. To this end the resonant frequencies for all methods were calculated based on the
model in Fig. 8.5(a) and are presented in Table 8.2.
Chapter 8
236
Table 8.1 Values of the Components in the Electrical Circuit Model
Non-resonant part First resonant mode (70.19 kHz)
C0(F) Rs )( Rp )( R1 )( L1(H) C1(F)
9103 10 5101 492.3 410665 1110006.9
Second resonant mode (48.6 kHz) Third resonant mode (38.81 kHz)
R2 )( L2(H) C2(F) R3 )( L3(H) C3(F)
640.0795 410834 0110184.2 582.744 410483 01103.414
Table 8.2 Estimated Resonant Frequencies in Simulation
Method
Resonant Frequency(kHz) Fr1 Fr2 Fr3
Impedance Measurement 70.19 48.6 38.81
Impulse Response (Step Excitation) 70.31 48.83 39.06
White Noise 70.31 48.83 39.06
Proposed 70.31 48.83 39.06
8.3.2. Experimental Results
Two different piezoelectric devices Type A and Type B were considered for experimental
evaluation. As was the case with the simulation testing, the impedance of both devices was
obtained in the frequency range of 30 kHz to 80 kHz. For the proposed method a power
converter was used for generating pulses. For the traditional and broadband excitation
methods a G5100A function waveform generator was used as the signal generator source and
the signals were amplified using an OPA549. Fig. 8.6 shows the experimental setup for the
proposed method. All signals were captured using a RIGOL DS1204B oscilloscope. As the
frequency was between 30 kHz and 80 kHz, the cut-off frequency of the filter was set to 100
kHz. Hence, according to equation (8-5), capacitor and resistor values of Cf =1nF and Rf
=1.59K respectively were selected for the low pass filter. The measured impedance
frequency behaviour of the Type A piezoelectric device for all methods is shown in Fig.
8.7(a). Table 8.3 also shows the extracted frequencies for the three strongest resonant
frequencies.
Chapter 8
237
Fig. 8.7(b) shows the impedance frequency response of the second piezoelectric device. The
estimated frequencies for the first three resonant frequencies are shown in Table 8.4. As can
be seen from the measured results, the white noise method did not result in accurate estimation
of Fr2 and Fr3 compared to the other methods. The main reason is that in each experiment the
level of power in white noise changes randomly over time, and may even go to zero at
particular frequencies. Therefore for resonances which are not especially strong (as was the
case for Fr2 and Fr3 in the Type B device) the results are unreliable.
It should be noted that the electrical circuit model of the piezoelectric device is a simple
model which is not able to perfectly model the piezoelectric device nonlinearity, time
variations and high frequency behaviour. That is why there are small differences between the
simulation and test results. The advantages and drawbacks of the various methods are
summarized in Table 8.5. The proposed method offers simplicity and high performance in
addition to its ability to be used for online systems.
Fig. 8.6 Experimental setup for the proposed method.
Chapter 8
238
(a)
(b)
Fig. 8.7 Experimental results obtained for the piezoelectric devices impedance response: (a) Type A, (b)
Type B.
Table 8.3 Estimated resonant frequencies of the Type A piezoelectric device
Method
Resonant Frequency(kHz) Fr1 Fr2 Fr3
Network Analyser 70.18 48.62 38.82
Impulse Response (Step Excitation) 70.31 48.83 39.06
White Noise 70.31 48.83 39.06
Traditional 70 49 39
Proposed 70.31 48.83 39.06
Chapter 8
239
Table 8.4 Estimated resonant frequencies of the Type B piezoelectric device
Method
Resonant Frequency(kHz) Fr1 Fr2 Fr3
Network Analyser 78.08 47.48 66.94
Impulse Response (Step Excitation) 78.13 46.88 66.41
White Noise 78.13 44.92 68.36
Traditional 78 47 67
Proposed 78.13 46.88 66.41
Table 8.5 Survey on advantages and drawbacks of mentioned methods
Method Advantages Drawbacks
Traditional Easy to implement
Low expense
Direct frequency output
Circuit stability required
Labour intensive
Not applicable for online process
Network Analyzer Provides complete impedance
frequency response
High accuracy due to calibration
High expense
Not applicable for online process
Not applicable on high power converter
Slow due to the complex measurement
procedure
Impulse Response
(Step Excitation)
Provides impedance frequency
response
Easy to generate with power
converter
Can be generated at high power
Can be applied for online process
Slower than the proposed method (because the
applied step is longer than the pulse).
