Escolar Documentos
Profissional Documentos
Cultura Documentos
030601-10-T
ROBUST
HYBRID
FINITE
FOR ANTENNAS
AND
J. Gong
Volakis
and
National
J.L.
Aeronautics
Space
Langley
Hampton,
ELEMENT
MICROWAVE
METHODS
CIRCUITS
and
Administration
Research
Center
VA
23681-0001
October
1996
THE
UNIVERSITY
Radiation
Department
and
Ann
USA
MICHIGAN
Laboratory
of Electrical
Computer
Arbor,
OF
Engineering
Science
Michigan
48109-2122
Report
NASA
Grant
Langley
Grant
NAG
#030601-10-T
1-1478
Title:
Simulation
of Conformal
Antennas
Report
Title:
Report
Authors:
Primary
University
Primary
tor:
on Composite
Robust
Development
nite Element
Methods
and
Microwave
J. Gong*
Collaborator:
NASA-Langley
J.L.
John
L. Volakis
Platforms
of Hybrid
FiFor Antennas
Fred
Beck
Volakis
Collabora-
Address:
(804)
Radiation
ing and
864-1829
Laboratory
Department
of Electrical
Computer
Engineer-
Science
Ann Arbor,
Michigan
48109
Email:
volakis_umich.edu
Phone:
Date:
*This
ment
October
report
of the
requirements
Slot
Circuits
and
Telephone:
University
Spiral
for the
(313)764-0500
1996
degree
at the
2122
toward
University
the
fulfill-
of Michigan.
TABLE
DEDICATION
OF
CONTENTS
..................................
ACKNOWLEDGEMENTS
LIST
OF
FIGURES
LIST
OF
TABLES
LIST
OF
APPENDICES
ii
..........................
iii
...............................
ix
................................
xv
............................
xvi
CHAPTER
I. Introduction
II.
..............................
1.1
Overview
1.2
Fundamentals
Finite
2.1
Maxwell
Boundary
1.2.3
Uniqueness
1.2.4
Integral
Element
2.2
Galerkin
2.3
Total
and
Theorem
Equation
and
and
Value
Equivalence
Problems
Principle...
Green's
in Electromagnetics
Function
..........
Pertinent
Functional
-- II
..........................
For
Formulation
21
Lossy/Anisotropic
Media
25
Lossy/Anisotropic
Media
27
Field
2.3.1
2.3.2
Scattered/Incident
Galerkin's
Method
2.3.3
Variational
Extraction
'28
Formulations
Fields and
...................
Method
Radiation
and
2.4.2
Gain
Axial
RCS
..................
Ratio
vi
Pattern
.........
Boundary
Conditions
'.30
31
34
36
......................
2.4.1
13
20
......................
Scattered
7
10
. .
.....................
For
and
5
6
Boundary
Dyadic
Pertinent
Functional
-- I ...........................
and
...........
..................
Conditions
Formulation
Field
Parameler
Theory
Equations
Analysis
Functional
of Electromagnetic
1.2.2
2.1.'2
2.4
.............................
1.2.1
2.1.1
38
.............
.................
38
10
III.
Edge-Based
3.1
Inlroduction
3.2
Hybrid
3.3
3.4
IV.
V.
VI.
FE-BI
Efficient
...................
l:_
...........................
S\slenl
13
Functional
....................
I1
:1.:2.1
FK.M Subs\stem
....................
15
3.:2.2
3.2.3
Boundary
('ombined
System
Numerical
Selected
Integral
FE-Bl
Subs\slenl
lmt_lementation
Numerical
Results
Boundary
4.1
Introduction
4.2
Application
Integral
-IT
5(}
............
...............
...................
...................
Subsystem
31
3t
--
I ...........
(il
...........................
of Conjugate
4.9.1
BiCG
4.:2.:2
BiCG-FFT
4.2.3
Convolutional
6I
Gradient
Algorithm
With
........
Preconditioning
Algorithm
Form
Algorithms
For Linear
of Boundary
6:2
.......
(;:2
System
......
6-1
Integral
......
65
4.3
Mesh Overlay
Scheme
......................
4.3.1
Field Transformations
.................
71
71
4.4
Results
76
..............................
Efficient
Finite
5.1
Introduction
,5.2
Hybrid
,5,3
Edge-Based
5.4
Applications
5.,5
Concluding
Antenna
6.I
6.2
6.3
Element
Feed
Probe
Remarks
Simple
Probe
6.1.2
Voltage
Gap
Aperture-coupled
Coax Cable Feed
Modeling
II
............
Motivation
Hybrid
S0
80
...................
Elements
S1
................
85
87
.......................
95
.......................
98
............................
6.1.1
Conclusion
--
...........................
Feed
6.3.3
6.3.4
Circuit
Formulation
Prismatic
Modeling
6.3.2
6.4
Subsystem
...........................
FE-BI
6.3.1
VII.
Technique
98
Feed
Feed
..................
...................
Microstrip
Model
.........................
99
..............
.......................
FE-BI
System
...........................
vii
99
102
10:2
................
Proposed
Coax Feed Model
..............
Results
and Conclusion
................
............................
98
103
]03
10S
109
115
7.1
7.2
7.3
7.4
VIII.
AWE:
I] ",
117
12tJ
I:21
1_._'
121
Asymptotic
133
8.1
Brief
8.2
Theory
8.3
IX.
[m rodu(liolJ ...........................
.Nunwrhal Do en,}wdding ....................
TrullcatioiJ l_si_g I).MT . ....................
]rullcation l'six_gP.XIL .....................
7.4.1 -I'heorv . ........................
7.4._ Results .........................
Overview
of AWE
.....................
8.2.1
FEM
8.2.2
Asymptotic
Numerical
.............
1:_3
System
13-1
Recast
..................
\Vaveform
Implementation
Evaluation
...................
...............................
9.1
Discussion
9.2
Suggestions
9.3
Modular
BIBLIOGRAPHY
Evaluation
. .............................
Conclusions
APPENDICES
Waveform
on the
Development
136
138
142
Research
for Future
13-1
..........
Work
Tasks
..................
.......................
..................................
................
142
145
14,5
147
................................
159
Vlll
LIST
OF
FIGURES
Figure
1.1
(a)
Recessed
cavity
a dielectrically
Illustrations
shown
1.3
PEC
of equivalence
Illustrations
(by Recessed
cavity
in
when
12
applied
to tile
structure
14
principle
when
applied
to the
structure
Examples
of protruding
conformal
Illustration
of a scattering
lation ...................................
of a typical
A tetrahedron
3.3
Pair
3.4
A typical
and
of triangles
A flow
chart
its local
the
BiCG
solution
Comparison
of the
the
angle
30 off the
was
ith
.......
setup
for scattered
field
edge
major
finally
scattering
and
for the
ground
16
'2"2
formu-
data
.......
scheme
45
......
46
.................
49
circular
implementation
output
measured
shown
circular
plane ...................
ix
problem
numbering
preprocessors,
the
computed
of frequency
configuration
node/edge
the
and
(by
principle.
antenna
and
a few data
a function
equivalence
for a cavity-backed
describes
generation,
platform
30
geometry/mesh
mesh
on a planar
of the
problem
radiation
sharing
the
(a)
in consideration
of a typical
Illustration
15
configurations
platform
Illustration
3.6
plane.
plane ................
principle
of equivalence
on a curved
3.5
ground
ground
shown
1.4
in a PEC
coated
patch
antenna.
procedures
the
FE-BI
from
kernel,
to
.............
53
aoo backscatter
patch.
51
The
RCS
as
incidence
56
3.7
('onlpari-wl,
circular
palch
center
3._
:howll
of lh(. patch
Illustration
spiral
3..9
of the
o = 90-plane.
3.10
3.11
Visualization
of the
mesh
Illustration
of two
Printed
the
t_f the
paltern
for the
tm_.-arIik
c,,_ical
5",
(Ec,).
one-arm
lakel,
iit lhe
conical
spiral
FE-BI
shown
right
with
the
property
antenna
cavil v-
the
lower
laver
of a
60
triangles
...........
corresponding
is modeled
boundary
using
the
algorithm.
(a)
of the BiCG-FFT
presented
triangular
here
at
of the
technique
of a structured
mesh,
distribution
the BiCG-FFT
(b) Comparisons
ordinary
illust ral_'d
.....................
of equal
convolutional
59
field
triangles
patch
calculat
antenna
consists
circular
Overlay
tured
n_'_,h
radiation
in reference
near
patch
Structured
of the
,xv,,t)T tronl
5_
circular
to incorporate
configuration:
4.5
data
Comparison
of input impedance
backed
slot .................................
to prove
4.4
a1,d
wa,
calculated
with
I_,l 1t_,'
in fig..3.,_
stacked
4.3
contiguratiotl
illl_Ul il_ll,,'_l,Ll_,'
tl,(, fr(,(lu(,n(-v
('omparison
shown
of the
used
al_,l lll,'a.._ir_.cl
indices
to help
integral
.....
recessed
mesh
to a circular
70
scheme
Illustration
of the
result with that
in chapter
aperture
to conform
66
3 ........
over
patch
an
72
unstruc-
........
7:3
Illustration
of the parameters
and geometry
used in constructing
the
transformation
matrix elements
between
the structured
and unstructured
mesh ................................
4.6
Illustration
4.7
Comparisons
2 triangular
of the
cavity-backed
of the monostatic
patch
using
the
regular
using
the
BiCG-FFT
74
array
BiCG
shown
FE-BI
triangular
radar
cross
in fig. 4.6.
technique
proposed
in this
patch
section
The
.......
scattering
reults
described
chapter
array.
were
by a 2
computed
in chapter
.............
77
3 and
78
-t..',
('on_parison._
empty
of I}_e nlollO_.lalic
aperlure
with
4.9
Ill(' regular
using
the
BiCG
regular
BiCG
transform
5.1
Geometry
5.2
Illustration
5.3
Right
5.4
Geometry"
of the
and
FE-BI
servation
cut
slot antenna
RCS patterns
code
and
prismatic
wave
is polarized
the
FE-BI
along
the
to that
lines
whereas
the
dotted
Upper
code
periodic
element
Upper
face
FE-BI
transmission
setup
and
coefficient
Illustration
of a typical
5.10
Radiation
Pattern
axial
is achieved
from
are
norinallv
and
the
are
probe
the
ob-
X-pol
and
annular
slot
computed
lines
excitation
using
computed
is placed
the
us-
at the
89
computing
periodic
the
element
FSS
(top
(cross-sectional
view);
view)
of transmission
transmission
co-
Lower
figure:
...........
through
90
..........................
of the
5.9
ratio
for
comparisons
geometry
The
Radiation:
_ = 0 plane
The
the tetra-
5.4 .....................
in cavity
in fig. 5.6
figure:
(FSS)
code.
in fig.
figure:
cm in
using
code.
plane.
solid
marked
23.7
o = 0 plane
The
FE-BI
by a cavity
computed
in the
efficient
8-1
88
(co-pol)
of the
51
........................
patterns
shown
o\cr-
...........
backed
in fig. 5.4.
Calculations
wilh
S-I
shown
Illustration
usil_g
................
radiation
ture
5.8
prisms
is perpendicular
(y=0)
modeled
algorilhm
antennas
antenna
point
:{ al_d
7,9
using
annular
5.7
alltcnna
Co-pol
tetrahedral
ilt clial,lCl
...........................
Bistatic
plane
!i
............
BiCG-EFT
microstrip
3 cm deep
incident
_.=
xv,lc alcutal,'_t
described
chal)ICr.
patch
},_ a:_
ric lilliz_
.............................
prism
Scattering:
the
dich'cl
_IL(' resull._
in this
and
of tessellation
diameter
Again.
a crcular
of cavity-backed
angled
hedral
by
FE-BI
size and
BI Iechni(lUe
proposed
R(_'S scattering
laying
5.6
t'E
cavity
Bi('G-FFT
Bistatic
the
5.5
shown
usillg
radar
t}l_' same
92
multilayer
lower
through
the
figure:
FSS structure
2-arm
slot-spiral
at f=l.lGHz
(center
up to 60 degree
xi
frequency
measured
design
frequency
.................
selective
and
sur-
calculated
93
..........
...........
design).
96
A good
96
5.11
Hadiation
Pallern
II call
t,e seen
compared
3dB
5.12
6.1
turns
can
seen
shows
increased
to insure
center
frequency
the
aperture
computation
(c) uniform
of the
distribution
element
method
using
(a)
most
is the
point
coordinate
Field
calculated
Measured
and
lar patch
r=13mm;
antenna
cavity
Our
and
on
the
modeling
feed
!_7
raIlgCi.
frequency.
slill
Ihib
needs
lo
l,c
97
showing
II for two
the
different
100
line
with
with
surface)
cable
expansion
a coax
A/4 from
impedance
100
cable
cable
feed.
feed:
junction
(the
--:
the
by the
analytical:
of the
Field
110
(b)
field
110
as computed
(b)
from
.............
the length
short);
mesh
region ......
a coax
(6.18).
e0 along
input
typical
cavity-cable
coax
of the
(b)
antenna
at the
at a distance
calculated
....
is deleriorated
antenna,
region
antenna
coefficient
location
loops
microstrip
conductor
the
sutticien_
of frequency
_l,,'
...............
aperture:
patch
mesh
center
likel\
lower
patch
from
in a shorted
numerical.
based
mesh
FEM
at the
l,uxnt,'vt,f
pattern
and
line
slot
of a cavity-backed
is set to zero
e_=2.4;
(a)
of a cavity-backed
Illustration
Field
coupled
th,'
.......................
its discretization
Side view
end
pattern
microstrip
domains
wiIl_i_
of inner
quality
the
litlgCl
reniai_>
of the
number
a good
},ccolIl','.-
thai
(higher
ial_,.:,,',.
])atlcrll
is mosl
at the
of an
(a)
design
of frcqucIic.',
}nil still
in(licales
ratio
I and
region:
Ihis
axial
cavity
and
of lh_'
the
that
cmt
flequcltcv.
colllour
('ross-section
slot
ratio
at f=l.256GHz
certainly
Illustration
6.6
range.
that
to those
region
lh,xvvr
ccnler
spiral
Pattern
be
cavity
6.5
angle
compared
FEM
6.2
the. axial
at tile
in the
Radiation
It
that
to thal
for a wide
outer
at f=(I.!144(;tlz
xxx:
cable
along
the
shorted
finite
(leftradial
termination.
for a cavity-backed
111
circu-
having
the following
specifications:
patch
radius
radius
R=21.1mm;
substrate
thickness
t=4.1mm:
location
simple
probe
retains
the
x/=0.8
model
vertical
xii
cm
are
wire
field
distance
also
from
shown
connection
for excitation.
center.
for
to the
(a)
Results
comparison.
patch
Real
and
part:
112
(;.7
Measured
/an/_=(}.0012.
pro})(, model.
(a)
-"
Ileal
part
7.I
Illustration
of a shielded
7.2
Illustration
of the
7.3
7.4
A rectangular
perfectly
Plane
wave incidence
half-spaces
7.5
Typical
7.7
Field
field
7.11
7.13
line.
(b)
]aver.
....
1:21
truncal('d
........
1:22
two diagonally
inside
matched
TErn
mode
inside
uniaxial
element
per
wavelength
Reflection
aniso)VOl)iC
layer
used
vs 2i3t/Ag
layer
absorber
128
terminated
by a
is 10 elements
translates
to about
thick
13 samples
used
with
to terminate
the
vs 2_t/Ag
by the
coefficient
to the
by the
calculations
of a meander
line
...............................
Comparison
of calculated
and
perfectly
shown
matched
in fig. 7.6.
with
matched
. . 129
microstrip
laver ........
_3, for
matched
PML
129
in fig. 7.6.
shielded
uniaxial
geometry
shown
for the
perfectly
reference
Illustration
measurement
in fig.7.12
waveguide
for the
theoretical
1) for the
matched
vs 2/3t/Ag
line terminated
impedance
with
perfectly
128
the
vs 2/3t/Ag,
coefficient
Reflection
(o
to terminate
coefficient
Reflection
laver ...........
a waveguide
The
wavegui(h'
coefficient
Reflection
layer.
a rectangular
uniaxial
matched
shown
line
absorbing
mode
each
compared
7.12
microstrip
between
and
Input
113
115
a microstrip
perfectly
crostrip
c:
123
of TErn
of the
line terminated
7.10
and
on an interface
values
values
uniaxial
7.9
,7 c
line ................
uniaxiaI
by a perfectly
uniaxial
7.8
lll('l]lt)d:
l)arl ............
of a shielded
(a)
matched
llli_,
................................
terminated
7.6
lmagilmry
section
waveguide
the
(b)
xxx:
microstrip
cross
using
nleasur(,nl('lll:
the
shielded
uniaxial
terminated
130
mi-
layer.
microstrip
130
as
data ...............
used
131
for comparison
with
132
measured
.............................
results
for the
meander
line
132
xiii
8.1
l ttl
b.2
8.3
4th order and Sth order AWE implenwnta_ionsusing one I)oim expansionat 1.78GHz are shown to comparewith the referencedata.
With the 4th order AWE solutions. 5(;Zand :13_7_
bandwidtl_ agreement can be achievedfor the real (a) and imaginary (b) parts of
impedancecomputations, respectively. It is also shown that the _th
order solutions agree excellently with the refe,'encedata over the
entire band. (a) Real Part (b) Imaginary Part computations ....
1-tl
9.1
146
A.1
148
xiv
LIST
OF
TABLES
Table
5.1
Comparisons
6.1
The
of gain
correspondence
for each
parameters
coordinate(r,o
and
axial
between
ratio
the
or z) along
in (6.18) ...........................
XV
at different
edge
numbers
with
the
operating
and
definition
frequencies
tilt'
node
pairs
of tilt'
tildcd
95
108
LIST
OF
APPENDICES
Appendix
A.
Evaluation
of Matrix
B.
Evaluation
of the
C.
Formulation
D.
System
Elements
Boundary
for Right
Derivation
Integral
Angle
From
for Tetrahedrals
Prisms
A Functional
xvi
System
............
.Matrix
...................
..................
..........
t-t'_
152
154
157
CHAPTER
Introduction
One
of the
of robust
and
hybrid
design
integral
on
primary
finite
recently
wave
the
chapter
with
are
(FE-BI)
techniques
shape.
overviewed
FE-BI
of the
of certain
important
Both
in this
methodologies
structure
discussions
with
integral
of arbitrary
The
which
development
on vehicles
are
restricted
have
loaded
with
gain
of simulation
is a challenging
niques
and
hybrid
is concerned
for
finite
chapter
with
antennas,
dissertation
develol)_n('nt
for modeling
the
element
and
an emphasis
microwave
is then
fundamental
to this
the
and
outlined.
concepts
and
We
methods
study.
Overview
The
ods
dissertation
will be first
applications.
in electromagnetics,
1.1
antennas
methods
developed
millimeter
in this
element-boundary
of conformal
equation
conclude
goals
to
difficulty
planar
in being
dielectrics
performance
a sophisticated
task.
feeding
and
[1-4].
and
techniques
By and
mostly
extended
comprised
large,
existing
rectangular
antennas
analysis
patch
of intricate
such
practical
as coaxial
shapes
antenna
cable,
typically
and
to attain
designs
microstrip
mounted
design
antennas.
to non-rectangular/non-planar
Moreover,
structure,
for conformal
These
methtech-
configurations
larger
may
bandwidth
also
line.
require
stripline.
proxin_itv or al,erlur,,
are
riot
easi]\
co,,]4e_t
adaptal,le
sized <tMecTric
(e.g.
elemenl
and finite
unbouilded
field
modeling
scattering.
velop
The
antennas
loading-.
personal
cellular
GPS
for
such
techniques
are
being
re-designed
and
other
since
even
diverse
traditional
this
can be applied
to medical
complexity
veloped
able.
based
with
finite
element
tenna
structure
design,
analvsis
the
most
bination
feed
application
more
been
method
variety
of existing
is indeed
(multi-layer
network
methods
accurate
of the
and
substrate,
cavity
in engineering.
element
and
that
versatile,
finite
approach
the
and
other
hand.
boundary
anlennas
of new
have
in
shown
exterior
method
quite
volume
is one
of the
the
boundary
integral
be deadapt-
when
antenna.
(FE-BI)
[9]. The
of the
stacked
most
methods
an-
element
celebrated
integral
Thus
cou-
attractive
configurations
loading,
to the
software
geometrically
interior
size dielectric
etc.)
design
becomes
modeling
the
which
element
for
fields
need
for a host
and
antenna
of the
of
electromagnetics
and
emerging
the
growth
rapid
is an increasing
treatment
and
finite
the
low frequency
analysis
the
equation
volume,
On
representation
finite
ideal
('onformal
[7, 8].
that
integral
aim
simulate
requirements
and
robust,
demonstrated
traditional
a wide
are
there
protruding
diagnosis
demands
that
\Vith
of computational
potential
of new antennas
it has
the
for modeling
development
on methodologies
Recently'
pled
and
accurately
to meet
radiathm
schemes.
and
_cct,,,icim._.
ditticultiv-izl
in a,:enlm
systems,
and
most
The
feeding
for conformality
research
can
,,_,'_it,,,l-
al>o ('t..riciic_.
is therefol'e
communication
tremendous
niav
which
_.,l_.,li,,,_
_.(lUa_iv,_ I1_I)I]!
applications
subject
method_.l
techniques
with
i_I,.._,,,l
e>t..ci_,llv
(litt'crciltial
of this dissertation
shape
a,_l
_.truclu,,.-.
Partial
difference
analysis
of arbitrary
lh,'.-v
problems,
motivation
general-purpose
nelwork.,.I<,
to nlo&'li,,g
of finitely
finite
circuit
offers
the
com-
provides
for thehandlinaoftlle_eometricalcoml_h'xitvwithout
cunlpr_,llli_inaaccut_,,x. [t_i-
has
designed
Frequency
design.
and
waveform
been
successfully
shall
investigate
of the
using
the
out
employed
used
the
suitability
important
finite
(AWE;),
issues
element
An intensive
selection
in shielded
the
are
needed,
A method,
can be used
to alleviate
circuit
and
validity
in antenna
method
analysis.
they
PML's
parameters
In the
referred
this
has
context
analysis
is the
feed
engineering
quickly
become
to as the
asymp-
It has
already
issue.
for simulating
a challenging
for
can
of AWE
is indeed
of the
modeling.
information
techniques.
and
study
of PML
structure
necessary
responses
domain
in VLSI
walls.
the optimal
provide
wideband
to time
evaluation
and
herewith
methods
when
compared
totic
One
carried
domain
However,
expensive
feed
been
truncation
of the
FEM,
MMIC
devices.
design.
problem,
we
Modeling
and
a sin>
pliii('d prol,e
other
hand.
the
is modeled
numerical
without
careful
will he investigated
and
vohage
consid('rations.
generators,
In till,
of accuracy
stripline,
iom.d
wll('ll
ef[icienc\.
line. coaxial
()I_ _it,'
a f('('(I l_,,lx_,I-k
<li.-._er_atiolJ xari_m>
and
microstrip
ilIIt,(',tatl,,'.
f_,('(l lla,(],,l-
"]-tl(,v ilhclmh'
cabh,,
<urn,hi
etc.
In regards
test
and
and
circular)
cone
(all
in consideration
gap
microstrip,
svsten_
to the
design
development
benchmark
patch
antenna,
models
antennas,
etc.
and
applications
of the
of particular
dual-stacked
]t is noted
that.
dissertation
structure,
and
proceed
interest
patch
some
simulation
are
inicrostrip
antenna,
of thern
ring
are
techniques.
not
_he
(rectai_gular
slot
antenna,
necessarily
and
planar
or
conformal.
Referring
magnetic
to the
fundamentals
alence
principle,
element
method
subsequently
tional
the
Dyadic
Galerkin
inclusion
eralization
Green's
as applied
described
and
lossy
materials
are
edge-based
FE-BI
feed
network
methodologies
Chapter
for
guide
basic
structures
The
present.
The
the
to circuit
are
and
emphasis
electromagnetic
fields
presented.
are
in these
computing
where
equiv-
The
and
from
finite
waves
both
is
varia-
in algebraic
form
allowing
discussion
is on
the
techniques
6 discuss
efficiency
The
derived
of the
and
of electro-
conditions,
theorems.
Galerkin
modeling
boundary
is given
3,4,5
a description
related
emphasis
significant
required
the
equations
Chapter
with
with
the
derivation
functional
modeling.
7 is devoted
and
FEM
The
techniques
to minimize
wave
the
anisotropy.
variational
and
functions
techniques.
of general
to discuss
to time-harmonic
and
of the
and
then
we begin
when
the
is on
anisotropic
development
improvement
chapters
gen-
of
for antennas
developing
novel
resources.
specialized
perfectly
matched
truncations
laver
suited
(PML).
an
5
anb,olropic aNiticial absorber usedfor nlesb truncalhm, i, inve',_lieal_'t]irl 1,'Illl'- L,I
perfornlanceand applications.
\\ide}_and svslem responses])lompt us to look at lnore c[tichql/ allal\-i_ t_,t,lto replacetile
discusses
last
the
list of suggested
capability
topics
ployed,
This
man5
will also
domain
of the
and
is included
FEM
and
anal\sis
describe
discuss
tile
ensure
with
recommendations.
and
pertinent
consistency'
anticipaled
of the
specific
('ltat_I'i
Wil}l lhe
applications
concepts
the
apl)roaches.
in connection
robusl
of Electromagnetic
fundamental
we will
frequency
we summarize
Fundamentals
Since
force
development
chapter,
to extend
1.2
brute
a preliminary
In the
work
current
.-\\V[:'.
research
future
FEM
,',
deve]opz_enl.
Theory
theorems
ones
of electromagnetics
in this
in nomenclature
section
and
for
will be em-
reference
conventions
purpose.
throughout
the
(PDE)
con-
dissertation.
The
sidered
vector
wave
in this
research
boundary
conditions
boundary
value
the
half-space
in radiation
equation
--
and
will
the
(BVP).
Green's
scattering
only
be first
will be studied
problems
dyadic
--
partial
derived
to establish
The
functions
problems.
equivalence
are
then
differential
from
the
Maxwell
general
principle,
briefly;
equation
equations.
Various
mathematical
uniqueness
discussed
models
theorem
for EM
of
and
solutions
1.2.1
Maxwell
Tinle
Equations
harnxoriic
uniform
I_ledinm
E and
radian
frequency
dissertation;
to serve
denote
the
and
Mi and
VE
-j_'_.H-M,
t l.1)
VH
j.:_.E+J,
(1.2)
V'._-E
p,
(1.3/
pm
(1.4/
the
electric
the
and
magnetic
which
medium
charge
the
_=
respectively,
e0L
#o_
#o
_o and/_o
being
the
free
space
may
_23
_31
_32
_33
p_ and
fictitious
problems
and
when
_ represent
in general,
p_,
the
as
_13
tA12
_13
/_22
/123
P32
F33
and
are
_ and
be written,
(22
permittivity
p_
this
respectively.
finally
and
tensors
e21
_u21
current,
of physical
_712
F31
with
M,
(11
_11
_u =
Both
_ is the
throughout
consideration;
material
and
e0
electric
formulation
The
respectively.
suppressed
and
under
density.
facilitate
intensity,
and
magnetic
is employed.
permeability,
field
e a'_t is assumed
in the
magnetic
principle
and
and
factor
sources
quantities,
equivalence
permittivity
electric
as possible
non-physical
the
H are
;,i,ci
t)5 [10]
V. .H
where
permeability.
(1.5)
(1.6)
The
the
procedure
two
of (1.1)
to (h'rixe
lhe
fi('ld quanlilies
from
with
_-l
the
tensor
Substitution
(1.7)
(--,
p,.
and
then
(1.2}.
take
equation
-v'g
I,t,oitl>
I,x t'lillliILalitlC
the curl
E=-j_'poV
of (1.'2)into
V" x
(l.lt
and
V p,l.V
way('
v(,ctor
on I,oth
H-V
lake" a dot
sides
H_
t_It_' _,f
cd tl.l!
i,v_,tul
to ol,_ail_
L.7]
-M,
yields
=_2,Uo_o_,..E-j,_',,o.li-K
(--,)
H,
.M,
or
X-7
where
_:_.V
ko = Wv//2oeo
and
can
vector
1.2.2
carefully
types
finite
Dirichlet
medium
space
of the
of boundary
two media
1 to medium
-H=
with
desired
number.
(--,)
P,.
The
dual
"M,
of (1.8)
-jwe0Mi+Vx
(')
(1.2).
(1.8)
_,
Equations
1.S)
is given
by
.J,
and
(1.9)
(1.9)
are
the
form.
and
Boundary
conditions
these
follows
-j,_/.t0J,-V
wave
starting
method,
In what
.E=
-koCh,
Conditions
Boundary
Consider
free
derived
element
treated.
-k02_,.
xH
equations
Boundary
Three
of the
_-'.V
be similarly
wave
is the
V x
Value
are typically
boundary
we shall
Problems
encountered,
conditions
discuss
these
and
must
in the
context
be considered
and
conditions.
Condition
separated
2. The
by a surface
fields
on two sides
(E2 -
El)
F whose
unit
of the
interface
= -Ms
normal
satisfy
_ points
the
from
relation
(1.10)
nlediuml
conductor
aad medium
11'.\I('1.
magnetic
then
current
(induced)
_:_x E2 also
vanishes
Et vanish,'_
and
and
J,
denotes
currents,
given
the
surface
PMC
thus
M_
magnetic
The
They
the
become
homogeneous
tangential
introduced
special
field
where
later
Neumann
In formulating
work
with
then
(1.11)
where
the
(_-'-V
boundary
must
H1)
electric
nlav
r_,l,t_'-t.I_T
conducl_r
(Pl(('t.
surface.
(1.11t
The
H1 is zero
PEC
surface
within
the
currents,
when
the
i.e.
can
support
conductor.
electric
By duality.
V x H2.
Dirichlet
J_ = 0, and
dielectric
boundary
in those
interfaces.
conditions.
cases
Often,
they
J,
applying
the
equivalence
principle
a priori).
The
implication
of this
Boundary
and
imply
M,
are
(except
in
issue
will be
problem
Condition
using
hybrid
El or H field.
be rewritten
as
(from
"lhc-_ll,_'
development.
Mixed
x E)
= -M,.
= J_
M_ = 0 and
are specified
a physical
either
electric
across
currents
in the
and
when
they
l>_'Jt_'ctlv lll_l,,'_i_
(excilationtor
continuity
as fictitious
cases
discussed
and
PE('
current.
since
not support
(1.10)
source
:_ , E,
1 is a perfectly
= 0 on the
surface
by _ x H2 = J,,
relations
I,_.come,
I i:-a
field.
an electric
does
(1.101
If mediunl
x (H2
where
curry'hi
be an impressed
current.
for the
surface
2. r_'spectivel.v.
M_ can either
a secondary
Similarly.
n_agnetic
the
i=
ith
finite
we choose
just
If medium
inside
element
methods,
El as the
the
1 is a PEC.
working
ith medium
then
we usually
quantity,
approaching
V x El
= 0. and
to _ standard
on the derivative
Xeunlann
boundary
of E at the inlerface
colldi_i_lz,
is defim'd.
l)kt'
and
this condition
single
field
in this
use
formulation
manner.
of the
is rather
is used
when
However.
second
order
desired
straightforward
with
the
seen
that
referred
to impose
the H field.
since
it is already
conditions
dual of t l.12*
lJ
working
is often
system
the
Ill many
size may
single
field
be kept
conditiom_,
of conditions
in regard
_hc
millimum
formulation
to as nat_lral
this type
at)t)licatioil_.
implios
l:ortunalelv,
to finite
it
element
simulations.
As for mixed
electric
and
boundary
magnetic
conditions,
fields
satisfy
with
R being
difference
the
above
effective
and
geneous
boundary
mission
line
problem
electric
and
magnetic
the
_ E+
the
condition.
R_
[H] +=0
of the
surface
surface.
