Chapter 15

Filter Analysis and Design
Butterworth Filters
Butterworth filters have a transfer function whose squared
1
2
magnitude has the form H a ( j ) =
2n .
1 + ( / c )
M. J. Roberts - All Rights Reserved
Butterworth Filters
The poles of a lowpass Butterworth filter lie on a semicircle of radius c in the
open left half-plane. The number of poles is n and the angular spacing between
the poles is always / n. For n odd there is one pole on the negative real axis
and all the others occur in complex conjugate pairs. For even n all the poles occur
in complex conjugate pairs. The form of the transfer function is
n
n
1
1
pk
Ha ( s) =
=
=
(1 s / p1 )(1 s / p2 )(1 s / pn ) k=1 1 s / pk k=1 pk s
Frequency Transformation
A normalized lowpass filter can be transformed into an un-normalized
lowpass filter or into a highpass, bandpass or bandstop filter by
frequency transformation.
Lowpass to lowpass s s / c
Lowpass to highpass s c / s
s 2 + L H
Lowpass to bandpass s
s ( H L )
Lowpass to bandstop s
s ( H L )
s 2 + L H
c is the cutoff frequency in rad/s of a lowpass or highpass filter and

and are the lower and upper cutoff frequencies in rad/s of a
H
L
bandpass or bandstop filter.
*
Example

Design a first-order bandstop Butterworth filter with a cutoff frequencies
of 1 kHz and 2 kHz.
1
1
s 2 + L H
H norm ( s ) =
H BS ( s ) =
= 2
s ( H L )
s +1
s + s ( H L ) + L H
+1
2
s + L H
L = 2 1000 = 2000 , H = 4000
s 2 + 8 2 10 6
s 2 + 7.896 10 7
H BS ( s ) = 2
2
2
6
s + 2000 s + 8 10
s + 6283s + 7.896 10 7
Zeros at s = j8886 which correspond to = 8886 rad/s and f = 1414 Hz
Poles at s = 3142 j8886.
Example

Chebyshev, Elliptic and Bessel

Filters
The Butterworth filter is "maximally flat" in its passband. This means
that its frequency response in the passband is monotonic and the slope
approaches zero at the maximum response. The Chebyshev filter has
ripple in either its passband or stopband depending on which type of
Chebyshev filter it is. Ripple is a variation of the frequency response
between two limits. Allowing this ripple in the passband or stopband
lets the transition from pass to stop band be faster than for a Butterworth
filter of the same order. The Elliptic (or Cauer) filter has ripple in both
the passband and stopband and also has an even faster transition between
passband and stopband than the Chebyshev filters.
Chebyshev, Elliptic and Bessel

Filters
Impulse and Step Invariant

Design
Impulse Invariant Design

Let h ( t ) = h a ( t ) Ts ( t ) . Then
H ( j ) = fs
H ( j ( k ))
a
k=
L
where h a ( t )
Ha ( s ).
Also let h d [ n ] = h a ( nTs ) . Then, using /fs = = fs
( )
H d e j = fs
H ( jf ( 2 k ))
a
k=
10

In the expression
( )
H d e j = fs
H ( jf ( 2 k ))
k =
it is clear that the digital filter's frequency response consists of

multiple scaled aliases of the analog filter's frequency response
and, to the extent that the aliases overlap, the two frequency
responses must differ.
11

As an example of impulse invariant design let
A c2
Ha ( s) = 2
.
2
s + 2 c s + c
Then h a ( t ) = 2A c e c t / 2 sin ct / 2 u ( t ) . Sample fs times
per second to form h d [ n ] = 2A c e c nTs / 2 sin c nTs / 2 u [ n ].

