Sie sind auf Seite 1von 33

EE247

Lecture 4
Convert to integrator based form- example shown

Then will attend to high order ladder type filters

incorporating zeros
Implement the same 7th order elliptic filter in the form of
Find level of sensitivity to component mismatch

Effect of Integrator Non-Idealities on Filter

Frequency Characteristics
EECS 247

2007 H.K. Page 1

Vin

C1

Vo

L4

L2

Rs

C3

C5

RL

Design:
Filter tables
A. Zverev, Handbook of filter synthesis, Wiley, 1967.
A. B. Williams and F. J. Taylor, Electronic filter design, 3rd edition,
McGraw-Hill, 1995.

Matlab
Spice
EECS 247

2007 H.K. Page 2

Design a LPF with maximally flat passband:

f-3dB = 10MHz, fstop = 20MHz

Rs >27dB

Stopband Attenuation dB

Maximally flat passband F Butterworth

Find minimum filter order :
- Use of Matlab
- or Tables
Here tables used
-3dB

fstop / f-3dB = 2
Rs >27dB
Minimum Filter Order
F5th order Butterworth
EECS 247

rmalized
From: Williams and Taylor, p. 2-37

2007 H.K. Page 3

Find values for L & C from Table:
Note L &C values normalized to

-3dB =1
Denormalization:
Multiply all LNorm, CNorm by:
Lr = R/ -3dB
Cr = 1/(RX -3dB )
R is the value of the source and
termination resistor
(choose both 1 for now)
Then: L= Lr xLNorm
C= Cr xCNorm
EECS 247

From: Williams and Taylor, p. 11.3

2007 H.K. Page 4

Find values for L & C from Table:
Normalized values:
C1Norm =C5Norm =0.618
C3Norm = 2.0
L2Norm = L4Norm =1.618

Denormalization:
Since -3dB =2x10MHz
Lr = R/-3dB = 15.9 nH
Cr = 1/(RX-3dB )= 15.9 nF
R =1

FC1=C5=9.836nF, C3=31.83nF
FL2=L4=25.75nH
EECS 247

Example:

2007 H.K. Page 5

Last Lecture:
Order Butterworth Filter

5th

L2=25.75nH

L4=25.75nH

Vo

Rs=1Ohm
C1
9.836nF

Specifications:
f-3dB = 10MHz,

fstop = 20MHz
Rs >27dB
Used filter tables to obtain
Ls & Cs

C3
31.83nF

Magnitude (dB)

Vin

0
-5
-10
-20

RL=1Ohm

30dB

-30
-40
-50
0

EECS 247

C5
9.836nF

10

20

Frequency [MHz]

30

2007 H.K. Page 6

Conversion to Integrator Based Active Filter

Vin

+ V3

+ V1 V2
Rs I 1

I3

L2

C3

I4

I7

V6

Vo

I5

L4

C1

I2

+ V5

V4

C5

I6

RL

Use KCL & KVL to derive equations:

I
V1 = Vin V2 , V2 = 2
sC1
I
V4 = 4
, V5 = V4 V6
sC3
V
I1 = 1
Rs

, I 2 = I1 I 3

I4 = I3 I5

, I5 =

EECS 247

, V3 = V2 V4
, V6 =

V5
sL4

I6

Vo =V6

sC5
V3

I3 =

I 6 = I 5 I7

sL2

V
I7 = 6
RL

2007 H.K. Page 7

Signal Flowgraph
I
V1 = Vin V2 , V2 = 2
sC1
I
V4 = 4
, V5 = V4 V6
sC3

Vin

V
I1 = 1
Rs

, I 2 = I1 I 3

I4 = I3 I5

, I5 =

1 V1

1
Rs
I1

EECS 247

V2

V5
sL4

V3

, V3 = V2 V4
, V6 =

I 6 = I 5 I7

V4

1
sC3

I4
SFG

sL2

s L2

I3

V3

I3 =

Vo =V6

sC5

sC1
I2

I6

V
I7 = 6
RL

V5

V6

s L4

sC5

I5

I6 1

Vo

1
RL

I7

Signal Flowgraph

Vin

Vin

+ V3

V2
+ V1
I
Rs 1

1 V1

V2

V3

V4

1
1

sC1

s L2

sC3

I3

I4

C5

I6

I5

C3

I2

I7

V6

L4

I4

1
Rs
I1

I3

L2
C1

I2

+ V5

V4

V6

V5

RL

s L4

sC5

I6 1

I5

Vo

1
RL

I7

SFG
EECS 247

Normalize
Vin

1 V1

V2

1
Rs

Vin

I1

1 V1

EECS 247

V4

V5

V6

sC1

s L2

sC3

s L4

sC5

I2

V2

V2'

