Modeling and Analysis of Near-field Probes and Electromagnetic Radiation from PCB-Chassis Structure

Save this PDF as:
 WORD  PNG  TXT  JPG

Size: px
Start display at page:

Download "Modeling and Analysis of Near-field Probes and Electromagnetic Radiation from PCB-Chassis Structure"

Transcription

1 Modeling and Analysis of Near-field Probes and Electromagnetic Radiation from PCB-Chassis Structure Hiroki Funato A thesis submitted for the degree of Doctor of Philosophy at Tokyo Metropolitan University 2014

2 首都大学東京博士 ( 工学 ) 学位論文 ( 課程博士 ) 論文名 近傍電磁界プローブと PCB- 筐体構造における電磁界 放射のモデル化と理論解析 ( 英文 ) 著者 船戸裕樹 審査担当者 主査 委員 委員 委員 上記の論文を合格と判定する平成年月日首都大学東京大学院理工学研究科教授会研究科長

3 DISSERTATION FOR A DEGREE OF DOCTOR OF PHILOSOPHY IN ENGINEERING TOKYO METROPOLITAN UNIVERSITY TITLE : Modeling and Analysis of Near-field Probes and Electromagnetic Radiation from PCB-Chassis Structure AUTHOR: Hiroki Funato EXAMINED BY Examiner in chief Examiner Examiner Examiner Dean QUALIFIED BY THE GRADUATE SCHOOL OF SCIENCE AND ENGINEERING TOKYO METROPOLITAN UNIVERSITY Date

4 Abstract Electromagnetic compatibility (EMC) is essential to achieve safety and reliability in the development of electronics. However, the rapid growth of electronics has led to a variety of issues and challenges in the field of EMC. This thesis focuses on two major challenges for today s EMC: one is overcoming problems caused by the complexity of system, and the other is the changes in requirements for EMC due to the miniaturization of electronics in high frequency over GHz. To keep contributing to the growth of electronics, measurement and simulation technologies for EMC must be developed and improved. In this thesis, modeling on the basis of physical structure of various probes for near-field measurements is presented in order to optimize probe design. Measurement techniques to improve the accuracy and the spatial resolution are also proposed and verified. A calculation model for estimating electric field radiation from PCB-chassis structure, which is a typical configuration representing the complex system as stated above, is then proposed using three different approaches and discussed. This thesis consists of two main chapters. The first covers measurement technologies, and the second covers the modeling and analysis of the radiation from PCB-chassis structure. Subsection 2.1 clarifies the effect of the electric near-field on the magnetic near-field probe and provides the equivalent circuit of the probe with consideration of this effect. In subsection 2.2, the dual-loop magnetic field probe is proposed to achieve more accurate measurements in the system that has electromagnetic disturbances from other components around the probe. Subsection 2.3 describes a new double position/signal difference (DPSD) method proposed to obtain a tangential electric near-field as well as an improved position/signal difference (PSD) method for more accurate measurements. Chapter 3 describes modeling and analysis on the electric-field radiation from the PCB-chassis structure. Subsection 3.1 provides measurement results of an electric far-field and reveals the increase in the radiation caused by integrating components. In subsection 3.2, a specialized thin current probe is proposed and applied for measuring the current flowing through the screw between PCB and chassis. It is revealed that the screw current stimulates the cavity consisting of PCB with chassis. Subsection 3.3 describes cavity resonator modeling using a double summation method with the screw current as the stimulus. Subsection 3.4 describes a radiation model using an equivalent source from a parallel-plates structure with consideration of asymmetric sizes for PCB and chassis. In subsection 3.5, an inductive network method is employed to calculate the cavity for fast calculation time, and the results are verified. Subsection 3.6 describes the modeling of a 1

5 PCB-chassis structure using the multilayered finite difference method (MFDM), which can be integrated with a SPICE circuit and can be performed without measured screw current as a stimulus. Subsection 3.7 explains the model-based analysis for reducing radiation from the PCB-chassis structure. As stated, this thesis describes the modeling and the improvement techniques for near-field measurements as well as the theoretical modeling and analysis of the PCB-chassis structure. These studies contribute to realizing EMC design for miniaturized electronics at high frequency with consideration of system-level behavior. 2

6 Preface EMC has a short history as a field of study. IEEE, which is the world s largest organization dedicated to the development of technologies on electronics, established the EMC society in 1978, although it had originated from the radio frequency interference group back in The EMC came about after multiple wireless communication devices had been produced. Since then, as the technologies on electronics have been developed, the achievements of EMC have become more important. Thus, EMC is considered an essential discipline in electrical engineering, which itself is based on other fundamental sciences. Hence, during my life as a researcher, I studied electronic devices during my master s degree at university and have been engaged in EMC since joining Hitachi. Throughout my research, whenever I was about to lose direction, I always tried return to my starting point: the definition of "engineering". "Engineering" is defined as the application of scientific, economic, social, and practical knowledge to design, build, maintain, and improve structures, machines, devices, and systems 1. Based on this definition, my understanding is that engineering should create value by utilizing fundamental sciences. So, what do we mean by "value"? The meaning or definition of value depends on the viewpoint: those of industrial companies, people who use the technologies, the environment surrounding us. It may be difficult to define all the applications, but we must always try to clarify what value the technology can create and how the technologies can be used for people and their environment. I believe these questions should motivate our research. With such considerations, this thesis summarizes all the work I have done during my time at Tokyo Metropolitan University. I hope this study will contribute to creating value in the future. 1 Wikipedia: 3

7 Contents 1. Introduction Modeling and Improvement of Near-field Measurements Equivalent Circuit of Magnetic Near-Field Probe in GHz Band and Improvement of Spatial Resolution Evaluation of Sensitivity to Magnetic and Electric Near-Field Equivalent Circuit for Sensitivity to Electric Near-field Equivalent Circuit for Sensitivity to Magnetic Near-field Effective Height for Magnetic Near-field Measurement Application of Position Signal Difference Method Measurement-Based Modeling of Dual Loop Magnetic Near-field Probe Fabricated Probes for Evaluation Modeling of Probes Modeling and Improvement of Electric Near-field Measurements Introduction of Position/Signal Difference Method Probe Structure and Theory of Proposed Double Position/Signal Difference Method Measurements and Validation of PSD and DPSD Method Accurate Modeling with Fringe Capacitance Proposal and Validation of Improved PSD Method Role of Probe Displacement (d p ) for DPSD Modeling and Analysis of Radiation from PCB-Chassis Structure Radiation from PCB-Chassis Structure Proposal and Application of Screw Current Probe Structure of Screw Current Probe Measurements of Screw Current Cavity-Mode Modeling using Double Summation Method Calculation of Radiation from Parallel-Plate Structure Asymmetric Size Coefficients Fast Calculation using Inductive Network Method Modeling by Multilayered Finite Difference Method Investigation on the Reduction of Radiation Additional Bypass Capacitor at Screw Model-Based Analysis for the Reduction of Radiation

8 4. Conclusion Appendix A: Stand-alone Electric-field Probe Prototype design and function test D Simulation and calculation Electric field measurement inside enclosure Appendix B: EM Simulation of On-Glass Antenna with Vehicle Body Background of simulation Simulation model Comparison with measurement Bibliography Acknowledgements List of related publications

9 List of figures Figure 1. Schematic process of the components integration Figure 2. Magnetic field strength of mobile electronics Figure 3. Structure of probes Figure 4. Fabricated probe (left: Type A, right: Type B) Figure 5. Measurement setup Figure 6. Direction of induced current for electric field and magnetic field Figure 7. Evaluation results of induced voltage by electric near-field ( V E ) and by magnetic near-field ( V H ) for type A Figure 8. Evaluation results of V H and V E for type B Figure 9. Probe height dependence of induced voltage by electric near-field ( V E ) Figure 10. Schematic diagram of equivalent circuit for electric near-field (type A) Figure 11. Schematic diagram of equivalent circuit for electric near-field (type B) Figure 12. Calculation and measurement results of V E for type A Figure 13. Calculation and measurement result of V E for type B Figure 14. equivalent circuit of probe for magnetic near-field (type A) Figure 15. equivalent circuit of probe for magnetic near-field (type B) Figure 16. height dependence for magnetic near-field inside loop Figure 17. Calculation result of effective measurement height (h eff ) with loop center height (h center ) Figure 18. Calculation and measurement result of induced voltage by magnetic near-field ( V H ) for type A Figure 19. Calculation and measurement result of induced voltage by magnetic near-field ( V H ) for type B Figure 20. Schematic diagram of spatial resolution improvement method Figure 21. Calculation results of induced voltage ratio for S 1 to S Figure 22. Comparison between calculation and measurement using type B with/without applying subtraction Figure 23. Schematic of problem in actual application Figure 24. Schematic structures of two dual loop magnetic near-field probes Figure 25. Magnetic near-field and far-field coupling to dual loop probes Figure 26. Measurement setup for probe evaluation Figure 27. Measurement results of S 21 for probes A and B Figure 28. Equivalent circuit of dual loop magnetic near-field probe

10 Figure 29. Schematic of magnetic flux paths during S 21 and S 22 measurements Figure 30. simulation model of a dual loop magnetic near-field probe Figure 31. Simulation results for magnetic near-field distribution during S 21 and S 22 measurements Figure 32. Frequency dependence of input impedance for probe B Figure 33. Comparison of calculation and measurement of S 21 for probe B Figure 34. Comparison of calculation and measurement of S 21 for probe A Figure 35. Equivalent circuit for probe A that considers capacitive coupling between the probe and the microstrip line Figure 36. Frequency dependence of input impedance for probe A Figure 37. Comparison of calculation and measurement results for S 21 for probe A when the electric near-field is considered Figure 38. Configuration of calculation for different probe type Figure 39. Comparison of calculated scanned results between probes Figure 40. Schematic structure of printed monopole electric near-field probe Figure 41. Schematic of position/signal difference method Figure 42. Capacitive couplings to the probe in position/signal difference method Figure 43. Setup for electric near-field measurement Figure 44. Normalized magnitude of induced voltage at h = 0 and 0.35 mm Figure 45. PSD and DPSD results Figure 46. HFSS simulation model of microstrip line for electric near-field analysis.. Figure 47. Measured and simulated magnitude of electric near-field component... Figure 48. PSD applied result and simulated normal component of electric near-field51 Figure 49. DPSD applied result and simulated tangential component of electric near-field 51 Figure. FSV applied results for simulated and PSD/DPSD Figure 51. PSD applied results with different probe heights comparing to simulated normal component of electric field Figure 52. Probe and setup for PSD/DPSD measurements Figure 53. PSD measurement results with changing shift parameter d p at 1 GHz Figure 54. Fringe capacitance model between probe element and trace Figure 55. Calculated FWHM wfw compared to measurements with various dp Figure 56. Procedure of proposed method Figure 57. (a) Definitions of parameters and (b) Comparison between conventional PSD and proposed method Figure 58. DPSD measurement results at different dp comparing to Ex obtained by 7

11 electromagnetic simulation. (a) d p = 0.1 mm, E x at z = 0.45 mm (b) d p = 0.4 mm, E x at z = 0.75 mm (c) d p = 0.7 mm, E x at z = 1.05 mm Figure 59. (a) Configuration for measurements and simulation (b) Comparison of peak-to-peak distance d t between measurements at various d p and simulation results 59 Figure inch hard disk drive chassis and PCB for evaluation Figure 61. Schematic circuit diagram for evaluation board Figure 62. Pattern layouts of evaluation board Figure 63. Setup for electric far-field measurements Figure 64. Measured vertical electric far-fields for four different screw connections, PCB without a chassis, and PCB mounted on a chassis but with no screw Figure 65. Measured horizontal electric far-fields for four different screw locations, PCB without a chassis, and PCB mounted on a chassis but with no screw Figure 66. Difference in measured electric far-fields between the PCB only and the PCB mounted on a chassis for vertical (left) and horizontal (right) electric far-fields.. 65 Figure 67. Schematic diagram of screw current measurement Figure 68. Fabricated screw current probe Figure 69. Frequency characteristics of induced power for fabricated junction current probe by 4.48 ma current Figure 70. Measurement results of screw current for each screw under operating Figure 71. Magnetic near-field distribution on PCB with connection for each screw Figure 72. Calculation model for the radiation (a) Schematic diagram of the parallel plane cavity resonator (b) Equivalent magnetic current loop Figure 73. Electromagnetic simulation model used to calculate the asymmetric parallel plane coefficient Figure 74. Calculated vertical and horizontal electric far-fields at a distance of 3 m for different chassis sizes Figure 75. Asymmetric size coefficient for horizontal electric far-field with different screw locations Figure 76. Comparison of calculated and measured vertical electric far-fields for four different screw locations Figure 77. Comparison of calculated and measured horizontal electric far-fields for four different screw locations Figure 78. Comparison of calculated and measured vertical electric far-fields for four different screw locations Figure 79. Comparison of calculated and measured horizontal electric far-fields for four different screw locations

12 Figure 80. Multilayered equivalent circuit of PCB with a chassis Figure 81. Comparison of calculated and measured frequency spectra of screw current for four different locations of the screws connecting the PCB and the chassis Figure 82. Comparison of calculated and measured vertical electric far-fields for four different screw locations Figure 83. Comparison of calculated and measured horizontal electric far-fields for four different screw locations Figure 84. Additional bypass capacitor located close to the screw for experiment Figure 85. Frequency spectra of electric far-field with/without additional bypass capacitor. 85 Figure 86. Improvement of radiated emission with additional capacitors Figure 87. Measurement results of screw current for screw 1 and 4 with connections of both screw 1 and 4 with/without 1000 pf capacitors Figure 88. Calculated impedance distribution with no screw connection Figure 89. Calculated screw current corresponding to the impedance at each node Figure 90. Calculated impedance distribution with single screw connection Figure 91. Total screw current at 0.3 GHz with different screw connections Figure 92. Calculated horizontal electric far-field with different screw connections Figure 93. Measurement results of radiated electric field Figure 94. Reduction of the calculated screw(s) current and the measured radiation. 92 Figure 95. Schematic circuit diagram of prototype probe Figure 96. Pattern layout design of prototype printed circuit board Figure 97. Prototype stand-alone electric-field probe Figure 98. Setup for testing function of prototype device Figure 99. Converted received power to the electric field for various frequencies Figure 100. Directivity of prototype probe in different planes Figure D simulation model created in CST Microwave Studio Figure 102. Comparison of the simulated directivity (normalized with the maximum electric field strength) with the measurement results Figure 103. Comparison between calculated received power as a function of electric field strength and measurement results Figure 104. Schematic of experiment setup for the measurement of electric field inside enclosure with small slit Figure 105. Comparison of experimental electric field measurements inside an enclosure by prototype and conventional probes Figure 106. GPS antenna simulation model

13 Figure 107. Left: small cut glass model. Right: windshield model Figure 108. Left: return loss of antenna. Right: VSWR of antenna Figure 109. Radiation pattern in y-z plane (Left), and in x-z plane (Right) Figure 110. Vehicle model consists of parts used in simulation Figure 111. Left: return loss of antenna. Right: VSWR of antenna Figure 112. Radiation pattern in y-z plan (Left), and in x-z plane (Right) Figure 113. Radiation pattern in y-z plan (Left), and in x-z plane (Right) Figure 114. GPS antenna location on windshield used for measurement Figure 115. Left: radiation pattern in y-z plan. Right: radiation pattern in x-z plane Figure 116. Elevation angle dependence of linear average gain for measurement result and simulation result in x-y plane

14 List of tables Table 1 Scope of this thesis Table 2 Summary of this thesis Table 3 Derived parameters Table 4 Derived Values for RLC mesh Table 5 Comparison of different approaches

15 1. Introduction 1. Introduction Electromagnetic compatibility is mandatory for all the electronics on the market in order to ensure the devices or systems function without electromagnetically interfering with each other. EMC regulations are globally standardized to validate performance before the product comes onto the market. EMC can be categorized into two parts: emission and immunity. Emission is defined as the electromagnetic noise generated by the device, and the conducted or radiated noise must be lower than the specified limit. Immunity means to check that the device/system functions even it is exposed to the various sorts of electromagnetic disturbances. The work in this thesis is mainly related to the emission of EMC, although the improvements of and technologies for the emission are generally also directly applicable to the immunity. The challenges of EMC in developments of electronics are to reduce countermeasures and realize EMC design. In the product development process, a prototype is typically fabricated before the final design is fixed in order to validate its function/performance. If the prototype fails to comply with the EMC requirements, then the root cause must be found and the design changed appropriately. However, the time needed to investigate and the huge man-hours needed to modify the design increase the product development time as well as product development cost, which could result in critical problems such as longer time-to-market. Therefore, the root-cause needs to be clarified and determined effectively and speedily when the device/system fails to meet requirements. Moreover, it is important to design the product for not only the basic function but also EMC behavior in order to achieve further optimizations. Technologies to overcome such problems can be classified into two types. One is measurement. The most important aspect to improve electronics from the viewpoint of EMC is to understand the phenomena i.e. fact-finding or clarification of mechanism, since this understanding allows effective and optimized design. Measurement is a direct method for revealing what is occurring in the device or system. There are generally three things to be clarified from the EMC viewpoint: noise source, propagation path, and radiation part. Required or appropriate methods depend on the objective of the measurements. By revealing those points, proper countermeasures and feedback to the design can be determined. In other words, those countermeasures or the design changes cannot be eliminated by only improving measurement since the measurements can be performed only after prototyping. Therefore, simulation technology is also crucial. There are two main advantages of employing simulation-based analysis. One is that the 12

