Chapter 4

DISCRETE-TIME SYSTEMS
4.6 Convolution Summation
4.7 Stability

Copyright c 2005 Andreas Antoniou

Victoria, BC, Canada
Email: [email protected]

July 14, 2018

Frame # 1 Slide # 1 A. Antoniou Digital Signal Processing – Secs. 4.6, 4.7

Introduction

N The convolution summation is of considerable importance for

the characterization, representation, analysis, and design of
discrete-time systems.
N This presentation will deal with the derivation, properties, and
applications of the convolution summation.

Frame # 2 Slide # 2 A. Antoniou Digital Signal Processing – Secs. 4.6, 4.7

Derivation

N An arbitrary excitation x(nT ) can be considered to be made

up of a series of impulses as shown:
x(2T )
x(nT ) x(T )
x(0)

0 T 2T 3T nT

=
x1(nT )
x(0)

nT
+
x2(nT ) x(T )

nT
+
x(2T )
x3(nT )

nT

Frame # 3 Slide # 3 A. Antoniou Digital Signal Processing – Secs. 4.6, 4.7

Derivation Cont’d

N What has been done graphically can now be done in terms of

algebra.

Frame # 4 Slide # 4 A. Antoniou Digital Signal Processing – Secs. 4.6, 4.7

Derivation Cont’d

N What has been done graphically can now be done in terms of

algebra.
N An arbitrary signal can be written as
∞
X
x(nT ) = xk (nT )
k=−∞
(
x(kT ) for n = k
where xk (nT ) =
0 otherwise
= x(kT )δ(nT − kT )

Frame # 4 Slide # 5 A. Antoniou Digital Signal Processing – Secs. 4.6, 4.7

Derivation Cont’d

N What has been done graphically can now be done in terms of

algebra.
N An arbitrary signal can be written as
∞
X
x(nT ) = xk (nT )
k=−∞
(
x(kT ) for n = k
where xk (nT ) =
0 otherwise
= x(kT )δ(nT − kT )
N Hence
∞
X
x(nT ) = x(kT )δ(nT − kT ) (A)
k=−∞

Frame # 4 Slide # 6 A. Antoniou Digital Signal Processing – Secs. 4.6, 4.7

Derivation Cont’d

··· ∞
X
x(nT ) = x(kT )δ(nT − kT ) (A)
k=−∞

N Consider a linear time-invariant system and assume that its

impulse response is given by
h(nT ) = Rδ(nT )

Frame # 5 Slide # 7 A. Antoniou Digital Signal Processing – Secs. 4.6, 4.7

Derivation Cont’d

··· ∞
X
x(nT ) = x(kT )δ(nT − kT ) (A)
k=−∞

N Consider a linear time-invariant system and assume that its

impulse response is given by
h(nT ) = Rδ(nT )
N From Eq. (A), we have
∞
X
y (nT ) = Rx(nT ) = R x(kT )δ(nT − kT )
k=−∞

Frame # 5 Slide # 8 A. Antoniou Digital Signal Processing – Secs. 4.6, 4.7

Derivation Cont’d

··· ∞
X
x(nT ) = x(kT )δ(nT − kT ) (A)
k=−∞

N Consider a linear time-invariant system and assume that its

impulse response is given by
h(nT ) = Rδ(nT )
N From Eq. (A), we have
∞
X
y (nT ) = Rx(nT ) = R x(kT )δ(nT − kT )
k=−∞

N Since the system is linear,

X∞
y (nT ) = x(kT )Rδ(nT − kT )
k=−∞

Frame # 5 Slide # 9 A. Antoniou Digital Signal Processing – Secs. 4.6, 4.7

Derivation Cont’d

··· ∞
X
y (nT ) = x(kT )Rδ(nT − kT )
k=−∞

N The system is also time-invariant and hence we get

∞
X
y (nT ) = x(kT )h(nT − kT )
k=−∞

Frame # 6 Slide # 10 A. Antoniou Digital Signal Processing – Secs. 4.6, 4.7

Derivation Cont’d

··· ∞
X
y (nT ) = x(kT )Rδ(nT − kT )
k=−∞

N The system is also time-invariant and hence we get

∞
X
y (nT ) = x(kT )h(nT − kT )
k=−∞

N This relation is known as the convolution summation.

