Circuit Modeling For EMC
Circuit Modeling For EMC
Very simply, electromagnetic interference (EMI) costs money, reduces profits, and
generally wreaks havoc for circuit designers in all industries. This book shows
how the analytic tools of circuit theory can be used to simulate the coupling
of interference into, and out of, any signal link in the system being reviewed.
The technique is simple, systematic and accurate. It enables the design of any
equipment to be tailored to meet EMC requirements.
Every electronic system consists of a number of functional modules interconnected
by signal links and power supply lines. Electromagnetic interference can be
coupled into and out of every conductor. A review of the construction of the wiring
assemblies and the functions of the signals they carry will allow critical links to be
identified. Circuit modeling can be used to simulate the electromagnetic coupling
mechanism of each critical link, allowing its performance to be analyzed and
compared with the formal requirements. Bench testing during the development
of any product will allow any interference problem to be identified and corrected,
long before the manufactured unit is subjected to formal testing.
KEY FEATURES
A fully outlined, systematic and dramatically simplified process of designing
equipment to meet EMC requirements.
Focuses on simplifications which enable electrical engineers to singularly
handle EMC problems.
Helps minimize time-to-market of new products and reduces the need for
costly and time-consuming modifications.
Outlines how general purpose test equipment (oscilloscopes and signal
generators) can be used to validate and refine any model.
Discusses how to use Mathcad or MATLAB to perform analysis and
assessment.
ABOUT THE AUTHOR
Ian B. Darney was awarded a BSc degree in Electrical Engineering at the University
of Glasgow in 1960. He joined the Guided Weapons Division of British Aerospace
and worked on the circuit design of equipment for missiles, ground equipment,
submersibles, and spacecraft. After transferring to the Airbus Division he carried
out certification work associated with lightning indirect effects, electrostatics and
intrinsic safety. He was a member of the European Organisation for Civil Aviation
Equipment (EUROCAE) committee which defined the requirements for the
protection of aircraft from the indirect effects of lightning. Since his retirement,
he has continued to work as an EMC consultant, and has written two technical
papers and numerous magazine articles on EMC.
Circuit Modeling for
Electromagnetic Compatibility
Other titles in the series
Designing Electronic Systems for EMC (2011)
by William G. Duff
Ian B. Darney
Edison, NJ
scitechpub.com
Published by SciTech Publishing, an imprint of the IET.
www.scitechpub.com
www.theiet.org
No part of this publication may be reproduced, stored in a retrieval system or transmitted in any form or by any
means, electronic, mechanical, photocopying, recording, scanning or otherwise, except as permitted under Sections
107 or 108 of the 1976 United Stated Copyright Act, without either the prior written permission of the Publisher, or
the authorization through payment of the appropriate per-copy fee to the Copyright Clearance Center, 222 Rosewood
Drive, Danvers, MA 01923, (978) 750-8400, fax (978) 646-8600, or on the web at copyright.com. Requests to the
Publisher for permission should be addressed to The Institution of Engineering and Technology, Michael Faraday
House, Six Hills Way, Stevenage, Herts, SG1 2AY, United Kingdom.
While the author and publisher believe that the information and guidance given in this work are correct, all parties
must rely upon their own skill and judgement when making use of them. Neither the author nor the publisher
assumes any liability to anyone for any loss or damage caused by any error or omission in the work, whether such an
error or omission is the result of negligence or any other cause. Any and all such liability is disclaimed.
10 9 8 7 6 5 4 3 2 1
Preface xiii
Acknowledgments xvii
1 Introduction 1
1.1 Background 1
1.1.1 The need for EMC 1
1.1.2 Pragmatic approach 1
1.1.3 Academic approach 2
1.1.4 Managerial approach 2
1.1.5 Misleading concepts 2
1.1.6 Circuit modeling 3
1.1.7 Computations 3
1.1.8 Testing 3
1.1.9 Essence of the approach 4
1.2 Developing the model 4
1.2.1 Basic model 4
1.2.2 Parameter types 5
1.2.3 Derivation process 6
1.2.4 Composite conductors 7
1.2.5 Proximity effect 8
1.2.6 Electrical length 8
1.2.7 Distributed parameters 9
1.3 Intra-system interference 11
1.3.1 The signal link 11
1.3.2 Simulating the structure 11
1.3.3 Equivalent circuits 12
1.3.4 Conducted emission 12
1.3.5 Conducted susceptibility 13
vii
viii Contents
Back in the 1960s, the author was a member of a team designing a Flight Trainer. In this
equipment, an analog computer generated a set of waveforms which resulted in a trapezoidal
raster being displayed on the screen of a flying-spot scanner. The light from the screen
illuminated a continuously moving film of a five-mile wide strip of terrain. The light which
penetrated the film was focused by a collimator lens onto a photomultiplier. The video
output simulated that of a camera mounted on a low-flying missile.
The system worked reasonably well, but was plagued by a wide range of interference
problems which were never satisfactorily solved. The underlying reason was the fact that, at
the outset of the project, the customer insisted that a single-point ground terminal be located
at the bottom of the rack of equipment and that three wire-braids be connected to that point.
These were designated the analog ground, logic ground, and power ground, and the star-
point was specified as the only place where ohmic contact was allowed. At all other locations
in the equipment, the reference grounds were isolated from each other. This was, and is, the
worst possible configuration to adopt.
Even so, the concept of the star-point ground has gained widespread acceptance by the
engineering community. Other misleading concepts in vogue are the equipotential ground
and the need to avoid earth loops.
This book started out as a study report that advocated the use of a set of guidelines which
could replace these misleading concepts. Since engineers are skeptical individuals, there was
always someone who could point out a defect in the reasoning. So more background material
was gathered, tests were carried out, and further analyses performed. It eventually became
clear that circuit modeling could be used to analyze the coupling mechanisms.
But there were still critics who pointed out that such an approach could not be used to
handle high-frequency simulations. So the modeling technique was developed further to
cater for transmission-line effects and to take into account the action of cables as antennae.
The end result is a technique that can be used to assess and analyze the mechanisms usually
associated with electromagnetic interference (EMI). That is
common impedance,
electric fields (capacitive induction),
magnetic fields (magnetic induction), and
electromagnetic fields (plain waves).
xiii
xiv Preface
The following pages provide many circuit models which can simulate the various
conducted EMI and radiated EMI problems. The approach is unique in that it uses simple
analytical methods. It is easy to implement.
The contents are useful to practical design engineers at various levels, such as circuit
designers, printed circuit board designers, electronic system engineers, power system engi-
neers, EMC engineers, and EMC consultants. Time is precious to such individuals, so it is
recommended that the busy designer first reads Chapter 9, which describes a top-down
approach and provides a set of simple guidelines. If this systematic approach is implemented,
then the design can be made fundamentally sound. Then it is worth reading Chapter 8, which
identifies most of the techniques which reduce the level of EMI coupling and describe the
mechanisms involved. The preceding chapters can be regarded as material which justifies the
detailed recommendations.
Lecturers who teach subjects such as electronic circuit design (analogue, digital, switched-
mode, radio-frequency, etc.) should find it useful, since it relates fundamental concepts to
the considerations of practical design.
Students of electrical engineering will benefit from this book, since EMC is no longer an
optional topic and the approach described in the following pages is the simplest possible.
It should also be useful to universities who provide special courses on the subject of
EMC, since it identifies a different approach to the analysis of EMI. Since it does not require
an ability to manipulate the mathematics of electromagnetic field theory, it is understandable
to a wider range of engineers.
One of the tests in Chapter 7 identifies the fact that antenna-mode current propagates
faster than differential-mode current, and shows how the two velocities can be measured.
This should be of interest to researchers.
There are many books which describe the various interference coupling mechanisms, and
which identify practical design solutions. Others delve into the analysis of electromagnetic
field propagation. Since these aspects are well-covered elsewhere, there is no need to reprise
their contents. Such a policy keeps this book relatively short.
The first chapter identifies the underlying concepts and summarizes the approach.
Chapter 2 defines the building blocks of all circuit models and derives simple models
of familiar configurations such as the coupling between common-mode circuits and
differential-mode circuits. These models are useful in providing an insight into the coupling
mechanisms. They are amenable to analysis using SPICE software. The simulated response
is reasonably accurate up to the frequency at which the wavelength of the signal is one-tenth
the length of the cable.
Chapter 3 develops the process to allow the electromagnetic coupling in complex
assemblies such as aircraft wings or multilayer boards to be simulated. Although the fre-
quency response of such models is subject to the same limitation as that of the simpler
configurations, the range of possible applications is vastly extended.
An open-circuit line will resonate at a frequency where the quarter wavelength of the
signal is equal to the length of the line. A short-circuited line will resonate at the half-wave
frequency. At resonance the level of interference will reach a peak value. If it is hoped to
simulate the interference-coupling characteristics of any signal link, then the model should
be capable of handling signals up to, and beyond, the half-wave frequency of the line.
Chapter 4 achieves this objective by invoking the relationships of transmission-line theory.
Preface xv
Chapter 5 takes the process one step further, to simulate the behavior of cables as
antennae.
Chapter 6 derives a circuit model that can replicate the transient behavior of a twin-
conductor cable as an antenna.
Chapter 7 shows how circuit models can be used to simulate the response of bench tests
on actual hardware. This establishes the all-important connection between theory and
practice.
Chapter 8 describes a number of techniques that have been used by engineers to improve
EMC, and relates these designs to the interference coupling mechanisms identified in the
previous chapters.
Chapter 9 outlines a systematic method of analyzing the EMC characteristics of the
system-under-development. It establishes a clear link between the formal EMC design
requirements and the performance of the equipment. The S.I. system of units is used
throughout the book.
Although the analytical process is dramatically simpler than one based on the use of
electromagnetic field theory, the calculations still require the use of a computer. Simulation
Programs with Integrated Circuit Emphasis (SPICE) can deal with the simple configurations
described in Chapter 2, but cannot handle the computations described in the later chapters.
Mathematical software is needed.
It was found that Mathcad software was ideal for the purpose since it can combine the
equations of circuit analysis with those of electromagnetic field propagation. It can accept
input data in the form of geometrical measurements of the hardware-under-review and
combine this with data derived from tests on that hardware. Appendix A provides a brief
introduction to this software. Copies of the worksheets can be downloaded from the website
www.designemc.info.
Subsequent to the completion of the first draft of this book, an exercise was carried out to
translate the Mathcad worksheets into MATLAB m-files. These also can be downloaded
from the website. Appendix B identifies the relationships between the two software packages
which help engineers who are familiar with MATLAB to read and understand the contents of
the Mathcad worksheets.
One of the key features of the analysis is the use of a transformation formula derived
from the equations of transmission-line theory. Appendix C provides a succinct introduction
to the concept of distributed parameters and derives the hybrid equations used as a starting
point in Chapter 4.
Although many of the concepts used in this book are familiar to electrical and electronic
engineers, some are new. So a set of definitions is provided in Appendix D.
Reports on further tests and analyses will be filed at www.designemc.info as and when
they are completed. The website also has a page for feedback from readers.
Acknowledgments
Thanks are due to the many old colleagues in British Aerospace who provided encourage-
ment and criticism in equal measure, to Alistair Duffy who promoted the book, to Dudley
Kay who agreed to publish it, and to my wife Frances for her patience and support.
xvii
CHAPTER 1
Introduction
1.1 Background
1.1.1 The need for EMC
The development of electronic equipment has come a long way since the invention of valves
and transistors, to the extent that modern society is highly dependent on the smooth
functioning of the myriad systems that myriad systems now in operation.
Concurrently with that development, Electromagnetic Interference (EMI) has also
increased, both in the number of daily incidents and in the severity of the possible con-
sequences. Initially, most of the effects were annoying; for example, crackles on the radio
due to a nearby thunderstorm were something one learned to accept. Latterly, some of the
effects could be life threatening. The phenomenon described as sudden unintended accel-
eration could be a case in point.
A succinct definition of Electromagnetic Compatibility (EMC) is the ability of a device,
unit of equipment, or system to function satisfactorily in its electromagnetic environment
without introducing intolerable electromagnetic disturbances to anything in that environ-
ment [1.1].
1
2 CHAPTER 1 Introduction
design techniques are identified in books that adopt this approach [1.4]. This being so, there
is no need to reprise the material they provide.
While such an approach leads to many useful design techniques, it is essentially hit or
miss. A technique that works well in one application can cause disastrous effects in another.
Since it involves subjective judgment, there is still plenty of scope for disagreement between
designers.
This is in marked contrast to requirements such as functional performance, frequency
response, power consumption, reliability, mass, and size, which are all amenable to rigorous
analysis. Care is taken at every stage of the design process to ensure that these other require-
ments will be met. Bench testing is carried out on prototype equipment. In situ tests are carried
out on assembled systems. Test results are compared with predicted performance. Regular
design reviews are carried out. If there are any problems, modifications are implemented.
reference point for all signals in the system. Guidance for the design of UK military
equipment formalizes this concept and provides detailed requirements for its implementation
[1.6]. If this concept is implemented into the design of any electronic equipment, it can be
guaranteed that the equipment will suffer from intractable interference problems.
Closely related to the concept of the single-point ground is the stricture to avoid earth
loops.
Although none of these concepts is to be found in any textbook on electromagnetic
theory, they have become firmly entrenched beliefs in the engineering community. There has
to be a better way of approaching the task of achieving EMC.
1.1.7 Computations
Although the simplifications achieved through the use of circuit modeling provide a dramatic
reduction in the complexity of the mathematics, it is still necessary to employ the use of
personal computers. Since many of the calculations are beyond the capability of Simulation
Programs with Integrated Circuit Emphasis (SPICE), another type of general-purpose soft-
ware is needed mathematical software.
A Mathcad worksheet is used to illustrate the details of each computation. Since
mathematical notation is used throughout in these programs, they are much easier to follow
than ones written in, say, the JAVA language. A few special features of Mathcad are
described in Appendix A. After reading this appendix, it should not be difficult for
any reader to understand the contents of the worksheets. Since every worksheet is fully
explained, the reader is not left to formulate his or her own program from a set of mathe-
matical relationships.
In the following pages, the parameter dimensions adhere to those of the SI System.
1.1.8 Testing
The ability to analyze the mechanisms involved in coupling interference into and out of the
signal link-under-review leads to the ability to devise tests which measure the actual cou-
pling parameters and correlate them with those of the relevant circuit model. A connection
has been established between test and analysis.
4 CHAPTER 1 Introduction
Since circuit theory is capable of analyzing signals in either the time domain or the
frequency domain, it becomes possible to identify the most probable cause of unexpected
interference during product development. It is but a small step to modify the design of the
signal link, build a prototype of the new link, check its performance using bench test
equipment, create a representative circuit model, and prepare a progress report. More than
that, it becomes possible to check functional performance against formal requirements.
That is, EMC can be subjected to the same design process that applies to every other
performance requirement.
Lc1
near end far end
Vc1
Vc2
Ic2 Ic1
Since current must flow along a conductor in order to flow into the capacitance, and since
a cable behaves in the same manner no matter at which end voltages are applied, the most
logical way of simulating the combined effect of inductive and capacitive coupling would be
to use the circuit of Figure 1.2.3.
This model includes resistors to represent the effect of the series resistance of each
conductor.
3 6
using one theory are always distinguishable from equations derived from the other. This can
be done by defining different types of parameter and by assigning a unique symbol to each.
Four distinct types of parameter can be identified; primitive, partial, loop, and circuit.
A primitive parameter is one which relates the current in a conductor of circular section
to the energy level of the electromagnetic field associated with that current.
A partial parameter is one which relates the current in a conductor of any cross-section to
the energy level of the electromagnetic field associated with that current.
Loop parameters are derived from primitive parameters and are used in equations which
relate loop voltages to loop currents. They are the parameters which can be measured
directly by electronic test equipment.
Circuit parameters are those which appear in circuit diagrams.
These parameters are described more fully in Chapters 2 and 3.
Electromagnetic theory invokes the concept of distributed parameters; resistance per
meter, capacitance per meter, inductance per meter, and conductance per meter. Chapter 4
identifies a simple transformation formula which obviates the need to use these
parameters.
Reflections can occur at transmission line terminations. Incident current flows in one
direction along a conductor while reflected current flows in the opposite direction. Total
current is the sum of these two partial currents. The term partial is also used to identify the
associated voltages.
1. Define the length of the assembly, the radii of the conductors, and the spacing between
the centers.
2. Establish a set of three primitive equations relating the voltage on each conductor to the
current in all three conductors.
3. Define the loops in terms of conductor pairs.
4. Derive a set of loop equations.
5. Define the loop inductors and loop capacitors in terms of the primitives.
6. Postulate the existence of a circuit model which creates two mesh equations.
7. Relate the components of the circuit model to the constants in the loop equations.
8. Relate the components of the circuit model to the primitive parameters.
The key feature to note in the above process is that it contains a discontinuity. Step (6)
does not follow logically from step (5). Lateral thinking is needed. The purpose of the circuit
model is to create a set of mesh equations which correlate precisely with those of the loop
equations. Mesh equations are derived using the rules of circuit theory. Loop equations are
derived from the relationships of electromagnetic field theory.
Setting up such a relationship is conceptually the same as defining x as an unknown
variable. If circuit theory is treated in this way, it can be utilized to simulate all types of
electromagnetic coupling.
It is useful to always bear in mind the fact that circuit theory does not define the physical
mechanisms. Concepts such as the equipotential ground plane and the single point
reference are convenient assumptions which allow circuit models to simulate the behavior
of complex printed circuit boards in signal processing equipment. In so doing, they
carry with them the unstated assumption that there is no such thing as electromagnetic
interference.
Since mesh analysis caters for the fact that partial currents can flow in both directions
along a conductor at the same time, this form of analysis is used in the following pages.
short
wavelength
The velocity of propagation of a signal along the conductor of a transmission line is the same
as that of the associated electromagnetic wave. The electrical length k of a conductor of
length l can be defined as:
l
k
l
The usual textbook definition of a short electrical length is that:
k < 0:1
Figure 1.2.5 illustrates the relationship between short electrical length and wavelength.
Experience has shown that the circuit model of Figure 1.2.3 is reasonably accurate, provided
the conductor is electrically short. Equally, it can be said that the model is reasonably
accurate up to a frequency at which the wavelength is ten times the length of the assembly-
under-review.
Vn Z2 Vf
In If
1 4
Z11 Z11
Ic1 Ic1 Ic3
Vc1
2 5
Z12 Z12
Ic2 Ic4
Vc2 Ic2
Z13 Z13
3 6
Although these impedances are derived in an unconventional way, they still evaluate to the
form R j X , where X is the value of the reactance at frequency f.
Since there is a clear correlation between the models of Figures 1.2.3 and 1.2.7, the
transformation process is extremely simple. The three-conductor line can be defined in terms
of the lumped parameter model and analyzed using the distributed parameter model. This
means that the designer retains visibility of the properties of the model throughout the
analytical process. This method of simulation is much simpler than stringing dozens of
lumped parameter models in series.
Using the distributed parameter model, the maximum frequency of the simulation is no
longer limited by the length of the cable. But there is still a limit. It is assumed that action
and reaction between adjacent conductors in any cross section of the cable is instantaneous.
By analogy with the limitation defined in section 1.2.6, the maximum frequency is that at
which the wavelength is ten times the maximum spacing between conductors.
Best accuracy is obtained if the cross section of the cable is exactly uniform along its
entire length and that the lengths of the three conductors are identical. If these requirements
are not met because different types of cable are used, or because intermediate connectors are
included in the wiring harness, then the accuracy of the simulation will be reduced. How-
ever, accuracy can be restored by carrying out bench testing on a representative assembly
and creating a circuit model which replicates the test results.
1.3 Intra-system interference 11
1 send conductor 4
2 differential-mode loop 5
Vsig return conductor
3 common-mode loop 6
structure
return conductor
send conductor
structure
image conductors
conduit
A more accurate method of deriving values for the inductance and capacitance of the
structure is to use the method of composite conductors, as shown by Figure 1.3.3. In this
simulation, all other conductors in vicinity of the send and return conductors are assumed to
form a single composite conductor; and the properties of that conductor are assigned to the
structure.
As far as this method of simulation is concerned, it does not matter that the other
signals in the system are interfering with each other. All that matters is how they affect
the link-under-review and how the signal in the link under-review couples with the
structure.
Chapter 3 derives computer programs for calculating values for the components of
Figure 1.2.3, using data on the geometry of the structure and cables.
send
Ic2 Ic4
ratio of current in the common-mode loop to the voltage in the differential-mode loop gives a
value for the transfer admittance YT.
Ic2
YT
Vc1
Ic1 Ic3
return
Ic2 Ic4
structure
Vc2
used to simulate the effect of this interference. Again, the transfer admittance characteristic
can be used to define the susceptibility of the link:
Ic1
ZT
Vc2
However, it will always be possible to modify a few parameters which define the
component values of the model to achieve close correlation between the two curves. The
number of independent variables is much fewer than the number of components.
Experience has shown that it does not take many iterations of the simulation to achieve
this objective.
The end result of such an exercise is the creation of a circuit model which defines the
coupling characteristics of the signal link. Such a model can be utilized in much the same
way as one of the modules in the library of components available in most SPICE software
packages.
Advantage can be taken of the fact that every parameter (R, L, C, and G) of the model is a
function of length. Tests can be carried out on a long test rig at low frequencies and used to
create a model of that rig. Knowledge of the length of that rig can be used to modify the
component values in the model, to allow the modified model to simulate the behavior of a
much shorter link. Tests at relatively low frequencies can be used to predict performance at
much higher frequencies. Such an approach is analogous to the use of wind-tunnel tests on a
small-scale model to predict the behavior of the actual aircraft.
This means that a circuit model for a 10-m line, tested at frequencies up to 20 MHz, can be
used to create a model for a 1-m line, valid at frequencies up to 200 MHz. Care would need to be
taken that the cross section of the test rig is constant over the entire length and that the com-
ponents at the interfaces are suitable for use at the frequency at which the equipment will operate.
The method can be used to simulate the cross-coupling between two 50 mm lengths of
printed circuit track, valid up to 4 GHz. Testing the accuracy of this particular model would
require the use of sophisticated test equipment.
Most of the tests described in Chapter 7 are limited to the short-wave band. But this is
only because the test equipment available was limited to 20 MHz.
The model itself is limited only by the assumptions inherent in the theory of transmission
lines. That is, it is assumed that action and reaction between adjacent conductors is instan-
taneous. In practice, the upper frequency limit is that at which the wavelength is ten times the
maximum dimension of any section of the assembly-under-review.
Section 7.7 shows that the technique of circuit modeling can be used to characterize small
components such as capacitors, at frequencies over the range 200 kHz to 1 GHz.
transmitter
75
co-axial cable
monopole monopole
A commonly accepted formula for the average radiated power Pt over a spherical
surface is:
1
Pt Rrad Ip2
2
where Ip is the peak amplitude of the current. The parameter Rrad is not a resistor in the
conventional sense of the word, but a mathematical constant derived from a complex process
of integration which happens to have the dimensions of resistance. For a half-wave antenna
the radiation resistance, Rrad, is 73 W.
Since the current distribution along a monopole of a half-wave dipole is identical to
that along an open-circuit transmission line operating at the quarter-wave frequency and since the
T-network model of Figure 1.2.6 simulates the behavior of the line, it is logical to assume
that this model can also simulate the behavior of a monopole. This leads to the model of Fig-
ure 1.4.1, where Lp and Cp are the primitive inductance and primitive capacitance of a monopole.
Tests on a single length of isolated conductor, when it is acting like a dipole, are described
in section 7.4. The fact that there is close correlation between the responses of the model and
the hardware provides a high level of confidence in the soundness of the technique.
virtual conductor
Although most of the transmitted energy is picked up by the receiving antenna, a significant
proportion constitutes radiated interference.
Analysis of the geometry of the twin-conductor cable provides formulae for the induc-
tance Lrad and capacitance Crad. The other components remain exactly the same as those of
the transmission-line model of the cable.
Since the components Lrad, Crad, and Rrad are also configured as a T-network, it is
logical to give it the name virtual conductor, since it does not simulate an actual piece if
hardware. It represents the effect of the environment.
The amplitude of the current radiated into the environment Irad gives a measure of
the radiated field strength. The maximum strength of the magnetic field H at a distance
r is:
Irad
H
2pr
This formula is recognisable as that derived by integrating the Biot-Savart equation along an
infinitely long, straight conductor, as (2.2.1) would indicate. However, it can also be derived
from the equations of antenna theory. Section 5.7 shows that it is related to the maximum
power delivered to the environment by a wire pair routed along the structure. It can be used
to predict the results of formal EMC testing of radiated emission of the equipment-under-
review, by indicating the maximum strength of the signal which would be picked up by the
monitor antenna.
The amplitude of the threat voltage Vthreat can be calculated by integrating the value of
l
the electric field strength E over the length l of the cable. In the case where l :
4
l
Vthreat E
p
Vthreat is the maximum voltage which can be injected into the antenna-mode loop by an
electromagnetic field of electrical field strength E.
Analysis of this model allows the differential-mode current caused by an external field to
be calculated.
voltage source Vthreat in series with Rrad. One way of simulating the effect of the threat
environment is to insert both these components in series with the structure. Section 5.5
illustrates how the response of such a signal link can be analyzed.
Similarly, if the structure provides no shielding, then current in the structure can be
regarded as the source of interference. Since this conductor carries the common-mode cur-
rent, then knowledge of the amplitude of the common-mode current will provide a first
estimate of the level of emission which can occur.
1.5 Transients
Transients are an ever-present source of glitches in electronic systems. Sources can be relays,
switches, motors, and power supplies. They can easily corrupt the data streams handled by
microprocessors. Depending on the criticality of the processing circuitry, such events could
be inconsequential, annoying, dangerous, or catastrophic.
Since most signal processing is now carried out by digital signals, it is essential that this
topic be included in any analysis of interference. The lumped parameter models developed
for frequency response analysis can also be used for analysis in the time domain. The text-
book approach to such analysis is to invoke the use of Fourier transforms, Laplace trans-
forms, and even more complex techniques. The approach adopted here follows the example
of SPICE programs and uses time-step analysis.
However, the problems encountered with the analysis of the frequency response of a
transmission line also appear in the use of transient analysis. It takes a finite time for a signal
to traverse from one end of the line to another. With frequency analysis, it was possible to
transform the lumped parameter model into a distributed parameter model. With transient
analysis, another solution is called for.
In concept, the solution is much simpler; use the computer memory to store the signal
applied to the near end of the line for a fixed number of time steps before delivering it to the
terminals at the far end.
Books on electromagnetic theory introduce the concept of partial currents and partial
voltages to explain the behavior of transient signals at the interface between cables and
equipment terminations. Incident current flows toward the interface; some of it is absorbed
by the equipment and some is reflected back down the line. The total current at any section
of the cable is the sum of the incident and reflected currents at that location. Section 6.2
describes the phenomenon in more detail.
The reflected signal is also delayed a fixed number of time steps before appearing back at
the near end. A simple program to simulate the propagation of incident and reflected currents
is also described in section 6.2.
An experiment was carried out by applying a square wave to one end of a twin-core
cable via a 5-W resistance, leaving the far end open-circuit and monitoring the current
flowing in the line. Since the impedance at the near end was much lower than the char-
acteristic impedance of the line, it was expected that there would be multiple reflections.
Attention was focused on the response to the leading edge; that is, on the response to a step
input.
Textbook theory predicted that the incident current would be inverted by the open-circuit
terminals at the far end and reflected straight back to the near end. At the near end, the
20 CHAPTER 1 Introduction
inverted current would almost double in amplitude. The response to a step input was
expected to be a square waveform of slowly diminishing amplitude. It wasnt.
During the time the leading edge of the pulse took to make the return trip, the current
remained steady. This was in accordance with classical theory. However, the trailing edge was
not sharp, as one would expect. It took the form of an exponential decay. This process continued
and the waveform underwent a metamorphism, rapidly changing shape into a sine wave.
A trial-and-error process ensued in an effort to create a circuit model which replicated the
observed waveform and which could be explained in terms of electromagnetic theory. It was
eventually reasoned that, as the step waveform propagated along the line, it left a residual
charge on the conductors and that this charge gradually decayed away via current flow into
the environment. This charging current could be simulated.
In addition, it proved to be possible to measure and simulate the proportion of the current
which departed into the environment. Results from these transient tests could also be cor-
related with observations and analysis of tests which had been made using sinusoidal
waveforms. Section 6.6 explains the reasoning.
In hindsight, the phenomena can be explained by the fact that the antenna-mode waves
propagate mainly in the air, whereas the differential-mode waves are mostly in the cable
insulation. From the viewpoint of field propagation, this explanation comes quite readily to
mind. In this case, however, the evidence comes from observations on the behavior of cur-
rents and voltages rather than the analysis of H-field and E-fields.
To someone who thinks in terms of system function, this is something of a revelation.
Although the electromagnetic fields propagate in the insulating medium, the currents which
create them flow in the conductors. The antenna-mode current and the differential-mode
current are separate entities, just as surely as the incident and reflected currents in trans-
mission lines are independent of each other.
This conclusion is supported by the analysis of the transient tests described in section 7.6.
In the setup of Figure 7.6.1, a step voltage is injected into the signal conductor of a conductor
pair. This creates an antenna-mode current which propagates in the same direction down the
pair of conductors, from near end to far end. Current in the signal conductor also causes a
current to flow back along the return conductor.
Just behind the leading edge of the antenna-mode current step, current is flowing in both
directions along the return conductor. Current flowing in opposite directions in a pair of
conductors constitutes differential-mode current. The leading edge of the differential-mode
step follows behind the leading edge of the antenna-mode current, at a lower velocity. From
the results of the tests of section 7.5, the two velocities are:
Antenna-mode current: 230 m/ms
Differential-mode current: 170 m/ms
That is, the use of circuit modelling techniques provides the user with an ever-improving
understanding of the mechanisms involved in interference coupling.
The key point here is that progress can only be made by engineers who carry out mea-
surements for themselves.
An equally important aspect of testing is that it provides firm evidence of the validity of
the model. Being able to verify theoretical results by comparing them with other theoretical
results is encouraging; but it does not provide the confidence achievable when practical
measurements correlate with theoretical predictions.
interface circuit in the light of the underlying physics. Given this ability, it is possible to
identify groups of techniques which have the same objective. These are:
Common-mode rejection
Differential-mode damping
Common-mode damping
Shielding
The design of the interface circuitry and system shielding is critically dependent on an
understanding of cable-coupling mechanisms. Every circuit described in Chapter 8 includes
a definition of the complete signal link; there are no loose ends where the circuit terminates
at a set of plug pins or socket inserts. This allows the design of the interface circuitry to be
related to the coupling characteristics of the cable. For example, the performance char-
acteristics of grounded and floating configurations are compared.
Packing density on printed circuit boards (pcb) precludes the option of carrying out a
detailed analysis of every signal link on a board. Even so, anyone who understands the
method of modeling will also understand the interference coupling mechanisms, and will
avoid most of the obvious errors.
It would be good practice to fit a buffer circuit at each pcb interface to decouple the signals
on the interconnecting cables from the multiple branches and stubs that must exist on the board.
A variety of such buffer circuits is described in Chapter 8. Any interference experienced would
then be due to internal coupling. A well-designed board would not experience such coupling;
and this could be confirmed the first time the board is functionally checked.
Since there are books which provide excellent advice on the detailed design of pcbs,
there is no point in duplicating their content. One book which stands out by virtue of the
wealth of detailed design information it provides is EMC Design Techniques for Elec-
tronic Engineers [1.10]. Another book which is well worth purchasing is Introduction to
Electromagnetic Compatibility [1.11], since it provides detailed information on the the-
ory underlying electromagnetic field propagation and relates this to practical design
problems. The chapters on shielding and electrostatic discharge also provide valuable
information.
4. Design the interface circuitry to enhance the balance between the current in the send
conductor and the current in the return conductor. Ideally, the net current flow through
any cable section should be zero.
5. Treat every signal link as a transmission line and use resistors similar in value to the
characteristic impedance in the interface circuitry; as far as is practicable.
6. Implement common-mode damping on critical signal links.
7. Test and analyze critical signal links.
A review of the method used to calculate circuit component values for three-phase power
lines leads to the identification of the first step; the derivation of a formula relating the
capacitance of an isolated conductor to its length and radius. This formula can be regarded as
a basic building block from which the capacitive parameters of multi-conductor assemblies
can be derived.
In computer terminology, a low-level object or operation from which higher-level, more
complex objects or operations can be constructed is termed a primitive. So it seems rea-
sonable to use the term primitive capacitance to identify the basic relationship.
The first equation quoted in section 2.1 can be found in textbooks on electromagnetic
theory [2.1]. Although the derivation involves some complex integration, the end result is a
formula which defines the primitive capacitance, Cpi,j. This relates the voltage on conductor
i to the charge on conductor j.
Primitive inductance, Lpi,j, can be regarded as the twin of primitive capacitance. In this
case, textbooks can be found which outline the derivation process [2.2]. However, these
earlier derivations do not include the contribution made by internal linkages. Nor do they
cater for the fact that in any circuit analysis the current must flow in a complete loop. The
derivation provided by section 2.2 includes these considerations.
In these earlier textbooks, the name given to the derived parameter is partial inductance.
However, in the analysis of Chapter 3, an array of elemental conductors is used to represent a
composite conductor. In this treatment, the partial parameters associated with each com-
posite conductor are derived from a collection of primitive parameters. With any computa-
tion process, it is necessary to distinguish between the two types. So they have been given
different names.
Section 2.3 identifies the duality between primitive inductance and primitive capacitance.
The square root of the product of these two parameters is the propagation delay; the time it
takes for a transient pulse to propagate along the length of a conductor. Primitive parameters
define the properties of conductors when they are acting as antennae.
Section 2.4 describes how primitive parameters can be combined to derive the capaci-
tance between a pair of parallel conductors, and the inductance of a loop formed when the
terminals at the far end are connected. Loop parameters are those which can be measured
with electronic test equipment.
25
26 CHAPTER 2 Lumped parameter models
However, the screens of most co-axial cables are braided. At higher frequencies a number
of capacitive and inductive effects due to the screens construction come into play and
dominate the effect that is caused by screen resistance. The effect can be simulated by
representing the screen as a T-network, with very small values assigned to the reactive
components. These values can be derived from manufacturers data on transfer impedance,
or from tests on a cable sample.
Analyzing the coupling between two conductors over a ground plane is a problem faced
by most designers at one time or another. The method of images can be used to derive
formulae for the components of the three-conductor model. The formulation is described in
section 2.11.
A basic assumption inherent in the textbook theory of circuits is that resistors, inductors,
capacitors, voltage sources, and current sources are lumped parameters; that is, they are
discrete devices which link two nodes of a network. Another assumption is that action and
reaction are instantaneous everywhere in the network. Since conventional circuit theory does
not cater for the fact that it takes time for current to propagate along a conductor, its use is
limited to low-frequency applications, where the length of the cable is less than one-tenth the
wavelength of the signal.
This limitation applies to the models derived in this chapter and in Chapter 3. Subsequent
chapters describe how this restriction can be overcome.
1
l P(r, z)
2
l/2
r
r1,1
The radial component of the electric field strength at the point P is [2.1]:
r 1
Er sina1 sina2 V=m 2:1:1
4pe r
where e is the permittivity.
Since the point P lies on the plane intersecting the mid-point of the conductor, a1 is equal
to a2. Re-writing (2.1.1) in terms of radial and axial components gives:
r l=2
Er q 2:1:2
2pe
r l=22 r2
At all points on the surface defined by the equation z 2l the Ez component of the electric
field is zero. If unit charge were to follow a path along this surface from infinity to the radius
r1,1 then no axial force would be involved; all the electric forces would be radial. The work
done on conductor 1 due to the charge on conductor 1 would be:
1
r l=2
Vp1;1 q dr V 2:1:3
2pe
r1;1 r l=22 r2
The capacitance of the conductor is the ratio of charge to voltage. That is:
rl
Cp1;1 F 2:1:6
Vp1;1
r coulombs/metre Vp2
conductor 2
conductor 1
r1,1 r2,2
r1,2
where i and j are integers which identify the conductors. If j i then the value is the pri-
mitive capacitance of conductor i due to the charge it carries.
This formulation depends on the assumption that the length is much greater than any
radius and that the charge is uniformly distributed along the length. It has the practical
advantage that it is simple. It works well in practice, in that it can be used as a building block
in the construction of all circuit models. Moreover, any inaccuracies in the derivation are
completely masked by the uncertainty in the value of the relative permittivity.
Capacitance values are a function of length, radii, and permittivity. Length and radii can
be determined simply by measuring the physical parameters. Permittivity is a combination of
two parameters:
e eo e r F=m 2:1:11
where eo is defined as the permittivity of free space, 8.854 pF/m, and er is the relative
permittivity.
In classical theory, the value of the relative permittivity is different for each type of
material. When dealing with EMC problems, a variety of insulating materials are present in
the vicinity of the conductor. Rather than taking the individual relative permittivity of each
material and using data on the cross section of the assembly to calculate the overall effect,
30 CHAPTER 2 Lumped parameter models
an approximate overall or effective value can be used. In this book er is defined as the
effective value of the relative permittivity. Its value can be determined by tests such as those
described in sections 7.4 and 7.5.
Ip
1
P(r, z)
2
l
r
r1,1
and
f B ds Wb 2:2:4
The above derivation can be found in at least one textbook [2.2]. Inductance is defined as the
ratio of magnetic flux to the current creating that flux. Since the flux f is external to the
conductor:
f
Lexternal H 2:2:5
Ip
To a first approximation:
ml l
Lp1;1 ln 2:2:9
2p r1;1
When the pair of conductors is analyzed as part of a complete loop, it must be assumed that
the return path for the current Ip is along conductor 2. Magnetic flux due to current Ip in
conductor 1 will link the region between infinity and the outer surface of conductor 2. It will
also counteract the internal linkages of the return current Ip in conductor 2.
Hence, it is necessary to add the component Linternal to the value of the mutual
inductance.