Then
H d ( z ) = 2A c
ze c Ts / 2 sin cTs / 2
z 2 2e c Ts /
cos cTs / 2 z + e2 c Ts /
and
( )
H d e j = 2A c
*
e j e c Ts / 2 sin cTs / 2
e j 2 2e c Ts /
cos cTs / 2 e j + e2 c Ts /
12

The digital filter's frequency response is
( )=
Hd e
2A c
e j 2 2e c Ts /
( )
and, from a previous slide, H d e j = fs
H ( jf ( 2 k )).
k =
Therefore
A c2
( )= f
jf ( 2 k )
Hd e
k =
= 2A c
2
+
j
2
k
+
(
)
c s
c
e j 2 2e c Ts /
13

Let A = 10 and c =100 and fs =200. Then
( )
Hd e
= 2000
k =
j2 ( 2 k ) + j2 2 ( 2 k ) + 1
2
and
( ) = 1000
Hd e
e j e1/2 2 sin 1 / 2 2
e j 2 2e1/2
cos 1 / 2 2 e j + e1/
At = 0,
1
= 1958.5
2 2
k = 1 16 k j4 2 k
H d (1) = 2000
and
H d (1) =
*
343.825
= 1958.5
1 1.31751 + 0.49306
14

By design the digital filter's impulse response consists of
samples of the analog filter's impulse response.
15

The frequency response of the digital filter differs from the
frequency response of the analog filter due to aliasing.
16

Direct Form II Realization of the Digital Filter Designed by the
Impulse Invariant Technique
17

Let H a ( s ) =
A c
A c
A c
H a ( j2 f ) =
H a ( j ) =
.
s + c
j2 f + c
j + c
Then h a ( t ) = A c e c t u ( t ) . Sample at rate fs to form h d [ n ] = A c e c nTs u [ n ]

z
e j
j
and H d ( z ) = A c
H d e = A c j
c Ts
ze
e e c Ts
j
e
c
H d e j = fs
= A c j
e e c Ts
k = jf s ( 2 k ) + c
( )
( )
Let A = 10, c = 50 and fs = 100 and check equality at = 0.
fs
k =
A c
50000
=
= 1020.7
jfs ( 2 k ) + c k = j200 k + 50
e j
1
A c j
= 500
= 1270.7
c Ts
1/2
e e
1 e
*
18

Why are the two results
fs
k =
A c
50000
=
= 1020.7
jfs ( 2 k ) + c k = j200 k + 50
and
e j
1
A c j
=
500
= 1270.7
c Ts
1/2
e e
1 e
different by almost 25%? The difference comes from sampling h a ( t ) to form
h d [ n ] = A c e c nTs u [ n ]. h a ( t ) has a discontinuity at t = 0, a sample time. If we
( )
let the first sample be A c / 2 instead of A c the two expressions for H d e j
are now exactly equal. A c / 2 is the average of the two limits at the point of
discontinuity. This is consistent with Fourier transform theory which says that
the Fourier representation at a discontinuity is the average of the two limits.
*
19

9.87 10 4 s 2
Let H a ( s ) = 4
s + 444.3s 3 + 2.467 10 6 s 2 + 5.262 10 8 s + 1.403 1012
and let fs = 1000.
h a ( t ) = 246.07e122.41t cos(1199.4t 1.48) + 200.5e99.74t cos(977.27t + 1.683) u ( t )
h d [ n ] = 246.07e0.12241n cos(1.1994n 1.48) + 200.5e0.09974 n cos(0.97727n + 1.683) u [ n ]

48.4z 3 107.7z 2 + 51.46z
Hd ( z) = 4
z 1.655z 3 + 2.252z 2 1.319z + 0.6413
20
21
22
23
Step Invariant Design

In step invariant design the analog unit step response is sampled to
form the digital unit sequence response.
Ha ( s)
h 1a ( t ) = L
h 1d [ n ] = h 1,a ( nTs )
s
1
z
z 1
Z ( h 1d [ n ]) =
Hd ( z) Hd ( z) =
Z ( h 1d [ n ])
z 1
z
9.87 10 4 s 2
Let H a ( s ) = 4
s + 444.3s 3 + 2.467 10 6 s 2 + 5.262 10 8 s + 1.403 1012
and let fs = 1000.
24

