Contents

Contents 1

1 Transmission line equations and their solution 4

1.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.2 Electromagnetism background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.3 Circuit model of a transmission line . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.4 Lossless lines. Wave equations and their solutions . . . . . . . . . . . . . . . . . . . 11
1.5 Review of Fourier transforms and phasors . . . . . . . . . . . . . . . . . . . . . . . 14
1.6 Transmission line equations in the frequency domain . . . . . . . . . . . . . . . . . 16
1.7 Propagation of the electric state and geometrical interpretations . . . . . . . . . . 21
1.8 Solution of transmission line equations by the matrix technique . . . . . . . . . . . 23

2 Parameters of common transmission lines 27

2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
2.2 Coaxial cable . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
2.3 Two-wire line . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
2.4 Wire on a metal plane . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
2.5 Shielded two-wire line . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
2.6 Stripline . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
2.7 Microstrip . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32

3 Lossless transmission line circuits 38

3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
3.2 Definition of local impedance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
3.3 Reflection coefficients . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
3.4 Energy balance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
3.5 Line voltage, current and impedance diagrams . . . . . . . . . . . . . . . . . . . . . 47
3.6 The Smith Chart . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
3.7 Analysis of simple circuits . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57

1
 CONTENTS

4 Energy dissipation in transmission lines 61

4.1 Dielectric losses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
4.2 Conductor losses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
4.3 Loss parameters of some transmission lines . . . . . . . . . . . . . . . . . . . . . . 68
4.3.1 Coaxial cable . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
4.3.2 Two-wire line . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70

5 Lossy transmission line circuits 72

5.1 Solution of transmission line equations . . . . . . . . . . . . . . . . . . . . . . . . . 72
5.2 Computation of the power flow . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
5.3 Frequency dependence of phase constant and characteristic impedance . . . . . . . 80

6 Matching circuits 84
6.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
6.2 Types of impedance matching . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
6.3 Impedance matching devices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
6.3.1 L cells with lumped reactive elements . . . . . . . . . . . . . . . . . . . . . 87
6.3.2 Resistive matching pad . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
6.3.3 Single stub matching network . . . . . . . . . . . . . . . . . . . . . . . . . . 91
6.3.4 Double stub matching network . . . . . . . . . . . . . . . . . . . . . . . . . 96
6.3.5 λ/4 matching networks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98

7 The Scattering matrix 101

7.1 Lumped circuits . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
7.2 Distributed parameter circuits . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103
7.3 Relationship between [S] and [Z] or [Y ] . . . . . . . . . . . . . . . . . . . . . . . . 104
7.4 Computation of the power dissipated in a device . . . . . . . . . . . . . . . . . . . 105
7.5 Properties of the scattering matrix [S] of a device . . . . . . . . . . . . . . . . . . . 106
7.6 Change of reference impedances . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106
7.7 Change of reference planes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107
7.8 Cascade connection of structures . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108
7.9 Scattering matrix of some devices . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112
7.9.1 Ideal attenuator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112
7.9.2 Isolator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112
7.9.3 Circulator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112
7.9.4 Ideal directional coupler . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113
7.10 Examples of analysis of structures described by S matrices . . . . . . . . . . . . . 114

2
 CONTENTS

7.10.1 Cascade connection of a two-port and a load . . . . . . . . . . . . . . . . . 115

7.10.2 Interconnection of two two-ports by means of a length of transmission line . 116
7.10.3 Change of reference impedance for a one-port load . . . . . . . . . . . . . . 117
7.11 Transmission matrix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118

8 Time domain analysis of transmission lines 122

8.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122
8.2 The group velocity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123
8.3 Distortions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127
8.4 Digital communication . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 130
8.5 Mismatched ideal transmission lines . . . . . . . . . . . . . . . . . . . . . . . . . . 131
8.5.1 General solution of transmission line equations . . . . . . . . . . . . . . . . 131
8.5.2 Mismatched ideal lines . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 132
8.5.3 Real interconnections . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 140

Bibliography 141

3
Chapter 1

Transmission line equations and

their solution

1.1 Introduction
Electromagnetic energy, once generated in one place, has a natural tendency to spread in the
whole space at a speed close to 300.000 Km/s. In telecommunications this behavior can be useful
when the user position is not known in advance, as in a broadcasting system or in a cell phone
network. In other applications, instead, electromagnetic energy must be transferred from one place
to the other along a well defined path without any spreading at all: an example is the cabling of
a building.
In the most general terms, a transmission line is a system of metal conductors and/or dielectric
insulating media that is capable of “guiding” the energy transfer between a generator and a load,
irrespective (at least with a good approximation) of the bends that the line undergoes because
of installation needs. From this point of view, a one dimensional propagation phenomenon takes
place on a transmission line.
There are many types of transmission lines, some examples of which are shown in Fig. 1.1.
The various line types are used for different applications in specific frequency ranges. Striplines
and microstrips are used only inside devices, such as amplifiers or filters, and their lengths never
exceeds some centimeters. Twisted pairs and coaxial cables are used for cabling a building but
coaxial cables can also be used for intercontinental communications. Hollow metal pipes, known as
waveguides, are used to deliver large amounts of microwave power over short to moderate distance.
Waveguides can also be made of dielectric materials only, as in the case of optical fibers. In
this text we will deal only with structures consisting of two metal conductors, such as coaxial
cables, microstrips and striplines. These can be defined transmission lines in strict sense, whereas
the others are more appropriately called metal or dielectric waveguides. More rigorously, all the
structures of Fig. 1.1 are waveguides, but those of the first type are characterized by the fact that
their fundamental propagation mode is TEM (transverse electromagnetic) - or quasi-TEM in the
case of microstrips - since they consist of two conductors. This implies that they can be used also
at very low frequency - even at dc - irrespective of their size. Waveguides, in general, have a lowest
frequency of operation, which depends on their transverse size. In conclusion, transmission lines
are waveguides whose behaviour, at sufficiently low frequency, is related to the TEM mode only.

4
Renato Orta - Transmission Line Theory (Nov. 2012)

n3
n2
n1
b
a c

d e

Figure 1.1. Examples of transmission lines: (a) coaxial cable, (b) two wire line, (c) optical
fiber, (d) microstrip , (e) stripline.

1.2 Electromagnetism background

The physical phenomena that take place in a transmission line belong to the realm of electromag-
netism and hence, from a quantitative point of view, they are completely described by four vector
fields: the electric field E(r,t), the magnetic field H(r,t), the electric displacement (or electric
induction) D(r,t) and the magnetic induction B(r,t). The relationships between these fields and
the sources (described by the current density J (r,t)) are specified by Maxwell equations, that are
written in MKSA units as
∂
∇ × E(r,t) = − B(r,t)
∂t
(1.1)
∂
∇ × H(r,t) = D(r,t) + J c (r,t) + J (r,t)
∂t
A general reference for electromagnetism is [1]. Let us review the meaning of the symbols and the
relevant measurement units.

E(r,t) electric field V/m

H(r,t) magnetic field A/m

D(r,t) electric induction C/m2

B(r,t) magnetic induction Wb/m2

J (r,t) current density (source) A/m2

J c (r,t) (conduction) current density [A/m2 ]

These equations must be supplemented with the constitutive relations, that describe the link

5
Renato Orta - Transmission Line Theory (Nov. 2012)

between fields and inductions. The simplest case is that of free space in which
B(r,t) = µ0 H(r,t)
(1.2)
D(r,t) = ²0 E(r,t)
where ²0 , dielectric permittivity, and µ0 magnetic permeability, have the values

µ0 = 4π · 10−7 H/m
1 1
²0 = 2
≈ · 10−9 F/m
µ0 c 36π
where the speed of light in free space c has the value

c = 2.99792458 · 108 m/s.

Moreover, in the case of a plane wave, the ratio between the magnitudes of the electric and magnetic
fields is called wave impedance and has the value
r
µ0
Z0 = ≈ 120π ≈ 377 Ω
²0

In the case of linear, isotropic, non dispersive dielectrics, the constitutive relations (1.2) are
substituted by
B(r,t) = µ H(r,t)
(1.3)
D(r,t) = ² E(r,t)
where

µ = µ0 µr
² = ²0 ²r

and µr , ²r (pure numbers) are the relative permittivity and permeabilities. All non ferromagnetic
materials have values of µr very close to 1.
When the dielectric contains free charges, the presence of an electric field E(r,t) gives rise to a
conduction current density J c (r,t):

J c (r,t) = γ E(r,t)

where γ is the conductivity of the dielectric, measured in S/m.

Even if an electromagnetic field can have an arbitrary time dependance, the time harmonic
(sinusoidal) regime with frequency f is very important, both from a theoretical and from an
application point of view. In these conditions, electromagnetic waves are characterized by a spatial
period λ0 = c/f , called wavelength, which is a sort of characteristic length of the field spatial
structure. It is known from Mathematics that a field with “arbitrary” time dependence can be
represented as a summation of sinusoidal fields with frequencies contained in a certain band (Fourier
theorem). In this case λ0 denotes the minimum wavelength, i.e. the one that corresponds to the
maximum frequency.
The size L of the structures with which the electromagnetic field interacts must always be
compared with wavelength. The ratio L/λ0 is defined electrical length of the structure and is a
pure number. Depending on the value of L/λ0 , essentially three regimes can be identified:

6
Renato Orta - Transmission Line Theory (Nov. 2012)

• quasi-static regime, with L/λ0 ¿ 1, typical of lumped parameter circuits

• the resonance regime , with L/λ0 ∼ 1, typical of distributed parameter circuits, analyzed in
this text

• the optical regime, with L/λ0 À 1, typical of the usual optical components studied by
classical optics (lenses, mirrors, etc...)

The solution technique of electromagnetic problems and even their modeling is different de-
pending on the regime of operation.
Lumped parameter circuit theory deals with the dynamics of systems made of elements of
negligible electrical size. The state variables employed in the model are the potential difference
vrs (t) between two nodes Pr and Ps of a network and the electric current irs (t) that flows in the
branch defined by the same two nodes. Rigorously, these quantities are defined uniquely only in
static conditions, i.e. at DC, but they are commonly used also in the frequency band for which the
electrical size of the network is very small. This condition can be reformulated in terms of transit
time. Indeed
L L L1 τ
= = =
λ0 c/f cT T
where T is the period of an oscillation with frequency f = 1/T and τ is the time that an electro-
magnetic wave requires to go from one end of the network to the other. Hence, an electromagnetic
system can be considered lumped provided the propagation delay is negligible with respect to the
period of the oscillations. For this reason one says that a lumped parameter circuit operates in
quasi-static regime.
Consider now one of the transmission lines shown in Fig. 1.1. Typically, their transverse size
is small with respect to wavelength but their length can be very large. Then, while a lumped
parameter circuit is modeled as point like, a transmission line is a one dimensional system, in
which voltage and currents depend on time and on a longitudinal coordinate that will always be
indicated with z. The state variables of such a system are then v(z,t) and i(z,t).
A circuit containing transmission lines is often called “distributed parameter circuit”to under-
line the fact that electromagnetic energy is not only stored in specific components, e.g. inductors,
capacitors, but also in the space surrounding the conductors of a line. As a consequence, a trans-
mission line is characterized by inductance and capacitance per unit length.
The equations that determine the dynamics of a transmission line could be obtained directly
from Maxwell equations, but for teaching convenience we will proceed in circuit terms, by gener-
alizing the properties of lumped parameters networks.

1.3 Circuit model of a transmission line

Consider a length of uniform transmission line, i.e. with a transverse cross section that is indepen-
dent of the longitudinal coordinate z. In Fig. 1.2a a coaxial cable is shown as an example. Fig. 1.2b
shows its symbol, i.e. its schematic and conventional representation in the form of two parallel
“wires” in which a current flows and between which a potential difference exists. It is evident that
all two conductor transmission lines have the same circuit symbol shown in Fig. 1.2b.
As previously remarked, a transmission line can be long with respect to wavelength, hence
its behavior cannot be predicted by Kirchhoff laws, that are applicable only to lumped parameter

7
Renato Orta - Transmission Line Theory (Nov. 2012)

(a) (b)

Figure 1.2. (a) Length of coaxial cable and (b) its symbolic representation

circuits. However we can subdivide the line in a large number of sufficiently short elements ∆z ¿ λ,
derive a lumped equivalent circuit for each of them and then analyze the resulting structure by
the usual methods of circuit theory. This is actually the modeling technique used in some circuit
simulators. We will instead follow a different route because we are interested in an analytical
solution of the problem. To this end we will let ∆z go to zero, so that we will be able to derive a
set of partial differential equations that can be solved in closed form.
Fig. 1.3 shows an element of the line with its equivalent circuit. To obtain the equivalent circuit

i(z,t) /∆z 5∆z i(z+∆z,t)

∆z

v(z,t) &∆z *∆z v(z+∆z,t)

(a) (b)

Figure 1.3. a) Element ∆z of a coaxial cable. The surface used to define L is shown
dashed. b) Equivalent circuit

of the element we use physical arguments; we make reference to the coaxial cable, but for the other
transmission lines one can proceed similarly . We start by observing that the current flowing in the
conductors produces a magnetic field with force lines surrounding the conductors. This field gives
rise to a linked flux through the rectangle shown in Fig. 1.3. The proportionality factor relating
the flux to the current is, by definition, the inductance of the element that we can write as L∆z
because the surface of the rectangle is clearly proportional to ∆z. Hence, L, measured in H/m is
the inductance per unit length of the line.
Analogously, power is dissipated in the metal conductors because of their limited conductivity:
hence, the equivalent circuit contains a series resistance with value R∆z, where R is the resistance
per unit length of the line, expressed in Ω/m.
Moreover, as a consequence of the potential difference maintained between the inner and outer
conductors, a charge is induced on them. The proportionality constant that relates the charge on

8
Renato Orta - Transmission Line Theory (Nov. 2012)

the ∆z element to the potential difference is, by definition, the capacitance of the element, that
we write C∆z, where C is the capacitance per unit length of the line, measured in F/m.
Finally, the dielectric between the conductors has a non zero conductivity, which is responsible
of a current flowing from one conductor to the other through the insulator. From a circuit point
of view, this phenomenon is accounted for by the conductance G∆z, where G is the conductance
per unit length of the line, measured in S/m.
Since ∆z ¿ λ Kirchhoff laws can be applied to the circuit of Fig. 1.3b:

 ∂

 v(z,t) − v(z + ∆z,t) = R ∆z i(z,t) + L ∆z ∂t i(z,t)

(1.4)



 i(z,t) − i(z + ∆z,t) ∂
= G ∆z v(z + ∆z,t) + C ∆z v(z + ∆z,t)
∂t
Next divide both sides by ∆z and take the limit for ∆z → 0. The incremental ratios in the left
hand side become partial derivatives with respect to z and, noting the continuity of v(z,t), we
obtain the transmission line equations (Telegrapher’s equations, Heaviside 1880):

 ∂ ∂

 − ∂z v(z,t) = R i(z,t) + L ∂t i(z,t)

(1.5)


 ∂
 − i(z,t) = G v(z,t) + C ∂
v(z,t)
∂z ∂t
It is to be remarked that any other disposition of the circuit elements, such as those of Fig. 1.4,
leads exactly to the same differential equations.

Figure 1.4. Alternative equivalent circuits of an element of transmission line.

Equations (1.5) are a system of first order, coupled, partial differential equations, that must
completed with boundary and initial conditions. Usually, a line connects a generator to a load, as
sketched in Fig. 1.5, where, for simplicity, both the load impedance and the internal impedance of
the generator have been assumed real. This is the simplest circuit comprising a transmission line.
It is clear that the boundary conditions to be associated to (1.5) are:

In z = 0 e(t) − Rg i(0,t) = v(0,t) ∀t ≥ 0

(1.6)
In z = L v(L,t) = RL i(L,t) ∀t ≥ 0

where e(t) is a given causal function. Moreover, the initial condition that specify the initial state
of the reactive components (only of the line, in this case) is

v(z,0) = v0 (z) 0≤z≤L

i(z,0) = i0 (z) 0≤z≤L

9
Renato Orta - Transmission Line Theory (Nov. 2012)

Rg

+
e(t) RL

0 L

Figure 1.5. Fundamental circuit comprising a generator and a load connected by a transmission line.

where v0 (z) e i0 (z) are known (real) functions. Typically, at t = 0 the line is at rest and, hence,

v0 (z) ≡ 0 e i0 (z) ≡ 0 0 ≤ z ≤ L

We observe that (1.5) is a system of homogeneous equations, i.e. without forcing term. Concerning
the boundary conditions (1.6), the first is nonhomogeneous, the second is homogeneous. In the
case the line is initially at rest, we can say that the system is excited via the boundary condition
in z = 0.
In the case the load network contains reactive elements, the boundary condition is not of
algebraic type, but is formulated as an ordinary differential equation of the type
d d
D( ) v(L,t) = N ( ) i(L,t) (1.7)
dt dt
to be completed with the initial conditions for the reactive components of the load network. D
and N are two formal polynomials in the operator d/dt. For example, if the load network is that
of Fig. 1.6, eq. (1.7) takes the form:

d d d2 1
v(L,t) = R i(L,t) + L 2 i(L,t) + i(L,t)
dt dt dt C
The initial conditions to be specified are vc (0) and i(0), which express the voltage across the
capacitor and the current in the inductor at the time t = 0.

i(L,t)
R

v(L,t)
L

C vC(L,t)

Figure 1.6. Load network with reactive components, consisting of a series connection of a resistor
R, a capacitor C and an inductor L.

In the applications, not always is a transmission line excited only at its ends. In problems of
electromagnetic compatibility one studies the effect of a wave that impinges on the transmission

10
Renato Orta - Transmission Line Theory (Nov. 2012)

line: the phenomenon is not a point-like excitation and can be modeled by means of a set of
voltage and current generators “distributed” along the the line with a density per unit length
◦ ◦
v (z,t) e i (z,t). In this case the equivalent circuit of a line element has the form shown in Fig. 1.7
and correspondingly eq. (1.5) become

 ∂ ∂ ◦
 i(z,t)+ v (z,t)
 − ∂z v(z,t) =
 R i(z,t) + L
∂t
(1.8)


 − ∂ i(z,t)

= G v(z,t) + C
∂ ◦
v(z,t)+ i (z,t)
∂z ∂t
◦ ◦
The functions v (z,t) and i (z,t) describe source terms and therefore are to be considered as known.

o
/∆z 5∆z v ]
++

o
i ] &∆z *∆z

Figure 1.7. Equivalent circuit of a line element ∆z when distributed generators are
present on the transmission line.

Eq. (1.8) define a non-homogeneous problem, since they contain a forcing term.
It is well known that the general solution of a linear non-homogeneous differential equation
is given by the sum of a particular solution of the non-homogeneous equation and the general
solution of the associated homogeneous equation. We are going to focus first on the homogeneous
equation. We will find that the general solution is the linear combination of two normal modes
of the system, called forward wave and backward wave. Other common names are free evolutions,
resonant solutions, proper evolutions.

1.4 Lossless lines. Wave equations and their solutions

A transmission line is called ideal when the ohmic losses in the conductors and in the insulators
can be neglected. The line equations, without sources, become in this case

 ∂v ∂i

 ∂z + L ∂t
 = 0
(1.9)


 ∂i + C ∂v

= 0
∂z ∂t

From this system of first order partial differential equations we can obtain a single second order
equation for the voltage v(z,t) alone. Differentiate the first equation with respect to z and the

11
Renato Orta - Transmission Line Theory (Nov. 2012)

second with respect to t:  2

 ∂ v ∂2i

 + L = 0
 ∂z 2 ∂z ∂t



∂2i ∂2v
+C 2 = 0
∂t ∂z ∂t
The two mixed derivatives are equal under the usual regularity conditions for i(z,t) and we obtain

∂2v ∂2v
− LC =0
∂z 2 ∂t2
This equation is known as wave equation (in one dimension) because its solutions √ (obtained by
d’Alembert in 1747) are waves propagating along the line with speed ±vph = ±1/ LC. Obviously
one of the two (1.9 ) must be associated to (1.10), in order to obtain the current i(z,t). Recall in
fact that on a transmission line, voltage and current are inextricably linked.
Observe that also the current i(z,t) obeys a wave equation identical to (1.10). To obtain it,
differentiate the first of (1.9) with respect to t and the second with respect to z.
The wave equation for an infinitely long ideal transmission line, with the initial conditions

v(z,0) = v0 (z), i(z,0) = i0 (z) (1.10)

can be solved by a change of variable technique. Define the new independent variables

ξ = z − vph t, η = z + vph t

The old variables are expressed in terms of the new ones as

1 1
z= (ξ + η), t= (η − ξ).
2 2vph

Now rewrite the wave equation in the new variables. We need the chain rule of multivariable
calculus.
∂v ∂v ∂ξ ∂v ∂η ∂v ∂v
= + = +
∂z ∂ξ ∂z ∂η ∂z ∂ξ ∂η
µ ¶
∂v ∂v ∂ξ ∂v ∂η ∂v ∂v
= + = −vph −
∂t ∂ξ ∂t ∂η ∂t ∂ξ ∂η
and also µ ¶ µ ¶
∂2v ∂ ∂v ∂v ∂ ∂v ∂v ∂2v ∂2v ∂2v
= + + + = + 2 +
∂z 2 ∂ξ ∂ξ ∂η ∂η ∂ξ ∂η ∂ξ 2 ∂ξ∂η ∂η 2
· µ ¶ µ ¶ ¸ µ ¶
∂2v ∂ ∂v ∂v ∂ ∂v ∂v 2 ∂2v ∂2v ∂2v
= v ph − vph − − v ph = v ph − 2 +
∂t2 ∂η ∂η ∂ξ ∂ξ ∂η ∂ξ ∂η 2 ∂ξ∂η ∂ξ 2
Using these two last expressions, the wave equation in the new variables becomes

∂2v
=0
∂ξ∂η
that is µ ¶
∂ ∂v
=0
∂η ∂ξ

12
Renato Orta - Transmission Line Theory (Nov. 2012)

whose solution is
∂v
= f (ξ)
∂ξ
where f is a constant with respect to η, i.e. an arbitrary function ξ. By integrating the previous
equation, we get Z
v(ξ,η) = f (ξ)dξ + f2 (η)

where f2 is an arbitrary function of η. Rewrite the previous equation as

v(ξ,η) = f1 (ξ) + f2 (η)

This is the general solution of the wave equation. We have introduced the symbol f1 (ξ) to denote
the integral of the arbitrary function f (ξ). Returning to the original variables, we get

v(z,t) = v + (z − vph t) + v − (z + vph t) (1.11)

where the more appropriate symbols v + e v − have been introduced in place of f1 e f2 .

To derive the expression of the current, consider (1.9) from which
∂i 1 ∂v
=−
∂t L ∂z
that is Z
1 ∂
i(z,t) = − v(z,t)dt.
L ∂z
From (1.11) we compute
∂v
= v +0 (z − vph t) + v −0 (z + vph t)
∂z
and
½Z Z ¾
1 +0 −0
i(z,t) = − v (z − vph t)dt + v (z + vph t)dt
L
½ Z Z ¾
1 1 +0 1 −0
= − − v (ξ)dξ + v (η)dη
L vph vph
= Y∞ {v + (z − vph t) − v − (z + vph t)}
p
were the quantity Y∞ = C/L is called characteristic admittance of the line and is measured in
Siemens, S.
In conclusion, the general solution of the transmission line equations can be written as

v(z,t) = v + (z − vph t) + v − (z + vph t)

(1.12)
i(z,t) = Y∞ v + (z − vph t) − Y∞ v − (z + vph t).

To complete the solution of the initial value problem, we must obtain the functions v + (ξ) and
v − (η) in such a way that the initial conditions (1.10) are satisfied. Now, eq. (1.12), written for
t = 0, yield
v0 (z) = v + (z) + v − (z)

i0 (z) = Y∞ v + (z) − Y∞ v − (z).

13
Renato Orta - Transmission Line Theory (Nov. 2012)

Solving by sum and difference, we find

1
v + (z) = [v0 (z) + Z∞ i0 (z)],
2
1
v − (z) = [v0 (z) − Z∞ i0 (z)].
2
In this way the functions v + e v − are determined. The solution for t > 0 is obtained by substituting
the argument z with z − vph t in v + and z + vph t in v − , as it follows from (1.12):

1 1
v(z,t) = [v0 (z − vph t) + Z∞ i0 (z − vph t)] + [v0 (z + vph t) − Z∞ i0 (z + vph t)] ,
2 2
Y∞ Y∞
i(z,t) = [v0 (z − vph t) + Z∞ i0 (z − vph t)] + [v0 (z + vph t) − Z∞ i0 (z + vph t)] .
2 2
Alternatively, these equations can be rewritten
1 Z∞
v(z,t) = [v0 (z − vph t) + v0 (z + vph t)] + [i0 (z − vph t) − i0 (z + vph t)] ,
2 2
Y∞ 1
i(z,t) = [v0 (z − vph t) + v0 (z + vph t)] + [i0 (z − vph t) − i0 (z + vph t)] .
2 2
one can immediately verify that these expression satisfy the initial conditions.
Recall that the general solution of an ordinary differential equation contains arbitrary constants,
whereas a partial differential equation contains arbitrary functions. The arbitrariness is removed
when a particular solution is constructed, which satisfies initial/boundary conditions. Note that
the electric state on the line depends on z e t only through the combinations t − z/vph e t + z/vph :
this is the only constraint enforced by the wave equation.
The solution method just presented is the classical one, obtained for the first time by d’Alembert.
It is possible also to employ another method, based on the use of Fourier transforms. This is the
only possible one in the case of finite length lossy lines and will be presented now after a short
review of phasors and Fourier transforms.

1.5 Review of Fourier transforms and phasors

It is known that for every absolutely integrable function of time f (t), i.e.
Z ∞
| f (t) | dt < ∞
−∞

the spectral representation exists:

Z ∞
1
f (t) = F (ω) ejωt dω (1.13)
2π −∞

where F (ω) is the Fourier transform, or spectrum, of f (t), defined by

Z ∞
F (ω) = f (t) e−jωt dt = F{f (t)} (1.14)
−∞

14
Renato Orta - Transmission Line Theory (Nov. 2012)

The meaning of (1.13) is that the function f (t) can be represented as a (continuous) sum of sinu-
soidal functions, each one with (infinitesimal) amplitude F (ω) dω. This representation underlines
the importance of sinusoidal functions in the analysis of linear systems. A very useful property of
Fourier transforms is the following:
½ ¾
df
F = jω F{f (t)} = jω F (ω) (1.15)
dt
In other words, there is a one-to-one correspondence between the derivative operator in time
domain and the multiplication by jω in the frequency domain.
Even if the Fourier transform is defined for complex time functions, provided they satisfy
(1.13), the physical quantities such as voltage and current are real functions. This implies that the
following relation holds:
F (−ω) = F ∗ (ω) (1.16)
i.e. the spectrum of a real function is complex hermitian; the part of spectrum corresponding to
the negative frequencies does not add information to that associated with the positive frequencies.
In the applications, very often signals are sinusoidal (i.e. harmonic), that is of the type
f (t) = F0 cos(ω0 t + φ) (1.17)
Let us compute the spectrum of this signal by means of (1.14); by Euler’s formula
Z ∞
F (ω) = F0 cos(ω0 t + φ) e−jωt dt =
−∞
Z ∞ Z ∞
F0 F0
= ej(ω0 t+φ) e−jωt dt + e−j(ω0 t+φ) e−jωt dt =
2 −∞ 2 −∞

= πF0 ejφ δ(ω − ω0 ) + πF0 e−jφ δ(ω + ω0 ) (1.18)

This spectrum consists of two “lines” (Dirac δ functions) at the frequencies ±ω0 , so that the signal
(1.17) is also called monochromatic.

F(ω)

-ω0 ω0 ω

Figure 1.8. Spectrum of a sinusoidal signal.

Let us now proceed in the opposite direction and derive the time domain signal from its spec-
trum (1.18) through the inverse transform formula (1.13):
½ Z ∞ Z ∞ ¾
F0
f (t) = ejφ δ(ω − ω0 ) ejωt dω + e−jφ δ(ω + ω0 ) ejωt dω =
2 −∞ −∞

F0 © jφ jω0 t ª
= e e + e−jφ ejω0 t =
2
© ª
= Re F0 ejφ ejω0 t (1.19)

15
Renato Orta - Transmission Line Theory (Nov. 2012)

The quantity F = F0 exp(jφ) is generally called phasor of the harmonic signal f (t) and coincides,
apart from the factor π, with the coefficient of the Dirac δ function with support in ω = ω0 .
Moreover, eq. (1.19) can be defined as the inverse transform formula for phasors.
Observe further that, calling Ph the one-to-one correspondence that associates a time-harmonic
signal to its phasor,
F = Ph{f (t)}
the following property holds ½ ¾
df
Ph = jω0 F
dt
This equation is formally identical to (1.15); Note, however, that ω denotes a generic angular
frequency, whereas ω0 is the specific angular frequency of the harmonic signal under consideration.
Because of the very close connection between phasors and Fourier transforms, we can say that
any equation in the ω domain can be interpreted both as an equation between transforms and as
an equation between phasors and this justifies the use of the same symbol F for the two concepts.
It is important to remember, however, that phasors and transforms have different physical
dimensions:

• phasors have the same dimensions as the corresponding time harmonic quantity
• transforms are spectral densities.

For example, the phasor of a voltage is measured in V, whereas its transform is measured in V/Hz.
This is obvious if we consuider eq. (1.18) and note the well known property
Z ∞
δ(ω) dω = 1
−∞

which implies that the Dirac function δ(ω) has dimensions Hz−1 .

