Embed Size (px)
Citation preview
DOCTORAL THESIS
Department of Computer Science and Electrical Engineering Division of EISLAB
2004:07 • ISSN: 1402 - 1544 • ISRN: LTU - DT - - 04/07 - - SE
2004:07
Characterization of Components and Materials for EMC Barriers
Urban Lundgren
Characterization of components andmaterials for EMC barriers
Urban Lundgren
EISLABDept. of Computer Science and Electrical Engineering
Lulea University of TechnologyLulea, Sweden
Supervisors:
Professor Jerker Delsing and Professor Dag Bjorklof
ii
To my dear wife Karinand to my parents
Erland and Margaretha Lundgren
iv
Abstract
This thesis presents contributions to work for better methodologies for addressing Elec-tromagnetic Compatibility (EMC) issues. In particular measurement methods are re-viewed and devised for acquiring data on barriers used for EMC. Such data is used forcharacterization, modeling and model verification of barriers.
The concept of EMC barriers is introduced as a general view of filter components,separation of conductors (crosstalk problems), electromagnetic shielding etc. The aimis to find methodologies to help engineers to identify EMC problems and to include themanagement of EMC in the design of a electrical circuit in a practical and effectivemanner.
Methodologies for generation of EMC barrier modeling techniques have been devel-oped. This work have resulted in design tools for electronic design engineers to includeEMC considerations at an early design stage of a new product.
Problems with existing barrier characterizing measurement methods have been identi-fied. By comparison of far field and near field shielding effectiveness measurement meth-ods, data for shielding thermoplastic materials was acquired. Considering the purposeof studied shielding materials in an application the usefulness of the far field shieldingeffectiveness measurement method is questioned.
EMC barrier measurement methodologies of interest in this thesis includes shield-ing effectiveness measurements, transfer impedance measurements, scattering parametermeasurements, measurements of material permittivity and permeability and near fieldscanning techniques for analysis of current distributions.
v
vi
Contents
Chapter 1 - Thesis Introduction 1
1.1 The Need for Electromagnetic Compatibility in the Information Technol-ogy Society . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 EMC - a real and a legal requirement on electronics . . . . . . . . . . . . 21.3 A Motivation to Work on EMC Barriers . . . . . . . . . . . . . . . . . . 4
Chapter 2 - EMC Barriers 5
2.1 Interpretation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52.2 Types of EMC Barriers . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52.3 Characterization of EMC Barriers . . . . . . . . . . . . . . . . . . . . . . 82.4 EMC Barrier Modeling . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
Chapter 3 - Measurement Methods Essential to Work on Barriers 13
3.1 Introduction to measurement methodologies . . . . . . . . . . . . . . . . 133.2 Shielding Effectiveness Measurement Methods . . . . . . . . . . . . . . . 143.3 Transfer Impedance Methods . . . . . . . . . . . . . . . . . . . . . . . . 183.4 Characterization of Permittivity and Permeability . . . . . . . . . . . . . 203.5 Near Field Scanning of Printed Circuit Boards and Conductive Surfaces . 22
Chapter 4 - Thesis Summary 23
4.1 Summary of Contributions . . . . . . . . . . . . . . . . . . . . . . . . . . 23
Chapter 5 - Conclusions - Suggested Further Work 31
References 33
Paper A 43
Paper B 51
Paper C 59
Paper D 67
Paper E 75
Paper F 83
Paper G 91
viii
Preface
This thesis is a summary of the research I have been participating in at Lulea Universityof Technology since 1996, when a professorship in EMC Technology was inaugurated atthe University and the present EMC Laboratory was built.
It was not a coincidence that the EMC directive became mandatory in Sweden in1996, making EMC a hot topic for all manufacturers of electronic products. Today LuleaEMC Center is a competence center for education, R&D and a support center for SMEsin the region.
In 1996 I started on my master thesis project studying possible EMC problems relatedto the use of high speed modems (VDSL) connected to overhead telecommunication lines.That was my first encounter with the concept of EMC and I really found a lot of newinteresting experiences. The new knowledge I acquired during the master thesis projectgave me a desire to find out more. This experience was making me start thinking thatEM compatibility is something I want to work with.
After receiving the master of science degree in 1997 I came to a decision based onmany circumstances at that time, to consider staying at the university as a PhD student.The occupation of people near me and discussions with friends doing PhD studies, mademe take the step towards a PhD degree.
I want to thank my supervisors, Professor Jerker Delsing and Professor Dag Bjorkloffor their guidance and kind support. I want to thank all my friends and colleagues atLulea University of Technology and especially Ake and Jonas for assisting me and sharingtheir experiences.
I also want to thank my wife, Karin Lundgren for helping me finding out what reallycounts and my family for their support.
ix
x
Part I
xii
Chapter 1
Thesis Introduction
1.1 The Need for Electromagnetic Compatibility in
the Information Technology Society
In later years embedded systems and smart sensors have been of increasing interest formachine to machine (sometimes abbreviated M2M) communication and automation ofservices. In such devices it is sometimes desired to use wireless communication for transferof control and data.
People are experiencing that banking services and a variety of E-commerce opportu-nities exists through the use of modern information technology. Embedded systems alsoincreases the standard of living by offering a higher level of functionality to the home,workplace and to the environment we live in.
In literature the future of digital services and media is expected to have a greatimpact on the society [65, 2]. As a consequence this will make it hard or even fatalfor companies in many business areas to neglect the changing situation. When digitalinformation becomes more important, and more and more of products and services oftoday becomes digitalized, digital communication between machines is essential.
For the access to digital services networked computers are used. To reduce the effortmade and improve the availability to gain access to digital services, mobile terminals suchas mobile phones and handheld computers (PDAs) can be used. Such mobile terminalstypically use wireless communication to maintain mobility while communication can takeplace.
The use of robust wireless communication can give an advantage in electromagneticcompatibility (EMC) issues if this implicate a reduction of cabling and galvanic intercon-nections in a system. For a system consisting of many parts with a substantial amountof interconnections, electrical currents in ground conductors and cable shields can causeelectromagnetic interference problems.
However when the density of devices equipped with wireless communication capabili-ties increases, it is also important to realize that EMC considerations becomes central. Adevice with wireless communication capabilities is a transceiver of electromagnetic energybut may at the same time affect other radio receivers and be subject to electromagnetic
1
2 Introduction
interference (EMI) from other devices emitting electromagnetic energy.To assure a high quality of communication and reliable operation while using a digital
service the electronic devices involved must be adapted to the electromagnetic environ-ment where they will be used. More advanced services such as video communication willrequire terminals with high performance (high clock frequency).
This scenario shows a number of issues that is known to influence on the requirementsfor EMC.
• Wireless communication devices emits electromagnetic energy that may cause EMI
• High clock frequencies in electronic circuits requires better knowledge of EMC todesign devices with reliable functionality
• high level of integration of electronic circuits is used in handheld computers andmobile phones, this may cause sensitivty to EMI
• plastic encapsulation of electronics offers limited shielding performance and there-fore increases the risk of EMI
• larger number of electronic devices is expected and increases the risk of EMI
To manage the situation with an increasing density of mobile electronic devices wheremany are equipped with wireless communication technology, EMC issues must be con-sidered at at planning stage. Using electromagnetic simulation techniques to identifysituations where EMC problems may occur, the EMI threats can be reduced by EMCbarriers in the design of electronic devices.
There is a desire to come to better understanding how to make small electronic devicesthat uses wireless communication but is protected from EMI. It is also an advantage ifthis can be combined with an encapsulation technique that offers robust mechanical andweatherproof capabilities.
As an example application where hard conditions must be met is electronic devicesfor identification and monitoring of cargo containers. A module is attached to a cargocontainer and by wireless communication it reports its whereabout and the status of thecargo to the owner.
This kind of automated services will bring forward a big demand on reliable wirelessdevices that must face hard condition including the threat of EMI.
1.2 EMC - a real and a legal requirement on elec-
tronics
The necessity of electromagnetic compatibility (EMC) is not that obvious for peoplewith no experience from electronic design. It is however necessary for modern electronicdevices to be designed with electromagnetic compatibility issues in mind to fulfill
all safety and protection requirements and offer the functionality and quality expectedfrom the device. An apparatus both emits and receives electromagnetic energy conducted
1.2. EMC - a real and a legal requirement on electronics 3
on attached cables or radiated from enclosures and cables. That is just a side effect whena electronic device is offering some functionality and can not be completely avoided.Because of this undesired side effect there are legislated demands on the device regardingits electromagnetic properties.
The European EMC directive (directive 89/336/EEC [21]) states under article 4 thatan electrical apparatus (covered by the directive) should be constructed so that:
• the electromagnetic disturbance it generates does not exceed a level allowing radioand telecommunications equipment and other apparatus to operate as intended
• the apparatus has an adequate level of intrinsic immunity of electromagnetic dis-turbance to enable it to operate as intended.
Which means that a manufacturer of a apparatus covered by the directive must designthe apparatus and specify how it should be used so that it does not cause electromag-netic disturbance harming other devices that do comply with the directive. Also, theapparatus must be designed to be immune to a normal level of electromagnetic energyin the environment it is intended to be used.
These statements must be respected by all manufacturers of electric and electronicequipment to be placed on the internal European market or they may experience legalactions from market surveillance authorities. The way the directive is written (calledthe new approach) gives the manufacturer responsibility of all products placed on themarket even if a product is approved by a third party, e.g. an accredited test lab or acompetent/notified body. The European EMC directive which was introduced in 1992makes it necessary for manufacturers with intent to put their products on the Europeanmarket to address EMC issues.
In the United States the FCC regulations (FCC part 15 [24]) states limits on themaximum acceptable levels (section 15.109) of emitted electromagnetic disturbance froman apparatus working with signals in radio frequency (RF) range. However there areno general regulations addressing a required level of a immunity against electromagneticdisturbance as the case is in Europe. This issue is in the United States left for the marketitself to take care of, except for certain medical technical devices which are regulated byFDA.
The EMC regulations are an important step since they are helping people to realizethat it is necessary to address electromagnetic compatibility problems. The problemshave to be solved before products involving electronic circuits can be put on the market[9, 10].
Because of the increasing frequencies for the electromagnetic energy due to the useof high clock frequencies and fast digital logic circuits in modern electronic designs, theEMC requirements gets even harder to meet. It is necessary for the electronic designerto incorporate EMC into the device at an early stage of the design process. If the EMCissues are neglected until a working prototype is ready, the cost for fixing eventual EMCproblems and non-complience may be severe. It is sometimes necessary to start over,designing the device from scratch again. A good approach to reach EMC is usually touse different kinds of electromagnetic barriers such as filters and shielding enclosures with
4 Introduction
conductive gaskets, or just a well planned layout of the circuit. It is much easier andmuch less expensive to meet the regulations if the apparatus is well designed regardingEMC aspects to start with, than trying to fix an insufficient design.
1.3 A Motivation to Work on EMC Barriers
It is well understood by most electronic design engineers that it is important to meetrequirements on EMC. However the lack of easy to use design tools to address EMCproblems often makes EMC fixing necessary at a late stage in the design process. Betterknowledge of EMC barriers as components and design blocks can simplify EMC consid-erations at an early design stage.
This work aims in general to a better understanding of EMC barriers. From such un-derstanding we can devise improved design and verification methodologies. Thus cuttingtime to market and development costs.
My focus have been on:
• measurement methodologies evaluating barriers
• measurement methodologies characterizing barriers
• measurement methodologies to produce barrier modeling input data
• measurement methodologies for verification and validation of barrier modeling andcomputer simulations
• measurement methodologies for barrier material data such as permittivity and per-meability
• material data
Investigated barriers include filters, shielded cables, shielded connectors and shieldingmaterials for use in electronic device enclosures.
Following is a short review of barrier concepts and methods to model and experimen-tally characterize barriers.
Chapter 2
EMC Barriers
2.1 Interpretation
While designing an electronic circuit the engineer is trying to interconnect integratedcircuits or different kinds of discrete components so that the components and integratedcircuits interact to give the desired functionality. Designing circuits that make use ofhigh frequency electromagnetic signals puts extra demands on the interconnections.
A good approach when designing electronic circuits is to use an electromagnetic topol-ogy approach by separating the EM environment into zones with different signal qualityor different signal amplitudes.
The boundaries between the zones represents barriers and reduces the influence ofcharges and charge movements in one interconnection path on charges in another in-terconnection path. Such barriers often needs to be frequency selective so that desiredfunction of interconnections can be maintained. The electromagnetic barriers thereby of-fers a certain attenuation of electromagnetic energy that reduces the threat of electronicequipment in one zone interfering with equipment in another zone.
Barriers can be divided into physical barriers and geometrical barriers. The physi-cal barrier category contains standard construction elements like filters, where internalcomponent values and geometrical design usually are unknown. The characteristics forthese barriers can be found on datasheets from the manufacturer or have to be obtainedby measurements.
The geometrical barriers can be characterized completely by the geometrical designand material properties, here of course the exact material properties can sometimes behard to find. Examples of this kind is for instance enclosures made of homogeneous con-ductive materials like metals and plastic materials, printed circuit board trace separationand separation between pins in a multiconductor connector.
2.2 Types of EMC Barriers
Electromagnetic barriers used in EMC problem solutions may take many different forms.A short overview will be given with examples on barriers and how they are used. The
5
6 EMC Barriers
following examples are considered:
• Physical or generalized screens (Shields)
• Conductive gaskets
• Shielded cables and connectors
• AC line filters
• Feed through filters
• Conductor spacing/routing
To reduce radiated interference or radiated emissions, a barrier in the form of aconductive enclosure can be used as a shield, enclosing the victim or the radiating source.The design of a shielding enclosure can be a tricky task since the enclosure often playmany roles and must fulfill requirements other than those on electromagnetic shielding.When designing shielding enclosures thus compromises must be made. The design shouldoffer:
• Low production cost for enclosure
• Customer appealing and user friendly product
• Ventilation or cooling for interior parts
• Communication with user and/or other devices
• Mechanical protection for interior parts
• Moisture and dust seal
• Electromagnetic shielding
• Care must be taken when choosing materials and production techniques so that thenecessary enclosure properties lasts for the specified lifetime of the product
There are of course many techniques available in each aspect of the enclosure designproblem. Considering the electromagnetic shielding aspect, the enclosure material canbe metallic, nonconductive covered with a conductive layer or conductive composite orpolymer. The choice of material influences on the possible techniques available for meet-ing the overall specification on an enclosure. It is important that a good compromisecan be reached between material and production cost, and enclosure design regardingshielding performance and user friendliness.
The electromagnetic shielding offered by a shielding material used in an enclosure isa combination of shielding effects where each can be physically explained by theory. Forelectromagnetic waves the electric field component E and the magnetic field component
2.2. Types of EMC Barriers 7
H(f)
I(f)
Figure 2.1: As an example a shielding box of high conductivity material can offer good shieldingeven for time varying magnetic fields H(f) if induced currents I(f) are free to flow all overthe surface. An aperture forces currents to a longer path around the aperture which causes avoltage potential difference across the aperture that can couple into the shielding box
H are orthogonal and related by a wave impedance Z depending on the medium wherethe wave is propagating.
Z =E
H
In free space the wave impedance is constant and equal to 377Ω. This is the case whenfar field shielding is referred to and then the shielding effects can be described by thefollowing losses:
• Absorption losses due to ohmic losses in shielding material.
• Reflection losses due to impedance change for propagating EM wave crossing mediumboundary.
For near field shielding (distance from source less than λ2π
) the electric field andthe magnetic field components must be considered separately and can be treated usingdifferent shielding approaches. The field type and frequency decides which materials andtechniques to use to achieve desired shielding performance:
• Faraday’s cage shielding, high conductivity material, effective for electric fields
• Low-reluctance path shielding, high permeability material, effective for magneticfields
• Induced eddy current shielding, high conductivity material, effective for high fre-quency magnetic fields
For an enclosure where a high level of electromagnetic shielding is desired, a metallicmaterial is the best choice since metals offers the best conductivity. The enclosure willthereby offer good shielding effectiveness for both electric and magnetic field at reasonablehigh frequencies.
The desired lifecycle of shielding enclosure is a main factor when the designer ischoosing the materials to be used. If metallic materials are used to make an enclosure
8 EMC Barriers
offering a high level of shielding it may be necessary to include bolts, rivets and gasketsin the enclosure design to ensure high conductivity between the parts. High conductivitypaths between the parts are necessary for the shielding performance, see Figure 2.1,but is also a problem when considering the life cycle of the shielding enclosure. Highconductivity paths between different metals can accelerate corrosion and can also causenoise in sensitive circuits [54].
Filters are used for separating zones in electronics with different electromagnetic envi-ronment (EME). On a printed circuit board ground planning is necessary when differentparts of the circuit have different signal quality for instance analog ground and digitalground. When a signal trace is crossing a boundary between electromagnetic zones sur-face mounted filters can be used to keep the integrity of the boundary and the signalin the more sensitive circuit. It is also wise to make sure that the board is routed togive adequate separation between traces especially for traces with analog signals withlow signal levels and traces that is crossing zone boarders.
2.3 Characterization of EMC Barriers
It is desired to find ways to describe barriers so that the efficiency of different barriersolutions can be compared. That enables manufacturers of filters, gaskets and otherbarrier components to improve component designs and compare material choices. Itwould also be a great advantage if circuit simulations were possible with the barriercomponents included. That would give the electronic design engineer tools to evaluatethe EMC performance of a design in an early design stage. In order to make such barrierdescriptions, techniques to characterize the electromagnetic properties of the barriers areneeded.
A barrier is often described in terms of:
• scattering parameters
• shielding effectiveness
– insertion loss
– attenuation
• circuit elements
– transfer impedance and transfer admittance
– lumped circuit network
2.3.1 Scattering parameters
Scattering parameters is a commonly used tool in RF and microwave design. The scat-tering parameters (s-parameters) represents the reflection coefficients and transmissioncoefficients for a two port circuit. The two involved reflection coefficients are denoted S11
and S22. The two transmission coefficients are denoted S12 and S21.
2.3. Characterization of EMC Barriers 9
aa22
aa11
bb11
bb22
SS22
SS21
SS12
SS11Port 1 Port 2
S11=b1a1
S21=b2a1
S12=b1a2
S22=b2
a2a2= 0 a2= 0 a1= 0 a1= 0
Figure 2.2: Using scattering parameters to describe signal paths in a two-port
Scattering parameters can be explained by studying a passive two-port shown in figure2.2. If a electromagnetic wave enters the two-port via a connected transmission line, thewave is scattered and both transmitted trough the two-port and reflected back towardsthe source of the wave. The relationships between the incident wave, the transmittedwave and the reflected wave is represented by the s-parameters.
For a forward direction of wave propagation the incident wave is denoted a1, thetransmitted wave b2 and the reflected wave b1. For a backward direction of wave propa-gation the incident wave is denoted a2, the transmitted wave b1 and the reflected wave b2.These representations of waves are actually complex entities and so are the correspond-ing s-parameters. A complex s-parameter holds information of the magnitude ratio andphase angle difference between the two waves that is considered.
[b1
b2
]=
[S11 S12
S21 S22
] [a1
a2
](2.1)
If a complete four s-parameter description is available it is often presented in ma-trix form, equation (2.1). By transformation of the s-parameter matrix into the whatis called the chain-matrix the effect of cascading networks can easily be calculated bymatrix multiplication. Generalizations of the s-parameter description can be made forcharacterization of networks with three ports or more [41, 50].
2.3.2 Shielding Effectiveness
Shielding enclosure materials and conductive gasket performance can be specified bymeans of shielding effectiveness that is obtained by an insertion loss measurement. Theinsertion loss is the difference between attenuation measurements with and without theobject present in a fixture.
For insertion loss measurements the attenuation of the test object equals the inverseof the magnitude of the scattering parameter representing transmission, S21 (or S12).
Insertion loss measurements are usually made with far field shielding conditions usingtransmitting and receiving antennas or in transmission line fixtures.
10 EMC Barriers
2.3.3 Circuit Elements
Barriers characteristics can also be described by a circuit schematic. The circuit elementsand their values describing the barrier is sometimes easily obtained from datasheets,sometimes they are possible to extract from the geometry or by measurements. It issometimes a great advantage to use the circuit description because of the possibilityto include the barrier in a circuit simulation of the entire system. The effect of thebarrier then can be studied under correct drive and load conditions in time domain or infrequency domain [14, 46].
Shielding performance is sometimes given in form of the transfer impedance. This istypically the case for shielded cables, shielded connectors and conductive gaskets but itcan also be used for shielded enclosures in general. To completely describe the shieldingeffect on electromagnetic fields the transfer impedance description is not enough, there isalso a transfer admittance that can be causing leakage through an electromagnetic shield.
The transfer impedance Zt represents two coupling mechanisms namely the diffusionthrough the thickness of the shield and the magnetic field coupling through imperfectionsin the shield [49], see left part of figure 2.3. It can be obtained from measurements as theratio of a potential difference across an electric field Eout on the secondary side of a shielddue to a current density Jin on the primary side, see equation (2.2). The current densityused in this context is the current per unit length of a gasketed seam or per width of ashielding surface rather than the current per area commonly used in other contexts. Thecurrent density is therefore given the dimension of [A/m] rather than [A/m2].
Zt =Eout
Jin
(2.2)
Yt =Jout
Ein
(2.3)
Transfer admittance Yt that represents electric field coupling through imperfectionsin the shield, [49], is instead the ratio of the current density Jout on the secondary sideof the shield by the potential difference across the primary side caused by an incidentelectric field Jin, see equation (2.3). This is shown in the right part of figure 2.3.
However the effect of the transfer impedance on the shield leakage is often dominantover the transfer admittance which thereby often can be neglected [49, 55]. The transferimpedance and transfer admittance are usually considered as being a distributed ho-mogeneously over the length of a gasketed seam or similarly for other components. Itis therefore practical to use the transfer impedance/admittance description only at fre-quencies where the wavelength is much greater than the size of the shielding componentto be characterized.
Using transfer impedance fixtures, shielding components can be characterized by mea-surements with good repeatability [58] from really low frequencies (dc) up to a frequencydependent on the size of the fixture. Transfer impedance measurement results character-izing conductive gaskets at frequencies up to 10 GHz have been published[37].
2.4. EMC Barrier Modeling 11
Primary side
Secondary side
Primary side
Secondary side
Transfer impedance Transfer admittance
Electric field, EEin
Electric field, EEout
Current density, JJout
Current density, JJin
Figure 2.3: To the left the two coupling mechanisms represented by transfer impedance is shown.To the right the coupling mechanism represented by transfer admittance is shown.
2.3.4 Other measures to characterize and evaluate EMC barri-ers
Material electromagnetic properties such as complex permittivity and complex permit-tivity are another description from which an EMC barrier can be characterized. Usingtraditional absorbtion loss and reflection loss calculations the shielding effectiveness of amaterial can be estimated under certain assumptions [48].
For dielectric materials included in EMC barriers, the knowledge of the materialcomplex permittivity and complex permittivity may be necessary for accurate modelingof those barriers. Barrier models can then by computer simulation give an estimation onshielding effectiveness etc.
For some EMC barriers an approach for characterization is to study the currentdistribution that results from some excitation. It may for instance be of interest tostudy the current distribution over a gasketed seam in a shielding enclosure applicationto understand the effect of different conductive gaskets.
In the layout of printed circuit boards (PCBs) the effect of separated ground segmentscan be analyzed by studying the ground layer current distribution and can adjustmentscan be made to solve signal integrity issues. In addition electromagnetic radiation froma conductive surface can be estimated if the surface current distribution is known.
On approach to investigate a current distribution by measurement is by near fieldscanning of magnetic field component over a conductive surface and then estimate thecorresponding surface currents.
2.4 EMC Barrier Modeling
The need to address EMC issues at an early stage of product development is drivingresearch on computer modeling techniques to analyze design solutions. Replacing EMCmeasurements on product prototypes with computer simulations also makes it possibleto try varied designs at a lower cost. For development of numerical modeling techniquesfor use in computer simulations, comparisons with measured data and analytical models
12 EMC Barriers
are essential for verification and validation.Depending on the complexity of an EMC barrier its characterization may be obtain-
able by modeling the structure of the barriers and applying a numerical electromagneticfield solver that can extract the s-parameters or circuit component values from the ge-ometry and material data given in the model. This approach is of course desired and itsometimes gives opportunities to study internal characteristics that hardly can be studiedby other means.