Sensitive to the noise
White Noise
Excitation
Provides impedance frequency
response
Can be applied for online process
Not easy to generate especially at high power
Not possible to generate with power converter
Sensitive to the noise
Proposed Provides impedance frequency
response
Easy to generate with power
converter
Can be generated at high power
Can be applied for online processes
Sensitive to the noise
Chapter 8
240
8.3.3. Ultrasound Interface
The performance of the proposed frequency estimation algorithm was evaluated in the
previous sections, but as already mentioned, exciting the device at its resonant frequency is not
enough to achieve maximum power conversion. The excitation signal harmonics also need to
be considered.
In this section the advantage of exciting a piezoelectric transducer using a multi-level
waveform at the resonant frequency compared with a uni-polar waveform is illustrated. For
the comparison, one multi-level waveform and one uni-polar waveform were generated with
peak to peak voltage of 120V at 39 kHz (Type A device resonant frequency).
To perform the evaluation, one pair of the Type A piezoelectric transducers was placed face
to face as sender and receiver. For the first experiment, a unipolar pulse was applied to one of
the piezoelectric devices and the voltage across the other one was captured. Fig. 8.8(a) shows
the applied voltages. To show their influence on piezoelectric devices the frequency responses
of applied and captured voltages are illustrated in Fig. 8.8 (b). As can be seen from Fig. 8.8(b)
the captured response contains several harmonics, due to the excitation signal having much
energy away from the fundamental frequency.
For the next test, a multi-level waveform was applied to the piezoelectric device. As can be
seen from Fig. 8.8(c), harmonic levels are attenuated significantly. This was due to the use of a
multi-level waveform which damped the harmonics in the vast area around the fundamental
frequency.
Comparing the Fig. 8.8(b) with Fig. 8.8(c), illustrates that the maximum energy is achieved
at the resonant frequency. Higher efficiency is obtained for multi-level signal. In particular the
method is quite effective for reducing the harmonic content. In addition, the presence of
harmonics not only adversely affects frequency sensitive applications, but also causes an
increase in temperature and an increase in power loss. Moreover, when using filters for
attenuating remaining harmonics the filter cost and size decreases when multi-level topology
is employed.
Chapter 8
241
(a)
(d)
Input voltage (unipolar)
(b)
Input voltage (multilevel)
(e)
Output voltage (unipolar)
(c)
Output voltage (multilevel)
(f)
Fig. 8.8 Obtained results for the ultrasound interface: (a) applied uni-polar pulse at 39 kHz in time domain, (b)
frequency response of the applied uni-polar pulse (input signal), (c) frequency response of the output signal when
the input is a uni-polar pulse, (d) applied multi-level pulse at 39 kHz in time domain, (e) frequency response of the
applied multi-level pulse (input signal), (f) ) frequency response of the output signal when the input is a multi-level
pulse.
8.4. Conclusions
In this paper, a new method is proposed for improving the efficiency of high power
ultrasound applications. It has been seen that the resonant frequencies vary under different
Chapter 8
242
system conditions. Moreover, the advantage of exciting the piezoelectric device at the exact
resonant frequency using multi-level waveform generation has been illustrated. An algorithm
which extracts the resonant frequencies and updates the system is therefore needed. The
proposed method has been compared with different methods and excitation signals. The
comparative results have been conducted based on experimentation and simulation. The
results obtained demonstrated the high performance of the proposed method.
8.5. References
[1] W. Heywang, K. Lubitz, and W. Wersing, Piezoelectricity. New York: Springer, 2008.
