Вы находитесь на странице: 1из 33

Digital IIR Filter Design

Rajesh M Hegde

Department of Electrical Engineering, IIT Kanpur


Lecture @ HBTI Kanpur
January 8, 2014

Lecture at HBTI Kanpur

1 / 33

Digital IIR Filter Design

Nyquist Sampling Theorem


I Let xc(t) be a band limited signal with
Xc(j) = 0; || N
I Then xc(t) can be uniquely determined by its samples x[n] = xc(nT ), n =
0, 1, 2, ..
if
s = 2
2N
T
Here N : Nyquist frequency
s : Nyquist rate

Lecture at HBTI Kanpur

2 / 33

Digital IIR Filter Design

Frequency domain illustration of sampling in the time domain

Figure (a) Spectrum of the original signal, (b) Spectrum of the sampling function, (c) Spectrum of the
sampled signal with s > 2N , (d) Spectrum of the sampled signal with s < 2N

Lecture at HBTI Kanpur

3 / 33

Digital IIR Filter Design

Figure illustrates exact recovery of a continuous-time signal from its samples using an ideal low pass
filter.
Lecture at HBTI Kanpur

4 / 33

Digital IIR Filter Design

Reconstruction of a Bandlimited Signal

Lecture at HBTI Kanpur

5 / 33

Digital IIR Filter Design

Reconstruction of a Bandlimited Signal from its Samples

? Given x[n], How do we get xc(t) ?


From x[n] form xs(t) as follows
xs(t) =

x[n](t nT )

nth sample associated with impulse at t = nT where T : sampling period.


Input xs(t) to an ideal LPF with frequency response Hr (j) and impulse
response hr (t)
X
xr (t) =
x[n]hr (t nT )
(1)
n

Lecture at HBTI Kanpur

6 / 33

Digital IIR Filter Design

Reconstruction of a Bandlimited Signal from its Samples


? Note : (a) Hr (j) has gain 0T 0, to compensate for factor 1/T in Equation
below
jT

X(e
(b) c =

s
2

Lecture at HBTI Kanpur

)=

1X
T

Xc(j( ks)

; Nyquist rate sampling and > 2N

7 / 33

Digital IIR Filter Design

Reconstruction of a Bandlimited Signal from its Samples


Corresponding hr (t)[I.F.T.of Hr (j)]
hr (t) =

sin( t
)
T

(2)

t
T

from equation (1) and (2)


xr (t) =

x[n]

)
sin[ (tnT
]
T

(3)

(tnT )
T

However one can not infer xr (t) = xc(t) ; looking at Equation (3) alone.
So consider hr (t) as in Equation 2
Here
hr (0) = 1;

(4)

and
(5)

hr (nT ) = 0;
for n = 1, 2, ...
Lecture at HBTI Kanpur

8 / 33

Digital IIR Filter Design

Reconstruction of a Bandlimited Signal from its Samples

From Equation (1), (2), and (5)


given that x[n] = xc(nT )
xr (mT ) = xc(mT )
where 0m0 is an integer.
I Signal reconstructed xr (t) in Equation (3) , is same as the original xr (t) at
the sampling instants.
I At non sampling instants band limited interpolation is used for reconstruction.
Consider xc(t), xs(t)
Generate several terms of the following expression

x[n]

Lecture at HBTI Kanpur

)
sin[ (tnT
]
T

(6)

(tnT )
T

9 / 33

Digital IIR Filter Design

Reconstruction of a Bandlimited Signal from its Samples

Ideal bandlimited interpolation

I Note that ideal LPF interpolates between impulses of xs(t) to reconstruct


xr (t)
Lecture at HBTI Kanpur

10 / 33

Digital IIR Filter Design

Filter Design Techniques


I IIR filter design
I FIR filter design
I Steps in filter design :
. Specification of desired frequency response.
. Approximation of specification using causal discrete time system.
. Realization of system.

Lecture at HBTI Kanpur

11 / 33

Digital IIR Filter Design

Impulse Invariance Method

I Overall system response





jT
H(e
) , || < T
Hef f (j) =
0
, || > T
I The relation between and is = T , for computing a digital filter from
a CT filter.
Effective H(ej ) = Hef f (j T ), || <

Lecture at HBTI Kanpur

12 / 33

Digital IIR Filter Design

Filter Specifications
I Specification of overall system
I Corresponding discrete-time filter specification, where 1 and 2 are ripple
parameters.
I p and p are analog and digital passband cut-off frequency.
I s and s are analog and digital stopband cut-off frequency.

Lecture at HBTI Kanpur

13 / 33

Digital IIR Filter Design

Lecture at HBTI Kanpur

14 / 33

Digital IIR Filter Design

IIR Filter: Concept of Impulse Invariance

Lecture at HBTI Kanpur

15 / 33

Digital IIR Filter Design

IIR Filter Design (Impulse Invariance Method)

Butterworth, Chebyshev(I,II), Elliptic Filter


I Digital filter specifications Analog counterpart digital filter
I H() hc(t) H(t)
I where Ha(s) is analog counterpart of hc(t).
IIM
I In the IIM method, continuous time filters are transformed to discrete time
filters as shown below
. = T
I Hence, the impulse response of the discrete time filter is given by
. h(n) = Td.hc(nTd)
I Here, Td is sampling interval and has no role in the design process.
Lecture at HBTI Kanpur

16 / 33

Digital IIR Filter Design

Impulse Invariance Method

I Frequency response of the discrete time filter is therefore given by


P
j
. H(e ) = k Hc(j Tw + j 2
k)
T
d

I If the C.T. filter is band limited,


. Hc(j) = 0 ; ||

Td

I then, H(ej ) = Hc(j T ) ; ||


d

I and (D.T. and C.T. frequency variables) are related as


. = Td for ||

Lecture at HBTI Kanpur

17 / 33

Digital IIR Filter Design

Impulse Invariance Method

I In Practice, we may not be able to get a Band limited filter T.F. resulting in
aliasing.

