3/4/2009 4_3 The Scattering Matrix 1/3

4.3 – The Scattering Matrix

Reading Assignment: pp. 174-183

Admittance and Impedance matrices use the quantities I (z),

V (z), and Z (z) (or Y (z)).

Q: Is there an equivalent matrix for transmission line

activity expressed in terms of V + ( z ) , V −( z ) , and Γ( z ) ?

A: Yes! Its called the scattering matrix.

HO: THE SCATTERING MATRIX

Q: Can we likewise determine something physical about our

device or network by simply looking at its scattering matrix?

A: HO: MATCHED, RECIPROCAL, LOSSLESS

EXAMPLE: A LOSSLESS, RECIPROCAL DEVICE

Q: Isn’t all this linear algebra a bit academic? I mean, it

can’t help us design components, can it?

A: It sure can! An analysis of the scattering matrix can tell

us if a certain device is even possible to construct, and if so,
what the form of the device must be.

HO: THE MATCHED, LOSSLESS, RECIPROCAL 3-PORT NETWORK

Jim Stiles The Univ. of Kansas Dept. of EECS

3/4/2009 4_3 The Scattering Matrix 2/3

HO: THE MATCHED, LOSSLESS, RECIPROCAL 4-PORT NETWORK

Q: But how are scattering parameters useful? How do we

use them to solve or analyze real microwave circuit problems?

A: Study the examples provided below!

EXAMPLE: THE SCATTERING MATRIX

EXAMPLE: SCATTERING PARAMETERS

Q: OK, but how can we determine the scattering matrix of a

device?

A: We must carefully apply our transmission line theory!

EXAMPLE: DETERMINING THE SCATTERING MATRIX

Q: Determining the Scattering Matrix of a multi-port device

would seem to be particularly laborious. Is there any way to
simplify the process?

A: Many (if not most) of the useful devices made by us

humans exhibit a high degree of symmetry. This can greatly
simplify circuit analysis—if we know how to exploit it!

HO: CIRCUIT SYMMETRY

EXAMPLE: USING SYMMETRY TO DETERMINING A SCATTERING

MATRIX

Jim Stiles The Univ. of Kansas Dept. of EECS

3/4/2009 4_3 The Scattering Matrix 3/3

Q: Is there any other way to use circuit symmetry to our

advantage?

A: Absolutely! One of the most powerful tools in circuit

analysis is Odd-Even Mode analysis.

HO: SYMMETRIC CIRCUIT ANALYSIS

HO: ODD-EVEN MODE ANALYSIS

EXAMPLE: ODD-EVEN MODE CIRCUIT ANALYSIS

Q: Aren’t you finished with this section yet?

A: Just one more very important thing.

HO: GENERALIZED SCATTERING PARAMETERS

EXAMPLE: THE SCATTERING MATRIX OF A CONNECTOR

Jim Stiles The Univ. of Kansas Dept. of EECS

2/25/2009 The Scattering Matrix present 1/14

The Scattering Matrix

At “low” frequencies, we can completely characterize a linear device or network using an
impedance matrix, which relates the currents and voltages at each device terminal to the
currents and voltages at all other terminals.

But, at microwave frequencies, it is difficult to measure

total currents and voltages!

* Instead, we can measure the magnitude and phase of each of the two transmission
line waves V + (z ) and V − (z ) .

* In other words, we can determine the relationship between the incident and
reflected wave at each device terminal to the incident and reflected waves at all
other terminals.

These relationships are completely represented by the scattering matrix. It completely

describes the behavior of a linear, multi-port device at a given frequency ω , and a given
line impedance Z0.
Jim Stiles The Univ. of Kansas Dept. of EECS
2/25/2009 The Scattering Matrix present 2/14

Consider now the 4-port microwave device shown below:

Note that we have now

V2− ( z 2 ) Z0 V2+ ( z 2 )
characterized transmission line
activity in terms of incident and
port
“reflected” waves. Note the z 2 = z 2P
2
negative going “reflected” waves
V1 + ( z1 ) port 1 port 3 V3− ( z 3 )
can be viewed as the waves exiting
the multi-port network or device. 4-port
Z0 microwave Z0
Æ Viewing transmission line device
activity this way, we can fully V1 − ( z1 ) z1 = z1P z3 = z3P V3+ ( z 3 )
characterize a multi-port device
port
by its scattering parameters! z 4 = z 4P
4

Z0
V4+ ( z 4 ) V4− ( z 4 )

Jim Stiles The Univ. of Kansas Dept. of EECS

2/25/2009 The Scattering Matrix present 3/14

Say there exists an incident wave on port 1 (i.e., V1 + ( z1 ) ≠ 0 ), while the incident waves on
all other ports are known to be zero (i.e., V2+ ( z 2 ) =V3+ ( z 3 ) =V4+ ( z 4 ) = 0 ).

V1 + ( z1 ) port 1

+ Say we measure/determine the voltage of the wave flowing

Z0 V1
+
(z 1
= z1 p ) into port 1, at the port 1 plane (i.e., determine
− V1 + ( z1 = z1P ) ).
port 2 V2− ( z 2 )
z 1 = z 1P

+
Say we then measure/determine the voltage of the wave
flowing out of port 2, at the port 2 plane (i.e., determine V2
−
(z 2
= z2p ) Z0
V2− ( z 2 = z 2P ) ). −

z 2 = z 2P

The complex ratio between V1 + (z1 = z1P ) and V2 − (z 2 = z 2P ) is know as the scattering
parameter S21:
V2− (z 2 = z 2P ) V02− e + j β z P V02− + j β (z P +z P )
2

S21 = + = = e 2 1

V1 (z1 = z1P ) V01+ e − j β z P V01+

1

Likewise, the scattering parameters S31 and S41 are:

V3− (z3 = z3P ) V4− (z 4 = z 4P )

S31 = + and S41 = +
V1 (z1 = z1P ) V1 (z1 = z1P )

Jim Stiles The Univ. of Kansas Dept. of EECS

2/25/2009 The Scattering Matrix present 4/14

We of course could also define, say, scattering parameter S34 as the ratio between the
complex values V4+ (z 4 = z 4P ) (the wave into port 4) and V3 − (z 3 = z 3P ) (the wave out of port
3), given that the input to all other ports (1,2, and 3) are zero.

Thus, more generally, the ratio of the wave incident on port n to the wave emerging from
port m is:

Vm− (z m = z mP )
Smn = + (given that Vk + ( z k ) = 0 for all k ≠ n )
Vn (zn = znP )

Note that frequently the port positions are assigned a zero value (e.g., z1P = 0, z 2P = 0 ).
This of course simplifies the scattering parameter calculation:
Microwave
Vm (z m = 0) V0m e
− − +j β 0
V − lobe
Smn = = = 0m

Vn + (zn = 0) V0n+ e − j β 0
V +
0n

We will generally assume that the port locations are defined as znP = 0 ,
and thus use the above notation. But remember where this expression
came from!

Jim Stiles The Univ. of Kansas Dept. of EECS

2/25/2009 The Scattering Matrix present 5/14

Q: But how do we
ensure that only
one incident wave Γ2 L = 0
is non-zero ?

V2− ( z 2 ) Z0 V2+ ( z 2 ) = 0

A: Terminate all other ports with a matched load!

V1 + ( z1 ) V3− ( z 3 )
4-port
Z0 microwave Z0 Γ3L = 0
device
V1 − ( z1 ) V3+ ( z 3 ) = 0

V3+ ( z 3 ) = 0 Z0 V4− ( z 4 )

Γ 4L = 0

Jim Stiles The Univ. of Kansas Dept. of EECS

2/25/2009 The Scattering Matrix present 6/14

Note that if the ports are terminated in a matched load (i.e., Z L = Z 0 ), then ΓnL = 0 and
therefore:
Vn+ ( zn ) = 0

In other words, terminating a port ensures that there will be no

signal incident on that port!

Q: Just between you and me, I think you’ve messed this up! In all previous handouts you
said that if ΓL = 0 , the wave in the minus direction would be zero:

V − (z ) = 0 if ΓL = 0

but just now you said that the wave in the positive direction would be zero:

V + (z ) = 0 if ΓL = 0

Of course, there is no way that both statements can be correct!

A: Actually, both statements are correct! You must be careful to understand the physical
definitions of the plus and minus directions—in other words, the propagation directions of
waves Vn+ ( zn ) and Vn− ( zn )!

Jim Stiles The Univ. of Kansas Dept. of EECS

2/25/2009 The Scattering Matrix present 7/14

For example, we originally analyzed this case:

V + (z )

Z0 ΓL V − (z ) = 0 if ΓL = 0

V − (z )

In this original case, the wave incident on the load is V + ( z ) (plus direction), while the
reflected wave is V − ( z ) (minus direction).

Contrast this with the case we are now considering:

port n Vn − ( zn )
N-port
Microwave Z0 ΓnL
Network

Vn + ( zn )

Jim Stiles The Univ. of Kansas Dept. of EECS

2/25/2009 The Scattering Matrix present 8/14

For this current case, the situation is reversed. The wave incident on the load is now
denoted as Vn− ( zn ) (coming out of port n), while the wave reflected off the load is now
denoted as Vn+ ( zn ) (going into port n ).

As a result, Vn+ ( zn ) = 0 when ΓnL = 0 !

Perhaps we could more generally state that for some load Γ L :

V reflected ( z = z L ) = ΓL V incident ( z = z L )

For each case, you must be able to correctly identify the

mathematical statement describing the wave incident on, and
reflected from, some passive load.

Like most equations in engineering, the variable names can

change, but the physics described by the mathematics will not!

Jim Stiles The Univ. of Kansas Dept. of EECS

2/25/2009 The Scattering Matrix present 9/14

Now, back to our discussion of S-parameters. We found that if znP = 0 for all ports n,
the scattering parameters could be directly written in terms of wave amplitudes V0n+ and
V0−m .

V0−m
Smn = + (when Vk + ( z k ) = 0 for all k ≠ n )
V0n

Which we can now equivalently state as:

V0−m
Smn = + (when all ports, except port n , are terminated in matched loads)
V0n

One more important note—notice that for the ports terminated in matched loads (i.e.,
those ports with no incident wave), the voltage of the exiting wave is also the total
voltage!

Vm ( z m ) =V0+m e − j β zn +V0−m e + j β zn
= 0 +V0−m e + j β zm
=V0−m e + j β zm (for all terminated ports)

Jim Stiles The Univ. of Kansas Dept. of EECS

2/25/2009 The Scattering Matrix present 10/14

Thus, the value of the exiting wave at each terminated port is likewise the value of the
total voltage at those ports:

Vm ( 0 ) =V0+m +V0−m
= 0 +V0−m
=V0−m (for all terminated ports)

And so, we can express some of the scattering parameters equivalently as:

Vm ( 0 )
Smn = (for terminated port m , i.e., for m ≠ n )
V0n+

You might find this result helpful if attempting to determine scattering parameters
where m ≠ n (e.g., S21, S43, S13), as we can often use traditional circuit theory to easily
determine the total port voltage Vm ( 0 ) .

