Integrated Frequency-Selective Conduction Transmission-Line EMI Filter. Yan Liang

Transcription

1 Integrated Frequency-Selective Conduction Transmission-Line EMI Filter by Yan Liang 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 Khai D. T. Ngo Jacobus Daniel van Wyk Guo-Quan Lu Shuo Wang Yilu Liu Carlos T. A. Suchicital December 8, 28 Blacksburg, VA Keywords: Integrated DM Transmission-Line EMI Filter, Integrated CM&DM Transmission-Line EMI Filter, Discrete LC EMI Filter, Reflective Filter, Absorptive Filter, Electromagnetic Compatibility, Integrated Passives, Transmission Line, Multi- Conductor Lossy Transmission-Line Theory, Two-Conductor Lossy Transmission-Line Theory, Finite Element Analysis, Broadband DM EMI Filter. Copyright 28, Yan Liang

2 Integrated Frequency-Selective Conduction Transmission-Line EMI Filter Yan Liang ABSTRACT The multi-conductor lossy transmission-line model and finite element simulation tool are used to analyze the high-frequency attenuator and the DM transmission-line EMI filter. The insertion gain, transfer gain, current distribution, and input impedance of the filter under a nominal design are discussed. In order to apply the transmission-line EMI filter to power electronics systems, the performance of the filter under different dimensions, material properties, and source and load impedances must be known. The influences of twelve parameters of the DM transmission-line EMI filter on the cut-off frequency, the roll-off slope, and other characteristics of the insertion gain and transfer gain curves are investigated. The most influential parameters are identified. The current sharing between the copper and nickel conductors under different parameters are investigated. The performance of the transmission-line EMI filter under different source and load impedances is also explored. The measurement setups of the DM transmission-line EMI filter using a network analyzer have been discussed. The network analyzer has a common-ground problem that influences the measured results of the high-frequency attenuator. However, the commonground problem has a negligible influence on the measured results of the DM transmission-line EMI filter. The connectors and copper strips between the connectors and the filter introduce parasitic inductance to the measurement setup. Both simulated and measured results show that transfer gain curve is very sensitive to the parasitic inductance. However, the insertion gain curve is not sensitive to the parasitic inductance. There are two major methods to reduce the parasitic inductance of the measurement setup: using small connectors and applying a four-terminal measurement setup. The transfer gain curves of three measurement setups are compared: the two-terminal measurement setup with BNC connectors, the two-terminal measurement setup with Sub Miniature version B (SMB) connectors, and the four-terminal measurement setup with SMB connectors. The four-terminal measurement setup with SMB connectors is the most accurate one and is applied for all the transfer gain measurements in this dissertation.

3 This dissertation also focuses on exploring ways to improve the performance of the DM transmission-line EMI filter. Several improved structures of the DM transmissionline EMI filter are investigated. The filter structure without insulation layer can greatly reduce the thickness of the filter without changing its performance. The meander structure can increase the total length of the filter without taking up too much space and results in the cut-off frequency being shifted lower and achieving more attenuation. A prototype of the two-dielectric-layer filter structure is built and measured. The measurement result confirms that a multi-dielectric-layer structure is an effective way to achieve a lower cut-off frequency and more attenuation. This dissertation proposes a broadband DM EMI filter combining the advantages of the discrete reflective LC EMI filter and the transmission-line EMI filter. Two DM absorptive transmission-line EMI filters take the place of the two DM capacitors in the discrete reflective LC EMI filter. The measured insertion gain of the prototype has a large roll-off slope at low frequencies and large attenuation at high frequencies. The dependence of the broadband DM EMI filter on source and load impedances is also investigated. Larger load (source) impedance gives more attenuation no matter it is resistive, inductive or capacitive. The broadband DM EMI filter always has more highfrequency attenuation than the discrete reflective LC EMI filter under different load (source) impedances. iii

4 ACKNOWLEDGEMENT Foremost, I would like to extend my greatest thanks to both of my advisors, Dr. J. D. van Wyk and Dr. Khai D. T. Ngo for their guidance, encouragement and support. Dr. van Wyk is my first advisor at CPES. Your endless knowledge, creative ideas, and enthusiasm with research have always impressed me. I am very lucky to have Dr. Ngo as my second advisor at CPES. I have learned so much from the weekly discussions with you, from the papers that you passed me, from the red marks that you labeled on the draft of my paper. I am honored to have had the privilege to study under both of your supervision; what I have learned from you is invaluable for the rest of my life as well. I would like to thank my committee member Dr. Shuo Wang. Your knowledge in EMI field has helped me to solve a lot of problems with my work. I want also to thank all the other professors in my degree committee: Dr. G.Q. Lu, Dr. Yilu Liu and Dr. Carlos Suchicital. You have provided, with kindness, your insight and suggestions which are very precious to me, improved my understanding, and broadened my vision. I would also like to thank all of the staff at CPES for their assistance. I want also to express my appreciation to all my colleagues at CPES for your support and friendship. I would like to thank Dr. David Arnold and Mr. Mingliang Wang in University of Florida for their help with the VSM measurement. Last but not least, I would like to thank my parents, my parents-in-law, and my husband for their tremendous support. Without them, I could never finish the study. Great thanks to my daughter Evelyn, too, you have brought me so much happiness and strength. This work was supported primarily by the Engineering Research Center Program of the National Science Foundation under NSF Award Number EEC This work was conducted with the use of Maxwell 2D and Maxwell 2D Extractor, donated in kind by the Ansoft Corporation of the CPES Industrial Consortium. Yan Liang Dec. 28 iv

5 TABLE OF CONTENTS Abstract... ii Acknowledgement... iv Table of contents... v List of figures... ix List of tables...xxiii List of symbols... xxiv Chapter 1: Introduction Integration technology in power electronics EMI filters for switch mode power supplies EMI in power electronics system Discrete reflective LC EMI filter Integrated reflective LC EMI filter Absorptive filter Integrated absorptive transmission-line EMI filter Motivation and objective Dissertation outline Chapter 2: Electromagnetic modeling of integrated transmission-line EMI filter Introduction Two-conductor lossy transmission-line model Application of two-conductor lossy transmission-line model to the filter Multi-conductor lossy transmission-line model Application of multi-conductor lossy transmission-line model to the filter Model parameter extraction v

6 Parameter extraction based on analytical method Parameter extraction based on Maxwell 2D Parameter extraction based on Maxwell 2D Extractor Comparison of two-conductor and multi-conductor transmission-line models Summary Chapter 3: Parametric characterization of high-frequency attenuator Introduction Insertion gain, transfer gain, currents and input impedance Parametric study of the high-frequency attenuator Influence of dimensions Influence of conductor s properties Influence of dielectric material s properties Dependence on source and load impedances Summary... 7 Chapter 4: Parametric characterization of DM Transmission-Line EMI filter Introduction Insertion gain, transfer gain, current and input impedance Parametric study of DM transmission-line EMI filter Influence of dimensions Influence of inner conductor s properties Influence of dielectric material s properties Influence of insulation material s properties Dependence on source and load impedances Summary vi

7 Chapter 5: Measurement setup of integrated DM transmission-line EMI filter Introduction Measurement setup using network analyzer Measurement setup for insertion gain Measurement setup for transfer gain Common-ground problem of network analyzer Influence of common-ground problem to high-frequency attenuator Influence of common-ground problem to DM transmission-line EMI filter Measurement board design Summary Chapter 6: Experimental verification Introduction Material properties of DM transmission-line EMI filter Nickel s properties Dielectric s properties Manufacturing process of DM transmission-line EMI filter Experimental verification on high-frequency attenuator Experimental verification on DM transmission-line EMI filter Summary Chapter 7: Improved structures of DM transmission-line EMI filter Introduction Different connections of DM transmission-line EMI filter DM transmission-line EMI filter without insulation layer DM transmission-line EMI filter with meander structure vii

8 7.5. DM transmission-line EMI filter with multi-dielectric-layer structure A Broadband DM EMI filter combining reflective and absorptive effects Summary Chapter 8: Conclusion and future work Introduction High-frequency attenuator DM transmission-line EMI filter Measurement setup Improved structures of the DM transmission-line EMI filter Future work References Appendix I: Manufacturing process of DM transmission-line EMI filter viii

9 LIST OF FIGURES Fig Discrete AC/DC converter for DPS [5]... 2 Fig Integrated AC/DC converter for DPS [5]... 2 Fig D model of integrated AC drives [15]... 3 Fig Integrated AC drive converter [15] Fig EMI filter in power electronics system Fig Discrete reflective LC EMI filter [23]... 6 Fig Integrated reflective LC EMI filter [26]... 7 Fig Low-pass coaxial line... 8 Fig Structures of transmission line low-pass filters [35]... 8 Fig Meander structure of conductor [35]... 9 Fig Cross-sectional view of integrated DM transmission-line EMI filter... 1 Fig Integrated transmission-line EMI filter prototype [38]... 1 Fig Measured transfer gain of the DM transmission-line filter prototype [38]... 1 Fig Cross-sectional view of integrated CM & DM transmission-line EMI filter [39] Fig Integrated CM & DM transmission-line EMI filter prototype Fig Measured DM transfer gain of the integrated CM & DM EMI filter prototype Fig Measured CM transfer gain of the integrated CM & DM EMI filter prototype Fig Equivalent circuit of an infinitesimal section Fig Two-port network Fig Schematic of transfer gain and insertion gain measurement circuit for transmission-line EMI filter... 2 Fig Equivalent circuit of infinitesimal sections of inside and outside transmissionline structures ix

10 Fig Equivalent circuit of the DM transmission-line EMI filter Fig Multi-conductor transmission-line configuration parallel to the z axis Fig Equivalent circuit of an infinitesimal section Fig Cross-sectional view of integrated DM transmission-line EMI filter Fig Equivalent circuit of an infinitesimal section Fig Schematic of insertion gain measurement circuit for integrated DM transmission-line EMI filter Fig Cross-sectional view of DM transmission-line EMI filter Fig Maxwell 2D model of the DM transmission-line EMI filter Fig Current density in copper and nickel conductors at 1 khz from Maxwell 2D Fig Current density in copper and nickel conductors at 1MHz from Maxwell 2D Fig Resistances vs. frequency from Maxwell 2D Extractor Fig Inductances vs. frequency from Maxwell 2D Extractor Fig Comparison of resistances from Maxwell 2D and Maxwell 2D Extractor Fig Comparison of Inductances from Maxwell 2D and Maxwell 2D Extractor Fig Insertion gains of DM transmission-line EMI filter Fig Transfer gains of DM transmission-line EMI filter Fig Schematic of measurement setup for high-frequency attenuator Fig Insertion gain of high-frequency attenuator Fig Transfer gain of high-frequency attenuator Fig Insertion gain of high-frequency attenuator with copper conductor Fig Transfer gain of high-frequency attenuator with copper conductor Fig Input impedance of high-frequency attenuator Fig Input and output currents of high-frequency attenuator x

11 Fig Dimensions of high-frequency attenuator Fig Insertion gain of high-frequency attenuator vs. length Fig Transfer gain of high-frequency attenuator vs. length Fig Insertion gain of high-frequency attenuator vs. width Fig Transfer gain of high-frequency attenuator vs. width Fig Insertion gain of high-frequency attenuator vs. length & width Fig Transfer gain of high-frequency attenuator vs. length & width Fig Insertion gain of high-frequency attenuator vs. thickness of conductor Fig Skin depth of nickel Fig Transfer gain of high-frequency attenuator vs. thickness of conductor... 6 Fig Insertion gain of high-frequency attenuator vs. thickness of dielectric layer Fig Transfer gain of high-frequency attenuator vs. thickness of dielectric layer Fig Insertion gain of high-frequency attenuator vs. thickness of dielectric layer & length Fig Insertion gain of high-frequency attenuator vs. relative permeability of the conductor Fig Transfer gain of high-frequency attenuator vs. relative permeability of the conductor Fig Insertion gain of high-frequency attenuator vs. conductivity of the conductor Fig Transfer gain of high-frequency attenuator vs. conductivity of the conductor Fig Insertion gain of high-frequency attenuator vs. relative permittivity of dielectric Fig Transfer gain of high-frequency attenuator vs. relative permittivity of dielectric Fig Insertion gain of high-frequency attenuator vs. loss factor of dielectric xi

12 Fig Transfer gain of high-frequency attenuator vs. loss factor of dielectric Fig Insertion gain of high-frequency attenuator vs. resistive load impedance Fig Insertion gain of high-frequency attenuator vs. inductive load impedance Fig Insertion gain of high-frequency attenuator vs. capacitive load impedance Fig Insertion gain of high-frequency attenuator vs. resistive source impedance Fig Insertion gain of high-frequency attenuator vs. inductive source impedance Fig Insertion gain of high-frequency attenuator vs. capacitive source impedance Fig Schematic of measurement setup of DM transmission-line EMI filter Fig Insertion gains of DM transmission-line EMI filter and high-frequency attenuator Fig Transfer gains of DM transmission-line EMI filter and high-frequency attenuator Fig Input impedances of DM transmission-line EMI filter and high-frequency attenuator Fig Input and output currents of DM transmission-line EMI filter Fig Dimensions of DM transmission-line EMI filter Fig Insertion gain of DM transmission-line EMI filter vs. length... 8 Fig Transfer gain of DM transmission-line EMI filter vs. length Fig Input currents of DM transmission-line EMI filter vs. length Fig Insertion gain of DM transmission-line EMI filter vs. width Fig Transfer gain of DM transmission-line EMI filter vs. width Fig Input currents of DM transmission-line EMI filter vs. width Fig Insertion gain of DM transmission-line EMI filter vs. length & width Fig Transfer gain of DM transmission-line EMI filter vs. length & width Fig Insertion gain of DM transmission-line EMI filter vs. thickness of nickel layer xii

13 Fig Input currents of DM transmission-line EMI filter vs. thickness of nickel layer Fig Transfer gain of DM transmission-line EMI filter vs. thickness of nickel layer Fig Insertion gain of DM transmission-line EMI filter vs. thickness of copper layer Fig Transfer gain of DM transmission-line EMI filter vs. thickness of copper layer Fig Input currents of DM transmission-line EMI filter vs. thickness of copper layer Fig Insertion gain of DM transmission-line EMI filter vs. thickness of dielectric layer Fig Transfer gain of DM transmission-line EMI filter vs. thickness of dielectric layer Fig Input currents of DM transmission-line EMI filter vs. thickness of dielectric layer Fig Insertion gains of DM transmission-line EMI filter vs. thickness of dielectric layer & filter length Fig Insertion gain of DM transmission-line EMI filter vs. thickness of insulation layer Fig Transfer gain of DM transmission-line EMI filter vs. thickness of insulation layer Fig Input currents of DM transmission-line EMI filter vs. thickness of insulation layer... 9 Fig Insertion gain of DM transmission-line EMI filter vs. relative permeability of nickel Fig Transfer gain of DM transmission-line EMI filter vs. relative permeability of nickel Fig Input currents of DM transmission-line EMI filter vs. relative permeability of nickel Fig Insertion gain of DM transmission-line EMI filter vs. conductivity of nickel xiii

14 Fig Transfer gain of DM transmission-line EMI filter vs. conductivity of nickel Fig Input currents of DM transmission-line EMI filter vs. conductivity of nickel Fig Insertion gain of DM transmission-line EMI filter vs. relative permittivity of dielectric Fig Transfer gain of DM transmission-line EMI filter vs. relative permittivity of dielectric Fig Input currents of DM transmission-line EMI filter vs. relative permittivity of dielectric Fig Insertion gain of DM transmission-line EMI filter vs. loss factor of dielectric Fig Transfer gain of DM transmission-line EMI filter vs. loss factor of dielectric Fig Input currents of DM transmission-line EMI filter vs. loss factor of dielectric Fig Insertion gain of DM transmission-line EMI filter vs. relative permittivity of insulation Fig Transfer gain of DM transmission-line EMI filter vs. relative permittivity of insulation Fig Input currents of DM transmission-line EMI filter vs. relative permittivity of insulation Fig Insertion gain of DM transmission-line EMI filter vs. loss factor of insulation Fig Transfer gain of DM transmission-line EMI filter vs. loss factor of insulation Fig Input currents of DM transmission-line EMI filter vs. loss factor of insulation Fig Insertion gain of DM transmission-line EMI filter vs. resistive load impedance Fig Insertion gain of DM transmission-line EMI filter vs. inductiveload impedance Fig Insertion gain of DM transmission-line EMI filter vs. capacitive load impedance Fig Insertion gain of DM transmission-line EMI filter vs. resistive source impedance xiv

15 Fig Insertion gain of DM transmission-line EMI filter vs. inductive source impedance Fig Insertion gain of DM transmission-line EMI filter vs. capacitive source impedance Fig Schematic of transfer gain and insertion gain measurement circuit for transmission-line EMI filter Fig Agilent 4395A network/spectrum/impedance analyzer Fig Input/output terminals of Agilent 4395A Fig Voltage measurement without DM transmission-line EMI filter Fig Voltage measurement with the DM transmission-line EMI filter Fig Insertion gain measurement setup using Agilent 4395A Fig Agilent 87512A transmission / reflection test kit Fig Internal circuit of Agilent 87512A transmission / reflection test kit Fig Schematic of measurement setup with Agilent 87512A transmission / reflection test kit Fig Simulated insertion gains of DM transmission-line EMI filter with/without test kit Fig Measured insertion gains of DM transmission-line EMI filter with/without test kit Fig Transfer gain measurement setup with external load Fig Transfer gain measurement setup with internal load Fig Measured transfer gains of transmission-line EMI filter with inside/outside load Fig Measurement setup of high-frequency attenuator with to-ground impedances Fig Insertion gains of high-frequency attenuator under different SG and LG Fig Transfer gains of high-frequency attenuator under different SG and LG xv

16 Fig Input currents of high-frequency attenuator when SG = LG = Fig Output currents of high-frequency attenuator when SG = LG = Fig Insertion gain of high-frequency attenuator vs. length when SG = LG = Fig Transfer gain of high-frequency attenuator vs. length when SG = LG = Fig Insertion gain of high-frequency attenuator vs. width when SG = LG = Fig Transfer gain of high-frequency attenuator vs. width when SG = LG = Fig Insertion gain of high-frequency attenuator vs. length & width when SG = LG = Fig Transfer gain of high-frequency attenuator vs. length & width when SG = LG = Fig Insertion gain of high-frequency attenuator vs. thickness of nickel when SG = LG = Fig Transfer gain of high-frequency attenuator vs. thickness of nickel when SG = LG = Fig Insertion gain of high-frequency attenuator vs. thickness of dielectric when SG = LG = Fig Transfer gain of high-frequency attenuator vs. thickness of dielectric when SG = LG = Fig Insertion gain of high-frequency attenuator vs. relative permeability of nickel when SG = LG = Fig Transfer gain of high-frequency attenuator vs. relative permeability of nickel when SG = LG = Fig Insertion gain of high-frequency attenuator vs. conductivity of nickel when SG = LG = Fig Transfer gain of high-frequency attenuator vs. conductivity of nickel when SG = LG = xvi

17 Fig Insertion gain of high-frequency attenuator vs. relative permittivity of dielectric when SG = LG = Fig Transfer gain of high-frequency attenuator vs. relative permittivity of dielectric when SG = LG = Fig Insertion gain of high-frequency attenuator vs. loss factor of dielectric when SG = LG = Fig Transfer gain of high-frequency attenuator vs. loss factor of dielectric when SG = LG = Fig Schematic of measurement setup of DM transmission-line EMI filter Fig Insertion gains of DM transmission-line EMI filter under different SG and LG Fig Transfer gains of DM transmission-line EMI filter under different SG and LG Fig Input and ground currents of DM transmission-line EMI filter when SG = LG = Fig Output and ground currents of DM transmission-line EMI filter when SG = LG = Fig Measurement setup of DM transmission-line EMI filter with parasitic inductance Fig Insertion gains of DM transmission-line EMI filter under different parasitic inductance Fig Transfer gains of DM transmission-line EMI filter under different parasitic inductance Fig SMB and BNC connectors Fig Measurement board with BNC connectors Fig Connection of BNC connector to board copper strip Fig Two-terminal measurement board with SMB connectors Fig Four-terminal measurement board with SMB connectors xvii

18 Fig Connection of SMB connector to filter Fig Measured transfer gains using different measurement setups Fig Measured insertion gains using different measurement setups Fig Dimension of electroplated nickel strip Fig Resistance of electroplated nickel strip Fig Electrical conductivity of electroplated nickel strip Fig B-H hysteresis loops of nickel core at different frequencies [36] Fig vibrating sample magnetometer [55] Fig Measured B-H loop of nickel using VRM Fig Measured relative permeability of electroplated nickel Fig Measured capacitance and loss factor vs. frequency Fig Measured relative permittivity of Y5V vs. electric field strength and temperature Fig Measured and simulated insertion gains of high-frequency attenuator under nominal design Fig Measured and simulated transfer gains of high-frequency attenuator under nominal design Fig Measured and simulated insertion gains of high-frequency attenuator vs. length Fig Measured and simulated transfer gains of high-frequency attenuator vs. length Fig Measured and simulated insertion gains of high-frequency attenuator vs. width Fig Measured and simulated transfer gains of high-frequency attenuator vs. width Fig Measured and simulated insertion gains of high-frequency attenuator vs. thickness of nickel Fig Skin depth of nickel xviii

19 Fig Measured and simulated transfer gains of high-frequency attenuator vs. thickness of nickel Fig Measured and simulated insertion gains of high-frequency attenuator vs. thickness of dielectric Fig Measured and simulated transfer gains of high-frequency attenuator vs. thickness of dielectric Fig Measured and simulated insertion gains of high-frequency attenuator vs. dielectric material Fig Measured and simulated transfer gains of high-frequency attenuator vs. dielectric material Fig Measured and simulated insertion gains of DM TL EMI filter under nominal design Fig Measured and simulated transfer gains of DM transmission-line EMI filter under nominal design Fig Measured and simulated insertion gains of DM transmission-line EMI filter vs. length Fig Measured and simulated transfer gains of DM transmission-line EMI filter vs. length Fig Measured and simulated insertion gains of DM transmission-line EMI filter vs. width Fig Measured and simulated transfer gains of DM transmission-line EMI filter vs. width Fig Measured and simulated insertion gains of DM transmission-line EMI filter vs. nickel thickness Fig Measured and simulated transfer gains of DM transmission-line EMI filter vs. nickel thickness xix

20 Fig Measured and simulated insertion gains of DM transmission-line EMI filter vs. copper thickness Fig Measured and simulated transfer gains of DM transmission-line EMI filter vs. copper thickness Fig Measured and simulated insertion gains of DM transmission-line EMI filter vs. dielectric thickness Fig Measured and simulated transfer gains of DM transmission-line EMI filter vs. dielectric thickness Fig Measured and simulated insertion gains of DM transmission-line EMI filter vs. insulation thickness Fig Measured and simulated transfer gains of DM transmission-line EMI filter vs. insulation thickness Fig Measured and simulated insertion gains of DM transmission-line EMI filter vs. dielectric material Fig Measured and simulated transfer gains of DM transmission-line EMI filter vs. dielectric material Fig Different connections of DM transmission-line EMI filter Fig Measurement setup of DM transmission-line EMI filter with new connection method Fig Simulated insertion gains of DM transmission-line EMI filter under different connections Fig Measured insertion gains of DM transmission-line EMI filter under different connections Fig Measured insertion gains of DM transmission-line EMI filters with/without insulation layer Fig DM transmission-line EMI filter with meander and long-strip structures xx

21 Fig Measured insertion gains of DM transmission-line EMI filters with meander and long-strip structures Fig Cross-sectional view of two-dielectric-layer DM transmission-line EMI filter Fig Connection of two-dielectric-layer DM transmission-line EMI filters Fig Measured insertion gains of one-dielectric-layer and two-dielectric layer DM transmission-line EMI filters Fig Discrete one-stage LC EMI filter [23] Fig Equivalent circuit of the discrete one-stage LC EMI filter Fig Equivalent circuit of the discrete one-stage DM LC EMI filter Fig DM Insertion gain of the discrete reflective LC EMI filter [23] Fig Insertion gain of one DM absorptive transmission-line EMI filter Fig Structure of the broadband EMI filter Fig Prototype of the broadband DM EMI filter Fig Measured insertion gains of the broadband DM EMI filter and discrete reflective LC EMI filter Fig Broadband EMI filter - three two-port networks in cascade Fig Two-port network of two DM inductors Fig Simulated insertion gain of the broadband DM EMI filter prototypes Fig Schematic of discrete reflective LC EMI filter in Saber simulation Fig Insertion gains of the broadband DM EMI filter vs. resistive load impedance Fig Insertion gains of the discrete reflective LC EMI filter vs. resistive load impedance Fig Insertion gains of the broadband DM EMI filter at vs. inductive load impedance Fig Insertion gains of the discrete reflective LC EMI filter vs. inductive load impedance xxi

22 Fig Insertion gains of the broadband DM EMI filter vs. capacitive load impedance Fig Insertion gains of the discrete reflective LC EMI filter vs. capacitive load impedance Fig. AI-1. Dielectric sheet during manufacturing process of DM transmission-line EMI filter Fig. AI-2. FR4 frame for nickel electroplating Fig. AI-3. Bulk Nickel as anode during electroplating Fig. AI-4. Nickel electroplating times for various thickness and current densities... 2 xxii

23 LIST OF TABLES Table 2-1 Technical parameters of integrated DM transmission-line EMI filter Table 2-2 Comparison of Two-conductor and Multi-conductor Model Table 3-1 Nominal design of an high-frequency attenuator... 5 Table 3-2 Nominal value and sweep range of an high-frequency attenuator Table 3-3 Summary of parametric study results on high-frequency attenuator Table 4-1 Nominal design of DM transmission-line EMI filter Table 4-2 Nominal design of DM transmission-line EMI filter Table 4-3 Summary of parametric study results Table 6-1 Properties of Nickel 2, commercially pure grade (99.6% Ni) Table 6-2 Electrical resistivities of electroplated nickel from various solutions Table 6-3 Magnetic property of electroplated nickel Table 6-4 Calculated average relative permeability of nickel core Table 6-5 EIA three-character code Table 6-6 Properties of dielectric materials Table 6-7 Parameters of high-frequency attenuator prototypes Table 6-8 Parameters of DM transmission-line EMI filter prototypes Table 7-1 Parameters of DM absorptive transmission-line EMI filter xxiii

24 LIST OF SYMBOLS R ( Ω/m): Resistance per unit length; L ( H/m): Inductance per unit length; C ( F/m): Capacitance per unit length; G (S/m ): Conductance per unit length; R Cu L Cu C Cu ( Ω/m): Resistance per unit length of copper conductors; ( H/m): Inductance per unit length of copper conductors; ( F/m): Capacitance per unit length between two copper conductors; G Cu (S/m ): Conductance per unit length between two copper conductors; R Ni ( Ω/m): Resistance per unit length of nickel conductors; L Ni ( H/m): Inductance per unit length of nickel conductors; C Ni ( F/m): Capacitance per unit length between two nickel conductors; G Ni (S/m ): Conductance per unit length between two nickel conductors; R ii ( Ω/m): Self resistance per unit length; R ij ( Ω/m): Mutual resistance per unit length; L ii ( H/m): Self inductance per unit length; L ij ( H/m): Mutual inductance per unit length; C i ( F/m): Capacitance per unit length between the ith conductor to ground; C ij ( F/m): Capacitance per unit length between the ith and jth conductors; G i (S/m ): Conductance per unit length between the ith conductor to ground; G ij (S/m ): Conductance per unit length between the ith and jth conductors; ( Ω/m): Impedance per unit length; Y (S/m ): Admittance per unit length; Cu Ni ( Ω/m): Impedance per unit length of copper conductors; ( Ω/m): Impedance per unit length of nickel conductors; : Characteristic impedance of the two-conductor transmission-line structure; L ( Ω ): Load impedance; S ( Ω ): Source impedance; xxiv

25 in ( Ω ): Input impedance of the two-conductor transmission-line structure; f (Hz): Frequency; ω (rad/s): angular frequency; l ( m ): Length; γ : Propagation constant; α (db/m): Attenuation constant; β : Phase constant; ρ v : Voltage reflection coefficient at load side; V in (V): Input voltage of the transmission-line EMI filter; V out V L _ w V L _ w / o (V): Output voltage of the transmission-line EMI filter; (V): Load voltage with the transmission-line EMI filter; (V): Load voltage without the transmission-line EMI filter; t V = V V ) (V): Natural voltage vector; I = ( 1 2 L ( 1 2 L I n V n t I I ) (A): Natural current vector; V S (V): Source voltage vector; Vˆ (V): Modal voltage vector; Î (A): Modal current vector; V ˆ ( i ) (V): Incident voltage vector in the modal space; V ˆ ( r ) (V): Reflected voltage vector in the modal space; I ˆ( i ) (A): Incident current vector in the modal space; I ˆ( r ) (A): Reflected current vector in the modal space; ( Ω/m): Impedance per unit length matrix; Y (S/m ): Admittance per unit length matrix; R ( Ω/m): Resistance per unit length matrix; L ( H/m): Inductance per unit length matrix; C ( F/m): Capacitance per unit length matrix; G (S/m ): Conductance per unit length matrix; xxv