16 1. Introduction simulation can increase understanding of the behavior of electromagnetic fields since the simulation is able to provide and visualize the field distribution at arbitrary locations to which even the measurements cannot be applied. The other advantage is that the simulation can provide the estimation results of EMC performance by using only design information before prototyping, which can reduce the risk of failure after the fabrication of the sample. Also, even when the countermeasures are required, the simulation can effectively be applied without modifying the actual sample to verify the ideas by sweeping design parameters. However, such simulations always require the model to be validated. Therefore, it is concluded that both measurement and simulation are essential technologies and necessary for EMC development of electronics. A number of studies have been done on measurement and simulation modeling so far, especially for integrated circuits (IC) and printed circuit boards (PCB). For instance, a lot of research has been done on electromagnetic near-field measurements around IC or PCB. Modeling and simulation have also been investigated mainly for those components in order to clarify the conducted path as well as the electromagnetic radiation from the components. However, in recent years, even if the components such as the LSI or PCB have been designed properly at component-level, the EMC issue can sometimes be observed at system-level, which means integrated components. This is because the integration of components can change the RF behavior of components or generate an unexpected path or radiation from the system. Figure 1 shows a schematic example of the integration of components. EMC compliances or requirements described above generally define each component level, but they cannot exactly simulate each system at component-level testing. Therefore, designing components with considerations of system-level EMC behavior has become more important. Two other recent trends in electronics are obtainment of higher frequency and shrinkage of devices. Since the operation of electronics is becoming faster, noise generated by those electronics are reaching frequencies higher than the GHz band as shown in Figure 2. Therefore, to adopt such a high frequency band, even more accurate measurement and simulation must be realized and maintained by considering the high frequency characteristics. Also, the number of physically small mobile electronics such cell phones or tablets has been increasing, and the electromagnetic interferences between digital circuit and RF components have become a major issue in such high density packaging electronics [1][2]. Thus, measurement technologies must be capable of higher spatial resolution to identify finer sources or paths of electromagnetic propagations. As described above, the main challenges of today s EMC are application to system-level, accurate measurements at high frequency, and high spatial resolution for application to 13

17 1. Introduction small integrated devices. This thesis covers both approaches: measurementt and model-based calculation for these challenges, as summarized in Table 1. Note that calculation is used instead of simulation to avoid misunderstanding in the table. The PCB-chassis structure is the typical configuration of the integration ass a system and is used for the investigationn in this thesis. The modeling and calculation techniques of electric-field radiation forr just PCB or just the chassis such as near-field to far-field transformationn have been widely studied by other research groups and soo are not included in this thesis, as shown in Table 1. Figure 1. Schematic process of the componentss integration 14

18 1. Introduction Mgnetic field [db A/m] Lower power & Higher frequency Digital TV Wireless LAN Bluetooth Cell phone GPS UWB Frequency [GHz] High density packaging Calculations are based on: 1. Half-wave dipole antenna 2. Plane wave ( o=120p) 3. General spec of measurement (Spectrum Analyzer 8561EC LN1000A SMA cable) 4. Signal to Noise Ratio > 10dB Figure 2. Magnetic field strength of mobile electronics Table 1 Scope of this thesis Target PCB Chassis Enclosure PCB + Chassis Measurement (chapter 2) Higher accuracy, higher spatial resolution Clarification of source and its contribution to the radiation Calculation (chapter 3) Near-Far transformation Modeling of radiation 15

19 2. Modeling and Improvement of Near-field Measurements 2. Modeling and Improvement of Near-field Measurements Electric and magnetic near-field measurements are crucial in the field of EMC. They are generally utilized for identifying radiation sources or noise propagation paths [3]- [5]. Also a number of studies for calculating and estimating electric field radiation from electrical devices from near-field distribution has been done so far [6]-[9]. Magnetic near-field measurements using a scanning system were firstly reported by Whiteside and R.W.P. King in 1964 [10]. Since then, as technologies on electronics grow, requirements for measurement technologies have also been changing as describe in chapter 1. Following sections in this chapter describe various modeling and measurement techniques for near-field, which are all based on the requirements for such growth of electronics. Table 2 describes the scope of this thesis with problems focused on in the studies. This chapter covers the three different measurement problems: increase of the measurement error in high frequency due to the electric near-field for magnetic near-field measurement in section 2.1, the error due to the disturbance from distant source in section 2.2 and the quantitative measurement condition for electric near-field measurement in section 2.3 respectively with the coverage of PCB as well as chassis as the target toward the realizations of higher accuracy and higher spatial resolution in high frequency band for near-field measurements. 16

20 2. Modeling and Improvement of Near-field Measurements Measurement Table 2 Summary of this thesis Target PCB Chassis Enclosure PCB + Chassis Purpose Higher accuracy, higherspatial resolution Clarification of radiation behavior Field Electric field Magnetic field Current Probe /Method Problem Method Error due to Change of the gap between Error due to Unclear measurement condition electric nearfield insertion of PCB and chassis due to the disturbance Section 2.3 probe PSD Higher Objective spatial resolution Method Modeling improved PSD Higher accuracy DPSD Separation of horizontal component Capacitive coupling model PSD Higher spatial resolution Effective height Probe B dual loop Probe A Cancel of the disturbance Equivalent circuit Effective permiability Thin current probe No modification of the gap Conventional Objective Optimized measurement condition Optimized probe structure - Calculati on Method Near-Far transformation (not studied in this thesis) Cavity radiation model 2.1. Equivalent Circuit of Magnetic Near-Field Probe in GHz Band and Improvement of Spatial Resolution A number of studies have reported magnetic near-field probes for high frequencies (e.g., over 1 GHz) for various applications [11][12]. Besides these, an electro-optic probe has been proposed for frequencies above 1 GHz [13]. A miniaturized loop probe has been fabricated and used to measure current distributions in large scale integration (LSI) circuits [14][15]. For such applications, we proposed simplified lumped equivalent circuit models for the loops of magnetic near-field probes to enable the probe characteristics to be calculated [16]. A major source of error in measurements using loop probes is the voltage induced by the electric near-field since the probe has to be positioned very close to the target device or PCB. In this chapter, we first describe the fabrication of two types of magnetic near-field loop probes that have different structures of transmission line and loop element. Frequency characteristics of induced voltage due to the electric and magnetic near-field of the probes are then evaluated separately and we propose a lumped equivalent circuit to analyze the characteristic differences between the probes. Parameters of the equivalent circuit are theoretically calculated and the frequency responses are compared to the measurement 17

21 2. Modeling and Improvement of Near-field Measurements results. Since all the parameters of the equivalent circuit of the probe are calculated based on the structural information of the probe, the probe structure can be designed using the equivalent circuit with consideration of high frequency characteristics. Another important parameters for near-field measurements is, as already described, the broadband spatial resolution. Due to the continually increasing levels of integration of electrical components, higher spatial resolutions are required [17][18]. The spatial resolution in near-field measurements is basically determined by the size and structure of the probe. However, the required spatial resolution is typically not the same over the entire surface of the target device; dense regions require a higher resolution than sparse regions. Therefore, different probes need to be prepared for different requirements, which is not a very efficient approach. In addition, miniaturized probes such as those fabricated using semiconductor processes are difficult to manufacture. To overcome these difficulties, the probe/signal difference (PSD) method proposed by Kantor and Shvets for electric near-field measurement is applied to magnetic near-field measurements for the first time [19][20]. This method achieves higher spatial resolution by subtracting measurement results obtained at two different probe heights above the target surface. Near-field measurements at an arbitrary resolution can be achieved by varying the heights of the two measurements. One advantage of this method is that the probe does not need to be physically miniaturized in order to obtain a high spatial resolution. Only fine control of the probe height is required, which allows the use of a conventional loop probe during measurements, leading to greater simplicity and efficiency Evaluation of Sensitivity to Magnetic and Electric Near-Field Figure 3 shows the two types of magnetic near-field probe structures that have been fabricated in order to evaluate the effect on the sensitivity due to the difference of shielding around the loop. As shown in the figure, type A has no shielding around the loop while type B has shield. The shielding part for type B has been design to have characteristic impedance as ohm of strip line structure. The loop area for magnetic flux across is 1 mm by 1 mm square for both types. The transmission line part which is between the loop and the connector for cable is also design as ohm strip line which is designed by using the 2 dimensional transmission line impedance calculation tool Lipar, which is based on MoM (method of momentum), and its length is 17 mm. The SMA connector used for the probe covers the frequency up to 10 GHz. The probe has been made of three layered FR-4 based printed circuit board that has GND layer for top and bottom with signal layer for middle. 18

22 2. Modelingg and Improvement off Near-fieldd Measurements Cu FR µm µm 100 µ m Shield GND µm 330 µm SMA connector 15 mm 1 mm via 1 mm via 1 mm 1 mm Type A Type B Figure 3. Structure of probes Figure 4. Fabricated probe (left: Type A, right: Type B) Figure 5 shows the measurement setup for the evaluation of induced voltage to the probe by near-field. In this section, the targeting trace used for f the probe evaluation is microstrip line with mm line width. The port 1 of networkk analyzer (Agilent, 8510C) is connected to the MSL and the port 2 is connected to the t probe which position is manipulated by near-field scanner. With setting the probe close c to the MSL, magnitude and phase of S21 with frequencies from 45 MHz to 10 GHz were measured. Magnetic near-field probe was set to t have the loop parallel to the trace to get the magnetic flux across the loop. In the measurements, the probe height fromm the surface of MSL to t the center of the loop was set to t 1.1 mm for type A while the height of type B is set to make the induce voltage due to the magnetic near-field become same as type A. 19

23 2. Modeling and Improvement of Near-field Measurements Network Analyzer Agilent 8510 port 1 port 2 f = 45 MHz ~ 10 GHz Vs = 0 dbm magnetic near-field probe hp 100 mm micro strip line Ω chip resister Figure 5. Measurement setup In this evaluation, induced voltage due to the electric near-field V E and magnetic near-field V H are separately extracted by following procedure as shown in Figure 6. Fixing the probe height, induced voltage to the probe has been measured twice: original and 180 degree rotated. By this way, induced current due to magnetic near-field I H is considered to be differential to the probe although the induced current due to the electric near-field I E is considered to be common mode. Therefore, the induced current with probe at original angle I S0 can be described as I S0 = I E - I H, and at rotated I S180 is I S180 = I E + I H. From these characteristics, the induced voltage V H and V E can be separately evaluated by the addition and the subtraction of these two results as following equations. (1) I H 180 degree rotation I H I E I E H H current current Figure 6. Direction of induced current for electric field and magnetic field 20

24 2. Modeling and Improvement of Near-field Measurements Figure 7 shows the frequency dependence of the probe A for the induced voltage due to the magnetic near-field V H and electric near-field V E. In the result, V E increased and becomes similar level as V H with the difference less than 6 db between them. The result indicates that the magnetic near-field measurement results of ohm MSL using the probe type A could include the error due to the electric near-field up to 6 db. Figure 8 shows the evaluated V H and V E results with same manner as type A. Notable aspect is that the results for type B has obvious dips around at 3.5 GHz and 9 GHz, which result in V H / V E > 6 db. It has to be noted that these results are based on the ohm MSL and the results should depend on the impedance of target trace. Figure 9 shows measurement results of the probe height dependence of V E for type B. The probe height h p is varied from 0.6 mm to 1.4 mm with 0.2 mm step. The results suggest that the dip frequencies at 3.5 GHz and 9 GHz have been independent from the height, which means that the dips depend on only the probe characteristic not on the coupling with the targeting race. Therefore, modeling and design of the probe including the dip should be able to suppress the effect due to the electric near-field. V H, V E [db V] 80 V H dB V E Frequency [GHz] Figure 7. Evaluation results of induced voltage by electric near-field ( V E ) and by magnetic near-field ( V H ) for type A 21

25 2. Modeling and Improvement of Near-field Measurements V H, V E [db V] 80 V H 70 6dB 60 V E Frequency [GHz] Figure 8. Evaluation results of V H and V E for type B h p = 0.6 mm 0.8 mm V E [db V] mm 1.2 mm 1.4 mm Frequency [GHz] Figure 9. Probe height dependence of induced voltage by electric near-field ( V E ) Equivalent Circuit for Sensitivity to Electric Near-field Lumped element equivalent circuit model of the probe including the MSL to describe the measured results of V H is created as shown in Figure 10 for probe type A. Electric near-field coupling to the probe is expressed as capacitive coupling Cs to describe the electric flux from MSL to the loop element of the probe. The value of Cs was calculated using Linpar which is the transmission line analysis software based on the method of momentum (MoM) [21]. The transmission part of the probe is divided into 8 parts of LCRG ladder of which wavelength is shorter than the one tenth of the wavelength at upper frequency 10 GHz. Self-inductance at the transmission part and the loop element 22

26 2. Modeling and Improvement of Near-field Measurements Lts, Ltg, Lsv, Lsh and Lex are calculated using following equation which is for rectangular conductor [22] (2) where, l is the length, W is the width of conductor and thickness t has to be t << W. Lsv is the partial inductance for the vertical part of loop element, Lsh is the horizontal part, Lex is also the partial inductance of the trace connecting the loop with GND via. The capacitance C and the conductance G are calculated with using Linpar. The coupling coefficient k1 between signal and GND traces for transmission line and k2 between GNDs are calculated by three dimensional electromagnetic analysis. Cc is the stray capacitance between the pad for ohm termination chip resistor (1.6 mm by 0.8 mm) and GND, which value is calculated as the parallel plate capacitance using the pad size. Lc is the parasitic inductance of SMA connector for the connection to the cable and the value is obtained from impedance measurement result of the connector itself. loop microstrip line transmission line Lsv Rsv Cc k2 Ltg 8 ladder Rtg k2 Ltg Rtg Lex Rex Lsh Rsh k1 k1 Ct Gt k1 Ct Gt Lts Rts Lts Rts Ct Gt k1 Ct Gt Ltg Rtg Ltg Rtg Lsh Rsh Lsh Rsh Cs Lsv Rsv V E 0 dbm Lc network analyzer connector Figure 10. Schematic diagram of equivalent circuit for electric near-field (type A) 23

27 2. Modeling and Improvement of Near-field Measurements Figure 11 shows the equivalent circuit for type B. Type B has the shielded loop which is different from type A. Therefore, the capacitance and conductance Cl and Gl for loop element have been added. Other elements for the loop Lsv, Lgv, Lsh and Lgh are calculated using the equation (2) same as type A. Also since the shielded GND of the loop is close to the MSL, the capacitance between the shielding and the MSL is considered as Cg. These Cl, Gl and Cg are calculated using also the Linpar. Table 1 shows the derived values for each parameter. k2 k1 k1 V E transmission line Rgh 8 ladder Lgh Ltg Rtg k2 Ltg Rtg Ct Gt k1 Ct Gt Lts Rts Lts Rts Ct Gt k1 Ct Gt k2 Rgh Lgh Lsh Ltg Rtg Ltg Rtg Lgh k2 Rgh Rsh 0 dbm loop Lgv k1 Cl Lsv k1 Cl Lgv Lgh Rgh Lgv k1 Lsv Cl k1 Cl Lgv microstrip line Rgv Gl Cc Rsv Cg Gl Rgh Cg Rgv Lgh Cs Rgh Rgv Cg Gl Lgh Rsv Gl Cg Rgv Lc network analyzer connector Figure 11. Schematic diagram of equivalent circuit for electric near-field (type B) 24

28 2. Modeling and Improvement of Near-field Measurements Table 3 Derived parameters Type A Type B Type A Type B Lsv 0.35nH 0.42nH Lts 0.68nH 0.66nH Rsv 9.6mΩ 4.4mΩ Ltg 0.33nH 0.57nH Lsh 0.14nH 0.16nH Ct 0.16pF 0.14pF Rsh 4.8mΩ 2mΩ Gt 56kΩ 62kΩ Lex 0.24nH - Rts 7mΩ 6.2mΩ Lgv nH Rtg 0.5mΩ 1.8mΩ Rgv - 0.9mΩ Cs 0.02pF 0.02pF Lgh nH Cg pF Rgh - 0.6mΩ Cc 0.5pF Cl - 0.1pF Lc 3nH Gl - 88kΩ k k Calculated results of V E using the equivalent circuit model described above are compared with measured results in Figure 12 and Figure 13 for types A and B, respectively. A notable aspect in these results is that the calculation for type A has no dip although that for type B does due to the resonance around 4.2 GHz and 8.8 GHz, which shows good correlation with measurements. However, the measured dips are 3.6 GHz and 9.1 GHz, which are slightly different from the calculations. It is considered that the accuracy for these dips can be improved by considering the frequency dependency of permittivity of the board or by considering the parasitic parameters for the via between signal trace and GND. V E [db V] measurement calculation Frequency [GHz] Figure 12. Calculation and measurement results of V E for type A 25

