Berkeley

Scattering Parameters

Prof. Ali M. Niknejad

U.C. Berkeley
Copyright © 2023 by Ali M. Niknejad

February 6, 2025 1 / 81
Scattering Parameters

2 / 81
Scattering Matrix

Voltages and currents are difficult to measure directly at microwave freq. Z

matrix requires “opens”, and it’s hard to create an ideal open (parasitic
capacitance and radiation). Likewise, a Y matrix requires “shorts”, again ideal
shorts are impossible at high frequency due to the finite inductance.
Many active devices could oscillate under the open or short termination.
S parameters are easier to measure at high frequency. The measurement is direct
and only involves measurement of relative quantities (such as the SWR or the
location of the first minima relative to the load).

3 / 81
S-Parameters
V1+

1
V1−

V3−
3

2 V3+
V2−

V2+

Scattering parameters represent the flow of power into and out of ports of an
arbitrary N-port
It’s important to realize that although we associate S parameters with high
frequency and wave propagation, the concept is valid for any frequency.

4 / 81
Power Flow in an One-Port
We begin with the simple observation that the power flow into a one-port circuit
can be written in the following form
Pin = Pavs − Pr

where Pavs is the available power from the source. Unless otherwise stated, let us
assume sinusoidal steady-state. If the source has a real resistance of Z0 , this is
simply given by
V2
Pavs = s
8Z0

Of course if the one-port is conjugately matched to the source, then it will draw
the maximal available power from the source. Otherwise, the power Pin is always
less than Pavs , which is reflected in our equation. In general, Pr represents the
wasted or untapped power that one-port circuit is “reflecting” back to the source
due to a mismatch. For passive circuits it’s clear that each term in the equation is
positive and Pin ≥ 0.
5 / 81
Power Absorbed by One-Port
The complex power absorbed by the one-port is given by
1
Pin = (V1 · I1∗ + V1∗ · I1 )
2

which allows us to write

Vs2 1
Pr = Pavs − Pin = − (V1 I1∗ + V1∗ I1 )
4Z0 2

the factor of 4 instead of 8 is used since we are now dealing with complex power.
The average power can be obtained by taking one half of the real component of
the complex power. If the one-port has an input impedance of Zin , then the power
Pin is expanded to
 
1 Zin Vs∗ Zin∗ ∗ Vs
Pin = Vs · + V ·
2 Zin + Z0 (Zin + Z0 )∗ (Zin + Z0 )∗ s (Zin + Z0 )

6 / 81
(cont.)
The previous equation is easily simplified to (where we have assumed Z0 is real)
 
|Vs |2 Z0 Zin + Zin∗ Z0
Pin =
2Z0 |Zin + Z0 |2

With the exception of a factor of 2, the premultiplier is simply the source available
power, which means that our overall expression for the reflected power is given by
 
Vs2 Z0 Zin + Zin∗ Z0
Pr = 1−2
4Z0 |Zin + Z0 |2

which can be simplified

2
Zin − Z0
Pr = Pavs = Pavs |Γ|2
Zin + Z0

7 / 81
Definition of Reflection Coefficient

2
Zin − Z0
Pr = Pavs = Pavs |Γ|2
Zin + Z0

We have defined Γ, or the reflection coefficient, as

Zin − Z0
Γ=
Zin + Z0

From the definition it is clear that |Γ| ≤ 1, which is just a re-statement of the
conservation of energy implied by our assumption of a passive load.
This constant Γ, also called the scattering parameter of a one-port, plays a very
important role. On one hand we see that it is has a one-to-one relationship with
Zin .
8 / 81
Scattering Parameter

Given Γ we can solve for Zin by inverting the above equation

1+Γ
Zin = Z0
1−Γ
which means that all of the information in Zin is also in Γ. Moreover, since
|Γ| < 1, we see that the space of the semi-infinite space of all impedance values
with real positive components (the right-half plane) maps into the unit circle.
This is a great compression of information which allows us to visualize the entire
space of realizable impedance values by simply observing the unit circle. We shall
find wide application for this concept when finding the appropriate load/source
impedance for an amplifier to meet a given noise or gain specification.

9 / 81
Scattering Parameter as Power Flow

More importantly, Γ expresses very direct and obviously the power flow in the
circuit. If Γ = 0, then the one-port is absorbing all the possible power available
from the source. If |Γ| = 1 then the one-port is not absorbing any power, but
rather “reflecting” the power back to the source. Clearly an open circuit, short
circuit, or a reactive load cannot absorb net power. For an open and short load,
this is obvious from the definition of Γ. For a reactive load, this is pretty clear if
we substitute Zin = jX
q
jX − Z0 X 2 + Z02
|ΓX | = = q =1
jX + Z0 X2 + Z2 0

10 / 81
Relation between Z and Γ
The transformation between impedance and Γ is the well known Bilinear
Transform. It is a conformal mapping (meaning that it preserves angles) from
vertical and horizontal lines into circles. We have already seen that the jX axis is
mapped onto the unit circle.
Since |Γ|2 represents power flow, we may imagine that Γ should represent the flow
of voltage, current, or some linear combination thereof. Consider taking the
square root of the basic equation we have derived
p p
Pr = Γ Pavs

where we have retained the positive root. We may write the above equation as

b1 = Γa1

where a and b have the units of square root of power and represent signal flow in
the network. How are a and b related to currents and voltage?
11 / 81
Definition of a and b
Let
V1 + Z0 I1
a1 = √
2 Z0
and
V1 − Z0 I1
b1 = √
2 Z0
It is now easy to show that for the one-port circuit, these relations indeed
represent the available and reflected power:

|V1 |2 Z0 |I1 |2 V1∗ · I1 + V1 · I1∗

|a1 |2 = + +
4Z0 4 4
Now substitute V1 = Zin Vs /(Zin + Z0 ) and I1 = Vs /(Zin + Z0 ) we have

|Vs |2 |Zin |2 Z0 |Vs |2 |Vs |2 Zin∗ Z0 + Zin Z0

|a1 |2 = + +
4Z0 |Zin + Z0 |2 4|Zin + Z0 |2 4Z0 |Zin + Z0 |2

12 / 81
a/b and Power Flow
We have now shown that a1 is associated with the power available from the
source:  
2 |Vs |2 |Zin |2 + Z02 + Zin∗ Z0 + Zin Z0
|a1 | =
4Z0 |Zin + Z0 |2
 
|Vs |2 |Zin + Z0 |2
= = Pavs
4Z0 |Zin + Z0 |2

In a like manner, the square of b is given by many similar terms

 
2 |Vs |2 |Zin |2 + Z02 − Zin∗ Z0 − Zin Z0
|b1 | = =
4Z0 |Zin + Z0 |2
2
|Zin − Z0
Pavs = Pavs |Γ|2
Zin + Z0