Frame # 6 Slide # 11 A. Antoniou Digital Signal Processing – Secs. 4.6, 4.7

Alternative Form

N If we let k = n − k 0 in the convolution summation

∞
X
y (nT ) = x(kT )h(nT − kT )
k=−∞

then k 0 = n − k.

Frame # 7 Slide # 12 A. Antoniou Digital Signal Processing – Secs. 4.6, 4.7

Alternative Form

N If we let k = n − k 0 in the convolution summation

∞
X
y (nT ) = x(kT )h(nT − kT )
k=−∞

then k 0 = n − k.

N If
k → ∞ then k 0 → −∞
and if
k → −∞ then k0 → ∞

Frame # 7 Slide # 13 A. Antoniou Digital Signal Processing – Secs. 4.6, 4.7

Alternative Form

N If we let k = n − k 0 in the convolution summation

∞
X
y (nT ) = x(kT )h(nT − kT )
k=−∞

then k 0 = n − k.

N If
k → ∞ then k 0 → −∞
and if
k → −∞ then k0 → ∞

N Hence the convolution summation can also be expressed as

−∞
X
y (nT ) = x(nT − k 0 T )h(k 0 T )
k 0 =∞

Frame # 7 Slide # 14 A. Antoniou Digital Signal Processing – Secs. 4.6, 4.7

Alternative Form Cont’d

··· −∞
X
y (nT ) = x(nT − k 0 T )h(k 0 T )
k 0 =∞

N Dropping the primes and reversing the order of summation,

we obtain the identity
∞
X ∞
X
y (nT ) = x(kT )h(nT − kT ) = h(kT )x(nT − kT )
k=−∞ k=−∞

Frame # 8 Slide # 15 A. Antoniou Digital Signal Processing – Secs. 4.6, 4.7

Special Cases

N Two special cases of the convolution summation are of

particular interest.

Frame # 9 Slide # 16 A. Antoniou Digital Signal Processing – Secs. 4.6, 4.7

Special Cases

N Two special cases of the convolution summation are of

particular interest.
N If the system is causal, we have h(nT ) = 0 for n < 0, and so
n
X ∞
X
y (nT ) = x(kT )h(nT − kT ) = h(kT )x(nT − kT )
k=−∞ k=0

Frame # 9 Slide # 17 A. Antoniou Digital Signal Processing – Secs. 4.6, 4.7

Special Cases

N Two special cases of the convolution summation are of

particular interest.
N If the system is causal, we have h(nT ) = 0 for n < 0, and so
n
X ∞
X
y (nT ) = x(kT )h(nT − kT ) = h(kT )x(nT − kT )
k=−∞ k=0

N If, in addition, x(nT ) = 0 for n < 0, we have

n
X n
X
y (nT ) = x(kT )h(nT − kT ) = h(kT )x(nT − kT )
k=0 k=0

Frame # 9 Slide # 18 A. Antoniou Digital Signal Processing – Secs. 4.6, 4.7

Important Property

···
n
X n
X
y (nT ) = x(kT )h(nT − kT ) = h(kT )x(nT − kT )
k=0 k=0

Evidently, if the impulse response h(nT ) of a discrete-time system

is known, its response to an arbitrary excitation can be readily
determined by using the convolution summation.

Frame # 10 Slide # 19 A. Antoniou Digital Signal Processing – Secs. 4.6, 4.7

Graphical Representation
h(kT ) x(kT )

kT kT
(a) (b)

×
h(−kT )

h(nT − kT )

kT kT
nT
(c) (d )
=
y(nT )

}
x(kT )h(nT − kT )

kT
(e)

Frame # 11 Slide # 20 A. Antoniou Digital Signal Processing – Secs. 4.6, 4.7

Example

Using the convolution summation, find the unit-step response of a

discrete-time system characterized by the equation

y (nT ) = x(nT ) + py (nT − T )

The system has an impulse response

h(nT ) = u(nT )p n

and is initially relaxed (i.e., y (nT ) = 0 for n < 0).