29 2. Modeling and Improvement of Near-field Measurements V E [db V] measurement calculation Frequency [GHz] Figure 13. Calculation and measurement result of V E for type B Equivalent Circuit for Sensitivity to Magnetic Near-field Next, the equivalent circuit for the calculation of induced voltage due to the magnetic near-field V H is studied. Figure 14 and Figure 15 show the equivalent circuit for type A and B respectively. The values for parasitic parameters are all same as the equivalent circuit as described. The induced voltage at loop due to the magnetic flux is expressed as V loop. The induced voltage e can be described using the magnetic flux across the loop as e = -d /dt. In the static state at AC, phasor expression is as follows using angular frequency, (3) It can be also expressed as (4) where, is permeability, S is the loop area. However, since the magnetic flux across the loop is not uniform, the distribution of magnetic field has to be considered for accurate calculation. 26

30 2. Modeling and Improvement of Near-field Measurements loop transmission line 8 ladder Lex Lsv Rsv Lsh k2 Ltg Rtg k2 Ltg Rtg Rex Rsh V H k1 Ct Gt k1 Ct Gt Vloop k1 Lts Rts Lts Rts k1 Ct Gt Ct Gt Ltg Rtg Ltg Rtg Lsh Rsh Lsv Rsv Lsh Rsh Figure 14. equivalent circuit of probe for magnetic near-field (type A) loop transmission line Rgh 8 ladder Lgh k2 Ltg Rtg k2 Ltg Rtg k2 Lgv k1 Cl Lsv k1 Cl Rgh Lgv Lgh Gl Gl Rsv Rgh Lgh V H k1 Ct Gt k1 Ct Gt Vloop Lsh Lgh Lts Rts Lts Rts Rgh Rgh k1 Ct Gt k1 Ct Gt Lgv k1 Lgh Ltg Rtg Ltg Rtg Cl Gl Lgh k2 Rsv Rsh Lsv Rgh k1 Cl Gl Lgv Figure 15. equivalent circuit of probe for magnetic near-field (type B) Effective Height for Magnetic Near-field Measurement Because the loop is placed very close to the target current, the magnetic field in the loop is not uniform as shown in Figure 16. The distributed field must be considered for accurate calculation. In this study, considering the signal and return current as shown in Figure 16, the total magnetic flux with consideration of such non uniform field is calculated and the height which gives same total flux without consideration of non-uniform field is defined as effective probe height h eff derived from following equation. 27

31 2. Modeling and Improvement of Near-field Measurements (5) where, a is the distance between the bottom of loop and signal current, b is the distance between top of loop and the signal current, w is the distance between signal current and return current. For this calculation, it assumed that the target trace length is long enough than the probe height and the magnetic field H from the current I is calculated using the relation H = I/2 r. From the equation, h eff can be expressed as (6) Figure 17 shows the calculation result of heff with w = 0.3 mm, loop size is 1 mm by 1 mm using equation (6) comparing the height of loop center h center. The calculated h eff is 0.4 mm when a is 0.1 mm while the h center is 0.6 mm, which difference is almost 30 %. This result indicate that consideration of field distribution inside loop avoid the error 30 %. height Magnetic field w MSL a b current Figure 16. height dependence for magnetic near-field inside loop 28

32 2. Modeling and Improvement of Near-field Measurements [mm] 5 Loop center: h center Effective height: h eff h center h eff Loop height: a [mm] Figure 17. Calculation result of effective measurement height (h eff ) with loop center height (h center ) By using the h eff and the equivalent circuit, induced voltage due to the magnetic field VH are calculated as shown in Figure 18 and Figure 19 for types A and B, respectively. Both results show good correlations, and the maximum difference is 5 db around 7.5 GHz for both types. Since the difference is observed at higher frequency, it could be caused by unconsidered parameters such impedance mismatching at the connector V H [db V] measurement calculation Frequency [GHz] Figure 18. Calculation and measurement result of induced voltage by magnetic near-field ( V H ) for type A 29

33 2. Modeling and Improvement of Near-field Measurements V H [db V] measurement calculation Frequency [GHz] Figure 19. Calculation and measurement result of induced voltage by magnetic near-field ( V H ) for type B Application of Position Signal Difference Method The spatial resolution of magnetic near-field measurement depends on the loop size of probe. It is obvious that the smaller size of loop can measure the field with higher spatial resolution. However, as described, the probe characteristics rely on its structure and it is complicated due to many parasitic parameters. Therefore it is not efficient to fabricate probes for the specific spatial resolution required for each measurement. Here new measurement technique to control spatial resolution without fabricating probes on the basis of position/signal difference (PSD) method is described. PSD method can be done by subtracting two measured results obtained at two different probe heights as shown in Figure 20, which method has been proposed by Kantor and Shvets for electric field measurement [19]. It can be applied for magnetic near-field measurement by following assumption. When the loop area at probe height A defined as S 1, induced voltage due to the magnetic field across the loop depends on the S 1. With slight shift z of probe and considering the magnetic field across the loop S 2, subtracted results can be separately described as S 1 and S 2 as shown in Figure 20 since the common loop area S C disappear. This can be express as following equations. (7) 30

34 2. Modelingg and Improvement off Near-fieldd Measurements S 2 S 2 S C S 1 Position A Position B z S 1 Figure 20. Schematicc diagram of spatial resolution improvement method In the equation, the induced voltage due to the S 2 can be neglected when it is smaller enough than the induced voltage due to the S 1. The ratio of these t induced voltages due to S 1 and S 2 is shown in Figure 21 withh various z from 30 m to 0 m with the condition that the probe height is fixed to 0.1mmm and the loop size is 1 mm m by 1mmm which is same as type A and B. The targett trace length is assumed to be longer enoughh than the probe height in this calculation. The figure suggests that the ratio is more thann 17 db for z = 30 mm and 13 db even for z = 0 m. Therefore, it is suggestedd that neglecting the effect of S 2 can be lesss than 2 db, which is little effect for the measurements. Figure 21. Calculation results of inducedd voltage ratio for S 1 to S 2 The PSD method is then applied to actual magnetic near-field measurement for the first time. The fabricated probe type B was used for the measurement. The target was the same MSL as above that had a 0- m-wide signal trace. The probe height was fixed to 0.1 mm for the original position, and z was set to 30 m, which is the condition in which 31

35 2. Modeling and Improvement of Near-field Measurements the ratio of S 1 to S 2 is 17 db. The probe manipulator used for the measurement was EMV-200 from Peritec, Inc. (formerly Hitachi Displays, Inc.). The positioning accuracy of the manipulator is within 4 m, which is sufficiently small and accurate enough for this measurement. Figure 22 shows the measurement results of magnetic near-field using the probe type B with and without applying the PSD. The results are normalized to the peak value at position = 0 mm. The full width at half maximum (FWHM) without PSD is 2.4 mm, and it was improved to 1.4 mm by applying the PSD, which is a 40 % improvement. Also, the result with PSD shows good agreement with the result calculated with a 30 m loop. By applying the PSD, the effective probe height h eff in Figure 17 improved from 0.9 mm for the original to 0.53 mm with PSD. This indicates that equivalent loop size is miniaturized by using the PSD method. By employing the PSD method, the spatial resolution can be controlled by changing the probe shift value, and this enables the equivalent loop size in one-time near-field scanning to be varied depending on the portion of target PCB. Normalized amplitude Original With subtraction :Calculation :Measurement Position [mm] Figure 22. Comparison between calculation and measurement using type B with/without applying subtraction 2.2. Measurement-Based Modeling of Dual Loop Magnetic Near-field Probe The error due to background electromagnetic fluctuations is not problematic for measurements of single PCBs or LSI circuits because experiments are generally performed in shielding chambers and a manipulator is sometimes used to scan the probe across the surface of the device being tested. Furthermore, other devices in the vicinity of the device being tested are turned off when performing measurements in a shielding chamber; this ensures relatively low background noise to enable accurate measurements 32

36 2. Modelingg and Improvement off Near-fieldd Measurements to be performed. However, in actual systems that contain many electronic devices (e.g., automotive vehicles), it is usually extremely difficult to operate just one device due to the complexity of the system. Consequently, in magnetic-near field measurements off actual systems, magnetic fields from other devicess besides the target device could cause critical measurement errors. For hybrid electric vehicles, the main electronics e (which include the inverter and/or converter) ) generate high magnetic fields around the system, which could result in inaccurate measurements as shown in Figure 23. Shielding the probe is i one potential way to overcome this problem. However, a shielding conductor can adversely affect the sensitivity and the t spatial resolution of the probe, especially e att high frequencies, and an improperly designed probe could reducee the qualityy of measurements. Another solution is to use a probe that t has twoo coils connected in series with opposite polarities to prevent disturbance from a distant source, which gives high h immunity only against background magnetic disturbance [4]. This technique uses a simpler probe. However, the probe design has not been investigated sufficiently. In this section, the investigation on two probes with different structures, namely two separate and serially connected ferrite-core coils is described. We studied them experimentally and theoretically too understand the details of magnetic near-field measurementss and to determine the optimal probe design. Figure 23. Schematic of problem in actual application 33

37 2. Modeling and Improvement of Near-field Measurements Fabricated Probes for Evaluation Two different probe structures were prototyped and evaluated. Figure 24 schematically depicts the two probes. Probe A in the figure consists of two ferrite-core coils stacked vertically, while probe B has a coil with two cores that are positioned adjacent to each other. In both probes, the two coils are serially connected. The cores are made from ferrite (HF-90, TDK) and have dimensions of 8.2 mm (thickness) by 20 mm by 20 mm (see Figure 24). In both probes, there is a 5-mm gap between the cores. Three coil turns are made from wire (AWG 20) with an insulating outer layer. In both probes, two coils are connected in series to cancel out the voltages induced by magnetic fields from distant sources that generate a uniform magnetic field (see Figure 25). To determine the magnetic near-field generated by the current flowing in the target or a conductor, probe A measures the difference between the voltage induced in the upper and lower coils. Therefore, the distance to the source of the magnetic near-field is determined by the gap between the two coils. Probe A generally has a lower sensitivity than probe B since probe B measures the sum of the voltages induced by the magnetic near-field in the two coils without subtraction. However, the spatial resolution in the scanning plane is determined by the gap between the two coils for probe B, whereas the resolution can be controlled by changing the probe scanning step for probe A. By considering these features of the probes, an appropriate structure has to be selected based on the required sensitivity and spatial resolution for a specific application. In this study, two separate cores are used, but a single core would give a better sensitivity unless the two coils are not balanced. 3 turns Ferrite 20 mm 5 mm 8.2 mm 10 mm 10 mm Probe A 20 mm 20 mm 8.2 mm 5 mm Probe B 3 turns 2 mm Figure 24. Schematic structures of two dual loop magnetic near-field probes 34

38 2. Modeling and Improvement of Near-field Measurements Magnetic field disturbance Magnetic near-field Enclosure Surface current (distributed) Current PCB Probe A Probe B Figure 25. Magnetic near-field and far-field coupling to dual loop probes. The frequency characteristics of the magnetic near-field were measured using a microstrip line and a network analyzer (8753D, Agilent) (see Figure 26). The measurement frequency range was 30 khz to 0.1 GHz, which includes the low-frequency service bands (e.g., AM and FM) since the present study focuses on large-scale systems such as automotive systems and such systems often have major problems in low-frequency bands. The microstrip line board has a 0.5-mm-wide signal line on the top layer and a GND plane on the back with a 1.6-mm-thick FR-4 board. The two port S-parameters were measured for an input power of 0 dbm using two probes at the center of the microstrip line (MSL) board. Probe A contacted the board; thus, the gap between the probe wire and the MSL signal line is considered to be the same as the wire insulation thickness, which is less than 1.0 mm. Figure 27 shows the measurement results for a magnitude of S 21 for both probes. As expected, the measurement results indicate that probe B has a better sensitivity than probe A. The slope at low frequencies is 20 db/decade, but it decreases around 10 MHz because the impedance of the probe inductance increases and the voltage drop at the probe inductance becomes dominant. The two magnetic near-field probes with canceling structures for distant disturbances exhibit different frequency characteristics. This difference needs to be explained theoretically and calculations need to be performed to design an appropriate probe that satisfies the target specifications and to determine what the probes measure. 35

39 2. Modeling and Improvement of Near-field Measurements Network analyzer (8753D) Port 1 Port 2 Probe Microstrip line ohm termination Figure 26. Measurement setup for probe evaluation -20 S21 [db] Probe B Probe A Frequency [Hz] Figure 27. Measurement results of S 21 for probes A and B Modeling of Probes Figure 28 shows a simplified equivalent circuit of the probe for the voltage V H induced by the magnetic near-field across the loop. The induced voltage V H can be expressed as in the equation (4), V H = e 0 NSH where S is the area of the loop, N is the number of turns, and H is the magnetic field strength across the loop. The effective permeability e is a key parameter for describing the probe characteristics; it can be calculated from the probe inductance L C using the theoretical expression for the inductance. The probe inductance L C was first determined from the measurement result for the input impedance obtained using the S 22 parameter. However, the magnetic field distribution around the probe must be considered in this process. 36

40 2. Modeling and Improvement of Near-field Measurements Network analyzer C S Probe V H L C Figure 28. Equivalent circuit of dual loop magnetic near-field probe Figure 29 shows a schematic diagram of a probe with magnetic flux across the coil. The magnetic field distribution when the probe is excited by port 2 of the network analyzer is assumed to be the same or very similar to that the probe detects for the magnetic near-field generated by MSL excited by port 1 for probe B. However, the magnetic field distribution for probe A when it is excited by port 2 (i.e., the magnetic field distribution when measuring the input impedance of probe) is assumed to differ from that when the probe measures the magnetic near-field from MSL (see Figure 29). Therefore, the effective permeability e for calculating V H can be extracted using the probe inductance L C obtained from the S 22 result, just as for probe B. These assumptions regarding the magnetic field distribution have been confirmed by 3D electromagnetic simulation software HFSS from ANSYS. The model shown in Figure 30 has been constructed with some simplifications (e.g., direct excitation of the tracing edge without using a connector). Two ferrite-cored coils are modeled and set at the same locations as in the measurements. Simulations were performed at 0.1 GHz for both probes. Figure 31 shows the magnetic field distributions obtained for the cross-section of the center of the MSL board. As expected, the magnetic field distributions obtained by exciting the probe (during S 22 measurement) and by exciting the MSL (during S 21 measurement) for probe B are very similar, whereas they clearly differ for probe A. 37

41 2. Modelingg and Improvement off Near-fieldd Measurements Stimulus source of network analyzer for S21 measurement Magnetic field during S22 measureme ent PCB Cur rrent Probe A Magnetic near-field by target current Probe B Figure 29. Schematic of magnetic flux paths during S 21 and a S 22 measurements Figure 30. Simulation model of a dual loop magneticc near-field probe Figure 31. Simulation results for magnetic near-field distribution during S 21 and S measurements S 22 38

42 2. Modeling and Improvement of Near-field Measurements Based on these results, e to obtain the V H can be calculated from the probe inductance L C just for probe B. Figure 32 shows the frequency dependence of the input impedance of probe B. Probe B has a calculated inductance L C of 900 nh and a stray capacitance C S parallel to the inductance of 5 pf. Next, to obtain the induced voltage V H, the effective e was calculated from the measured L C. Theoretical expressions for the coil inductance are given below in terms of the core length D L [23]. (8) where (9) With (10) (11) (12) Here, r is the core radius; it is calculated by assuming that the cross-sectional area of the rectangular core (20 mm by 8.2 mm) is equal to that of the cylinder core, which has r = 7.2 mm. From these equations and the calculated coil inductance L C = 900 nh, the effective e was calculated to be 6.4. Using the effective e, the induced voltage V H for probe B was obtained using equation (1). Figure 33 compares the calculation results obtained using the effective e with the measurement results. This figure shows a good correlation at frequencies over 1 MHz. The same approach cannot be applied to probe A since the magnetic field distributions at the core during S 22 measurements (i.e., the input impedance measurement) and during S 21 measurements (i.e., the magnetic near-field measurement) obviously differ. Therefore, the effective e for a single-core coil is used for probe A since the dominant magnetic flux path contributing to V H for probe A contains just one core (see Figure 29 and Figure 31), which is similar to the situation for a single-core coil. e for a single-core coil can be theoretically obtained, but here it has been obtained from the input impedance. The obtained e of 6.8 was used to calculate the frequency dependence of S 21 for probe A, as shown in Figure 34. It agrees well with the measurement 39

43 2. Modeling and Improvement of Near-field Measurements results except for the resonance near 40 MHz. Generally, the non-uniformity of the magnetic field across the loop and the effect of electric near-field are major sources of error in magnetic near-field measurements [16][17]. In this case, the peak around 40 MHz is considered to be due to the electric near-field because the non-uniformity of the magnetic field across the loop affects frequencies below 10 MHz, where the slope is 20 db/decade. In addition, the difference from the measurement results below 1 MHz is assumed to be due to the frequency dependence of the effective permeability e. In this study, a constant e was used in the calculations, whereas in reality it could vary with frequency; we intend to investigate the frequency dependence of the coil inductance in a future study. Magnitude of impedance [ ] : Calculation : Measurement Frequency [Hz] Figure 32. Frequency dependence of input impedance for probe B : Calculation : Measurement S21 [db] Frequency [Hz] Figure 33. Comparison of calculation and measurement of S 21 for probe B 40

