Adapted from notes by
ECE 5317-6351 Prof. Jeffery T. Williams
Microwave Engineering
Fall 2019
Prof. David R. Jackson
Dept. of ECE
Notes 22
Filter Design Part 1:
Insertion Loss Method
1
Filters
Consider the following circuit:
Rs = R0
+- Filter RL
This is the basic filter circuit that we will study.
The filter is assumed to be lossless (consists of L and C elements).
2
Filters (cont.)
We can think of transmission lines connected at the input and output
of the filter.
Pinc Ptrans = PL
Rs = R0
+
-
Z 0 = R0 Filter Z 0 = RL RL
Γ
Z in − Z 0
Γ= Lossless filter:
Z in + Z 0
Power loss ratio:
Pinc
PL= Pin= Pinc 1 − Γ ( 2
)
PLR ≡
PL
Pinc 1
=PLR =
(
Pinc 1 − Γ
2
) (1− Γ
2
)
3
Filters (cont.)
Also
2P2− Ptrans PL 1 1
S21= = = = PLR = 2
P1+ Pinc Pinc PLR S21
Insertion Loss (dB): IL ≡ 10 log10 PLR =
−20 log10 S 21
2 2
From conversation of energy: S21 = 1 − S11
Pinc Ptrans = PL
Rs = R0
+
- Z 0 = R0 Filter Z 0 = RL RL
Γ
4
Filters (cont.)
The “insertion loss method” aims at designing a filter to achieve a
particular type of response for the function PLR(ω).
Pinc 1 1
P= = =
LR
PL (
1− Γ
2
) S21
2
IL ≡ 10 log10 PLR =
−20 log10 S 21
2 2
S11 = 1 − S21
Pinc Ptrans = PL
Rs = R0
+- Z 0 = R0 Filter Z 0 = RL RL
Γ
5
Low-Pass Filter
Ideal low-pass filter response
∞
1.0
PLR (ω ) S21 (ω )
Passband Passband Stopband
Stopband
ω 0 ω
1.0
ωc ωc
1
Recall : PLR = 2
S21
6
Filters (cont.)
Types of filters:
Low-pass
High-pass
Bandpass
Bandstop
Approach:
Design the low-pass filter first. The other types of filters
come from this by using a frequency transformation.
(This is discussed in the next set of notes.)
7
Filters (cont.)
Common types of filters:
Butterworth (binomial)*
Chebyshev (type I)*
Chebyshev (type II)
Linear phase**
Elliptic
*Discussed in this set of notes and the next set of notes.
** Discussed in the next set of notes.
8
Note on Reflection Coefficient
Property of the Power Loss Function
Γ ( −ω ) =Γ* (ω ) (This property holds for the Fourier transform of any real-valued
signal. A reflected time-domain signal must be real valued.)
Γ R ( −ω ) + jΓ I ( −ω ) =Γ R (ω ) − jΓ I (ω )
Γ R (ω ) =
even function of ω
Γ I (ω ) =
odd function of ω
2
Γ (ω ) =
even function of ω
PLR (ω ) = even function of ω
(When we design a filter, we have to keep this in mind!)
9
Low-Pass Filter
Low-pass filter response:
PLR (ω )
Passband
Stopband
1+ k2
1.0 ω
ωc
Cutoff frequency
The constant k is somewhat arbitrary, and defines the cutoff frequency.
PLR (ωc )= 1 + k 2
10
Low-Pass Filter
Normalized low-pass filter response:
PLR (ω )
Passband
Stopband
1+ k2
1.0 ω
1
Cutoff frequency
We design the normalized low-pass filter first (ωc = 1), and then use scaling and frequency
transformation to obtain the final filter (low-pass, high-pass, bandpass, bandstop).
(This is done in the next set of notes.)
11
Low-Pass Filter
Low-pass filter responses:
PLR (ω ) Butterworth
Chebyshev
Elliptic
1+ k2
1.0 ω
ωc
Note:
The Chebyshev response shown is called “type 1”, with ripple in the
passband. The “type 2” Chebyshev response has ripple in the stopband.
