Chapter 2

Characterization of Microwave Networks

2.1
2.1

Introduction
Scattering Parameters
2.1.1 Definition
2.1.2 Advantages of Scattering Matrix Representation
2.1.3 Some Characteristics of S-matrices
2.2
Scattering Matrices for Some 2-port Networks
2.2.1 S-matrices of Symmetrical 2-port Networks
2.2.2 General Method for Determining S-matrices
2.2.3 Analysis of arbitrarily connected 2-ports
2.2.4 Analysis of Arbitrarily Connected Networks using Multiport
Connection Method
2.3
Shift in Port Reference Planes
2.4
Signal-flow Graph
2.3.1 Rules for Constructing Signal-flow Graphs
2.3.2 Simplification of Signal-flow Graphs
2.3.3 Applications of Signal-flow Graphs
2.5
Cascaded Two-ports (ABCD Parameters)
2.6
Immittance Matrices
2.7
Measurement of S-parameters using Vector Network Analyzer
2.7.1 Calibration of Network Analyzer
2.7.2 De-embedding
Appendix-2A
Appendix-2B
Appendix-2C

Chapter 2
Characterization of Microwave Networks
The microwave circuits have the following distinguishing features primarily due to
the their size being comparable to the wavelength of operation.
(a) Radiation from the circuit: We know from radiation theory that power radiated by
a short wire varies as ( l/ )2 . Therefore, longer the size of the circuit element, larger
will be the power radiated. The radiated power not only gives rise to loss of signal it
may give rise to undesired coupling between neighboring circuit elements affecting
the circuit performance.
(b) Coupling: Any two open transmission lines running in close proximity to each
other gives rise to coupling of fields between them. This results in transfer of power
from one line to the other. The coupling is not significant at low frequencies because
the coupling length is a small fraction of wavelength. However, the coupling becomes
significant at microwave frequencies and this phenomenon is used to design
directional couplers, power splitters and combiners.
(c) Parasitic: Parasitic is defined as something undesirable. The microwave circuits
include junctions and bends of transmission lines, open ends, shorts and so many
other types of discontinuities. These discontinuity effects can be characterized in
terms of additional reactances. Since these reactances are unintentional at the initial
circuit design stage, they are called parasitics and affect the accuracy of design. In the
accurate or integrated circuit design, the parasitics are included in the design process.
(d) Skin effect and losses: The skin depth is inversely proportional to the square root
of frequency of operation. As a result, the penetration of fields in the good conductors
is limited to a few skin depths only. The conductors therefore have higher loss at
microwave frequencies than at lower frequencies.
(e) Reference planes: The circuits at microwave frequencies are enclosed between
reference planes at the ports of the network. This is due to the fact that voltages and
currents have spatial variations, and therefore the network parameters [Z], [Y], or [S]
depend on the reference planes. Any shift from the reference planes results in change
in network characteristics.
At lower frequencies, networks are characterized invariably in terms of voltages and
currents at various terminals. This leads to the formulation of immitance (impedance
or admittance) parameters or ABCD parameters, more formally known as Aparameters. However, at microwave frequencies it is difficult to measure immitance
matrices of circuits. Therefore, microwave circuits are described theoretically by
wave variables a and b at various ports, rather than by voltages and currents.
Relationships among variables a and b are expressed by means of a scattering matrix,
whose elements are known as S-parameters. It does not mean that immitance or
ABCD-parameters are not used at all at microwave frequencies. We shall see,
however, that ABCD-parameter representation is very useful for analyzing two-port
circuits.
2.1
2.1.1

Generalized Scattering Parameters

Definition
2

The microwave networks are best analyzed in terms of wave propagation. It is

therefore appropriate to characterize microwave networks using wave amplitudes at
the ports. Consider an N-port network or component shown in Fig. 2.1. Let
Z 01 , Z 02 , ..
Z 0 N be the characteristic impedances of the transmission lines
connected to these ports. In place of the voltages and currents at the terminals of low
frequency networks, we employ the incident and the reflected wave amplitudes at the
ports to determine the characteristics of microwave networks. The incident and
reflected wave amplitudes at the n-th port is denoted by an and bn , respectively;
which may be defined in terms of the voltages as
V +
V
n
n
,
(2.1)
Z
Z
0n
0n
an =
bn =
+

where V and V are the voltage wave amplitudes in the transmission line (or
n
n
waveguide) connected to the nth port. The voltages are complex numbers with
magnitude and phase. The generalized S-matrix refers to the S-matrix of a network
with different port impedances, that is, in general Z 01 Z 02 Z 0 N Z 0 . The Smatrix of a network with Z 01=Z 02= Z 0 N =Z 0 is readily obtained from the
generalized S-matrix.
To justify the definition (2.1) for wave variables we compute power flow into and out
of ports. For this purpose, port voltage V n and port current I n at the nth port
(see Fig. 2.1) may be written as: (Chapter1, (1.10), (1.11) and (1.17) with z = 0 at the
port plane),
= Z 0 n ( a n+ bn )
+
(2.2)
V n=V n +
V n
and
V

+ - V /Z = a b / Z
(2.3)
n
0n ( n
n) 0n

I n=
The negative sign in (2.3) is due to the direction assigned to the backward current as a
convention. The net power flowing into the nth port may be expressed as
1
1
W n= ( V n I n )= ( an anbn bn )
(2.4)
2
2
This shows that W n is equal to the power incident at the nth port via the incident
1
1

an a n) less the power sent back

bn b n) via the reflected wave traveling
wave
(
(
2
2
away from the port. Relation (2.4) shows that the definition (2.1) for a and b is
consistent with the calculation of power flow in the network.
Mathematically, scattering matrix [S] expresses the relationship between as and bs at
various ports of a network through the matrix equation
b = [S]a
(2.5)
a
b
where and are n-element column vectors (n1matrices) and [S] is an nn square
matrix. For a two-port network, (2.5) may be written as
b1=S 11 a1 +S 12 a 2
(2.6a)

b2=S 21 a1 + S22 a2
(2.6b)
It may be noted that the coefficients along the main diagonal of the scattering matrix
are voltage reflection coefficients. For an n-port network,
+
Vn
V
n

(2.7)
bn
S nn=
=
an a =0(i n)
Equation (2.7) indicates that the voltage reflection coefficient S nn may be
determined by finding bn for a unit value of an when the incident wave
amplitudes at all other ports are set to zero. Off-diagonal terms of the scattering
matrix represent the voltage transmission coefficients of the network. For example,

+/ Z 0m

Vm
Z
V /
n
(2.8)

b
S nm= n
=
am a =0(i m)
Thus S nm represents the voltage transmission coefficient from port m to port n, and
may be determined by finding bn for am =1 , with incident wave amplitudes at
all other ports equal to zero. Note that the order of subscripts in S nm is opposite to
the direction of power transmission. For a network with ports impedances same,
+

/ V m .

S nm=V n
0n

2.1.2 Advantages of Scattering Matrix Representation

Apart from the physical interpretation of scarring matrix parameters in terms of
voltage reflection and transmission coefficients as discussed above, the scattering
matrix representation is convenient also because of the following considerations.
Existence of S-matrix: A scattering matrix exists for every linear, passive and timeinvariant network. This is not true of admittance or impedance matrices. We will
come across an example of this while discussing the representation of a transmission
line section later.
Convenience of measurement: One of the main advantages associated with the use of
S-parameters is the ease with which they can be measured at microwave frequencies.
At lower frequencies, measurements of Z- or Y-parameters are carried out by using
open-circuit or short-circuit terminations at some of the ports and measuring the
voltage and/or current at the port(s) associated with the particular matrix element
being measured. At microwave frequencies, an ideal open-circuit or short-circuit is
difficult to realize because of the parasitic reactance associated with such conditions.
Also, for some active circuits such as transistor amplifiers, open-circuit and short
circuit terminations may disturb the stability of the amplifier. The S-parameter
measurements on the contrary, are carried out by terminating the ports with matched
loads (usually the normalizing impedance Z 0 n ) so that the input wave amplitudes
at these ports are zero. The matched loads are easier to realize and do not in general
affect the circuit stability.
Shift of reference plane: Another important advantage of the S-parameter approach
emerges from the fact that the S-parameters are defined in terms of traveling waves;
and unlike currents and voltages, the wave magnitude does not vary along a uniform
lossless line. This feature allows the S-parameters of a network or device to be
measured relatively far away from the physical location of its ports because at times
the physical ports are not accessible. This is equivalent to shifting the reference planes
from the port locations and can be compensated for by modifying the phase of the
measured data. On the other hand, if ABCD or Z- or Y-parameters are used, they
would vary with distance away from the port terminals not only in phase but in
magnitude also.
2.1.2

Characteristics of S-matrices
6

Some of the important characteristics of S-matrices are listed below:

(i)
For a reciprocal network, the S-matrix is symmetric, that is,
(2.9)
[ S ] =[ S ] t
where the superscript t indicates the transpose of a matrix. This relationship can be
proved by starting from the fact that for reciprocal networks the Z-matrix is
symmetrical. Then use the inter-relationship between Z- and S-parameters.
(ii)
For lossless passive networks, the power conservation property leads to
two interesting characteristics of S-matrices*. Power conservation implies
that the sum of the net power flowing into or out of the network equals
zero, that is,
( an anbn bn ) =0
(2.10)
n

(a) One of the results that can be derived from the above relation is
N

|S | = S S =1
n=1

(for all i = 1,2,3N)

(2.11a)

n=1

that is, the inner product of any column of the scattering matrix with its
own conjugate equals unity i.e. the sum of reflected and transmitted power
coefficients equals unity for unit power fed to the port.
(b) Another property of S-parameters arising from power conservation is
N

Sns S nr=0

(for all s, r = 1,2,3N, sr) (2.11b)

n=1

This orthogonality constraint implies that the inner product of any column
of a scattering matrix with the complex conjugate of any other column is
zero.
Derivations of (2.11a) and (2.11b) are given in the Appendix-2A. These two relations
are useful in the verification of the correctness of S-matrices for lossless, passive
networks, and are also helpful in deriving the values of certain elements of S-matrix.
2.2 Scattering Matrices for Some 2-port Networks
Two-port networks are relatively easy to design (if the network consists of sections of
transmission lines/ waveguides connected in cascade) and constitute a major building
block of microwave networks. Analysis for the S-matrix of 2-ports is discussed next.
2.2.1 S-matrices of Symmetrical 2-port Networks
A 2-port network is symmetrical when the interchange of ports 1 and 2 does not alter
its characteristics in any manner. A simple example of such a network is a section of
uniform transmission line shown in Fig. 2.2(a). The symmetry results in the following
relationship between S-parameters:
S11 S 22 ,
S12 S 21
(2.12)
that is, only the two out of total four parameters need to be determined. Let us
consider two different modes of excitation of a symmetrical 2-port network shown in
Fig. 2.2(b) and Fig. 2.2(c). The excitation in Fig. 2.2 (b) is called in-phase excitation
because the ports are excited by sources with identical magnitude and phase. Because
of the symmetry of the network and excitations, the voltage distribution across the
network has even symmetry about the plane ss. Consequently, the gradient of voltage
at the plane of symmetry is zero and no current crosses ( I dV / dz ) this plane.
7

Thus the plane ss may be replaced by an open circuit plane without disturbing the
voltage and current distribution anywhere in the network. The open circuit condition
is equivalent to magnetic wall boundary condition at plane ss in the field theory.
* For example, see R.E. Collin, Foundations for Microwave Engineering,
McGraw Hill, 1966, pp. 174-175.
Introducing an open-circuit plane (or a magnetic wall) at ss allows us to determine
b1 e by considering the corresponding one-port network problem shown in Fig. 2.3.
For the even-voltage distribution or in-phase excitation, half-section of a transmission
line of length , S 11e (reflection coefficient at port 1) or b1 for a1=1 , may
be determined from the input impedance Z of transmission line of length /2 and
using (1.25) to calculate . We get,
Z ine = j Z 0 cot ( l/2 )
and
j Z 0 cot ( l/ 2 )Z 0 jcot ( l /2 ) 1
S 11e = e =
=
j Z 0 cot ( l /2 ) + Z 0 jcot ( l/2 )+ 1
(2.13)
The second mode of excitation shown in Fig. 2.2(c) is out-of-phase excitation since
the voltage sources are out of phase. This excitation produces a voltage distribution
that is anti-symmetric with respect to the plane ss. Therefore, the voltage along ss is
zero which can be replaced by a short circuit plane. The short circuit condition is
equivalent to electric wall boundary condition at plane ss in field theory. The half
sections of the networks corresponding to this excitation scheme are shown in Fig.
2.4. The reflection coefficient S 11o for the anti-symmetric distribution is found as
follows:
Z ino = j Z0 tan ( l/2 )
and
j Z tan ( l/2 )Z0 jtan ( l/2 )1
S 11o = o= 0
=
j Z 0 tan ( l/2 ) +Z 0 jtan ( l/2 ) +1
(2.14)
The characteristics of even-mode (in-phase) and odd-mode (out-of-phase) half
sections may be combined to yield the characteristics of the 2-port symmetrical
networks. If the excitations shown in Fig 2.2(b) and Fig. 2.2(c) are superimposed, we
obtain the input and output wave amplitude values shown in Fig. 2.5. For this
excitation, we can write S 11 for the 2-port network as
b
b +b
S +S
S 11= 1
= 1e 1o
= 11e 11o
(2.15)
a1 a =0
2
2
a =0
When we substitute the values of S 11e S11 o from (2.13) and (2.24), respectively

and simplify (see Appendix-2B), we obtain S 11=0 .

From the values of a1 and b2 shown in Fig. 2.5, the value of S 21 may be
obtained as:

b2
b +b
b b
S S
= 2e 2o
= 1e 1 o
= 11e 11o
(2.16a)
a1 a =0
2
2
2
a =0
a =0
Again substituting the values of S 11e and S 11o , and simplifying (as given in
Appendix-2B),
(2.16b)
S 21=e j l
Thus, the S-matrix of the transmission line section shown in Fig. 2.2 may be written
as
0
e j l
[S] =
(2.17)
j l
e
0
Various terms in (2.17) may be interpreted as follows. Zeros along the diagonal
indicate that the reflection coefficients at ports 1 and 2 are zero. This is obvious from
the configuration since the normalizing impedances at ports 1 and 2 are Z 0 . The
transmission coefficients (from port 1 to port 2 and vice versa) are given by e j l ,
which implies a phase delay of l radians as the signal propagates through length
l of the line.
S 21 =

10

11

The above example of a transmission line section illustrates a useful method for
determining the S-parameters of symmetrical 2-port networks. Some of the networks
whose S-parameters may be determined in a similar manner are shown in Fig. 2.6.
The concept of symmetry may also be used for simplifying the analysis of
symmetrical 4-port (in general 2n-port) networks. For a symmetrical 4-port network,
one has to analyze only two 2-port even and odd-mode half-sections. Examples of this
kind of network would be discussed later when we describe the design of branchline
and coupled line directional couplers in Chapter 8.
Evaluation of S-matrices of symmetrical networks can also be described in terms of
the solution of eigenvalue equation given below
[S] U n=s n U n
or
( [ S ] sn [ I ]) U n=0
(2.18)
where U n is the nth eigenvector and s n is the corresponding eigenvalue.
Comparing (2.18) with (2.5), it can be seen that the eigenvector U n represents a
possible excitation of the network at the terminal planes with the excitation voltages
proportional to the elements of the eigenvector. The eigenvalue s n represents a
reflection coefficient measured at any terminal plane. The elements of the S-matrix
are obtained as linear combination of eigenvalues. Details of this approach are
described in the book by Altman. [J.A. Altman, Microwave Circuits, Van Nostrand,
New York, 1964].
2.2.2 General Method for Determining S-matrices
Another approach for calculating the scattering matrices is based on finding the
relationships among wave variables by expressing them in terms of port voltages and
port currents. We will illustrate this method by considering an ideal junction of two
transmission lines of different impedances as shown in Fig. 2.7. In general, a junction
discontinuity gives rise to disturbance in rf voltage and current across the junction and
the effect of this disturbance may be expressed as a series reactance. Neglecting this
junction effect, we can write using the Kirchoffs laws:
V 1=V 2 ,
I 1 =I 2
(2.19)
From (2.2) and (2.3) we obtain
Vn
+I Z
2 an =
Z0 n n 0 n
(2.20a)
and
Vn
I Z
2 bn =
Z0 n n 0 n
(2.20b)
I n=( a nbn ) / Z 0 n
V n= Z 0n ( a n+ bn )
or
(2.21)
Thus, for the circuit in Fig. 2.7
V1
+I Z
2 a1=
Z1 1 1
Using (2.19) in (2.22) gives
V2
I Z
2 a1=
Z1 2 1

(2.22)

(2.23)

12

Also
2 b2=

V2
I Z
Z2 2 2

(2.24)

By definition
V2
I 2 Z 2
b2
Z2

S 21 =
=
a1 a =0 V 2
I Z
Z1 2 1

(2.25)

a2=0

The condition a2=0 can be ensured by terminating port 2 in matched load

Z L=Z 0 . This gives V 2=I 2 Z 2 . Substituting this in (2.25) yields
Z 2
Z 2
Z2
2 Z1 Z 2