Another
a coax
fields
is the
resistive
surface
where
the
condition
resistivity
below
(e.g.
an example
This
and
[H] +=
is a typical
mixed
example
cable,
(1.14)
of a mixed
or other
at a cross-section
guided
of the
line
H + - H(third
condition
wave
type)
occurs
structures),
are given
the
field
homoin trans-
where
the
by
Eie -_
+ FEiC '_
(1.15)
Hie -'_
(1.16)
rHie
"_
and
h x E i = -ZH
where
or load
i_ = -3. and
along
the
line.
Also,
fields
(1.17)
before
Z is the
wave
encountering
impedance
a discontinuity
associated
with
llJ
the _ransrnissionline me,dr of the _uide wav_,_.Iructur,'. i_tit_ii_ati,,..:'l troIl, 1.1"_i
and (1.1(;) in view of I 1.17i yields._h_'relatioll
;t _: E -
which
is an example
becomes
the
tities.
The
of inhomogeneous
apparent
therefore
(In this
mixed
when
left
case.
a form
of absorbing
structures
uniqueness
itly applied
and
tions,
it becomes
associated
intent
Partial
the
with
convenient
various
the
far-field
and
derivalive
and
very
useful
condition.
(curl)
Thi.-
of E. 1 and
undifferentiated
considered
excitation
as a known
when
applying
simultaneously.
quanfunctioil.)
the
FE.XI to
It is basically
Principle
principle
integral
pattern.
these
equations
concepts
principle
will be explicitly"
Together
in radiation
the
1,oundarv
to terminate
with
dyadic
to construct
and
(without
or implicthe
Green's
integral
scattering
for later
func-
equations
problems.
proof)
FEM
It is our
applications.
Theorem
differential
equations
results
conditions.
etc.)
is t, suallv
with
to apply
theorem
of lhe
differentiated
the equivalence
geometries
type)
Equivalence
dealing
the
1.1,',t
(ABC).
and
and
(third
and
condition
when
evaluating
corresponding
boundary
ation,
when
to discuss
Uniqueness
theorem
mixed
is found
for truncation
to this work
mesh
side
(1.18)
Theorem
_: E',-':
both
hand
boundary
= 2b
in terms
contains
condition
Uniqueness
The
side
the right
boundary
wave
1.2.3
H is expressed
hand
guided
ZH
conditions
can
(PDE)
also
can
be solved
be represented
Moreover,
many'
of PDE
models
(boundary',
can
be
using
various
in numerous
initial,
extracted
approaches
forms
natural,
from
given
essential,
the
and
certain
radi-
mathematical
11
si,ecification._of well (letin('d l_lly.'.i(altm,l,h'n_.. lh(' (tm'_.lit,tl 1tL_'llati-.,'- a- _, I,_,x_
to relate the solulion_ and how n_anv(,J_dition_.arc
rect"
solution.
Uniqueness
electromagnetics,
region
plus
the
components
Equivalence
the
tangential
uniqueness
(either
exists
slab.
half
space
in the
The
presence
typical
exterior
analysis,
the
containing
the
expressed
the
currents
are
used
through
the
uniqueness
EM field
When
for the
dete r'mi,ed
field
on bouT_daries.
EM
problem
"cor-
(iUi'SliOll.._i),'citicallv
ii_
II_ a 91t'_ J_
o12 bour_dari.,,
field
and
equivalent
near
or l_lu., th,
in terms
the
theorem
may
currents.
[10].
From
currents
the
region
substitution
the
represented
upper
in integral
or (1.11)
discontinuity
Ji.
the
tangenmay
It can
ensures
be
By 'equiva-
fictitious
volume).
indeed
the
region)
and/or
when
(cavity
a diehw-
region.
(1.10)
same
is l)r ,-
with
we consider
be
Mi
if t t_,
inhomogeneous
coated
interior
can
surface
this
surface
be shown
an identical
region.
is excluded
However,
the
region
remain
that
be arbitrarily
cavity
solved
boundary
a dielectric
probably
(or the
interior
at the
where
aperture
distribution
region
component
sources.
of the
be uniquely
exterior
the
can
platform,
the
current
in the exterior
interior
(1.11)
and
in the
field
field
PEC
are shown
region
to replace
distribution
the
of a large
fields
field
is an EM problem
equivalent
equivalently
we demand
an
geometries
or magnetic
(1.10)
thi-
10
electric
or magnetic)
of interest
tial electric
lence',
theorem,
to be the
In EM
form
of the
magnetic
the electric
In this work,
tric
uniquely
componenl._
of the
lh(' aliswer
_, ;,c[_i,.v_. tl_'
Principle
From
region
offer
the tangential
tangential
scribed.
theorenls
...uilicic_
from
chosen
in our
consideration,
leading
work
the
to an infinite
the
field
current
sources
number
of choices
behavior
in the
interior
in
equivalent
currents
cavity
ground plane
(a)
equivalent
currents
l
cavity
Figure
1.1" (a)
Recessed
dielectricallv
cavity
in a PEC
coated
PEC
ground
ground
plane.
plane.
(b)
Recessed
cavity
in a
13
regi(m
is a]_o need_,(l,
outer
cavity
electric
regions
and
m(,sl
i._ d(,sired.
or magnetic
field
spe(iti('a]ly,
It i, t]l('i'('for_'
Io specify
M+=E
This
choice
implies
is considered,
and
(1.3)
affect
and
region
observed
that
coupling
1.2.4
Integral
recessed
The
lations.
choice
of the
the
containing
a possible
has
been
dyadic
region
used
on
Green's
formulation,
magnetic
of the
aperture.
protrusion
the
outer
contours
FEM
(1.2)
ROI.
It is
interior/exterior
applications.
for constructing
platforms.
infinite
For a planar
ground
depending
varies
end,
plane
structure,
in which
depending
seeking
the
a cavity
on the
using
us start
structures
FEM
an appropriate
region
let
integral
on applications.
of the
a convenient
in the exterior
as shown
Fig.
outside
convenient
To this
regiott
Function
we are
field
('xterior
is nee(le(t.
where
in hybrid
function
the
region
permits
canonical
PEC
when
EM fields
Green's
or absorption
field
zero
are particularly
is the
loading
to find the
the
Dyadic
i1.191
fields
the principle,
with
a.,,
interior
formulation"
of certain
electric
on or near
ple
and
the
currents
field
interest
dielectric
only
functions
presence
with
representation
"total
Green's
of particular
For
(ROI)
Equation
in the
platform
of applying
tail_'lltial
;, >,,H.
interior
when
currellts
,],=
of zero
of equivalent
for the
an(l
fields
details
of interest
Dyadic
equations
the
(ottw'niCllt
e(luixalent
i_
exterior
this choice
system
the
assumption
zero
illustrate
the
The
the
the
the
with
integral
information
the
equivalence
t.o obtain
formu-
the
structure
princiequivalent
currents.
Consider
the
wave
equation
V x V G(r,r')-
..'2poeoG(r,r')=
-I_5(r-
r')
(1.20)
is
11
equivalent
currents
(E,H)=(O,O)
ground
plane
(ROI):
Exterior
Region of Interest
(a)
equivalent
currents
(E,H)=(O,O)
cavity
Region of Interest
(ROI): Interior
(b)
Figure
1.2:
Illustrations
in fig. 1.1a.
of equivalence
principle
when
applied
to the structure
shown
15
equivalent
currents
(E,H)=(O,O)
I
ground
Region of Interest
plane
(ROI): Exterior
(a)
equivalent
currents
(E,H)=(O,O)
cavity
Region
of Interest
(ROI): Interior
(b)
Figure
1.3: Illustrations
in fig. 1.lb.
of equivalence
principle
when
applied
to the structure
shown
Me
Je
(a)
Je
Me
(b)
Figure
1.4
Examples
curved
where
G is the
0), and
of protruding
platform
dyadic
i is the
idem
configurations
in consideration
Green's
factor
function
defined
fff {p.(v
upon
setting
{H.
a planar
with
(1.9)
and
(1.20),
LHS
(assuming
note
the
right
hand
left
hand
= -H(r')
RHS
side
[[
JJS
can
H.
the
M,
identity
(1.21)
N)
(V
H)-G}
side
[[[
dV
(1.2"))
dX
(LHS)
of
V" x J.
(1.22)
reduces
to
G(rfr')dV
JJJ_
and
on
Q = G, we get
(V
the
(b)
P).Q}
=-J_h.[HVg+(VH)_
From
(1.9)
platform
principle.
[v v _+ (v P) q--]as'
P = H and
fff
on
G in association
= - fla.
and
(a)
of tile equivalence
be rearranged
[fi VxG----]
as
+(V
H)-[;,
G----]dS
17
|':quating
the
IAtS and
H(r)
R|lS
vield._
//S
V'
xJ.G(r'irtdl"
(H.
[5 x V'>,
"t
tf
G'--; + (V',
Hi.
tl.2:_,
J ,) S !
where
r and
_"' is the
surface
r' have
volume
been
containing
enclosing
the
To eliminate
the
the
curl
where
infinity.
divergence
the
cavity
since
inserting
to represent
(1.24)
H(r)
into
electric
currenl
can
source
be realized.
and
S'
is i])(,
space.
the
=(V
radiation
field
loss of generality..-ks
dyadic
identity.
xJ)-G-J.
(V
x G)
to get
S' is only
this
half
on .1. we use
theorem
Therefore,
without
distributed
upper
(J x G)
Sommerfeld
It remains
the
entire
V.
and
interchanged
condition
over
the
outer
surface
is typically
(1.23),
the
was
surface
integrals
the
invoked
of the
in terms
computable
to eliminate
the
integral
at
body.
of the
quantity.
electric
This
field
is carried
near
out
the
bv
yielding
i/iv J.(V'xG)dV'
-SL,{H'(fixV'xG)+(V'xH-J)'(_xG)}
-
dS'
/iS,, J-(V'xG)dV'
- if
JTS
where
the
Maxwell
above
field representation
{H'(h
xV'xG)+j_0E-(fi
xG)}
dS'
(1.25)
equation
(1.2)
is general,
has
i.e.
been
used.
not restrictive
It should
be remarked
to planar
or conformal
that
the
cases.
For iil_.talkce,
tiguration.out
over
The
and
in tho
as shown
the
step
-/t,
Vx/Z
T denotes
of (1.26)
vanishes
dielectric
protrusion,
function
which
definition
of the
kind
Green's
G is referred
of both
definitions
functions
[11].
PEC
Go is the
free
platform,
space
to the
condition
the
al. c,rii(.(t
as
>'-G)T(i_
x _)r
leads
function
with
only
in the
the
over the
an electric
=
0.
to the
of first
properties
As
same
in the
dyadic
term
integrand
For
outer
contour
dyadic
Green's
be seen.
vanishing
of the
las_
Green's
platform.
can
kind.
d'q'_:t'o
(1.2l;)
integral
electric
first
> E)}
dS'
the
) is the
symmetry
The
this
term
in
equivalence
dyadic
Green's
to
Go(rlr ) +
function
given
4 lr-
(1.27)
by
e-ik01r-r'l
Go(rlr') -
and
0. the
(V"
Green's
= Go(rtr')-
dyadic
_,,_
for t't)lll't)l'illltl
i_ltegral
(V'
is to define
fix
G reduces
Green's
J-_V
integration
equivalently
from
surface
S' is coincident
An alternative
can be proved
G(rlr')
where
reduces
dyadic
lhe
If G(rlr'
provided
function
ai_d (ontt)arod
of the
as _ x G
platform,
ii11(',.zrati_,ll-
_-'t'oJ(h E).(V'-G)}
field.
defined
to as the
>,lt'/It'l IlI'l'>.
[tit'
x H)d'q'
electric
the
of
we rewrite
operation
platform.
11.2-,!
l,e examined
xG)
this term
satisfies
(1.26).
For a planar
dS',
{H.(i,
on the
to the
shall
x-G)T'(i'
to the
first
('OlllOlli'>
To this end.
a transpose
of the
conformal
(1.25)
{ (_
is proportional
platf.r1_t.
the outer
structures.
function
not
plus
representation
protruding
h,t_,:_
of a P["('
in fig. 1.4.
platform
field
where
im'_.ence
r'l
1!5
and
G-o(rlr')
Inserting
(1.26)
and
H(r)
(1.27)into
H"_(r)
1
I + _-_-_-
(1.25).
+ H"f(r)
) (,o(r,r
.
we obtaill
+ 2jL'}0//
Go(r[r')
J j.
This
is the
boundary
desired
integral
form
equation
of tile
magnetic
for a planar
,t
field
platform.
- (;_ _ E)d.<"
(1.2S)
representation
used
to eslablish
lh(-
CHAPTER
Finite
The
since
Element
finite
element
several
it is observed
engineering
netic
that
grew
problems
approaches
This
chapter
indispensable
associated
merical
with
and
electromagnetic
in the
have
these
literature
been
used
Galerkin's
problems.
related
for domain
[13].
Of interest
This
technique,
the
stage,
usually
some
especially
is the discussion
to general
of the
least
(refer
2O
to Chapter
the
With
has
its
been
theoretical
considered
interesting
of the
and
therefore
the
issues
in terms
Anisotropic
7) and
analytical
FEM
anisotropic
studied
etectromagnetics.
and
[13].
2 describe
though
electromag-
needs.
applications
are
nineties,
in electrical
because
practical
1 and
applied
is one
to computational
truncations
EM
even
when
FEM
techniques,
These
in this context
the
(EM)
early
complex
meet
in development
topic
with
and
is primarily
Section
equations.
techniques
eighties
increasingly
for various
FEM
are still
This
numerical
as follows.
to electromagnetics
late
no longer
other
of the
applied
associated
become
exploited
the
basics
implementation.
functional
over
is organized
been
ill the
techniques
features
fundamentals
pace
designs
numerical
to construct
has
volume
rapid
and
in Electromagnetics
Especially
publication
in engineering
investigated
formulations
(FEM)
[12].
in a fairly
attractive
extensively
ago
the
or other
numerous
Analysis
method
decades
II
of nu-
variational
and
lossv
documented
materials
the general
21
veclor
or _ensor form
for the
t-E3I.
sp('cial
('ascs.
With
The chapter
analysis
given
this
require
Functional
ally,
FEM
many
the
standard
(1.S)
the
or (1.9)
conditions,
FEM.
the
feature
of merit
method
in a final
be used
_x,t,rll,
m,tv
of the physical
of effort
_h.ri_._i,,_-
},. r,',_,_r_t,',l
a>
Th(' formulas
fox computations.
(BVP)
of the
cards
convenient
may
in general
section,
other
and
sometimes
represent
Also,
Last,
the
a subset
but
not
expressions
the
function
can
readily
incorporate
hence
as can
cases
physical
this
be seen
may
of which
the
wave
least,
the
variational
based
on Galerkin's
equations
formulation.
quantity
(e.g.
treated
functional
method.
in low
provides
chapter,
to that
It is
approach.
formulation
be rigorously
is identical
t.o
boundary
meaningful
in this
those
approaches
vector
into
be
a discrete
to soh'e
important
physically
and
be used
can
With
onto
for the
constraints
a true
space
can
two
FEM
which
applications)
system,
functional.
of the
and
non-self-adjoint
variational
method
version
Tradition-
in practice
a continuous
A functional
analysis.
encountered
of a certain
one
in this
of functional
functional
to validate
aid
becomes
transmission
technique.
the
therefore
resistive
symmetric
Galerkin's
1_.cc>-arv
of e]('ctromagnetic
problems
variational
functional
power
value
to project
and
the
with
procedure
as a natural,
frequency,
developed
is discussed
Furthermore,
i>_,tro[,ic
amount
extremization
sources,
regarded
minimum
to the
space,
problems
formulate
i1_ It.,
the discussion
the computation
boundary
related
expansion
physical
tional
first
Rayleigh-Ritz
finite
with
p_,:>il,t_,
Formulation
was
equivalently
type' of formulation>,
with
in this context
The
be used whencw'r
is concluded
in association
2.1
can
will
the
varia-
to result
obtained
from
formulation
.).)
('om.ider
a tyifica]
radiati_m
or _catt('ri.g
pr.[,h'n_-l.,wn
ill
ll_. 2. I. _l_'I,'
tl..
Z
Radiating
element
Ground
plane
._ Y
Figure
radiating
the
2.1:
elements
Illustration
(or array)
radiation/scattering
shape
of a typical
(cylinder/sphere),
function
is not
available.
or (1.9),
which
can
are
conformal
enclosed
geometry
can
or even
a doubly
be a planar
described
configuration.
Q. The
platform
in a region
curved
In Q, electromagnetic
be concisely
antelma
ground
surface
fields
using
plane,
certain
in which
case
satis_"
a linear
surrounding
the
operator
wave
(I) denotes
Ki
is the
and
may
the
field E or H,
V x/a_--1 T
/;
source
(V
term
be explicitly
Vx
x =-1
_
.Vx
associated
given
K,
=-j_'poJ,-V
K,
= -j_'_oMi
the
Green's
equation
L: given
Z:O = K;
where
canonical
(1.8)
by
(2.1)
and
with
(k02_-)
for
electric
(ko2_ .)
for
magnetic
the
impressed
electric
field
and
field
magnetic
(2.2)
(2.3)
currents
by
(_-1.
+ V
Mi)
for
electric
. J,
for
magnetic
field
field
(2.4)
(2.5)
2:1
As already menliolu'd. is a linear opera_or arid
ore'
sxnlmotric
tuxlc_ional
has
diolectric
lensorb
_ an_l ?. the
perlinem
1
= 7_ < 0-
> -
_ll_,x_ _ti,T
of lhc t,ri_itml
t,,r
t'[)li
tit(' form
.T()
where
the
inner
product
< A.B
(with
B" being
the
complex
for both
The
lossless
equivalent
functional
BC's).
variational
of B)
for lossless
media,
or more
generally
the
with
the
boundary
essential
value
as the extremization
boundary
problem
conditions
is equivalent
of the
(e.g.
to the
Dirichfollowing
model
=o
Essential
Because
interest
the
to us,
effect
boundary
approach
of (2.9)
physical
mathematical
of a numerical
less storage
of complex
we restrict
Dirichlet
many
as
('_,.s)
>=fnA-BdI:
problem
in conjunction
Specifically,
(2.(;)
media.
variational
(2.6)
>
as
conjugate
and lossv
< qb.K,
>=fnA.B'dI
<A,B
let
cat_ readily
most
conditions
ensures
problems
models
system
requirements.
(2.9)
Boundary
materials
of our
unless
should
a certain
therefore
is always
on
the
discussions
otherwise
a symmetric
retain
Conditions
desirable
in this
specified.
numerical
symmetry
reflect
since
resultant,
this
system.
system
chapter
it leads
to homogeneous
Therefore,
the
This
is significant
property
property.
is of primary
and
the
Moreover.
to more
efficient
variational
since
corresponding
the
symmetry
solution
and
2_
"lh(' ,,xist('n('vof th(" fulJ('tional (2.(;) re(luir_.>tl,.
proper_y
usually
defined
(I) and
does
the
qJ represent
numerical
any
system
concern
is no longer
the
medium,
type
whose
methods
from
is the situation
where
the
),lathematicallv
dielectric
material
tensors
is more
often
problems
seen
is still
partial
(2.(;)
holds,
remain
differential
exists
is the
are
no such
presence
ll()t o111\
SVlnlm,tri(.
The
an early'
operalor
llle
functional
of a lossv and
because
in
anisotropic
or Hermilia,.
development
stage
of a svslenl
a l_alural
not symmetric
nowadays.
at
If (2.10)
meaningful.
there
example
,2.1()b
funclional
physically
A typical
for those
,l,.!:,,:,:.
>
functions.
the
to (2.6).
of problems
(I). _
admissible
becomes
self-adjoint.
case similar
two
derived
minimization/maximization
Of most
/.." I,_...,//
to sat isf',
where
_,i,rralor
and
of finite
it involves
this
element
numerous
challenges.
Traditionally,
using
with
these
available
three
loss-free
A few
approaches
The
later
with
[15, 16]).
tensors
were
in 1980's,
the
usually
doubled
in size.
Konrad
eigenvalue
worse,
the
As indicated
to reduce
to the
and
to formulate
fields
in anisotropic
subject
structures.
Unfortunately,
the
during
period
systems
(even
systems
in [17]. when
form,
consistently
with
the
derived
a non-standard
the
size of the
aid
in this
but
increased
variational
led to non-
of an
manner
eigenvalue
system
with
in his study'.
in this
that
dealt
a 3-D FEM
to be Hermitian
of publications
numerical
standard
simplified
assumed
number
authors
fictitiously'
electromagnetic
therefore
to waveguide
by different
Even
were
to represent
non-Hermitian
system
was manipulated
approaches.
applications
reported
and
problems
components
media.
typically
standard
numerical
vector
years
physical
adjoint
were
system
was doubled
25
again. Similar reports wer_'al_,_('enin Inlet pal_'r>!l_
propagation
reporled
inside
in ibis
In what
anisotropic
we generalize
problems.
the
to these
two variational
Pertinent
operators
(2.8)),
the
since
one
design.
In these
the
is the
natural
PDE
for
what
_ can
lwo differem
with
tla-
alld
anis_trotm
nwlhods,
the Lagrange
Galel'kin's
functiona]
a self-adjoint
melhod
I.','i,
of,'
wil Ii
nlull iplier,
is thell
qb,_
function
is given
for
operator
> -- < O, Ka
solution
Media
no longer
a generalized
function
as in (2.1).
ca_
col_ll_a1,,d
exists
inner
--
for non-Hermit
products
necessary
for
(see
71alural
ian
(2.7)
or
functional
functional
> -
of the
< k0,K
>
original
Similarly,
(2.11
PDE
q_ is the
problem
solution
and
function
Is
of the
that
be derived
= K_
(2.12)
from
< O,k_
with
aIlalv_i-
Io lossv
Lossy/Anisotropic
/;_
where
lhal
the other
definition
we consider
unknown
such
formulation
functional.
variational
obtain
cases,
right-hand-side
adjoint
amt ,calleriIl-
techniques.
_-=<
where
and
pertinent
no matter
cannot
FE.M
we show
system
Functional
As is known,
1.he
Specifically.
auxiliary
be used
2.1.1
radia_ioll
_,_'
context.
follows,
electromagnetic
media..X_,
>=<
,,,qJ
>
(2.1:3)
- _.
It should
possible
be remarked
constant
coefficient)
that
the
functional
if is self-adjoint.
(2.11)
reduces
Also.
the
to (2.6)
(except
original
PDE
for a
and
its
_(t
_1(ljoilll
COIIIITf'II,aI[
(_.I_P ('_I!I)('
l('(t,V('l('d
|[|Ft,Holl
|h_'\iLliil|it,l_,1]
[_It,_X'---\viIll
respecl
here).
This
Io the
functions
_ and
.-kfler discretizalion
can
be shown
qJ. resp_,ctixcly
is carrh,d
as follows.
(I:) =
oul.
lh,.
ql},,. >iillph,
[iIlal
V,
is the
corresponding
basis
Z.r,V,.
function
expansion
nunwrica]
i- t,II,ill,._l
Let
,
where
d,'ri',ati_,l,
yjVj
('_.l-l)
j
used
coefficient
for both
s. Insert
unknown
function.,
and
.r,..tt,
are
the
(2.11 ) yi,'lds
t,l',)
i
Upon
the
performing
the
two decoupled
differentiation
systems
with
of linear
respect
to x, and
Yj individually,
equations
(0:)(:)(,)
=
Qy
where
the
matrices
In general,
These
holds
Even
this
"-'uOx#
relations
for the
The
overall
system
does
the
column
vectors
K _. K _ are
O_
._-- 13
<LIV3.V,
Q_
K_
<Vi,
Ky,
<Vi,K.>
a loss of symmetry
given
by
>
K>
Oz..,.,,7 _ O_a.
However.
of the original
OT./ =
problem,
QYi = (O,y)r.
but the
symmetry
system!
requirement
there
(2.16)
K_
indicate
storage
though
Q_, Q_ and
we get
is a function
is an auxiliary'
not
require
system
storage.
of :/2.
needed
where
N is the dimension
to complete
the
analysis,
of (2.16).
in practice
2.1.2
NIedia--
II
conslrucling
to incorporate
consider
the
same
the pertinent
additional
PDE
functional.
constraints
model
Tim l.a,.,rallgc
to a system.
as described
in (2.1).
n_ultit,licr
"lo i[luslxalc
and
we firsl
i> i>u,ll\
l]_c lcc[llliqm'.
rmvrilc
il as
qb __ K, = 0
Next.
we assume
an expansion
unknown
function
If (2.17)
is regarded
and
q) and
the
space
where
multiplier
as a "constraint".
the
(2.17)
solution
function
is defined
A are expanded
we try to add
and
in the
the constraint
solxcd.
same
-l'he
space.
to a "'null'" system
get
functional
is now used
the
Rayleigh-Ritz
procedure
to formulate
to both
the
FEM.
_5 and
A using
(2.18)
As described
the
same
above,
set
on applying
of basis
functions.
viz.
we obtain
(2.2o)
i
Carrying
out
differentiation
with
j
respect
to xi and
5'j, individually,
yields
(2.21)
where
Or
decoupled
= OU. as in (2.16).
subsystems
of the
\Ve observe
same
size.
that
The
(2.21)
properties
is similar
of the
to (2.16)
subsystems
with
Iwo
are
also
.),,_
sinlilar
Io lho_e
in (2.1_;i.
may
t,e regarded
now
virtually
self-adjoint
and
the
auxiliary
necessary
(for
considered
tions.
assumed
and
the
as a special
problem
approach
ob_cv\_,
ca_'
(i.e.
K,
= 0).
multiplier
subsystem
becomes
multiplier
to the
results
in an
.-\gain.
unknown
fornmlal
either
redundant
ftmction
interesting
iol_:
coincidence
with
ul,'ll
svsl_'_l
nmltil)li,r
the
Galerkin
i>
atq)roac]_)oi
the
using
t'l)tl
-Wl,'_ll
in a svmlnetric
case.
expanded
tidal
[1 l tie adjoi,1
([_oI
(for adjoin_
later
_,,, t_l_i_t/,.
a,li,,iI_l
I,_, leI_latkc_t
will result
In the
ii_ullil,li,._
;1 la,l_,,_,_'l,'_,tl-
il sh,lltd
technique)
technique).
tile l.a,_'ranm'
wild're
the above
Lagrange
the
tha_
I, 12.11!
is considered,
identical
This
\Vc further
sanw
,illtalk
[Jasi,
technique
),'
t_J_c-
h'scril,ed
next.
2.2
Galerkin
Galerkin's
method
ditionally.
the
testing
case
determines
unlike
by a testing
in the
functions
FEM,
and
method
prescribed
q_.
one
constraints
integral
asvmmetric
Galerkin's
method
from
boundary
the
symmetry
feature
approach,
Galerkin's
conditions,
of the
equation
does
the
not
linear
Tra-
emph)ys
dense
resulting
method
method.
numerical
always
operator
lead
of a
system.
solves
the weighted
PDE
as
_ are both
seeks
with
element
to obtain
Apart
< q/,/Z_
where
the finite
in conjunction
of the
the variational
process
to formulate
used
expansion
system.
problem
Also,
technique
and
However,
to asvmmetric
PDE
is now considered
Galerkin's
same
system.
Formulation
the
defined
solution
and
(2.22),
>=<
in the same
for the
with
_,Ki
function
unknown
the
>
(2.22)
space.
function
aid of another
Specifically.
which
arbitrarily
in Galerkin's
satisfies
chosen
certain
function
2_
Similarly
Io tim variational
used
to project
The
malhematical
entries
are
the
be defined
proper
the
coefticients
of the
a tini_e
rephrased
expansion.
space
t{ii:., tm>_cdur,'z_l,_\
discr(qc
T(, seek
"lhe
to ensure
a di:cze1,
testing
that
,-cparal,h'
the
_l-,,
Ililtwrl
-1,a_,'.
_t_lulio_J sol
funclion
qJ
original
PI)E
t,,,
u ij_,.,,
O[ ('t)Ul'SC.
IllU.';I.
is solved
with
if the
linear
operator
in a symmetric
is defined,
the
is self-adjoint,
numerical
final
the choice
svstem.
discrete
system
Otherwise.
in general
of the
_esting
no mailer
does
nol
how
exhibit
property.
We observe
even
when
cases
(as
leads
to the
can
that
product
symmetry
onto
is thcJ_
discrete
_ = results
inner
space
1}_(' t{avh,i,..,h
conditions.
It is obvious
the
same
boundary
function
contilmous
problem
in the
approach.
that
the
Galerkin's
method
corresponding
considered
natural
when
same
be demonstrated
system
as follows.
functional
describing
numerical
x,yj
can usually
the
as the
Inserting
be applied
does
functional
desired
< Vj,V,
>=
exist.
Also
approach).
portion
(2.14)
not
to any linear
into
Yi < Vj,K,
in the
Galerkin's
in (2.16)
(2.22),
operators
and
we readily
general
method
(2.21).
This
obtain
>
or
As assumed,
vanish,
is an arbitrary
function.
the
term
in the
curl3'
bracket
should
yielding
Q_x
which
Thus
is exactly
the
same
as the
= K"
subsystem
(2.24)
derived
from
the
variational
approach.
II i_ holed
as the
testing
though,
the
This
is due
expected,
2.3
that
function
entire
system
to the
obtained
Total
Field
whose
relative
is twi(,
the
size
difference
numerical
systems
Scattered
we focus
on a general
conducting
electric
permeability
1_l_]tii,lh'r,in
mlmerica[
and
In this section,
a perfectly
.,,a_.
fundamental
the
where
qJ. the
Lacran_'
and
>v_.len_
of ll,al
two
of interest
r,._ll.
\ia
,_I,],T ,_,<lJ
.\_La]_ ,,llx
(.;ah'lkill'_
lechlLiques,
are
Field
virt uallv
lll,'T[a,,l
l.aidh,>>,
a,
idenlical!
Formulations
scattering
permittivity
would
derived
of the
(PEC)
I},_, variati<,_,a',
problem
body
as illuslrated
is coaled
with
in tig. 2.2.
a dielectric
_d. respectively.
laver
(Note
that
f2t
_d:
dielectric
_:-: free
_:
coated
space
absorbing
region
(_d,_d)
(_l = P/ = 1)
layer
(_,_)
Fp: boundary
of the
PEC
Fd: boundary
of the
dielectric
body
coating
and
F/: boundary
of the absorber
and free
F0" PEC
boundary
of the outer
absorber
Figure
2.2:
for the
purpose
The
main
has
Illustration
situation
been
of a scattering
of generality,
with
analyzed
the
absorbing
before
problem
medium
boundary
(see e.g.
free
space
space
region
setup
for scattered
is assumed
anisotropic.)
conditions
[24] or [25]).
region
field
for truncating
However.
formulation.
the
two issues
FE.M
do-
associated
:31
with this
of them
the
type
ofl_roblem.
is the equivalence
scattered
tedious
one
field
is used
to assuming
that
relates
including
accuracy
inclusion
and
care
tion,
there
be taken
although
the
conformalitv,
when
conditions
a metal-backed
absorber
cussions
on the
FEM
to scattered
boundary
and
transition
contains
Fields
we begin
be readily
with
ing the
proof.
ma_ched
absorbh_g
male-
handled
absorbing
materials.
etc.
tlowever.
are enforced
in the
FEM
formulation.
the
FEM
implementa-
simplifies
inhomogeneity,
which
we extend
where
our
the
absorber
again
laver
requires
theoretical
treatment
disof the
Boundary
wave
by duality).
Conditions
equation
To proceed
in terms
with
of H
Galerkin's
(the
method,
H tt as
H *_=t and
source
wilholll
be described.
the
H ** = H _t
where
equivalenl
the
high
1,,,
lnforlunalvlx.
treatment,
representations,
and
ca_
inside
layer
will also
wt_,'l_
conditions
To this end,
field
conditions
Scattered/Incident
write
are
of boundary
artificial
those
region
ease
nwll_,d
issue,
to use artificial
of the formulation.
we first
an hnportanl
formulations
()lu'
of lhi., equivalc_l('c
advantages
presentation
may
l'roof
are several
a careful
formulation
II_c Gah'rkil_'s
perfectly
additional
to fig. 2.2,
ill tI.'lil,'raturc.
introduced
FEM
Referring
alld
variabh',
two
multilayered
2.3.1
variational
it is nevertheless
these
addr_>.od
recently
control,
introduces
must
Moreover,
to the
out.
the
as the _orking
but
As it turns
their
is used
cumbersome
issue
betweeIL
and
Another
riM.