However, we cannot use the expression above to determine the scattering parameters
when m = n (e.g., S11, S22, S33).

Jim Stiles The Univ. of Kansas Dept. of EECS

2/25/2009 The Scattering Matrix present 11/14

Think about this! The scattering parameters for these cases are:

V0n−
Snn = +
V0n
Γ2 L = 0
V1 ( 0 ) =V1 + ( 0 ) +V1 − ( 0 ) V3 ( 0 ) = V3− ( 0 )

Therefore, port n is a port V2− ( z 2 ) Z0 V2+ ( z 2 ) = 0

where there actually is some
incident wave V0n+ (port n is not
terminated in a matched V1 + ( z1 ) ≠ 0 V3− ( z 3 )
load!).
4-port
Z0 microwave Z0 Γ 3L = 0
And thus, the total voltage is
device
not simply the value of the
exiting wave, as both an V1 − ( z1 ) V3+ ( z 3 ) = 0
incident wave and exiting wave
exists at port n.
V4+ ( z 4 ) = 0 Z0 V4− ( z 4 )

Γ 4L = 0

Jim Stiles The Univ. of Kansas Dept. of EECS

2/25/2009 The Scattering Matrix present 12/14

Typically, it is much more difficult to determine/measure the scattering parameters of

the form Snn , as opposed to scattering parameters of the form Smn (where m ≠ n ) where
there is only an exiting wave from port m !

We can use the scattering matrix to determine the solution for a more general
circuit—one where the ports are not terminated in matched loads!

Q: I’m not understanding the importance scattering

parameters. How are they useful to us microwave engineers?

A: Since the device is linear, we can apply superposition. The output at any port
due to all the incident waves is simply the coherent sum of the output at that port
due to each wave!

For example, the output wave at port 3 can be determined by (assuming znP = 0 ):

V03− = S34V04+ + S33V03+ + S32V02+ + S31V01+

Jim Stiles The Univ. of Kansas Dept. of EECS

2/25/2009 The Scattering Matrix present 13/14

More generally, the output at port m of an N-port device is:

N
V0m = ∑ Smn V0n+
−
( znP = 0)
n =1

This expression can be written in matrix form as:

V− = S V+
Where V − is the vector:

T
V − = ⎡⎣V01− ,V02− ,V03− , … ,V0−N ⎤⎦
and V + is the vector:

T
V + = ⎡⎣V01+ ,V02+ ,V03+ , … ,V0+N ⎤⎦

Therefore S is the scattering matrix:

⎡ S11 … S1n ⎤
S = ⎢⎢ ⎥
⎥
⎢⎣Sm 1 Smn ⎥⎦

Jim Stiles The Univ. of Kansas Dept. of EECS

2/25/2009 The Scattering Matrix present 14/14

The scattering matrix is a N by N matrix that completely characterizes a linear, N-port

device. Effectively, the scattering matrix describes a multi-port device the way that ΓL
describes a single-port device (e.g., a load)!

But beware! The values of the scattering matrix for a particular device or
network, just like ΓL , are frequency dependent! Thus, it may be more
instructive to explicitly write:

⎡ S11 (ω ) … S1n (ω ) ⎤
S (ω ) = ⎢⎢ ⎥
⎥
⎢⎣Sm 1 (ω ) Smn (ω ) ⎥⎦

Also realize that—also just like ΓL—the scattering matrix is dependent on both the
device/network and the Z0 value of the transmission lines connected to it.

Thus, a device connected to transmission lines with Z 0 = 50Ω will have a completely
different scattering matrix than that same device connected to transmission lines
with Z 0 = 100Ω !!!

Jim Stiles The Univ. of Kansas Dept. of EECS

2/26/2009 Matched reciprocal lossless present 1/11

Matched, Lossless, Reciprocal Devices

As we discussed earlier, a device can be lossless or reciprocal. In addition, we can
likewise classify it as being matched. Let’s examine each of these three characteristics,
and how they relate to the scattering matrix.

Matched

A matched device is another way of saying that the input impedance at each port is equal
to Z0 when all other ports are terminated in matched loads. As a result, the reflection
coefficient of each port is zero—no signal will be come out of a port if a signal is incident
on that port (but only that port!).

In other words, we want:

Vm− = Smm Vm+ = 0 for all m

a result that occurs when:

Smm = 0 for all m if matched

Jim Stiles The Univ. of Kansas Dept. of EECS

2/26/2009 Matched reciprocal lossless present 2/11

We find therefore that a matched device will exhibit a scattering matrix where all
diagonal elements are zero.

Therefore:
⎡ 0 0.1 j 0.2 ⎤
⎢ ⎥
S = 0.1
⎢ 0 0.3 ⎥
⎢ ⎥
⎢⎣ j 0.2 0.3 0 ⎥⎦

is an example of a scattering matrix for a matched, three port device.

Lossless

For a lossless device, all of the power that delivered to each device port must eventually
find its way out!

In other words, power is not absorbed by the network—no power to be converted to

heat!

Recall the power incident on some port m is related to the amplitude of the incident wave
(V0+m ) as:
2
V0+m
Pm+ =
2Z 0
While power of the wave exiting the port is:

Jim Stiles The Univ. of Kansas Dept. of EECS

2/26/2009 Matched reciprocal lossless present 3/11

2
V0−m
Pm− =
2Z 0

Thus, the power delivered to (absorbed by) that port is the difference of the two:

2 2
V0+m V0−m
∆Pm = Pm+ − Pm− = −
2Z 0 2Z 0

Thus, the total power incident on an N-port device is:

N N
1 2
P +
= ∑ Pm +
=
2Z 0 ∑ V0+m
m =1 m =1
Note that:
N
2 H
∑ V0+m = ( V+ ) V+
m =1

where operator H indicates the conjugate transpose (i.e., Hermetian transpose)

H
operation, so that ( V+ ) V+ is the inner product (i.e., dot product, or scalar product) of
complex vector V+ with itself.

Thus, we can write the total power incident on the device as:

Jim Stiles The Univ. of Kansas Dept. of EECS

2/26/2009 Matched reciprocal lossless present 4/11

N H
1 2 ( V+ ) V+
2Z 0 m∑
P +
= V0+m =
=1 2Z 0

Similarly, we can express the total power of the waves exiting our M-port network to be:

N H
1 2 ( V− ) V−
2Z 0 m∑
P −
= V0−m =
=1 2Z 0

Now, recalling that the incident and exiting wave amplitudes are related by the scattering
matrix of the device:
V− = S V+

Thus we find:
H H
( V− ) V− ( V+ ) S H S V+
P −
= =
2Z 0 2Z 0

Now, the total power delivered to the network is:

M
∆P = ∑ ∆P = P+ −P−
m =1

Or explicitly:

Jim Stiles The Univ. of Kansas Dept. of EECS

2/26/2009 Matched reciprocal lossless present 5/11

∆P = P + − P −
H H
( V+ ) V+ ( V+ ) S H S V+
= −
2Z 0 2Z 0
1 H
= ( V+ ) ( I − S H S ) V+
2Z 0

where I is the identity matrix.

Q: Is there actually some point to this long, rambling, complex presentation?

A: Absolutely! If our M-port device is lossless then the total power exiting the device
must always be equal to the total power incident on it.

If network is lossless, then P + = P − .

Or stated another way, the total power delivered to the device (i.e., the power absorbed
by the device) must always be zero if the device is lossless!

If network is lossless, then ∆P = 0

Thus, we can conclude from our math that for a lossless device:

Jim Stiles The Univ. of Kansas Dept. of EECS

2/26/2009 Matched reciprocal lossless present 6/11

1 H
∆P = ( V+ ) ( I − S H S ) V+ =0 for all V+
2Z 0

This is true only if:

I − SH S = 0 ⇒ SH S = I

Thus, we can conclude that the scattering matrix of a lossless device has the
characteristic:

If a network is lossless, then SH S = I

Q: Huh? What exactly is this supposed to tell us?

A: A matrix that satisfies S H S = I is a special kind of matrix known as a unitary

matrix.

If a network is lossless, then its scattering matrix S is unitary.

Q: How do I recognize a unitary matrix if I see one?

Jim Stiles The Univ. of Kansas Dept. of EECS

2/26/2009 Matched reciprocal lossless present 7/11

A: The columns of a unitary matrix form an orthonormal set!

⎡ S11 S12 S13 S14 ⎤

⎢ ⎥
S S22 S23 S22 ⎥
S = ⎢⎢ 21
S S33 S33 S33 ⎥
⎢ 31 ⎥
⎢⎣S41 S42 S43 S44 ⎥⎦

matrix
columns

In other words, each column of the scattering matrix will have a magnitude equal to one:

∑ Smn
2
=1 for all n
m =1

while the inner product (i.e., dot product) of dissimilar columns must be zero.

∑
n
Sni Snj
=1
∗
=S1i S1∗j + S2i S2∗j + " + SNi SN∗j = 0 for all i ≠ j

In other words, dissimilar columns are orthogonal.

Jim Stiles The Univ. of Kansas Dept. of EECS

2/26/2009 Matched reciprocal lossless present 8/11

Consider, for example, a lossless three-port device. Say a signal is incident on port 1, and
that all other ports are terminated. The power incident on port 1 is therefore:

2
V01+
P1 + =
2Z 0

while the power exiting the device at each port is:

2 2
V0−m Sm 1V01− 2
Pm− = = = Sm 1 P1 +
2Z 0 2Z 0

The total power exiting the device is therefore:

P − = P1 − + P2− + P3−
2
= S11 P1 + + S21 2 P1 + + S31 2 P1 +
= ( S11 2
+ S21 2
+ S31 2
) P1 +

Since this device is lossless, then the incident power (only on port 1) is equal to exiting
power (i.e, P − = P1 + ). This is true only if:

2 2 2
S11 + S21 + S31 =1

Jim Stiles The Univ. of Kansas Dept. of EECS

2/26/2009 Matched reciprocal lossless present 9/11

Of course, this will likewise be true if the incident wave is placed on any of the other
ports of this lossless device:

2 2 2
S12 + S22 + S32 =1
2 2 2
S13 + S23 + S33 =1

We can state in general then that:

∑
2
Smn =1 for all n
m =1

In other words, the columns of the scattering matrix must have unit magnitude (a
requirement of all unitary matrices). It is apparent that this must be true for energy to
be conserved.