To provide barrier component- and material data for EMC barrier modeling it isoften necessary to characterize physical samples by measurements. This is particularlythe case for components with unknown internal geometries and when new materials withunknown electromagnetic properties are involved. For this kind of measurements aninstrument known as a vector network analyzer is used because of its ability to collecta full s-parameter description of a studied two-port. It is often necessary to design acustom made fixture for this measurement depending on the design and shape of thebarrier and the frequency range that is considered for the measurement.
Earlier research in this field have investigated discrete component models for improv-ing the high frequency accuracy of circuit simulations [70]. By including stray capaci-tances and pin inductances realistic simulation results were obtained at frequencies upto 1 GHz for resistors, capacitors and coils.
More recent research have diversified and is covering a wide range of aspects. Numeri-cal modeling techniques for extracting lumped component values or transmission line pa-rameters from geometrical structures has been an expanding research area [60, 43, 69, 5].Finite Difference Time Domain (FDTD), Partial Equivalent Element Circuit (PEEC)and Method of Moments (MOM) are the most commonly used numerical techniques [5].
Generation of a lumped circuit model followed by a circuit simulation using SPICEor equivalent software tools is an approach many have chosen to use [60, 18]. For circuitsimulations of crosstalk in multi conductor cables and on printed circuit boards modelsbased on transmission lines and multi conductor transmission lines are commonly used[43, 69]. Earlier work usually focuses on a specific barrier or design element and do notdevelop generic approaches to generate lumped element circuit models.
Chapter 3
Measurement Methods Essential toWork on Barriers
3.1 Introduction to measurement methodologies
The concept of EMC barriers is a general view of techniques to reduce interference be-tween electronic devices or interference within a device (signal integrity). Since a barriercan be in the form of an electromagnetic shielding enclosure as well as a filter com-ponent, different measurement methods must be used for characterization of differentbarrier types.
• Shielding effectiveness (SE) measurements are typically used for characterizationof shielding materials and components of a shield such as conductive gaskets.
• Scattering parameters (s-parameters) measurements are used for characterizationof filters and can also be used to study signal integrity and crosstalk problems.
• Transfer impedance (Zt) measurements are typically used for characterization ofshielded cables, shielded conductors and conductive gaskets.
• Permittivity and permeability measurements are used for characterization of ma-terials used in barriers. This data can then be used in barrier modeling.
• Near field scanning measurements may be used for analysis of current distributionin a conductive surface. For instance divided ground planes can be analyzed.
Development of computer simulation techniques needs verification and validation toprove the technique to be accurate when solving real design problems. This is usuallyaccomplished by comparisons of simulation results with analytical models and with mea-surements on prototypes. Different measurements methods are used for this purpose tocompare an observable parameter with the corresponding parameter from a simulation.
13
14 Measurement Methods
3.2 Shielding Effectiveness Measurement Methods
The methods to measure the shielding effectiveness of an enclosure or a conductive gasketor other shielding components includes a variety of standard methods and variation onstandard methods. A performance value can be obtained using radiated measurementsmethods, transmission line fixture methods and so on. Each measurement method is pro-ducing a performance value that is hard to relate to the performance value obtained usinganother measurement method. For example a measurement method where a shieldingmaterial sample is mounted inside a coaxial transmission line, will not necessarily exposethe sample for the same electromagnetic field impedance as a method where an incidentplane wave is radiated from an antenna [68].
There exists standardized measurement methods for obtaining a performance valuefor shielding enclosures and shielding enclosure components such as conductive gaskets.However, these standardized methods can not always be used because they need welldefined samples that is sometimes impossible to make due to the properties of the objectto be measured. In these cases it is necessary to design measurement equipment thatdiffers from standardized equipment. The custom made measurement equipment can betailored to be used for measurements on specific gaskets or sheets of material with aparticular shape and size.
Unfortunately it is very hard to relate the data obtained with a tailored fixture tocorresponding data as obtained according to standards and thereby the effort is notalways made. Another reason why this relationship is seldom established is the factthat it is well known that some standardized measurement techniques struggles withrepeatability problems, for example [68].
3.2.1 MIL-STD-285 Type Methods
The foundation of shielding effectiveness measurements has earlier been the Americanmilitary standard MIL-STD-285 from 1956 [51]. The standard is now withdrawn butimproved methods of this type have made new standards evolve. The method uses twoscreened rooms with one common wall as shown in figure 3.1. This wall has an aperturewhere test objects are mounted.
In one room is the transmitter antenna located and the receiver antenna is locatedin the other room, the antennas are directed towards each other and at a fix distancefrom each other. The transmitter is transmitting at constant power and the receivermeasure the transferred power with and without test object mounted in the aperture.The difference between these measurements is the insertion loss (IL) for the test object.Measurements according to MIL-STD-285 have been used to examine new shielding ma-terials as conductive composites and performance of conductive gaskets. Drawbacks withthe method are that measured insertion loss is dependent on the antenna placement andthe reflections of the electromagnetic wave inside the screened rooms. This makes therepeatability poor.
Improved versions of the method in MIL-STD-285 have developed were the problemswith reflections have been minimized by the use of absorbing material in the chambers
3.2. Shielding Effectiveness Measurement Methods 15
Test object
Transmit antenna Receive antenna
Figure 3.1: Shielding effectiveness measurement according to the standard MIL-STD-285
[11]. New improved version of the method in MIL-STD-285 can also be found in thestandard IEEE-STD-299 from 1997 [33]. Frequency range a few MHz to 18 GHz [58].
3.2.2 Dual Mode Stirred Chamber
A measurement method called the dual mode stirred chamber or dual reverberationchamber method is being developed by FOI, the Swedish Defence Research Agency inLinkoping, Sweden, National Institute of Standards and Technology (NIST) in the USAand others [58, 56, 28, 26, 67, 57]. It uses two mode stirred chambers next to each otherwith a common wall having an aperture for mounting of test objects.
A mode stirred chamber is a shielded room with a mode stirring arrangement tocreate electromagnetic fields with a large number of modes within the chamber, seefigure 3.2. The mode stirring arrangement can be a mechanical paddle wheel or someother highly conductive structure that can be rotated stepwise or continuously. It canalso be implemented by modulating the signal source and create frequency mode stirring[63], also referred to as electronic mode stirring.
The multimode electromagnetic field in the chambers can be of high amplitude usinga small not to expensive power amplifier. A 1 Watt amplifier can be used to create afield in the chamber with electric components of 100 Volts per meter or more, dependingon the quality factor (Q) of the chamber [63].
The facility can be used for mode tuning when the stirring mechanism is stepped be-tween measurements so that several measurements are obtained in different electromag-netic environments. A mode stirring action can also be used when the stirring mechanismis causing a continuously changing electromagnetic environment in the chamber.
This makes the measurements considerably easier, the placement of antennas is forinstance not critical here. Drawbacks are that the needed equipment to control the modestirring can be expensive and the method does not show if the test object is particularlysensitive to a certain field polarity from a certain direction. The lower frequency limit forthe method is about 500 MHz and it is dependent on the smallest size of the mode stirredchambers (the smallest distance inside a chamber should be seven times the wavelength of
16 Measurement Methods
Test objectTransmitantenna
Receive antenna
Stirring paddleStirring paddle
Figure 3.2: Shielding effectiveness measurement using a dual mode stirred chamber facility
the lowest frequency [56], i.e. 500 MHz gives a distance of more than four meters). Otherways to derive the lowest useful frequency [31, 15, 17] gives different lower frequency limitand to really know the lower frequency limit, measurement of the field uniformity [31] inthe completed chamber is the best way to get an exact realistic value [29].
3.2.3 Apertured TEM Cell in a Reverberating Chamber
This is a simplified method [58, 67] that makes use of one mode stirred chamber. Insidethe chamber a transmitting antenna is located and an apertured TEM cell as receiver[58, 40]. The location of the transmitting antenna and the TEM cell is not critical butthe antenna should not be aiming towards the TEM cell and the TEM cell should not beclose to the walls or other reflecting objects [31, 15]. A TEM cell is an expanded sectionof a rectangular co-axial transmission line. The sample is mounted over an aperturein the TEM cell. An electromagnetic field is then generated with the antenna and areceiving instrument connected to the TEM cell is used to measure the leakage throughthe sample. The usable frequency range is 200 MHz to 1 GHz and dynamic range about100 dB [58].
3.2.4 Dual TEM Cell
A related and even cheaper method is the dual TEM cell method [58, 67, 53, 35]. Heretwo TEM cells are connected together in a ”piggy-back” manner, see figure 3.3. TheTEM cells are coupled trough an aperture in the common wall. An important featurewith this measurement method is that near field SE measurements are obtained (E-fieldand H-field shielding effectiveness).
One TEM cell is connected to a signal source and terminated in the other end. Thesecond TEM cell have two outputs where electric field coupling and magnetic field cou-pling can be measured respectively. The aperture is covered with the sample to beinvestigated. A drawback with this method is that the polarization of the electric fieldis normal to the sample [58]. A nice feature with this method is that both electric- and
3.2. Shielding Effectiveness Measurement Methods 17
Termination
Far end measurementNear end measurement
Signalsource
Test sample
Figure 3.3: Shielding effectiveness measurement using a dual transverse electromagnetic (TEM)cell
magnetic field shielding simultaneously can be investigated by measuring the receivedpower in both ends of the secondary TEM cell. The frequency range for this method is1 MHz to 200 MHz and the dynamic range is about 60 dB.
3.2.5 Split TEM Cell (TEM-T Cell)
Another usage of the TEM cell is the split TEM cell [58, 35, 27], it is also called rectan-gular split transmission line holder. The split TEM cell is made as an ordinary TEM cellin two halves. A sample is characterized by the insertion loss calculated from a measure-ment of attenuation through the empty cell with the halves joined and a measurement ofattenuation through the shorted cell with the halves joined with the sample in between.Both the center conductor and the outer conductor must make good contact with thesample on both sides so that the cell is shorted by the sample.
The receiving half of the TEM cell can be modified to measure the magnetic fieldshielding efficiency. Then a loop antenna is combined with a box equipped with a 90-degree angle reflector on one wall. The loop antenna is mounted trough the reflectorsuch that three quarters of the loop is inside the box and one quarter outside. Whenperforming a measurement the wall with the reflector and quarter loop antenna is joinedwith the half TEM cell and the sample in between. The frequency range for this methodis 1MHz to 1 GHz (1 MHz to 400 MHz for H-field) and the dynamic range about 70-80dB[58].
3.2.6 Circular Coaxial Holder
There are two quite different versions of this kind of test fixture [27, 58]. The continuousconductor (cc) version is an expanded 50Ω coaxial transmission line with tapered ends tofit standard 50Ω coaxial connector. The sample has to have an annular washer shape tofit between the inner and outer conductor and thereby short the transmission line. Thecontinuous conductor test fixture has an operating frequency range of dc to 1 GHz anda dynamic range of 90 - 100 dB [58].
18 Measurement Methods
The other version called split conductor (sc) is described in the standard ASTM-D4935 and has a similar design but is split in two halves. This simplifies mounting oftest samples. Test samples are needed both for the reference measurement and for theactual measurement of the material.
For the reference measurement a small disc shaped piece is fitted between centerconductors and a large washer shaped piece is fitted between outer conductors. In theactual measurement a large disc shaped piece is fitted between the two halves of thefixture. This method keeps distance and material between the two parts of the fixtureconstant except for the region between center and outer conductor. The outer conductoris equipped with flanges to offer good capacitive coupling between the two halves of thefixture.
The insertion loss is calculated as the difference between the measurement of the largedisc sample and the reference measurement. The split conductor fixture has a frequencyrange of 1 MHz to 1.8 GHz and a dynamic range of 90 - 100 dB [58].
3.2.7 Dual Chamber Test Fixture (ASTM ES7-83)
A box split into two sections. Each section has an antenna fixed on the inside. A sheet ofthe sample material is sandwiched between the two sections and the transmission throughthe material is measured. As a reference the transmission between the antennas aremeasured without the sample present. From these measurements the shielding efficiencyis calculated as the insertion loss. The test method has been used for frequencies from100 kHz to 1 Ghz and gives a dynamic range of 80 dB.
3.3 Transfer Impedance Methods
The transfer impedance or surface transfer impedance is a way to describe the highfrequency characteristics of an electromagnetic shield in terms of lumped or distributedcircuit elements. Properties influencing the performance of an electromagnetic shieldare the skin depth, geometrical shape among others. By determination of the transferimpedance these properties are modelled by a circuit element giving the correspondingelectric field on the secondary shield surface for a certain current on the primary shieldsurface. This model are then very suitable for generation of a SPICE model making itpossible to simulate the penetration of an electromagnetic shield in a circuit simulator[46].
3.3.1 Transfer Impedance of Coaxial Cables
For measuring the efficiency of a coaxial cable shield a method is commonly used weretransfer impedance is measured. The standard IEC 96-1A [32] describes this procedureand the design of the test fixture. The triaxial fixture setup used in this standard can fora characterization of a 50Ω schematically be shown as in figure 3.4. The fixture is usuallyused with a network analyzer and transmission through the fixture (S21) is recorded. Thetransfer impedance of the cable shield can then be calculated be a formula (3.1) given in
3.3. Transfer Impedance Methods 19
VV~~UU
50ΩΩ 50ΩΩ 50ΩΩ
II
Figure 3.4: Schematic measurement setup example for the triaxial transfer impedance fixtureused in standard IEC 96-1A
the standard.
Zt = abs(S21 · 2 · 1.4 · 60 · ln(60
d)) (3.1)
where d is the outer diameter of the cable screen under test.
Other standards exists for transfer impedance measurements on shielded cables. Theuseful frequency range for the IEC 96-1A fixture is DC to 30 MHz. More advancedfixtures using quadraxial and quintaxial setups are useful up to 1 GHz [30].
3.3.2 Transfer impedance of conductive gaskets
The standard SAE ARP 1705 [61] (from 1981) and the revised version SAE ARP 1705A[62] (from 1997) describes how transfer impedance measurements is used to determinethe performance of EMC gaskets. Transfer impedance is defined as the voltage Uout onthe secondary side of the shield divided by the current density Jin (in Ampere per meter)on the primary side of the shield, see figure 3.5 . The current density Jin may for instancebe induced by an incident electromagnetic field. The higher the transfer impedance fora gasket is the lower is the performance of the gasket since an electromagnetic field onthe front side of the shield easier can give raise to electromagnetic fields on the backsideof the shield.
The measurement procedure is straightforward and to be able to determine the trans-fer impedance from the measured entities a simplified low frequency model of the testfixture is used, by improvements of this model more exact results would be obtained[56, 13]. The test fixture in the standard SAE ARP 1705 is stated to be useful for ab-solute measurements up to 700MHz. There are improved versions of transfer impedancetest fixtures for use at higher frequencies [36, 37].
The contact pressure between the gasket and the mating surface is an importantparameter that has to be controlled when a gasket is examined. In the SAE ARP 1705fixture this is accomplished by pneumatic pressure behind a membrane involving the”cover” that rests on the gasket. This method has its drawbacks, especially when testing
20 Measurement Methods
++
−−UU
out
JJin
Figure 3.5: The transfer impedance measurement on a conductive gasket
fingerstock gaskets that needs a small contact pressure and long compression height [64].The newer SAE ARP 1705A the membrane is removed and the new construction is usingpneumatic cylinders to set the contact pressure.
3.3.3 Transfer Impedance of Shielded Connectors
For shielded connectors the transfer impedance characterization is also commonly used.Different fixtures are used in published studies [22, 66].
3.3.4 Relating Transfer Impedance to Other Performance Mea-sures
The transfer impedance is a circuit theory description of the shielding performance. In afixture for transfer impedance measurement the purpose of the fixture design is usuallyto establish an uniform current distribution over the test object.
In measurement methods for shielding effectiveness in terms of insertion loss, the at-tenuation of an electromagnetic wave is studied. In this case no special effort is made tocontrol the current distribution caused by the incident field. Here the focus is on stan-dardized antenna positions and other measures for the repeatability of the measurement.
There is no easy way to convert transfer impedance to the insertion loss measuredusing a MIL-STD-285 type measurement since that would require the geometry of thetest fixture and the incident angle of the field towards the gaskets to be taken intoaccount [20, 57]. Experimental comparisons have been performed to relate the differentrepresentations of shielding performance [58, 25, 26, 1, 38].
3.4 Characterization of Permittivity and Permeabil-
ity
Several methods exists for the measurement of permeability and permittivity. The mostcommon measurement methods can be categorized as:
3.4. Characterization of Permittivity and Permeability 21
• Loaded resonant waveguide cavity
• Open ended coaxial line
• Loaded coaxial transmission line
Loaded cavity resonance measurement methods are commonly used because of thegood accuracy offered, particularly for measurement of imaginary part of the permittivityfor determining the losses in a material [6, 7]. That sample material is mounted insidea resonant cavity and thereby loading the cavity. The sample piece must be of certaindimensions to fit. However the resonance of a cavity makes the method only covering anarrow frequency band. This can be improved to some extend by including higher orderresonant modes.
The open ended coaxial line has a main advantage that it is a non destructive testmethod. The coaxial probe is pressed against a specimen and the reflection coefficientis measured. From the reflection coefficient the permittivity can be determined. [6, 19].One disadvantage is that the method is sensitive to air gaps that disturb the electric fieldthe probe and the sample [6]. The measurement accuracy for this method is not as goodas the other methods mentioned.
In the loaded coaxial transmission line method the material under test is placed tofill the volume between the inner and outer conductor in a section along the transmissionline. The dimensions of the sample piece are critical to ensure a precise fit. The materialmay load the line and cause a change of characteristic impedance. Both reflection andtransmission through the fixture is used when calculating the test material data [6, 8].The measurement accuracy for this method is not as good as that of loaded cavityresonance measurement methods. An advantage of the loaded coaxial transmission linemethod is that it offers the possibility to perform measurements in a broad frequencyband.
The advantages and disadvantages of the methods gives a guideline on what mea-surement method to use depending on restrictions on sample preparation and desiredfrequency range and accuracy of the data. The loaded cavity resonance measurementmethod gives high accuracy data for narrowband measurements at frequencies of over100 GHz. The open ended coaxial line offers a method where only a flat surface of stud-ied material is needed. The loaded coaxial transmission line method makes it possibleto measure over wide frequency range with better accuracy than what the open endedcoaxial line method offers [6].
There exists a large number of publications on electromagnetic properties of plas-tic materials. Older publication (before 1985) often focused on plastic material data atfrequencies below 100 MHz [12]. In later years plastic material data at microwave fre-quencies have been of major concern because the increased need for accurate data whendesigning devices working at higher frequencies [59]. It is then also important to studythe temperature variations in the electromagnetic properties of the material under test[12, 59]. Because of temperature variations the dielectric material can cause impedancemismatch losses and degrade the performance of a device. A good overview of availablemeasurement methods are given in publications [6, 7].
22 Measurement Methods
3.5 Near Field Scanning of Printed Circuit Boards
and Conductive Surfaces
Near field scanning is a useful tool in EMC for finding radiating sources on a printedcircuit board (PCB). A broken component or a poorly routed conductor trace can therebybe identified and fixed.
In particular near field scanning of magnetic field over a conductive surface gives ahint of the surface current distribution. In this case a magnetic field probe is stepwisepositioned over the conductive surface and connected to one port of a vector networkanalyser. The other port of the vector network analyser is connected to the conductivesurface for excitation of surface currents. The coupling between the surface and the fieldprobe is measured to magnitude and phase. From this magnetic field distribution thesurface current distribution can be estimated [34].
An investigation on near field radiation from traces on a printed circuit board showsgood agreements between measured and simulated fields [16]. Both E-field and H-fieldprobes was used and the magnitude of the field was studied. The near field scan werecollected using a computer controlled three dimensional positioning mechanism. A similarstudy were done using two probes scanning simultaneously at different height to obtaina phase difference and a amplitude difference [42]. The measured values compares wellwith simulated results using a thin-wire structure analysis program (NEC). The methodof near field scanning used in the study is suggested to be accurate enough for far fieldpredictions of the radiation from a printed circuit board.
For antenna design work near field scanning is also used. The current distribution ina printed antenna can be analyzed and the far field radiation pattern can be estimated[34]. By scanning of magnetic field over a ground plane on a printed circuit board thecurrent distribution can be analyzed and it can be decided if the ground planning designstage have been done properly. This type of analysis usually better suited to be run in acomputer simulation. However it is of great value to able to verify computer simulationresults with measured data.
Chapter 4
Thesis Summary
4.1 Summary of Contributions
My research is documented in seven papers of which two are submitted to scientificjournals and five have been presented at scientific conferences and published in conferenceproceedings.
Paper A:Carlsson J. and Lundgren U., ”An Approach to the Generation of SPICE Models Fea-sible for EMC Problems”, Symposium Record, 2000 IEEE International Symposium OnElectromagnetic Compatibility, (Washington, D.C., USA), 2000
Paper B:Jenvey S. and Lundgren U., ”A comparison of measured and simulated current distribu-tions on a printed log-periodic antenna”, Symposium Record, Antenn 00, Nordic AntennaSymposium, (Lund, Sweden), 2000
Paper C:Ekman J. and Lundgren U., ”Analysis of Printed Antenna Structures using the PartialElement Equivalent Circuit (PEEC) Method”, Symposium Record, EMB01, (Uppsala,Sweden), 2001
Paper D:Lundgren U., Carlsson J. and Delsing J., ”SPICE models of barrier compared to measureddata”, Symposium Record, 2001 IEEE International Symposium On ElectromagneticCompatibility, (Montreal, Canada), 2001
Paper E:Lundgren U., Ekman J. and Delsing J., ”Characterization of Conductive ThermoplasticComposite Materials Using Multiple Measurements Methods”, Symposium Record, EMCEurope 2002, (Sorrento, Italy), 2002
23
24 Thesis Summary
Paper F:Lundgren U., Ekman J. and Delsing J., ”Shielding Effectiveness Data on CommercialThermoplastic Materials”, Submitted to Electromagnetic Compatibility, IEEE Transac-tions on
Paper G:Lundgren U. and Delsing J., ”Electromagnetic properties of thermoplastic material forvarying temperatures”, Submitted to Microwave Theory and Techniques, IEEE Trans-actions on
4.1.1 Paper A: An approach to the generation of SPICE modelsfeasible for EMC problems
Considering electromagnetic barriers and the desire to be able to simulate barriers incircuit simulators, work have been done to develop methods to generate SPICE models fora few barrier kinds. The barriers considered are divided into two groups, physical barriersand geometrical barriers. In the physical barrier group we can find filters and othercommercially available components with unknown interior structure. The geometricalbarriers are geometrically well known structures like a shielding enclosure or separationof traces on a printed circuit board.
It has been shown that for a geometrical barrier where the structure has a fixedcross section, the barrier characteristics can be calculated with inexpensive tools if thematerial properties are known. Barriers with stepwise changing cross sections were alsoapplicable when different circuit model segments were cascaded. A two-dimensional finitedifference program was developed in which the cross-section of the barrier is defined usinga CAD-like user interface. The program computes the per-unit length parameters of themulti-conductor transmission line assumption of the barrier.
Prototype printed circuit boards incorporating the studied barriers was subject tomeasurements. Good agreement was obtained between measurements and circuit simu-lation results with the generated circuit model of the barrier. The highest useful frequencyfor models created with this approach depends on the number of segments in the lumpedcircuit model.
For physical barriers another technique was developed based on measurements on thebarrier followed by a error minimization procedure of an assumed lumped componentcircuit to the measured data. Models generated this way was also verified by comparisonof measured values to simulation results of the generated model. Good agreement wasobtained for all comparisons at frequencies up to 1 GHz. Some models were performingwell up to 4 GHz (limited by lab equipment), while some models could be improved toperform well in the frequency range 1 GHz to 4 GHz.
4.1. Summary of Contributions 25
4.1.2 Paper B: A comparison of measured and simulated cur-rent distribution on a printed log-periodic antenna
The current distribution on a log periodic dipole antenna (LPDA) have been studied usingnear field measurements and simulations based on method of moments. The antenna wasconstructed on a printed circuit board.