_ have
wave
scattered
equation
L{:'
V-Vxe_
with
(225)
+ H mc
and
incident
field,
the
testing
function
.VxH-k02_.H
respectively.
Next.
weight-
V yields
df_=0
(2.26)
with the derivatioi_ of the weak f_,rm wart, (,(ittatit)tt. it i> tlt'(t,s,,_trx
certain
constraints
First.
since
well as the
on the scatter(,_t
the
incident
metal
back
field
wall
a11ct ill('i(lerll
is _ol allowed
F0, we note
ti(,](ls wilhilk
to pass
it, itl_r_,(lu_,'
through
tl_(' absort_('r
lav('r
a,,
that
r E .Qd + -Qj
(2.27)
Htt ( r ) =
with
the
incidem
HSCa
H
*e=t_
r E otherwise
field satisfying
{V x =-'
e_ .9"x
=0
'k
It is thus
regions
evident
_
The
the
+ fli
boundary
that
the
and
the
conditions
field satisfies
wave
Note
that
(2.28)
r E otherwise
the
homogeneous
equation
on H *_ can be readily
(2.25).
decomposed
:_ 0
inhomogeneous
field decomposition
is likewise
scattered
+
wave
equation
in _a.
derived
in accordance
by consistently
with
(2.27),
the
applying
electric
tions
one
in the
should
be cautioned
dielectric
region.
that
That
E i_ and
(2._9)
H in_ do
field
conflicts
is assumed
dielectric
that
(2.30)
there.
with
what
to exist
After
is against
one would
in the
a quick
Maxwell
not satisfy
Maxwell
equa-
is,
which
field
as
However,
in
r e fie
intuitively
dielectrically
glance,
theory.
one
assume.
coated
would
In reality,
(2.30)
Conventionally,
region
immediately
it can
the
f_d as if there
arrive
be proved
incident
was
no
at a conclusion
thai
if E ''_' and
33
H'": ir)(leedsatisfied Maxwell equatior_._
in .(),.,,.
a cor)/zadiclo)v t)ouzl(larvc_)_lditi_)r_
would
]rnmediatelv
result.
This
of (2.2.5).
In view of (2.29)
Imposing
the
condition
(aJ) be seen
and
._laxwe]l
of total
field
I)v taking
equations
tangential
a el)r] _,p('raIion
in di(,h,ctric
continuity
nwdia,
at the
w(, would
boundary
Ilav_,
I'_ would
yield
which
is a homogeneous
(2.27),
upon
tangential
which
This
tion
_.VxH
Neumann
the
continuity
at the
inconsistency
and
boundary
boundary
and
Neumann
boundary
(2.32) was
equations.
the
However,
to keep
in mind
derived
the
if we start
with
of total
field
condition
on the
field
only
basis
within
decomposition
that
....
condition.
incident
the
=0
However.
imposing
Fa, we get
that
It;
condition.
boundary
assumption
necessary
'Vx
operation
is because
the
Maxwell
is therefore
(2.27)
of the
dielectric
(2.29)
holds
decomposiregions
is artificial
true
when
also
and
it
deriving
conditions.
As a rule of thumb,
the
curl
is an inhomogeneous
(2.29)
satisfies
taking
Ir.+-_f
incident
as if the free
field
an appropriate
inside
space
dielectric
was replaced
interpretation
media
with
ezisted
the
media.
of the phenomenon
in the same
fashion
Mathematically,
should
read:
as i7_ fre_
spac_
this
implies
the
condition
inc = =-1
3_'_oE
ed V x Hinc
r E ft_
(2.34)
('onsistentlv
formulalion
mann
conditions
tile same
i_ x H __' = K:
- f_ x H ''_ (K_
unknown
cmrenl)
F/
alict N ,ll-
(K,
unknown
current)
Conditions
Boundaries
Conditions
=-1
e4
lPp
Fa
fix
Fj
fix
Fo
fix
and
.V
=-1 .VX
%
negative
Hint
=-1
a .Vx
=fi
H,cat
{} ]+-{}
sides
xej
.V
=0
_ and
e_ and
of a specific
H *_=' should
take
the
values
at
boundary.
Method
to (2.26)
and
A.V
divergence
HSCat =-fixe
.Vx
q. 4
{e-(1. V x H "_' }i+ -h x =-,
+.V x
{} + denotes
Galerkin's
the
as l)irhllhq
{w_'l+- I-V_'l_
}=o
Neumann
and
cla._siticd
Fe
Fo
Returning
"l-hey arc
r_.quirvd
('onditions
Fp
2.3.2
procedure.
comtition,
Conditions
Boundaries
positive
bouI_dar_
as follows.
Dirichlet
where
the other
theorem,
in view
xB=(V
of the
vector
identity
xA).B-V-(A
we obtain
Intd + Int/+
the
corresponding
Int_
(2.35)
xB)
= 0
weak
form
wave
equation
(2.36)
3?)
wherc
Iilt4
--
j_
(H '-_
-/<_V-_..IH'
+ H'"
,/(_.)
:'-H"
V.
b o x _e
.V
>:H
d.';'+
(:,
V.
[_1 x t e
._"
H scat
(2.37)
I (=-'
V"
_2
x _I
VxV.%
i (:'
V.
.V
rio, ill.
fi2 and
respectively.
They
fi3 x %
-_"
x H "_*
fia being
the
unit
all point
Invoking
the
boundary
integrals
on Fv in (2.37)
integrals
on Ff in
away
C/
=_|
H ''_'_)
d.N'
the
reduces
x (_
V x H '__'')
ds'
(2.:3<,))
at the
center
boundaries
of the
above,
PEC
F v, Fa, Ff and
body
(i.e.
we observe
vanish.
Also,
the
outwards).
that
sum
F0.
the
surface
of the
surface
to
xef
V.fi2
-i_2
dS
F0 in (2.39)
(2.39)
V-
as tabulated
on
(2.38) and
(2.3S)
d.'_'
normals
from
conditions
and
df_+
with
-ill
Into
H .... )
d'2;
V.
.+
=-'
.Vx
dS
(2.40)
Similarly,
the
integrals
(involving
H sc_t) on Fd in (2.37)
and
(2.38)
becomes
Note
the
that
both
excitation
disappear
an electric
integrals
on the
(2.40)
right
if the
permitivity
field
formulation,
and
hand
(2.41)
side.
Of course,
is continuous
a term
will not
similar
across
contribute
the
the
to (2.41)
to the
excitation
boundary
will appear
system,
term
Fd.
(2.41)
By
but
to
would
duality,
for discontinuous
for
3(;
pern,eabililv.
enl
ItLi> observation
incidenl
shown
the
field definition
in (2.27).
inlegral
system
Furthermore.
on
reduces
two sides
the
Ve involving
second
of th,. t,oul,(larv
lern_
H '_: in (2.37).
il_ (2.-11)
After
I,.
ml Ill,, i_llL,'I-
a> nlatl,,'llla_icall\
a >inlph'
llmnilmlaliolk.
oul
wiltl
_hc
tiltal
to
= ....
/,.0V.
_7.
H ''_"
V.fil
xe/
.Vx
dS+
V._72
where
?_ is again
is the
desired
in [24] using
2.3.3
the
weak
the
relative
form
corresponding
to
representation
and
where,
mebility,
with
in terms
of the
different
form
presented
the
wave
with
equation,
formulation
to employ
To
use the
respect
and
to the
a similar
for isotropic
functional
end,
specific
region.
This
equation
was obtained
to obtain
the
media.
and
_
Also
of linear
denote
similar
decompostion
scattered
field.
system
D.
formulation
it is intuitive
to begin
with
the
equation
total
field
functional
:,
VH.%
field
in Appendix
the
this
.+n1+a.
respectively.
proceed
of the
(2.42).
1
= ,_
as before,
.V
Method
of interest
_'(H)
permittivity
functional
Variational
It is now
x(/
the
.VH-k0_H._
corresponding
to Galerkin's
relative
method,
(2.42)
and
that
the
this
detailed
dQ
(2.43)
permittivity
one
It is unfortunate
than
.H
would
express
approach
now
the
and
logically
functional
will result
mathematical
per-
proof
in a
is
:17
The db,crepancy
a valid expre.,sion
call
in section
the
application
functional
with
this
explicitly
arises
why.
2.1.1
that
the
adjoint
is given
bv a generalized
generalized
functional
the
variation
variational
technique
rather
to note
(2.44)
that
Ha "
seems
retically
adjoint
on the
Apart
required
subsection,
that
from
the
V in that
original
will not
system
-V
adjoint
l_c-
nl_'_lia r_'cluirc-
which
the'
f_vr'tirwrlt
necessary
In a source-free
lo })egin
regioil.
(2.11)
PDE
does
difference
the
FEM
variable
from
9.-14)
2.45)
proves
function
the
is similar
field
the
to that
(2.45)
to Galerkin's
interesting
quantity
However.
as a solution
above,
to avoid
and
identical
It is also
adjoint
latter
to be a solution
as mentioned
here
and
V in (2.11).
whereas
not have
valid
problems.
is defined
problem,
(2.44)
dl
Ho to get
method,
former
system
be discussed
obtained
x H -/"0P_ " H
_H=0 = 0
H)
testing
the
that
concept
to derive
final
(2.43).
than
in electromagnetic
of the
function
the
and
with
in (2.11 ). It is lherefore
to Galerkin's
place
of the
testing
aTld to>>\
[ogether
of this version
operators
compared
from
system
anisotropic
V" c,
method
linear
to take
differs
admissible
dure
functional
once
form
is imposed
for any
tq)erator
system.
form
&T(H_,
The
_ PI)E
thepreselweofge_wral
of ari auxilarv
takes
a>_.uml)_iolk
_h("
the corresl_,ndin
_-(H_ H)
and
fronl
to the
the
is of course
adjoint
for the
system.
proce-
in the
One
theo-
an arbitrary
mathematical
presented
repetition.
Ho
function
is just
H= in
previous
can
be assured
identical
to (2.42).
3"
2.4
Parameter
An
Extraction
accurate
full
methods)
or current
to obtain
certain
terns,
far
involved
practical
network
via
feed
full
or circuit
platform
the
chapter
used
in the
1, where
to consider
2.4.1
the
patterns
and
RCS
such
as input
in later
chapters).
characterized
Consider
distribution
the
with
planar
we can then
ward
approach
plane
is use
for this
of the
includillg
a de-embedding
of the
dvadic
and the
near
free
field
7.
the
For
fox feed
a non-planar
general
Green's
space
distribution
is required
in chapter
in terms
interested
such
as gain
feedline
and
radar
respect
to the
3-D
cavity-backed
antenna
S of the conformal
to evaluate
computation
equivalence
extraction,
process
from
radiation
proceed
and
the
obtained
parameters
on the aperture
T',t,,'
discussion
function
Green's
in
must
be
function.
Pattern
impedance,
Both
I'l)V
t"or in_rica1,.sw-
data
after
be
currents
,f_,I
simulations.
will be discussed
can
ti,'l,l
platforms
de-embedding
we are mostly
related
for output
evaluated
The
formulation
and
other
quantities
analysis,
analysis.
case of antennas,
and
readily
readily
equivalent
Radiation
In the
can
ll'al
I <ti_tri}_ulio11:.
of non-canonical
evaluation
the
ltle
on application.-,
be needed
or circuit
wave
field
lechni(lu_,>
network
be
pr,dict
depending
may
presence
modeling
far
only
based
models
parameters
is achieved
can
parameters
numerical
for antenna
Antenna
analvsi-
(for integral
field evaluation
process
wave
in the
principle.
in their
and
radiation
axial
S-parameters,
cross
spherical
presence
patterns
0 and
in fig. 2.1.
The
of an infinite
To do so. we define
the
near
field
will be disccussed
(RCS)
is obtained
scattering
(The
coordinates
as shown
antenna
ratio.
etc.,
section
and
most
be
0.
Once
from
can
the
field
conducting
ground
magnetic
current
normal
to the al,ertuw
surface.
Tlw
eh'c_ric
F(O.o)
the
function
electric
field
./_
(Oe-Jk:
4rrr
/L
is typically
S,_. Introducing
t,olcrJtial
4=r
where
vector
expanded
this expansion.
4rrr
(OC
-jkr
( E'C x
in terms
(2.46)
(2.-161
d.g"
: )erkr'
of the
surface
vector
t,a>i>
becomes
E_
__
(".4T)
4 rcr
m
where
elements
denotes
on the
the
sum
aperture.
of the
The
surface
far zone
integral
magnetic
over
field
each
of the
discretization
H becomes
HT = --j,ZFT
where
HT
represents
0 and
0 projections
the
transverse
are
given
Ho = -j..
J kr
4rrr
magnetic
field,
whose
by
'ce-_k_
4rrr
_,,f.oe-
component
( 2.-1S )
- sin 01:- }
Po
(2.4._))
O(_--J kr
in which
in the
in terms
Po = -j
curly
47rr
{'} and
brackets.
of Po and
The
kr
p,
Pe = -j
RCS
{.},
of the
and
{.} stands
for the
t (t = 0 or 0) component
corresponding
can
terms
be represented
Po to yield
,2.2
_ __01p,l__
4_
A2-Z_ IPe
i_
(2.50)
t(I
whereA and Z.
respectively.
it,,.
= \/)+u,/_t,
In (2.,501.
In practice,
radiation
above.
The
imum
this,
field
case.
may
we represent
o"rcs to avoid
here
The
source
rather
therefore
the
far
a repetition
value.
Of interest
to expres.,,
in procedure
by an interior
tering
wave
ta)tIt+ali/t'(l
first.
pattern
radiation
excited
meter
difference
free space'
Ill(' incident
it is customary
to &2 or to squared
The
are the
reason
the
of post
field
in terms
processing.
manner
with
in radiation
incident
relative
intensity
sanle
normalization
is that
than
is the
field
is the
in the
wave
intensity
the
to a nlax-
the
antentta
H + as in the
in the
above
Specifically,
respect
mode.
plane
of the
a_, nwlltk)m+(l
fat zone.
calculated
i>
scat-
To get
quantity
fornmla
FCS
o.rad
__
CTt
(O.fCS)max
is used
for radiation
2.4.2
Gain
Gain
and
(G) and
analysis.
Axial
Axial
antenna's
performance.
terize
far
the
zone
By definition,
(2.51)
Ratio
Ratio
It is also
features
the
(AR)
gain
of the
that
these
parameters
two
which
quantities
indicate
typically
the
charac-
antenna.
G for a lossless
antenna
(with
100%
efficiency)
is given
by
2wUmax
G=
for a cavity-backed
denotes
noted
the
that
structure,
total
this
radiated
definition
where
power
Prad
(2.52)
T
bm_x
is the
from
of G is identical
the
maximum
antenna
to that
in the
of directivitv
power
upper
density
half
for lossless
and
space.
frad
(It
antennas.)
is
tl
_2.',:{
the
antenna
gain
becolne_
Zo(ere + ao)
G =
(2.7,.1
2P,_a
Thus.
the
previous
input
has
computation
of G is rather
subsection.
at the
In reality.
input
resistance
However,
where
this
on
an accurate
far
field
the
consistency
far
full
and
the
is directly
To avoid
this accuracy
by integrating
model
field
current
near
inconsistency,
such as circular
space.
about
offer
different
It is obvious
the
the
radiated.
the
in lilt'
that
all
In this case.
oxle
feed,
and
Ri,, is the
plane.
gain
(more
precisely
accuracy.
G, a far
gain
Specifically.
relies
known
This
zone
without
the
di_'ec'-
computation
It is well
field computation
pattern.
assumption
resistance
gain
calculate
as given
that
the
accuracy
in-
pattern
whose
averaging
effect.
or the directivity
by
intensity
=
half
input
an
on
the
prediction.
one may
f / [
ddJ2
the
since
field
radiation
and
reference
the
to determine
by the
P_a
over
work
while
computations
a is found
on
source
at the
for near
once
element(s)
always
behavior,
governed
from
the
not
if Rin is used
accuracy
known
done
be evaluated
measured
may
field
wave
near
arises
P_d
antenna
scheme
may
to antenna
I is the
of the
tivit!l ) reflects
evaluating
P_a
feed is transferred
P, ad = I2Ri_,
trivially
vertical
g dft
7r
Zo/ff2,(, e
4-7
that
if a certain
Z axis,
then
symmetry
g(O,o)
reduces
of the
to [:(0)
pattern
and
remains
the above
12
integration
can
be effecti,.elv
evaluated.
Olherwi_,e.
a nutll'ric.]
2 l) itlt_'_l'a_i_,IL
i-
required.
Axial
circular
AR also
with
ratio
(.4f{)
polarization
features
co.
('P)
the
a minimum
o0 and
ix another
of antenna's
far zone
effect
comt)utationa]
one is again
ilnportant
able
antenna
performance
of the
load.
to determine
.4R-
is of theprimarvcon-erll.
antenna,
It is noted
.4R
ull,n
it is desirable
that
given
uniquely
Silw,
to determine
the above
AI{
pre-cahulat'd
by
lio._
.
short
(2..5t;)
with
4 +
9
2
_0.00.
2
COS ,3]
(2.57)
_ho_t=
where/3
field
(2.50).
readily
= 2(_Ho
--
Since
It
these
represented
the
_H,),
components.
,7_+ qo --
can
two
be
twice
of the phase
obtained
quantities
Im {.} being
from
are
difference
the
complex
quantities
numbers,
between
the
Po and
the
phase
two magnetic
P defined
in
difference
is
as
fl=-jim
with
% + -o'o%cosd]_
the
imaginary
part.
{In p_}
(2.58)
CHAPTER
Edge-Based
3.1
years
methods
in antenna
junction
with
However,
have
analysis
various
Green's
their
versatility.
(not
configuration
Furthermore,
in the
context
ing simplified
models
that
to the
measures
singularity
must
was
Technique
be taken
recently
of canonical
Based
on the
(IE)
with
specific
leads
differ
field
[26] for
are
scattering
success
when
the
to inaccuracies
antenna
distribution
method
a model
which
applied
rectangular
methods
43
are
actual
alleviates
patch
for antenna
the
[1 3].
at)pro-
this
limits
assumption
deviates
bandwidth
excitations
the
the
in coilrole
and
on
from
antennas.
represented
us-
configurations.
feed
In contrast,
was
a major
formulated
near
modeling.
method
be employed
the
for many
in which
for larger
or less from
must
usually
technique
moment
played
substrate,
proper
(FE-BI)
the
researchers
representations
geometry
more
current
them,
and
formulations
of IE methods,
Integral
past
equation
and
engineers
Among
inhomogeneous)
demonstrated
shape
the
IE techniques
of the
Element-Boundary
design.
for the
Moreover,
layered
practical
and
serving
are associated
function
of an infinite
been
integral
IE methods
priate
due
FE-BI
Introduction
Numerical
the
III
these
junction(s),
the
to inhomogeneous
anah'sis,
special
hybrid
difficulties
antennas
Also,
Finite
and
this
objects
[9].
it is desirable
to extend
hybrid
tinite
(']('nwrl_-boundarv
ization
of arbitrarily
configuration
plane enclosing
excited
practical
FEM,
rally
cavity
is firsl
discretized
into
to triangles
based
configurations,
the
Hybrid
readily
the
different
extend
complete
configuration
probe,
aperture.
of modeling
substrate
elements
a number
on the aperture
may t.,
line.
patch
arbitrarily
frill gen_'rator,
In the
contexl
elemenls
non-rectangular
the
t_f ..u,l_ a
ta,u_,_l
a magnelic
For
inhomogeneities,
_'a_lll,h'
in a n_'tallic
of tetrahedral
and
to any
.\ll
i_. rv(ess_,d
or a CP\V
non-uniform
applies
[29 i.
exact
As a result,
shaped
cavity-backed
anisotropies,
of the
t]lal
llalu-
patches
this
integral
for-
boundary
shape.
as well
the
hybrid
antenna
as various
schemes.
System
section,
variational
mesh
is capable
excitation
In this
this
slot
cavitv's
is. in general,
upon
technique
on the
a (axitv
as a simple
line.
gridding
practical
such
microstrip
triangular
3.2
schemes,
,_ll ,'_1,_',.ira-,.,]
azl_'_l_a_.
The antenlla
cable,
reduce
FE-BI
coaxial
the
mulation
cavilv-lmcke(t
by different
i_lTegra] f_rlllula_it,l,
shaped
is shown
the
principle,
the
edge-based
where
formulae
functional
shown
Functional
hybrid
matrix
to the
pertinent
in fig. 3.1 may
FE-BI
algebra
general
to the
notation
anisotropic
scattering
be written
method
as
and
will be formulated
is employed
case.
radiation
As
using
so that
one can
presented
in [9].
by a cavity-backed
[5
z
&
Patch
Ground
plane
Aperture
D, Y
Cavity
x
Coax
Figure
F(E)
3.1:
Illustration
of a typical
the
J,
and
cavity
Mi
the
c_ and
function
3.2.1
with
FEM
In proceeding
r and
.E
electric
field,
aperture
tt_ denote,
wave
interior
incident
cavity
scattering
t)roblem.
dv
H i) _ dS
represent
V; H i is the
S encompasses
elements;
and
(VE)-I(vE)-ko2_E
+ 2jkoZo//(E
where
radiation
cable
r' denoting
if any,
i the
the
the
the
unit
magnetic
from
excluding
respectively,
number,
and
the
dyad,
observation
exterior
portion
relative
and
and
current
sources
region;
occupied
the
by the
permittivity
and
G0(r,
free space
r') the
integration
within
surface
antenna
permeability;
Green's
points.
Subsystem
with
the
discretization
of (3.1),
F = Fv + Fs
it is convenient
to decompose
it as
(3.2)
I('
wh(.re
surface
['+.. denote+,
integral
elements
1_
th(,
vc+,ltttne
integral
contributions.
(e=l.2
....
N).
and
c,.,nt
T].'
cavity
witliin
each
rilmti,.,tl-
vo]utllt+
atld
:.itttilarl\
l.
i:.....ul,(li\i(h.d
t('tral,,<hut_
a;
ac<.t,ttttt>
iiitu
.,,]iowti
itl
tig.
fur
tltc
tt'tr,tli<.llal
:1.2
ill,'
tit,I,t
1
I
nodes/vertices
Figure
is expanded
3.2:
A tetrahedron
using edge-based
and
its
elements
local
node/edge
numbering
scheme
as
E = [v]T{E}+
(3.3)
with
IV]+
Vu2
11 =
E2
{E}+
x,y,z
(:t.4)
tT
in whi(tl
1 _, i._ the
along
the
dron
edge.
ith
edge.
(ttere
Inserting
u I_ = .r. 9 or z_ ('omi_t_lwnt
The
unknown
we use square
(:/.3)into
(3.1).
and
vector
brackets
taking
{1:'}.
of lhe vo[unw
ha,
six enlrie_.<,ll_'
for matrices
the
first
\eclol t,a_i-
alld
variation
curl\
fux_c_i_._-
f_,zcach
t,racket-
of _\. wittl
l_'tla}_.
for x,'dor-I.
rcstwc_
Io {t.}
yields
where
I
Jiz
{h'}_
J,_
+ V x
d,,
!lli_
(3.V)
M,:
o v;}
[DVl y =
{v;}
o{_i}_
To carry
out
the
above
or shape
functions
V_.
shape
functions
matrix
entries
in Appendix
integrations,
For our
for tetrahedral
associated
with
A for reference.
(3.s)
- o{v,;}
o{
oy v;}
it remains
to introduce
implementation
we employed
elements
given
in [30,
a typical
tetrahedron
the
the
volume
linear
expansion
edge-based
finite
in fig. 3.2)
element
are
given
3.2.2
Boundary
To discretize
gular
elements
triang]e,
the
Integral
Subsystem
the surface
since
field
these
integrals
iu (3.1).
correspond
is represented
to th('
flu' al)vr_ur(,
i)_t_, lriai_
\\'illlill
ca<l)
as
E = [5"]_r{ E_},
1:3.9)
where
SU
u = x.y
(3.10)
Su3
Es3
in which
S,,
the
ith edge.
first
variation
is the
u(u
= x, Y) component
On substituting
of Fs with
(3.9)
respect
into
e
of the
the
to {E,},
surface
surface
vector
integrals
basis
of (3.1)
functions
and
along
taking
the
we obtain
where
Go(r,
r') dS dS'
(3.]2)
and
(3.13)
.Nole
that
vector
r.
in (3.12)
w}lerea._
A suitable
the
the
eh,ment_
of the
e]ementsof
set of linear
'_,.!z
edge-based
S,(r)
array
ar_" witll
surface
(referring
fui_ction_,
r(,_t_,<t
to
basi._ functions
2.4,
/, 5 >: (r
r,)_(r)
to fig. :3.3)./,
denotes
the
iz_t,tratit_i_
l_oill_
r'.
is
/:l.t.l
In this expression
i?"._ arc
i th
r_.,
ot herwise
the
length
and
r, i.-
edge
0
Figure
the position
triangles,
e(r)
vector
one
is given
3.3:
Pair
of the vertex
is defined
as the
of triangles
opposite
plus
to the ith
and
the
minus
triangle
yields
the
of boundary
equation
ith
edge
edge.
Since
as the
minus
each
edge
triangle.
shares
two
Therefore.
(3.15)
1
-1
S,
other
the
by
_(r)
where
sharing
= S + + S/.
The
depending
basis
functions
integral
subsystem
constant
A,
on whether
used
equations.
is given
by Rao,
The
rES
rES[
in (3.14)
the
area
of the
r E S + or r E S_-.
We
note
that
moment
method
solution
for the
boundary
integral
explicit
in Appendix
+
denotes
expressions
]3 for reference.
plus
S,(r)
or
-_()
3.2.3
Combined FE-BI
System
{Ix'}
the
and
exterior
consists
coinciding
element(s)
with
surface
pins
fields
[A] and
whose
fully
The
entries
explicit
respect
[A] exhibits
high
(PEC)
walls.
the
are part
of those
in { E} with
sparsity
due
vector
{E,}
(see also
isotropic
FEM
sources
vector
PEC
except
amenna
represents
and
the
corresponding
vectors
[33]).
{El
edges
their
to the
field
element
Finally,
of the assumed
current
to the
cavity.
expressions
as a result
interior
electric
conducting
to the
unknown
with
electrically
the
due
The
coefficients
inside
extracted
In addition,
vectors
respectively.
perfectly
on the aperture.
can be readily
excitation
expansion
or PEC
unknown
are the
excitation,
of all field
those
edges
{L}
(3.16)
in (3.1(;)
It is evident
medium
that
and reciprocity.
formulation
whereas
[B] is
populated.
Two
approaches
subsystems
(BiCG)
when
method
matrix-vector
products
[A] and
execution
of the
carrying
approach
matrix
an iterative
[34]. These
matrices
after
may' be followed
out
is more
and
storing
BiCG
adding
individual
only
is employed
such
in terms
the
differ
in the
algorithm,
efficient
out
approaches
called
[B] by
the
solver
in carrying
up
the
manner
steps.
matrix-vector
of computation
elements.
the
solution
as the
One
corresponding
or instead
non-zero
the
resulting
products.
sum
This
is because,
prior
may
We observe
reordering
of
the coefficient
entries
vectors
after
gradient
could
time
combined
biconjugate
used
matrix
of the
to the
be summed
that
the
first
the combined
in the
conlext
of
-, 1
this ._chenl<
iteration,
where
tlowever,
the
integral
The
tim conlbination
t"FT
the
thus
Bi('G-FFT
3.3
on
and
scattering
etry/mesh
Figure
input
ap[_roach
elin_inating
i.- COml)a_il_h'
to exploit
_o ('xpli(iIly
will be discussed
in chapter
above
presented
a computer
FE-B1
was
by cavity-backed
patch
antennas
is first
as shown
generated
A typical
file which,
xxiTtl the
l_i('(;-Vl"l
tlt_' coilvolutional
a need
program
3.4:
it T,_.rforn_,d
s_ore
,clwllw.
l,r_,i_crt5
tb_' etltire
_f ll_'
Ill matr}x.
4.
hnplementation
the
plemented
Ixvt, mallicc>
is employed
technique
Numerical
Based
second
algorithm
operator,
of the
geometry/mesh
as a minimum,
(a)
the
nodes
and
(b)
the
tetrahedral
(c)
the
nodes
their
developed
the
for the
of arbitrary
hvl)rid
analysis
shape.
supplied
for a cavity-backed
method
The
to this
circular
was
im-
of radiation
antenna
program
patch
and
geomin an
antenna.
contains
(x, y, z) coordinates;
elements
identifying
formulation,
the
and
the
cavity
corresponding
aperture;
nodes
forming
each
element,:
7,2
For arbitrary
mesh
each
antenna
generation
of these
the
routine
is used
usually
refer
of the
FEM
boundary
cavity
integral
be independently
svstems
and
method
but
also
two
and
and
accounts
even
for
complex
the
task.
subsvstems
can
in fig. :l..'_
a few data
the
data
output.
is tile
It is noted
space,
be dominant
non-metallic
hybrid
\Vc
implementation
can
radiating
FEM
shown
less storage
the
It,
an imerpretation
finally
subsystems.
when
t)reproce.,.,ed
generation,
technique
purpose
is a relatively
these
solution
antenna
of the
This
the mesh
but
I5 lficallv
to edge-elenmnts.
from
in size
in a general
that
them
is that
the
be readily
flow chart
requires
x,]ll_ll_'
comnmrciall\.
list of data.
The
separate
true
developed.
however
two
is particularly
over
_l St,ldli>th'al,_]
node-elements
hybrid
small
from
combining
in the
the
subsystem
independent
BiCG
is always
predominates
boundary
to the
in dimension
subsystem
This
from
preprocessing.
complications
demand.
aperture
kernel,
is large
the above
procedures
of combining
integral
is entirely
the
primary
matrix
of memory
the
FE-BI
treatment
sparse
the information
implementation
the
list. Given
as data
t, 'nlploy
a "'universal
input
procedure
major
it is necessary
a number
generates
to convert
this
preprocessors,
efficient
and
aforementioned
the
One
package
packages
extract
describes
geometries,
One
while
tile
in terms
on the
Furthermore.
FE-BI
two
the
portion
elements.
the
thai
implementation
basis
functions
arbitrary
numbering
major
be developed
can
advantage
and
of
validated
individually.
Once
complished
the
integral
both
of the
subsystems
by enforcing
representation
the
are
verified,
boundary
and
explicitly
conditions
the
coupling
implicitly
on tangential
of the
subsystems
on tangential
E fields
over
the
is ac-
H fields
interior
via
and
MESHER //
V
[
[-
Universal File
..........................
Pre-Processors/
I
FOR
FEMSystem
II
BI System
Combinations
FEM System
BI System
Core Program//
Feed Models
w
Combinations
procedures
from tile
FE-BI kernel, to the
.5 t
manner
so lhal
Specifically.
the
to condense
the
the
sparse
.\_. is the
system
finite
number
for the
be carried
systems,
the
the repeated
ation
is avoided.
FEM
and
ent
when
3.4
a special
and
Selected
and
and
radiation
computed
before
alternative
this
below
demonstrating
by different
results
via
The
is performed
will be investigated
some
the
and
is to carry
separately.
treatment
execution
FE
and
was
If the
solulion
the
FFT).
wa._ lo
then
and
the
about
moderately
BiCG
algorithm
reduces
of operations
within
matrix-vector
of the
tl_e
in differem
in a 1-D array
BI subsystems
advantage
denole.,
s_ored
In that
a number
out, the
.\.,
spirals,
of the
Bi('(;
resulting
scheme.
is because
of the
Numerical
We present
dating
The
BI n_atrix
cavity-backed
to use this
This
combinations
BI subsystems
consideration
matrix
volume
array
of h'ngll_
as incorporating
FE
including
requirements.
with
to the
t., I_il_i_l_i:_'d.
array
products.
