Chapter 5 Scattering Matrix
Chapter 5 Scattering Matrix
Figure 1. N-port network with input and output waves at each port.
Let us assume that Vi+ and Vi- are the forward and backward voltage waves on port i (for i = 1, 2, ….. N)
and corresponding to that we have current waves as Ii+ and Ii-. The convention we are using is putting a
different z-axis on every transmission line with z=0 at the port. Thus, the voltage Vi and current Ii we will
get at every single port will be given as:
𝑉𝑖 = 𝑉𝑖 + + 𝑉𝑖 −
1
𝐼𝑖 = 𝐼𝑖 + − 𝐼𝑖 −
We can relate the voltage Vi and current Ii with a matrix as:
𝑉1 𝑍11 𝑍12 … 𝑍1𝑁 𝐼1
𝑉2 𝑍21 𝑍22 … 𝑍2𝑁 𝐼2
[ ]=[ ][ ]
.. … … … … ..
𝑉𝑁 𝑍𝑁1 𝑍𝑁2 … 𝑍𝑁𝑁 𝐼𝑁
Or in short form as:
[𝑉] = [𝑍][𝐼]
[𝑍] is called the impedance matrix.
What does this mean in terms of measurements? We apply currents to the different ports and measure the
voltages. Let us open V1 up and see how we can measure the various components of [𝑍].
𝑉1 = 𝑍11 𝐼1 + 𝑍12 𝐼2 + ⋯ + 𝑍1𝑁 𝐼𝑁
We can write 𝑍11 as:
𝑉1
𝑍11 = ]
𝐼1 𝐼
2 ,𝐼3 …..𝐼𝑁 =0
To measure Z11, we keep ports I2, I3, …. IN open so no current can pass through those ports and measure the
voltage on V1. While we are doing this, we can also measure the matrix component Zi1 by measuring the
voltage Vi on the ith port. Thus, in general we can write:
𝑉𝑖
𝑍𝑖𝑗 = ]
𝐼𝑗 𝐼
𝑘 =0 𝑓𝑜𝑟 𝑘≠𝑗
𝑍𝑖𝑗 is the “transfer” impedance between port “i” and “j” when all other ports are open circuited.
𝑍𝑖𝑗 is the input impedance seen looking into port “i” when all other ports are open-circuited.
By doing these measurements and characterizing the impedance matrix, we can know how the circuit will
respond to the electrical elements connected outside. We can know everything there is to know about a
circuit.
We could also have done the opposite and applied voltage instead of current and then we get an admittance
matrix given as:
[𝐼] = [𝑌][𝑉]
For calculating admittance matrix, we short the ports (so that voltage on them goes to 0) instead of keeping
them open-circuit. Obviously, the two matrices are related by inverse relations such that:
[𝑌] = [𝑍]−1
You should have seen these matrices in your circuits courses, so we are not going to spend much time on
these but can we really do the above experiments in high frequency to measure the impedance or admittance
matrices? Remember any scopes we use have to be connected to the ports with transmission lines and the
values change when we go to different lengths of the transmission lines. We need a new way forward.
2
5.3 Scattering Matrix
Looking at the network in Figure 1, instead of measuring voltages and currents, we can also measure the
forward and backward waves in the ports and relate them by a matrix. Such a matrix is called the Scattering
matrix and is written as:
For the time being for the same of developing an understanding, we are measuring these waves right at the
port itself. We will correct this later. We are also assuming that we all connecting all ports with transmission
lines of same characteristics impedance, Z0. Now practically this is what will be done (we won’t keep
different types of cables just for fun, will we) but we can extend the S parameters to generalized form,
which will be covered in ECE 373.
Let us see what the scattering parameters, S, imply. Let us open up 𝑉1 − and we get:
It can be considered as the reflection coefficient on port 1 provided there are no waves entering the network
at any other port. Can we achieve this by keeping the other ports open circuit like we did for impedance
matrices or closed circuit like we did for admittance matrices. Absolutely not. Some of 𝑉1 + may exit
the other ports and if there is an open circuit or short circuit there, will get reflected back and viola, the
condition is broken. So how can we achieve no incoming waves at other ports?