I3

V3

sC1R*
1

V3

R*
Rs

V1'

1 V3'

I4 1

V4

I5

V5

R*

R*

sC3R*

s L2

V4'

V6

1
1

sC5 R*

s L4

V5'

I6 1

V6' 1

Vo

1
RL

I7

Vo

R*
RL

V7'

Synthesize
V2

V2'

1 V3'

V2
+

s 2

sC5 R

V6' 1

V6

s 3

V3'
EECS 247

s 4

R*
RL

s L4 1
V5'

Vo

V4

s1 1

V4'

Vin R*
Rs

sC3R

R*

s L2

V6

V5

1
1

R*

sC1R*
1

V4

V3

R*
Rs

V1'

1 V1

Vin

Vo
R* R

s 5

V7'

V5'

2007 H.K. Page 11

Integrator Based Implementation
V2
+

s 2

s1 1

V6

V4
1

s 3

V3'

1 = C1 .R*

L
, 2 = 2 = C2 .R*
R*

s 4

Vo
R* R

s 5

V5'
, 3 = C3 .R*

Building Block:
RC Integrator

EECS 247

Vin R*
Rs

L4
, 4 =
= C4 .R*
R*

, 5 = C5 .R*

V2
= 1
V1
sR C

2007 H.K. Page 12

Negative Resistors
V1Vo+

V2+

V1V2-

Vo+

V2+

Vo-

V1+
EECS 247

Synthesize
V2

V3'
V4
V5'

Vo
EECS 247

2007 H.K. Page 14

Frequency Response
1

0.5
V2

V3'

0.1
0.5

V4

Vo

10MHz
EECS 247

V5'

10MHz
Lecture 4: Active Filters

Scale Node Voltages

Scale Vo by factor s

EECS 247

Node Scaling
V2

X 1.2
V3 X 1.2

X 1/1.2
X 1.6/1.2

X 1.2/1.6

V4 X 1.6

X 1.8/1.6

X 1.6/1.8

V5 X 1.8

X 2/1.8

X1.8/2

VO X 2
EECS 247

Maximizing Signal Handling by Node Voltage Scaling

Before Node Scaling

After Node Scaling

Scale Vo by factor s

EECS 247

2007 H.K. Page 18

Filter Noise
Total noise @ the
output: 1.4 V rms
(noiseless opamps)
Thats excellent, but the
capacitors are very
large (and the resistors
small high power
dissipation). Not
possible to integrate.
Suppose our application
allows higher noise in
the order of 140 V rms

EECS 247

Scale to Meet Noise Target

Scale capacitors and resistors
to meet noise objective

s = 10-4
Noise: 141 V rms (noiseless opamps)
EECS 247

2007 H.K. Page 20

Completed Design
V2

V3'

V4

Final design utilizing:
-Node scaling
V5'
-Final R & C
scaling based on
noise
considerations
EECS 247

Vo

2007 H.K. Page 21

Sensitivity
C1 made (arbitrarily) 50% (!) larger than
its nominal value
0.5 dB error at band edge
3.5 dB error in stopband
Looks like very low sensitivity

EECS 247

Differential 5th Order Lowpass Filter

Vin

Vo
Since each signal and its inverse readily available, eliminates the need for
negative resistors!
Differential design has the advantage of even order harmonic distortion
components and common mode spurious pickup automatically cancels
Disadvantage: Double resistor and capacitor area!
EECS 247

2007 H.K. Page 23

Including Transmission Zeros
All poles

Vin

C1

Vo

L4

L2

Rs

C5

C3

C6

C4

C2

RL

Vo
Rs

Vin

EECS 247

L4

L2
C1

C3

L6
C5

C7

RL

2007 H.K. Page 24

Example
Design a baseband filter for CDMA IS95 cellular phone receive
path with the following specs.
Filter frequency mask shown on the next page
Allow enough margin for manufacturing variations
Assume pass-band magnitude variation of 1.8dB
Assume the -3dB frequency can vary by +-8% due to
manufacturing tolerances & circuit inaccuracies

Assume any phase impairment can be compensated in the

digital domain
* Note this is the same example as for cascade of biquad while
the specifications are given closer to a real product case
EECS 247

