Modelling and Analysis of DC-DC Converters For Bidirectional EV Charging Applications
Modelling and Analysis of DC-DC Converters For Bidirectional EV Charging Applications
Modelling and Analysis of DC-DC Converters For Bidirectional EV Charging Applications
By:
Lais Farias Martins
November 2019
Abstract
This thesis is focused on the modelling and analysis of DC-DC converter
topologies used for bidirectional charging of electric vehicles. Bidirectional converters
are used in vehicle-to-grid (V2G) systems to allow bidirectional power transfer between
the vehicle and the grid. Following the investigation in the literature review of potential
converter topologies used in V2G applications and modelling techniques, this research
proposes the application of the cyclic-averaging method for analysis of the Dual Active
Bridge, 4th order resonant CLLC converter, and series compensated Inductive Power
Transfer (IPT) converter.
Following the analysis of the DAB, a 4th order CLLC converter is evaluated
considering frequency and phase-shift modulations. The cyclic-averaging model is
derived to model the behaviour of the converter’s output and state-variables in steady
state. Additionally, a Fundamental Mode Approximation (FMA) model and a novel
piecewise-linear state-variable model are also implemented for comparison. The
models obtained are validated using Spice and, for the phase-shift modulated converter,
experimental results.
Overall, for all converters analysed in this research, the cyclic-averaging method
showed good performance with considerably fast execution and accuracy similar to
Spice simulations.
i
List of publications
ii
Acknowledgments
First, I would like to thank my supervisors Prof. David Stone and Prof. Martin
Foster for all the support, guidance and patience throughout my studies. Many thanks to
the EMD group colleagues and staff for the friendly and supportive environment.
Finally, I would also like to thank the Brazilian National Council for Scientific and
Technological Development (CNPq) for the financial support with the scholarship
201065/2015-0.
iii
Table of Contents
Abstract ................................................................................................................................. i
List of publications ..............................................................................................................ii
Acknowledgments ............................................................................................................. iii
Table of contents ................................................................................................................ iv
Acronyms ...........................................................................................................................vii
List of Symbols ................................................................................................................ viii
List of Figures ..................................................................................................................... xi
List of Tables ................................................................................................................. xviii
1 Introduction ...................................................................................................................... 1
1.1 Background and motivation ........................................................................................ 1
1.2 Thesis structure ........................................................................................................... 3
1.3 References ................................................................................................................... 5
2 Literature Review ............................................................................................................ 6
2.1 Introduction ................................................................................................................. 6
2.2 Bidirectional DC-DC converters ................................................................................. 7
2.2.1 Dual Active Bridge .............................................................................................. 7
2.2.2 Resonant topologies ........................................................................................... 11
2.2.3 Wireless power transfer systems ........................................................................ 15
2.3 Modelling techniques review .................................................................................... 19
2.3.1 Fundamental Mode Approximation (FMA)....................................................... 19
2.3.2 State-variable ..................................................................................................... 20
2.3.3 Cyclic-averaging ................................................................................................ 22
2.4 Conclusions ............................................................................................................... 30
2.5 References ................................................................................................................. 31
3 Cyclic-Averaging Analysis of Dual Active Bridge Converter .................................... 40
3.1 Introduction ............................................................................................................... 40
3.2 State-variable description .......................................................................................... 41
3.3 Cyclic-averaging analysis ......................................................................................... 45
3.4 Simulation results ...................................................................................................... 49
3.5 Conclusions ............................................................................................................... 56
3.6 References ................................................................................................................. 57
iv
4 Modelling of Frequency Modulated CLLC Converter .............................................. 58
4.1 Introduction ............................................................................................................... 58
4.2 Operation ................................................................................................................... 59
4.3 Fundamental Mode Approximation (FMA) .............................................................. 59
4.4 State-variable model .................................................................................................. 64
4.4.1 Implementation in Simulink .............................................................................. 70
4.5 Cyclic-averaging model ............................................................................................ 73
4.5.1 Estimation of duty cycle .................................................................................... 79
4.6 Case study ................................................................................................................. 82
4.7 Simulation results ...................................................................................................... 85
4.8 Conclusions ............................................................................................................... 91
4.9 References ................................................................................................................. 92
5 Modelling of Phase-shift Modulated CLLC Converter .............................................. 93
5.1 Introduction ............................................................................................................... 93
5.2 Phase-shift modulation technique ............................................................................. 93
5.3 Fundamental Mode Approximation (FMA) .............................................................. 95
5.4 State-variable analysis ............................................................................................... 98
5.4.1 Implementation in Simulink ............................................................................ 102
5.5 Cyclic-averaging analysis ....................................................................................... 104
5.6 Design and simulation ............................................................................................. 112
5.6.1 Converter design .............................................................................................. 112
5.6.2 Simulation results............................................................................................. 117
5.7 Conclusions ............................................................................................................. 128
5.8 References ............................................................................................................... 129
6 CLLC Converter Design and Prototype .................................................................... 131
6.1 Introduction ............................................................................................................. 131
6.2 Converter design ..................................................................................................... 132
6.2.1 Switching circuit .............................................................................................. 132
6.2.2 PCB: Deadtime, gate driver and H-bridge circuits .......................................... 135
6.2.3 Resonant tank design ....................................................................................... 137
6.2.4 Load ................................................................................................................. 141
6.3 Experimental results ................................................................................................ 142
6.4 Conclusions ............................................................................................................. 157
v
6.5 References ............................................................................................................... 158
7 CLLC Converter Design and Prototype .................................................................... 159
7.2 Fundamental Mode Approximation (FMA) applied to inductive power transfer ... 160
7.3 State-variable and cyclic-averaging analysis .......................................................... 163
7.4 Simulation results .................................................................................................... 164
7.4.1 Converter design based on FMA model .......................................................... 164
7.4.2 Verification of FMA model ............................................................................. 165
7.4.3 Verification of state-variable and cyclic-averaging method ............................ 169
7.5 Magnetics design of coil pad structures .................................................................. 172
7.6 Experimental results ................................................................................................ 183
7.6.1 Construction of coil pad structures .................................................................. 183
7.6.2 Evaluation of converter operation .................................................................... 192
7.7 Conclusions ............................................................................................................. 203
7.8 References ............................................................................................................... 204
8 Conclusions and Further Work .................................................................................. 205
8.1 Conclusions ............................................................................................................. 205
8.2 Further work ............................................................................................................ 208
8.3 References ............................................................................................................... 209
Appendix A.1 ................................................................................................................... 210
Appendix A.2 ................................................................................................................... 213
vi
Acronyms
AC - Alternating Current
BBP - Bipolar Pad
CPS - Conventional Phase-Shift
DAB - Dual Active Bridge
DBSRC - Dual-Bridge Series Resonant Converter
DC - Direct Current
DDP - Double D Pad
DDQP - Double D Quadrature Pad
DPS - Dual Phase-Shift
EPS - Extended Phase-Shift
EV - Electric vehicle
FEA - Finite Element Analysis
FMA - Fundamental Mode Approximation
GaN - Gallium Nitride
IPT - Inductive Power Transfer
MOSFET - Metal Oxide Semiconductor Field-Effect Transistor
PCB - Printed Circuit Board
PPM - Pulse-Phase Modulation
RCFMA - Rectifier Compensated Fundamental Mode Approximation
RMS - Root Mean Square
RTFMA - Rectifier Transformed Fundamental Mode Approximation
SiC - Silicon Carbide
SPS - Single Phase-Shift
SRDAB - Series Resonant Dual Active Bridge
TPS - Triple Phase-Shift
V2G - Vehicle-to-grid
WPT - Wireless Power Transfer
ZCS - Zero-Current Switching
ZVS - Zero-Voltage Switching
vii
Symbol list
viii
𝑋𝐿1 - Reactance of primary series inductor
𝑋𝐶1 - Reactance of primary series capacitor
𝑋𝐿𝑚 - Reactance of magnetizing inductor
𝑋𝐶2′ - Reactance of secondary series capacitor reflected to primary
𝑣1 - Output voltage of bridge 1
𝑣2 - Output voltage of bridge 2
𝑣𝑖 - Input voltage of resonant tank for FMA analysis
𝑣𝑖,𝐹𝑀𝐴 - Fundamental component if input voltage of resonant tank
𝑉𝑖,𝑅𝑀𝑆 - RMS input voltage of resonant tank
𝑣𝑜 - Output voltage of resonant tank for FMA analysis
𝑣𝑜,𝐹𝑀𝐴 - Fundamental component if output voltage of resonant tank
𝑉𝑜,𝑅𝑀𝑆 - RMS output voltage of resonant tank
𝑣𝐶𝑓 , 𝑣𝐶𝑖 - Filter capacitors voltage
𝑣𝐿 - Inductor voltage for DAB converter
𝑣𝐿𝑠1 - Primary series inductor voltage
𝑣𝐿𝑚 - Primary magnetizing inductor voltage
𝑣𝐶𝑠1 - Primary capacitor voltage
𝑣𝐶𝑠2 - Secondary capacitor voltage
𝑉𝑏𝑎𝑡 - Battery voltage
𝑉𝑑𝑐 - DC bus voltage
𝑉𝑑 - Diode forward voltage
𝑉𝑜𝑢𝑡 - Output voltage of converter
𝐼𝑜𝑢𝑡 - Output current of converter
𝑖𝐿 - Current through inductor for DAB converter
𝑖𝐿𝑠1 - Current through series inductor
𝑖𝐿𝑚 - Current through magnetizing inductor
𝑖𝐶𝑠1 - Current through primary capacitor
𝑖𝐶𝑠2 - Current through secondary capacitor
𝑖𝐶𝑓 , 𝑖𝐶𝑖 - Current through filter capacitors
𝑖𝑏𝑟𝑖𝑑𝑔𝑒 - Input current to slow subsystem
𝑖𝑟𝑒𝑐𝑡 - Current through bridge rectifier
𝐼1 - Primary current phasor
𝐼2 - Secondary current phasor referred to primary
ix
𝐼𝑠𝑒𝑐 - Secondary current phasor
x
List of Figures
Figure 1.1: Typical V2G system ......................................................................................... 2
Figure 2.1: Typical vehicle-to-grid system ......................................................................... 7
Figure 2.2: Wireless charging system ................................................................................. 7
Figure 2.3: Dual Active Bridge converter .......................................................................... 8
Figure 2.4: Diagrams for (a) Single Phase-Shift modulation (b) Extended Phase-Shift
modulation and (c) Dual Phase-Shift or Triple Phase-Shift modulation .......................... 10
Figure 2.5: Resonant tank of Series-Resonant Dual Active Bridge converter ................. 12
Figure 2.6: Resonant tank of LLC converter .................................................................... 13
Figure 2.7: Resonant tank of CLLC converter ................................................................. 14
Figure 2.8: IPT converter .................................................................................................. 16
Figure 2.9: LC-series compensation circuit ...................................................................... 16
Figure 2.10: LC-parallel compensation circuit ................................................................. 16
Figure 2.11: LCL compensation circuit ............................................................................ 17
Figure 2.12: Pad structures comparison: (a) Double D pad (DDP), (b) Double D
Quadrature pad (DDQP) and (c) Bipolar pad (BPP) ........................................................ 18
Figure 3.1: Dual Active Bridge converter ........................................................................ 40
Figure 3.2: Single Phase-Shift (SPS) modulation, forward mode .................................... 41
Figure 3.3: Equivalent circuits for DAB converter in (a) forward mode and (b) reverse
mode.................................................................................................................................. 42
Figure 3.4: Typical waveforms of bridge voltages for DAB converters operating in (a)
forward mode and (b) reverse mode ................................................................................. 45
Figure 3.5: Simulation results for SPS modulation (a) forward mode and (b) reverse
mode.................................................................................................................................. 51
Figure 3.6: Bridge voltages and inductor current in forward operation for (a) ϕ = 90°
and (b) ϕ = 45° ................................................................................................................ 51
Figure 3.7: Bridge voltages and inductor current in reverse operation for (a) ϕ = 90° and
(b) ϕ = 45° ....................................................................................................................... 53
Figure 3.8: Inductor current waveform from (a) Spice (b) cyclic method with 1st, 3rd and
5th harmonics considered and (c) cyclic method with 1st, 3rd, 5th, 7th and 9th harmonics
considered ......................................................................................................................... 55
Figure 4.1: CLLC topology .............................................................................................. 59
xi
Figure 4.2: Equivalent circuit for FMA analysis in forward mode, referred to primary .. 60
Figure 4.3: Equivalent circuit for FMA analysis in reverse mode, referred to primary ... 60
Figure 4.4: Equivalent circuit for state-variable analysis, forward mode ......................... 64
Figure 4.5: DC voltage gain versus switching frequency for CLLC converter in forward
operation ........................................................................................................................... 66
Figure 4.6: Typical bridge voltages and rectifier current in forward mode (a) fsw ≥ fli
(b) fsw < fli ...................................................................................................................... 67
Figure 4.7: Equivalent circuit during non-conduction mode, forward operation ............. 67
Figure 4.8: Equivalent circuit for state-variable analysis, reverse mode .......................... 68
Figure 4.9: Equivalent circuit during non-conduction mode, reverse operation .............. 69
Figure 4.10: Fast subsystem, forward mode ..................................................................... 71
Figure 4.11: Slow subsystem, forward mode ................................................................... 72
Figure 4.12: Rectifier and coupling equations, forward mode ......................................... 72
Figure 4.13: Typical voltage and current sequence for operation in region I, forward
mode.................................................................................................................................. 75
Figure 4.14: Typical voltage and current sequence for operation in region II, forward
mode.................................................................................................................................. 76
Figure 4.15: Typical voltage and current sequence for operation in region I, reverse mode
.......................................................................................................................................... 78
Figure 4.16: Typical voltage and current sequence for operation in region II, reverse
mode.................................................................................................................................. 78
Figure 4.17: Battery current versus battery voltage for forward operation ...................... 83
Figure 4.18: DC link current versus battery voltage for reverse operation ...................... 84
Figure 4.19: DC voltage gain versus switching frequency for converter operating in (a)
forward mode and (b) reverse mode ................................................................................. 85
Figure 4.20: Switching frequency for different battery voltage operation points for (a)
forward and (b) reverse operation ..................................................................................... 87
Figure 4.21: Waveforms comparison for (a) State-variable and (b) Spice simulations in
forward mode and Vbat = 350V ...................................................................................... 89
Figure 4.22: Waveforms comparison for (a) State-variable and (b) Spice simulations in
reverse mode and Vbat = 350V ....................................................................................... 90
Figure 5.1: CLLC topology .............................................................................................. 94
xii
Figure 5.2: Bridge output voltages for (a) SPS modulation, operating in reverse mode
and (b) PPM modulation, operating in forward mode ...................................................... 95
Figure 5.3: Simplified circuit for FMA analysis, referred to primary .............................. 96
Figure 5.4: Equivalent circuit for state-variable analysis, forward mode ......................... 99
Figure 5.5: Equivalent circuit for state-variable analysis, reverse mode ........................ 100
Figure 5.6: Definition of bridge voltages, forward mode ............................................... 102
Figure 5.7: Coupling equation, forward mode ................................................................ 102
Figure 5.8: Slow subsystem, forward mode ................................................................... 103
Figure 5.9: Fast subsystem, forward mode ..................................................................... 103
Figure 5.10: Typical bridge voltage and current sequence for SPS operation in (a)
forward mode and (b) reverse mode ............................................................................... 106
Figure 5.11: Typical bridge voltage and current sequence for PPM operation and 90° ≤
α ≤ 180° in (a) forward mode and (b) reverse mode ..................................................... 107
Figure 5.12: Typical bridge voltage and current sequence for PPM operation and α <
90° in (a) forward mode and (b) reverse mode............................................................... 107
Figure 5.13: Final circuit for simulation of CLLC converter ......................................... 113
Figure 5.14: Output power versus α1 and α2 for (a) ϕ = 90° (b) ϕ = 22.5° ............... 117
Figure 5.15: Primary current magnitude versus α1 and α2 for (a) ϕ = 90° (b) ϕ = 22.5°
........................................................................................................................................ 118
Figure 5.16: Secondary current magnitude versus α1 and α2 for (a) ϕ = 90° (b) ϕ =
22.5° ............................................................................................................................... 118
Figure 5.17: Magnetizing branch current magnitude versus α1 and α2 for (a) ϕ = 90°
(b) ϕ = 22.5° .................................................................................................................. 118
Figure 5.18: Output power for (a) FMA and (b) Spice simulations ............................... 119
Figure 5.19: Primary current for (a) FMA and (b) Spice simulations ............................ 120
Figure 5.20: Secondary current for (a) FMA and (b) Spice simulations ........................ 120
Figure 5.21: Magnetizing branch current for (a) FMA and (b) Spice simulations ......... 120
Figure 5.22: Efficiency from Spice simulation............................................................... 122
Figure 5.23: Simulation results for SPS modulation (a) forward mode and (b) reverse
mode................................................................................................................................ 123
Figure 5.24: Simulation results for PPM modulation (a) forward mode and (b) reverse
mode................................................................................................................................ 124
xiii
Figure 5.25: Waveform comparison for αratio = 0.25 (a) state-variable and (b) Spice,
forward mode .................................................................................................................. 126
Figure 6.1: Simplified block diagram for prototype construction .................................. 131
Figure 6.2: Switching circuit diagram ............................................................................ 132
Figure 6.3: Inverting circuit ............................................................................................ 132
Figure 6.4: Implementation of 90 degrees phase-shift, Ch2_count = 52 ...................... 134
Figure 6.5: Implementation of 180 degrees phase-shift, Ch1_count = 104 .................. 134
Figure 6.6: Implementation of angles greater than 180 degrees phase-shift, Ch0_count =
0 , Ch3_count = 104 and Inv_flag = 1 ......................................................................... 134
Figure 6.7: Implementation of CLLC converter ............................................................. 135
Figure 6.8: Deadtime circuit ........................................................................................... 136
Figure 6.9: Gate driver circuit......................................................................................... 136
Figure 6.10: Resonant tank ............................................................................................. 137
Figure 6.11: Inductor non-ideal model ........................................................................... 138
Figure 6.12: Series resistance of inductor in function of frequency ............................... 139
Figure 6.13: Capacitor non-ideal model ......................................................................... 140
Figure 6.14: Transformer non-ideal model ..................................................................... 141
Figure 6.15: Load resistor for forward operation ........................................................... 142
Figure 6.16: Experimental set-up.................................................................................... 145
Figure 6.17: Resonant tank circuits employed for (a) Spice simulation and (b) cyclic-
averaging simulation ....................................................................................................... 146
Figure 6.18: Spice, cyclic-averaging and prototype results comparison for (a) PPM
modulation and (b) SPS modulation in forward operation ............................................. 147
Figure 6.19: Spice, cyclic-averaging and prototype results comparison for (a) PPM
modulation and (b) SPS modulation in reverse operation .............................................. 147
Figure 6.20: CLLC converter circuit .............................................................................. 148
Figure 6.21: Drain source voltages for 10° phase-shift test case in forward mode ........ 148
Figure 6.22: Inductor equivalent circuit for cyclic-averaging implementation .............. 149
Figure 6.23: Secondary capacitor equivalent circuit for cyclic-averaging implementation
........................................................................................................................................ 149
Figure 6.24: Optimized Spice, cyclic-averaging and prototype results comparison for (a)
PPM modulation and (b) SPS modulation in forward operation .................................... 150
xiv
Figure 6.25: Optimized Spice, cyclic-averaging and prototype results comparison for (a)
PPM modulation and (b) SPS modulation in reverse operation ..................................... 151
Figure 6.26: Comparison between (a) Spice and (b) experimental results for the primary
current ............................................................................................................................. 152
Figure 6.27: Comparison between (a) Spice and (b) experimental results for the
secondary current ............................................................................................................ 152
Figure 6.28: Comparison between (a) Spice and (b) experimental results for the primary
capacitor voltage ............................................................................................................. 152
Figure 6.29: Comparison between (a) Spice and (b) experimental results for the
secondary capacitor voltage ............................................................................................ 153
Figure 7.1: Series compensated IPT converter ............................................................... 159
Figure 7.2: Equivalent circuits for series compensated IPT converter with (a) coupled
inductors representation and (b) T-model representation ............................................... 160
Figure 7.3: Equivalency between series compensated IPT converter and CLLC converter
........................................................................................................................................ 163
Figure 7.4: Spice and FMA results for RMS value of primary current in function of (a)
coupling coefficient and (b) self-inductance .................................................................. 166
Figure 7.5: Spice and FMA results for secondary current in function of (a) coupling
coefficient and (b) self-inductance ................................................................................. 168
Figure 7.6: Spice and FMA results for output power in function of (a) coupling
coefficient and (b) self-inductance ................................................................................. 169
Figure 7.7: Equivalent circuits for Spice simulation (a) series compensated IPT converter
and (b) CLLC equivalent circuit ..................................................................................... 170
Figure 7.8: Model comparison SPS modulation (a) forward operation and (b) reverse
operation ......................................................................................................................... 171
Figure 7.9: Model comparison PPM modulation (a) forward operation and (b) reverse
operation ......................................................................................................................... 171
Figure 7.10: 3D model of double D coil (a) top view of coil pad (b) primary and
secondary coil pads ......................................................................................................... 173
Figure 7.11: Mesh plot for coil pad structures, cross section area perpendicular to Y axis
........................................................................................................................................ 174
Figure 7.12: Coupling coefficient and self-inductance in function of coil length .......... 175
Figure 7.13: Coupling coefficient and self-inductance in function of coil width ........... 176
xv
Figure 7.14: Coupling coefficient and self-inductance in function of number of ferrite
cores per row ................................................................................................................... 176
Figure 7.15: Coupling coefficient and self-inductance in function of space between rows
of ferrite bars ................................................................................................................... 177
Figure 7.16: Coupling coefficient and self-inductance in function of pitch ................... 177
Figure 7.17: Implementation of offset in the X axis ....................................................... 178
Figure 7.18: Coupling coefficient in function of distance between pads in the X axis .. 179
Figure 7.19: Implementation of offset in the Y axis ....................................................... 179
Figure 7.20: Coupling coefficient in function of distance between pads in the Y axis .. 179
Figure 7.21: Implementation of offset in the Z axis ....................................................... 180
Figure 7.22: Coupling coefficient in function of distance between pads in the Z axis .. 180
Figure 7.23: Flux density distribution in the cross-section area perpendicular to Y axis (a)
with aluminium shield (b) no aluminium shield placed ................................................. 182
Figure 7.24: Current density distribution for primary and secondary coils and aluminium
shields ............................................................................................................................. 183
Figure 7.25: Coil pad structure ....................................................................................... 184
Figure 7.26: 3D printed structure to implement 4mm pitch ........................................... 184
Figure 7.27: Gain curve measured with Bode 100 ......................................................... 185
Figure 7.28: Inductance measurement in (a) series-aiding and (b) series-opposing
configurations ................................................................................................................. 186
Figure 7.29: Experimental measurement of coupling coefficient in function of
misalignment in the X axis ............................................................................................. 188
Figure 7.30: Experimental measurement of coupling coefficient in function of
misalignment in the Y axis ............................................................................................. 188
Figure 7.31: Experimental measurement of coupling coefficient in function of distance
between pads in the Z axis .............................................................................................. 188
Figure 7.32: Coupling coefficient in function of (a) offset in X axis (b) offset in Y axis
and (c) distance between pads in Z axis ......................................................................... 189
Figure 7.33: Primary and secondary compensation capacitors....................................... 192
Figure 7.34: Experimental and theoretical coupling coefficients in function of (a) offset
in X axis (b) offset in Y axis ........................................................................................... 193
Figure 7.35: Experimental and simulation results for 48-12V SPS modulated converter
operating in (a) forward and (b) reverse modes .............................................................. 196
xvi
Figure 7.36: Experimental and simulation results for 48-12V PPM modulated converter
operating in (a) forward and (b) reverse modes .............................................................. 197
Figure 7.37: Experimental and simulation results for 24-24V SPS modulated converter
operating in (a) forward and (b) reverse modes .............................................................. 200
Figure 7.38: Experimental and simulation results for 24-24V PPM modulated converter
operating in (a) forward and (b) reverse modes .............................................................. 200
Figure 7.39: Prototype tests: input voltage and current .................................................. 202
Figure 7.40: Prototype tests: output voltage and current ................................................ 202
Figure 8.1: Coil resistance versus frequency .................................................................. 209
xvii
List of Tables
Table 3.1: Mode descriptions for DAB converter operating under SPS modulation,
forward mode .................................................................................................................... 46
Table 3.2: Mode descriptions for DAB converter operating under SPS modulation,
reverse mode ..................................................................................................................... 46
Table 3.3: Simulation parameters ..................................................................................... 49
Table 3.4: Simulation results for SPS modulation............................................................ 50
Table 3.5: Inductor current states across a cycle considering forward operation and ϕ =
90° .................................................................................................................................... 52
Table 3.6: Inductor current states across a cycle considering forward operation and ϕ =
45° .................................................................................................................................... 52
Table 3.7: Inductor current states across a cycle considering reverse operation and ϕ =
90° .................................................................................................................................... 53
Table 3.8: Inductor current states across a cycle considering forward operation and ϕ =
45° .................................................................................................................................... 53
Table 3.9: Harmonics analysis of inductor current iL, forward mode .............................. 54
Table 4.1: Mode descriptions for converter operating in region I, forward mode ........... 75
Table 4.2: Mode descriptions for converter operating in region II, forward mode .......... 76
Table 4.3: Mode descriptions for converter operating in region I, reverse mode............. 78
Table 4.4: Mode descriptions for converter operating in region II, reverse mode ........... 79
Table 4.5: Simulation parameters ..................................................................................... 83
Table 4.6: Equivalent output load resistor across operating range ................................... 84
Table 4.7: Operation frequency forward mode ................................................................. 86
Table 4.8: Operation frequency reverse mode .................................................................. 86
Table 4.9: Percentage error between proposed models and Spice, forward mode ........... 87
Table 4.10: Percentage error between proposed models and Spice, reverse mode .......... 88
Table 4.11: Comparison of duty estimation techniques for cyclic-averaging
implementation, forward mode ......................................................................................... 88
Table 4.12: Execution time comparison ........................................................................... 91
Table 5.1: Mode descriptions for SPS modulation, forward mode…………………….106
Table 5.2: Mode descriptions for SPS modulation, reverse mode……………………..106
xviii
Table 5.3: Mode descriptions for PPM modulation and 90° ≤ α ≤ 180°, forward mode
........................................................................................................................................ 108
Table 5.4: Mode descriptions for PPM modulation and 90° ≤ α ≤ 180°, reverse mode
........................................................................................................................................ 108
Table 5.5: Mode descriptions for PPM modulation and α ≤ 90°, forward mode .......... 109
Table 5.6: Mode descriptions for PPM modulation and α ≤ 90°, reverse mode ........... 109
Table 5.7: Mode descriptions for PPM modulation and α = 90°, forward mode .......... 111
Table 5.8: Mode descriptions for PPM modulation and α = 90°, reverse mode ........... 111
Table 5.9: Resonant tank parameters calculated for different values of k ...................... 115
Table 5.10: Spice simulation results for different values of k ........................................ 116
Table 5.11: Design parameters ....................................................................................... 116
Table 5.12: Influence of voltage variations on output power ......................................... 117
Table 5.13: Simulation results for SPS modulation........................................................ 124
Table 5.14: Simulation results for PPM modulation ...................................................... 125
Table 5.15: State-variable model results for variables states considering αratio = 0.25,
forward mode .................................................................................................................. 127
Table 5.16: Cyclic-averaging model results for variables states considering αratio =
0.25, forward mode ........................................................................................................ 127
Table 5.17: Spice model results for variables states considering αratio = 0.25, forward
mode................................................................................................................................ 127
Table 5.18: Execution time comparison ......................................................................... 128
Table 6.1: Values of ChN_count and Inv_flag for different modulation cases, ............. 133
Table 6.2: Ls1 (design value: 54.04 µH) ........................................................................ 138
Table 6.3: Lm (design value: 27.02 µH)......................................................................... 139
Table 6.4: Cs1 (recalculated design value: 29.70 nF)..................................................... 140
Table 6.5: Cs2 (recalculated design value: 1.25 µF) ...................................................... 140
Table 6.6: Measured transformer parameters ................................................................. 141
Table 6.7: Inductors and capacitor equivalent parameters referred to primary .............. 150
Table 6.8: Difference between Spice and experimental results in waveform analysis... 153
Table 6.9: Influence of additional resistances on output current .................................... 154
Table 6.10: Influence of capacitor Cs1 on output current .............................................. 155
Table 6.11: Influence of capacitor Cs2 on output current .............................................. 155
Table 6.12: Influence of inductor Ls1 on output current ................................................ 155
xix
Table 6.13: Influence of inductor Lm on output current ................................................ 155
Table 6.14: Influence of frequency on output current .................................................... 156
Table 6.15: Influence of output resistor on output current in forward mode.................. 156
Table 6.16: Influence of output resistor on output current in reverse mode ................... 156
Table 7.1: Design parameters ......................................................................................... 165
Table 7.2: Comparison between Spice and FMA models for primary current (I1)........ 167
Table 7.3: Comparison between Spice and FMA models for current I2 ........................ 168
Table 7.4: Simulation parameters for IPT and CLLC-equivalent converters ................. 170
Table 7.5: Initial parameters for Double D coil in the FEA simulations ........................ 175
Table 7.6: Parameters for Double D coil in the FEA simulations .................................. 178
Table 7.7: Parameters used in Double D coil construction............................................. 181
Table 7.8: Coupling coefficient when primary and secondary pads are separated by 150
mm .................................................................................................................................. 187
Table 7.9: Coupling coefficient (k) in function of misalignment on X axis ................... 190
Table 7.10: Coupling coefficient (k) in function of misalignment on Y axis ................. 191
Table 7.11: Coupling coefficient (k) in function of distance between pads ................... 191
Table 7.12: Equivalent circuit parameters measured with Bode 100 ............................. 194
Table 7.13: Equivalent circuit parameters for reduced IPT converter and CLLC converter
........................................................................................................................................ 195
Table 7.14: Load definition for forward and reverse tests .............................................. 195
Table 7.15: Results for SPS modulation (48 – 12 V) ..................................................... 197
Table 7.16: Results for PPM modulation (48 – 12 V) .................................................... 198
Table 7.17: Influence of converter voltage ratio in the output power and efficiency of the
system at maximum modulation ..................................................................................... 199
Table 7.18: Load definition for forward and reverse tests for 24-24V tests ................... 199
Table 7.19: Results for SPS modulation (24 – 24 V) ..................................................... 201
Table 7.20: Results for PPM modulation (24 – 24 V) .................................................... 201
Table A1.1: Output power relative error (%) between FMA and Spice considering ϕ =
90°, where: Green: <5%, Orange : 5-10% and Red: >10% ............................................ 210
Table A1.2: Primary current (|I1|) relative error (%) between FMA and Spice,
considering ϕ = 90°, where: Green: <5%, Orange: 5-10% and Red: >10% ................. 211
Table A1.3: Secondary current (|I2|) relative error (%) between FMA and Spice
considering ϕ = 90°, where: Green: <5%, Orange: 5-10% and Red: >10% ................. 211
xx
Table A1.4: Magnetizing current (|Im|) relative error (%) between FMA and Spice
considering ϕ = 90°, where: Green: <5%, Orange: 5-10% and Red: >10% ................. 212
Table A1.5: Efficiency in Spice considering ϕ = 90°, where: Green: ≥0.98, Orange:
0.95-0.98 and Red <0.95 ................................................................................................. 212
Table A2.1: Results for single phase-shift modulation forward operation ..................... 213
Table A2.2: Results for single phase-shift modulation reverse operation ...................... 214
Table A2.3: Results for pulse-phase modulation forward operation .............................. 214
Table A2.4: Results for pulse-phase modulation reverse operation ............................... 215
Table A2.5: Measured input voltage SPS tests ............................................................... 215
Table A2.6: Measured input voltage PPM tests ............................................................. 216
xxi
1 Introduction
Typical EV charging systems are unidirectional, where the power flow direction is
from the grid to the battery. Vehicle-to-grid systems enable a bidirectional power
transfer between the grid and the battery pack of an EV. Since electric vehicles are
usually connected to the grid for long periods of time while charging (during the night
or when the owner is at work), the vehicle’s battery may work as a temporary energy
storage element. If the vehicle-to-grid interface allows bidirectional power flow, the
battery may be available for grid support during peak times, when the energy demand is
high. During off-peak time the battery may also absorb the excess generation. This
way, the battery may work as support for the grid, contributing to the grid stability and
efficient use of energy. All these benefits contribute to a reduction of the impact of
EV’s on the power grid.
