Huang D D 2014

Investigation of Topology and Integration for
Multi-Element Resonant Converters
Daocheng Huang
Dissertation submitted to the faculty of the

Virginia Polytechnic Institute and State University
in partial fulfillment of the requirements for the degree of
Doctor of Philosophy
in
Electrical Engineering
Fred C. Lee, Chair

Dushan Boroyevich
Paolo Mattavelli
Jaime De La Ree
Dwight Viehland
December 6th, 2013

Blacksburg, Virginia
Keywords: Integration, Topology, LLC resonant converter
© 2013, Daocheng Huang

Investigation of Topology and Integration for
Multi-Element Resonant Converters
Daocheng Huang
ABSTRACT
With the fast development of communication systems, computers and consumer
electronics, the power supplies for telecoms, servers, desktops, laptops, flat-panel TVs,
LED lighting, etc. are required for more efficient power delivery with smaller spaces. The
LLC resonant converter has been widely adopted for these applications due to the
advantages in high efficiency, high power density and holdup time operation capability.
However, LLC resonant converter meets some issues, especially in high output
current applications. Those issues include magnetic design, start-up, short-circuit
protection, synchronous rectifier drive, EMI noise and integration, etc.
To solve those issues, like start-up and short-circuit protection, SR driving and EMI, etc., a
synthesis method is proposed to find the similar resonant topologies like LLC. Based on
this method, lots of multi-element resonant converters are found to solve the issues that
LLC resonant converter cannot handle.
To evaluate the performance of found numerous valuable topologies. Thus, a general
evaluation system is required. State-plane analysis with new normalization factors is
utilized. Based on it, the voltage stress, current stresses and apparent power of resonant
converters are easy to compare. This method can help select suitable circuit topology for
certain applications. Meanwhile, it also can help resonant converters’ design. The
important performance factors, like start-up, short-circuit protection, SR driving,
integration and EMI performance, are also taken into account for the whole evaluation
system.
The high switching frequency is needed recently for high power density requirement.
However, LLC resonant converter suffers high transformer loss. Matrix transformer is
introduced to reduce winding loss and total volume. Flux cancellation method is utilized to
reduce core size and loss. Synchronous Rectifier (SR) devices and output capacitors are
integrated into secondary windings to eliminate termination related winding losses, via loss and
reduce leakage inductance.
The passive integration is necessary for high power density resonant converter,
especially for high order system. Based on stress, suitable passive components are chosen
for integration. Then, the magnetic integration method is shown based on multi-winding
transformer structure. The passive integration principles are discussed. A novel passive
integration method is proposed for multi-elements resonant converters.
In conclusion, this work is focus on the topology analysis and integration of resonant
converters. Searching the suitable topologies for certain application, and evaluate the
performance of them. Then, improve the system power density by integration techniques.
iii
To My Family:
My parents: Liugen Huang
Fengling Luo
iv
Acknowledgments
First I would like to express my sincere gratitude to my advisor, Dr. Fred C. Lee. It
is him who shows me how to climb the mountain of research and see the beautiful
scenery of power electronics. It is his depths of knowledge and experience that gave me
the power to break the rock blocking my research path. Besides his board knowledge, his
rigorous research attitude and logical way of thinking have been a great value for me.
I would like to thank Dr. Dushan Boroyevich for his valuable guidance during my
research. I am also grateful to his enthusiastic help during the CPES annual conference
when I was the fresh chairman. I also want to acknowledge the other members of my
advisory committee, Dr. Paolo Mattavelli, Dr. Jaime De La Ree and Dr. Dwight Viehland
for their support, comments, and suggestions.
I am especially indebted to my colleagues in the Power Management Mini-
Consortium Group. It has been a great pleasure to work with the talented, creative,
helpful and dedicated colleagues. I would like to thank all the members of our teams: Dr.
Ming Xu, Dr. Shuo Wang, Dr. Julu Sun, Dr. Authur Ball, Dr. Yan Jiang, Dr. Jian Li, Dr.
Yan Dong, Dr. Chuanyun Wang, Dr. Dianbo Fu, Dr. Pengju Kong, Dr. David Reusch,
Dr. Qiang Li, Dr. Zheng Zhao, Dr. Mingkai Mu, Dr. Yingyi Yan, Dr. Weiyi Feng, Mr.
Doug Sterk, Mr. Bin Huang, Mr. Yucheng Ying, Mr. Yi Sun, Mr. Ya Liu, Mr. Clark
Person, Mr. Zheng Luo, Mr. Pengjie Lai, Mr. Zijian Wang, Mr. Qian Li, Ms. Yin Lu, Mr.
Chanwit Prasantanakorn, Mr. Shu Ji, Mr. Feng Yu, Mr. Haoran Wu, Mr. Yipeng Su, Mr.
Wei Zhang, Mr. Li Jiang, Mr. Shuilin Tian, Mr. Pei-Hsin Liu, Dr. Dongbin Hou, Mr.
v
Zhiqiang Wang, Mr. Xiucheng Huang, Mr. Zhengyang Liu, Mr. Yucheng Yang, Mr.
Yang Jiao, Mr. Sizhao Lu, Mr. Chao Fei, Mr. Xuebing Chen, Ms. Yincan Mao, Mr.
Zhongsheng Cao, Mr. Syed Bari. Dr. Marcos Alonso, Dr. Fang Luo, Dr. Ke Jin, Dr.
Xiaoyong Ren, Dr. Fanghua Zhang, Dr. Yuling Li, Dr. Shaojun Xie, Dr. Xinke Wu, Dr.
Brian Cheng, Dr. Feng Wang, Dr Xin Ming, Dr. Feng Zheng.
My thanks also go to all of the other students I have met in CPES, especially to Dr.
Jing Xu, Dr. Michele Lim, Dr. Yan Liang, Dr. Rixi Lai, Dr. Honggang Sheng, Dr. Di
Zhang, Dr Puqi Ning, Dr. Tim Thacker, Dr. Carson Baisden, Dr. Xiao Cao, Dr. Zhiyu
Shen, Dr. Ruxi Wang, Dr. Dong Jiang, Dr. Dong Dong, Dr. Hemant Bishnoi, Dr. Zhen
Chen, Ms. Yiying Yao, Mr. Jing Xue, Ms. Zhuxian Xu, Mr. Zheyuan Tan, Mr. Xuning
Zhang, Mr. Bo Wen, Mr. Lingxiao Xue, Mr. Zhemin Zhang, Mr. Bo Zhou, Mr. Ying
Wang, Mr. Li Jiang, Mr. Marko Jaksic, Ms. Han Cui.
I would also like to thank all the wonderful staffs in CPES: Ms. Linda Gallagher,
Ms. Teresa Rose, Ms. Marianne Hawthorne, Ms. Linda Long, Mr. Robert Martin, Mr.
Jamie Evans, Mr. Dan Huff, Dr. Wenli Zhang and Mr. David Gilham for their support
and help over the past six years. I also hope give my special thanks to Ms. Teresa Shaw
for her sincere help during the CPES annual conference.
With much love and gratitude, I want to thank my parents: my mother Fengling Luo
and my father Liugen Huang for their endless love, encouragement, and support
throughout more than twenty years of school.
Special thanks go out to the CPES PMC mini-consortium members for the funding
of my research: Chicony Power, CSR Zhuzhou Institute Co., Ltd., Delta
vi
Electronics, Huawei Technologies, Infineon Technologies, International Rectifier, Linear
Technology, Murata Manufacturing Co., Ltd., NEC TOKIN Corporation, NXP
Semiconductors, Richtek Technology, Texas Instruments.
vii
Table of Contents
Chapter 1. Introduction ................................................................................... 1
1.1 Background and Motivation........................................................................................... 1
1.2 Issues of PWM DC-DC Power Conversions ................................................................. 4
1.2.1 Pros and Cons of Soft-switching PWM DC-DC Converter .....................................4
1.2.2 Limitation of PWM DC-DC Converter with Hold-up Time Requirement ..............6
1.3 Opportunity of Resonant DC-DC Converters ................................................................ 8
1.3.1 Basic Concepts of PWM Converter and Resonant Converters ................................8
1.3.2 Advantages of LLC Resonant Converter (LLC) with Hold-up Time Requirement13
1.4 Challenges of LLC Resonant Converter ...................................................................... 18
1.4.1 Start-Up and Short Circuit Protection ....................................................................18
1.4.2 Synchronous Rectifiers (SR) Driving Scheme .......................................................19
1.4.3 Chanllenges for High Output CurrentApplication .................................................23
1.4.4 EMI Performance of LLC Resonant Converter .....................................................25
1.5 Multi-Element Resonant Converters ............................................................................ 26
1.6 Opportunity of Emerging Technologies ...................................................................... 27
1.6.1 New System Architectures .....................................................................................28
1.6.2 Benefits and Chanllenges of LLC-DCX ................................................................30
viii
1.6.3 Wide Band Gap Devices —GaN HEMT ...............................................................31
1.7 Dissertation Outline ..................................................................................................... 32
Chapter 2. Synthesis of Multi-Element Resonant Converters........................ 36
2.1Fundamental Concept of Resonant Converters ............................................................. 36
2.1.1 Basic Resonant Cells (Two-Element Resonant Tanks) ..........................................38
2.1.2 Synthesis of Basic Resonant Cells .........................................................................41
2.3 Synthesis of Three-Element Resonant Converters....................................................... 42
2.4 Inherent Issues of LLC Resonant Converter ................................................................ 49
2.5 LLC-like Resonant Converters with Notch Filter ........................................................ 54
2.6 LLC-like Resonant Converters with Notch Filter and Two Band-pass Filters ............ 62
2.7 Conclusion ................................................................................................................... 66
Chapter 3. Evaluation of LLC-like Resonant Converters .............................. 67
3.1 Introduction .................................................................................................................. 67
3.2 State-Trajectory Analysis of LLC Resonant Converter ............................................... 67
3.3 New Normalization for State-plane ............................................................................. 79
3.4 Evaluation based on Startup, Synchronous Rectifier and EMI .................................... 89
3.5 Conclusion ................................................................................................................... 95
Chapter 4. Integrated System Design for High Current Application ............. 96
4.1 Introduction .................................................................................................................. 96
ix
4.2 Matrix Transformer with Flux Cancellation .............................................................. 101
4.3 Winding Structure Integrated with Synchronous Rectifier Devices .......................... 109
4.4 Matrix Transformer Design Procedure ...................................................................... 117
4.5 Experimental Results ................................................................................................. 120
4.6 Conclusion ................................................................................................................. 125
Chapter 5. Design Considerations for Passive Integration............................127
5.1 Introduction ................................................................................................................ 127
5.2 Basic Characteristics of Multi-elements resonant converters .................................... 128
5.3 Magnetic integration for multi-elements resonant converters ................................... 132
5.4 Passive Integration for Multi-element Resonant Converters ..................................... 140
5.5 Measurement of Transformer parameters .................................................................. 147
5.6 Thermal Improvement by Heat Extractor .................................................................. 152
5.7 Experimental Results ................................................................................................. 156
5.8 Conclusion ................................................................................................................. 157
Chapter 6. Conclusion and Future Work ......................................................159
6.1 Conclusion ................................................................................................................. 159
6.2 Future Work ............................................................................................................... 160
References ...................................................................................................162
x
List of Figures
Fig. 1.1 Ac-dc front-end converter ........................................................................................................... 2
Fig. 1.2 Efficiency trend for ac-dc front-end converters of redundant, datacenter applications .................. 3
Fig. 1.3 Power density trend for ac-dc front-end converters...................................................................... 3
Fig. 1.4 Typical Soft-switching PWM converters ..................................................................................... 4
Fig. 1.5 Waveforms of typical Soft-switching PWM converters ............................................................... 5
Fig. 1.6 Holdup time operation of ac-dc front-end converters ................................................................... 7
Fig. 1.7 Holdup time capacitance vs. input voltage range for dc-dc converters of the front-end converter . 7
Fig. 1.8 Duty cycle range for PWM converter with hold up requirement. ................................................. 8
Fig. 1.9 Simplified general isolated dc/dc PWM converter block diagram ................................................ 9
Fig. 1.10 Simplified general isolated dc/dc resonant converter block diagram .......................................... 9
Fig. 1.11 Band-pass filter concept for resonant converters...................................................................... 10
Fig. 1.12 Series resonant converter ........................................................................................................ 11
Fig. 1.13 Gain curves of series resonant converter ................................................................................. 11
Fig. 1.14 Parallel resonant converter ...................................................................................................... 12
Fig. 1.15 Gain curves of parallel resonant converter ............................................................................... 13
Fig. 1.16 LLC resonant converter .......................................................................................................... 14
Fig. 1.17Voltage gain of LLC resonant converter................................................................................... 14
Fig. 1.18 Waveforms of LLC resonant converter at resonant frequency.................................................. 15
Fig. 1.19 Commercial products comparison with LLC DC/DC and PWM DC/DC ................................. 17
Fig. 1.20 Market trend for LLC resonant Converter ............................................................................... 17
Fig. 1.21 LLC soft start-up in above resonant frequency region.............................................................. 18
Fig. 1.22 LLC soft starts process by commercial IC chip ....................................................................... 19
Fig. 1.23 PSFB circuit and devices’ driving signals................................................................................ 20
Fig. 1.24 Desired SR gate driving signals in different switching frequency regions ................................ 20
xi
Fig. 1.25 SR gate driving signals by secondary current sensing .............................................................. 21
Fig. 1.26 SR gate driving signals by primary current sensing ................................................................. 22
Fig. 1.27 The influence of package inductance ....................................................................................... 22
Fig. 1.28 Circuit schematic of the conventional transformer and SR structure. ....................................... 24
Fig. 1.29 Conventional transformer and the secondary side structure...................................................... 24
Fig. 1.30 CM noise spectrum comparison of PWM and LLC resonant converters. .................................. 25
Fig. 1.31 Desired voltage gain for LLC start-up. .................................................................................... 26
Fig. 1.32 Traditional AC distribution system for data center................................................................... 27
Fig. 1.33 380V DC distribution system for data center ........................................................................... 28
Fig. 1.34 DC power-train for IBM Supercomputer ................................................................................. 29
Fig. 1.35 Improved DC power-train for 400V to CPU ............................................................................ 30
Fig. 1.36 LLC resonant converter as DCX ............................................................................................. 31
Fig. 1.37 Specific on-resistance versus breakdown voltage for GaN and Si ............................................ 32
Fig. 2.1 Simple Block diagram for the basic concept of resonant converter ............................................ 37
Fig. 2.2 Zero current switching and zero voltage switching principles .................................................... 37
Fig. 2.3 Two-element resonant tanks...................................................................................................... 38
Fig. 2.4 Voltage gains of two-element resonant tanks............................................................................. 40
Fig. 2.5 Circuit and voltage gains of LCC resonant converter ................................................................. 41
Fig. 2.6 Circuit and voltage gains of LLC resonant converter ................................................................. 42
Fig. 2.7 Three-element resonant topology networks ............................................................................... 43
Fig. 2.8 Three-element resonant tanks.................................................................................................... 44
Fig. 2.9 Three-element resonant tanks for voltage source ....................................................................... 44
Fig. 2.10 Voltage gains and tanks of SRF+Notch ................................................................................... 45
Fig. 2.11 Voltage gains and tanks of Notch+SRF ................................................................................... 46
Fig. 2.12 Voltage gains and tanks of PRF+Notch ................................................................................... 46
xii
Fig. 2.13 Voltage gains and tanks of Notch+PRF ................................................................................... 47
Fig. 2.14 Voltage gains and tanks of SRF+PRF (LCC-like).................................................................... 47
Fig. 2.15 Voltage gains and tanks of PRF+SRF (LLC-like) .................................................................... 48
Fig. 2.16 DC characteristic of the LLC resonant converter ..................................................................... 49
Fig. 2.17 Soft start-up with low switching frequency ............................................................................. 50
Fig. 2.18 Soft start-up with a high switching frequency.......................................................................... 51
Fig. 2.19 Phase-shifted control in full-bridge LLC (fs=2f0, Φ=π/2) ......................................................... 52
Fig. 2.20 Optimal trajectory soft start-up ............................................................................................... 53
Fig. 2.21 Desirable Gain Curves for LLC-like resonant converters ......................................................... 54
Fig. 2.22 One example of selected 4-elements resonant converters ......................................................... 55
Fig. 2.23Four-element resonant topology networks ................................................................................ 57
Fig. 2.24 Gain curve characteristics vs. networks ................................................................................... 58
Fig. 2.25 Four-element resonant topology networks including Network 2(SRF) ..................................... 58
Fig. 2.26 Four-element resonant topology networks including Network 2(SRC) and Network 1(PRF).... 59
Fig. 2.27 Four-element resonant topology networks including Network 2(SRC), Network 1(PRF) and
Network 3 or 4 (NRF) ........................................................................................................................... 59
Fig. 2.28 Four-element resonant topologies including Network 2 (SRC), Network 1(PRF) and Network 3
or 4 (NRF)............................................................................................................................................. 60
Fig. 2.29 Four-element resonant topologies with desired resonant frequencies ....................................... 61
Fig. 2.30 One example of 5-element resonant converters ....................................................................... 62
Fig. 2.31 Power Delivery Difference between LLC & 5-e ...................................................................... 63
Fig. 2.32 Gain Curves of selected 5-elements resonant converter ........................................................... 63
Fig. 2.33 Five element resonant tank cell based on LLC with notch filter and 3rd harmonics band filter .. 64
Fig. 2.34 Experimental results for 5-element resonant converters under full load.................................... 66
Fig. 2.35 Experimental results of 5-element resonant converters under short-circuit condition................ 66
xiii
Fig. 3.1 Equivalent circuit of stage I ...................................................................................................... 68
Fig. 3.2 State-trajectory of stage I .......................................................................................................... 68
Fig. 3.3 Equivalent circuit of stage II ..................................................................................................... 70
Fig. 3.4 State-trajectory of stage II ......................................................................................................... 70
Fig. 3.5 Equivalent circuit of stage III .................................................................................................... 71
Fig. 3.6 State-trajectory of stage III........................................................................................................ 71
Fig. 3.7 Equivalent circuit of stage IV.................................................................................................... 73
Fig. 3.8 State-trajectory of stage IV ....................................................................................................... 74
Fig. 3.9 Equivalent circuit of stage V ..................................................................................................... 74
Fig. 3.10 State-trajectory of stage V ....................................................................................................... 75
Fig. 3.11 Equivalent circuit of stage VI.................................................................................................. 75
Fig. 3.12 State- trajectory of stage VI .................................................................................................... 76
Fig. 3.13 Steady-state time-domain waveforms when fs=f0 ..................................................................... 77
Fig. 3.14 Steady-state trajectory when fs=f0............................................................................................ 77
Fig. 3.15 Steady-state time-domain waveforms when fs<f0 ..................................................................... 78
Fig. 3.16 Steady-state trajectory when fs<f0............................................................................................ 78
Fig. 3.17 Steady-state time-domain waveforms when fs>f0 ..................................................................... 79
Fig. 3.18 Steady-state trajectory when fs>f0............................................................................................ 79
Fig. 3.19 LLC and CLL state-plane comparison based on conventional normalization ........................... 81
Fig. 3.20 Series resonant converter operates at resonant frequency ......................................................... 83
Fig. 3.21 State-plane of SRC operates at resonant frequency .................................................................. 83
Fig. 3.22 Steady-state trajectory when fs=f0............................................................................................ 84
Fig. 3.23 V-I stresses of LLC resonant converters for Ln=12.5 ............................................................... 85
Fig. 3.24 V-I stresses of LLC resonant converters for different Ln .......................................................... 85
Fig. 3.25 LCL Resonant Converter ........................................................................................................ 86
xiv
Fig. 3.26 V-I stresses of LCL resonant converters for different Ln .......................................................... 87
Fig. 3.27 CLL Resonant Converter ........................................................................................................ 87
Fig. 3.28 V-I stresses ofCLL resonant converters for different Ln ........................................................... 88
Fig. 3.29 V-I stresses ofthree-element resonant converters for different Ln ............................................. 88
Fig. 3.30 V-I stresses of three-element resonant converters for different Ln ........................................... 89
Fig. 3.31 Startup example of LLC Resonant Converter .......................................................................... 90
Fig. 3.32 Startup Comparison of LLC, CLL & LCL Resonant Converters .............................................. 91
Fig. 3.33 SR drive Comparison of LLC, CLL & LCL Resonant Converters ........................................... 92
Fig. 3.34 CM noise spectrum of LLC Resonant Converter ..................................................................... 93
Fig. 3.35 CM noise path of LLC Resonant Converter............................................................................. 93
Fig. 3.36 CM noise spectrum of LLC, CLL & LCL Resonant Converters............................................... 94
