ELE635 Communication Systems

Chapter 1 : Introduction

1
 Chapter 1. Introduction

A course on communication systems is primarily concerned with transmission of

information.

Figure 1: Example applications of communication systems.

2
 Main topics

In this course, we will investigate techniques applicable to the analysis and synthesis
of analog communication systems. In particular, we will study
• mathematical tools (Fourier analysis)
• Basic signal transmission and reception techniques.
• Transmission media (communication channels)
• Performance analysis of standard communication systems.
• Implementation examples, AM, FM, TV broadcasting.

Distortion, Noise, ...

Input Output
Message Message
Input Output
TX Channel RX
Transducer Transducer
x(t) y(t)

Figure 2: A generic communication system.

3
 Diagram of a Communication System

In this course, we will investigate “Tx-Channel-Rx” as shown in Fig.2. The basic

functionality of the components are as follows
• Input Message: This is signal containing the raw information like human voice,
picture, etc.;
• Input Transducer: Convert the raw signal into an electrical signal (voltage,
current, etc.). The resulting signal is referred as baseband/input/message signal;
• Transmitter (TX): Convert the input signal (electronically, mechanically,
physically · · · ) to a format that best matches the channel characteristics;
• Channel: This is the physical medium over which the signal will travel. Here, in
the channel a lot of ugly, nasty things happen, such as distortion, noise,
interference, fading, etc.;
• Receiver (RX) and the output Transducer: These components undo/reverse the
modifications implemented by the transmitter (TX) and the input transducer,
respectively.

4
 Objective

Our goal is to transmit the information contained within the input signal x(t) to the
destination such that the information contained in y(t) is either identical (ideally) or
closely resembles the information in the input signal. We want to achieve this
objective most efficiently, economically, and effectively.

Some Remarks:
• In this course, we will work primarily with analog communication systems.
• Digital communication systems will be covered in the 4th year.
• Even digital communication systems will use ANALOG signals which in turn
rely on analog communication techniques. Hence, this is indeed an introductory
course of universal communication fundamentals.

5
 Section 1.1: Fourier Transform

Let x(t) be a (possibly non-periodic) function satisfying the Dirichlet conditions. We

can determine the Fourier transform of x(t) to be represented by X(f ) using the
Fourier integral:
∫ ∞
X(f ) = F[x(t)] = x(t)e−j2πf t dt (1)
−∞
We will also use the notation
x(t) =⇒ X(f )
to represent the origin/source of the Fourier transform X(f ). We can “recover” x(t)
from its Fourier transform X(f ) by using the inverse-Fourier transform operation:
∫ ∞
x(t) = F −1 [X(f )] = X(f )ej2πf t df (2)
−∞

Equivalently, we will use the symbolic representation:

x(t) ⇐⇒ X(f )

to show that x(t) and X(f ) form a Fourier transform pair.

6
 1
2

Magnitude
cos 2πf0 t −f0 0 f0 Spectrum

sin 2πf0 t π

.
−f0
2

0
.
f0

π
Phase
Spectrum
−
2

Figure 3: Fourier spectrum of sinusoid signals.

7
Frequently Used Fourier Transform Table:
x(t) X(f )
δ(t) 1
1 δ(f )
δ(t − t0 ) exp(−j2πf t0 )

2 [δ(f − f0 ) + δ(f + f0 )]
1
cos(2πf0 t)

2j [δ(f − f0 ) − δ(f + f0 )]
1
sin(2πf0 t)
exp(j2πf0 t) δ(f − f0 )
(t)
rect τ τ sinc(πf τ )
( )
1 f
sinc(2πBt) rect
( ) 2B
( 2B )
∆ fτ τ
sinc 2 πf τ
2
( )2
f
sinc2 (πBt) 1
B ∆ 2B

g(t − t0 ) G(f )e−j2πf t0

g(t)ej2πf0 t G(f − f0 )
∑∞ ∑∞ ( )
n=−∞ δ(t − nT0 ) δ f−
1 n
T0 n=−∞ T0

8
 Section 1.2: Special Functions

Section 1.2.1: Unit Impulse Function δ(t)

The unit impulse signal δ(t) is one of the most useful signals that will allow us to
formulate many operations frequently encountered in the analysis of LTI systems.
The unit impulse signal δ(t) (also known as the Dirac-delta function, named after the
quantum mechanist P.A.M. Dirac) has zero amplitude everywhere except at t = 0
where its magnitude is infinitely large such that the area underneath is unity. As
such δ(t) belongs to the category of generalized functions.

