Fourier Analysis

(Chapter 3)

These overheads were originally developed by Mark Fowler at Binghamton University,

State University of New York.
 Ch. 3: Fourier Series & Fourier Transform
(This chapter is for C-T case only) i.e., sinusoids
3.1 Representation in terms of frequency components
- Section 3.1 motivates the following important idea:
“signals can be built from sinusoids”
- Then Sections 3.2 – 3.3 take this idea further to precisely answer
How can we use sinusoids to build periodic signals?
- Then Section 3.4 – 3.7 take this idea even further to precisely answer
How can we use sinusoids to build more-general non-periodic signals?

LTI System
Q: Why all this attention to sinusoids?
A: Recall “sinusoidal analysis” in RLC circuits:
Fundamental Result: Sinusoid In ⇒ Sinusoid Out

It is easy to find out how sinusoids go through an LTI system!

Why Do We Study System Response to Sinusoids?
LTI: Linear, Time-Invariant

Q: How does a sinusoid go through an LTI System?

Consider:

x  t = A cos ω 0 tθ  h(t) y  t =?

To make this easier to answer we use Euler’s Formula:

A j  ω tθ  A − j  ω0 t θ 
0
x  t = A cos ω 0 tθ = e  e
2 2

The input is now viewed as the sum of two parts…

By linearity of the system we can find the response
to each part and then add them together.

So we now re-form our question …

Q: How does a complex sinusoid go through an LTI System?
A j  ω t θ 
Consider: 0
x 1  t = e h(t) y  t =?
2

With convolution as a tool we can now easily answer this question:

Plug in our input for x(t-τ)
∞
y  t = x 1  t ∗ht =∫−∞ x 1  t −τ  h τ  dτ
∞ A j [ ω  t−τ θ ] A j [ ωo t θ ] − jω o τ
∞
o
=∫−∞ e h τ dτ =∫−∞ e e h  τ  dτ
2 2
A j [ ωo t θ ] ∞ − jω τ
o
= e ∫
 h τ e dτ

2 −∞

Δ
Use rules for exponentials

= H  ωo  Evaluates to some
complex number that
Pull out part that does depends on h(t) and ωo
not depend on variable
of integration… So … the output is just this complex
Note that it is just x1(t)
sinusoidal input multiplied by some
complex number!
 A j  ω tθ 
o
So … y  t =H  ωo  e
2
Complex-valued

Let’s work this equation a bit more to get a more useful, but equivalent form …
j∠ H ω 
Because it is complex we can write H ω =∣H  ω ∣e o
o o


So using this gives: y t = ∣H  ωo ∣e
j ∠ H ω 
o
 A j ω o tθ 
2
e
j  ω tθ +∠ H  ω 

= ∣H ω o ∣
A
2  e o o

j  ω t θ +∠ H  ω o 
A o
y  t =∣H  ω o ∣ e
2
System shifts the
System scales the
phase by this!
amplitude by this!
 Now … we can re-visit our first question…
Q: How does a sinusoid go through an LTI System?
Consider:
x  t = A cos ω tθ 
0 h(t) y  t =?

This is equivalent to:

A j  ω 0 t θ  A − j  ω 0 t θ 
x  t = e  e y  t =?
2 2 h(t)
And due to linearity and the previous result used twice we have:

A j  ωo tθ +∠ H  ω o 
j −ω t−θ +∠ H −ω 
A o o
y  t =∣H  ωo ∣ e ∣H −ω o ∣ e
2 2

Later we’ll see that ∣H  ω ∣=∣H −ω ∣ ∠ H −ω =−∠ H  ω 

o o o o

[
1 j  ωo t θ +∠ H  ωo  1
]
− j  ω t θ +∠ H ω 
o o
So we get: y  t =∣H  ωo ∣A e  e
2  2
cos ωo tθ +∠ H  ωo 
So … How does a sinusoid go through an LTI System?
Consider:

x  t = A cos ω0 t θ  h(t) y  t = A∣H  ω o ∣cos  ω o tθ +∠ H  ωo  

The only thing an LTI system does to a sinusoid is

change its amplitude and its phase!

But what about when we have more complicated input signals???

We’ve already seen that we have to do convolution to solve that
case!
But … if we have a signal that is a sum of sinusoids then we could
use this easy result because of linearity and superposition!
x  t = A cos  ω t θ 
1 1 1
y  t = A1∣H  ω1 ∣cos  ω1 t θ 1 +∠ H  ω 1  
h(t)
A cos ω t θ  A2∣H  ω 2 ∣cos  ω 2 tθ 2 +∠ H  ω2 
2 2 2
So … breaking a signal into sinusoidal parts makes our job EASY!!
(As long we know what the H(ω) function looks like… that is Ch. 5’s
Problem)

What we look at in Ch. 3 is…

What kind of signals can we use this trick on?
Or in other words… What kinds of signals can we build by adding
together sinusoids?

Let ω0 be some given “fundamental” frequency.

Q: What can I build from building blocks that looks like:

A k cos  kω 0 θ k  ?

