EE247 - Lecture 2

Filters
• From last lecture:
– Dynamic range of analog circuits

• Filters:
– Nomenclature
– Specifications
• Quality factor
• Frequency characteristics
• Group delay
– Filter types
• Butterworth
• Chebyshev I & II
• Elliptic
• Bessel
– Group delay comparison example
– Biquads
EECS 247 Lecture 2: Filters © 2008 H.K. Page 1

Nomenclature
Filter Types wrt Frequency Range Selectivity

Lowpass Highpass Bandpass Band-reject

(Notch)
All-pass
H ( jω )
H ( jω ) H ( jω ) H ( jω ) H ( jω )

ω ω ω ω ω

Provide frequency selectivity Phase shaping

or equalization

EECS 247 Lecture 2: Filters © 2008 H.K. Page 2

 Filter Specifications
• Frequency characteristics (lowpass filter):
– Passband ripple (Rpass)
– Cutoff frequency or -3dB frequency
– Stopband rejection
– Passband gain
• Phase characteristics:
– Group delay
• SNR (Dynamic range)
• SNDR (Signal to Noise+Distortion ratio)
• Linearity measures: IM3 (intermodulation distortion), HD3
(harmonic distortion), IIP3 or OIP3 (Input-referred or output-
referred third order intercept point)
• Area/pole & Power/pole

EECS 247 Lecture 2: Filters © 2008 H.K. Page 3

Lowpass Filter Frequency Characteristics

H ( jω ) [ d B ]
Passband Ripple (Rpass) f−3dB
H (0 )
Passband 3dB
Transition
Gain H ( jω ) Band
Stopband
Rejection

H ( jω )
f
0
fc f stop
Frequency (Hz)
x 10
Stopband
Passband
Frequency

EECS 247 Lecture 2: Filters © 2008 H.K. Page 4

 Quality Factor (Q)

• The term quality factor (Q) has different

definitions in different contexts:
– Component quality factor (inductor &
capacitor Q)
– Pole quality factor
– Bandpass filter quality factor

• Next 3 slides clarifies each

EECS 247 Lecture 2: Filters © 2008 H.K. Page 5

Component Quality Factor (Q)

• For any component with a transfer function:

H ( jω ) = 1
R (ω ) + jX (ω )

• Quality factor is defined as:

X (ω ) Energy Stored
Q= → p e r u n i t t i me
R (ω ) Aver age Po w er D i s s i pa t i o n

EECS 247 Lecture 2: Filters © 2008 H.K. Page 6

 Component Quality Factor (Q)
Inductor & Capacitor Quality Factor
• Inductor Q :

YL = 1 QL = ω L
Rs + jω L Rs Rs L

• Capacitor Q :
Rp
ZC = 1 QC = ω CR p
1 + jωC
Rp C

EECS 247 Lecture 2: Filters © 2008 H.K. Page 7

Pole Quality Factor

jω
• Typically filter
s-Plane
singularities include
pairs of complex
conjugate poles.
ωx ωP
• Quality factor of
complex conjugate
poles are defined as: σx σ

ωp
QP o l e =
2σ x

EECS 247 Lecture 2: Filters © 2008 H.K. Page 8

 Bandpass Filter Quality Factor (Q)
H ( jf )
Q= fcenter /Δf


0
Magnitude [dB]



-3dB


Δf = f2 - f1



 

0.1 f1 fcenter
1 f2 Frequency
10

EECS 247 Lecture 2: Filters © 2008 H.K. Page 9

What is Group Delay?

• Consider a continuous time filter with s-domain transfer function G(s):
jθ(ω)
G(jω) ≡ ⏐G(jω)⏐e

• Let us apply a signal to the filter input composed of sum of two

sinewaves at slightly different frequencies (Δω<<ω):

vIN(t) = A1sin(ωt) + A2sin[(ω+Δω) t]

• The filter output is:

vOUT(t) = A1 ⏐G(jω)⏐ sin[ωt+θ(ω)] +

A2 ⏐G[ j(ω+Δω)]⏐ sin[(ω+Δω)t+ θ(ω+Δω)]

EECS 247 Lecture 2: Filters © 2008 H.K. Page 10

 What is Group Delay?
θ(ω)
vOUT(t) = A1 ⏐G(jω)⏐ sin {ω[ t+ ω ]} +
θ(ω+Δω)
+ A2 ⏐G[ j(ω+Δω)]⏐ sin
{ (ω+Δω) [ t + ω+Δω ]}
Δω
then Δω 2
Since
ω <<1 ω [ ]
Æ0

θ(ω+Δω) dθ(ω)
ω+Δω
≅
[ θ(ω)+
dω
Δω
][ ω1 (1 - Δωω ) ]
θ(ω) dθ(ω) θ(ω)
≅
ω + ( dω - ω ) Δωω
EECS 247 Lecture 2: Filters © 2008 H.K. Page 11

What is Group Delay?