Fig. 3.37 CM noise spectrum of LLC, CLL & LCL Resonant Converters............................................... 95
Fig.4.1 Circuit schematic of the transformer and SR in [D.16]. .............................................................. 99
Fig.4.2 Circuit schematic and transformer structure in [D.17]. ............................................................. 100
Fig.4.3 Vicor BCM series. ................................................................................................................... 101
Fig. 4.4 LLC resonant converter with four sets output .......................................................................... 102
Fig. 4.5 12 layer PCB structure with one core and its MMF ................................................................. 103
Fig. 4.6 12 layer PCB structure with two cores and its MMF................................................................ 104
Fig. 4.7 LLC Resonant Converter with Matrix Transformer Structure .................................................. 105
Fig. 4.8 Four layer PCB structure with four cores and its MMF ........................................................... 106
Fig. 4.9 Primary side winding pattern with four U-I cores matrix transformer ...................................... 106
Fig. 4.10 Rearranged primary side winding pattern with four U-I cores ................................................ 107
Fig. 4.11 Primary side winding pattern for two E-I core matrix transformer ......................................... 107
Fig. 4.12 Primary side winding pattern for two core matrix transformer with flux cancellation ............. 108
Fig. 4.13 LLC resonant converter with proposed transformer structure................................................. 108
xv
Fig. 4.14 Top view of 1st winding arrangement ................................................................................... 110
Fig. 4.15 Cross view of the first winding arrangement and its MMF .................................................... 111
Fig. 4.16 AC current distribution for secondary winding of the first winding arrangement.................... 111
Fig. 4.17 FEA simulation for the first winding arrangement ................................................................. 112
Fig. 4.18 FEA simulation for the first winding arrangement with output shorted .................................. 113
Fig. 4.19 FEA simulation for the first winding arrangement with secondary winding shorted ............... 114
Fig. 4.20 Top view of the second winding arrangement........................................................................ 115
Fig. 4.21 Cross view of the second winding arrangement ..................................................................... 115
Fig. 4.22 FEA simulation for the second winding arrangement ............................................................ 116
Fig. 4.23 Loss density vs. Cross area. .................................................................................................. 118
Fig. 4.24 Core and winding dimensions ............................................................................................... 118
Fig. 4.25 AC & DC resistance of Sec winding. .................................................................................... 119
Fig. 4.26 AC & DC resistance of Pri winding. ..................................................................................... 119
Fig. 4.27 The circuit of 1kW, 400V/12V LLC resonant converter prototype ......................................... 120
Fig. 4.28 The prototype of 1kW, 400V/12V LLC resonant converter ................................................... 121
Fig. 4.29 Experimental waveforms for 15% load condition .................................................................. 122
Fig. 4.30 Experimental waveforms for 50% load condition .................................................................. 122
Fig. 4.31 Experimental waveforms for 100% load condition ................................................................ 123
Fig. 4.32 The efficiency comparison for CPES prototype and industry product .................................... 124
Fig. 5.1 LLC resonant converter .......................................................................................................... 128
Fig. 5.2 Circuit and voltage gains of LLC resonant converter ............................................................... 129
Fig. 5.3 One example of selected 4-elements resonant converters......................................................... 129
Fig. 5.5 V-I stresses of LLC and 4-element resonant converters ........................................................... 131
Fig. 5.6 One example of 5-element resonant converters ....................................................................... 132
xvi
Fig. 5.8 Cantilever model for LLC transformer .................................................................................... 134
Fig. 5.9 Cantilever model for proposed transformer ............................................................................. 135
Fig. 5.10 Simplified model for proposed transformer ........................................................................... 135
Fig. 5.11 Equivalent circuit for discrete Cr, Cp and integrated magnetics .............................................. 137
Fig. 5.12 The gain curves of passive integration model ........................................................................ 137
Fig. 5.13 The realization of magnetic integration module ..................................................................... 138
Fig. 5.14 The transformation of equivalent circuit ................................................................................ 139
Fig. 5.15 Volume comparison of proposed and traditional magnetic integration method ....................... 139
Fig. 5.16 Stress comparison for integrated magnetic module ................................................................ 140
Fig. 5.17 LC cell structure and equivalent circuit ................................................................................. 141
Fig. 5.18 Interconnection of LC cells ................................................................................................... 142
Fig. 5.19 Interconnections and winding structure of an integrated LLCT module ................................. 143
Fig. 5.20 Exploded view of an LLCT module with the stacked structure .............................................. 144
Fig. 5.21 Winding model for 4-winding transformer ............................................................................ 145
Fig. 5.22 The passive integration model ............................................................................................... 146
Fig. 5.23 Transformer structure for integration..................................................................................... 147
Fig. 5.24 Transformer model of proposed structure .............................................................................. 148
Fig. 5.25 Equivalent circuits for 6 cases ............................................................................................... 149
Fig. 5.26 Calculated results of integrated transformer module .............................................................. 150
Fig. 5.27 The final parameters of integrated transformer module .......................................................... 151
Fig. 5.28 The volume comparison of passive module and discrete components .................................... 151
Fig. 5.29 Integrated passive module with heat extractors ...................................................................... 153
Fig. 5.30 Diagram of heat paths on cross section .................................................................................. 154
Fig. 5.31 Core dimensions ................................................................................................................... 155
xvii
Fig. 5.32 Picture of cores and heat extractors ....................................................................................... 155
Fig. 5.33 The volume comparison of passive module and discrete components .................................... 156
Fig. 5.34 Experimental results for five-element resonant converter with passive integration module..... 157
xviii
List of Tables
Table 4.1The AC resistance of pri and sec windings vs. thickness ........................................................ 120
Table 4.2 Loss breakdown of prototype under 100% load .................................................................... 124
Table 4.3 Loss breakdown of prototype under 15% load ...................................................................... 125
xix
Daocheng Huang Chapter 1
Chapter 1. Introduction
This chapter presents the motivations, objectives and overview of this dissertation. The
advantages of LLC resonant converters and its wide applications are investigated. The challenges
in LLC resonant converter are described. A review in this field is provided, followed by the
dissertation outline and the scope of research.
1.1 Background and Motivation
Over the past decade energy efficiency and power density have become the top concerns for
power conversions. Rising energy intensity leads to a higher cost for delivering power.
Meanwhile, the demand for compact power supplies grows significantly. It requires power
supplies with high efficiency, low profile and high power density.
With the explosive development of information technology, the communication and
computing systems, such as data centers, telecoms, servers, desktops, laptops and notebooks
have become a large market for the power supply industry. For the advance of the integrated
circuit technology, the density and functionality of computer systems are continually increasing,
which requires more efficient power delivery while smaller size on power suppliers [A.1]-
[A.11]. For consumer electronics, such as flat-panel TVs and LED lighting, high reliability, low
cost and high efficiency power supplies are demanded [A.12]-[A.15]. More ever, for the
prosperousness of the cloud computing and online service, energy saving and cost reduction
become major concerns.
1
Fig. 1.1 Ac-dc front-end converter
Ac-dc front-end converters as shown in Fig. 1.1 are widely adopted by these applications.
Driven strongly by economic and environmental concerns, high efficiency is becoming more and
more desired for ac-dc front-end converters. The 80 PLUS programs provide a class of efficiency
requirements as shown in Fig. 1.2[A.9]. Take the 80 PLUS TITANIUM program as an example,
it provides the highest efficiency requirement for the front-end converters nowadays. The
certified power supplies are expected to achieve 91% efficiency at 100% load, 96% efficiency at
50% load, 94% efficiency at 20% load, and 90% efficiency at 10% load when converting 230V
ac power from electric utilities to dc power used in most electronics. Right now, some industry
benchmarks are beyond TITANIUM. The efficiency requirement at the dc-dc stage is even
higher to meet the overall efficiency from ac input to the load.
In addition to the efficiency requirement, the power density of front-end converters is
continuously increasing as well [A.5]. Higher power density will eventually reduce the converter
cost and allow for accommodating more equipment in the existing infrastructures. As shown in
Fig. 1.3, in the past years, the power density of front-end commercial converters has increased by
more than twenty times. This will continuously increase in the future because of the fast
developments in new devices, magnetic materials and so on [A.16]-[A.19].
2
Fig. 1.2 Efficiency trend for ac-dc front-end converters of redundant, datacenter applications
Fig. 1.3 Power density trend for ac-dc front-end converters
3
1.2 Issues of PWM DC-DC Power Conversions
1.2.1 Pros and Cons of Soft-switching PWM DC-DC Converter
Hard switching pulse-width-modulation (PWM) converters are common in power supplies,
however, they suffer high switching losses. Soft switching techniques can help the PWM circuit
achieve zero-voltage-switching (ZVS) so that lower switching loss and higher frequency can be
accomplished. Soft-switching PWM circuits, such as the phase-shift full-bridge PWM converter
(PSFB)[A.20]-[A.23]and the asymmetrical half-bridge PWM converter (AHB)[A.24]-[A.28], are
widely used for front-end dc-dc conversion. These topologies are plotted in Fig. 1.4, respectively.
(a) Phase shift full bridge PWM converter (b) Asymmetry half bridge PWM converter
Fig. 1.4 Typical Soft-switching PWM converters
ZVS lost at light load is one drawback of soft-switching PWM converters. For PSFB, large
leakage inductance is needed to achieve ZVS. With large leakage inductance, the duty cycle loss
due to charge and discharge leakage inductance will be significant. This will limit the choice of
transformer turns ratio, which will affect the performance of whole converter. Even with large
leakage inductance, still ZVS cannot be achieved at light load. There are many different methods
4
to expend ZVS region for phase shift full bridge[A.29], but they are not been widely adopted due
to complexity. Since AHB also utilizes leakage inductance to achieve soft switching, there is
similar problem as discussed for PSFB, which lost ZVS during light load condition.
High turnoff loss for primary devices is another drawback. As dash circled in Fig. 1.5, the
turnoff current of primary devices is around the highest point due to the trapezoid or square
shape primary current.
No ZCS for synchronous rectifier (SR) generates large reverse recovery loss. As circled in
Fig. 1.5, the turnoff current of SR device is around the highest point due to the trapezoid or
square shape current.
(a) Phase shift full bridge PWM converter (b) Asymmetry half bridge PWM converter
Fig. 1.5 Waveforms of typical Soft-switching PWM converters
5
Passive components are hard to reduce volume. As mentioned above, the switching loss of
soft-switching PWM converters is not small, which limit the switching frequency increasing.
That means passive components are not easy to reduce volume. On the other hand, there are
usually two individual magnetic components in the circuit, the transformer and the output
inductor. Sometimes, an extra inductor is in series with transformer to help achieving ZVS. It is
very challenge to integrate them all to shrink the volume.
1.2.2 Limitation of PWM DC-DC Converter with Hold-up Time Requirement
The holdup time operation poses a major design challenge in front-end converters. As
shown in Fig. 1.6, the front-end ac-dc converters are required to maintain the regulated output
voltage for about 20ms when the input ac line is lost[A.30].As shown in Fig. 1.1, during the
holdup time, all the energy transported to the load comes from the bulk capacitor. The
requirement of this capacitor is determined by the system power level and the input voltage
range of the dc-dc converter. The relationship between holdup time capacitor requirement and
minimum dc-dc stage input voltage for 1kW front-end converter is shown in Fig. 1.7.Apparently,
the wider the dc-dc stage input voltage range is, the fewer bulky holdup time capacitors are
required, and the higher power density will be. Therefore, to meet the holdup time operation
requirement, the capability of operating within a wide input voltage range is required for the dc-
dc stage in front-end converters.
6
Holdup time
VNORM
VBUS_MIN
20ms
Fig. 1.6 Holdup time operation of ac-dc front-end converters
Fig. 1.7 Holdup time capacitance vs. input voltage range for dc-dc converters of the front-end converter
However, conventional PWM converters have to sacrifice normal operation efficiency to
extend their operation range. It is difficult to design a wide-operation-range PWM converter with
high efficiency. Normally, the duty cycle is designed to be as high as possible to handle the
holdup time operation, shown in Fig. 1.8. Thus, at normal conditions, the duty cycle is much
smaller. The primary side to secondary side transformer turns ratio is small, which leads to high
7
primary side current, as equation (1.1) and (1.2) indicated. Both conduction loss and switching
loss increase. Consequently, efficiency suffers at normal operation conditions.
Fig. 1.8 Duty cycle range for PWM converter with hold up requirement.
np Vin _ min
N = = D max (Full bridge) (1.1)
ns Vo
np Vin _ min
N = = D max (Half bridge) (1.2)
ns 2V o
1.3 Opportunity of Resonant DC-DC Converters
1.3.1 Basic Concepts of PWM Converter and Resonant Converters
Conceptually, PWM converters control the power flow by controlling the duty circle of
switches. When switch is on, the power flows from the source to the load. When switch is off,
8
the power flow is blocked. The output filter of PWM converters is for switching frequency and
higher order harmonics, as shown in Fig. 1.9.
Fig. 1.9 Simplified general isolated dc/dc PWM converter block diagram
Different with PWM converter, resonant converters have quite different way to control
power flow. In general, resonant converters can be presented in a simplified block diagram form
as shown inFig. 1.10. The input can be either voltage or current source. The output sink is
implemented by rectifier circuits. The output circuit consists of either a series inductor or a
parallel capacitor with the load resistance in order to give current sink or voltage sink,
respectively. The basic resonant cell consists of frequency selective network whose function is to
transfer the energy from the source to the load. In other words, it works like a band-pass filter.
Fig. 1.10 Simplified general isolated dc/dc resonant converter block diagram
As shown in Fig. 1.11, resonant converter controls the power flow by adjusting the switching
frequency. When the switching frequency is around the resonant frequency, most of the input
9
power pass through the band-pass filter. When the switching frequency deviates from the
resonant frequency, the input power is attenuated by the filter, thus, only part of it flow to the
load. This mechanism is quite different from PWM converter.
Fig. 1.11 Band-pass filter concept for resonant converters
A good band-pass filter is desirable for efficient energy delivery. Here, good band-pass filter
means the least circulating energy in the converter. Nevertheless, which resonant cell contributes
good band-pass filter?
a. Series Resonant Converter (SRC)
Series resonant converter (SRC) is one of the basic resonant converters[A.31]-[A.36]. The
schematic of series resonant converter is Fig. 1.12. The voltage gain curves of the SRC is shown
in Fig. 1.13. The benefits are ZVS achievable above resonant frequency, ZCS achievable below
10
resonant frequency. The operating region is on the right side of resonant frequency owing to
preferred ZVS operation.
Fig. 1.12 Series resonant converter
Fig. 1.13 Gain curves of series resonant converter
Series resonant frequency is most efficient point. From the gain curves, we can see the gain
equals one at resonant frequency. The total energy from the source is delivered to the load.
However, the resonant frequency is the boundary of ZVS and ZCS region. It is very risky to
operate at this point. Due to the holdup time requirement, the operation point is far away from
resonant point at nominal input voltage (Vin=400V), and the voltage gain is much smaller than
11
one. Meanwhile, the turn-off current increases significantly due to large phase shift between
excitation voltage and current of the resonant tank. The switching loss will be high.
Light load regulation is bad for SRC. It can be seen from the operating region that at light
load, the switching frequency needs to increase to a very high value to keep the output voltage
regulated due to the flat voltage gain beyond resonant frequency.
SRC is not a good candidate for front end DC-DC converter. The major problems are: light
load regulation, high circulating energy and turn off current at nominal input voltage condition.
b. Parallel Resonant Converter (PRC)
Parallel resonant converter is another basic resonant converter[A.37]-[A.40]. The schematic
of parallel resonant converter is Fig. 1.14. The voltage gain curves of the SRC is shown in Fig.
1.15. The operating region of the PRC is shown inFig. 1.15as shaded area. Similar to SRC, the
operating region is also designed on the right hand side of resonant frequency to achieve ZVS.
Fig. 1.14 Parallel resonant converter
Compared with SRC, the voltage gain of PRC can be larger than one. That means PRC can
boost voltage. Meanwhile, the operating region of PRC is much narrower. At light load, the
frequency doesn't need to change too much to keep output voltage regulated.
12
Fig. 1.15 Gain curves of parallel resonant converter
However, the circulating energy is very high for PRC, even at light load. For PRC, since
the load is in parallel with the resonant capacitor, even at a no-load condition the input still sees a
pretty small impedance of the series resonant tank. This will induce pretty high circulating
energy, even when the load is zero.
The major problems of PRC are: high circulating energy, high turn off current at high input
voltage condition.
In sum, to deal with a wide input voltage range, all these traditional resonant converters
encounter some problems. To achieve higher efficiency, other resonant topologies should be
considered.
1.3.2 Advantages of LLC Resonant Converter (LLC) with Hold-up Time Requirement
LLC resonant converters (LLC) is drawn a lot of attention since 1990 [A.41]-[A.48].The
half bridge LLC resonant converter is illustrated in Fig. 1.16. The voltage gain of the LLC
13
resonant converter is drawn in Fig. 1.17. When switching frequency lower than resonant
frequency (fs<fo), the voltage gain is more like parallel resonant converter. When switching
frequency higher than resonant frequency (fs>fo), the voltage gain is more like series resonant
converter. With variable frequency control, the voltage gain of LLC resonant converter can be
controlled as boost mode or buck mode. Based on the combination of two resonant frequencies,
LLC resonant converter has more advantages than previous introduced converters.
Fig. 1.16 LLC resonant converter
Fig. 1.17Voltage gain of LLC resonant converter
14
ZVS is achievable from no load to full load. With the help of parallel resonant frequency,
LLC can operate around series resonant frequency without the risk falling into ZCS region. In
addition, voltage gains of different Q converge at this point. Thus, the LLC would not change
frequency too much as load changes. The voltage second on Lm, which determines the turnoff
current, is relatively constant. And this current independent of load helps primary devices charge
and discharge junction capacitance to achieve ZVS. In other words, ZVS is independent of load.
Low turnoff loss is achievable for primary devices due to low turnoff current. The turnoff
current could be controlled by Lm to be as low as just fulfill ZVS requirement.
Fig. 1.18 Waveforms of LLC resonant converter at resonant frequency
ZCS for synchronous rectifier (SR) is naturally achieved if it operates at or below series
resonant frequency. When switching period equals the series resonant period, the primary current
(iLr) resonances to touch magnetizing current (iLm). Meanwhile, secondary side current reaches
zero. When switching period becomes larger, the discontinuous period of secondary current
extends.
15
Magnetic integration is easily achieved for LLC. Because of soft-switching for primary and
secondary devices, LLC can increase switching frequency to push power density. On the other
hand, high switching frequency introduces small magnetic component, which makes them easy
integration. For LLC, Lr could be the leakage inductance of transformer, while Lm is the
magnetizing inductance, as shown in Fig. 1.16. Naturally, LLC has the potential to integrate all
the magnetic components into transformer, which continue push power density.
The holdup time extension capability is accomplished without sacrificing the efficiency
under the nominal operating condition. As mentioned before, series resonant frequency is the
optimal operation point for LLC. By reducing the LLC switching frequency, high voltage gain is
achievable, which is desired for the holdup time operation. Therefore, the LLC resonant
converter has been a very popular topology for the front-end power supplies in computing
systems and consumer electronics systems.
Soft-switching diminish most of switching loss, thus, LLC can reach high efficiency and be
easy to push high frequency. High frequency and easy integration structure give LLC the
potential to reach high power density. Fig. 1.19 shows the volume and efficiency comparison of
two commercial front-end products. It is obvious that the front-end converter with LLC has
better power density and efficiency than the one with PWM converter.
For the benefits of LLC resonant converter as mentioned above, firstly it is adapted by flat-
panel display. Then, telecom quickly accepts it as main stream solution. Right now, LLC
gradually takes the markets of LED lighting, server and datacenter, etc.
16
Fig. 1.19 Commercial products comparison with LLC DC/DC and PWM DC/DC
Fig. 1.20 Market trend for LLC resonant Converter
17
1.4 Challenges of LLC Resonant Converter
However, there are many issues for the LLC resonant converter to deal with, like start-up,
short-circuit protection, SR driving method, EMI noise reduction and magnetic design, etc.
1.4.1 Start-Up and Short Circuit Protection
The first issue is start-up process. Although LLC has parallel resonant frequency to help
achieve holdup requirement, the voltage gain above resonant frequency is very flat due to the
characteristics of series resonant frequency, as shown in Fig. 1.21.
Fig. 1.21 LLC soft start-up in above resonant frequency region
During the soft start-up process, the switching frequency needs to be pushed to a high value
(fs>f0), and then decreased step by step, in order to minimize the voltage and current stress in the
resonant tank, meanwhile building up the output voltage[A.49]-[A.55].
18
If the start-up frequency (fmax) is not high enough, there will be large voltage and current
stress in the resonant tank at the start-up moment. Fig. 1.22shows the experimental test of [A.49]
where the voltage and current stress is large in the resonant tank if starting-up with 1.6 times of
resonant frequency.
Fig. 1.22 LLC soft starts process by commercial IC chip
Thus, start-up is an inherent issue for LLC resonant converter.
1.4.2 Synchronous Rectifiers (SR) Driving Scheme
For isolated dc/dc converter, synchronous rectifier is necessary to improve the efficiency of
DC/DC converter. The SR driving signals for PWM dc/dc converter are very easy to get from
primary side device signal. Take phase shift full bridge converter as example[A.56], the circuit is
19
shown in Fig. 1.23. The driving signal of SR2 is the same as SA, and the driving signal of SR1 is
the same as SB.
Fig. 1.23 PSFB circuit and devices’ driving signals
LLC resonant converter is quite different with PWM converter for SR driving scheme. The
driving signal of primary side and secondary side devices are not in phase, as shown in Fig. 1.24.
f S<f 0 f S=f 0 f S>f 0