44 2. Modeling and Improvement of Near-field Measurements : Calculation : Measurement S21 [db] Frequency [Hz] Figure 34. Comparison of calculation and measurement of S 21 for probe A The peak observed around 40 MHz in the frequency response to the magnetic near-field of probe A is considered to be a resonance due to the electric near-field, which can be described as capacitive coupling between the coil winding and the signal trace. Figure 35 shows the equivalent circuit for probe A used to calculate the voltage V E induced by capacitive coupling C C. The coil inductance is divided into two components in order to consider the capacitive coupling C C. The coil inductance was estimated to be 1.1 H by fitting the measured input impedance of the probe shown in Figure 36. The parallel capacitance of the coil C Y was estimated to be 10 pf by fitting the measured resonant frequency at 48 MHz. Figure 37 shows the calculation results of S 21 for probe A with consideration of V E when C C is varied between 0.3 pf and 1 pf; these calculations reveal a peak at 40 MHz and a dip around 80 MHz, which provide better correlation with the measurement results. The voltage induced by capacitive coupling becomes comparable to the voltage induced by the magnetic near-field and significantly affects the result above the resonant frequency. Thus, it is important to reduce the capacitive coupling by miniaturizing the probe for high-frequency applications (the present study considers only low-frequency applications). The remaining disparity from measurements at frequencies over 40 MHz for probe A is assumed to be due to unconsidered stray capacitance around the probe and also the frequency dependence of the effective permeability, as described above. We intend to investigate this in a future study. 41

45 2. Modeling and Improvement of Near-field Measurements Probe C Y C Y V E C C 1.41 V Microstrip line Figure 35. Equivalent circuit for probe A that considers capacitive coupling between the probe and the microstrip line Magnitude of impedance [ ] : Calculation : Measurement Frequency [Hz] Figure 36. Frequency dependence of input impedance for probe A S21 [db] C C = 1 pf 0.5 pf 0.3 pf Measurement Frequency [Hz] Figure 37. Comparison of calculation and measurement results for S 21 for probe A when the electric near-field is considered With the configuration shown in Figure 38, the calculations of probe scanning over 1 ma current for different probe structures studied above have been conducted comparing with 42

46 2. Modeling and Improvement of Near-field Measurements the result by conventional single loop. The source as disturbance is considered as 100 ma at 1 meter away from the targeting current. Figure 39 shows the calculated results for different probe structures. The results without disturbance give obvious peak around center. However, in the results with disturbance, the result for single conventional loop shows the deterioration of signal-to-noise ratio clearly, which could lead to incorrect result. As demonstrated such typical calculation, the probes investigated in this section can be advantageous for the measurement under strong magnetic disturbances. Probe A Probe B scan 100 ma(source of disturbance) z 1mA x H x1 H x2 H z1 H z2 1mA Probe 1mA x=0 1000mm GND plane 0mm Figure 38. Configuration of calculation for different probe type without disturbance with disturbance Level [dbm] Probe B Single loop Probe A Level [dbm] Probe B Single loop Probe A Probe position [mm] Probe position [mm] Figure 39. Comparison of calculated scanned results between probes 2.3. Modeling and Improvement of Electric Near-field Measurements Electric near-field measurements using a scanning system were first published by Barrett and Barnes in 19 [24]. Since then, major research topics have included sensitivity improvements and extension of such measurements to higher frequencies [25]-[28]. Besides these, an electro-optic probe has been proposed for frequencies above 1 GHz [29]. One of the most important parameters for electric near-field measurements is the 43

47 2. Modeling and Improvement of Near-field Measurements broadband spatial resolution. As described in introduction, due to the continually increasing levels of integration of electrical components, higher spatial resolutions are required also for electric near-field measurements [30]-[32]. In this section, quantitative model of electric near-field measurements is proposed. On the basis of that, various techniques to improve the measurements in terms of spatial resolution are described Introduction of Position/Signal Difference Method The spatial resolution in electric near-field measurements is basically determined by the size and structure of the probe. However, the required spatial resolution is typically not the same over the entire surface of the target device; dense regions require a higher resolution than sparse regions. Therefore, different probes need to be prepared for different requirements, which is not a very efficient approach. In addition, miniaturized probes such as those fabricated using semiconductor processes are difficult to manufacture. To overcome these difficulties, the probe/signal difference (PSD) method proposed by Kantor and Shvets can be used [19][20]. This method achieves higher spatial resolution by subtracting measurement results obtained at two different probe heights above the target surface. Near-field measurements at an arbitrary resolution can be achieved by varying the heights of the two measurements. One advantage of this method is that the probe does not need to be physically miniaturized in order to obtain a high spatial resolution. Only fine control of the probe height is required, which allows the use of a single conventional monopole probe during measurements, leading to greater simplicity and efficiency. However, there are still some issues that need to be addressed with regard to the PSD method. First, the theoretical mechanism needs to be clarified quantitatively. Since the measurements are made in the electric near-field region, the probe characteristics must be described based on near-field considerations. Second, the extracting process of the polarization of the measured electric near-field also needs to be studied. Although a method has been proposed [19] that allows extraction of the normal and tangential electric field components, theoretical support and validation studies are still insufficient. We have proposed and validate a PSD method for carrying out high-spatial-resolution magnetic near-field measurements using a small loop probe, and have provided a theoretical validation for this method in section 2.1. In this section, electric near-field measurements using the PSD method are carried out based on capacitive coupling considerations between the probe and target device. The PSD method is then extended to a double position/signal difference (DPSD) method that allows the normal and tangential components of the electric near-field to be separately 44

48 2. Modelingg and Improvement off Near-fieldd Measurements determined. The proposed method is simulations. validated both experimentally andd using full-wave Probe Structure and Theoryy of Proposed Double Position/Signal Difference Method The electric near-field probe used in the present study was fabricated using a three-layer printed circuit board (PCB) as shown in Figure 40. The PCB was w made from FR-4. The 1st and 3rd layers contain a 2-mm-wide ground, and the 2nd layer l contains a signal trace with a width of 0.28 mm and a thickness of about m. All the conductors are made of copper. The length of the probe tip is 1 mm, and this is the part that detects the electric near-field. There is a gap of 0.35 mmm between the probe tipp and the edge of the signal s trace. The thickness of the FR-4 dielectric layer is 0.3 mm, which givess a characteristic impedance of for the transmission line. The length of the transmission line is 34 mm and it is connected to the cable c SMA connector. Figure 40. Schematic structure off printed monopole electric near-field probe The PSD method is a post-processingg method that uses twoo different electric near-field measurementss obtained at different probe heights. A typical probe scanning system has a manipulator that is capable of high-resolution positioning inn the heightt direction, which allows the two different heights to be precisely set. Assumingg that the first measurement is made at position A and the second measurement at position B in Figure 41, subtracting the second measurement from the first gives the difference due to the electric near-field detected only by the probe tip. The contributions from other parts of thee probe effectively cancel out since the slight upward shift of the probe has little l effect except in the tip 45

49 2. Modeling and Improvement of Near-field Measurements region. Figure 41 illustrates the electric near-field coupling between the probe tip and the target conductor (e.g., circuit board trace). The electric near-field that the probe measures contains both normal and tangential components. The PSD method mainly measures the coupling between the probe tip and the target conductor, which is considered to reflect the normal electric near-field component. Considering this coupling to be capacitive, the measured electric current I due to the electric near-field can be expressed as [26] (13) where C is the capacitance between the probe tip and the target conductor, and A is a probe constant. In Eq. (14), the measured current includes a dependency on a probe-structure related constant. However, if the same probe is being used for the measurements, the constant A is fixed, and the capacitance is the only variable. The currents measured at two different probe heights I 1 and I 2 are then expressed in Eq. (2), where the different capacitances are illustrated in Figure 42. (14) C z1 is the capacitance between the target and the probe tip, which is related to the normal component of the electric near-field. C xa and C xb are the capacitances between the sides of the probe and the target, and are assumed to be related to the tangential component. C xa is associated only with the tip region of the probe that is shifted between measurements. In contrast, C xb is associated with side regions of the probe other than the lower tip region, and is assumed to be unchanged between measurements. Thus, in the second measurement, C xa disappears and only C xb remains. By applying the PSD method, the C xb component is removed, so that only the terms C z1 -C z2 and C xa remain, as shown in Eq. (16) and Figure 42. (15) Here, C z1 -C z2 is related to the normal component of the electric near-field, whereas C xa is considered to be related to the tangential component. Since the difference in probe height 46

50 2. Modelingg and Improvement off Near-fieldd Measurements between the two measurements is sufficiently small to regardd C xb as being unchanged, C xa is also expected to be very small so that the term C z1 -C z1 dominates. GND Probe Signal Target h = 0 h Original Position A Position B Subtracted Figure 41. Schematic of position/signal difference methodd Figure 42. Capacitive couplings to the probe in position/signal difference method Next, in orderr to obtain the t tangential component of the electric e near-field, the DPSD method is proposed. This method simply involves subtracting the PSD result (I 1 -I 2 ) from the measured current I 2 att position B.. (16) Based on the assumption that C xa andd C xb are both related to the tangential component of the electric near-field, the t normal component can be eliminated when 2 C z2 -C z1 is negligibly small. Since these capacitances are determined by thee probe heights, measurementss can be carried out with a suitable choice of probe p heights to ensuree that 2 CC z2 -C z1 is negligibly small. One of advantages of this method iss that only two measurementss are required, which is also the case for the PSD method. 47

51 2. Modeling and Improvement of Near-field Measurements Measurements and Validation of PSD and DPSD Method Actual measurements were next carried out using a fabricated printed monopole probe. The measurement setup is shown in Figure 43. A network analyzer (Agilent 8753D) was used, connecting port 1 to the microstrip line and port 2 to the probe to measure the S 21 characteristics. A microstrip line was used as the target, and had a 0- m-wide trace with a characteristic impedance and a chip resistor termination. The probe was manipulated by a probe scanning system specially designed for near-field measurements. The S 21 measurements were made at two different probe heights of h = 0 mm and 0.35 mm at a frequency of 1 GHz. The probe scanning pitch was mm. Figure 44 shows the S 21 magnitude normalized by the maximum value at h = 0 mm. It can be seen that the measurement results at h = 0 mm appear narrower than those at h = 0.35 mm, which is reasonable. Both the PSD and DPSD methods were then applied, and the results are shown in Figure 45. Based on the assumptions described above, the PSD results represent the distribution of the normal component of the electric near-field and the DPSD results represent the tangential component distribution. Network analyzer (8753D) Port 1 Port 2 Probe Microstrip line termination Figure 43. Setup for electric near-field measurement 48

52 2. Modeling and Improvement of Near-field Measurements 1 Normalized amplitude h = 0 mm h = 0.35 mm Position [mm] [m] Figure 44. Normalized magnitude of induced voltage at h = 0 and 0.35 mm Normalized amplitude PSD DPSD Position [mm] [m] Figure 45. PSD and DPSD results Three-dimensional full-wave electromagnetic simulations were performed in order to validate the measurement results. The simulation model is shown in Figure 46. The simulation software used was HFSS from ANSYS. The physical dimensions of the model are exactly the same as those for the actual board. Figure 47 shows a comparison between the measurement results for a probe height h = 0.35 mm, the edge of signal trace height is 0.7 mm, and the simulation results for a position 0.7 mm above the trace. It can be seen that is reasonably good agreement between them, which validates the concept that the measurement results contain both normal and tangential components of the electric near-field. Figure 48 shows a comparison between the experimental PSD applied results and simulated results for the normal component of the electric field. Again, good agreement is obtained. Figure 49 shows a comparison between the experimental DPSD applied results and simulated 49

53 2. Modelingg and Improvement off Near-fieldd Measurements results, which confirms that the tangential component can bee extracted as predicted using the proposed DPSD method. Moreover, in order to conduct the quantitative comparison with simulated results, feature selective validation (FSV) has been used for the results. FSV is the algorithm to show how correlate between two results by using score [33][34]. Figure shows the FSV applied results for each PSD/DPSD applied results comparing with simulated results above. GDM tot is the score representing the correlation. The results suggest that it gives good correlation although the results for original and PSD are relatively better. Figure 46. HFSS simulation model off microstrip line for electric near-field analysis Normalized amplitude :Simulation :Measurem ment Position [mm] [m] 4 Figure 47. Measured and simulatedd magnitude of electricc near-field componentt

54 2. Modeling and Improvement of Near-field Measurements Normalized amplitude :Simulation :Measurement Position [mm] [m] Figure 48. PSD applied result and simulated normal component of electric near-field Normalized amplitude :Simulation :Measurement Position [mm] [m] Figure 49. DPSD applied result and simulated tangential component of electric near-field 51

55 2. Modelingg and Improvement off Near-fieldd Measurements Figure. FSV applied results for simulated and PSD/DPSD Figure 51 gives the PSD applied results with different probe heights shown in the figure. It also includes the normal componentt of electric field obtained by HFSS simulation at h = h 1. It clearly shows that the t changing the height of h 1 related to the electric near-field height. Figure 51. PSD applied results with different probe heightss comparingg to simulated normal component of electric field Accurate Modeling with Fringe Capacitance As described in previous section, s electric near-field measurement can bee expressed using 52

56 2. Modelingg and Improvement off Near-fieldd Measurements capacitive coupling model. In this section, quantitative fringe capacitance model between the probe and target trace is described. The probe and measurement setup used for the modeling and validations are shown in Figure 52. A three-layered printed circuit board is used for the monopole electric near-field has a signal element tip, which is 0.28-mm wide, 1-mmm long probe. The surface layers have a 2-mm-wide ground and the inner layer and 18- m thick. The conductors are made of copper and the substrate is FR-4 with 0.3-mm between the layers. The probe is connected to port 2 of networkk analyzer 8753D and the microstrip line (MSL) with a -ohm termination is connected to port 1 during the measurements. The frequency response of S 21 is then observedd at 1 GHz with scanning the probe along the x-axis, as shown in Fig. 1, at 0.1-mm steps. The measurements were performed at various probe heights too obtain the PSD-processed result I PSD based on the following definition of PSD. (17) where the measured value is expressed as electricc current based on the considerationn that the current is induced due to the electric near-field the height for I 2 iss set to h p + d p, where the. The probe height h p for I 1 in i the measurementss is fixed to 0 mm, and displacement d p is varied from f 0.1 to 0.7 mm. Figure 52. Probe and setup for PSD/DPSD measurementss Figure 53 shows an example of the PSD-processed results r for different probe displacements d p. It can be seen that the full width at half maximum m (FWHM) w fw, which is generally used to evaluate the spatial resolution of the measurements, is varied depending on the dp. Since the d p is the key parameter for PSD, it is very important to understand this aspect. 53

57 2. Modeling and Improvement of Near-field Measurements Normalized amplitude of I PSD w fw Microstrip line d p = 0.1 mm 0.4 mm 0.7 mm x-position [mm] Figure 53. PSD measurement results with changing shift parameter d p at 1 GHz We assume the amplitude of the signal detected by the probe can be described by focusing on several of the fringe capacitances, as shown in Figure 54. The figure shows the definitions of the capacitance between each portion of the probe tip and the trace, where C s is the capacitance between the side of probe and the surface of the trace and C b is between the bottom of the probe and the side of the trace. These fringe capacitances are calculated using the following equations [35]. where, (18) (19) where W p is the width of the probe tip (0.28 mm), is the permeability, s is the separation of the probe from the trace edge, and h is the probe height. L e is the length of the probe tip element (1 mm) for the calculation of C s and the thickness of MSL ( m) for C b, Wm is the trace width of MSL (0.5 mm) for C s and the thickness of the probe element (18 m) for C b. Therefore, C s and C b can be calculated as (20) 54

58 2. Modelingg and Improvement off Near-fieldd Measurements C f is the capacitance between the corners of the probe tip and the trace, which can be expressed as follows [35]. (21) Figure 55 indicates the calculated FWHM using the fringe capacitance model above while comparing it to the measurement results with different d p. The difference between the calculated and measured results is within 13%, which is a good g correlation. The notable aspect is that selecting a greater d p value makes the spatial resolution r worse for PSD due to the increasee in fringe capacitance between the probe element and thee trace. It iss clear from the equations above that t this fringe capacitance depends on not onlyy the shape of the probe tip but also the dimensions of the targeting trace. Therefore, it is important to select an appropriate d p in order to obtain the required spatial resolution with respect to t the finest trace in the device under testing, which can be derived using the model addressed. Figure 54. Fringe capacitance model between probe element e andd trace 55