Frame # 12 Slide # 21 A. Antoniou Digital Signal Processing – Secs. 4.6, 4.7

Example Cont’d

Solution The convolution summation gives

∞
X
y (nT ) = Ru(nT ) = u(kT )p k u(nT − kT )
k=−∞

k=−1 k=0 k=1

z }| { z }| { z }| {
= · · · + u(−T )p −1 u(nT + T ) + u(0)p 0 u(nT ) + u(T )p 1 u(nT − T )
k=n k=n+1
z }| { z }| {
+ · · · + u(nT )p n u(0) + u(nT + T )p n+1 u(−T ) + · · ·

For n < 0, the unit step assumes a value of zero and hence we get
y (nT ) = 0 since all the terms are zero.

Frame # 13 Slide # 22 A. Antoniou Digital Signal Processing – Secs. 4.6, 4.7

Example Cont’d

··· ∞
X
y (nT ) = Ru(nT ) = u(kT )p k u(nT − kT )
k=−∞

k=−1 k=0 k=1

z }| { z }| { z }| {
= · · · + u(−T )p −1 u(nT + T ) + u(0)p 0 u(nT ) + u(T )p 1 u(nT − T )
k=n k=n+1
z }| { z }| {
+ · · · + u(nT )p n u(0) + u(nT + T )p n+1 u(−T ) + · · ·

For n ≥ 0, we obtain
n
X 1 − p (n+1)
y (nT ) = 1 + p 1 + p 2 + · · · + p n = 1 + pn =
n=1
1−p

since this is a geometric series with a common ratio p.

Frame # 14 Slide # 23 A. Antoniou Digital Signal Processing – Secs. 4.6, 4.7

Example Cont’d

···
For n < 0, y (nT ) = 0.
For n ≥ 0,
1 − p (n+1)
y (nT ) =
1−p
Therefore, the response can be expressed in closed form as

1 − p (n+1)
y (nT ) = u(nT )
1−p

Frame # 15 Slide # 24 A. Antoniou Digital Signal Processing – Secs. 4.6, 4.7

Example

1 − p (n+1)
y (nT ) = Ru(nT ) = u(nT )
1−p
and is initially relaxed (i.e., y (nT ) = 0 for n < 0).
Find the response produced by the excitation
(
1 for 0 ≤ n ≤ 4
x(nT ) =
0 otherwise

Frame # 16 Slide # 25 A. Antoniou Digital Signal Processing – Secs. 4.6, 4.7

Example Cont’d

Solution We observe that

(
1 for 0 ≤ n ≤ 4
x(nT ) = = u(nT ) − u(nT − 5T )
0 otherwise

and so

y (nT ) = Rx(nT ) = Ru(nT ) − Ru(nT − 5T )

Since
1 − p (n+1)
y (nT ) = Ru(nT ) = u(nT )
1−p
we get

1 − p (n+1) 1 − p (n−4)
y (nT ) = u(nT ) − u(nT − 5T )
1−p 1−p

Frame # 17 Slide # 26 A. Antoniou Digital Signal Processing – Secs. 4.6, 4.7

Example

An initially relaxed causal nonrecursive system was tested with an

input (
0 for n < 0
x(nT ) =
n for n ≥ 0
and found to have the response given by the following table:

n 0 1 2 3 4 5 6 7
y (nT ) 0 1 4 10 20 30 40 50

(a) Find the impulse response of the system for values of n over
the range 0 ≤ n ≤ 5.
(b) Using the result in part (a), find the unit-step response for
0 ≤ n ≤ 5.

