Signals and Systems

Lecture 9: Infinite Impulse Response Filters

Dr. Guillaume Ducard

Fall 2018

based on materials from: Prof. Dr. Raffaello D’Andrea

Institute for Dynamic Systems and Control

ETH Zurich, Switzerland

G. Ducard 1 / 41
Outline

1 Infinite Impulse Response Filters

IIR filter : difference equation
Transfer function

2 First-Order Low-Pass Filter

Definition
Properties
Design considerations
Connection to CT systems

3 IIR Filter Design

Methodology
CT Butterworth filter design
Bilinear transform
G. Ducard 2 / 41
 Infinite Impulse Response Filters
IIR filter : difference equation
First-Order Low-Pass Filter
Transfer function
IIR Filter Design

Outline

1 Infinite Impulse Response Filters

IIR filter : difference equation
Transfer function

2 First-Order Low-Pass Filter

Definition
Properties
Design considerations
Connection to CT systems

3 IIR Filter Design

Methodology
CT Butterworth filter design
Bilinear transform
G. Ducard 3 / 41
 Infinite Impulse Response Filters
IIR filter : difference equation
First-Order Low-Pass Filter
Transfer function
IIR Filter Design

IIR: Difference equation

The class of causal infinite impulse response (IIR) filters can be

captured by the difference equation
M
X −1 N
X −1
y[n] = bk u[n − k] − ak y[n − k],
k=0 k=1

Characteristics :
M input coefficients bk ∈ R,
N − 1 output coefficients ak ∈ R.
filter order: is given by max(M − 1, N − 1) and corresponds
to the number of delay elements an implementation of the
filter would require;
it is also the size of the state in a state-space description
of the system. G. Ducard 4 / 41
 Infinite Impulse Response Filters
IIR filter : difference equation
First-Order Low-Pass Filter
Transfer function
IIR Filter Design

FIR vs. IIR

Key differences :
1 the output of a causal IIR filter is dependent on both the filter’s input
and on previous outputs (if one or more coefficients ak are non-zero).
2 Dependence on previous output(s) generally implies that the impulse
response has infinite length (hence the name: IIR filter).
3 IIR filters are not necessarily stable: the stability depends on the
coefficients ak .

Advantages of IIR filters :

1 they usually meet filter specifications with a lower filter order,
2 this corresponds to lower computation and storage cost compared to FIR
filters.

G. Ducard 5 / 41
 Infinite Impulse Response Filters
IIR filter : difference equation
First-Order Low-Pass Filter
Transfer function
IIR Filter Design

Outline

1 Infinite Impulse Response Filters

IIR filter : difference equation
Transfer function

2 First-Order Low-Pass Filter

Definition
Properties
Design considerations
Connection to CT systems

3 IIR Filter Design

Methodology
CT Butterworth filter design
Bilinear transform
G. Ducard 6 / 41
 Infinite Impulse Response Filters
IIR filter : difference equation
First-Order Low-Pass Filter
Transfer function
IIR Filter Design

IIR filter : transfer function

Transfer function and frequency response calculated from difference equation:
MP−1 MP−1
bk z −k bk e −jΩk
jΩ
k=0 z=e k=0
H(z) = N−1
−−−−→ H(Ω) = N−1
P P
1+ ak z −k 1+ ak e−jΩk
k=1 k=1

The goal of IIR filter design : find coefficients ak and bk such that the filter
meets given specifications and is stable.
IIR filter design :
often employs established continuous-time (CT) filter design methods, for
example Butterworth filter design,
and then transforms the resulting CT filter into DT.

In this lecture, we introduce:

1 the concepts underlying IIR filters;

2 how to design a CT Butterworth filter; and finally,

3 how to convert a CT filter into DT using the bilinear transform.