[2] C. Steinem and A. Janshoff, Piezoelectric Sensors. New York: Springer, 2007.
[3] T. Hall and C. Cain, "A Low Cost Compact 512 Channel Therapeutic Ultrasound System For
Transcutaneous Ultrasound Surgery," AIP Conference Proceedings, vol. 829, pp. 445-449,
05/08/ 2006.
[4] Y. Kui, K. Uchino, X. Yuan, D. Shuxiang, and L. Leong Chew, "Compact piezoelectric
stacked actuators for high power applications," Ultrasonics, Ferroelectrics and Frequency
Control, IEEE Transactions on, vol. 47, pp. 819-825, 2000.
[5] R. Li, N. Frohleke, and J. Bocker, "DESIGN AND IMPLEMENTATION OF A POWER
INVERTER FOR A HIGH POOWER PEIZOELECTRIC BRAKE ACTUATOR IN
AIRCRAFTS," presented at the 9th Brazilian Power Electronics Conference, 2003.
[6] S. Priya, "High power universal piezoelectric transformer," Ultrasonics, Ferroelectrics and
Frequency Control, IEEE Transactions on, vol. 53, pp. 23-29, 2006.
[7] A. M. Sanchez, M. Sanz, R. Prieto, J. A. Oliver, P. Alou, and J. A. Cobos, "Design of
Piezoelectric Transformers for Power Converters by Means of Analytical and Numerical
Methods," Industrial Electronics, IEEE Transactions on, vol. 55, pp. 79-88, 2008.
[8] C. Volosencu, "Methods for Parameter Estimation and Frequency Control of Piezoelectric
Transducers, Automation Control - Theory and Practice," in Automation Control - Theory and
Practice, A. D. Rodić, Ed., ed: InTech, 2009, pp. 115-136.
[9] E. Dallago and A. Danioni, "Resonance frequency tracking control for piezoelectric
transformer," Electronics Letters, vol. 37, pp. 1317-1318, 2001.
[10] D. Huijuan, W. Jian, Z. Hui, and Z. Guangyu, "Measurement of a piezoelectric transducer's
mechanical resonant frequency based on residual vibration signals," presented at the
Information and Automation (ICIA), 2010 IEEE International Conference on, 2010.
[11] T. Saar, O. Martens, M. Reidla, and A. Ronk, "Chirp-based impedance spectroscopy of piezo-
sensors," presented at the Electronics Conference (BEC), 2010.
[12] K. Agbossou, J. L. Dion, S. Carignan, M. Abdelkrim, and A. Cheriti, "Class D amplifier for a
power piezoelectric load," Ultrasonics, Ferroelectrics and Frequency Control, IEEE
Transactions on, vol. 47, pp. 1036-1041, 2000.
[13] C. Kauczor and N. Frohleke, "Inverter topologies for ultrasonic piezoelectric transducers with
high mechanical Q-factor," in Power Electronics Specialists Conference, 2004. PESC 04.
2004 IEEE 35th Annual, 2004, pp. 2736-2741 Vol.4.
Chapter 8
243
[14] T. Sai Chun and G. T. Clement, "A harmonic cancellation technique for an ultrasound
transducer excited by a switched-mode power converter," Ultrasonics, Ferroelectrics and
Frequency Control, IEEE Transactions on, vol. 55, pp. 359-367, 2008.
[15] A. Nami and F. Zare, "Multilevel Converters in Renewable Energy Systems," in Renewable
Energy, T. J. Hammons, Ed., ed: InTech, 2009, pp. 271-296.
[16] J. Rodriguez, L. Jih-Sheng, and P. Fang Zheng, "Multilevel inverters: a survey of topologies,
controls, and applications," Industrial Electronics, IEEE Transactions on, vol. 49, pp. 724-
738, 2002.
[17] F. Zare and G. Ledwich, "A hysteresis current control for single-phase multilevel voltage
source inverters: PLD implementation," Power Electronics, IEEE Transactions on, vol. 17,
pp. 731-738, 2002.
