The scattering matrix

Shouri Chatterjee
September 21, 2014

As a student of circuit theory, I have always found the scattering matrix

to be shrouded in mystery. Reading books, unfortunately, did not help me.
This article is a ground-up attempt to uncover the mysteries behind the
scattering matrix, and to provide a physical insight into its importance.
Consider the two-port network shown below. The two ports, 1 and 2,
have voltages of V1 and V2 , and currents of I1 and I2 .

I1 I2
+ +
V1 V2
- -

Figure 1: Two port network

1 Voltage and current waves

In the spirit of transmission-line analysis, let us hypothesize that any voltage
can be declared to be a sum of a forward moving voltage wave, V + , and a
backward moving voltage wave, V . At the same time, the current will be
the sum of a forward moving current wave, V + /Z0 , and a backward moving
current wave, V /Z0 .
For the n-th port of an n-port network, Vn and In can therefore be written
as:

Vn = Vn+ + Vn
In = (Vn+ Vn )/Z0 (1)

1
If we know Vn and In , then given the Z0 , it is possible to compute unique
values for Vn+ and Vn . (A pair of linear simultaneous equations, two un-
knowns.)
For the two-port network shown above, the forward and backward moving
voltage waves at the two ports are related to each other through the scattering
matrix.

Definition 1 (The scattering matrix). The scattering matrix is defined as

the relationship between the forward and backward moving waves. For a
two-port network, like any other set of two-port parameters, the scattering
matrix is a 22 matrix.
  
S11 S12 V1+
 
V1
= (2)
V2 S21 S22 V2+

It is instructive to point out here that the scattering matrix for a two-
port network is always for a specified characteristic impedance, Z0 . In many
cases, the characteristic impedance, Z0 for different ports could be different.
For example, if port-1 has a characteristic impedance of Z01 , and port-2
has a characteristic impedance of Z02 , then all that changes is the pair of
simultaneous equations (1), for the two individual ports.

Corollary 1. The individual S-parameters can be computed from the

definition of the scattering matrix using the following relationship:

Vi 

Sij = +  (3)
Vj V + =0 k6=j
k

2 Termination
Let us consider port-1 of the two-port network. To compute a parameter
such as S22 , V1+ has to be forced to be 0. However, from (1), if V1+ is 0, the
voltage at port-1 is just V1 , and the current is V1 /Z0 . This means that
port-1 is terminated with a resistor of value Z0 .
To force Vk+ to 0, the k-th port has to be terminated with the correspond-
ing characteristic impedance. (In most cases the characteristic impedance is
specified to be the same for all ports. However, in general, each port is
allowed to have its own characteristic impedance.)

2
 Corollary 2. The computation of individual S-parameters in (3) can be
re-written as:
Vi 

Sij = +  (4)
Vj All other ports terminated by Z0

3 Impedance looking into a port

Let us now consider the driving-point impedance looking into a port, say
port-1. If we apply a voltage, V1 , and measure a current, I1 , then the driving-
point impedance looking into port-1 is given by Zin = V1 /I1 .
However, from (1), V1 and I1 are related to the forward and backward
moving voltage waves. Zin is therefore given by:

V1+ + V1
Zin = Z0 (5)
V1+ V1

If port-2 (and all other ports for an n-port network) is terminated by its
characteristic impedance, then V1 would be given by S11 V1+ .

Corollary 3. The driving point impedance looking into port-1, with all
other ports terminated with their characteristic impedances, is given by:
1 + S11
Zin = Z0 (6)
1 S11

A componendo-dividendo operation on (6) will enable us to compute S11

from Zin .
Zin Z0
S11 = (7)
Zin + Z0
where Zin is the driving-point impedance looking into port-1, and Z0 is the
characteristic impedance for port-1. The same result is applicable for any
other port in an n-port network.
The results in (6) and (7) are similar to formulations for transmission
lines, where we obtain the impedance looking into a transmission line when
the reflection coefficient is , or where we obtain the reflection coefficient at
the load-end of the transmission line. After all, S11 is a reflection coefficient.