G. Ducard 7 / 41
 Definition
Infinite Impulse Response Filters
Properties
First-Order Low-Pass Filter
Design considerations
IIR Filter Design
Connection to CT systems

Outline

1 Infinite Impulse Response Filters

IIR filter : difference equation
Transfer function

2 First-Order Low-Pass Filter

Definition
Properties
Design considerations
Connection to CT systems

3 IIR Filter Design

Methodology
CT Butterworth filter design
Bilinear transform
G. Ducard 8 / 41
 Definition
Infinite Impulse Response Filters
Properties
First-Order Low-Pass Filter
Design considerations
IIR Filter Design
Connection to CT systems

IIR 1st order low-pass filter

Consider the causal, first-order, low-pass IIR filter, which has the
difference equation

y[n] = α y[n − 1] + (1 − α) u[n],

where 0 ≤ α < 1.

Intuition :
For α 6= 0, this is an infinite impulse response filter.
If α = 0 the output is equal to the input and no filtering
occurs.
As α → 1, the output becomes increasingly constant.

G. Ducard 9 / 41
 Definition
Infinite Impulse Response Filters
Properties
First-Order Low-Pass Filter
Design considerations
IIR Filter Design
Connection to CT systems

IIR 1st order low-pass filter

Transfer function
1−α
H(z) = .
1 − α z −1

Stability discussion:
The filter has a single pole at z = α.
It immediately follows that the filter is stable if 0 ≤ α < 1.

Frequency response
1−α
H (Ω) = .
1 − α e−jΩ

G. Ducard 10 / 41
 Definition
Infinite Impulse Response Filters
Properties
First-Order Low-Pass Filter
Design considerations
IIR Filter Design
Connection to CT systems

Outline

1 Infinite Impulse Response Filters

IIR filter : difference equation
Transfer function

2 First-Order Low-Pass Filter

Definition
Properties
Design considerations
Connection to CT systems

3 IIR Filter Design

Methodology
CT Butterworth filter design
Bilinear transform
G. Ducard 11 / 41
 Definition
Infinite Impulse Response Filters
Properties
First-Order Low-Pass Filter
Design considerations
IIR Filter Design
Connection to CT systems

IIR 1st order low-pass filter : behavior

Low - frequency signals remain unaltered since
1−α
H(Ω = 0) = = 1.
1 − αe−j0
The magnitude response is:
1−α
|H (Ω)| = q .
(1 − α cos Ω)2 + α2 sin2 Ω
Furthermore, one can show that the magnitude is monotonically non-increasing:
d|H(Ω)|
≤ 0, for 0 ≤ Ω ≤ π.
dΩ
The phase is
Always negative
 z }| { 
−α sin Ω
∠H(Ω) = arctan , for 0 ≤ Ω ≤ π.
1 − α cos Ω
| {z }
Always positive

Therefore
π
− < ∠H(Ω) ≤ 0.
2 G. Ducard 12 / 41
 Definition
Infinite Impulse Response Filters
Properties
First-Order Low-Pass Filter
Design considerations
IIR Filter Design
Connection to CT systems

A plot of the magnitude and phase response follows:

1 α = 0.3
α = 0.5
α = 0.8
|H(Ω)|

α = 0.9

0.5

Ω
0
0 π/2 π

0
∠H(Ω)

Ω
−π/2
0 π/2 π
G. Ducard 13 / 41
 Definition
Infinite Impulse Response Filters
Properties
First-Order Low-Pass Filter
Design considerations
IIR Filter Design
Connection to CT systems

Outline

1 Infinite Impulse Response Filters

IIR filter : difference equation
Transfer function

2 First-Order Low-Pass Filter

Definition
Properties
Design considerations
Connection to CT systems

3 IIR Filter Design

Methodology
CT Butterworth filter design
Bilinear transform
G. Ducard 14 / 41
 Definition
Infinite Impulse Response Filters
Properties
First-Order Low-Pass Filter
Design considerations
IIR Filter Design
Connection to CT systems

Choice of parameter α
To choose n , let :
Ts be the sampling time,
and T0 be the desired time for the continuous process to decay to e−1 ,
i.e., T0 = n Ts .
T
T0 1 − s
n= ⇒ α = e− n = e T0 ,
Ts
we assume that T0 is an integer multiple of Ts (if this is not the case, n may be rounded).