2007 H.K. Page 25

Magnitude (dB)

+1
0
-1

-44
-46
600k 700k 900k 1.2M
Frequency [Hz]
EECS 247

2007 H.K. Page 26

Since phase impairment can be corrected for, use filter type with
max. roll-off slope/pole
Filter type Elliptic
Design filter freq. response to fall well within the freq. mask
Allow margin for component variations & mismatches
For the passband ripple, allow enough margin for ripple change
due to component & temperature variations
Design nominal passband ripple of 0.2dB
For stopband rejection add a few dB margin 44+5=49dB
Final design specifications:
fpass = 650 kHz Rpass = 0.2 dB
fstop = 750 kHz Rstop = 49 dB
Use Matlab or filter tables to decide the min. order for the filter
7th Order Elliptic
EECS 247

2007 H.K. Page 27

7th order Elliptic

C6

C4

C2

Vo
Vin

L4

L2

Rs
C1

C3

L6
C5

C7

RL

EECS 247

2007 H.K. Page 28

Specifications
fpass = 650 kHz
Rpass = 0.2 dB
fstop = 750 kHz
Rstop = 49 dB
Use filter tables to determine LC values
Table from: A. Zverev, Handbook of filter synthesis, Wiley,
1967
Elliptic filters tabulated wrt reflection coeficient

Rpass = 10 log 1 2

Since Rpass=0.2dB =20%

Use table accordingly
EECS 247

Table from Zverev book page

#281 & 282:
Since our spec. is Amin=44dB
add 5dB margin & design for
Amin=49dB

EECS 247

Table from Zverev

page #281 & 282:
Normalized
component values:
C1=1.17677
C2=0.19393
L2=1.19467
C3=1.51134
C4=1.01098
L4=0.72398
C5=1.27776
C6=0.71211
L6=0.80165
C7=0.83597

EECS 247

Frequency
superimposed
Magnitude (dB)

Frequency
response well
within spec.

-5

-15

-25

-35

-45

-55

-65
200

300

400

500

600

700

800

900

1000

1100

1200

Frequency [kHz]

EECS 247

Passband Detail
-5

Passband
well within
spec.
Magnitude (dB)

-5.5

-6

-6.5

-7

-7.5
200

300

400

500

600

700

800

Frequency [kHz]
EECS 247

2007 H.K. Page 33

The design has the same specifications as the
To compare the sensitivity of RLC ladder versus
Changed all Ls &Cs one by one by 2% in order to change the
Found frequency response most sensitive to L4 variations
Note that by varying L4 both poles & zeros are varied

EECS 247

2007 H.K. Page 34

Component mismatch in RLC filter:
Increase L4 from its nominal value by 2%
Decrease L4 by 2%
-5

L4 nom
L4 low
L4 high

Magnitude (dB)

-15
-25
-35
-45
-55
-65
200

300

400

500

600

700

800

900

1000

1100

1200

Frequency [kHz]
EECS 247

-5.7

-5.9

-6.1

-5

0.2dB

-6.3

Magnitude (dB)

-15

-6.5
200

-25

300

400

500

600

700

-50

1.7dB

-35
-55

-45
-60

-55
-65
600

700

800

900

1000

1100

1200

-65
200

300

400

500

600

700

800

900

Frequency [kHz]

EECS 247

2007 H.K. Page 36

Component mismatch in Biquad 4 (highest Q pole):
Increase p4 by 1%

2.2dB

Decrease z4 by 1%

Magnitude (dB)

0
-10
3dB

-20
-30
-40
-50
200kHz

EECS 247

600kHz
Frequency [Hz]

1MHz

2007 H.K. Page 37

7th Order elliptic filter
1% change in pole & zero pair
Passband
deviation

2.2dB
(29%)

0.2dB
(2%)

Stopband
deviation

3dB
(40%)

1.7dB
(21%)

EECS 247

2007 H.K. Page 38

7th order Elliptic

C6

C4

C2

Vo
Vin

L4

L2

Rs
C1

L6
C5

C3

C7

RL

transmission zeros
Question:
o How do we implement the transmission zeros in the integratorbased version?
o Preferred method no extra power dissipation no active
elements
EECS 247

2007 H.K. Page 39

How Do to Implement Transmission zeros?
Ca

+ V1 I 1

V2

L2

Rs

Vin

+ V3

I2

C1

Use KCL & KVL to derive :

