Basics Equations for Filters

Synthesis

Giuseppe Macchiarella
Polytechnic of Milan, Italy
Electronic and Information Department
Scattering parameters for lumped-elements filters

2-port
R0 R0
network

For linear, lumped-element networks the scattering

parameters can be always represented as a polynomial
ratio of the complex frequency p ( p    j , with   2 f ) :
N11 ( p ) N12 ( p ) N 22 ( p )
S11  , S 21  S12  , S 22 
D( p) D( p) D( p)

Roots of N11 (N22) : reflection zeros

Roots of N21 : transmission zeros
Roots of D : poles (natural frequencies)

The above polynomials are called characteristic polynomials

 Properties of lossless networks

For synthesis purposes the 2-port network representing a

filter is assumed lossless (i.e. composed by lossless elements)

The scattering matrix of a lossless network is unitary:

S11  p   S11*  p   S 21  p   S 21
*
 p  1
S  S *  U 2
S 22  p   S 22
*
 p   S12  p   S12*  p   1

As a consequence the following properties hold for the

polynomials defining the scattering parameters:
N11 ( p )  N ( p ), N 22 ( p )   1 N * ( p )
n

N ( p )  N * ( p )  N 21 ( p )  N 21
*
( p )  D ( p )  D* ( p )

The latter expression is know as Feldtkeller equation

 Properties of characteristic polynomials

 The roots of D(p) are real or conjugate pairs. The real

part must be negative (strict Hurwitz polynomial)
 The roots of N(p) can be everywhere in the complex
plane (if complex they occur as conjugate pairs)
 The roots of N12(p) can be on the imaginary axis
(conjugate pair) or on the real axis (pairs with opposite
values) or as a complex quad in the p plane
 The coefficients of all the polynomials are real
numbers
 Steps of the synthesis process

 Assignment of selectivity specifications (Attenuation

mask)
 Selection of a suitable approximating function and
evaluation of its parameters in order to satisfy the
specifications
 Evaluation of the characteristic polynomials from the
approximating function
 Synthesis of the network from the polynomials
 Synthesis in a transformed frequency domain

In order to simplify the design process, the synthesis can be

performed in a normalized low-pass domain, analytically
defined by a frequency transformation. The synthesized
network is then de-normalized to the band-pass frequency
domain with a suitable circuit transformation

Requirements in Synthesis in the

Specifications the normalized transformed
domain domain

De-normalization of
the synthesized
network
 Low-pass  Band-pass transformation

Let be s=+j the normalized low-pass domain where

the filter passband is defined B=-1+1.
The following equation relates the normalized domain
with a pass-band domain suitably defined:
f0  p f0 
s   
B  f0 p 
f0 and B are the frequency parameters defining the
transformation:

f0  f p1  f p 2 , B  f p 2  f p1
 Properties of the transformation (imaginary axis)

For s=j and p=jf it has

f0  f f0 
   
B  f0 f 
s -1 0 1 s


fs1 fp1 f0 fp2 fs2 f (>0)

f p1  f p 2  f s1  f s 2  f 02
 Conservation of response: A(j)=A(j)
80

60
dB

40

20

0
-2 -1.5 -1 -0.5 0 0.5 1 1.5 2
Normalized

80

60

B=55 MHz
dB

40
f0=1000 MHz
20

0
950 960 970 980 990 1000 1010 1020 1030 1040 1050
MHz
 Circuital property of the transformation:
Conservation of Z and Y

Low pass Band pass

gc Ceq

Leq
f0  p f0 
Y=s.gc s    f 0  p 0  1
B  f0 p  Y   
 C g  p  C 
B  0 p  p  Leq
eq

g 1
gL Ceq  C , Leq  2
2 B 0  Ceq
Leq Ceq
Z=s.gL

f 0 f 0  p 0  1
Z   
 L g  p  L 
B B  0 p  p  Ceq
eq

g 1
Leq  L , Ceq  2
2 B 0  Leq
 Normalized low-pass prototype

 Ladder network synthesized in the normalized domain

assuming unitary reference load (generator side):

Lowpass

Bandpass

Bandpass de-normalization:

g1, g3, …  shunt resonators with Ck=gk/(R1.)

g2, g4, …  series resonators with Lk=R1.gk/
 Specifications in the lowpass domain

Amin Amin

f s ,1  f s ,2  f 02

Stopband Stopband

Passband As3

A
(dB)
Amax

fs1 fp1 fp1 fs2 Freq.

