Digital Filters: Vedat Tavşanoğlu

DIGITAL FILTERS
Vedat Tavanolu
Frequency Response of Ideal Filters

Low-pass Filters
H ILP ( e j )
1 -4 -3 -2 - c 0 c 2 3 4
H ILP ( e
1 for c c )= 0 for c
and c
H ILP ( e j ) = H ILP ( e j ( 2 k ) ) ; k = 1,2,........

DIGITAL FILTERS Vedat Tavanolu 2

High-pass Filters 1 -4 -3 -2 - c 0 c 2 3 4
H ILP ( e j )
H IHP ( e j )
1 -4 -3 -2 - l 0 l 2 3 4
H IHP ( e j ) = H ILP ( e j ( ) ) ;
DIGITAL FILTERS Vedat Tavanolu
c = l
3

Band-pass Filters 1 -4 -3 -2 - c 0 c 2 3 4
H ILP ( e j )
H IBP ( e j )
1 -4 -3 -2 - 0 0 0 2 3 4
H IBP ( e j ) = H ILP ( e j ( 0 ) ) + H ILP ( e j( +0 ) )


Band-stop Filters -4 -3 -2 - 0
H IBP ( e j )
0 0 2 3 4
H IBP ( e j )
-4 -3 -2 - 0 0 0 2 3 4
H IBS ( e j )
-4 -3 -2 - 0 0 0 2 3 4
H IBS ( e
) = 1 H IBP ( e
)
5

All-pass Filters
| H IAP ( e j ) |
-4 -3 -2 - 0 0 0 2 3 4
Impulse Response of Ideal Low-pass Filters

H ILP ( e
j
1 for c c )= 0 for c
and c
H ILP ( e j ) = H ILP ( e j ( 2 k ) ) ; k = 1,2,........
hILP ( n ) = F { H ILP ( e
)}
7

1 hILP ( n ) = 2
H ILP ( e
)e
jn
1 d = 2
jn e d
1 sin c n j c n j c n e hILP ( n ) = (e )= 2jn n

close all n=-20:1:20; wc=pi/2; x=sin(wc*(n+.0001)); h=x./(pi*(n+.0001)); stem(n,h) title('h(n)=sin(wc*n)/(pi*n); wc=pi/2') xlabel('n')
10
11

Amplitude values of
sin c n hILP ( n ) = n
3
-0.11 0.08 0.11
for c=/2, /4, and /6
c
2
0
0.5 0.25
1
0.32 0.23 0.16
2
0 0.16 0.14
4
0 0 0.07
5
0.06 -0.05 0.03
6
0 -0.05 0
7
-0.05 -0.03 -0.02
8
0 0 -0.03
9
0.04 0.03 -0.04
10
0 0.03 -0.03
4
1/6
12
Design of Low-pass Filters
h( n ) = { h( 0 ), h( 1 ),........, h( N 1 ), h( N )}
1 0 x(n) h(N) h(N-1) h(N-2) . . . . h(1) h(0)
Let x(n)=(n),i.e., a unit sample, then the output of the system Shown below is given as:
h(0) 0
h(1) 1
.................
h(N) N
Ts
Ts
Ts
13
y(n)

The two features of such a system are: The response to the unit-sample is of finite duration. The response to the unit-sample starts at time zero,i.e., at n=0.
14

However, the impulse response of the ideal low-pass filter,
is of infinite duration and extends symmetrically to infinity on both sides of n=0.

This implies that the impulse response of the ideal low-pass filter is not realisable with such a system. The question is: How can we make such an impulse response realisable?

Obviously by making it: to have finite duration. to start at time zero. This can be acomplished by: first truncating it on both ends to make it finite duration. secondly shifting to the right to make it start at time zero .
17

The Fourier transform of
is given as:
H ILP ( e
)=
n =
ILP
( n )e
jn
18

This can also be interpreted as the Fourier series expansion of HILP(ej), which is the periodic square waveform with duty cycle 2c. Since HILP(ej) is symmetrical around n=0, we can write: j ILP ILP ILP n =0
(e
) = h ( 0 ) + 2 h ( n ) cos n
Now consider the truncation of hILP(n) at both ends.