= |a1 |2 |Γ|2
as expected.
13 / 81
One-Port Equation

We can now see that the expression b = Γ · a is analogous to the expression

V = Z · I or I = Y · V and so it can be generalized to an N-port circuit. In fact,
since a and b are linear combinations of v and i, there is a one-to-one relationship
between the two. Taking the sum and difference of a and b we arrive at
2V1 V1
a1 + b1 = √ = √
2 Z0 Z0
which is related to the port voltage and
2Z0 I1 p
a1 − b1 = √ = Z0 I1
2 Z0
which is related to the port current.

14 / 81
Incident and Scattered Waves

15 / 81
Incident and Scattered Waves

Let’s define the vector v + as the incident “forward” waves on each transmission
line connected to the N port. Define the reference plane as the point where the
transmission line terminates onto the N port.
The vector v − is then
the+ 
reflected or “scattered” waveform
 −  at the location of the
V1 V1
V +  V − 
 2   2 
port. v + = V +  v − = V − 
 3   3 
.. ..
. .

16 / 81
Scattering Waves (cont)

Because the N port is linear, we expect that scattered field to be a linear function
of the incident field
v − = Sv +

S is the scattering matrix

 
S11 S12 · · ·
 .. 
S =
 S 21 . 

..
.

17 / 81
Relation to Voltages

The fact that the S matrix exists can be easily proved if we recall that the voltage
and current on each transmission line termination can be written as
Vi = Vi+ + Vi− Ii = Y0 (Ii+ − Ii− )

Inverting these equations

Vi + Z0 Ii = Vi+ + Vi− + Vi+ − Vi− = 2Vi+

Vi − Z0 Ii = Vi+ + Vi− − Vi+ + Vi− = 2Vi−

Thus v + ,v − are simply linear combinations of the port voltages and currents. By
the uniqueness theorem, then, v − = Sv + .

18 / 81
Measure Sij

1
V1+
4 Z0
The term Sij can be computed directly by the
V1− following formula
V2−
Z0 2 5 Z0 Vi−
Sij =
Vj+
V3− Vk+ =0 ∀ k̸=j
Z0 3 6 Z0

In other words, to measure Sij , drive port j with a wave amplitude of Vj+ and
terminate all other ports with the characteristic impedance of the lines (so that
Vk+ = 0 for k ̸= j). Then observe the wave amplitude coming out of the port i

19 / 81
Termination
It’s important to realize that our definition of scattering parameters is
independent of transmission lines and can be defined completely in terms of
voltage and currents.
So then why do we terminate the line? Because to make V + = 0 we solve:
1
V + = (Vi + Z0 Ii ) = 0
2

Solving this equation we find the conditions

Vi = −Z0 Ii
or
Vi
ZL = = Z0
−Ii

We see that there are no transmission lines to terminate, this rather follows from
the definition of scattering parameters.
20 / 81
S Matrix for a 1-Port Capacitor

Note you can solve this problem with T-lines or by definition of the S matrix:

Let’s calculate the S parameter for a

capacitor
Z0 C V−
S11 = 1+
V1

This is of course just the reflection coefficient for a capacitor

1
ZC − Z0 jωC − Z0
S11 = ρL = = 1
ZC + Z0 jωC + Z0

1 − jωCZ0
=
1 + jωCZ0

21 / 81
S Matrix for a 1-Port Cap (cont)
Let’s calculate the S parameter for a capacitor directly from the definition of S
parameters

V1−
S11 =
C V1+

Substituting for the current in a capacitor

V1− = V − IZ0 = V − jωCV = V (1 − jωCZ0 )

V1+ = V + IZ0 = V + jωCV = V (1 + jωCZ0 )

We arrive at the same answer as expected

1 − jωCZ0
=
1 + jωCZ0
22 / 81
S Matrix for a 2-Port Shunt Element
Consider a shunt impedance connected at the junction of two transmission lines.
The voltage at the junction is of course continuous. The currents, though, differ
V1 = V2

Z0 ZL Z0
I1 + I2 = YL V2

To compute S11 , enforce V2+ = 0 by terminating the line. Thus we can be

re-write the above equations

V1+ + V1− = V2−

Y0 (V1+ − V1− ) = Y0 V2− + YL V2− = (YL + Y0 )V2−

23 / 81
Shunt Element (cont)
We can now solve the above eq. for the reflected and transmitted wave
Y0
V1− = V2− − V1+ = (V + − V1− ) − V1+
YL + Y0 1

V1− (YL + Y0 + Y0 ) = (Y0 − (Y0 + YL ))V1+

V1− Y0 − (Y0 + YL ) Z0 ||ZL − Z0

S11 = + = Y + (Y + Y ) = Z ||Z + Z
V1 0 L 0 0 L 0

The above eq. can be written by inspection since Z0 ||ZL is the effective load seen
at the junction of port 1.
Thus for port 2 we can write
Z0 ||ZL − Z0
S22 =
Z0 ||ZL + Z0

24 / 81
Shunt Element (cont)
Likewise, we can solve for the transmitted wave, or the wave scattered into port 2

