Module D: Fourier Transform of Continuous-Time Signals

We will now consider continuous-time signals that are not necessarily peri-
odic.
We start with a motivating example.
Consider an aperiodic signal x(t) that has finite duration, i.e., x(t) = 0 for
|t| > S and x(t) = 1 for |t|  S.
Clearly, x(t) is not a periodic signal.

We construct a periodic signal x̃(t) with period T :

P1
It is easy to see that x̃(t) := k= 1 x(t + kT ).

1
 Fourier Transform: Motivation

P1
Since x̃(t) = k= 1 x(t+kT ) is periodic, we may express x̃(t) using Fourier
Series:
1
X 2⇡
x̃(t) = ak ejk T t ,
k= 1
Z T Z S
1 2
j 2⇡ 1 j 2⇡
where ak = x̃(t)e T kt dt = e T kt dt
T T
2
T S

sin( 2⇡kS
T ) 2 sin(!0 Sk)
= = ,
⇡k T !0 k
2S 2⇡
a0 = , !0 = .
T T

T T
Note that over the interval [ 2 , 2 ], x̃(t) coincides with x(t). Therefore,
Z T Z 1
1 2
j 2⇡ 1 j 2⇡
ak = x̃(t)e T kt dt = x(t)e T kt dt.
T T
2
T 1

2
 Fourier Transform

What happens whern T increases?

sin( 2⇡kS
T ) 2 sin(!S)
ak = =
⇡k T !

3
 Fourier Transform

Define
Z 1
j!t
X(j!) = x(t)e dt,
1

then
Z T
1 2
j 2⇡
ak = x̃(t)e T kt dt
T T
Z 12
1 jk!0 t
= x(t)e dt
T 1
X(jk!0 ) 2⇡
= , !0 = .
T T
Substituting this in the synthesis equation, we get
1 1
1 X jk!0 t !0 X
x̃(t) = X(jk!0 )e = X(jk!0 )ejk!0 t
T 2⇡
k= 1 k= 1

4
 Fourier Transform

When the period T ! 1 (!0 ! 0), the periodic signal x̃(t) approaches x(t).
That is,

1 Z
1 X jk!0 t
1
1
x(t) = lim X(jk!0 )e !0 = X(j!)ej!t d!.
!0 !0 2⇡ 1 2⇡
k= 1

Hence, this is called the Synthesis Equation because we are gathering the Fourier
domain information to reconstruct the time signal.

The Analysis Equation, because we are analyzing the time signal in the Fourier
domain, is given by
Z 1
X(j!) = x(t)e j!t dt.
1

5
 Fourier Transform of an Aperiodic CT Signal

The Fourier Transform X(j!) is given by the

Analysis Equation:
Z 1
j!t
X(j!) = x(t)e dt.
1

and the inverse Fourier Transform is given by the

Synthesis Equation:
Z 1
1
x(t) = X(j!)ej!t d!.
2⇡ 1

X(j!) is called the spectrum of the signal and it represents the contribution of
frequency ! to the signal x(t).

6
 Example 1

at
Consider the signal x(t) = e u(t), for a > 0. Find its Fourier transform.

Z 1
j!t
X(j!) = x(t)e dt
1
Z 1
at j!t
= e u(t)e dt
1
Z 1
at j!t
= e dt
0
1 (a+j!)t
1
= e
a + j! 0
1
= .
a + j!

To visualize X(j!), we need to plot its magnitude and phase with respect to !
on separate plots. We will revisit this later.

7
 Example 2

Consider the unit impulse signal x(t) = (t). Find its Fourier transform.

The Fourier Transform is

Z 1
j!t
X(j!) = x(t)e dt
1
Z 1
j!t
= (t)e dt
1
= 1.

In other words, the spectrum of the impulse signal has equal contribution from
all frequencies.

8
 Example 3

Find the Fourier transform of x(t) which takes value 0 for |t| > S and x(t) = 1
for |t|  S.

9
 Example 4

Find the signal whose Fourier transform is given by:

(
1, |!|  W,
X(j!) =
0, |!| > W.