This leads to:
L2;1 Lmutual Linternal
" p! p! #
ml l l2 r2;1 2 r2;1 l2 r2;1 2 1
ln 2:2:11
2p r2;1 l 4
Ip
Ip
Ip
r1,1 r2,2
r2,1
mo mr l l
Lpi;j ln H 2:2:13
2p ri;j
m mo m r H=m 2:2:14
where mo 4 p 107 H/m. The parameter mr is a pure number, similar in concept to er. It
can be defined here as the effective value of the relative permeability of the loop-under-
review.
In this book, it is assumed that the value of mr is unity because in the majority of
applications there is no magnetic material in the cable assembly. In those situations
where magnetic material is present, the value of mr can be established by referring
to data on the properties of materials, or by carrying out tests of the assembly-under-
review.
Such tests would involve measurements of the frequency response of the current in a
short-circuited loop. A circuit model of the assembly can be created by using data on the
physical construction. This will produce a similar curve. The initial slope of both frequency
response curves will be 20 dB per decade. By modifying the value of mr in the model, both
curves can be made to coincide over this range. The value of the relative permeability of the
loop-under-test will be that of the model.
The test described in section 7.5 illustrates the process used to measure er, using an open-
circuit line. If both ends of the line are short-circuited, the same process can be used to
measure mr .
34 CHAPTER 2 Lumped parameter models
and
mo mr l l
Lpi;j ln H 2:3:2
2p ri;j
where i and j are integers which identify the conductors. For example, Lpi;j gives the value of
the primitive inductance of conductor i due to current in conductor j. If j i then the value is
the primitive inductance of conductor i due to the current it carries.
Since ri;j rj;i , then Lpi;j Lpj;i and Cpi;j Cpj;i .
Multiplying gives:
where l is the wavelength and f is the frequency. At the quarter-wave frequency, fq, of a
monopole antenna, the current amplitude will peak. If the wavelength at this frequency is lq.
then the length of the cable will be:
lq
l m 2:3:7
4
p
Taking the square root of (2.3.3) and using (2.3.4) to substitute for mo mr eo er gives:
p l
Lpi;j Cpi;j s 2:3:8
v
The ratio l/v is the propagation delay; the length of time it takes a transient pulse to propagate
along the length of the conductor assembly. It can also be defined as the time constant T,
2.4 Loop parameters 35
as used in transient analysis. Section 6.3 derives its relationship to the inductance, capaci-
tance, and characteristic impedance of the line.
Using (2.3.6) to substitute for v and (2.3.7) to substitute for l provides a relationship
between the circuit components and the first resonant frequency:
p 1
Lpi;j Cpi;j 2:3:9
4 fq
This means that, if the values of Cpi,j and the frequency of quarter-wave resonance, fq, are
known, then the value of Lpi,j can be calculated.
Rearranging (2.3.7) gives:
lq 4 l
Vp1 Ip1
Vp2 Ip2
It is possible to deal with inductive and capacitive effects separately. Voltages Vp1 and
Vp2 can initially be defined as the energy levels due to magnetic effects on conductors
1 and 2. If it is assumed that currents and voltages are sinusoidal functions of time, then the
relationship between them can be defined. Restricting consideration to magnetic effects, the
primitive equations for two conductors are:
Vp1 j w Lp1;1 Ip1 j w Lp1;2 Ip2
2:4:1
Vp2 j w Lp2;1 Ip1 j w Lp2;2 Ip2
Primitive equations define the behavior of the conductor assembly as an antenna. When
analyzing the behavior of the assembly as part of a circuit, it is necessary to deal with loop
currents and loop voltages, as illustrated by Figure 2.4.2.
The relationship between voltages is:
Vl Vp1 Vp2 2:4:2
Using (2.4.3) to substitute loop currents for primitive currents in (2.4.1) and then invoking
(2.4.2) leads to:
Vl j w Lp1;1 Lp1;2 Lp2;1 Lp2;2 Il 2:4:4
This gives the value of the loop inductance of the conductor pair:
Ll Lp1;1 2 Lp1;2 Lp2;2 2:4:5
Vl Il
This relationship takes into account the fact that Lp1;2 Lp2;1 . Using (2.3.2) to substitute for
the primitive inductors:
m0 mr l r1;2 r1;2
Ll ln 2:4:6
2p r1;1 r2;2
The step between (2.4.5) and (2.4.6) is highly significant, in that the length parameter dis-
appears from the logarithmic term. There is no longer any inaccuracy due to variation in the
ratio between length and separation distance at any particular cross section of the assembly-
under-review. It does assume that the cross section is uniform along the length l and that
l > 10 r.
If l 10 r, then propagation time needs to be considered and it becomes necessary to
invoke the concepts of electromagnetic field theory. Even then, however, any inaccuracy in
the predicted value of inductance does not negate the fact that loop inductance exists. The
inaccuracy can be catered for by carrying out tests on the assembly-under-review.
Loop capacitance can be derived the same way. The primitive equations are:
1 1
Vp1 Ip1 Ip2
j w Cp1;1 j w Cp1;2
1 1 2:4:7
Vp2 Ip1 Ip2
j w Cp2;1 j w Cp2;2
Using (2.4.3) to substitute loop currents for primitive currents in (2.4.7) and then invoking
(2.4.2) leads to:
1 1 1 1 1
Vl Il 2:4:8
j w Cp1;1 Cp1;2 Cp2;1 Cp2;2
Hence:
1 1 1 1 1
2:4:9
Cl Cp1;1 Cp1;2 Cp2;1 Cp2;2
So:
2 p eo er l
Cl 2:4:10
r1;2 r1;2
ln
r1;1 r2;2
Equations (2.4.6) and (2.4.10) define the loop inductance and loop capacitance of a con-
ductor pair. These are the values which could be measured with an LCR meter.
This derivation makes the assumption that all the current flowing in one conductor
returns via the other. The nature of electromagnetic coupling ensures that this never happens
38 CHAPTER 2 Lumped parameter models
in practice. Nevertheless, (2.4.6) and (2.4.10) can be regarded as providing a good approx-
imation to the properties of an isolated twin-conductor cable. Section 5.2 develops these
relationships to deal with the properties of such a cable as an antenna.
2.5.1 Inductance
The primitive equations for inductance define the voltage developed along each conductor in
terms of the primitive currents. Reproducing (2.4.1):
Vp1 j w Lp1;1 Ip1 j w Lp1;2 Ip2
2:4:1
Vp2 j w Lp2;1 Ip1 j w Lp2;2 Ip2
Equations (2.4.1) and (2.5.1) were derived using different criteria. Equation (2.4.1) was
derived from electromagnetic theory, while (2.5.1) came from circuit theory. Different
assumptions were involved in their derivation. The equations can be correlated by defining
the relationship between voltages and currents:
Vp1 Vc1
2:5:2
Vp2 Vc2
Lc1
Ic1
Lc2
Vc2
and
Ip1 Ic1
2:5:3
Ip2 Ic1
Making these substitutions allows the circuit parameters to be derived from the
primitives:
Lc1 Lp1;1 Lp1;2
2:5:4
Lc2 Lp2;2 Lp1;2
Using (2.3.2) to relate primitive inductors to physical parameters gives formulae for the
circuit inductors for a conductor pair:
m m l r1;2
Lc1 o r ln
2p r1;1
2:5:5
mo mr l r1;2
Lc2 ln
2p r2;2
2.5.2 Capacitance
Figure 2.5.2 illustrates the capacitive parameters associated with each conductor.
The primitive equations for capacitance define the voltage developed between each
conductor and a theoretical surface at zero voltage in terms of the current flowing into the
environment or out of the environment. Reproducing (2.4.7):
1 1
Vp1 Ip1 Ip2
j w Cp1;1 j w Cp1;2
2:4:7
1 1
Vp2 Ip1 Ip2
j w Cp2;1 j w Cp2;2
Cc1 Vc1
Ic1
zero voltage
Cc2 Vc2
The circuit equations for capacitive coupling derived from Figure 2.5.2 are:
1
Vc1 Ic1
j w Cc1
2:5:6
1
Vc2 Ic1
j w Cc2
Invoking (2.5.2) and (2.5.3) to substitute circuit parameters in the primitive equations
leads to:
1 1 1
Vc1 Vp1 Ic1
j w Cp1;1 Cp1;2
2:5:7
1 1 1
Vc2 Vp2 Ic1
j w Cp2;2 Cp2;1
Correlating (2.5.6) and (2.5.7) allows circuit capacitors to be defined in terms of primitive
capacitors:
1 1 1
Cc1 Cp1;1 Cp1;2
2:5:8
1 1 1
Cc2 Cp2;2 Cp1;2
Using (2.3.1) to relate primitive capacitors to length and radii gives formulae for the circuit
capacitors for a conductor pair:
2 p eo er l
Cc1 r1;2
ln
r1;1
2:5:9
2 p eo er l
Cc2 r1;2
ln
r2;2
2.5.4 Resistance
The resistance of each conductor is a function of its cross-sectional area and its length. If
conductor i is circular in section, the steady-state resistance is:
rl
Rssi ohm 2:5:11
p ri;i 2
where r is the resistivity of the conducting material. Skin effect makes resistance a function
of frequency. Textbooks on electromagnetic theory derive the formula [2.4]:
r
l mf
Rskini ohm 2:5:12
2 ri;i ps
1
r ohm m 2:5:13
s
At high frequencies, resistance increases as the square root of the frequency; that is, at
10 dB per decade. The crossover point occurs when Rskin is equal to Rss. From (2.5.11),
(2.5.12) and (2.5.13):
r
l m r Fx rl
2 ri;i p p ri;i
2
The graph of Figure 2.5.3 illustrates the variation of resistance with frequency. The curves for
Rc and Rskin were derived from (2.5.15) and (2.5.12), respectively. These give the relationships
for a 15-m length of 1-mm diameter copper conductor. In this case the crossover frequency is 69
kHz. This leads to a third representation of a twin-conductor cable; Figure 2.5.4.
Unlike inductance and capacitance, resistance is independent of the current in other
conductors. This means that:
Rci Rpi;i 2:5:16
42 CHAPTER 2 Lumped parameter models
10
Rc
Rskin
0.1
0.01
1 103 1 104 1 105 1 106 1 107
f Hz
Rc1
Rc2
Cc1 Cc2
node at zero voltage
every conductor possesses these properties, it is certainly not valid to assume that a zero-
impedance path for current exists anywhere in the system.
Since the simplest configuration capable of efficiently carrying an electronic signal from
one location to another is the twin-core cable, the best place to start is to analyze the prop-
erties of such a cable. Figure 2.6.1 departs from the conventional approach by recognizing
the fact that both conductors have the same set of properties.
The most significant property of this model is the duality of the inductors and capacitors.
If a voltage source is connected to the near end and the terminals at the far end are short-
circuited, then, as far as reactive parameters are concerned, the configuration behaves as a
bridge network.
From (2.5.10):
Hence:
Lc1 Cc2
2:6:1
Lc2 Cc1
I1 I2
V1 2:6:2
j w C1
I1 I2
V2 2:6:3
j w C2
Lc1
V3 j w I2 2:6:4
2
Lc2
V4 j w I2 2:6:5
2
44 CHAPTER 2 Lumped parameter models
Lc1 V3
near end far end
2
Cc1 Lc1
Vin V1 2
Cc2 V2 Lc2
2
Lc2
V4
2
Hence:
V1 Cc2
2:6:6
V2 Cc1
and
V3 Lc1
2:6:7
V4 Lc2
From (2.6.1):
V1 Cc1 Lc2 V3
2:6:8
V2 Cc2 Lc1 V4
Since the voltage at the junction of the capacitors is zero, the voltage at the junction of the
inductors is also zero. Hence the bridge circuit can be redrawn, as shown in Figure 2.6.3.
This form of the circuit model indicates that it is valid to represent the reactive compo-
nents of each conductor as a single impedance; a parallel combination of inductance and
capacitance in series with an inductance. Such a representation is extremely useful in sim-
plifying the analysis of the transfer admittance in section 2.9 and in deriving parameter
values for the virtual conductor in section 5.2.
Lc1 Lc1
near end far end
2 2
Vin Cc1
Cc2
Lc2 Lc2
2 2
Figure 2.7.2 illustrates a configuration in which the cable assembly is terminated by short-
circuits at both ends, with voltage sources Vl1 and Vl2 inserted in the loops at the near end.
Comparing Figures (2.7.1) and (2.7.2) allows the relationships between primitive para-
meters and loop parameters to be defined. Voltages are related by:
Ip1 Il1
Ip2 Il2 Il1 2:7:3
Ip3 Il2
Ip2
Vp2
Ip3
Vp3
Vl1 Il1
Vl2 Il2
Vp1 Vp2 Zp1;1 Zp1;2 Zp2;1 Zp2;2 Il1 Zp1;2 Zp1;3 Zp2;2 Zp2;3 Il2
Vp2 Vp3 Zp2;1 Zp2;2 Zp3;1 Zp3;2 Il1 Zp2;2 Zp2;3 Zp3;2 Zp3;3 Il2
2:7:5
where:
The circuit model which can simulate the behavior of the loop equations is shown in
Figure 2.7.3. The circuit equations for three conductors are:
A comparison between (2.7.6) and (2.7.9) shows clearly that there is a one-to-one correlation
between loop equations and circuit equations Hence, loop impedances can be related to circuit
impedances:
Zc1
Vc1 Ic1
Zc2
Vc2 Ic2
Zc3
Using (2.7.7) to substitute primitive impedances for loop impedances gives the circuit
impedances for three conductors in terms of the primitive impedances:
Zc1 Zp1;1 Zp2;1 Zp1;3 Zp2;3
Zc2 Zp2;2 Zp1;2 Zp2;3 Zp1;3 2:7:11
Zc3 Zp3;3 Zp3;1 Zp2;3 Zp2;1
If the radial dimensions of the conductor assembly are defined as shown in Figure 2.7.4, then
circuit components can be defined in terms of spatial parameters.
If the impedances are assumed to be purely inductive, that is Zpi;j j w Lpi;j , then
(2.3.2) can be used to define the circuit inductors for three conductors:
mo mr l r1;2 r1;3
Lc1 ln
2p r1;1 r2;3
m m l r1;2 r2;3
Lc2 o r ln 2:7:12
2p r2;2 r1;3
mo mr l r1;3 r2;3
Lc3 ln
2p r3;3 r1;2
r2,2
r1,2
r1,1
r2,3
r1,3
r3,3
Vc1
Zc2 Zc3
Zc1
Ic1 Ic2
Making this substitution in (2.7.11) and then invoking (2.3.1) leads to formulae for the
circuit capacitors for three conductors:
2 p eo er l
Cc1 r1;2 r1;3
ln
r1;1 r2;3
2 p eo er l
Cc1 r1;2 r2;3
ln 2:7:13
r2;2 r1;3
2 p eo er l
Cc1 r1;3 r2;3
ln
r3;3 r1;2
Equations (2.7.12) and (2.7.13) can be correlated with the formulae developed for phase
inductance [1.7] and phase capacitance [1.8] of three-phase power lines. In fact, the deri-
vation above is simply a formalization of the process described in that basic textbook.
interface
near end connector
far end signal
receiver
signal conductor
signal
signal current
driver return conductor
interference current
shield or structure
differential-mode current
Lc3
Given this set of equations, it can be seen that reducing r12 would reduce the value of Lc1 and
Lc2, while increasing the value of Lc3. That is, altering just one physical dimension will
affect the value of all three inductors. Reducing the spacing between the signal and return
conductors will have a significant effect in improving EMC.
2.8 Optimum coupling 51
r1,2
r1,3 r2,3
If r1;3 r2;3
r1;3 r2;3
then : ln ln 0
r2;3 r1;3
That is, if the conductors carrying the differential-mode current are equidistant from the
structure, then these logarithmic factors can be deleted from (2.8.1). This means that Lc1 and
Lc2 can be reduced further.
If, in addition, the radii of the signal and return conductors were the same, then:
r1;2 r1;2
r1;1 r2;2
Cc3
0 Z2 I1 Z2 Z3 I2 2:9:2
Lc1 Lc1
2 2
Lc2 Lc2
2 2
Lc3 Lc 3
2 2
Lc1 Lc1
near end far end
2 2
V1
Cc1
Lc2
2 Lc2
2
Cc2
Lc3
2 Lc3
2
Cc3
Z1
V1 I1
Z2
I2
Z3
from (2.9.2):
Z2 Z3
I1 I2 2:9:3
Z2
substituting for I1 in (2.9.1):
Z1 Z2 Z2 Z3
V1 I2 Z2 I2
Z2
this expands to:
Z1 Z2 Z1 Z3 Z2 Z2 Z2 Z3 Z2 Z2
V1 I2
Z2
giving:
I2 Z2
2:9:4
V1 Z1 Z2 Z1 Z3 Z2 Z3
54 CHAPTER 2 Lumped parameter models
Z1
I1
Z2
I2
Z3
V2
If the voltage source is placed in the second loop, as shown in Figure 2.9.4, then the same
process leads to the ratio:
I1 Z2
2:9:5
V2 Z1 Z2 Z1 Z3 Z2 Z3
This identifies the fact that there is a duality in any circuit network. Comparing Figures
(2.9.3) and (2.9.4)
I1 I2
YT 2:9:6
V2 V 1
Since the ratio of current to voltage is termed admittance, and since the ratio is between
parameters in different loops, then the parameters of (2.9.4) and (2.9.5) are described as
transfer admittance, YT.
The transfer admittance provides a direct measure of the ratio of unwanted signal in the
victim loop to the source voltage in the culprit loop. It is the basic parameter which enables
the interference characteristic of any assembly to be defined.
Figure 2.9.3 is effectively a circuit model of the conducted emission test carried out by
engineers at an EMC Test House. It allows the ratio of common-mode current to injection
voltage to be simulated over a range of frequencies. Figure 2.9.4 simulates the setup used in
a conducted susceptibility test. It defines the ratio of unwanted signal current to a source
voltage in the common-mode loop.
This reasoning leads to a definition of transfer admittance:
Transfer admittance is the ratio of the current in the victim loop to the source voltage
in the culprit loop when there are no other voltage sources.
Most significantly, transfer admittance has the same value for conducted emission as for
conducted susceptibility. This means that a conducted susceptibility test on a cable assembly
can be used to predict the response of a conducted emission test, and vice versa. It also
2.10 Co-axial coupling 55
identifies a useful check which can be made on the integrity of any program used to analyze
any circuit model:
Interchange the locations of source voltage and monitored current and re-run the
program. If the results from both runs are not identical, then there is an error
somewhere.
In practice, there are many factors which tend to compromise this analysis of transfer
admittance. Most notable is the fact that that the test methods and test equipment used with
the two types of test are different. Even so, it is fair to say that a cable configuration which
exhibits high susceptibility at a particular frequency will also create a high level of emission
at that frequency.
and
r1;3 r3;1 r2;3 r3;2 2:10:2
r3,2
r2,1
r3,1
r1,1 r3,3
r2,2
The circuit model for a three-conductor cable is shown in Figure 2.7.6 and the formula for
inductance values is given by (2.7.12). When the relationships of (2.10.1) and (2.10.2) are
incorporated, the values of the inductors become:
mo mr l r2;2
Lc1 ln
2p r1;1
Lc2 0 2:10:3
m m l r3;2 r2;3
Lc3 o r ln
2p r3;3 r2;2
The value of Lc1 is the familiar formula for the loop inductance of co-axial cable. The value
of Lc2 turns out to be zero, while the value of Lc3 is the loop inductance of conductors 2 and
3 acting as a conductor pair. Since Lc2 is zero, this inductance disappears from the circuit
model.
Values for the capacitors of the circuit model are related to the inductors by (2.3.3). This
means that the theoretical value of Cc2 is infinite. In effect, it becomes a short-circuit. When
these modifications are carried out on the model of Figure 2.7.6, it changes to the circuit
model of co-axial coupling shown by Figure 2.10.2.
The circuit model of Figure 2.10.2 is that of an ideal co-axial cable, where the screen is
constructed of solid conducting material. For this ideal case, the only component at the
interface between common-mode loop and the differential-mode loop is the resistance of the
screen. This means that the use of co-axial cable for the signal link will ensure that the value
of the transfer admittance is extremely small. Put another way, the common-mode rejection
will be extremely high.
However, it is usually impractical to use such a cable. In the vast majority of assemblies,
the outer shield of the screened cables is made from thin wires inter-wound to form a flexible
braid. Gaps in the braid allow penetration of external field and emission of internal field. At
high frequencies, the shielding effectiveness of the screen deteriorates.
A more realistic model for a braided co-axial cable is shown in Figure 2.10.3. The
magnetic field coupling between inner and outer loops is simulated by Lc2, while Cc2
represents the effect of the electric field coupling. Effectively, the model reverts to that
shown in Figure 2.7.6. There is a significant difference. The values of the reactive
solid inner
Lc1 Rc1 conductor Lc1 Rc1
near end 2 2 2 2 far end
Cc1
cylindrical screen
Rc2 Rc2
2 2
Cc3
Lc3 Rc3 structure Lc3 Rc3
2 2 2 2
Cc2
Cc3
parameters associated with the screen conductor are much less than those of the separate
return conductor derived in section 2.7.
A rule-of-thumb estimate for the value of these parameters would be to assume:
Lc1 Cc1
Lc2 and Cc2
10 10
The best way of establishing more accurate values is to carry out tests on a representative
assembly. Such tests usually result in a parameter described as the transfer impedance, and
are defined in terms of a frequency response relating impedance to frequency. The rela-
tionship between transfer admittance and transfer impedance indicates that Z2 in Figure 2.9.3
can be defined as the transfer impedance of the assembly-under-review.
Some manufacturers provide data on the transfer impedance in the form of a frequency
response curve for a 1 m length. It should not be too difficult to create a model which
replicates this curve and to derive values for the resistance, inductance, and capacitance.
In any event, it can be said that co-axial cable will provide much better EMC than the
twin-conductor cable, since the transfer impedance is much less. It may be that a shielded
twisted pair could provide even better immunity to interference. However, it is more difficult
to derive parameter values for such a configuration. This topic is dealt with in detail in the
next chapter.
conductor 3
conductor 1
s conducting
surface;
r r conductor 2
r1,1 r4,4
r1,4 non-conducting
surface
r1,3
2*h
r1,2 r3,4
r2,4
r2,3
r2,2 r3,3
I1
I2
2V1
If these four conductors are connected as two separate circuit loops, then the picture
changes to that shown by Figure 2.11.3. Here, the coupling is between two isolated loops. As
far as magnetic coupling is concerned, this is essentially a transformer; a transformer where
the primary and secondary have one turn each and there is no magnetic material.
The circuit model for this configuration is the familiar transformer model of
Figure 2.11.4. Here, Lc1 represents the leakage inductance of the primary, Lc2 represents the
mutual inductance, and Lc3 represents the leakage inductance of the secondary.
2.11 The ground plane 59
Lc1
2V1 I1
mutual inductance
Lc2
I2
Following the procedure outlined in section 2.7, the three inductors can be defined for a
length l of the assembly:
mo mr l r1;2 r2;1 r1;4 r2;3
Lc1 ln
2p r1;1 r2;2 r1;3 r2;4
mo mr l r1;3 r2;4
Lc2 ln 2:11:1
2p r1;4 r2;3
mo mr l r3;2 r4;1 r3;4 r4;3
Lc3 ln
2p r3;1 r4;2 r3;3 r4;4
The next step is to use the transformer coupling model to derive a model which simulates the
coupling between the conductors of Figure 2.11.1. The relationships between the radial
parameters of Figure 2.11.2 and the dimensions of Figure 2.11.1 are:
The voltage source is defined as 2 V1 in Figure 2.11.3 because the voltage between any
conductor and its image is twice that which would exist between conductor and ground
plane. Since the formulae of (2.11.1) have been derived for a voltage source of 2 V1 , the
inductor values are twice that associated with conductors over a plane.
Dividing each of the inductive parameters of (2.11.1) by two and invoking the relation-
ships of (2.11.2) leads to:
Lc1 mo mr l 2hs
Ld1 ln p
2 2p r s2 4 h2
p
Lc2 mo mr l s2 4 h2
Ld2 ln 2:11:3
2 2p s
Lc3 mo mr l 2hs
Ld3 ln p
2 2p r s2 4 h2
60 CHAPTER 2 Lumped parameter models
Duality between inductance and capacitance can be used to calculate values for the asso-
ciated capacitors:
m0 mr e0 er l2
Cdi 2:11:4
Ldi
If the spatial dimensions of the setup are known, then a one to one correlation can be
established between the conductors of Figure 2.11.1 and the components of the triple-T
circuit model of Figure 2.7.6. Inductance Ld1 and capacitance Cd1 can be assigned to con-
ductor 1, while inductance Ld3 and capacitance Cd3 can be assigned to conductor 3.
The most significant feature of (2.11.3) is that a value can be assigned to Ld2. This is the
inductance of the ground plane. Similarly, the capacitance of the ground plane is Cd2.
The existence of Lc2 means that any transient current in the plane will create an end-to-
end voltage along the surface. If a voltage exists along the surface, then that surface cannot
possibly be equipotential. To assume that the conducting plane is an equipotential surface is
to ignore the lessons learnt from electromagnetic theory.
The relationship between separation distance and inductance values can be determined by
inspection of (2.11.3). As the separation between conductors 1 and 2 reduces, the values of
Ld1 and Ld3 reduce while the value of Ld2 increases. When the spacing is at a minimum, the
assembly behaves like a transformer. Conversely, as the spacing increases, the coupling
between the two loops decreases, and the value of Ld2 reduces.
The reasoning applied to inductor values can also be applied to capacitor values. As the
spacing between the conductors increases, capacitive coupling reduces.
This means that, if the ground plane is acting as a return conductor for signals in con-
ductors 1 and 3, (as with a printed circuit board) then interference coupling can be reduced
by separating conductors 1 and 3 as much as possible and by reducing the separation
between these conductors and the ground plane. The reasoning applies to both inductive
effects and capacitive effects. The action of the ground plane in printed circuit boards is
analyzed further in sections 8.2 and 9.3.
If the ground plane is used to represent the properties of the structure when it is acting as
a shield, then it is desirable for conductors 1 and 3 to be as close together as possible. This
will increase the coupling between the send and return conductors and enhance the balance
between the currents in those conductors, while reducing the amplitude of common-mode
current in the ground conductor.
CHAPTER 3
61
62 CHAPTER 3 Other cross sections
Section 3.3 develops the technique one stage further; to enable a circuit model to be
created for a three-conductor assembly. The example chosen is that of a screened pair cable.
The program defines the geometry of the elemental conductors of the three composite
conductors, creates an array of primitive inductors, and converts this to an array of loop
inductors. If it is assumed that a sinusoidal voltage of one volt is applied between the screen
and the inner pair, the loop currents in the elemental conductors can be calculated.
Then the primitive current in every elemental conductor is calculated. That is, current
flow from left to right is deemed to be positive current. Negative current flows from right to
left. Summing the currents in the appropriate elements allows the current in each composite
to be calculated.
Knowledge of currents and impedance values allows a three-by-three matrix of partial
voltages to be determined. From this it is possible to calculate the values for a three-by-three
array of partial inductors. Then the loop inductance values for the composite conductors are
calculated. This is a two-by-two matrix.
Using this information, the value of each of the three inductors of the circuit model can
be calculated. Invoking the duality between inductance and capacitance gives the value of
each of the capacitors. This allows a circuit model to be defined for a 10 meter length of
twin-core, screened cable.
The Mathcad worksheet which carries out all the necessary computations is reproduced.
This can be hand copied into a new worksheet by the reader. Alternatively, the worksheet file
can be downloaded from the website at www.designemc.info. MATLAB files which
replicate the calculations in every worksheet are also available for download. Modifying this
program, virtually any cross section can be modeled. The only changes would be a re-
definition of the geometry of the cross section and of the length of cable.
Intermediate results from the calculation are used to create a bubble chart which defines
the current distribution in the cable. This is more meaningful to a circuit designer than a
diagram of the magnetic field distribution. It also illustrates the fact that skin effect is not
uniform. Differential-mode current concentrates on adjacent surfaces.
Figure 3.3.5 illustrates the fact that common-mode current tends to flow on surfaces of
the signal-carrying conductors that are as far apart as possible. This is because the common-
mode current in the wire pair is flowing in the same direction. Since this current returns via
the outer shield, it concentrates on that part of the shield which is closest to the current
concentration in the wire pair.
Current in any conductor causes a magnetic field which has both internal and external
linkages with that conductor, and the formula derived for primitive inductance takes account
of both. By representing the surface of the conductor as an array of elemental conductors,
this treatment enables the effect of internal linkages to be included, even though they are
concentrated on that surface.
Rad
s
q
radius = r
Rad
Figure 3.1.1 illustrates a section of the conductor to be analyzed, while Figure 3.1.2
shows the method of simulation. Each elemental conductor represents a small segment, s, of
the surface. Hence:
2 p r s q Rad
Defining the primitive equations for elemental conductors in terms of vector algebra:
Vp Zp Ip 3:1:5
Ip lsolveZp, Vp 3:1:6
The total current in the composite conductor is the sum of the currents in the elemental
conductors. That is:
Xn
Iq Ipi 3:1:7
i1
The parameter Lq can be correlated with the primitive inductance Lp defined by (2.2.9) in
that it defines the inductance of a composite conductor when that conductor is acting as an
antenna. To avoid confusion between the two terms, it is useful to describe Lq as a partial
inductance.
It follows that Vq, Iq, and Zq can be described as partial voltage, partial current, and
partial impedance.
The set of equations developed above is amenable to the creation of a computer program
to calculate the partial inductance of virtually any cross section. Two stages are involved
defining the co-ordinates of the elemental conductors,
calculating the values of the electrical parameters.
Figure 3.1.3 is a copy of a Mathcad worksheet which carries out the computations for the
first stage. In this case, the process is quite simple. For a more complex cross section, the
2p
qi := n (i 0.5) angle at which each element is located, radian
10
plot of co-ordinates
y 103 0
10
10 0 10
x 103
definition of the co-ordinates would probably take the form of a three-column table of
values.
Figure 3.1.4 invokes equations (3.1.4) to (3.1.9) to determine the value of Lq, the partial
inductance of the composite conductor. It then carries out a check to confirm the resultant
value is the same as the primitive inductance of a conductor with the same radius as the
composite. This check provides a level of confidence that the process is valid.
Finally, the value of the partial capacitance is calculated, using the relationship between
inductance and capacitance defined by (2.3.3).
mo := 4 p 107 H/m mr := 1
mo mr l
K := = 2 106 H
2 p K is a constant for this computation
l
Lpi, j K ln see (2.3.2)
rad
Lp
n
Iq :=
i=1
Ipi = 7.238 104 sum of currents in elements
Vp1 w
w := 1 radian/s Lq := Lq = 1.382 105 H
Iq
l
Check:- Lp = 2 107 l ln Lp = 1.382 105 H
Rad
eo er mo mr l 2
Cq := Cq = 8.053 1011 F
Lq
Inspection of (2.7.7) reveals a clear correlation between primitive impedances and loop
impedances. The general formula for loop impedance can be defined as:
Zli; j Zpi; j Zpi; j1 Zpi1; j Zpi1; j1 3:2:3
Using this relationship, it is possible to create a matrix of loop impedances Zl. The loop
equations can then be set up. Again in vector notation:
Vl Zl Il 3:2:4
It is clear from Figure 3.2.2 that the number of elements in the Vl and Il vectors is N 1.
Voltages in every loop can be defined:
Vli 1 if i n1
3:2:5
Vli 0 if i 6 n1
Since all the voltages and impedances have been defined for the loop equations, it is possible
to determine the loop currents. The relevant function in Mathcad is:
Il lsolveZl; Vl 3:2:6
Ip1
Ip2 Il1
Ipn12
Ipn11 Iln12
Ipn1 Iln11
Vln1 Iln1
Ipn1+1
Ipn1+2 Iln1+1
Ipn1+n21
Ipn1+n2 Iln1+n21
Once the loop currents have been determined, it is a simple matter to calculate the values of
the primitive currents. The relationship can be derived from Figure 3.2.2.
Ip1 Il1
Ipi Ili for i 2 to N 1 3:2:7
IpN IlN 1
It now becomes possible to calculate every voltage component in every elemental conductor.
vi;j Zpi;j Ipj 3:2:8
If the variables are set out in tabular form, the result would be an array of N by N values.
Since there are only two composite conductors, the objective is to reduce this to an array of
2 by 2 values. This can be done by dividing the array into four sections, each representing the
contribution made by the current in a composite conductor. Figure 3.2.3 provides a picture of
the resultant sub-matrices. Since there are n1 elemental conductors in composite 1, the
average value of the contributions of current in these conductors to the voltage experienced
by composite 1 is:
1 Xn1 Xn1
vq1;1 vi;j 3:2:9
n1 i1 j1
Similar reasoning applies to the other three sections of Figure 3.2.3, giving:
1 X X
n1 n1n2
vq1;2 vi;j 3:2:10
n1 i1 jn11
X X
1 n1n2 n1
vq2;1 vi;j 3:2:11
n2 in11 j1
3.2 The composite pair 69
vq1,1 vq1,2
vq2,1 vq2,2
X n1n2
1 n1n2 X
vq2;2 vi;j 3:2:12
n2 in11 jn11
Calculating the average value of the current in each conductor is much simpler:
X
n1
Iq1 Ipi 3:2:13
i1
X
n1n2
Iq2 Ipi 3:2:14
in11
The four voltage components can be set out in the form of the small array of Figure 3.2.4,
where:
vq1,1 is the voltage in composite 1 due to Iq1
vq1,2 is the voltage in composite 1 due to Iq2
vq2,1 is the voltage in composite 2 due to Iq1
vq2,2 is the voltage in composite 2 due to Iq2
70 CHAPTER 3 Other cross sections
The integers h and k are used to identify the two composite conductors.
The number of equations has now been reduced from N to 2:
Comparing this equation with (2.4.1) or (2.4.7) indicates that the parameters Vq, Iq, and Zq
can be treated as primitives. As indicated in section 3.1, a clear distinction between the
properties of elemental conductors and those of composite conductors can be achieved by
invoking the term partial to describe the parameters involved. Hence (3.2.16) can be
described as the partial equations for the composite pair.
If the impedances Zqi,j are assumed to be due to inductive effects, then the relationship of
(2.5.4) can be used to determine the circuit inductance associated with each conductor.
Hence:
In summary, the process above has established a relationship between the primitive para-
meters of the elemental conductors and the inductances of a circuit model of the composite
pair. Since the primitive parameters of the elemental conductors can be related to the para-
meters of radius and length, it becomes possible to derive the inductive components from
knowledge of the structure of the assembly. Circuit capacitances can be derived by invoking
the duality between inductance and capacitance described in section 2.3.
A three-page worksheet derived from this set of equations is illustrated by Figures 3.2.5
to 3.2.7. It is assumed that the diameter of the conductors is 2 mm, that the centers are spaced
4 mm apart, and that the length of the cable is 1 m.
Page 1 of the worksheet derives the co-ordinates of the elemental conductors, Figure 3.2.5.
Page 2 calculates the value of the current in each elemental conductor, Figure 3.2.6.
Page 3 processes this data to derive component values for the circuit model, Figure 3.2.7.
It is useful to print out the intermediate results in the computation. If there are any errors
in the program, then the results would become implausible. For example, the sum of the
currents in the Iq vector should always be zero.
The Vq vector indicates that the input voltage is split equally between the two con-
ductors. This is intuitively correct since the two conductors are defined to be identical. Vq2 is
negative because Iq2 is negative. Figure 3.2.8 is a circuit model which effectively sum-
marizes the results of the computations.
It is possible to check this result by comparing it with that derived from a textbook
on electromagnetic theory [3.1]. Figure 3.2.9 is a copy of the final three steps in the
3.2 The composite pair 71
1 103
y 0
1 103
0 2 103 4 103
x
Mathcad worksheet. The fact that the results agree to three decimal places provides a high
degree of confidence in the method. The fact that the value of the loop capacitance is the last
parameter to be computed indicates that all the preceding results are also correct.
However, it is worth bearing in mind that the relative permittivity and the relative per-
meability are both assumed to be unity. It is best to confirm the component values using
electrical tests. Alternatively, estimated values for these parameters can be defined at the
start of the program.
A noteworthy feature of this configuration is the distribution of current in the conductors,
illustrated by the bubble plot of Figure 3.2.10. The diameter of each circle is proportional to
the amplitude of the current in that conductor. Current in the left-hand conductor is assumed
72 CHAPTER 3 Other cross sections
rad h2 + v2
rad ri if rad = 0
l
Lpi, j K ln
rad
Lp
to be flowing down into the page, while current in the right-hand conductor is assumed to
flow up out of the page.
This illustration shows that the current is concentrated in the two facing surfaces. This
asymmetrical distribution means that conductor resistance will probably increase more
rapidly with frequency than the increase predicted in section 2.5. However, tests indicate that
this aspect of the response can be catered for by assuming a slightly higher value for the
steady-state resistance than that given by (2.5.11) and by retaining the use of (2.5.15).
3.2 The composite pair 73
mo mr eo er l 2 4.224 1011
Cc := see (2.3.3) Cc = F
Lc
4.224 1011
132 nH 132 nH
132 nH 132 nH
42 pF 42 pF
near end far end
Figure 3.2.10 Bubble plot for composite pair, showing distribution of current.
If it were assumed that the two conductors were isolated at the far end, then the plot
would show the distribution of charge on the conductors, where the left-hand conductor is
positively charged and the right-hand conductor is assumed to hold negative charge.
l := 10
X1 := 0 X2 := 0.6 103 X3 := 0.6 103
Y1 := 0 Y2 := 0 Y3 := 0
N := n1 + n2 + n3
2 p
i := 1 ..n1 qi := n (i 0.5)
1
R1
xi := X1 + R1cos(qi) yi := Y1 + R1sin(qi) ri :=
n1
2p
i := n1 + 1.. n1 + n2 qi := n (i n1 0.5)
2
R2
xi := X2 + R2 cos(qi) yi := Y2 + R2 sin(qi) ri :=
n2
2 p
i := n1 + n2 + 1 ..N qi :=
n2 (i n1 n2 0.5)
R3
xi := X3 + R3 cos(qi) yi := Y3 + R3 sin(qi) ri :=
n3
1 103
y 0
1 103
2 103
2 103 1 103 0 1 103
x
1 n1
pointers to nine sub-matrices.