26 T : Transformation matrix relating natural voltage V and modal voltage Vˆ ; W : Transformation matrix relating natural current I and modal current Î ; γ 2 = diag{γ i 2, γ k 2, γ n 2 }: diagonal matrix of eigen values of Y and Y; γ i : eigen value of Y and Y; W : Characteristic impedance in the natural space; Y W : Characteristic admittance in the natural space; Ẑ W : Characteristic impedance in the modal space; Ŷ W : Characteristic admittance in the modal space; Γ : Square-root matrix of Y ; sinh: hyperbolic sine; SINH(A): hyperbolic-sine matrix function of A; cosh: hyperbolic cosine; COSH(A): hyperbolic-cosine matrix function of A; coth: hyperbolic cotangent; COTH(A): hyperbolic- cotangent matrix function of A; csch: hyperbolic cosecant; CSCH(A): hyperbolic-cosecant matrix function of A; T A : transfer matrix; LG ( Ω ): To-ground impedance at load side; SG ( Ω ): To-ground impedance at source side; L ( Ω ): Load impedance matrix; S ( Ω ): Source impedance matrix; I ( Ω ): Input impedance matrix; ESL ( Ω ): Impedance of parasitic inductance; TG: Transfer gain; IG: Insertion gain; δ Cu (m): Skin depth of copper; δ Ni (m): Skin depth of nickel; xxvi

27 t Cu (m): Thickness of copper conductor; t Ni (m): Thickness of nickel conductor; t die (m): Thickness of dielectric layer; t ins (m): Thickness of insulation layer; w Cu (m): Width of copper conductor; w Ni (m): Width of nickel conductor; d Cu (m): Distance between two copper conductors; d Ni (m): Distance between two nickel conductors; σ Cu (S/m): Electrical conductivity of copper; σ Ni (S/m): Electrical conductivity of nickel; µ (H/m): Permeability of free space; µ Ni : Nickel s relative permeability; ε (F/m): Permittivity of free space; ε die : Relative permittivity of dielectric material; ε ins : Relative permittivity of insulation material; tanδ die : Loss factor of dielectric layer; tanδ ins : Loss factor of insulation layer; L int_cu (H/m): Internal inductance per unit length between two copper conductors; L int_ni (H/m): Internal inductance per unit length between two nickel conductors; L ext_cu (H/m): External inductance per unit length between two copper conductors; L ext_ni (H/m): External inductance per unit length between two nickel conductors; C die (F/m): Capacitance per unit length of dielectric layer; G die (S/m): Conductance per unit length of dielectric layer; xxvii

28 Chapter 1: Introduction 1.1. Integration technology in power electronics Power electronic products, to date, are essentially custom designed with a long design cycle time. They are based on assembling pre-manufactured discrete components. Each component consists of a number of parts, manufactured in a variety of manufacturing processes. This has resulted in a diversity of construction parts and mutually incompatible manufacturing processes in a typical power electronic converter and has brought power electronics to the edge where it becomes extremely difficult to reduce the cost and size of power electronic converters [1]-[2]. The recent trends in power electronics pursue higher levels of integration and better packaging techniques in order to meet the system requirements imposed on power electronic converters related to cost, size and power density [3]. For example, CPES has developed an integrated system approach via integrated power electronic modules (IPEMs) that can lead to a standardized power electronics systems approach which is suitable for automated manufacturing and mass production. This approach will drive the development of standardized systems design via functional blocks, as well as the modularization and standardization of functional modules. The IPEM approach enables dramatic improvement in performance and cost-effectiveness of power electronic systems. Among the technologies being developed in CPES are planar metalization device interconnects, allowing 3-D integration of power devices, as well as the integration of power passives to increase the power density, as these dominate the physical size of the 1

29 system. The technologies being developed will ultimately span a wide range of applications from distributed power systems to motor drives [4]-[8]. Fig. 1-1 shows the discrete version of a 1kW AC/DC converter for the distributed power system (DPS). Fig. 1-2 is the integrated version with the active, passive and EMI filter modules. Due to the large transformer and inductor size, discrete version can t meet 1U profile requirement and has a 1.5U profile, which results in a 7.5W/in 3 power density. Meanwhile, because of low profile active and passive IPEMs, the IPEM version converter can easily meet U profile requirement, and achieve 11.7W/in 3 power density. Thus 56% power density improvement has been achieved by replacing the discrete devices with IPEMs. Fig Discrete AC/DC converter for DPS [5]. Fig Integrated AC/DC converter for DPS [5]. The researchers in ECPE, Delft also developed a system integration approach that allows for achieving high power densities [9]-[15]. This approach is also hybrid integration with integrated power electronics modules (IPEMs) as basic building blocks, together with integrated thermal and spatial design with optimized air flow. In [15], the integrated electrical, spatial and thermal design of a case study 2.2kW converter for AC 2

30 drives is presented. The integrated converter has a power density of 62.3W/in 3, which is almost four times larger than that of commercially available drives. Fig. 1-3 shows the 3D model of the integrated AC drives. Fig. 1-4 shows the prototype of the integrated AC drive converter. Fig D model of integrated AC drives [15]. Fig Integrated AC drive converter [15]. 3

31 1.2. EMI filters for switch mode power supplies EMI in power electronics system Electromagnetic interference (EMI) is a very important issue not only for power electronics, but for all electronic and electrical equipments. The EMI noise frequency range of highest concern for power electronics systems is conducted radio frequency disturbance, gauged from 15 KHz to 3 MHz [16]. As power electronics systems move toward integration, modularization, standardization and planarization to improve the electric and thermal performance with reduced size, weight and cost, the EMI issue is getting more severe. Higher switching frequency helps to reduce converter size, weight and cost, it also increases EMI concerns. Advanced packaging and integration technologies make it possible to squeeze more components into a small space, while more considerations regarding EMI are required for power electronics circuit design [17]-[18]. The switching function and the corresponding high di/dt loops and high dv/dt nodes are the major mechanism of electromagnetic noise generation in power electronics converters. To reduce the noises, there are different approaches, such as filtering, shielding, grounding, isolation, separation and orientation [16]. For switch mode power supplies, EN5522 class B is one of the generic standards for conducted EMI for most of the AC/DC converters in both commercial and industrial applications in the United States. Most of AC/DC power products need to meet the EMI standards before coming to market. EMI filters are necessary to attenuate noise at switching frequencies and the harmonics so as to maintain electromagnetic compatibility. 4

32 As shown in Fig. 1-5, EMI filters are placed between the power line and the input of the converter. There are several requirements for the EMI filter. First, the EMI filter must be able to carry full load power. Second, the EMI filter must produce sufficient attenuation for a wide range of source and load impedances. Third, the EMI filter must provide high enough attenuation at high frequency. SOURCE EMI + Power Converter _ LOAD Filter Fig EMI filter in power electronics system Discrete reflective LC EMI filter Fig. 1-6 shows a typical EMI filter for AC/DC converters. It is a reflective type of filter comprised of inductors and capacitors. The reflective LC EMI filter attenuates the high-frequency noise by reflecting the noise back to the source out-of-phase where cancellation occurs. There are some problems with this type of discrete reflective LC EMI filter. First, the low-frequency power current and high-frequency noise current share the same path in this type of filter, thus, the filter must be of the same power level as the subsequent power converter. Thus the EMI filter is very large for high-power systems. Second, due to the reflective nature of the filter, the attenuation of the filter depends a lot on source and load impedances. Many commercial EMI filters give the attenuation curve of the filter under the condition that the source and load impedances are both 5Ω. In reality, the power converter input impedance can be far away from 5Ω. The attenuation of the EMI filter can also be quite different from the given attenuation curve. Third, the high-frequency attenuation of the discrete reflective LC filter mainly depends on the parasitics of the discrete components, such as the equivalent parallel capacitance (EPC) 5

33 of the inductors and the equivalent series inductance (ESL) of the capacitors. The parasitics caused by the filter layout further impairs the filter performance. The effective filter frequency range is normally below a few MHz. Fourth, the discrete EMI filter consists of a fairly large number of components, each of which involves different processing techniques. Some of them may require labor-intensive processing steps. These components are functionally and structurally separated. This requires excessive material and manufacturing time. Last, because the components of a discrete EMI filter vary in type, value, size and form factor, considerable space is taken by interconnections between components, which leads to inefficient utilization of space [2]-[23]. Fig Discrete reflective LC EMI filter [23] Integrated reflective LC EMI filter In order to improve the high-frequency characteristics with a compact size and low profile, and to achieve structural, functional, processing and mechanical integration to reduce manufacturing time and cost, planar electromagnetic integration technology was proposed for reflective LC EMI filter design. The details about the technology of integrating reflective LC EMI filters are discussed in [23] and [24]. In [25] and [26], winding capacitance cancellation method with embedded conductive shield layer is applied to integrated reflective LC EMI filter structure to reduce the parasitic capacitance, shrink the filter size and improve the filter s high-frequency performance. 6

34 Fig. 1-7 shows the prototype of the integrated LC EMI filter. Compared with discrete reflective LC EMI filter, the integrated reflective LC EMI filter has better highfrequency performance and around 5% size reduction. However, same as discrete reflective LC EMI filter, the integrated reflective LC EMI filter has the drawbacks as large dependence on source and load impedances and same paths for power and noise currents. Fig Integrated reflective LC EMI filter [26] Absorptive filter Unlike reflective filters that consist only of inductors and capacitors, absorptive filters have lossy elements (electric, magnetic or dielectric) to purposely generate more losses in the desired frequency range to realize unconditional EMI filters. An unconditional EMI filter provides a certain minimum attenuation regardless of the source and load impedances. The first absorptive (dissipative) low-pass filter, in the form of electrical wires and cables, using the principle of magnetic absorption, appeared in 1957 [27]. Fig. 1-8 shows the basic structure of an absorptive low-pass cable: a layer of absorptive flexible composite surrounds the conductor, covered either internally or externally by a thin layer of insulating material, which gives the wire its dielectric strength. Since 196 s, such cables have been on the market, particularly for military, aeronautical, and space applications. To further increase high-frequency attenuation of these cables, the skin 7

35 effect can be enhanced. A thin resistive and/or magnetic layer creates an artificial skineffect, resulting in more high-frequency losses [28]-[33]. Proximity effect can also be used to enhance high-frequency losses, as demonstrated in [34] for multilayer strip conductors. An absorptive transmission-line low-pass filter for switching power supplies is proposed in [35]. The filter had metallic soft magnetic layer, dielectric layer and conductor line, as shown in Fig The conductor layer had a planar meander pattern to increase the total conductor length. The filter is for low power application. Its effectiveness was verified in [35] as output noises were filtered for a 2W 5MHz DC/DC converter. Absorptive composite Shield Copper conductor Thin insulating layer Fig Low-pass coaxial line. a. Conductor inside b. Conductor outside Fig Structures of transmission line low-pass filters [35]. 8

36 a. Conductor inside b. Conductor outside Fig Meander structure of conductor [35] Integrated absorptive transmission-line EMI filter The integrated absorptive transmission-line EMI filter structure that will be discussed throughout this dissertation was first proposed by van Wyk Jr. [36][37]. He studied 17 different kinds of filter structures, including some complex structures with embedded ferrite. As a result, the simple structure as shown in Fig exhibited the highest attenuation Integrated absorptive DM transmission-line EMI filter As shown in Fig. 1-11, the integrated differential-mode (DM) transmission-line EMI filter has two kinds of conductors in parallel: nickel as the inner conductor and copper as the outer conductor. Between two nickel layers is the high-relative-permittivity dielectric material, such as the BaTiO 3 based dielectric materials. An insulation layer is placed between nickel and copper layers. 9

37 25um 65um 17um 15um 1mm Insulation High e r dielectric ceramic Insulation Cu Ni Ni Cu Fig Cross-sectional view of integrated DM transmission-line EMI filter. The first DM transmission-line EMI filter prototype is built in 22, as shown in Fig [38]. It is made of a U-shaped structure for two reasons. First, the filter has to be sufficiently long to get a low enough cut-off frequency. Second, an integrated active power module could be put inside the U-shaped structure to save more space. Fig shows the measured transfer gain of the transmission line EMI filter given by [38]. Fig Integrated transmission-line EMI filter prototype [38]. 18 1dB db -1dB -2dB -3dB -4dB db _ 4dB/decade Gain Phase MHz Fig Measured transfer gain of the DM transmission-line filter prototype [38]. 1

38 Compared to reflective LC EMI filter, this absorptive transmission-line EMI filter has four major advantages. First, this transmission-line EMI filter can realize frequency selectivity. Low-frequency power current and high-frequency noise current go through different conductors. Second, because the transmission-line EMI filter converts the highfrequency noise current to losses, it is inherently less dependent on source and load impedances. Third, this transmission-line EMI filter has high attenuation at high frequencies. Fourth, the transmission-line EMI filter is a planar structure, which is compatible with other integrated planar modules that have been developed Integrated absorptive CM & DM transmission-line EMI filter If two more high-frequency attenuators are added to the DM transmission-line EMI filter; two more insulation layers are added between nickel and copper conductors; at the same time the outside nickel layers are connected to ground; this new structure is an integrated common mode and differential mode (CM & DM) transmission-line EMI filter, as shown in Fig It is first proposed in [39]. Due to the CM leakage current requirements, the relative permittivity of the dielectric material used in the two outside CM high-frequency attenuators is much lower than relative permittivity of the dielectric material in DM high-frequency attenuator. For example, N125 dielectric material with relative permittivity around 186 and Y5V dielectric material with relative permittivity around 14 are used in CM and DM high-frequency attenuators, respectively. 11

39 15um 17um 65um 25um 65um 17um 15um dielectric ceramic Insulation Insulation dielectric ceramic Insulation Insulation dielectric ceramic Ni Ni Cu Ni Ni Cu Ni Ni Fig Cross-sectional view of integrated CM & DM transmission-line EMI filter [39]. Fig shows the integrated CM & DM transmission-line EMI filter prototype that was built in 24 by Rengang Chen [4]. Fig and Fig show the measured CM and DM transfer gains of this filter prototype. Fig Integrated CM & DM transmission-line EMI filter prototype. 12

40 Fig Measured DM transfer gain of the integrated CM & DM EMI filter prototype. Fig Measured CM transfer gain of the integrated CM & DM EMI filter prototype Motivation and objective The previous work only gave one design of the transmission-line EMI filter. In order to apply the transmission-line EMI filter in different power electronics systems, the performance of the filter under different materials, dimensions, and source and load impedances must be known. This dissertation focuses on exploring the working mechanisms of the DM transmission-line EMI filter and providing design guidelines for the DM transmission-line EMI filter. The investigation is based on multi-conductor lossy transmission-line model and finite element simulation, and is verified by experimental results. In order to get accurate measurement results of the DM transmission-line EMI filter, the measurement setup and measurement board design are also discussed. This dissertation also focuses on exploring ways to improve the performance of the DM transmission-line EMI filter. Some improved filter structure prototypes of the DM transmission-line EMI filter are built and measured. A broadband DM EMI filter combining the advantages of discrete reflective LC EMI filter and absorptive transmission-line EMI filter is proposed, built and measured. 13

41 1.4. Dissertation outline The remainder of this dissertation will be organized as follows: Chapter 2 will discuss the electromagnetic modeling and finite element simulation of the transmission-line EMI filter. In Chapter 3, the dependence of the high-frequency attenuator s (nickel-dielectricnickel) performance on material properties, dimensions and source and load impedances are studied. In Chapter 4, the dependence of the DM transmission-line EMI filter s performance on material properties, dimensions and source and load impedances are studied. In Chapter 5, the measurement setup of the transmission-line EMI filter is discussed. In Chapter 6, experimental results are given to verify the electromagnetic modeling results in Chapter 3 and Chapter 4. Based on the previous electromagnetic modeling and experimental results, some improved transmission-line EMI filter structures are proposed and discussed in Chapter 7. A broadband DM EMI filter combining the advantages of discrete reflective LC EMI filter and absorptive transmission-line EMI filter is also proposed in this chapter. In Chapter 8, the conclusions are drawn and some future works are suggested. 14

42 Chapter 2: Electromagnetic modeling of integrated transmission-line EMI filter 2.1. Introduction To study the behavior of the integrated transmission-line EMI filter, it is essential to create an electromagnetic model of the filter. This model can be used in a parametric study to identify the most influential factors on the filter s performance. A good model can facilitate the design process of the filter. For power electronics circuits, either the lumped model or the distributed model can be used. If the dimensions of the circuit are much smaller than the wavelength, using a lumped model of the circuit elements is appropriate, since voltage and current waves affect the entire circuit at the same time. If the dimensions of the circuit are comparable to the wavelength, the distributed model is required, since voltage and current waves do not affect the entire circuit at the same time. The circuit must be broken down into unit sections within which the circuit elements are considered to be lumped [41]. For the modeling of the integrated transmission-line EMI filter, a large frequency range (from hundreds of Hz to at least 3MHz) needs to be considered. The wavelength in the structure is greatly reduced due to the application of dielectric material with very high relative permittivity. Therefore, a distributed model is necessary. The two-conductor lossy transmission-line model and the multi-conductor lossy transmission-line model can both be used to model the integrated transmission-line EMI filter. In [36] and [38], the two-conductor lossy transmission-line model is used to model the integrated transmission-line EMI filter. In [4] and [42], the multi-conductor lossy 15

43 transmission-line model is used to model the transmission-line EMI filter. In this chapter, these two modeling methods are reviewed and compared. The results from these two modeling methods are also compared with measurement results. The multi-conductor lossy transmission-line model is chosen for the following parametric study of the filter due to its comprehensiveness and accuracy. The parameters of the transmission-line EMI filter are needed for both two-conductor and multi-conductor transmission-line models. The parameters can be acquired using either the analytical method or finite element analysis. References [36] and [38] propose an analytical model to obtain the parameters for the two-conductor lossy transmissionline model. For the multi-conductor lossy transmission-line model, using an analytical method is very difficult, and finite element analysis is needed to extract the transmissionline parameters. Both Maxwell 2D and Maxwell 2D Extractor simulation software from Ansoft can be used to extract the parameters. Maxwell 2D gives the transmission-line s parameters at one specific frequency for each simulation. Maxwell 2D Extractor can give the parameters for a frequency range for each simulation. The results from Maxwell 2D Extractor are compared with those from Maxwell 2D. The results are very close. Because the Maxwell 2D simulation takes several hours to perform the simulation at one frequency point, it is too time-consuming to perform Maxwell 2D simulation at a wide frequency range, Maxwell 2D Extractor is chosen as the simulation software due to its efficiency and accuracy Two-conductor lossy transmission-line model The transmission-line model represents the filter structure as an infinite series of twoport elementary components, each representing an infinitesimally short segment of the 16

44 filter. Fig. 2-1 is the schematic representation of the elementary component. R is the resistance in both conductors per unit length in Ω/m. L is the inductance in both conductors per unit length in H/m. G is the conductance of the dielectric media per unit length in S/m. C is the capacitance between the conductors per unit length in F/m[36][38]. Fig Equivalent circuit of an infinitesimal section. Applying Kirchhoff s current and voltage laws (KCL & KVL), the following partial differential equations can be obtained for the sinusoidal steady-state condition: dv ( z) = ( R + jω L) I ( z) = I ( z) (2-1) dz di ( z) = ( G + jω C) V ( z) = YV ( z) (2-2) dz After separation of variables and substitution, (2-1) and (2-2) become: 2 d I( z) 2 γ I( z) = (2-3) 2 dz 2 d V ( z) 2 γ V ( z) = (2-4) 2 dz 17

45 γ is the propagation constant of the line: γ = Y = α + jβ. The real part of the propagation constant α is the attenuation constant, representing the losses along the line. The imaginary part β is the phase constant, representing the phase shift along the line. One possible set of solutions for (2-3) and (2-4) is: V ( z) (2-5) + + γ z + γz = V ( z) + V ( z) = V e + V e + + V γ z V +γz I( z) = I ( z) + I ( z) = e e (2-6) R + jωl = = is the characteristic impedance of the two-conductor Y G + jωc transmission-line structure. Boundary conditions to the solutions are determined by the excitation and the source and load impedances connected to the transmission line. The solutions shown in (2-5) and (2-6) have two parts. The first term in both equations is the voltage and current of the incident wave, which is the electrical signal moving from source to load, and the second term is for the reflected wave, which is the signal moving from load to source. The voltage across the line and the current in the line are the sum of all incident and reflected waves. The line can be terminated at any distance from the input with a load impedance equal to the characteristic impedance of the line without any reflections. A reflection occurs at an impedance discontinuity or mismatch. A part of the incident energy is transmitted beyond the discontinuity, and part is reflected. The signal ratio of the incident voltage to the reflected voltage at the load is given by the voltage reflection coefficient: 18

46 V L ρ v = V + = (2-7) L + in which L is the load impedance. This equation shows that for a load of Ω, or a short circuit, the reflection coefficient will be -1 and the voltage at the load is, as it should be. For a load of open circuit, the reflection coefficient is 1, and the voltage at the load is doubled. The input impedance seen by the source depends on the load impedance and the length of line l. The input impedance seen by the source is: in L + tanh γl = (2-8) + tanh γl L For a short circuit load, the input impedance is: in = tanh γl (2-9) and for an open circuit load, it s: in = coth γl (2-1) Transmission line structures can be considered as a two-port network, as shown in Fig The transfer matrix, or ABCD matrix, is one of the most commonly used circuit parameter matrices for two-port networks. ABCD matrix shows the relationship between the output and input voltages and currents. It is very convenient for cascaded structures. + I 1 I 2 + Two port V 1 - network V 2 - Fig Two-port network. 19

47 V1 A = I1 C B V D I 2 2 cosh γl = 1 sinh γl sinh γl V cosh γl I 2 2 (2-11) γl γl e + e where coshγl =, and 2 γl γl e e sinh γl =. 2 The transfer voltage gain (TG) and insertion voltage gain (IG) are often used to describe the attenuation effect of an EMI filter. As shown in Fig. 2-3, the transfer voltage gain is defined as the ratio of the output voltage (V out ) to the input voltage (V in ) of the EMI filter. Insertion gain is defined as the ratio of the load voltage with the filter (V L_w ) to the load voltage without the filter (V L_w/o ): V V out TG = (2-12) in IG L _ w = (2-13) V V L _ w/ o Fig Schematic of transfer gain and insertion gain measurement circuit for transmission-line EMI filter Application of two-conductor lossy transmission-line model to the filter Some approximations have to be made to model the integrated transmission-line EMI filter based on the two-conductor lossy transmission-line theory. First, the high-frequency attenuator (nickel-dielectric-nickel) and the outside copper conductors are taken as two 2

48 two-conductor transmission-line structures; the inside transmission-line structure and the outside transmission-line structure, as shown in Fig. 2-4 [36][38]. R Ni z L Ni z C Ni z G Ni z (a) Outside (b) Inside Fig Equivalent circuit of infinitesimal sections of inside and outside transmission-line structures. The nickel conductors and the copper conductors are connected at the terminal of the filter, forcing the voltages on these two conductors to be the same at the terminal. At high frequencies, eddy currents in the two sets of conductors will be different. This leads to a localized voltage difference on the copper and nickel conductors along the z axis. The magnitude of this voltage difference is assumed to be much smaller than the voltage across the dielectric layer, and could be neglected. The relative permittivity of Al 2 O 3 is much smaller than the relative permittivity of the dielectric (ε Al2O3 =9.8 << ε Y5V =14). Loss generated in the insulation layer is neglected. Any capacitance of the Al 2 O 3 layer is also neglected. Thus: C =, G =. (2-14) Cu Cu The outside transmission-line structure is represented by series impedance only: Cu = R + jωl (2-15) Cu Cu The inside and outside transmission-lines are combined into a single model, as shown in Fig The equivalent series impedance is calculated as: Cu Ni = (2-16) Cu + Ni 21

49 Fig Equivalent circuit of the DM transmission-line EMI filter. The characteristic impedance is: 1 Cu Ni = = (2-17) Y Cu + Ni GNi + jωcni The propagation constant is: Cu Ni γ = Y = ( GNi + jωcni ) (2-18) + Cu Ni The transfer gain and the insertion gain of the DM transmission-line EMI filter are: V TG = V out in = V out Vout 1 = cosh γl + I l out sinh γ cosh γl + L sinh γl (2-19) V IG = V L _ w L _ w/ o = V S Vout L + S L = V in V S + in out in L + S L S + = TG L L in + S in (2-2) with in L + tanh γl =. + tanh γl L Multi-conductor lossy transmission-line model The multi-conductor lossy transmission-line model was used to model the transmission-line EMI filter in [4] and [42]. The model is reviewed here. Let s consider the n-conductor coupled transmission-lines shown in Fig There are n conductors 22

50 from 1 to n. Conductor is a perfect ground plane. The whole system is assumed to be uniform along the z axis [41][43][44]. (a) Longitudinal view (b) Cross-sectional view Fig Multi-conductor transmission-line configuration parallel to the z axis. The equivalent circuit of an infinitesimal cell of the multi-conductor transmission-line system is shown in Fig Fig Equivalent circuit of an infinitesimal section. Apparently, the model should be based on the matrix representation, due to the magnetic and capacitive coupling between more than two conductors. Applying KVL and KCL, the basic voltage and current equations for a multi-conductor transmission-line system can be obtained in matrix form: 2 d V 2 dz = Y V (2-21) 23

51 2 d I 2 dz = Y I (2-22) The voltage and current vectors are: t V = ( V 1 V2 L V n ) (2-23) t I = ( I 1 I 2 L I n ) (2-24) The impedance matrix is: R11 R12 R1n L11 L12 L1n R21 R22 R2n L21 L22 L2n = R + jωl = + jω (2-25) Rn1 Rn2 Rnn Ln 1 Ln2 Lnn where R ii and L ii are self resistance and inductance per unit length, and R ij and L ij are mutual resistance and inductance per unit length. The admittance matrix is: G11 G12 G1 n C11 C12 C1 n G21 G22 G2n C21 C22 C2n Y = G + jωc = + jω (2-26) Gn 1 Gn2 Gnn Cn 1 Cn2 Cnn n where Gii = Gi + Gij, G i is the conductance between the ith conductor to ground, j= 1, j i G ij is the conductance between the ith and jth conductors, C = C + C, C i is the ii i ij j= 1, j i capacitance between the ith conductor to ground, and C ij is the capacitance between the ith and jth conductors. n 24

52 Decoupling of these equations can be achieved through the use of suitable modal transformation matrices. The transformation matrix relating natural voltage V and modal voltage ^ V is introduced as: ^ V = TV (2-27) and similarly, a transformation W between natural current I and modal current ^ I is: ^ I= WI (2-28) Substitution of (2-27) into (2-21) and substitution of (2-28) into (2-22) yields ^ 2 ^ 1 d V = ( ) 2 T YT V (2-29) dz ^ 2 I ^ 1 ( ) 2 d dz = W YW I (2-3) For effective decoupling of equations to take place, the products in parentheses must lead to diagonal matrices. -1 T YT= γ = diag { γ,... γ,... γ } (2-31) k n -1 W YW = γ = diag { γ,... γ,... γ } (2-32) k n When γ 2 is a diagonal matrix and γ i 2 is the eigen value of Y and Y, the following relationship between T and W can be found. 25

53 -1t W=T (2-33) Using (2-31) and (2-32) in (2-29) and (2-3), modal quantities are decoupled: ^ 2 ^ 2 = 2 d V γ V (2-34) dz ^ 2 d I 2 dz ^ 2 = γ I (2-35) The solutions of (2-34) and (2-35) yield ^ ^ ^ -γz z e () i e + γ ( r) V = V + V (2-36) ^ ^ ^ -γz γz () i ( r) I= e I + e + I (2-37) ^ ^ where V () i and V ( r) are the incident and reflected voltage vectors in the modal space, ^ ^ and I () i and I ( r) are the incident and reflected current vectors in the modal space. The modal voltage and current vectors are related through the modal characteristic impedance and admittance matrices. ^ ^ ^ I () i =+Yw V (2-38) () i ^ ^ ^ I =-Y V ( r) w ( r) (2-39) where 26