1.6 Transmission line equations in the frequency domain

Let us apply now these concepts to the ideal transmission line equations, that we rewrite here for
convenience: 
 ∂v ∂i

 ∂z + L ∂t = 0



 ∂i + C ∂v = 0

∂z ∂t
Take the Fourier transforms of both sides, observing that z is to be considered as a parameter in
this operation: 
 d
 − V (z,ω) = jω L I(z,ω)

 dz
(1.20)


 d
 − I(z,ω) = jω C V (z,ω)
dz
where V (z,ω) = F{v(z,t)} and I(z,ω) = F{i(z,t)} are the Fourier transform of voltage and
current. Note that the transmission line equations have become ordinary differential equations

16
Renato Orta - Transmission Line Theory (Nov. 2012)

in the spectral domain. Moreover, the spectral components of voltage and current at different
frequencies are uncoupled, as it is obvious since transmission lines are a linear time-invariant (LTI)
system.
Proceeding in a similar way on the wave equation (1.10), we obtain
d2
V (z,ω) + k 2 V (z,ω) = 0
dz 2
and
d2
I(z,ω) + k 2 I(z,ω) = 0
dz 2
√
where the quantity k = ω LC, with the dimensions of the inverse of a length, has been introduced.
These equations can be called Helmholtz equations in one dimension. Their counterpart in two or
three dimensions are very important for the study of waveguides and resonators. These equations
have constant coefficients (because of the assumed uniformity of the transmission line) and their
general solution is a linear combination of two independent solutions. As such one could choose
sin kz and cos kz but exp(+jkz) and exp(−jkz) have a nicer interpretation. Hence, we can write
V (z,ω) = V0+ (ω) e−jkz + V0− (ω) e+jkz
(1.21)
I(z,ω) = I0+ (ω) e−jkz + I0− (ω) e+jkz
where V0± (ω) and I0± (ω) are arbitrary constants with respect to z (but dependent on ω, of course,
which is a parameter). We must remember, however, that the transmission line equations are a
2×2 first order system (see eq. (1.9)) and hence, its solution contains only two arbitrary constants.
Then, between V0± (ω) and I0± (ω) two relations must exist, which we can find by obtaining I(z,ω)
from the first of (1.20) by substituting (1.21):
µ ¶
1 dV
I(z) = − =
jωL dz
(1.22)
1 ¡ + −jkz − +jkz
¢
= jkV0 e − jkV0 (ω) e
jωL
Note that √ r
k ω LC C 1
= = = Y∞ =
ωL ωL L Z∞
where we have introduced the characteristic admittance and characteristic impedance of the line.
The characteristic impedance is denoted by the symbol Z∞ since it coincides with the input
impedance of a semi-infinite line, as it will be shown in section 5.1. Eq. (1.22) can be rewritten as
I(z,ω) = Y∞ V0+ (ω) e−jkz − Y∞ V0− (ω) e+jkz
From the comparison between this equation and the second one of (1.21), it follows
I0+ (ω) = Y∞ V0+ (ω) e I0− (ω) = −Y∞ V0− (ω)
which are the desired relations. In conclusion, the general solution of transmission line equations
in the spectral domain are
V (z,ω) = V0+ (ω) e−jkz + V0− (ω) e+jkz
(1.23)
I(z,ω) = Y∞ V0+ (ω) e−jkz − Y∞ V0− (ω) e+jkz

17
Renato Orta - Transmission Line Theory (Nov. 2012)

To understand fully the meaning of these two equations, it is necessary to transform them back
to time domain. Consider first the simplest case, in which only one spectral component at ω0 is
present, so that the signals are monochromatic. We can use the inverse transform rule of phasors

f (t) = Re{F ejω0 t } (1.24)

so that we obtain:

v(z,t) = v + (z,t) + v − (z,t) =

= | V0+ | cos(ω0 t − k0 z + arg(V0+ )) +
+ | V0− | cos(ω0 t + k0 z + arg(V0− )) (1.25)

i(z,t) = Y∞ v + (z,t) − Y∞ v − (z,t) =

= Y∞ | V0+ | cos(ω0 t − k0 z + arg(V0+ )) +
− Y∞ | V0− | cos(ω0 t + k0 z + arg(V0− )) (1.26)
√
where k0 = ω0 LC. Consider the first term of the expression of v(z,t). It is a function of z and of
t, sketched in Fig. 1.9, called wave.
The propagation velocity of a wave (phase velocity) can be defined as the velocity an observer
must have in order to see the wave phase unchanging. It is clear that the value of the cosine
function is constant if the argument is constant. By enforcing its differential to be zero

d (ω0 t − k0 z + arg(V0+ )) = ω0 dt − k0 dz = 0

we find the condition that must be satisfied:

dz ω0 ω0 1
= = √ =√ = vph
dt k0 ω0 LC LC

Hence we say that the first term

√ of (1.25) represents a forward wave because it moves with positive
phase velocity equal top 1/ LC. Note that also the first term of the expression of the current
describes a forward wave: in particular, the current is proportional to the voltage via the char-
acteristic admittance. It is to be remarked that when the dielectric is homogeneous, so that the
propagation mode is rigorously TEM, it can be shown that
c
vph = √
²r

and, as a consequence,
²r
LC =
c2

Consider now the plots of Fig. 1.10. The first (a) shows the time evolution of the forward
voltage in a specific point of the line z = z0 . The second (b) shows the distribution of the forward
voltage on the line at a specific time instant t = t0 . The two curves are obviously periodic and we
can define two periods:

• the temporal period T = 2π/ω0 is the time interval during which the wave phase changes of
2π radians (note that ω0 is the time rate of change of the wave phase)

18
Renato Orta - Transmission Line Theory (Nov. 2012)

Figure 1.9. Tree dimensional representation of (a) a forward wave, (b) a backward wave and (c) a
stationary wave on a short circuited transmission line.

• the spatial period or wavelength λ = 2π/k0 is the distance over which the wave phase changes
by 2π radians (note that k0 is the space rate of change of the wave phase)

From this definition and from that of k0 we find at once

ω0 2π ω0 1
fλ = = =√ = vph
2π k0 k0 LC
and also T vph = λ: in other words, a wave moves over the distance of a wavelength during the
time interval of a temporal period. In the spacetime plot of Fig. 1.9 the straight lines z = vph t are
clearly recognizable as the direction of the wave crests .
Consider now the second term of the expression of the voltage (1.25), plotted in Fig. 1.9b. We
find immediately, with similar argument as above, that it describes a backward wave, moving with
negative phase velocity
ω0 1
vph = − = −√
k0 LC

19
Renato Orta - Transmission Line Theory (Nov. 2012)

λ
2 2
+
T +
v ( z0 , t ) v ( z, t 0 )
V0+
1 1
V0+
0 0

-1 -1
0 5 10 15 0 5 10 15
t z
(a) (b)
Figure 1.10. (a) Time evolution of the forward wave in a fixed point of the line and (b) distribution
of the forward voltage on the line at a specific time instant.

Moreover, the current is proportional to the voltage via the factor −Y∞ . Also in this case, the
wave crests are aligned on the straight lines z = −vph t.
In conclusion, we find again the result of Section 1.4: the general solution of the transmission
line equations is expressed as linear combination of two waves, a forward one propagating in the
direction of increasing z and a backward one, moving in the opposite direction. Each wave is made
of voltage and current that, in a certain sense, are the two sides of a same coin. It is important
to observe that the two waves are absolutely identical since the transmission line is uniform and
hence is reflection symmetric. The proportionality between voltage and current of the same wave
(called impedance relationship)
I0+ (ω) = Y∞ V0+ (ω) e I0− (ω) = −Y∞ V0− (ω)
is only apparently different in the two cases. The minus sign in the impedance relation for the
backward wave arises because the positive current convention of the forward wave is used also for
the backward one.
Forward and backward waves on the line are the two normal modes of the system. They are
independent (uncoupled) if the line is of infinite length, whereas they are in general coupled by the
boundary conditions (generator and load) if the line has finite length.
When on a transmission line both the forward and the backward wave are present with the same
amplitude, we say that a (strictly) stationary wave is present. This definition, even if ordinarily
used, is improper since a wave is always travelling at the phase speed. Actually, what is referred to
by the term stationary wave is the interference pattern of two waves. In any case, the name given
to the phenomenon is related to the fact that eq. (1.25), with | V − |=| V + | can be rewritten in
factorized form:
1 1
v(z,t) = 2 | V0+ | cos[ω0 t + (arg(V0+ ) + arg(V0− ))] · cos[k0 z − (arg(V0+ ) − arg(V0− ))] (1.27)
2 2
and
1 1
i(z,t) = 2Y∞ | V0+ | sin[ω0 t + (arg(V0+ ) + arg(V0− ))] · sin[k0 z − (arg(V0+ ) − arg(V0− ))] (1.28)
2 2
i.e. as a product of a function of z and of a function of t. Fig. 1.9c shows a spacetime plot
of v(z,t). Whereas Figs. 1.9a e b suggest, even at intuitive level, an idea of movement, this

20
Renato Orta - Transmission Line Theory (Nov. 2012)

plot is clearly characteristic of a stationary phenomenon. Further considerations will be made in

Section 3.5.

1.7 Propagation of the electric state and geometrical inter-

pretations
We have obtained the general solution of the transmission line equations in the form
V (z) = V0+ e−jkz + V0− e+jkz
(1.29)
I(z) = Y∞ V0+ e−jkz − Y∞ V0− e+jkz

where the two arbitrary constants V0+ e V0− appear. In order to understand better the meaning of
these equations, we solve the initial value problem associated to eq. (1.20). Suppose then that the
electric state of the line is given at z = 0, i.e. V (0) = V0 and I(0) = I0 are given: we want to find
the state V (z), I(z) in an arbitrary point z.
Equations (1.29) hold in any point z and, in particular, also in z = 0:
V (0) = V0+ + V0− = V0
(1.30)
I(0) = Y∞ V0+ − Y∞ V0− = I0

from which V0+ e V0− can be obtained:

V0+ = 1
2 (V0 + Z∞ I0 )
(1.31)
V0− = 1
2 (V0 − Z∞ I0 )
Substituting these relations into (1.29) we find
1
V (z) = 2 (V0 + Z∞ I0 ) e−jkz + 12 (V0 − Z∞ I0 ) e+jkz
(1.32)
1
I(z) = 2 (Y∞ V0 + I0 ) e−jkz − 12 (Y∞ V0 − I0 ) e+jkz
i.e., via Euler’s formula,
V (z) = V0 cos kz − jZ∞ I0 sin kz
(1.33)
I(z) = I0 cos kz − jY∞ V0 sin kz
This form of the solution is called stationary wave type solution whereas eq. (1.29) is called travelling
wave type solution.
It is useful to describe the propagation phenomenon on the transmission line in geometric terms.
Since voltage and current in a point of the line define the system state, we can introduce a two
dimensional complex state space (isomorphic to C2 ) each point of which correspond to a possible
operation condition of the transmission line. The state is a function of z and the corresponding
point moves on a trajectory in the state space.
In the light of these considerations, we can rewrite (1.29) in vector form:
µ ¶ µ ¶ µ ¶
V (z) + 1 −jkz − 1
= V0 e + V0 e+jkz (1.34)
I(z) Y∞ −Y∞

21
Renato Orta - Transmission Line Theory (Nov. 2012)

In other words, the state in a generic point z is obtained as a linear combination of two basis states
µ ¶ µ ¶
1 1
ψ1 = , ψ2 = (1.35)
Y∞ −Y∞

with complex coefficients V0+ e−jkz e V0− e+jkz , respectively. Obviously the two basis states are
the forward and backward waves discussed before. As in the cartesian plane of analytic geometry
different reference systems can be used, in the state space we can describe the excitation of the
line with reference to the “natural basis” V e I or to the vectors ψ1 e ψ2 . Forward and backward
voltages are then interpreted as excitation coefficients of these waves. Assuming for simplicity of
drawing that in a point of the line voltage and current are real, the situation is that sketched in
Fig. 1.11. In the general case, four real dimensions would be necessary for this type of plot.

ψ1
ψ2

Figure 1.11. Geometric representation of the electric state of a transmission line.

It is convenient to rewrite also eq. (1.33) in vector form:

µ ¶ µ ¶µ ¶
V (z) cos kz −jZ∞ sin kz V0
= (1.36)
I(z) −jY∞ sin kz cos kz I0
| {z }
[T (z,0)]

where we have introduced the matrix [T (z,0)] which relates the state in a generic point z to that
in the origin z = 0. This matrix is known as transition matrix in the context of dynamical systems
(in which the state variables are real and the independent variable is time) but coincides with the
chain matrix (ABCD) of the transmission line length, viewed as a two-port device.
The basis of the two vectors ψ1 e ψ2 has peculiar properties with respect to all the other bases
that could be introduced in the state space. Assume for instance that the backward wave is not
excited in the point z = 0: it will be absent on the whole transmission line. Indeed, in the origin
µ ¶ µ ¶
V0 1
= V0+ (1.37)
I0 Y∞

By means of (1.36) we find immediately

µ ¶ µ ¶
V (z) 1
= V0+ e−jkz (1.38)
I(z) Y∞
In geometric terms, we can say that in the propagation the state vector remains parallel to it-
self since it is only multiplied by the scalar exp{−jkz}. In algebraic terms this state vector is

22
Renato Orta - Transmission Line Theory (Nov. 2012)

eigenvector of the transition matrix [T (z,0)], with eigenvalue exp{−jkz}. A completely analogous
property holds for the backward wave (vettore ψ2 ). For comparison, notice that if the total voltage
is zero in a point, it is not identically zero on the line (apart for the trivial case of a non excited
line). Conversely, if we wish that on a transmission line only one of the basis states is excited,
it is necessary that V0 /I0 = ±Z∞ . Otherwise, both modes are excited, with coefficients given by
(1.31). Hence these equations describe the change of basis. Note that ψ1 , e ψ2 are not orthogonal
(if Z∞ 6= 1Ω).

1.8 Solution of transmission line equations by the matrix

technique
In the previous sections we have found the solution of transmission line equations from the second
order equation. In this section we obtain the same result directly from the first order system, with
a more abstract technique, which has the advantage that the geometrical interpretation of forward
and backward waves as modes of the system is almost automatic.
Consider again the transmission line equations in the spectral domain

 d

 − dz V (z,ω) = jω L I(z,ω)




 − d I(z,ω)

= jω C V (z,ω)
dz
The system can be rewritten as a single differential equation for the state vector ψ(z), whose
components in the natural basis are total voltage and current. Suppose we know voltage and
current in the point z0 of the line and we want to compute the corresponding values in an arbitrary
point z. In other words, we want to solve the initial value problem
 d

 − ψ(z,ω) = jω A · ψ(z,ω)

 dz
µ ¶ (1.39)

 V0

 ψ(z,ω)|z=z0 = ψ0 =
I0

where we use a double underline to denote matrices and

µ ¶
0 L
A=
C 0

It is well known that the solution of this problem can be written in the form
© ª
ψ(z,ω) = exp −jωA(z − z0 ) · ψ 0 (1.40)

where the exponential of the matrix is defined by the series expansion:

© ª 1
exp −jωA(z − z0 ) = I − jωA(z − z0 ) − ω 2 A2 (z − z0 )2 + . . . (1.41)
2!
where I is the identity matrix.

23
Renato Orta - Transmission Line Theory (Nov. 2012)

It is simple to verify that (1.40) satisfies (1.39). Indeed, by differentiating (1.41) term by term,
(which is allowed by the fact that the series converges uniformly for all matrices A and all (complex)
z) we find
d © ª © ª
− exp −jωA(z − z0 ) = jωA exp −jωA(z − z0 )
dz
so that
d ³ © ª ´ d ¡ © ª¢
− exp −jωA(z − z0 ) · ψ 0 = − exp −jωA(z − z0 ) · ψ 0
dz dz
© ª
= jωA exp −jωA(z − z0 ) · ψ 0

= jωA · ψ
The matrix exponential can be computed directly by eqs. (1.41) and (1.40). Note first that
µ ¶2n
0 L √
A2n = = ( LC)2n I
C 0
and µ ¶2n+1 µ ¶
2n+1 0 L √ 2n 0 L
A = = ( LC)
C 0 C 0
Hence the series (1.41) reduces to
© ª 1 √ 1 √
exp −jωA(z − z0 ) = [1 − (ω LC(z − z0 ))2 + (ω LC(z − z0 ))4 + . . .] I+
2! 4!

1 √
− j[ω(z − z0 ) − (ω(z − z0 ))3 ( LC)2 +
3!

1 √
+ (ω(z − z0 ))5 ( LC)4 + . . .] A
5!
We modify slightly the previous equation as follows
© ª 1 √ 1 √
exp −jωA(z − z0 ) = [1 − (ω LC(z − z0 ))2 + (ω LC(z − z0 ))4 + . . .] I+
2! 4!

1 √ 1 √
− j √ [ω(z − z0 ) LC − (ω(z − z0 ))3 ( LC)3 +
LC 3!

1 √
+ (ω(z − z0 ))5 ( LC)5 + . . .] A
5!
In the first square parenthesis we recognize the Taylor expansion of cos k(z − z0 ) and in the second
one the expansion of sin k(z − z0 ). Moreover
 q 
L µ ¶
1 0 C  0 Z∞
√ A= q =
LC C
0 Y∞ 0
L

so that, in conclusion,
µ ¶
© ª 0 Z∞
exp −jωA(z − z0 ) = cos k(z − z0 )I − j sin k(z − z0 )
Y∞ 0

24
Renato Orta - Transmission Line Theory (Nov. 2012)

i.e. µ ¶
© ª cos k(z − z0 ) −jZ∞ sin k(z − z0 )
exp −jωA(z − z0 ) =
−jY∞ sin k(z − z0 ) cos k(z − z0 )
Even if we are now in the position to obtain the solution of the initial value problem (1.39), we
will use instead a different method that allows a more fruitful physical interpretation. Indeed,
it is known that a function of a (diagonalizable) matrix is easily computed in the basis of its
eigenvectors, because in this basis the matrix is diagonal. Hence we compute first the eigenvectors
of A, by solving
½µ ¶ µ ¶¾ µ ¶
0 L 1 0 u1
−λ =0
C 0 0 1 u2
We find immediately
 µ ¶
√


 λ1 = LC [u1 ] = p1
λ= µ C/L ¶
√


 λ2 = − LC [u2 ] = p1
− C/L

The eigenvectors have an arbitrary norm, since they are solutions of a homogeneous problem; we
have chosen to set to one their first component (i.e. the “voltage” component). Notice that they
coincide with the basis states of (1.35).
Define the modal matrix M , whose columns are the two eigenvectors :
 
r1 1
r
M = C C 
−
L L
The matrix M , together with the eigenvalue diagonal matrix, satisfies
µ ¶ µ ¶
0 L λ1 0
M =M . (1.42)
C 0 0 λ2

It can be shown that if f (x) is an analytic function, then

µµ ¶¶ µ ¶
0 L f (λ1 ) 0
f M =M
C 0 0 f (λ2 )

from which, by left multiplication by M −1 ,

µµ ¶¶ µ ¶
0 L f (λ1 ) 0
f =M M −1 .
C 0 0 f (λ2 )

Applying this property to the exponential of the matrix in (1.40), we obtain:

µ ¶ µ ¶ µ ¶
V (z,ω) exp{−jk(z − z0 )} 0 −1 V (z0 ,ω)
=M M (1.43)
I(z,ω) 0 exp{+jk(z − z0 )} I(z0 ,ω)

where µ ¶
exp{−jk(z − z0 )} 0
Td = (1.44)
0 exp{+jk(z − z0 )}

25
Renato Orta - Transmission Line Theory (Nov. 2012)

√
is the evolution matrix in the modal basis and k = ω LC. The inverse of M is
 r 
L
1 1 
[M ]−1 =  rC 
2 L 
1 −
C
so that (1.43) is rewritten as
µ ¶ µ ¶µ ¶
V (z,ω) cos k(z − z0 ) −jZ∞ sin k(z − z0 ) V (z0 ,ω)
= (1.45)
I(z,ω) −jY∞ sin k(z − z0 ) cos k(z − z0 ) I(z0 ,ω)
| {z }
[T (z,z0 )]

This equation is identical to (1.36), apart from the fact that the initial point is in z = z0 instead
of the origin. Eq. (1.45) is the final result of the computation, but (1.43) is fundamental for the
interpretation, because it makes explicit the change of basis, from the natural basis V , I to the
modal basis of forward and backward waves. Fig. 1.12 shows pictorially the method described.

natural V0  V ( z ) 
   
basis I   
 0  I ( z) 

−1
M M

+
Td +
modal V0  V0 ( z ) 
   
basis V −  V − ( z ) 
 0   0 

evolution
initial point z0 final point z

Figure 1.12. Method of solution of transmission line equations

26
Chapter 2

Parameters of common
transmission lines

2.1 Introduction
In chapter 1 we have obtained the transmission line equations on the basis of a phenomenological
model that contains four primary parameters: L (inductance per unit length, p.u.l.), R (resistance
p.u.l.), C (capacitance p.u.l.), G (conductance p.u.l.). The expressions that yield these parameters
as a function of the geometry of the structure require the solution of Maxwell equations for the
various cases. In this chapter we limit ourselves to a list of equations for a number of common
structures: the reader can consult the books in the bibliography for further details . In particular,
we show only the expressions of the inductance and capacitance p.u.l. The parameters related to
the losses will be shown in chapter 4.

2.2 Coaxial cable

The coaxial cable is a transmission line consisting of two coaxial cylindrical conductors, separated
by a dielectric (see Fig. 2.1). The two conductors, here shown as homogeneous, are often made of
braided small diameter copper wires.
If ²r denotes the relative permittivity of the insulator, the line parameters are given by:
µ ¶
2π²0 ²r µ0 D
C= , L= log , (2.1)
log(D/d) 2π d
r µ ¶
µ0 1 D 60 D
Z∞ = log ≈ √ log( ), (2.2)
²0 ²r 2π d ²r d
c
vf = √ , (2.3)
²r
where the logarithms are natural (basis e). Fig. 2.2 shows a plot of Z∞ , L e C versus the ratio of
the conductor diameters. Fig. 2.1 shows the field lines of the electric and magnetic fields of the
TEM mode, the fundamental one of this structure viewed as a waveguide. We can observe that

27
Renato Orta - Transmission Line Theory (Nov. 2012)

d D

Figure 2.1. Coaxial cable. The field lines of the electric field are shown by solid lines, those
of the magnetic field by dashed lines.

Figure 2.2. Parameters of the coaxial cable vs.the geometrical dimensions.

the electric field configuration is that of a cylindrical capacitor, consistently with the fact that the
TEM mode has zero cutoff frequency. If the operation frequency increases, a point is reached in
which higher order modes start to propagate. The maximum frequency for which the coaxial cable
is single mode is approximately
2vf
fmax = , (2.4)
π(D + d)
The corresponding minimum wavelength is
π
λmin = (D + d). (2.5)
2
The electric field in the cable is radial and its magnitude is given by
V (z) 1
E(ρ,ϕ,z) =
log(D/d) ρ

28
Renato Orta - Transmission Line Theory (Nov. 2012)

where V (z) is the voltage. Hence the maximum electric field, not to be exceeded in order to avoid
sparks, is on the surface of the inner conductor and has the value

V (z) 1
Emax =
log(D/d) d

Example

Compute the parameters of a cable, with inner conductor diameter d =1.6 mm, outer conductor diameter
D = 5.8 mm, ²r = 2.3.
√
Applying the previous formulas we get L = 0.26 µH/m, C = 99.35 pF/m, Z∞ = 50.92 Ω, vf /c = 1/ ²r
= 65.9%, fmax = 17.0 GHz. The normalized maximum electric field is Emax = 485.3V/m if the voltage V
is 1V.

It is to be remarked that the coaxial cable is an unbalanced line, which means that the return conductor
is connected to ground. Hence the voltage of the inner conductor is referred to ground.

2.3 Two-wire line

The two-wire line consists of two parallel cylindrical conductors. This structure has a true TEM mode only
if the dielectric that surrounds the conductors is homogeneous and the formulas reported hereinafter refer
to this case. In practice, of course, the conductors are embedded in a thin insulating support structure,
which causes the fundamental mode to be only approximately TEM.
The parameters of the two-wire transmission line, whose geometry is shown in Fig. 2.3 are:
π²0 ²r µ0
C= , L= cosh−1 (D/d), (2.6)
cosh−1 (D/d) π
r µ ¶ µ ¶
1 µ0 D 120 D
Z∞ = cosh−1 ≈ √ cosh−1 , (2.7)
π ²0 ²r d ²r d
c
vf = √ . (2.8)
²r
It may be useful to recall that
p
cosh−1 x = log(1 + x2 − 1) ≈ log(2x), se x À 1. (2.9)

Figure 2.3. Two-wire transmission line. The field lines of the electric field are shown solid,
those of the magnetic field dashed.

29
Renato Orta - Transmission Line Theory (Nov. 2012)

Example
Compute the parameters of a two-wire line, in which the wires have a diameter of 1.5 mm and a separation
of 5.0 mm and are located in air.
we find that C = 14.84 pF/m, L = 750 nH/m, Z∞ = 224.71, vf = c.

It is to be remarked that the TEM fields are non negligible up to large distance from the line itself,
so that the two-wire line is never isolated from the other nearby conductors, which entails problems of
electromagnetic compatibility. On the contrary, in a coaxial cable with sufficiently good outer conductor,
the operation of the line is completely shielded from external interference. For this reason, the two-wire
line is always used in a balanced configuration, i.e. the two wires have opposite potentials with respect to
ground.

2.4 Wire on a metal plane

This line consists of a single wire running parallel to a grounded metal plate, see Fig. 2.4a. If the metal
plate were infinite, this line would be perfectly equivalent to a two-wire line, because of the image theorem
(Fig. 2.4b). When the ground plane is finite, the equivalence is only approximate, but if its size is much
larger than the distance h between the wire and the plane, the errors are negligible.

d d

D= 2h

(a) (b)

Figure 2.4. (a) Wire on a metal plane and (b) equivalent two-wire transmission line.

The parameters of the two-wire line are:

π²0 ²r µ0
C= , L= cosh−1 (2h/d), (2.10)
cosh−1 (2h/d) π
r µ ¶ µ ¶
1 µ0 2h 120 2h
Z∞ = cosh−1 ≈ √ cosh−1 , (2.11)
π ²0 ²r d ²r d
c
vf = √ . (2.12)
²r

Example
Consider a wire with diameter d = 3.2 mm in air, placed at an height h = 5.74 cm on a ground plane.
We find C = 6.51 pF/m, L = 1.71 µH/m and Z∞ = 512.4 Ω.

30
Renato Orta - Transmission Line Theory (Nov. 2012)

2.5 Shielded two-wire line

To avoid the electromagnetic compatibility problems of the two-wire line, the structure of Fig. 2.5 can be
used. Note that this is a three conductor line (two plus a grounded one). In this case there are two TEM

2h

Electric field
D Magnetic field
d

Figure 2.5. Shielded two-wire line and field configuration of the symmetric (balanced) TEM mode.

modes, a symmetric (balanced) one where the potentials of the two inner conductors are symmetric with
respect to that of the outer one, connected to ground, and an asymmetric (unbalanced) one, with different
parameters. The parameters for the symmetric mode can be computed from the following equations:
µ ¶
π²0 ²r µ0 2h(D2 − h2 )
C= µ ¶, L= log , (2.13)
2h(D2 − h2 ) π d(D2 + h2 )
log
d(D2 + h2 )
r µ ¶
1 µ0 2h(D2 − h2 )
Z∞ = log , (2.14)
π ²0 ²r d(D2 + h2 )
c
vf = . (2.15)
²r

Example
Consider a shielded two-wire line with diameter of the outer conductor D = 100 mm, inner conductors
with diameter d = 15 mm e spacing 2h = 50 mm.
Using the previous formulas we get: C= 25.77 pF, L = 0.43 µH, Z∞ = 129.39 Ω.

2.6 Stripline
The stripline consists of a metallic strip placed between two grounded metal planes (Fig. 2.6). This is clearly
an unbalanced structure, which is used only inside components and devices. Since the two planes have the
same potential, this is a two conductor line and the fundamental mode is TEM. The relevant parameters
cannot be expressed in terms of elementary functions. We report below an approximate expression for the
characteristic impedance, which is valid in the case the strip thickness is negligible:

30π b
Z∞ ≈ √ (2.16)
²r weff + 0.441b

31
Renato Orta - Transmission Line Theory (Nov. 2012)

w
b

Figure 2.6. Stripline geometry .

where the equivalent strip width weff is computed from

(
weff w 0 if w/b > 0.35,
= − ³ w ´2 (2.17)
b b 0.35 − if w/b < 0.35
b
The phase velocity, as for all TEM structures is given by
c
vf = √ . (2.18)
²r
The previous equations are appropriate in an analysis problem, in which the dimensions of the structure
are known. For the design activity, in which the dimensions are to be determined in order for the line
to have a desired characteristic impedance, we can use the following equations, obtained by inversion of
(2.16) and (2.17): ½ √
w x √ if ²r Z∞ < 120 Ω,
= √ (2.19)
b 0.85 − 0.6 − x if ²r Z∞ > 120 Ω
where
30π
x= √ − 0.441 (2.20)
² r Z∞

Example
Design a stripline with characteristic impedance Z∞ = 50 Ω, separation between the ground planes b =
0.32 cm, ²r = 2.2. Find then the value of the propagation constant and the wavelength at the frequency
f = 10 GHz and the delay τ = l/vf introduced by line lentgth l = 5 cm.
√
Since Z∞ ²r = 74.2 Ω (< 120 Ω) we compute x = 0.830 by means of (2.20) and this is already the value
of w/b. Hence w = 0.266 cm. then the propagation constant is computed from
√
ω ω 2πf ²r
k= = √ = = 3.1065 cm−1
vf c/ ²r c
and √
2π l l ²r
λ= = 2.0212 cm, τ = = = 0.247 ns.
k vf c

Fig. 2.7 shows plots of the characteristic impedance of a stripline where the strip thickness f is non
negligible.

2.7 Microstrip
A microstrip consists of a conducting strip deposited on a dielectric layer, whose lower face is covered with
a metal ground plane, as shown in Fig. 2.8. Since the transverse cross section is not homogeneous, the

32
Renato Orta - Transmission Line Theory (Nov. 2012)

Figure 2.7. Characteristic impedance of a stripline vs. its dimensions.

h εr Ground conductor

Figure 2.8. Microstrip geometry.

fundamental mode is not rigorously TEM. In practice, the longitudinal field components are very small
with respect to the transverse ones and the so called “quasi-TEM approximation” is used. Even in this
case, only approximate formulas are available for the characteristic impedance. In an analysis problem, in
which the dimensions of the line are known, we compute first an equivalent dielectric constant ²eff , which
is a weighted average of the permittivities of air and of the substrate:
à !
²r + 1 1
²eff = 1+ p . (2.21)
2 1 + 12h/w

The phase speed is computed as always, but exploiting this effective permittivity
c
vf = √ (2.22)
²eff

33
Renato Orta - Transmission Line Theory (Nov. 2012)

and the characteristic impedance is given by

 µ ¶

 √60 log 8h + w if
w
< 1,

 ²eff w 4h h

Z∞ = (2.23)

 120π w
 √ ³w
 ³w ´´ if >1
 ²eff + 1.393 + 0.667 log + 1.44 h
h h
where natural logarithms are used.
For design these formulas are not convenient and the following are used instead. First of all, three
auxiliary quantities are computed:
r µ ¶
Z∞ ² r + 1 ²r − 1 0.11
A= + 0.23 + (2.24)
60 2 ²r + 1 ²r
377π
B= √ (2.25)
2Z∞ ²r
0.61
C = log(B − 1) + 0.39 − (2.26)
²r
Next 

 8eA w

 if < 2,
 e −2
2A h
w
= · ¸ (2.27)
h 
 2 ²r − 1 w


 B − 1 − log(2B − 1) + C if >2
π 2²r h

Example
Compute the width w and length l of a microstrip with characteristic impedance Z∞ = 50 Ω, which
introduces a phase shift of 90◦ at the frequency f = 2.5 GHz. The substrate thickness is 1/2000 and ²r =
2.2.
We compute A = 1.159, B = 7.985 and C = 2.056. Moreover, from the first of (2.27) we get w/h = 3.125.
Since this result is greater than 2, it is not acceptable. From the second, instead, we get w/h = 3.081,
which is in the domain of vality of the equation and hence it is acceptable. From this w = 0.391 cm results.
Next, from (2.21) the effective dielectric constant is computed, ²eff = 1.88. Then the propagation constant
is given by √
2πf ²eff
k= = 71.87 rad/m = 41.18◦ /cm.
c
If the phase shift must be kl = π/2, we obtain l = 2.19 cm.