10
 Duality

F.T.

Dual
y (t ) Y ( jω )
F.T.

F.T.

Dual

y (t ) Y ( jω )

F.T.

11
 Properties of Fourier Transform: Duality

Duality: FT and IFT are very similar. Mathematically, for a signal x(t)

F T (F T (x(t))) = 2⇡x( !).

Example: We know that (t) ! 1. What is the IF T { (!)}?

1
– By duality: it is 2⇡ .
1 1
Example 2: We know that e t u(t) ! 1+j! . What is F T { 1+jt }?
1
– Using duality: F T { 1+jt } = 2⇡e! u( !).

12
 Existence of Fourier Transform

A CT signal x(t) has a Fourier Transform if all three of the following condi-
tions are satisfied.
1. The signal is absolutely integrable:
Z 1
|x(t)|dt < 1.
1

2. In any finite interval of time, x(t) has bounded variation, i.e., it only
have a finite number of maxima and minima during any finite interval
of time.
3. In any finite interval of time, there are only a finite number of discon-
tinuities and each of these discontinuities are finite.

The above conditions are called Dirichlet conditions.

The above conditions are only sufficient, not necessary.
An alternative sufficient condition is that the signal has finite energy, i.e., it
is square integrable: Z 1
|x(t)|2 dt < 1.
1

13
 Fourier Transform of a Periodic CT Signal

Do all periodic signals have Fourier Transforms?

We start backwards, start with a frequency domain signal and find its inverse
Fourier Transform.
Let the Fourier Transform of a signal x(t) be given by

X(j!) = 2⇡ (! !0 ).

Then,

x(t) =

Thus, we have
F.T.
ej!0 t ! 2⇡ (! !0 )

For a general periodic signal:

1
X
x(t) = ak ejk!0 t ,
k= 1

we have
1
X
X(j!) = 2⇡ak (! k!0 ).
k= 1

Thus, for a periodic signal, the FT consists of a sequence of impulse functions

at multiples of !0 with height 2⇡ak .

14
 Example 5

Find the Fourier transform of x(t) = cos(!0 t).

Since this a sinusoidal signal, we can use the synthesis equation to obtain the
Fourier series coefficients:
ej!0 t + e j!0 t
cos(!0 t) = ,
2
1
which implies a1 = a 1 = 2 and ak = 0 otherwise.

The Fourier transform is:

1
X
X(j!) = 2⇡ak (! k!0 ) = ⇡ (! + !0 ) + ⇡ (! !0 )
k= 1

15
 Example 6

P1
Find the Fourier transform of x(t) = n= 1 (t nT ).

16
 Properties of Continuous-time Fourier Transform

F.T.
Notation: x(t) ! X(j!)
We also write
X(j!) = F{x(t)}
and
1
x(t) = F {X(j!)}
as alternative notations.

Linearity: FT and IFT are both linear:

F.T.
↵x(t) + y(t) ! ↵X(j!) + Y (j!).

Time-Shifting:
j!t0
x(t t0 ) !e X(j!).

This holds as:

Z 1 Z 1
1 j!(t t0 ) 1 j!t0
x(t t0 ) = X(j!)e d! = (e X(j!)) ej!t d!.
2⇡ 1 2⇡ 1 | {z }
FT of x(t t0 )

Frequency-shift:

ej!0 t x(t) ! X(j(! !0 )).

17
Properties of Fourier Transform and Inverse Fourier
Transform cont.

Conjugate and Conjugate Symmetry:

x(t) ! X(j!) =) x⇤ (t) ! X ⇤ ( j!).

In particular for real-valued signals:

x(t) ! X(j!) =) x(t) ! X ⇤ ( j!).

This implies the Conjugate Symmetry Property: X ⇤ ( j!) = X(j!).

In other words, for any !, X(j!) has the same magnitude as X( j!), and
the phase of X(j!) is negative of the phase of X( j!).

If x(t) is even, show that X( j!) = X(j!).