Truncated impulse response is given by:
h t ( n ) = hILP ( n )[ u( n + N ) u( n N )]
The Fourier transform of ht(n) is obtained as:
Ht ( e
) = hILP ( 0 ) + 2 hILP ( n ) cos n

n =1

We have
Ht ( e
Assuming
c 2 N sin c n + )= cos n n=1 n
c = 2
Ht ( e
1 1 2 1 ) = + (cos cos 3 + .... + cos N ) 2 3 N


In the following we will deal with the Frequency Response of Truncated Impulse Response or equivalently we can say Truncated Fourier Series Expansion of Ideal Filter Frequency Response.

MATLAB Program to Compute the Frequency Response of Truncated Impulse Response from n = 6 close all title('Truncated Fourier Series w=-2*pi:pi/64:2*pi; Expansion of Ideal Filter wc=pi/2; Frequency Response'); h=wc/pi; xlabel('w'); n=1; ylabel('|HILP(expjw)|') while n<6 x=sin(wc*(n+.0001)); x1=x./(pi*(n+.0001)); x2=2*x1*cos(w*n); h=h+x2; n=n+1; end figure,plot(w,h);

Frequency Response of Truncated Impulse Response from n = 6 plotted for 2<<2
24

25

26

Frequency Response of Truncated Impulse Response from plotted for <<
n = 32
27

What remains to be done to obtain a realisable frequency response is to shift ht(n) to the right by N, i.e., the realisable impulse response is given as:
Since:
h( n ) = ht ( n N )
F { ht ( n N )} = e
jN
Ht( e
j
we obtain
F { h( n )} = e
jN
Ht ( e
)
28

H( e
j
) = ( hILP ( 0 ) + 2 hILP ( n ) cos n )e

n =0
j
jN
H( e
c 2 N sin c n jN + )=( cos n )e n =1 n
Given c , the design of the truncated low-pass filter is straightforward.

Design of HP, BP and BS Filters

The low-pass design is also used in the design of HP,BP and BS filters. For such filters the frequency specifications are first translated to the frequency specifications of the corresponding lowpass filter.Then the impulse response of the required filter is obtained through relationships given above:

High-pass Filters
The relation between the frequency responses of highand low-pass filters was given as:
H HP ( e
We know that:
) = H LP ( e
j ( )
)
jn
H LP ( e
)=
n =
LP
( n )e
Using the above relationship we can write

H HP ( e =
) = H LP ( e ( n )e
jn
j ( )
)=
n = jn
LP
( n )e
j ( ) n
n =
LP
n =
( 1 ) h
n n
LP
( n )e
jn
where we used the identity
jn
= ( 1 )
32

We can conclude that:
hHP ( n ) = ( 1 ) hLP ( n )
n
33
34
35
36

Amplitude values of
sin c n hIHP ( n ) = ( 1 ) for c = / 2 , / 4, / 6 n

n
c
2
0
0.5 0.25
1
-0.32 -0.23 -0.16
2
0 0.16 0.14
3
-0.11 -0.08 -0.11
4
0 0 0.07
5
-0.06 0.05 -0.03
6
0 -0.05 0
7
0.05 0.03 0.02
8
0 0 -0.03
9
-0.04 -0.03 0.04
10
0 0.03 -0.03
4
1/6
37

close all w=-2*pi:pi/64:2*pi; wc=pi/2; h=wc/pi; n=1; while n<6 x=((-1).^n)*sin(wc*(n+.0001)); x1=x./(pi*(n+.0001)); x2=2*x1*cos(w*n); h=h+x2; n=n+1; end figure,plot(w,h); title('Truncated Fourier Series Expansion of Ideal HP Filter Frequency Response; wc=pi/2'); xlabel('w'); ylabel('|HIHP(expjw)|')