Frame # 18 Slide # 27 A. Antoniou Digital Signal Processing – Secs. 4.6, 4.7

Example Cont’d

Solution (a) Since the system is causal and x(nT ) = 0 for n < 0, the
convolution summation assumes the form
n
X
y (nT ) = Rx(nT ) = x(kT )h(nT − kT )
k=0
= x(0)h(nT ) + x(T )h(nT − T ) + · · · + h(0)x(nT )

Frame # 19 Slide # 28 A. Antoniou Digital Signal Processing – Secs. 4.6, 4.7

Example Cont’d

Solution (a) Since the system is causal and x(nT ) = 0 for n < 0, the
convolution summation assumes the form
n
X
y (nT ) = Rx(nT ) = x(kT )h(nT − kT )
k=0
= x(0)h(nT ) + x(T )h(nT − T ) + · · · + h(0)x(nT )

Evaluating y (nT ) for n = 1, 2, . . ., we get

y (T ) = x(0)h(T ) + x(T )h(0) = 0 · h(T ) + 1 · h(0) = 1 or h(0) = 1

y (2T ) = x(0)h(2T ) + x(T )h(T ) + x(2T )h(0)
= 0 · h(2T ) + 1 · h(T ) + 2 · h(0)
= 0 + h(T ) + 2 = 4 or h(T ) = 2
y (3T ) = x(0)h(3T ) + x(T )h(2T ) + x(2T )h(T ) + x(3T )h(0)
= 0 · h(3T ) + 1 · h(2T ) + 2 · h(T ) + 3 · h(0)
= h(2T ) + 2 · 2 + 3 · 1 = 10 or h(2T ) = 3

Frame # 19 Slide # 29 A. Antoniou Digital Signal Processing – Secs. 4.6, 4.7

Example Cont’d

y (4T ) = x(0)h(4T ) + x(T )h(3T ) + x(2T )h(2T ) + x(3T )h(T )

+x(4T )h(0)
= 0 · h(4T ) + 1 · h(3T ) + 2 · h(2T ) + 3 · h(T ) + 4 · h(0)
= h(3T ) + 2 · 3 + 3 · 2 + 4 · 1 = 20 or h(3T ) = 4
y (5T ) = x(0)h(5T ) + x(T )h(4T ) + x(2T )h(3T ) + x(3T )h(2T )
+ x(4T )h(T ) + x(5T )h(0)
= 0 · h(5T ) + 1 · h(4T ) + 2 · h(3T ) + 3 · h(2T ) + 4 · h(T ) + 5 · h(0)
= 0 + h(4T ) + 2 · 4 + 3 · 3 + 4 · 2 + 5 · 1 = 30 or h(4T ) = 0
y (6T ) = x(0)h(6T ) + x(T )h(5T ) + x(2T )h(4T ) + x(3T )h(3T )
+ x(4T )h(2T ) + x(5T )h(T ) + x(6T )h(0)
= 0 · h(6T ) + 1 · h(5T ) + 2 · h(4T ) + 3 · h(3T ) + 4 · h(2T )
+5 · h(T ) + 6 · h(0)
= h(5T ) + 2 · 0 + 3 · 4 + 4 · 3 + 5 · 2 + 6 · 1
= 40 or h(5T ) = 0

Frame # 20 Slide # 30 A. Antoniou Digital Signal Processing – Secs. 4.6, 4.7

Example Cont’d

Summarizing the results so far, we have

h(0) = 1 h(T ) = 2 h(2T ) = 3

h(3T ) = 4 h(4T ) = 0 h(5T ) = 0

Frame # 21 Slide # 31 A. Antoniou Digital Signal Processing – Secs. 4.6, 4.7

Example Cont’d

(b) Using the convolution summation again, we obtain the

unit-step response as follows:
n
X n
X
y (nT ) = Rx(nT ) = u(kT )h(nT − kT ) = h(nT − kT )
k=0 k=0

Hence

y (0) = h(0) = 1
y (T ) = h(T ) + h(0) = 2 + 1 = 3
y (2T ) = h(2T ) + h(T ) + h(0) = 3 + 2 + 1 = 6
y (3T ) = h(3T ) + h(2T ) + h(T ) + h(0) = 10
y (4T ) = h(4T ) + h(3T ) + h(2T ) + h(T ) + h(0) = 15
y (5T ) = h(5T ) + h(4T ) + h(3T ) + h(2T ) + h(T ) + h(0) = 21

Frame # 22 Slide # 32 A. Antoniou Digital Signal Processing – Secs. 4.6, 4.7

Alternative Classification of Discrete-Time Systems

Discrete-time systems can also be classified on the basis of the

duration of the impulse response as
– finite-duration impulse response (FIR) systems
– infinite-duration impulse response (IIR) systems

