CH 4CT Fourier Transform
CH 4CT Fourier Transform
Lecture 4 – Continuous-Time
Fourier Transform
1
Lecture 4 Topics
• Representation of Aperiodic Signals:
Continuous-Time Fourier Transforms
• Fourier Transform for Periodic Signals
• Properties of Continuous-Time Fourier
Transforms
• Convolution Property
• Multiplication Property
• Systems Characterized by LCCDEs
2
Lecture Overview
• Developed a representation of periodic signals as linear
combinations of complex exponentials
• Used the same representation to describe the effects of
LTI systems on signals
• Extend these concepts to signals that are not periodic
• We develop a representation, called the Fourier
Transform, in the form of an integral rather than a sum
(as in the periodic signals case)
• The inverse Fourier transform uses the Fourier transform
to represent the signal as a linear combination of
complex exponentials
3
CT Fourier Transform
4
Motivating Example: Square Wave
⎧⎪1, t < T1
x(t ) = ⎨
⎪⎩0, T1 < t < T / 2
increases
kept fixed
Discrete
frequency
points
become
denser in
ω as T
increases
Note: Envelope is
independent of T 5
Motivating Example: Square Wave
6
Motivating Example: Square Wave
7
Motivating Example: Square Wave
Each term in the summation for x(t ) is the area of a rectangle of height X ( jkω0 )e jkω0t and 8
width ω0 . As ω0 → 0, the summation converges to the integral of X ( jω ) e . jω t
Motivating Example: Square Wave
9
Kinds of Signals with CT FT
(1) It works also even if x(t) is infinite duration, but satisfies:
a) Finite energy
b) Dirichlet conditions
11
Convergence of FT
i Dirichlet conditions for convergence of Fourier Transform
1. x(t ) be absolutely integrable:
∞
∫ | x(t ) | dt < ∞
−∞
12
Example 4.1 – Decaying Exponential
∫ ∫ dt + ∫ e − at e− jωt dt
−a t − jωt at − jωt
X ( jω ) = e e dt = e e
−∞ −∞ 0
1 1 2a
= + = 2
a − jω a + jω a + ω 2
i In this case X ( jω ) is real.
14
Even Symmetry Even Symmetry
Example 4.3 – Unit Impulse
(a)
(b)
15
Example 4.4 – Square Pulse
i Consider the rectangular pulse signal:
⎧1, | t |< T1
x(t ) = ⎨
⎩0, | t |> T1
i The Fourier transform of this signal is:
T1
sin(ωT1 )
X ( jω ) = ∫e
− jωt
dt = 2
−T1
ω
16
Example 4.4 – Square Pulse in Time
Note the inverse relation between the two widths => Uncertainty principle
17
Example 4.4 – Square Pulse Convergence
i The rectangular pulse signal can be thought of as the limiting form
of a periodic square wave as the period becomes arbitrarily large.
i Would expect convergence of the synthesis equation would behave
like convergence for square wave.
i Consider the inverse Fourier transform for the rectangular pulse signal:
1
∞
sin (ωT1 )
xˆ (t ) =
2π ∫
−∞
2
ω
e jωt d ω
∫
2
x(t ) − xˆ (t ) dt = 0
−∞
19
Sinc Functions
i Functions of the form:
sin(ωT ) sin(Wt )
or
ω πt
occur often in Fourier analysis and in LTI system analysis
and are referred to as sinc functions, and are of the form :
sin(πθ )
sinc (θ ) =
πθ
20
Uncertainty Principle
21
Example
−at 2
x(t) = e — A Gaussian, important in
probability, optics, etc.
— periodic in t with
frequency ωo
This is exactly the Fourier series representation of a periodic signal. Thus the
Fourier transform of a periodic signal can be interpreted as a train of impulses 23
occurring at the set of harmonically related frequencies.