VgsQ1
VgsQ2
iLr
iLm
iSEC iSR1 iSR2 iSR1 iSR2 iSR1 iSR2
VgsS1
VgsS2
phase difference phase difference
Fig. 1.24 Desired SR gate driving signals in different switching frequency regions
20
When fs<f0, the SR should be turned off earlier than the main switch. Otherwise, the SR
would conduct circulating energy, namely a reverse current from the load to the source, thus
causing a greater increase in the RMS currents and turn-off current, and causing efficiency to
deteriorate dramatically. When fs>f0, the SR should be turned off a bit later than the main switch.
Otherwise, the sharply decreased current would go through the paralleled body diode, resulting
in a serious reverse recovery. Therefore, the driving signals of SRs and main switches cannot
derive from the same PWM signal for control of the conduction times as those in the isolated
PWM converters.
One solution [A.57] is based on the SR current iSRsensing by transformers to generate SR
driving signal. The method is precise, but due to the large current on the secondary side, it
requires a large size current transformer and it presents a lower efficiency due to the extra
resistance of the transformer windings.
Fig. 1.25 SR gate driving signals by secondary current sensing
An alternative solution [A.58] is sensing current through the transformer’s primary side
winding. The primary side current is a precise replica of the secondary side current. Although a
21
smaller loss could be achieved when compared to the secondary side current sensing, three
magnetic components are needed, losing the integration of leakage, magnetizing inductors and
transformer in a single element.
Fig. 1.26 SR gate driving signals by primary current sensing
d’ iSR
Vth_off
Lpackage ideal Vds
sensed Vd’s’
Vth_on
SR duty loss
VgsSR Actual ton
s’
Fig. 1.27 The influence of package inductance
A promising driving method is based on the SR drain to source voltage. The sensed SR Vds
is processed by the control circuits as [A.59]. However, the accuracy of this driving scheme is
highly affected by the SR package [A.60]. Due to the inevitable package inductance, the sensed
22
terminal drain to source voltage of the SR is actually the sum of the MOSFET’s on status
resistive voltage drop and the package inductive voltage drop. Fig. 1.27shows that the sensed Vd's'
of the SR terminal deviates greatly from the purely resistive voltage drop Vds of the MOSFET.
Thus, the actual SR drive signal VgsSR is significantly shorter than the expected value.
Therefore, an accurate SR driving scheme under high frequency is required, which can
control the SR gate driving signal in the whole switching frequency regions with simple
implementation. This method should not increase the volume of whole system too.
1.4.3 Chanllenges for High Output CurrentApplication
Isolated high output current dc/dc converter may suffer very high conduction loss at
secondary side even synchronous rectifier is applied [A.52] [A.53].
High conduction loss of secondary windings and SR devices is the first issue for low-voltage
high-current applications. To deal with this issue, a paralleled winding structure with paralleled
SR devices is a common practice. However, this structure is not good enough. The conventional
circuit schematic of the transformer and SR structure is shown in Fig. 1.28. The conventional
planar transformer and the secondary side structure are depicted in Fig. 1.29.
The current sharing among secondary windings and SR devices becomes an issue for this
parallel structure. The current distribution through secondary side is determined by many aspects,
including the resistance and leakage inductance of windings, the package inductance and
resistance of SR devices, the trace and connection inductance and resistance, etc. Thus, it is hard
to keep even current distribution by traditional transformer structure.
23
Fig. 1.28 Circuit schematic of the conventional transformer and SR structure.
Fig. 1.29 Conventional transformer and the secondary side structure.
High termination losses is the another issue. Due to the proximity effect and skin effect, the
currents in adjacent terminals with opposite directions attract each other. Thus, very high losses
and hot spots are generated. This termination loss deteriorates efficiency significantly. For the
conventional design, all SR devices are placed on the motherboard. Physically speaking, a large
loop of the secondary side rectifiers is inevitable. Consequently, a large distribution loss will be
24
generated. In addition, for large number of SR devices, it is extremely difficult toachieve a
symmetrical layout for each SR. Hence, current-sharing of the SR is a severe problem.
Thus, how to improve the efficiency and power density of high step-down dc/dc converter is
a key issue.
1.4.4 EMI Performance of LLC Resonant Converter
In EMC standards, such as FCC part 15 in the United States and EN55022, CISPR 22 in
Europe, the conducted EMI noise emissions of devices are limited. It is required to meet the EMI
standard for front-end converters.
Fig. 1.30 CM noise spectrum comparison of PWM and LLC resonant converters.
Because of the bulk capacitor between PFC and dc-dc converter, only common mode noise
of dc-dc converter will influence the grid and EMI filter. The measured common mode EMI
noise spectrum of LLC resonant converters is shown in Fig. 1.30. Compare with PWM converter,
it is a stringent challenge to reduce the EMI noise due to the EMI standard requirement [A.61].
25
1.5 Multi-Element Resonant Converters
Since LLC resonant converter has so many issues, how to deal those issues but still keep the
merits of LLC is a challenge. Some issues come from the fundamental characteristics of LLC
resonant converter. How to overcome them remains a problem. As mentioned in [A.62]-[A.64],
there are more than hundreds multi-element resonant topologies. Is there any resonant topology
similar or better than LLC resonant converter?
Based on the previous sections, the merits of LLC come from the combination of two
resonant frequencies. Thus, we want to keep the characteristics below series resonant frequency
of LLC gain curves in Fig. 1.17. However, we do want to change the gain curves of LLC resonant
converter above resonant frequency.
For the start-up and short-circuit protection issue, the fundamental reason is the flat
voltage gain above resonant frequency. To deal with this issue, the resonant topology has sharp
gain above resonant frequency is desired as presented in Fig. 1.31[A.64].
Fig. 1.31 Desired voltage gain for LLC start-up.
26
For the SR driving scheme, the resonant topology, which can simply utilize primary side
current information to drive SR without suffering the benefit of integration, is desirable.
Low EMI noise by nature is also desired for resonant topology.
The question is if there is any resonant topology having those benefits among the ocean of
resonant topologies. How to find it and how to compare them with LLC is very challenging.
1.6 Opportunity of Emerging Technologies
The exponential rise in “Big Data” generation, processing and storage highlights the
growing demand for datacenter and cloud computing power worldwide. The subsequent task is
to maximize energy efficiency, thus saving money and natural energy resource, minimizing
pollution and meeting the US Department of Energy’s “Exascale” challenge.
Fig. 1.32 Traditional AC distribution system for data center
27
For the power semiconductor industry, the challenge is to provide efficient and high quality
power conversion from 480V 3-phase AC power entering the datacenter, all the way to 1V,
100A+ processors. Fig. 1.32 shows the typical AC distribution system structure for data center.
1.6.1 New System Architectures
The ever-present trend for datacenter distributed power systems is the pursuit of higher
efficiency and power density. Existing solutions within the AC-DC conversion-system topology
are struggling to provide even a fewpercentage points of large-scale improvement despite
localized improvements in performance.
Fig. 1.33 380V DC distribution system for data center
The answer may be to look instead to a very different AC-DC structure, based on new
approaches rather than merely incremental ones. Using high-voltage DC for power transmission,
in conjunction with new conversion approaches, offers tangible and significant benefits for both
sourcing options and system end-to-end performance. Recently, Intel proposed a 400 Vdc direct-
28
distribution bus for data center applications in order to reduce the number of power conversion
stages and the power loss in the power distribution path, as shown in Fig. 1.33. By eliminating an
AC uninterruptible power supply (UPS) and a transformer in the power distribution unit (PDU),
the overall efficiency can be improved. In fact, some work from France Telecom and China
Mobile estimate that between 8% and 10% across the board can be saved by going to DC
distribution[A.7]-[A.8], [A.65]-[A.67].
Continue pushing whole system efficiency, one effective way is to reduce the regulated
DC/DC stage. For new DC distribution system, the holdup requirement is not valid due to the
battery connecting to 380V bus. IBM supercomputer utilizes the DC system illustrated in Fig.
1.34. The efficiency of each module is also given in each box. There is only one regulation stage:
PRM (Pre-Regulated Module) regulator. Bus converter and VTM (Voltage Transformer Module)
are both unregulated DC/DC converter. The efficiency of the power train can be as high as 87%.
Fig. 1.34 DC power-train for IBM Supercomputer
Another possible power train structure is shown in Fig. 1.35. The DC/DC converter from
400V to 12V only needs to deal with small ripple on the bus, which can be easily deal with next
regulated stage (VR). It is obvious that two regulated stages are less efficient than one regulated
stage with one unregulated stage.
29
Fig. 1.35 Improved DC power-train for 400V to CPU
Thus, how to improve the efficiency and power density of high ratio step-down DCX is one
of the key issues for the whole performance of datacenter.
1.6.2 Benefits and Chanllenges of LLC-DCX
Compared with regulated LLC-D2D, LLC-DCX has better performance and less issues
especially at high switching frequency.
Firstly, its switching frequency is fixed exactly at the resonant frequency point (fs=f0) to
achieve the best efficiency[A.68]. Meanwhile, the SR current is in phase with primary side
driving signal just like PWM converter. It is very easy to achieve SR driving by utilizing primary
side driving signal. Due to new system architecture, LLC-DCX do not need to worry about start-
up problem, the regulator before or after LLC-DCX can help.
Thus, start-up, short-circuit protection and SR driving are not issues for LLC-DCX
anymore.
30
Fig. 1.36 LLC resonant converter as DCX
Usually, large Lr is designed for LLC-D2D to help achieve holdup time and regulation
requirement, while LLC-DCX does not have that. Mostly large Lr needs extra inductor in series
with leakage inductance of transformer, which increases the volume and loss of whole system.
For LLC-DCX, Lr is only the leakage inductance of transformer, and can be as small as possible.
LLC-DCX has potential to be high efficiency and high power density. For high efficiency
and easy magnetic integration, LLC-DCX could operate at high switching frequency. Meanwhile,
the high frequency transformer design and EMI are still issues for LLC-DCX.
1.6.3 Wide Band Gap Devices —GaN HEMT
High performance active devices have been the first force to increase power density to meet
the requirements of modern systems including datacenter. Silicon devices have dominated for
power management since the late 1950s. However, due to continuous device optimization and
31
improvements in the production process, material properties of Si are increasingly becoming the
limiting factor.
Fig. 1.37 Specific on-resistance versus breakdown voltage for GaN and Si
In recent years, the GaN HEMT has emerged as a promising device for high frequency, high
efficiency, high density power conversion due to a better figure of merit than comparable Si and
SiC transistors [A.69]-[A.71]. A progressively larger number of GaN devices have been
manufactured for in field applications ranging from low power voltage regulators to high power
infrastructure base-stations[A.72]-[A.80]. The switching frequency has been continuously
pushed up to several MHz to both reduce passive components size and increase power density.
1.7 Dissertation Outline
Taking into account the challenges raised during the review of the existing literature on the
LLC resonant converters, several issues have been addressed in this dissertation.
32
Chapter 1:
Research background, motivation, and literature review.
The advantages of LLC resonant converter over PWM converter and traditional resonant
converters are presented, especially under hold up time requirement. However, LLC has its own
issues, like startup, short-circuit protection, SR driving, transformer design and EMI
performance. Multi-element resonant converters are one possible way to deal with them.
Meanwhile, with the new system architecture of datacenter, there is opportunity to use
unregulated LLC, which has less issues and better performance than regulated LLC resonant
converter. This new structure and emerging GaN HEMT promote the requirement of high
frequency high efficiency transformer design. To be more advanced, passive integration for
multi-element resonant converter is also achievable.
Chapter 2:
One effective way to solve the fundamental issues of LLC resonant converter is to find
alternative resonant converters.
The resonant tank acts like band-pass filter in resonant converters. The four type basic
resonant cells are studied. A synthesis of resonant topology based on basic resonant cells is
proposed. Every two basic resonant cells are combined to form a new resonant cell. Three three-
element LLC-like resonant converters are discovered.
For four-element resonant converters, a notch filter is introduced to LLC-like resonant
converters helping limit the inrush current during startup or short-circuit protection period.Eight
LLC-like resonant converters with notch filter are discovered based on previous synthesis.
33
To enhance power delivery efficiency, an extra element is added to four-element resonant
tank for 3rd harmonics band-pass filter. Twenty five-element resonant converters with 3rd
harmonics band-pass filter are gotten.
Chapter 3:
Lots of LLC-like resonant converters are discovered based on the method in chapter 2. How
to evaluate each topology still remains an issue. To find out if new resonant topologies have
similar or better performance than LLC, an evaluation system is necessary.
To utilize state-plane for evaluation, traditional normalization factors are not good enough.
The new normalization factors are proposed based on the apparent power of converter. Utilizing
the state-plane with new normalization factors, the apparent power, voltage and current stresses
of each multi-element resonant converter can be easily compared.
Current limit capability, SR driving scheme, magnetic integration and EMI noise are very
important to resonant converters. All of them are taken into account to evaluate the performance
of resonant converters.
Based on the pros and cons of each LLC-like resonant converter, several possible
application examples are proposed as reference.
Chapter 4:
To deal the high output current conduction loss, the resonant topology can hardly help. The
circuit topology can help. To alleviate the heavy output current, the matrix transformer instead of
the traditional transformer is proposed. Matrix transformer can help split transformer secondary
current to reduce conduction loss. With the matrix transformer winding structure, series at
34
primary side and parallel at secondary side, the current sharing problem among windings and SR
devices can be solved.
The price is the multiple cores of matrix transformer. With the concept of flux cancellation,
the number of core reduces into half. In other words, the core loss and core volume are greatly
reduced without suffering conduction loss.
Meanwhile, a system method is proposed to integrate transformer winding with SR devices
to eliminate termination loss. After that, the tradeoff of termination loss and interleaving is
discussed. Then, the optimal winding structure is figured out.
Finally, a detail design procedure for transformer dimensions is given.
Chapter 5:
To be more advance integration, magnetic integration method for multi-element resonant
converter is proposed based on transformer cantilever model. A measuring method is given to
identify the parasitic inductance value in complex cantilever model.
For better power density, passive integration is necessary. However, capacitance integration
is risky due to life time constraint. An evaluation criteria is proposed to determine which
components integrated in the system based on lowest voltage and current stress.
A passive integration method is given to integrate all the magnetic components and
capacitor with lowest stresses.
Chapter 6:
Conclusion and future work.
35
Chapter 2. Synthesis of Multi-Element Resonant
Converters
In this chapter, a general and systematic method to explore the characteristics of multi-
element resonant converters is proposed. Each resonant converter is synthesized by basic
resonant cells. The resonant converter inherits the characteristics of parental resonant cells. LLC
resonant converter is one of the most popular topologies in DC/DC converters. The LLC-like
resonant converters are found by the synthesis of basic resonant cells. Furthermore, the LLC-like
resonant converters are improved by synthesizing more resonant cells.
2.1Fundamental Concept of Resonant Converters
As mentioned in chapter one, dc-dc resonant converters can be presented in a simplified
block diagram form as shown in Fig. 2.1.
The input DC voltage or DC current source is converted into square waveform by primary
switches. The output sink is implemented by rectifier circuits. The output circuit consists of
either a series inductor or a parallel capacitor with the load resistance in order to give current
sink or voltage sink, respectively.
The basic resonant cell consists of frequency selective network whose function is to transfer
the energy from the source to the load, in other words, band-pass filter. The passing-through
energy is controlled by changing the switching frequency. The band-pass filter characteristic of
each resonant tank could be represented by its voltage transfer functions [B.1], [B.2].
36
Fig. 2.1Simple Block diagram for the basic concept of resonant converter
Most of dc-dc applications are voltage source, like front-end dc-dc converters. Thus, the
voltage source is taken into account in the following analysis. The similar concept can also be
applied to current source application.
Fig. 2.2Zero current switching and zero voltage switching principles
37
When the operation point at resonant frequency (fo, in Fig. 2.1), the input voltage (Vac) is in
phase with input current (iin). The output power reaches its maximum value if input power is
fixed. All of the input power is directly delivered to the load. When operation below fo, the
voltage is leading the current, so the tank is capacitive. The up slop of voltage gain indicate ZCS
operation, as shown in Fig. 2.2.When operation above fo, the current is leading the voltage, so the
tank is inductive. The down slop of voltage gain indicates ZVS operation.
2.1.1 Basic Resonant Cells (Two-Element Resonant Tanks)
Two-element resonant tanks are basic resonant cells of multi-element resonant converters.
For two energy storage elements, there are totally eight resonant tank configurations as shown in
Fig. 2.3. Tank A, B, C and D are suitable for voltage source, and Tank E, F, G and H are suitable
for current source. Here, not “suitable” means no regulation capability (like Tank E to voltage
source), and/or no resonance (like Tank F to voltage source).
Fig. 2.3 Two-element resonant tanks
38
Tank A is series resonant tank, which is the resonant tank of series resonant converter (SRC).
It behaves like band-pass filter. The resonant frequency in Fig. 2.4 (a) is named series resonant
frequency (SRF). When the frequency of input energy is the same as SRF, the whole energy will
pass through it, while the energy of other frequencies is attenuated. However, this band-pass
filter is not good at light load. The voltage gain becomes flat. Hence, a very high switching
frequency is required for series resonant converter light load regulation.
The resonant frequency in Fig. 2.4 (b) is named parallel resonant frequency 1 (PRF1). It is
not good band-pass filter either. The high frequency input energy will pass through it, but low
frequency energy will be blocked. Thus, it performs more like high-pass filter.
The resonant frequency in Fig. 2.4 (c) is named parallel resonant frequency 2 (PRF2). It is
the resonant tank of parallel resonant converter (PRC). It is not good band-pass filter either. The
low frequency input energy will pass through it, but high frequency energy will be blocked. Thus,
tank C performs more like low-pass filter.
Both of two PRFs have high gains at light load around resonant frequency. Steep slope close
to resonant frequency come with this.
The resonant frequency in Fig. 2.4 (d) is named notch resonant frequency (NRF). This tank
is quite different from previous tanks. It behaves like a band-stop filter. When the frequency of
input energy is the same as NRF, the tank impedance is infinite high, the whole energy will be
blocked. Meanwhile, the energy of high order harmonics pass through the whole tank. Thus,
Tank D alone can hardly form a resonant converter.
All of basic resonant cells are only have one resonant frequency. The resonant frequency is
the most efficient point for the resonant tank except band-stop filter, because all the fundamental
39
of input power is delivered to the load. Otherwise, part of it is attenuated by the tank. However,
they cannot operate at resonant frequency in fear of falling into different soft-switching region,
like from ZVS into ZCS or from ZCS into ZCS, which is detrimental to devices.
Fig. 2.4 Voltage gains of two-element resonant tanks
40
2.1.2 Synthesis of Basic Resonant Cells
From the discussion above, two-elements resonant converters have their inherent limitations.
Because each resonant frequency has its own characteristics, the synthesis of two or more
resonant frequencies may give resonant converter more merits.
LCC resonant converter (LCC) and LLC resonant converter (LLC) are very good examples
to illustrate the concept of resonant converter synthesis.
LCC resonant converter (LCC) is a popular resonant converter in many applications. It is the
combination of series resonant converter (tank A) and parallel resonant converter (tank C). SRF
is at the left side of PRF2, in other words, ZCS region of PRC. Around SRF, most region is ZCS
region, which is not preferred. Series resonant frequency (SRF) (black dot in Fig. 2.5 (a)) helps
in operation; however, the operation region (shadow in Fig. 2.5 (a)) is away from SRF. Thus,
LCC is still dominated by parallel resonant frequency (PRF), and have similar pros and cons as
PRC. It cannot operate at resonant frequency either.
Fig. 2.5 Circuit and voltage gains of LCC resonant converter
41
Fig. 2.6 Circuit and voltage gains of LLC resonant converter
LLC is the combination of tank A and tank B. It is different from LCC due to the different
order of two resonant frequencies. PRF2 is at the left side of SRF, in other words, the ZCS
region of SRF. PRF2 improves part of this region into ZVS region. Thus, it is safe to operate
around SRF, where is the optimal point. However, the right side of SRF is still flat [B.3]-[B.10].
The synthesis of SRF and PRF1 generates many merits : (1) ZVS capability for zero to full
load range; (2) low turn off current for primary side switches; (3) Zero-current-switching (ZCS)
for synchronous rectifier devices; (4) voltage gain boost capability without deterioration of
efficiency at normal condition.
Three-element resonant converters have more resonant frequencies than two-element. The
synthesis of resonant frequencies could provide extra benefits to the resonant converter.
However, the synthesis should be the right types and the right order of resonant frequencies to
improve the performance, or the improvement could be very little or worse.
2.3 Synthesis of Three-Element Resonant Converters
42
As mentioned before, LLC has many advantages due to the synthesis of resonant
frequencies. A systematic analysis is given below to search LLC-like resonant converters.
All the three-element resonant topology networks are given in Fig. 2.7 [B.11]. The topology
networks 15, 16, 17 are degraded to two-elements if the blocks replaced by an inductor or
capacitor.
Fig. 2.7 Three-element resonant topology networks
Replacing each block of networks with an inductor or capacitor, 36 topology candidates are
gotten in Fig. 2.8.
If the input source types are taken into account, there are some tanks are not suitable. Firstly,
if one or more elements are clamped to input voltage source, the whole resonant tank will be
degraded to less order system. Secondly, capacitors cannot be in parallel with voltage source.
Based on these two criteria, there are only 13 topologies left, shown in Fig. 2.9.
43
Fig. 2.8 Three-element resonant tanks
Fig. 2.9 Three-element resonant tanks for voltage source
44
The tanks in Fig. 2.9 are classified by combination of the resonant frequencies defined in
section 2.1.1.
From Fig. 2.10 to Fig. 2.14, the dot lines show the boundaries of ZVS and ZCS regions.
As shown in Fig. 2.10, Tank D and Tank H1 are the synthesis of SRF and NRF. Thus, this
group has similar characteristics like SRC. NRF sharps flat voltage gains above SRF, so ZVS
region of SRC is improved. Voltage regulation is achievable in a narrow frequency range.
However, resonant frequencies are the boundaries of ZVS and ZCS regions. Operation at
resonant frequency is still not achievable in this topology group.
Fig. 2.10 Voltage gains and tanks of SRF+Notch
As shown in Fig. 2.11, Tank C and Tank G1 are the synthesis of SRF and NRF. Thus, this
group has similar characteristics like SRC. In this group, NRF sharps flat voltage gains below
SRF, so ZCS region of SRC is improved.
45
Fig. 2.11 Voltage gains and tanks of Notch+SRF
As shown in Fig. 2.12, Tank K is the synthesis of PRF and NRF. Thus, this group has similar
characteristics like PRC. NRF sharps voltage gains above PRF, so ZVS region of PRC is
improved. However, the ZVS region of PRC is sharp before. Thus, the improvement is limited.
Meanwhile, PRF and Notch filter are the boundaries of ZVS and ZCS regions. Operation at
resonant frequency is still not achievable in this topology group either.
Fig. 2.12 Voltage gains and tanks of PRF+Notch
46
Fig. 2.13 Voltage gains and tanks of Notch+PRF
As shown in Fig. 2.13, Tank L and Tank O are the synthesis of PRF and NRF. Thus, this
group has similar characteristics like PRC. Similarly, NRF sharps voltage gains below PRF, so
ZCS region of PRC is improved. Operation at resonant frequency is still not achievable in this
topology group either.
Fig. 2.14 Voltage gains and tanks of SRF+PRF (LCC-like)
As shown in Fig. 2.14, Tank H, Tank Z and Tank X are the synthesis of SRF and PRF. Thus,
this group has both characteristics of SRC and PRC.PRF locates at higher frequency and SRF at
47
lower frequency. Tank H in this group is the resonant tank of LCC resonant converter. The
typical voltage gain curves are very similar as LCC resonant converter. Thus, this group is
named LCC-like resonant converters.
As shown in Fig. 2.15, Tank G, Tank W and Tank U are the synthesis of SRF and PRF, like
LCC. The difference is PRF locates at lower frequency and SRF at higher frequency. Tank G in
this group is the resonant tank of LLC resonant converter. The typical voltage gain curves are
very similar as LLC resonant converter. In this group, operation at resonant frequency is durable
due to the help of PRF. The resonant converter in this group can operate around series resonant
frequency without the concern falling into the ZCS region under given load condition.
Furthermore, they have the similar merits of LLC, like low loss at nominal conditions, ZVS for
whole load range, ZCS for output rectifier, etc. They can also achieve good efficiency at high
input voltage, where the front-end dc-dc works for 99% time, with wide input range. Thus, this
group is named LLC-like resonant converters [B.12].
Fig. 2.15 Voltage gains and tanks of PRF+SRF (LLC-like)
48
For the inherent problem from SRF, the startup and short-circuit protection of LLC are still
an issue due to the flat voltage gains above SRF.
2.4 Inherent Issues of LLC Resonant Converter
Fig. 2.16 shows the DC characteristic of the LLC resonant converter. Normally, the LLC
operates near the resonant frequency point (fs=f0) to keep the best efficiency. During the hold-up
time operation, the input voltage decreases while the output voltage should stay constant, thus
the switching frequency decreases (fs<f0) to boost the voltage-gain. Under the soft start-up
condition, the output voltage needs be built up from zero to the steady-state value, in order to
prevent from entering the zero-current-switching (ZCS) zone, the LLC should work in above
resonant frequency region (fs>f0).
2.0
hold-up
Voltage gain nVo/Vin
1.5
normal operation point
1
0.01
0.2
0.3
ZVS 0.6
0.5
ZCS 1.0
3.0 start-up
Q=20
vo=0
0
0 1.0 2.0 3.0 4.0
f n=f s/f 0
Fig. 2.16 DC characteristic of the LLC resonant converter
However, as shown in Fig. 2.17, if the start-up switching frequency (fs) is not high enough,
large voltage and current stress can be observed at the start-up moment [A.49]-[A.54].
49
large current and