Signal Magnitude and Phase Impairment

θ(ω)
vOUT(t) = A1 ⏐G(jω)⏐ sin { ω [t + ω ]} +
θ(ω) Δω
{ (ω+Δω) [ t + θ(ω)
dθ(ω)
+ A2 ⏐G[ j(ω+Δω)]⏐sin
ω +( dω - ω ω ) ]}
• τPD ≡ -θ(ω)/ω is called the “phase delay” and has units of time

• If the second term is zero, then the filter’s output at frequency ω+Δω and
the output at frequency ω are each delayed in time by -θ(ω)/ω

• If the second term in the phase of the 2nd sin wave is non-zero, then the
filter’s output at frequency ω+Δω is time-shifted differently than the
filter’s output at frequency ω
Æ “Phase distortion”

EECS 247 Lecture 2: Filters © 2008 H.K. Page 12

 What is Group Delay?
Signal Magnitude and Phase Impairment

• Phase distortion is avoided only if:

dθ(ω) θ(ω)
dω - ω = 0

• Clearly, if θ(ω)=kω, k a constant, Æ no phase distortion

• This type of filter phase response is called “linear phase”
ÆPhase shift varies linearly with frequency
• τGR ≡ -dθ(ω)/dω is called the “group delay” and also has units
of time. For a linear phase filter τGR ≡ τPD =k
Æ τGR= τPD implies linear phase
• Note: Filters with θ(ω)=kω+c are also called linear phase filters, but
they’re not free of phase distortion

EECS 247 Lecture 2: Filters © 2008 H.K. Page 13

What is Group Delay?

Signal Magnitude and Phase Impairment

• If τGR= τPD Æ No phase distortion

[ (
vOUT(t) = A1 ⏐G(jω)⏐ sin ω t - τGR)] +
+ A ⏐G[ j(ω+Δω)]⏐ sin [ (ω+Δω) ( t - τ )]
2 GR

• If also⏐G( jω)⏐=⏐G[ j(ω+Δω)]⏐ for all input frequencies within

the signal-band, vOUT is a scaled, time-shifted replica of the
input, with no “signal magnitude distortion” :

• In most cases neither of these conditions are exactly realizable

EECS 247 Lecture 2: Filters © 2008 H.K. Page 14

 Summary
Group Delay

• Phase delay is defined as:

τPD ≡ -θ(ω)/ω [ time]
• Group delay is defined as :
τGR ≡ -dθ(ω)/dω [time]

• If θ(ω)=kω, k a constant, Æ no phase distortion

• For a linear phase filter τGR ≡ τPD =k

EECS 247 Lecture 2: Filters © 2008 H.K. Page 15

Filter Types wrt Frequency Response

Lowpass Butterworth Filter
0
Magnitude (dB)

• Maximally flat amplitude -20

within the filter passband
-40
N
d H( jω )
=0 -60
dω
Normalized Group Delay

0 5
ω =0
Phase (degrees)

-200 3

• Moderate phase distortion

-400 1
0 1 2
Normalized Frequency
Example: 5th Order Butterworth filter

EECS 247 Lecture 2: Filters © 2008 H.K. Page 16

 Lowpass Butterworth Filter

• All poles jω

s-plane
• Number of poles equal to
filter order

• Poles located on the unit σ

circle with equal angles

pole

Example: 5th Order Butterworth Filter

EECS 247 Lecture 2: Filters © 2008 H.K. Page 17

Filter Types
Chebyshev I Lowpass Filter

0
Magnitude [dB]

• Chebyshev I filter
– Ripple in the passband -20

– Sharper transition band

compared to Butterworth (for -40
the same number of poles)
0 35
Normalized Group Delay
Phase [degrees]