Example 4.6 – Square Wave
i Consider a periodic square wave of period T = ( 2π / ω0 ) of the form:
⎧⎪1, t < T1
x(t ) = ⎨
⎪⎩0, t > T1
i The Fourier Series coefficients and Fourier transform for this signal are:
sin ( kω0T1 )
ak =
πk
∞ 2 sin ( kω0T1 )
X ( jω ) = ∑
k =−∞ k
δ (ω − kω0 )
T = 4T1
24
Example 4.7
“Line spectrum”
1 jω0t 1 − jω0t
x(t ) = sin(ω0t ) = e − e ;
2j 2j
π π
X ( jω ) = δ (ω − ω0 ) − δ (ω + ω0 )
j j
i Fourier Series Coefficients:
a1 = 1/ (2 j ); a−1 = −1/ (2 j ); ak = 0, k ≠ ±1 25
Example 4.8 – Impulse Train
— Sampling function
k = −∞
Same function in
the frequency-domain!
Note: (period in t) T
⇔ (period in ω) 2π/T
Inverse relationship again!
26
Notation
x(t ) ←⎯
F
→ X ( jω )
Some Examples
1
= F {e − at u (t )}
a + jω
− at ⎧ 1 ⎫
−1
e u (t ) = F ⎨ ⎬
⎩ a + jω ⎭
− at 1
e u (t ) ←⎯→F
a + jω
27
Properties of the CT Fourier Transform
1) Linearity:
2) Time Shifting:
FT magnitude unchanged
28
Example 4.9 – Use of Linearity and Time Shift
i Consider the signal x(t ) which can be decomposed into the signals:
1
x(t ) = x1 (t − 2.5) + x2 (t − 2.5)
2
i The FTs of x1 (t ) and x2 (t ) are readily obtained as:
2 sin (ω / 2 ) 2 sin ( 3ω / 2 )
X 1 ( jω ) = , X 2 ( jω ) =
ω ω
i Using linearity and time-shifting properties, we can determine X ( jω ) as:
− jω 2.5 sin ( ω / 2 ) + 2 sin ( 3ω / 2 )
1 − jω 2.5 − jω 2.5
⎧ ⎫
X ( jω ) = X 1 ( jω )e + X 2 ( jω )e =e ⎨ ⎬
2 ⎩ ω ⎭
29
Properties of the CT Fourier Transform
i Conjugation property states that if:
F
x(t ) ←⎯ → X ( jω )
i then
F
x∗ (t ) ←⎯ → X ∗ ( − jω )
i Derivation:
∗
⎡ ∞
⎤ ∞
X ( jω ) = ⎢ ∫ x(t )e dt ⎥ = ∫ x∗ (t )e jωt dt
∗ − jωt
⎣ −∞ ⎦ −∞
i Replacing ω by − ω gives:
∞
X ∗ ( − jω ) = ∫ x∗ (t )e − jωt dt ←⎯
F
→ x∗ (t )
−∞
3) Conjugate Symmetry:
Even
Odd
Even
Odd
31
Example 4.10
i Consider the signal x(t ) = e − a|t| , a > 0. Use symmetry properties of Fourier
transform to evaluate Fourier transform of x(t ).