An example of a (unitary) scattering matrix for a lossless device is:

⎡ 0 1
2 j 3
2 0 ⎤
⎢ 1 0 j 3 2 ⎥⎥
0
S=⎢ 3
2

⎢j 2 0 0 1
2 ⎥
⎢ ⎥
⎣ 0 j 3
2
1
2 0 ⎦

Jim Stiles The Univ. of Kansas Dept. of EECS

2/26/2009 Matched reciprocal lossless present 10/11

Reciprocal

Recall reciprocity results when we build a passive (i.e., unpowered) device with simple
materials.

For a reciprocal network, we find that the elements of the scattering matrix are related
as:
Smn = Snm

For example, a reciprocal device will have S21 = S12 or S32 = S23 . We can write
reciprocity in matrix form as:

ST = S if reciprocal

where T indicates (non-conjugate) transpose.

Jim Stiles The Univ. of Kansas Dept. of EECS

2/26/2009 Matched reciprocal lossless present 11/11

An example of a scattering matrix describing a reciprocal, but lossy and non-matched

device is:

⎡ 0.10 −0.40 − j 0.20 0.05 ⎤

⎢ −0.40 j 0.20 0 j 0.10 ⎥⎥
S=⎢
⎢ − j 0.20 0 0.10 − j 0.30 −0.12 ⎥
⎢ ⎥
⎣ 0.05 j 0.10 −0.12 0 ⎦

Jim Stiles The Univ. of Kansas Dept. of EECS

2/23/2007 Example A Lossless Reciprocal Network 1/4

Example: A Lossless,
Reciprocal Network
A lossless, reciprocal 3-port device has S-parameters of
S11 = 1 2 , S31 = 1 2 , and S33 = 0 . It is likewise known that all
scattering parameters are real.

Æ Find the remaining 6 scattering parameters.

Q: This problem is clearly

impossible—you have not provided
us with sufficient information!

A: Yes I have! Note I said the device was lossless and

reciprocal!

Start with what we currently know:

⎡ 1 2 S12 S13 ⎤
S = ⎢⎢S21 S22 S23 ⎥⎥
⎢⎣ 1 2 S32 0 ⎥⎦

Because the device is reciprocal, we then also know:

S21 = S12 S13 = S31 = 1

2
S32 = S23

Jim Stiles The Univ. of Kansas Dept. of EECS

2/23/2007 Example A Lossless Reciprocal Network 2/4

And therefore:

⎡ 1 2 S21 1 2 ⎤
S = ⎢⎢S21 S22 S32 ⎥⎥
⎢⎣ 1 2 S32 0 ⎥⎦

Now, since the device is lossless, we know that:

2 2 2
1 = S11 + S21 + S31
= ( 1 2 ) + S21 + ( 1 )
2 2 2
2

2 2 2
Columns have 1 = S12 + S22 + S32
2 2 2
unit magnitude. = S21 + S22 + S32

2 2 2
1 = S13 + S23 + S33
= ( 1 2 ) + S32 + ( 1 )
2 2 2
2

and:

0 = S11S12∗ + S21S22∗ + S31S32∗

= 1
2 S21∗ + S21S22∗ + 1
2
S32∗
Columns are
0 = S11S13∗ + S21S23∗ + S31S33∗ orthogonal.

= 1
2 ( )+S
1
2 21S32∗ + 1
2 (0)

0 = S12S13∗ + S22S23∗ + S32S33∗

= S21 ( 1 2 )+S 22 S32∗ + S32 ( 0 )

Jim Stiles The Univ. of Kansas Dept. of EECS

2/23/2007 Example A Lossless Reciprocal Network 3/4

These six expressions simplify to:

S21 = 1
2

2 2 2
1 = S21 + S22 + S32

S32 = 1
2

0= 1
2 S21 + S21S22 + 1
2
S32

0= (2 2 ) + S21S32
1

0 = S21 ( 1 2 )+S S32

22

where we have used the fact that since the elements are all
real, then S21∗ = S21 (etc.).

Q Q: I count the expressions and find 6 equations yet

only a paltry 3 unknowns. Your typical buffoonery
appears to have led to an over-constrained condition
for which there is no solution!

A: Actually, we have six real equations and six real

unknowns, since scattering element has a magnitude and
phase. In this case we know the values are real, and thus
the phase is either 0D or 180D (i.e., e j 0 = 1 or
e j π = −1 ); however, we do not know which one!

From the first three equations, we can find the magnitudes:

Jim Stiles The Univ. of Kansas Dept. of EECS

2/23/2007 Example A Lossless Reciprocal Network 4/4

S21 = 1
2 S22 = 1
2 S32 = 1
2

and from the last three equations we find the phase:

S21 = 1
2 S22 = 1
2 S32 = − 1
2

Thus, the scattering matrix for this lossless, reciprocal device

is:
⎡ 12 1
2
1
2
⎤
S = ⎢⎢ 1 2 1 2 −1 2 ⎥⎥
⎢⎣ 1 2 −1 2 0 ⎥⎦

Jim Stiles The Univ. of Kansas Dept. of EECS

3/4/2009 MLR 3 port network present 1/2

A Matched, Lossless,
Reciprocal 3-Port Network
Consider a 3-port device. Such a device ⎡S11 S12 S13 ⎤
would have a scattering matrix : S = ⎢⎢S21 S22 S23 ⎥⎥
⎢⎣S31 S32 S33 ⎥⎦

Assuming the device is passive and made of simple (isotropic) materials, the device will be
reciprocal, so that:
S21 = S12 S31 = S13 S23 = S32

Likewise, if it is matched, we know that S11 = S22 = S33 = 0 .

As a result, a lossless, reciprocal device would have a scattering matrix of the form:

⎡ 0 S21 S31 ⎤
S = ⎢⎢S21 0 S32 ⎥⎥
⎢⎣S31 S32 0 ⎥⎦

Just 3 non-zero scattering parameters define the entire matrix!

Jim Stiles The Univ. of Kansas Dept. of EECS

3/4/2009 MLR 3 port network present 2/2

Likewise, if we wish for this network to be lossless, the scattering matrix must be
unitary, and therefore:
2 2
S21 + S31 = 1 S31∗ S32 = 0
2 2
S21 + S32 = 1 S21∗ S32 = 0
2 2
S31 + S32 = 1 S21∗ S31 = 0

Since each complex value S is represented by two real numbers (i.e., real and
imaginary parts), the equations above result in 9 real equations. The problem is,
the 3 complex values S21, S31 and S32 are represented by only 6 real unknowns.

We have over constrained our problem ! There are no solutions to these

equations !

As unlikely as it might seem, this means that a matched, lossless,

reciprocal 3-port device of any kind is a physical impossibility!

You can make a lossless reciprocal 3-port device, or a matched

reciprocal 3-port device, or even a matched, lossless (but non-
reciprocal) 3-port network.

But try as you might, you cannot make a lossless, matched, and
reciprocal three port component!

Jim Stiles The Univ. of Kansas Dept. of EECS

3/4/2009 MLR 4 port network present 1/3

The Matched, Lossless,

Reciprocal, 4-Port Network
Guess what! I have determined that—unlike a
3-port device—a matched, lossless, reciprocal
4-port device is physically possible! In fact,
I’ve found two general solutions!

The first solution is referred to as the symmetric solution:

⎡0 α jβ 0 ⎤
⎢α 0 0 jβ⎥
S=⎢ ⎥
⎢j β 0 0 α ⎥
⎢ ⎥
⎣0 jβ α 0 ⎦

Note for this symmetric solution, every row and every column of the scattering matrix
has the same four values (i.e., α, jβ, and two zeros)!

Jim Stiles The Univ. of Kansas Dept. of EECS

3/4/2009 MLR 4 port network present 2/3

The second solution is referred to as the anti-symmetric solution:

⎡0 α β 0⎤
⎢α 0 0 −β ⎥
S=⎢ ⎥
⎢β 0 0 α ⎥
⎢
⎣ 0 −β α 0 ⎥⎦

Note that for this anti-symmetric solution, two rows and two columns have the same four
values (i.e., α, β, and two zeros), while the other two row and columns have (slightly)
different values (α, -β, and two zeros)

It is quite evident that each of these solutions are matched and reciprocal. However, to
ensure that the solutions are indeed lossless, we must place an additional constraint on
the values of α, β. Recall that a necessary condition for a lossless device is:

∑
2
Smn =1 for all n
m=1

Applying this to the symmetric case, we find:

α2+ β 2
=1

Jim Stiles The Univ. of Kansas Dept. of EECS

3/4/2009 MLR 4 port network present 3/3

Likewise, for the anti-symmetric case, we also get

α2+ β 2
=1

It is evident that if the scattering matrix is unitary (i.e., lossless), the values α and β
cannot be independent, but must related as:

α2+ β 2
=1

Generally speaking, we will find that α ≥ β . Given the constraint on these two values, we
can thus conclude that:

0≤ β ≤ 1 and 1 ≤ α ≤1
2 2

Jim Stiles The Univ. of Kansas Dept. of EECS

2/23/2007 Example Scattering Parameters 1/4

Example: Scattering
Parameters
Consider a two-port device with a scattering matrix (at some
specific frequency ω0 ):

⎡ 0.1 j 0.7 ⎤
S ( ω = ω0 ) = ⎢
⎣ j 0.7 −0.2 ⎥⎦

and Z 0 = 50Ω .

Say that the transmission line connected to port 2 of this

device is terminated in a matched load, and that the wave
incident on port 1 is:

V1 + ( z1 ) = − j 2 e − j β z 1

where z1P = z 2P = 0 .

Determine:

1. the port voltages V1 ( z1 = z1P ) and V2 ( z 2 = z 2P ) .

2. the port currents I1 ( z1 = z1P ) and I2 ( z 2 = z 2P ) .

3. the net power flowing into port 1

Jim Stiles The Univ. of Kansas Dept. of EECS

2/23/2007 Example Scattering Parameters 2/4

1. Since the incident wave on port 1 is:

V1 + ( z1 ) = − j 2 e − j β z 1

we can conclude (since z1P = 0 ):

V1 + ( z1 = z1P ) = − j 2 e − j β z P 1

= − j 2 e − j β (0)
= −j 2

and since port 2 is matched (and only because its matched!),

we find:

V1 − ( z1 = z1P ) = S11 V1 + ( z1 = z1P )

= 0.1 ( − j 2 )
= − j 0.2
The voltage at port 1 is thus:

V1 ( z1 = z1P ) =V1 + ( z1 = z1P ) +V1 − ( z1 = z1P )

= − j 2.0 − j 0.2
= − j 2.2
−j π 2
= 2.2 e
Likewise, since port 2 is matched:

V2+ ( z 2 = z 2P ) = 0

Jim Stiles The Univ. of Kansas Dept. of EECS

2/23/2007 Example Scattering Parameters 3/4

And also:

V2− ( z 2 = z 2P ) = S21 V1 + ( z1 = z1P )

= j 0.7 ( − j 2 )
= 1.4
Therefore:

V2 ( z 2 = z 2P ) =V2+ ( z 2 = z 2P ) +V2− ( z2 = z2P )

= 0 + 1 .4
= 1.4
= 1.4 e − j 0

2. The port currents can be easily determined from the

results of the previous section.

