From last lecture: Dynamic range of analog circuits Filters: Nomenclature Specifications

Quality factor Frequency characteristics Group delay

EE247 - Lecture 2 Filters

Filter types
Butterworth Chebyshev I & II Elliptic Bessel

Group delay comparison example Biquads

EECS 247 Lecture 2: Filters 2007 H.K. Page 1

Nomenclature Filter Types

Lowpass Highpass H ( j ) Bandpass H ( j ) Band-reject (Notch) H ( j )

All-pass
H ( j )

H ( j )

Provide frequency selectivity

Phase shaping or equalization

2007 H.K. Page 2

EECS 247

Lecture 2: Filters

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) Power/pole & Area/pole
EECS 247 Lecture 2: Filters 2007 H.K. Page 3

Lowpass Filter Frequency Characteristics

H ( j ) [ d B ] Passband Ripple (Rpass)
H (0 )
Passband Gain

f3dB

3dB H ( j )

Transition Band

Stopband Rejection

H ( j )
Passband

fc

Frequency (Hz)

f stop

Stopband Frequency

x 10

EECS 247

Lecture 2: Filters

2007 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

2007 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=

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

p e r u n i t t i me

EECS 247

Lecture 2: Filters

2007 H.K. Page 6

Inductor & Capacitor Quality Factor

Inductor Q :
YL = 1 Rs + j L QL = L Rs
Rs L

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

Rp

QC = CRp
C

Lecture 2: Filters

2007 H.K. Page 7

Pole Quality Factor

j
s-Plane

QP o l e =

p 2 x

x
x

EECS 247

Lecture 2: Filters

2007 H.K. Page 8

Bandpass Filter Quality Factor (Q)

H ( jf )

Q= fcenter /f

0 Magnitude [dB] -3dB

f = f2 - f1

0.1

f1

1 fcenter

f2

10 Frequency
2007 H.K. Page 9

EECS 247

Lecture 2: Filters

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 2007 H.K. Page 10

What is Group Delay?

vOUT(t) = A1 G(j) sin

{[ t+

()

]} +
(+) +

+ A2 G[ j(+)] sin Since then 2 0 <<1

{ (+) [ t +

]}

[ ]
()+

(+) +

d() d

1 ][ (1 - )]

() +

() d() d -

)
2007 H.K. Page 11

EECS 247

Lecture 2: Filters

What is Group Delay? Signal Magnitude and Phase Impairment

vOUT(t) = A1 G(j) sin

{ [t +

()

]} +
d()

+ A2 G[ j(+)]sin

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

()

) ]}

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 If the second term is zero, then the filters output at frequency + and the output at frequency are each delayed in time by -()/

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

EECS 247

Lecture 2: Filters

2007 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
Lecture 2: Filters

EECS 247

2007 H.K. Page 13

What is Group Delay? Signal Magnitude and Phase Impairment

If GR= PD No phase distortion vOUT(t) = A1 G(j) sin t - GR
2

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

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

2007 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 2007 H.K. Page 15

Filter Types Lowpass Butterworth Filter

0

Maximally flat amplitude within the filter passband

Magnitude (dB)

-20 -40 -60

Phase (degrees)

=0

-200

Moderate phase distortion

-400 0 1 2

Normalized Frequency Example: 5th Order Butterworth filter

EECS 247 Lecture 2: Filters 2007 H.K. Page 16

Normalized Group Delay

H( j ) d

=0

Lowpass Butterworth Filter

All poles Poles located on the unit circle with equal angles
j s-plane

Example: 5th Order Butterworth Filter

EECS 247 Lecture 2: Filters 2007 H.K. Page 17

Filter Types Chebyshev I Lowpass Filter

0

-200

-400 0 1 2 0

Normalized Frequency
Example: 5th Order Chebyshev filter EECS 247 Lecture 2: Filters 2007 H.K. Page 18

Normalized Group Delay

Phase (degrees)

Chebyshev I filter Ripple in the passband Sharper transition band compared to Butterworth Poorer group delay As more ripple is allowed in the passband: Sharper transition band Poorer phase response

Magnitude (dB)

-20

-40 0 35

Chebyshev I Lowpass Filter Characteristics

j All poles Poles located on an ellipse inside the unit circle Allowing more ripple in the passband: Narrower transition band Sharper cut-off Higher pole Q Poorer phase response
Chebyshev I LPF 3dB passband ripple Chebyshev I LPF 0.1dB passband ripple Example: 5th Order Chebyshev I Filter
EECS 247 Lecture 2: Filters 2007 H.K. Page 19

s-plane

Filter Types Chebyshev II Lowpass Chebyshev II filter

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) -90 -180 -270 -360 0 0.5 1 1.5 2

Magnitude (dB)

Normalized Frequency

Example: 5th Order Chebyshev II filter

EECS 247 Lecture 2: Filters 2007 H.K. Page 20

Filter Types Chebyshev II Lowpass

Both poles & zeros No. of poles n No. of finite zeros n-1 Poles located both inside & outside of the unit circle Complex conjugate zeros located on j axis Nulls in stopband
Example: 5th Order Chebyshev II Filter
EECS 247 Lecture 2: Filters

s-plane

poles zeros
2007 H.K. Page 21

Filter Types Elliptic Lowpass Filter

0

Magnitude (dB)
Phase (degrees)