1
In Figure 1.1 a simplified diagram for a V2G system is presented. An AC-DC
converter is connected to the grid for power factor correction and voltage/current AC-
DC or DC-AC conversion, depending on the power flow direction. The DC-DC
converter connects the DC bus at the output of the AC-DC converter to the battery and
it is mainly used for voltage and current regulation and galvanic isolation [3].
The majority of EV’s charging systems are plug-in, where a cable is used to
establish a wired connection between vehicle and charging station. However, with the
performance improvement of wireless power transfer (WPT) systems in the recent
years, bidirectional wireless chargers are a great option for use in V2G applications,
especially due to flexibility, reliability and safe operation under harsh environmental
conditions.
The analysis of high order resonant converters with circuit simulation software
(Spice, Simulink) can be very time-consuming, therefore, research of more
computationally efficient modelling techniques is essential. For validation of the cyclic-
averaging model, simulations of Fundamental Mode Approximation (FMA), state-
variable and Spice models will also be evaluated along this thesis and performance will
be compared.
2
The main novelty of this thesis is the use of the cyclic-averaging technique to
model the behaviour of the bidirectional DAB, CLLC and series compensated IPT
converters with high accuracy and faster execution compared to more traditional
models, as Spice, FMA and state-variable. Therefore, the models developed in this
thesis are useful during the converter design and control processes, where simulations
are performed multiple times to evaluate the influence of components values or control
variables and, consequently, a reduced execution time is crucial.
The main contribution topics for this thesis are summarized below.
3
A literature review is conducted in Chapter 2. Prior research on bidirectional
topologies of DC-DC converters, as Dual Active Bridge, resonant variants of the DAB
(series resonant DAB, LLC and CLLC) and topologies for wireless power transfer
applications, will be analysed. Additionally, potential modulation techniques used for
the converter control are discussed. A review on modelling techniques applied for
converter analysis is also provided, including a detailed equation description and
implementation methodology for the cyclic-averaging method.
In Chapter 7, the models developed for the phase-shift modulated CLLC converter
in Chapters 5 and 6 are adapted to model a series compensated wireless converter, also
operating with phase-shift modulation. The design and construction of the IPT
converter prototype followed by the experimental verification are also presented in this
chapter.
The thesis conclusions and further work possibilities are presented in Chapter 8.
4
1.3 References
[1] “Plan for roadside NO2 concentrations published,” Press release: Department for
Environment, Food & Rural Affairs, 2017. [Online]. Available:
https://www.gov.uk/government/news/plan-for-roadside-no2-concentrations-
published. [Accessed: 22-Oct-2019].
[5] C. Gould, “Rapid Analysis and Design of CLL Resonant Power Converters,” PhD
Thesis. University of Sheffield, 2006.
5
2 Literature Review
2.1 Introduction
In the previous chapter an introduction to vehicle-to-grid charging systems was
given and the importance of bidirectional DC-DC converters for this application was
identified. Additionally, the structure and novelty elements of this thesis were
presented. Following the background, motivations and structure of this work previously
discussed, a review on the base subjects for this research will be conducted in this
chapter.
6
Figure 2.1: Typical vehicle-to-grid system
Lastly, after the analysis of wireless power transfer systems, a review on the
modelling methods used to describe DC-DC converters will be performed. The main
techniques analysed are Fundamental Mode Approximation (FMA), cyclic-averaging
and methods based on state-variable description.
Potential topologies suitable for high power V2G applications are reviewed in this
section. Due to the safety requirement in V2G systems and consequent need of galvanic
isolation, only isolated topologies are analysed. The converters evaluated are divided in
two main categories: the converters suitable for conventional bidirectional charging
systems, which includes the non-resonant Dual Active Bridge and its resonant variants,
and topologies suitable for wireless charging systems.
7
The Dual Active Bridge (DAB) topology was first proposed in 1988 [4]. This soft-
switching DC-DC converter was considered a suitable option for high power
applications and, initially, only unidirectional power transfer was analysed. In [5] the
bidirectional power flow and converter’s performance are analysed and experimental
results are presented. At first, the efficiency of this converter was low due to the
switching technology in the early 1990’s, however, with the evolution of the switching
devices technology and development of new modulation and control techniques, the
efficiency was reported as 92.2% for a 1 kW prototype in 2012 [6]. Also, in 2016, an
efficiency value of nearly 98% was obtained for a 1.6 kW converter [7].
The DAB topology is presented in Figure 2.3, where n is the transformer turns
ratio, 𝑣1 is the output voltage of bridge 1 and 𝑣2 the output voltage of bridge 2. The
converter is composed of two full-bridge circuits connected by an isolated transformer
and an additional inductor, 𝐿𝑠 . The left-side full-bridge (bridge 1) is connected to a
high-voltage DC bus while the right-side bridge (bridge 2) is connected to an energy
storage device, which can be a battery or ultracapacitor. The converter operates in
“forward mode” when power flows from the DC bus to charge the energy storage
device. When operating in “reverse mode” the power direction is from the energy
storage system to the DC bus, in a process of discharging the battery.
The main benefits of the topology are: bidirectional power transfer, galvanic
isolation, reduced number of components, symmetric structure, high power density,
soft-switching operation, high-frequency operation and reduced size [5], [6], [8]. The
performance of the DAB converter was previously analysed for aircraft energy storage
applications in [9] and for vehicle-to-grid/automotive applications in [6], [10], [11].
8
Various modulation techniques were proposed to operate the DAB converter. The
Conventional Phase-Shift (CPS) or Single Phase-Shift (SPS) modulation [5], [12], [13]
is the simplest and easiest method to implement, where the power can be controlled
using a simple PI-based controller. The switching circuit is implemented to generate a
high-frequency square-wave voltage with 50% duty cycle at each bridge terminal
(voltages 𝑣1 and 𝑣2 in Figure 2.3), as shown in Figure 2.4-a, and the magnitude and
direction of power transferred is controlled by the phase shift angle between the
primary and secondary bridge voltages (𝜙), where the power flows from the bridge that
generates the leading square-wave.
Despite the simple and fast implementation, the main drawback of the SPS method
is the limited voltage operation range. As the input or output voltages deviate from the
nominal operation values (𝑣1 ≠ 𝑛𝑣2 ) the soft-switching range is reduced, a high
circulating energy between the bridges appears and conduction losses increase,
resulting in significant reduction in efficiency [7], [14]–[16].
The Extended Phase-Shift (EPS) control technique proposed in [17] adds a degree
of freedom to the control, utilising the inner phase-shift ratio of a bridge (Figure 2.4-b).
The output voltage of one of the bridges is maintained as a square wave with 50% duty
cycle, as in the SPS modulation case. For the other bridge, an inner phase shift angle,
𝛼1 for bridge 1 or 𝛼2 for bridge 2, is introduced between the two legs of the bridge,
resulting in a variable duty cycle that is used as an additional control variable.
The Dual Phase-Shift (DPS) control [18]–[20] is an extension of the EPS method,
where the inner phase-shift of both bridges is modified but maintaining the same value
for both primary and secondary bridges, as shown in Figure 2.4-c, where 𝛼1 = 𝛼2 . In
the Triple Phase-Shift (TPS) or Pulse-Phase Modulation (PPM) control, shown in
Figure 2.4-c, the inner phase-shift ratios of the two bridges (𝛼1 and 𝛼2 ) may have
different values (𝛼1 = 𝛼2 or 𝛼1 ≠ 𝛼2 ) [6], [21]–[24]. Therefore, the DPS is a particular
case of TPS modulation that occurs when 𝛼1 = 𝛼2 .
9
It was observed that the use of EPS, DPS and TPS modulation techniques results in
lower peak currents, reduced conduction losses, increased soft-switching range and
improvement of efficiency [25], [26]. Although best results are obtained with the TPS
method, the manipulation of three control variables significantly increases both the
system and control complexity [7].
(a)
(b) (c)
Figure 2.4: Diagrams for (a) Single Phase-Shift modulation (b) Extended Phase-
Shift modulation and (c) Dual Phase-Shift or Triple Phase-Shift modulation
10
Advanced methods are proposed where the switching sequence is manipulated in
order to obtain triangular or trapezoidal waveforms for the transformer current [6], [7],
[27], [28]. Switching losses are reduced with the triangular method (TRM) but this
technique should only be used when input and output voltage are significantly different,
since the triangular current slope is defined based on the input and output bridge
voltages [27]. For 𝑣1 ≈ 𝑛𝑣2 the positive slope of the triangular current becomes zero
𝑣1 −𝑛𝑣2
( = 0) and the trapezoidal method (TZM) is preferred [28]. Improved efficiency
𝐿
and lower peak and rms currents are obtained with the trapezoidal method, however, for
low voltage applications, when the input voltage is not sufficiently high, the trapezoidal
current becomes triangular and this technique is no longer recommended [15]. Despite
the reduction of switching losses, the use of trapezoidal or triangular methods results in
higher rms current and a consequent increase of conduction losses when compared to
SPS modulation. A hybrid method combining the triangular and trapezoidal methods
depending on the voltage range is proposed in [28], performance is improved but
increased computational effort on behalf of the controller is required.
The main phase-shift modulation techniques proposed for the DAB topology are
reviewed and compared in [26]. Overall, the improvement of modulation techniques
proposed in the literature results in higher efficiency and optimized operation of the
DAB, but also leads to complex control systems, increasing the difficulty of physical
implementation.
The most recent research on Dual Active Bridge converters is focused on transient
analysis of the converter and development of novel modulation techniques to improve
the operation and efficiency of the DAB [29].
The DAB converter presents large peak currents and limited operation range when
operating with simpler modulation and control methods. To overcome the flaws of the
DAB topology, alternative converters were proposed using the DAB topology as a base
but incorporating a resonant tank in the interface between the two bridges. These
resonant topologies can achieve higher efficiency, reduced peak currents and improved
range of soft-switching operation [30]. Various resonant converters are analysed in
[31] for renewable energy applications and in [1] for high voltage gain applications.
11
A simple resonant topology is obtained when a capacitor is added in series to the
inductor of the conventional DAB topology, as shown in Figure 2.5. This converter is
named Dual-Bridge Series Resonant converter (DBSRC) or Series-Resonant Dual
Active Bridge (SRDAB) and its operation is analysed in [31]–[35]. The addition of a
capacitor causes reduction of harmonics in the high-frequency transformer current,
resulting in a waveform closer to a sine wave, while in a standard DAB converter the
current has a triangular waveform. The capacitor is also responsible for blocking DC
currents that could lead to saturation of the transformer [36].
The SRDAB and DAB topologies were compared in [37]. Efficiency improvement
and reduction of rms currents were noticed for the resonant topology. However, both
topologies presented higher efficiency when the primary voltage was close to the
reflected secondary voltage and performance degradation was observed for operation
outside this range. Therefore, both DAB and DBSRC converters are not appropriate to
operate in applications with a wide voltage range when using less complex modulation
techniques.
Similar to the DAB converter, advanced control and modulation techniques may be
applied to improve the operation of the resonant converter. The use of Single Phase-
Shift modulation results in poor efficiency, therefore, various phase-shift modulation
techniques are applied in [32], [35], [36] and [38] to reduce current stress and improve
the efficiency of the SRDAB. Although most research is focused on phase-shift
modulated SRDAB, variable frequency control is analysed in [37] and [39]. A
drawback of the frequency modulation technique is the fact that, for resonant
converters, maximum efficiency is achieved for operation close to the resonant
frequency. Therefore, for converters operating with a wide voltage range, frequency
modulation is not indicated since the switching frequency would significantly deviate
from the resonant point during control.
12
The LLC converter [40], [41] is an alternative resonant topology where a shunt
inductor (𝐿𝑚 ) is added to the resonant tank of the SRDAB topology, as presented in
Figure 2.6.
The unidirectional version of the LLC converter is a popular topology for wide
voltage range applications due to its ability of achieving soft-switching over the full
operation range for all switching devices [42]. However, the behaviour of the
bidirectional version is equal to the unidirectional converter only in forward mode
(charging) operation. For reverse mode operation (discharging), the shunt inductor is
clamped and the system behaves as a conventional series resonant converter, resulting
in large losses, especially when switching frequency and resonant frequency are
significantly different. The difference between the forward and reverse mode operations
causes problems of voltage regulation, power control, waveform distortions and large
circulating current, making this topology not suitable for bidirectional wide voltage
range applications [42]–[44].
To overcome the limitations of the 3rd order LLC converter for bidirectional
operation, a 4th order CLLC converter is proposed in [45]. For this topology, a series
capacitor is added in the secondary side of the LLC converter resonant tank, as shown
in Figure 2.7. Here, soft-switching is achieved for both power flow directions, zero-
voltage switching (ZVS) for the primary side and zero-current switching (ZCS) for the
secondary side, resulting in a minimization of switching losses and an overall
improvement of efficiency. A maximum efficiency of 97.9% is obtained in [44] for a
3.5kW converter.
13
Figure 2.7: Resonant tank of CLLC converter
14
with low switching frequency and high tank currents the method proposed in [46],
where 𝜙 = ±90°, is more appropriate.
In the field of V2G systems, the optimization of the bidirectional charging process
is a key point of research and is the basis for several studies focused on improvement of
factors like cost, flexibility, efficiency and complexity of the charging/discharging
system. Wireless Power Transfer (WPT) methods have been used in V2G systems with
the objective to meet these demands avoiding the need of plugs or any physical
connections for the charging process. The use of WPT systems is particularly important
to guarantee the safety isolation between the vehicle and the grid and reliability in harsh
environment conditions, when the systems are exposed to rain, dirt, chemicals or dust
for example. The benefits of WPT make it suitable for different areas, including
aerospace, automotive and biomedical applications.
Inductive Power Transfer (IPT) systems consists of two or more loosely coupled
coils used for wireless power transfer by induction. Although most of research is
focused on unidirectional systems, bidirectional IPT systems for V2G applications were
analysed in [48]–[50].
A conventional IPT system is shown in Figure 2.8. The system is symmetrical and
composed by a primary and secondary (or pickup) side, where the primary side is
placed on the ground and the secondary side is placed inside the vehicle. Similar to the
DAB converter, an active full-bridge (bridge 1) is connected to a DC-link while the
secondary active bridge (bridge 2) is connected to the battery. In IPT systems, primary
and secondary coil pads are separated by a large air gap. According to the Society of
Automotive Engineers (SAE) standard for WPT for light duty EV’s [51], the distance
between the lower surface of the primary coil pad and ground surface must be in a
range from 100 to 250 mm.
The primary and secondary compensation circuits are implemented to minimize the
volt-ampere (VA) rating of the power supply, improve power factor in the primary and
secondary circuits and achieve maximum power transfer capability, with consequent
improvement of efficiency [52], [53].
15
Figure 2.8: IPT converter
Various configurations were proposed for the compensation circuit, the series
compensated, LC-series or Series-Series (SS) circuit [54]–[56], shown in Figure 2.9, is
the most popular configuration, where a series capacitor is connected in series to the
inductor in both primary and secondary sides.
The LCL compensation circuit [57], [58], shown in Figure 2.11, is more complex
due to the increased number of components. In this topology, filter inductors (𝐿𝑠1 and
𝐿𝑠2 ) are added to the LC-parallel configuration in order to reduce harmonics of the
inverter current.
16
Figure 2.11: LCL compensation circuit
For all configurations, the mutual inductance M is defined based on the coupling
coefficient k and coil self-inductances 𝐿1 and 𝐿2 . It is important to note that the
coupling coefficient decreases as the distance between the primary and secondary pads
increases.
𝑀 = 𝑘√𝐿1 𝐿2
(2.1)
1 1
𝜔𝑟 = =
√𝐿1 𝐶1 √𝐿2 𝐶2
(2.2)
The most used compensation circuits, LC-series and LCL, are evaluated in [59]
while in [52] and [60] the LC-series, LC-parallel and LCL configurations are analysed
and compared, where it was shown that LCL compensation is less sensitive to
misalignments between the primary and secondary pads. However, better efficiency
and higher power factor are achieved with the LC-series circuit. Due to the presence of
large current spikes, the use of LC-parallel configuration results in the lowest efficiency
and power factor. It was also observed that for an ideal converter, as the coupling
coefficient reduces, the output power of the converter increases when using LC-series
compensation but reduces when using LC-parallel or LCL. Experimental results
obtained in [56] show that this behaviour is only observed for a higher range of
coupling factor, as the coupling coefficient significantly decreases, the output power
also starts to reduce for the series compensated converter.
17
For the coil pad design, various configurations have been proposed in the literature
and the most popular structures are compared in [61]–[63]. Finite element analysis is
often employed for the magnetic design process of the pad structures, as in [56], [64],
[65].
Circular pads are simple and extensively researched structures but show poor
coupling and high sensitivity to misalignment. The Double D pad (DDP) was proposed
in [66] as an alternative to the circular pad, improving coupling and tolerance for
misalignments in the Y axis. The charge zone of DDP structures is nearly two times
larger than that of equivalent circular pads.
The Bipolar pad (BPP) analysed in [67], has similar performance to a DDQ coil.
The advantage of using this structure instead of a DDQP is the less amount of copper
required for construction, since the third quadrature coil is not implemented and only
two coils are used, as shown in Figure 2.12 where the DDP, DDQP and BPP structures
are compared.
Figure 2.12: Pad structures comparison: (a) Double D pad (DDP), (b) Double D
Quadrature pad (DDQP) and (c) Bipolar pad (BPP)
Phase-shift modulation techniques are usually applied for the analysis of IPT
converters instead of frequency modulation, since best performance is obtained when
operating around the resonant frequency.
18
Despite the safety, flexibility and reliability benefits, the main drawbacks of IPT
systems are the reduced efficiency compared to the wired systems and the dependency
on the misalignment between primary and secondary coils, which also affects the
system overall performance and efficiency. Still, high efficiency systems were
previously achieved with maximum of 95.4% at 3kW with 150mm gap in [68], 95.3%
for a 6kW prototype with 150mm air gap in [69] and 97.4% for 22kW and gap of
135mm in [70].
Fundamental Mode Approximation (FMA) [71] is one of the most used frequency-
domain techniques for analysis and design of resonant DC-DC converters due to its
simplicity and fast execution. Following this method, a linear equivalent circuit is
obtained from the analysis of a non-linear converter. The behaviour of the input source
combined with the switching devices is typically modelled as a sinusoidal voltage
source, representing the fundamental component of the resonant tank input voltage,
while the output is modelled as an equivalent resistor, representing the behaviour of
transformer, output filter, output bridge and load. Conventional AC circuit analysis is
utilized to solve the system and a transfer function is obtained to describe the
converter’s behaviour and frequency response [72], [73].
According to the FMA method, only the fundamental component of voltages and
currents is considered. Therefore, despite the rapid and simple analysis, the method
suffers from accuracy problems when the actual resonant tank currents/voltages are
19
more distant from perfect sine waves, and harmonics influence must be considered.
Additionally, only steady-state results are obtained when using FMA.
The FMA technique was previously applied for analysis of LCC [73]–[75], LLC
[76], [77] CLLLC converters [44], [78] and CLLC converters [45], [79] operating
under frequency modulation. For the phase-shift modulated variant of the CLLC
converter, the fundamental assumption is employed in [47] for calculation of resonant
tank currents. FMA was also used to analyse wireless power converters in [52], [59],
[80], [81]. In [31] the FMA method is applied to evaluate and compare five popular
DC-DC resonant topologies: the series, parallel, series-parallel, LLC and CLLC
converters.
2.3.2 State-variable
𝑑𝑥(𝑡)
= 𝐴𝑥(𝑡) + 𝐵𝑢(𝑡)
{ 𝑑𝑡
𝑦(𝑡) = 𝐶𝑥(𝑡) + 𝐷𝑢(𝑡)
(2.3)
20
where 𝑥(𝑡) is the state-vector that contains all state-variables, 𝐴 is the state (or
dynamic) matrix, 𝐵 is the input matrix, 𝐶 is the output matrix and 𝐷 is the direct
transmission (or feedthrough) matrix.
The differential equations are solved based on the initial value of the state-
variables and system inputs. From the state-variable description, the transient and
steady state characteristics are obtained, therefore, this method is an important tool to
analyse bidirectional DC-DC converters, serving as base to numerous modelling
techniques.
The waveform relaxation method [89], [90] is used in [91] for fast time-domain
analysis of non-linear power electronics systems. According to this technique, a non-
linear model is decomposed into two subsystems (separating the fast and slow
dynamics elements) that are described by a set of linear equations and connected by a
coupling equation, where each subsystem is integrated over its own time-step. The
decoupling of the circuit into subsystems reduces the complexity of the equations to be
solved, resulting in decreased simulation time. This decomposition method is used to
21
analyse 3rd order LCC resonant converters in [92], 4th order LCLC converters in [93]
and to model LLC converters in [94], in all these analyses the converter’s circuit is
divided into a fast subsystem, representing the power switches and resonant tank
behaviours, and a slow subsystem that represents the output filter and load. The fast and
slow subsystem are then connected by a set of coupling equations that describe the
action of the output rectifier. Similar analysis will be adopted here in the next chapters
to analyse the DAB, CLLC and series compensated IPT converters.
2.3.3 Cyclic-averaging
In this section, the equation description for this method will be presented, this will
serve as basis for the analysis performed in Chapters 3 to 7 to model the DAB
converter, CLLC converter and the series resonant wireless converter using the cyclic
technique.
A system presents cyclic behaviour when the state-vector at the beginning and at
the end of the period are equal, as defined in (2.4).
where T is the period of a cycle calculated based on the operating frequency and c
is an integer representing the number of cycles.
𝑑𝑥(𝑡)
= 𝐴𝑖 𝑥(𝑡) + 𝐵𝑖
𝑑𝑡
22
(2.5)
where 𝑥(𝑡) is the state-vector, 𝐴𝑖 is the dynamic matrix and 𝐵𝑖 represents the
excitation term for the 𝑖 𝑡ℎ mode.
The circuit operates in each mode for a fixed period, the calculation of this time is
performed by analysing the circuit and input voltages behaviour over a cycle, and will
depend on the type of modulation employed. Considering the system operates in mode i
when 𝑡𝑖−1 ≤ 𝑡 ≤ 𝑡𝑖 , the evolution of the state-vector through mode i is calculated as
follows.
(2.6)
The duty cycle 𝑑𝑖 is defined as the normalized time interval for each operation
mode and T is the period of a cycle. Therefore, the length of each mode is defined as
∆𝑡𝑖 = 𝑡𝑖 − 𝑡𝑖−1 = 𝑑𝑖 𝑇 and the following notation is adopted:
𝑡
𝛤𝑖 = ∫𝑡 𝑖 𝑒 𝐴𝑖 (𝑡𝑖 −𝜏) 𝐵𝑖 𝑑𝜏 = (𝑒 𝐴𝑖 𝑑𝑖 𝑇 − 𝐼)𝐴−1
𝑖 𝐵𝑖 , if 𝐴𝑖 is invertible
𝑖−1
(2.7)
The system can be solved recursively combining the initial condition 𝑥(𝑡0 ) with
equations (2.6) and (2.7). However, due to the complexity of the integral term, this
calculation is considerably complicated, especially when 𝐴𝑖 is a singular matrix. To
simplify and speed up the analysis, the integration term in (2.7) can be eliminated by
introducing an augmented vector that combines the dynamic and excitation matrices.
The new augmented matrices are defined as follows.
𝑑
𝑥̂(𝑡) = 𝐴̂𝑖 𝑥̂(𝑡)
𝑑𝑡
(2.8)
where:
23
𝑥(𝑡)
𝑥̂(𝑡) = [ ]
1
𝐴𝑖 𝐵𝑖
𝐴̂𝑖 = [ ]
0 0
(2.9)
𝑥(𝑡0 )
where 𝑥̂(𝑡0 ) = [ ] is the initial condition.
1
The augmented matrix 𝜙̂𝑖 is also defined combining matrices 𝜙𝑖 and 𝛤𝑖 from (2.7).
𝜙𝑖 𝛤𝑖
𝜙̂𝑖 = 𝑒 𝐴̂𝑖 𝑑𝑖 𝑇 = [ ]
0 1
(2.12)
The periodic solution is obtained using the definition of periodic system in (2.4).
24
𝑥(𝑡0 ) = (𝐼 𝑛 − 𝜙𝑡𝑜𝑡 )−1 𝛤𝑡𝑜𝑡
(2.14)
where:
𝜙𝑡𝑜𝑡 = 𝜙𝑚 𝜙𝑚−1 … 𝜙1
Thus, with knowledge of the circuit modes and duration times, the cyclic-mode
initial condition, 𝑥(𝑡0 ), can be precisely calculated. The values of all state-variables
through a cycle are then calculated recursively using (2.9), (2.10) and (2.14).
The cyclic method is also used to calculate the average value of the state-variables
over a cycle. The averaged state-vector 𝑥𝑎𝑣𝑔 is obtained using the periodic solution
previously defined and the average value definition in (2.16).
1 𝑡0 +𝑇
𝑥𝑎𝑣𝑔 = ∫ 𝑥(𝑡)𝑑𝑡
𝑇 𝑡0
(2.16)
𝑥̇ (𝑡) = 𝐴𝑖 𝑥(𝑡) + 𝐵𝑖
{ 1
𝑦̇ (𝑡) = 𝑥̇ 𝑎𝑣𝑔 = 𝑥(𝑡)
𝑇
(2.17)
The augmented-vector technique is used once more to obtain a simplified and rapid
solution. The new state-vector is obtained combining the variables 𝑥(𝑡) and 𝑦(𝑡),
resulting in the following system.
𝑑
𝑧(𝑡) = 𝐴̃𝑖 𝑧(𝑡)
𝑑𝑡
(2.18)
where:
25
𝐴𝑖 𝐵𝑖 0
𝐴̃𝑖 = [ 0 0 0]
𝐼⁄ 0 0
𝑇
𝑥(𝑡)
𝑧(𝑡) = [ 1 ]
𝑥𝑎𝑣𝑔 (𝑡)
(2.19)
𝑥(𝑡0 )
𝑧(𝑡0 ) = [ 1 ]
0
(2.20)
𝑥(𝑡0 )
𝑧(𝑡0 + 𝑇) = 𝜙̃𝑚 𝜙̃𝑚−1 … 𝜙̃1 𝑧(𝑡0 ) = [ 1 ]
𝑥𝑎𝑣𝑔
(2.21)
where:
𝜙̃𝑖 = 𝑒 𝐴̃𝑖 𝑑𝑖 𝑇
(2.22)
The last application of the cyclic method here evaluated is the harmonics analysis.
Using the Fourier theorem, a periodic function can be decomposed into an infinite sum
of complex exponentials functions. Therefore, the periodic state-vector previously
obtained is decomposed as follows:
26
∞
𝑥(𝑡) = ∑ 𝑐𝑘 𝑒 𝑗𝑘𝜔𝑡
𝑘=−∞
(2.23)
2𝜋
where 𝜔 = , and 𝑐𝑘 is the Fourier coefficient vector for the 𝑘 𝑡ℎ harmonic defined
𝑇
by:
𝑡0 +𝑇
1
𝑐𝑘 = ∫ 𝑥(𝑡)𝑒 −𝑗𝑘𝜔𝑡 𝑑𝑡
𝑇
𝑡0
(2.24)
Since the system analysed here is piecewise linear, the integral from (2.24) is
separated into a sum of integrals for each mode i and the state-vector 𝑥(𝑡) is replaced
by the augmented vector 𝑥̂(𝑡). Consequently, for a system with m modes:
𝑡𝑖 𝑚
1
𝑐̂𝑘 = ∑ ∫ 𝑒 𝐴̂𝑖 (𝑡−𝑡𝑖−1 ) 𝑥̂(𝑡𝑖−1 )𝑒 −𝑗𝑘𝜔𝑡 𝑑𝑡
𝑇 𝑡𝑖−1
𝑖=1
(2.25)
Equation (2.25) is reorganized splitting the complex exponential 𝑒 −𝑗𝑘𝜔𝑡 into two
parts, resulting in the following equation:
𝑚 𝑡𝑖
1 −𝑗𝑘𝜔𝑡 ̂𝑖 −𝑗𝑘𝜔𝐼)(𝑡−𝑡𝑖−1 )
𝑐̂𝑘 = ∑ ∫ 𝑒 𝑖−1 𝑒 (𝐴 𝑥̂(𝑡𝑖−1 )𝑑𝑡
𝑖=1 𝑡𝑖−1 𝑇
(2.26)
From (2.26) new variables 𝑣𝑘 (𝑡) and 𝑤𝑘 (𝑡) are defined in (2.27) in (2.28).
𝑡𝑖
1 −𝑗𝑘𝜔𝑡 ̂𝑖 −𝑗𝑘𝜔𝐼)(𝑡−𝑡𝑖−1 )
𝑤𝑘 (𝑡𝑖 ) = ∫ 𝑒 𝑖−1 𝑒 (𝐴 𝑥̂(𝑡𝑖−1 )𝑑𝑡
𝑡𝑖−1 𝑇
(2.27)
27
𝑣̇ 𝑘 (𝑡) = (𝐴̂𝑖 − 𝑗𝑘𝜔𝐼)𝑣𝑘 (𝑡)
{ 1 −𝑗𝑘𝜔𝑡
𝑤̇𝑘 (𝑡) = 𝑒 𝑖−1 𝑣 (𝑡)
𝑘
𝑇
(2.28)
(2.29)
where:
𝑣𝑘 (𝑡)
𝛿𝑘 (𝑡) = [ ]
𝑤𝑘 (𝑡)
𝐴̂𝑖 − 𝑗𝑘𝜔𝐼 0
𝐴̂𝑖,𝑘 = [ 1 ]
𝑒 −𝑗𝑘𝜔𝑡𝑖−1 𝐼 0
𝑇
(2.30)
𝑣𝑘 (𝑡𝑖 )
[ ] = 𝑒 𝐴̂𝑖,𝑘𝑑𝑖 𝑇 𝜁(𝑡𝑖−1 )
𝑤𝑘 (𝑡𝑖 )
(2.31)
where
𝑥̂(𝑡𝑖 )
𝜁(𝑡𝑖 ) = [ ]
0
(2.32)
Comparing (2.26) and (2.27) it is possible to calculate the Fourier coefficients from
the sum of 𝑤𝑘 (𝑡𝑖 ) for all m modes.
28
𝑚
𝑐̂𝑘 = ∑ 𝑤𝑘 (𝑡𝑖 )
𝑖=1
(2.33)
𝑐𝑘
where 𝑐̂𝑘 = [ ].
0
Based on the equation description presented in this section, the methodology for
application of the cyclic-averaging method is summarized in the following steps:
7. The values of the state-variables at the transition of each mode are defined,
𝑥(𝑡𝑖 )
since 𝑥̂(𝑡𝑖 ) = [ ].
1
To calculate the average value of the state-variables in a full cycle the following
steps are furtherly taken:
12. Calculation of the Fourier coefficient vector for the 𝑘 𝑡ℎ harmonic, 𝑐𝑘 , from
(2.31) and (2.33).