There is only one way. We have to match the other ports by terminating them with a load equal to the
transmission line characteristic impedance, Z0. No matter if a wave is transmitted to another port, there will
be no reflections at the port and hence, no wave entering the network at that port.
So S11 can be defined as the reflection coefficient on port 1 provided all other ports are terminated with a
matched load. This termination is very important and implied in the S-parameter definition. What is
S12. It is the transmission coefficient from port 2 to port 1 when all other ports except for 2 are terminated
with a matched load. We apply a source to port 2 and measure how much voltage wave exists on port 1.
Remember port 1 has to be matched also.
In general, we can write a S parameters as:
𝑉𝑖 −
𝑆𝑖𝑗 = + ]
𝑉𝑗 𝑎𝑙𝑙 𝑝𝑜𝑟𝑡𝑠 𝑒𝑥𝑐𝑒𝑝𝑡 𝑓𝑜𝑟 𝑗 𝑡𝑒𝑟𝑚𝑖𝑛𝑎𝑡𝑒𝑑 𝑤𝑖𝑡ℎ 𝑎 𝑚𝑎𝑡𝑐ℎ𝑒𝑑 𝑙𝑜𝑎𝑑
In practice, though not necessary, the source impedance on port j is also matched to the transmission line.
If it is not done so, we will have to set up a calculation for V+.
Example 1. Let us calculate the S parameters for the following circuit.
3
Solution: First question, how many ports does the network have? We have 2 ports and thus, we will have
a 2 x 2 matrix consisting of S11, S12, S21, S22. To calculate S11 and S21, we apply our signal to port 1 and
measure the reflections on port 1 and transmission to port 2. However, we have to make sure, we are have
connected a matched load to port 2. Thus, the setup looks like:
We have connected transmission lines to our ports and terminated the transmission line on port 2 with a
matched load. Since, port 2 is matched, the input impedance we will see will be Z0 and the circuit looking
from port 1 will look as:
The input impedance we see on port 1 will be Z2 in series with Zo the combination of which is parallel with
Z3, and the combination in series with Z1. This is load which is seen on the transmission line. So we get:
𝑍3 (𝑍2 + 𝑍0 )
𝑍𝑖𝑛 = 𝑍1 + 𝑍3 //(𝑍2 + 𝑍0 ) = 𝑍1 +
𝑍3 + 𝑍2 + 𝑍0
The reflection created at port 1, will be the value of S11 and can be written as:
𝑉1 − 𝑍𝑖𝑛 − 𝑍0
𝑆11 = + =
𝑉1 𝑍𝑖𝑛 + 𝑍0
4
Thus, S11 can be calculated by terminating port 2, and then seeing what equivalent impedance looks at port
1. That impedance acts like a load on the transmission line and thus, reflection coefficient can be calculated.
To calculate S21, we have to calculate the voltage which is dropped on port 2. That voltage will excite a
wave on port 2 (remember port 2 is matched and hence, there is no reflected wave). We follow the rules
you learned in ECE 140. Let us see the steps and we refer to the figure below:
To calculate S12 and S22, we repeat the process with matched load on port 1 and wave applied on port 2.
For this course, we will only limit ourselves to calculating S-parameters to 2 port circuits but you should
be able to extend it to N port circuits with the same thinking.
Practice the method with the following circuit using a transmission line of characteristics impedance of 50
:
5
Important question: Will the S-parameters change if the characteristics impedance of the
transmission line was 75 ?
Remember, S-parameters are defined with respect to the transmission lines used in the test system.
Note: In general, Sij is a complex number and normally written as
which basically means:
|𝑺𝒊𝒋 |𝒆𝒋𝜽
𝑉𝑖 − + 𝑉𝑖 + = 𝑉𝑖
and
𝐼𝑖 + − 𝐼𝑖 − = 𝐼𝑖
However, the waves are related to each other by the characteristics impedance of the transmission line.
𝑉𝑖 + − 𝑉𝑖 −
𝐼𝑖 =
𝑍0
We know that the voltage and current at the ports are related by the characteristics impedance. Thus, we
get:
[𝑉] = [𝑍][𝐼]
[𝑍]
[𝑉 + ] + [𝑉 − ] = [𝑍]{[𝐼𝑖 + ] − [𝐼𝑖 − ]} = {[𝑉 + ] − [𝑉 − ]}
𝑍0
To work with matrix, we need to multiply the right hand side terms with the unitary matrix [U].