59 2. Modeling and Improvement of Near-field Measurements Full widh at half maximum w fw [mm] 1.5 Calculation 1 Measurement 0.5 w fw Displacement d p [mm] Figure 55. Calculated FWHM wfw compared to measurements with various dp Proposal and Validation of Improved PSD Method It was revealed that the fringe capacitance between the probe element and target is the factor that causes the degradation of the spatial resolution discussed in the previous section. Since the fringe capacitance to the side of the probe is considered to correspond to the tangential component of the electric near-field, it is important to use a smaller dp displacement value for the measurement of the normal component of an electric field. However, the limit of the finer displacement is normally decided due to the performance of the manipulation hardware and its repeatability/stability, e.g., the probe manipulator used for this study can handle a displacement value dp down to 100 m in order to maintain good repeatability. Therefore, it is required to eliminate the fringe capacitance without needing the finer displacements of probe. The new measurement technique for eliminating the fringe capacitance is proposed in this section. Figure 56 shows the procedure for taking the measurements. It requires two sets of PSD results obtained by using the same displacement d p, which are defined as I PSD1 and I PSD2. When taking into account that the remaining fringe capacitance C a is nearly equal to C b, only the capacitances related to the normal component remain by subtracting I PSD2 from I PSD1, as can be determined using the following equation. (22) where I 1, I 2, and I 3 are the measured results at different probe heights. The fringe capacitance to the corner of the probe tip is neglected since it is sufficiently small. As in 56

60 2. Modelingg and Improvement off Near-fieldd Measurements the equation, this method can be processed using the threee sets of results I 1, I 2, and I 3 acquired using the same displacemen nts d p. The processed result when using the proposed method is shown in Figure 57 (b) with the comparison too conventional PSD and the normal component of electric field E z obtained by electromagnetic simulation at z = 0.35 mmm as definedd in Figure 57 (a), where d p was set to 0.1 mm for I PSD1 and I PSD2 and the separation of the probe tip from the trace was 0.35 mm. It is obvious that the proposed method shows better matching with the simulated E z than the conventional PSD without using the finer displacements of probe. The only disadvantag e of using this method is that the signal level is also reduced by thee repeated subtractions, but this would not become a serious problem since the field of interest is normally strong for EMC applications. Figure 56. Procedure of proposed method Figure 57. (a) Definitions of parameterss and (b) Comparison between b conventional PSD and proposed method 57

61 2. Modeling and Improvement of Near-field Measurements Role of Probe Displacement (d p ) for DPSD Since the probe displacement d p is a common parameter for PSD and DPSD, it is very important to understand its meaning and effect not only for PSD but also for DPSD. The DPSD measurement results can be obtained by using the process in the following equation along with the same definitions as in Eq. (1). (23) DPSD is also the post process for eliminating the normal component of the electric near-field coupling between the probe tip and targeting trace, and the fringe capacitance remains. Therefore, we believe changing the probe displacement d p is equivalent to changing the measurement height of the tangential electric field. This assumption is verified by comparing the DPSD processed results at different d p to the tangential component of electric field E x obtained from an electromagnetic simulation at the corresponding heights. The results are shown in Figure 58, where the definition of the parameter is as shown in Figure 59 (a). In order to compare these results, the peak-to-peak distances d t as shown in the inset of Figure 59 (b) were evaluated as one feature parameter and summarized in Figure 59 (b). From this comparison, the change in probe displacement d p provides a good correlation to the change in observation height for a tangential electric field Ex. Normalized amplitude p of I DPSD Measured by DPSD E x obtained by simulation 0.2 MSL x-position [mm] Normalized amplitude of I DPSD x-position [mm] Normalized amplitude of I DPSD x-position [mm] Figure 58. DPSD measurement results at different dp comparing to Ex obtained by electromagnetic simulation. (a) d p = 0.1 mm, E x at z = 0.45 mm (b) d p = 0.4 mm, E x at z = 0.75 mm (c) d p = 0.7 mm, E x at z = 1.05 mm 58

62 2. Modelingg and Improvement off Near-fieldd Measurements Figure 59. (a) Configuratio on for measurements and simulation (b) Comparison of peak-to-peak distance d t between measurements at various d p andd simulation results 59

63 3. Modeling and Analysis of Radiation from PCB-Chassis Structure 3. Modeling and Analysis of Radiation from PCB-Chassis Structure Several techniques and design guidelines have been developed for reducing electromagnetic interference (EMI), especially for printed circuit boards (PCBs). Such techniques are based on the characterization of noise sources, paths, and radiation [36] [40]. However, in product development of electronic devices, electromagnetic compatibility (EMC) problems that exceed these requirements/regulations at the system level are sometimes encountered even when each component has been designed properly based on EMC design guidelines or rules. Such system-level problems are generally found at a very late stage in product development and can thus cause major problems in terms of the timing and/or cost of product launching [41]. Some of the most important components that need to be accounted for when considering the system-level EMC behavior are GND connections between a PCB and an enclosure or a metal chassis since they can increase the radiation [42][43]. Several approaches have been studied to reduce or minimize emission in order to overcome system grounding problems [44] [46]. Grounding configurations also affect the immunity performance [47], [48]. A rapid, quantitative design method based on characterization of PCB chassis systems is required to implement all EMC considerations for complex circuit and grounding structures. In this chapter, specially designed current probe to determine the contributions of the GND current to the radiation is proposed followed by the measurements of radiation from PCB-chassis structure which is the typical configuration of integrated components as indicated in table 2. Then the analytical modeling to clarify and understand the radiation behavior as well as the guideline to lower the radiation is studied Radiation from PCB-Chassis Structure A 3.5 hard disk drive with a PCB mounting on its chassis was used for these measurements (Figure 60). The chassis has dimensions of 147 mm (length) 102 mm (width) 22 mm (thickness), while the PCB is 93 mm in width and 47 mm in length. The spacing between the PCB and the chassis is approximately 1 mm, although the chassis surface is not perfectly flat. The PCB used for the measurements has a specially designed 60

64 3. Modeling and Analysis of Radiation from PCB-Chassis Structure circuit for the experiment that simulates typical problems that are likely to occur in an actual product. The circuit is schematically depicted in Figure 61 and has a 10-MHz crystal and an oscillator, as well as high-speed buffers (74AC541). Bypass capacitors between the power supply and GND are located close to the buffers and the connector of the power supply harness to suppress power bus noise. The PCB is a four-layered board that consists of a top layer for mounting components, GND, Vcc, and a bottom layer as shown in Figure 62. The spacing between the GND and Vcc layers is 0.15 mm, while the spacings between the GND and top layers and between the Vcc and bottom layers are 0.22 mm. An external battery is connected to supply 5 V through a 90 mm long twisted wire harness clamped with common-mode chokes to eliminate its contribution to the radiation. The PCB has a screw (length: 5 mm; radius: 1 mm) that connects the PCB GND to the chassis at four different locations (see Figure 61) 102 mm 95 mm Screw1 147 mm Screw3 Screw2 Screw4 mm Figure inch hard disk drive chassis and PCB for evaluation 0 ferrite cores GND 10 F 0.1 F Vcc 5V 787RH3 10MHz 0 74AC F 0.1 F 0 74AC pF 0 10pF 0.1 F 10pF 10pF Figure 61. Schematic circuit diagram for evaluation board 61

65 Setup for electric far-field measurements 3. Modeling and Analysis of Radiation from PCB-Chassis Structure Figure 62. Patternn layouts of evaluation board Figure 63 shows the measurement setup. The vertical and horizontal h electric fields 3 m from the device were measured at frequencies of from 30 MHz to 1 GHz by an EMI receiver while the device was powered on. The device was placed on a rotating table and the peak values of the electric field were measured by rotating the table through 360. Figure 64 and Figure 65 respectivelyr show the vertical and horizontal h components of the frequency spectra of the measured electric field. Both figures include the measured spectra for the PCB without a chassis and for the PCB with the chassis connected by screw 1, together with the envelope of the spectra for all the configurations to facilitate comparison of the configurations. Figure

66 3. Modeling and Analysis of Radiation from PCB-Chassis Structure The measured electric field from the PCB without the chassis shown in Figure 64 exhibits two peaks at approximately 0.12 and 0.3 GHz up to 32 dbµv/m. These peaks are considered to be related to the circuit because the board size is electrically smaller than the first resonance of the PCB. The spectrum of the PCB mounted on the chassis but with no electrical connection is almost identical to that of the PCB without a chassis (see Figure 64). This indicates that capacitive coupling between the PCB and the chassis makes a negligible contribution to the radiation in this frequency range. Comparison of these results with a screw connection reveals that the peak at approximately 0.3 GHz increased to 45 dbµv/m for screws 1 and 4 (see Figure 64), but that the peak at around 0.12 GHz changed little. When measurements with a single screw connection were performed, other screws were replaced with non-conducting screws. The horizontal electric field measurement results shown in Figure 65 exhibit the same tendency whereby a peak around 0.3 GHz is increased by more than 20 db by screws 1 and 4, whereas they make little contribution to the peak at 0.12 GHz. Figure 66 shows the difference in the measured electric far-field between the PCB only and the PCB mounted on a chassis for each screw connection. The results clearly indicate that the radiation due to the mounting of the PCB on the chassis is primarily 0.3 GHz radiation rather than 0.12 GHz radiation, which means that the 0.12 GHz component for the cavity between the PCB and the chassis is slight. Since the cavity between the PCB and the chassis is the target to be modeled in the following section, the highest peak at 0.3 GHz is the peak to be estimated. Another remarkable aspect of the measurement results is that the peak level at 0.3 GHz depends on the position of the screw that connects the PCB GND to the chassis. The highest electric field was obtained with screw 4, while the lowest was with screw 2; the difference was up to 8 db for both vertical and horizontal far fields. This increase in the emission on adding the chassis represents a possible problem that could be observed in an assembled system during EMC testing. The PCB circuit design is typically optimized at the PCB level using PCB design guidelines since it is difficult to predict the assembly-level EMC behavior during circuit design. However, as mentioned above, it is becoming increasingly important to consider the EMC of an assembled system at the PCB design level (although PCB designers are not responsible for fully integrated system design). 63

67 3. Modeling and Analysis of Radiation from PCB-Chassis Structure Ev [db V/m] Ev [db V/m] 60 PCB without chassis Screw Frequency [GHz] Frequency [GHz] 60 Ev [db V/m] PCB without chassis Screw 4 Screw 1 Screw 3 Screw 2 20 PCB mounted on chassis but no electrical connection Frequency [GHz] Figure 64. Measured vertical electric far-fields for four different screw connections, PCB without a chassis, and PCB mounted on a chassis but with no screw Eh [db V/m] Eh [db V/m] 60 PCB without chassis Screw Frequency [GHz] Frequency [GHz] Eh [db V/m] PCB without chassis Screw 4 Screw 1 Screw 3 Screw Frequency [GHz] PCB mounted on chassis but no electrical connection Figure 65. Measured horizontal electric far-fields for four different screw locations, PCB without a chassis, and PCB mounted on a chassis but with no screw 64

68 3. Modeling and Analysis of Radiation from PCB-Chassis Structure Ev [ V] Screw4 Screw1 Screw3 Screw2 Eh [ V] Screw4 Screw1 Screw2 Screw Frequency [GHz] Frequency [GHz] Figure 66. Difference in measured electric far-fields between the PCB only and the PCB mounted on a chassis for vertical (left) and horizontal (right) electric far-fields Proposal and Application of Screw Current Probe This section describes the measurement technique focusing on the relationship between PCB design and radiated emissions from the chassis. As described in introduction, measurement technique to reveal RF behavior of integrated system is important. We assume the screw current flowing through the screw which connects the PCB GND with chassis is the key to estimate the radiation from such structure. The thin current probe for the measurement of screw current is proposed and used for the investigation on the PCB-chassis structure in this section Structure of Screw Current Probe The current that causes radiated emissions from the chassis must flow through the screw connecting the PCB GND to the chassis GND. To verify this assumption, we fabricated a specially thin current probe to measure the screw current by placing the probe between the PCB and the chassis without changing the gap as shown in Figure 67. The external diameter of the probe was 12 mm, the internal diameter 8 mm, and thickness 1 mm. The probe has a 9-turn coil to detect magnetic fields originating from current through the screw, connected to a thin coaxial cable running to the spectrum analyzer. We fabricated the thin-current-probe by the standard PCB process and were able to measure current through the screw during actual board operations without modification of setup. Figure 68 shows the actual fabricated screw current probe. The surface of the probe is covered with resist and insulating tape to prevent contact with the PCB pattern. Capacitive coupling between the probe and PCB pattern can be neglected because the output from probe was hardly changed when the probe was set to plastic screw. Figure 69 shows the frequency 65

69 3. Modeling and Analysis of Radiation from PCB-Chassis Structure dependence of the magnitude of output power from the probe as a dot, measured with network analyzer using a 0 dbm calibrated signal source to estimate the probe characteristics. The solid line in Figure 69 shows the calculation results for inductance L = 9 nh for the loops and the size of one turn for magnetic field detection for an area S measuring 1 mm by 3 mm. The output voltage V of the probe can be calculated from the relation as in following equation, where N is the number of turns and is termination impedance of network analyzer. The calculation results in Figure 69 are consistent with estimates. (24) PCB chassis chassis current Radiation screw GND PCB To spectrum analyzer 1 mm Magnetic field current To chassis 8 mm 12 mm Figure 67. Schematic diagram of screw current measurement 66

70 3. Modeling and Analysis of Radiation from PCB-Chassis Structure Figure 68. Fabricated screw current probe Output power [dbm] -20 measurement theory Frequency [MHz] Figure 69. Frequency characteristics of induced power for fabricated junction current probee by 4.48 ma current Measurements of Screw Current Figure 70 shows the frequency spectra of the screw currentt measured with a fabricated thin-current-probe. In this measurement, plastic screws aree used for all but one screw electrically connected to the t chassis GND when current measurements s were performed. The unit of magnitude is converted from voltage to ampere usingg the calibration coefficient. The peak frequency around 0.3 GHz and the envelope shapee of the spectrum are similar to the EMI spectrum. The results suggest that the t current that causess EMI flows from the chassis PCB GND to the chassis through thee screw. Figure 71 shows the measurement results for magnetic near-field distributions measured at 1.5 mm above the PCB at 0.32 MHz with the configuration of each screw connection. The measurements were performed using 1 mm by 1 mmm single loop probe with setting of the probe to detect 67

71 3. Modeling and Analysis of Radiation from PCB-Chassis Structure tangential component of magnetic near-field. These results can visualize the sources of EMI on PCB, and slightly higher current distributions and density around the screw, but it is difficult to estimate the path of current flowing into chassis. It is not best way to determine the current path of EMI from chassis in this case. As described, conventional magnetic near-field scanning can help to locate the EMI source on PCB but the screw current probe can be more straightforward technique for identifying the path of chassis EMI currents. Based on the measured screw current, including magnitude, phase, and position allows the modeling and the estimation of radiation from the chassis by using an appropriate model of chassis. Current [ A] Screw1 Current [ A] Screw2 Current [ A] Screw Frequency [MHz] Frequency [MHz] Current [ A] Frequency [MHz] Screw Frequency [MHz] Figure 70. Measurement results of screw current for each screw under operating db( A/m) 40 screw screw mm screw3 20 screw mm Figure 71. Magnetic near-field distribution on PCB with connection for each screw 68

72 3. Modeling and Analysis of Radiation from PCB-Chassis Structure 3.3. Cavity-Mode Modeling using Double Summation Method The PCB-on-chassis structure is a common configuration in real products, and it has been demonstrated that the cavity model can be used to describe the RF behavior for the configuration [54], [55]. Regarding the parallel plane structure, a lot of studies in terms of EMC have been made so far for high frequency behavior [53], application for integration with enclosure [56], lossy structure [57], arbitrary shape of parallel plates [58], populated conditions [59], and so on. Recently, meshed equivalent circuit modeling has also been studied for multilayered structure [60]. Another important aspect to achieve better EMC performance at system level is to provide not only radiation estimation but also a design guideline/direction to show how to lower the radiation. Since the study of EMC is a more industry-related field, just the analysis is sometimes not sufficient from the engineering point of view. In this section we describe the approach to obtain the electric field radiation using the measured screw current. This approach assumes the board-chassis is a cavity and uses theoretical expression of an impedance network for the cavity-mode model. Modeling the cavity resonator is one theoretical approach for characterizing rectangular PCB power bus structures [61]. The impedance (Z) distribution between the two plates is given by the solution of the two-dimensional Helmholtz equation [53] and is used to calculate the edge voltage V. (25) The constant mn depends on indices as follows: (26) where sinc(x) = sin(x)/x and x i, y i, x j, and y j are the coordinates of the ith and jth ports in the 69

73 3. Modeling and Analysis of Radiation from PCB-Chassis Structure plane, respectively. denotes the permeability, which is assumed to equal that of vacuum 0, and is the angular frequency. k xm and k yn are respectively the coefficients for the m th and n th modes and are given by (27) The electric field distribution along the boundary of parallel plates shown in Figure 72 (a) can be obtained by using the impedance distribution above with screw current as the source. In this study, the screw current measured in the previous section is used as the source Calculation of Radiation from Parallel-Plate Structure For this parallel plate structure with an electrically small separation, the radiation from the edge field can be calculated from the equivalent magnetic current at the boundary [52]. Figure 72 shows the calculation model used in this study (width: 47 mm; length: 93 mm; separation: 1 mm; relative permittivity: 1). The frequency range considered is 30 MHz to 1 GHz, which is lower than the first resonance for the present case. The tangential magnetic field is set to zero; this approximation is valid when there is a small gap h between the PCB and the chassis. We used an equivalence principle that was based on Huygens equivalence principle and had reportedly been used for a PCB power bus structure [53]. The electric current is considered to be zero (J = 0) at the edge of the parallel planes; consequently, the magnetic current was used as the source. The magnetic current at the edge was obtained from the electric field E, the magnitude of which can be calculated from the voltage V at the edge and the separation h as V/h. The PCB conductor planes and the chassis are regarded as perfect conductors and the four sides of the PCB have no tangential magnetic field. Therefore, the magnetic current M can be expressed as (28) where nˆ is a normal unit vector on the boundary surface. Equation (5) was used to obtain the tangential magnetic currents at the four side walls and the vector potential F was then calculated using Equation (27) as given in [52]: 70