29
The cyclic-averaging is an attractive technique due to the rapid and accurate
analysis. As previously shown in [96], the accuracy of the cyclic method is close to
component-based simulations (Spice) but requiring 1/10,000th of the computation time
when analysing 3rd order LCC converters. The main drawback of this method, however,
is the need for a priori knowledge of the circuit’s behaviour during a cycle for the
determination of the operation modes and calculation of duty cycles [76]. This analysis
is required since the converter’s behaviour and number of modes change depending on
the choice of modulation and various circuit’s variables, as the operating frequency and
bridge/legs phase-shift angles.
2.4 Conclusions
In this chapter a literature review was conducted in fundamental topics for the
development of this research. The most popular topologies of DC-DC converters
applied to wired and wireless bidirectional charging applications were described and
potential modelling techniques used to describe the behaviour of these converters were
evaluated.
Based on the review of the DC-DC converters, the CLLC resonant converter was
selected as a potential topology to be furtherly analysed due to its improved
performance in bidirectional operation. From the modelling techniques described, the
cyclic-averaging method, previously applied to analyse LCLC, LCC and LLC
unidirectional converters, stands out for the accuracy and low execution time.
Therefore, the cyclic analysis will be applied here to model a bidirectional CLLC
converter operating under frequency and phase-shift modulations in Chapters 4 and 5,
respectively. Due to the low complexity of the topology and lack of this analysis in the
literature, the cyclic-averaging method will be applied first to model a Dual Active
Bridge converter in the next chapter.
30
2.5 References
[1] K. L. Jorgensen, M. C. Mira, Z. Zhang, and M. A. E. Andersen, “Review of high
efficiency bidirectional DC-DC topologies with high voltage gain,” in 2017 52nd
International Universities Power Engineering Conference, UPEC 2017, 2017,
pp. 1–6.
[3] Y. Du, S. Lukic, B. Jacobson, and A. Huang, “Review of high power isolated bi-
directional DC-DC converters for PHEV/EV DC charging infrastructure,” in
2011 IEEE Energy Conversion Congress and Exposition, 2011, pp. 553–560.
[7] G. Xu, D. Sha, J. Zhang, and X. Liao, “Unified boundary trapezoidal modulation
control utilizing fixed duty cycle compensation and magnetizing current design
for dual active bridge DC-DC converter,” IEEE Transactions on Power
Electronics, vol. 8993, no. c, pp. 1–1, 2016.
31
[11] W. Yen-Ching, W. Yen-Chun, and L. Tzung-Lin, “Design and implementation of
a bidirectional isolated dual-active-bridge-based DC/DC converter with dual-
phase-shift control for electric vehicle battery,” 2013 IEEE Energy Conversion
Congress and Exposition, pp. 5468–5475, 2013.
[12] S. Inoue and H. Akagi, “A bidirectional DC-DC converter for an energy storage
system with galvanic isolation,” IEEE Transactions on Power Electronics, vol.
22, no. 6, pp. 2299–2306, 2007.
[14] G. Xu, D. Sha, Y. Xu, and X. Liao, “Dual-transformer-based DAB converter with
wide ZVS range for wide voltage conversion gain application,” IEEE
Transactions on Industrial Electronics, vol. 65, no. 4, pp. 3306–3316, 2018.
[15] S. Zengin and M. Boztepe, “A novel current modulation method to eliminate low
frequency harmonics in single-stage dual active bridge AC/DC converter,” IEEE
Transactions on Industrial Electronics, vol. 67, no. 2, pp. 1048–1058, 2019.
[16] F. Krismer and J. W. Kolar, “Accurate small-signal model for the digital control
of an automotive bidirectional dual active bridge,” IEEE Transactions on Power
Electronics, vol. 24, no. 12, pp. 2756–2768, 2009.
[20] H. Bai and C. Mi, “Eliminate reactive power and increase system efficiency of
isolated bidirectional dual-active-bridge dc-dc converters using novel dual-phase-
shift control,” IEEE Transactions on Power Electronics, vol. 23, no. 6, pp. 2905–
2914, 2008.
32
[22] A. Tong, L. Hang, G. Li, X. Jiang, and S. Gao, “Modeling and analysis of a Dual-
Active-Bridge-isolated bidirectional DC/DC converter to minimize RMS current
with whole operating range,” IEEE Transactions on Power Electronics, vol. 33,
no. 6, pp. 5302–5316, 2018.
[26] N. Hou and Y. Li, “Overview and Comparison of Modulation and Control
Strategies for Non-Resonant Single-Phase Dual-Active-Bridge dc-dc Converter,”
IEEE Transactions on Power Electronics, pp. 1–23, 2019.
[29] J. Hong, Y. Liu, and J. Yin, “Modulation Strategy for Optimized Transient
Performance of Dual Active Bridge Converter under Reverse of Power Flow
Direction,” 2019 IEEE International Conference on Industrial Technology
(ICIT), pp. 510–514, 2019.
[32] B. Y. Luan and X. Li, “A new control strategy to reduce circulation current for a
33
bidirectional resonant converter,” 2016 IEEE 8th International Power Electronics
and Motion Control Conference, IPEMC-ECCE Asia 2016, pp. 3477–3482, 2016.
[38] W. Han and L. Corradini, “General closed-form zvs analysis of dual-bridge series
resonant DC-DC converters,” IEEE Transactions on Power Electronics, vol. 34,
no. 9, pp. 9289–9302, 2019.
[39] M. Yaqoob, K. H. Loo, and Y. M. Lai, “Modeling the effect of dead-Time on the
soft-switching characteristic of variable-frequency modulated series-resonant
DAB converter,” 2017 IEEE 18th Workshop on Control and Modeling for Power
Electronics, COMPEL 2017, 2017.
[40] G. Pledl, M. Tauer, and D. Buecherl, “Theory of operation, design procedure and
simulation of a bidirectional LLC resonant converter for vehicular applications,”
2010 IEEE Vehicle Power and Propulsion Conference, VPPC 2010, pp. 1–5,
2010.
[42] S. Zong, H. Luo, W. Li, Y. Deng, and X. He, “High-power bidirectional resonant
34
DC–DC converter with equivalent switching frequency doubler,” IET Renewable
Power Generation, vol. 10, no. 6, pp. 834–842, 2016.
[45] W. Chen, P. Rong, and Z. Lu, “Snubberless bidirectional DC-DC converter with
new CLLC resonant tank featuring minimized switching loss,” IEEE
Transactions on Industrial Electronics, vol. 57, no. 9, pp. 3075–3086, 2010.
[51] SAE International, “J2954TM Wireless Power Transfer for Light-Duty Plug-In/
Electric Vehicles and Alignment Methodology,” Surface Vehicle Information
Report, 2016.
35
[53] W. Zhang, S. C. Wong, C. K. Tse, and Q. Chen, “Design for efficiency
optimization and voltage controllability of series-series compensated inductive
power transfer systems,” IEEE Transactions on Power Electronics, vol. 29, no. 1,
pp. 191–200, 2014.
[54] Y. Zhang, T. Kan, Z. Yan, and C. C. Mi, “Frequency and Voltage Tuning of
Series-Series Compensated Wireless Power Transfer System to Sustain Rated
Power under Various Conditions,” IEEE Journal of Emerging and Selected
Topics in Power Electronics, vol. 7, no. 2, pp. 1311–1317, 2019.
[56] S. Wang, J. Chen, Z. Hu, C. Rong, and M. Liu, “Optimisation design for series–
series dynamic WPT system maintaining stable transfer power,” IET Power
Electronics, vol. 10, no. 9, pp. 987–995, 2017.
[60] S. Li and C. Mi, “Wireless Power Transfer for Electric Vehicle Applications,”
IEEE Journal of Emerging and Selected Topics in Power Electronics, vol. 3, no.
1, pp. 4–17, 2014.
[61] C. Liu, C. Jiang, and C. Qiu, “Overview of coil designs for wireless charging of
electric vehicle,” 2017 IEEE PELS Workshop on Emerging Technologies:
Wireless Power Transfer, WoW 2017, pp. 15–18, 2017.
[62] G. A. Covic and J. T. Boys, “Modern trends in inductive power transfer for
transportation applications,” IEEE Journal of Emerging and Selected Topics in
Power Electronics, vol. 1, no. 1, pp. 28–41, 2013.
36
4, pp. 1937–1955, 2015.
[65] M. Budhia, G. Covic, and J. Boys, “A new IPT magnetic coupler for electric
vehicle charging systems,” IECON Proceedings (Industrial Electronics
Conference), pp. 2487–2492, 2010.
[68] F. Lu, H. Zhang, H. Hofmann, and C. Mi, “A high efficiency 3.3 kW loosely-
coupled wireless power transfer system without magnetic material,” 2015 IEEE
Energy Conversion Congress and Exposition, ECCE 2015, pp. 2282–2286, 2015.
[69] W. Li, H. Zhao, S. Li, J. Deng, T. Kan, and C. C. Mi, “Integrated LCC
Compensation Topology for Wireless Charger in Electric and Plug-in Electric
Vehicles,” IEEE Transactions on Industrial Electronics, vol. 62, no. 7, pp. 4215–
4225, 2015.
[72] C. Gould, “Rapid Analysis and Design of CLL Resonant Power Converters,”
PhD Thesis. University of Sheffield, 2006.
37
[76] C. Gould, C. M. Bingham, D. A. Stone, and M. P. Foster, “CLL resonant
converters with output short-circuit protection,” IEE Proceedings-Electric Power
Applications, vol. 152, no. 5, pp. 1296–1306, 2005.
[77] J. Jung and J. Kwon, “Theoretical analysis and optimal design of LLC resonant
converter,” 2007 European Conference on Power Electronics and Applications,
pp. 1–10, 2007.
[86] R. Dirkman, “Generalized state space averaging,” 1983 IEEE Power Electronics
Specialists Conference, pp. 283–294, 1983.
38
K. Madawala, “Modeling and Control of a CLC Resonant Dual Active Bridge,”
IEEE Transactions on Power Electronics, vol. 31, no. 10, pp. 7297–7310, 2016.
[88] Y. Zhang and P. C. Sen, “D-Q models for resonant converters,” PESC Record -
IEEE Annual Power Electronics Specialists Conference, vol. 3, pp. 1749–1753,
2004.
[90] R. A. Saleh and A. Richard Newton, “The Exploitation of Latency and Multirate
Behavior Using Nonlinear Relaxation for Circuit Simulation,” IEEE Transactions
on Computer-Aided Design of Integrated Circuits and Systems, vol. 8, no. 12, pp.
1286–1298, 1989.
39
3 Cyclic-Averaging Analysis of Dual Active
Bridge Converter
3.1 Introduction
Following the literature review presented in the last chapter, the cyclic-averaging
modelling method described in Chapter 2 will now be employed to analyse a Dual
Active Bridge (DAB) converter, shown in Figure 3.1. One of the main objectives of this
research is to employ this modelling method for analysis of resonant CLLC converters.
As explained in the previous chapter, the CLLC is a resonant variant of the DAB
converter, therefore the cyclic analysis is performed here first for the DAB converter, a
simpler topology.
In this chapter, the DAB converter is considered to be operating under the simplest
type of phase-shift modulation: the Single Phase-Shift (SPS) [1], [2]. The drain-to-
source voltages of switches S1.3 (𝑣𝑆1.3) and S1.4 (𝑣𝑆1.4) and resulting primary and
secondary bridge voltages (𝑣1 and 𝑣2 ) for a converter operating in forward mode are
presented in Figure 3.2.
The bridge waveforms have fixed frequency and duty cycle of 50%. The phase-
shift angle between primary and secondary bridges (𝜙) is the only variable manipulated
40
to control the power flow direction and magnitude. The power flow direction is
determined by the signal of the bridge phase-shift angle, forward operation occurs when
𝑣1 leads 𝑣2 by an angle 𝜙, while reverse operation is achieved when the bridge voltage
𝑣1 lags 𝑣2 by an angle 𝜙, with 0 ≤ |𝜙| ≤ 90°. Maximum output power is reached
when 𝜙 = ±90°.
41
(a)
(b)
Figure 3.3: Equivalent circuits for DAB converter in (a) forward mode and (b)
reverse mode
Here, the two cases of power flow direction (forward and reverse modes) are
analysed separately and an equivalent circuit and equation description are obtained for
each case. In forward operation, the fast subsystem is referred to primary while the
slow subsystem is referred to the secondary of the transformer. For reverse operation,
both fast and slow subsystems are referred to primary. The resistors 𝑟1 and 𝑟𝐶𝑓 , 𝑟𝐶𝑖
represent the series resistance of the inductor and filter capacitors, respectively. The
resistances of switching devices may be incorporated into the value of 𝑟1 to increase
accuracy. Resistors 𝑟𝑑𝑐 and 𝑟𝑏𝑎𝑡 are placed in series with the DC source and battery,
respectively, representing their internal resistances or auxiliary resistances placed in the
circuit for current measurement. These resistors are useful to simplify the calculation of
the output current in the cyclic-averaging simulations and have a low value to prevent
significant influence on the converter’s output.
𝑣1 = 𝑣𝑟1 + 𝑣𝐿 + 𝑛𝑣2
42
𝑑𝑖𝐿
𝑣𝐿 = 𝐿 = 𝑣1 − 𝑣𝑟1 − 𝑛𝑣2
𝑑𝑡
𝑑𝑖𝐿 𝑣1 − 𝑟1 𝑖𝐿 − 𝑛𝑣2
=
𝑑𝑡 𝐿
(3.1)
Repeating the circuit analysis for the reverse mode circuit, the fast subsystem
equation is also defined for reverse operation:
𝑑𝑖𝐿 𝑛𝑣2 − 𝑟1 𝑖𝐿 − 𝑣1
=
𝑑𝑡 𝐿
(3.2)
(3.3)
(3.4)
𝑑𝑣𝐶𝑓
𝑖𝐶𝑓 = 𝐶𝑓 = 𝑖𝑏𝑟𝑖𝑑𝑔𝑒 − 𝐼𝑜𝑢𝑡
𝑑𝑡
(3.5)
The state equation for the slow subsystem in forward mode is obtained substituting
(3.4) in (3.5):
(3.6)
43
Repeating the analysis for the reverse mode circuit results in:
(3.7)
(3.8)
Similar analysis is performed for reverse operation, now the output bridge is bridge
1 and the input current to the slow subsystem will depend on the state of voltage 𝑣1 (𝑡).
Note that in this case the slow subsystem is referred to the primary side of the
transformer.
(3.9)
The differential equations obtained for fast and slow subsystems are now
reorganized resulting in the following state-space representation for forward mode:
(3.10)
44
𝑟1 𝑛𝑣2 (𝑡) − 𝑣1 (𝑡)
𝑖𝐿 (𝑡) − 0 𝑖𝐿 (𝑡) 𝐿
𝑑 𝐿
[ ]=[ 1 ][ ]+
𝑑𝑡 𝑣 (𝑡) 0 − 𝑣𝐶𝑖 (𝑡) 𝑉𝑑𝑐 + 𝑟𝑑𝑐 𝑖𝑏𝑟𝑖𝑑𝑔𝑒 (𝑡)
𝐶𝑓
𝐶𝑖 (𝑟𝑑𝑐 + 𝑟𝐶𝑖 ) [ 𝐶 (𝑟 + 𝑟 ) ]
𝑖 𝑑𝑐 𝐶𝑖
(3.11)
The coupling equations for forward and reverse operation are given by equations
(3.8) and (3.9), respectively. These are used to define the current 𝑖𝑏𝑟𝑖𝑑𝑔𝑒 (𝑡) based on
the polarity of bridge voltage 𝑣2 (𝑡) in forward operation or 𝑣1 (𝑡) in reverse operation.
The behaviour of the bridge voltages is evaluated in the next section, where the
converter operation will be analysed and the state-variable description here obtained
will serve as base for application of the cyclic-averaging method.
(a) (b)
Figure 3.4: Typical waveforms of bridge voltages for DAB converters operating
in (a) forward mode and (b) reverse mode
45
The states of variables 𝑣1 , 𝑣2 and 𝑖𝑏𝑟𝑖𝑑𝑔𝑒 during each mode are defined in Tables
3.1 and 3.2. Since the resistances 𝑟𝑑𝑐 and 𝑟𝑏𝑎𝑡 are very small, the bridge voltages can be
approximated to 𝑣1 (𝑡) = ±𝑉𝑑𝑐 and 𝑣2 (𝑡) = ±𝑉𝑏𝑎𝑡 for the calculation of the inductor
current in the fast subsystem. Alternatively, the bridge voltage at the output side may
also be approximated to 𝑣2 (𝑡) = ±𝑣𝐶𝑓 (𝑡) (or 𝑣1 (𝑡) = ±𝑣𝐶𝑖 (𝑡) in reverse mode) when
neglecting the filter capacitor resistance. Initially, the assumption of 𝑣1 (𝑡) = ±𝑉𝑑𝑐 and
𝑣2 (𝑡) = ±𝑉𝑏𝑎𝑡 will be considered, later on, the validity of these approximations will be
evaluated in the cyclic-averaging simulations and compared to the assumption of
𝑣2 (𝑡) = ±𝑣𝐶𝑓 (𝑡) or 𝑣1 (𝑡) = ±𝑣𝐶𝑖 (𝑡) to determine how this choice of approximation
affects the results.
Table 3.1: Mode descriptions for DAB converter operating under SPS modulation,
forward mode
Table 3.2: Mode descriptions for DAB converter operating under SPS modulation,
reverse mode
The values of 𝑣1 (𝑡), 𝑣2 (𝑡) and 𝑖𝑏𝑟𝑖𝑑𝑔𝑒 (𝑡) defined in the mode descriptions tables
are now incorporated into equations (3.10) and (3.11) to obtain a linear state-variable
description for each mode 𝑖 as follows.
46
𝑑𝑥(𝑡)
= 𝐴𝑖 𝑥(𝑡) + 𝐵𝑖
𝑑𝑡
(3.12)
𝑖𝐿 (𝑡)
where 𝑖 = 1, 2, 3,4 and the state-vector is 𝑥(𝑡) = [ ] for forward mode or
𝑣𝐶𝑓 (𝑡)
𝑖𝐿 (𝑡)
𝑥(𝑡) = [ ] for reverse mode.
𝑣𝐶𝑖 (𝑡)
|𝜙|
𝑑1 =
360
(3.13)
For the remaining modes, the duties are calculated from waveform symmetry:
𝑑2 = 0.5 − 𝑑1 , 𝑑3 = 𝑑1 and 𝑑4 = 𝑑2
(3.14)
𝑟1 𝑉𝑑𝑐 + 𝑛𝑉𝑏𝑎𝑡
− 0
𝐿 𝐿
𝐴1_𝑓 = ; 𝐵1_𝑓 =
𝑟𝑏𝑎𝑡 𝑛 1 𝑉𝑏𝑎𝑡
− −
[ 𝐶𝑓 (𝑟𝑏𝑎𝑡 + 𝑟𝐶𝑓 ) 𝐶𝑓 (𝑟𝑏𝑎𝑡 + 𝑟𝐶𝑓 )] [𝐶𝑓 (𝑟𝑏𝑎𝑡 + 𝑟𝐶𝑓 )]
47
𝑟1 𝑉𝑑𝑐 + 𝑛𝑉𝑏𝑎𝑡
− 0
𝐿 𝐿
𝐴1_𝑟 = ; 𝐵1_𝑟 =
𝑟𝑑𝑐 𝑛 1 𝑉𝑑𝑐
− −
[ 𝐶𝑖 (𝑟𝑑𝑐 + 𝑟𝐶𝑖 ) 𝐶𝑖 (𝑟𝑑𝑐 + 𝑟𝐶𝑖 )] [𝐶𝑖 (𝑟𝑑𝑐 + 𝑟𝐶𝑖 )]
𝑇
Mode 𝑀2 : when 𝑡1 < 𝑡 < 𝑡2 = :
2
𝑟1 𝑉𝑑𝑐 − 𝑛𝑉𝑏𝑎𝑡
− 0
𝐿 𝐿
𝐴2_𝑓 = ; 𝐵2_𝑓 =
𝑟𝑏𝑎𝑡 𝑛 1 𝑉𝑏𝑎𝑡
−
[𝐶𝑓 (𝑟𝑏𝑎𝑡 + 𝑟𝐶𝑓 ) 𝐶𝑓 (𝑟𝑏𝑎𝑡 + 𝑟𝐶𝑓 )] [𝐶𝑓 (𝑟𝑏𝑎𝑡 + 𝑟𝐶𝑓 )]
𝑟1 −𝑉𝑑𝑐 + 𝑛𝑉𝑏𝑎𝑡
− 0
𝐿 𝐿
𝐴2_𝑟 = ; 𝐵2_𝑟 =
𝑟𝑑𝑐 𝑛 1 𝑉𝑑𝑐
−
[𝐶𝑖 (𝑟𝑑𝑐 + 𝑟𝐶𝑖 ) 𝐶𝑖 (𝑟𝑑𝑐 + 𝑟𝐶𝑖 )] [ 𝐶𝑖 (𝑟𝑑𝑐 + 𝑟𝐶𝑖 ) ]
𝑇
Mode 𝑀3 : when 𝑡2 < 𝑡 < 𝑡3 = 𝑑1 𝑇 + :
2
𝑟1 −𝑉𝑑𝑐 − 𝑛𝑉𝑏𝑎𝑡
− 0
𝐿 𝐿
𝐴3_𝑓 = ; 𝐵3_𝑓 =
𝑟𝑏𝑎𝑡 𝑛 1 𝑉𝑏𝑎𝑡
−
𝐶 (𝑟
[ 𝑓 𝑏𝑎𝑡 + 𝑟𝐶𝑓 ) 𝐶 (𝑟
𝑓 𝑏𝑎𝑡 + 𝑟 )
𝐶𝑓 ] [𝐶𝑓 (𝑟𝑏𝑎𝑡 + 𝑟𝐶𝑓 )]
𝑟1 −𝑉𝑑𝑐 − 𝑛𝑉𝑏𝑎𝑡
− 0
𝐿 𝐿
𝐴3_𝑟 = ; 𝐵3_𝑟 =
𝑟𝑑𝑐 𝑛 1 𝑉𝑑𝑐
−
[𝐶𝑖 (𝑟𝑑𝑐 + 𝑟𝐶𝑖 ) 𝐶𝑖 (𝑟𝑑𝑐 + 𝑟𝐶𝑖 )] [ 𝐶𝑖 (𝑟𝑑𝑐 + 𝑟𝐶𝑖 ) ]
𝑟1 −𝑉𝑑𝑐 + 𝑛𝑉𝑏𝑎𝑡
− 0
𝐿 𝐿
𝐴4_𝑓 = ; 𝐵4_𝑓 =
𝑟𝑏𝑎𝑡 𝑛 1 𝑉𝑏𝑎𝑡
− −
[ 𝐶𝑓 (𝑟𝑏𝑎𝑡 + 𝑟𝐶𝑓 ) 𝐶𝑓 (𝑟𝑏𝑎𝑡 + 𝑟𝐶𝑓 )] [𝐶𝑓 (𝑟𝑏𝑎𝑡 + 𝑟𝐶𝑓 )]
48
𝑟1 𝑉𝑑𝑐 − 𝑛𝑉𝑏𝑎𝑡
− 0
𝐿 𝐿
𝐴4_𝑟 = ; 𝐵4_𝑟 =
𝑟𝑑𝑐 𝑛 1 𝑉𝑑𝑐
− −
[ 𝐶𝑖 (𝑟𝑑𝑐 + 𝑟𝐶𝑖 ) 𝐶𝑖 (𝑟𝑑𝑐 + 𝑟𝐶𝑖 )] [𝐶𝑖 (𝑟𝑑𝑐 + 𝑟𝐶𝑖 )]
Once the duty ratios and dynamic and input matrices are defined for each mode
and for both power flow directions, the equation description presented in Chapter 2 for
application of the cyclic-averaging method can now be used, following the
methodology described at the end of section 2.3.3. The simulation of the cyclic-
averaging model is performed in the MATLAB environment.
Parameter Value
DC link voltage
390 V
(𝑽𝒅𝒄 )
Battery voltage
180 V
(𝑽𝒃𝒂𝒕 )
Transformer 1
turns ratio (n)
L 61.2 𝜇𝐻
𝒓𝟏 0.11 Ω
𝒓𝒅𝒄 , 𝒓𝒃𝒂𝒕 0.01 Ω
𝑪𝒊 , 𝑪𝒇 3000 𝜇𝐹
f 20 kHz
49
Firstly, the average output current is measured considering forward and reverse
operations, Spice and cyclic-averaging results are presented in Table 3.4. The phase-
shift angle range analysed is from 10° to 90° with steps of 10 degrees. For the cyclic-
averaging model implementation, the average value of the output filter capacitor
voltage (state-variable 𝑣𝐶𝑓 or 𝑣𝐶𝑖 ) is calculated and the output current is determined
using equation (3.15).
𝑣𝐶𝑓_𝑎𝑣𝑔 − 𝑉𝑏𝑎𝑡
𝐼𝑜𝑢𝑡_𝑎𝑣𝑔 = 𝑟𝑏𝑎𝑡
, for forward operation
𝑣𝐶𝑖_𝑎𝑣𝑔 − 𝑉𝑑𝑐
𝐼𝑜𝑢𝑡_𝑎𝑣𝑔 = , for reverse operation
𝑟𝑑𝑐
(3.15)
The results from Table 3.4, also plotted in Figure 3.5, show that the cyclic-
averaging model could successfully model the behaviour of the converter over the full
range of SPS modulation, with accuracy similar to the Spice model. A maximum error
of 0.28% was obtained when the cyclic-averaging model output current results were
compared to Spice.
50
(a) (b)
Figure 3.5: Simulation results for SPS modulation (a) forward mode and (b)
reverse mode
The cyclic-averaging method is not only used to calculate the averaged state-vector
and estimate the output current of the converter but can also be employed to measure
the values of the state-variables over a full cycle. Following the verification of the
converter’s output current, the behaviour of the inductor current (𝑖𝐿 (𝑡)) is analysed at
the points of transition between the operation modes. The bridge voltages and inductor
current waveforms obtained in Spice are presented in Figure 3.6 considering forward
operation and two cases of modulation angles: 𝜙 = 90° and 𝜙 = 45°.
(a) (b)
Figure 3.6: Bridge voltages and inductor current in forward operation for (a) 𝝓 =
𝟗𝟎° and (b) 𝝓 = 𝟒𝟓°
51
The magnitude of the inductor current is measured at the times of transition
between modes (𝑡0 , 𝑡1 , 𝑡2 , 𝑡3 and 𝑡4 in Figure 3.6). The comparison between Spice and
cyclic-averaging results is shown in Table 3.5 for full modulation and in Table 3.6
when the modulation angle is reduced to 45°. The high accuracy of the cyclic-averaging
model is confirmed again since results are close to Spice with maximum difference of
53mA in the instant current measurements. Therefore, this method can also be useful to
quickly estimate the peak of the inductor current, which occurs at times 𝑡0 , 𝑡2 and 𝑡4 .
Table 3.5: Inductor current states across a cycle considering forward operation
and 𝝓 = 𝟗𝟎°
Spice
𝒕𝟎 𝒕𝟏 𝒕𝟐 𝒕𝟑 𝒕𝟒
𝒊𝑳 (A) -79.213 37.693 79.204 -37.702 -79.213
Cyclic-averaging
𝒕𝟎 𝒕𝟏 𝒕𝟐 𝒕𝟑 𝒕𝟒
𝒊𝑳 (A) -79.230 37.653 79.230 -37.653 -79.230
Table 3.6: Inductor current states across a cycle considering forward operation
and 𝝓 = 𝟒𝟓°
Spice
𝒕𝟎 𝒕𝟏 𝒕𝟐 𝒕𝟑 𝒕𝟒
𝒊𝑳 (A) -60.912 -2.350 60.903 2.341 -60.912
Cyclic-averaging
𝒕𝟎 𝒕𝟏 𝒕𝟐 𝒕𝟑 𝒕𝟒
𝒊𝑳 (A) -60.956 -2.390 60.956 2.390 -60.956
The inductor current is also verified for a converter operating in reverse mode. The
Spice waveforms for bridge voltages and inductor current are shown in Figure 3.7.
Similar to forward operation, the current values for cyclic-averaging and Spice models
are very close, as observed in Tables 3.7 and 3.8.
52
(a) (b)
Figure 3.7: Bridge voltages and inductor current in reverse operation for (a) 𝝓 =
𝟗𝟎° and (b) 𝝓 = 𝟒𝟓°
Table 3.7: Inductor current states across a cycle considering reverse operation
and 𝝓 = 𝟗𝟎°
Spice
𝒕𝟎 𝒕𝟏 𝒕𝟐 𝒕𝟑 𝒕𝟒
𝒊𝑳 (A) -35.816 80.080 35.822 -80.071 -35.815
Cyclic-averaging
𝒕𝟎 𝒕𝟏 𝒕𝟐 𝒕𝟑 𝒕𝟒
𝒊𝑳 (A) -35.864 80.056 35.864 -80.056 -35.864
Table 3.8: Inductor current states across a cycle considering forward operation
and 𝝓 = 𝟒𝟓°
𝒕𝟎 𝒕𝟏 𝒕𝟐 𝒕𝟑 𝒕𝟒
𝒊𝑳 (A) 3.787 61.623 -3.778 -61.613 3.787
Cyclic-averaging - reverse mode, 𝝓 = 𝟒𝟓°
𝒕𝟎 𝒕𝟏 𝒕𝟐 𝒕𝟑 𝒕𝟒
𝒊𝑳 (A) 3.732 61.575 -3.732 -61.575 3.732
The cyclic-averaging method is also useful to calculate the Fourier coefficients for
all state-variables. Here, this method will be used to calculate the magnitude of
53
fundamental, third and fifth harmonics for the inductor current 𝑖𝐿 . Due to waveform
symmetry, the even order harmonics are null.
The RMS values of fundamental, third and fifth components obtained from the
Spice and cyclic-averaging simulations are shown in Table 3.9 considering two phase-
shift angles (𝜙 = 90° and 𝜙 = 45°) and forward operation. The results for both models
are very close, confirming the accuracy of the cyclic-averaging method for harmonic
analysis. Results also show that the magnitude of the Fourier coefficients decreases as
the harmonic order increases.
Once the coefficients are obtained, the waveform of the inductor current can be
reconstructed following equation (3.16).
∞ ∞
(3.16)
54
The shape and peak of the current waveform in Figure 3.8-c, considering 1st, 3rd,
5th, 7th and 9th harmonics, is clearly closer to the current waveform measured in Spice.
Therefore, the results shown in Figure 3.8 confirm that accuracy increases as the
number of harmonics accounted in the reconstruction process increases. The
consideration of additional components beyond the 9th harmonic does not significantly
affect the inductor current waveform since the magnitude of the high order coefficients
is reduced.
(a)
(b) (c)
Figure 3.8: Inductor current waveform from (a) Spice (b) cyclic method with 1st,
3rd and 5th harmonics considered and (c) cyclic method with 1st, 3rd, 5th, 7th and 9th
harmonics considered
The cyclic-averaging model results presented in this section were obtained using
the approximation of 𝑣2 (𝑡) = ±𝑉𝑏𝑎𝑡 in forward mode and 𝑣1 (𝑡) = ±𝑉𝑑𝑐 in reverse
mode. Simulations were also performed considering 𝑣2 (𝑡) = ±𝑣𝐶𝑓 (𝑡) (and 𝑣1 (𝑡) =
±𝑣𝐶𝑖 (𝑡) for reverse mode) and no significant difference was noticed. For the
calculation of average output current a maximum difference of 0.001A was observed
55
when using 𝑣2 (𝑡) = ±𝑣𝐶𝑓 (𝑡) instead of 𝑣2 (𝑡) = ±𝑉𝑏𝑎𝑡 . A maximum error of 0.081A
for the inductor current values over a cycle, and error of 0.023A for the harmonic
components calculation were also observed when comparing the two possibilities of
approximation. Therefore, it was confirmed that when resistances 𝑟𝑏𝑎𝑡 and 𝑟𝑑𝑐 have
small values, in the order of milliohms, the influence of these resistors on converter
operation is very reduced and either approximation can be used. In a case where the
resistors have high values, the assumption of 𝑣2 (𝑡) = ±𝑣𝐶𝑓 (𝑡) (or 𝑣1 (𝑡) = ±𝑣𝐶𝑖 (𝑡))
would give more accurate results. Those resistors, however, represent internal or
measurement resistances which typically have low values.