54 ^ t -1 Y w =(T T)γ (2-4) ^ w ^ w -1 =Y (2-41) Applying (2-27) and (2-28), we obtain ^ -1 (),( i r) = (),( i r) V T V (2-42) ^ -1 (),( i r) = (),( i r) I W I (2-43) The voltages and currents in natural space can be solved by reverse transformation. V = Te T V + Te T V γz -1 + γz -1 () i ( r) ^ I = WYw T ( Te T V Te T V ) -1 γz -1 + γz -1 () i ( r) (2-44) where γz e = diag e e e γ1z γkz γnz {,...,,..., } + γz e = diag e e e + γ1z + γkz + γnz {,...,,..., } (2-45) Defining the characteristic admittance matrix Y w in natural space, ^ Y w = WYwT = Γ =YΓ (2-46) where -1 Γ =TγT = SQRT(Y) (2-47) Inversion of (2-46) yields the characteristic impedance matrix in natural space. 27

55 = Y = Γ =ΓY (2-48) w w We can find the relationship between the voltages and currents in natural space as I I = + Y V () i w () i = Y V ( r) w ( r) V V =+ I () i w () i = I ( r) w ( r) (2-49) If we define EXP( ± z) = e ± γz 1 Γ T T (2-5) employing the new matrices introduced above, we can present a new version of (2-44): w I = EXP(-Γz) V( i) EXP( + V = EXP(-Γz) V ( i) + EXP( + Γz) V ( r) Γz) V ( r) (2-51) The boundary condition at z=, V=V() and I=I(), allows us to determine V () i and V ( r) : 1 V() i = [ V () + wi ()] 2 1 V( r ) = [ V () wi ()] 2 (2-52) Substituting (2-52) into (2-51) yields 1 1 V( z) = [EXP( + Γz) + EXP( Γz)] V() [EXP( + Γz) EXP( Γz)] wi() I( z) = Yw [EXP( + Γz) EXP( Γz)] V() + Yw [EXP( + Γz) + EXP( Γz)] wi() 2 2 (2-53) Particularization at z = l allows us to determine V(l) and I(l) in terms of V() and I(): 28

56 V() l COSH( Γl) SINH( Γl) w V() = () l SINH( l) COSH( l) () I Yw Γ Yw Γ w I (2-54) where the hyperbolic functions of Γz are defined as: 1 COSH( Γz) = [EXP( + Γz) + EXP( Γz)] = T(cosh γz) T 2 1 SINH( Γz) = [EXP( + Γz) EXP( Γz)] = T(sinh γz) T (2-55) The square matrix of order 2n in (2-54) establishes the relationship between the output and input voltages and currents. The transfer matrix or ABCD matrix is: T A A B = C D (2-56) where the sub-matrices are A = COSH( Γl) B = SINH( Γl) C= Y SINH( Γl) w D= Y COSH( Γl) = A w w w t (2-57) Inversing T A, the input voltages and currents can be expressed by those of the output. V( ) COSH( ) SINH( ) 1 V( l) Γl Γl w V( l) = T = (2-58) A I() I( l) YwSINH( Γl) YwCOSH( Γl) w I( l) From (2-54), we can derive the immittance matrix I() V() Y11 Y12 V() = = () l Y () l () l I V Y21 Y22 V (2-59) 29

57 where Y Y B A DB Y COTH( Γl) = 22 = = = w Y Y B DB A C Y CSCH( Γl) = 21 = = = w (2-6) The matrix hyperbolic functions are defined as: COTH( Γz) = SINH ( Γz)COSH( Γz) = T(coth γz) T 1 1 CSCH( Γz) = SINH ( Γz) = T(csch γz) T 1 1 (2-61) Inversion of (2-59) leads to the impedance matrix: V() I() I() = = () l () l () l V I I (2-62) where = = COTH( Γl) w = = CSCH( Γl) w (2-63) Considering the transmission-lines are terminated at z = l by a group of loads defined by the load impedance or admittance matrices, V() l = I() l L I V Y V 1 () l = L () l = L () l (2-64) The input impedance matrix at z = can be easily derived by using the T A matrix. V() = I I () (2-65) Where I is the line input impedance: 3

58 = Γl + Γl Γl + Γl 1 I {[COSH( ) L SINH( ) w] [COSH( ) w SINH( ) L] } w = 1 ( A LC) ( LD B) (2-66) Application of multi-conductor lossy transmission-line model to the filter Let s consider the cross-section of an integrated DM transmission-line EMI filter, shown in Fig. 2-8; this structure can be treated as a four-conductor coupled transmissionline. The equivalent circuit of an infinitesimal section of this four-conductor coupled transmission-line is shown in Fig To solve this circuit, the input source vector, input source impedance matrix, and load impedance matrix need to be found using the boundary conditions, as shown in Fig µm 65µm 17µm 15µm 1mm In su lation High e r dielectric ceramic In su lation Cu Ni Ni Cu Fig Cross-sectional view of integrated DM transmission-line EMI filter. Fig Equivalent circuit of an infinitesimal section. 31

59 Fig Schematic of insertion gain measurement circuit for integrated DM transmission-line EMI filter. The boundary conditions at z = for the four-conductor transmission-line can be expressed by the following equations: V () = V () = V 1 S [ I () + I ()] V () = V () = [ I () + I () + I () + I ()] S [ I () + I () + I () + I ()] SG 3 4 SG (2-67) Rewriting (2-67) in matrix form yields V ( ) = V I 4 () (2-68) S S where the input source vector is: [ V ] t V = (2-69) S S V S The source impedance matrix is: S + SG S + SG SG SG S + SG S + SG SG SG = (2-7) S4 SG SG SG SG SG SG SG SG The boundary condition at load side is: 32

60 V ( l) = V ( l) = [ I ( l) + I ( l)] 1 V ( l) = V ( l) = [ I ( l) + I ( l) + I ( l) + I ( l)] L + [ I ( l) + I ( l) + I ( l) + I ( l)] LG 3 4 LG (2-71) Rewriting (2-71) in matrix form yields V( l) = L4 I( l) (2-72) Where the load impedance matrix is: L + LG L + LG LG LG L + LG L + LG LG LG = (2-73) L4 LG LG LG LG LG LG LG LG The input impedance matrix at z = can be found as: 1 I4 = {[COSH( Γ4l) L4 + SINH( Γ4l) w4] [COSH( Γ4l) w4 + SINH( Γ4l) L4] } w4 (2-74) where w4 is the characteristic impedance matrix of the four-conductor transmission-line structure, defined by (2-48). Knowing the input impedance I4, the input currents and voltages can be solved by: I() = ( + ) V 1 S4 I4 S V() = I() = ( + ) V 1 I4 I4 S4 I4 S (2-75) The output current and voltage can be found using the ABCD matrix: V() l COSH( Γ4l) SINH( Γ4l) w 4 V() () l = 4SINH( 4l) 4COSH( 4l) 4 () I Yw Γ Yw Γ w I (2-76) The voltage on load is: 33

61 V load = V l) V ( ) (2-77) 1( 3 l The transfer gain of the EMI filter is: V1 ( l) V3( l) TG = (2-78) V () V () 1 3 The insertion gain of the EMI filter is: V IG = V 1 ( 3 S l) V ( l) L + S L (2-79) 2.2. Model parameter extraction As stated at the beginning of this chapter, to get the attenuation of the DM transmission-line EMI filter from the two-conductor or multi-conductor transmission-line model, the model parameters, such as the R Ni and L Ni of the two-conductor transmission-line model and the impedance and conductance matrices of the multiconductor transmission-line model, must be obtained. Although the analytical method can be used for two-conductor transmission-lines, an analytical solution is difficult for multiconductor transmission-lines. The finite element simulation software Maxwell 2D and Maxwell 2D Extractor can be used to extract the parameters for the multi-conductor lossy transmission-line model. 34

62 Parameter extraction based on analytical method The skin depth of a conductor is a measure of the distance an AC current can penetrate beneath the surface of a conductor [36][38]. It is given by: 1 δ = (2-8) σπµ f µ r The AC resistance of the two copper conductors per unit length is approximated as: R Cu = σ Cu 2 w Cu _ eff t Cu (2-81) where w Cu _ eff w, = 2δ Cu 2δ Cu w, with δ Cu =, 2δ < w Cu σ Cu 1 πµ f The AC resistance of the two nickel conductors per unit length is approximated as: R Ni = 2 σ w t (2-82) Ni Ni _ eff wheret Ni _ eff t Ni, δ = δ Ni, δ Ni Cu t < t Ni Ni, with δ Ni = σ Ni 1 πµ µ Ni f The inductance per unit length for two planar conductors is comprised of internal inductance and external inductance, which are given as: L int 2 t = µ µ (2-83) r1 3 w where µ r1 is the relative permeability of the conductor, t is the thickness of the conductor, and w is the width of the conductor. 35

63 L d = µ µ (2-84) r w ext 2 where µ r2 is the relative permeability of the material between the two conductors, and d is the distance between the two conductors, as shown in Fig The above equations are valid only if the width is much greater than the thickness, so that the magnetic path length can be approximated as two times the width. The inductance per unit length for copper conductors of the DM transmission-line EMI filter is: L 2 tcu dcu = Lint_ Cu + Lext _ Cu = µ + µ (2-85) 3 w w Cu under the assumption that the relative permeability of the layers between copper conductors is 1. The inductance per unit length for nickel conductors is: L 2 t Ni d Ni = Lint_ Ni + Lext _ Ni = µ µ Ni + µ (2-86) 3 w w Ni where d = t. Ni die Insulation Dielectric Insulation Fig Cross-sectional view of DM transmission-line EMI filter. 36

64 The capacitance per unit length of the inside transmission-line structure is: C die w = ε ε die (2-87) t die The conductance per unit length of the dielectric material is: G = 2π tan δ (2-88) die fc die die where tan δ die is the loss factor of the dielectric material Parameter extraction based on Maxwell 2D Maxwell 2D is an interactive software package for analyzing electric and magnetic fields in structures with uniform cross-sections or full rotational symmetry, where the field patterns of the entire device can be analyzed by modeling the field patterns within its cross-section. Maxwell 2D can be used to extract the parameters of the integrated DM transmission-line EMI filter under two assumptions. The first is that the length of the filter is at least one order of magnitude larger than the dimensions of the cross-section, namely the width and the thickness. Second, the cross-sections are uniform along the length of the filter. Fig shows the cross sectional view of the DM transmission-line EMI filter in Maxwell 2D. Table 2-1 shows the required material properties and dimensions of the filter used in Maxwell 2D simulation. 37

65 Fig Maxwell 2D model of the DM transmission-line EMI filter. TABLE 2-1 TECHNICAL PARAMETERS OF INTEGRATED DM TRANSMISSION-LINE EMI FILTER PARAMETERS VALUE Width (mm) 1 Copper conductor Thickness (µm) 25 Conductivity (Ω -1 /m) 5.8e7 Relative permeability 1 Width (mm) 1 Nickel conductor Dielectric Y5V Thickness (µm) 17 Conductivity (Ω -1 /m) 1.45e7 Relative permeability 6[51] Thickness (µm) 15 Relative permittivity 14 Loss factor 4% Thickness (µm) 65 Insulation Al 2 O 3 Relative permittivity 9.8 Loss factor.1% 38

66 Fig shows the Maxwell 2D simulation result of the current density distribution within the copper and nickel conductors of the DM transmission-line EMI filter at 1kHz. The graphs are enlarged to show the details at the edge and the corner. The Al 2 O 3 insulation layer between the copper and nickel conductors is omitted from the graph. The current density is uniform in both the copper and nickel conductors. The current density in copper is higher than that in nickel because copper has higher conductivity. Fig shows the enlarged current density distribution in the copper and nickel conductors of the DM transmission-line EMI filter at 1MHz. The insulation layer is not shown in the graph. Fig shows that the current density distributions are very different in the copper and nickel conductors. In the copper conductor, the current crowds around the vertical edges. However, in the nickel conductor, the current density is the highest at the inside horizontal edge. Copper Nickel Fig Current density in copper and nickel conductors at 1 khz from Maxwell 2D. 39

67 Copper Nickel Fig Current density in copper and nickel conductors at 1MHz from Maxwell 2D. After the simulation, the extracted capacitance (F/m) matrix is obtained: 1.84e e e 1 3.7e e e e e 1 C = (2-89) 2.47e e e e 9 3.7e e e e 9 The conductance is calculated according to: G i = G ( f) = ωc ( f)tan δ ( f) ij i j ij ij G = G + G ii i ij j= 1, j i n (2-9) The resistance matrix (Ω/m) and inductance matrix (H/m) are dependent upon frequency. The matrices at 1MHz are: R = (2-91)

68 8.74e e e e e e 7 8.3e e 7 L = (2-92) 7.72e 7 8.3e e e e e 7 8.3e e 7 According to the symmetry of the structure, the resistance and inductance matrices are symmetric: R = R, L = L R = R = R = R, L = L = L = L R = R = R = R, L = L = L = L R = R, L = L R = R, L = L R = R, L = L (2-93) The Maxwell 2D simulation must run many times at different frequencies to get the resistance and inductance matrices over a wide frequency range. This requires a lot of simulation time. To simplify the simulation process, Maxwell 2D Extractor (Maxwell Q2D), which can give the resistance and inductance matrices over a wide frequency range in one simulation, are investigated Parameter extraction based on Maxwell 2D Extractor The Maxwell 2D Extractor (Maxwell Q2D) is quick 2D simulation software that can give parameters at different frequencies in one simulation. Fig and Fig show the resistances and inductances curves from Maxwell 2D Extractor. The mutual resistance has negative values, such as R13, R24. This is because the current in one conductor will make the current distribution in another conductor more even, reducing 41

69 the total loss generated in the other conductor. This is related to the relative position and the material properties of the conductors. Resistance per unit length Ohm/m R11=R44 R12=R21=R34=R43 R13=R31=R24=R42 R14=R41 R22=R33 R23=R E+4 1.E+5 1.E+6 1.E+7 1.E+8 Fig Resistances vs. frequency from Maxwell 2D Extractor for DM transmission-line EMI filter. Inductance per unit length H/m 1.2E-6 1.1E-6 1.E-6 9.E-7 8.E-7 7.E-7 L11=L44 L12=L21=L34=L43 L13=L31=L24=L42 L14=L41 L22=L33 L23=L32 6.E-7 1.E+4 1.E+5 1.E+6 1.E+7 1.E+8 Fig Inductances vs. frequency from Maxwell 2D Extractor for DM transmission-line EMI filter. The Maxwell 2D Extractor performs one field solution at the specified maximum (nominal) frequency. From this initial solution, the software then uses AWE [6] 42

70 (asymptotic waveform evaluation) to extrapolate an entire bandwidth of solution information by determining the poles and zeros of the field solution. While solutions can be viewed at any frequency, the accuracy of solutions above the nominal frequency cannot be guaranteed. Here, the accuracy of the Maxwell 2D Extractor simulation is verified by comparing the results with the results from Maxwell 2D simulation. The specified nominal frequency in Maxwell 2D Extractor is the highest frequency of interest, 1MHz. Fig and Fig show that the resistances and inductances from Maxwell 2D Extractor agree very well with the results from Maxwell 2D. Since Maxwell 2D Extractor is much faster and just as accurate, it s a much more efficient way to extract the parameters of the filter. Ohm/m Resistance R11=R44 R12=R21=R34=R43 R13=R31=R24=R42 R14=R41 R22=R33 R23=R32 1kHz(Maxwell 2D) 1kHz(Maxwell 2D) 1MHz(Maxwell 2D) 1MHz(Maxwell 2D) 1MHz(Maxwell 2D) E+4 1.E+5 1.E+6 1.E+7 1.E+8 Fig Comparison of resistances from Maxwell 2D and Maxwell 2D Extractor. 43

71 1.2E-6 Inductances L11=L44 1.1E-6 L12=L21=L34=L43 L13=L31=L24=L42 L14=L41 1.E-6 L22=L33 L23=L32 H/m 9.E-7 1kHz (Maxwell 2D) 1kHz (Maxwell 2D) 8.E-7 1MHz (Maxwell 2D) 1MHz (Maxwell 2D) 1MHz (Maxwell 2D) 7.E-7 6.E-7 1.E+4 1.E+5 1.E+6 1.E+7 1.E+8 Fig Comparison of Inductances from Maxwell 2D and Maxwell 2D Extractor Comparison of two-conductor and multi-conductor transmission-line models Compared with multi-conductor transmission-line models, the two-conductor transmission-line model is relatively simple and easy to understand. However, it neglects several aspects of the filter, such as the magnetic coupling of the conductors, the proximity effect, and the capacitance due to the insulation layer. The multi-conductor transmission-line model takes all these into account, and is therefore a more complete and more accurate model, as summarized in Table 2-2. TABLE 2-2 COMPARISON OF TWO-CONDUCTOR AND MULTI-CONDUCTOR MODEL Magnetic coupling Proximity effect Insulation capacitance Simplicity Two-conductor model Multi-conductor model 44

72 Fig and Fig. 2-2 show the measured insertion gain and transfer gain of a DM transmission-line EMI filter prototype. The measurement setup is discussed in detail in Chapter 5. The insertion gains and transfer gains from two-conductor and multiconductor transmission-line models are drawn in the same graph for comparison. It shows that the results from the multi-conductor transmission-line model agree very well with the measurement result. However, the results from the two-conductor transmissionline model have some discrepancies from the measured results. The multi-conductor transmission-line model is chosen to perform the parametric analysis of the DM transmission-line EMI filter. Gain (db) Insertion Gain E+3 1.E+4 1.E+5 1.E+6 1.E+7 1.E+8 Measurement Two-conductor TL model Multi-conductor TL model Phase (Degree) Insertion Gain 1.E+3 1.E+4 1.E+5 1.E+6 1.E+7 1.E+8 Measurement Two-conductor TL model Multi-conductor TL model Fig Insertion gains of DM transmission-line EMI filter. 45

73 Gain (db) Transfer Gain 1.E+3 1.E+4 1.E+5 1.E+6 1.E+7 1.E+8 Measurement Two-conductor TL model Multi-conductor TL model Phase (Degree) Transfer Gain E+3 1.E+4 1.E+5 1.E+6 1.E+7 1.E+8 Measurement Two-conductor TL model Multi-conductor TL model Fig Transfer gains of DM transmission-line EMI filter Summary The two-conductor lossy transmission-line model and the multi-conductor lossy transmission-line model were reviewed and compared in this chapter. The two-conductor lossy transmission line model was relatively simple and easy to understand. However, it neglected several aspects of the filter, such as the magnetic coupling of the conductors, the proximity effect, and the capacitance due to the insulation layer. The multi-conductor transmission-line model took all these into account, and was therefore a more complete and more accurate model. The comparison of the models with the measured results showed that the result from the multi-conductor transmission-line model was more accurate than the result from the two-conductor transmission-line model. The parameters of the two-conductor transmission-line model can be obtained by an analytical method. The parameter matrices of the multi-conductor transmission-line model can be obtained using finite element simulation software Maxwell 2D or Maxwell 2D Extractor. The results from these two softwares were compared. The Maxwell 2D Extractor is chosen because compared with Maxwell 2D, it is much faster and just as 46

74 accurate. However, for Maxwell 2D Extractor, the simulation frequency has to be set as the maximum interested frequency, since the accuracy of the results above the simulation frequency can not be guaranteed. 47

75 Chapter 3: Parametric characterization of high-frequency attenuator 3.1. Introduction It has been shown in Chapter 2 that the multi-conductor lossy transmission-line model is an accurate and effective model for an integrated transmission-line EMI filter. It can help us understand the working mechanisms of the integrated transmission-line EMI filter. It can also be used to show the influence of dimensions and material properties on the filter s performance. These are essential for the design of the filter. The high-frequency attenuator (nickel-dielectric-nickel) carries the high-frequency noise current and gives the attenuation of the filter. Before using the multi-conductor lossy transmission-line model to analyze the DM transmission-line EMI filter, it is necessary to first, analyze the high-frequency attenuator. In this chapter, the insertion gain, transfer gain, input and output currents and input impedance of the high-frequency attenuator under nominal design are shown. Parametric study of the high-frequency attenuator is also carried out based on the multi-conductor lossy transmission-line model. There are totally eight parameters that are studied about the high-frequency attenuator: The length and width; The thicknesses of the nickel layer and the dielectric layer; The conductivity and relative permeability of nickel material; The relative permittivity and loss factor of the dielectric material. 48

76 3.2. Insertion gain, transfer gain, currents and input impedance The high-frequency attenuator is a structure with dielectric material of high relative permittivity and lossy conductors. It has a large capacitance. It can also generate highfrequency losses. The insertion gain, transfer gain, input and output currents and input impedance of the high-frequency attenuator under nominal design are given below. The dimensions and material properties of the high-frequency attenuator under nominal design are shown in Table 3-1. The nominal value of the length is chosen as 1cm, since the longest dielectric sheet that is available in the packaging lab of CPES is 12cm. The width is chosen to be one order of magnitude smaller than the length. The nickel conductor thickness is chosen as 17µm according to the design of the first prototype of the DM transmission-line EMI filter. Nickel s relative permeability can be any value between 5 and 6 [45]. It depends a lot on the deposition condition for electroplated nickel, such as the solution concentration, current density, and temperature [46][47]. The maximum value 6 is chosen as the nominal value for nickel s relative permeability. BaTiO 3 based dielectric material Y5V is used as the high-relative-permittivity dielectric material in the first transmission-line filter prototype. Among all dielectric materials that are available in the packaging lab of CPES, Y5V has the highest relative permittivity, which is 14, under room temperature. However, its relative permittivity decreases rapidly with increased electric field strength and temperature [48]. 14 is chosen as the nominal value for the relative permittivity of the dielectric material. The thinnest dielectric sheet in lab is 15µm. Therefore, 15µm is chosen as the nominal value for the dielectric layer s thickness. The nominal values of the conductivity of nickel and the loss factor of the 49

77 dielectric material are chosen according to the measurement results in Chapter 6. More discusses about the material properties are also in Chapter 6. TABLE 3-1 NOMINAL DESIGN OF AN HIGH-FREQUENCY ATTENUATOR PARAMETERS VALUE Length (cm) 1 Dimensions Width (mm) 1 Conductor thickness (µm) 17 Dielectric thickness (µm) 15 Conductor Dielectric Conductivity (Ω -1 /m) 1.45e7 Relative permeability 6 Relative permittivity 14 Loss factor 4% Fig. 3-1 shows the high-frequency attenuator in a circuit. S and L are the source and load impedances, which are assumed to be 5Ω unless otherwise specified. The insertion gain of the high-frequency attenuator is shown in Fig F c1 is the cut-off frequency of the high-frequency attenuator, at which the insertion gain of the filter is -3dB. Starting from f c1, the effect of the large capacitor starts to show, the roll-off slope of the curve between f c1 and f c2 is -2dB/dec, which is the roll-off slope of a first-order capacitive filter. In the frequency range below f c2, the high-frequency attenuator can be modeled as a simple capacitor. The capacitance can be obtained from the simple parallel-plate capacitor equation A C = ε, where A is the area of the nickel conductor, ε die and t die ε die tdie are the relative permittivity and thickness of the dielectric layer. 5

78 Above f c2, the effect of the losses in nickel conductors starts to show, this increases the attenuation of the high-frequency attenuator. The roll-off slope is greatly increased above f c2. The transfer gain of the high-frequency attenuator is shown in Fig Since the transfer gain is the ratio of the output voltage to input voltage of the high-frequency attenuator, the effect of the large capacitance is not shown. The attenuation shown in the transfer gain curve comes only from the losses in the nickel conductors. The insertion gain and transfer gain of a high-frequency attenuator with copper conductors are shown in Fig. 3-4 and Fig. 3-5 respectively. The transmission-line effects at high frequencies are shown in these two graphs. The insertion gain and transfer gain curves of the high-frequency attenuator with nickel conductors are very smooth, because the high losses in nickel conductors damp the transmission-line effects. The highfrequency attenuation of the copper-conductor attenuator is greatly reduced, which confirms that the high-frequency attenuation of the nickel-conductor attenuator comes from the losses in the nickel conductors. Fig Schematic of measurement setup for high-frequency attenuator. 51

79 Insertion Gain: Gain 2 Insertion Gain: Phase Magnitude (db) Phase (d) f c1 f c2-2 f c1 f c Fig Insertion gain of high-frequency attenuator. 2 Transfer Gain: Gain 2 Transfer Gain: Phase Magnitude (db) -4-6 Phase (d) Fig Transfer gain of high-frequency attenuator. Insertion Gain: Gain 2 Insertion Gain: Phase Magnitude (db) Phase (d) Fig Insertion gain of high-frequency attenuator with copper conductor. 52

80 14 Insertion Gain: Gain 2 Insertion Gain: Phase Magnitude (db) Phase (d) Fig Transfer gain of high-frequency attenuator with copper conductor. The input impedance in of the high-frequency attenuator with 5Ω load is shown in Fig At low frequency, the high-frequency attenuator is like a large capacitor. The input impedance is equivalent to a RC parallel circuit. At high frequency, the transmission line effect shows. The input impedance is equivalent to the input impedance of a lossy two-conductor transmission-line structure with a 5Ω load. Fig. 3-7 shows the input and output currents of the high-frequency attenuator over frequency, under the assumption that the voltage source is 1V. 1 2 Input impedance: Magnitude Input impedance: Phase -1 Magnitude (A) Phase (d) Fig Input impedance of high-frequency attenuator. 53

81 .2 DM Current Distribution: Magnitude 2 DM Current Distribution: Phase Magnitude (A) I in I out Phase (d) Fig Input and output currents of high-frequency attenuator Parametric study of the high-frequency attenuator To facilitate the design of the transmission-line filter, it is necessary to know the dependence of the filter performance on dimensions and material properties. Before the parametric study on the DM transmission-line EMI filter, the parametric study of the high-frequency attenuator is carried out based on the multi-conductor lossy transmissionline model. There are eight parameters of the high-frequency attenuator that can be studied. The nominal values and sweep range of these parameters are summarized in Table 3-2. For the parametric study of the high-frequency attenuator, only one parameter changes every time, the other parameters keep the nominal values. TABLE 3-2 NOMINAL VALUE AND SWEEP RANGE OF AN HIGH-FREQUENCY ATTENUATOR PARAMETERS NOMINAL VALUE SWEEP RANGE Length (cm) 1 5~2 Dimensions Width (mm) 1 1~1 Conductor thickness (µm) 17 1~25 Dielectric thickness (µm) 15 15~6 54

82 Conductor Dielectric Conductivity (Ω -1 /m) 1.45e7 1.e6~5.8e7 Relative permeability 6 1~6 Relative permittivity 14 1~14 Loss factor 4%.1%~1% Influence of dimensions The length, width, thicknesses of the nickel and dielectric layers of the highfrequency attenuator can influence its attenuation. Fig. 3-8 illustrates the dimensions of the high-frequency attenuator. Fig. 3-9 shows the insertion gains of the high-frequency attenuator for different lengths. The length influences both f c1 and f c2. The roll-off slope of the insertion gain in the frequency range above f c2 also depends on the length. For a longer structure, f c1 and f c2 move to lower frequency; the roll-off slope increases. Longer structures give more attenuation. Fig. 3-1 shows the transfer gains of the high-frequency attenuator for different lengths. Longer structures give lower cut-off frequency and larger roll-off slope. Fig Dimensions of high-frequency attenuator. 55

83 Insertion Gain: Gain 2 Insertion Gain: Phase Magnitude (db) Phase (d) length=.5m length=.1m length=.2m Fig Insertion gain of high-frequency attenuator vs. length. 5 Transfer Gain: Gain 2 Transfer Gain: Phase 15 1 Magnitude (db) -5-1 Phase (d) Length=.5m Length=.1m Length=.2m Fig Transfer gain of high-frequency attenuator vs. length. Fig shows the insertion gain curves for high-frequency attenuators of different widths. A wider structure has a larger capacitance, thus, gives a lower f c1. However, f c2 hardly changes for different widths. The roll-off slope doesn t change much either. The wider structure has more attenuation if the other parameters are kept the same. Fig shows the transfer gain curves of different widths. For widths changing from 1mm to 1mm, the transfer gain curves are similar. 56

84 Insertion Gain: Gain 2 Insertion Gain: Phase Magnitude (db) Phase (d) width=1mm width=2mm width=5mm width=1mm Fig Insertion gain of high-frequency attenuator vs. width. 2 Transfer Gain: Gain 2 Transfer Gain: Phase Magnitude (db) -4-6 Phase (d) width=1mm width=2mm width=5mm width=1mm Fig Transfer gain of high-frequency attenuator vs. width. If the area (length width) of the high-frequency attenuator stays the same, as shown in Fig. 3-13, the capacitance of these high frequency attenuators are the same. F c1 doesn t move, but f c2 is lower for a narrower and longer structure. The roll-off slope is larger too for a narrower and longer structure. Thus, a narrower and longer structure gives much more attenuation than the wider and shorter structure with the same area. The same conclusion can be drawn from the transfer gain curves in Fig