voltage stress
iLr
VO
Vcr
full-load steady state
fmax=1.5f0
(a) Voltage and current stress of resonant tank during soft start-up
(b) Output current and voltage during soft start-up
Fig. 2.17 Soft start-up with low switching frequency
50
large current and

voltage stress
no stress
iLr
Vcr
VO
fmax=5f0 full-load steady state
(a) Voltage and current stress of resonant tank during soft start-up
(b) Output current and voltage during soft start-up
Fig. 2.18 Soft start-up with a high switching frequency
51
Otherwise, as shown in Fig. 2.18, if the LLC starts up with a very higher, this is about 5
times of normal operation frequency. At the beginning, the voltage and current stress will be not
large. However, during the soft start-up process, if fs decreases too quickly, large voltage and
current stress can still happen, which will trigger the over-current-protection. As a result, fs will
increase a bit to limit the stress, and then decreases smoothly to finish the soft start process. If fs
decreases slowly, the stress will stay small, but the output voltage will build up slowly, which
means, the soft start-up process will take much longer.
Thus, if frequency control is applied start-up process, the operation frequency of LLC
resonant converter should be very high to attenuate the voltage and current stress at the
beginning of start-up. Even start up at high frequency, the high voltage and current stresses still
high.
VgsQ1
VgsQ2
VgsQ3
VgsQ4
VN
VCr
ILr
I1
I2
ILm
Fig. 2.19 Phase-shifted control in full-bridge LLC (fs=2f0, Φ=π/2)
52
Besides the frequency control, phase-shifted control can be used in full-bridge LLC to limit
the voltage and current stress during the soft start process. Fig. 2.19 shows steady state
waveforms when fs=2f0 with π/2 phase shift. For the leading leg (Q1 and Q3), it is easy to realize
ZVS, since the resonant current is large at the turn-off moment. However, for the lagging leg (Q2
and Q4), it is hard to achieve ZVS due to the lower current to charge the junction capacitors at
the turn-off instant.
Based on frequency control and phase-shift control, [B.13] proposed an optimal trajectory
control for the LLC soft start-up. The resonant voltage and current stress can be limited, the
output voltage should be built up quickly and smoothly, and the ZVS of all switches can be
guaranteed.
Fig. 2.20 shows the experiment results. The start-up frequency is more than 2 times of
normal operation frequency.
Vo (2V/div)
+IMAX +Ipeak
-IMAX iLr (5A/div) -Ipeak
2ms 5ms/div
Fig. 2.20 Optimal trajectory soft start-up
53
However, all above control methods need very high switching frequency to attenuate the
voltage and current stress during the process of start-up. This issue comes from the inherent
problem of series resonant converter.
2.5 LLC-like Resonant Converters with Notch Filter
As discussed in previous sections, three-element topologies have more merits than two-
elements, but three-element topologies still have their own limitations. With poor frequency
selectivity, namely, very wide bandwidth, the frequency has to be increased very high to achieve
enough damping effect.
To improve the flat voltage gain above series resonant frequency of LLC-like resonant
converters, additional resonant elements are needed to create sharp gain. The desirable voltage
gain curves are plotted in Fig. 2.21. One possible way building sharp voltage gains is making a
notch filter.
Fig. 2.21 Desirable Gain Curves for LLC-like resonant converters
54
The resonant tank with notch filter characteristic can create zero voltage gain. Conceptually,
infinite impedance can be created when frequency operates at the resonant frequency f02 of the
notch filter. As a result, if overload or even shorted output condition occurs, the current can be
limited inherently by the resonant tank characteristic. The notch filter center frequency f02 can be
designed to achieve the over load protection.
Fig. 2.22 One example of selected 4-elements resonant converters
One of these resonant converters is depicted in Fig. 2.22. The notch filter section is created
with Lp and Cp. The band pass filter section consists of Lp, Cr and Cp.
In general, with variable frequency control, the resonant tank input voltage is excited as a
square waveform. According to the Fourier analysis, the voltage excitation only consists of odd
harmonics. Therefore, to avoid circulating energy at nominal condition, notch filter resonant
frequency f02should be designed at even times of nominal resonantfrequencyf01. Meanwhile, f02 is
preferred close to f01.Thus, f02 = 2f01 is a good choice.
When converter operates at f01, the notch filter represents inductive. The impedance of Cp is
much larger than Lp. The major part of primary side current flows through Lp, not Cp. The current
stress on Cp is much less than Cr.
55
However, the number of four-element resonant topologies is numerous. A systematic way
for classification and selection can save a lot effort and energy to find desired topologies.
All four-element networks are listed in [B.14], [B.15], as shown in Fig. 2.23. Each square
block represents one energy storage element. The networks in the dot line pane in Fig. 2.23 are
not suitable for voltage source, for they cannot regulate voltage.
Fig. 2.24 shows the relationship between resonant frequency types and topology networks.
Each square block in Fig. 2.24 represents one or more energy storage elements. The rectangle
block indicates input source or output sink or other energy storage elements. If the resonant
characteristics shown in Fig. 2.24 are wanted, the network below gain curves should be part of
resonant tanks.
As discussed before, series resonant frequency (SRF) is needed to achieve low switching
loss and low circulating energy at nominal conditions where the operation point is close to it.
Parallel resonant frequency is required for wide range voltage regulation. With the combination
of these two resonant frequencies, zero-current switching (ZCS) is achieved for the output
rectifiers, and no reverse recovery appears. A notch filter is also needed for startup and short-
circuits protection. Thus, the candidate topology should include PRF, SRF and NRF. The
desired gain curves should be similar as Fig. 2.21.
56
Fig. 2.23Four-element resonant topology networks
57
Fig. 2.24 Gain curve characteristics vs. networks
Fig. 2.25 Four-element resonant topology networks including Network 2(SRF)
58
Fig. 2.25 shows the topology networks including network 2 structure based on Fig. 2.23 and
Fig. 2.24. The number is 24.
Fig. 2.26 Four-element resonant topology networks including Network 2(SRC) and Network 1(PRF)
Fig. 2.26 shows the topology networks including network 1 and network 2 structures based
on Fig. 2.24 and Fig. 2.25. The number is 14.
Fig. 2.27 Four-element resonant topology networks including Network 2(SRC), Network 1(PRF) and
Network 3 or 4 (NRF)
59
Fig. 2.27 shows the topology networks including network 1, network 2 and network 3 (or
network 4) structures based on Fig. 2.26 and Fig. 2.24. The number is 7.
In topology network shown in Fig. 2.27,, each energy storage element can be replaced with
inductor or capacitor. The result is shown in Fig. 2.28,, ignoring which are not su
suitable for
voltage source.
Fig. 2.28 Four-element

element resonant topologies including Network 2 (SRC), Network 1(PRF) and
Network 3 or 4 (NRF)
60
Fig. 2.29 Four-element resonant topologies with desired resonant frequencies
Three different resonant frequencies should be arranged in order. Operation point is near
series resonant frequency (SRF) to achieve low switching loss and low circulating energy.
Parallel resonant frequency (PRF) should locate at lower frequency for wide range voltage
regulation. Notch resonant frequency (NRF) should be placed at higher frequency for start-up
and short-circuit protection. Thus, the order of resonant frequencies is PRF, SRF and NRF from
61
low frequency to high frequency (shown in Fig. 2.29). Eight topologies are gotten based on this
rule shown in Fig. 2.29. We can see that they also can be deduced from three-element LLC-like
resonant converters.
Similarly, the synthesis method based on resonant frequencies can be utilized in three, four,
five, or higher elements resonant converters.
2.6 LLC-like Resonant Converters with Notch Filter and Two Band-pass
Filters
With leakage inductance Lr, a second band pass filter is created. A 5-elements resonant tank
is given as an example, shown in Fig. 2.30. The second band pass filter is designed at the three
times of the nominal frequency. The third order harmonics energy of primary side can be
delivered through this band pass filter as shown in Fig. 2.31. Obviously, the power delivery is
more efficient.
Fig. 2.30 One example of 5-element resonant converters
62
Fig. 2.31 Power Delivery Difference between LLC & 5-e
Fig. 2.32 Gain Curves of selected 5-elements resonant converter
63
Fig. 2.33Five element resonant tank cell based on LLC with notch filter and 3rd harmonics band filter
64
However, the number of five-element resonant topologies is even much more than four-
element resonant converters. A same systematic way is used for searching resonant topologies
with similar gain curves like Fig. 2.32.
Based on the searching results of previous section, one more inductor should be added into
resonant tanks in Fig. 2.29. Fig. 2.33 shows the searching results of 5-element resonant converters
with desired voltage gain curves.
Each resonant frequency expression is shown in equation (3.1), (3.2) and (3.3) for the
example circuit is shown in Fig. 2.30.
When switching frequency is very low, the effect of Cp is weak, Lr and Lp are in series to
resonance with Cr. Lp resonances with Cp to get notch filter. If switching frequency reach very
high, the impedance of Lp is very high. Thus, only Cr and Cp are in series to resonance with Lr.
= ≈ + (2.1)
= (2.2)
= ≈ (2.3)
Fig. 2.34shows the waveforms of under full load condition. The tank current is more like
square waveform which combines fundamental and 3rd harmonics. Fig. 2.35 shows the tank
waveforms under short-circuit condition. The output current is limited to less than half load.
65
Fig. 2.34 Experimental results for 5-element resonant converters under full load
Fig. 2.35 Experimental results of 5-element resonant converters under short-circuit condition
2.7 Conclusion
A systematic method to classify and select resonant tanks is proposed. Three-element, four-
element and five-element resonant tanks are analyzed based on the proposed principle.
66
Chapter 3. Evaluation of LLC-like Resonant Converters

In this chapter, a general evaluation system is proposed to help evaluate the performance of
resonant converters. State-plane analysis with new normalization is proposed to evaluate the
resonant converter performance. Based on it, the voltage stress, current stresses and apparent
power of resonant converters are easy to compare. This method can help select suitable circuit
topology for certain application. Meanwhile, it also can help resonant converters’ design. The
important performance factors, like start-up, short-circuit protection, SR driving and EMI, are
also taken into account for the whole evaluation system.
3.1 Introduction
There are still many candidate topologies which fits required characteristics based on
previous selection methodology. It is not enough to choose resonant topologies based on the
combination of resonant frequencies. Thus, several evaluation methods are proposed. First, the
current stress of primary devices is major concern, because it will determine the conduction of
primary side. Secondly, the voltage stress on resonant capacitor is still important due to the life
time of capacitor. Then, the integration is taken into consideration, which includes the ease of
magnetic integration and the impacts of integration, including synchronous rectifier drive, EMI
noise, transformer loss, etc.
3.2 State-Trajectory Analysis of LLC Resonant Converter
67
In order to determine the optimal trajectory under transition conditions, the state-plane
analysis is introduced. In [C.1]-[C.3], the state plane is presented to analyze the SRC steady state
and transient behavior. In this section, we extend this approach to the LLC resonant converter.
To simplify the analysis, it is assumed that the output capacitor is large enough that the
output voltage (Vo) can be approximated to be constant. A second assumption is that the circuit
losses, including the resonant tank, switch, and filter losses, are negligible.
Under these conditions, the LLC resonant converter can be viewed as a piecewise linear
system switching among six possible linear stages.
• Operation stage I
+ vCr - iLr
Fig. 3.1 Equivalent circuit of stage I
iLrN
vCrN
(1-nVON, 0)
Fig.3.2 State-trajectory of stage I
68
The primary main switch Q1 turns on, and the resonant current flows through the secondary-
side synchronous rectifier S2. In this stage, the magnetizing inductor Lm is clamped by the output
voltage. The equivalent circuit is shown in Fig. 3.1, where the resonant capacitor Cr and the
resonant inductor Lr form the series resonant tank. The source voltage across the resonant tank is
Vin-nVo. Thus, we can write:
1
v&Cr = iLr (3.1)
Cr
i&Lr = − vCr + (Vin − nVo )

1 1
(3.2)
Lr Lr
WherevCr and iLr are the resonant capacitor voltage and resonant inductor current. Combining
(5.1) and (5.2), vCr and iLr can be expressed as:
vCr − (Vin − nVo ) = ILr0 ⋅ Z0 ⋅ sin[ω0 (t −t0 )] + [VCr0 − (Vin − nVo )] ⋅ cos[ω0 (t −t0 )] (3.3)
VCr 0 − (Vin − nVo )

i Lr = I Lr 0 ⋅ cos[ω 0 (t − t 0 )] − ⋅ sin[ω 0 (t − t 0 )] . (3.4)
Z0
where VCr0 and ILr0, are the initial conditions at time t=t0, Z0 = Lr Cr is the characteristic
impedance, and ω0 = 1 Lr Cr is the resonant frequency.
By normalizing all voltages with the voltage factor Vin, and all currents with the current
factor Vin/Z0 in (5.3) and (5.4), the following expressions are derived:
69
vCrN − (1− nVoN ) = ILr0N ⋅ sin[ω0 (t −t0 )] +[VCr0N − (1− nVoN )] ⋅ cos[ω0 (t −t0 )] (3.5)
iLrN = ILr0N ⋅ cos[ω0 (t −t0 )] −[VCr0N −(1− nVoN)] ⋅ sin[ω0 (t −t0 )] . (3.6)
The subscript N in (5.5) and (5.6) refers to normalized circuit variables.
From (5.5) and (5.6), the trajectory equation is given by:
[vCrN − (1− nVoN )]2 + iLrN2 = [VCr0N − (1− nVoN )]2 + ILr0N 2 , (3.7)
which behaves as a circle as shown in Fig. 5.2, whose center (1-nVoN, 0) is determined by the
source voltage across the tank, and radius depends on the initial conditions: VCr0 and ILr0.
• Operation stage II
Fig. 3.3 Equivalent circuit of stage II
iLrN
vCrN
(1+nVON, 0)
Fig. 3.4State-trajectory of stage II
70
In operation stage II, Q1 is on, but the current flowing through the secondary side changes its
direction as shown in Fig. 3.3. This occurs when the switching frequency is larger than the
resonant frequency (i.e. fs>f0). This stage is similar to Stage I, but the source voltage across the
resonant tank changes to Vin+nVo.
Similar to Stage I, the trajectory expression is given by:
[vCrN − (1+ nVoN )]2 + iLrN2 = [VCr0N − (1 + nVoN )]2 + I Lr0N 2 , (3.8)
which behaves as a circle as well, but the center changes to (1+nVoN, 0) as shown in Fig. 3.4.
• Operation stage III:
Fig. 3.5Equivalent circuit of stage III
iLrN
vCrN
(1, 0)
Fig. 3.6 State-trajectory of stage III
71
In operation stage III, Q1 is on, but there is no current flowing through the secondary side.
This stage occurs when fs<f0. In this stage, the magnetizing inductor Lm joins the resonance. The
equivalent circuit is shown in Fig. 3.5, where Cr, Lr, and Lm form the series resonant tank. The
source voltage across the resonant tank is Vin, and:
1
v&Cr = iLr (3.9)
Cr
1 1
i&Lr = − vCr + Vin (3.10)
Lr + Lm Lr + Lm
From (5.9) and (5.10), vCr and iLr in Stage III are given by:
vCr −Vin = I Lr0 ⋅ Z1 ⋅ sin[ω1(t − t0 )] + (VCr0 −Vin ) ⋅ cos[ω1(t − t0 )] (3.11)
VCr 0 − Vin
i Lr = I Lr 0 ⋅ cos[ω1 (t − t 0 )] − ⋅ sin[ω1 (t − t 0 )] , (3.12)
Z1
where Z1 = (Lr + Lm ) Cr is the characteristic impedance, and ω1 = 1 (Lr + Lm )Cr is the
resonant frequency when Lm is in the resonance.
With the same normalization factors, the normalized forms of vCr and iLr are
⋅ sin[ω1 (t − t 0 )] + (VCr0 N − 1) ⋅ cos[ω1 (t − t 0 )]

I Lr0 N
vCrN − 1 = (3.13)
Z 0 Z1
72
= Lr 0 N ⋅ cos[ω1 (t − t 0 )] − (VCr 0 N − 1) ⋅ sin[ω1 (t − t 0 )] .