V2−
S21 =
V1+

Since V2− = V1+ + V1− , we have

2Z0 ||ZL
S21 = 1 + S11 =
Z0 ||ZL + Z0

By symmetry, we can deduce S12 as

2Z0 ||ZL
S12 =
Z0 ||ZL + Z0

25 / 81
Conversion Formula
Since V + and V − are related to V and I , it’s easy to find a formula to convert
for Z or Y to S
Vi = Vi+ + Vi− → v = v + + v −

Zi0 Ii = Vi+ − Vi− → Z0 i = v + − v −

Now starting with v = Zi, we have

v + + v − = ZZ0−1 (v + − v − )

Note that Z0 is the scalar port impedance

v − (I + ZZ0−1 ) = (ZZ0−1 − I )v +

v − = (I + ZZ0−1 )−1 (ZZ0−1 − I )v + = Sv +

26 / 81
Conversion (cont)
We now have a formula relating the Z matrix to the S matrix
S = (ZZ0−1 + I )−1 (ZZ0−1 − I ) = (Z + Z0 I )−1 (Z − Z0 I )

Recall that the reflection coefficient for a load is given by the same equation!
Z /Z0 − 1
ρ=
Z /Z0 + 1

To solve for Z in terms of S, simply invert the relation

Z0−1 ZS + IS = Z0−1 Z − I

Z0−1 Z (I − S) = S + I Z = Z0 (I + S)(I − S)−1

As expected, these equations degenerate into the correct form for a 1 × 1 system
1+S11
Z11 = Z0 1−S 11
27 / 81
Properties of S-Parameters

28 / 81
Shift in Reference Planes
Note that if we move the reference planes, we can easily recalculate the S
parameters.
We’ll derive a new matrix S ′ related to S. Let’s call the waves at the new
reference ν
v − = Sv +

ν − = S ′ν +

Since the waves on the lossless transmission lines only experience a phase shift, we
have a phase shift of θi = βi ℓi
νi− = v − e −jθi

νi+ = v + e jθi

29 / 81
Reference Plane (cont)

Or we have
 jθ   −jθ 
e 1 0 ··· e 1 0 ···
 0 e jθ2 ···   0 e −jθ2 ··· 
  −   +
 0 0 e jθ3 
· · · ν = S  0 0 e −jθ3 · · · ν
  
.. ..
. .

So we see that the new S matrix is simply

 −jθ   −jθ 
e 1 0 ··· e 1 0 ···
 0 e −jθ ···   e −jθ2 ··· 
  0
2
 
S′ =  0 0 e −jθ3 · · · S  0 0 e −jθ3 · · ·
   
.. ..
. .

30 / 81
Normalized S-Parameters
a1 b2
b1 [S] a2

Let’s introduce normalized voltage waves

v + (x) v − (x)
a(x) = √ b(x) = √
Z0 Z0

So now |a|2 and |b|2 represent the power of the forward and reverse wave. Define
the scattering matrix as before
b = Sa

For a 2 × 2 system, this is simply

    
b1 S11 S12 a1
=
b2 S21 S22 a2
31 / 81
Generalized Scattering Parameters
We can use different impedances Z0,n at each port and so we have the generalized
incident and reflected waves
v+ v−
an = p n bn = p n
Z0,n Z0,n

The scattering parameters are now given by

p
bi Vi− Z0,j
Sij = Sij = + p
aj ak̸=j =0 Vj Z0,i
Vk̸+=j =0

Consider the current and voltage in terms of a and b

p
Vn = vn+ + vn− = Z0,n (an + bn )

1
1
In = vn− − vn− = p (an − bn )
Z0,n Z0,n

32 / 81
(cont)

The power flowing into this port is given by

1 1
1

ℜ (Vn In∗ ) = ℜ |an |2 − |bn |2 + (bn an∗ − bn∗ an ) = |an |2 − |bn |2
2 2 2

33 / 81
Scattering Transfer Parameters
b2 a3
a1 b4
b1 [T] [T] a4
a2 b3

Up to now we found it convenient to represent the scattered waves in terms of the

incident waves. But what if we wish to cascade two ports as shown?
Since b2 flows into a1′ , and likewise b1′ flows into a2 , would it not be convenient if
we defined the a relationship between a1 ,b1 and b2 ,a2 ?
In other words we have     
a1 T11 T12 b2
=
b1 T21 T22 a2

Notice carefully the order of waves (a,b) in reference to the figure above. This
allows us to cascade matrices
       
a1 b2 a3 b
= T1 = T1 = T1 T2 4
b1 a2 b3 a4

34 / 81
Reciprocal Networks

35 / 81
Reciprocal Networks
Suppose the Z /Y matrix are symmetric. Now let’s see what we can infer about
the S matrix.
1
v + = (v + Z0 i)
2
1
v − = (v − Z0 i)
2

Substitute v = Zi in the above equations

1 1
v + = (Zi + Z0 i) = (Z + Z0 )i
2 2
1 1
v − = (Zi − Z0 i) = (Z − Z0 )i
2 2

Since i = i, the above eq. must result in consistent values of i. Or

2(Z + Z0 )−1 v + = 2(Z − Z0 )−1 v −
36 / 81
Reciprocal Networks (cont)
From the above, we have
S = (Z − Z0 )(Z + Z0 )−1

Consider the transpose of the S matrix

t
S t = (Z + Z0 )−1 (Z − Z0 )t

Recall that Z0 is a diagonal matrix

S t = (Z t + Z0 )−1 (Z t − Z0 )

If Z t = Z (reciprocal network), then we have

S t = (Z + Z0 )−1 (Z − Z0 )

37 / 81
(cont)
Previously we found that
S = (Z + Z0 )−1 (Z − Z0 )

So that we see that the S matrix is also symmetric (under reciprocity)S t = S

Note that in effect we have shown that
(Z + I )−1 (Z − I ) = (Z − I )(Z + I )−1

This is easy to demonstrate if we note that

Z 2 − I = Z 2 − I 2 = (Z + I )(Z − I ) = (Z − I )(Z + I )