δ(t) 1
ǫ
ǫ→0

ǫ ǫ
0 t − t
2 2

(a) (b)

Figure 4: (a) Unit impulse and (b) its approximation.

9
δ(t) is defined as:
δ(t) = 0 t ̸= 0 (3)
∫ ∞
δ(t) dt = 1 (4)
−∞

δ(t) is normally considered as a limiting case obtained by considering the limit of a

rectangle function: ( )
1 t
δ(t) = lim rect
τ →0 τ τ
( )
where “rect τt ” is a rectangular pulse of τ -duration centered at the origin.

10
Important properties:
1. Even symmetry: δ(t) = δ(−t);
2. For a constant K and an arbitrary function x(t):
∫
Kδ(t)dt = K (5)

x(t)δ(t) = x(0)δ(t) x(t)δ(t − t0 ) = x(t0 )δ(t − t0 ) (6)

∫ ∫
x(t)δ(t)dt = x(0) x(t)δ(t − t0 )dt = x(t0 ) (7)

3. Convolution with a δ-function:

δ(t) ∗ x(t) = x(t) δ(t − t0 ) ∗ x(t) = x(t − t0 )

4. Properties of δ-function allow us to determine the Fourier transform of some

interesting functions
∫ ∞
F[δ(t)] = δ(t)e−j2πf t dt = 1 (8)
−∞
∫ ∞
F −1 [δ(f )] = δ(f )ej2πf t df = 1
−∞

11
Therefore, we have following Fourier transform pairs:
δ(t) → 1 1 → δ(f )

δ(t)
G(f ) = 1
1

0 t 0 f

(a) (b)

Figure 5: (a) Unit impulse and (b) its Fourier Transform.

g(t) = 1

G(f ) = δ(f )
1

0 t f

(a) (b)

Figure 6: (a) Constant (dc) and (b) its Fourier Transform.

12
 Section 1.2.2: Unit Rectangular and Triangular Pulse Functions

We will use the notation rect(t) to refer to the unit rectangular pulse function of unit
amplitude and unit width, centered at the origin


 1 |t| < 1
 2
rect(t) = 1
|t| = 1 (9)


2 2

0 |t| > 1
2
( )
Frequently, we will use the scaled version of the unit rectangular pulse as rect τt
where the scaling parameter τ will represent the width of the pulse centerer at the
origin.

13
 ! "
t
rect(t) rect
τ
1 1

-1/2 1/2 τ τ
0 t − t
2 2

(a) (b)

Figure 7: Rectangular pulses.

! "
t G(f) = Aτ · sinc(πfτ)
g(t) = A · rect
τ
Aτ
A

τ 0
τ t 0 1 f
−
2 2 1 τ
−
(a) τ (b)

Figure 8: (a) Rectangular pulse and (b) its Fourier spectrum.

14
 Unit triangular pulse function ∆(t)

The unit triangular pulse ∆(t) is a triangular shaped pulse of unit width and unit
amplitude centered at the origin:

 1 − 2|t| |t| < 12
∆(t) = (10)
 0 |t| > 21

We have FT transform pair:

( ) ( )
t Aτ πf τ
A·∆ =⇒ · sinc2
τ 2 2

! "
t
∆(t) A·∆
τ

1 A

-1/2 0 1/2 t −
τ τ t
2 2

(a) (b)

15
 Section 1.2.3: Sinc(x) Function

The “sinc” function is defined as:

sin(x) sin(πf τ )
sinc(x) = , sinc(πf τ ) = (11)
x πf τ

sinc(x) sinc(πfτ)

x
−2π −π 0 π 2π 3π 4π 0 1 f
τ

sinc(2W πt) 1
rect
!
f
"

2W 2W

1/2W

f
0 t −W 0 W
1/2W

16
Inspection of Equation (11) reveals the following key characteristics of the sinc
function:
• sinc(t) has even symmetry, i.e.., sinc(t) = sinc(−t).
• sinc(0) = 1. sinc(0) can be determined using the L’Hospita’s rule.
• sinc(t) = 0 when t = ±π, ±2π, ±3π, · · · .
• sinc(t) is the product of an oscillating signal sin(t) of period 2π and a
monotonically decreasing function (1/t). Therefore, sinc(t) is an oscillating
function with even symmetry and decreasing amplitude. It has a unit amplitude
at t = 0 and zero crossings at integer multiples of π.
Observe that
( )
1 t
lim rect = δ(t) in time-domain
τ →0 τ τ
lim sinc(πf τ ) = 1 in frequency domain
τ →0

