Introduction to RF

Filter Design
RF Electronics
Spring, 2018
Robert R. Krchnavek
Rowan University
 Objectives

• Understand the fundamental concepts and

definitions for filters.

• Know how to design filters using tabulated

parameters for common filter types.

• Know how to convert lumped-element filter designs

into distributed-element filters.
 Filter Configurations

ω where ωc is defined as the cutoff frequency for

Ω= low-pass and high-pass filters and the center
ωc frequency for bandpass and bandstop filters.
 Low-Pass Filters
Profiles for Three Common Types

• Binomial or Butterworth - easy to implement; monotonic profile; requires

numerous elements to get step profile.
• Chebyshev - equal amplitude variations; steeper profile than Butterworth.
• Elliptic or Cauer - amplitude variations in both stopband and passband; steepest
profile; complicated to design.
Bandpass Filter - Profile
 Filter Definitions

• Insertion loss - how much power is

IL = 10 log
Pin
lost in going through the filter. PL
• Ripple - the flatness of the signal
in the passband.
• Bandwidth - the width of the BW 3dB
= 3dB
fU − 3dB
fL
passband.
• Shape factor - the sharpness of
60dB
the filter response. BW
SF =
• Rejection - the attenuation of the BW 3dB

undesired signals. 60dB 60dB

fU fL
• Quality factor - see next slide.
−
= 3dB 3dB
fU − fL
 Q - Quality Factor
The quality factor, or Q, is a parameter that is used to
describe the selectivity of the filter.

The unloaded Q is defined as

! "
maximum energy stored in the filter at fC
Q = 2πfC
power lost in the filter

The loaded Q is defined as

! "
maximum energy stored in the filter at fC
QLD = 2πfC
power lost in the filter and to the external circuit

A higher Q indicates a more selective filter. Details to follow.

Series RLC Bandpass Filter

Find VR1 for this circuit.

 Series RLC Bandpass Filter
! "
maximum energy stored in the filter at fC
Q = 2πfC
power lost in the filter
1 2
LIp = maximum energy stored in the filter at fC
2

2
Irms R = power lost in the filter
The resonant frequency is the
1 2
2
Irms = Ip frequency where the imaginary
2 component of the impedance is
L L equal to 0: 1
Q = 2πfC = ωC ȷωC L + =0
R R ȷωC C

1
ωC = √
LC
 Series RLC Bandpass Filter
Solving for the frequencies at which VR1 is down 3 dB yields
!" #2 !" #2
R R 1 −R R 1
ωU = + + and ωL = + +
2L 2L LC 2L 2L LC

The bandwidth is given by

R 1 R
BW = ωU − ωL = and BW = fU − fL =
L 2π L

And, using our previous result for unloaded Q, we see the

relationship between Q and BW is given by
L fC fC
Q = 2πfC = or BW =
R BW Q
Series and Parallel Resonators
Quantitative Analysis of a Series RLC
Bandpass Filter

VL
H(ω) =
VG

VL ZL
H(ω) = =
VG (ZG + ZL ) + R + ȷ [ωL − 1/(ωC)]

This filter is different from the previous one

because of the addition of ZG and ZL.
 Series RLC Bandpass Filter

where RE = ZG + ZL

Define three different Q factors:

ωC L
Unloaded, internal, or filter Q QF =
R
ωC L
External Q QE =
RE
ωC L
Loaded Q QLD =
R + RE
 Series RLC Bandpass Filter

Note: ZL=ZG=50Ω, R=20Ω, L=5 nH, and C=2 pF.

 Series RLC Bandpass Filter

• Series RLC bandpass filter is easy to analyze.

• Minimum attenuation at the resonance point.

• HOWEVER, the transition from passband to stop

band is not very sharp (large shape factor.)
 Butterworth Filter

• One of a series of special filter designs that

consist of more elements than a simple RLC and
give better control over the filter parameters.

• Also known as a maximally flat filter - no ripple.

• Strategy

• First, do the normalized, low-pass filter.

• Second, implement the desired form through a

frequency scaling.

• Third, if necessary, create distributed elements.

 Butterworth Filter

⎧ ⎫
⎨ inductance for series inductor ⎬
gm = capacitance for shunt capacitor
(m ≡ 1, . . . , N )
⎩ ⎭
! "
load resistance if the last element is a shunt capacitor
gN +1 =
load conductance if the last element is a series inductor

Two different networks that are used to implement the

Butterworth filter.
 Butterworth Filter
Pin ! 2 2N
"
IL = 10 log = 10 log 1 + a Ω
PL

The coefficient “a” is usually taken to be 1 so that the IL is 3

dB at the cutoff frequency.
 Butterworth Filter

The attenuation vs frequency as a function of the number

of stages. Note: this design does NOT result in a linear
phase relationship.
 Butterworth Filter

Coefficients for a maximally flat response.

 Butterworth Filter

Coefficients for a linear phase response.

 Comments

• Coefficients for a 3 dB Chebyshev filter design are

in Table 5.4 (a).

• Coefficients for a 0.5 dB Chebyshev filter design

are in Table 5.4 (b).

• The generic, multisection, normalized element

circuits designs are the same for Butterworth and
Chebyshev filters.
 Butterworth Example

Design a 4th-order, low-pass, standard (maximally flat),

3 dB Butterworth filter.
 Frequency and Impedance
Transformations

• The normalized values need to be modified to

produce:

• The desired response (low-pass, high-pass, etc.)

• At the desired center frequency.

• With an impedance that is realistic.

 Frequency Transformation

• All of the different filter types are derived from

the low-pass filter.