S 21 =
=
Z 2
Z 1 +Z 2
Z 1
Z1
(2.26)
Similarly,
V1
I 1 Z 1
b1
Z1

S 21 =
=
a1 a =0 V 1
+I Z
Z 1 1 1 V =I Z

I 2 Z 2
+I 2 Z 1
Z1
Z Z

S 11=
= 2 1
I 2 Z 2
Z 2 +Z 1
I 2 Z 1
Z1

(2.27)
Thus the S-matrix of the transmission line junction of Fig. 2.7 may be written as

[S] =

Z 2Z 1
Z2 + Z 1
2 Z1 Z2
Z1 + Z 2

2 Z1 Z2
Z1 + Z 2
Z 1Z 2
Z2 + Z 1

(2.28)
In (2.28), S 22 is obtained from S 11 by interchanging Z1 and Z2.
The method used above can be extended for the circuits shown in Fig. 2.8. As we will
discuss later in Chapter 4, the configuration in Fig. 2.8(a) is an equivalent circuit for a
step discontinuity in strip lines. It may be noted that these networks are not
symmetrical, and hence the in-phase and out-of-phase excitation approach cannot be
used here. For these networks, relationships similar to (2.19) are obtained from
Kirchoffs laws. For the series impedance network of Fig. 2.8(a) we have
I 1 =I 2 , and
V 1=V 2 +Z I 1
(2.29)
whereas for the shunt admittance network of Fig. 2.8(b) we have

13

V 1=V 2 ,

and

I 1 =Y V 1 I 2

(2.30)
These equations, along with (2.20) and (2.21) yield S-matrices for these networks.
2.2.3 Analysis of Arbitrarily Connected 2-ports [CAD of Microwave Circuits*]
2-port components can be combined in series or parallel to yield a 2-port network.
Four different configurations, which are possible for such connections, are shown in
Fig. 2.9. These circuits can be analyzed by using Y-, Z-, H-, or G-matrices* also. The
Y-matrix characterization for the parallel-parallel connection of Fig. 2.9(a) can be
obtained simply by adding Y-matrices of the two components. For the series-series
connection of Fig. 2.9(b), the individual Z-matrices are added to obtain the overall Zmatrix. Similarly, for the series-parallel connection of Fig. 2.9(c) and parallel-series
connection of Fig. 2.9(d). Even though very little effort is spent in adding these
matrices, these representations have the drawback that Y-, Z-, H- or G-matrices are
computed from the S-matrices and vice-versa.

Fig. 2.10: Combinations of 2-port components

(a) A parallel-parallel connection of two 2-port components
(b) A series-series connection of two 2-port components
(c) A series-parallel connection of two 2-port components
(d) A parallel-series connection of two 2-port components
2-port components can be combined arbitrarily to yield a multiport network. Some
examples of these connections are shown in Fig. 2.10. These circuits can be analyzed
by characterizing such a connection as a dummy multiport component (indicated by
dotted lines in the Figure), and then using any of the multiport connection methods*;
one such method is described in the next section. To do so, the S-matrix of the
multiport connection/junction is required.

14

Fig. 2.10: Some examples of arbitrarily connected 2-port components to yield

multiport networks.
2.2.4 Analysis of Arbitrarily Connected Networks using Multiport Connection
Method
In this method, the S-matrix of the overall network is determined from the S-matrices
of individual components. This method is applicable when the network contains
arbitrarily interconnected multiport components without independent generators.
When one or more independent generators are present, these can be treated as existing
outside the remaining network N as shown in Fig. 2.11, and the present method yields
the S-matrix for the network N.

Fig. 2.11: An arbitrarily connected network with p-external ports and c-internal ports

15

Consider a network N of multiport components shown in Fig. 2.11. The ports in the
network are classified as internal ports or c-ports which are connected, and p-ports
which are external to it. If there are m components in the network, the governing
relations for all the components can be written together as
[b] = [S] [a]
(2.31)
Since the components are not yet connected, the associated S-parameter entries are
filled by 0s.The rows and columns in (2.31) can be re-ordered so that the wave
variables are separated into two groups; the first corresponding to the p-external ports,
and the second to c-internally connected ports. Equation (2.31) can now be written as
bp
S
S pc a p
= pp
(2.32)
bc
S cp S cc ac
where b p and a p are the wave variables at the p external ports and bc and
ac are the wave variables at the c internal ports. The interconnection constraints
for the c-internal ports can be written as
(2.33)
[ bc ]=[ ] [ ac ]
where is the connection matrix and is defined next.

[ ][

][ ]

*Wienberg, L., Scattering matrix and transfer scattering matrix, in

Amplifiers, R. F. Shea, Ed., McGraw Hill, New York, 1966.
Interconnection matrix: Let us consider cascade of two 2-port components as in
Fig. 2.12. This connection results in a pair of connected ports as shown there. Across
the junction, the outgoing wave variable at one port must equal the incoming wave
variable at the other port provided the wave variables at the two connected ports are
similarly normalized. Applying this argument to the pair of connected ports of Fig.
2.12, we have

2 X

Fig. 2.12: A cascade connection of two 2-port components showing the

dummy 2-port model for interconnection constraints.
b2x =a1y and
b1y =a2x
(2.34)
or

[ ] [ ][ ]
b2x
ax2
0
1
=
1 0 a1y
b1y

(2.35)

The matrix on the right hand side of (2.35) may be defined as the connection matrix ,
i.e.
bc = ac
(2.36)
The resultant network is supposed to have a different S-matrix than that of either
component. The interconnection constraints are expected to play an important role in
16

this respect; and we may model the interconnection constraints as a dummy 2-port as
shown Fig. 2.12. The S-matrix description for this dummy 2-port may be written as
x
a2x
I b2
=
[S
]
or ac =S I bc
(2.37)
y
y
a1
b1
The LHS in (2.37) is the outgoing wave vector on dummy 2-port and the RHS is the
incident wave vector. Comparing (2.37) and (2.36) we may write that S-matrix for the
interconnection is the inverse of connection matrix, i.e.
1
I
1
0
1
(2.38)
S = =
1 0
If the normalizing impedances for the S-matrices across the interconnection are not
same, say Z 1 and Z 2 for components X and Y, respectively, the interconnection
matrix in (2.36) is given by the inverse of S-matrix of (2.28).

[] []

[ ]

The concept of interconnection matrix can be extended to more than one junction,
e.g. Fig. 2.11. For this, the relations similar to (2.36) are written together for all the
interconnected ports in the network. The interconnection matrix now describes the
topology of the network. In each row of , all elements are zero except an
entry 1 in the column indicating the interconnection. If the element
(j, k) of is 1, it implies that the port j is connected to the port k.
The size of is equal to the number of connected ports. The interconnection matrix
is therefore highly sparse, and the zero, nonzero pattern depends on the topology and
is independent of the component characterizations.
The interconnection matrix for the network N of Fig. 2.11 may be written as
b
a
(2.39)
[ ] [ c ]
[ c]=

From (2.39) and lower half of (2.32) we get

ac = S cp a p + S cc a c
or
ac = ( S cc )1 S cp a p
(2.40)
Substituting this in upper half of (2.32), we get
b p = { S pp + S pc ( S cc )1 Scp } a p
(2.41)
or, the network scattering matrix S p is given by
S p = S pp +S pc ( S cc )1 S cp
(2.42)
After having obtained the S-matrix of a multiport network, the wave variables at the
internal ports for an arbitrary excitation at the p-external ports a p may be obtained
from (2.39) and (2.40).
An example
In the example of a simple circuit shown in Fig. 2.13, the circuit contains a 2-port
component A characterized by S A ; a 3-port component B characterized by S B ;
and a 1-port component C characterized by S C . The ports are numbered as shown
in the figure; local and global port numbers are used for components and assembled

17

circuit, respectively. Characterizations for the three components are put together as in
(2.43). The vectors a and b employ global port numbers whereas the component Smatrices employ local port numbers.

[][

A
A
S11
0 S 12
b1
0 S B22 0
b4
A
A
b2
S22
= S21 0
b6
0 0
0
B
b5
0 S 32
0
B
b3
0
0 S 12

0
0
0
sC
0
0

][ ]

0
0 a1
B
S 23 S B21 a4
0 0 a2
0 0 a6
B
S 33 S B31 a5
S B S B a3
13

11

(2.43)

The interconnection matrix defines the topology of the assembled network and can
be written as

[ ] [ ][ ]
b2
0 0
b6
0 0
=
0 1
b5
1 0
b3

0
1
0
0

1
0
0
0

a2
a6
a5
a3

(2.44)

Using (2.43), (2.44) in (2.42), the S-matrix of the composite network is given by

S p=

A
11

S
0

][

A
12

0
S
+
B
0
S22

0 0
B
0 S23

A
S22
0
0
C
1
0
s
0
B
B
S 21
S 33
0 1
1 0
S B13

][ ]
1

1
0
S B31
S B11

A
S21
0
0
0

0
0

SB32
SB12

(2.45)

Fig. 2.13: The network considered in the example.

