Convolution and Fourier

Transform
Dr. Hariharan Muthusamy,
School of Mechatronic Engineering,
Universiti Malaysia Perlis.
Convolution
Consider a linear system whose behaviour is specified by the impulse
response h(n). To find the output signal be y(n), resolve the signal x(n) into
a weighted sum of impulses, and the linearity and time shift property of the
LTI system.
The above expression gives the response y(n) of the LTI system as a function
of the input signal x(n) and impulse response h(n) is called the convolution
sum.
Steps involved in finding out the convolution sum
1. Folding: Fold the signal h(k) about the origin, ie at k=0;
2. Shifting: Shift h(-k) to the right by n
o,
if n
o
is positive or shift h(-k) to the
left by n
o
, if n
o
is negative to obtain h(n
o
k).
3. Multiplication: Multiply x(k) by h(n
o
k) to obtain the product sequence
y
o
(k) = x(k) h(n
o
k)

=
=
k
k) x(k)h(n y(n)
4. Summation: Sum all the values of
the product sequence y
o
(k) to
obtain the value of the output at
time n = n
o
Find the convolution of two finite
duration sequences
otherwise
1 n 1
0
1
x(n)
s s

=
otherwise
1 n 1
0
1
h(n)
s s

=
2. Find the convolution of the two signals x(n) = u(n) and h(n) =
a
n
u(n), |a|<1, n>=0.
Fourier Transform
When the repetition period T becomes infinity, the waveform f(t) become
non-periodic, the separation between two adjacent harmonics will be zero.
Assuming f(t) is initially periodic, we have
dt e f(t) Tc ) F(jn
Let
, e c T Tf(t)
dt e f(t)
T
1
c
where
, e c f(t)
T/2
T/2
t jn
n o
t jn
n
T/2
T/2
t jn
n
t jn
n
o
o
o
o
}

= =
=
=
=

As T ->, the fundamental

frequency becomes infinitely
small, hence
o
-> d
dt e f(t) ) F(j
d e ) F(j
2
1
f(t)
e ) F(jn
2
1
e c T
T
1
f(t)
t j
t j
o
t jn
o
t jn
n
o
o
}
}

=
=
=
=

Properties of Fourier Transform

Linearity
FT obeys superposition and homogeneity principles.
According to this property FT of the sum of two signals is
equal to sum of FT of individual signals.
F[f
1
(t)]= F
1
(j) F[f
2
(t)]= F
2
(j)
F[a
1
f
1
(t)+b
1
f
2
(t)] = a
1
F
1
(j) + b
1
F
2
(j)
Where a
1
and b
1
are constants.
Time Shifting Time Reversal
F[f(t)]= F(j) F[f(t)]= F(j)
F[f(t-t
0
) = e
-jt0
F(j) F[f(-t)]= F(-j)
Frequency Shifting
According the this property multiplication of the signal e
jot
with f(t)
shifts the frequency spectrum by
o
F[f(t)]= F(j)
F[f(t) e
jot
]= F[j(-
o
)]
Time Scaling:
F[f(t)]= F(j) F[f(at)]= 1/|a| F(j /a), when a>1 f(at) is a compressed
version of the signal f(t) by a factor a.
When a < 1, f(at) is an expanded version of the signal x(t) by a factor
a
Differentiation in time
Differentiating a signal results in a multiplication of the Fourier
transform by j, F[f(t)]= F(j) then F[d/dt(f(t))]= j F(j)
Differentiation in Frequency
Multiplication of signal x(t) with time t results in differentiation of
frequency spectrum F(j)
F[f(t)]= F(j) then F[t f(t)]= j d/d F(j)
Time Integration
Integration of a signal results in a division of Fourier transform by j.
However to account for the dc or average value that can result from
integration, we must add the term F(0)()
F[f(t)] = F(j), then
) F(j ) (j
dt
f(t) d
F
n
n
n
=
(

F(0)() F(j(
j
1
f(() F
t
+ =
(

}

Conjugation
F[f(t)]= X(j), F[f*(t)] = F*(-j)
Fourier transform of complex and real functions
f(t) =f
R
(t)+jf
I
(t)
Fourier transform of f(t) is given by
F[f(t)]=
Auto-correlation
F[R
ff
()] = F[f(t) * f*(-t)] =
ff
(j) = |F(j|
2
Duality
F[f(t)]= F(j)
F[f(t)]= 2f(-j)
dt e f(t) ) F(j
t j
}

Convolution
The convolution of two signals results in the multiplication of their Fourier
transforms in the frequency domain.
F[f(t)*h(t)]= F(j) H(j)
The output of a system can be obtained from the convolution of input
signal and the system impulse response
}

= d ) )h(t f( y(t)
Fourier transform of Single Gate Function
otherwise
T/2 t T/2 for
0
1,
f(t)
< <

=
| |
|
.
|

\
|
=
|
.
|

\
|
=

=
=
=

}
}
2
T
Tsinc
2
T
2
T
sin
T.
e
j
1
dt .e 1
dt e f(t) ) F(j
T/2
T/2
t j
t j
T/2
T/2
t j