85 Insertion Gain: Gain 2 Insertion Gain: Phase Magnitude (db) Phase (d) W=1mm, L=.5m W=5mm, L=.1m W=2mm, L=.25m Fig Insertion gain of high-frequency attenuator vs. length & width. 5 Transfer Gain: Gain 2 Transfer Gain: Phase 15 1 Magnitude (db) -5-1 Phase (d) width=1mm, length=.5m width=5mm, length=.1m width=2mm, length=.25m Fig Transfer gain of high-frequency attenuator vs. length & width. Fig shows the influence of the conductor layer s thickness on the highfrequency attenuator s insertion gain curve. The transmission line effect starts to show from around the MHz range. If the conductor thickness is always larger than the skin depth from several MHz to 1MHz, the attenuation curves are all the same. As shown in Fig. 3-15, for all conductor thicknesses larger than 5µm, the attenuation curves are almost the same. For conductor thickness smaller than 5µm, the attenuation curves show a difference at the frequency range where the thickness of the conductor is smaller than the 58

86 skin depth. Fig shows the skin depth of the nickel conductor based on the conductivity value 1.45e7 and relative permeability value 6. The conclusion is to have a conductive thickness larger than 5µm with the attenuations all the same. For conductive thickness smaller than 5µm, a smaller thickness gives more attenuation. The same conclusions can be drawn from the transfer gain curves in Fig Insertion Gain: Gain 2 Insertion Gain: Phase Magnitude (db) t Ni =25um t Ni =17um t Ni =1um t Ni =5um t Ni =3um t Ni =2um t Ni =1.5um t Ni =1um Phase (d) Fig Insertion gain of high-frequency attenuator vs. thickness of conductor. 1 3 skin depth of nickel conductor 1 2 (um) Fig Skin depth of nickel. 59

87 2 Transfer Gain: Gain 2 Transfer Gain: Phase 15 1 Magnitude (db) t Ni =25um t Ni =17um t Ni =1um t Ni =5um t Ni =3um t Ni =2um t Ni =1.5um t Ni =1um Phase (d) Fig Transfer gain of high-frequency attenuator vs. thickness of conductor. Fig shows the influence of dielectric layer s thickness on the insertion gain. A thinner dielectric layer will increase the capacitance value. Dielectric layer thickness has influence on f c1, f c2, and the roll-off slope. A thinner dielectric layer gives more attenuation. With a thinner dielectric layer, the length of the structure can be reduced to get the same attenuation. In Fig. 3-2, the insertion gain curve for the dielectric thickness of 15µm and length of 1cm is shown. If both the dielectric thickness and the length are reduced to one tenth of the original values, f c1 stays the same. However, f c2 shifts to a higher frequency and the roll-off slope decreases. If the length is one fourth of the original value, f c2 is close to the original value and the attenuation at the high-frequency range is close to the original design, although f c1 is lower. 6

88 Insertion Gain: Gain 2 Insertion Gain: Phase Magnitude (db) t die =5um Phase (d) t die =15um t die =3um -1 t die =6um Fig Insertion gain of high-frequency attenuator vs. thickness of dielectric layer. 2 Transfer Gain: Gain 2 Transfer Gain: Phase Magnitude (db) t die =5um t die =15um t die =3um t die =6um Phase (d) Fig Transfer gain of high-frequency attenuator vs. thickness of dielectric layer. Insertion Gain: Gain 2 Insertion Gain: Phase 15 1 Magnitude (db) -5-1 t die =15um, length=.1m t die =15um, length=.1m Phase (d) t die =15um, length=.25m Fig Insertion gain of high-frequency attenuator vs. thickness of dielectric layer & length. 61

89 Influence of conductor s properties Nickel is used as the conductor in the high-frequency attenuator. The conductor s conductivity, relative permeability will influence the attenuation of the high-frequency attenuator. Fig shows influence of the conductor s relative permeability on the attenuation. It is clearly shown that the conductor s relative permeability has no influence on f c1. However, it will influence f c2 and the roll-off slope. A higher relative permeability gives a lower f c2 and larger roll-off slope. For non-magnetic conductor (µ r =1), the attenuation is very limited at high frequency range. This is easy to understand, since the higher the relative permeability, the smaller the skin depth of the conductor 1 ( δ = ). Smaller skin depth will give larger loss in the conductor, which will σπµ f µ r give more attenuation to the high-frequency attenuator. Fig shows influence of the conductor s relative permeability on the transfer gain. Insertion Gain: Gain 2 Insertion Gain: Phase Magnitude (db) µ r =1 µ r =2 µ r =5 Phase (d) µ r =1 µ r = µ r = Fig Insertion gain of high-frequency attenuator vs. relative permeability of the conductor. 62

90 2 Transfer Gain: Gain 2 Transfer Gain: Phase 15 1 Magnitude (db) µ r =1 µ r =2 µ r =5 Phase (d) µ r =1 µ r =3 µ r = Fig Transfer gain of high-frequency attenuator vs. relative permeability of the conductor. Fig and Fig shows the influence of the conductor s conductivity on the attenuation of the high-frequency attenuator. The highest conductivity studied is 5.8e7 S/m, which is the conductivity of copper. However, the red curve is not the insertion gain of the high-frequency attenuator with copper conductor, because the relative permeability is 6, not e7 S/m is the nominal conductivity of nickel. 1e6 S/m is about one order smaller than the nominal conductivity of nickel. The results show that lower conductivity gives more attenuation, simply due to the fact that lower conductivity gives more losses in the conductor. Insertion Gain: Gain 2 Insertion Gain: Phase Magnitude (db) Phase (d) σ Ni =5.8e7 σ Ni =1.45e7 σ Ni =1e Fig Insertion gain of high-frequency attenuator vs. conductivity of the conductor. 63

91 2 Transfer Gain: Gain 2 Transfer Gain: Phase Magnitude (db) Phase (d) σ Ni =5.8e7 σ Ni =1.45e7 σ Ni =1e Fig Transfer gain of high-frequency attenuator vs. conductivity of the conductor Influence of dielectric material s properties A high-relative-permittivity dielectric material is used in the high-frequency attenuator structure. It gives the large capacitance in the structure. Fig and Fig shows the influence of the dielectric material s relative permittivity on the insertion gain and transfer gain curves. F c1, f c2 and the roll-off slope all depend on the relative permittivity of the dielectric material. Higher relative permittivity gives lower f c1 and f c2, and larger roll-off slope. Insertion Gain: Gain 2 Insertion Gain: Phase Magnitude (db) ε r =1 ε r =3 ε r =5 ε r =1 ε r =14 Phase (d) Fig Insertion gain of high-frequency attenuator vs. relative permittivity of dielectric. 64

92 2 Transfer Gain: Gain 2 Transfer Gain: Phase 15 1 Magnitude (db) ε r =1 ε r =3 ε r =5 ε r =1 ε r =14 Phase (d) Fig Transfer gain of high-frequency attenuator vs. relative permittivity of dielectric. There are also losses generated in the dielectric layer. The loss factor of the dielectric material is a parameter describing how much loss is generated in the dielectric material. Fig shows the insertion gain curve versus the loss factor of the dielectric material. A higher loss factor gives more loss, thus, more attenuation to the high-frequency attenuator. However, losses in the dielectric layer are much smaller compared with the losses generated in the nickel layer. The differences in the attenuations are very small with two orders of difference in the loss factor. We can draw the conclusion that the influence of the dielectric material s loss factor on the attenuation of the high-frequency attenuator is negligible. Same conclusion can be drawn from the transfer gain curves shown in Fig

93 Insertion Gain: Gain 2 Insertion Gain: Phase Magnitude (db) Phase (d) Loss factor=.1% Loss factor=1% Loss factor=1% Fig Insertion gain of high-frequency attenuator vs. loss factor of dielectric. 2 Transfer Gain: Gain 2 Transfer Gain: Phase Magnitude (db) -4-6 Phase (d) Loss factor=.1% Loss factor=1% Loss factor=1% Fig Transfer gain of high-frequency attenuator vs. loss factor of dielectric Dependence on source and load impedances The above parametric study is under the condition that the source and load impedances are both 5Ω. When the source and load impedances change, the filter performance will be different. In real applications, the input impedance of the converter is never 5Ω. Therefore, it is important to know the performance of the filter under different source and load impedances. 66

94 Fig shows the insertion gain curves of the high-frequency attenuator at different resistive loads. A larger resistance gives more attenuation. For resistance that is larger than 5Ω, the differences of the attenuations are very small. As the load resistance reduces, f c1 gets higher, while f c2 stays the same. The effect of the large capacitor gradually decreases. For a load that is close to short circuit (1-6 Ω), the load impedance is smaller than the impedance of the large capacitor. Thus, the attenuation from the large capacitance doesn t exist any more. However, the high frequency attenuation from the losses in nickel layers is still there. The attenuation in the high frequency range is not influenced by the load impedance. 2 Insertion Gain: Gain 2 Insertion Gain: Phase Magnitude (db) L =1 6 Ohm L =5 Ohm L =1 Ohm L =1 Ohm L =.1 Ohm L =1-6 Ohm Phase (d) Fig Insertion gain of high-frequency attenuator vs. resistive load impedance. Fig. 3-3 shows the insertion gain curves of the high-frequency attenuator at different inductive loads. The insertion gain curve with 5Ω load is put in the same graph for comparison. The larger inductance gives more attenuation. For inductive load as small as 1nH, there is still some attenuation in the high frequency range that comes from the losses in the nickel conductors. 67

95 2 Insertion Gain: Gain 2 Insertion Gain: Phase 15 Magnitude (db) L =5 Ohm L =1mH L =1uH L =1uH L =1nH L =1nH Phase (d) Fig Insertion gain of high-frequency attenuator vs. inductive load impedance. Fig shows the insertion gains of the high-frequency attenuator at different capacitive loads. Smaller capacitance gives more attenuation. For a large capacitive load, for example, the input capacitance of the PFC circuit, which is usually several µf, there is still some attenuation in the high frequency range that comes from the losses in the nickel conductors. 2 Insertion Gain: Gain 2 Insertion Gain: Phase Magnitude (db) L =5Ohm L =1mF L =1uF L =1nF L =1nF L =1nF L =1pF Phase (d) Fig Insertion gain of high-frequency attenuator vs. capacitive load impedance. Fig. 3-32, Fig and Fig show the dependence of insertion gain of highfrequency attenuator on resistive, inductive and capacitive source impedances. The

96 insertion gain curve keeps the same when the source and load impedances switch. Fig. 3-32, Fig and Fig are the same as Fig. 3-29, Fig. 3-3, and Fig respectively. The conclusions for the source impedances are all the same as the load impedances. 2 Insertion Gain: Gain 2 Insertion Gain: Phase Magnitude (db) S =1 6 Ohm S =5 Ohm S =1 Ohm Phase (d) S =1 Ohm S =.1 Ohm S =1-6 Ohm Fig Insertion gain of high-frequency attenuator vs. resistive source impedance. 2 Insertion Gain: Gain 2 Insertion Gain: Phase Magnitude (db) S =5 Ohm S =1mH S =1uH S =1uH S =1nH S =1nH Phase (d) Fig Insertion gain of high-frequency attenuator vs. inductive source impedance. 69

97 2 Insertion Gain: Gain 2 Insertion Gain: Phase 15 Magnitude (db) S =5Ohm S =1mF S =1uF S =1nF S =1nF S =1nF S =1pF Phase (d) Fig Insertion gain of high-frequency attenuator vs. capacitive source impedance Summary In this chapter, the multi-conductor lossy transmission-line model was used to study the high-frequency attenuator (nickel-dielectric-nickel) structure of the DM transmissionline EMI filter. The insertion gain, transfer gain, input and output currents and input impedance of the high-frequency attenuator under nominal design were first shown. The high-frequency attenuator could be modeled as a simple capacitor in the low frequency range. In higher frequency range, the multi-conductor transmission-line model had to be used to analyzer the performance of the high-frequency attenuator. Parametric study of the high-frequency attenuator was also carried out based on the multi-conductor lossy transmission-line model. The influences of the parameters of the high-frequency attenuator on f c1, f c2 and the roll-off slope above f c2 were summarized in Table 3-3: 7

98 Table 3-3 SUMMARY OF PARAMETRIC STUDY RESULTS ON HIGH-FREQUENCY ATTENUATOR Length Width t die µ Ni ε die tanδ die f c1 f c2 slope The following conclusions can be drawn from the parametric study of the highfrequency attenuator: A longer structure gives lower f c1, lower f c2, and more attenuation; A wider structure gives lower f c1 and more attenuation; A long and narrow structure gives the same f c1, but lower f c2 and more attenuation than a short and wide structure with the same conductor area; Under the nominal relative permeability and conductivity of nickel, nickel thickness larger than 5µm provides the same attenuation. For nickel thickness smaller than 5µm, smaller thickness gives more attenuation. For different nickel property values, the result will be different; Thinner dielectric material gives lower f c1, lower f c2, and more attenuation; The higher relative permeability of nickel gives lower f c2 and larger roll-off slope; The lower conductivity of nickel gives more attenuation; The higher relative permittivity of dielectric material gives lower f c1, lower f c2, and more attenuation; 71

99 The loss factor of the dielectric material has little influence on attenuation; Larger load (source) impedance gives more attenuation. For very small load (source) impedance, for example 1-6 Ω, there is still some attenuation coming from the loss in the nickel conductors. 72

100 Chapter 4: Parametric characterization of DM Transmission-Line EMI filter 4.1. Introduction As shown in section 2.3, the simulated insertion gain and transfer gain curves from the multi-conductor lossy transmission-line model agree well with the measured results of the DM transmission-line EMI filter prototype. In this chapter, the multi-conductor lossy transmission-line model is used to model the DM transmission-line EMI filter. The insertion gain, transfer gain, input and output currents and input impedance of the DM transmission-line EMI filter under nominal design are shown. The parametric study of the DM transmission-line EMI filter is also carried out based on multi-conductor lossy transmission-line model. The insertion gain and transfer gain of the DM transmission-line EMI filter under different dimensions, material properties, and source and load impedances are shown in this chapter. For the DM transmission-line EMI filter, copper conductor and nickel conductor are in parallel. The input current sharing between the copper and nickel conductors is essential to understand the working mechanisms of the DM transmission-line EMI filter. Therefore, the input currents of copper and nickel conductors under different dimensions and material properties are shown for the DM transmission-line EMI filter. filter: There are totally twelve parameters that are studied for the DM transmission-line EMI The length and width; 73

101 The thicknesses of the nickel, copper, dielectric, and insulation layers; The conductivity and relative permeability of nickel material; The relative permittivity and loss factor of the dielectric material; The relative permittivity and loss factor of the insulation material Insertion gain, transfer gain, current and input impedance Table 4-1 lists the dimensions and material properties of the DM transmission-line EMI filter under nominal design. The thickness of the copper layer is 25µm in the first prototype of the DM transmission line EMI filter. This thickness is chosen as the nominal value for copper layer thickness. Al 2 O 3 sheet with a thickness of 65µm is used as the insulation material in the design of the first prototype. 65µm is chosen as the nominal value of the insulation layer thickness. The relative permittivity 9.8 and loss factor.1% of Al 2 O 3 are chosen as the nominal values of the relative permittivity and loss factor of the insulation material. Fig. 4-1 shows the DM transmission-line EMI filter in a circuit. S and L are the source and load impedances, which are assumed to be 5Ω unless otherwise specified. The insertion gain, transfer gain, currents of the inner and outer conductors and input impedance of the DM transmission-line EMI filter are shown below from Fig. 4-2 to Fig

102 TABLE 4-1 NOMINAL DESIGN OF DM TRANSMISSION-LINE EMI FILTER PARAMETERS VALUE Length (cm) 1 Width (mm) 1 Overall dimension Outer conductor thickness (µm) 25 Inner conductor thickness (µm) 17 Dielectric thickness (µm) 15 Insulation thickness (µm) 65 Outer conductor Conductivity (S/m) 5.8e7 Inner conductor Conductivity (S/m) 1.45e7 Relative permeability 6 Dielectric Relative permittivity 14 Loss factor 4% Insulation Relative permittivity 9.8 Loss factor.1% Fig Schematic of measurement setup of DM transmission-line EMI filter. Fig. 4-2 shows the insertion gain of the DM transmission-line EMI filter under nominal design. The insertion gain of the high-frequency attenuator is put in the same graph for comparison. Same as the insertion gain curve of the high-frequency attenuator, the insertion gain curve of the DM transmission-line EMI filter shows two major turning frequency points: f c1 and f c2. F c1 is the cut-off frequency of the DM transmission-line EMI filter, at which the insertion gain of the filter is -3dB. The roll-off slope of the curve between f c1 and f c2 is -2dB/dec, which is the roll-off slope of a first-order capacitive 75

103 filter. Same as the high-frequency attenuator, the DM transmission-line EMI filter can be modeled as a simple capacitor in the frequency range below f c2. The capacitance is the same as the high-frequency attenuator. Above f c2, the effect of the losses in nickel conductors starts to show, this increases the attenuation of the DM transmission-line EMI filter. The roll-off slope is greatly increased above f c2. The insertion gain curve of the DM transmission-line filter is almost the same as the insertion gain curve of the high-frequency attenuator below 5MHz. At frequencies above 5MHz, DM filter has less attenuation than the high-frequency attenuator. Same conclusion can be drawn from the transfer gain curves shown in Fig This is due to the fact that the copper conductors of the DM transmission-line EMI filter provide another current path. Although above 5MHz, most of the current is now in the nickel conductor, the magnitude of the current in the copper conductor is very small, as shown in Fig. 4-5, this still results in an appreciable difference in the attenuation. Insertion Gain: Gain 2 Insertion Gain: Phase Magnitude (db) Phase (d) Attenuator DM filter f c1 f c Fig Insertion gains of DM transmission-line EMI filter and high-frequency attenuator. 76

104 2 Transfer Gain: Gain 2 Transfer Gain: Phase Magnitude (db) -4-6 Phase (d) Attenuator DM filter Fig Transfer gains of DM transmission-line EMI filter and high-frequency attenuator. Fig. 4-4 shows the input impedance of the DM transmission-line filter together with the input impedance of the high-frequency attenuator. Fig. 4-5 shows the input and output currents on the copper and nickel conductors. At lower frequency range, input current in the copper conductor is much larger than the input current in the nickel conductor. The current sharing between these two conductors is determined by the conductivities and dimensions of the copper and nickel conductors. As frequency goes up, more and more current goes to the nickel layer. At around 1kHz, the magnitudes of the input currents in the copper and nickel conductors are about the same. The input current of the nickel conductor continues increasing until at around 1MHz. Most of the currents are now in nickel layer; copper carries very little current. 77

105 1 2 Input impedance: Magnitude 2 Input impedance: Phase 1 1 in (Attenuator) in (DM filter) -2 Magnitude (A) 1 Phase (d) Fig Input impedances of DM transmission-line EMI filter and high-frequency attenuator..25 DM Current Distribution: Magnitude I in (Ni) 2 DM Current Distribution: Phase.2 I in (Cu) I out (Ni) I out (Cu) 15 1 Magnitude (A).15.1 Phase (d) Fig Input and output currents of DM transmission-line EMI filter Parametric study of DM transmission-line EMI filter The parametric study of the DM transmission-line EMI filter is carried out based on the multi-conductor lossy transmission-line model. There are totally twelve parameters. The nominal values and sweep range of these parameters are summarized in Table 4-2. For the parametric study, only one parameter changes every time, the other parameters keep the nominal values. 78

106 TABLE 4-2 NOMINAL DESIGN OF DM TRANSMISSION-LINE EMI FILTER Overall dimension Inner conductor Dielectric Insulation PARAMETERS NOMINAL VALUE SWEEP RANGE Length (cm) 1 5~2 Width (mm) 1 2~1 Outer conductor thickness (µm) 25 1~75 Inner conductor thickness (µm) 17 1~34 Dielectric thickness (µm) 15 15~6 Insulation thickness (µm) 65 1~65 Conductivity (Ω -1 /m) 1.45e7 1.e5~5.8e7 Relative permeability 6 1~6 Relative permittivity 14 1~14 Loss factor 4%.1%~2% Relative permittivity ~1 Loss factor.1%.1%~1% Influence of dimensions The dimensions of the DM transmission-line EMI filter are illustrated in Fig The influences of the length, width, and thicknesses of different layers on the filter performance are investigated. Fig. 4-7 shows the insertion gains of the DM transmissionline for different lengths. The length influences both f c1 and f c2. For a longer structure, f c1 and f c2 move lower. A longer structure gives more attenuation. Fig. 4-8 shows the transfer gain curves for different lengths. Fig. 4-9 shows the input currents of copper and nickel conductors for different lengths. The solid line is the input current in nickel conductor; the dotted line is the input current in copper conductor. For longer filters, the migration of the input current from copper to nickel starts at a lower frequency. 79

107 Fig Dimensions of DM transmission-line EMI filter. Insertion Gain: Gain 2 Insertion Gain: Phase Magnitude (db) Phase (d) length=.5m length=.1m length=.2m Fig Insertion gain of DM transmission-line EMI filter vs. length. 1 Transfer Gain: Gain 2 Transfer Gain: Phase Magnitude (db) Length=.5m Length=.1m Length=.2m Phase (d) Fig Transfer gain of DM transmission-line EMI filter vs. length. 8

108 .25 DM Current Distribution: Magnitude 14 DM Current Distribution: Phase i Ni 8 Magnitude (A).15.1 Phase (d) i cu (a) Magnitude Fig Input currents of DM transmission-line EMI filter vs. length. Fig. 4-1 shows the insertion gain curves for DM transmission-line EMI filter of different widths. A wider structure gives a lower f c1. However, f c2 hardly changes for different widths. A wider structure has more attenuation if the other parameters are kept the same. Fig shows the transfer gain curves of different widths. The transfer gain curves are similar for these three different widths, except some difference in attenuation at frequencies above 4MHz. Fig shows the input currents of copper and nickel conductors for different filter widths. Insertion Gain: Gain 2 Insertion Gain: Phase Magnitude (db) Phase (d) width=1mm width=5mm width=2mm Fig Insertion gain of DM transmission-line EMI filter vs. width. 81

109 1 Transfer Gain: Gain 2 Transfer Gain: Phase Magnitude (db) Phase (d) width=1mm width=5mm width=2mm Fig Transfer gain of DM transmission-line EMI filter vs. width..25 DM Current Distribution: Magnitude 6 DM Current Distribution: Phase i Ni.15-2 (A).1 (Degree) i cu (a) Magnitude Fig Input currents of DM transmission-line EMI filter vs. width. If the area (length width) of the conductors stays the same, as shown in Fig. 4-13, f c1 doesn t move because the capacitance value of the dielectric layer doesn t change. However, f c2 is lower for a long and narrow structure. If we want a smaller first-order (- 2dB/dec) bandwidth (f c2 -f c1 ), a long and narrow structure is better than a short and wide structure with the same area. Same conclusion can be drawn from the transfer gain curves in Fig

110 Insertion Gain: Gain 2 Insertion Gain: Phase Magnitude (db) Phase (d) W=2mm, L=.25m W=5mm, L=.1m W=1mm, L=.5m Fig Insertion gain of DM transmission-line EMI filter vs. length & width. 1 Transfer Gain: Gain 2 Transfer Gain: Phase Magnitude (db) -2-3 Phase (d) W=2mm, L=.25m W=5mm, L=.1m W=1mm, L=.5m Fig Transfer gain of DM transmission-line EMI filter vs. length & width. Nickel is the inner conductor of the DM transmission-line EMI filter. Fig shows the influence of nickel layer s thickness on the filter s performance. The insertion gain curves of the filters with nickel thicknesses larger than 5µm are almost the same. From Chapter 3, we know that for high-frequency attenuators with nickel s thicknesses smaller than 5µm, the attenuation is larger for a thinner nickel layer. The result is different for the DM transmission-line filter. For a very thin nickel layer, such as 1µm as shown in Fig. 4-15, the high impedance of the nickel layer will hinder the current from migrating from 83

111 copper to nickel, as illustrated in Fig Thus the attenuation of the filter with 1µm nickel layer is even smaller than the attenuation of filter with thicker nickel layer. However, the difference is small. We can draw the conclusion that the influence of nickel thickness on the filter s attenuation is negligible. Same conclusion can be drawn from the transfer gain curves shown in Fig Transfer Gain: Gain 2 Transfer Gain: Phase Magnitude (db) t Ni =1um t Ni =2um t Ni =5um t Ni =1um t Ni =17um t Ni =25um t Ni =34um Phase (d) Fig Insertion gain of DM transmission-line EMI filter vs. thickness of nickel layer..25 DM Current Distribution: Magnitude 2 DM Current Distribution: Phase.2 i Ni 15 Magnitude (A).15.1 Phase (d) i cu (a) Magnitude Fig Input currents of DM transmission-line EMI filter vs. thickness of nickel layer. 84

112 1 Transfer Gain: Gain 2 Transfer Gain: Phase Magnitude (db) t Ni =1um t Ni =2um t Ni =5um t Ni =1um t Ni =17um t Ni =25um Phase (d) t Ni =34um Fig Transfer gain of DM transmission-line EMI filter vs. thickness of nickel layer. Copper is the outside conductor in the filter structure. It carries the low-frequency power current. Fig shows the insertion gain curves for DM transmission-line filters with different copper layer thicknesses. Copper layers with different thicknesses have different resistance. This will influence the current migration from copper to nickel in the low frequency range, as shown in Fig However, copper thickness has little influence on the attenuation of the filter, as shown in Fig and Fig This means that the DM transmission-line filter can be used in different power level applications without changing its performance. Magnitude (db) Insertion Gain: Gain tcu=1um tcu=12um tcu=25um tcu=5um tcu=75um Phase (d) Insertion Gain: Phase Fig Insertion gain of DM transmission-line EMI filter vs. thickness of copper layer. 85

113 1 Transfer Gain: Gain 2 Transfer Gain: Phase Magnitude (db) tcu=1um tcu=12um tcu=25um tcu=5um tcu=75um Phase (d) Fig Transfer gain of DM transmission-line EMI filter vs. thickness of copper layer..25 DM Current Distribution: Magnitude 14 DM Current Distribution: Phase 12 Magnitude (A) i Ni Phase (d) i cu (a) Magnitude Fig Input currents of DM transmission-line EMI filter vs. thickness of copper layer. Fig and Fig show the insertion gain and transfer gain curves for DM transmission-line EMI filter with different dielectric layer thicknesses. A thinner dielectric layer gives more attenuation. Fig shows the input currents of copper and nickel conductors for different dielectric layer thicknesses. With a thinner dielectric layer, the length of the filter can be reduced to get the same attenuation. In Fig. 4-24, the red insertion gain curve is for a filter with dielectric thickness of 15µm and length of 1cm. 86

114 If a thinner dielectric layer is used, for example 15µm, the length can be reduced to around 1.8cm to get similar attenuation, as shown by the green curve in the same graph. Insertion Gain: Gain 2 Insertion Gain: Phase Magnitude (db) Phase (d) 5-5 tdielectric=5um -1 tdielectric=15um tdielectric=3um tdielectric=6um Fig Insertion gain of DM transmission-line EMI filter vs. thickness of dielectric layer. 1 Insertion Gain: Gain 2 Insertion Gain: Phase Magnitude (db) Phase (d) tdielectric=5um tdielectric=15um tdielectric=3um tdielectric=6um Fig Transfer gain of DM transmission-line EMI filter vs. thickness of dielectric layer. 87

115 .25 DM Current Distribution: Magnitude 14 DM Current Distribution: Phase i Ni 8 Magnitude (A).15.1 Phase (d) i cu (a) Magnitude Fig Input currents of DM transmission-line EMI filter vs. thickness of dielectric layer. Insertion Gain: Gain 2 Insertion Gain: Phase Magnitude (db) -6-8 Phase (d) t die =15um, length=.1m t die =15um, length=.1m t die =15um, length=.18m (a) Magnitude Fig Insertion gains of DM transmission-line EMI filter vs. thickness of dielectric layer & filter length. In the original DM transmission-line EMI filter structure, two Al 2 O 3 layers are used as the insulation between the nickel and copper conductors. Fig and Fig shows that the thickness of the insulation layer has no influence on the filter s performance. Fig shows that for different insulation layer thicknesses, the input currents to copper and nickel conductors keep the same. Actually, the insulation layer can be removed from the DM transmission-line EMI filter structure. The copper can be directly deposited onto the nickel layer. Eliminating the insulation layer can greatly 88

116 reduce the filter s thickness, while keeping the attenuation of the filter the same. The experimental demonstration of this design will be shown later in Chapter 7. Insertion Gain: Gain 2 Insertion Gain: Phase Magnitude (db) tinsulation=1um tinsulation=2um tinsulation=3um -1 tinsulation=4um tinsulation=5um tinsulation=65um Phase (d) Fig Insertion gain of DM transmission-line EMI filter vs. thickness of insulation layer. 1 Transfer Gain: Gain 2 Transfer Gain: Phase Magnitude (db) tinsulation=1um tinsulation=2um tinsulation=3um tinsulation=4um tinsulation=5um tinsulation=65um Phase (d) Fig Transfer gain of DM transmission-line EMI filter vs. thickness of insulation layer. 89

