Impulse Response Impulse Response

• Example - The impulse response of the • Example - The impulse response of the
system discrete-time accumulator
y[n] � �1x[n] � � 2 x[n � 1] � � 3 x[n � 2] � � 4 x[n � 3] n
is obtained by setting x[n] = �[n] resulting
y[n] � � x[�]
� � ��
in
is obtained by setting x[n] = �[n] resulting
h[n] � �1� [n] � � 2� [n � 1] � � 3� [n � 2] � � 4� [n � 3] in n
• The impulse response is thus a finite-length h[n] � � � [�] � � [n]
sequence of length 4 given by � � ��
{h[n]} � {�1, � 2 , � 3 , � 4}
43
� Copyright © 2005, S. K. Mitra
44
Copyright © 2005, S. K. Mitra

Impulse Response Time-

Time-Domain Characterization
• Example - The impulse response {h[n]} of of LTI Discrete-
Discrete-Time System
the factor-of-2 interpolator • Input-Output Relationship -
1
y[n] � xu [n] � ( xu [n � 1] � xu [n � 1]) A consequence of the linear, time-
2
• is obtained by setting xu [n] � � [n] and is invariance property is that an LTI discrete-
given by time system is completely characterized by
1 its impulse response
h[n] � � [n] � (� [n � 1] � � [n � 1])
2
• Knowing the impulse response one
• The impulse response is thus a finite-length
can compute the output of the system for
sequence of length 3:
any arbitrary input
45
{h[n]} � {0.5, 1 0.5} 46
� Copyright © 2005, S. K. Mitra Copyright © 2005, S. K. Mitra

Time-
Time-Domain Characterization Time-
Time-Domain Characterization
of LTI Discrete-
Discrete-Time System of LTI Discrete-
Discrete-Time System
• Let h[n] denote the impulse response of a • Since the system is time-invariant
LTI discrete-time system input output
• We compute its output y[n] for the input: �[n � 2] � h[n � 2]
x[ n] � 0.5�[ n � 2] � 1.5�[n � 1] � �[ n � 2] � 0.75�[ n � 5]
�[n � 1] � h[n � 1]
• As the system is linear, we can compute its
�[n � 2] � h[n � 2]
outputs for each member of the input
separately and add the individual outputs to �[n � 5] � h[n � 5]
determine y[n]
47 48
Copyright © 2005, S. K. Mitra Copyright © 2005, S. K. Mitra

8
 Time-
Time-Domain Characterization Time-
Time-Domain Characterization
of LTI Discrete-
Discrete-Time System of LTI Discrete-
Discrete-Time System
• Likewise, as the system is linear
input output • Now, any arbitrary input sequence x[n] can
0.5�[n � 2] � 0.5h[n � 2] be expressed as a linear combination of
1.5�[n � 1] � 1.5h[n � 1] delayed and advanced unit sample
� �[n � 2] � � h[n � 2] sequences in the form
�
0.75�[n � 5] � 0.75h[n � 5] x[n] � � x[k ] �[n � k ]
• Hence because of the linearity property we k � ��
get • The response of the LTI system to an input
y[n] � 0.5h[n � 2] � 1.5h[n � 1]
x[ k ] �[n � k ] will be x[k ] h[n � k ]
49
� h[n � 2] � 0.75h[n � 5] 50
Copyright © 2005, S. K. Mitra Copyright © 2005, S. K. Mitra

Time-
Time-Domain Characterization
Convolution Sum
of LTI Discrete-
Discrete-Time System
• Hence, the response y[n] to an input • The summation
� � �
x[ n] � � x[k ] �[ n � k ] y[n] � � x[k ] h[n � k ] � � x[n � k ] h[n]
k � �� k � �� k � ��
will be � is called the convolution sum of the
y[n] � � x[k ] h[ n � k ] sequences x[n] and h[n] and represented
k � ��
which can be alternately written as compactly as
� y[n] = x[n] * h[n]
y[n] � � x[n � k ] h[k ]
51 k � �� 52
Copyright © 2005, S. K. Mitra Copyright © 2005, S. K. Mitra

