Communication Systems

Lecture #2: Review of the Fourier Series

Overview

† The Objectives of Today’s Lecture

„ Overview/Motivation for Fourier Theory
„ Review of the Fourier Series
„ Relationship to the Fourier Transform
Motivation

† If a system is linear, the response due to a sum of

signals is the sum of the responses to each individual
signal
† System analysis can be simplified by decomposing an
input signal into a sum of simpler signals
„ The system output can then be found as the sum of
the system responses to these simpler signals
† A physically meaningful way of decomposing signals is
to represent them as a sum (or integral) of sinusoids
„ Periodic signals – Fourier Series
„ Aperiodic signals – Fourier Transform
† Periodic signals can also be represented using the
Fourier Transform
† This gives rise to the idea of the frequency domain
Fourier Theory

† Two basic types of signals

„ Power signals
„ Energy signals
† We can represent a signal in time or in frequency
† Fourier Representations
„ Fourier Series
† Representation valid for all time if signal is periodic
(i.e., power signals)
† Representation is valid only over a certain interval for
aperiodic signals
„ Fourier Transform
† Applies directly to energy signals
† Requires introduction of the impulse for application to
power signals
Fourier Theory (cont.)

† Fourier Theory tells us that signals can be represented

as weighted sums (or integrals) of sinusoids.
† The “amount” of each sinusoid is equivalent to the
“frequency domain” information of a particular signal
† If the signal is periodic, the signal can be represented
as an infinite sum of sinusoids whose frequencies are
integer multiples of the fundamental frequency, fo.
† If a signal is aperiodic we can take the limit of the
Fourier Series as the period goes to infinity. The
result is the Fourier Transform
† The Fourier Transform doesn’t technically apply to
periodic signals.
„ However we can create a FT through the use of the
delta function
Motivation (revisited)

† Assume that I want to transmit a signal

through a linear system
† I know that the system output is
y (t ) = x (t ) ⊗ h (t )
† If I know the time domain description of
the signal, x(t) and the impulse response
of the channel h(t) can I mentally
determine y(t)?
x ( t ) = cos ( 200π t )
h ( t ) = sinc ( 50t )
y (t ) = ?
Motivation (cont.)

† Now assume that I know the frequency

domain description of the signal and the
channel.
Y ( f ) = X ( f )H ( f )
† Knowing X(f) and H(f) can I easily
determine Y(f)?

X ( f ) = {δ ( f − 100 ) + δ ( f + 100 )}
1
2
1 ⎛ f ⎞
H ( f ) = rect ⎜ ⎟
50 ⎝ 50 ⎠
Y( f )=?
Motivation (cont.)
† Many aspects of communication system
analysis are simplified through the use of the
frequency domain
† The Fourier Series represents any periodic
signal as an infinite sum of sinusoids. The
weights of the sinusoids provide “frequency
domain” information of the signal
† Unfortunately, the Fourier Series is not
applicable to energy signals (e.g., aperiodic
signals)
† Thus, while we will review Fourier Series and
the concept is important, we will typically use
a more general technique of converting time
domain signals into the frequency domain –
the Fourier Transform
Continuous Time Fourier Series
† We can represent a periodic signal x(t) with period T0
exactly by the sum of complex sinusoids
∞
x (t ) = ∑ cn e j 2π nf0t
1
where n =−∞ f0 =
to +T0 T0
1
cn = ∫ x ( t ) e− j 2π nf0t dt
T0 to

† The above integral must converge

† This is termed the Continuous Time Fourier Series or
simply the Fourier Series
† We can represent the relationship between x(t) and cn as
FS
x ( t ) ↔ cn
Example 2.1
†Let’s examine a square wave which
can be written as ⎛ 1⎞∞
x ( t ) = 2 ∑ rect ⎜ 2t − 2i − ⎟ −1
⎝ 2⎠
i =−∞

†Examining one period (TF = 1) 0<t<1

the Fourier Series can be written as
∞
x (t ) = ∑
n =−∞
cn e j 2π nf0t

to +T0 To / 2 T
1 1 1 o − j 2π nfot
∫ x (t ) e ∫ ∫
− j 2π nf o t − j 2π nf o t
cn = dt = e dt − e dt
To to
To 0
To To / 2
Example 2.1 (cont.)

