Micro and mm wave Circuit Design

ENG5056
2020

Lecture 3 (Part 1)

Network analysis theory

 Network analysis theory
• Standard circuit theory (l >> lc , f < 100 MHz)
Circuit is seen as interconnection of lumped active/passive elements with unique
voltage/current defined at any point in the circuit (Dj < 0) KVL and KCL (Kirchhoff)

• Microwave/mm wave network analysis (l ~ lc )

Circuit and Network concepts extension to Microwave/mm wave analysis and design avoiding
solving of Maxwell’s equations (much more complex, too much detail and information)
Sometime is an oversimplification! be aware of approximation.....
Characteristic impedances

Matrix Circuit theory

 Network analysis theory: matrices
For a linear circuit, the I-V relationship is linear and can be written in
the form of matrix equations.

A simple example of linear 2-port circuit is shown below. Each port is

associated with 2 parameters, the V and I.

Convention for positive

R polarity current and voltage
I1 I2
+

Port 1 V1 C V2 Port 2

-
 Network analysis theory: matrices
• For this 2 port circuit we can easily derive the I-V relations.
R I1 I2
I1 I1 + I 2 = jwCV2
I1 = 1 (V1 - V2 )
R
V1 V2 R
(R )
Þ I 2 = - 1 V1 + 1 + jwC V2 jwCV2 C V2

• We can choose V1 and V2 as the independent variables, the I-V relation

can be expressed in matrix equations. Network parameters
(Y-parameters)
é I1 ù éê R ù éV ù
1 -1
R ú 1 é I1 ù é y11 y12 ù éV1 ù
êI ú = ê 1
ë 2 û ë- R ( R
)
1 + jwC ú êV ú
ûë 2 û
êI ú = ê y
ë 2 û ë 21 y22 úû êëV2 úû

R
I1 I2 I1 I2

Port 1 V1 C V2 Port 2
V1 2 - Ports V2
 Network analysis theory: matrices
• To determine the network parameters, the following relations can be used:

I I
é I1 ù é y11 y12 ù éV1 ù y11 = 1 y12 = 1
êI ú = ê y
V1 V = 0 V 2 V =0
y ú êV ú 2 1
ë 2 û ë 21 22 û ë 2 û I
y21 = 2 I
y22 = 2
or V1 V = 0 V 2 V =0
2 1

I = Y ×V This means we short circuit the port

• For example to measure y11, the following setup can be used:

I1 I2
Short circuit

V1 2 - Ports V2 = 0
Network analysis theory: matrices
• By choosing different combination of independent variables, different
network parameters can be defined. This applies to all linear circuits
no matter how complex.
• Furthermore this concept can be generalized to more than 2 ports,
called N - port networks.
I1 I2

I1 I2 V1 2 - Ports V2

V1 V2
éV1 ù é z11 z12 ù é I1 ù
êV ú = ê z z ú êI ú
ë 2 û ë 21 22 û ë 2 û

Linear circuit, because all éV1 ù é h11 h12 ù é I1 ù

ê I ú = êh úê ú
elements have linear I-V relation ë 2 û ë 21 h22 û ëV2 û
 Network analysis theory: matrices
• Of particular interest in RF and microwave systems is the transmission or
ABCD or transmission parameters. ABCD parameters are the most useful for
representing the line and other linear microwave components in general.

Take note of the

éV1 ù é A B ù éV2 ù direction of positive current!
ê I ú = êC D ú ê I ú I1 I2
ë 1û ë ûë 2 û
Þ V1 = AV2 + BI 2
I1 = CV2 + DI 2
V1 2 -Ports V2

V V1 I I1
A= 1 B= C= 1 D=
V2 I = 0 I 2 V =0 V2 I = 0 I 2 V =0
2 2 2 2

Open circuit Port 2 Short circuit Port 2

 Network analysis theory: matrices
• The ABCD matrix is useful for characterizing the overall
response of 2-port networks that are cascaded to each other.