3.5 Conclusions
In this chapter the cyclic-averaging modelling method was applied for analysis of a
Dual Active Bridge converter operating under single phase-shift modulation. The
model was employed for verification of the converter’s output current, the inductor
current cyclic behaviour and harmonic analysis. Simulations were performed under
nearly ideal conditions (considering transformer and switches ideal) and results were
verified against a Spice component-based simulation.
It was observed that the cyclic-averaging model could successfully predict the
converter’s behaviour for both power flow directions, showing similar accuracy to
Spice. Additionally, the cyclic-averaging model benefits from significantly reduced
execution time compared to the Spice.
56
3.6 References
[1] R. T. Naayagi and A. J. Forsyth, “Bidirectional DC-DC converter for aircraft
electric energy storage systems,” in 5th IET International Conference on Power
Electronics, Machines and Drives (PEMD 2010), 2010, pp. 1–6.
[4] C. Gould, “Rapid Analysis and Design of CLL Resonant Power Converters,” PhD
Thesis. University of Sheffield, 2006.
57
4 Modelling of Frequency Modulated
CLLC Converter
4.1 Introduction
The cyclic-averaging method was applied in the previous chapter to model a Dual
Active Bridge (DAB) converter operating under single phase-shift modulation. The
application of this technique resulted in an accurate model of the converter’s behaviour
with the advantage of fast execution time compared to Spice.
Although this work is aimed at fixed-frequency converters, the models are applied
to frequency modulated CLLC converters as a first step, since it is a more popular
variant and due to a lack of this analysis in the literature. Additionally, since the focus
of this research is on the phase-shift modulation variant, the design process will not be
analysed for the frequency-modulated converter. For model validation, the FMA, state-
variable and cyclic-averaging models will be developed and applied to the specification
and design presented in [1].
58
4.2 Operation
The circuit diagram of a CLLC resonant converter is presented in Figure 4.1. The
DC bus is represented by the voltage source 𝑉𝑑𝑐 and the battery is simply modelled also
as a voltage source 𝑉𝑏𝑎𝑡 . Voltages 𝑣1 and 𝑣2 represent the output voltage of each
bridge.
For forward operation switches S1.1 to S1.4 form a full-bridge inverter while
switches S2.1 to S2.4 work as a rectifier. For reverse operation the input inverter is
formed by switches S2.1 to S2.4 and the output rectifier by S1.1 to S1.4. The switching
signal is sent to the input stage switches while the output stage switches are always
open, consequently, the diodes associated to the output switches form a rectifier bridge.
Between the two bridges, a CLLC resonant network and a transformer are connected.
The transformer is used to achieve galvanic isolation between input and output sides
and the CLLC resonant tank is composed by inductances 𝐿𝑠1 and 𝐿𝑚 and capacitors 𝐶𝑠1
and 𝐶𝑠2 . The resonant capacitors also serve as blocking capacitors, eliminating DC
voltage offset on voltages 𝑣1 and 𝑣2 [2].
The switching signal has a 50% fixed duty cycle and 180° phase-shift between
switches in the same leg. Therefore, the input voltage to the resonant tank is a square
wave with variable frequency. Here, the frequency is manipulated to control the
converter’s output current.
59
simplified to the equivalent circuit shown in Figure 4.2. For the reverse mode analysis,
the equivalent circuit is presented on Figure 4.3. The circuits are referred to primary,
′ 𝐶𝑠2⁄ 𝑁1
therefore 𝐶𝑠2 = 𝑛2 , considering an ideal transformer with turns ratio 𝑛 = ⁄𝑁2 ,
with 𝑁1 and 𝑁2 being the number of turns on primary and secondary windings
respectively. The resistances of switches and resonant tank components are neglected
for simplification of the circuit analysis. The input to the resonant tank is represented as
a voltage source 𝑣𝑖 and the output filter, rectifier and load are modelled as an equivalent
resistor.
Figure 4.2: Equivalent circuit for FMA analysis in forward mode, referred to
primary
Figure 4.3: Equivalent circuit for FMA analysis in reverse mode, referred to
primary
Based on the converter operation described in the previous section, the input
voltage to the resonant tank is a square waveform 𝑣𝑖 (𝑡) that alternates between +𝑉𝑖𝑛
and −𝑉𝑖𝑛 . According to the Fourier series representation, 𝑣𝑖 (𝑡) can be expressed as in
(4.1), where 𝑖 is harmonic number and 𝑓𝑠 the switching frequency.
∞
4𝑉𝑖𝑛 1
𝑣𝑖 (𝑡) = . ∑ sin(2𝜋𝑖𝑓𝑠 𝑡)
𝜋 𝑖
𝑖=1,3,5
(4.1)
60
Additionally, the output voltage 𝑣𝑜 (𝑡) of the resonant tank is also represented as a
square wave phase-shifted from the input voltage and alternating between +𝑉𝑜𝑢𝑡 and
−𝑉𝑜𝑢𝑡 , as shown in equation (4.2).
∞
4𝑉𝑜𝑢𝑡 1
𝑣𝑜 (𝑡) = . ∑ sin(2𝜋𝑖𝑓𝑠 𝑡 − 𝑖𝜙)
𝜋 𝑖
𝑖=1,3,5
(4.2)
Notice that according to the topology diagram in Figure 4.1, for forward mode
𝑣𝑖 = 𝑣1 , 𝑉𝑖𝑛 = 𝑉𝑑𝑐 and 𝑣𝑜 = 𝑣2 , 𝑉𝑜𝑢𝑡 = 𝑉𝑏𝑎𝑡 , while for reverse mode 𝑣𝑖 = 𝑣2 , 𝑉𝑖𝑛 =
𝑉𝑏𝑎𝑡 and 𝑣𝑜 = 𝑣1 , 𝑉𝑜𝑢𝑡 = 𝑉𝑑𝑐 .
Following the FMA method, the resonant circuit is analysed assuming the
waveforms of current and voltage can be approximated to its fundamental components
and the load is modelled as an equivalent resistor. Therefore, the input voltage is
reduced to the fundamental harmonic of 𝑣𝑖 (𝑡). The fundamental component of 𝑣𝑖 (𝑡)
and the RMS value are shown in equation (4.3):
4𝑉𝑖𝑛
𝑣𝑖,𝐹𝑀𝐴 (𝑡) = sin(2𝜋𝑓𝑠 𝑡)
𝜋
4𝑉𝑖𝑛 2√2𝑉𝑖𝑛
𝑉𝑖,𝑅𝑀𝑆 = =
𝜋√2 𝜋
(4.3)
4𝑉𝑜𝑢𝑡
𝑣𝑜,𝐹𝑀𝐴 (𝑡) = sin(2𝜋𝑓𝑠 𝑡 − 𝜙)
𝜋
4𝑉𝑜𝑢𝑡 2√2𝑉𝑜𝑢𝑡
𝑉𝑜,𝑅𝑀𝑆 = =
𝜋√2 𝜋
(4.4)
After the full wave rectifier stage, the average value of the output current, 𝐼𝑜𝑢𝑡 , can
be found calculating the average of |𝑖𝑟,𝐹𝑀𝐴 | as in (4.5) for forward mode and (4.6) for
reverse mode.
61
1 𝑇 2𝑛 2√2𝑛𝐼𝑟,𝑅𝑀𝑆
𝐼𝑜𝑢𝑡 = ∫ 𝑛|𝑖𝑟,𝐹𝑀𝐴 | = 𝐼𝑟,𝑃𝐸𝐴𝐾 =
𝑇 0 𝜋 𝜋
(4.5)
1 𝑇 2 2√2𝐼𝑟,𝑅𝑀𝑆
𝐼𝑜𝑢𝑡 = ∫ |𝑖𝑟,𝐹𝑀𝐴 | = 𝐼𝑟,𝑃𝐸𝐴𝐾 =
𝑇 0 𝜋 𝜋
(4.6)
The equivalent load resistor referred to primary, at the output of the resonant tank,
can be calculated as in (4.7) for forward mode and (4.8) for reverse mode.
𝑉𝑜𝑢𝑡
where the output is represented as a load resistor 𝑅𝑙𝑜𝑎𝑑 = .
𝐼𝑜𝑢𝑡
From circuit analysis, a transfer function can be calculated for forward and reverse
modes, as in equations (4.9) and (4.10) respectively. The inclusion of components’
resistances and parasitic inductances/capacitances would result in an increasingly
complex circuit and transfer function, therefore, for simplification, these elements are
not considered for the FMA analysis here.
(4.9)
(4.10)
From the transfer functions, the DC gain can be calculated as in (4.11) for forward
mode and (4.12) for reverse mode.
62
𝑉𝑜𝑢𝑡 1
𝐷𝐶𝐺𝑎𝑖𝑛𝑓 = | |=
𝑉𝑖𝑛 √𝐴2 + 𝐵 2
(4.11)
𝑉𝑜𝑢𝑡 1
𝐷𝐶𝐺𝑎𝑖𝑛𝑟 = | |=
𝑉𝑖𝑛 √𝐶 2 + 𝐷2
(4.12)
where:
1 1
𝐴= +1−
ℎ ℎ. 𝜔𝑛2
1
𝐶 = 1−
ℎ. 𝑔. 𝜔𝑛2
1 𝑄1 . (1 + ℎ) 𝑄1
𝐵 = 𝐷 = 𝑄1 ( − 𝜔𝑛 ) + −
𝜔𝑛 𝑔. ℎ. 𝜔𝑛 𝑔. ℎ. 𝜔𝑛3
Given:
1 𝜔𝑠𝑤
𝜔𝑠𝑟 = 𝜔𝑛 =
√𝐿𝑠1 .𝐶𝑠1 𝜔𝑠𝑟
𝐿𝑚 𝐶𝑠2
ℎ= 𝑔=
𝐿𝑠1 𝐶𝑠1
𝐿𝑠1
√
𝐶𝑠1
𝑄1 =
𝑅𝑒𝑞
where 𝜔𝑠𝑟 is the series resonant frequency, 𝜔𝑠𝑤 is the switching frequency, 𝜔𝑛 is
normalized frequency, 𝑄1 quality factor, h is inductance ratio and g capacitance ratio.
63
monotonically decreasing to guarantee a linear control, reducing the complexity for the
controller implementation.
The FMA method allows easy implementation and results in a rapid solution but
suffers from accuracy problems due to the circuit simplifications and the dominant
fundamental harmonic assumption.
From the converter diagram presented in Figure 4.1, an equivalent circuit for state-
variable analysis in forward mode is obtained, as shown in Figure 4.4. To facilitate the
analysis, the technique previously employed in [3]–[5] for partitioning the converter
circuit into decoupled subsystems will also be used here, with the circuit divided in two
subsystems. The fast subsystem represents the dynamics of resonant tank and switching
devices. The output filter capacitor composes the slow subsystem due to its slow
response. The two subsystems are connected by coupling equations that represents the
non-linear behaviour of the rectifier.
The output voltage of the primary bridge is represented by 𝑣1 (𝑡) and the input
voltage to the rectifier, referred to the primary side of the transformer, is 𝑛𝑣2 (𝑡), where
𝑛 is the transformer’s turns ratio. The primary resistance, secondary capacitance
referred to primary and secondary resistance referred to primary are defined as follows:
64
𝑟1 = 𝑟𝐿𝑠1 + 𝑟𝐶𝑠1 + 2. 𝑟𝑆𝑊𝐼𝑇𝐶𝐻
𝐶𝑠2
𝐶𝑠2 ′ =
𝑛2
𝑟2 ′ = 𝑛2 𝑟𝐶2
(4.13)
From circuit analysis of the fast subsystem the following equations are derived:
𝑑𝑖𝐿𝑠1 𝑑𝑖𝐿𝑠1 1
𝑣𝐿𝑠1 = 𝐿𝑠1 → = (𝑣 − 𝑟1 𝑖𝐿𝑠1 − 𝑣𝐶𝑠1 − 𝑣𝐶𝑠2′ − 𝑟2 ′ 𝑖𝐿𝑠1 + 𝑟2 ′ 𝑖𝐿𝑚 − 𝑛𝑣2 )
𝑑𝑡 𝑑𝑡 𝐿𝑠1 1
(4.14)
𝑑𝑣𝐶𝑠1 𝑑𝑣𝐶𝑠1 1
𝑖𝐶𝑠1 = 𝐶𝑠1 → = (𝑖 )
𝑑𝑡 𝑑𝑡 𝐶𝑠1 𝐿𝑠1
(4.15)
𝑑𝑖𝐿𝑚 𝑑𝑖𝐿𝑚 1
𝑣𝐿𝑚 = 𝐿𝑚 → = (−𝑟𝐿𝑚 𝑖𝐿𝑚 + 𝑣𝐶𝑠2′ + 𝑛𝑣2 + 𝑟2 ′ 𝑖𝐿𝑠1 − 𝑟2 ′ 𝑖𝐿𝑚 )
𝑑𝑡 𝑑𝑡 𝐿𝑚
(4.16)
𝑑𝑣𝐶𝑠2′ 𝑑𝑣𝐶𝑠2′ 1
𝑖𝐶𝑠2′ = 𝐶𝑠2 ′ → = (𝑖 − 𝑖𝐿𝑚 )
𝑑𝑡 𝑑𝑡 𝐶𝑠2 ′ 𝐿𝑠1
(4.17)
𝑑𝑣𝐶𝑓 𝑑𝑣𝐶𝑓 1
𝑖𝐶𝑓 = 𝐶𝑓 → = (𝑟𝑏𝑎𝑡 |𝑖𝑟𝑒𝑐𝑡 | − 𝑣𝐶𝑓 + 𝑉𝑏𝑎𝑡 )
𝑑𝑡 𝑑𝑡 𝐶𝑓 (𝑟𝑏𝑎𝑡 + 𝑟𝐶𝑓 )
(4.18)
𝑉𝑜𝑢𝑡 − 𝑉𝑏𝑎𝑡
𝐼𝑜𝑢𝑡 =
𝑟𝑏𝑎𝑡
(4.19)
65
The rectifier operation must be analysed to define the coupling equations. For
CLLC converters operating under frequency modulation, the rectifier behaviour
depends on the operating frequency. A typical graph of DC conversion gain in function
of switching frequency for a CLLC converter, obtained from FMA analysis, is shown in
Figure 4.5. It was observed in simulations that when the switching frequency, 𝑓𝑠𝑤 , is
higher or equal to the frequency at the load independent point, 𝑓𝑙𝑖 , the rectifier will
always operate in the conduction state. In this case the output voltage of the tank is a
perfect square wave. When the converter operates at frequencies lower than 𝑓𝑙𝑖 , a non-
conduction mode occurs when the output voltage of the resonant tank is not sufficient
to turn on the diodes (|𝑣2 (𝑡)| < 𝑉𝑜𝑢𝑡 + 2𝑉𝑑 , where 𝑉𝑑 is the diode forward voltage), in
this case the bridge rectifier is not conducting and the secondary current is zero.
Figure 4.5: DC voltage gain versus switching frequency for CLLC converter in
forward operation
For the LLC converter the load independent and unity gain points coincide and
always occurs at the series resonant frequency but this premise cannot be applied to the
CLLC converter case [6], [7]. In the example from Figure 4.5, the series resonant
frequency is 73.16 kHz, the load independent point occurs around 91 kHz and the unity
gain point is load dependent and placed in a range of 96 to 98 kHz for forward
operation.
In Figure 4.6 typical inverter and rectifier bridge voltages and rectifier input
current are shown for a converter operating with a switching frequency higher or equal
to the load independent frequency (𝑓𝑠𝑤 ≥ 𝑓𝑙𝑖 ) and with switching frequency lower than
66
the load independent frequency (𝑓𝑠𝑤 < 𝑓𝑙𝑖 ). The input voltage of the resonant tank,
𝑣1 (𝑡) for forward operation, is always a square wave with amplitude 𝑉𝑑𝑐 , but the
behaviour of voltage 𝑣2 (𝑡) depends on the switching frequency range.
(a) (b)
Figure 4.6: Typical bridge voltages and rectifier current in forward mode (a)
𝒇𝒔𝒘 ≥ 𝒇𝒍𝒊 (b) 𝒇𝒔𝒘 < 𝒇𝒍𝒊
For continuous conduction mode operation, |𝑣2 | ≥ 𝑉𝑜𝑢𝑡 + 2𝑉𝑑 , the coupling
equations are defined as:
(4.20)
During the non-conduction mode, the equivalent circuit of the fast subsystem
behaves as shown in Figure 4.7. Since the diodes are not conducting in this case, no
current flows through the secondary capacitor.
67
From circuit analysis, the input current and voltage of the rectifier are defined by
equation (4.21).
𝐿𝑚 𝐿𝑠1
𝑛𝑣2 (𝑡) = [ (𝑣1 (𝑡) − 𝑣𝑐𝑠1 (𝑡) − 𝑟1 𝑖𝐿𝑠1 (𝑡) + 𝑟 𝑖 (𝑡))] − 𝑣𝐶𝑠2′ (𝑡)
𝐿𝑚 + 𝐿𝑠1 𝐿𝑚 𝐿𝑚 𝐿𝑠1
(4.21)
The same analysis performed for converters operating in forward mode is now
repeated for reverse mode operation. Here the input of the resonant tank is 𝑛𝑣2 while 𝑣1
now is the rectifier input voltage, the equivalent circuit is shown in Figure 4.8:
From circuit analysis, the equations for the fast subsystem are defined as follows:
𝑑𝑖𝐿𝑠1 𝑑𝑖𝐿𝑠1 1
𝑣𝐿𝑠1 = 𝐿𝑠1 → = (𝑛𝑣2 − 𝑣𝐶𝑠2′ − 𝑟2 ′ (𝑖𝐿𝑠1 + 𝑖𝐿𝑚 ) − 𝑟1 𝑖𝐿𝑠1 − 𝑣𝐶𝑠1 − 𝑣1 )
𝑑𝑡 𝑑𝑡 𝐿𝑠1
(4.22)
𝑑𝑣𝐶𝑠1 𝑑𝑣𝐶𝑠1 1
𝑖𝐶𝑠1 = 𝐶𝑠1 → = (𝑖 )
𝑑𝑡 𝑑𝑡 𝐶𝑠1 𝐿𝑠1
(4.23)
𝑑𝑖𝐿𝑚 𝑑𝑖𝐿𝑚 1
𝑣𝐿𝑚 = 𝐿𝑚 → = (𝑛𝑣2 − 𝑣𝐶𝑠2′ − 𝑟2 ′ (𝑖𝐿𝑠1 + 𝑖𝐿𝑚 ) − 𝑟𝐿𝑚 𝑖𝐿𝑚 )
𝑑𝑡 𝑑𝑡 𝐿𝑚
(4.24)
68
𝑑𝑣𝐶𝑠2′ 𝑑𝑣𝐶𝑠2′ 1
𝑖𝐶𝑠2′ = 𝐶𝑠2 ′ → = (𝑖 + 𝑖𝐿𝑚 )
𝑑𝑡 𝑑𝑡 𝐶𝑠2 ′ 𝐿𝑠1
(4.25)
𝑑𝑣𝐶𝑖 𝑑𝑣𝐶𝑖 1
𝑖𝐶𝑖 = 𝐶𝑖 → = (𝑟 |𝑖 | − 𝑣𝐶𝑖 + 𝑉𝑑𝑐 )
𝑑𝑡 𝑑𝑡 𝐶𝑖 (𝑟𝐶𝑖 + 𝑟𝑑𝑐 ) 𝑑𝑐 𝑟𝑒𝑐𝑡
(4.26)
(4.27)
When the converter is operating in reverse mode, and |𝑣1 | ≥ 𝑉𝑜𝑢𝑡 + 2𝑉𝑑 , the
diodes are in a conduction state and the coupling equations are defined as:
(4.28)
When non-conduction mode occurs |𝑣1 | < 𝑉𝑜𝑢𝑡 + 2𝑉𝑑 and the equivalent circuit is
shown in Figure 4.9.
In this case the diodes are not conducting, the inductor current is constant and
equal to zero and consequently the primary resistor and inductor voltages are also equal
to zero. From circuit analysis the coupling equations are given by:
69
𝑣1 (𝑡) = 𝑛𝑣2 (𝑡) − 𝑟2 ′ (𝑖𝐿𝑠1 (𝑡) + 𝑖𝐿𝑚 (𝑡)) − 𝑣𝐶𝑠2′ (𝑡) − 𝑣𝐶𝑠1 (𝑡)
(4.29)
Once the equation description is obtained, the piecewise linear state-variable model
is implemented using the graphical programming environment Simulink for modelling
and analysis of transient and steady-state behaviour of the converter. The block
diagrams for the fast and slow subsystems, coupling equations and rectifier operation
are shown in Figures 4.10, 4.11 and 4.12.
The fast and slow subsystem equations are implemented using mainly sum, gain
and integration blocks. As shown in Figure 4.12, to implement the rectifier, the input
current and voltage to the rectifier are monitored to determine if non-conduction mode
occurs.
70
Figure 4.10: Fast subsystem, forward mode
71
Figure 4.11: Slow subsystem, forward mode
72
4.5 Cyclic-averaging model
Based on the state-variable description obtained in section 4.4 and cyclic analysis
presented in Chapter 2, the cyclic-averaging model is developed for steady-state
analysis of the converter.
Combining the equations for fast and slow subsystems from the state-variable
model previously obtained, the equation description can be presented also in matrix
form as in (4.30):
𝑑
𝑥(𝑡) = 𝐴. 𝑥(𝑡) + 𝐵
𝑑𝑡
(4.30)
Where the state-vector 𝑥(𝑡), dynamic matrix 𝐴𝑓 and input matrix 𝐵𝑓 for forward
mode are defined as follows:
𝑖𝐿𝑠1 (𝑡)
𝑣𝐶𝑠1 (𝑡)
𝑥(𝑡) = 𝑖𝐿𝑚 (𝑡)
𝑣𝐶𝑠2′ (𝑡)
[ 𝑣𝐶𝑓 (𝑡) ]
(𝑟1 + 𝑟2 ′ ) 1 𝑟2 ′ 1
− − − 0
𝐿𝑠1 𝐿𝑠1 𝐿𝑠1 𝐿𝑠1
1
0 0 0 0
𝐶𝑠1
𝑟2 ′ (𝑟𝐿𝑚 + 𝑟2 ′ ) 1
𝐴𝑓 = 0 − 0
𝐿𝑚 𝐿𝑚 𝐿𝑚
1 1
0 − 0 0
𝐶𝑠2 ′ 𝐶𝑠2 ′
1
0 0 0 0 −
[ 𝐶𝑓 (𝑟𝑏𝑎𝑡 + 𝑟𝐶𝑓 )]
73
𝑣1 (𝑡) − 𝑛𝑣2 (𝑡)
𝐿𝑠1
0
𝑛𝑣2 (𝑡)
𝐵𝑓 =
𝐿𝑚
0
𝑟𝑏𝑎𝑡 |𝑖𝑟𝑒𝑐𝑡 (𝑡)| + 𝑉𝑏𝑎𝑡
[ 𝐶𝑓 (𝑟𝑏𝑎𝑡 + 𝑟𝐶𝑓 ) ]
Due to the presence of a non-conduction mode, the analysis is split in two cases:
when the converter is operating in region I (switching frequency is higher or equal to
load independent frequency: 𝑓𝑠𝑤 ≥ 𝑓𝑙𝑖 ) and when it operates in region II (switching
frequency is lower than load independent frequency: 𝑓𝑠𝑤 < 𝑓𝑙𝑖 ). Therefore, the coupling
equations, are defined as follows:
a. Conduction mode (occurs when operating in regions I or II): |𝑣2 | ≥ 𝑉𝑜𝑢𝑡 + 2𝑉𝑑
b. Non-conduction mode (occurs only when operating in region II): |𝑣2 | < 𝑉𝑜𝑢𝑡 + 2𝑉𝑑
𝐿𝑚 𝐿𝑠1
𝑛𝑣2 (𝑡) = [ (𝑣1 (𝑡) − 𝑣𝑐𝑠1 (𝑡) − 𝑟1 𝑖𝐿𝑠1 (𝑡) + 𝑟 𝑖 (𝑡))] − 𝑣𝐶𝑠2′ (𝑡)
𝐿𝑚 + 𝐿𝑠1 𝐿𝑚 𝐿𝑚 𝐿𝑠1
74
Figure 4.13: Typical voltage and current sequence for operation in region I,
forward mode
From waveform analysis and coupling equations, the bridge voltages and rectified
current can be defined for each mode as shown in Table 4.1.
Table 4.1: Mode descriptions for converter operating in region I, forward mode
The values of variables 𝑣1 (𝑡), 𝑣2 (𝑡) and 𝑖𝑟𝑒𝑐𝑡 (𝑡) for each mode 𝑀𝑛 are
incorporated into the state-variable equation (4.30) in order to obtain a set of piecewise
linear equations for each mode. A set of matrices 𝐴𝑛_𝑓 and 𝐵𝑛_𝑓 will be used with the
cyclic-averaging method equations from Chapter 2 to solve the system, where 𝑛=1, 2,
3, 4.
Similar analysis is performed for converters operating in region II. The typical
voltage and current sequence is shown in Figure 4.14, note that non-conduction state
occurs during two modes, 𝑀2 and 𝑀4 . Here the start of a cycle is determined by the
transition from negative to positive of input voltage 𝑣1 (𝑡).
75
Figure 4.14: Typical voltage and current sequence for operation in region II,
forward mode
The mode descriptions necessary to define the system over a cycle and apply the
cyclic method equations are presented in Table 4.2.
Table 4.2: Mode descriptions for converter operating in region II, forward mode
𝑖𝐿𝑠1 (𝑡)
𝑣𝐶𝑠1 (𝑡)
𝑥(𝑡) = 𝑖𝐿𝑚 (𝑡)
𝑣𝐶𝑠2′ (𝑡)
[ 𝑣𝐶𝑖 (𝑡) ]
76
(𝑟1 + 𝑟2 ′ ) 1 𝑟2 ′ 1
− − − − 0
𝐿𝑠1 𝐿𝑠1 𝐿𝑠1 𝐿𝑠1
1
0 0 0 0
𝐶𝑠1
𝑟2 ′ −(𝑟𝐿𝑚 + 𝑟2 ′ ) 1
𝐴𝑟 = − 0 − 0
𝐿𝑚 𝐿𝑚 𝐿𝑚
1 1
0 0 0
𝐶𝑠2 ′ 𝐶𝑠2 ′
1
0 0 0 0 −
[ 𝐶𝑖 (𝑟𝐶𝑖 + 𝑟𝑑𝑐 )]
𝑣1 (𝑡) = 𝑛𝑣2 (𝑡) − 𝑟2 ′ (𝑖𝐿𝑠1 (𝑡) + 𝑖𝐿𝑚 (𝑡)) − 𝑣𝐶𝑠2′ (𝑡) − 𝑣𝐶𝑠1 (𝑡)
The voltages and current sequence for a converter operating in regions I and II for
reverse mode are shown in Figures 4.15 and 4.16. Here the start of a cycle is defined by
the transition from negative to positive of the input voltage 𝑣2 .
Based on the coupling equations and waveform analysis, the mode descriptions for
converters operating in reverse mode are shown in Tables 4.3 and 4.4.
77
Figure 4.15: Typical voltage and current sequence for operation in region I,
reverse mode
Table 4.3: Mode descriptions for converter operating in region I, reverse mode
Figure 4.16: Typical voltage and current sequence for operation in region II,
reverse mode
78
Table 4.4: Mode descriptions for converter operating in region II, reverse mode
Once the steady-state operating modes, and state-variable equations for each mode,
are defined for forward and reverse operation, the next step is to determine the duration
of each mode and calculate the duty cycles.
In order to define the duty cycle for each mode, the phase-shift between the bridge
voltages must be calculated. As demonstrated in the previous chapter, for phase-shift
modulated converters, both primary and secondary bridges are active and the duties are
easily calculated based on the defined modulation angle. When operating under
frequency modulation, the secondary bridge functions as a diode rectifier bridge and the
duration of each mode depends on the rectifier’s operation and, consequently, the
operating frequency range.
The phase-shift angle between the input and output of the resonant tank can be
estimated based on FMA analysis. From the FMA equation description developed in
section 4.3, the input and output voltage phasors are defined:
2√2
𝑽𝒊 = 𝑉𝑖,𝑅𝑀𝑆 ∠0 = 𝑉 ∠0
𝜋 𝑖𝑛
2√2
𝑽𝒐 = 𝑉𝑜,𝑅𝑀𝑆 ∠𝜙 = 𝑉 ∠𝜙
𝜋 𝑜𝑢𝑡
(4.31)
And based on the transfer functions defined in (4.9) and (4.10), the phase-shift
angle between the input and output voltages can be calculated:
𝐵
𝜙 = tan−1 (− 𝐴), for forward operation
79
𝐷
𝜙 = tan−1 (− 𝐶 ), for reverse operation
Considering a converter operating with continuous rectifier current (region I), the
duty for the first mode is calculated from equation (4.32) while the duties for modes
M2, M3 and M4 can be determined by symmetry, from 𝑑1 .
𝜙
𝑑1 = ; 𝑑2 = 0.5 − 𝑑1 ; 𝑑3 = 𝑑1 and 𝑑4 = 𝑑2
2𝜋
(4.32)
When the converter is operating in region II the duties are calculated as in equation
(4.33).
𝜙
𝑑1 = 0.5 + 2 ; 𝑑2 = 0.5 − 𝑑1 ; 𝑑3 = 𝑑1 and 𝑑4 = 𝑑2
2𝜋
(4.33)
Due to accuracy limitations, the FMA analysis only gives an approximate angle
value, therefore, an optimization method must be employed to obtain more precise
values of the duties. From Figure 4.13, 4.14, 4.15 and 4.16 it is observed that the
change from mode M1 to M2 occurs when 𝑖𝑟𝑒𝑐𝑡 = 0, which corresponds to 𝑖𝐿𝑠1 −
𝑖𝐿𝑚 = 0 for forward operation, or 𝑖𝐿𝑠1 = 0 for reverse operation. Numerical analysis
will be used here to find a more accurate duty value that fulfils the above mode
transition conditions.
Based on the equations of cyclic-averaging model, the duty values are used to
calculate the values of 𝑖𝐿𝑠1 and 𝑖𝐿𝑚 at each mode transition point and therefore depend
on the value 𝑑1 . For a converter operating in forward mode, a function can be defined
as:
(4.34)
The Newton-Raphson recursive method will be used here to find the root of
equation (4.34), that corresponds to the value of 𝑑1 for 𝑖𝐿𝑠1 = 𝑖𝐿𝑚 . Using the cyclic
model equations described in Chapter 2 for mode M1, 𝑖𝐿𝑠1 and 𝑖𝐿𝑚 are first calculated
80
using the initial guess of 𝑑1 obtained from the FMA analysis in equations (4.32) and
(4.33).
To calculate the derivative of 𝑓(𝑑1 ), the function is next calculated for a small
increment of 𝛿 to 𝑑1 . Here the increment is defined as 𝛿 = 10−5 .
(4.35)
𝑓(𝑑1 + 𝛿) − 𝑓(𝑑1 )
𝑓 ′ (𝑑1 ) =
𝛿
(4.36)
𝑓(𝑑1 [𝑘])
𝑑1 [𝑘] = 𝑑1 [𝑘 − 1] −
𝑓′(𝑑1 [𝑘 − 1])
(4.37)
Where 𝑑1 [𝑘] is the present time estimate for 𝑑1 and 𝑑1 [0] is the initial condition.