(such
ca_
compoilents.
cavity
Tl,e
treatments
preferable
FE-BI
the
entries.
is added
surface-
as a si_gh,
within
combined
metallic
matrix-vector
special
it was found
computational
row
lv_luir_q_'_ll-
nonzero)ti('ld
was stored
of the
matrix
the
only
of unknowns
without
3,'_.._,_ + Ny long.
of a single
matrix
evaluation
out
on
involves
of nonzero
N_ A'_ BI integral
sized
which
number
compulalional
conditi(ms
element
total
maximum
ways
boundary
and
associated
the
BiCG
products
scheme
on to the
numerical
at certain
depth
in chapter
numerical
results
for the
becomes
system
iter-
for the
appar-
for efficiency
5.
Results
representative
the robustness
of the tetrahedral
configurations
of cavity-backed
FE-BI
method
are
compared
purpose
formulation
antennas.
with
of vali-
for scattering
In each
reference
case the
measured
or
55
calculated data.
Scattering and radiation by a circular patch:
Vig.
3.6 il[ustrales
substrate
having
2.6 cm and
and
the
plane
and
substrate
patch
frequency
a relative
the
comparison
were
surements
the
is very
and
in this
was
measurements
by
are
of the
illustrated
follow
denote
the
the
has
whose
For this
placed
FE-BI
throughout
calculations
modeling
method.
the
radiator
and
and
spiral
was
plane
radiation
base.
is given
cross
of 9.1.78 cm.
situated
This
section
pattern
in fig. 3.9.
taken
spiral
has
in the
It is seen
measurements
impedance
and
mea-
The
it is again
a+ are
arm
probe
seen
that
this
and
strip
equal
cm
.X = 30 cm
deep.
pattern
The
using
the
(f
is in good
spiral
(p, o, z)
0.02,57k.
(9.24
cross
cm
section
= 1 GHz)
computed
a probe
and
cone
top
mesh
the
where
on an inverted
ver-
surface
forming
to 0.0832_
of 1.68 cm and
_ = 90-plane.
that
geometrical
radiator
of the
resides
calculations
10.01
the
z = a exp(0.2210),
a diameter
cavity
spiral
edges
2.66)],
For our
in a circular
of the incident
in fig. a.7.
center
of the
coordinates,
bottom
a diameter
cylindrical
of
Input
to demonstrate
bottom
lines p = 0.0503)_exp[0.221(gS+
standard
as a funclioJ_
agreement.
Two projections
top
band.
<avilx
:\
between
are displayed
patch's
]l_is
i_-
ils ]{('S.
tile direction
GHz
diallwtcr
cn_ wi&,.
aae_ ]{('S
agreemenl
4-9
patcl_:
for m_'asuring
computation
patch
l[w
lllick
spiral:
of this
The
body
the
6.292
backscatler
are in good
conical
>urfacc
cavity
section
the
same
tlw
of _r = '2.9.
as seen
0.8 cm from
a one-arm
the
and
for the
in fig. 3.8.
respectively,
tall)
in fig. a.6.
good
and
We considered
satility
calculated
calculations
case
Radiation
and
normal,
on
in a circular
in a low cross
measured
60 from
calculations
residing
constallt
is enclosed
is also shown
was
l,atch
dielectric
recessed
of the
wave
feed
a circular
and
the
E, principal
feed at the
agreemen_
cavity
with
the
U.UtJO
f
i .............
-tO.O
u,
FE- BI Me_m:l
-20.0
ee
t_
-30.0
.r-
-40.0
-50.0
Ii
.........
-60.0
2_
314bern
I .......
4.00
I .......
5.00
3.6:
Comparison
of the
computed
function
of frequency
for the
was 30 off the ground
plane.
600
frequency
Figure
7.00
80(2,
(GHz)
and
measured
shown
circular
eree backscatter
patch.
The
R('S
incidence
as a
anglo
1.0
0.5 _2.0
0.5 _/2.0
1.0
Figure
3.7:
Comparison
circular
center
patch
of the
of the
shown
patch
computed
in fig.
and
the
and
3.6.
measured
The
frequency
feed
was
input
was
placed
swept
from
impedance
0.S cm
for
the
from
the
3 to 3.8 GHz.
.)i
dala
gixen
the
nwasured
cavity
not
in [35].
data
housing
part
a large
circular
Annular
slot
cular
3.10
24.7
cm
and
[14].
agree
For these
constant
of the
circular
a qualitative
stacked
of the
and
stacked
annular
fi'_,lll
ti_,il' circular
]]m
spiral
Because
the
slot situated
annular
slm
in the
algorithm.
small
BiCG
aperture
Input
oi,
calculated
calculations,
the
to account
via
frequency
filling
method
is
without
a need
for
as a function
of
the
swept
slot,
are
shown
from
in
integral
700-1000
MHz.
presence
the
as a result
a modal-boundary
was
the cavity
for the
for holding
across
_hc
FE-BI
calculations
placed
in a ci>
is narrow.
and
The
configurations
impedance
by a probe
values
model
wide)
the
value
circular
bandwidth
two
consisled
is verb' small
of the material
the
with
FFT
with
patch
study
(0.75
3 cm deep.
excited
measurement
To demonstrate
to the
nmdel.
cm
in modeling
well
this is an effective
nificant
in the analytical
circular
the
radiator,
and
configuration
[ltJIIi lilt
which
considerations.
dielectric
backed,
of interfereI_cc
configuration
BI subsystem
effective
for this
Stacked
wa.- included
and
to invoke
The
as part
wide
computational
fig. 3.10,
method
a narrow
of the
quite
frequency
which
because
impedance:
is no need
special
horizon
plate.
implementation
basically
the
measurement
shows
cavity
there
near
the spiral
of the
Fig.
of a dielectric
slot cover
used
plate.
antenna:
capability
of the
visualization
patch
rectangular
increase.
patches.
of the
antenna
patch
This
The
developed
near
as shown
shape
has
is because
circular
hybrid
field
distribution
in fig. 3.11.
been
of the
patches
technique,
are
Note
investigated
dual
more
frequency
attractive
we now present
inside
that
and
a cavitythe
similar
found
a sig-
resonance
than
due
slacked
5",
for the
computation
," ........
I .........
the one-arm
conical
spiral
of fig. 3.9.
1.........
I .........
I .........
I .........
I .........
I .........
O.
._
-10.
E
_
-20.
-30.
_e
-40.
-50.
Figure
3.9:
Comparison
90-plane,
in fig. 3.8
of the
with data
-60.
_0.
calculated
in reference
O.
radiation
_0.
60.
pattern
90,
(E_),
taken
conical
in the
spiral
o =
shown
a = 12.35 cm
"i
b = 0.75 cm
Po"
7.7 cm
d=3cm
i.0
0.5
1.0
Figure
3.10:
Comparison
backed
slot.
of input
impedance
calculations
for the
illustrated
cavity-
rectangular
quencv
patche_
[3?].
lnfortunatelv,
in the literature
It turns
this
fed with
base
the
an offset
lower
PDE
patches
electric
from
Figure
3.11:
from
(The
and
One
respect
excitation
Visualization
circular
the
the
laboratory
near
patch
the
ensures
near
antenna.
patch.
antenna.
the
has
upper
[38].)
linear
may
feed,
and
The
the
interesting
the
energy
the
other
the
polarization
can
complete
two regions
Although
field distribution
This
the
be clearly
only
via the
image
of the
is that
is concentrated
is near
at
the opposite
act as out-of-phase
patches
lower
and
on
point
are
in radiated
at the
cavity
uplmr
completely
provide
Another
the
for
alllenna
to link the
to rely
and
_,uiled
patch
between
patch.
probe,
is well
patch
contact
is available
offset
the probe
of the
I,,,,11 i_,l,_,II_,_l
circular
lower
measurement
a single
excite
of the
to the
a microstrip
center
the
transfer
which
technique
stacked
electric
power
lower
is around
to the
to effectively
offset
thus
FE._I
underneath
No direct
regions.
with
the
plane).
hybrid
a cavity-backed,
post
lla-
tool.
field visualization,
above
antenna
pole
we chose
and
techniques.
the
presented
vertical
exists
the near
related
shape,
this.
coupling
two distinct
location
the above
single
field distribution
though
no sufticient
as a ground
electromagnetic
verified
that
To show
(viewed
out
study.
laver
circular
fields.
of a stacked
in
CHAPTER
Efficient
4.1
Boundary
IV
Integral
Subsystem
Introduction
As is known,
the
capable
of handling
sociated
with
this
it less attractive.
tenna
systems
much
hybrid
technique
This
with
useful
approach.
element-boundary
of conformal
and
any
is especially
the
small
global
if one
FE-BI
aperture
isotropy/anisotropy,
the
One
solution
However.
truncation
method
and
is the
is accurate
and
drawback
as-
the
approaches
possibly"
layers
reduce
CG-FFT
CPU
cavity
etc.),
an-
for the
design
it would
requirement
technique
make
large
suited
complex
of metals,
the
can
in modeling
is particularly
size and
speed
method
is interested
other
to accelerate
possible
integral
antennas.
other
true
Although
relatively
feedlines,
more
finite
a variety
(arrays).
configurations
(including
hybrid
be
for" the
discussed
in this
chapter.
The
whose
this
boundary
integral
size is determined
analysis
ments,
and
By resorting
a discrete
becomes
convolutional
equation
subsystem
leads
to a full)'
by the number
of aperture
mesh
edges.
impractical
to overcome
to the
(BI)
in terms
this inefficiency,
structured
form,
mesh,
thus
of storage
a uniform
the
boundary
permitting
61
the
and
zoning
populated
For large
computation
computation
matrix
apertures.
time
of the aperture
integral
matrix
can
of the
require-
is required.
be cast
matrix-vector
into
('2'
l)roducls
via
full BI matrix.
involving
discrete
the
Ttlis
menlorv
rectangular
for triangular
surface
we first
that
show
triangular
ferred
to as the
are
4.2
Application
been
presented
Conjugate
cially
problems
numerical
4.2.1
BiCG
systems,
the
CG
symmetric
may
the
without
With
(CG)
of the
primary
and
to costly
oI1 uniform
triangular
m('sh('s
treatn)enl
in section
4.3.
rt,A few
validity.
Algorithms
of linear
systems
references
not
lend
of equations
are
collected
themselves
of handling
for direct
system
one seeks
the
computing
has
in [41].
as a robusl
large-scale
solvers.
solutions
com-
It is espeof large-scale
resources.
Preconditioning
have
applications
advantage
Also.
discussed
capable
been
that
of different
preconditioning
developed
nowadays
It is noteworthy
taking
sparsity.
do
when
algorithms
as aforementioned.
algorithms,
algorithms
resorting
a sl)ecia]
representative
be impossible
implenmnted
geometries,
solution
are indeed
In thi_. ('llal)It'r.
and
method's
algorithms
wa_. alst)v('t)ollt'(t
rectangular
and
I}_(,
t,, IIi-_,l_i,,_.-
approxinlations.
Gradient
the
CG
which
gradient
one
the
iterative
and
Algorithm
Conjugate
is proposed
[42], they
to employ
systems
(CG)
state-of-the-art
desirable
antenna
demonstrate
I,('('1_ al)l)li('d
be preci..;ely
the
of Conjugate
"black-box"
putational
between
scheme
which
can
iTllt,h'llmIIlatioll
inherent
solver
differences
investigated
the
input/output
BiCG-FFT
Gradient
extensively
Although
[40] involving
) ;lt_tl a_t,i(]ill_
ha> alr('a,l_
a similar
For non-rectangular
results
The
grids
overlaying
( l)l"-I
sch('nl('
sax'ilkc
[:)._).3-11. and
The
described.
t ral_sfornl
grids
the
meshes.
are also
Fourier
is to solve
there
properties
is often
for over
used
exist
large
various
of the
to speed
forty
years
scale
linear
versions
matrix
such
up convergence.
of
as
.-ks su,,oe_ted.,._
.
ixl [45.4(; i. the
algorithm
used
i> a- f_,ltoxvs:
Given
Pl =rl
For
/,'=
= b-A.xl
1.'2.3 ....
_.
Ok
r k
_-Z_- A p j,.
rk+l
= rk -
ol,.A p_..
G+I
= rk -
flk
--
_k+l
" Fk+l
--m
r k rk
xk+l
where
* denotes
iterative
the
algorithm
is usually
referred
systems.
If the
r_, the
per
algorithm
iteration,
complex
conjugate
is quite
general
to as the
matrix
since
and
in terms
Biconjugate
can then
= Xk + akpk
is symmetric
be shortened
in each
step
_: and
T is the
of the
gradient
and
system
(BiCG)
the
to require
Pk
transpose.
are
initial
only
complex
This
matrix
of the
to be solved
method
and
for unsymmetric
value
one
version
is chosen
as rl
matrix-vector
conjugate
product
of rk and
Pk,
respectively.
The
the
ordinary
BiCG
be shortened
amenable
when
conjugate
gradient
A is Hermitian
to have
about
to a straightforward
(i.e.
50_
algorithm
can
A = A'T).
Again
less computational
interpretation
be considered
as a special
in this case,
effort.
The
of its convergence
the algorithm
CO algorithm
principle.
case
of
can
is also
Basically.
(,1
the algorithm nlinimizes the funclion
1
f(x)=
Hence
x obtained
solution
from
the
-x.A.x-b-x
.)
('G
algorithm
The
CG
type
introduced
convergence
matrix.
of algorithms
to improve
rate.
4.2.2
One
with
In our
to alleviate
memory
and
integral
sparse.
The
CPU
equation
of moment
sit,l)>
Iw'ozlms
lhc
may
where
is the
counts
will of course
plished
by' recasting
fast
BiCG
solver.
Fourier
condition
is the
since
manner
boundary
the
dominate
leads
significantly
of the
quite
CPU
requires
reduce
original
the
diagonal
successful
and
in size.
partially
O(N
(FFT)
onto
type
matrix
demand.
This
2) order
solution
out
the
can
requires
due
is used
Reduction
time
and
submatrices
matrix-vector
be used
boundary
the
Solving
of operations
full
excessive
to the
is because
by its nature.
a few Toeplitz
to carry
of algorithms
dense
dimension.
the
large
LU decomposition
system
system
decrease
is usually
(CG)
the
to a dense
BI system
transform
inverse
preconditioning
System
partially
the
integral
significantly
Linear
gradient
However,
always
until
and
has been
of equations
conjugate
still
the
For
requirement
in a traditional
svstem
competilive
algorithm.
system
time.
(MoM)
of the
BiCG
linear
memory
become
this preconditioning
the
the
not
preconditioner
Algorithm
work,
partially
did
simple
the
BiCG-FFT
system
/, iteralioli
A-x-b=O
the original
In our applications
in conjunction
and
the
of the equation
...Xf(x)=
was
after
method
the
dense
per
iteration,
of the
operation
this
can be accomand
making
products
use
in the
.-ksdescribedin the
a convoluTional
prising
since
using
the
To this
end.
data
In contrast
order
algorithm
is employed.
nique
the
fast
has
lowest
multipole
4.2.8
the
and
the
next
relation
section.
integral
We recognize
as shown
3).
include
These
the
uniformly
spaced.
numbered
in accordance
carry
the
numbering
along
the
?,, direction.
that
thus
FFT
with
p-directed
(rn, n) if the
The
indices
of
_he product
of
yields
the
operation
result.
producl
counts
significant
if the
and
solvers,
in
FFT
the
tech-
including
the
preferred.
the
structured
structured
mesh
triangular
is described
grid.
in the
consists
of equal
right
triangles
different
classes
of edges
(class
1.2
the
diagonal
edges,
each
class
and
geometric
edge
lral>fonn
grid
implementation
their
Fourier
prodtlc!
equation
using
and
three
ilcralhm.
eat])
lntegral
triangular
involves
z-directed,
For the
integral
iIl_t,
t_,I >ur-
for a malrix-\ector
2 ,\')
is discretized
the
of the
is alwavs
unstructured
a!
t_ecomes
is indeed
of Boundary
the
tile
convolution
reduction
among
equation
between
the
lt,i-i.
convolut}ol_
by taking
O(.log
[47], thus
Form
the
transform
needs
demand
(FMM)
out
},_' Oh-1
illl_'gl'al
requirement
operation
CPU
method
boundary
scheme
The
Convolutional
The
An inverse
of O(.T 2) CPU
fashion,
ill the
in which
Call
for discreli:_a_i_ll.
convolution
(arrays)
_'(_llaliOl|
ilJl_'_la]
is applied
Io carry
the
sequences.
to the
a traditional
calculale
sequences
I_oul,dar\
i> involved
i! remains
may
two "frequency"
lhc
_rid
function
algorithm,
one
seclk)n,
if a uniform
the Green's
('G
two spatial
the
form
next
location.
is the
rnth
(m, n) take
the
along
values
rn
0,1,2
....
,M i
0,1.2
.....
:yi
of edges
Specifically,
the
all of which
and
are
is independently
the
a" direction
ith class
will
and
nth
the
{,t,
///////
/////
/
////////
///////
///////
;a
m: 0
Figure
with
4.1"
i = 1 for the
x-directed
edges.
and
Consequently,
denote
Structured
y-directed
M'=
where
mesh
edges,
.........
consists
of equal
i = 2 for the
right
diagonal
triangles
edges
and
i = 3 for the
we find that
M-1
i=2
M-1
i=3
the
,_"'_=
numbers
of elements
N-1
i=2
N-2
i=3
along
the
(4.1)
and
y directions.
respectively.
To
perform
elements,
new
indices
the
it is now
(re, n).
integrations
for the
convenient
to rewrite
We readily
find
that
evaluation
of the
the
basis
the
edge-based
boundary
functions
basis
(3.14)
integral
in terms
functions
matrix
of the
associated
t;7
with each of t[w aforenwnliolwd cla,>of vde,'>vaill,_.
r,,wril_t'i_
a-
t.r
q l ._
_,'"
ill
"-
%',,,(._.:/_-
.-._J"
I-t.21
oi ]wrwise
",''-
(y-(,_
3',-(1.:_)
-,,/(Z._'): + (Ay) 2
_'v_,,(x.y) =
_Xa'Ay
+ 1)_xy)._. + (,,,A.,.-.,.)k
((n + 2).Xy
ot herwise
- y).i" + (a'-
(,,, + 1)_X.r)_)
(.r.y)
E +
(.r.y)
S(
I
S_,(x,y)
Ay
(y-
nay)/+
(m_Xx-
a')_
where
the
superscripts
processes,
of the
one
refer
obtains
boundary
otherwise
to the
edge
a discretized
matrix-vector
class
version
product
After
of the
can
now
{Blsubsystem}-[Bl{E,}=_-'_
(m, n) are
whereas
(m',n')
ements.
Thus,
entry
ki =naI
matrix-vector
the
are
the
the
same
of the
indices
the
class
of the
column
resulting
It is readily
found
S_
(
u)8
,AxAo,
indices
iL
and
which
asseml)lv
each
entrv
as
EJrn',n'
(4.,5)
n_=O
for the
edges
ith
the
class
belonging
i, rn and
after
observation
to integration
n completely
execution
edgos
of the
defines
eltile
boundary
that
S_,.,
from
Bmn,m'm'
o
ml=O
jth
discretization
be calculated
location
for
specification
_+ M
product.
geometric
the
BI system,
_
j=l
in which
(-1.1)
iis
Go(r,
ei(r)ej(r)liljGo(r,r')dxdydx'dtj
4.6)
with
=i
_
")
l.d
= :_
.More importantly,
lutiona]
it can
property
be shown
B_,_.,_, n, :
that
BI___,.,__,_,
tile BI subsvstenl
(-l.(i) exhibit_
) and
rewrhe
thus
we can
tilt'
(-t.5)
c<)lJxo-
a.,,
[B]{E,} = Z B',,
(4.sl
./=t
where
the
section
that
* denotes
to ensure
the
BI
the
smooth
computation
employing
the
convolution.
the
matrix
property
discussion
of the
2-D
boundary
discrete
other
Fourier
than
of Bi__m,,__,v)
The
those
proof
will
of the
remaining
matrix-vector
transform
entries
is also
be presented
which
invoked,
procedure.
product
(DFT),
thus
are
in the
can
be performed
a need
When
the
that
the
total
symmetry
implying
h'; = _
should
be compared
be required
thy
that
B(m__,,__,v
)i;
and
zero
Specifically,
if the
._
is much
_- B_;(ff_, _)
padding
the
may
was
smaller
in the
Ni(M'
+ M j-
not
cast
are
1)
(4.10)
To
in a 2-D
also
be required
matrix-vector
product
whose
storage
in convolutional
in number.
be cast
BI matrix
j=l
to the
BI system
entries
i=1
This
I .9/
non-redundant
3
by
to store
,,_.,,=
it is concluded
of this
It is now seen
avoiding
unique.
end
array
to make
(4.5)
avoid
aliasing,
which
use
form
has
is executed
by
and
normally
it is notewor-
it is necessary
the
of the
would
usual
standard
using
periodic
FFT
the
that
form.
routines.
MFTNFT
(;D
arra\
B'_(-
fi_.-Tt).
B'J(-F_.
-_t.
B'J(ff_._
B'a(fft
0
with
the
MFT0<
M'+
fi < .\J
0 < ff_ < M'
1 - NFT).
- 1 - MFT,
(-1.11)
NFT-
fi - 1 - NFT).
MFT-
NFT-
otherwise
corresponding
field vector
given
by
t
0<7_
o<
,= J
and
MFT
and
In the
NFT
must
BiCG-FFT
< :y2
< M,,
(4.12)
otherwise
be powers
algorithm
the
of 2 if a radix
2 FFT
algorithm
is used.
BI subsystem
vector
is symbolically
computed
as
3
{BI subsystem}
= E
{DFT-'
{DFT{B_J}
(4.1,3)
DFT{Esa}}}
i=1
The
presence
which
results
indexing
the
(4.6).
first
the
integrals
is necessary
conclude
after
operator
for addition
to [40] the
As
of the
inverse
to the
(4.6)
to preserve
promised,
this
integral
In addition,
FFT
FEM
the
(4.8),
with
The
the
same
of the
or the
proof
no loss of generality
a single
relation
we refer
can
2-D
arrav
with
the
be remarked
that
in contrast
proper
any approximation.
global
combined
B_,_,_,,v
to the
proof,
the
This
system.
BI__,_,.___,
be appied
for the
of the
column
introducing
feature
proof,
reordering
It should
without
symmetry,
show
necessary
into
subsystem.
To simplify
in (4.6).
the
operation
are evaluated
we now
section.
06 indicates
consider
second
i =
t to
only
integral
1 class
in
edges
7(J
/
(m,n)
%,
- /
%.
(m',n')
Figure4.2:
Illustration
prove
(y-directed)
for the
function
the
?/"
=
trial
first
function
this
C J JsH+ J JsH'_
-[(n/ky
[n-_y,
Note
Therefore,
(4.14)
boundary
and
the
in (4.6)
indices
Io help
Io
integral.
i = '2 class
edges
point , substitutng
(diagonal)
in 5"7
1
- + and
_c, m,_(x,
y) in 5'_
yields
-y)(n'/ky-
coefficient
that
(n + 1)A9]
corresponding
?/"
C is a constant
proof.
the
of the
To this
term
Go [(x -
where
with
property
will be used.
y) in 5;[ into
Intmn,m'n'
triangles
convolutional
testing
S_,n,(x,
of two
the
the
and
unprimed
will be simplified
x = max
+ (x-
m _2._x)(rn2Xx
-Jr-2.._x-
integration
for the
y')
its
detailed
form
limits
should
be set
coordinates
if the
and
following
x
of our
as [m_x,
transforms
= m-Xx
(-1.14)
is not
similarly
,r')]
concern
for
(m + 1).sxJ
and
for the
are
primed
introduced,
ones.
viz.
(4.].5)
y = nay
Indeed,
on substituting
Int_,_.,,,,_,
= C
the
transforms,
[AxAy
d
dO
[[&xAy
d
y' = nAy
+ 7?
one
+ 7?'
obtains
Jr/r/' + _(2Zx
- {')] Go {.i'[(m
rn')Za'+
dO
(_-
_')] + 0[(n -
n')_y
+ (r/-
r/')]}
d&l,ld,_'dr/'
(4.16)
71
is howeveras._uredlhal
7/)] no matter
theresultiIl_ext>ressio_lnmsl
what
function
Intm,.m,,_'
This
is the
in (4.8)
desired
relation.
Mesh
As
Overlay
described
so that
the
achieved
during
mesh
which
material
4.3(b)
shows
Field
we can
conclu<h'lhal
the
in block
circulant
form.
whenever
the patches
at. some
latter
distance
case,
edge/corner
especially
example
lhc
<on\olulio_l
when
of this
might
of the
small)
need
cavity
the
This
aperture
below
the
size
be always
in shape
to add
near
gridding
can
are rectangular
(usually
one
aperture
the aperture
an
appropriale
aperture
is not
or in
to avoid
large
enough.
implementation.
Transformations
natural
mesh
unstructured
free
surface
BiCG-FFT
algorithm,
another
subsystem
triangular,
to construct
aperture
grid
be put
the
effects,
the
integral
proof,
uniform
In the
for circular,
it is not
aperture
the
requires
can
around
edge/corner
However,
ory
i-l.
IT_
solver
are placed
absorbing
ically,
i_ [_,'a1_-
= lnt,,__m,.,__,_,
BiCG-FFT
generation
in fig. 4.3(a).
possible
the
BI subsystem
as shown
ture,
Scheme
above,
case of radiators
4.3.1
t,-_,'l.lt_-
holds.
4.3
Fig.
From
I,eatuiLctit_ltt,f
grid.
non-rectangular
a uniform
is necessary
In this
case,
using
to the
to make
is proposed
structured
constructed
mesh
to conform
an approach
coincident
is then
or other
grid
by using
use
patches
the
patch
of the
to overlay
as shown
the overlaid
mesh
on the
generator.
shape,
leading
efficient,
on the
in fig. 4.4.
uniform
aperTypto an
low mem-
unstructured
The
grid
boundary
whose
edge
uniformgrid
recessedpatch
cavity
(+,)
Circular
.30.
Patch
........
Recessed
I .........
10. L
1.........
Underneath
I .........
I .........
the
I.........
I .........
Ground
I .........
Plane
I.........
""_
..
er_
O.
20.
-10.
,m
-20._J
e,,,
permittivity
= 4
permittivity
= I
z_
permittivity
= 4 (with FFT)
,,
permittivity
= 1 (with FF'q')
"
-30.
rj
........
,+,..+
.g
:I
-40.
;.
c_
-50.
-60.
'....... t.........
0.
10.
,......... i.........
20.
30.
,.........
40.
i......... l.........
50.
60.
J.........
70.
l........ i
80.
90.
(b)
Figure
4.3:
Printed
circular
incorporate
tion;
FE-BI
the
patch
BiCG-FFT
(b) Comparisons
technique
antenna
is modeled
algorithm.
(a)
of the BiCG-FFT
presented
in chapter
using
recessed
Illustration
result
3.
the
with
of the
that
of the
scheme
configuraordinary
to
7:1
2>
m
Figure4.4:
Overlay
can
shown
be related
matrices.
That
of a structured
mesh,
fields
here
to those
triangular
to conform
on the
is, it is necessary
aperture
to a circular
unstructured
to append
grid
to the
mesh
over
an
unstructured
patch.
via
system
two sparse
(3,11)
transformation
the
relations
(4.18)
where
the
subscripts
non-uniform
and
aperture
backward
numbers
these
of the
Given
grids,
respectively.
Also,
matrices,
respectively,
uniform
the
points
three
nu
transformation
To derive
at three
u and
elements
on each
points
the
and
fields
at the
refer
to the
non-uniform
belonging
center
at these
and
points,
with
to the
ends
of the
coefficients
[TF] and
mesh
of [TF], we begin
edge
field
with
[TB] refer
N,, and
edges
on the
the
expansion
uniform
edge
we can interpolate
of the
grid.
cavity
uniform
to the
N,_
and
forward
denoting
the
aperture.
(3.14)
and
enforce
We conveniently
it
place
along
the
,l
e u
overlaid
_"
|'.
_ _ _-,
r...2
mesh
triangles
a__-
(m,n)
0 _
"
_ __-I.
....
unstructured
uniform
grid edge
Figure4.5:
Illustration
transformation
of the parameters
matrix
elements
and geometry
used in constructing
the
between
the structured
and unstructured
mesh.
of the
uniform
grid
using
the
(E,,)(m,n)
weighted
average
1.
_-e_,.
k
E._(r_.d,)
Ne
k=ln d
2 _'rendl
]Vmid
--
1 _
+ Aria
Enu(rmid)
k=l
-+ 2Aend2
1
/_'re n d _
Enu (rend_)
k
(4.19)
}
k=l
in which
of edge
being
grid
denotes
triangles
a point
denote
the
rends,
considered.
with
specify
the
shared
number
respectively,
After
the
assembling
unit
vector
The
superscript
by more
into
E_,, represent
k being
than
will typically
(4.19)
x, y or the diagonal,
quantities
of non-uniform
and
along
one
grid
a sum
triangle.
triangles
be equal
(4.18)
depending
the
variable
fields
sharing
in the
in case
Obviously,
on the
non-uniform
J_endl, Armid
the
node
class
at rends,
and
:V,,,d2
rmid and
to unity.
we find
that
the
elements
of the
forward
tran_fornlalio1_
n_atrix
are
giveiL ],\
1
I
--_endl
k=l
,:l
\'m,d
]
+-\'rni----_
('JtS_(r'nid)
k=l
_=1
l
+-
Z',,,S;(r_,,d:
,"_'en d
_2"_"end2
k=l
(-t.2(li
(=1
in which
(
j
_idt
and
the global
indices
grid
edge.
The
uniform
grid
edges,
whereas
in the
definition
storage
efficiency
that
because
the
arrays
subscript
the
basis
it avoids
storage
to only provide
element
Following
the
backward
transformation
grid
edge
traverses
each
midpoint
and
non-zero
entries
endpoints
in each
elements
assembly
data
we can
matrix.
used
edge
However,
on an edge
during
matrix
flags the
user-oriented
grid
they
of the
index
used
computation
increase
iteration
uniform
step
that
the
assuming
the
patch
edges
entries
each
number
lying
of the
uniform
entries
a maximum
typical
user
formulation.
non-zero
grid.
in
and
The
grid
of the
for
non-uniform
the
non-
is avoided.
feature
reach
the
non-rectangular
triangles,
can
explicit
each
expression
be noted
edge
in a substantial
BI system
the
local
the
of large
obtain
It should
results
hi numbering
analysis
file which
or less non-uniform
reside
grid
that
process
the
is an important
procedure,
index
We remark
populated
on a PEC
three
global
St.
allows
an additional
ith uniform
( (-- 1, :2 or 3) is the
matrix
scheme
this
to the
functions
of fully
and
otherwise
subscript
the usual
overlay
same
3_ is the
transformation
proposed
because
needs
of the
of the
i and j correspond
uniform
and
I j = j,
in
of 18 if the
The
maximum
is much
less.
.i,
4.4
Results
Figure
angled
triangle.
Since
angled
triangles,
it is used
with
the
is first
BiCG-FFT
over
the
using
space
the
structure
Figure
with
regular
array.
the
array,
about
23 dB/A 2 lower.
To
varify
the
bistatic
RCS
tenna.
In this
The
version
used,
respectively,
with
proposed
FE-BI
FE
of the
agrees
nwsll
with
cross
very
I'i_tll
rigtJl
i]lcorporated
with
FE-BI
radar
i_" _t
[_}tltll
technique
BI code
hybrid
monostatic
o = 0 plane
t'_t(tl
to a uniforlll
developed
and
scattered
to find
RCS
of the
overlaying
the
the
the
as shown
dielectric
the
those
It is observed
FE-BI
methods.
reaches
scattering
scheme
results
computed
that
by the
the
l{i('G-
section
FI:T
descril,ed
lec]lnique
([{('S)
well with
that
for
fillings
paltern
compuled
using
the
using
the
agreement
cavity-backed
array
to scattering.
with
both
the
the
be seen,
aperture
at the
a cavity-backed
the
level
of the
the
the presence
incidence
is
we evaluated
circular
patch
(_ = l0 inside
the cavity
regular
FE-BI
BiCG-FFT
is quite
with
BiCG
the overlaying
satisfactory
of
BiCG
same
geometry,
absence
regular
of e_ = 4 and
obtained
of the
using
nonrectangular
in fig. 4.9 by
same
As can
above
by the
by the
aperture
were obtained
incidence
the
fields
contribution
patterns
the computations
whereas
scattering
case,
array
monostatic
versions
patch
analysis.
patch
formations.
solver.
wllt,l_.,
solver.