i LrN I
(3.14)
Z 0 Z1 Z 0 Z1
The trajectory is obtained from (4.13) and (4.14):
2 2
 i   i 
(v CrN − 1) +  LrN
2
 = (V Cr 0 N − 1)2 +  Lr 0 N  , (3.15)
 Z 0 Z1   Z 0 Z1 
which behaves as an ellipse with a center (1,0) as shown in Fig. 3.6. Since Lm participates in the
resonance, the characteristic impedance Z1 becomes large (Z1>Z0). If it is normalized with the
same current factor Vin/Z0, the trajectory will be an ellipse.
If switch Q1 is off and Q2 turns on, there are another three operation stages:
• Operation stage IV
Fig. 3.7 Equivalent circuit of stage IV
73
iLrN
vCrN
(nVON, 0)
Fig. 3.8 State-trajectory of stage IV
In operation stage IV, the primary-side main switch Q2 turns on, and the resonant current
flows through the secondary-side SR S1 as shown in Fig. 3.7. The only difference from Stage I is
the source voltage across resonant tank becomes nVo. Deriving the normalized equation for the
state trajectory, we get:
(vCrN − nVoN )2 + iLrN2 = (VCr0N − nVoN )2 + I Lr0N 2 , (3.16)
and thus, the circle center changes to (nVoN, 0) as shown in Fig. 3.8.
• Operation stage V
Fig. 3.9 Equivalent circuit of stage V
74
iLrN
vCrN
(-nVON, 0)
Fig. 3.10State-trajectory of stage V
In operation stage V, Q2 is on, but the current on the secondary side changes its direction,
flowing through S2, as shown in Fig. 3.9. This occurs when fs>f0, similar to stage II. The
normalized expression for the trajectory is:
(vCrN + nVoN )2 + iLrN2 = (VCr0N + nVoN )2 + I Lr0N 2 . (3.17)
Now the circle center changes to (-nVoN, 0) as shown in Fig. 3.10.
• Operation stage VI
Fig. 3.11Equivalent circuit of stage VI
75
iLrN
vCrN
(0 ,0)
Fig. 3.12State- trajectory of stage VI
In operation stage VI, Q2 is on, but there is no current flowing through the secondary side.
This stage happens when fs<f0. Like in stage III, Cr, Lr and Lm form the series resonant tank. But
the source voltage across the resonant tank changes to 0. The trajectory after normalization is:
2 2
 i   I 
+  LrN  = V Cr 0 N +  Lr 0 N 
2 2
v CrN , (3.18)
 Z 0 Z1   Z 0 Z1 
which is an ellipse as in stage III, but with a center (0, 0) shown in Fig. 3.12.
The LLC steady-state trajectory is the combination of the above six operation stages. Fig.
3.13 shows the time-domain waveforms when the LLC operates at the resonant frequency point
(fs=f0). Fig. 3.14shows the corresponding trajectory locus when mapping the normalized resonant
current iLrN and resonant voltage vCrN to the state plane. To the half-bridge LLC resonant
converter, when fs=f0, the voltage gain is one; that is, nVoN=0.5. Therefore, the two half circles of
Stage I and Stage IV have the same center (0.5, 0), which combine to form one whole circle.
76
Stage I Stage IV
VgsQ1
VgsQ2
iLr
iLm
VCr
t0 t1 t2
Fig. 3.13Steady-state time-domain waveforms when fs=f0
stage I
ρ (0.5,0)
t1
ρ
t0
(t2) stage IV
Fig. 3.14Steady-state trajectory when fs=f0
When the LLC operates below the resonant frequency region (fs<f0), the magnetizing
inductor Lm participates in the resonance from times t1 to t2 and from t3 to t4, as shown in Fig.
3.15. When mapping to the state plane in Fig. 3.16, there are two partial ellipse trajectory loci of
Stage III and Stage VI, connecting two half circles of Stage I and Stage IV with split centers. In
fact, when fs<f0, the voltage gain is larger than one (i.e., nVoN>0.5), and the circle center of
77
StageI is on the left of (0.5, 0), since 1-nVoN<0.5; and the circle center of Stage IV is on the right
of (0.5, 0), since nVoN>0.5.
I III IV VI
VgsQ1
VgsQ2
iLr
iLm
VCr
t0 t1 t2 t3 t4
Fig. 3.15 Steady-state time-domain waveforms when fs<f0
I
III
ρ t1 t2
(0,0) ρ (1,0)
t0 t3
(t4) VI IV
Fig. 3.16 Steady-state trajectory when fs<f0
In Fig. 3.17and Fig. 3.18, the LLC operates above the resonant frequency region (fs>f0). From
time t1 to t2, it operates in Stage V, which corresponds to the partial locus of the large circle,
whose center is located near (-0.5, 0). Similarly, from times t3 to t4, it operates in Stage II,
corresponding to the partial locus of another large circle whose center is located near (1.5, 0).
78
I V IV II
V gsQ1
V gsQ2
iLr
iLm
VCr
t0 t1 t2 t3 t4
Fig. 3.17 Steady-state time-domain waveforms when fs>f0
ρ' I
t
ρ 1
t2 V
(t4) t0
II t3 ρ
ρ'
IV
Fig. 3.18 Steady-state trajectory when fs>f0
3.3 New Normalization for State-plane
State-plane analysis was found to be a simple, yet powerful method which can clearly
portray the steady-state and transient operation of resonant converters. Also, the current stress
and voltage stress of resonant components can be observed in one single figure. Significantly the
state plane directly indicates the resonant tank energy level at which the system is operating.
79
As stated above, state plane of resonant converters is good tool to analyze the performance
of whole converter. However, if we want to compare different design of same topologies or
different topologies, there are some revisions needed. Previous state-plane analysis is based on
same normalize factors, which are Vin for voltage, Vin/Zo for current, as shown in Equation (3.19).
Zo is the characteristic impedance of resonant converter.
= I = V!"⁄Z$ (3.19)
However, Zo is varied in different designs or different topologies. For resonant converters
having two or more resonant frequencies, Zo is even varied at different time period. It is hard to
compare the performance based on Vin/Zo.
Fig. 3.19 gives a comparison example of LLC and CLL resonant converters. Both converters
work at same condition: Vin/Vo=400V/12V, output power Po=250W, switching frequency
fs=1Mhz. The input current waveforms of two converters are the same. However, the normalized
currents do not represent the real current value due to the different characteristic impedance Zo.
In this example, Z0 = Lr Cr = 88 for LLC, and Z0 = L1 ⋅ L2 /(L1 + L2 ) Cr = 131 for CLL.
80
Fig. 3.19LLC and CLL state-plane comparison based on conventional normalization
For given application, the output power Po and input voltage Vin are fixed. These two values
are independent with resonant topology. Po/Vin also means the average input current assuming no
loss in the converter. Thus, Po/Vin is an ideal normalized factor for current. Meanwhile, Vin is
the voltage normalizing factor, as shown in Equation (3.20). In the resonant converters, primary
side current represents the conduction loss of primary side devices, and the voltage cross
81
capacitor express the life time of it. Thus, the state-plane of them can illustrate those key
performances.
V = V!" I = P$ ⁄V!" (3.20)
Because of full bridge structure, the rms value of input voltage equals its average value. At
the same time, normalized input current IinNRMS equals the apparent power Pappr over output
power Po.as shown in Equation (3.21).
I!" '() = I!"'() ⁄*P$ ⁄V!" + = I!"'() × V!" ⁄P$ = P-../ ⁄P$ (3.21)
Series resonant converter is taken as an example to illustrate how to utilize the new
normalization factors.
When SRC operates at resonant frequency, the tank impedance is zero. The input voltage of
resonant tank is square waveform, while the input current is sinusoidal waveform, as shown in
Fig. 3.20. Only fundamental of input voltage contributes real power, while high order harmonics
of input voltage do not. The RMS value for fundamental of input voltage is shown in Equation
(3.22).
= ∙ V3
√
_!"'() (3.22)
The apparent power of whole converter with new normalization factors is
56778 :; ∙<=>?@A
P4 _ = 59
= √
= √
≈ 1.11 = I!" '() (3.23)
∙:; ∙<=>?@A
B B
That means 11% extra power is the minimal value for resonant converter operation.
82
(a) Series Resonant Converter
(b) Fourier transform of input voltage vin and input current iin
Fig. 3.20Series resonant converter operates at resonant frequency
Fig. 3.21State-plane of SRC operates at resonant frequency
83
Fig. 3.21 shows the state-plane of SRC at resonant frequency. The amplitude of normalized
current is 1.57 = 1.11 × √2, which is the minimal value of normalized current.
Design factors for LLC are shown in Equation (3.24). Ln is the ratio of magnetizing
inductance Lm and resonant inductor Lr. Q is the quality factor. Cr is the resonant capacitor. RL
represents the full load. n is the transformer turns ratio.
L" = LI ⁄L/ Q = KL/ ⁄C/ P*n R O + f$ = 1⁄ 2πKL/ ⁄C/ (3.24)
Fig. 3.22(a) is the voltage gain curves of LLC, when Ln=12.5. Fig. 3.22(b) plots the steady
state-plane under new normalized factor for different Q and same Ln under full load. The voltage
and current stresses of different design can be easily compared in one coordinate system.
(a) Voltage gain curves of LLC (b) State-plane of given design @full load
Fig. 3.22 Steady-state trajectory when fs=f0
84
Fig. 3.23plots the steady state-plane under new normalized factor for different Q and same Ln
under full load. The voltage and current stresses of different design can be easily compared in
one coordinate system.
Fig. 3.23V-I stresses of LLC resonant converters for Ln=12.5
Fig. 3.24V-I stresses of LLC resonant converters for different Ln
85
Fig. 3.24depicts the voltage and current stress of LLC for different Ln and Q. Each line in Fig.
3.24means those designs sweep Q for one Ln. As Q increases, the current stress decrease, and the
voltage stress rises. The optimal design for each Ln should be around knee point of according
line. It is also known that the large the Ln is, the smaller voltage and current stresses the knee
point locates. And it is also known that the minimal apparent power of LLC is 1.11 times of
output power.
Based on chapter 4, there are three good candidates in three-element resonant topologies.
The V-I stress figure is easily drawn for other two resonant topologies.
Fig. 3.25gives the LCL resonant converter. Fig. 3.26shows the whole picture of LCL resonant
converter V-I stress.
Fig. 3.25LCL Resonant Converter
86
Fig. 3.26V-I stresses of LCL resonant converters for different Ln
Fig. 3.27gives the CLL resonant converter. Fig. 3.28shows the whole picture of LCL resonant
converter V-I stress.
Fig. 3.27CLL Resonant Converter
87
Fig. 3.28V-I stresses ofCLL resonant converters for different Ln
Fig. 3.29V-I stresses ofthree-element resonant converters for different Ln
88
Fig. 3.29 gives the V-I stress comparison of all three 3-element resonant converters. It is
obvious that LCL resonant converter have larger voltage and current stress than CLL and LLC
resonant converters. The shadow area shows the difference.
Fig. 3.30 V-I stresses of three-element resonant converters for different Ln
One design for each topology is chosen to depict in state-plane based on proposed
normalized factor in Fig. 3.30. Each design is one dot shown in Fig. 3.29.
3.4 Evaluation based on Startup, Synchronous Rectifier and EMI
Startup and short-circuit protection are not good in LLC resonant converter. Fig. 3.31gives
one example of LLC resonant converter startup procedure. The LLC resonant converter starts up
at one higher frequency than resonant frequency, where current stress depends on current gain.
The output voltage gradually increases at this frequency. When it reaches one preset voltage, the
89
switching frequency slowly decrease to resonant frequency. The whole start up progress is
shown as dot line in Fig. 3.31.
Fig. 3.31Startup example of LLC Resonant Converter
The key problem in LLC startup is the current stress is very high.
Fig. 3.32presents the voltage gain and current gain curves comparison of three-element
resonant converters. It is clear that CLL resonant converter has the same problem as LLC.
However, the LCL resonant converter is quite different. The startup current of LCL resonant
converter is small compare with the full load current.
90
Fig. 3.32Startup Comparison of LLC, CLL & LCL Resonant Converters
The accurate synchronous rectifier drive is very critical in high output application for
resonant converters. Utilizing primary side current of transformer is very easy way to get precise
SR drive signal.
The LLC resonant converter normally integrates the magnetizing inductance into the
transformer. The magnetizing current will give phase shift between primary and secondary
current. That is the reason LLC resonant converter cannot use primary side current for SR drive.
In another side, LLC can use primary side current without integrate Lm and Lr, which cause
power density suffers.
91
CLL and LCL resonant converter is different as shown in Fig. 3.33. Both of them can
integrate the leakage inductance without suffer SR drive.
Fig. 3.33SR drive Comparison of LLC, CLL & LCL Resonant Converters
92
Fig. 3.34CM noise spectrum of LLC Resonant Converter
LLC resonant converter has high common mode noise at and beyond switching frequency.
One example is shown in Fig. 3.34compared with EMI standard. The common mode noise path is
given in Fig. 3.35.
Fig. 3.35CM noise path of LLC Resonant Converter
93
Fig. 3.36CM noise spectrum of LLC, CLL & LCL Resonant Converters
Compare with LLC, CLL has lower CM noise at very high frequency spectrum due to soft
voltage edge. The LCL resonant converter has very different EMI characteristics. The voltage
cross transformer is like sinusoidal, thus, there is high peak at switching frequency but very low
for higher frequencies. Fig. 3.36 shows the CM noise spectrum comparison.
Fig. 3.37shows the comparison results of all three resonant converters.
94
Fig. 3.37CM noise spectrum of LLC, CLL & LCL Resonant Converters
3.5 Conclusion
A method based on state-plane is proposed to help evaluate the performance of resonant

converters. Based on it, the voltage stress, current stresses and apparent power of resonant
converters are easy to compare. This method can help select suitable circuit topology for certain
application. Meanwhile, it also can help resonant converters’ design.
95
Chapter 4. Integrated System Design for High Current
Application
In this chapter, LLC resonant converter with matrix transformer is introduced. Flux
cancellation method is utilized to reduce core size and loss. Synchronous Rectifier (SR) devices
and output capacitors are integrated into secondary windings to eliminate termination related
winding losses, via loss and reduce leakage inductance.
4.1 Introduction
As mentioned before, LLC type resonant converters are excel in both of efficiency and
power density [D.1]-[D.8]. Firstly, it has ZVS capability for zero to full load range. Then, low
turn off current is achievable for primary side switches. At the same time, synchronous rectifier
(SR) devices are zero-current-switching (ZCS). At the end, it gets voltage gain boost capability
without efficiency deterioration at normal condition, suitable for application with hold-up time
requirement. In comparison with soft switching PWM converters, LLC resonant converters can
achieve higher frequency and higher power density with better efficiency.
High switching frequency introduces high power density. However, the switching loss and
magnetic components losses are increased. The emerging GaN devices [D.9]-[D.11] give the
opportunity to diminish the switching related loss. Hence, the magnetic design becomes very
critical.
96
The isolated high output current DC/DC applications, like server, suffer very high
conduction loss, especially for the transformer and SR devices [D.12], [D.13]. Paralleling SR
devices are required to reduce device conduction loss. For winding loss, matrix transformer is
good candidate for its high current capability [D.14]. However, the price paid for matrix
transformer is multiple cores. Reference [D.15] uses flux cancellation method to reduce the core
number for less core volume and core loss. The drawbacks are expensive 12 layers PCB
transformer and the large conduction loss from alternative current through vias among PCB
layers.
To eliminate the AC conduction loss among layers, one effective way is mounting MOSFET
device on PCB board [D.17], [D.16]. This structure can help reduce conduction loss between
layers dramatically. Both transformer structures have litz wires for primary side windings and
PCB board for secondary side windings. Nevertheless, its primary winding is complicated and
hard to manufacture.
High termination losses are the third concern. Due to the proximity effect and skin effect, the
currents in adjacent terminals with opposite directions attract each other. Thus, very high losses
and hot spots are generated. This termination loss deteriorates efficiency significantly. For the
conventional design, all SRs are placed on the motherboard. Physically speaking, a large loop of
the secondary side rectifiers is inevitable. Consequently, a large distribution loss will be generated.
In addition, for large number of SR devices, it is extremely difficult to achieve a symmetrical
layout for each SR. Hence, current-sharing of the SR is a severe problem.
97
For high frequency high step down LLC resonant converters, the transformer loss dominates
the whole converter loss [D.12]-[D.16], thus, the transformer design is very critical.
For conventional wired transformer, the primary side windings are solid or multi-strand
wires for low switching frequency and low cost design. At high frequency, litz wires are chosen
for lower conduction loss. For off-line dc-dc converters, the secondary side current is much
higher than the primary side. Therefore, copper foils are selected for the secondary side winding.
To alleviate ac winding loss, the primary side and the secondary side windings are interleaved as
a sandwich structure. Mostly, the secondary side windings are connected with the motherboard
via copper poles. However, this structure suffers high termination losses. Due to proximity effect,
the current surges in ac connection terminals. Meanwhile, very little current goes through the
center of the terminals because of the skin effect. Thus, very high losses and hot spot are
generated [D.17].
In reference [D.16], the secondary side windings are integrated with the SR devices on PCB
board, as shown in Fig.4.1. Virtually, termination is achieved at dc output. Thus, the termination
loss can be reduced considerably. However, there are some limitations. The primary side
windings are interleaved with the secondary side windings as one. For each cell, the primary side
winding is connected in series. To finish the whole transformer, multiple cells are connected in
parallel. As a result, this structure becomes very complicated. In addition, since all primary side
and secondary side windings are paralleled, current sharing between windings might be a
potential issue for some circumstances, especially for high current applications.
98
Fig.4.1 Circuit schematic of the transformer and SR in [D.16].
This transformer structure is simplified by [D.17]. The primary side winding are placed in
series, and secondary windings in parallel, as shown in Fig.4.2. However, the manufacture
process is still complicated for half litz wires half PCB board transformer. The current sharing of
paralleled secondary side windings is not guaranteed neither.
99
Fig.4.2 Circuit schematic and transformer structure in [D.17].
The planar transformer is easy to be adapted to an automatic manufacturing process [D.13].
It can achieve high power density and low profile. Mostly, multi-layer PCB board is used as
planar transformer windings. The primary side and secondary side windings interleaves as
sandwich structure. Nevertheless, it is hard to integrate SR devices on secondary side windings
inside PCB board.
The state-of-art industry product is Vicor BCM series, as shown in Fig.4.3. By the soft-
switching topology called Sine Amplitude Converter (SAC), which is very similar as LLC, the
module can achieve ZVS-ZCS operation. Thus, high efficiency and power density is achievable.
There are two half bridge stacked together at primary side, thus low voltage device can be utilized.
Two core transformer structure is used to help split secondary side current. The secondary side
100
conduction loss is reduced. The typical power density of this series is about 1000W/in3, while the
peak efficiency is around 95.5%. However, this module utilizes 12 MOSEFTs, 14 layers PCB
board to achieve this performance.
Fig.4.3Vicor BCM series.
Thus, how to improve the efficiency and power density of high step-down dc/dc converter is
the key issue for the whole performance of datacenter.
4.2 Matrix Transformer with Flux Cancellation
The matrix transformer could help to fix the dilemma. The matrix transformer is defined as
an array of elements interwired so that the whole functions as a single transformer [D.14]. Each
element being a single transformer that contains a set turns ratio, i.e. 1:1, 2:1 ….n:1. The desired
101
turns ratio is obtained by connecting the primary windings of the elements in series or parallel
and the secondary’s in series or parallel. For the high output current cases only a single turn
secondary will be considered. The benefits of the matrix transformer are that it can split current
between secondary windings connected in parallel, reduce leakage inductance by lowering the
N2 value of the secondary loop inductance, and improve thermal performance by distributing the
power loss throughout the elements. On the other hand, the matrix transformer structure also can
effectively reduce MMF (magnetomotive force) of windings, especially for PCB winding. That
also means leakage inductance and winding AC resistance reduction.
Fig. 4.4LLC resonant converter with four sets output
Fig. 4.4 shows one typical example of 400V/12V 1kW LLC resonant converter. The GaN
HFETs Q1 and Q2 are primary devices. SR1, SR2, etc are synchronous rectifier (SR) devices.
102
The resonant capacitor is Cr. Lr is the leakage inductance of transformer, while Lm is the
magnetizing inductance. The transformer turns ratio is 16:1:1. Based on optimized loss of
synchronous rectifier devices, four sets of center-tap output are chosen. The detail reasons are
discussed in [D.18].
Fig. 4.512 layer PCB structure with one core and its MMF
Fig. 4.5 shows the planar transformer structure employing Fig. 4.4 circuit by one core
structure and 12 layer PCB board as transformer windings. The primary side winding is in series,
and secondary side in parallel. All the windings are wrapped around the single core. 12 layer
PCB windings can be divided into 4 sections. Each section has three PCB layers with two layers
for secondary winding and the center one for primary side. As indicated in Fig. 4.5, when
secondary winding 1 conducts, the secondary winding 2 does not conduct due to center-tap
structure. The current direction for first and second layer PCB board is marked. The other
103
sections are similar. The magnetomotive force (MMF) between primary and secondary windings
is four times of primary side current, as shown Fig. 4.5.
Fig. 4.6 shows the PCB winding structure employing matrix transformer concept. The
current direction for the first three PCB layers are shown, other layers are similar. The primary
and secondary windings wrap on two cores. The MMF between primary and secondary windings
are only twice of primary current, as indicated in Fig. 4.6. Low MMF means low leakage
inductance and low AC resistance of windings.
Fig. 4.612 layer PCB structure with two cores and its MMF
However, 12 layers PCB board is seldom used in most front end applications, like server,
telecom, etc. Four layer PCB board is widely employed. Thus, matrix transformer based on 4
layer PCB board is adopted. The whole transformer is split into four small transformers, whose
primary side are in series and secondary side in parallel. Each small transformer has four turns
104
primary windings and two turns center-tap secondary windings as indicated in Fig. 4.7. As
shown in Fig. 4.8, two PCB layers are taken for primary windings at top and bottom, and the
inner two layers for secondary windings. At this instant, the secondary winding 1 and primary
windings are conducted. The MMF between primary and secondary windings are only twice of
primary current, that is very similar as the MMF of 12 layer structure in Fig. 4.6. Consequently,
low MMF means low AC winding resistance and low leakage inductance of transformer. At the
same time, the current sharing among secondary side are automatically achieved by the structure
of primary winding in series and secondary winding in parallel.
Fig. 4.7LLC Resonant Converter with Matrix Transformer Structure
When low MMF is achieved, the core loss is increased because of multiple cores. The flux
excursion for each transformer core is the same in each matrix element as the traditional
transformer as mentioned in Fig. 4.5. The primary winding pattern with four magnetic cores in
105
Fig. 4.8 is shown in Fig. 4.9. Just one of two primary winding layers is shown for simplicity. The
other layer winding pattern is similar.
Fig. 4.8Four layer PCB structure with four cores and its MMF
Fig. 4.9 Primary side winding pattern with four U-I cores matrix transformer
106
It has been known that using integrated magnetic structures can allow for flux cancellation
[D.15]. The matrix transformer configuration offers an ideal case for magnetic integration as a
result of the cores being excited with identical voltage-second generating the same flux in each
core. This allows for almost complete flux cancellation if designed properly. This can decrease
overall size and core loss.
Fig. 4.10 Rearranged primary side winding pattern with four U-I cores
Fig. 4.11Primary side winding pattern for two E-I core matrix transformer
The primary side winding rearranged to reverse flux direction of core 2 and core 4 inFig.
4.10. Hence, the U-I core 1 and core 2 can both turn 90 degree and merge into E-I core 1’, while
core 3 and core 4 can do into core 2’ as shown inFig. 4.11. The flux density in each core is the
same due to same voltage-second, the magnetic flux in center leg of core 1’ can be cancelled, so
107
can core 2’. Thus, the flux density in the center leg is nearly zero. The center leg of two E-I cores
actually can be reaped, so two U-I cores are gotten, as pictured inFig. 4.12. By flux cancellation,
four U-I cores of matrix transformer can be reduced to two U-I cores. The core loss and core size
can be reduced into half. Then, the whole circuit is shown inFig. 4.13.
Fig. 4.12Primary side winding pattern for two core matrix transformer with flux cancellation
Fig. 4.13LLC resonant converter with proposed transformer structure
108
4.3 Winding Structure Integrated with Synchronous Rectifier Devices
As mentioned above, the planar matrix transformer has only four layer PCB board as
windings, and easy manufacture. It also can effectively reduce leakage inductance and winding
AC resistance with low MMF compared to the traditional single transformer design. Meanwhile,
flux cancellation method will help reduce magnetic cores’ size and loss.
For better efficiency, the transformer winding structure needs more improvement. In [D.15]
and [D.17], we know the termination of secondary side windings introduces large AC conduction
loss. To eliminate the AC loss introduced by termination, SR devices need be integrated into
secondary side windings [D.16]. However, [D.16] did not givethe detail loss analysis and
optimal winding design. The following discussion gives insight analysis and optimal transformer
design of previous proposed matrix transformer structure.
The matrix transformer structure shown in Fig. 4.8has two parts. Each part has 8 turns
primary windings in series, two sets center-taped secondary windings in parallel and one set of
U-I core. These two parts are also in series at primary side and in parallel at secondary side.
There are two options for winding arrangement.
The first one shown in Fig. 4.13has good interleaving structure. The winding structure of
one U-I core indicated in dashed box is shown in Fig. 4.14andFig. 4.15. The top view is in Fig.
4.14, and the side view in Fig. 4.15. Primary windings locate top and bottom side of PCB board.
For one half circle of switching period, primary winding conducts current, when secondary
winding 1 (Sec1) conducts and secondary winding 2 (Sec2) does not. The magenetomotive force
109
(MMF) of this structure is given in Fig. 4.15. However, the secondary windings need vias and
external terminal to connect the SR devices on top and bottom layers, as indicated in Fig. 4.14.
Fig. 4.14 Top view of 1st winding arrangement
The Maxwell 3D Finite Element Analysis (FEA) simulation is applied to estimate the
leakage inductance and AC resistance of transformer windings. Fig. 4.16 shows the FEA
simulation results of AC current distribution for secondary side of transformer. The dash arrows
express the AC current direction. It is clear that there is severe current crowding along the
terminal connecting SR device and capacitors. The reason is that the termination does not
interleave with primary winding, thus, the AC currents in opposite direction will attract each
other along the edge. The doted box is where vias connecting the middle layer secondary
windings and the top layer. The current through vias is not even either.
110
Fig. 4.15Cross view of the first winding arrangement and its MMF
Fig. 4.16 AC current distribution for secondary winding of the first winding arrangement
111
(a) Current Density Distribution
(b) Magnetic Field Intensity Plot
Fig. 4.17 FEA simulation for the first winding arrangement
The simulation result of one U-I core transformer is given in Fig. 4.17. The secondary side
AC resistance including termination is Rsec=6.81mΩ, and the leakage inductance is Lk = 176.4nH.
112
Fig. 4.18 FEA simulation for the first winding arrangement with output shorted
If the transformer secondary side is cut off as dash line indicated in Fig. 4.17(a), and shorted
by copper bar at the cut port. Fig. 4.18is the simulation result. The secondary side AC resistance
is Rsec=3.35mΩ, and the leakage inductance Lk = 58.0nH. This phenomenon means the AC
resistance of termination including via is 3.46 mΩ, and the leakage inductance 118.4nH.
However, there are opposite direction currents flow in and out the shorted port, which will also
cause current crowding.
113
Fig. 4.19 FEA simulation for the first winding arrangement with secondary winding shorted
Fig. 4.19is the simulation result for the secondary winding totally shorted, that means no AC
current flow out and in secondary winding, and no termination loss. The AC current direction is
shown in solid line in Fig. 4.19. The AC resistance is Rsec=1.37mΩ, and the leakage inductance
Lk=36.6nH. Compared with Fig. 4.17, the AC resistance of secondary winding with termination
is 4 times larger than the one without termination.
114
Based on previous analysis, the lossy parts of the first winding arrangement are the terminal
and the vias which connect the secondary windings, SR devices and output capacitors. To
eliminate this termination related loss, the second winding arrangement is proposed whose SR
devices and output capacitors are integrated into secondary windings. For this one, no extra
terminal needed for devices and capacitors connection. The devices and capacitors are naturally
part of secondary side windings. The AC current loop of secondary winding is exactly matching
those of primary side winding. There is no AC current flowing out and in secondary side
windings, thus no extra AC current loss is introduced.
Fig. 4.20 Top view of the second winding arrangement
Fig. 4.21 Cross view of the second winding arrangement
115
The second winding arrangement is proposed in Fig. 4.20 and Fig. 4.21. The secondary
windings locate top and bottom side of PCB board. The SR MOSFETs and output capacitors are
part of secondary windings. In Fig. 4.21, the MMF between windings is higher than that of the
first winding arrangement in Fig. 4.15, which means this winding structure has worse
interleaving.
Fig. 4.22 FEA simulation for the second winding arrangement
116
The FEA simulation results are given in Fig. 4.22. The secondary side AC resistance is
Rsec=2.08mΩ, and the leakage inductance Lk = 63.4nH.Although the first winding arrangement
has better interleaving, its AC resistance and leakage inductance are much higher than the second
one. That means the termination loss is much larger than the loss saved by better interleaving
structure.
In the first winding arrangement, SR devices connect secondary side winding byvias and
external connection, which cause huge AC conduction loss. The second reason is that its primary
winding and secondary winding are not exactly match each other. When the windings are not
overlapped very well, the leakage flux induced by those parts will force the current crowded to
the edge of windings. In the second winding arrangement, the SR devices are integrated into
secondary windings, thus, there is no termination loss. The primary and secondary windings are
exactly overlapped, thus, there is nearly no current crowding anywhere in this case as shown in
Fig. 4.22(a). Only little crowding current is induced by SR device package.
4.4 Matrix Transformer Design Procedure
At the beginning of detail transformer design, the transformer width is constrained by the
dimension of synchronous rectifier devices. For this case, we choose Infineon device SuperSO8
package, the minimal width of whole transformer is 70mm. Then, 1 MHz is picked as the
switching frequency. The core material 3F45 is chosen based on performance factor [B.25]. Then,
the loss density is shown in Fig. 4.23. As the cross area increases, the core loss drops
dramatically. The curve is reach diminishing returns around shadow area. One point (Ae=44mm2)
is chosen in shadow area based on the tradeoff of core size and loss.
117
Fig. 4.23 Loss density vs. Cross area.
Fig. 4.24 Core and winding dimensions
As shown in Fig. 4.24, the core shape does not only influence the core loss but also winding
loss. Because the transformer width is determined, we can only change the a and b value of core
to get the minimal winding loss. Fig. 4.25 and Fig. 4.26 show the trend of AC and DC resistance
118
of secondary winding and primary winding, which changes as a and b ratio changes. At last, we
can choose the a and b values based on minimal resistance.
Fig. 4.25 AC & DC resistance of Sec winding.
Fig. 4.26 AC & DC resistance of Pri winding.
119
The last step is chosen suitable winding thickness. The simulation results are given in Table
2.1. The skin depth at 1 MHz is around 66um. However, the results show the optimal thickness is
around 3 oz copper. 4 oz copper is chosen for better efficiency.
Table 4.1The AC resistance of pri and sec windings vs. thickness
4.5 Experimental Results
An 1MHz 390V/12V 1kW LLC resonant converter prototype with proposed matrix
transformer structure is built. The second winding arrangement is applied. The circuit is given in
Fig. 4.27, while the hardware picture is shown in Fig. 4.28.
Fig. 4.27The circuit of 1kW, 400V/12V LLC resonant converter prototype
120
Fig. 4.28The prototype of 1kW, 400V/12V LLC resonant converter
The primary devices are Transphorm GaN HEMTs. For better efficiency, two GaN devices
is paralleled. Thus, there are four devices at primary side. SR devices are BSC010N04LSI from
Infineon. The core material is 3F45.The magnetizing and leakage inductances are used as
resonant elements, which are integrated into matrix transformer. The leakage inductance of
matrix transformer is 220nH, while the magnetizing inductance is 17.6uH. The resonant
capacitor is70nF.The power density of power stage is around 830W/in3.
121
Fig. 4.29Experimental waveforms for 15% load condition
Fig. 4.30Experimental waveforms for 50% load condition
122
Fig. 4.31 Experimental waveforms for 100% load condition
Fig. 4.29, Fig. 4.30 and Fig. 4.31show the waveforms under 15%, 50% and 100% load.
Vds_Q2 is the drain-source voltage waveform of primary device Q2. ICr is the current through
resonant capacitor Cr. Even as high as 1 MHz switching frequency, the primary side devices can
easy achieve Zero-Voltage-Switching (ZVS). The oscillation during dead time is caused by the
leakage inductance and secondary rectifier’s junction capacitance. The turnoff current is as low
as 2A.
One of the best industry products on market is Vicor module, whose power density is around
1050W/in3. To achieve the efficiency and power density, it has 14 layer PCB board, 12
MOSFETs for 300W output power. Compared with this product, our prototype just has 4 layer
PCB and 12 devices for 1kW to reach similar power density and efficiency. The prototype
efficiency can reach as high as 95.4%. The efficiency comparison is shown in Fig. 4.32.
123
Fig. 4.32The efficiency comparison for CPES prototype and industry product
One more thing needs to clarify is that the advantage of GaN devices is smaller Coss
compared with Si MOSFET having the same Rdson. That will result in lower turn off current
which help achieve zero-voltage switching. Lower turn off current means less circulating energy
in the circuit. Thus, the major loss reduction parts caused by GaN device are the conduction loss
and driving loss of primary devices, the conduction loss of primary winding of transformer.
While the major loss saved by new transformer structure is secondary side winding conduction
loss and termination loss. The detail analysis for GaN device benefits is described in [D.20].
Table 4.2 Loss breakdown of prototype under 100% load
Loss Value (W)

