Digital Signal Processing

Convolution
(Adapted from Oppenheim, Alan V., lecture notes)
 STM rEcY:
c-7T: %clecorpose pt 5 pwL
Strategy
X1I
invo GLLineer comet'ncL4Zo1.

o C baSic
0'sM Sina
V
Tkiv%

-tha. respolase eqs to

\ r - L ,tY%3

/9 vA LT I SWs ens-

6.~i Co,e g Convo +

Discrete-time signal expressed as a superposition of
weighted, delayed unit impulses I

x[-I] x[0] 1 x[2]

-lIOJ
fr2 TRANSPARENCY
x[0] 4.1
A general discrete-
x[O]8a[n] time signal expressed
as a superposition of
-.- e--.-0- n weighted, delayed unit
-1 0 I2
x[o]8[n]+x(I] 8[n -1] impulses.

X[1] x[1]8[n-1] + x [-I]8[n+ ]+.--

+X kr
-1 0 I 2
=2 x[k]8[n-k]
k= -c
x[-I]8[n+1] n
0-0 -- *-

-1 0 I 2

X[-] x [-2]8[n +2]

0--0-0 n
-1 0 1 2
Convolution Sum x[n] =E x[k] S[n-k]
k = -o

TRANSPARENCY
4.3 Linear System:
One interpretation of +0n
the convolution sum
for an LTI system. y [n] =E x[k] hk [n]
Each individual k = - 010
sequence value can be
viewed as triggering a
response; all the 5 [n - k] - h [n]
responses are added
to form the total
output. If Time-invariant:
hk [n] = h [n-k]

LTI: y[n]
+
=E
o0 x[k] h[n - k]I
Convolution Sum:
+0o

x[n] =E x[k] S[n-k]

k= -0ok TR
4.8
Co
con
dis
y [n] = x [k] h[n-k] =x[n] * h[n] sys
k= -o00 con
for
LT

Convolution Integral:
 for the superposition involved in generating y[n]. These echoes are displayed in
ure 2.3(b). By summing the two echoes for each value of n, we obtain y[n], whi
Example
shown in Figure 2.3(c).

h[n]
n K h n 11
y
Cos In a iii 11 II
as him 2
• • 0 2
• • n
o.sn I
0.5h n

x[n]
0.5
21
MAKE or • • T
0
• • • n
3
(a)
 (a)

Solution 0.5
0.5h[n]

• • T
0
T T • •
2 n

2h[n-1]

21
• • •0 11 •
2 3 n
(b)

251
Ir
y[n]

o.5T
• • 0 2 3
• n

(c)
 k h 4 k
n

2 1 0

K
y n 0
0ps 134

hÉi
K
O I

O
yen no
I O S N O
YEO
0 5 1 2 2.5 4 1
1 I
y 4 2
I 0.5 1 2 2.5
2
shifts y
1 2 2 n 3
y 3
 and, finally, h[n- k] for an arbitrary positive value of nand an arbitrary negative value
ider an input of n.and
x[n] Fromathis figure,
unit we note that
impulse for n < 0, h[n]
response there isgiven
no overlap
by between the nonzero
points in x[k] and h[n - k]. Thus, for n < 0, x[k]h[n- k] = 0 for all values of k, and
Ex. 2 hence, from eq. (2.6), we see that
x[n] = y[n] = 0, n < 0. For n 2:: 0,
anu[n],
x[k]h[n - k] = { ak,
h[n] = u[n], 0, 0 s;. k. s;. n . 0<⍺<1
otherwise

0 < a < 1. These signals are illustrated in Figure 2.5. Also, to help us in visualizing
calculating the convolution of thex[n]signals, 1
= anu[n] in Figure 2.6 we have depicted the signal
followed by h[- k], h[ -1- k], and h[1- k] (that is, h[n- k] for n = 0, -1, and+ 1)
finally, h[n- k] for an arbitrary positive value
........... of nand an arbitrary negative value
llliiiiilttttttttt
From this figure, we note that for n < 0,0 there (a)
is no overlap between
n the nonzero

s in x[k] and h[n - k]. Thus, for n < 0, x[k]h[n- k] = 0 for all values of k, and
e, from eq. (2.6), we see that y[n] = 0, n < 0. For n 2:: 0,

x[k]h[n - k] = { ak, 0 s;. k. s;. n .