17
 Section 1.3: Fourier Transform Properties

Table 1: Properties of Fourier Transform Operations

Operation g(t) G(f )

Linearity a1 g1 (t) + a2 g2 (t) a1 G1 (f ) + a2 G2 (f )
Scalar multiplication kg(t) kG(f )
Duality G(t) g(−f )
( )
1 f
Time scaling g(at) |a| G a

Time shifting g(t − t0 ) G(f )e−j2πf t0

Frequency shifting g(t)ej2πf0 t G(f − f0 )
Time convolution g1 (t) ∗ g2 (t) G1 (f ) · G2 (f )
Frequ. convolution g1 (t) · g2 (t) G1 (f ) ∗ G2 (f )

18
 Time Shifting

Time shifting brings a phase change as

g(t − t0 ) =⇒ G(f )e−j2πf t0

g(t) g(t − T0 ) g(t + T0 )

t t t
T0 −T0
0

δ(t − t0 ) x(t)
x(t − t0 )
b + t0
t0 t * a
b
t = a + t0 t

19
 Duality Property

If g(t) =⇒ G(f ), then G(t) =⇒ g(−f )

! "
t
G(f) = τ · sinc(πfτ)
g(t) = rect
τ τ

t
τ 0 τ 0 1 2 f
−
2 2 τ τ

τ · sinc(πtτ ) ! "
f
τ rect
τ

f
0 1 2 t τ 0 τ
−
τ τ 2 2

20
 Time-scaling Property

Time-scaling leads to inverse operation in the spectrum:

( )
1 f
g(at) =⇒ G
|a| a

α<1 g(αt)

a/α b/α

a b
g(βt)
β>1

21
 ! "
t
G(f) = τ · sinc(πfτ)
g(t) = rect
τ τ

t
τ 0 τ 0 1 2 f
− τ τ
2 2
! " ! "
1 t
g t = rect
2 2τ 2τ

t 0 1 1 f
−τ 0 τ 2τ τ

22
 Frequency Shift Property

We can continue with the computation of other FT pairs by using some FT properties
∫ ∞
F −1 [δ(f − f0 )] = δ(f − f0 )ej2πf t df = ej2πf0 t (12)
−∞
∫ ∞
F −1 [δ(f + f0 )] = δ(f + f0 )ej2πf t df = e−j2πf0 t
−∞
{ }
1 1 ( j2πf0 t )
F −1 [δ(f − f0 ) + δ(f + f0 )] = e + e−j2πf0 t = cos 2πf0 t
2 2
which results in the following Fourier transform pair
1
cos(2πf0 t) → [δ(f + f0 ) + δ(f − f0 )] (13)
2

Similarly, by first expressing sin(2πf0 t) as (ej2πf0 t − e−j2πf0 t )/2j, we can show that
1 [ j2πf0 t ] j
sin(2πf0 t) = e − e−j2πf0 t → [δ(f + f0 ) − δ(f − f0 )] (14)
2j 2
As expected the Fourier transforms of the trigonometric functions cos(·) and sin(·)
have the same magnitude but different phase spectra.

23
 Frequency Shifting/Modulation

Fourier Transform Property:

x(t) → X(f ) x(t)ej2πf0 t → X(f − f0 )

Proof:
∫ ∞ ∫ ∞
j2πf0 t −j2πf t
F[x(t)e j2πf0 t
] = x(t)e e dt = x(t)e−j2π(f −f0 )t dt (15)
−∞ −∞
= X(f − f0 )

Thus, the time-domain multiplication of x(t) with the complex exponential ej2πf0 t
will result in frequency shifting of X(f ) such that it will be centered on f0 . Similarly,
if were to multiply x(t) with e−j2πf0 t then X(f ) will be shifted in frequency and
centered on −f0 . We can now use this property together with the linearity of the
Fourier transform to determine the effect of multiplying a signal with a sinusoid.
Using Euler’s identity we first expand
1 [ j2πf0 t ]
cos(2πf0 t) = e + e−j2πf0 t
2
Then, using the frequency-shifting property together with the linearity of the Fourier