Primary MOSFET Conduction Loss 8.2
Primary MOSFET Turnoff Loss 2.2
Primary MOSFET Driving Loss 0.6
Resonant Capacitor (Cr) Loss 0.46
Transformer Core Loss 5.10
Transformer Primary Winding Loss 4.49
Transformer Secondary Winding Loss 7.24
124
Secondary MOS Conduction Loss 6.6

Secondary MOS Driving Loss 3.6
Secondary MOS Body Diode Loss 2.53
Output Capacitor Loss 1.95
Connection, Trace & Solder Loss 3.56
Calculated Total Loss 46.53
Measured Total Loss 51
Table 4.3 Loss breakdown of prototype under 15% load
Loss Value (W)

Primary MOSFET Conduction Loss 0.27
Primary MOSFET Turnoff Loss 2.2
Primary MOSFET Driving Loss 0.6
Resonant Capacitor (Cr) Loss 0.015
Transformer Core Loss 5.10
Transformer Primary Winding Loss 0.15
Transformer Secondary Winding Loss 0
Secondary MOS Conduction Loss 0
Secondary MOS Driving Loss 3.6
Secondary MOS Body Diode Loss 0
Output Capacitor Loss 0
Connection, Trace & Solder Loss 0.24
Calculated Total Loss 12.17
Measured Total Loss 15.75
Based on previous Maxwell 3D simulation, the new winding structure can save more than
60% secondary side winding loss under full load. More than 20% of total loss is reduced. The
detail loss breakdown for 100% load and 15% load are given in Table 4.2 and Table 4.3. It is
clear to see the major loss is distributed around transformer and secondary rectifier devices at
full load.
4.6 Conclusion
125
To deal with the high output current applications, matrix transformer structure help reduce
the leakage inductance and winding AC resistance of transformer. The price is the rising of cores
loss and volume due to multiple magnetic cores. Flux cancellation reduces core size and core
loss. Then, SR devices and output capacitors are integrated into secondary side windings to
eliminate the termination loss. A high efficiency high power density matrix transformer structure
based on previous discussion is proposed.
126
Chapter 5. Design Considerations for Passive Integration

In this chapter, the multi-elements resonant converters’ characteristics are reviewed. The
voltage and current stresses of all resonant elements are examined. Based on stress, the principles
of choosing suitable passive components for integration are proposed. Then, the magnetic
integration method is shown based on multi-winding transformer structure. The basic principles
of passive integration are discussed. A novel passive integration method is proposed for multi-
elements resonant converters. A 1000 W, 800 kHz, 400 V/48 V test board for resonant tanks is
built to verify the proposed method.
5.1 Introduction
To make practical use of LLC and CLL resonant converters, the over current protection has
to be fulfilled, especially for startup and over load conditions. The purpose of over current
protection is to limit the current stress in the system during over load or short circuit condition,
also to limit the inrush current during start up so that the power converter can be protected from
destructive damage. Multi-element resonant converters were proposed to solve this issue [E.1]-
[E.2]with more passive components added to the resonant tank. However, the additional passive
components impose the demand for integration efforts in order to alleviate their impact on
converter size.
At a high switching frequency, the passive components’ volume can be reduced. However,
in the state-of-the-art resonant converters, the discrete passive components and the heat sink still
occupy more than 80% of the total system volume [E.3]. To reduce the volume of the magnetic
127
components, magnetic integration has been proposed. However, the passive components still
have large profile. To further reduce the overall volume, passive integration technology is
applied to integrate the resonant inductor (Lr) and capacitor (Cr), the magnetizing inductor (Lm),
and the transformer (Tx) (L-L-C-T) into one module [E.4]-[E.10]. There is still a challenge to
employ passive integration techniques in multi-elements resonant converters. This paper
proposes a novel passive integration method to solve this problem.
5.2 Basic Characteristics of Multi-elements resonant converters
The LLC resonant tank can be considered as a band pass filter which includes series
resonant frequency (SRF) and parallel resonant frequency (PRF), as shown in Fig. 5.1 and Fig. 5.2.
The series resonant frequency is created by Lr and Cr, black dot in Fig. 5.2. It is the optimal point
at nominal operation. The fundamental component of input energy can be delivered from the
source to the load very efficiently. Therefore, high efficiency can be achieved. The hold-up time
capability is given by the parallel resonant frequency, dash line oval in Fig. 5.2. This resonant
frequency is determined by Lr, Lm and Cr.
Fig. 5.1LLC resonant converter
128
Fig. 5.2Circuit and voltage gains of LLC resonant converter
However, with poor frequency selectivity, namely, very wide bandwidth, the frequency has
to be increased very high to achieve enough damping effect. To improve the characteristics of
LLC resonant converters, additional resonant elements are needed. With additional resonant
element component, a multi-element resonant tank, which combines band pass filter and notch
filter, is proposed to solve the current protection issues. One of these resonant converters is
depicted in Fig. 5.3. The notch filter section is created with Lp and Cp. The band pass filter section
consists of Lp, Cr and Cp.
Fig. 5.3One example of selected 4-elements resonant converters
129
Fig. 5.4Gain Curves of selected 4-elements resonant converter
The voltage gain of this four-element resonant tank is illustrated in Fig. 5.4. The resonant
tank with notch filter characteristic can create zero voltage gain. Conceptually, infinite
impedance can be created when frequency operates at the resonant frequency f02 of the notch
filter. As a result, if overload or even shorted output condition occurs, the current can be limited
inherently by the resonant tank characteristic. The notch filter center frequency f02 can be
designed to achieve the over load protection.
In general, with variable frequency control, the resonant tank input voltage is excited as a
square waveform. According to the Fourier analysis, the voltage excitation only consists of odd
harmonics. Therefore, to avoid circulating energy at nominal condition, notch filter resonant
frequency f02should be designed at even times of nominal resonantfrequencyf01. Meanwhile, f02 is
preferred close to f01.Thus, f02 = 2f01 is a good choice.
130
As shown in Fig. 5.5, the voltage and current stresses of LLC and proposed 4-element
resonant converter are very similar.
Fig. 5.5 V-I stresses of LLC and 4-element resonant converters
When converter operates at f01, the notch filter represents inductive. The impedance of Cp is
much larger than Lp. The major part of primary side current flows through Lp, not Cp. The current
stress on Cp is much less than Cr.
With leakage inductance Lr, a second band pass filter is created. A 5-elements resonant tank
is given as an example, shown in Fig. 5.6. The second band pass filter is designed at the three
times of the nominal frequency. The third order harmonics energy of primary side can be
delivered through this band pass filter. Obviously, the power delivery is more efficient.
However, there are 6 passive components in resonant tank. The high number of passive
components might increase the size considerably. It is very desirable to integrate the passive
components. Hence, higher power density can be achieved.
131
Fig. 5.6One example of 5-element resonant converters
Fig. 5.7Gain Curves of selected 5-elements resonant converter
5.3 Magnetic integration for multi-elements resonant converters
To improve the power density, the magnetic integration method is widely applied in
resonant converters. For LLC resonant converters, Lr is the leakage inductance, and Lm is the
magnetizing inductance, as shown inFig. 5.1.
132
For 5-elements resonant converters, there are four magnetic components: Lr, Lp, Lm, and
transformer. It is hard to integrate all components into one module by traditional transformer
structure. The transformer structure needs to be modified.
In this section, a different transformer structure is proposed to integrate all magnetic
components into one.
As mentioned in [E.9], cantilever model is suited for multiple-winding magnetic model. Fig.
5.8(a) gives the cantilever model for LLC center-tap transformer. Winding 1 is the primary side
winding, and winding 2, 3 are two secondary side windings. The effective leakage inductance l12,
l13, l23 represent the couple relationship between winding 1, 2, and 3. n2, n3 represent the
effective turns ratio of winding 2, 3 with winding 1.
Normally, the two secondary side windings are very close, and have same turns (n2=n3),
thus, l23 is negligible. Thus, l12, l13 can be combined as l123, as shown in Fig. 5.8 (b). Then, this
model can be transferred into equivalent model in Fig. 5.8(c).Here, L112/ (L11+l123)represents the
magnetizing inductance (Lm), sodoes (L11×l123)/ (L11+l123) the leakage inductance (Lr).
(a) Cantilever model for LLC transformer
133
(b) Simplified model for LLC transformer
L 11 l123  l 
L 11 + l 123 1 : n 2  1 + 123 
 L 11 
L 112
1 L 11 + l 123 2
 l 
1 : n 3  1 + 123 
 L 11 
(c) Equivalent model for LLC transformer
Fig. 5.8Cantilever model for LLC transformer
With traditional transformer structure, only series resonant inductor and magnetizing
inductor can be integrated. To integrate the third resonant inductor, one more winding (winding 4)
is added into transformer.
134
The effective leakage inductance l14, l24, l34 represent the couple relationship between
winding 1, 2, 3 and 4. n4 represent the turns ratio of winding 4 with winding 1. The cantilever
model is shown in Fig. 5.9.
Fig. 5.9Cantilever model for proposed transformer
As mentioned above, l23 is small enough to be negligible, and l12, l13 can be combined as l123.
If winding 4 is not coupled tightly with winding 2 and 3, l42 and l42 would be large enough to be
ignored. The equivalent circuit is in Fig. 5.10.
Fig. 5.10Simplified model for proposed transformer
135
The resonant tank connection is shown in Fig. 5.11 (a). The dash lines are the connecting
wires among integration module and discrete capacitors Cr and Cp. The winding 1 and 4 share
the same ground. When n4=1, the node A and B have the same electrical potential, so they can
be treated as virtually shorted as in Fig. 5.11 (b). Thus, the final resonant tank is Fig. 5.11 (c). The
leakage inductance between winding 1 and 4 (l14) creates Lp.
(a) Connection for discrete Cr, Cp and integrated magnetics
(b) Equivalent circuit when n4=1
136
(c) Modified equivalent circuit when n4=1
Fig. 5.11 Equivalent circuit for discrete Cr, Cp and integrated magnetics
The gain curves for this multi-element resonant converter are shown in Fig. 5.12, which is
very similar as Fig. 5.9. Namely, superior merits, full load range ZVS, good frequency
controllability for current limit and voltage gain, can be successfully obtained.
Fig. 5.12The gain curves of passive integration model
137
Each parameter of the cantilever model can be directly measured. All magnetic components
in resonant tank are integrated in one transformer module.
For simplification, an easy way to understanding the integration method is based on the
magnetic relationship between windings. Fig. 5.13 (a) shows the integrated magnetic module
which includes transformer (Tx), magnetizing inductance (Lm), leakage inductance (Lr) and Lp.
The equivalent circuit is given in Fig. 5.13 (b). Lr is the leakage inductance indicating the
magnetic decoupling of winding 2, 3 and winding 1, 4. Lp is the leakage inductance indicating
the magnetic decoupling of winding 1 and winding 4. Magnetizing inductance (Lm) is determined
by the air gap.
(a) Crossview of integrated magnetics (b) Equivalent circuit
Fig. 5.13The realization of magnetic integration module
138
The turns of winding 1 and winding 4 is the same, and these two windings are connected at
one terminal. Thus, we can transform Fig. 5.13 (b) to Fig. 5.14 (a). While Lr is determined by the
distance between primary winding (1,4) and secondary winding (2, 3), so Fig. 5.14 (a) can be
transferred into Fig. 5.14 (b), which is similar as the resonant tank in Fig. 5.6.
(a) Transformation 1 (b) Transformation 2
Fig. 5.14The transformation of equivalent circuit
Fig. 5.15 Volume comparison of proposed and traditional magnetic integration method
139
One magnetic module is built to verify the volume improvement. The previous magnetic
integration method utilizing magnetizing inductance and leakage inductance of transformer, but
cannot integrate Lp. The proposed method can integrate all the magnetic components. The
previous magnetic components volume is 39.73cm3. The new magnetic module is 16.23cm3.
The experimental waveforms are provided in Fig. 5.16 (a). It is clear that the fundamental go
through winding 1 and the 3rd harmonic through winding 4. For fundamental, the impedance of
Cp and Lr is very high, while Lp and Lr is relative low. Thus, major part of fundamental flew
through winding 1. For 3rd harmonic, the impedance of Cp and Lr is very low, while Lp and Lr is
relative high. Thus, major part of 3rd harmonic flew through winding 4.
(a) Experiment results for winding stresses (b) Equivalent circuit
Fig. 5.16 Stress comparison for integrated magnetic module
5.4 Passive Integration for Multi-element Resonant Converters
In order to further improve the power density, passive integrationincluding the resonant
capacitors is required.
140
Firstly, there are two resonant capacitors (Cr and Cp) in the resonant tank. Which capacitor
integrated into module should

ld be chosen carefully. Voltage and current stresses are very critical
to capacitor’s lifetime. High voltage stress or high current stress will cause short life time of
capacitors, in other words, short life of whole converter. Thus, low stress resonant capacitor
ca
should be picked.
Based on the analysis of Fig. 5.16, we can see all current including fundamental and 3rd
harmonic flow through Cr, only small 3rd harmonic through Cp. For half bridge structure, there is
half DC input voltage