3
S11 is the reflection coefficient looking into port-1, when the other ports are
terminated with their appropriate characteristic impedances.

4 Transmission
Let us now consider the voltage at port-2, when we apply a voltage at port-1.
Without any loss of generality, let us consider a two-port network. Futher,
let us assume that port-2 is terminated with its characteristic impedance, i.e,
V2+ is 0.
Termination at port-2 implies that V2 is just V2 . V2 /V1+ when port-2 is
terminated with Z0 is the definition for S21 . Therefore, if we apply a forward
moving voltage wave V1+ at port-1, the voltage at port-2 is S21 V1+ . However,
there is also a reflected wave V1 at port-1, given by S11 V1+ . Therefore, the
voltage gain, V2 /V1 is given by:

V2 V V S21
= + 2 = + 2 = (8)
V1 V1 + V1 V1 (1 + S11 ) 1 + S11
If instead, we consider the current gain, I2 /I1 is given by:

I2 V Z0 S21 Z01
= + 2 1 = (9)
I1 V1 V1 Z02 1 S11 Z02
The voltage gain in (8) could be used to compute the value of S21 . If
we terminate port-2 with the appropriate characteristic impedance, apply a
voltage V1 at port-1, and measure a voltage V2 at port-2, then S21 is given
by:
V2
S21 = (1 + S11 ) (10)
V1
For circuits which have a driving-point impedance for port-1 exactly equal
to the corresponding Z0 , i.e, S11 = 0, the voltage gain is equal to S21 .

5 Voltage and current gain

Let us now look at an experiment where the two-port network is being excited
by a voltage source on port-1, with a source-impedance of Z01 . Further, the
load on port-2 is exactly equal to Z02 .
The input impedance of the two-port network is given by (6). This gives
us an input current of:
Vin 1 S11
Iin = I1 = = Vin (11)
Zin + Z01 2Z01

4
 Z01 I1 I2
+ + I +
out
Vin V1 V2 Z02 V
out

- - -

Figure 2: The two-port network being excited by a source of source resistance

Z01 , and loaded with a load resistance Z02 .

Further, we can compute the voltage at port-1:

1 + S11
V1 = I1 Zin = Iin Zin = Vin (12)
2
(8) and (12) can be combined to obtain the overall voltage gain from the
input to the output.
Vout V2 V2 V1 S21
= = = (13)
Vin Vin V1 Vin 2

Corollary 4. When a two-port network is excited on port-1 with a volt-

age source of source-impedance Z01 , and when there is a load of Z02 on
port-2, the voltage gain from the input to the output is directly given by
S21 /2.

The current gain can also be computed similarly. Iout is given by I2 ;

I2 /I1 is given by (9).
Iout I2 S21 Z0
= = 1 (14)
Iin I1 1 S11 Z02

6 Power
With the help of the scattering matrix, we would like to compute the power
delivered to the two-port network, and the power delivered to the load, Z02 .

5
 With the help of (13), the power delivered to the load is given by:1

S21 S |Vin |2
Pout = Vout Iout = Vin Vin 21 = |S21 |2 (15)
2 2Z02 4Z02
|Vin |2 /4Z01 is the maximum available power from the voltage source Vin .
(Recall the maximum power transfer theorem. For completeness, the maxi-
mum power transfer theorem is being discussed in Appendix B.)

Corollary 5. The power delivered to the load, through the two-port net-
work, is given by the maximum available power from the voltage source,
Pmax , times |S21 |2 , times a ratio of the characteristic impedance of the
two ports. That is:
Z0
Pout = Pmax |S21 |2 1 (16)
Z02

Corollary 6. When all the ports have equal characteristic impedance,

the power delivered to the load is simply given by:

Pout = Pmax |S21 |2 (17)

The power delivered to the two-port network, into port-1, is given with
the help of (12) and (11), by:

1 + S11 1 S11 |Vin |2
P1 = V1 I1 = Vin Vin = (1 + 2j Im(S11 ) |S11 |2 ) (18)
2 2Z01 4Z01
P1 includes both real power and reactive power. The real (active) and
reactive powers delivered to the two-port network are given by:
|Vin |2
Re(P1 ) = (1 |S11 |2 ) = Pmax (1 |S11 |2 ) (19)
4Z01
|Vin |2
Im(P1 ) = 2 Im(S11 ) = Pmax 2 Im(S11 ) (20)
4Z01

1
Revelation! All through we have been assuming that all the voltages and currents are
phasors. However, we are also assuming that Z01 and Z02 are real quantities. After all,
these are the characteristic impedances of the two ports, and can be chosen arbitrarily.
These are also the source and load resistances respectively, in our experiment of Fig. 2.
Power in phasors is discussed in Appendix A.

6
 Z0 Z0

|S81 |2 Pmax |S71 |2 Pmax

|S11 |2 Pmax |S61 |2 Pmax
Pmax
Z0 Port 8 Port 7
Port 1 Port 6 Z0

Z0 Port 2 Port 5 Z0

Port 3 Port 4

|S21 |2 Pmax 2 2
|S51 |2 Pmax
|S31 | Pmax |S41 | Pmax

Z0 Z0

Figure 3: Power distribution in an 8-port network. All ports have a charac-

teristic impedance of Z0 .

(19) is significant. It tells us that the real power delivered to the two port
network is (1 |S11 |2 ) times the maximum available power from the source.
If S11 is 0, we get maximum power transfer, and Pmax is delivered to the
two-port network. Otherwise, Pmax |S11 |2 is the power reflected back to the
Z
source. Further, to recapitulate, Pmax |S21 |2 Z001 is the power delivered to the
2
load.
Fig. 3 shows an 8-port network, where all the ports have the same char-
acteristic impedance, Z0 . The scattering matrix of this 8-port network is an
88 matrix. All the ports are terminated in Z0 . Now, let us consider an
experiment where the 8-port network is excited on port-1 with a source of
source impedance Z0 , and maximum available power of Pmax . The scattering
matrix tells us how power is distributed to the 8 ports. That is, a power of
|S11 |2 Pmax is reflected back to the source, a power of |S21 |2 Pmax is delivered
to port-2, a power of |S31 |2 Pmax is delivered to port-3, etc., as shown in the
figure.

7
7 Lossless passive two-port networks
Let us further consider a passive two-port network which has only ideal in-
ductors, capacitors and mutual inductances, with any arbitrary configuration
within it. The components of the two-port network themselves will not con-
sume any real (active) power.
In such a case, the real (active) power delivered to the two-port network
into port-1 will be completely delivered to the load on port-2. From (19) and
(15), this means:
Z01
Pmax (1 |S11 |2 ) = Pmax |S21 |2
Z02

Corollary 7. For a lossless two-port network, therefore, from the above:

Z01
|S11 |2 + |S21 |2 =1 (21)
Z02

Corollary 8. When the two ports of a lossless two-port network have a

uniform characteristic impedance,

|S11 |2 + |S21 |2 = 1 (22)

8 Reciprocity
The reciprocity theorem tells us the following for a passive two-port network:

We apply a voltage, V , at port-1 and measure the short-circuit current

at port-2 to be I.

If we apply the same voltage, V , at port-2, we will find the short-circuit

current at port-1 to be I.

If a two-port network is reciprocal (i.e., passive), the two-port network shown

in Fig. 4 is also reciprocal.
If we apply a voltage V at port-1 of the bigger two-port network, and short
circuit port-2 of the bigger two-port network, then the voltage at port-2 of

8
 Z01 I1 I2 Z02
+ +
V1 V2
- -

Figure 4: Larger two-port network, including the terminations

the smaller two-port network will be given, with the help of (13), as:

V
V2 = S21
2
This gives a short-circuit current at port-2 as:
V
I2 = S21
2Z02

Conversely, if we apply the same voltage at port-2 of the bigger two-port

network and measure the short circuit current at port-1 of the bigger two-
port network, the short circuit current will be given by:
V
I1 = S12
2Z01

By the reciprocity theorem, the short circuit currents for the two experiments
are equal.

