Introduction to Filters

Filtering = frequency-selective signal processing

Its the most common type of signal processing Examples:
Extract desired signal from many (radio) Separating signal and noise Amplifier bandwidth limitations

Where to start
Perfectionist: ideal (low-pass) filter Engineer: continuous time, first-order low-pass filter

A/D DSP

EECS 247 Lecture 2: Introduction to Filters

2002 B. Boser 1

First-Order RC Filter (LPF1)

Steady-state frequency response:

H ( s) = Vout ( s) 1 = Vin ( s ) 1 + s o o = 1 = 2 100kHz RC
2002 B. Boser 2

with

A/D DSP

EECS 247 Lecture 2: Introduction to Filters

Poles and Zeros

s-plane (pzmap):

H (s) =

1 s 1+ o p = o z
p=-o

Pole : Zero :

A/D DSP

EECS 247 Lecture 2: Introduction to Filters

2002 B. Boser 3

Magnitude Response
Magnitude Response (s-plane) 3 2.5 Magnitude [linear] 2 1.5 1 0.5 0 5 5 x 10
5

0 0 x 10 Sigma [Hz] -5 -5 Frequency [Hz]

5

L02_bode3_lpf1.m

A/D DSP

EECS 247 Lecture 2: Introduction to Filters

2002 B. Boser 4

Frequency Response
Asymptotes: - 20 dB/dec rolloff - 90 degrees phase shift per 2 decades
Magnitude (dB) Bode Diagram 0 -20

H ( s = j ) = 0 = 1 H ( s = j ) = 0

-40

-60

-80

-100

-120 0

Matlab code (L02_bode_lpf1.m):

wo = 2*pi*100e3; s = tf('s'); h = 1 / (1+s/wo); bodehz(h, logspace(1, 10, 100));
Phase (deg) -30 -60

-90

Note: bodehz is same as bode, but frequency axis is in Hz, rather than rad/s.

10

10

10

10

10

10

10

10

10

10

10

Frequency [Hz]

A/D DSP

EECS 247 Lecture 2: Introduction to Filters

2002 B. Boser 5

Parasitics
Can we really get 100dB attenuation at 10GHz?
Probably not Parasitics limit the performance of analog components E.g.
Shunt capacitance Feed-through capacitance Finite inductor, capacitor Q

A/D DSP

EECS 247 Lecture 2: Introduction to Filters

2002 B. Boser 6

LPF2

1 + sRCP H ( s) = 1 + sR (C +C P )
A/D DSP

Pole : Zero :

p= z=

1 1 R (C + CP ) RC 1 RCP
2002 B. Boser 7

EECS 247 Lecture 2: Introduction to Filters

Frequency Response
LPF2 0 -10 -20

H ( j ) =0 = 1 H ( j ) CP = C + CP C P C = 10 3 = 60dB

Magnitude (dB) Phase (deg)

-30 -40 -50 -60 -70 -80 0

-45

-90 10
2

10

10

10

10

10

10

10

10

10

Frequency [Hz]

Why not just make C larger? Beware of other parasitics not included in this model
A/D DSP

EECS 247 Lecture 2: Introduction to Filters

2002 B. Boser 8

Continuous Time Analog

Analog passive components arent ideal
Extra real poles/zeroes result from parasitics Parasitic effects begin to appear 50dB beyond desired component characteristics Common sense helps you anticipate them

Digital filters do not suffer from these effects

A/D DSP

EECS 247 Lecture 2: Introduction to Filters

2002 B. Boser 9

Second-Order LPF
Improved attenuation (compared to 1st order) Complex poles (rather than multiple real ones)
Why? Visualize 3D s-plane plot!

Biquadratic (2nd order) transfer function:

H (s) = 1+

1 s s2 + 2 PQP P

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

A/D DSP

EECS 247 Lecture 2: Introduction to Filters

2002 B. Boser 10

Biquad Poles
H (s ) = 1 s s2 + 2 1+ PQP P P 2 1 1 4QP 2QP

has poles at

s=

for

QP

1 2

poles are real, complex otherwise

A/D DSP

EECS 247 Lecture 2: Introduction to Filters

2002 B. Boser 11

Complex Poles
QP >
1 2

s=

P 2 1 j 4QP 1 2QP

Distance from origin in s-plane:

P 2 d2 = 2Q (1 + 4QP 1) P 2 = P

A/D DSP

EECS 247 Lecture 2: Introduction to Filters

2002 B. Boser 12

s-Plane
j
radius = P
poles

real part = -

P 2QP

A/D DSP

EECS 247 Lecture 2: Introduction to Filters

2002 B. Boser 13

LPF3
Magnitude Response (s-plane) 3 2.5 Magnitude [linear]