Fig. 2.9 and Fig. 2.10 show the plots of ²eff versus w/h in the two ranges of wide and narrow strip, for
various values of ²r of the substrate. Fig. 2.11 and Fig. 2.12 show the analogous plots of the characteristic
impedance Z∞ .
Note that the effective permittivity ²eff given by (2.21) does not depend on frequency, as it is to be
expected in the case of a TEM mode. If we desire a more accurate model, which takes into account the
frequency dispersion of ²eff due to the longitudinal field components, we can use the approximate formula
(Getzinger, 1973)
²r − ²eff (0)
²eff = ²r − (2.28)
1 + (f 2 /fp2 ) G
where ²eff (0) is the zero frequency value given by (2.21) and the other parameters are

fp = Z∞0 /(2µ0 h) (2.29)

34
Renato Orta - Transmission Line Theory (Nov. 2012)

or
fp (GHz) = 0.398Z∞0 /h(mm) (2.30)
and
G = 0.6 + 0.009Z∞0 . (2.31)
where Z∞0 is the zero frequency characteristic impedance (in Ω). The characteristic impedance at the
operating frequency is then computed by (2.23) with this value of ²eff (f ).

Figure 2.9. Effective permittivity ²eff versus microstrip dimensions (wide strip approximation).

35
Renato Orta - Transmission Line Theory (Nov. 2012)

Figure 2.10. Effective permittivity ²eff versus microstrip dimensions (narrow strip approximation).

Figure 2.11. Characteristic impedance Z∞ versus microstrip dimensions (wide strip approximation).

36
Renato Orta - Transmission Line Theory (Nov. 2012)

Figure 2.12. Characteristic impedance Z∞ versus microstrip dimensions (narrow strip approximation).

37
Chapter 3

Lossless transmission line circuits

3.1 Introduction
In Chapter 1 we have obtained the general solution of the transmission line equations. With this result
in our hands, we can start to study some simple circuits. The fundamental concepts we are going to
introduce are the local impedance on a line and the reflection coefficient. The relationship between these
two quantities is displayed in graphic form by means of a famous plot, called Smith chart, which can be
considered the trademark of microwave circuits. Next we discuss the power flow on the transmission line.
Finally, we indicate how a shorted transmission line of suitable length can be used to realize capacitors,
inductors or resonators that can work at high frequencies, where ordinary lumped parameter components
are not available.

3.2 Definition of local impedance

In the analysis of lumped parameter circuits a fundamental quantity is the impedance of an element,
defined as the ratio between the phasors of the voltage at the terminals and that of the ingoing current.
In the case of a transmission line terminated with a load ZL we can define a local impedance Z(z), whose

IL I(z)

VL V ( z)
VL ZL = Z ( z) = V(z) ZL
IL I ( z)

(a) (b) 0 z

Figure 3.1. (a) Impedance of a one-port circuit element and (b) local impedance on a transmission line.

value depends on the longitudinal coordinate z:

V (z)
Z(z) = (3.1)
I(z)

38
Renato Orta - Transmission Line Theory (Nov. 2012)

Substitute in this equation the expressions (1.33) of voltage and current on the line:
V0 cos kz − jZ∞ I0 sin kz
Z(z) =
I0 cos kz − jY∞ V0 sin kz
(3.2)
V0 − jZ∞ I0 tan kz
=
I0 − jY∞ V0 tan kz
Note that the origin has been placed on the load, so that V0 and I0 are the load voltage and current.
Then the local impedance Z(0) = V0 /I0 coincides with the load impedance ZL , and the previous equation
becomes
ZL − jZ∞ tan kz
Z(z) = (3.3)
1 − jY∞ ZL tan kz
It is convenient to introduce the normalized impedance ζ(z) = Z(z)/Z∞ . Its transformation law is easily
deduced from the previous equation:
ζL − j tan kz
ζ(z) = (3.4)
1 − jζL tan kz
Obviously this formula allows the computation of the input impedance of a transmission line length loaded
by the normalized impedance ζL . This equation defines a curve in the complex plane ζ with z as parameter.
It is clearly a closed curve, due to the periodicity of the tangent function, which is completed when the
variable z increases by λ/2. This curve is shown in Fig. 3.2 and it can be shown to be a circumference.
The intersections with the real axis, rmax e rmin have the property

rmax rmin = 1

Consider now some particularly important examples.

x ζ0

rmin 1 rmax r

Figure 3.2. Representation in the complex plane of the normalized impedance ζ = r+jx
of the curve ζ(z) defined by (3.4).

Example 1
Shorted piece of lossless transmission line of length l, as shown in Fig. 3.3a.
We have
ζL = 0
ζ(z) = −j tan kz (3.5)
Zing = jXing = jZ∞ tan kl

Note that this input reactance is purely imaginary, as it is to be expected in the case of a lossless circuit
of finite size.

39
Renato Orta - Transmission Line Theory (Nov. 2012)

10
X ing
Z∞ 5

-5
Zing
-10
0 0.25 0.5 0.75 1 1.25
kl
(a) (b) 2π

Figure 3.3. (a) Shorted transmission line and (b) corresponding input reactance.

If we choose the line length conveniently, we can obtain any input reactance, either inductive or
capacitive. If the line is λ/4 long, the input impedance is that of an open circuit. We observe that the
input impedance is a periodic function of kl with period π.
Suppose now to fix a certain value of the line length, say l0 . Recalling that k = ω/vf , we note that the
input reactance is a function of frequency:

ωl0
Xing = Z∞ tan
vf

and, obviously, the plot of this function is still given by Fig. 3.3b.
Note also that Xing (ω) is an ever increasing function of frequency, as typical of all lossless circuits,
lumped or distributed (Foster theorem). Typical of the distributed parameter circuits is that Xing (ω) is a
periodic meromorphic function. On the contrary, the input impedance of a lumped parameter circuit is a
rational function, i.e. can always be written as the ratio of two polynomials.
We can observe that in the neighborhood of f0 = vf /(2l0 ), i.e. of that frequency for which the line
is half wavelength long, the input reactance Xing (ω) has a behavior similar to that of the reactance of a
series LC resonator. Analogously, in the neighborhood of f0 = vf /(4l0 ), for which the line is λ/4 long, the
line behaves as a shunt resonator.

Example 2
Length of lossless transmission line terminated with an open circuit.
We have
ζL → ∞
ζ(z) = j cot kz
Zing = jXing = −jZ∞ cot kl
The behavior is analogous to that of the shorted line, apart from a kl = π/2 translation of the plot.

Example 3
Length of lossless transmission line terminated with a reactive load.

40
Renato Orta - Transmission Line Theory (Nov. 2012)

10
X ing
Z∞ 5

-5
Zing
-10
0 0.25 0.5 0.75 1 1.25
kl
(a) (b) 2π

Figure 3.4. (a) Open circuited length of lossless transmission line and (b) correspond-
ing input reactance.

We find
ZL
ζL = jxL =
Z∞
xL − tan kz
ζ(z) = j
1 + xL tan kz
It is useful to set xL = tan φL because the previous equation becomes

tan φL − tan kz
ζ(z) = j = j tan(φL − kz)
1 + tan φL tan kz
from which we get
Zing = jXing = jZ∞ tan(kl + φL )
We see that changing the load produces a rigid displacement of the input reactance plots.

X ing 10
Z∞ 5

XL xL 0

Zing -5

-10
0 0.375 0.875 1.25
kl
(a) (b) 2π

Figure 3.5. (a) Length of lossless transmission line closed on a reactive load and (b)
corresponding input reactance.

Example 4
Length of lossless transmission line, terminated with the characteristic impedance Z∞ .

41
Renato Orta - Transmission Line Theory (Nov. 2012)

We find
ζL = 1
ζ(z) = 1
Zing = Z∞

The line is said to be matched and this is the only case in which the input impedance does not depend
on the line length.

2
Ring
Z ∞ 1.5
1
Z∞
Zing 0.5

0
0 0.25 0.5 0.75 1 1.25
kl
(a) (b) 2π

Figure 3.6. (a) Transmission line terminated with the characteristic impedance and (b) corre-
sponding input resistance (Xing = 0).

Example 5
l = λ/4 length of lossless transmission line, terminated with a generic impedance ZL .
If l = λ/4, the argument of the tangent in (3.4) is π/2 and we are in presence of an undetermined form. If
we evaluate the limit of ζ(z) for z → −λ/4 by de l’Hospital rule we find

2
1 Z∞
ζing = , Zing = (3.6)
ζL ZL

This length of transmission line behaves as a normalized impedance inverter and is commonly employed
to realize impedance transformers, discussed in Section 6.3.5.

ZL
Zing

λ 4

Figure 3.7. λ/4-length of lossless transmission line, terminated with a generic impedance ZL .

42
Renato Orta - Transmission Line Theory (Nov. 2012)

+ Zg
Vg ZL

ZA A B

Figure 3.8. Circuit consisting of a generator and a load, connected by a transmission line.

Example 6
Analysis of a complete circuit.
We can now perform the complete analysis of a simple circuit, consisting of a generator and a load,
connected by a transmission line. Compute the impedance seen by the generator, ZA . This is also the
input impedance of a piece of transmission loaded by ZL , hence it is
ZL + jZ∞ tan kl
ZA =
1 + jY∞ ZL tan kl
So we are left with the lumped parameter circuit of Fig. 3.9. We find immediately
ZA
VA = Vg
ZA + Zg
Vg
IA =
ZA + Zg
Voltage and current in all points, hence also on the load, can be computed by the (ABCD) chain matrix,
computed in Section 1.7 µ ¶ µ ¶
V (z) VA
= [T (z,zA )]
I(z) IA
where µ ¶
cos k(z − zA ) −jZ∞ sin k(z − zA )
[T (z,zA )] =
−jY∞ sin k(z − zA ) cos k(z − zA )
We will see that, in practice, another procedure is more convenient.

+ Zg
Vg ZA

Figure 3.9. Lumped equivalent circuit.

Example 7
Measurement of the parameters Z∞ and k of a length of transmission line.
The results of Examples 1 and 2 can be used as a basis for a measurement technique of the parameters
Z∞ e k of a length l of line. Recall that the input impedance Zsc of this piece, when it is shorted, is

Zsc = jZ∞ tan kl

43
Renato Orta - Transmission Line Theory (Nov. 2012)

whereas Zoc given by

Zoc = −jZ∞ cot kl
is the corresponding input impedance of the length of line when it is open. These equations can easily be
solved with respect to Z∞ e k in the form
√
Z∞ = Zcc Zca
" r #
1 Zcc
k= arctg − + nπ
l Zca
The presence in this formula of the integer n is related to the fact that the tangent function is periodic
with period π. Its value can be determined only if we know an estimate of the wavelength on the line. If
the line is shorter than λ/4, n = 0.

3.3 Reflection coefficients

In lumped circuits the state variables are voltages and currents and circuit elements are characterized by
their impedance (or admittance) that plays the role of transfer function. As discussed in section 1.7, total
voltages and currents are not the most convenient quantities for the description of the electric state on a
transmission line. The natural state variables are instead the amplitudes of forward and backward waves,
since these are the basis states of the system. Hence we must define the behavior of a generic load in the
basis of forward and backward waves.
Consider a transmission line with characteristic impedance Z∞ and phase constant k, loaded by the
impedance ZL , excited by a generator that produces a forward wave incident on the load, as depicted in
Fig. 3.10:

V+ ZL
V−

z
0

Figure 3.10. Scattering description of a load.

V inc (z) = V0+ e−jkz

I inc (z) = Y∞ V0+ e−jkz
Saying that the line in z = 0 is loaded by ZL is equivalent to saying that in this point voltage and current
are related by V (0) = ZL I(0). Obviously V inc and I inc satisfy this relation only if ZL = Z∞ : in this case
the forward wave alone is capable of satisfying the boundary condition. If instead the load impedance is
arbitrary, necessarily on the load a backward (reflected) wave must be generated
V ref (z) = V0− e+jkz
I ref (z) = −Y∞ V0− e+jkz
with a suitable amplitude V0− in such a way that the total voltage and current, sum of the forward and
backward components, satisfy the boundary condition:
V inc (0) + V ref (0) = ZL (I inc (0) + I ref (0))

44
Renato Orta - Transmission Line Theory (Nov. 2012)

that is
V0+ + V0− = ZL Y∞ (V0+ − V0− )
From this the unknown amplitude V0− is immediately deduced as

ZL Y∞ − 1 +
V0− = V
ZL Y∞ + 1 0

The proportionality coefficient that relates the backward voltage to the forward one is called voltage
reflection coefficient
−
V def V ZL Y∞ − 1 ζL − 1 ZL − Z∞
ΓL = V Γ0 = 0+ = = =
V0 ZL Y∞ + 1 ζL + 1 ZL + Z∞
This voltage reflection coefficient is the transfer function of the circuit element when forward and backward
voltages are used as state variables. Obviously also the forward and backward currents I0+ e I0− could be
used as state variables: this choice would lead to the definition of a current reflection coefficient:

I I0− −Y∞ V0−

ΓL = I Γ0 = = − V Γ0
def
+ =
I0 Y∞ V0+

Hence, the same circuit element can be characterized either:

• by the impedance ZL or the admittance YL (with ZL = 1/YL )

V
• or by the voltage Γ or current I Γ reflection coefficient (with V
Γ = − I Γ)

The equations that relate the reflection coefficients to the corresponding normalized impedances and ad-
mittances are
V ζ −1 y−1 1 + VΓ 1 − IΓ
Γ= =− , ζ= = (3.7)
ζ +1 y+1 1 − VΓ 1 + IΓ
V
I ζ −1 y−1 1− Γ 1 + IΓ
Γ=− = , y= VΓ
=
ζ +1 y+1 1+ 1 − IΓ
with y = 1/ζ. All these relations are fractional bilinear transformations of the general type

az + b
w=
cz + d

This class of complex variable mappings are well known and have a number of properties that will be
discussed later on.
We have seen that the transformation law of the local impedance on a transmission line is fairly
complicated. By the way, it is a fractional bilinear transformation. Since the reflection coefficients are
defined with reference to the forward and backward waves, which are the basis states of the line, it is to
be expected that their transformation law is simple. We prove now that this is the case.
The local voltage reflection coefficient in a point z is defined as the ratio of the backward and forward
voltages in that point:
V V − (z) V − e+jkz
Γ(z) = + = 0+ −jkz = V Γ0 e+j2kz (3.8)
V (z) V0 e
√
In the case of a lossless transmission line, for which k = ω LC is real the magnitude of the reflection
coefficient is independent of z, whereas its phase is proportional to z. In other words, V Γ(z) moves on a
circumference with center in the origin of the complex plane.

45
Renato Orta - Transmission Line Theory (Nov. 2012)

3.4 Energy balance

In the study of lumped circuits, the energy considerations play an important role. We extend them to the
realm of distributed circuits containing transmission lines.
Consider an ideal transmission line, in harmonic regime, loaded with the impedance ZL (see Fig. 3.11).
If in the point z voltage and current are V (z) e I(z), in the same point we can define an ingoing active
power P (z)
1
P (z) = Re{V (z) I ∗ (z)}
2
as well known from circuit theory. This power, if positive, is absorbed by the part of the circuit lying to
the right of z (note the current sign convention) and is interpreted as the power flowing in the line in the
point z. It is useful to express this power in terms of the amplitudes of the forward and backward waves,

ZL

z
z’ 0

Figure 3.11. Transmission line terminated with a generic load impedance.

since these variables yield the most natural description of the system. Recalling (1.23), we have
P (z) = 1
2
Re{[V + (z) + V − (z)] Y∞
∗
[V +∗ (z) − V −∗ (z)]} =

∗
= 1
2
Re{Y∞ [|V + (z)|2 − |V − (z)|2 ] + Y∞
∗
(V − (z)V +∗ (z) − V + (z)V −∗ (z))} =

∗
= 1
2
Re{Y∞ [|V + (z)|2 − |V − (z)|2 ] − j2Y∞
∗
Im{V + (z)V −∗ (z)}}
For an ideal line Y∞ is real and then
1 1 |V + |2
P (z) = Y∞ |V + |2 − Y∞ |V − |2 = (1 − | V Γ|2 )
2 2 2Z∞
We can make the following remarks:
• Since | V Γ| = constant on an ideal line, the net power flowing is the same in every point of the line.
This is obviously related to the fact that an ideal line is lossless . Hence the power absorbed by the
load impedance ZL is PL = P (0) = P (z).
• In a lossless line the net active power flowing in a point is the difference between the active powers
flows associated to the forward and backward waves. We can also say that the net power is the
difference between the incident and the reflected power. Hence the two waves are power-orthogonal
(i.e. power uncoupled).
• If V Γ = 0 (i.e. ZL = Z∞ ), the whole incident power is absorbed by the load, which is said to be a
“matched” load for the line. Conversely, one says that, in these conditions, the line is matched. The
net power coincides with the incident one, due to the absence of the reflected wave.
• If | V Γ| = 1, the reflected power is equal to the incident one, and consequently the net flowing power
is zero. This condition takes place when the load is a pure reactance. Indeed, if ZL = jXL
V jXL − Z∞
Γ=
jXL + Z∞
and numerator and denominator have the same magnitude.

46
Renato Orta - Transmission Line Theory (Nov. 2012)

• For a passive load, the reflected power is smaller or equal to the incident one, hence | V Γ| ≤ 1. This
condition is equivalent to Re{ZL } ≥ 0, as we will prove in Section 3.6.

A quantity frequently used in practice to characterize a load is the return loss RL, defined as

RL = −10 log10 |Γ|2

It yields the ratio (in dB) between the reflected power (which is “lost”, from the point of view of the load)
and the incident one. Hence, RL = 0 dB for a reactive load and RL → ∞ dB for a matched load. Return
loss e Standing Wave Ratio (VSWR), which will be introduced in the next section, express the mismatch
of the load with respect to the line in different but equivalent manners. Both of them are very often used
in practice (see Table 3.1).
The quantity 1 − | V Γ|2 is called power transmission coefficient, because it is equal to the ratio of
the power absorbed by the load and the incident power. The same coefficient, expressed in dB, is called
reflection loss.

3.5 Line voltage, current and impedance diagrams

Consider a loaded transmission line, as shown in Fig. 3.12. Voltage and current on the line can be expressed
in the following way in terms of forward and backward waves:
V
V (z) = V + (z) + V − (z) = V + (z)(1 + Γ(z))

V
I(z) = I + (z) + I − (z) = Y∞ V + (z)(1 − Γ(z))

Since the reflection coefficients for voltage and current are just opposite one of the other, for simplicity we

ZL

v z
L

Figure 3.12. Ideal transmission line terminated with a generic load impedance.

will always use the one for voltage also in the current expression. Since there is no ambiguity, the voltage
reflection coefficient will be written Γ(z) without superscripts.
Our goal now is to obtain plots of the magnitude and phase of voltage, current and impedance on the
line. Let us start with the magnitude plot, shown in Fig. 3.13. This shape is easily explained.
The magnitude of voltage and current is given by

|V (z)| = |V + (z)| |1 + Γ(z)|

|I(z)| = |Y∞ V + (z)| |1 − Γ(z)|

The first factor |V + (z)| is constant on an ideal line. As for the second, recall that Γ(z) = Γ0 exp{+j2kz}
(see Fig. 3.14). The analytic expression of |V (z)| is then

47
Renato Orta - Transmission Line Theory (Nov. 2012)

2 V ( z)
V+

0
kz (2π )
-1.5 -1 -0.5 0

2 I ( z)
I+

0
-1.5 -1 -0.5 kz (2π ) 0

3 Z (z)
Z∞

0
-1.5 -1 -0.5 0
kz (2π )

Figure 3.13. Plot of the magnitude of voltage, current, local impedance on a transmis-
sion line loaded by ZL = (1 + j)Z∞ .

Γ (z)
1 + Γ ( z)

-1 1
1 − Γ ( z)
Γ (z )
−Γ

Figure 3.14. Plot of the local reflection coefficient in the complex plane.

|V (z)| = |V + | |1 + |Γ0 | exp(j(arg(Γ0 ) + 2kz))| =

£ ¤1/2
= |V + | 1 + |Γ0 |2 + 2|Γ0 | cos(arg(Γ0 ) + 2kz)

48
Renato Orta - Transmission Line Theory (Nov. 2012)

Note that the curve is only apparently sinusoidal. It is evident from the figure that |V (z)| and |I(z)| reach
the maximum and minimum value when Γ(z) is real, and moreover:

|1 + Γ(z)|max = 1 + |Γ|

|1 + Γ(z)|min = 1 − |Γ|

In Fig. 3.14 the vectors 1 + Γ e 1 − Γ are shown. The ratio between the maximum and minimum voltage
magnitude is called VSWR (Voltage Standing Wave Ratio)

Vmax 1 + |Γ|
S= =
Vmin 1 − |Γ|

Clearly, the ratio between the maximum and minimum current magnitude is also S.
Since the magnitude of Γ(z) of a passive load is always comprised between 0 (matched load) and 1
(reactive load) (see Section 3.4), the VSWR is always greater than 1. The VSWR is normally used in
practice to specify the mismatch of a load with respect to a reference resistance (Z∞ ). Hence, VSWR,
Return loss, reflection loss and magnitude of the reflection coefficient express the mismatch in equivalent
manners. Table 3.1 yields examples of correspondences.

Table 3.1. Correspondence between values of return loss, magnitude of the reflection
coefficient, VSWR and reflection loss

Return Loss (dB) |Γ| VSWR Reflection Loss (dB)

0 1 ∞ 0
3 0.7079 5.8480 3.0206
5 0.5623 3.5697 1.6508
10 0.3162 1.9249 0.4575
15 0.1778 1.4325 0.1395
20 0.1 1.2222 0.0436
30 0.0316 1.0653 0.0043

We have seen in Fig. 3.2 that the normalized local impedance ζ(z) moves on a circumference in the complex
ζ plane. Hence, the magnitude of the impedance is an oscillating function and the maxima and minima
are reached when ζ(z) is real and their value is

Vmax |V + |(1 + |Γ|)

Rmax = = = Z∞ S
Imin |Y∞ V + |(1 − |Γ|)

Vmin |V + |(1 − |Γ|) Z∞

Rmin = = =
Imax |Y∞ V + |(1 + |Γ|) S

On the basis of these results, we find that the circumference of Fig. 3.2 has center in ζc and radius R given
by µ ¶
1 1 1 + |Γ|2
ζc = S+ =
2 S 1 − |Γ|2
and µ ¶
1 1 2|Γ|
R= S− =
2 S 1 − |Γ|2

49
Renato Orta - Transmission Line Theory (Nov. 2012)

10

+
arg(I(z)/I0 )
+ 5
arg(V(z)/V0)

0
-1.5 -1 -0.5 kz (2π ) 0

1
Z (z)
arg
Z∞

-1
kz (2π )
-1.5 -1 -0.5 0

Figure 3.15. Plot of the phase of voltage, current and local impedance on a trans-
mission line loaded by ZL = (1 + j)Z∞ .

It is clear that the circumference degenerates in the imaginary axis if |Γ| → 1, i.e. when the load becomes
purely reactive.
With a little algebra it is possible to find the expressions of the phase of voltage, current and normalized
impedance:
|Γ0 | sin(2kz + arg Γ0 )
arg V (z) = arg V0+ − kz + arctan
1 + |Γ0 | cos(2kz + arg Γ0 )
|Γ0 | sin(2kz + arg Γ0 )
arg I(z) = arg I0+ − kz − arctan
1 − |Γ0 | cos(2kz + arg Γ0 )

|Γ0 | sin(2kz + arg Γ0 )

arg ζ(z) = arctan
1 + |Γ0 | cos(2kz + arg Γ0 )
|Γ0 | sin(2kz + arg Γ0 )
+ arctan
1 − |Γ0 | cos(2kz + arg Γ0 )

The phases of voltage and current are decreasing functions for increasing z, that tend to resemble a
staircase when |Γ0 | → 1, i.e. the load becomes reactive. The normalized impedance, as already shown by
(3.4), has a periodic behavior, as it is shown in Fig. 3.15).

3.6 The Smith Chart

The Smith Chart is a graphical tool of great importance for the solution of transmission line problems.
Nowadays, since computers are widespread, its usefulness is no longer that of providing the numerical
solution of a problem, but that of helping to set up a geometrical picture of the phenomena taking place
on a transmission line. Hence, all modern codes for the Computer Aided Design (CAD) of distributed
parameter circuits. as well as measurement instruments such as the Network Analyzer, display the results
on a Smith Chart.

50
Renato Orta - Transmission Line Theory (Nov. 2012)

Mathematically, the Smith chart consists of a portion of the complex V Γ plane, on which suitable
coordinate curves are displayed. In particular, it is based on the two relations, shown in Section 3.3:
V
V ζ −1 1+ Γ
Γ= , ζ= VΓ
ζ +1 1−

where ζ = Z/Z∞ = r + jx is the normalized impedance and V Γ = V Γr + j V Γi = | V Γ| exp{j arg( V Γ)} is

the voltage reflection coefficient. Both of them are complex variables and in order to provide a graphical
picture of the previous equations, we can draw two sets of curves in the complex V Γ plane: on the curves
of the first set the real part of the impedance is constant (constant resistance curves), on those of the
second the imaginary part of the impedance is constant (constant reactance curves). In this way the
transformation ζ → V Γ and its inverse are geometrically straightforward.
It can be shown [2] that the bilinear fractional transformation (3.9) has the following property: if the
variable ζ moves on a circumference in its complex plane, also the corresponding V Γ values are located
on a circumference. For this to be true without exceptions we must regard straight lines as (degenerate)
circumferences of infinite radius.
With some algebraic manipulation it can be shown that

• The right half plane Re{ζ} ≥ 0, corresponding to passive loads, is mapped onto the unit radius
circle, | V Γ| ≤ 1; the left half plane Re{ζ} < 0, is mapped onto the region external to the unit circle,
| V Γ| > 1;
• the vertical lines of the ζ plane, (r=const.) are mapped onto the circumferences with equation
µ ¶2 µ ¶2
V r V 2 1
Γr − + Γi =
1+r 1+r
V
All of them pass through the point Γ = 1, which is a singular point of the mapping, and have the
centers on the real axis (Fig. 3.16);
• the horizontal lines of the ζ plane, (x=cost.) are mapped onto the the circumferences with equation
³ ´2 µ 1
¶2 µ ¶2
1
V
Γr − 1 + V Γi − =
x x
Also these circumferences pass through the singular point, but have their centers on a vertical line,
parallel to the imaginary axis, passing through the singular point V Γ = 1; (Fig. 3.17);
• The two sets of circumferences meet always at right angle (except at V Γ = 1), because the straight
lines r = constant and x = constant are orthogonal in the ζ plane and the mapping (3.9) is analytic
in the whole complex plane, apart from V Γ = 1).

An example of Smith chart, equipped with all the necessary scales, is shown in Fig. 3.18. Because of the
form of the evolution law of the reflection coefficient on a line, the complex number V Γ is always given
in polar form, i.e. V Γ = | V Γ| exp{j arg( V Γ)}. The Smith chart is equipped with scales to measure
magnitude and phase of V Γ.
We have seen (Eq. (3.7)) that the relation between V Γ and ζ is formally the same as that between I Γ
and y. Hence, the Smith chart can be considered equally well as:
V
• The complex Γ plane, on which constant resistance and constant reactance curves are drawn;
I
• The complex Γ plane, on which constant conductance and constant susceptance curves are drawn.

If we recall the relation between the two types of reflection coefficients, I Γ = − V Γ, it is clear that we can
exploit the Smith chart to compute the admittance corresponding to a given impedance and viceversa.
Indeed, if we know the normalized impedance ζA , we can place it on the chart by viewing the set of lines as
constant resistance and constant reactance circles: in this way V ΓA is automatically defined. The opposite

51
Renato Orta - Transmission Line Theory (Nov. 2012)

ζ PLANE Γ PLANE
5 1

0.5

Imag( Γ )
Imag( ζ )
0 0

−0.5

−5 −1
0 2 4 −1 0 1
Real( ζ ) Real( Γ )

V
Figure 3.16. Constant resistance lines in the ζ plane and their image in the Γ plane.

ζ PLANE Γ PLANE
5 Imag( Γ ) 1
Imag( ζ )

0.5
0 0
−0.5
−1
−5
0 2 4 −1 0 1
Real( ζ ) Real( Γ)

V
Figure 3.17. Constant reactance lines in the ζ plane and their image in the Γ plane, limited
to the region inside the unit circle.

point with respect to the center is I ΓA and, by reading its coordinates with respect to the set of lines,
viewed this time as constant conductance and constant susceptance circles, we obtain the desired value
of yA . This property is clearly very useful when we have to analyze transmission line circuits containing
series and parallel loads.
A more complex problem, which is solved with the same simplicity is the following. Suppose we must
find in the V ΓA plane the set of impedances with conductance greater than one. Fig. 3.20a shows hatched
the region of the I ΓA plane where g ≥ 1, Fig. 3.20b displays the symmetric region with respect to the origin.
Using the standard curves, labeled now with resistance and reactance values, to read the coordinates of
the points, solves the problem.
Now let us see how the use of the Smith chart simplifies the analysis of the circuit of Fig. 3.21, already
solved in 3.2.
From the load impedance ZL and the line characteristic impedance Z∞ , compute the normalized
impedance ζB at point B. Place ζB on the Smith chart, so that V ΓB is determined. The reflection
coefficient at point A is given by

V V V 4π
ΓA = ΓB exp(−j2klAB ) = ΓB exp(−j lAB )
λ

52
Renato Orta - Transmission Line Theory (Nov. 2012)

Figure 3.18. An example of Smith chart that can be used for analysis and design purposes

V V
Hence ΓA is on the circumference, with center in the origin, passing through ΓB and with a phase
lAB
arg( V ΓA ) = arg( V ΓB ) − 4π (3.9)
λ

53
Renato Orta - Transmission Line Theory (Nov. 2012)

V
A

I
A

Figure 3.19. Computation of impedances and admittances.

(a) (b)
Figure 3.20. Regions of the Smith chart: (a) loads with conductance g ≥ 1, (b)
impedances of the loads with g ≥ 1.