At the end of each iteration, the difference between 𝑖𝐿𝑠1 and 𝑖𝐿𝑚 is verified to
check if this value is within a small error tolerance. For the converter analysis
performed in this chapter, a tolerance of 0.001, or 1 mA, is considered sufficient to
obtain accurate results.
The analysis for reverse mode is similar, the only difference is that the function is
now defined as in (4.38), since now 𝑖𝑟𝑒𝑐𝑡 = 𝑖𝐿𝑠1 .
(4.38)
The recursive method is applied to find the value 𝑑1 that results in 𝑖𝐿𝑠1 = 0, which
corresponds to the transition from mode M1 to M2.
81
When an acceptable value of duty is found, the recursive calculations are
interrupted and cyclic-averaging equations are applied to calculate the average steady-
state value of state-variables and output current. Since the series resistances of the filter
capacitors are very small, they are considered negligible and the output current is
approximated to:
𝑣𝐶𝑓,𝑎𝑣𝑔 −𝑉𝑏𝑎𝑡
𝐼𝑜𝑢𝑡,𝑎𝑣𝑔 = , for forward operation
𝑟𝑏
𝑣𝐶𝑖,𝑎𝑣𝑔 −𝑉𝑑𝑐
𝐼𝑜𝑢𝑡,𝑎𝑣𝑔 = , for reverse operation
𝑟𝑑𝑐
(4.39)
Using this technique, the initial estimation from the FMA equations is optimized to
obtain accurate values of duty cycle for the cyclic-averaging calculations.
For simulation purposes, the transformer and diodes are assumed to be ideal and
capacitors’ resistances are neglected, hence the converter is validated considering
nearly ideal conditions. The converter’s parameters used for all simulations in this
chapter are listed on Table 4.5. Based on the defined parameters, the series resonant
frequency (𝑓𝑠𝑟 ) for this converter is 73.16 kHz, where 𝜔𝑠𝑟 = 2𝜋𝑓𝑠𝑟 = 1⁄ .
√𝐿𝑠1 𝐶𝑠1
82
Table 4.5: Simulation parameters
Parameter Value
𝑽𝒅𝒄 400 V
𝑽𝒃𝒂𝒕 250 to 450 V
𝑪𝒇 , 𝑪𝒊 300 µF
𝑳𝒔𝟏 34.8 µH
𝑪𝒔𝟏 136 nF
𝑳𝒎 78.28 µH
𝑪𝒔𝟐 200 nF
n (turns ratio) 0.8333
𝒓𝒃𝒂𝒕, 𝒓𝒅𝒄 0.01 Ω
𝒓𝑺𝑾𝑰𝑻𝑪𝑯 0.044 Ω
𝒓𝑳𝒔𝟏 , 𝒓𝑳𝒎 0.1 Ω
The operation specification for forward and reverse modes is shown in Figure 4.17
and 4.18, where the output current is plotted against battery voltage. Here the frequency
is the variable used to control the output current. In forward mode the battery is charged
at a constant current of 10A whilst the battery voltage is lower than 350V. As the
voltage increases beyond this point the current is reduced to maintain the converter
maximum output power of 3.5 kW.
Figure 4.17: Battery current versus battery voltage for forward operation
83
In reverse mode the battery is supplying power to the DC bus, hence, the output
voltage is now constant at 400V. While the battery voltage is in the range of 400 to
450V the output power is maintained close to maximum, at 3.4 kW. As the input
battery voltage decreases the output power is reduced to avoid high RMS current in the
resonant tank.
Figure 4.18: DC link current versus battery voltage for reverse operation
Based on the converter operation just described, five points are chosen for
simulation and model comparison. Using the defined output voltage and current for
each point, the output can be represented as a load resistor for the FMA analysis, where
𝑉𝑜𝑢𝑡
𝑅𝑙𝑜𝑎𝑑 = ⁄𝐼 . The values of 𝑅𝑙𝑜𝑎𝑑 for the full range of operation points is shown
𝑜𝑢𝑡
in Table 4.6.
Operation points
𝑽𝒃𝒂𝒕 =250V 𝑽𝒃𝒂𝒕 =300V 𝑽𝒃𝒂𝒕 =350V 𝑽𝒃𝒂𝒕 =400V 𝑽𝒃𝒂𝒕 =450V
𝑹𝒍𝒐𝒂𝒅 (Ω)
25 30 35 45.71 60
Forward mode
𝑹𝒍𝒐𝒂𝒅 (Ω)
80 66.67 57.14 47.06 47.06
Reverse mode
84
4.7 Simulation results
Once the test points are defined, the models are simulated and validated using a
Spice circuit-based simulation. The models are used to estimate the switching
frequency necessary to provide the required output current for different values of
battery voltage following the converter’s designed operation.
Based on the FMA equation description from section 4.3, the DC voltage gain
curves are plotted for all load configurations considering forward and reverse operation,
as shown in Figure 4.19. Since the DC link voltage is constant (400V), the gain
requirement for forward mode is in a range of 0.625 to 1.125, based on the battery
voltage range. For reverse mode, the gain operation range is from 0.89 to 1.6. From the
results shown, the gain requirement is accomplished and, for this operation range, the
gain curves are monotonically decreasing which is an important characteristic for
control purposes.
(a) (b)
Figure 4.19: DC voltage gain versus switching frequency for converter operating
in (a) forward mode and (b) reverse mode
85
At first, the results obtained from model simulations are compared to Spice.
Additionally, for the 350V and 450V operation points in forward mode and 250, 350
and 450V points in reverse mode, the results are also compared to the experimental
results previously obtained in [1]. Results are shown in Tables 4.7 and 4.8 and plotted
on Figure 4.20. The frequency at the load independent point, that marks the transition
between regions I and II, is approximately 91.5 kHz in the Spice simulations, a result
close to the value of 91 kHz obtained in the FMA analysis. Therefore, in forward mode,
the converter operates in region II (𝑓𝑠𝑤 < 𝑓𝑙𝑖 ) only when 𝑉𝑏𝑎𝑡 = 450𝑉. For reverse
mode, the converter operates mostly in region II, switching to region I (𝑓𝑠𝑤 ≥ 𝑓𝑙𝑖 )
when 𝑉𝑏𝑎𝑡 = 450𝑉.
Frequency (kHz)
𝑽𝒃𝒂𝒕 =250V 𝑽𝒃𝒂𝒕 =300V 𝑽𝒃𝒂𝒕 =350V 𝑽𝒃𝒂𝒕 =400V 𝑽𝒃𝒂𝒕 =450V
FMA 147.80 134.80 118.10 97.77 78.10
Spice 136.10 123.14 109.30 94.81 79.75
State-variable 136.50 123.50 109.90 95.43 80.17
Cyclic-
136.35 123.45 109.83 95.40 80.18
averaging
Experimental
- - 114.39 - 74.34
(literature)
Frequency (kHz)
𝑽𝒃𝒂𝒕 =250V 𝑽𝒃𝒂𝒕 =300V 𝑽𝒃𝒂𝒕 =350V 𝑽𝒃𝒂𝒕 =400V 𝑽𝒃𝒂𝒕 =450V
FMA 44.16 48.57 56.92 80.12 113.10
Spice 47.60 54.26 64.50 81.38 104.27
State-
47.71 54.45 64.88 82.22 105.06
variable
Cyclic-
47.77 54.49 64.91 82.24 105.00
averaging
Experimental
45.27 - 59.85 - 97.75
(literature)
86
(a)
(b)
Figure 4.20: Switching frequency for different battery voltage operation points
for (a) forward and (b) reverse operation
The error between the frequency obtained from FMA, state-variable and cyclic-
averaging models simulation and Spice is calculated and presented on Table 4.9 for
forward operation and Table 4.10 for reverse operation.
Table 4.9: Percentage error between proposed models and Spice, forward mode
Frequency (kHz)
𝑽𝒃𝒂𝒕 =250V 𝑽𝒃𝒂𝒕 =300V 𝑽𝒃𝒂𝒕 =350V 𝑽𝒃𝒂𝒕 =400V 𝑽𝒃𝒂𝒕 =450V
FMA 8.60% 9.47% 8.05% 3.12% 2.07%
State-variable 0.29% 0.29% 0.55% 0.65% 0.53%
Cyclic-averaging 0.18% 0.25% 0.48% 0.62% 0.54%
87
Table 4.10: Percentage error between proposed models and Spice, reverse mode
Frequency (kHz)
𝑽𝒃𝒂𝒕 =250V 𝑽𝒃𝒂𝒕 =300V 𝑽𝒃𝒂𝒕 =350V 𝑽𝒃𝒂𝒕 =400V 𝑽𝒃𝒂𝒕 =450V
FMA 7.23% 10.49% 11.75% 1.55% 8.47%
State-variable 0.23% 0.35% 0.59% 1.03% 0.76%
Cyclic-averaging 0.36% 0.42% 0.64% 1.06% 0.70%
The models also showed reduced error when compared to the literature
experimental results, with a maximum error of 8.45% for cyclic-averaging and 15.7%
for FMA. The experimental results are affected by circuit resistances and parasitic
elements while the model simulations were performed under nearly ideal conditions,
resulting in increased error between simulations and practical results.
The closeness of the results for cyclic-averaging method and Spice show that the
technique adopted in section 4.5.1 for calculation of duties, with initial estimation of
duty value using FMA analysis and the Newton-Raphson method for optimization,
gives accurate results. In Table 4.11 the results previously obtained are compared to a
cyclic-averaging simulation where the duties are calculated considering only the FMA
estimation. It is verified that estimation using only FMA is not accurate.
Frequency (kHz)
𝑽𝒃𝒂𝒕 =250V 𝑽𝒃𝒂𝒕 =300V 𝑽𝒃𝒂𝒕 =350V 𝑽𝒃𝒂𝒕 =400V 𝑽𝒃𝒂𝒕 =450V
Cyclic-averaging
149.75 139.07 125.33 108.01 60.21
(only FMA)
Cyclic-averaging
136.35 123.45 109.83 95.40 80.18
(FMA+Newton)
Spice 136.10 123.14 109.30 94.81 79.75
88
As a further investigation, the waveforms obtained with the proposed state-variable
model are compared to Spice results. The currents flowing through inductors 𝐿𝑠1 and
𝐿𝑚 , the rectifier voltage and the input current to the rectifying bridge, 𝑖𝑟𝑒𝑐𝑡 , defined as
𝑖𝑟𝑒𝑐𝑡 = 𝑛(𝑖𝐿𝑠1 − 𝑖𝐿𝑚 ) for forward operation, are shown in Figure 4.21 for a converter
operating at 𝑉𝑏𝑎𝑡 =350V. As confirmed by the current waveforms, the converter is
operating in region I, consequently, non-conduction mode will not occur and the input
and output of the resonant tank are square waves.
(a)
(b)
Figure 4.21: Waveforms comparison for (a) State-variable and (b) Spice
simulations in forward mode and 𝑽𝒃𝒂𝒕 = 𝟑𝟓𝟎𝑽
In Figure 4.22 the results are shown for a converter operating in reverse mode, also
at 𝑉𝑏𝑎𝑡 = 350𝑉, when the converter is operating in region II. Here the rectifier’s input
current is equal to 𝑖𝐿𝑠1 and, due to non-conduction mode occurrence, the rectifier
89
voltage is not a square wave as in the previous case. For the Spice simulation, noise is
observed on the input voltage to the rectifier (𝑣1 ) during the non-conduction period.
For all cases analysed, waveforms obtained from Spice simulation are very similar
to the state-variable results.
(a)
(b)
Figure 4.22: Waveforms comparison for (a) State-variable and (b) Spice
simulations in reverse mode and 𝑽𝒃𝒂𝒕 = 𝟑𝟓𝟎𝑽
After verifying the accuracy of the models through simulation, the execution time
for each method is compared. The average execution time for each model analysed is
shown in Table 4.12. It is verified that cyclic-averaging is the fastest method, being
nearly 37 times faster than Spice, the slowest method. Both Spice and state-variable
90
simulations are performed considering a simulation time of 3ms and maximum step
size of 10 ns. Note that while FMA and cyclic-averaging methods directly calculate
steady-state results, Spice and state-variable are time-domain models that include the
transient response and simulation time must be adequately chosen to allow the system
to reach steady-state.
Modelling technique
State- Cyclic-
FMA Spice
variable averaging
Execution time (s) 2.045 5.888 0.165 6.084
4.8 Conclusions
In this chapter FMA, state-variable and cyclic-averaging models were developed
for analysis of a 3.5kW CLLC bidirectional converter operating under frequency
modulation.
In Chapter 5 the models developed here will be adapted for the analysis of a phase-
shifted modulated variant of the CLLC converter.
91
4.9 References
[1] Z. U. Zahid, Z. M. Dalala, R. Chen, B. Chen, and J. S. Lai, “Design of bidirectional
DC-DC resonant converter for Vehicle-to-Grid (V2G) applications,” IEEE
Transactions on Transportation Electrification, vol. 1, no. 3, pp. 232–244, 2015.
[2] W. Chen, P. Rong, and Z. Lu, “Snubberless bidirectional DC-DC converter with
new CLLC resonant tank featuring minimized switching loss,” IEEE Transactions
on Industrial Electronics, vol. 57, no. 9, pp. 3075–3086, 2010.
[3] J. Sun and H. Grotstollen, “Symbolic Analysis Methods for Averaged Modeling of
Switching Power Converters,” IEEE Transactions on Power Electronics, vol. 12,
no. 3, pp. 537–546, 1997.
[4] C. Gould, “Rapid Analysis and Design of CLL Resonant Power Converters,” PhD
Thesis. University of Sheffield, 2006.
[6] J. Deng, S. Li, S. Hu, C. C. Mi, and R. Ma, “Design Methodology of LLC
Resonant Converters for Electric Vehicle Battery Chargers,” IEEE Transactions on
Vehicular Technology, vol. 63, no. 4, pp. 1581–1592, 2014.
[7] B. Sen, Chaohui Liu, Jiabin Wang, C. Gould, and K. Colombage, “A CLLC
Resonant Converter Based Bidirectional EV Charger with Maximum Efficiency
Tracking,” 8th IET International Conference on Power Electronics, Machines and
Drives (PEMD 2016), pp. 1–6, 2016.
92
5 Modelling of Phase-shift Modulated
CLLC Converter
5.1 Introduction
In the previous chapter the FMA, state-variable and cyclic-averaging modelling
techniques were applied to analyse a bidirectional CLLC resonant converter operating
under frequency modulation. From comparison of the results taken from each
modelling approach with a Spice-based simulation, it was observed that cyclic-
averaging and state-variable were the most accurate models, with the cyclic-averaging
having the advantage of being the method with fastest execution time.
In this chapter similar analysis will be performed for the phase-shift modulated
variant of the CLLC converter, considering bidirectional power transfer and two types
of phase-shift modulation: Single Phase-Shift (SPS) and Pulse-Phase Modulation
(PPM), also called Triple-Phase-Shift modulation (TPS).
From the results obtained in the previous chapter, it was observed that the FMA
method suffers from accuracy problems, but due to its simple and easy implementation,
this analysis will be applied here again to study the influence of phase-shift modulation
angles on the converter’s behaviour. A low-power converter design is presented and the
state-variable and cyclic-averaging methods will be employed to estimate the average
output current of a converter operating under SPS and PPM modulation. The results
obtained from the aforementioned modelling methods will be verified against a Spice
simulation and the execution time will be compared.
93
Figure 5.1: CLLC topology
In this chapter, two types of phase-shift modulation are considered for the
operation of the CLLC converter. The first and simpler method analysed is the Single
Phase-Shift (SPS) [1]–[3], the main advantage of this being easy implementation. In
this configuration the switching frequency is fixed, duty cycle is kept constant at 50%
for all switches, and a phase-shift angle (−90° ≤ 𝜙 ≤ 90°) is introduced between the
output voltages of bridge 1 and bridge 2 from Figure 5.1. Typical waveforms of bridge
voltages 𝑣1 and 𝑣2 for a converter operating in reverse mode are shown in Figure 5.2-a.
The output power magnitude and power flow direction are controlled using only the
phase-shift angle, where positive values of 𝜙 (primary voltage 𝑣1 lags secondary
voltage 𝑣2 by 𝜙) result in forward operation while negative values (primary voltage 𝑣1
leads secondary voltage 𝑣2 by 𝜙) are used for reverse mode implementation. The major
drawback of SPS modulation is the high circulating current and, consequently,
conduction losses when operating under partial loading [4], [5]. As the phase-shift
angle decreases, the output current decreases but circulating current increases, resulting
in increased losses.
94
voltages 𝑣1 and 𝑣2 from Figure 5.1 are shown in Figure 5.2-b. For a simpler control, 𝛼1
is maintained equal to 𝛼2 in many applications.
Apart from the FMA analysis, where the influence of the phase-shift angles will be
evaluated, when analysing the PPM modulation case for state-variable and cyclic-
averaging models in this chapter the following considerations are made:
𝛼1 = 𝛼2 = 𝛼 and 𝜙 = ±90°
(a) (b)
Figure 5.2: Bridge output voltages for (a) SPS modulation, operating in
reverse mode and (b) PPM modulation, operating in forward mode
95
Based on the converter design procedure proposed in [10], inductor 𝐿𝑚 resonates
′
with 𝐿𝑠1 𝐶𝑠1 and 𝐶𝑠2 , as a result, the reactances in each leg of the T-network in Figure
5.3 are represented using a base reactance 𝑋𝑛 , where 𝑋𝑛 = 𝑋𝐶1 − 𝑋𝐿1 = 𝑋𝐿𝑚 = 𝑋𝐶2′ .
Using the Fourier series representation, the bridge voltages referred to the primary
are defined in (5.1). For the case of a converter operating under SPS modulation, the
angles 𝛼1 and 𝛼2 are fixed and equal to 180°.
∞
4𝑉𝐷𝐶 1 𝑖𝛼1
𝑣1 (𝑡) = ∑ cos(𝑖𝜔𝑡) sin ( )
𝜋 𝑖 2
𝑖=1,3,…
∞
4𝑛𝑉𝑏𝑎𝑡 1 𝑖𝛼2
𝑣2 (𝑡) = ∑ cos(𝑖𝜔𝑡 + 𝑖𝜙) sin ( )
𝜋 𝑖 2
𝑖=1,3,…
(5.1)
Applying FMA, only the fundamental component is considered, and the voltage
phasors are defined as:
4𝑉𝐷𝐶 𝛼1
𝑽𝟏 = sin ( )
𝜋√2 2
4𝑛𝑉𝑏𝑎𝑡 𝛼2
𝑽𝟐 = sin ( ) (cos(𝜙) + 𝑗 sin(𝜙))
𝜋√2 2
(5.2)
From circuit analysis, the resonant tank currents are calculated.
𝑽𝟐
𝑰𝟏 =
𝑗𝑋𝑛
𝑽𝟏
𝑰𝟐 = −
𝑗𝑋𝑛
96
𝑽 𝟏 + 𝑽𝟐
𝑰𝒎 =
𝑗𝑋𝑛
(5.3)
The RMS values of currents are obtained calculating the magnitude of the phasors
defined in (5.3). A generic phasor Z is defined, composed of a real component a and
imaginary part b.
𝒁 = 𝑎 + 𝑗𝑏
(5.4)
|𝒁| = √𝑎2 + 𝑏 2
(5.5)
The magnitude of primary and secondary currents are presented in (5.6). Note that
phasor 𝑰𝟐 is referred to the primary side of the circuit, therefore, the magnitude of the
secondary current (𝑰𝒔𝒆𝒄 ) must be calculated considering the transformers turns ratio as
in (5.6).
4𝑛𝑉𝑏𝑎𝑡 𝛼2
|𝑰𝟏 | = sin ( )
𝜋𝑋𝑛 √2 2
4𝑛𝑉𝐷𝐶 𝛼1
|𝑰𝒔𝒆𝒄 | = 𝑛|𝑰𝟐 | = sin ( )
𝜋𝑋𝑛 √2 2
(5.6)
For the FMA analysis, parasitic resistances are not considered and, consequently,
the transferred power is calculated using (5.2), (5.3) and the active power equations
𝑃 = 𝑅𝑒[𝑽𝟐 𝑰𝟐 ∗ ] or 𝑃 = 𝑅𝑒[𝑽𝟏 𝑰𝟏 ∗ ], resulting in equation (5.7). Also, it was previously
shown in [10] that, for a phase-shift modulated CLLC converter, most of the power is
transferred by the fundamental component.
97
8𝑛𝑉𝐷𝐶 𝑉𝑏𝑎𝑡 𝛼1 𝛼2
𝑃𝑜𝑢𝑡 = 2
sin(𝜙) sin ( ) sin ( )
𝜋 𝑋𝑛 2 2
(5.7)
From (5.7) it is confirmed that the maximum value of output power is obtained
when 𝜙 = ±90° and 𝛼1 = 𝛼2 = 180°, as discussed in the previous section. In section
5.6.2 the equation description here obtained will be implemented using MATLAB and
the output power and resonant tank currents will be calculated for different angles and
compared to a Spice simulation.
𝑟2 ′ = 𝑛2 (𝑟𝐶2 + 2. 𝑟𝑆𝑊𝐼𝑇𝐶𝐻 )
(5.8)
98
Figure 5.4: Equivalent circuit for state-variable analysis, forward mode
Similar to the bridge voltage definition performed in Chapter 3 for the DAB
converter, for operation under SPS modulation, the output voltages of the H-bridges are
simplified to square waves with amplitude dependent of the DC bus or battery voltages,
as defined in (5.9). For the PPM modulation case, a zero-voltage level is added
depending on the values of 𝛼1 and 𝛼2 , as previously shown in Figure 5.2.
𝑣1 = ±𝑉𝑑𝑐
𝑣2 = ±𝑉𝑏𝑎𝑡
(5.9)
To obtain the state-variable piecewise linear equation description, the fast and slow
subsystems are analysed and the coupling equation is determined based on the
equivalent circuit in Figure 5.4. From circuit analysis, the fast subsystem equations are
given by:
𝑑𝑣𝐶𝑠1 𝑖𝐿𝑠1
=
𝑑𝑡 𝐶𝑠1
(5.11)
99
For the slow subsystem:
Once the equations for fast, slow and coupling systems are defined, the output
voltage and current are calculated.
𝑉𝑜𝑢𝑡 − 𝑉𝑏𝑎𝑡
𝐼𝑜𝑢𝑡 =
𝑟𝑏𝑎𝑡
(5.16)
100
Using basic circuit analysis, the fast subsystem equations are obtained:
𝑑𝑣𝐶𝑠1 𝑖𝐿𝑠1
=
𝑑𝑡 𝐶𝑠1
(5.18)
The coupling equation is now defined based on the state of the output bridge
voltage 𝑣1 (𝑡), the case of 𝑣1 (𝑡) = 0 will only happen during PPM modulation.
𝑉𝑜𝑢𝑡 − 𝑉𝑑𝑐
𝐼𝑜𝑢𝑡 =
𝑟𝑑𝑐
(5.23)
101
5.4.1 Implementation in Simulink
After the set of piecewise linear equations is obtained for forward and reverse
operation, the state-variable model is implemented in Simulink.
Firstly, the bridge voltages are defined as shown in Figure 5.6. The “pulse
generator” and “variable time delay” blocks are used to implement the waveforms of
𝑣1 (𝑡) and 𝑣2 (𝑡) with a phase-shift between the legs of each bridge (angles 𝛼1 and 𝛼2 )
and between the primary and secondary bridge (angle 𝜙).
Once the output voltages of the bridges are defined, the fast and slow subsystems
and coupling equation are implemented as shown in Figures 5.7, 5.8 and 5.9.
102
Figure 5.8: Slow subsystem, forward mode
103
5.5 Cyclic-averaging analysis
In the previous section a piecewise linear state-variable model was obtained, which
will serve as base for application of the cyclic-averaging method. The set of equations
for forward mode are reorganized for a matrix form representation:
104
𝑖𝐿𝑠1 (𝑡) 𝑖𝐿𝑠1 (𝑡)
𝑣𝐶𝑠1 (𝑡) 𝑣𝐶𝑠1 (𝑡)
𝑑
𝑖𝐿𝑚 (𝑡) = 𝐴𝑟 𝑖𝐿𝑚 (𝑡) + 𝐵𝑟
𝑑𝑡
𝑣𝐶𝑠2′ (𝑡) 𝑣𝐶𝑠2′ (𝑡)
[ 𝑣𝐶𝑖 (𝑡) ] [ 𝑣𝐶𝑖 (𝑡) ]
(5.25)
where:
(𝑟1 + 𝑟2′ ) 1 𝑟𝐶2 ′ 1
− − − − 0
𝐿𝑠1 𝐿𝑠1 𝐿𝑠1 𝐿𝑠1
1
0 0 0 0
𝐶𝑠1
𝑟𝐶2 ′ (𝑟𝐿𝑚 + 𝑟𝐶2 ′ ) 1
𝐴𝑟 = − 0 − − 0
𝐿𝑚 𝐿𝑚 𝐿𝑚
1 1
0 − 0 0
𝐶𝑠2 ′ 𝐶𝑠2 ′
1
0 0 0 0 −
[ 𝐶𝑖 (𝑟𝑑𝑐 + 𝑟𝐶𝑖 )]
The typical bridge voltage and current sequence for a converter operating under
SPS modulation is shown in Figure 5.10. The bridge voltages are square waves with
two possible states, ±𝑉𝑑𝑐 for 𝑣1 and ±𝑉𝑏𝑎𝑡 for 𝑣2 (𝑡), and the bridge current is
calculated using (5.15) or (5.22) depending on the power flow direction. Four operation
modes are identified for a cycle, considering the beginning of a cycle at the transition of
the output bridge voltage (𝑣2 for forward operation or 𝑣1 for reverse operation) from
105
negative to positive. The mode descriptions, containing the state of the bridge voltages
and output current during each mode, is presented in Table 5.1 for forward operation
and in Table 5.2 for reverse operation.
(a) (b)
Figure 5.10: Typical bridge voltage and current sequence for SPS operation in
(a) forward mode and (b) reverse mode
For the PPM modulation case, the bridge voltages are no longer perfect square
waves and have three possible states, 0 and ±𝑉𝑑𝑐 for 𝑣1 (𝑡), and 0 and ±𝑉𝑏𝑎𝑡 for 𝑣2 (𝑡).
The occurrence and duration of the zero-voltage level depends on the phase-shift angle
106
𝛼 between the legs of each H-bridge, where 𝛼1 = 𝛼2 = 𝛼. The phase-shift between the
bridge output voltages, 𝜙, is kept constant at ±90°, where the sign determines the
power flow direction.
(a) (b)
Figure 5.11: Typical bridge voltage and current sequence for PPM operation
and 𝟗𝟎° ≤ 𝜶 ≤ 𝟏𝟖𝟎° in (a) forward mode and (b) reverse mode
(a) (b)
Figure 5.12: Typical bridge voltage and current sequence for PPM operation
and 𝜶 < 𝟗𝟎° in (a) forward mode and (b) reverse mode
For both cases, eight operation modes are identified and, considering the beginning
of a cycle when the output bridge voltage (𝑣2 for forward and 𝑣1 for reverse operation)
107
becomes positive, the mode descriptions for the full 𝛼 range are presented in Tables
5.3, 5.4, 5.5 and 5.6.
Table 5.3: Mode descriptions for PPM modulation and 𝟗𝟎° ≤ 𝜶 ≤ 𝟏𝟖𝟎°, forward
mode
Table 5.4: Mode descriptions for PPM modulation and 𝟗𝟎° ≤ 𝜶 ≤ 𝟏𝟖𝟎°, reverse
mode
108
Table 5.5: Mode descriptions for PPM modulation and 𝜶 ≤ 𝟗𝟎°, forward mode
Table 5.6: Mode descriptions for PPM modulation and 𝜶 ≤ 𝟗𝟎°, reverse mode
Based on the mode descriptions, the values of 𝑣1 (𝑡), 𝑣2 (𝑡) and 𝑖𝑏𝑟𝑖𝑑𝑔𝑒 (𝑡) are
substituted on the state-variable equations (5.24) or (5.25) and the dynamic and input
matrices, 𝐴𝑖 and 𝐵𝑖 respectively, are determined for each mode i, where 𝑖 = 1, 2, 3, 4
for SPS modulation and 𝑖 = 1, 2, 3, 4, 5, 6, 7, 8 for PPM modulation cases.
Once the state-variable description is determined for all modes, the duration of
each mode, 𝑑𝑖 𝑇, must be calculated based on the period T calculated from the switching
frequency and the duty cycle, or normalised time interval, for each mode, 𝑑𝑖 .
109
angles 𝛼 and 𝜙. From waveform analysis, the duty cycles are calculated solely based on
the leg and bridge phase-shift angles. Therefore, the duty cycle calculation here is
significantly simpler when compared to the frequency modulated variant previously
analysed in Chapter 4.
When operating under SPS modulation, as shown in Figure 5.10, the leg phase-
shift angle, 𝛼, is kept constant at 180° and the duration of each mode is determined
based on the phase-shift angle between the primary and secondary bridge, where 0° ≤
|𝜙| ≤ 90°. For the first mode:
|𝜙|
𝑑1 =
360
(5.26)
The remaining duty cycles are determined by waveforms symmetry:
𝑑2 = 0.5 − 𝑑1 , 𝑑3 = 𝑑1 and 𝑑4 = 𝑑2
(5.27)
For the PPM modulation case, the normalised time duration of each mode is
determined analysing the waveforms from Figures 5.11 and 5.12. The bridge angle 𝜙 is
now constant and, as a result, the duty values are calculated based on the values of the
leg phase-shift angle 𝛼. When 90° ≤ 𝛼 ≤ 180°, the duties are determined as in (5.28).
−90 + 𝛼
𝑑1 = = 𝑑3 = 𝑑5 = 𝑑7
360
180 − 𝛼
𝑑2 = = 𝑑4 = 𝑑6 = 𝑑8
360
(5.28)
Now considering the case of 𝛼 ≤ 90°, the duty cycle values are calculated as
follows:
𝛼
𝑑1 = = 𝑑 3 = 𝑑5 = 𝑑7
360
90 − 𝛼
𝑑2 = = 𝑑4 = 𝑑6 = 𝑑8
360
(5.29)
110
When 𝛼 = 90° operation is reduced from eight to four modes. In this case either
set of mode descriptions and duty calculation presented for PPM modulation can be
used, since the extra modes will be eliminated in the calculation of the modes duration.
According to the duty calculation equation (5.28), when 𝛼 = 90° the values of duties
𝑑1 , 𝑑3 , 𝑑5 and 𝑑7 will be equal to zero, cancelling modes 𝑀1 , 𝑀3 , 𝑀5 and 𝑀7 from
Tables 5.3 and 5.4. Similarly, from (5.29) modes 𝑀2 , 𝑀4 , 𝑀6 and 𝑀8 , from Tables 5.5
and 5.6, are cancelled, resulting in the same mode description, as shown in Tables 5.7
and 5.8.
Table 5.7: Mode descriptions for PPM modulation and 𝜶 = 𝟗𝟎°, forward mode
Table 5.8: Mode descriptions for PPM modulation and 𝜶 = 𝟗𝟎°, reverse mode
𝑴𝟏 𝑉𝑑𝑐 0 𝑖𝐿𝑠1
𝑴𝟐 0 𝑉𝑏𝑎𝑡 0
𝑴𝟑 −𝑉𝑑𝑐 0 −𝑖𝐿𝑠1
𝑴𝟒 0 −𝑉𝑏𝑎𝑡 0
After the state-variable equations, mode descriptions and duty cycles are
determined for both types of phase-shift modulation, the cyclic method equations from
Chapter 2 are used to model the converter and calculate the average value of the state-
variables. Considering the resistances of the output filter capacitors (𝑟𝐶𝑓 and 𝑟𝐶𝑖 )
negligible, the average output current is calculated.
