Sei sulla pagina 1di 10

Digital Signal Processing

5/2/2015

Overview
Comparison of FIR and IIR
Digital Signal Processing: Conversion Analog to Digital Filter
Bilinear & Impulse Invarient
IIR filter design

Designing Butterworth Filter


Analog and Digital Filter
Designing Chebyshev Filter
Nikesh Bajaj Analog and Digital Filter
nikesh.14730@lpu.co.in
Digital Signal Processing
nikeshbajaj.in 2 By: Nikesh Bajaj

FIR & IIR IIR


Recursive & Non recursive Many ways to design IIR Filter
Stable and Unstable Robust and Widely used way is
Design Analog filter then convert to
digital

3 By: Nikesh Bajaj 4 By: Nikesh Bajaj

Analog to Digital Filter Analog to Digital Filter


Lets consider you already have analog filter Analog Filter
How would you convert analog to digital
filter?

By: Nikesh Bajaj 6 By: Nikesh Bajaj

1
nikeshbajaj.in
Digital Signal Processing

5/2/2015

S-Plane to Z-Plane : Mapping Impulse Invariant Method


H(s) > h(t) > h(nT)=h(n) > H(Z) H(s) > h(t) > h(nT)=h(n) > H(Z)
Poles and Zeros

7 By: Nikesh Bajaj 8 By: Nikesh Bajaj

Impulse Invariant Method: Conclusion Filters


N N
Ck Ck
H a ( s )= ---- >H ( z )= pkT 1
k=1 s p k k=1 1 e z
s+a
H a ( s )=
(s+a )2 +b 2
1 e aT cos(bT ) z 1
H ( z )= aT 1 2aT 1
1 2 e cos(bT ) z +e z
b
H a ( s )=
(s+a )2 +b 2
e aT sin (bT ) z 1
H ( z )= aT 1 2aT 1
1 2 e cos(bT ) z +e z
9 By: Nikesh Bajaj 10 By: Nikesh Bajaj

Filters Characteristics Bilinear Transformation

11 By: Nikesh Bajaj 12 By: Nikesh Bajaj

2
nikeshbajaj.in
Digital Signal Processing

5/2/2015

Conclusion Bilinear Transformation: Problem


Convert H(s) to H(z) with resonant frequency at wr = pi/4

13 By: Nikesh Bajaj 14 By: Nikesh Bajaj

Impulse Invariant Method Butterworth Filter


Analog to Digital All pole Filter*
Maximally Flat response Filter

15 By: Nikesh Bajaj 16 Nikesh Bajaj

Butterworth Filter

17 Nikesh Bajaj 18 By: Nikesh Bajaj

3
nikeshbajaj.in
Digital Signal Processing

5/2/2015

Poles: Butterworth Filter: Transfer Function


N
c
H a ( s )= N /2
N even
( s2 +b k c s+ 2c )
i=1

N
c
H a ( s )= ( N 1 )/ 2
N Odd
( s+ c ) ( s 2 +b k c s+ 2c )
i=1

b k = 2sin (2k-1
2N
)

19 By: Nikesh Bajaj 20 Nikesh Bajaj

Butterworth Filter: N & c Butterworth Filter: N & c


s 2 1

N B=
log 10
( )
p 2 1 p = 10 Ap /20
2log 10 ( s / p ) s = 10 As/20

log 10 ( s 2 1 )
N=
2log 10 ( s / c )

21 By: Nikesh Bajaj 22 By: Nikesh Bajaj

Designing Analog filter and


converting to digital Problems
Given Determine Order and Poles of Low pass Butterworth filter
Specification
for that has -3dB bandwidth at 500Hz and attenuation of 40
of Analog
Filter dB at 1000Hz
Convert same filter to Digital filter using Bilinear Method
Compute with sampling frequency of 4KHz
Order and
3dB freq. of
filter N, c

Compute
H(z)
Digital Compute
Filter H(s)
Analog
filter By: Nikesh Bajaj 24 By: Nikesh Bajaj