If x(t) is real-valued and even, then X(j!) is real-valued and even.

If x(t) is real-valued and odd, then X(j!) is purely imaginary and odd.

18
 Example 7

a|t|
Verify the above statements for x(t) = e where a is a positive real number.

19
Properties of Fourier Transform and Inverse Fourier
Transform cont.

Di↵erentiation: Suppose that x(t) is a di↵erentiable signal. Then:

dx(t)
! j!X(j!).
dt

Time and Frequency Scaling:

1 !
x(at) ! X(j ).
|a| a
Implication: x( t) ! X( j!). Shrinking a time-domain signal expands

it in the frequency domain.

20
 Key CT FT Pairs

ej!0 t ! 2⇡ (! !0 )

at 1
e u(t) ! a+j! for a with Re(a) > 0

2 sin(!T1 )
rectT1 (t) ! ! = 2T1 sinc( T⇡1 !) where
(
1 |t|  T1 sin(⇡✓)
rectT1 (t) = , sinc(✓) =
0 otherwise ⇡✓

(t) !1

1
u(t) ! j! + ⇡ (!)
– We cannot use analysis equation.
at
– We see u(t) as lima!0 u(t)e
at 1
– u(t)e ! a+j! and:
1 a j! a j!
= 2 2
= 2
a + j! a +! a + !2 a2 + !2
j! 1
– lima!0 a2 +! 2 = j!
a
– For the first term: lima!0 a2 +! 2 = 1 for ! = 0 and otherwise, it is zero
R1 a
– On the other hand, 1 a2 +!2 d! = tan 1 !a |11 = ⇡
a
– Therefore, lima!0 a2 +! 2 = ⇡ (!)

21
Properties of Fourier Transform and Inverse Fourier
Transform: Example

Example: We know that rectT1 (t) ! 2T1 sinc( T⇡1 !). Using this, calculate
the FT of the following signal.
8
< 1 t 2 [ 1, 0]
>
x(t) = 1 t 2 (0, 1] .
>
:
0 otherwise

t
3 2 1 1 2 3

Figure 1: Plot of x(t).

22
 Properties of Fourier Transform and Inverse Fourier
Transform: Example

2
1
t
2 2
1
2

Solution: Note that x(t) = rect1 (t) 2 rect0.5 (t + 0.5) (why?).

X(j!) = F T {rect1 (t))} 2F T {rect0.5 (t + 0.5)}

! ! 1 !
= 2 sinc( ) 2ej 2 (2 sinc( ))
⇡ 2 2⇡
! j !2 !
= 2 sinc( ) 2e sinc( ).
⇡ 2⇡

23
 Properties of Fourier Transform cont.

Parseval’s Theorem:
Z 1 Z 1
2 1
|x(t)| dt = |X(j!)|2 d!
1 2⇡ 1

Convolution:
x(t) ⇤ y(t) ! X(j!)Y (j!)

24
 Properties of Fourier Transform cont.

Rt
Integration: Let y(t) = 1 x(⌧ )d⌧ . Then:
1
y(t) ! X(j!) + ⇡X(0) (!).
j!

Multiplication:
1
x(t)y(t) ! X(j!) ⇤ Y (j!)
2⇡

25
Properties of FT

26
 Fourier Transform and LTI Systems

We know:

ej!t H(j!)ej!t
h(t)

R1 j!t
where H(j!) = 1 h(t)e dt is the frequency response of the system.
Interestingly: h(t) ! H(j!)
Does H(j!) exist for all LTI systems?
Therefore:

ej!t d! H(j!)ej!t d!
h(t)

Hence:

1
R1 j!t 1
R1 j!t
x(t) = 2⇡ 1 X(j!)e d! 2⇡ 1 H(j!)X(j!)e d!
h(t)

As a result: y(t) = x(t) ⇤ h(t) ! X(j!)H(j!) = Y (j!)