12
Low-Pass Filter
Comments:
The Butterworth design has the flattest response.
The Chebyshev (type 1) design has a constant ripple in the passband.
The Chebyshev (type 2) has a constant ripple in the stopband.
The elliptic filter has ripple in both the passband and the stopband.
The elliptic filter has the sharpest transition from the passband to the
stopband, for given levels of ripple in the passband and stopband.
13
Low-Pass Filter
Comparison of Filter Responses
Butterworth Chebyshev (type 1)
S21 (ω )
ωc = 0.5 ωc = 0.5
Chebyshev (type 2) Elliptic
S21 (ω )
ωc = 0.5 ωc = 0.5
https://en.wikipedia.org/wiki/Electronic_filter 14
Butterworth Low-Pass Filter
Choose:
2N
ω
PLR (ω ) = 1 + k 2 N = order of the filter*
ωc
The first N-1 derivatives of PLR with respect to ω are zero.
At the band edge: PLR (ωc )= 1 + k 2
Common choice: k =1 ( PLR (ωc ) 2,=
= IL (ωc ) 3dB )
* As seen later, this will also be the number of (L,C) elements in the filter.
15
Butterworth Low-Pass Filter (cont.)
High-frequency limit: ω / ωc → ∞
2N
ω
PLR (ω ) k 2
ωc
Hence
ω 2 N
IL ≡ 10log10 PLR 10log10 k 2
ωc
so
ω
IL (ω ) 20log10 k + 20 N log10
ωc
Conclusion: IL increases at 20N dB/decade in the stopband.
16
Chebyshev Low-Pass Filter
Choose:
ω
PLR (ω )= 1 + k T
2 2
N N = order of the filter*
ωc
There is equal ripple in the passband.
At the band edge: PLR (ωc )= 1 + k 2 Note : Tn (1) = 1
Ripple level: 1 + k 2 ( dB = 10log10 (1 + k 2 ) )
T1 ( x ) = x
cos ( n cos −1 x ) , x ≤ 1 T2 (=
x ) 2 x2 − 1
Tn ( x ) ≡
cosh ( n cosh x ) , x ≥ 1 T3 (=
x ) 4 x 3 − 3x
−1
Tn ( x ) 2 x Tn-1 ( x ) − Tn-2 ( x )
=
* As seen later, this will also be the number of (L,C) elements in the filter. 17
Chebyshev Low-Pass Filter (cont.)
High-frequency limit: ω / ωc → ∞
2N
k 2N ω
2
PLR (ω ) 2
4 ωc
Note : Tn ( x ) 2n −1 x n , x → ∞
Hence
k2 ω
IL ≡ 10log10 PLR 10log10 + 20 N log10 ( 2 ) + 20 N log10
4 ωc
ω
= 20log10 ( k ) − 10log10 ( 4 ) + 20 N log10 ( 2 ) + 20 N log10
ωc
so
ω
IL 20log10 ( k ) + 20 ( N − 1) log10 ( 2 ) + 20 N log10
ωc
18
Chebyshev Low-Pass Filter (cont.)
Comparison:
ω
Butterworth: IL (ω ) 20log10 k + 20 N log10
ωc
ω
Chebyshev: IL 20log10 ( k ) + 20 ( N − 1) log10 ( 2 ) + 20 N log10
ωc
Conclusion:
In the stopband, the IL is larger than for the Chebyshev filter,
but it increases at the same rate for both filters.
19
Two Element Low-Pass Filter
Original circuit:
Rs = R0 L
C
+- RL
RLn = RL / R0
Normalize impedances by R0: Ln = L / R0
Cn = CR0
1 Ln
This
normalization
does not Cn
change the +- RLn
reflection
coefficient.
20
Two Element Low-Pass Filter (cont.)
Normalized circuit:
1 Ln
Cn
+- RLn
Z in ,n
1 Z −1
Z in ,n jω Ln +
= Γ n = in ,n
1 Z in ,n + 1
jωCn +
RLn
RLn 1
= jω Ln + PLR =
1 + jω RLnCn (
1− Γ
2
)
21
Two Element Low-Pass Filter (cont.)
