Professional Documents
Culture Documents
Microwaves 2011 2012 Locked
Microwaves 2011 2012 Locked
MICROWAVES #46216
LECTURE NOTES
based upon lectures delivered by
Prof. L. Schchter
W
Wiinntteerr 22001111--22001122
This work is subject to copyright. All rights reserved, whether the whole part or part of the material is concerned,
specifically the rights of translation, reprinting, reuse of illustrations, recitation, broadcasting, reproduction on
microfilm, or in any other way, and storage in data banks and electronic storage. Duplication of this publication or
parts thereof is permitted only with the explicit written permission of the authors. Violations are liable for
prosecution.
C ONTENTS
1 Transmission Lines
1.1 Simple Model
1.2 Coaxial Transmission Line
1.3 Low Loss System
1.4 Generalization of the Transmission Line Equations
1.5 Non-Homogeneous Transmission Line
1.6 Coupled Transmission Lines
1.7 Microstrip
1.8 Stripline
1.9 Resonator Based on Transmission Line
1.9.1
Short Recapitulation
1.9.2
Short-Circuited Line
1.9.3
Open-Circuited Line
1.10 Pulse Propagation
1.10.1 Semi-Infinite Structure
1.10.2 Propagation and Reflection
1
2
7
11
13
21
24
26
38
46
46
50
52
53
53
59
1.11 Appendix
1.11.1 Solution to Exercise 1.10
1.11.2 Solution to Exercise 1.11
1.11.3 Solution to Exercise 1.15
1.11.4 Solution to Exercise 1.17
64
64
65
66
68
2 Waveguides Fundamentals
2.1
General Formulation
2.2
Transverse Magnetic (TM) Mode [Hz = 0]
2.3
Transverse Electric (TE) Mode [Ez = 0]
2.4
Power Considerations
2.4.1
Power Flow
2.4.2
Ohm Loss
2.4.3
Dielectric Loss
2.5
Mode Comparison
2.6
Cylindrical Waveguide
2.6.1
Transverse Magnetic (TM) Mode [Hz = 0]
2.6.2
Transverse Electric (TE) Mode [Ez = 0]
2.6.3
Power Considerations
71
71
79
82
85
85
88
90
93
95
95
97
100
ii
2.6.4
Ohm Loss
Pulse Propagation
Waveguide Modes in Coaxial Line
104
107
113
114
115
125
130
130
139
146
146
153
157
164
165
166
176
181
2.7
2.8
iii
3.9
185
185
187
191
193
194
197
197
204
207
207
4 Matrix Formulations
4.1
Impedance Matrix
4.2
Scattering Matrix
4.3
Transmission Matrix
4.4
Wave-Amplitude Transmission Matrix
4.5
Loss Matrix
4.6
Directional Coupler
211
211
215
220
221
223
224
3.10
3.11
iv
4.7
Coupling Concepts
4.7.1 Double Strip Line: Parameters of the Line
4.7.2 Equations of the System
4.7.3 Simple Coupling Process
225
225
230
233
5 Nonlinear Components
5.1
Transmission Line with Resistive Nonlinear Load
237
237
6 Periodic Structures
6.1
The Floquet Theorem
6.2
Closed Periodic Structure
6.2.1 Dispersion Relation
6.2.2 Modes in the Groove
6.2.3 Spatial Harmonics Coupling
6.3
Open Periodic Structure
6.3.1 Dispersion Relation
6.4
Transients
241
242
252
254
259
261
264
266
271
7 Generation of Radiation
7.1
Single-Particle Interaction
7.1.1 Infinite Length of Interaction
7.1.2 Finite Length of Interaction
7.1.3 Finite Length Pulse
7.1.4 Cerenkov Interaction
7.1.5 Compton Scattering: Static Fields
7.1.6 Compton Scattering: Dynamic Fields
7.1.7 Uniform Magnetic Field
7.1.8 Synchronism Condition
7.2
Radiation Sources: Brief Overview
7.2.1 The Klystron
vi
276
277
277
280
281
282
284
285
286
287
290
290
291
293
295
297
299
308
(1.1.1)
0.
y
z
y
4. The
of the vacuum:
constitutive
relations
z
y
(1.1.3)
0 0 z = j B 1y : z Ex
= j By
Ex
Ey
1z
: 0
= 0
E y = jo H x .
z
(1.1.4)
z
y
0 0 z = j D 1y : z H x = j Dy (1.1.5)
Hx 0 0
1z : 0
= 0
hence
(1.1.6)
H x = j o E y .
z
From these two equations [(1.1.4) and (1.1.6)] it can be readily seen that we obtain the
wave-equation for each one of the components:
H x = j o E y
2 2
z
(1.1.7)
2 2 Ey = 0
z
c
E y = jo H x
z
which has a solution of the form
(1.1.8)
E y = A exp j z B exp j z
c
z
y
reads
z
y
(1.1.9)
V ( z ) = E y ( z )a.
In order to
we recall that based on the boundary conditions we have
define
the current
n ( H1 H 2 ) = K where K is the surface current. Consequently, denoting by w the
height of the metallic plates, the local current is I = K z w or
(1.1.10)
I ( z ) = H x ( z )w.
Based on these two equations [(1.1.9)(1.1.10)] it is possible to write
V
I
a
E y jo H x
j
V
j
o
o w I
z
z a
z
w
(1.1.11)
V
w
j
I
j
H x j o E y
o V
o
a
z
z
z w
a
The right hand side in both lines of (1.1.11) represent the so-called transmission line
equation also known as telegraph equations.
w
d
V z j L I ( z )
dz
(1.1.12)
a
d
I z jC V ( z )
dz
C being the capacitance per unit length whereas L is the inductance per unit length;
a
w
L o and C o . As expected, these two equations lead also to the wave equation
w
a
dV
d2
2
= j LI ( z )
V ( z ) = 0
dz 2
dz
(1.1.13)
dI
= jCV ( z )
LC
dz
j z
j z
The general solution is V ( z ) = Ae
and correspondingly, the expression for
Be
the current is given by
1 dV
j z
j z
I ( z) =
Ae
=
(1.1.14)
Be
j L dz L
defining the characteristic impedance Z c1 =
LC
L
=
or Z c
, we get
L
L
C
z
y
1 j z
j z
Ae
Be
.
Zc
In the specific case under consideration
a
o
w = a , = LC = .
Zc =
o
w
w
c
o
a
I ( z) =
(1.1.15)
(1.1.16)
2R int
2R ext
1
1
(1.2.2)
Er =
= Vo
.
r ln( Rint / Rext )
r
The charge per unit surface at r = Rext is calculated based on n ( D1 D2 ) = s and it is
given by
V
1
(1.2.3)
s = 0 0
.
Rext ln Rext / Rint
Based on this result, the charge per unit length ( z ) may be expressed as
oVo 2
1
2 Rext =
.
ln Rext /Rint
R
ln ext
Rint
Consequently, the capacitance per unit length is given by
V
Q
= s 2 Rext = 0 o
Rext
z
(1.2.4)
2 o
Q/ z
=
.
C
ln( Rext /Rint )
Vo
2R int
(1.2.5)
2R ext
In a similar way, we shall calculate the inductance per unit length. Assuming that
the inner wire carries a current I , based on Ampere law the azimuthal magnetic field is
I
H ( r ) = o .
(1.2.6)
2 r
With this expression for the magnetic field, we can calculate the magnetic flux. It is given
by
R
R
I
(1.2.7)
= o z ext drH ( r ) = o z
ln ext .
Rint
2 Rint
The inductance per unit length z is
/ z o Rext
=
ln
.
I
2 Rint
(1.2.8)
Io
9
L
1 Rext
= o
ln
C
2 Rint
2R int
2R ext
(1.2.9)
c
Exercise 1.1: Determine Z c and for a coaxial line filled with a material ( r , r )?
10
z
y
d
H = j o E I ( z ) = jCV ( z ),
dz
(1.3.1)
where we assumed a line without dielectric ( r ) and Ohm ( ) loss. In the case of
dielectric loss we have
(1.3.2)
r = j
or in our case
(1.3.3)
j C j C G Y .
In a similar way based on Faraday's law
d
E = jo r H V ( z ) = j LI ( z )
(1.3.4)
dz
and the magnetic losses
r = r j r
allows us to extend the definition according to
(1.3.5)
j L j L R Z
hence the equations
11
d
I ( z ) = YV ( z )
dz
(1.3.6)
d
a
V ( z ) = ZI ( z )
dz
may be conceived as a generalization of the transmission line equations in the presence of
loss. The characteristic impedance for small loss line is
Z
L
R
G
(1.3.7)
1
Zc =
j
Y
C
2C 2 L
and the wave number, assuming a solution of the form exp z ,
= j
R
GZ o
2Z o
2
(1.3.8)
RG
G2
R2
2 2 2 2
LC 1 2
4
8 C
8 L
LC
12
z
y
z Ey
1
1x 1y 1z
x ( )
(1.4.1)
x y z = 1y
z Ex
Ex E y 0
1z
x E y y Ex
thus
1x : z E y = jo r H x
1y : z E x = jo r H y
1z : x E y y E x = 0
1 : 1z
= jo r H
z
1z : E = 0.
13
(1.4.2)
= j o r E
1 : 1
H = j o r E z z
(1.4.3)
1 : H
= 0.
z
Fromthe two
curl equations
E = 0 E = g ( z ) ( x, y )
2 = 0
(1.4.4)
2
H = 0 H = h( z ) ( x, y )
= 0,
we conclude that the transverse variations of the transverse field components are
determined by 2D Laplace equation justifying the use of DC quantities adopted above
(capacitance and inductance per unit length). From the other two equations we get the
wave equation
E
H
j
1
o r
z
z
z
z
(1.4.5)
2 E
j
1z 1z
1
o r z
= jo r [ j o r E ],
2
z
z
or explicitly
14
2 2
(1.4.6)
z 2 c 2 r r E = 0.
o r
jo r H 1z E
1z
H
o r
z
j o r E 1z H o r E
1z
o r
z
15
(1.4.7)
s1
V V0
x
As in the previous two cases we shall see next how y
V
s2
the electric parameters can be calculated in the
general case and for the sake of simplicity we shall z
assume that the medium has uniform transverse and
longitudinal properties. The electric field in the entire space is given by
j z
whereas the magnetic field is
E = ( )e
o r
j z
(1.4.8)
H =
( 1z )e
.
o r
Note that associated to this electric field, one can define the voltage
s2
s2
s2
Vo = E d = d = d
s1
s1
s1
(1.4.9)
j z
. On the two (ideal) conductors the electric field generates a
such that V ( z ) = Vo e
surface charge given
by
(1.4.10)
s = o r n E ,
therefore the charge per unit length is
Q
(1.4.11)
= dl s =
dl o r ( n E ).
z
Since by virtue of linearity of Maxwell's equations
16
y
z
s1
V V0
s2
o r
Since it was shown that H =
1z E we conclude that
o r
J s n H = n [1z E ] o r
o r
(n E )1z
o r
o r
o r
o r
dl n E
(1.4.14)
(1.4.15)
At this point rather than calculating the inductance per unit length we combine the
previous result for the charge per unit length and (1.4.15) the result being
17
V0
o r
dl n E
Io
c
o r
=
=
.
Q/ z
r r
o r dl n E
y
z
s1
V V0
s2
(1.4.16)
However, having established this relation between the current and the charge per unit
length we may use again the linearity of Maxwell's equations and express Q / z = CV0 .
Substituting in Eq. (1.4.16) we get
I0
c
(1.4.17)
=
CVo
r r
but by definition
V0
= Zc ,
(1.4.18)
I0
which finally implies that
1
c
=
.
CZ c
r r
18
V0
y
z
r r
s1
V V0
s2
(1.4.19)
.
Cc 2
It is important to re-emphasize that this relation is valid only if the electromagnetic
properties ( r , r ) are uniform over the cross-section.
L=
Exercise 1.3: Calculate the capacitance per unit length of two wires of radius R which
o r
1
1
P = Re dxdy E H 1z = Re dxdy E 1z E 1z
o r
2
2
1
= Re
2
=
*
dxdy E E
o o r r
We =
o r
2
o r o r
[We Wm ].
r r
o r
o r
19
*
1
4 dxdy o r E E
(1.4.20)
V0
Exercise 1.4: In the last expression we used the fact that We = Wm -- prove it.
1
1
Z c | I o |2 = Vo I o* .
2
2
Finally, we may define the energy velocity as the average power propagating along the
transmission line over the total average energy per unit length
P
c
=
.
Ven =
We Wm
r r
Exercise 1.6: Show that the material is not frequency dependent, this quantity equals
exactly the group velocity. What if not? Namely r ( ) .
Exercise 1.5: Show that the power can be expressed as P =
20
I
(
z
)
Z
(
z
)
dz 2
dz
dz
(1.5.2)
d
dV
ln Z ( z )
Z ( z )Y ( z )V ( z )
dz
dz
A solution of a general character is possible only using
numerical methods. However, an analytic solution is
possible if we assume an "exponential'' behavior of the
21
form
Z ( z ) j Lo exp qz
(1.5.3)
Y z jCo exp qz
Substituting these expressions in Eq. (1.5.2) we get
d 2V ( z )
dV
q
2 LoCoV = 0
2
dz
dz
z
therefore assuming a solution of the form V ( z ) = Vo e 1 we conclude that
1
1
1 = q q 2 4 2 LoCo q q 2 4 2 LoCo .
2
2
In a similar way, the equation for the current is given by
d 2 I dI d
ln[Y ( z )] YZI ( z ) = 0.
dz 2 dz dz
z
Assuming a solution of the form I = I o e 2 we obtain
1
1
2 = q q 2 4 2 LoCo q q 2 4 2 LoCo .
2
2
It is convenient to define
q2
2
c
,
4 L0C0
22
(1.5.4)
(1.5.5)
(1.5.6)
(1.5.7)
(1.5.8)
which sets a "cut-off'' in the sense that for < c both 1 and 1 are real.
The second important result is that the impedance along the transmission
line
z
V ( z ) V0e 1
qz
Z
e
=
=
(0)
c
I ( z ) I 0e 2 z
(1.5.9)
is frequency independent.
Exercise 1.7: Plot the average power along such a transmission line as well as the
average electric and magnetic energies. What is the energy velocity?
23
(1.6.2)
= L n I n .
z n =1
24
Having these two equations [(1.6.1)(1.6.2)] in mind, we may naturally extend the
telegraph equations to read
N
N
d
d
(1.6.3)
I ( z ) = j C nVn ( z ).
V ( z ) = j L n I n ( z )
dz
dz
n =1
n =1
Subsequently we discuss in more detail phenomena linked to this coupling process
however, at this point we wish to emphasize that the number of wave-numbers ( 2 )
corresponds to the number of ports. This is evident since
j z
j z
,
(1.6.4)
V ( z ) = V0 e
I ( z ) = I0 e
enabling to simplify
(1.6.3) to read
(1.6.5)
V0 = LI 0
I 0 = CV0
=
= =
(1.6.6)
wherein is the unity matrix. Clearly the normalized wave number 2 c / are
2
the eigen-values of the matrix LC ( = C L since both matrices are symmetric) and if the
= =
= =
25
1.7 Microstrip
r
dielectric layer. On the other side of the latter, there h
w
is a ground metal; the side walls have been
introduced in order to simplify the analysis and the
width w is large enough such that it does not affect the physical processes in the vicinity
of the strip. We shall examine a simplified model of this system and for this purpose we
make the following assumptions: (i) The width of the device is much larger than the
height ( w h ) and the width ( w ) of the strip.
(ii) The charge on the strip is distributed uniformly.
Our goal is to calculate the two parameters of the transmission line: capacitance
and inductance per unit length. With this purpose in mind we shall start with
evaluation of the capacitance therefore let us assume a general charge distribution on the
26
strip
r
w
x
x
(
)
|
|<
2 2
(1.7.1)
s ( x) =
w
0
| x |>
2 2
With the exception of y = h the potential is given by
nx
ny
A
sin
sinh
0 yh
w
w
n =1
( x, y ) =
n ( y h)
nx
w
y h.
Bn sin
e
w
n =1
The continuity of the potential at y = h implies
x
nh
nx
=
s
i
n
( x, y = h) = An sin n sinh
B
n
w
w n
w
n
nh
An sinh
= Bn .
w
27
(1.7.2)
(1.7.3)
y
x
r
w
w w
w
n
(1.7.4)
n nx
n
sin
exp
(
)
.
Dy ( x, y > h) o Bn
y
h
w
w
w
With these expressions, we can write the boundary conditions i.e., n ( D1 D2 ) = s in
the following form
w
x
x
(
)
|
|<
n nx
nh
2 2
(1.7.5)
o sin
Bn r An cosh
=
w
w
w
w
n
0
| x |> .
2 2
Using the orthogonality of the sin function we obtain [for this reason the two side walls
were introduced]
w
nh 1 2
2 dx ( x)sin nx .
(1.7.6)
=
Bn r An cosh
w
n
w
o
2
28
y
x
w
The next step is to substitute (1.7.3) into the last expression. The
result is
w
nh
nh 1 2
2 dx ( x )sin nx .
(1.7.7)
An sinh
cosh
=
r
w
w
n
w
o
2
Consequently, subject to the assumption that x is known, the potential is known in the
entire space and specifically at y = h is given by
w
1
1 2
nx
2 dx ( x)sin nx . (1.7.8)
( x, y = h) = sin
w 1 ctanh nh o n w
w
n
r
2
w
In principle, this is an integral equation which can be solved numerically since the
potential on the strip is constant and it equals V0 ; many source solution.
At this point we shall employ our second assumption namely that the charge is
uniform across the strip and determine an approximate solution. The first step is to
average over the strip region, | x w / 2 | / 2 . The left hand side is by definition constant
thus
29
y
x
w
w
1
Vo = 2 dx ( x, y = h)
w
2
w
(1/ o )(2/ n) 1
2 dxsin nx
2 dx ( x)sin nx .
=
w w
w
nh w
1 r ctanh
2
2
(1.7.9)
Explicitly our assumption that the charge is uniformly distributed, implies Q / z ,
therefore
2
w
1
Q 1
2
1
2 dxsin nx
Vo =
z 0 n n
w
nh w
1 r ctanh
2
30
y
x
r
w
C=
Q/ z
Vo
o /2
1/n
nh
1 r ctanh
2 n
2 n
(1.7.10)
sin
sinc
2w
With this expression we can, in principle calculate all the parameters of the
microstrip. For evaluation of the inductance per unit length we use the fact that the
dielectric material cannot have any impact on the DC inductance. Moreover, we know
that in the absence of the dielectric ( r = 1), the propagation number is / c and the
characteristic impedance satisfies
1
1
L( r = 1)
(1.7.11)
1=
.
Zc =
C ( r = 1) c
C ( r = 1)
Since the DC magnetic field is totally independent of the dielectric coefficient of the
medium (electric property), we deduce from the expression of above that
1
(1.7.12)
,
L( r = 1) = 2
c C ( r = 1)
or explicitly
31
y
x
r
w
1
h
h
2
exp
(2
1)
sinh
(2
1)
sinc
(2
1)
2 w
. (1.7.13)
=0 2 1
w
w
With the last expression and (1.7.10) we can calculate the characteristic impedance of the
microstrip
1/2
1/2
sinc 2 ( )
2 e sinh(h )sinc ( )
L
= o
Zc =
, (1.7.14)
2 1
(2
1)[1
ctanh(
)]
=0
C
h
=0
r
where h (2 1)h / w and
(2 1) . The next parameter that remains to be
2w
determined is the phase velocity. Since L and C are known, we know that = LC
implying that
sinc 2 ( )
1
2 1 1 ctanh(h )
1
=0
r
=c
.
Vph =
(1.7.15)
2
sinc ( )
LC
exp h sinh(h )
2
1
=0
L = o
32
y
x
r
w
Contrary to cases encountered so far the dielectric material fills only part of the entire
volume. As a result, only part of the electromagnetic field experiences the dielectric. It is
therefore natural to determine the effective dielectric coefficient experienced by the
field. This quantity may be defined in several ways. One possibility is to use the fact that
c
when the dielectric fills the entire space we have the phase velocity in Vph =
it
2 1
= =0
.
sinc 2 ( )
1
2 1 1 ctanh(h )
=0
c2
2 thus
Vph
eff
(1.7.16)
The following figures illustrate the dependence of the various parameters on the
geometric parameters.
33
y
x
7.0
0.39
50
eff
Vph / c
Zc ohm
60
0.38
40
30
1.0
7.2
0.40
70
6.8
6.6
6.4
2.0
3.0
mm
4.0
0.35
1.0
2.0
mm
(a)
(b)
3.0
4.0
6.2
1.0
2.0
3.0
mm
(c)
34
4.0
60
0.40
50
40
30
1.0
7.2
7.0
eff
0.41
Vph / c
Zc ohm
70
0.39
0.38
2.0
3.0
h mm
4.0
0.37
1.0
6.6
6.4
2.0
3.0
h mm
(a)
6.8
(b)
4.0
6.2
1.0
2.0
3.0
h mm
(c)
35
4.0
36
confinement.