2.3 Applications of S-Parameters
The S-matrix of a network may be employed to determine (i) the wave variables at
various ports when the network is subjected to an excitation at one or many external
ports. One of the applications of S-matrix is to determine the (ii) new S-matrix when
one or more of the reference planes of the network are shifted. This may be required
for measurements or analysis of network. Another common application is (iii) to
determine the S-matrix of loaded network e.g. for input impedance calculation. These
types of applications are discussed next.

18

2.3.1 Shift in Reference Planes

S-parameters relate magnitude and phase of the travelling waves incident on and
reflected/transmitted from microwave networks. Therefore, reference planes for phase
determination of S-parameters must be specified for each port of the network. Any
change in any of the reference planes gives rise to a change in the whole of the Smatrix of the network. However, this change in S-matrix can be determined
analytically and the S-matrix for the new port planes can be obtained from the Smatrix for the old port planes.
Consider a 2-port microwave network of Fig. 2.14 where the original reference port
planes are located at z 1 = 0 and z 2 = 0, for the 2-port. Then [b] = [S][a], where
[S] is the scattering matrix of the network for this set of reference planes.
z2

z1
1

l1

l2

Fig. 2.14: Shift of port planes of a network

Consider, now a new set of port planes at z 1=l 1 and z 2=l 2 on the
transmission lines connected to the ports. Let the new S-matrix for the same network
be defined as [ b' ] =[ S' ] [ a' ] . Then, in terms of the incident and reflected port
voltages we have

(2.41a)
+

'
V

(2.41b)
'+
V

where the unprimed quantities refer to the original port planes at z 1 = 0 and
0, and the primed quantities are referred to the new planes at z 1=l 1 and
z 2=l 2 .

z2 =

From the theory of traveling waves on lossless lines, the incident wave amplitudes are
related as
j
'+ e

+=V n

Vn
or
n

19

+
V

(2.42)
' +

where n= l n . Similarly, the reflected wave amplitudes are related as

j n

jn

=V n

V 'n

or

' e

=V n

Vn

'
V

(2.43)

Substituting for

+
V and V in (2.41a) gives

'
V

' +
V
e j 0

0 e j
1

or

'

(2.44)
'+

Comparing this expression with (2.41b) gives

j
0
e j
0
[ S ' ]= e
[S
]
j
j
0
e
0 e
j 2
S 11 e

j( 1+2 )

j2
[ j ( 1+2 ) S12 e S21 e S 22 e
]

] [

(2.45)

2.3.2 S-Matrix of a Loaded Network

One of the most common uses of S-parameters is to determine the S-parameters of
reduced network when the network is loaded at one or many ports, e.g. input
impedance of a transmission line section when loaded at the other port. The effect of
port loadings on the S-matrix can be included by the constraint between a and b

20

vectors imposed by the load/loads. The reduced S-matrix may be obtained by

including the effect of load as a = L b.
2.4
Signal-Flow Graph
The signal-flow graph is a convenient technique to graphically represent and analyze
the transmission and reflection of waves in a microwave component. It permits the
derivation of expressions such as power gains, voltage gains of complex microwave
amplifiers, reflection coefficient, transmission coefficient, etc.
2.4.1 Rules for Constructing Signal-flow Graphs
1. A flow graph has two nodes for each port, one for the incident wave and the
other for the reflected wave. The incident waves are independent wave
amplitude nodes and the reflected waves are dependent amplitude nodes.
2. The S-parameters and reflection coefficients are represented by branches.
3. Branches enter dependent amplitude nodes and emanate from independent
amplitude nodes.
4. A dependent node is equal to the sum of the products of the branches
terminating at the node with the independent nodes they emanate from.
5. Flow graph of cascaded networks is the cascade of flow graphs of individual
networks comprising it.
Example 1: b1 S11a1 S12 a 2
Here b1 is the dependent node, a1 and a 2 are independent nodes, and S11 and S12 are
represented by the appropriate branches. The signal flow graph is shown in Fig. 2.9.
Example 2: b2 S 21a1 S 22 a 2
Proceeding in the same way as for example 1 we obtain the signal flow graph is
shown in Fig. 2.10.
Example 3: Two-port network
The signal flow graphs of Fig. 2.9 and Fig. 2.10 can be combined to yield the signal
flow graph of a two-port network. This is shown in Fig. 2.11.
Example 4: Voltage-source with impedance Z S
The voltage-source with an interconnecting line is shown in Fig. 2.12(a). From here,
the voltage at the terminal AA can be written as
V g VS I g Z S
(2.31)
In terms of traveling waves on the connecting line,
V g V g V g
(2.32)
and
V g V g
Ig

Z0 Z0
(2.33)
Then (2.31) becomes
V g V g

Z
V g V g VS

Z0 Z0 S

(2.34)
21

Substitute for

If we denote
V g
bg
Z0

Vg

Ig

and
in (2.31) and solving for

Vg
V g Z S Z 0
VS Z 0

Z0 ZS Z0
Z0 ZS Z0
V g

ag

bs

VS Z 0

Z0
Z S Z 0 , and
,
,
equation (2.35) becomes
bg bs s a g

V g

, we obtain
(2.35)

ZS Z0
ZS Z0

(2.36)

(2.37)
Equation (2.37) can also be obtained by using multiple reflections procedure*. This
b
equation suggests that the incident wave g is not equal to bs unless the source is
a 0 s 0
non-reflecting, which would give g
or
(that is, a matched source with
ZS Z0
). The signal flow graph for bg bs s a g is shown in Fig. 2.12(b).

Example 3: Load impedance terminating a line

For the load impedance Z L terminating a transmission line of characteristic
impedance Z 0 , as shown in Fig. 2.13(a), we can write
*G.D. Vendlin, Design of Amplifiers and Oscillators by the S-parameter Method,
John Wiley, New York, 1982.

VL Z L I L

(2.38)

In terms of traveling waves on the feeding line, we can write

V L V L V L

(2.39)

and

V L V L
IL

Z0 Z0

(2.40)

Substituting for VL and I L gives

V L VL

VL V L Z L

Z
Z
0
0
or
Z Z 0 VL
VL
L
Z0 ZL Z0 Z0

(2.41)

(2.42)

Writing

aL
we obtain

VL
Z0

bL
,

VL
Z0

and

bL a L L

Z L Z0
ZL Z0

(2.43)
(2.44)

The signal flow graph for (2.44) is shown in Fig. 2.13(b).

22

We can combine the signal flow graph for the two-port network, with the signal flow
graphs of signal generator and the load to yield the signal flow graph of a typical
b a
microwave amplifier. This is shown in Fig. 2.14. Here g , g , bL and a L are replaced
by a1 , b1 , a 2 and b2 , respectively. S11 , etc. are the S-parameters of the transistor
amplifier network.
2.4.2 Simplification of Signal-flow Graphs
The flow graph representation can be used to determine the ratio or transfer function
of a dependent to an independent variable. For this we can use Masons non-touching
loop rule*. An alternative procedure, that is more easy to remember, is the flow graph
simplification technique proposed by Kuhn**. The simplification can be achieved
using four rules that are stated as follows:
Rule 1 (Node elimination): The transfer function of two similarly directed branches in
series is the product of the two as shown in Fig. 2.15. This eliminates the common
node between the branches.

* S.J. Mason, Feedback theory-further properties of signal flow graphs, Proc. IRE,
vol. 44, July 1956, pp. 920-926.
** N. Kuhn, Simplified signal flow graphs analysis, Microwave J., vol. 6, Nov.
1963, pp. 59-66.
Rule 2 (Branch elimination): The transfer function of two similarly directed branches
in parallel is the sum of the two as shown in Fig. 2.16. This step eliminates a branch.
Rule 3 (Loop elimination): When a node possesses a self-loop with transfer function
S, the self-loop can be eliminated by dividing the transfer function of all the input
branches by (1-S) as shown in Fig. 2.17
Rule 4 (Node duplication): The transfer function remains unchanged if a node with an
input branch and N output branches is split into two nodes. The input branch goes to
each of the new nodes. There will be K<N output branches leaving one node and the
remaining (N-K) branches leaving the second node. The dual graph with one output
branch and several input branches also holds. This is shown in Fig. 2.18.
The rules, given above, can be proved very easily as done by Adam*. Simplifying
flow graphs in this manner is particularly helpful in finding the transfer function of a
circuit, e.g. gain of a transistor amplifier. Some of the applications of the flow graph
are described next.
2.4.3

Applications of Signal-flow Graphs

Example 1: Input reflection coefficient

Figure 2.19 shows the flow graph for a load connected at the output of a two-port
network. The input reflection coefficient for this network is defined as
b
in 1
a1
(2.45)

23

To determine this ratio we can simplify the flow graph as shown in Fig. 2.19. From
the final flow graph we obtain using Rule 1
S S
b1 S11a1 12 21 L a1
1 S 22 L
(2.46)
or

in

b1
S S
S11 12 21 L
a1
1 S 22 L

(2.47)

Example 2: Output reflection coefficient

Consider a two-port network with a load connected at the input as shown in Fig. 2.20.
The corresponding signal flow graph is also shown there, which can be simplified by
splitting node a1 to obtain

out

S S
b2
S 22 12 21 s
a2
1 S11s

(2.48)

* S.F. Adam, microwave Theory and Applications, Prentice Hall, New Jersey, 1969.