Start := n1 + 1 End := n1 + n2
n1 + n2 + 1 N
3.597 105
Figure 3.3.2 Computing values for voltages and currents in composite conductors.
Calculating the current in each elemental conductor is carried out by the second page
of the worksheet. Since this page is identical to that of Figure 3.2.6, there is no need to
replicate it.
Once the currents in all the elemental conductors have been defined, the next stage is to
compute the values of the currents and voltages in the composite conductors, and this set of
subroutines is illustrated in Figure 3.3.2. This process is very similar to that of Figure 3.2.7,
3.3 The screened pair 77
er := 1
6.386 109
mo mr eo er l 2
Cc := Cc = 4.575 1010 F
Lc
4.575 1010
the main difference being that the number of composites has increased from two to three.
Two sets of intermediate results are worthy of note; the vectors for partial voltage Vq and
partial current Iq.
The value of Vq1 illustrates the fact that the sum of the voltages induced in the screen
is effectively zero. This is essentially the same as a statement that all internally generated
electromagnetic fields are contained within the confines of the screen.
78 CHAPTER 3 Other cross sections
458 pF 458 pF
87 nH 6.39 nF 87 nH
screen
y 0 + +
2
1.5 1.0 0.5 0.0 0.5 1.0 1.5
x
Figure 3.3.5 Bubble plot for screened pair, showing distribution of current.
The net effect of currents in the three conductors is to balance out the voltages induced in
the screen. As far as external circuits are concerned, the surface of the screen is at zero voltage.
The screen behaves in exactly the same way as the outer conductor of a co-axial cable.
The voltages on conductors 2 and 3 are both 1 V in amplitude, since these conductors are
shorted together and the voltage source is set at 1 V. The sign is negative because the
currents Iq1 and Iq2 are both negative.
An important check to make is to add all the partial currents in the composite conductors
together. If the sum of the currents is not zero, then there can be no confidence in any
subsequent calculations. It is to be expected that the returning current is equally split
between composites 2 and 3, so this is also a useful check.
Figure 3.3.3 illustrates the final set of calculations in the worksheet. The partial induc-
tance values are derived from current and voltage data, and these values are used to create
3.3 The screened pair 79
the matrix of loop inductors. For this particular cross section, the loop matrix L_loop is
symmetrical, so there is no problem in proceeding to the final two stages; the determination
of circuit inductance and circuit capacitance values. Since half-values of inductors are used
in the circuit model, then these are the values printed out. This leads to the model of
Figure 3.3.4.
The only parameters left to define are the resistors, and this can be done by using general
purpose test equipment to measure conductor resistance, or by invoking (2.5.11).
If the cross section of the cable assembly had been asymmetrical, the loop matrix would
also have been asymmetrical, leading to a problem in determining a circuit model. Due to the
nature of circuit theory, the impedance matrices derived from mesh equations are always
symmetrical. It would not have been possible to achieve a one-to-one correlation between an
asymmetrical matrix of loop impedances and a symmetrical matrix of circuit impedances.
This does not preclude the creation of circuit models, but it does mean that the models for
susceptibility and emission would be different.
A useful check on the plausibility of the computations is to produce a bubble plot of the
distribution of current in the assembly. This is essentially the same as the cross section
illustrated in Figure 3.3.1, but with the radius at each location proportional to the amplitude
of the current rather than the actual radius.
Figure 3.3.5 illustrates the resultant plot. It is assumed that the current in the outer screen
is flowing out of the page, while the current flow in the inner conductors is into the page.
Since the signal and return conductors are shorted together at each end, the picture that
emerges is one showing the distribution of common-mode current. In contrast to
Figure 3.2.10, the current is concentrated on the outer surfaces of the wire pair.
CHAPTER 4
It is often required to extend the model to frequencies well beyond the one-tenth wavelength
limit of the lumped component model. Such a task can be carried out by invoking the
concepts of transmission line theory. This involves the need for circuit components to be
defined in terms of W/m, F/m, H/m, and S/m.
Using this approach, transmission line theory derives a pair of hybrid equations relating
current and voltage at the sending end of the line to the current and voltage at the far end.
Although at least one book on electromagnetic theory derives these equations [4.1], some
books do not. So a derivation is provided by Appendix C.
In section 4.1, it is postulated that a T-network circuit model can be used to replicate the
relationships of the hybrid equations. A pair of loop equations is derived from this model. It
is assumed that a one-to-one relationship exists between the currents and voltages of the two
sets of equations. This allows formulae to be developed which define the impedances of the
T-network model. These impedances can be described as distributed parameters since they
are based on the assumption that R, L, and C are distributed along the length of the cable.
Since this model represents a defined length of cable, the impedances of the branches can
be defined in terms of the lumped parameters; that is, in terms of resistors, capacitors,
inductors, and frequency, as used in textbooks on circuit theory. A one-to-one correlation is
thus established between the components of the distributed parameter model and the lumped
parameter model. The distributed parameter model is defined by Figure 4.1.2 and the rela-
tionships are defined by (4.2.1)(4.2.4). The key feature of this set of relationships is that it
eliminates the need for the circuit designer to use per unit length parameters.
There is still a fundamental limitation. The formulation assumes that action and reaction
are instantaneous between adjacent conductors. An electromagnetic wave takes a finite time
to cross the gap between these conductors. Section 4.1 provides a way of estimating the
maximum usable frequency of the model.
Section 4.2 develops this model to represent a three-conductor line, by representing each
conductor as a T-network. Although the impedance values of each branch are not simple
functions of inductance, capacitance, and resistance, the distributed parameter model is still
amenable to analysis using circuit theory. (This means that it should be possible to develop
SPICE software to implement this approach.)
In section 4.3, a Mathcad program is used to simulate the coupling between two con-
ductors over a ground plane. This is done for three situations; where the terminations at the
far end are open-circuit, short-circuit, and loaded by the characteristic impedance. The
81
82 CHAPTER 4 Transmission line models
resultant graphs are extremely informative in comparing the frequency response character-
istics of the grounded configuration and the floating configuration.
Section 4.4 shows how the circuit model can be used in conjunction with conducted
susceptibility tests.
Cc1 Cc2
Cc 4:1:3
Cc1 Cc2
where Lc1 and Lc2 are defined by (2.5.5) and where Cc1 and Cc2 are as defined by (2.5.8).
This simplification to the model does not mean that the reference terminals at the near
and far ends are at the same voltage. It merely reduces the number of parameters that the
mathematics has to handle.
A two-conductor transmission line can be simulated by connecting a large number of
such models in series. However, there is a better way. The model can be constructed using
distributed parameters, as shown in Figure 4.1.2.
Here, Vs and Is are the voltage and current at the sending end, while Vr and Ir are the
voltage and current at the receiving end.
Appendix C shows how a pair of hybrid equations relating these four parameters can be
derived from fundamental concepts:
Vr
Is sinhg l Ir coshg l 4:1:5
Zo
signal signal
Lc Rc Lc Rc
2 2 2 2
Cc
reference reference
Vs Z2 Vr
Is Ir
The circuitry at each interface can be anything from a short-circuit to an open-circuit; it can
be resistive, capacitive, inductive, or any combination of all three. It can also be another
transmission line. Appendix C is well worth reading, since it recognises the fact that the
equations relate to conducting loops. It does not assume the existence of a zero-volt surface.
The propagation constant g is a function of distributed parameters; resistance per meter,
inductance per meter, conductance per meter, and capacitance per meter:
s
Rc j w Lc Gc j w Cc
g
l l l l
s
Rc j w Lc l
Zo
l Gc j w Cc
The parameter Gc is assumed to represent the conductance of the insulation. That is, the
inverse of the insulation resistance. Since most cables use high quality insulation, it can be
assumed that Gc 0. However, this can only be an initial assumption.
Circuit theory gives the relationships between currents and voltages of Figure 4.1.2 as:
Vs Z1 Z2 Is Z2 Ir 4:1:8
Vr Z2 Is Z1 Z2 Ir 4:1:9
84 CHAPTER 4 Transmission line models
Simplifying further:
Z1 Z1
Vs 1 Vr Z1 2 Ir 4:1:11
Z2 Z2
Equations (4.1.11) and (4.1.10) form a pair of equations which can be correlated with (4.1.4)
and (4.1.5).
Comparing (4.1.4) with (4.1.11) gives:
Z1
1 coshg l 4:1:12
Z2
and
Z1
Z1 2 Zo sinhg l 4:1:13
Z2
Z1
1 coshg l
Z2
To make subsequent equations less unwieldy, let:
qgl 4:1:15
The parameter q can be described as a phase variable since it effectively determines the
phase shift between input and output signals of a transmission line and since it varies with
frequency. From (4.1.12) and (4.1.15):
Z1
cosh q 1
Z2
4.1 Single-T model 85
Z1
Substituting for in (4.1.13) gives:
Z2
Z1 1 cosh q Zo sinh q
Rearranging (4.1.14):
Z2 Zo cosech q 4:1:17
Equations (4.1.16) and (4.1.17) relate the Z-parameters of Figure 4.1.2 to the characteristic
impedance and the phase variable of the transmission line.
Finally, from (4.1.6) and (4.1.15):
p
q Rc j w Lc Gc j w Cc 4:1:18
Using (4.1.7) and (4.1.18), Zo and q can be calculated. These can then be used in (4.1.16) and
(4.1.17) to obtain values for the components of Figure 4.1.2. This means that the lumped
parameter model of Figure 4.1.1 can easily be transformed into one which used distributed
parameters. The transformation would be:
1 q
Rc j w Lc ! Zo tanh
2 2
4:1:19
1
! Zo cosech q
Gc j w Cc
Invoking this transformation allows the circuit model of Figure 4.1.2 to simulate the
response of a transmission line over the same range of frequencies at which the hybrid
equations are valid.
The characteristic impedance Zo is independent of the length of the line. Given knowl-
edge of the values of Rc, Lc, Gc, and Cc, the phase variable q is also independent of the line
length. The variable l does not appear in (4.1.18). Although it is necessary to invoke the
concept of distributed parameters to derive the hybrid equations, it is not necessary to use
the concept when calculating impedance values for the model of Figure 4.1.2.
86 CHAPTER 4 Transmission line models
An electromagnetic wave takes a finite time to cross the gap between the conductors of
any cableform. Just as SPICE modelling is limited by the electrical length of the assembly,
modeling using distributed parameters is limited by the radial separation of the adjacent
conductors. It can be reasoned that, if rmax is the maximum separation between the centers
of two conductors in the cross section of the assembly-under-review, then the maximum
frequency at which the distributed parameter model is usable is fmax, where:
vmin
fmax 4:1:20
10 rmax
and vmin is the propagation velocity of the electromagnetic wave in the insulating medium
with the highest relative permittivity in the section-under-review.
That is, if the velocity of propagation is 200 m/ms and the maximum spacing between
conductors is 10 mm, then the maximum frequency at which this modelling is usable would
be about 2 GHz.
When distributed parameters are used, there is no theoretical limit to the length of the
cable which can be modeled. It is assumed that the cross section of the cable is constant and
that the lengths of the conductors are identical.
p
qi Rci j w Lci Gci j w Cci 4:2:2
qi
Z1;i Zoi tanh 4:2:3
2
These four equations can be collected together to compute the values of a pair of
Z-parameters from a set of R, L, C, and G values. Indices are used to retain correlation
between the parameters and the conductors.
Given the definitions above, the result of the computation would be an array of two rows by
three columns. The first row would hold values correlating with horizontal branches of Fig-
ure 4.2.2. The second row would hold values for the vertical branches. Each column would be
associated with a conductor. The conductance parameter Gi is included in the formulae to
allows losses in the insulation to be simulated. It is fair to make an initial assumption that all
conductance values are zero.
4.2 Triple-T model 87
Z1,2 Z1,2
Z1,3 Z1,3
One feature of the Triple-T model is that it only includes components representing the
cable assembly.
It is necessary for the components at the cable terminations to be included before any
simulation is possible. This results in the full model of Figure 4.2.3.
Impedances within the unit at the near end are identified as Zni, while those within the
equipment at the far end are designated Zfi.
Four separate loops can be identified. Currents in each loop have been defined, as well as
all possible voltage sources. It is assumed that conductors 1 and 2 carry the differential-mode
signal current. So voltage sources V1 and V3 are located within the equipment units. Voltage
sources V2 and V4 represent possible interference sources in the common-mode loop.
The four loop equations for the circuit model of Figure 4.2.3 are:
V1 I1 Zn I3 V3
2 Z1,2 Z1,2 Zf2
Z2,2
V2 V4
near end Z2,1 Z2,3 far end
I1 I2
I3 I4
Figure 4.2.3 Full circuit model, using lumped and distributed parameters.
The loop impedances are derived by inspection of Figure 4.2.3. For example, Z11 is the sum
of the impedances in the loop carrying the current I1, while Z34 is the sum of the impedances
which share currents I3 and I4. The negative sign indicates that these two currents flow in
opposite directions.
Equations (4.2.1)(4.2.6) provide a traceable relationship between conventional circuit
components and the array of loop impedances. If all the component values are defined, and
all the source voltages are defined, then vector algebra can be used to derive all the currents.
It then becomes a routine matter to calculate the voltage between any two points or the
current in any component, at any particular frequency.
Since there is now a one-to-one correlation between the branches of the lumped para-
meter and distributed parameter networks, it becomes possible to define the circuit model in
terms of the lumped parameters, but analyze it using the loop impedances of (4.2.6). This
leads to the general circuit model of Figure 4.2.4. It is entirely possible to replace the cir-
cuitry at the near and far ends with components representing the interface circuits of the
equipment-under-review.
In this model, the values for the components of the triple-T network can be derived from
geometrical data of the cable and structure, while values for the components of the interface
circuitry at the near end and far end of the cable are defined by the designer.
4.3 Cross-coupling 89
It is also possible to assign values to all the components of the model by carrying out
electrical tests on an actual assembly. Chapter 7 describes how this can be done.
4.3 Cross-coupling
Enough analytical tools now exist to allow an initial assessment to be made of a wide range
of intra-system interference problems. Perhaps the simplest illustration of the process lies in
the analysis of the cross-coupling between two conductors over a ground plane.
Figure 4.3.1 illustrates the configuration. It is assumed that the conductors are both 1 mm
in diameter, and that each is spaced 1 mm above the plane. The separation between them is
set at 4 mm and the length of the assembly is assumed to be 1 m.
Figure 4.3.2 sets out the calculations involved in determining the component values for
the circuit model. It is a copy of the first page of a four-page Mathcad worksheet. The first
line records the values of the physical constants involved in the derivation; the permeability,
permittivity, and the speed of light. It is assumed that the values of relative permittivity and
permeability are both unity. The other physical constant is r, the resistivity of copper.
The second line of the worksheet defines values for the spatial parameters of the con-
figuration. These correlate with the parameters of Figure 2.11.1.
Steady-state resistance values Rss for the two circular-section conductors are determined
using (2.5.11). A guess-value of 5 mW is assigned to the ground plane. Above a certain
frequency, skin effect will cause an increase in the resistance of the conductors. The cross-
over frequency Fx is determined using (2.5.14) and the relationship between Rc and f is
defined by (2.5.15).
Inductance values for the circuit model are derived from (2.11.3). Since the two wires
are equidistant from the plane, Lc3 is equal to Lc1. The three values are stored in the vector
Lc. Capacitor values are obtained quite simply, by invoking (2.3.8).
Another significant parameter is Fq, the frequency at which quarter-wave resonance
occurs. This is determined using (2.3.9). Knowledge of this frequency allows subsequent
analysis to include maximum values of the peaks in the response curves.
90 CHAPTER 4 Transmission line models
l=1m
s = 4 mm
r = 0.5 mm
h = 1 mm
Since half-values of resistance and inductance are used in the circuit model of
Figure 4.2.1, these are the values printed out in the worksheet. Adding the values of the three
capacitors provides sufficient information to draw the representative circuit model for the
conductor assembly.
Before the response can be analyzed, it is necessary to include values for the interface
components of Figure 4.2.4. Invoking worst-case conditions, it is assumed that the compo-
nents at the near end are short-circuits and that the wire terminals at the far end are open-
circuit. Following normal practice used in SPICE analysis, the open-circuits are represented
by 10 MW resistors. To allow for the option of simulating losses in the insulation, con-
ductance values Gc are also defined. In this model, all three are assumed to be zero.
Data derived from this first page of the worksheet can be illustrated by the circuit model
of Figure 4.3.3. It is assumed that the assembly of Figure 4.3.1 is configured as a pair of
transmission lines; with the ground plane acting as the return conductor for both signals. It is
also assumed that there is only one voltage source, between conductor 1 and the ground
plane. This model represents the situation where the culprit circuit uses conductor 1, while
the victim circuit uses conductor 3.
The purpose of the analysis is to simulate the current that would flow in the victim loop
over a range of frequencies. Since the far end is open circuit, only the current at the near end
will be of any significance. Hence the objective is to simulate the current I2 when voltage V1
is applied. If V1 is held at a constant value of one volt, the result will be the frequency
response of the transfer admittance.
Figure 4.3.4 sets out the two program functions derived in section 4.2. It constitutes page 2
of the Mathcad worksheet.
The function Zbranch( f ) supplies one input variable; the frequency f. It calculates the
values of the two distributed impedances associated with each conductor, and assembles
them in the output array Z. As indicated in section 4.2, this contains two rows of three
columns. A feature worth noting is the fact that the value of each resistor is calculated at each
frequency. This allows skin effect to be incorporated into the analysis.
In addition to supplying frequency as an input variable, the function Zloop( f , Zf ) also
includes the vector containing component values at the far interface. This allows any of these
components to be varied, subsequent to the definition of the function.
The first program step in the Zloop() function is to call up the Zbranch( f ) function and
place the results in the local variable Z. This allows component values of individual branches
to be accessed as required when determining values for the loop impedances. These impe-
dance values are then placed in the four-by-four array at the end of the subroutine.
4.3 Cross-coupling 91
4 r
Fx := = 6.89 104 Hz see (2.5.14)
mo p r 2
mo l 2 h s
Lc1 := ln H
2 p 2 2
r s + 4 h
mo l s2 + 4 h2
Lc2 := ln H see (2.11.3)
2 p s
Lc3 := Lc1
2
1 l
Cc := F see (2.3.8)
Lc c
1
Fq := = 7.495 107 Hz see (2.3.9)
4 Lc1 Cc1
0 107 0
Zn = 0 Zf = 0 Gc = 0 S
0 10 7 0
Creating a graph of the response involves the definition of two vectors; the frequency and
the output variable, and this is the task performed by page 3 of the worksheet. Figure 4.3.5
illustrates the process.
A set of spot frequencies is defined in the first line. The integer n is used to set the
number of spot frequencies in the range between zero and the frequency of quarter-wave
resonance, Fq. To cover the range up to a full wavelength, the control variable s is set to
make four times this number of steps. This allows each spot frequency, Fs, to be a multiple of
the quarter-wave frequency. Using this method of defining the spot frequencies ensures that
92 CHAPTER 4 Transmission line models
all the resonant frequencies are selected and that the amplitude of every peak and every
trough in the response is calculated.
In the model of Figure 4.3.3, the source voltage generator is located in loop 1. Setting the
amplitude of loop 1 at 1 V and the other three sources at zero voltage allows the voltage
vector V to be defined.
Since all the inputs have now been defined, it becomes possible to create the main pro-
gram, and this is illustrated on the third line of the worksheet of Figure 4.3.5.
The first program step is to select a spot frequency from the vector F and assign it to the
local variable f. The second line calls up the function Zloop( f, Zf ). Since Zf has already been
defined, (see Figure 4.3.2) both variables are visible to the function. The output is stored in
the local variable Z.
The next step is to calculate the values of all four loop currents and place them in the
vector I.
The final step in the main subroutine is to select I2, the current at the near end of the
victim loop, and determine its amplitude. Dividing this current by the input voltage V1 gives
the value of the transfer admittance. Since the value of V1 is unity, the values of current and
transfer admittance are identical. Finally, the output is stored in the relevant location in the
vector Yoc; the transfer admittance when the line terminations are open-circuit. The graph of
this function is shown on Figure 4.3.6.
Although normal convention is to display such a function using logarithmic scale for both
parameters, in this case a linear scale is used for frequency. This highlights the symmetrical
nature of the response. Symmetry is evident in the fact that the response between 150 MHz
and 300 MHz is almost the same as between zero and 150 MHz; almost the same, but not
identical. Skin effect causes the resistance to increase, reducing the amplitude of the second
peak and increasing the amplitude of the second trough.
By resetting all the components of the impedance vector Zf to zero and re-running the
main program, the vector Ysc can be created to define the response of the assembly when the
terminations are short-circuited. The response of the assembly when it is critically damped
can also be determined. The program steps are illustrated by Figure 4.3.7.
Figure 4.3.8 provides a further illustration of the symmetry of the response, by comparing the
open-circuit response with the response when the terminations at the far end are short-circuited.
Transfer admittance under short-circuit conditions, Ysc, is illustrated by the dotted curve.
4.3 Cross-coupling 93
Zbranch( f ) := 2 f
for i 1.. 3
f
Rci Rssi 1 + see (2.5.15)
Fx
Zo Rci + j Lci
see (4.2.2)
Gci + j Cci
Z1, i Zo tanh
2 see (4.2.3)
Z2, i Zo csch()
see (4.2.4)
Z
When each conductor is terminated by its characteristic impedance, the response is perfectly
flat over the entire frequency range. This is illustrated by the Ycrit curve; the dashed line.
The bottom half of the graph, below the Ycrit curve, is almost the mirror image of the top
half.
Many lessons can be derived from this set of curves, not least among them the fact that
the only configuration which is free from peaks in the transfer admittance is also the one
which provides the highest efficiency in terms of signal transmission.
94 CHAPTER 4 Transmission line models
A/V
0.1
0.01
Yoc 1 103
1 104
1 105
1 106
0 1 108 2 108 3 108
F Hz
In practice, the peaks will be much lower and much more rounded than those illustrated,
since this model does not take account of losses due to radiated emission. Chapter 5 develops
the use of models to cater for this effect.
76.432
Lc
Zf := Zf = 6.69 Ycrits := f Fs
Cc
76.432 Z Zloop( f, Zf )
Figure 4.3.7 Calculating frequency response for short-circuited and critically damped lines.
A/V
0.1
0.01
Yoc
Ysc 1 103
Ycrit
1 104
1 105
1 106
0 1 108 2 108 3 108
F Hz
Figure 4.3.8 Transfer admittance for open-circuit, short-circuit, and critically damped lines.
modified in the light of more accurate information. It allows corrections to be made to the
design to improve its performance. This is true of requirements such as system function,
response time, reliability, size, and cost.
It can also apply to the requirements of electromagnetic compatibility. Since the
EMC Test Houses have already acquired a wide experience in meeting the regulatory
requirements, it is a logical exercise to tap into that experience to identify the relevant test
equipment and test methods.
96 CHAPTER 4 Transmission line models
Two problems immediately emerge. The equipment is extremely expensive, and the test
procedures require the services of highly skilled engineers. It would be far too costly for a
small organization to attempt to emulate this approach.
However, the product designer also has one advantage; a big advantage. It is possible to
design and use interface equipment which allows direct visibility of the signals being pro-
cessed. This is not an option available to the Test Houses, since any special modification of
the equipment-under-test would render the test results invalid.
Another advantage is the fact that General Purpose Test Equipment is already in hand to
carry out functional tests of the assembly-under-review. This could also be used for EMC
testing. Also, the frequency range of the tests can be tailored to the range over which the
equipment-under-test normally function. There is no need to cover the range covered by the
Test Houses.
The simplest set of bench tests would be to analyze the performance of the inter-
connecting cables; specifically, to measure the frequency response of the transfer admit-
tance. The results could be compared directly with those obtained by circuit modeling. But
first, the representative circuit model needs to be created.
Figure 4.4.1 illustrates a setup which can be used to measure the conducted emission of
any wiring configuration. In this case, it is assumed that the assembly-under-review is a 10 m
length of a conductor pair routed 5 mm above a ground plane.
An input signal can be applied to the near-end terminals via a splitter box. This allows the
input voltage to be monitored on channel 1 of the oscilloscope. Channel 2 is used to monitor
the common-mode current. The 100 W load at the far end of the cable is floating.
The ratio of output to input gives the value of the transfer admittance at the frequency of
the signal generator output. Repeating the measurements over a range of spot frequencies
enables the frequency response characteristic of the conducted emission to be determined.
If the physical geometry of the signal link is as shown by Figure 4.4.2, then the circuit
model can be derived by invoking the calculations recorded on the worksheet of Figure 4.3.2.
As with SPICE modeling, it is necessary to simulate the open-circuit as a high value resis-
tance; in this case 10 MW.
signal oscilloscope
generator channel 1 channel 2
50 50
splitter
box
twin core cable
current
l = 10 m
transformer
far end floating
18
18 18 100
I
Iout
ground plane
2 mm
conductor 1
0.5 mm
conductor 2
5 mm
ground plane
Vin 100
1367 nH 0.108 1367 nH 0.108
407 pF 407 pF
10 M
Iout 342 pF
1629 nH 1629 nH 0.0025
0.0025
The resultant model is shown in Figure 4.4.3. It is worth noting that the source voltage is
in the differential-mode loop while the output current is in the common-mode loop.
Conducted susceptibility measurements would involve a setup similar to that illustrated
by Figure 4.4.4. Here, the input voltage is applied to the common-mode loop via an injection
transformer. Since any transformer has an output impedance, the applied voltage can reduce
as the load current increases. This effect is catered for by using a separate turn on the
transformer to monitor the actual voltage applied. One channel of the oscilloscope provides
the means of measurement. This enables the amplitude of the input voltage to be determined.
The second channel of the oscilloscope allows the resultant current in the differential-
mode loop to be measured. Again, the ratio of output to input provides a value for the
transfer admittance. Measurement over a range of spot frequencies provides the frequency
response of the conducted susceptibility characteristic.
A prediction of the response of this configuration can be achieved with the model of
Figure 4.4.5. Since the assembly-under-review is exactly the same as that used for the con-
ducted emission test, the passive components of the two models are identical. The only
difference lies in the fact that the voltage source is now in the common-mode loop, while the
output current is now in the differential-mode loop.
An analysis of the frequency response of either model can be carried out by utilizing the
program set out by Figures 4.3.4 and 4.3.5. It is only necessary to redefine the component
values, the input voltage, the output current, and the frequency range.
98 CHAPTER 4 Transmission line models
oscilloscope signal
channel 2 channel 1 generator
twin core cable
l = 10 m
50 50 far end floating
50 voltage injection
transformer
18
18 18 100
V
ground plane
100
Iout 1367 nH 0.108 1367 nH 0.108
407 pF 407 pF
10 M
342 pF
1629 nH 1629 nH 0.0025
0.0025
Analyzing the response of Figure 4.4.3 gives the characteristic shown on Figure 4.4.6.
Analyzing the response of Figure 4.4.5 gives exactly the same characteristic. This illustrates
the fact that the transfer admittance for conducted susceptibility is exactly the same as for
conducted emission. With the circuit models, the responses are identical. This confirms the
conclusion of section 2.9.
Actual tests would show some differences, due to the effect of the test equipment and due
to radiated emission.
The response curve also illustrates that although the floating configuration gives excel-
lent performance at low frequencies, it actually amplifies the level of interference at the
quarter-wave frequency.
If the configuration-under-review is changed, to add a link between the ground plane and
the return conductor at the far end, then the circuit model for the susceptibility test changes
to that shown on Figure 4.4.7.
The response of this setup is given by Figure 4.4.8. This illustrates the fact that, for the
grounded configuration, minimum interference occurs at the quarter-wave frequency, while
the peak level occurs at the half-wave frequency.
4.4 Bench test models 99
0.1
A/V
0.01
YToc
1 103
1 104 0
1 107 2 107 3 107
F Hz
100
Iout 1367 nH 0.108 1367 nH 0.108
407 pF 407 pF
342 pF
1629 nH 1629 nH 0.0025
0.0025
This response can be compared with that of Figure 4.4.6, where peak interference with
the floating configuration is experienced at quarter-wave frequency, while the minimum
level occurs at the half-wave frequency.
There is another significant difference in the responses. The difference between maxima
and minima of the response curve of the grounded configuration is much less. This is
because the 100-W load resistance is damping the amplitude of the reflections in the com-
mon-mode loop as well as in the differential mode loop. With the floating configuration,
there is no damping whatsoever in the common-mode loop.
It can be concluded that the performances of the grounded and floating configurations
tend to complement each other. At the frequency where one gives poor performance, the
100 CHAPTER 4 Transmission line models
0.01
A/V
YTsc
1 103
0 1 107 2 107 3 107
F Hz
performance of the other is excellent. It is also true that it never gives good performance at
all frequencies.
This section has identified the relationships between the bench test equipment, the con-
figuration-under-review and the general circuit model. It has developed a computer program
which can be used to analyze the response of that model over the range of frequencies at
which interference problem are most critical.
Practical examples of actual tests are provided by Chapter 7 and methods of dealing with
interference problems are described in Chapter 8.
CHAPTER 5
Antenna models
EMI extends well beyond the confines of the assembly-under-review. Analysis of radiated
emission and radiated susceptibility calls for a review of the properties of antennae. For-
tunately, this review allows many simplifying assumptions to be invoked.
Section 5.1 provides a brief overview of the textbook analysis of the half-wave dipole and
identifies a very significant parameter; when connected to the output of a radio transmitter
which is generating a signal at the half-wave frequency, the dipole exhibits the same char-
acteristics as a resistive load. This has been named the radiation resistance and its value has
been calculated to be 73 W.
Combining the radiation resistance with the primitive capacitance and primitive induc-
tance of each monopole allows a circuit model to be created; a model which simulates the
coupling between the dipole and the environment.
Developing the model to simulate antenna-mode coupling with a twin-conductor cable
involves the same process that was used to derive the three-conductor model in section 2.7.
This represents each conductor as a T-network and the environment as another T-network.
Since the environment must now be assigned values for inductance, capacitance, and resis-
tance, it is fair to say that it behaves like a virtual conductor. Section 5.2 derives formulae for
all the components of the circuit model.
Antenna-mode current is induced in any cable which is exposed to external radiation.
Simulating the effect of this radiation calls for the introduction of a voltage source in
series with the virtual conductor. Section 5.3 relates this threat voltage to the electric field
strength of the radiation.
If the power density of the external radiation is constant, but the frequency varies, then
the graph of the amplitude of the threat voltage will exhibit a series of peaks and troughs. As
far as the analysis of interference coupling is concerned, worst-case conditions can be
defined as the envelope curve which touches the peaks.
Section 5.4 provides an illustration of the differential-mode current which would be
induced in a twin-conductor cable when the frequency of the external radiation is varied over
a wide range.
If the structure is treated as a perfectly reflecting ground plane, then this will represent worst-
case conditions. Since a circuit model can be created for this configuration, then such a model can
be used to simulate the structure. Section 5.5 provides an example which assigns values to the
components of this model. The effect of the external field can be simulated by connecting a
voltage source and a 50-W resistor in series with the structure. Unlike a dipole receiving antenna
101
102 CHAPTER 5 Antenna models
which directs input power to a 75 W resistor in the receiver, the incoming energy is stored in the
assembly. Since more energy is delivered, the effective value of the radiation resistance decrea-
ses. The test described in section 7.5 measures this value to be 50 W. This value is a good round
figure; and can be modified if experience shows it to be too pessimistic.
For radiation susceptibility assessment, it is possible to relate the power output of the test
transmitter to the threat voltage applied to the equipment-under-test. Section 5.6 provides the
relevant formulae. Since mathematical software is not restricted to the analysis of circuit models,
it can be used to compute the response of the equipment to the electromagnetic environment.
It is also possible to relate the antenna-mode current created by the equipment-under-test
to the power received by a test receiver during a radiated emission test. Relevant formulae
are provided by section 5.7.
Since the approach used in this chapter is always to assume worst-case conditions, the
simulated results are likely to err on the side of caution. The most significant benefit derived
from the use of worst-case conditions is that the relevant formulae are dramatically simpler
than those necessary to derive field distribution patterns.
Current in a small element dz will create a magnetic potential A at the point P. This magnetic
potential is a vector, and its direction is parallel to the axis of the conductor. A relationship
can be established between A and the current in the element dz.
Spherical co-ordinates, centred at the mid-point of the antenna, are then used to analyze
the vectors involved in propagating the field. Figure 5.1.3 illustrates this. Since the vector A
is aligned in the same direction as the current, there is no longitudinal component. That is,
Af 0. Equations for the radial and latitudinal components Ar and Aq are derived. It is then
possible to define formulae for the magnetic field vector Hf . This is a longitudinal vector; in
Figure 5.1.1 it would be directed into the page. Then, using one of Maxwells equations, the
components of the E-field are calculated.
Since the only components of the total field which are capable of propagating power are
those which are at right angles to each other, the analysis focusses on the latitudinal electric
5.1 The half-wave dipole 103
A = magnetic potential
z = /4
P
R F Transmitter dz
Rt
Io*cos(.t)
z = /4
field Eq and the longitudinal magnetic field Hf . Multiplying the amplitudes of these com-
ponents together gives an expression for the power density vector Sr. Integrating the value of
this vector over the surface of the sphere gives a value for the total radiated power Pt:
1
Pt Rrad Io2 W 5:1:2
2
The parameter Rrad is named the radiation resistance. It is not a resistor in the conventional
sense of the word, but a mathematical constant which happens to have the dimensions of
resistance. Its value can be calculated, and for the half-wave dipole:
Rrad 73 W 5:1:3
This value for Rrad is derived by assuming that the medium is lossless; meaning that the
total radiated power does not decrease with distance from the antenna. Nor can it increase
with distance. This being so, (5.1.2), with a value of 73 W assigned to Rrad, defines the
maximum power which can be delivered to the environment by a dipole antenna.
It is useful to quote an extract from the Ordnance Board Pillar Proceeding at this point:
At distances closer to about two wavelengths from a single linear aerial, the radiation
field is accompanied by further electric and magnetic components the strengths of
which fall as the square or cube of the distance from the aerial. These components are
not related by Zo and do not radiate power away from the aerial. Although power can
be extracted from them by very close receiving aerials, this power is always less than
predicted by the extrapolation of far field theory. [5.2]
This means that (5.1.2) will provide a worst-case estimate of the power that would be
delivered to a monitor antenna, whether that antenna is located in the near-field or the far-
field. For the purposes of EMC analysis, it is desirable to have available a worst-case
estimate.
This leads to the circuit model of Figure 5.1.2, where the antenna acts as a resistive load
to the co-axial transmission line. It should be emphasized that this relationship applies at only
104 CHAPTER 5 Antenna models
Rt Rt = Rrad
Io*cos( . t)
V Rrad
one frequency; one which is slightly higher than that of half-wave resonance. At all other non-
resonant frequencies, the antenna impedance is a resistance in series with a reactance.
Equation (5.1.2) defines the average power which can be transmitted by the antenna
when the frequency is such that this power is at a maximum value. Under all other condi-
tions, the transmitted power is less than this maximum.
If Irms is the root-mean-square value of the current, then:
p
Irms 2 Io A 5:1:4
and
Since optimum transmission of power is obtained when the output impedance of the trans-
mitter is the same as that of the load, the optimum value for Rt and the characteristic
impedance of the cable is 73 W: making 75 W co-axial cable a good choice for signal links
between radio-frequency equipment and antennae.
: latitudinal
: longitudinal
Sr
Hf
r
Eq
y
where E is the electric field strength and H is the magnetic field strength. The relationship
between E, H, and the power density is:
Hence:
E2
S W=m2 5:1:11
Zo
and
S Zo H 2 W=m2 5:1:12
This means that if the value of either E or H is used to specify the strength of the electro-
magnetic field in the far field, it is always possible to determine the value of the power
density.
106 CHAPTER 5 Antenna models
R F Receiver
/2
Rrec
Rrad
Vrms Vrms
Rrec 2
Rrec = Rrad = 73
For a half-wave dipole receiving antenna, the radiation resistance is 73 W, exactly the same
as for a transmitting antenna.
73
Cp1
I
Cp1
Rrad
V
2 Lp1 Rp1 Lp1 Rp1
2 2 2 2
Rrad Cp1
0V I
2
0V Rrad
Cp1 I 2
and the radiation resistance. At resonance, the circuit acts as a series tuned circuit where the
only limit to the current is provided by the resistive components.
Figure 5.2.2 illustrates the situation where the co-axial link to the transmitter has been
removed and the two halves of the dipole are shorted together to form a single conductor.
Signals can be introduced into this conductor by threading it through a ferrite core which acts
as a step-down voltage transformer. In the figure, this is illustrated as a voltage source V.
With such a configuration, there will be a flow of current forward and backward along
the conductor. Any current flowing into the right-hand monopole can only come from the
left-hand monopole and vice versa. Since the currents and voltages in the two halves of the
dipole are balanced it is useful to redraw the model, giving the circuit of Figure 5.2.3.
This is effectively a bridge circuit with voltages in one arm balancing voltages in the
other arm. If the voltage at the center of the conductor is defined as zero volts, then the
voltage at the mid-point of Rrad must also be zero.
Hence it is legitimate to join these two points together in the model. When this is done
and attention is focused on the right-hand monopole, the model simplifies to that shown by
Figure 5.2.4.
Now, if the single conductor were to be replaced by a pair of parallel conductors con-
nected together at the mid-point and a voltage source inserted in series with both conductors,
then the circuit model of the right-hand monopole would be as shown in Figure 5.2.5.
5.2 The virtual conductor 109
V
2 Lp1 Rp1 Lp1 Rp1
2 2 2 2
I
Cp1
Rrad
2
virtual conductor
This correlates closely with Figure 2.7.6; with the characteristics of the environment repla-
cing the third conductor. The environment is acting as a virtual conductor.