Convolution Sum Convolution Sum

• Properties -
• Interpretation -
• Commutative property:
• 1) Time-reverse h[k] to form h[� k ]
x[n] * h[n] = h[n] * x[n]
• 2) Shift h[� k ] to the right by n sampling
• Associative property :
periods if n > 0 or shift to the left by n
(x[n] * h[n]) * y[n] = x[n] * (h[n] * y[n])
sampling periods if n < 0 to form h[n � k ]
• Distributive property : • 3) Form the product v[k ] � x[k ]h[n � k ]
x[n] * (h[n] + y[n]) = x[n] * h[n] + x[n] * y[n] • 4) Sum all samples of v[k] to develop the
n-th sample of y[n] of the convolution sum
53 54
Copyright © 2005, S. K. Mitra Copyright © 2005, S. K. Mitra

9
 Convolution Sum Convolution Sum
• Schematic Representation - • We illustrate the convolution operation for
the following two sequences:
h[n � k ] v[k ]
h[� k ] zn � � y[n] �1, 0 � n � 5
k x[ n] � �
x[k ] �0, otherwise
• The computation of an output sample using �1.8 � 0.3n, 0 � n � 5
h[ n] � �
the convolution sum is simply a sum of � 0, otherwise
products • Figures on the next several slides the steps
• Involves fairly simple operations such as involved in the computation of
55
additions, multiplications, and delays 56
y[n] = x[n] * h[n]
Copyright © 2005, S. K. Mitra Copyright © 2005, S. K. Mitra

Convolution Sum Convolution Sum

Plot of x[-4- k] and h[k] h[k]x[-4- k]
2 3
1.5
2
Amplitude

Amplitude
1
0.5 1

0
0
-0.5
-10 0 10 k� -10 0 10 k �
y[-4] y[n]
8 8

6 6
Amplitude

Amplitude
4 4

2 2

0 0
-10 0 10 -10 0 10
n n
57 58
Copyright © 2005, S. K. Mitra Copyright © 2005, S. K. Mitra

Convolution Sum Convolution Sum

Plot of x[-1- k] and h[k] h[k]x[-1- k] Plot of x[0- k] and h[k] h[k]x[0- k]
2 3 2 3
1.5 1.5
2 2
Amplitude

Amplitude

Amplitude

Amplitude

1 1
0.5 1 0.5 1

0 0 0
0
-0.5 -0.5
-10 0 10 -10 0 10 -10 0 10 -10 0 10
y[-1] k� y[n] k� y[0] k� y[n] k�
8 8 8 8

6 6 6 6
Amplitude

Amplitude
Amplitude

Amplitude

4 4 4 4

2 2 2 2

0 0 0 0
-10 0 10 -10 0 10 -10 0 10 -10 0 10
59 n n 60 n n
Copyright © 2005, S. K. Mitra Copyright © 2005, S. K. Mitra

10
 Convolution Sum Convolution Sum
Plot of x[1- k] and h[k] h[k]x[1- k] Plot of x[3- k] and h[k] h[k]x[3- k]
2 3 2 3
1.5 1.5
2 2
Amplitude

Amplitude

Amplitude

Amplitude
1 1
0.5 1 0.5 1

0 0 0 0
-0.5 -0.5
-10 0 10 -10 0 10 -10 0 10 -10 0 10
k� k� k� k�
y[1] y[n] y[3] y[n]
8 8 8 8

6 6 6 6
Amplitude

Amplitude

Amplitude

Amplitude
4 4 4 4

2 2 2 2

0 0 0 0
-10 0 10 -10 0 10 -10 0 10 -10 0 10
61 n n 62 n n
Copyright © 2005, S. K. Mitra Copyright © 2005, S. K. Mitra