24
transform we can express the effect of multiplying a signal with a sinusoid:
∫
1 ∞ ( j2πf0 t −j2πf0 t
) −j2πf t
F[x(t) cos(2πf0 t)] = x(t) e +e e dt (16)
2 −∞
∫
1 ∞
= x(t)e−j2π(f −f0 )t + x(t)e−j2π(f +f0 )t dt
2 −∞
1
= [X(f − f0 ) + X(f + f0 )]
2
Similarly, the effect of multiplying x(t) with sin(2πf0 t) becomes
1
F[x(t) sin(2πf0 t)] = [X(f − f0 ) − X(f + f0 )] (17)
2j
1
= [X(f − f0 )e−jπ/2 + X(f + f0 )ejπ/2 ]
2
We will see that the spectra of both x(t) cos(2πf0 t) and x(t) sin(2πf0 t) are centered
at ±f0 , occupies the frequency band [f0 + B, f0 − B] (x(t) occupies [0, B]).
This concept of shifting the signal up/down in frequency band is known as
Modulation. The method of transmitting several signals simultaneously over a
channel by sharing its frequency band is known as frequency-division multiplexing
(FDM).

25
 Modulation with cos(2πf0 t)

1
x(t) cos(2πf0 t) → [X(f − f0 ) + X(f + f0 )]
2

26
 Modulation with sin(2πf0 t)

1
x(t) sin(2πf0 t) → [X(f − f0 )e−jπ/2 + X(f + f0 )ejπ/2 ]
2

27
 Section 1.4: Fourier Series

Given a time-limited signal x(t), we can construct a periodic signal by periodic

extension of x(t) as:

x(t)

t
0
!
xp (t) = x(t − nT0 )
n

t
0 T0

Figure 9: Construction of a periodic signal by periodic extension of x(t).

28
The obtained periodic waveform, denoted as xp (t), has period T0 and the
corresponding fundamental frequency f0 = 1/T0 (ω0 = 2πf0 ), we can expand this
waveform in a Fourier series as
∞
∑
xp (t) = Dn ej2πnf0 t (18)
n=−∞

where ∫
1 1
Dn = xp (t) e−j2πnf0 t dt and f0 = (19)
T0 T0 T0

We will refer to the expansion of the periodic waveform xp (t) in terms of

harmonically-related complex exponentials as shown in Equations (18) and (19) as
the complex Fourier series representation of xp (t). In general, the {Dn } coefficients
of the complex Fourier series expansion will be complex valued. We will frequently
use the notation:
xp (t) ⇐⇒ {Dn } (20)
to represent the equivalence between the time domain (the function xp (t) itself) and
the corresponding Fourier domain representation (the {Dn } coefficients).

29
 Fourier Series: Impact of Period T0

When the period T0 increases, the fundamental frequency f0 = 1/T0 reduces, leading
to denser line spectrum. When T0 turns to infinity, the line spectrum becomes a
continuous Fourier transform.

30
 FS in Trigonometric Form

In many cases (particularly, if the waveform xp (t) has even or odd symmetry), it is
convenient to expand xp (t) in terms of an equivalent trigonometric Fourier Series
given as:
∞
∑
xp (t) = a0 + (an cos nω0 t + bn sin nω0 t)
n=1
∫
1
a0 = xp (t)dt,
T0 T 0
∫
2
an = xp (t) cos nω0 tdt, t = 1, 2, · · ·
T0 T 0
∫
2
bn = xp (t) sin nω0 tdt, t = 1, 2, · · ·
T0 T 0
The integrals used in the evaluation of the complex or trigonometric Fourier series
coefficients can be taken over any interval of T0 duration as both the waveform xp (t)
and the complex/trigonometric basis functions of the Fourier Series are periodic with
period T0 .

31
Example 1 ∑
pulse train xp (t) = δ(t − nT0 ). Then xp is periodic, and its FS as
∫ T0 /2
1 1 1
Dn = xp (t)e−j2πnf0 t dt = (f0 = )
T0 −T0 /2 T0 T0

As a result, the pulse train can be written as:

∞
1 ∑ j2πnf0 t
xp (t) = e (21)
T0 n=−∞

Considering the Fourier Transform pair:

e(j2πf0 t) ⇐⇒ δ(f − f0 )

Taking Fourier transform to both sides, we obtain the FT of comb function:

∞
1 ∑
F(xp (t)) = δ (f − nf0 ) (22)
T0 n=−∞

32
 g(t)
1
... ... ... ...
−T0 0 T0 2T0 t −f0 0 2f0
f

Figure 10: (a) Impulse train and (b) its spectrum.