i From earlier examples we have:
1
e − at u (t ) ←⎯
F
→
a + jω
i Note that for t > 0, x(t ) = e − at u (t ) while for t < 0, x(t ) takes on mirror image
values, i.e.,
x(t ) = e − a|t| = e − at u (t ) + e at u (−t )
⎡ e − at u (t ) + e at u (−t ) ⎤
= 2⎢ ⎥ = 2 Even {e − at u (t )}
⎣ 2 ⎦
i Since e − at u (t ) is real-valued, the symmetry properties of Fourier transforms give
⎧ 1 ⎫
Even {e − at u (t )} ←⎯ F
→ Real ⎨ ⎬
⎩ a + jω ⎭
⎧ 1 ⎫ 2a
X ( jω ) = 2 Real ⎨ ⎬= 2
⎩ a + jω ⎭ a + ω
2
32
Properties of the CT Fourier Transform
dx(t ) FT
←⎯→ jω X ( jω )
dt
i if we integrate both sides of the FT synthesis eqn. we get:
t
1
∫ x(τ )dτ ←⎯→ X ( jω ) + π X (0)δ (ω )
FT
-∞
jω
i The impulse term reflects the average value that results from integration
33
Example 4.11 – Unit Step
i Determine the Fourier Transform of a unit step from the
transform of an impulse by using the integration property:
g (t ) = δ (t ) ←⎯→
FT
G ( jω ) = 1
i Noting that:
t
x(t ) = ∫ g (τ )dτ
−∞
36
Properties -- Time and Frequency Scaling
i If:
F
x(t ) ←⎯ → X ( jω )
i then:
F 1 ⎛ jω ⎞
x(at ) ←⎯ → X⎜ ⎟ ; a real constant
a ⎝ a ⎠
i Derivation:
∞
F { x(at )} = ∫ x(at )e− jωt dt
−∞
i Letting a = −1 gives:
F
x(−t ) ←⎯ → X ( − jω ) 37
Properties of the CT Fourier Transform
4) Time-Scaling:
c)
38
Properties of the CT Fourier Transform
6) Duality :
i symmetry between time and frequency transform pairs
39
Example 4.13
i Use duality to find the Fourier transform, G ( jω ), of the signal:
2
g (t ) =
1+ t2
i Earlier we showed that the signal x(t ) = e −2|t| had Fourier transform
2
X ( jω ) =
1+ ω2
i The synthesis equation for this Fourier transform pair is:
∞
⎛ 2 ⎞ jωt
1
∫−∞ ⎜⎝ 1 + ω 2 ⎟⎠ e dω
−2|t |
e =
2π
i Multiply by 2π and replace t by − t , giving
∞
⎛ 2 ⎞ − jωt
2π e ∫−∞ ⎜⎝ 1 + ω 2 ⎟⎠ e dω
−2|t |
=
2π
i The bracketed term is the Fourier transform of x(t ), giving
∞ ∞
1
∫ ∫
2 2
x(t ) dt = X ( jω ) dω
−∞
2π −∞
41
Example 4.14
i For each of the Fourier transforms shown in the
figure, evaluate the following time-domain expressions:
∞
∫
2
E= x(t ) dt
−∞
d
D= x(t ) |t =0
dt
i Evaluate E in the frequency domain, using Parseval;s
relation:
∞
1
∫
2
E= X ( jω ) dω
2π −∞
i which gives 5/8 for part (a) and 1 for part (b).
i To evaluate D in the frequency domain, use the differentiation
property giving:
d
g (t ) = x(t ) ←⎯→
FT
jω X ( jω ) = G ( jω )
dt
i Noting that
∞ ∞
D = g (0) = ∫ G ( jω ) dω = ∫
−∞ −∞
jω X ( jω ) dω
i which give zero for part (a) and - π for part (b) 42
Review -- CT Fourier Transform
Pair
(Analysis Equation)
(Synthesis Equation)
43
Convolution Property
Synthesis equation
for y(t)
44
Convolution – More Formal Derivation
⎡∞ ∞
⎤ − jωt
Y ( jω ) = FT { y (t )} = ∫ ⎢ ∫ x(τ )h(t − τ )dτ ⎥ e dt
−∞ ⎣ −∞ ⎦
∞
⎡∞ ⎤
= ∫ x(τ ) ⎢ ∫ h(t − τ )e − jωt dt ⎥ dτ
−∞ ⎣ −∞ ⎦
e − jωτ H ( jω )
∞ ∞
45
Convolution System Order
46
Example 4.15 -- Convolution
47
Frequency Response
impulse response
frequency response
Example #1:
48
Differentiator Example 4.16
i Consider a differentiator with input-output relationship
of the form
dx(t )
y (t ) =
dt
i From the differentiator property we get
Y ( jω ) = jω X ( jω )
i with frequency response
H ( jω ) = jω
49
Example #2 -- Differentiator
Differentiation property:
50
Integrator Example 4.17
i Consider an integrator specified as
t
y (t ) = ∫ x(τ )dτ
−∞
51
Example 4.18 – IR of Ideal LPF
Questions:
1) Is this a causal system? No.