Deriving formulae for the components of this conductor is simply a matter of repeating
the procedure introduced in Chapter 2. The starting point is to set out the primitive equations
for two conductors:
Vp1 Zp1;1 Ip1 Zp1;2 Ip2
5:2:1
Vp2 Zp2;1 Ip1 Zp2;2 Ip2
These are illustrated in diagram form by Figure 5.2.6, which also relates primitive currents
and voltages to loop currents and voltages.
From the diagram:
Ip1 Il1
5:2:2
Ip2 Il2 Il1
and
Vl1 Vp1 Vp2
5:2:3
Vl2 Vp2
110 CHAPTER 5 Antenna models
Zp1,1 Zp1,2
Vp1
Ip1
Il1
Vp2
Ip2
Zp2,1 Zp2,2
Vl2 Il2
Ip1 + Ip2
Rearranging:
Vp1 Zp1;1 Zp1;2 Il1 Zp1;2 Il2
Vp2 Zp2;1 Zp2;2 Il1 Zp2;2 Il2
From (5.2.4):
Zl1;2 Zl2;1 5:2:6
5.2 The virtual conductor 111
Zc1
Ic1
Zc2
Vc2 Ic2
Zc3
Figure 5.2.7 shows the circuit model which creates a similar pair of equations:
Vc1 Zc1 Zc2 Ic1 Zc1;2 Ic2
5:2:7
Vc2 Zc1;2 Ic1 Zc2 Zc3 Ic2
Comparing this set of equations with that of (2.5.5) reveals that the inductance value asso-
ciated with each conductor is the same, whether the conductor pair is viewed as a trans-
mission line or as an antenna. The new parameter introduced by (5.2.9) is the circuit
inductance of the virtual conductor, Lc3. This is the primitive inductance of a conductor with
the same radius as the separation between the conductors.
1
Letting Zpi;j and invoking (2.3.1) gives:
j w Cpi;j
2 p eo er l
Cc1 r1;2
ln
r1;1
2 p eo er l
Cc2 r1;2 5:2:10
ln
r2;2
2 p eo er l
Cc3
l
ln
r1;2
Comparing this set of equations with that of (2.5.9) reveals that the capacitance value
associated with each conductor is the same, whether the conductor pair is viewed as a
transmission line or as an antenna. The new parameter introduced by (5.2.10) is the circuit
capacitance of the virtual conductor, Cc3.
As would be expected, there is a duality between the primitive capacitance and the
primitive inductance of the cable assembly.
A necessary condition for the derivation of a value for the radiation resistance Rrad is
that the length of the conductor is much greater than its radius. Since the separation between
conductors of a transmission line is much less than the length, the same condition is met for a
transmission line as for an antenna. At this point, it is reasonable to assume that the radiation
resistance of the cable is 73 W, the same as defined by (5.1.3).
Having derived formulae for all the parameters associated with the conductor pair, it
becomes possible to construct a general circuit model which simulates the coupling between
an isolated length of twin-conductor cable and the environment (Figure 5.2.8).
Following on from the creation of this model, it is possible to define the virtual conductor
as an imaginary conductor which enables the coupling between cable and environment to be
simulated. It behaves as a return conductor for antenna-mode current.
It has the same properties as an actual conductor capacitance, inductance, and resis-
tance. Numerical values for the reactive parameters can be derived from data on the physical
construction of the cable. An initial assumption can be made that the resistance is 73 W, the
same as that of the half-wave dipole. Tests, such as those described in Chapter 7, can be
devised to refine the values of the three parameters.
As far as differential-mode signal transmission is concerned, the values of the reactive
components of the circuit model remain exactly the same as those of a pair of conductors. As
far as antenna-mode coupling is concerned, the cable can be treated as a single conductor
with the same radius as the separation between the conductor pair.
In Figure 5.2.8, it is assumed that two identical voltage sources are used to inject a signal
into the cable. Such a configuration could be implemented by clamping a toroidal transfor-
mer round the cable at its mid-point.
5.3 The threat voltage 113
Cc3 Cc3
remote
twin-conductor transducer
signal processing unit
cable
structure
l l
signal conductor
Lc1 Rc1 Lc1 Rc1
2 2 2 2
structure
Lc4 Rc4 Lc4 Rc4 Lc2 Rc2 Lc2 Rc2
2 2 2 2 2 2 2 2
Rrad Lc3
Vthreat 2 virtual conductor
This model differs from that shown by Figure 5.2.8 in that the voltage sources in the
conductors have been replaced by a single source located in series with the virtual conductor.
The source in Figure 5.3.2 now represents the effect of the external electromagnetic field.
The triple-T network representing the left half of the twin-conductor cable has been replaced
by a single-T network representing the structure.
Such a representation allows the circuit to be analyzed from the point of view of radiated
susceptibility. But before such an analysis can proceed, it is necessary to relate the amplitude
of the voltage source to the strength of the field in which the system is immersed.
Figure 5.3.3 depicts a length of conductor which is subjected to an external field and the
associated circuit model. An incremental voltage dV is induced in series with each element
dz of the conductor.
dV E dz 5:3:1
Over the length of the cable and structure, the voltage is:
l
2p
Vthreat Emax cos z dz
l
l
This leads to the relationship between threat voltage and electric field strength
5.3 The threat voltage 115
dz
z
dV
Emax
z
l l
l l
4 4
l 2p
Vthreat Emax sin l 5:3:3
p l
Figure 5.3.5 illustrates the relationship when the length l is less than a quarter wavelength.
From (5.1.11):
p
Emax S Zo
5:3:4
where Zo 377 W
It is clear from (5.3.5) and (5.3.3) that the threat voltage is a function of both frequency f and
length l. Equation (5.3.4) demonstrates that the electric field strength is proportional to the
square root of the power density. Combining these equations in the way illustrated by the
subroutine Vthreati in Figure 5.3.6 allows the relationship between power density and threat
voltage to be calculated over a wide range of frequencies. This response is illustrated by the
solid curve of Figure 5.3.7. It is assumed that the power density S is 1 W/m2and that the
length l is 15 m.
116 CHAPTER 5 Antenna models
Vthreat
2
z
l l l l
4 4
Vthreat
2
i := 1..100 fi := i 106 Hz
c c
Vthreati := l Venvi := l
fi fi
l l
Va S Zo p Va S Zo p
l l
Vb Va sin 2p Vb Va sin 2 p
l l
Vc Vb l
Vb Va if l >
4
Vc 10.1 if Vc < 10
Figure 5.3.6 Calculating the relationship between threat voltage and frequency.
This curve is a series of peaks, each peak occurring at a resonant frequency. The first
resonance occurs when the length l is equal to a quarter wavelength. At this frequency and at
higher resonant frequencies:
l
Vthreat Emax 5:3:6
p
Joining these points together creates an envelope curve, Venv. Since this envelope is of threat
voltages which are equal to, or higher than, all voltage points on the Vthreat curve, it
represents worst-case conditions. This curve is defined by that of the Venvi function of
Figure 5.3.7 and is illustrated by the dashed curve.
When the frequency is lower than that of quarter-wave resonance both curves are co-
incident.
Assigning this envelope voltage to the source of the circuit model of Figure 5.3.2 will
allow the response of the assembly-under-review to be simulated.
5.4 The threat current 117
1 103
A/V 15 m cable
Vthreat
100
Venv
10
1 106 1 107 1 108
f Hz
signal conductor
Lc1 Rc1 Lc1 Rc1
2 2 2 2
return
Rrad
conductor
Cc3
Lc3
virtual conductor
2
r := 1.7 108 m l := 15 m
mo l r1, 2 mo l r1, 2 mo l l
Lc1 := ln Lc2 := ln Lc3 := ln
2 p r1, 1 2 p r2, 2 2 p r1, 2
r l
Rss1 := Rss2 := Rss1 Rss3 := 0
p (r1, 1)2
4 r
Fx := = 1.077 105 Hz
mo p (r1, 1) 2
2
0
l 1
Cc := Gc := 0
c Lc
0
65.873
Lc
Zo := Zo := 65.873 Rn := Zo1 + Zo2 Rf := Rn
Cc
565.632 Rn = 131.746
1
Fq := = 4.997 106 Hz
4 Lc1 Cc1
are obtained from (2.5.11). Since the virtual conductor does not have the properties of a
conventional conductor, Rss3 is set to zero. The crossover frequency Fx is obtained from
(2.5.14) and the capacitor values are derived from (2.3.8). Conductance values are set at
zero, since it is assumed that there are no losses in the insulation.
Values for inductance, capacitance, and resistance of the representative circuit model are
then displayed on the sheet. To minimize reflections at the near and far ends of the line, the
values of Rn and Rf are set at a value equal to the characteristic impedance. If the values of
the terminating impedances are any other value, the peaks in the response curve would be
higher. Terminating the line with perfectly matched impedances does not eliminate
interference.
The final parameter to be calculated on this page is the value of the quarter-wave
frequency Fq. This invokes the use of (2.3.9).
Assigning these values to the general circuit model of Figure 5.4.1 leads to the repre-
sentative circuit model of Figure 5.4.3.
Re-drawing Figure 5.4.1 in terms of distributed parameters results in Figure 5.4.4.
Inspection of this model allows the equations of the Zloop( f ) function to be defined, and
this function is listed in Figure 5.4.5. It invokes the Zbranch( f ) function, which is identical to
that in Figure 4.3.4.
0.25
132 132
1.65 H 0.25 1.65 H
0.25
Vthreat 760 pF 760 pF
73 14 H 88 pF
Z1,1 Z1,1
Rn I1 I2 Rf
Z1,2
Z2,2 Z1,2
I3 Z2,1
V3 I2
I1
Zbranch( f ) := 2 f
for i 1.. 3
f
Rci Rssi 1+
Fx
Rci + j Lci
Zo
Gci + j Cci
Z1, i Zotanh
2
Z2, i Zocsch( )
Z
Zloop( f ) = Z Zbranch( f )
Z11 Z1, 1 + Z2, 1 + Z2, 2 + Z1, 2 + Rn
Z12 (Z2, 2 + Z2, 1)
Z13 (Z1, 2 + Z2, 2)
Z22 Z1, 2 + Z2, 2 + Z2, 1 + Z1, 1 + Rf
Z23 Z2, 2
Z33 Z1, 2 + Z2, 2 + Z2, 3 + Z1, 3 + Rrad
Z11 Z12 Z13
Z12 Z22 Z23
Z13 Z23 Z33
c
S := 1 Zo := 377 Vthreat( f ) :=
f
Va S Zo
l
Vb Va sin 2
Vb Va if l >
4
Figure 5.4.5 Calculating values for impedance matrix and threat voltage at each frequency.
5.4 The threat current 121
10
A/V 15 m cable
Ithreat
0.1
0.01
1 106 1 107 1 108
F Hz
The curve itself is very similar to that of Figure 5.3.7, the main difference being that the
response falls sharply as the frequency reduces below that of quarter-wave resonance.
This means that the configuration of Figure 5.3.1 is most susceptible to interference at
about 5 MHz. At that frequency, an incoming wave with a power density of 1 W/m2 will
generate a threat current of about 1 A in the signal circuitry. This current could trigger an
electro-explosive device, or could cause overheating in transducers or semiconductor
devices with consequential damage to the system.
122 CHAPTER 5 Antenna models
The subsequent peaks at higher frequencies could also pose a threat; they could induce
false signals into the system, or prevent actual signals from reaching the intended input.
System upset could occur. However, the designer is now in possession of information not
previously available. Given a detailed understanding of both the effect of the interference on
the signal-under-review and the bandwidth, amplitude, and function of that signal, it should
be possible to design the system to be immune to radiation in any defined EMI environment.
external interference
conducting surface
structure
Figure 5.5.1 Interface diagram for the signal link under review.
s = 1.2 mm
r = 0.4 mm
h = 10 mm
ground plane
In most situations, the routing of the cable is fairly convoluted. So, the geometry of the
assembly becomes complicated and the task of defining the various cross sections becomes
onerous.
One simple way of overcoming this problem is to invoke the concept of the ground
plane in its intended sense; as a purely reflecting surface. The ground plane does not need to
be infinite. It just needs to be able to create an image of the conductors routed above it. A
glass mirror does not need to be infinite in size to provide a good quality image of the object
in front. An estimate can be made of the average separation between cable and structure and
that value used to define the cross section. Figure 5.5.2 illustrates the geometry of a section
of such an assembly.
Given knowledge of the cross section of the cable assembly and the details of the
interface circuitry, enough information now exists to create a circuit model. Since a repre-
sentative circuit model of a similar assembly has already been derived in the section on
cross-coupling, a Mathcad program is available to calculate the component values. Fig-
ure 5.5.3 is essentially a copy of the worksheet page from that section, modified slightly to
utilize input data from the link-under-review. This constitutes the first page of worksheet 5.5.
124 CHAPTER 5 Antenna models
4
Fx := = 1.077 105 Hz see (2.5.14)
o r 2
o l 2 h s
Lc1 := ln Lc2 := Lc1
2
r s2 + 4 h2
o l s2 + 4 h2
Lc3 := ln
2 s see (2.11.3)
2
1 l
Cc := see (2.3.8)
Lc c
1
Fq := = 4.997 106 Hz see (2.3.9)
4 Lc1 Cc1
Component values for three-conductor assembly of Figure 5.5.2:
132 132 0
Zn := 0 Zf := 0 Gc := 0 S Rrad := 50
0 0 0
Figure 5.5.3 Calculating parameter values for the three-conductor circuit model.
In this setup, the objective is to analyze the response of the circuitry to an external field
rather than the cross-coupling between two signals. So the program identifies signal, return,
and structure as conductors 1, 2, and 3, respectively.
Data from this first page of the worksheet can be used to assign values to most of the
components of the triple-T model; most, but not all. It is still necessary to relate the model to
the external field.
The action of the incoming electromagnetic field is to create a current in the structure,
and this current develops a voltage in the common-mode loop. If it is assumed that the
shielding effectiveness of the structure is zero, then it can also be assumed that all the power
5.5 Coupling via the structure 125
common-mode
Rrad
loop 4.22 mH
50 structure
4.22 mH 0.0025
Vthreat
Figure 5.5.4 Representative circuit model of radiation coupling via the structure.
in the interference field is transferred to the structure. Such a transfer of power can be
represented by a voltage source Vthreat in series with the radiation resistance Rrad and with
the structure. Under worst-case conditions the amplitude of the voltage source Vthreat can be
defined by the envelope curve of Figure 5.3.6.
In section 5.2 on the virtual conductor it is assumed that the value of the radiation
resistance is 73 W; the same as that of a half-wave dipole. However, an actual test on a single
length of conductor revealed that, with an assumed value of 73 W, the response of the model
was much lower than the actual response. Figure 7.4.8 illustrates this discrepancy. It was
reasoned that this discrepancy was due to storage of energy in the conductor rather than the
transfer of power to a 73 W load.
A subsequent test on a twin-conductor cable revealed that the measured value of the
radiation resistance was indeed much less than 73 W. Figure 7.5.10 illustrates the response of
the cable-under-test on the same graph as the response of the model, and shows that the peak
values coincide when it is assumed that the radiation resistance is 50 W. This is the value
assumed for Rrad in Figure 5.5.4.
Having established a rationale for a method of linking the circuit model to the external
field, it becomes possible to complete the model. Figure 5.5.4 shows the representative
circuit model for a twin conductor cable routed along the structure.
It is a fairly simple step to convert this lumped parameter model to one which uses
distributed parameters. This is illustrated by Figure 5.5.5. This distributed parameter model
also defines the four circuit loops used in the subsequent mesh analysis.
There is now enough data to compile the second page of the worksheet, and this page is
illustrated by Figure 5.5.6. The Zbranch(f) function calculates numerical values for the six
distributed parameters of Figure 5.5.5. This data is provided as input to the Zloop( f ) function
which compiles a four-by-four array of loop impedance values for the circuit at the spot
frequency f. This page of the worksheet is a slightly modified version of page 2 of the
worksheet 4.3.
Figure 5.5.7 illustrates the final set of computations on page 3 of the worksheet.
The power density S of the incoming interference is set at a constant level of 1 W/m2.
126 CHAPTER 5 Antenna models
I1 I3
V2
I1 I2
I3 I4
The Vthreat( f ) function then calculates the value of the threat voltage at the given frequency f.
This function is a copy of the envelope function for the threat voltage, derived in section 5.3
and recorded in Figure 5.3.6.
The integer n is used to set the number of spot frequencies in the frequency range
between zero and the frequency of quarter-wave resonance Fq. It has already been calculated
(in page 1 of the worksheet) that the value of Fq is approximately 5 MHz. The integer s is
used to set the total range of frequencies to twenty times the quarter-wave frequency. The
variable F is a vector which defines every spot frequency in the range.
The main program is defined by the Iout function. This picks out a value for the fre-
quency f from row s of the F vector, assigns a four-by-four matrix of loop impedance values
to the variable Z, sets the voltage V2 to the value of the threat voltage at that frequency,
calculates values for the four loop currents, provides the magnitude of the current I2 as an
output variable, and stores the results in the vector Iout.
Since I2 is a measure of the current at the near end of the common-mode loop, the vector
Iout is effectively a table of the common-mode current at every spot frequency. The two
vectors Iout and F can then be used to display a graph which defines the frequency response
of the common-mode current. This is replicated by Figure 5.5.8.
This response follows the expected pattern; a null at the quarter-wave frequency followed
by a peak at the half-wave frequency. This is followed by a series of peaks and nulls of ever-
decreasing amplitude.
By selecting the value of I1 as the output variable of the main program, a graph can be
created which defines the frequency response of the differential-mode current. This is illu-
strated by Figure 5.5.9.
Several features of this response are worth noting:
Between 100 kHz and 1 MHz, the amplitude of the current in the differential-mode loop
rises. This is because the current in the common-mode loop is relatively constant. Constant
current at increasing frequency flowing in the inductance of the return conductor causes the
5.5 Coupling via the structure 127
Zbranch( f ) := w 2 p f
for i 1.. 3
f
Rci Rssi 1+
Fx
Rci + j w Lci
Zo
Gci + j w Cci
q
Z1, i Zo tanh
2
Z2, i Zo csch(q)
Zloop( f ) := Z Zbranch( f )
Z11 Zn1 + Z1, 1 + Z2, 1 + Z2, 2 + Z1, 2 + Zn2
Z14 Z2, 2
Z23 Z2, 2
Figure 5.5.6 Calculating the values of the branch and loop parameters.
voltage across that inductance to rise. Since this voltage is applied to the differential-mode
loop, the current in that loop increases.
As the frequency rises above 2.5 MHz, the response flattens out. This is because the
132 W resistors act to minimize the amplitude of the resonant peaks. It is worth noting that
2.5 MHz corresponds to one eighth of a wavelength and that the first crossover point in
Figure 4.3.8 also occurs at one eighth of a wavelength.
128 CHAPTER 5 Antenna models
Fq
n := 100 s := 1.. 20 n Fs := s
n Defining the frequency range
100
A 10
Iout 1
0.1
0.01
1 105 1 106 1 107 1 108
F Hz
Iout 0.1
0.01
1 105 1 106 1 107 1 108
F Hz
easiest simplest most idealised solution. It also represents worst-case conditions, since it
does not take into account the fact that other equipment in the system is also absorbing
interference energy.
If the cross section of the cable is not uniform, it would be necessary to analyse the
properties of each section individually and invoke the concept of equivalent circuits, as
described in section 1.3.3. The general circuit model of Figure 5.5.5 can cater for impe-
dances of any value at the terminations. If any doubt exists as to the accuracy of any parti-
cular simulation, then electrical measurements can be carried out on a test rig.
transmitter
near-end unit
far-end unit
structure
Grouping these equations together allows the threat voltage to be related to the
transmitter power:
Gt Pt
S
4 p r2
p
E Zo S
l l 5:6:1
Vthreat E if l >
p 4
l 2pl l
Vthreat E sin if l
p l 4
The maximum power which can be delivered to the victim loop is limited only by the
radiation resistance and the threat voltage. From Figure 5.5.4:
Vthreat 2
Pthreat 5:6:2
Rrad
Using (5.3.6) to substitute for Vthreat:
2
l E2
Pthreat
p Rrad
Substituting for E:
l2 S Zo
Pthreat 5:6:3
p p Rrad
Hence, if Ge is defined as:
Zo
Ge 5:6:4
p Rrad
then, using (5.1.9) and (5.1.3) to substitute for Zo and Rrad in (5.6.4), the value of the gain
Ge can be calculated to be 1.64. This correlates precisely with (5.1.8). Equation (5.6.3) can
be re-written as:
Ge l2 S
Pthreat 5:6:5
p
Comparing (5.6.5) with (5.1.15) leads to the conclusion that the power delivered to the
victim loop can, in theory, be four times that which appears at the receiver terminals of a
matched antenna/load assembly. Such a conclusion would imply that the assumed value of
the radiation resistance should be 36.5 W, rather than the 73 W defined in (5.1.3).
Although the tests described in sections 7.4 and 7.5 measure the value of the radiation
resistance to be much less than 73 W, neither of them achieve the theoretically minimum value
of 37.5 W. It can be surmised that the discrepancy is due to re-radiation of the stored energy.
It can also be reasoned that the test of section 7.5 represents worst-case conditions, since
all the antenna-mode power is delivered to the cable. This would not be the case in a prac-
tical situation. Some of the power of the input radiation would be absorbed by the adjacent
wiring in the system. Since the test on the twin-conductor cable results in a measured value
of 50 W for Rrad, this would seem to be a reasonable starting point for any analysis of
radiation susceptibility.
132 CHAPTER 5 Antenna models
From (5.1.6), the power density at a monitor antenna located a distance r away from the
assembly would be:
Ge Prad
Sm 5:7:2
4 p r2
Since the assembly under test is the same as that which had been subjected to the radiation
susceptibility test, then the gain Ge would be the same as defined by (5.6.4). Substituting for
Ge and Prad gives:
Zo Irad 2 Rrad
Sm
p Rrad 4 p r2
Irad 2
Zo
4 p2 r 2
receiver
near-end far-end
unit unit
structure
Irad
structure 4.22 mH
4.22 mH 0.0025
761 pF
761 pF 296 pF
It is worth noting that this relationship does not depend on the value of Rrad. From (5.1.12)
the magnetic field strength at the surface of the monitor antenna is:
r
Sm
Hm
Zo
This establishes a relationship between the current radiated into the environment by the
equipment under test and the power delivered to the input of the monitor receiver.
Adding this set of equations to the subroutine which calculates the value of the common-
mode current will provide a clear relationship between signal output voltage of the near-end
unit of Figure 5.7.1 and the level of interference measured by the test receiver.
CHAPTER 6
Transient analysis
135
136 CHAPTER 6 Transient analysis
The equations in section 6.2 are based on the concepts of partial currents and partial
voltages used in textbook analysis of transients in transmission lines. A program is described
which simulates the response of a twin conductor line when a step pulse is applied to the near
end. It is possible to develop this model to include the response of interface circuitry at the
terminations.
Section 6.3 examines the characteristics of the delay-line model, and shows that the
properties of inductance and capacitance are still present.
When a sinusoidal source at the quarter-wave frequency is switched on to a twin-
conductor cable which is open-circuit at the far end, the current will gradually build up as
the energy in successive half-cycles is stored in the cable. This is illustrated by Figure 6.3.10.
If a signal at half-wave frequency is connected, the current delivered to the cable will
gradually decrease. This is because the stored energy in the cable increases to a level where
energy delivered from source to cable is balanced by energy delivered from cable to source.
Figure 6.3.11 illustrates this.
These two illustrations show that an analysis of the transient response of a system to a
sinusoidal signal helps us to better understand frequency response analysis.
Section 6.4 tackles the problem of analyzing antenna-mode transients. An experiment is
described where a step pulse was injected into a twin-conductor cable which was open-
circuit at the far end. After a trial-and-error process, a circuit model was identified; one
which provided a fair reproduction of the response. The existence of such a model raised
several questions about the mechanisms involved, questions which led to a further devel-
opment of the model.
In section 6.5, the picture emerges of wavefront propagating along the signal conductor
in the form similar to the bow wave of a ship. Given a separation of 2 mm between the
conductors, the wavefront does not arrive at the return conductor until the forward edge has
progressed a further 2 mm along the signal conductor. Electric charge is deposited on the
return conductor. Induced voltage causes a current to flow back toward the source. This
current creates an electromagnetic field that couples with the signal conductor and that also
radiates outwards. Since the current in the return conductor does not completely balance that
in the signal conductor, there is an unbalanced charge which propagates along the cable, just
behind the wavefront.
Since the return current is always less than the signal current, there must be a net flow of
aerial-mode current out toward the far end. This outgoing current must be balanced by an
inflow of current from the ground connections. Effectively, the system is behaving like a
dipole, with the cable acting as one monopole and the ground conductors acting as the other.
This means that there are at least three current components involved
that which carries differential-mode signal,
that which deposits unbalanced charge on the cable, and
that which flows out radially in the form of electromagnetic radiation.
Section 6.6 derives a general circuit model which allows the amplitude of all three of these
components to be calculated. A program is described which carries out the computations.
Comparing the response of the model with the waveform displayed on an oscilloscope
allows the two waveforms to be aligned. The values so derived for component values allow a
representative model to be derived for the cable-under-test; Figure 6.6.5. The tests them-
selves are described in section 7.6.
6.1 Time-step analysis 137
For a capacitor:
Q
V 6:1:3
C
where:
t
Q I dt 6:1:4
0
given that
t elapsed time
dt finite increment of time
dI finite increment of current
138 CHAPTER 6 Transient analysis
At the end of that increment of time the value of the current will have changed. In a Mathcad
computer program such as that illustrated in Appendix A, the new value for the current
would be defined as:
I I dI 6:1:7
During this time, the charge on the capacitor will have changed, and the program statement
which updates the value of Q would be:
Q Q I dt 6:1:8
These last three equations form the basis of a simple subroutine next(D) in the Mathcad
worksheet of Figure 6.1.2. The variable D is a two-element vector containing the value of the
current and the charge in the circuit of Figure 6.1.1 at any particular instant.
The subroutine picks out these two values, uses (6.1.6)(6.1.8) to update them, and then
returns the new value of I and Q. The value of the time increment dt has been set at 1 ms on
the top line of the worksheet, together with the other circuit parameters.
C R
100 nF 20
L
V I 1 mH
Worksheet 6.1.1
L := 1103 H C := 100 109 F R := 20
V := 1 V dt := 106 s N := 100
0
D := next(D) := I D1
0
Q D2
dt Q
dI V RI
L C
I I + dI
Q Q + Idt
I
Q
0.01
Iout 0
5 103
0.01
0 2 105 4 105 6 105 8 105 1 104
t s
The control variable i for the main program is set to vary from 2 to N, where N has been
preset at 100. The main program itself is defined immediately below the subroutine.
It simply updates the data variable D at each time step, picks out the value of D1 (the
current I), and transfers this to the element Iouti, where the vector Iout is the record of
the output current.
The variable t records the time of each calculation, and allows the graph of Iout versus
time to be displayed; Figure 6.1.3. Not surprisingly, this is a damped sine wave.
140 CHAPTER 6 Transient analysis
Q1 Q2
Vgen R1 R2 I1 R2 I2 6:1:9
C1 C1
Q1 Q2
0 R2 I1 R2 R3 I2 R3 I3 6:1:10
C1 C1
dI3
0 R3 I2 R3 I3 L1 6:1:11
dt
In formulating the program which calculates the values of the variables at each time step,
it is useful to take into consideration the characteristics of the step pulse. Theoretically,
it changes from zero volts to Vgen in zero time. As far as the initial step is concerned, C1
behaves as a short-circuit and L1 behaves as an open-circuit. Most of the initial surge current
I1 flows through C1 and R2 in series. The resultant voltage across R2 creates a current I2.
Current I2 flowing in R3 creates a voltage across L1.
This means that, initially, I1 is the most significant current, I2 is less significant, while the
value of I3 is inconsequential. This effectively determines the order in which parameter
values should be calculated.
Since the charge on the capacitor is always due to I1 I2 , (6.1.9) and (6.1.10) can be
simplified by defining:
Q Q1 Q2 6:1:12
It is assumed that the initial amplitudes of I1, I2, and I3 are zero.
Current I1 can be calculated by rearranging (6.1.9). This gives:
1 Q
I1 Vgen R2 I2 6:1:13
R1 R2 C1
R1
5000
C1
100 nF L1
R3
Vgen
R2 1M 1 mH
I1 20 I2 I3
Since the value of I1 is now known, it can be used to calculate a new value of I2:
1 Q
I2 R2 I1 R3 I3 6:1:14
R2 R3 C1
The new value of I2 can be used to calculate the rate of change of I3:
dt
dI3 R3 I2 I3 6:1:15
L1
Worksheet 6.1.2
R1 := 5000 R2 := 20 R3 := 106
dt := 108 s
0
0
D :=
0
0
next(D) := I2 D2
I3 D3
Q D4
1 Q
I1 Vgen + R2.I2
R1 + R2 C1
1 Q
I2 R2.I1 + R3. I3 +
R2 + R3 C1
dt
dI3 R3(I2 I3)
L1
I3 I3 + dI3
Q Q + (I1 I2) dt
I1
T := 100 106 s
I2
I3 T
N := ceil
Q dt
Figure 6.1.5 Calculating the amplitude of the transient current in the parallel LCR circuit.
142 CHAPTER 6 Transient analysis
It is then a routine matter to assemble (6.1.13)(6.1.17) into a subroutine which updates the
values of current and charge at every time step. This is illustrated in Figure 6.1.5.
The simplest way of checking the accuracy of the calculations is to vary the time step dt
and recalculate. If a reduction in the amplitude of the time step results in no perceptible
change to the output waveform, then the simulation can be assumed to be reasonably
accurate.
In the worksheet of Figure 6.1.5, the simulation time T is set at 100 ms, and the number of
iterations is defined as:
T
N ceil 6:1:18
dt
(In Mathcad, the function ceil(z) returns the smallest integer greater than or equal to z. ) No
matter what value is selected for dt, the simulation will always run for 100ms.
The waveform displayed in Figure 6.1.6 is that of the current in the inductance L1. This is
a ringing transient, similar to that of Figure 6.1.3. The main difference between this and the
previous response is the fact that it approaches a steady-state amplitude of 200 mA. That is,
the current in a 5 kW resistor when a voltage of 1 V is applied.
In formulating programs similar to the two illustrated in this section, the first objective is
to identify the parameter with the greatest change during the first time increment. In the first
example, this is the voltage across the inductor; in the second example it is the current in the
capacitor.
4 104
A
Iout 2 104
1 104
0
0 2 105 4 105 6 105 8 105 1 104
t s
Figure 6.1.6 Waveform of the current in the inductor of the parallel LCR circuit.
6.2 Delay-line model 143
Ifr Ifa RL
Ifi
far end
Figure 6.2.2 Currents and voltages at the far end of the line.
144 CHAPTER 6 Transient analysis
parameter is absorbed, reflected, or incident. Hence, Ifa is the current which is absorbed by
the load resistor in the receiver at the far end.
A noteworthy feature of the parameters Ifi and Ifr is that they are partial currents. The
total current in the line at any particular location is the sum of Ifr and Ifi at that location.
Similarly, Vfr and Vfi can be described as partial voltages.
Currents and voltages in Figure 6.2.2 can be related:
where Ro is the characteristic impedance of the line and RL is the load resistor. Subtracting
(6.2.5) from (6.2.4) gives:
Vfi Vfr Ro Ifi Ifr
Using (6.2.3):
Vfi Vfr Ro Ifa 6:2:7
Using (6.2.10) to substitute for Ifa in (6.2.11) leads to the standard equation for the reflection
coefficient found in any textbook on electromagnetic theory.
Ifr Ro RL
K
Ifi Ro RL
6.2 Delay-line model 145
near end
Vna
Rg
Ini
Vgen Ina
Inr
Vni Vnr
Figure 6.2.3 Currents and voltages at the near end of the line.
This coefficient is not utilized here because the load at the far end could be any mixture of
resistors, inductors, and capacitors. It is important to know the value of the current actually
delivered to the receiver, Ifa. The magnitude of this current is a function of the state of the
interface circuitry at the far end. Once Ifa is known, the value of Ifr can be obtained from (6.2.11).
Current reflected by the far-end termination, Ifr, travels back along the line and manifests
itself as incident current Ini at the terminals at the near end, as shown in Figure 6.2.3. Since
positive loop current is defined as clockwise on all diagrams, and since power is delivered by
the line to the near-end terminals, the incident voltage Vni must necessarily be inverted with
respect to the voltage Vfr.
At any instant, the voltages at the near end are:
Using the same process that was used to analyze currents and voltages at the far end, the
relationships at the near end become:
2 Ro Ini Vgen
Ina 6:2:17
Ro Rg
and
Inr Ina Ini 6:2:18
Unlike the computations involving lumped parameters, where the status of the system at time
t is dependent on its status at time t dt, the status of the equipment interface at each end of
the transmission line is dependent on the status of the interface at the other end of the line at
time t T.
146 CHAPTER 6 Transient analysis
Since the current at each end of the line is time-dependent, it is necessary to store data on
instantaneous values for a large number of time steps. Such a requirement calls for a table of
values to be recorded, each record holding data on parameter values for a defined instant in
time. An array needs to be created to hold this data.
Fortunately, it is unnecessary for this array to store a large number of variables for each
instant. If the interface circuits are purely resistive, then only two parameters are involved,
the reflected current at the far end, Ifr, and the reflected current at the near end, Inr. This
limits the number of columns of the array to two. The number of rows can also be limited,
since the state of the transmission line before t T plays no part in the computation. If the
duration T is divided into N steps of duration dt then:
T
dt 6:2:19
N
Defining the value for dt in this way ensures that a computation is carried out for every
instant a transient arrives at a termination.
This means that the table should contain N records and that each record should contain
two values. Since it is necessary to scan the table several times during any particular simu-
lation, there needs to be some way of relating the time of each event to the appropriate record.
If the time interval between each sample is always dt, then the relationship between
computation number n and time t is given by:
tn n dt 6:2:20
Table 6.2.1 illustrates the necessary correlation between computation number, n, and the
relevant column, p, of the table, given the assumption that the number of columns, N, is 10.
The diagram of Figure 6.2.4 illustrates a configuration in which a signal is transmitted
from one end of a twin-conductor line to the other. The time taken for a transient to pro-
pagate from one end to the other is 100 ns and the characteristic impedance is 100 W. The
signal source at the near end is a voltage generator Vgen with an output impedance of 10 W.
The resistor RL at the far end provides a load impedance of 1000 W.
Determining the response of this signal link to a step voltage is simply a matter of
assembling the relevant equations into an ordered sequence. The worksheet of Figure 6.2.5
illustrates one method.
The top two lines of the worksheet is derived directly from the characteristics of the
configuration-under-review. On the third line, the number N1 is set at 100. This defines
the number of calculations to be carried out during a single traverse of a transient pulse. The
value of N1 is selected to allow the simulation of 30 such sweeps.
n 1 2 3 4 5 6 7 8 9 10
n 11 12 13 14 15 16 17 18 19 20
n 21 22 23 24 25 26 27 ... ... ...
p 1 2 3 4 5 6 7 8 9 10
Inr
Ifr
6.2 Delay-line model 147
Rg
10
Vin
Vgen Ina Ini Ro = 100 RL
Ifr
10 V T = 100 nS Ifi Ifa 1K
Inr
Ro := 100 Rg := 10 RL := 1000
Vgen := 10 V T := 100 10 s 9
T
N1 := 100 dt := N2 := 30 N1
N1
send(INPUT, Vgen) := Ini INPUT2 recv(INPUT ) := Ifi INPUT1
2Ro Ini + Vgen 2Ro Ifi
Ina Ifa
Rg + Ro RL + Ro
Inr Ina Ini Ifr Ifa Ifi
Ina Ifa
Inr Ifr
Ina := data2, N1 0
for i 1..N 2
p point(i)
INPUT data p
Ina
send(INPUT, Vgen)
Inr
Ifa
recv (INPUT )
Ifr
Inr
OUTPUT
Ifr
data p OUTPUT
Ii Ina
I
n := 1..N2 tn := (n 1). dt
Figure 6.2.5 Calculating the waveform of the current at the near end of the line.
148 CHAPTER 6 Transient analysis
The function send (INPUT,Vgen) carries out the computations which define the currents
existing at the near end of the line at any particular instant the time t. It responds to two
input variables.
The voltage Vgen is the amplitude of the voltage source. For a step pulse, its value is zero
at all time before t 0, and 10 V at all times thereafter. For a sine wave, it would be a
sinusoidal function of time.
The variable INPUT is a two element vector containing the values of Inr and Ifr at time t T.
The first line of this subroutine picks out the value of Ifr. Since it is assumed that there
are no losses in the line, then this value is the amplitude of the current Ini arriving at the near-
end terminals at time t.
Given that the values of Vgen and Ini are now available to the subroutine, the values
of Ina and Ini can be calculated using (6.2.17) and (6.2.18). The output of the subroutine is a
two-element vector containing the values of these two parameters at time t.
The subroutine recv(INPUT) performs a similar set of calculations for currents at the far
end of the line and returns a two element vector with the values of Ifa and Ifr.
The point(n) function takes as input the number n of the sequence of computations and
uses it to calculate a value for p, a pointer to the appropriate record in Table 6.2.1.
In Mathcad, the function mod(n,N) returns the remainder of n when divided by N.
The main program is used to calculate a set of values for Ina, the current delivered to the
input terminals of the transmission line.
The first action of the main program is to define the array data as having two rows and
N1 columns. Initially, all the values are set to zero.
A control variable i is set to run from 1 to N2, the total number of computations involved.
For each computation, the value of the integer p is calculated. This points to the
appropriate column in the data array, and the record stored in that column is defined as
the vector INPUT. This is a two-element vector holding the values of Inr and Ifr at time
t T.
These values are treated as Ifi and Ini at time t by the two subroutines send(INPUT,Vgen)
and recv(INPUT), and values for Inr and Ifr are calculated. These are placed in a two element
vector, OUTPUT, and the contents of this vector are used to overwrite the record in column
p of the data array.