Exercise 1.9: What fraction of the energy is confined in the dielectric and how the
37
1.8 Stripline
Being open on the top side, the microstrip has limited ability to confine the
electromagnetic field. For this reason we shall examine now the stripline which has a
metallic surface on its top. The basic configuration of a stripline is illustrated below
y
x
The model we shall utilize first replaces the central strip with a wire as illustrated below
and as in Section 1.7 our goal is to calculate the parameters of the line.
y
x
h
38
y
x
For evaluation of the capacitance per unit length it is first assumed that the charge density
is given by
Q
(1.8.1)
( x, y ) = ( x ) ( y h ),
z
and we need to solve the Poisson equation subject to trivial boundary conditions on the
two electrodes. Thus
t2 =
o r
ny
(1.8.2)
( y = 0) = 0 ( x, y ) = n ( x)sin
.
d
n
( y = d) = 0
Substituting the expression in the right hand side in the Poisson equation we have
2
d2
ny
Q
n
(1.8.3)
( x ) ( y h )
(
x
)
(
x
)
sin
=
n dx 2 n
d
d
o r z
and the orthogonality of the trigonometric function we obtain
39
d2 n
2 nh
Q
(
)
=
(
)
sin
x
x
2
n
= Qn ( x ),
dx
d
d
d
o r z
where
2Q
nh
Qn
sin
.
o r d z
d
exp
A
n
n
x>0
n ( x ) =
B exp n x
x<0
n
d
and since the potential has to be continuous at x = 0 then
An = Bn ,
integration of (1.8.4) determines the discontinuity:
d n
d
n
= Qn
[ An Bn ] = Qn .
n
dx x =0 dx x =0
d
From (1.8.6) and (1.8.7) we find
2
40
(1.8.4)
(1.8.5)
(1.8.6)
(1.8.7)
y
x
Qn
Q h
h
=
sinc n .
(1.8.8)
n o r z d
d
2
d
This result permits us to write the solution of the potential in the entire space as
nx
Q h
nh ny
d
sinc
x0
sin
e
d
d
d
n =1
(1.8.9)
( x, y ) = o r z
nx
Q
h
nh
ny
sinc
sin
e d
x 0.
o r z d n =1
d d
An = Bn =
At this stage we can return to the initial configuration and assume that the central
strip is a superposition of charges Qi located at xi and since the system is linear, we
apply the superposition principle thus
h/d
| x xi |
nh ny
(1.8.10)
( x, y ) =
sinc
sin
Q
exp
n
.
d d i i
o r z n
d
In the case of a continuous distribution we should replace
41
n x x
Qie d
i
n|x x|
1
dxQ( x)e d
(1.8.11)
and consequently
h/ d
nh ny 1
n
( x, y ) =
sinc
sin
(
)exp
|
|
dx
Q
x
x
x
d d
, (1.8.12)
o r z n
d
which again leads us to an integral equation; note that the surface changes density as
s ( x) = Q( x) / z . As in the microstrip case, we shall assume uniform distribution
therefore
h/ d
nh ny Qo /2
n
( x, y ) =
dx
|
x
x
|
sinc
sin
exp
d d /2
. (1.8.13)
o r z n
d
42
y
x
Vo =
1 /2
dx ( x, y = h)
/2
(1.8.14)
(h/d )Qo
1 /2
nh
n
2 nh 1 /2
exp
dx
dx
x
x
|
|
sin
o r z n =1 d
d /2 /2
d
n
n
dx
dx
exp
x
x
=
1
exp
sinhc
/2 /2
d
2 d
2 d
2d
(1.8.15)
such that
2h
(h/d )Qo
sinc 2 nh 1 exp n sinhc n .
(1.8.16)
Vo =
d
d
d
2
2
o r z
n =1
The last result enables us to write the following expression for the capacitance per unit
length
1
Qo / z
1 d
nh
(1.8.17)
= o r 2 sinc 2
C=
1 exp n sinhc( n )
Vo
2 h n
d
43
1 exp n sinhc( n ) .
CV ph
r d n
d
The two frames show the impedance dependence on the width and height of the strip
10
Z( h d ) 377
50
150
40
100
Z 1 hn 430
Z n 2 4
Z 2 hn 4
Z 3 hn 4
0.0
20
50
10
0.5
1
n
1.5
44
2
hn
Exercise 1.15: Determine the inductivity per unit length and analyse the dependence of
the various characteristics on the geometric parameters. (For solution see Appendix 11.3)
Exercise 1.16: Compare the dependence of the various characteristics of the stripline
and microstrip as a function of the geometric parameters.
Exercise 1.17: Compare micro-strip and strip-line from the perspective of sensitivity to
the dielectric coefficient. (For solution see Appendix 11.4)
Exercise 1.18: Determine the error associated with the assumption that the charge is
uniform across the strip.
Exercise 1.19: Analyze the effect of a strip of finite thickness. Remember that
throughout this calculation the strip was assumed to have a negligible thickness.
45
Resonant circuits are of great importance for oscillator circuits, tuned amplifiers,
frequency filter networks, wavemeters for measuring frequency. Electric resonant circuits
have many features in common, and it will be worthwhile to review some of these by
using a conventional lumped-parameter RLC parallel network as an example, the Figure
illustrates a typical low-frequency resonant circuit. The resistance R is usually only an
equivalent resistance that accounts for the power loss in the inductor L and capacitor C
as well as the power extracted from the resonant system by some external load coupled to
the resonant circuit. One possible definition of
I
resonance relies on the fact that at resonance the
input impedance is pure real and equal to R
Z in
V
R
L
C
implying
P 2 j (Wm We )
(1.9.1)
Z in = l
.
*
II / 2
Although this equation is valid for a one-port
circuit, resonance always occurs when Wm = We , if we define resonance to be that
condition which corresponds to a pure resistive input impedance or explicitly
0 = 1/ LC ; note that these are the lumped capacitance (C ) and inductance ( L) .
46
= 1 =
0
2Q
R
0 .707 R
90
Z in
(1.9.5)
90
47
BW
Z in
The fractional bandwidth BW between the 0.707 R points is twice this value, hence
1
(1.9.6)
Q= 0 =
.
2 BW
If the resistor R in Fig. 8 represents the loss in the resonant circuit only, the Q give by
(1.9.3) is called the unloaded Q . If the resonant circuit is coupled to an external load that
absorbs a certain amount of power, this loading effect can be represented by an additional
resistor RL in parallel with R . The total resistance is now less, and consequently the new
Q is also smaller. The Q , called the loaded Q and denoted QL , is
RRL / ( R RL )
(1.9.7)
QL =
.
0 L
The external Q , denoted Qe , is defined to be the Q that would result if the resonant
circuit were loss-free and only the loading by the external load was present. Thus
R
(1.9.8)
Qe = L
0 L
leading to
1
1 1
(1.9.9)
.
=
QL Qe Q
Another parameter of importance in connection with a resonant circuit is the decay factor
. This parameter measures the rate at which the oscillations would decay if the driving
48
source were removed. Significantly, with losses present, the energy stored in the resonant
circuit will decay at a rate proportional to the average energy present at any time (since
Pl VV * and W VV * , we have Pl W ), so that
dW
2
t
(1.9.10)
= W W = W0 exp 2
dt
where W0 is the average energy present at t = 0 . But the rate of decrease of W must
equal the power loss, so that
dW 2
= W = Pl
dt
and consequently,
P
P
1
(1.9.11)
= l = 0 l = 0.
2 0W 2Q
2W
Thus, the decay factor is proportional to the Q . In place of (1.9.10) we now have
(1.9.12)
W = W0 exp 0 t .
Q
49
l
Z in
Zc , ,
l j / 0
Zin = Z c
Z c l j
(1.9.14)
1 j l / 0
0
50
R0
Z in
1
Ll
C
R
j
l
LC
(1.9.15)
= Rl jlL.
C 2
L
2
It is of interest to compare (1.9.15) with a series R0 L0C0 circuit illustrated
above. For this circuit
Z in = R0 j L0 1 1/ 2 L0C0 .
Zin =
then
(1.9.16)
Z in R0 2 jL0 .
By comparison with (1.9.15), we see that in the vicinity of the frequency for which
l = 0 / 2 , the short-circuited line behaves as a series resonant circuit with resistance
R0 = Rl / 2 and inductance L0 = Ll / 2 . We note that Rl , Ll are the total resistance and
inductance of the line; so we might wonder why the factors 1/ 2 arise: recall that the
current on the short-circuited line is half sinusoid, and hence the effective circuit
parameters R0 , L0 are only one-half of the total line quantities.The Q of the shortcircuited line may be defined as for the circuit
L L
Q= 0 0 = 0 =
.
(1.9.17)
R0
R
2
51
L0
C0
l
Z in
Zc
R0
L0
C0
By means of an analysis similar to that used earlier, it is readily verified that an opencircuited transmission line is equivalent to a series resonant circuit in the vicinity of the
frequency for which it is an odd multiple of a quarter wavelength long. The equivalent
relations are
Z in l j / 2 / 0 Z c = Rl / 2 j Ll
(1.9.18)
2
l 0 / 4, R0 Rl / 2, L0 Ll / 2, 0 1/ L0C0
Comment: Note that formally from (1.9.13) in the lossless case i.e., Zin = jZ c tan l we
conclude that there are many (infinite) resonances
since tan( l ) vanishes for l = but also for
l = n , n = 1,2,3 corresponding to all the
"series'' resonances. In case of "parallel'' resonances
the condition tan l is satisfied for
l =
2
resonance is used since beyond that the validity of
the approximations leading to the equations are
questionable.
52
So far the discussion has focused on solution of problems in the frequency domain.
In this section we shall discuss some time-domain features. Let us assume that at the
input of a semi-infinite and lossless transmission line we know the voltage pulse
V ( z = 0, t ) = V0 (t ) . In general in the absence of reflections
jt j ( ) z
(1.10.1)
V ( z , t ) = d V ( ) e
and specifically
V ( z = 0, t ) = d V ( ) e
jt
(1.10.2)
or explicitly, the voltage spectrum V ( ) is the Fourier transform of the input voltage
1
1
jt
jt
V ( ) =
dtV
z
t
e
dtV
t
e
(1.10.3)
(
=
0,
)
=
(
)
.
0
2
2
Consequently, substituting in Eq.(1.10.1) we get
jt j ( ) z 1
jt
V ( z , t ) = d e
dt
V
z
t
e
(
=
0,
)
2
(1.10.4)
1
) j ( ) z
(
j
t
t
= dt V ( z = 0, t )
d e
53
r which leads to
z
z
r = V z = 0, t = t r (1.10.5)
V ( z , t ) = dt V ( z = 0, t ) t t
c
c
implying that the pulse shape is preserved as it propagates in the z -direction. If the
phase velocity is frequency-dependent, then different frequencies propagate at different
velocities and the shape of the pulse is not preserved. As a simple example let us assume
that the transmission line is filled with gas
2p
( ) = 1 2 .
Since
(1.10.6)
1
z
2
2
(1.10.7)
d
exp
j
(
t
t
)
p
2
c
it is evident that sufficiently far away from the input, the low frequencies ( < p ) have
no contribution and the system acts as a high-pass filter.
V ( z , t ) = dt V z = 0, t
The dispersion process may be used to determine the frequency content of a signal.
In order to envision the process let us assume that the spectrum of the signal at the input
is given by
54
(1.10.8)
V ( ) = a exp
v =1
where all the parameters are known and that / 1 ; note that at the limit that
/ 0 the Gaussian function behaves very similar to Dirac delta function. In such
a case it is convenient to expand
1
( )
(1.10.9)
( ) ( )
( )
2 ( )
=
and also to use the fact that
( )
( )
1
1
=
=
(
)
=
Vgr c
c
c 2 ( )
(1.10.10)
1 1 1 ( )
=
.
V ph c 2 ( )
With these two observations, we now aim to develop an analytic expression for the
voltage at any location z and at any time t . Specifically, to demonstrate that the peak of
the Gaussian pulses depends on t = z / Vgr (r ) and reveal the way this signal varies in
space.
M
55
jt j ( ) z
V ( z , t ) = d V ( ) e
jt j ( ) z
= a d e
exp
(1.10.11)
( ) therefore,
V ( z , t ) = a d exp
exp jt j z ( ) ( )
2
(
)
56
(1.10.12)
V ( z , t ) = a exp j t
( )
c
z
z
d
d exp j ( ) t
( )
c
c
d
2
(
)
1
z
1
d
2
j
exp ( )
.
c
d
2 ( )
z
exp j ( ) t
V
gr
c 2 ( ) d =
Vgr V ph
(1.10.13)
(1.10.14)
(1.10.15)
and define
1
1
1
j
2 ( z ) 2
2 c
c
c
z
Vgr , V ph ,
57
(1.10.16)
Consequently,
2
z
z
( ) d exp j ( ) t
V ( z , t ) = a exp j t
Vgr
z
z
= a exp j t
.
d exp j t
(
)
V
V
ph
gr
(1.10.17)
2
The integral may be evaluated analytically since d e = such that finally we
get
2
z
z
z
V ( z , t ) = a ( z ) exp j t
t
. (1.10.18)
exp
2
V
Vgr ,
=1
gr ,
According to (1.10.16) the width of the pulse varies with the distance from the input.
M
58
Other phenomena which may significantly affect the propagation of a pulse along a
transmission line are discontinuities. For a glimpse into those phenomena let us consider
three transmission lines connected in series.
Z1
Z2
Z3
z d
z 0
59
2 Z1Z 3
cos ( Z1 Z 3 ) Z 2 j sin ( Z1Z 3 Z 22 )
(1.10.19)
= 2d.
With these coefficients we aim to determine the transmitted signal in the time domain.
Assuming that the incoming voltage is
V ( in ) ( z , t ) = d V ( ) exp jt j 1 z
(1.10.20)
wherein
1
jt
.
dtV ( in ) ( z = 0, t ) e
2
Based on this expression, the transmitted signal is
V ( ) =
V ( tr ) ( z , t ) = d V ( ) exp jt j 3 ( )( z d ) ( )
60
(1.10.21)
(1.10.22)
1
V ( tr ) t , z = d
dt V ( in ) ( z = 0, t ) e
jt
z d
4 Z1Z 3 exp j exp j t
V
3
.
2
1 exp 2 j ( Z1 Z 3 ) Z 2 1 exp 2 j ( Z1Z 3 Z 2 )
We now define
4 Z1Z 3
( Z1 Z 3 ) Z 2 Z1Z 3 Z 22
,=
=
( Z1 Z 3 ) Z 2 Z1Z 3 Z 22
Z 2 ( Z1 Z 3 ) Z1Z 3 Z 22
(1.10.23)
enabling us to write
z d d
exp j t t
V
V2
1
3
( tr )
( in )
.
V ( z , t ) = dt V ( z = 0, t )
d
1 exp 2 j d
V2
61
(1.10.24)
1
= u
1 u =0
(1.10.25)
V ( tr ) ( z , t ) = dt V ( in ) ( z = 0, t )
1
2
z d d
d
exp
j
t
t
V3
V2
(1.10.26)
exp 2 j d .
V2
=0
The integration over is straight forward resulting in a Dirac delta function therefore
zd
d
( tr )
(1.10.27)
V ( z , t ) = V ( in ) z = 0, t
(2 1) .
V3
V2
=0
A delay in the occurrence of the pulse reflects in the third term. The term = 0
represents the first pulse which reaches the output end. After a full round-trip, the second
contribution occurs. Effects of round-trip time and pulse duration as well as the sign of
are revealed by the following frames.
62
1.0
2.0
Z 1 200
2 0.75
0.143
Z 2 100
3 0.9
Z 3 0.4
d 1.0
T p 2.0
0.5
1.0
0.8
V z / d , 9T p
V z / d , 9T p
1.5
0.6
0.4
0.2
0.0
-0.5
Z 1 200
2 0.75
0.286
Z 2 40
3 0.9
Z 3 100
d 1.0
T p 2.0
10
15
20
25
10
z/d
Z 1 200
3 0.85
0.143
Z 2 100
3 0.9
Z 3 0.4
d 0 .5
T p 2.0
0.5
1.0
Z 1 200
2 0.85
0.286
Z 2 40
3 0.9
Z 3 100
d 0 .5
T p 2.0
1.2
V z / d , 4T p
25
1.4
1.5
V z / d , 4T p
20
z/d
2.0
1.0
0.8
0.6
0.4
0.0
-0.5
15
0.2
0
6
z/d
10
-0.5
12
63
6
z/d
10
12
1.11 Appendix
1.11.1 Solution to Exercise 1.10
40
1
2
3
[mm]
V ph / c
40
30
6.8
eff
0.39
0.38
6.4
2
3
[mm]
0.42
7.2
0.40
6.8
0.38
0.36
h [mm]
V ph / c
104
68
7
10
7.0
0.82
0.65
0.47
0.30
1
4.6
2.2
r
64
h [mm]
1.00
2
3
[mm]
h [mm]
140
6.4
eff
Z c []
70
50
7.2
0.37
60
0.40
eff
50
30
Z c []
Microstrip:
dependence of
various
parameters on
the geometric
parameters and
the dielectric
coefficient. For
all the graphs the
paramters are
(when they are
not the variable):
= 2 mm,
w = 20 mm,
h = 2 mm,
r = 10 and
60
V ph / c
Z c []
70
10
10
= 0,1100 .
60
Z c []
52
44
0.415
n=50
36
28
0
100
200
w[mm]
300
V ph / c
20
0.408
0.400
6.75
6.70
eff
w . [ = 2 mm, r = 10 , h = 2
6.65
6.60
n=50
n=100
n=150, 200
6.55
6.50
655
70
135
w[mm]
0.393
200
0.385
36.67 68.33
w[mm]
100
Determine the inductivity per unit length and determine its dependence on the
various parameters.
The inductivity per unit length of the stripline can be determined according to Eq.
(1.7.12) L = 1/ c 2C ( r = 1) substituting the expression for the capacitance per unit length
h2
h
sinh( n )
2
d
n
d n =1
where n = n .
d
2
0.36
L[H / m]
0.34
0.32
0.30
0.28
1.0
66
1.5
2.0
h[mm]
2.5
3.0
0.8
L[H / m]
n = 1,2, 200 ].
0.6
0.4
0.2
0
[mm]
Z c []
60
40
20
0
10
d [mm]
67
15
20
n = 1,2, 200 ].
150
Z c []
100
Microstrip
50
Stripline
0
68
10
15
0
Stripline
Microstrip
dZ c
d r
-20
-40
-60
30
15
Stripline
25
20
15
69
10
35
dC
[pF / m ]
d r
Microstrip
10
15
0.8
V ph / c
0.6
Stripline
0.4
0.2
10
15
10
15
-0.2
Stripline
Microstrip
-0.4
-0.6
70
0
1 dV ph
c d r
Microstrip
71
72
0 r
0r
Rectangular waveguide; a and b are the dimensions of the rectangular cross section. a corresponding to
the x coordinate, b to the y coordinate.
The first step in our analysis is to establish the basic assumptions of our approach:
a) The electromagnetic characteristics of the medium: = 0 r and = 0 r .
b) Steady state operation of the type exp ( jt ) .
c) No sources in the pipe.
d) Propagation in the z direction -- exp ( jkz z ) ; kz can be either real or imaginary or
complex number.
e) The conductivity ( ) of the metal is assumed to be arbitrary large ( ).
73
0 r
0r
a
Subject to these assumptions Maxwell's Equations may be
written in the following
form
E = j H
H = j E
(2.1.1)
Substituting
one
equationinto the other we
obtain
the
wave equation
( E ) = j ( H )
( H ) = j ( E )
2
2
( E ) E = j ( j E )
( H ) H = j ( j H )
2
2
=
( E ) E E
v
E = 0
2 2
=
+ E 0
v
2
2
=
( H ) H H
v
H = 0
y
x
(2.1.2)
2 2
+ H = 0
v
74
0 r
0r
As a second step, it will be shown that assuming the longitudinal components of the
electromagentic field are known, the transverse components are readily established. For
this purpose we observe that Faraday's Law reads
1x 1y 1z
(2.1.4)
E = j H x y jkz = j H .
Ex
Ey
Ez
(i) 1x :
= j H x jkz E y + j H x
y Ez + jkz E y
(ii) 1y : ( x Ez + jkz Ex ) = j H y jkz Ex + j H y
(iii) 1z :
= j H z x E y y Ex
x E y y Ex
In a similar way, Ampere's law reads
1x 1y 1z
H = j E x y jkz = j E ,
Hx
Hy
Hz
75
= y Ez
=
x Ez
= j H z .
(2.1.5)
0 r
0r
or explicitly
(iv) 1x :
y H z + jkz H y
= j Ex
(v) 1y : ( x H z + jkz H x ) = j E y
(vi) 1z :
xH y yHx
= j Ez
From equations (ii) and (iv) we obtain
j Ex jk z H y
j E y + jkz H x
xH y yHx
=
yHz
= xHz
=
j Ez .
jkz 1
1
Hy
E
H
=
x z
jkz Ex + j H y =
x Ez
kz 2 ( /v) 2 kz
y z
j Ex jkz H y =
yHz
j
E
k E + y H z
=
x
2
2 z x z
kz ( /v)
It is convenient at this point to define the transverse wavenumber
k
2
2
2
kz 2
v
That as we shall shortly see, has a special physical meaning.
76
(2.1.6)
(2.1.7)
y
x
0 r
0r
y Hz.
Hy =
E
x z
2
2
k
k
In a similar way, we use equations (i) and (v) and obtain
jk
j
E y = 2 z y Ez + 2 x H z
k
k
j
jk
H x = 2 y Ez 2z x H z .
k
k
Equations (2.1.8),(2.1.10) and (2.1.9),(2.1.11) can be written in a vector form
jkz
j
E = 2 Ez + 2 1z H z
k
k
jkz
j
H = 2 H z 2 1z Ez .
k
k
77
y
x
(2.1.10)
(2.1.11)
(2.1.12)
(2.1.13)
Comments:
1. The wave equations for E z and H z with the corresponding boundary conditions and
3. Therefore it is important to note within the framework of the present notation that
2
2
E = 0;
H = 0.
(2.1.14)
4. By the superposition principle and the structure of ((2.1.12)--(2.1.13)), the
transverse field components may be derived from the longitudinal ones.
Complete Solution = ( Ez = 0) and( H z =/ 0) + ( Ez =/ 0) and ( H z = 0) .