†Continuing …
To / 2 T
1 1 o − j 2π nfot
∫ ∫
− j 2π nf o t
cn = e dt − e dt
To 0
To To / 2

=
1 1
To − j 2π nf o
{
e− j 2π nfoTo / 2 − 1 −
1 1
}
To − j 2π nf o
{
e − j 2π nfoTo − e− j 2π nfoTo / 2 }
=
1
j 2π n
{e − j 2π n − 2e− jπ n + 1}

†We can simplify this for n even and n

odd… ⎧ 0 n even
⎪
cn = ⎨ 2 j
⎪⎩− π n n odd
Example 2.1 (cont.)
† Since the terms for positive and negative n
can be combined using Euler’s Identity we
can write the equation for the original signal
as
∞
2 − 2 cos (π n )
x (t ) = ∑ sin ( 2π nf 0t )
n =1 πn

† Since the even terms are zero we could also

rewrite this as
∞
sin ( 2π ( 2n − 1) f 0t )
4 1
x (t ) = ∑
π n =1 ( 2n − 1)
Example 2.1 - Plots
First 10 terms † As we continue to add the
sinusoidal terms our
representation is closer and
closer to the original signal
over the interval of interest.
† When the number of terms
goes to infinity the
representation is exact

First 100 terms First 1000 terms

Table of Fourier Series Coefficients

Time Domain Function Fourier Series Coefficients

x ( t ) = e j 2π f o t cn = δ [ n − 1]

c n = {δ [ n + 1] − δ [ n − 1]}
j
x ( t ) = sin ( 2π f ot ) 2
c n = {δ [ n + 1 ] + δ [ n − 1 ]}
1
x ( t ) = cos ( 2π f o t ) 2
x (t ) = 1 cn = δ [ n ]
∞
⎛ t − nτ ⎞ T ⎛T ⎞
x ( t ) = ∑ rect ⎜ ⎟ cn = sinc ⎜ n⎟
n =−∞ ⎝ T ⎠ τ ⎝τ ⎠
∞
⎛ t − nτ ⎞ T ⎛T ⎞
x ( t ) = ∑ tri ⎜ ⎟ cn = sinc ⎜ 2
n⎟
n =−∞ ⎝ T ⎠ τ ⎝τ ⎠
Properties of the Fourier Series

† Important Properties of the Fourier Series

„ Linearity
† The FS of a sum of two functions is the sum of
the individual FS
„ Time-shifting
† Time-shift corresponds to phase shift in FS
coefficients
„ Frequency-shifting
† Frequency shift corresponds to multiplying the
time signal by a complex sinusoid
„ Time reversal
„ Time scaling
 Properties of the Fourier Series
† Time differentiation
„ Differentiating in time corresponds to multiplying the FS
coefficients by f
† Time integration
„ Integrating in time corresponds to dividing the FS
coefficients by f
† Multiplication-Convolution
„ Convolving two functions in time corresponds to a
convolution sum of Fourier Series coefficients
† Parseval’s Theorem – Avg. power is the same in both
domains
∞
1
∫ x ( t ) dt = ∑ cn
2 2

T0 T0 n =−∞

Average power of x(t)

Use of Tables
† In general we use properties and tables of pairs
to find new pairs.
Property
Conjugation x* (t )←⎯→
FS
c−n
*

Linearity α x (t ) + β y (t )←⎯→ αc + β cFS x

n
y
n
Time-shifting x (t − to )←⎯→
FS
e − j 2 πnf 0to
cn
Frequency-
shifting e j 2 πno f0t x (t )←⎯→
FS
cn−no
Time reversal x (−t )←⎯→
FS
c−n
Time-derivative d
{ x (t )} ←⎯→
FS
( j 2πnf 0 ) cn
dt
t
Time-integration ∫ x (τ ) d τ ←⎯→
FS 1
cn
Intro to Comm. Fall
−∞2007
j 2πnf 0
.
Properties (cont.)

Property

Time-scaling
⎧ n
⎪c = integer
(integer a)
FS
x ( at ) ↔ ⎨ n a a
⎪⎩ 0
TF = To else

Multiplication ∞
x (t ) y (t )←⎯→
FS
∑cc
q=−∞
y x
q n− q

Convolution
x (t ) ⊗ y (t )←⎯→
FS
T0 cnx cny
Conjugation
x* (t )←⎯→
FS
c−* n
Example 2.2
†Find the Continuous Time Fourier
Series of
x (t ) = cos (50πt − π / 4)

for T0 = To=1/fo=1/25

We can use FS 1 1
cos (2π f ot )↔ δ [ n −1] + δ [ n + 1]
2 2

and x (t − to )←⎯→ e
FS − j 2 πnf 0to
cn
Example 2.2 (cont.)