Again.
BiCG-FFT
compared
our
to compare
the
of the
are
to verify
original
informative
4.8 shows
scattered
is adaptal)le
BiCG
array,
geometry
without
the patch
the
3. As shown
It is also
and
system
compared
in chapter
this
2 x 2 palch
anwere
method
trans-
in scattering
77
Patch
Figure
For
the
radiation
near-zone
significantly
felds
integral.
couplings
coulings
(see
analysis
e.g.
control
are
the
(e.g.
may
input
be required,
the
accuracy
the
trial-testing
it is suggested
from
the
far
are treated
using
the
normal
the
fast
algorithm.
and
accuracy
once
of the
combined
FE-BI
with
technique
zone
the
t)atch
where
an accurate
of the
overlaying
element's
to separate
elements
using
triangular
impedance)
by considering
computed
[47]),
of tile cavity-backed
Specifically,
closed-region
region
Illustration
enhanced
boundary
the
4.6:
Array
weak
method
the
couplings.
approach
overlaying
scheme,
in an adaptive
prediction
scheme
has
been
it can
manner.
between
strong
whereas
be
in tile
interactions
The
of
can
interactions
of moments,
This
array.
close-
the
weak
reported
be used
to
10.
.........
I .........
I .........
| .........
I .........
I ....
| ....
[ ....
I -
"
"
O.
-I0.
-20.
_.
-30.
4o.
._
-SO.
5tlpm_/'l_ _/_
40.
_I
-7o.
_n
.......
-80.
I .........
O.
Figure
4.7:
Comparisons
of the
triangular
patch
the regular
BiCG-FFT
i .........
I0.
I .........
20.
30.
! .........
monostatic
array
I ........
40.
50.
radar
shown
I .........
60.
cross
in fig. 4.6.
section
The
BiCG FE-BI
technique
described
proposed
in this chapter.
-10.0
."
....
I .....
I .....
....
I .....
....
t .........
70.
reults
I .......
80.
90.
scattering
were computed
in chapter
3 .....
[ .....
by a 2 x 2
3 and
using
using
the
I .....
-20.0
-300
cxl
-40.0
o_
........
-50.0
v
-600
_rJ'ef_c
(H-.pol)
_'
_.Juencc('E-pol)
-700
1 ....
-80 0 .....
0 (.w)1000
Figure
4.8:
Comparisons
aperture
structure
the
regular
BiCG-FFT
l .....
20.00
....
1"
30.00
of the monostatic
with
the
shown
BiCG
same
FE-BI
proposed
size
Again,
technique
in this
"L
chapter.
....
5000
radar
cavity
in fig. 4.6.
' ""
4000
I .....
60.1)0
[ .....
70(X)
cross section
and
the
dielectric
results
described
I'"
801111 t,uIg_
scattering
filling
were
in chapter
by an empty
(_=1)
as the
calculated
using
3 and
using
the
........
! .........
t .........
I .........
I .........
I.........
I .........
I .........
t .........
-8
-_
-16.
e,,.,,
-24.
"_'_'_r'"_'"_'_'_"Cf'_"_'_'_e"Cr--+.._. .....
e,-
"-
-32.
_e ,_ -_-. . .. .
-40.
r,..)
._
-48.
With
No
Transform
(permtttivity=10_
With
No
Transform
(permtttivlty-.._,}
........
-56.
-64.
t_
x,
With Transform(permittivity=l
_r
0)
-72.
......
-80.
O.
Figure
4.9:
I .........
10.
I .........
20.
Bistatic
RCS
scattering
regular
BiCG
FE-BI
transform.
I .........
30.
I .........
40.
I .........
50.
by' a crcular
and
the
I .........
60.
patch
BiCG-FFT
I .........
70.
I ........
80.
antenna
algorithm
90.
modeled
with
using
the
overlaying
CHAPTER
Efficient
5.1
Finite
Element
Subsystem
II
Introduction
As demonstrated
hybrid
finite
acterizing
in the
previous
element-boundary
conformal
integral
antennas
cable
is because
of the
plementation.
sampling
variety
rates
where
due
to the
or inter-distance
over
involve
only
severe
for 3-D
tetrahedrals.
a large
system
presence
(spirals,
slot width
eter
10-20
The
of slots).
higher
slot
of fine
where
to the
spirals,
antennas
this
details.
Among
slots,
dimensions
cases,
the
mesh
is extremely
the
assembled
degraded
8O
geometrical
from
mesh
this
quality.
typical
sampling
details
type
for tile
imhigh
them
are
slots.
(cavity
diam-
dense
(with
discretizations
rate
is especially
usually
of mesh
Also.
This
log-periodic
other
dense
probe.
extremely
the
This
with
used
require
whereas
for char-
formulation.
elements
cross
planar/non-planar.
than
wavelength),
meshes,
due
with
geometrical
smaller
wavelength.
system
slot
configurations
slot
per
numerical
or spiral
in [13.29.48]).
be employed
Indeed.
of tetrahedral
certain
per
can
[29].
be simulated
In these
samples
elements
condition
can
rings,
is much
tetrahedral
ring
reported
[48.13]
shape
adaptability
in practice,
antennas
the
feeds
geometrical
However,
of slot
etc.),
line
(also
technique
designs,
or microstrip
chapters
of arbitrary
rectangular/non-rectangular
coax
two
mesh
distort
often
leads
generation
the
to
is
",1
Ground
Plane
Thin
Slot
V
Sf
Figure
tedious
and
In this
the
gular
slots.
prism
for planar
Among
is the
simplicity
required
for
slot
CPU
antenna
for prisms
of mesh
generation.
is then
the
applications
and
scattering
5.2
Consider
recessed
presented
are
FE-BI
the
cavity-backed
in a ground
plane.
It can
prismatic
Also,
much
modeling
to generate
hybrid
given
that
containing
of the
physically
integral
be shown
the
to conclude
the
most
(FE-BI)
choice
and
method
trian-
is
Below.
formulation
we
for
functions
of equations.
to various
is
important
set of edge-based
system
u._ing
of unknowns
geometries.
integral
the
element
circular
number
of complex
discrete
FE-BI
this
elements,
meaningful,
fox'mulation
spirals,
smaller
element-boundary
proposed
element-boundary
efficient
A new,
problems
Hybrid
and
the finite
of the
large.
configurations
advantages
modeling.
is unacceptably
elements.
many
by first outlining
microsl
a finite
antenna
an accurate
of cavit v-backed
time
we propose
triangular
suited
begin
solution
chapter,
edge-based
ideally
5.1 : Geometry
antenna
Finally,
radiation
chapter.
Formulation
slot
antenna
To solve
for the
shown
E-field
in fig. 5.1
inside
and
where
the
on the aperture
cavilv
is
of the
cavilv,
a .'.tandard
axJl scatl,.'rillg
al.,l_roach
l,roblenl.-.,
F(E)
where
_r and
medium.
tively,
_r
Z0 and
So and
///, (.
jkoZo
in (5.1)
denote
outer
is required
replaced
in terms
space
aperture
(or
slots)
the
surface
solution
of G.
So and
S:
Specifically,
integral
(BI)
on the
feed models,
such as aperture
coupled
coax
boundary
integral
implementation
context
of the
method
indeed
FE-BI.
meets
the
accuracy
H is represented
JJS
satisfied
G is the electric
on the
(planar.
dyadic
Green's
spherical
by the
f/-
of the
allows
the
Go(r,r
without
portions
and
I _ is the
volume
explicit
field on
knowledge
of H
truncation
field H over
.-q0may
boundary
given
structure.
inclusion
of various
6 for details).
efficiency,
therefore
3. It will be seen
extra
CPU
be
condition
feeding
(see chapter
FEM
Y,':
to as mesh
an easy
etc.
as in chapter
need
the
cavil\
respec-
magnetic
or absorbing
cable,
on improving
will be employed
H = H " + 2jko}o
where
slot,
Also.
magnetic
basis
to (3.1)
metallic
referred
the
of the
constant,
corresponding
(also
as compared
we concentrate
the
by ft. As before,
H on S.f is determined
since
propagation
structures.
H is the
of the functional
chapter
parameters
excluding
feeding
or M i) and
is given
and
version
In this
d/
constilutive
impedance
to the guided
normal
. M'
x fi)dS;
tensor
of E via a boundary
whereas
+ X-- i_
E-(H
free
(J'
over
E)-I,.t_E._.E}dl
o+$1
relative
opening
for a unique
(ABC),
the
a>
:-,)
koZoJ'
[[
rids
the
the
surfaces)
This
E-
source(s)
SI whose
q3.1,
l/j'/- .{(VxE).jTr-'.(V,
1% are
by the
be _,'neralized
the junction
occupied
may
_c,'
0 represents
denotes
i> loe.xtretllizetllefutlclh,l_al
burden.
the
that
this
In the
integral
' ). (_ x E(r'))
ds'
(5.2)
function
or cylindrical)
platform
such that
(refer
h x G = 0 is
to chapter
1 ).
',3
l"or
tho antonna
plane.
pvol,lem
becomes
the
shown
half
space
G =
with
r and
r' being
unit
in a ground
for scattering
To discretize
prismatic
can
the
i+
VV
observation
and
dvad.
p]ane.
elements
functional
(5.1),
as shown
be approximated
using
linear
our
sum
of the
edge-based
The
incidenl
expansion
] =
of a (a\i_v
and
retle('_od
evaluations.
the volume
field
and
i.e. that
parameter
to subdivide
fig. 5.3,
respeclive]y,
problem,
for antenna
we choose
t.5.3)
points,
with
to the
or zero
,'t_,utl_t
l"
integration
H a is equal
i>- ,_ i)).tiar
fullctioll
4=[r-r'
In connection
computations,
the
(;leell's
tt)(' l)l.tf_)rlll
region
in each
of the
using
prisms
[49-51]
Z gv; =
j=l
where[V],
= [{14},{t';},
{l_';}],and
x, Y, z, are
of dimension
ra = 9 and
with
edge
V_ associated
based
jth
functions,
edge
of the
the jth
the
eth
unknown
{E *} = {E_,E_
they
of the
simply'
eth
element.
E;} T. The
represent
element.
coefficients
.....
the
Since
E._ represent
A corresponding
representation
{I_},_,
x, y, z components
V_ are
the
vectors
chosen
average
for the
=
of
to be edge-
field
along
aperture
the
fields
is
E(r) =
S (r) =
i=l
where
IS],
= [_5':_,Sy], and
V(r)
reduces
to S*(r)
when
the
position
vector
is on the
slot.
To generate
and
subsequently
the
discrete
F(E)
system
is differentiated
and
respect
(5.5)
are substituted
to each
unknown
into
E_.
\Vith
(5.1)
the
0
3
9
7
--Q
.....
Z-Z c
i=1,2,3
6
j--4,5,6
Figure
k=7,8,9
5.3:
Right
angled
prism
coettici('ll_,
+ li,"
_=l
where
the
matrix
sums
,=l
are over
elements
the total
are given
A,}
jloSo/il
---/L,/L,
where
the
subscript
is removed
in case
scattering
5.3
problem
are parallel
iL
2jkoZo
[j
iL
_'
q_ " (Hi
of radiation
triangular
is any
position
based
field
xl_ ,
_--1
.V
x M'
"
,,t"
(:)._)
" S_(r')G(r'r')d'_d'_'
to take
problems
Prismatic
angled
(right-angled
edge
(,_.n)
S,_(r')]=Go(r.r')dsd,,'
(5.1o)
the
and
z component.
that
the
same
that
It is noted
holds
for
L_
h'7 when
the
is considered.
the right
ri denotes
']cll,'nts.
x _)ds
Elements
prism
prism).
shown
the
opposite
vector
representation
location
We now design
where
+V
[V S:(r)]:[V'
z in Bi_ denotes
Edge-Based
Consider
-Y" x V a-
2kaS_(r)
+2/L
L_
,_
yr x V,-p_
B'"
number
_.5.t) i
I..'}=()
_=l
by
]-\.ie
1-
wc _,l,_,_il_
of the
to the
terminated
for the
ith
ri)
ith
node,
vertical
two geometric
(z-directed)
quantities
sides
as
q:_ = /-'h
_,
9
'l
node,
gi is the
li denotes
inside
the
triangle.
prism
is to utilize
the
One
the
(.5.11)
unit
length
wav
nodal
vector
along
of this
to obtain
basis
the
edge
ith
and
an edge-
functions
[52]
,,,{_
and
t l,en app]y
more
_],_. proc_.dure
physically
edge
meaningful
elements,
t{eferring
S,_e in (.5.11)
joining
the
vector
gives
the
nodes
:2 -
di-cus_ed
approach
3 and
it is evideEll
of another
triangle
equal.
ll
where
_,.
is equal
S, = _7"
to the
(5.12)
is the
where
integrals
the
if r i- ill ltl,'
12'3'
such
this
that
the
x v t,lan,,,
lengths
definition
_,t'_Iw
t],,'ll
of e<tge.,
of r. we define
integrals
numerically',
by Wilton
expansion
12'3'
for the
in (5.5).
The
corresponding
1._,') 3
= (z_+Az--Z)si
j=4
Vk
= _.k
k = 7,8,9
coordinate
in closed
from
the
When
the
self-cells
associated
We observe
elements
[32] and
volumetric
5.6
with
h = Az represent
(5.13)
form
substitution
which
which
basis
viz.
Vs
respectively.
of 123.
triangular
i=
simplex
to that
(:-'YZ)Si
:hz
resulting
[55].
triangle
by inspection,
be evaluated
except
of the
along
As illustrated
can
(5.12)
V,
triangle
height,
areas
r,)
component
to be used
be obtained
(r-
vector
the edge-based
expansion
_k is the
prism
is. the
of the
is simply,
can
(Xk, yk).
That
ratio
appropriate
functions
the
lhal
\Vilh
all ;t]t,.rllali\_.al.]
a vector
Si = _z
2S _
that
]]o_vvvcr.
to fig..5.2,
area
in [-19.53.51.
are
substituted
as given
of (5.12)
can
in the
into
be performed
(5.13)
the
kth
the offset
into
must
analytically
vertex
coordinate
(5.7),
Appendix
(5.9)
prism
at
and
the
resulting
C.
However,
be carried
out
as discussed
5T
5.4
Applications
Thin
slot antenna
technique
and
strate
validity
the
Radiation
certain
wide)
are
shown
in diameter
ometry
because
shifted
on
whereas
cavity
were
slot is very
narrow
the
finite
the
element
out
using
other.
the
more
Moreover,
tetrahedrals
than
only
analysis
investment
sectioll
of the
system.
[:I(
I_l
to delmm-
prisms
distortion.
narrow
for this
as a criterion
in
As seen.
for modeling
homogeneously
the
filled
structure
for mesh
for modeling
generated
given
for generating
ge-
shown
[29].
be needed
is required
2-1.7
requirements
calculation
needed
slot
for this
elements
were
afore-
cavity
computational
The
the
annular
circular
If a multi-layered
was used
is trivially
and
attractive
tetrahedral
elements
would
of the
is quite
of the
and
mesh
by a metallic
most
1024
prisms
mesh
ill thi_
prismatic
method
prismatic
condition
prismatic
the
was
generation.
such
a struc-
slot outline.
and
In
post-processing
mesh.
Selective
Frequency
and
element
system
time
FE-BI
of the tetrahedral
for acceptable
substantial
Frequency
the
the
slot is backed
portion
However,
number
the
tetrahedral
consider
the
or a similar
much
contrast,
we first
The
2898
considered,
of the
3 cm deep.
each
cavity,
efl:iciencv
and
they
overlay
Slot:
The
carried
ture.
and
formulated
approach.
Annular
in fig. 5.4.
then
accuracy
an
will be presented
of the
by
implementation,
(0.75cm
cm
the
results
capability
scattering
To evaluate
modeling
and
and
mentioned
structures
Surfaces:
selective
periodic
elements
periodic
elements
surfaces
which
may
are
(FSS)
typically
be of printed
structures
sandwiched
form
[56,57]
are
between
or slot configurations
arrays
of tightly
dielectric
designed
layers.
packed
The
to resonate
a=l
2.35
cm
b=0.75
po=
cm
7.7
cm
0.7<f<lGHz
.3cm_L
[
FigureS.4:
Geometry
of the
diameter
at specific
and
and
frequencies.
completely
specifications,
stacked
layer
we shall
FE-BI
surface,
FSS
antenna
backed
by
a cavity
23.7
cm
in
3 cm deep
they
reflecting
are
penetrable
at other
element
the
finite
structures,
arrays
the
Because
element
but
interest
as a function
single
consider
method.
particular
around
frequencies.
may
method
use
be
used
the
element
To meet
resonances
bandwidth
in conjunction
in FSS design
FSS.
impinging
at normal
incidence.
yet
since
with
the
the
array
Assuming
FSS screen
will retain
be used
to find the
scattered
that
its TEM
oT2
E$
1 -oR
near
character,
field
for the
this
design
dielectric
easier
it suffices
T, the
is excited
resonance
the
the
via
the
FSS
characterization
transmission
coefficient
elements)
with
a much
of the
structure
slot
associated
is periodic,
transmission
the
(with
detail
makes
determination
the
structures
to be applied
elements
For computing
as shown
through
has
is the
and
in a cavity
of FSS
fine geometrical
of prismatic
of frequency,
cell of the
analysis
of the
is placed
can
slot
loading.
Here
the
annular
As such,
become
3s
task.
transmission
Of
coefficient
to consider
periodic
by a plane
wave
of
a
cell
wave
transmitted
line concept
,,I I
-20
Prism
Elements
--
Tetrahedral
do
Elements
"_ -50
-60
-70
'
-80
10
'
40
5
theta
10
'
60
'o
'
80
90
IO0
(degree)
-10
_" -20
-3o
o.
g -40
"o
o
n- -50
-6O
-7O
-80
-loo -8'0 -60 -,o -2
i
theta
Figure
5.5:
Scattering:
dral
Bistatic
FE-BI
plane
code
wave
(co-pol)
and
the
is polarized
to that
patterns
in the
fig. 5.4.
The
solid
whereas
the
dotted
The
excitation
0 plane
lines
probe
lines
patterns
prismatic
the
plane.
are
is placed
8'0 ,oo
computed
FE-BI
code.
= 0 plane
Radiation:
from
computed
are
8;
(degree)
RCS
along
is perpendicular
2'0 ;o
computed
at the
the
The
and
X-pol
annular
using
using
point
using
the
and
incident
observation
Co-pol
antenna
tetrahedral
tile prismatic
(y=0)
tetrahe-
normally
the
slot
the
marked
cut
radiation
shown
FE-BI
FE-BI
in fig.
in
code
code.
5.4.
cover
radome
(on e=4.5 substrate)
ES
0.0762cm
0.0762cm
1.2cm
metal
backwall
_I
absorber
where
7 is Ill(' Iransn_i-_.ion
tion
coef[icient
within
the
reducing
associaled
cavity,
the
wilh
of o to less than
0.1.
considering
the
next
the
FSS
direct
and
layer
integral
over
FSS
multi-layer
configuration.
and
e_ = 4.5.
our
calculations
the
more
The
are
traditional
geometry
the
witt_ sonw
cavit\
is also small
I,_ullil_lc
iI_T_'I.Lcli_,t,-
al,>,,vl,,,1,
m'av resmall<v,
tl_a g,t,d
thickness
the
same
geometry)
order
However,
lower
surface
we have
-- T Im )].
of Tds would
[58].
of the
slot. array
was
this
involve
requires
of the
single
on
using
slot,
the
placement
enforcement
leading
of
of the
to a much
of moments
layer
with
more
is placed
the
periodic
layer
element
element
is shown
0.0762
elements
the
periodic
cm thick
is given
measurements
of single
was
in fig. 5.6
and
in fig. 5.7.
and
data
a slot
having
Clearly,
based
on
[59, 60].
radome
considered
6.3072
within
cases,
prismatic
method
sandwiched
for a characterization
a dielectric
agreement
FSS was
applied
In both
in good
absorber
interactions,
[1 -- a(1
approach
calculation
of the
cavity
implementation.
of the multilayer
total
thick
entire
geometry
FE-BI
The
a 1.54cm
lt_'
+lOlg
slot
modeling
of a planar
The
-_
FSS structures.
The
consists
the
intensive
above
and
E*
--
computation
in a thick
computationally
The
traditional
element
boundary'
To reduce'
Since/1'
lOlog
higher
Tds re.as
rio)
A more
base.
H = 1 -
for T is
* dB
upon
}:S_.
to terminale
r)
and
of tlw
lh(' cavilv
it is appropriate
value
approximation
cocfficienl
cm and
the
below
in our study
is comprised
dielectric
the
FSS
is given
of two slot
layers.
For modeling
as shown
in fig 5.8.
in fig. 5.8.
arrays
(of
purpose,
From
the
_l_NmSu_r
I
FE-BI
i
=
.......I
-_,o
-15
L
, .......
....
-25
.................
0
'
2
10
Frequency
Figure
5.7:
calculated
good
Calculations
shown
in fig. 5.6
results,
it is seen
agreement
Radiation
with
is built
on
differs
the
from
fabricated.
a bandwidth
fabrication
used
other
The
complex
results
16
18
2C
through
generated
by the
Slot
PCB
with
and/3
Spiral
antenna
its two
the
FSS structure
FE-BI
method
are
in
design
for it relies
30%,
and
that
geometric
feed network.
shown
= 2.8595cm.
in that
the
on the
the
One
following
One
central
facts
and
in fig. 5.9.
arm
It is noted
the
portion
still
of our
that
goal
the
can
usually
the
is to study
structure
spiral
is designed
requires
central
an-
expression:
this
of the
antenna
This
be determined
that
portion
that
central
details,
Antenna:
arms
180 counterclockwisely.
conventional
of the
of transmission
slot-spiral
o = 0.1333cm
reasoning
(GHz)
of Conformal
a double-sided
because
the
Archemidean
less than
for possibly
that
I
14
measurements.
by rotating
the
comparisons
study
a typical
r = aO + 'fl, where
from
the
Property
Consider
tenna
and
'
12
is not
with
a careful
space
the effect
may
be
of this
slot
array
60mils
placed
belov,
top
surface
of 90rruls
layer
top
mils
of 90mils
laver
Er =
O,
-5
-10
;_-15
I,/
-20
--
measured
calculated
-25
-300
Figure
5.8:
kipper
figure:
0.2
'
0.4
06
geometry
08
of the
'
1.4
multilayer
lower figure:
the
'
12
16
frequency
measured
FSS structure
'
1.8
selective
and calculated
surface
trans-
!)t
spiral shal)con it- performance.
A benchmark t('s) model is d('<ignatedto Ol)(,ral(,fron) l Is ._llll I(, I.')7 .kill,,.
to replace tile conventional protruding blade an/('nua. Th(' siz(' how(,v('r i_.lllll('}l
compact
with
operate
its conformalitv
at 944MHz
o and
3 above
of turns
to 1256.MHz
were
(4.5).
property.
determined
The
cavity
to approximately
The
directivity
higher
frequencies
Figure
MHz,
- 5.12
respectively.
is understood
slot
spiral.
fact,
the
the
both
E-plane
cases
really
center
pattern
the
are
plays
radiation
the
principle
clearly
an important
frequencies,
the EM fields
turns
inside.
entire
designated
case).
is most
The
moves
exhibit
affected
AR deterioration
It is seen,
nevertheless,
bandwidth
In practice,
and
with
we notice
a good
quality
has
that
a CP
a wide
that
944,
azimuthal
absorbing
and
and
1256
1100
the
strong
why
materials
along
the
axial
the
spiral
shape
At
When
the
profile.
The
closer
radiated
AR increases
a couple
(as wide
may
fox.
pattern.
becomes
the
(In
ratio
and
be achieved
angle
It
achieved.
by adding
can
plotted.
E-plane.
radiation
been
explains
mode
center
travels
that
to the center
can be avoided
zlunil)(,r
frequency.
at lower,
optimum
it shows
this
the
center
with
The
a comparatively
and
of
at the
region
and
inwards
also
0 = 90 are
active
3 dB AR
value_
Th('
of 0.3 cm del)th.
be coincident
5.1,
t_,
5.1.
plane
case.)
I)v 1.',
= 2.2)
for frequency
the
in this
and
pattern
in Table
not
less than
region
likely
radiated
varies,
to insure
(_
wavelength
principle
in Table
role
where
therefore
the
band
slab
patterns
may
defined
also
the active
feeds
at
plane
tabulated
lower
the
i.- .-cal'd
1100.MHz.
frequency
tabulated
frequency
increases,
high frequency.
optimum
the
model
frequency
a dielectric
from
show
Eo and
center
free space
are
is not
and
frequency
to the
the
0.011
results
when
Thus
three
design
that
with
the
The
simulation
on this
is analyzed
and
5.10
the
based
is filled
corresponding
antenna's
with
Our
at
of spiral
within
the
as 60 in the
be needed
lo
the
number
of _urns
al
,-\xial
5.5
('omparisons
Concluding
A hybrid
finite
for modeling
narrow
in connection
with
elements,
these
advantages
better
system
The
explicit
tabulated
and
were
presented
elJd-
_f
-tfiral.
the'-It_t
_'-I_'_ i,_lt_
(dB)
of gain
and
[ 1.100
i 6.66
2.7
axial
! 1.23{i :
= "" '
l.(I
ratio
::l
at different
operating
frequencie,-
Remarks
element-boundary
slots
the
meshes
Other
(dB)
Ratio
FE-BI
conditions
mesh
prismatic
and
faster
results
for
the
formulation
cavities.
Prismatic
and
Among
over
the
pre/post
data
processing.
implementation
validity
and
capability
were
tetrahedral
frequency
of the
used
tetrahedral
rates
are needed
simplified.
elements
of prismatic
and
presell_ed
is substantially
elements
antennas
to the
low sampling
process
slot
was
elements
in contrast
them,
generation
for FE-BI
to demonstrate
(FE-BI)
implementation,
the
expressions
numerical
backed
advantages.
and
of the
integral
in metal
offer several
for generating
or
0.9-t4
-"')
I_._,
((',|lz)
(lain
5.1
[,_,oillililt_
is minin_ized.
]"requellcy
Table
llle
elements
selective
technique.
include
were
surfaces
!i(,
/
II
,",.' ,_./',/'/'___.'_\_"_
' "
' ":
of a typical
frequency=l.
/
/
(
\
-/
',
2-arm
k\
slot-spiral
design
1GHz
-t0
-20
-30
)_
/
i"
line: E_theta
frequency
design).
A good axial
freQuenc'y=O
944G_z
[
r
'
-20
'
i I
'
solid
FigureS.1l:
Radiation
Pattern
can be seen
to that
at the
range.
spiral
contour
line:
E_phi,
dashed
hne:
E_theta
at f=0.944GHz
that
angle
-30
the axial
center
This
ratio
(lower
of the
frequency,
indicates
design
but
that
is most
of frequency
pattern
becomes
still remains
the
likely
frequency=1.256
end
number
range).
larger
within
of the
It
compared
outer
turns
in the
sufficient.
GHz
i
\
Figure
5.12:
Radiation
Pattern
solid
line:
that
to those
the
center
the
number
shows
that
insure
a good
dashed
line:
at f=l.256GHz
can be seen
at
E_phi.
the axial
quality
ratio
E_theta
(higher
end
of the pattern
frequency
of inner
pattern.
and
loops
lower
still
of frequency
range).
is deteriorated
frequency.
needs
It
compared
This
to be
certainly
increased
to
CHAPTER
Antenna
For scattering
the
'source',
(5.10)
problems
the right-hand-side
proposed
practical
feed
6.1
and
thin
filament
inside
Probe
model.
cavity',
the
has
incidence
been
However,
several
is usually
explicitly
schemes
Some
considered
as
in (3.7)
and
expressed
for antenna
feeding
antennas.
the
first
the
hi
where
r is the
edges
and
jm
with
coaxial
l carrying
Specifically',
associated
wave
impedance
evaluations.
[61] associated
of these
are
with
discussed
various
below.
Feed
substrates
of length
the
plane
further.
for microstrip
Modeling
Feed
Simple
For
the
excitation
reported
designs
Probe
6.1.1
Feed
where
we have
VI
location
is the
of the
(global)
= jkoZoI
of the
global
m element
feed
an electric
term
ith
cable
edges,
current
integral
entry
Since
in (3.7)
needs
of the
this
as a thin
filament
is located
to be considered
excitation
vector
current
for this
Ki becomes
i = j, ,j2, ...,jm
rn is the
numbering
and
be simplified
I[.
1. V;(r),
filament,
edge
may
thus
98
number
index.
the
index
of (non-metallic)
In general,
i goes
from
m such
jl
element
entries
up to j,,.
are
This
expre._sion
call
b_. further
coincident
with
the current
6.1.2
Voltage
This
Gap
excitation
reduced
lo E,
I. l_r_,vid,'d
l}lal
I}."
Illl
_'*tt,'
i.
to .-t_ecifyiilg
filamelll.
Feed
is also referred
electric
= jlq,Z.l
priori
the
voltage
Since
to as a gap 9_ncvalor
1 across
the
(.] is a vector
opening
whose
and
of the
amounts
coax
magnitude
cable
is the
gap
or any
olher
width,
and
gap.
E lhe
I
electric
field across
ith edge
setting
is parallel
the
to zero.
the
edge
entries
6.2
we have
term
the
associated
gap
underneath
treatment
of the
is because
the
compared
to the
the
3-D
entire
feed
edges
when
not
the
ground
structure
microstrip
gap
voltage
model
can
and
only
and
cavity's
equal
gap
are
.4,
corresponding
to the
value
(i j)
to the
ith
E, whereas
all
set to zero.
antenna
(cavity's
must
be considered
base)
designed
Hence,
type
terms
by tirs_
Model
plane
a single
entry
Io 1 if the
be realized
off-diagonal
set
microstrip
geometries.
using
the
the
in the
line is usually
domain
cosO, is equal
Microstrip
6.1,
the
where
to unity
is specified
Aperture-coupled
network
dcosO,"
vector,
with
in fig.
Ei -
this
equal
Aii
right-hand-side
across
As shown
that
to d. Numerically.
diagonal
For
(global)
other
the gap.
the
is fed with
a microstrip
via a coupling
in the
to have
FEM
is not
special
formulation.
different
conventional
of elements
aperture,
line
size
and
simulation
efficient
This
shape
as
of treating
or appropriate
of the
difficulty
coupling
to fig. 6.1,
small
element
encountered
through
it is appropriate
size
when
the
required
this
aperture.
to separate
in modeling
decomposition
As
the
an
example
computational
the
guided
is implemented
let
us consider
domains
feed
is how
structure.
to model
a rectangular
ltJtJ
aw'rlure
which
(li._cretized
gular
and
using
bricks
the
are
t etrahe(lral
the
substrate
be_t
elnl,luvecl
eh'nl<,l_t>,
candidate,
is of constant
itl l,ractic<
wh_.rea,
since
1 ll_' cavitx
thicktless.
:\Ithottgh
l,otlt
t_'hl-
lit.'
r_'_i_,l_ t,'_t,tlL
i+,, t+_'ctatl_nlar
tyl..>
rll,_\ I.,
if, >llal,'
of ('h'tll_'tlt..