I 1 ( z 1 = z 1 P ) = I 1+ ( z 1 = z 1 P ) − I 1− ( z 1 = z 1 P )
V1 + ( z1 = z1P ) V1 − ( z1 = z1P )
= −
Z0 Z0
2.0 0.2
= −j +j
50 50
1.8
= −j
50
= − j 0.036
−jπ 2
= 0.036 e

and:

Jim Stiles The Univ. of Kansas Dept. of EECS

2/23/2007 Example Scattering Parameters 4/4

I2 ( z 2 = z 2P ) = I2+ ( z2 = z2P ) − I2− ( z2 = z 2P )

V2+ ( z 2 = z2P ) V2− ( z2 = z2P )
= −
Z0 Z0
0 1. 4
= −
50 50
= −0.028
= 0.028 e + j π

3. The net power flowing into port 1 is:

∆P1 = P1 + − P1 −
2 2
V01+ V01−
= −
2Z 0 2Z 0
2 2
(2 ) − ( 0.2 )
=
2 ( 50 )
= 0.0396 Watts

Jim Stiles The Univ. of Kansas Dept. of EECS

2/23/2007 Example The Scattering Matrix 1/6

Example: The
Scattering Matrix
Say we have a 3-port network that is completely characterized
at some frequency ω by the scattering matrix:

⎡0.0 0.2 0.5 ⎤

S = ⎢⎢ 0.5 0.0 0.2 ⎥⎥
⎢⎣ 0.5 0.5 0.0 ⎥⎦

A matched load is attached to port 2, while a short circuit has

been placed at port 3:
Z = Z0

zP 2 = 0
Z0
V2− (z) V2+ (z)
port
2
z P 1 = 0 V + (z) V3− (z) z P 3 = 0
1

3-port
Z0 port 1 microwave port 3 Z0 Z =0
device
V1 − (z) V3+ (z)

Jim Stiles The Univ. of Kansas Dept. of EECS

2/23/2007 Example The Scattering Matrix 2/6

Because of the matched load at port 2 (i.e., ΓL = 0 ), we know

that:
V2+ (z 2 = 0) V02+
= =0
V2− (z 2 = 0) V02−

and therefore:
V02+ = 0

You’ve made a terrible mistake!

Fortunately, I was here to
correct it for you—since ΓL = 0 ,
the constant V02− (not V02+ ) is
equal to zero.

NO!! Remember, the signal V2− (z ) is incident on the matched

load, and V2+ (z ) is the reflected wave from the load (i.e., V2+ (z )
is incident on port 2). Therefore, V02+ = 0 is correct!

Likewise, because of the short circuit at port 3 ( ΓL = −1 ):

V3+ (z 3 = 0) V03+
= = −1
V3− (z 3 = 0) V03−

and therefore:

V03+ = −V03−

Jim Stiles The Univ. of Kansas Dept. of EECS

2/23/2007 Example The Scattering Matrix 3/6

Problem:

a) Find the reflection coefficient at port 1, i.e.:

V01−
Γ1 +
V01

b) Find the transmission coefficient from port 1 to port 2, i.e.,

V02−
T21 +
V01

I am amused by the trivial

problems that you apparently
find so difficult. I know that:

V01−
Γ1 = + = S11 = 0.0
V01
and

V02−
T21 = + = S21 = 0.5
V01

NO!!! The above statement is not correct!

Remember, V01− V01+ = S11 only if ports 2 and 3 are

terminated in matched loads! In this problem port 3
is terminated with a short circuit.

Jim Stiles The Univ. of Kansas Dept. of EECS

2/23/2007 Example The Scattering Matrix 4/6

Therefore:
V01−
Γ1 = + ≠ S11
V01
and similarly:

V02−
T21 = + ≠ S21
V01

To determine the values T21 and Γ1 , we must start with the

three equations provided by the scattering matrix:

V01− = 0.2V02+ + 0.5V03+

V02− = 0.5V01+ + 0.2V03+

V03− = 0.5V01+ + 0.5V02+

and the two equations provided by the attached loads:

V02+ = 0

V03+ = −V03−

Jim Stiles The Univ. of Kansas Dept. of EECS

2/23/2007 Example The Scattering Matrix 5/6

We can divide all of these equations by V01+ , resulting in:

V01− V02+ V03+

Γ1 = + = 0.2 + + 0.5 +
V01 V01 V01

V02− V03+
T21 = + = 0.5 + 0. 2 +
V01 V01

V03− V02+
= 0.5 + 0.5 +
V01+
V01

V02+
=0
V01+

V03+ V03−
=− +
V01+ V01

Look what we have—5 equations and 5 unknowns! Inserting

equations 4 and 5 into equations 1 through 3, we get:

V01− V03+
Γ1 = + = −0.5 +
V01 V01

V02− V03+
T21 = + = 0.5 − 0.2 +
V01 V01

V03−
= 0.5
V01+

Jim Stiles The Univ. of Kansas Dept. of EECS

2/23/2007 Example The Scattering Matrix 6/6

Solving, we find:

Γ1 = −0.5 ( 0.5 ) = −0.25

T21 = 0.5 − 0.2 ( 0.5 ) = 0.4

Jim Stiles The Univ. of Kansas Dept. of EECS

2/23/2007 Example Determining the Scattering Matrix 1/5

Example: Determining the

Scattering Matrix
Let’s determine the scattering matrix of this two-port device:

Z0 2Z0 Z0

z1 z2
z 1P = 0 z 2P = 0

The first step is to terminate port 2 with a matched load, and

then determine the values:

V1 − ( z1 = z P 1 ) and V2− ( z 2 = z P 2 )

in terms of V1 + ( z1 = z P 1 ) .

+ +
Z0 V1 ( z 1 ) 2Z0 V2 ( z 2 = 0 ) Z0
− −

z1
z 1P = 0 z 2P = 0

Jim Stiles The Univ. of Kansas Dept. of EECS

2/23/2007 Example Determining the Scattering Matrix 2/5

Recall that since port 2 is matched, we know that:

V2+ ( z 2 = z 2P ) = 0
And thus:

V2 ( z 2 = 0 ) = V2+ ( z2 = 0 ) +V2− ( z2 = 0 )
= 0 +V2− ( z 2 = 0 )
= V2− ( z 2 = 0 )

In other words, we simply need to determine V2 ( z 2 = 0 ) in order

to find V2− ( z 2 = 0 ) !

However, determining V1 − ( z1 = 0 ) is a bit trickier. Recall that:

V1 ( z1 ) =V1 + ( z1 ) +V1 − ( z1 )

Therefore we find V1 ( z1 = 0 ) ≠ V1 − ( z1 = 0 ) !

Now, we can simplify this circuit:

+
Z0 2
V1 ( z 1 ) Z0
3
−

z1
z 1P = 0

And we know from the telegraphers equations:

Jim Stiles The Univ. of Kansas Dept. of EECS

2/23/2007 Example Determining the Scattering Matrix 3/5

V1 ( z1 ) =V1 + ( z1 ) +V1 − ( z1 )
=V01+ e − j βz1 +V01− e + j βz1

Since the load 2Z 0 3 is located at z1 = 0 , we know that the

boundary condition leads to:

V1 ( z1 ) =V01+ (e − j βz + Γ L e + j βz
1 1
)
where:
( 23 ) Z 0 − Z 0
ΓL =
( 23 ) Z 0 + Z 0
( 23 ) − 1
=
( 23 ) + 1
− 13
=
5
3

= −0.2
Therefore:

V1 + ( z1 ) = V01+ e − j βz 1
and V1 − ( z1 ) = V01+ ( −0.2 ) e + j βz1

and thus:
V1 + ( z1 = 0 ) =V01+ e − j β( 0 ) = V01+

V1 − ( z1 = 0 ) = V01+ ( −0.2 ) e + j β( 0 ) = −0.2V01+

We can now determine S11 !

V1 − ( z1 = 0 ) −0.2V01+
S11 = + = = −0.2
V1 ( z1 = 0 ) V01+

Jim Stiles The Univ. of Kansas Dept. of EECS

2/23/2007 Example Determining the Scattering Matrix 4/5

Now its time to find V2− ( z 2 = 0 ) !

Again, since port 2 is terminated, the incident wave on port 2

must be zero, and thus the value of the exiting wave at port 2 is
equal to the total voltage at port 2:

V2− ( z 2 = 0 ) =V2 ( z 2 = 0 )

This total voltage is relatively easy to determine. Examining

the circuit, it is evident that V1 ( z1 = 0 ) = V2 ( z 2 = 0 ) .

+ +
Z0 V1 ( z 1 = 0 ) V2 ( z 2 = 0 ) Z0
2Z0
− −

z1
z 1P = 0 z 2P = 0
Therefore:

V2 ( z 2 = 0 ) =V1 ( z1 = 0 )
(
= V01+ e − j β( 0 ) − 0.2 e + j β( 0 ) )
= V01+ (1 − 0.2 )
= V01+ ( 0.8 )

And thus the scattering parameter S21 is:

V2− ( z 2 = 0 ) 0.8V01+
S21 = + = = 0.8
V1 ( z1 = 0 ) V01+

Jim Stiles The Univ. of Kansas Dept. of EECS

2/23/2007 Example Determining the Scattering Matrix 5/5

Now we just need to find S12 and S22 .

Q: Yikes! This has been an awful lot of work, and you mean that
we are only half-way done!?

A: Actually, we are nearly finished! Note that this circuit is

symmetric—there is really no difference between port 1 and
port 2. If we “flip” the circuit, it remains unchanged!

Z0 2Z0 Z0

z2 z1
z 2P = 0 z 1P = 0

Thus, we can conclude due to this symmetry that:

S11 = S22 = −0.2

and:
S21 = S12 = 0.8

Note this last equation is likewise a result of reciprocity.

Thus, the scattering matrix for this two port network is:

⎡ −0.2 0.8 ⎤
S=⎢
⎣ 0.8 −0.2⎥⎦

Jim Stiles The Univ. of Kansas Dept. of EECS

3/4/2009 Circuit Symmetry present 1/16

Circuit Symmetry
One of the most powerful concepts in for evaluating circuits is
that of symmetry. Normal humans have a conceptual
understanding of symmetry, based on an esthetic perception
of structures and figures.

On the other hand, mathematicians (as they are wont to do) have
defined symmetry in a very precise and unambiguous way. Using a branch
of mathematics called Group Theory, first developed by the young
genius Évariste Galois (1811-1832), symmetry is defined by a set of
operations (a group) that leaves an object unchanged.