Convolution Sum Convolution Sum

Plot of x[5- k] and h[k] h[k]x[5- k] Plot of x[7- k] and h[k] h[k]x[7- k]
2 3 2 3
1.5 1.5
2 2
Amplitude

Amplitude

Amplitude

Amplitude
1 1
0.5 1 0.5 1

0 0 0 0
-0.5 -0.5
-10 0 10
k�
-10 0 10
k�
-10 0 10 k� -10 0 10 k�
y[5] y[n] y[7] y[n]
8 8 8 8

6 6 6 6
Amplitude

Amplitude

Amplitude

Amplitude
4 4 4 4

2 2 2 2

0 0 0 0
-10 0 10 -10 0 10 -10 0 10 -10 0 10
63 n n 64 n n
Copyright © 2005, S. K. Mitra Copyright © 2005, S. K. Mitra

Convolution Sum Convolution Sum

Plot of x[9- k] and h[k] h[k]x[9- k] Plot of x[10- k] and h[k] h[k]x[10- k]
2 3 2 3
1.5 1.5
2 2
Amplitude

Amplitude

Amplitude

Amplitude

1 1
0.5 1 0.5 1

0 0 0 0
-0.5 -0.5
-10 0 10 -10 0 10 -10 0 10 -10 0 10
k� k� k� k�
y[9] y[n] y[10] y[n]
8 8 8 8

6 6 6 6
Amplitude

Amplitude

Amplitude

Amplitude

4 4 4 4

2 2 2 2

0 0 0 0
-10 0 10 -10 0 10 -10 0 10 -10 0 10
65 n n 66 n n
Copyright © 2005, S. K. Mitra Copyright © 2005, S. K. Mitra

11
 Convolution Sum Convolution Sum
Plot of x[12- k] and h[k] h[k]x[12- k] Plot of x[13- k] and h[k] h[k]x[13- k]
2 3 2 3
1.5 1.5
2 2
Amplitude

Amplitude

Amplitude

Amplitude
1 1
0.5 1 0.5 1

0 0 0 0
-0.5 -0.5
-10 0 10
k�
-10 0 10
k�
-10 0 10 k� -10 0 10 k�
y[12] y[n] y[13] y[n]
8 8 8 8

6 6 6 6
Amplitude

Amplitude

Amplitude

Amplitude
4 4 4 4

2 2 2 2

0 0 0 0
-10 0 10 -10 0 10 -10 0 10 -10 0 10
67 n n 68 n n
Copyright © 2005, S. K. Mitra Copyright © 2005, S. K. Mitra

Time-
Time-Domain Characterization Time-
Time-Domain Characterization
of LTI Discrete-
Discrete-Time System of LTI Discrete-
Discrete-Time System
• In practice, if either the input or the impulse • If both the input sequence and the impulse
response is of finite length, the convolution response sequence are of infinite length,
sum can be used to compute the output convolution sum cannot be used to compute
sample as it involves a finite sum of the output
products • For systems characterized by an infinite
• If both the input sequence and the impulse impulse response sequence, an alternate
response sequence are of finite length, the time-domain description involving a finite
output sequence is also of finite length sum of products will be considered
69 70
Copyright © 2005, S. K. Mitra Copyright © 2005, S. K. Mitra

Time-
Time-Domain Characterization Time-
Time-Domain Characterization
of LTI Discrete-
Discrete-Time System of LTI Discrete-
Discrete-Time System
• Example - Develop the sequence y[n] • As can be seen from the shifted time-
generated by the convolution of the reversed version {h[n � k ]} for n < 0, shown
sequences x[n] and h[n] shown below below for n � �3 , for any value of the
x[n]
sample index k, the k-th sample of either
3 h[n] {x[k]} or {h[n � k ]} is zero
2
1 1 h[�3 � k ]
0
2
3
4
n 0 1 2
3
n 2
1
–1 –1 1
–2
–6
–5 –4 –3 –2 –1 0
k
–1
71 72
Copyright © 2005, S. K. Mitra Copyright © 2005, S. K. Mitra