Detailed measurements of the magnetic field was obtained with 5 mm stepping twodimensional scanning just above the surface of the antenna. The field was sampledusing a loop probe connected to a vector network analyzer. The probe was positionedby a computer controlled scanning mechanism. Two polarizations of the magnetic fieldwas acquired with magnitude and phase. The current distribution on the antenna werestudied by examining the magnetic fields obtained by the scanning procedure.
The use of the printed circuit board to support the radiating elements and the parallelwire transmission feeder line led to a mixed dielectric environment. This affected thecurrent distributions on the feeder and the radiating elements and hence the radiationpatterns and the impedance characteristics of the antenna.
Measured current distributions were compared with predicted distributions obtainedfrom Method of Moments (MOM) analysis of the LPDA structure. Comparison resultsshow that magnetic field scanning of a printed antenna is a useful tool for getting a betterunderstanding of the real performance of the antenna. Measured and predicted far fieldradiation patterns are also compared.
4.1.3 Paper C: Analysis of Printed Antenna Structures usingthe Partial Element Equivalent Circuit (PEEC) Method
In this paper, the partial element equivalent circuit (PEEC) method is illustrated andapplied to printed antenna structures where measurements are compared to simulationsand analytical solutions. The possibility to use simplified PEEC models to decreasecomputation time is discussed with illustrative examples. The PEEC method is a fullwave technique for the solution of mixed circuit and field problems in both the time andfrequency domain.
The international interest for the method has been gaining rapidly for the past yearsbut in the Nordic countries the research effort has been low. This paper can be consideredas an fundamental introduction to this electromagnetic computation technique.
By using a specialized discretization, the original structure is converted into a net-work of discrete inductances, capacitances and resistances, called the partial elements.The partial elements are calculated either by using numerical integration techniques orsimplified closed form equations. The resulting equivalent circuits are solved by using acommercial circuit simulation program like SPICE. The use of SPICE-like circuit solversfacilitates the inclusion of discrete components, transmission lines, current/voltage sourceetc in the resulting PEEC model.
The PEEC method has been shown to be a very powerful simulation technique forcombined circuit and electromagnetic field problems. The first example displays thepossibility to make PEEC models by using closed form equations to calculate the partialelements and a free version of SPICE as the solver. This feature makes the method
26 Thesis Summary
possible to use in education and for simple design tasks. However, the method require aretarded circuit solver to be considered a full wave method comparable with a methodof moments solution.
The application of PEEC’s to antennas is a valuable tool in many areas where antennaresonance frequencies are of importance. The paper shows that antennas are modelledwith good agreement compared to analytical solutions and measurements.
4.1.4 Paper D: SPICE models of barrier compared to measureddata
SPICE models of electromagnetic zone barrier are devised. The models are based ondata from 2D and 3D Maxwell equation solvers. A transfer impedance approach mod-elled incident electromagnetic waves in SPICE. Test systems using D-sub connectors,passive surface mounted filters and encapsulation was designed. Test system verificationmeasurements were made in a fully anechoic chamber. The coupling through the zonebarriers was measured. Good agreement was found between simulated and measureddata. The focus for electronic system designers is on product functionality. Here EMCaspects are hard to approach using for the electronic engineer well known tools such asSPICE.
This work thus focuses on building SPICE models for electromagnetic zone barri-ers enabling SPICE simulations of incident radiated power and immunity to incomingdisturbances.
SPICE models has been developed for commercially available components that canbe regarded as EMC barriers. Surface mounted filters and shielded connectors and cablesare examples of such components. Models have also been developed for EMC barriersthat appear in a circuit due to the layout of circuit.
The measurement results obtained in the unechoic chamber shows a poor agreementwith the SPICE simulations in the frequency range 30 MHz to 60 MHz where overlappingdata is available. Differences between 10 and 40 dB in the two configurations is found.However when comparing with other measurements or by extrapolation of the simulationresults a better agreement is found in the higher frequency range towards 100 MHz.
With the EM clamp injection method a much better agreement is obtained betweenthe measured system transfer impedance and the SPICE simulation. By combiningmeasurements for the parts of the system into a complete system transfer impedancegood agreement is obtained with SPICE simulation, in some instances a remaining offsetis found.
The measurements for the desired verification are hard to do with one single approach.The frequency range for the SPICE simulation was limited by the transfer impedance dataobtained for the coaxial cable shield. It is reasonable to believe that better agreementwould be found if the SPICE simulation could be done at 100 MHz. The deviationbetween the obtained results seems to decrease when frequency increases in the frequencyrange of this study.
4.1. Summary of Contributions 27
4.1.5 Paper E: Characterization of Conductive ThermoplasticComposite Materials Using Multiple Measurements Meth-ods
In a study of conductive thermoplastic composite materials, samples were manufacturedand measured. Samples with different base polymers, filler materials and different amountof filler made it possible to generate data for many combinations. The samples werecharacterized in terms of their complex permittivity and complex permeability, planewave shielding effectiveness (SE) and near electric field shielding effectiveness. As canbe expected materials that shows a relatively high shielding effectiveness for a incidentplane wave also in general offers shielding in the near field situation that was studied. Acorrelation between SE and complex permittivity was also found.
This paper describes the work to compare measurement methods to acquire electro-magnetic shielding effectiveness of conductive thermoplastic materials. The frequencyrange in this study is 150 MHz to 1 GHz. Three different measurement methods arecompared to test the validity of the methods. For the comparison of measurement meth-ods a number of different composite materials were analyzed in the study. Data fromthree of those are here used for the comparison of the three methods. Measured dataare also compared to give an indication how shielding effectiveness is affected by incidentfield impedance and by the permittivity and permeability of the material.
A measurement method for studying near field shielding effectiveness was developed.A battery powered square wave generator with dipole antenna was enclosed in a box madeof the tested material. Several boxes of different materials were specially manufacturedfor this study. The repeatability for these measurements are very good and the resultspresented are from one measurement occasion but the results must of course be regardedas unique to this test set-up. This is because the shielding material is in close proximityof the transmitting antenna so that the input impedance of the transmitting antennamay change when changing material. Also the near field impedance, that is the relationof electric field strength to magnetic field strength is unknown.
As a second measurement method a modified MIL-STD-285 type method was used.Measured shielding effectiveness with this far field method show the same trend in fre-quency response as the near field method but with an offset in some cases. When studyinga larger number of different materials than presented in this paper, materials that per-forms well in the plane wave case usually also offers good shielding in the near fieldcase.
The third method was to use a loaded coaxial transmission line fixture for measure-ment of complex permittivity and complex permeability. Using traditional techniquesthis data can be used for estimation of SE for an infinitely large plane electromagneticshield. Calculated SE based on measured material properties does not agree well withmeasured near field shielding effectiveness. The largest deviation found in this compari-son is almost 20 dB.
By analyzing the measured permittivity and permeability further a large difference isfound in imaginary permittivity for the materials. The imaginary part of the permittivityincludes the effect of conductivity in the material. The cause of the losses in a dielectricmaterial is usually that the conductivity is large. In conclusion the thermoplastic ma-
28 Thesis Summary
terials with high imaginary part of the permittivity seem to give an improved shieldingeffectiveness compared to materials with small imaginary part of permittivity. The realpart of the permittivity does not correlate well with shielding effectiveness.
4.1.6 Paper F: Shielding Effectiveness Data on Commercial Ther-moplastic Materials
Ten different commercially available conductive thermoplastic materials have been testedfor near- and far-field shielding effectiveness. Far field shielding effectiveness was testedusing a modified standard measurement technique to provide results comparable withcompany provided data. Further, housings of the different thermoplastic materials wasconstructed and equipped with a EMI source to model a realistic near field shielding effec-tiveness situation. Shielding effectiveness data up to 1GHz is presented. The conductivethermoplastic material Faradex XP211 (with filling of stainless steel fibre) and RTP EMI283 (with filling of nickel coated carbon fibre) were the two materials offering the best farfield shielding performance. For near field shielding, Faradex XX711 and Bekaert Beki-Shield (both with filling of stainless steel fibre) were the two best performing. FaradexXX711 showed the best combined far field and near field shielding results.
One problem that arises for the EMC engineer is to select an encapsulation techniquethat offers a desired degree of electromagnetic shielding for a new electronic device. Themanufacturers of different conductive filler materials sometimes specifies the shieldingperformance of their material in an application according to standardized measurementmethod but deviations from the exact standard often occurs. This makes comparisonsbetween different manufacturers hard. Further the standardized method just give a hintof what the shielding performance can be for the same material in an electronic deviceencapsulation application.
Thus it was decided to evaluate electromagnetic shielding effectiveness for commer-cially available thermoplastic materials. Ten materials were chosen and samples manu-factured for analysis using two measurement methods. This paper describes the compo-sition of the chosen materials, the measurement techniques are discussed, the recordedresults are presented and conclusions from the comparisons are drawn. In some instancesmanufacturer data were available for comparison.
This paper describes the composition of the chosen materials, the measurement tech-niques are discussed, the recorded results are presented and conclusions from the com-parisons are drawn. In some instances manufacturer data were available for comparison.
When comparing the near- and far- field shielding effectiveness for the thermoplasticmaterials the following is noted.
• Faradex XP211 offer the best far field shielding effectiveness.
• Faradex XX711 and Beki-Shield offer the best near field shielding effectivenessknocking the signal from the EMI source (transmitter) down below the noise floor.
• Faradex XA611 is the material with the lowest level of shielding effectiveness forboth near- and far- field.
4.1. Summary of Contributions 29
• Faradex XX711 is the best material for the combined shielding effectiveness.
Considering that the near field shielding measurement imitates the use of material inan application it is disappointing to se how poor guide the far field shielding effectivenessresults are when a material selection for an enclosure must be made. In cases wheremanufacturer data were available, agreement was quite good with the near field methodin one instance while the difference was close to 30 dB in the other three instances.
4.1.7 Paper G: Electromagnetic properties of thermoplastic ma-terial for varying temperatures
A thermoplastic material is examined. The complex permittivity and complex perme-ability are obtained while the temperature is varied from 20 to 60 Celsius. It wasnecessary to perform instrument calibration at each temperature to cancel temperatureeffects on cables and connectors. The thermoplastic material is used in a laminate withmetallic foil for encapsulation of electronic circuits. The laminated technique offers agood barrier against moisture and good shielding for electromagnetic energy. To be ableto design stripline transmission lines in the laminate it is necessary to know the electro-magnetic behavior of the isolating plastic material. Real part of the relative permittivitywas found to be 2.0± 0.1 in the frequency range 100 MHz to 2.5 GHz. This value showsa very small dependence of temperature changes in the range 20 Celsius to 60 Celsius.
It is of interest to explore the usage of a particular laminated encapsulation materialwith integrated antennas in environments with changing temperature. The major prob-lem is to establish a test methodology where temperature effects other than these of thematerial under test are sufficiently suppressed. Thus series of experiments have been con-ducted to obtain complex permittivity and complex permeability for the thermoplasticover temperature 20 − 60C and frequencies from 100 MHz to 2.5 GHz.
It was found that the material under test have a real relative permittivity of 2.0 atroom temperature (20C) and it is independent of temperature in the range 20C to 60C.Imaginary relative permittivity was slightly smaller for the material under test than forpolyethylene. The results for the permeability was as expected for the material undertest. The real part of the relative permeability is close to 1 and imaginary part close to0.
The measurement setup was very sensitive to temperature variations. The first at-tempt to cover the temperature range 0 Celsius to 80C with a single calibration wasinsufficient. Measurements were then done at three temperature points, 20, 40 and60C, with careful calibration at each temperature to cancel the temperature effects onthe cables and connectors.
The good results from this work led to a patent application for an encapsulationtechnique with integrated patch antenna and antenna feed transmission line [44].
30 Thesis Summary
Chapter 5
Conclusions - Suggested FurtherWork
This thesis concludes with some reflections on achieved results and interesting angles tobe aimed at for further work.
Techniques to generate lumped circuit models for electromagnetic barriers have beendeveloped and methods have been devised for generating models for transmission likebarriers as well as for barriers with unknown geometrical shape (Papers A and D).
The developed methods have been verified successfully for frequencies up to 1 GHz,for some barrier models up to 4 GHz. It is desired to find methodologies that increasesthe useful frequency range of the generated models further and that would enable barriermodeling for addressing electromagnetic shielding for instance in shielding enclosures.
Modeling of printed antenna structures have been done using the Partial ElementEquivalent Circuit (PEEC) method (Paper C). From SPICE computer simulations withthe generated antenna models good agreement is found when comparing with analyt-ical solutions and s-parameter measurements. It would also be of interest to comparePEEC model simulations of current distributions with measured current distributionssince this is an important step towards barrier modeling including coupling to radiatedfields. A comparison of current distributions on a printed antenna attained by Methodof Moment (MoM) simulations and scanning measurements of magnetic field distributiondemonstrates this technique (Paper B).
A near field shielding effectiveness (SE) measurement method has been developedimitating the use of shielding material in an application (Paper E and F). Measured nearfield SE for conductive thermoplastic materials are compared with results from a far fieldSE measurement method. The measured near field SE results deviates substantially frommeasured far field SE. Thus it is questioned how useful far field (plane wave) SE data iswhen a material selection for an enclosure must be made.
In a study for characterization of a thermoplastic material in terms of permittivityand permeability measurements, the temperature dependence of the material parameterswas of interest (Paper G) . It was noted that the measurement cables were very sensitiveto temperature variations. To acquire reliable data instrument calibration were necessaryat each temperature point to cancel the temperature effects on the cables and connectors.
31
32 Conclusions
Further work is desired in the area of EMC barrier characterization and modeling andcould be summarized:
• Development of improved lumped circuit model generation techniques to increasethe useful frequency range for generated barrier models.
• Work to establish barrier modeling techniques for electromagnetic shielding whereradiating sources are considered. This would enable circuit simulation using SPICEsoftware to estimate shielding effectiveness of shielding enclosures. Here the transferimpedance and transfer admittance description of an electromagnetic shield maybe an important approach. PEEC modeling is also an interesting and powerfultechnique for this purpose.
• Work on PEEC modeling for analysis of current distribution in conductive surfacessuch as patch antennas and PCB ground planes. This is an important step to-wards the ability to include barriers for electromagnetic field shielding in circuitsimulation. Near field scanning measurements of magnetic fields could be used forverification in such work.
• Better methods are desired for specification of electromagnetic shielding materialssince far field SE measurements may fail to reflect the application of a shieldingmaterial in an electronic device enclosure. This may require better understandingof impedance of radiated field from a device and electric field SE and magnetic fieldSE given separately for a shielding material.
References
33
34 References
References
[1] Acuna V. E., ”Transfer impedance measurement as a test for electromagneticcompatibility”, IEEE 1975 Int. Symp. on Electromagnetic Compatibility, Sympo-sium Record, 1975
[2] Allmendinger G., ”LET THE CIRCLE BE UNBROKEN, How the InformationCircle Created by Device Networking / M2M and the Internet Will Automate theGlobal Enterprise”, Harbour Research Inc., Boston USA, July, 2003
[3] Archambeault B. R., ”EMI fingerstock performance when mated to varioussurfaces”, IEEE 1987 Int. Symp. on Electromagnetic Compatibility, SymposiumRecord, p. 161-163, 1987
[4] Archambeault B. and Thibeau R., ”Effects of corrosion on the electrical propertiesof conducted finishes for EMI shielding”, IEEE 1989 Int. Symp. on ElectromagneticCompatibility, Symposium Record, p. 46-51, 1989
[5] Archambeault B., ”Modeling and Simulation Validation for EMC Applications”,IEEE 1999 Int. Symp. on Electromagnetic Compatibility, Symposium Record, p.492-496, 1999
[6] Baker-Jarvis J., Janezic M.D., Riddle B., Holloway C.L., Paulter N.G., BlendellJ.E., ”Dielectric and Conductor-Loss Characterization and Measurements on Elec-tronic Packaging Materials”, NIST Technical Note 1520, July 2001, CODEN: NT-NOEF
[7] Baker-Jarvis J., Geyer R.G., Grosvenor J.H., Jr.; Janezic M.D., Jones C.A.,Riddle B., Weil C.M., Krupka J.,”Dielectric characterization of low-loss materials acomparison of techniques”, Dielectrics and Electrical Insulation, IEEE Transactionson, Volume: 5 Issue: 4 , Aug. 1998 Page(s): 571-577
[8] Barry W., ”A broad-band, automated, stripline technique for the simultaneousmeasurement of complex permittivity and permeability”, Microwave Theory andTechniques, IEEE Transactions on, Volume: 34, Issue: 1, January 1986
[9] Bjorklof D., ”EMC Standards and Their Application”, Compliance Engineering,1999
[10] Bjorklof D., ”Guidelines for Application of the EMC Directive”, ComplianceEngineering, 1999
35
36 References
[11] Bodnar D. G., Denny H. W. and Jenkins B. M., ”Shielding effectiveness measure-ments on conductive plastics”, IEEE 1979 Int. Symp. on Electromagnetic Compat-ibility, Symposium Record, p. 27-33, 1979
[12] Bur A. J., ”Dielectric properties of polymers at microwave frequencies: A review”,Polymer, Volume: 26, Page(s): 963-977, July 1985
[13] Carlsson E. J. and Kasevich R. S., ”Analysis of transfer impedance technique:equivalent circuit model and network analyzer measurements”, IEEE 1987 Int.Symp. on Electromagnetic Compatibility, Symposium Record, 1987
[14] Carlsson J. and Lundgren U., ”An Approach to the Generation of SPICE ModelsFeasible for EMC Problems”, Symposium Record, 2000 IEEE International Sympo-sium On Electromagnetic Compatibility, (Washington, D.C., USA), 2000
[15] Crawford M.L. and Koepke G.H., ”Design, Evaluation and use of a Reverbera-tion Chamber for Performing Electromagnetic Susceptibility/Vulnerability Measure-ments”, National Bureau of Standards, NBS Technical Note 1092, April 1986
[16] Criel S., Haelvoet K. and Martens L., ”Theoretical and Experimental Quantita-tive Characterization of the Near-Fields of Printed Circuit Board InterconnectionStructures”, IEEE 1995 Int. Symp. on Electromagnetic Compatibility, SymposiumRecord , p. 471-474, 1995
[17] Csenki W., Lunden O. and Backstrom M., ”Transient Behaviour of Large MultiMode Resonant Cavities”, 97 MSC, Anechoic chamber and OATS Users Meeting &Tutorial, Apr 28 - May 2, 1997, Vail, Colorado
[18] Cui W., Shi H., Luo X., Sha F., Drewniak J. L., Van Doren T. P. and Anderson T.,”Lumped-element Sections for Modeling Coupling Between High-Speed Digital andI/O Lines”, IEEE 1997 Int. Symp. on Electromagnetic Compatibility, SymposiumRecord, p. 260-265, 1997
[19] Devendra K. M., ”Method and apparatus for measuring the permittivity of ma-terials”, United States Patent, Pat. no. 5233306, Aug. 3, 1993
[20] Dike G., Wallenberg R. and Birkin J., ”Electromagnetic relationships betweenshielding effectiveness and transfer impedance”, IEEE 1979 Int. Symp. on Electro-magnetic Compatibility, Symposium Record, p. 133-138, 1979
[21] Directive 89/336/EEC, ”Electromagnetic Compatibility (EMC) Directive89/336/EEC”, The Council of the European Communites
[22] Dunwoody S. and VanderHeyden E., ”Transfer impedance testing of multi-conductor shielded connectors of arbitrary cross-section”, IEEE 1990 Int. Symp.on Electromagnetic Compatibility, Symposium Record, 1990
References 37
[23] Ekman J. and Lundgren U., ”Analysis of Printed Antenna Structures using thePartial Element Equivalent Circuit (PEEC) Method”, Symposium Record, EMB01,(Uppsala, Sweden), 2001
[24] FCC Part 15, ”Part 15 - Radio Frequency Devices”, Federal CommunicationsCommission
[25] Freyer G. J. and Hatfield M. O., ”Comparison of gasket transfer impedanceand shielding measurements part I”, IEEE 1992 Int. Symp. on ElectromagneticCompatibility, Symposium Record, p. 139-141, 1992
[26] Freyer G. J., Rowan J. and Hatfield M. O., ”Gasket shielding performance mea-surements obtained from four test techniques”, IEEE 1994 Int. Symp. on Electro-magnetic Compatibility, Symposium Record, p. 279-284, 1994
[27] Hariya E. and Masahiro U., ”Instruments for measuring the electromagneticshielding effectiveness”, IEEE 1984 Int. Symp. on Electromagnetic Compatibility,Symposium Record, p. 800-805, 1984
[28] Hatfield M. O. and Freyer G. J., ”Comparison of gasket transfer impedanceand shielding measurements part II”, IEEE 1992 Int. Symp. on ElectromagneticCompatibility, Symposium Record, p. 142-148, 1992
[29] Hatfield M. O., Godfrey E. A. and Freyer G. J., ”Investigations to Extend theLower Frequency Limit of Reverberation Chambers”, IEEE 1998 Int. Symp. onElectromagnetic Compatibility, Symposium Record, 1998
[30] Hoeft L. O. and Hofstra J. S., ”Measurement of Surface Transfer Impedance ofMulti-Wire Cables, Connectors and Cable Assemblies”, IEEE 1992 Int. Symp. onElectromagnetic Compatibility, Symposium Record, p. 308-314, 1992
[31] IEC 77B/215/CD, ”Amendment to IEC 61000-4-3: Electromagnetic Compat-ibility (EMC) - Part 4: Testing and measurement techniques - Section 3: Radi-ated, radio-frequency, electromagnetic field immunity test - Annex XX: Alternativemethod - Reverberation chamber method”, September 1997
[32] IEC 96-1A, ”Radio Frequency Cables, Part I: General Requirements and Mea-suring Methods”, International Electrotechnical Commission, IEC Standard Publi-cation 96-1A, 1976
[33] IEEE-STD-299, ”IEEE Standard Method for Measuring the Effectiveness of Elec-tromagnetic Shielding Enclosures”, Dec 09, 1997
[34] Jenvey S. and Lundgren U., ”A comparison of measured and simulated currentdistributions on a printed log-periodic antenna”, Symposium Record, Antenn 00,Nordic Antenna Symposium, (Lund, Sweden), 2000
38 References
[35] Kashyap S., ”Shielding effectiveness measurements with a dual TEM cell and asplit TEM cell”, IEEE 1986 Int. Symp. on Electromagnetic Compatibility, Sympo-sium Record, p. 262-264, 1986
[36] Kunkel G. M., ”Introduction to the testing for the shielding quality of EMI gas-kets and gasketed joints”, IEEE 1992 Int. Symp. on Electromagnetic Compatibility,Symposium Record, p. 134-138, 1992
[37] Kunkel G. M., ”Design of transfer impedance test fixture accurate through 10GHz”, IEEE 1990 Int. Symp. on Electromagnetic Compatibility, Symposium Record,p. 628-633, 1990
[38] Kunkel M. and Kunkel G., ”Comparison Between Transfer Impedance and Shield-ing Effectiveness Testing”, IEEE 1992 Int. Symp. on Electromagnetic Compatibility,Symposium Record, p. 149-153, 1992
[39] Kunkel G., ”Corrosion effects on field penetration through apertures”, IEEE 1978Int. Symp. on Electromagnetic Compatibility, Symposium Record, p. 96-99, 1978
[40] Kunkel G., ”Corrosion effects on EMI gasketed joints”, IEEE 1979 Int. Symp. onElectromagnetic Compatibility, Symposium Record, p. 198-203, 1979
[41] Langdon H.S., Luebbers R., ”Efficient FDTD calculation of multi-port S pa-rameters for microstrip and stripline circuits”, Antennas and Propagation SocietyInternational Symposium, 1997. IEEE., 1997 Digest , Volume: 2 , 13-18 July 1997Pages:998 - 1001 vol.2
[42] Laurin J. J., ”Near-field characterization of PCBs for radiated emissions predic-tion”, IEEE 1993 Int. Symp. on Electromagnetic Compatibility, Symposium Record,p. 322-326, 1993
[43] Lee K., Lee S. J., Park D. C. and Chung Y. C., ”Equivalent Circuit Model for theTime-domain Analysis of Multiconductor Transmission Lines by the Implicit FDTDMethod”, IEEE 1997 Int. Symp. on Electromagnetic Compatibility, SymposiumRecord, p. 287-292, 1997
[44] Leeb K.-E., Bjorklof D., Lundgren U., ”Anordning med integrerad antenn forkapsling av radioelektronik samt satt att tillverka anordningen”, Swedish Patent,Pat. No. SE0201263, Oct. 27, 2003
[45] Lessner P. and Inman D., ”Quantitative measurement of the degradation of EMIshielding and mating flange materials after environmental exposure”, IEEE 1993Int. Symp. on Electromagnetic Compatibility, Symposium Record, p. 207-213, 1993,ISBN: 0 7803 1304 6
[46] Lundgren U., Carlsson J. and Delsing J., ”SPICE models of barrier comparedto measured data”, Symposium Record, 2001 IEEE International Symposium OnElectromagnetic Compatibility, (Montreal, Canada), 2001
References 39
[47] Lundgren U. and Delsing J., ”Electromagnetic properties of thermoplastic ma-terial for varying temperatures”, Submitted to Microwave Theory and Techniques,IEEE Transactions on
[48] Lundgren U., Ekman J. and Delsing J., ”Characterization of Conductive Thermo-plastic Composite Materials Using Multiple Measurements Methods”, SymposiumRecord, EMC Europe 2002, (Sorrento, Italy), 2002
[49] Madle P. J., ”Transfer impedance and transfer admittance measurements ongasketted panel assemblies, and honeycomb air-vent assemblies”, IEEE 1976 Int.Symp. on Electromagnetic Compatibility, Symposium Record, p. 204-209, 1976
[50] van der Merwe, J., Reader H.C., Cloete J.H., ”S-parameter measurements yieldingthe characteristic matrices of multiconductor transmission lines”, ElectromagneticCompatibility, IEEE Transactions on , Volume: 40 , Issue: 3 , Aug. 1998 Pages:249- 256
[51] MIL-STD-285, ”Attenuation measurements for enclosures, electromagnetic shield-ing, for electronic test purposes, method of,”, 25 June 1956
[52] Molyneux-Child J. W., ”EMC Shielding materials, A designer’s guide”, Oxford:Newnes, 1997, ISBN 0-7506-3548-7
[53] Ondrejka A. R. and Adams J. W., ”Shielding effectiveness (SE) measurementtechniques”, IEEE 1984 Int. Symp. on Electromagnetic Compatibility, SymposiumRecord, p. 249-256, 1984
[54] Ott H. W., ”Noise reduction techniques in electronic systems”, New York: JohnWiley & Sons, Inc., 1988, ISBN 0-471-85068-3
[55] Peel R. J., ”Relationships of shielding effectiveness to transferimpedance/admittance for shielded cables”, IEEE 1982 Int. Symp. on Elec-tromagnetic Compatibility, Symposium Record, p. 312-317, 1982
[56] Quine J. P., ”Characterization and testing of shielding gaskets at microwavefrequencies”, IEEE 1993 Int. Symp. on Electromagnetic Compatibility, SymposiumRecord, p. 306-308, 1993
[57] Quine J. P. and Pesta A. J., ”Shielding effectiveness of an enclosure employinggasketed seams - relation between SE and gasket transfer impedance”, IEEE 1995Int. Symp. on Electromagnetic Compatibility, Symposium Record, p. 392-395, 1995
[58] Rahman H., Saha P. K., Dowling J. and Curran T., ”Shielding effectiveness mea-surement techniques for various materials used for EMI shielding”, IEE Colloquiumon ’Screening of Connectors, Cables and Enclosures’ (Digest No.012) London, p.9/1-9/6 of 68, 1992
40
[59] Riddle B., Baker-Jarvis J., Krupka J.,”Complex Permittivity Measurements ofCommon Plastics Over Variable Temperatures”, Microwave theory and techniques,IEEE Transactions on, Volume: 51 Issue: 3 , March. 2003 Page(s): 727-733
[60] Rumsey I. and Piket-May M., ”Application of the Finite Difference Time Do-main (FDTD) Method to a Challenging Real-World EMC Problem”, IEEE 1999Int. Symp. on Electromagnetic Compatibility, Symposium Record, p. 679-683, 1999
[61] SAE ARP 1705, ”Coaxial test procedure to measure the RF shielding characteris-tics of EMI gasket materials”, Society of Automotive Engineers, Committee AE-4G,June, 1981
[62] SAE ARP 1705A, ”Coaxial test procedure to measure the RF shielding char-acteristics of EMI gasket materials”, Society of Automotive Engineers, CommitteeAE-4G, Jan, 1997
[63] Spiegelaar H. and VanderHeyden E., ”The Mode Stirred Chamber - A Cost Ef-fective EMC Testing Alternative”, IEEE 1995 Int. Symp. on Electromagnetic Com-patibility, Symposium Record, 1995
[64] Stickney W., ”Improved testing for EMC gasket performance”, R.F. Design, vol.7, no. 1, p. 40-44, 1984, ISSN: 0163-321x
[65] Strom P., ”Vinna eller forsvinna i IT-aldern”, Liber Ekonomi, Malmo, 1999,ISBN: 91-47-04533-7
[66] Van Horck F. B. M., et al, ”A Rapid Method for Measuring the TransferImpedance of Connectors”, IEEE transactions on Electromagnetic Compatibility,Vol. 40, No. 3, pp. 193-200, Aug. 1998
[67] Wilson P. F. and Ma M. T., ”Techniques for measuring the shielding effectivenessof materials”, IEEE 1987 Int. Symp. on Electromagnetic Compatibility, SymposiumRecord, p. 547-552, 1987
[68] Wilson P. F., Ma M. T., ”Factors influencing material shielding effectiveness mea-surements”, IEEE 1985 Int. Symp. on Electromagnetic Compatibility, SymposiumRecord, Page(s): 29-33, 1985
[69] Wollenberg G. and Gorisch A., ”Coupling of PEEC Models with TransmissionLine Models for Simulation of Wiring Structures”, IEEE 1999 Int. Symp. on Elec-tromagnetic Compatibility, Symposium Record, p. 848-853, 1999
[70] Woody J. A. and Paludi C. A., ”Modeling techniques for discrete passive com-ponents to include parasitic effects in EMC analysis and design”, IEEE 1980 Int.Symp. on Electromagnetic Compatibility, Symposium Record, p. 39-45, 1980
Part II
42
Paper A
An Approach to the Generation ofSPICE Models Feasible for EMC
Problems
Authors:Jan Carlsson and Urban Lundgren
Reformatted version of paper originally published in:Symposium Record, 2000 IEEE International Symposium On Electromagnetic Compat-ibility, (Washington, D.C., USA), 2000
c© 2000, IEEE, Reprinted with permission.