– Poorer group delay

compared to Butterworth -200

– More ripple in passband Æ

poorer phase response -400
0
0 1 2
Normalized Frequency
Example: 5th Order Chebyshev filter

EECS 247 Lecture 2: Filters © 2008 H.K. Page 18

 Chebyshev I Lowpass Filter Characteristics

jω
s-plane
• All poles

• Poles located on an ellipse

inside the unit circle
σ
• Allowing more ripple in the
passband:
Narrower transition band
Sharper cut-off
Higher pole Q
Chebyshev I LPF 3dB passband ripple
Poorer phase response Chebyshev I LPF 0.1dB passband ripple
Example: 5th Order Chebyshev I Filter

EECS 247 Lecture 2: Filters © 2008 H.K. Page 19

Filter Types
Chebyshev II Lowpass

• Chebyshev II filter 0
Magnitude (dB)

– No ripple in passband
-20
– Nulls or notches in -40
stopband
-60
– Sharper transition band 0
compared to
-90
Phase (deg)

Butterworth
-180
– Passband phase more -270
linear compared to
-360
Chebyshev I 0 0.5 1 1.5 2
Normalized Frequency

Example: 5th Order Chebyshev II filter

EECS 247 Lecture 2: Filters © 2008 H.K. Page 20

 Filter Types
Chebyshev II Lowpass
jω
• Poles & finite zeros
– No. of poles n
(n −> filter order) s-plane
– No. of finite zeros: n-1

• Poles located both inside

& outside of the unit circle
σ
• Complex conjugate zeros
located on jω axis

• Zeros create nulls in

Example:
stopband pole
5th Order
Chebyshev II Filter zero

EECS 247 Lecture 2: Filters © 2008 H.K. Page 21

Filter Types
Elliptic Lowpass Filter
0

• Elliptic filter
Magnitude (dB)

-20

-40
– Ripple in passband
-60
– Nulls in the stopband 0
Phase (degrees)

– Sharper transition band

compared to Butterworth & -200
both Chebyshevs

– Poorest phase response -400

0 1 2
Normalized Frequency
Example: 5th Order Elliptic filter

EECS 247 Lecture 2: Filters © 2008 H.K. Page 22

 Filter Types
Elliptic Lowpass Filter

• Poles & finite zeros jω

– No. of poles: n s-plane
– No. of finite zeros: n-1

• Zeros located on jω axis

σ
• Sharp cut-off
Narrower transition
band
Pole Q higher
compared to the
previous filter types Pole
Zero
Example: 5th Order Elliptic Filter

EECS 247 Lecture 2: Filters © 2008 H.K. Page 23

Filter Types
Bessel Lowpass Filter

• Bessel
jω s-plane

– All poles

– Poles outside unit circle

σ
– Relatively low Q poles

– Maximally flat group delay

Pole
– Poor out-of-band attenuation
Example: 5th Order Bessel filter

EECS 247 Lecture 2: Filters © 2008 H.K. Page 24

 Magnitude Response Behaviour
as a Function of Filter Order
Example: Bessel Filter
0
d
-10
se
rea
-20 c
In n=1
-30 er
rd
Magnitude [dB]

-40 r O
lte nÆ Filter order
-50 Fi 2
-60
3
-70
4
-80
5
6
-90
7
-100
0.1 1 10 100
Normalized Frequency

EECS 247 Lecture 2: Filters © 2008 H.K. Page 25

Filter Types
Comparison of Various Type LPF Magnitude Response

0
Magnitude (dB)

-20

-40

-60

0 1 2
Magnitude (dB)

Normalized Frequency

Bessel
All 5th order filters with same corner freq. Butterworth
Chebyshev I
Chebyshev II
Elliptic

EECS 247 Lecture 2: Filters © 2008 H.K. Page 26

 Filter Types
Comparison of Various LPF Singularities
jω
Poles Bessel
Poles Butterworth
Poles Elliptic
s-plane
Zeros Elliptic
Poles Chebyshev I 0.1dB