33
 Section 1.5: How to relate FS to FT?

Since x(t) is an aperiodic function, we can compute its FT as

∫ ∞
X(f ) = F[x(t)] = x(t)e−j2πf t dt (23)
−∞

By using Dn definition, we have

∫ T0 /2 ∫ T0 /2
1 1 1
Dn = xp (t) e−j2πnf0 t dt = x(t)e−j2π(n/T0 )t dt = X(nf0 )
T0 −T0 /2 T0 −T0 /2 T0
Furthermore, from linearity property,
[ ]
∑ ∑ [ j2πnf0 t ] ∑
F[xp (t)] = F Dn e j2πnf0 t
= Dn F e = Dn δ(f − nf0 )
n n n

Hence,
1 ∑
F[xp (t)] = X(nf0 )δ(f − nf0 )
T0 n
The FT of a periodic function is a set of δ-functions spaced at multiples of f0 . The
weight of each component is given by T10 X(nf0 ).

34
FT Can be shown to be a limiting case of the FS as T → ∞, xT approaches to a
non-periodic function. In this case, FS → FT, such that
∫ ∫
1
X(f ) = lim x(t)e−j2πf t dt x(t) = X(f )ej2πf t dt
T0 →∞ T 2π
0

while ∫
1
Dn = xp (t)e−j2πnf0 t dt
T0 T0
we can have
1
Dn = X(f )|f =nf0
T0

35
Example 2
rectangular pulse train with width τ and period T0 . Since
( )
t
A · rect → Aτ sinc (πf τ )
τ
we have
( )
Aτ Aτ Aτ nπτ
Dn = sinc(πf τ )|f =nf0 = sinc(π · nf0 · τ ) = sinc
T0 T0 T0 T0

w(t)

A
... ...
t
−T0 τ τ T0
−
2 2
Dn

n
0

36
Let’s exercise A = 3, τ = 1 msec, T0 = 2 msec.
Since ( )
t
A · rect → Aτ sinc(πf τ )
τ

1 Aτ ( nπ )
Dn = X(nf0 ) = · sinc(πnf0 τ ) = 1.5sinc
T0 T0 2

37
Example 3
A truncated sinusoid can be
( )
t
x(t) = 1.5 cos(1000πt) × rect
1 × 10−3
1.5
1.5 cos(1000πt) → [δ(f − f0 ) + δ(f + f0 )]
2
( )
t
rect → τ sinc (πf τ )
τ
where T0 = 2 msec, f0 = 500 and τ = 1 msec. We have Fourier transform of x(t) as
1.5τ 1.5τ
X(f ) = sinc[π(f − f0 )τ ] + sinc[π(f + f0 )τ ] (24)
2 2

38
Therefore,
1
Dn = X(f )|f =nf0
T0
1.5τ 1.5τ
= sinc[π(nf0 + f0 )τ ] + sinc[π(nf0 − f0 )τ ]
2T0 2T0
[ ] [ ]
1.5 π(n + 1)τ 1.5 π(n − 1)τ
= sinc + sinc
4 T0 4 T0
1.5 [π ] 1.5 [π ]
= sinc (n + 1) + sinc (n − 1)
4 2 4 2

39
Some Remarks:
• From {Dn }, we can have {|Dn |}-magnitude spectrum, and {∠Dn }– phase
spectrum. In the examples given, all Dn s are real values, real value corresponds
to 0 phase angle, and negative value corresponds to ±π angle.
• What is happening with negative frequencies? If we have a real value signal, it
has Fourier Series terms

Dn ej2πnf0 t and D−n e−j2πnf0 t

with D−n = Dn∗ . Let

Dn = |Dn |ejθn D−n = Dn∗ = |Dn |e−jθn

then,
[ j(2πnf0 t+θn ) −j(2πnf0 t+θn
]
e +e )
Dn ej2πnf0 t + D−n e−j2πnf0 t = 2|Dn |
2
= 2|Dn | cos(2πnf0 t + θn )