Frame # 23 Slide # 33 A. Antoniou Digital Signal Processing – Secs. 4.6, 4.7

Alternative Classification Cont’d

N If the impulse response of a discrete-time system is of finite

duration such that h(nT ) = 0 for n > N, then the convolution
summation gives
N
X
y (nT ) = h(kT )x(nT − kT )
k=0

Frame # 24 Slide # 34 A. Antoniou Digital Signal Processing – Secs. 4.6, 4.7

Alternative Classification Cont’d

N If the impulse response of a discrete-time system is of finite

duration such that h(nT ) = 0 for n > N, then the convolution
summation gives
N
X
y (nT ) = h(kT )x(nT − kT )
k=0

N This equation is of the same form as the difference equation

of a nonrecursive system, i.e.,
N
X
y (nT ) = ai x(nT − iT )
i=0

with
h(0) = a0 , h(T ) = a1 , . . . , h(NT ) = aN

Frame # 24 Slide # 35 A. Antoniou Digital Signal Processing – Secs. 4.6, 4.7

Alternative Classification Cont’d

Thus we conclude that

N the impulse response of a nonrecursive system is always of
finite duration, and

Frame # 25 Slide # 36 A. Antoniou Digital Signal Processing – Secs. 4.6, 4.7

Alternative Classification Cont’d

Thus we conclude that

N the impulse response of a nonrecursive system is always of
finite duration, and
N given an impulse response of finite duration, a nonrecursive
system can be obtained.

Frame # 25 Slide # 37 A. Antoniou Digital Signal Processing – Secs. 4.6, 4.7

Alternative Classification Cont’d

N An impulse response of infinite duration could be achieved

with a nonrecursive system of infinite order or with a recursive
system.

Frame # 26 Slide # 38 A. Antoniou Digital Signal Processing – Secs. 4.6, 4.7

Alternative Classification Cont’d

N An impulse response of infinite duration could be achieved

with a nonrecursive system of infinite order or with a recursive
system.
N Since infinite-order systems are not feasible, an
infinite-duration impulse response can only be achieved with a
recursive system.

Frame # 26 Slide # 39 A. Antoniou Digital Signal Processing – Secs. 4.6, 4.7

Alternative Classification Cont’d

N An impulse response of infinite duration could be achieved

with a nonrecursive system of infinite order or with a recursive
system.
N Since infinite-order systems are not feasible, an
infinite-duration impulse response can only be achieved with a
recursive system.
N To confuse the issue somewhat, it is possible to construct a
recursive system that has a finite-duration impulse response!

Frame # 26 Slide # 40 A. Antoniou Digital Signal Processing – Secs. 4.6, 4.7

An FIR Recursive System

N To illustrate that an FIR system can be represented by a

recursive equation, or by a network with feedback, consider an
FIR system represented by the equation

y (nT ) = x(nT ) + 3x(nT − T )

Frame # 27 Slide # 41 A. Antoniou Digital Signal Processing – Secs. 4.6, 4.7

An FIR Recursive System

N To illustrate that an FIR system can be represented by a

recursive equation, or by a network with feedback, consider an
FIR system represented by the equation

y (nT ) = x(nT ) + 3x(nT − T )

N If we premultiply both sides of the equation by the operator

(1 + 4E −1 ), we get

(1 + 4E −1 )y (nT ) = (1 + 4E −1 )[x(nT ) + 3x(nT − T )]

Frame # 27 Slide # 42 A. Antoniou Digital Signal Processing – Secs. 4.6, 4.7

An FIR Recursive System

N To illustrate that an FIR system can be represented by a

recursive equation, or by a network with feedback, consider an
FIR system represented by the equation

y (nT ) = x(nT ) + 3x(nT − T )

N If we premultiply both sides of the equation by the operator

(1 + 4E −1 ), we get

(1 + 4E −1 )y (nT ) = (1 + 4E −1 )[x(nT ) + 3x(nT − T )]