In general matrix multiplication does not commute, but here it does

(Z − I ) = (Z + I )(Z − I )(Z + I )−1 (Z + I )−1 (Z − I ) = (Z − I )(Z + I )−1

38 / 81
S-Parameters of a Lossless Network
Consider the total power dissipated by a network (must sum to zero)

1

Pav = ℜ v t i ∗ = 0
2

Expanding in terms of the wave amplitudes

1

= ℜ (v + + v − )t Z0−1 (v + − v − )∗
2

Where we assume that Z0 are real numbers and equal. The notation is about to
get ugly
1  
t ∗ t ∗ t ∗ t ∗
= ℜ v+ v+ − v+ v− + v− v+ − v− v−
2Z0

39 / 81
Lossless (cont)

Notice that the middle terms sum to a purely imaginary number. Let x = v + and
y = v−

y t x ∗ − x t y ∗ = y1 x1∗ + y2 x2∗ + · · · − x1 y1∗ + x2 y2∗ + · · · = a − a∗

We have shown that

 
1  t t ∗ =0
Pav = v| +{zv +} − v| −{zv −}
2Z0
total incident power total reflected power

40 / 81
(cont)

This is a rather obvious result. It simply says that the incident power is equal to
the reflected power (because the N port is lossless). Since v − = Sv +
t ∗ t ∗
v + v + = (Sv + )t (Sv + )∗ = v + S t S ∗ v +

This can only be true if S is a unitary matrix

StS∗ = I

S ∗ = (S t )−1

41 / 81
Orthogonal Properties of S
Expanding out the matrix product
X X
δij = (S t )ik Skj∗ = Ski Skj∗
k k

For i = j we have X
Ski Ski∗ = 1
k

For i ̸= j we have X
Ski Skj∗ = 0
k

The dot product of any column of S with the conjugate of that column is unity
while the dot product of any column with the conjugate of a different column is
zero. If the network is reciprocal, then S t = S and the same applies to the rows of
S.
Note also that |Sij | ≤ 1.
42 / 81
S-Parameter Representation of a Source

43 / 81
Representation of Source
ZS IS

+
Vi = Vs − Is Zs
VS Vi
−
The voltage source can be represented directly for s-parameter analysis as follows.
First note that  + 
+ − Vi Vi−
Vi + Vi = Vs + − Zs
Z0 Z0

Solve these equations for Vi− , the power flowing away from the source
Zs − Z0 Z0
Vi− = Vi+ + Vs
Zs + Z0 Z0 + Zs
√
Dividing each term by Z0 , we have
√ p
Vi− Vi+ Z0 bi = ai Γs + bs bs = Vs Z0 /(Z0 +Zs )
√ = √ Γs + Vs
Z0 Z0 Z0 + Zs
44 / 81
Available Power from Source
A useful quantity is the available power from a source under conjugate matched
conditions. Since
Pavs = |bi |2 − |ai |2

If we let ΓL = Γ∗S , then using ai = ΓL bi , we have

bi = bs + ai ΓS = bs + Γ∗S bi ΓS

Solving for bi we have

bs
bi =
1 − |ΓS |2

So the Pavs is given by  

2 2 2 1 − |ΓS |2
Pavs = |bi | − |ai | = |bs |
(1 − |ΓS |2 )2
|bs |2
=
1 − |Γ |2 45 / 81
Signal Flow Analysis

46 / 81
Signal-Flow Analysis

a1 S21 b2

S11 S22

b1 S12 a2

Each signal a and b in the system is represented by a node. Branches connect

nodes with “strength” given by the scattering parameter. For example, a general
two-port is represented above.
Using three simple rules, we can simplify signal flow graphs to the point that
detailed calculations are done by inspection. Of course we can always “do the
math” using algebra, so pick the technique that you like best.

47 / 81
Series and Parallel Rules

SA SB SA SB

a1 a2 a3 a1 a3

Rule 1: (series rule) By inspection, we have the cascade.

SA SA + SB

a1 a2 a1 a2

SB

Rule 2: (parallel rule) Clear by inspection.

48 / 81
Self-Loop Rule
SB

SA SC

a1 a2 a3
SA
1 − SB SC

a1 a2 a3

Rule 3: (self-loop rule) We can remove a “self-loop” by multiplying branches

feeding the node by 1/(1 − SB ) since
a2 = SA a1 + SB a2

a2 (1 − SB ) = SA a1

SA
a2 = a1
1 − SB
49 / 81
Splitting Rule

SB a1 SA
a2 a2

a1 a3 a3
SA SB

SC SA
a4 SC a4
a!2

We can duplicate node a2 by splitting the signals at an earlier phase

50 / 81
Example: Signal Flow Analysis
S22 ΓL
a1 S21 b2
a1 S21

S11 S22 ΓL b2
S11 ΓL

b1 S12 a2
b1 S12 a2

Using the above rules, we can calculate the input reflection coefficient of a
two-port terminated by ΓL = b1 /a1 using a couple of steps.
First we notice that there is a self-loop around b2 .
S21
a1 1 − S22 ΓL

b2
S11 ΓL

b1 S12 a2

Next we remove the self loop and from here it’s clear that the
b1 S21 S12 ΓL
Γin = = S11 +
a1 1 − S22 ΓL
51 / 81
Mason’s Rule
 P1 a1 S21 b2 P2
bS

ΓS S11 S22 ΓL

b1 S12 a2

Using Mason’s Rule, you can calculate the transfer function for a signal flow
graph by “inspection”
P P
P

P1 1 − L(1)(1) + L(2)(1) − · · · + P2 1 − L(1)(2) + · · · + · · ·
T = P P P
1 − L(1) + L(2) − L(3) + · · ·

Each Pi defines a path, a directed route from the input to the output not
containing each node more than once. The value of Pi is the product of the
branch coefficients along the path.
For instance the path from bs to b1 (T = b1 /bs ) has two paths, P1 = S11 and
P2 = S21 ΓL S12 52 / 81
Loop of Order Summation Notation
 a1 S21 b2
bS
P
ΓS S22 ΓL
The notation L(1) is the sum
S11
over all first order loops.
b1 S12 a2