[18] F. Zare and G. Ledwich, "A New Predictive Current Control Technique for Multilevel
Converters," presented at the TENCON IEEE Region 10 Conference, 2006.
[19] J. George K Lewis, George K Lewis, Sr, and William Olbricht, "Cost-effective broad-band
electrical impedance spectroscopy measurement circuit and signal analysis for piezo-materials
and ultrasound transducers," MEASUREMENT SCIENCE AND TECHNOLOGY, vol. 19, pp.
1-13, 2008.
[20] I. Getman and S. Lopatin, "Matching of series and parallel resonance frequencies for
ultrasonic piezoelectric transducers," presented at the Proc. Int. Symp. Applications of
Ferroelectrics.
[21] J. S. Kim, K. Choi, and I. Yu, "A new method of determining the equivalent circuit
parameters of piezoelectric resonators and analysis of the piezoelectric loading effect,"
Ultrasonics, Ferroelectrics and Frequency Control, IEEE Transactions on, vol. 40, pp. 424-
426, 1993.
[22] K. Kin Wing, H. L. W. Chan, and C. L. Choy, "Evaluation of the material parameters of
piezoelectric materials by various methods," Ultrasonics, Ferroelectrics and Frequency
Control, IEEE Transactions on, vol. 44, pp. 733-742, 1997.
[23] Y. Y. Lim, S. Bhalla, and C. K. Soh, "Structural identification and damage diagnosis using
self-sensing piezo-impedance transducers," Smart Materials and Structures, vol. 15, pp. 987-
995, 2006.
[24] A. L. Lopez-Sanchez and L. W. Schmerr, "Determination of an ultrasonic transducer's
sensitivity and impedance in a pulse-echo setup," Ultrasonics, Ferroelectrics and Frequency
Control, IEEE Transactions on, vol. 53, pp. 2101-2112, 2006.
[25] V. Loyau and G. Feuillard, "Relationship between electrical impedance of a transducer and its
electroacoustic behavior: Measurement without primary source," Journal of Applied Physics,
vol. 100, pp. 034909-034909-7, 2006.
[26] G. Mingjie and L. Wei-Hsin, "Studies on the circuit models of piezoelectric ceramics,"
presented at the Proc. Int. Conf. Information Acquisition, 2004.
[27] T. S. O. Märtens, M. Min, R. Land, M. Reidla, "Fast Impedance Spectroscopy of
Piezosensors for Structural Health Monitoring," ELECTRONICS AND ELECTRICAL
ENGINEERING, vol. 7, pp. 31-34, 2010.
[28] L. W. Schmerr, A. Lopez-Sanchez, and R. Huang, "Complete ultrasonic transducer
characterization and its use for models and measurements," Ultrasonics, vol. 44, pp. e753-
e757, 2006.
[29] R. Pintelon and J. Schoukens, "Basic Choices in System Identification," in System
Identification: A Frequency Domain Approach, ed: IEEE Press and John Wiley, 2001, pp.
351-375.
Chapter 8
244
[30] P. Davari and H. Hassanpour, "Designing a new robust on-line secondary path modeling
technique for feedforward active noise control systems," Signal Processing, vol. 89, pp. 1195-
1204, 2009.
[31] B. C. Baker, "Anti-aliasing, analog filters for data acquisition systems," in Application Note
no. AN699, Microchip Technology Inc., 1999.
245
Chapter
9 . C o n c l u s i o n s a n d F u r t h e r R e s e a r c h
9
Chapter 9
246
9.1. Conclusions
In the past decades, high power converters have penetrated into more and more applications,
playing as key role in industry. Regarding this, covering wide range of all related areas is not
practical, therefore in this thesis two main areas were considered. The main intention of this
research project was pulsed power technology. Investigating pulsed power area aimed at
applying power electronics techniques in order to develop solid-state pulsed power supplies
to meet application demands. On the other hand, high power ultrasound application regarding
to the high frequency issues was aimed as the second aspect of this PhD research. Three main
research objectives were identified as below:
Improving solid-state pulsed power supplies regarding to the low power rating and
operating speed limits of the current switching devices.