78
0 r
0r
y
x
In this section our attention will be focused on a specific case where H z = 0 . This
step is justified by the fact that equations ((2.1.12)--(2.1.13)) are linear, therefore by
virtue of the superposition principle (e.g. circuit theory) and regarding H z and Ez as
sources of the transverse field, we may turn off one and solve for the other and vice
versa. As indicated in the last comment of the previous section, the overall solution is,
obviously the superposition of the two. The boundary conditions impose that the
longitudinal electric field Ez vanishes on the metallic wall therefore
mx ny jkz ,n ,m z
(2.2.1)
.
Ez = Anm sin
sin
e
a b
n ,m
This further implies that the transverse wave vector, k , is entirely determined by the
geometry of the waveguide (substitute in (2.1.3))
2
2
2
m
n
2
2
(2.2.2)
k =
+
= 2 kz .
a
b
v
79
0 r
0r
kz, n2 ,m =
v2
k2 =
m n
2
v a b
2
2 m
n
kz =
.
2
v a b
2
(2.2.3)
This expression represents the dispersion equation of the electromagnetic wave in the
waveguide.
Exercise 2.1: Analyze the effect of the material characteristics on the cut-off frequency.
Exercise 2.2: What is the impact of the geometry?
Exercise 2.3: Can two different modes have the same cut-off frequency?
80
y
x
0 r
0r
Comments:
b) There is an angular frequency c,n ,m for which the wavenumber kz vanishes. This is
called the cutoff frequency.
1 m n
m n
f
=
v + ,
c , m ,n vk = v
+
(
)
c m ,n
a
b
2
a b
where v = c/ r r .
2
(2.2.4)
c) Below this frequency the wavenumber kz is imaginary and the wave decays or grows
exponentially in space.
c12
c11
= k zv
81
kz
0 r
0r
e
a
b
m,n
The expression for the transverse wavenumber k is identical to the TM case and so is
the dispersion relation. However, note that contrary to the TM mode where if n or m
were zero the field component vanishes, in this case we may allow n = 0 or m = 0
without forcing a trivial solution.
82
y
x
0 r
0r
TM mode
TE mode
mx ny jkzmn z
Ez = mn Amn sin
sin
e
a b
j n mx
ny
H x = mn Amn 2
sin
cos
k b a
b
j m
mx ny
H y = mn Anm 2
cos
sin
k a
a
b
Hz = 0
jk m mx ny
E x = mn Amn 2 z
cos
sin
k a a b
jk n mx ny
E y = mn Amn 2 z
sin
cos
k b a b
mx ny jkmn z
H z = mn Bmn cos
cos
e
a b
j n mx ny
E x = mn Bmn 2
cos
sin
k b a b
j m mx
ny
E y = mn Bmn 2
sin
cos
k a a
b
Ez = 0
+ jk m mx
ny
H x = mn Bmn 2 z
sin
cos
k a a
b
+ jk n
mx ny
H y = mn Bmn 2 z
cos
sin
k b
a b
(TM )
(TM )
Ex ,mn = Z mn
H y ,mn ; Z
=
mn
( TM )
E y ,mn = Z mn
H x ,mn
1/2
f c ,m ,n 2
1
f
H x ,mn =
H y ,mn =
E y ,mn
(TE )
Z mn
f
(TE )
; Z mn
=1 c ,m ,n
f
E x ,mn
( TE )
Z mn
Exercise 2.4: Check all the expressions presented in the table above.
83
1/ 2
y
x
0 r
0r
b
a
Comments:
1. The phase velocity of the wave is the velocity an imaginary observer has to move
1/2
f
(2.3.2)
= v 1 .
kz
f
2. The phase velocity is always faster than v !! Specifically, in vacuum the phase
velocity of a wave is larger than c . In fact close to cutoff this velocity becomes
"infinite"!!
3. The group velocity is defined from the requirement that an observer sees a constant
envelope in the case of a relatively narrow wave packet. At the continuous limit this
is determined by
v ph
2
c
2
1/2
f c2
(2.3.3)
v gr =
= v 1 2 .
kz
f
84
y
x
0 r
0r
Let us now consider the power associated with a specific TM mode; say
mx ny jkz ,n ,m z
.
Ez = Amn sin
sin
e
a b
At this stage, for simplicity sake, we assume that this is the only mode in the
waveguide. Based on Poynting's theorem, the power carried by this mode is given by
Pmn = Re
dx dy Sz,mn .
0
(2.4.1)
85
0 r
0r
b
1 (TM ) a
H 2 + H 2
a
Pmn = Z
dx
dy
y
x
0
0
2
2
2 2
1 (TM )
n
2 a b m
= Z
Amn
+
2
k4 2 2
a
b
2 2
1 (TM )
2 ab
(2.4.3)
Pmn = Z
Amn
.
k2 4
2
The last expression represents the average power carried by the specific mode.
functions, the total average power carried by the wave in the forward direction is a
superposition of the average power carried by each individual mode separately. In other
words show that P = n ,mPmn .
Exercise 2.7: Show that below cutoff, the power is identically zero although the field is
not zero.
86
y
x
0 r
0r
average magnetic energy per unit length. Calculate this quantity. Compare it with the
average electric energy per unit length.
Exercise 2.9: Calculate the energy velocity of a specific mode v EM = P /WEM . Compare
87
y
x
0 r
0r
b
a
So far it was assumed that the walls are made of an ideal metal ( ) . If this is
not the case ( ) a finite amount of power is absorbed by the wall. In order to calculate
this absorbed power we firstly realize that the magnetic field is "discontinuous" which is
compensated by a surface current
J s = n H.
(2.4.4)
This current flows in a very thin layer which is assumed to be on the scale of the skindepth [ J = J s / ] therefore, the dissipated average power per unit length is given by
2
11
(2.4.5)
PD =
dxdy J ,
2
or explicitly
2
2 Rs
2
Js
1
1
(2.4.6)
=
PD =
dl
dl J s = dl J s ,
2
2
2
88
y
x
0 r
0r
2
1
x
a
(2.4.7)
PD = Rs dl J s .
2
This is the average electromagnetic power per unit length which is converted into heat
(dissipation) due to Ohm loss. Based on Poynting's theorem we may deduce that the
spatial change in the electromagnetic power is given by
d
(2.4.8)
P = PD
dz
and since in case of a single mode both P and PD are proportional to | A |2 ,
(2.4.9)
P A
and
PD A
we conclude that the change in the amplitude of the mode is given by
2
d A
2
2
2
(2.4.10)
= 2 A A( z ) = A( z = 0) e2 z .
dz
The coefficient represents the exponential decay of the amplitude and based on the
arguments of above is given by
P
(2.4.11)
D.
2P
2
89
0 r
0r
Exercise 2.11: Based on the previous calculation of the power show that this parameter is
given by
( TM )
m,n
1 m 2b 3 + n 2 a 3
=
2 2
2 3
1 ( fc / f )2 b m b a + n a
2 Rs
(2.4.12)
Explain the
If the dielectric coefficient of the material is not ideal, in other words, it has an
imaginary component r = ' j '' , then the wavenumber is given by
1
(2.4.13)
kz kz (0) j ''
,
(0)
2 c ckz
where we assumed that (i) the dielectric loss is small and (ii) the system operates remote
from cutoff conditions [i.e. kz (0) '' / c ].
90
0 r
0r
0
0
a
b
2
1
1 2
.
(2.4.14)
S z ,mn = ( Ex H *y E y H x* ) =
E
E
+
x
y
(
)
TE
mn
2
2 Z mn
mn
2
2
2 2
1 1
2 2 m 1 1
2 n 1 1
Pmn =
Bmn 4
a b+ 4
a b
(TE )
2 Z mn
k a 2 2
k b 2 2
1 ab
2
B
mn
(TE )
8 Z mn
k2
2
(TE)
mn
=
2 Rs
(2.4.15)
2
2
1 b f c b g n f c m 2 ab + n 2 a 2
1 + + 2 2
. (2.4.16)
2 2
2 b
+
a
f
a
f
m
b
n
a
2
f c
91
Repeat all the exercises from the above (TM mode) for the TE mode. Make a comparison
table where relevant.
and (TM ) are large close to cutoff and increase as
for large frequencies. In between there is a minimum loss for an optimal
frequency. Calculate it.
Exercise 2.14: Calculate the loss very close to cutoff.
Exercise 2.13: Note that both
( TE )
92
93
TE10
TE11
3
2
1
1
y
x
TE21
z
1
2
y
x
z
1
TM 11
TE20
3 2
TM 21
3
2
2
1
1
y
x
y
z
x
y
x
z
1
94
where k =
2
v2
k z2 or explicitly
2 1 1 2
2
(2.6.2)
k
+
+
+
E z = 0.
r 2 r r r 2 2
An , s cos(n ) + Bn , s sin(n ) ,
(2.6.3)
E z = J n ps , n e
R
n,s
where J n (u ) is the n 'th order Bessel function of the first kind. This function behaves
similar to a trigonometric function (sin or cos ). It has zeros, denoted by ps ,n i.e.,
ps ,n : J n ( ps ,n ) 0.
95
The first few zeros of the Bessel function are tabulated next.
n=0
n=1
n=2
s=1
2.405
3.832
5.135
s=2
5.52
7.016
8.417
s=3
8.654
10.174
11.62
r
r
r
r
R
thus
2
2
2
p
s ,n
2 ps , n
2
2
2
(2.6.5)
k 2 k z =
kz, s ,n = 2 2 .
R
v
v
R
Based on this expression the characteristic impedance of the TM mode is given by
kz,s ,n v
(TM )
(2.6.6)
.
Z s ,n =
96
An , s cos(n ) + Bn , s sin(n ) ,
H z = J n ps ,n e
R
s ,n
where
H z
= 0 ps ,n : J n ( ps ,n ) = 0.
r r = R
The first few zeros of the derivatives of the Bessel function are
s=1
s=2
s=3
n=0
3.832
7.016
10.174
n=1
1.841
5.331
8.536
n=2
3.054
6.706
9.970
thus
)
Z s(,TE
n =
vk z , s ,n
, kz,s ,n
2
ps ,n
=
v R
97
(2.6.7)
(2.6.8)
(2.6.9)
(2.6.10)
Summary
TE modes
TM modes
Hz =
r jksn z cos n
J n psn e
sin n
Ez =
r jksn z cos n
J n psn e
sin n
jk p
r jksn z cos n
Hr =
sn2 sn J n' psn e
Rk ,sn
R
sin n
jnk sn
r jksn z sin n
H=
J
p
n sn
e
rk2 ,sn
R
cos n
Er
Z sn(TE ) H
E =
Z sn(TE ) H r
E / Z sn(TM )
Er / Z sn(TM )
2
Rk ,sn
R
sin n
jnksn
r jksn z sin n
J n psn e
2
rk ,sn
R
cos n
1/2
1/2
2 psn 2
ksn =
2
v R
2 psn 2
2
v R
vksn
Z sn(TM )
Z sn(TE ) =
vksn
k ,sn = psn / R
psn / R
98
TE modes
k0 ksn
Power=
( psn'2 n 2 ) J n2 ( psn )
4
2k ,sn g 0,n
2
2
Rs k,sn v
=
1
2
R
1/2
k2 ,sn v 2
n2
+
2
psn'2 n 2
0 r
v=
, =
0 r
0 r 0 r
1
TM modes
k0 ksn 2
psn [ J n (k ,sn R)]2
4
2k ,sn g 0,n
g 0,n
2
2
Rs k,sn v
1
2
R
1 n = 0
=
2 n =/ 0
99
1/2
jk
E
E = 2z Ez ,
H = r ,
H r = E / ZTM .
k
ZTM
Consequently, the average Poynting vector is
*
1
S z = E H 1z
2
and the average power flowing in the waveguide
* 1
1
P = Re da E H = Re da Er H * E H r*
c.s .
z
2 c.s .
2
2
E
1
|
|
E
E 2 + E 2
= 1 Re
= Re da r +
da
c.s. r
ZTM 2 ZTM
2 c.s. ZTM
kz
1
1 k z2
2
2
(2.6.11)
.
=
Re da 2 Ez =
Re
da
E
z
4
c.s.
c.s .
k
Z
k
2 ZTM
2
TM
In order to further simplify the last expression let us examine the wave equation:
(2.6.12)
( 2 + k2 ) Ez = 0
100
(2.6.13)
2
2
z da ( Ez* Ez ) Ez + k2 Ez = 0.
*
*
da
E
=
d
Ez = 0
(2.6.14)
hence
2
da Ez = k da Ez .
2
(2.6.15)
Now back to the propagating power for a superposition of modes starting from the
expression for a single mode we get
101
1 k z2
Ps ,n =
Re da Ez
(TM ) 4
2 Z s ,n k
1 k z2
=
da Ez
(TM ) 2
2 Z s ,n k
kz
A
J
p
n
cos(
)
k n,s n n,s R
2
R
s ,n
1
P=
d drr
0
/ v
2 0
s ',n '
The integration over the angle is straight forward
n=0
1
2
d
cos(
n
)cos(
n
'
)
2
g
,
g
=
1
n n ,n '
n
0
n =/ 0
2
and after integration over r we get
2
2
1 k z , s ,n 2 R 2
P =2 g n
A
J
p
(
)
s ,n
n
n , s s , s
(TM ) 2
k
2
2 Z s ,n , s ,n
n , s , s
where we used
R
2
r
r R2
dr
r
J
p
J
p
J
p
=
(
)
n n,s
n
n,s s ,s '
n n,s '
0
R
R 2
102
(2.6.16)
(
)
J
p
1
2 pn , s
n
n,s
Ps ,n =
RAs ,n R g n
(2.6.17)
Re R 2 2 .
v
R
2
v
pn , s
Note that there is power flow only if the wave is above cutoff and as in the rectangular
case, the the total power is the superposition of the power in each mode separately
2
Exercise 2.15: Calculate the average energy per unit length stored in the
electromagnetic field.
103
Now to the general expression for the losses. Starting from the dissipated power
2
PD , s ,n =
Er ,ns
2
Rs
1
1
(s) 2
J
d
R
H
d
R
d
=
=
z
s
s
(TM )
bound
2
2
2
Z ns
1 Rs
2
2 2 J n ( pns )
k
R
Rg
A
R
=
2
n
ns
p
2 RZ (TM ) 2
ns
ns
2
1 Rs J n ( pns )
2
R
Rg
A
R
=
2
n
ns
2 R 2 v pns
which finally entails for a single mode
1 Rs v ph , sn
ns(TM ) =
.
R v
Exercise 2.16: Check Eq. (2.6.19).
2
(TE )
104
(2.6.18)
(2.6.19)
TE11
Attenuation, decibels/meter
0.010
0.008
0.006
0.004
0.002
0
TE11
0
1
TM 01
TE 21 TE 01 TE 31 TE 41 TE12
2
TM 11
3
TM 21 TM 02
TM 01 TE 01
0.4 0.8
1.2
1.6
20
24
28
32
36
40
Diameter
Wavelength
Attenuation due to copper losses in circular waveguides at 3000 Mc/sec.
0.07
Attenuation, decibels/meter
0.06
TM 01
0.05
0.04
0.03
TE 01
0.02
TE11
0.01
1000
5000
9000
13000
Frequency, megacycles/second
Attenuation due to copper losses in circular
waveguides; diameter = 2 in.
105
Wave Type
TM 01
TM 02
TM 11
TE 01
Field distribution
in cross-sectional
plane, at plane of
maximum
transverse fields
TE 11
Distributions
Below Along
T his Plane
Distributions
Below Along
T his Plane
Field distribution
along guide
Field components
present
p or p
( kc )
( c )
( fc )
Attenuation due to
imperfect
conductors
E z , Er , H
E z , Er , H
Ez , Er , E , H r , H
H z , H r , E
H z , H r , H , Er , E
2.405
2.405
5.52
5.52
3.83
3.83
3.83
3.83
1.84
1.84
2.61
0.383
1.14
0.877
1.64
0.609
1.64
0.609
3.41
0.293
Rs
n 1 ( f c / f )2
Rs
n 1 ( f c / f )2
Rs
n 1 ( f c / f )2
106
Rs
( f c / f )2
n 1 ( f c / f )2
Rs
n 1 ( f c / f )2
r
(2.7.1)
Ez (r , z , t ) = J 0 ( ps ) d exp [ jt s z ] s ( )
R
s =1
2
2
wherein s = ps / R 2 2 / c 2 and s ( ) is the Fourier transform of this field component
at z = 0
R
1
r 1
(2.7.2)
dr
r
J
dt exp ( jt ) Ez (r , z = 0, t ).
s ( ) = 2
0 ps
0
R 2
R 2
J 1 ( ps )
2
Let us now calculate the energy associated with the radiation field as it propagates in
an empty waveguide. The transverse field components are
rp
Er (r , z , t ) = J1 ps s d exp ( jt s z ) s s 2
R R
s =1
2s + 2
c
s
1 j
r ps
(2.7.3)
H (r , z , t ) = J1 ps d exp ( jt s z )
2
2
R
R
c
s =1
0
2s + 2
c
therefore, the z -component of the Poynting vector is
107
r
R
S z (r , z , t ) = J1 ps d exp ( jt s z ) s ( ) s
ps
s =1 R
r
R
1
J1 p d exp ( jt () z ) () j
0 cp
=1 R
Using the orthogonality of Bessel functions
R
r R2 2
r
0 drrJ1 ps R J1 p R = 2 J1 ( ps ) s
the power propagating is
R
R2 2
P ( z , t ) = 2
J1 ( ps ) d exp ( jt s z ) s ( ) s
2
s =1
ps
R
1
d exp ( jt s z ) s () j
cp
s
0
108
. (2.7.4)
(2.7.5)
(2.7.6)
WR ( z ) = dtP( z , t )
R2 2
= (2 )
J1 ( ps ) d s ( )exp s ( ) z d s ()exp s ( ) z
s =1 2
R1 1
s ( ) j R dt exp [ j ( + )t ].
ps 0 c ps
(2.7.7)
1
j ( +)t
we have
dt
e
R2 2
WR ( z ) = (2 )
J1 ( ps ) d s ( )s ( )exp z s ( ) + s ( )
2
s =1
2
(2.7.8)
( ) R 1 1
s
j R .
p
c ps
s
o
For proceeding it is important to emphasize two features: since the field components are
real functions, it is evident that the integrand ought to satisfy s ( ) = s* ( ) .
Consequently
(2.7.9)
s ( ) = s* ( )
and
(2.7.10)
s ( ) = *s ( ).
109
This last conclusion implies that if the frequency is below cut-off (| |< ps c / R ) then
(2.7.11)
s ( ) =| s ( ) |
whereas if | |> cps / R
(2.7.12)
s ( ) = j | s ( ) |.
With these observations we conclude that
cps /R
2 s ( ) z
(2 ) 2 R 2 2
2
WR ( z ) =
J1 ( ps )2Re d | s ( ) | e
0
0 s =1 2
(2.7.13)
j R | s ( ) | R
2 R | s ( ) | R
+
d
|
(
)
|
s
cp /R
.
ps
ps
s
cps
cps
Clearly the first integrand is pure imaginary therefore, its contribution is identically
zero and as a result, in the lossless case considered here, the energy associated with the
propagating signal does not change as a function of the location
(2 )2 R 2 2
2 R | s ( ) | R
(2.7.14)
(
)
2
|
(
)
|
W=
J
p
d
s
cp
.
p
0 s =1 2 1 s cps / R
s
s
110
Comments:
1. Only the propagating components contribute to the radiated energy.
2. The non-propagating components are confined to the close vicinity of the input
(2.7.14).
Let us now simplify the discussion and focus on a source which excites only the first
mode ( s = 1) i.e., Ez (r , z = 0, t ) = J 0 ( p1 r/R ) E0 (t ) therefore according to Eq.(2.7.2) we get
R
1
r r 1 jt
dr
r
J
J 0 p1
dt e
E0 (t ')
s ( ) = 2
0 ps
R 2
R R 2
J 1 ( ps )
(2.7.15)
2
jt
E0 (t ')
= s ,1 dt e
2
111
T
2
1 + 2 jT + T ( )
Substituting in Eq.(2.7.14)
W=
1
2 4 J1 ( p1 )
E
0R
2
0c
p
1
( T ) ( )
d
1 + T ( ) + ( 2T )
1 = cp1/R
2
1
(2.7.17)
2
2
2
2
2
p1
T + u + 4u T
Its numerical analysis reveals that
2
1 <1
2 3 J1 ( p1 )
(2.7.19)
W 0 E0 R
2
p1 4
T > 1
Exercise 2.20: Compare with the dependence of the propagating energy in the frequency
when the signal prescribed by Eq.(2.7.16) propagates in free space.
2
0
112
113
114
x
y nx ,n y z
Ez ( x, y, z; ) = An ,n sin nx sin n y e
x y
ax
a y
nx , n y
x
y mx ,m y z
(3.1.1)
H z ( x, y, z; ) = Bm ,m cos mx cos m y e
x y
a
a
mx , m y
x
y
wherein = k r r ( / c )
2
and k = ( nx / ax ) + ( n y / a y )
2
j
j
E = 2 Ez + 2 1z H z , H = 2 H z 2 1z Ez (3.1.2)
k
k
k
k
115
Let us now assume that the relative dielectric coefficient ( r ) depends on the
transverse coordinates (e.g. x and y in a rectangular geometry) and so is the relative
permeability coefficient ( r ) , the goal being to determine the character of the modes
supported by the
again is Maxwell's equations
structure. Our
starting point
E = j0 r H
( 0 r H ) = 0
(3.1.3)
H = j o r E
( 0 r E ) = 0
which lead to the following
wave equations
( E ) = j0 ( r H ) = j0 [ r H + r H ]
2
E + ( E ) = j0 [ r H + r ( j 0 r E )]
2 2
(3.1.4)
+ c 2 r r Ez = ( E ) + j0 r H 1z
2
2
2
(3.1.5)
+ + r r c 2 Ez = ( E ln r ) + j0 r ( ln r H ) 1z
116
2
2
2
+ + r r c 2 H z = ( H ln r ) j 0 r ( ln r E ) 1z
117
(3.1.6)
no longer
are linear
frequently
y
b
r
2 r 2
2
+
+
2
Ez = 0
1
c
(3.1.7)
2
m 2
2
A sinh x
0 xd
r 2
b
c
z
sin m
Ez = e
2
b
m
2
2
B sinh ( x a )
2 d x a
c
+
+
2
Hz = 0
1c
and
118
(3.1.8)
(3.1.9)
y
b
r
2
m 2
2
C cosh x
0 xd
d
r 2
b
c
m
z
(3.1.10)
cos
Hz = e
y
2
b
m
2
2
D cosh ( x a )
2 d x a
c
Imposing the boundary condition at x=d leads us to the dispersion relation. In order to
emphasize the coupling between TE and TM modes we shall develop the dispersion
relation in two steps.