†Re-writing the original equation

x (t ) = cos (50πt − π / 4)
⎛ ⎛ π / 4 ⎟⎞⎞⎟
⎜
= cos ⎜50π ⎜⎜t −
⎜⎝ ⎜⎝ 50π ⎠⎟⎟⎠⎟⎟
⎛ ⎛ 1 ⎞⎞⎟
⎜
= cos ⎜50π ⎜⎜t − ⎟
⎜⎝ ⎜⎝ 200 ⎠⎟⎟⎠⎟⎟

− j 2 π n*25(1/ 200) ⎧
⎪1 1 ⎫⎪
cn = e ⎨ δ [ n −1] + δ [ n + 1]⎬
⎪
⎪
⎩2 2 ⎪⎪⎭
⎧⎪ 1 1 ⎫⎪
⎨ δ [ n −1] + δ [ n + 1]⎬
− j 2 πn / 8
=e
⎪⎩⎪ 2 2 ⎪⎭⎪
 Example 2.2 (cont.)
⎧⎪ 1 1 ⎫⎪
†Continuing cn = e − j 2 πn / 8
⎨ δ [ n −1] + δ [ n + 1]⎬
⎪⎩⎪ 2 2 ⎪
⎪
⎭
1 j 2π / 8 1
=e − j 2π / 8
δ [ n −1] + e δ [ n + 1]
2 2
− jπ / 4 1 jπ / 4 1
=e δ [ n −1] + e δ [ n + 1]
2 2
(1− j ) 1 (1 + j ) 1
= δ [ n −1] + δ [ n + 1]
2 2 2 2

(1− j ) 1 (1 + j ) 1
cos (50πt − π / 4)←⎯→
FS
δ [ n −1] + δ [ n + 1]
2 2 2 2
Energy Signals

† The Fourier Series applies to periodic

signals which are also power signals.
† However, we would like to analyze
both power signals and energy
signals.
† Thus, we need a more powerful
analysis tool.
† The Fourier Transform is the answer
The Fourier Transform

† Let’s examine the Fourier Series representation of a

periodic signal with fundamental frequency
∆f = f0 = 1/T0
∞
x (t ) = ∑
n =−∞
cn e j 2π n∆ft 1
cn = ∫ x ( t ) e − j 2π n∆ft dt
To To
† Substituting cn into the equation for x(t)
∞ ⎧⎪ 1 ⎫⎪ j 2π n∆ft
x ( t ) = ∑ ⎨ ∫ x (τ ) e − j 2π n∆f τ
dτ ⎬e
n =−∞ ⎪⎩ T0 T0 ⎪⎭
† Since the exact starting and stopping points for the
integration are arbitrary (provided that we integrate
over a period), let us define the integration period to
be
−T0 / 2 ≤ t ≤ T0 / 2
The Fourier Transform (cont.)
† Using this region of integration
∞ ⎧⎪ T0 / 2 ⎫⎪ j 2π n∆ft
x ( t ) = ∑ ⎨∆f ∫ x (τ ) e − j 2π n∆f τ
dτ ⎬e
n =−∞ ⎪⎩ −T0 / 2 ⎪⎭
† Now in the limit as T0 approaches infinity,
the periodic signal becomes aperiodic, ∆f
approaches the differential df, n∆f becomes
a continuous variable f and
⎧⎪ ∞ ⎪⎧ T0 / 2 ⎪⎫ j 2π n∆ft ⎪⎫
x ( t ) = lim ⎨ ∑ ⎨∆f ∫ x (τ ) e − j 2π n∆f τ
dτ ⎬e ⎬
TF →∞
⎪⎩ n =−∞ ⎪⎩ −T0 / 2 ⎪⎭ ⎭⎪
∞
⎧∞ ⎫ j 2π ft
= ∫ ⎨ ∫ x (τ ) e − j 2π f τ
dτ ⎬e df
−∞ ⎩ −∞ ⎭
The Fourier Transform
† We then define the Continuous Time
Fourier Transform as
∞
X(f )= ∫ x ( t ) e − j 2π ft dt
−∞

= F { x ( t )}
and the original signal can be written in
terms of the Fourier Transform as
∞
x (t ) = ∫ X ( f )e j 2π ft df
−∞

= F −1 { X ( f )}
This is typically called the inverse Fourier
Transform
The Frequency Domain
† The original signal x(t) is said to be in the time domain since its
argument represents time
† The Fourier Transform X(f) representation is said to be in the
frequency domain since its argument f represents frequency
† Notes:
„ Frequency is the reciprocal of time
„ The Fourier Transform is referred to as an analysis of the
signal x(t) since it extracts the components of x(t) at each
value of f
„ The Inverse Fourier Transform is referred to as synthesis
since it recombines the components X(f) to obtain the
original signal x(t)
„ The physical meaning of X(f) depends on the meaning of
x(t). If x(t) has units of volts, X(f) has units volts/Hz.
† Thus it represents how much of the overall voltage signal is
present at each frequency.
Summary
† Today we have begun our discussion of the
frequency domain by briefly reviewing the
Fourier Series
† The Fourier Series allows us to represent any
periodic signal (power signals) as a weighted
sum of sinusoids.
„ The weights associated with each sinusoid can be
interpreted as the frequency domain information
† The Fourier Transform is a generalization of the
Fourier Series that allows us to analyze aperiodic
signals (such as energy signals)
„ We will spend more time on this tool next lecture.