Amin

Passband

f 0  f s ,1 f 0 
Stopband

a    

B  f0 f s ,1  Amax

1 a 
 Approximation of the ideal Lowpass response

General expression for attenuation in the normalized

domain:
A     1   2Cn2   
Cn is called Characteristic Function and define the
approximation of ideal filter response. The parameter
n represents the order of the characteristic function
and corresponds to the number of resonators in the
de-normalized network

Properties:
-1<Cn<1 for -1<<1
|Cn|=1 for =±1
A(±1)=1+’2
 Characteristic function and S parameters

Assuming lossless condition (|S11|2+ |S21|2=1):

2
1 1 S11
A  2
 1   2Cn2      2Cn2     2
1 
S 21 S 21 S 21

1 S11  j  1 S11  j 
2

C   2
2
 Cn  j   
n
  S 21  j    S 21  j 

’ can be expressed as function of Return Loss (RL):

RL
2 
1 S11 max 10 10
1    2
  2

1  S11 max 1  S11 max
2 2 RL

1  10 10
 The characteristic function in s domain

Being s=j,it has:

analytic continuation
Cn2     Cn  j   Cn   j   Cn2  s   Cn  s   Cn   s   Cn  s   Cn*  s 

As a function of the scattering parameters:

1 S11  s 
Cn  s  
  S 21  s 

The S parameters can be expressed as ratio of

characteristic polynomials (monic):

P( s)   F s
S11  s  
F (s)
S 21  Cn  s  
E ( s)
,
E ( s)   P s
 Properties of characteristic polynomials

Roots of F(s): reflection zeros

Roots of P(s): transmission zeros
Roots of E(s): poles

For the unitary of S matrix:

P( s )  P( s )   2  F  s   F   s    E  s   E   s 

E(s) is found from the roots with negative real part of right
hand side (Hurwitz polynomial)

In conclusion: given Cn(), the transmission zeros and the

requested return loss at =±1, the polynomials P(s), F(s),
E(s) can be analitically evaluated
 All-pole characteristic functions (P=cost, Cn=F)

 Butterworth function:
Cn      n
Main property: maximum flatness for =0, monotonic
for |>0
 Chebycheff function:

cos  n  cos 1      1
Cn    
cosh  n  cosh 1      1
Main property: oscillates between ±1 for |<1 (n peaks),
increases monotonically for |>1
Comparison between Butterworth and Chebicheff
70

60 n=5
50
Chebycheff
40
Butterworth
30

20

10

0
0 0.5 1 1.5 2 2.5

Attenuation of Chebycheff function is 6(n-1) dB larger than

Butterworth function for ||>>1
 Evaluation of ’ and n

 Assigned parameters:
- Passband Return Loss in dB (RL)
- Filter bandwidth (B) and center frequency (f0)
- Minimum attenuation (dB) in stopband (AM)
- Frequencies at the beginning of stopband fs1,fs2 with fs1.fs2=(f0)2.

f 0  f s1 f 0 
Evaluation of fs1 in the lowpass domain: s    
B  f 0 f s1 
m
Evaluation of    with  m  10 RL 20
1   m2
AM  RL
n Butterworth characteristic
20 log   s 
AM  RL  6
n Chebycheff characteristic
20 log   s   6
 Synthesis of lowpass prototype

 Butterworth characteristic
  2q  1  
rn  1, g q  2 aq  ,
n
aq  sin  
 2 n 
 Chebycheff characteristic
2
rn  1 (n odd), rn   1   2    (n even)
 
2a1 4aq 1  aq   2q  1  
g1  , gq  , aq  sin  
 bq 1  g q 1  2 n 
 1  1  2 1    q 
  sinh  ln    , bq   2  sin 2  
 2n  1   2  1    n 
  
 De-normalized (bandpass) network with only
series or shunt resonators
For microwave frequencies implementation is requested to
have only one type of resonator (series or shunt). This is
obtained with the introduction of impedance (admittance)
inverters
 General design equation

Introduction of inverter determines additional degrees of

freedom which allows the arbitrary assignments of some
parameters (inverters or resonators parameter).
The condition to be satisfied are expressed by the
following equations:
Series Network Shunt Network
B 0 Ls ,1 B 0Cs ,1
K 01  R1 J 01  G1
f 0 g1 f 0 g1

B 0 Ls ,q  0 Ls ,q 1 B 0Cs ,q  0Cs ,q 1
K q ,q 1  J q ,q 1 
f0 g q  g q 1 f0 g q  g q 1

The bandpass network is assumed symmetric respect

the central inverter (n even) or resonator (n odd)
 Universal coupling parameters

Coupling coefficient:

K q ,q 1 J q ,q 1 B 1
kq ,q 1   
0 Ls ,q  0 Ls ,q 1 0Cs ,q  0Cs ,q 1 f0 g q  g q 1

External Q:
0 Ls ,1 0Cs ,1 g1
QE  2
 2

K 01 R1 J 01 G1 B f 0

0 Ls ,1  Equivalent reactance of series resonators

0Cs ,1  Equivalent susceptance of shunt resonators