117 .25 DM Current Distribution: Magnitude 14 DM Current Distribution: Phase Magnitude (A).15.1 i Ni Phase (d) i cu (a) Magnitude Fig Input currents of DM transmission-line EMI filter vs. thickness of insulation layer Influence of inner conductor s properties Nickel is used as the inner (low-conductivity) conductor in the DM transmission-line EMI filter. Its conductivity, relative permeability may influence the attenuation of the filter. Fig shows the influence of nickel s relative permeability on the filter s attenuation. It is clearly shown that nickel s relative permeability doesn t influence f c1. However, it has influence on f c2 and the roll-off slope of the curve in the frequency range above f c2. A higher relative permeability gives lower f c2 and larger roll-off slope. For a non-magnetic conductor (µ r =1), the attenuation is very limited in the high frequency range, as shown by the red insertion gain curve in Fig Thus the magnetic property of nickel is essential for high attenuations of the filter. Fig shows the transfer gain curves for different relative permeability of nickel. Fig. 4-3 shows the input currents in copper and nickel for different nickel s relative permeability. 9

118 Insertion Gain: Gain 2 Insertion Gain: Phase Magnitude (db) u r =1 u r =2 u r =5 u r =1 u r =3 u r =6 Phase (d) Fig Insertion gain of DM transmission-line EMI filter vs. relative permeability of nickel. 1 Transfer Gain: Gain 2 Transfer Gain: Phase 15 Magnitude (db) µ r =1 µ r =2 µ r =5 µ r =1 µ r =3 µ r =6 Phase (d) Fig Transfer gain of DM transmission-line EMI filter vs. relative permeability of nickel..25 DM Current Distribution: Magnitude 2 DM Current Distribution: Phase Magnitude (A).15.1 i Ni Phase (d) i cu (a) Magnitude Fig Input currents of DM transmission-line EMI filter vs. relative permeability of nickel. 91

119 Fig shows the influence of nickel s conductivity on the DM transmission-line filter s attenuation. The previous result on the high-frequency attenuator shows that lower nickel s conductivity gives more attenuation. However, for DM transmission-line filter, the result is different. The highest conductivity in Fig is 5.8e7S/m, which is the conductivity of copper. As the conductivity gets smaller, more losses are generated in nickel layer. Thus the filter has more attenuation. However, if nickel s conductivity is too low, such as 1.e5S/m, as the brown curve shown in Fig. 4-31, the high impedance of nickel conductor will hinder the current from migrating from copper to nickel. Copper conductor still carries a lot of current until at very high frequency. Fig shows that for nickel s conductivity 1.e5S/m, the current in copper keeps an appreciable magnitude until several MHz. Thus there is an optimal range for nickel s conductivity to get the largest attenuation. If all other parameters are kept at the nominal values, the optimal range for nickel s conductivity is between 5.e6S/m and 5.e5S/m. Same conclusion can be drawn from the transfer gain curves in Fig Insertion Gain: Gain 2 Insertion Gain: Phase Magnitude (db) σ Ni =5.8e7 σ Ni =1.45e7 σ Ni =5.e6 σ Ni =1.e6 Phase (d) 5-5 σ Ni =5.e σ Ni =1.e Fig Insertion gain of DM transmission-line EMI filter vs. conductivity of nickel. 92

120 2 Transfer Gain: Gain 2 Transfer Gain: Phase 15 1 Magnitude (db) σ Ni =5.8e7 σ Ni =1.45e7 σ Ni =5.e6 σ Ni =1.e6 σ Ni =1.e5 Phase (d) Fig Transfer gain of DM transmission-line EMI filter vs. conductivity of nickel..25 DM Current Distribution: Magnitude 15 DM Current Distribution: Phase.2 i Ni 1 Magnitude (A).15.1 Phase (d) i cu (a) Magnitude Fig Input currents of DM transmission-line EMI filter vs. conductivity of nickel Influence of dielectric material s properties Fig and Fig shows the influence of dielectric material s relative permittivity on the DM transmission-line filter s attenuation. Fig shows the input currents under different relative permittivity values. Higher relative permittivity is highly desired to get high attenuation. Fig and Fig show the insertion gain and transfer gain curves of the DM transmission-line filter for different loss factors of the dielectric material. Fig shows the input currents for different loss factors of the 93

121 dielectric material. The loss factor of the dielectric material has almost no influence on the filter attenuation, because compared to the loss generated in the nickel conductors; the loss generated in the dielectric layer is negligible. Insertion Gain: Gain 2 Insertion Gain: Phase Magnitude (db) ε =1 r ε =1 r ε =1 r ε =4 r ε =14 r Phase (d) Fig Insertion gain of DM transmission-line EMI filter vs. relative permittivity of dielectric. 1 Transfer Gain: Gain 2 Transfer Gain: Phase Magnitude (db) ε r =1 ε r =1 Phase (degree) ε r =1 ε r = ε r = Fig Transfer gain of DM transmission-line EMI filter vs. relative permittivity of dielectric. 94

122 .25 DM Current Distribution: Magnitude 14 DM Current Distribution: Phase i Ni 8 Magnitude (A).15.1 Phase (d) i cu (a) Magnitude Fig Input currents of DM transmission-line EMI filter vs. relative permittivity of dielectric. Insertion Gain: Gain 2 Insertion Gain: Phase Magnitude (db) loss factor=.1% loss factor=.1% -1 loss factor=1% loss factor=1% loss factor=2% Phase (d) Fig Insertion gain of DM transmission-line EMI filter vs. loss factor of dielectric. 1 Transfer Gain: Gain 2 Transfer Gain: Phase Magnitude (db) Phase (d) loss factor=.1% loss factor=.1% -6 loss factor=1% loss factor=1% loss factor=2% Fig Transfer gain of DM transmission-line EMI filter vs. loss factor of dielectric. 95

123 .25 DM Current Distribution: Magnitude DM Current Distribution: Phase.2 1 Magnitude (A).15.1 i Ni Phase (d) i cu (a) Magnitude Fig Input currents of DM transmission-line EMI filter vs. loss factor of dielectric Influence of insulation material s properties The insulation layer is between nickel and copper conductors in the DM transmissionline EMI filter. Al 2 O 3 is used as the insulation material in the original design. The relative permittivity of Al 2 O 3 is 9.8. The loss factor of Al 2 O 3 is.1%. Fig. 4-4 to Fig show that the relative permittivity and loss factor of the insulation material have no influence on the filter s attenuation. Insertion Gain: Gain 2 Insertion Gain: Phase Magnitude (db) Phase (d) ε ins =9.8 ε ins =1 ε ins = Fig Insertion gain of DM transmission-line EMI filter vs. relative permittivity of insulation. 96

124 1 Transfer Gain: Gain 2 Transfer Gain: Phase Magnitude (db) Phase (d) ε r =9.8 ε r =1 ε r = Fig Transfer gain of DM transmission-line EMI filter vs. relative permittivity of insulation..25 DM Current Distribution: Magnitude 14 DM Current Distribution: Phase i Ni 8 Magnitude (A).15.1 Phase (d) i cu (a) Magnitude Fig Input currents of DM transmission-line EMI filter vs. relative permittivity of insulation. Insertion Gain: Gain 2 Insertion Gain: Phase Magnitude (db) Phase (d) 5-5 loss factor=.1% -1 loss factor=.1% loss factor=1% loss factor=1% Fig Insertion gain of DM transmission-line EMI filter vs. loss factor of insulation. 97

125 1 Insertion Gain: Gain 2 Transfer Gain: Phase Magnitude (db) Phase (d) loss factor=.1% loss factor=.1% loss factor=1% loss factor=1% Fig Transfer gain of DM transmission-line EMI filter vs. loss factor of insulation..25 DM Current Distribution: Magnitude 14 DM Current Distribution: Phase i Ni 8 Magnitude (A).15.1 Phase (d) i cu (a) Magnitude Fig Input currents of DM transmission-line EMI filter vs. loss factor of insulation Dependence on source and load impedances The above insertion gain curves are all under the condition that the source and load impedances are both 5Ω. When the source and load impedances change, the insertion gain curve will be different. Fig shows the insertion gain curves at different resistive loads. A larger load resistance gives more attenuation. For load resistances larger than 5Ω, the attenuations are almost the same. As the load resistance reduces, the attenuation reduces. The worst case is for a load that is close to a short circuit (for 98

126 example 1-6 Ω). The filter still has some attenuation under the worst condition. The attenuation comes from the loss generated in the nickel conductors. 2 Insertion Gain: Gain 2 Insertion Gain: Phase Magnitude (db) L =1 6 Ohm L =1 3 Ohm L =5 Ohm L =1 Ohm L =1 Ohm L =.1Ohm L =.1 Ohm L =1-6 Ohm Phase (d) Fig Insertion gain of DM transmission-line EMI filter vs. resistive load impedance. Fig shows the insertion gain curves under different inductive loads. The 5Ω case is drawn in the same graph for comparison. A larger inductance gives more attenuation. For an inductive load as small as 1nH, there is still some attenuation at high frequency that comes from the losses in nickel layer. 4 Insertion Gain: Gain 2 Insertion Gain: Phase Magnitude (db) L =5 Ohm L =1mH L =1uH Phase (d) L =1uH L =1nH L =1nH Fig Insertion gain of DM transmission-line EMI filter vs. inductiveload impedance. 99

127 Fig shows the insertion gains at different capacitive loads. A smaller capacitance gives more attenuation. The worst condition is for a large capacitance, such as 1mF, as shown by the blue curve in Fig Insertion Gain: Gain 2 Insertion Gain: Phase Magnitude (db) L =5 Ohm L =1mF L =1uF L =1nF L =1nF L =1nF L =1pF Phase (d) Fig Insertion gain of DM transmission-line EMI filter vs. capacitive load impedance. Fig. 4-49, Fig. 4-5 and Fig show the insertion gain curves of DM transmissionline EMI filter under different resistive, inductive and capacitive source impedances. The results are all the same as the load impedances. 2 Insertion Gain: Gain 2 Insertion Gain: Phase Magnitude (db) -4-6 S =1 6 Ohm S =1 3 Ohm S =5 Ohm S =1 Ohm S =1 Ohm Phase (d) S =.1Ohm S =.1 Ohm -1-1 S =1-6 Ohm Fig Insertion gain of DM transmission-line EMI filter vs. resistive source impedance. 1

128 4 Insertion Gain: Gain 2 Insertion Gain: Phase Magnitude (db) S =5 Ohm S =1mH S =1uH Phase (d) S =1uH S =1nH -1-1 S =1nH Fig Insertion gain of DM transmission-line EMI filter vs. inductive source impedance. 2 Insertion Gain: Gain 2 Insertion Gain: Phase Magnitude (db) S =5 Ohm S =1mF S =1uF S =1nF S =1nF S =1nF S =1pF Phase (d) Fig Insertion gain of DM transmission-line EMI filter vs. capacitive source impedance Summary In this chapter, the multi-conductor lossy transmission-line model was used to study the DM transmission-line EMI filter. The insertion gain, transfer gain, input and output currents and input impedance of the DM transmission-line EMI filter under nominal design were first shown. The DM transmission-line EMI filter can be modeled as a simple capacitor in the low frequency range. In higher frequency range, the multi- 11

129 conductor transmission line model has to be used to analyzer the performance of the DM transmission-line EMI filter. Parametric study of the DM transmission-line EMI filter is also carried out based on the multi-conductor lossy transmission-line model. The influences of the parameters of the DM transmission-line EMI filter on f c1, f c2 and the roll-off slope are summarized in Table 4-3: TABLE 4-3 SUMMARY OF PARAMETRIC STUDY RESULTS Length Width t Ni t Cu t die t ins µ Ni ε die tanδ die ε ins Tanδ ins f c1 f c2 slope The following conclusions can be drawn from the parametric study of the DM transmission-line EMI filter: A longer structure gives lower f c1, lower f c2 and more attenuation; A wider structure has lower f c1 and more attenuation; A long and narrow structure gives the same f c1, but lower f c2 and more attenuation than a short and wide structure with the same conductor area; The nickel thickness has little influence on the filter attenuation; The copper thickness has little influence on the filter attenuation; Thinner dielectric material gives lower f c1, lower f c2 and more attenuation; The Insulation thickness has little influence on filter attenuation; 12

130 The higher relative permeability of nickel gives lower f c2, larger roll-off slope and more attenuation; The conductivity of nickel has an optimum value range: 5.e6S/m ~ 5.e5S/m. This is different from the high-frequency attenuator. Lower conductivity of nickel gives more attenuation for the high-frequency attenuator. The higher relative permittivity of dielectric material gives lower f c1, lower f c2 and more attenuation; The loss factor of dielectric material has little influence on filter attenuation; The relative permittivity of insulation material has little influence on filter attenuation; The loss factor of insulation material has little influence on filter attenuation; The larger load (source) impedance gives more attenuation. For very small load (source) impedance, for example 1-6 Ω, there are still some attenuation coming from the loss in the nickel conductors. 13

131 Chapter 5: Measurement setup of integrated DM transmission-line EMI filter 5.1. Introduction The previous chapters have analyzed the performance of the high-frequency attenuator and the DM transmission-line EMI filter under different dimensions and material properties. Measurement results are needed to verify the results from the multiconductor lossy transmission-line model. The insertion voltage gain and transfer voltage gain of the transmission-line EMI filter can be measured using a network analyzer. However, it is not a trivial issue to get accurate measurement results for the filter, because of the parasitics introduced by the connector and other parts of the measurement setup, especially at frequencies above several MHz. Great care must be taken when design the measurement setup and choose connectors for the transmission-line EMI Filter. The measurement setups for both insertion gain and transfer gain have a commonground problem, which is inherent with the network analyzer. This common-ground problem will influence the measurement results of the high-frequency attenuator in the high frequency range. However, the measurement results of the DM transmission-line EMI filter are almost not influenced, as shown in section The parametric study of the high-frequency attenuator is carried out again taking into account the commonground problem of the network analyzer. 14

132 5.2. Measurement setup using network analyzer The definition of the transfer voltage gain and insertion voltage gain are repeated here. As shown in Fig. 5-1, the transfer voltage gain is defined as the ratio of the output voltage to the input voltage of the EMI filter. Insertion gain is defined as the ratio of the load voltage with the filter to the load voltage without the filter: V V out TG = (5-1) in VL _ w IG = (5-2) V L _ w/ o Fig Schematic of transfer gain and insertion gain measurement circuit for transmission-line EMI filter. The Agilent 4395A network/spectrum/impedance analyzer, shown in Fig. 5-2, is used to measure the insertion gain and transfer gain of the transmission-line EMI filter [49][5]. The frequency range of the Agilent 4395A is 1Hz - 51MHz. There are four input/output terminals: RF out, R, A, and B, as shown in Fig The input impedances of these terminals are all 5Ω. Thus the measurement results from network analyzer are all under the condition that the source and load impedances are both 5Ω. 15

133 Fig Agilent 4395A network/spectrum/impedance analyzer. Fig Input/output terminals of Agilent 4395A Measurement setup for insertion gain There are two measurement methods to measure the insertion gain using network analyzer: two-measurement method and one-measurement method. The twomeasurement method measures the load voltage with and without the filter separately and then gets the ratio of these two load voltages. The one-measurement method uses the transmission/reflection test kit and can get the insertion gain from one-single measurement. For the two-measurement method, the first measurement is shown in Fig The R terminal connects to the RF out terminal by a direct through connector. The voltage level of the R terminal is measured. This is the case without any filter. The first equivalent circuit is shown in the left part of Fig The second measurement is shown in Fig The signal coming from the RF out terminal is connected to the DM transmission-line 16

134 EMI filter. The output of the filter is connected to the B terminal, which is connected to a 5Ω load. The equivalent circuit of the second measurement is shown in the right part of Fig Thus the insertion gain of the transmission-line EMI filter is the ratio of the two measured voltages, as described in Equation (5-3). L _ w IG = = V V L _ w / o V V B R (5-3) Fig Voltage measurement without DM transmission-line EMI filter. Fig Voltage measurement with the DM transmission-line EMI filter. Insertion gain can also be measured from a single measurement. Fig. 5-6 shows the measurement setup for measuring insertion gain using the transmission/reflection test kit 17

135 Agilent 87512A. The test kit is a resistor network. Its internal circuit is shown in Fig The test kit provides a convenient means of measuring the reflection and transmission characteristics of a device in one direction. The frequency range of the 87512A is DC up to 2GHz. This test kit routes a portion of the signal from the RF out terminal to the R terminal of the network analyzer. This signal is used as the reference signal in this ratio measurement. The reflected signal from the filter is returned to input terminal A of the network analyzer. The following matrix shows the theoretical value of A/R when Open, Short or 5Ω is connected at the test port. Open Short 5Ω A/R =.954 (-.41 db) A/R = (- Infinity db) A/R =.477 ( db) When Open, Short or 5Ω is connected at the test port, the values of A/R become approximately 1,,.5 respectively. The transmitted signal from the filter is measured from the input terminal B of the network analyzer. The insertion gain is measured as the voltage ratio of the signals from terminal B and terminal R. The test kit can be simulated together with the transmission-line EMI filter using the multi-conductor lossy transmission-line model. Fig. 5-9 shows the schematic of the measurement setup with the test kit. Fig. 5-1 shows the simulated V B /V R curve. It is identical with the insertion gain curve that is defined in Fig Thus the simulated results from the model confirm that the one-measurement setup with the test kit gives the same insertion gain curve as the two-measurement setup. Fig compares the measured insertion gains of the transmission-line EMI filter from the two-measurement and one-measurement methods. The results from these two measurement setups are identical. The one-measurement setup with the transmission/reflection test kit is chosen due to its simplicity. 18

136 Fig Insertion gain measurement setup using Agilent 4395A. RF-out R A B RF-in R Test A Fig Agilent 87512A transmission / reflection test kit. R2(1) R3(23.7) R R4(4.2) RF IN R1(82.5) Test R5(9.9) R6(14.7) A R7(82.5) Fig Internal circuit of Agilent 87512A transmission / reflection test kit. 19

137 Fig Schematic of measurement setup with Agilent 87512A transmission / reflection test kit. 1 Insertion Gain: Gain 2 Insertion Gain: Phase Magnitude (db) Phase (d) V L (with filter)/v L (without filter) V B /V R with test kit Fig Simulated insertion gains of DM transmission-line EMI filter with/without test kit. Gain (db) Insertion Gain 1.E+3 1.E+4 1.E+5 1.E+6 1.E+7 1.E+8 Two measurements without test kit One measurement with test kit Phase (Degree) Insertion Gain E+3 1.E+4 1.E+5 1.E+6 1.E+7 1.E+8 two measurement without test kit one measurement with test kit Fig Measured insertion gains of DM transmission-line EMI filter with/without test kit. 11

138 Measurement setup for transfer gain The transfer gain of the DM transmission-line EMI filter can also be measured using a network analyzer. Fig shows one setup to measure the transfer gain of the DM transmission-line EMI filter. The voltage source with source impedance 5Ω is connected to the filter through the RF-output terminal of the network analyzer. One external 5Ω resistor is connected to the filter as the load. The voltages of the filter at the input and output terminals are obtained by measuring the voltages at terminal R and terminal B. Two input adapters (Agilent 4182A) are used to convert the input impedances of the input terminals R and B from 5Ω to 1MΩ. This will reduce the input currents of terminal R and terminal B to near zero. The ratio of the voltages at terminal B to the voltages at terminal R gives the transfer gain of the EMI filter. The reason to use the four-terminal measurement setup for the transfer gain is to reduce the parasitic inductance, introduced by the connectors and copper strips between the filter and the connectors. The transfer gain of the EMI filter is very sensitive to the parasitic inductances, which will be discussed in detail later in this chapter. As mentioned above, the input impedances of all the terminals of the network analyzer are 5Ω. If an external 5Ω resistor is not available, the internal 5Ω of the network analyzer can also be used as the load for the transmission-line filter. As shown in Fig. 5-13, the output of the filter is connected to the A terminal of the network analyzer. The measured transfer gain curve is the same as that from the measurement setup with external 5Ω load, as shown in Fig

139 Fig Transfer gain measurement setup with external load. Fig Transfer gain measurement setup with internal load. 1 Transfer Gain db E+3 1.E+4 1.E+5 1.E+6 1.E+7 1.E+8 Outside 5Ohm load Inside 5Ohm load Fig Measured transfer gains of transmission-line EMI filter with inside/outside load. 112

140 5.3. Common-ground problem of network analyzer For the measurement setups of insertion gain and transfer gain, there is one common problem that comes from the network analyzer. For the network analyzer, the ground of the input and output terminals RF-output, R, A, and B are all connected together. This provides another current path connecting the input and output of the filter. As shown in Fig. 5-15, SG and LG are the to-ground impedances. In real DM transmission-line EMI filter applications, SG or LG should be very high impedance, SG = LG = 1MΩ is assumed. However, for the measurement setups using network analyzer, the ground of the input and output are connected. The exact values of the to-ground impedance are not available. SG = LG = is assumed for the network analyzer measurement. The multiconductor lossy transmission-line model is again used to show the influence of SG and LG on the high-frequency attenuator and the DM EMI filter s performance Influence of common-ground problem to high-frequency attenuator Fig shows the insertion gains of the high-frequency attenuator under different values of SG and LG. As long as either SG or LG is high impedance, there is no current path, the insertion gains are the same as what is shown in Fig However, if SG = LG =, the attenuation of the high-frequency attenuator at high frequencies is reduced, as shown by the red curve of Fig The same conclusion can be drawn from the transfer gain curves shown in Fig The input and output currents of the high-frequency attenuator when SG = LG = are shown in Fig and Fig The to-ground current is drawn in the same graph. Although the magnitude of the to-ground current at high frequencies is small, it still reduces the high-frequency attenuation of the filter. 113

141 Fig Measurement setup of high-frequency attenuator with to-ground impedances. Insertion Gain: Gain 2 Insertion Gain: Phase Magnitude (db) Phase (d) sg=lg= sg=1 6, lg= sg=, lg=1 6 sg=lg= Fig Insertion gains of high-frequency attenuator under different SG and LG. 2 Transfer Gain: Gain 2 Transfer Gain: Phase Magnitude (db) -4-6 SG =, LG = Phase (d) SG =1 6, LG = SG =, LG = SG =1 6, LG = Fig Transfer gains of high-frequency attenuator under different SG and LG. 114

142 .25.2 Input Current Distribution: Magnitude Iin ni1 Iin ni2 Iground 5 Input Current Distribution: Phase Magnitude (A).15.1 Phase (d) (a) Magnitude Fig Input currents of high-frequency attenuator when SG = LG =..1.8 Output Current Distribution: Magnitude Iout ni1 Iout ni2 Iground Output Current Distribution: Phase Magnitude (A).6.4 Phase (d) (a) Magnitude Fig Output currents of high-frequency attenuator when SG = LG =. The parametric study of the high-frequency attenuator in Chapter 3 is performed under the condition SG = LG = 1MΩ. The insertion gain and transfer gain curves from the parametric study can not be compared with the measurement results from the network analyzer. The parametric study is performed again here under the condition SG = LG =. The resulted insertion gain and transfer gain curves are shown from Fig. 5-2 to Fig

143 Insertion Gain: Gain 2 Insertion Gain: Phase Magnitude (db) Phase (d) length=.5m length=.1m length=.2m Fig Insertion gain of high-frequency attenuator vs. length when SG = LG =. 1 Transfer Gain: Gain 2 Transfer Gain: Phase Magnitude (db) Phase (d) Length=.5m Length=.1m Length=.2m Fig Transfer gain of high-frequency attenuator vs. length when SG = LG =. Insertion Gain: Gain 2 Insertion Gain: Phase Magnitude (db) -6-8 Phase (d) width=1mm width=2mm width=5mm width=1mm Fig Insertion gain of high-frequency attenuator vs. width when SG = LG =. 116

144 1 Transfer Gain: Gain 2 Transfer Gain: Phase Magnitude (db) Phase (d) width=1mm width=2mm width=5mm width=1mm Fig Transfer gain of high-frequency attenuator vs. width when SG = LG =. Insertion Gain: Gain 2 Insertion Gain: Phase Magnitude (db) Phase (d) W=1mm, L=.5m W=5mm, L=.1m W=2mm, L=.25m Fig Insertion gain of high-frequency attenuator vs. length & width when SG = LG =. 1 Transfer Gain: Gain 2 Transfer Gain: Phase Magnitude (db) Phase (d) W=1mm, L=.5m W=5mm, L=.1m W=2mm, L=.25m Fig Transfer gain of high-frequency attenuator vs. length & width when SG = LG =. 117

145 Insertion Gain: Gain 2 Insertion Gain: Phase Magnitude (db) tni=1um -8 tni=1.5um tni=2um tni=3um -1 tni=5um tni=1um tni=17um tni=25um Phase (d) Fig Insertion gain of high-frequency attenuator vs. thickness of nickel when SG = LG =. 1 Transfer Gain: Gain 2 Transfer Gain: Phase Magnitude (db) tni=1um tni=1.5um tni=2um tni=3um tni=5um tni=1um tni=17um tni=25um Phase (d) Fig Transfer gain of high-frequency attenuator vs. thickness of nickel when SG = LG =. Insertion Gain: Gain 2 Insertion Gain: Phase Magnitude (db) -6-8 t die =5um Phase (d) t die =15um t die =3um t die =6um Fig Insertion gain of high-frequency attenuator vs. thickness of dielectric when SG = LG =. 118

146 1 Transfer Gain: Gain 2 Transfer Gain: Phase Magnitude (db) t die =5µm Phase (d) t die =15µm -1-6 t die =3µm t die =6µm Fig Transfer gain of high-frequency attenuator vs. thickness of dielectric when SG = LG =. Insertion Gain: Gain 2 Insertion Gain: Phase Magnitude (db) u r =1 u r =2 u r =5 u r =1 Phase (d) u r =3 u r = Fig Insertion gain of high-frequency attenuator vs. relative permeability of nickel when SG = LG =. 1 Transfer Gain: Gain 2 Transfer Gain: Phase Magnitude (db) µ r =1 µ r =2 µ r =5 µ r =1 Phase (d) µ r =3 µ r = Fig Transfer gain of high-frequency attenuator vs. relative permeability of nickel when SG = LG =. 119

147 Transfer Gain: Gain 2 Insertion Gain: Phase Magnitude (db) σ Ni =5.8e7 σ Ni =1.45e7 σ Ni =1.e6 Phase (d) Fig Insertion gain of high-frequency attenuator vs. conductivity of nickel when SG = LG =. 1 Transfer Gain: Gain 2 Transfer Gain: Phase Magnitude (db) Phase (d) σ Ni =5.8e7 σ Ni =1.45e7 σ Ni =1.e Fig Transfer gain of high-frequency attenuator vs. conductivity of nickel when SG = LG =. Insertion Gain: Gain 2 Insertion Gain: Phase Magnitude (db) Er=1 Er=3-1 Er=5 Er=1 Er= Phase (d) Fig Insertion gain of high-frequency attenuator vs. relative permittivity of dielectric when SG = LG =. 12

148 1 Transfer Gain: Gain 2 Transfer Gain: Phase Magnitude (db) ε r =1 ε r =3 Phase (d) ε r =5 ε r =1 ε r = Fig Transfer gain of high-frequency attenuator vs. relative permittivity of dielectric when SG = LG = Insertion Gain: Gain 2 Insertion Gain: Phase Magnitude (db) Loss factor=.1% Loss factor=.4% -1 Loss factor=1% Loss factor=4% Loss factor=1% Phase (d) Fig Insertion gain of high-frequency attenuator vs. loss factor of dielectric when SG = LG =. 1 Transfer Gain: Gain 2 Transfer Gain: Phase Magnitude (db) loss factor=.1% -6 loss factor=.4% loss factor=1% -7 loss factor=4% loss factor=1% Phase (d) Fig Transfer gain of high-frequency attenuator vs. loss factor of dielectric when SG = LG =.