74 3. Modeling and Analysis of Radiation from PCB-Chassis Structure (29) where r is the distance from the origin to the field observation point, r' is the distance from the origin to the source, and is the angle between r and r'. The electric far-field E is defined in terms of the vector potential F for magnetic sources and A for electric sources by using Equation (28) as given in [52]: (30) By using the approximation J 0, A is set to 0. Therefore, the electric field can simply be expressed as (31) (a) a = 93 mm b = 47 mm (x j, y j ) Screw PCB (x i, y i ) h = 1 mm Current: I S V i Ez Chassis (b) b = 47 mm a = 93 mm M h r 0 z y x Observation point r Figure 72. Calculation model for the radiation (a) Schematic diagram of the parallel plane cavity resonator (b) Equivalent magnetic current loop 71

75 3. Modeling and Analysis of Radiation from PCB-Chassis Structure Asymmetric Size Coefficients The electric field radiation was calculated for parallel planes of the same size, which is applicable when modeling most PCB power-ground structures using the cavity resonator model. Since the electric field source at the edge of the parallel planes in the calculation is vertical, the radiated electric far-field will predominantly be vertical. However, the horizontal component of the measured electric field shown in Figure 65 and Figure 66 is higher than the vertical component. This is assumed to be due to the chassis plane being 100 mm longer than the other parallel plane. Most PCB chassis systems have asymmetric dimensions, which should be considered when calculating the far-field. In the present study, the radiated electric field was calculated by using a correction coefficient that incorporates modification of the far-field due to the asymmetric dimensions; the coefficient is obtained using the three-dimensional electromagnetic simulation software HFSS. Figure 73 shows the three-dimensional model used in the simulation of the electric far-field. It is a simplified model that has just two different sizes of parallel perfectly conducting planes (separation: 1 mm) and a 1 A current source as the excitation source. Frequencies used in the simulation were in the range 0.01 to 1 GHz. The electric field was evaluated at 3 m from the device over 360 in the x y plane, just as in the measurements. To avoid introducing unintentional errors due to the mesh used in this simulation, a finer mesh was employed until the calculation results became stable and did not vary greatly. In the simulation, the asymmetric dimension was evaluated by changing the extension length of the chassis L ex (see Figure 73). Figure 74 shows the variation in the vertical and horizontal components of the electric field when the extension length of the chassis L ex was varied from 0 to 100 mm with the excitation source at the center of plane 1. The results shown in Figure 74 reveal that the vertical field varies little, whereas the horizontal component varies drastically with the asymmetric dimension. When L ex = 100 mm (the approximate size of a hard disk drive chassis), the horizontal component exceeds the vertical component. These simplified simulations reveal that the horizontal component can exceed the vertical component due to the asymmetric dimensions of the parallel plane. Therefore, in the present study, the changes in the horizontal electric field due to extension of the chassis length were implemented using the correction coefficient C E in far-field calculations: Ec E 1 C E (32) 72

76 3. Modeling and Analysis of Radiation from PCB-Chassis Structure where Ec is the horizontal electric field considering the correction coefficient, which is calculated from the original horizontal electric field E. Figure 75 shows the calculated frequency dependence of the correction coefficient C E for four different excitation locations (S 1 to S 4 ); the dependence was calculated using the same dimensions as the device under test (i.e., L ex = 100 mm). The results indicate that excitation at location S 2 gives the highest correction coefficient, although otherr locations give similar values of approximately 0.3 GHz. Since the source located at S 2 is closest to the center of the t PCB, thee radiated electric field for parallel planes with equal sizes is considerably lowerr than that at the other three source locations. Therefore, the asymmetric dimensions affect the result much more for asymmetric parallel planes. Since thee calculated coefficient depends d only on the physical dimensions of the device, far-field calculations to evaluatee the effect of changing the design of the PCB or circuit can be performed more rapidlyy than full-wave simulations with all of the circuitry at every design change. Figure 73. Electromagnetic simulation model used to calculate the asymmetric parallel plane coefficient Horizontal E-field [V/m] L ex = 100 mm mmm 10 mm 0 mm 0.1 Frequency [GHz] 1 Vertical E-field [V/m] L ex = 100 mm mm 10 mm 0 mm 0.1 Frequency [GHz] 1 Figure 74. Calculated vertical and horizontal electric far-fields at a distance of 3 m for different chassis sizes 73

77 3. Modeling and Analysis of Radiation from PCB-Chassis Structure Asymmetric size coefficient C E 10 2 Screw Screw 4 Screw 3 Screw Frequency [GHz] Figure 75. Asymmetric size coefficient for horizontal electric far-field with different screw locations Using the calculated coefficient C E, the electric field at 3 m from the device was calculated and compared with the measurement results (see Figure 76 and Figure 77 for vertical and horizontal components, respectively). Comparison of the results of the vertical component for S 1 and S 2 reveals that the calculation results for these two screw locations clearly differ. Furthermore, for all four screw locations, the horizontal component exceeds the vertical components when the correction coefficient is used, which agrees with the measurement results. The reason why the peak at about 0.12 GHz does not appear in the calculations is thought to be because the cavity which was modeled here does not contain the 0.12 GHz component, as mentioned above. At frequencies over around 0.4 GHz in the calculation results especially for screw location S 2, there is a maximum deviation of over 10 db for the magnitude. This error is considered to be due to the asymmetric size coefficient and its implementation which is the correction for maximum electric far-field strength but the radiation at those higher frequencies may have deviations on radiation angle due to the asymmetric size of planes, which was not considered in the coefficients. Another possible cause may be the simplified simulation model because the coefficients were derived with flat rectangle parallel planes but actual product has more complex chassis structure. Since the simple image magnetic current on the boundary of cavity for asymmetric edge cannot express the results observed in these measurements, imperfect cavity wall on the side or the effect of electric current on the corner of asymmetric chassis may need to be considered for more accuracy. The approach described here requires full-wave simulation only one time to prepare the coefficients and it can be useful for this kind of PCB-chassis system such as hard disk drive that has standardized dimension and has no frequent change of shape. On the other hand, from the practical point of view, such a full-wave simulation could include most of radiation behavior without the combination of an analytical expression such as the radiation from cavity wall. However, breaking down the behavior into detailed functions 74

78 3. Modeling and Analysis of Radiation from PCB-Chassis Structure to understand the mechanism that how big the contribution of asymmetric size, screw location and how big the contribution of the radiation from cavity itself is most important to get feedback to design knowledge. Ev [db V/m] Screw 1 Calculated 60 Screw 2 Calculated Measured Measured Frequency [GHz] Ev [db V/m] Frequency [GHz] Ev [db V/m] Screw 3 70 Calculated 60 Screw 4 40 Measured 30 Measured 20 Ev [db V/m] Calculated Frequency [GHz] Frequency [GHz] Figure 76. Comparison of calculated and measured vertical electric far-fields for four different screw locations 75

79 3. Modeling and Analysis of Radiation from PCB-Chassis Structure Eh [db V/m] Eh [db V/m] Screw 1 Calculated 60 Screw 2 Measured Frequency [GHz] 70 Screw 3 Calculated 60 Screw 4 Measured Frequency [GHz] Eh [db V/m] Eh [db V/m] 40 Calculated Measured Frequency [GHz] Calculated Measured Frequency [GHz] Figure 77. Comparison of calculated and measured horizontal electric far-fields for four different screw locations 3.5. Fast Calculation using Inductive Network Method The other approach to obtain electric field strength at the boundary of cavity for the estimation of radiation is impedance network calculation with screw current as stimulus. From previous study the current flowing through screw connecting PCB and chassis is considered to be the excitation source for radiation. Also the current probe specially designed for screw with small gap was described as shown in Figure 67. It is placed between the PCB and the chassis and is used to measure the magnetic near-field around the screw with little interference to the system. By making this probe sufficiently thin, the probe does not alter the gap between the PCB and the chassis. In this section, inductive network method which is applicable when the physical size of cavity resonator is electrically smaller than first resonance is employed [54]. The interest frequency 0.3 GHz is lower than the first resonance due to the device size. The advantage of inductive network method is fast calculation time, which is important when it is applied for actual product design. The voltage V i at the observation point due to the current I S is described as following equation [54]. 76

80 3. Modeling and Analysis of Radiation from PCB-Chassis Structure (33) where G A is the Green's function for the static magnetic vector potential, which is as follow based on the condition that the aspect ratio b/a should be greater than 1, where a and b are width and length of PCB respectively. (34) where, (35) C 0 is the parallel planes capacitance with consideration of fringing at the open boundaries. (36) The advantage of inductive network model employed here is shorter calculation time than conventional double summation method as stated above. In case of these calculations the result for one screw location can be obtained more than hundreds times faster than conventional method. From the equations above with measured screw current in Figure 70 as stimulus, electric vertical and horizontal far-field radiations for different screw locations were calculated as in Figure 78 and Figure 79. Electric far-field calculation from cavity resonator model used is same approach described in section 3.4 as well as asymmetric size corrections. In calculated results, the peak at 0.3 GHz and the order depending on the screw locations showed good agreement with measured far-field. For horizontal polarization in Figure 79 over 0.4 GHz contains more error than lower band possibly due to the asymmetric size coefficients as described. Overall results obtained by screw current measurements with inductive network model showed similar accuracy to the results of double summation method. 77

81 3. Modeling and Analysis of Radiation from PCB-Chassis Structure Ev [db V/m] Ev [db V/m] Screw 1 Inductive network 60 Screw 2 Measured Ev [db V/m] Frequency [GHz] Frequency [GHz] 70 Screw 3 60 Screw 4 Inductive network Inductive network Measured Ev [db V/m] Inductive network Measured Measured Frequency [GHz] Frequency [GHz] Figure 78. Comparison of calculated and measured vertical electric far-fields for four different screw locations. Eh [db V/m] Eh [db V/m] Screw 1 Inductive network Screw 2 60 Inductive network 40 Measured Measured Frequency [GHz] 70 Screw 3 Inductive network 60 Screw 4 Measured Frequency [GHz] Eh [db V/m] Eh [db V/m] Frequency [GHz] Inductive network Measured Frequency [GHz] Figure 79. Comparison of calculated and measured horizontal electric far-fields for four different screw locations. 78

82 3. Modeling and Analysis of Radiation from PCB-Chassis Structure 3.6. Modeling by Multilayered Finite Difference Method As describe above, it is revealed that the cavity consists of PCB with screw current as stimulus can be considered as the source of radiation. Also the study clarified that the radiation from the PCB-chassis system can be calculated the electric field distribution between the PCB and the chassis along the PCB edge as the source of radiation. In this section we introduce the multilayered finite difference method [49] to obtain the distribution. The PCB planes and chassis were created as a meshed plane with RLC passive components as shown in Figure 80. The plane contains 9 4 mesh. The size of one mesh is about 10 mm, which is one tenth the wavelength at the highest frequency of 1 GHz. Table 4 lists the parameters of the passive elements used for signal tracing and planes; they were calculated using the following equations (37) obtained from [49] with separation between layers h. In Table 4, Rp is the resistance of the cell, Cp is the capacitance between the GND and Vcc, Cc is the capacitance between the Vcc and chassis, Lp is the inductance of the Vcc and Lc is that for the chassis treating the GND as a reference, k is the coupling factor for the inductances, and Gp and Gc are the conductance between planes. Rpe, Lpe, Cpe, Lce, and Cce are parameters at the boundary of planes corresponding to the parameters described above, which were calculated for half of the cell []. In addition, the screw was modeled as having 1.8 nh inductance, which was calculated as that for a circular cylinder model [51]. (37) Figure 80 also shows the equivalent circuit parameters derived from its dimensions. The chassis was modeled as one of the layers, but only the part of it that was coupled to the PCB was implemented in the meshed equivalent circuit model since the coupling between the PCB plane and the chassis outside the counter surface was assumed to have a negligible effect on the electric field strength at the boundary. 79

83 3. Modeling and Analysis of Radiation from PCB-Chassis Structure P1 1nH P2 3.5 nh 2 nh 14 nh P3 10 pf 10 pf 8 nh F 0.1 F 5 nh 1 nh G1 G2 0.1 F 0.1 F 8nH 1 74AC nh 2 nh 10 pf 10 pf G3 Screw 1 93 mm Screw 3 Cpe Cp Gp Gp Lpe Rpe Lp Rp Gc Cce Cc k Gc Lce Lc Rpe Rp Screw 2 P1 P2 P3 Screw 4 G1 G2 G3 47 mm GND Vcc Chassis Figure 80. Multilayered equivalent circuit of PCB with a chassis Table 4 Derived Values for RLC mesh Symbol Value Symbol Value Rp Rpe Lp 0.18 nh Lpe 0.36 nh Cp 30 pf Cpe 15 pf Lc 1.44 nh Lce 2.88 nh Cc 1.1 pf k 0.38 Cce 0.6 pf Gp 2.8 k Gc M To ascertain the model s validity, we calculated the spectra of the screw current from PSPICE equivalent circuit simulation, and measured them by using the screw current probe as described in section 3.2. The calculated and measured results are compared in Figure 81. The peak at around 0.3 GHz was observed in both the measurement and calculated results. Screw 4 gives the highest screw current at 0.3 GHz, while screw 2 gives the lowest screw current; this agrees with the equivalent circuit calculations and the measurements. The difference between the calculated and measured values below 0.2 GHz is due to the noise floor, which appeared only in the measurement results. This good agreement between the results is a good indication of the model s validity. From the 80

84 3. Modeling and Analysis of Radiation from PCB-Chassis Structure validated model, the electric field distribution along the PCB edge can be obtained using the difference in voltage between the PCB and the chassis layer, and its separation can be used to calculate the radiation. Screw Current [A] Screw Current [A] Frequency [GHz] Frequency [GHz] Screw Current [A] Screw Current [A] 10-1 Screw 1 Calculation Screw 2 Measurement Frequency [GHz] 10-1 Screw 3 Screw 4 Calculation Calculation 10-2 Measurement Measurement Calculation Measurement Frequency [GHz] Figure 81. Comparison of calculated and measured frequency spectra of screw current for four different locations of the screws connecting the PCB and the chassis. The electric field at 3 m from the device was calculated and compared with the measurement results (see Figure 82 and Figure 83 for vertical and horizontal components, respectively). In calculated results, the peak at 0.3 GHz and the order depending on the screw locations showed very good agreement with measured far-field. These results obtained by multilayered finite difference model showed similar or better accuracy than the results of cavity resonator model with measured screw current as stimulus. 81

85 3. Modeling and Analysis of Radiation from PCB-Chassis Structure Ev [db V/m] Ev [db V/m] Screw 1 Screw 2 MFDM 60 MFDM Measured Measured Frequency [GHz] Frequency [GHz] 70 Screw 3 60 Screw 4 MFDM Measured Frequency [GHz] Ev [db V/m] Ev [db V/m] Measured MFDM Frequency [GHz] Figure 82. Comparison of calculated and measured vertical electric far-fields for four different screw locations Eh [db V/m] Eh [db V/m] Screw 1 MFDM 60 Screw 2 Measured Frequency [GHz] Eh [db V/m] 70 Screw 3 Screw 4 MFDM 60 MFDM 40 Measured 30 Measured Frequency [GHz] Frequency [GHz] Eh [db V/m] MFDM Measured Frequency [GHz] Figure 83. Comparison of calculated and measured horizontal electric far-fields for four different screw locations As described above, two different approaches to calculate the cavity modeling as well as 82

86 3. Modeling and Analysis of Radiation from PCB-Chassis Structure the MFDM-based calculation were described. The comparison of these approaches is described in table 5. The advantage of inductive network method and double summation approach are that the creation of equivalent circuit is not required and it can be faster way to know the radiation without performing radiated emission tests but this is more suitable for later stage of product development since the actual device is necessary to perform screw current measurements. Hence, the advantage of the MFDM is that active or nonlinear components can easily be integrated by using SPICE circuit simulation, which enables application at the design stage, i.e., before the prototype stage. Therefore, as the table shows, the suitable product development phases differ for each approach, but the results obtained from each approach were sufficiently similar as studied in this chapter. Table 5 Comparison of different approaches Approach (1) (2) (3) Suitable Design Fabrication & evaluation product phase Source Active components (SPICE) Screw current (I S ) Frequency Less than first Depends on cell size Broadband limitation resonance Impedance Double summation Inductive network* MFDM (Sec. 3.2) distribution (Sec ) (Sec ) Object E-field radiation (Sec. 3.3) Advantage No measured information required Fast calculation time & no size limitation Fast calculation time 3.7. Investigation on the Reduction of Radiation In this section, the study to lower the radiation from PCB-chassis structure is described. The theoretical modeling and calculation can provide the estimation results that how high the electromagnetic radiation will be but it normally does not provide how to lower the radiation. Therefore it is required to clarify the guidelines or some design concepts to achieve the reduction of the radiation from the design point of view. 83