I 2 = I1 I 3 I C ,
a

V4

I3

I C = (V2 V4 ) sCa ,
a

I4

C3

I
V2 = 2 ,
sC1

Vo

I5

RL

V2 =

I1 I 3 I C
a
sC1

a
Ca
I1 I 3
V2 =
+ V4
s ( C1 + Ca )
C1 + Ca
EECS 247

2007 H.K. Page 40

How Do to Implement Transmission zeros?
Ca

+ V1 I 1

V2

Vin

V4

C1

I2

Vo

I3

L2

Rs

+ V3

I4

I5

C3

RL

Ca
I1 I 3
V2 =
+ V4
s ( C1 + Ca )
C1 + Ca

Frequency independent constants

Can be substituted by:
Voltage-Controlled Voltage Source

Ca
I3 I5
V4 =
+ V2
s ( C3 + Ca )
C3 + Ca
EECS 247

2007 H.K. Page 41

Transmission zeros
Ca

+ V1 I 1

+ V3

V2

L2

Rs

(C1 + Ca )

Vin

I2

+
-

Ca
V4
C1 + Ca

V4

I3

I5

(C3 + Ca )

RL

+
-

Ca
I 4 V2 C + C
3
a

Replace shunt capacitors with voltage controlled voltage sources:

Ca
I1 I 3
+V
V2 =
s ( C1 + Ca ) 4 C1 + Ca

Ca
I3 I5
+V
V4 =
s ( C3 + Ca ) 2 C3 + Ca
EECS 247

3rd Order Lowpass Filter

All Poles & No Zeros
+ V1 I 1

+ V3

V2

Vin

C1

I2
Vin

1 V1
1
Rs

I1

EECS 247

V3 1

I2

RL

V4

s C1

s L2

sC3

Vo

C3

I4

1 V2

I5

V4

I3

L2

Rs

I3

Vo

1
RL
1

I4

Transmission Zero Implementation

W/O Use of Active Elements
+ V1 I 1

+ V3

V2

(C1 + Ca )

Vin

C
+ V
- 4 C aC
1+ a

1
Rs
I1
EECS 247

Ca
C3 + Ca
1 V3 1

1 V2

V4

s (C1 + Ca )

s L2

s (C3 + Ca )

I2

I3

Lecture 4: Active Filters

RL

Ca
+
- V2
C
I4
3 + Ca

Ca
C1 + Ca
1 V1

Vo

(C3 + Ca )

I2

Vin

I5

V4

I3

L2

Rs

I4

Vo

1
RL

1
2007 H.K. Page 44

Higher Order Transmission zeros
Ca

V2

V6
C5

C3

C1

Convert zero
generating Cs
in C loops to
voltagecontrolled
voltage sources

Cb

V4

V6

V2

V4

(C1 + Ca )

(C3 + Ca + Cb )

(C5 + Cb )

C
+ V
- 4 C aC
1+ a

a
+ V2
C3 + Ca
Cb
+
- V6

+
-

Cb
V4
C3 + Cb

C3 + Cb

EECS 247

Higher Order Transmission zeros

+ V1
I1

Rs

Vin

+ V3

V2

I3

L2

(C1 + Ca )

I4

+
+
-

Ca
+ V4
C1 + Ca

I2

1 V1

V1'
EECS 247

1 V3 1

1 V2

R*
Rs

(C3 + Ca + Cb )

V4

R*

s (C1 + Ca ) R*

s L2

s R* (C3 + Ca + Cb )

V3'

V4'

I7 Vo

(C5 + Cb )

RL

Cb
+ V4
C5 + Cb

I6

Cb
C3 + Cb

V2'

V6

L4

Ca
V2
C3 + Ca
Cb
V6
C3 + Cb

Ca
C3 + Ca

Ca
C1 + Ca
Vin

I 5 + V5

V4

Cb
C5 + Cb
V5

1
R*
s L4

V5'

V6

Vo

1
*
s R ( C5 + Cb )
1 V'
6

R*
RL

V7'

2007 H.K. Page 46

Example:
5th Order Chebyshev II Filter

EECS 247

Table from: Williams &

Taylor book, p. 11.112

f-3dB =10MHz

V1 =

1
Vi V1 V2 + Ca V
Ca + C1 3
s (Ca + C1 ) Rs
R*

-Rs

Rs

R*

EECS 247

V2

From:
Lecture 4
page 14

V3'
V4
V5'