Notice that the phase values in this equation must be expressed in radians. After identifying V ΓA it is
enough to read the coordinates of this point on the basis of the constant resistance and constant reactance
circles to obtain ζA and therefrom ZA = ζA Z∞ . Find then the (total) voltage at the line input
ZA
VA = Vg
Zg + ZA
The voltage at point z is given by
V
V (z) = V + (z)(1 + Γ(z))
where
VA
V + (z) = VA+ e−jk(z+lAB ) = e−jk(z+lAB )
1 + V ΓA
and
V V
Γ(z) = ΓB e+j2kz
assuming that the origin has been chosen in B, so that the coordinate of A is z = −lAB . In conclusion

ZA e−jklAB −jkz V
V (z) = Vg (e + ΓB e+jkz )
Zg + ZA 1 + V ΓA
and
ZA e−jklAB −jkz V
I(z) = Y∞ Vg (e − ΓB e+jkz )
Zg + ZA 1 + V ΓA
To write the expression of the current, we have used
+
IA = Y∞ VA+

and
I(z) = I + (z)(1 + I Γ(z))) = I + (z)(1 − V
Γ(z)))
From a graphical point of view, it is straightforward to draw the plots of magnitude and phase of voltage
and current, taking into account that it is just necessary to study the behavior of 1 ± V Γ(z), as explained

54
Renato Orta - Transmission Line Theory (Nov. 2012)

ζB
+ Zg
Vg ZL rm l AB
ZA
λ

A B ζA

− l AB 0 z

(a) (b)

Figure 3.21. (a) Complete circuit, consisting of generator, transmission line and load
and (b) Smith chart solution.

in Section 3.5. This quantity is called Transmission coefficient for a reason explained in the next section.
Appropriate scales are provided on the chart to simplify these operations. In particular, the magnitude of
1 ± V Γ(z), in the range [0,2], is to be read on the scale with the label Transmission coefficient E or I. The
phase is read by means of an angular scale with the label Angle of Transmission Coefficient in degrees,
and with the ticks pointing toward the point V Γ = −1, see Fig. 3.14 and Fig. 3.22.
Two scales drawn on the periphery of the chart simplify the evaluation of eq.(3.9). The outer one has
the label Wavelengths toward generator and displays the quantity
µ ¶T G
l def arg( V ΓB )
= 0.25 −
λ eqB 4π

A second one, concentric with the first, is labeled Wavelengths toward load and displays the quantity
µ ¶T L
l def arg( V ΓB )
= 0.25 +
λ eqB 4π

both of them being measures of the phase of the reflection coefficient, even if the symbol suggests an
interpretation as equivalent electrical length. The presence of the 0.25 shift is related to the fact that the
origin of these scales is on the negative real axis. Moreover, the first is a clockwise scale, the second a
counterclockwise one. In this way eq. (3.9) becomes
µ ¶T G µ ¶T G
l l lAB
= + (3.10)
λ eqA λ eqB λ

The rotation sense on the chart is clockwise, as specified by the sign of the exponent in (3.9) (remember
that the phase of complex numbers increases counterclockwise). Note that the “generator” in the label
has nothing to do with the one present in the circuit, but is the driving point impedance generator that
one imagines to connect in the point of interest of a circuit to define the relevant impedance.
The second scale, wavelengths toward load, has values that increase counterclockwise and is useful
when the input impedance is known and the load impedance value is desired:
µ ¶T L µ ¶T L
l l lAB
= +
λ eqB λ eqA λ

55
Renato Orta - Transmission Line Theory (Nov. 2012)

ζB
1+ ΓB
V
arg(1+V ΓB )

Figure 3.22. Magnitude and phase of the transmission coefficient 1+ V ΓB , measured

 on the Smith Chart.

In this way, only sums are carried out.

It is useful to clarify the reasons for using the symbol (l/λ)TeqG as a measure of arg( V Γ). In Fig. 3.21
the intersection of the circle | V Γ| = | V ΓB | with the negative real axis is the point labeled rm , because in
this point the normalized local impedance ζ(z) has the minimum real part and zero imaginary part (see
Section 3.5). Clearly, from the picture, V ΓB can be viewed as the input reflection coefficient of a line
with electrical length (l/λ)TeqB
G
, terminated with a resistor of value RL = Z∞ rm . A similar interpretation
TL
holds for (l/λ)eq . Moreover, eq.(3.10) has the appearance of a sum of homogeneous quantities, more than
eq. (3.9).

56
Renato Orta - Transmission Line Theory (Nov. 2012)

3.7 Analysis of simple circuits

Sometimes two transmission lines with different characteristics are cascaded or lumped loads are connected
in series or in shunt with the transmission line. Let us see how the analysis is carried out in such cases.

Cascade connection of transmission lines Consider first the cascade connection of two lines with
different characteristic impedance. Notice that the picture uses the symbols of the transmission lines:

Z ∞1 Z∞2

-
A+

Figure 3.23. Cascade connection of two lines with different characteristic impedance.

hence the different size of the “conductors” is just a graphical convention to denote lines with different
characteristic impedance, but it has nothing to do with the actual geometry of the lines.
The very circuit scheme adopted implies that both the voltage and the current are continuous at point
A:
VA− = VA+ IA− = IA+
Dividing both sides of the first equations by IA− = IA+ yields the continuity of the local impedance
ZA− = ZA+ . The normalized impedance is instead discontinuous (ζA− 6= ζA+ since Z∞1 6= Z∞2 ).
V
As for the forward voltage, by recalling the general formula V (z) = V + (z)(1 + Γ(z)), we find
V V
VA+− (1 + ΓA− ) = VA++ (1 + ΓA+ )

that is
VA++ 1+ V
ΓA−
=
VA+− 1+ VΓ
A+

Also the ratio of the forward currents is obtained immediately:

+
IA+ Y∞2 VA++ Y∞2 1 + V
ΓA−
+ = =
IA − Y∞1 VA+− Y∞1 1 + VΓ
A+

V
Suppose that the second line is matched, so that ΓA+ = 0. Then
V
VA+ = VA++ = VA+− (1 + ΓA− )

Defining a Transmission Coefficient

def VA++
TV =
VA+−
in this case is
V
TV = 1 + ΓA−
V
For this reason, on the Smith chart, the quantity 1 + Γ is always called Transmission coefficient.
Shunt connection of a lumped load Consider now the case of of a line with the lumped load Yp
connected in shunt at A. Apply Kirchhoff laws at the node A:

57
Renato Orta - Transmission Line Theory (Nov. 2012)

Ip
Yp

Figure 3.24. Shunt connection of a lumped load on a transmission line.

VA− = VA+

IA− = IA+ + Ip

YA− = YA+ + Yp
Exploiting the continuity of the total voltage at A, we can obtain, as in the previous case, the relation
between the forward voltages
VA++ 1 + V ΓA−
+ =
VA− 1 + V ΓA+
and therefrom the corresponding one for the forward currents
+ V
IA + Y∞2 1 + ΓA−
+ = VΓ
IA − Y∞1 1 + A+

Series connection of a lumped load Consider now the case of a lumped load Zs connected in series
on a transmission line at A. Kirchhoff law at node A yield

Vs

Zs

A- A+

Figure 3.25. Series connection of a lumped load on a transmission line.

VA− = VA+ + Vs

IA− = IA+

ZA− = ZA+ + Zs
To find the link between forward and backward waves, it is convenient to work on the current, which is
continuous:
+
IA + 1 + I ΓA− 1 − V ΓA−
+ = =
IA − 1 + I ΓA+ 1 − V ΓA+
As for the voltage
VA++ Z∞2 1 − V
ΓA−
=
VA+− Z∞1 1 − VΓ
A+

58
Renato Orta - Transmission Line Theory (Nov. 2012)

Note that in these cases the use of Kirchhoff laws is completely justified, since they have been applied to
lumped elements. It is interesting to note that the loads in the circuits above are lumped in the z direction
but not necessarily in others. In other words, Zs could be the input impedance of a distributed circuit,
positioned at right angle with respect to the main line, as shown in Fig. 3.27. Likewise, Yp could be the
input admittance of a distributed circuit positioned at right angle with respect to the main line, as in
Fig. 3.26. We will see examples of such circuits in Chapter 6 on impedance matching.

Yp

Figure 3.26. Shunt connection of a distributed load on a transmission line.

Zs

A- A+

Figure 3.27. Series connection of a distributed load on a transmission line.

Transmission line length as a two-port device Two analyze more complex cases, it may be conve-
nient to represent a transmission line length as a two-port device, characterized via its matrices Z, Y , or
ABCD, and then apply the usual lumped circuit theory. We derive now the expression of these matrices
for a length l of transmission line with characteristic impedance Z∞ and propagation constant k. See also
Chapter 7 for a review of these matrices.

Open circuit impedance matrix [Z]

Defining equations µ ¶ µ ¶µ ¶
V1 Z11 Z12 I1
=
V2 Z21 Z22 I2
We get µ ¶
cot kl csc kl
Z = −jZ∞ (3.11)
csc kl cot kl
Note that the current I2 is assumed to be positive when it enters into the port.

59
Renato Orta - Transmission Line Theory (Nov. 2012)

Short circuit admittance matrix [Y ]

Defining equations µ ¶ µ ¶µ ¶
I1 Y11 Y12 V1
=
I2 Y21 Y22 V2
We get µ ¶
− cot kl csc kl
Y = jY∞ (3.12)
csc kl − cot kl
Note that, also in this case, the current I2 is assumed to be positive when it enters into the port. Moreover,
obviously, Y = Z −1 .

Chain matrix ABCD

Defining equations µ ¶ µ ¶µ ¶
V1 A −B V2
=
I1 C −D I2
We get µ ¶
cos kl jZ∞ sin kl
ABCD = (3.13)
jY∞ sin kl cos kl
Note that, differently from before, the current I2 is assumed to be positive when it goes out of the port.
This is the reason of the minus signs in the defining equations.
Also useful are the T and Π equivalent circuits, shown in Figure 3.28. The values of the elements are
kl
ZT 1 = ZT 2 = jZ∞ tan , ZT 12 = −jZ∞ csc kl
2
kl
YP 1 = YP 2 = jY∞ tan , YP 12 = −jY∞ csc kl
2
Note that all the matrix elements are periodic functions, as it is typical of distributed parameter circuits.

Z T1 Z T2 YP12

Z T12 YP1 YP2

(a) (b)

Figure 3.28. (a) T equivalent circuit and (b) Π equivalent circuit of a transmission line length.

60
Chapter 4

Energy dissipation in transmission

lines

Wave propagation in real world transmission lines is always affected by attenuation. This attenuation has
two origins: one is the energy loss in the dielectrics, which have a small but not negligible conductivity,
the other is the energy loss in the conductors, which have very high but not infinite conductivity. The
detailed study of these phenomena requires the solution of Maxwell’s equations in the structures of interest.
In accordance with the circuit point of view, adopted in these notes, we limit ourselves to a qualitative
discussion of the subject. A much more detailed treatment can be found in [3].

4.1 Dielectric losses

The phenomenon of energy dissipation in insulators is the simplest to describe. In every real dielectric
there are electrons that are not strictly bound to atoms and are set in motion by an applied electric field:
in this way an electric current is produced. From this point of view, the material is characterized by a
conductivity γd , measured in S/m, defined by

Jc = γd E (4.1)

where E is the applied electric field, Jc is the resulting current density per unit surface and the subscript
“c” reminds us that this is not an independent source but a conduction current, caused by the applied
field. It is useful to note that this equation is the microscopic form of Ohm’s law. Indeed, consider a metal
wire of length L, and cross section S, for each point of which (4.1) applies. Assuming that the current
density is constant in the cross section and the electric field is constant along the wire, we can write

I V
|Jc | = |E| =
S L
where I is the current in the wire and V the potential difference between the two wire ends. Substituting
we get
γd S
I= V = GV
L
which we recognize as the macroscopic form of Ohm’s law.
Usually the conduction current, which is in phase with the applied electric field because γd in (4.1)
is real, is summed with the displacement current, which is in quadrature, so that a complex dielectric

61
Renato Orta - Transmission Line Theory (Nov. 2012)

constant is introduced. Indeed, recall the second Maxwell’s equation

∂E (r,t)
∇ × H (r,t) = ε + γd E (r,t) + J e (4.2)
∂t
which, in the spectral domain, becomes

∇ × H (r,ω) = jωεE (r,ω) + γd E (r,ω) + Je (r,ω)

³ γd ´
= jω ε − j E (r,ω) + Je (r,ω) (4.3)
ω
= jω ε̃E (r,ω) + Je (r,ω)

In very straightforward way we have introduced a complex equivalent permittivity ε̃, whose real part is
the usual dielectric constant and whose imaginary part is related to the conductivity. Generally, the loss
angle δ, is introduced: it is defined as the argument of the complex number ε̃:
µ ¶
γd
ε̃ = ε0 − jε00 = ε0 εr 1 − j = ε0 εr (1 − jtgδ) (4.4)
ωε0 εr
Then the relationship between loss angle and conductivity is
γd
tgδ = (4.5)
ωε0 εr
Obviously, for a good conductor the loss angle δ → π/2. Observe that if the frequency behavior of ε̃ is of
interest, we must keep in mind that both εr , and γd are functions of frequency. Finally it is to be noted
that the symbol ε̃ has been introduced only for clarity. Indeed ε is always understood to be complex unless
specific indications are given.
We have seen in Chapter 1 that dielectric losses are accounted for in circuit form by means of the
conductance per unit length G. The computation of this quantity, as well as that of all line parameters,
starting from the geometry and the physical parameters of the materials, requires the solution of Maxwell’s
equations for the structure under consideration. From the knowledge of the fields it is possible to derive
the values of the line parameters. This procedure will be briefly illustrated in the next section, where
conductor losses are analyzed. The formulas that allow the computation of G for some examples of lines
are reported in Section 4.3.

4.2 Conductor losses

The complex dielectric permittivity can describe also a good conductor. Actually, in conductors such as
copper, for frequencies up to the millimeter wave range, the displacement current is negligible with respect
to the conduction current, so that ε is assumed to be pure imaginary.
In a transmission line in which the conductors can be assumed perfect, the electromagnetic field is
different from zero only in the insulators. In these conditions, on the very surface of the conductors there
is an electric current strictly related to the electromagnetic field. It is a surface current, whose density
per unit length Jσ , measured along the boundary of the conductor cross-section, has a magnitude equal to
that of the tangential magnetic field in the points of the dielectric facing the conductor, see Fig. 4.1. Its
direction is orthogonal to that of the magnetic field.
If now we imagine that the metal conductivity is very large but finite, it can be shown that the current
is no longer confined to the conductor surface but is distributed also inside the metal, with a density per
unit surface of the cross-section that decays exponentially toward the inside of it. Also the magnetic field
penetrates the metal, with a similar exponential decay. This phenomenon has two consequences:

• energy is dissipated in the metal because the electric field and the conduction current are in phase,
according to (4.1).

62
Renato Orta - Transmission Line Theory (Nov. 2012)

Jσ

ds

Figure 4.1. Perfect conductor and surface current on it. Its density Jσ is the current that
flows through the line element ds.

• the magnetic field in the conductor produces an induction flux described by means of an internal
inductance, to be summed to the external one, which accounts for the flux in the insulator.

A case that lends itself to a simple analysis is that of a planar transmission line, shown in Fig. 4.2.
Assume w/h >> 1, so that the y-variations of fields and current can be neglected, so that they depend

z
x
w
y

Figure 4.2. Planar transmission line.

only on z e da x. Here we focus on the x dependance, since we want to obtain the line parameters per unit
length. It can be shown that the current density per unit surface in the left conductor, has the direction
ẑ and is given by h ³x ´i
I cosh T h h − 1
Jz (x) = T (4.6)
w sinh (T h)
where T = (1 + j) /δ and I/w is the total current (per unit length along the y direction) flowing in the
conductor. In the right conductor the current flows in the opposite direction.
Fig. 4.3 shows a 3D plot of the current density per unit surface Jz (x,ω) versus the normalized depth x/h
and the parameter h/δ. This corresponds to showing the frequency dependance, since the skin depth δ can
be shown to be related to frequency by r
2
δ= (4.7)
ωµγ
We see that if h/δ → 0 the current is uniformly distributed in the conductor. On the contrary, if h/δ is
large, Jz (x) decays exponentially with decay rate δ. In these conditions, the current flows essentially in a
thin film, adjacent to the interface between the metal and the insulator, which justifies the name of the
phenomenon. This behavior is analyzed in greater detail below. The skin depth δ is inversely proportional
to the square root of frequency and of metal conductivity. Table (4.1) shows the data for some common
conductors.

63
Renato Orta - Transmission Line Theory (Nov. 2012)

Figure 4.3. Plot of the current density Jz vs. depth x and frequency (through the parameter
h/δ, which is small at low frequency).

Table 4.1. Characteristics of some good conductors

Skin Depth
1 1
Material γ [S/m] δf 2 [m Hz 2 ] 50 Hz 1 kHz 1 MHz 3 GHz
[cm] [mm] [mm] [µm]
Aluminum 3,54 ×107 0,085 1,19 2,7 0,085 1,6
7
Silver 6,15 ×10 0,064 0,90 2,03 0,064 1,2
Chromium 3,8 ×107 0,081 1,15 2,6 0,081 1,5
Graphite 1,0 ×105 1,59 22,50 50,3 1,59 29
7
Nickel 1,3 ×10 0,014 0,19 4,4 0,014 0,26
7
Gold 4,50 ×10 0,075 1,06 2,38 0,075 1,4
7
Brass 1,59 ×10 0,126 1,78 3,98 0,126 2,3
7
Copper 5,80 ×10 0,066 0,93 2,1 0,066 1,2
7
Tin 0,870×10 0,171 2,41 5,41 0,171 3,12
7
Zinc 1,86 ×10 0,117 1,65 3,70 0,0117 2,14

Starting from the expressions of the current density and of the electric field it is possible to compute
the metal surface impedance, defined as the ratio between the electric field Ez at the interface x = 0 and
the current density per unit length along y, (I/w). The electric field Ez (x = 0) is found from (4.6) and

64
Renato Orta - Transmission Line Theory (Nov. 2012)

(4.1):
1 T I
Ez (x = 0) =
Jz (x = 0) = coth (T h) (4.8)
γ γ w
Hence the surface impedance, per unit length in the z direction and per unit width in the y direction, is
T
Z = R + jωLi = 2 coth (T h) =
γ
· ¸ · ¸ (4.9)
1+j h h
= 2 coth (1 + j) = 2Rs (1 + j) coth (1 + j)
γδ δ δ
where we have introduced the parameter Rs
r
1 ωµ
Rs = = (4.10)
γδ 2γ
called surface resistance, which actually coincides with R{Z} only if h À δ. This surface resistance
depends on frequency and is measured in Ω. At the end of this chapter we will see that traditionally its
numerical value is expressed in “Ω per square”, (Ω/t).
u Finally, the factor 2 in Eq. (4.9) takes into account
the presence of two identical conductors. If the conductor has width w, the impedance per unit length
along z has the value Z/w, since the various elements of the conductor are in parallel.
It is to be noted that this impedance per unit length coincides with the series impedance of the
equivalent circuit of an element ∆z of transmission line (see Fig. 1.3), apart from the contribution of the
external inductance, related to the magnetic flux in the dielectric between the conductors. The imaginary
part of Z in (4.9) is proportional to the internal inductance, associated to the magnetic flux inside the
conductor.

15 1.025

1.02
J z (x )
J z (x )
I (wh ) 10 1.015
I (wh )
1.01

5 1.005

0 0.995
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
x/h x/h

Figure 4.4. Plot of Jz (x) for h/δ = 10. The Figure 4.5. Plot of Jz (x) for h/δ = 0.5. Note
normalization quantity I/(wh) is the average the range on the vertical axis, which is much
current density. smaller than in the left figure.

The expression of the conduction current density (4.6) is valid for all frequencies. In particular, if the
frequency is very high, the skin depth δ is small and h/δ À 1, so that the conductor behaves as if it had
infinite thickness. The expression (4.6) simplifies and becomes an exponential,
I I (1 + j) h xi
Jz (x) ≈ T exp (−T x) = exp − (1 + j) (4.11)
w w δ δ
(see also Fig. 4.4), while (4.9) becomes

2Rs
Z = R + jωLi ≈ (1 + j) (4.12)
w

65
Renato Orta - Transmission Line Theory (Nov. 2012)

Conversely, when the frequency is sufficiently low, h/δ ¿ 1. In these conditions the current flows with
almost uniform density in the whole conductor cross-section (see Fig. 4.5) and the impedance per unit
length can be obtained from (4.9) by recalling the small argument expansion of the hyperbolic cotangent
1 z
coth (z) ∼
= + . (4.13)
z 3
If we substitute this into (4.9), we get
· ¸
∼ 2 (1 + j) δ 1 h
Z = + (1 + j) =
w γδ h (1 + j) 3 δ
µ ¶ µ ¶ (4.14)
1 2h 1 1 ωµh
= 2 +j 2 =2 +j = R + jωLi
γhw 3γ w γhw 3 w
where the expression (4.7) of δ has been used and the factor 2 refers always to the presence of two identical
conductors that contribute to the result. In this low frequency condition the resistance per unit length has
the value
1
R= (4.15)
γwh
for each conductor. Since wh is the conductor cross-section area, this result coincides, as is to be expected,
with the direct current resistance Rdc . On the contrary, at high frequency, the resistance per unit length
of each conductor is given by (4.12)
Rs 1
R= = (4.16)
w γδw
By comparing (4.15) and (4.16) we can derive the following interpretation of the skin depth δ: at high
frequency, i.e. when the conductor thickness is much larger than δ, the resistance per unit length is the
same as that a direct current would feel flowing with uniform density in layer with thickness δ.
Fig. 4.6a shows a plot of the impedance per unit length, given by (4.9), normalized to the surface
resistance Rs versus the normalized thickness h/δ. Fig. 4.6b shows a similar plot, but the impedance is
normalized to the dc resistance Rdc = 1/ (γwh). We note that the normalized resistance becomes very

3 3

2.5 2.5

2 2
Zs / Rdc
Zs / Rs

1.5 1.5

1
1
0.5
0.5
0
0.5 1 1.5 2 2.5 3 0.5 1 1.5 2 2.5 3
Normalized thickness h /δ Normalized thickness h /δ
(a) (b)

Figure 4.6. Normalized series impedance of the planar line. Solid line: real part;
dashed line: imaginary part. The normalization impedance is the surface resistance
Rs in (a) and the dc resistance Rdc in (b).

large at low frequency. Actually, the absolute resistance tends to the finite value Rdc (as it is evident from

66
Renato Orta - Transmission Line Theory (Nov. 2012)

3 3

2.5 2.5

2 2
s
γhR

δ/h
1.5 1.5

1 1

0.5 0.5

0 0
0.0 2.5 5.0 7.5 10 0.0 2.5 5.0 7.5 10.
Normalized frequency Normalized frequency
(a) (b)

Figure 4.7. (a) Normalized surface resistance Rs γh = h/δ. (b) Normalized skin depth δ/h. The
frequency on the horizontal axis is normalized to the demarcation frequency fd .

Fig. 4.6b) whereas the surface resistance Rs goes to zero, as shown in Fig. 4.7a. This figure shows a plot
of the normalized surface resistance
Rs h
= Rs γh =
Rdc δ
versus the normalized frequency f /fd , where the normalization frequency fd , also called demarcation
frequency, has been chosen to be that for which δ = h:
1
fd = (4.17)
πµγh2
By exploiting the previous equations, we find that
s
Rs f
=
Rdc fd

Fig. 4.7b shows a plot of the skin depth vs. the normalized frequency. As far as the series reactance is
concerned, Eq.(4.14) shows that it approaches zero as ω → 0.
Fig. 4.8 shows again the plot of Fig. 4.6b, but the asymptotes relative to the low and high frequency
behavior are added. They cross at h/δ = 1. Since δ is a function of frequency, this condition determines
the demarcation frequency fd that separates the low and high frequency regimes.
Fig. 4.9 shows a plot of the internal inductance, normalized to the dc value, versus the normalized
thickness h/δ. Since δ depends on ω, this equivalent inductance depends on frequency. Note that the
internal inductance is always small with respect to the external one. Indeed, the external inductance is
given by
µd
Le = (4.18)
w
whereas the dc internal inductance is
1µ
Li0 = h (4.19)
3w
and even smaller if the frequency increases. Since in general d À h, the internal inductance is negligible
with respect to the external one.

67
Renato Orta - Transmission Line Theory (Nov. 2012)

2.5

2
R / Rdc
1.5

0.5

0
0 0.5 1 1.5 2 2.5 3
Normalized thickness h /δ

Figure 4.8. Real part of the series impedance per unit length, normalized to Rdc . The asymptotic
behaviors are also plotted and define the demarcation frequency.

0.8

0.6

0.4

0.2

0
0 2 4 6 8 10
Normalized thickness h/δ

Figure 4.9. Internal inductance Li /Li0 of a planar line, normalized to the dc value, versus
the normalized thickness h/δ.

4.3 Loss parameters of some transmission lines

4.3.1 Coaxial cable
Dielectric losses

2πγd
G= (4.20)
D
log
d

68
Renato Orta - Transmission Line Theory (Nov. 2012)

d D

De

Figure 4.10. Coaxial cable.

Metal losses
• Low frequency

Resistance per unit length: · ¸

1 1 1
R= + 2 π (4.21)
d2 De − D 2 γc
4
Internal inductance:
" µ ¶2 #−2 "µ ¶2 µ ¶#
µ D D D
Li = 1− − 1 − 2 log (4.22)
4π De De De

• Medium frequency

3
³ ´
√ 1√
³ ´
Rs j 2 2J0 de Rs j 2 2K0 De
R + jωLi = ³ ´ + ³ ´ (4.23)
πdJ1 de πDK1 D e

where
3 d e = j 12 √D
de = j 2 √ , D
2δs 2δs
and J0 , J1 are Bessel functions of first kind and K0 , K1 are modified bessel functions.

• High frequency

µ ¶
1+j 1 1
R + jωLi = Rs + (4.24)
π d D

This formula has a simple interpretation. When the skin effect is well developed, the series impedance
is the same as the one we would have if the whole current flew with uniform density in a layer one skin
depth thick. The equivalent width of the conductor is 1/πd for the inner conductor and 1/πD for the
outer one: these quantities are obviously the circumferences of the conductors. The same interpretation
was already given in connection with Eq.(4.16), in the case of a planar line.

69
Renato Orta - Transmission Line Theory (Nov. 2012)

4.3.2 Two-wire line

Dielectric losses

πγd
G= µ ¶ (4.25)
−1 D
cosh
d

Figure 4.11. Two-wire line.

Metal losses
• Low frequency
2 µ
R= Li = (4.26)
πd2 γc 4π
• Medium frequency (if D >> d)
3 √ ³ ´
Rs j 2 2J0 de
R + jωLi = ³ ´ (4.27)
πdJ1 de

where
3 d
de = j 2 √
2δs
e J0 , J1 are bessel functions of first kind.
• High frequency
1+j
R + jωLi = 2Rs (4.28)
πd

Finally, in order to understand why the surface resistance Rs is measured in Ω/t,

u refer to Fig. 4.12
where a prismatic conductor with length and width w and thickness δ is considered.

J w
w

Figure 4.12. Prismatic conductor with length and width w and thickness δ.

70
Renato Orta - Transmission Line Theory (Nov. 2012)

We said that the surface impedance is the same we would have for a uniform current flow in a layer of
thickness δ. In such conditions, the surface resistance has the value
l w 1
R= = =
γS γδw γδ
which is independent from the sides of the square. Hence, every square, with arbitrary side, has the same
resistance.

71
Chapter 5

Lossy transmission line circuits

5.1 Solution of transmission line equations

After explaining in detail the analysis technique of circuits containing ideal transmission lines, i.e. without
losses, we turn back to the complete equations, to understand the role played by the parameters R (con-
ductor resistance per unit length) and G (dielectric conductance per unit length). The relevant equations
are 
 ∂ ∂
 − ∂t v(z,t) = R i(z,t) + L ∂t i(z,t)

(5.1)


 − ∂ i(z,t) = G v(z,t) + C ∂ v(z,t)
∂t ∂t
Fourier transforming both sides, we get the real transmission line equations in the spectral domain, that is
 d

 − dz V (z,ω)
 = (R + jω L) I(z,ω)
(5.2)


 − d I(z,ω) = (G + jω C) V (z,ω)
dz
We could now repeat step by step the analysis carried out for the ideal lines, but it is simpler to resort to
the trick of introducing a complex inductance and capacitance per unit length

R R
Lc = L+ =L−j
jω ω

G G
Cc = C+ =C−j
jω ω

in such a way that Eq.(5.2) take the form

 d

 − dz V (z,ω)
 = jω Lc I(z,ω)



 − d I(z,ω) = jω Cc V (z,ω)
dz
formally identical to that of ideal lines. It is just enough to take the solution of the ideal case and obtain
its “analytic continuation” from the real values L e C to the complex ones (Lc and Cc ). Note that the

72
Renato Orta - Transmission Line Theory (Nov. 2012)

τs τp

L∆z R∆z
∆ R G ∆z C ∆z

Figure 5.1. Time constants of the RL and RC groups of an elementary length of transmission line

equivalent inductance and capacitance Lc e Cc can be written

µ ¶
1
Lc = L 1 − j
ωτs
µ ¶
1
Cc = C 1 − j
ωτp
where τs = L/R and τp = C/G can be viewed as the time constants of the series RL group and of the
parallel RC group, respectively, in the equivalent circuit of an elementary length ∆z, shown in Fig. 5.1.
Obviously the time constants τs , τp go to infinity for an ideal transmission line.
Hence, the expressions of the voltage and current on a lossy line are given by
V (z,ω) = V0+ (ω) e−jkz + V0− (ω) e+jkz
(5.3)
I(z,ω) = Y∞ V0+ (ω) e−jkz − Y∞ V0− (ω) e+jkz
where the (complex) propagation constant is
√
k = ω Lc Cc (5.4)
and the (complex) characteristic admittance is
Ãr !−1
1 Lc
Y∞ = = (5.5)
Z∞ Cc

and, substituting the expressions of the complex inductance and capacitance,

sµ ¶µ ¶ sµ ¶µ ¶
R G √ 1 1
k = ω L−j C−j = ω LC 1−j 1−j
ω ω ωτs ωτp
(5.6)
v³ ´
s
G
r uu 1−j 1
1 C −jω Cu ωτ p
Y∞ = = = t³ ´
Z∞ L−jR
ω
L 1−j 1 ωτs

Observe that even if the solution (5.3) holds for any value of R,L,C,G, the case of practical interest is that
in which Lc e Cc have very small imaginary parts.
Let us analyze now the properties of (5.3) when k and Y∞ are complex. As for the propagation constant,
the quantity below the square root sign in (5.4) is given by the product of two factors with phase between
−π/2 and 0, and hence has a phase between −π and 0:
−π < arg(k2 ) ≤ 0

73
Renato Orta - Transmission Line Theory (Nov. 2012)

Im Im
k2 k

−k
Re Re

k2 k

Figure 5.2. k2 complex plane (left) and k complex plane (right)

Im Im
2
Y ∞
Y∞
Y∞2
Y∞
Re Re

2
Figure 5.3. Y∞ complex plane (left) and Y∞ complex plane (right)

Computing the square root yields the values ±k = ±(β − jα) with β ≥ 0 and α ≥ 0. To ascertain whether
k belongs to the fourth or to the second √ quadrant, we can take the limit for R,G → 0, which is the ideal
line case, in which we had chosen k = ω LC, i.e. Re{k} > 0, see Fig. 5.2. Hence, by continuity, in the
lossy case k belongs to the fourth quadrant. This means also that Im{k} < 0: this choice agrees also with
the fact that the forward wave must attenuate for increasing z. As for the characteristic admittance, the
radicand in (5.5) belongs to the right halfplane: for continuity with the case of a lossless line, we choose
Y∞ with positive real part, as shown in Fig. 5.3. Moreover, we recall that Y∞ is the input admittance of
a semi-infinite line: since it is a passive load, the real part of its admittance must be positive.
Sometimes, instead of the propagation constant k = β − jα, one introduces

γ = jk = α + jβ

in terms of which the general expression of the line voltage, for example, is

V (z) = V0+ e−γz + V0− eγz

Moreover, the elements of the matrices Z,Y ,ABCD of a line length, given in (3.11) - (3.13), become
hyperbolic functions of γl, instead of circular trigonometric of kl. This choice is natural when transients
are studied and the line equations are solved by the Laplace transform technique instead of the Fourier
transform. In these notes we will always use the phase constant k.