Initially, the symmetric “objects” under consideration by Galois were

Évariste Galois polynomial functions, but group theory can likewise be applied to
evaluate the symmetry of structures.

For example, consider an ordinary equilateral triangle; we find that it

is a highly symmetric object!

Jim Stiles The Univ. of Kansas Dept. of EECS

3/4/2009 Circuit Symmetry present 2/16

Q: Obviously this is true. We don’t need a mathematician to tell us that!

A: Yes, but how symmetric is it? How does the symmetry of an equilateral triangle
compare to that of an isosceles triangle, a rectangle, or a square?

To determine its level of symmetry, let’s first label each corner as corner 1, corner 2, and
corner 3.
2

1 3

First, we note that the triangle exhibits a plane of reflection symmetry:

1 3

Jim Stiles The Univ. of Kansas Dept. of EECS

3/4/2009 Circuit Symmetry present 3/16

Thus, if we “reflect” the triangle across this plane we get:

3 1

Note that although corners 1 and 3 have changed places, the triangle itself remains
unchanged—that is, it has the same shape, same size, and same orientation after
reflecting across the symmetric plane!

Mathematicians say that these two triangles are congruent.

Note that we can write this reflection operation as a permutation (an exchange of
position) of the corners, defined as:

1→3
2→2
3→1

Jim Stiles The Univ. of Kansas Dept. of EECS

3/4/2009 Circuit Symmetry present 4/16

Q: But wait! Isn’t there is more than just one plane of reflection symmetry?

A: Definitely! There are two more:

2 1
1→2
2→1
1 3 3→3 2 3

2 3
1→1
2→3
1 3 3→2 1 2

In addition, an equilateral triangle exhibits rotation symmetry! Rotating the triangle 120D
clockwise also results in a congruent triangle:

2 1
1→2
2→3
1 3 3→1 3 2

Jim Stiles The Univ. of Kansas Dept. of EECS

3/4/2009 Circuit Symmetry present 5/16

Likewise, rotating the triangle 120D counter-clockwise results in a congruent triangle:

3
2
1→3
2→1
3→2 2 1
1 3

Additionally, there is one more operation that will result in a congruent triangle—do
nothing!
2 2
1→1
2→2
1 3 3→3 1 3

This seemingly trivial operation is known as the identity operation, and is an element of
every symmetry group.

These 6 operations form the dihedral symmetry group D3 which has order six (i.e., it
consists of six operations). An object that remains congruent when operated on by any
and all of these six operations is said to have D3 symmetry.

An equilateral triangle has D3 symmetry!

Jim Stiles The Univ. of Kansas Dept. of EECS

3/4/2009 Circuit Symmetry present 6/16

By applying a similar analysis to a isosceles triangle, rectangle, and square, we find that:

D1 An isosceles trapezoid has D1 symmetry, a dihedral group of order 2.

D2 A rectangle has D2 symmetry, a dihedral group of order 4.

D4 A square has D4 symmetry, a dihedral group of order 8.

Thus, a square is the most symmetric object of the four we have discussed; the isosceles
trapezoid is the least.

Q: Well that’s all just fascinating—but just what the heck does this have to do with
microwave circuits!?!

A: Plenty! Useful circuits often display high levels of symmetry.

Jim Stiles The Univ. of Kansas Dept. of EECS

3/4/2009 Circuit Symmetry present 7/16

For example, consider these D1 symmetric multi-port circuits:

Port 1
1→2 50Ω Port 2

2→1 200Ω 200Ω

3→4
100Ω
4 →3
Port 3 Port 4

1→3 Port 1 Port 2

50Ω
2→4 100Ω 200Ω
3→1
4 →2 50 Ω

Port 3 Port 4

Jim Stiles The Univ. of Kansas Dept. of EECS

3/4/2009 Circuit Symmetry present 8/16

Or this circuit with D2 symmetry:

Port 1 Port 2
50Ω
200Ω 200Ω

50 Ω

Port 3 Port 4

which is congruent under these permutations:

1→3 1→2 1→4

2→4 2→1 2→3
3→1 3→4 3→2
4 →2 4 →3 4 →1

Jim Stiles The Univ. of Kansas Dept. of EECS

3/4/2009 Circuit Symmetry present 9/16

Or this circuit with D4 symmetry:

Port 1 50Ω Port 2

50 Ω 50 Ω

50 Ω

Port 3 Port 4

which is congruent under these permutations:

1→3 1→2 1→4 1→4 1 →1

2→4 2→1 2→3 2→2 2→3
3→1 3→4 3→2 3→3 3→2
4 →2 4 →3 4 →1 4 →1 4→4

The importance of this can be seen when considering the scattering matrix, impedance
matrix, or admittance matrix of these networks.

Jim Stiles The Univ. of Kansas Dept. of EECS

3/4/2009 Circuit Symmetry present 10/16

For example, consider again this symmetric circuit:

Port 1 Port 2
50Ω
200Ω 200Ω

100Ω

Port 3 Port 4

This four-port network has a single plane of reflection symmetry (i.e., D1 symmetry), and
thus is congruent under the permutation:

1→2
2→1
3→4
4 →3

So, since (for example) 1 → 2 , we find that for this circuit:

S11 = S22 Z 11 = Z 22 Y11 = Y22

must be true!

Jim Stiles The Univ. of Kansas Dept. of EECS

3/4/2009 Circuit Symmetry present 11/16

S13 = S24 Z 13 = Z 24 Y13 = Y24

Or, since 1 → 2 and 3 → 4 we find:
S31 = S42 Z 31 = Z 42 Y31 = Y42

Continuing for all elements of the permutation, we find that for this symmetric circuit,
the scattering matrix must have this form:

⎡ S11 S21 S13 S14 ⎤

⎢S S11 S14 S13 ⎥
S=⎢ 21
⎥
⎢S31 S41 S33 S43 ⎥
⎢S S S S ⎥
⎣ 41 31 43 33 ⎦

and the impedance and admittance matrices would likewise have this same form.

Note there are just 8 independent elements in this matrix. If we also consider
reciprocity (a constraint independent of symmetry) we find that S31 = S13 and S41 = S14 , and
the matrix reduces further to one with just 6 independent elements:

⎡ S11 S21 S31 S41 ⎤

⎢S S11 S41 S31 ⎥
S=⎢ 21
⎥
⎢S31 S41 S33 S43 ⎥
⎢S ⎥
⎣ 41 S31 S43 S33 ⎦

Jim Stiles The Univ. of Kansas Dept. of EECS

3/4/2009 Circuit Symmetry present 12/16

Or, for circuits with this D1 symmetry:

Port 1 Port 2
50Ω
100Ω 200Ω
1→3
2→4
50 Ω
3→1
4 →2 Port 3 Port 4

⎡ S11 S21 S31 S41 ⎤

⎢S S22 S41 S31 ⎥
S=⎢ 21
⎥
⎢S31 S41 S11 S21 ⎥
⎢S ⎥
⎣ 41 S31 S21 S22 ⎦

Q: Interesting. But why do we care?

A: This will greatly simplify the analysis of this symmetric circuit, as we need to
determine only six matrix elements!

Jim Stiles The Univ. of Kansas Dept. of EECS

3/4/2009 Circuit Symmetry present 13/16

For a circuit with D2 symmetry:

Port 1 Port 2
50Ω
200Ω 200Ω

50 Ω

Port 3 Port 4

we find that the impedance (or scattering, or admittance) matrix has the form:

⎡ Z 11 Z 21 Z 31 Z 41 ⎤
⎢Z Z 11 Z 41 Z 31 ⎥
Z = ⎢ 21 ⎥
⎢ Z 31 Z 41 Z 11 Z 21 ⎥
⎢
⎣Z 41 Z 31 Z 21 Z 11 ⎥⎦

Note here that there are just four independent values!

Jim Stiles The Univ. of Kansas Dept. of EECS

3/4/2009 Circuit Symmetry present 14/16

For a circuit with D4 symmetry:

Port 1 50Ω Port 2

50 Ω 50 Ω

50 Ω

Port 3 Port 4

we find that the admittance (or scattering, or impedance) matrix has the form:

⎡Y11 Y21 Y21 Y41 ⎤

⎢Y Y11 Y41 Y21 ⎥
Y=⎢ ⎥
21

⎢Y21 Y41 Y11 Y21 ⎥

⎢
⎣Y41 Y21 Y21 Y11 ⎥⎦

Note here that there are just three independent values!

Jim Stiles The Univ. of Kansas Dept. of EECS

3/4/2009 Circuit Symmetry present 15/16

One more interesting thing (yet another one!); recall that we earlier found that a
matched, lossless, reciprocal 4-port device must have a scattering matrix with one of two
forms:

⎡0 α jβ 0 ⎤
⎢α 0 0 jβ⎥
S=⎢ ⎥ The “symmetric” solution
⎢j β 0 0 α ⎥
⎢ ⎥
⎣0 jβ α 0 ⎦

⎡0 α β 0⎤
⎢α 0 0 −β ⎥
S=⎢ ⎥ The “anti-symmetric” solution
⎢β 0 0 α ⎥
⎢ ⎥
⎣ 0 −β α 0⎦

Compare these to the matrix forms above. The “symmetric solution” has the same form
as the scattering matrix of a circuit with D2 symmetry!

⎡0 α jβ 0 ⎤
⎢α 0 0 jβ⎥
S=⎢ ⎥
⎢j β 0 0 α ⎥
⎢ ⎥
⎣0 jβ α 0 ⎦

Jim Stiles The Univ. of Kansas Dept. of EECS

3/4/2009 Circuit Symmetry present 16/16

Q: Does this mean that a matched, lossless, reciprocal four-port device with the
“symmetric” scattering matrix must exhibit D2 symmetry?

A: That’s exactly what it means!

Not only can we determine from the form of the scattering matrix whether a particular
design is possible (e.g., a matched, lossless, reciprocal 3-port device is impossible), we can
also determine the general structure of a possible solutions (e.g. the circuit must have D2
symmetry).

Likewise, the “anti-symmetric” matched, lossless, reciprocal four-port network must

exhibit D1 symmetry!
⎡0 α β 0 ⎤
⎢α 0 0 − β ⎥
S=⎢ ⎥
⎢β 0 0 α ⎥
⎢ ⎥
⎣ 0 −β α 0 ⎦

We’ll see just what these symmetric, matched, lossless, reciprocal four-port circuits
actually are later in the course!

Jim Stiles The Univ. of Kansas Dept. of EECS

3/6/2009 Example Using Symmetry to Determine S 1/3

Example: Using Symmetry

to Determine a
Scattering Matrix
Say we wish to determine the scattering matrix of the simple
two-port device shown below:

port port
Z0 , β Z0, β Z0 , β
1 2

z = −A z =0

We note that that attaching transmission lines of

characteristic impedance Z 0 to each port of our “circuit”
forms a continuous transmission line of characteristic
impedance Z 0 .