• The key is to determine a transformation function

that maps the normalized, low-pass design into the
appropriate (low-pass, high-pass, etc.) at the
desired frequency.

• New values for L and C are obtained by maintaining

the same Z through the transformation.
Frequency Transformation – Low Pass
ωc is the new cutoff frequency

! = ⌦!c

The impedances should remain the same:

!
ZL = |⌦L = | L = |!Lnew
!c
L
Lnew =
!c

1 !c 1
ZC = = =
|⌦C |!C |!Cnew

C
Cnew =
!c
Normalized Low-Pass Low-Pass

ω = Ωωc
ω
ZL = ȷΩL = ȷ L = ȷωLnew
ωc
L
Lnew =
ωc
1 ωc 1
ZC = = =
ȷΩC ȷωC ȷωCnew
C
Cnew =
ωc
Frequency Transformation – High Pass
ωc is the new cutoff frequency

!c
!=
⌦

Again, the impedances should

remain the same giving . . . .
Normalized Low-Pass High-Pass

−ωc
ω=
Ω
1
Cnew =
ωc L
1
Lnew =
ωc C
 Frequency Transformations
Bandpass and Bandstop

• These transformations are more complex. See the

textbook for both.
Frequency Transformations
Summary
 Impedance Transformation

• For the Butterworth designs, the source and load

resistances have a value of 1.

• For the Chebyshev designs, even-number ordered

designs have a non-unity load.

• Impedance transformation is the process of

adjusting all the elements to account for different
source and load impedances.
 Impedance Transformation

Assume the source impedance, RG, is scaled from 1 in the

original design to RG,new. Then, the new values are:

RG, new = 1RG, new

Lnew = LRG, new
C
Cnew =
RG, new
RL, new = RL RG, new
 Distributed-Element Filters

• Above approximately 1 GHz, lumped-element filter

design is problematic because the elements are
approaching a significant fraction of λ.

• Distributed-element filters are common.

• One approach is to design the lumped-element

filter and then convert it to a distributed-element
realization.
 Distributed-Element Filters

• Assume you have a lumped-element filter design

that you want to build as a distributed-element
filter.

• Recall our expression for the impedance of a

terminated (lossless) transmission line:

ZL + ȷZ0 tan βd
Z(d) = Z0
Z0 + ȷZL tan βd
• If ZL = 0, then

Z(d) = ȷZ0 tan βd

• If ZL = ∞, then

Z(d) = −ȷZ0 cot βd

• The electrical length, βd, can be put in the

following form
2π 2π 2πf
βd = d= d= d
λ vp /f vp
• Assume we chose a line that is 1/8 of a wavelength
d = λ/8
• The expression for impedance explicitly in terms of
frequency is then (ZL = 0)

2πf vp π f
Z(d) = ȷZ0 tan = ȷZ0 tan
vp f0 8 4 f0
• The impedance of the stub must equal the impedance of
the lumped element

π f
ȷωL = ȷZ0 tan
4 f0
• For a capacitive element, you could use the open-
circuited transmission line.

• One significant difference is the frequency range is

shortened because the tan function is periodic.
• For a d = short-circuited line, we have the
following for the Richard’s transform:

• For a d= open-circuited line, we the following

for the Richard’s transform:
Distributed-Element Filters - Physical
Realization

• Using transmission line sections to build the filter

may require sections of line that separate elements
from each other. These are called unit elements.

• The unit elements have an electrical length of

π f
βd =
4 f0
• We also need to be able to create distributed
element sections for difficult-to-replace lumped
elements such as series inductors. We use Kuroda’s
identities for this.
 Distributed Filter Implementation
Design a 4th-order, low-pass, standard
(maximally flat), 3 dB Butterworth filter. It
should have a cutoff frequency of 1 GHz.

1. Select the normalized filter order and parameters to

meet the design criteria.

2. Replace inductances and capacitances with equivalent

λ/8 transmission lines.

3. Convert series stub lines to shunt stub lines through

Kuroda’s identities.

4. Denormalize and select equivalent microstrip lines.

Distributed Filter Implementation

R=1Ω 1.8478 H 0.7654 H

G=1
Ω

0.7654 F 1.8478 F

1. Select the normalized filter

order and parameters to meet
the design criteria.
Distributed Filter Implementation
Z0=1.8478 Z0=0.7654

R=1Ω

G=1
Ω

Y0=0.7654 Y0= 1.8478

2. Replace inductances and

capacitances with equivalent λ/
8 transmission lines.
Distributed Filter Implementation
Z0=1.8478 Z0=0.7654

ZUE=1Ω ZUE=1Ω
R=1Ω

G=1
Ω
UE UE

Y0=0.7654 Y0=1.8478

3. Convert series stub lines to shunt stub lines

through Kuroda’s identities.
Distributed Filter Implementation
Z0=1.8478 Z0=0.7654

ZUE=1Ω
R=1Ω

G=1
Ω
UE UE

Y0=1.8478

3. Convert series stub lines to shunt stub lines

through Kuroda’s identities.
Distributed Filter Implementation
Z0=1.8478

R=1Ω

G=1
Ω
UE UE

Y0=1.8478

3. Convert series stub lines to shunt stub lines

through Kuroda’s identities.
 Distributed Filter Implementation
Z0=1.8478

ZUE=1Ω
R=1Ω

G=1
Ω
UE UE UE

Y0=1.8478

3. Convert series stub lines to shunt stub lines

through Kuroda’s identities.
 Distributed Filter Implementation

R=1Ω

G=1
Ω
UE UE UE

Y0=1.8478

3. Convert series stub lines to shunt stub lines

through Kuroda’s identities.