Hence, negative frequencies are essential, they are the product of the
mathematical operations. For real-valued periodic functions, we have FS

40
expansions
∞
∑ ∞
∑
j2πnf0 t
xp (t) = Dn e = C0 + Cn [cos(2πnf0 t + θn )] (25)
n=−∞ n=1

where 
 |D | n=0
0
Cn =
 2|Dn | n = 1, 2, · · ·
Hence,
– Two sided line spectrum → {Dn } or {|Dn |} and {∠Dn }
– One-sided line spectrum → {Cn } and {θn }
– One-sided RMS line spectrum → {Rn } and {αn }, where


 C0 = |D0 | n=0
 √
Rn = C
√n = 2|Dn | n = 1, 2, · · ·

 2

αn = θn = ∠Dn n = 0, 1, 2 · · ·

– |Rn | is the RMS value of the cosine component corresponding to the

frequency f = nf0 .

41
– The spectrum analyzer in the lab displays the one-sided RMS spectrum.

42
 T0=6, τ=2 T0=12, τ=2

0.3 0.3
1 1

Fourier Series Dn
Fourier Series D
0.8 0.2 0.8 0.2

0.6 0.6
0.1 0.1
0.4 0.4
0 0
0.2 0.2

0 −0.1 0 −0.1
−10 −5 0 5 10 −10 −5 0 5 10 −15 −10 −5 0 5 10 15 −20 −10 0 10 20

T0=12, τ=2 T0=12, τ=4

0.3 0.3
1 1
n

Fourier Series Dn
Fourier Series D

0.8 0.2 0.8 0.2

0.6 0.6
0.1 0.1
0.4 0.4
0 0
0.2 0.2

0 −0.1 0 −0.1
−10 −5 0 5 10 −20 −10 0 10 20 −15 −10 −5 0 5 10 15 −20 −10 0 10 20

(a) Period changes, width fixed (b) Width changes, period fixed

Figure 11: Fourier Series for Periodic Rectangular Pulse Train

43
 Section 1.6: Signal Energy and Signal Power

Given the fact that most signals of interest are rapidly varying functions of time t,
how can we measure the “size” of a given waveform?
For a DC signal such as g(t) = 5 V, the concept of the size of the signal is well
understood. Even for a sinusoidal signal such as g(t) = A cos 2πf0 t, we can
unambiguously refer to the size of signal by referencing the amplitude of the
sinusoidal oscillations. We have to introduce a new measure to compare the “size” of
an arbitrary waveform g(t).
• Signal energy:
∫ ∞
Eg = |g(t)|2 dt
−∞

– for real-valued g(t), |g(t)|2 = g 2 (t);

– for complex-valued g(t), |g(t)|2 = g(t) g ∗ (t), where * denotes complex
conjugate.
• There are a large class of important signals for which Eg is not finite power
signals (eg. sinusoid, infinite duration). For these we will use signal power as a

44
 measure of the “size” of the signal.
∫ T
1 2
Pg = lim |g(t)|2 dt
T →∞ T − T2

For example, periodic signals are typically power signals.

• Both Eg and Pg do not indicate the true/actual energy and power. Since
energy/power depends on signal AND the load. Hence, if g(t) is a voltage
waveform, then Eg and Pg represent UNIT energy/power deliverred across a 1-Ω
resistor.

Figure 12: Examples of signals: (a) signal with finite energy; (b) signal with finite
power.

45
 Parseval’s Theorem

Rayleigh’s Energy Theorem : given x(t) → X(f ), then the energy of x(t) is
∫ ∞ ∫ ∞ ∫ ∞
1
Ex = x2 (t) dt = |X(f )|2 df = |X(ω)|2 dω
−∞ −∞ 2π −∞

Given the periodic function g(t) with the FS expansion {Dn }, then the average
power is
∞
∑ ∑∞
1
g(t) = C0 + Cn cos(2πnf0 t + θn ) → Pg = C02 + Cn2
n=1
2 n=1
∞
∑ ∞
∑
g(t) = D0 + Dn e j2πnf0 t
→ Pg = |Dn |2
n=−∞(̸=0) n=−∞
∞
∑
for real g(t) → Pg = D02 + 2 |Dn |2
n=1
∞
∑
Rn → Pg = |Rn |2 (26)
n=0

