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)

H ( j )

H ( j )

H ( j )

H ( j )

Provide frequency selectivity

EECS 247

Lecture 2: Filters

All-pass
H ( j )

Phase shaping
or equalization
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 outputreferred third order intercept point)
Area/pole & Power/pole
EECS 247

Lecture 2: Filters

2008 H.K. Page 3

Lowpass Filter Frequency Characteristics

H ( j ) [ d B ]
Passband Ripple (Rpass)

f3dB

H (0 )
3dB

Passband
Gain

H ( j )
Passband

EECS 247

Transition
Band

H ( j )

fc

f stop

Frequency (Hz)

Lecture 2: Filters

Stopband
Frequency

Stopband
Rejection

x 10

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 ( j ) =

1
R ( ) + jX ( )

Quality factor is defined as:

Q=

EECS 247

X ( )
Energy Stored

Aver age Po w er D i s s i pa t i o n
R ( )

Lecture 2: Filters

p e r u n i t t i me

2008 H.K. Page 6

Component Quality Factor (Q)

Inductor & Capacitor Quality Factor
Inductor Q :
YL =

1
Rs + j L

QL = L
Rs

Rs

Capacitor Q :
1
ZC =
1 + jC
Rp
EECS 247

Rp

QC = CR p
C

Lecture 2: Filters

2008 H.K. Page 7

Pole Quality Factor

Typically filter
singularities include
pairs of complex
conjugate poles.

QP o l e =

EECS 247

Quality factor of
complex conjugate
poles are defined as:

s-Plane

p
2 x

Lecture 2: Filters

2008 H.K. Page 8

Bandpass Filter Quality Factor (Q)

H ( jf )

Q= fcenter /f

Magnitude [dB]

0


-3dB

f = f2 - f1





 

f1

0.1

EECS 247

1
fcenter

f2

Lecture 2: Filters

10
Frequency
2008 H.K. Page 9

What is Group Delay?

Consider a continuous time filter with s-domain transfer function G(s):

G(j) G(j)e

j()

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

sinewaves at slightly different frequencies (<<):

vIN(t) = A1sin(t) + A2sin[(+) t]

The filter output is:

vOUT(t) = A1 G(j) sin[t+()] +

A2 G[ j(+)] sin[(+)t+ (+)]
EECS 247

Lecture 2: Filters

2008 H.K. Page 10

What is Group Delay?

vOUT(t) = A1 G(j) sin

]} +

{ (+) [ t +

+ A2 G[ j(+)] sin
Since

()

{[ t+

(+)
+

]}

then 2
0
<<1

[ ]

(+)
+

()+
()
+

EECS 247

d()

][ 1 (1 - ) ]

()
d()
d -

Lecture 2: Filters

2008 H.K. Page 11

What is Group Delay?

Signal Magnitude and Phase Impairment
vOUT(t) = A1 G(j) sin

{ [t +

+ A2 G[ j(+)]sin

()

]} +

{ (+) [ t + ()
+( d

d()

()

) ]}

PD -()/ is called the phase delay and has units of time

If the second term is zero, then the filters output at frequency + and
the output at frequency are each delayed in time by -()/
If the second term in the phase of the 2nd sin wave is non-zero, then the
filters output at frequency + is time-shifted differently than the
filters output at frequency
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:
()
d()
d - = 0
Clearly, if ()=k, k a constant, no phase distortion
This type of filter phase response is called linear phase
Phase shift varies linearly with frequency
GR -d()/d is called the group delay and also has units
of time. For a linear phase filter GR PD =k
GR= PD implies linear phase
Note: Filters with ()=k+c are also called linear phase filters, but
theyre 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 GR= PD No phase distortion

[ (

)] +
+ A G[ j(+)] sin [ (+) ( t - )]

vOUT(t) = A1 G(j) sin t - GR

2

GR

If alsoG( j)=G[ j(+)] 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:
PD -()/ [ time]
Group delay is defined as :

GR -d()/d [time]
If ()=k, k a constant, no phase distortion
For a linear phase filter GR PD =k
EECS 247

Lecture 2: Filters

2008 H.K. Page 15

Filter Types wrt Frequency Response

Lowpass Butterworth Filter

H( j )
d

=0

-20
-40
-60

=0

Moderate phase distortion

-200

-400
0

Normalized Group Delay

Maximally flat amplitude

within the filter passband

Phase (degrees)

Magnitude (dB)

Normalized Frequency
Example: 5th Order Butterworth filter
EECS 247

Lecture 2: Filters

2008 H.K. Page 16

Lowpass Butterworth Filter

j

All poles

s-plane

Number of poles equal to

filter order

Poles located on the unit

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

-20

-40
35

Poorer group delay

compared to Butterworth
More ripple in passband
poorer phase response

Phase [degrees]