43
44 Paper A
An Approach to the Generation of SPICE Models Feasible forEMC Problems
Jan Carlsson Urban LundgrenSP Swedish National Testing & Research Institute Luleå University of Technology
Brinellgatan 4 RegnbågsallénSE-501 15, Borås, SWEDEN SE-971 87, Luleå, SWEDEN
Abstract: A method to describe barriers such as filters, cables,connectors etc. with circuits consisting of linear discretecomponents is presented. The circuit is constructed by viewingthe barrier as a multi-conductor transmission line for whichthe per-unit length parameters have to be determined. Adeveloped two-dimensional finite difference program in whichthe cross-section is defined by drawing it using the CAD-likeuser interface computes these. For barriers that cannot beviewed as transmission lines a method for determiningequivalent circuits outgoing from measured S-parameters hasbeen developed. Derived models have been used in SPICE andvalidated by comparison with measurements.
Introduction
When analyzing EMC problems for complex systems it isnecessary to break down the system and characterize eachcoupling path or barrier. By describing a barrier with a circuitconsisting of discrete components the propagation ofdisturbances in the system can be computed by the use of anordinary circuit simulator as e.g. SPICE [1]. However, themain problem is to find a circuit representation and componentvalues that describe the behavior of the barrier. Two differentapproaches to accomplish this have been used.
For barriers that can be viewed as multi-conductortransmission lines the cross-section, which is assumed to beuniform, is divided into small elements and the Laplaceequation is solved. From the solution the charges on eachconductor can be computed and thereby the per-unit lengthinductance and capacitance matrices. When the matrices areknown it is a straightforward task to construct a circuitrepresentation that can be used in a standard circuit simulator.
Barriers such as e.g. commercial filters that cannot be viewedas transmission lines must be treated in another way. In themethod that we have used the first step is to set up a discretecircuit and then compute the S-parameters. The computed S-parameters are then compared with measured for thefrequency range of interest and the weighted difference isminimized by adjusting component values. This is done in aniterative scheme searching among component values in agiven range.
Determination of the per-unit length parameters
For some type of barriers it is sufficient to have knowledge ofthe geometrical shape and the material in the cross-section in
order to describe its electrical characteristics. One examplewould e.g. be the barrier between two parallel conductors on aprinted circuit board for which the crosstalk could becomputed with the knowledge of the geometry and thematerial properties of the circuit board. The approach that wehave used for these types of barriers is to first determine theper-unit length parameters and then create a discrete circuitrepresentation. The circuit can then be used in a standardcircuit simulator such as e.g. SPICE to obtain the desiredresponses either in the time or in the frequency domain. Therequirement for this method to be successful is that the barrierunder consideration can be viewed as a multi-conductortransmission line, i.e. it must have a uniform cross-sectionwith an extent that is small compared to the wavelength. Forbarriers that don’t have a uniform cross-section it is sometimespossible to describe them as a number of cascaded sectionswith uniform cross-sections, i.e. a staircase approximation.
Since the approach that we have used is based on multi-conductor transmission line theory we have to determine theper-unit length parameters in order to arrive at the wantedcircuit representation. Looking at the circuit representation ofa short section of a multi-conductor transmission with threewires (reference not counted), Figure 1, the following relationsbetween the entries in the per-unit length capacitance andinductance matrices and the values of the circuit elements canbe found.
[ ]
−−
−−
−−
=
∑
∑
∑
=
=
=
3
1
33231
23
3
1
221
1312
3
1
1
k
k
k
k
k
k
ccc
ccc
ccc
C, [ ]
=
333231
232221
131211
lll
lll
lll
L
z z+dz
c11dz c22dz c33dz
c12dz
c23dz
c13dz
l33dz
l22dz
l11dz
l13dz
l12dz
l23dz
ref.
1
2
3
Figure 1 Circuit representation of a short section of athree-wire transmission line.
The per-unit length parameters can be computed by usingnumerical methods such as the method of moments (MoM)[2], the finite element method (FEM) [3] etc. For some simplecases it is even possible to use analytical formulas. Themethod that we have used is the finite difference method(FDM) [4, Sec. 3], mainly because it is simple to implementand that it easily can handle complicated cross-sections withdifferent materials. For the simple example shown in Figure 2the dimension of the per-unit length capacitance andinductance matrices will be two by two, since we have twoconductors and a common reference (the ground plane). Forthe general case theij -element in the capacitance matrix canbe determined by letting the potential on all conductors exceptthe j:th be equal to zero and evaluating the charge on the i:th
conductor, i.e. [5], jmVj
iij mV
QC ≠== ,0 . Thus, in order to
determine the capacitance matrix for the PCB in Figure 2, wehave to solve the Laplace equation for the configuration twotimes with different boundary conditions.
Reference (ground plane)
Conductor 1 Conductor 2
Figure 2 Cross-section of a printed circuit board with twoconductors.
The solution of the Laplace equation gives the potentialdistribution in the region and we can by applying Gauss’ lawdetermine the charge per unit-length on conductori as:
∫ ⋅∇−=
il
ii dlnVQ ˆε where il is a closed line around conductori,
n is an outward directed unit vector andV is the potentialdistribution. The inductance matrix can be computed by theknowledge of the capacitance matrix for the case when allmaterial in the cross section is free space, i.e. [6],
[ ] [ ] 1000
−= CL εµ where [ ]0C is the capacitance matrix when all
dielectric material in the cross section is replaced by freespace. The remaining problem now is to determine thepotential distribution by solving the Laplace equation. Bystarting with Maxwell’s equations for the two-dimensionalelectrostatic case and approximating the derivatives with finitedifferences we can quite easily write down the followingrelation between the potential in neighboring nodes in thefinite difference mesh, see Figure 3.
( )( )
( )( )
( )( )
( )( )DCBA
DAji
DCBA
CBji
DCBA
DCji
DCBA
BAjiji
VV
VVV
εεεεεε
εεεεεε
εεεεεε
εεεεεε
+++
++
+++
++
+++
++
+++
+=
−+
−+
22
22
1,1,
,1,1,
(1)
By giving all nodes in the cross-section an initial estimate,0, jiV , and by scanning through the nodes by an iterative
procedure we can determine the potential distribution in thewhole region and thereby are we able to compute the per-unitlength parameters.
Computer code for determining per-unit length parameters
Based on the method for computing the per-unit lengthparameters described above a computer code called FD2D wasdeveloped. The code has a Windows user interface where thecross-section of the barrier easily can be defined by simplydrawing it on the screen, see Figure 5. The code has beenvalidated against several test cases and the agreement has beenfound to be good with previously published results. As anexample of a simple validation the characteristic impedancefor an air coaxial cable with an inner to outer conductor radiusratio of five was computed using a grid size of 200 by 200nodes. The computed impedance was 96.85Ω, which shouldbe compared to the exact value of 96.57Ω.
Vi-1,j Vi,j
Vi,j-1
Vi+1,j
Vi,j+1
εB
εAεD
εC
Figure 3 A part of the finite difference mesh.
Figure 4 Printed circuit board for measurements ofcrosstalk between adjacent conductors.
By defining the length of the analyzed barrier the code iscapable of generating a representative SPICE circuit file orcomputing the scattering parameters (S-parameters) for agiven frequency range. The S-parameters are determined byfirst setting up the chain matrix for the analyzed device [7] andthen using relations between the chain matrix and thescattering matrix. These relations can easily be found byexpressing the total voltages and currents in terms of thescattering voltages and currents, see (2).
Validation against measurements
In order to validate the computed response of a barrier byusing the FD2D program for generating circuit files and thenusing them in SPICE a number of measurements have beenconducted. As an example the crosstalk between conductorson a printed circuit board was analyzed. For the measurementof near end and far end crosstalk a vector network analyzerwas used to get the scattering parameters from the four-portmade of two adjacent conductors with the length 100 mmacross the circuit board. All four ports were connected to 50Ωduring all measurements. Configurations with differentspacing between the conductors were measured and comparedwith simulations on the corresponding circuit models. Alsoconductors with a non-uniform cross section along the lengthwere considered, see Figure 4.
300k 1M 10M 100M 1G-90
-80
-70
-60
-50
-40
-30
-20
-10
0
Near end
Far end
MeasuredComputed
UN
E/U
g,
UF
E/U
g[d
B]
Frequency, Hz
Figure 6 Crosstalk between conductors on a printed circuitboard determined by modeling the barrier as a discrete
circuit compared to measurements.
One example of the agreement between the computedcrosstalk and the measured crosstalk is shown in Figure 6. Ascan be seen in the figure the agreement is good up to 1 GHz,although the resonance frequencies are not accuratelypredicted. One reason for this is that in the circuit modelapproach that we have used radiation losses are not taken intoaccount.
Figure 5 User interface of the finite difference program for determination of per-unit length parameters. In theexample a D-sub connector is analysed.
Barriers that cannot be viewed as transmission lines
For some types of barriers measurement is the only possibleway of characterizing the behavior. For these cases a methodby which it is possible to deduce a network from measureddata has been developed. In this approach we take themeasured S-parameters and compute the same for an assumednetwork. The next step is to seek for optimal values forcomponents in the network so that a best fit, in some respect,is found. By this procedure the wanted network representingthe measured device can be determined. Of course, the successof this method is only guaranteed if the assumed networkactually can represent the device under consideration. Thus, abasic knowledge of circuit theory and some experience arerequired.
Since measurements with a network analyzer usually gives theS-parameters of a network it is natural to use these parametersas the basis for the comparison with the assumed network.Thus, we have to compute the S-parameters for the network.This can be done by first computing the chain matrix for thenetwork and then convert to S-parameters. The relationsbetween the S-parameters and the chain parameters are:
[ ] [ ] [ ] [ ] [ ] [ ]( )
[ ] [ ] [ ] [ ]( ) [ ]
−+
+
∗
+
++=
−
−−
11
11
10
0
11
00
11
DCZBZ
A
DCZBZ
AS
[ ] [ ] [ ] [ ] [ ] [ ]( )
[ ] [ ] [ ] [ ] [ ]( ) [ ] [ ]( )
−+
+−
−
∗
+
++=
−
−−
DCZDCZZ
ABZ
A
DCZBZ
AS
01
000
11
00
12
11
11
[ ] [ ] [ ] [ ]( ) [ ] [ ] [ ] [ ]( ) 10
1
0
1021
112 −
−− +
+++= DCZB
ZADCZS
[ ] [ ] [ ] [ ]( ) [ ] [ ] [ ] [ ]( )
[ ] [ ] [ ] [ ]( )
−+
−
∗+
+++−= −
−−
DCZBZ
A
DCZBZ
ADCZS
00
10
1
0
1022
1
11
(2)
where [A], [B], [C] and [D] are the usual sub-matrices in thegeneral chain matrix.
When the S-parameters have been determined for allfrequencies of interest we have to compare the values withmeasured S-parameters. Figure 7 shows an example of such acomparison.
In order to find component values in the network that willmake the network representing the measured device weminimize the average difference between measured and
computed S-parameters, i.e. we minimize the
function ( )∑ ∆=i
i fQ .
1M 10M 100M 1G-80
-70
-60
-50
-40
-30
-20
-10
0
∆i(f)
|S
21
|[d
B]
Frequency [Hz]
ComputedMeasured
Figure 7 Example of computed and measured S-parameter.
Figure 8 Printed circuit board for the characterization ofcommercial surface mounted filters with three calibration
traces and four different filters.
In order to validate the method a number of different surfacemounted filters were studied. The filters were placed on aspecial circuit board, Figure 8, on which also conductors forcalibrating the network analyzer were present. Themeasurement procedure that was used involves calibration tothe footprint of the examined component by a through,reflection and match calibration method. When the calibrationwas done the four S-parameters S11, S21, S12 and S22 for eachsurface mounted filter were measured.
Lumped circuit generation for a feed through capacitor filter
For one of the studied commercial filters of the type feedthrough capacitor the error function minimization procedureexplained above gave component values to the assumednetwork resulting in a lumped circuit model shown in Figure9. The schematic also shows a 50Ω generator and a matchedload used when running the SPICE circuit simulation.
0 0
C1
220nF
R3
50
V1
1V
R1
50
R2
5M
L4
280PH
Figure 9 A derived lumped circuit model of a feed throughfilter used in the computation (note that in SPICE syntax
M or m stands for the prefix milli-).
1M 10M 100M 1G 3G-80
-70
-60
-50
-40
-30
-20
-10
|S21
|[dB
]
Frequency [Hz]
ComputedMeasured
Figure 10 Insertion loss for a feed through filterdetermined using a circuit model and measurements.
With this developed lumped circuit description of the feedthrough capacitor filter the behavior of the filter can easily besimulated in an electronic circuit simulator such as SPICE. InFigure 10 the results from a simulation of insertion loss isshown. For comparison the measured insertion loss for thecommercial filter is shown in the same graph.
Lumped circuit generation for a series inductance filter
A surface mounted series inductor filter was included in thestudy. The assumed network for this filter was enhancedstepwise by adding discrete components one at a time. The
lumped circuit model for the filter finally chosen is shown inFigure 11 together with the generator and load used insimulation.
0 0
C2
1.15pF
L2
2.9uH
V2
1V
R4
50
R5
50
R6
2000
Figure 11 A derived lumped circuit model of a seriesinductance filter used in the computation.
MC3resultD.opj
100k 1M 10M 100M 1G 5G-80
-60
-40
-20
0
Inse
rtio
nL
oss
[dB
]
Frequency [Hz]
Filter L1, L model
Filter L1, L-C model
Filter L1, L-C-R model
measured data
Figure 12 Insertion loss for a series inductance filterdetermined using different circuit models and
measurements.
The different lumped circuit models with increasingcomplexity were used when simulating the insertion loss ofthe filter in a circuit simulator. The resulting simulatedinsertion loss is shown in Figure 12 together with measuredvalues. As can been seen in Figure 12 the first two attempts toa network design both failed to agree with measured data at afrequency close to 100 MHz. The third attempt to choose anetwork capable of adapting to the correct behavior was moresuccessful. After minimizing the error function for thisnetwork a reasonable good fit to the measured insertion losswas obtained. In this case there is a very good agreementbetween the developed lumped circuit model of the filter andthe measurements up to at least 4 GHz.
Conclusions
The developed methods for generating lumped circuit modelsfor transmission like barriers as well as for barriers withunknown geometrical shape have been verified againstmeasurements on prototype printed circuit boards. Goodagreement between SPICE simulation and measurement wasfound for frequencies up to 1 GHz. For models of transmissionlike barriers the upper frequency limit for good agreement isdepending on the number of cascaded sections in the circuitmodel. For the 100 mm parallel conductor crosstalk, anagreement within a few dB:s up to 1 GHz was achieved with16 cascaded sections. The method of adapting an assumednetwork to measured S-parameters gave an good agreement upto 1 to 4 GHz when simple networks were assumed. Betteragreement in the high frequency range can of course bereached by assuming a more complex network.
Acknowledgements
Thanks to Lennart Hasselgren at IVF (The Swedish Instituteof Production Engineering Research) for running SPICEsimulations with the generated lumped circuit models.
References
[1] L. W. Nagel, ”SPICE2: A Computer Program to SimulateSemiconductor Circuits”, Memorandum No. M520, May1975.
[2] R.F. Harrington, ”Field computation by moment methods",Macmillan New York 1968.
[3] P.P. Silvester and R.L. Ferrari, ”Finite elements forelectrical engineers", 2nd ed., Cambridge University Press,1990.
[4] Pei-bai Zhou, ”Numerical Analysis of ElectromagneticFields”, Springer-Verlag 1993.
[5] R. Laroussi and G. I. Costache, ”Finite-Element MethodApplied to EMC Problems”, IEEE Transactions onElectromagnetic Compatibility, Vol. 35, No. 2, pp. 178-184,May 1993.
[6] C. Wei, R. F. Harrington, J. R. Mautz and T. K. Sarkar.,”Multiconductor Transmission Lines in MultilayeredDielectric Media”, IEEE Transactions on Microwave Theoryand Techniques, Vol. 32, No. 4, pp. 439-450, April 1984.
[7] F. M. Tesche, M. V. Ianoz and T. Karlsson, ”EMCanalysis methods and computational models”, John Wiley &Sons, inc., 1997.
Paper B
A comparison of measured andsimulated current distributions on
a printed log-periodic antenna
Authors:Stewart Jenvey and Urban Lundgren
Reformatted version of paper originally published in:Symposium Record, Antenn 00, Nordic Antenna Symposium, (Lund, Sweden), 2000
c© 2000, SNRV, Reprinted with permission.
51
52 Paper B
A COMPARISON OF MEASURED AND SIMULATED CURRENT DISTRIBUTIONON A PRINTED LOG-PERIODIC ANTENNA
Stewart Jenvey* and Urban Lundgren†
*Department of Electrical and Computer Systems Engineering,Monash University, Wellington Rd, Clayton, Australia 3168
† Systemteknik, Luleå Tekniska Universitet,Universitetsområdet, Porsön, Luleå, Sweden
SUMMARY
This paper describes the use of a position scanned magnetic field probe to investigate thecurrent distributions on the surface of a log periodic dipole antenna (LPDA) which wasconstructed on printed circuit board. Measurements of the magnetic field magnitude andphase at each point of the scan were used to derive the time variant instantaneous values ofthe magnetic field and the instantaneous currents on the LPDA. These were used in theanalysis of the antenna design. The wave nature of the current distribution can be readilyobserved and problems with the design such as standing waves on the feeder lines arehighlighted for attention in a revised design. Measured current distributions are comparedwith predicted distributions obtained from Method of Moments (MOM) analysis of the LPDAstructure. Measured and predicted far field radiation patterns are also compared.