[𝑍]
[𝑈][𝑉 + ] + [𝑈][𝑉 − ] = {[𝑉 + ] − [𝑉 − ]}
𝑍0
Reworking the terms, we get:
6
[𝑍] [𝑍]
{ + [𝑈]} [𝑉 − ] = { − [𝑈]} [𝑉 + ]
𝑍0 𝑍0
So we get:
−1
[𝑍] [𝑍]
[𝑆] = { + [𝑈]} { − [𝑈]}
𝑍0 𝑍0
and
∗
∑𝑁
𝑘=1 𝑆𝑘𝑖 𝑆𝑘𝑗 = 0 for 𝑗 ≠ 𝑖
In words, this operation is that we take a column and do a dot product of the column with itself (multiply
all a row term with its complex conjugate and add the results), we should get a sum of 1. Every column
7
should do that. But that is not all. When we take a column and take a dot product with another column, we
should get a 0. This should happen for every pair of columns we have.
Let us consider a two port network with the following S-matrix:
A rule is broken in the first step itself and we do not need to go any further. The network is not loss-less.
5.7 Transmission and Reflection Coefficients for terminated ports
Please remember again that S parameters only represent the reflection coefficients on a port and
transmission coefficients between ports when the ports are terminated with matched loads. If we have
networks which are terminated with loads other than matched loads, the value of the reflection and
transmission coefficients will change. We can use the S-parameters but will have to add other equations at
the ports which are terminated. Let us consider an example to understand how. Let us take the network
represented by the S parameters given in the matrix above i.e.:
We terminate port 2 with a short circuit. What is the reflection coefficient seen at port 1?
We start by writing the relations between the reflected waves and the transmitted waves. I am doing it in
the general form. These can be written as:
𝑉1 − = 𝑆11 𝑉1 + + 𝑆12 𝑉2 +
𝑉2 − = 𝑆2 𝑉1 + + 𝑆22 𝑉2 +
However, because port 2 is short circuited, the wave 𝑉2 − which leaves the network will get reflected back
and come into the network as 𝑉2 + . Thus,
𝑉2 + = −1 ∗ 𝑉2 −
8
We can substitute the equation in the second equation above to get:
Or
−𝑆21 𝑉1 +
𝑉2 + =
(1 + 𝑆22 )
We can substitute the term in our first equation to get:
You can substitute the values to get the exact numbers. The reflection coefficient will be:
𝑉1 − 𝑆12 𝑆21
+ = (𝑆11 − )
𝑉1 1 + 𝑆22
Let us try to understand what do these terms physically tell us (we are going back to our transient thinking
to understand). A wave 𝑉1 + hits the network and an immediate reflection of S11 is created. That is the first
term. Another part of the wave gets transmitted to port 2 with a transmission coefficient of S 21. There the
wave bounces back and forth between the termination with a coefficient of -1 and the network with a
coefficient of S22. As it bounces back and forth, it creates a geometric series of the form 1 + a + a 2 + a3 +
𝑆21
…. where “a” is Γ𝐿 𝑆22 , 𝑤ℎ𝑒𝑟𝑒 Γ𝐿 = −1 and thus, we get a term which will be 1+𝑆 , S21 due to how much
22
1
was transmitted and 1+𝑆22
due to the infinite bouncing back and forth in a cavity that is created. Part of this
total is transmitted to port 1 with transmission coefficient of 𝑆12. That is what the second term is
representing.
The network of port 2 is acting like a cavity, like a drum for example and the results we see are the beats
of the drum.
5.8 Shift in Reference Plane
Now let us consider what happens when we have transmission lines of lengths added to the ports as shown
in the figure below. Instead of measuring right at the port, we will be actually measuring at the ends of the
transmission lines.
9
We can write the relationship between the waves on a transmission line as:
What we want to do is relate the S matrix measured right at the network port (as we have been doing) with
the S matrix measured at the end of the transmission line. Let us call that S’. We have:
10
Thus, the shifted reference plane S’ matrix can be written as:
The matrices around [S] account for the phase delays the wave has to take through the transmission line.
11