0.2041e122.408t cos(1199.4t + 3.1312)
h 1a ( t ) =
u (t )
99.74t
cos(977.27t + 0.01042)
+0.2041e
0.2041( 0.8847 )n cos(1.1994n + 3.1312)
h 1d [ n ] =
u[n]
n
+0.2041( 0.9051) cos(0.97727n + 0.0102)
0.03443z 3 0.03905z 2 0.02527z + 0.02988
Hd ( z) =
z 4 1.655z 3 + 2.252z 2 1.319z + 0.6413
y [ n ] = 0.03443x [ n 1] 0.03905 x [ n 2 ] 0.02527 x [ n 2 ]
+ 0.02988 x [ n 4 ] +1.655y [ n 1] 2.252 y [ n 2 ]
+ 1.319 y [ n 3] 0.6413y [ n 4 ]
25
26
27
28
Finite-Difference Design
Finite-difference design is based on approximating continuous-time
1
derivatives by finite differences. The transfer function H a ( s ) =
s+a
d
implies the differential equation ( y a ( t )) + a y a ( t ) = x a ( t ) . The
dt
y d [ n + 1] y d [ n ]
d
derivative can be approximated by ( y a ( t ))
. Then
dt
Ts
y d [ n + 1] y d [ n ]
+ a y d [ n ] = x d [ n ] and the differential equation in
Ts
continuous time has become a difference equation in discrete time. Then
Yd ( z )
Ts
we can form the discrete-time transfer function H d ( z ) =
=
.
X d ( z ) z (1 aTs )
*
29
A derivative can be approximated by a forward difference
y d [ n + 1] y d [ n ] Z z 1
d
y a ( t ))
Yd ( z )
(
dt
Ts
Ts
a backward difference
y d [ n ] y d [ n 1] Z 1 z 1
d
z 1
y ( t ))
Yd ( z ) =
Yd ( z )
(
dt
Ts
Ts
zTs
or a central difference
y d [ n + 1] y d [ n 1] Z z z 1
d
z2 1
y a ( t ))
Yd ( z ) =
Yd ( z ) .
(
dt
2Ts
2Ts
2zTs
Since every multiplication by s represents a differentiation in time,
finite difference design can be done by simply replacing s with a
z-transform expression for a difference that approximates a derivative.
*
30
Using a forward difference,
1
1
1
Hd ( z) = Ha ( s) =
=
= Ts
1
z
1
z (1 aTs )
s + a s
+
a
Ts
Ts
Using a backward difference,
1
Ts
z
1
Hd ( z) = Ha ( s) =
=
=
1
1
z
1
1
+
aT
s + a s
s
+a
z
zTs
zTs
1 + aTs
Using a central difference,
1
z
1
Hd ( z) = Ha ( s) =
= 2
= 2Ts 2
2
z
1
z 1
z + 2aTs z 1
s + a s
+
a
2 zTs
2zTs
*
31
Using the finite-difference technique it is possible to approximate
a stable continuous-time filter by an unstable discrete-time filter.
1
In H a ( s ) =
the pole is at s = a. In the digital filter
s+a
1
H d ( z ) = Ts
the pole is at z = (1 aTs ) . If H a ( s ) is
z (1 aTs )
stable, a > 0 and 1 aTs is on the real axis of the z plane. If aTs 2
the digital filter is unstable. One way to stabilize the design is to
use a smaller value for Ts , which implies a higher sampling rate. This
design was based on using a forward difference.
32
If we use a backward difference instead of a forward difference
we get the digital filter
1
Hd ( z) = Ha ( s) =
s + a s z 1
zTs
Ts
z
.
1 + aTs z 1 / (1 + aTs )
The pole is now at z = 1 / (1 + aTs ) and

is guaranteed to be inside the unit circle
in the z plane, making this filter stable for
any a > 0 and any Ts . Using a backward
difference in any digital filter design
guarantees stability, but that guarantee comes
at a price. The pole locations are restricted to
a small region inside the unit circle.
*
33
9.87 10 4 s 2
Let H a ( s ) = 4
s + 444.3s 3 + 2.467 10 6 s 2 + 5.262 10 8 s + 1.403 1012
and design a digital filter using the finite-difference technique with
fs = 1000.
0.169z 2 ( z 1)
Hd ( z) = 4
z 1.848z 3 + 1.678z 2 0.7609z + 0.1712
2
34
35
36
Matched z and Direct