74
Renato Orta - Transmission Line Theory (Nov. 2012)

v + ( z0 , t )
V 0+ T

Figure 5.4. Time evolution of the forward voltage wave at point z = z0

To understand better the meaning of the solution of lossy transmission line equations, we compute the
time evolution of voltage and current relative to the first term of (5.3), which represented a forward wave
in the case of ideal lines. We interpret (5.3) as phasor equations, for which the following inverse transform
formula holds:
v + (z,t) = Re{V + (z,ω) ejω0 t }
In this way we obtain the expression of the forward wave in the form
+
v + (z,t) = Re{| V0+ | ej arg(V0 )
e−j(β0 −jα0 )z ejω0 t }
(5.7)
= | V0+ | cos(ω0 t − β0 z + arg(V0+ )) e −α0 z

and, for the forward current:

+
i+ (z,t) = Re{| Y∞ | ej arg(Y∞ ) | V0+ | ej arg(V0 )
e−j(β0 −jα0 )z ejω0 t }
(5.8)
= | Y∞ || V0+ | cos(ω0 t − β0 z + arg(Y∞ ) + arg(V0+ )) e−α0 z

assuming k0 = k(ω0 ) = β0 − jα0 , where the real and imaginary parts of the complex propagation constant
have been introduced.
Fig. 5.4 shows a plot of the time evolution of the forward voltage wave in a specific point z = z0 . Note
that it is identical to the plot of Fig. 1.10, which refers to a loss-less line. Fig. 5.5 shows instead a plot of
the same wave vs. z at time t = t0 .
From the analysis of (5.8) and (5.9) we can conclude that:

• The first term of (5.3) represents a wave traveling in the direction of increasing z with phase velocity
ω0 ω0
vf = =
β0 Re{k0 }

Note that vf depends on the value of ω0 , since Re{k0 } is a nonlinear function of frequency.
• The wave amplitude has an exponential decrease vs. z, as it can be expected, because of the power
dissipation taking place on the lossy line. The inverse of the imaginary part of the propagation
constant (α0 ) is the distance over which the amplitude undergoes a decrease of the factor 1/e =
0.36788 = 8.68589 dB of voltage or current (see Fig. 5.6).

75
Renato Orta - Transmission Line Theory (Nov. 2012)

v + ( z, t0 )
V 0+ λ

Figure 5.5. Forward voltage wave vs z at time t = t0

V + ( z ,ω )
V0+
e −1

z
1 α0

Figure 5.6. z-plot of the amplitude of the forward voltage wave on a lossy line

• The wavelength, defined as usual as the spatial period of the wave, is given by
2π 2π
λ= =
β0 Re{k0 }

• The current is proportional to the voltage, but shows the phase shift arg(Y∞ ) with respect to it.
Note also that Y∞ depends on frequency.
• The measurements units of β0 and α0 are
β0 → rad/m

α0 → Np/m or dB/m
As for α0 , since ¯ + ¯
¯ V (z) ¯ def | V + (z) |
¯ ¯ = ln = ln e−α0 z = −α0 z
¯ V + (0) ¯ | V + (0) |
Np
it is natural to express α0 in Np/m. If we express the voltage ratio in dB, we have
¯ + ¯
¯ V (z) ¯ def | V + (z) | def
¯ ¯ = 20 log10 = 20 log10 e−α0 z = −α0 z 20 log10 e = −α0dB z
¯ V + (0) ¯ | V + (0) |
dB

76
Renato Orta - Transmission Line Theory (Nov. 2012)

v − (z , t 0 )
V0−

Figure 5.7. Plot of the backward voltage wave vs z

+ Zg
Vg ZL

A B

Figure 5.8. Lossy transmission line loaded with a generic impedance

Hence, the conversion factor for the attenuation constant is 20 log10 e = 8.68589.

The same considerations can be carried out for the second term of (5.3), which represents a backward
wave identical to the forward one (apart, of course, from the propagation direction), because of the reflection
symmetry of the transmission line.
The time domain expressions of the voltage and current backward wave are

v − (z,t) = | V0− | cos(ω0 t + β0 z + arg(V0− )) eα0 z

i− (z,t) = − | Y∞ || V0− | cos(ω0 t + β0 z + arg(Y∞ ) + arg(V0− )) eα0 z

The plot of the backward voltage wave vs. z at the time t = t0 is shown in Fig. 5.7.
The presence in these expressions of an exponential that increases with z seems to contradict the
dissipative character of the lossy line. Actually, in Fig. 5.8 the forward wave is created in A by the
generator, whereas the backward wave is excited in B at the load position and then it propagates in the
backward direction −ẑ. It is in this direction, in which the natural evolution of the phenomenon takes place,
that the amplitude of the backward wave reduces. The same conclusion can be reached by introducing
the reference in which the backward wave is at rest, z = −vf t = −ω0 t/β0 ; in this reference the amplitude
decays as exp{−ω0 α0 t/β0 }.
Fig. 5.9 shows the space-time plots of the forward and backward voltage waves. We can observe that
the crests are parallel to the straight lines z = ±vf t where the upper sign refers to the forward wave and
the lower to the backward one.

77
Renato Orta - Transmission Line Theory (Nov. 2012)

1
5
v(z, t) progr.

v(z,t) regr.
0 0

−5
−1
2 2

1 1

z/λ 0 1 2 3 z/λ 0 1 2 3
0 t/T 0 t/T

Figure 5.9. Space-time plots of the forward and backward voltage waves

(a ) (b ) (c )

V
Figure 5.10. Plot of the curve Γ(z) for α/β = 0.2 (a), 0.1 (b), 0 (c)

In Chapter 3 we have seen that in the analysis of circuits containing transmission lines, it is useful to
introduce the notion of reflection coefficient:
V V − (z)
Γ(z) =
V + (z)
V
In the case of a lossy transmission line, the transformation law of Γ becomes
V V V
Γ(z) = Γ(0) ej2kz = Γ(0) ej2βz e2αz
We can observe that if we move from the load toward the line input, both the phase and the amplitude
of the reflection coefficient decrease, so that V Γ traces a logarithmic spiral in the complex plane, with the
origin as pole, as shown in Fig. 5.10. For this reason, we can say that the input impedance of a semi-infinite
real (lossy) line coincides with its characteristic impedance. This fact justifies the use of the symbol Z∞
. This result has also an intuitive explanation. Indeed, the fact that the input impedance of a line is
different from Z∞ means that in A, apart from the forward wave, originally produced by the generator,
there is also an appreciable contribution of the backward wave, created in B by the load mismatch. If the
product of the attenuation constant times the line length is very large (α0 l → ∞), the backward wave in A
is negligible and the line appears to be matched. Actually the generator power is only partially delivered
to the load: the rest is dissipated in the line.

78
Renato Orta - Transmission Line Theory (Nov. 2012)

+ Zg
Vg ZL

A B

Figure 5.11. Length of lossy transmission line terminated with an arbitrary load impedance

We have seen in Chapter 3 that when an ideal line is connected to a reactive load, a purely stationary
wave is established on it and the net power flux is zero. We can ask ourselves if also on a lossy transmission
line, connected to a reactive load, a purely stationary wave can be formed. The answer is no, because it is
algebraically impossible to write the voltage on the line as the product of a function of t times a function
of z. There is also a physical explanation: in each point of the line an active power flow exists even if the
load is lossless. This power, obviously, is dissipated in the line length comprised between the point under
consideration and the load.

5.2 Computation of the power flow

The general formula that allows the computation of the power flow in each point of any transmission line
has been derived in Chapter 3 and is reported her for sake of convenience:

1 1© £ + ¤ ª
P (z) = Re{V (z) I ∗ (z)} = G | V (z) |2 − | V − (z) |2 + 2B Im{V +∗ (z)V − (z)} =
2 2
(5.9)
1 n h i o
V
= | V + (z) |2 G 1− | Γ(z) |2 + 2B Im{ V Γ(z)}
2
Where Y∞ = G + jB is the characteristic admittance.
A line is defined to be a low-loss line if B << G, and the characteristic admittance can be taken as
real. The power flow in this case can be computed by the formula that, rigorously, holds only in the case
of ideal lossless lines:

1 h i
P (z) = | V + (z) |2 G 1− | V Γ(z) |2 (5.10)
2
As for the propagation constant, we note that it is always multiplied times the line length, hence we must
always consider the quantity αl. If this is small (αl << 1) then the effects of losses can be neglected
altogether, since
e−αl ' 1
V
When this condition is not satisfied losses must be accounted for and both | V + (z) | and | Γ(z) | are
functions of z, so that also the power flow changes from point to point of the line.
Apply now this formula to the circuit of Fig. 5.11. Denote by PA (PB ) the net power flowing in A(B);
obviously, PB is also the power delivered to the load ZL . The ratio PB /PA is found readily by applying
Eq.(5.10) twice, in the points A and B:
V
PB
1
G | VB+ |2 (1− | Γ B |2 ) 1− | V
Γ B |2
= 2
= e−2αl (5.11)
PA 1
2
G | VA+ |2 (1− | V Γ |2 )
A 1− | VΓ
A |
2

79
Renato Orta - Transmission Line Theory (Nov. 2012)

where we have used the equation

| VB+ |=| VA+ | e−αl
Moreover
V V
| ΓA |=| ΓB | e−2αl
If the load impedance ZL coincides with the characteristic impedance of the line (matched line), V ΓB = 0
and the ratio PB /PA equals the factor exp(−2αl), which is defined nominal attenuation.
If the load impedance is arbitrary (mismatched line) the ratio PB /PA reduces, since the fraction that
describes the additional attenuation due to the mismatch is always less than 1. The amount of power
dissipated in the line length AB is readily found by taking into account the energy conservation:
µ ¶
PB
Pdiss = PA 1 −
PA
Often the expression (5.11) is expressed in dB:
µ ¶
PB V V
= −αdB l + (1− | ΓB |2 )dB − (1− | ΓA |2 )dB
PA dB
The Smith chart is provided with a scale that allows the fast evaluation the attenuation increase due to
the line mismatch.
Finally, if αl ¿ 1 the line has negligible losses and PB /PA ∼ 1, as it would be rigorously true only in
the case of a lossless line.

5.3 Frequency dependence of phase constant and character-

istic impedance
The expressions (5.7) show clearly that both the phase constant and the characteristic admittance have
a frequency dependence. In this section we analyze it, by assuming that the primary constants L, C, R,
G do not depend on frequency. This amounts to neglecting the dielectric dispersion and the frequency
dependence of the skin effect (see Chapter 4).
Considering the equations (5.7), we recognize that we can define two frequency ranges:

• a high frequency range, where ωτs À 1 and ωτp À 1, i.e.

R G
¿1 ¿1
ωL ωC
• a low frequency range, where ωτs ¿ 1 and ωτp ¿ 1

By recalling that if | x |¿ 1:
1 1
(1 + x) 2 ' 1+ x
2
1 1
(1 + x)− 2 ' 1− x
2
it is possible to obtain simple approximate expressions valid for each range. For the high frequency range
we find
sµ ¶µ ¶
√ R G
k = ω LC 1−j 1−j
ωL ωC
· µ ¶¸
√ 1 R G
' ω LC 1 − j +
2 ωL ωC

80
Renato Orta - Transmission Line Theory (Nov. 2012)

and hence
√
β ' ω LC

à r r !
1 C L
α ' R +G
2 L C
As for the characteristic admittance, always in the high frequency range, we find
r s
G
C 1 − j ωC
Y∞ = R
'
L 1 − j ωL
r · µ ¶¸
C 1 R G
' 1+j −
L 2 ωL ωC
from which
r
C
G '
L
r µ ¶
C 1 R G
B ' j −
L 2 ωL ωC
We see that the terms R/ωL and G/ωC, beside being small by hypothesis, are summed in the expression
of α but subtracted in that of Z∞ .
To obtain the expressions for the low frequency range it is convenient to first rewrite (5.7) as follows:
p
k = −(R + jωL)(G + jωC)
(5.12)
r
G + jωC
Y∞ =
R + jωL
from which we get
sµ
¶µ ¶
√ ωL ωL
k = −j RG 1+j 1+j '
R G
· µ ¶¸
√ 1 ωL ωL
' −j RG 1 + j +
2 R G
From this equation the expressions of β and α follow by inspection
à r r !
ω R G
β (ω) = C +L
2 G R
√
α (ω) = RG (5.13)
Note that β (ω) is linear at both high and low frequency, but the slope of the two straight lines is different.
A simple computation shows that the low frequency slope is larger than the high frequency one if τp > τs ,
which usually holds true in practice.
As for the low frequency approximation of the characteristic admittance, from Eq. (5.13) we find
r s ωC
G 1+j G
Y∞ = ωL
R 1+j R
r · µ ¶¸
G ω C L
' 1+j −
R 2 G R

81
Renato Orta - Transmission Line Theory (Nov. 2012)

and hence the real and imaginary parts of the characteristic admittance are
r
G
G '
R
r µ ¶
ω G C L
B ' j −
2 R G R

We notice that both at low and high frequency the real part of the characteristic admittance is essentially
frequency independent, but the two constant values are different. The imaginary part instead tends to
zero in both regimes.
In the intermediate frequency range no approximation is possible and the general expressions (5.7)
must be used.
Fig. 5.12 shows plots of β (ω), α (ω), G (ω), B (ω) for a realistic transmission line with the following
values of the primary constants:

R = 25 Ω/m L = 2.5 mH/m G = 0.3 µS/m C = 5 nF/m

The time constants have the values

τp = 0.0167s τs = 10−4 s

The plot of β(ω) is of log-log type, so that the different slopes in the two frequency ranges is represented
as a vertical translation. The other plots are instead of semi-log type. We note that the imaginary part of
Y∞ is maximum when the real part has the maximum slope. This is a general property (Kramers Krönig
relations), related only to the fact that Z∞ (ω) can be considered to be a transfer function, that is the
Fourier transform of the impulse response, which is a causal time function.
It is important to note that when τp = τs = τ , the curve β (ω) becomes a straight line, while α (ω) is
a constant. Indeed, from Eq. (5.7) it follows
µ ¶
√ 1
k = ω LC 1 − j (5.14)
ωτ

from which
√
β (ω) = ω LC

√
α (ω) = RG

The condition τs = τp is called Heaviside condition and is very important since it guarantees distortion
free propagation, as it will be discussed in Chapter 8. Since in practice the line parameters do not fulfill
this condition, one can load the line with periodically spaced series inductors. If the spacing is much
smaller than the wavelength, it can be shown that the line inductance per unit length is increased by the
quantity L/d, where L is the inductor value and d their spacing. If the Heaviside condition is fulfilled, Z∞
is frequency independent, since r r
R L
Z∞ = =
G C
Note that if the losses are so large that the imaginary part of the characteristic admittance B cannot
be neglected, the forward and backwards waves are no longer power orthogonal This implies that the
amplitude of the reflection coefficient has no energy interpretation.

82
Renato Orta - Transmission Line Theory (Nov. 2012)

β =Re(k) ( rad / Km ) α =−Im(k) ( Np / Km )

0.015

0
10
0.01

0.005

0
10
0 τ−1 τ−1 5 0 τ−1
−1
τs 105
p s 10 10 p
FREQUENCY ( Hz ) FREQUENCY ( Hz )
−3 Re(Y∞) ( S ) −4 Im(Y∞) (S)
x 10 x 10
1.5
5

4
1
3

2
0.5

0 0
10
0 τ−1 τ−1 105 10
0 τ−1 τ−1 105
p s p s
FREQUENCY ( Hz ) FREQUENCY ( Hz )

Figure 5.12. Plots of β (ω), α (ω), G (ω), B (ω) for a realistic transmission line. The values of the
primary constants are specified in the text

83
Chapter 6

Matching circuits

6.1 Introduction
In this chapter we address a subject with great practical importance in the field of distributed parameter
circuits, i.e. the notion of impedance matching. Actually there are two types of matching, one is matching
to the line, the other is matching to the generator.
When a transmission line must be connected to a load with an impedance different from the charac-
teristic impedance of the line, it is necessary to introduce a matching device, capable of eliminating the
presence of reflected waves on the line. The other type of matching, not specific of distributed parameter
circuits, has the property of allowing a generator to deliver its available power.
These two objectives can be reached by means of lossless impedance transformers, which can be realized
either in lumped or distributed form. As for the latter, several solutions will be described.

6.2 Types of impedance matching

Consider the circuit of Fig. 6.1, where a real generator and an arbitrary load are connected by a transmission
line with negligible losses. We have already analyzed this circuit in Section 3.7, to find voltages and

+ Zg
Vg ZL

ZA A B

Figure 6.1. Circuit consisting of a lossless transmission line, connected to a generator and a load.

currents in every point of the line. The power delivered to the load coincides with that absorbed by the
input impedance Zin = ZA , since the line is lossless:
1 ∗ 1
PB = PA = Re{VA IA } = |IA |2 Re{Zin } =
2 2

84
Renato Orta - Transmission Line Theory (Nov. 2012)

1 |Vg |2
= Re{Zin } (6.1)
2 |Zg + Zin |2
The standing wave ratio (VSWR) on the line is given by

Vmax 1 + | V ΓB |
S= =
Vmin 1 − | V ΓB |

The power absorbed by the load can also be expressed in terms of the maximum voltage on the line.
Indeed, express this power first in terms of the forward voltage

1 |VB+ |2 ³ ´
PB = 1 − | V ΓB |2 .
2 Z∞
The maximum voltage on the line is

Vmax = |VB+ |(1 + | V ΓB |).

Eliminating |VB+ | between the two equations we find:

2
1 Vmax 1
PB = . (6.2)
2 Z∞ S
Depending on the values of the internal impedance of the generator ZG and of the load impedance, two
different cases can be considered:

A) Matching of the load to the line

If ZL = Z∞ , the reflection coefficient in B is zero, as well as that in A, so that Zin = Z∞ . In this situation,
defined as uniformity matching, the standing wave diagram is flat (VSWR = 1) since only the forward
wave is present on the line. The power delivered to the load is

1 Z∞
PB = |Vg |2 .
2 |Z∞ + Zg |2

Observe that, fixing the active power delivered to the load PB , the maximum line voltage Vmax has the
minimum value when the load is matched to the line. Alternatively, we can say that fixing the maximum
voltage on the line, the power delivered to the load is maximum when the load is matched. This remark
is important in high power applications, since for every transmission line there is maximum voltage that
must not be exceeded in order to avoid sparks that would destroy the line. From (6.2) we recognize the
importance of a VSWR as close to one as possible.

B) Generator matching
Suppose that in the circuit of Fig.6.1 the generator is fixed but the value of the input impedance Zin = ZA
can be changed at will. We can ask what is the optimum value of Zin that allows the maximum power to
∗
be extracted from the generator. Rewrite (6.1) recalling that che Re{Zin } = (Zin + Zin )/2:

|Vg |2 Zin + Zin∗

PB = ∗ ∗
4 (Zg + Zin )(Zg + Zin )

and differentiate with respect to Zin :

∂PB |Vg |2 1 ∗
Zin + Zg − Zin − Zin
= ∗ ∗ 2
∂Zin 4 Zg + Zin (Zg + Zin )
|Vg |2 1 ∗
Zg − Zin
= ∗
4 Zg∗ + Zin (Zg + Zin )2

85
Renato Orta - Transmission Line Theory (Nov. 2012)

This derivative is zero for Zin = Zg∗ , a condition defined as conjugate matching. It can be readily
checked that it corresponds to a maximum. The power delivered in this case is the available power of the
generator and has the value
1 |Vg |2
Pav =
2 4Rg
where Rg = Re{Zg } is the internal resistance of the generator.
It is interesting to note that the power delivered by a generator in an arbitrary load condition can be
written as
PB = Pav (1 − | k Γin |2 )
where k Γin is a generalized reflection coefficient of the impedance Zin with respect to the internal generator
impedance, introduced by Kurokawa:
k Zin − Zg∗ Zin + jXg − Rg
Γin = =
Zin + Zg Zin + jXg + Rg
Note that when Zg is real, the Kurokawa reflection coefficient is coincident with the ordinary one, whereas
it is a different concept when Zg is complex. However, it is recognized as the usual reflection coefficient of
the equivalent impedance Zeq = Zin + jXg with respect to Rg and hence can be determined graphically
by means of the Smith chart.
If we now set
V
Γin = x + jy,
V Zg − Z∞
Γg = = a + jb,
Zg + Z∞
it can be proved that the locus in the plane V Γin of the points for which it is PB /Pav = m with m a
constant, is the circumference with equation (see Fig. 6.2):
x2 + y 2 − 2αx − 2βy + γ = 0
where
ma
α = ,
1 − (1 − m) (a2 + b2 )
mb
β = − ,
1 − (1 − m) (a2 + b2 )
m − 1 + a2 + b 2
γ = ,
1 − (1 − m) (a2 + b2 )
V ∗
The center is in the point with coordinates (α,β), lying on the segment joining the point Γg , with
coordinates (a, − b), to the origin. The radius is
√
1 − m(1 − a2 − b2 )
r= .
1 − (1 − m) (a2 + b2 )
In particular, when m = 0, i.e. the delivered power is zero, the locus is the unit circumference with center
in the origin of the plane V Γin : this result is obvious, since the corresponding Zin is a pure reactance.
When m = 1, i.e. the generator delivers its available power, the locus reduces to the point V Γ∗g .
Note that when the energy matching condition holds, the VSWR can be greater than one, since it
is related to | V ΓB |. In other words, energy matching and line matching are independent. The optimum
operating condition for the circuit of Fig. 6.1 is that both the load and the generator are matched to the
line. Indeed, in these conditions the generator delivers the maximum power. Moreover, because of the
line matching, the voltage on the line is the minimum for that value of active power flow. If the losses
were not negligible, the line attenuation would be the minimum one and would be coincident with nominal
one. Finally, as it will be discussed in Chapter 8, the line matching condition is essential to minimize
distortions.
In the rest of this chapter we will show how to design impedance transformers that allow the matching
condition to be reached.

86
Renato Orta - Transmission Line Theory (Nov. 2012)

m=0

m=0.2

m=0.4

m=0.6
m=0.8

m=1

V
Figure 6.2. PB /Pav = m loci on the Smith chart, with Γg = 0.5(−1 + j).

6.3 Impedance matching devices

First of all, we observe that a matching network must be formed by at least two components, since two
conditions must be enforced, one on the real and one on the imaginary part of the input impedance. If
the network contains more than two independent elements, multiple matching conditions can be enforced,
i.e. at several frequencies or on a frequency band. First we address the simplest case of single frequency
matching.
We have seen in the previous section that for several reasons it is useful to be able to design impedance
transformers that perform as indicated in Fig. 6.3. In the case of matching to the line, Zin is the charac-

ZL

Z ing

Figure 6.3. Scheme of impedance transformer.

teristic impedance Z∞ of the feeding line. In the case of conjugate matching, Zin is the complex conjugate
of the generator internal impedance. There are various solutions to this problem, all consisting of ideally
lossless networks. We will discuss

• “L” cells with lumped reactive

• Single stub cells
• Double and triple stub cells
• λ/4 impedance transformers

6.3.1 L cells with lumped reactive elements

Two structures are possible (see Fig.6.4) For certain combinations of Zin and ZL both can be used, for

87
Renato Orta - Transmission Line Theory (Nov. 2012)

jX jX
jB ZL jB ZL

Z in Z in

(a) (b)

Figure 6.4. The two possible schemes of L matching networks.

others only one of them. Consider the first configuration. The condition to be enforced at the input
terminals is
1
Rin + jXin = jX + 1
jB + RL +jXL

This is a complex equation in the two real unknowns B and X, which can be solved by separating real
and imaginary part of the right hand side (RHS). After some algebra, we find
RL + jXL
RHS = jX + =
1 − BXL + jBRL
(RL + jXL )[(1 − BXL − jBRL ]
= jX + =
(1 − BXL )2 + (BRL )2
2
XL (1 − BXL ) − BRL
= jX + j +
(1 − BXL )2 + (BRL )2
RL (1 − BXL ) − BRL XL
+ =
(1 − BXL )2 + (BRL )2
RL
= 2 +
B 2 (RL + XL2 ) − 2BXL + 1
½ 2
¾
XL − B(RL + XL2 )
+ j X+ 2 2 2
B (RL + XL ) − 2BXL + 1

Enforcing the equality of left and right hand side yields

RL
Rin = 2
B 2 (RL + XL2 ) − 2BXL + 1
2
(6.3)
XL − B(RL + XL2 )
Xin = X+ 2 2 2
B (RL + XL ) − 2BXL + 1

From the first we obtain a quadratic equation in B:

2 RL
(RL + XL2 )B 2 − 2XL B + (1 − )=0
Rin
with solutions ( r )
q
1 RL 2
B= 2 XL ± RL + XL2 − Rin RL (6.4)
RL + XL2 Rin
The corresponding X values are found from the second of (6.3). Obviously the square root must be real:
if this condition is not satisfied, we must use the configuration of Fig. 6.4b. In this case the condition to

88
Renato Orta - Transmission Line Theory (Nov. 2012)

enforce is
1 1
= jB +
Rin + jXin jX + (RL + jXL )
from which
Rin − jXin RL − j(X + XL )
2 2 = jB + 2
Rin + Xin RL + (X + XL )2
By equating real and imaginary parts of the two sides we get the two equations
Rin RL
2 2 = 2
Rin + Xin RL + (X + XL )2
(6.5)
Xin X + XL
− 2 2 = B− 2
Rin + Xin RL + (X + XL )2

From the first we get X:

RL 2
(X + XL )2 = 2
(Rin + Xin 2
) − RL
Rin
from which r
p RL
2 2
X = −XL ± Rin + Xin − RL Rin (6.6)
Rin
Finally, from the second of (6.5) we get B. Also in this case, the radicand of the square root must be
positive. It is interesting to ascertain for which combinations of load and input impedance each form of
the L circuit can be used. Suppose that ZL is specified. From eq. (6.4) we see that the radicand is positive,
and hence the circuit of Fig. 6.4a can be used, if
2
RL + XL2
Rin ≤ Rc =
RL
Geometrically, in the plane Rin , Xin , this region is the strip comprised between the imaginary axis and the
vertical line Rin = Rc .
Next, consider the circuit of Fig. 6.4b. From eq.(6.6), the radicand is positive for
2 2
Rin + Xin − RL Rin ≥ 0

Geometrically, this region is the part of the right half-plane lying outside of the circle with radius RL /2
and center in the point (RL ,0). Fig. 6.5 shows these regions. We see that matching is possible only with
the circuit of type a for Zin inside the circle and only with the circuit of type b for Zin to the right of the
vertical line. For other values of desired input impedance, both circuits can be used.
In this way we have solved the matching problem in the most general case. Obviously, in the case
of line matching, the formulas will simplify because Xin = 0. It is interesting to note that the problem
can also be solved graphically by means of the Smith chart. The susceptance B and the reactance X
can be realized by lumped elements (inductors and capacitors) if the frequency is low enough. The upper
limit can be identified as the frequency for which the component size is of the order of λ/10. This means
that with present day technology this matching technique can be used up to some GHz (see [4] p. 287).
Alternatively, for frequencies in the microwave range, B e X can be realized with transmission line lengths,
terminated in short circuit or open circuit which, as discussed in section 3.2, have a purely reactive input
impedance.

6.3.2 Resistive matching pad

The matching network described in the previous section has the advantage of being lossless (in the case
of ideal components) and the disadvantage of being narrowband. In some cases, it is so important to
have a wideband behavior that a certain amount of losses is accepted. This is the case when two matched
transmission lines of different characteristic impedances have to be connected, as in Fig.6.6, and matching

89


Renato Orta - Transmission Line Theory (Nov. 2012)

y 
ŝŶ

ď
Ă
Z
Đ Z 
ŝŶ

Figure 6.5. Realizability of L matching networks. For Zin in the circle, only network a can be
used, for Zin to the right of the vertical line, only network b.

is required in the two directions, i.e. from left to right and from right to left. The two equations to be
enforced are (
Z∞1 = R1 + R2 //Z∞2
(6.7)
Z∞2 = R2 //(R1 + Z∞1 )
that is 
 R2 Z∞2

 Z∞1 = R1 + R + Z
2 ∞2

 R2 (R1 + Z∞1 )
 Z∞2 =
R2 + (R1 + Z∞1 )
Solve the first with respect to R1 :
R2 Z∞2
R1 = Z∞1 −
R2 + Z∞2
and substitute in the second. After a bit of algebra, we get
(Z∞1 − Z∞2 )R22 − Z∞1 Z∞2
2
=0
and finally  r

 Z
 R1 = Z∞1 1 − ∞2

 Z∞1
Ãr !−1

 Z∞2

 R2 = Z∞2
 1−
Z∞1

R1
Z ∞1 Z∞2
R2

Figure 6.6. Resistive matching pad

90
Renato Orta - Transmission Line Theory (Nov. 2012)

Obviously, the configuration of Fig.6.6 can be used only if Z∞1 > Z∞2 , so that the resistance values turn
out to be real. In other words, the series branch of the matching device must be connected to the line with
larger characteristic impedance.
In order to compute the attenuation introduced by this device, assume that a voltage wave of amplitude
V in is incident from the left. The net active power absorbed is P in = Y∞1 |V in |2 /2 since there is no
reflection. Likewise, the active power leaving the device is P out = Y∞2 |V out |2 /2 and the two voltages are
related by µ ¶
R2 //Z∞2 V in R1
V out = V in = R2 //Z∞2 = V in 1 −
R1 + R2 //Z∞2 Z∞1 Z∞1
where the first of (6.7) has been used. The power transmission coefficient is then
à r !2 Ãr r !2 Ãr r !−2
P out Z∞1 Z∞2 Z∞1 Z∞1 Z∞1 Z∞1
= 1− = − −1 = + −1
P in Z∞2 Z∞1 Z∞2 Z∞2 Z∞2 Z∞2

From the last expression it is evident that the transmission coefficient is always less than 1, as it is to be
expected. As a numerical example, consider the connection between a 75 Ω line to a 50 Ω one. Using the
previous expressions, one gets 
 R1 = 43.30 Ω



R2 = 86.60 Ω

 out
P

= 0.27 → −5.72dB
P in
The configuration of Fig.6.6, called “minimum loss matching pad”, is not the only possible one: sometimes
T and π configurations are used.

6.3.3 Single stub matching network

The matching networks of this type consist essentially of a a transmission line length and a reactance,
that can be connected in shunt or in series to the line itself. This reactance is realized by another length
of transmission line, terminated with an open or short circuit, called “stub”. Suppose that a shunt stub
matching network is to be designed, to match a load to the feeding transmission line, (Fig. 6.7). Assume
also that the matching network employs transmission lines with the same characteristic impedance as the
feeding line. The matching network is an impedance transformer: its normalized input admittance must

Z∞ jbs Z∞ ZL

-
A+ B

Figure 6.7. Matching network with shunt stub: matching to the line.

be yA− = 1 when it is terminated with yL . We know that the locus on the admittance Smith chart of
yA+ when lAB is changed is a circumference with center in the origin and radius equal to |ΓL |. This
circumference intersects the constant conductance circumference g = Re{yA+ } = 1 in the points I1 e I2
(see Fig. 6.8). Both points refer to values of yA+ that have the required real part. From these points the

91
Renato Orta - Transmission Line Theory (Nov. 2012)

yL
I1

y=1

I2
zL

Figure 6.8. Smith chart relative to the design of the matching network of Fig. 6.7

origin of the Smith chart ( I Γ = 0) is reached by selecting the susceptance bs :

bs = −Im{yA+ }.

Using again the Smith chart, the stub length is readily found as soon as its termination (short or open
circuit) has been chosen. The length of the line AB is deduced from the angle between I ΓL and I1 (or I2 ).