Example :

sampling time Ts = 0.01 s

chosen decay time T0 = 1 s,

α = exp(−0.01) ≈ 0.99.

Usually, T0 is large relative to Ts and we may use a first-order approximation

to obtain
Ts
α≈1− .
T0 G. Ducard 15 / 41
 Definition
Infinite Impulse Response Filters
Properties
First-Order Low-Pass Filter
Design considerations
IIR Filter Design
Connection to CT systems

Outline

1 Infinite Impulse Response Filters

IIR filter : difference equation
Transfer function

2 First-Order Low-Pass Filter

Definition
Properties
Design considerations
Connection to CT systems

3 IIR Filter Design

Methodology
CT Butterworth filter design
Bilinear transform
G. Ducard 16 / 41
 Definition
Infinite Impulse Response Filters
Properties
First-Order Low-Pass Filter
Design considerations
IIR Filter Design
Connection to CT systems

A connection to CT systems can be made: In CT, the transfer function of a

first-order low-pass filter is given by
1
H(s) = .
τs + 1
The differential equation for the output is
1
ẏ(t) = − (y(t) − u(t)) . (1)
τ
Assuming u(t) = 0 for t ≥ 0, we obtain the time-domain system response
t
y(t) = y(0)e− τ .
Choosing y(0) = 1, we obtain the system response:

e−1
t
τ

Therefore, the time to reach e−1 is τ . G. Ducard 17 / 41

 Definition
Infinite Impulse Response Filters
Properties
First-Order Low-Pass Filter
Design considerations
IIR Filter Design
Connection to CT systems

Another connection to CT can be made by discretizing

1
ẏ(t) = − (y(t) − u(t))
τ
assuming the input u is constant over a sample period Ts , i.e. a zero-order
hold device is used:
   1 1 
ẏ −τ τ y
= 0 ≤ t < Ts ,
u̇ 0 0 u
which we solve using the matrix exponential and obtain
   Ts Ts    " Ts T
# 
y(Ts− ) −τ y(0) − τ − τs y(0)
= exp τ = e 1 − e .
−
u(Ts ) 0 0 u(0) 0 1 u(0)

As discussed in Lecture 1, this solution is valid on any time interval because the
system is time invariant. Substituting the decay time T0 for the time constant
τ , the resulting difference equation becomes
T T
− Ts − Ts
y[n] = e 0 y[n − 1] + (1 − e 0 )u[n − 1] = αy[n − 1] + (1 − α)u[n − 1]
which closely resembles the first-order, low-pass IIR filter, except that the input
is delayed by one sample. G. Ducard 18 / 41
 Infinite Impulse Response Filters Methodology
First-Order Low-Pass Filter CT Butterworth filter design
IIR Filter Design Bilinear transform

Outline

1 Infinite Impulse Response Filters

IIR filter : difference equation
Transfer function

2 First-Order Low-Pass Filter

Definition
Properties
Design considerations
Connection to CT systems

3 IIR Filter Design

Methodology
CT Butterworth filter design
Bilinear transform
G. Ducard 19 / 41
 Infinite Impulse Response Filters Methodology
First-Order Low-Pass Filter CT Butterworth filter design
IIR Filter Design Bilinear transform

IIR filter design

FIR filter design only makes sense in DT.

In contrast,
1 IIR filters are often designed in CT using an established
method, for example
a Butterworth,
or Chebyshev method.
2 they are then converted to DT using the bilinear transform
(sometimes also called Tustin method).

We will see in the following section,

the bilinear transform has properties that make it a useful tool
for the above design procedure.