ltage cross Cr, while no DC voltage bias on Cp. It is obvious that Cp has much
less stress than Cr. Thus, Cp should be taken into integrated module.
(a) Physical structure of LC cell (b) Lossless symmetric model
A D
C (L+M)/2
(c) 1st order model for fundamental design
Fig. 5.17LC cell structure and equivalent circuit
141
LC cell is the basic element in passive integration. Its structure is shown in Fig. 5.17(a). The
dielectric layer is covered by copper layers at the both side. The lossless symmetric model of this
structure is given in Fig. 5.17(b). C is the capacitance of LC cell, while L is the self-inductance of
one side copper strip and M is the mutual inductance. Its equivalent circuit measured between
terminal A and D is a series of inductor and capacitor, as shown in Fig. 5.17(c).The value of
inductance and capacitance can be controlled by cascading LC cells, as shown in Fig. 5.18. The
equivalent capacitors are in parallel and the equivalent inductors are in series.
Fig. 5.18Interconnection of LC cells
The capacitance of each LC cell can be calculated by following formula.
ε 0 ⋅ ε r ⋅ Acu
C= (5.1)
Tdi
C is the capacitance. Acu is the area of each plate. εr is the relative static permittivity
(dielectric constant) of the material between the plates. ε0 is the permittivity of free space. Tdi is
the separation between the plates.
The passive integration with a stacked structure for resonant converters has been proposed to
integrate the LLCT module to achieve high power density. Fig. 5.19gives the winding structure of
LLCT module. The exploded view of an LLCT module with the stacked structure is shown in Fig.
5.20. Two sets of C core and I core constitute the two winding windows. The secondary winding
142
is made of thin copper strips. The primary winding is implemented with cascading LC cells in
order to integrate the resonant capacitor in to the transformer. The cells are connected at the
terminals through the interconnections on the vertical surface at the two ends of the module as
shown in Fig. 5.20. The cell dimensions and the properties of the dielectric substrate determine
the resonant capacitance (Cr). The magnetizing inductance (Lm) can be adjusted by changing the
length of the air gap. By inserting leakage layer between the primary and secondary windings,
the leakage inductance (Lr) can also be controlled.
Fig. 5.19Interconnections and winding structure of an integrated LLCT module
For the 5-element resonant tank, it is difficult to integrate both capacitors. The choice of
capacitor to be integrated is determined by the consideration of reliability. Large voltage and
current stress will cause short life time of capacitor. In this particular topology, Cr has very high
143
DC voltage bias while Cp only has small AC voltage. Moreover, Cr carries all the primary
current while Cp only carries the third order harmonics as discussed before. Thus, Cr has much
higher voltage and current stress than Cp, in other words, less reliable. Thus, Cp is chosen to be
integrated into the transformer.
Fig. 5.20Exploded view of an LLCT module with the stacked structure
The integrated transformer structure for Lp,Lr,Lm,Cp, and transformer is similar to the
integration technique for magnetic integration except that one winding (winding 1) is replaced by
LC cell, the connection is shown in Fig. 5.21(a). It is a 4-winding structure and the cantilever
model is introduced to derive its equivalent circuit. Winding 1 and 4 are primary windings
constructed by LC cell and copper wire, respectively. The LC cell interconnection is like Fig. 14.
That means the equivalent capacitor of LC cell is in series of parasitic inductance of winding 1,
like Fig. 5.21(b). Winding 2 and 3 are secondary side windings. L11 is the self-inductance of
144
winding 1. The effective leakage inductance l12, l13, l14, l23, l24, l34 represent the couple
relationship between winding 1, 2, 3 and 4. n2, n3, n4 represent the turns ratio of winding 2, 3, 4
with winding 1, as shown in Fig. 5.21(c).
(a) (b)
(c)
Fig. 5.21Winding model for 4-winding transformer
Normally, the two secondary side windings are very close, and have same turns (n2=n3),
thus, l23 is negligible. Thus, l12, l13 can be combined as l123, so can l24, l34 as l423, as shown in Fig.
5.22. The dash lines represent connection wire between integration module and Cr. If winding 4
is not coupled tightly with winding 2 and 3, l423 would be large enough to be ignored. Then, it is
145
the same topology as in Fig. 5.12 (a). Thus, it also can be modified as the resonant tank in Fig.
5.12(c). They are also have the same gain curves.
Fig. 5.22The passive integration model
Thus, all passive components are integrated except Cr (the high stress component). All the
leakage inductance can be tuned by adjusting the distance between windings or with magnetic
shunt. L11 can be controlled by air gap.
The transformer structure in Fig. 5.23is given for example. Cp is determined by LC cell in
winding 1. L11 is controlled by air gap. l14 (Lp) is controlled by magnetic shunt between winding
1 and 4. l123 (Lr) is controlled by magnetic shunt between winding 1 and 2. Winding 4 can be
either round copper wire or copper strips.
146
Fig. 5.23Transformer structure for integration
5.5 Measurement of Transformer parameters
Different with the transformer module before, passive integrated transformer including LC
cell is not suited to use the same measure method. The capacitance of LC cells makes the
measurement method invalid. A new measure method is proposed based on cantilever model.
The transformer model of proposed passive integration structure is shown in Fig. 5.24.
147
Fig. 5.24Transformer model of proposed structure
First, the self-inductance L11and capacitance C are measured based on impedance analyzer
(Agilent 4294A), by open-circuiting windings 1, 2 and 3.
To measure the effective turns ratios n2, n3,...,nk, and AC voltage V1is applied to winding 1,
with the other windings open circuited. The effective turns ratio nk, is given by
1
+ j ω L11
V k jω C V ω 2 L11 C − 1
nk = = k
V1 jω L11 V 1 ω 2 L11 C (5.2)
Vk is the measured voltage of winding k. ω is the angular frequency of AC voltage applied to
winding 1. A negative value of nk indicates that the winding polarity marks should be reversed.
Direct measurement is hard for this transformer structure due to LC cell. All the inductance
can be calculated based on impedance measurement. Six measurement cases are needed for
effective leakage inductance calculation.
148
(a) Case 1 (b) Case 2
(c) Case 3 (d) Case 4
(e) Case 5 (f) Case 6
Fig. 5.25. Equivalent circuits for 6 cases
Case 1 is measuring the impedance of winding 1 port by shorting winding 2, 3 and 4. The
equivalent circuit is shown in Fig. 5.25(a).
Case 2 is measuring the impedance of winding 1 port by shorting winding 2, 3 and opening
winding 4. The equivalent circuit is shown in Fig. 5.25 (b).
149
winding 2. The equivalent circuit is shown in Fig. 5.25 (c).
winding 3. The equivalent circuit is shown in Fig. 5.25 (d).
winding 1. The equivalent circuit is shown in Fig. 5.25 (f).
Based on above 6 cases, all effective leakage inductances can be calculated. The calculated
results based on real integrated module are shown in Fig. 5.26.
Fig. 5.26Calculated results of integrated transformer module
The final parameters of integrated module after simplified are shown in Fig. 5.27.
150
Fig. 5.27 The final parameters of integrated transformer module
The volume comparison is given inFig. 5.28. The volume of previous magnetic components
and Cp is 40.48cm3. The passive integration module is 16.79cm3.
Fig. 5.28 The volume comparison of passive module and discrete components
151
5.6 Thermal Improvement by Heat Extractor
Associated with the increasing power density, loss density increases. Although the loss
density in the passive component is much lower than that in power semiconductor devices, the
lower thermal conductivities of most materials applied in passive components leads to high
temperature gradient inside the component. The thermal issues limit the increase of power
density. The maximum allowed temperature inside an integrated module is one of the most
important design criteria.
Although the surface temperature of a passive module can be controlled by the external
cooling conditions, the temperature rise inside the module is determined by the loss distribution,
the material properties and the structure. For a practical design, the structure is basically
determined by the considerations of electromagnetic function, and the materials are generally
selected for their electromagnetic characteristics to meet the required parameter values. When
the structure and materials are fixed, the loss distribution is determined by the circuit operation,
which is normally also fixed. Thus, itis difficult to improve the thermal performance of an
existing design without changing the structure or materials. The most common approach to
reduce the inside temperature gradient is to increase the dimensions, which reduces the loss
density and increase the heat dissipation area, however, sacrifices the power density. It is clear
that in cases where the thermal constraint is the dominant design criterion, the heat dissipating
ability of a module is directly related to the power density. In other words, by improving the heat
dissipating ability, the power density can be improved. Being aware of this, the heat extraction
technology is developed[E.12].
152
To apply heat extractors in power electronics module, the heat extractor plate should be
parallel to the magnetic flux to avoid influencing the electromagnetic performance. The structure
shown in Fig. 5.29can embed heat extractors without change the electromagnetic performance,
but the inside filed distribution changes due to the redistribution of the magnetic flux. The
induced fringing effect may even influence the conduction loss. If the heat extractors occupy a
small relative volume, the change of the magnetic field is neglected.
Fig. 5.29Integrated passive module with heat extractors
The embedding of heat extractors also changes the distribution of the thermal field.
Originally regarded as a uniform heat source, the core is now several individual heat sources
with higher heat generation.
In the integrated module, there are two heat sources: the winding block and the magnetic
core. By applying the superposition principle, the amounts of heat flux generated by these two
heat sources can be studied separately, and the results can be combined linearly. For the heat
generated inside the winding block, both the core and heat extractors are conduction channels,
through which heat can be conducted to the heat sink. The internal and external thermal contact
153
resistance and the equivalent thermal resistance of the combination of the core and heat
extractors are the barriers on this heat path.
For heat generated in the core, the winding block functions as a conductor connecting the
core and heat extractors. The internal thermal contact resistance functions twice on one of the
heat paths. Because of the different heat distribution, the equivalent thermal resistance of the
combination of core and heat extractors is different from that for heat coming from inside the
winding block.
The equivalent structures of both cases are shown in Fig. 5.30, where the core and the
winding block are replaced by equivalent spot heat sources, and the corresponding equivalent
thermal resistance is applied. Here, the thermal conductivity of core material (MnZn ferrites) is
about (3.5~5.0) x 10-3K·m·W−1, while the thermal conductivity of heat extractor (aluminum
nitride), is about 170~190 K·m·W−1. The thermal conductivity of heat extractor is much larger
than core material. Thus, large portion of heat generated by windings and cores would pass
through heat extractor to the heatsink.
Fig. 5.30 Diagram of heat paths on cross section
154
The core dimension is given by Fig. 5.31, while the heat extractors are made of 0.5mm
aluminum nitride substrates and have the same cross-sectional dimensions with the core. The
core material is 3F4. The whole core is combined by 9 pieces of 3F4 ferrite and 8 pieces of heat
extractor. The real picture is shown in Fig. 5.32.
Fig. 5.31Core dimensions
Fig. 5.32Picture of cores and heat extractors
155
The volume comparison after adding heat extractor is given in Fig. 5.33.
Fig. 5.33 The volume comparison of passive module and discrete components
5.7 Experimental Results
An 800 kHz 1.2kW 400V/48V five-element integration prototype is shown in Fig. 5.34 (a).
The parameters in resonant tank are Cr=12nF, Cp=5nF, Lp=1.2uH, Lr=0.9uH, Lm=10uH. The
turns ratio is 4:1:1. The no load and 50% load waveforms are shown in Fig. 5.34 (b), (c). The
waveforms for secondary side short-circuit is given in Fig. 5.34 (d). The output current is
suppressed to 20A at 1.6 times of resonant frequency, which is less than full load current. The
power density is nearly 171W/inch3.
156
400V No load
Vds
Vgs
0V
4A
0A
-4A icr
(a) Power stage (b) No load waveform
400V 50% load

Vds Vds
Vgs 8A
0V 0V
4A 0A
0A
-4A iPri -8A
iPri
(c) 50% load waveform (d) Short-circuit waveform
Fig. 5.34Experimental results for five-element resonant converter with passive integration module
5.8 Conclusion
A novel passive integration method is proposed for multi-element resonant converters.
Based on this method, the passive integration for 3-element, or higher order elements resonant
converters are achievable. Most of passive components (capacitors, inductors and transformer)
are integrated into one module to achieve high power density. The proposed multi-winding
transformer structure with integrated inductor and capacitor are analyzed with a detailed
157
cantilever model. Design guidelines are summarized for the control of the component values.
One prototype of 1.2kW, 800 kHz 400V/48V 5-element resonant converter is built and tested to
verify the proposed structure. The power density can be achieved as high as 171W/inch3.
158
Chapter 6. Conclusion and Future Work
6.1 Conclusion
The LLC resonant converter exhibits higher efficiency and higher power density than
conventional PWM converters. Thus, it has been widely adopted as a desirable power supply for
telecoms, servers, desktops, laptops, flat-panel TVs, LED lighting, etc. To further improve the
LLC performance, several novel concepts and techniques have been developed in this
dissertation.
The high switching frequency is needed recently for high power density requirement.
However, LLC resonant converter suffers high transformer loss. Matrix transformer is
introduced to reduce winding loss and total volume. Flux cancellation method is utilized to
reduce core size and loss. Synchronous Rectifier (SR) devices and output capacitors are
integrated into secondary windings to eliminate termination related winding losses, via loss and
reduce leakage inductance.
To solve the inherent issues, like start-up and short-circuit protection, SR driving and EMI,
etc., a systematic method is proposed to find the similar resonant topologies like LLC. Based on
this method, lots of high order resonant converters are found to solve the issues LLC type
resonant converters cannot handle.
To evaluate the performance of found valuable topologies, a general evaluation system is
proposed. State-plane analysis with new normalization factors is utilized. Based on it, the voltage
159
stress, current stresses and apparent power of resonant converters are easy to compare. This
method can help select suitable circuit topology for certain application. Meanwhile, it also can
help resonant converters’ design. The important performance factors, like start-up, short-circuit
protection, SR driving and EMI, are also taken into account for the whole evaluation system.
The passive integration is necessary for high power density resonant converter, especially
for high order system. Based on stress, suitable passive components are chosen for integration.
Then, the magnetic integration method is shown based on multi-winding transformer structure.
The passive integration principles are discussed. A novel passive integration method is proposed
for multi-elements resonant converters.
In conclusion, this work is focus on the topology analysis and integration of resonant
converters. Searching the suitable topologies for certain application, and evaluate the
performance of them. Then, improve the system power density by integration techniques.
6.2 Future Work
All of these novel implementations proposed in this work will help find valuable resonant
converters and evaluate their performance. Following this research, there are still some
remaining works can be done:
1. The proposed matrix transformer is high power density and high efficiency, but its EMI
performance is unknown. The continuous work may study the EMI noise reduction for matrix
transformer.
160
2. By passive integration, the power density of five-element resonant converter is improved
a lot. If passive integration techniques cooperates with matrix transformer, how this impact the
efficiency and power density of whole system is interesting for future study.
3. The state-plane with new normalization can be extent to not only resonant converters, but
also can be used for PWM converters.
161
Daocheng Huang References
References
A. Introduction
[A.1] W. A. Tabisz, M. M. Jovanovic, and F. C. Lee, "Present and future of distribution power
systems", in Proc. APEC 1992, pp. 11-18.
[A.2] G. C. Hua, W. A., Tabisz, C. S. Leu, N. Dai, R. Watson, F. C. Lee, “Development Of a
dc distributed power system,” in Proc. IEEE APEC 1994, vol.2, pp. 763 -769.
[A.3] F.C. Lee, P. Barbosa, P. Xu, J. Zhang, B. Yang, F. Canales, “Topologies and design
considerations for distributed power system applications,” in Proc. of the IEEE, Volume 89
Issue: 6, June 2001, pp: 93 –950.
[A.4] S. K. Mazumder, K. Shenai, “On the reliability of distributed power systems: a macro- to
micro- level overview using a parallel DC-DC converter,” in Proc. IEEE PESC, 2002, pp.
809-814.
[A.5] M. M. Jovanovic, "Technology drivers and trends in power supplies for
computer/telecom," APEC 2006, Plenary session presentation.
[A.6] A. Pratt, P. Kumar, T. V. Aldridge, “Evaluation of 400V DC distribution in telco and data
centers to improve energy efficiency,” in Proc. IEEE INTELEC, 2007, pp. 32-39.
[A.7] M. Ton, B. Fortenbery, and W. Tschudi, “DC power for improved data center
efficiency," [Online]. Available: http://hightech.lbl.gov/dc-powering/about.html, Jan. 2007.
[A.8] D. J. Becker, "400Vdc power distribution: Overcoming the challenges," in Proc. IEEE
INTELEC, 2010, pp. 1-10.
[A.9] 80Plus Program [Online]. Available: http://www.80plus.org.
[A.10] Energy Start Program [Online]. Available: http://www.energystar.gov, 2012.
[A.11] Climate Savers Computing Initiative [Online]. Available: http://www.climatesavers
computing.org, 2012.
[A.12] Texas Instruments User’s Guide (Nov. 2009). UCC28810EVM-003 110W multiple string
162
LED driver with universal line input and PFC [Online]. Available: http://www.ti.com
/lit/ug/sluu380b/ sluu380b.pdf.
[A.13] H. Chiu, Y. Lo, J. Chen, S. Cheng, “A high-efficiency dimmable LED driver for low-
power lighting applications,” IEEE Trans. on Industrial Electronics, vol. 57, Issue 2, pp. 735-
1743, Feb. 2010.
[A.14] Q. Hu, R. Zane, “LED driver circuit with series-input-connected converter cells operating
in continuous conduction mode,” IEEE Trans. on Power Electronics, vol. 25, Issue 3, pp.
574-582, Mar. 2010.
[A.15] H. Wu, S. Ji, F. C. Lee, X. Wu, “Multi-channel constant current (MC3) LLC resonant
LED driver,” in Proc. IEEE ECCE’11, pp. 2568- 2575, 2011.
[A.16] D. Reusch, F. C. Lee, D. Gilham, Y. Su, "Optimization of a high density gallium nitride
based non-isolated point of load module", in Proc. IEEE ECCE 2012, pp. 2914-2920.
[A.17] X. Huang, Z. Liu, Q. Li, F.C. Lee, "Evaluation and application of 600V GaN HEMT in
cascode structure," in Proc. IEEE APEC 2013.
[A.18] S. Ji, D. Reusch, F. C. Lee, "High-frequency high power density 3-D integrated gallium-
nitride based point of load module design," IEEE Trans. on Power Electronics, vol. 28, Issue
9, pp. 4216-4226, Sep. 2013.
[A.19] Y. Su, Q. Li, F. C. Lee, "Design and evaluation of a high-frequency LTCC inductor
substrate for a three-dimensional integrated DC/DC converter," IEEE Trans. on Power
Electronics, vol. 28, Issue 9, pp. 4354-4355, Sep. 2013.
[A.20] W. Chen, F.C. Lee, M.M. Jovanovic, J.A. Sabate, “A Comparative Study of A Class of
Full Bridge Zero Voltage Switched PWM Converters,” Proc. IEEE APEC, 1995, pp. 893-899.
[A.21] A.W. Lotfi, Q. Chen, F.C. Lee, “A Nonlinear Optimization Tool For The Full Bridge
Zero Voltage Switched DC-DC Converter,” Proc. IEEE PESC ‘92, 1992, pp. 1301-1309.
[A.22] Xinbo Ruan; Yangguang Yan, “Soft-Switching Techniques for PWM Full Bridge
Converters,” Proc. IEEE PESC ’00, 2000, pp. 634 -639 vol.2.
[A.23] J.G. Cho, J.A. Sabate, G. Hua, and F.C. Lee, “Zero Voltage and Zero Current Switching
Full Bridge PWM Converter for High Power Applications,” Proc. IEEE PESC ’94, 1994.
163
[A.24] J.H. Liang, Po-chueh Wang, Kuo-chien Huang, Cern-Lin Chen, Yi-Hsin Leu, Tsuo-Min
chen, “ Design Optimization For Asymmetrical Half Bridge Converters,” Proc. IEEE
APEC ’01, 2001, pp. 697-702, vol.2.
[A.25] Y. Leu, C. Chen, “Analysis and Design of Two-Transformer Asymmetrical Half-Bridge
Converter,” Proc. IEEE PESC ’02, 2002, 943-948.
[A.26] S. Korotkov, V. Meleshin, R. Miftahutdinov, S. Fraidlin, “Soft Switched Asymmetrical
Half Bridge DC/DC Converter: Steady State Analysis. Analysis of Switching Process,” Proc.
Telescon ’97,. 1997, pp. 177-184.
[A.27] L. Krupskiy, V. Meleshine, A. Nemchinov, “Unified Model of Asymmetrical Half Bridge
for Three Important Topological Variations,” Proc. IEEE INTELEC ’99, 1999, pp.8.
[A.28] W. Chen, P. Xu, F.C. Lee, “The Optimization Of Asymmetrical Half Bridge Converter,”
Proc. IEEE APEC ’01, 2001, pp. 603-707, vol.2.
[A.29] Bo Yang, “Topology Investigation for Front End DC-DC Power Conversion for
Distributed Power System," PhD Dissertation, Dept. ECE., Virginia Tech, 2003.
[A.30] Intel Corporation, NEC Corporation, Dell Computer Corporation, Data General a
division of EMC Corporation, Compaq Computer Corporation, Silicon Graphics Inc., and
International Business Machines Corporation, “A server system infrastructure (SSI)
specification for entry chassis power supplies," 1998.
[A.31] R. Oruganti and F.C. Lee, “Effect of Parasitic Losses on the Performance of Series
Resonant Converters,” Proc. IEEE IAS, ’85, 1985.
[A.32] R. Oruganti, J. Yang, and F.C. Lee, “Implementation of Optimal Trajectory Control of
Series Resonant Converters,” Proc. IEEE PESC ’87, 1987.
[A.33] A.K.S. Bhat, “Analysis and Design of a Modified Series Resonant Converter,” IEEE
Trans. on Power Electronics, 1993, pp. 423-430.
[A.34] J.T. Yang and F.C. Lee, “Computer Aided Design and Analysis of Series Resonant
Converters,” Proc. IEEE IAS ’87, 1987.
[A.35] V. Vorperian and S. Cuk, “A Complete DC Analysis of the Series Resonant Converter,”
Proc. IEEE PESC’82, 1982.
164
[A.36] F.S. Tsai, and F.C. Lee, “A Complete DC Characterization of a Constant- Frequency,
Clamped-Mode, Series Resonant Converter,” Proc IEEE PESC’88, 1988.
[A.37] R. Oruganti, J. Yang, and F.C. Lee, “State Plane Analysis of Parallel Resonant
Converters,” Proc. IEEE PESC ’85, 1985.
[A.38] R. Liu, I. Batarseh, C.Q. Lee, “Comparison of Capacitively and Inductively Coupled
Parallel Resonant Converters,” IEEE Trans. On Power Electronics, 1993, pp. 445-454, vol.8,
issue 4.
[A.39] M. Emsermann, “An Approximate Steady State and Small Signal Analysis of the Parallel
Resonant Converter Running Above Resonance,” Proc. Power Electronics and Variable
Speed Drives ’91, 1991, pp. 9-14.
[A.40] Y.G. Kang, A.K. Upadhyay, D. Stephens, “Analysis and Design of a Half Bridge Parallel
Resonant Converter Operating Above Resonance,” Proc. IEEE IAS ’98, 1998, pp. 827-836.
[A.41] A.K.S. Bhat, “Analysis and design of LCL-type series resonant converter,” in Proc. IEEE
INTELEC, 1990, pp: 172 - 178.
[A.42] B. Yang, F.C. Lee, A.J. Zhang and G. Huang, “LLC Resonant Converter For Front End
DC-DC Conversion,” in Proc. IEEE APEC, 2002, vol. 2, pp. 1108-1112.
[A.43] B. Lu, W. Liu, Y. Liang, F. C. Lee, and J. D. van Wyk, “Optimal Design Methodology
for LLC Resonant Converter,” in Proc. IEEE APEC, 2006, pp. 533–538.
[A.44] D. Fu; Y. Liu; F.C. Lee, M. Xu, “A Novel Driving Scheme for Synchronous Rectifiers in
LLC Resonant Converters,” IEEE Trans. on power electron. vol. 24, iss. 5, pp: 1321 – 1329,
May 2009.
[A.45] H. Groot, E. Janssen, R. Pagano, and K. Schetters , “Design of a 1-MHz LLC Resonant
Converter Based on a DSP-Driven SOI Half-Bridge Power MOS Module”, IEEE Trans. on
power electron., vol. 22, no. 6, pp. 2307-2320, Nov. 2007.
[A.46] X. Xie, J. Zhang, C. Zhao, Z. Zhao, and Z. Qian, “Analysis and Optimization of LLC
Resonant Converter with a Novel Over-Current Protection Circuit”, IEEE Trans. on power
electron., vol. 22, no. 2, pp. 435-443, Mar, 2007.
[A.47] Y. Gu, Z. Lu, L. Hang, Z. Qian, and G. Huang, “Three-Level LLC Series Resonant DC-
165
DC Converter,” IEEE Trans. on power electron. vol. 20, no. 4, pp. 781–789, Jul. 2005.
[A.48] K. Yi, G. Moon, “Novel Two-Phase Interleaved LLC Series-Resonant Converter Using a
Phase of the Resonant Capacitor,” IEEE Trans. on Ind. Electron., vol. 56, iss. 5, pp: 1815 –
1819, May 2009.
[A.49] On Semiconductor, "NCP1395LLCGEVB: 240W LLC evaluation board," [online],
available:http://www.onsemi.com/PowerSolutions/evalBoard.do?id=NCP1395LLCGEVB,
Feb, 2006.
[A.50] Texas Instrument, "UCC25600EVM: LLC resonant half bridge converter 300 W
evaluation module," [online], available: http://www.ti.com/lit/ug/sluu361/sluu361.pdf, Apr,
2009.
[A.51] STMicroelectronics, "STEVAL-ISA018V1 demonstration board based on resonant half-
bridge SMPS for industrial applications," [online], available:
http://www.st.com/internet/com/TECHNICAL_RESOURCES/TECHNICAL_LITERATUR
E/USER_MANUAL/CD00238992.pdf, May, 2010.
[A.52] Power Integration, "Reference design report for a 150 W LLC high-voltage DC-DC
resonant converter using hiperLCSTM LCS702HG," [online], available:
http://www.powerint. com/sites/default/files/PDFFiles/rdr239.pdf, Sep, 2011.
[A.53] Infineon, "ICE2HS01G: High performance resonant mode controler," [online],
available:http://www.infineon.com/dgdl/Design+Guide+for+LLC+Converter+with+ICE2HS
01G+05072011.pdf?folderId=db3a304412b407950112b408e8c90004&fileId=db3a304330f6
8606013103ebd94f3e98, May, 2011.
[A.54] NXP Semiconductors, "TEA1713T: Resonant power supply control IC with PFC,"
[online], available: http://www.nxp.com/documents/data_sheet/TEA1713T.pdf, Feb, 2011.
[A.55] Fairchild Semiconductor, "FAN7621: PFM controller for half-bridge resonant
converters," [online], available: http://www.fairchildsemi.com/ds/FA/FAN7621.pdf, July,
2010.
[A.56] Vishay Siliconix ,” Application notes 833: Switching Analysis of Synchronous Rectifier
MOSFETs With Phase-Shifted Full-Bridge Converter and Current Doubler, ”[online],
available: http://www.vishay.com/docs/69747/answitch.pdf, Oct, 2007
166
[A.57] X. Xie, J. Liu, F.N.K. Poon, M. Pong, “A novel high frequency current-driven SR
applicable to most switching topologies”, IEEE Trans. on Power Electronics, vol. 16, issue. 5,
2001, pp. 635-648.
[A.58] X. Wu, G. Hua, J. Zhang, Z. Qian, “A new current-driven synchronous rectifier for
series–parallel resonant (LLC) DC-DC converter”, IEEE Trans. on Industrial Electronics, vol.
58, issue. 1, 2011, pp. 289-297.
[A.59] International Rectifier, “IR11672AS: Advanced smart rectifier control IC” [Online].
Available: http://www.irf.com/product-info/datasheets/data/ir11672aspbf.pdf.
[A.60] D. Fu, Y. Liu, F. C. Lee and M. Xu, “A Novel Driving Scheme for Synchronous
Rectifiers in LLC Resonant Converters,” IEEE Trans. Power Electron., vol. 24, no. 5, pp.
1321–1329, May 2009.
[A.61] Dianbo Fu; Pengju Kong; Shuo Wang; Lee, F.C.; Ming Xu, "Analysis and suppression of
conducted EMI emissions for front-end LLC resonant DC/DC converters," Power Electronics
Specialists Conference, 2008. PESC 2008. IEEE , vol., no., pp.1144,1150, 15-19 June 2008
[A.62] R. P. Severns, “Topologies for three-element resonant converters,” IEEE Trans. Power
Electron., vol. 7, no. 1, pp. 89–98, Jan. 1992.
[A.63] I. Batarseh, “Resonant converter topologies with three and four energy storage elements,”
IEEE Trans. Power Electron., vol. 9, no. 1, pp. 64–73, Jan. 1994.
[A.64] Dianbo Fu; Lee, F.C.; Ya Liu; Ming Xu, "Novel multi-element resonant converters for
front-end dc/dc converters," Power Electronics Specialists Conference, 2008. PESC 2008.
IEEE , vol., no., pp.250,256, 15-19 June 2008
[A.65] Vicor Corporation, “High-voltage DC distribution is key to increased systemefficiency
and renewable-energy opportunities, ” [online], available:
http://www.vicr.com/documents/whitepapers/wp-High-voltage-DC-istribution.pdf
[A.66] IBM, “A New Redundancy Strategy for High-Availability Power System,” [online],
available: http://www-
03.ibm.com/procurement/proweb.nsf/7a84535a0acd580885256b3f000e250a/3a70bbaee2280
48085257ac20066b558/$FILE/8-A%20New%20Redundancy%20Strategy%20for%20High-
Availability%20Power%20Systems%20(Covi).pdf
167
[A.67] Vicor Corporation, "From 48 V direct to Intel VR12.0: Saving 'Big Data' $500,000 per
datacenter, per year, "[online]
[A.68] C. Prasantanakorn, "Current sharing method for resonant DC-DC transformers," Master
thesis in Virginia Tech, 2011.
[A.69] U. K. Mishra, P. Parikh, and Y. Wu, “AlGaN/GaN HEMTs – an overview of device
operation and applications,” Proceedings of the IEEE, vol. 90, no. 6, pp.1022-1031, Jun.
2002.
[A.70] M. A. Khan, G. Simin, S. G. Pytel, A. Monti, E. Santi, and J. L. Hudgins, “New
developments in gallium nitride and the impact on power electronics,” in proc. IEEE Power
Electronics Specialists Conference, 2005, pp. 15-26.
[A.71] N. Kaminski, “State of the art and the future of wide band-gap devices,” in Proc. IEEE
Power Electronics and Applications, 2009, pp. 1–9.
[A.72] Y. Wu, M. J. Mitos, M. Moore, and S. Heikman, “A 97.8% Efficient GaN HEMT boost
converter with 300W output power at 1 MHz,” IEEE Electron Device Letters, vol. 29, no. 8,
pp. 824–826, Aug. 2008.
[A.73] W. Saito, T. Nitta, Y. Kakiuchi, Y. Saito, K. Tsuda, I. Omura, and M. Yamaguchi, “A
120-W boost converter operation using a high-voltage GaN-HEMT,”IEEE Electron Device
Letters, vol. 29, no. 1, pp. 8–10, Jan. 2008.
[A.74] W. Saito, T. Domon, I. Omura, T. Nitta, Y. Kakiuchi, K. Tsuda, and M. Yamaguchi,
“Demonstration of resonant inverter circuit for electrodeless fluorescent lamps using high
voltage GaN-HEMT,” in proc. IEEE Power Electronics Specialists Conference, 2008, pp.
3324-3329.
[A.75] W. Chen, K. Wong, and K. J. Chen, “Single-chip boost converter using monolithically
integrated AlGaN/GaN lateral field-effect rectifier and normally off HEMT, ” IEEE Electron
Device Letters, vol. 30, no. 5, pp. 430–432, May. 2009.
[A.76] D. Costinett, H. Nguyen; R. Zane, and D. Maksimovic, “GaN-FET based dual active
bridge DC-DC converter,” in proc. IEEE Applied Power Electronics Conference, 2010, pp
1425-1432.
[A.77] M. J. Scott, K. Zou, J. Wang, C. Chen, M. Su, and L. Chen, “A Gallium-Nitride
168
switched-capacitor circuit using synchronous rectification,” in proc. IEEE Energy