1. INTRODUCTION
Log periodic dipoles are a common, linearly polarised, broadband type of antenna. Thedesign, construction and testing of a printed LPDA and the use of magnetic scanning to aid inoptimising the design are discussed in this report.
The antenna examined was made using printed circuit technology so that a prototype could bequickly created and that subsequent modifications to the design based on the investigationsdescribed herein could be quickly and easily incorporated in the antenna design. New versionsof the prototype antenna could then be quickly and cheaply produced. The basic designparameters and construction of the printed of the antenna are addressed Section 2.
The use of the printed circuit board to support the radiating elements and to separate the twostrips forming the parallel wire transmission feeder line created a mixed dielectricenvironment. Consequently the current distribution on the antenna, and the LPDA impedanceand radiation patterns, are different from those that would exist for a wire LPDA operating atthe same frequencies in a free space environment. These current distributions on the printedantenna were studied experimentally by examining the magnetic field of the antenna asdescribed in Section 3. The measurement results are presented in Section 4.
Section 5 discusses the MOM modelling of the LPDA and Section 6 then compares themeasured currents with those predicted by the MOM analysis. The computed and predictedradiation patterns resulting are also compared.
2. DESIGN AND CONSTRUCTION OF THE LPDA
The design principles of the LPDA are well established [1, 2]. The LPDA (as shown in Figure1.) is an array of dipoles connected to a common transmission line fed from the apex of thearray.
Feed Lndn
Xn+1
Xn
Figure 1. A typical LPDA
The geometry of an LPDA is defined in terms ofXn (the distance of thenth dipole in the arrayfrom the array apex), andLn (the length of thenth dipole), and by the relationships
n
n
n
n
n
n
L
d
L
L
X
X
2
11
=
== ++
σ
τ
which ensures that the structure will thus have impedance and radiation patterns that scale asf0τn , wheref0 is a reference frequency. Ifτ andσ are small (σ around 0.06-0.22 andτ in therange 0.8-0.97) then the impedance and the radiation patterns remain reasonably constant withfrequency over the frequency range in which the shortest to longest elements become half-wave resonant. Values ofτ = 0.8 andσ = 0.16 were used in this case.
The transmission line from the feed must alternate which side of the line connects to whichside of the dipole in order to get the correct phasing to create an antenna that radiates in thedirection of the array apex. By using double-sided printed circuit board to construct the LPDAthe transmission line could consist of two strip conductors, one on either side of the board. Byputting one half of each dipole on either side of the board and connecting it to thetransmission line strip, and alternating which half dipole went on which side of the board, thealternating feed connection was obtained (see Figure 2).
Fibreglass board (εr = 4.5) 1/16 inch thick was used to construct the LPDA. The requiredfrequency range of the LPDA was 900 MHz to 3GHz which meant that the dipole elements,based on a free space wavelength, were of a length that could easily be accommodated on the200 mm by 300 mm printed circuit board used. The exact resonant length of the individualdipoles was initially unknown due to the mixed dielectric environment in which the dipolestanding waves exist. The effective dielectric constant for the LPDA transmission line couldbe calculated from the formula for the effective wavelength [3] in microstrip of the same trackwidth and half the dielectric thickness of the board used for the LPDA.
Figure 2 The LPDA studied
3. MAGNETIC FIELD DISTRIBUTION MEASUREMENT TECHNIQUE
The magnetic field distribution on the LPDA was measured with the aid of a magnetic probe(a shielded loop antenna), a coordinate table and a vector network analyzer. The smallmagnetic field probe was positioned close to the upper surface of the printed antenna andmoved stepwise in a rectangular grid pattern. The center of the loop probe was about 10 mmabove the surface of the printed antenna. Because the coordinate table surface formed ametallic ground plane, the probe was attached to an extension arrangement making it possibleto scan the antenna to the side of the table (see Figure 3). This arrangement reduced theinfluence of the ground plane.
The LPDA was connected to port 1 of a Vector Network Analyser (VNA) and the loopantenna was connected to port 2. The LPDA was excited with CW from port 1 of the VNAand S21 was measured as the loop antenna was scanned in a raster pattern over the surface ofthe antenna.
Measurement equipment used:• Rohde & Schwarz ZVR Vector Network Analyzer, frequency range 9 kHz – 4 GHz• EMC-scanner, coordinate table manufactured by Detectus AB, Sweden• PC software for controlling coordinate table and network analyzer, made by Detectus AB• 903B, 1.0 cm diameter electrically shielded H-field loop probe from EMCO probe set
model 7405
The probe used was a 1 cm diameter loop antenna, with the polarization axis normal to theplane of the loop, positioned in parallel with the plane of the antenna. In Figure 3 the probe isX polarized to be sensitive to currents along the dipole. Rotating the loop ninety degreesabout the vertical axis to be Y polarised enabled the probe to respond to currents on thetransmission line joining the dipoles.
The PC software for controlling the coordinate table utilised modified GPIB instruction filesto enable automatic control of the vector network analyzer in order to make the entiremeasurement procedure automatic. Between each positional step the magnitude or phase forsignal transmission from the antenna to the probe was measured. Two measurement scans per
X
Y
polarisation were necessary to obtain both magnitude and phase information (due tolimitations in the control software). The procedure was repeated for each of the X and Ypolarisations of the magnetic field and at each measurement frequency.
Figure 3. Scanning the LPDA using the coordinate table.
4. MANETIC FIELD MEASUREMENT RESULTS
Figure 4 is a three dimensional plot of the X directed magnetic field measured at 1132 MHzclose to the surface of the LPDA. As the tangential magnetic field at the surface of aconductor is a measure of the current density, Figure 4 therefore graphically shows the currentdistribution on the dipoles of the LPDA.
Figure 4 X directed magnetic field at 1132 MHz, close to the LPDA surface
Figure 5 shows a two dimensional contour plot of the X and Y directed components of thesurface magnetic field at 1132 MHz. Dark areas surround the strongest fields. Thiscorresponds to the current distributions on the dipoles and the feed line respectively.
The measured magnitude of magnetic field at 1.132 GHz
Figure 5(a) X Polarised Figure 5(b) Y Polarised
5. METHOD OF MOMENTS MODELLING OF THE LPDA
MOM modelling of a wire LPDA in free space is straightforward, but for analysis of thisantenna the MOM package was unable to handle the mixed dielectrics (air and PC boardsubstrate).
The current distributions on the dipoles are dependent on the antenna near fields that exist inthe mixed dielectric environment. An effective uniform dielectric constant was used with theMOM program to represent this mixed dielectric environment in order to calculate the currentdistributions. The value for this effective dielectric constant was determined from themagnetic scanning of the LPDA by observing which dipoles of what length went resonant atwhich frequency. This gave a value ofεr = 1.84
The current distributions determined in the MOM analysis were then used to calculate theradiation patterns assumingεr = 1.0 (as the radiation takes place principally in free space).The match between the measured and predicted radiation patterns is shown in Figure 6.
6. COMPARISON OF MEASURED AND PREDICTED RESULTS
6.1 Far Field Azimuth and Elevation Radiation Patterns
Figure 6 E Plane and H Plane radiation patterns of the LPDA at 1132 MHz
The computed and measured E plane and H plane radiation patterns of the LPDA are shownin Figure 6. There is good agreement between them for the main lobe of the pattern but somevariance is seen between the back lobes. Differences between the actual current distributionand that derived from the MOM analysis must be resolved to improve this back lobe match.
6.2 Current Distribution on the Dipole Elements
The currents on the individual dipoles are shown in Figure 6. The currents are plottednormalised and in dB as the surface magnetic fields were measured in relative levelsexpressed in dB.
LPDA E Plane Pattern
-30
-25
-20
-15
-10
-5
0
-200 -100 0 100 200Degrees
dB
Az-measAz-calc
LPDA H Plane Radiation Pattern
-30
-25
-20
-15
-10
-5
0
-200 -100 0 100 200Degrees
dB El-measEl-calc
Figure 7Current distributions on two of the dipole elements (Solid-measured, broken-predicted)
The shapes of the current distributions match except for an asymmetry in them which isassumed to be due to the presence of the co-ordinate table (see Figure 3).
Some of the absolute levels of the individual dipole currents do not correspond well. This maybe due to a non-proportional relationship between the magnetic field strength (measured witha finite sized loop antenna just above conductors of varying width) and the total current onthat conductor. It is expected that investigations with a smaller loop antenna passed closer tothe surface of the antenna and analysis with other numerical methods (eg Finite DifferenceTime Domain) will reveal the reason for this mismatch.
7. CONCLUSIONS
Magnetic scanning of the LPDA has been used in order to get an understanding of the LPDAoperation from its current distributions. The current distributions observed on the prototypeantenna were used to see where problems existed with the design (such as the standing wavesseen on the feed line in Figure 5(b)). They were also used to determine information on theeffective relative dielectric constant to be used in MOM analysis of the antenna and itsradiation characteristics. That the patterns predicted matched reasonably well givesconfidence in this approach.
Observed differences in the magnitude of the measured and computed currents on the dipolesrequire further investigation.
8. REFERENCES
1. R. C. Johnson and H. Jasik,Antenna Engineering Handbook, McGraw-Hill, 2nd Ed.,1984.2. W. L. Stutzman and G. A. Thiele,Antenna Theory and Design,John Wiley and Sons, 2nd
Ed.3. S. Y. Liao,Microwave Circuit Analysis and Amplifier Design, Prentice-Hall, New Jersey,
1987.
Normalised Current Element 3
-25
-20
-15
-10
-5
0
-0.1 -0.05 0 0.05 0.1metres
dB
Normalised Current Element 5
-25
-20
-15
-10
-5
0
-0.08 -0.03 0.02 0.07metres
dB
Paper C
Analysis of Printed AntennaStructures using the PartialElement Equivalent Circuit
(PEEC) Method
Authors:Jonas Ekman and Urban Lundgren
Reformatted version of paper originally published in:Symposium Record, EMB01, (Uppsala, Sweden), 2001
c© 2001, SNRV, Reprinted with permission.
59
60 Paper C
Analysis of Printed Antenna Structures using thePartial Element Equivalent Circuit (PEEC) Method
Jonas Ekman Urban Lundgren
EISLAB, Lulea University of Technology, SE-97187 Lulea, Sweden
[email protected] [email protected]
Abstract| The partial element equivalent circuit(PEEC) method is a electromagnetic simulationtechnique suitable for mixed circuit and eld prob-lems. The technique is numerically equivalent to amethod of moments solution using Galerkin solution.In this paper, the PEEC method is illustrated andapplied to printed antenna structures where mea-surements are compared to simulations and analyti-cal solutions. The possibility to use simplied PEECmodels to decrease computation time is discussedwith illustrative examples.
I. Introduction
The Partial Element Equivalent Circuit (PEEC)method is a full wave technique for the solution ofmixed circuit and eld problems in both the timeand frequency domain. The international interestfor the method has been gaining rapidly for thepast years but in the Nordic countries the researcheort has been low. This paper can be consideredas an fundamental introduction to this electromag-netic computation technique.The method is based on the conversion of the
Mixed Potential Integral Equation (MPIE) to thecircuit domain. By using a specialized discretiza-tion, the original structure is converted into a net-work of discrete inductances, capacitances and re-sistances, called the partial elements. The ca-pacitive(electric eld) and inductive(magnetic eld)couplings are modelled using partial mutual ele-ments which results in a electromagnetic correctmodel. The partial elements are calculated ei-ther by using numerical integration techniques orsimplied closed form equations. The resultingequivalent circuits are solved by using a commer-cial circuit simulation program like SPICE. Theuse of SPICE-like circuit solvers facilitates the in-clusion of discrete components, transmission lines,current/voltage source etc in the resulting PEEC
model.Since the method is based on an integral equation
formulation the analysis of 'open air' problems likeradiation from antennas are computationally eÆ-cient. To have eÆcient electromagnetic computa-tion techniques for printed antenna structures canbe important to speed up the development of forexample mobile embedded internet systems.
II. Derivation of the PEEC model
The starting point for the theoretical derivationis the summation of the electric eld, E, at a eldpoint, r, in a multiconductor system expressed interms of the vector magnetic potential A and thescalar electric potential at a source point r0.
E(r; t) = @
@tA(r; t)r(r; t) (1)
For a system containing K conductors the free-space Green's function retarded potentials are givenby
A(r; t) =KXk=1
4
Zvk
J(r0; t0)
jr r0jdvk (2)
and
(r; t) =KXk=1
1
4"
Zvk
q(r0; t0)
jr r0jdvk (3)
, where
t0 = tjr r0j
v(4)
, denotes the retardation time in the medium withpropagation speed v. The charge density q con-siders both the bound charges and the chargesbounded in the dielectric regions. The expressionfor the current density J must be modied [12]to include the conduction current density JC and
the polarization current density in the dielectricmedium according to
J = JC + "o("r 1)@E
@t(5)
Since the total electric eld at the surface of a con-ductor, E(r; t), can be expressed using the currentdensity and conductivity, ; and if no incident eldis considered equation (6) is obtained by combiningequations (1), (2), (3).
J(r; t)
+
KXk=1
4
Zvk
@
@t
J(r0; t0)
jr r0jdvk+
KXk=1
"o("r 1)
4
Zvk
@2
@t2E(r0; t0)
jr r0jdvk+ (6)
KXk=1
1
4"or
"Zvk
q(r0; t0)
jr r0jdvk
#= 0
To solve the system of equations in (6), the currentand charge densities are discretized into volume andsurface cells respectively, Figure 1(top). The solu-tion requires also that the surface cells are shiftedhalf a cell length to the volume cells as indicatedin Figure 1(top). The current volume cells leadthe current between the nodes and the charge sur-face cells represent the node charge. Inside the cellsthe variables are constant. Applying the Galerkin
RmLp i(t)
Cp12Cp12
Cp11Cp11 Cp22Cp22
Fig. 1. Conductor discretisation(top) and correspondingPEEC model(bottom)
method K equations are obtained for the K vol-ume cells of the structure. The interpretation ofthese equations as a loop [15] leads to the structureof the equivalent circuit for a PEEC cell, Figure1(bottom).The practical implications of this solution re-
quires that xed nodes are placed on the structureunder test. From these nodes two partitions are
made. The inductive partition is based on the vol-ume cells between two subsequent nodes. And, thecapacitive partition is based on the surface cells as-sociated to each node. This is illustrated in Figure2. The two partitions are used for the calculation
Fig. 2. PEEC method procedure
of the partial elements according to the follwowingsection.
A. Partial elements
The discrete components in Figure 1 and 2 aredenoted partial elements and the calculations areperformed based on the geometrical shape of theinductive and capacitive partitions using numericalintegration or closed-form equations.
A.1 Partial resistance
In Figure 1(bottom), Rm is the volume cell dcresistance of the volume cells in the inductive par-titions. The resistance in the PEEC model corre-spond to conductive losses and the inclusion resultsin a (R)PEEC model. The resistance is simply thedc resistance of the conductor calculated as:
Rm =lm
am (7)
where lm is the volume cell length in the directionof the current ow, am is the cell cross-section and is the conductivity of the specic material.
A.2 Partial inductance
In Figure 1 and 2 Lp is the partial inductance cal-culated using the inductive partition and equation(8).
Lpmn =
aman
1
4 jr r0j
Zam
Zan
Zlm
Zln
dlm dlndamdan
(8)If n = m, equation (8) represents the partial selfinductance of the m : th cell. This is the internalinductance of the volume cell and is connected inseries with the partial resistance between the nodes.
If n 6=m, equation (8) represents the partial mu-tual inductance between the m : th and the n : thcell. This corresponds to the magnetic eld cou-pling between the volume cells. In SPICE this eectis modeled using the K command. This represen-tation of the partial mutual inductance is instanta-neous, meaning that the current in one conductoraects all other conductors at once. For volumecells that are far apart, or for high frequencies, thisrepresentation becomes unvalid since the eld cou-plings occur at nite travel times. This problemis solved by the use of delayed, retarded, current-or voltage sources [11]. This is described in Fig-ure 3(left) where current i2 1 is instantly inducedin conductor 2 due to the current in conductor onethrough the partial mutual inductance Lp12. In Fig-ure 3(right) the partial mutual inductance is cou-pled to a current source with a specied delay time.
But since delayed sources are not supportedby most commercial circuit simulation tools likeSPICE a specialized solver must be used, as de-scribed in [16]. The inclusion of partial inductancesand retardation in a PEEC model is denoted a(Lp,)PEEC model.
Since the calculation of the partial inductancesusing equation (8) is both complex and time con-suming a set of closed form equations has been pre-sented, see [1] and [7].
Lp1 2Lp1 2
i1i1 i1i1
Lp2 2Lp2 2
Lp11Lp11
Lp2 2Lp2 2
Lp11Lp11
i2 _ 1i2 _ 1
i2 _ 1i2 _ 1
Fig. 3. Magnetic eld coupling using delayed current sources
A.3 Partial capacitance
Cp in Figure 1 and 2 is the partial capacitancethat can be calculated from the coeÆcients of po-tential Pij [5], where
Pij =1
"oasiasj
1
jr r0j
Zasi
Zasj
dasidasj (9)
The coeÆcients of potential is an alternative capaci-tive representation relating the surface potential, V,to the surface charge, Q, according to
V = PQ (10)
The regular capacitance notation, C, is the inverseto the coeÆcients of potential since
CV = Q (11)
If i = j in equation (9), the partial self coeÆcient ofpotential of the j : th surface cell is calculated. Thiscapacitance is connected between the correspondingnodes, surface cell, and the node at innity.If i 6=j, equation (9) represents the partial mu-
tual coeÆcient of potential between the i : th andthe j : th surface cell. This correspond to the elec-tric eld coupling between the surface cells. Sinceequation (10) is a instantaneous relationship the in-version of the coeÆcients of potential matrix, P , tothe capacitance matrix, C, can only be done whenretardation times are neglected. Then the diago-nal elements in the C matrix represents the partialself capacitance to each surface cell, Cpii in Figure1 and 2. The o diagonal elements are the partialmutual capacitances between the surface cells, Cpijin Figure 1 and 2.When retardation times must be considered, de-
layed sources are used in a similar manner as forthe partial inductances [11]. The inclusion of ca-pacitances or coeÆcients of potential in a PEECmodel is denoted a (C)PEEC or (P)PEEC modelrespectively.Closed form equations for the calculation of the
partial coeÆcients of potential, thus also partial ca-pacitances, has been presented in [2].
A.4 Extended PEEC models
There are numerous features with the PEECmethod that is not covered in this paper, as an ex-ample:
For PEEC models where dielectric regions mustbe considered, the PEEC method has been ex-tended through the use of dielectric cells [12]. The excitation of the structure is usually per-formed using current- or voltage sources. In [13]the formulation was extended to include excitationusing incident elds. An eÆcient Skin eect model were presented in[17].
III. Examples
In this section two examples are presented andthe corresponding PEEC models are discussed. Inthe PEEC model examples the partial elementshave been calculated using the closed form equa-tions suggested in the previous section. For theretarded current/voltage source simulation a spe-cialized solver has been used [16]. The accuracy ofthe PEEC models are compared against analyticalsolutions and measurements.
A. Half wavelength dipole
The 2dipole and monopole antennas are com-
mon radiating structures in electronic systems andis therefor of great importance. The simplicity ofthese structures makes it an ideal example for sim-ple PEEC models.To model the resonance frequency of an 40 cm
free space 2dipole, the upper frequency limit, cor-
responding min, must be specied. This must beknown to assure that the dimension of the volumeand surface cells in the PEEC model does not ex-ceed min
10. The use of 20 mm volume elements makes
the PEEC model valid up to 1.5 GHz and since thetheoretical resonance frequency for the dipole is 375MHz, this is choosen.
The rst step in the modeling is to make the in-ductive and capacitive partition based on the celllengths. This is shown in Figure 4 where all induc-tive volume elements are 20 mm long (the dipolecross-section was choosen to 1x1 mm). Since thecapacitive cells are shifted half a cell length to theinductive cells the two antenna elements consists of11 capacitive cells each (two 10 mm and nine 20mm). Second, all partial elements are calculatedusing the proposed closed form equations. Third,the inductive and capacitive elments are combinedinto the PEEC model, as in Figure 4, and the de-sired analsis is performed.
Dipole
Inductive partition Capacitive partition
Combine into complete PEEC model
Place nodes
+ -
+ -
Fig. 4. The inductive and capacitive partitions for the
2dipole example
At rst a simplied (R,L,C)PEEC model for the2dipole was implemented without the partial mu-
tual inductive and capacitive elements. The resultsis shown in Figure 5 where the resonance frequencyis low by 20 % if no magnetic eld coupling(dottedline) is included or high by 32 % if no electric eldcoupling(dash-dot line) is included compared to thecorrect solution, straight line in Figure 5. The per-formance of the model is improved by the use ofall the partial mutual elements and the resonancefrequency is predicted to 358 MHz (4.5 % o), Fig-ure 5 (dash). If this PEEC model is upgraded toa retarded formulation using nite trave times be-tween the partial mutual elements the resonancefrequency is predicted to 376 MHz (< 0.3 % o),Figure 5 (solid).The inclusion of time retardation in the PEEC
model also introduces damping in the equivalentcircuit. This is clearly visible in Figure 5 where thedrive current for the non retarded formulation hasa very high Q-value compared to the retarded for-mulation. Since the focus is on the PEEC method,the actual Q-value and drive current amplitude hasnot been investigated.
B. Patch antenna
The patch antenna is a very common structurein antenna applications where they appear as sin-gle patches or in array formulations. This struc-ture is more complicated compared to the dipoleantenna since the PEEC model must include (1) a
Fig. 5. Simulated resonance frequency for
2dipole using
dierent PEEC models compared to analytical solution(line)
nite ground plane, (2) the dielectric medium be-tween the planes and (3) a two dimensional currentdistriubution to correctly model the patch antenna.
To model the ground plane and the patch antennausing two dimensional current distribution [6], onex-directed and one y-directed inductive partition ismade using the xed nodes in a similar manner asdescribed in Figure 2. This model could also be ex-tended to include currents in the thickness of thepatch and ground plane. But since three dimen-sional PEEC models are complicated the two di-mensional representation was choosen.
In this example a 35 m thick 62 99 mm patchantenna located on a 1.55 mm dielectric substrate,"r = 4:5, over a ground plane is modeled. Thepatch is shown i Figue 6 where the feeding pointis marked with a -symbol. The basics of the twodimensional equivalent circuit for the patch and theground plane is shown, in the gure top right cor-ner, where the partial self inductances and the pa-tial resistances are displayed. Note that the partialself capacitances and all partial mutual elementsare excluded in the gure, for simplicity reasons,but not in the PEEC model.
The dielectric medium is not modelled using cells,instead an eective "r has been used [10]. As forthe one dimensional case, the dipole, the partial el-ements were calculated using closed form equations.
Fig. 6. Example patch antenna with cut to show the discretecomponents in the PEEC model
For this PEEC model the re ection coeÆcient,S11, was measured using a Rhode & Schwarz ZVRnetwork analyzer. The measurements are comparedto a (Lp,P,R,)PEEC model as shown in Figure 7.The discretisation of the patch antenna into a
max cell size of 20 20 mm results in a upper fre-quency limit of 1.5 GHz. This cell size is ne enoughto model the rst resonance around 1.2 GHz, indi-cated by measurements. As can be seen in the g-
Fig. 7. Re ection coeÆcient for patch antenna where thedashed line is measurements and the solid line is simu-lation
ure the maesurements and simulations are in closeagreement, except for the rippel and the higher Q-value for the measured re ection coeÆcient.The PEEC model used in the simulations consists
of : 154 partial self coeÆcients of potential 1542 partial mutual coeÆcients of potential 132 x-directed and 140 y-directed partial self in-ductances 1322 + 1402 partial mutual inductances 272 partial resistancesAnd takes approximately 45 minutes to run on aPIII/750 MHz laptop computer. To speed up thecomputation time it is possible to exclude partialmutual elements with weak coupling coeÆcients, <0.15, without eecting the predicted resonance fre-quency more than +/- 5 %. The speed up cut com-putation times by approximately 25 minutes. Toimprove the accuracy of the PEEC model, dielec-tric cells and a ner partition could be used.