The final action of each computation is to select the local variable Ina and place the value
in element i of the vector I.
The output of the main program is a vector containing all the values computed for
Ina. This parameter can be regarded as the current delivered to the input terminals of the
transmission line.
The waveform of this current is illustrated by Figure 6.2.6. The amplitude of the initial
step current is due to a 10 V supply loaded by 10 W and 100 W in series. It is only after
several reflections have occurred that the current settles down to its steady-state value the
current due to a 10 V supply loaded by 1000 W and 10 W in series. During the settling-in
period, there is a burst of high frequency oscillation in the system.
The configuration depicted by Figure 6.2.4 is indicative of many signal links in the
average system. In many configurations, the output impedance of the sender is less than 10 W
and the input impedance of the load is greater than 1000 W. This means that there is a brief
burst of oscillation on the line every time a voltage step occurs.
6.3 Line characteristics 149
0.1
A
See Figure 6.7
0.05
Ina
0.05
0 1 106 2 106 3 106
t s
Step changes in voltage occur every time a logic signal changes state and every time a
device is switched on or off. If the interface circuitry is not matched to the line, then it will
also carry a great number of high frequency transient currents. Moreover, the frequency
will be close to that of quarter-wave resonance, the ideal frequency for creating maximum
emission.
and
p
qjw La Ca 6:3:2
150 CHAPTER 6 Transient analysis
where La and Ca are the loop inductance and the loop capacitance of the line. From
(2.3.8):
p l
La Ca T 6:3:3
v
It follows that:
qjwT 6:3:4
From (6.3.1):
La
Ca 6:3:5
Ro2
Substituting for Ca in (6.3.3):
La
T
Ro
This removes the j operator and the w parameter from the equations and allows inductance
to be defined in terms of T and Ro:
La T Ro 6:3:6
Rg Ro = 100
Vgen 10 Iline
T = 100 ns
10 V
Rg L
Vgen 10 Iind
10 V 10 H
Rg := 10 RL := 0 dt := 1109 L := T Ro L := 1105
dt
I := 0 next(I ) := dI (Vgen Rg I ) Iindn := I next(I )
L
I
I I + dI
1
A
0.8
0.6
Iline
Iind
0.4 see Figures 6.3.1 and 6.3.2
0.2
0
0 1 106 2 106 3 106
t s
Rg
1k near far
Vgen Ro = 100 RL
Iline
10 V T = 100 ns 10 M
Rg
1k near
Icap C
Vgen
1 nF
10 V
Comparing the responses of the open-circuit line with that of the RC circuit can be carried
out by invoking a similar procedure to that employed with the assessment of inductance.
Figure 6.3.7 is a copy of the second page of the relevant worksheet. The new values of Rg
and RL are recorded at the top of the page, and the core calculation of the program is derived
from (6.3.11).
Figure 6.3.8 shows the results of the computations. As expected, the initial amplitude of
the current into the capacitor C is 10 mA, the current flowing in 1 kW due to a voltage
of 10 V. The stepped curve which tracks this response is due to current delivered to the open-
circuit transmission line of Figure 6.3.5. The correlation between the two curves could not be
closer. This demonstrates that the transmission line possesses capacitive properties, and that
the value of the capacitance is given by (6.3.7).
Normally, a capacitor is constructed by winding a length of two closely spaced con-
ductors into a tight spiral. Coupling between the turns of the spiral results in multiple
6.3 Line characteristics 153
0.01
A
8 103
6 103
Iline
Icap
4 103
2 103
0
0 1 106 2 106 3 106
t s
reflections in the assembly, the net result being a curve that is indistinguishable from that of
the lumped parameter model.
Changing Vgen from a step-function generator to a sinusoidal voltage source enables the
response of the open-circuit line to be assessed at any frequency. Figure 6.3.9 illustrates the
configuration. The quarter-wave frequency of the line is:
1
fq 2:5 MHz 6:3:12
4T
If this is the frequency of the source Vgen, then the current delivered would be as shown
in Figure 6.3.10. This illustrates the fact that the amplitude of the current gradually builds
up. At every half-cycle, the amplitude is increased by a small increment. A limit is
reached when the peak current is 1 A, the current in Rg when a peak voltage of 10 V
is applied.
154 CHAPTER 6 Transient analysis
Rg
10 near far
Vgen Ro = 100 RL
Iline
10 V T = 100 ns 10 M
A 0.5
Iline 0
0.5
1
0 1 106 2 106 3 106
t s
A 0.1
0.05
Iline 0
0.05
0.1
0 1 106 2 106 3 106
t s
For the first cycle of the waveform the current is due to a 10 V sinusoidal signal in a
resistance of 110 W. That is, the resistance value is equal to Rg Ro. Subsequent cycles of
the waveform are progressively reduced. After about 100 cycles, the current delivered to the
line is effectively zero. Such a response is due to the fact that the incident current at the near
end of the line generates a voltage across Rg that precisely balances the voltage delivered by
the source generator Vgen.
This means that, as far as the source generator is concerned, a transmission line that is
open-circuit at the far end will appear as an open-circuit at the near end. Again, this con-
clusion is the same as that predicted by ac analysis. However ac analysis does not predict the
existence of a brief burst of current into the line, and that a standing wave exists along the
line all the time an input signal is applied.
signal oscilloscope
generator
Ch1 Ch2
50 50
current
transformer
15 m of twin-core cable
46 46
signal conductor
I
4.7 return conductor far end
near end
Rg Rcable
4.7 5 from far end
Vna Ini = Ifr
after 83 ns
Ro
Vgen Ina Ini Vnr Inr
Vni 100
near end
Inr
to far end
Ifi = Int
Crad after 83 ns
Int Ro
Ins 250 pF
100
The first circuit model created to simulate this waveform was found to be wildly inac-
curate. A modified model was found to be equally unrepresentative. This was followed by a
trial-and-error process that eventually resulted in the model of Figure 6.4.2. This produced
a waveform that was a fair representation of the trace displayed on channel 2 of the
oscilloscope.
What follows is a description of the circuit model and an attempt to relate its operation to
the behavior of the system-under-review. The fact that this approach also seems to be very
trial and error is because thats just what it was. However, the end result is the identifi-
cation of features of electromagnetic coupling which could not be predicted by frequency
response analysis.
In this model, Rg represents the source impedance of the step-function generator Vgen,
while Rcable is a resistor that allows copper losses and dielectric losses to be catered for, and
Ro represents the characteristic impedance of the cable.
Equations for the near end of the cable are:
These are essentially the same as those defined in section 6.2, with the exception that the
parameter Rcable is included in (6.4.5). Using the process described in section 6.2, it can be
deduced that:
2 Ro Ini Vgen
Ina 6:4:6
Ro Rg Rcable
6.4 Antenna-mode current 157
Rearranging (6.4.2):
Inr Ina Ini 6:4:7
This current transient travels down the line at a velocity approaching that of light. During
this transit, most of the current flows from the signal conductor, through the characteristic
impedance Ro, and back via the return conductor. (Most of the current delivered to the signal
conductor is derived from current flow via the conductors of the structure and the source
resistance Rg. Section 6.5 deals with this aspect of the phenomenon in more detail.) How-
ever, not all of the differential-mode current is delivered to the terminations at the far end.
The current that does not arrive is represented by Ins flowing in the radiation capacitor Crad.
The energy apparently lost to the environment due to current flow in the resistor Ro is
effectively stored as a voltage across the capacitor.
The amplitude of Ins can be calculated using Figure 6.4.3, since a current source in
parallel with a resistor can be represented by a voltage source in series with that resistor. The
loop equation for this circuit is:
Qns
Inr Ro Ins Ro 6:4:8
Crad
Hence:
Qns
Ins Inr 6:4:9
Ro Crad
After each time step, the value of Qns changes. Using Mathcad terminology:
Qns Qns Ins dt 6:4:10
After a delay of T seconds, this transmitted current arrives at the far end. Again using
Mathcad terminology:
Ifi Int 6:4:12
In the model, it is assumed that the open-circuit is represented by a high value resistor RL,
and a value of 10 MW is assigned to this parameter.
Ro
100
Crad
Inr Ro Ins
250 pF
After a further time delay of T seconds, the incident current arriving at the near end is given
by:
Ini Ifr 6:4:15
Assembling the relevant equations into a software program leads to the creation of a
Mathcad worksheet. The first page is illustrated in Figure 6.4.4.
The top two lines of the program are derived from component values of the circuit model.
The function send(near, INPUT, Vgen) is a modified version of send(near, Vgen) of
Figure 6.2.5. The modifications are essentially the inclusion of (6.4.9)(6.4.11).
Since the states of the currents and voltages at the near end of the line are now a function
of Ins and Qns, the values of these parameters need to be available as an input to the
subroutine. This is done through the use of the vector near. Since this vector is also used to
provide updated values of Ina and Int to the main program, these parameters are also
included as input variables (even though they are not actually used by the subroutine).
The function recv(INPUT) is essentially the same as that defined in the worksheet of
Figure 6.2.5.
Figure 6.4.5 illustrates the main program. It is a development of the main program of
Figure 6.2.5. To correlate the simulated waveform with the actual waveform displayed on
channel 2 of the scope, it was necessary to define the time of the leading edge T1 and the
sweep time T2. Since all the time steps in the program are equal in value, it is a simple matter
to identify the step counts N1 and N2 at which they occur.
Since the vector near is used in the near-end calculations, this vector needs to be
declared at the start of the main program. It contains four variables. So the number of rows
is four.
The first step of the iterative subroutine of the main program is to set the initial value of
Vgen to zero. At time T1 it is switched to the value Vg. The value of Vg is defined on the first
page of the worksheet. The most significant modification to the subroutine of Figure 6.2.5 is
due to the need to update the values of the parameters Ins and Qns. So the number of rows in
the near vector changes from two to four.
After all the values have been updated the final action of the iterative subroutine is to
place the value of Ina in the appropriate row of the vector Out. When the iterations have been
completed, the contents of this vector are transferred to the vector Idiff.
Figure 6.4.6 shows the waveform of the current Ina of the circuit model of Figure 6.4.2.
This simulates the current that is being monitored by the current transformer in the setup of
Figure 6.4.1. This waveform is extremely informative.
The first rising edge is exactly as would be predicted by the simple delay-line model of
section 6.2. After this first step, the source delivers a constant current to the line. The
amplitude is due to the application of the voltage Vgen to Rg, Rcable, and Ro in series.
6.4 Antenna-mode current 159
Ifi
recv(INPUT ) := INPUT
Ini
2 Ro Ifi
Ifa
RL + Ro
Ifr Ifa Ifi
Ifa
Ifr
This constant current is maintained throughout the time taken by the transient edge to
travel to the far end and for the reflected pulse to travel back to the near end. During this
period, current delivered to the capacitance Crad is totally invisible to the source generator at
the near end. It does not appear in the waveform monitored by channel 2 of the oscilloscope.
When the first leading edge reaches the far end, all the current is reflected straight back
into the line. Its amplitude is unchanged, but its sign is reversed. Positive current leaving the
near end appears as a negative current on its return.
160 CHAPTER 6 Transient analysis
INPUT data p
near send(near, INPUT, Vgen)
Ina
Int
near
Ins
Qns
Ifa
recv(INPUT)
Ifr
Int
OUTPUT
Ifr
datap OUTPUT
Outi Ina
Out
n := 1.. N 2 tn := (n 1) dt
Figure 6.4.5 Calculating the waveform of the current at the near end of the line.
Since the load at the near end is much less than Ro, the amplitude of the reflected current
is almost doubled.
If the simple delay-line model had been used, then this current would have caused a step
change in the line current from positive to negative. This does not happen in practice. The
trailing edge is an exponential curve, providing a clear indication that some of the current did
not reach the far end. The lost energy is effectively stored as a voltage across Crad.
When the voltage step reaches the far end, current Inr of Figure 6.4.3 ceases to flow, and
the amplitude of the voltage source Inr Ro drops to zero. Antenna mode current Ins now
flows back from the cable into the terminals at the near end.
Since the edge of the waveform is now distributed in time, and since subsequent traverses
of the transients are subjected to the same physical phenomena, the result is a gradual
6.5 Radiated emission 161
0.01
A 5 103
Idiff 0
5 103
0.01
0 5 107 1 106 1.5 106 2 106
t s
transition from a square waveform to a sinusoidal waveform. After only a few cycles, the
mismatched line is resonating at its natural frequency.
The amplitude of the waveform gradually decays and after about 3 ms is virtually
undetectable.
It should be emphasized at this point that the component values of the circuit model were
adjusted iteratively until the simulated waveform bore a close resemblance to the waveform
monitored on the scope. In fact, the development of the model and the explanation of how it
simulated the physical phenomena went hand-in-hand. Although the explanation so far
provided is reasonably plausible, there are still some puzzling questions raised by the test
results. These questions are posed and answered in the next section.
Earlier tests on the same cable, using a sinusoidal source, had demonstrated quite clearly
that the coupling between the conductors is close enough to guarantee that most of the
current emanating from the signal conductor is picked up by the return conductor and flows
in the opposite direction. (See section 7.5.)
Also, the test of a single conductor, described in section 7.4, illustrates the fact that all the
current transmitted along one monopole is provided by current delivered from the other
monopole. With the setup shown on Figure 7.6.1, the signal conductor can be regarded as
one monopole, while the other monopole exists in the form of the assembly of earthed
conductors of the test equipment.
It is not plausible to assume that the send and return components of the differential-mode
current propagate simultaneously along the cable. It is true that if the signal source in
Figure 7.6.1 had been completely isolated from structural conductors, then the voltages
applied to the input terminals of the cable would have been equal in magnitude and opposite
in sign. But it is not isolated.
Since the return terminal is connected to both the return conductor and to ground,
it follows that the voltages applied to signal and return conductors are unbalanced. Since
the primitive capacitance of the ground conductors is much greater than that of the return
conductor, that unbalance must be significant. The amplitude of the current flowing from
the return conductor into the ground conductors has to be much less than that delivered into
the signal conductor.
However, the fact that the antenna-mode current must emanate from the ground con-
ductor does not explain why it is flowing in the same direction as the signal current.
A process of elimination means that the energy causing this outward flow can only come
from the voltage source, via the signal conductor.
The picture emerges of a current transient propagating along the surface of the signal
conductor and creating the wavefront of an electromagnetic field. This spreads out in the
same way as the bow wave of a ship. Since the conductors of the transmission line are 2 mm
apart, the wavefront does not reach the return conductor until the current pulse has propa-
gated at least 2 mm along the signal conductor.
When the wavefront does reach the return conductor, it induces a voltage on the surface.
This voltage creates a current that flows back toward the near end. Current in the return
conductor creates an electromagnetic field of its own, a field that spreads out to enclose
the signal conductor. This field tends to neutralize the field emanating from the signal
conductor; but not all of it. The coupling cannot be 100 per cent efficient. So the amplitude
of the induced current is less.
By the time the field from the return conductor reaches the signal conductor, the
advancing wavefront has progressed a further 2 mm. Since this happens for every increment
of length, it must happen along the whole length.
Behind the wavefront, there is an increasing length of cable in which current is flowing in
one direction in the signal cable, and in the opposite direction in the return conductor. Since
the current in the return conductor is slightly less, there must be an unbalanced current flow
in the cable. Since both conductors are now acting as transmitting antennae, the net result is
that there is a residual current flow from the near end. Mutual coupling ensures that this
antenna-mode current is shared equally between the two conductors.
Hence, there are two current components flowing in each conductor; the differential-
mode current and the antenna current. Figure 6.5.1 illustrates the current flow, viewed from
6.5 Radiated emission 163
Iaerial
Idiff +
2
signal conductor
Iaerial
Idiff +
2
return conductor
Iaerial
Idiff +
2
signal conductor
Iaerial
Idiff
2
return conductor
the point of view of unidirectional current. This shows that there is a net flow outward of
Iaerial. At the wavefront, antenna-mode current is flowing in the forward direction along
both conductors.
Figure 6.5.2 views the same two conductors, this time from the point of view of loop
current. Magnetic material in the transformer responds only to the current which encloses the
core. Since the antenna-mode current in one conductor balances the antenna-mode current in
the other, this part of the current flow is invisible to the transformer.
In classical transmission-line theory, the concept of partial currents was introduced to
explain the behavior of reflections at the terminations. The above reasoning indicates that
there is at least one more partial current to add to incident current and reflected current
antenna-mode current.
To summarize: Electric charge is deposited on the return conductor. Induced voltage
causes a current to flow back along the return conductor toward the source. This current
creates an electromagnetic field that couples with the signal conductor and that also radiates
outwards. Since the current in the return conductor does not completely balance that in the
signal conductor, there is an unbalanced charge that propagates along the cable, just behind
the wavefront. In addition, there is antenna-mode current which propagates outwards along
both conductors and is converted into an electromagnetic wave. This means that there are at
least three current components involved:
that which carries differential-mode signal,
that which deposits unbalanced charge on the cable, and
that which flows out radially in the form of electromagnetic radiation.
It is useful at this point to make a distinction between common-mode current and antenna-
mode current:
Common-mode current is that which flows in the loop formed by the return conductor
and the adjacent structure.
Antenna-mode current is that which flows between the cable and the environment when
there is no nearby structure.
164 CHAPTER 6 Transient analysis
dQ V dC 6:5:1
where dC is the capacitance of that element. Since the cross section of the cable is constant,
the capacitance is proportional to the length which has been charged. Since the propagation
velocity is constant, the rate of change of length with time is constant. So the apparent rate of
change of capacitance with time is constant. From (6.5.1):
dQ dC
V 6:5:2
dt dt
dQ dV
C 6:5:3
dt dt
Rg
Vgen Vin
return conductor receiving aerial
ground Idiff
conductor
Iaerial
conducting material. This being so, it must have a high capacitance, a low inductance, and
hence, a low characteristic impedance.
This means that, at the output terminals of the supply, all the antenna-mode current is
flowing in the signal conductor. The signal conductor is behaving as a transmitting antenna
and the return conductor as a receiving antenna. Differential-mode current is delivered to the
return conductor, while the rest is converted into electromagnetic radiation.
From Figure 6.5.3, the voltage applied to the input terminals of the cable is Vin, where:
Comparing Figure 6.5.3 with Figure 6.4.2 allows the parameters of the two models to be
correlated. That is, the current Ina used in the calculations corresponds to Idiff, and the
antenna-mode current Ins corresponds to Iaerial. This gives:
Rg Ro1
Ro2
Irad
virtual conductor
Ro3
Calculating the value for these three components is relatively simple, using the equation:
r
Lci
Roi 6:5:6
Cci
where i is an integer identifying the conductor, and Lci and Cci are defined by (5.2.9) and
(5.2.10).
The resistance Ro as seen by the output terminals of the source at the near end is:
Ro2 Ro3
Ro Ro1 6:5:7
Ro2 Ro3
Ro2
Irad Ina 6:5:8
Ro2 Ro3
Since the far end is isolated from ground, there is no third conductor available to deliver
extra charge to the cable. So the primitive voltages existing at the far-end terminals are
perfectly balanced with respect to the environment. Even so, both conductors act as trans-
mitting antennae, as far as returning current from the far end is concerned.
There could be losses in the dielectric material of the insulation. These could be simu-
lated by placing a high-value resistor between the two conductors at the generator end of
the line.
Some of the transient differential-mode current radiates away. When the polarity of the
radiation from one conductor is positive, the polarity of the radiation from the other is
negative. If the cable is twisted, this differential-mode radiation tends to cancel itself out
at a short distance from the line.
There is also a transient antenna-mode current Irad which emanates from the cable in
the form of electromagnetic radiation. Figure 6.5.4 shows how this can be simulated.
Current flows into the cable from the ground conductor to replace this loss. (This
is the same phenomenon that was analyzed in section 5.2, using sinusoidal
waveforms.)
Section 6.4 has shown that, as well as propagating down the line, the current transient leaves
a residual charge on the line. The amplitude of the stored charge can be determined by the
model of Figure 6.4.2. If a step voltage is applied to the input terminals, current Ins flows
into Crad for 2 T seconds, that is, the time it takes for the step to be reflected back to the
input terminals. Since this charge is eventually converted back into a current and the energy
dissipated in the resistors Rg and Rcable, this stored energy also represents a loss.
Vna
from far end
Rg Rcable Ro1 Ini = Ifr
after T second
Vgen Ina Vin Ini Inr
Ro2
near end
Ine
Ro3
conductors. In the model of Figure 6.4.2, all the losses are simulated by a single resistor,
Rcable.
A comparison between Figures 6.4.2 and 6.6.1 reveals that the significant modification is
the inclusion of a new circuit loop carrying the current Ine. This is the current which flows
into the environment in the form of electromagnetic radiation.
The impedance presented by the cable to the output terminals of the source generator is:
Ro2 Ro3
Ro Ro1 6:6:1
Ro2 Ro3
This is also the impedance presented to the terminals at the far end.
From inspection of Figure 6.6.1, the equation for the antenna-mode loop is:
Hence:
Ro2
Ine Ini Inr
Ro2 Ro3
So:
Ine Loss Ina 6:6:2
where:
Ro2
Loss 6:6:3
Ro2 Ro3
The current flowing from the near end toward the far end is now:
The current which stores antenna-mode charge on the cable is similar to (6.4.9):
Qns
Ins Inf 6:6:5
Rno Crad
And the current which is actually transmitted to the far-end terminals is:
In section 6.5, it was reasoned that the current Ins used to create antenna-mode charge on the
cable must flow through the source resistance Rg. This means that (6.4.5) needs to be
modified. The modified relationship becomes:
Since the current used to deliver stored charge varies along the length of the cable, it is
assumed that no voltage drop is incurred along the length of the cable by Ins. Taking this
new relationship into account, (6.4.6) changes to:
There is now sufficient information available to modify the worksheet of section 6.4 to
analyze the response of the model of Figure 6.6.1. However, several more relationships are
needed before the response of the model can be correlated with that of the setup of
Figure 6.4.1.
The differential-mode current delivered by the source generator to the cable is:
That is:
Idiff Ina Ine 6:6:9
It can be assumed that the relationship between the voltage monitored by channel 1 of the
oscilloscope and the voltage Vin is:
Vch1
K 6:6:11
Vin
Similarly, it can be assumed that the relationship between current monitored by the current
transformer, Imon, and the voltage displayed by channel 2 of the scope is:
Vch2
ZT 6:6:12
Imon
Having established all the necessary relationships, the next step is to assemble them into a
program that simulates the transient response of the configuration. The first page of the
worksheet is devoted to the definition of the input variables associated with the model of
Figure 6.6.1 and with the test setup illustrated by Figure 6.4.1. This page is reproduced by
Figure 6.6.2.
As with the frequency-response models, this time-domain response model can also be
used to correlate the theoretical response with that of the actual hardware. Relevant variables
are identified by the phrase adjust to suit.
The second page of the worksheet, Figure 6.6.3, defines the two subroutines used to
calculate the responses at the near and far ends of the line after each time step. It is a
modified version of the subroutines illustrated by Figure 6.4.4.
6.6 Transient emission model 171
Ro1 := 50 Ro2 := Ro1 Ro3 := 600 See Figure 6.6.1 Adjust to suit
Ro2
Loss := See (6.6.3)
Ro2 + Ro3
Ro2 Ro3
Ro := Ro1 + = 96.154 See (6.6.1)
Ro2 + Ro3
The third page of the worksheet, illustrated by Figure 6.6.4, defines the main program
used to calculate the response of the selected variable over the defined time.
The output variable can be the following:
Vch1: The input voltage waveform, as observed on channel 1 of the oscilloscope.
The leading edge is defined by T1, the trailing edge by T2.
172 CHAPTER 6 Transient analysis
point(n) := m mod(n, N )
m N if m = 0
Vch1 := data2, N 0
near5 0
for i 1..N3
Vgen Vg if i > N1
Vgen 0 if i > N2
p point(i)
INPUT data p
near send(near, INPUT, Vgen)
Ina
Int
Ine near
Ins
Qns
Ifa
recv(INPUT )
Ifr
Int
OUTPUT
Ifr
data p OUTPUT
Idiff Ina Ine See (6.6.9)
Vin Vgen Rg(Ina + Ins) See (6.6.10)
Vch1 K Vin See (6.6.11)
Vdiff ZTIdiff See (6.6.12)
Vrad ZTIne See (6.6.12)
Vi Vch1
Channel 1 selected as output variable
V
alteration was made to the input variables. The program was run three more times, each time
with a different output variable selected.
Then the input variables were assigned to the general circuit model of Figure 6.6.1 to
create the representative circuit model of the assembly-under-test (Figure 6.6.5).
Section 7.6 demonstrates the fact that the same circuit model can simulate three dif-
ferent responses of the cable-under-test. This indicates that it is a useful predictor of
transient radiation. Worksheet 7.6.9 demonstrates that there is a clear correlation between
174 CHAPTER 6 Transient analysis
the frequency response model of Figure 7.5.12 and the transient response model of
Figure 6.6.5.
Both of these representative models have been related to electromagnetic phenomena and
both replicate the actual response of the hardware. So it is fair to claim that the general
circuit models on which they are based can be used with a high degree of confidence. Further
testing and modeling should confirm their reliability. Sufficient information has been pro-
vided for this to be done.
CHAPTER 7
Bench testing
It is normal practice to carry out bench tests of the functional behavior of prototype circuitry
during the development of any new product. This identifies problems that had not been
predicted during the feasibility study and provides an opportunity to rectify those faults. It
also provides an early opportunity to check that the design requirements are being met.
Since EMC is also a functional requirement, it is logical to expect that this aspect of
design should be checked at the prototype stage.
Two items of essential equipment are the voltage transformer and the current transformer.
Several manufacturers produce such items, but they are highly expensive, and not really
suitable for general-purpose use. So the transducers described in this section were assembled
from components obtained from a supplier of electronic components.
Section 7.1 describes the construction of a low-cost voltage transformer. This consists of
ten turns of wire wound on a spit-core ferrite assembly. A low-value resistor in parallel with
the primary winding ensures that the transducer is a wide-band device. The secondary
winding is the loop-under-test. A monitor winding enables the amplitude of the injected
voltage to be measured.
The frequency response of the transformer was checked by applying a known voltage to
the primary winding and measuring the output of the secondary. A circuit model was sub-
jected to analysis using a Mathcad program. Only a few iterations of this program were
needed to establish a one-to-one correlation between the test results and the response of
the model. By this means, a model was created that could be used to simulate the response of
the voltage transformer. This model can be used for analysis in the frequency domain or the
time domain.
A current transformer was assembled in much the same way. This time, the primary was
the loop-under-test and the secondary provided an output voltage proportional to the input
current. A circuit model was created to simulate the device characteristics. This circuit
model provides the calibration curve for the device. The process is described in section 7.2.
Section 7.3 describes the construction of a triaxial cable that can be used to minimize
interference between cables of the test equipment.
Section 7.4 describes a test on an isolated conductor, where a known voltage is injected
into the center of the conductor using the voltage transformer and the current monitored with
the current transformer. Data on the length and diameter of the conductor was used to assign
value to the components of the dipole model of section 5.2. A close match was obtained
between the response of the conductor and the model. However, the peak of the emitted
175
176 CHAPTER 7 Bench testing
current was significantly higher than that of a dipole antenna. This is probably because there
was no resistance in the center of the cable to damp down the oscillations (as there would be
in a dipole antenna).
Section 7.5 describes similar test on a twin-conductor cable, where the terminations are
open-circuit. A circuit model was developed, which simulates both the differential-mode
current and the antenna-mode current.
Section 7.6 provides details of the transient test on the 15 m cable described in sections
6.4 and 6.5. Photographs of the actual waveforms are compared with graphs of the simulated
waveforms, providing persuasive evidence of the validity of the model. It is not claimed that
this model is correct in every detail. However,
the one single model simulates the waveforms of three separate signals reasonably
accurately,
every parameter is shown to be related to electromagnetic phenomena,
it reveals aspects of transmission-line behavior that are not identified in textbooks on
electromagnetic theory,
sufficient information is provided to allow any electronic designer to replicate the tests
and perform the analyses,
clear correlation can be established between the results obtained from the frequency
response tests of section 7.5 and those obtained from the transient tests of section 7.6.
Although the test equipment and tests described in sections 7.17.5 are limited to a band-
width of 20 kHz to 20 MHz, the technique and approach are applicable to a much wider
range of frequencies. Each designer will have test equipment which is suitable for use with
the equipment under development. It is simply a matter of adapting the available equipment
to carry out the type of testing described in this chapter.
The technique of circuit modeling can also be applied to the high-frequency character-
ization of components such as capacitors, inductors, and filters. Section 7.7 provides an
example of the characterization of a capacitor over the range 200 kHz to 1 GHz.
primary
winding
signal
generator
R1 Vin
loop-
oscilloscope 50
Vin under-
Vch test
51
monitor turn
requirement for a wide bandwidth, the impedance of the primary winding should be higher
than R1 at the lowest operating frequency.
The requirement to measure the voltage induced in the loop-under-test is met by the
monitor turn. If the loop-under-test is a short-circuit, then a high current will be induced, and
the applied voltage will be reduced due to the output impedance of the transformer. The
voltage detected by the monitor turn will also drop. This means that the monitor turn will
measure the actual voltage applied to the loop-under-test.
From the circuitry of Figure 7.1.1, the relationship between the voltage Vin applied to the
loop-under-test and the voltage monitored at the appropriate channel of the oscilloscope is:
50 51
Vin Vch 7:1:1
50
The 50 W input impedance at the oscilloscope is provided by a BNC adaptor and a BNC
terminator.
178 CHAPTER 7 Bench testing
Such a transformer was constructed from parts purchased from suppliers of electronic
components, and is illustrated by Figure 7.1.2. The core itself is a cable suppression core
assembly part number 04 31 173 551, supplied by Fair Rite Products Corp.
The primary winding was 10 turns of 22 SWG enameled copper wire wound round one of
the split cores. The monitor winding was a single turn of enameled copper wire wound round
the other core. There were two reasons for this:
It ensures that the magnetic field being coupled was the same as that which was coupling
the loop-under-test.
It minimizes capacitive coupling with the primary winding, since it is wound on the other
half of the split-core ferrite. Figure 7.1.2 illustrates this separation.
The windings were terminated in a standard terminal block, which provided a mount for the
other components. A 68 W, 2 W resistor was connected in parallel with a 240 W, 0.6 W
resistor and the primary winding. This provided a 53 W resistor in parallel with the primary.
Ideally, the co-axial cable should have been terminated by a 50 W resistor to match its
characteristic impedance. However, it is inevitable that reflected impedance of the loop-
under-test will act as a load in parallel with the termination resistor. To compensate for this
loading effect, a value of 53 W was fitted.
Two BNC connectors were connected to the terminal block to provide an interface
with 50 W co-axial cable. A section of a plastic box was used to provide more rigidity to the
assembly.
Although this assembly does not meet the quality standard of a professionally engineered
item of equipment, it does the job intended and has proved to be reliable over a period of
more than 5 years.
Testing was carried out using the setup of Figure 7.1.3. A splitter box was used to provide
an input to channel 1 of the oscilloscope. This allowed the input to the voltage transformer to
be monitored. The output of the monitor turn was monitored by channel 2. Each 16.5 W
resistor is a parallel combination of two 33 W resistors.
The signal generator was set to provide a sinusoidal output at a particular frequency, and
the amplitudes of the peak-to-peak signals were measured on the oscilloscope. This process
was carried out over a number of defined frequencies and the results were tabulated. Column 1
of the table of Figure 7.4 gives the frequency in megahertz, column 2 gives the amplitude of
the channel 1 in volts, and column 3 gives the amplitude of the signal at channel 2 in millivolt.
Hameg
voltage transformer HM604
53
50
51
10:1 channel 2
0.01 2.2 50
Column 1: frequency, mhz
0.02 2.6 90 Column 2: peak-to-peak voltage on channel 1, V
0.05 3.4 146 Column 3: peak-to-peak voltage on channel 2, mV
0.1
TFt
Ref
0.01
1 104 1 105 1 106 1 107 1 108
f Hz
For each spot frequency, the ratio of output voltage Vch2 to input voltage Vch1 was
calculated and this ratio gave the value of the transfer function at that frequency. Plotting this
parameter against the frequency, as shown in Figure 7.1.5, gives an indication of the band-
width of the voltage transformer.
The solid line on the graph defines the 3 dB margin. Comparing the dotted curve with
the solid curve gives a clear indication of the bandwidth. In this case, it is from 20 kHz to
20 MHz.
180 CHAPTER 7 Bench testing
At this point, the characteristics of the device have been defined in terms of a specified
test and a table of results. A more compact and informative characterization is to define the
transformer in terms of a circuit model. Figure 7.1.6 is such a model. It is based on the
textbook treatment of transformers.
Mutual inductance (the inductance shared by primary and secondary) is represented by
L2. Primary inductance (the inductance due to magnetic flux which links only with the
primary winding) is represented by L1.
Since the turns-ratio is 10 to 1, the impedance ratio of the secondary reflected in the
primary is 100 to 1. Resistance R3 is the reflected value of the series resistance of the
secondary circuit, and resistance R4 is the reflected value of the load resistance at channel 2
input terminals.
Since the simulation is only concerned with amplitudes of the signals, there is no need to
know the phase relationship between currents and voltages. So, as far as this model of the
setup is concerned, the co-axial cable is transparent.
Resistance R2 is the parallel combination of the 68 and 240 W resistors. The 16.5 W
resistors represent the splitter-box components, while the 50 W resistors represent the output
impedance of the signal generator and the input impedance of channel 2 of the oscilloscope.
The capacitance C1 represents the capacitance of the primary winding.
The voltage V1 represents the voltage as monitored by channel 1 of the scope, V2 is the
voltage at the junction of the 16.5 W resistors, V3 is the voltage applied to the input terminals
of the transformer, V4 is the voltage across the mutual inductance, and V5 is 10 times the
voltage monitored by channel 2.
Simulating the frequency response of this model over the range 10 kHz to 20 MHz
requires that the frequency steps are spaced logarithmically. So a vector needs to be created
that defines the set of frequencies to be used. A set of 100 frequencies provides enough data
points to ensure a smooth curve on the graph. Figure 7.1.7 shows the section of the Mathcad
worksheet which performs this function.
L1
R1 R3
16.5 6 H
50 16.5 5.1 K
16.5 R2 L2 R4
Vgen C1 V3 V4 V5
V2 53 58 H 5K
V1 50 520 p
y2 y1
y1 := log(10103) y2 := log(20106) m :=
100
i := 1 .. 101 Fi := y m (i 1) + y1
10 y
Figure 7.1.7 Calculating a set of equally spaced frequencies over a logarithmic scale.
7.1 Voltage transformer 181
Calculating the transfer function of the circuit model TFm is performed by the section of
the worksheet illustrated in Figure 7.1.8. The objective is to determine the ratio of V5 to V1,
since the former simulates the voltage at channel 2, while the latter simulates the voltage at
channel 1. So the amplitude of V1 is set at unity. This means that the transfer function of the
model at any frequency is the amplitude of V5 divided by the turns-ratio.
Equations used in the subroutine illustrated in Figure 7.1.8 are derived from inspection of
the circuit model, using the technique of equivalent circuits.
It is now possible to compare the transfer function derived from test results, TFt, with that
derived from the circuit model, TFm, and this is done in the graph of Figure 7.1.9.
This graph illustrates the fact that the curve produced by the circuit model intersects all
the data points derived from testing the assembly. Since the component parts of the model
68 240
R1 := 16.5 R2 := = 52.987
68 + 240
R3 := 5100 R4 := 5000
50 + R1
V1 := 1 V V2 := V1 = 1.33
50
Turns := 10 C1 := 520 1012 F
TFmi := w 2 p Fi
1 1
Y2 +
j w L2 R3 + R4
1
Z2
Y2
1 1
Y1 + + j w C1
R2 Z2 + jw L2
1
Z1
Y1
Z1
V3 V2
Z1 + R1
Z2
V4 V3
Z2 + jw L1
R4
V5 V4
R3 + R4
V5
Turns
0.1
TFt
TFm
0.01
1 104 1 105 1 106 1 107 1 108
f, F Hz
simulating the voltage transformer can be identified, it means that those parts can be used to
define its characteristics.
It is useful to describe the process by which the two curves of Figure 7.1.9 were brought
into co-incidence. It was basically an iterative process.
Initially, guess values were assigned to Lp, L1, and C1. Values of other circuit compo-
nents were fixed. Then the value of one of the parameters at the top of the worksheet was
altered slightly and the page was scrolled up to display the final graph. The curve on the
graph did not change until the mouse was clicked. This allowed the change in the position of
the curve to be noted. If the curves moved closer together, then the same incremental change
was made to the parameter.
Initially, it was assumed that capacitor C1 was very small, and attention was focused on
the low-frequency response. The value of Lp (see third line of Figure 7.1.8) was adjusted to
line up the two curves at the low end of the frequency range, and L1 was adjusted to line
up the mid-frequency response.
It is useful to note that L2 is a dependent variable. This means that the total inductance of
the primary winding of the transformer model can be altered with a single parameter change.
Surprisingly little iteration was needed to achieve coincidence over the lower and mid-
frequency ranges. Finally, the value of C1 was adjusted to line up the high-frequency roll-off.
The useable range of this transformer is from 20 kHz to 20 MHz. It is reasonable to
expect that transformers can be constructed to match the operating frequency range of any
particular system-under-review, for example, by using a smaller transformer to extend the
frequency range upwards or by adding more turns on the secondary to extend the range
downwards.
7.2 Current transformer 183
current
transformer
50
loop- oscilloscope
under-
test R1
51
Iprim
Isec
R1 R2
51 50
R1
51
R2
R1 Isec Vscope
50
From the viewpoint of the input terminals of the oscilloscope, the signal is generated by a
voltage source, as illustrated by Figure 7.2.3. In all the tests described in this chapter, a 50 W
resistor is inserted in parallel with each scope input.