Substitution
The matched-z and direct substitution digital filter design techniques use the
correspondence z = e sTs to map the poles and zeros of the analog filter to
corresponding locations in the z plane. If the analog filter has a pole at s = a,
the corresonding digital filter pole would be at z = eaTs . The matched-z method
aTs
z
e
replaces every factor of the form s a with the factor 1 eaTs z 1 or
and
z
the direct substitution method replaces it with the factor z eaTs . Zeros are mapped
in the same way.
s=a
*
z = eaTs
37

Substitution
9.87 10 4 s 2
Let H a ( s ) = 4
.
3
6 2
8
12
s + 444.3s + 2.467 10 s + 5.262 10 s + 1.403 10
Then, using the matched-z technique
z ( z 1)
z 4 1.655z 3 + 2.252z 2 1.319z + 0.6413
y [ n ] = 98700 x [ n ] 197400 x [ n 1] + 98700 x [ n 2 ]
H d ( z ) = 98700
+1.655 y [ n 1] 2.252 y [ n 2 ] + 1.319 y [ n 3] 0.6413y [ n 4 ]
38

Substitution
39

Substitution
40

Substitution
41
The Bilinear z Method

One approach to digital filter design would be to use the mapping
z = e sTs s (1 / Ts ) ln ( z ) to convert the s-domain expression to the
z-domain expression
H d ( z ) = H a ( s ) s (1/T ) ln ( z ) .
s
Unfortunately this creates a transcendental function of z with infinitely

many poles. We can simplify the result by making an approximation to
the exponential function. The exponential function is defined by
2
3
k
x
x
x
ex = 1 + x +
+ + =
2! 3!
k = 0 k!
42
The Bilinear z Transform

We can approximate the exponential function by truncating its
series definition to two terms e x 1 + x. Then z = 1 + sTs or
z 1
s
. This is exactly the same as the finite difference technique
Ts
using forward differences and has the same drawback, the possibility
of converting a stable analog filter into an unstable digital filter. A
simple but very significant variation on this idea is to express e sTs in
e sTs /2
the form sTs /2 and then approximate both exponentials with the first
e
1 + sTs / 2
2 z 1
two terms of the series. Then z =
or s
. This is
1 sTs / 2
Ts z + 1
the basis of the bilinear digital filter design technique.
*
43

2 z 1
let s = j . Then
Ts z + 1
2 + sTs 2 + j Ts
j 2 tan 1 ( Ts /2 )
1
z=
=
= 1 tan ( Ts / 2 ) = e
2 sTs 2 j Ts
In the mapping s
The axis of the s plane maps into the unit circle of the
z plane. If s = + j , < 0, the mapping is from the left
half of the s plane to the interior of the unit circle in the z
plane. If s = + j , > 0, the mapping is from the right
half of the s plane to the exterior of the unit circle in the z
plane.
44

For real discrete-time frequencies the correspondence between
2 e j 1
2

the s and z planes is s =
= j tan . Then, since
j
2
Ts e + 1
Ts
s = + j , = 0 and = ( 2 / Ts ) tan ( / 2 ) or = 2 tan 1 ( Ts / 2 ) .
45
9.87 10 4 s 2
Let H a ( s ) = 4
.
3
6 2
8
12
s + 444.3s + 2.467 10 s + 5.262 10 s + 1.403 10
Then H d ( z ) = 12.38
( z + 1)2 ( z 1)2
z 4 1.989z 3 + 2.656z 2 1.675z + 0.711
46
47
48
49
FIR Filter Design

An analog filter can be approximated by
an FIR digital filter. The impulse response
of the analog filter has infinite duration in
time but beyond some time it has dropped
to a very low value and the remainder
beyond that time can be truncated with
little change in the filter characteristics.
We can create a digital FIR filter by
sampling over that time.
*
50
FIR Filter Design
We can also approximate ideal filters,

which are non-causal, by truncating that
part of the impulse response occurring
before time n = 0 and later when the
impulse response has fallen to a very low
value.
51
FIR Filter Design