-200

-400
0

Normalized Group Delay

Chebyshev I filter
Ripple in the passband
Sharper transition band
compared to Butterworth (for
the same number of poles)

Magnitude [dB]

Normalized Frequency
Example: 5th Order Chebyshev filter
EECS 247

Lecture 2: Filters

2008 H.K. Page 18

Chebyshev I Lowpass Filter Characteristics

j

All poles

Poles located on an ellipse

inside the unit circle

s-plane

Allowing more ripple in the

passband:
Narrower transition band
Sharper cut-off
Higher pole Q

Chebyshev I LPF 3dB passband ripple

Chebyshev I LPF 0.1dB passband ripple

Poorer phase response

Example: 5th Order Chebyshev I Filter

EECS 247

Lecture 2: Filters

2008 H.K. Page 19

No ripple in passband
Nulls or notches in
stopband
Sharper transition band
compared to
Butterworth
Passband phase more
linear compared to
Chebyshev I

0
-20
-40
-60
0

Phase (deg)

Chebyshev II filter

Magnitude (dB)

Filter Types
Chebyshev II Lowpass

-90
-180
-270
-360
0

0.5

1.5

Normalized Frequency

Example: 5th Order Chebyshev II filter

EECS 247

Lecture 2: Filters

2008 H.K. Page 20

Filter Types
Chebyshev II Lowpass

Poles & finite zeros

No. of poles n
(n > filter order)
No. of finite zeros: n-1

s-plane

Poles located both inside

& outside of the unit circle

Complex conjugate zeros

located on j axis
Zeros create nulls in
stopband

EECS 247

Example:
5th Order
Chebyshev II Filter

pole
zero

Lecture 2: Filters

2008 H.K. Page 21

Filter Types
Elliptic Lowpass Filter
0

Magnitude (dB)

Elliptic filter
Ripple in passband
Nulls in the stopband

-20
-40
-60

Sharper transition band

compared to Butterworth &
both Chebyshevs
Poorest phase response

Phase (degrees)

-200

-400
0

Normalized Frequency

Example: 5th Order Elliptic filter

EECS 247

Lecture 2: Filters

2008 H.K. Page 22

Filter Types
Elliptic Lowpass Filter
j

Poles & finite zeros

No. of poles: n
No. of finite zeros: n-1

s-plane

Zeros located on j axis

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

s-plane

All poles
Poles outside unit circle

Relatively low Q poles

Maximally flat group delay
Poor out-of-band attenuation

Pole
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
-10

Magnitude [dB]

-20
-30
-40

lte
Fi

-50

er
rd
O
r

d
se
ea
r
c
n=1
In

n Filter order

-60

-70

-80

-90

-100
0.1

10

100

Normalized Frequency

EECS 247

Lecture 2: Filters

2008 H.K. Page 25

Filter Types
Comparison of Various Type LPF Magnitude Response

Magnitude (dB)

-20
-40

Magnitude (dB)

-60
2

Normalized Frequency

All 5th order filters with same corner freq.

EECS 247

Lecture 2: Filters

Bessel
Butterworth
Chebyshev I
Chebyshev II
Elliptic
2008 H.K. Page 26

Filter Types
Comparison of Various LPF Singularities
j
Poles Bessel
Poles Butterworth
Poles Elliptic
Zeros Elliptic
Poles Chebyshev I 0.1dB

s-plane

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
1
12
10

Butterworth

4
1
1

EECS 247

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

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

Magnitude (dB)

-20

-40

-60

4th Order Chebychev 1

4th Order Bessel
4

10

10

10

Frequency [Hz]
EECS 247

Lecture 2: Filters

2008 H.K. Page 30

Phase Response
4th Order Chebyshev I versus Bessel
0

Phase [degrees]

-50
-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

Group Delay [usec]

14
12
10
8

4th order
Chebyshev 1

6
4

4th order Bessel

2
0
10

EECS 247

100
Frequency [kHz]
Lecture 2: Filters

1000

2008 H.K. Page 32

Step Response
4th Order Chebyshev I versus Bessel
1.4

Amplitude

1.2

4th order Bessel

1
4th order
Chebyshev 1

0.8
0.6
0.4
0.2
00

EECS 247

10

Time (usec)

Lecture 2: Filters

15

20

2008 H.K. Page 33

Intersymbol Interference (ISI)

ISI 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

Output

0.5

0.5

-0.5

-0.5

-1

-1

-1.5

1.1

1.2

1.3

1.4

1.5

1.6

1.7

1.8

1.9
x 10

-1.5

1.1

1.2

1.3

1.4

1.5

1.6

1.7

1.8

-4

4th order Chebyshev I

4th order Bessel

1.9
2
-4
x 10

Note that in the case of the Chebyshev filter not only the pulse has broadened but it
also has a long tail
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