I1 I2’ I2 I3
éV1 ù é A1 B1 ù é A2 B2 ù éV3 ù
é A1 B1 ù é A2 B2 ù ê I ú = êC D ú êC D ú ê I ú
V1 êC D ú V2 êC D ú V3 ë 1û ë 1 1ûë 2 2 ûë 3 û
ë 1 1û ë 2 2û
éV1 ù é A3 B3 ù éV3 ù
Þê ú=ê úêI ú
I C
ë 1û ë 3 D 3 ûë 3 û

𝐴#𝐴$ + 𝐵#𝐶$ 𝐴#𝐵$ + 𝐵#𝐷$ Overall ABCD matrix

𝐶#𝐴$ + 𝐷#𝐶$ 𝐶#𝐵$ + 𝐷#𝐷$
Network analysis theory: ABCD matrix
 Network analysis theory: Matrices
Impedance and admittance Port (2 terminal pair) and reference planes t defined
matrices

ABCD matrix
 Network analysis theory: Matrices
Port (2 terminal pair) and reference planes t defined
Scattering matrix

measurable with VNA

 Network analysis theory: matrices
• Usually we use Y, Z, H or ABCD parameters to describe a linear
two port network.
• These parameters require us to open or short a network to find
the parameters.
• At radio frequencies it is difficult to have a proper short or open
circuit, there are parasitic inductance and capacitance in most
instances.
• Open/short condition leads to standing wave, can cause
oscillation and destruction of device.
• For non-TEM propagation mode, it is not possible to measure
voltage and current. We can only measure power from E and H
fields.
 Network analysis theory: matrices
• Hence a new set of parameters (S) is needed which
– Do not need open/short condition.
– Do not cause standing wave.
– Relates to incident and reflected power waves, instead of
voltage and current.

• As oppose to V and I, S-parameters relate the reflected and

incident voltage waves.

• S-parameters have the following advantages:

1. Relates to familiar measurement such as reflection coefficient,
gain, loss etc.
2. Can cascade S-parameters of multiple devices to predict system
performance (similar to ABCD parameters).
3. Can compute Z, Y or H parameters from S-parameters if needed.
 Scattering S-parameter matrix
• There is a voltage and current on each port.
• This voltage (or current) can be decomposed into the incident (+) and reflected
component (-).

V (z ) = V2+ e - jbz + V2- e jbz

V (0) = V2 = V2+ + V2- + V1+

V1 + V1 -
I1 V1 = +
I (z ) = I 2+ e - jbz - I 2-e jbz Port 1
- V1-
I (0) = I 2 = I 2+ - I 2- Port
V1 n
V1 = V1+ + V1-
+z
z=0
I1 = I1+ - I1-
Port 2
Linear
n - port
Port 1
= 1 V1+ - V1-
Z c1
( )
Network
 Scattering S-parameter matrix
Thus in general:
V =sV
- +

1 12 2

V1+ V1-
V =s V
- +
Port 1
2 22 2
Zc1 Vn+
Vn- Port n
=s V
- +
V n n2 2

Zcn
Port 2
Linear
V2+ Constant that
V2- n - port depends on the
Zc2 Network network construction

T-line or
waveguide Vi+ and Vi- are propagating
voltage waves, which can
be the actual voltage for TEM
modes or the equivalent
voltages for non-TEM modes.
 Scattering S-parameter matrix
Considering that we can send energy into all ports, this can be generalized
to:
V = s V + s V + s V +!+ s V
- + + + +

1 11 1 12 2 13 3 1n n

V = s V + s V + s V +!+ s V
- + + + +

2 21 1 22 2 23 3 2n n

V = s V + s V + s V +!+ s V
- + + + +

n n1 1 n2 2 n3 3 nn n

Or written in Matrix equation:

éV ù és s ... s ù éV ù
- +

Vi -
11 12 1n

ê
1

ú ê ê ú
1

ú = ês s ... s úú êV ú
- +

êV Sij = +
V = SV
- +
or
2 21 22 2n 2

ê: ú ê: : ! : úê : ú Vj
ê ú ê úê ú Vk+ = 0 , for ® k ¹ j
û ës s ... s û ëV û
- +

ëV n n1 n2 nn n

Where sij is known as the generalized Scattering (S) parameter, or just S-

parameters for short. Each port i can have different characteristic
impedance Zci
 Scattering S-parameter matrix
For 2-port networks, (4.4) reduces to:

éV ù é s s ù éV ù éV ù
- + +

ê ú=ê ê ú = Sê ú
1 11 12 1 1

ú
ëV û ës s û ëV û
- + +

2 21 22 ëV û
2 2

- - - -
V V V V
s = 1
+
s = 2
+
s = 2
+
s = 1
+
V V V V
11 21 22 12

1 1 2 2
V2+ = 0 V2+ = 0 V1+ = 0 V1+ = 0

Note that Vi+ = 0 implies that we terminate i th port with its characteristic
impedance.