N After simplification, we have

y (nT ) + 4y (nT − T ) = x(nT ) + 3x(nT − T )

+4x(nT − T ) + 12x(nT − 2T )

Frame # 27 Slide # 43 A. Antoniou Digital Signal Processing – Secs. 4.6, 4.7

An FIR Recursive System Cont’d

···
y (nT ) + 4y (nT − T ) = x(nT ) + 3x(nT − T )
+4x(nT − T ) + 12x(nT − 2T )

N Thus the FIR system can be represented by the recursive

equation

y (nT ) = x(nT ) + 7x(nT − T ) + 12x(nT − 2T ) − 4y (nT − T )

Frame # 28 Slide # 44 A. Antoniou Digital Signal Processing – Secs. 4.6, 4.7

An FIR Recursive System Cont’d

···
y (nT ) + 4y (nT − T ) = x(nT ) + 3x(nT − T )
+4x(nT − T ) + 12x(nT − 2T )

N Thus the FIR system can be represented by the recursive

equation

y (nT ) = x(nT ) + 7x(nT − T ) + 12x(nT − 2T ) − 4y (nT − T )

N Evidently, the manipulation has actually increased the order of

the difference equation and, therefore, no obvious advantage
is gained by converting an FIR system into a recursive one.

Frame # 28 Slide # 45 A. Antoniou Digital Signal Processing – Secs. 4.6, 4.7

An FIR Recursive System Cont’d

···
y (nT ) + 4y (nT − T ) = x(nT ) + 3x(nT − T )
+4x(nT − T ) + 12x(nT − 2T )

N Thus the FIR system can be represented by the recursive

equation

y (nT ) = x(nT ) + 7x(nT − T ) + 12x(nT − 2T ) − 4y (nT − T )

N Evidently, the manipulation has actually increased the order of

the difference equation and, therefore, no obvious advantage
is gained by converting an FIR system into a recursive one.
N For most practical purposes nonrecursive systems are FIR
systems and recursive systems are IIR systems.
Frame # 28 Slide # 46 A. Antoniou Digital Signal Processing – Secs. 4.6, 4.7
Alternative Classification Cont’d

Nonrecursive FIR

Impossible Possible but

unnecessary

Recursive IIR

Note: An IIR system cannot be a nonrecursive system and

vice-versa. However, a recursive system can be constructed that is
also an FIR system but such a system would serve no useful
purpose.
Frame # 29 Slide # 47 A. Antoniou Digital Signal Processing – Secs. 4.6, 4.7
Stability

N A discrete-time system is said to be stable if and only if any

bounded excitation results in a bounded response, i.e.,

if |x(nT )| < ∞ then |y (nT )| < ∞

Frame # 30 Slide # 48 A. Antoniou Digital Signal Processing – Secs. 4.6, 4.7

Stability

N A discrete-time system is said to be stable if and only if any

bounded excitation results in a bounded response, i.e.,

if |x(nT )| < ∞ then |y (nT )| < ∞

N For a linear and time-invariant system, the convolution

summation gives
∞
X
y (nT ) = h(kT )x(nT − kT )
k=−∞

Frame # 30 Slide # 49 A. Antoniou Digital Signal Processing – Secs. 4.6, 4.7

Stability

N A discrete-time system is said to be stable if and only if any

bounded excitation results in a bounded response, i.e.,

if |x(nT )| < ∞ then |y (nT )| < ∞

N For a linear and time-invariant system, the convolution

summation gives
∞
X
y (nT ) = h(kT )x(nT − kT )
k=−∞

N Hence
∞
X ∞
X
|y (nT )| = h(kT )x(nT − kT ) ≤ |h(kT )·x(nT −kT )|
k=−∞ k=−∞

Frame # 30 Slide # 50 A. Antoniou Digital Signal Processing – Secs. 4.6, 4.7

Stability Cont’d

···
∞
X ∞
X
|y (nT )| = h(kT )x(nT − kT ) ≤ |h(kT ) · x(nT − kT )|
k=−∞ k=−∞

N For example,
X
2 · 3 + (−1) · 4 + 2 · (−2) + (−3) · (−3) = 7
X
≤ |2 · 3| + |(−1) · 4| + |2 · (−2)| + |(−3) · (−3)| = 23