Thus, the voltage all along this transmission line thus has the
form:
V ( z ) =V0+e − j β z + V0−e + j β z

Jim Stiles The Univ. of Kansas Dept. of EECS

3/6/2009 Example Using Symmetry to Determine S 2/3

We begin by defining the location of port 1 as z1P = −A , and

the port location of port 2 as z 2P = 0 :

We can thus conclude:

V1 + ( z ) =V0+e − j β z ( z ≤ −A )

V1 −( z ) =V0−e + j β z ( z ≤ −A )

V2+ ( z ) =V0−e + j β z (z ≥ 0)

V2−( z ) =V0+e − j β z (z ≥ 0)

V1 +( z ) V0+e − j β z V2−( z )

+
V1 −( z ) V e− + j βz V2+( z )
V (z ) 0

z = −A z =0
Say the transmission line on port 2 is terminated in a matched
load. We know that the –z wave must be zero (V0 − = 0 ) , and so
the voltage along the transmission line becomes simply the +z
wave voltage:
V ( z ) =V0+e − j β z
and so:

Jim Stiles The Univ. of Kansas Dept. of EECS

3/6/2009 Example Using Symmetry to Determine S 3/3

V1 + ( z ) =V0+e − j β z V1 −( z ) = 0 ( z ≤ −A )

V2+ ( z ) = 0 V2−( z ) =V0+e − j β z (z ≥ 0)

Now, because port 2 is terminated in a matched load, we can

determine the scattering parameters S11 and S21 :

V1 − ( z = z1P ) V − ( z = −A ) 0
S11 = + = + = =0
V1 ( z = z1P ) V + V ( z = −A ) V + V0+ e − jβ ( − A )
2 =0 2 =0

V2− ( z = z2P ) V2− ( z = 0 ) V0+ e − jβ ( 0 ) 1

S21 = + = + = + − jβ ( − A ) = + jβ A = e − jβ A
V1 ( z = z1P ) V + V1 ( z = −A ) V + V0 e e
2 =0 2 =0

From the symmetry of the structure, we can conclude:

S22 = S11 = 0

And from both reciprocity and symmetry:

S12 = S21 = e − j β A

⎡ 0 e −j βA ⎤
Thus: S= ⎢ −j βA ⎥
⎢⎣e 0 ⎥⎦

Jim Stiles The Univ. of Kansas Dept. of EECS

3/6/2009 Symmetric Circuit Analysis present 1/10

Symmetric Circuit Analysis

200Ω
Consider the following D1 symmetric two-port device:

I1 100Ω 100Ω I2

Q: Yikes! The plane of reflection symmetry

slices through two resistors. What can we do + +
about that? 50Ω V2
-

A: Resistors are easily split into two equal pieces: the 200Ω resistor into two 100Ω
resistors in series, and the 50Ω resistor as two 100 Ω resistors in parallel.
100Ω 100Ω

I1 100Ω 100Ω I2

+ +
V1 100Ω 100Ω V2
- -

Jim Stiles The Univ. of Kansas Dept. of EECS

3/6/2009 Symmetric Circuit Analysis present 2/10

Recall that the symmetry of this 2-port device leads to simplified network matrices:

⎡S11 S21 ⎤ ⎡ Z 11 Z 21 ⎤ ⎡Y11 Y21 ⎤

S= ⎢ Z= ⎢ Y= ⎢
⎥
⎣S21 S11 ⎦ ⎣Z 21 Z 11 ⎥⎦ ⎥
⎣Y21 Y11 ⎦

Q: Yes, but can circuit symmetry likewise simplify the procedure of determining these
elements? In other words, can symmetry be used to simplify circuit analysis?

A: You bet!

First, consider the case where we attach sources to circuit in a way that preserves the
circuit symmetry:

100Ω 100Ω

I1 100Ω 100Ω I2

+ +
+
Vs
+
V1 100Ω 100Ω V2 -
Vs
-
- -

Jim Stiles The Univ. of Kansas Dept. of EECS

3/6/2009 Symmetric Circuit Analysis present 3/10

Or,
100Ω 100Ω

I1 100Ω 100Ω I2

+ +
Is V1 100Ω 100Ω V2 Is
- -

Or, 100Ω 100Ω

I1 100Ω 100Ω I2

Z0 Z0
+ +
Vs
+
V1 100Ω 100Ω V2
+
-
Vs
-
- -

But remember! In order for symmetry to be preserved, the source values on both sides
(i.e, Is,Vs,Z0) must be identical!

Jim Stiles The Univ. of Kansas Dept. of EECS

3/6/2009 Symmetric Circuit Analysis present 4/10

Now, consider the voltages and currents within this circuit under this symmetric
configuration:
I1a I2a

+ V1a - - V2a +
I1 I2b I2
I1b I1d I2d

+ V1b - - V2b +
+ + + +
+
Vs
+
V1 V1c V2c V2 -
Vs
-
- - - -
I1c I2c

Since this circuit possesses bilateral (reflection) symmetry ( 1 → 2, 2 → 1 ), symmetric

currents and voltages must be equal:

I1 = I 2
V1 = V2
I 1a = I 2a
V1a =V2a
I1b = I2b
V1b = V2b
I1c = I2c
V1c =V2c
I1d = I2d
Q: Wait! This can’t possibly be correct! Look at currents I1a and I2a, as well as currents
I1d and I2d. From KCL, this must be true:

I1a = −I2a I1d = −I2d

Jim Stiles The Univ. of Kansas Dept. of EECS

3/6/2009 Symmetric Circuit Analysis present 5/10

Yet you say that this must be true:

I 1 a = I 2a I1d = I2d

There is an obvious contradiction here! There is no way that both sets of equations can
simultaneously be correct, is there?

A: Actually there is! There is one solution that will satisfy both sets of equations:

I 1a = I 2a = 0 I1d = I2d = 0
The currents are zero!

If you think about it, this makes perfect sense! The result says that
no current will flow from one side of the symmetric circuit into the
other.

If current did flow across the symmetry plane, then the circuit
symmetry would be destroyed—one side would effectively become the
“source side”, and the other the “load side” (i.e., the source side
delivers current to the load side).

Thus, no current will flow across the reflection symmetry plane of a symmetric circuit—
the symmetry plane thus acts as a open circuit!

Jim Stiles The Univ. of Kansas Dept. of EECS

3/6/2009 Symmetric Circuit Analysis present 6/10

The plane of symmetry thus becomes a virtual open!

+ V1a - - V2a +
I1 I2b I2
I1b

+ V1b - - V2b +
+ + + +
+
Vs
+
V1 V1c V2c V2 -
Vs
-
- - - -
I1c I2c

Q: So what? Virtual Open

I=0
A: So what! This means that our circuit can be split apart into two separate but
identical circuits. Solve one half-circuit, and you have solved the other!
I1a

I1 = I 2 = Vs 200
+ V1a - V1 = V2 = Vs
I1
I1b I 1a = I 2a = 0
V1a = V2a = 0
I1b = I2b = Vs 200
+ V1b - V1b = V2b = Vs 2
+ + I1c = I 2c = Vs 200
+
Vs - V1 V1c V1c = V2c = Vs 2
- I1d = I 2d = 0
- I1c

Jim Stiles The Univ. of Kansas Dept. of EECS

3/6/2009 Symmetric Circuit Analysis present 7/10

Now, consider another type of symmetry, where the sources are equal but opposite (i.e.,
180 degrees out of phase).
100Ω 100Ω

I1 100Ω 100Ω I2

+ +
+
Vs
+
V1 100Ω 100Ω V2 -
-Vs
-
- -

Or,
100Ω 100Ω

I1 100Ω 100Ω I2

+ +
Is V1 100Ω 100Ω V2 -Is
- -

Jim Stiles The Univ. of Kansas Dept. of EECS

3/6/2009 Symmetric Circuit Analysis present 8/10

Or, 100Ω 100Ω

I1 100Ω 100Ω I2

Z0 Z0
+ +
+
Vs
+
V1 100Ω 100Ω V2 -
-Vs
-
- -

This situation still preserves the symmetry of the circuit—somewhat. The voltages and
currents in the circuit will now posses odd symmetry—they will be equal but opposite
(180 degrees out of phase) at symmetric points across the symmetry plane.

I1a I2a

+ V1a - - V2a +
I1 I2b I2 I1 = −I2
I1b I1d I2d V1 = −V2
I1a = −I2a
+ V1b - - V2b + V1a = −V2a
+ + + + I1b = −I2b
+
V2
+
-Vs V1b = −V2b
Vs - V1 V1c V2c - I1c = −I2c
- - - - V1c = −V2c
I1c I2c I1d = −I2d

Jim Stiles The Univ. of Kansas Dept. of EECS

3/6/2009 Symmetric Circuit Analysis present 9/10

Perhaps it would be easier to redefine the circuit variables as:

I1a I2a

+ V1a - + V2a - I1 = I 2
I1 I2b I2 V1 = V2
I1b I1d I2d
I 1 a = I 2a
V1a =V2a
+ V1b - + V2b - I1b = I2b
+ + - - V1b =V2b
Vs
+
V2
+
-Vs I1c = I2c
- V1 V1c V2c - V1c =V2c
- - + + I1d = I2d
I1c I2c

Q: But wait! Again I see a problem. By KVL it is evident that:

V1c = −V2c

Yet you say that V1c = V2c must be true!

A: Again, the solution to both equations is zero!

V1c =V2c = 0

For the case of odd symmetry, the symmetric plane must be a plane of constant potential
(i.e., constant voltage)—just like a short circuit!

Jim Stiles The Univ. of Kansas Dept. of EECS

3/6/2009 Symmetric Circuit Analysis present 10/10

Thus, for odd symmetry, the symmetric plane forms a virtual short.

I1a I2a

+ V1a - + V2a -
I1 I2b I2
I1b I1d I2d

+ V1b - + V2b -
+ + - -
+
Vs
+
V1 V1c V2c V2 -
-Vs
-
- - + +
I1c I2c

Virtual short
V=0

This greatly simplifies things, as we can again break the circuit into two independent and
(effectively) identical circuits!
I1a

+ V1a - I1 = Vs 50
I1 V1 = Vs
I1b I1d I1a = Vs 100
V1a = Vs
I1b = Vs 100
V1b =Vs
+ V1b - I1c = 0
+ + V1c = 0
+ I1d = Vs 100
Vs - V1 V1c
- -
I1c

Jim Stiles The Univ. of Kansas Dept. of EECS

3/6/2009 Odd Even Mode Analysis present 1/11

Odd/Even Mode Analysis

Q: Although symmetric circuits appear to be plentiful in microwave engineering, it seems
unlikely that we would often encounter symmetric sources . Do virtual shorts and opens
typically ever occur?