The amplitude spectrum is shown in Fig. At = 0, sin

c(T/2) = T, therefore F(j) at = 0 is equal to T.
At T/2 = n, sin c(T/2) = 0,
Fourier Transform of Rectangular Pulse
otherwise
T t 0 for
0
1,
f(t)
< <

=
| |
| |
(
(
(
(

|
.
|

\
|
|
.
|

\
|
=
(

=
=
=

}
}
2
T
2
T
sin
e T
2j
e e 2e
e e
j
e
e
j
1
dt .e 1
dt e f(t) ) F(j
T/2 j -
T/2 j T/2 j T/2 j
T/2 j T/2 j
T/2 j
T
0
t j
t j
T
0
t j

|
.
|

\
|
=

2
T
sinc Te
T/2 j

Fourier transform of a rectangular pulse 2 seconds

long with a magnitude of 10 volts
| |

c sin e 20
sin
e 20
2j
e e e
20
e e
j
e
10
j
1 e
10
j
e
10
dt 0.e 1
dt e f(t) ) F(j
j
j
j j j -
j j
j
j2 -
t j
t j
T/2
T/2
t j

=
=
(

=
(

=
(

=
=
=
}
}
Exponential Pulse
0 t for
0 t for
0
, e
f(t)
-at
<
>

=
|
.
|

\
|
Z
+
=
+
=
=
< = + =
+ =
=

}
}
} }
}
a
tan
a
1
j a
1
dt e
0 t for 0 f(t) since , dt e e 0
dt e f(t) dt e f(t)
dt e f(t) ) F(j
1
2 2
0
t ) j (a
0
t j at
0
t j
0
t j
t j

Fourier transform for the double exponential pulse

0 t for
0 t for
e
, e
f(t)
at
-at
<
>

=
2 2

) (

e
e
+
=
+
+

=
+ =
+ =
+ =
=
} }
} }
} }
}

a
j F
a 2
j a
1
j a
1
dt e dt e
dt e e dt e e
dt e f(t) dt e f(t)
dt e f(t) ) F(j
0
t ) j (a
-
t ) j (a
0
t j at
0
t j at
0
t j
0
t j
t j
Fourier transform of triangular pulse
f(t) = A (1+ 2t/T), -T/2<t<=0 f(t) = A(1-2t/T) 0<=t <=T/2
} } } }
} }
} }
} }

+ + =
|
.
|

\
|
+
|
.
|

\
|
+ =
+ =
= =
T/2
0
t j
0
T/2
t j
0
T/2
T/2
0
t j t j
t j
0
T/2
T/2
0
t j
0
T/2
T/2
0
t j t j
T/2
T/2
t j t j
dt e t
T
2A
dt e t
T
2A
dt e A dt e A
dt e t
T
2
1 A dt e t
T
2
1 A
dt e f(t) dt e f(t)
dt e f(t) dt e f(t) ) F(j

=
|
.
|

\
|
=
(
(
(
(

=
=
(

+
(

+ =
}
} }
} } } }

4
T
si n 2
T
4A
2
T
cos 1
T
4A
dt

si ni

2
T
si n
2
T
T
4A

2
T
si n
2A
dt 2t cos t
T
2A
dt coso 2 A
dt e t dt e t
T
2A
dt e dt e A F

obt ai n we i nt egr al s, t hir d and first t he i n t - t o t Changi ng

2
2
2
T/ 2
0
T/ 2
0
T/ 2
0
T/ 2
0
j
T/ 2
0
j
T/ 2
0
T/ 2
0
j j
Impulse Function (Unit Impulse)
The impulse function, which has an infinite amplitude and is infinitely narrow. This is
defined as (t) = 0 for all values except at t = 0
The Fourier transform or the impulse function (t) is obtained as
The frequency spectrum of the impulse function
(t) has a constant amplitude and extends
over positive and negative frequencies.
The inverse Fourier transform of the
unit impulse is given by
1 dt e (t) ) F(j
t j
= =
}

o
) (
2
1
.

e o
t t
=
Z =
= =
} }

Therefore
2
1
f(t)
t
2
1
d ) (
2
e
d e ) (
2
1
f(t)
t j
t j
Hence, the Fourier transform of the constant
function is an impulse at the origin with an area
equal to 2
SignumFunction
The signumfunction denoted by sgn(t) is defined by
0 t if
0 t if
0 t if
1
0
1
f(t)
>
=
<

j
2
j
1
j
1
j
e
j
e
dt e (1) dt e 1) (
dt e sgn(t) ) F(j
0
t j
0
t j
0
0
t j t j
t j
= + =
(

+
(

=
+ =
=

} }
}
Unit Step Function
The unit step function is obtained by suddenly closing a switch of a DC
circuit. For easier analysis, the waveform of the unit step function split
into two component waveforms.
The first waveform is similar to the signumfunction with half amplitude.
Therefore, the Fourier transform function is given by
F
1
(j) = (2/j) = 1/j
The second waveform is related to the unit impulse function and hence its
Fourier transform is given by
F
2
(j) = [2 ()] = ()
Therefore, the Fourier transform of the step function becomes
F(j) = F
1
(j) + F
2
(j)
=1/j + ()
Sinusoidal Functions
The Fourier transforms of the sinusoidal functions cos
o
t and sin
o
t are
obtained as given below.
F[cos
o
t] =
Using the transform pair
F[cos
o
t] = [(
o
) + (+
o
)]
Similarly
F[sin
o
t] = [( +
o
) + ( -
o
)]
(