Step I: we impose the continuity of Ez and H z facilitating to redefine the amplitudes
thus A and D according to
sinh( x)
0 xd
sinh( d )
y
z
(3.1.11)
sin m
Ez = Ae
b sinh [ v (a x) ]
d xa
sinh [ v (a d ) ]
and
119
y
b
r
cosh( x)
0 xd
cosh( d )
d
z
(3.1.12)
cos m
H z = De
cosh
(
)
a
x
b
[
]
v
d xa
cosh [ v (a d ) ]
2
2
wherein 2 ( m / b ) 2 ( / c ) is the horizonthal wave number in the dielectric
Step II: In our second step if 0 < d < a we need to impose continuity of E y and H y at
x = d . According to (3.1.2) the transverse components are
j0
Ey =
E
+
x Hz
y z
2
2
2 + r 2
2 + r 2
c
c
j 0 r
(3.1.13)
Hy =
H
x Ez
y
z
2
2
2 + r 2
2 + r 2
c
c
120
y
b
r
m
d
E y ( x = d 0) =
A
D tanh( d )
+
2
2
b
2 + r 2
2 + r 2
c
c
j0
m
(3.1.14)
E y ( x = d + 0) =
A
D v tanh[ v (a d )]
2
2
b
2+ 2
2 + 2
c
c
thus their continuity implies
m
v tanh [ v (a d ) ] tanh( d )
b
b
+
= j0 D
A
(3.1.15)
2
2
2
2
2
2
2 + 2
2 + r 2
+ r 2 + 2
c
c
c
c
121
y
b
r
Similarly
j 0 r
d
m
(3.1.16)
coth(
)
H y ( x = d 0) =
D
A
d
2 b 2
2
2
+ r 2
+ r 2
c
c
j 0
m
H y ( x = d + 0) =
D
A ( v )coth[ v ( a d )] (3.1.17)
2
2
2 + 2
2 + 2
c
c
thus
m
m
j 0
j 0 r
b
v coth[ v (a d )] +
coth( d ) + D
+ b 2 = 0.
A
2
2
2
2
2
2 +
2 +
+
+ 2
r 2
r 2
2
c
c
c
c
(3.1.18)
The dispersion relation, in a matrix form, may by now be expressed using (3.1.15) and
(3.1.18)
122
y
b
r
TE
tanh( d )
v tanh[ v ( a d )]
2
2 2 1
2 + 1
r
+ c 2 j
c 2 j0
1
1
m
+
2
2
b 2
2
+ 2
+ r 2
c
c
D
= 0
A
TM
j 0 v
j 0 r
coth[
(
a
d
)]
coth(
d
)
v
2
2
2 + 2
2 + r 2
c
c
m
1
1
2
2
b 2
2
+ r 2 + 2
c
c
(3.1.19)
or formally
11 12 D
= 0
(3.1.20)
21 22 A
implying that for a non trivial solution the determinant of the matrix must be zero
11 12
= 0 + 2 = 0.
(3.1.21)
det
( 12 )
11 22
21 22
123
y
b
r
Comments:
124
r
r
k
+
r dr dr c 2 r ( )
E (r ) = 0
125
Rd
(3.2.1)
2
1 d d
2
r + 2 k E1 (r ) = 0
c
r dr dr
1 d d 2
2
r + 2 k E2 (r ) = 0.
c
r dr dr
v2
(3.2.2)
(3.2.3)
126
Rd
R
AS0 (r ) 0 < r < Rd
Ez ( r ) =
BT0 (r ) Rd < r < R
J 0 ( r )Yo ( R ) Yo ( r ) J 0 ( R )
2 > 0
(3.2.4)
T0 (r )
2
<
I
r
K
R
K
r
I
R
(|
|
)
(|
|
)
(|
|
)
(|
|
)
0
0
0
0
0
J 0 ( v r ) v2 > 0
S0 (r ) =
2
I 0 (| v | r ) v < 0
As in the previous section the dispersion relation will be now determined by imposing
the boundary conditions at the interface between the two regions. Firstly, we impose the
continuity of Ez at r = Rd :
(3.2.5)
Ez AS0 ( Rd ) = BT0 ( Rd )
j 0 r
k r ( / c )
2
| |
| v |
Ez
r BT 0 ( Rd ) = 2 A S 0 ( Rd )
2
v
r
wherein
127
(3.2.6)
Rd
Rd
J ( r )Y ( R ) + Y1 ( r ) J 0 ( R)
> 0
T 0 (r ) 1 o
2
I1 (| | r ) K 0 (| | R ) + K1 (| | r ) I 0 (| | R ) < 0
(3.2.7)
2
J ( r ) v > 0
S 0 (r ) = 1 v
2
I1 (| v | r ) v < 0
From the previous two equations[(3.2.5), (3.2.6)] we obtain the dispersion equation
| |
| v |
(3.2.8)
r S0 T 0 = 2 S 0 T0
2
R
20
15
f [GH z ]
empty waveguide
10
5
0
0
128
dielectric filled
waveguide
1
3
kR
Exercise 3.4: Analyse the effect of Rd on the phase velocity, group velocity and ohm
dielectric
air
129
metal
130
d
z
y
z=0
d
a
z
a
z=0
2
2
nx
(ref )
(3.3.2)
H z = H 0 n cos
z
n
a
c
exp
;
=
/
/
(
)
(
)
(
)
n
n
n =0
131
d
a
z
x
z=0
2
2
mx
( tr )
(3.3.3)
H z = H 0 m cos
z
m
d
c
exp
;
=
/
/
(
)
(
)
(
)
m
m
m =0
m representing the transmitted coefficient of each mode. Three observations at this
stage:
1. The amplitude of both reflected ( m ) and transmitted ( m ) will be determined by the
boundary conditions to be imposed shortly.
2. The character of the mode (propagating or evanescent) is determined by the angular
frequency ( ) of the incident wave.
3. The choice of TE 01 is dictated by the need for a simple analysis. If a higher mode is
launched (e.g. TE 11 ) there will be coupling to TM modes.
132
a
(inc)
y
(sc)
y
z =0
z
x
a mx
E y(sc) |z =0 = j0 H o m
sin
a
m
d mx
E y( tr ) |z =0 = j0 H o m
sin
.
m d
m
Substituting in (3.3.4) we get
d
mx
sin
0< x<d
a
mx
[
]sin
=
+
m
d
m m m,1 m a m
0
d < x < a.
133
(3.3.4)
z=0
(3.3.5)
(3.3.6)
d
a
a
d
1
1 d
mx mx
m dxsin
[ m,1 + m ] =
sin
,
0
m
2 m m a
d a
or
m ,1 + m = Amm m ,
z=0
(3.3.7)
(3.3.8)
wherein the matrix Amm is proportional to the overlap integral of the modes from both
sides of the discontinuity
2
d m 1 d
m m
(3.3.9)
Amm 2 dx sin
x sin
x .
0
a m d
d a
Step 2: Continuity of the x -component of the magnetic field in the region 0 < x < d
implies
(H
( inc )
x
+ H x( sc ) |z =0 = H x( tr ) |z =0
and since H = 0 or explicitly z H z + x H x = 0 we get at z = 0
134
(3.3.10)
d
a
x
z=0
H
= H 0 sin
x
( jkz01 )
a
a
mx
sin
H x(sc) = H 0 m
m
m
a
m
d
mx
(3.3.11)
sin
H x( tr ) = H 0 m
( m ).
m
d
m
Thus, in the region z = 0 and 0 < x < d
a
d
mx
mx
(3.3.12)
[
]sin
=
sin
;
m 0m m,1 m a
om m
d
m m
this time we use the orthogonality of sin ( mx / d ) in the range 0 x d we get
1 d
1 d
m m
x sin
x , (3.3.13)
m
m =
m m ,1 m dx sin
0
d
2 m
m m
d a
or
(3.3.14)
m = Bmm [ m ,1 m ],
(inc)
x
135
d
a
a m m 1 d
m m
Bmm 2
dx
x sin
x .
sin
0
d m m d
d a
We can now write (3.3.8) and (3.3.14) in avector form
+ = A and = B( )
(3.3.15)
z=0
(3.3.16)
1
= [ I + AB] [ AB I ]
(3.3.17)
= 2[ I + B A]1 .
dx
m
m
m
m
sin
sin
=
sinc
sinc
+
. (3.3.18)
0
d
a
a
d a
136
d
a
z
z=0
1 2 a
a
Pref = H 0 b | m |2
Re( j0 m )
m
2
2 m =1
1 2 d
d
(3.3.19)
Ptr = H 0 b | m |2
Re( j0 m )
m
2
2 m =1
Assuming that at the operating frequency there is only a single propagating mode in
each region, then the average power may be written as in a transmission line
1
k a ab
, I1 = H 0 1
Pinc = | I1 |2 Z1 Z1 = 0
2
2
ck1
1
| I1 |2 | 1 |2 Z1
2
1
kd
, I2 = H0 2
Ptr = | I 2 |2 | 1 |2 Z 2 Z 2 = 0
2
ck2
Pref =
137
db
2
(3.3.20)
( / c )
( / a ) and k2 =
2
( / c )
( / d ) .
Reflection coefficient
wherein k1 =
1.2
1.0
0.8
frequency:
1. The power can be injected at frequencies
0.6
above 6GHz (a=2.5cm).
0.4
2. To the point whereby the first mode in the
4 6 8 10 12 14 16 18
second part is above cutoff
f [GH z ]
f c 8.485 GHz ( d = a / 2 )
the reflection coefficient of the first mode is unity.
3. Above 8.485GHz power can be transferred to the second part therefore the
inc
ref
tr
reflected power drops yet P1( ) = P1( ) + P1( ) .
4. Above 12 GHz for which the second mode in the left-hand side may carry
power, the power reflected in the first mode decreases further yet beyond
inc
ref
tr
inc
ref
ref
tr
this point P1( ) P1( ) + P1( ) but rather P1( ) = P1( ) + P2( ) + P1( ) .
138
r
Ez(inc) (r , z ) = E0 J 0 p1
exp ( 1 z )
Rext
Rint
d
Rext
z
z=0
r
(3.3.21)
(r , z ) = E0 s J 0 ps
exp ( s z ),
R
s =1
ext
(ref )
z
139
Rint
d
r
(d )
Ez (r , z < d ) = E0 J 0 pn
An exp ( n z ) + Bn exp ( n z )
n =1
Rint
Rext
z
z=0
(3.3.22)
r
Ez( tr ) (r , z > d ) = E0 s J 0 ps
exp [ s ( z d ) ]
R
s =1
ext
amplitudes of the transmitted and reflected waves ( s and s ). For this purpose let us
concentrate on the field components at z = 0 .
H (left )
Rext j 0
r s ,1 exp ( 1 z ) + s exp ( s z )
E
J
p
=
(left )
0
1 s
exp
z
exp
z
p
R
E
(
)
(
)
s
s
s
1
s
1
ext s ,1
(3.3.23)
(d)
H
Rint j 0
r An exp ( n z ) + Bn exp ( n z )
E
J
p
=
(d)
0
1 n
A
z
B
z
exp
exp
p
R
E
(
)
(
)
n
n
n
n
n
n
int n
140
Rint
d
(3.3.24)
Er (r , z = 0 ) =
Rint < r < Rext
0
Using the orthogonality of the Bessel functions
2
Rext
r
r Rext
2
=
( ps ) ss
drr
J
p
J
p
J
s
s
1
1
1
0
2
Rext Rext
we obtain from (3.3.24)
Rint
Rint
1
r
r
[
]
s ,1 s =
A
B
drr
J
p
J
p
1 n
n
n
n 0
1 s
2
s Rext Rext
R
2
n =1 pn
ext
Rint
J 1 ( ps )
2
ps
or
Rext
z
z=0
(3.3.25)
(3.3.26)
s ,1 s = sn[ An Bn ],
(3.3.27)
n =1
ps n 1 Rint
2 Rint
r
,
U
U
dr
r
J
p
sn
s ,n
s ,n
1 s
3
pn s J12 ( ps ) Rext
Ri2nt 0
Rext
141
r
J
p
1 n
R
int
(3.3.28)
Rint
d
Rext
The continuity of the azimuthal magnetic field in the region 0 < r < Rint is
H ( d ) (r , z = 0 ) = H ( d ) (r , z = 0+ ) . Again we use the orthogonality of the
Bessel functions in the inner region
2
Rint
Rint
R Rint
r
r
2
[ An + Bn ]
J1 ( pn ) = [ s ,1 + s ] ext drr J1 ps
J
p
1 n
0
2
pn
p
R
R
s =1
s
ext
int
hence
An + Bn = ns ( s ,1 + s ),
z
z=0
(3.3.29)
(3.3.30)
An exp ( n d ) + Bn exp ( n d ) = ns s
s =1
142
(3.3.32)
Rint
d
Rext
(3.3.33)
n =1
z
z=0
B ( + ) B I I
( 3.3.30 ) A + =
1
1
A e d
( 3.3.32 ) Aed + Bed =
0 I I 1
=
d
d
B 0 e d I I
( 3.3.33) ( Ae Be ) =
1
0
(3.3.34)
Substituing we get
143
Rint
d
Rext
z
z=0
1 1 I
=
I
or
1
1 1 I
I e d
I 0
0 I
e d I
I 1 1
I 0
(3.3.35)
I e d
I 0
0 I
e d I
I 1 1
0 =
I
or using a short notation
Q11 Q12
=
(3.3.36)
Q
21 Q22 0
thus Q11 + Q12 = and Q21 + Q22 = 0 . Consequently,
(3.3.37)
= Q221Q21 , = [Q11 Q12Q221Q21 ] .
The Figure below illustrates the reflection coefficient and power conservation as a
function of the frequency. We observe that below the cut-off of the second mode all the
propagating power is in the first mode. As the frequencies exceed the cut-off of the
second mode, part of the power leaves the system through the second mode.
144
1.1
0.9
11 + 11
0.7
1.0
11
0.8
0.9
0.6
0.5
5
10
15
0.8
20
10
15
f [GH z ]
f [GH z ]
The reflection coefficient as a function of the frequency and energy conservation of the first mode. Note that as
long as the higher mode is below cutoff,
first mode to the second and therefore
145
20
146
J y , s ( x ) = I ( x d ) = U m sin m
a
m =1
(3.4.3)
hence
2
d
(3.4.4)
I sin m .
a
a
(sec)
(3.4.5)
E y ( x, z ) = m sin m exp ( m z )
a
m =1
where 2m = ( m / a ) 2 ( / c) 2 . In order to establish the relation between the current's
"harmonics'' and the secondary field we observe that based on Maxwell's equations the
magnetic field
x
1
(sec)
(3.4.6)
H x ( x, z ) =
sin
m
m m
exp ( m z ) sgn( z )
j0 m =1
a
147
I
d j0
(3.4.8)
.
m = sin m
a
a m
We have yet to determine the relation between the current ( I ) in the filament and the
amplitude of the incident wave. This is done by relaxing the condition of our infinitely
small filament and imposing that the tangential electric field vanishes on the wall of the
post. Specifically, averaging the electric field over the azimuthal coordinate at r = R we
must get zero total field, namely
(3.4.9)
z = Rsin
explicitly
E0 m ,1 exp ( m R sin ) m
2
d
sin
d
R
cos
0
+
=
(
)
m =1
+m exp ( m R sin ) a
which formally may be rewritten by defining
1 2
m
(1)
exp
sin
sin
cos
d
R
d
R
+
(
)
(
)
m
2 0
a
and
1 2
m
exp
sin
sin
cos
d
d
R
(2)
+
(
)
(
)
m
m
a
2 0
148
(3.4.10)
(3.4.11)
(3.4.12)
hence
1
d j0 (2)
m = 0 I =
E I sin m
a
a
m
m =1
(1)
0 1
1(1)
E0 a
( j / c ) sin ( md / a )
(2)
m
m =1
(3.4.13)
At this stage we may define the reflection coefficient
(1)
m 1 ( j / c m ) sin ( md / a )
=
m
E0
(2)n ( j / c n ) sin ( nd / a )
(3.4.14)
n =1
1 =
1(1) ( / c ) sin ( d / a )
(3.4.15)
n =2
At this point it can be shown that since m = 1 is a propagating mode 1(1) = Re( 1(2) ) and
the second term in the denominator is pure imaginary
1
(3.4.16)
.
1 =
(2)
(2)
n sin ( nd / a )
Im(1 )
j
1+ j
+
(1)
1(1)
n =2 1 n sin ( d / a )
149
Based on this last result, we conclude that the post may be represented by a reactance
since according to the equivalent scheme below
1
1
1 =
=
1 + 2Z L 1 + 2 j X L
Z0
Z0
ZL
and
1 Im[1(2) ] 1 (2)
sin ( nd / a )
n
XL =
+
.
(1)
(1)
2 1
2 n =2 1 n sin ( d / a )
Further simplification, is achieved assuming that R d we get
d 1
d
(1)
(3.4.17)
exp
sin
sin
d
R
m
m sin m
(
)
m
I0 ( m R )
2
a
a
d 2
(2)
m sin m
d exp ( m R sin )
0
2
a
(3.4.18)
d
k
1
sin m I 0 ( m R ) + 2 ( 1) sinc k I k ( m R )
a
2
k =1
thus
150
I 0 ( n R ) 4 ( 1)k I 2 k +1 ( n R )
d
2
sin
n
R
R
a
n =2
k = 0 2k + 1
n
n
(3.4.19)
XL =
J0 ( R) 2 d
sin
R
a
R 1 and n R 1 thus
At the limit R 0 or more precisely assuming
I ( x 1) ( x / 2 ) / ! we get
a 1
Int
XL =
n =2
2k
k
2 k +1
1
4 ( 1) ( n R ) 1
d
2
sin
n
a
n R k =0 2k + 1 ( 2k + 1)! 2
1
d
sin 2
R
a
151
(3.4.20)
Exercise 3.8: Analyze (numerically) the dependence of the normalized reactance on three
parameters; the angular frequency ( ) , its location ( d ) and its radius ( R ) . It is possible
to simplify the two quantities defined in (3.4.11) and (3.4.12) by using
z 1
exp t = t k J k ( z ) ; J k ( z ) = (1) k J k ( z ).
2 t k =
152
3.4.2 Diaphragm
x
x
+
+
exp
j
exp
j
z
sin
exp
z
sin
m
(
) 1 ( ) m ( m )
z<0
a m =2
E y = E0
1
a
a
m =2
(3.4.21)
Let us denote by E y ( x, =
z 0=
) E0 ( x) the transverse electric field in the aperture, thus
153
2 d
x
(3.4.22)
dx
x
(
)
sin
0
a
a
x
2 d
(3.4.23)
m 2:
m = m = dx ( x ) sin m .
0
a
a
1 + 1 = 1 =
a m =2
(3.4.24)
Substituting (3.4.23) we get
m
x4 d
x
x
2 1 sin =
(3.4.25)
sin m a a 0 dx ( x ) sin m a
a
j
m =2
and further multiplying and deviding by 1 + 1 and using (3.4.22) we finally obtain
m
2 1
x d
x
x d
x'
sin dx ( x )sin = 2 j
sin
m
dx
(
x
)sin
m
0
.
0
1 + 1
a
a
a
a
m=2
(3.4.26)
154
2 1
in the left hand side may be rewritten
1 + 1
m
x d
x
x d
x'
sin
(
)sin
YL sin dx ( x)sin = 2 j
m
dx
x
m
0
. (3.4.27)
0
a
a
a
a
m=2
So far the formulation is exact (of course subject to the specified assumptions). At this
stage we multiply both sides by ( x) and integrate over the cross section of the aperture
and get
2
d
x
dx ( x)sin m
0
a
(3.4.28)
.
YL = 2 j m
2
m=2 d
x '
(
)sin
dx
x
0
a
which evidently entails that the distribution of the electric field in the aperture
determines the admittance of the diaphragm as experienced by the wave. For a rough
approximation we may consider the simplest form of ( x ) that satisfies the boundary
conditions at x = 0 and x = d but it ignores edge effect namely, ( x) x ( d x ) . An
Since the admittance is Y L =
155
a
2 a
Exercise 3.12: What is the effect of the edge of the diaphragm?
156
y
h
a
b
d
x
z
t
157
In order to have the antenna radiate in one direction only, a short-circuiting plunger is
placed at a distance L to the left of the probe. The basic concept is to generate a current
( I ) at a location where the parallel electric field is maximum -- quarter wavelength
from a short circuiting plane.
y
L
b
z
2Rext
2Rint
The total field in the waveguide could be found if Ea and H a , the total tangential
fields in the aperture were known. Unfortunately, these aperture fields are unknown,
but, if the coaxial-line opening is small, we may, to a first approximation, assume that
the higher-order coaxial-line modes which being excited, are negligible. The field in the
aperture will then be that associated with the incident and reflected TEM modes in the
coaxial line.
158
y
L
b
z
+ ( z L) 2 .