A “first order loop” is defined as product of the branch values in a loop in the
graph. For the given example we have Γs S11 , S22 ΓL , and Γs S21 ΓL S12 .
A “second order loop” L(2) is the product of two non-touching first-order loops.
For instance, since loops S11 Γs and S22 ΓL do not touch, their product is a second
order loop.
A “third order loop” L(3) is likewise the product of three non-touching first order
loops.
P
The notation L(1)(p) is the P
sum of all first-order loops that do not touch the
path
P p. (2)For path P1 , we have L(1)(1) = ΓL S22 but for path P2 we have
L(1) = 0.
53 / 81
Example: Input Reflection of Two-Port
a1 S21 b2

S11 S22 ΓL

b1 S12 a2

Using Mason’s rule, you can quickly identify the relevant paths for a Γin = b1 /a1 .
There are two paths P1 = S11 and P2 = S21 ΓL S12
P
There is only one first-order loop: L(1) = S22 ΓL and so naturally there are no
higher order loops.
P
Note that the loop does not touch path P1 , so L(1)(1) = S22 ΓL .
Now let’s apply Mason’s general formula
S11 (1 − S22 ΓL ) + S21 ΓL S12 S21 ΓL S12
Γin = = S11 +
1 − S22 ΓL 1 − S22 ΓL

54 / 81
Example: Transducer Power Gain
 a1 S21 b2
bS

ΓS S11 S22 ΓL

b1 S12 a2

By definition, the transducer power gain is given by

2
PL |b2 |2 (1 − |ΓL |2 ) b2
GT = = |b | 2 = (1 − |ΓL |2 )(1 − |ΓS |2 )
PAVS s
2
bS
1−|ΓS |

By Mason’s Rule, there is only one path P1 = S21 from bS to b2 so we have

X
L(1) = ΓS S11 + S22 ΓL + ΓS S21 ΓL S12

X X
L(2) = ΓS S11 ΓL S22 L(1)(1) = 0

55 / 81
Transducer Gain (cont)
The gain expression is thus given by
b2 S21 (1 − 0)
=
bS 1 − ΓS S11 − S22 ΓL − ΓS S21 ΓL S12 + ΓS S11 ΓL S22

The denominator is in the form of 1 − x − y + xy which allows us to write

|S21 |2 (1 − |ΓS |2 )(1 − |ΓL |2 )
GT =
|(1 − S11 ΓS )(1 − S22 ΓL ) − S21 S12 ΓL ΓS |2

Recall that Γin = S11 + S21 S12 ΓL /(1 − S22 ΓL ). Factoring out 1 − S22 ΓL from the
denominator we have
 
S21 S12 ΓL
den = 1 − S11 ΓS − ΓS (1 − S22 ΓL )
1 − S22 ΓL
  
S21 S12 ΓL
den = 1 − ΓS S11 + (1 − S22 ΓL )
1 − S22 ΓL

= (1 − ΓS Γin )(1 − S22 ΓL ) 56 / 81

Transducer Gain Expression

This simplifications allows us to write the transducer gain in the following

convenient form
1 − |ΓS |2 2 1 − |ΓL |
2
GT = |S21 |
|1 − Γin ΓS |2 |1 − S22 ΓL |2

Which can be viewed as a product of the action of the input match “gain”, the
intrinsic two-port gain |S21 |2 , and the output match “gain”. Since the general
two-port is not unilateral, the input match is a function of the load.
Likewise, by symmetry we can also factor the expression to obtain

1 − |ΓS |2 2 1 − |ΓL |
2
GT = |S21 |
|1 − S11 ΓS |2 |1 − Γout ΓL |2

57 / 81
Stability From Another Perspective
We can also derive stability in terms of the input reflection coefficient. For a
general two-port with load ΓL we have
v2− = Γ−1 + + +
L v2 = S21 v1 + S22 v2

S21
v2+ = v−
Γ−1
L − S22 1
 
S12 S21 ΓL
v1− = S11 + v1+
1 − ΓL S22
S12 S21 ΓL
Γ = S11 +
1 − ΓL S22

If |Γ| < 1 for all ΓL , then the two-port is stable

S11 (1 − S22 ΓL ) + S12 S21 ΓL S11 + ΓL (S21 S12 − S11 S22 )
Γ= =
1 − S22 ΓL 1 − S22 ΓL
S11 − ∆ΓL 58 / 81
Stability Circle
To find the boundary between stability/instability, let’s set |Γ| = 1

S11 − ∆ΓL
=1
1 − S22 ΓL

|S11 − ∆ΓL | = |1 − S22 ΓL |

After some algebraic manipulations, we arrive at the following equation

∗ − ∆∗ S
S22 |S12 S21 |
11
ΓL − 2 2
=
|S22 | − |∆| |S22 |2 − |∆|2

This is of course an equation of a circle, |ΓL − C | = R, in the complex plane with

center at C and radius R
Thus a circle on the Smith Chart divides the region of instability from stability.
59 / 81
Example: Stability Circle

RS
In this example, the origin of the
circle lies outside the stability circle
CS
but a portion of the circle falls
re gion
n
re

inside the unit circle. Is the region

gio

le
tab
uns table
s

of stability inside the circle or

outside?
|S11 | < 1

This is easily determined if we note

that if ΓL = 0, then Γ = S11 . So if
S11 < 1, the origin should be in the
stable region. Otherwise, if S11 > 1,
the origin should be in the unstable
region.