G. Ducard 20 / 41
 Infinite Impulse Response Filters Methodology
First-Order Low-Pass Filter CT Butterworth filter design
IIR Filter Design Bilinear transform

Outline

1 Infinite Impulse Response Filters

IIR filter : difference equation
Transfer function

2 First-Order Low-Pass Filter

Definition
Properties
Design considerations
Connection to CT systems

3 IIR Filter Design

Methodology
CT Butterworth filter design
Bilinear transform
G. Ducard 21 / 41
 Infinite Impulse Response Filters Methodology
First-Order Low-Pass Filter CT Butterworth filter design
IIR Filter Design Bilinear transform

CT design: Butterworth filter

It is a general purpose low-pass filter. The starting point is the desired
frequency response, with corner frequency 1 rad/sec:
1
R(ω) = √ ,
1 + ω 2K
where K is the order of the filter.
0

−40
|R(ω)| (dB)

−80

−120 K=2
K=3
K=8
−160
10−2 10−1 100 101 102
ω
G. Ducard 22 / 41
 Infinite Impulse Response Filters Methodology
First-Order Low-Pass Filter CT Butterworth filter design
IIR Filter Design Bilinear transform

Butterworth filter design: properties

This frequency response has two very desirable properties:
1 it has no ripples,
2 and is maximally flat.

1 1
R(ω) = √ = (1 + ω 2K )− 2 ,
1+ω 2K

First, let us calculate the derivative dR/dω.

1 3
dR(ω)/dω = − (1 + ω 2K )− 2 2Kω 2K−1
2
= −KR3 ω 2K−1 ≤ 0 for all ω ≥ 0.

Conclusion
This means that the Butterworth filter has no ripples.
In other words, all derivatives of R up to 2K − 1 are 0 at 0 and the filter is said
to be maximally flat.
G. Ducard 23 / 41
 Infinite Impulse Response Filters Methodology
First-Order Low-Pass Filter CT Butterworth filter design
IIR Filter Design Bilinear transform

Transfer function
Let H(s) be the transfer function of a filter with frequency response R(ω).
We then have that

|H(jω)|2 = R(ω)2 = (1 + ω 2K )−1 .

The only stable transfer function that achieves this is

1
H(s) = K
,
Q
(s − sk )
k=1

j(2k+K−1)π
where sk = e 2K , k = 1, . . . , K.

Filters with a transfer function of this structure are known as Butterworth

filters.

G. Ducard 24 / 41
 Infinite Impulse Response Filters Methodology
First-Order Low-Pass Filter CT Butterworth filter design
IIR Filter Design Bilinear transform

Second-order Butterworth low-pass filter

Consider one of the most common Butterworth filters: a

second-order (K = 2) low-pass. We have,

−1 + j
s1 = ej3π/4 = √ , (135◦ )
2
−1 − j
s2 = ej5π/4 = √ , (225◦ )
2
which results in the transfer function
1 1
H(s) = 1−j 1+j
= √ .
(s + √ )(s + √ ) s2 + 2s + 1
2 2

G. Ducard 25 / 41
 Infinite Impulse Response Filters Methodology
First-Order Low-Pass Filter CT Butterworth filter design
IIR Filter Design Bilinear transform

Second-order Butterworth low-pass filter

The poles of the CT filter lie on the unit circle in the s-plane (not to be
confused with the z-plane) and are represented by the black crosses below.
Remark: The gray crosses represent the poles of H(−s) and are useful to visualize the pole-placement pattern.

Im(s)
K =2

135◦

Re(s)

−135◦

G. Ducard 26 / 41
 Infinite Impulse Response Filters Methodology
First-Order Low-Pass Filter CT Butterworth filter design
IIR Filter Design Bilinear transform

Third-order Butterworth low-pass filter

If, instead, a third-order low-pass filter is chosen (K = 3), the pole plot looks
as follows:

Im(s)
K=3

120◦

Re(s)

−120◦

G. Ducard 27 / 41
 Infinite Impulse Response Filters Methodology
First-Order Low-Pass Filter CT Butterworth filter design
IIR Filter Design Bilinear transform

Corner frequency specification

The design method introduced above assumed a corner frequency

of 1 rad/sec. However, other corner frequencies ωc can be chosen.