IV. Conclusions
The PEEC method has been shown to be a verypowerful simulation technique for combined circuitand electromagnetic eld problems. The rst ex-ample displays the possibility to make PEEC mod-els by using closed form equations to calculate thepartial elements and a free version of SPICE as thesolver. This feature makes the method possible touse in education and for simple design tasks. How-ever, the method require a retarded circuit solver tobe considered a full wave method comparable witha method of moments solution. The application ofPEEC's to antennas is a valuable tool in many areaswhere antenna resonance frequencies are of impor-tance. The paper shows that antennas are modelledwith good agreement compared to analytical solu-tions and measurements.
References
[1] A. E. Ruehli, "Inductance Calculations in a Complex Inte-grated Circuit Environment", IBM Journal of Research andDevelopment, vol. 16, no. 5, sep. 1972, p. 470-81.
[2] P. A. Brennan, A. E. Ruehli, "EÆcient Capacitance Cal-culations for Three-Dimensional Multiconductor Systems",IEEE Transactions on microwave Theory and Techniques,vol. MTT 21, no. 2, feb. 1973, p. 76-82.
[3] A. E. Ruehli, "Equivalent Circuit Models for Three-Dimensional Multiconductor", IEEE Transactions on mi-crowave Theory and Techniques, vol. MTT 22, no. 3, march1974, p. 216-21.
[4] L. W. Nagel, SPICE2: A Computer Program to SimulateSemiconductor Circuits, Memorandum No. M520, May 1975.
[5] P. A. Brennan, A. E. Ruehli, "Capacitance Models for Inte-grated Circuit Metallization Wires", IEEE Journal of SolidState Circuits, vol. SC 10, no. 6, dec. 1975, p. 530-6.
[6] P. A. Brennan, N. Raver, A. E. Ruehli, "Three-DimensionalInductance Computations with Partial Element EquivalentCircuits", IBM Journal of Research and Development, vol.23, no. 6, nov. 1979, p. 661-68.
[7] A. E. Ruehli, P. K. Wol, "Inductance Computations forComplex Three-Dimensional Geometries", IEEE Interna-tional Symposium on Circuits and Systems, IEEE, NewYork, 1981, 3 vol. p. 16-19, vol. 1.
[8] S. Daijavad, H. Heeb, S. Janak, A. Ruehli, "SimulatingElectromagnetic Radiation of Printed Circuit Boards", 1990IEEE International Conference on Computer-Aided Design.Nov. 11-15 1990, p. 392-395. ASBN : 0-8186-2055-2.
[9] H. Heeb, A. E. Ruehli, "Approximate Time-Domain Mod-els of Three-Dimensional Interconnects", Proceedings, 1990IEEE International Conference on Computer Design, VLSIin Computers and Processors, IEEE Comput. Soc. Press, LosAlamitos, CA, USA, 1990, p. 201-5.
[10] S. Jenvey, U. Lundgren, "A Comparison of Measured andSimulated Current Distributions on a Printed Log-PeriodicAntenna", Symposium Record, Antenn 00 Nordic AntennaSymposium. Lund, Sweden, 2000.
[11] A. E. Ruehli, H. Heeb, "Retarded Models for PC Board In-terconnects - or How the Speed of Light Aects Your SPICECircuit Simulation", 1991 IEEE International Conferenceon Computer-Aided Design. IEEE Comput. Soc. Press, LosAlamitos, CA, USA, 1991, p.70-3.
[12] A. E. Ruehli, "Circuit Models for Three-Dimensional Ge-ometries Including Dielectrics", IEEE Transactions on Mi-crowave Theory and Techniques, vol. 40, no. 7, july 1992, p.1507-16.
[13] J. Garrett, C. Paul, A. E. Ruehli, "Circuit Models for 3DStructures with Incident Fields", 1993 International Sympo-sium on Electromagnetic Compatibility Symposium Record,IEEE, New York, NY, USA, aug. 1993, p. 28-32.
[14] J. Garrett, C. Paul, A. E. Ruehli, "Inductance Calculationsusing Partial Inductances and Macromodels", Atlanta 1995.EMC - a Global Concern. IEEE 1995 International Sympo-sium on Electromagnetic Compatibility, Symposium Record.IEEE, New York, NY, USA, 1995, p. 23-8.
[15] J. E. Garrett, Advancements of the Partial Element Equiva-lent Circuit Formaulation, Ph.D. Dissertation, The Univer-sity of Kentucky, 1997.
[16] A. Gorisch, G. Wollenberg, "Analysis of 3-D InterconnectStructures with PEEC Using SPICE", IEEE Transactionson EMC, vol. 41, no. 4, nov. 1999, p. 412-7.
[17] A. Cangellaris, K. M. Coperich, A. E. Ruehli, "EnhancedSkin Eect for Partial Element Equivalent Circuit (PEEC)Models", IEEE Transactions on Microwave Theory andTechniques , v48, n9, sep, 2000, IEEE, Piscataway, NJ, USA,p 1435-1442.
Paper D
SPICE models of barrier comparedto measured data
Authors:Urban Lundgren, Jan Carlsson and Jerker Delsing
Reformatted version of paper originally published in:Symposium Record, 2001 IEEE International Symposium On Electromagnetic Compat-ibility, (Montreal, Canada), 2001
c© 2001, IEEE, Reprinted with permission.
67
68 Paper D
SPICE models of barrier compared to measured dataUrban Lundgren Jan Carlsson Jerker Delsing
Luleå University of Technology SP Swedish National Testing & Research Institute Luleå University of TechnologyRegnbågsallén Brinellgatan 4 Regnbågsallén
SE-971 87, Luleå, SWEDEN SE-501 15, Borås, SWEDEN SE-971 87, Luleå, SWEDEN
Abstract: SPICE models of electromagnetic zone barrier aredevised. The models are based on data from 2D and 3DMaxwell equation solvers. A transfer impedance approachmodeled incident electromagnetic waves in SPICE. Testsystems using D-sub connectors, passive surface mountedfilters and encapsulation was designed. Test systemverification measurements were made in a fully anechoicchamber. The coupling through the zone barriers wasmeasured. Good agreement was found between simulated andmeasured data.
Introduction
The focus for electronic system designers is on productfunctionality. Here EMC aspects are hard to approach usingfor the electronic engineer well known tools such as SPICE.This work thus focuses on building SPICE models forelectromagnetic zone barriers enabling SPICE simulations ofradiated power and immunity to incoming disturbances. Thedatasheets from the manufacturer seldom gives a completedescription of the electrical characteristics of a component.For instance for a filter the insertion loss is often the onlygiven performance parameter. To be able to make accurateSPICE models for a component it is necessary to know theentire scattering parameter matrix. It is then desired to find alumped component circuit with the behavior described by theS-parameters.
SPICE models has been developed for commercially availablecomponents that can be regarded as EMC barriers. Surfacemounted filters and shielded connectors and cables areexamples of such components. Models have also beendeveloped for EMC barriers that appear in a circuit due to thelayout of circuit. This includes separation of traces on aprinted circuit.
The modeling approach for the different kinds of barriers haveto by chosen with care to obtain useful data with a limitedamount of effort put in. Here some a priori knowledge isnecessary. A skilled RF designer can from the geometry of abarrier judge if it can be regarded as a multi-conductortransmission line (MTL), cascaded MTLs or if a more generalapproach has to be chosen.
SPICE model generation techniques
For barriers that can be viewed as multi-conductortransmission lines the cross-section is divided into smallelements and the Laplace equation is solved [1]. Anassumption is made that the cross-section is uniform. From the
solution the charges on each conductor can be computed andthereby the per-unit length inductance and capacitancematrices. When the matrices are known it is a straightforwardtask to construct a circuit representation that can be used in astandard circuit simulator, see fig. 1.
z z+dz
c11dz c22dz c33dz
c12dz
c23dz
c13dz
l33dz
l22dz
l11dz
l13dz
l12dz
l23dz
ref.
1
2
3
Figure 1 Circuit segment representing a multi conductortransmission line
Barriers with complex or unknown geometry, for instancecommercial filters, cannot be viewed as transmission lines.Here another method has been used. The S-parameters of thecomponent are measured using a vector network analyzer. Thenext step is to set up a discrete circuit and then compute the S-parameters. The computed S-parameters are then comparedwith measured data for the frequency range of interest. Theweighted difference is minimized by adjusting componentvalues in an iterative scheme, searching among componentvalues in a given range [1]. To verify the generated SPICEmodels results from SPICE simulations were compared withmeasurements on the barriers. Good agreement was found [1].
Simulations
For further verification of the generated models a test systemwas designed involving several combined barriers and manyconnection ports. This made it possible to analyze the systemin many configurations. The test system consists of aterminated coaxial cable (RG58), a shielded 9-pin D-subconnector, four parallel traces on a printed circuit board and asurface mounted filter, fig. 2. The traces on the printed circuitboard were designed as 50Ω microstrips to minimizereflections and standing waves in the system causingunpredictable results in the measurements.
Two measurement ports on the PCB utilizing SMA connectorsmakes it possible to monitor both transmitted signals in amicrostrip and signals in an adjacent microstrip due tocrosstalk on the PCB. Moreover the unused measurement portcan be left open or terminated with matched impedance whichrenders 16 different test configurations that has been used inthe comparison.
Using the generated SPICE models of the EMC barriers todescribe the essential electromagnetic behavior of the testsystem in a circuit simulation, the different configurationswere analyzed in the frequency domain.
To model the coaxial cable shield leakage and D-subconnector shield leakage the transfer impedance for thosecomponents were used. For the cable the IEC 96-1A standardwas used to obtain transfer impedance values and for the D-sub connector a method was used that is described in [2].
Figure 3 The test system consisting of a printed circuitboard with parallel traces in a shielded enclosure. Asurface mounted filter is included on one microstrip and acoaxial cable is via a 9-pin D-sub connected to the PCB.
Measurements
The simulated system was implemented on a PCB andshielded by a metallic box with the shield connected to thegroundplane of the PCB microstrip lines. The SMA ports ofthe PCB were extended to the wall of the box (see fig. 3). A 1meter RG58 coaxial cable was terminated in the far end in 50Ω and in the near end connected to 4 different pins of the D-sub in sequence. The outer conductor (shield) of the cable wasin good connection over 360° with the backshell of theshielded D-sub connector. The shell of the D-sub connectorwas then grounded in the shielded box.
Figure 4 Setup of the system inside the unechoic chamber
Measurements were performed inside a fully anechoicchamber where the excitation of the test system was done withan incident electromagnetic plane wave (fig. 4). The couplingfrom the radiating antenna through the cable and connectorshield and through filters and crosstalk between parallel
00 0
0 0
0
00
0
0
0 0
0
00
00
Track2
DPAT4P16
inrefin1
outrefout1
Track3
DPAT4P16
inrefin1
outrefout1
Track1
DPAT3P16
inref
in1in2in3
outref
out1out2out3
R1
100MEG
R2
100MEG
DSUB1
DSUB9
in1in2in3in4in5in6in7in8in9
inref
out1out2out3out4out5out6out7out8out9
outref
R5
100MEG
R6
50
R3
100MEG
R4
100MEG
R7
100MEG
R8
100MEGR11
100MEG
R13
100MEG
R16
50
R18
5mL1
0.28n
R19
50
R15
50
R17
50
R14
100MEG
C2
220n
V1
1V
R10
100MEG
R20
100MEG
R21
100MEG
Filter SMA1
SMA2
Figure 2 SPICE simulation model of the system (the models for the cable and connector shield is not included in thisfigure).
Trace connected to pin 7
Trace connected to pin 8
SMA 1
SMA 2
Filter
microstrips was measured using a vector network analyzer.The electromagnetic field was generated with a bilog antennaat a distance of 3 meters from the system. The coupling for thetest system was and compared to the coupling for a singleconductor that replaced the system in the same position.
The lower frequency limit of the antenna is 30 MHz and ofcourse the corresponding 10 meter wavelength does not givegood far field conditions at the 3 meter antenna distance.Therefore the lower end of the frequency range in thismeasurement should be regarded with some caution.
First the coupling between the antenna and the singleconductor was measured as S21single, then the coupling wasmeasured with the system S21system. Because the incident fieldis giving raise to a surface current on the cable shield thesystem transfer impedance was calculated by including the 50Ω measurement system impedance:
gle
systemsystemT S
SZ
sin21
21_
50Ω⋅= (1)
The system transfer impedance was obtained for 16configurations of the system, with the center conductor of thecoaxial cable connected to each of the 4 pins of the D-subconnected to the traces on the PCB, leaving unused pins open.For each of these measuring output on SMA1 and SMA2leaving the unused connector open or with a matched load.
To have better control of the current distribution anothermeasurement was made according to the sketch in fig. 5. Thesource port of the network analyzer connected to the cabletermination causing a direct injection of current on the cableshield. The current (Iprobe) was monitored using a calibratedcurrent probe while the cable was positioned at a fix distanceof 20 mm over a groundplane. After measuring the voltagelevel at the SMA ports (USMA), the system transfer impedanceis easily calculated as:
probe
SMAsystemT I
UZ =_ (2)
Figure 5 Setup for measurement with direct injection ofcurrent onto the cable shield
Again this measurement was repeated for all 16configurations. A problem with this method is however thepoorly defined input impedance of the current injecting point.
To improve the method an EM clamp was used for injection ofcurrent on the cable shield outer surface, see fig. 6. Themetallic box and the termination of the coaxial cable were isgood contact with the groundplane making a closed loop forthe current. A calibrated current probe was used to monitor thecurrent and calculations were carried out as described above.
Figure 6 Measurement setup were the current injetion wasimproved by the use of an EM clamp (the current probe ismissing in this figure).
As a comparison measurement were made on the shielded boxonly, including the PCB inside and the D-sub and SMAconnectors. The obtained coupling values were combined withthe transfer function due to the transfer impedance data for thecoaxial cable and D-sub connector shield. This offers anotherway to calculate the system transfer impedance.
Results
The measurement results obtained in the unechoic chambershows a poor agreement with the SPICE simulations in thefrequency range 30 MHz to 60 MHz where overlapping data isavailable. Differences between 10 and 40 dB in the twoconfigurations shown in fig. 7 and fig. 8. However whencomparing with other measurements or by extrapolation of thesimulation results a better agreement is found in the higherfrequency range towards 100 MHz.
When analyzing the results from the method using directcurrent injection there seem to be a resonant behavior at about25 MHz probably due to the uncertain input impedance at thecurrent injection point (the far end of the coaxial cable). Bythe using the EM clamp for the injection of current smootherfrequency dependence is obtained.
Figure 8 Comparison of SPICE simulation results and measured data with input on D-sub pin 8 andoutput on SMA 2, SMA 1 terminated in 50 ohm
Figure 7 Comparison of SPICE simulation results and measured data with input on D-sub pin 7 andoutput on SMA 2, SMA 1 terminated in 50 ohm
System transfer impedance - Dsub pin 7 - SMA 2
-100-90-80-70-60-50-40-30-20-10
0
0.0E+00 2.0E+07 4.0E+07 6.0E+07 8.0E+07 1.0E+08
Frequency (Hz)
Impe
danc
e(d
Boh
m)
SPICE simulationDirect current injectionEM clamp current injectionCombined transfer functionsIncident plane wave
System transfer impedance - Dsub pin 8 - SMA 2
-100
-90
-80
-70
-60
-50
-40
-30
-20
-10
0
0.0E+00 2.0E+07 4.0E+07 6.0E+07 8.0E+07 1.0E+08
Frequency (Hz)
Impe
danc
e(d
Boh
m)
SPICE simulationDirect current injectionEM clamp current injectionCombined transfer functionsIncident plane wave
With the EM clamp injection method a much better agreementis obtained between the measured system transfer impedanceand the SPICE simulation, shown in fig. 7. In fig. 8 the samecomparison is not so good but exhibits an almost constantoffset.
By combining measurements for the parts of the system into acomplete system transfer impedance good agreement isobtained with SPICE simulation in fig. 8 but a remainingoffset is shown in fig. 7.
Conclusion
The measurements for the desired verification are hard to dowith one single approach. For the measurements with incidentplane wave in the unechoic chamber the problems at 30 MHzare likely to be caused by the short distance between the setupand the radiating antenna compared to the wavelength. It mayalso be affected by the poor performance of the antenna at thisfrequency.
The frequency range for the SPICE simulation was limited bythe transfer impedance data obtained for the coaxial cableshield. It is reasonable to believe that better agreement wouldbe found if the SPICE simulation could be done at 100 MHz.
The deviation between the obtained results seems to decreasewhen frequency increases in the frequency range of this study.
References
[1] Carlsson J. and Lundgren U., “An Approach to theGeneration of SPICE Models Feasible for EMC Problems”, 2000 IEEE International Symposium on ElectromagneticCompatibility, Washington, D.C., USA, 21-25 Aug, 2000
[2] Van Horck F. B. M., et al, “ A Rapid Method forMeasuring the Transfer Impedance of Connectors”, IEEEtransactions on Electromagnetic Compatibility, Vol. 40, No. 3,pp. 193-200, Aug. 1998
74 Paper E
Paper E
Characterization of ConductiveThermoplastic CompositeMaterials Using MultipleMeasurements Methods
Authors:Urban Lundgren, Jonas Ekman and Jerker Delsing
Reformatted version of paper originally published in:Symposium Record, EMC Europe 2002, (Sorrento, Italy), 2002
c© 2002, IEEE, Reprinted with permission.
75
76 Paper E
CHARACTERIZATION OF CONDUCTIVE THERMOPLASTIC COMPOSITE MATERIALS USING MULTIPLE MEASUREMENT
METHODS
Urban Lundgren Jonas Ekman Jerker Delsing EISLAB Luleå University of Technology 971 87 Luleå, Sweden Abstract - In a study of conductive thermoplastic composite materials, samples were manufactured and measured. Samples with different base polymers, filler materials and different amount of filler made it possible to generate data for many combinations. The samples were characterized in terms of their complex permittivity and complex permeability, plane wave shielding effectiveness and near electric field shielding effectiveness. As can be expected materials that shows a relatively high shielding effectiveness for a incident plane wave also in general offers shielding in the near field situation that was studied. A correlation between SE and complex permittivity was also found. I. INTRODUCTION This paper describes the work to compare measurement methods to acquire electromagnetic shielding effectiveness (SE) of conductive thermoplastic materials. The frequency range in this study is 150 MHz to 1 GHz. It is desired to find a method to establish the shielding effectiveness for a box encapsulating electronic equipment. An easy to use measurement method would allow the study of new material mixtures and the effect on shielding effectiveness of varying process parameters. As a tool for the plastic manufacturing industry this will enable improvement of electromagnetic shielding effectiveness in plastic enclosures for electronics. A good method should also be as simple as possible regarding instrumentation so that necessary investments to be able to do the measurement are kept small. Previously published work related to this problem has been found in [1] which reviews a large number of papers and summarizes in four measurement methods. As a near field measurement methods suitable for materials with insulative surface standards ASTM-ES-7 and modified MIL-STD-285 are suggested. Those two methods are also reviewed in [2]. In [2] and [3] two measurement methods are suitable involving a TEM cell. However the TEM cell limits the upper frequency to less than 1 GHz. Another method developed from the MIL-STD-285 is shown in [4]. This measurement method is suitable for both near field and far field measurements. For measurement of permittivity and permeability on a material, methods were found in [5]
and [6]. The methods are using cavity resonance or transmission line loading techniques. In [2] a method is reviewed for estimating shielding effectiveness when knowing complex permittivity and permeability of a material. Three different measurement methods are compared to test the validity of the methods. For the comparison of measurement methods a number of different composite materials were analysed in the study. Data from three of those are here used for the comparison of the three methods. Measured data are also compared to give an indication how shielding effectiveness is affected by incident field impedance and by the permittivity and permeability of the material. II. EXPERIMENTAL SETUP AND MATERIALS The three thermoplastic composite materials used in the comparison have the base polymers Polycarbonate / Acrylnitrilbutadienstyren (PC/ABS) or Acrylnitril-butadienstyren (ABS). The materials used are listed in Table I. Table I Materials for which measured data are presented in this paper Base polymer Main additive PC/ABS 1 vol. % Stainless steel fibre ABS 1 vol. % Stainless steel fibre ABS 1.5 vol. % Stainless steel fibre For each material test samples were made as boxes for near field SE, square plates for the far field SE measurement and two smaller rectangular plates for the permittivity and permeability measurement. II.1 Shielding of a Electric Near Field Source This measurement method mimics a shielding enclosure application by enclosing a radiating dipole with a box made of the sample material. This method was selected as reference since it provides the closest match to the desired application of the shielding material. The main
1
drawback of this method is the cost of making the boxes used as test samples and the expensive anechoic shielded room facility used for the measurement. Because of the desire to evaluate the usefulness of the materials in an electronic equipment encapsulation application, near field measurements were made. A high impedance source generating mainly electric field was designed and used for measurement of the near field shielding effectiveness. If the near field radiation mainly is an electrical field (high impedance) then conductive thermoplastic materials are expected to offer better shielding effectiveness than for mainly magnetic field (low impedance).
Figure 1 Transmitter for near field SE measurements
The battery powered comb generator was constructed (Fig. 1) with a fundamental frequency of 20 MHz and does produce strong frequency components well above 2 GHz. In this work the frequency range that have been used is 150 MHz to 1 GHz. The comb generator alone were placed on a table in a anechoic room and the emitted free-space signal spectra at 3 m distance were measured, see Fig. 2. This measurement data is used as the baseline in the insertion loss calculation. A Rohde & Schwarz ESPC EMC/EMI test receiver was used for this measurement. The instrument causes a step in the baseline at 500 MHz, this is due to an internal compensation in the instrument and should not affect the measured values at the peaks according the manufacturer. However, even if the levels of the peaks are affected the step is about 4 dB and this error would not alter the conclusions in this paper. The battery powered generator was enclosed in a box of the sample material with the dimensions 18 x 11 x 12 cm. To be able to access the interior of the box the boxes were made in two halves, see Fig. 3. To ensure good seal when closing the box the meeting surfaces were designed to have a male and female configuration, see Fig. 4, that offered a wave trap function. The thickness of the material is 5.0 mm in the bottom and 3.0 mm in the walls of each half of a box. The wall with thickness 3.0 mm was facing the receiving antenna.
dBµV
80
-20
-10
0
10
20
30
40
50
60
70
MHz
1000.020.0 100.0
Figure 2 Radiated electric field from battery powered
comb generator
Figure 3 Transmitter placed in box made of a
thermoplastic material
Figure 4 Cross section of box wall showing male –
female arrangement to ensure good seal when closing a box
The emitted signal spectra were measured with the generator in box located at the same position as in the baseline measurement. Finally, the shielding effectiveness is calculated as the difference between this measurement and the baseline. In this method the thermoplastic material is very close to
2
the signal source like in most shielding applications. This means that the barrier, the thermoplastic material, is in the near field of the source and the impedance of the emitted electromagnetic wave is unknown. The results from these measurements are therefore ’unique’ to this test set-up and can not be directly applied to other shielding applications. II.2 Far Field Shielding Effectiveness Measurement Another method was used to measure the plane wave shielding effectiveness (shielding material located in far field from radiator). It is implemented in a nested anechoic chamber set-up, similar to [4]. The test samples in this case are simple flat slabs but this method uses an expensive anechoic shielded room facility.
Figure 5 Test set-up for far field measurements
The plane wave shielding effectiveness measurements were carried out in an anechoic chamber with a miniature anechoic chamber inside. Samples are fitted in an aperture in the miniature chamber. The methodology is similar to that of standards MIL-STD-285, IEEE-STD-299 and MIL-G-83528B. In this improved set-up both the primary and secondary chamber are fully anechoic. A miniature anechoic chamber is located inside a large anechoic chamber with 3 meter measurement distance, see Fig. 5. The miniature chamber is a cube of brass with the sides 0.6 meter. It is lined on all inside surfaces with ferrite tiles that absorbs radio frequency electromagnetic fields and reduces reflections. A top loaded monopole antenna is used as receive antenna and is mounted over a 20 x 20 cm ground plane. This arrangement is located in the centre of the cube. The transmit antenna located in the larger chamber 3 meters from the aperture wall of the cube is a wideband CHASE bilog antenna often used for EMC testing. The attenuation of electromagnetic plane waves is measured as the insertion loss when closing an aperture with a test sample. The aperture size is 90 x 90 mm corresponding to the sample size of 95 x 95 mm. The shielding effectiveness is obtained as the difference between the antenna coupling with open aperture and
the antenna coupling with a sample fitted in the aperture. The samples were squeezed between the wall of the brass cube and a 20 x 20 cm brass frame with a 90 x 90 mm aperture. Between the sample and the metallic surfaces a fabric over foam conductive gasket was used. A Rohde & Schwarz ZVR vector network analyzer was used in conjunction with an Amplifier Research power amplifier to measure the attenuation in the transmission between the antennas. The frequency range was chosen to 150 MHz to 1 GHz because of the restrictions induced by the small aperture and limitations of the power amplifier used.