Frame # 31 Slide # 51 A. Antoniou Digital Signal Processing – Secs. 4.6, 4.7

Stability Cont’d

···
∞
X ∞
X
|y (nT )| = h(kT )x(nT − kT ) ≤ |h(kT ) · x(nT − kT )|
k=−∞ k=−∞

N For example,
X
2 · 3 + (−1) · 4 + 2 · (−2) + (−3) · (−3) = 7
X
≤ |2 · 3| + |(−1) · 4| + |2 · (−2)| + |(−3) · (−3)| = 23

N
If |x(nT )| ≤ P < ∞ for all n
∞
X
we have |y (nT )| ≤ P |h(kT )|
k=−∞

Frame # 31 Slide # 52 A. Antoniou Digital Signal Processing – Secs. 4.6, 4.7

Stability Cont’d

··· ∞
X
|y (nT )| ≤ P |h(kT )|
k=−∞

N Clearly, if
∞
X
|h(kT )| < ∞ (B)
k=−∞

then |y (nT )| < ∞ for all n

Frame # 32 Slide # 53 A. Antoniou Digital Signal Processing – Secs. 4.6, 4.7

Stability Cont’d

··· ∞
X
|y (nT )| ≤ P |h(kT )|
k=−∞

N Clearly, if
∞
X
|h(kT )| < ∞ (B)
k=−∞

then |y (nT )| < ∞ for all n

N Therefore, Eq. (B) constitutes a sufficient condition for

stability.

Frame # 32 Slide # 54 A. Antoniou Digital Signal Processing – Secs. 4.6, 4.7

Stability Cont’d

N A discrete-time system can be classified as stable if and only if its

response is bounded for all possible bounded excitations.

Frame # 33 Slide # 55 A. Antoniou Digital Signal Processing – Secs. 4.6, 4.7

Stability Cont’d

N A discrete-time system can be classified as stable if and only if its

response is bounded for all possible bounded excitations.
N Let us consider a bounded excitation of the form
(
P if h(kT ) ≥ 0
x(nT − kT ) =
−P if h(kT ) < 0
where P is a positive constant.

Frame # 33 Slide # 56 A. Antoniou Digital Signal Processing – Secs. 4.6, 4.7

Stability Cont’d

N A discrete-time system can be classified as stable if and only if its

response is bounded for all possible bounded excitations.
N Let us consider a bounded excitation of the form
(
P if h(kT ) ≥ 0
x(nT − kT ) =
−P if h(kT ) < 0
where P is a positive constant.
N From the convolution summation, we get
∞
X
|y (nT )| = x(nT − kT )h(kT )
k=−∞

Frame # 33 Slide # 57 A. Antoniou Digital Signal Processing – Secs. 4.6, 4.7

Stability Cont’d

N A discrete-time system can be classified as stable if and only if its

response is bounded for all possible bounded excitations.
N Let us consider a bounded excitation of the form
(
P if h(kT ) ≥ 0
x(nT − kT ) =
−P if h(kT ) < 0
where P is a positive constant.
N From the convolution summation, we get
∞
X
|y (nT )| = x(nT − kT )h(kT )
k=−∞

N Hence
∞
X ∞
X
|y (nT )| = P · |h(kT )| = P |h(kT )|
k=−∞ k=−∞

Frame # 33 Slide # 58 A. Antoniou Digital Signal Processing – Secs. 4.6, 4.7

Stability Cont’d

···
∞
X
|y (nT )| = P |h(kT )|
k=−∞

N Evidently, at least for the type of signal under consideration,

the response will be bounded if and only if
∞
X
|h(kT )| < ∞
k=−∞

which implies that this condition is also a necessary condition

for stability.

Frame # 34 Slide # 59 A. Antoniou Digital Signal Processing – Secs. 4.6, 4.7

Stability Cont’d

N Summarizing, the condition

∞
X
|h(kT )| < ∞
k=−∞

is both a necessary and sufficient condition for stability.