The truncated impulse response
has the form
hN [n] =
N 1
a [n m]
m=0
where N is the number of samples

retained. This type of FIR digital
filter can be realized in Direct
Form II. Its transfer function is
Hd ( z) =
N 1
m
a
z
m .
m=0
52
FIR Filter Design

h d [ n ] , 0 n < N
Let h N [ n ] =
= h d [ n ] w [ n ] be the
, otherwise
0
impulse response of the digital filter where h d [ n ] is the
sampled ideal analog filter impulse response and w [ n ] is

a window function. Then the frequency response is
( )
( ) ( )
The periodic convolution with W ( e ) changes the
H N e j = H d e j W e j .
j
frequency response from ideal and introduces ripple

in the frequency response.
*
53
FIR Filter Design

Impulse response and frequency response of an FIR lowpass
digital filter designed by truncating the ideal impulse response.
54
FIR Filter Design

55
FIR Filter Design

56
FIR Filter Design

The digital filter impulse and frequency responses on the
previous three slides were all formed by using a rectangular
1 , 0 n < N
window of the form, w [ n ] =
. Other window
0 , otherwise
functions could be used to reduce the ripple in the frequency
response.
57
FIR Filter Design

Frequently-Used Window Functions
von Hann
Bartlett
Hamming
w[n] =
1
2 n
1
cos
, 0 n < N
2
N 1
N 1
2n
,
0
N 1
2
w[n] =
2 2n , N 1 n < N
N 1
2
2 n
w [ n ] = 0.54 0.46 cos
, 0n<N
N 1
2 n
4 n
Blackman w [ n ] = 0.42 0.5 cos
+
0.08
cos
, 0n<N
N 1
N 1
Kaiser
2
2
N 1
N 1
I0 a
n

2
2
w[n] =
N 1
I0 a
58
FIR Filter Design
59
FIR Filter Design
60
FIR Filter Design

In studying the common types of windows in the last two slides
it becomes apparent that in designing FIR filters two design
goals are in conflict. The ideal filter has no ripple in the passband
and the transition from the passband to the stopband is abrupt in
frequency; it has no width. For a fixed window width, if we
optimize the transition to be as quick as possible we must accept
large ripple in the pass and stop bands. If we want to minimize
passband and stopband ripple we must accept a slower transition
between the passband and stopband.
61
FIR Filter Design

The general form of an FIR impulse response is
h d [ n ] = h d [ 0 ] [ n ] + h d [1] [ n 1] + + h d [ N 1] n ( N 1)
and its z transform is
H d ( z ) = h d [ 0 ] + h d [1] z 1 + + h d [ N 1] z ( N 1)
Let N be an even number and let the coefficients be chosen such
that
h d [ 0 ] = h d [ N 1] , h d [1] = h d [ N 2 ] , , h d [ N / 2 1] = h d [ N / 2 ]
62
FIR Filter Design

The frequency response is
j ( N 1)
j ( N 2 )
j
h
0
+
h
0
e
+
h
1
e
+
h
1
e
+
[
]
[
]
[
]
[
]
d
d
d
d
j
Hd e =
j ( N /2 1)
jN /2
+ h d [ N / 2 1] e
+ h d [ N / 2 1] e
which can be expressed in the form
( )
( )
H d e j = 2e
N 1
j
N 1
N 3
+ h d [1] cos
+
h d [ 0 ] cos
2
2
+ h [ N / 2 1] cos ( )
The phase of the frequency response is linear. This is usually a desirable

result because it is the same as an ideal filter. By a similar process it can
be shown that if the coefficients are anti-symmetric meaning
h d [ 0 ] = h d [ N 1] , h d [1] = h d [ N 2 ] , , h d [ N / 2 1] = h d [ N / 2 ]
that the phase is also linear and this holds true for N even or odd.
*
63
Optimal FIR Filter Design

