You are on page 1of 17

IIR Filters

Impulse Invariant and Bilinear Transform


Methods of Design

© 2019 Arm Limited


Recursive Filters
Unlike non-recursive FIR filters, Infinite Impulse Response (IIR) filters use previous output samples as well as
previous input samples in order to calculate their output.

x(n) + + y(n)
b0
+ +
FIR filter IIR filter
z -1 z -1
M N

b x ( n  k )   al y ( n  l )
+ +
x(n-1) b1 -a1 y(n-1) y ( n)  k
+ + k 0 l 1

z -1 z -1
+ +
x(n-2) b2 -a2 y(n-2)
IIR filterFIR filter
+ +

Y ( z) b0  b1 z 2    bM z  M
H ( z)  
X ( z) 1  a1 z 1    a N z  N
x(n-N) bM -aN y(n-N)

FIR filter IIR filter


2 © 2019 Arm Limited
Impulse Response

FIR filter IIR filter


stable potentially unstable
impulse response equal to filter impulse response computed recursively
coefficients from filter coefficients

3 © 2019 Arm Limited


Stable Impulse Response Example
y ( n)  0.5 x ( n  1)  0.7 y ( n  1)  0.8 y ( n  2)

Recursive calculation of impulse response


n x(n) x(n-1) y(n) y(n-1) y(n-2)
0 1 0 0 0 0
1 0 1 0.5 0 0
2 0 0 0.35 0.5 0
3 0 0 -0.155 0.35 0.5
4 0 0 -0.3885 -0.155 0.35
5 0 0 -0.148 -0.3885 -0.155
6 0 0 0.207 -0.148 -0.3885

4 © 2019 Arm Limited


Stable Impulse Response Example
y ( n)  0.5 x ( n  1)  0.7 y ( n  1)  0.8 y ( n  2)

y(n)
0.5

1 2 6 7 n

Exponentially decaying impulse response

5 © 2019 Arm Limited


Finite Impulse Response Example
y ( n)  0.5 x ( n)  y ( n  1)  0.5 x ( n  2)

Recursive calculation of impulse response


n x(n) x(n-1) x(n-2) y(n) y(n-1)
0 1 0 0 0.5 0
1 0 1 0 0.5 0.5
2 0 0 1 0 0.5
3 0 0 0 0 0
4 0 0 0 0 0
5 0 0 0 0 0
6 0 0 0 0 0

6 © 2019 Arm Limited


Finite Impulse Response Example
y ( n)  0.5 x ( n)  y ( n  1)  0.5 x ( n  2)

y(n)
0.5

1 2 6 7 n

Finite impulse response computed recursively

7 © 2019 Arm Limited


Unstable Impulse Response Example
y ( n)  0.5 x ( n  1)  0.5 y ( n  1)  2.0 y ( n  2)

Recursive calculation of impulse response


n x(n) x(n-1) y(n) y(n-1) y(n-2)
0 1 0 0 0 0
1 0 1 0.5 0 0
2 0 0 0.25 0.5 0
3 0 0 -0.875 0.25 0.5
4 0 0 -0.9375 -0.875 0.25
5 0 0 1.2813 -0.9375 -0.875
6 0 0 2.5156 1.2813 -0.9375
7 0 0 -1.3 2.5156 1.2813

8 © 2019 Arm Limited


Unstable Impulse Response Example
y ( n)  0.5 x ( n  1)  0.5 y ( n  1)  2.0 y ( n  2)

y(n)
2.0

1 2 5 6 n

Unstable impulse response computed recursively

9 © 2019 Arm Limited


Design of IIR Filters
• Traditionally, IIR filters are based on analogue filter designs described by continuous-
time s-transfer functions H(s).
• A vast amount of knowledge about analogue filter design exists.
• Once an analogue filter prototype has been chosen, the design of an IIR filter comprises
transforming to the discrete-time or z-domain.
• More recently, iterative computer-based techniques for finding the coefficients of an IIR
filter that will satisfy a particular specification have taken over.

10 © 2019 Arm Limited


Impulse Invariance Method
• Based on mapping of s-plane poles into the z-plane.
• Partial fraction expansion of H(s) separates it into terms corresponding to poles.
 pT 1
• Replace each denominator term ( s  p ) in H(s) with a term (1  e s z ) in H(z).
• Alternatively, use Laplace and z-transform tables.
• Discrete impulse response of resulting IIR filter comprises samples of the continuous
impulse response of the analogue prototype (scaled by Ts).
• Frequency response of IIR filter may not match that of analogue prototype.

11 © 2019 Arm Limited


Aliasing in the Impulse Invariant Design Method
Time-domain Frequency-domain

Continuous
impulse response
fs
h(t) -fs

Multiply Convolve

Discrete
impulse response
h(n)

12 © 2019 Arm Limited


Impulse Invariance Method
• Simple concept
• Lots of algebra
• Aliasing problems

13 © 2019 Arm Limited


Bilinear Transform Method
• Substitute for s in H(s)

H ( z )  H ( s) s  2 ( z 1)
Ts ( z 1)

• Effectively, the s-plane is mapped into the z-plane such that the imaginary axis in the s-
plane is mapped onto the unit circle in the z-plane and the left hand half of the s-plane
is mapped inside the unit circle in the z-plane.

14 © 2019 Arm Limited


Bilinear Transform Method
0.5

Normalized digital frequency (wd/ws)


0.4

0.3

0.2

0.1

0.0
0 0.5 1 1.5 2

Normalized analog frequency


(wa/ws)
  DT 
 A  K tan  
 2 
15 © 2019 Arm Limited
50
-5

-25
-35
-15
Bilinear Transform Method

Bilinear transformation of analogue


filter H(s) to digital filter H(z)

4000
digital frequency (Hz)

3000

2000

1000

0
0 1000 2000 3000 4000 5000 6000 7000 8000

magnitude (dB)
5
magnitude (dB)

-5

-15

-25

-35
0 1000 2000 3000 4000 5000 6000 7000 8000
analogue frequency (Hz)
16 © 2019 Arm Limited
Bilinear Transform Method
• Simple algebra
• No aliasing problems
• Warped frequency response

17 © 2019 Arm Limited

You might also like