Only frequencies that are integer multiples of ω0

Ex.: ω0 = 30 rad/sec then consider 0, 30 60, 90, …

Ex. 3.1 (Motivation to answer this question!) My notation is a bit different than the book
A little experiment:
π π
x  t = A1 cos t A 4 cos  4t   A8 cos  8t 
3 2
ω 0 =1 4ω 0 8ω 0

A1 = 0.5
A4 = 1
A8 = 0.5

A1 = 1
A4 = 0.5
A8 = 0.5

A1 = 1
A4 = 1
A8 = 1
Q: How can we easily convey the information about this signal model?
A: Give a plot that shows the amplitude and phase at each frequency!
Amplitude Spectra Phase Spectra
All three cases are the same

Book uses degrees although it is

more correct to plot radians…
So we can write this sum of sinusoids like this:
N
x  t = ∑ A k cos kω 0 t θ k  (Ak Real)
k =1

Like (3.1) except mine is less general… I force: ω k =kω 0

So I’ve got a given set of “frequency components” and my model

consists of setting the amplitudes and phases to “desired” values

What if we also let k = 0, then we get: A0 cos 0θ 0 

Its frequency is 0 rad/sec (0 Hz) = constant, so just use

⇒ It is a “DC term” A0 & θ0 = 0

N This is called
x  t = A0  ∑ A k cos  kω 0 t θ k  “Trigonometric Form”
k =1 (Later we’ll let it have an
infinite # of terms)
“DC offset” (A0 & Ak Real)
⇒ Adding a DC Offset term just moves the whole signal up or down
 Trigonometric Form makes the most physical sense but mathematically we
often prefer the equivalent “complex exponential form”

  
N −N
jk ω t jk ω t
x  t =c 0  ∑ cke 0
 ∑ ck e 0

k =1 k =−1

c0 is Real
Positive freq. terms Negative freq. terms
ck is Complex

N
jk ω t This is called
x  t = ∑ ck e 0
“Complex Exponential
k =− N Form”

Q: How do we get the Complex Exponential Form from the Trigonometric

Form?
A: Euler’s Formula!
Each Term in the Trigonometric Form gives …
Two Terms in the Complex Exponential Form (Except the A0 term)
The details on how to get the Complex Exponential Form:
From direct application of Euler’s Formula to each term in the Trigonometric Form
of the Fourier Series we get:
j  kω tθ  − j kω t θ 
Ak e 0 k
 Ak e 0 k
k = 1, 2, 3, . . .
Ak cos  kω 0 tθ k =
2 ω0 > 0

= [ Ak
2
e
jθ
k
] e
jk ω t
0
 [
2
e e ]
Ak − jθ k − jk ω0 t

=
2 [ 
2
e ]
Ak jθ k jk ω 0 t Ak − jθ k jk −ω0 t
e e e [ ] Negative-Frequency Term

Positive-Frequency Term

Every physical sinusoid consists of …

one positive-frequency term and one negative-frequency term!
So for this complex exponential form we need a slightly different spectrum plot.
Must show both positive and negative frequencies
Called “Double-Sided Spectrum”
Example: Consider x  t =cos  t 0 . 5cos  4tπ / 3 cos  8tπ / 2
which is already in Trigonometric Form of the Fourier Series with ω0 = 1 :
A1 = 1 A4 = 0.5 A8 = 1 (all others are 0)
θ1 = 0 θ4 = π/3 θ8 = π/2
Using the results in the previous slides we can re-write this in Complex
Exponential Form of the FS as:

x  t = [ 0 . 5e 0 .5 e ][ 0 . 25 e ]
jt − jt jπ /3 j4t − jπ /3 − j4t
e 0 . 25 e e

[ 0. 5 e ]
jπ /2 j8t − jπ /2 − j8t
e 0 . 5 e e

c1 = 0.5 c4 = 0.25e jπ/3 c8 = 0.5e jπ/2

c-1 = 0.5 c-4 = 0.25e-jπ/3 c-8 = 0.5e -jπ/2 (all others are 0)
Both Forms Tell Us the Same Information… sometimes one
or the other is more convenient
 Single-Sided Spectra for Trigonometric Form of FS
Ak
Amplitude 1
0.5 etc.
Spectrum
1 2 3 4 5 6 7 8 9 10
frequency (rad/sec)
θk
Phase π/2
Spectrum etc.
π/3
x  t =cos  t 0 . 5cos  4tπ / 3 cos  8tπ / 2 1 2 3 4 5 6 7 8 9 10
frequency (rad/sec)

Double-Sided Spectra for Complex Exp. Form of FS

|ck|
Amplitude 1
Spectrum 0.5 etc.
0.25
Phase -8 -7 -6 -5 -4 -3 -2 -1 1 2 3 4 5 6 7 8 9 10
Spectrum frequency (rad/sec)
∠ck
x  t =[ 0 . 5 e jt 0. 5 e− jt ] π/2
π/3 etc.
[ 0 . 25 e jπ /3 e j4t 0 . 25e − jπ / 3 e− j4t ]
-π/3 1 2 3 4 5 6 7 8 9 10
[ 0 .5 e jπ / 2 e j8t 0 .5 e − jπ / 2 e− j8t ] -π/2 frequency (rad/sec)
What do these complex exponential terms look like?
Well … at any fixed time t the function ejωt is a complex number with unit
amplitude and angle ωt … so we can view it a a vector in the complex plane:

Im e jωt
ωt Re

Now… if we let the time variable “flow”… then this vector will rotate:
If ω > 0 then this vector rotates counter-clockwise
Im

If ω < 0 then this vector rotates clockwise

e jωt

Re

ω controls the angular rate…

it has units of rad/sec