EECS 247 Lecture 2: Filters © 2008 H.K. Page 27

Comparison of Various LPF Groupdelay

5
28
Bessel Chebyshev I
0.5dB Passband Ripple

1
12
10 Butterworth

4
1

1
Ref: A. Zverev, Handbook of filter synthesis, Wiley, 1967.

EECS 247 Lecture 2: Filters © 2008 H.K. Page 28

 Group Delay Comparison
Example

• Lowpass filter with 100kHz corner frequency

• Chebyshev I versus Bessel

– Both filters 4th order- same -3dB point

– Passband ripple of 1dB allowed for Chebyshev I

EECS 247 Lecture 2: Filters © 2008 H.K. Page 29

Magnitude Response
4th Order Chebyshev I versus Bessel

0
Magnitude (dB)

-20

-40

-60 4th Order Chebychev 1

4th Order Bessel
4 5 6
10 10 10
Frequency [Hz]

EECS 247 Lecture 2: Filters © 2008 H.K. Page 30

 Phase Response
4th Order Chebyshev I versus Bessel
0

-50
Phase [degrees]
-100

-150 4th order Bessel

-200

-250
4th order Chebyshev 1
-300

-350 0 50 100 150 200

Frequency [kHz]

EECS 247 Lecture 2: Filters © 2008 H.K. Page 31

Group Delay
4th Order Chebyshev I versus Bessel

14

12
Group Delay [usec]

10

8 4th order
Chebyshev 1
6

4 4th order Bessel

0
10 100 1000
Frequency [kHz]

EECS 247 Lecture 2: Filters © 2008 H.K. Page 32

 Step Response
4th Order Chebyshev I versus Bessel

1.4

1.2
4th order Bessel
1
Amplitude

0.8 4th order

Chebyshev 1
0.6

0.4

0.2

00 5 10 15 20
Time (usec)

EECS 247 Lecture 2: Filters © 2008 H.K. Page 33

Intersymbol Interference (ISI)

ISIÆ Broadening of pulses resulting in interference between successive transmitted
pulses
Example: Simple RC filter

EECS 247 Lecture 2: Filters © 2008 H.K. Page 34

 Pulse Impairment
Bessel versus Chebyshev
1.5 1.5
Input
1 Output 1

0.5 0.5

0 0

-0.5 -0.5

-1 -1

-1.5 -1.5
1.1 1.2 1.3 1.4 1.5 1.6 1.7 1.8 1.9 2 1.1 1.2 1.3 1.4 1.5 1.6 1.7 1.8 1.9 2
-4 -4
x 10 x 10

4th order Bessel 4th order Chebyshev I

Note that in the case of the Chebyshev filter not only the pulse has broadened but it
also has a long tail
Æ More ISI for Chebyshev compared to Bessel

EECS 247 Lecture 2: Filters © 2008 H.K. Page 35

Response to Psuedo-Random Data

Chebyshev versus Bessel
1.5

1
1111011111001010000100010111101110001001

Input Signal: 0.5

Symbol rate 1/130kHz 0

-0.5

-1

-1.5
0 0.2 0.4 0.6 0.8 1 1.2 1.4
-4
x 10

1.5 1.5

1 1

0.5 0.5

0 1111011111001010000100010111101110001001 0
1111011111001010000100010111101110001001

-0.5 -0.5

-1 -1

-1.5 -1.5
0 0.2 0.4 0.6 0.8 1 1.2 1.4 0 0.2 0.4 0.6 0.8 1 1.2 1.4
-4 -4
x 10 x 10
4th order Bessel 4th order Chebyshev I

EECS 247 Lecture 2: Filters © 2008 H.K. Page 36

 Summary
Filter Types
– Filters with high signal attenuation per pole poor
phase response
– For a given signal attenuation, requirement of
preserving constant groupdelay ÆHigher order filter
• In the case of passive filters higher component count
• For integrated active filters higher chip area &
power dissipation

– In cases where filter is followed by ADC and DSP

• Possible to digitally correct for phase impairments incurred
by the analog circuitry by using digital phase equalizers &
thus reducing the required filter order

EECS 247 Lecture 2: Filters © 2008 H.K. Page 37

RLC Filters
R Vo
• Bandpass filter (2nd order):
Vin L C
s
Vo = RC
Vin s2 + ωo s + ωo2
Q
jω
ωo = 1 LC
s-Plane
Q = ωo RC = R
Lω o
σ

Singularities: Pair of complex conjugate poles

Zeros @ f=0 & f=inf.