In the particular transformer described here, the core is exactly the same as that used for
the voltage transformer, and the secondary winding comprises 10 turns of 22 SWG enam-
elled copper wire. Figure 7.2.4 illustrates the assembly. The tie wrap is removable, and is
used to ensure that the two halves of the core are tightly clamped together.
To characterize this transformer, a simple coupling jig was assembled. This delivered a
primary current of known amplitude and known frequency, and was designed to ensure that
the loop-under-test was tightly coupled to the transformer.
The test setup is shown in Figure 7.2.5. Channel 1 of the oscilloscope was used to
measure the current delivered to the transformer, that is, the current in the 50 W resistor
placed at the scope input connector. Channel 2 was used to monitor the output of the
transformer assembly.
As with the test on the voltage transformer, measurements were taken of the peak-to-peak
amplitude of the sine waves displayed on the screen of the scope.
Figure 7.2.6 tabulates the results and illustrates how the transfer impedance ZTt is cal-
culated, by dividing the voltage at channel 2 input by the current in the primary winding.
It is worth noting that the parameter resulting from these calculations is derived from the
ratio of two measurements. It can be assumed that the amplifiers for channels 1 and 2 in the
7.2 Current transformer 185
coupling jig
50
signal channel 1
generator
oscilloscope
51 channel 2
transformer
assembly 50
resistor
0.005 8 98
0.01 8 175
0.02 8 270 column 1: frequency, MHz
column 2: channel 1 voltage, V
0.05 8 340
column 3: channel 2 voltage, mV
0.1 8 360
0.2 8 365
data := 0.5 8 375
s := 1..rows (data) fs = datas, 1 106
1 8 370
ZTts := Vch1 datas, 2
2 8 370
5 7.9 365 Vch2 datas, 3 103
Figure 7.2.6 Using the test results to calculate the transfer impedance.
oscilloscope are identical. This being so, most of the errors in the absolute measurements are
cancelled out.
Mathcad software was then used to display the results. As with the response of the
voltage transformer, this was a set of points that could be joined up to create a frequency
response characteristic.
Again, the use of circuit theory enables the creation of circuit model of the link between
the loop-under-test and the oscilloscope input. Figure 7.2.7 is a development of the simple
model of Figure 7.2.2. R2 and R3 represent the 51 W resistor in the transformer assembly and
the resistor at channel 2 input connector, respectively.
L1 represents the inductance of the transformer winding, while R1 represents transformer
losses. These losses could be due to the magnetic field from the loop-under-test which does
not link with the transformer core. Another cause of losses is the eddy current in the core.
186 CHAPTER 7 Bench testing
Isec C1
R1 60 pF R2 R3
L1 Vch2
300 R4 50
200 H 51
850
y2 y1
y1 := log(5 103) y2 := log(20106) m :=
100
i := 1 .. 101
R1 := 300 R2 := 51 R3 := 50
6
R4 := 850 L1 := 20010 H C1 := 60 1012 F
Turns := 10
ZTmi := w 2 p Fi
1
Z1 R4 +
j w C1
1 1 1 1 1
Y2 + + + +
R1 R2 R3 j w L1 Z1
1
Z2
Y2
Z2
ZT
Turns
Capacitance C1 and resistor R4 were added to the model to simulate additional losses at
frequencies over 2 MHz. This model is amenable to the use of Simulation Programs with
Integrated Circuit Emphasis (SPICE) software. It would have been possible to use such
software to produce a frequency response curve similar to that of ZTt. Achieving close
correlation between theoretical results and test results would have involved downloading the
results of the analysis to a computer file, picking up that file with Mathcad software, and then
using Mathcad to compare the two curves.
Such a process was avoided by carrying out the frequency response analysis in the same
worksheet that was used to display the test results. Figure 7.2.8 illustrates the program
involved.
7.2 Current transformer 187
This first calculates a set of 101 frequencies, on a logarithmic scale, between 5 KHz and
20 MHz, and stores them in the vector F. It then defines the component values of the circuit
model as well as the number of turns on the secondary.
The equations used in the function ZTm are derived from an inspection of the circuit
model. The impedance Z2 is the impedance as seen by the current generator. It defines the
ratio of Vch2 to Isec. Dividing Z2 by the number of turns gives the ratio of Vch2 to Iprim,
that is, the transfer impedance of the circuit model.
Figure 7.2.9 illustrates the correlation between the test results and the response of
the model. Although there was an initial discrepancy between the two curves, a few
adjustments of L1 and R1 led to a curve which intersected the data points at the low fre-
quencies. Varying the values of C1 and R4 led to a curve which intersected the data point
above 2 MHz as well.
The existence of this model makes it possible to deduce the amplitude of the current in
the loop-under-test by noting the amplitude and frequency of the signal observed on the
oscilloscope. That is, the response of the model of Figure 7.2.8 provides the calibration curve
for the current transformer.
The fact that the frequency response of the transformer assembly is flat over a wide
bandwidth means that it can also be used to monitor the amplitude and waveform of transient
currents. There is one important proviso: the bandwidth of the waveform being monitored
must lie within the bandwidth of the device.
10
ZTts
1
ZTmi
0.1
1 103 1 104 1 105 1 106 1 107 1 108
fs, Fi Hz
Figure 7.2.9 Transfer impedance from test results, ZTt, and from circuit model, ZTm.
and
1
Rsec 7:2:3
Ysec
This gives:
Vch Rsec Isec 7:2:4
Voltage at the oscilloscope input can be related to the current in the primary loop by
invoking (7.2.1):
Rsec
Vch Isec 7:2:5
Turns
For this particular device, the transfer impedance to use with transient test analysis is:
Rsec
RT 2:27 7:2:6
Turns
The inner and outer braids act as a transmission line, and the characteristic impedance
can be calculated using:
r
1 mo mr r3;3
Ro ln 7:3:1
2p eo er r2;2
where r2,2 and r3,3 are the radii of the inner and outer braids, respectively.
For RG58 cable, the diameter of the screen is 3.3 mm and that of the outer sheath is
5 mm. Allowing for a loose fitting braid, it can be assumed that the braid diameter is 6 mm.
If it is also assumed that the relative permittivity is 4, then invocation of (7.3.1) gives the
value of Ro as 18 W.
In the presence of an external field, antenna-mode current will flow in the outer skin of
the outer braid. Voltage developed along the length of the braid will cause common-mode
current to flow in the loop formed by the inner and outer braid. Since these two conductors
act as a transmission line, any signal arriving at either end is absorbed in an 18 W resistor.
Since these resistors are the same as the characteristic impedance of the line, there will be
little or no reflection. This means that most of the energy delivered to the cable in the form of
electromagnetic radiation will be dissipated in a resistive load. To achieve good performance
at very high frequencies, several resistors would need to be assembled in an annular ring at
each termination to minimize the inductance of the resistive elements.
As far as radiation from the differential-mode signal is concerned, any voltage developed
along the length of the inner braid will cause a circulating current in the outer transmission
line. Again, the energy of the unwanted signal will be absorbed by the two end resistors.
The net result is a significant improvement in the shielding effectiveness of the RG58
cable.
signal oscilloscope
generator channel 1 channel 2
50 50
isolated conductor:
14/0.2 mm equipment wire
1 mm diameter
V I
7.5 m 7.5 m
Monitoring the waveforms provides a visible indication of any distortion which might be
present.
The user is not inundated with vast quantities of data to be processed and interpreted.
Oscilloscopes and signal generators are ubiquitous items of equipment in any electronics
laboratory.
All the items of test equipment can be provided by designers with a limited budget.
Figure 7.4.2 displays the recorded data in the form of a three-column array and defines a
function Ytest(s) which can be used to calculate the admittance characteristic. This con-
stitutes the first page of the Mathcad worksheet which analyzes the results.
The range variable s is used to identify a particular row in the array. The relationship
between the voltage Vin induced in the conductor and the voltage at channel 2 of the
oscilloscope Vch2 is defined by equation (7.1.1). The admittance of the secondary winding
of the current transformer is calculated by using the set of equations defined in the subroutine
of Figure 7.2.8. Multiplying this value by the channel 2 voltage Vch2 gives the value of the
current Isec in the secondary winding. Multiplying Isec by the number of turns gives the
current in the conductor, Iprim.
The ratio of the current in the conductor to the applied voltage Vin gives a value for the
admittance at the spot frequency fs. All the values are recorded in the vector Yt. Figure 7.4.3
illustrates the frequency response of this parameter over the range 120 MHz.
As far as the predicted response was concerned, the shape of this curve was extremely
encouraging. There is a single peak at just below 10 MHz and an indication that there could
be a second peak at just below 30 MHz. This would relate to resonances at the quarter-wave
and three-quarter-wave frequency.
It could reasonably be expected that the dipole model of Figure 5.2.2 would be capable of
simulating the response. Such a task would be just a matter of assigning numerical values to
the components, defining the relevant equations, and setting these out in the worksheet. The
general circuit model is reproduced in Figure 7.4.4.
A brief examination of the tabulated data reveals that the quarter-wave frequency fq of the
conductor-under-review is 7.83 MHz. Given knowledge of his frequency and the measured
length of the cable, it is possible to use (2.3.10) to calculate the propagation velocity.
7.4 The isolated conductor 191
1 400 0.4
column 1: frequency, MHz
2 400 0.4 column 2: channel 1 voltage, mV
column 3: channel 2 voltage, mV
3 400 1
4 395 1.9 Yt is the admittance derived from test results
5 390 3.4
6 390 5.9
6.5 385 8.4 From Figure 7.2.7:
16 330 2.4 1
Z1 R4 +
j w C1
17 320 1.4
1 1 1 1 1
18 315 2.2 Y2 + + + +
R1 R2 R3 j w L1 Z1
19 310 3.4
Isec Y2 Vch2
19.6 305 4.2
Iprim Isec Turns
Iprim
Vin
Yts = Ytest(s)
Since there was no magnetic material in the cable, the value of mr can be assumed to be unity.
So (2.3.11) can be used to derive a value for the relative permittivity.
Equations (2.3.1) and (2.3.2) can then be used to calculate theoretical values for the
primitive capacitance Cp and primitive inductance Lp . Since (5.1.3) gives the theoretical
value for the radiation resistance Rrad, there is now enough information to assign values to
the components of the circuit model of Figure 7.4.4.
192 CHAPTER 7 Bench testing
A/V
0.02
Yt
0.01
0
0 5 106 1 107 1.5 107 2 107
f Hz
Lp Rp Lp Rp Vsource Lp Rp Lp Rp
2 2 2 2 2 2 2 2
Rsource
Cp
Cp I
Rrad
Figure 7.4.5 is a copy of the page of the worksheet that calculates the values of the
reactive and resistive parameters of the conductor. The frequency at which skin effect comes
into play is calculated to be 6.89 kHz. The only other parameters to be defined are the
radiation resistance and the amplitude of the voltage source. The former is set at the same
value as a conventional dipole, 73 W. The latter is set at unity to allow the response to be
defined as amperes per volt.
Assigning these values to the general circuit model leads to the model of Figure 7.4.6.
The Mathcad function to calculate the value of the admittance at each frequency is
illustrated in Figure 7.4.7. This is an adaptation of the function Zbranch(f) depicted in
Figure 4.3.4.
Distributed parameters Z1 and Z2 are determined, and then used to calculate the loop
impedance Z3. Dividing the amplitude of the voltage source Vsource by the magnitude of the
loop impedance gives a value for the current which will be flowing in the loop.
Since the magnitude of Vsource is unity, the output variable of the function Ymodel(i,
Vsource) is the admittance of the loop shown, at the frequency Fi. The input variable i is an
integer which points to the relevant row of the frequency vector F. The other input variable
7.4 The isolated conductor 193
mo mr l l
Lp := ln = 1.442 105 see (2.3.2)
2p r
Lp inductive component value
= 7.212 106
2 for model
2 p eo er l
Cp := Cp = 7.077 1011 see (2.3.1)
ln l
r
see (5.1.3)
Rrad := 73
r l steady-state resistance;
Rss := = 0.162
p r 2 see (2.5.11)
Vsource
1V
7.21 H 7.21 H 7.21 H 7.21 H
Fi
Ymodel(i, Vsource) := Rp Rss 1 +
Fx
w 2 p Fi
Figure 7.4.7 Calculating the admittance of the circuit model over a range of frequencies.
Vsource allows the response to be determined when the source voltage itself is a function of
frequency. (Figure 5.3.7 illustrates how the threat voltage of the incoming wave can vary.)
The vector Ym stores the results of the calculations over the same range of frequencies
that were used in the test procedure. Displaying the response of this model and those derived
from test results gives the two curves of Figure 7.4.8.
The close correspondence in the region between 1 and 5 MHz indicates that the value
calculated for the primitive capacitance Cp was fairly accurate. The fact that the peaks of
both models occur at the same frequency indicates that the value of Lp was also accurate.
However, there is an obvious discrepancy between the magnitudes of the two peaks.
Discovering the reason for this discrepancy involved several attempts to vary component
values of the model and note the effect this had on its response. At this point in the assess-
ment, it was assumed that the value of 73 W for Rrad was sacrosanct.
Eventually, it was found that increasing the amplitude of the voltage source to 1.65 V
brought the two curves into near-coincidence over the critical range between 5 and 10 MHz.
Figure 7.4.9 is a copy of the final page of the worksheet, and illustrates the resultant two
curves.
In hindsight, the explanation as to why the source voltage needed to be increased is fairly
simple. Unlike the conventional dipole antenna which is connected to the source termination
resistance of a transmitter or the load termination of a receiver, an isolated length of wire
contains no resistors to absorb the electromagnetic energy. So energy which is not radiated
out into the environment is stored as a voltage across the capacitors. This immediately
creates a reflected current which flows back through the transformers, and this reflected
current is a source of further radiation.
7.4 The isolated conductor 195
A/V
0.02
Yt
Ym
0.01
0
0 5 106 1 107 1.5 107 2 107
f, F Hz
Figure 7.4.8 Comparing response of the test with that of the initial circuit model.
Hence, the final circuit model turns out to be very similar to that of the initial model.
Figure 7.4.10 shows that the only change is the addition of a second voltage source to
represent the effect of stored energy.
Since the total energy is shared equally between current and voltage and since the current
has departed from the conductor, it follows that half the energy is stored in the capacitors.
Since voltage is proportional to the square root of power, it is reasonable to assume that the
maximum value for the voltage developed across the capacitors is:
p
Vstored 2 Vsource 0:71 Vsource 7:4:1
This value of 0.71 represents a system in which there are no other losses. So the observed
value of 0.65 is entirely plausible.
However, the addition of the source Vstored has caused the theoretical curve to be higher
than the actual response over the range 05 MHz. This deviation was corrected by removing
the extra voltage source and reducing the value of the radiation resistance to 42 W.
Figure 7.4.11 shows the resultant curve. This correlates well with the actual response in the
196 CHAPTER 7 Bench testing
0.02
Yt
Ym
0.01
0
0 5 106 1 107 1.5 107 2 107
f, F Hz
Vsource
1V
7.21 H 7.21 H 7.21 H 7.21 H
range 05 MHz, as well as replicating the amplitude of the peak response. This test can be
used to measure the radiation resistance of the conductor.
Deviations in the region between 10 and 20 MHz can be explained by the fact that, above
the quarter-wave frequency, current is flowing backwards and forwards along the cable,
rather like waves in a harbor. Simulating this effect would require extra complexity to be
introduced into the model.
7.5 Cable characterization 197
A/V
0.02
Yt
Ym
0.01
0
0 5 106 1 107 1.5 107 2 107
f, F Hz
However, as far as EMC is concerned, the critical frequencies are when the peaks are
high. So minor deviations between reality and model at other frequencies are of little concern
when assessing the probability that the equipment will pass the formal EMC tests.
The close correlation between the response of the actual system and the theoretical model
effectively validates the concepts introduced in the previous chapters. Specifically, these are:
the derivation of the formulae for the primitive inductance Lp in section 2.2,
the derivation of the formulae for the primitive capacitance Cp in section 2.1,
the distributed impedance circuit model, derived in section 4.1,
the circuit model for the dipole, postulated in section 5.2.
The close correlation also demonstrates that simple test equipment is capable of providing
extremely accurate measurements. For example, the fact that the response peaks at 7.83 MHz
means that, for this particular setup, the propagation velocity of antenna-mode current is 235
m/ms. (See the value for v in Figure 7.4.5.)
signal oscilloscope
generator channel 1 channel 2
50 50
V
I
7.5 m 7.5 m
Figure 7.5.1 illustrates the first setup. This configuration is essentially a practical
implementation of the general circuit model of an isolated cable, derived from theoretical
considerations in the section on the virtual conductor. Figure 5.2.8 illustrates this model and
(5.2.9) and (5.2.10) define formulae for component values. Since each half of the 15 m cable
is represented by a triple-T network, the length l of each network is 7.5 m.
A voltage transformer is used to inject a voltage in series with one conductor of the cable.
Both ends of the cable are open-circuit. If these terminations had both been short-circuited,
then it would be clear that the transformer was inducing a voltage in the loop formed by the
signal and return conductors (a differential-mode voltage). The existence of open-circuit
terminations does not change the fact that a differential-mode voltage is being injected.
This injected voltage creates a current that flows along the signal conductor. Electro-
magnetic coupling between the two conductors creates a current that flows in the opposite
direction along the return conductor (the differential-mode current). The injected voltage
also creates an antenna-mode current that flows along the cable and converts into an elec-
tromagnetic wave.
A current transformer clamped round both conductors was used to measure the amplitude
of the antenna-mode current. The range of frequencies was from 1 to 20 MHz and the output
displayed in Figure 7.5.2 is the transfer admittance YT, in amperes per volt. A worksheet
similar to that illustrated by Figure 7.4.2 was used to create this graph.
Figure 7.5.3 illustrates the second setup. Again, the voltage transformer injects a differ-
ential-mode voltage into the cable. But this time the monitored current is the differential-
mode current. In Figure 7.5.4 there are two peaks, at about 5.5 MHz and at about 16.8 MHz.
The general circuit model for antenna-mode coupling of a transmission line is derived in
the section on the virtual conductor, and illustrated by Figure 5.2.8. Rearranging this in the
form of a bridge circuit leads to Figure 7.5.5. This layout simplifies the definition of the two
current loops. In the configuration-under-review the voltage source is in series with con-
ductor 1; so this fact is reflected in the location of Vsource in the diagram.
Transforming the circuit components into distributed impedances leads to Figure 7.5.6.
Component values can be assigned by following a systematic process, starting with the
measurement of the conductor radii r11 and r22, the spacing between conductors, r12, and the
length l of the cable. Using equations available in previous chapters, initial values can be
7.5 Cable characterization 199
0.01
A/V
8 103
6 103
YTt
4 103
2 103
0
0 5 106 1 107 1.5 107 2 107
f1 Hz
signal oscilloscope
generator channel 1 channel 2
50 50
I
7.5 m 7.5 m
assigned to the resistive and inductive components. This is carried out by the page of the
worksheet illustrated by Figure 7.5.7.
Since the conductors of all cabling in electrical systems are sheathed or supported by
insulating material, and since this material has a significant effect on the propagation velo-
cities of the signals, it is necessary to determine the value of the relative permittivity of the
dielectric before values for the capacitors can be defined. This is done on the page of the
worksheet illustrated on Figure 7.5.8.
Having provided initial values for all the components of the model, the next step is to
simulate the responses of both tests. This is carried out by the main program, illustrated by
Figure 7.5.9. There are two outputs: the vector YTm, which simulates the response of the
radiated emission test, and Ym, which simulates wire-to-wire coupling.
Response characteristics of the model are compared with those of the actual tests in the
final page of the worksheet. These are displayed here as Figures 7.5.10 and 7.5.11.
200 CHAPTER 7 Bench testing
0.08
A/V
0.06
Yt 0.04
0.02
0
0 5 106 1 107 1.5 107 2 107
f Hz
Cc1 Cc1
I1
Lc3 Lc3
Cc3
2 Rrad 2 Cc3
Cc2
Cc2 I2
Vsource
Z2,1 I1 Z2,1
Rrad
Z2,3 Z1,3 Z1,3 Z2,3
Z2,2 Z2,2
I2
o r l r12
Lc1 := ln
2 r11
Calculating inductance values for the
o r l r12
Lc2 := ln circuit model and placing these values
2 r22 in a three-element vector
see (5.25)
o r l l
Lc3 := ln
2 r12
1.051 106
Lc
= 1.051 106 H Values of inductors of circuit model
2
6.191 106
0.375
Rss
= 0.375 Values of resistors of circuit model
2
0
Radiation resistance
Value selected during analysis
Rrad := 50 Initially set at 73
On the first run of the program, there were noticeable errors in the response of the model.
However,
varying the value of the conductor resistance Ra (see Figure 7.5.7) brought the amplitude
of Ym closer to that of Yt,
varying the value of the radiation resistor Rrad brought the amplitude of the peak of YTm
closer to that of YTt,
202 CHAPTER 7 Bench testing
c
2 Relative permittivity of dielectric of cable when it is
erb := acting as an aerial. see (2.3.11)
vb
Value of relative permittivity to be used to define
erb = 1.754
capacitor value for monopole
2p eo era l
Cc1 :=
r12
ln
r11
Calculating capacitance values for the circuit model
Cc2 := Cc1 and placing these values in a three-element vector.
see (5.2.10)
2 p eo erb l
Cc3 :=
l
ln
r12
9.291 1010
Cc = 9.291 1010 Values of capacitors of circuit model
8.867 1011
varying the value of the frequency fqa (see Figure 7.5.8) brought the peak of Ym closer to
that of Yt,
varying the value of fqb brought the peak of YTm closer to that of YTt.
Of course, varying any of the above parameters modified the responses of both curves. Even
so, these side effects were minimal and very few iterations of the program were needed to
achieve the correlation depicted in these two final graphs.
7.5 Cable characterization 203
Zbranch(s) := 2 Fs
for k 1.. 3
Fs
Rck Rssk 1+
Fx
(Rck + j Lck) j Cck
Equations for loop impedances, derived from inspection of see Figure 7.5.6
Zloop(s) := Z Zbranch(s)
Z11 2(Z1, 1 + Z2, 1 + Z1, 3 + Z2, 3) + Rrad
Z12 2(Z1, 3 + Z2, 3) Rrad
Z22 2(Z1, 2 + Z2, 2 + Z1, 3 + Z2, 3) + Rrad
Z11 Z12
Z12 Z22
1
V := V YTmi := Z Zloop(i) Calculating response of transfer
0 admittance for radiated emission
I lsolve(Z, V )
from the cable
I1 I2
Figure 7.5.9 The main program: calculating the response of the circuit model.
It is worth noting that although there are 25 different components in the circuit model, only
four independent variables were required to correlate the curves in both graphs. The component
values used in the final run of the program are recorded at the bottom of Figures 7.5.7 and 7.5.8.
Assigning these values to the variables of the general circuit model of Figure 7.5.5 leads to the
representative circuit model for the configuration-under-review (Figure 7.5.12).
There was no need to include a second voltage source in the model as was done for the
model of the single wire in section 7.4. All that needed to be done to line up the related peaks
was to reduce the value of the radiation resistance from 73 W to 50 W.
204 CHAPTER 7 Bench testing
0.01
A/V
8 103
6 103
YTt
YTm
4 103
2 103
0
0 5 106 1 107 1.5 107 2 107
f 1, F Hz
0.08
A/V
0.06
Yt
0.04
Ym
0.02
0
0 5 106 1 107 1.5 107 2 107
f, F Hz
The ability to create a circuit model of such a cable leads to several advantages:
The basic feature of a circuit model of a length of cable is that component values are
proportional to length of the cable. Per-unit-length parameters can be derived and later
7.5 Cable characterization 205
Vsource
1.05 H 1.05 H 1.05 H 1.05 H
929 pF 929 pF
6.19 H 6.19 H
Rrad 929 pF
88.7 pF 88.7 pF
929 pF 50
used to define models for any length of cable used in the system-under-review. (It is
important that the cross section of the cable is uniform.)
Measurements can be carried out on long cables at relatively low frequencies and used to
predict the performance of shorter cable at higher frequencies.
Having determined a circuit model for a particular cable, the file can be stored in a
library, similar to that used by SPICE software for devices such as operational amplifiers
or Schmitt triggers.
Combining the models of the interface equipment with that of the cable will allow the
radiated emission of the system-under-review to be simulated.
Since the same model is valid no matter where the location of the voltage source, it is also
possible to simulate the response of the system to external interference.
It can be reasoned that the field pattern of antenna-mode propagation is almost totally
external to the cable and so includes more air than it does plastic insulation. The differential-
mode propagation is more confined to the plastic insulation, so travels slower.
However, this set of tests does more than provide a reason for the difference. It defines
the propagation velocities and provides an actual measurement of the relative permittivities.
From Figure 7.5.8:
Propagation of antenna-mode current, vb 227 m/ms
Propagation velocity of differential-mode current, va 170 m/ms
Relative permittivity of cable, acting as an antenna, erb 1.76
Relative permittivity of transmission line, era 3.12
It is also worth re-emphasising the fact that the value of the radiation resistor of Figure 7.5.12
was obtained by varying the value of Rrad in the model to line up the amplitude of the peak
values of Yt and Ym in Figure 7.5.11. Effectively, the value of Rrad was derived from the test
results. For any particular cable, the radiation resistance at the half-wave frequency is
assumed to be a constant.
From the worksheet of Figure 7.5.7:
measured value of radiation resistance, Rrad 50 W
206 CHAPTER 7 Bench testing
This particular model was derived from tests of a particular cable. Tests on another cable
would result in a different set of values to assign to the general model of Figure 5.2.8. The
significant feature of this approach is the existence of a general circuit model that will reduce
the time it takes for a representative model to be created for any particular assembly.
Since the cross section of the cable is constant, the capacitors, inductors, and resistors are
all proportional to length of the cable. Since the terminals are open-circuit, there are no
components at the interfaces to upset the measurements. This means that the representative
model of Figure 7.5.12 can be extrapolated to allow the performance of a longer or shorter
cable to be assessed.
The theory of antennae suggests that a value of 73 W should be assigned to the radiation
resistance. This particular test suggests that a value of 50 W would be more appropriate when
carrying out a worst-case analysis of radiation susceptibility. Sections 5.6 and 5.7 give more
information on this aspect. In the end, the choice of the value to use for Rrad in any system
analysis is a matter of engineering judgment.
signal
oscilloscope
generator
Vch1 Vch2
50 50
current
transformer
15 m of twin-core cable
46
46 signal conductor
I
4.7 Vin return conductor far end
near end
46.2
50 46.2
4.7
Vin
Vsource 50 Vch1
0.1
V
0
0.1
Vch1 0.2
0.3
0.4
0.5
0 1 106 2 106 3 106 4 106 5 106
t s
Figure 7.6.4 Simulated waveform of channel 1 input voltage. Setup as Figure 7.6.1.
208 CHAPTER 7 Bench testing
0.02
waveform of differential-mode current
V
Vch1 0
0.02
Figure 7.6.2 illustrates how the voltage monitored at channel 1 input terminals is related
to the voltage delivered to the cable. Inspection of this model gives:
50
Vch1 Vin 7:6:1
96:2
Figure 7.6.3 is a photograph of the actual waveform monitored by channel 1. The current
delivered to the line causes minor perturbations to the input voltage for about 2 ms after each
step change, after which Vin reaches its open-circuit value. The period for the square wave
was about 8 ms. That is, the oscillator frequency was set at about 125 kHz.
7.6 Cable transients 209
The program described in section 6.6 was used to create the waveform of Figure 7.6.4.
Input variables were set to simulate an initial falling edge T1 at 150 ns and a rising edge T2 at
4.1 ms (see Figure 6.6.2). Output variable was selected to be Vch1 (see Figure 6.6.4).
Figure 7.6.5 is a photograph of the waveform displayed by channel 2 of the oscilloscope.
The oscillations during the first 200 ns could be due to reflections emanating from the
ground conductors of the test equipment. They could also be due to cable or connector
discontinuities.
The program of section 6.6 was run again, with input variables set to simulate an initial
falling edge T1 at 20 ns and a sweep time T3 of 1 ms, and the output variable was selected to
be Vdiff. Figure 7.6.6 was the result. The period between the first and second edges of the
waveform is twice the differential-mode propagation delay of the 15 m cable the time
taken for the reflected wave to return to its source.
The setup of Figure 7.6.1 was modified to route both conductors of the cable through the
core of the current transformer. Antenna-mode current was now being monitored.
Running the program of worksheet 6.6 a third time, with the time T1 of the initial rising
edge set at 120 ns, the time T2 of the falling edge set at 1.65 ms, and with Vrad selected as the
output variable, gave Figure 7.6.8. The rise and fall times of the simulation were set by the
value of dt. From the worksheet of Figure 6.6.2:
T 83109
dt 830 ps
N 100
Comparing Figure 7.6.5 with Figure 7.6.6 reveals that the simulated waveform of the dif-
ferential-mode current Idiff is very similar to the actual waveform. The most striking aspect
is the fact that the waveform gradually changes from a square wave to a sine wave.
There are some deviations, however:
The leading edge of the step on the observed signal is not quite as fast as the theoretical
step. This is almost certainly due to the finite response times of the signal generator and
oscilloscope.
There is a ripple on the actual waveform which lasts for about 200 ns. This is probably
due to reflected current returning from the conductors of the structure.
The similarity between Figure 7.6.4 and Figure 7.6.3 is just as significant. These compare the
monitored voltage at channel 1 with the simulated voltage Vch1. The ripple during the first
2 ms following each step change is due to voltage drop in the source resistance Rg caused by
current delivered to the cable.
Most notable is the fact that the corners of the leading and trailing edges are rounded. If
the loading effect had been due solely to the current waveform displayed on Figure 7.6.6,
then the corners would have been much sharper. The fact that they are rounded is a clear
demonstration that the source resistance is carrying antenna charging current as well as
differential-mode current.
There is one significant difference between Figure 7.6.3 and Figure 7.6.4. There is a
sharp spike after each step change of the monitored waveform. This is due to capacitive
coupling between the output of the signal generator and the input to channel 1 of the scope.
The two terminals connecting co-axial inner conductors to the 46 W resistors shown in
Figure 7.6.1 were very close to each other. However, the existence of this spike does not
obscure the fact that the ends of each step change are rounded.
210 CHAPTER 7 Bench testing
5 103
Vch1 0
5 103
Figures 7.6.7 and 7.6.8 also display significant similarities. These compare the waveform
of the antenna-mode current carried by the cable with the waveform generated by the
simulation. It is quite clear that the waveform resulting from the simulation follows the same
basic pattern as that observed on the oscilloscope, that is, a damped oscillation that is a
miniature version of Idiff.
However, there is also a higher frequency component superimposed on top of the damped
oscillation. This lasts for many more cycles than the underlying waveform of Figure 7.6.8,
indicating that the damping it experiences is insignificant. The probable cause is a current
component that is oscillating backwards and forwards between cable and structure. This
component is detectable as a ripple superimposed on the differential-mode waveform of
Figure 7.6.5.
7.6 Cable transients 211
The fact that the test results and the simulated waveforms are closely correlated provides
a fair degree of confidence in the reliability of the model. The waveform of figure 7.6.6 is
certainly more representative of reality than that of Figure 6.2.6. A point worth remembering
is that, although the capacitors and inductors of that simple model are replaced by the
parameters of propagation time and characteristic impedance, these properties continue to
exist. (Section 6.3 covers this aspect in more detail.)
Since the delay-line model has introduced the parameter of time into the computations, it
becomes possible to think in terms of the energy stored in the line. That is, the static energy
of electric charge and the kinetic energy of moving charges. In fact, (6.1.5) for an RLC
circuit is very similar to the equation of motion of a mass on a spring.
The inclusion of the network of Crad, Ro, and Inf of Figure 6.6.1 enables the effect of
energy storage caused by antenna-mode current to be separated from the effect of storage of
differential-mode energy in the delay line. When a step voltage is applied to the open-circuit
line, the differential-mode current delivered to that line is constant for a period equal to twice
the propagation time, and so is the antenna-mode current. After that period, the capacitor
Crad has been charged up and contains a lot of potential energy. The initial flow of current
into Crad represents the kinetic energy required to deliver that charge.
The fact that the current waveform reverts rapidly from a square wave to a sinusoidal wave
at the natural frequency of the assembly shows that the transient energy of the initial step has
been converted into the dynamic energy of an L-C circuit. From Figure 7.6.5, the time for one
cycle of this waveform is about 0.34 ms, corresponding to about 2.9 MHz. That is, a frequency
that is close to the quarter-wave frequency of a 15 m cable acting as a monopole antenna.
Figure 7.5.2 shows that, at the frequency of resonance (of a 15 m dipole), the radiation emitted
by the assembly reaches a peak value. Since the amplitude of this sinusoidal wave of Fig-
ure 7.6.5 decays rapidly, it is reasonable to assume that this loss of energy is mostly due to
radiated emission.
The inclusion of Ro3 in the model of Figure 6.6.1 allows the amplitude of the antenna-
mode current itself to be evaluated.
The setup of Figure 7.6.1 offers another way of deriving values for cable parameters,
simply by using data from the transient response. Since the voltage and current are almost
constant during the first 166 ms, the value of the characteristic impedance Ro can be calcu-
lated quite quickly. Since the transit time is also known, the loop inductance La and loop
capacitance Ca of the 15 m cable can also be derived, as illustrated in the worksheet of
Figure 7.6.9. The worksheet also derives values for these reactive components, using data
from the frequency response tests on the same cable.
Reasonably close correlation between these results shows that it is possible to obtain a
first estimate of the values of both the inductance and the capacitance of a cable, just by
using data from a couple of transient waveforms.
212 CHAPTER 7 Bench testing
Worksheet 7.6.4
From the waveform of Figure 7.6.2, the amplitude of the step voltage measured by channel 1, at
the end of the first exponential rise, is:
Vch1 := 0.39 V
The voltage applied to the input terminals of the line is related to Vch1 by Equation (7.6.1)
96.2
Vin = Vch1 = 0.75 V
50
From the waveform of Figure 7.47, the voltage measured by channel 2 is:
Vch2 := 0.016 V
The ratio of the amplitude of Vch2 to the amplitude of the current monitored by the current
transformer is given by Equation (7.2.6)
RT := 2.27
Vch2
Invoking Equation (7.2.5) Iout = = 7.048 103 A
RT
The ratio of Vin to Iout gives a measure of the characteristic impedance of the cable:
Vin
Ro := = 106
Iout
The waveform of Figure 7.6.5 also gives a measure of the time taken for a current step to
propagate to the far end and return to the near end. This is the time difference between the
first and second edges of the waveform. This gives:
166 109
T := sec
2
The capacitance and inductance of the cable can be determined by invoking Equations
(6.3.7) and (6.3.6)
T
Ca := = 7.8 1010 F
Ro
La := T Ro = 8.8 106 H
These results can be compared with those obtained from cable characterisation tests.
From Figure 7.5.12
1
Ca_ := 2 929 pF = 9.29 1010 F
2
La_ := 81.05 H = 8.4 106 H
Figure 7.6.9 Calculating values of components derived for 15 m cable, using data from
transient tests and frequency response tests.
It is useful to summarize the reasoning involved in the transient analysis and to identify
its significance in the design of electronic equipment.
Any step voltage between two terminals of a cable will cause a transient wavefront to
propagate along the cable at a velocity somewhat less than the speed of light. During the
transit time, some of the energy will be radiated into the environment and some will be
stored temporarily in the form of static charge.
The static charge between cable and environment will disappear in the form of a high-
frequency signal which rapidly decays in amplitude (see Figure 7.6.7). This means that both
the static energy in the charge and the dynamic energy used to create that charge are con-
verted into radiated interference.
7.7 Capacitor characterization 213
The charge between the conductors will eventually settle down to a steady-state value, but
not before some of the energy is converted into differential-mode radiation (see Figure 7.6.5).
This will happen with every step change in voltage. Since any signal is effectively a
continuous series of step changes, it is inevitable that a significant portion of the signal on a
twin-conductor cable will radiate away in the form of EMI.
If the separation between send and return conductors is uncontrolled, then most of the
transient energy will be converted into unwanted radiation. This is very useful for covert
surveillance, but not much else.
The only way of minimizing the amount of electromagnetic energy in a transmission line
which is lost to the environment is to use a co-axial cable, a screened pair, or a waveguide. If
an unscreened twin-conductor cable is used, the only way of reducing the level of radiated
energy is to absorb it in a resistive component. Sections 8.5 and 8.6 describe a range of
design techniques that enable this objective to be achieved.
50 50
co-axial cable co-axial cable
test jig
signal signal
generator monitor
Rn
50
C1
Vin Rf
L1 I2 Vout
I1 50
R1
between source and jig is effectively transparent. The same applies to the output link, that
between jig and signal monitor.
Since all capacitors possess inductive and resistive properties, the simplest model for this
particular component-under-test is the series LCR circuit. For this type of test, the relevant
parameter is the transfer impedance.
Vout
ZT 7:7:1
I1
So the test procedure would be to apply a constant-voltage variable frequency to the test jig
and note the amplitude of the output signal at a set of spot frequencies. This gives the
frequency response of the transfer impedance ZTt, that is, the transfer impedance derived
from test results.
It is a simple matter to write a program to calculate the frequency response of the transfer
impedance ZTm of the circuit model. Both responses can then be plotted on the same graph.
Initial guess values are assigned to C1, L1, and R1, and these parameters are then treated as
input variables. Adjusting C1 will bring the negative slopes of both curves into close cor-
100
10
ZTm
1
ZTt
0.1
0.01
1 105 1 106 1 107 1 108 1 109
F, f Hz
Rn
50
55 nF
Vin 850 pH Rf
I1 I2 50 Vout
50 m-ohm
Figure 7.7.5 Calculating frequency response of transfer impedance from test data.
Rn := 50 Rf := 50
1
i := 2..10000 Fi := i 105 V := V
0
ZTmi := w 2 p Fi
1
Z11 Rn + R1 + + jw L1
j w C1
1
Z12 R1 + + jw L1
j wC1
1
Z22 Rf + R1 + + jw L1
j wC1
Z11 Z12
Z
Z12 Z22
I lsolve(Z, V )
Vout I2Rf
Iin I1
Vout
Iin
relation, adjusting L1 will do the same for the positive slopes, and adjusting R1 will align the
minimum values.