4
nikeshbajaj.in
Digital Signal Processing

5/2/2015

s 2 1
log 10
( )
p 2 1
Problems N B=
2log 10 ( s / p ) Designing Digital filter
Design Low Pass Butterworth filter to meet the following Given
Specification Convert
specifications of Digital Specification
Passband gain = 0.89 Filter of Digital
filter to
Pass band frequency edge =30Hz
Specification
Attenuation = 0.20 of Analog Compute
Stop band edge =75Hz Filter Order and
3dB freq. of
filter N, c

Compute
H(z)
Digital Compute
Filter H(s)
Analog
25 Nikesh Bajaj filter By: Nikesh Bajaj

2 1
log 10 s 2
1
NB = p
2log 10 ( s / p )
Given
Problems Specification Convert
Specification
of Digital
Filter of Digital
Design a Digital lowpass Butterworth filter to meet the filter to
Steps
following specifications, use bilinear transformation Specification
Digital Filter Design of Analog
Passband ripple < = 1.25dB
1. Normalize frequency with Fs Filter
Pass band edge =200Hz
Stop band Attenuation > = 15dB 2. prewarp -Find analog frequency
Stop band edge =300Hz 3. Find N and c of analog filter Compute
Order and
Sampling frequency = 2KHz 4. find the Transfer Function of analog filter 3dB freq. of
filter N, c
5. find transfer function of digital filter
Compute
H(z)
Compute
Digital
H(s)
Filter
Analog
27 Nikesh Bajaj filter

Problems Chebyshev Filter


MAKE sure to understand term Optimum, except Elliptical filters
Design DIGITAL Filter Narrowest Tr. Band
Design ANALOG Filter Equi-ripple, monotonous
Type 1 and Type II

Steps are different in both cases


See both kind of example in Proakis at end of
book. Example 24 and 25 pg : 1112-1116

29 By: Nikesh Bajaj 30 Nikesh Bajaj

5
nikeshbajaj.in
Digital Signal Processing

5/2/2015

Chebyshev Filter Chebyshev Filter


1
Type 1 and Type 2 | H a ( j )|2 =
1+ C 2N
2
p ( )
Chebyshev Poly.
C(x) = cos(Ncos-1(x)) |x|<1
C(x) = cosh(Ncosh-1(x)) |x|>1
1
| H a ( j )| 2 1 for 0 p
1 + 2

31 Nikesh Bajaj 32 By: Nikesh Bajaj

Chebyshev Filter Chebyshev Filter


1
| H a (jj |2 = CN(x) = 2xCN-1(x) - CN-2(x) CN(0) = 0 N odd

1+ 2C N2 =+-1 N Even

p Behavior at 0
Chebyshev Poly.
CN(x) = cos(Ncos-1(x)) |x|<1 CN(0) = 0 N odd
CN(x) = cosh(Ncosh-1(x)) |x|>1 =+-1 N Even
C0(x) =
C1(x) =

CN(x) = 2xCN-1(x) - CN-2(x)


33 Nikesh Bajaj 34 Nikesh Bajaj

Chebyshev Filter: Transfer Function


(N 1 ) / 2
pN C0 C k
k=1
H a (s)= (N 1 ) / 2
N Odd
2
(s + pC0 ) (s + bk p s + Ck p2 )
k=1

N /2
1
pN C k .
H a (s) = k =1 1+ 2 N Even
N /2
2
(s + bk p s + C k p2 )
k =1

35 Nikesh Bajaj

6
nikeshbajaj.in
Digital Signal Processing

5/2/2015

Chebyshev Filter and N


C0 = yN
2k 1
Ck = yN2 + cos2
2N
2k 1
bk = 2 yN sin
2N
1/ N 1/ N
1 1 1 1 1
yN = 1+ 2 + 1+ 2 +
2

37 Nikesh Bajaj

Order of Filter N Let's take N =4 and = 0.8

s2 1
cosh 1
p 2 1
NC =
cosh 1 ( s / p )

p = 10 Ap /20 Nc = integer so Nc >=N


s = 10 As/20

By: Nikesh Bajaj 39 By: Nikesh Bajaj 40

Buttorworth Vs Chebyshev Butterworth vs Chebyshev Design


1
| H a (jj |2=

1+ 2C N2

p
2 1 s2 1
log 10 s 2 cosh1
1
p p2 1
NB = NC =
2log 10 ( s / p ) cosh1(s / p )