39

Frequency Response of Truncated Impulse Response from plotted for 2<<2
n = 16
40

n = 32
41

n = 32
42

Q.1 An analogue high-pass filter approximating the frequency characteristic given in the figure below is to be designed using the system given on the next page which is composed of an A/D converter, a 5-coefficient causal Finite Impulse Response (FIR) digital filter and a D/A converter, where A/D and D/A converters are assumed to be perfect.
The lower cutoff frequency of the analogue high-pass filter and the maximum frequency in the analogue input signal are given as: Fl=60 kHz and Fmax=80kHz.
43

(a)Find and plot the impulse response hILP(n) of the corresponding ideal low-pass filter. (b)Find and plot the impulse response hLP(n) of the 5-coefficient causal filter. (c)Give the expression for HLP(ej) and |HLP(ej)| and sketch |HLP(ej)| roughly. (d)Give hhP(n) and HHP(ej). Plot |HLP(ej)| roughly.
44

H HP ( j )
1
45
Analogue Filter
xa (t )
A/D
x (n)
Digital Filter
y (n)
D/A
y a (t )
Fs
Fs
46

Solution: (a)Minimum sampling frequency is found as:
Fs = 2 Fmax = 80kHz
Using this frequency we find:
Fl 20 1 = = fl = Fs 80 4
l = 2
47

H ILP ( e
1 -3 -2 - l 0 l 2 3 4
j
The cutoff frequency of the corresponding prototype lowpass filter can be found as:
The impulse response of the prototype lowpass filter:
c = l = 2
48
sin c n sin 2 n = hILP ( n ) = n n
49

(b)The impulse response of the 5-coefficient (truncated) causal filter is given as:
sin ( n 2 ) 2 hLP ( n ) = [ u( n ) u( n 2 )] ( n 2 )
50
51

(c)
H LP ( e
) = [ hLP ( 0 ) + 2hLP ( 1 ) cos ] e
j 2
= [ 0.5 + 0.64 cos ] e | H LP ( e

j
j 2
) |=| 0.5 + 0.64 cos |

53

(d)
hHP ( n ) = ( 1 )( n 2 ) sin ( n 2 ) 2 [ u( n ) u( n 2 )] ( n 2 )
j 2
H HP ( e
) = [ hLP ( 0 ) 2hLP ( 1 ) cos ] e
= [ 0.5 0.64 cos ] e
j 2
| H HP ( e
) |=| 0.5 0.64 cos |

55

Band-pass Filters
The relation between the frequency responses of ideal bandand low-pass filters was given as:
H IBP ( e
) = H ILP ( e
j ( 0 )
) + H ILP ( e
j ( + 0 )
Using the above relationship we can write
H IBP ( e j ) =
n =
j ( 0 ) n h ( n ) e + ILP
n =
j ( + 0 ) n h ( n ) e ILP
56

H IBP ( e
j
)= = = =
n =
ILP
( n )e
j ( 0 ) n
n =
ILP
( n )e
j ( + 0 ) n
n =
jn j0 n + h ( n ) e e ILP
n =
jn j0 h ( n ) e e ILP
n =
j 0 n j 0 jn + h ( n ) e ( e e ) ILP
n =
( 2h
ILP
( n ) cos 0 n ) e
jn
where we used the identity
57

We can make the following identification:
hIBP = 2hILP ( n ) cos 0 n
58

MATLAB Program Computing the Impulse Response of BP Filter with w0=pi/3
close all n=-20:1:20 wc=pi/4 w0=pi/3 x=sin(wc*(n+.0001)); h=x./(pi*(n+.0001)); h=2*h.*cos(w0*n) stem(n,h) title('hIBP(n)=2*cos(w0*n)*sin(wc*n)/(pi*n); wc=pi/4, w0=pi/3') xlabel('n')

Impulse Responses of BP Filters with w0=pi/3
60

close all w=-2*pi:pi/64:2*pi; wc=pi/6; w0=2*pi/3; h=2*wc/pi; n=1; while n<33 x=sin(wc*(n+.0001)); x1=x./(pi*(n+.0001)); x1=2*x1.*cos(w0*n) x2=2*x1.*cos(w*n); h=h+x2; n=n+1; end figure,plot(w,h); title('Truncated Fourier Series Expansion of Ideal BP Filter Frequency Response; wc=pi/6,w0=2pi/3'); xlabel('w'); ylabel('|HIBP(expjw)|')
61