87 3. Modeling and Analysis of Radiation from PCB-Chassis Structure Additional Bypass Capacitor at Screw Since it has been revealed that the screw current is the stimulus for PCB-chassis cavity in previous section, the bypass capacitor to reduce the screw current is considered to be one of the approach when the capacitor is placed very close to the screw. The effect of the additional capacitor which is located close to the screw is experimentally verified in actual measurement using the PCB which has special pads for the capacitor as shown in Figure 84 as the example of screw 1. The size of via for this special land is 0.3 mm diameter and of anti-pad is 1.3 mm. All other configurations for measurement were same as original setup described in Figure 63. The capacitor value used for the measurement is 1000 pf surface mount type. The impedance specification of capacitor has self-resonant frequency around at 160MHz due to the parasitic inductance which is about 1 nh. However, since the impedance around the target frequency 300 MHz is still lower than the impedance between Vcc and GND, it is considered to be effective to lower the screw current by placing the capacitor. But it is more effective reduce the parasitic inductance for lowering the screw current especially in high frequency. Vcc via to Vcc plane screw GND bypass capacitor between Vcc and GND Figure 84. Additional bypass capacitor located close to the screw for experiment Figure 85 shows the envelope of measured frequency spectra of electric far-field with and without 1000 pf capacitor close to screw 1 and 4 respectively. The peak level around at 0.3 GHz is lowered down by up to 10 db due to the additional 1000 pf capacitor for both configurations: screw 1 and screw 4. However, another peak which is exist around at 0.13 GHz is increased by about 7 db in the measured results, which is considered to be related to the resonance due to the circuitry on the PCB with impedance between Vcc and GND planes. As described, since the cavity model does not include the radiation from the circuitry itself, the increase at 0.13 GHz cannot be calculated in the model described in previous section. The implementation of that radiation from the PCB is part of the future work. 84

88 3. Modeling and Analysis of Radiation from PCB-Chassis Structure Figure 86 shows the measured frequency spectra for the condition with connections using both screw 1 and 4 comparing with and without 1000 pf bypass capacitors close to screws. In the figure, the result with 1000 pf indicates almost 19 db lower than the result without capacitors at peak frequency 0.3 GHz. As it has been seen in the results for single connection at screw 1 and 4, the peak around 0.13 GHz is increased by 6 db comparing to the original result. In order to verify these are due to the reduction of screw current, frequency spectra of screw current with and without 1000 pf capacitors have been measured as shown in Figure 87. The results show that the current for each screw is lowered more than 10 db at 0.3 GHz, which is good correlation with the radiation results. These demonstrated results suggest that the reduction of screw current, which is considered as the stimulus for the cavity of PCB-chassis structure, can be one approach to reduce the radiation. Although the capacitance value was fixed to 1000 pf in this study, the optimum value in terms of radiation should depend on the cavity and PCB design. In order to eliminate such peak around 0.3 GHz in this case, band eliminate filter design could be applicable instead of just bypass capacitor which is placed next to the screw. E 3m away [db V/m] Screw 4 (plastic screws for others) Screw 4 with 1000pF Screw 4 (plastic screws for others) Screw 1 with 1000pF Frequency [MHz] Figure 85. Frequency spectra of electric far-field with/without additional bypass capacitor. 85

89 3. Modeling and Analysis of Radiation from PCB-Chassis Structure E 3m away [db V/m] E 3m away [db V/m] Frequency [MHz] E 3m away [db V/m] Screw 1 and 4 with 1000pF for both 1 and Screw 1 and 4 without capacitor Frequency [MHz] 19 db lowered at 0.3 GHz Frequency [MHz] Figure 86. Improvement of radiated emission with additional capacitors Screw current [ A] Screw 1 Screw 4 Screw 4 (both 1 and 4 with 1000 pf) Screw 1 (both 1 and 4 with 1000 pf) Frequency [ z] Figure 87. Measurement results of screw current for screw 1 and 4 with connections of both screw 1 and 4 with/without 1000 pf capacitors Model-Based Analysis for the Reduction of Radiation In this section, the relation between the impedance between the ground and chassis and the screw current has been investigated for the PCB-on-chassis structure to lower the screw current that is considered as the stimulus of the cavity consists of the PCB and 86

90 3. Modeling and Analysis of Radiation from PCB-Chassis Structure chassis. The impedances at each node were then investigated using the created equivalent circuit model. Figure 88 (a) shows the impedance distribution between PCB GND and Vcc observed at each node with no screw connection but showing the location of the screw, and Figure 88 (b) shows the impedance distribution between GND and chassis at 0.3 GHz, which is the peak frequency of the noise that the device under test has. For these results, we assumed that there is some correlation between the impedance and the screw current that is the stimulus for the PCB-chassis cavity. Figure 89 shows the calculation results of the magnitude of the screw current at 0.3 GHz when the screw was placed at each node corresponding to the impedance between GND and chassis at same node without any screw connection. The results indicate that there is a correlation between impedance and screw current: higher impedance leads to lower screw current. Also, the figure indicates the calculation results for screw 1 to 4, which suggests screw 2 and 3 give lower screw current than screw 1 and 4. Figure 90 (a) shows the impedance observed at each node with screw connection at S 2. Since it is shorted at S 2, the impedance at S 2 was the lowest. From the correlation shown in Figure 89, it is better to place the screw at the node that has higher impedance, i.e. the node at S 3 in Figure 90 (a), in order to achieve lower screw current. Also, Figure 90 (b) shows the result with the connection at S 1, which indicates that the impedance at S 4 became the highest. Figure 91 shows the summations of screw current at 0.3 GHz by changing the configuration of the screws. As shown in this figure, even multiple connections, e.g. S 2 +S 3, provide lower total current than the current of the single screw connection S 2 derived from the calculations. These model-based investigations reveal that the single screw connection at S 2 or S 3 can provide lower screw current, resulting in lower radiation from the PCB-chassis system than at S 1 or S 4. Also, the screw current of the multiple connection at S 1 +S 4 can be lower than that at S 1 or S 4, and the screw current of the multiple connection at S 2 +S 3 can be lower than that at S 2 or S 3. Another notable aspect is that impedance calculations can be a guideline to decide which screw location should be selected to achieve lower radiation from the system. 87

91 3. Modeling and Analysis of Radiation from PCB-Chassis Structure (a) Between ground layer and Vcc layer (b) Between ground layer and chassis Figure 88. Calculated impedance distribution with noo screw connection 3 S 4 S 1 Current [ma] 2 1 S 3 S Input impedancee [ohm] Figure 89. Calculated screw currentt corresponding to the impedance i at each node 88

92 3. Modeling and Analysis of Radiation from PCB-Chassis Structure (a) Between ground layer and chassis with screw s at S 2 (b) Between ground layer and chassis with screw s at S 1 Figure 90. Calculated impedancee distribution with single screw connection 3 Screw Current [ma] S1 S S4 S1+S4 S2 S3 S2+S3 Connected screw(s) Figure 91. Total screw current att 0.3 GHz with different screw connections Figure 92 shows the calculation results of the horizontal electric far-field with different screw connections. As estimated from the screw current calculations, the result for S1+S 1 4 is lower than that of the single screw connection of S 1 or S 4 at around 0.3 GHz by almostt up to 10 db. But the results of S 2, S 3, and S 2 +S 3 do not clearly show a difference from the calculations although the results suggest the peak at around GHz is lowerr than that of the S 1 or S 4 configuration. 89

93 3. Modeling and Analysis of Radiation from PCB-Chassis Structure Lastly, the actual measurements were performed on the PCB-chassis structure in an anechoic chamber for the horizontal component in the same manner as shown in Figure 72. The results with different screw configurations are shown in Figure 93. Figure 93 (a) shows the connections at S 1, S 4, and S 1 +S 4, and Figure 93 (b) shows the connections at S 2, S 3, and S 2 +S 3. In (a), the result for S 1 +S 4 shows a lower peak than that of S 1 or S 4 by up to 8 db, which correlates well with the calculation results of screw current and of radiation. Also, the results of S 2 and S 3 indicate a lower peak than that of S 1 or S 4, which is as estimated in the calculation. Figure 93 (b) indicates that S 2 +S 3 are slightly lower than S 2 or S 3 although the calculation did not show such a clear trend. The reason is considered to be the coupling between the PCB and the harness that is connected to the PCB at a location closer to S 3, which is not considered in this study. These relations between the calculated screw current and the measured radiation are plotted in Figure 94 as the reduction for each configuration with reference to S 1 connection. The figure shows good correlation and it is verified that even multiple connections S 2 + S 3 can be lower than single connection S 3 which is correlated with screw current results. The results also suggest that the reduction of radiation can be achieved by the appropriate selections of screw locations. In this study, although the impedance is considered as the parameter related to the screw current, the electromagnetic field existing in the cavity of PCB with chassis obtained theoretically in previous section can also be used. Electric Field [db V/m] S4 S1 S1+S Frequency [GHz] Electric Field [db V/m] 60 S2 40 S3 S2+S Frequency [GHz] Figure 92. Calculated horizontal electric far-field with different screw connections 90

94 3. Modeling and Analysis of Radiation from PCB-Chassis Structure E 3m away [db V/m] E 3m away [db V/m] Frequency [MHz] 60 S1 S4 S1+S E 3m away [db V/m] Frequency [MHz] Frequency [MHz] Electric Field [db V/m] Frequency [GHz] (a) With S 1, S 4 and both S 1 and S 4 E 3m away [db V/m] Frequency [MHz] 60 S2 S3 S2+S3 E 3m away [db V/m] E 3m away [db V/m] Frequency [MHz] Frequency [MHz] Electric Field [db V/m] Frequency [GHz] (b) With S 2, S 3 and both S 2 and S 3 Figure 93. Measurement results of radiated electric field 91

95 3. Modeling and Analysis of Radiation from PCB-Chassis Structure Difference of screw current and radiation with reference to S1 [db] :Calculated screw(s) current :Measured radiation S4 S1+S4 S2 S3 S2+S3 Connected screw(s) Figure 94. Reduction of the calculated screw(s) current and the measured radiation 92

96 4. Conclusion 4. Conclusion This thesis covered the modeling and analysis of various measurements near-field probes in chapter 1. To apply magnetic near-field measurements to highly integrated devices such as mobile phones in a high frequency band over GHz, a detailed equivalent circuit was proposed on the basis of the physical structure of the probe with consideration of parasitic elements around a loop in section 2.1. The characteristics of induced voltage due to electric near-field and magnetic near-field were measured and compared with the calculated results obtained by using the equivalent circuit model. The results showed good correlation for both electric and magnetic fields with an accuracy of 6 db up to 6 GHz. The calculated and the evaluated results of probes also revealed that the induced voltage due to the electric near-field, which is generally treated as an error for magnetic near-field measurement, can be reduced by properly designing the probe structure to have the resonance only for electric near-field sensitivity. Furthermore, by using the fabricated near-field probe, a position/signal difference (PSD) method that enables control of the spatial resolution without miniaturizing the probe has been applied to magnetic near-field measurement for the first time. It was theoretically confirmed that the PSD is applicable for the targeting range with the fabricated 1 mm squared probe. The PSD applied results proved that the equivalent loop size can be downsized from 1 mm to 30 m without the fabrication of a miniaturized probe. These results in section reveal that the magnetic near-field measurements with arbitrary spatial resolution in GHz band can be realized. The probe studied in section 2.2 is a dual loop probe that has two coils in opposite polarities to cancel the magnetic field disturbance from surrounding electronics. It is applicable for the magnetic near-field measurements in complex systems such as electric vehicles. For a magnetic near-field probe that has a non-closed magnetic core, the derivation of effective permeability is key to calculating the probe characteristics. This thesis described the procedure to obtain the effective permeability from the impedance measurement results of the probe. The calculated results using the established equivalent circuit model showed good correlation with measured probe characteristics. In this thesis, electric near-field measurements were also covered in section 2.3. First, a double position signal difference (DPSD) method was proposed to obtain the tangential electric near-field with arbitrary equivalent probe size on the basis of the PSD method. The DPSD applied results using a microstrip line as DUT showed very good correlation with simulated results of the electric near-field. Since the electric-near field measurements using a monopole probe can be expressed as capacitive coupling between 93

97 4. Conclusion the probe and DUT, the fringe capacitance model to calculate full-width at half maximum (FWHM), which is equivalent to spatial resolution, was then analytically described. The calculated FWHM using the analytical model provides good matching with measured results with error less than 13 %. On the basis of the capacitive coupling model, an improved PSD method was then proposed to obtain a more accurate normal component of electric near-field. The proposed method was experimentally tested and obtained better correlation with simulated results than conventional PSD. Chapter 3 described the modeling and analysis of radiation from PCB-chassis structure. It was clarified that mounting the PCB on a chassis increases the radiation from PCB-chassis compared with PCB itself by up to 13 db. The thin current probe optimized for screw current measurements was then described in section 3.2. The probe was proposed for investigating the correlation between the radiation from PCB-chassis structure and the screw current. The measured frequency spectra of the screw current showed very good correlation with those of increased radiation due to the PCB being mounted on the chassis, which suggests that the screw current is a key parameter to estimate how the radiation from PCB-chassis structure changes due to the integration. A cavity resonator model using a double summation method as well as an inductive network method was then applied to model the PCB-chassis structure in order to obtain the electric field at the boundary of PCB, which was considered as the source of radiation. By considering the PCB-chassis structure as a parallel plate structure of cavity, the radiation was calculated using a magnetic current loop at the boundary edge of the cavity as well as asymmetric size coefficient to compensate for the difference in physical size between the PCB and chassis. The calculated results showed good correlation with measured results with each screw connection. As an alternative method to obtain the electric field at the boundary, a multilayered finite difference method (MFDM) was applied to the PCB-chassis structure in section 3.6. The results of radiation obtained by MFDM correlated quite well with measured results. In the last section, advantages of each method were clarified and discussed. With these developed models, the techniques to lower the radiation from PCB-chassis structure were investigated in section 3.7. The bypass capacitor placed close to the screw was proposed and verified as an approach to reduce the radiation. Also, it was revealed that the impedance distribution between PCB GND and chassis correlates to the screw current and experimentally verified that the higher impedance position provides lower screw current, which results in lower radiation. From the results above, modeling and improvements of near-field probes and electromagnetic radiation from PCB-chassis structure have been developed for realizing quantitative EMC design for growing electronics. 94

98 5. Appendix A: Stand-alone Electric-field Probe 5. Appendix A: Stand-alone Electric-field Probe A technique that scans an electric or magnetic field to determine the sensitive pattern on a printed circuit board (PCB) has been developed for investigating immunity failure. Such studies commonly use a magnetic near-field probe to locate the propagation path of an injected current and/or to inject current into a device by simulating electrostatic discharge or immunity testing such as bulk current injection. However, it is much more difficult to measure the electromagnetic field around a PCB in an enclosure than to measure the electric field outside the enclosure when performing radiated immunity testing of a system enclosed in a conducting case. This is because the electric field outside the enclosure is much higher than that inside and the probe and/or cable or the measurement equipment is sensitive to those unwanted electric fields. To overcome this problem, an optical fiber is generally used to eliminate direct interference with the cable attached to the probe. However, an optical cable can be used only if the device being tested has an appropriately sized gap or hole for routing the cable and if there is sufficient space inside the enclosure. If the enclosure does not have a wide enough gap or hole for the cable to pass through (which is usually the case for components used in moist environments such as automotive applications), the cover of the device will need to be opened or be modified to permit routing of the cable, which may affect the immunity testing result. Therefore, a measurement technique is required that does not use a cable connected to the probe. In addition, the electric-field probe needs to be miniaturized to enable measurements to be performed on small electronic components. In this section, a stand-alone electric-field probe has been evaluated. The probe monitors the maximum electric field detected by a printed bent monopole antenna and stores the value. A prototype probe was tested by applying an electric field. In addition, the prototype probe was used to measure the electric field inside an enclosure and the results obtained were compared with measurements by a conventional calibrated electric-field probe Prototype design and function test In immunity testing, the frequency is generally swept while monitoring the function of the 95