𝑣𝐶𝑓,𝑎𝑣𝑔 −𝑉𝑏𝑎𝑡
𝐼𝑜𝑢𝑡,𝑎𝑣𝑔 = , for forward operation
𝑟𝑏𝑎𝑡
𝑣𝐶𝑖,𝑎𝑣𝑔 −𝑉𝑑𝑐
𝐼𝑜𝑢𝑡,𝑎𝑣𝑔 = , for reverse operation
𝑟𝑑𝑐
111
(5.30)
The FMA equations will be used to evaluate the influence of phase-shift angles on
the converter operation, while the state-variable and cyclic-averaging models are used
to estimate the converter’s output current based on a predetermined range of phase-shift
angles. The results obtained from the models’ simulations are validated against a
component-based Spice simulation.
In this chapter, at simulation stage the circuit is considered nearly ideal, the
influence of resistances and parasitic elements will only be analysed in the next chapter
with the construction of a prototype and comparison between simulation and
experimental results.
The circuit used for state-variable, cyclic-averaging and Spice simulations is shown
in Figure 5.13. The converter is considered nearly ideal, only a small resistance of 0.1Ω
is considered at the input and output of the resonant tank (𝑟1 and 𝑟2 ). Resistances 𝑟𝑏𝑎𝑡
and 𝑟𝑑𝑐 are used for measurement of the output current in the cyclic-averaging method
112
and have small values (0.01Ω) to not significantly influence the output power
magnitude. Note that, as previously explained in section 5.3, for FMA analysis all
resistances are neglected.
Once the input and output voltages are specified, the DC voltage conversion ratio
is calculated:
𝑉𝑏𝑎𝑡
𝐷𝐶𝑟𝑎𝑡𝑖𝑜 = 𝑛
𝑉𝑑𝑐
(5.31)
For the resonant frequency calculation, inductor 𝐿𝑚 forms a resonator with 𝐿𝑠1 𝐶𝑠1
𝐶𝑠2⁄
and 𝑛2 , consequently the resonant network is tuned to the switching frequency as
follows.
1 𝑛2
𝜔𝑟2 2
= (2𝜋𝑓𝑠 ) = =
(𝐿𝑠1 + 𝐿𝑚 )𝐶𝑠1 𝐿𝑚 𝐶𝑠2
(5.32)
113
Additionally, each leg of the T-network is designed to have the same reactance, as
defined in (5.33).
𝑋𝐿𝑚 = 𝑛2 𝑋𝐶2 = 𝑋𝑛
𝑋𝐶1 = (1 + 𝑘) 𝑋𝑛
𝑋𝐿1 = 𝑘𝑋𝑛
(5.34)
8𝑛𝑉𝐷𝐶1 𝑉𝐷𝐶2
𝑃𝑜𝑢𝑡_𝑚𝑎𝑥 =
𝜋 2 𝑋𝑛
(5.35)
Considering the nominal voltage, turns ratio, and output power previously
specified, the base reactance is calculated from (5.35) and 𝑋𝑛 = 16.98 Ω.
Substituting the value of 𝑋𝑛 in (5.34) and using the reactance definition (𝑋𝐿 = 𝜔𝐿
and 𝑋𝐶 = 1⁄𝜔𝐶 ), the values of inductance and capacitances are calculated considering
𝑘 = 1, 𝑘 = 2 and 𝑘 = 5, as shown in Table 5.9.
114
Table 5.9: Resonant tank parameters calculated for different values of k
Once the value of k is defined, all circuit elements are determined and the
parameters obtained at the end of the design process are listed in Table 5.11.
115
Table 5.10: Spice simulation results for different values of k
Parameter Value
𝑽𝒅𝒄 48 V
𝑽𝒃𝒂𝒕 12 V
𝑪𝒇 , 𝑪𝒊 300 µF
𝑳𝒔𝟏 54.04 µH
𝑪𝒔𝟏 31.24 nF
𝑳𝒎 27.02 µH
𝑪𝒔𝟐 1.5 µF
𝒓𝟏 , 𝒓𝟐 ′ 0.1 Ω
𝒓𝒅𝒄 , 𝒓𝒃 0.01 Ω
Similar to the phase-shift angles, the input and output voltage sources also affect
the output power calculation and could be used as extra control variables. In Table 5.12
it is shown how the output power changes when the DC bus and battery voltages
116
deviate from nominal values, where a variation of 15% in the DC bus and battery
voltages, simultaneously, results in a change of nearly 32% in the output power.
Equations (5.3), (5.6) and (5.7) obtained during FMA analysis are now used to
calculate the resonant tank currents and output power, evaluating the influence of
phase-shift angles 𝛼1 , 𝛼2 and 𝜙 on the converter operation.
The angles 𝛼1 and 𝛼2 are tested for a range from 0 to 180°, with a step size of 5°.
Three values of phase-shift angle 𝜙 were evaluated: 22.5°, 45° and 90°. The 3D graphs
obtained for a converter operating in forward mode, with maximum (𝜙 = 90°) and
minimum (𝜙 = 22.5°) values of bridge phase-shift angle are presented in Figures 5.14,
5.15, 5.16 and 5.17.
(a) (b)
Figure 5.14: Output power versus 𝜶𝟏 and 𝜶𝟐 for (a) 𝝓 = 𝟗𝟎° (b) 𝝓 = 𝟐𝟐. 𝟓°
117
(a) (b)
Figure 5.15: Primary current magnitude versus 𝜶𝟏 and 𝜶𝟐 for (a) 𝝓 = 𝟗𝟎° (b) 𝝓 =
𝟐𝟐. 𝟓°
(a) (b)
Figure 5.16: Secondary current magnitude versus 𝜶𝟏 and 𝜶𝟐 for (a) 𝝓 = 𝟗𝟎° (b) 𝝓 =
𝟐𝟐. 𝟓°
(a) (b)
Figure 5.17: Magnetizing branch current magnitude versus 𝜶𝟏 and 𝜶𝟐 for (a) 𝝓 =
𝟗𝟎° (b) 𝝓 = 𝟐𝟐. 𝟓°
118
The phase-shift angle between primary and secondary bridges, 𝜙, functions as a
control variable for the output current since, considering fixed input and output
voltages, the output power decreases as the angle 𝜙 decreases. From the results
presented it is also observed that the magnitude of primary and secondary currents is
independent of the angle 𝜙. The magnitude of the primary current only depends on the
phase-shift angle between the legs of the secondary bridge (𝛼2 ) while the secondary
current is dependent of 𝛼1 , the phase-shift angle between the legs of the primary bridge.
This behaviour is one of the main disadvantages of the SPS modulation technique,
where 𝛼1 and 𝛼2 are maintained at the maximum value (180°) and 𝜙 is the only control
variable. In this case, for low values of 𝜙 the converter’s output current decreases but
the input and output currents of the resonant tank are not affected, their magnitude is
always at maximum value, resulting in an increase of the circulating current, system
losses and stress on switching devices.
To evaluate the accuracy of the FMA method, the results obtained are compared to
a nearly ideal Spice simulation. Contrary to FMA, the Spice simulation is not ideal
since this model includes the resistances 𝑟1, 𝑟2 , 𝑟𝑏𝑎𝑡 and 𝑟𝑑𝑐 , as explained in section
5.6.1. The two models are compared considering the maximum bridge phase-shift (𝜙 =
90°) and results are shown in Figures 5.18, 5.19, 5.20 and 5.21. Since when 𝛼1 = 0 or
𝛼2 = 0 the output power is nearly zero, the range here analysed is defined based on a
converter operating with angles 𝛼1 and 𝛼2 from 18° to 180°, with a step size of 18°.
(a) (b)
Figure 5.18: Output power for (a) FMA and (b) Spice simulations
119
(a) (b)
Figure 5.19: Primary current for (a) FMA and (b) Spice simulations
(a) (b)
Figure 5.20: Secondary current for (a) FMA and (b) Spice simulations
(a) (b)
Figure 5.21: Magnetizing branch current for (a) FMA and (b) Spice simulations
120
Compared to FMA, the output power in Spice simulations decreases more rapidly
as the phase-shift angles decrease. The magnitude of primary current is very similar for
both models, but Spice results for secondary current are slightly dependent of 𝛼2 ,
especially for low values of 𝛼1 , while in FMA simulations the secondary current
magnitude is independent of 𝛼2 . A difference in shape was also noticed between Spice
and FMA results for the magnitude of the magnetizing current in Figure 5.21. The
differences between Spice and FMA are mainly due to the influence of the circuit
resistances and the approximation adopted in the FMA method that ignores the
influence of harmonics.
The relative error between Spice and FMA results was calculated for each point
using equation (5.36) and results are shown in Tables A1.1, A1.2, A1.3 and A1.4, in
Appendix A.1. Note that, in general, reduced error is obtained in the region of 𝛼1 =
𝛼2 , and the discrepancy between Spice and FMA results tends to increase as the
difference between 𝛼1 and 𝛼2 increases, especially for output power and magnetizing
current results.
𝑋𝑠𝑝𝑖𝑐𝑒 − 𝑋𝑓𝑚𝑎_𝑒𝑞
Relative error (%) =
𝑋𝑓𝑚𝑎_𝑒𝑞
(5.36)
where X is the variable under analysis, which can be |𝐼1 |, |𝐼2 |, |𝐼𝑚 | or 𝑃𝑜𝑢𝑡 .
Overall, FMA is a simple and useful method for fast analysis but, as also observed
in the previous chapter for the frequency modulated CLLC converter, this technique
suffers from accuracy problems due to the fundamental and ideal circuit
approximations.
Following the FMA analysis, to further evaluate the influence of the leg phase-shift
angles 𝛼1 and 𝛼2 on the converter operation, a Spice simulation is used to calculate the
converter’s efficiency for the full operation range of 𝛼1 and 𝛼2 . The 3D plot obtained is
shown in Figure 5.22 and the data points are listed in Table A1.5, in Appendix A.1.
The results show that the highest values of efficiency are obtained near the region
of 𝛼1 = 𝛼2 and a considerable drop in efficiency is observed as the difference between
𝛼1 and 𝛼2 significantly increases. Therefore, for the next step of analysis and
121
verification of state-variable and cyclic-averaging models, the converter will be
considered operating with 𝛼1 = 𝛼2 = 𝛼 when applying the PPM modulation technique.
Additionally, the efficiency also reduces as the bridge phase-shift angle decreases.
When 𝜙 drops from 90° to 45°, while maintaining 𝛼1 = 𝛼2 = 90°, the efficiency
reduces from 98.56% to 97.83%. At 22.5° the efficiency drops to 95.89%.
Following the FMA and efficiency analysis, the state-variable and cyclic-averaging
models, obtained in sections 5.4 and 5.5 respectively, will be implemented and verified
against a Spice simulation. The converter is analysed considering SPS and PPM
modulation techniques and forward and reverse power flow directions.
For the SPS modulation analysis the angle 𝛼 is fixed at 180° and the output current
is measured as the phase-shift angle 𝜙 increases from 10° to 90° in steps on 10°. For the
PPM modulation implementation, the angle 𝜙 is kept at its maximum value (90°) and
the variation of the angle 𝛼 is implemented using the normalized angle 𝛼𝑟𝑎𝑡𝑖𝑜 , defined
in (5.37), where the range of 𝛼𝑟𝑎𝑡𝑖𝑜 evaluated is from 0.1 (𝛼 = 18°) to 1 (𝛼 = 180°),
with a step size of 0.1.
122
𝛼
𝛼𝑟𝑎𝑡𝑖𝑜 =
180
(5.37)
The results obtained for the SPS modulation case are shown in Figure 5.23. The
three models have very similar results and differences are hard to identify in the graph,
therefore, for a meticulous analysis, the values of average output current for each value
of modulation angle are listed in Table 5.13. The graphs obtained for PPM modulation
are shown in Figure 5.24 while the data points are presented in Table 5.14.
(a) (b)
Figure 5.23: Simulation results for SPS modulation (a) forward mode and (b)
reverse mode
123
(a) (b)
Figure 5.24: Simulation results for PPM modulation (a) forward mode and (b)
reverse mode
124
Table 5.14: Simulation results for PPM modulation
For further verification, the behaviour of the circuit variables during a cycle is
analysed and waveforms obtained from the state-variable simulation are compared to
Spice. All state-variables and bridge voltages were verified for different points across
the full range of phase-shift angles, considering forward and reverse operation. The
graphs containing the bridge voltages 𝑣1 and 𝑣2 , primary inductor current 𝑖𝐿𝑠1 and
magnetizing inductor current 𝑖𝐿𝑚 for 𝛼𝑟𝑎𝑡𝑖𝑜 = 0.25 and 𝜙 = 90° in forward mode are
shown in Figure 5.25. From the results analysis, it is noticed a very similar behaviour
between Spice and state-variable waveforms, confirming the accuracy of the state-
variable method.
125
(a) (b)
Figure 5.25: Waveform comparison for 𝜶𝒓𝒂𝒕𝒊𝒐 = 𝟎. 𝟐𝟓 (a) state-variable and (b)
Spice, forward mode
Based on the cyclic analysis and mode descriptions performed in section 5.5, a
cycle is divided in 8 modes when operating under PPM modulation. To verify the
accuracy of the cyclic-averaging method during a full cycle, the steady-state values of
the resonant tank state-variables (𝑖𝐿𝑠1 , 𝑣𝐶𝑠1 , 𝑖𝐿𝑚 and 𝑣𝐶𝑠2 ) are calculated at the
beginning of each mode, points 𝑡0 , 𝑡1 , 𝑡2 , 𝑡3 , 𝑡4 , 𝑡5 , 𝑡6 and 𝑡7 from Figure 5.25. The
values obtained from the cyclic-averaging simulation (Table 5.16) are compared to
state-variable (Table 5.15) and Spice (Table 5.17) results considering 𝛼𝑟𝑎𝑡𝑖𝑜 = 0.25
and 𝜙 = 90°, in forward mode. This investigation, as well as the state-variable
waveform analysis, were performed for different values of 𝛼𝑟𝑎𝑡𝑖𝑜 and 𝜙 and results
were similarly accurate across the full modulation range. Part of the error from Spice
and state-variable results compared to cyclic-averaging is due to the dependency on the
simulation step size and consequent difficulty in measuring the current and voltage
values in the Spice and state-variable graphs at the exact point in time each mode starts.
126
Table 5.15: State-variable model results for variables states considering
𝜶𝒓𝒂𝒕𝒊𝒐 = 𝟎. 𝟐𝟓, forward mode
State-variable model
𝒕𝟎 𝒕𝟏 𝒕𝟐 𝒕𝟑 𝒕𝟒 𝒕𝟓 𝒕𝟔 𝒕𝟕
𝒊𝑳𝒔𝟏 (A) -0.038 0.258 1.050 1.739 0.052 -0.256 -1.049 -1.742
𝒗𝑪𝒔𝟏 (V) -66.970 -60.630 -31.930 29.720 66.980 60.720 32.070 -29.560
𝒊𝑳𝒎 (A) -2.454 -0.071 0.596 1.588 2.473 0.073 -0.595 -1.588
𝒗𝑪𝒔𝟐 (V) -2.098 2.851 4.459 6.021 2.203 -2.858 -4.465 -6.025
Cyclic-averaging model
𝒕𝟎 𝒕𝟏 𝒕𝟐 𝒕𝟑 𝒕𝟒 𝒕𝟓 𝒕𝟔 𝒕𝟕
𝒊𝑳𝒔𝟏 (A) -0.047 0.254 1.049 1.744 0.047 -0.254 -1.049 -1.744
𝒗𝑪𝒔𝟏 (V) -66.913 -60.695 -31.819 29.383 66.913 60.695 31.819 -29.383
𝒊𝑳𝒎 (A) -2.474 -0.080 0.591 1.580 2.474 0.080 -0.591 -1.580
𝒗𝑪𝒔𝟐 (V) -2.204 2.830 4.471 6.036 2.204 -2.830 -4.471 -6.036
Table 5.17: Spice model results for variables states considering 𝜶𝒓𝒂𝒕𝒊𝒐 = 𝟎. 𝟐𝟓,
forward mode
Spice model
𝒕𝟎 𝒕𝟏 𝒕𝟐 𝒕𝟑 𝒕𝟒 𝒕𝟓 𝒕𝟔 𝒕𝟕
𝒊𝑳𝒔𝟏 (A) -0.051 0.256 1.057 1.748 0.050 -0.256 -1.054 -1.750
𝒗𝑪𝒔𝟏 (V) -67.174 -60.934 -31.837 29.322 67.134 60.955 32.065 -29.451
𝒊𝑳𝒎 (A) -2.476 -0.076 0.598 1.584 2.477 0.075 -0.593 -1.584
𝒗𝑪𝒔𝟐 (V) -2.217 2.834 4.475 6.033 2.201 -2.829 -4.463 -6.040
After the accuracy of the models is verified, the execution time for the three
methods is compared, as shown in Table 5.18. Six measurements were performed for
each method and the average execution time is calculated. Both state-variable and Spice
simulations are performed with simulation time of 8ms and step size of 10 ns. The
127
values calculated with the FMA and cyclic-averaging equation descriptions are directly
steady-state, while Spice and state-variable simulations reach steady-state around 7 ms.
The execution time for the FMA and cyclic-averaging methods was measured
considering the first time the code is compiled, to include the memory allocation time.
After the first run, the variables are already defined and stored in memory and
execution time drops to an average of 0.005 seconds for the cyclic method, nearly 4000
times faster than Spice, and 0.238 seconds for FMA. Since the execution of these
methods is considerably fast, this allocation time significantly influences the calculation
of average execution time, while for state-variable and Spice simulations no significant
difference in execution time was observed between the first and subsequent executions.
Modelling technique
Cyclic- State-
FMA Spice
averaging variable
Average 0.594 0.136 24.583 19.568
execution time (s)
5.7 Conclusions
In this chapter three modelling techniques were employed to describe the operation
of a bidirectional CLLC converter operating under phase-shift modulation. The
equation descriptions for FMA, state-variable and cyclic-averaging models were
obtained, and simulations were performed considering two types of phase-shift
modulation techniques: Single Phase-Shift (SPS) and Pulse-Phase Modulation (PPM).
For accuracy verification, the results were compared to a component-based Spice
simulation.
From FMA analysis, it was shown that for higher values of 𝛼1 and 𝛼2 and in a
range around 𝛼1 = 𝛼2 this technique shows accurate results compared to Spice. Still,
the FMA method is not as accurate as the state-variable and cyclic-averaging
techniques. For low values of 𝛼1 and 𝛼2 and as the difference between these angles
increases the accuracy of the FMA model tends to reduce.
128
From Spice results it was also shown that efficiency reduces as the difference
between 𝛼1 and 𝛼2 significantly increases and as 𝜙, the bridge phase-shift angle,
decreases. Therefore, it is concluded that better operation is obtained when using PPM
modulation and with equal values of leg phase-shift angles of the primary and
secondary bridge (𝛼1 = 𝛼2).
The simulation results also show that state-variable and cyclic-averaging models
could accurately predict the output current behaviour for both SPS and PPM
modulation cases. As also observed for the frequency-modulated variant in the previous
chapter, the cyclic-averaging method had the fastest execution time.
5.8 References
[1] S. Inoue and H. Akagi, “A bidirectional DC-DC converter for an energy storage
system with galvanic isolation,” IEEE Transactions on Power Electronics, vol.
22, no. 6, pp. 2299–2306, 2007.
[4] H. Bai and C. Mi, “Eliminate reactive power and increase system efficiency of
isolated bidirectional dual-active-bridge dc-dc converters using novel dual-phase-
shift control,” IEEE Transactions on Power Electronics, vol. 23, no. 6, pp. 2905–
2914, 2008.
129
[7] A. A. S. Mohamed, A. Berzoy, F. G. N. Nogueira de Almeida, and O.
Mohammed, “Modeling and Assessment Analysis of Various Compensation
Topologies in Bidirectional IWPT System for EV Applications,” IEEE
Transactions on Industry Applications, vol. 53, no. 5, pp. 4973–4984, 2017.
130
6 CLLC Converter Design and Prototype
6.1 Introduction
In the previous chapter FMA, state-variable and cyclic-averaging models were
developed for a CLLC resonant converter operating under phase-shift modulation. It
was verified that state-variable and cyclic-averaging are very accurate methods to
model the converter’s behaviour when compared to a Spice simulation. It was also
previously observed that the use of the cyclic-averaging method results in the fastest
execution for both frequency and phase-shift modulated converters.
Following the verification through simulation, the construction process for a low
power prototype will be discussed in this chapter. The converter will be tested
considering bidirectional power flow and operation under Single Phase-Shift (SPS) and
Pulse-Phase (PPM) modulation. The experimental results obtained will be compared to
a Spice and cyclic-averaging simulation.
A simplified diagram for the prototype construction is shown in Figure 6.1. The
switching signals are produced using a microcontroller and inverting circuit. The H-
bridge 1 connects a power supply to one side of the resonant tank while H-bridge 2
connects the other side of the resonant tank to a resistor load.
131
6.2 Converter design
In this section the design procedure for each prototype subsystem from Figure 6.1
will be explained.
The block diagram from Figure 6.2 illustrates how the switching circuit is
implemented. The pulse signals are generated using the FRDM-KL25Z microcontroller
development board. Each channel (Ch0 to Ch3) generates the phase-shifted switching
signals to each leg of the two H-bridges. An inverting circuit is used to implement
phase-shift angles greater than 180°. In Figure 6.3 the inverting circuit is presented in
detail. The development board output has a voltage level of 3.3 V but the input signals
for deadtime circuit in the PCB must be at 15V level. Therefore, before the switching
signals are sent to the deadtime circuits, a voltage level shifter (CD4504B) is added.
The FRDM-KL25Z board uses an ultra low-power KL25 microcontroller from the
Kinetis L-series of ARM Cortex MCUs with Processor Expert support. Processor
Expert is a tool with graphical user interface that generates code from embedded
components as building blocks, allowing a fast and simplified development. To
132
generate the phase-shifted pulse waveforms the timer units are used in Output Compare
mode, where the output is toggled when the counter matches a set value. To obtain a
100 kHz switching signal the timer frequency is set to 200 kHz. The timer counter goes
from 0 to 104 and resets to zero, therefore, a full count is completed in 5 µs for a
frequency of 200 kHz.
For the four channels used (Ch0 to Ch3) the output value is zero at start. The
counter starts at 0 and when it reaches the value of variable ChN_count the channel
output value is inverted.
𝑃𝑆. 104
𝐶ℎ𝑁_𝑐𝑜𝑢𝑛𝑡 =
180
(6.1)
For phase-shift angles greater than 180 degrees, the inverting circuit is activated,
variable Inv_flag, from Figure 6.3, is changed from low to high and:
(𝑃𝑆 − 180).104
𝐶ℎ𝑁_𝑐𝑜𝑢𝑛𝑡 =
180
(6.2)
In Table 6.1 the values of Inv_flag, ChN_count and the resulting phase-shift angle
relative to Ch0 are presented for different modulation cases. Note that variable
ChN_count only accepts integer values, therefore an exact angle of 40° (𝐶ℎ2_𝑐𝑜𝑢𝑛𝑡 =
23.11) cannot be implemented. Choosing the closest integer value (23) the phase-shift
angle effectively implemented is 39.81°.
Table 6.1: Values of 𝑪𝒉𝑵_𝒄𝒐𝒖𝒏𝒕 and 𝑰𝒏𝒗_𝒇𝒍𝒂𝒈 for different modulation cases,
133
The pulse waveform diagrams for phase-shift angles of 90, 180 and 270 degrees
are shown in Figures 6.4, 6.5 and 6.6 respectively. This implementation results in a
pulse waveform with period of 10 µs and consequently, a frequency of 100 kHz.
134
Following the switching circuit implementation, the output signals will be
connected to a voltage level shifter, producing 15V switching signals to the input of the
deadtime circuit.
As shown in Figure 6.7 and from the CLLC topology description in Chapters 4 and
5, the CLLC converter is composed by two H-bridges connected to a resonant tank.
Each leg of the bridge consists of two complementary switches. To avoid the activation
of both switches in the same leg at the same time during transition, and consequently a
short circuit, a deadtime circuit is used to introduce a delay.
The deadtime circuit is shown in Figure 6.8, one circuit is used for each leg of the
two bridges. From experimental tests, a deadtime of approximately 100 ns is sufficient
to avoid short circuit. To obtain that delay, resistors 𝑅1 = 𝑅2 = 3.3 kΩ and capacitors
𝐶1 = 𝐶2 = 2.2pF were chosen for this application.
135
Figure 6.8: Deadtime circuit
The input signals for the deadtime circuit are provided by the switching circuit
(microcontroller board + inverting circuit + voltage level shifter) described in the
previous section. The output signals from the deadtime circuits are low power,
therefore, a gate driver circuit must be used before connection to the MOSFET
switches.
The IR2011 high power MOSFET driver is used to control the MOSFET switches.
The output of the deadtime circuit is used as input to the gate driver circuit, as shown in
Figure 6.9.
For the implementation of the H-bridges, IRFP250 Power MOSFETs are chosen
with 200V maximum drain source voltage, 30A maximum continuous drain current and
𝑅𝐷𝑆(𝑜𝑛) = 0.085Ω. A high voltage/current rating was chosen with the intention of
perform future tests at higher power levels after the validation of the 100W case. To
reduce conduction losses, a device with low value of 𝑅𝐷𝑆(𝑜𝑛) must be chosen.
136
Therefore, to further improve efficiency, devices using wide bandgap materials as
Silicon Carbide (SiC) and Gallium Nitride (GaN) could be used due to their reduced
on-resistance [1], [2].
The resonant tank circuit for a CLLC converter is shown in Figure 6.10.
Considering the operating frequency of 100 kHz, ETD cores with ferrite material 3C90
are used for the construction of inductors and transformer. This material is
recommended for applications with a frequency limit of 200 kHz.
The voltage and current values used in the design equations in this section are
chosen based on the simulation results of the ideal converter presented in the previous
chapter and considering a converter operating under maximum modulation.
For the inductors construction, the minimum number of turns is calculated based
on equation (6.3).
𝐿𝐼
𝑁𝑚𝑖𝑛 =
𝐵𝑚𝑎𝑥 𝐴𝑐𝑜𝑟𝑒
(6.3)
where L is the inductance, 𝐵𝑚𝑎𝑥 is the saturation flux density, 𝐴𝑐𝑜𝑟𝑒 is the effective
area of the core and 𝐼 is the inductors’ peak current.
𝑁 2 𝐴𝑐𝑜𝑟𝑒 𝜇0
𝑙𝑔 =
𝐿
(6.4)
The equivalent circuit for a real inductor is shown in Figure 6.11. Once the
inductors are built, the inductance and parasitic components values are measured using
the Bode 100 Vector Network Analyzer. The measurements were performed based on
the Bode 100 application note for power inductor modelling [3]. The parameters values
measured for inductor 𝐿𝑠1 are shown in Table 6.2 while the values for inductor 𝐿𝑚 are
shown in Table 6.3.
Parameter Value
𝑳 51.620 µH
𝑹𝒔 0.25 Ω
𝑹𝒑 1.9 kΩ
𝑪𝒑 2.52 pF
138
Table 6.3: 𝑳𝒎 (design value: 27.02 µH)
Parameter Value
𝑳 32.089 µH
𝑹𝒔 0.073 Ω
𝑹𝒑 1.96 kΩ
𝑪𝒑 2.394 pF
The inductor’s series resistance represents the winding losses. From the
measurements performed with the Bode 100, shown in Figure 6.12, it is possible to
observe that this resistance value is frequency-dependent, significantly increasing as the
frequency increases. The series resistances were measured at the operating frequency,
100 kHz.
For the resonant tank capacitors, the values of capacitance were calculated in
Chapter 5 considering the ideal values of inductance with the resonant circuit tuned to
100kHz. Based on the measured values of inductance, the values of capacitors 𝐶𝑠1 and
𝐶𝑠2 are recalculated to keep the resonant frequency at 100kHz. The value of 𝐶𝑠1 drops
from originally 31.24 nF to 29.70 nF, and 𝐶𝑠2 changes from 1.5 µF to 1.25 µF.
Polypropylene film capacitors were used due to the low-ESR, high power and
frequency characteristics.
139
Here the real capacitor is modelled considering only the capacitance and an
equivalent series resistance (ESR), as shown in Figure 6.13. The values of capacitance
and series resistance are measured with the Bode 100 vector analyser, based on the
Capacitor ESR measurement application note [4]. The results obtained are presented in
Table 6.4 for the primary capacitor and in Table 6.5 for the secondary capacitor.
Parameter Value
𝑬𝑺𝑹 0.044Ω
𝑪𝒔 28.354 nF
Parameter Value
𝑬𝑺𝑹 0.004 Ω
𝑪𝒔 1.326 µF
For the transformer design, the minimum number of turns in the primary coil is
calculated from equation (6.5).
𝑉. 𝛿
𝑁1 >
𝑓. 2. 𝐵𝑚𝑎𝑥 . 𝐴𝑐𝑜𝑟𝑒
(6.5)
where V is the primary RMS voltage, f is the switching frequency and 𝛿 is the
maximum duty cycle.
The transformer is built using an ETD-44 core and 20 turns on the primary winding
and 5 turns on secondary, resulting in a turns ratio of 4. A short circuit test is performed
with the Bode 100 to measure the leakage inductance and total winding resistance
reflected to the primary at 100kHz. The open circuit test is employed to measure the
140
primary inductance, 𝐿𝑝 . A gain measurement is used for calculation of the turns ratio
and based on this value the secondary inductance, 𝐿𝑠 , can be estimated for the Spice
𝐿𝑝
simulation as 𝐿𝑠 = . The transformer equivalent circuit is shown in Figure 6.14 and
𝑛2
Parameter Value
Gain 0.2501 (𝑛 = 3.998)
𝑹𝒕 0.92 Ω
𝑳𝒑 1.374 mH
𝑳𝒍𝒆𝒂𝒌 13.28 µH
6.2.4 Load
The main objective in this chapter is to verify the accuracy of the models
developed in Chapter 5 using an experimental set-up. Therefore, to simplify the model-
prototype comparison, a load resistor, which is simpler to model, is used in place of the
output battery. In Chapter 5 the battery is simply modelled as a voltage source. Using a
resistor as load, the simulation and prototype results can be accurately compared
without the need of developing a more accurate battery model for the simulations. The
equivalent resistor is defined based on the output voltage and current for the maximum
modulation case (𝛼𝑟𝑎𝑡𝑖𝑜 = 1 and 𝜙 = 90°). From Spice simulation results obtained in
Chapter 5, the output resistors are calculated for forward (𝑅𝑓𝑜𝑟 ) and reverse (𝑅𝑟𝑒𝑣 )
operation:
141
𝑉𝑜𝑢𝑡 12.093
𝑅𝑓𝑜𝑟 = = = 1.298Ω
𝐼𝑜𝑢𝑡 9.3162
𝑉𝑜𝑢𝑡 48.023
𝑅𝑟𝑒𝑣 = = = 20.805Ω
𝐼𝑜𝑢𝑡 2.3082
(6.6)
The HS100 series are aluminium housed resistors designed to be used with
heatsink for maximum performance, hence, as shown in Figure 6.15, the resistors are
directly mounted in a heatsink with thermal compound. Wirewound resistors have a
small inductance associated, however, it was observed in simulation that the inclusion
of this value in the model did not significantly affect the average output current value.
142
models are modified to include the output resistor load and the real values of the
resonant circuit components (capacitors, inductors and transformer) measured in the
previous section.
(6.7)
143
where:
𝑣1 (𝑡) = ±𝑉𝑑𝑐 or 0 and 𝑣2 (𝑡) = ±𝑣𝐶𝑓 (𝑡) or 0
(6.8)
(6.10)
144
Once the experimental set-up is built, as shown in Figure 6.16, the system is tested
using a 0-60V, 100A power supply as input.