149 Comparing these insertion gain and transfer gain curves with the previous curves in Chapter 3, we can see that the attenuations at higher frequencies are all reduced. The curves here show more pronounced transmission-line effect: some bumps (resonant points) along the line. The insertion gain and transfer gain curves in Chapter 3 are very smooth, because the high losses in the nickel layers damped the transmission-line effect. When SG = LG =, the common grounding provides another current return path that is lossless. Thus, the transmission line effect is more pronounced. Despite these differences, the influences of the parameters on the performance of the high-frequency attenuator show the same trend. The conclusions drawn from the previous parametric study still apply Influence of common-ground problem to DM transmission-line EMI filter The measurement setup of the DM transmission-line EMI filter is illustrated in Fig Insertion gains and transfer gains of the DM transmission-line EMI filter under different to-ground impedances SG and LG are shown in Fig and Fig As shown in Fig. 5-39, SG and LG have a very small influence on the insertion gain of the DM transmission-line EMI filter. The insertion gain curves are almost identical, except for a small difference above 5MHz. Such a small difference is hard to measure using a network analyzer at such a high frequency. Same phenomenon can be seen in the transfer gain curves in Fig The reason SG and LG have a larger influence on highfrequency attenuator performance than they have on DM transmission-line EMI filter performance is that the DM filter has two outside copper conductors, which are also lowimpedance paths. The input and output currents, as well as the to-ground current of the 122

150 DM transmission-line EMI filter when SG = LG =, are shown in Fig and Fig Fig Schematic of measurement setup of DM transmission-line EMI filter. Insertion Gain: Gain 2 Insertion Gain: Phase Magnitude (db) Phase (d) SG = LG =1MOhm SG = LG = Fig Insertion gains of DM transmission-line EMI filter under different SG and LG. 1 Transfer Gain: Gain 2 Transfer Gain: Phase Magnitude (db) Phase (d) SG = LG =1MOhm SG = LG = Fig Transfer gains of DM transmission-line EMI filter under different SG and LG. 123

151 Magnitude (A) Input Current Distribution: Magnitude Iin cu1 Iin ni2 Iin ni3 Iin cu4 Iin ground Phase (d) Input Current Distribution: Phase (a) Magnitude Fig Input and ground currents of DM transmission-line EMI filter when SG = LG =. Magnitude (A) Output Current Distribution: Magnitude Iout cu1 Iout ni2 Iout ni3 Iout cu4 Iout ground Phase (d) Output Current Distribution: Phase (a) Magnitude Fig Output and ground currents of DM transmission-line EMI filter when SG = LG = Measurement board design To measure the insertion gain and transfer gain of the transmission-line EMI filter, connectors and coaxial cables must be used to connect the filter with the inputs of the network analyzer. The connectors and copper strips between the connectors and the filter will introduce parasitic inductance to the measurement setup, as shown in Fig The influence of the parasitic inductance to the measured insertion gain and transfer gain is studied using the multi-conductor lossy transmission-line model. 124

152 Fig Measurement setup of DM transmission-line EMI filter with parasitic inductance. With the parasitic inductance, the source and load impedance matrices of the multiconductor transmission-line model needs to be modified. We assume the impedance of the parasitic inductance is ESL. The boundary conditions at z = for the four-conductor transmission-line can be expressed by the following equations: V () = V () = V 1 S [ I () + I ()]( V () = V () = [ I () + I () + I () + I ()] S + 4 ESL ) [ I () + I () + I () + I ()] SG 1 [ I () + I ()] ESL 4 SG (5-4) Rewriting (5-4) in matrix form yields V( ) = V 4I() (5-5) S S where the input source vector is: [ V ] t V = (5-6) S S V S The source impedance matrix is: 125

153 S + SG + ESL S + SG + ESL SG SG = S SG ESL S SG ESL SG SG (5-7) S4 SG SG SG + ESL SG + ESL SG SG SG + ESL SG + ESL The boundary conditions at the load side are: V ( l) = V ( l) = [ I ( l) + I ( l)]( 1 L + ESL V ( l) = V ( l) = [ I ( l) + I ( l) + I ( l) + I ( l)] ) + [ I ( l) + I ( l) + I ( l) + I ( l)] 4 1 LG 2 + [ I ( l) + I ( l)] ESL LG (5-8) Rewriting (5-8) in matrix form yields V( l) = L4 I( l) (5-9) where the load impedance matrix is: L + LG + ESL L + LG + ESL LG LG = L LG ESL L LG ESL LG LG (5-1) L4 LG LG LG + ESL LG + ESL LG LG LG + ESL LG + ESL The voltage on the load is: V = )] (5-11) load V1 ( l) V3( l) [ I1( l) + I 2( l) I 3( l) I 4( l ESL The transfer gain of the EMI filter is: TG V ( l) V ( l) [ I ( l) + I ( l) I ( l) I ( l)] V () V () + [ I () + I () I () I ()] ESL = (5-12) ESL The insertion gain of the EMI filter is: 126

154 V IG = l) V3( l) [ I1( l) + I 2( l) I L VS + S L ( l) I ( l)] 1 ( 3 4 ESL (5-13) Fig shows the insertion gains of the DM transmission-line EMI filter under different parasitic inductances. Insertion gain is not sensitive to parasitic inductance. With 1nH parasitic inductance, the insertion gain curve is almost identical to the one without any parasitic inductance. As long as the parasitic inductance is smaller than 1nH, the insertion gain curve is very close to the one without any parasitic inductance. However, transfer gain curve will change under different parasitic inductances. A larger parasitic inductance will shift the cut-off frequency of the transfer gain to a lower frequency. The roll-off slope is also changed by the parasitic inductance, as shown in Fig Thus, to measure an accurate transfer gain curve using a network analyzer, it is essential to eliminate the parasitic inductance. Insertion Gain: Gain 2 Insertion Gain: Phase Magnitude (db) ESL= ESL=1nH -1 ESL=5nH ESL=1nH ESL=5nH Phase (d) Fig Insertion gains of DM transmission-line EMI filter under different parasitic inductance. 127

155 4 Insertion Gain: Gain 2 Insertion Gain: Phase Magnitude (db) Phase (d) ESL= ESL=1nH ESL=1nH ESL=1nH Fig Transfer gains of DM transmission-line EMI filter under different parasitic inductance. Two kinds of connectors are tried on the measurement board of the DM transmissionline EMI filter: a BNC connector and an SMB connector, as shown in Fig The BNC connector is much larger than the SMB connector. The BNC connector also has much longer wires than the SMB connector. Fig SMB and BNC connectors. Fig shows the measurement board of DM transmission-line EMI filter with BNC connectors. Fig shows the connection of the BNC connector to the copper strip on the PCB board. The long wire of the BNC connector and the copper strips on the PCB board introduce parasitic inductances to the measurement setup. 128

156 Fig Measurement board with BNC connectors. Fig Connection of BNC connector to board copper strip. Fig and Fig. 5-5 show two measurement boards with SMB connectors. Fig shows the connection of the SMB connector to the filter. The central pin of the SMB connector is directly connected to the top conductor of the filter. The SMB connector is soldered onto a copper strip on the PCB, which connects to the bottom conductor of the filter. Great care has been taken to reduce the parasitic inductance as much as possible. Besides using smaller connectors, applying a four-terminal measurement setup can also reduce the parasitic inductance. Fig shows a two-terminal measurement setup for the DM transmission-line EMI filter. The source and load are connected to these two terminals and the voltages of these two terminals are measured. Fig. 5-5 shows a fourterminal measurement setup. The source and load are connected to the A and B terminals, respectively; however, the voltages of the C and D terminals are measured. As discussed in 5.2.2, C and D are connected to high-impedance (1MΩ) adapters. The currents drawn from C and D terminals are close to zero. The voltage drops on the parasitic inductances 129

157 of C and D terminals are close to zero. Although the voltage drops on the parasitic inductances of A and B terminals are not zero; it is not included in the measured results. Fig Two-terminal measurement board with SMB connectors. C A B D Fig Four-terminal measurement board with SMB connectors. Fig Connection of SMB connector to filter. The measured transfer gains using BNC connectors, two-terminal SMB connectors and four-terminal SMB connectors are compared in Fig The measurement board with BNC connectors has the largest parasitic inductance, and the transfer gain curve is shifted to a much lower frequency. The two-terminal measurement board with SMB connectors still has an appreciable parasitic inductance. Its transfer gain is also shifted to lower frequency. The four-terminal measurement board with SMB connectors gives the most accurate transfer gain measurement result. The four-terminal measurement board 13

158 with SMB connectors is used to measure the transfer gains of all the filter prototypes that will be shown in Chapter 6. For the insertion gain measurement, if the test kit is used, the four-terminal measurement board can not be used because there is only one terminal connecting the test kit and the filter. The four-terminal measurement board can be used for the twomeasurement method without the test kit. Fig shows the measured insertion gains for two-terminal and four-terminal SMB measurement boards of the two-measurement method without test kit and the two-terminal SMB board of one-measuremnt method with test kit. The insertion gain curves are almost the same, except small difference at very high frequency. This confirms the result from the model that the insertion gain curve is not sensitive to the parasitic inductance. The two-terminal measurement setup with SMB connectors is used to measure all the insertion gains of the filter prototypes. 2 Transfer Gain -2 db E+3 1.E+4 1.E+5 1.E+6 1.E+7 1.E+8 BNC connectors 2-terminal SMB connectors 4-terminal SMB connectors Fig Measured transfer gains using different measurement setups. 131

159 Insertion Gain db E+3 1.E+4 1.E+5 1.E+6 1.E+7 1.E+8 Frequency two-terminal SMB, two-measurement without test kit four-terminal SMB, two-measurement without test kit two-terminal SMB, one-measurement with test kit Fig Measured insertion gains using different measurement setups Summary The measurement setups of the DM transmission-line EMI filter using network analyzer were discussed in this chapter. There were two measurement setups for the insertion gain: two-measurement method without the transmission/reflection test kit and one-measurement method with the transmission/reflection test kit. The results from these two measurement setups were identical. The one-measurement method with the transmission/reflection test kit was chosen due to simplicity. The network analyzer had a common-ground problem that will influence the measured results of the high-frequency attenuator. However, the common-ground problem had negligible influence on the measured results of the DM transmission-line EMI filter. In order to compare the parametric study results with the measured results, the parametric study was performed again for the high-frequency attenuator taking into 132

160 account the common-ground problem. The new insertion gain and transfer gain curves showed less high-frequency attenuations and more pronounced transmission-line effect than the insertion gain and transfer gain curves in Chapter 3. However, the conclusions from the parametric study on the high-frequency attenuator still applied. The connectors and copper strips between the connectors and the filter introduced parasitic inductances to the measurement setup. The multi-conductor lossy transmissionline model was used to investigate the influence of the parasitic inductances on the measured insertion gain and transfer gain curves. The results showed that the transfer gain was very sensitive to the parasitic inductances, while the insertion gain curve was not. There were two major methods to reduce the parasitic inductance of the measurement setup: use small connectors and apply four-terminal measurement. The transfer gain curves of three measurement setups are compared: two-terminal BNC connector, twoterminal SMB connector, and four-terminal SMB connector. There were appreciable differences between each two of these transfer gain curves. The four-terminal SMB measurement setup was the most accurate one and will be used for all the following transfer gain measurements. For the insertion gain curves, four-terminal measurement setup gave the same result as the two-terminal measurement setup. Thus the two-terminal measurement setup could be still be used. 133

161 Chapter 6: Experimental verification 6.1. Introduction In Chapter 3 and chapter 4, the parametric study of the high-frequency attenuator and the DM transmission-line EMI filter are performed based on the multi-conductor lossy transmission-line model. In this chapter, experimental results are given to verify the results from the electromagnetic modeling. Eight high-frequency attenuator prototypes and twelve DM transmission-line EMI filter prototypes with different materials and dimensions are built and measured using the measurement setup described in Chapter 5. Before comparing the measured results with the simulated results from the model, all the material properties of the prototypes must be known. The measurement of some material properties are performed, such as the conductivity and relative permeability of nickel, the relative permittivity and loss factor of Y5V dielectric material. For every prototype, the measured insertion gain and transfer gain curves are shown together with the simulated insertion gain and transfer gain curves for better comparison Material properties of DM transmission-line EMI filter From the parametric study in Chapter 3 and Chapter 4, the influence of the material properties on the filter performance is clearly shown. Some material properties are essential to the attenuation of the filter, such as the relative permeability of nickel, and the relative permittivity of dielectric material. Some material properties depend on the process of how the material is acquired. Other material properties depend on the working environment, such as the applied electric field, environmental temperature, etc. It is 134

162 important to get an accurate value of the material properties before the experimental verification Nickel s properties Nickel is a soft, ordered, ferromagnetic material. The properties of commercially pure grade nickel are listed in Table 6-1 [51]. TABLE 6-1 PROPERTIES OF NICKEL 2, COMMERCIALLY PURE GRADE (99.6% NI). PROPERTIES VALUE Annealed Tensile Strength at 2 C 45Mpa Annealed.2% Proof Stress at 2 C 15Mpa Elongation (%) 47 Density 8.89g/cm 3 Melting Range C Specific Heat 456 J/kg. C Curie Temperature 36 C Relative Permeability Initial 11 Maximum 6 Co-Efficient of Expansion (2-1 C) 13.3x1-6 m/m. C Thermal Conductivity 7W/m. C Electrical Resistivity.96x1-6 ohm.m Electroplated nickel is commonly used for corrosion prevention, controlling hardness and magnetic properties, etc. Electroplated nickel can be deposited soft or hard, dull or bright. There are several different kinds of plating baths, some examples include sulfamate, Watts, bright and, Chloride, etc [46][52][53]. Nickel sulfamate bath is most widely used. Sulfamate nickel is a pure deposit that allows soldering and brazing during later assembly steps. It is normally a dull grey to dull silver color depending upon the finish of the plated part [52]. 135

163 Watts nickel is deposited from a nickel sulfate bath. Watts nickel normally yields a brighter finish than does sulfamate nickel since even the dull Watts bath contains a grain refiner to improve the deposit. Watts nickel may also be deposited as a semi-bright finish. Semi-bright Watts nickel achieves a brighter deposit because the bath contains organic and/or metallic brighteners. The brighteners may lead to problems with soldering and brazing, and this should be considered when selecting a finish for the present application [52][54]. A bright nickel bath contains a higher concentration of organic brighteners, which have a levelizing effect on the deposit. Organic brighteners yield a lustrous deposit; however, the negative effects on later assembly operations such as brazing and soldering should be considered [46][52]. The properties of nickel, such as density, specific heat, CTE, electrical resistivity, relative permeability, corrosion resistance, and hardness vary with solution composition, PH value of solution, and type and amount of organic additives. Operating variables such as current density and temperature also alter the properties of electroplated nickel. [46][47]. For a transmission-line EMI filter, as discussed in Chapter 3 and Chapter 4, the electrical resistivity and relative permeability of nickel are of most interest. Table 6-2 shows the electrical resistivity of electroplated nickel from various solutions at 2 C [46]. Comparative values of some magnetic properties of electroplated nickel and metallurgical nickel are given in Table 6-3. The coercive force of electroplated nickel with a thickness larger than 1µm shows a nearly constant coercive force of 56 A/m. Below 1µm, the coercive force depends on the thickness of electroplated nickel. 136

164 TABLE 6-2 ELECTRICAL RESISTIVITIES OF ELECTROPLATED NICKEL FROM VARIOUS SOLUTIONS TYPE OF BATH RESISTIVITY(microhm-cm) at 2 C Acetate 9.18 Chloride 8.24 Chloride, cobalt-free 7.71 Co-bright Fluoborate 8.34 Organic bright 1. Sulfamate 8.6 Watts 7.76 Watts, cobalt-free 7.44 Watts plus Na 2 SO Watts plus (NH 4 ) 2 SO TABLE 6-3 MAGNETIC PROPERTY OF ELECTROPLATED NICKEL MAGNETIC PROPERTY Electroplated nickel Metallurgical nickel Coercive force, A/m 32~96 27 Residual induction, Tesla.2~.45 - Saturation magnetization, Tesla ~.7.61 Nickel sulfamate solution from Technic Inc. is used during the manufacturing process of the transmission line EMI filter. The temperature of the nickel electroplating process is at 12 F (49 C). The electrical conductivity and relative permeability are measured for the electroplated nickel. The electrical conductivity of electroplated nickel is obtained by measuring the resistance of an electroplated nickel strip. Fig. 6-1 shows the dimension of the 137

165 electroplated nickel sample. The resistance of the nickel strip is measured using an impedance analyzer (Agilent 4294A). The measurement is done for a frequency range 1Hz to 1 khz. For frequencies higher than 1kHz, an accurate resistance value is difficult to get due to the existence of ωl. Fig. 6-2 shows the measured resistance of the electroplated nickel strip. Fig. 6-3 shows the calculated electrical conductivity of electroplated nickel from the resistance measurement. The average value is around 1.45e7 S/m. For the electroplated nickel in the experimental condition of the packaging lab in CPES, 1.45e7 S/m is chosen as the electrical conductivity. t=5um L=9.cm W=1.2cm Fig Dimension of electroplated nickel strip. Ohm Resistance 1.E+2 1.E+3 1.E+4 Fig Resistance of electroplated nickel strip. 138

166 S/m 2.E+7 1.9E+7 1.8E+7 1.7E+7 1.6E+7 1.5E+7 1.4E+7 1.3E+7 1.2E+7 1.1E+7 1.E+7 Nickel Conductivity 1.E+2 1.E+3 1.E+4 Fig Electrical conductivity of electroplated nickel strip. The measurement of nickel s relative permeability is not a trivial issue. In [45], a toroid core of electroplated nickel was made. A signal generator was directly coupled to the wire wound on the core. The B-H hysteresis loop of the core was measured, as shown in Fig However, the signal generator was not able to provide enough current to saturate the core. The B-H hysteresis is only partial. The average relative permeability of nickel was calculated from the hysteresis loop, as shown in Table 6-4. Fig B-H hysteresis loops of nickel core at different frequencies [36]. 139

167 TABLE 6-4 CALCULATED AVERAGE RELATIVE PERMEABILITY OF NICKEL CORE There are some commercially available equipments that can measure material s magnetic properties. The vibrating sample magnetometer (VSM) has become a widely used instrument for determining magnetic properties of a large variety of materials: diamagnetics, paramagnetics, ferromagnetics, ferromagnetics and antiferromagnetics. The VSM machine in University of Florida was used to measure the relative permeability of the electroplated nickel produced in the CPES packaging lab. Fig. 6-5 shows the vibrating sample magnetometer equipment. It operates based on Faraday s law of induction. If a sample of any material is placed in a uniform magnetic field, created between the poles of an electromagnet, a dipole moment will be induced. If the sample vibrates with sinusoidal motion a sinusoidal electrical signal can be induced in suitably placed pick-up coils. The signal has the same frequency of vibration and its amplitude will be proportional to the magnetic moment, amplitude, and relative position with respect to the pick-up coils system. The sample is fixed to a small sample holder located at the end of a sample rod mounted in an electromechanical transducer [55]. Fig. 6-6 shows the measured B-H loop of the electroplated nickel using VRM. This B-H loop is for a frequency of 9Hz. Fig. 6-7 shows the calculated relative permeability of nickel from Fig The relative permeability of the electroplated nickel is around

168 This value is used for the frequency range from DC to 1MHz, assuming that in this frequency range, nickel s relative permeability is not very sensitive to frequency. Fig vibrating sample magnetometer [55]. B (Tesla) E+4-2.E+4.E E+4 4.E H (A/m) Fig Measured B-H loop of nickel using VRM. 141

169 4 35 Relative permeability E+4-2.E+4.E+ 2.E+4 4.E+4 H (A/m) Fig Measured relative permeability of electroplated nickel Dielectric s properties Y5V is used as the ceramic material in the prototype of the DM transmission-line EMI filter. It is one of the Electronic Industries Alliance (EIA) class 2 dielectric materials. The EIA Class 2 dielectrics are usually based on formulas with a high content of barium titanate (BT), possibly mixed with other dielectric electroceramics. The EIA three-character code is derived from the low and high temperature limit, and the range of capacitance change, as shown in Table 6-5 [56]. In comparison with the EIA Class 1 dielectrics, the EIA Class 2 dielectrics tend to have severe temperature drift, high dependence of capacitance on applied voltage, a high voltage coefficient of dissipation factor and a high-frequency coefficient of dissipation. For Y5V, with temperature range of -3 C to +85 C, capacitance changes with temperature of +22/-82%. Some electrical, mechanical and thermal properties of commonly used dielectric materials are listed in Table 6-6 [57]. 142

170 TABLE 6-5 EIA THREE-CHARACTER CODE LOW TEMP HIGH TEMP CAPACITANCE CHANGE RANGE X: -55 C 4: +65 C A: +/- 1.% Y: -3 C 5: +85 C B: +/- 1.5% : 1 C 6: +15 C C: +/- 2.2% 7: +125 C D: +/- 3.3% 8: +15 C E: +/- 4.7% 9: +2 C F: +/- 7.5% P: +/- 1% R: +/- 15% S: +/- 22% T: -33%, +22% U: -56%, +22% V: -82%, +22% TABLE 6-6 PROPERTIES OF DIELECTRIC MATERIALS Material Relative permittivity Breakdown E (kv/mm) Density (kg/m 3 ) Young s modulus (GPa) Y5V X5U X7R X5S N NP In [4], the capacitance and loss factor of a capacitor with Y5V dielectric was measured. Fig. 6-8 shows the measured results under different frequencies. This is a 143

171 small signal measurement using an impedance analyzer. As frequency goes up, the capacitance decreases, however, the loss factor increases. Fig Measured capacitance and loss factor vs. frequency. A large signal dielectric characterization method was proposed and the details about the measurement circuits can be found in [48]. The D-E hysteresis loop, relative permittivity, and dielectric loss density of several dielectric materials were given. The temperature dependency of different dielectric materials was also discussed. Fig. 6-9 shows the measured relative permittivity of Y5V under different electric field strength and temperature. As the electric field strength or temperature increases, the relative permittivity of Y5V reduces rapidly. 144

172 C Relative Permittivity C 9 C 12 C Electric field strength (V/m) x 1 6 Fig Measured relative permittivity of Y5V vs. electric field strength and temperature Manufacturing process of DM transmission-line EMI filter The major manufacturing processes of the DM transmission-line EMI filter include cleaning, sputtering, electroplating, photolithography, etching, laser cutting, and etc. The details of the manufacturing processes are described in Appendix I Experimental verification on high-frequency attenuator Chapter 3 and Chapter 4 have discussed the parametric study of the high frequency attenuator and the DM transmission-line EMI filter based on the multi-conductor lossy transmission-line model. The dependencies of the filter performance on dimensions and material properties are identified. To verify the results from the model, a series of prototypes with different dimensions and materials are manufactured; the insertion gains and transfer gains of these prototypes are measured. Table 6-7 is a summary of the parameter values that have been tried on the highfrequency attenuator prototypes. There are totally eight prototypes. One is built based on 145

173 the nominal values. Please note that the nominal values here are not exactly the same as the nominal design in Chapter 3. For example, the nominal value for nickel s relative permeability is 6 in Chapter 3, which is the maximum relative permeability of nickel [51]. However, the relative permeability of the electroplated nickel of the prototypes is 27, according to the VRM measurement result in Fig For the other seven prototypes, there is only one parameter different from the nominal value; all the other parameters keep the nominal values. Some material properties, such as nickel s conductivity and relative permeability and, dielectric s loss factor are difficult to change or control in the experiments. These parameters keep the nominal values in all the prototypes. TABLE 6-7 PARAMETERS OF HIGH-FREQUENCY ATTENUATOR PROTOTYPES PARAMETERS NOMINAL VALUE OTHER VALUES Length (cm) 12 3, 6 Dimensions Width (mm) 1 5, 25 Conductor thickness (µm) 25 5 Dielectric thickness (µm) 15 3 Conductor Dielectric Conductivity (Ω -1 /m) 1.45e7 * Relative permeability 27 * Relative permittivity Loss factor 4% * Fig. 6-1 and Fig show the measured insertion gains and transfer gains of the high-frequency attenuator prototype under nominal design. The results from the model are plotted in the same graph for comparison. There is good agreement between the results from the measurement and results from the model. 146

174 Gain (db) Insertion Gain of HF Attenuator E+3 1.E+4 1.E+5 1.E+6 1.E+7 1.E+8 Measurement Multi-conductor TL model Phase (Degree) Insertion Gain of HF Attenuator E+3 1.E+4 1.E+5 1.E+6 1.E+7 1.E+8 Measurement Multi-conductor TL model Fig Measured and simulated insertion gains of high-frequency attenuator under nominal design. Gain (db) Transfer Gain of Attenuator E+3 1.E+4 1.E+5 1.E+6 1.E+7 1.E+8 Measurement Multi-conductor TL model Phase (Degree) Transfer Gain of Attenuator E+3 1.E+4 1.E+5 1.E+6 1.E+7 1.E+8 Measurement Multi-conductor TL model Fig Measured and simulated transfer gains of high-frequency attenuator under nominal design. Fig shows the measured insertion gains of three high-frequency attenuator prototypes with length 3cm, 6cm and 12cm. A longer structure has a lower cut-off frequency and more attenuation. The simulated results are plotted in the same graph for comparison. Fig shows the measured and simulated transfer gain curves for these three prototypes. 147

175 Gain (db) Insertion Gain of HF Attenuator vs. Total Length 1.E+3 1.E+4 1.E+5 1.E+6 1.E+7 1.E+8 L=12cm_measurement L=6cm_measurement L=3cm_measurement L=12cm_model L=6cm_model L=3cm_model Phase (Degree) Insertion Gain of HF Attenuator vs. Total Length 1.E+3 1.E+4 1.E+5 1.E+6 1.E+7 1.E+8 L=12cm_measurement L=6cm_measurement L=3cm_measurement L=12cm_model L=6cm_model L=3cm_model Fig Measured and simulated insertion gains of high-frequency attenuator vs. length. Gain (db) Transfer Gain of HF Attenuator vs. Total Length 1.E+3 1.E+4 1.E+5 1.E+6 1.E+7 1.E+8 L=12cm_measurement L=6cm_measurement L=3cm_measurement L=12cm_model L=6cm_model L=3cm_model Phase (Degree) Transfer Gain of HF Attenuator vs. Total Length 1.E+3 1.E+4 1.E+5 1.E+6 1.E+7 1.E+8 L=12cm_measurement L=6cm_measurement L=3cm_measurement L=12cm_model L=6cm_model L=3cm_model Fig Measured and simulated transfer gains of high-frequency attenuator vs. length. Fig shows the measured insertion gains of three high-frequency attenuator prototypes with width 5mm, 1mm and 25mm. The wider structure has a lower f c1. However, f c2 keeps almost the same for these three prototypes. The same conclusion was drawn from the parametric study in Chapter 3. The simulated results for these three prototypes are also plotted in Fig for comparison. Fig shows the measured and simulated transfer gain curves for these three prototypes. The transfer gain curves are 148

176 the same for these three prototypes with different widths, both from the measurement and from the model. Gain (db) Insertion Gain of HF Attenuator vs. Width Insertion Gain of DM TL EMI Filter vs. Width E+3 1.E+4 1.E+5 1.E+6 1.E+7 1.E+8 1.E+3 1.E+4 1.E+5 1.E+6 1.E+7 1.E+8 width=5mm (measurement) width=1mm (measurement) width=5mm (measurement) width=1mm (measurement) width=25mm (measurement) width=5mm (model) width=25mm (measurement) width=5mm (model) width=1mm (model) width=25mm (model) width=1mm (model) width=25mm (model) Fig Measured and simulated insertion gains of high-frequency attenuator vs. width. Phase (Degree) -1 1 Transfer Gain of HF Attenuator vs. Width 25 Transfer Gain of HF Attenuator vs. Width Gain (db) -2-3 Phase (Degree) E+3 1.E+4 1.E+5 1.E+6 1.E+7 1.E+8 width=5mm (measurement) width=1mm (measurement) width=25mm (measurement) width=5mm (model) width=1mm (model) width=25mm (model) E+3 1.E+4 1.E+5 1.E+6 1.E+7 1.E+8 width=5mm (measurement) width=1mm (measurement) width=25mm (measurement) width=5mm (model) width=1mm (model) width=25mm (model) Fig Measured and simulated transfer gains of high-frequency attenuator vs. width. Two high-frequency attenuator prototypes with nickel thickness 5µm and 25µm are built. Fig shows insertion gains of these two prototypes. The prototype with a 5µm nickel thickness has higher attenuation than the 25µm one in the frequency range near f c2. The analysis in Chapter 3 showed that for a nickel thickness larger than 5µm, the 149

177 attenuation curves are the same. The previous results are under the assumption that nickel s relative permeability is 6. If nickel s relative permeability changes to 27, the skin depth of nickel changes too, as shown in Fig The skin depth at 1MHz is about 24µm, at 1MHz it is about 7.6µm, and at 3MHz it is about 5µm. The attenuation for the prototype with 5µm nickel has a lower attenuation than the one with 25µm nickel until around 3MHz. The measured and simulated transfer gains of these two prototypes are shown in Fig From 1MHz to 3MHz, the prototype with 5µm nickel has a larger attenuation than the one with 25µm nickel. There is good agreement between the measured and simulated curves for both insertion gain and transfer gain. Gain (db) Insertion Gain of HF Attenuator vs. Nickel Thickness E+3 1.E+4 1.E+5 1.E+6 1.E+7 1.E+8 tni=25um (measurement) tni=5um (measurement) tni=25um (model) tni=5um (model) Phase (Degree) Insertion Gain of HF Attenuator vs. Nickel Thickness E+3 1.E+4 1.E+5 1.E+6 1.E+7 1.E+8 tni=25um (measurement) tni=5um (measurement) tni=25um (model) tni=5um (model) Fig Measured and simulated insertion gains of high-frequency attenuator vs. thickness of nickel. 15