12
 Time-
Time-Domain Characterization Time-
Time-Domain Characterization
of LTI Discrete-
Discrete-Time System of LTI Discrete-
Discrete-Time System
• As a result, for n < 0, the product of the k-th • The product sequence {x[k ]h[�k ]} is plotted
samples of {x[k]} and {h[n � k ]} is always below which has a single nonzero sample
zero, and hence x[0]h[0] for k = 0
y[n] = 0 for n < 0 x[k ]h[ � k ]
0
k
• Consider now the computation of y[0] –5 –4 –3 –2 –1 1 2 3

• The sequence h[ � k ] –2

{h[� k ]} is shown
2
1
–3
• Thus y[0] � x[0]h[0] � �2
k
on the right –1
–6 –5 –4 –2 –1 0 1 2 3
73 74
Copyright © 2005, S. K. Mitra Copyright © 2005, S. K. Mitra

Time-
Time-Domain Characterization Time-
Time-Domain Characterization
of LTI Discrete-
Discrete-Time System of LTI Discrete-
Discrete-Time System
• For the computation of y[1], we shift {h[�k ]} • To calculate y[2], we form {h[2 � k ]} as
to the right by one sample period to form shown below on the left
{h[1 � k ]} as shown below on the left • The product sequence {x[k ]h[2 � k ]} is
• The product sequence {x[k ]h[1 � k ]} is plotted below on the right
shown below on the right h[ 2 � k ]
h[1 � k ] x[k ]h[1 � k ] 2
x[ k ]h[ 2 � k ]
0
2
2 3
k –1
1 1
k
–5 –4 –3 –2 –1 1
1
–4 –3 –2 0 1 2 3 4 5
k –3 –2 –1 0
–2 1 2 3 4 5 6
–5 –4 –3 –1 0 1 2 3
k –1
–1
–4
75 • Hence, y[1] � x[0]h[1] � x[1]h[0] � �4 � 0 � �4 76 y[2] � x[0]h[2] � x[1]h[1] � x[2]h[0] � 0 � 0 � 1 � 1
Copyright © 2005, S. K. Mitra Copyright © 2005, S. K. Mitra

Time-
Time-Domain Characterization Time-
Time-Domain Characterization
of LTI Discrete-
Discrete-Time System of LTI Discrete-
Discrete-Time System
• Continuing the process we get • From the plot of {h[n � k ]} for n > 7 and the
y[3] � x[0]h[3] � x[1]h[2] � x[2]h[1] � x[3]h[0] plot of {x[k]} as shown below, it can be
� 2 � 0 � 0 �1 � 3 seen that there is no overlap between these
two sequences
y[4] � x[1]h[3] � x[2]h[2] � x[3]h[1] � x[4]h[0]
� 0 � 0 � 2 � 3 �1 • As a result y[n] = 0 for n > 7
y[5] � x[2]h[3] � x[3]h[2] � x[4]h[1] x[k] h[8 � k ]
� �1 � 0 � 6 � 5 3
2
y[6] � x[3]h[3] � x[4]h[2] � 1 � 0 � 1 0
1
3
k
5
1
k
y[7] � x[4]h[3] � �3
1 2 4 2 3 4 6 7 8 9 10 11
77 78 –1 –1
Copyright © 2005, S. K. Mitra –2 Copyright © 2005, S. K. Mitra

13
 Time-
Time-Domain Characterization
Time-
Time-Domain Characterization of LTI Discrete-
Discrete-Time System
of LTI Discrete-
Discrete-Time System
• Note: The sum of indices of each sample
• The sequence {y[n]} generated by the product inside the convolution sum is equal
convolution sum is shown below to the index of the sample being generated
y[n] by the convolution operation
5
• For example, the computation of y[3] in the
3
previous example involves the products
0 1
1 1 1
7
n
x[0]h[3], x[1]h[2], x[2]h[1], and x[3]h[0]
2 3 9
• The sum of indices in each of these
–2 –1 4 5 6 8