(
)
int
2
Rint
This current density excites a TE nm mode satisfying
2 2
1 2
(3.5.3)
+ c 2 E y = j0 J y j y 2 J y
0
which in terms of Green's function reads
c2 2
E y ( x, y, z ) = j0 dxdydzG y ( x, y, z | x, y, z) J y ( x ', y ', z ') + 2
J
x
y
z
',
',
'
.
(
)
y
2
y '
(3.5.4)
In turn Green's function for our configuration is given by
159
2Rext
2Rint
y
L
b
z
G y (r | r ') =
x
x
+
n
n
(1
)sin
sin
m ,0
ab
2
a
(
)
n =1, m =0
n ,m
2Rext
2Rint
y
y
}
(3.5.5)
2 2
+ 2
c
G y = ( x x) ( y y) ( z z)
G y
= 0, G y ( x, y, z = 0) = 0.
G y ( x = 0, a ) = 0,
y y =0,b
Substituting (3.5.4) and (3.5.2) in (3.5.1) as well as taking advantage of the probe being
thin we have
j0
c2 2
*
+
(
)
(
|
)
(
)
(
)
dyI
y
dy
g
y
y
I
y
I
y
= P + 2 j[WM WE ] (3.5.6)
2
'2
wherein
160
y
L
b
z
g ( y | y) G ( x = h, y, z = L | x = h, y, z = L ) =
n =1
m =0
( 2 ) ab
n ,m
2Rext
(1 + m ,0 )
2Rint
(3.5.7)
y
y
h
In terms of the current at the aperture of the coax (interface coax-waveguide) the current
is I (0) therefore, we may define the input impedance
P + 2 j[WM WE ]
(3.5.8)
Z in
1
2
| I (0) |
2
hence
d
d
*
c2 2
Z in j0 dyI ( y ) dy g ( y | y) I ( y ') + 2 '2 I ( y) .
(3.5.9)
0
0
y
At the end of the probe the current is zero, therefore, for simplicity sake we assume
2
I ( y) y d
(3.5.10)
=
I ( y) =
d
I (0)
or
161
2
y d 2 2c 2
yd d
(3.5.11)
Z in = j0 dy
0 dy g ( y | y)
+ 2 2
0
d
d
d
Let us assume now that except the first mode, all the modes are evanescent. Clearly
their contribution to the impedance is pure imaginary (inductive). Only the propagating
mode has both imaginary and real contribution to the impedance
d
0 d 2
Z in =
sin 2
18 ab c
h
sin ( L ) sin ( L ) + j cos ( L ) 1 + 6 d (3.5.12)
a
c
y
L
k0 d
z
2Rext
2Rint
3
R = 100
R = 70
R = 50
2
X =0
162
/4
/2
3 / 4
10 L
intersection of the two curves determines the required parameters. From the figure it is
seen that the self-resonant length of the antenna, corresponding to L = 0 or / 2 , is
0.280 . Also it is seen that, for k0 d less than 1.22 ( d < 0.61 centimeter), a value of L
which will make X = 0 does not exist. The reason is that the TE 10 mode standing wave
between the antenna and the short-circuiting plunger cannot provide enough inductive
reactance to counterbalance the high capacitive reactance of a short probe.
163
164
z
(a)
(b)
z
(c)
(d)
165
Although surface waveguides have several features which are similar to those of the
cylindrical conducting-tube guide, they have many characteristics which are quite
different. Some of the outstanding differences are:
1. Propagation with no low-frequency cutoff.
2. Finite number of discrete modes of propagation at a given frequency.
3. Phase velocity smaller than that of plane wave in vacuum.
3.7.1 Dielectric Layer above a Metallic Surface
Consider a TM mode and for the sake of simplicity, yet without significant loss of
generality, we shall examine the case when y = 0 , or more specifically TM 01 . This mode
may be derived from the longitudinal electric field
x
r
z
Basic configuration of the open dielectric waveguide.
166
r
z
2
A sin x 2 r k
0< x<d
(3.7.1)
Ez ( x, z ) = exp ( jkz )
2
2
x>d
Bexp ( x d ) k c 2
The x -component of the electric field may be derived using Gauss' law ( E = 0)
jkA
2
c
2
c 2 r k
(3.7.2)
Ex ( x, z ) = exp ( jkz )
2
jkB
2
exp ( x d ) k 2
x>d
2
c
k 2
c2
167
r
z
2
j A
2
0 r
cos x r 2 k
0< x<d
2
c
k
r
c2
H y ( x, z ) = exp ( jkz )
j 0 B
2
2
exp ( x d ) k 2 d < x < .
2
c
k 2
c2
(3.7.3)
Continuity of the tangential electric field implies
2 2
Ez : A sin d r 2 k = B
c
r
2
1
2
Hy :
A cos d r 2 k =
B.
2
2
c
r 2 k2
k2 2
c
c
These two equations lead to the following dispersion relation
168
(3.7.4)
(3.7.5)
tan( ) = r d k
2
where d r
2
c
2
c
(3.7.6)
d r 1 therefore, we
may write
2
2 2 2
2
= d ( r 1) 2 + 2 k d k 2 = 02 2
c c
c
(3.7.7)
tan( ) = r 02 2 .
x
(3.7.8)
The two sides of the equation are plotted in the figure below
169
r
z
<
1
.
r 1
We shall examine what happens at low frequencies i.e. 0, 0 0 . At this limit the
dispersion relation reads
(3.7.9)
2 = r 02 2
The solution of the fourth order polynomial which corresponds to a low-frquency
propagating wave is
02
1 2
2
2
= r + r 1 + 4 2 (3.7.10)
tan ( )
2
r
and now substituting the definitions of and 0 we r 03
tan ( )
get
r 02
2
r 01
(3.7.11)
k 2 = eff
c
which is only weakly dependent on the thickness
(provided d ) but is dependent on the dielectric
01 02 03 3
2
170
2
r2
4
2
r 2 k = 2 1 + 1 + 2 d ( r 1)
2d
c
r c
k =
c2
where the effective dielectric coefficient is unity.
171
r
z
(3.7.12)
kd
r = 3
d = 0.5cm
d r
c
172
r
z
d r ( / c ) k 2 and d k 2 ( / c )
2
2
2
kd
(
)
2
2 x
2 x
=
dx A sin + A
2 cos
0
4
d
d
2
A2 sin(2 ) kd sin(2 )
+ 1 +
= 0 r d 1
8
2
2
0 r
k 2d 2
xd
W
= A sin ( ) dx 1 + 2 exp 2
d
d
4
k 2d 2 d
2 0
2
= A sin ( )1 + 2 .
4
2
(3.7.13)
173
(3.7.14)
2 2
2 2
sin ( 2 )
2 02 r 2 d 2 2 2 x 1
2 0 r d 1
=
1
dx
A
A
d
+
cos
2
2
2
2
d 4
(d )
M
(a)
M
2 2 2
1
d
2
2 0 d
= 0 A sin
4
2 2
(3.7.15)
1.0
0.8
Wd
WT
r = 3
d = 0.5 cm
0.6
W m ,d
WT
0.4
W e ,d
0.2
0
WT
0
10
K i , 0
174
15
20
Exercise 3.14: Show that for a dielectric layer ( r ) of thickness 2d there are even TE
tan = 02 2
(3.7.16)
(3.7.17)
What happens at low frequencies in each case? Calculate the energy ratio in each case.
Exercise 3.15: A metallic "wire'' of radius Rint is coated with a dielectric layer ( r ) the
external radius being Rext . Show that the radial behavior of a TM on mode is described by
B[ J 0 (r )Y0 (Rint ) Y0 (r ) J 0 (Rint ) Rint < r < Rext
Ez =
r > Rext
A K 0 ( r )
wherein = k 2 2 / c 2 and = 2 / c 2 k 2 . Demonstrate that the dispersion
relation is
K1 (Rext )
J1 (Rext )Y0 (Rint ) Y1 (Rext ) J 0 (Rint )
=
.
K 0 (Rext )
J 0 (Rext )Y0 (Rint ) Y0 (Rext ) J 0 (Rint )
Analyze the mode.
175
176
surface, and the axial propagation constant is only slightly larger than k0 . As the radius
increases, the field is confined closer and closer to the rod, and approaches = k0 r
in the limit of infinite radius. Since > k0 , the phase velocity is less than that of plane
waves in free space. In all of the above respects, the dielectric rod does not differ from
the plane-dielectric sheet. Omitting the term exp ( jn j z ) , the field expansion
components are
r<R
r>R
E z = Cn K n ( r )
Ez = An J n (r )
n
n
j
j
Er =
Cn K n (r ) + 2 0 Dn K n (r )
Er =
An J n (r ) 2 0 Bn J n (r )
r
j0
j0
n
n
E = 2 Cn K n (r )
Dn K n (r )
E = 2 An J n (r ) +
Bn J n (r )
r
H z = Dn K n (r )
H z = Bn J n (r )
n
j
A
J
r
Bn J n (r )
(
)
n n
2r
j
n
H =
An J n (r ) 2 Bn J n (r )
r
Hr =
n 0
j
+
(
)
C
K
r
Dn K n (r )
n n
2r
j 0
n
H =
Cn K n (r ) + 2 Dn K n (r )
r
Hr =
177
=
2
2
2
and =
2
2
2
c
c
prime indicates differentiation with respect to the arguments r or r .
Imposition of the boundary conditions at r = R leads to equations for determining
the relative amplitudes of the coefficients, and also the eigenvalue equation. The
eigenvalue equation is
2
J n (u1 )
K n (u2 ) J n (u1 )
K n (u2 ) n (u22 + u12 )
(3.7.18)
2 2
u1 J n (u1 ) u2 K n (u2 ) u1 J n (u1 ) u2 K n (u2 ) k0 u1 u2
where u1 = R, u2 = R . When n = 0 , the right-hand side vanishes, and each factor on
the left-hand side must equal zero. These two factors give the eigenvalue equations for
the axially symmetric TM and TE modes:
K (u )
J 0 (u1 )
(3.7.19)
= 0 2
TM modes
u1 J 0 (u1 ) u2 K 0 (u2 )
J 0 (u1 )
K 0 (u2 )
(3.7.20)
=
TE modes.
u1 J 0 (u1 )
u2 K 0 (u2 )
In addition, u1 and u2 are related by the equation
u12 + u22 =( 1) R .
c
2
178
(3.7.21)
1.5
= 2.56
1.4
/ k0
1.3
TE 1
TM 1
1.2
1.1
Comments:
HE11
1.0
0.613
179
J1 02 2
J 0
2
0
wherein 0 =
2
0
2
1
1 1 K1 ( )
.
r K 0 ( )
(3.7.22)
c
several phase velocities reads
1 1 1 K1 ( )
(3.7.23)
=
.
2 r K 0 ( )
The solution denoted by c ( r ) implies
2
2 c2 ( r )
2
2 2
2
(3.7.24)
c = R k 2 k = 2 +
2
c
c
R
and since for example for r = 3, the solution is c = 0.963 , we conclude that there is no
solution which is consistent with the assumption 0 . Consequently, there is no
solution at low frequencies.
180
181
(3.8.1)
= 0
+ s (t t ) (r r')
t
where t is a unit vector giving the direction of the current element, and r designates the
field point ( x, y , z ) , while r' designates the source point or location. If the Laplace
transform of these equations is taken, we get
E(r, s ) = 0 sH (r, s )
(3.8.2)
H (r, s ) = 0 sE(r, s ) + se st (r r')
with a similar definition for H(r , s ) . These equations are formally the same as those
obtained by assuming a time dependence e jt , and the solution may be obtained by the
methods we have previously discussed. All our previous solutions may be converted into
jt
1/2
2
2
2
a n =1
(n /a ) + s / c
where n ( x ) = sin( n x / a ) . The inversion of this expression gives us the time-dependent
Green's function corresponding to the disturbance set up in the waveguide by an impulse
line current located at x , z . A typical term from (3.8.3) gives
1
2 j
1/2
} ds
exp s (t t ) (n /a ) 2 + s 2 / c 2 | z z |
1/2
(n /a ) 2 + s 2 / c 2
n 2
| z z |
2
2 1/2
cJ
c
t
t
z
z
t
t
[
(
)
(
)
]
0
<
<
(3.8.4)
= 0 a
c
0
otherwise
where J 0 is the Bessel function of the first kind. At any given distance | z z | from the
source, the disturbance is zero until a time t = t + | z z | /c is reached when the presence
of the signal first becomes known to the observer at this position. No information reaches
the observer in a time interval less than the time required to propagate a disturbance with
183
the velocity of light. The velocity of light is, therefore, the wavefront velocity.
The time derivative of the above function is
c 3 (n /a )(t t )
n 2
2
2 1/2
(3.8.5)
J
[
c
(
t
t
)
(
z
z
)
]
1
[c 2 (t t ) 2 ( z z) 2 ]1/2 a
where J1 is the Bessel function of the first kind and order 1. The solution for the timedependent Green's function which is equal to y (r , t ) becomes
J1 {(n /a )[c 2 (t t ) 2 ( z z) 2 ]1/2 }
1 n
(3.8.6)
(r | r', t t ) =
n ( x)n ( x) [ c(t t ) ]
2
2
2 1/2
0 a n =1 a
[c (t t ) ( z z ) ]
for t < t < t + | z z | /c and zero otherwise.
184
The goal of this sub-section is to determine a general relation between power, energy and
impedance at the input of an electromagnetic device. We shall consider the power in a
volume V and an envelope S ; through this envelope an average power P crosses
1
(3.9.1)
P = Re E H * ds .
2
Thus based on the complex Poynting theorem
1 *
1 * 1 * 1 *
(3.9.2)
( E H ) ds = 2 j dV B H E D E J dV .
V
V
2
4
4
2
Assuming linear medium characterized by = j , = j and we can
separate the real and imaginary part of this theorem
1
1
1 *
1
2
2
2
Re
=
2
|
|
|
|
|
|
E
H
ds
dV
E
dV
V 4
4
2
2
1 *
2 2
(3.9.3)
Im
E H ds = 2 dV | H | | E | .
V
4
2
185
2
2
2
jC
where PL = R | I |2 / 2 represents all losses in the system, WM = L | I |2 / 4 represents the
average magnetic energy stored in the system whereas WE = C | V |2 / 4 represents the
average electric energy stored in the system hence
1 * 1
(3.9.5)
VI = Z | I |2 = PL + 2 j(WM WE )
2
2
or
P + 2 j[WM WE ]
(3.9.6)
Z= L
.
1 2
|I|
2
This can be conceived as a generalization of the impedance concept.
186
d
z
2 n m
(3.9.7)
nm = r r 2
.
c a b
187
c
1
n m l
(3.9.9)
f n ,m,l =
+
+ .
2 r r a b d
We shall examine next electromagnetic field characteristics in a cavity. In particular we
shall calculate the quality factor of a TE101 . Choosing d > a > b , this is the lowest mode.
It has the following field (non-zero) components
x
j z
j z
H z Ae
cos
=
+ Be
a
j a j z
j z x
(3.9.10)
Hx
Ae
sin
=
Be
j a
x
Ey =
r r 0 Ae j z + Be j z sin
a
c
Bearing in mind that E y vanishes at z = 0 we obtain A + B = 0 and applying the
j d
j d
same condition at z = d we get Ae
Be
= 0 2 jA sin( d ) = 0 therefore
for a non trivial solution has to satisfy d = l which is just (3.9.8).
188
a d
a
x
z
H x ( x, z; ) = 2 jA sin
cos
d
a
d
1/2
(3.9.11)
2
a a 2
x z
E y ( x, z ; ) =
sin
2 A
+
sin
a
d d 2
d
The electric and the magnetic energy stored in the system
2
2
b
d
1 a
1
2 a d
2
2 a
2
WE = o =
dx dy dz| E y |
o | 2 A | 2 2 + 2 b = WM
0
0
0
4
4
d 2 2
a
Clearly at resonance the electric energy equals the magnetic energy stored in the system
therefore the total average energy stored in the cavity is
a2 2
1
(3.9.12)
WT = 2WE = o (bad ) 1 + 2 | A | .
d
2
Next we shall calculate the loss. The surface current is given by J s = n H therefore the
dissipated power is given by
189
Ploss
| H x |2 Rs
J2 1
1
Js 1 1
= da
= da
= da
= da | H x |2
2
2
2
2
=| A |2 Rs ( 2a 3b + 2d 3b + ad 3 + da 3 ) / d 2
(3.9.13)
(3.9.14)
Q=
=
=
.
2
3
3
3
3
Ploss
Ploss
Rs 2 (2a b + 2d b + a d + d a )
As an example consider the following parameters
Cu 5.8 107 [ Ohm/m ] , a = b = d = 3[ cm ] f = 7.0 [ GHz ] , T = 1/ f 0.14[ nsec]
Rs 0.022, Q 12700, = 2Q / 0 = Q / f 0.58[ msec].
In the case of dielectric losses
1
2WE
1
1 1
2
Ploss,d = dV | E | Qd =
=
=
+
V
2
Ploss ,d
Qeff Qd Q
(3.9.15)
since the total power loss is the sum of the two mechanisms.
Exercise 3.18: Check whether (3.9.13) is correct. In particular check if H z does not
190
2R
H z = J1 p11
cos
p11
= +
(3.9.18)
.
c d R
191
Exercise 3.19: Prove that the quality factor for a TE mode is given by
2
n 2
l
2
) + R
1
( pnm
d
pnm
1
(TE )
Qnml =
2
2
R
l
R
nl
R
2
s ( p ) + 2 R + 1 2
nm
c
d d
d pnm
d
For a TM mode the quality factor is given by
2 1/2
l
2
pnm
+ R
d
l>0
R
1
(TM )
1+ 2
=
Qnml
d
0
s
pnm
c
l=0
R
1+
d
3/2
(3.9.19)
(3.9.20)
192
In many cases the cavity relies on evanescent waves that decay in one direction and they
do not carry any power in that direction. One example is illustrated below: A cylindrical
waveguide of radius R with a short-circuiting plane at z=0, a dielectric ( r ) of thickness
d.
Exercise 3.21: Determine the resonant frequency and the
193
Coaxial
line
Coaxial
line
(a)
(b)
Cavity
2r0
z=0
(a)
jX
194
Zc =1
z=d
(b)
8 r02
In case of a TE 10 the normalized impedance of the aperture is X L =
r0
3 ab
revealing that this is proportional to the areas ratio. For low loss
( j X L ) j tan( L)
(3.9.23)
Z in
j X L + j tan + tanh( d )
and the condition for resonance
(3.9.24)
( ) X L + tan = 0
implies that at resonance ( = 0 )
2
XL
.
Z in ( = 0 )
d
In the vicinity of the resonance
Z in ( )
d + j
(3.9.25)
2
L
( 0 )
=
Z in ( = 0 )
j
1+
( 0 )
d =
(3.9.26)
Since for an RLC circuit the impedance in the vicinity of the resonance is
1
Z in ( ) Z in ( = 0 ) [1 + j ( 0 ) / ] we conclude by analogy that the bandwidth is
=
d / ( = 0 ) implying that the quality factor Q= 0 / = 0 ( = 0 ) / d .
195
X L =Ud
( = 0 ) =
=
1
[U d + tan( d )] =
d U +
cos d =0
(3.9.27)
d
d
U + 1 + tan 2 ( d )
=
[U + 1 + (U d ) 2 ]
=0
Vgr
Vgr
8 r02 r0
where U =
196
The goal of this section is to analyze a wave guide of a circular cross-section but which
lacks a slice as illustrated below
In order to keep the problem relatively simple, we shall limit the discussion to the
evaluation of the cut-off frequency. For this purpose we recall that in Section 2.1 we
197
found that
jkz
j
E = 2 Ez + 2 1z H z
k
k
jkz
j
H = 2 H z 2 1z Ez
k
k
(3.10.1)
k2 2 k 2
=
And both longitudinal components satisfy the wave equation. Limitint the analysis to the
cut-off frequencies k z = 0 these equations simplify
1 1
=
E
Hz
r
j r
1
1z H z
E =
j
E = 1 H
z
j r
(3.10.2)
1
1
Hr =
Ez
j r
1
H = +
1z Ez
j
H = 1 E
j r z
and the wave equation simplifies
198
1 1 2
Ez
2
(3.10.3)
0
r + 2 2 + =
r
r
r
r
z
Obviously, the solution of this set of equations describe a resonator
whereby the field does not vary in the z-direction.
For the TM modes the solution is of the form
Ez ( r , ) J ( qr ) A cos ( ) + B sin ( )
subject to the boundary conditions
E=
R, / 2 < < 2 =
/ 2) 0
z (r
Ez =
/ 2) 0
( r , =
(3.10.4)
(3.10.5)
Ez ( r , =
2 / 2 ) =
0
The first condition implies that
(3.10.6)
J ( qR ) = 0
wherein q is yet to be determined and from the other two conditions we will specify :
/ 2) =
Ez ( r , =
0 A cos ( / 2 ) + B sin ( / 2 ) =
0
(3.10.7)
Ez ( r , =
2 / 2 ) =
0 A cos ( 2 / 2 ) + B sin ( 2 / 2 ) =
0
or explicitly
199
A
cos ( / 2 )
sin ( / 2 )
(3.10.8)
= 0
cos
/
2
sin
/
2
)
) B
(
(
2 /
Jn( 1 , x)
0.5
Jn( , x)
Jn( 2 , x)
0
0.5
200
5
x
10
Explicitly
Ez ( r , ) =
pn , s : J
n , s =1
n,s J
1
2 /
r / 2
p
n , s sin n
R
2
( p) = 0
2 /
(3.10.11)
Ez
(3.10.12)
The top frame in the right reveals the contours of constant Ez for
=
s 1and
=
n 1. The central frame illustrates the case
=
s 1and
=
n 2
whereas the lower frame corresponds
to s 2and
=
=
n 1.
