Upload
others
View
0
Download
0
Embed Size (px)
Citation preview
Analog Filters (Ch. 14)g
김 영 석김 영 석
충북대학교 전자정보대학
2012.9.1.9.
Email: [email protected]
전자정보대학 김영석 14-1
Contents14.1 General Considerations
14.2 First-Order Filters
14.3 Second-Order Filters
14 4 A ti Filt14.4 Active Filters
14.5 Approximation of Filter Response
전자정보대학 김영석 14-2
14.1 General ConsiderationsWhy We Need Filters?
In order to eliminate the unwanted interference that accompanies a signal, a filter is needed.
전자정보대학 김영석 14-3
14.1.1 Filter Characteristics
Ideally, a filter needs to have a flat pass band and a sharp roll-off in its transition band.
Realistically, it has a rippling pass/stop band and a transition band.
전자정보대학 김영석 14-4
Ex14.1/2: Filter I
Ex14.1
Given: Adjacent channel Interference is 25 dB above the signal
Design goal: Signal to Interference ratio of 15 dB
S l ti A filt ith t b d Att ti f 4 dBSolution: A filter with stop band Attenuation of 40 dB
Ex14.2
Given: Adjacent channel Interference is 40 dB above the signal
Design goal: Signal to Interference ratio of 20 dB
Solution: A filter with stop band of 60 dB at 60 Hz
전자정보대학 김영석 14-5
Ex14.3: Filter III
A bandpass filter around 1.5 GHz is needed to reject the dj t C ll l d PCS i ladjacent Cellular and PCS signals.
전자정보대학 김영석 14-6
14.1.2 Classification of Filters
전자정보대학 김영석 14-7
14.1.3 Filter Transfer Function
Filter a) has a transfer function with -20dB/dec roll-off
A B
) s s
Filter b) has a transfer function with -40dB/dec roll-off, better selectivity.
Zm=m’th zero( )( ) ( )1 2( ) ms Z s Z s ZH s α
− − −=
L
Pn =n’th pole( )( ) ( )1 2( )
mH s
s P s P s Pα=
− − −L
전자정보대학 김영석 14-8
Ex14.4 Pole-Zero Diagram
전자정보대학 김영석 14-9
Ex14.5 Position of the Poles
Poles on the RHPUnstable
Poles on the jω axisOscillatory
Poles on the LHPDecaying
(no good)y
(no good)y g
(good)
Imaginary zero is used to create a null at certain frequency.g y s s q y
전자정보대학 김영석 14-10
14.1.4 SensitivitySensitivity measures the variation of a filter parameter due to
dP dCvariation of a filter component.
P=ParameterPC
dP dCSP C
=P Parameter
C=ComponentP C C Component
Ex14 6: Sensitivity ( )/1=ω CREx14.6: Sensitivity ( )1
/1
20
110
−=
=ω
ω
CRdRd
CR
10
111
−=ω
RdRdCRdR
10
1
10
−=ω
ω
RS
R
전자정보대학 김영석 14-11
14.2 First-Order Filters
1( ) s zH s α +=
1( )H s
s pα=
+
L /hi h filt b li d b h i th l tiLow/high pass filters can be realized by changing the relative positions of poles and zeros.
전자정보대학 김영석 14-12
KnowhowPole 구하는 법: Vin=0, Vout=∞, Output Node에서의 시정수
))(||(1
2121 CCRRwp +
=
Zero 구하는 법: Vout= 0, Input Node에서의 시정수
))(||( 2121
1
11
1CR
wz =
|H(w=0)|=? C open, L short
2RVout =
|H(w=∞)|=? C short, L open
21 RRVin +=
|H(w ∞)| ? C short, L open
21
1
CCC
VV
i
out
+=
전자정보대학 김영석 14-13
21 CCVin +
Ex14.8: First-Order Filter I
R1*C1: zeroR1 C1: zero
R2C2 < R1C1 R2C2 > R1C1
전자정보대학 김영석 14-14
Ex14.9: First-Order Filter II
R1*C1: zeroR1 C1: zero
R2C2 < R1C1 R2C2 > R1C1
전자정보대학 김영석 14-15
14.3 Second-Order Filters
22
2
)(wsws
sssHn ++
++=
γβα
nws
Qs ++
1,2 211
2 4n
np jQ Qω ω= − ± − 22 4Q Q
Second-order filters are characterized by the “biquadratic” equation with two complex poles shown above.