Corollary 9. For a reciprocal network:

S21 S12
= (23)
Z02 Z01

If the characteristic impedances for the two ports are equal, S21 = S12 .

9
Appendices
A Power computations with phasors
Let us compute the power consumed by a device, where the current through
the device is given by i(t) = 2 I cos(t + 1 ), and the voltage across the
device is v(t) = 2 V cos(t + 2 ). The phasor corresponding to i(t) is
I 6 1 , or, Iej1 , or, I cos 1 + jI sin 1 . The phasor corresponding to v(t) is
V 6 2 , or, V ej2 , or, V cos 2 + jV sin 2 .
The average power consumed over one period is given by:
Z 2/ Z 2/

P = v(t)i(t)dt = V I cos(t + 1 ) cos(t + 2 )dt
2 0 0
Z 2/

= VI (cos(2t + 1 + 2 ) + cos(2 1 ))dt
2 0
Z 2/

= VI cos(2 1 )dt
2 0
= V I cos(2 1 ) (24)
This also happens to be the real part of the product of the phasor corre-
sponding to v(t) and the conjugate of the phasor corresponding to i(t).

P = V I cos(2 1 ) = Re(V ej2 Iej1 ) = Re(V I ) (25)

P = Re(V I ), is also known as the active power consumed by the
device. The imaginary part, Pr = Im(V I ) is called the reactive power
consumed by the device. If the device is split into a real resistance and a
reactance, the resistance will consume the active power. The reactance will
store energy. The rate of change of stored energy in the reactance will have an
amplitude equal to the reactive power. The total reactive power in a circuit
will be equal to 0, and can be proved with the help of Tellegens theorem.

B Maximum power transfer theorem

The maximum power transfer theorem is often misunderstood. This ap-
pendix is to prove and clarify what the maximum power transfer theorem is
all about.
Fig. 5 shows a voltage source, of source impedance ZS , trying to deliver
power to a load of load impedance ZL . The basic premise of the rest of
the discussion is that the source impedance is fixed. No matter what
you do, you are not allowed to change the source impedance.

10
 ZS
ZL
V

Figure 5: Source trying to deliver power to a load

Theorem 1. A source with source impedance ZS is able to deliver maximum

power to a load when the load impedance, ZL , is equal to ZS .
Proof. The current through the circuit is V /(ZS + ZL ). The voltage across
ZL is V ZL /(ZS + ZL ). Therefore the power delivered to ZL is:
ZL V ZL
PL = V
= |V |2
ZS + ZL ZS + ZL |ZS + ZL |2
If ZS is RS + jXS , and if ZL is RL + jXL , the power delivered to the load
ZL will become:
RL + jXL
PL = |V |2
(RS + RL )2 + (XS + XL )2
The real (active) part of this power is:
RL
Re(PL ) = |V |2 (26)
(RS + RL )2 + (XS + XL )2
If we need maximum power delivered to the ZL , the derivatives of the (26)
with respect to RL and XL will have to be 0. The derivative with respect to
RL being 0 implies:
 
d RL
= 0
dRL (RS + RL )2 + (XS + XL )2
2RL (RS + RL ) (RS + RL )2 = 0
RL = RS (27)
Further, the derivative with respect to XL being 0 implies:
 
d RL
= 0
dXL (RS + RL )2 + (XS + XL )2
2RL (XL + XS ) = 0
XL = XS (28)

11
 (27) and (28) indicate that ZL = RS jXS , or in other words, ZL has to
be ZS .
The maximum power that can be transferred from a source with a real
source resistance of RS , is given by the power delivered to a load of value
RS . If the voltage of the source is V , the maximum power that it can deliver
is therefore:
|V |2
Pmax = (29)
4RS
In many situations, a source of power is defined by the maximum power
it can deliver. Instead of specifying its voltage (or current) and source
impedance, often one can specify the maximum power that it can deliver,
or the maximum available power.

12