Figure 6 Dynamic range for far field SE measurement
test setup
The useful dynamic range for this set-up was investigated by first measuring a baseline attenuation with the aperture open. Then closing the aperture with a 5 mm thick brass plate and using conductive tape to seal thoroughly around the plate. The attenuation was measured again and the insertion loss was calculated as the difference between this reading and the baseline. The dynamic range is more than 50 dB in the chosen frequency range, see Fig. 6. II.3 Transmission line technique for complex permittivity and permeability measurements The third method offers a compact way to estimate the shielding effectiveness by theoretical calculations based on measured complex permittivity and complex permeability. This is the least costly method in terms of sample preparation and test facility. The complex permittivity and complex permeability were measured indirectly using the transmission line technique described in [6]. The transmission line was constructed using two coaxial cables with characteristic impedance, Z0 = 50 Ω connected to a rectangular metallic housing as seen in Fig. 7. The metallic housing impedance, ZL, is 50 Ω when the medium between the center conductor and the
3
housing is air (see Fig. 8) and is thereby matched to the coaxial cables.
(2)
Figure 7 Transmission line fixture for εr and µr
measurements
Figure 8 Unloaded test fixture
Figure 9 Test fixture loaded with Polyethylene
The transmission line fixture was loaded with test samples resulting in a change in characteristic impedance, ZL, proportional to the material properties. The dimensions of the test samples were t x 43 x 0.48 mm where t was 20 mm and 40 mm and represents the length along the propagation direction in the transmission line. The samples are mounted in the fixture in pairs with one on each side of the flat center conductor. See Fig. 9. The change in ZL introduces reflections in the transmission line structure that was be measured using a Rohde & Schwarz ZVR vector network analyzer. From the measured scattering parameters, S11 and S12, the complex properties are calculated (1) – (5). For a complete theoretical derivation, see [3].
+−
=RR
kk
r 11
0
ε (1)
−+
=RR
kk
r 11
0
µ
where
000 εµω=k (3)
( )t
SSeklkj 2
112
1241 0cos −+
=−−
(4)
( )jktlkj eSeSR −− −
=12
211
0 (5)
To verify this technique, measurements on polyethylene were performed. In Fig. 10 the real part of relative permittivity εr for polyethylene is shown. The dotted lines indicates the published constant value εr = 2.25 for Reεr [7]. The measured values differs from the correct by ± 7%.
0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 20
0.5
1
1.5
2
2.5
3
3.5
rRe
GHz Figure 10 Real part of permittivity for polyethylene
The measured data of the permittivity and permeability is used to calculate theoretical plane wave shielding effectiveness for an infinite flat shield. Since the reference case in this comparison is the near field shielding with a box a thorough theoretical model would be too complex and the assumptions is made that this simplified case is a good approach. The total shielding effectiveness is the sum of absorption (A) , reflection (R) and correction (B) also called re-reflection loss. See (6) – (9) found in [2]:
BRASE ++= dB (6) where
( )rrrA εεµ ′−= 1285.0 dB (7)
⋅=
4log20 10
rRε
dB (8)
rl fjeB εγγ ⋅⋅=−⋅= − 021.0,1log20 2 dB (9)
4
The re-reflection loss B corrects for the multiple reflections inside the barrier and is always a negative value since re-reflections degrade SE. This term is very small and have been neglected in the calculations. III. RESULTS Data from the three methods were collected and are here shown for the three sample materials in Fig. 11, 12 and 13. In Fig. 11, 12 and 13 the stars shows near field shielding effectiveness of an electric field source. In the figures, the stars indicates the insertion loss at frequencies where peaks were present in the transmitter free-space signal spectra (Fig. 2). The repeatability for these measurements are very good and the results presented are from one measurement occasion but the results must of course be regarded as unique to this test set-up. This is because the shielding material is in close proximity of the transmitting antenna so that the input impedance of the transmitting antenna may change when changing material. Also the near field impedance, that is the relation of electric field strength to magnetic field strength is unknown.
Figure 11 PC/ABS with 1% Stainless steel
The results for the far field shielding effectiveness measurements are presented as solid lines in Fig. 11, 12 and 13 for the different thermoplastic materials. The results are for 3 mm thick sample plates and are averaged results for 4 measurement occasions performed over a period of 2 weeks. The dotted lines in Fig. 11, 12 and 13 shows the calculated shielding effectiveness values based on measured permittivity and permeability for the different thermoplastic materials. For the materials used in Fig. 11 and Fig. 12 the measured imaginary part of the permittivity is shown in Fig. 13 and Fig. 14, respectively.
Figure 12 ABS with 1% Stainless steel
Figure 13 ABS with 1.5% Stainless steel
100 200 300 400 500 600 700 800 900 1000
Frequency /MHz
0
40
80
120
160
200
Figure 14 Imaginary part of the measured complex
permittivity for PC/ABS with 1% Stainless steel
5
100 200 300 400 500 600 700 800 900 1000
Frequency /MHz
0
20
40
60
Figure 15 Imaginary part of the measured complex
permittivity for ABS with 1% Stainless steel
IV. DISCUSSION Measured shielding effectiveness with the far field method show the same trend in frequency response as the near field method but with an offset in some cases. When studying a larger number of different materials than presented in this paper, materials that performs well in the plane wave case usually also offers good shielding in the near field case. The largest deviation found in this comparison is about 20 dB (Fig. 11). This could indicate the effect of the incident field impedance on SE. However the deviation would then be expected to be in the opposite, that is a larger value of SE were expected for near field SE than for far field SE indicating expected good electric field shielding. Calculated shielding effectiveness based on measured material properties does not agree well with measured near field shielding effectiveness. The largest deviation found in this comparison is almost 20 dB (Fig. 12). In conclusion the measurement of material properties does not give good information on near field shielding effectiveness (SE). Further the approach of far field measurement agrees well in most cases but deviations with no reasonable explanation are found. The base PC/ABS offers 8 dB to 10 dB better far field SE than the pure ABS base with vol. 1% stainless steel fibre as the additive. This is also supported by calculated shielding effectiveness for an infinite flat shield based on the measured electrical properties. For the near field case, no noticeable difference can be seen. The difference in calculated shielding effectiveness (dotted lines) between Fig. 11 and Fig. 12 can be explained by studying the properties of the materials. By analyzing the measured permittivity and permeability further a large difference is found in imaginary
permittivity, see Fig. 14 and Fig. 15, for the materials. The imaginary part of the permittivity includes the effect of conductivity in the material. The cause of the losses in a dielectric material is usually that the conductivity is large [7]. In conclusion the thermoplastic materials with high imaginary part of the permittivity seem to give an improved shielding effectiveness compared to materials with small imaginary part of permittivity. The real part of the permittivity can not alone be correlated to a good SE. V. REFERENCES [1] Mottahed B. D., Manoochehri S., "A review of materials, modeling and simulation, design factors, testing, and measurements related to electromagnetic interference shielding", Polymer-Plastics Technology and Engineering, vol. 34, no. 2, p. 271-346, Mars 1995, ISSN: 0360-2559 [2] Rahman H., Saha P.K., Dowling J. and Curran T., “Shielding effectiveness measurement techniques for various materials used for EMI shielding”, IEE Colloquium on Screening of Connectors, Cables and Enclosures Digest No.012, London, p. 9/1-9/6 of 68, 1992 [3] Wilson P. F., Ma M. T., “Techniques for measuring the shielding effectiveness of materials”, IEEE EMC Society Symposia Records, p. 547-552, 1987 [4] Bodnar D. G., Denny H. W., Jenkins B. M., “Shielding effectiveness measurements on conductive plastics”, IEEE EMC Society Symposia Records, 1979, pp. 27-33. [5] Bush G.G., “Measurement techniques for permeability, permittivity and EMI shielding: a review”, IEEE International Symposium on Electromagnetic Compatibility, 1994. Symposium Record., pp. 333 – 339, 22-26 Aug. 1994, Chicago, USA, ISBN: 0-7803-1398-4 [6] Barry W., “A Broad-Band, Automated, Stripline Technique for the Simultaneous Measurement of Complex Permittivity and Permeability”, IEEE Trans. on MTT, vol. 34, no. 1, Jan. 1986, pp. 80-84. [7] Liao S. Y.,”Engineering Applications of Electromagnetic Theory”, West Publishing Company, St. Paul, USA, ISBN: 0-314-60175-9
6
Paper F
Shielding Effectiveness Data onCommercial Thermoplastic
Materials
Authors:Urban Lundgren, Jonas Ekman and Jerker Delsing
Reformatted version of paper originally published in:Submitted to Electromagnetic Compatibility, IEEE Transactions on
c© 2004, IEEE, Reprinted with permission.
83
84 Paper F
Shielding Effectiveness Dataon Commercial Thermoplastic Materials
U. Lundgren, J. Ekman, J. DelsingEISLAB
Lulea University of TechnologyLulea, SE-971 87
Sweden
Abstract— Ten different commercially available conductivethermoplastic materials have been tested for near- and far-field shielding effectiveness. Far field shielding effectiveness wastested using a modified standard measurement technique toprovide results comparable with company provided data. Further,housings of the different thermoplastic materials was constructedand equipped with a EMI source to model a realistic near fieldshielding effectiveness situation. Shielding effectiveness data upto 1GHz is presented. The conductive thermoplastic materialFaradex XP211 (with filling of stainless steel fibre) and RTPEMI 283 (with filling of nickel coated carbon fibre) were thetwo materials offering the best far field shielding performance.For near field shielding, Faradex XX711 and Bekaert Beki-Shield (both with filling of stainless steel fibre) were the twobest performing. Faradex XX711 showed the best combined farfield and near field shielding results.
I. INTRODUCTION
Shielding effectiveness data of commercially available con-ductive thermoplastic materials are seldom published otherthan in product specification from the manufacturer. Alsodifferent manufacturers uses different measurement methodswhich makes comparisons hard.
For encapsulating electronic devices, boxes of metal sheetshave earlier been used. During the last two decades, plasticmaterials have replaced metal in encapsulation. The last coupleof years, internal clock frequencies in electronic circuits haveincreased to more than 1 GHz. This makes it important tocontrol emitted radio frequency electromagnetic emissions.Also circuits must be protected from electromagnetic energyin its environment. This is made by including filters in thecircuit design and to use encapsulation with electromagneticshielding capabilities.
There are a number of methods available to make en-capsulations of plastic materials offering some degree ofelectromagnetic shielding [7]. Those include surface treatmentof non conductive plastic such as conductive paint, vacuummetallization etc. Another method is to mix conductive parti-cles into the plastic before the encapsulation product is formed.A great advantage with this method is that the number ofproduction steps are reduced.
One problem that arises for the EMC engineer is to selectan encapsulation technique that offers a desired degree ofelectromagnetic shielding for a new electronic device. Themanufacturers of different conductive filler materials some-times specifies the shielding performance of their material in
an application according to standardized measurement methodbut deviations from the exact standard often occurs. Thismakes comparisons between different manufacturers hard.Further the standardized method just give a hint of what theshielding performance can be for the same material in anelectronic device encapsulation application.
To measure plane wave shielding effectiveness for a smallsample of a material several methods are published [4] and[5]. A circular split coaxial transmission line (flanged coax-ial holder) is described in the standard ASTM-D4935. Thismethod is often referred to by conductive composite materialmanufacturers. A problem with this kind of measurementmethod is that it seldom reflects the real situation. When amaterial is used in an enclosure for an electronic device it isusually exposed to a different electromagnetic field conditionthan in the standardized test method.
Thus it was decided to evaluate electromagnetic shieldingeffectiveness for commercially available thermoplastic materi-als. Ten materials were chosen and samples manufactured foranalysis using two measurement methods. This paper describesthe composition of the chosen materials, the measurementtechniques are discussed, the recorded results are presentedand conclusions from the comparisons are drawn. In someinstances manufacturer data were available for comparison.
II. PRESENTATION OF THERMOPLASTIC MATERIALS
This paper presents measurements on ten thermoplasticmaterials. The raw materials was bought from three largemanufacturers and prepared according to manufacturer in-structions. Materials from the following manufacturers wereselected.
Manufacturer LNP Engineering Plastics, Inc offers Faradex[1] compounds that combine thermoplastics, ABS, PC/ABS,PC, and PP, with stainless steel fibres in relatively low concen-trations. The materials are typically used for EMI protection(housing) of electronic components/systems including LANconnection boxes and credit card payment devices. The fol-lowing materials were used:
• XA711 Acrylonitrile-Butadiene-Styrene (ABS) withstainless steel content of 1.5%-vol. corresponding to10%-wt, contains flame retardant.
• XP211 Polypropylene (PP) with stainless steel content of1.5%-vol. corresponding to 12%-wt.
• XX711 Polycarbonate (PC) with stainless steel contentof 1.5%-vol. corresponding to 9%-wt, contains flameretardant.
• XC711 Polycarbonate/Acrylonitrile-Butadiene-Styrene(PC/ABS) with stainless steel content of 1%-vol.corresponding to 7%-wt, contains flame retardant.
• XA611 Acrylonitrile-Butadiene-Styrene (ABS) withstainless steel content of 1%-vol. corresponding to 7%-wt, contains flame retardant.
Manufacturer RTP Company [2] offer different kinds of plas-tic products to match specific applications demanding EMIshielding, high temperature protection, and wear durability.Typically, carbon fibre, stainless steel fiber, or nickel coatedcarbon fibre are used in a thermoplastic matrix to provide thenecessary shielding. The following materials were used:
• EMI 2583 C FR Polycarbonate/Acrylonitrile-Butadiene-Styrene (PC/ABS) with nickel coated carbon fibre contentof 20%-wt, contains flame retardant.
• EMI 162 Polypropylene (PP) with stainless steel fibrecontent of 15%-wt.
• EMI 330 F FR Polycarbonate (PC) with stainless steelfibre content of 15%-wt, contains flame retardant.
• EMI 283 NYLON-6,6 Polyhexamethylene-adipamide(PA) with nickel coated carbon fibre content of 20%-wt.
Manufacturer Bekaert [3] offer a stainless steel fibre (Beki-Shield) that is used as filler for electrically conductive plastics.The chopped fibres are bound together with polymeric bindersspecific for various polymer resins. For the test Beki-Shieldwas mixed in Acrylonitrile-Butadiene-Styrene (ABS) with afibre content of 15%-wt.
III. SHIELDING EFFECTIVENESS MEASUREMENT
TECHNIQUES
For the measurement of electromagnetic shielding effective-ness various techniques are offering different dynamic rangesin different frequency intervals. A number of standardizedmethods exists, however in practical use tailor made testfixtures often deviates from the design in standards whichmakes measured data specific to the modified fixture andcomparisons becomes less significant. Also the demands onpreparation of test samples may influence when choosingmethod of measurement.
We have chosen to use two measurements methods wherethe first method (A) is a far field shielding measurement thatis based on standardized methods. The second measurementmethod (B) is a generic approach to totally encapsulate atransmitter in its near field with the material under test. Theintention with this second method is to resemble an applicationof the material where it is used for the encapsulation of aworking electronic device. A comparison have been madebetween the far field shielding effectiveness measurementmethod and the application imitating near field method [6].
A. Far field shielding effectiveness measurements, modifiedMIL-STD 285 Type Measurement
The foundation of shielding effectiveness measurements hasearlier been the American military standard MIL-STD-285from 1956 (now withdrawn) [4], [7] and [8]. The method usestwo shielded rooms with one common wall. This wall has anaperture were test objects can be placed. The transmitter andreceiver antenna is located in separate rooms, directed towardseach other at a fix distance. The transmitter is transmitting atconstant power and the receiver measure the transferred powerwith and without test object mounted in the aperture. Thedifference between these measurements is the insertion loss(IL) for the test object. Measurements according to MIL-STD-285 have been used to examine new shielding materials likeconducting composites and performance of conducting gas-kets. Drawbacks with the method are that measured insertionloss is dependent on the antenna placement and orientation andthe reflections of the electromagnetic wave inside the shieldedrooms [9].
Improved versions of the method in MIL-STD-285 havedeveloped were the problems with reflections have been mini-mized. A proposed method [10] involves a box, with absorberson the inside walls, containing the transmitting antenna. Thereis an aperture in one wall of the box where test objectscan be mounted. The box is placed in an anechoic shieldedroom together with the receiving antenna. The insertion lossis determined in the same manner as above. This improvedmethod have been used in this paper for the far field shieldingeffectiveness measurements with the modifications that thereceiving antenna is placed inside the box and the transmittingantenna in the room outside the box. Also the box apertureand sample holder has been adapted to the material samplesavailable to us.
New improved versions of the method in MIL-STD-285can also be found in the standards IEEE-STD-299 from 1991and MIL-G-83528B from 1992. Frequency range from a fewMHz to many GHz, dynamic range about 50dB [4]. Usuallymeasurements are done in the range 200 MHz to 18 GHz. Inthis paper to characterize the far field shielding effectivenessfor the ten different conductive thermoplastics we used themethod described below.
Inside an anechoic shielded room a brass box containingthe receiver antenna is placed on a wooden table. The boxis damped inside with ferrite tiles and has a 90 x 90 mmaperture on one side. The aperture size is determined by thethermoelectric material sample size which is 95 x 95 mm.A transmitting antenna were placed three meters from theaperture and the receiver antenna measured the transmittedpower with and without a thermoplastic material in the aper-ture. The shielding effectiveness for that specific thermoplasticmaterial is then calculated as the difference between these twomeasurements.
The frequency range for this test set-up is 150-1000 MHz.The lower frequency limit is set to ensure far field conditionsat the front of the metallic box and by the aperture size.
Fig. 1. Dynamic range for far field SE measurement test set-up.
This aperture acts like a wave guide and attenuates the elec-tromagnetic field under a specific cut-off frequency resultingin measurement problems and a reduced dynamic range forlower frequencies. The upper frequency limit is set to 1000MHz by the upper frequency limit for the transmitter poweramplifier. The transmitting antenna was a Chase 6112B BiLog.The dynamic range for the test set-up can be seen in Figure1 and is over 50 dB for all frequencies. The main drawbackfor this set-up is the small aperture size, only 90x90mm. Thedynamic range could be increased and the lower frequencylimit could be decreased if a bigger aperture could be used,however larger material samples would then be required. Othersuggestions for improving the technique can be summarizedas follows:
• better cables connecting the measurement receiver andthe receiver antenna offering a higher level of shieldingagainst the radiated incident field,
• improved sample holder design to simplify mounting oftest sample with repeated measurements.
B. Near field shielding effectiveness measurements, applica-tion specific measurement
The second method used a more ’realistic’ test set-up tocharacterize the shielding effectiveness for the different ther-moplastic materials. A battery powered transmitter, is enclosedin a 18 x 11 x 12 x cm box, see Figure 2, made out of thedifferent materials. The transmitter were placed on a table in aanechoic room and the emitted free-space signal spectra at a 3m distance were measured, see Figure 3. Then, the transmitterwere enclosed in a box, made of each material and theemitted signal spectra were measured. The boxes were madeof two halves joined together with a slit. Finally, the shieldingeffectiveness is calculated as the difference between these twomeasurements. In this case the thermoplastic material is veryclose to the signal source like in most shielding applications.Meaning that the barrier, the thermoplastic material, is in thenear field of the source and the nature, high or low impedance,
Fig. 2. Transmitter placed in box made of a thermoplastic material.
dBµV
80
-20
-10
0
10
20
30
40
50
60
70
MHz
1000.020.0 100.0
Fig. 3. Radiated electric field from battery powered transmitter.
of the emitted electromagnetic wave is unknown. This meansthat the results from these measurements are ’unique’ to thistest set-up and can not be directly applied to other shieldingapplications. The low level of output power from the batterypowered transmitter give an undesired consequence of smalldynamic range for this method. The measured data from thismethod presented in this paper cover the frequency interval200 MHz to 800 MHz. Above 800 MHz the output powerwas not sufficient.
IV. RESULTS
A. Measurement results
The results from the two different measurements are dis-played in parallel in Figures 4 to 13.
The measurements shows clearly the different materialsshielding effectiveness, from 17 dB to more than 50 dB forsome materials. Four of the ten materials offer a far fieldshielding effectiveness of better than 40 dB in the entirefrequency range.
Among these materials, the near field shielding effectivenessdiffers significantly. All of the curves shows an increased SEfor higher frequencies and the reason can be the test set-up
and the small aperture size. As can be seen in Figure 1 thedynamic range for the set-up is also increasing with frequency.
The measurement accuracy for the far field shielding effec-tiveness measurements is hard to estimate because only twomeasurements were made for each material sample. Deviationsof up to ±6 dB was observed when measurements wasrepeated. This random error may to some part be a explainedby non repeatable effects from the mounting of the testsamples over the aperture. Other problems with this type ofmeasurement method is also known [9]. A bias error of ±2dB is expected in this measurement. The measurements wasnot repeated enough to do a proper statistical analysis of thedistribution of the measurement results because the sampleholder was demanding to work with.
0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 10
10
20
30
40
50
60
Sh
ield
ing
effe
ctiv
en
es
s (
dB
)
Frequency (GHz)
Far field SENear field SE
Fig. 4. Measured shielding effectiveness for Faradex XA711. Shieldingeffectiveness data for plane wave shielding is shown with circles. The resultsfor near field shielding effectiveness of an electric field source is shown withfilled circles.
B. Comparison with manufacturer data
Manufacturer data for some thermoplastics can be found atthe manufacturers webpages.
For Faradex XX711 (3 mm thick sample), the far fieldshielding effectiveness is reported by manufacturer to be 50 to65 dB using the ASTM D4935 test method [1]. The frequencyrange is unspecified but 1 MHz to 1.8 GHz is a typicalrange for this method [4]. The measured far field shieldingeffectiveness data presented in this paper lies in the span 40to 48 dB for a 3 mm thick sample in the frequency range 200MHz to 1 GHz.
For RTP EMI 2583 C FR (0.120 inch thick correspondingto 3.05 mm), the far field shielding effectiveness is reportedby manufacturer to be 42-50 dB using ASTM D4935 testmethod [2]. The frequency range is 100 MHz to 1.5 GHz.The measured far field shielding effectiveness data presentedin this paper lies in the span 35 to 49 dB for a 3 mm samplein the frequency range 200 MHz to 1 GHz.
For RTP EMI 283 (0.110 inch thick corresponding to2.79 mm), the far field shielding effectiveness is reported
0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 10
10
20
30
40
50
60
Sh
ield
ing
effe
ctiv
en
es
s (
dB
)
Frequency (GHz)
Far field SENear field SE
Fig. 5. Measured shielding effectiveness for Faradex XP211. Shieldingeffectiveness data for plane wave shielding is shown with circles. The resultsfor near field shielding effectiveness of an electric field source is shown withfilled circles.
0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 10
10
20
30
40
50
60
70S
hie
ldin
g e
ffec
tive
ne
ss
(d
B)
Frequency (GHz)
Far field SENear field SE
Manufacturer data
Fig. 6. Measured shielding effectiveness for Faradex XX711. Shieldingeffectiveness data for plane wave shielding is shown with circles. The resultsfor near field shielding effectiveness of an electric field source is shown withfilled circles.
by manufacturer to be 26-30 dB using ASTM D4935 testmethod [2]. The frequency range is 100 MHz to 1.5 GHz.The measured far field shielding effectiveness data presentedin this paper lies in the span 40 to 53 dB for a 3 mm samplein the frequency range 200 MHz to 1 GHz.