_,ttlldoy
Antenna Elments
("
St
(II)
Truncation Plane
"\\
I
Coupling Aperture
Figure6.1"
Cross-section
region
tation
of an aperture
I and the
domains.
coupled
microstrip
patch
line region
antenna,
II for two
showing
different
the
FEM
cavity
compu-
J
(a)
I I
(b)
Figure
6.2:
slot and
its discretization
region;
edge-based
are
aperture.
mav
again
(c) uniform
field expansions,
different,
and
this
However,
be assumed
applied
(c)
at
mesh
the
causes
since
the
from
meshes
difficulty,
the
any' given
to relate
aperture
frequency.
fields
below
microstrip
across
the
in enforcing
is very
above
typical
mesh
from
cavity
line region.
common
field
narrow,
Therefore.
and
(b)
area
continuity'
a 'static'
the
the
(coupling
potential
aperture.
aperture)
across
field
the
slot
distribution
concept
may
Specifically.
be
the
llll
"eqtti potential" continuitv condition i>.etiforc<'d,at_tltt, t,t+t,c<'cd
tt, tit,>t,. I,'t tl>.lit-t
classifv
Tetrahedral
as follows
Mesh
(Cavity
Region
I):
E_ 1
j = 1.2.3
....
verticaledges
E_ 2
j = 1.2.3
....
diagonal
Brick
Mesh
E_
Then
(Feed
Region
j = 1,2.3
the
II):
....
vertical
'equi-potential'
edges
edges
continuity
Eb 2
condition
,jE
requires
that
(6.1)
c
(6.2)
in which
_:j
whereas
That
t and
d are
is, t is simply
is equal
and
to +1
below
The
the
the
the
the
implementation
would
the
structure.
number
generalizing
be significantly
narrow
and
rectangular
sign conventions
for
this
entirely
diagonal
edges,
aperture.
associated
different
concept.
It
straightforward
inefficient
In addition,
of distorted
vertical
The
with
the
respectively.
coefficient
meshes
above
aperture.
scheme
by
+I
--1
of the
of the
on the
coupling
extended
FEM
width
depending
connectivity
lengths
elements
the
is apparent
for different
if only
one type
technique
ensures
in the
computational
mesh
are
that
minimized.
this
were
system
max" be
approach
geometry/size
of elements
a good
domains
makes
domains
used
that
for modeling
condition
since
the
l(J2
6.3
Coax
6.3.1
Cable
Motivation
The
backed
coax
cable
patch
abundant
cable
literature
feeds
only
integral
pertinent
the
for different
proposed
a hybrid
a constant
element
for thin
substrates
To model
an
modeling
scheme
the
element
and
electrically
The
mode
excitation.
Model
and
proposed
in this
junction
The
in the
model
is then
does
not
the
application
of
must
be
to Green's
function,
we recently
[29] which
is described
the
grounded
base
to the
is only
valid
probe
feed
Method
a more
on the
basis
when
feed
(FEM)
entire
necessary
model
results.
method
of the
of the
presented
(MM)
sophisticated
element
formulation
hold.
is
probe
finite
funclion
(;teen's
a simple
the
section
presence
in
that.
Moment
of the
elements.
feed
the
tec]mique,_
formulation
Green's
But
with
context
described
the
linking
efficient.
of coax
the
approach
conductor
investigation
corresponding
it is observed
substrate,
improvements
cavity-cable
integral
is consistent
tetrahedral
Also,
of the
lnlecl.
limits
configurations
radiation,
inner
this
or ca\i_x-
radiation,
equation
However,
and
complexity'
in the
integral
designs.
- boundary
the
thick
will be first
for feedline.
at the
this
is proposed
edge-based
system
along
low spuriou,-
functions.
antenna
h_l nlicrt_:tril,
experimental
of geometries,
is straightforward
term
assumption
avoid
and
t)resent
geometry
4. For antenna
current
antenna
numerical
To
finite
3 and
linear
class
recoded
variations.
struclure
and
papers
Green's
to those
function
using
a certain
a h'ecting
theoretical
of these
techniques
and
as
of its simpliciu
on the
with
modified
in chapter
exists
Most
for
used
because
[62-64].
available
with
is widely
anlennas
conjunction
the
Feed
hybrid
functional
of a TEM
the
TEM
103
6.3.2
Hybrid FE-BI
System
feed
(as shown
in fig.
6.3)
F(E)
is given
E)-k_E.E}
(E x _:)-
[f(E
-jkoZo
;_ c_,_x
by
E)-I(V
--/%
i,7_t_.tlz_i, wilh
(1)
,(E
dc
i + .=vv
d .'4
r. r')d.q"}
x H) ._dS',
((i.3)
JJc
where
V refers
excluding
to the cavity
the
portion
respectively,
number,
G0(r,
and
the
Z0
r')
is the
of the
by the
space
space
points;
antenna
C is the
elements:
impedance.
with
r and
cross
the cavity
_ and
t_
k0 is the
free
space
I is the
unit
r' denoting
section
aperl
of the
the
coax
ure
denote.
wave
dyad.
and
observation
cable
at
the
procedure
[29], we obtain
a system
of equa-
form
Us_
{[A'5]{E-7}
} + Z
e=l
expressions
surface
Proposed
To proceed
Nc_ OFc(E,
{[BiS]{E;}}
+ Z
e_.S
the explicit
Fc is the
6.3.3
function
S encompasses
permeability;
wave
discretization
N_
term
and
surface
the standard
where
surface
metallic
intrinsic
Green's
the
the
junction.
Following
tions
and
permittivity
free
free
integration
cavity-cable
occupied
relative
is the
volume
with
-0,
(6.4)
eEC
integral
H)
OEi
Bij
may
be found
in [29] and
the functional
on C in (6.3).
Coax
Feed
the
evaluation
Model
Fc = -jhoZo
of
/fc(E
x H) 5. dS,
(6.5)
a boundary
n,odeoll
con_,train_
(' and
rela_in_
consequently
E Io H i.. J,_','ch'd.
lhe
field>
wiTllill
"]o t],i-
cl,_t,
w_,,,..uIll,,,_
]llXl
I,,' cxl,rc>-,'d
a> ,-,','
fi_. 6.4 )
E-
loZo
1+F)1-?.
2rr _7_c
where
(_
is the
coefficient
location.
center
relative
measured
Also.
permlttivitv
are
at r = 0. To simpli_"
the
I0 is the
the
polar
the
analysis,
coax
given
cable:
input
coordinales
g dela)l('s
curren!
of a point
we introduce
IoZo
e0-
(ii.t;i
27
inside
at z = 0 and
(r.o.z)
H = I(1-I'_-o,
r
the
thv
source
in the
relh'clion
al the
cable
sanlv
wilh
lhc
qnanlities
10
2_-_-/_
(1 + F),
ho = :Z_(1-
F).
(6.7)
Hence,
eO ^
E=--r,
and
from
(6.7)
is the
ho and
fields
coefficients.
of the
desired
Co. Note
IS and
(6.8)
it follows
,e/g7
" -Co+--,
ho-
which
h0
H=--o,
H,
constraint
e0 and
and
it is therefore
field
cable
junction
in terms
To do so, we substitute
axisvmmetric
(6.9)
Z0
at the
that
I0
property
appropriate
(6.7)
as new
to rewrite
and
(6.9)
into
of the
new
unknowns
in place
Fc in terms
(6.._5) and
quantities
upon
of the
of these
new
making
use
we obtain
vc = -2,,jkoZo oh;W,,(
),
(6.1o)
(/
where
a and
src stands
functional.
b are the
to indicate
radii
that
of the
inner
h0 is treated
and
outer
as a source
cable
term
conductors.
in the
The
superscript
extremization
of the
I(l'l
XVe choose
and
the
Ill(' corresponding
fornmlation,
mesh
the
edges
inside
1/r
this
and
continuity
these
the
bv
the
fields
of the
it then
follows
cable.
This
field and
the
denotes
excitation
mesh
inside
Fc in terms
into
the
mesh
simply'
cable
modal
the
hybrid
cable
(6.9)
edges
and
field.
finite
on the
cable
element
(6.10).
junction,
-aJk0Z0(b-o){ Iore
the
field,
we made
rather
use of the
than
representing
average
field
the
(').
"I"o overcome
inust
be the
Specifically.
the
cable,
and
coax
upon
fi'om
is
6.11)
outer
surface
constant
to extend
cable
taking
into
Ei.
edge
it introduces
It remains
the
of the
cavity
Fc,
a need
probe.
Then
same
region.
functional
bordering
that
Nc).
the
edges
tield
by recognizing
without
current
as
tangential
condition
inner
into the
the
to rewrite
and
account
to do so
all
Nc
we obtain
vYgTb-aEi
Zo
In b_
a
In this expression,
bohave
across
1,2 .....
system
of the
fiel<ts t(;.(i)
cable
between
used
a fictitious
into
Np(p=
ca\ilv
CXl)al>ion
conductors
equi-potential
I1_ l}li-
edge-l,ased
junction
is also
,,_vil\
nulnt,erof
aperture
in the
a relation
value
cable
between
When
is lhc
to apply
the
tl,c
1_ Jig. i;.t_},_.
modal
outer
and
i=
difference
(6.11)
the
appropriate
provides
over
or those
eoln b,
a
field
TE.M
and
_ti_ctcli:_'
linear
it difficult
borders
the
Vc (.\c
by the
( i.e.
cavity
or to employ
the
......
across
(' i, sl,own
cable
inner
region
potential
of Ei, i.e.
we substitute
cavity
the
coax
the
Ei(b-a)=
condition
the
the
the fields
that
p=l.2
makes
of the
cavity
s('<tioll
junction
between
elcilwn_-
as dictated
cable
we can relate
5_th edge
AV
t/_ edge.
However.
at the
AV=
where
the
inconsistency
difference
(6.6)-(6.8)
tlw cross
constant
cavity.
condition
as computed
tmrahedral
on
to a
modeling
potential
mesh
is set
difficulties,
if the
edge-based
field across
on (').
function
the
linear
the functional
across
the
cable
src
Nc
i=Np
Fc in terms
as computed
(6.12)
of a single
by the
edge
totality
of tile
equal
ele.wnt
properly,
all
the curly
brackels
field_
elenwnt,
extremization
oil
fields
the
cabh.'._
at_ert_lr_ , ,I,_'ca_l>,.
at the
cabh"_
aperture'
of i6.12),
process.
with
Hence.
OE,
the
superscript
3"zjkoZo(b-
(,E,-
a)
li.
at,' _'qual !.
.-q'c. functioll.-
the extremization
_,f tl,c
of (6.12!
I,
=
Zu
i=.\'v(p=
aXi_vlllltlt'tri,
Ill,'
tact,t
it_,,i,l,,
viel(l._
I,_
E,
(6.13)
where
(b - a) 2
(6.14
fl
1
1;
We observe
is just
tion
the
by the
system
entries
where
except
is solved
with
element
Also,
for the
[29], the
cavity
equi-potential
finite
cable.
jskoZo(b-
the
Nc
input
a)Io.
field'
field representation
together
of the
coax
the 'constant
a fictitious
is governed
FEM
that
along
and
(6.11).
(6.13).
Specifically,
matrix
which
entries
admittance
}'+_
are
each/7,
of the
set
at z=0
to
where
Z_ is the
In the
mode
at
above
the
characteristic
feed
model
cavity-cable
we assumed
junction,
an
proceed,
hybrid
we assemble
to the
the
the
Nc
Arc. edges
system
Once
interpretation
diagonal
bordering
is nullified
hybrid
the
FE-BI
everysystem
from
H . rdo
7T
eoln_
impedance
physical
with
to t'i.
210
-
edge
is added
is calculated
If
To
is associated
column
which
each mesh
its meaningful
constraint
excitation
(6.15
Zc"
of the
the
coax
presence
assumption
(6.16)
cable.
of only
which
may
the
not
dominant(TEM)
be
suitable
for
107
certairl
alq)licatiolls.
rnesh
(say.
]o
a distance
applied
at z=-d.
suitable
expansion
OVel'COl]]e
d) into
where
equi-potential
feed model.
far below
distribution
mode.
coax
near
With
cable
the
this
(using
in most
of the
first
understanding,
cases
where
q=r,
12 edges
0 or z, i=1.2,3
(3 directionsx
N i
with
q_,qb
q_, q_band
along
and
_'_ represent
the
coordinate,
and
each
correspondence
between
or z) is given
in Table
When
an axisymmetric
(6.17)
reduces
independent
sion
is the
accuracy
parameters
6.1 along
field
property
0 variable.
introduction
of (6.17)
of the
is demonstrated
brick
Note
1/r
of_lw
lit a frequency
tlw field
fundamental
I'I_XI
by the
caleb',
for tile
field
in the
is
'
(6.17)
-- g)(q_
z in
and
cyclic
factor,
given
for each
of tile
by
(6.18)
rotation
and
correspondingly
7.
Also,
i denotes
the
of the
edge
along
0_ direction.
the node
is assumed,
the
the
for each
tilded
the
we show
that
The
and
in (6.18).
in
z components,
cable
only
expansion
property
coaxial
number
coordinate(r,
of the
radial
particular
edge
parameters
numerator
for the
simulating
the
pairs
of the
format
also that
are
function
-- g)gl_
the definition
element
shape
They
width
numbers
with
lhe ellicicncx
expansion
0, z) is the
_-, 0 and
Aq= is the
to the standard
of the
N;(r,
r,_b and
the edge
a lllt)l,'
the con_l,licalion
of tile coax
tetrahedrals)
per direction).
q_ representing
is operated
mode
E;Nv(r''Z)r
gi
Aqb,__qc(qb
qa
retain
_llt'
or 4 and
4 edges
t'xl_,'lltt
Z
q
to avoid
antenna
suitable
lt,
be dominated
than
i',
cotMitit,ii
thus.
order
a possible
E=
and.
the
higher
rather
it])l)lt)it(tl
(7 will still
shell elements
vanish.
Since
junction
-d
t)II('
,,qui-l)olcntia]
nlodc_
in the
mesh
the cut-off
The'
order
the tetrahedral
ii|]li[_tTit)II,
tile cable.
all higher
of extending
range
lhi>
of this expanmode.
The
2-3 elements
]{r',
TABLE1
Ill,dr
_.Jr,llH,_l,'-
'
p,rlT d[
pint-
_1-2,
il-3,
,:,
o:
.:
Ac,
ol
i
flt_l t'T-
:_ -
A:
-_ -
-_:
21
Ol
Z_
".
Table
6.1:
The correspondence
coordinate(r,
(6.18).
are
needed
field
along
the
efficiency
to model
6.3.4
of this
the
and
our
used
of a circular
0.41
cm
input
input
potential
this
was
model
are
curate
results
Figure
near
6.7 shows
definition
for
the
accurate
to the
(i.e..
and
the node
of the
tilded
linear
many
more
......
prediction
conventional
is apparent
feed
simulation,
One
patch,
feed
over
with
the
agreement
data
two circular
patch
cm)
the
measured
Clearly,
in excellent
the
(radius=2.1
impedance
feed model.
with
02
-_
pairs
for each
parameters
of the
dominant
tetrahedral
tetrahedrals
in
elements,
are
needed
region).
proposed
For
numl)ers
..'"
"-
Conclusion
cavity
impedance
measured
compared
for calculation.
atop
deep.
cable
the edge
direction
expansion
same
To validate
were
radial
When
Results
tions
o or z) along
the
distribution.
between
antenna
filled
was
the
band
from
whereas
the
0.8 cm
on
the
measurements
probe
model
antenna
of radius
a dielectric
2 - 5 GHz.
computed
results
was
with
placed
patch
from
configura-
1.3 cm
(e,=2.9)
the
center
printed
material
and
the
of the
and
yields
proposed
the
equi-
the equi-potential
substantially
inac-
resonance.
the
comparison
between
measurements
and
calculations
for an-
1()!_
other
palch
anlenna
"l-his patch
had
a radius
a loss langent
our
FE-BI
As shown
Ian_=0.0012.
calculation
measurements,
the
accurate
in the context
to the
presented
case
was
[ocaI(,d
coax
also
by the
formulation
is very
demonstrated
how
feed
has
Moreover.
the
a_ld t_)_
agrc('lll('tlt
wit Ii
}>eezl s]_owrl
Io },('
in [65].
mode]
simple
,_l_t
model
structures
Ila_t _.=2.:_:1
in excclh'Ht
probe
feed
IIic c(,Iil(,r,
ca\'i_v
is agai_l
feed structures.
of asymmetric
.-,ll,_,Ital,'
in a circular
model
results
t,\ .\t,_'_h'
<hi tllick
equi-potential
element
It was
to the
feed
0.215
was placed
in modeling
feed.
the
equi-potential
of a finite
probe
generalized
the
pa_ch
as opposed
In conclusion,
as the
The
the
in fig. 6.7,
extremely
of :2.0 cm and
wa_, ill('a_l,r_'d
and
its implenl('))I;)I
as easy
proposed
where
feed
i_,))
to imph'm('nt
model
evanescent
can
modes
b(,
may
be present.
6.4
Conclusion
In developing
lenging
problems
is primarily
on one
results
is usually
around
The
techniques
to solve
in consideration
because
tainties
rate
numerical
the
side.
antenna
On
the
extremely
the
feed
proposed
point,
numerical
for antennas,
feed
to the
in this
as possible,
and
with
a thorough
tations,
we realize
that
they
can
to the
which
with
coincides
group
domain
has
and
certain
the
model,
is one
simplicity.
instrumental
accuracy
the
of chal-
feed
This
uncer-
of numerical
location,
sampling
so on.
closely
As an addition
feed
feed network
efiq.ciency
as aforementioned,
sensitive
feeds
of accuracy,
in fabrication
other,
and
the
be used
of feed
chapter
the
consideration
of their
for mostly
encountered
models,
truncations.
resemble
we also developed
Since
this model
practical
numerical
antenna
a circuit
has to do with
systems
as
implemenproblems.
modal
feed
microwave
ll()
Z
A
Patch
Ground
Aperture
plane
Cavity
Coax cable
Figure
6.3:
Illustration
of a cavity-backed
cavity
patch
antenna
with
a coax
cable
feed.
patch
T
2b
A
2a
cavity-cable junction
(a)
Figure
6.4:
(a)
Side
view
Illustration
to zero
(b)
of a cavity-backed
of the
at the
FEM
center
mesh
conductor
antenna
at the
with
cavity-cable
surface).
a coax
cable
junction
(the
feed:
(b)
field is set
Ill
Co_I:larlso-
oq E-F_elO
A,_or_i,I -Wavelen_l!_
Coax
Cable
150
100
50
=
"o
ft.
u_
-50
-100
-150
2i
3i
4i
5i
6o
segment
7t
8i
9i
110
11
number
(a)
Companson
O! E-Field
O_stnbution
AJong R
6OO0
55OO
-_5ooo
4500
el
z: 4000
35OO
,;,3O00
o
o
2500
g
2000
1500
100_; 02
i
0.04
i
0.06
O.OB
(b)
Figure
6.5:
Field
distribution
ment
method
(a) Field
location
a distance
in a shorted
using
coefficient
of the short);
_/4
from
coax
the expansion
eo along
the
(b) Field
the
shorted
cable
(6.18).
length
along
as computed
--:
of the
analytical;
cable
the radial
termination.
by the
xxx:
(leftmost
coordinate
finite
ele-
numerical.
point
calculated
is the
at
1]:?
-I
I _Lp
E
.......
140
i Z72,
100
=-
""
6O0
20O
.'N
...........
-20 0
I ...........
1.00
I ............
200
I ...........
300
frequency
',
400
500
(GHz)
(a)
180
"w
............
mca.su
rcrncnl
-_
"O
'
Ibis
100
pmhc
method
model
C-
60.0
200
C
-200
-60.0
...........
_............
1.00
_ ...........
2.00
_...........
3.00
frequency
4.0O
5.00
(GHz)
(b)
Figure
6.6:
Measured
and
calculated
patch
antenna
having
cavity
radius
R=21.1mm;
location
probe
vertical
field
x/=0.8
model
wire
input
substrate
cm distance
are
also
connection
for excitation.
impedance
the following
(a)
shown
to the
Real
from
for
thickness
center.
part;
and
patch
t=4.1mm;
Results
for comparison.
patch
a cavity-backed
specifications:
Our
uses the
(b) Imaginary
radius
_=2.4;
based
modeling
incoming
part.
circular
r= 13ram
on
and
the
simple
retains
coaxial
feed
the
mode
120
I00
o
E
=:
80
o
} 6o
"6
-_
20
$
j
2.55
285
27
frequency
2.75
28
285
29
(GHz)
(a)
10(]
8C
6C
_E 2o
=o a
-2C
_ .4Q
_E
o
o
o
2.55
2 6
2.65
2,7
2.75
frequency
2 8
2,9
(GHz)
(b)
Figure6.7:
Measured
tenna
and
having
thickness
following
d=0.21844cm;
tan6=O.O012.
model
calculated
the
[65].
--:
input
impedance
specifications:
feed
location
measurement;
(b) Imaginary
for
patch
from
center
xxx:
this
part.
circular
radius
patch
r=2cm:
xj:0.Tcm;
method;
an-
substrate
_=2.33_
o o o:
probe
subjecls.
CHAPTER
Circuit
Many
netic
designs
signals
to these
After
from
wave
an overview
topic
7.1
cation
element
problems.
This
mesh
certain
for the
for circuit
feedlines
methods
suited
is devoted
techniques,
finite
to carry
are also
chapter
termination
appropriate
truncations
most
of the
truncation
date,
a variety
employed
are not
and
reflected
must
element
electx'omag-
and
applicable
to circuit
we discuss
analysis,
modeling.
a numerical
and
then
turn
to
simulation.
reported.
applied
the
surface,
The
An
simulate
a surface
are typically
mesh
termination
11,5
is the
trun-
scheme
ensure
that
mesh
termination
which
actually
boundary
volume
applications
implementations
truncation
at the
computational
and
element
ideal
or absorbing
ABCs
at the
of finite
backwards
of non-reflecting
for truncating
are
aspects
volume.
scheme
or scattering
been
important
computational
waves
mesh
tions
require
Finite
of recent
of mesh
of the
outgoing
also
antennas
source.
process
Modeling
Introduction
One
ating
the
propagation
de-embedding
the
of microstrip
VII
conditions
at some
to microwave
second
surface
must
surface,
does
(ABCs)
distance
circuits
or higher
order
to truncate
not
from
and
the
exist.
To
have
been
the
radi-
devices
boundary
the
i.e.
have
condi-
computational
llt,
the oIle
way
Wilcox
tion
wave
Epansion
and
traditional
meshes
An
them
control,
when
Basically,
instead
of an ABC,
t,o truncate
the
and
performance
considered
[7'2_,7.3].
not)
s_ill exhibit
cently_
_hough,
absorber
that
sions_
a non-zero
Berenger
his approach
tivity.
In this
_o simulate
is lossy,
the
la_er
found
property
an Mditional
degree
have
ma_ched
Mready
for _runca_ing
highly
medium
is maintained
_s a perfectly
dimensions
sorber
assigning
successful.
been
finite
_o each
l_yer(PML)
difference-time
Nevertheless.
and
from
Re-
In two dimeninvolving
normal
a different
layer.
or
an artificial
conduc-
that
Provided
is chosen
the
Berenger
domain(FDTD)
solutions
medium
refers
of his idea
implementations
that
been
normal.
angles.
is introduced
be noted
is used
(homogeneous
it should
for mesh
have
for modeling
generMiza_ion
<_t"
packages.
material
away
angles.
th__
knowledge'
absorbers
component
thickness
Also.
absorber
components
of freedom
all incidence
.-\B(_'._.
o prior;
for a finite
considered
order
of such
angles
wi_h
eas_, of p_ralleliz_-
absorbers
feld
p_rli_ul_,rlv
approach
of the
and
a reflectionless
this
_he spli_ting
derivatives
manner
a new
a_ i_s interface
requires
incidence
It_,.
ditticul_i_,._
of absorbing
lossy-artificial
at
u._'d,
an artificiM
a _hin laver
[74] introduced
is reflectionless
not available
_i',ll
iti_,.TtIit_,rl_lul,,-
several
requires
for a variety
these
reflection
are
wil h higher
is to employ
truncation.
lligdon_
conformali_v,
modeling
are typically
Nevertheless,
There
dealing
circuit
ABCs
u_i1_
is accuracy
which
the
.-\l_('_
[71].
to traditional
mesh.
structures
in microwave
constants
order
AB('s
difficulties
of AB("s
alternative
l_igher
in guided
Among
propagation
[(i(;.(;7] aml
numerical
implementation
applications
Also.
specific
AB('s.
tion and
nwth_,(I
[(;$.:_1].
problenl
for truncating
the
equatiolj
_o _hree
of the
has
Berenger's
to
ab-
so far been
PML
does
117
nol salisfv Maxw(,ll equationsand
( t: E.M
anisoiropic(uniaxial)artificial
truncating
angles.
FEM
wave
ficient
meshes.
Basically.
the
and
inll,h'_llcntc<t
solution.
.-\new
sors.
}}c ca*itv
canIlot
by making
medium
artificial
angle.
loss in the
finite
can
choices
be made
A PML
material
absorber
appropriate
impedance
incidence
for truncating
This
absorber
layer
properties.
element
imroduccd
rcccntl\
for
al all il_cidcncc
for t he constit
t>arame!er
independent
can then
The
meshes
[771 was
ulive
of frequency,
be constructed
implelnentation
and.
polarization.
by introducing
of this
is straightforward
t_'l_-
artificial
moreover,
suf-
absorber
the
absorber
is Maxwellian.
7.2
Numerical
De-embedding
quantities.
presented
Specifically,
line terminated
from
with
line-to-line
is assumed
any loads
near
shielding
analysis
along
the
line.
with
respect
is a numerical
we are interested
denoting
One
to the
line
scheme,
provides
the
the
is therefore
locations
process
the possible
The
of certain
length
able
circuit
transmission
which
transmission
may
arise
line
mode
discussion.
as shown
in the
distribution
to represent
certain
for a uniform
dominant
5'_j" in this
line is included
to extract
discontinuities,
plane
E field
used
in S-parameters
couplings.
a reference
a transmission
appropriate
wave
here
or line-to-antenna
at and
Consider
full
De-embedding
in fig.
7.1.
computational
anywhere
E field
along
an
domain.
including
the
With
the
The
region
transmission
line
to get
(7.])
where
the
V is proportional
reflected
wave
to the
amplitude,
magnitude
z is measured
of E with
from
the
I;
being
reference
the
incoming
plane
and
_;
11"
Microstrip
Shielded
Line
FEM Region
Substrate
Probes
Figure
propagation
or
7.1:
constant
Illustration
of a shielded
to be determined
and
microstrip
R = _/Ii
line.
is tile
reflection
coefficient.
S11-
in (7.1)
Since
wave
propagation
values
for l,"(z)
"t, ti and
and
at three
reflection,
to determine
points
z_, z0 and
t'(__)
center
FEM
the
line.
point
To simplify
such
computations,
that
the
--
Z0
z0 = 0 and
one
z+ of equal
2,0 --
we choose
z+ = -z_
that
needs
characterize
to specie"
the
the
field
inter-distances
v(=+)
problem,
quantities
them
v(_0)
Z+
along
independent
(7.2)
Z_
the
= d. Given
reference
the
three
plane
field
right
values
at the
from
it follows
V(d)
(7.3)
t,'(o)
t'i + t.;
(7.4)
(7.5)
V(-d)
add
(7.3)
to (7.5)
to get
lherJ
olin_inating
Ii-4-
I; from
17.4_ and
cosh(':d,
from
which
dielectric
-,. can
constant
6_fJ may
1l-d!
i7.7
21 (0_
Tile
then
we (,blaill
I(d)-r
be determined.
(7.(;i.
effective
guided
be calculated
waveh,nglh
A:: and
ct[cclivc
by
"2r,
with
and
,'3 = lm {')}
I;- are
and
expressed
Ao being
the
Ag
free
space
V(0)e
the
reflection
de-embedding
technique
assumption
may
coefficient
As mentioned
Low quality
able.
next
be readily
this
(7.-1),
_,i
before,
high
issue.
is suited
extended
where
performance
(7.s)
(7.9)
_
t'i
(7.10)
for one
port
to two-port
or match
evaluations
in prediction
termination
network
networks.
at port
S-parameter
result
_d -- V(d)
becomes
termination
by' _o/Vi,
terminations
Therefore,
discuss
process
of a perfect
can be obtained
and
v(0) -
$11
This
From(7.3)
2 sinh('yd)
t,; =
the
wavelength.
as
t,
Therefore,
2, once
wave
depend
errors
and
methods
at port
analysis.
For instance,
on the
l.'} is determined.
2 predicted
on termination
make
However.
the
are always
analysis
desirable
<c,'21
by FEM.
methods.
less reliand
we
7.3
Truncation
As already
transmission
only
Using
indicated.
line
may
a dominant
output
given
port
For
(similar
a possil,h,
at a distain
shielded
to that
ctisccmlilmilv
referenccldalw,
microstrip
lines
in fig. 7.1).
the
at
wlwrc
rite
modes
r,'git,xl
input
llwrc
port
underneath
atoI,_
cxist..
(#1)
the
aim
lines
are
by
where
Eo(x)
denotes
(port
1) and
R is the
mission
coefficient
propagation
= {
FI, field
assuming
E = -V.
Eo(.r)(e-"'-" + Re"':)
T Eo(.r )o _:
-
field
at the
FEM
across
model
Sove
mesh
the
port
port.
Vo
this standard
at
the
incident
plane.
T represents
2). and
it is necessary
structure.
(7.1])
plane
the
the
trans-
effective
2, respectively.
at a specific
V20
wave
same
So_,t (port
shielded
shown
found
at the
plane
1 and
- E .':,',,,
- E 2"o_,t
of incident
coefficient
at port
the
it is finally
distribution
reflection
measured
pattern
a static
derivation,
the
constants
For truncating
where
front
be extracted
E_(z)
the
S-parameters
mode.
(#2).
DMT
This
the
static
on metallic
line
PDE
model,
and
with
can
to first
be
determine
accomplished
potential
by
satisfies
(7.12)
a tedious
mathematical
that
'/27/"
sinh
En=I
,odd
A"sinh(
( 7.13 )
( _ d)
"_/b-7,_ d))COS(_ax)sinh(n--_-_(b-y))
where
( Yl Tf tP _
.*.i n
= I'b
sin \ '_-_a / 1
.2A
z<d
121
x
v
0 -_2
Figure
7.2:
Illustration
a/2
of the
cross
section
of a shielded
microstrip
lii_e.
with
-1
G
fn
{r2
cosh
7271"
--a--(b-
d))
7"/71"
-t- _rlCOSh
A complete
at
FEM
fii,_ to truncate
troduces
before.
7.4
Truncation
Below,
followed
structures
ent
the
the
volume
absorber's
loss factors.
performance
This
and
at the
by introducing
truncation
the
the
simultaneously
S-parameters
input
and
EM fields
may
output
inthen
ports
be
as men-
PML
a brief
meshes
system
distributions
Using
with
domain.
numerical
field
an examination
and
nov,, be constructed
computational
the
we begin
by
may
to the
bv measuring
tioned
show
system
an excitation
extracted
(-_-d)
presentation
of the
absorber's
in scattering
as a function
of the
artificial
performance
problems.
absorber,
and
in terminating
Results
of thickness/frequency
are
presented
and
this
is
guided
which
for differ-
122
7.4.1
Theory
is to mode]
the
wave propagation
in t}u,-,(..-,lrucl_lr(,-
ll.,,ill_
ti,_.
7.:1.
I}_, fl]lil('
Electric
Probe
-1
d+t=40cm t=5cm
cross-section: 4.755x2.215 cm
i TM
(a).
waveguide
Absorbing
t,=
layer
_r
h"
d+t=12.0cm
]
p,I--H
L = 2.38
h = 0.21
Microstrip__\ line
,
_?_
H = 1.06
-- - _"
(b).
7.3:
A rectangular
the
element
perfectly
method.
waveguide
matched
_f
cm
(a)
uniaxial
VxE
"-
and
a microstrip
absorbing
medium,
VE)
(_:1
e = 3.2
w = 0.548 cm
Microstrip Line
uniaxial
For a general
._
cm
Figure
cm
line
the
functional
- ko_
E.
2=
E x (_-1.