–2
–3 products is equal to 3
79 –4 80
Copyright © 2005, S. K. Mitra Copyright © 2005, S. K. Mitra

Time- Tabular Method of

Time-Domain Characterization
of LTI Discrete- Convolution Sum Computation
Discrete-Time System
• Can be used to convolve two finite-length
• In the example considered the convolution sequences
of a sequence {x[n]} of length 5 with a • Consider the convolution of {g[n]}, 0 � n � 3 ,
sequence {h[n]} of length 4 resulted in a with {h[n]}, 0 � n � 2 , generating the
sequence {y[n]} of length 8 sequence y[n] = g[n] * h[n]
• In general, if the lengths of the two • Samples of {g[n]} and {h[n]} are then
sequences being convolved are M and N, multiplied using the conventional
then the sequence generated by the multiplication method without any carry
convolution is of length M � N � 1 operation
81 82
Copyright © 2005, S. K. Mitra Copyright © 2005, S. K. Mitra

Tabular Method of Tabular Method of

Convolution Sum Computation Convolution Sum Computation
n:
g[ n ]:
0
g[ 0 ]
1
g[1]
2
g[ 2 ]
3
g[ 3 ]
4 5
• The samples of {y[n]} are given by
y[0] � g[0]h[ 0]
h[ n ]: h[ 0 ] h[1] h[ 2 ]
g[ 0 ]h[ 0 ] g[1]h[ 0 ] g[ 2 ]h[ 0 ] g[ 3]h[ 0 ]
g[ 0 ]h[1] g[1]h[1] g[ 2 ]h[1] g[ 3 ]h[1]
g[ 0 ]h[ 2 ] g[1]h[ 2 ] g[ 2 ]h[ 2 ] g[ 3]h[ 2 ]
y[1] � g[1]h[ 0] � g[ 0]h[1]
y[ n ]: y[ 0 ] y[1] y[ 2 ] y[ 3 ] y[ 4 ] y[ 5 ] y[2] � g[2]h[0] � g[1]h[1] � g[0]h[2]
• The samples y[n] generated by the y[3] � g[3]h[0] � g[2]h[1] � g[1]h[2]
convolution sum are obtained by adding the y[ 4] � g[3]h[1] � g[2]h[2]
entries in the column above each sample y[5] � g[3]h[2]
83 84
Copyright © 2005, S. K. Mitra Copyright © 2005, S. K. Mitra

14
 Tabular Method of Tabular Method of
Convolution Sum Computation Convolution Sum Computation
• The method can also be applied to convolve • Finally, the decimal point is inserted
two finite-length two-sided sequences according to the rules of conventional
• In this case, a decimal point is first placed multiplication
to the right of the sample with the time • The sample immediately to the left of the
index n = 0 for each sequence decimal point is then located at the time
• Next, convolution is computed ignoring the index n = 0
location of the decimal point
85 86
Copyright © 2005, S. K. Mitra Copyright © 2005, S. K. Mitra

Simple Interconnection
Convolution Using MATLAB
Schemes
• The M-file conv implements the convolution
sum of two finite-length sequences • Two simple interconnection schemes are:
• If a �[� 2 0 1 � 1 3] • Cascade Connection
b �[1 2 0 -1] • Parallel Connection
then conv(a,b) yields
[� 2 � 4 1 3 1 5 1 � 3]