Input Signal:
Symbol rate 1/130kHz

1111011111001010000100010111101110001001

0.5

-0.5

-1

-1.5
0

0.2

0.4

0.6

0.8

1.2

1.4
x 10

1.5

1.5

0.5

0.5

1111011111001010000100010111101110001001

-0.5

-0.5

-1

-1

-1.5
0

-4

0.2

0.4

0.6

0.8

4th order Bessel

EECS 247

1.2

-1.5
0

1.4
x 10

-4

1111011111001010000100010111101110001001

0.2

0.4

0.6

0.8

1.2

1.4
x 10

-4

4th order Chebyshev I

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 Higher order filter
In the case of passive filters
For integrated active filters

higher component count

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

Vin

Vo =
RC
Vin s2 + o s + o2

Vo

Bandpass filter (2nd order):

o = 1 LC
Q = o RC = R
L

s-Plane

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:
Vo

Center frequency of 1kHz

Quality factor of 20

Vin

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
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
Vo

Vin

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= ( 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
LC filters in the monolithic form feasible: freq>350MHz
(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
Biquadratic

Order Transfer Functions (Biquads)

(2nd

order) transfer function:

H( s ) =
1+

H( j ) =

s2

PQP P2

H ( j ) =0 = 1

1
2

2 2
1 2 +
PQP

Bi quad pole s @: s =
Note : for QP 1
2

EECS 247

H ( j ) = 0
H ( j ) = = QP
P

P
2
1 1 4QP

2QP

pole s are re al, c o mplex othe rwise

Lecture 2: Filters

2008 H.K. Page 44

Biquad Complex Poles

QP >

1
2

s=

Complex conjugate poles:

P
2

1 j 4QP 1

2QP

S-plane
poles

d
Distance from origin in s-plane:

d 2 = P (1 + 4QP2 1)
2QP
= P2

EECS 247

Lecture 2: Filters

2008 H.K. Page 45

s-Plane
j

radius = P
arccos

1
2QP

real part = -

EECS 247

poles

P
2Q P

s=
Lecture 2: Filters

P
2

1 j 4QP 1

2QP

2008 H.K. Page 46

Implementation of Biquads
Passive RC: only real poles cant implement complex conjugate
poles
Terminated LC
Low power, since it is passive
Only fundamental noise sources 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

R1
Vin

R2
C2

1+

Vout

P =
QP =

H ( s) =

C1

s
s2
+
PQP P2

1
R1C1R2C2

1
1
1 G
+
+
R1C1 R2C1 R2C2
Single gain element
Can be implemented both in discrete & monolithic form
Parasitic sensitive
Versions for LPF, HPF, BP,
Advantage: Only one opamp used
Disadvantage: Sensitive to parasitic all pole no zeros

EECS 247

Lecture 2: Filters

2008 H.K. Page 48

Addition of Imaginary Axis Zeros

Sharpen transition band

Can notch out interference
High-pass filter (HPF)
Band-reject filter

s
1+

H( s ) = K
1+

s
+

PQP P

H( j ) = K P
Z

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
f P = 100kHz

Zeros substantially sharpen transition band

At the expense of reduced stop-band attenuation at
high frequencies

QP = 2
fZ = 3 fP

6
x 10
2

Pole-Zero Map

1.5

With zeros
No zeros

Imag Axis

Magnitude [dB]

10

0.5

-10
-20

-0.5

-30

-1

-40
-50
4
10

EECS 247

-1.5
5
10

6
10

Frequency [Hz]

7
10

-2

-2

-1.5

-1

-0.5

0.5

Real Axis
Lecture 2: Filters

1.5

2
6
x 10

2008 H.K. Page 50

Moving the Zeros

f P = 100kHz
QP = 2

5
x10

fZ = fP

6
4
Imag Axis

20

Magnitude [dB]

Pole-Zero Map

10
0
-10

2
0

-2

-20

-4

-30

-6

-40
-50 4
10

105

106

Frequency [Hz]

EECS 247

107

Lecture 2: Filters

-6

-4

-2

Real Axis

6
5
x10

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
(b a b )s + (b2a0 b0 )
Vo1
= 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
s 2 + a1s + a0
k1 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
given ai , bi , ki , C1 , C2 and R8

b0 =

R8
R3 R5 R7C1C2

b1 =

1 R8 R1R8

R1C1 R6 R4 R7

R2 =

b2 =

R8
R6

R3 =

a0 =

R8
R2 R3 R7C1C2

1
k1k2

a1 =

1
R1C1

R4 =

1
1
1
k2 a1b2 b1 C1

R5 =

k1 a0
b0C2

R6 =

R8
b2

k1 =
k2 =

R2 R8C2
R3 R7C1
R7
R8

EECS 247

R1 =

1
a1C1
k1
a0 C2
1
a0 C1

it follows that
P =

R8
R2 R3 R7C1C2

QP = P R1C1

R7 = k2 R8

Lecture 2: Filters

2008 H.K. Page 54