Ez2
201
Ez3
(3.10.13)
(3.10.14)
2 / 2 ) =
0
Er ( r , =
The first condition [Eq. (3.10.2)] implies that
d
(3.10.15)
J
qr
(
)
dr
= 0
r=R
wherein q is yet to be determined and from the other two conditions we specify :
Er ( r , = / 2 ) = 0 A sin ( / 2 ) + B cos ( / 2 ) = 0
(3.10.16)
Er ( r , = 2 / 2 ) = 0 A sin ( 2 / 2 ) + B cos ( 2 / 2 ) = 0
or explicitly
A
sin ( / 2 )
cos ( / 2 )
(3.10.17)
= 0
sin
/
2
cos
/
2
B
(
)
(
)
n
1
(3.10.18)
= n
2
2 /
wherein
=
n 1, 2,3.... . Now that we have determined we may proceed and establish
the zeros of the Bessel function:
(3.10.19)
J 1 ( p) = 0
2 /
H z ( r , ) =
pn , s : J
U n,s J
n , s =1
( p) = 0
1
2 /
r
/ 2
p
cos
n,s
n
R
2
(3.10.20)
2 /
Exercise 3.22: Draw the contours of constant H z . Compare the field distribution of the
TE and TM modes. For both modes compare the magnetic and electric energy per-unit
length.
203
2R
r = 1
r = 1
( r , r )
Vacuum
Slice
1 1 2 2
1 1 2 2
0
r
+
+
E
=
r
+
+
r r Ez =
2
2
2 z
2
2
2
r
r
c
r
r
r
1 1
1 1
Hr =
Ez
Hr =
Ez
j0 r
j0 r r
1
Ez
H =
Ez
j0 r
j0 r r
A differential approach (separation of variables) is not applicable since the radial
H =
204
variation needs to be the same in both regions but if this is the case the function does not
satisfy the wave equation. We must seek an integral approach. For determining the cutoff frequencies associated with the TM-like modes we need to solve
1 1 2 2
2
(3.10.21)
2 f ( ) E z
r + 2 2 + 2 Ez =
c
c
r r r r
wherein f ( ) is a function which equals r r 1 in the region(s) with material and zero
otherwise. A solution which satisfies the boundary conditions
(3.10.22)
Ez ( r , ) = n , s J n pn , s exp ( jn )
R
n = s =1
substituting in (3.10.21) we get
pn2, s 2
2
s ,s
n , s n ,n 2 + 2 =
2 m , m ,,ns Fn ',m
c
c m = =1
R
s ,s '
n ,n '
s ,s '
n ,n
1
Fn ',m
=
2
s , s ' dx xJ n ( pn , s x ) J n ( pn , s x ) (3.10.23)
=
d f ( ) exp j ( n ' m )
205
Rewriting
R
c
m =
=1
,s
s ,s 2
F
+
=
m , m , n n , m
n ,m , s
n , s n , n pn , s
(3.10.24)
thus defining = R / c .
C{
{m , }
m ,,ns
p
s ,s
n ,n
2
n,s
Fn ,m + n ,m , s
n , s},{m , } {m , }
I{
{m , }
{m , }
n , s},{m , }
(3.10.25)
(3.10.26)
implying that the (square of the ) cut-off frequencies are the inverse of the eigen-values
of the matrix C
(3.10.27)
0.
C 2I =
Exercise 3.23: Calculate the first cut-off frequency of a waveguide with dielectric slice
identical with the metallic slice. Compare the two results. Examine the convergence of
the solution.
Exercise 3.24: Repeat Exercise 3.23 for the TE mode including the formulation.
206
3.11 Appendix
3.11.1 Solution to Exercise 3.8
0.6
XL
0.5
0.4
0.3
10
f [GHz ]
207
6
10[GHz]
XL
8[GHz ]
2
0
10
15
20
d [mm]
208
209
1.5
10[GHz ]
XL
1.0
0.5
8[GHz ]
0.0
0.5
1.0
1.5
R[mm]
210
Chapter 4
Matrix Formulations
4.1 Impedance Matrix
Consider an N port systems that is characterized at a given frequency by set of complex
numbers Zij i.e. 2N 2 parameters
V1
Z11 Z1N I1
V
2
=
(4.1.1)
VN
Z N 1 Z NN I N
or, explicitly, Z ij =
Vi
Ij
.
Ik j =0
(4.1.2)
V
1
I1
V2
I2
IN
211
VN
I
Y
N
N 1 YNN VN
Example: For the sake of simplicity, let us determine the Z matrix of a section of
transmission line.
This two port system is characterized by
V1
Z11 Z12 I1
(4.1.4)
=
V
Z
I
Z
2
21
2
22
and for example the first term is
V
Z11 = 1
(4.1.5)
I1 I =0
2
I2
I1
V1
Z0
V2
212
Z 0
V ( z ) = V e
j z
I ( z ) = V e
j z
V e
j z
V e
j z 1
Z
(4.1.6)
Z o
hence
2 j
1 e
cos( )
= Zo
= jZ o ctan( )
Z11 = Z o
2 j
sin
j
1 e
Z11 = jZ o ctan( ).
(4.1.7)
(4.1.8)
(4.1.9)
(4.1.10)
(4.1.11)
213
In a similar way we may establish the other parameters of the matrix. For example,
V
V V
Z12 = 1
= Z0
(4.1.12)
j
j
I 2 I =0
V e
() V e
1
1
and since I1 = V V V = V we get
Zo
2
jZ 0
(4.1.13)
Z12 = Z 0
=
j sin( )
j
e
e
jZ
ctan(
sin
Z=
.
jZ
o
jZ o ctan( )
sin
214
(4.1.14)
V1
V1
V2
V N
V2
v1
S11 S1N
v 2
vN
N 1 S NN
v1
v 2
vN
v1
v1
v
v1
Zo
Zo
the reason for this normalization becomes evident since
1
V (V )* 1 *
2
P=
V =
= v (v ) .
2Zo
2Zo
2
215
(4.2.1)
(4.2.2)
(4.2.3)
V N
Sij =
vi
v v
k j =0
(4.2.4)
v k 0
Z0
Z0
v k
v = v v and I = v v .
Using now the definition of the Z matrix
v = v v = Z v v
216
(4.2.5)
[ I Z ]v = [ I Z ]v
thus since v = Sv
(4.2.6)
S = ( Z I )1 ( Z I ).
(4.2.7)
(4.2.8)
1
| v n |2 i.e. characterization
2
impedance at all ports is the same. In order to prove this statement we recall that based on
Lorentz reciprocity theorem (in absence of sources)
(4.2.9)
E
H
E
H
nds
=
0
E
H
da
=
E
H
da
1 2 2 1
i j j i
and since the electric field is associated with the voltage whereas the local magnetic
field is linked to the current E1 V1 , H 2 I 2 thus
V i I j = Vj I j.
(4.2.10)
217
v v v v 0 v
I v Sv
Sv 0
v I v v S t S * v 0 v
I S S v
t
218
(4.2.13)
(4.2.14)
v1 S
= 11
v 2 S 21
S12 v1
S 22 v 2
(4.2.15)
Z Zo v 2
= .
SL = L = L
Z L Z o v 2
Combining the two entails
(4.2.16)
(4.2.17)
(4.2.18)
v1
v2
ZL
v1
219
v2
(4.3.1)
V1 = Z11I1 Z12 I 2
(4.3.2)
V2 = Z 21I1 Z 22 I 2
Thus
Z11
V1 Z12
I =
1
1
Z12
Z11Z 22 Z122
V
Z12
2
I 2
Z 22
Z12
Det{ ABCD} 1
(4.3.3)
(4.3.4)
I1
I2
v2
v1
220
v1 T T v 2
= 11 12
v1 T21 T22 v 2
Relation to S matrix
S S
S
S
1
T11 = S12 11 22 , T12 = 11 , T21 = 22 , T22 =
S 21
S 21
S 21
S 21
In a reciprocal system S is symmetrical and S12 = S21 . Therefore
S
Det(T ) = 12 = 1
S21
(4.4.2)
(4.4.3)
v1
v1
(4.4.1)
v2
221
v2
v1 S11
=
v2 S 21
S12 v1
S 22 v2
(4.4.4)
v1
v2
v1
v2
j
j
S11 =
= 0, S 22 =
= 0, S12 =
=e
, S 21 =
=e
v1 v =0
v2 v =0
v2 v =0
v1 v =0
2
1
1
2
(4.4.5)
0
S =
j
e
j
e j
T =
0
v1
v1
j
e
(4.4.6)
Z0
222
v2
v2
1 T
* T
(
)
Pdis = v
I
S
S
v
2
the matrix defined above determines this dissipated power
1 T
Pdis = v
Qv .
2
223
(4.5.1)
(4.5.2)
(4.5.3)
input
direct
1
the
other are loaded with Z o
The power entering (1) splits
equally between (3) and (4).
No power reaches (2).
insulated
Performance parameters:
Coupling
: C = 10log P1 / P4
Directivity : D = 10log P4 / P2
S matrix:
Isolation condition : S12 = S 21 = S34 = S 43 = 0
Matching condition : S11 = S 22 = S33 = S 44 = 0
Symmetrical condition : S13 = S31 , S 23 = S32 , S14 = S 41 , S 24 = S 42
0
0
S
S13
S14
0
0
S13
S 23
S 23
S 24
0
0
S14
S 24
0 224
coupled
We shall discuss several aspects of the coupling effect. Here we shall investigate
coupling in transmission line configuration.
4.7.1 Double Strip Line: Parameters of the Line
x=a
(1)
(2)
a/2
x=0
z=0
225
x
z
The basic configuration under consideration is illustrated above. Each one of the wires is
charged thus the charge distribution in the system is assumed to be
a
a
1 ( x, z ) = 1 x z , 2 ( x, z ) = 2 x z
2
2
2
2
nx
( x, z ) = n ( z )sin
n =1
Which satifies the boundary conditions on the two plates. This potential satisfies the
Poisson equation i.e.,
2
2
1
(
,
)
=
x
z
( x, z )
2
2
x
z
o
d 2 n 2
1 2
nx
z
dx
x
z
(
)
=
(
,
)sin
2
o a
a
dz a
n
1 2
=
z
sin
2 2
o a
where = 1,2 and 1 = , 2 = .
226
(4.7.1)
1
n
n
nx
n
( x, z ) = sin
sinc
exp
exp
z
2
(4.7.2)
2 n =1 a
a
2
a
2
2
n
1
d
)exp
z
1
a
1
nx
2 2
( x, z ) = sin
(4.7.3)
sinc n
2 n =1 a
2 2
1
d 2 ( )exp
z
2 2
n
2 n =1 a
2
2 2
(4.7.4)
2 2
z
1 d 2 d exp
a
2 2
2 2
227
2 2
a
dz
(
x
, z)
2
2 2
= 1
2
2 2
1
n
n
n
dz
d
z
sinc
exp
2
2
sin
n =1
2 2
2 2
2
2
2 2
2 2
2 2
2 2
2 2
1
n
sin
sinc
n
dz
2
2
n =1
(4.7.5)
d
exp
z
sinc
1
exp
sinhc
2
2 n =1
2a
2a
2
(4.7.6)
n
1a
n n
2 n
sinc
sinhc
exp
a
2 n =1
2 2 a
2a
Similar considerations can be applied to the second strip hence V2 = U 211 U 222
U12 =
228
and detailed calculation shows that U 21 = U12 and U 22 = U11 . Note that the off
diagonal term is proportional to
(4.7.7)
exp
n
a ( )n
which indicates that the coupling drops rapidly. Furthermore, U11 are independent!
If the strips have different geometries U11 = U 22 but still U12 = U 21 . In a matrix form:
V1 U11 U12 1 1 C11 C12 V1
=
=
(4.7.8)
V U
2 21 U 22 2 2 C12 C22 V2
A similar approach can be applied for evaluation of the inductance
1 L11 L12 I1
=
(4.7.9)
L
2 12 L22 I 2
Exercise 4.1: Analyze the coupling term.
Exercise 4.2: Calculate the inductance matrix and analyze the coupling term.
Exercise 4.3: Analyze the case when the two strips are not equally wide.
Exercise 4.4: Analyze the case when the two strips are not equally wide and are not in
the same plane.
229
d
V = j LI
= j
dz
L
dV2
12 L22 I 2
dz
(4.7.10)
dI1
dz
C11 C12 V1
d
I = jCV
= j
dz
C
dI 2
12 C22 V2
dz
It is convenient in this case to investigate the waves which propagate in this system
d2
dI
(4.7.11)
V
=
L
= j L( j )CV = 2 LCV
2
dz
dz
Let us define the wavenumber matrix
230
K112 K122
2
K 2 LC =
2
2
K12 K 22
j z
Assume a solution of the form e
2
0 K112
2
( 2 I K )V = 0
2
K122
0
(4.7.12)
K122 V1
= 0
K 222 V2
(4.7.13)
K112 2
K122
V1
2
2
2
2
4
=
0
(
K
)(
K
)
K
(4.7.14)
11
22
12 = 0
K 222 2 V2
Let us assume for simplicity sake that the two lines are identical i.e. K11 = K 22 hence
(4.7.15)
( K112 2 ) 2 K124 = 0 2 = K121 K122
The eigen-vectors corresponding to 2 may be derived from one of the two equations
K122
231
K112 2
eg., ( K )V1 K V = 0 threfore choosing V1 = 1 we get V2 =
= 1 implying
K122
that the normalized egen-vector is
2
11
2
12 2
1 / 2
V =
1 / 2
V =
1/ 2
232
(4.7.16)
(4.7.17)
(4.7.18)
j z
j z
V ( z) =
Ae
Be
1 / 2
1 / 2
(4.7.20)
1 / 2 Ae j z 1 / 2 Be j z
1 / 2 Ae j z 1 / 2 Be j z
233
1
0
jK11z
jK11z
(4.7.21)
V ( z ) = C e
D e
0
1
Now since V ( z = 0) = V0 for line 1 and V ( z = 0) = 0 for line 2 we get C = V0 and D = 0
implying that the voltage on line 2 is always zero.
Let us reinstate the coupling process: the voltage along the transmission lines is
given by
V0 A/ 2 B/ 2
V0
=
=
V ( z = 0) = =
A
B
(4.7.22)
2
0
/
2
/
2
A
hence
1
1
V
j
exp
cos
z
e j z e j z o
2
2
Vo
=
V ( z) =
e j z e j z
1
Vo exp j z j sin z
2
2
In order to illustrate the coupling process, we examine the voltage on each one of the
234
lines
1
L =
2
2
the voltage is maximum on line 2 and zero on
line 1. Recall = K112 K122 . For simplicity we
shall assume K112 K122 hence
1 K122
= K11
2 K11
and
235
(4.7.23)
(4.7.24)
in the right. If we
(4.7.25)
(4.7.26)
1
1
1 K122
1 K122
1 K122
L
K11
L K11
L =
2
2
2 K11
2 K11
2 K11
(4.7.27)
K
L = 211
=
2
K12
This clearly indicates that the coupling is directly related to the off diagonal term in the
2
matrix K .
Exercise 4.5: Analyze L as a function of the geometry (separation , width and
planes separation a ) in the previous section. Make sure that the result is independent of
the number of harmonics.
236
Chapter 5
Nonlinear Components
5.1 Transmission Line with Resistive Nonlinear Load
237
1 2
V (t ).
Z0
1 z
z
I ( z, t ) =
V
t
V
t + .
+
Z0 v
v
At the location of the load
=
VL (t ) V ( z = 0, t ) = V+ (t ) + V (t )
1
[V+ (t ) V (t )]
Z0
thus after substituting in the constitutive relation
1
[V+ (t ) V (t )].
V+ (t ) + V (t ) = V0 tanh
Z0 I 0
It is convenient to define
V
V
V =
and Z L = 0
Z0 I 0
Z0 I 0
hence
V + + V = Z L tanh(V + V ).
=
I L (t ) I ( z = 0, t ) =
238
(5.1.3)
(5.1.4)
(5.1.5)
(5.1.6)
(5.1.7)
10
V+
Z L = 15
5
Z L = 10
ZL =8
0
-5
ZL =4
Z L = 0. 8
V +
-10
-3
-2
-1
Another aspect of interest is the energy balance. The incident and reflected energy may
be evaluated by
239
1 2
1
V+ (t ), W = dt V2 (t ).
Z0
Z0
In a linear system the ratio of the two is
2
W Z L 1
W
=
W+ Z L + 1
wherein Z L = Z L / Z0 . The figure in the right illutrates W
when the maximum value of the incident (Gaussian) wave
is V+( max ) = 2,5,10 . This figure clearly reveals that there is
full reflection when Z L is zero or tends to infinity. Zero
reflection occurs when
(max )
V0
V+(max )
V+(max ) = V0 .
=
ZL =V+
Z0 I0
Z0 I0
W+ = dt
(5.1.8)
(5.1.9)
1.0
Energy Ratio
0.8
Linear load
0.6
V max = 2
0.4
V max = 5
0.2
0
0
V max = 10
5
10
ZL
240
15
20
Chapter 6
Periodic Structures
In a variety of applications it is necessary to control the phase velocity of the wave
or the propagation time. Periodic metallic or dielectric structures play an important role in
implementation of these devices. The periodic geometry can be conceived as a set of
obstacles delaying the propagation of the wave due to the multi-reflection process and, as
will be shown during our discussion on Floquet's theorem, an infinite spectrum of spatial
harmonics develops. A few of which cases these harmonics may propagate with a phase
velocity larger or equal to c but the absolute majority have a smaller phase velocity.
In the first section we present the basic theorem of periodic structures namely,
Floquet's theorem. This is followed by an investigation of closed periodic structures in
Section 6.2 and open structures in the third. Smith-Purcell effect is considered as a
particular case of a Green's function calculation for an open structure and a simple
scattering problem is also considered. The chapter concludes by presenting a simple
transient solution in a periodic structure.
241
j 2 nz / L
(6.1.2)
.
f ( z) = fn e
n =
The amplitudes f n are determined by the value of the function f ( z ) in a single cell.
j 2 mz / L
Specifically, we multiply (6.1.2) by e
and integrate over one cell i.e.,
L
L
j 2 nz / L
j 2 mz / L
j 2 mz / L
(6.1.3)
(
)
=
.
dzf
z
e
dze
f
e
n
n =
242
different cells is identical. It can not describe a propagation phenomenon, thus it can not
represent a dynamic system. In the latter case the function f ( z ) has to satisfy
(6.1.5)
f ( z ) = f ( z + L),
which means that the value of the function is proportional to the value of the function in
the adjacent cell up to a constant, , whose absolute value has to be unity otherwise at
z the function diverges or is zero as can be concluded from
(6.1.6)
f ( z ) = n f ( z + nL),
where n is an arbitrary integer. Consequently, the coefficient can be represented as a
phase term of the form = exp( j ) hence
j
(6.1.7)
f ( z ) = e f ( z + L);
is also referred to as the phase advance per cell. Without loss of generality one can
redefine this phase to read = kL . Since a-priori we do not know , this definition does
not change the information available. Nonetheless based on the Fourier series in (6.1.2)
we can generalize the representation of a dynamic function in a periodic structure to
j 2 nz / L jkz
(6.1.8)
,
f ( z) = fn e
e
n =
jkL
f ( z + L),
243
(6.1.9)
which is identical with the expression in (6.1.7). The last two expressions are different
representations of the so-called Floquet's Theorem. Later we shall mainly use the form
presented in (6.1.8), however in order to illustrate the use of Floquet's theorem in its latter
representation, we investigate next the propagation of a TM wave in a periodically loaded
waveguide.
L
2R
244
z
+ z
r
Az (r , z ) = J 0 ps As e d , s + Bs e d , s ,
R
s =1
+d , s z
c 2 ps
r d , s z
J
,
Er ( r , z ) =
p
A
e
B
e
d ,s 1 s
s
s
j r s =1 R
R
+d , s z
r d , s z
2
J
,
p
A
e
B
e
d ,s 0 s
s
s
j r s =1
R
+d , s z
1 ps r d , s z
(6.1.10)
J
H (r , z ) =
p
A
e
B
e
+
s
,
0 s =1 R 1 s R s
where 2d , s = ( ps / R )2 r ( / c )2 . In a similar way, we have in the vacuum ( g < z < L):
Ez (r , z ) =
c2
245
( z g )
+ ( z g )
r
+ Ds e s
,
Az (r , z ) = J 0 ps Cs e s
R
s =1
c 2 ps
+s ( z g )
r s ( z g )
J
,
Er (r , z ) =
p
C
e
D
e
s 1 s R s
s
j s =1 R
c 2 2 r s ( z g )
+ ( z g )
s J 0 ps C s e
+ Ds e s
,
Ez (r , z ) =
j s =1
R
H (r , z ) =
s =1
ps r s ( z g )
+ ( z g )
J1 ps Cs e
,
+ Ds e s
R R
(6.1.11)
with 2s = ( ps / R )2 ( / c )2 . At this point we shall consider only the TM 01 mode
( s = 1 ) thus the continuity of the radial electric field at z = g implies
g
+ g
1
(6.1.12)
d ,1 A1e d ,1 B1e d ,1 = 1 [C1 D1 ],
r
and in a similar way the continuity of the azimuthal magnetic field reads
g
+ g
A1e d ,1 + B1e d ,1 = C1 + D1 .
(6.1.13)
Last two equations express the relation between the amplitudes of the field in the
dielectric and vacuum.
246
In the dielectric filled region of next cell ( L z L + g ) the field has a similar form
as in (6.1.10) i.e.,
+ ( z L)
r ( z L)
Az (r , z ) = J 0 ps As' e d , s
+ Bs' e d , s
,
R
s =1
c2
r ' d , s ( z L)
' +d , s ( z L )
Er (r , z ) =
p
R
J
p
A
e
B
,
s
d ,s 1 s
se
s
j r s =1
R
r ' d , s ( z L)
' +d , s ( z L )
2
p
A
e
B
,
d ,s 0 s
se
s
j r s =1
R
1
r ' d , s ( z L)
' +d , s ( z L )
+
H (r , z ) =
p
R
p
A
e
B
J
se
0 s =1 s 1 s R s
(6.1.14)
Accordingly, the boundary conditions at z = L read
1
( L g )
( L g )
(6.1.15)
d ,1 A1' B1' = 1 C1e 1
D1e 1
,
and
( L g )
( L g )
(6.1.16)
A1' + B1' = C1e 1
.