60 / 81
Stability: Unilateral Case
Consider the stability circle for a unilateral two-port
∗ − (S ∗ S ∗ )S
S11 ∗
S11
11 22 22
CS = =
|S11 |2 − |S11 S22 |2 |S11 |2

1
RS = 0 |CS | =
|S11 |

The cetner of the circle lies outside of the unit circle if |S11 | < 1. The same is
true of the load stability circle. Since the radius is zero, stability is only
determined by the location of the center.
If S12 = 0, then the two-port is unconditionally stable if S11 < 1 and S22 < 1.
This result is trivial since
ΓS |S12 =0 = S11

The stability of the source depends only on the device and not on the load.
61 / 81
Mu Stability Test
If we want to determine if a two-port is unconditionally stable, then we should use
the µ test
1 − |S11 |2
µ= ∗ | + |S S | > 1
|S22 − ∆S11 12 21

The µ test not only is a test for unconditional stability, but the magnitude of µ is
a measure of the stability. In other words, if one two port has a larger µ, it is
more stable.
The advantage of the µ test is that only a single parameter needs to be evaluated.
There are no auxiliary conditions like the K test derivation earlier.
The derivation of the µ test proceeds as follows. First let ΓS = |ρs |e jϕ and
evaluate Γout
S22 − ∆|ρs |e jϕ
Γout =
1 − S11 |ρs |e jϕ

62 / 81
Mu Test (cont)
Next we can manipulate this equation into the following circle |Γout − C | = R
p
∗ ∆−S
|ρs |S11 |ρs ||S12 S21 |
22
Γout + 2
=
1 − |ρs ||S11 | (1 − |ρs ||S11 |2 )

For a two-port to be unconditionally stable, we’d like Γout to fall within the unit
circle
||C | + R| < 1

∗
p
||ρs |S11 ∆ − S22 | + |ρs ||S21 S12 | < 1 − |ρs ||S11 |2

∗
p
||ρs |S11 ∆ − S22 | + |ρs ||S21 S12 | + |ρs ||S11 |2 < 1

The worse case stability occurs when |ρs | = 1 since it maximizes the left-hand side
of the equation. Therefore we have
1 − |S11 |2
µ= ∗ >1 63 / 81
K-∆ Test
The K stability test has already been derived using Y parameters. We can also do
a derivation based on S parameters. This form of the equation has been
attributed to Rollett and Kurokawa.
The idea is very simple and similar to the µ test. We simply require that all points
in the instability region fall outside of the unit circle.
The stability circle will intersect with the unit circle if
|CL | − RL > 1
or ∗ − ∆∗ S | − |S S |
|S22 11 12 21
>1
|S22 |2 − |∆|2

This can be recast into the following form (assuming |∆| < 1)
1 − |S11 |2 − |S22 |2 + |∆|2
K= >1
2|S12 ||S21 |

64 / 81
Two-Port Power and Scattering Parameters
The power flowing into a two-port can be represented by
|V1+ |2
Pin = (1 − |Γin |2 )
2Z0

The power flowing to the load is likewise given by

|V2− |2
PL = (1 − |ΓL |2 )
2Z0

We can solve for V1+ using circuit theory

Zin
V1+ + V1− = V1+ (1 + Γin ) = VS
Zin + ZS

In terms of the input and source reflection coefficient

1 + Γin 1 + ΓS
Zin = Z0 ZS = Z0
1 − Γin 1 − ΓS
65 / 81
Two-Port Incident Wave
Solve for V1+
VS (1 + Γin )(1 − ΓS )
V1+ (1 + Γin ) =
(1 + Γin )(1 − ΓS ) + (1 + ΓS )(1 − Γin )
VS 1 − ΓS
V1+ =
2 1 − Γin ΓS

The voltage incident on the load is given by

V2− = S21 V1+ + S22 V2+ = S21 V1+ + S22 ΓL V2−

S21 V1+
V2− =
1 − S22 ΓL
2
|S21 |2 V1+
1 − |ΓL |2
PL =
|1 − S22 ΓL |2 2Z0
66 / 81
Operating Gain and Available Power
The operating power gain can be written in terms of the two-port s-parameters
and the load reflection coefficient
PL |S21 |2 (1 − |ΓL |2 )
Gp = =
Pin |1 − S22 ΓL |2 (1 − |Γin |2 )

The available power can be similarly derived from V1+

+ 2
V1a
Pavs = Pin |Γin =Γ∗ = (1 − |Γ∗S |2 )
S 2Z0

+ VS 1 − Γ∗S
V1a = V1+ Γin =Γ∗S
=
2 1 − |ΓS |2

|VS |2 |1 − ΓS |2
Pavs =
8Z0 1 − |ΓS |2

67 / 81
Transducer Gain

The transducer gain can be easily derived

PL |S21 |2 (1 − |ΓL |2 )(1 − |ΓS |2 )

GT = =
Pavs |1 − Γin ΓS |2 |1 − S22 ΓL |2

Note that as expected, GT is a function of the two-port s-parameters and the load
and source impedance.
If the two port is connected to a source and load with impedance Z0 , then we
have ΓL = ΓS = 0 and
GT = |S21 |2

68 / 81
Unilateral Gain
Z0
+ ! "
vs S11 0
M1 M2 Z0
− S21 S22

2
GS |S21 | GL

If S12 ≈ 0, we can simplify the expression by just assuming S12 = 0. This is the
unilateral assumption

1 − |ΓS |2 1 − |ΓL |2
GTU = 2
|S21 |2 2
= GS |S21 |2 GL
|1 − S11 ΓS | |1 − S22 ΓL |

The gain partitions into three terms, which can be interpreted as the gain from
the source matching network, the gain of the two port, and the gain of the load.
69 / 81
Maximum Unilateral Gain
We know that the maximum gain occurs for the biconjugate match
∗
ΓS = S11

∗
ΓL = S22

1
GS,max =
1 − |S11 |2
1
GL,max =
1 − |S22 |2
|S21 |2
GTU,max =
(1 − |S11 |2 )(1 − |S22 |2 )

Note that if |S11 | = 1 of |S22 | = 1, the maximum gain is infinity. This is the
unstable case since |Sii | > 1 is potentially unstable.
70 / 81
Design for Gain

So far we have only discussed power gain using bi-conjugate matching. This is
possible when the device is unconditionally stable. In many case, though, we’d like
to design with a potentially unstable device.
Moreover, we would like to introduce more flexibility in the design. We can trade
off gain for
bandwidth
noise
gain flatness
linearity
etc.
We can make this tradeoff by identifying a range of source/load impedances that
can realize a given value of power gain. While maximum gain is acheived for a
single point on the Smith Chart, we will find that a lot more flexibility if we
back-off from the peak gain.