P = 2 100kHz QP = 10

2 1.5 1 0.5 0 2 1 2 0 0 -1 Sigma [Hz] -2 -2 -1 Frequency [Hz] x 10 1

5

x 10

A/D DSP

EECS 247 Lecture 2: Introduction to Filters

2002 B. Boser 14

Frequency Response
Bode Diagram 50 0 Magnitude (dB) -50

-40 dB/dec

-100

-150

-200 0

-45 Phase (deg)

-90

-180o
-135

-180 10
2

10

10

10

10 10 Frequency (rad/sec)

10

10

10

10

A/D DSP

EECS 247 Lecture 2: Introduction to Filters

2002 B. Boser 15

Varying Q Magnitude
LPF3 Magnitude Response 40 Q = 0.5 Q = 10.0 Q = 100.0 30

20

10 Magnitude [dB]

-10

Gain at p: 20 log Q [dB]

-20

-30

-40

-50 4 10

10 Frequency [Hz]

10

A/D DSP

EECS 247 Lecture 2: Introduction to Filters

2002 B. Boser 16

Phase
LPF3 Phase Response 0 Q = 0.5 Q = 10.0 Q = 100.0 -20

-40

-60 Phase [degrees]

-80

-100

Slope at p : -45 Q deg/decade

-120

-140

-160

-180 4 10

10 Frequency [Hz]

10

A/D DSP

EECS 247 Lecture 2: Introduction to Filters

2002 B. Boser 17

Implementation of Biquads
Passive RC: only real poles Terminated LC
lowest power (well its passive!) No noise (except load and source)

Active Biquad
Filter texts give you dozens of topologies. Who needs or wants that many choices? Single-opamp biquad: Sallen-Key Two-opamp biquad: Tow-Thomas
A/D DSP

EECS 247 Lecture 2: Introduction to Filters

2002 B. Boser 18

Sallen-Key LPF
H ( s) = G s s2 + 2 1+ PQP P 1 R1C1R2C2 P 1 1 1G + + R1C1 R2C1 R2C2

P =

Single gain element Parasitic sensitive Versions for LPF, HPF, BP,

QP =

Ref: K. L. Su, Analog Filters, Chapman & Hall, 1996, pp. 215.

A/D DSP

EECS 247 Lecture 2: Introduction to Filters

2002 B. Boser 19

Component Sizing Choice 1

4 unknowns: R1, R2, C1, C2 2 knowns: P, QP problem is underdetermined Choice 1: minimum component spread
C1 = C2 = 1nF R1 = R2 = G = 3 1 = 1.6k PC1

1 = 2 .9 QP
2002 B. Boser 20

A/D DSP

EECS 247 Lecture 2: Introduction to Filters

SK Magnitude Response 1
Sallen-Key Choice 1 Magnitude Response 40 nominal R1 R1 10% large 30

20

10% increase of R1 more than doubles Q P! The circuit is very sensitive to component variations.

10 Magnitude [dB]

-10

-20

-30

-40

-50 4 10

10 Frequency [Hz]

10

A/D DSP

EECS 247 Lecture 2: Introduction to Filters

2002 B. Boser 21

Component Sizing Choice 2

Choice 2: minimum sensitivity
G =1 R1 = R2 = 2 k 2QP C1 = = 16nF P R1 C2 = 1 = 40 pF 2QP P R1

Note also:

C1 2 = 4QP = 400 C2
Huge element spread This topology is suitable only for low-Q filter implementations.

A/D DSP

EECS 247 Lecture 2: Introduction to Filters

2002 B. Boser 22

SK Magnitude Response 2
Sallen-Key Choice 2 Magnitude Response 30 nominal R1 R1 10% large 20

10

10% increase of R1 has only small effect on response! The circuit is NOT very sensitive to component variations.

0 Magnitude [dB]

-10

-20

-30

-40

-50 4 10

10 Frequency [Hz]

10

A/D DSP

EECS 247 Lecture 2: Introduction to Filters

2002 B. Boser 23

Sensitivity
Definition with y x = S xy y x x dy y Sx = y dx QP QP R1 = SR 1 QP R1
QP SR = QP 0.5 = 9.5 1

Example

Choice 1

QP R 9.5 1 = 95% QP R1 Choice 2

QP SR =0 1

Implementation and component sizing have huge impact on sensitivity High-sensitivity circuits are problems in practice No theory for finding a lowsensitivity architecture Ladder filters are usually low sensitivity Use proven circuits & check!

Common sense: Sensitivity is a first order approximation, correct only for infinitesimally small errors

A/D DSP

EECS 247 Lecture 2: Introduction to Filters

2002 B. Boser 24

Summary
Frequency Response
Poles and zeros are like tent poles and pegs Frequency response is evaluated on j axis Poles and zeros close to j axis dominate resonse

Practical Implementation Constraints

Components are not ideal Avoid solutions requiring large element spread Beware of high-sensitivity architectures

A/D DSP

EECS 247 Lecture 2: Introduction to Filters

2002 B. Boser 25