Z -Z
-
V (1)

S = 1
=G (1)
= in 0

Z +Z
+
V V2+ =0 V2+ =0
11 (1) Z0

1 in 0
 Networks and matrices properties
A reciprocal network is one in which the transmission of a signal
between any two ports does not depend on the direction of
propagation- input and output ports are interchangeable (all passive
elements using isotropic materials: TL, filters, splitters...)

For a lossless network, all of the power (or energy) of a wave

incident at any one port must be equal to the sum of the powers of
the waves exciting from the other ports plus the power of the
reflected wave at the incident port - no dissipation (heat, radiation...)

By imposing these conditions on to the respective electric and

magnetic fields => Voltage and current waves you infer properties of
matrices describing the system!
 Matrices properties
Z and Y Matrices

ABCD Matrix

for any i,j

S Matrix
All the matrices are related to each other
(look for the conversion table on Pozar)
Micro and mm wave Circuit Design
ENG5056
2020

Lecture 3 (Part 2)

Filters theory
 Filtering by using frequency response of reactive
(lumped ) components
Capacitor:

= j2!"L

#
=
$%&'(

Capacitors used in μ-Wave amplifier Used to ground RF signal at

DC decoupling between stages + biasing circuit

As frequency (f) gets higher Impedance(Zc) of a series capacitance gets smaller ⇒

Capacitor becomes leaky
 Filtering by using frequency response of reactive
(lumped ) components
Inductor:

= j2!"L

#
=
$%&'(

Inductors used in μ-Wave amplifier RF decoupling at dc biasing circuits as a choke

As frequency (f) gets higher Impedance(ZL) of a series inductor gets bigger ⇒ Inductor
becomes an open circuit
Resonators
Resonators
 Infinite periodic structure
Stop-band and pass-band properties of a periodically loaded with lumped
susceptances infinite transmission line
b is normalized respect to Z0

éV1 ù é A1 B1 ù é A2 B2 ù éV3 ù
ê I ú = êC D ú êC D ú ê I ú
ë 1û ë 1 1ûë 2 2 ûë 3 û
éV1 ù é A3 B3 ù éV3 ù
Þê ú=ê úêI ú
I C
ë 1û ë 3 D 3 ûë 3 û
Infinite periodic structure
 Infinite periodic structure
For an in0inite periodic loaded line (no reflection)

non trivial solution given by det M = 0

AD-BC = 1 due to the symmetry of the system (reciprocity)

𝛼 = 0, 𝛽 ≠ 0
pass-band of structure
The frequency and b can determine the pass-band and
stop-band characteristic of the line: filtering.
𝛼 ≠ 0, 𝛽 = 0, 𝜋
stop-band of structure
@A # DEF
(terminated line with matching ZL = ZB ≡ 𝑍0 B = )
CA #
B
GH I#
 Signal Filtering methods:
• Image parameters method:
A cascade of simple two-ports filters sections to provide desired cut-off frequency
and attenuation characteristics No specific response over the complete operating
range (large number of iterations to obtain desired results).

• Insertions losses method:

Design of low-pass filters that are normalized in terms of impedance and frequency.
Transformation are then applied to obtain filters with desired frequency range and
impedance level.

Both methods provide Lumped-element circuits.

Richard’s transformation and Kuroda’s identities allow to modify the design for mm and
microwave applications using distributed elements consisting of transmission line sections.
 Image parameter method:
Specification of pass-band and stop-band characteristics (no specific frequency included in design)

2-port reciprocal network characterized by ABCD matrix

Image impedances:

Zi1 ≡ input impedance port 1 when line terminated with Zi2

Zi2 ≡ input impedance port 2 when line terminated with Zi1

𝑍JA# = 𝑍J# 𝑎𝑛𝑑 𝑍JA$ = 𝑍J$

By matching the symmetric (A=D) 𝐴𝐵 𝐵𝐷

reciprocal network (AD-BC=1) 𝑍J# = 𝑎𝑛𝑑 𝑍J$ =
equating the input and image 𝐶𝐷 𝐴𝐶
impedances
𝐵
symmetric 𝑍J# = 𝑍J$ =
Voltage/current transfer function network 𝐶
terminated in its image impedances:

𝑒 IR = 𝐴𝐷 − 𝐵𝐶
or
cosh 𝛾 = 𝐴𝐷
Image parameter method: filtering
Constant k (low-pass) filters in T form

Determine ABCD parameters as function of L and C (use

tables) and then:

U XHUV U $
𝑍JT = 1− , 𝑅F = = 𝑘 and 𝜔] ≡
V Y V UV

XH $XH $X XH 𝛾 = 𝛼 + 𝑗𝛽
Finally 𝑍JT = 𝑅F 1 − H and 𝑒R =1− H + H −1 as usual
X^ X^ X^ X^

𝜔 < 𝜔V ∶ 𝜔 > 𝜔V ∶

𝑍JT is real 𝑍JT is imaginary

𝛾 is imaginary 𝛾 is real
𝜔 →∞,𝛼 → ∞
attenuation rate
|𝑒 R | = 1 40dB/decade (𝜔>> 𝜔𝐶 )
Image parameter method: filtering
Constant k (low-pass) filters in p form

Determine ABCD parameters as function of L and C (use

table) and then:

U $ hi
𝑅F = = 𝑘 and 𝜔] ≡ , 𝑍Jg =
V UV jH
#I H
j ^

Constant k (High-pass) filters in T and p form

U #
For both: 𝑅F = =𝑘 and 𝜔] ≡ ….
V $ UV

Drawbacks: relatively slow attenuation rate passing 𝜔V and 𝑍JT,g (𝜔) non constant.
 Image parameter method: filtering
m-derived low/high pass filters (in T form)
𝑍#m /2 =m 𝑍# /2 𝑍#m /2

𝑍$m

We set 𝑍#m = 𝑗𝜔 𝑚 𝐿 (= m Z1 ) and chose 𝑍$m so that 𝑍JT is the same as for the k – constant:
1 (1 − 𝑚$ )$ 𝑍$ 1 − 𝑚$
𝑍$m = + 𝑗𝜔𝐿 ( = + 𝑍# )
𝑗𝜔 𝑚 𝐶 4𝑚 𝑚 4𝑚

2 k-const + m-derived
𝜔] ≡
If 0 < m < 1 𝑒 R is real and 𝑒 R > 1 𝑓𝑜𝑟 𝜔 > 𝜔V 𝐿𝐶

Stop-band at 𝜔 = 𝜔V 𝑎𝑛𝑑 𝑖𝑛𝑓𝑖𝑛𝑡𝑒 𝑎𝑡𝑡𝑒𝑛𝑢𝑎𝑡𝑖𝑜𝑛 𝑎𝑡:

X^
𝜔y = (it can be controlled by m)
#IzH
Attenuation decreases for 𝜔 > 𝜔y
𝑍JT (w) still not constant!
Image parameter method: filtering
m-derived low/high pass filters (in 𝜋 form)

To obtain equivalent 𝜋 section you consider it

as a piece of T - sections

$
1 − 1 − 𝑚 $ (𝜔 | $ ) 𝐿
𝜔V
𝑍Jg = 𝑅F 𝑅F =
𝜔$ 𝐶
1− |𝜔$
V

It’s is possible to minimize the variation of

𝑍Jg over the pass-band of the filter

To match the 𝑍Jg = 𝑍JT we can use a bisection of 𝜋,

finding the ABCD parameters and finding 𝑍J# and 𝑍J$
Image parameter method: filtering
Composite filters

Low pass
filter

Pass-band and notch filters can be

obtained opportunely combining low-
pass and high pass filters with right
input impedance matching
Micro and mm wave Circuit Design
ENG5056
2020

Lecture 3 (Part 3)

Filters theory
 Insertion losses method
Perfect filter: infinite attenuation in the stop–band, and linear phase response for no
dispersion (avoid signal distortion) and Insertion losses IL = 0 dB in the pass-band

~•€•‚ ƒ„ƒJ…ƒ†…• ‡‚•z ˆ•‰‚]• ~Œ•Ž #

𝑃Uh = ~•€•‚ Š•…J„•‚Š ‹• …•ƒŠ
= ~••‘’
= #I|“(X)|H
(in the pass-band)

real polynomial
•(XH )
|Γ(𝜔)|$ = (even function of 𝜔 and can be expressed as polynomial in 𝜔$ )
• XH B–(XH )

• maximally flat (binomial or Butterworth)

𝑀(𝜔$ ) • Equal ripple (Chebyshev polynomial)

𝑃Uh 𝜔 =1+
𝑁(𝜔 $ )
• Elliptic function (Specify min and max attenuation)

• Linear phase
 Insertion losses method
Rsource = 1, wC = 1 => deriving normalized L and C components