EECS 247 Lecture 2: Filters © 2008 H.K. Page 38

 RLC Filters
Example
• Design a bandpass filter with:
R Vo
ƒ Center frequency of 1kHz
Vin C
ƒ Quality factor of 20 L

• First assume the inductor is ideal

• Next consider the case where the inductor has series R
resulting in a finite inductor Q of 40
• What is the effect of finite inductor Q on the overall filter
Q?

EECS 247 Lecture 2: Filters © 2008 H.K. Page 39

RLC Filters
Effect of Finite Component Q
1 Qfilt.=20 (ideal L)
= 1 + 1
Q filt Qid ea l Qin d .
filt
Qfilt. =13.3 (QL.=40)

Need to have component Q much higher

compared to desired filter Q

EECS 247 Lecture 2: Filters © 2008 H.K. Page 40

 RLC Filters

R Vo

Vin L C

Question:
Can RLC filters be integrated on-chip?

EECS 247 Lecture 2: Filters © 2008 H.K. Page 41

Monolithic Inductors
Feasible Quality Factor & Value

Typically, on-chip
inductors built as
spiral structures out
of metal/s layers

QL= (ω L/R)

QL measured at
frequencies of
operation ( >1GHz)

Feasible monolithic inductor in CMOS tech. <10nH with Q <7

™Ref: “Radio Frequency Filters”, Lawrence Larson; Mead workshop presentation 1999

EECS 247 Lecture 2: Filters © 2008 H.K. Page 42

 Integrated Filters
• Implementation of RLC filters in CMOS technologies requires
on-chip inductors
– Integrated L<10nH with Q<10
– Combined with max. cap. 20pF
ÆLC filters in the monolithic form feasible: freq>350MHz
Æ(Learn more in EE242 & RF circuit courses)

• Analog/Digital interface circuitry require fully integrated filters

with critical frequencies << 350MHz

• Hence:

Need to build active filters built without inductor

EECS 247 Lecture 2: Filters © 2008 H.K. Page 43

Filters
2nd Order Transfer Functions (Biquads)
• Biquadratic (2nd order) transfer function:
1
H( s ) =
s s2
1+ +
ωPQP ωP2
H ( jω ) ω =0 = 1
1
H( jω ) = H ( jω ) ω →∞ = 0
2
⎛ ω 2 ⎞ ⎛ ω ⎞2
⎜1− 2 ⎟ + ⎜ H ( jω ) ω =ω = QP
⎜ ω ⎟ ⎝ ωPQP ⎟⎠ P
⎝ P⎠

ωP ⎛ 2⎞
Bi quad pole s @: s = − ⎜ 1 ± 1 − 4QP ⎟
2QP ⎝ ⎠

Note : for QP ≤ 1 pole s are re al, c o mplex othe rwise

2

EECS 247 Lecture 2: Filters © 2008 H.K. Page 44

 Biquad Complex Poles
QP > 1 → Complex conjugate poles:
2

ωP ⎛ S-plane
s=− 2 ⎞ jω
⎜1 ± j 4QP − 1 ⎟
2QP ⎝ ⎠
poles
d
Distance from origin in s-plane:

⎛ω ⎞
2 σ
d 2 = ⎜⎜ P ⎟⎟ (1 + 4QP2 − 1)
⎝ 2QP ⎠
= ω P2

EECS 247 Lecture 2: Filters © 2008 H.K. Page 45

s-Plane
jω

radius = ω P poles
1
arccos
2QP σ

ωP
real part = - ωP ⎛
2Q P s=− 2 ⎞
⎜1 ± j 4QP − 1 ⎟
2QP ⎝ ⎠

EECS 247 Lecture 2: Filters © 2008 H.K. Page 46

 Implementation of Biquads
• Passive RC: only real poles Æ can’t implement complex conjugate
poles

• Terminated LC
– Low power, since it is passive
– Only fundamental noise sources Æ load and source resistance
– As previously analyzed, not feasible in the monolithic form for
f <350MHz

• Active Biquads
– Many topologies can be found in filter textbooks!
– Widely used topologies:
• Single-opamp biquad: Sallen-Key
• Multi-opamp biquad: Tow-Thomas
• Integrator based biquads