After some algebra:
(1 − RLn )2 1 2 2
2
PLR = 1 +
4 RLn
+ ω ( RLnCn + Ln − 2 LnCn RLn )
2 2
4 RLn
4 1 2 2 2
+ω ( LnCn RLn )
4 RLn
or
(1 − RLn )2 ω 2 ω 2 2 2
PLR = 1 +
+ c
4 RLn ωc 4 RLn
( RLnCn + Ln − 2 LnCn RLn )
2 2
4
ω ωc4 2 2 2
+ ( LnCn RLn )
ωc 4 RLn
ωc = specified cutoff frequency
22
Two Element Low-Pass Filter (cont.)
Hence we have:
2 4
ω ω
PLR =
A0 + A2 + A4
ωc ωc
where
(1 − RLn )2
A= 1 +
0
4 RLn
ωc2 2 2
=A2 ( R
Ln nC + L2
n − 2 Ln n Ln )
C R 2
4 RLn
ωc4 2 2 2
A4 = ( LnCn RLn )
4 RLn
23
Two Element Low-Pass Butterworth
2 4
ω ω
PLR =
A0 + A2 + A4
ωc ωc
Choose Butterworth response (k = 1):
2N
ω
PLR = 1 + N =2
ωc
Hence
Note:
The order N of the filter is the same
as the number of (L,C) elements.
(1 − RLn )2
1 +
A0 = = 1
4 RLn
ωc2 2 2
( RLnCn + Ln − = ) 0
2 2
=A2 2 LnCn RLn
4 R
Ln
ωc4 2 2 2
=A4 = ( LnCn RLn ) 1
4 R
Ln
24
Two Element Low-Pass Butterworth (cont.)
Three equations:
(1 − RLn )2
1 +
A0 = = 1
4 RLn
ωc2 2 2
( RLnCn + Ln − = )
2 2
=A2 2 LnCn RLn 0
4 RLn
ωc4 2 2 2
=A4 = ( LnCn RLn ) 1
4 RLn
Solution:
A0 equation: A2 equation: A4 equation:
2
RLn = 1 Ln = Cn L=
n C=
n
ωc
25
Two Element Low-Pass Butterworth (cont.)
Unnormalizing:
RLn = RL / R0 RL = RLn R0
RLn = 1
Ln = L / R0 L = Ln R0 where
L= C= 2 / ωc
Cn = CR0 C = Cn / R0 n n
Hence:
Rs = R0 L
RL = R0
2 C
L = R0 +- RL
ωc
1 2
C=
R0 ωc
26
Two Element Low-Pass Butterworth (cont.)
Example
Rs = R0 L
Design a low-pass
Butterworth filter C
+- RL
N =2
k =1
f c = 1 GHz
R=
0 R=
L 50 Ω 1 ⇒ PLR (ωc ) =
Recall : k = 2 ⇒ IL (ωc ) =
3dB
RL = R0 R= R= 50 [ Ω ]
L 0
2 2 R=
L R=
0 50 [ Ω ]
L = R0
9 [ ]
L = 50 H
ωc 2π 10 L = 11.25 [ nH ]
1 2 1 2 C = 4.50 [ pF]
9 [ ]
C= C= F
R0 ωc 50 2π 10
27
Two Element Low-Pass Butterworth (cont.)
Results from Ansys Designer
PLR 40 dB/decade = 12dB/octave
PLR = 2 ( 3dB)
PLR [ dB]
f [ GHz ]
28
Two Element Low-Pass Butterworth (cont.)
Results from Ansys Designer
S21 ( dB)
S21 = −3dB
S11 ( dB)
S 21 −40 dB/decade =
−12dB/octave
f [ GHz ]
1 − S 21 ( lossless )
dB 2 2
Recall : IL =
− S 21 , S11 =
29
Two Element Low-Pass Chebyshev
2 4
ω ω
PLR =
A0 + A2 + A4
ωc ωc
Choose Chebyshev response (arbitrary k value):
ω
PLR (ω )= 1 + k T
2 2
N N =2
ωc
so
Note :
ω 4 ω
2
PLR (ω ) =
1 + k 4 − 4 + 1
2
T2 (=
x ) 2 x2 −1
ωc ωc
T22 ( x ) = 4 x 4 − 4 x 2 + 1
Hence:
A0 = 1 + k 2
A2 = −4k 2
A4 = 4k 2
30
Two Element Low-Pass Chebyshev (cont.)