전자정보대학 김영석 14-16
Second-Order Low-Pass Filter
α=β=022
)(ws
Qws
sHn ++
=γ
nQ0)( ,)0( 2 =∞= H
wH
n
γ
전자정보대학 김영석 14-17
Ex14.10: Second-Order LPF
3Q2)(n
wQwH n
γ=
2
3
/ 1 1/(4 ) 3
Q
Q Q
=
− ≈
2
/ 1 1/(4 ) 3
1 1/(2 )n n
Q Q
Qω ω
≈
− ≈
전자정보대학 김영석 14-18
Second-Order High-Pass Filter
2
2 2( )
n
sH s αω= β=γ=0
2 2nns s
Qω+ +
α=∞= )( ,0)0( HH
전자정보대학 김영석 14-19
Second-Order Band-Pass Filter
2 2( )
n
sH s βω= α=γ=0
2 2nns s
Qω ω+ +
0)(,)( ,0)0( =∞== HwQwHHn
nβ
전자정보대학 김영석 14-20
14.3.2 LC Realization of Second-Order Filters
L 1) ( 0)0( 11 shortLZ =
11 2
1 1 1L sZ
L C s=
+ ) ( 0)(
)2
1(
11
1
shortCZLC
fZ
=∞
∞==π
An LC tank realizes a second-order band-pass filter with two imaginary poles at ±j/(L1C1)1/2 , which implies infinite i d t 1/(L C )1/2impedance at ω=1/(L1C1)1/2.
전자정보대학 김영석 14-21
RLC Realization of Second-Order Filters
1 12 2
R L sZ =2 21 1 1 1 1R L C s L s R+ +
11 1 1 Lp j= − ± −) ( 0)0( 12 shortLZ =1,2 2
1 1 1 11 11
2 4p j
R C R CL C= − ± −
)2
1( 111
2
12
RCL
fZ ==π
) ( 0)( 12 shortCZ =∞
With a resistor, the poles are no longer pure imaginary which implies there will be no infinite impedance at any ω.
전자정보대학 김영석 14-22
Voltage Divider Using General Impedances
( )out PV Zs =( )in S P
sV Z Z+
Low-pass High-pass Band-pass
전자정보대학 김영석 14-23
Low-pass Filter Implementation with Voltage Divider
( ) 12
1 1 1 1 1
out
in
V RsV R C L s L s R
=+ +1 1 1 1 1in
0),(pen )0( :),0( :0
11
1
=∞==∞====
outsp
inouts
VZoLorZshortCfVVZshortLf
전자정보대학 김영석 14-24
Ex14.14: Frequency Peaking
( ) 1outV Rs( ) 12
1 1 1 1 1
out
ins
V R C L s L s R=
+ +
1Q > Peaking exists2
Q > Peaking existsVoltage gain larger than unity
전자정보대학 김영석 14-25
Ex14.15 Low Pass Circuit Comparison
Good BadGood Bad
pen, :0 11= shortCoLf pen,:0 11= shortCoLf0
p , 11
=outVf
방해유한 0,,p ,
1
11
== outVRButf
The circuit on the left has a sharper roll-off at high frequency than the circuit on the right.than the circuit on the right.
전자정보대학 김영석 14-26
High-pass Filter Implementation with Voltage DividerDivider
( )2
1 1 12
1 1 1 1 1
out
in
V L C R ssV R C L s L s R
=+ +
out
VVoLshortCfVshortLoCf
=∞===
pen:0, pen, :0 11
inout VVoLshortCf =∞= ,pen, : 11
전자정보대학 김영석 14-27
Band-pass Filter Implementation with Voltage Divider
2V L s( ) 12
1 1 1 1 1
out
in
V L ssV R C L s L s R
=+ +
, /:2
10,:0
11
1
==
==
inout
out
VVopenLCCL
f
VshortLf
π0, :
2
1
11
=∞= outVshortCfCLπ
전자정보대학 김영석 14-28
14.4 Active Filters14.4.1 Sallen and Key (SK) Filter: Low-Pass
11 CQ R R 11 2
1 2 2Q R R
R R C=
+
1nω =
( )( )2
1 2 1 2 1 2 2
11
out
in
V sV R R C C s R R C s
=+ + +
1 2 1 2n R R C C
Sallen and Key filters are examples of active filters. This ti l filt i l t l d d t fparticular filter implements a low-pass, second-order transfer
function.