A: One word—superposition!

If the elements of our circuit are independent and linear, we can apply superposition to
analyze symmetric circuits when non-symmetric sources are attached.

For example, say we wish to determine the admittance matrix of this circuit. We would
place a voltage source at port 1, and a short circuit at port 2—a set of asymmetric
sources if there ever was one!
100Ω 100Ω

I1 100Ω 100Ω I2

+ +
+
Vs1=Vs
+
- V1 100Ω 100Ω V2 -
Vs2=0
- -

Jim Stiles The Univ. of Kansas Dept. of EECS

3/6/2009 Odd Even Mode Analysis present 2/11

Here’s the really neat part. We find that the source on port 1 can be model as two equal
voltage sources in series, whereas the source at port 2 can be modeled as two equal but
opposite sources in series.

+ Vs
- 2
+
-
Vs
+ Vs
- 2

Vs +
2 -
+
- 0
+
− V2s -

Jim Stiles The Univ. of Kansas Dept. of EECS

3/6/2009 Odd Even Mode Analysis present 3/11

Therefore an equivalent circuit is:

100Ω 100Ω

I1 100Ω 100Ω I2

+ Vs +
Vs 2
- -
2 100Ω 100Ω
+
+ Vs − V2s -
- 2

Now, the above circuit (due to the sources) is obviously asymmetric—no virtual ground,
nor virtual short is present. But, let’s say we turn off (i.e., set to V =0) the bottom
source on each side of the circuit:
100Ω 100Ω

I1 100Ω 100Ω I2

+ Vs +
Vs 2
- -
2 100Ω 100Ω

Jim Stiles The Univ. of Kansas Dept. of EECS

3/6/2009 Odd Even Mode Analysis present 4/11

Our symmetry has been restored! The symmetry plane is a virtual open.

This circuit is referred to as its even mode, and analysis of it is known as the even mode
analysis. The solutions are known as the even mode currents and voltages!
100Ω
Evaluating the resulting even mode half circuit we find:
I1e 100Ω
Vs 1 V
I1e = = s = I2e
2 200 400
+
Vs/2 -
100Ω

Now, let’s turn the bottom sources back on—but turn off the
top two! 100Ω 100Ω

I1 100Ω 100Ω I2

100Ω 100Ω
+
+
-
Vs
2
− V2s -

Jim Stiles The Univ. of Kansas Dept. of EECS

3/6/2009 Odd Even Mode Analysis present 5/11

We now have a circuit with odd symmetry—the symmetry plane is a virtual short!

This circuit is referred to as its odd mode, and analysis of it is known as the odd mode
analysis. The solutions are known as the odd mode currents and voltages!

Evaluating the resulting odd mode half circuit we find:

100Ω
I1o

100Ω
Vs 1 Vs
I1 =
o
= = −I2o
Vs +
2 50 100 2 - 100Ω

Q: But what good is this “even mode” and “odd mode” analysis? After all, the source on
port 1 is Vs1 =Vs, and the source on port 2 is Vs2 =0. What are the currents I1 and I2 for
these sources?

A: Recall that these sources are the sum of the even and odd mode sources:

Vs Vs Vs Vs
Vs 1 =Vs = + Vs 2 = 0 = −
2 2 2 2

and thus—since all the devices in the circuit are linear—we know from superposition that
the currents I1 and I2 are simply the sum of the odd and even mode currents !

Jim Stiles The Univ. of Kansas Dept. of EECS

3/6/2009 Odd Even Mode Analysis present 6/11

I1 = I1e + I1o I2 = I2e + I2o

100Ω 100Ω

I1 = I1o + I1e 100Ω I2 = I2o + I2e

100Ω

+ Vs +
Vs 2
- -
2 100Ω 100Ω
+
+ Vs
2
− V2s -
-

Thus, adding the odd and even mode analysis results together:

I1 = I1e + I1o I2 = I2e + I2o

Vs Vs Vs Vs
= + = −
400 100 400 100
Vs 3V
= =− s
80 400

And then the admittance parameters for this two port network is:

I1 Vs 1 1 I2 3Vs 1 −3
Y11 = = = Y21 = =− =
Vs 1 Vs 2 =0
80 Vs 80 Vs 1 Vs 2 =0
400 Vs 400

Jim Stiles The Univ. of Kansas Dept. of EECS

3/6/2009 Odd Even Mode Analysis present 7/11

And from the symmetry of the device we know:

1 −3
Y22 = Y11 = Y12 = Y21 =
80 400

Thus, the full admittance matrix is:

−3
⎡ 1 80 400 ⎤
Y = ⎢ −3 1 ⎥
⎣ 400 80 ⎦

Q: What happens if both sources are non-zero? Can we use symmetry then?

A: Absolutely! Consider the problem below, where neither source is equal to zero:

100Ω 100Ω

I1 100Ω 100Ω I2

+ +
+
Vs1
+
- V1 100Ω 100Ω V2 -
Vs2
- -

Jim Stiles The Univ. of Kansas Dept. of EECS

3/6/2009 Odd Even Mode Analysis present 8/11

In this case we can define an even mode and an odd mode source as:

Vs 1 +Vs 2 Vs 1 −Vs 2
Vs e = Vs o =
2 2

+
- Vs e
+
-
Vs1 Vs 1 =Vs e +Vs o
+
-
Vs o

Vs e +
-
+
Vs 2 Vs 2 =Vs e −Vs o
-
−Vs o +
-

Jim Stiles The Univ. of Kansas Dept. of EECS

3/6/2009 Odd Even Mode Analysis present 9/11

We then can analyze the even mode circuit:

100Ω 100Ω

I1 100Ω 100Ω I2

+
Vs e Vs e +
-
-
100Ω 100Ω

And then the odd mode circuit: 100Ω 100Ω

I1 100Ω 100Ω I2

100Ω 100Ω
+
+
-
Vs o −Vs o -

And then combine these results in a linear superposition!

Jim Stiles The Univ. of Kansas Dept. of EECS

3/6/2009 Odd Even Mode Analysis present 10/11

Q: What about current sources? Can I likewise consider them to be a sum of an odd
mode source and an even mode source?

A: Yes, but be very careful! The current of two source will add if they are placed in
parallel—not in series! Therefore:

Is 1 + Is 2 Is 1 − Is 2
Ise = Iso =
2 2

Is1 Iso Ise Is 1 = Ise + Iso

Is2 −Iso Ise Is 2 = Ise − Iso

Jim Stiles The Univ. of Kansas Dept. of EECS

3/6/2009 Odd Even Mode Analysis present 11/11

One final word (I promise!) about circuit symmetry and even/odd mode analysis: precisely
the same concept exits in electronic circuit design!

Specifically, the differential (odd) and common (even) mode analysis of

bilaterally symmetric electronic circuits, such as differential amplifiers!

Hi! You might remember differential and common mode

analysis from such classes as “EECS 412- Electronics
II”, or handouts such as “Differential Mode Small-
Signal Analysis of BJT Differential Pairs”

BJT Differential Pair

Differential Mode Common Mode

Jim Stiles The Univ. of Kansas Dept. of EECS

3/6/2009 Example Odd Even Mode Circuit Analysis 1/5

Example: Odd-Even Mode

Circuit Analysis
Carefully (very carefully) consider the symmetric circuit
below.

-
+
4.0 V Z 0 = 50 Ω

50 Ω λ 50 Ω
+
50 Ω 50 Ω
v1
λ -
2

50 Ω Z 0 = 50 Ω 50 Ω

The two transmission lines each have a characteristic

impedance of .

Use odd-even mode analysis to determine the value of

voltage v1 .

Jim Stiles The Univ. of Kansas Dept. of EECS

3/6/2009 Example Odd Even Mode Circuit Analysis 2/5

Solution

To simplify the circuit schematic, we first remove the bottom

(i.e., ground) conductor of each transmission line:

Z 0 = 50 Ω 50 Ω

+
50 Ω λ
4.0 V - +
50 Ω 50 Ω
v1
λ -
2

50 Ω Z 0 = 50 Ω 50 Ω

Note that the circuit has one plane of bilateral symmetry:

50 Ω λ 50 Ω

+
50 Ω 50 Ω
v1
-

50 Ω λ 50 Ω
2

Thus, we can analyze the circuit using even/odd mode analysis

(Yeah!).

Jim Stiles The Univ. of Kansas Dept. of EECS

3/6/2009 Example Odd Even Mode Circuit Analysis 3/5

The even mode circuit is:

50 Ω λ 50 Ω

+ +
2.0 V 2.0 V
- + -
50 Ω 50 Ω v1e
-

50 Ω λ 50 Ω
2

I=0

Whereas the odd mode circuit is:

50 Ω λ 50 Ω

+ +
2.0 V - + -2.0 V -
50 Ω 50 Ω v1o
-

50 Ω λ 50 Ω
2
V =0
We split the modes into half-circuits from which we can
determine voltages v1e and v1o :

Jim Stiles The Univ. of Kansas Dept. of EECS

3/6/2009 Example Odd Even Mode Circuit Analysis 4/5

λ 50 Ω
2
Recall that a A = λ 2 transmission
line terminated in an open
+
+ 2.0 V - circuit has an input impedance
50 Ω
v 1
e
of Z in = ∞ —an open circuit!
-
Likewise, a transmission line A = λ 4
λ
4 50 Ω terminated in an open circuit has an
input impedance of Z in = 0 —a short
circuit! 50 Ω

Therefore, this half-circuit simplifies to: +

And therefore the voltage v1e is easily + - 2.0 V
50 Ω v1e
determined via voltage division:
-
⎛ 50 ⎞
v1e = 2 ⎜ ⎟ = 1.0 V
⎝ 50 + 50 ⎠
50 Ω

Now, examine the right half-circuit of the odd mode:

λ
2 50 Ω
Recall that a A = λ 2 transmission
line terminated in a short
+
-2.0 V circuit has an input impedance
+ -
50 Ω of Z in = 0 — a short circuit!
v1o
-
Likewise, a transmission line A = λ 4
λ 50 Ω terminated in an short circuit has an
4
input impedance of Z in = ∞ — an open
circuit!