Hence:
(1 − RLn )2
1 +
A0 = = 1+ k 2
4 RLn
ωc2 2 2
( RLnCn + Ln − 2 LnCn RLn ) =
2 2
A2 =
−4k 2
4 RLn
ωc4 2 2 2
( LnCn RLn ) 4k
2
=A4 =
4 RLn
Solution (please see the Appendix):
k 1
2
1 + 2k ± 2k 1 + k
RLn = 2 LnCn = 4 2
ωc RLn
2 4 RLn
2
2 1 2 pR 2 − 4k 2 4 RLn 2 2 2
C
= n 2 Ln 2 ± 4k 2 − 2 pRLn − 4 RLn p
2 RLn ωc ωc
31
Appendix
Chebyshev solution:
(1 − RLn )2
1 +
A0 = = 1+ k 2
4 RLn
ωc2 2 2
( RLnCn + Ln − 2 LnCn RLn ) =
2 2
A2 =
−4k 2
4 RLn
ωc4 2 2 2
( LnCn RLn ) 4k
2
=A4 =
4 RLn
2
1 − 2 RLn + RLn k 2 4 RLn
=
A0 equation:
(1 − RLn ) − RLn ( 2 + 4k 2 ) + 1 =
2 2
RLn 0
1+ 1+ k 2
=
4 RLn
1 + 2k 2 ± 2k 1 + k 2
RLn =
(1 − RLn )
2
= k2
4 RLn Note :
RL ≠ R0 for N = even.
(1 − RLn )
2
k 2 4 RLn
= (It turns out that RL = R0 for N = odd.)
32
Appendix (cont.)
(1 − RLn )2
1 +
A0 = = 1+ k 2
4 RLn
ωc2 2 2
( RLnCn + Ln − 2 LnCn RLn ) =
2 2
A2 =
−4k 2
4 RLn
ωc4 2 2 2
( LnCn RLn ) 4k
2
=A4 =
4 RLn
A4 equation:
2 2 k2 1
L C = 16 4
n n
ωc RLn
k 1
LnCn = 4 2
ωc RLn
33
Appendix (cont.)
(1 − RLn )2
1 +
A0 = = 1+ k 2
4 RLn
ωc2 2 2
( RLnCn + Ln − 2 LnCn RLn ) =
2 2
A2 =
−4k 2
4 RLn
ωc4 2 2 2
( LnCn RLn ) 4k
2
=A4 =
4 RLn
A2 equation:
ωc2 2 2
( RLnCn + Ln − 2 LnCn RLn ) =
2 2
−4k 2
4 RLn
4 RLn
( R C + L − 2 LnCn R ) =
2
Ln
2
n
2
n
2
−4k 2
Ln
2
ωc
4R
(R 2
Ln Cn2 + L2n ) =
−4k 2 2Ln + 2 ( LnCn ) RLn
2
(The RHS is known.)
ωc
34
Appendix (cont.)
A2 equation (cont.):
4 RLn k 1
( 2
R C +2
L ) =
−42
k 2
2 2
+ 2 pRLn
4 2
p ≡ LnCn =
ωc RLn
Ln n n
ωc
2 2 p2 2 4 RLn 2
−4k 2 + 2 pRLn
RLnCn + 2 =
Cn ωc
( R C + p ) = C −4k 2 4ωR2Ln
2
Ln
4
n
2 2
n
2
+ 2 pRLn
c
2 4 RLn
2 2
R x + x 4k 2
Ln
2 2
− 2 pRLn + p =0 (x ≡ C )
2
ωc
n
2 4 RLn
2
1 2 2 4 RLn 2 2 2
x
= 2
2 pRLn − 4 k 2 ± 4k 2 − 2 pRLn − 4 RLn p
2 RLn ωc ωc
35