2) What is h(0)?
52
Example #4 – Cascading Filter Operations
H(jω)
53
Ideal LPF Design Issues
• Ideal LPF impulse response is not zero for
negative time => ideal filter not causal
• Ideal filter not easy to approximate closely
• Ideal LPF impulse response may oscillate
• There exist filter design methods to
combat each of these design issues and
still yield a good LPF solution
54
Example 4.19
i Consider the response of an LTI system with impulse response
h(t ) = e − at u (t ), a > 0
i to the input signal
x(t ) = e − bt u (t ), b > 0
i The Fourier transforms of x(t ) and h(t ) are
1 1
X ( jω ) = , H ( jω ) =
b + jω a + jω
i Therefore
1
Y ( jω ) =
( a + jω )( b + jω )
i With b ≠ a we can do a partial fraction expansion for Y ( jω )
A B
Y ( jω ) = +
a + jω b + jω
i We can solve for A, B giving
1
A= = −B
b−a
i thereby giving
1 ⎡ 1 1 ⎤
Y ( jω ) = − What changes when b=a?
b − a ⎣ a + jω b + jω ⎥⎦
⎢
1
y (t ) = ⎡⎣e − at u (t ) − e − bt u (t ) ⎤⎦
b−a
55
Example
56
Example #2 Revisited
57
Examples #5 and #6
Example #5:
Example #6:
Multiplication Property
FT is highly symmetric,
— A consequence of Duality
60
Examples of Multiplication Property
⇓
For any s(t) ...
61
Example 4.21
i Let s (t ) be a signal whose spectrum S ( jω ) is shown in part (a) of the figure on this slide.
i Consider the signal:
p (t ) = cos(ω0t )
i with Fourier transform:
P ( jω ) = πδ (ω − ω0 ) + πδ (ω + ω0 )
i as shown in part (b) of the figure.
i The spectrum R ( jω ) of the multiplicative signal r (t ) = s (t ) p (t ) is of the form:
1
R ( jω ) = S ( jω ) ∗ P ( jω )
2π
1 1
= S ( j (ω − ω0 )) + S ( j (ω + ω0 ))
2 2
i as shown in part (c) of the figure (assuming ω0 > ω1 )
62
Example 4.22
g (t ) = r (t ) ⋅ p (t )
G ( jω ) = R ( jω ) * P ( jω )
63
Example 4.23
i Determine the Fourier transform of the signal:
sin ( t ) sin ( t / 2 )
x(t ) =
πt2
i Write x(t ) as the product of two sinc functions:
⎛ sin ( t ) ⎞ ⎛ sin ( t / 2 ) ⎞
x(t ) = π ⎜ ⎟⎜ ⎟
⎝ π t ⎠⎝ π t ⎠
i Using the multiplication property we get:
1 ⎧ sin ( t ) ⎫ ⎧ sin ( t / 2 ) ⎫
X ( jω ) = F⎨ ⎬ ∗ F ⎨ ⎬
2 ⎩ πt ⎭ ⎩ π t ⎭
i Since the FT of each sinc function is a rectangular
pulse, we can convolve those pulses to obtain X ( jω )
1 1
* =
−1 0 1 ω −1/2 0 1/2 ω 64
Frequency Selective Filtering
Y ( jω ) = δ (ω − ωc ) ∗ X ( jω ) = X ( j (ω − ωc ))
F ( jω ) = W ( j (ω + ωc ))
65
More Complicated Example
Y ( jω ) = δ (ω − ωc ) ∗ X ( jω ) = X ( j (ω − ωc ))
F ( jω ) = W ( j (ω + ωc ))
66