Considering load characteristics and effect of pulsed power supply
Increasing the efficiency of high power ultrasound applications
Considering the mentioned objectives, this dissertation set out to investigate power
converters at system and application levels for the two areas of pulsed power and high power
ultrasound. At system level different topologies were proposed to improve the system
efficiency regarding to the existing device’s limits and load demands, and at application level
the performance of the developed method was evaluated using real-world applications. In
each investigated area, the studies were started at system level and ended up at application
stage.
9.1.1. Improving solid-state pulsed power supplies regarding to the
low power rating and operating speed limits o f the current
switching devices
For the pulsed power area, employing different combination of flyback converter was
considered as the first step. The flyback converter was selected as it can generate high level
of the output voltages using a low input voltage, provides isolation and gives the ability of
controlling the input energy. To take advantage of current and voltage sharing, two flyback
configurations were considered. The proposed idea depicted the ability to increase the
performance of a pulsed power supply in terms of output voltage level and rate of rise based
on the simulation results. The preliminary concept of parallel and series configurations are
introduced in Chapter 2.
Chapter 9
247
To evaluate the proposed methods, laboratory prototypes were implemented with ability of
generating up to 4kV. The experimentations demonstrated the performance of the proposed
method in boosting up the voltage level and rate of rise comparing with the single module.
The further analysis and experimentations are depicted in Chapter 3. Analysing varied
operating conditions and effect of the load, illustrated that this technique can be utilized in
high impedance applications with the advantage of generating both DC and AC voltages.
The idea along with the simulation results are published in a conference paper entitled
“Parallel and Series Configurations of Flyback Converter for Pulsed Power Applications” at
The 7th
IEEE Conference on Industrial Electronics and Applications in July 2012 in
Singapore. The hardware implementation and experimental results are presented as a part of a
journal paper entitled “High Voltage Modular Power Supply Using Parallel and Series
Configurations of Flyback Converter for Pulsed Power Applications” published in IEEE
Transaction on Plasma Science Journal, Oct 2012.
The achieved outcomes illustrated that the series configuration is more practical. However,
two series modules were not enough to provide required voltage level and rate of rise. To
investigate the possibility of extending the proposed idea and benefiting from modularity
concept, a laboratory prototype based on 10 series flyback modules was implemented. This
research work is presented in Chapter 3. The capability of generating up to 40kV with
improved rate of rise (ten times comparing to a single module) completely proved the
effective performance of the proposed method. The analysis and experimentations were
published as a part of a journal paper entitled “High Voltage Modular Power Supply Using
Parallel and Series Configurations of Flyback Converter for Pulsed Power Applications”
published in IEEE Transaction on Plasma Science Journal, Oct 2012.
To preserve the modularity concept while making the pulsed power supply load independent
inspired the idea of employing switch-capacitor units supplying by a current source. The
ability of supplying the low impedance loads at high voltage levels and fast rise time made
the proposed topology capable of covering wide range of applications. Chapter 4 presented
extended analyses of this research. Moreover, employing a smart controlling algorithm made
the proposed method to prevent from any possible faults and easy diagnosis.
The proposed method was published and presented entitled “A Flexible Solid-State Pulsed
Power Topology” at the 15th
International Power Electronics and Motion Control Conference
and Exposition, Novi Sad, Serbia 2012.
Chapter 9
248
9.1.2. Considering load characteristics and effect of pulsed power
supply
To explore load characteristics and evaluate the designed power supply three real-world
applications were considered. The first application was investigating pulsed power supply for
exhaust gas treatment. In this regard DBD load characteristics were considered and based on
that a push-pull based power supply was implemented. Moreover, the effect of plasma
regarding to the generated voltage level on particle matters were studied. The measured
results depicted the availability of optimum operating point where gas treatment can be
performed while small size particle production is prevented. Chapter 5 presented extended
description about this work. The pulsed power supply take advantage of resonance
phenomena happening between the converter and the DBD capacitance to generate high level
of voltages. This idea made the developed pulsed power supply suitable for high impedance
capacitive loads. This study entitled “Effect of Pulsed Power on Particle matter in Diesel
Engine Exhaust Using a DBD Plasma Reactor” is published in IEEE Transactions on
Plasma Science Journal, Aug 2013.