When the two curves are as closely correlated as possible, the new values of C1, L1, and
R1 can then be assigned to the relevant component in the general circuit model. This gives
the representative circuit model for the capacitor-under-test. This analytical process was
carried out using data from a test on a 60 nF capacitor, one of many tests carried out by Roy
Ediss [7.1].
Figure 7.7.3 illustrates the final graph and Figure 7.7.4 shows the representative
circuit model. A copy of the Mathcad worksheet used to create the graph is provided by
Figures 7.7.5 and 7.7.6.
CHAPTER 8
Practical design
There are a number of design concepts and techniques in existence which are aimed at
improving EMC. Many of them are supported by the analytical approach. But some are not.
A concept that can be traced back for many decades is that of the equipotential ground
plane. There is no such thing. Since the method of images of electromagnetic theory is
often quoted as the basis for this concept, it is useful to identify the fallacies underlying the
correlation. Section 8.1 does just that. Designers should resist the temptation to treat
the structure as a convenient return path for signals and power. The most effective use of the
conducting structure is as a shield.
One deduction which is continually reinforced during the process of creating circuit
models is the fact that the signal and return conductors should be as close together as
possible.
Section 8.2 shows that this requirement is implemented to good effect in the design of
printed circuit boards by routing the signal tracks as close to the ground plane as possible. It
also shows that, for signals between boards and between equipment, this requirement is best
achieved by allocating a return conductor to every signal or power conductor.
In terms of EMC, one of the most counterproductive concepts is that of the single-
point ground. Section 8.3 illustrates how this technique guarantees the creation of
intractable problems. In contrast, the existence of a multitude of ground loops causes a
significant reduction in the level of interference.
Improvements achieved by the inclusion of dedicated return conductors can be nullified
by inappropriate design of the interface circuitry of equipment units. Section 8.4 identifies
the need for current balance in the interconnecting cables, and illustrates a few methods of
achieving this objective.
Section 8.5 focuses on the fact that optimum transmission is achieved by terminating
transmission lines with a resistance equal in value to the characteristic impedance. If the
signal is transmitted efficiently, it follows that emissions due to interference are minimal.
Circuit modeling confirms that a configuration which exhibits low emission is less suscep-
tible to external interference.
Whether the common-mode loop is open-circuit at one end or short-circuited at both
ends, it is inevitable that there will be reflections. Electrical energy will be stored tem-
porarily in this loop and will quickly depart into the environment in the form of interference.
It is possible to absorb this energy and damp down the oscillations. Section 8.6 describes
ways of doing this.
217
218 CHAPTER 8 Practical design
Section 8.7 deals briefly with the subject of equipment shielding and identifies the basic
requirements. Measures used to protect buildings and equipment from the indirect effects of
lightning are also described.
Every technique described in this chapter defines the interface circuitry at both ends of
the signal link. This provides visibility of both the signal loop and the common-mode loop.
Without such visibility, there is no point in even trying to assess the interference coupling
characteristics, let alone analyze them.
By defining the interface circuits at both ends of the link, an important feature of printed
circuit board design is identified; the interface circuits provide a buffer between the internal
wiring of each board and the cabling used to carry signals between boards. This means that
the interface circuits can be designed to handle the higher levels of interference which are
bound to exist on the longer conductors.
The length of the signal link is an important parameter. The longer the link, the lower is
the frequency at which interference reaches its highest peak. Since the power available is
inversely proportional to the square of the frequency, the longest links deliver the highest
level of interference energy. The buffer circuits on the printed circuit boards should be able
to absorb whatever unwanted energy arrives via the signal link.
8.1 Grounding
Reliance on the use of the conducting structure as the universal return path for all signals and
all supplies is probably the most prevalent cause of EMC problems. This could be due to the
widespread belief in the existence of the equipotential ground plane. There is no such thing.
Ground planes are an extremely useful design feature of printed circuit boards and
integrated circuits. But this does not mean that they are equipotential surfaces. Nor does
it mean that a conductor designated as ground or earth is automatically a zero voltage
reference point for all signals in the system.
The concept of a surface at which all points are at zero voltage probably arises from a
misunderstanding of the method of images used in electromagnetic field theory. So it is a
useful exercise to review the reasoning behind the technique.
The image solution method is usually illustrated by considering an infinite length of line
charge parallel to a grounded conductor of infinite length [3.1]. It is possible to simulate
the electric field distribution by replacing the ground plane with an insulating surface and
placing an image conductor the same distance below the plane as the actual conductor is
above the surface.
Figure 8.1.1 illustrates a configuration where two conductors are routed over a plane
surface. A steady voltage is applied between the near end of conductor 1 and the near end of
the ground plane (conductor 3). The far end of conductor 1 is open-circuit. Conductor 2 is
short-circuited to ground at both ends. Contour lines, depicting intermediate voltages, are
also shown on the figure.
Since the voltage applied between conductor 1 and the ground plane is constant, no
current is flowing. Under these circumstances, the ground plane is indeed an equipotential
surface. But with no current flowing there can be no signal and no interference. It is also
useful to note that the contour lines do not intersect conductor 2. The field pattern is
asymmetrical.
8.1 Grounding 219
conductor 1
10 V 4V equipotential contours
6V
8V conductor 2
2V 0V
If, now, the far end of conductor 1 is short-circuited to the ground plane and a current at
constant rate of change is caused to flow in that conductor, returning via the ground plane,
then, although the actual voltages would be different, the pattern of contour lines would
remain exactly the same. However, there would be voltage difference between each cross
section.
If the method of images had been used to create this field pattern, it would have been
assumed that the image of conductor 1 had been carrying exactly the same current. This
means that the voltage developed along conductor 1 would be balanced by a voltage along
the image conductor. Since this image conductor represents the effect of the plane, it follows
that the voltage developed along the ground would be the same magnitude as the voltage
along conductor 1.
The coupling mechanism can be analyzed by assuming that a low-frequency sinusoidal
waveform is applied to the near-end terminals of conductors 1 and 3, and that the short-
circuit link between the near end of conductor 2 and ground is replaced by a voltmeter.
Figure 8.1.2 is a circuit diagram depicting the inductive properties of the three conductors. A
sinusoidal voltage V2 would be developed between the near-end terminals of conductor 2
and ground. Since the impedance of a voltmeter is high, then minimal current would flow in
conductor 2, and V2 would be a measure of the voltage along L3, the inductance of the
ground plane.
If a voltage is developed along the ground plane, then that conductor cannot possibly be
regarded as an equipotential surface.
V1 L1
ground plane
L3
V2
conductor 2
L2
If the frequency of the voltage source is increased, capacitive coupling becomes more
significant. Also, proximity effects will cause the return current in conductor 3 to flow through a
restricted cross section. The resistance presented by the ground will also increase. The end
result is that the simple model of Figure 8.1.2 can best be modeled by the triple-T network of
Figure 2.7.6. The ground plane can be a source of both electric and magnetic field coupling.
Acceptance of the concept of the equipotential ground has inevitably led to further
misconceptions. Notable among these is the belief that all connections made to the ground
conductor are automatically held at zero voltage.
In circuit diagrams, it is often desired to define that a connection exists between two
separate points without the need to draw a line between those points. Normal convention is to
identify those points with a unique symbol. This avoids the use of tramlines crisscrossing the
diagram, and allows the reader to focus attention on the functional purpose of the circuit. The
most frequently used symbols are for ground or earth. From a functional point of view, it is
assumed that any voltage difference between such points can only have a marginal effect on
circuit function. It is then assumed that, for all practical purposes, there is no voltage difference.
However, such an assumption is only valid if the area enclosed by the relevant circuit
loop is extremely small. This means that there are few problems when the convention is used
with dual-layer or multilayer boards. In these assemblies, the traces used to link components
are very close to the ground plane. Hence, the area enclosed by the differential-mode loop is
always very small. The ground plane provides a return path for the current, just where it is
needed. Section 8.3 provides more information on this aspect of design.
But that is not the case with single-sided boards. When such a board is used, great care is
needed to ensure that the loop area involved is as small as possible. To do this, it is necessary
to use solid lines to depict conductors on the board. Symbols indicating the presence of
hidden connecting links have no place on such a diagram.
If it transpires that the conductive trace linking two points on a single-sided board follows
a long circuitous route, then there is an increased probability that the signal in question will
be prone to interference problems. If the designer does not have visibility of this track, then
the existence of an interference problem will come as a surprise.
At the time when electronic circuitry was based on the use of valves, the components
were usually mounted on an aluminum chassis and the circuit diagram depicted this con-
ductor as a solid line at the bottom of the page. The power supply rail was also included at
the top of the page, identifying the R-C networks used to minimize the interaction between
each stage of the supply from the next. Visibility of every circuit path was maintained. This
allowed the designer to minimize the area of every circuit loop, whether that loop carried
power supply current, signal current, or both. Care was taken to ensure that high current in
the output stages did not affect the sensitive circuitry at the front end.
If this disciplined approach was utilized during the design of single-sided printed circuit
boards, then a significant improvement in the EMC of that particular circuit function could
be achieved. In the case of single-sided boards, such an approach might be desirable. In cases
where the connecting links are between boards of an equipment unit, this approach would be
extremely important. In cases where the link is between equipment units, such an approach is
essential.
Assumption of the existence of an equipotential conductor is evident in every circuit
diagram which includes the ground symbol or any of its variants. This may be reasonable
if the purpose is to describe the function of the circuit, but totally counterproductive if the
8.2 Conductor pairing 221
objective is to analyze EMC. The existence of these symbols means that the return path for
every signal current is undefined. The same applies to power supply currents.
If half the circuitry is undefined at the initial design stage, then subsequent stages such as
detailed drawing, component manufacture, wiring, and assembly are left to the discretion of
whoever is carrying out the task in question. The EMC of the system is effectively out of
control. By the time the completed system is submitted for EMC testing, the only thing to do
is cross ones fingers and hope for the best.
If the physical relationships between the signal and return conductors are defined at the
initial stage of the project, then the designer is able to use all the analytical tools described in
the preceding chapters and all the techniques described in the following sections. Visibility
of this relationship can be maintained at all subsequent stages of development.
It can be concluded that the conductors of the structure should not be used as a con-
venient return path for signals and power. Any current flowing along the structure will create
a voltage along that length and any voltage between different locations on the structure is a
source of interference.
Up to now, the focus of this section has been on what the structure should not be used for.
To identify just what it should be used for, it is necessary to go back to Figure 8.1.1. Here, it
shows that neither the equipotential contours of an electric field nor the magnetic lines of
force of a magnetic field penetrate the ground plane. Any circuitry on the underside of this
barrier is shielded from the worst effects of the electromagnetic field above. From the point
of view of EMC, this is a desirable design feature.
Another significant aspect of the magnetic field distribution emerges. The lines of force
do not link with conductor 2. This is because a loop has been formed between this conductor
and ground. Transformer action ensures that the current in this loop creates an opposing
magnetic field that precisely balances the field emanating from conductor 1. That is, this
secondary loop tends to act as a shield that minimizes the penetration of the magnetic field to
the right of conductor 2.
This action is put to good use in the design of Faraday cages and the assembly of
conductors used to protect buildings from the worst effects of lightning. So, it can also be
reasoned that, even where the conductors of the structure do not present an impenetrable
surface, they still perform a very useful shielding function.
If the structure is designed to form a lattice network of conducting loops, then any inter-
ference will create a circulating current in each loop, and the field created by these circulating
currents will tend to balance the incoming interference. Such a structure is behaving as shield
which radiates unwanted electromagnetic fields back into the environment.
reason for this is not immediately apparent, it is useful to review the behavior of the ground
plane in a little more detail.
Figure 8.2.1 illustrates part of a printed circuit board assembly where two adjacent tracks
are connected to two voltage sources and the far ends are short-circuited to the ground plane.
Using the technique of composite conductors described in Chapter 3, it is possible to
determine the distribution of current in a cross section of the printed circuit board.
If it is assumed that the voltage sources apply step voltages of 1 V of opposite polarity
simultaneously to both tracks, then the rate of change of current in all three conductors can
be calculated using the process described in Chapter 3. The resultant distribution in all three
conductors is illustrated in Figure 8.2.2.
It is evident that the current in the left-hand track is positive and that in the right-hand
track is negative. Also, the rate of change of current in each track is unevenly distributed,
with the greatest rate of change occurring at the outer edges.
However, the most significant feature of the illustration is the fact that most of the current
in each track returns via the area of the ground plane that is immediately adjacent. Effec-
tively, the ground plane provides a return path for the current precisely where it is needed.
ground plane
V
V
A/s
1
current distribution in
left-hand track current distribution in
0.5 ground plane
0.5
current distribution in
right-hand track
1
0 2 4 6 8 10
mm
signal
receiver
signal driver
suppressor
signal conductor
conduit conduit
structure
signal
receiver
common-mode choke
signal driver
signal conductor
return conductor
structure
This effect is also present in systems where the supporting framework is constructed
of conducting material. In such a system, it is possible to route interconnecting cables along that
framework to utilize the structure as a return conductor. Figure 8.2.3 illustrates such a setup.
However, it is never possible to match the close coupling achieved on a printed circuit
board. It is inevitable that wide gaps will occur between cable and structure, and the area
enclosed by the current loop can become quite large. Stray coupling is increased. Reducing
the level of this unwanted coupling can only be achieved by minimizing the separation
between cable and structure. Adding cable trays or cable conduits as shown on Figur 8.2.3
will help toward this objective. Effectively, the structure is being brought closer to the cable.
Even so, the signal link is still quite susceptible to interference transients. The longer the
cable, the higher the level of interference. It becomes advisable to include suppressors at
the interface circuitry to prevent high-voltage transients from damaging the more sensitive
semiconductor devices.
Although suppressors can protect the equipment from damage, they cannot prevent
interference pulses from getting through to the processing circuitry. Incorrect messages can
be received. Data can be corrupted.
If, at the initial design stage, it is decided to allocate a return conductor to every signal
conductor, then this will create two separate loops, the differential-mode loop and the
common-mode loop. Analysis of this configuration is provided in section 2.8, where it is
concluded that the signal and return conductors should be held as closely together as possible
along the entire route from driver to receiver. It is much easier to hold two conductors
together than to minimize the spacing between cable and structure.
Given this configuration, there are many ways of minimizing interference coupling.
For example, the suppressor can be replaced by a common-mode choke, as shown in Figure 8.2.4.
224 CHAPTER 8 Practical design
This significantly reduces the amplitude of interference pulses in the common-mode loop while
having minimal effect on performance.
Also, it is very helpful to reduce the area of the common-mode loop by routing the signal/
return pair very close to the structure. It generally reduces the antenna efficiency of that loop.
+15 V supply
single
point
ground
0 V 15 V
stabilized
supply
module
2 3
4
1
those braids are bonded at each end to the units they interconnect, then the entire system will
be shielded. Circuitry within the system will be protected from the worst effects of external
fields, and radiated emission will be attenuated by the shield barrier. In such a configuration,
there will be a multitude of ground loops.
It does not matter whether the link-under-review is on a printed circuit board, inter-
connects two boards, or interconnects two units of equipment, the existence of the ground
loop is an essential part of good EMC design.
In large systems, it is also good practice to route the cables as close to the conducting
structure as possible. Any high-amplitude transient in a conducting member of the structure,
226 CHAPTER 8 Practical design
such as those due to a lightning strike, will create an electromagnetic field that will couple
with every conductor routed alongside that member. This will induce essentially the same
voltage in series with every conductor of the cable. Differential voltages will be minimal,
and interference will be reduced.
That is, the output voltage of the amplifier in the signal receiver is a function of the voltage
difference between the two input terminals. Equally important is the relationship between the
signal and return current. Ideally, this is:
Isignal Ireturn 8:4:2
If this condition is achieved, then any voltage developed along the signal conductor is
precisely balanced by an equal and opposite voltage developed along the return conductor.
In theory, this means that the signal passed between the transmitter and the receiver is
immune to any interference signal appearing in the common-mode loop. That is, the source
Vcm has no effect on the transmitted signal.
In practice, this is not the case. Several factors are involved. All four resistors should be
the same value, within a close tolerance. To minimize reflections, the value chosen for R1
should be the same as the characteristic impedance of each conductor. At high frequencies,
the level of rejection also depends on the length of the two cable conductors being identical.
If the amplitude of the interference exceeds the linear range of the amplifier, signal distortion
will occur. Common-mode rejection is also limited by the bandwidth of the amplifier.
However, these considerations do not really diminish the usefulness of the differential
amplifier. Protection from common-mode interference is provided over a wide range of
frequencies. The useful bandwidth is from zero frequency to the upper frequency limit of the
amplifier, and some amplifiers have a very fast response indeed.
Such a configuration is amenable to analysis using the techniques described in the
previous chapters. Figure 2.7.6 illustrates how the connections between the two units can be
modeled.
ground Vexternal
It will be noted that the screen is connected to the shield of both units of equipment as
well as to the grounded conductors. Since this balance depends on the cable conductors being
of equal length and the values of the resistors matching precisely, there will always be some
coupling of common-mode voltage into the output signal.
+Vcc
R1
+ R1
+
R1
R1
Vcc
Vcm
+Vcc +Vcc
+ signal T1
+
return
Icm R1
structure
Since the transformer is transparent to differential-mode signals, its effect on the common-
mode loop can be modeled by inserting an inductance in series with the structure.
There will be a small amount of flux linking each conductor which does not link the
other. In the classical model of a transformer there would be a primary inductance and a
secondary inductance. It is important that a bifilar winding is used because this will mini-
mize the primary and secondary inductances. Also, the balancing action is enhanced by the
capacitive coupling.
Since the transformer behaves as an inductor in the common-mode loop, its effectiveness
will increase with frequency. However, this also means that there will be little or no com-
mon-mode rejection at the low-frequency end of the response curve.
At the upper end of the response, capacitive effects come into play, and these will
limit the level of rejection offered to unwanted signals. There is also a distinct possibility
that resonances will occur; a transformer acts as an open-circuit at high frequencies. Soft
ferrites which behave resistively can help to ameliorate this effect. Hence, it is a useful
exercise to identify the frequencies at which the common-mode rejection is minimal.
Section 4.3 helps with this task.
The common-mode choke can sometimes be useful in minimizing troublesome effects
after the equipment has been manufactured. Split-core ferrites are available, and these can be
used to clamp round cables.
+Vcc
low pass filter
ground
ground
Vcc Vexternal
The response to interference in the common-mode loop is very similar to the curve of
Figure 4.3.6. To avoid problems with high-frequency coupling, it is necessary to ensure that
the upper frequency of the signal is significantly less than the quarter-wave resonance of the
cable. Also, it is good idea to include a low-pass filter at the receiver to limit its bandwidth to
that of the expected signal.
8.4.7 Opto-isolator
With logic signals, isolation between the return conductor and local ground can be achieved
by using an opto-coupler, as illustrated in Figure 8.4.7.
Again, it is necessary to guard against high levels of interference at the quarter-wave
frequency of the cable.
8.5 Differential-mode damping 231
+Vcc
+Vcc
+
Vout
Vexternal ground
pcb1
120
+Vcc
stabilized
supply
module
Vtransient
pcb2
120
+Vcc
supply cable
Zo = 120
If there are no reflections, then there will be no ringing transients in the supply cable.
Removing these ringing pulses will remove a potential source of interference.
mains filter
switched
substation socket
C1
live R1
neutral
earth C2
R2
supply SW1
C2 C1
RL1 drive RL1
R2 R1
ground
RL1 return R3
structure
will be no reflection from the initial edge. If the value of the capacitor is selected such that
the current waveform is critically damped, then the waveshape of the current in the cable will
be a single pulse, decaying to a constant value. High-frequency ringing should be minimal.
More importantly, when the switch opens, most of the energy stored by current in the
inductance of the solenoid will be absorbed by R1. Most of the energy stored in the capa-
citance of the line will be absorbed by R2.
It is also a fact that when SW1 is closing, the contacts will bounce several times. The
input supply to the solenoid is not a clean step function. Rather, it is a series of pulses of
varying mark/space ratio. There is no way of avoiding the creation of some level of inter-
ference. However, the amplitude of the radiated field can be minimized by using a conductor
pair to carry the supply current and by using damping resistors at each end, as Figure 8.5.3
illustrates.
If the structure had been used to carry the return current, then the amplitude of the
radiated field would have been much larger.
+Vcc +Vcc
R2
+ signal
+
return
T1 R1
Icm
ground Vexternal ground
structure
there will be a voltage induced in this third winding proportional to the rate of change of this
current. The choke now really needs to be viewed as a transformer. Since the third winding
is loaded by the resistor R2, the effect is to insert a resistance in series with the common-
mode loop. If the third winding has the same number of turns as the bifilar winding, then the
value of the resistance inserted into the common-mode loop will be R2.
At all frequencies, this resistor is transparent to the signal current. But at the frequencies
for which the transformer is effective, the device acts to absorb the unwanted energy in
the common-mode loop. The value of the resistor is determined by the characteristic impe-
dance of that loop. Damping action is effective at all frequencies where the impedance of the
winding is greater than the resistance. Figure 7.5 illustrates the fact that the frequency range
can be quite wide.
One benefit from this technique is that the value of the common-mode resistor can now
be defined by the designer. The other is that it can be used at relatively low frequencies.
common-mode resistor
speaker
amplifier
microphone
mixer
clamp-on toroid structure
antenna-mode current. If the device were to be located near the microphone it would be
totally ineffective.
An alternative solution to such a problem would be to use a shorter length of microphone
cable to change its quarter-wave frequency. This was the solution adopted for a public
address system in a church, when it had been possible to hear faint voices coming from the
speakers during quiet times of an evening service.
+ R1
+
R2 R2 R1
ground ground
some cases even touching. Although an outer woven braid of insulating material will prevent
direct contact, electromagnetic coupling cannot be avoided.
If the setup of Figure 8.6.3 is now viewed as a victim of such stray coupling, then the
effect of this coupling can be represented by a source voltage Vcross_coupling in series with
the outer shield. This will cause a current to flow in the outer loop. Since the resistors R2
limit the amplitude of this current, the voltage developed along the length of the inner screen
will be extremely small.
This means that the interference caused by the coupling of signals between cables of the
test setup will be extremely small. The introduction of damping resistors into the screen/
shield loops represents an improvement on the level of shielding that can be provided by a
configuration where the outer shields are simply shorted to the connector shells.
inter-winding
substation live
neutral
earth
R1
structure structure
Vthreat1 Vthreat2
operate under conditions where the electromagnetic environment is extremely harsh, such as
when it is located near a high-power transmitter.
A fair level of protection is provided by the use of a differential-mode supply, the
installation of a low-pass filter at the igniter end of the line, and the shielding afforded by
the screen of the twisted pair. However, the firing line is still prone to resonate at some
frequencies, and at these frequencies, the protection can be compromised.
The purpose of the buffer box is to ensure that the high peaks due to resonance never
occur even in the most severe environmental conditions. Under such worst-case conditions, a
high threat voltage Vthreat1 appears in the common-mode loop between control unit and
buffer box. Since transformer T1 acts as an inductor to common-mode current, it limits the
common-mode current very significantly.
A threat voltage Vthreat2 also exists between the buffer box and the igniter. This creates
a current in the loop formed by screen and structure. Current in the screen develops a voltage
along its resistance, and this small voltage appears as a voltage source in the common-mode
loop formed by the screen and the firing line. Since the resistors in the buffer box match
the characteristic impedances of the conductors they terminate, there is no resonance. Also,
the resistor network acts as a bridge circuit to balance the current in the two conductors of the
firing line.
The result is a firing circuit that is safe to use in very hostile environments.
filter
motor
R1 R2
substation SW1 SW2 C1
L
N T1 T2
E
C2
other switched loads on the supply. One resistor absorbs differential-mode interference,
while the other minimizes the common-mode current.
Transformer T1 is trifilar wound, and T2 is bifilar wound. The resistor R1 is approxi-
mately equal in value to the characteristic impedance associated with the common-mode
loop, while R2 is the same as the characteristic impedance of the differential-mode loop. The
mutual inductance of each transformer is determined by the lowest frequency at which the
resistor is effective. Such a circuit can help to reduce the level of environmental pollution
caused by switching transients in heavy-duty loads.
It could be used to prevent the nuisance tripping of the residual current detector in the
mains switch box. Such tripping can occur at switch-on if there is a large value capacitance
between the live and earth conductors at the load end.
8.7 Shielding
Since every book on EMC carries a section on shielding, there is no need for an extensive
treatment of the subject in these pages. However, it is useful to apply the lessons learnt in the
previous chapters to review the shielding properties of the structure.
The action of any electromagnetic field on a conductor is to induce a current along that
conductor. If the frequency of the radiation is high, then most of this current will be con-
centrated on the surface of the conductor; the skin effect predominates.
With a dipole antenna, this current is routed to a receiver via a transmission line. If the
conductor is a square, flat surface such as an equipment panel, then the current travels to the
edge of the square. Since this represents a discontinuity in the current flow, some of it is
reflected back. Some of it continues to flow, down the edge and back along the opposite face
of the panel.
Any transient current flowing in a conductor will emit a radiated field. That is, the panel
acts as a repeater and radiates a significant proportion of the transient energy back into the
environment. Since current is flowing on both sides of the panel, radiation is in all directions.
As far as shielding is concerned, such a panel is highly inefficient.
If the shield were to be spherical, then there would be no direct path between the inner
and outer surfaces. At the outer surface of the conductor, part of any external field would be
reflected back into the environment, but some would be absorbed into the material. At the
inner surface of the conductor, some energy would be reflected back to the outer surface and
some would propagate into the volume enclosed by the sphere. The ratio between external
and internal power levels is described as the shielding effectiveness of the material.
8.7 Shielding 241
This shielding action is equally useful in preventing internally generated field from
radiating out into the environment. This means that a conducting surface can provide an
effective barrier to the propagation of electromagnetic energy at all frequencies, provided
that there are no gaps in the surface.
300 106
flimit 8:7:1
4Lgap
Below this frequency, it can be assumed that the shielding effectiveness is 20 dB, that is, a
voltage or current ratio of 10 to 1. Above this frequency, it is better to assume that the
shielding effectiveness is zero. When assessing the EMC of any particular signal link, the
approach could be to implement the analytical techniques described in Chapters 26, and
then to invoke the shielding factor as described above. This would give a worst-case figure.
In cases where the structure does not meet the rigorous guidelines of Figure 8.7.1, it still
provides some measure of shielding by virtue of the fact that there will be circulating
242 CHAPTER 8 Practical design
SHIELDED ENCLOSURE
Crimped seam
currents in the conducting loops which absorb the energy of the incoming radiation, in much
the same way that a breakwater protects boats inside a harbor. Chapter 3 describes a method
that has been successfully used by researchers to locate regions in aircraft structures where
the indirect effects of a lightning strike are least severe. This identifies routes along which
cables can safely be laid. Such an approach can apply to any structural assembly.
Carrying out tests of actual performance, as described in Chapter 7, would allow a more
accurate figure to be assigned to the shielding provided to any particular signal link.
Other techniques for calculating shield effectiveness can be found in Reference 1.5.
live signal
neutral return
earth lightning conductor
linking conductor earth
buried
earthing mesh
other conductor. So it can be expected that a proportion will be carried by the earth con-
ductor of the power line.
High transient current in the conductors inevitably creates high transient voltages
between these and any other conductors in the near vicinity, with the risk of arcing. This can
happen at ground level as well as at height.
To prevent damage to telecommunications cables, surge protection devices can be fitted
at the point of entry to the building. These are usually assemblies of metal oxide varistors.
At voltages below a pre-determined threshold, a varistor behaves like a pair of back-to-
back Zener diodes. If the threshold voltage is exceeded, the current increases rapidly, and
avalanche breakdown occurs. Some devices are capable of handling extremely high currents,
of the order of 40 kA, albeit for a short period, but long enough to survive a lightning current
surge. During a lightning transient, the reaction time is fast enough to prevent the voltage
transient from exceeding the voltage withstand capability of the equipment being protected.
Current into the telecommunications line is limited by the inductance and resistance
of the line when it is acting as an antenna. The cable acts to carry a small proportion of
lightning current away from the building, and, hopefully, continues to serve its normal
function after the event.
A similar surge protection device can be fitted at the point where the service power cable
enters the building. Although the earth conductor of the power cable will carry the larger
proportion of the current flowing into the cable, electromagnetic coupling will create a high
transient voltage between the earth and live conductors, and between the earth and neutral
conductors. When this happens, the components of the surge protection device change into
the conducting state. This has two effects; the live and neutral conductors now act to carry
some of the transient current away from the building, and the transient voltages between the
conductors are held to a level that the insulation can withstand.
244 CHAPTER 8 Practical design
If lightning strikes either the telecommunications cable or the power cable outside the
building, the relevant surge protection device will act to ensure that the electrical equipment
inside the building remains safe. A more detailed description of the method used to protect
buildings from the effects of lightning is provided by a manufacturer of surge protection
devices [8.3].
The measures illustrated in Figure 8.7.2 can provide protection from the worst effects of
lightning, and injury to humans, fire, and structural damage. Preventing damage to electrical
and electronic systems requires further measures, such as other surge suppression devices
between the main switchboard and the equipment. Lightning protection zones, such as
screened rooms, can also be installed.
It should be noted that surge protection devices can only provide protection from
damage. They cannot prevent interference from upsetting the normal functioning of the
equipment.
To prevent upset due to interference, one or more of the other techniques illustrated in
this chapter need to be included in the system design.
System design
Chapters 2 and 3 have used the concepts of electromagnetic theory to create circuit models
that allow electromagnetic coupling between adjacent circuits to be simulated using circuit
theory. The resultant simplification in the mathematics comes at a price. Action and reaction
anywhere in the system is assumed to be instantaneous. This means that the wavelength of
the maximum frequency of the simulation must be greater than 10 times the maximum
dimension of the assembly-under-review.
Chapter 4 shows how this limitation can be relaxed by invoking the concept of distributed
parameters. Even so, the limitation is that the wavelength of the maximum frequency of the
simulation must be greater than 10 times the maximum size of the cable section. However,
this means that the maximum frequency of the simulation is no longer limited by the length
of the cable.
Chapters 5 extends the application to analyze coupling between cable and environment.
Here, the simplifications are achieved by assuming worst-case analysis. This is acceptable when
analyzing EMC, where the criterion is to ensure that the actual interference is always less than
that predicted.
The transient analysis of Chapter 6 reveals many features of electromagnetic coupling
that are not generally known and which provide a much improved understanding of the
phenomena.
The development of the modeling process was made possible by the fact that many tests
were carried out on many cable assemblies. Chapter 7 describes some of them. Correlating
the results of each test with those of the model allowed defects in the model to be corrected.
This was a more unforgiving approach than one based purely on theory. Even so, every
equation is traceable back to formulae derived in textbook theory.
Chapter 8 has identified several techniques which can be used to minimize the level of
interference coupled into and out of a signal link. It is possible to glean a few simple
guidelines that identify the basic concepts used in all these techniques, and these are listed in
section 9.1. If implemented at the initial stage of a project, these guidelines should ensure
that the EMC of the system can be achieved in a cost-effective way.
In all the previous chapters, the approach has been to build up circuit models from basic
building blocks and to develop those models to deal with increasingly complex coupling
mechanisms. Having developed these models to a point where they provide a reasonably
accurate simulation of the performance of actual hardware, it is possible to reverse the
process and describe a top-down approach. This is the objective of section 9.2.
245
246 CHAPTER 9 System design
Section 9.3 identifies the function of the interface circuits as buffers between the signal
processing functions on printed circuit boards and the signal distribution function of the
wires and cables. Chapter 8 provides details of a wide range of such interface circuits.
Having established a clear set of relationships between currents and voltages in the signal
link and the electric and magnetic fields in the near vicinity, it is possible to extend those
relationships to include the EMC requirements. That is, the design of the system can be
tailored to meet those requirements. Section 9.4 establishes a relationship between the threat
environment and the level of interference experienced by the victim circuit. Section 9.5
relates the design of a potential source of interference to the maximum level of radiated
interference, as specified by the formal requirements.
One of the responsibilities of every engineer is to ensure adequate preparations are made
before embarking on a design project. Section 9.6 identifies those aspects of the planning
necessary to implement the approach described here. In particular, it recommends that
enough bench test equipment be provided to meet the needs of the project and that mathe-
matical software be installed on the personal computers. It goes on to describe methods of
identifying those signal links that could pose problems and of characterizing those links
during the initial development of the system. It concludes by stating that the best way of
dealing with an EMC problem is to create a circuit model and review it in the light of the
guidelines of section 9.1.
However, the floating configuration acts to amplify interference levels at the quarter-wave
frequency. In such configurations, it is necessary to ensure that the interface circuitry
does not respond to frequencies in this range. This can be done by the use of bandwidth
filtering.
If the single-point ground configuration is implemented, then the system will exhibit the
worst EMC characteristics possible. Section 8.3 identifies the essential difference between a
configuration that uses ground loops and one that is based on the concept of the single-point
ground.
+15 supply
+15 return
15 return
audio input + 15 supply
audio return
speaker drive
speaker return
zero volt
structure
No component values are shown in this particular figure because the diagram is for
illustrative purposes only. If component values had been defined, it would have been pos-
sible to analyze the performance of this module. Simulation Programs with Integrated Cir-
cuit Emphasis (SPICE) are designed specifically to deal with such a task. All that the
equipment designer needs to do is draw the circuit on a computer screen, define the com-
ponent values, add a voltage source to the input terminals, and invoke one of the analysis
options. Very significantly, SPICE software is capable of analyzing both transient behavior
and frequency response of the system.
Although SPICE can be used to simulate the behavior of extremely complex circuitry on
printed circuit boards, that facility comes at a price. The previous chapters have shown that
the analysis of interference coupling is based on the assumption that voltages exist between
reference or ground conductors. SPICE analyses of the functional performance of com-
plex assemblies are based on the assumption that all points connected to the reference con-
ductor are at zero volts. Given such an assumption, there is no point in attempting to use such
software to analyze interference coupling in complex assemblies.
Section 8.2 shows that although voltages exist along the surface of a ground plane, the
simultaneous existence of circulating currents allows that plane to be regarded as an equi-
potential surface. Hence, SPICE analysis of complex circuitry on a printed circuit board with
a ground plane can provide a reliable simulation of the behavior of that circuitry. This makes
the software a valuable analytical tool. The words integrated circuit emphasis indicate the
intended application of the software, the analysis of small sized modules.
In this particular example, the signals being processed are in the audio range, up to about
20 kHz. At this frequency the quarter-wavelength is more than 3.5 km. Since the largest
dimension on the printed circuit board is of the order of 10 cm (corresponding to a quarter-
wave frequency of 7.5 GHz), it is not likely that antenna-mode interference will have any
direct effect on the board-mounted circuitry.
Although it is possible for high levels of radio frequency (RF) interference to arrive via
the input (or output) connectors, the design of the interface circuitry should be such that all
signals outside the operating range of the amplifier are severely attenuated. This is probably
the most important precaution.
It is worth noting that non-linearity in the relationship between voltage and current in
transistor and diode junctions can cause the modulation of an RF signal to be picked up. The
9.2 Relating the diagrams 251
same is true if the center frequency of the RF signal lies on the sloping part of the frequency
response curve of a filter. Care should be taken to select transistors and diodes that are not
capable of responding to high-frequency RF signals outside the bandwidth of the signal
being processed.
It is also advisable to keep the signal loops on the board as small as possible, either by
using a ground plane or by routing the tracks carrying each signal and its return current
closely together.
Given such precautions, it is reasonable to assume that there will be no internal inter-
ference between components on the board. By ignoring effects that are judged to be irrele-
vant, the computational power of SPICE software can be put to good use in analyzing the
behavior of extremely complex circuit boards.
Even though the main purpose of SPICE analysis is to simulate system function, it is still
possible to assess and analyze the interference coupling characteristics of critical signal links
on the board by reviewing them in the light of the guidelines of section 9.1.
power
microphone pre- speaker
amplifier
amplifier
+15 V
supply
15 V
supply
audio
amplifier
filter
live
neutral
earth
illustrates the relationship. That is, the longer the cable, the higher the voltage which can be
induced.
Figure 5.3.7 also illustrates the fact that the frequency at which maximum voltage can be
induced reduces as the length of the cable increases. For a long cable, the design of the
interface circuitry should ensure that the signal passed on the processing circuitry due to the
threat voltage is within acceptable limits. This design would almost certainly include a low-
pass filter, ensuring that the level of attenuation increases as the frequency increases.
This being so, the focus of attention shifts to the design of the interconnecting cables.
Any external electromagnetic field will link with the cables and will cause antenna-mode
current to flow in the outer surface of every exposed conductor. Equally, any antenna-mode
current created by differential-mode current in the cables will radiate interference away from
the system.
For such a system, the assumption that every point on a conductor is at the same voltage
is no longer valid. It becomes necessary to represent the coupling to and from the signal loop
as a triple-T network, as illustrated by Figure 2.7.6.
audio
microphone pre-amp
amplifier
power
+
+
common-mode
loop
structure
Figure 9.2.4 Interface diagram of link between pre-amplifier and power amplifier.
interface diagram for each signal link, the same block diagram can be used to focus attention
on the signal links, that is, on the lines on the block diagram, rather than the boxes.
The interface diagram identifies the cable conductors, the connector terminals, and the
circuitry at each end of the cable, and indicates how the structure interconnects the ground
reference conductors at each end. It identifies the common-mode loop and the differential-
mode loop. It defines precisely how the transmitter circuitry interfaces with the cable and
precisely how the buffer at the receiving end is designed.
This diagram illustrates clearly that three conductors are involved in carrying a signal
from one location in the system to another. In this case, it is the inner conductor of the
co-axial cable, the screen of that cable, and the structure. Differential-mode current is carried
by loop formed by the signal and return conductors (in this case, the core and screen of the
co-axial cable); common-mode current is carried by the loop formed by the return conductor
and structure. In this context, the structure represents the combined effect of all other con-
ductors linking the two equipment units.