Example 1
Design a line matching network, having a shunt short circuited stub. The data are: ZL = 125 − 125j Ω e
Z∞ = 50 Ω.
We find ζL = ZL /Z∞ = 2,5 − 2,5j and, read on the chart, yL = 0,2 + 0,2j, | I ΓL | = 0,67 and equivalent
electrical length (TG) (l/λ)TeqB
G
= 0,0326 (see Fig. 6.9). From the intersection of the constant | I Γ| circle
with the Re{y} = 1 circle we read yA+ = 1 + 1,84j and then bs = −1,84. Because bs < 0 we say that the
stub is inductive. Since (l/λ)TeqA
G
+ = 0,1843, the length of AB has the value (0,1843 − 0,0326)λ = 0,1517λ.
The length of the stub is found from the Smith chart of Fig. 6.10. The rotation takes place on the
unit circle from the short circuit point y → ∞ to the point ys = 0 + jbs . The stub length is then
ls = (0,329 − 0,25)λ = 0.079λ.
Of course there is also the solution bs = 1,84 (capacitive stub) corresponding to yA+ = 1 − 1,84j. In
this case the length of AB becomes 0.283λ and that of the stub, still short circuited, is ls = 0.421λ. The
relevant Smith charts are shown in Fig. 6.9 and Fig. 6.10

The line AB could be lengthened by any multiple of λ/2. The input impedance of the matching network
would still be Z∞ at the design frequency, but its bandwidth would be smaller. There is indeed a general
rule: the bandwidth of a device is inversely related to its electrical length. A similar remark holds for the
stub.
If the stub were to be connected in series to the main line, the design procedure would be only slightly
modified. In this case we would have employed an impedance Smith chart: no other change would be
required.
The procedure described above to design a line matching network can be generalized to solve the
problem of designing a conjugate matching network. In this case the arrival point on the Smith chart
is not the origin but a generic point, corresponding to the complex conjugate of the generator internal
impedance. Let us make reference to a shunt stub. The matching network structure is the same as before:

• a transmission line length that allows the desired real part of the input admittance to be obtained;

92
Renato Orta - Transmission Line Theory (Nov. 2012)

l AB
λ

0,1843

yA +

0,0326

yL
l AB
yA- λ

ζL

yA +

Figure 6.9. Smith chart relative to the design of the matching network of Example 1

• a shunt susceptance that modifies the imaginary part of the input admittance, so that it has the
desired value.

The problem, now, is that the first step is not always successful. Indeed, it is evident from Fig. 6.11
that, for a given load admittance YL , with a matching network of the type shown in Fig. 6.7 only the
points of the region Rd can be reached. In fact, when the line length AB is changed, the real part of the
input admittance is always comprised between gm and gM defined by the intersection of the |Γ| =constant

93
Renato Orta - Transmission Line Theory (Nov. 2012)

yA +

0,25

ls
λ
bs

0,329

Figure 6.10. Smith chart relative to the design of the stub for the matching network of Example 1

circle through yL and the real axis. We find readily

1 1 − |ΓL | 1
gm = = = ,
gM 1 + |ΓL | S

where S is the load VSWR. Hence the region Rd is the part of plane internal to the circle Re{y} = gm
and external to the circle Re{y} = gM . Incidentally, it is simple to recognize that whatever the value of
yL , the origin belongs always to Rd , so that the line matching is always possible.
When the admittance yg∗ to be reached lies outside of the region Rd , we can still use a stub matching

94
Renato Orta - Transmission Line Theory (Nov. 2012)

y
L

gM
gm

Rd

Figure 6.11. The points of the region Rd represent the input admittances of a matching network
of the type of Fig. 6.7, loaded by yL .

network, provided the structure is reversed, as shown in Fig. 6.12, i.e. a “reversed L matching network”
is used, being that of Fig. 6.7 a “straight L ”. Indeed, in this case, the values of yA− that can be obtained

Z∞ Z∞ jbs ZL

- +
A B

Figure 6.12. “Reversed L” stub matching network.

from yL are those belonging to the region Rr (see Fig. 6.13). The domain Rr is the annular region between
the concentric circles with radii |Γ| = 1 and |Γ| = |gL − 1|/|gL + 1|. We see that the union of Rr and
Rd equals the entire Smith chart, hence every matching problem can be solved by a stub network (either
straight or reversed L). Moreover the intersection of Rr e Rd is not empty, so that the solution for certain
values of yA− can be obtained with both types of networks.

Example 2
Design a conjugate matching network with an open circuit shunt stub. The data are: ZL = 75 + 75j Ω,
Zg = 150 − 300j Ω and Z∞ = 75 Ω.
We compute ζg∗ = 2 + 4j and ζL = 1 + j, then the corresponding admittances are read on the Smith chart:
yg∗ = 0.1 − 0.2j and yL = 0.5 − 0.5j. Let us try a reversed L configuration. The intersection between
the constant | I Γ| circle through yg∗ with the constant conductance circle through yL defines two points,
of which, for example, yB − = 0.5 + 2j, bs = 2.5 (capacitive stub). The length of AB is 0.289λ and that
of the stub is ls = 0.190λ. If we choose the other intersection yB − = 0.5 − 2j, we obtain bs = −1.5
(capacitive stub), and the length of AB becomes 0.147λ and ls = 0.344λ. The relevant Smith chart is
shown in Fig. 6.14. We see easily that the constant | I Γ| circle through yL and the constant conductance
circle through yg∗ have no intersections, hence only the reversed L configuration is possible.

95
Renato Orta - Transmission Line Theory (Nov. 2012)

yL

gL

Rr

Figure 6.13. The points belonging to the region Rr represent input admittances of a stub
matching network loaded by yL .

The reason for which only examples of shunt stubs have been discussed is that this type of connection
is more common, because it is easier to realize, for example by the microstrip technology

6.3.4 Double stub matching network

Even if the straight or reversed L matching networks can solve any practical problem, sometimes double
stub networks are used. Because of their form, shown in Fig. 6.15 in the case of a shunt stubs, they are
also called Π networks. It is clear that the stub susceptances and their separation can be chosen in an
infinite number of different ways. Sometimes the distance AB is fixed a priori. In this case the solution
are at most two, but are not guaranteed to exist. The design can be carried out in two different ways,
starting from the load or from the generator.
If we start from the load, (see Fig. 6.16) the procedure is the following:

1. draw the constant conductance circle through yL : this is the locus of all possible yB − as the stub
susceptance in B is changed;
2. rotate the whole circle by d/λ toward the generator: the locus of the corresponding yA+ is obtained;
3. this circle intersects the constant conductance circle through yA− in the points I1 and I2 . From either
of these points move to yA− by means of the stub with susceptance b1 = Im{yA− } − Im{yA+ };
4. having fixed the value of b1 , obtain yB − and then b2 = Im{yB − } − Im{yL }.

The procedure for the design of the same matching network, but starting form the generator (see Fig. 6.17),
is the following:

1. draw the constant conductance circle through yA− : this is the locus of all possible yA+ as the stub
susceptance in A is changed;
2. rotate the whole circle by d/λ toward the load: the locus of the corresponding yB − is obtained;
3. this circle intersects the constant conductance circle through yL in the points I3 and I4 . These points
define yB − , from which b2 = Im{yB − } − Im{yL } is found;
4. having found b2 , obtain yA+ and therefrom b1 = Im{yA− } − Im{yA+ }.

Obviously, even if the diagrams on the Smith chart are different in the two cases, the values of b1 e b2 turn
out to be the same.

96
Renato Orta - Transmission Line Theory (Nov. 2012)

0.179

l AB
λ
y B−
ζL

ζ g∗

yg∗

yL

y B−
0.468

l AB
0.321
λ

Figure 6.14. Smith chart relative to the design of the conjugate matching network (re-
versed L) discussed in Example 2.

If the distance between the stubs is fixed a priori, the solution for certain load and input admittances
is not guaranteed to exist. This limitation is not present in the case of a triple stub matching network,
even if the relative distances are fixed a priori (see Fig. 6.18). In this case, in fact, the desired matching
can always be obtained, provided convenient stub lengths are selected. This device can be useful in the
laboratory: indeed, some implementation exist, where the stub lengths are changed by means of sliding
pistons.

If we want to design such a matching network, we can use the method described above. The detailed

97
Renato Orta - Transmission Line Theory (Nov. 2012)

A B

ZL

Figure 6.15. Double stub matching network

yi
yL
I1

I2

d/λ

Figure 6.16. Design of a double stub matching network, starting form the load.

procedure is the following:

1. draw the constant conductance circles through yL and yA− ;
2. rotate the first toward the generator by d2 /λ and the second toward the load by d1 /λ. They
represent the loci of yB − and yB + , respectively. They meet in two points and the value of b2 is given
by b2 = Im{yB − } − Im{yB + };
3. having defined yB − and yB + , obtain yA+ and yC − and, therefrom, b1 e b3 .
From these nontrivial examples we can appreciate the power of the Smith cart as a design tool.

6.3.5 λ/4 matching networks

This type of matching network, in its simplest form, can be used only to match real impedances and
consists of a λ/4 length of transmission line of suitable characteristic impedance. The wavelength is to be
evaluated at the design frequency. The scheme is shown in Fig. 6.19. The normalized input impedance is
the inverse of the normalized load impedance (see Eq.(3.6)):
1
ζA = ,
ζL
from which
2
Z∞
ZA = .
RL

98
Renato Orta - Transmission Line Theory (Nov. 2012)

d/λ

yi
yL I
1

I2

Figure 6.17. Design of a double stub matching network, starting form the generator.

A B C

ZL

jb1 jb2 jb3

Figure 6.18. Triple stub matching network.

By enforcing the condition that the input impedance ZA coincides with the desired input resistance Ri ,
we find √
Z ∞ = RL Ri .
In conclusion, the characteristic impedance of the line must be the geometric mean of the two resistances
to be matched.
In the case the two impedances to be matched are complex, we can use a matching network consisting
of a λ/4 line length, inserted between two line lengths (of arbitrary characteristic impedance Z∞ ), as
shown in Fig. 6.20. Their purpose is that of transforming the complex impedances into pure resistances, as
indicated in Fig. 6.21. If these resistances are called RB = Z∞ rB − and RC = Z∞ rC + , the characteristic
impedance of the central line BC is √
Z∞m = RB RC

99
Renato Orta - Transmission Line Theory (Nov. 2012)

Z∞ RL

Ri A B

Figure 6.19. λ/4 matching network between real impedances.

d1 λ 4 d2

Z1 Z∞ Z2 ZL

Z in A B C D

Figure 6.20. Structure of the λ/4 matching network for complex impedances.

ζ
L

rB−
rC+

ζin

Figure 6.21. Smith chart for the design of the λ/4 matching network for complex impedances.

100
Chapter 7

The Scattering matrix

In this Chapter we develop a convenient formalism to describe distributed parameter circuits containing
multiport devices. First we review the matrix characterization of multiport devices based on the use of
total voltage and total current as state variables. This description is appropriate to the case of lumped
networks. As discussed at length in the previous Chapters, in the case of distributed parameter circuits
a change of basis is highly convenient: we will introduce the so called power waves, a normalized form of
forward and backward waves.

7.1 Lumped circuits

The simplest two-lead circuit element is characterized by its impedance ZL , (or its inverse, i.e. the
admittance YL ), defined as the ratio between the voltage V at its leads and the absorbed current I.
Suppose this element is linear, so that the impedance does not depend on the excitation (I) but only on
frequency. Often a couple of leads of a device is called a “port”: hence, such a circuit element is also called
a one-port device.
As known in circuit theory, these concepts can be generalized to the case of devices with several ports,
say N . We start this presentation by focusing our attention to the important case of two-port devices (see
Fig.7.1).

I1 I2

V1 V2

Figure 7.1. Two-port device with the definitions of voltage and current at the ports

In general, the two voltages V1 and V2 depend on both I1 and I2 :

(
V1 = Z11 I1 + Z12 I2
(7.1)
V2 = Z21 I1 + Z22 I2
where Zij are only functions of frequency, in the case of linear networks, to which we will limit our attention.
The relation (7.1) can be written in matrix form:
[V ] = [Z][I]

101
Renato Orta - Transmission Line Theory (Nov. 2012)

where [V ] = [V1 V2 ]T , [I] = [I1 I2 ]T are column vectors and the 2 × 2 matrix [Z] is called open circuit
impedance matrix. The name is justified by the definition of its elements, derived from (7.1):
¯
Vi ¯¯
Zij =
Ij ¯Ik =0,k6=j

In other words, all ports, except for the j-th, at which the exciting current is applied, must be open
circuited. The advantage of the matrix notation is that (??) can describe also a N -port structure; in this
case [Z] is a N × N complex matrix. Note that the matrix [Z(ω)] can be interpreted as a set of transfer
functions between the applied currents (inputs) and the voltages at all ports (outputs). The diagonal
elements are input impedances, the other elements are trans-impedances.
As in the case of a one-port device we can introduce the admittance YL = 1/ZL , also for an N -port
structure we can introduce a short circuit admittance matrix [Y ]. In the N = 2 case, the linear dependence
between currents and voltages is expressed in the form:
(
I1 = Y11 V1 + Y12 V2
I2 = Y21 V1 + Y22 V2
that is, in matrix form,
[I] = [Y ][V ]
From the comparison with (??) we get [Y ] = [Z]−1 . The name given to the matrix [Y ] comes from the
definition of its elements ¯
Ii ¯¯
Yij =
Vj ¯Vk =0,k6=j

An important role in circuit theory is played by reciprocal and lossless networks. Recall that a circuit
made of resistors, capacitors, inductors, transmission lines is always reciprocal. On the contrary, an
amplifier is non reciprocal, as well as devices containing a magnetic material maintained in a static magnetic
field (e.g. ferrite devices). It can be shown [4] that the matrices [Z] and [Y ] of reciprocal devices are
symmetrical.
The total power dissipated in the device is the sum of the powers entering through the various ports:
N
1 1 X 1
Pdiss = <{V1 I1∗ + V2 I2∗ + ... + VN IN
∗
} = <{ Vi Ii∗ } = <{[V ]T [I]∗ }
2 2 i=1 2

If the device is lossless, this dissipated power is zero for any excitation. Then
1 1
Pdiss = 0 = <{[V ]T [I]∗ } = <{[V ]T [Y ]∗ [V ]∗ }
2 2
Due to the arbitrariness of [V ], it follows
<{[Y ]∗ } = 0
Hence, the [Y ] and [Z] matrices of lossless devices are pure imaginary.
Another useful matrix characterization of two-port devices is that based on the equations
(
V1 = AV2 − BI2
I1 = CV2 − DI2
or à ! à !à !
V1 A B V2
=
I1 C D −I2
The relevant matrix is called ABCD matrix and is a kind of transmission matrix. In fact it relates the
electric state at the input to that at the output of the device. It can be shown that the ABCD matrix of
a reciprocal device has unit determinant.

102
Renato Orta - Transmission Line Theory (Nov. 2012)

The various matrices are useful for the characterization of the connection of multi-port devices. For
example, if two two-port devices are connected in series, a single two-port is obtained, whose open circuit
impedance matrix is the sum of the ones of the two sub-blocks. If the two two-port devices are connected
in parallel, the short circuit admittance matrix of the resulting circuit is the sum of the ones of the two
sub-blocks. Finally, if the two two-port devices are connected in cascade, the ABCD matrix of the resulting
structure is the product of the ones of the sub-blocks.

7.2 Distributed parameter circuits

Suppose now that each port i of an N -port device is connected to a transmission line, with characteristic
impedance Z∞i and propagation constant ki . We have seen in Chapter 1 that the electric state of a
transmission line is specified in the more natural (and simplest) way by giving the amplitudes of the
forward and backward waves V + and V − . Consequently, a one-port device connected to the line is more
conveniently described in terms of a reflection coefficient rather than in terms of impedance or admittance.
In this Chapter we are going to generalize this concept to N -port networks, by introducing a matrix
reflection coefficient, which is generally known as scattering matrix or S matrix. In this case, however, the
amplitudes of the forward and backward waves on the line are specified as power waves amplitudes a e b,
instead of voltages or currents. To explain this name, recall that in Chapter 3 we have shown that the net
active power flowing on a line with real characteristic impedance is given by:
1 1
Pt = Y∞ |V + |2 − Y∞ |V − |2
2 2
If we set √ √
a= Y∞ V + , b= Y∞ V −
the previous equation is rewritten as:
1 2 1 2
Pt = |a| − |b|
2 2
so a and b are directly related to the power flow. Obviously, the z dependence of the signals a and b is the
same as that of V + and V − :
a(z) = a(0)e−jkz
b(z) = b(0)e+jkz
Moreover, they are related by
b(z) = Γ(z)a(z)
where Γ(z) is the local reflection coefficient. In order to define power waves independently from the
orientation of the z axis, which is arbitrary, it is useful to describe the signal a as a wave incident on the
load and b as a wave scattered from the load. Note also that the line characteristic impedance plays the
role of a reference impedance, with respect to which the reflection coefficient is computed.
Let us generalize these concepts to the case of a device with N ports, labeled with the subscript
i = 1,2...N . To each port we assign a reference impedance Zri that can be interpreted as the characteristic
impedance of a transmission line connected to the port. On this line we define power waves:
√ √
ai = Yri Vi+ , bi = Yri Vi−

where the z axis points always into port i, in order to guarantee that ai is actually incident on the device.
In the case of a two-port device, (see Fig.7.2) we have:
(
b1 = S11 a1 + S12 a2
(7.2)
b2 = S21 a1 + S22 a2

i.e. the scattered waves on the various lines depend in general on the incident waves at all ports. Introduce

103
Renato Orta - Transmission Line Theory (Nov. 2012)

a1
a2
Z r1 Zr2

b1 b2

Figure 7.2. Two-port device and definition of the relevant power waves

the column vectors [a] = [a1 a2 ]T and [b1 b2 ]T , so that Eq.(7.2) can be rewritten in matrix form:

[b] = [S][a] (7.3)

where [S] is a 2 × 2 complex matrix, called Scattering Matrix. Obviously, the form of (7.3) is valid also in
the case of a N -port network, [S] being a N × N matrix. From (7.2) we find that the elements are defined
as: ¯
bi ¯¯
Sij = (7.4)
aj ¯ak =0,k6=j
from which it is evident the character of generalized reflection coefficients of the elements of the S matrix.
The condition ak = 0 at port k is obtained by terminating the access transmission line, with characteristic
impedance Zrk , with a load impedance numerically equal to Zrk itself. In this way the access line is
matched and only an outgoing wave is present on it. The terms on the main diagonal (i = j) of [S] are
the usual reflection coefficients at port i when all the others are terminated with the relevant reference
impedances. The terms out of the main diagonal are usually called transmission coefficients from port j
to port i.
Even if the characterization of a device by means of its scattering matrix [S] is, from a theoretical point
of view, completely equivalent to that in terms of the matrices [Z] or [Y ] (apart from the singular cases), it
is in practice the only one to be employed in the microwave field. There are several reasons, among which:

• voltage and current are not always well defined quantities, for example in a waveguide
• the power waves a e b can be measured directly by means of an instrument called Network Analyzer.
In general a wide band characterization of the devices is of interest and, in practice, it is easier to
construct a wide band matched load rather than an open circuit, that is the reference load for the
[Z] matrix.

7.3 Relationship between [S] and [Z] or [Y ]

In (7.4) we have given an explicit definition of the [S] matrix elements. There is also a noteworthy
relation between the [S] matrix and the open circuit impedance matrix [Z], which generalizes the scalar
one (eq.(3.7)).
Let us start from the device characterization in terms of [Z] matrix:

[V ] = [Z][I] (7.5)

Express now [V ] and [I] in terms of power amplitudes [a] e [b]:

[V ] = [V + ] + [V − ] = [Zr ]1/2 ([a] + [b])

¡ ¢
[I] = Y∞ [V + ] − [V − ] = [Yr ]1/2 ([a] − [b])

104
Renato Orta - Transmission Line Theory (Nov. 2012)

where [Zr ] is the diagonal matrix constructed with

√ the reference impedances of all ports. Remember that
[Zr ]1/2 is the diagonal matrix having the values Zri on the main diagonal. Substitute into (7.5):

[Zr ]1/2 ([a] + [b]) = [Z][Yr ]1/2 ([a] − [b])

Expanding the products and factoring [Zr ]1/2 at the left of both sides, we find
³ ´ ³ ´
[Zr ]1/2 [Yr ]1/2 [Z][Yr ]1/2 + [1] [b] = [Zr ]1/2 [Yr ]1/2 [Z][Yr ]1/2 − [1] [a]

Canceling the common factor [Zr ]1/2 we get

[S] = {[ζ] − [1]}{[ζ] + [1]}−1 (7.6)

where [1] is the identity matrix of size N and [ζ] is the normalized open circuit impedance matrix of the
device:
[ζ] = [Zr ]−1/2 [Z][Zr ]−1/2 (7.7)

The inverse relation of (7.6) is

[Z] = [Z∞ ]1/2 {[1] + [S]}{[1] − [S]}−1 [Z∞ ]1/2

In a similar way we can obtain the following relations between the scattering matrix and the short circuit
admittance matrix:
[S] = {[1] − [y]}{[1] + [y]}−1
where the normalized admittance matrix is defined as:

[y] = [Zr ]1/2 [Y ][Zr ]1/2

The inverse relation is :

[Y ] = [Z∞ ]−1/2 {[1] − [S]}{[1] + [S]}−1 [Z∞ ]−1/2

7.4 Computation of the power dissipated in a device

Consider a N -port device, characterized by its scattering matrix [S]. The power dissipated in it is the sum
of the net active powers flowing into all ports:

N
1X 1
Pd = (|ai |2 − |bi |2 ) = ([a]T ∗ [a] − [b]T ∗ [b])
2 i=1 2

Now recall that

[b] = [S][a], [b]T ∗ = [a]T ∗ [S]T ∗
hence
1 T∗
Pd = [a] ([1] − [S]T ∗ [S])[a] (7.8)
2
It is straightforward to realize that this equation, for a one-port device (N = 1) reduces to

1 |V + |2
Pd = (1 − |Γ|2 )
2 Z∞

105
Renato Orta - Transmission Line Theory (Nov. 2012)

7.5 Properties of the scattering matrix [S] of a device

• A reciprocal device has a symmetric [S] matrix: [S] = [S]T .
The proof can be found in [5].
• A lossless device has unitary scattering matrix: [S]T ∗ [S] = [1].
This follows immediately from (7.8), by requiring that Pd = 0 for arbitrary excitation [a].
In the case N = 2 the previous equation yields:

|S11 |2 + |S21 |2 = 1
2 2
|S12 | + |S22 | = 1
2 2
|S11 | + |S12 | = 1
2 2
|S21 | + |S22 | = 1
∗ ∗
S11 S12 + S21 S22 = 0
∗ ∗
S11 S21 + S12 S22 = 0

These relations have a geometrical interpretation: the rows and the columns of the unitary matrix
[S] form an ortho-normal basis in the N dimensional complex linear vector space CN .
• A passive device has a scattering matrix [S] such that all the eigenvalues of [S]T ∗ [S] have a magnitude
less than (or at most equal to) 1.
• an active device has a scattering matrix [S] such that at least one eigenvalue of [S]T ∗ [S] has magnitude
greater than 1.

It is to be remarked that the eigenvalues of [S]T ∗ [S] are the squares of the singular values of [S], [6]

7.6 Change of reference impedances

Let [So ] be the scattering matrix of a device, with respect to a given set [Zro ] of reference impedances.
Suppose that we want to compute the scattering matrix [Sn ] of the same device, with respect to a new set
of reference impedances [Zrn ].
Let us address the problem gradually, by considering first the case of a one-port device, which has
reflection coefficient Γo with respect to Zro . Its reflection coefficient Γn with respect to Zrn is derived as:

Z − Zrn
Γn =
Z + Zrn
with
1 + Γo
Z = Zro
1 − Γo
Substituting and with a little of algebra, we find:

1 + Γo
Zro − Zrn
1 − Γo (Zro − Zrn ) + Γo (Zro + Zrn ) Γno + Γo
Γn = = =
1 + Γo (Zro + Zrn ) + Γo (Zro − Zrn ) 1 + Γno Γo
Zro + Zrn
1 − Γo
where we have induced the reflection coefficient of the old reference impedance with respect to to the new
one.
Zro − Zrn
Γno =
Zro + Zrn

106
Renato Orta - Transmission Line Theory (Nov. 2012)

We can now address the case of an N -port device along the same lines. From (7.6), we can express the
desired scattering matrix as
[ζn ] − [1]
[Sn ] = (7.9)
[ζn ] + [1]
Notice that the equation can be written in this form because both ([ζn ]−[1]) and ([ζn ]+[1])−1 are functions
of the same matrix [ζn ], hence they commute. Moreover, from (7.7),

[ζn ] = [Zrn ]−1/2 [Z][Zrn ]−1/2 = [Zrn ]−1/2 [Zro ]1/2 [ζo ][Zro ]1/2 [Zrn ]−1/2 =
= [R][ζo ][R]

where we have introduced the diagonal matrix [R] = [Zrn ]−1/2 [Zro ]1/2 . Notice that these matrices are
diagonal, hence they commute. Express now [ζo ] in terms of [So ] and substitute in (7.9). We find

[1] + [So ] [1] + [So ]

[R] [R] − [1] [R] − [R]−1
[1] − [So ] [1] − [So ]
[Sn ] = = =
[1] + [So ] [1] + [So ]
[R] [R] + [1] [R] + [R]−1
[1] − [So ] [1] − [So ]
([R] − [R]−1 ) + ([R] + [R]−1 )[So ] [S]no + [So ]
= =
([R] + [R]−1 ) + ([R] − [R]−1 )[So ] [1] + [S]no [So ]

where we have introduced the diagonal scattering matrix of the old reference impedances with respect to
the new ones
[R] − [R]−1 [R]2 − 1
[S]no = =
[R] + [R]−1 [R]2 + 1

7.7 Change of reference planes

The scattering matrix [S] of a device describes its input/output characteristic on the basis of the incident
and scattered waves on the transmission lines attached to each port. The complex amplitudes of these
waves are function of the longitudinal coordinate. Hence, a given S matrix refers always to a specific choice
of reference planes. We want now to examine the transformation of the [S] matrix induced by a change of
these planes.
Consider a N -port device with scattering matrix S0 (see Fig.7.3). The amplitudes of incident and
reflected waves at port i (i = 1,2, . . . ,N ) are indicated by a0i and b0i , respectively. The same amplitudes
at a distance li (> 0) away form the device are denoted by ai , bi , given by

ai = a0i ejki li , bi = b0i e−jki li

where ki is the propagation constant on the line connected to port i. Introduce now the column vectors
[a], [b], [a0 ], [b0 ], with components ai , bi , a0i , b0i , so that the previous equations can be written in matrix
form :
[a] = exp{+j[k]l} [a0 ]
(7.10)
[b] = exp{−j[k]l} [b0 ]
where exp{±j[k]l} is the diagonal matrix

exp(±j[k]l) = diag{ exp{±jk1 l1 } ... exp{±jkN lN } }

Denote with [S] the scattering matrix of the new structure obtained by the translation of the reference
planes. If [S0 ] is the scattering matrix of the original structure, we have:

[b0 ] = [S0 ][a0 ]

107
Renato Orta - Transmission Line Theory (Nov. 2012)

ai a0i
bi b0i port i

li a0j aj
port j b0j bj

lj
[S0]

[S]

Figure 7.3. Change of reference planes

Substitute (7.10):
[b] = [exp{−jkl}][S0 ][exp{−jkl}][a]
In other words
[S] = [exp{−jkl}][S0 ][exp{−jkl}] (7.11)
Make this equation explicit. For the elements on the main diagonal

Sii = S0ii e−2jki li

We recognize at once the analogy with the transformation rule of reflection coefficients (3.8). There should
be no surprise, since these matrix elements are indeed reflection coefficients at port i (when all the other
ports are terminated with the respective characteristic impedances). For the other elements the previous
relation becomes
Sij = S0ij e−j(ki li +kj lj ) , i 6= j
because the incident and scattered waves propagate on different transmission lines.

7.8 Cascade connection of structures

Very often it is convenient to view a complex system as made out of interconnected simpler blocks. As-
suming that their scattering matrices are known, the problem arises of computing the scattering matrix of
the complete structure.
With reference to Fig. 7.4, consider two structures, the first with N + K ports, the second with K + N
ports, which must be connected through the K ports, so that the complete structure has N + M ports.
In defining the power waves at the various ports, it is convenient to use a matrix formulation, separating
the ones referring to the ports that are going to be connected from the others.
This operation implies a partition in blocks of the scattering matrices of the two structures. The first
is written as  
  0
.. 0
 
N{ [b1 ]  [S 11 ] . [S 12 ]  [a1 ] }N
 ···  =  ··· ···   ···  (7.12)
 
K{ [b0 ] 0
.. 0
[a0 ] }K
[S21 ] . [S22 ]

108
 Renato Orta - Transmission Line Theory (Nov. 2012)

1 2

[a1] [a’] [a’’] [a2]

S’(N+K)(N+K) S’’(K+M)(K+M) [b2]
[b’] [b’’]
[b1]
M ports
K ports K ports
N ports S(N+M)(N+M)

Figure 7.4. Connection of two structures

0
where the blocks [Sij ] have the dimensions
0
[S11 ] → N ×N
0
[S12 ] → N ×K
0
[S21 ] → K ×N
0
[S22 ] → K ×K
Implicit in this partition is the assumption that the ports to be connected are the last K ones. This
condition can always be obtained by suitable exchanges of rows and columns. Indeed:

• the exchange of bi with bj requires the exchange of the rows i and j in [S 0 ]

• the exchange of ai with aj requires the exchange of the columns i e j in [S 0 ]

Likewise, the second structure is characterized by the scattering matrix [S 00 ] in the following way
 .. 
 00  00 00
 
K{ [b ]  [S 11 ] . [S 12 ]  [a00 ] }K
 ···  =  ··· ···   ···  (7.13)
 
M{ [b2 ] . [a2 ] }M
00
[S21 ] .. [S2200
]
00
where the blocks [Sij ] have the dimensions
00
[S11 ] → K ×K
00
[S12 ] → K ×M
00
[S21 ] → M ×K
00
[S22 ] → M ×M
In this case, it has been assumed the the ports to be connected are the first K ones.
Suppose also that the ports to be connected have the same reference impedances. In these conditions,
the equations that define the connection are
[a0 ] = [b00 ]
(7.14)
[a00 ] = [b0 ]

109
Renato Orta - Transmission Line Theory (Nov. 2012)

After the connection, the structure is accessible from the external world through N ports on the “1”
side and M ports on the “2” side and, hence, it is described by a scattering matrix [S] with dimension
(M + N ) × (M + N ), obviously partitioned in the following way
 .. 
   