0, otherwise
0 n
(b)
 I
-1 0 Linear Time-Invariant Systems k Chap. 2
84 (c) Linear Time-Invariant Systems
h[-k] Chap. 2

I
h[1-k]
x[k] a'u[k]

Solution ...... JiliiJiliil

Jliillliiil I....................
I
x[k] a'u[k]

................... .
.
n α u n
.............. llliiiiiJitttttttT
.............. llliiiiiJitttttttT ...
0

001
...
(b)
(d)
kk
k

(a)
Afn Ucn
0 k

........... Illllll
(a)

JIIIIII....................
IIIIIJHill. IIIJI... .........
.. ... .. ...
h[-k]
h[-k]

YEN o 9 0
IJiliiJiliil-1 0
n:O ••.
k
n 0
0
I
JiliiJiliil .................... .
(c) n k
4cm
E α 0 (e) k
(b)

Illllll.....Hill.I
0 h[1-k] k
(b)
h[n-k]

yen
no
•. •.IIIIII
... Jliillliiil
n
j
Hill.I...................
... .. .. ... .. .. .
01
0
-1 0
u u
(d) ... .. .. ... .. .. k
k
k

0 n LO -1 0
(f)
(c)
k

II
(c)

I
Figure 2.6 Graphical interpretation of the calculation of the convolution
h[1-k]
sum for Example 2.3.
......IJliillliiil
IIIIIJJIII I...................
IIIJI ....... .
h[1-k]

n:O ••.

yen
uh ... Jliillliiil I................... .
0
01 (e)
(d)
n
k
k
 Figure 2.5 The signals x[n] and h[n] in Example 2.3.

Thus, for n 2:: 0,

Solution y[n] = Lak,

n

k=O

and using the result of Problem 1.54 we can write this as

n 1 -an+!
y[n] = Lak = - - - for n 2:: 0.
k=O 1 -a
Discrete-Time LTI Systems: The Convolution Sum 85
Thus, for all n,
y[n] =
1 -a n+1) u[n]
( 1-a 1 an+!)
y[n] = _ a u[n].
( 1

The signal y[n] is sketched in Figure 2.7.

1-a

0 n
ws:

Ex. 3 x[n] = 2"u[ -n], (2.17

h[n] = u[n]. (2.18
Sec. 2.1 1
interval Discrete-Time LTI Systems: The Convolution Sum 89
no y D
1
1
2 x[k] = 2ku[ -k]
1 1 1
16
•
8

'l l
4
t
-2
I
-1
10 • • • • k

if
l l l
2

11 111 • r n
h[n-k]

• k•
snitt
IT
(a)
 lem considered
Figure in Example
2.11 2.5;sequences
(a) The (b) the resulting output
x[k] and h[n- the y[n].
signal
k] for convolution prob-
lem considered in Example 2.5; (b) the resulting output signal y[n].

Solution
The sequences
x[k]
x[k] and h[n- k] are plotted as functions of kin Figure 2.11(a). Note that
The sequences
x[k] is zero fork> 0 and
and
h[n-
is zero fork> 0 and
x[k]
k]
are plotted as functions of kin Figure 2.11(a). Note that
h[n- k]
is zero fork>
is zero
h[n- k] fork> n. WeWe
n. also
alsoobserve
observe that, regardlessofof
that, regardless
the value the
of n, the of
value sequence x[k]h[n-
n, the sequence k] always
x[k]h[n- hashas
k] always nonzero
nonzerosamples
samples along thek-axis.
along the k-axis.
When n :2: 0, x[k]h[n
When - k] has
n :2: 0, x[k]h[n nonzero
- k] samples
has nonzero in the
samples interval
in the 0. It follows
intervalk k :::;:::; 0. followsthat,
that,
for n :2: 0,for n :2: 0,
0 0

y[n] =
y[n] L0
L= x[k]h[n- = L 2k.
k] = L 2k.
x[k]h[n- k]
0
(2.19)
(2.19)
k=-X k=-X
k=-X k=-X
To evaluate the infinite sum in eq. (2.19), we may use the infinite sum formula,
To evaluate the infinite sum in eq. (2.19), we may use the infinite sum formula,
1
= --,
1 Lak
X

0 < lal < 1. (2.20)

k=O
Lak
X

k=O
1 -a
= --,
1 -a
0< lal < 1. (2.20)
Changing the variable of summation in eq. (2.19) from k tor = - k, we obtain