The Parks-McClellan algorithm is a technique for directly designing
a filter by specifying its frequency response by multiple linear
functions of frequency in multiple frequency bands separated by
transition bands. In MATLAB the syntax is B = firpm(N,F,A) where
is a vector of N + 1 real symmetric coefficients in the impulse response
of the FIR filter which is the best approximation to the desired frequency
response described by F and A. (See the MATLAB help file for a more
detailed description.)
64

Example
Design an optimal FIR filter to meet this specification.
65

Example
The band edges are = {0, 0.6, 0.7, 2.2, 2.3, } and the desired
amplitude responses at those band edges are A = {0, 0,1,1, 0, 0} .
The vector F should be F = / = {0, 0.191, 0.2228, 0.7003, 0.7321,1} .
N is the length of the filter and can be found by trying different
values until filter specification is met. In this case N = 70 meets

the specification.
66
67

Chapter 15

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Chapter 15

Uploaded by

Copyright:

Available Formats

Filter Analysis and Design

M. J. Roberts - All Rights Reserved

M. J. Roberts - All Rights Reserved

c is the cutoff frequency in rad/s of a lowpass or highpass filter and

M. J. Roberts - All Rights Reserved

M. J. Roberts - All Rights Reserved

M. J. Roberts - All Rights Reserved

Chebyshev, Elliptic and Bessel

M. J. Roberts - All Rights Reserved

Chebyshev, Elliptic and Bessel

M. J. Roberts - All Rights Reserved

Impulse and Step Invariant

M. J. Roberts - All Rights Reserved

Impulse Invariant Design

Also let h d [ n ] = h a ( nTs ) . Then, using /fs = = fs

M. J. Roberts - All Rights Reserved

Impulse Invariant Design

it is clear that the digital filter's frequency response consists of

M. J. Roberts - All Rights Reserved

Impulse Invariant Design

Then h a ( t ) = 2A c e c t / 2 sin ct / 2 u ( t ) . Sample fs times

per second to form h d [ n ] = 2A c e c nTs / 2 sin c nTs / 2 u [ n ].

M. J. Roberts - All Rights Reserved

Impulse Invariant Design

and, from a previous slide, H d e j = fs

M. J. Roberts - All Rights Reserved

Impulse Invariant Design

Impulse Invariant Design

M. J. Roberts - All Rights Reserved

Impulse Invariant Design

M. J. Roberts - All Rights Reserved

Impulse Invariant Design

M. J. Roberts - All Rights Reserved

Impulse Invariant Design

Then h a ( t ) = A c e c t u ( t ) . Sample at rate fs to form h d [ n ] = A c e c nTs u [ n ]

Let A = 10, c = 50 and fs = 100 and check equality at = 0.

M. J. Roberts - All Rights Reserved

Impulse Invariant Design

h d [ n ] = A c e c nTs u [ n ]. h a ( t ) has a discontinuity at t = 0, a sample time. If we

let the first sample be A c / 2 instead of A c the two expressions for H d e j

M. J. Roberts - All Rights Reserved

Impulse Invariant Design

h d [ n ] = 246.07e0.12241n cos(1.1994n 1.48) + 200.5e0.09974 n cos(0.97727n + 1.683) u [ n ]

M. J. Roberts - All Rights Reserved

Impulse Invariant Design

M. J. Roberts - All Rights Reserved

Impulse Invariant Design

M. J. Roberts - All Rights Reserved

Impulse Invariant Design

M. J. Roberts - All Rights Reserved

Step Invariant Design

M. J. Roberts - All Rights Reserved

Step Invariant Design

M. J. Roberts - All Rights Reserved

Step Invariant Design

M. J. Roberts - All Rights Reserved

Step Invariant Design

M. J. Roberts - All Rights Reserved

Step Invariant Design

M. J. Roberts - All Rights Reserved

M. J. Roberts - All Rights Reserved

M. J. Roberts - All Rights Reserved

M. J. Roberts - All Rights Reserved

M. J. Roberts - All Rights Reserved