N{ [b1 ]  [S 11 ] . [S 12 ]  [a1 ] }N
 ···  =  ··· · · ·  ···  (7.15)
 
M{ [b2 ] .. [a2 ] }M
[S21 ] . [S22 ]

where the blocks [Sij ] have the dimensions

[S11 ] → N ×N
[S12 ] → N ×M
[S21 ] → M ×N
[S22 ] → M ×M

In order to determine the resulting [S] matrix, it is necessary to eliminate the variables [a0 ], [a00 ], [b0 ], [b00 ]
from (7.12) and (7.13), via (7.14).
The steps to be performed are the following:

1. Substitute (7.14) into (7.13), the first of which becomes

[a0 ] = [S11 ][b0 ] + [S12

00
][a2 ] (7.16)

2. The second of (7.12) is

[b0 ] = [S21
0 0
][a1 ] + [S22 ][a0 ] (7.17)
0
Eliminating [a ] between (7.16) and (7.17), yields
© 00 0 ª
[b0 ] = [S21
0 0
][a1 ] + [S22 ] [S11 00
][b ] + [S12 ][a2 ]

from which we find the expression of [b0 ] as a function of [a1 ] e [a2 ]

¡ 00 ¢−1 © 0 ª
[b0 ] = [1] − [S22
0
][S11 ] 0
[S21 ][a1 ] + [S22 00
][S12 ][a2 ] (7.18)

3. Substituting (7.18) into (7.16) we find the expression of [a0 ] as a function of [a1 ] and [a2 ]:
−1
[a0 ] = 00
[S11 ] ([1] − [S220 00
][S11 ]) [S210
][a1 ]+
n o (7.19)
00 0 00 −1 0 00
+ [S11 ] ([1] − [S22 ][S11 ]) [S22 ] + [1] [S12 ][a2 ]
00 00 0
The curly parenthesis that multiplies [S12 ][a2 ] can be simplified. Set [X] = [S11 ] and [Y ] = [S22 ] for
a simpler reading. The expression to be rewritten is:

[X] ([1] − [Y ][X])−1 [Y ] + [1] =

factor [X] to the left and [Y ] to the right:

© ª
= [X] ([1] − [Y ][X])−1 + [X]−1 [Y ]−1 [Y ] =

factor ([1] − [Y ][X])−1 to the left:

© ª
= [X] ([1] − [Y ][X])−1 [1] + ([1] − [Y ][X]) [X]−1 [Y ]−1 [Y ] =

factor [X]−1 [Y ]−1 to the right:

= [X] ([1] − [Y ][X])−1 {[Y ][X] + [1] − [Y ][X]} [X]−1 [Y ]−1 [Y ] =

110
Renato Orta - Transmission Line Theory (Nov. 2012)

Simplify:
= [X] ([1] − [Y ][X])−1 [X]−1 =
rewrite as the inverse of a matrix:
© ª−1
= [X] ([1] − [Y ][X])−1 [X]−1 = {[1] − [X][Y ]}−1

In conclusion, and resuming the usual notation, (7.19) is rewritten as

−1
[a0 ] = 00
[S11 0
] ([1] − [S22 00
][S11 ]) 0
[S21 ][a1 ]+
−1
(7.20)
00 0 00
+ ([1] − [S11 ][S22 ]) [S12 ][a2 ]

4. Sostitute (7.20) in the first of (7.12), to obtain [b1 ] as a function of [a1 ] e [a2 ]:
n o
0 0 00 0 00 −1 0
[b1 ] = [S11 ] + [S12 ][S11 ] ([1] − [S22 ][S11 ]) [S21 ] [a1 ]+
0 00 0 −1 00
+ [S12 ] ([1] − [S11 ][S22 ]) [S12 ][a2 ]

From the comparison of this equation with (7.15), we get the expressions of [S11 ] e [S12 ].
5. Now substitute in the second of (7.13) [a00 ] = [b0 ], the latter being given by (7.18):
00 0 00−1 0
[b2 ] = [S21 ] ([1] − [S22 ][S11 ]) [S21 ][a1 ]+
n o
00 0 00 −1 0 00 0
+ [S21 ] ([1] − [S22 ][S11 ]) [S22 ][S12 ] + [S22 ] [a2 ]

Comparing this equation with the second of (7.15) we derive the expressions of the remaining elements
[S21 ] e [S22 ]. For reference sake, we collect the expressions of the four blocks:
0 0 00 0 00 −1 0
[S11 ] = [S11 ] + [S12 ] [S11 ] ([1] − [S22 ] [S11 ]) [S21 ]
0 00 0 −1 00
[S12 ] = [S12 ] ([1] − [S11 ] [S22 ]) [S12 ]
−1
(7.21)
00 0 00 0
[S21 ] = [S21 ] ([1] − [S22 ] [S11 ]) [S21 ]
00 00 0 00 −1 0 00
[S22 ] = [S22 ] + [S21 ] ([1] − [S22 ] [S11 ]) [S22 ] [S12 ]

In some applications, it is useful to know the value of [b0 ], as given by (7.18); we write it here again for
ease of reference: ¡ 00 ¢−1 © 0 ª
[b0 ] = [1] − [S22
0
][S11 ] 0
[S21 ][a1 ] + [S22 00
][S12 ][a2 ]

Note that if the structures of interest are two-ports to be connected through one of their ports, all the
sub-matrices are scalars.
Another particular case is the one in which all the ports of the second structure are connected, so that
it behaves as a K-port load. In this case M = 0 and the whole structure, having only N access ports, is
characterized by [S11 ] alone.
The computation of the comprehensive scattering matrix requires the inversion of a matrix with di-
mension equal to the number of ports K that are connected. This matrix is singular, hence non invertible,
when the comprehensive structure that originates from the connection is resonant.
If the constituents structures are passive, this happens only for complex frequency values, located in
the upper half-plane for stability reasons.
If the Taylor expansion is used to compute the inverse matrix

([1] − [A])−1 = [1] + [A] + [A]2 + . . . + [A]n + . . .

which is convergent if the structures are passive, so that the eigenvalues of [A] have amplitude less than 1,
we obtain the characterization of the comprehensive structure in terms of the multiple reflection series.

111
Renato Orta - Transmission Line Theory (Nov. 2012)

7.9 Scattering matrix of some devices

In this section we present the scattering matrix of some devices of common use in the microwave technology.
In the drawings we will use the common convention of using one line as the symbol of a port, instead of
two leads, as done previously.

7.9.1 Ideal attenuator

An ideal attenuator is a reciprocal device, matched at both ports, that produces an attenuation AdB on
the signals through it. Its scattering matrix is then:
· ¸
0 Ae−jϕ
[S] =
Ae−jϕ 0

where A = 10−AdB /20 . The phase shift is related to the physical size of the device.

7.9.2 Isolator
An ideal isolator, whose symbol is shown in Fig. 7.5, is a nonreciprocal device, matched at both ports,
that allows the undisturbed passage of a signal from port 1 to 2, but prevents it in the opposite direction.
The S matrix of the device is: · ¸
0 0
[S] =
e−jϕ 0
The non-reciprocity is evident from the fact that S12 6= S21 . An isolator contains a magnetic material

1 2

Figure 7.5. Symbol of an ideal isolator.

(ferrite), maintained in a static magnetic field. The power incident on port 2 is completely dissipated in
the device.

7.9.3 Circulator
An ideal circulator is a nonreciprocal device, whose symbol is shown in Fig. 7.6. It is matched at all ports
and the power propagating in the direction of the arrow suffers no attenuation, while the one flowing in
the opposite direction is completely dissipated. Its S matrix is then:
 
0 0 e−jϕ1
[S] =  e−jϕ2 0 0 .
−jϕ3
0 e 0

This matrix is clearly non symmetrical, because of the non reciprocity of the device. It can be shown that
a matched, lossless three-port structure is necessarily non reciprocal and is a circulator. Also in this case,
the behavior is due to a magnetized ferrite.
Note that if port 3 is terminated with a matched load, we obtain a two-port structure that behaves as
an isolator, see Fig. 7.7. Indeed, an incident signal at port 1 goes to port 2 without being influenced by
the load on port 3. On the contrary, a signal incident on port 2 is directed to port 3, where it is dissipated
in the matched load, without any power coming out of port 1.

112
Renato Orta - Transmission Line Theory (Nov. 2012)

1 3

Figure 7.6. Symbol of an ideal three port circulator.

1 2

Figure 7.7. Isolator constructed by terminating port 3 of a circulator with a matched load.

A circulator can be used to realize a “diplexer”, when in a communication system both the transmitter
and the receiver are connected to the same antenna, as it is shown in Fig. 7.8. It is clear that in practice
this configuration works only if the transmission and reception frequencies are different and a bandpass
filter is inserted between the circulator and the receiver. Indeed, S31 of practical devices is not small
enough.

TX
1 2

RX

Figure 7.8. Circulator used as a diplexer.

7.9.4 Ideal directional coupler

An ideal directional coupler, whose symbol is reported in Fig. 7.9, is a matched, reciprocal four port
structure, in which the two ports on each side are uncoupled. With a suitable choice of the reference
planes, its S matrix can be written:
 √ 
√ 0 1 − k2 jk 0
 1 − k2 0 0 
[S] =  √ jk .
 jk 0 0 1 − k 2 
√
0 jk 1 − k2 0
The quantity
C = −20 log10 k = −20 log10 |S31 |

113
Renato Orta - Transmission Line Theory (Nov. 2012)

is called coupling coefficient: there are 3 dB-couplers (called hybrids), 10 dB-, 20 dB-couplers, etc.

1 2
1− k 2

jk

4 1 − k2 3

Figure 7.9. Symbol of directional coupler.

Other parameters used to characterize a real directional coupler are the directivity D:
|S41 |
D = −20 log10
|S31 |
and the isolation I:
I = −20 log10 |S14 |.
The coupling C denotes the fraction of the incident power at port 1 that is transferred to port 3. The
directivity measures the ability of the directional coupler to discriminate the incident waves at port 1 from
those incident at port 2, by specifying how port 4 is isolated from 1.
The isolation I is related to the same concept, hence the following relation holds

I =D+C (in dB).

An ideal coupler has infinite isolation and directivity.

A directional coupler is the heart of the Network Analyzer, an instrument capable of measuring directly
the scattering parameters of a device. The concept of the measurement is illustrated in Fig. 7.10. From
the ratio (in amplitude and phase) of the signals coming out of ports 3 and 4, it is possible to derive the
value of the reflection coefficient.

jk L 1 − k 2V1+ 4 3 jkV1+

1 − k2
jk

V
1
+ 1 2

Figure 7.10. Measurement of the reflection coefficient of a load by means of a directional coupler.

7.10 Examples of analysis of structures described by S ma-

trices
In this section we illustrate the use of the scattering matrix for the analysis of simple networks.

114
Renato Orta - Transmission Line Theory (Nov. 2012)

7.10.1 Cascade connection of a two-port and a load

We want to compute the reflection coefficient at the input of a two-port whose port 2 is terminated with
a load, as shown in Fig. 7.11. The comprehensive structure is a one-port load and is characterized by a
1 × 1 scattering matrix, i.e. by the scalar Γin , that can be found by using equations (7.21):

S12 S21 ΓL
Γin = S11 + =
1 − S22 ΓL
S11 − S11 S22 ΓL + S12 S21 ΓL
= = (7.22)
1 − S22 ΓL
S11 − det [S] ΓL
= .
1 − S22 ΓL

where ΓL is computed with reference to Zr2 and Γin is referred to Zr1 .

Alternatively, it can also be useful to derive it directly. The equations of the structure are:


 b1 = S11 a1 + S12 a2

b2 = S21 a1 + S22 a2



a2 = Γ L b2

The first two equations describe the general operation of the two-port, the third one stipulates that port
2 is loaded. The desired result Γin is obtained by eliminating a2 and b2 . To this end, substitute a2 in the
first two equations
(
b1 = S11 a1 + S12 ΓL b2
b2 = S21 a1 + S22 ΓL b2
From the second we get the important relation

S21
b2 = a1
1 − S22 ΓL

Substitute this b2 into the first of (??) and get

S21 S12 ΓL
b1 = S11 a1 + a1
1 − S22 ΓL

From this, Eq.(7.22) is immediately derived.

It can be remarked that when ΓL = 0, Γin = S11 , as it is obvious from the definition of scattering
parameters.

S' ΓL
Γi

Figure 7.11. Two-port network, whose port 2 is loaded with an impedance

Observe finally that also the relation between ΓL and Γin is a bilinear fractional transformation.

115
Renato Orta - Transmission Line Theory (Nov. 2012)

7.10.2 Interconnection of two two-ports by means of a length of trans-

mission line
Consider the structure of Fig. 7.12, in which two two-ports are connected by means of a length of trans-
mission line. In practice, several devices can be modeled in this way, for example a one-resonator filter or
a Fabry-Perot interferometer. The S matrix of the comprehensive structure can be obtained in two steps:

l
S’ S’’
_
A S’’ B
S

Figure 7.12. Interconnection of two two-ports by means of a length l of transmission line.

• shift the reference plane of the right two-port from B to A, by means of (7.11)
µ −jθ ¶ µ −jθ ¶ µ 00 −2jθ 00 −jθ ¶
e 0 e 0 S11 e S12 e
[S̄ 00 ] = S 00 = 00 −jθ 00
0 1 0 1 S21 e S22

• use (7.21) to cascade the two structures.

We obtain
0 00 −2jθ 0
0 S12 S11 e S21
S11 = S11 + 0 00 −2jθ
1 − S22 S11 e
0 00 −jθ
S21 S21 e
S21 = 0 00 −2jθ (7.23)
1 − S22 S11 e
where θ = kl = ωl/vf is the electrical length of the line.
Suppose that the S matrices of the two two-ports do not depend on frequency, i.e. on θ. In practice
this is not strictly true, but if the devices are not resonating or very large, their frequency dependence is
much weaker than that of the exponential. In these conditions it is simple to obtain a plot of the amplitude
of the total transmission coefficient S21 (θ).
Setting ¯ 00 ¯ jϕ11 ¯ 0 ¯ jϕ22
00
S11 = ¯S11 ¯e 0
S22 = ¯S22 ¯e
the denominator of (7.23) becomes
¯ 00 ¯ ¯ 0 ¯ −j2(θ−(ϕ11 +ϕ22 )/2
D(θ) = 1 − ¯S11 ¯ ¯S22 ¯ e
00 0
The plot of D(θ) in the complex plane is clearly a circle with center in 1 + j0) and radius |S11 | |S22 |, as
shown in Fig.7.13a. Moreover, it is known that the operation of inversion transforms circles into circles, [2].
In particular, D−1 (θ) traces a circle, symmetrical with respect to the real axis with center in
1
C= 0 2 00 2
1 − |S22 | |S11 |
and radius
0 00
|S22 | |S11 |
R= 0 2 00 2
1 − |S22 | |S11 |

116
Renato Orta - Transmission Line Theory (Nov. 2012)

ℑm ℑm

1 ℜe ℜe

1-|S22'||S11''| 1+|S22'||S11''|
(1+|S22'||S11''|)-1 (1-|S22'||S11''|)-1

Figure 7.13. Plot of D(θ) (left) and of D−1 (θ) (right).

|S21|

|S21|max

|S21|min

θmin θmax θ

Figure 7.14. Plot of |S21 (θ)|.

Hence the plot of |S21 (θ)| is the oscillating curve shown in Fig. 7.14 where the minimum and maximum
values of the transmission coefficient are
0 00
|S21 | |S21 |
|S21 |MIN = 0 00
1 + |S22 | |S11 |
0 00
|S21 | |S21 |
|S21 |MAX = 0 00
1 − |S22 | |S11 |
and their position is
1
θMIN = (ϕ11 + ϕ22 + (2m + 1) π)
2
1
θMAX = (ϕ11 + ϕ22 + 2mπ)
2
It is also clear that from the shape of the curve |S21 (θ)| (for example obtained through a measurement) it
is possible to infer the characteristics of the discontinuities present on the line and their separation.
If the discontinuities on the line had a very high reflection coefficient, the structure would behave as
pass-band filter.

7.10.3 Change of reference impedance for a one-port load

Consider a load characterized by means of its reflection coefficient with respect to the reference impedance
Zr1 . We want to compute the reflection coefficient of the same load with respect to another reference
impedance Zr2 . We can make reference to Fig.7.15.

117
Renato Orta - Transmission Line Theory (Nov. 2012)

Zr2 Zr1 ΓL

-
A A+

Figure 7.15. Change of reference impedance.

We obtain the result by passing through the impedance:

1 + ΓL
ZA+ = Zr1
1 − ΓL

ZA+ − Zr2
ΓA− =
ZA+ + Zr2
Substituting the first equation into the second one, we can obtain a direct link between ΓA− and ΓL .
However, this can also be obtained by applying (7.21) to analyze the two cascaded discontinuities. The
first, the junction between the two lines, has the scattering matrix
 q 
Zr2
£ 0¤ ΓF Zr1
(1 + ΓF )
S = q 
Zr2
Zr1
(1 − ΓF ) −ΓF

where ΓF is the reflectin coefficient

Zr1 − Zr2
ΓF =
Zr1 + Zr2
This is called Fresnel reflection coefficient because of its analogy with the reflection coefficient defined in
optics at the interface between two semi-infinite media.
0 0
Note that S12 and S21 are only apparently different: if the expressions of ΓF are substituted, we can
verify that they are equal, as required by the reciprocity of the structure.
The second discontinuity is a one-port load, hence its scattering matrix S reduces to the reflection
coefficient ΓL .
By means of (7.21) we find
¡ ¢
1 − Γ2F ΓL ΓF + Γ L
ΓA− = ΓF + =
1 + Γ F ΓL 1 + Γ F ΓL

which is the desired equation.

7.11 Transmission matrix

Structures in which we can identify two “sides” with the same number N of ports, which can be viewed
as generalizations of two-port structures, are often characterized by means of a transmission matrix.
The transmission matrix relates the electrical states at the two sides of the structure in the power
wave basis. Referring to Fig. 7.16, we introduce a single z axis for the two sides and define forward and
backward waves in accordance with it.

118
Renato Orta - Transmission Line Theory (Nov. 2012)

The characterization of the device by means of the transmission matrix [T ] is then the following:
 .. 
   + 
[c+
1 ]  [T11 ] . [T12 ]  [c2 ]
 ···  =  ··· ···   ··· 
 
[c−
1 ]
.. [c−
2 ]
[T21 ] . [T22 ]

where all blocks are square with dimension N × N .

side 1 side 2

1 1
+
[c ]
1 [c2+ ]
[c1− ] [c2− ]

N N

Figure 7.16. Generalized two-port structure: definition of forward and backward

waves on the two sides of it.

When two structures of the type of Fig. 7.16, with transmission matrices [T1 ] and [T2 ] are connected
via N ports as shown in Fig. 7.17 the transmission matrix of the comprehensive structure is found by
means of
[T ] = [T 0 ][T 00 ]
The convenience of the transmission matrix is related to the simplicity of this composition law.

[T’] [T’’]

[T]

Figure 7.17. Cascade connection of two generalized two-ports.

A characteristic of the transmission matrix is that its elements cannot be defined directly by circuit
type equations but must be obtained by other groups of parameters such as the scattering ones via algebraic
relations.

119
Renato Orta - Transmission Line Theory (Nov. 2012)

To derive the relation between the [T ] and [S] matrices of the same device, observe first that the
forward and backward power waves are related to the incident and scattered ones by
£ +¤ £ +¤
c1 = [a1 ] c2 = [b2 ]
£ −¤ £ −¤ (7.24)
c1 = [b1 ] c2 = [a2 ]

The characterization of the device via the scattering matrix [S] is

 .. 
   
[b1 ]  [S11 ] . [S12 ]  [a1 ]
 ···  =  ··· ···   ··· 
 
[b2 ] . [a2 ]
[S21 ] .. [S22 ]
Substitute (7.24) into this
[c− + −
1 ] = [S11 ][c1 ] + [S12 ][c2 ]
(7.25)
[c+ + −
2 ] = [S21 ][c1 ] + [S22 ][c2 ]

Since the blocks [Sij ] are square, we can obtain [c+

1 ] from the second of (7.25)

[c+
1 ] = [S21 ]
−1 +
[c2 ] − [S21 ]−1 [S22 ][c−
2 ] (7.26)
Since [c+
1 ]
is expressed as a function of [c+
2 ]
and [c−
2 ],
we can identify the expressions of [T11 ] and [T12 ].
Substituting (7.26) into the first of (7.25), we obtain
¡ ¢
[c−
1 ] = [S11 ][S21 ]
−1 +
[c2 ] + [S12 ] − [S11 ][S21 ]−1 [S22 ] [c−
2 ]

from which the expressions of the other elements of the [T ] matrix can be obtained. In conclusion, the
transformation formulas are:
[T11 ] = [S21 ]−1
[T12 ] = −[S21 ]−1 [S22 ]
(7.27)
[T21 ] = [S11 ][S21 ]−1
[T22 ] = [S12 ] − [S11 ][S21 ]−1 [S22 ]
Obviously, if the submatrix [S21 ] of a structure is not invertible, such a structure cannot be represented
by a transmission matrix. An example is the structure of Fig. 7.18 if K < N .
The comprehensive structure is perfectly defined by its scattering matrix [S]. However, it can be
verified that the submatrix [S21 ] is not invertible, so that the transmission matrix [T ] does not exist.
Indeed, consider the expression of the block [S21 ] given by (7.21). Observe that the dimensions of the
various factors are
0
[S21 ] → K ×N
00
[S21 ] → N ×K
0 00 −1
([1] − [S22 ][S11 ]) → K ×K
hence the dimension of [S21 ] is N × N . However, this matrix has rank K at most, hence it has zero
determinant if K < N . From the linear algebra point of view, [S21 ] represents an operator that maps
vectors belonging to the complex vector space C N into vectors of C N through C K . If K < N , we are in
presence of a projection, a notoriously non invertible operation.
For completeness we list also the inverse relations of(7.27), which allow the computation of [S] from
[T ]:
[S11 ] = [T21 ][T11 ]−1
[S12 ] = [T22 ] − [T21 ][T11 ]−1 [T12 ]
[S21 ] = [T11 ]−1
[S22 ] = −[T11 ]−1 [T12 ]

120
Renato Orta - Transmission Line Theory (Nov. 2012)

N ports K ports N ports

Figure 7.18. Structure with 2N ports that arises from the “back-to-back” connection of
two substructures with N + K ports each. If K < N the transmission matrix of the
comprehensive structure is not defined.

121
Chapter 8

Time domain analysis of

transmission lines

8.1 Introduction
In the previous chapters we have discussed in detail the frequency domain analysis of transmission line
circuits. By this technique, we can compute directly the circuit response to time harmonic signals. In this
chapter we consider instead signals with arbitrary time dependence.
Take the case of a simple circuit, consisting of a generator and a load, connected by a transmission line,
as shown in Fig. 8.1. Suppose that the line is characterized by a phase constant k (ω) and characteristic
impedance Z∞ (ω), in general complex functions of frequency. Also the load impedance ZL and the
internal impedance of the generator are generic complex functions of frequency. We want to compute the
load voltage vB (t) that is produced by a generator with open circuit voltage e (t).
This problem can be conveniently described, as shown in Fig. 8.2, in the language of system theory.
The generator waveform e (t) is the system input, the load voltage vB (t) is the output. The system is:

• linear, because the line parameters k and Z∞ , as well as the impedances Zg and ZL are independent
of the voltages and currents in the circuit.
• time-invariant, because these parameters do not depend on time.

It is well known that for linear time invariant systems (LTI), the input-output relation can be expressed
in time domain as a convolution product
Z +∞
vB (t) = tv (t − τ ) e (τ ) dτ (8.1)
−∞

or in the frequency domain as an algebraic product as

VB (ω) = TV (ω) E (ω) (8.2)

where the impulse response tv (t) is related to the transfer function by a Fourier transform:
Z +∞
1
tv (t) = TV (ω) ejωt dω (8.3)
2π −∞

122
Renato Orta - Transmission Line Theory (Nov. 2012)

Zg

+
e(t) ZL

A B

Figure 8.1. Circuit consisting of a generator and a load, connected by a transmission line AB

tv(t)
e(t) TV(ω) vB(t)

Figure 8.2. Linear system interpretation of the circuit of Fig. 8.1

Eq. (8.2) yields the solution to our problem in the form

Z Z +∞
1 1
vB (t) = VB (ω) ejωt dω = TV (ω) E (ω) ejωt dω (8.4)
2π 2π −∞

Obviously, the transfer function TV (ω) is obtained without difficulty by the methods described in the
previous chapters. We find

ZA (ω) 1
TV (ω) = e−jk(ω)l (1 + ΓB (ω)) (8.5)
ZA (ω) + Zg (ω) 1 + ΓA (ω)

It is useful to note the following interpretation of (8.4):

• the input is decomposed into a linear combination of harmonic signals ejωt

Z +∞
1
e (t) = E (ω) ejωt dω (8.6)
2π −∞

each with amplitude E (ω) dω.

• since the system is LTI, the output signal is still harmonic, with amplitude
VB (ω) dω = TV (ω) E (ω) dω.
• vB (t) is obtained by summing all these signals.

Time-harmonic signals ejωt are “eigensignals” of LTI systems, in the sense that they travel through
the system unchanged, except for the multiplication by a complex number, which is the transfer function
evaluated at the frequency ω, [7]. This property explains the usefulness of the Fourier transform technique
in the analysis of LTI systems.

8.2 The group velocity

In general, the integral (8.4) must be evaluated by numerical techniques, such as the Fast Fourier Transform
(FFT). However, here we wish to focus on the role of the transmission line, hence we simplify the model

123
Renato Orta - Transmission Line Theory (Nov. 2012)

e(t)

Figure 8.3. Example of an amplitude modulated signal. Note that the envelope is slowly
varying in comparison with the carrier

of Fig. 8.1, by assuming that the generator is ideal (Zg = 0) and the line is matched (ZL = Z∞ ), so that
the transfer function becomes
TV (ω) = e−jk(ω)l (8.7)
Moreover, we assume that the signal e (t) is not too different from a time-harmonic signal. In particular,
we choose
e (t) = m (t) cos ω0 t (8.8)
where m (t) (envelope) is a slowly varying signal that is almost constant in a period T = 2π/ω of the cosine
(carrier). This signal, of the type shown in Fig. 8.3, is amplitude modulated and, in these conditions, is
quasi-monochromatic, as it can be ascertained by computing its spectrum. We find
1
E (ω) = = {m (t) cos ω0 t} = = {m (t)} ∗ = {cos ω0 t} =
2π
1
= M (ω) ∗ π {δ (ω − ω0 ) + δ (ω + ω0 )} =
2π
1 1
= M (ω − ω0 ) + M (ω + ω0 ) = (8.9)
2 2
where M (ω) is the Fourier transform of m (t). These functions are plotted in Fig. 8.4. Requiring that the
envelope is slowly varying with respect to the carrier is equivalent to assuming ωc ¿ ω0 , where ωc is the
highest frequency present in the spectrum of m(t). Hence the signal e(t) is indeed quasi-monochromatic.
Observe that vB (t) is computed from (8.4) as an integral over both negative and positive frequencies.
However, since vB (t) is real, its spectrum is hermitian, i.e.
VB (−ω) = VB∗ (ω) (8.10)
and the spectral integral can be limited to the positive omega half-axis. Indeed, by decomposing the
integration domain into two parts, we have
Z 0 Z +∞
1 ¡ ¢ 0 1
vB (t) = VB ω 0 ejω t dω 0 + VB (ω) ejωt dω (8.11)
2π −∞ 2π 0
where the integration variable in the first integral has been called ω 0 . Letting now ω = −ω 0 and using
(8.10), we find
Z +∞ Z +∞
1 1
vB (t) = VB∗ (ω) e−jωt dω + VB (ω) ejωt dω =
2π 0 2π 0

124
Renato Orta - Transmission Line Theory (Nov. 2012)

|Ε(ω)|

Μ(ω)

−ω0 −ωc ωc ω0

Figure 8.4. Spectrum of e (t). Also the spectrum of m (t) is shown. Since ωc ¿ ω0 ,
e (t) is almost time-harmonic

½ Z +∞ ¾
1
= 2Re VB (ω) ejωt dω (8.12)
2π 0

This transformation is equivalent to introducing the analytic signal associated to vB (t), whose spectrum
is zero for ω < 0 and 2VB (ω) for ω > 0.
At this point the problem is reduced to the evaluation of
½ Z +∞ ¾
1
vB (t) = Re M (ω − ω0 ) e−jk(ω)l ejωt dω (8.13)
2π 0
Taking into account that the signal is quasi-monochromatic, that is the support of its spectrum is a small
neighborhood of ω0 , we can think of substituting the function k (ω) with its Taylor expansion around
ω = ω0 ¯ ¯
dk ¯¯ 1 d2 k ¯¯
k (ω) = k (ω0 ) + (ω − ω ) + (ω − ω0 )2 + ... (8.14)
dω ¯ω0 2 dω 2 ¯ω0
0

In general, the propagation constant is complex, k (ω) = β (ω) − jα (ω). Suppose we truncate the previous
expansion at the second term for the real part and at the first term for the imaginary one, i.e. assume
β (ω) ' β (ω0 ) + β 0 (ω0 ) (ω − ω0 )
(8.15)
α (ω) ' α (ω0 )
This truncation, apparently asymmetrical, is justified by the actual behavior of k(ω) in the usual cases.
Substitute (8.15) into (8.13)
½ Z +∞ ¾
1 0
vB (t) ' Re ej(ω0 t−β(ω0 )l) e−α(ω0 )l M (ω − ω0 ) e−jβ (ω0 )(ω−ω0 )l ej(ω−ω0 )t dω (8.16)
2π 0
where actually the integral receives contribution only in the support of M (ω − ω0 ), which is by hypothesis
a small neighborhood of ω = ω0 . Letting Ω = ω − ω0 , the previous equation can be rewritten
½ Z +∞ ¾
j(ω0 t−β(ω0 )l) −α(ω0 )l 1 +j (t−β 0 (ω0 )l)Ω
vB (t) ' Re e e M (Ω) e dΩ (8.17)
2π −∞
where the lower limit has been shifted to −∞, without changing the value of the integral, so that we can
recognize the inverse Fourier transform of M (Ω) evaluated in t − β 0 (ω0 ) l:
¡ ¢
vB (t) ' Re{ej(ω0 t−β(ω0 )l) e−α(ω0 )l m t − β 0 (ω0 ) l } =

125
Renato Orta - Transmission Line Theory (Nov. 2012)

¡ ¢
= m t − β 0 (ω0 ) l e−α(ω0 )l cos (ω0 t − β (ω0 ) l) (8.18)
0
The quantity β (ω0 ) l has the dimensions of a time and is called group delay τg (ω0 )

l
τg (ω0 ) = β 0 (ω0 ) l = (8.19)
vg (ω0 )

with the consequent definition of the group velocity vg (ω0 )

¯
1 dω ¯¯
vg (ω0 ) = ¯ = (8.20)
dβ ¯¯ dβ ¯β(ω0 )
¯
dω ω0

These quantities get their names from the fact that the signal e (t) consists of a “group” of frequencies. If
we recall the definitions of phase velocity
ω0
vph (ω0 ) = (8.21)
β (ω0 )

and of the corresponding phase delay τph (ω0 ) = β (ω0 ) l/ω0 = l/vph (ω0 ), eq.(8.18) can be rewritten
µ ¶ µ ¶
l ω0 l
vB (t) ' m t − e−α(ω0 )l cos ω0 t − =
vg (ω0 ) vph (ω0 )

= m (t − τg (ω0 )) e−α(ω0 )l cos (ω0 (t − τph (ω0 ))) (8.22)

This equation has the following interpretation. The signal e (t) is not monochromatic but consists of
a “packet” of harmonic components, each of which appears in the point B weighted by the transfer
function. As a consequence of the constructive and destructive interference phenomena that here take
place, everything happens as if the envelope m (t) moved at the group velocity and the carrier with the
phase velocity. Obviously, this is only an interpretation of (8.22), since envelope and carrier are not signals
with an independent existence.
In the applications, the information is associated to the envelope. We see also that, in the limit in
which (8.15) hold, the envelope in B is an attenuated and delayed replica of the envelope in A, so that the
propagation can be defined as distortion free. Hence, a transmission line does not introduce distortions if,
at least in the band of the signal that is propagating, the transfer function has constant magnitude and
linear phase. The higher order terms in the expansion (8.14) are responsible for the distortions. Obviously,
they cannot be neglected when the bandwidth of e (t) is not small.
The concepts of phase and group velocity lend themselves to a geometrical interpretation. Consider a
dispersion curve as the one sketched in Fig. 8.5. On the basis of the previous definitions, we can write

vph (ω0 ) = tan ϕph vg (ω0 ) = tan ϕg (8.23)

Note that the concept of group velocity is the most physically important of the two. Indeed, it is the
propagation velocity of information and of energy on the line. It turns out to be always smaller than the
speed of light in vacuum, as required by the theory of relativity.
Note, finally, that the concept of group delay can be defined both for lumped and distributed devices.
The general definition is in fact
d
τg = − arg (H (ω)) (8.24)
dω
where H (ω) is the transfer function of the device. Recalling (8.7), we observe that (8.19) is in agreement
with this definition. In general terms, a group delay appears in the cases in which the device can store
energy. Obviously a resistor network has a real transfer function and τg = 0 according to (8.24).