In that case, we proceed with the following change of

variable
s
s→ .
ωc

For example, the second order filter becomes

ωc 2
H(s) = √ ,
s2 + 2ωc s + ωc 2
which has the same response
√ as a mass-spring-damper system with
sub-critical damping 1/ 2 and natural frequency ωc .
G. Ducard 28 / 41
 Infinite Impulse Response Filters Methodology
First-Order Low-Pass Filter CT Butterworth filter design
IIR Filter Design Bilinear transform

Outline

1 Infinite Impulse Response Filters

IIR filter : difference equation
Transfer function

2 First-Order Low-Pass Filter

Definition
Properties
Design considerations
Connection to CT systems

3 IIR Filter Design

Methodology
CT Butterworth filter design
Bilinear transform
G. Ducard 29 / 41
 Infinite Impulse Response Filters Methodology
First-Order Low-Pass Filter CT Butterworth filter design
IIR Filter Design Bilinear transform

Bilinear transform or Tustin’s method

Once a CT filter has been designed, ⇒ the bilinear transform (also

known as Tustin’s method) can be used to convert it into a DT
filter.
The bilinear transform uses the substitution that
 
2 z−1
s= ,
Ts z + 1
where Ts is the sampling time.

G. Ducard 30 / 41
 Infinite Impulse Response Filters Methodology
First-Order Low-Pass Filter CT Butterworth filter design
IIR Filter Design Bilinear transform

Bilinear transform or Tustin’s method

Recall that z can be given the interpretation of a DT shift operator. If

Y (z) = zU (z),

then
y[n] = u[n + 1].
sTs
Similarly, in CT, e can be given the interpretation of a time shift operator. If

Y (s) = esTs U (s)

where Y (s) and U (s) are the Laplace transform of y(t) and u(t), respectively,
then
y(t) = u(t + Ts ).
Therefore, the two operators are equivalent:

z = esTs .

G. Ducard 31 / 41
 Infinite Impulse Response Filters Methodology
First-Order Low-Pass Filter CT Butterworth filter design
IIR Filter Design Bilinear transform

Bilinear transform or Tustin’s method

A rational approximation for the relation between z and s will map a rational
CT transfer function to a rational DT transfer function. This is equivalent to
converting differential equations to difference equations. We therefore use the
approximation
Ts
sTs es 2 1 + s T2s
e = ≈ ,
e−s 2
Ts
1 − s T2s
which is valid if sTs is small. We call
1 + s T2s
z=
1 − s T2s

the bilinear transform. The inverse is

 
2 z−1
s=
Ts z+1
and is straightforward to verify by substitution.

G. Ducard 32 / 41
 Infinite Impulse Response Filters Methodology
First-Order Low-Pass Filter CT Butterworth filter design
IIR Filter Design Bilinear transform

DC- CT mapping
We now evaluate the bilinear transform along the imaginary axis of the s-plane;
that is, let s = jω. Using the bilinear transform, this point maps to

1 + jω T2s
z= .
1 − jω T2s

Note that  
 1 + jω Ts 
 2 
|z| =   = 1.
 1 − jω T2s 

Conclusion
The bilinear transform therefore maps the imaginary axis of the s-plane to the
unit circle in the z-plane.

G. Ducard 33 / 41
 Infinite Impulse Response Filters Methodology
First-Order Low-Pass Filter CT Butterworth filter design
IIR Filter Design Bilinear transform

DC- CT mapping
We therefore write
1 + jω T2s
z = ejΩ = ,
1 − jω T2s
and calculate the mapping of a CT frequency ω to a DT frequency Ω as:
∠ejΩ = ∠(1 + jω T2s ) − ∠(1 − jω T2s )
Ω = arctan(ω T2s ) − arctan(−ω T2s )
= 2 arctan(ω T2s ).

Conclusion
The frequency response of the CT system at ω (the CT transfer function
evaluated on the imaginary axis at s = jω)
directly corresponds to
the frequency response of the resulting DT system at Ω = 2 arctan(ω T2s ) (the
DT transfer function evaluated on the unit circle at z = ejΩ ).