71 / 81
Unilateral Design
No real transistor is unilateral. But most are predominantly unilateral, or else we
use cascades of devices (such as the cascode) to realize such a device.
The unilateral figure of merit can be used to test the validity of the unilateral
assumption
|S12 |2 |S21 |2 |S11 |2 |S22 |2
Um =
(1 − |S11 |2 )(1 − |S22 |2 )

It can be shown that the transducer gain satisfies the following inequality
1 GT 1
< <
(1 + U)2 GTU (1 − U)2

Where the actual power gain GT is compared to the power gain under the
unilateral assumption GTU . If the inequality is tight, say on the order of 0.1 dB,
then the amplifier can be assumed to be unilateral with negligible error.
72 / 81
Gain Circles
We now can plot gain circles for the source and load. Let
GS
gS =
GS,max

GL
gL =
GL,max

By definition, 0 ≤ gS ≤ 1 and 0 ≤ gL ≤ 1. One can show that a fixed value of gS

represents a circle on the ΓS plane
√
∗ g
S11 S 1 − gS (1 − |S11 |2 )
ΓS − =
|S11 |2 (gS − 1) + 1 |S11 |2 (gS − 1) + 1

More simply, |ΓS − CS | = RS . A similar equation can be derived for the load.
Note that for gS = 1, RS = 0, and CS = S11 ∗ corresponding to the maximum gain.

73 / 81
Gain Circles (cont)
0.12 0.13
0.11 0.14
0.38 0.37 0.15
0.1 0.39 0.36
0.4 0.35 0.16
0.09

45
50
0.41 40 0.34

1.0
0.9

1.2
0.17

55

0.8
0.08 35

1.4
0.33

0.7
0.42

0.6 60
0.1

1.6
7
0.0 o) 30 8
3 jB/Y 0.3
E (+

1.8
0.4 0.2 2
NC
TA

65
EP
SC

2.0
0.5
SU

0.1
0.0
VE 25
ITI

9
4

0.3
C

0.4
PA

1
A

70
RC 0.4
), O

5
0.4

0.2
0.0

o
/Z
5

0.3
20

jX
0.4

(+
NT
3.0

75

NE
0.6

PO
0.04

0.21
OM
0.3

0.46

0.29
EC
C
0.8 15

AN
4.0

80

CT
—>
gS = 0 dB

REA
1.0

TOR

0.22
IV E
0.47

0.28
ERA
5.0

UCT

1.0
0.2

85
GEN

IND
0.8 10

AR D

0.23
S T OW

0.27
0.48

90
0.6

ENGTH

ANGLE
10

ANGLE
0.1
0.4

—> WAVEL

OF

0.24
0.26
0.49

OF REFLECTION OEFFICIENT IN DEGREES

TRANSMISSION COEFFICIENT IN DEGRE
20
∗ 0.2
̸ S11 50

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1.0

1.2

1.4

1.6

1.8

2.0

3.0

4.0

5.0

10

20

50

0.25
0.25
0.0
0.0

C
RESISTANCE COMPONENT (R/Zo), OR CONDUCTANCE COMPONENT (G/Yo)
50
0.2

LOAD <—
20

0.24
0.26
0.49
0.1 gS = −1 dB 0.4

WARD
10

ES
TO
0.6
-90

0.23
GTHS

0.27
0.48

gS = −2 dB
ELEN

0.8 -10

o)
-85
WAV

jB/Y
0.2

1.0

0.22
5.0

E (-
<—
0.47

0.28
NC gS = −3 dB
1.0

TA
EP
-80

C 4.0
US 0.8 -15
ES
0.04

0.21
0.3
0.46

TIV

0.29
UC
ND
0.6
-75

3.0
RI
), O
5

0.2
0.0

o
-20
/Z
5

0.3
jX

0.4
0.4

- (
NT 0.4
NE
-70

PO
6

0.1
M
0.0

CO

9
CE -25
4

0.3
0.4

0.5

AN

1
2.0
CT
EA
-65

ER 0.1
7 ITIV

1.8
0.2 8
0.0 AC
CAP
0.6

0.3
3
0.4 -30 2

1.6
-60

0.17
0.7
0.08

1.4
-35 0.33

0.8
0.42

1.2
-55

0.9
0.16

1.0
0.09 0

0
-4

-5

5
0.34

-4
0.41 0.1 0.15

0.11 0.14 0.35

0.4 0.12 0.13
0.39 0.36
0.38 0.37

All gain circles lie on the line given by the angle of Sii∗ . We can select any desired
value of source/load reflection coefficient to acheive the desired gain. To minimize 74 / 81
Extended Smith Chart

For |Γ| > 1, we can still employ the Smith Chart if we make the following
mapping. The reflection coefficient for a negative resistance is given by

−R + jX − Z0 (R + Z0 ) − jX
Γ(−R + jX ) = =
−R + jX + Z0 (R − Z0 ) − jX

1 (R − Z0 ) + jX
=
Γ∗ (R + Z0 ) + jX

We see that Γ can be mapped to the unit circle by taking 1/Γ∗ and reading the
resistance value (and noting that it’s actually negative).

75 / 81
Potentially Unstable Unilateral Amplifier

For a unilateral two-port with |S11 | > 1, we note that the input impedance has a
negative real part. Thus we can still design a stable amplifier as long as the source
resistance is larger than ℜ(Zin )

ℜ(ZS ) > |ℜ(Zin )|

The same is true of the load impedance if |S22 | > 1. Thus the design procedure is
identical to before as long as we avoid source or load reflection coefficients with
real part less than the critical value.