Frequency Selective Filtering
Y ( jω ) = δ (ω − ωc ) ∗ X ( jω ) = X ( j (ω − ωc ))
F ( jω ) = W ( j (ω + ωc ))
68
FT Pairs
69
LTI Systems Described by LCCDE’s
70
LCCDE Example 4.24
i Consider a stable LTI system characterized
by the differential equation:
dy (t )
+ ay (t ) = x(t ), a > 0
dt
Y ( jω ) 1
H ( jω ) = =
X ( jω ) a + jω
h(t ) = e − at u (t )
71
LCCDE Example 4.25
i Consider a stable LTI system characterized by
the differential equation:
d 2 y (t ) dy (t ) dx(t )
2
+ 4 + 3 y (t ) = + 2 x(t )
dt dt dt
Y ( jω ) jω + 2
H ( jω ) = =
X ( jω ) ( jω )2 + 4( jω ) + 3
jω + 2
=
( jω + 1)( jω + 3)
i Using the method of partial fraction expansion, we get:
1/ 2 1/ 2
H ( jω ) = +
jω + 1 jω + 3
1 1
h(t ) = e −t u (t ) + e −3t u (t )
2 2
72
LCCDE Example 4.26
i Consider the system of Example 4.25 with frequency response:
jω + 2
H ( jω ) =
( jω + 1)( jω + 3)
i and with input:
x(t ) = e − t u (t )
i The output is then of the form:
⎡ jω + 2 ⎤⎡ 1 ⎤
Y ( jω ) = H ( jω ) X ( jω ) = ⎢ ⎥⎢ ⎥
⎣ ( jω + 1)( jω + 3 ) ⎦ ⎣ jω + 1 ⎦
jω + 2 A11 A12 A21
= = + +
( jω + 1) ( jω + 3) jω + 1 ( jω + 1) jω + 3
2 2
∞
X ( jω ) = ∫
−∞
x(t )e − jωt d ω Analysis Equation
74
Lecture 4 Lessons
i Fourier Transform Pairs:
1
x(t ) = e − at u (t ) ←⎯
→ = X ( jω ) One-Sided Decaying Exponential
a + jω
2a
= X ( jω )
−a t
x(t ) = e ←⎯ → 2 Two-Sided Decaying Exponential
a + ω2
→1 = X ( jω )
x(t ) = δ (t ) ←⎯ Impulse Function
⎧⎪1, t < T1 sin(ωT1 )
x(t ) = ⎨ ←⎯
→2 = X ( jω ) Pulse in Time
⎪⎩0, t > T1 ω
⎧⎪1, ω < W sin(ωt )
X ( jω ) = ⎨ ←⎯
→ = x(t ) Pulse in Frequency
⎪⎩ 0, ω > W π t
ω2
π −
= X ( jω )
2
x(t ) = e − at ←⎯
→1 = e 4a
Gaussian Signal
a
→ πδ (ω − ω0 ) + πδ (ω + ω0 ) = X ( jω ) Cosine
x(t ) = cos(ω0t ) ←⎯
π π
x(t ) = sin(ω0t ) ←⎯→ δ (ω − ω0 ) − δ (ω + ω0 ) = X ( jω ) Sine
j j
1
x(t ) = u (t ) ←⎯→ + πδ (ω ) = X ( jω ) Step Function
jω
75
Lecture 4 Summary
• Developed Fourier transform representation for
CT signals
• Examined the properties of the Fourier transform
• Derived Fourier transfrom for aperiodic signals
from the Fourier series representation for
periodic signals
• Special emphasis on the convolution and
multiplication properties of Fourier transforms
• Special emphasis on defining and using
Frequency Response properties
76
Review Exercise - Prob. 4.2
Solution
−2 0 2 t
78
Review Exercise - Prob. 4.4
Solution
79
Review Exercise - Prob. 4.25
Solution
81
Review Exercise - Prob. 4.27
X ( jω )
Solution
82
Review Exercise - Prob. 4.27
83