전자정보대학 김영석 14-29
Ex14.16 Sallen and Key (SK) Filter: Band-pass
31 R+
Pass-band gain>1
( )3
4
2 3
1
1
out
in
V RsV RR R C C s R C R C R C s
+=
⎛ ⎞+ + +⎜ ⎟1 2 1 2 1 2 2 2 1 1
41R R C C s R C R C R C s
R+ + − +⎜ ⎟⎝ ⎠
전자정보대학 김영석 14-30
Ex14.17: SK Filter Poles
C1=C2
R1=R21 2
전자정보대학 김영석 14-31
Sensitivity in Band-Pass SK Filter
1 2 1 2
12
n n n nR R C CS S S Sω ω ω ω= = = = −
1 2
2 2 1 212
Q QC C
R C R CS S QR C R C
⎛ ⎞= − = − + +⎜ ⎟⎜ ⎟
⎝ ⎠1 2 1 2 2
2 21Q Q R C
1 21 1 2 12C C R C R C⎜ ⎟
⎝ ⎠
Q R C1 2
2 2
1 1
12
Q QR R
R CS S QR C
= − = − + 1 1
2 2
QK
R CS QKR C
= K=1+R3/R4
전자정보대학 김영석 14-32
Ex14.18: SK Filter Sensitivity I
1 2R R R= =
1 2C C C= =
1 2
1 12 3
Q QR RS S
K= − = − +
−
1 2
1 22 3
Q QC CS S
K= − = − +
−
3QK
KSK
=−
전자정보대학 김영석 14-33
Ex14.19: SK Filter Sensitivity II
2Q =2K =
1
2 2
1 1
3 14
QR
R C SR C
= ⇒ =
1
1 1
2 2
1 58 4
QC
R C SR C
= ⇒ =2 2 8 4
81 5
QK
R C
S =1.5
전자정보대학 김영석 14-34
14.4.2 Integrator-Based Biquads
( )2
2 2
out
nin
V ssV s s
αω ω
=+ +
( ) ( ) ( ) ( )2
21.n n
out in out outV s V s V s V sQ s sω ωα= − −
ns sQ
ω+ +
It is possible to use integrators to implement biquadratic transfer s p ss s g s p q s functions.
The block-diagram above illustrates how.
전자정보대학 김영석 14-35
KHN Biquads
( ) ( ) ( ) ( )21ω ω( ) ( ) ( ) ( )2
1.n nout in out outV s V s V s V s
Q s sω ωα= − −
5 6
4 5 31R R
R R Rα
⎛ ⎞= +⎜ ⎟+ ⎝ ⎠
4
4 5 1 1
1.n RQ R R R Cω
=+
2 6
3 1 2 1 2
1.nRR R R C C
ω =4 5 3⎝ ⎠ 4 5 1 1Q
전자정보대학 김영석 14-36
Versatility of KHN Biquads
2V sα
High-Pass2 1V sα
Band-Pass
( )2 2
out
ninn
V ssV s s
Q
αω ω
=+ +
( )2 2 1 1
1.X
ninn
V ssV R C ss s
Q
αω ω
−=
+ +
2 1V sα
Low-Pass
( ) 22 2 1 2 1 2
1.Y
ninn
V ssV R R C C ss s
Q
αω ω
=+ +
전자정보대학 김영석 14-37
Sensitivity in KHN Biquads
0 5nR R C C R R R RSω = 0 5QS1 2 1 2 4 5 3 6, , , , , , , 0.5R R C C R R R RS
1 2 1 2, , , 0.5QR R C CS =
3 6
3 6 2 2,
5 3 6 1 12 1
QR R
R R R CQS R R R R CR
−=
+ 4 5
5,
4 51Q
R RRS
R R= <
+4R
Low sensitivities
전자정보대학 김영석 14-38
Tow-Thomas Biquad
Merge the adder and the first integratorMerge the adder and the first integrator
3 42
1 2 3 4 1 2 2 4 2 3
1.Y
in
R RVV R R R R C C s R R C s R
=+ +
2 3 4 22
1 2 3 4 1 2 2 4 2 3.out
in
V R R R C sV R R R R C C s R R C s R
= −+ +
Band-Pass Low-Pass
전자정보대학 김영석 14-39
Differential Tow-Thomas Biquads
By using differential integrators, the inverting stage is eliminated.
전자정보대학 김영석 14-40
14.4.3 Simulated Inductor (SI)
1 35in
Z ZZ ZZ Z
= 52 4
in Z Z
It is possible to simulate the behavior of an inductor by usingIt is possible to simulate the behavior of an inductor by using active circuits in feedback with properly chosen passive elements.
전자정보대학 김영석 14-41
Simulated Inductor I
2in X YZ R R Cs=
By proper choices of Z1-Z4, Zin has become an impedance that increases with frequency, simulating inductive effect.
Z5=RX, Z2=1/(Cs), Z1=Z3=Z4=RY
전자정보대학 김영석 14-42
Ex14.21: Simulated Inductor II
1 2 3
1YZ Z Z R
Z
= = =
4
2
ZCs
Z R R Cs
=
=in X YZ R R Cs=
전자정보대학 김영석 14-43
High-Pass Filter with SI
( )2
12
1 1 1 1 1
out
in
V L ssV R C L s L s R
=+ +
With the inductor simulated at the output, the transfer function resembles a second-order high-pass filter.