Elliptic filter
Ripple in passband Nulls in the stopband Sharper transition band compared to Butterworth & both Chebyshevs Poorest phase response

-20 -40 -60 0

-200

-400 0

Normalized Frequency

Example: 5th Order Elliptic filter

EECS 247

Lecture 2: Filters

2007 H.K. Page 22

Filter Types Elliptic Lowpass Filter

Both poles & zeros No. of poles: n No. of finite zeros: n-1 Zeros located on j axis Sharp cut-off Narrower transition band Pole Q higher compared to the previous filters j s-plane

Pole Zero Example: 5th Order Elliptic Filter

EECS 247

Lecture 2: Filters

2007 H.K. Page 23

Filter Types Bessel Lowpass Filter Bessel All poles Maximally flat group delay Poor out-of-band attenuation Poles outside unit circle Relatively low Q poles
j s-plane

Pole Example: 5th Order Bessel filter

EECS 247

Lecture 2: Filters

2007 H.K. Page 24

Magnitude Response of a Bessel Filter as a Function of Filter Order (n)

0 -10 -20

Magnitude [dB]

-30 -40 -50 -60 -70 -80 -90 -100 0.1 1

lte Fi

er rd O r

d se ea r c n=1 In

2 3 4 6 7
10 100

Normalized Frequency

EECS 247

Lecture 2: Filters

2007 H.K. Page 25

Filter Types Comparison of Various Type LPF Magnitude Response

0

Magnitude (dB)

-20 -40 -60

Magnitude (dB)

Normalized Frequency
Bessel Butterworth Chebyshev I Chebyshev II Elliptic
2007 H.K. Page 26

All 5th order filters with same corner freq.

EECS 247

Lecture 2: Filters

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

2007 H.K. Page 27

Comparison of Various LPF Groupdelay

5
Bessel

28
Chebyshev I 0.5dB Passband Ripple

1 1 12 10
Butterworth

4 1 1

Ref: A. Zverev, Handbook of filter synthesis, Wiley, 1967. Lecture 2: Filters 2007 H.K. Page 28

EECS 247

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

2007 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

10

10

10

Frequency [Hz]
EECS 247 Lecture 2: Filters 2007 H.K. Page 30

Phase Response
4th Order Chebyshev I versus Bessel
0 -50

Phase [degrees]

-100 -150 -200 -250 -300 -350 0 50 100 150 200 4th order Chebyshev 1 4th order Bessel

Frequency [kHz]
EECS 247 Lecture 2: Filters 2007 H.K. Page 31

Group Delay
4th Order Chebyshev I versus Bessel
14 Group Delay [usec] 12 10 8 6 4 2 0 10 100 Frequency [kHz]
Lecture 2: Filters

4th order Chebyshev 1 4th order Bessel

1000

EECS 247

2007 H.K. Page 32

Normalized Group Delay

4th Order Chebyshev I versus Bessel
3

Group Delay [normalized]

2.5 2 1.5 1 0.5 0 10 100 1000 4th order Chebyshev 1 4th order Bessel

Frequency [kHz]
Lecture 2: Filters

EECS 247

2007 H.K. Page 33

Step Response
4th Order Chebyshev I versus Bessel
1.4 1.2 4th order Bessel

Amplitude

1 0.8 0.6 0.4 0.2 00 5 10 15 20 4th order Chebyshev 1

Time (usec)

EECS 247

Lecture 2: Filters

2007 H.K. Page 34

Intersymbol Interference (ISI)

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

EECS 247

Lecture 2: Filters

2007 H.K. Page 35

Pulse Impairment Bessel versus Chebyshev

1.5

Input

1.5

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

2
-4

-1.5

1.1

1.2

1.3

1.4

1.5

1.6

1.7

1.8

8th order Bessel

4th order Chebyshev I

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 compared to Bessel
EECS 247 Lecture 2: Filters 2007 H.K. Page 36

Response to Psuedo-Random Data Chebyshev versus Bessel

1.5 1

1111011111001010000100010111101110001001

Input Signal: Symbol rate 1/130kHz

0.5

-0.5

-1

-1.5 0

0.2

0.4

0.6

0.8

1.2 x 10

1.4
-4

1.5

1.5

0.5

0.5

1111011111001010000100010111101110001001

1111011111001010000100010111101110001001

-0.5

-0.5

-1

-1

-1.5 0

0.2

0.4

0.6

0.8

1.2 x 10

1.4
-4

-1.5 0

0.2

0.4

0.6

0.8

1.2 x 10

1.4
-4

4th order Bessel

EECS 247

4th order Chebyshev I

Lecture 2: Filters

2007 H.K. Page 37

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
EECS 247 Lecture 2: Filters 2007 H.K. Page 38

RLC Filters
Bandpass filter:
Vo = RC 2 Vin s2 + o s + o
Q s
Vin
R L

Vo
C

o = 1 LC Q = o RC = R L

j
s-Plane

Singularities: Pair of complex conjugate poles Zeros @ f=0 & f=inf.