+ D1e 1
Ez (r , z ) =
c2
247
The relation between the amplitudes of the wave in the second cell ( L < z < 2 L ) and
the first cell can be represented in a matrix form
(6.1.17)
a' = Ta,
where the components of a' are A1' and B1' and similarly the components of a are A1 and
B1 . According to Floquet's theorem (6.1.9) the two vectors are expected to be related by
jkL
(6.1.18)
a' = e
a,
jkL
thus e
represents the eigen-values of the single cell transmission matrix T :
jkL
(6.1.19)
|Te
I |= 0.
Explicitly this reads
2 jkL jkL
(6.1.20)
e
e
(T11 + T22 ) + T11T22 T12T21 = 0.
For a passive system the determinant of the matrix T is unity (prove this statement!!),
thus
2 jkL jkL
(6.1.21)
e
e
(T11 + T22 ) + 1 = 0.
The fact that the last term in this equation is unity indicates that if k is a solution of
(6.1.21) k is also a solution. Consequently, we can write
1
(6.1.22)
cos( kL) = (T11 + T22 ).
2
248
2
1
0
Note that this is an explicit expression for k as a function of the frequency and the other
geometric parameters. In principle there are ranges of parameters where the right-hand
side is larger than unity and there is no real k which satisfies this relation. If only the
frequency is varied then this result indicates that there are frequencies for which the
solution of the dispersion relation is real thus a wave can propagate, or the solution is
imaginary and the amplitude of the wave is zero. The frequency range for which the wave
is allowed to propagate is called the passband. Explicitly, the right-hand side of (6.1.22)
reads
( Z1 + Z 2 )2
( Z1 Z 2 )2
1
(T11 + T22 ) =
cosh( + )
cosh( ),
2
4 Z1 Z 2
4 Z1 Z 2
(6.1.23)
where = 1 ( L g ) , = d ,1 g , the characteristic impedances are
0 c d ,1
c
(6.1.24)
Z1 =
,
Z2 = 0 1 ,
j r
j
and 0 = 377 is the impedance of the vacuum.
-1
-2
0 5 10 15 20 25
f (GH z )
249
20
f (GHz )
250
j (k0 +2 / L) L
jk0 L j 2
,
e
jk L
(6.1.25)
= f ( z + L)e 0 = f ( z ).
Consequently, since the dispersion relation is periodic in k , it is sufficient to represent
its variation with k in the range / L k / L ; this k domain is also called the first
Brillouin zone.
Comment 2. Bearing in mind the last comment, we can re-examine the expression in
(6.1.8) and realize that f ( z ) is represented by a superposition of spatial harmonics
exp( jkn z ) where
2
(6.1.26)
kn = k +
n,
L
which all correspond to the solution of the dispersion relation of the system. According
to this definition the phase velocity of each harmonic is
f ( z + L )e
vph , n =
= f ( z + L )e
(6.1.27)
ckn
and for a high harmonic index, n , this velocity decreases as n 1 . Furthermore, all
harmonics with negative index correspond to backward propagating waves. In addition,
note that the zero harmonic ( n = 0 ) has a positive group velocity for / L > k > 0 and
negative in the range / L < k < 0 . This is a characteristic of all spatial harmonics.
251
Since the group velocity is related to the energy velocity, one can conclude that although
the the wave number of a particular space harmonic is positive, the power it carries may
flow in the negative direction (if the group velocity is negative).
252
d
6.2 Closed Periodic Structure
2R
2R
int
ext
Based on what was shown in the previous section
L
one can determine the dispersion relation of a TM 01
mode which propagates in a corrugated waveguide
[Brillouin (1948)]. Its periodicity is L , the inner
radius is denoted by Rint and the external by Rext ; the
distance between two cavities (the drift region) is d . Using Floquet's Theorem (6.1.8)
we can write for the magnetic potential in the inner cylinder ( 0 < r < Rint ) the following
expression
jkn z
(6.2.1)
I0 ( n r ),
Az ( r , z ) = An e
n =
253
c2
jkn z
(
)
I1 ( n r ),
Er ( r , z ) =
jk
A
e
n n n
j n =
c2
jkn z
( 2n ) An e
I0 ( n r ),
Ez (r , z ) =
j n =
H (r , z ) =
n Ane
0 n =
jkn z
I1 ( n r )
(6.2.2)
In these expressions,
=k
2
n
2
n
2
2
(6.2.3)
c
and I0 ( x ), I1 ( x ) are the zero and first order modified Bessel functions of the first kind
respectively. This choice of the radial functional variation is dictated by the condition of
convergence of the electromagnetic field on axis.
In each individual groove (superscript ) the electromagnetic field can be derived
from the following magnetic vector potential:
( )
z
( r , z ) = B( ) cos [ q ( z z d )] t0, ( r ),
(6.2.4)
=0
where q = / ( L d ) ,
t0, ( r ) = I0 ( r )K 0 ( Rext ) K 0 ( r )I0 ( Rext ),
254
(6.2.5)
j =0
E
( )
z
( )
c2
( r, z ) =
( 2 ) B( ) cos[ q ( z z d )]t0, ( r ),
j =0
H ( r , z ) =
( )
cos[ q ( z z d )]t1, ( r ),
(6.2.6)
=0
(6.2.7)
255
c2
jkn z
2
(
)
I0 ( n Rint )
A
e
n
n
j n =
0
for z < z < z + d ,
(6.2.8)
=
c2
j =0
and the azimuthal magnetic field [ H ( r = Rint , z + d < z < z + L) ] reads
1
1
jkn z
I1 ( n Rint ) = B( ) cos[q ( z z d )]t1, ( Rint ). (6.2.9)
n Ane
0 n =
0 =0
From these boundary conditions the dispersion relation of the structure can be
developed. For this purpose we analyze the solution in the grooves having Floquet's
theorem in mind. The latter implies that the longitudinal electric field in the 's groove
has to satisfy the following relation:
jkL
2 ( )
B cos[q ( z z d )]t0, (r ) = 2 B( +1)e cos[q ( z + L z +1 d )]t0, (r ).
=0
=0
256
(6.2.10)
L
jk z
2
2
(6.2.12)
I
(
)
(
)
A
L
R
B
t
R
n n n,m 0 n int 0, int dze m cos[q ( z d )],
d
=0
n =
here n , m is the Kroniker delta function which equals 1 if n = m and zero otherwise. We
also used the orthogonality of the Fourier spatial harmonics. We follow a similar
procedure when imposing the continuity of the magnetic field with one difference, (6.2.9)
is defined only in the groove aperture thus we shall utilize the orthogonality of the
trigonometric function cos[ q ( z d )] . Accordingly, (6.2.9) is multiplied by
cos[ q ( z d )] and we integrate over d < z < L; the result is
n =
An I1 ( n Rint ) dz cos q ( z d ) e
jkn z
= B t1, ( Rint )( L d ) g , ,
=0
(6.2.13)
257
d
)
e
,
[
]
d
Ld
which allows us to write (6.2.12) as
1
Ld 2
(6.2.15)
=
t0, ( Rint )n , ( k ) B ,
An
2
n I0 ( n Rint ) L =0
and (6.2.13) as
1
(6.2.16)
B
An n I1 ( n Rint )*n , ( k ).
=
t1, ( Rint ) g n =
These are two equations for two unknown sets of amplitudes ( An , B ) and the
dispersion relation can be represented in two equivalent ways: One possibility is to
substitute (6.2.16) in (6.2.15) and get
n,m
n , m , Am =
2
L n I0 ( n Rint ) =0 t1, ( Rint ) g
m =
258
0.
,
n , n , B =
L t1, ( Rint ) g n= n I0 ( n Rint )
=0
In both cases the dispersion relation is calculated from the requirement that the
determinant of the matrix which multiplies the vector of amplitudes, is zero.
Although the two methods are equivalent, we found that the latter expression is by
far more efficient for practical calculation because of the number of modes required to
represent adequately the field in the groove compared to the number of spatial harmonics
required to represent the field in the inner section. In the case of single mode operation
we found that 1 to 3 modes are sufficient for description of the field in the grooves and
about 40 spatial harmonics are generally used in the inner section. As indicated by these
numbers it will be much easier to calculate the determinant of a 3 3 matrix rather than
40 40 one; we shall quantify this statement later. At this point we shall discuss the
design of a disk-loaded structure assuming that the number modes in the grooves and
harmonics in the inner space are sufficient.
Let assume that we want to determine the geometry of a disk-loaded structure which
enables a wave at 10 GHz to be in resonance with electrons with = 0.9 and the phase
advance per cell is assumed to be kL = 2 / 3. These two conditions determine the period
of the structure. In our case L = 9 mm. There are three additional geometric parameters
to be determined: Rext ,Rint and d . The last two have
259
11
10
f (GHz )
R ext = 14 mm
Rint = 8mm
L = 9mm
d = 2 mm
kL = 2 / 3
9
Beam Line
= 0.9
8
0.0
0.4
0.8
kL /
260
f (GHz )
f (GHz )
261
(6.2.20)
sinc kn ( L d ) exp j kn ( L + d ) .
n ,0 ( k ) =
L
2
Let us compare the first few spatial harmonics relative to the zero harmonic. For this
purpose we take f = 10 GHz, v0 = 0.9c,Rint = 8 mm, L = 9 mm and d = 2 mm. The ratio
of the first two amplitudes is
262
A1
A
A
A
(6.2.21)
= 8 103 , 1 = 3 106 , 2 = 2 106 , 2 = 1 108
A0
A0
A0
A0
This result indicates that on axis, the amplitude of the interacting harmonic is dominant.
At the interface with the grooves ( r = Rint ) the ratio between the contribution of the zero
and n th harmonic is much closer to unity and it can be checked that it reads
| Ez , n ( r = Rint ) | | sinc( kn ( L d ) / 2) |
(6.2.22)
=
,
| Ez ,0 ( r = Rint ) | | sinc( k0 ( L d ) / 2) |
which is a virtually unity.
A more instructive picture is obtained by examining the average power flowing in
one time and spatial period of the system:
Rint
1 L 1
(6.2.23)
P = 2 drr dz Er ( r , z ) H* ( r , z ) .
0
0
L
2
0 n =
0
the integral can be calculated analytically [Abramowitz and Stegun (1968) p.484] and it
reads
263
1
(6.2.25)
U ( ) dxxI12 ( x ) I0 ( )I1 ( ) + 2 [I12 ( ) I02 ( )].
0
2
Based on these definitions we can calculate the average power carried by each harmonic
as
ck
(6.2.26)
Pn
=
| cAn |2 n U ( n Rint ),
264
2Rint
L
In this section an analysis similar to
that in Section 6.2 is applied to an open
periodic structure. As we shall see the
d
number of modes which may develop in
such a structure is small and therefore
mode competition is minimized.
We shall consider a system in which the wave propagates along the periodic
structure which consists of a disk-loaded wire, as illustrated in Figure 5 whose periodicity
is L, the inner radius is denoted by Rint , the external by Rext and the distance between two
cavities is d . Floquet's theorem as formulated in (6.1.8) allows us to write for the
magnetic potential in the external region ( > r Rext ) the following expression
jkn z
(6.3.1)
K 0 ( n r ),
Az (r , z ) = An e
n =
265
c2
jkn z
( jkn n ) An e
K1 ( n r ),
Er ( r , z ) =
j n =
c2
jkn z
2
(
)
K 0 ( n r ),
Ez (r , z ) =
A
e
n
n
j n =
H (r , z ) =
( ) A e
0 n =
jkn z
(6.3.2)
K1 ( n r )
In these expressions K 0 ( x ), K1 ( x ) are the zero and first order modified Bessel functions
of the second kind respectively and 2n = kn2 ( / c )2 . This choice of the radial functional
variation is dictated by the condition of convergence of the electromagnetic field far
away from the structure.
In each individual groove the electromagnetic field can be derived from the
following magnetic vector potential:
( )
z
( r , z ) = B( ) cos[ q ( z z d )]t0, ( r ),
(6.3.3)
=0
where q = / ( L d ) ,
t0, ( r ) = I0 ( r )K 0 ( Rint ) K 0 ( r )I0 ( Rint ),
and 2 = q2 ( / c )2 . The electromagnetic field reads
266
(6.3.4)
( )
r
c2
( )
q
B
( r, z ) =
(
sin[ q ( z z d )]t1, ( r ),
j =0
( )
z
c2
( r, z ) =
( 2 ) B( ) cos[ q ( z z d )]t0, ( r ),
j =0
( )
H ( r , z ) =
( )
cos[ q ( z z d )]t1, ( r ).
(6.3.5)
=0
267
c2
jkn z
2
(
)
K 0 ( n Rext ) =
A
e
n
n
j n =
0
for z < z < z + d ,
(6.3.7)
c2
j =0
and the azimuthal magnetic field, H ( r = Rext , z + d < z < z + L) , reads
1
1
jkn z
(6.3.8)
K1 ( n Rext ) B( ) cos[q ( z z d )]t1, ( Rext ).
n Ane
0 n =
0 =0
L
jkm z
2
2
(6.3.9)
=
K
(
)
(
)
cos[q ( z d )].
A
L
R
B
t
R
dze
n n n,m 0 n ext 0, ext
n =
=0
268
We follow a similar procedure when imposing the continuity of the magnetic field; the
difference in this case is that (6.3.8) is defined only in the groove's aperture thus we shall
utilize the orthogonality of the trigonometric function cos[ q ( z d )] . Accordingly,
(6.3.8) is multiplied by cos[ q ( z d )] and we integrate over d < z < L; the result is
n =
jkn z
B t1, ( Rext )( L d ) g , .
=0
(6.3.10)
In this expression g0 = 1 and g n 0 = 0.5 . It is convenient to define the quantity
L
1
jkn z
(6.3.11)
=
n , (k )
dz
cos[
q
(
z
d
)]
e
,
d
Ld
by whose means, (6.3.9) reads
1
Ld 2
(6.3.12)
An
t0, ( Rext )n, ( k ) B ,
2n K 0 ( n Rext ) L =0
whereas (6.3.10)
1
(6.3.13)
B
An n K1 ( n Rext )*n , ( k ).
t1, ( Rext ) g n =
These are two equations for two unknown sets of amplitudes ( An , B ) . The
269
2
L n K 0 ( n Rext ) =0 t1, ( Rext ) g
m =
The other possibility is to substitute (6.3.12) in (6.3.13) and obtain one equation for the
amplitudes of the various modes in the groove
2
L
t
(
R
)
g
K
(
R
)
=0
1,
n = n 0
n ext
ext
In both cases the dispersion relation is calculated from the requirement that the
determinant of the matrix which multiplies the vector of amplitudes, is zero. As in the
closed structure the two methods are equivalent, but the last expression is by far more
efficient for practical calculation.
There is one substantial difference between open and closed periodic structures. In the
latter case, the radiation is guided by the waveguide and there is an infinite discrete
spectrum of frequencies which can propagate along the system. In open structures, modes
can propagate provided that the projection of the wavenumbers of all harmonics in the
first Brillouin zone corresponds to waves whose phase velocity is smaller than c ; in other
270
ph = 1
ph = 1
words, no radiation propagates outwards
(radially). The figure illustrates the two
regions of interest: in the shadowed region no
solutions are permissible and in the remainder
the solution is possible with an adequate
choice of the geometric parameters. It is
evident from this picture that waves at
frequencies higher than
-4
-2
1c
,
2L
0
50
271
40
f (GHz )
-6
30
20
4 kL /
(6.3.16)
Rint = 15 mm
R ext = 21mm
L = 3mm
d = 1mm
ph = 1
10
0
0.0 0.2 0.4 0.6 0.8 1.0
kL /
6.4 Transients
In order to illustrate the effect of the periodicity on the propagation of a wave packet
we shall consider at t = 0 the same wavepacket a ( z ) , in vacuum and in a periodic
structure. The propagation in vacuum will be represented by a dispersion relation
k 2 = 2 / c 2 , therefore a scalar wave function ( z , t ) is given by
jkz 1 jkct
jkct
(6.4.1)
e
.
( z , t ) = dk ( k )e
+e
2
Since at t = 0 this function equals a ( z ) , the amplitudes ( k ) can be readily determined
using the inverse Fourier transform hence
1
jkz
(6.4.2)
dza
z
e
(k ) =
(
)
.
2
Substituting back into (4.1) we find that
1
(6.4.3)
( z , t ) = [ a ( z ct ) + a ( z + ct )],
2
which basically indicates that the pulse moves at the speed of light in both directions and
asymptotically, it preserves its shape. In a periodic structure the description of the
wavepacket is complicated by the dispersion relation which in its lowest order
approximation (e.g., first TM symmetric mode in a waveguide) can be expressed as
272
( k ) = cos( kL),
(6.4.4)
where = (0 + ) / 2 is the average frequency between the low ( kL = 0 ) cut-off
denoted by 0 and the high ( kL = ) cut-off denoted by . The quantity
= ( 0 ) / 2 is half the passband width and L is the period of the structure.
Contrary to the previous case k here denotes the wavenumber in the first Brillouin zone.
In the framework of this approximation we can use Floquet's representation to write
/L
j( k )t jkn z
(6.4.5)
( z , t ) = Re dk n ( k )e
,
/ L
n =
jk z
(6.4.6)
n ( k ) = 12 dza ( z )e n .
/L
1
jt[ cos( kL)] jkn ( z )
( z , t ) = Re d a ( ) dke
.
L
/
n =
2
273
(6.4.7)
1
( 1/ )
e2
J ( ),
(6.4.8)
/L
1
jt
j ( kL /2) jkn ( z )
,
( z , t ) = Re d a ( ) dke
J (t )e
e
L
/
n =
=
2
(6.4.9)
which after the evaluation of the integrals and summation (over n ) reads
( z, t ) =
274
(6.4.10)
275
(z, t )
(z, t )
(z, t )
(z, t )
0.5
0.5
The figure illustrates the propagation of two
t = 6L / c
t = 3L / c
wavepackets in vacuum (dashed line) and in a
periodic structure. The latter is characterized by
0.0
0.0
= 2 10 GHz, = / 30 and a spatial
periodicity of L = 1 cm. At t = 0 the distribution is a
Gaussian, a ( z ) = exp[ ( z / L)2 ] . In each one of the
-0.5
-0.5
frames ( z , t ) was plotted at a different time as a
0 5 10 15
0 5 10 15
z/L
z/L
function of z . Characteristic to all the frames is the
relatively large peak following the front of the pulse.
0.5
0.5
t = 12 L / c
t = 9L / c
It is evident that although the front of the pulse
propagates at the speed of light (as in vacuum) the
main pulse propagates slower. In fact, a substantial
0.0
0.0
fraction of the energy remains at the origin even a
long time after t = 0 . For the parameters used, the
amplitude of the signal at the origin ( z = 0 ) is
-0.5
-0.5
dominated by the zero order Bessel function i.e.,
0 5 10 15
0 5 10 15
z/L
z/L
J 0 (t ) therefore the energy is drained on a time
scale which is determined by the asymptotic behavior
of the Bessel function namely 1 / t . Clearly the wider the passband the faster the
energy is drained from the origin.
Exercise 6.1: Based on the solution for Az ( r , z ) in Section 6.1 determine the Floquet
276
Chapter 7
Generation of radiation
Throughout these notes it was always assumed that somewhere there is a radiation
source providing us with the necessary energy. In what follows we shall skim through the
fundamentals of generation of radiation. Our discussion will be limited to "vacuum''
devices, this is to say that the energy is extracted from free electrons in vacuum.
Employed on medium ( < 1kW) and high ( > 1kW) power devices, the concepts to be
discussed rely solely on energy and momentum conservation. Another very important
family of devices relying on electrons moving in "solid state'', will not be discussed here.
276
277
z
(7.1.1)
E x ( z , t ) = E0 cos t .
c
If a charged particle moves at v parallel to z axis, then the electric field this charge
experiences (neglecting the effect of the charge on the wave) is given by
z (t )
(7.1.2)
Ex [ z (t ), t ] = E0 cos t
.
c
A crude estimate for the particle's trajectory is
(7.1.3)
z (t ) vt,
therefore if the charge moves in the presence of this wave from t to t then
the average electric field it experiences is zero,
v
(7.1.4)
dt
cos
t 1 c = 0,
even if the particle is highly relativistic. The lack of interaction can be illustrated in a
clearer way by superimposing the dispersion relation of the wave and the particle on the
same diagram. Firstly, the relation between energy and momentum for an electron is
given by
(7.1.5)
E = c p 2 + ( mc )2 ,
where m = 9.1094 1031 Kg is the rest mass of the electron. Secondly, the corresponding
278
279
pf
pi
v
(7.1.9)
= E0 sinc T 1 ;
c
here sinc( x ) = sin( x ) / x . That is to say that if the time the electron spends in the
interaction region, is small on the scale of the radiation period T0 = 2 / then the net
electric field it experiences, is not zero. From the perspective of the conservation laws,
the interaction is possible since although the energy conservation remains unchanged i.e.,
(7.1.10)
Ei = Ef + ,
the constraint on momentum conservation was released somewhat and it reads
280
E
Ei
| pi pf
E = c p 2 + m 2c 2
|< cT ,
(7.1.11) E
f
c
which clearly is less stringent than in (7.1.8) as also illustrated
E = cp
in the figure; = 1.05457 1034 Jsec is the Planck constant. The
operation of the klystron to be discussed subsequently relies on
the interaction of an electron with a wave in a region which is
shorter than the radiation wavelength.
pf
pi
pi
(7.1.12)
E x ( z , t ) = E0 e ( t z / c ) / .