Frame # 35 Slide # 60 A. Antoniou Digital Signal Processing – Secs. 4.6, 4.7

Stability Cont’d

N Summarizing, the condition

∞
X
|h(kT )| < ∞
k=−∞

is both a necessary and sufficient condition for stability.

N Note: In nonrecursive systems, the impulse response is both

finite in value and also of finite duration and hence the above
condition is always satisfied, i.e., nonrecursive systems are
always stable.

Frame # 35 Slide # 61 A. Antoniou Digital Signal Processing – Secs. 4.6, 4.7

Example

A first-order system is characterized by the equation

y (nT ) = x(nT ) + py (nT − T )
has an impulse response
h(nT ) = u(nT )p n
Check the stability of the system.

Frame # 36 Slide # 62 A. Antoniou Digital Signal Processing – Secs. 4.6, 4.7

Example

A first-order system is characterized by the equation

y (nT ) = x(nT ) + py (nT − T )
has an impulse response
h(nT ) = u(nT )p n
Check the stability of the system.
Solution We can write
X∞
|h(kT )| = 1 + |p| + · · · + |p k | + · · ·
k=−∞

Frame # 36 Slide # 63 A. Antoniou Digital Signal Processing – Secs. 4.6, 4.7

Example

A first-order system is characterized by the equation

y (nT ) = x(nT ) + py (nT − T )
has an impulse response
h(nT ) = u(nT )p n
Check the stability of the system.
Solution We can write
X∞
|h(kT )| = 1 + |p| + · · · + |p k | + · · ·
k=−∞

This is a geometric series and has a sum

∞
X 1 − |p|(n+1)
|h(kT )| = lim
n→∞ 1 − |p|
k=−∞

Frame # 36 Slide # 64 A. Antoniou Digital Signal Processing – Secs. 4.6, 4.7

Example Cont’d

If p > 1,
∞
X 1 − |p|(n+1)
|h(kT )| = lim →∞
n→∞ 1 − |p|
k=−∞

Frame # 37 Slide # 65 A. Antoniou Digital Signal Processing – Secs. 4.6, 4.7

Example Cont’d

If p > 1,
∞
X 1 − |p|(n+1)
|h(kT )| = lim →∞
n→∞ 1 − |p|
k=−∞
and if p = 1,
∞
X
|h(kT )| = 1 + 1 + 1 + · · · = ∞
k=−∞

Frame # 37 Slide # 66 A. Antoniou Digital Signal Processing – Secs. 4.6, 4.7

Example Cont’d

If p > 1,
∞
X 1 − |p|(n+1)
|h(kT )| = lim →∞
n→∞ 1 − |p|
k=−∞
and if p = 1,
∞
X
|h(kT )| = 1 + 1 + 1 + · · · = ∞
k=−∞

On the other hand, if p < 1,

∞
X 1 − |p|(n+1) 1
|h(kT )| = lim → =K <∞
n→∞ 1 − |p| 1 − |p|
k=−∞

where K is a positive constant. Therefore, the system is stable if

and only if
|p| < 1
Frame # 37 Slide # 67 A. Antoniou Digital Signal Processing – Secs. 4.6, 4.7
Example

A discrete-time system has an impulse response

nπ
h(nT ) = u(nT )e 0.1nT sin
6
Check the stability of the system.

Frame # 38 Slide # 68 A. Antoniou Digital Signal Processing – Secs. 4.6, 4.7

Example

A discrete-time system has an impulse response

nπ
h(nT ) = u(nT )e 0.1nT sin
6
Check the stability of the system.
Solution We can write
∞ ∞
X X kπ
|h(nT )| = u(kT )e 0.1kT sin
6
k=0 k=0
∞ ∞
X
0.1kT
X kπ
= e + e 0.1kT sin →∞
6
k=3,9,15,... k6=3,9,15,...

Therefore, the system is unstable.

Frame # 38 Slide # 69 A. Antoniou Digital Signal Processing – Secs. 4.6, 4.7

 This slide concludes the presentation.
Thank you for your attention.

Frame # 39 Slide # 70 A. Antoniou Digital Signal Processing – Secs. 4.6, 4.7