EECS 247 Lecture 2: Filters © 2008 H.K. Page 47

Active Biquad
Sallen-Key Low-Pass Filter
G
H ( s) =
C1 s s2
1+ +
ω PQP ωP2
R1 R2
G 1
ωP =
Vin C2 Vout R1C1R2C2
ωP
QP =
1 1 1− G
+ +
• Single gain element R1C1 R2C1 R2C2
• Can be implemented both in discrete & monolithic form
• “Parasitic sensitive”
• Versions for LPF, HPF, BP, …
Æ Advantage: Only one opamp used
Æ Disadvantage: Sensitive to parasitic – all pole no zeros

EECS 247 Lecture 2: Filters © 2008 H.K. Page 48

 Addition of Imaginary Axis Zeros
2
⎛ s ⎞
• Sharpen transition band 1+ ⎜ ⎟
• Can “notch out” interference H( s ) = K ⎝ ωZ ⎠
2
• High-pass filter (HPF) ⎛ s ⎞
s
1+ +⎜ ⎟
• Band-reject filter ωPQP ⎝ ωP ⎠
2
⎛ω ⎞
H( jω )ω →∞ = K ⎜ P ⎟
⎝ ωZ ⎠

Note: Always represent transfer functions as a product of a gain term,

poles, and zeros (pairs if complex). Then all coefficients have a
physical meaning, and readily identifiable units.

EECS 247 Lecture 2: Filters © 2008 H.K. Page 49

Imaginary Zeros
• Zeros substantially sharpen transition band f P = 100kHz
• At the expense of reduced stop-band attenuation at QP = 2
high frequencies
fZ = 3 fP
6
x 10
2
Pole-Zero Map
10 1.5
With zeros
No zeros 1
0
Magnitude [dB]

Imag Axis

0.5
-10
0
-20
-0.5
-30
-1
-40
-1.5
-50 -2
4 5 6 7 -2 -1.5 -1 -0.5 0 0.5 1 1.5 2
10 10 10 10
Frequency [Hz] 6
Real Axis x 10

EECS 247 Lecture 2: Filters © 2008 H.K. Page 50

 Moving the Zeros
f P = 100kHz
QP = 2 5
x10 Pole-Zero Map
fZ = fP
6
4
20
2

Imag Axis
10
Magnitude [dB]

0 0
-10 -2
-20
-4
-30

-40
-6
-6 -4 -2 0 2 4 6
-50 4
105 106 107
5
10
Real Axis x10
Frequency [Hz]

EECS 247 Lecture 2: Filters © 2008 H.K. Page 51

Tow-Thomas Active Biquad

• Parasitic insensitive
• Multiple outputs

Ref: P. E. Fleischer and J. Tow, “Design Formulas for biquad active filters using three
operational amplifiers,” Proc. IEEE, vol. 61, pp. 662-3, May 1973.

EECS 247 Lecture 2: Filters © 2008 H.K. Page 52

 Frequency Response
Vo1 (b a − b )s + (b2a0 − b0 )
= −k2 2 1 2 1
Vin s + a1s + a0
Vo 2 b2 s 2 + b1s + b0
= 2
Vin s + a1s + a0
Vo 3 1 (b0 − b2a0 )s + (a1b0 − a0b1 )
=−
Vin k1 a0 s 2 + a1s + a0

• Vo2 implements a general biquad section with arbitrary poles and zeros

• Vo1 and Vo3 realize the same poles but are limited to at most one finite zero

EECS 247 Lecture 2: Filters © 2008 H.K. Page 53

Component Values
R8 given ai , bi , ki , C1 , C2 and R8
b0 =
R3 R5 R7C1C2 1
R1 =
a1C1
1 ⎛ R8 R1R8 ⎞
b1 = ⎜ − ⎟
R1C1 ⎜⎝ R6 R4 R7 ⎟⎠
k1
R2 = it follows that
a0 C2
R8
b2 = 1 1 R8
R6 R3 = ωP =
k1k2 a0 C1 R2 R3 R7C1C2
R8
a0 = 1 1 1 QP = ω P R1C1
R2 R3 R7C1C2 R4 =
1 k2 a1b2 − b1 C1
a1 =
R1C1 k1 a0
R5 =
R2 R8C2 b0C2
k1 = R8
R3 R7C1 R6 =
b2
R7
k2 = R7 = k2 R8
R8

EECS 247 Lecture 2: Filters © 2008 H.K. Page 54