This diagram is arguably the most important from the point of view of EMC design. The
designer is able to define what sort of cable to use wire pair, co-axial cable, screened pair,
multiconductor assembly, or something else. He or she is free to use any of the interface
circuits described in Chapter 8, or to define a completely new interface. It is also possible to
decide whether or not to enclose the cable with an overall screen or to route it along conduits.
Anyone viewing the interface diagram is provided with a complete picture of the features
of that link which are relevant to EMC. If the information is contained in several drawings,
each with a generous sprinkling of earth symbols, it is rather like peering into the mist.
L1 L1
2 co-axial inner 2
Vsignal C1 Vout
screen
L2 C2 L2
2 2
C3
L3 structure
L3
2 Vthreat 2
Figure 9.2.5 Circuit model of link between pre-amplifier and power amplifier.
A circuit model for a signal link can be defined as one that simulates the interference
coupling mechanisms. It allows both the susceptibility and emission characteristics to be
analyzed.
In many cases, the mere act of creating a circuit model allows the characteristics of the
signal link to be assessed. It is possible to view such a model in much the same way that an
experienced designer can look at the circuit diagram of a functional module and identify its
strengths and weaknesses.
In the case of the link-under-review, the only coupling between differential-mode and
common-mode loops is via the transfer impedance of the screen. It would be useful to know
the frequency response characteristic of this impedance. If this information is available, then
the component values for the circuit model can be deduced. Section 2.10 gives details on the
method of modeling co-axial cables.
There may be problems when resonance conditions arise and the current in the screen
reaches a high amplitude. If, as in this example, the bandwidth of the differential-mode
signal is restricted to audio frequencies, then there should be no problems. There should be a
low-pass filter at the audio amplifier interface to ensure that any high-frequency content is
rejected.
co-axial cable
Lp1
Lp4
common-mode Lp2
loop
mains supply cable mains supply cable
to pre-amplifier to audio amplifier
Lp3
mains trunking
Since the value of capacitance C3 is closely related to the inductance L3, this parameter
can also be defined. Section 2.3, on the duality between L and C, provides the relevant
formulae.
With properly terminated co-axial cable, the impedance common to the common-mode
and differential-mode loops is the transfer impedance of the screen. As indicated previously,
the model can be created if information on the frequency response of this parameter is
available. Such information could be provided by manufacturers data sheets.
Given knowledge of the diameter of the two conductors of the co-axial cable, the
inductance per meter of the differential-mode loop, L1, can be defined. Since the char-
acteristic impedance of the cable is (to a first approximation) equal to the square root of the
ratio between inductance and capacitance, it is possible to derive a value for C1 in the circuit
model.
If the signal link is more complex (for example, if the conductors form part of a multi-
conductor cable with an overall screen), then the technique described in Chapter 3 can be
used to determine the coupling parameters.
Alternatively, data from a set of bench tests can be used to construct the model. This
could be provided by tests carried out on the prototype equipment, or by data from a library
of representative circuit models as described in section 9.6.6.
Values for the components at each interface are either defined by the designer or speci-
fied in the relevant drawings of each equipment unit.
The concept of segregation of noisy and quiet circuits is often implemented on printed
circuit boards. If one section of the board is used to mount logic circuitry and another section
holds analog components, then it is not a good idea to mount each set of components on its
own separate ground plane. The ground plane should cover the entire board, since it provides
a return path for each signal trace at precisely the right location. Any separation of ground
planes on a board will disrupt the paths for return currents, with an increased probability of
interference.
Since the wiring on printed circuit boards is much shorter than that used externally, the
frequencies at which problems occur will be much higher. Since the threat level reduces with
frequency, interference from distant sources is unlikely (unless the board is completely
exposed to the environment). This means that interference problems at very high frequencies
are more likely to be due to emanations from nearby boards. Sniffers are available to detect
sources of high-level emissions [9.1].
Sections 8.48.6 are devoted to describing ways of designing the interface circuitry. The
most important consideration is that the design of the circuitry at both ends of the cable
assembly should be considered at the same time. Books on circuit design [9.29.4] can
provide guidance on the design of the functional circuitry, while the book EMC Design
Techniques [1.10] has an excellent chapter on printed circuit boards.
10 kHz100 kHz 10
100 kHz500 kHz 10
500 kHz1.6 MHz 10
1.6 MHz5 MHz 560
5 MHz10 MHz 380
10 MHz30 MHz 200
30 MHz100 MHz 200
100 MHz200 MHz 60
200 MHz700 MHz 70
700 MHz1 GHz 60
258 CHAPTER 9 System design
1 103
V/m
100
E(i)
10
1
1 105 1 106 1 107 1 108
Fi Hz
Figure 9.4.1 Example of military threat environment, 100 kHz to 100 MHz.
1 105
1 104
Vthreat (i)
1 103
100
1 105 1 106 1 107 1 108
Fi Hz
Figure 9.4.2 Threat voltage induced in the common-mode loop for a 15 m link.
In section 5.5, a method of relating the power density of the external interference to the
differential-mode current induced in a signal link was derived. A Mathcad worksheet was
used to calculate the frequency response of induced current in that link to a field of constant
power density. Electric field strength is related to power density by:
p
E Zo S 9:4:1
This being so, the response of Figure 5.5.9 could also be described as the response of the
assembly to a field of constant electric field strength. Since the contractual requirements
usually define the threat environment in terms of a field of varying intensity, it is necessary
to include this parameter as a variable.
Given knowledge of the length of the assembly-under-review, it is a simple matter to
define the threat in terms of the voltage that will be induced in the common-mode loop.
Figure 9.4.2 illustrates the variation of that voltage with frequency for a signal link of 15 m.
This is the same length as the link described in section 5.5, a twin conductor cable routed
9.4 Susceptibility requirements 259
over a conducting structure. The computations are carried out by page 1 of Worksheet 9.4
and illustrated by Figure 9.4.3. It is worth noting that the threat voltage peaks at over 10 kV.
A representative circuit model was created in section 5.5 to simulate the coupling
between the common-mode loop and the differential-mode loop, and a worksheet was
derived to calculate the amplitude of the interference current in the signal circuitry. By
applying the threat voltage Vthreat to the culprit loop, the current Iout can be calculated. The
response is illustrated by Figure 9.4.4.
The computations were carried out by adding pages 1, 2, and 3 of Worksheet 5.5 to
Worksheet 9.4. (See figures 5.5.3, 5.5.6, and 5.5.7)
0.1 10
l := 15 m Length of cable
0.1 10
c
0.5 10 fq := = 4.997 106 Hz
4 l
0.5 10
1.6 10 n := 100 N := 20 n i := 1 .. N
1.6 560
fq
5 560 Fi := i
n
5 380
10 380
data = fnd(thres) := j 1
10 200
while dataj, 1 thres
30 200
30 200 jj+1
100 60 datak, 2
100
Iout 1
0.1
0.01
1 105 1 106 1 107 1 108
F Hz
It is clear from this graph that the critical range of frequencies lies between 1.5 and
10 MHz. So the design of the interface circuits, the shielding, and the signal processing will
need to focus on providing protection to the system function over this range of frequencies.
The above example is intended for illustrative purposes only, to show how system design
can be related to the susceptibility requirements.
It is also worth noting that this is a worst-case analysis. In practice, there will be a
significant reduction in threat level due to shielding afforded by the structure as well as
re-radiation of the interference back into the environment. Although this is a limitation of the
circuit modeling approach, any errors will be on the side of caution.
20 Hz 130
2 kHz 130
50 kHz 50
500 kHz 20
150 MHz 20
9.5 Emission requirements 261
The test involves the use of a current transformer clamped round the cable-under-test and
connected to a current measuring device. The values in the second column of the table define
points on the characteristic. Figure 9.5.1 illustrates that characteristic. It defines the upper
limit for measured emissions.
Creating this graph was a simple matter of converting the values of dBmA in the second
column of the table to values in amperes and plotting these values at the appropriate fre-
quency. Figure 9.5.2 reproduces the first section of Worksheet 9.5 because some readers
might be unfamiliar with the relationship between dBmA and A.
10
A
1
0.1
0.01
Ilimit
1 103
1 104
1 105
1 106
10 100 1 103 1 104 1 105 1106 1 107 1 108 1 109
F1 Hz
dBA
y
20
A 10 y
20 130
2000 130
500 103 20
i := 1.. rows(data) Ilimiti := Amps(dBi)
200 106 20
The emission requirements illustrated by Figure 9.5.1 can be compared to the predicted
response of any given signal link, for example, the link depicted by Figure 9.5.3.
This particular link is used as an example because it is the one which was subjected to a
susceptibility analysis in the previous section.
Analyzing the emission characteristics of this link can be carried out by copying pages 1,
2, and 3 of Worksheet 5.5, (Figures 5.5.3, 5.5.6, and 5.5.7) adding them to Worksheet 9.4,
and making a few minor changes to the program. Basically, these changes are:
setting the value of the radiation resistance Rrad to zero (Inserting a 50 W resistor in
series with the structure is only necessary when susceptibility is being analyzed.),
setting the voltage source in loop 2 of the model to zero (When simulating emission, it is
assumed that there is no voltage source in the common-mode loop.),
setting the voltage source in loop 1, Vout, to unity (The voltage source is assumed to be in
the differential-mode loop.).
radiated field
structure
10
A
1 maximum acceptable radiation
0.1
1 104
1 105
1 106
1 103 1 104 1 105 1 106 1 107 1 108
F1 , F2 Hz
Figure 9.5.4 Comparing the response of the proposed design with the emission requirements.
9.6 Planning 263
It is clear from this analysis that the configuration-under-review is non-compliant with the
requirements at frequencies above 50 kHz.
However, this analysis is of the worst-case conditions.
The design can be remedied by
limiting the bandwidth and the amplitude of the signal to that which is essential for the
intended function, and
implementing one or more of the techniques described in Chapter 8.
9.6 Planning
9.6.1 Performance requirements
These usually take the form of a set of EMC test requirements for the particular system being
developed. Several groups have developed independent sets of such requirements: military,
aircraft, spacecraft, automotive, ships, commercial, etc. Different countries have different
regulations. It is usually the case that the manufacturer involved in designing the product
already has a full set of regulatory documents in place. If not, then there are several con-
sultancies that specialize in offering advice. The International Journal of Electromagnetic
Compatibility [9.7] identifies a number of such organizations.
It is sometimes the case that external EMC requirements are inadequate to meet the needs
of safety or reliability. In which case, in-house EMC requirements may be called for.
Once the EMC requirements have been defined, the task is to design the equipment to meet
those requirements. Normal procedure is to carry out formal EMC testing after all other tests
have been completed. If the result of the formal test process is a number of non-compliances,
then the equipment will need to be modified and resubmitted. This can be an expensive process,
since at this stage all the manufacturing drawings will have been completed, all the manu-
facturing processes will have been written, and all the components will have been purchased.
Also, there is not any guarantee that the modified equipment will be compliant.
The previous chapters of this book have shown that a clear correlation can be established
between test requirements and the detailed design of the equipment. This allows EMC to be
treated in the same way as requirements such as frequency response, functional performance,
mass, size, reliability, and cost. This being so, analysis and test of the EMC requirements can
be integrated into normal design process.
sections 7.1 and 7.2 use 10-turn windings on the toroids to ensure that the impedance
reflected by the test equipment into the loop-under-test is as small as possible.
Some commercial transformers extend the frequency range by employing just one turn on
the primary winding.
It would be a worthwhile exercise to manufacture a set of triaxial cables, as described in
section 7.3. It is important that the test equipment does not create spurious coupling with the
equipment-under-test.
The use of such devices need not be restricted to bench testing during the development
process; they can also be used for in situ testing during trials and trouble-shooting. The
availability of bench test equipment to check EMC enables newly purchased equipment to be
assessed before it is integrated into the system.
9.6.3 Software
Any organization involved in the design of electronic equipment will already have a number
of personal computers devoted to the design and manufacturing process. SPICE software
will be installed on many of these to analyze the functional performance of the circuitry on
the printed circuit boards.
Such software is based on the use of nodal analysis and the existence of a reference node
at zero volts. If it is assumed that all connections to the reference plane are at zero volts, then
extremely complex boards can be analyzed. However, this assumption carries with it the
hidden assumption that there is no such thing as interference.
This does not mean that SPICE software is incapable of analyzing interference coupling.
It is entirely possible for this software to analyze any of the lumped parameter models of
Chapter 2. Moreover, such models can be combined with models of complex printed circuit
boards. Provided the user is aware of the limitations of such software, valuable information
can be gleaned from the exercise.
However, there will always be occasions where such an approach is inadequate, making
it necessary to use mathematical software such as Mathcad. This type of software has the
advantage that the computations are not restricted to circuit models, as many of the work-
sheets in this book can confirm. If both types of software are installed on the same computer,
it becomes possible for them to exchange data.
As far as susceptibility is concerned, the most critical links are often, but not always,
those that carry very low level signals, such as those from remote transducers.
As far as emission is concerned, the most critical links are those that carry the highest
transient currents. Power supplies fall into this category, since they almost always contain
switch-mode converters. The transmission of fast digital signals between separate items of
equipment can also create high levels of emission.
It is normal practice to create a block diagram at the early stage of a project to identify the
main items of equipment and to define their function. Since the links between blocks at this
level of definition are likely to be the longest, as well as carrying the most important signals,
such a diagram can be used to identify the most critical links.
9.6.6 Characterization
Although the methodology described in this book can be used to create circuit models of any
particular type of cable coupling using manufacturers data, there are always uncertainties.
For example, the relative permittivity and the radiation resistance are usually unknown
variables (unless one has to hand a field solver that can accurately predict permittivity and
radiation resistance from dimensions and material characteristics).
If the link-under-review is constructed from different types of cable, or if the effect of the
terminations is uncertain, tests on a representative rig can be used to create a representative
circuit model that can be used with confidence.
However, it is not necessary to characterize every signal link in the system. The repre-
sentative circuit model of a particular type of cable can be applied to every cable of that
type. Since the L, C, R, and G parameters of a cable are proportional to length, tests on a long
cable can be used to define the characteristics of a short cable.
Tests of long cables over a low frequency range can be used to predict the response at
high frequencies. When carrying out such a test, care should be taken that the cross section is
266 CHAPTER 9 System design
uniform along the entire length and that the effect of the terminations is minimal. The test
described in section 7.5 is an example.
One observation made during these tests was that it did not seem to matter how the cable
was laid out, or whether or not there were tight bends. These features might cause local hotspots
in the field distribution, but they did not seem to have any effect on the overall response.
Another observation was that when two conductors are twisted together to create a twin-core
cable and subsequently separated, the lengths of the two conductors were different. The more
turns per meter, the greater the difference. If the lengths are different, then so are the impedances.
If the impedances are different, then the coupling between common-mode and differential-mode
signals must increase. Specifying a high number of turns per meter is not a good idea. There are
other ways of achieving close spacing between conductors, for example, ribbon cable.
Capacitors are really open-circuit transmission lines coiled into a small volume. Just like
transmission lines, they have inductive properties. Unlike transmission lines, there is only
one resonant frequency, as the response of Figure 7.7.3 illustrates. This could be due to
capacitive coupling between adjacent turns. Interference within the component seems to
cancel out the higher frequency resonances that would occur with a normal transmission line.
Tests can be devised to measure the value of each component-within-a-component as
illustrated by Figure 7.7.4. By applying an input signal via a 50 W cable and monitoring the
output via another 50 W cable, the component can be measured over a range that includes a
peak or a null in the frequency response. A circuit model can be postulated to simulate the
component. Analysis of this model allows the theoretical curve to provide a faithful simu-
lation of the measured response. Adjusting component values of the model allows the actual
and simulated responses to coincide. The end result is a representative circuit model of the
component, valid over the frequency range of the tests carried out.
By creating a library of representative circuit models to simulate signal links and com-
ponents, it becomes possible to simulate any signal link in the system. This is essentially the
approach used with SPICE software, but adapted to meet the needs of EMC.
Several examples of representative circuit models are provided in the previous pages, and
these are listed in Table 9.6.1.
These have been derived from the small set of general circuit models listed in Table 9.6.2.
Description Figure
Description Figures
exercise to spend some time assessing the possible problems that could arise due to EMI. The
section Banana Skins in every issue of the EMC Journal [9.8] provides some hair-raising
examples of what can go wrong.
When dealing with any problem relating to electromagnetic interference, the approach of
the circuit designer should always be to create an interface diagram of the link-under-review
such as the one illustrated by Figure 9.2.4, and assess its characteristics using the guidelines
of section 9.1. This task should take less than a day for even the most convoluted link.
If this does not immediately identify the cause of the problem, the diagram can be used to
create and assess a circuit model of the link. The results of this initial assessment can
determine whether there is a simple solution or whether more detailed analysis is necessary.
Tests on a prototype assembly can be devised to check that the solution is viable before
implementing the modification.
The objective should be to acquire a high degree of confidence that the system-under-
review will meet the defined requirements without over-designing the protection measures.
It is much easier to do this when the design is based on the results of circuit modeling than
when it is based on the consensus of opinion at a design review meeting.
If further development is needed to improve the design of a critical assembly, then the
information acquired by the circuit modeling approach can be used as a baseline for more
sophisticated analyses using three-dimensional field solvers.
APPENDIX A
Mathcad worksheets
As far as circuit modeling is concerned, Mathcad software eliminates most of the tedium
involved in developing programs. It can handle tasks that include the calculation of values of
circuit components, analysis of the response of the circuit model, analysis of test data to
define the response of an assembly under test, and display a single graph showing the
responses of model and hardware. This can all be done on the same worksheet; that is,
without the need to invoke subroutines that are recorded on separate files. Moreover, the
equations used in the computations are displayed in exactly the same way as they would
appear in a textbook on circuit theory.
In a programming language, equations look something like:
x B SQRTB 2 4A C=2A
With Mathcad software, the same equation looks the same way as it does in a reference book:
p
b b2 4 a c
x
2a
This makes the programs much easier to read and understand. Mathcad software also avoids
the need to use a rigorous procedure to write a program. If the programmer gets the syntax
wrong, the software refuses to accept the entry; and provides a message indicating what is
wrong.
On a worksheet, the equations can be set out from left to right, top to bottom. Text can be
included anywhere on the page.
Doubtless there are other software packages which provide the same sort of facility.
However, for consistency in presentation, Mathcad is the one used here.
Since this software has been available for over a decade, there is a fair probability that the
reader is already using it. If not, it should be fairly easy task to modify the programs in the
worksheets of this document to convert them into programs which work with other software
packages.
A few symbols in Mathcad have a special meaning:
269
270 APPENDIX A Mathcad worksheets
If the equals sign is in bold typeface, the software interprets this as the boolean equal
and returns a zero or a one.
The programming operator looks like that shown on Figure A.1:
root(a, b, c) = discr b2 4 a c
num b + discr
denom 2 a
num
denom
The programming operator behaves like a function, taking input variables and returning
an output. This output is the last variable to be declared. In the example above it is the value
of the ratio of num divided by denom. The program can return a single variable, a vector,
or an array.
Local variables defined in the function are not visible outside. However, variables
declared in the worksheet above the program function are visible within the function.
Included in the software are a number of built-in functions. As far as circuit modeling is
concerned, the most important is lsolve(M, v). The argument M is a square array, while
v is a vector. M contains as many rows as there are elements in v. This function returns a
solution vector x such that Mx v.
Another useful characteristic of the software is that it can distinguish between the j
operator as used with complex numbers and the variable j. It can be defined as an ima-
ginary number by typing the characters 1j. It appears on the worksheet as j.
It is also possible to intermingle text with variables. This facility is put to good use in
defining the units appropriate to each variable. To avoid cluttering up the worksheets, this is
usually restricted to places where a variable is defined or a final result is displayed.
Armed with this information, there should be no real problems in understanding the
programs contained in the Mathcad worksheets of this book. Since the figures in the book
which depict Mathcad worksheets have been hand-copied from copies of the actual work-
sheets, it is possible that errors have been introduced in the transcription process. Further
errors could be introduced by the reader hand-copying the text onto his or her own computer.
To prevent the chance of such errors creeping in, copies of the original Mathcad files are
available in a zip folder which can be downloaded from www.designemc.info. These files
can be run on any computer which has Mathcad version 15.0 (or higher) software installed.
The techniques described in the previous pages can be used to simulate the interference-
coupling characteristics of any signal link. Every electrical or electronic system will have its
own particular set of interference problems. It should be possible for individual designers to
carry out tests and create circuit models of their own critical links. Such information can be
shared. Mathcad users have access to PlanetPTC, a mix of dynamic channels that enables PTC
customers and product development professionals to actively participate in exchanging ideas.
However, it does not really matter which software is used to carry out the calculations.
The key feature of any circuit model is the fact that it is a shorthand method of defining a set
of equations. It becomes possible to describe the hardware, the tests, the model, and to
illustrate the results in a single report. There are many forums and many communities which
can be used to publish such reports.
APPENDIX B
MATLAB
Translating Mathcad worksheets into MATLAB m-files was a fairly simple task. That being
so, it should not be difficult for MATLAB users to understand the function of the Mathcad
worksheets replicated in this book. There are some differences which could be puzzling at first.
With MATLAB, there are five windows; Command Window, Command History, Cur-
rent Folder, the Workspace, and the Editor. The program is really a text file in the Editor
Window, and needs to be saved to a file before it is run, a task performed by typing and
entering a command in the Command Window. All variables and their computed values are
accessible in the separate Workspace. The Current Folder holds the files which are acces-
sible to the Command Window, and the Command History keeps a record of all the actions
taken to date. If a computation is carried out to produce a graphical picture, the picture is
displayed in a separate Figure Window.
With Mathcad, there is only one window; the Worksheet. Variables and expressions can
be defined anywhere on the page and intermingled with text, in the same way that calcula-
tions are recorded in a laboratory notebook. There is no list of numbers down the left-hand
side of the page. Of course, there is a basic rule; the sequence of commands must be left to
right; top to bottom. Graphs are created and modified by invoking menu commands. The
run command is invoked by clicking anywhere on an empty portion of the worksheet. This
means that everything; definitions, expressions, functions, numerical results, graphs, and
explanatory text, is displayed on the worksheet. Any part of the worksheet can be copied and
pasted onto a Microsoft word document. Copies of the Mathcad worksheets are available in
rich text format (.rtf) at www.designemc.info.
Both Mathcad and MATLAB work with arrays and vectors, and the related operations are
called up by similar expressions. For example; Ai,j in a Mathcad worksheet can be translated
into A(i, j) in MATLAB, V k : 0 into V zeros(k,1), lsolve(Z, p V ) into linsolve(Z, V ), etc.
Similarly, ln( f ) can be translated into log( f ), jxj into abs(x), x into sqrt(x), etc.
In Mathcad, all vectors are column vectors; in MATLAB, vectors can be defined as row
vectors or column vectors.
In Mathcad function statements, the parameters on the worksheet are visible to the
computations in that function. In MATLAB, they are not; every variable used by a function
must be included in the set of input variables.
In Mathcad, the function statements form part of the worksheet. In MATLAB, these are
stored as separate Function Files. This means that those MATLAB m-files which invoke the
use of special functions must have the relevant Function Files available in the Current Folder.
271
272 APPENDIX B MATLAB
If in doubt about the method of computation employed by any worksheet, the reader who
is familiar with MATLAB can easily clarify its purpose by examining the text of the
equivalent m-file. Each Mathcad worksheet has been translated into one or more MATLAB
files and the set of files stored in a zip folder. This folder can be downloaded from www.
designemc.info.
In the preceding pages of the book, each page of a worksheet is presented in the form of a
figure, and these figures are interspersed with diagrams and descriptive text. To ease the task
of relating the MATLAB files to the Mathcad worksheets, the text of each worksheet is
available in its entirety in portable document format (pdf) file, and these files are also
available at the above website.
APPENDIX C
It is entirely possible that the hybrid equations of section 4.1 are unfamiliar to engineers,
even those who are experienced in the application of electromagnetic theory. If the starting-
point equations are viewed with scepticism, then designers are unlikely to have any con-
fidence in the succeeding mathematics. That being so, it is necessary to provide proof of the
validity of (4.1.4) and (4.1.5).
It is even more likely that the average circuit designer has never seen them before. This
section provides a succinct derivation of these equations. It is also made clear that both
conductors of the line possess the properties of inductance, resistance, capacitance, and
conductance. By defining the variables in terms of loop parameters, it avoids the use of
the concept of the equipotential ground. It is basically a set of lecture notes copied from a
blackboard at Glasgow University in 1959.
Figure C.1 illustrates how voltage and current can vary along the length of a transmission
line.
For the purpose of deriving the transmission line equations, the following definitions
apply:
Voltage and current are functions of both x and t. For the section dx:
dV dV
R j w L dx I V V dx dx C:1
dx dx
dI dI
G j w C dx V I I dx dx C:2
dx dx
Hence:
dV
R j w L I C:3
dx
273
274 APPENDIX C The hybrid equations
dI
G j w C V C:4
dx
V dx
VV+
x
I dx
I I+
Vsource x
x
dx
d2 V dI
R j w L R j w L G j w C V C:5
dx2 dx
d2 I dV
G j w C R j w L G j w C I C:6
dx2 dx
d2 V
g2 V C:7
dx2
This can be confirmed by differentiating (C.8) twice. Another way of defining this rela-
tionship is:
This is because:
gx
egx egx e egx
V A B
2 2
A B gx A B gx
e e
2 2
gx gx
Ce De
APPENDIX C The hybrid equations 275
Now, since:
s
R j w L RjwL RjwL
p Zo
g R j w L G j w C GjwC
then:
1
I A sinhg x B coshg x C:10
Zo
Vs Is I Ir Vr
x l
V Vr at x l
I Ir at x l
Hence:
B Vr sinhg l Zo Ir coshg l C:15
Similarly:
A Vr coshg l Zo Ir sinhg l C:16
Hence:
V Vr coshgl x Zo Ir sinhgl x C:17
Similarly, substituting A and B in (C.12) and reducing the hyperbolic products gives:
Vr
I Ir coshg l x sinhg l x C:18
Zo
At the sending end, x 0; V Vs; I Is. Substituting these values in (C.17) and (C.18)
gives:
Vs Vr coshg l Zo Ir sinhg l
Vr C:19
Is sinhg l Ir coshg l
Zo
It is worthwhile emphasizing the fact that, in this derivation, the parameters R, L, G, and C
are defined in terms of W/m, H/m, S/m, and F/m. The concept of per-unit-length parameters
is inherent in all the transmission line equations to be found in textbooks. Even so, this
appendix is the only place in this book where such parameters are invoked. Section 4.1
shows that the analysis of transmission line behavior can be related to the actual resistance,
inductance, capacitance, and conductance of the conductors.
APPENDIX D
Definitions
These definitions are those used in this book. They are not necessarily the same as those
appearing in other documents.
antenna-mode current Unidirectional current in the conductors of a cable. That which flows
between the cable and the environment.
buffer circuit The circuit which forms an interface between the conductors of a signal
link and the processing circuitry of the equipment unit.
circuit equations A set of equations which relate voltages to currents in a circuit model.
circuit model A model which obeys the rules of circuit theory and which simulates the
behavior of the assembly under review.
circuit parameter A parameter used in a circuit model.
common-mode current Current which flows in the loop formed by the cable and the structure.
common-mode gain The ratio of the output voltage of a buffer circuit to the common-mode
input voltage when the differential-mode input is zero.
common-mode rejection The ratio of the differential-mode gain to the common-mode gain.
Usually quoted in terms of decibels.
composite conductor A set of elemental conductors, aligned in parallel, which enables the
distribution of currents or voltages in the actual conductor to be
simulated.
conducted emission The current induced in the common-mode loop by a voltage source in the
differential-mode loop. See Figure 4.4.1.
conducted susceptibility The current induced in the differential-mode loop by a voltage source in
the common-mode loop. See Figure 4.4.5.
culprit loop The loop in which the interfering voltage source is located.
current transformer A transducer which monitors the magnetic field surrounding a group of
conductors and generates a voltage proportional to the sum of the
currents in those conductors.
delay-line model A circuit model which simulates the differential-mode behavior of a
transmission line.
differential-mode current Current which flows in the loop formed by the signal and return
conductors.
differential-mode gain The ratio of the output voltage of a buffer circuit to the differential-mode
input voltage when the common-mode input voltage is zero.
277
278 APPENDIX D Definitions
distributed parameters Parameters which are derived from the use of per-unit-length parameters.
earth The conductor(s) designated to carry fault current in an AC distribution
network. Always connected to the conducting structure.
elemental conductor A conductor which represents a small segment of the surface of a
composite conductor.
EMC Electromagnetic compatibility. The ability of a device, unit of equip-
ment, or system to function satisfactorily in its electromagnetic
environment without introducing intolerable electromagnetic dis-
turbances to anything in that environment.
EMI Electromagnetic interference.
EUT Equipment under test.
floating configuration A wiring configuration where the signal link is isolated from the structure
at one (or both) ends.
general circuit model A model which includes interface circuit components as well as cable
coupling components, but which does not specify component values.
ground Another name for the conducting structure. The terms ground, earth
and structure are interchangeable in the analyses described in this
book.
grounded configuration A wiring configuration where the return conductor of the signal link is
bonded to local structure at each end.
loop equations A set of equations, derived from the primitive equations, which relate
loop voltages to loop currents.
loop parameter Parameter which has been derived from groups of primitive parameters
or partial parameters. Loop parameters can be measured by electrical
test equipment.
lumped parameter A resistor, capacitor, or inductor which represent the relevant properties
of a defined length of conductor. The term also applies to circuit
components.
partial capacitance The capacitance of a composite conductor, or the capacitance of a loop
segment.
partial current A portion of the current flowing in a conductor.
partial inductance The inductance of a composite conductor, or the inductance of a loop
segment.
partial inductor Inductor associated with partial current or partial voltage.
partial parameters Those parameters associated with the behavior of composite conductors
as antennae. The term is also used to distinguish between incident and
reflected currents in transmission lines.
partial voltage A portion of the voltage in a circuit loop.
per-unit-length parameter Resistance, inductance, capacitance, or conductance which is defined in
terms of W/m, H/m, F/m, or S/m.
primitive capacitance A component which relates the voltage on an isolated conductor to the
energy contained by the electric field in which it is immersed.
primitive current Unidirectional current flow in a single conductor of a multi-conductor
assembly.
primitive equations A set of equations relating primitive voltages to primitive currents.
primitive inductance A component which relates the current in the conductor to the energy
contained by the magnetic field in which it is immersed.
primitive parameters Parameters associated with conductors which are treated as antennae.
APPENDIX D Definitions 279
Chapter 1
1.1. Williams, T. EMC for Product Designers (Section 1.1.1). 2nd edn. Jordan Hill, Oxford: Newnes;
1996. p. 4. ISBN: 0-7506-2466-3.
1.2. Europe EMC guide. The International Journal of Electromagnetic Compatibility. Retrieved from
http://www.interferencetechnology.eu
1.3. Armstrong, K.: EMC design of SMP and PWM power converters. EMC Journal. 2011, March:
p. 28.
1.4. Williams, T. EMC for Product Designers. 4th edn. Jordan Hill, Oxford: Newnes; December 2006.
ISBN: 0-750-68170-5.
1.5. Tesche, F., Ianoz, M., Karlsson, T. EMC Analysis Methods and Computational Models. New
York, NY, USA: John Wiley & Sons, Inc., 1997. ISBN: 0-471-15573-X.
1.6. Defence Standard 59-411, Part 5, Issue 1, Amendment 1. Electromagnetic Compatibility. Part 5.
Code of Practice for Tri-Service Design and Installation. (Section 8.9. Single Point Reference
Connection). Glasgow, UK: Ministry of Defence; January 2007. p. 29.
1.7. Shepherd, J., Morton, A.H., Spence, L.F. Higher Electrical Engineering (Section 7.28. Equivalent
Phase Inductance of a Three-Phase Line). Pitman, London, UK: Pitman Publishing Limited;
1985. pp. 234235. ISBN: 0-273-40063-0.
1.8. Shepherd, J., Morton, A.H., Spence, L.F. Higher Electrical Engineering (Section 7.16. Equivalent
Phase Capacitance of an Isolated Three-Phase Line). Pitman, London, UK: Pitman Publishing
Limited; 1985. pp. 216219. ISBN 0-273-40063-0.
1.9. Burrows, B.J.C. The computation and prediction of induced voltages in aircraft wings. CLSU
memo 18. April 1974. Culham Lightning, Units 13/15, Nuffield Way, Abingdon.
1.10. Armstrong, K. EMC Design Techniques for Electronic Engineers. Armstrong/Nutwood, UK:
Nutwood UK Limited, Cornwall, UK; 2010. ISBN: 978-0-9555118-4-4. Retrieved from http://
www.emcacademy.org/books.asp
1.11. Paul, C.R. Introduction to Electromagnetic Compatibility. 2nd edn. Hoboken, NJ, USA: Wiley-
Interscience; January 2006. ISBN: 978-0-471-75500-5.
281
282 References
Chapter 2
2.1. Skitek, G.G., Marshall, S.V. Electromagnetic Concepts and Applications (Section 2.5.
Electric Field Intensity of a Line of Charge). Englewood Cliffs, N.J., USA: Prentice Hall; 1982.
ISBN 0-13-248963-5.
2.2. Page, L., Adams, N. I., Principles of Electricity: Inductance of Straight Conductors. New York,
USA: D Van Nostrand; 1958. p. 325.
2.3. Skitek, G.G., Marshall, S.V. Electromagnetic Concepts and Applications (Section 8.3. Magneto-
static Field Intensity from the Biot-Savart Law: Magnetic Field due to a Filamentary Current
Distribution of Finite Length). Englewood Cliffs, USA: Prentice Hall; 1982. ISBN 0-13-248963-5.
2.4. Skitek, G.G., Marshall, S.V. Electromagnetic Concepts and Applications (Section 12.12. Skin
Effect, and High and Low Loss Approximations). Englewood Cliffs, USA: Prentice Hall; 1982.
ISBN 0-13-248963-5.
Chapter 3
3.1. Skitek, G.G., Marshall, S.V. Electromagnetic Concepts and Applications (Section 7.4. Image
Solution Method: Capacitance between two Cylindrical Conductors). Englewood Cliffs, USA:
Prentice Hall; 1982. ISBN 0-13-248963-5.
Chapter 4
4.1. Skitek, G.G., Marshall, S.V. Electromagnetic Concepts and Applications (Section 12.2. General
Equations for Line Voltage and Current). Englewood Cliffs, USA: Prentice Hall; 1982. ISBN
0-13-248963-5.
Chapter 5
5.1. Skitek, G.G., Marshall, S.V. Electromagnetic Concepts and Applications (Section 14.4. The Half-
Wave Dipole). Englewood Cliffs, USA: Prentice Hall; 1982. ISBN 0-13-248963-5.
5.2. Ordnance Board Pillar Proceeding P101 (Issue 2). Principles for the design and assessment
of electrical circuits incorporating explosive components. (Annex. E. Appendix 1. The Radio
Frequency Environment). p. E1-3. Bristol, UK.
Chapter 6
6.1. Savant, C. J, Jr., Roden, M.S., Carpenter, G. L., Electronic Design Circuits and Systems
(Appendix A. SPICE. Section A.2.4.3 Transient Analysis). 2nd edn. Redwood City, California:
The Benjamin-Cummings Publishing; 1991. ISBN 0-8053-0292-1.
References 283
Chapter 7
7.1. Ediss Electric Ltd. Totton, Hampshire, UK: Ediss Electrical Ltd., Retrieved from http://www.
ediss-electric.com
Chapter 8
8.1. Gnecco, L.T. The Design of Shielded Enclosures. Woburn, MA, USA; Newnes: 2000. ISBN
0-7506-7270-6.
8.2. Ordnance Board Pillar Proceeding P101 (Issue 2). Principles for the design and assessment of
electrical circuits incorporating explosive components. (Annex. E. Appendix 1. Section 28:
Shielding Assessment). p. E114. Bristol, UK.
8.3. Thomas & Betts Limited. A Guide to BS EN 62305:2006. Protection against Lightning.
Nottingham, UK: Thomas & Betts Limited; 2008.
Chapter 9
9.1. EMC probes. Magnetic Sciences. Retrieved from http://www.magneticsciences.com/
EMCProbes.html
9.2. Horowitz, P., Hill, W. The Art of Electronics. 2nd edn. Cambridge, CB2 1RP, UK: Cambridge
University Press; 1989. ISBN 0-521-37095-7.
9.3. Savant, C. J, Jr., Roden, M.S., Carpenter. G. L., Electronic Design Circuits and Systems. 2nd
edn. Redwood City, California. The Benjamin Cummings Publishing; 1991. ISBN 0-8053-029-1.
9.4. Ludwig, R., Bretchko, P. RF Circuit Design Theory and Applications. Upper Saddle River, New
Jersey: Prentice Hall; 2000. ISBN 0-13-122475-1.
9.5. Defence Standard 59-411, Part 2, Issue 1, Amendment 1. Electromagnetic Compatibility. Part 2.
The Electric, Magnetic & Electromagnetic Environment. Table 18. Front Line and Operational
Support Equipment Field Strength. Ministry of Defence; January 2008. p. 28.
9.6. Defence Standard 59-411, Part 3, Issue 1, Amendment 1. Electromagnetic Compatibility. Part 3.
Test Methods and Limits for Equipment and Sub Systems. Appendix B.2. DCE02.B Conducted
Emissions Control, Signal and Secondary Power Lines. 20 Hz150 MHz. Figure 51. DCE02
Limit for Air Service Use. Ministry of Defence; January 2008. p. 84.
9.7. The International Journal of Electromagnetic Compatibility. 1000 Germantown Pike, F-2 Plymouth
Meeting, PA 19462, USA: ITEMTM. www.interferencetechnology.com
9.8. The EMC Journal. Eddystone Court, De Lank Lane, St Breward, Bodmin, Cornwall, UK. Nutwood
UK Ltd. www.theemcjournal.com
Index
285
286 Index