178 1 3 skin depth of nickel conductor µ Ni =6 µ Ni = (um) Fig Skin depth of nickel. Gain (db) 1 Transfer Gain of HF Attenuator vs. Nickel Thickness 25 Transfer Gain of HF Attenuator vs. Nickel Thickness E+3 1.E+4 1.E+5 1.E+6 1.E+7 1.E+8 1.E+3 1.E+4 1.E+5 1.E+6 1.E+7 1.E+8 tni=25um (measurement) tni=5um (measurement) tni=25um (measurement) tni=5um (measurement) tni=25um (model) tni=5um (model) tni=25um (model) tni=5um (model) Phase (Degree) Fig Measured and simulated transfer gains of high-frequency attenuator vs. thickness of nickel. Fig shows the measured and simulated insertion gains for the two highfrequency attenuator prototypes with Y5V dielectric thicknesses of 15µm and 3µm respectively. The thinner dielectric gives a lower cut-off frequency and more attenuation. The transfer gain curves are shown in Fig

179 Gain (db) Insertion Gain of HF Attenuator vs. Dielectric Thickness 1.E+3 1.E+4 1.E+5 1.E+6 1.E+7 1.E+8 tdie=15um_measurement tdie=3um_measurement tdie=15um_model tdie=3um_model Phase (Degree) Insertion Gain of HF Attenuator vs. Dielectric Thickness 1.E+3 1.E+4 1.E+5 1.E+6 1.E+7 1.E+8 tdie=15um_measurement tdie=3um_measurement tdie=15um model tdie=3um model Fig Measured and simulated insertion gains of high-frequency attenuator vs. thickness of dielectric. Gain (db) 1 Transfer Gain of HF Attenuator vs. Dielectric Thickness 25 Transfer Gain of HF Attenuator vs. Dielectric Thickness E+3 1.E+4 1.E+5 1.E+6 1.E+7 1.E+8 1.E+3 1.E+4 1.E+5 1.E+6 1.E+7 1.E+8 tdie=15um_measurement tdie=3um_measurement tdie=15um_measurement tdie=3um_measurement tdie=15um_model tdie=3um_model tdie=15um_model tdie=3um_model Phase (Degree) Fig Measured and simulated transfer gains of high-frequency attenuator vs. thickness of dielectric. One high-frequency attenuator prototype is built using the X7R dielectric material. The relative permittivity of X7R is around 35. Fig shows the measured and simulated insertion gains of two high-frequency attenuator prototypes with Y5V and X7R respectively. Lower relative permittivity gives a higher cut-off frequency and less attenuation. The transfer gain curves are shown in Fig

180 Gain (db) Insertion Gain of HF Attenuator vs. Dielectric Constant 1.E+3 1.E+4 1.E+5 1.E+6 1.E+7 1.E+8 Y5V_measurement X7R_measurement Y5V Model X7R model Phase (Degree) Insertion Gain of HF Attenuator vs. Dielectric Constant 1.E+3 1.E+4 1.E+5 1.E+6 1.E+7 1.E+8 Y5V_measurement X7R_measurement Y5V model X7R model Fig Measured and simulated insertion gains of high-frequency attenuator vs. dielectric material. Gain (db) Transfer Gain of HF Attenuator vs. Dielectric Constant E+3 1.E+4 1.E+5 1.E+6 1.E+7 1.E+8 Y5V_measurement X7R_measurement Y5V model X7R model Phase (Degree) Transfer Gain of HF Attenuator vs. Dielectric Constant E+3 1.E+4 1.E+5 1.E+6 1.E+7 1.E+8 Y5V_measurement X7R_measurement Y5V model X7R model Fig Measured and simulated transfer gains of high-frequency attenuator vs. dielectric material. For the high-frequency attenuator prototypes, the measured insertion gain and transfer gain curves agree well with the results from the model. The influence of dimensions and material properties on high-frequency attenuator performance draws the same conclusions as in Chapter

181 6.5. Experimental verification on DM transmission-line EMI filter Twelve DM transmission-line EMI filters are built to verify the parametric study results from the model in Chapter 4. Table 6-8 summaries the parameter values that have been tried on the DM transmission-line EMI filter prototypes. One prototype is built based the nominal values. All the other prototypes have only one parameter different from the nominal values. TABLE 6-8 PARAMETERS OF DM TRANSMISSION-LINE EMI FILTER PROTOTYPES Overall dimension Inner conductor Dielectric Insulation PARAMETERS NOMINAL VALUE OTHER VALUES Length (cm) 12 3, 6 Width (mm) 1 5, 25 Copper thickness (µm) 25 75, 15 Nickel thickness (µm) 25 5 Dielectric thickness (µm) 15 3 Insulation thickness (µm) 5 125, 25 Conductivity (Ω -1 /m) 1.45e7 * Relative permeability 27 * Relative permittivity Loss factor 4% * Relative permittivity 9.8 * Loss factor.1% * Fig and Fig show the measured insertion gains and transfer gains of the DM transmission-line EMI filter prototype under nominal design. The results from the model are plotted in the same graph for comparison. The results from the measurement and the model are in agreement. 154

182 Gain (db) Insertion gain of DM TL EMI filter 1.E+3 1.E+4 1.E+5 1.E+6 1.E+7 1.E+8 Measurement Multi-conductor TL model Phase (Degree) Insertion Gain of DM TL EMI filter 1.E+3 1.E+4 1.E+5 1.E+6 1.E+7 1.E+8 Measurement Multi-conductor TL model Fig Measured and simulated insertion gains of DM TL EMI filter under nominal design. Gain (db) Transfer gain of DM TL EMI filter E+3 1.E+4 1.E+5 1.E+6 1.E+7 1.E+8 Measurement Multi-conductor TL model Phase (Degree) Transfer gain of DM TL EMI filter 1.E+3 1.E+4 1.E+5 1.E+6 1.E+7 1.E+8 Measurement Multi-conductor TL model Fig Measured and simulated transfer gains of DM transmission-line EMI filter under nominal design. Fig shows the measured insertion gains of three DM transmission-line EMI filter prototypes with length 3cm, 6cm and 12cm. The insertion gain curves from the model are plotted in the same graph for comparison. Fig shows the measured and simulated transfer gain curves for these three prototypes. 155

183 Gain (db) Insertion Gain of DM TL EMI filter vs. Total Length E+3 1.E+4 1.E+5 1.E+6 1.E+7 1.E+8 L=12cm_measurement L=6cm_measurement L=3cm_measurement L=12cm_model L=6cm_model L=3cm_model Phase (Degree) Insertion Gain of DM TL EMI filter vs. Total Length E+3 1.E+4 1.E+5 1.E+6 1.E+7 1.E+8 L=12cm_measurement L=6cm_measurement L=3cm_measurement L=12cm_model L=6cm_model L=3cm_model Fig Measured and simulated insertion gains of DM transmission-line EMI filter vs. length. Gain (db) Transfer Gain of DM filter vs. Total Length 1.E+3 1.E+4 1.E+5 1.E+6 1.E+7 1.E+8 L=12cm_measurement L=6cm_measurement L=3cm_measurement L=12cm_model L=6cm_model L=3cm_model Phase (Degree) Transfer Gain of DM filter vs. Total Length 1.E+3 1.E+4 1.E+5 1.E+6 1.E+7 1.E+8 L=12cm_measurement L=12cm_measurement L=12cm_measurement L=12cm_model L=6cm_model L=3cm_model Fig Measured and simulated transfer gains of DM transmission-line EMI filter vs. length. Fig shows the measured insertion gains of three DM transmission-line EMI filter prototypes of width 5mm, 1mm and 25mm. The wider structure has a lower f c1. However, f c2 stays mostly the same for these three prototypes. The same conclusion was drawn from the parametric study in Chapter 4. The simulated results for these three prototypes are also plotted in the same graph for comparison. Fig shows the measured and simulated transfer gain curves for these three prototypes. The transfer gain 156

184 curves are almost the same for these three prototypes with different widths, both from measurement and from model. Gain (db) Insertion Gain of DM TL EMI Filter vs. Width 1.E+3 1.E+4 1.E+5 1.E+6 1.E+7 1.E+8 width=5mm (measurement) width=1mm (measurement) width=25mm (measurement) width=5mm (model) width=1mm (model) width=25mm (model) Phase (Degree) Insertion Gain of DM TL EMI Filter vs. Width 1.E+3 1.E+4 1.E+5 1.E+6 1.E+7 1.E+8 width=5mm (measurement) width=1mm (measurement) width=25mm (measurement) width=5mm (model) width=1mm (model) width=25mm (model) Fig Measured and simulated insertion gains of DM transmission-line EMI filter vs. width. Gain (db) Transfer Gain of DM TL EMI filter vs. Width 1.E+3 1.E+4 1.E+5 1.E+6 1.E+7 1.E+8 width=5mm (measurement) width=1mm (measurement) width=25mm (measurement) width=5mm (model) width=1mm (model) width=25mm (model) Phase (Degree) Transfer Gain of DM TL EMI filter vs. Width 1.E+3 1.E+4 1.E+5 1.E+6 1.E+7 1.E+8 width=5mm (measurement) width=1mm (measurement) width=25mm (measurement) width=5mm (model) width=1mm (model) width=25mm (model) Fig Measured and simulated transfer gains of DM transmission-line EMI filter vs. width. Two DM transmission-line EMI filter prototypes with nickel thickness 5µm and 25µm are built. Fig shows the insertion gains of these two prototypes. The prototype with the 5µm nickel thickness has a higher attenuation than the 25µm one in the frequency range near f c2. The measured and simulated transfer gains of these two 157

185 prototypes are shown in Fig There is good agreement between the measured and simulated curves for both the insertion gain and the transfer gain. Gain (db) Insertion Gain of DM Filter vs. Nickel Thickness E+3 1.E+4 1.E+5 1.E+6 1.E+7 1.E+8 tni=25um (measurement) tni=5um (measurement) tni=25um (model) tni=5um (model) Gain (db) Insertion Gain of DM Filter vs. Nickel Thickness E+3 1.E+4 1.E+5 1.E+6 1.E+7 1.E+8 tni=25um (measurement) tni=5um (measurement) tni=25um (model) tni=5um (model) Fig Measured and simulated insertion gains of DM transmission-line EMI filter vs. nickel thickness. Gain (db) Transfer Gain of DM Filter vs. Nickel Thickness 1.E+3 1.E+4 1.E+5 1.E+6 1.E+7 1.E+8 tni=25um (measurement) tni=5um (measurement) tni=25um (model) tni=5um (model) Phase (Degree) Transfer Gain of DM Filter vs. Nickel Thickness 1.E+3 1.E+4 1.E+5 1.E+6 1.E+7 1.E+8 tni=25um (measurement) tni=5um (measurement) tni=25um (model) tni=5um (model) Fig Measured and simulated transfer gains of DM transmission-line EMI filter vs. nickel thickness. Three DM transmission-line EMI filter prototypes with copper thickness 25µm, 75µm and 15µm are built. Fig shows the insertion gains of these three prototypes. From these curves, the same conclusion can be drawn as in Chapter 4: the copper 158

186 thickness has little influence on the filter performance. Fig shows the transfer gains of these three prototypes. Gain (db) Insertion Gain of DM Filter vs. Copper Thickness Insertion Gain of DM Filter vs. Copper Thickness E+3 1.E+4 1.E+5 1.E+6 1.E+7 1.E+8 1.E+3 1.E+4 1.E+5 1.E+6 1.E+7 1.E+8 tcu=25um (measurement) tcu=75um (measurement) tcu=25um (measurement) tcu=75um (measurement) tcu=15um (measurement) tcu=25um (model) tcu=15um (measurement) tcu=25um (model) tcu=75um (model) tcu=15um (model) tcu=75um (model) tcu=15um (model) Phase (Degree) -1 Fig Measured and simulated insertion gains of DM transmission-line EMI filter vs. copper thickness. Gain (db) Transfer Gain of DM Filter vs. Copper Thickness 1.E+3 1.E+4 1.E+5 1.E+6 1.E+7 1.E+8 tcu=25um_measurement tcu=75um_measurement tcu=15um_measurement tcu=25um_model tcu=75um_model tcu=15um_model Phase (Degree) Transfer Gain of DM Filter vs. Copper Thickness 1.E+3 1.E+4 1.E+5 1.E+6 1.E+7 1.E+8 tcu=25um_measurement tcu=75um_measurement tcu=15um_measurement tcu=25um_model tcu=75um_model tcu=15um_model Fig Measured and simulated transfer gains of DM transmission-line EMI filter vs. copper thickness. Fig shows the measured and simulated insertion gains of two DM transmissionline EMI filter prototypes with Y5V dielectric thicknesses 15µm and 3µm respectively. Thinner dielectric gives a lower cut-off frequency and more attenuation. The transfer gain curves are shown in Fig

187 Gain (db) Insertion Gain of DM TL EMI Filter vs. Dielectric Thickness E+3 1.E+4 1.E+5 1.E+6 1.E+7 1.E+8 tdie=15um_measurement tdie=3um_measurement tdie=15um_model tdie=3um_model Phase (Degree) Insertion Gain of DM TL EMI Filter vs. Dielectric Thickness E+3 1.E+4 1.E+5 1.E+6 1.E+7 1.E+8 tdie=15um_measurement tdie=3um_measurement tdie=15um_model tdie=3um_model Fig Measured and simulated insertion gains of DM transmission-line EMI filter vs. dielectric thickness. Gain (db) Transfer Gain of DM TL EMI Filter vs. Dielectric Thickness 1.E+3 1.E+4 1.E+5 1.E+6 1.E+7 1.E+8 tdie=15um_measurement tdie=3um_measurement tdie=15um_model tdie=3um_model Phase (Degree) Transfer Gain of DM TL EMI Filter vs. Dielectric Thickness 1.E+3 1.E+4 1.E+5 1.E+6 1.E+7 1.E+8 tdie=15um_measurement tdie=3um_measurement tdie=15um_model tdie=3um_model Fig Measured and simulated transfer gains of DM transmission-line EMI filter vs. dielectric thickness. Three DM transmission-line EMI filter prototypes with insulation thickness 125µm, 25µm and 5µm are built and measured. Fig shows the measured and simulated insertion gains. Insulation thickness has little influence on the filter performance. The transfer gain curves are shown in Fig

188 Gain (db) Insertion Gain of DM TL EMI Filter vs. Insulation Thickness E+3 1.E+4 1.E+5 1.E+6 1.E+7 1.E+8 tinsu=125um (measurement) tinsu=25um (measurement) tinsu=5um (measurement) tinsu=125um (model) tinsu=25um (model) tinsu=5um (model) Phase (Degree) Insertion Gain of DM TL EMI Filter vs. Insulation Thickness E+3 1.E+4 1.E+5 1.E+6 1.E+7 1.E+8 tinsu=125um (measurement) tinsu=25um (measurement) tinsu=5um (measurement) tinsu=125um (model) tinsu=25um (model) tinsu=5um (model) Fig Measured and simulated insertion gains of DM transmission-line EMI filter vs. insulation thickness. Gain (db) Transfer Gain of DM TL EMI Filter vs. Insulation Thickness 1.E+3 1.E+4 1.E+5 1.E+6 1.E+7 1.E+8 tinsu=125um_measurement tinsu=25um_measurement tinsu=5um_measurement tinsu=125um_model tinsu=25um_model tinsu=5um_model Phase (Degree) Transfer Gain of DM TL EMI Filter vs. Insulation Thickness 1.E+3 1.E+4 1.E+5 1.E+6 1.E+7 1.E+8 tinsu=125um_measurement tinsu=25um_measurement tinsu=5um_measurement tinsu=125um_model tinsu=25um model tinsu=5um model Fig Measured and simulated transfer gains of DM transmission-line EMI filter vs. insulation thickness. One DM filter prototype is built using X7R (ε r =35) dielectric material. Its insertion gain curve is shown in Fig in comparison with the insertion gain of the filter with Y5V (ε r =14) dielectric. The transfer gain curves are compared in Fig

189 Gain (db) Insertion Gain of DM TL EMI Filter vs. Dielectric Constant 1.E+3 1.E+4 1.E+5 1.E+6 1.E+7 1.E+8 Y5V_measurement X7R_measurement Y5V model X7R model Phase (Degree) Insertion Gain of DM TL EMI Filter vs. Dielectric Constant 1.E+3 1.E+4 1.E+5 1.E+6 1.E+7 1.E+8 Y5V_measurement X7R_measurement Y5V_model X7R_model Fig Measured and simulated insertion gains of DM transmission-line EMI filter vs. dielectric material. Gain (db) Transfer Gain of DM TL EMI Filter vs. Dielectric Constant 1.E+3 1.E+4 1.E+5 1.E+6 1.E+7 1.E+8 Y5V_measurement X7R_measurement Y5V model X7R model Phase (Degree) Transfer Gain of DM TL EMI Filter vs. Dielectric Constant 1.E+3 1.E+4 1.E+5 1.E+6 1.E+7 1.E+8 Y5V_measurement X7R_measurement Y5V model X7R model Fig Measured and simulated transfer gains of DM transmission-line EMI filter vs. dielectric material. The experimental results on the high-frequency attenuator and DM transmission-line EMI filter verified the parametric study results in Chapter 3 and Chapter 4. There is good agreement between the results from experiments and from the model for these prototypes, which confirms that the multi-conductor lossy transmission line model is an accurate and effective model for the transmission-line filter. 162

190 6.6. Summary In this chapter, some important material properties were measured and discussed, such as the conductivity and relative permeability of nickel, the relative permittivity and loss factor of the Y5V dielectric material. Eight high-frequency attenuator prototypes and twelve DM transmission-line EMI filter prototypes with different materials and dimensions were built and measured. The measured results on the high-frequency attenuator and DM transmission-line EMI filter verified the parametric study results in Chapter 3 and Chapter 4. There was good agreement between the measured insertion gain and transfer gain curves and the simulated insertion gain and transfer gain curves from the multi-conductor lossy transmission-line model. 163

191 Chapter 7: Improved structures of DM transmission-line EMI filter 7.1. Introduction The previous chapters have shown how to improve the performance of the DM transmission-line EMI filter by changing its dimensions and materials. In this chapter, some additional ways to improve the performance of the filter are explored. First, a different connection method for the DM transmission-line EMI filter is investigated. Next, some other structures of the DM transmission-line EMI filter are studied, such as the structure without insulation layer, the meander structure, and the multi-dielectric-layer structure. Then, a broadband DM EMI filter is proposed, which combines the advantages of the discrete reflective LC EMI filter and the transmission-line EMI filter. As discussed in Chapter 1, the high-frequency attenuation of the discrete reflective LC EMI filter is impaired by the parasitics within the structure. The absorptive transmission-line EMI filter has large high-frequency attenuation. However, its roll-off slope at low frequencies is only -2dB/dec. The new broadband DM EMI filter has the advantages of both filters: large roll-off slope at low frequency and large attenuation at high frequency. The insertion gain curve of the proposed broadband DM EMI filter is predicted by the multi-conductor lossy transmission-line model. One prototype of the broadband DM EMI filter is built and the insertion gain curve is measured. Finally, the dependence of the broadband DM EMI filter s performance on source and load impedances is discussed in this chapter. 164

192 7.2. Different connections of DM transmission-line EMI filter When the DM transmission-line EMI filter is connected to power electronics converters, the top copper and nickel conductors should be connected at the source and load terminals, as should the bottom copper and nickel conductors, as shown in Fig. 7-1 (a). Another connection method that merits investigation is shown in Fig. 7-1(b). The copper and nickel conductors are connected only at the source side. At the load side, the copper conductor is connected to the load, while the nickel conductors are left open. (a) Original connection (b) New connection Fig Different connections of DM transmission-line EMI filter. As mentioned above, nickel only carries the high-frequency noise current. The current in nickel has been greatly attenuated from source to load, so there should be very little current at the load side in the nickel conductor. Disconnecting the nickel conductor from the load is explored here to see if it will affect the attenuation of the filter. The multi-conductor lossy transmission-line model is used to predict the performance of the filter with this new connection method. As shown in Fig. 7-2, ter is the termination impedance of the two nickel conductors. For the new connection method, the two nickel conductors are left open. ter should be a large impedance, for example 1MΩ. high is the impedance between the bottom nickel and copper conductors. It is also a large impedance 1MΩ. 165

193 Fig Measurement setup of DM transmission-line EMI filter with new connection method. For the new connection, the source and load impedance matrices of the multiconductor transmission-line model needs to be modified. The boundary conditions at z = can be expressed by the following equations: V () = V () = V 1 S [ I () + I ()] V () = V () = [ I () + I () + I () + I ()] S [ I () + I () + I () + I ()] SG 3 4 SG (7-1) Rewriting (2-67) in matrix form yields V( ) = V 4I() (7-2) S S where the input source vector is: [ V ] t V = (7-3) S S V S The source impedance matrix is: S + SG S + SG SG SG S + SG S + SG SG SG = (7-4) S4 SG SG SG SG SG SG SG SG The boundary conditions at the load side are: 166

194 167 LG LG high LG high ter LG L l I l I l I l I l V l I l I l I l I l I l I l V l I l I l I l I l I l I l I l V l I l I l I l I l I l V )] ( ) ( ) ( ) ( [ ) ( )] ( ) ( ) ( ) ( [ )] ( ) ( [ ) ( )] ( ) ( ) ( ) ( [ )] ( ) ( [ ) ( ) ( )] ( ) ( ) ( ) ( [ ) ( ) ( = = = = (7-5) Rewriting (2-71) in matrix form yields ) ( ) ( 4 l l I V L = (7-6) where the load impedance matrix is: = LG LG LG LG LG LG high LG high LG LG LG high LG high ter LG LG LG LG LG L L4 (7-7) The voltage on the load is: ) ( ) ( 4 1 l V l V V load = (7-8) The transfer gain of the EMI filter is: () () ) ( ) ( V V l V l V TG = (7-9) The insertion gain of the EMI filter is: L S L S V l V l V IG + = ) ( ) ( 4 1 (7-1)

195 Fig. 7-3 shows the simulated insertion gains for these two connection methods from the model. Apparently, the insertion gain of the new connection method is very different from the insertion gain of the original connection. The attenuation of the filter is greatly reduced at high frequencies. There is more pronounced transmission-line effect at high frequencies for the new connection method. The losses in the nickel conductors are not effectively used to attenuate the high-frequency noises for the new connection method. Fig. 7-4 shows the measured insertion gains of these two connection methods, which confirm the result from the model. -1 Insertion Gain: Gain Original connection New connection 2 15 Insertion Gain: Phase Magnitude (db) Phase (d) Fig Simulated insertion gains of DM transmission-line EMI filter under different connections. db Insertion gain of DM TL EMI filter E+3 1.E+4 1.E+5 1.E+6 1.E+7 1.E+8 Original connection New connection Phase (Degree) Insertion gain of DM TL EMI filter 1.E+3 1.E+4 1.E+5 1.E+6 1.E+7 1.E+8 Original connection New connection Fig Measured insertion gains of DM transmission-line EMI filter under different connections. 168

196 7.3. DM transmission-line EMI filter without insulation layer The parametric study in Chapter 4 and the experiment results in Chapter 6 both show that the thickness and material properties of the insulation layer have no influence on the performance of the DM transmission-line EMI filter. Actually, the insulation layer can be removed from the structure. The copper conductor can be directly deposited onto the nickel conductor. The performance of the filter stays the same. One DM transmission-line EMI filter prototype is built without the insulation layer, and the insertion gain curve is measured. As shown in Fig. 7-5, the insertion gain of the DM filter without the insulation layer is the same as the insertion gain of the DM filters with 125µm or 25µm insulation layers. Getting rid of the insulation layer can greatly reduce the thickness of the DM transmission-line EMI filter without affecting its attenuation. Insertion Gain 25 Insertion Gain Gain (db) -6-8 Phase (Degree) E+3 1.E+4 1.E+5 1.E+6 1.E+7 1.E+8 no insulation tinsu=125um tinsu=25um E+3 1.E+4 1.E+5 1.E+6 1.E+7 1.E+8 no insulation tinsu=125um tinsu=25um Fig Measured insertion gains of DM transmission-line EMI filters with/without insulation layer DM transmission-line EMI filter with meander structure As shown in the parametric study, a longer structure has a lower cut-off frequency and more attenuation. The meander structure is attempted to increase the total length of 169

197 the filter without taking up too much space. Fig. 7-6 shows the DM transmission-line EMI filter prototype with a meander structure and the original long-strip structure. The equivalent total length of the meander structure is 26cm. The measured insertion gain is shown in Fig. 7-7 together with the insertion gain of the original long-strip structure, whose length is 1cm. The meander structure has a much lower cut-off frequency and more attenuation than the original long-strip structure. The insertion gain curve of the meander structure doesn t continue going down above 3MHz because it has already hit the noise floor of the network analyzer. Fig DM transmission-line EMI filter with meander and long-strip structures. Gain (db) Insertion Gain E+3 1.E+4 1.E+5 1.E+6 1.E+7 1.E+8 Meander Long-strip Phase (Degree) Insertion Gain 1.E+3 1.E+4 1.E+5 1.E+6 1.E+7 1.E+8 Meander Long-strip Fig Measured insertion gains of DM transmission-line EMI filters with meander and long-strip structures. 17

198 7.5. DM transmission-line EMI filter with multi-dielectric-layer structure As discussed in previous chapters, the cut-off frequency f c1 is determined by the capacitance of the dielectric layer and the load impedance. In order to lower the cut-off frequency, the capacitance of the structure should be increased. One way to increase the capacitance is to use a multi-dielectric-layer structure. Fig. 7-8 shows the cross-sectional view of a DM transmission-line EMI filter with a two-dielectric-layer structure. Fig. 7-9 shows the connection of the two-dielectric-layer transmission-line EMI filter in a circuit, while Fig. 7-1 shows the measured insertion gain of the two-dielectric-layer transmission-line filter with the insertion gain of the original one-dielectric-layer structure. The f c1 of the two-dielectric-layer structure is lower than the f c1 of the insertion gain of the one-dielectric-layer structure. The attenuation is also increased in most of the frequency range, but not above 4MHz, where the attenuations of these two structures are about the same. Applying the multi-dielectric-layer structure is an effective way to lower the cut-off frequency and increase the attenuation of the filter. 65um 15um 65um 15um 65um Insulation dielectric ceramic Insulation dielectric ceramic Insulation Cu=25um Ni =17um Ni =17um Ni =17um Ni=17um Cu=25um Fig Cross-sectional view of two-dielectric-layer DM transmission-line EMI filter. 171

199 Fig Connection of two-dielectric-layer DM transmission-line EMI filters. Gain (db) Insertion gains of DM TL EMI filter E+3 1.E+4 1.E+5 1.E+6 1.E+7 1.E+8 one-dielectric-layer Two-dielectric-layer Phase (Degree) Insertion Gain of DM TL EMI filter 1.E+3 1.E+4 1.E+5 1.E+6 1.E+7 1.E+8 One-dielectric-layer Two-dielectric-layer Fig Measured insertion gains of one-dielectric-layer and two-dielectric layer DM transmission-line EMI filters A Broadband DM EMI filter combining reflective and absorptive effects As discussed in Chapter 1, discrete reflective LC EMI filters with lumped inductors and capacitors are usually used to attenuate the noise in power electronics systems. However, the attenuation of these filters at high frequencies is greatly reduced due to the high-frequency parasitics within the filter structure. Fig shows a discrete one-stage LC EMI filter [23]. Fig shows the equivalent circuit of this discrete one-stage LC EMI filter. This filter has one common-mode choke. The leakage inductance of the common-mode choke is used as the DM inductance. C x1 and C x2 are the two DM capacitors, and C y is the common-model capacitor. Fig shows the DM equivalent 172

200 circuit, while Fig shows the DM insertion gain of the discrete reflective LC EMI filter. From point f 1, instead of continuing to go down, the insertion gain curve flattens out, and even goes up at f 2. The high-frequency attenuation is impaired by the parasitics of the filter, such as the ESL of the DM capacitors. Fig Discrete one-stage LC EMI filter [23]. L CM L DM C y C x1 C x2 C y L DM Fig Equivalent circuit of the discrete one-stage LC EMI filter. 2L DM C x1 C x2 Cy /2 Fig Equivalent circuit of the discrete one-stage DM LC EMI filter. 173