Example 3: Available power from a generator

The power available from a generator or a source is defined as the power delivered by
the generator to a conjugately matched load. The signal flow graph of a generator
*
connected to a conjugately matched load, i.e. L s , is shown in Fig. 2.21. The
power available from the generator (or the power delivered to the load through the
terminals AA) is given by
2

PAVS bg
Also, because

ag

bg bs s a g

(2.49)

and

ag b

*
s g

bg
ag

Substitution in

PAVS

PAVS

(Rule 1)

(2.50)

bs
1 s

(2.51)

bs

*
s

1 s

(2.52)

gives
2
2
bs 1 s

2 2
1 s

or
PAVS

bs

1 s

(2.53)

24

Substituting for bs and s from (2.36) in (2.55) gives

Vs2 Z 0
PAVS
2

Z s Z 0
2
Zs Z0 1

Zs Z0

or
Vs2 Z 0
Vs2
PAVS

2
2
4 Re( Z s )
Zs Z0 Zs Z0

(2.54)

where Vs represents the rms voltage of the signal generator.

Example 4: Power delivered to a load L
The power delivered by the generator to the load is given by the difference between
the incident and reflected powers, that is,
2
2
PL a L bL
or
2

PL a L 1 L

(2.55)

a bg
From the flow graph of Fig. 2.22, L
and
bs
bs
bg
aL
1 s L
1 s L

(2.56)

Therefore,
PL

bs

1 s L

1
2

(2.57)

For conjugately matched load L , the above relation reduces to (2.53) that is
power delivered to the load is equal to the power available from the source.
*
s

Example 5: Transfer function of a two-port network with signal generator at the input
and a load at the outport port
Signal flow graph for this problem is shown in Fig. 2.23. It will be simplified to
determine the transfer function T defined as T b2 / bs . Various steps in the analysis
are shown in Fig. 2.23. One obtains
b2
S 21

bs 1 S11s 1 S 22 L S12 S 21s L

(2.58)

2.5
Analysis of Cascaded Two-ports (ABCD Parameters)
In several microwave circuits, two-port components are connected in cascade, that is,
port 2 of one component is connected to the port 1 of next component (and so on), as
shown in Fig. 2.24. If we consider a cascade combination of only two components A
and B, S-matrix of the combination can be written in terms of the S-matrices of the
components A and B. Let us write

25

 S SS
A

S12A

S 22A

A
11
A
21

S SS
B

and

B
11
B
21

S12B

S 22B

(2.59)

We have
b1A S11A a1A S12A a 2A

(2.60a)

b S a S a

A
2

(2.60b)

b1B S11B a1B S12B a 2B

(2.61a)

b S a S a

(2.61b)

A
2

A A
21 1

A
22

and
B
2

B B
21 1

B
22

B
2

At the common port

b2A a1B ,

a 2A b1B
and
(2.61)
The S-matrix of the combination circuit can be obtained by eliminating these common
A
B
A
B
wave amplitudes, that is, b2 , a1 , a 2 and b1 from (2.60) and (2.61). We obtain

S12A S11B S 21A

1 S 22A S11B
B
S 21A S 21
1 S 22A S11B

A
S11

b1A
B
b2

S12A S12B

A
A B
1 S 22 S11 a1
S B S A S B aB
B
S 22
21 22A 12B 2
1 S 22 S11

(2.62)
This relation can be used repeatedly to obtain S-matrix of a cascade of more than two
components.
The analysis of cascaded two-ports can be simplified, compared to (2.62), by using
another representation for two-port components. This is known as ABCD or A-matrix
and is defined as follows in terms of voltages and currents at the ports of the
component.
V1 A B V2
I C D I

2
1
(2.63)
I
I
The directions of 1 and 2 are as shown in Figs. 2.7 and 2.8, and explains the negative
sign with I 2 in (2.63). When two components x and y are cascaded as in Fig. 2.25 the
interconnection relationships may be written as
V1y V2x ,
I 1y I 2x
and
(2.64)
From (2.63) and (2.64) we can write
V1x A x B x V2x A x B x V1 y
x x

D x I 2x C x D x I 1y
I1 C
A x B x A y B y V2y
x

D x C y D y Iy
C

(2.65)
x
Thus ABCD-matrix of the combination xy is given by the product of [ ABCD ] and
[ ABCD y ] or

A xy
xy
C

B xy A x

D xy C x

Bx Ay

Dx C y

By

Dy

(2.66)

26

The relationship (2.68) is very convenient from the computational point of view since
it involves only multiplication of the matrices of individual components. This
property makes ABCD-matrix representation very attractive for cascade connection of
large number of two-port components.
The ABCD-matrix for a transmission line section may be derived from (1.19) and
(1.20), which relate the voltage and current at one end of the line to those at the other
end. We have*
Z 0 sinh V2
V1 A B V2 cosh

I C D I sinh / Z
cosh I 2

2
0
1

(2.67)

0
,

cosh(
j

cos(

)
For a lossless line
and since
, and
sinh( j) j sin( ) , we have

cos

[ ABCD ]

jZ 0 sin
cos

j sin / Z 0

(2.68)

------------------------* It may be noted that the direction of I 2 used for defining [ABCD] is opposite to that
used in (1.19) and (1.20).
ABCD-matrices for other networks may be obtained by deriving relationships
between voltages and currents at the two ports. Alternatively, if the S-matrix for a
given network is known, ABCD-matrix of the network may be derived there-from by
using the following equivalences:
A 1 S11 S 22 S Z 01 / Z 02 / 2 S 21
(2.69a)
B 1 S11 S 22 S Z 01Z 02 / 2S 21
(2.69b)
C 1 S11 S 22 S / 2 S 21 Z 01 Z 02
(2.69c)
D 1 S11 S 22 S Z 02 / Z 01 / 2S 21
(2.69d)
where S S11S 22 S12 S 21 and Z 01 and Z 02 are the impedances of the feed lines

connected to the network. Alternatively, if the ABCD-matrix is known, S-matrix may

be derived using the following formulas*:
S11 AZ 02 B CZ 01 Z 02 DZ 01 / A
(2.70a)
S12 2 AD BC Z 01 Z 02 / A

(2.70b)

S 21 2 Z 01 Z 02 / A

(2.70c)

S 22 AZ 02 B CZ 01Z 02 DZ 01 / A

where

(2.70d)

A AZ 02 B CZ 01Z 02 DZ 01

Physical interpretation of ABCD or A-parameters

The ABCD-parameters may be interpreted physically by noting that

27

V1
V2

B
I 2 0

V1
I2

I1
V2

I1
I2

V2 0
I 2 0
V2 0
,
,
, and
(2.71)
We note that A is a dimensionless parameter, being the ratio of the input voltage at
port 1 to the voltage appearing at port 2 when port 2 is open circuited. D is likewise a
dimensionless parameter, being the ratio of current flowing in port 1 to the current
flowing out of port 2 when it is short circuited. Parameter B has the dimension of
impedance, and is the ratio of voltage at port 1 to the current flowing out of port 2
when it is short circuited. Finally, C is an admittance given by the ratio of the current
flowing into port 1 to the voltage appearing at port 2 when it is open circuited.