As shown in the previous section, the real inductor equivalent circuit comprises a
series resistance, parallel resistance and capacitance. For the cyclic-averaging model
developed in the previous chapter, the only loss element considered in the resonant
circuit was the series resistance and the transformer was assumed to be ideal. To avoid
further modifications in the model equations with the addition of these parallel and
leakage elements, the cyclic averaging simulations are initially performed not
considering the parallel capacitances and resistances, and the transformer’s leakage
inductance.
Since the modifications on the Spice model are easily implemented, simply adding
the new components blocks, the Spice simulations are performed considering all
measured parasitic elements. The resonant tank circuit diagrams used for Spice and
cyclic-averaging simulations at this stage are presented in Figure 6.17.
145
(a)
(b)
Figure 6.17: Resonant tank circuits employed for (a) Spice simulation and (b)
cyclic-averaging simulation
Simulation and experimental results are compared in Figures 6.18 and 6.19, where
the output DC current is plotted as function of the alpha ratio for PPM modulation and
in function of the bridge phase-shift angle for SPS modulation.
146
(a) (b)
Figure 6.18: Spice, cyclic-averaging and prototype results comparison for (a)
PPM modulation and (b) SPS modulation in forward operation
(a) (b)
Figure 6.19: Spice, cyclic-averaging and prototype results comparison for (a)
PPM modulation and (b) SPS modulation in reverse operation
For a converter operating under SPS modulation, as the bridge phase-shift angle
decreases, the error between prototype and Spice results significantly increase. During
experimental validation it was observed that implementation of small values of phase-
shift angle 𝜙 are subject to reduced precision from the microcontroller and higher
influence of deadtime and delays, contributing to increased errors between simulation
and prototype results. Also, the operation of a converter under SPS modulation
degrades as the phase-shift angle reduces and as the voltage mismatch between primary
and secondary increases resulting in high circulation current, hard switching operation
and large current spikes. For a converter operating under PPM modulation results were
147
not significantly affected by the reduction of leg phase-shift angles, even when testing
small values of 𝛼𝑟𝑎𝑡𝑖𝑜 .
The simulations were performed considering the bridge phase-shift angle value set
by the microcontroller, thus to improve the accuracy of the simulations, the difference
between the phase-shift angle of the switching signals and the real phase-shift at the
output of the bridges must be incorporated into the models. The drain-source voltages
𝑣𝑑𝑠13 and 𝑣𝑑𝑠23 from Figure 6.20 are analysed with the oscilloscope, as shown in
Figure 6.21. The phase-shift is measured for different SPS modulation cases and an
average difference of 90.28ns is obtained between the set and real time difference,
which corresponds to a 3.24° phase difference. Therefore, Spice and cyclic-averaging
simulations are performed again adding 3.24° to the phase-shift angle 𝜙 previously
used.
Figure 6.21: Drain source voltages for 10° phase-shift test case in forward
mode
148
It was also observed in the first set of results in Figures 6.18 and 6.19 that the
cyclic-averaging model was not as accurate as Spice, especially for high modulation
angles. To further improve the accuracy of the cyclic-averaging method without
changing the equation description, the non-ideal inductor circuit can be reduced to an
equivalent inductive reactance in series with a resistor, as in Figure 6.22, and the
leakage inductance of the transformer is incorporated to the reactance value of the
secondary capacitor, Figure 6.23. This way, all parasitic elements considered in the
Spice simulation are now also incorporated to the cyclic method, improving the
comparison between the models.
The equivalent impedances for the inductor, 𝑍𝑙𝑒𝑞 , and for the secondary capacitor,
𝑍𝑐𝑒𝑞 , are defined in equation (6.11).
1
𝑍𝑙𝑒𝑞 = 𝑅𝑙𝑒𝑞 + 𝑗𝜔𝐿𝑒𝑞 =
1 1
+ + 𝑗𝜔𝐶𝑝
𝑅𝑝 𝑅𝑠 + 𝑗𝜔𝐿
1 𝑗
𝑍𝑐𝑒𝑞 = 𝑅𝑐𝑒𝑞 − 𝑗 = 𝑅𝑠 + 𝑗𝜔𝐿𝑙𝑒𝑎𝑘 −
𝜔𝐶𝑒𝑞 𝜔𝐶𝑠2 ′
(6.11)
where, in the equivalent circuit for the secondary capacitor, 𝑅𝑠 is obtained adding
the capacitor’s ESR to the transformer resistance, all referred to primary.
149
The calculated values for the equivalent circuit parameters, referred to the primary
side of transformer, are shown in Table 6.7.
(a) (b)
Figure 6.24: Optimized Spice, cyclic-averaging and prototype results comparison
for (a) PPM modulation and (b) SPS modulation in forward operation
150
(a) (b)
Figure 6.25: Optimized Spice, cyclic-averaging and prototype results comparison
for (a) PPM modulation and (b) SPS modulation in reverse operation
The results for the SPS modulation case improved compared to the previous set of
results, especially for small phase-shift angles. The simulation results are now closer to
experimental due to the phase-shift angle compensation implemented. No significant
difference is observed for the PPM modulation case from the addition of the
compensation angle. Even with the basic delay compensation implemented on
simulation models, errors are still higher for SPS modulation when operating with small
values of phase-shift angle 𝜙.
151
(a) (b)
Figure 6.26: Comparison between (a) Spice and (b) experimental results for
the primary current
(a) (b)
Figure 6.27: Comparison between (a) Spice and (b) experimental results for
the secondary current
(a) (b)
Figure 6.28: Comparison between (a) Spice and (b) experimental results for
the primary capacitor voltage
152
(a) (b)
Figure 6.29: Comparison between (a) Spice and (b) experimental results for
the secondary capacitor voltage
Overall, the experimental waveforms are also close to the simulation results. The
values of rms voltages and currents obtained with the Spice and experimental
measurements are compared in Table 6.8. From the waveforms results, it is observed
that the peak value of the secondary current is higher for the experimental case,
however, the experimental rms current value is lower. Furthermore, the experimental
waveform for the secondary capacitor voltage shows increased noise, while in
simulations this waveform is cleaner.
153
the primary or secondary side of the resonant tank. Results are presented in Table 6.9,
where the percentage error is calculated as follows:
(6.12)
The secondary is the low voltage/high current side of the circuit, therefore,
increased resistances on this side results in higher current drop. From the results it is
also verified that reverse mode results are more significantly affected by additional
resistances in the secondary side of the resonant tank, in this case the resistor is
introduced in the high current side which is also the input side.
From the results obtained it is noticed that results are slightly affected by variations
in the primary capacitor, inductors and frequency while variations on the secondary
capacitor do not significantly affect the output current, with a difference in current
inferior to 1% for a change of ±5% in the capacitor value.
154
Table 6.10: Influence of capacitor 𝑪𝒔𝟏 on output current
155
Table 6.14: Influence of frequency on output current
Forward mode
Parameter 𝑰𝒐𝒖𝒕 (A) Error (%)
Maximum (1.9 Ω) 6.7244 -0.88
𝑹𝒇𝒐𝒓 Average (1.7 Ω) 6.7844 -
Minimum (1.5 Ω) 6.8454 0.90
Reverse mode
Parameter 𝑰𝒐𝒖𝒕 (A) Error (%)
Maximum (22.65 Ω) 1.5443 -0.22
𝑹𝒓𝒆𝒗 Average (22.30 Ω) 1.5477 -
Minimum (22.10 Ω) 1.5497 0.13
156
MOSFET including parasitic elements, instead of only 𝑅𝐷𝑆(𝑜𝑛) and associated body
diode, but this would also increase the complexity of the system, especially for cyclic-
averaging implementation.
6.4 Conclusions
In this chapter a prototype for the CLLC converter was built to verify the accuracy
of the Spice and cyclic-averaging models. The results show that both models analysed
could successfully predict the converter’s behaviour, with the confirmation that the
phase-shift angles of the two H-bridges can be used to control the output current.
Therefore, it is possible to use this converter as a charging/discharging system when
connecting a battery to the output.
The models adopted in this chapter for representation of real inductors, capacitors,
transformer, the associated resistances and parasitic elements are approximations and
cannot perfectly represent the real system behaviour. Also, measurement errors,
parameters variation and environment factors, as temperature, affect the results.
Therefore, discrepancies between the models and experimental results still occur.
Errors between simulation and practical results are lower for the PPM modulation
case, due to the increased sensitivity to delays and performance deterioration as the
value of bridge phase-shift angle decreases when operating under SPS modulation, and
finally, the difficulty to accurately incorporate the system’s delays and all parasitic
elements to the simulation models.
The influence of parasitic elements and transformer leakage inductance affects the
resonant state, where a high value of transformer leakage inductance could result in a
deviation from the point of resonance. Therefore, to improve performance, in [5] a
construction process is suggested where the transformer is designed to incorporate the
values of the resonant inductors. Consequently, to reduce the influence of the non-ideal
157
transformer to this system, the transformer could be designed to have leakage
inductance equal to 𝐿𝑠1 and magnetizing inductance equal to 𝐿𝑚 . In the next chapter, a
wireless variant of this converter will be analysed, where the resonant tank inductors
originate from the wireless transformer equivalent circuit and no extra inductor is
needed.
6.5 References
[1] F. Xue, R. Yu, and A. Q. Huang, “A 98.3% Efficient GaN Isolated Bidirectional
DC-DC Converter for DC Microgrid Energy Storage System Applications,” IEEE
Transactions on Industrial Electronics, vol. 64, no. 11, pp. 9094–9103, 2017.
[3] M. Bitschnau, “Power Inductor Modelling,” Bode 100 Application Note, 2017.
158
7 Analysis of Series Compensated IPT
converter
7.1 Introduction
The cyclic-averaging analysis approach was successfully applied in the previous
chapters to model the bidirectional CLLC resonant converter. Here, this analysis is
extended to evaluate an Inductive Power Transfer (IPT) wireless system. In the
literature review conducted in Chapter 2, three topologies of compensation circuits
were compared. The series compensated topology, shown in Figure 7.1, is chosen due
to its simplicity, good performance and similarity to the previously analysed CLLC
converter. FMA, state-variable and cyclic-averaging methods will now be applied to
this circuit and verified against a Spice simulation. To simplify the analysis and
construction processes, the IPT converter analysed in this chapter has a turns ratio, 𝑛,
equal to one, therefore 𝐿1 = 𝐿2 .
159
The conventional transformer utilized in the previous chapters is now substituted
by primary and secondary IPT coils, separated by a large air gap of 150mm. The values
of inductance and coupling coefficient depends on the coil pad dimensions, material
and the distance, or misalignment, between the pads. Using a 3D finite element analysis
(FEA) simulation tool, the Double D coil pad structures will be designed and the
influence of construction parameters and misalignment will be evaluated. The coil
structures are built, tested for misalignment conditions, and results are compared to the
FEA simulation.
Once the coil structures are validated, the pads are connected to the compensation
capacitors and the prototype of the IPT converter is tested for experimental verification
of the cyclic-averaging and Spice simulations.
(a) (b)
Figure 7.2: Equivalent circuits for series compensated IPT converter with (a)
coupled inductors representation and (b) T-model representation
𝑀 = 𝑘√𝐿1 𝐿2
160
(7.1)
The resonant frequency is defined in (7.2) based solely on the values of self-
inductance and compensation capacitors, therefore, this frequency is independent from
the coupling coefficient.
1 1
𝜔0 = 2𝜋𝑓0 = =
√𝐿1 𝐶𝑠1 √𝐿2 𝐶𝑠2
(7.2)
From analysis of the equivalent circuit, Figure 7.2-b, the circuit impedances in the
T-network are defined as follows:
1
𝑍1 = 𝑗𝜔(𝐿1 − 𝑀) − 𝑗
𝜔𝐶𝑠1
1
𝑍2 = 𝑗𝜔(𝐿2 − 𝑀) − 𝑗
𝜔𝐶𝑠2
𝑍𝑚 = 𝑗𝜔𝑀
(7.3)
Since the converter operates under phase-shift modulation (SPS or PPM), the
bridge voltages waveforms and the phasor voltages are defined the same way as in
Chapter 5 (equations (5.1) and (5.2)), applying the fundamental approximation to the
Fourier representation of functions 𝑣1 (𝑡) and 𝑣2 (𝑡). Therefore, the voltage phasors are
given by:
4𝑉𝐷𝐶 𝛼1
𝑽𝟏 = sin ( )
𝜋√2 2
4𝑉𝑏𝑎𝑡 𝛼2
𝑽𝟐 = sin ( ) (cos(𝜙) + 𝑗 sin(𝜙))
𝜋√2 2
(7.4)
Once the voltage phasors are defined, the following equations are obtained
analysing the circuit in Figure 7.2-b.
161
𝑽𝟏 = 𝑍1 𝑰𝟏 + 𝑍𝑚 𝑰𝒎
𝑰𝟏 = 𝑰 𝟐 + 𝑰𝒎
𝑍𝑚 𝑰𝒎 = 𝑍2 𝑰𝟐 + 𝑽𝟐
(7.5)
After solving the system of equations, the currents are obtained, as in (7.6).
𝑽𝟏 (𝑍2 + 𝑍𝑚 ) − 𝑽𝟐 𝑍𝑚
𝑰𝟏 =
𝑍1 𝑍2 + 𝑍1 𝑍𝑚 + 𝑍2 𝑍𝑚
𝑽𝟏 𝑍𝑚 − 𝑽𝟐 (𝑍1 + 𝑍𝑚 )
𝑰𝟐 =
𝑍1 𝑍2 + 𝑍1 𝑍𝑚 + 𝑍2 𝑍𝑚
𝑽𝟏 𝑍2 + 𝑽𝟐 𝑍1
𝑰𝒎 =
𝑍1 𝑍2 + 𝑍1 𝑍𝑚 + 𝑍2 𝑍𝑚
(7.6)
At simulation stage, MATLAB “abs()” function will be used with (7.6) to calculate
the rms value of the primary and secondary currents.
No resistive losses are considered in the FMA model, therefore, the magnitude of
the output active power is obtained calculating 𝑅𝑒(𝑽𝟏 𝑰∗𝟏 ) or 𝑅𝑒(𝑽𝟐 𝑰∗𝟐 ), resulting in the
following equation:
𝛼 𝛼
8𝑉𝑑𝑐 𝑉𝑏𝑎𝑡 𝑋𝑚 sin(𝜙) sin ( 1 ) sin ( 22 )
𝑃𝑜𝑢𝑡 = 2
𝜋 2 (𝑋1 𝑋2 + 𝑋1 𝑋𝑚 + 𝑋2 𝑋𝑚 )
(7.7)
The power equation can be simplified by substituting equation (7.2), which defines
the resonant condition, into (7.7). The final expression for active power calculation is
given by (7.8).
𝛼1 𝛼2 𝛼1 𝛼2
8𝑉𝑑𝑐 𝑉𝑏𝑎𝑡 sin(𝜙) sin ( ) sin ( ) 8𝑉𝑑𝑐 𝑉𝑏𝑎𝑡 sin(𝜙) sin ( ) sin ( )
𝑃𝑜𝑢𝑡 = 2 2 = 2 2
𝜋 2 𝜔𝑀 𝜋 2 𝜔𝑘√𝐿1 𝐿2
(7.8)
162
The converter operates at resonant frequency 𝜔 = 𝜔0 and maximum output power
is achieved when 𝜙 = ±90° and 𝛼1 = 𝛼2 = 180°. The output power equation is often
used in the design process to dimension the inductors based on a defined range of
input/output voltages, coupling coefficient and operating frequency.
(𝐿𝑎 + 𝑀)(𝐿𝑏 + 𝑀) − 𝑀2
𝐿𝑠 =
(𝐿𝑏 + 𝑀)
𝑀2
𝐿𝑚 =
(𝐿𝑏 + 𝑀)
𝑀
𝑛=
(𝐿𝑏 + 𝑀)
(7.9)
Once the equivalent CLLC converter is defined, the state-variable and cyclic-
averaging models obtained in Chapter 5 for phase-shift modulated converters are
163
applied to describe the behaviour of the IPT converter. At simulation stage, the models
will be used here to calculate the converter’s output current, however, both state-
variable and cyclic-averaging simulations can also be employed to analyse the
behaviour of the primary and secondary currents and voltages across the primary and
secondary capacitors, since the primary and secondary currents (𝐼1 and 𝐼2 ) in the IPT
resonant tank and CLLC resonant tank are equal.
The simulation results for the FMA, state-variable and cyclic-averaging models
are discussed in the next section.
As discussed in the previous chapters, the FMA method is often used in the design
process due to its simple and fast analysis. The design process here performed consists
of the following steps:
164
5. The equation for resonance condition (7.2) is used to calculate the values of the
compensation capacitors 𝐶1 and 𝐶2 ;
Here, the converter is designed considering the same operating frequency and
voltage specification adopted for the verification of the CLLC converter in Chapters 5
and 6. Based on literature analysis and the previous work performed on loosely coupled
transformers in [2], the design value of 0.3 was chosen for the coupling coefficient. The
parameters used for the design process, as the inductors and capacitors calculated in
steps 4 and 5, are listed in Table 7.1.
Parameter Value
𝑷𝒐𝒖𝒕 110 W
𝑽𝒅𝒄 48 V
𝑽𝒃𝒂𝒕 12 V
𝒇 100 kHz
𝒌 0.3
𝑳𝟏 , 𝑳𝟐 21 𝜇H
𝑪𝒔𝟏 , 𝑪𝒔𝟐 120 nF
For the FMA simulations all resistances are neglected. The Spice, state-variable
and cyclic-averaging simulations are conducted under nearly ideal conditions and the
only resistances considered are the input and output resistances associated with the
voltage sources (𝑟𝑏𝑎𝑡 = 𝑟𝑑𝑐 = 0.01Ω) and series resistances of primary and secondary
inductors (𝑟1 = 𝑟2 = 0.03Ω).
In the previous section the converter was designed to operate with coupling
coefficient of 0.3 and primary and secondary self-inductances (𝐿 = 𝐿1 = 𝐿2 ) equal to
21𝜇𝐻 in order to obtain output power close to 110W. Now the FMA model is used to
estimate how the output power, primary and secondary currents behave with variations
in the coupling coefficient and self-inductance. The results obtained will also be
compared to a Spice simulation to verify the accuracy of the FMA model.
165
Figure 7.4 shows the primary current results obtained with the Spice and FMA
models considering forward operation and maximum modulation. In Figure 7.4-a, the
self-inductance of primary and secondary coils is kept constant at 21𝜇𝐻 while the
coupling coefficient varies in a range from 0.05 to 0.6, in steps of 0.05. In Figure 7.4-b,
to analyse the effect of variations in the inductance value, the results were obtained
maintaining the coupling coefficient constant at 0.3 and inductance values from 10𝜇𝐻
to 50𝜇𝐻 were tested in steps of 5𝜇𝐻. For the results shown in Figure 7.4-b, the
capacitances are changed according to the values of inductances tested to maintain the
resonant frequency constant.
(a) (b)
Figure 7.4: Spice and FMA results for RMS value of primary current in
function of (a) coupling coefficient and (b) self-inductance
From the results analysis it is observed that, as the values of coupling coefficient
and self-inductance increase, the magnitude of the primary current decreases,
confirming the previous analysis of series compensated IPT converters performed in
[3], [4]. Additionally, the error between Spice and FMA results is higher for lower
values of coupling coefficient and inductance. However, when the coupling coefficient
and self-inductance start to increase, the difference between the models reduces and
results become very close. As stated in the previous section, the Spice model is
simulated considering low values of circuit resistances, while in FMA these resistances
are completely neglected. Therefore, the errors may be due to the fundamental
approximation in the FMA model and to the increased resistances considered in the
Spice model.
166
To identify the main source of error in the lower range of coupling coefficients and
self-inductances, the Spice simulation is performed again reducing the values of
resistances to get closer to an ideal circuit, 𝑟𝑏𝑎𝑡 and 𝑟𝑑𝑐 are reduced from 0.01 to 0.001
and 𝑟1 and 𝑟2 are reduced from 0.03 to 0.0003. Results are obtained for the first point of
each graph (𝑘 = 0.05 in Figure 7.4-a and 𝐿 = 10𝜇𝐻 in Figure 7.4-b), where the error
is the highest. Additionally, the fundamental component of the primary current is also
measured in the Spice simulation to verify the contribution of the remaining harmonics.
The results obtained are presented in Table 7.2.
Table 7.2: Comparison between Spice and FMA models for primary current (𝑰𝟏 )
After analysing the data in Table 7.2 it is possible to conclude that when using low
resistances the RMS value of the primary current is very close to the magnitude of the
fundamental component. Once the circuit resistances were reduced to nearly zero, the
difference between the magnitude of the primary current and its fundamental
component significantly increases, indicating that other harmonics have higher
contribution to the current waveform. Note that the value of the fundamental
component in Spice when the resistances are closer to zero is now very close to the
FMA result. Therefore, it is confirmed that the errors observed in Figure 7.4 are due to
not only the increased resistances in the Spice model, but also to the lack of accuracy of
the fundamental approximation to represent the primary current when the circuit
resistances are very close to zero. This behaviour, however, was only significantly
observed in the lower range of coupling coefficients and self-inductances analysed.
167
Following the analysis of the primary current, the Spice and FMA results for the
secondary current are shown in Figure 7.5. In this case, FMA model shows good
accuracy for the full range of coupling coefficients and inductances analysed.
(a) (b)
Figure 7.5: Spice and FMA results for secondary current in function of (a)
coupling coefficient and (b) self-inductance
Table 7.3: Comparison between Spice and FMA models for current 𝑰𝟐
168
The results for output power are shown in Figure 7.6. Contrary to the parallel or
LCL compensation topologies, for the series compensated converter ideal behaviour, an
increase in the output power is observed as the coupling coefficient decreases,
confirming the previous analysis found in the literature [3], [5]. For a practical
converter, however, experimental results show that the power increases with the
reduction of coupling coefficient until a certain point, but as the coupling furtherly
decreases the output power starts to significantly drop, as discussed in [5].
(a) (b)
Figure 7.6: Spice and FMA results for output power in function of (a)
coupling coefficient and (b) self-inductance
169
(a)
(b)
Figure 7.7: Equivalent circuits for Spice simulation (a) series compensated
IPT converter and (b) CLLC equivalent circuit
The Spice simulation of the IPT converter is performed considering the circuit
parameters from Table 7.1, obtained at the end of the design stage described in section
7.4.1. Substituting the IPT converter parameters in equation (7.9), the parameters for
the CLLC equivalent circuit are calculated and listed in Table 7.4. The simulations are
performed considering nearly ideal conditions, using ideal switches and low values of
resonant tank resistances.
Parameter Value
𝑪𝒔𝟏 , 𝑪𝒔𝟐 120 nF
𝑳𝒔 19.11 𝜇𝐻
𝑳𝒎 1.89 𝜇𝐻
𝒏 0.3
𝒓𝟏 , 𝒓𝟐 0.03 Ω
𝒓𝒃𝒂𝒕 , 𝒓𝒅𝒄 0.01 Ω
170
The results obtained for SPS and PPM modulation are shown in Figures 7.8 and
7.9, respectively.
(a) (b)
Figure 7.8: Model comparison SPS modulation (a) forward operation and (b)
reverse operation
(a) (b)
Figure 7.9: Model comparison PPM modulation (a) forward operation and
(b) reverse operation
From results analysis it is possible to observe that the transformation of the IPT
converter into a CLLC equivalent converter was accurate, where, between the two
Spice simulations performed, the maximum difference in the output current was in the
order of 30 mA. Furthermore, both state-variable and cyclic-averaging simulations
could accurately model the behaviour of the converter’s output current. The maximum
error between the cyclic-averaging and the Spice simulation of the IPT converter is
2.23%, between state-variable and Spice, the maximum error is slightly lower, at
1.61%.
171
7.5 Magnetics design of coil pad structures
After the simulation analysis, a prototype is built for further verification of the
cyclic-averaging and Spice models. The first step in the prototype construction process
is the magnetic design of the primary and secondary coil pad structures. This design is
performed using Ansys Maxwell, a simulation software that uses finite element analysis
(FEA) to solve the electromagnetic field.
Based on the literature review conducted in Chapter 2, the Double D coil topology
was chosen due to its simplicity and reduced sensitivity to misalignment compared to
circular coils. The 3D model of the structures simulated in Ansys Maxwell is shown in
Figure 7.10. The pad structure is shown in Figure 7.10-a and it is composed by two
coils connected in parallel and mounted in an aluminium base. Rows of ferrite I cores
(Figure 7.10-a, in black) with dimensions of 93x28x16 mm are used below the coil to
increase the transferred power and direct the magnetic flux [6], [7].
Each coil has 6 turns and uses a 2.6 mm diameter litz wire (Figure 7.10-a, in
green), therefore, in the current excitation settings the option of stranded wire was
chosen for the conductor type. The aluminium plate (Figure 7.10-a, in grey) is used as
an electromagnetic shield and it also serves as support for the coil and ferrite bars. The
aluminium shield is effective reducing leakage magnetic fields, however, the electrical
performance will be affected since circulating eddy currents will be induced in the
aluminium plate. The negative effects of the metallic shielding are minimized when
using a ferrite layer between the coil and shield, with a metal shield thickness superior
to the skin depth [8].
Following the SAE standard J2954 [9], the primary and secondary pads are
designed to operate separated by a distance of 150 mm (Figure 7.10-b), this
corresponds to a Z2 class, where the vertical distance between ground surface and the
furthest coil is between 140 and 210 mm. The axis orientation for the 3D model
implemented is shown in Figure 7.10-b.
In Fig 7-c the region of simulation surrounding the coil pad structures is shown.
The material of the region is set to air and the size of the region must be large enough to
not impact the calculation of the electromagnetic fields near the structures
implemented, here, accurate results were obtained using a region with volume around
seven times the volume of the coil pad structures separated by 150mm air gap.
172
(a)
(b)
(c)
Figure 7.10: 3D model of double D coil (a) top view of coil pad (b) primary and
secondary coil pads (c) coil pads + region
173
Once the 3D finite element model is implemented in Ansys Maxwell, the FEA
analysis is employed to calculate the coupling coefficient and self-inductances using the
eddy current solver, which gives a steady state solution for AC magnetic fields at the
operating frequency of 100kHz. A sinusoidal current with peak value of 7A and
100kHz is assigned as excitation for each coil. Contrary to the magnetostatic solver, the
operation frequency and skin effect are taken into consideration when using the eddy
current solver. The solver also uses adaptive mesh refinement, where the mesh is
automatically refined at each pass. Over-defining mesh can lead to long execution times
without significant improvement of accuracy, therefore, accurate solutions can be
obtained with adaptive mesh with faster execution. Here, two types of tests were
performed, first using adaptive mesh and, secondly, manually defining the mesh. The
solution using adaptive mesh showed good accuracy and improved execution time,
therefore, this technique was chosen for the simulations presented in this section. The
final mesh plot in the ZX plane obtained after the adaptive mesh refinement is shown in
Figure 7.11. It is possible to observe how the mesh size is optimized along the region,
where finer mesh is attributed to smaller elements and in the areas closer to the pad
structures.
Figure 7.11: Mesh plot for coil pad structures, cross section area perpendicular to
Y axis
Simulation is conducted using the initial pad dimensions listed in Table 7.5. The
influence of each construction parameter on the inductance and coupling coefficient
values will be evaluated for an optimization process to achieve the values of inductance
and coupling coefficient defined in the converter specification, 21 𝜇H and 0.3
respectively.
174
Table 7.5: Initial parameters for Double D coil in the FEA simulations
Parameter Value
Coil width (𝑾𝒄𝒐𝒊𝒍 ) 300 mm
Coil length (𝑳𝒄𝒐𝒊𝒍 ) 300 mm
Number of ferrite cores per row 6
Distance between rows of ferrite cores 20 mm
Pitch 10 mm
The first parameters analysed are coil length and width, as shown in Figures 7.12
and 7.13. For each parameter variation, the remaining variables are kept constant
following the values in Table 7.5. From the coil length and coil width results, it is
possible to observe that inductance increases almost linearly as these parameters
increase. The coupling coefficient is nearly constant for coil length higher than 375 mm
and for coil width in a range of 275 to 375 mm. To maximize the coupling coefficient
and increase the inductance, the values of 425 mm and 325 mm are chosen for coil
length and coil width respectively.
175
Figure 7.13: Coupling coefficient and self-inductance in function of coil width
The simulation is updated with the optimized coil dimensions and the influence of
the number of ferrite cores per row and the distance between these rows is evaluated in
Figures 7.14 and 7.15, respectively. The ferrite material is used in wireless power
transfer applications to increase the power transfer capability and, as observed in Figure
7.14, the number of ferrite cores per row considerably affects the coupling coefficient,
where the peak value is reached at 6 bars of ferrite cores per row. Therefore, this is the
optimal number of bars, resulting in a total ferrite length of 558 mm per row, nearly
86% of the double D coil total width. When the total number of bars is increased to 7,
the total ferrite length starts to surpass the double coil width and the coupling
coefficient slightly decreases. As observed in Figure 7.15, the space between the rows
of ferrite bars does not significantly influence the values of inductance and coupling
coefficient, however, the peak value for both coupling coefficient and inductance
occurs around 25mm, therefore, this value is preferred for the construction instead of
the initial estimation of 20 mm.
176
Figure 7.15: Coupling coefficient and self-inductance in function of space between
rows of ferrite bars
The last parameter analysed is the pitch. As shown in Figure 7.16, the coupling
coefficient is not significantly influenced by the pitch value, however, the inductance
notably increases as the pitch decreases. The pitch value of 8 mm is chosen to keep the
coupling coefficient close to the maximum value and achieve a value of inductance that
meets the specification, resulting in pad structures with self-inductance of 21.792 𝜇H
and coupling coefficient of 0.398. Therefore, after the optimization process, the
specification values of 21 𝜇H for inductance and coupling coefficient of 0.3 are met. A
comparison between the initial estimation and the optimized parameters is shown in
Table 7.6.
177
Table 7.6: Parameters for Double D coil in the FEA simulations
Another factor analysed in the magnetics design is the misalignment in the X and
Y axis and influence of distance between the pads in the Z axis. Offsets in the X, Y or Z
axis (Figure 7.10) do not have large influence in the inductance but the coupling
coefficient is substantially affected.
First, the results for an offset in the X axis (Figure 7.17) are presented in Figure
7.18. It is possible to observe that, for an offset higher than approximately 80 mm, the
coupling coefficient steeply drops. As previously discussed in Chapter 2 during the
literature review, the Double D coil has improved performance for misalignment in the
Y axis. To improve the behaviour of the coupling coefficient for misalignments in the
X axis, Double D Quadrature or Bipolar coils should be used.
178
Figure 7.18: Coupling coefficient in function of distance between pads in the X
axis
An offset in the Y axis (Figure 7.19) also influences the coupling coefficient, as
shown in Figure 7.20, but not as significantly as in the X axis. For an offset of 200 mm
in the Y axis, the coupling coefficient drops from 0.398 to 0.272. The same offset in the
X axis causes a drop in the coupling coefficient from 0.398 to 0.124.
179
The converter is designed to operate at a fixed distance of 150 mm between the
primary and secondary coil pads. The influence of a variation of this distance (Figure
7.21) is shown in Figure 7.22. As expected, the coupling coefficient is increased when
reducing the distance between the primary and secondary pads and, as this distance
increases, the coupling coefficient decreases.
In the construction process, due to market availability and price, ferrite plates with
dimensions 43x28x4.1 mm and 3C95 material were used instead of the I93x28x16
cores used in the previous simulations. Since the new ferrite plates have reduced length
compared to the I core, a higher number of cores per row (13 cores) is necessary to
180
obtain a length closest to the equivalent of using 6 I cores (558mm). Furthermore, due
to availability in the university lab, the litz wire used in the construction stage has
diameter of 4 mm (with strands of 0.4 mm). Consequently, the 3D FEA model was
readjusted for the new dimensions of the ferrite core and wire.