126
Renato Orta - Transmission Line Theory (Nov. 2012)

ϕg
ω0
ϕf
β

Figure 8.5. Geometrical interpretation of phase and group velocity

8.3 Distortions
In the previous section we have seen that a quasi monochromatic signal is not distorted (more precisely,
its envelope is not distorted) when the group velocity is constant on the bandwidth of the signal itself. In
this section we will discuss the distortions caused by a transfer function with constant magnitude and a
phase curve that is non linear but can be approximated by a parabola. Hence, we will assume that the
propagation constant can be expressed in the form
1 00
β (ω) = β0 + β00 (ω − ω0 ) + β0 (ω − ω0 )2 (8.25)
2
where ¯ ¯
dβ ¯¯ d2 β ¯¯
β0 = β (ω0 ) β00 = β000 =
dω ¯ω0 dω 2 ¯ω0
and we will assume α (ω) = 0 for simplicity. This assumption implies that the group velocity (and the
group delay) are linear functions of frequency.
By repeating with minor modifications the computations that lead to (8.16) we find
½ Z +∞
1 1 1 00 2
vB (t) ' Re ej(ω0 t−β0 l) [M (ω + ω0 ) + M (ω − ω0 )] e−j 2 β0 l(ω−ω0 ) ·
2 2π 0
o
·ej (t−β0 l)(ω−ω0 ) dω
0
(8.26)
We see clearly that the quadratic phase term causes a distortion, but the computation can no longer be
carried out for a generic envelope m (t). The simplest case for which an analytic expression can be obtained
is that of a gaussian pulse, in which µ ¶
t2
m (t) = exp − 2 (8.27)
2T0
The standard deviation of the gaussian T0 can be used as a conventional measure of the duration of the
pulse, see Fig. 8.6. Making use of the integral
Z +∞ √ δ2
2 2 π − (2α)
e−α x ejδx dx = e 2
(8.28)
−∞ α

we obtain the spectrum M (ω) µ ¶

√ T 2 ω2
M (ω) = 2πT0 exp − 0 (8.29)
2

127
Renato Orta - Transmission Line Theory (Nov. 2012)

m(t)
1

e-1/2

T0 t

Figure 8.6. Envelope of a gaussian pulse and definition of its conventional duration T0

Note that the spectrum of the envelope is still gaussian with standard deviation T0−1 : in accordance with
the uncertainty principle for the Fourier transform, a short pulse has a large bandwidth and viceversa.
The integrand in (8.26) is the sum of two terms. In normal applications the pulse duration is much
larger of the carrier period, so that ω0 T0 >> 1 and the first term gives a completely negligible contribution.
For the same reason, as for the second, the lower limit of the integral can be shifted from 0 to −∞ without
changing its value.
Rewrite (8.26) in the form:
½ Z +∞ ¾
1 1
M (Ω) e−j 2 β0 lΩ ej (t−β0 l)Ω dΩ =
1 00 2 0
vB (t) ' Re ej(ω0 t−β0 l)
2 2π −∞
½ Z +∞ ¾
1 1
e− 2 (T0 +jβ0 l)Ω ej (t−β0 l)Ω dΩ
1 2 00 2 0
= Re ej(ω0 t−β0 l) (8.30)
2 2π −∞
By applying (8.28) with
¡ 2 ¢
α= 1
2
T0 + jβ000 l δ = t − β00 l (8.31)
we find ( ( ))
2
T0 (t − β00 l)
vB (t) ' Re p exp {j (ω0 t − β0 l)} exp − (8.32)
2
T0 + jβ000 l 2 (T02 + jβ000 l)
Transform now the expression in the curly brackets in such a way that the real part is obtained simply.
Note that the algebraic term can be rewritten as
"  1
· µ 00 ¶¸− 1  µ 00 ¶2 # 12 ½ µ 00 ¶¾− 2
T0 β0 l 2 β0 l β0 l
p = 1+j = 1+ exp j arctan =
T02 + jβ000 l T02  T02 T02 

" µ ¶2 #− 14 ½ µ 00 ¶¾
β000 l 1 β0 l
= 1+ exp −j arctan (8.33)
T02 2 T02
If we separate magnitude and phase, we arrive at the following final expression
µ ¶1 ½ ¾
T0 2 (t − l/vg (ω0 ))2
vB (t) = exp − cos (ϕ (t)) (8.34)
T (l) 2T (l)2

128
Renato Orta - Transmission Line Theory (Nov. 2012)

Figure 8.7. Distorted gaussian pulse at the line end. Observe the spurious frequency modulation

where the phase ϕ (t) is

µ ¶ √ µ√ ¶
l (t − l/vg (ω0 ))2 3(l/ld )sign (β000 ) 1 3l
ϕ (t) = ω t − + 2 − arctan (8.35)
vph (ω0 ) 2T02 1 + 3 (l/ld ) 2 ld

and s µ ¶2
l √ T02
T (l) = T0 1+3 ld = 3 00 (8.36)
ld |β0 |
Note that the signal at the far end of the line is still gaussian: this is related to the invariance of gaussians
with respect to Fourier transforms and to the fact that by expanding the phase constant k(ω) to the second
order also the line transfer function turns out to be a gaussian, albeit with imaginary variance.
The envelope moves at the group velocity but its variance increases from T02 to T 2 (l). Even if the
pulse retains the gaussian shape, it gets distorted because its standard deviation increases.
The quantity ld is defined doubling distance because T (ld ) = 2T0 . Note that ld increases with the initial
pulse duration T0 , because its bandwidth becomes narrower. Moreover it increases if the line dispersivity
becomes smaller, i.e. if |β000 | decreases. Note, moreover, that the maximum value of the envelope becomes
smaller and smaller during the propagation: it is simple to verify that the energy in the pulse does not
depend on the length of the line l, in accordance with the fact that the line has been assumed to be lossless.
Consider now the phase term and note the quadratic dependance on t − l/vg (ω0 ). Compute the instan-
taneous frequency √ 00
dϕ t − l/vg (ω0 ) 3(l/ld )sign(β0 )
ω (t) = = ω0 + 2 2 (8.37)
dt T0 1 + 3 (l/ld )
It changes linearly and increases or decreases depending on the sign of β000 . The output signal is affected
by a spurious frequency modulation, called chirp, depicted in Fig. 8.7.
Note that the transmission line is a symmetric device, hence if the signal (8.34) is made to propagate
from z = l to z = 0, the output signal will be
µ ¶
(t − l/vg (ω0 ))2
vA (t) = exp − cos(ω0 t)
2T02

In other words, it is possible to compress a pulse by exploiting the dispersivity of the line on which it
propagates. Obviously it is necessary that the signal to be compressed has a frequency modulation (chirp)
and that the sign of β000 is appropriate. Indeed, the compression is obtained by removal of the frequency
modulation.

129
Renato Orta - Transmission Line Theory (Nov. 2012)

z=0

    t

z=l

Figure 8.8. Intersymbol interference in a digital link on a dispersive optical fiber. Received pulses
are so distorted that the transmitted word is no longer recognizable

The considerations made for the gaussian pulse hold qualitatively for any pulse waveform. Basically,
the pulse duration increases during the propagation; however, since the pulse form changes, it is difficult
to define precisely the pulse duration.

8.4 Digital communication

As an example of the results presented above, consider an optical fiber link using a digital modulation such
that the transmission of a pulse is associated to the logical value 1 and the absence of a pulse to the value
0. If in the course of the propagation pulses suffer a distortion and increase their duration, an intersymbol
interference can take place such that pulses are no longer recognized by the receiver, as shown in Fig. 8.8.
To quantify the phenomenon, suppose that the receiver is able to recognize two pulses provided their time
separation is greater than the pulse duration multiplied by a factor r typical of the receiver.
Let BT be the bit rate, i.e. the number of bits transmitted per second. The condition of good operation
can be written
1
≥ rT (l)
BT
that is, recalling (8.36)
1 1
BT ≤ BT max = = s µ 00 ¶2 (8.38)
rT (l) β0 l
rT0 1 +
T02
If the characteristics of the line and of the receiver are specified, we can try to select the pulse duration
in such a way as to maximize BT max . Fig. 8.9 shows a plot of BT max (T0 ). The general characteristics of
the plot are easily explained. If T0 is small, the doubling length ld is also small, hence the bit rate must
be kept very low if the intersymbol interference is to be avoided. At the other extreme, if T0 is large the
doubling length is large, hence the distortion is small; however the bit rate is trivially low, in order to
avoid the interference already at the transmitter side.
Clearly, there is an optimum pulse duration that can be found by setting to zero the derivative of (8.38)
p
T0opt = |β000 | l

1
BT max (T0opt ) = √ p 00
2r |β0 | l

130
Renato Orta - Transmission Line Theory (Nov. 2012)

BTmax

T0opt T0

Figure 8.9. Plot of the maximum bit rate BT max on a digital link versus the pulse duration

Note that T0opt does not depend on the receiver characteristics,

√ but only on those of the fiber. In these
conditions, the pulse duration increases by the factor 2 in the propagation between transmitter and
receiver.
A remarkable property of silica optical fibers is that the parameter β000 is zero at the wavelength
λ = 1.3µm, so that the doubling length becomes infinite. Obviously, this does not mean that the pulse
are not distorted, but that the analysis carried out above is no longer applicable. Indeed, if β000 6= 0 it is
possible to neglect the higher order terms in (8.25), but if β000 = 0 it is necessary to take into account at
least the cubic term, which gives rise to a different type of distortion. In any case, it is always convenient
to operate the link at this wavelength, and for this reason the band centered around λ = 1.3µm is called
second window for its importance in the optical communications. Unfortunately, ordinary silica fibers
show their minimum attenuation (about 0.2 dB) in the third window, centered at λ = 1.55µm. Various
technological solutions have been devised to unify at the same wavelength the properties of low losses and
low dispersion.

8.5 Mismatched ideal transmission lines

In the previous sections we have discussed the behavior of transmission lines alone, in the assumption that
the terminations were matched. In this section we examine the effects of load and generator mismatch
but, in order to proceed by small steps, we assume first that the line is ideal, i.e. non dispersive.

8.5.1 General solution of transmission line equations

As an example of the technique that we are going to use, let us obtain first of all the general solution of
the time domain transmission line equations. In section 1.4 we presented the classical d’Alembert solution,
based on a suitable change of variable, here we obtain the desired result by the Fourier transform technique.
We know that the general solution in the ω-domain is
V (z,ω) = V0+ (ω) exp(−jkz) + V0− (ω) exp(jkz)
(8.39)
I (z,ω) = Y∞ V0+ (ω) exp(−jkz) − Y∞ V0− (ω) exp(jkz)
with r
√ ω C
k = ω LC = Y∞ = (8.40)
vph L

131
Renato Orta - Transmission Line Theory (Nov. 2012)

Rg l

+
e(t) Z∞ , vf RL

A B

Figure 8.10. Basic circuit, comprising a mismatched ideal line.

Note that V0+ (ω) and V0− (ω) are two arbitrary constants with respect to z ma can certainly depend on the
parameter ω. Their explicit expression can be defined when the load and generator are specified. Compute
then the time domain voltage as
Z +∞
1
v (z,t) = V (z,ω) exp(jωt)dω =
2π −∞

Z +∞ · ¸
1 + z − z
= V0 (ω) exp(−j ω) + V0 (ω) exp(+j ω) exp(jωt)dω =
2π −∞ vph vph
Z +∞ · µ µ ¶ ¶ µ µ ¶ ¶¸
1 z z
= V0+ (ω) exp j t − ω + V0− (ω) exp j t + ω dω =
2π −∞ vph vph
µ ¶ µ ¶
z z
= v0+ t − + v0− t + (8.41)
vph vph
where v0± are the inverse Fourier transforms of V0± (ω).
Moreover, as for the current, since Y∞ does not depend on ω, it is straightforward to write
µ ¶ µ ¶
z z
i (z,t) = Y∞ v0+ t − − Y∞ v0− t + (8.42)
vph vph

Hence, we have obtained again, by the Fourier transform technique, the same expressions of section 1.4.
Note that the general solution is clearly constituted by two waves propagating in opposite directions. In the
ω-domain this behavior is indicated by the fact that the two wave components have a phase proportional
to ω. In turn, this is a consequence of the fact that the phase velocity is frequency independent.

8.5.2 Mismatched ideal lines

Consider now the main problem of this section, shown in Fig. 8.10. Note that for simplicity we have
assumed that both the generator internal impedance and the load impedance are pure resistances, hence
frequency independent. Suppose we want to compute the load voltage vB (t). In section 8.1, we obtained
vB (t) in the form of an inverse Fourier transform:
Z Z +∞
1 1
vB (t) = VB (ω) ejωt dω = TV (ω) E (ω) ejωt dω (8.43)
2π 2π −∞

with
ZA (ω) 1
TV (ω) = e−jωτ (1 + ΓB (ω)) (8.44)
ZA (ω) + Zg (ω) 1 + ΓA (ω)

132
Renato Orta - Transmission Line Theory (Nov. 2012)

where E(ω) is the Fourier transform of the open circuit generator voltage e(t) and τ = l/vph is the
transit time on the line. The previous equations contain both reflection coefficients and impedances. It is
convenient to eliminate the latter to obtain a homogeneous expression. Define
1 + Γg 1 + ΓA
Rg = Z ∞ ZA = Z∞ (8.45)
1 − Γg 1 − ΓA

where Γg is the voltage reflection coefficient of the generator internal impedance. It can be verified that
the following relation holds
ZA 1 − Γg 1 + ΓA
= (8.46)
ZA + Zg 2 1 − ΓA Γg
and, recalling that ΓA = ΓB exp(−j2ωτ ), we obtain

1 − Γg 1
VB (ω) = E(ω) (1 + ΓB ) exp(−jωτ ) (8.47)
2 1 − Γg ΓB exp(−j2ωτ )

Note that because of the assumptions we made on load and generator, Γg and ΓB are frequency indepen-
dent. To compute the inverse Fourier transform, two alternative routes can be followed, which give rise
to two radically different forms of writing the solution. The first displays the dynamic evolution of the
phenomenon, the second yields a description in terms of resonances, i.e. of stationary states.

The lattice diagram

Consider the last fraction in (8.47) and note that it can be expanded by the binomial theorem:

(1 − Γg ΓB exp(−j2ωτ ))−1 = 1 + Γg ΓB exp(−j2ωτ ) + Γ2g Γ2B exp(−j4ωτ ) + ... (8.48)

The expansion is certainly convergent if the load is passive, because in this case |Γg ΓB exp(−j2ωτ )| < 1.
Substituting the expansion into (8.47) we find

1 − Γg
VB (ω) = E(ω) (1 + ΓB ) exp(−jωτ )·
2
© ª
· 1 + Γg ΓB exp(−j2ωτ ) + Γ2g Γ2B exp(−j4ωτ ) + ... (8.49)
and taking the inverse Fourier transform termwise, we obtain:
1 − Γg © ª
vB (t) = (1 + ΓB ) e(t − τ ) + Γg ΓB e(t − 3τ ) + Γ2g Γ2B e(t − 5τ ) + ... (8.50)
2
Apparently the solution is given in the form of an infinite series. In reality, in general, we are interested to
compute vB (t) for 0 ≤ t ≤ tmax , i.e. within a certain observation window. Obviously, the function e(t) is
causal, i.e. it is zero for negative argument. Hence, for a fixed time t ≤ tmax , there is only a finite number
of terms that contribute. This is easily explained, by noting that each term of the sum
1 − Γg n n
Γg ΓB e(t − (2n + 1)τ ) (1 + ΓB ) (8.51)
2
represents a wave that has travelled 2n + 1 times the length AB in the forward and backward direction,
with n + 1 reflections at the far end B (load) and n at the near end A (generator). Moreover, the factor
(1 + ΓB ) originates from the fact that the total voltage in B is the sum of the forward and backward
components. Hence, if n is sufficiently large, tmax − (2n + 1)τ is negative and this and the subsequent
terms do not give any contribution.
To interpret the first factor (1 − Γg )/2 it is convenient to rewrite it in terms of impedances. We find

1 − Γg Z∞
= (8.52)
2 Rg + Z ∞

133
Renato Orta - Transmission Line Theory (Nov. 2012)

Rg

+
e(t)

Figure 8.11. Definition of the surge impedance

This function is clearly the partition factor vA (t) /e(t) for the circuit of Fig. 8.11 in which the line is
infinitely long. This circuit is applicable also in the case of Fig. 8.10, but only for t ≤ 2τ , because in this
case the signal launched from A, even if it has reached the far end B, produces an echo that reaches A
no sooner than t = 2τ . Hence, before this time, the generator cannot ”know” if the line is infinite or not.
We can say that Z∞ is the input impedance of the line but only for t ≤ 2τ and this justifies the name of
”surge impedance” that sometimes is used to denote Z∞ . In this light, we can introduce the voltage in A
at the time t = 0+
Z∞
vA0 (t) = e(t)
Rg + Z ∞
so that (8.50) can be rewritten
© ª
vB (t) = (1 + ΓB ) vA0 (t − τ ) + Γg ΓB vA0 (t − 3τ ) + Γ2g Γ2B vA0 (t − 5τ ) + ... (8.53)

On the basis of this interpretation it is possible to draw a space-time plot, called “lattice diagram”, shown
in Fig. 8.12, which allows to write directly the expression of the transient response without computing first

Zg

+
e(t) ZL

A B
z
1-Γg e(t)
2
ΓB

Γg

ΓB

Γg

Figure 8.12. Lattice diagram for the circuit of Fig.8.10

134
Renato Orta - Transmission Line Theory (Nov. 2012)

Time domain response Max [ Vg(t) ] = 1

0.5
VL(t) (nat)

-0.5

-1

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

t / T0

1
Figure 8.13. Load voltage in the multiple echo conditions (τ > T ,
2 0
Zg = 10Ω,
Z∞ = 150Ω, ZL = 300Ω)

the frequency response and then evaluating the inverse Fourier transform of it.

Suppose that the open circuit voltage of the generator is a signal of duration T0 . Depending on the
line length, i.e. on the transit time τ , two different situations can appear.

− If τ < 12 T0 the supports of the functions e(t − (2n + 1)τ ), for successive values of n, overlap partially:
this is called a reverberation condition.

− If τ > 12 T0 the supports of the functions e(t − (2n + 1)τ ), for successive values of n, are disjoint: this
is called a multiple echo conditions.

Figs. 8.13 and 8.14 show the two conditions in the case e(t) is a rectangular pulse with value 1V. The
dashed line is the plot of the voltage on the load if this were directly connected to the generator. The
continuous line is the plot of the load voltage vB (t) when the transmission line is present. Its length is
deducible from the delay τ of the pulse front. Note that the successive echoes have decreasing amplitude,
because the common ratio of the geometrical series is smaller than one.

Having introduced (and proved) the lattice diagram method for the computation of vB (t), we can use
it for the computation of the voltage at the near end A and in a generic intermediate point C. Fig. 8.15
shows the relevant diagram. The voltage vA (t) is written immediately

1 − Γg
vA (t) = {e(t) + ΓB (1 + Γg )e(t − 2τ )+
2
ª
+Γ2B Γg (1 + Γg )e(t − 4τ ) + . . . (8.54)

135
Renato Orta - Transmission Line Theory (Nov. 2012)

Time domain response Max [ Vg(t) ] = 1

0.5
VL(t) (nat)

-0.5

-1

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

t / T0

Figure 8.14. Load voltage in the reverberation conditions (τ < 21 T0 , Zg = 10Ω, Z∞ =

150Ω, ZL = 300Ω)

A C B
z
τC 1-Γg
(1+ΓB)e(t-τ)
2

1-Γg
ΓB(1+Γg)e(t-2τ)
2
1-Γg
ΓgΓB(1+ΓB)e(t-3τ)
2

1-Γg Γ2 Γ (1+Γ )e(t-4τ)

B g g
2

Figure 8.15. Lattice diagram showing the computation of the voltages in vari-
ous points of the line (A, B, C).

136
Renato Orta - Transmission Line Theory (Nov. 2012)

whereas the voltage vC (t) is given by:

1 − Γg
vC (t) = {e(t − τc ) + ΓB e(t − (2τ − τc ))+
2
ª
+ΓB Γg e(t − (2τ + τc )) + Γ2B Γg e(t − (4τ − τc )) + . . . (8.55)

where τC = lAC /vph is the travel time from A to C. Note that the forward and backward voltages in C
are not simultaneous and this explains the absence of a factor of the type 1 + ΓB o 1 + Γg .
Sometimes we can be interested in computing the current in the load iB (t). The easiest method is just
to apply Ohm’s law:
vB (t)
iB (t) = (8.56)
RL
with vB (t) given by (8.50). The current at the near end iA (t) can be found similarly, by applying the
Kirchhoff loop law:
e(t) − vA (t)
iA (t) =
Rg
where vA (t) is computed by (8.54).
Not so easy is the case of iC (t), where C is an intermediate point. Indeed, here it is not possible to
apply Kirchhoff laws. However, the solution is readily found by reconsidering eq.(8.55) and noting that
this expression consists of forward and backward waves. The relevant impedance relations are
+ −
vC (t) vC (t)
i+
C (t) = i−
C (t) = −
Z∞ Z∞

hence we can write the desired current as

1 − Γg
iC (t) = {e(t − τc ) − ΓB e(t − (2τ − τc ))+
2Z∞
ª
+ΓB Γg e(t − (2τ + τc )) − Γ2B Γg e(t − (4τ − τc )) + . . . (8.57)

Since this is a general formula, it should hold also if C −→ B:

1 − Γg
iB (t) = {e(t − τ ) − ΓB e(t − τ )+
2Z∞
ª
+ΓB Γg e(t − 3τ ) − Γ2B Γg e(t − 3τ ) + . . . (8.58)

i.e.
1 − Γg
iB (t) = (1 − ΓB ) {e(t − τ ))+
2Z∞
+ΓB Γg e(t − 3τ ) + . . . } (8.59)

This expression is apparently different from (8.56), but if the reflection coefficients in (8.59) are substituted
with their expressions in terms of impedances, the identity of the two expression is readily proved.
It is convenient to analyze in greater detail the definition of the characteristic impedance of the line
as surge impedance, introduced in connection with eq.(8.52). Indeed, we must clarify the actual meaning
of an impedance in time domain. Recall that impedance is a frequency domain concept, introduced to
characterize the linear behavior of a one port device:

VL (ω) = ZL (ω)IL (ω)

This equation has the circuit translation and the input-output system interpretation of Fig. 8.16. The
impedance then can be considered as the transfer function of the LTI system. We know that the same

137
Renato Orta - Transmission Line Theory (Nov. 2012)

IL
VL
ZL I L (ω ) Z L (ω ) VL (ω )

Figure 8.16. Definition of impedance of a one port device: circuit (left) and LTI system
(right) points of view.

system is characterized in time domain by its impulse response z(t), obtained as inverse Fourier transform
of the transfer function ZL (ω). In terms of that, the model equation is
Z ∞
vL (t) = zL (t) ∗ iL (t) = zL (t − t0 )iL (t0 )dt0
−∞

This is the correct definition of impedance in time domain and certainly, in general, zL (t) 6= vL (t)/iL (t).
However, in the case of a (frequency independent) ideal resistor RL , zL (t) = RL δ(t) since it is the inverse
transform of a constant, and we recover the standard Ohm’s law
Z ∞
vL (t) = RL δ(t − t0 )iL (t0 )dt0 = RL iL (t)
−∞

The reason is that a resistor does not store energy (as capacitors and inductors do), hence the system is
without memory and the response is instantaneous, without delay.
Analogously, if the one port device is characterized by its reflection coefficient ΓL (ω), the input-output
equation is
VL− (ω) = ΓL (ω)VL+ (ω)
in frequency domain and
Z ∞
−
vL (t) = γL (t − t0 )vL
+ 0
(t )dt0
−∞

where γL (t), the inverse transform of ΓL (ω), is the impulse response of the system, expressed in the basis
of forward and backward voltage waves instead of total voltage and current.
As an application, let us compute the time domain input impedance of an ideal line with travel time τ ,
terminated by the resistive load RL .

1 + ΓA
ZA = Z∞ = Z∞ (1 + ΓA )(1 + ΓA + Γ2A + Γ3A + . . .) =
1 − ΓA
£ ¤
= Z∞ 1 + 2ΓL exp(−j2ωτ ) + 2Γ2L exp(−j4ωτ ) + 2Γ3L exp(−j6ωτ ) + . . .

from which we obtain easily (note that ΓL is frequency independent)

£ ¤
zA (t) = Z∞ δ(t) + 2ΓL δ(t − 2τ ) + 2Γ2L δ(t − 4τ ) + 2Γ3L δ(t − 6τ ) + . . .

From this expression it is evident, also from the mathematical point of view, that the terminated line
behaves as a simple resistor of value Z∞ for 0 < t < 2τ .
If the circuit contains more discontinuities, such as changes of characteristic impedance, lumped loads
connected in series and parallel, etc, the lattice diagram complexity increases exponentially and the method
becomes useless.

138
Renato Orta - Transmission Line Theory (Nov. 2012)

ℑm ω


ℜe ω

Figure 8.17. Integration path in the complex ω-plane and position of the singularities
of the integrand function

Solution in terms of resonances

Consider again (8.47) and compute directly the inverse transform integral by the complex analysis methods.
We must evaluate
Z +∞
1 E(ω) exp(jωt)
vB (t) = (1 − Γg ) (1 + ΓB ) exp(−jωτ ) dω (8.60)
2π −∞ 2 1 − Γg ΓB exp(−j2ωτ )

The integration path runs along the real axis as indicated in Fig. 8.17, but can be transformed into a
closed path by the addition in the upper halfplane of a half circle with radius tending to infinity, in order
to apply the residue theorem. The added half circle does not contribute to the integral because of Jordan’s
lemma.
The singularities of the integrand, beyond those of E(ω), are indicated by a cross. They are simple poles,
located in those ω values in which the denominator of (8.60) vanishes:

1 − Γg ΓB exp(−j2ωτ ) = 0 (8.61)

It is important to note that the quantity Γg ΓB exp(−j2ωτ ) is the system loop gain. As well known, the
poles identify the free oscillations. We see here clearly the two ways in which the frequency of a free
oscillation can be defined: either as a pole of the transfer function or as the frequency for which the loop
gain has magnitude one and phase n2π, for arbitrary integers n.
The solution of (8.61) is
1
ωn = [log |ΓB Γg | + j (arg (ΓB Γg ) + n · 2π)] =
2jτ
1 1
= (arg (ΓB Γg ) + n · 2π) − j log |ΓB Γg | (8.62)
2τ 2τ
with n = 0, ±1, ±2 . . .
Remember that |ΓB Γg | ≤ 1, so that these singularities, infinite in number, have all the same imaginary
part, which is non negative, coherently with the absence of gain in the system.
Observe that when ΓB Γg > 0 (i.e. if Rg and RL are both greater or smaller than Z∞ ), the pole for
n = 0 has zero real part. On the contrary, in the case ΓB Γg < 0 (i.e. Z∞ belongs to the interval [Rg , RB ]),
the poles are in symmetrical positions with respect to the imaginary axis and thei separation is π/τ .
Compute now the residues in the poles ωn . Applying the de l’Hôspital rule, we find
ω − ωn 1 1
Rn = lim = lim = (8.63)
ω→ωn 1 − Γg ΓB e−j2ωτ ω→ωn −Γg ΓB e−j2ωτ (−j2τ ) j2τ

If we assume that E(ω) is an entire function, i.e. without singularities in any finite region of the complex
ω-plane, which happens if e(t) has finite duration, then there are no other singularities and the response

139
Renato Orta - Transmission Line Theory (Nov. 2012)

Z∞ C RL

Figure 8.18. Transmission line connected to a logical gate

vB (t) can be written

X +∞
X
1
vB (t) = 2πj Res(Integranda; ωn ) = (1 − Γg ) (1 + ΓB ) E(ωn )e+jωn (t−τ ) (8.64)
n
2τ n=−∞

When the solution is expressed in this way, the dynamical response of the system is represented in terms
of the system normal modes. It is an alternative representation that is completely equivalent to that of
(8.50) in terms of multiple reflections.

8.5.3 Real interconnections

In a real world interconnection problem the various distortions mechanisms, which we have considered
separately, are simultaneously present. Limiting ourselves to a short list, we can mention:

• The load which terminates the line is not frequency independent. This implies that the various
echoes have a different shape one from the other and from the incident signal. A typical example is
that of a line connected to a logical gate, characterized by an input capacity, as shown in Fig. 8.18.
The resistance in parallel to C is Req = (RL Z∞ ) / (RL + Z∞ ), hence the time constant of the RC
group is CReq . In general we can say that the echoes have rising and descending fronts smoother
than those of the incident pulse, since the load behaves essentially as a low pass filter.
• If the line is not ideal, both the signal incident on the load and the successive echoes suffer distortions
in the course of their propagation.
• Often, as in printed circuit boards (PCB) there are several lines on the same board. These lines can
be considered as independent only in a first approximation. A more accurate model describes them
as multiconductor transmission lines, in which cross-talk effects appear.
• Finally, if the loads that terminate the lines are non linear, different types of distortion arise, which
in general can be studied only by sophisticated numerical techniques

140
Bibliography

[1] S. Ramo, J.R. Whinnery, Th. van Duzer, Fields and waves in communication electronics, New York,
Wiley, 1994.
[2] T. Needham, Visual complex analysis, Oxford: Clarendon, 1997.
[3] P.C. Magnusson, G.C. Alexander, V.K. Tripathi, Transmission lines and wave propagation, Boca Ra-
ton, CRC 1992
[4] D. M. Pozar, Microwave engineering, Reading: Addison-Wesley, 1990.
[5] R. E. Collin , Foundations for microwave engineering, New York: McGraw-Hill 1992.
[6] G. H Golub, C. F. Van Loan, Matrix computations, Baltimore: The Johns Hopkins University Press,
1983.
[7] A. Papoulis, “The Fourier integral and its applications”, McGraw-Hill Book Company Inc., New York
1962, pp. 81-44.

141