Illustration aliasing in impulse invariance design technique

Lecture at HBTI Kanpur

18 / 33

Digital IIR Filter Design

Impulse Invariance Method


Assuming that aliasing in the transformation of Hc(j) H(ej ) is negligible. Specifications of Hc(j) can be obtained as
=

(7)

Td

Step 1: Specification of H(ej ) Hc(j)


Next we proceed to the step 2 or conversion/transformation of Hc(s) to H(z).
Step 2:Transformation of Hc(s) H(z)
Let us start with a Hc(s) as
Hc(s) =

N
X

Ak

k=1

s sk

(8)

The corresponding impulse response is


 PN
Ak esk t; t 0
k=1
hc(t) =
0;
t<0
Lecture at HBTI Kanpur

19 / 33

Digital IIR Filter Design

Impulse Invariance Method


The impulse response of the D.T. filter is obtained as

h[n] = Tdhc(nTd) =
h[n] =

N
X

TdAk esk nTd u[n]

(9)

TdAk (esk Td )nu[n]

(10)

k=1
N
X
k=1

Computing H(z) from here, we have

H(z) =

N
X
k=1

Hc(s) =

Lecture at HBTI Kanpur

N
X

Ak

k=1

s sk

TdAk
1

H(z) =

N
X
k=1

20 / 33

(11)

esk Td z 1
TdAk
1

esk Td z 1
Digital IIR Filter Design

(12)

Impulse Invariance Method

Mapping from S
| plane
{z } Z
| plane
{z }
C.T. filter

D.T. filter

1. S-plane pole at S = Sk maps to Z-plane pole at


Z = eSk Td

(13)

2. Coefficient of partial fraction expansion of Hc(s) is/are same as H(z) except for Td.
3. If C.T. filter is stable then real part of Sk < 0, then eSk Td will be less than
unity, and corresponding pole in the D.T. filter is stable.
4. Mapping from S Z plane

Lecture at HBTI Kanpur

Zk = eSk Td

21 / 33

(14)

Digital IIR Filter Design

Impulse Invariance Method

Lecture at HBTI Kanpur

22 / 33

Digital IIR Filter Design

Impulse Invariance Method

Lecture at HBTI Kanpur

23 / 33

Digital IIR Filter Design

Impulse Invariance Method

Hence, mapping needs non-linear mappings


Solution: BLT

Lecture at HBTI Kanpur

24 / 33

Digital IIR Filter Design

Bilinear Transformation

Lecture at HBTI Kanpur

25 / 33

Digital IIR Filter Design

Mapping from S-plane to Z-plane

I A transformation that satisfies this mapping constraint is


w = c tan(

w1 T

or c = w cot(

(15)

2
w1 T

2
Here = for analog filter maps to w1 = 21 ws for digital filter.

Lecture at HBTI Kanpur

26 / 33

Digital IIR Filter Design

(16)

Selection of c

I Such that w = w1 correspondence can be established at any desired frequency. Put w = w1 = wr and find c.
c = wr cot(
For small wr ,

wr T
2

wr T
2

(17)

<< 1
c = wr (

2
wr T

)=

(18)

If c = 1/T then = T1 tan 12T

Lecture at HBTI Kanpur

27 / 33

Digital IIR Filter Design

Diagram illustrating warping

Lecture at HBTI Kanpur

28 / 33

Digital IIR Filter Design

Relationship between s and s1

tan x =

sin x
cos x

jx

= j

ejx

jx

e
+

!
(19)

ejx

Let S = jw and S1 = jw1 in


w = c tan

js = jc

S = c

Lecture at HBTI Kanpur

w1 T

S1 T
2

S1 T
2

S1 T
2

S1 T
2

29 / 33

(20)

2
e
+e

e
+e

S1 T
2
S1 T
2

S1 T
2
S1 T
2

(21)

(22)

Digital IIR Filter Design

Relationship between s and z: BLT

I let z = eS1T
S1 T

s=c

1e
1+

eS1T


s = c

Lecture at HBTI Kanpur

30 / 33


=c

1z
1+

1z

1+
1 

1 

(23)

z 1

(24)

z 1

Digital IIR Filter Design

Structure for IIR filters

I.I.M. Design of a analog filter : Ha(s) =

Lecture at HBTI Kanpur

31 / 33

Pm

Ki
i=1 S+Si

Digital IIR Filter Design

Time domain invariance of Analog and Digital IIR filters


Input:
t-domain: xc(t)
s-domain: Xc(s)

Sample

Output:
t-domain: y(t) = xc(t) * ha(t)
s-domain:Y(s) = Xc(s)H(s)

Time-domain
Invariant
Equivalents

Input:
t-domain: xc(nT)
s-domain: X(z)

Sample

Output:
t-domain: y(nT) = xc(nT) * h(nT)
z-domain: Y(z) = X(z)H(z)

Time domain invariance

Lecture at HBTI Kanpur

32 / 33

Digital IIR Filter Design

Thank You
Dept. of EE, IIT Kanpur
rhegde@iitk.ac.in
http://www.iitk.ac.in/

Вам также может понравиться