99 5. Appendix A: Stand-alone Electric-field Probe device being tested. Injection at a specific frequency is repeated if a deviation is observed with increasing injection power as a threshold test. When investigating the failure mechanism during threshold testing, an electric field probe is required to monitor only the electric field and not the frequency since the injection frequency is fixed. To simplify the circuit design of the prototype probe used in the present study, the probe monitors only the maximum electric field; it does not measure frequency data during injection. antenna pf MAX9933 PIC12F683 4 k output 2400 pf 1 k 33 pf 1 k 0.01 µf 1 k 1.5 V 1.5 V pre-amp (not populated) MAX6008 Figure 95. Schematic circuit diagram of prototype probe Figure 95 shows a schematic circuit diagram of the prototype probe. An RF power detector (MAX9933) is used to convert the RF power detected by the antenna into a DC voltage. The IC has a frequency range of 2 MHz to 1.6 GHz. A programmable IC (PIC; PIC12F683) is used to monitor the output DC voltage from the power detector and to store the maximum value measured when the device is powered up. When the external trigger for the PIC is activated, the PIC output pulses whose widths depend on the DC voltage stored by the PIC. Therefore, the electric field that the device detects can be determined by checking the pulse width of the output voltage from the device. The RF power detector provides a reference voltage of 1.25 V for the A/D converter. All the devices were powered by a 3.0 V battery. Figure 96 shows the board layout. The board is made from FR-4 and has dimensions of 30 mm 30 mm. It has two sides: the top side contains ICs and passive components and the bottom side has two 1.5 V coin batteries. The antenna used to detect the RF power is printed on the board as a pattern. It is a bent monopole antenna and has a total length of 30 mm. Figure 97 shows prototype probe. The shielding box for the board is plastic covered with a Zn-coated copper sheet to prevent RF waves interfering directly with the detection circuit. In the measurements, the device starts monitoring the electric field and storing the maximum electric field after the device has been powered up. The stored maximum electric field is read from the pulse width of the device output. The actual power at the RF detector IC can be derived from the pulse output using the data 96

100 5. Appendix A: Stand-alone Electric-field Probe sheet for the IC. Since the device has no metal/optical wires, it i can be used to measure the electric field inside an enclosure without any gaps. Figure 96. Pattern layout design of prototype printed circuit board Figure 97. Prototype stand-alone electric-field probe First, to test the function of prototype unit, the device was operated in a known electric field. Figure 98 shows the experimental setup. The experiment wass performed in a semi-anechoic chamber whose floor is a conducting ground plane. The frequency of the electric field was increased from to 1.5 GHz in 0.25 GHz steps andd the electricc field strength was varied from 10 to 100 V/m for each frequency. A bi-log antenna located 1 m from the device was used to t apply thee electric field. Both the antenna and the devicee were positioned 1 m above the floor. f The electric field was calibrated and a conventional optical probe (HI-6005) was placed next to the device being tested too measure the actual electric 97

101 5. Appendix A: Stand-alone Electric-field Probe field applied to the device. The antenna for injection was set to apply horizontal polarization perpendicular to the xy plane of the device and the device was set to align y axis to the horizontal polarization (see Figure 98(b)). Anechoic chamber Probe being tested V/m 1 m Calibrated E-field probe 1 m Monitor (a) Antenna and probe location in anechoic chamber antenna case antenna case antenna case y x z y z z x y x (b) Axes of the electric-field probe Figure 98. Setup for testing function of prototype device Figure 99 shows the experimental results for the received power converted from the device output as a function of the electric field strength for various frequencies. The received power increases exponentially with the applied electric field for frequencies up to 1 GHz. However, at frequencies over 1 GHz, the received power obtained from the output pulse width becomes saturated below 70 V/m so that no further variation with applied electric field strength can be observed. This is because the dynamic range of the RF power detector is too low for the electric field strengths used. The results of this experiment demonstrate that the device functions correctly up to a frequency of 1 GHz for electric field strengths up to 100 V/m. The electric field strength working range could be extended by using an additional amplitude adjustment circuit. Next, the directivity of the device was evaluated using the same testing setup as that shown in Figure 98. These tests were performed at 1 GHz. The angle of the antenna used 98

102 5. Appendix A: Stand-alone Electric-field Probe was fixed and the orientation of the device being tested was varied (see Figure 98(b)). Figure 100 shows the evaluation results. The directivity is not isotropic since the antenna printed on the PCB is just a single element of a bent monopole. The sensitivity to RF waves applying to the xz plane is considerably lower than that in the other plane. It will be important to improve the directivity because it directly affects the measurement accuracy for an unknown incident RF wave. It is also important to discriminate the polarization of the incident electric field, which can be achieved by using three antennas for the three axes. Converted received power [dbm] : 1.5 GHz : 1.25 GHz : 1.0 GHz : 0.75 GHz : 0.5 GHz : 0.25 GHz Applied electric field [V/m] Figure 99. Converted received power to the electric field for various frequencies 99

103 5. Appendix A: Stand-alone Electric-field Probe Received power [dbm] Angle [deg] : x-y plane : x-z plane : y-z plane Figure 100. Directivity of prototype probe in different planes D Simulation and calculation A 3D electromagnetic simulation was performed to verify the directivity of the device. Figure 101 shows the model created in CST Microwave Studio. The PCB model consists of only a bent monopole element with a 1 k termination resistor and the planar pattern to simulate the GND plane on a board with no ICs or precise patterning (see Figure 101). This is because the purpose of the simulation is to determine whether the measured directivity of the device is theoretically reasonable. If there is electromagnetic interference or coupling between the RF waves around the device and the PCB circuit, the simulated directivity will differ from the measured directivity. In the simulation, the shielding box for the board has the same dimensions as the experimental one and it is covered with a copper sheet. FR-4 and copper are assigned to the PCB and the pattern (including the antenna element), respectively. The 3D simulation was performed at 1 GHz and the directivity of the device was calculated. An electric field of 100 V/m was applied as a plane wave to the device and the voltage generated across the termination resistor was monitored in the simulation. Figure 102 shows the simulation results in the xy plane. The electric field strength at each angle was normalized by the maximum electric field strength. The simulation results show good agreement with measurements, even at the null angles of 90 and 270. This 100

104 5. Appendix A: Stand-alone Electric-field Probe indicates that the bent monopole antenna with the shieldingg box functions correctly and does not have unwanted coupling to the PCB. In addition, the dependence of the received power on the electric field strength applied to the bent monopole antenna was calculated. For a sinusoidal current distribution, the received voltage Vr from a simple monopole antenna is given by (38) where Z R is the termination resistance, Z A is the antenna impedance, E iss the electricc field strength, L is the antenna length, andd is the wavelength. The antenna impedance Z A was considered to be the same as the inputt impedancee calculated from f the simulation since the antenna loss was assumed to be negligible. The terminationn resistancee is 1 k and the wavelength was calculated by considering the shortening effect due to the dielectric constant r = 4. Figure 103 compares the calculated voltage from the RF power detector IC with the measurement results. r There is good agreement between the measured and theoretical results. Figure D simulation model created in CST Microwave Studio 101

105 5. Appendix A: Stand-alone Electric-field Probe normalized level Angle [deg] z x y Figure 102. Comparison of the simulated directivity (normalized with the maximum electric field strength) with the measurement results Received power [dbm] : Measured : Theory Applied electric field [V/m] Figure 103. Comparison between calculated received power as a function of electric field strength and measurement results 5.3. Electric field measurement inside enclosure The prototype device was used to measure the electric field inside an enclosure. A metal box with dimensions of 120 mm (height) 1 mm (width) 200 mm (length) (see Figure 104) was used for these measurements. The box has a slit that is mm long and 10 mm wide that permits an electric field to be generated within the enclosure. It is used to simulate an actual device that has a slit or a hole for a connector or a harness. A wired probe cannot be used to measure the electric field inside such an enclosure in practical 102

106 5. Appendix A: Stand-alone Electric-field Probe applications because the connector generally blocks the slit or hole. The prototype probe was placed at the center of the bottom plane inside the enclosure (see Figure 104). The electric field was also measured using a calibrated commercial probe (HI-6005) that has an optical cable; it was used as a reference and was located at the same position. A vertically polarized 200 V/m electric field was applied using a horn antenna located 0.9 m from the box surface. The measurement was performed for frequencies from 0.2 to 1.0 GHz in 0.2 GHz steps. Figure 105 compares the measurement results of the received electric field strength (normalized by the maximum electric field strength) obtained using the prototype probe and the calibrated probe. Both measurements have a peak at 0.6 GHz and the electric field strength drops to half at the dip at 0.8 GHz, which suggests good correlation between the measurements by the prototype and conventional probes. These results demonstrate the potential of the prototype stand-alone electric field probe. 200 mm 1 mm 100 mm 100 mm 10 mm mm 120 mm z y x probe 0.9 m Horn antenna y z probe x y z x 75 mm 75 mm Figure 104. Schematic of experiment setup for the measurement of electric field inside enclosure with small slit 103

107 5. Appendix A: Stand-alone Electric-field Probe Normalized electric field strength : HI-6005 : Prototype probe Frequency [GHz] Figure 105. Comparison of experimental electric field measurements inside an enclosure by prototype and conventional probes A stand-alone electric field probe was developed to measure the electric field inside an enclosure for investigating immunity testing at high electric fields. The prototype probe consisted of a bent monopole antenna, an RF power detector IC, and a PIC in a shielding box. Its function was verified by directly applying an electric field to the probe. The measurements confirmed that the probe works up to 1 GHz without saturation by the electric field. The directivity of the probe was also evaluated. 3D electromagnetic simulations and theoretical calculations of the probe were performed to verify its directivity and sensitivity. The simulation and calculation results showed good agreement with the experimental results; there was a maximum error of 6 db for the directivity, which indicates that the probe functions properly. Moreover, the prototype probe was used to measure an electric field at 1 GHz inside an enclosure with a small slit. The measured electric field strength normalized by the maximum electric field strength agreed well with the results obtained by a calibrated commercial probe. The stand-alone probe can measure the electric field inside an enclosure without modifications (such as an opening hole/slit for wire routing). It can thus be utilized to investigate immunity mechanisms. 104

108 6. Appendix B: EM Simulation of On-Glass Antenna with Vehicle Body 6. Appendix B: EM Simulation of On-Glass Antenna with Vehicle Body This chapter describes the results of a study using electromagnetic (EM) simulation for a GPS antenna mounted on the front windshield along with other parts of the vehicle. Due to the rapid growth of electromagnetic simulation technology, the approach based on the commercial simulation software is also necessary for the investigation and the development of products. In this chapter, through electromagnetic simulation such large scale model in high frequency, the study on the capability of the simulation is described Background of simulation GPS and cellular antenna are key components for telematics application which is rapidly growing in automotive industry. In product development for the telematics, antenna location is one of the most important design parameter since the antenna location affects on the antenna performance which is directly related to system performance, and it could affect on the vehicle aesthetic design as well [62]. Usually GPS antenna prefers to be located on the roof of vehicle from the engineering point of view since there is no obstruction for the communication with satellite and the roof works as huge ground plane [63], which means vehicle design e.g. difference of coupe and sedan has less effect on the antenna performance. However, from the viewpoint of vehicle design, it is preferred to locate the antenna inside vehicle due to some reasons such as antenna security or vehicle appearance design etc. [64]. To find better location inside vehicle for radiation pattern of antenna, electromagnetic simulation can be effective alternative way [65], [66]. GPS antenna simulation with vehicle has been studied with variety of respects [67], [68] but still needs to be investigated to clarify what the important parameter is in order to use simulation for total design because the antenna inside vehicle is more sensitive to vehicle parts or vehicle design than roof antenna due to the reason described above. This chapter summarizes the study on GPS antenna simulation with more-vehicle-like car model. In the simulation, mesh GPS antenna which we have reported [69] is used. First the GPS performance with windshield is calculated. Second, the effects of vehicle parts have been analyzed separately. Lastly the whole vehicle simulation results have 105

109 6. Appendix B: EM Simulation of On-Glass Antenna A with Vehicle Body been compared with actual measurement resultss to give thee propriety of the simulation model Simulation model Meshed GPS antenna which is mountable on glass is used in i this simulation. The peak gain direction in radiation pattern of this antenna can be controlled by just changing the mesh pattern without changing the size of antenna, which will w be significant advantage when the mesh GPS antenna is installed on windshield or rear r glass because the tilt of those glass are usually different between vehicles. Fig. 1 shows the antenna structure in simulation model. The dimensions off antenna are 22 mm width, w 39 mmm length, 100 µm thickness and antenna element usess copper in this simulation to be tuned at GPS L1 frequency GHz. In order to obtain the correlation with actual a antenna prototype, the antenna is formed on FR-4 substrate but it is possible to print on windshield directly without PCB depending on the necessity of application. Due to thee mesh design of antenna and the small size of element, it can be more transparent than conventional patch antenna, which can be considered as benefit in termss of vehicle design because it does not interfere with both driver s eyesight and vehicle appearance. Figure 106. GPS antenna simulation model Since the antenna is designed to be placed on windshield, the antenna performance can c be verified only with glass model. First the simulation has been performed with small cut glass as shown in figure 2 and with windshield model to checkk the antenna characteristics then. Frequency range is from 0.8 GHz to 2.2 GHz. The size of cut glass is 1 mm by 100mm. In the windshieldd model, thee antenna is placed at middle m of windshield with 40 mmm distance from top edge of windshield. The simulation results of return loss (S11) and VSWR are shown in figure 3, and radiation patterns for RHCPP gain (dbic) in y-z plane and x-z plane are shown in figure 4 respectively. 106

110 6. Appendix B: EM Simulation of On-Glass Antenna A with Vehicle Body The resonant frequency off antenna with cutglasss 1.64 GHz is i slightly higher than target t GHz but the result with windshield 1.59 GHz is closer. This result suggests that the glass size affects on the frequency characteristics of antenna impedance and the design needs to be optimized with consideration of the size of glass. The T peak gain direction in y-z plane radiation shown in figure f 4 for small glass model around 330 degrees is as designed. In the windshield result, null around 270 degrees and the peak around 240 degrees can c be considered as the effect of wind shield. Figure 107. Left: small cut glass model. Right: windshield model 0 4 Return Loss [db] with just small glas s with windshield Freq quency [GHz] VSWR with just small glass with win ndshield Frequen ncy [GHz] Figure 108. Left: return loss of antenna. Right: VSWR of antenna 107

111 6. Appendix B: EM Simulation of On-Glass Antenna with Vehicle Body z x y x z y with small glass at 1.575GHz with small glass at 1.640GHz with windshield at 1.575GHz 180 Figure 109. Radiation pattern in y-z plane (Left), and in x-z plane (Right) Next, parts of vehicle have been added to the windshield model one by one, including A-pillars, roof, front vehicle body, and chassis with front of dash which is metal part between engine room and passenger room. Those parts model of vehicle are shown in figure 5. Fig. 6 shows result of return loss and VSWR for windshield with A-pillars model comparing to whole vehicle model. The difference between two models is less than 0.1 GHz which is quite small and suggests the frequency characteristics can be designed with only windshield model in this case. Radiation patterns in figure 7 show the effect of A-pillars and roof in comparison to just windshield result. In y-z plane, A-pillars increase the gain between 290 and 360 degrees by up to 6 db, but roof reduces gain in the same direction. Also the roof increases the gain at 30 degrees which is the direction of the roof is located. On the other hand, gain around 100 and 270 degrees in x-z plane for A-pillars and roof model increased. Figure 8 shows results for rest of parts including front of dash with chassis, hood with side body and rear body. Front of dash and chassis have significantly increase null points in the range from 300 to 30 degrees in y-z plane which is quite important direction for GPS performance. These null are considered to be related with the reflection from those parts. While the radiation pattern has been changed with front of dash and chassis, the effect of hood with side body and rear body is quite small. This is because those parts are not close to windshield and located at the position which does not affect on radiation from GPS antenna. 108

112 6. Appendix B: EM Simulation of On-Glass Antenna A with Vehicle Body Figure 110. Vehicle model consists of parts used in i simulation 0 4 Return Loss [db] VSWR Windshield & A pillars Whole Vehicle Frequency [GHz] 2 1 Windshield & A-pillars Whole Vehicle Frequency [GHz] 2 Figure 111. Left: return loss of antenna. Right: VSWR of antenna z x y x z y with windshield 210 with windshield & A-pillars 1 with windshield & A-pillars & roof Figure 112. Radiation pattern inn y-z plan (Left), and inn x-z plane (Right) 109

113 6. Appendix B: EM Simulation of On-Glass Antenna with Vehicle Body z x y x z y Add instrument panel & chassis Add hood and side body with whole vehicle 180 Figure 113. Radiation pattern in y-z plan (Left), and in x-z plane (Right) 6.3. Comparison with measurement Actual antenna measurement sample has been done using sedan vehicle which has very similar shape to the simulation model in figure 5 in order to verify simulation results. GPS antenna is put on the windshield at the same position as simulation which is 40 mm from top as shown in figure 9. The vehicle experiment was performed in an anechoic chamber. Horn antenna was used for measurement and was located 4.2 m away from GPS antenna in vehicle. Radiation pattern measurement results are shown in figure 10 comparing with simulation results. A loss of cable for feeding antenna is considered as 3 db in the simulation results. In the comparison, there is close correlation between simulation and measurement results. For quantitative evaluation, linear averages of RHCP gain in each x-y plane at different elevation angle have been calculated which is shown in figure 11. Simulation results have good correlation with measurement results especially at the elevation angle from 0 to 60 degrees. The difference at higher elevation angle more than 60 could come from uncertainty of measurement or the effect of other vehicle parts which are not considered in this simulation. 110

114 6. Appendix B: EM Simulation of On-Glass Antenna A with Vehicle Body Figure 114. GPS antenna location on windshield used for measurement z x 60 y z x y Simulation Measuremen nt Figure 115. Left: radiation pattern in y-z plan. Right: radiation patternn in x-z plane Linear Average Gain [dbic] Measurement Simulation Elevation Angle [deg] Figure 116. Elevation angle dependence of linear average gain for f measurement result and simulation result in x-y planee 111