201 Fig DM Insertion gain of the discrete reflective LC EMI filter [23]. The DM absorptive transmission-line EMI filter doesn t have the high-frequency parasitic problem. Fig shows the insertion gain curve of one DM absorptive transmission-line EMI filter. The curve continues to go down at high frequencies. However, the roll-off slope of the insertion gain curve between f c1 and f c2 is only - 2dB/dec. The slope is too small for real applications. A broadband EMI filter that has a large roll-off slope as well as large high-frequency attenuation is needed. Insertion gain of DM TL EMI filter Gain (db) f c1 1.E+3 1.E+4 1.E+5 1.E+6 1.E+7 1.E+8 Measurement Multi-conductor TL model f c2 Fig Insertion gain of one DM absorptive transmission-line EMI filter. 174

202 A filter that can combine the advantages of the discrete reflective LC EMI filter and the DM absorptive transmission-line EMI filter is proposed here. Fig shows the structure of the proposed broadband DM EMI filter. Two DM absorptive transmissionline EMI filters take place of the two DM capacitors in the discrete reflective LC EMI filter. Fig shows a prototype of the broadband DM EMI filter that has been built in the packaging lab of CPES. This prototype uses the same CM choke as that used in the discrete LC EMI filter prototype. The leakage inductance of this CM choke is used as the DM inductance. Two DM absorptive transmission-line EMI filter structures are connected to the CM choke. The absorptive DM transmission-line EMI filter was designed to have the same capacitance as the DM capacitors of the discrete reflective LC EMI filter. Fig Structure of the broadband EMI filter. Fig Prototype of the broadband DM EMI filter. TABLE 7-1 summarizes the design parameters of the DM absorptive transmission-line EMI filter. Y5V is used as the dielectric material in the structure. The thickness of the Y5V dielectric sheet is 15µm. Kapton is used as the insulation material between the 175

203 copper and nickel. Its relative permittivity and loss factor are 3.5 and.2% respectively [59], and its thickness is 5µm. Fig shows the measured insertion gains of the broadband DM EMI filter prototype. The insertion gain of the discrete reflective LC EMI filter is drawn in the same graph for comparison. The insertion gain curve of the broadband DM EMI filter is very close to the insertion gain curve of the discrete reflective LC EMI filter in the low-frequency range. The insertion gain curve of the broadband DM EMI filter doesn t continue going down in the MHz range because it has already hit the noise floor of the network analyzer. Fig also shows that the broadband DM EMI filter keeps high attenuation to at least 1MHz. The difference between the attenuations of the broadband DM absorptive transmission-line EMI filter and the discrete reflective LC EMI filter is around 6dB at 3MHz. TABLE 7-1 PARAMETERS OF DM ABSORPTIVE TRANSMISSION-LINE EMI FILTER Overall dimension Nickel Dielectric Insulation PARAMETERS VALUE Length (cm) 8 Width (mm) 1 Copper thickness (µm) 5 Nickel thickness (µm) 25 Dielectric thickness (µm) 15 Insulation thickness (µm) 5 Conductivity (Ω -1 /m) 1.45e7 Relative permeability 27 Relative permittivity 14 Loss factor 4% Relative permittivity 3.5 Loss factor.2% 176

204 Insertion gain -2 db dB E+3 1.E+4 1.E+5 1.E+6 1.E+7 1.E+8 Broadband DM EMI filter Discrete LC EMI filter Fig Measured insertion gains of the broadband DM EMI filter and discrete reflective LC EMI filter. The multi-conductor lossy transmission-line model can be used to model the broadband EMI filter too. The broadband EMI filter can be looked as three two-port networks in cascade, as shown in Fig As mentioned in Chapter 2, the transfer matrix or ABCD matrix is very convenient for cascaded structures. The transfer matrix of the DM transmission-line EMI filter is: T A = COSH ( Γl) YwSINH ( Γl) - SINH ( Γl) w Y COSH ( Γl) w w (7-11) T A is an 8 8 matrix. Because the top copper and nickel conductors are connected, so are the bottom copper and nickel conductors. The transfer matrix can be transformed to a 4 4 matrix: 177

205 = A A T T (7-12) The transfer matrix of the two DM inductors in the middle is also a 4 4 matrix. The voltages and currents of this two-port network are defined in Fig The relationship between the voltages and currents is: = I I V V I I V V B T (7-13) = DM DM L j L j ω ω B T (7-14) Fig Broadband EMI filter - three two-port networks in cascade. V 12 L DM L DM V 11 V 22 V 21 I 11 I 12 I 21 I 22 Fig Two-port network of two DM inductors. The transfer matrix of the broadband DM EMI filter is

206 A B T = T T T = (7-15) A B A C D The input impedance matrix at z = can be easily derived by using the T matrix. V() = I I () (7-16) Where I is the line input impedance: I 1 = (A C) ( D B) (7-17) L L Where L is the load impedance matrix. Fig shows the simulated insertion gain of the broadband DM EMI filter from the multi-conductor lossy transmission-line model together with the measured results. The insertion gain curve from the model shows that the attenuation of the broadband DM EMI filter continues going down to at least 1MHz. The attenuation at 1MHz is around 2dB. 5 Insertion gain db dB E+3 1.E+4 1.E+5 1.E+6 1.E+7 1.E+8 Broadband DM EMI filter (Measurement) Discrete LC EMI filter (Measurement) Broadband DM EMI filter (Model) Fig Simulated insertion gain of the broadband DM EMI filter prototypes. 179

207 The above measured and simulated insertion gain curves in Fig and Fig are under the condition that the source and load impedances are both 5Ω. The dependence of the insertion gain of the broadband DM EMI filter on source and load impedances is discussed below. The insertion gain of the broadband DM EMI filter is compared with the insertion gain of the discrete reflective LC EMI filter under different source and load impedances. The insertion gain curves of the discrete reflective LC EMI filter are obtained from the Saber simulation tool. Fig shows the schematic of the discrete LC EMI filter in the Saber simulation. The broadband DM EMI filter has more attenuation than the discrete reflective LC EMI filter not only for 5Ω source and load, but also under other source and load impedances. LDM1 ESL1 ESL2 RS2 RS1 RL2 + + ESR1 CDM1 ESR2 CDM2 RL1 LDM2 Fig Schematic of discrete reflective LC EMI filter in Saber simulation. Fig shows the insertion gains of the broadband DM EMI filter at different resistive loads. The source impedance stays at 5Ω for all the curves. The resistive load changes from 1kΩ to 1mΩ. It is clearly shown that the filter has more attenuation for a larger resistive load. Fig shows the insertion gains of the discrete reflective LC EMI filter under different resistive loads. For all the resistive load values, the attenuation of the broadband DM EMI filter is higher than the attenuation of the discrete reflective LC EMI filter in the high frequency range. 18

208 The insertion gain curve stays the same when the source and load impedances switch for both the broadband DM EMI filter and the discrete reflective LC EMI filter. Fig and Fig also show the insertion gains at different resistive source impedances when the load is 5Ω. The filters have more attenuation for larger resistive source impedance. 5 Insertion Gain: Gain 2 Insertion Gain: Phase 15 1 Magnitude (db) L =5Ohm L =1 3 Ohm L =1Ohm L =1Ohm L =.1Ohm L =1mOhm Phase (d) Fig Insertion gains of the broadband DM EMI filter vs. resistive load impedance. Fig Insertion gains of the discrete reflective LC EMI filter vs. resistive load impedance. Fig shows the insertion gains of the broadband DM EMI filter at different inductive load impedances. The insertion gain curve of the 5Ω load is drawn in the same 181

209 graph for comparison. The source impedance stays at 5Ω for all the curves. The inductive load changes from 1nH to 1mH. The broadband DM EMI filter has more attenuation for a larger inductive load. 5 Insertion Gain: Gain 2 Insertion Gain: Phase 15 1 Magnitude (db) L =5Ohm L =1mH L =1uH L =1uH L =1nH L =1nH Phase (d) Fig Insertion gains of the broadband DM EMI filter vs. inductive load impedance. Fig shows the insertion gains of the discrete reflective LC EMI filter under different inductive loads. For all the inductive load values, the broadband DM EMI filter has more attenuation in the high-frequency range than the discrete reflective LC EMI filter. The same conclusion can be drawn for the inductive source impedances. Fig Insertion gains of the discrete reflective LC EMI filter vs. inductive load impedance. 182

210 Fig shows the insertion gains of the broadband DM EMI filter at different capacitive loads. The insertion gain curve of a 5Ω load is drawn in the same graph for comparison. The source impedance stays at 5Ω for all the curves. The capacitive load changes from 1pF to 1mF. The broadband DM EMI filter has more attenuation for a load with a smaller capacitance value. 5 Insertion Gain: Gain 2 Insertion Gain: Phase 15 1 Magnitude (db) L =5Ohm L =1mF L =1uF L =1nF L =1nF L =1pF Phase (d) Fig Insertion gains of the broadband DM EMI filter vs. capacitive load impedance. Fig shows the insertion gains of the discrete reflective LC EMI filter under different capacitive loads. For all the capacitive load values, the broadband DM EMI filter has more attenuation than the discrete reflective LC EMI filter in the highfrequency range. The same conclusion can be drawn for the capacitive source impedances. 183

211 Fig Insertion gains of the discrete reflective LC EMI filter vs. capacitive load impedance Summary In this chapter, disconnecting the load from the nickel conductors of the DM transmission-line EMI filter was investigated. The measured insertion gain curve showed much less high-frequency attenuation than the original connection method. Eliminating the insulation layer could greatly reduce the thickness of the DM transmission-line EMI filter without changing its performance. A meander structure for the DM transmission-line EMI filter could increase the total length of the filter without taking up too much space. The cut-off frequency was shifted lower and greater attenuation was achieved. A prototype of the two-dielectric-layer filter structure was built and measured. The measurement result confirmed that multi-dielectric-layer structure is an effective way to achieve lower cut-off frequency and more attenuation. A broadband DM EMI filter combining the advantages of the discrete reflective LC EMI filter and the transmission-line EMI filter was proposed. Two DM absorptive transmission-line EMI filters took the place of the two DM capacitors in the discrete 184

212 reflective LC EMI filter. A prototype of the broadband DM EMI filter was built, and found to have a measured insertion gain with a large roll-off slope at low frequencies and large attenuation at high frequencies. The dependence of the broadband DM EMI filter on source and load impedances was also investigated in this chapter. A larger load (source) impedance gave more attenuation regardless of whether it was resistive, inductive or capacitive. The broadband DM EMI filter always had more high-frequency attenuation than the discrete reflective LC EMI filter under different load (source) impedances. 185

213 Chapter 8: CONCLUSION AND FUTURE WORK 8.1. Introduction This dissertation focused on exploring the working mechanisms and providing design guidelines for the DM transmission-line EMI filter. It consisted of four major sections. The first section discussed the working mechanisms and parametric dependence of the high-frequency attenuator based on the multi-conductor lossy transmission-line model and finite element analysis, followed by experimental verification. The second section discussed the working mechanisms and parametric dependence of the DM transmissionline EMI filter. These results were also verified by experimental results. The measurement setups and the measurement board designs were discussed in the third section. Some improved structures of the DM transmission-line EMI filter were discussed in the fourth section. A broadband DMl EMI filter combining the advantages of discrete LC EMI filter and transmission-line EMI filter was proposed. The experimental results were also shown in this section. The conclusions drawn for each of these sections are summarized below High-frequency attenuator The high-frequency attenuator was modeled as a simple capacitor in the lowfrequency range. In a higher-frequency range, the multi-conductor transmission line model had to be used to analyze the performance of the high-frequency attenuator. The following conclusions were drawn from the parametric study of the high- frequency attenuator: A longer structure gave lower f c1, lower f c2 and more attenuation; 186

214 A wider structure gave lower f c1 and more attenuation; A long and narrow structure gave the same f c1, but lower f c2 and more attenuation than the short and wide structure with the same conductor area; Under the nominal relative permeability and conductivity of nickel, nickel thickness larger than 5µm provided the same attenuation. Nickel thickness smaller than 5µm gave more attenuation; Thinner dielectric material gave lower f c1, lower f c2 and more attenuation; The higher relative permeability of nickel gave lower f c2 and larger roll-off slope; The lower conductivity of nickel gave more attenuation; The higher relative permittivity of the dielectric material gave lower f c1, lower f c2 and more attenuation; The loss factor of the dielectric material has little influence on attenuation; Larger load (source) impedance gave more attenuation. For a very small load (source) impedance, for example 1-6 Ω, there was still some attenuation coming from the loss in the nickel conductors DM transmission-line EMI filter The DM transmission-line EMI filter was modeled as a simple capacitor in the lowfrequency range. In a higher-frequency range, the multi-conductor transmission line model had to be used to analyzer the performance of the DM transmission-line EMI filter. The following conclusions were drawn from the parametric study of the DM transmission-line EMI filter: 187

215 A longer structure gave lower f c1, lower f c2 and more attenuation; A wider structure had lower f c1 and more attenuation; A long and narrow structure had a smaller first order bandwidth (slope = -2dB/dec part) than short and wide structure of same conductor area; A thinner dielectric material gave lower f c1, lower f c2 and more attenuation; The higher relative permeability of nickel gave lower f c2, larger roll-off slope and more attenuation; The conductivity of nickel had an optimum value range: 5.e6S/m ~ 5.e5S/m. This was different from the high-frequency attenuator. The lower conductivity of nickel gave more attenuation for the high-frequency attenuator. The higher relative permittivity of the dielectric material gave lower f c1, lower f c2 and more attenuation; The thicknesses of nickel layer, copper layer and insulation layer had little influence on filter attenuation; The loss factor of the dielectric material, relative permittivity and loss factor of insulation material had little influence on filter attenuation; A larger load (source) impedance gave more attenuation. For a very small load (source) impedance, such as 1-6 Ω, there was still some attenuation coming from the loss in the nickel conductors. 188

216 8.4. Measurement setup The measurement setups of the DM transmission-line EMI filter using a network analyzer were discussed. The network analyzer had a common-ground problem that influenced the measured results of the high-frequency attenuator. However, the commonground problem had a negligible influence on the measured results of the DM transmission-line EMI filter. To compare the parametric study results with the measured results, the parametric study was performed again for the high-frequency attenuator, taking into account the common-ground problem. The new insertion gain curves showed less high-frequency attenuation and a more pronounced transmission-line effect than the original insertion gain curves. However, the conclusions from the parametric study on the high-frequency attenuator still applied. Both the simulated and measured results showed that the transfer gain was very sensitive to parasitic inductance, introduced by the connectors and copper strips between the connectors and the filter. However the insertion gain curve was not sensitive to this parasitic inductance. There were two major methods to reducing the parasitic inductance of the measurement setup: using small connectors and applying a four-terminal measurement setup Improved structures of the DM transmission-line EMI filter Different connection methods and improved structures for the DM transmission-line EMI filter were investigated. The measured insertion gain curve shows that disconnecting the load from the nickel conductors of the DM transmission-line EMI filter impairs the high-frequency attenuation of the filter. Eliminating the insulation layer greatly reduced 189

217 the thickness of the DM transmission-line EMI filter without changing its performance. The meander structure of the DM transmission-line EMI filter increased the total length of the filter without taking up too much space, and resulted in the cut-off frequency being shifted lower and achieving more attenuation. A prototype of the two-dielectric-layer filter structure was built and measured. The measurement results confirmed that a multidielectric-layer structure was an effective way to achieve a lower cut-off frequency and more attenuation. This dissertation proposed a broadband DM EMI filter combining the advantages of the discrete reflective LC EMI filter and the transmission-line EMI filter. Two DM absorptive transmission-line EMI filters take the place of the two DM capacitors in the discrete reflective LC EMI filter. The measured insertion gain had a large roll-off slope at low frequencies and large attenuation at high frequencies. The broadband DM EMI filter always had more high-frequency attenuation than the discrete reflective LC EMI filter under different load (source) impedances Future work The following subjects have not been studied in this dissertation and are suggested to be the topics of future work: Large signal characterization of broadband DM EMI filter; Integrated transmission-line EMI filter with other low-conductivity conductor than nickel; Design methodology of the transmission line EMI filter under different source and load impedances; 19

218 3D simulation model; Thermo-mechanical stress and intermetallic diffusion if Cu directly on Ni; Effect of ambient changes on filter s performance, such as bias voltage, temperature; Stat-of-the-art processes for the transmission line EMI filter 191

219 REFERENCES [1] van Wyk, J.D., Lee, F.C., Boroyevich D., Power Electronics Technology: Present Trends and Future Developments, in Scanning the Issue, Proceedings of the IEEE, Vol. 89, Issue 6, June 21, pp [2] J. Popovic, J.A. Ferreira, "An approach to deal with packaging in power electronics", IEEE Trans. Power Electronics, Volume 2, Issue 3, May 25, Page(s): [3] Popovic, J., Ferreira, J.A., Concepts for high packaging and integration efficiency, Power Electronics Specialists Conference, IEEE 35th Annual, Volume 6, 2-25 June, 24, Page(s): [4] Van Wyk, J.D., Lee, F.C., Power electronics technology at the dawn of the new millenium-status and future Power Electronics Specialists Conference, 3th Annual IEEE, Volume 1, 27 June-1 July, 1999, Page(s):3 12. [5] van Wyk, J.D., Lee, F.C., henxian Liang, Rengang Chen, Shuo Wang; Bing Lu, Integrating active, passive and EMI-filter functions in power electronics systems:a case study of some technologies Power Electronics, IEEE Transactions on, volume 2, Issue 3, May 25, Page(s): [6] Lee, F.C., van Wyk, J.D., Boroyevich, D., Guo-Quan Lu, henxian Liang, Barbosa, P., Technology trends toward a system-in-a-module in power electronics, Circuits and Systems Magazine, IEEE, Volume 2, Issue 4, Fourth Quarter 22, Page(s):4 22. [7] Lee, F.C., van Wyk, J.D., Liang,.X., Chen, R., Wang, S., Lu, B., An integrated power electronics modular approach: concept and implementation IPEMC 24. Volume 1, Page(s):1 13. [8] Strydom, J.T., van Wyk, J.D., de Rooij, M.A., Integration of a 1 MHz converter with active and passive stages, APEC 21. Volume 2, Page(s): [9] J. Popovic, J.A. Ferreira, "An approach to deal with packaging in power electronics", IEEE Trans. Power Electronics, Volume 2, Issue 3, May 25, Page(s): [1] J. Popovic, J.A. Ferreira, "Converter concepts to increase the integration level", IEEE Trans. Power Electronics, Volume 2, Issue 3, May 25, Page(s):

220 [11] Popovic, J., Ferreira, J.A., Design and evaluation of highly integrated dc-dc converters for automotive applications IAS 25, Volume 2, Page(s): [12] Popovic, J., Ferreira, J.A., Waffenschidt, E., PCB embedded DC/DC 42V/14V converter for automotive applications, Power Electronics and Applications, 25 European Conference on, Page(s):1 [13] Popovic, J., Ferreira, J.A., Gerber, M.B., Konig, A., de Doncker, R., Integration technologies for high power density power electronic converters for AC drives, Power Electronics, Electrical Drives, Automation and Motion, 26. International Symposium on, Page(s): [14] Gerber, M., Ferreira, J.A., A System Integration Philosophy for Demanding Requirements in Power Electronics, Industry Applications Conference, 27, Page(s): [15] Jelena Popovic-Gerber, Mark Gerber and Braham Ferreira, New system integration concept for high power density drives PESC 28. [16] Henry, W. Ott, Noise reduction techniques in electronic systems, 2nd edition, New York : Wiley, c1988; [17] Liang,.X.; Lee, F.C.; van Wyk, J.D.; Boroyevich, D.; Scott, E.; Chen, J.; Lu, B.; Pang, Y., Integrated packaging of a 1 kw switching module using planar interconnect on embedded power chips technology, APEC '3, vol. 1, pp [18] van Wyk J. D., Fred C. Lee, Dushan Boroyevich, henxian Liang, IEEE and Kaiwei Yao, A Future Approach to Integration in Power Electronics Systems, proc. of IEEE IECON 23; [19] Mark J. Nave, Power line filter design for switched-mode power supply, van Nostrand Reinhold, New York, [2] Shuo Wang, Lee FC, Odendaal WG., Improving the performance of boost PFC EMI filters, APEC 23, IEEE. Part vol.1, 23, pp [21] Shuo Wang. Lee FC. Chen DY. Odendaal WG, Effects of parasitic parameters on EMI filter performance, PESC 23, IEEE. Part vol.1, 23, pp

221 [22] Shuo Wang, F. C. Lee and W. G. Odendaal, Controlling the parasitic parameters to improve EMI filter performance, in Proc. IEEE APEC, Anaheim, California 22-26, Feb. 24, Volume 1, pp [23] Rengang Chen, Shuo Wang, van Wyk, J.D., Odendaal, W.G., Integration of EMI filter for distributed power system (DPS) front-end converter, PESC 23, Volume 1, Page(s): [24] Rengang Chen, van Wyk, J.D., Wang, S., Odendaal, W.G., Planar electromagnetic integration technologies for integrated EMI filters IAS 23, Volume 3, Page(s): [25] Chen, R., van Wyk, J.D., Wang, S., Odendaal, W.G., Application of structural winding capacitance cancellation for integrated EMI filters by embedding conductive layers IAS 24, Volume 4, Page(s): [26] Rengang Chen, van Wyk, J.D., Shuo Wang, Odendaal, W.G., Improving the Characteristics of integrated EMI filters by embedded conductive Layers Power Electronics, IEEE Transactions on Volume 2, Issue 3, May 25, Page(s): [27] F. Mayer, Absorptive low-pass cables: state of the art and an outlook to the future, IEEE Trans on EMC, vol. EMC-28, No. 1, Feb [28] F. Mayer, Improvements on the MIL-STD low-pass cable line, IEEE EMC symposium 1996, proc., pp [29] F. Mayer, Electrical power and signal distribution in modern aircraft, combines weight advantages and EMC compatibility, IEEE EMC symposium, 1998, proc., pp [3] F. Mayer, F. Heather, HF lossy line IEEE EMC symposium, 1998, proc., pp [31] F. Mayer, Electromagnetic compatibility: Anti-interference wires, cables and filters, IEEE Trans on EMC, vol. EMC-8, No. 3, Sep. 1966, pp [32] H.M. Schlicke, H. Weidmann, Compatible EMI filters, IEEE Spectrum, October 1967, pp [33] H.M. Hoffart, Electromagnetic interference reduction filters, IEEE Trans on EMC, vol. EMC-1, no.2, pp , June [34] I.W. Ha, R.B. Yarbrough, A lossy element for EMC filters, IEEE Trans on EMC, vol. EMC-18, no. 4, Nov. 1976, pp

222 [35] T. Sato, S. Ikeda, K. Yamasawa, T. Mizoguchi, Transmission line low-pass filter for switching power supplies, IEEE PESC 1998 Proceedings, pp [36] J.D. van Wyk jr. Power Electronic Interconnects: Frequency Selective Electromagnetic Propagation PhD dissertation, Rand Afrikaans University [37] van Wyk, J.D., Jr., Cronje, W.A., van Wyk, J.D., Campbell, C.K., Wolmarans, P.J.; Power electronic interconnects: skin- and Proximity-Effect-based frequency Selective-Multipath propagation, IEEE Transactions on Power Electronics, vol. 2, issue 3, pp. 6-61, May 25. [38] P.J. Wolmarans, J.D. van Wyk, J.D. van Wyk Jr. and C.K. Campbell, Technology for integrated RF- EMI transmission line filters for integrated power electronics modules, Proc. of IEEE IAS 22, vol. 3, pp [39] hao, L.; Chen, R.; van Wyk, J.D., An integrated common mode and differential mode transmission line RF-EMI filter, proc. of IEEE PESC 4, Volume: 6, Pages: [4] Rengang Chen, Integrated EMI filters for Switch Mode Power Supplies PhD dissertation, Virginia Polytechnic Institute and State University [41] David M. Pozar, Microwave Engineering, 2 nd edition, New York, John Wiley & Sons, Inc [42] hao, L., Generalized frequency plane model of integrated electromagnetic power passives, PhD dissertation, dept. of Electrical Engineering, Virginia Polytechnic Institute and State University, May 24. [43] Lingyin hao; van Wyk, J.D., Electromagnetic modeling of an integrated RF EMI filter, proc. of IEEE IAS 23, Volume: 3, pp [44] J.A. Brandao Faria, Multiconductor transmission-line structures: modal analysis techniques, John Wiley & Sons, Inc., [45] Pieter Johannes Wolmarans, Investigation of a class of distributed planar conducted RF-EMI filters for integration in power electronics converters Master thesis, Jan 23, Engineering, Rand Afrikaans University, Johannesburg, South Africa [46] William H.Safranek, The properties of electrodeposited metals and alloys, American Elsevier Publishing Co., Inc.,

223 [47] J. K. Dennis and T. E. Such, Nickel and chromium plating, Butterworth & Co Ltd, [48] Y. Liang, R. Chen, and J. D. van Wyk, Large signal dielectric characterization for integrated electromagnetic power passives, in Proc. IEEE-APEC, 25. [49] Agilent 4395A Network/Spectrum/Impedance Analyzer Operation Manual [5] Agilent 87512A/B transmission/reflection test kit operation and service manual [51] [52] [53] William H.Safranek, The properties of electrodeposited metals and alloys, American Elsevier Publishing Co., Inc., [54] J. K. Dennis and T. E. Such, Nickel and chromium plating, Butterworth & Co Ltd, [55] [56] [57] [58] W. Betteridge, Nickel and its alloys, Ellis Horwood Limited, [59] Krishnamurthy, V.B.; Cole, H.S.; Sitnik-Nieters, T.; Use of BCB in high frequency MCM interconnects, Components, Packaging, and Manufacturing Technology, Part B: Advanced Packaging, IEEE Transactions on, Volume 19, Issue 1, Feb Page(s): [6] Maxwell Q2D manual, Ansoft corporation 196

224 APPENDIX I: MANUFACTURING PROCESS OF DM TRANSMISSION-LINE EMI FILTER The DM transmission-line EMI filter can be built using the facilities in the packaging lab of CPES. The major manufacturing processes include cleaning, sputtering, electroplating, photolithography, etching, laser cutting, etc. Step 1: Cleaning A 12mm X 6mm high permittivity dielectric sheet (Y5V) is used to build the DM transmission-line EMI filter. The first step is to clean the dielectric sheet using acetone, alcohol and DI water. The dielectric sheet is rinsed by these three solutions in turn. Acetone is used to remove any organics on the substrate; alcohol and DI water are used immediately after acetone to prevent the organics residue on the substrate. The sheet is then blown dry using nitrogen gas. Don t let the sheet dry naturally in the air, because that will leave residues on the dielectric sheet. Step 2: Sputtering The dielectric sheet is put into the sputtering chamber after cleaning. The sputtering process is under a pressure of 5mTorr in argon gas plasma. The chamber is first pulled to a pressure less than 2µTorr to remove impurities from the chamber. A layer of Ti is first sputtered onto the dielectric sheet as seed layer to achieve good attachment. The sputtering of Ti takes 15 minutes. Ni is sputtered after Ti. Ni is sputtered for two 15- minutes periods. The substrate is then flipped over and the same sputtering process is repeated on the other side of the dielectric sheet. Step 3: electroplating 197

225 The thickness of the sputtered metal is less than 1µm. Electroplating is used after sputtering to get thicker metal layer. Electroplating will plate metal from an anode to cathode through an ionic solution when excited by an external current. Nickel sulfamate solution from Technic Inc. is used as the electroplating solution. The solution needs to be heated to 12 o F before electroplating. The anode is a piece of bulk nickel. The cathode is the substrate with sputtered Ti and Ni. A frame is used to hold the substrate. The frame is comprised of two pieces of FR4 board with Cu on one side. The filter sample is put between the two FR4 boards. The Cu on FR4 boards contacts with the metal on the substrate. The FR4 board is connected to the cathode terminal. To plate nickel to both sides of the filter sample, two bulk nickel anodes are put on each side of the substrate. Fig. AI-4 shows the time need to electroplate nickel to different thicknesses under different current density. The nickel layer is more even for smaller current density. However, it takes longer for smaller current density to get the same nickel thickness. 1A/ft 2 is chosen for the nickel plating. The area of the dielectric sheet is 12mm 6mm=72mm 2 =.775ft 2. The current for electroplating both sides of dielectric sheet is 1A/ft 2.775ft 2 2=1.55A. The time needed to get 17µm nickel thickness is around 7 minutes. 198

226 (a) Dielectric sheet Y5V (b) after sputtering (c) after electroplating Fig. AI-1. Dielectric sheet during manufacturing process of DM transmission-line EMI filter. Fig. AI-2. FR4 frame for nickel electroplating. Fig. AI-3. Bulk Nickel as anode during electroplating. 199