Some important properties of ABCD-matrices may be summarized as follows:

i)
For reciprocal networks
AD BC = 1
(2.72)
Unlike [S], [Z] and [Y] matrices, ABCD-matrix need not be symmetric for reciprocal
networks.
ii)
For symmetrical networks (which remain unaltered when the two ports are
interchanged,
A=D
(2.73)
--------------------------* K.C. Gupta et. al., CAD of Microwave Circuits, Artech House, Mass., 1981, p.34.
2.6
Immittance Matrices
Impedance and admittance matrices are also useful at microwave frequencies.
Immittance description of transmission line components and discontinuities may be
used to derive their lumped equivalent circuits. Also, impedance matrix description is
convenient for two-dimensional (planar) components because of the availability of
impedance Greens functions for such components.
The impedance matrix Z for an n-port component is defined as
V1 Z 11 Z 12 Z 1n I 1
V Z

2 21 Z 22 Z 2 n I 2




Vn Z n1 Z n 2 Z nn I n
(2.74)
The directions of currents are as shown in Fig. 2.7. For a two=port network,
V1 Z11 I 1 Z 12 I 2
(2.75a)

V2 Z 21 I 1 Z 22 I 2

The entries in Z-matrix are the open circuit impedances, since

V
V
V
V
Z 11 1
Z 12 1
Z 21 2
Z 22 2
I 1 I 0
I 2 I 0
I 1 I 0
I2
2
1
2
,
,
, and
For a reciprocal network,
Z 12 Z 21
[ Z ] [ Z ]t

(2.75b)

I1 0

(2.76)
(2.77)

and for a symmetrical two-port (where ports 1 and 2 are interchangeable)

Z 11 Z 22
(2.78)

28

The following relationships holds between Z-matrix and ABCD-matrix for a two-port
network (See Appendix-2C)
1 A AD BC
[Z ]

D
C 1

(2.79)
For a transmission line section, the ABCD-matrix is given in (2.68). Using (2.79), the
Z-matrix may be written as
cot cos ec
[ Z ] jZ 0

cos ec cot
(2.80)
When / 4, / 2 ,
0
[Z ]
jZ 0

jZ 0
0

(2.81)

/4

But when / 2, , cot and 1 / sin tend to infinity, with the result that
Z-matrix cannot be written for a / 2 line length. This illustrates the point that Zmatrix might not exist for every circuit. On the other hand, S-matrix exists for every
linear, passive, time invariant network.
One of the advantages of Z-matrix representation is that once the Z-matrix is known,
a lumped equivalent circuit for the network can be obtained. For a symmetrical,
reciprocal two-port the equivalent lumped network is shown in Fig. 2.26(a); when the
two-port is not symmetrical, the equivalent circuit gets modified to that shown in Fig.
2.26(b). These equivalent circuits may be verified by the equivalent Z-parameters
using (2.75) and (2.76).
Based on this approach we can write an equivalent lumped T-network for a given
transmission line of length , with Z-parameters given in (2.79). We get
Z 1 Z 2 Z 11 Z 12 jZ 0 cot jZ 0 / sin
jZ 0 cot 1 / sin jZ 0

and

1 cos

jZ 0 tan

sin
2

Z 3 Z 12 jZ 0 / sin

(2.82)
(2.83)

When , the series elements are purely inductive with equivalent L1 given by
Z

L1 0 tan

2
(2.84a)
and shunt capacitance
C 3 sin( ) /(Z 0 )
(2.84b)
The equivalent circuit is shown in Fig. 2.27. When the line length is so small that
1 , (2.84) can be approximated as

L1

Z 0
2 ,

C3

Z 0

(2.85)

A further approximation can be made for low values of Z 0 . In this case, the
equivalent circuit can be approximated by a shunt capacitor only since the series
inductors are negligibly small.
29

Admittance Matrix Representation

Y-matrix is inverse of the Z-matrix, and is defined as
1
I1
Z 11 Z 12 Z 1n V1
Y11 Y12 Y1n V1
I
Z

2 21 Z 22 Z 2 n V2 Y21 Y22 Y2 n V2

In
Z n1 Z n 2 Z nn Vn
Yn1 Yn 2 Ynn Vn
For a two-port network,
I 1 Y11V1 Y12V2

I 2 Y21V1 Y22V2

(2.86)
(2.87a)
(2.87b)

Y-matrix and ABCD-matrix are related by (see Appendix-2D)

1 D AD BC
[Y ]

A
B 1

(2.88)
The Y-matrix for a transmission line section may be derived from (2.88). We obtain
cos ec
cot
[Y ] jY0
cot
cos ec
(2.89)
When / 4, / 2 ,
jY0
0
[Y ]
0 / 4
jY0
(2.90)
Like the Z-matrix, the Y-matrix also becomes indeterminate for a half-wave
transmission line section.

Once the Y-matrix of a network is known, a section equivalent network may be

drawn as shown in Fig. 2.28. The admittances Y1 ,Y2 and Y3 are given by

Y1 Y11 Y12 , Y2 Y22 Y12 , Y3 Y12

(2.91)

For a transmission line section, the lumped element equivalent -network consists of
Y1 Y2 jY0 cot jY0 / sin
1 cos

jY0 cot 1 / sin jY0
jY0 tan

sin
2 (2.92)
and
Y3 Y12 jY0 / sin
(2.93)
The lumped elements corresponding to a line length with 1 are shunt
capacitances and a series inductor with the values given by

Y
L3
C1 0
Y0
2 ,
(2.94)

The equivalent circuit is shown in Fig. 2.29. It may be observed from (2.85) and
(2.94) that these equations are dual of each other with L C and Y0 Z 0 since the
- and T-networks are also dual of each other.
When the characteristic impedance of the transmission line section is very high, the
equivalent circuit can be approximated by a series inductor only.
30

The lumped element equivalent networks obtained from the Y-matrix or Z-matrix
representations are strictly true over a very narrow band of frequencies. Otherwise,
the values of lumped elements vary with frequency.
2.6.1 Conversion between S-, Z- and Y-matrices
For some circuit configurations it is easier to derive impedance or admittance
matrices. These matrices can be converted into S-matrix by using the relationship
(2.20) and (2.21) between the wave amplitudes and the voltages and currents. The
general method of determining the S-matrices is based on this approach.
We can write, for the normalized wave amplitudes
a 12 (V I )
and

b 12 (V I )

(2.95)
(2.96)

S-matrix from Z- or Y-matrix:

For this conversion replace b by [S ]a and V by [Z ]I in the above equations to obtain
[ S ]a 12 ([ Z ] [ I ])I
(2.97)
Using (2.95) in (2.97) gives
[ S ] [ Z ] [ I ] I ([ Z ] [ I ])I
implying

[ S ] [ Z ] [ I ] ([ Z ] [ I ])

Post multiply (2.99) by [ Z ] [ I ] to yield

1
[ S ] ([ Z ] [ I ]) [ Z ] [ I ]

(2.98)
(2.99)

(2.100)

By virtue of [Y ] [ Z ] , we obtain from above

1
[ S ] ([ I ] [Y ]) [ I ] [Y ]

(2.101)
Equations (2.100) and (2.101) provide conversion from Z- and Y-matrices to Smatrix.
Z- or Y-matrix from S-matrix:
From (2.95) and (2.96) we can write
V ab
and
I ab
Again replace b by [S ]a in (2.102) to obtain
and

(2.102a)
(2.102b)

V [ I ] [ S ] a

(2.103)

I [ I ] [ S ] a

(2.104)

Now replace V [Z ]I in (2.103) and use (2.104) giving

[ Z ] [ I ] [ S ] a [ I ] [ S ] a

(2.105)

implying

31

[ Z ] [ I ] [ S ] [ I ] [ S ]
Post multiply (2.106) by [ I ] [ S ] to give
1
[ Z ] ([ I ] [ S ]) [ I ] [ S ]

(2.106)

or
or

[Y ] 1 ([ I ] [ S ]) [ I ] [ S ]
[Y ] ([ I ] [ S ]) [ I ] [ S ]

(2.107)

(2.108)
Equations (2.107) and (2.108) can be used to convert S-matrix to Z- and Y-matrices,
respectively.

32

Appendix-2A: Derivation of the S-matrix Relations (2.11) for Lossless Networks

The power conservation condition (2.10) may be expressed as
N

n 1

n 1

bn bn* a n a n*

(2A.1)

From (2.2)
N

bn S ni ai

(2A.2)

i 1

The power conservation condition (2A.1) may now be written as

N

S ni ai

n 1 i 1

a n a n*
n 1

(2A.3)

Since a n s are independent, we may choose all a n s except ai to be zero. This leads to
2

S ni ai

a i ai*

n 1

or
2

ni

n 1

S ni S ni* 1

(2A.4)

n 1

Since the index i is arbitrary, the above equation is valid for all values of i.
The other set of constraints (2.11b) may be derived from (2A.3). By selecting all a n s
except a s and a r to be zero ( r s ), the relation (2A.3) may be written as
2

ns

S ns a s S nr a r S ns a s S nr a r a s a r

a s S nr a r

(2A.5)
Expanding the left side, and using (2A.4) to equate two of the terms to zero gives
n 1

n 1

S
N

n 1

ns

*
S nr* a s a r* S ns
S nr a s* a r 0

In view of the independent nature of

as ar
. This leads to
as

S
N

n 1

ns

as

and a r , we may choose in the first instance,

*
S nr* S ns
S nr 0

If, instead we choose a s jar with a r real, we obtain

j ar

S
N

n 1

ns

(2A.6)

S nr* S ns* S nr 0

(2A.7)

(2A.8)