This is a desirable property, as we will later see. G. Ducard 34 / 41

 Infinite Impulse Response Filters Methodology
First-Order Low-Pass Filter CT Butterworth filter design
IIR Filter Design Bilinear transform

DC- CT mapping
For small ωTs , the DT frequency is approximately Ω ≈ 2(ω T2s ) = ωTs . This is
also evident when we plot the mapping of CT frequencies to DT frequencies for
z = esTs and the bilinear transform:

Ω Ω = ωTs
π Ω = 2 arctan(ωTs /2)

2
π/2

ω
0
0 2/Ts π/Ts

Note how Ω asymptotically converges to π as ω → ∞: The bilinear transform

compresses the imaginary axis of the s-plane onto the unit circle. G. Ducard 35 / 41
 Infinite Impulse Response Filters Methodology
First-Order Low-Pass Filter CT Butterworth filter design
IIR Filter Design Bilinear transform

Frequency warping
The underlying nonlinear relation between ω and Ω is called frequency warping.
A few common values are:

ω=0 ⇒ Ω=0
ω=∞ ⇒ Ω=π
2 π
ω= ⇒ Ω=
Ts 2
jΩ
s = jω ⇒ z = e .

Other points are:

2 2
z=0⇒s=− , z=∞⇒s= .
Ts Ts
G. Ducard 36 / 41
 Infinite Impulse Response Filters Methodology
First-Order Low-Pass Filter CT Butterworth filter design
IIR Filter Design Bilinear transform

Region of stability mapping

Stable poles in the continuous domain are mapped to stable poles in the
discrete domain:
s-domain z-domain

This is desirable, as it means that a stable CT filter is transformed into a stable

DT filter.

Summary
The bilinear transform preserves stability and maps the imaginary axis in the
s-plane to the unit circle in the z-plane by compressing the CT frequencies
−∞ < ω < ∞ to DT frequencies −π < Ω < π.
G. Ducard 37 / 41
 Infinite Impulse Response Filters Methodology
First-Order Low-Pass Filter CT Butterworth filter design
IIR Filter Design Bilinear transform

Example: Converting a CT first-order low-pass filter

Consider the CT low-pass filter with time constant τ

1
H(s) = .
τs + 1

Using the bilinear transform, we obtain

1 1 − α 1 + z −1
H(z) =   =
1 + τ T2s z−1 1 − αz −1 2
z+1

with
Ts
1− 2τ
α= Ts
.
1+ 2τ

G. Ducard 38 / 41
 Infinite Impulse Response Filters Methodology
First-Order Low-Pass Filter CT Butterworth filter design
IIR Filter Design Bilinear transform

Example: Converting a CT first-order low-pass filter

Ts Ts
For small values of τ
, α ≈1− τ
, as before.

Let us compare different discretization methods:

Method Transfer function Filter parameter

Ts
1−α
Direct H(z) = 1−αz −1
α = e− τ (decay time τ )
− Tτs
−1
(1−α)z
Sample and Hold H(z) = 1−αz −1
α=e (time constant τ )
−1
(1−α)( 1+z2 ) 1− T s
Bilinear H(z) = α= 2τ
(time constant τ )
1−αz 1+ T
−1 s
2τ

G. Ducard 39 / 41
 Infinite Impulse Response Filters Methodology
First-Order Low-Pass Filter CT Butterworth filter design
IIR Filter Design Bilinear transform

Example: Converting a CT first-order low-pass filter

A nice property of the bilinear transform, for this example: is that

H(z = −1) = 0

This corresponds to
z = −1 = e−jπ
which is the highest possible DT frequency .

Remark: Observe that the CT first-order low-pass has

lim H(s) = 0
s→j∞

Conclusion :
the frequency responses at the highest possible frequencies are the same for the
CT and the DT system when using the bilinear transform.
⇒ It follows that their high-frequency behavior is similar, which is one of the
advantages of the bilinear transform. G. Ducard 40 / 41
 Infinite Impulse Response Filters Methodology
First-Order Low-Pass Filter CT Butterworth filter design
IIR Filter Design Bilinear transform

Example: Converting a CT first-order low-pass filter

This can be seen by looking at the frequency response plots of the resulting
filters for Ts = 1 and τ = 2:

1 Continuous-time
Bilinear transform
|H(Ω)|

ZOH
0.5 Direct

90
∠H(Ω)

0
−90

0 π/2 π
G. Ducard 41 / 41