To study another type of application, exciting plasma lamp was selected. As the plasma
lamp is a DBD load, the same push-pull pulsed power supply applied for the exhaust gas
treatment was considered here. The obtained results are discussed in Chapter 6. Controllable
parameters of the developed pulsed power supply (adjustable output voltage and repetition
rate) make it suitable for plasma lamp applications. The main reason is due to possibility of
exciting the plasma lamp at variety of power consumption and intensity levels by combining
different operating voltages and frequencies. Moreover, the measured results showed
nonlinear behaviour of plasma lamp which provides the feasibility of selecting an optimum
operating point.
This work entitled “Analysing DBD Plasma Lamp Intensity versus Power Consumption
Using a Push-Pull Pulsed Power Supply” is accepted in 15th
European Conference on Power
Electronics and Applications to be held in Lille, France, 2013.
Chapter 9
249
9.1.3. Increasing the efficiency of high power ultrasound applications
The efficiency of piezoelectric transducer can increases by generating a proper excitation
signal regarding to its resonance frequencies and operating conditions. Therefore, to indentify
the properties of the required excitation signal, as the first step, characteristics of a high
power piezoelectric transducer was investigated under varied situations. The complete
analysis of this research is presented in Chapter 7. The extensive experimentations under
varied mediums, power levels and voltage waveforms illustrated the nonlinear behaviour of
the piezoelectric transducer under different excited frequencies. This shows that performance
of a high power ultrasound system not necessarily increases with the increment in the input
power level.
This work was published and presented entitled “Power Electronic Converters for High
Power Ultrasound Transducers” at The 7th
IEEE Conference on Industrial Electronics and
Applications in July 2012 in Singapore.
Regarding to the obtained results from studying the piezoelectric transducer behaviour,
generating an excitation signal using multilevel converters found to be advantageous. In
addition, due to effect of environmental situation on resonance frequency variations
employing an adaptive frequency detection algorithm found to be beneficial. Further
description of this study is provided in Chapter 8. The proposed adaptive algorithm not only
follows the resonance frequency variations but also update the power converter in order to
generate the proper excitation signal.
This work entitled “Improving the Efficiency of High Power Piezoelectric Transducers for
Industrial Applications” was published in IET Science, Measurement and Technology
Journal, in Feb 2012.
Chapter 9
250
9.1.4. Summary of advantages and drawbacks of proposed topologies
and methods
Parallel Configuration of Flyback Converter for Pulsed Power Applications
Ll
Lm
Vs
1 : n
Co2
Ll
Lm
n : 1
Co1Load
Vo
+
-
Converter (1)Converter (2)
Advantages Improves rate of rise
Requires low input voltage
Provides insulation between the load side and the input side
Ability to control the energy flow
Provides both AC and DC output voltage
Requires low power rating switches as it provides current sharing at
the input side
Drawbacks Load dependent
Requires careful high voltage insulation consideration for the
transformer
The components at the load side should tolerate the whole generated
voltage
Chapter 9
251
Series Configuration of Flyback Converter for Pulsed Power Applications
Ll
Lm
Vs
1 : n
Co1
Load
Ll
Lm
1 : n
Co2
Vo
+
-
Converter (2)
Converter (1)
Advantages Improves rate of rise
Requires low input voltage
Provides insulation between the load side and the input side
Ability to control the energy flow
Provides both AC and DC output voltage
Extendable to higher number of modules
Requires low power rating switches as it provides:
Current sharing at the input side