Since neither a s nor a r is zero, both (2A.7) and (2A.8) are satisfied only if
N

ns

S nr* 0

sr
which is the orthogonality condition stated in (2.11).
n 1

(2A.9)

33

Appendix- 2B: Derivation of (2.17)

Let x , then from (2.13) and (2.14)

j cot x / 2 1 j tan x / 2 1
S11e S11o

j cot x / 2 1 j tan x / 2 1
1 j cot x / 2 j tan x / 2 1 1 j cot x / 2 j tan x / 2 1

0
2 j tan x / 2 cot x / 2
Let x/2 = y, then
j cot y 1 j tan y 1
S11e S11o

j cot y 1 j tan y 1
1 j cot y j tan y 1 1 j cot y j tan y 1 2 j cot y tan y

2 j tan y cot y
2 j tan y cot y (2B.1)
Since,
tan( x / 2) (1 cos x) / sin x and
cot( x / 2) (1 cos x) / sin x
tan( x / 2) cot( x / 2) 2 cos x / sin x
tan( x / 2) cot( x / 2) 2 / sin x
Therefore, (2B.1) can be expressed as
2 j 2 / sin x
2j
2
S11e S11o

jx 2e jx
2 j 2 cos x / sin x sin x j cos x e
or
S11e S11o
e j
2

34

Appendix-2C: Derivation of (2.79) and (2.88)

From the definition of ABCD-matrix
V1 AV 2 BI 2

I 1 CV2 DI 2
Equation (2C.2) can be written as
CV2 I 1 DI 2
or
V2 I 1 DI 2 / C

(2C.1)
(2C.2)

(2C.3)

Substituting for V2 in (2C.1) one obtains

A
AD

I1
B I2
C
C

1
D
V2 I 1 I 2
C
C
Equations (2C.4) and (2C.5) yield the impedance matrix as
1 A AD BC
[Z ]

D
C 1

V1

Also,

I 2 V1 / B ( A / B)V2

Substituting for I 2 in (2C.2) one obtains

A
D
AD
1

I 1 CV2 D
V1 V2 V1 C
V2
B
B
B
B

Equations (2C.7) and (2C.8) yield

1 D BC AD
[Y ]

A
B 1

(2C.4)
(2C.5)

(2C.6)
(2C.7)

(2C.8)

(2C.9)

35

Appendix-2D: ABCD, and S and T-Matrices for some commonly used two-ports

36

37

Important Results
1.

2.
3.

For a multi-port network, the wave amplitudes a n and bn are defined as

a n Vn / Z 0 n
bn Vn / Z 0 n
,
where Z 0 n represent normalizing impedances at various ports.
Scattering matrix [S] for an n-port network is defined by b [S ]a .
For the lossless, passive networks the following relationships exist
N

S
n 1

ni

S ni* 1

ns

S nr* 0

for all i = 1,2,3,N.

for all s, r = 1,2,N; s r .

For two-port symmetrical networks S11 S 22 , and S-parameters may be
n 1

4.

,
,

found from S11e and S11o for the even and odd-mode half-sections,
respectively.
S11 S 22 12 S11e S11o
S12 S 21 12 S11e S11o
,
5.

In general, S-parameters may be found from the voltages and currents at

the various ports by using
Vn

Vn

a n 12
I n Z 0n
bn 12
I n Z 0n
Z

0n
0n

, and

6.

ABCD-matrix representation is convenient for cascaded two-ports and is

defined as
V1 A B V2
I C D I

2
1

38

7.

Some general properties of ABCD-matrix are:

[ A] [ A]1 [ A] 2 ...[ A] N
i) For cascaded two-port networks
ii) For reciprocal networks

8.

A Ad BC 1

Z- and Y-parameters may be used for writing T- and -lumped equivalent

circuits for a network, as shown below

Z 11 Z 12

Z 22 Z12
Z12 Z 21

Y12 Y21
Y11 Y12
Y12 Y22

39

Problems for Chapter 2

2.1 A lossless transmission line of characteristic impedance Z and electrical length
is terminated at both ends by transmission lines of characteristic
impedance Z 0 . Derive the expressions for S-matrix for this two-port.
2.2 Consider a lumped impedance Z in series with a transmission line of
characteristic impedance Z 0 . The line extends on both the sides as shown in
Fig. 2.6(b). Evaluate two-port S-matrix of this configuration.
2.3 Consider a shunt connected admittance Y as shown in Fig. 2.6(c). Calculate
two-port scattering matrix for this configuration.
2.4 Find the scattering matrix for the symmetrical T-network shown in Fig. 2.6(f).
2.5 Find the scattering matrix for the symmetrical -network shown in Fig.
2.6(g).
2.6 Calculate the S-matrix for a series impedance Z connected between two lines
of characteristic impedances Z 1 and Z 2 , respectively as shown in Fig. 2.8(a).
2.7 Find S-matrix for a shunt admittance Y connected as shown in Fig. 2.8(b).
2.8 Use the S-matrix to A-matrix transformation to find A-matrices for series
impedance connected as in Prob. 2.2, and also when connected as in Prob. 2.6.
Compare and comment on the two results obtained.
2.9 Calculate the A-matrix for an admittance Y connected in shunt between two
lines of characteristic admittance Y0 as shown in Fig. 2.6(c).
2.10

A transmission cavity can be represented by a section of transmission line

(characteristic admittance Y0 , length ) shunted by susceptance jB on either end as
shown in Fig. 2.6(d). Find admittance parameters and derive an equivalent circuit
therefrom.
A lossless transmission line of length 1 (characteristic impedance Z 0 and
phase constant ) is connected at the input terminals of a two-port network with S2.11

parameters S11 , S12 , S 21 , S 22 . Find the S-matrix for the composite network. If another
line of length 2 is connected at port 2, what will be the resulting S-matrix?
2.12 Show that a symmetrical 3-port lossless, reciprocal transmission line junction
cannot be matched at all the three ports. The properties of S-matrix for a lossless
network may be used.

40

2.13 Consider a 3-port network consisting of a 3-db power divider matched at the
input port (say port 1). The reference planes at ports 2 and 3 are chosen such that S 21
and S 31 are real. Use the properties of S-matrix for lossless reciprocal networks to
show that

S 22 S 33 S 23 1 / 2

A two-port network with S-parameters S11 , S12 , S 21 , S 22 is terminated with an

impedance Z L at port 2. Calculate the input reflection coefficient at port 1.

2.14

2.15

Can the measurement of input reflection coefficient in Prob.2.14 be used for

the measurement of S-parameters by varying Z L ? How many known values of

Z L will be required? Can all S-parameters be determined completely by this

method?
Consider a shunt connected tee-junction of three different impedances Z 1 , Z 2
and Z 3 as shown in Fig. 2.30. Derive the S-matrix (ignoring parasitic reactance).

2.16

2.17 Consider a junction of three series-connected transmission lines of characteristic

impedances Z 1 , Z 2 and Z 3 , respectively, as shown in Fig. 2.31. Derive the S-matrix
(parasitic reactance may be ignored).
Four transmission lines of different characteristic admittances Z L Z 0 , Y3 and

2.18

Y4 are connected in series at a junction J as shown in Fig. 2.32.. Derive the Smatrix of the junction assuming that the parasitic reactance at the junction may be
ignored.
2.19 Four transmission lines of different characteristic impedances Z 1 , Z 2 , Z 3 and Z 4
are connected in series at a junction J as shown in Fig. 2.33.. Find the S-matrix of the
junction when the parasitic reactances are ignored.
For a three-port network, S-matrix is expressed in terms of S11 , S12 , S13 etc.
The port 3 is terminated with Z L . Find the S-matrix of the resulting two-port

2.20

network when
(a) Z L Z 0 , where Z 0 is the reference characteristic impedance at the ports.
(b) Z L 0 , and
(c) Z L

(a) Show that for a series connected impedance of Fig. 2.6(b), S 21 1 S11 .

2.21

(b) Show that for a shunt connected admittance of Fig. 2.6(c), S 21 1 S11 .

41

2.22

The circuit configuration shown in Fig. 2.34 is used to measure the Sparameters of a connector when two identical connectors are available. Here, is
the uniform transmission line of characteristic impedance Z 0 . The signal flow
graph of the circuit is also shown. Verify using the signal flow graph that the
reflection and transmission coefficients can be expressed as
S S S e 2 j
S11 12 22 2 212 j
1 S 22 e
, and

S12 S 21e j
2 2 j
1 S 22
e
2.23 Determine the S-matrix of an ideal T-junction consisting of transmission lines
of characteristic impedance Z 0 .
T

Ans.