62

63

n = 32
64

n = 32
65

Band-stop (reject) Filters
The relation between the frequency responses of ideal band-pass and band-stop filters was given as:
H IBS ( e
) = 1 H IBP ( e
Taking the inverse Fourier transform of both sides,we obtain,
hIBS ( n ) = ( n ) hIBP ( n )
hIBS = ( n ) 2hILP ( n ) cos 0 n


n = 6
67

68

69

n = 32
70

Digital Filters: Vedat Tavşanoğlu

Diunggah oleh

Informasi Dokumen

Judul Asli

Hak Cipta

Format Tersedia

Bagikan dokumen Ini

Bagikan atau Tanam Dokumen

Opsi Berbagi

Apakah menurut Anda dokumen ini bermanfaat?

Apakah konten ini tidak pantas?

Hak Cipta:

Format Tersedia

Digital Filters: Vedat Tavşanoğlu

Diunggah oleh

Hak Cipta:

Format Tersedia

DIGITAL FILTERS

Frequency Response of Ideal Filters

H ILP ( e j ) = H ILP ( e j ( 2 k ) ) ; k = 1,2,........

Frequency Response of Ideal Filters

Frequency Response of Ideal Filters

H IBP ( e j ) = H ILP ( e j ( 0 ) ) + H ILP ( e j( +0 ) )

Frequency Response of Ideal Filters

Frequency Response of Ideal Filters

DIGITAL FILTERS Vedat Tavanolu

Impulse Response of Ideal Low-pass Filters

H ILP ( e j ) = H ILP ( e j ( 2 k ) ) ; k = 1,2,........

Impulse Response of Ideal Low-pass Filters

1 sin c n j c n j c n e hILP ( n ) = (e )= 2jn n

DIGITAL FILTERS Vedat Tavanolu

Impulse Response of Ideal Low-pass Filters

DIGITAL FILTERS Vedat Tavanolu

Impulse Response of Ideal Low-pass Filters

DIGITAL FILTERS Vedat Tavanolu

Impulse Response of Ideal Low-pass Filters

DIGITAL FILTERS Vedat Tavanolu

Impulse Response of Ideal Low-pass Filters

for c=/2, /4, and /6

DIGITAL FILTERS Vedat Tavanolu

Design of Low-pass Filters

Design of Low-pass Filters

DIGITAL FILTERS Vedat Tavanolu

Design of Low-pass Filters

Design of Low-pass Filters

Design of Low-pass Filters

DIGITAL FILTERS Vedat Tavanolu

Design of Low-pass Filters

DIGITAL FILTERS Vedat Tavanolu

Design of Low-pass Filters

Now consider the truncation of hILP(n) at both ends.

Design of Low-pass Filters

) = hILP ( 0 ) + 2 hILP ( n ) cos n

Design of Low-pass Filters

c 2 N sin c n + )= cos n n=1 n

1 1 2 1 ) = + (cos cos 3 + .... + cos N ) 2 3 N

Design of Low-pass Filters

Design of Low-pass Filters

Design of Low-pass Filters

DIGITAL FILTERS Vedat Tavanolu

Design of Low-pass Filters

DIGITAL FILTERS Vedat Tavanolu

Design of Low-pass Filters

DIGITAL FILTERS Vedat Tavanolu

Design of Low-pass Filters

DIGITAL FILTERS Vedat Tavanolu

Design of Low-pass Filters

DIGITAL FILTERS Vedat Tavanolu

Design of Low-pass Filters

) = ( hILP ( 0 ) + 2 hILP ( n ) cos n )e

c 2 N sin c n jN + )=( cos n )e n =1 n

Given c , the design of the truncated low-pass filter is straightforward.

Design of HP, BP and BS Filters

Design of HP, BP and BS Filters

Using the above relationship we can write

Design of HP, BP and BS Filters

where we used the identity

DIGITAL FILTERS Vedat Tavanolu

Design of HP, BP and BS Filters

DIGITAL FILTERS Vedat Tavanolu