Vo
EECS 247

5th Order Chebyshev II Filter

Opamp-RC Simulation
V2

V3'
V4

V5'
Vo
EECS 247

+
+

Vin

7th Order Differential Lowpass Filter

Including Transmission Zeros

Vo

Transmission zeros implemented with

pair of coupling capacitors
EECS 247

Effect of Integrator Non-Idealities on Filter

Frequency Characteristics
In the passive filter design (RLC filters) section:
Reactive element (L & C) non-idealities expressed in the
form of Quality Factor (Q)
Filter impairments due to component non-idealities explained
in terms of component Q

In the context of active filter design (integrator-based

filters)
Integrator non-idealities Translated to have form of Quality
Factor (Q)
Filter impairments due to integrator non-idealities explained in
terms of integrator Q
EECS 247

Effect of Integrator Non-Idealities on

Filter Performance
Ideal integrator characteristics
Real integrator characteristics:
Effect of opamp finite DC gain
Effect of integrator non-dominant poles

EECS 247

Effect of Integrator Non-Idealities on Filter Performance

Ideal Integrator
Ideal
Intg.Intg.
Ideal
l o g H (s)

Vin

Vo
0dB

I d e al o pa m p D C ga i n =
Sing le po le @ D C
n o no n- d om in an t po le s
o
H( s )=
s
o = 1/ RC
EECS 247

-90o

Ideal Integrator Quality Factor

Ideal intg. transfer function:

H( s )=

H ( j ) =
Q=

Then:

=
j

1
R ( ) + j X ( )

X ( )
R ( )

t g. =
Qiidneal

EECS 247

Real Integrator Non-Idealities

Ideal Intg.
l o g H (s)

P1 =

-90o

-90o

H( s )=
EECS 247

o
s

Real Intg.
l o g H (s)

H( s )
Lecture 4: Active Filters

0
a

P2P3

1+ s a

)(1 + p2s )(1 + p3s ) .. .

2007 H.K. Page 56

Effect of Integrator Finite DC Gain on Q

l o g H (s)
a

0
P1 =

+ Arct an P1

0
a

-89.5

-90

P1

P h a s e l ea d @ o

-90o

Example: P1/ 0 = 1/100

phase error +0.5degree
EECS 247

Effect of Integrator Finite DC Gain on Q

Magnitude (dB)

Ideal intg
Intg with finite DC gain

Droop in the passband

Normalized Frequency
EECS 247

l o g H (s)

P2P3
-90

Ar ct an po
i =2 i

o Phase l ag @
o
p
i =2 i

-90.5

-90o

Example: 0 /P2 =1/100

phase error 0.5degree
EECS 247

Effect of Integrator Non-Dominant Poles

Magnitude (dB)

Ideal intg
Opamp with finite bandwidth

Phase lag @ 0

Peaking in the passband

In extreme cases could
result in oscillation!

1
Normalized Frequency
EECS 247

Effect of Integrator Non-Dominant Poles &

Finite DC Gain on Q
l o g H (s)
a

o
P1 =

0
a

P2P3

-90

+ Arctan P1

o

Arctan o
i = 2 pi

-90o
Note that the two terms have different signs
Can cancel each others effect!
EECS 247

Integrator Quality Factor

Real intg. transfer function:

H( s )

1+ s a

Based on the definition of Q

and assuming that:

o
<< 1
p2,3,... ..
g.
Qrint
eal

It can be shown that in the

vicinity of unity-gain-frequency:

Phase le

EECS 247

&

a >> 1

1 1
o
a
i = 2 pi

Phase

l ag @

2007 H.K. Page 62

Example:
Effect of Integrator Finite Q on Bandpass Filter Behavior
0.5 excess @ ointg

Ideal

Ideal

Integrator P2 @ 100.o

Integrator DC gain=100

EECS 247

2007 H.K. Page 63

Example:
Effect of Integrator Q on Filter Behavior
error @

excess )

ointg

ointg ~ 0
Ideal

EECS 247

2007 H.K. Page 64

Summary
Effect of Integrator Non-Idealities on Q
int g. =
Qideal
int g.
Qreal

1
1
a o

1
p
i =2 i

Amplifier DC gain reduces the overall Q in the same manner as

series/parallel resistance associated with passive elements
Amplifier poles located above integrator unity-gain frequency enhance
the Q!
If non-dominant poles close to unity-gain freq. Oscillation
Depending on the location of unity-gain-frequency, the two terms can
cancel each other out!

EECS 247