87 88
Copyright © 2005, S. K. Mitra Copyright © 2005, S. K. Mitra

Cascade Connection Cascade Connection

h1[n] h2[n] � h2[n] h1[n] • Note: The ordering of the systems in the
cascade has no effect on the overall impulse
� h1[n] * hh[n]�h2[n] [n]
1 response because of the commutative
• Impulse response h[n] of the cascade of two property of convolution
LTI discrete-time systems with impulse • A cascade connection of two stable systems
responses h1[n] and h2[n] is given by is stable
h[n] � h1[n] * h2[n] • A cascade connection of two passive
(lossless) systems is passive (lossless)
89 90
Copyright © 2005, S. K. Mitra Copyright © 2005, S. K. Mitra

15
 Cascade Connection
Cascade Connection
• An application of the inverse system
• An application is in the development of an concept is in the recovery of a signal x[n]
inverse system from its distorted version xˆ[n] appearing at
the output of a transmission channel
• If the cascade connection satisfies the
relation • If the impulse response of the channel is
h1[n] * h 2[n] � �[n] known, then x[n] can be recovered by
designing an inverse system of the channel
then the LTI system h1[n] is said to be the channel ^ inverse system
inverse of h 2[n] and vice-versa x[n ]
x[n ] h1[n] h2[n] x[n ]

91 92 h1[n] * h 2[n] � �[ n]
Copyright © 2005, S. K. Mitra Copyright © 2005, S. K. Mitra

Cascade Connection Cascade Connection

• Example - Consider the discrete-time
accumulator with an impulse response �[n] • Thus the impulse response of the inverse
• Its inverse system satisfy the condition system of the discrete-time accumulator is
given by
�[n] * h 2[n] � �[n]
h2[ n] � �[n] � �[n � 1]
• It follows from the above that h2[n] � 0 for
n < 0 and which is called a backward difference
h2[0] � 1 system
n
� h2[�] � 0 for n � 1
93 � �0 94
Copyright © 2005, S. K. Mitra Copyright © 2005, S. K. Mitra

Parallel Connection Simple Interconnection Schemes

• Consider the discrete-time system where
h1[n]
� � h1[n] � hh[n]�h2[n] [n] h1[ n] � �[n] � 0.5�[n � 1],
1
h2[n]
h2[ n] � 0.5�[n] � 0.25�[n � 1],
h3[n] � 2�[n], h1[n] �
• Impulse response h[n] of the parallel
n
connection of two LTI discrete-time h4[ n] � �2(0.5) �[ n] h2[n]
systems with impulse responses h1[n] and �
h3[n]
h2[n] is given by
h[n] � h1[n] � h2[n] h4[n]
95 96
Copyright © 2005, S. K. Mitra Copyright © 2005, S. K. Mitra

16
 Simple Interconnection Schemes Simple Interconnection Schemes

• Simplifying the block-diagram we obtain • Overall impulse response h[n] is given by

h[n] � h1[n] � h2[n] * (h3[n] � h4[n])
h1[n] � � h1[n] � h 2[n] * h3[n] � h 2[n] * h 4[n]
h2[n] h1[n] � • Now,
�
h 3[ n ] � h 4[ n ] h 2[ n ] * ( h3[ n ]� h 4[ n ]) h2[n] * h3[n] � ( 1 �[ n] � 1 �[n � 1]) * 2�[n]
2 4
� �[n] � 1 �[n � 1]
2

97 98
Copyright © 2005, S. K. Mitra Copyright © 2005, S. K. Mitra

Simple Interconnection Schemes

h2[n] * h4[n] � ( 1 �[n] � 1 �[n � 1]) * � 2( 12 ) n �[n]

2 4
� �
� � ( 12 ) n �[ n] � 12 ( 12 ) n �1�[n � 1]
� � ( 12 ) n �[n] � ( 12 ) n �[n � 1]
� � ( 12 ) n �[n] � � �[n]
• Therefore
h[ n] � �[n] � 12 �[n � 1] � �[n] � 12 �[n � 1] � �[n] � �[n]
99
Copyright © 2005, S. K. Mitra

17