Conversion Congress and Exposition, 2011, pp 2501-2505.
[A.78] B. Hughes, Y. Y. Yoon, D. M. Zehnder, and K. S. Boutros, “ A 95% efficient normally-
off GaN-on-Si HEMT hybrid-IC boost converter with 425-W output power at 1MHz,” in
proc. IEEE Compound Semiconductor Integrated Circuit Symposium, 2011, pp 1-3.
[A.79] B. Hughes, J. Lazar, S. Hulsey, D. Zehnder, D. Matic, and K. Boutros, “GaN HFET
Switching Characteristics at 350V-20A and Synchronous Boost Converter Performance at
1MHz,” in proc.IEEE Applied Power Electronics Conference, 2012, pp 2506-2508.
[A.80] J. Delaine, P. Olivier, D. Frey, and K. Guepratte, “High frequency DC-DC converter
using GaN device,” in proc. IEEE Applied Power Electronics Conference, 2012, pp 1754-
1761.
B.Synthesis of Multi-Element Resonant Converters
[B.1] V. Vorperian ,etc , “A complete DC analysis of the series resonant converter, ” Proc of
PESC, 1982, page 85-100
[B.2] Steigerwald, R.L, “A comparison of half-bridge resonant converter topologies,” Trans of
IEEE, Vol 3, issue 2, 1988, page 174-182
[B.3] J. Chen, etc, “Load Independent AC/DC Power Supply for Higher Frequencies with Sine-
Wave Output,” IEEE Trans. Indus Application, Vol IA-19, Issue 2, 1983, page 223-227
[B.4] Bo Yang, “Topology Investigation for Front End DC-DC Power Conversion for
Distributed Power System," PhD Dissertation, Dept. ECE. Virginia Tech, 2003.
[B.5] G. Huang, A. J. Zhang, and Y. Gu, “LLC Series Resonant DC-to-DC Converter,” U.S.
Patent 6 344 979, Feb. 5, 2002.
[B.6] B. Yang, F. C. Lee, and M. Cancannon, “Over current protection methods for LLC
resonant converter,” in Proc. IEEE APEC, 2003, pp. 605–609.
[B.7] B. Yang, F. C. Lee, A. J. Zhang, and G. Huang, “LLC resonant converter for front end
dc-dc conversion,” in Proc. IEEE APEC, 2002, pp. 1108–1112.
169
[B.8] B. Yang, Y. Ren, and F. C. Lee, “Integrated magnetic for LLC resonant converter,” in
Proc. IEEE APEC, 2002, pp. 346–351.
[B.9] Bo Yang, “Topology Investigation for Front End DC-DC Power Conversion for
[B.10] Y. Gu, Z. Lu, L. Hang, Z. Qian, and G. Huang, “Three-level LLC series resonant DC-DC
converter,” IEEE Trans. Power Electron., vol. 20, no. 4, pp. 781–789, Jul. 2005.
[B.11] R. P. Severns, “Topologies for three-element resonant converters,” IEEE Trans. Power
Electron., vol. 7, no. 1, pp. 89–98, Jan. 1992.
[B.12] Daocheng Huang; Dianbo Fu; Lee, F.C.; , "High switching frequency, high efficiency
CLL resonant converter with synchronous rectifier," ECCE 2009. IEEE , vol., no., pp.804-
809, 20-24 Sept. 2009
[B.13] Weiyi Feng; Lee, F.C., "Optimal Trajectory Control of LLC Resonant Converters for
Soft Start-Up," Power Electronics, IEEE Transactions on , vol.29, no.3, pp.1461,1468, March
2014
[B.14] Dianbo Fu; Lee, F.C.; Ya Liu; Ming Xu; , "Novel multi-element resonant converters for
front-end dc/dc converters," PESC 2008. IEEE , vol., no., pp.250-256, 15-19 June 2008
[B.15] I. Batarseh, “Resonant converter topologies with three and four energy storage elements,”
IEEE Trans. Power Electron., vol. 9, no. 1, pp. 64–73,Jan. 1994.
[B.16] Steigerwald, R.L.; , "A comparison of half-bridge resonant converter topologies," Power
Electronics, IEEE Transactions on , vol.3, no.2, pp.174-182, Apr 1988
[B.17] Dianbo Fu; Ya Liu; Lee, F.C.; Ming Xu; , "A Novel Driving Scheme for Synchronous
Rectifiers in LLC Resonant Converters," Power Electronics, IEEE Transactions on , vol.24,
no.5, pp.1321-1329, May 2009
[B.18] Feng, Weiyi; Huang, Daocheng; Mattavelli, Paolo; Fu, Dianbo; Lee, Fred C.; , "Digital
implementation of driving scheme for synchronous rectification in LLC resonant converter,"
ECCE 2010, vol., no., pp.256-263, 12-16 Sept. 2010
C. Evaluation System for Resonant Converters
170
[C.1] E. X. Yang, F. C. Lee, and M. M. Jovanovic, “Small-signal modeling of LCC resonant

converter,” in Proc. IEEE-PESC'92, pp.167-178, 1992.
[C.2] B. Yang and F. C. Lee, “Small-signal analysis for LLC resonant converter,” in CPES
Seminar, 2003, S7.3, pp.144-149, 2003.
[C.3] H.de. Groot, E. Janssen, R. Pagano, K. Schetters, "Design of a 1-MHz LLC resonant
converter based on a DSP-driven SOI half-bridge power MOS module," IEEE Trans. on
Power Electronics, vol.22, Issue.6, pp. 2307 - 2320, 2007.
[C.4] L. Hang; Z. Lu; Z. Qian, "Research of digital control strategy for multi-resonant LLC
converter," in Proc. IEEE Industrial Electronics, 2007, pp. 479 - 484, 2007.
[C.5] J. Jang, M. Joung, B, Choi, and H. Kim, “Dynamic analysis and control design of
optocoupler isolated LLC series resonant converters with wide input and load variations,” in
Proc. IEEE-ECCE'09, pp.758-765, 2009
[C.6] J. Jang, M. Joung, S. Choi, Y. Choi, and B. Choi, “Current mode control for LLC series
resonant DC-to-DC converters,” in Proc. IEEE-APEC'11, pp.21-27, 2011.
[C.7] R. Oruganti, and F.C. Lee, “Resonant power processors, Part I- state plane analysis”,
IEEE Trans. on Industry Application, vol.IA-21, Issue 6, pp. 1453-1460, 1985.
[C.8] R. Oruganti, and F.C. Lee, “Resonant power processors, Part II- methods of control”,
IEEE Trans. on Industry Application, vol.IA-21, Issue 6, pp. 1461-1471, 1985.
[C.9] R. Oruganti, “State-Plane Analysis of Resonant Converters,” PhD dissertation in Virginia
Tech, 1987.
[C.10] R. Oruganti, J.J. Yang, and F.C. Lee, “Implementation of optimal trajectory control or
series resonant converter”, IEEE Trans. on Power Electronics, vol. 3, Issue 3, pp. 318-327,
1988.
D. Integrated System Design for High Current Application
[D.1] G. Huang, A. J. Zhang, and Y. Gu, “LLC Series Resonant DC-to-DC Converter,” U.S.
171
Patent 6 344 979, Feb. 5, 2002.

[D.2] B. Yang, F. C. Lee, A. J. Zhang, and G. Huang, “LLC resonant converter for front end
dc/dc conversion,” in Proc. IEEE APEC 2002, pp. 1108–1112.
[D.3] B. Yang, Y. Ren, and F. C. Lee, “Integrated magnetic for LLC resonant converter,” in
Proc. IEEE APEC, 2002, pp. 346–351.
[D.4] Bo Yang, “Topology Investigation for Front End DC/DC Power Conversion for
[D.5] Y. Gu, Z. Lu, L. Hang, Z. Qian, and G. Huang, “Three-level LLC series resonant DC/DC
converter,” IEEE Trans. Power Electron., vol. 20, no. 4, pp. 781–789, Jul. 2005.
[D.6] D. Fu, Y. Liu, F. C. Lee and M. Xu, “A Novel Driving Scheme for Synchronous
Rectifiers in LLC Resonant Converters,” IEEE Trans. Power Electron., vol. 24, no. 5, pp.
1321–1329, May. 2009.
[D.7] D. Huang; D. Fu; F.C. Lee and P. Kong; , "High-Frequency High-
Efficiency CLL Resonant Converters With Synchronous Rectifiers," , IEEE Trans. Industrial
Electronics , vol.58, no.8, pp.3461-3470, Aug. 2011
[D.8] D. Huang; F.C. Le; D. Fu; , "Classification and selection methodology for multi-element
resonant converters," in Proc. IEEE APEC 2011, pp.558-565, 6-11 March 2011
[D.9] “GaNpowIR – An Introduction”, IRF.com. Feb 2010
[D.10] International Recitifier. “iP2010PbF- High Frequency GaN-Based Integrated Power
Stage”. www.IRF.com
[D.11] Efficient Power Conversion. “EPC1015- Enhancement Mode Power Transistor”.
www.EPC.com
[D.12] Prieto, R.; Cobos, J.A.; Garcia, O.; Alou, P.; Uceda, J., "Taking into account all the
parasitic effects in the design of magnetic components," Applied Power Electronics
Conference and Exposition, 1998. APEC '98. Conference Proceedings 1998., Thirteenth
Annual , vol.1, no., pp.400,406 vol.1, 15-19 Feb 1998
[D.13] Dai, N.; Lee, F. C., "Edge effect analysis in a high-frequency transformer," Power
Electronics Specialists Conference, PESC '94 Record., 25th Annual IEEE , vol., no.,
172
pp.850,855 vol.2, 20-25 Jun 1994

[D.14] Edward Herbert, “Design and Application of Matrix Transformers and Symmetrical
Converters,” a tutorial presented at the High Frequency Power Conversion Conference '90,
Santa Clara, CA, May 11, 1990.
[D.15] Reusch, D.; Lee, F.C.; , "High frequency bus converter with low loss integrated matrix
transformer," Proc. IEEE APEC 2012, pp.1392-1397, 5-9 Feb. 2012
[D.16] Chao Yan; Fan Li; Jianhong Zeng; Teng Liu; Jianping Ying; , "A Novel Transformer
Structure for High power, High Frequency converter," PESC 2007. IEEE , vol., no., pp.214-
218, 17-21 June 2007
[D.17] Fu, D.; Lee, F.C.; Shuo Wang; , "Investigation on transformer design of high frequency
high efficiency dc-dc converters," Proc. IEEE APEC 2010, pp.940-947, 21-25 Feb. 2010
[D.18] Dianbo Fu, PhD Dissertation, “Topology Investigation and System Optimization of
Resonant Converters”
[D.19] Ferroxcube Data Handbook 2009
[D.20] Huang, X.; Li, Q.; Liu, Z.; Lee, F.C., "Analytical loss model of high voltage GaN HEMT
in cascode configuration," Power Electronics, IEEE Transactions on , vol.PP, no.99
E. Passive Integration of Multi-element Resonant Converters
[E.1] Dianbo Fu; Lee, F.C.; Ya Liu; Ming Xu; , "Novel multi-element resonant converters for
front-end dc/dc converters," PESC 2008. IEEE , vol., no., pp.250-256, 15-19 June 2008
[E.2] Daocheng Huang; Lee, F.C.; Dianbo Fu; , "Classification and selection methodology for
multi-element resonant converters," APEC 2011 IEEE , vol., no., pp.558-565, 6-11 March
2011
[E.3] B. Yang, R. Chen and F. C. Lee “ Integrated Magnetic for LLC Resonant Converter,”
Applied Power Electronics Conference and Exposition, APEC‘OZ, vol. 1, pp. 346 - 351.
[E.4] R. Chen, J. T. Strydom and J. D. van Wyk, “Design of Planar Integrated Passive Module
for Zero-Voltage- Switched Asymmetrical Half-Bridge PWM Converter,” Industry
173
Applications, IEEE Transactions, vol. 39, No. 6, pp. 1648-1655

[E.5] R. Chen, J. T. Strydom and J. D. van Wyk, “Second Order Approximation Lumped
Parameter Model for Planar Integrated L-L-C-T Module,” Industry Application Conference,
IAS’02, vol 4, pp. 2419-2424.
[E.6] J. T. Strydomm, and J. D. van Wyk, “Electromagnetic Design Optimization of Planar
Integrated Power Passive Modules,” Power Electronics Specialists Conference, PESC’02, vol.
2, pp. 573-578.
[E.7] W. Liu, J. D. van Wyk and W. G. Odendaal, “Design and Evaluation of Integrated
Electromagnetic Power Passives with Vertical Surface Interconnections,” Applied Power
Electronics Conference and Exposition, APEC ’04, vol. 2, pp. 958-963.
[E.8] L. Zhao, “Generalized Frequency Plane Model of Integrated Electromagnetic Power
Passives,” Ph.D. Dissertation, Virginia Polytechnic Institute and State University, 2004.
[E.9] W. Liu, J. D. van Wyk and W. G. Odendaal, “High Density Integrated Electromagnetic
Power Passives with Vertical Interconnect and Stacked Structure,” Power Electronics
Specialist, PESC’03, vol. 1, pp.442-447.
[E.10] Erickson, R.W.; Maksimovic, D.; , "A multiple-winding magnetics model having directly
measurable parameters," PESC 98, vol.2, no., pp.1472-1478 vol.2, 17-22 May 1998
[E.11] Mogilevsky, B.M.; Shirn, G.A.; , "Accelerated life tests of ceramic capacitors,"
Components, Hybrids, and Manufacturing Technology, IEEE Transactions on , vol.11, no.4,
pp.351-357, Dec 1988
[E.12] Wenduo Liu; Dirker, J.; van Wyk, J.D., "Power Density Improvement in Integrated
Electromagnetic Passive Modules With Embedded Heat Extractors," Power Electronics,
IEEE Transactions on , vol.23, no.6, pp.3142,3150, Nov. 2008
[E.13] Mogilevski, B.M.; Shirn, G.A.; , "Accelerated life tests of ceramic capacitors,"
Electronics Components Conference, 1988., Proceedings of the 38th , vol., no., pp.362-370,
9-11 May 1988
174

Huang D D 2014

Uploaded by

Copyright:

Available Formats

Huang D D 2014

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Huang D D 2014

Uploaded by

Copyright:

Available Formats

Investigation of Topology and Integration for

Multi-Element Resonant Converters

Dissertation submitted to the faculty of the

Fred C. Lee, Chair

December 6th, 2013

Keywords: Integration, Topology, LLC resonant converter

© 2013, Daocheng Huang

With the fast development of communication systems, computers and consumer

current applications. Those issues include magnetic design, start-up, short-circuit

protection, synchronous rectifier drive, EMI noise and integration, etc.

LLC resonant converter cannot handle.

To evaluate the performance of found numerous valuable topologies. Thus, a general

evaluation system is required. State-plane analysis with new normalization factors is

important performance factors, like start-up, short-circuit protection, SR driving,

reduce leakage inductance.

integration method is proposed for multi-elements resonant converters.

for their support, comments, and suggestions.

I am especially indebted to my colleagues in the Power Management Mini-

Wang, Mr. Li Jiang, Mr. Marko Jaksic, Ms. Han Cui.

for her sincere help during the CPES annual conference.

throughout more than twenty years of school.

of my research: Chicony Power, CSR Zhuzhou Institute Co., Ltd., Delta

Technology, Murata Manufacturing Co., Ltd., NEC TOKIN Corporation, NXP

Semiconductors, Richtek Technology, Texas Instruments.

1.1 Background and Motivation........................................................................................... 1

1.2 Issues of PWM DC-DC Power Conversions ................................................................. 4

1.2.1 Pros and Cons of Soft-switching PWM DC-DC Converter .....................................4

1.3 Opportunity of Resonant DC-DC Converters ................................................................ 8

1.3.1 Basic Concepts of PWM Converter and Resonant Converters ................................8

1.4 Challenges of LLC Resonant Converter ...................................................................... 18

1.4.1 Start-Up and Short Circuit Protection ....................................................................18

1.4.2 Synchronous Rectifiers (SR) Driving Scheme .......................................................19

1.4.3 Chanllenges for High Output CurrentApplication .................................................23

1.4.4 EMI Performance of LLC Resonant Converter .....................................................25

1.5 Multi-Element Resonant Converters ............................................................................ 26

1.6 Opportunity of Emerging Technologies ...................................................................... 27

1.6.1 New System Architectures .....................................................................................28

1.6.2 Benefits and Chanllenges of LLC-DCX ................................................................30

1.7 Dissertation Outline ..................................................................................................... 32

Chapter 2. Synthesis of Multi-Element Resonant Converters........................ 36

2.1Fundamental Concept of Resonant Converters ............................................................. 36

2.1.1 Basic Resonant Cells (Two-Element Resonant Tanks) ..........................................38

2.1.2 Synthesis of Basic Resonant Cells .........................................................................41

2.3 Synthesis of Three-Element Resonant Converters....................................................... 42

2.4 Inherent Issues of LLC Resonant Converter ................................................................ 49

2.5 LLC-like Resonant Converters with Notch Filter ........................................................ 54

2.7 Conclusion ................................................................................................................... 66

Chapter 3. Evaluation of LLC-like Resonant Converters .............................. 67

3.1 Introduction .................................................................................................................. 67

3.2 State-Trajectory Analysis of LLC Resonant Converter ............................................... 67

3.3 New Normalization for State-plane ............................................................................. 79

3.4 Evaluation based on Startup, Synchronous Rectifier and EMI .................................... 89

3.5 Conclusion ................................................................................................................... 95

Chapter 4. Integrated System Design for High Current Application ............. 96

4.1 Introduction .................................................................................................................. 96

4.4 Matrix Transformer Design Procedure ...................................................................... 117

4.5 Experimental Results ................................................................................................. 120

4.6 Conclusion ................................................................................................................. 125

Chapter 5. Design Considerations for Passive Integration............................127