Jim Stiles The Univ. of Kansas Dept. of EECS

3/6/2009 Example Odd Even Mode Circuit Analysis 5/5

50 Ω
This half-circuit simplifies to Æ
+
It is apparent from the circuit - -2.0 V
+
that the voltage v1o = 0 ! 50 Ω
v1o
-

50 Ω

Thus, the superposition of the odd and even modes leads to

the result:
v1 = v1e + v1o = 1.0 + 0 = 1.0V

Jim Stiles The Univ. of Kansas Dept. of EECS

3/6/2009 Generalized Scattering Parameters present 1/9

Generalized Scattering Parameters

V2− ( z 2 ) Z02 V2+ ( z 2 )

Consider now this microwave network:
port
z 2 = z 2P
2
Q: Boring! We studied this before; V1 + ( z1 ) port 1 port 3 V3− ( z 3 )
this will lead to the definition of
scattering parameters, right? 4-port
Z01 microwave Z03
A: Not exactly. For this network, device
the characteristic impedance of V1 − ( z1 ) z1 = z1P z3 = z3P V3+ ( z 3 )
each transmission line is different port
(i.e., Z 01 ≠ Z 02 ≠ Z 03 ≠ Z 04 )!
z 4 = z 4P
4

Z04
V4+ ( z 4 ) V4− ( z 4 )

Jim Stiles The Univ. of Kansas Dept. of EECS

3/6/2009 Generalized Scattering Parameters present 2/9

Q: Yikes! You said scattering parameters are dependent on transmission line

characteristic impedance Z 0 . If these values are different for each port, which Z 0 do we
use?

A: For this general case, we must use generalized scattering parameters! First, we
define a slightly new form of complex wave amplitudes:

V0n+ V0n−
an = bn =
Z 0n Z 0n

So for example:
V01+ V03−
a1 = b3 =
Z 01 Z 03

The key things to note are:

A variable a (e.g., a1 , a2 , " ) denotes the complex amplitude of an incident (i.e.,

a
plus) wave.

A variable b (e.g., b1 , b2 , " ) denotes the complex amplitude of an exiting (i.e.,

b
minus) wave.

Jim Stiles The Univ. of Kansas Dept. of EECS

3/6/2009 Generalized Scattering Parameters present 3/9

We now get to rewrite all our transmission line knowledge in terms of

these generalized complex amplitudes!

First, our two propagating wave amplitudes (i.e., plus and minus) are compactly written as:

V0n+ = an Z 0n V0n− = bn Z 0n
And so:

bn + j 2 β zn
Vn + ( zn ) = an Z 0n e − j β zn Vn −( zn ) = bn Z 0n e + j β zn Γ( zn ) = e
an

Likewise, the total voltage, current, and impedance are:

Vn ( zn ) = Z 0n ( an e − j β zn + bn e + j β zn )

an e − j β zn − bn e + j β zn
In ( zn ) =
Z 0n

an e − j β zn + bn e + j β zn
Z ( zn ) =
an e − j β zn − bn e + j β zn

Jim Stiles The Univ. of Kansas Dept. of EECS

3/6/2009 Generalized Scattering Parameters present 4/9

Assuming that our port planes are defined with znP = 0 , we can determine the total
voltage, current, and impedance at port n as:

an − bn
Vn Vn ( zn = 0 ) = Z 0n ( an + bn ) In I n ( z n = 0 ) =
Z 0n

an + bn
Zn Z ( zn = 0 ) =
an − bn

Likewise, the power associated with each wave is:

2 2
V0n+ an
2
V0n− bn
2

Pn =
+
= Pn =
−
=
2Z 0n 2 2Z 0n 2

As such, the power delivered to port n (i.e., the power absorbed by port n) is:

2 2
an − bn
Pn = Pn − Pn =
+ −

Jim Stiles The Univ. of Kansas Dept. of EECS

3/6/2009 Generalized Scattering Parameters present 5/9

This result is also verified:

1
Pn = Re {Vn In∗ }
2
1
{
= Re ( an + bn ) ( an∗ − bn∗ )
2
}
1
= Re {an an∗ + bn an∗ − an bn∗ − bn bn∗ }
2
1
2
{
= Re an + bn an∗ − (bn an∗ ) − bn
2 ∗ 2
}
1
2
{
= Re an + j Im {bn an∗ } − bn
2 2
}
2 2
an − bn
=
2
Q: So what’s the big deal? This is yet another way to express transmission line activity.
Do we really need to know this, or is this simply a strategy for making the next exam even
harder?
a +b
Z1 = 1 1
a1 − b1
A:
You may have noticed that this notation ( an , bn ) provides descriptions that
are a bit “cleaner” and more symmetric between current and voltage.

However, the main reason for this notation is for evaluating the scattering
parameters of a device with dissimilar transmission line impedance (e.g.,
Z 01 ≠ Z 02 ≠ Z 03 ≠ Z 04 ).

Jim Stiles The Univ. of Kansas Dept. of EECS

3/6/2009 Generalized Scattering Parameters present 6/9

For these cases we must use generalized scattering parameters:

V0−m Z 0n
Smn = + (when Vk + ( z k ) = 0 for all k ≠ n )
V0n Z 0m

Note that if the transmission lines at each port are identical ( Z 0m = Z 0n ), the scattering
parameter definition “reverts back” to the original (i.e., Smn = V0−m V0n+ if Z 0m = Z 0n ). E.G.:

V02−
S21 = + when V02+ = 0
V01

V1 + ( z1 ) port 1 port 2 V2−( z 2 )

2-port
Z01 =50 Ω microwave Z02 = 50 Ω
device
V1 −( z1 ) V2+ ( z 2 )

Jim Stiles The Univ. of Kansas Dept. of EECS

3/6/2009 Generalized Scattering Parameters present 7/9

But, if the transmission lines at each port are dissimilar ( Z 0m ≠ Z 0n ), our original
scattering parameter definition is not correct (i.e., Smn ≠ V0−m V0n+ if Z 0m ≠ Z 0n )! E.G.:

V02−
S21 ≠ + when V02+ = 0
V01

V1 + ( z1 ) port 1 port 2 V2−( z 2 )

2-port
Z01 =50 Ω microwave Z02 = 75 Ω
device
V1 −( z1 ) V2+ ( z 2 )

V02− 50
S21 = + when V02+ = 0
V01 75

Note that the generalized scattering parameters can be more compactly written in terms
of our new wave amplitude notation:

V0−m Z 0n bm
Smn = + = (when ak = 0 for all k ≠ n )
V0n Z 0m an

Jim Stiles The Univ. of Kansas Dept. of EECS

3/6/2009 Generalized Scattering Parameters present 8/9

Remember, this is the generalized form of scattering parameter—it always provides the
correct answer, regardless of the values of Z 0m or Z 0n !

Q: But why can’t we define the scattering parameter as Smn = V0−m V0n+ , regardless of Z 0m
or Z 0n ?? Who says we must define it with those awful Z 0n values in there?

A: Good question! Recall that a lossless device is will always have a unitary scattering
matrix. As a result, the scattering parameters of a lossless device will always satisfy, for
example:
M
1 = ∑ Smn
2

m =1

This is true only if the scattering parameters are generalized!

The scattering parameters of a lossless device will form a unitary matrix only if defined
as Smn = bm an . If we use Smn = V0−m V0n+ , the matrix will be unitary only if the connecting
transmission lines have the same characteristic impedance.

Jim Stiles The Univ. of Kansas Dept. of EECS

3/6/2009 Generalized Scattering Parameters present 9/9

Q: Do we really care if the matrix of a lossless device is unitary or not?

A: Absolutely we do! The:

lossless device ⇔ unitary scattering matrix

relationship is a very powerful one. It allows us to identify lossless devices, and it allows
us to determine if specific lossless devices are even possible!

Jim Stiles The Univ. of Kansas Dept. of EECS

3/6/2009 The Scattering Matrix of a Connector 1/4

Example: The Scattering

Matrix of a Connector
First, let’s consider the scattering matrix of a perfect
connector—an electrically very small two-port device that
allows us to connect the ends of different transmission lines
together.
Port Port
I1( z 1 ) 1 2 I 2( z 2 )

+ +
V1( z1 ) Z0 Z0 V2( z 2 )
− −

z1 z2
z1 = 0 z2 = 0

If the connector is ideal, then it will exhibit no series

inductance nor shunt capacitance, and thus from KVL and KCL:

V1( z1 = 0 ) = V2( z2 = 0 ) I1( z1 = 0 ) = −I2( z 2 = 0 )

Terminating port 2 in a matched load, and then analyzing the

resulting circuit, we find that (not surprisingly!):

V01− = 0 and V02− = V01+

Jim Stiles The Univ. of Kansas Dept. of EECS

3/6/2009 The Scattering Matrix of a Connector 2/4

From this we conclude that (since V02+ = 0 ):

V01− 0 V02− V01+

S11 = + = + = 0. 0 S21 = + = + = 1. 0
V01 V01 V01 V01

This two-port device has D2 symmetry (a plane of bilateral

symmetry), meaning:

S22 = S11 = 0. 0 and S21 = S12 = 1. 0

The scattering matrix for such this ideal connector is

therefore:
⎡0 1 ⎤
S= ⎢ ⎥
⎣ 1 0⎦

As a result, the perfect connector allows two transmission

lines of identical characteristic impedance to be connected
together into one “seamless” transmission line.

Z0 Z0

Jim Stiles The Univ. of Kansas Dept. of EECS

3/6/2009 The Scattering Matrix of a Connector 3/4

Now, however, consider the case where the transmission lines

connected together have dissimilar characteristic impedances
(i.e., Z 0 ≠ Z 1 ):

Z 01 Z 02

Port Port
1 2

Q: Won’t the scattering matrix of this ideal connector

remain the same? After all, the device itself has not
changed!

A: The impedance, admittance, and transmission matrix will

remained unchanged—these matrix quantities do not depend
on the characteristics of the transmission lines connected to
the device.

But remember, the scattering matrix depends on both the

device and the characteristic impedance of the transmission
lines attached to it.

After all, the incident and exiting waves are traveling on

these transmission lines!

The ideal connector in this case establishes a “seamless”

interface between two dissimilar transmission lines.

Jim Stiles The Univ. of Kansas Dept. of EECS

3/6/2009 The Scattering Matrix of a Connector 4/4

Z 01 Z 02

Remember, this is the same

structure that we evaluated
in an earlier handout!

In that analysis we found that—when V02+ = 0 :

V01− Z 02 − Z 01 V02− 2Z 02
= and =
V01+ Z 02 + Z 01 V01+ Z 02 + Z 01

And so the (generalized) scattering parameters S11 and S21

are:

V01− Z 01 Z 02 − Z 01 V02− Z 01 2 Z 01Z 02

S11 = + = and S21 = + =
V01 Z 01 Z 02 + Z 01 V01 Z 02 Z 02 + Z 01

As a result we can conclude that the scattering matrix of the

ideal connector (when connecting dissimilar transmission lines)
is:
⎡ Z 02 − Z 01 2 Z 01Z 02 ⎤
⎢ ⎥
Z + Z Z + Z
S= ⎢
02 01 01 02 ⎥
⎢2 Z Z ⎥
⎢ 01 02 Z 01 − Z 02 ⎥
⎢⎣ Z 01 + Z 02 Z 01 + Z 02 ⎥⎦

Jim Stiles The Univ. of Kansas Dept. of EECS