E
A particle following the same trajectory as in (7.1.3) will clearly
experience an average electric field which is non-zero even when
Ei
the interaction duration is infinite. This is possible since the
E = c p 2 + m 2c2
spectrum of the radiation field is broad -- in contrast to Sect. 7.1.1
where it was peaked -- therefore again the constraint of the Ef
conservation laws is less stringent:
E = cp
281
pf
| Ei Ef |< ,
(7.1.13)
and
(7.1.14)
.
c
Schematics of this mechanism is illustrated in the figure. It should be also pointed out
that in this section we consider primarily the kinematics of the interaction and we pay no
attention to the dynamics. In other words, we examined whether the conservation laws
can be satisfied without details regarding the field configuration.
| pi pf |<
E
Ei
E = c p 2 + m2c2
Ef
E = cp
E = (c / n )p
282
pf
pi
283
E
Ei
E = c p 2 + m 2c 2
7.1.5 Compton Scattering: Static Fields
Ef
E = cp
nh / L
2 1/2
where [1 (v / c) ] . Note that the frequency of the emitted photon depends on the
284
velocity of the electron which means that by varying the velocity we can change the
operating frequency. A radiation source which possesses this feature is a tunable source.
Identical result is achieved if we assume a periodic electrostatic field and both field
configurations are employed in the so-called "free electron lasers''.
7.1.6 Compton Scattering: Dynamic Fields
Static electric or magnetic field can be conceived as limiting cases of a dynamic
field of zero or vanishingly small frequency and we indicated above that they facilitate
the interaction between an electron and a wave. Consequently we may expect that the
interaction of an electron with a wave will occur in the presence of another wave. Indeed,
if we have an initial wave of frequency 1 and the emitted wave is at a frequency 2 the
conservation laws read
(7.1.20)
Ei + 1 = Ef + 2 ,
and
pi = pf +
(7.1.21)
c
c
Following the same procedure as above we find that the ratio between the frequencies of
the two waves is
285
E1
Ei
E2
2
4 2 ,
1
(7.1.22)
Ef
E = cp
E = c p 2 + m 2c 2
pf
286
pi
E = c p 2 + m 2 c 2 2 neBh / 2
E
Ei
n 1
ec 2 B
ec 2 B
n
(7.1.24)
Ei n1
= Ef n2
+ Eph ,
Ei
Ef
Ef
and the momentum conservation remains unchanged i.e.,
n +1
(7.1.25)
pi = pf + pph .
E = cp
From these two equations we find that the frequency of the
emitted photon is
pf
pi
eB
2 eB
= 2
= 2
(7.1.26)
.
m
m
The last term is known as the relativistic cyclotron angular frequency, c, rel eB / m .
The figure illustrates schematically this type of interaction. It indicates that the dispersion
line of the electron is split by the magnetic field in many lines (index n ) and the
interaction is possible since the electron can move from one line to another. Gyrotron's
operation relies on this mechanism and it will be discussed briefly in the next section.
7.1.8 Synchronism Condition
All the processes in which the interaction of an electron with a monochromatic wave
extends to large regions, have one thing in common: the velocity of the electron has to
287
equal the effective phase velocity of the pondermotive wave along the electron's main
trajectory, namely
(7.1.27)
v = vph,eff .
Here by pondermotive wave we mean the effective wave along the longitudinal
trajectory of the particle which accounts for transverse or longitudinal oscillation. In the
Cerenkov case we indicated that the phase velocity is c / n and there is no transverse
motion therefore the condition for interaction implies n = c / v where n is the refraction
coefficient of the medium. In the presence of a periodic static field the wavenumber of
the pondermotive wave is / c + 2 / L therefore
vph,eff =
/ c + 2 / L
2 1
(7.1.28)
(7.1.29)
k2 + k1
where k1,2 = 1,2 / c are the wavenumbers of the two waves involved. Finally, in a
uniform magnetic field only the effective frequency varies
c,rel
(7.1.30)
vph,eff =
.
k
The reader can check now that within the framework of this formulation we obtain
288
(7.1.19) from (7.1.28), in the case of the dynamic field we have from (7.1.29) the 4 2
term as in (7.1.22) and finally (7.1.30) leads to the gyrotron's operation frequency
presented in (7.1.26).
289
290
Electrons move along a drift tube and its geometry is chosen in such a way that at
the frequency of interest it does not allow the electromagnetic wave to propagate. The
latter is confined to cavities attached to the drift tube. The wave which feeds the first
cavity modulates the velocity of the otherwise uniform beam. This means that after the
cavity, half of the electrons have a velocity larger than the average beam velocity
whereas the second half has a smaller velocity. According to the change in the (nonrelativistic) velocity of the electrons the beam becomes bunched down the stream since
accelerated electrons from one period of the electromagnetic wave catch up with the slow
electrons from the previous period. When this bunch enters the gap of another cavity it
may generate radiation very efficiently. In practice, several intermediary cavities are
necessary to achieve good modulation.
7.2.2 The Traveling Wave Tube
The traveling wave tube (TWT) is a Cerenkov device, namely the phase velocity of
the interacting wave is smaller than c and the interaction is distributed along many
wavelengths. Generally speaking, as the beam and the wave advance, the beam gets
modulated by the electric field of the wave and in turn, the modulated beam increases the
amplitude of the electric field. In this process both the beam modulation and the radiation
field grow exponentially in space.
291
cathode
collector
(a)
output
input
cathode
(b)
output
output
input
cathode
(c)
output
input
cathode
(d)
cathode
(e)
292
In the interaction process the electron oscillates primarily along the major axis ( z
direction ) and the interaction is with the parallel component of the electric field.
Correspondingly, the interaction occurs here with the transverse magnetic (TM) mode.
7.2.3 The Gyrotron
The gyrotron relies on the interaction between an annular beam, gyrating around the
axis of symmetry due to an applied uniform magnetic field, and a transverse electric (TE)
mode. The concept of generating coherent radiation from electrons gyrating in a magnetic
field was proposed independently by three different researchers in the late fifties, and it
has attracted substantial attention due to its potential to generate millimeter and
submillimeter radiation.
In this device electrons move in the azimuthal direction and they get bunched by the
corresponding azimuthal electric field. As in the case of the TWT the bunches act back
on the field and amplify it. In contrast to traveling wave tubes or klystrons in which the
beam typically interacts with the lowest mode, in the gyrotron the interaction is with high
modes therefore various suppression techniques are employed in order to obtain coherent
operation with a single mode.
The operation frequency is determined by the applied magnetic field, the energy of
the electrons and, in cases of high mode operation, also by the radius of the waveguide:
293
(7.2.1)
= c + c2 + c2o ,
where = v / c , c = eB / m and co is the cutoff frequency of the mode. The operating
frequency in this case can reach very high values: for a magnetic field of 1T and 2.5
the operation frequency is of the order of 150 GHz or higher according to the mode with
which the electrons interact.
Circuit Coil
First Anode
-350kV
Second Anode
Open Cavity
Cathode
500kV
Gun Coil
295
traveling wave tube. However, its major advantage is the fact that it does not require a
metallic (or other type of) structure for the interaction to take place. Consequently, it has
the potential to either generate very high power at which the contact of radiation with
metallic walls would create very serious problems, or produce radiation at UV, XUV or
X-ray where there are no other coherent radiation sources. The Figure illustrates the basic
configuration.
cathode
296
v
The magnetron was invented at the beginning
of the 20th century but because of its complexity
B
there is no analytical model, as yet, which can
describe its operation adequately as a whole. In
recent years great progress has been made in the
understanding of the various processes with the aid
of particle in cell (PIC) codes. Its operation combines potential and kinetic energy
conversion. The Figure illustrates the basic configuration. Electrons are generated on the
cathode (inner surface) and since a perpendicular magnetic field is applied they form a
flow which rotates azimuthally. The magnetic field and the voltage applied on the anode
are chosen in such a way that, in equilibrium, the average velocity of the electrons equals
the phase velocity of the wave supported by the periodic structure at the frequency of
interest.
A simplistic picture of the interaction can be conceived in the following way:
electrons which lose energy to the wave via the Cerenkov type interaction, move in
upward trajectories -- closer to the anode. Consequently, two processes occur. Firstly, the
closer the electron is to the periodic surface the stronger the radiation field and therefore
297
298
r
'
'
(7.3.1)
G ( r , z | r , z ) = Gs ( z | r' , z' )J 0 ps ,
R
s =1
substitute in (2.4.11) and use the orthogonality of the Bessel functions we find that
r'
1
'
'
(7.3.2)
Gs ( z | r , z ) = J 0 ps
g s ( z | z' ),
1
R R2 J2 ( p )
1
s
2
where g s ( z | z' ) satisfies
d2
1
'
2
g
(
z
|
z
)
=
( z z' ),
(7.3.3)
s s
dz 2
2
and
299
ps2
2
=
r 2 .
R2
c
'
For z > z the solution of (7.3.3) is
2
s
g s ( z | z ) = A+ e
and for z < z' the solution is
'
(7.3.4)
s ( z z' )
s ( z z' )
(7.3.5)
(7.3.6)
g s ( z | z ) = A e
.
Green's function is continuous at z = z' i.e.,
(7.3.7)
A+ = A ,
and its first derivative is discontinuous. The discontinuity is determined by integrating
(7.3.3)from z = z' 0 to z = z' + 0 i.e.,
1
d
d
'
'
(7.3.8)
g
(
z
|
z
)
g
(
z
|
z
)
.
dz s
'
dz s
'
2
z = z +0
z = z 0
Substituting the two solutions introduced above, and using (7.3.7)we obtain
1 s | z z' |
'
(7.3.9)
.
gs ( z | z ) =
e
4 s
Finally, the explicit expression for the Green's function corresponding to azimuthally
symmetric TM modes in a circular waveguide is given by
'
300
J 0 ( ps r / R )J 0 ( ps r' / R ) 1 s | z z' |
(7.3.10)
.
G( r, z | r , z ) =
e
1
4
2
2
s =1
s
R J1 ( ps )
2
In this expression it was tacitly assumed that > 0 and s [defined in (7.3.4)] is nonzero.
With Green's function established, we can calculate the magnetic vector potential as
generated by the current distribution described in (2.4.10); the result is
'
'
e
= 02
8
J 0 ( ps r / R )
2
j ( / v0 ) z
(7.3.11)
.
e
2
2
2
1
s =1
R 2 J12 ( ps ) s + / v0
2
It will be instructive to examine this expression in the time domain; the Fourier
transform is
e
2
Az ( r , z , t ) = 2
2 0 R 2 1 n 2 2
J ( p r / R)
0 2 s
J 1 ( ps )
s =1
j ( t z / v0 )
2 + 2s
301
(7.3.12)
where
p c
(7.3.13)
= s
.
2 2
R 1 n
The problem has been now simplified to the evaluation of the integral
e j
(7.3.14)
Fs ( = t z / v0 ) d 2
,
2
+ s
which in turn is equivalent to the solution of the following differential equation
d2
2
(2.4.39)
s Fs ( ) = 2 ( ).
d 2
If the particle's velocity is smaller than the phase velocity of a plane wave in the medium
( n < 1 ) then 2s > 0 and the solution for > 0 is
2
2
s
Fs ( > 0) = A+ e
(7.3.15)
(7.3.16)
or
Fs ( < 0) = A e
302
(7.3.17)
F
Fs ( )
=
2 .
(
)
s
d
=0+ d
=0
When the velocity of the particle is smaller than c / n (i.e., n < 1 ) the characteristic
frequency s is real, therefore
Fs ( ) =
and
2
Az ( r , z , t ) =
2 0 R 2 1 2 n 2
s | |
(7.3.18)
J 0 ( ps r / R ) s |t z / v0 |
(7.3.19)
e
.
2
s =1 J1 ( ps ) s
This expression represents a superposition of evanescent modes attached to the particle.
It is important to emphasize that since the phase velocity of a plane wave in the medium
is larger than the velocity of the particle there is an electromagnetic field in front ( < 0 )
of the particle. The situation is different in the opposite case, > 1 / n , since 2s < 0 . In
this case the waves are slower than the particle and there is no electromagnetic field in
front of the particle i.e.,
Fs ( < 0) = 0.
(7.3.20)
By virtue of the continuity at = 0 we have for > 0
(7.3.21)
Fs (=
> 0) A+ sin (| s | ).
e
303
sin (| s | )h( ),
| s |
and the magnetic vector potential reads
e
2
Az ( r , z , t ) =
0 R 2 n 2 2 1
(7.3.22)
J 0 ( ps r / R )
z
z
(7.3.23)
sin | s | t h t ,
2
v
v
p
J
(
)
|
|
s =1 1
0
0
s
s
where h( ) is the Heaviside step function. This expression indicates that when the
velocity of the particle is larger than c / n, there is an entire superposition of propagating
waves traveling behind the particle. Furthermore, all the waves have the same phase
velocity which is identical with the velocity of the particle, v0 . It is important to bear in
mind that this result was obtained after tacitly assuming that r is frequency independent
which generally is not the case, therefore the summation is limited to a finite number of
modes. The modes which contribute are determined by the Cerenkov condition
n( = s ) > 1.
After we established the magnetic vector potential, let us now calculate the average
power which trails behind the particle. Firstly, the azimuthal magnetic field is given by
304
H ( r , z , t ) =
=
1
Az ( r , z , t )
0 r
As
s =1
sin | s
ps r
J 1 ps
R R
z
z
| t h t
v0 v0
where
(7.3.24)
e
2
1
(7.3.25)
As =
.
0 R 2 n 2 2 1 J12 ( ps ) | s |
Secondly, the radial electric field is determined by the electric scalar potential which in
turn is calculated using the Lorentz gauge and it reads
305
( r , z , t )
r
ps r
c2
A
=
s J1 ps
r v0 s =1 R R
Er ( r , z , t ) =
z
z
(7.3.26)
sin | s | t h t .
v0 v0
With these expressions we can calculate the average electromagnetic power trailing the
particle. It is given by
e2 c
1
1
(7.3.27)
.
P=
2
2
2
2 0 r R r 1 s =1 J1 ( ps )
Note that for ultra relativistic particle ( 1) the power is independent of the particle's
energy. In order to have a measure of the radiation emitted consider a very narrow bunch
of N 1011 electrons injected in a waveguide whose radius is 9.2 mm. The waveguide is
filled with a material whose dielectric coefficient is r 2.6 and all electrons have the same
energy 450 keV. If we were able to keep their velocity constant, then 23 MW of power at
11.4 GHz (first mode, s = 1 ) will trail the bunch. Further examining this expression we
note that the average power is quadratic with the frequency i.e.,
306
| s |2
( Ne)2
(7.3.28)
P Ps =
.
2
2 0 r c s =1 [ ps J1 ( ps )]
s =1
In addition, based on the definition of the Fourier transform of the current density in
(2.4.10), we conclude that the current which this macro-particle excites in the s th mode
is I s eN s / 2 . With this expression, the radiation impedance of the first mode ( s = 1 ) is
P1
4
(7.3.29)
.
0
=
RC,1 =
2
1
r [ p1J1 ( p1 )]
| I1 |2
2
For a relativistic particle, 1 , a dielectric medium r = 2.6 the radiation impedance
corresponding to the first mode is 1200 which is one order of magnitude larger than
that of a dipole in free space or between two plates. Note that this impedance is
independent of the geometry of the waveguide and for an ultra-relativistic particle it is
independent of the particle's energy.
307
In this section we shall consider only the internal problem, ignoring the electromagnetic
phenomena outside the cavity. The boundary conditions on the internal walls of the
cavity impose that Ez ( r = R, z , t ) = 0 , Er ( r , z = 0, t ) = 0 and Er ( r , z = d , t ) = 0 therefore
the magnetic vector potential reads
r
n
(7.4.3)
Az ( r , z , t ) = As ,n (t )J 0 ps cos
z .
R
d
s =1,n =0
308
Using the orthogonality of the trigonometric and Bessel functions we find that the
amplitude As , n (t ) satisfies
d2
ev 0
1
1
2
A
(
t
)
+
s ,n s ,n
dt 2
2 0 1 R 2 J 2 ( p ) g n d
s
1
2
v
n
v0t h(t ) h t 0 ,
cos
d
d
(7.4.4)
where
1 for n = 0,
gn
0.5 otherwise,
(7.4.5)
and
p n
(7.4.6)
=
s ,n c s +
,
R d
are the eigen-frequencies of the cavity. Before the particle enters the cavity ( t < 0 ), no
field exists, therefore
(7.4.7)
As , n (t < 0) = 0.
For the time the particle is in the cavity namely, 0 < t < d / v 0 , the solution of (7.4.4)
2
309
d
As ,n 0 < t < =
B1 cos( s ,n t ) B2 sin( s ,n t )
v
0
+ s ,n cos(n t ),
(7.4.8)
where
s ,n =
ev 0
1
1
1
,
2
2
1
2 0 R 2 J 2 ( p ) g n d s ,n n
s
1
2
and
n =
(7.4.9)
(7.4.10)
v0.
d
Since both the magnetic and the electric field are zero at t = 0 , the function As , n (t ) and
its first derivative are zero at t = 0 hence
(7.4.11)
B1 s ,n = 0,
and
(7.4.12)
B2 = 0.
Consequently, the amplitude of the magnetic vector potential [ As , n (t ) ] reads
=
As ,n (t ) s ,n cos(n t ) cos( s ,n t ) .
310
(7.4.13)
Beyond t = d / v0 , the particle is out of the structure thus the source term in (7.4.4) is
zero and the solution reads
d
d
As ,n t > =
C
cos
s ,n
1
v
v
0
0
d
(7.4.14)
+ C2 sin s ,n t .
v
0
d
n
(7.4.15)
C1,
s ,n (1) cos s ,n =
v
0
d
(7.4.16)
C2 s , n .
=
v
0
For this time period, the explicit expression for the magnetic vector potential is
s ,n s ,n sin s ,n
311
d
d
d
n
As ,n t > =
t
(
1)
cos
cos
s ,n
s ,n
s ,n
v
v
v
0
0
0
d
d
(7.4.17)
+ s ,n sin s ,n sin s ,n t ,
v
v
0
0
The expressions in (7.4.7), (7.4.13), (7.4.17) describe the magnetic vector potential in the
cavity at all times. The Figure illustrates schematically this solution.
During the period the electron spends in the
cavity, there are two frequencies which are
excited: the eigen-frequency of the cavity s , n and
the "resonances" associated with the motion of the
particle, n . The latter set corresponds to the case
when the phase velocity, v ph / k , equals the
velocity v0 . Since the boundary conditions impose
k n / d and the resonance implies
2 d
=
v 0 v=
c 2
ph
,
c n
d
v0
R
r
z
d
v0
r
z
(7.4.18)
d
R
312
v0
r
z
d
r
Az r , z=
z cos (nt ) cos( s ,nt ) . (7.4.19)
,0 < t < s ,n J 0 ps cos
v 0 s =1,n =0
R
nd
Note that the upper limit in the double summation was omitted since in practice this limit
is determined by the actual dimensions of the particle, which so far was considered
infinitesimally small. In order to quantify this statement we realize that the summation is
over all eigenmodes which have a wavenumber much longer than the particle's dimension
i.e., s , n Rb / c < 1 .
According to Maxwell's equations, the longitudinal electric field is
1
(7.4.20)
0 E z (r , t ) =
rH (r , t ).
J z (r , t ) +
r r
t
Furthermore, the field which acts on the particle does not include the self field, therefore
we omit the current density term. Using the expression for the magnetic vector potential
[(2.1.36)], we have
1
(7.4.21)
Ez (r , t ) = c 2 dt
r Az (r , t ),
r r r
or explicitly,
313
2
n sin(nt ) sin( s ,nt )
cps r
J
p
cos(
z)
s ,n
0 s
d
s ,n
R R
s =1, n =0
n
(7.4.22)
In a lossless and closed cavity the total power flow is zero, therefore Poynting's theorem
in its integral form reads
R
d
dW
(7.4.23)
= 2 drr dzEz ( r , z , t )J z ( r , z , t ).
0
0
dt
Thus substituting the current density [(7.4.1)] we obtain
d
Ez r , z ,0 < t < =
v
0
W = ev 0
d / v0
dtEz (r , z = v 0t , t ),
(7.4.24)
.
n
0
s ,n
R
s =1, n =0
n
The time integral in this expression can be evaluated analytically. As can be readily
deduced, the first term represents the non-homogeneous part of the solution and its
contribution is identically zero whereas the second's reads
2
n
cps 1 (1) cos( s ,n d / v 0 )
(7.4.26)
W = ev 0 s ,n
.
2
2
s , n n
R
s =1, n =0
314
e2
W W
d
4
2 ps 1
s ,n
1
n
d . (7.4.27)
=
1 (1) cos
2
2 2
p
g
J
(
)
v
s =1, n =0 1
ps + ( nR / d )
s
n
0
In the Figure we illustrate two typical terms from the expression above as a function of
the particle's momentum. The (normalized) energy stored at 10.7 GHz (corresponding to
s = 1, n = 1) is shown in the left frame and we observe that for = 2.5 the energy
reaches
its
asymptotic
value
4.0 R = 1 .5 cm
4.0 R = 1.5 cm
[W ( s = 1, n = 1) 4 ]. This is in
d = 2.0 cm
d = 2.0 cm
contrast to the energy stored in the
3.0
3.0
35.5 GHz ( s = 3,n = 3 ) wave which at
the same momentum reaches virtually
2.0
2.0
zero level; the asymptotic value
[W ( s = 3, n = 3) 0.5] is reached for
1.0
1.0
a much higher momentum ( = 15 ).
W (1,1)
W (3,3)
0.0
0.0
315
1.0
2.0
0.0
0.0
1.0
2.0