(b)
truncated
using
layer.
to be minimized
is
EdV
V x E)-
dS,
(7.14)
,IS ,. + So,,t
in which
total
and
electric
evaluated
for this
evaluate
_ denote
field
in the
by introducing
purpose
the
but
the
permeability
medium.
an
independent
alternatively
performance
The
of an
and
surface
integrals
boundary
an absorbing
absorbing
permittivity
layer
laver
tensors
over
condition
may
whereas
S,_ and
E is the
S'o_, must
and
the
ABC
be used.
An
approach
for terminating
the
FE
mesh
be
serves
to
is to
12:t
extract the reflec[ion coef[icientcomp_zled
to t(,rminate
of a thin
and
the
comtmtational
uniaxial
a microstrip
considered
laver
line.
the
plane
(l(,lllain.
for
termillating
Such
wave
reflection
laver
from
Region
Reflected
xv,, (_,l_i(1,'l
a uniaxial
was
an anisotropic
})v Sackinterfac,'
_2-,',1
ill a r_.ctang_llar
prol)o._,(l
Regton
I,_'_
wav_,gui,h'
_t.al.
(se,,
|ig.
[77i wl,,
7..1 1. If
_a_e
TrznsrllJtled
_:_e
[,001
a.
Incldenl
b,O
wave
0 _ C_
Figure
7.4:
Plane
wave
incidence
on an interface
between
two diagonally
anisotropic
half-spaces.
ff_ and
_ are
the
relative
constitutive
parameter
tensors
of the
form
: /a00/
_=_=
the
TE
and
background
TM
reflection
material)
coefficients
be
c_
at
the
interface
(assuming
(7.15)
free
space
as the
become
cosOiR TE
V/-_cosOt
_._
cosOi + V/_cosO_
cosO_ R
TM
cosOi
___
cosOi + _cosO_
(7.16)
arl(I by <h(,o4no.,
= t,, and
c, =
i_ follows
tl,al
l,Z_
/;l xt
angles, irlIplying
a perfectlymalcl,_.dmalerial iI1_evfac'e.
11 xv_._.T ..,= _,-/
reflected
field
for a metal-backed
llniaxial
IR(0,)I
where
t is the
parameter
ok
metal
backed
55dB
if 3tco_O,
The
reflection
ABC
rather
eliminates
thickness
of the
is simi)ly
the
uniaxial
=
than
the
the
finite
and
i)latw
in the
absorber,
a reflectivitv
where
A is the
can
integrals
of -30
the
over
thickness
the
layer
this
the
t)r()l)Osc(l
or
material.
laver
although
is also
Ih('
0.275A
attractive
Clearly.
and
the
background
is more
surfaces.
angh'.
t3asicallv,
by backing
backing
is not
incidelwe
of the
further
PEC
wave
(IB if .:ttco.,O,
wavelength
be reduced
However.
(,.1,
0, is the
wavenumber
has
is
= e -2._t:''_'
laver
(7.17)
a PEC.
altogether
with
an
because
it
the interface
true
for Barenger's
absorber.
Below
uniaxial
we present
for
microstrip
and
a2
line
----- #rb(O
constitutive
7.4.2
a number
absorbing
frequency
layer
the
of results
as a function
guided
structures
it is necessary
-- j,'3)
for the
parameters
which
of the
shown
to let
the performance
parameter
3, the
in Figure
7.3.
a2 = _b(c_ -j_3)
permeability
of the
show
background
tensor,
where
material
(i.e.
of the proposed
laver
thickness
We remark
that
for the
permittivity
_b and
/_b are
the
t and
for the
tensor
the
relative
guide's
cross-
substrate).
Results
Rectangular
Let
0.SA.
coefficient
is reflectionless,
PML
laver
laver
_'.II_'
us first
section
mode.
Waveguide
consider
has dimensions
It is excited
the
rectangular
4.755cm
by an electric
waveguide
x2.215cm
probe
and
at the
shown
is chosen
left.
and
in fig. 7.3.
The
to propagate
fig. 7.6 shows
only
the
the
mode
TElo
field
strength
inside
uniaxial
laver..-\s
for
._ values
reflections
about
the
wax'e_uich'
expected,
less
than
from
the
t is the
here
1.0 and
wave
metal
travel
decav
is seen
the
decay
distance
factor
plot
mode
reflection
TElo
to normalize
absorber
with
respect
performance
and
an optimum
value
suggest
B must
that
decay
without
reflection
may
amplitude
value
of B changes
still
maintained
/3 values.
to 2/dt/Ag
many
absorber
be chosen
before
since
which
absorber
thickness
when
the
entire
the
backing.
broadband
case of f = 4.5GHz
of-45dB
coefficient
be noted
2era or 1/3
whereas
though
free space
of -37
that
the
the
That
line)
value
wavelengths.
of
the results
the
slowest
lowest
to reduce
this
of the
the
of the
the
optimum
absorber
are
for unoptimized
optimum
value
of
of _3 = 3 (corresponding
dB which
setting
we chose
is, the
is used
existence
to provide
be achieved
(dashed
rapid
to look
Basically,
properties
and
7.7 provides
As expected,
can still
the
where
and
width
where
7.7 is typical
backing.
absorber
absorber's
but
a reflection
It should
is about
the
absorber
low reflections
coefficient
= 2.3) gives
applications.
for a given
acceptable
a reflection
nature
coefficient.
frequency
in the
Figure
its broadband
the
fig.
of 231/Ag.
A,.
\'._\\'l{
prompled
and
reflection
it reaches
with
wavelength
from
values,
is therefore
the
reflections
For example,
/3 = 1 gives
guided
of
[_ = 2.]//Aj
to larger
thickness
lhe
= t'JE;_"_'l',
interface.
as a function
demonstrates
be achieved
wave
absorber
a \'S\\[_
1o unilv,
by (-_t.t_o,_,
One
allI
<],cax I_ ._lZl_l_I'>_
('onsequenll.v.
3 is increased
zl_a_cllcd
,or i_ cxl)t,llclllial
sufli(ieI_/
absorber
VS\R's.
of D for minimizing
causing
the
coefficient
to the
al)_of
laver.
given
from
for a given
}_\ a t,'dccllv
as 3 is increased
is precisely
when
the
haw'
5cm
However.
unacceptable
decay
nol
of this
that
Icrmillal'd
inskIe
does
measured
It is noted
to cause
decay
wave
backing
for an optimum
of the
field
for 3 = 0.5.
the wave
0i = 44.5 .
lhe
unilv
1.1 is observed
is nearly
is still
acceptable
3 = 3 allows
Also.
as can
use
for
of an
be realized
presented
curve:
refer
to a s_mpling
role
of aroultd
example,
the
1> ,\.
4t ,\. _l_!
t}ll>
example.
Not
surprisingly
important
setting
7.7.
always
Our
performance.
microstrip
constant.
show
that
performance
other
it is expected
it is therefore
of the
choices
that
will
as the
play
not
is demonstraled
the
a role
to choose
play
all
case of _ = 1 stlowll
of o give
recommended
and
contact
exhibits
coefficient
at the
a thicker
absorbing
example,
the
is illustrated
coefficient
the
the
the
optimum/3
similar
in lhe
al>orl)cr
])reseIic_'
o = .3 to ensure
value
of o plays
in fig. 7.10.
the
absorption
exists
However,
it was
behavior
_,f
I llal
role
is maximized
the
verified
is the
waveguide
for approximately
and
effective
dielectric
a 1.87
the
cm
absorber
thick..5
to avoid
waveguide,
we again
that
absorber
in the
The
performance
case
of the
absorber
behavior
of the
microstrip
same
reflection
is required.
of the
the
shielded
is plolled
in fig. 7.6.
as in the
performance
of importance
of 2/3t/A 9. For
to the
if better
Again
in the
inside
the
coefficient
is the
with
as shown
and
be required.
little
e_:l
Similarly
and
in absorbing
reflection
up to 4 layers
= 1 is now -42dB
may
the
is terminated
wall.
attenuation
laver
A0/ex/g777_
H and
line
metallic
/3 value
same
laver
Ag =
microstrip
the
uniaxial
line is extended
with
as a function
that
where
an optimum
wave
matched
is illustrated
case.
absorber
that
perfectly
of 2/3t/kg,
In this
electric
observe
and
line mode
as a function
observe
also
the same
and this
of _ does
line
performance
layered
of the absorber
value
are absorbed.
Mierostrip
the
tests
modes
all modes
an
for this
o = '3 gives
However.
attenuating
The
(6.6)).
As seen.
in fig.
(see
waveguide
and
reflection
examples,
value
this
we
of 2::1/A::
1:77
il should be noted thal the discrclizalioli rate t_la\.-an ,qu_ll\ iIlll_wl.tlt rL_h'aI_,l
this needsfurther investigation.
"l-heaccuracyand validity of the PNIL applications for circuit parameter (oultmrations
of the
input
can
optimized
impedance
measuring
VSWR
the
also
the
which
shielded
be seen
from
the
4.5 cm
PML
values.
The
vertical
field
is often
difficult
microstrip
resuh
layer,
illustrated
with
shown
microstrip
at the
probe's
with
in fig. 7.11,
1 and
line
location
unstructured
line dimensions
,t = 1. yields
impedances
without
finite
for the
data
lI is seen
element
are given
very
were
a need
that
usc
accurale
conlputcd
_v
to extract
tlw
meshes.
Note
that
in fig. 7.11.
Meanderline
Another
example
the structure
The
along
cavity
the
of thickness
agreement
is the
was placed
was
y-axis.
meander
domain
t = 0.12ram.
with
the
shown
in a rectangular
tessellated
The
line
measured
using
cavity
$11 results
data
[78].
For
of size 5.8mm
29 150 5 edges
was terminated
The
in fig. 7,12.
with
are
and
only
a 10 layer
shown
the
FEM
18.0mm
150 edges
PML.
in fig 7.13
each
and
simulation.
3.175ram.
were
laver
are
used
being
in good
E_ecl,,c
.....
Rill
PIII_
....
Imlglmlty
P_obe
,/
Pmrl
Mllgnitudl
o
o
L,
vll
13
'
,
\
\
\
/
'
"
_=c%=
,
\
'
t..t,=lay=
d+t=4Ocm
io
_c
ac
?o
I.e-
-1
Cz=
1-jl_
-1
_LL= 1-j[3
t,=lcm
ic
SamplesAlongWaveguide(14umplt_.)
Figure
7.5:
Typical
nated
field values
of TE]o
by a perfectly
mode
matched
inside
uniaxial
40
a rectangular
waveguide
termi-
laver.
35
--
beta=0.5
1.
30
--
3.
--10.
\
o
-2,
20
e_
"E
\
\
6'2
64'
)
66
Segments
Figure
7.6:
Field
values
fectly matched
of the
,
70
'
68
72
TE]o
uniaxial
mode
layer.
element
was 0.5 cm which
at 4.5 GHz.
_
74
(alpha=l.0,
inside
*_
_-_
78
...
80
f=45GHz)
a waveguide
The absorber
translates
-_
76
to about
terminated
is 10 elements
13 samples
by a per-
12!_
-5
-10
--
f=5,0GHz
-15
-20
-25
-- -30
tr
I
j
1//"
/
-35
-4O
-45
-So
_o
_2
,.
213t in kg (_=1.0)
Figure
7.7:
Reflection
coefficient
layer
to terminate
used
-5
vs 2'3t/kg
the
(o = 1) for the
waveguide
shown
perfectly
matched
uniaxial
in fig. 7.6.
--- f=4.0GHz
-10
A
"l_ -15
-20
i=
g-25
-- -30
g
-35
-4O
-45
-50
0
10
12
14
2_t in kg (c=_)
Figure
7.8:
Reflection
uniaxial
coefficient
laver
used
vs 2_t/,_g,
to terminate
with
the
shown
perfectly
matched
in fig. 7.6.
l:i(_
oj
t---
f=40GHz
i---
f=45GHz
--
t=5.0GHz
....
/I
J
-10
A
-15
o -20
E
D -25
"D
c: -3o
E:
-35
-4O
-45
-500
0.5
1.5
Figure
7.9:
Reflection
coefficient
terminated
by
-5
the
I=4.0GHz
---
f=4.5GHz
--
f=5.0GHz
-10
---
'
2 5
3 5
in kg ((z=1.0)
vs 2j3t/Ag
perfectly
2
2_t
with
o=1,
matched
for the
shielded
uniaxial
laver.
3.5
microstrip
_'_ -15
-20
o
E
m
_
--
-25
-30
-35
-4O
--45
-500
0 5
15
2 5
213Iin kg (a=13)
Figure
7.10:
Reflection
line
coefficient
terminated
by
vs 23t/Ag
the
perfectly
with
o=
matched
3,
for
uniaxial
the
shielded
laver.
microstrip
line
1:51
70
65
FEM
60
55
\,,.
r..,_w
_= _
_h;-
J.,._zo,,,.
50
_,
%,,
64s
_
i
mLII
"
H,IB_
40
_32
""_?"
35
25
20
0.2
Figure
7.11:
Input
'
0.4
impedance
compared
to the
0 6
'
0 8
1
width of microstrip (cm)
calculations
theoretical
for
reference
the
'
1.2
PML
data.
1.4
terminated
microstrip
as
.305
.305
1.611__1_1.s2s
i-
J--r
! 1.61
_i-_i
--
in
q
O'J
(unit: ram)
0
Figure
7.12:
Illustration
surement.
of a meander
line geometry
used
for comparison
with
mea-
meander
line
I
_+
0.9
08
f_
0.7
+
0.6
0.5
1
!
0.4
--
Measured
-
0.3
- Calculated
0.2
0.1
1'o
0_
1hS
20J
frequency
Figure
7.13:
Comparison
shown
of calculated
in fig.7.12.
and
25J
30
(GHz)
measured
results
for
the
CHAPTER
AWE:
8.1
Brief
Overview
Although
typically
full
only
of these
is therefore
needed
wave
for
electromagnetic
the
Waveform
Evaluation(AWE)
ing model
of the
relies
analysis
at each
and
in conjunction
impedance
frequency
the
finite
are large
bv the
designer
impedance,
or analyst.
S parameters,
of the
is one approach
or other
useful
of the given
value.
cumbersome
in minimizing
response
method
in this paper
with
and
consideration
frequency
on a Pad;_ approximation
of the system
cuit
input
systems
(input
important
generating
Evaluation
are needed
parameters
an
Waveform
of AWE
a few parameters
der modeling
etc)
Asymptotic
VIII
It has already
The
to construct
a reduced
its application
element
to simulate
pattern.
Asymptotic
order-
parameters.
to avoid
been
or-
requirements
parameter.
we demonstrate
method
CPU
electromagnetic
parameters
A reduced
far field
the
solve.
to
the
applied
repeated
AWE
solution
to problems
and
validity
full wave
in cir-
when
used
electromagnetic
problems.
The
eling
cuit
method
of Asymptotic
of a linear
analysis
ports/variables
system
and
to approximate
in circuit
Waveform
has
already
the
transfer
networks
Evaluation
been
successfully
function
[79-$2].
133
The
(AWE)
associated
basic
idea
is a reduced-order
used
in VLSI
with
of the
a given
method
modand
sel
cirof
is to de-
131
w'loI_ an apt_roximale
SlWclral
solutions.
whose
Typically.
coefficients
available
are
spectral
In this
imating
the
the
knowledge
of the
method
a priori
tern.
Given
ments
first
proceed
efficiency
Theory
8.2.1
FEM
in full
antenna
recasting
the
for the
AWE
the
fields
FEM
method
input
system
system
entire
bandwidth
frequency
with
parameter
(state
variable)
of the
grid.
AWE.
its application,
of a shielded
Also.
of that
of CPU
computational
for application
from
involve
reduction
for
to frequency
samples
systems
or
A\VE
points.
respect
a few data
demonstrate
impedance
of radiation
is to use
an
a dramatic
in the entire
and
_lll_'
for" apt)rox-
interest
a single)
with
nwthod
simulation
equation(PDE)
have
for a given
i- l,_,.lul,,l,'d
r_'l_r_'-'lltalioI_
AWE
pattern
system
indeed
of the
wave
over
a backscattering
differential
can
of the
at a few (even
partial
t l,_' t'adi'
Of particular
of the simulation
system
System
application
amounts
to fill-in
a response
for computing
8.2
tional
the
applicability
of the
fui_clit,ll
svs_en_.
parameter
solution
AWE
the
to describe
The
wave
practical
in generating
describe
tions
full
knowledge
to resolve
the
impedance
unknowns,
a need
complete
of lh_' trarl,.fl'r
I_v ll_atclli_lg
in electromagnetics.
input
that
of the
of a given
can be used
from
of a _ivct_ tim'al
deternlined
we investigate
problems
thousand
lhell
response
evaluating
fultctiorl
a Pa_lb ext)ansioi_
solutions
chapter
scattering
the
Iraxl.'.fer
microstrip
pat-
several
requirewithout
Below
we
We then
accuracy
and
stub.
Recast
of the
to generating
finite
element
a linear
method
system
to full wave
of equations
electromagnetic
by extremizing
soluthe
func-
[83]
_'=<VxE,_.VxE>-k2<E._.E>+kb.t.
(8.1)
135
wher*' <.> denole_all illllei-producl and b.t. i: a l>t,-_il,h' I>tmn,iar\
_,'liI_ x',},_,,,'
specific
7 ar,' ilialt.-
forlli
rim related
operaling
is not required
coefficienls.
where
A,
describing
the
putationally
response.
of (8.2)
is not
AWE
below
mentation
of the
(sparse)
matrices
direct
or iterative
using
though
and
must
be further
analyses
and
an efficient
approach
wavenumber
in our
{fl}
being
subsequent
(t>l'rl'>l/t>llttill_
A discr<'tized
[c)rni
<if tS.l)
= {./}
(S.2)
and
{f}
is a column
methods
for a given
to these
in connection
it turns
out
and
therefore
independent
presentation.
repeated
designs
in connection
can
solution
that
with
the
the
antenna
excitation
be stated
system
matrix
value
(8.2)
k to obtain
both
temporal
burdens
and
a repeated
is com-
and
An application
is an attractive
alternative
and
system
microwave
column
{.f}
(8.2),
whose
circuit
is a linear
This
of
and
imple-
problem.
function
as
= k {.f,}
of frequency.
fl'e-
solution
responses.
FEM
of
a frequency
require
these
responses
with
of the
for each
may
in generating
an approximation
AWE
the
computational
{f}
with
{.\'}
A, is sparse,
additional
problems
_' ib l lie
is [29]
square
is considered
For these
conditions
usual
certain
we formulate
ali(l
of lighi.
+/,'2A:)
placing
to achieve
speed
boundary
be solved
intensive
responses
the
+/,'A,
Even
quency
is the wa\etlllllliil'r
7-7aiid
excitation.
can
Also,
(A0
specific
wavenumber.
discussion.
c being
appropriate
denote
(8.2)
,k
with
the
the
Clearly
the
A"
frequelicy
incorporating
for thi,
(8.3)
observation
will be specifically
used
8.2.2
Asymptotic
%Vaveform Evaluation
expanding
Ill(, solution
{.\'}
in a "Ia\-l_)r :('ri('-
will,
al)olLt
IIw l:t_._l,
/,_, a>
{-\'0}
ducing
(8.3).
is the
solution
this expansion
after
some
of (8.2)
into
(8.2)
corresponding
and
manipulations
equating
we find
to the
equal
wavenuml_er/,'u.
powers
of/,'
By intro-
in conjunction
with
that
{Xo}
koA, o'{f,}
{X,}
Ao I [{/,}
{x_}
-Ao'
[A, {X,}
+ A2({Xo}
{x,}
--&-o'
[A, {X,_,
} + A2({X,_2}
(_..I j
A,
{Xo}
2koA2 {Xo}]
(s..5)
+ 2ko {X,})]
with
-&-o= Ao +
Expressions
system
may
be
at the
be needed
stored
impedance
of the
(the
(8.5)
computations
.\'f(/,').
to as the
wavenumber
as indicated
out-of-core
entry)
referred
prescribed
computational
pth
are
in (8.5),
for the
The
(k0).
this
inversion
an
only a single
entry
can
in the
of {X_(k)}
then
whereas
explicit
is used
of AWE.
interested
moments
moments
Although
typically
above
(s.G)
system
implementation
we are
domain,
_2
4- koA2
]goAl
inversion
repeatedly
Also,
given
field value
needs
be reduced
(8.6)
and
that
at one
is the
of Ao _
can
for input
location
be considered,
to scalar
thus
form
say
and
l:ff
the expansion:so]ution
(_.5i
at k0.
moments
h_+come e_scalar
"Io \ie[d
a Illore
to a Pad/_ expansion,
special
case
of the
qlh order
where
a, and
by partial
1 -4--b,(k
analysis,
fraction
cxl)r(,ssiolL,
an expansiotl
],'o) + a2(k
- k0) + b2(k
q) are
it is observed
raliutlal
is gi\etL
t'o) _ + ...
to as the
that
[82.84]
rcvctt
I[_,
itl fotnt.
:\
hv
t'_t "_
+ b_(]," - k,)_
Padb
the Padb
ilt>tt,at]
t'utlcliotl
+ a,_(t -
- k0) 2 + ...
referred
decomposition
wc carl
is a conx('nliotLal
of such
b+ (i = 0.1 .....
For transient
convergciit
which
ao + at(k
.;'(k)
(_.T)
coefficients.
expansion
as
q
= x+ 0+
(s.s)
i=l
where
Xq0 is the
tation
suitable
(r, and
and
important
a better
residues
and
k0 from
the
origin,
accurate
merical
svstem,
of Pad6
expansion
mav
hybrid
to instead
beauty
element
because
with
accuracy
a more
generate
plane
number
can
be found
involved
be associated
attractive
is the
can
finite
the
expansion
which
the
be much
of AWE
method.
for instance
-
boundary
the
smaller
relies
to the
point
that
and
on the
usually
dominant
on the
real
needed
of the
(Detailed
In gen-
analysis
axis
to obtain
original
and
nu-
theory
in [85].)
integral
full}- populated
complex
of AWE
residues)
poles
poles
system
approximation.
and
than
aim
physical
residues
(and
represen-
residues
original
of the
closest
of poles
this is the
The
of the
accuracv
system
in a complex
in practice
to those
more
Since
Clearh.
transformation.
the
contains
approximation.
located
to infinity.
correspond
in determining
expansion
poles
a sufficiently
is more
k tends
domain
or (8.8)
roles
order
provides
For
when
for time/frequency
higher
AWE
value
ko + hi) in (8.7)
play
eral,
limiting
submatrix
dependence
system,
the
implementation
of the
on frequency.
by introducing
overall
In this
a spectral
of
system
case
it is
expatlsiotl
conventional
approach
of ('l'l
to continuous]v
requirvlnenl.-
repealing
the
_vlmn collltmTc<
solulion
of t]u'
elixirs'
system.
8.3
Numerical
hnplementation
As an application
the
evaluation
rectangular
a strong
of A\VE
of the
cavity
input
of AWE's
The
cavity
shielded
resides
on a 0.3.5cm
0.79crn
wide
at the
perfectly
and
absorber
was
alreadv
demonstrated
The
finite
element
reference
in conjunction
to approximate
(PEC)
up
and
and
from
1 to 3 GHz
the
the
stub's
in a metallic
input
example
impedance
is therefore
4th
reference
chose
order
wall
is
a good
and
response.
the
stub
single
was first
8th
intervals
input
order
As
seen
the
cavity
In this
problem
AWE
is
having
the
artificial
than
FEM
to es-
system
analysis
was
[86].
using
a full wave
to serve
solution
representation
the
stub
is terminated
rather
(50 points)
in fig. 8.3.
The
study
computed
impedance
stub
absorber
numerical
circuit
microstrip
of 3.2.
by' an artificial
forced
at 40MHz
the
that
for microwave
shielded
the
constant
#_ = (1.0. -1.0).
accurate
of the
Vv'e then
back
Nevertheless
response
with
ion. we consider
shielded
in size and
We note
an appropriate
interface.
valid
solution.
this
a dielectric
1.8 GHz.
(3.2, -3.2)
for setting
code
and
the
1.06crn
having
conductor
matched
frequency
GHz
x 6.00cr,_
substrate
of e_ =
a perfectly
stub
As expected,
1-3
tablish
microstrip
simulat
capability.
thick
used
electromagnetic
for
8.1.
from
is 2.38em
electric
constants
in fig.
of frequency
demonstration
relative
impedance
as shown
function
to a full wave
at
1.78GHz
given
4th
as the
order
in (8.8)
A\VE
C_'])Ie'_'IIt atioll
i._ iT]
/-ton!,_'enwtll ',_,il h t }if" r_'al al](] rt'a(1 i',,_' t_a_l - t_l ltnt, l_'[t'I,'[_( _' iTlt_llt
iml,edarwesolutiol]
shows
that
accuracy
AWE
both
the
impedance
order
AWE
original
to the
real
and
computations
products.
are near]y
order
Thus,
magnetic
svstem
numerical
over
that
t_]e
etlt
the ('])I
of these product
q and therefore
that
to demonstrate
point.
simulation
is required.
systems
be considered
Surprisiligly.
loll
solut
li,i>
ch._,rl\
reactiw'com]_olwnts.
band.
poles
r_'_.l,_'_ti\_'l_,
ire
I-:{(;]1/
requircnlent,,,
for
a low
operati(ms
mnch
smaller
A\VE
efficiency
for
for-lt]_
n_ozo times
of
of
thaw the
size of the
system.
It is also apparent
compared
3:9_; t,al,duidtl,,
system
The number
approximation
when
and
the reference
components.
numerical
system
5(;';
of llw
matrix-vector
will
about
contributions
difference
and 8th
the
over
AWE
representation
development
and multiple
in the future.
the
time
and utility
parameter
useful
a wideband
of the
simulation
solved
the
is an extreme])"
when
we only
addition
frequency
method
for
can be readily
to electro-
response
more
of the
complex
extended
and
l ltl
1.06 cm
,--Jr
2.38 cm
Figure
S.l"
Illustration
of the
shielded
microstrip
stub
excited
with
a current
pro],('.
i
50
t/
45
1 /
/
4O
/
/"
_35
/t
l//
25
20
15
10'
,:2
,:,
1:_
1:8
frequency
Figure
8.2:
Impedance
calculations
shielded
microstrip
and
dashed
the
computations
using
stub
line
are used
traditional
shown
denotes
2:2
2:,
2:_
2:B
in GHz
in figure
the
as reference
imaginary
FEM
8.1.
frequency
Solid
part
for comparisons.
line
of the
analysis
is the
solutions.
real
for a
part
These
111
45
_25_
, ,"--__._________
-"
-_h /
/ /
I-_-o
L--'w_-'''_r
,/,'
15k
,oh, ........
1
1.2
1,4
1.6
"-.
i
--1
1.8
2
2.2
trequency in GHz
24
2,6
2.8
(a)
50
i-:o
45
40
gas
830
g
E 25
2O
15
IIIIII
10
1.2
1.4
24
frequency
26
2.8
in GHz
(b)
Figure
8.3:
order
achieved
tations,
for the
Imaginary
with
solutions,
real
respectively.
excellently,
(b)
AWE
the
Part
(a)
and
It is also
reference
computations
56% and
data
using one
reference
33% bandwidth
imaginary
shown
the
(b)
that
over
parts
the
the
8th
entire
point expandata.
With
agreement
can be
of impedance
order
band.
compu-
solutions
(a)
Real
agree
Part
CHAPTER
IX
Conclusions
9.1
Discussion
During
pected
the
and
period
tational
made
the
issues
of physical
of results,
lack
demands,
during
were
systems,
We can
course
Work
the
hybrid
frequently
finite
data
Among
of mathematical
state
Some
methods,
for comparison,
comfortably
of this work.
element
encountered.
development
of measurement
etc.
the
Research
of developing
unexpected
understanding
tation
on
that
ex-
are
tile
them
models,
and
many
interpre-
increased
significant
progress
of our accomplishments
colnpuhas
been
are
summarized
were
mathemati-
below.
GENERAL
Once
cally
PURPOSE
the
FE
and
formulated,
HYBRID
BI
The
(SDRC-IDEAS)
mesh
geometry
and
geometrical
trary
be used
meshing
and
It is this
to simulate
was
was
The
an 3 planar
then
developed
task
of the
conformal
142
hybrid
FE-BI
of printed
version
DEVELOPMENT
method
devoted
the
latter
meshing
general
the
between
generator
data.
METHOD
and
effort
interface
modeling
shape.
subsystems
a major
two subsystems.
FE-BI
to the
program
with
was
antenna.
and
minimum
important
antenna
FE-BI
integration
a commercial
but
sufficient
in permitting
configurations
code
of the
that
can
the
of arbi(in
theory)
1t3
I"rI-:RA'rI\'ESYSTEMSOI.\'ER
A nmn,orv saving algorithm I-I-PA('I'_was inte;twin,'d witiz t],," ltvt,_i,I tli_l
subsvslen_to regisler on]\ the non-zero FE.M entries. The l{i('(;
itcrativ,.
solver was independently developedfor partially sparseand Imrtialt\ full n_atrices in conjunction with the ITPA('K algorithm.
UNIFORMGRIDBI SUBSYSTEM
-- BICG-FFT
To facilitate the efficient storageand evaluation of the BI sub-system,a uniform right triangular zoning schemefor discretization of the boundary integral
equation wasintroduced by re-numberingthe triangle edgesasdictated by t heir
geometricallocations. This approachleads to a BI sub-systemwhich could be
cast as a 2-D discreteconvolution, thus allowing useof FFT for fast execution
of the iterative solver. This truncation/termination the "exact" evaluation of
rectangular and right-triangular patches:.
NON-UNIFORMBI SUBSYSTEM
-- OVERLAY-BICQ-FFT
For non-rectangularpatches,an interpolation schemewasproposedto makeuse
of the efficient BiCG-FFT techniqueby overlayinga fictitious uniform grid with
the original arbitrary mesh. The forward/backward transformation matrices
to account for field interpolations using localized basisfunctions were derived
and they wereindeedhighly sparse.
FEED MODELING
Feedmodelingis oneof the most important and challengingtasksin the context
of the general purpose FEM. To this end, a seriesof commonly used feed
structures weremodeledusing the hybrid technique,especiallyin consideration
of efficiency
and
accuracy.
These
include
probes/generators,
aperture
coupled
1t-t
slotline, microstrip line. coax cal_lc,etc.
PRISMATI("FE.M EI.EMENIS INCORPOltATI()\
A major problem in any hybrid FE._I analysis i> tlm re(tit,us i_l,.-l)I_,(v,>il_.,.:
for mesh generation. Thin laver substrates in the presenceof _hi(k
are
often
found
configuration
used.
leads
To
and
certain
to large
alleviate
developed
these
MESH
using
ability
of the
similar
FEM
ORDER
been
elemenls
does
are
elements
formulation
not.
It can
as a subsystem
technique
the
studies
were
exhil)ils
therefore
module.
be readily
accom-
adapt-
PML
were
circuit
may
was
carried
simulations.
truncation
truncation
APPROXIMATION--Ate,
reported
varying
the
promise
its
This
simulation
Hence
mode
mesh
letrahedral
typical
AND PML
to microwave
dominant
this
(DMT),
first
introduced
out
to optimize
In the
FEM
domain
and
VLSI
applications.
this
technique
the
meantime,
was proposed
of the
into
an
and
im-
for microstrip
structures.
highly
we consider
hybrid
PML
FEM
However.
edge-based
FEM
medium
performance
as an alternative
Given
tetrahedral
Various
plemented
has
tetrahedral
elements.
approach,
and
system.
tetrahedral
analytical
AWE
hybrid
anisotropic
of the
REDUCED
when
_3,rlWH DMT
application
and
systems
in the
the proposed
lines
numerical
prismatic
the
or other
plished
designs.
that
to compensate
uniaxial
antenna
the
incorporated
TRUNCATIONS
conformal
difficulties,
features/advantages
be used
The
in practical
spa('('l(sl
useful
frequency
of the
application
in RLC
responses,
method
rE
for broadband
to electromagnetic
is particularly
simulations
system.
For wideband
of VLSI
In
particular.
efficient.
circuits,
AWE
l -t7)
was
incorporated
the
air rac!ive
into
lilt" finile
feat ures
elemelll
of AWE
are
mmhod.