N=3
20N dB / decade

Elliptic functions

Maximally flat: chose order N => Derive Zin , G , and PLR and equate with polynomial to obtain
values of reactive elements. N can be chosen by specifying PLR at a certain w in stop-band

Equal Ripple: chose order N => Derive Zin , G , and PLR and equate with Chebyshev polynomial
to obtain values of reactive elements. N even gives R not unity whereas N odd does: for N
even an impedance matching should be considered. k is used to determine the ripple level as
1+k2 (usually 3 dB). N can be chosen by specifying PLR at a certain w in stop-band
 filter transformation
Chosen source
impedance Scaling of impedance and frequency norm Chosen wC
w
w¬
L' = R0 L wc
C
C =
'
w
R0 jX k = j Lk = jwL'k
wc
Rs' = R0 ( = Rsource) w
jBk = j Ck = jwCk'
wc
RL' = R0 RL
New elements:

Lk R0 Lk
L =
'
=
w wc
k

Ck Ck
C = =
'

w R0wc
k
 filter transformation
high pass and bandpass/bandstop transformation

w0 æ w w0 ö 1 æ w w0 ö -1
wc w¬ çç - ÷÷ = çç - ÷÷
w¬- w2 - w1 è w0 w ø D è w0 w ø
w w - w1 and
where D = 2 w0 = w1w2
w0
fractional bandwidth center frequency

1
C =
'

R0wc Lk
k

D
R0 L'k =
L =
' w0Ck
wc C k
k
Ck
Ck' =
Dw 0
 Insertion losses method
summary

Series

Shunt

Plus impedance scaling!

 Filter implementation
Lumped-element filter design work well at low frequencies but problem arise at
mm at microwave frequencies:

• Lumped elements are generally available for limited range of frequencies

(difficult to implement at mm/MW frequencies).

• Distance between elements between components is not negligible (no

lumped element approximation)

From lumped elements to distributed elements (TL sections):

Richard’s transformations

Separation of filters by using TL (filter response not affected):

Kuroda’s identities
 Filter implementation
Richard’s transformations

Unity filter impedance/cut off

=1
æp 𝜔 ö
jX L = jw L = jZ o tan ç W ÷ = SZ o
è 4 𝜔™ø
For l = l/8 , S = j at 𝜔 = 𝜔o = 𝜔c
æp 𝜔 ö
jBC = jw C = jYo tan ç 𝜔W ÷ = SYo
è 4 ™ø
Away from the cut-off frequency the impedances of the stub will no longer match the original
lumped element impedances and the filter response will differ from the prototype response
 Filter implementation
Kuroda’s Identities: uses TL to:
• Physical Separation of transmission line elements
achieved by using Unit Elements (UEs), transform
open circuit to shunt stubs, more practical Z

X
UE electrical length: q = (p /4)(X )
š
UE Characteristic Impedance ZUE

é cos q jZUE sinq ù é 1 jWZUE ù

éA Bù 1
êC D ú =ê j ú= ê jW ú
ë ûUE ê sinq cos q ú 2 ê 1 ú
êë ZUE úû 1 + W êë ZUE úû
The Four Kuroda’s Identities
Kuroda’s Equivalent Circuit

L = R for
Realizations of Distributed Filters
Realizations of Distributed Filters
Impedance and admittance inverters
g
l/4 resonator stubs in shunt along transmission (both 𝜃 = 𝛽𝑙 = $ )
They form the inverse of the load impedance or admittance
Microwave resonators (lumped elements)
Series RLC resonant circuit Parallel RLC resonant circuit
 Bandpass and Bandstop filters with l/4 TL

Parallel RLC
resonator
ℓ = l /4 at w0
w = w0 + Dw

( )

Series RLC
resonator

l/4 resonator can be use to implement bandpass and bandstop filters

 Bandpass and Bandstop filters with l/4 TL
g
l/4 resonator (S.C. or O.C.) stubs in shunt along transmission (both 𝜃 = 𝛽𝑙 = $ )

Bandstop filter

Series

Bandpass filter

Shunt
Bandpass and Bandstop filters with l/4 TL
g
l/4 resonator stubs in shunt along transmission (both 𝜃 = 𝛽𝑙 = $ )
Bandstop filter
 Filter Realization Procedure
• Select normalized filter parameters to meet
specifications
• Replace L’s and C’s by lo /8 transmission lines
• Convert series stubs to shunt stubs using Kuroda’s
Identities
• Rescale impedance and frequency and select
equivalent microstrip/CPW design and then use
properties of lo /4 as resonators or
impedance/admittance inverters