76 / 81
Pot. Unstable Unilateral Amp Example
Consider a transistor with the following S-Parameters
S11 = 2.02∠ − 130.4◦
S12 = 0
◦
S22 = 0.50∠ − 70
0.0
9
0.1
0.4
0.11
0.39
0.12
0.38

45
0.13
0.37
0.14
0.36 0.15
0.35 0.1
6
S21 = 5.00∠60◦
50

1 40 0.3
4
1.0

0.4
0.9

1.2
0.17
55

0.8

0.08 35

1.4
0.33
0.7

0.42
0.6 60

0.18

1.6
0.07 o) 30
jB/Y 0.32
0.43 E (+

1.8
0.2
NC
TA
65

EP
SC

2.0
SU
0.5
6

0.1
0.0

E 25
TIV

9
CI
4

0.3
0.4

PA

1
CA

region
70

R 0.4
), O

ble
0.05

0.4

0.2
Zo

Since |S11 | > 1, the amplifier is

0.45

X/

0.3
20

sta
+j
T(

3.0
EN
75

ON

0.6
MP
0.04

0.21
CO

0.3
0.46

0.29
CE

RS 0.8 15
AN

GS = 5 dB 4.0
80

CT
—>

REA

1.0

potentially unstable. We begin by

TOR

0.22
IVE
0.47

0.28
ERA

5.0
UCT

1.0

0.2
85
GEN

IND

0.8 10

CS
ARD

0.23
∗ to find the negative
S TOW

0.48

0.27
90

0.6

ΓS
plotting 1/S11
ENGTH

ANGLE
10

ANGLE
0.1
0.4
—> WAVEL

OF TRANSM

0.24
0.49

0.26
OF REFLECTION EFFICIENT IN DEGREES
20
0.2
50
ISSION COEFFICIENT IN D

real input resistance.

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1.0

1.2

1.4

1.6

1.8

2.0

3.0

4.0

5.0

10

20

50

0.25
0.25
0.0
0.0

CO
RESISTANCE COMPONENT (R/Zo), OR CONDUCTANCE COMPONENT (G/Yo)
50
0.2
<—

20
RD LOAD

0.24
0.49

0.26
EGREES

0.4

Now any source inside this circle is

0.1
TOWA

10

0.6
1
-90

0.23
GTHS
0.48

0.27
ELEN

0.8
∗
-10

S11
o)
-85
WAV

stable, since ℜ(ZS ) > ℜ(Zin ).

jB/Y

0.2
1.0

0.22

5.0
E (-
<—
0.47

0.28

1.0
NC
TA
EP
-80

4.0
SC

0.8 -15
SU
0.04

0.21

0.3
E
0.46

0.29
TIV
UC
ND

0.6
-75

3.0
RI

We also draw the source gain circle for

,O
0.05

o)

0.2

-20
/Z
0.45

0.3
jX

0.4 (-
NT 0.4
NE
-70

PO
6

0.1

M
0.0

CO
9

CE -25
4

0.3
0.4

0.5

AN
1
2.0

GS = 5 dB.
CT
EA
-65

ER 0.18
ITIV
1.8

0.2
AC 0.07
CAP
0.6

0.32
0.43 -30
1.6
-60

0.17
0.7

0.08
1.4

-35 0.33
0.8

0.42
1.2
-55

0.9

0.1
1.0

9 0 6
0.0
0

-4
-5

0.3
-4

77 / 81
0.4 0.15 4
0.1
0.11 0.14 0.35
0.4 0.12 0.13
0.39 0.36
Amp Example (cont)
∗ . Note the real part is
The input impedance is read off the Smith Chart from 1/S11
interpreted as negative
Zin = 50(−0.4 − 0.4j)

The GS = 5 dB gain circle is calculated as follows

gS = 3.15(1 − |S11 |2 )
√
1 − gS (1 − |S11 |2 )
RS = = 0.236
1 − |S11 |2 (1 − gS )
∗
gS S11
CS = = −.3 + 0.35j
1 − |S11 |2 (1 − gS )
We can select any point on this circle and obtain a stable gain of 5 dB. In
particular, we can pick a point near the origin (to maximize the BW) but with as
large of a real impedance as possible:
78 / 81
Bilateral Amp Design
In the bilateral case, we will work with the power gain Gp . The transducer gain is
not used since the source impedance is a function of the load impedaance. Gp , on
the other hand, is only a function of the load.
|S21 |2 (1 − |ΓL |2 )
Gp =  2
 = |S21 |2 gp
11 −∆ΓL
1 − S1−S 22 ΓL
|1 − S22 ΓL |2

It can be shown that gp is a circle on the ΓL plane. The radius and center are
given by q
1 − 2K |S12 S21 |gp + |S12 S21 |2 gp2
RL = 2
−1 − |S22 |2 gp + |∆|2 gp
∗ − ∆∗ S )
gp (S22 11
CL =
1 + gp (|S22 |2 − |∆|2 )
79 / 81
Bilateral Amp (cont)
We can also use this formula to find the maximum gain. We know that this
occurs when RL = 0, or
1 − 2K |S12 S21 |gp,max + |S12 S21 |2 gp,max
2
=0

1  p 
gp,max = K − K2 − 1
|S12 S21 |
S21  p 
Gp,max = K − K2 − 1
S12
The design procedure is as follows
1 Specify gp
2 Draw operating gain circle.
3 Draw load stability circle. Select ΓL that is in the stable region and not too close to
the stability circle.
4 Draw source stability circle.
5 To maximize gain, calculate Γin and check to see if ΓS = Γ∗in is in the stable region.
If not, iterate on ΓL or compromise. 80 / 81
Refs

“S Parameter Design,” Hewlett-Packard Application Note 154, April 1972.

Microwave Transistor Amplifiers, Analysis and Design, Guillermo Gonzalez,
Prentice Hall 1984.
Microwave Engineering, David Pozer, Third Edition, Wiley 2005.
Microwave Circuit Design Using Linear and Nonlinear Techniques, by George
Vendelin, Anthony M. Pavio, & Ulrich L. Rohde, Wiley 1995.

81 / 81