27
CT LTI Systems Described by Di↵erential Equations

Di↵erential equations provide a bridge between math (engineering) and phys-

ical world
Almost any engineering (and even many of the economical) systems behavior
is modeled by ODE
R

x(t) C y(t)

i(t)
Example 1: RLC networks:
In this case: x(t) = RC dy
dt + y(t)

28
 LTI Systems Described by Di↵erential Equations

x(t) C y(t)

i(t)
Example 1: RLC networks:
In this case: x(t) = RC dy
dt + y(t)
1
Therefore: X(j!) = RC(j!)Y (j!) + Y (j!) ) H(j!) = 1+RC(j!) =
1 1
RC 1
RC +(j!)

1 1
Hence: h(t) = RC t
RC e u(t)

29
 LTI Systems Described by Di↵erential Equations

Let S be a stable LTI system described by

N
X M
dk y(t) X dk x(t)
ak = bk
dtk dtk
k=0 k=0

Apply the CTFT to both sides

( N ) (M )
X dk y(t) X dk x(t)
F ak =F bk .
dtk dtk
k=0 k=0

From the Linearity Property and Di↵erentiation Property

N
X M
X
k
ak (j!) Y (j!) = bk (j!)k X(j!)
k=0 k=0

By the Convolution Property, the frequency response is

PM
k=0 bk (j!)k
H(j!) = Y (j!)/X(j!) = PN
k
k=0 ak (j!)

Question: Can we get the system impulse response from Table? Yes,
through partial fraction!

30
 Partial Fraction

Suppose H(j!) is a rational function or ratio of polynomials

Apply Partial Fraction Expansion to write H(j!) in a form that allows
us to determine h(t) from Table 4.2
Example: Stable LTI system described by
d2 y(t) dy(t) dx(t)
2
+4 + 3y(t) = + 2x(t)
dt dt dt
Then

j! + 2
H(j!) =
(j!)2 + 4j! + 3
j! + 2
=
(j! + 1)(j! + 3)

Rewrite H(j!) as
A B
H(j!) = +
j! + 1 j! + 3

31
 Partial Fraction Example cont.

We got
A B
H(j!) = +
j! + 1 j! + 3
Put over a common denominator

A(j! + 3) + B(j! + 1)
H(j!) =
(j! + 1)(j! + 3)
(A + B)j! + (3A + B) j! + 2
= =
(j! + 1)(j! + 3) (j! + 1)(j! + 3)

Equate real and imaginary parts of the numerators:

Real part: 3A + B = 2
Imaginary part: A+B =1
Solving for A and B
1 1
A= and B = )
2 ✓ 2 ◆ ✓ ◆
1 1 1 1 1 1
H(j!) = + ) h(t) = e t u(t) + e 3t
u(t).
2 j! + 1 2 j! + 3 2 2

32
 Partial Fraction General Case

For convenience: consider j! as a variable v:

bn 1 v n 1 + bn 2 v n 2 + · · · + b1 v + b0
H(v) = n
v + an 1 v n 1 + an 2 v n 2 + · · · + a1 v + a0
bn 1 v n 1 + bn 2 v n 2 + · · · + b1 v + b0
=
(v p1 ) 1 (v p2 ) 2 · · · (v pr ) r

We write this as a sum of simpler polynomial ratios as

A11 A12 A1 1
H(v) = + 2
+ ··· +
(v p1 ) (v p1 ) (v p1 ) 1

A21 A22 A2 2
+ + + · · · +
(v p2 ) (v p2 )2 (v p2 ) 2

···
Ar1 Ar2 Ar r
+ + + · · · +
(v pr ) (v pr )2 (v pr ) r

tn 1 at 1
We can inverse CTFT each term using (n 1)! e u(t) ! (a+j!)n for a > 0.

33
 Example

Calculate the output for the input x(t) = e t u(t), for a system with the input-
output relationship:
d2 y(t) dx(t)
+ 3y(t) = + 2x(t).
dt2 dt

34
 Example

Suppose a causal stable continuous-time LTI system has frequence response

4 + j!
H(j!) = .
6 ! 2 + j5!
Determine its impulse response in time-domain.

35