1/ 3 2 / 3 2 / 3
[S ] 2 / 3 1 / 3 2 / 3
2 / 3 2 / 3 1 / 3

2.24

Consider a symmetrical Tee-junction in the form of a 1:1 power splitter. The

through-arm has characteristic impedance Z 0 100 ohm and the stub-arm has
Z 0 50

ohm. Determine the S-matrix of the junction. The effect of discontinuity

may be ignored.
1/ 2 1/ 2 1/ 2

[S ] 1 / 2 1 / 2 1 / 2
1/ 2 1/ 2
0

Ans.
2.25

(a) Consider an inductor L 1/6 nH embedded in series in a transmission line

with Z 0 = 50 ohm as shown in Fig. 4(a). Determine the magnitude of
reflection coefficient and transmission phase (delay or advance) produced by it
9
at 30 10 rad/sec.
(b) A capacitor with C = 1/15 pF is mounted in shunt across a transmission
line with Z 0 = 50 ohm as shown in Fig. 4(b). Determine the magnitude of
reflection coefficient and transmission phase (delay or advance) produced by it
9
at 30 10 rad/sec.
(c) Now make a combination of L and C as shown in Fig. 4(c) and insert it in
the same transmission line and compute the magnitude of reflection coefficient
and transmission phase (delay or advance) produced by the combination at
30 10 9 rad/sec.
(d) Comment on the behavior of the circuit in (c) compared to its constituents
behavior in (a) and (b).

42

Fig. 4(a)
2.26

Fig. 4(b)

Fig. 4(c)

This problem emphasizes that an unduly long interconnect line between the
components may seriously affect the power transfer between them especially
if the characteristic impedance of the interconnect is different from the
impedances at its ends. A transmission line section of length is connected to
the source at one end and the load at the other end as shown below

Show that the voltage V L across the load is given by the expression

VL

V0 Z in
Z
cos( ) j sin( ) 0
RS Z in
Z in

where
Z in Z 0

R L jZ 0 tan( )
Z 0 jRL tan(

V
Compute and plot L for Z 0 50 , RL 30 , RS 20 2cm ,
2 / and VS 2volt . The frequency may be varied from 1MHz to 10
GHz.
See the solution at the end.

43

*
Table 2.1: Conversions between two-port parameters normalized to Z 0 1 with
K K 11 K 22 K 12 K 21
S
Z
Y

b1 S11
b S
2 21

S12 a1
S 22 a 2

Z 1 Z 22 1 Z 12 Z 21
S11 11
Z 11 1 Z 22 1 Z 12 Z 21

S11

1 Y11 1 Y22 Y12Y21

1 Y11 1 Y22 Y12Y21

S12

2Z 12
Z 11 1 Z 22 1 Z 12 Z 21

S12

2Y12
1 Y11 1 Y22 Y12Y21

S 21

2Z 21
Z 11 1 Z 22 1 Z 12 Z 21

S 21

2Y21
1 Y11 1 Y22 Y12Y21

S 22

Z11

1 S11 1 S 22 S12 S 21
1 S11 1 S 22 S12 S 21

Z12

2 S12
1 S11 1 S 22 S12 S 21

Z 21

2S 21
1 S11 1 S 22 S12 S 21

1 S11 1 S 22 S12 S 21
1 S11 1 S 22 S12 S 21
1 S11 1 S 22 S12 S 21
Y11
1 S11 1 S 22 S12 S 21

Z 11 1 Z 22 1 Z 12 Z 21
Z 11 1 Z 22 1 Z 12 Z 21

Y21

2 S12
1 S11 1 S 22 S12 S 21
2S 21

1 S11 1 S 22 S12 S 21
1 S11 1 S 22 S12 S 21
Y22
1 S11 1 S 22 S12 S 21

1 Y11 1 Y22 Y12Y21

1 Y11 1 Y22 Y12Y21

Y22
Y
Y
Z 12 12
Y
Y21
Z 21
Y
Y11
Z 22
Y
Z 11

V1
Z11
V Z
2
21

Z 22

Y12

S 22

Z 22
Z
Z 12
Y12
Z
Z 21
Y21
Z
Z 11
Y22
Z

Z12 I1
Z 22 I 2

Y11

A BC
S11
A BC
2( AD _ BC
S12
A BC D
2
S 21
A BC

A B C
A BC
A
Z 11
C
A
Z 12
C
1
Z 21
C
D
Z 22
C
S 22

D
B
A

B
1

B
A

Y11

I 1 Y11 Y12 V1
I Y

2 21 Y22 V 2

Y12
Y21
Y22

Table 2.1 (continued)

A
B

C
D

1 S11 1 S 22 S12 S 21
2S 21

1 S11 1 S 22 S12 S 21
2 S 21

1 S11 1 S 22 S12 S 21
2S 21

1 S11 1 S 22 S12 S 21
2 S 21

Z 11
Z 21

Y22
Y21

Z
Z 21

1
Y21

1
Z 21

Y
Y21

Z 22
Z 21

Y11
Y21

*All Z, Y, and A parameters are normalized with respect to Z 0 Y0 1 . This implies that
normalizing impedances at two ports (for S-matrix representation) are equal. The
normalized ABCD parameters given in this Table are defined as
Z 0 B V2
V1 A
I C/Z
D I 2
0
1

V1 A B V 2
I C D I

1

Need for distributed parameter approach

A lumped element circuit consists of inductor, capacitor, etc connected through wire or
transmission lines called interconnect. The bases of low frequency circuit design are the
Kirchhoffs voltage and current laws, and ignore the spatial variation of voltage. The voltage
variation on the interconnect between lumped elements may be described as
V ( z , t ) V0 cos(t z )
(2.1)
where V0 is the maximum voltage, is the radian frequency and is the phase constant of
the signal. At low frequencies, where the length of interconnect is negligible compared to the
wavelength, the associated voltage variation given by (2.1) is negligible and the circuit
behaves according to the design. However, as the frequency of operation is increased the
length of the interconnect begins to tell on the circuit performance. The deteriorating effect of
the interconnect on the circuit performance can be reduced by eliminating them as far as
possible. To understand the effect of interconnect with frequency, let us study the following
experiment.
The experiment consists of a sinusoidal voltage source V0 with internal resistance RS
connected to a lumped load R L by a length of wire as shown in Fig. 2.1. Let the wire or
interconnect length be 2cm and is assumed lossless. Let this circuit be realized on a printed
circuit board (PCB) with the interconnection wire in the form of a narrow strip of
metallization. The strip and the ground metallization of PCB may be modeled as a
transmission line of characteristic impedance Z 0 . Alternatively, if the hook-up wires are used
to realize the circuit, the combination of interconnection and ground wires may again be
modeled as parallel wire transmission line of characteristic impedance Z 0 . The circuit may
now be analyzed according to the transmission line theory of Sec. 1.2.

Fig. 2.1: A transmission line fed by a source at one end

and terminated in a load at the other end.
For RS R L Z 0 , it is found from the analysis of the circuit of Fig. 2.1 that the voltage
V
VL 0 e j
2
across the load,
, that is, interconnect affects the phase only and not amplitude.
For any other combination of Z 0 , RL , RS , and the voltage across the load will vary with
frequency. The circuit may be analyzed as follows:
From the transmission line theory, the input impedance seen by the source is given by

Z in Z 0

RL jZ 0 tan( )
Z 0 jRL tan(

(2.2)

and the voltage and current at the source end of the line is given by
V0
V0
IS
VS
Z in
RS Z in
RS Z in

(2.3)
The voltage and current can be translated to the load side of the line by using (1.20)
jZ 0 sin( ) V L
VS cos( )
I j sin( ) / Z
cos( ) I L
0
S
(2.4)
It yields
VZ
Z
VL 0 in cos( ) j sin( ) 0
RS Z in
Z in
(2.5)
V0 j
VL e
2
This expression yields
for RS R L Z 0 as expected. The voltage V L as a
function of frequency is plotted in Fig. 2.2 for Z 0 50 , RL 30 , RS 20ohm 2cm ,
and VS 2volt . It is seen that VL V0 / 2 at low frequencies and starts varying as the
frequency is increased.

The effect of interconnect on the circuit performance can be included if the line is divided
into a number of segments of sufficiently small lengths so that the voltage or current over
each segment can be assumed to be constant as shown in Fig. 2.2. Each line segment of
length z can be modeled by a series resistance R1 , series inductance L1 , and shunt
capacitance C1 . This equivalent circuit will account for voltage and current variation over the
segment length z . We can build the complete lumped circuit model for the interconnect by
combining the effect of all the segments into which the interconnect is divided. An integrated
approach is to consider the line as a single segment, characterize the line in terms of
distributed parameters R, L, C, and G where these line constants are given in terms of unit
length, and the sinusoidal voltage variation given by (2.1) is built into the analysis. The
propagation of voltage and current waves on the line is discussed in Sec 1.2 and the voltage
and current distributions are given by (1.20).