Low reflected output voltage across the switches
Drawbacks Load dependent
Requires careful high voltage insulation consideration for the
transformer
Chapter 9
252
Modular Pulsed Power Supply Based on Series Configuration of Flyback Converter
VS
Lo
ad
Control
protocol of
switches
VLoad
+
-
Vo1
+
-
Vo2
+
-CO
CO
CO Vo3
+
-
CO Vo4
+
-
CO Vo5
+
-
Vo6
+
-
Vo7
+
-
Vo8
+
-
Vo9
+
-
Vo10
+
-
CO
CO
CO
CO
CO
Me
asu
rin
g th
e C
urr
en
t
Advantages Improves rate of rise and output voltage level
Requires low input voltage
Easy diagnostic as it is modular
Provides insulation between the load side and the input side
Ability to control the energy flow
Provides both AC and DC output voltage
Requires low number of switches
Based on low power rating switches as it provides current and voltage
sharing
Drawbacks Only applicable for high impedance applications
Requires good synchronization between the gating signal
Chapter 9
253
Flexible Pulsed Power Topology Based on Switch-Capacitor Units
S11
S12
S21
S22
Sn1
Sn2
SL
Load
Si SVVS
L
D1
D2
C1
C2
Cn
Posi
tive
buck
-boost
conver
ter
Load Switch
Control
protocol of
switches
Sw
itch
-cap
acit
or
unit
s
Hysteresis
current control
Hysteresis
voltage control
Ov
er-r
ide
sig
nal
Ov
er-r
ide
sig
nal
Voltage
Feedback
Synchronizing
PWM Signal
Advantages Flexible in terms of providing:
Adjustable output voltage
Controllable repetition rate
Controlling the energy flow
Load independent (suitable for any kind of application regardless
of the load impedance)
Provides high dv/dt
Requires low voltage components
Insulates the load side from the input side
Transformer-less
Provides protection and fault detection
Easy diagnostic as it is modular
Drawbacks D2 should be stack of series diodes as it should tolerate the whole
generated voltage
Chapter 9
254
Utilizing Push-Pull Converter for Exhaust Gas Treatment
S2
S1
Vin
DB
D
NA
NB
NS
CS
CS
NA = NB << NS
Advantages Ability to generate high level of the voltages using low voltage
switches
Requires only two switches
Provides easily controllable parameters (output voltage and repetition
rate)
High dv/dt
Compact
Insulates the load side from the input side
Suitable for DBD loads as it can discharge the load
Effective for exhaust gas treatment
Drawbacks Load dependent
Requires careful high voltage insulation consideration for the
transformer
Chapter 9
255
Developing a Pulsed Power Supply for Plasma Lamp Applications
S2
S1
Vin DB
D L
am
p
NA
NB
NS
CS
CS
NA = NB << NS
Cm
Advantages Ability to generate high level of the voltages using low voltage
switches
Requires only two switches
Effective for plasma lamp applications as it provides easily
controllable parameters (output voltage and repetition rate) in order to
control plasma lamp intensity and power consumption
High dv/dt
Compact
Suitable for DBD loads as it can discharge the load
Drawbacks Load dependent
Requires careful high voltage insulation consideration for the
transformer
Chapter 9
256
Improving the Efficiency of High Power Piezoelectric Transducers
DC/AC
Inverter
AC
Current Sensor
Piezoelectric
Frequency
Estimation
Advantages Detects piezoelectric transducer resonance frequency variations and
updates power converter fundamental frequency
Provides complete impedance frequency response
Easy to generate with power converter
Can be generated at high power
Can be applied for online processes
Fast computation due to low complexity
Cost effective
Drawbacks Required filter as it is noise sensitive
9.2. Further Research
This research study has focused on developing high frequency high power converters with
intention of applying for pulsed power and high power ultrasound applications. Moreover, the
experiments and investigated applications also yielded a large amount of valuable data
concerning converter development under different pulse conditions and load configurations.
Regarding this suggestions for further research work are discussed in six specific areas.