1 1
3dB = pcosh cosh1
N
By: Nikesh Bajaj 41

7
nikeshbajaj.in
Digital Signal Processing

5/2/2015

Butterworth vs Chebyshev Design Chebyshev Filter


N /2
pN Ck .
1 C0 = yN
cN
H a (s) = N even H a (s) = k =1 1+ 2
N /2
N /2 2k 1
Ck = yN2 + cos2
2 2
(s + bk c s + )
c
(s 2
+ bk p s + Ck p2 )
2N
i=1
k =1

cN pNC0
(N1)/ 2
2k 1
Ha(s)= N Odd C k bk = 2 yN sin
(N1)/ 2
(s+c ) (s2 +bk cs+c2 ) Ha(s)= (N1)/ 2
k=1
2N
i=1
(s+pC0 ) (s2 +bkps+Ck p2 ) 1/ N 1/ N
k=1 1 1 1 1 1
yN = 1+ 2 + 1+ 2 +
2k - 1
2
bk = 2sin 2k 1
bk = 2 y N sin

2N 2N
43 Nikesh Bajaj 44 Nikesh Bajaj

Chebyshev Filter: 3dB frequency


1 1
3dB = p cosh cosh 1
N

46 By: Nikesh Bajaj

Problems Problems
Design Chebyshev filter with pass band frq =1.5KHz stop MAKE sure to understand term
band frequency =6KHz passband ripple =0.9 stop band
ripple =0.01 Design DIGITAL Filter
Design digital filter satisfying above specifications for Design ANALOG Filter
sampling frequency of 20KHz
Try Designing filters with and Notice the change
Steps are different in both cases
p =0.8 p = 1000 s =0.1 s 3000
See both kind of example in Proakis at end of
p =0.8 p = 1000 s =0.1 s 2000
book. Example 24 and 25 pg : 1112-1116
p =0.8 p = 1000 s =0.1 s 1.500

47 By: Nikesh Bajaj 48 By: Nikesh Bajaj

8
nikeshbajaj.in
Digital Signal Processing

5/2/2015

Comparison Order to compare NC and NB


Passband Att = -1dB
Stopband Att =-40dB
Passband Frq = 1KHz
Stopband Frq = 5KHz

49 By: Nikesh Bajaj 50 By: Nikesh Bajaj

Frequency Transformation
Normalized low pass filter NLPF NLPF Normalized Low Pass Filter
NLPF to Any LPF
p
NLPF to High Pass Filter
NLPF to Band Pass
NLPF to Band Reject s
s

51 Nikesh Bajaj By: Nikesh Bajaj

02 p1 p 2 s1 s 2 Normalized Low pass


Bilinear Convert Convert Analog Filter design
B p 2 p1 transformation Specification Specification
to Analog to Normalized
Q B / 0 Filter Low Pass
Filter
Compute
s Order N,
NLPF to Any LPF S= s
p Given
p Specification
NLPF to High Pass Filter S= of Digital Designing IIR
s Filter


Q s 2 02 Digital filter
NLPF to Band Pass S Compute
0s H(s) of
Frequency
Transformation Normalized
0s Compute
NLPF to Band Reject S H(z) Lowpass

Q s 2 02 Digital
Compute H(s)
of De-
Analog filter
Filter normalized
Analog Filter
53 By: Nikesh Bajaj Nikesh Bajaj

9
nikeshbajaj.in
Digital Signal Processing

5/2/2015

Simple Digital Filters FIR


Match the Column
BRF 1

H ( z )= (1 +z 1 )
2
1
BPF H ( z )= (1 z 1 )
2
1
LPF H ( z )= (1 z 2 )
2
1
H ( z )= (1 +z 2 )
HPF 2
By: Nikesh Bajaj 56 Nikesh Bajaj

Readings and Working v


http://www.analog.com/designtools/en/filter
wizard/#/specifications

57 By: Nikesh Bajaj 58 By: Nikesh Bajaj

Problems
Determine Order and Poles of Low pass Butterworth filter
for that has -3dB bandwidth at 500Hz and attenuation of 40
dB at 1000Hz
Convert same filter to Digital filter using Bilinear Method
with sampling frequency of 4KHz

59 By: Nikesh Bajaj 60 By: Nikesh Bajaj

10
nikeshbajaj.in

Potrebbero piacerti anche