Beki-Shield is stated by the manufacturer Bekaert to offermore than 60 dB electromagnetic shielding effectiveness inthe frequency range 30 MHz to 1 GHz [3]. This value iscorresponding to a conductive composite with 15% stainlesssteel fibre content (the Beki-Shield material). The test methodand sample thickness are not specified. The measured far fieldshielding effectiveness data presented in this paper lies in thespan 33 to 38 dB for a 3 mm sample in the frequency range
0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 10
10
20
30
40
50
60
Sh
ield
ing
effe
ctiv
en
es
s (
dB
)
Frequency (GHz)
Far field SENear field SE
Fig. 7. Measured shielding effectiveness for Faradex XC711. Shieldingeffectiveness data for plane wave shielding is shown with circles. The resultsfor near field shielding effectiveness of an electric field source is shown withfilled circles.
0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 10
10
20
30
40
50
60
Sh
ield
ing
effe
ctiv
en
es
s (
dB
)
Frequency (GHz)
Far field SENear field SE
Fig. 8. Measured shielding effectiveness for Faradex XA611. Shieldingeffectiveness data for plane wave shielding is shown with circles. The resultsfor near field shielding effectiveness of an electric field source is shown withfilled circles.
200 MHz to 1 GHz.
V. CONCLUSIONS
When comparing the near- and far- field shielding effective-ness for the thermoplastic materials the following is noted.
• Faradex XP211 offer the best far field shielding effective-ness.
• Faradex XX711 and Beki-Shield offer the best near fieldshielding effectiveness knocking the signal from the EMIsource (transmitter) down below the noise floor.
• Faradex XA611 is the material with the lowest level ofshielding effectiveness for both near- and far- field.
• Faradex XX711 is the best material for the combinedshielding effectiveness.
0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 10
10
20
30
40
50
60
Sh
ield
ing
effe
ctiv
en
es
s (
dB
)
Frequency (GHz)
Far field SENear field SE
Manufacturer data
Fig. 9. Measured shielding effectiveness for RTP EMI 2583 C FR. Shieldingeffectiveness data for plane wave shielding is shown with circles. The resultsfor near field shielding effectiveness of an electric field source is shown withfilled circles.
0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 10
10
20
30
40
50
60S
hie
ldin
g e
ffec
tive
ne
ss
(d
B)
Frequency (GHz)
Far field SENear field SE
Fig. 10. Measured shielding effectiveness for RTP EMI 162. Shieldingeffectiveness data for plane wave shielding is shown with circles. The resultsfor near field shielding effectiveness of an electric field source is shown withfilled circles.
Considering that the near field shielding measurement imitatesthe use of material in an application it is disappointing to sehow poor guide the far field shielding effectiveness results arewhen a material selection for an enclosure must be made. Incases where manufacturer data were available, agreement wasquite good with the near field method in one instance whilethe difference was close to 30 dB in the other three instances.
Polymers with a 20% fill of nickel coated carbon fibreperforms similar to polymers with 15% fill of stainless steelin far field measurement. Choice of base polymer seem toinfluence the performance of the conductive plastic material.Particularly the ABS base polymer gave lower far field shield-ing effectiveness values that other base polymers with the sameamount of conductive filler.
0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 10
10
20
30
40
50
60
Sh
ield
ing
effe
ctiv
en
es
s (
dB
)
Frequency (GHz)
Far field SENear field SE
Fig. 11. Measured shielding effectiveness for RTP EMI 330 F FR. Shieldingeffectiveness data for plane wave shielding is shown with circles. The resultsfor near field shielding effectiveness of an electric field source is shown withfilled circles.
0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 10
10
20
30
40
50
60
Sh
ield
ing
effe
ctiv
en
es
s (
dB
)
Frequency (GHz)
Far field SENear field SE
Manufacturer data
Fig. 12. Measured shielding effectiveness for RTP EMI 283. Shieldingeffectiveness data for plane wave shielding is shown with circles. The resultsfor near field shielding effectiveness of an electric field source is shown withfilled circles.
In this study only the Beki-Shield material was mixedwhen the plastic samples was produced, all other were pre-mixed compounded grains. This fact may have caused thatpoor fibre distribution in mould was achieved and give someexplanation why poor far field results were observed for theBeki-Shield material. The RTP EMI 162 material showedlower shielding effectiveness performance that expected whencompared to other materials with same base polymer andamount of conductive filler. The reason for this is unclear andhave not been studied.
Materials containing stainless steel fibres shows better nearfield shielding effectiveness than materials with nickel coatedcarbon fibre even though stainless steel fibre filler percentage islower. The material RTP EMI 162 show unexpected poor near
0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 10
10
20
30
40
50
60
Sh
ield
ing
effe
ctiv
en
es
s (
dB
)
Frequency (GHz)
Far field SENear field SE
Manufacturer data
Fig. 13. Measured shielding effectiveness for Beki-Shield. Shielding effec-tiveness data for plane wave shielding is shown with circles. The results fornear field shielding effectiveness of an electric field source is shown withfilled circles.
field shielding effectiveness in respect to the stainless steelfiller percentage. The reason for this have not been studied.
Compared to the far field shielding effectiveness the ma-terial containing Beki-Shield stainless steel fibre performedrelatively well in near field shielding effectiveness. The reasonfor this have not been studied but may be caused by ununiformfibre distribution in mould.
REFERENCES
[1] LNP. Faradex and EMI-X Electrically Conduc-tive Compounds. Nov. 2003. [Online]. Available:http://www.lnp.com/LNP/Products/PShieldingEmi.html
[2] RTP Company. Nov. 2003. [Online]. Availablehttp://www.rtpcompany.com/products/index.htm
[3] Bekaert. Nov. 2003. [Online]. Available http://www.bekaert.com[4] Rahman H., Saha P.K., Dowling J., Curran T., ”Shielding effectiveness
measurement techniques for various materials used for EMI shielding”,IEE Colloquium on Screening of Connectors, Cables and EnclosuresDigest No.012, London, Page(s): 9/1-9/6 of 68, 1992
[5] Kashyap S., ”Shielding effectiveness measurements with a dual TEMcell and a split TEM cell”, IEEE International Symposium on Electro-magnetic Compatibility - Symposium Record., Page(s): 262-264 ISSN:0190-1494, 1986.
[6] Lundgren U., Ekman J., Delsing J., ”Characterization of ConductiveThermoplastic Composite Materials using Multiple Measurement Meth-ods”, EMC Europe 2002, Sorrento 2002
[7] Mottahed B. D., Manoochehri S., ”A review of materials, modelingand simulation, design factors, testing, and measurements related toelectromagnetic interference shielding”, Polymer-Plastics Technologyand Engineering, Volume: 34, Issue: 2, Page(s): 271-346, Mars 1995,ISSN: 0360-2559
[8] Ondrejka A. R., Adams J. W., ”Shielding effectiveness (SE) measure-ment techniques”, IEEE EMC Society Symposia Records, Page(s): 249-256, 1984
[9] Wilson P. F., Ma M. T., ”Factors influencing material shielding effective-ness measurements”, IEEE EMC Society Symposia Records, Page(s):29-33, 1985
[10] Bodnar D. G., Denny H. W., Jenkins B. M., ”Shielding effectivenessmeasurements on conductive plastics”, IEEE EMC Society SymposiaRecords, Page(s): 27-33, 1979
Paper G
Electromagnetic properties ofthermoplastic material for varying
temperatures
Authors:Urban Lundgren and Jerker Delsing
Reformatted version of paper originally published in:Submitted to Microwave Theory and Techniques, IEEE Transactions on
c© 2004, IEEE, Reprinted with permission.
91
92 Paper G
Electromagnetic properties of thermoplastic materialfor varying temperatures
Urban Lundgren, Jerker DelsingEISLAB
Lulea University of TechnologyLulea, SE-971 87
Sweden
Abstract— A thermoplastic material is examined. The complexpermittivity and complex permeability are obtained while thetemperature is varied from 20 to 60 Celsius. It was necessaryto perform instrument calibration at each temperature to canceltemperature effects on cables and connectors. The thermoplasticmaterial is used in a laminate with metallic foil for encapsulationof electronic circuits. The laminated technique offers a goodbarrier against moisture and good shielding for electromagneticenergy. To be able to design stripline transmission lines in thelaminate it is necessary to know the electromagnetic behavior ofthe isolating plastic material. Real part of the relative permittivitywas found to be 2.0±0.1 in the frequency range 100 MHz to 2.5GHz. This value shows a very small dependence of temperaturechanges in the range 20 Celsius to 60 Celsius.
I. INTRODUCTION
Most electronic devices are encapsulated to protect theelectronic circuits from the environment. In many applicationsthe cost, weight, mechanical protection, dust and moisturebarrier and EMC properties of the encapsulation are criticalparameters. A material choice that have very interesting prop-erties in all these aspects are laminated materials. Due to thatthe laminated material is composed of foils of metallic andthermoplastic material, the encapsulation technique based onsuch laminates offers a good barrier against dust and moistureand good electromagnetic shielding. Further weight and costturns out to be advantageous.
Since many new electronics devices are mobile it is ofinterest to integrate electrical circuits for example antennasin the laminate. To be able to design a antenna on thelaminate it is necessary to know certain material parametersof the involved parts. For instance the complex permittivityof the isolating material effects the characteristic impedancewhich must be considered in the design of the geometry. Atfrequencies above 1 GHz the losses in the isolator starts tobecome an issue and this is also an effect of the complexpermittivity.
There exists a large number of publications on electromag-netic properties of plastic materials. Older publication (before1985) often focused on plastic material data at frequenciesbelow 100 MHz [1]. In later years plastic material data atmicrowave frequencies have been of major concern becausethe increased need for accurate data when designing devicesworking at higher frequencies [2]. It is then also important
to study the temperature variations in the electromagneticproperties of the material under test [1], [2].
It is of interest to explore the usage of a particular laminatedencapsulation material with integrated antennas in environ-ments with changing temperature. The major problem is toestablish a test methodology where temperature effects otherthan these of the material under test are sufficiently suppressed.
Thus series of experiments have been conducted to obtaincomplex permittivity and complex permeability for the ther-moplastic over temperature 20 − 60C and frequencies from100 MHz to 2.5 GHz.
The desired material characteristics consists of a real andimaginary part according to the following notation
ε = ε0 · εr = ε0 · (ε′r − jε
′′r )
andµ = µ0 · µr = µ0 · (µ′
r − jµ′′r )
where ε is the complex permittivity and µ is the complexpermeability.
It is expected that the real relative permittivity ε′r of the
studied thermoplastic material will be between 2.0 and 2.5because of preliminary tests on similar materials. The tem-perature stability is unknown but very important to judge theusefulness of the material.
II. MEASUREMENT METHOD
Several methods exists for the measurement of permeabilityand permittivity. Some of the most common are the loadedresonant waveguide cavity [3], [4], the open ended coaxial line[3] and the loaded coaxial transmission line [3], [5]. Cavityresonance measurement methods are commonly used becauseof the good accuracy offered, particularly for measurement ofimaginary part of the permittivity for determining the losses ina material [3]. However the resonance of a cavity makes themethod only covering a narrow frequency band. This can beimproved to some extend by including higher order modes.When studying temperature stability it is desired to use amethod that measures the permittivity and the permeability,not the effect on the transmission line that is caused by thermalexpansion and contraction of the plastic material. Due to itsuseful frequency range the loaded coaxial transmission linetechnique is used in this paper for the measurement of the
11 2 3 4 5
Fig. 1. Test fixture containing sample with following parts: 1. top lid, 2.SMA connector, 3. stripline center conductor, 4. material sample (2 pieces,one on each side of center conductor), 5. ferrite for reducing undesired cavitymodes in stripline chamber (one in each corner)
complex parameters. In the transmission line fixture the mate-rial under test is placed to fill the volume between the inner andouter conductor in a section along the line. The material mayload the line and cause a change of characteristic impedance.Both reflection and transmission through the fixture is usedwhen calculating the test material data [5], [6].
Reflection coefficients and transmission coefficients areoften given as scattering parameters (S-parameters) for atwo port circuit. The two involved reflection coefficients aredenoted S11 and S22. The two transmission coefficients aredenoted S12 and S21.
The technique [5], [6] used for the measurements in thissection utilizes the measured complex S-parameters for aloaded 50Ω transmission line seen in Figure 1.
A. Model for parameter extraction
The calculation of complex permittivity and complex per-meability The expressions for the calculation of the complexpermittivity and complex permeability are given in this part.For a complete theoretical derivation, see [5]. In the calcula-tions, only two S-parameters are used since the test fixture inFigure 1 are assumed to be symmetrical and reciprocal. Thismeans that
S21 = S12
and
S22 = S11
and simplifies the calculations considerably. The complexquantities, εr and µr, can be expressed using the reflectioncoefficient,
R =ZL − Z0
ZL + Z0(1)
between the unloaded and loaded region in the housing ac-cording to
εr =k
k0(1 − R
1 + R) (2)
and
µr =k
k0(1 + R
1 − R) (3)
where k is the propagation constant in the loaded region andk0 is the propagation constant in the unloaded region and givenby
k0 = ω√
µ0ε0
For a sample length of t meter and an unloaded region in thehousing of l meter, it is possible to express k and R by usingS11 and S12 according to
k =arccos(e−j4k0l + S2
12 − S211)
t(4)
andR =
S11
(e−j2k0l − S12e−jkt)(5)
which are used in equation 2 and 3 to solve for the unknownquantities. Care must be taken when calculating the propa-gation constant k using equation 4 due to the inverse cosineexpression. By using an alternative expression for the inversecosine, the propagation constant can be split in its real andimaginary part. If equation 4 is rewritten and the argumentfor the inverse cosine term is called Arg according to
kt = arccos(e−j4k0l + S212 − S2
11) = arccos(Arg)
thenkt = ktreal + j · ktimg
and
ktreal = arctan(Im(Arg +
√Arg2 − 1)
Re(Arg +√
Arg2 − 1)) ± 2nπ
ktimg = ln(√
[Re(u)]2 + [Im(u)]2)
whereu = Arg +
√Arg2 − 1
The principal branch, n = 0, of ktreal can be used when thesample length t is
0 ≤ t ≤ λm
2
where λm is the wavelength inside the material.
III. MEASUREMENTS
The coaxial fixture described above was placed in antemperature test chamber (Heraeus HT4010), see Figure 2.This chamber acts as a combined oven and freezer capableof keeping a programmed temperature stable. Temporal tem-perature fluctuations stay within ±1C according to chamberspecification. The scattering parameters in the fixture wasmeasured with a vector network analyser (Rohde & SchwarzZVR) , see Figure 3. After calibration the vector networkanalyser and attached cables introduces a random error ofless than ±1 dB. A possible remaining bias error may comefrom an unknown behavior of the coaxial fixture which werenot included in the calibration. How these errors on the S-parameters influences on the resulting εr and µr has not beenderived.
Fig. 2. Test fixture inside climate chamber
Fig. 3. Test setup including vector network analyser and climate chamber
The upper frequency limit of the measurement is determinedby the wavelength in the transmission line and the samplelength in the wave propagation direction. The upper frequencylimit is therefore dependent of the sample material. Thefrequency axis in Figures 4 to 7 continues up to 3 GHz butreliable data is limited to 2.5 GHz for the material under testin this paper and 2.4 GHz for polyethylene.
After some testing it was decided that 3 hours waiting aftersetting chamber temperature gave the coaxial fixture enoughtime to equalize the temperature. However the temperature ofthe material sample has not been monitored, it was assumedthat repeated identical measurement readings indicated thattemperature equalization was reached. A measurement readingat preprogrammed temperature could then be done.
For the first measurement attempt the vector network anal-yser was calibrated in room temperature and temperature wasset and settling for 3 hours, then measurement was taken. Thisproved to be an insufficient method because the measurementsetup was very sensitive to temperature changes and temper-ature response from measurement cables and connectors hadto be cancelled. The measurement scheme was redesigned to
0.5 1 1.5 2 2.5 3
x 109
0
0.5
1
1.5
2
2.5
3
Frequency (Hz)
Re
al p
art
of
rela
tive
pe
rmitt
ivity
at
roo
m t
em
pe
ratu
re
Material under tes t (MUT)Polyethylene (PE)Empty fixture
Fig. 4. Measured real part of relative permittivity at room temperature (20)Celsius for material under test (MUT), polyethylene reference (PE) and emptyfixture (air) reference.
0.5 1 1.5 2 2.5 3
x 109
-1
-0.8
-0.6
-0.4
-0.2
0
0.2
0.4
0.6
0.8
1
Frequency (Hz)
Ima
gin
ary
pa
rt o
f re
lativ
e p
erm
ittiv
ity a
t ro
om
te
mp
era
ture
Material under tes t (MUT)Polyethylene (PE)Empty fixture
Fig. 5. Measured imaginary part of relative permittivity at room temperature(20) Celsius for material under test (MUT), polyethylene reference (PE) andempty fixture (air) reference.
include three temperatures 20, 40 and 60C. Calibration isdone at each temperature and temperature and cable layoutwere kept fixed until measurements had been carried out. Thecalibration kit of the network analyser is specified to performaccurately for temperatures up to 50C. In our measurementsat 60C calibration was successful and good results wereacquired. For measurements at 80C proper calibration wasnot possible.
After setting the temperature in the chamber, the datarecording was done after waiting 3 hours for temperatureequalization in the coaxial fixture.
IV. RESULTS
The measured real part of relative permittivity for polyethy-lene is 2.3 ± 0.1 at room temperature (20C), see Figure 4.For the material under test the value is 2.0 ± 0.1 at room
0.5 1 1.5 2 2.5 3
x 109
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
1.8
2
Frequency (Hz)
Re
al p
art
of
rela
tive
pe
rme
ab
ility
at
roo
m t
em
pe
ratu
re
Material under tes t (MUT)Polyethylene (PE)Empty fixture
Fig. 6. Measured real part of relative permeability at room temperature (20)Celsius for material under test (MUT), polyethylene reference (PE) and emptyfixture (air) reference.
0.5 1 1.5 2 2.5 3
x 109
-1
-0.8
-0.6
-0.4
-0.2
0
0.2
0.4
0.6
0.8
1
Frequency (Hz)
Ima
gin
ary
pa
rt o
f re
lativ
e p
erm
ea
bilit
y a
t ro
om
te
mp
era
ture
Material under tes t (MUT)Polyethylene (PE)Empty fixture
Fig. 7. Measured imaginary part of relative permeability at room temperature(20) Celsius for material under test (MUT), polyethylene reference (PE) andempty fixture (air) reference.
temperature.A curve fitting method have been used in a least square
error manner. A polynomial of degree 1 (a straight line) wasfitted to the measured data and is plotted together with themeasured data. The measured temperature dependence forreference material (PE), material under test and air is shownfor three frequencies in Figures 8, 9 and 10.
The corresponding straight line equations are at 900 MHz:
ε′r−Polyethylene = 2.3321 − (9E − 4) ∗ t
ε′r−MUTsample = 2.0985 + (1E − 4) ∗ t
ε′r−Airreference = 1.0283 − (2E − 4) ∗ t
The corresponding straight line equations are at 1800 MHz:
ε′r−Polyethylene = 2.2289
20 25 30 35 40 45 50 55 600
0.5
1
1.5
2
2.5
3
Re
al p
art
of
rela
tive
pe
rmitt
ivity
at
90
0 M
Hz
Temperature in Celcius
PE meas ured dataPE fitted polynomialMUT meas ured dataMUT fitted polynomialAir reference meas ured dataAir reference fitted polynomial
Fig. 8. Measured temperature characteristics for polyethylene (PE), thermo-plastic material (MUT) and empty fixture (air) in same graph and correspond-ing fitted polynomials, at 900 MHz
20 25 30 35 40 45 50 55 600
0.5
1
1.5
2
2.5
3
Re
al p
art
of
rela
tive
pe
rmitt
ivity
at
18
00
MH
z
Temperature in Celcius
PE meas ured dataPE fitted polynomialMUT meas ured dataMUT fitted polynomialAir reference meas ured dataAir reference fitted polynomial
Fig. 9. Measured temperature characteristics for polyethylene (PE), thermo-plastic material (MUT) and empty fixture (air) in same graph and correspond-ing fitted polynomials, at 1.8 GHz
ε′r−MUTsample = 2.0183 + (5E − 4) ∗ t
ε′r−Airreference = 0.9909 + (2E − 4) ∗ t
The corresponding straight line equations are at 2400 MHz:
ε′r−Polyethylene = 2.2690 − (6E − 4) ∗ t
ε′r−MUTsample = 2.0690 − (5E − 4) ∗ t
ε′r−Airreference = 0.9661
where ε′r denotes the real part of the relative permittivity
for corresponding materials and t denotes the temperature indegrees Celsius.
The temperature dependence on the real relative permittivityis very small in temperature range and frequency range, whichcan be seen in small coefficients of the line equations. For thematerial under test a temperature change from 20C to 60C
20 25 30 35 40 45 50 55 600
0.5
1
1.5
2
2.5
3R
ea
l pa
rt o
f re
lativ
e p
erm
ittiv
ity a
t 2
40
0 M
Hz
Temperature in Celcius
PE meas ured dataPE fitted polynomialMUT meas ured dataMUT fitted polynomialAir reference meas ured dataAir reference fitted polynomial
Fig. 10. Measured temperature characteristics for polyethylene (PE),thermoplastic material (MUT) and empty fixture (air) in same graph andcorresponding fitted polynomials, at 2.4 GHz
is causing the real relative permittivity to decrease 1%. Thischange is smaller than what the accuracy of this measurementmethod can resolve and must be considered non significant.
The imaginary part of relative permittivity for the materialunder test is slightly lower (closer to zero) than correspondingmeasured value for polyethylene, see Figure 5. The sensitivityfor this measurement method is not enough to quantify themagnitude difference.
Relative permeability values are expected to be close to 1for the real part and close to 0 for the imaginary part in allmeasurements. This was also the case in the measured datashown in Figure 6 and 7.
The measured relative permeability data is stable undertemperature variations.
V. CONCLUSIONS
The material under test has a real relative permittivity of2.0 at room temperature (20C) and it is independent oftemperature in the range 20C to 60C.
Imaginary relative permittivity was slightly smaller for thematerial under test than for polyethylene.
The results for the permeability was as expected for thematerial under test. The real part of the relative permeabilityis close to 1 and imaginary part close to 0.
The measurement setup was very sensitive to temperaturevariations. The first attempt to cover the temperature range0 Celsius to 80C with a single calibration was insufficient.Measurements were then done at three temperature points, 20,40 and 60C, with careful calibration at each temperature tocancel the temperature effects on the cables and connectors.
VI. ACKNOWLEDGEMENT
Karl-Erik Leeb at ProofCap AB contributed to this studyby producing the material samples.
REFERENCES
[1] Bur A. J., ”Dielectric properties of polymers at microwave frequencies:A review”, Polymer, Volume: 26, Page(s): 963-977, July 1985
[2] Riddle B., Baker-Jarvis J., Krupka J.,”Complex Permittivity Measure-ments of Common Plastics Over Variable Temperatures”, Microwavetheory and techniques, IEEE Transactions on, Volume: 51 Issue: 3 ,March. 2003 Page(s): 727-733
[3] Baker-Jarvis J., Janezic M.D., Riddle B., Holloway C.L., Paulter N.G.,Blendell J.E., ”Dielectric and Conductor-Loss Characterization and Mea-surements on Electronic Packaging Materials”, NIST Technical Note1520, July 2001, CODEN: NTNOEF
[4] Baker-Jarvis J., Geyer R.G., Grosvenor J.H., Jr.; Janezic M.D., JonesC.A., Riddle B., Weil C.M., Krupka J.,”Dielectric characterization oflow-loss materials a comparison of techniques”, Dielectrics and Electri-cal Insulation, IEEE Transactions on, Volume: 5 Issue: 4 , Aug. 1998Page(s): 571-577
[5] Barry W., ”A broad-band, automated, stripline technique for the si-multaneous measurement of complex permittivity and permeability”,Microwave Theory and Techniques, IEEE Transactions on, Volume: 34,Issue: 1, January 1986
[6] Lundgren U., Ekman J., Delsing J., ”Characterization of ConductiveThermoplastic Composite Materials using Multiple Measurement Meth-ods”, EMC Europe 2002, Sorrento 2002