The updated simulation also includes additional construction parameters that were
not previously considered, as the thickness of a 1mm polycarbonate (PC) sheet that
serves as a holding structure for the coil and the thickness of tape used to attach the
ferrite cores and coil to the PC sheet.
The pitch value was readjusted to 4mm in order to maintain the inductance close to
21 𝜇H, resulting in an inductance of 21.49 𝜇H and coupling coefficient of 𝑘 =0.366 in
the final simulation. The updated construction parameters are shown in Table 7.7. In
case the values of inductance or coupling coefficient are too distant from the desired
values (inductance of 21 𝜇H and coupling coefficient of 0.3) during experimental
verification, an extra layer of ferrite plates could be added increasing the thickness of
the rows of ferrite bars or the pitch could be readjusted in order to get values closer to
the specification.
Parameter Value
Coil width (𝑾𝒄𝒐𝒊𝒍 ) 325 mm
Coil length (𝑳𝒄𝒐𝒊𝒍 ) 425 mm
Number of ferrite cores per row 13
Distance between rows of ferrite cores 25 mm
Pitch 4 mm
From the updated FEA simulation, the magnetic flux density distribution is
analysed along the XZ plane as shown in Figure 7.23. The results for the updated model
based on the construction parameters is shown in Figure 7.23-a, while in Figure 7.23-b
the aluminium shield is removed from the model to verify its influence. It is possible to
observe that the aluminium shield is effective supressing the leakage magnetic field,
since the flux density is significantly reduced in the areas below and above the inferior
and superior shields, respectively.
181
(a)
(b)
Figure 7.23: Flux density distribution in the cross-section area perpendicular to Y
axis (a) with aluminium shield (b) no aluminium shield placed
The current density distribution for the coils and aluminium shields in a cross-
section area in the XZ plane is shown in Figure 7.24, where it is verified the presence
induced currents in the aluminium shields (in green/yellow). Despite the eddy current
occurrence in the shield structures, the metal shield is a simple and popular solution that
still results in good performance when used in conjunction with a ferrite layer.
182
Figure 7.24: Current density distribution for primary and secondary coils and
aluminium shields
The pad structures built will be further analysed in the next section, where the
experimental results obtained for inductance and coupling coefficient will be compared
to the FEA simulation.
After the FEA simulations performed in section 7.5 for the magnetics design, the
construction and experimental verification of the coil pad structures are evaluated in
this section. The primary coil pad structure built is shown in Figure 7.25. The
secondary pad is built with the same dimensions as the primary.
183
Figure 7.25: Coil pad structure
To maintain a fixed distance of 4 mm between each coil turn, the structure shown
in Figure 7.26 was built using a 3D printer.
Before the coil pads are connected to the switching circuit, compensating
capacitors and load for experimental verification of the converter, the coupling
coefficient and primary and secondary inductances are measured to check if the coil
pads parameters are close to the design values.
First, the self-inductance is measured using the Bode 100. The distance between
the pads has a slight influence in the measured self-inductance as consequence of the
proximity to the metal and magnetic material of the opposite pad. When the inductance
is measured with primary and secondary pads considerably distant from each other, the
value of inductance for the primary pad is 17.748 µH and for the secondary is 18.251
184
µH. Placing the primary and secondary pads at the position the system is designed to
operate, separated by 150mm, the primary and secondary inductances raise to 18.462
µH and 18.952 µH respectively, resulting in a increase of nearly 0.7 µH.
For measurement of the coupling coefficient, three methods will be used and
compared. The first and simplest method consists of measurement of the secondary-
primary voltage ratio using the gain measurement in the Bode 100. The implementation
of this technique was based on a Bode 100 article [10], where the equipment set-up
information and derivation of the coupling factor equation are described. A typical gain
curve measured with the Bode 100 is shown in Figure 7.27 and the following steps are
undertaken for definition of the coupling coefficient using this method:
𝑉2 𝐿2
𝑔𝑎𝑖𝑛 = ≈ 𝑘√
𝑉1 𝐿1
(7.10)
185
The second technique employed for measurement of coupling factor is the series-
aiding series-opposing method, previously described in [11], [12]. The following
methodology is used for the coupling coefficient calculation.
𝐿𝑎𝑖𝑑 = 𝐿1 + 𝐿2 + 2𝑀
𝐿𝑜𝑝 = 𝐿1 + 𝐿2 − 2𝑀
𝐿𝑎𝑖𝑑 − 𝐿𝑜𝑝
𝑀=
4
𝑀
𝑘=
√𝐿1 𝐿2
(7.11)
(a) (b)
Figure 7.28: Inductance measurement in (a) series-aiding and (b) series-opposing
configurations
The third and last method is the short and open circuit technique, previously
analysed in [11], [12]. The coupling coefficient is calculated taking the following steps.
186
2. Measurement of open circuit inductances in the primary (𝐿1 ) and secondary (𝐿2 )
sides;
𝐿𝑠ℎ𝑜𝑟𝑡
𝑘 = √1 −
𝐿1
(7.12)
Table 7.8: Coupling coefficient when primary and secondary pads are separated
by 150 mm
Following the measurements for the standard, aligned placement of the pads, the
influence of misalignment on the coupling coefficient is analysed. Experimental results
considering the three measurement methods previously described are shown in Figure
7.29 for misalignment in the X axis, in Figure 7.30 for misalignment in the Y axis and
in Figure 7.31 for the Z axis influence.
The open/short circuit method is the only measurement technique that gives only
the absolute value of the coupling coefficient. Therefore, to facilitate the comparison
187
between the methods in Figure 7.29, the results for the open/short circuit method are
plotted incorporating the sign information acquired with the gain and series methods.
188
Analysing the misalignment graphs it is possible to conclude that results are very
close between the three methods. Also, the behaviour of the coupling coefficient when
subject to misalignment was similar to what was observed in the FEA simulations
conducted in section 7.5.
From the experimental results obtained with the three methods, an average value of
coupling coefficient was calculated and compared to results obtained in the updated
FEA simulation, as shown in Figure 7.32. The data for each test point is presented in
Tables 7.9, 7.10 and 7.11.
For the designed operation point, with no misalignment and distance of 150mm
between pads, the difference between simulation (0.366) and experimental results
(0.296) for coupling coefficient is 0.07. An average absolute difference of 0.078 is
observed for Z axis measurements, 0.067 for Y axis and 0.080 for X axis misalignment.
(a) (b)
(c)
Figure 7.32: Coupling coefficient in function of (a) offset in X axis (b) offset in Y
axis and (c) distance between pads in Z axis
189
A few practical factors contribute to the difference between simulation and
experimental results. Overall, due to instrument and human errors in the construction
process, the elements of the prototype will not have the exact same dimensions as in the
FEA model. In simulation, the Double D coils have the shape of perfect rectangles, as
previously shown in Figure 7.10, however, due to difficulties to reproduce that shape
with a thick wire, the corners of the coils are slightly curved in the practical set-up
(Figure 7.25). Also, during the misalignment tests, there is an error associated to the
placement of the pads; in simulation the misalignment implemented is an exact value
but in a practical situation there are small errors in the measurement of these distances
that may influence the results.
Misalignment 𝒌 𝒌
x axis experimental FEA simulation
0 0.296 0.366
50 0.269 0.345
100 0.205 0.283
150 0.118 0.198
200 0.031 0.104
250 0.057 0.017
300 0.109 -0.048
350 0.129 -0.082
400 0.119 -0.089
450 -0.099 -0.079
500 -0.075 -0.062
550 -0.054 -0.044
600 -0.035 -0.029
190
Table 7.10: Coupling coefficient (k) in function of misalignment on Y axis
Misalignment 𝒌 𝒌
Y axis experimental FEA simulation
0 0.296 0.366
50 0.271 0.355
100 0.248 0.329
150 0.210 0.293
200 0.174 0.251
250 0.129 0.205
300 0.088 0.160
350 0.056 0.120
400 0.035 0.089
450 0.023 0.066
500 0.019 0.051
Distance
𝒌 𝒌
between pads
experimental FEA simulation
Z axis
100 0.442 0.513
125 0.361 0.433
150 0.296 0.366
175 0.232 0.315
200 0.190 0.272
225 0.155 0.237
250 0.128 0.209
275 0.103 0.186
300 0.086 0.167
191
7.6.2 Evaluation of converter operation
Once the coils are tested and inductances and coupling coefficient are measured,
these structures are ready to be connected to compensation capacitors, H-bridges and
the switching circuit for the converter validation.
Since the experimental values of the inductors were slightly lower than what was
calculated in the design stage, the value of compensation capacitors is recalculated
using equation (7.2) to maintain the resonant frequency at 100 kHz. Consequently, the
design value increases from 120 nF to approximately 134 nF. The measured values for
total compensation capacitance (Figure 7.33) are 133 nF for primary and 132 nF to the
secondary side.
For the converter experimental validation, resistances and parasitic elements are
measured with the Bode 100 and incorporated to the circuit used for Spice and cyclic-
averaging simulations. The equivalent circuit for the resonant tank is shown in Figure
7.34-a. The equivalent circuit for the inductor, described by equation (7.13), is the same
used in the experimental analysis of the CLLC converter in Chapter 6, considering an
inductance connected to a series resistance, parasitic capacitance in parallel and parallel
resistance.
192
incorporated to the simulation model for comparison. As shown in equation (7.14),
these values are incorporated into the equivalent circuit of the capacitor. Simplifying
the equivalent circuit for the inductors to a 𝑍𝑙𝑒𝑞 , as in (7.13) and capacitors to a 𝑍𝑐𝑒𝑞 , as
in (7.14), the resonant tank circuit can be reduced to the equivalent shown in Figure
7.34-b. In this circuit, the equivalent resistances are obtained adding the resistance of
the equivalent inductor (𝑅𝑙𝑒𝑞 ) to the resistance of the equivalent capacitor (𝑅𝑐𝑒𝑞 ), as
shown in equation (7.15). Once this reduced IPT resonant circuit is defined, equations
(7.1) and (7.9) can be used to obtain the CLLC equivalent circuit for the cyclic-
averaging simulation.
(a)
(b)
Figure 7.34: Experimental and theoretical coupling coefficients in function of (a)
offset in X axis (b) offset in Y axis
1
𝑍𝑙𝑒𝑞 = 𝑅𝑙𝑒𝑞 + 𝑗𝜔𝐿𝑒𝑞 =
1 1
+ + 𝑗𝜔𝐶𝑝
𝑅𝑝 𝑅𝑠 + 𝑗𝜔𝐿
(7.13)
1 𝑗
𝑍𝑐𝑒𝑞 = 𝑅𝑐𝑒𝑞 − 𝑗 = 𝑅𝑐 + 𝑅𝑤 + 𝑗𝜔𝐿𝑤 −
𝜔𝐶𝑒𝑞 𝜔𝐶𝑠2 ′
(7.14)
𝑅𝑒𝑞1 = 𝑅𝑐𝑒𝑞1 + 𝑅𝑙𝑒𝑞1
193
𝑅𝑒𝑞2 = 𝑅𝑐𝑒𝑞2 + 𝑅𝑙𝑒𝑞2
(7.15)
The measured values for all elements of the circuit from Figure 7.34-a are listed in
Table 7.12. The elements of the simplified circuit from Figure 7.34-b and CLLC
equivalent circuit for the cyclic-averaging simulation are listed in Table 7.13.
Parameter Value
Primary 𝑪𝒔𝟏 133 nF
Capacitor 𝑹𝒄𝟏 0.06 Ω
Secondary 𝑪𝒔𝟐 132 nF
Capacitor 𝑹𝒄𝟐 0.04 Ω
𝑳𝟏 18.42 𝜇H
Primary 𝑹𝒔𝟏 0.155 Ω
Inductor 𝑹𝒑𝟏 2.1 kΩ
𝑪𝒑𝟏 205.88 pF
𝑳𝟐 18.87 𝜇H
Secondary 𝑹𝒔𝟐 0.166 Ω
Inductor 𝑹𝒑𝟐 2.75 kΩ
𝑪𝒑𝟐 201.44 pF
Wire 𝑳𝒘𝟏 1.09 𝜇H
primary 𝑹𝒘𝟏 0.140 Ω
Wire 𝑳𝒘𝟐 1.12 𝜇H
secondary 𝑹𝒘𝟐 0.120 Ω
194
Table 7.13: Equivalent circuit parameters for reduced IPT converter and CLLC
converter
Parameter Value
𝑳𝒆𝒒𝟏 18.44 μH
𝑹𝒆𝒒𝟏 0.420 Ω
For the experimental verification, the output of the converter (battery in forward
operation or DC bus in reverse operation) is modelled as an equivalent load. As in
Chapter 6, these values are calculated using the simulation results at maximum
𝑉𝑜𝑢𝑡
modulation, where 𝑅𝑙𝑜𝑎𝑑 = 𝐼𝑜𝑢𝑡
. The design and practical values of resistance obtained
Once all parameters are defined, experimental results are compared to Spice
simulations of the IPT converter and cyclic-averaging simulations of the equivalent
CLLC converter. Results for SPS modulation are presented in Figure 7.35 while the
PPM modulation case is analysed in Figure 7.36. The data for each test point is
presented in Tables 7.15 and 7.16.
From analysis of the results, it is possible to conclude that both Spice and cyclic-
averaging methods can accurately represent the converter’s behaviour. At maximum
195
modulation, the error between Spice and experimental results is 50 mA, or 0.47%, for
forward operation and 70 mA, or 5.43%, in reverse operation. Comparing experimental
to cyclic-averaging results, also at maximum modulation, the error is equal to 40 mA,
or 0.38%, in forward operation and 70mA, or 5.43%, in reverse operation.
Overall, the errors between simulation and experimental results are lower when
operating with PPM modulation. As the modulation angle decreases in operation with
SPS modulation, the difference between simulation and experimental results increases.
It was observed in simulation that this is not solely due to the influence of delays in the
system, as observed in Chapter 6. When using SPS modulation, the output current is
more sensitive to the resonant tank components, and consequently the primary and
secondary wire inductance values. For a phase-shift angle of 10 degrees, the output
current in forward mode would drop from 2.68 to 2.01 (670 mA difference) if the wire
inductance was not considered in the Spice simulation. For the PPM modulation case,
the same comparison, at the lowest modulation angle, would result in a difference of
only 25 mA.
The error between experimental and simulation results in this chapter was reduced
compared to the analysis made in the previous chapter for the CLLC converter. This is
mainly due to the addition of the primary and secondary wire resistances and
inductances to the simulation of the IPT converter, while in Chapter 6 the influence of
long wires was neglected.
(a) (b)
Figure 7.35: Experimental and simulation results for 48-12V SPS modulated
converter operating in (a) forward and (b) reverse modes
196
(a) (b)
Figure 7.36: Experimental and simulation results for 48-12V PPM modulated
converter operating in (a) forward and (b) reverse modes
197
Table 7.16: Results for PPM modulation (48 – 12 V)
During experimental verification it was observed that output power and efficiency
were significantly reduced when operating in reverse mode. At maximum modulation
in forward mode, the measured output power is 130.206 W with efficiency of 56.6%,
while in reverse mode the output power dropped to 44.65 W with efficiency of 40.6%.
While in a nearly ideal simulation both forward and reverse operations have similar
output power and high efficiency, when the resistances and parasitic elements from the
real circuit are added to the simulation model, the output power and efficiency in
reverse mode are considerably reduced. Tests were performed using the Spice
simulation with voltage sources at input and output sides, to verify the influence of the
input-output voltage ratio in the converter’s behaviour. According to the simulation
results shown in Table 7.17, when the input and output voltage match, the results for
forward and reverse operation are very similar and the efficiency is increased. As the
difference between input and output voltages increase, the output power in reverse
mode is considerably lower than in forward mode and the efficiency drops, more
significantly for reverse operation.
198
Table 7.17: Influence of converter voltage ratio in the output power and efficiency
of the system at maximum modulation
In this research project, the converters were not built to achieve maximum
efficiency. The primary objective here is to verify the accuracy of the models in the
representation of the converter behaviour. Therefore, provided that most of the losses
are incorporated into the simulation and the models give accurate results, it is
acceptable to have a low efficiency converter. For converters built with optimized
efficiency, the voltage mismatch would not be a problem, since when the losses are
significantly reduced both forward and reverse modes perform well. Improved
efficiency can be achieved reducing the quantity and length of wires in the circuit,
using SiC or GaN MOSFETs with reduced on-resistance and prioritizing efficiency in
the inductor design and construction process.
Table 7.18: Load definition for forward and reverse tests for 24-24V tests
Experimental and simulation results are compared, as shown in Figures 7.37 and
7.38, and the values of output current for each test point are presented in Tables 7.19
and 7.20. Now the graphs for forward operation are nearly identical to the reverse
199
operation results. At maximum modulation in forward mode, the current error between
experimental and Spice and cyclic-averaging simulations is 60 mA (1.32%), while in
reverse mode the difference is 30 mA (0.65%) to Spice results, and 40 mA (0.87%) to
the cyclic-averaging results.
(a) (b)
Figure 7.37: Experimental and simulation results for 24-24V SPS modulated
converter operating in (a) forward and (b) reverse modes
(a) (b)
Figure 7.38: Experimental and simulation results for 24-24V PPM modulated
converter operating in (a) forward and (b) reverse modes
200
Table 7.19: Results for SPS modulation (24 – 24 V)
The input and output signals measured at maximum modulation and forward
operation are shown in Figures 7.39 and 7.40. Due to inaccuracies inherent of
201
experimental tests, as instrument errors, influence of the temperature in the
measurements, and the fact the measured load value (5.39 Ω) is not exactly the same as
the theoretical load value (4.97 Ω), the experimental voltage conversion ratio is not
equal to one, but still very close, as shown in (7.16).
𝑉𝑜𝑢𝑡 24.559
𝑐𝑜𝑛𝑣𝑒𝑟𝑠𝑖𝑜𝑛𝑟𝑎𝑡𝑖𝑜 = = = 1.03
𝑉𝑖𝑛 23.795
(7.16)
202
7.7 Conclusions
In this chapter a series compensated IPT converter was analysed. A FMA model
was derived and employed in the converter design. From simulation results, it was
observed that FMA still suffers from accuracy problems due to the fundamental
approximation, especially when analysing the primary current. For the secondary
current and output power analysis in a low loss circuit, the FMA model performed well.
For the design of the primary and secondary coils, a finite element analysis (FEA)
simulation was performed and the influence of construction parameters and
misalignment was analysed. In both simulation and experimental results, it was verified
that the Double D coil is more sensitive to misalignments in the X axis. To achieve
good performance for misalignments in both X and Y axis, more complex coils can be
used, as Double D Quadrature or Bipolar coils.
Finally, the influence of the voltage conversion ratio in the output power and
efficiency of the converter was evaluated. It was observed that, when the converter has
increased loss, performance in reverse mode deteriorates as the conversion ratio
diverges from unity, with reduced output power and efficiency compared to forward
mode results. The prototype was then tested with an input-output voltage ratio close to
one, which resulted in better performance in reverse mode and overall improvement of
efficiency.
203
7.8 References
[1] Z. U. Zahid, Z. M. Dalala, R. Chen, B. Chen, and J. S. Lai, “Design of bidirectional
DC-DC resonant converter for Vehicle-to-Grid (V2G) applications,” IEEE
Transactions on Transportation Electrification, vol. 1, no. 3, pp. 232–244, 2015.
[2] A. A. S. Mohamed, A. Berzoy, and O. Mohammad, “Magnetic design
considerations of bidirectional inductive wireless power transfer system for EV
applications,” IEEE Transactions on Magnetics, vol. 53, no. 6, pp. 1–5, 2017.
[3] A. A. S. Mohamed, A. Berzoy, F. G. N. Nogueira de Almeida, and O. Mohammed,
“Modeling and Assessment Analysis of Various Compensation Topologies in
Bidirectional IWPT System for EV Applications,” IEEE Transactions on Industry
Applications, vol. 53, no. 5, pp. 4973–4984, 2017.
[4] Y. Zhang, T. Kan, Z. Yan, and C. C. Mi, “Frequency and Voltage Tuning of
Series-Series Compensated Wireless Power Transfer System to Sustain Rated
Power under Various Conditions,” IEEE Journal of Emerging and Selected Topics
in Power Electronics, vol. 7, no. 2, pp. 1311–1317, 2019.
[5] S. Wang, J. Chen, Z. Hu, C. Rong, and M. Liu, “Optimisation design for series–
series dynamic WPT system maintaining stable transfer power,” IET Power
Electronics, vol. 10, no. 9, pp. 987–995, 2017.
[6] A. Zaheer, D. Kacprzak, and G. A. Covic, “A bipolar receiver pad in a lumped IPT
system for electric vehicle charging applications,” 2012 IEEE Energy Conversion
Congress and Exposition, ECCE 2012, pp. 283–290, 2012.
[7] G. a. Covic and J. T. Boys, “Inductive Power Transfer,” Proceedings of the IEEE,
vol. 101, no. 6, pp. 1276–1289, 2013.
[8] J. Kim et al., “Coil design and shielding methods for a magnetic resonant wireless
power transfer system,” Proceedings of the IEEE, vol. 101, no. 6, pp. 1332–1342,
2013.
[9] SAE International, “J2954TM Wireless Power Transfer for Light-Duty Plug-In/
Electric Vehicles and Alignment Methodology,” Surface Vehicle Information
Report, 2016.
[10] L. Heinzle, “Measuring the Mutual Interaction between Coaxial Cylindrical Coils
with the Bode 100,” Bode 100 Article, pp. 1–11, 2013.
[11] K. Hata, T. Imura, and Y. Hori, “Simplified measuring method of kQ product for
wireless power transfer via magnetic resonance coupling based on input impedance
measurement,” IECON 2017 - 43rd Annual Conference of the IEEE Industrial
Electronics Society, pp. 6974–6979, 2017.
[12] B. Hesterman, “Analysis and Modeling of Magnetic Coupling,” Denver Chapter
Seminar, IEEE Power Electronics Society, pp. 1–93, 2007.
204
8 Conclusions and Further Work
8.1 Conclusions
In this thesis the cyclic-averaging method was successfully applied to model a
DAB converter, a 4th order CLLC resonant converter and a series compensated IPT
converter. The technique showed good performance, accuracy and significantly faster
execution time compared to more traditional methods of analysis as FMA, state-
variable and Spice. Here the novel cyclic-averaging model was mainly used to evaluate
the influence of the control variables (phase-shift angles in Chapters 3, 5, 6 and 7 or
frequency in Chapter 4) on the converter’s behaviour. However, the models developed
in this thesis can also be applied to analyse the influence of the circuit components in
order to obtain a faster design process.
The Dual Active Bridge converter operating under Single Phase-Shift (SPS)
modulation was analysed in Chapter 3, considering bidirectional operation. The
operation modes of the converter were defined and the cyclic-averaging method was
applied for calculation of the converter’s output current in function of the phase-shift
angles, resulting in a rapid analysis with accuracy similar to Spice. The cyclic method
205
was also successfully employed to perform a harmonic analysis of the state-variables,
where harmonic components were calculated and used for waveform reconstruction.
The waveforms obtained from the harmonic analysis were very close to the Spice
simulation results. The main benefit observed for the cyclic-averaging method was the
fast execution, being nearly 33 times faster than Spice for calculation of the converter’s
output and the state-variables harmonics or nearly 166 times faster when the harmonic
analysis and waveform reconstruction are not performed.
Analysing the cyclic-averaging implementation for the DAB and CLLC converters,
it was possible to conclude that the complexity of the cyclic-averaging model increases
for higher order resonant networks and is highly dependent of the modulation technique
207
chosen. More complex modulation techniques will result in more variables affecting the
converter’s behaviour and, consequently, more operating modes to be analysed.
Therefore, despite the benefits of rapid and accurate analysis, the implementation of
this method may become too complicated for converters with significantly more
complex resonant networks or when intricate modulation methods are used.
In this thesis the models were validated with the construction of a small-scale
prototype (100 W converter). The application of cyclic-averaging models for converters
operating at higher power levels was verified only through simulation in Chapter 3, for
a 7kW converter, and in Chapter 4, for a 3.5 kW converter, also resulting in an accurate
analysis. The cyclic-averaging models here developed are still valid for converters used
in EV charging applications that operate at higher power levels as long as the loss and
parasitic elements are measured and incorporated into the equivalent circuits following
the same methodology used to verify the low power prototypes in Chapters 6 and 7.
In this thesis the cyclic-averaging method was applied for the simplest type of
phase-shift modulation for the DAB converter and for the most popular modulation
techniques for the CLLC and IPT series compensated converters. Depending on the
modulation technique utilized, the converter operation must be reanalysed for the
definition of the operating modes and duty cycles. Therefore, the cyclic analysis here
developed has the potential to be adapted for various modulation techniques and
converters topologies that were not considered in this research.
208
frequency-domain analysis, the RTFMA method [1], [2], that propose modifications to
the conventional FMA technique to achieve increased accuracy, can be investigated for
the converter topologies here analysed.
8.3 References
[1] H. I. Sewell, M. P. Foster, C. M. Bingham, D. A. Stone, D. Hente, and D. Howe,
“Analysis of voltage output LCC resonant converters, including boost mode
operation,” IEE Proceedings - Electric Power Applications, vol. 150, no. 6, pp.
673–679, 2003.
[2] M. P. Foster, “Analysis and Design of High-order Resonant Power Converters,”
PhD Thesis. University of Sheffield, 2003.
209
Appendix A.1
The relative errors between FMA and Spice simulations from chapter 5, for the
following variables: output power, primary current, secondary current and magnetizing
current, are shown in Tables A1.1 to A1.4.
The efficiency as function of leg phase-shift angles 𝛼1 and 𝛼2 obtained from the
Spice simulation is presented in Table A1.5.
Table A1.1: Output power relative error (%) between FMA and Spice
considering 𝝓 = 𝟗𝟎°, where: Green: <5%, Orange : 5-10% and Red: >10%
𝜶𝟐
𝜶𝟏 18 36 54 72 90 108 126 144 162 180
18 14.48 9.71 0.64 -12.10 -27.57 -37.08 -43.40 -54.71 -69.95 -87.49
36 12.41 13.08 8.81 0.01 -12.50 -23.75 -30.99 -39.21 -51.74 -67.45
54 4.42 9.90 10.40 5.99 -2.79 -12.39 -20.62 -28.25 -37.88 -51.24
72 -8.06 1.54 6.51 6.60 1.99 -4.67 -11.61 -19.23 -27.61 -38.22
90 -23.72 -10.91 -2.06 2.25 1.88 -1.20 -5.63 -11.75 -19.52 -28.60
108 -33.43 -22.25 -11.64 -4.31 -1.05 -0.58 -1.99 -5.75 -11.88 -20.06
126 -39.53 -29.43 -19.83 -11.16 -5.37 -1.85 -0.22 -1.19 -5.00 -11.45
144 -50.24 -37.37 -27.30 -18.66 -11.36 -5.47 -1.03 0.91 -0.15 -4.25
162 -64.79 -49.50 -36.67 -26.87 -18.99 -11.47 -4.688 0.02 1.86 0.52
180 -81.96 -64.97 -49.84 -37.34 -27.99 -19.58 -11.07 -3.99 0.63 2.20
210
Table A1.2: Primary current (|𝑰𝟏 |) relative error (%) between FMA and Spice,
considering 𝝓 = 𝟗𝟎°, where: Green: <5%, Orange: 5-10% and Red: >10%
𝜶𝟐
𝜶𝟏 18 36 54 72 90 108 126 144 162 180
18 8.89 5.37 3.75 2.48 1.43 0.67 0.41 0.47 0.64 0.72
36 12.97 7.14 5.28 3.90 2.65 1.57 0.97 0.88 1.04 1.16
54 13.63 7.19 5.64 4.56 3.46 2.38 1.60 1.27 1.33 1.46
72 12.10 6.24 5.15 4.47 3.71 2.85 2.14 1.71 1.58 1.63
90 11.78 5.95 4.77 4.23 3.69 3.08 2.56 2.22 2.01 1.88
108 14.38 7.44 5.39 4.46 3.85 3.33 2.98 2.79 2.66 2.47
126 14.98 9.17 6.57 5.12 4.22 3.61 3.32 3.28 3.30 3.20
144 14.02 9.27 7.34 5.81 4.66 3.88 3.54 3.57 3.74 3.81
162 12.78 7.94 6.87 5.95 4.91 4.04 3.62 3.63 3.88 4.11
180 10.35 5.66 5.31 5.09 4.60 3.91 3.49 3.47 3.73 4.05
Table A1.3: Secondary current (|𝑰𝟐 |) relative error (%) between FMA and Spice
considering 𝝓 = 𝟗𝟎°, where: Green: <5%, Orange: 5-10% and Red: >10%
𝜶𝟐
𝜶𝟏 18 36 54 72 90 108 126 144 162 180
18 26.34 64.23 86.17 84.81 60.22 21.39 6.77 34.65 73.61 93.84
36 6.38 18.76 28.61 30.64 23.50 9.851 1.04 7.61 22.73 32.97
54 1.78 6.80 11.88 13.90 11.51 5.49 0.46 1.31 8.23 14.53
72 0.39 2.48 4.98 6.20 5.29 2.41 -0.15 -0.15 2.76 6.30
90 0.35 1.41 2.48 2.87 2.18 0.58 -0.58 -0.24 1.43 3.00
108 0.81 1.86 2.38 2.17 1.26 -0.004 -0.56 0.18 1.72 2.65
126 0.85 2.31 2.95 2.523 1.36 -0.01 -0.49 0.41 2.06 3.04
144 0.65 2.14 3.22 3.04 1.80 0.22 -0.48 0.28 1.98 3.18
162 0.44 1.64 2.83 3.15 2.17 0.52 -0.489 -0.04 1.52 2.86
180 0.31 1.17 2.14 2.63 2.16 0.72 -0.458 -0.36 0.91 2.21
211
Table A1.4: Magnetizing current (|𝑰𝒎 |) relative error (%) between FMA and Spice
considering 𝝓 = 𝟗𝟎°, where: Green: <5%, Orange: 5-10% and Red: >10%
𝜶𝟐
𝜶𝟏 18 36 54 72 90 108 126 144 162 180
18 6.28 12.49 13.62 13.44 13.11 13.05 13.36 13.96 14.65 15.16
36 -3.48 4.70 11.04 14.75 17.09 18.89 20.56 22.22 23.79 25.07
54 -8.31 -4.18 3.39 9.90 14.91 18.92 22.34 25.41 28.15 30.44
72 -10.88 -11.12 -5.02 2.39 9.17 15.02 20.11 24.60 28.56 31.92
90 -12.42 -16.28 -12.71 -5.79 1.73 8.83 15.26 21.02 26.11 30.46
108 -13.48 -20.24 -19.46 -13.86 -6.36 1.42 8.82 15.62 21.72 26.99
126 -14.30 -23.43 -25.41 -21.59 -14.67 -6.66 1.39 9.02 15.99 22.11
144 -14.99 -26.18 -30.83 -29.05 -23.07 -15.16 -6.71 1.56 9.28 16.18
162 -15.60 -28.62 -35.92 -36.40 -31.62 -24.01 -15.34 -6.55 1.81 9.41
180 -16.13 -30.85 -40.80 -43.79 -40.46 -33.28 -24.46 -15.234 -6.30 1.94
212
Appendix A.2
The experimental results obtained for the CLLC resonant converter operating
under SPS and PPM modulation are presented in Tables A2.1, A2.2, A2.3 and A2.4.
The input voltages applied to the system for each test point are shown in Tables A2.5
and A2.6.
Test A: real component loss elements not fully considered and deadtime not
considered (test case exclusive to cyclic-averaging simulations);
Test B: all measured loss elements incorporated into model, deadtime influence not
considered;
213
Table A2.2: Results for single phase-shift modulation reverse operation
214
Table A2.4: Results for pulse-phase modulation reverse operation
215
Table A2.6: Measured input voltage PPM tests
216