EECS 247 Lecture 2: Filters 2007 H.K. Page 39

RLC Filters
Design a bandpass filter with:
R

Vo
L C

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 Q?
EECS 247 Lecture 2: Filters 2007 H.K. Page 40

RLC Filters Effect of Finite Component Q

1 = 1 + 1 Q filt Qid ea l Qin d . filt
Qfilt.=20 (ideal L) Qfilt. =13.3 (QL.=40)

Need to have component Q much higher compared to desired filter Q

EECS 247 Lecture 2: Filters 2007 H.K. Page 41

RLC Filters
R

Vo
L C

Vin

Question: Can RLC filters be integrated on-chip?

EECS 247

Lecture 2: Filters

2007 H.K. Page 42

Monolithic Inductors Feasible Quality Factor & Value

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 2007 H.K. Page 43

Monolithic LC Filters
Monolithic inductor in CMOS tech.
L<10nH with Q<7

Max. capacitor size (based on realistic chip area)

C< 20pF

LC filters in the monolithic form feasible: - Frequency >350MHz - Only low quality factor filters
Learn more in EE242
EECS 247 Lecture 2: Filters 2007 H.K. Page 44

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 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

2007 H.K. Page 45

Filters
2nd
Biquadratic
H( s ) = 1+ 1 s
2 PQP P

Order Transfer Functions (Biquads)

order) transfer function:
s2

(2nd
+

H( j ) =

1 2 2 1 2 + PQP P
2

H ( j ) =0 = 1 H ( j ) = 0 H ( j ) = = QP
P

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

2
EECS 247

P 2 1 1 4QP 2QP

pole s are re al, c o mplex othe rwise

Lecture 2: Filters 2007 H.K. Page 46

Biquad Complex Poles

QP >
1 2

Complex conjugate poles:

s=

2QP

P 2 1 j 4QP 1

S-plane
poles

d
Distance from origin in s-plane:
P 2 d2 = 2Q (1 + 4QP 1) P 2 = P
2

EECS 247

Lecture 2: Filters

2007 H.K. Page 47

s-Plane
j

radius = P arccos 1 2QP

poles

real part = -

P
2Q P
Lecture 2: Filters

s=

2QP

P 2 1 j 4QP 1

2007 H.K. Page 48

EECS 247

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 <250MHz 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 2007 H.K. Page 49

Active Biquad Sallen-Key Low-Pass Filter

C1

H ( s) =
1+

G s s2 + 2 PQP P

R1 Vin

R2
C2

Vout

P =
QP =

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
Lecture 2: Filters 2007 H.K. Page 50

EECS 247

Addition of Imaginary Axis Zeros

Sharpen transition band Can notch out interference High-pass filter (HPF) Band-reject filter
H( s ) = K 1+
s 1+ Z
2

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

2007 H.K. Page 51

Imaginary Zeros
Zeros substantially sharpen transition band At the expense of reduced stop-band attenuation at high frequency
6 x 10 2 10 With zeros No zeros 1.5 1

f P = 100kHz QP = 2 fZ = 3 fP
Pole-Zero Map

Magnitude [dB]

0 -10 -20 -30 -40 -50 4 10 5 10

Imag Axis
6 10 7 10

0.5 0

-0.5 -1 -1.5 -2 -2 -1.5 -1 -0.5 0 0.5 1 1.5 2 6 x 10

Frequency [Hz]

Real Axis
Lecture 2: Filters

EECS 247

2007 H.K. Page 52

Moving the Zeros

f P = 100kHz QP = 2 fZ = fP
5 x10 6 4 Imag Axis 2 0 Pole-Zero Map

20

Magnitude [dB]

10 0 -10 -20 -30 -40 -50 4 10 105 106 107

-2 -4 -6 -6 -4 -2 0 2 4 Real Axis 6 5 x10

Frequency [Hz]

EECS 247

Lecture 2: Filters

2007 H.K. Page 53

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.
Lecture 2: Filters 2007 H.K. Page 54

EECS 247

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 2007 H.K. Page 55

Component Values
b0 = b1 = b2 = a0 = a1 = k1 = k2 = R8 R3 R5 R7C1C2 1 R8 R1R8 R1C1 R6 R4 R7 R8 R6 R8 R2 R3 R7C1C2 1 R1C1 R2 R8C2 R3 R7C1 R7 R8
given ai , bi , ki , C1 , C2 and R8
R1 = R2 = R3 = R4 = R5 = R6 = 1 a1C1 k1 a0 C2 1 k1k2 1 a0 C1

it follows that
P =
R8 R2 R3 R7C1C2

1 1 1 k2 a1b2 b1 C1 k1 a0 b0C2 R8 b2

QP = P R1C1

R7 = k2 R8

EECS 247

Lecture 2: Filters

2007 H.K. Page 56