Department of Electrical Engineering

University of Arkansas

ELEG 5173L Digital Signal Processing

Ch. 5 Digital Filters

Dr. Jingxian Wu
[email protected]
 2
OUTLINE
• FIR and IIR Filters

• Filter Structures

• Analog Filters

• FIR Filter Design

• IIR Filter Design

 3
FIR V.S. IIR
• LTI discrete-time system
– Difference equation in time domain
N M
y (n)   ak y (n  k )   bk x(n  k )
k 1 k 0

– Transfer function in z-domain

N M
Y ( z )    ak Y ( z ) z k
  bk X ( z ) z  k
k 1 k 0

Y ( z) k
b z k

H ( z)   k 0
N
1   ak z  k
X ( z)
k 1
 4
FIR V.S. IIR
• Finite impulse response (FIR)
– difference equation in the time domain
M
y (n)   bk x(n  k )
k 0

– Transfer function in the Z-domain

Y ( z) M
H ( z)    bk z  k
X ( z ) k 0
– Impulse response

h(n)  [b0 , b1 ,  , bM ]

• The impulse response is of finite length  finite impulse response

 5
FIR V.S. IIR
• Infinite impulse response (IIR)
– Difference equation in the time domain
N M
y (n)   ak y(n  k )   bk x(n  k )
k 1 k 0

– Transfer function in the z-domain

M

Y ( z) k
b z k

H ( z)   k 0
N
1   ak z  k
X ( z)
k 1

– Impulse response can be obtained through inverse-z transform, and it has

infinite length
 6
FIR V.S. IIR
• Example
– Find the impulse response of the following system. Is it a FIR or IIR
filter? Is it stable?
1
y ( n)  y (n  2)  x(n)
4
 7
FIR V.S. IIR
• Example
– Find the impulse response of the following system. Is it a FIR or IIR
filter? Is it stable?

y (n)  x(n)  2 x(n  2)  5 x(n  3)

 8
OUTLINE
• FIR and IIR Filters

• Filter Structures

• Analog Filters

• FIR Filter Design

• IIR Filter Design

 9
FILTER STRUCTURE: FIR
• A FIR filter can be implemented in different structures
– Direct-form structure
– Cascade-form structure
– Lattice structure
– Frequency sampling
– ……
 10
FILTER STRUCTURE: FIR
• FIR: Direct-form
– Also called tapped-delay-line filter, or transversal filter
M 1
y (n)   h(k ) x(n  k )
k 0

Order of the filter: M

 11
FILTER STRUCTURE: FIR
• FIR: Cascade-form
– Factor the transfer function into the product of second-order FIR systems

H ( z )  H1 ( z ) H 2 ( z )  H K ( z )
• Where
H k ( z )  bk 0  bk1 z 1  bk 2 z 2
 12
FILTER STRUCTURE: FIR
• Example
– Draw the direct-form and cascade-form structures of the following filter
H ( z )  (1  2 z 1  3z 2 )( 2  5z 2 )
 13
FILTER STRUCTURE: IIR
• An IIR filter can be implemented in various structures
– Direct-form I
– Direct-form II
– Transposed structure
– Cascade-form structure
– Parallel-form structure
– Lattice structure
– ……
 14
FILTER STRUCTURE: IIR
• IIR: Direct-form I M

k
b z k

H ( z)  k 0
N
 H1 ( z ) H 2 ( z )
1   ak z  k
k 1
M
1
H1 ( z )   bk z k
H 2 ( z)  N
k 0
1   ak z  k
k 1
 15
FILTER STRUCTURE: IIR
• IIR: Direct-form I

M
H1 ( z )   bk z k
k 0

1
H 2 ( z)  N
1   ak z  k
k 1
 16
FILTER STRUCTURE: IIR
• IIR: Direct-form II
– Put H 2 ( z ) before H1 ( z )

H ( z )  H 2 ( z ) H1 ( z )
1
• Difference equation of H 2 ( z)  N
1   ak z  k
k 1
N
w(n)   ak w(n  k )  x(n)
k 1
M
• Difference equation of H1 ( z )   bk z k
k 0
M
y (n)   bk w(n  k )
k 0
• Both involve the delay of w(n)
 17
FILTER STRUCTURE: IIR
• IIR: Direct-form II
 18
FILTER STRUCTURE: IIR
• IIR: Transposed structure
– Transposition theorem: For a given filter structure, if we apply the
following transposition: (1) reverse the direction of all signal flows; (2)
replace all the adders with intersections, and vice versa; and (3)
interchange the input and output, then the transposed system is the same
as the original system.
 19
FILTER STRUCTURE: IIR
• IIR: transposed structure of direct-form II
 20
FILTER STRUCTURE
• Example
– Draw the direct-form I, direct-form II, and transposed direct-form II of the
filter
1 1
y ( n )  y ( n  2)  x ( n )  x ( n  1)
4 3
 21
FILTER STRUCTURE
• Example
– Find the difference equation and transfer function of the filter
– Represent the filter in direct-form II
 22
OUTLINE
• FIR and IIR Filters

• Filter Structures

• Analog Filters

• FIR Filter Design

• IIR Filter Design

 23
ANALOG FILTERS
• Ideal filter
– Ideal analog filter

– Ideal digital filter

– In practice, we cannot implement filter characteristics with abrupt

transitions
• E.g. the impulse response of ideal low pass filter is sinc(t), which is
non-causal.
 24
ANALOG FILTERS
• Practical low pass filter
– Passband cutoff frequency:  p
– Stop band cutoff frequency:  s