⎛ ⎞Node 4 is also an output node
4 1 Yout
X
RV VR
⎛ ⎞= +⎜ ⎟
⎝ ⎠
Node 4 is also an output node
전자정보대학 김영석 14-44
⎝ ⎠
Low-Pass Filter with Super Capacitor
( )1
1inX
ZCs R Cs
=+( )X
Product of 2 C
=> Super Capacitor
1out inV Z= =
=> Super Capacitor
2 21 1 1 1in in XV Z R R R C s R Cs+ + +
Low-PassLow Pass
전자정보대학 김영석 14-45
Ex14.23: Poor Low Pass Filter
( )4 2out XV V R Cs= +
Node 4 is no longer a scaled version of the Vout. Therefore the output can only be sensed at node 1, suffering from a high impedance.p
전자정보대학 김영석 14-46
14.5 Approximation of Filter ResponseFrequency Response Template
With all the specifications on pass/stop band ripples and t iti b d l t filt t l t th ttransition band slope, one can create a filter template that will lend itself to transfer function approximation.
전자정보대학 김영석 14-47
14.5.1 Butterworth Response
12
1( )
1n
H jωω
=⎛ ⎞
+ ⎜ ⎟⎝ ⎠0ω⎜ ⎟⎝ ⎠
ω-3dB=ω0 for all n3dB 0,
The Butterworth response completely avoids ripples in the pass/stop bands at the expense of the transition band slope.pass/stop bands at the expense of the transition band slope.
전자정보대학 김영석 14-48
Poles of the Butterworth Response
2 1j kπ ⎛ ⎞0
2 1exp exp , 1, 2, ,2 2kj kp j k n
nπω π−⎛ ⎞= =⎜ ⎟
⎝ ⎠L
2nd‐Order nth‐Order
전자정보대학 김영석 14-49
Ex14.24: Butterworth Order
22 64 2
nf⎛ ⎞
=⎜ ⎟1
64.2f
=⎜ ⎟⎝ ⎠
2 12f f=2 1
n=3
The Butterworth order of three is needed to satisfy the filter response on the leftresponse on the left.
전자정보대학 김영석 14-50
Ex14.25: Butterworth Response
12 22 *(1.45 )* cos sin3 3
p MHz jπ ππ ⎛ ⎞= +⎜ ⎟⎝ ⎠
RC section
32 22 *(1.45 )* cos sin3 3
p MHz jπ ππ ⎛ ⎞= −⎜ ⎟⎝ ⎠ 2 2 *(1.45 )p MHzπ=
2nd-Order SK
전자정보대학 김영석 14-51
14.5.2 Chebyshev Response
( )2 2
1
1
H j
C
ωωε
=⎛ ⎞
+ ⎜ ⎟0
1 nCεω
+ ⎜ ⎟⎝ ⎠
Chebyshev Polynomial
The Chebyshev response provides an “equiripple” pass/stop band response.
전자정보대학 김영석 14-52
Chebyshev Polynomial
Chebyshev Polynomial for n=1,2,3
Resulting Transfer function forn=2,3, ,
10
0 0cos cos ,nC nω ω ω ω
ω ω−⎛ ⎞ ⎛ ⎞
= <⎜ ⎟ ⎜ ⎟⎝ ⎠ ⎝ ⎠
10
0cosh cosh ,n ω ω ω
ω−⎛ ⎞
= >⎜ ⎟⎝ ⎠
전자정보대학 김영석 14-53
Ex14.26: Chebyshev Attenuation
( )23
2
1
1 0 329 4 3
H jω
ω ω
=⎡ ⎤⎛ ⎞⎢ ⎥+ ⎜ ⎟
2
0 01 0.329 4 3
ω ω⎢ ⎥+ −⎜ ⎟⎢ ⎥⎝ ⎠⎣ ⎦
ω0=2π X (2MHz)
A third-order Chebyshev response provides an attenuation of -18.7 dB a 2MHz.
전자정보대학 김영석 14-54
Ex14.27: Chebyshev Order
Passband Ripple: 1 dB
Bandwidth: 5 MHz
Attenuation at 10 MHz: 30 dBAttenuation at 10 MHz: 30 dB
What’s the order?
( )2 2 1
1 0.03161 0 509 cosh cosh 2n −
=+
n>3.66( )1 0.509 cosh cosh 2n+
전자정보대학 김영석 14-55
Ex14.28: Chebyshev Response
( ) ( )1 12 1 2 11 1 1 1k kπ π− −⎛ ⎞ ⎛ ⎞( ) ( )1 10 0
2 1 2 11 1 1 1sin sinh sinh cos cosh sinh2 2k
k kp j
n n n nπ π
ω ωε ε
− −⎛ ⎞ ⎛ ⎞= − +⎜ ⎟ ⎜ ⎟⎝ ⎠ ⎝ ⎠
K=1 2 3 4K=1,2,3,42,3 0 00.337 0.407p jω ω= − ±
SK21,4 0 00.140 0.983p jω ω= − ±
SK1
전자정보대학 김영석 14-56