Changing the variable of summation in eq. (2.19) from k tor = - k, we obtain

1 = 2. (2.21)
1 - (112)
1 = 2. (2.21)
1 - (112)
 90 Linear Time-Invariant
Linear Time-Invariant Systems Systems
Chap. 2 Chap. 2
90 When n < 0, x[k]h[n - k] has1 nonzero samples for k :::::: n. ItSystems
Linear Time-Invariant follows that,Chap.
for 2
Solution
n < 0,
When n < 0, When n < -0, k]x[k]h[n
x[k]h[n has 1
1
- 2 k] has
nonzero nonzero
samples for
x[k]
k
= 2ku[ -k]
samples
:::::: n. for It follows
k :::::: n.that,
It follows for that, for
1 1
n < 0,
IL 1
8 4- k] has nonzero samples for k :::::: n. It follows that, for
II II
n < 0, When n < 0, 16 x[k]h[n
y[n] t= x[k]lz[n- kl = 2k. L (2.22)
n < 0,
y[n] =
•
II -2 -1
y[n] =
x[k]lz[n-
'L
k=-X 0II • • II • •k=-CC kII
klx[k]lz[n-
=
L kl =
2k. L 2k. L (2.22) (2.22)
II II
I + n, we can again make
By performing a change of variablek=-XI = - k and
L
then m =k=-CC
L
r
k=-X
y[n] =
k=-CC
x[k]lz[n- kl =
use of the infinite sum formula, eq. (2.20), to evaluate theh[n-k] sum2k. in eq. (2.22). The result (2.22)
isBy
1 1 1 1 1 m = I + n, we can again make
k=-X k=-CC
By performingtheperforming
afollowing
change of avariable
for change
n < 0: Iof=variable I = - k and then
- k and then m = I + n, we can again make
use of the infinite sum formula, eq. (2.20), 11to evaluate • the •sum in eq. (2.22). The result
111inm•eq.
use of the By
infinite sum formula, (2.20),
eq. of to evaluate
= - k the sum = k (2.22). Thecan result
.L: (1 )' .L -"' (1 (1 )- (1 )
performing a change variable I and n then I + n, we again make
is the.vrnl
is the following
use of for
following :c
=< 0: sum
the ninfinite
I= -11
for n < 0:
2 formula,
=
)m--Il
2eq. (2.20),
(a)= 2 to evaluate .L
:;:
2 the = 211 sum . 2in=eq.211+(2.22).
'· (2.23)
The result
11 111
(1(1 )' .L-"' (1(1)- (1(1))- .L2.11(1 )
m=O m=O
is the following for n < 0: 11 111
.L:
.vrnl =The complete (1
:c .vrnl =)'
2 =sequence-
.L:"'
:c

I= -11
2
.L(1 )' -"' (1 .L(1 )- (1 )
2 =
)m--Il
of y[n]= is 2
m=O
2 )m--Il
sketched
2
:;: = 2
in2Figure
11 =
:;:
211 .
2
(b).= =
2
111 211+
211 . 2 = (2.23)
'·
211+ '· (2.23)
:;: m=O
I= -11
.vrnl = .L:
:c m=O
2 = .L 2 .L
)m--Il m=O
= 2 2 y[n]= 211 . 2 = 211+ '· (2.23)
The
TheseThe
complete
examples
complete
sequence
illustrate
I=
of
sequencethe
-11
y[n] is
of
m=Ousefulness
sketched
1 2.11
y[n] is sketched
in Figure
2.11
of visualizing
in m=O Figure the calculation
(b).
(b). of the con-
volution sum graphically. Moreover,2 in addition to providing a useful way in which to
calculate The complete of
the response
These
sequence
examples illustratean LTI of y[n] is the
system, sketched
the usefulness
2.11 in Figure
convolutionof visualizing sum also (b).provides an extremely
the calculation of the con-
seuseful
examples illustratefortheLTI
representation usefulness
systems that of visualizing the calculation of the con-
volution sum graphically. Moreover, in allows
addition us to to examine
providing their properties
a useful wayiningreatwhich to
sum graphically.
These
detail. examples
In particular,
calculate the
Moreover,
response in illustrate
Section
of an LTI
in2.3
-3addition
thewe
-2 willtodescribe
usefulness
-1
system,
0 providing
of visualizing
some
2
the convolution
3a of useful
the sum the way
properties
n
also
in which
calculation ofto extremely
anthe
of convolution
provides con-