11 wa> i_d,,'d
mainlained
wh(,xl used
,,I,-,.r\,,,l
in (,h,ct
tt_al
i("
l'Olll_tll('|
problems.
9.2
Suggestions
The
following
element
for
is a list
ORDER
ADAPTIVE
MIXED
tasks
EDGE-BASED
ELEMENTS
for further
dexelot)mcnt
of the
tinile
MODULAR
finite
magnetics)
either
technique
loses
TRUNCATIONS
AND INTEGRATION
capacity
for the
its efficiency
It is anticipated
scale
INVESTIGATIONS/APPLICATIONS
ROBUST
methods
in practice
purpose
will be desired,
FEM
OF MORE
element
any general
limited
DEVELOPMENT
(WITH
USER INTERFACE)
Development
lems encountered
problems.
LOSS)
DEVELOPMENT
Modular
Hybrid
FEM
AND INTERFACE
(WITH
INCORPORATION
in large
of suggested
ELEMENTS
ANISOTROPY
the
Tasks
methods
HIGHER
9.3
Future
that
particularly
computations
and
way to reach
or becomes
efficiency
in engineering
stage
more
and
available
when
of the
and
speed
design.
electromagnetic
its maturity.
incapable
current
resources,
when
of various
as the commercially
at the
of computing
analysis
FEM
more
become
prob-
As well known,
software
in electro-
simulating
intricate
development
specialized
a key
with
techniques
consideration
1t(;
With
a whole
then
consider
this
FEM
tions.
robust
._'I of ._peciMly
to creale
modular
in the
an imegrat
environment
A well designed
developed
and
ion enviro,mwm.
for
modular
l_'dlI,iqm'>
finite
fulure
A_ .-hown
coml)utalional
element
nwt la,&,l_,_i_'>.
met hods
_,***'>l,_,,lld
in tig. _. t. wc pi,_l,_,._.
_'l('c'1 romagm'1
will be the
inosl
ic al_l_lica-
cal)al,h'
future!
FEM Multi-Module
Environment
User Interface
_.
Mesh
User's Modules
bricks
prisms
FEM Modules
tetras
/ Truncation
I
Modules
Modules
FeedLine )
[
_'_ost
Figure
9.1:
Processor_'_
Multi-modular
FEM
environment
alld
APPENDICES
147
ll"
APPENDIX
Evaluation
Referring
of Matrix
to Fig.
A.1
and
the
Elements
associated
table,
for
the
Tetrahedrals
fields
in the
_th
letral_edron
nodes/vertices
(a)
Table
Edge
Numbering
Vertex
Numbering
il
i2
1
1
1
2
4
3
2
3
4
3
2
4
(b)
Figure
A.I:
(a) A tetrahedron.
node/edge
numbering
scheme
are expanded
as
Cw;
E = _
where
the
basis
functions
+gr-,
>:r
by
rCl_
W_-_,(r)
I
fT-
outside
element
bT-i
r, 1. r, 2 position
61_ r,_ r, 2
vectors
of vertices
il and
i, (s<'c "lal,h,)
b,br_,
g7_2
---
--
ei
(r,_ -
ri, )
b,
]r, 2 - ri_] = length
I/_
We note
element's
of the
edge
(see Table)
volume
that
V.W_
indicating
ith
that
W_
=0
V" x W_
are divergenceless.
= 2gi
Furthermore,
(
)
w_(1J) _, = _/
where
that
the
r j has
the
coefficients
jth
edge
of the
E/_ = E _i represent
i=j
ij
tetrahedron.
the
average
This
field
last
value
property
at the
ensures
ith
edge
tetrahedron.
Using
elements
the
Ai_.
above
basis
functions,
we now proceed
with the
derivation
We have
/f/
Z(v
x W_)
(V x W_)=
4
--g,
g iI_
of the
matrix
of
.-\ ]so.
w:-w;
"p
a,.
r_-tg
=
(_(I_
r!}
di
+ 12 + 1::)
where
D=(f,
gj)+(fj
x g,)
and
Since
f is a constant
Ii
ffj
f,.f,d,'
12
j j/
r. Ddc
/3
j//
(g,
vector,
11 reduces
r)-(gj
to
11 = f,-fj
To evaluate
where
xi,
vertices
I,';
12 we first, set,
9i, zi (i =
and
Li are
That
is, Li is the
other
than
standard
r)dt'
the
formula
i=1
i=1
i=1
1,...
the
and
denote
simplex
normalized
ith,
,4)
the
for volume
the
(z,y,z)
coordinates
volume
point
integration
of the
functions
for the
or shape
of the
(z,y,z)
coordinates
tetrahedron
located
within
within
formed
the
a tetrahedral
Dx
xi + D v
i=1
by its three
element
y, + D:
i=1
z,
i=1
same
tetrahedron.
we have
12 = T
tetrahedron's
and
element.
corners
Using
the
simplifying,
w}lr,re /].,
use of basic
13
vector
g, ' g_ f
of D.
identities.
]rlZ dc - j
\%
1,_ it_,,
t_ave
_:,,,.
'<
where
can
gim represents
be easih
. ,
evaluated
analytically
of the vector
and
the
restllt
g,.
can
.
Each
of l}le al>o\e
be expressed
inl>gial>
in lhe
general
form
ala,_ dv = 20
"e
for l,m
variables
1,...,3.
x, y, z.
The
al, a_i +
i=1
parameters
at or a._ can
Oh
I=l
ami
/=1
represent
any
of the
rectilinear
15'_'
APPENDIX
Evaluation
The
explicit
by (3.12)
of the
representation
and
can
Bpqi'=2/fr
where
G0(r,r')
triangle
pair
it should
over
/fr
Integral
boundary
integral
System
subsystem
matrix
is given
as
(-k_Si.S,+VxS,.VSj)Go(r.r')dF,'dS
is tile free
5'p as shown
space
Green's
in fig. 3.3.
be recognized
the
Boundary
Matrix
of the
be rewritten
that
the
function,
Similar
definition
(B.1)
and
to the finite
T,p is the
element
of (B.1)
virtually
the
region
pth
triangle
assembly
involves
of the
procedure.
an assembling
triangles.
To proceed,
(3.14)
is used
to discritize
field
and
thus
its curl
is given
by
x Si(r)
where
located
Given
e(r)
is defined
inside
the
by' (3.15).
the pth
Green's
triangle
function,
'2_r' p
.4 i .4j
that
in a planar
when
surface
deriving
JA,
(B.2),
is considered
it is straightforward
_:gUj //
bTrAp.4q
Note
(B.2)
= e(r)A/--_'_p5
to express
//(r-r,').(_-r;)_,(r)_:(
and therefore
the
matrix
r) _
_i(r)%(r)
q'
fact
that
r is
V.r
= 2.
entries
as
_-jkoR
the
_'
dS' dS
(B.3)
17,3
in which
terms
Tf
B' =
ir -
r'i.
bv numerical
integrations.
. the integrands
the singularity
subtracting
/_T,
Tl,_,se integral>
become
should
/_T
singular
an additional
- ) ko R
; --_,(r)%(r)
one
first
integral
term.
dS' dS'
in (B.4)
is carried
out
That
If.
ff:
f:
[55].
-elf
ct,ll
witIi
Iii thi-
ca_c.
lhi.- is accoi_lpli>[,,d
lw
is
e-jk_n-l_'(r)_J(r)dS''d>',,
/IT _
H
)
is evaluated
analytically
R,v ta,Ii
inlegral,
The
_'valtlat_'_l
be removed.
and adding
-_,(r)_(r)dS"dS'
using
Similarly,
numerical
the first
(B.-t)
integrations
integral
and
in (B.3)
the
second
is rearranged
as
e-jkoR
(r-
r'i).
(r -
rj)G(r)o(r)_
dS'd_
/fT_fLq
in which
singularity
form
[55].
the
first
removed
(r -- r'i).
integral
and
dS' dS;
(r-r'i)'(r-r'J)ei(r)ej(r)e-jkR-lR
on
the
the
second
(r - rj)G(r)ej(r)_
right
hand
one again
side
may
dS' d_q
is numerically
be expressed
(B.5)
integratable
in a simple
with
analytical
15 1
APPENDIX
Formulation
For
FEM
for
implementation,
P_
where
the
curls
are given
Right
Angle
the
following
quantities
_,,VV_.VV_dI"
Prisms
are
required
by
li
v v,
vv,
-2s_
v x vk
i = 1,_.a
li
[(.-xj)_+(y_yj)9+_(:_+A:_:)]
j=4.5.6
(c.3)
To
this
represent
end,
we follow
the
k, k'=7,8,9
be analytically
top
stand
the
triangle
for the
evaluated
notation
edges,
vertical
and
defined
in (4.13)
j,j'=4,5,6
three
denote
edges.
we tabulate
k = 7, s, 9
the
Pii'
Cii,
Dii,2__z
P#,
Cjj,
[ Djj,A:+
and
the
It is found
results
(4.14),
bottom
that
where
triangle
(C.2)
and
i.i'=1.2.3
edges
and
(('.3)
can
as follows
4-_(_:)_ 1
q- 3 _
a4"_(A:)3]
_
(('.4)
(C5)
17"17)
..__2
tL._:, -
? . t_'
45'
t;:
Pzk
P:, =-C,,
l) ....
Pkz--
Pkj--
l;
_11
('._"
(-_. (_x:_ :_
z, [._-.?(_.v
.r,:V_
? >'_
'_"
4(_g"):
)+._" (
-'"
)]
'
- u;>_ )]
('.9
(('.10
(C.11
(___:)3
Qii,
QJJ'
--
C,i,Dii,
(C.1_')
('j:,Djj,
(c.la)
(A:) 3
3
(c.]4)
AzS_Tkk,
Q,j
QJi -- (_)3
Qik
Qk,=Qjk=Qkj=O
Qkk'
(C.15)
Ci.i Dij
(C.16)
(C.17)
where
The
Tkk,
Cii
D_j
remaining
1/6
fork=k';
1/12
for
kT/:k'
lilj
(c.]s)
4(S_Az) 2
_q.X
quantities
(x, + xj)SX
in the
+ x,xj,_
above
S_
SX
+ S}"}"
list of the
(yi + yj)5}
expressions
fs_
dxdy
x dxdy
are
+ yiyj5
defined
as
l')t,
.,4)
?".\
These integrals
('\",
a triangle
are
_,,
.5'.\')
_,.rgd.rd!l
in terms
_ )- Specifically.
in counterclockwise
of the global
assuming
rotation,
dxdy
= y
(X_ + Xj
a',
y,
:ra
Ya
.r,_
y_
, z dxdv
SYY
so,_xdv= _
""q_{(_
jfs
xydzdY=_2{(X,
the
three
of the three
llOd('s
],j
+ .\'., )
+ x, + xm )2 + (.v_+ _v,
_+ .,:.,)}
.2
+ __+ _:,)2+ (_?+ V- + _,,,)
-2 }
+ xj + xm)(X
+ _; + Y_)
llodcs
all(l in of
have.
coordinates
thai
we then
= 1
SXX
SXY"
,r 2 d.rd!/
SY
d,r d.q
S_=
SX
{,..q
.5'})
can be expressed
.\"
") 7
APPENDIX
System
Referring
5r(H)
to derive
sition
the
1
= 2
d+a1+a,
system
in terms
H = H _cat
5r(HSCat'HinC)
where
the
the fact
integral
assumed,
Derivation
that
of the
it then
in free
H inc,
with
the
of the
the
(HSC_t'H_t)
(H_t,H_nc)
Hinc does
same
form
invoking
the
dft
field.
On inserting
ad+aj+n_--'--4-tN_cat'
a, +(H'n_'H_at)
not
the
(D.1)
field decompo-
becomes
exist
in f_
as in (D.1).
has
H,,,c)
Hscat)
f_a
a,
been
Once
nd+( Hinc.
(D.2)
considered,
and
a self-adjoint
(,)
system
represents
operator
is
that
HS_t)
a, = (H'_t'
Hin_) nd
(D.3)
Hi'_c)lnl
(D.4)
Q f,
A Functional
functional
scattered
functional
follows
space
From
VxH.G-VxH-k2H'_,-H
(Hi'_'
Also,
divergence
theory,
(H_at,
we have
H '_td
--
H
l
scat"
( :-,
5_ s
"VH
52
" V
( y
i'_
Hinc
dS
dfi'
(D.5)
I2vid('ntlv.
via
for a _(,If-aljoinT
(;alerkiE)'s
n,etl,od.
operator.
II shoul<l
c(,)iditions,
the
In the
of a non-se]f-adjoint
case
system
given
terms
of the
?(H)
self-adjoinl
by (:2.42)
scattered
_>))(.v('a(lilx
l,e )a,ted
property
field
same
_l,' >\-_('ll_,
l,'>id(,-
)h('
Ol)(.rator
it is generally
manner.
is of the
This
l,_,tlmlar\
>illll,]y
not
2.12,
,)))tai),',l
_rall>i)i,)))
r('(luirv>
I)OSsil,]( , to r('('ox('r
is })ecause
the
functional
ilk('
(I).2)
in
as that
in
form
_+n:+no
-_
V x
)lla)
of a .-\slt,m
operator,
in the
r(',_,v('r_
.f,l-V
V
x
_)..
-"o*"
'Pd
'H"c
df_
_1 Jn
( Vx
Hin_ _:a
.Vx
H_, t - "o""
b21_ii,,c._
" HSCat)
dQ
H "_'_t"
_1
_::
V x H _''c
dS
It is observed
(2.42).
the
than
that
All other
two integrals
(2.42).
in (D.7)
over domain
to the
system
leading
system
excitation.
to a different
Apparently,
FEM
system
BIBLIOGRAPHY
159
BIBLIOGRAPHY
[1] E.H.
Newman
methods".
and
IEEE
P. Tulyathan.
"'Analysis
Trans..4ntennat_
and
of microstrip
Propagat..
[2] J.R.
[3}M.C.
Bailey
antennas
1985.
[4] C.
and
M.D.
using
Tsai,
Deshpande.
moment
H. Massoudi,
culating
fields inside
linear basis functions
ory
Tech.,
IEEE
Durney,
arbitrarily
with the
"Analysis
methods".
C.H.
and
and
and
shaped
moment
1131-1139,
anlennas
Jan.
numerical
scatterers".
of elliptical
Trans.
M.F.
Goto.
"The
April 1992.
[r] Christopher
R. Johnson.
Computational
[8] D.
Colton
impact
Science
and
of mobile
P. Monk.
"A new
Jin
radiation
Trans.
[10] R.F.
[11]
and
J.L.
Volakis.
by microstrip
Antennas
Harrington.
Co.,
1961.
C.T.
New
Tai.
York,
Dyadic
1994.
Propagat.,
Green
finite
antennas
and
Time-harmonic
Functions
"'A procedure
dielectric
bodies
Trans. Microwave
Trends
and
Sept.
for calusing
The-
leukemia:
Science
element
1598-1604.
in medicine".
Mag-
IEEE
1995.
to detecting
arrays
challenges".
IEEE-AP
problems
Computational
"A hybrid
patch
Winter,
approach
IEEE
microstrip
Propagat..
communications".
inverse
Engineering,
putational
electromagnetics".
Winter,
1995.
[9] J.M.
radio
"Computational
and
circular
1986.
Rappaport.
"Wireless
personal
communications:
IEEE-AP
Magazine,
Vol. 33, No. 5, October
1991.
azine,
and
Iskander.
[5] T.S.
[6] Naohisa
monwnt
techniques
in green's
IEE" Pro('.. PI. II.
Antennas
inhomogeneous
method".
IEEE
Nov
using
1.981.
and
method
residing
Nov.
Electromagnetic
in Electromagnetic
fields.
Using
con>
Engineering,
for scattering
in a cavity".
and
IEEE
1991.
McGraw-Hill
Theory.
IEEE
Book
Press.
161
P. 5ilvester.
[la]
"'Finite'
Alla
Fr_qu_nza.
P.P.
Silvester
3hthods
[14j A.
and
and
G.
Pelosi.
"'\eclor
variational
IEEE
magnetic
problems".
886. 1980.
and
combination
and
J.B.
dielectric
Vol.
lossv
mittivity
No.
9.
vector
eh'clro87S
of problems
hav-
Theo W T_ch..
IEEE
variational
Trans.
Vol. 26,
formulation
Microu,ave
analysis
of h-plane
IEEE
1343-1351,
for
Theorg
Tech.,
waveguide
Trans.
junctions
Microwave
by
Theory
1988.
M. Koshiba.
IEEE
"Full
Trans.
analysis
Trans.
and
J.Helszajn.
vector
finite
Microwave
element
Theow
of chirowaveguides
Microwave
"Finite
gyromagnetic
Theory
and
Juan
Zapata.
formulism
Tech.,
for
\:ol.
37,
using
Tech.,
Vol.
the
38,
finiteNo.
10.
10, October,
with
Microwave
Theory
"Efficient
tensors".
IEEE
of longitudinally
Trans.
Microwave
magTheory
finite
element
materials
IEEE
analvsis
characterized
Trans.
of waveguides
by arbitrary
Microwave
Theow
perTech.,
1995.
Zapata.
lossy
solution
1989.
anisotropic
permeability
element
waveguides".
999-100,5,
inhomogeneous
Microu'ace
elements".
IEEE
and
Trans.
formulation
adjoim
1986.
"Propagation
elliptical
21.
1990.
Gibson
Tech.,
Vol.
ill
1989.
method".
netized
t,c.,
ticld,-
for nonself
boundary
and
Svedin.
1488-1496,
variational
"Self-adjoint
waveguide".
[21] A.A.P.
T_ch..
Theo W Tech..
"Numerical
and
K. Miura,
loss3' anisotropic
pp.
princil)le
waveguide".
129-134,
of finite
[19] K. Hayata,
element
l'.'h ctl'omagn,
clectrolllagIt_'tic
Th_or.q
Microu'av
IEEE
Davis.
A.M.
liar_
of
.llicrou'ac_
"Self-adjoint
[2o] Jan
"'.
1978.
Tech.,
j,,
formnlalioxl
Trans.
operators".
91-94.
anisotropic
Vol.
l_rol,h.tl>
199-1.
"'The variational
A. Wexler.
Eh m_ ,t.,
Press.
Trans.
IEEE
ing nonself-adjoint
No. 1, pp.
wa\c_uich,
1976.
[16] G. Jeng
of honam_'m._m_
1,%9.
Finite
lEEK
media".
pp. 553-559.
solulion
7"_chnique,,.
Konrad.
anisotropic
dement
"An
inhomogeneous
Tech.,
efficient
finite
element
bi-anisotropic
formulation
materials".
1996.
to analyze
IEEE
Trans.
1(;2
[:2-1]
[26]
Inc.
D. Zheng
shape
[o7]
and
with
X. Yuan,
D.R.
Trans.
[32]
integral
Barton
[33]
J.M.
Wilton,
D.R.
T.K.
J.
D.W.
[36] H.
Gong,
"On
and
radiating
Morishita,
F.
Croq
coupled
\'ol.
and
and
,k
the
D.
stacked
M. Pozar.
microstrip
Dec.
finite
elenmnt-
of arbitrary
1994.
for three-dimensional
3919-21.
of cavity
Theory
"Electromagnetic
Antennas
element
Electromagnetics,
vol.
biconjugate
Apt
resonances
Techn..
Nov.
scattering
Propagat.,
"A finite
of the
objects".
antennas
Microwave
Glisson.
alld
1990.
"Computation
Trans.
Trans.
"Numerical
and
one
inhomogeneous
elements
vol.
element
May
boundary
integral
gradient
by
1982.
63-85.
method".
J.
1987.
line antennas".
with
finite
Volakis.
A.W.
IEEE
application
P.E. Mayes.
1991.
"'A hybrid
Sept.
Phys..
IEEE
analysis".
Appl.,
of arl,i_rarv
of finite
of cavity-backed
A. Alexanian.
K. Hiraswa,
[37]
and
RCS
Waves
Smith
polarized
J.L.
elements".
shape".
for antenna
Sarkar.
J. Appl.
and
from
H. G. Wang.
vector
Rao,
of arbitrary
and
"New
S.M.
Volakis,
Woo,
Cendes.
Jin,
"'('oupling
scattering
Propagat.,
finite
Appl..
and
edge-based
Electromagn.
[35]
Z.J.
f('d antenna.-
l|av_.,
Strohbehn.
Antenna
using
1992.
J.L.
kVih'\
1._)90.
computation".
A. Chatterjee,
method
1994.
[34]
and
field
J.W.
A.C.
method
Trans
of coaxial
10.53-10.58.
Propagat.,
\:olakis,
IEEE
surfaces
.h,llll
electromagnetic
scattering
frolll inhornogelleous
and finite element
method".
IE'E'E 7)'ans. .-I,-
for electromagnetic
J.L.
M.L.
ehctromag,,t,c,".
J. Ehctromagn.
and
,4ntnnas
boundary
magnetic
1987.
[31]
Lynch.
J. Gong,
shape".
me thodt,
"'Analysis
methods
IEEE
[30]
Michalski.
thicksubstrates".
Propagat..
moment
[29]
N.A.
X. Yuan.
"'Three
dimensional
objects
bv the hybrid
moment
tennas
[28]
_hm:,t
1993.
and
experimental
Universit9
of Illinois,
K. Fujimoto.
point
shorted".
"Analysis
IEEE
"Millimeter-wave
antennas".
1991.
analysis
of a cavity-backed
Trans.
design
IEEE
Tra_s.
of circularly
1990.
Antennas
of wide-band
Antennas
an-
Propagal..
aperturePropagal..
1t;3
[:1_: I.l'.
Budka.
"microwave
circui_ eh'etric
tatiol_. ]iadlatioll
Loborator!/.
lnlr_r..it!/
[::19] .I..M..tin
planar
[40] A.F.
[41]
and
S.L.Ray.
tior_s of T]_
Cor_jugat(
Scatlering".
Elsevier
Yousef
Saad.
differential
[45] C.F.
for
T.F.
Chan.
Sparsf
,-
for >calIcriIl,.:
"'_\um_ricaI
tJ/( ('G-t"FY
Inc..
t,x
f,)r
]mj,/,ml_t_tl:'l, rlrom,:_,,
t,c
1991.
Linear
J. Demmel-V.
of Linear
Systems:
"'An iteration
method
and
integral
Sgstfm._".
Eijkhout,
Blocl,'s for
operators",
Iteratiee
t)\\']i
])ublishillg
for the
solution
d. Res.
:Val.
R. Pozo,
M_thods".
of the
Bur.
Smith,
"Solution
of systems
of linear equations
Bur. Stand.,
Vol. 49, pp. aa-sa,
1952.
A.F.
Peterson,
electromagnetic
and
R. Mittra.
scattering".
a,d
('o.
I_,--,
19.q:2.
l{..Nlitlra.
Method
/'/,.1_
"']'mp/at_s
51.-1.11. 1994.
eigenvalue
Stared.,
\'ol.
of linear
1950.
C. Lanczos.
d. Res. Nat.
for
and
Publishing
.llethod,_
tit solutioI_
gradicllt
1:2. pl ). 1()5-11!1.
('ban.
(;radie_t
Science
M. Berry.
[43] C. Lanczos.
[44]
('.I-l.
"'lteraliet
the Solulio_
282,
vol.
>v>_en>".
199").
1996.
R. Barrett.
for
"'Biconjugale
E[clromag1_tlc...
Peterson.
Company,
[42]
.I.L. \olakis.
plates".
lU'hl inla_il_e
of .ll_cl/ql,ll.
IEEE
"The
Trans.
by minimized
biconjugate
Antennas
iterations".
gradient
method
Propagat.,
Vol.
:}8,
1990.
"'Numerical
Press. 2nd.,
1992.
[47]
V. Rokhlin,
S. Wandzura,
and
the wave equation:
a pedestrian
Magn.,
[48]
Vol.
June
1993.
integral
structures".
IEEE
1991.
Antennas
Webb.
Magnetics,
[51] A. Bossavit
Trans.
[52]
35, pp.7-12,
R. Coifman.
prescription".
Propagat.
"Mixed
"Edge
finite
1989.
O.C. Zienkiewicz.
edition.
1979.
The
and
1460-1465,
I.Mayergoyz.
Magnetics,
vol.
elements
elements
Magazine,
finite
in R 3''.
what
Numer.
they
can
Alath.,
do
for
Jun
you".
IEEE
Tra_s.
1993.
"Edge-elements
Element
Method.
for scattering
McGraw
problems".
Hill.
New
IEEE
York.
3rd
[531 :\.
Bossavit
current
and .I.('.
prot,len>",
kerite.
elements
for microwave
.527. Nov. 1995.
\Vilton.
gra]s
S.M.
for uniform
domains".
[sG] E.L.
dipoles".
[.ss] J.M.
Trans.
and
[6o] L.W.
J.L.
Science
"'Scattering
Antennas
Propagat..
Volakis.
slot
Personal
communication.
Lab.
Ohio
State
ble feeding
Dec. 1995.
structure
Damiano
probe-fed
[63] J. Zheng
[64] W.C.
and
feed
and
antennas".
and
Chang.
D.C.
paths
Chew,
Z. Nie, Q.H.
antennas",
Aberle,
radar
lEE
D.M.
cross
model".
and
and
Soc.
IEEE
L.N.
Amer..
arrays
I._I.
of
crossed
frequency
selec-
scattering
by
and
transmission
plane".
IEEE
7)'ans.
Research
Corp..
1995.
of arbitrary
shaped
slol and/or
Ph.D.
dissertation.
Electro-
and
IEEE
Trans
of analytical
Proc.
Trans.
and
Y.T.
accurate
Antenna
and
H., Feb.,
Antennas
of the coax
ca-
and Propagat..
numerical
models
for
of a coaxial
probe
for
1994.
network
Lo.
model
Propagat.,
"Analysis
January
of a probe-fed
1991.
microstrip
H, 1991.
and
C.R.
of probe-fed
Trans.
A. Majda.
of waves".
[67] L.Halpern
Acoust.
section
IEEE
[66] B. Engquist
simulation
Pozar,
iillt'-
and poly]wdra]
for analyzing
"End-Correction
Proc.
"'Potelltial
1983.
IEE
Liu,
1 t_t,. .51!1
1991.
"Survey
antennas".
periodic
conducting
"An efficient
A. Papiernik.
37..X,,.
1979.
microstrip
microstrip
[65] J.T.
from
Mission
University,
\ol.
in a thick
1!_":2.
on polygonal
"Electromagnetic
disk
Propagal..
Schaubert.
distributions
Henderson.
"'The scattering
of planar array
elements
in stratified
dielectric
medium".
wire
J.P.
f:.U('.
and D.II.
Munk.
dimensional
Propagat.,
[59] J. Berrie.
don_ain
a three
Antennas
source
llwlho_l
1>. pl _. 131-5.._lal
lime
7,o,.,.
Glisson.
Antenna.,,
B.A.
IEEE
Jin
through
Trans.
IEH:
A.W.
vul.
eh'ment
cavities".
Rao.
and
[.57] R. Mittra.
('.H.
tive surfaces
-a
"'A finile
and linear
IEEE
Pelton
]"l:._l-l'lll_.'xl
[.5.5]D.R.
"'A mied
Birtcher.
microstrip
Antennas
Comp.,
Trefethen.
\"ol. 84, pp.
patch
Propagat.,
"Absorbing
Math.
"Evaluation
Vol. 31,pp.
"Wide-angle
1397-1404.
elements
Dec.
boundary
of input
using
an accurate
1991.
conditions
fox" the
629-651,
1977.
one-way
wave
1988.
impedance
numerical
equations",
d.
1(;-_
.J.P. \Vebb
and
nite element
Letter...
t{.L.
\'..\.
Kat,._'lk, t,<mh)_.
solutioni
of llic \_'ctor
"'Absorbiilg
stratified
media".
[70] J. Fang.
""AB('s
circuits".
applied
IEEE
Tral_s..IITT.
packages
Theow
and
absorber
using
and
AppI.,
waves".
ified
Microwave
Chew
maxwell's
D.M.
"finite
microwave
pp.
[79] S. Kumashiro,
for transient
Design
A. Taflove.
Wave
Weedon.
with
R.
vol.
difference
time-domain
International
Engineering,
R. Rohrer,
analysis
Aug.
and
of 3-d
of Integrated
and
pp. 268-270,
extension
to three
for fd-td
August.
matched
J.F,
meshes".
1994.
medium
fi'om
Microu,ave
Lee.
"A
boundary
mod-
al_d Optical
perfectly
condition".
matched
IEEE
2}'am_.
1995.
simulation
Journal
I,as('(1
llac_s
1994.
and
condition
of electromagnetic
condition
coordinates".
Lee,
.5cir _ct.
1994.
to Radio
absorption
"Validation
stretched
an artiticial
J. Electroma.qr_.
boundary
Letters,
and
boundary
185-200,
pml absorbing
Kingsland
circuits".
Computer-Aided
of electromagnetic
of Microwave
fields
and
and Millimeter-lI'av_
1992.
A. Strojwas.
"Asymptotic
interconnect
structures".
Circuits
S!lstems,
and
waveform
IEEE
vo].
Trans.
evaluation
12, No.
Computer-
7. pp
988
996.
1993.
vol. 114,
pp.599-604,
Propagat.,
ITs] I. Wolf.
for the
an_l el,w-
Tram,..ll,,'ou'acr
of an AB('
absorbing
laver
1!)!_ 1.
1994.
computation".
matched
wave> iTi
slruclurcs
submitted
199:2.
Aug.
IEEE
study
Dec.,
and
equations
absorber
Antennas
"An
scattering
Guided
W.H.
Letters,
Sacks,
Aided
and
and
anisotropic
3ul,
Thiele,
of nlmic
.March,
meshes",
al_c1 elastic
150(i-151:1.
analysis
1621-1634,
of the berenger
Technolo99
[77] Z.S.
for em
Ph_lsics,
E.T.
dimensions
[76] W.C.
L. Bahrmasel.
7,,i_/.
in microwaveintcgralc(l
conditions".
"'A comparative
"'A perfectly
J. Comp.
element
element
C)/,t
11,',. 1992.
piol)agalion
boundary
finite
wave
"'Finite
\'olakis.
absorber
Vol.
IEEE
J.L.
and
on anechoic
for truncating
[73] C. Rappaport
\o1.
for acouslic
absorbing
T(chT_..
[72] T. Ozdemir
Pt;q.,..
a_d
f._,l fi-
I!)_.q.
condilions
to model
Oc1
boulhdarv
J. ('Oral,.
l,,.,ut_.dar", c,.,it,lili,_,r_-
wa\cequalit,lL..ll,r,,u'ac_
tligdon.
"'.\l,.<,rl_,ir_e
dc equations".
Universitg,
1996.
"Asymptotic
Technical
waveform
Report.
evaluation
Computer
Sgstems
for circuils
Laborator
with
re-
9. ,5"tal_-
l(,ti
L. Pillage
ca/ff_
and
port.
f/. Rohrer.
"'.-\\VE:
Asy,nl)1_,tic
('_ n te r for
\Vavrfor,_
li-tiIl_ati,,:',
li,t,,;,-
and
L. Volakis.
methods
1994.
M. Nakhla.
.4,alg._is.
A Chatterjee.
and
for electromagnetics:
J.
Kluwer
Gong.
"'A class
Acad.
]'ul,s..
of hybrid
Ehctroma9_.
,,!
1!_ 1.
finite
eh'nl('nt
ll'are...-tlq,l_catio,.,.
[85]
E. Chiprout
1981.
[s6]
J. Gong,
S. Legualt.
Y. Botros.
J. L. Volakis.
and P. Petre.
"Application
and
design guidelines
of the pml absorber
for finite element
simulations
of microwave
M. Nakhla.
IEEE/MTT
Pade
Sgmposiurn
.4ppro.rimanls.
Di9ensl.
1996.
and
.U,,,,,,
Joseph
Trans.
and
decompositions
a,d
t"ra/,at,,,;
Norwell.
a review"..].
fraction
195S.
ll,r@,rm
[84]
packages".
Lehner.
"Partial
Amer.
Math. Soc..
Asymptotic
expansions
Addison-\Veshw
of zeiU.
Pub.
('_,..