1  1  H ( )  1  1 ,    p

H ( )   2 ,    s
 25
ANALOG FILTERS
• Butterworth filter
1
H ( ) 
2

1  2N
 26
ANALOG FILTERS
• The Chebyshev Filter
1
H ( ) 
2

1   2C N ( )

 cos( N cos 1  ), |  | 1
C N ( )  
cosh( N cosh  ), |  | 1
1
 27
OUTLINE
• FIR and IIR Filters

• Filter Structures

• Analog Filters

• FIR Filter Design

• IIR Filter Design

 28
FIR FILTER DESIGN
• Linear phase filter
– A filter is called a linear phase filter if its transfer function can be
expressed as
H () | H () | e  jA

• Where A is a constant
• The phase:    A , is a linear function of 
– Example
• Is the following filter a linear phase filter?
y (n)  x(n  n0 )
 29
FIR FILTER DESIGN
• Linear phase filter (Cont’d)
H () | H () | e jA

– A linear phase filter introduces the same amount of delay, A, to all the
frequencies (the same group delay)
• If the phase is not linear, then frequency components at different
frequencies will experience different delays  phase distortion
– A subset of FIR filters can have linear phase
• IIR filters do not have linear phase  IIR filter will introduce phase
distortion
• Not all FIR filters have linear phase
 30
FIR FILTER DESIGN
• Linear phase filter
– If a length-N FIR filter is symmetric with the middle of the filter, i.e.
h( n)  h( N  1  n)
• Then the filter has linear phase
– Example
h(n) = [1 2 3 3 2 1], h(n) = [1 2 3 4 3 2 1], h(n) = [2, 0, 0, 2]

– Proof
 31
FIR FILTER DESIGN
• FIR filter design with window
– 1. Find the desired frequency response (DTFT) H d ()
– 2. Perform Inverse DTFT, find the impulse response hd (n)
• The impulse response usually has infinite length
– 3. truncate hd (n) to finite length by multiplying it with a window function w(n)
• E.g. rectangular window 1, 0  n  N  1
w(n)  
0, otherwise

h(n)  hd n  ( N  1) / 2w(n)

• Different windows can be used to reduce the negative effects of truncation

– Rectangular window, Hamming window, Hanning window, Bartlett
window, Kaiser window, etc.
– Hamming window
 2n
0.54  0.46 cos , 0  n  N 1
w(n)   N 1
 0, otherwise
 32
FIR FILTER DESIGN
• Example
– Design a 9-point FIR filter to approximate a low pass filter with cut-off
frequency  c  0.2
 33
FIR FILTER DESIGN
• Example

% order of the filter

N = 9;

% cut-off frequency
Omega_c = 0.2*pi;

n = [-(N-1)/2:(N-1)/2];
% impulse response
hdn = Omega_c/pi*sinc(Omega_c/pi*n);

%wd = window(@rectwin, N).';

wd = window(@hamming, N).';

hn = hdn.*wd;
 34
OUTLINE
• FIR and IIR Filters

• Filter Structures

• Analog Filters

• FIR Filter Design

• IIR Filter Design

 35
IIR FILTER DESIGN
• IIR Filter design methods
– 1. impulse invariance
– 2. bilinear transformation
• Method 1: Impulse invariance
– Directly sample the impulse response of an analog filter
– If the impulse response and frequency response of the analog filter are
ha (t ) H a ( )

• Then the impulse response and DTFT of the IIR filter are
h(n)  ha (nT )

1    2n 
H ()   H a   
T n    T T 

– It will cause spectrum aliasing if the analog filter has a

bandwidth larger than 1
2T
 36
IIR FILTER DESIGN
• Method 1: Impulse invariance
– Analog filter:  , s    j ,
j
– Digital filter: , z  e
– Relationship between s and z

z  exp(Ts )
1
s log z
T
 37
IIR FILTER DESIGN
• Method 2: Bilinear transformation
– Use bilinear transformation to overcome aliasing
2 1  z 1 1  (T / 2) s
s z
T 1  z 1 1  (T / 2) s

2 
 tan
T 2

– Procedure:
• 1. From the digital filter specifications, find the corresponding analog
filter specifications by using the bilinear transformation
2 
 tan
T 2

• 2. Find the corresponding analog filter H a (s )

• 3. Find the equivalent digital filter

H ( z )  H a ( s) | 2 1 z 1
s
T 1 z 1
 38
IIR FILTER DESIGN
• Example
– If the transfer function of an analog filter is
1
H ( s) 
s 1
• Find the equivalent digital filter by using bilinear transformation.
Assume the sampling frequency is 1000 Hz
 39
IIR FILTER DESIGN
• Example
% analog Butterworth filter, in the form of zeros and poles
% cut-off frequency 1 rad/s
[z,p,k] = buttap(2); % creates a 2-pole Butterworth filter

% transfer function
[num,den] = zp2tf(z,p,k); % poles & zeros to transfer
function

% need to transform the cutoff frequency to 100 Hz (200

rad/sec)
wc = 200*pi;
[num,den] = lp2lp(num,den,wc);

% sampling frequency 1000 Hz

Fs = 1000;

% generate the digital filter with the bilinear

transformation
[numd,dend] = bilinear(num,den,Fs);

freqz(numd, dend)