Chapter 9
257
Hardware implementation of the proposed flexible switch-capacitor unit based
pulsed power supply
To investigate the performance of the proposed flexible topology (Chapter 4) performance on
different applications, a hardware implementation is required. The implemented hardware
setup can cover wide range of applications especially liquid discharge applications (low
impedance load). This is quite important as recently, due to ability for employing them for
biomass production, liquid discharge applications have gained lot of interest.
Implement and investigate a flexible pulsed power supply with bipolar modulation
The proposed flexible method shows variety of advantages regarding to the pulsed power
applications. However the developed topology is capable of generating unipolar voltages,
while recent studies shows the advantages of using bipolar voltage waveform. This is due to
the fact that with bipolar modulation it is possible to generate high peak to peak value across
the load while using switches with the same break down voltage as in unipolar modulation.
Hence, studying bipolar modulation using switch-capacitor units is recommended for the
future research.
Investigating the effect of electrodes shape and different geometries on the pulsed
power performance
Optimizing the designed reactor by considering the electrodes shape and applied geometry is
effective on the generated plasma performance. This can reduces the required voltage level
and rate of rise as the reactor capacitance can adversely affect the pulsed power supply.
Investigating the effect of different geometry and electrodes shape combinations on electric
field distribution using Finite Element simulations is suggested as the first step. As the
second step the by employing the implemented hardware setup it is possible to optimize the
required reactor setup.
Study the effect of negative and positive discharges on different applications
Plasma discharge can be positive or negative. This is determined by the polarity of the
voltage on the electrode with a high potential gradient. The characteristics of positive and
negative plasma are totally unlike. This difference causes varied phenomena, regarding to the
proposed and implemented pulsed power supplies it is advantageous to study the effect of
positive and negative discharges in different applications.
Chapter 9
258
Pulsed Power Applications
Regarding to the presented information and experimentation of pulsed power applications
and considering the recent interests in bio-applications investigating the following
applications can be quite beneficial. This study not only can improve and evaluate the
performance of the designed pulsed power supplies, but also there is a high possibility of
commercialization due to recent interests and investments in these types of applications:
1-Plasma Gasification
Gasification is a process which relates to conversion of organic or fossil based carbonaceous
materials into syngas. In other word, it is a process of converting waste into gaseous fuel. In
recent years plasma gasification has gained lot of interest. However, increasing the system
performance and efficiency using power electronics techniques is one of the active areas.
QUT has conducted a cleantech project based on plasma gasification. One of the goals in this
undergoing project is finding an optimum operating point with respect to power consumption
point of view and process outcome.
2-Biomedical applications
The generated UV, ozone, and active radicals due to the presence of plasma found to be
extremely effective in killing harmful bacteria and generally sterilization. Applying pulsed
power supplies can increase compactness and reduces the time required for the process.
Investigating different types of plasma discharges, effect of AC or DC voltages, and effect of
dv/dt can be advantageous in order to provide a protocol for biomedical applications.
3- Liquid discharge applications
Production of biomass has recently gained lot of interest due to the intention of switching
from fossil fuels to biofuels. Hence, investigating an efficient method to ease pre-processing
stage of converting biological matter into energy products is extremely important. For the
pulsed power supply applications this process involves within the liquid discharge
applications. As improvements of solid-state topologies for liquid discharge applications
were quite slow (low impedance application drawbacks), there is a high research potential in
this field.
Chapter 9
259
4- Biofuel treatment
Recent applications show a high interest of using biofuels instead of diesel. Investigating the
effect of plasma on particle matter distribution and mass is an interesting research, as the
diesel exhaust gas treatment has been studied in this research.
Applying current source based converters using adaptive algorithm for high power
piezoelectric transducer applications
The extensive experiments conducted on the piezoelectric transducer showed capacitive
behaviour of the piezoelectric device. Therefore, applying current source based converter is
more efficient and can reduce lot of spikes generated across the output. In addition, the
efficiency of the ultrasound system can be improved by employing a resonance frequency
detection algorithm. Hence, developing a current source converter along with an adaptive
algorithm is recommended as a potential future work in this field.