+

2
e e
F
t j t j
o o

get we ), ( 2 e
o
t j
o

FOURIER TRANSFORM OF POWER AND ENERGY
SIGNALS
The average power of a signal x(t) over a single period (t
1
, t
1
+T)
is given by
Where x(t) is a complex periodic signal.
A signal f(t) is called a power signal, if the average power
expressed by
If x(t) is bounded, P

is finite. Every bounded and periodic signal is

a power signal. But it is true that a power signal is not
necessarily a bounded and periodic signal.
dt x(t)
T
1
P
2
T t
t
av
1
i
}
+
=
}

>
=
T
T
2
t
x
x(t)
2T
1
Lt P
Energy Signal
A signal x(t) is called an Energy signal, if its total energy over the interval
(- , ) is finite, that is
For a digital signal, the energy is defined by
As n ->, the energy of period signals becomes infinite, whereas the energy
of aperiodic pulse-signals have a finite value.
< =
}

>
dt x(t) Lt E
T
T
2
T
x
2
n
x(n) E

=
Determine the signal energy and signal power for (a). f(t) = e
-3|t|
, (b).
f(t) = e
-3t
| |
| |
3
1
e
6
2
dt e 2
dt e dt e
dt e E
0
6t
0
6t
0
6t
0
6t
2
| t | 3
= =
=
+ =
=

}
} }
}
The signal power P

= 0,
since E

is finite. Hence,
the signal f(t) is an energy
signal
( )
| |
T
T
T
T
T
e
6
2
=
=
=

}
}
6T
T
6t -
3t -
e
6
1
dt e
dt e E
As T -> , E
T
approaches infinity.
Its average power is
Hence, e
-3t
is neither an energy signal nor a power
signal.
=

= =

> >

T
e e
Lt E
T
Lt P
T T
T
T
T
8 2
1
6 6
Compute the signal energy for x(t) = e
-4t
u(t)
Find the Fourier transform of the signal f(t) shown in Fig.
f(t) = (A/T) t, for 0 < t < T, A , for T < t < 2T
| |
| |
T j2 T/2 j T/2 j T/2 j
2
T j2 T j
2
2T
T
t j
T
0
2
t j t j
2T
T
t j
T
0
t j -
e
A
j e e e
T
A
e
A
j 1 e
T
A
j
e
A
) j (
e
j
e
t
T
A
dt e A dt e t
T
A
E

+ =
+ =
(

+
(

)
`

)
`

=
+ =
} }
(

=
+ =
+ =



T j2 T/2 j
T j2 T/2 j
T j2 T/2 j
2
e
2
T
sinc e
j
A
e
A
j
2
T
sinc e
A
j
e
A
j
2
T
sin e
T
2A
j

Obtain the Fourier transform of the trapezoidal pulse is shown in Fig.

f(t) = At/(t
p
t
a
) + At
p
/(t
p
t
a
), for t
p
< t < -t
a
= A, for t
a
< t < t
a
= At
p
/(t
p
t
a
) - At/(t
p
t
a
), for t
p
< t < -t
a
Therefore,
} } }
} }

p
a
p
a
a
p
a
p
a
p
t
t
a p
t j
t
t
a p
t j
p
t
t
a p
t j
t
t
a p
t j
p
t
t
a p
t j
dt
t t
Ate
dt
t t
e t A
dt
t t
Ate
dt
t t
e At
dt
t t
Ate
) F(j

{ } { }
(

+
(

+ + +
)
`

=
(

+
(
(

)
`

+
)
`

+
(
(

)
`

)
`

+

=
+
(
(

+
(
(

} } }
} }












j
e e
A
j
e e
t t
At
e e
1
e e
1
j
e e
t
t t
A
j
e
A
j
e
j
e
t t
At
e
j
te e
j
te
t t
A
dt e A dt e dt e
t t
At
dt e t dt e t
t t
A
a a a a
p p
a a
a a
a
a
p
a
a
p
p
a
a
p
a
a
a
p
p
a
a
p
p
a
t j t j t j t j
a p
p
t j t j
2
t j t j
2
t j t j
a
a p
t
t
t j
t
t
t j
t
t
t j
a p
p
t
t
2
t j t j
t
t
2
t j t j
a p
t
t
t j
t
t
t
t
t j t j
a p
p
t
t
t
t
t j t j
a p
| |
| |
p a
2
p a
2
a p
p
a p
p
a
a a
a p
p
p
2
a
2
a
a
a p
t cos t cos
2
t cos t cos
2
t t
t
t t
t
1 t sin
2A
t 2sin
A
t 2sin
1
t t
At
t 2cos
1
t 2cos
1
t sin 2
t
t t
A

=
+
(

=
+