46
 Examples

Find energy (for finite energy signals) and power (finite power signals) (from either
time domain or frequency domain):
(1) A cos(2πfc t)
( )
(2) A · rect τt
(3) A · sinc(πW t)
(4) A · sinc2 (πW t)
(t)
(5) A · ∆ τ
(6) Ae−λt (where λ > 0, 0 ≤ t ≤ T0 )
(7) Ae−λt (where λ > 0, 0 ≤ t)

47
 ∑∞
(1) Fourier Series for period function is xp (t) = n=−∞ Dn ej2πnf0 t .
A j2πfc t A −j2πfc t
x(t) = A cos(2πfc t) =
e + e
2 2
we have Fourier Series in exponential form, trigonometric form, and RMS form
respectively as
A
D1 = D−1 =, D0 = 0, Dn = 0 for |n| > 1
2
A
Cn = 2|Dn |, n = 1, 2, 3, · · · → C1 = 2 · = A
2
√ Cn A
Rn = 2|Dn |2 = √ n = 1, 2, 3, · · · → R1 = √
2 2

C1 = A √
D1 = A/2 R1 = A/ 2

-2 -1 0 1 2 0 1 2 3 0 1 2 3

P = 2D12 = A2 /2 1
P = C12 = A2 /2 P = R12 = A2 /2
2

48
(3) A · sinc(πW t)
From the Aid Sheet, we have following Fourier transform pair:
( )
1 f
sinc(2W πt) =⇒ rect
2W 2W

We can rewrite the target problem as (let W = 2B)

( )
A f
A · sinc(πW t) = A · sinc(2Bπt) =⇒ rect
2B 2B
( )
A 2 A2 A2
The Energy is: E = 2B · 2B = 2B = W

! "
1 f
rect
Part (3) sinc(2W πt) 2W 2W
1/2W

t f

0 −W 0 W
1/2W

! "
A f
Asinc(2Bπt) 2B
rect
2B
A/2B

f
0 t −B 0 B

49
(4) A · sinc2 (πW t)
From the Aid Sheet, we have following Fourier transform pair:

( ) 
2 f 1 − |f
W
|
|f | ≤ W
W sinc (πW t) =⇒ ∆ =
W  0 |f | > W

The target transform pair is

 ( )
( )  |f |
A f
A
W 1− W |f | ≤ W
Asinc2 (πW t) =⇒ G(f ) = ∆ =
W 2W  0 |f | > W

Then |G(f )|2 is shown in the figure below. The energy is the area of the curve:
∫ W( 2 ) ( 3 )
A2 f 2f 2A2 f f2
E = 2· 2 − + 1 df = − + f | W
0
W 0 W2 W W 2 3W 2 W
2A2 W 2A2
= =
W2 3 3W

50
 Part (4) ! "
A f
G(f) = ∆
W 2W |G(f)|2
! "
A 1
A/W G(f) = − f +1 A2 /W 2
W W

f f

−W 0 W −W W
"2
A2 A2 f 2
! ! "
2 1 2f
|G(f)| = 2 − f + 1 = 2 − +1
W W W W2 W

(5) Similar as triangle function in (4), the linear part (0 ≤ t ≤ τ /2) as

g(t) = − 2A
τ t + A. Energy is then
∫ τ /2 ∫ τ /2 ( 2 2
)
4A 4A
E = 2· g 2 (t)dt = 2 · 2
t 2
− t + A 2
dt
0 0 τ τ
( 2 3 2 2
)
4A t 4A t τ /2
= 2· − + A 2
t | 0
τ2 3 τ 2
A2 A2 τ
= τ −A τ +A τ =
2 2
3 3

51
(6) g1 (t) = Ae−λt , 0 ≤ t ≤ T0 , energy signal

∫ ∫ ( )
T0 ( )
−λt 2
T0
e−2λt
E = Ae dt = A2 e−2λt dt = A2 |T0 0
0 0 −2λ
A ( −2λt ) 0
2
A2 ( )
= e | T0 = 1 − e−2λT0
2λ 2λ
(7) g2 (t) = Ae−λt , 0 ≤ t < ∞, energy signal We can write

g2 (t) = lim g1 (t)

T0 →∞

Hence,
A2 ( −2λT0
) A2
E g2 = lim 1−e =
T0 →∞ 2λ 2λ

52
 Section 1.7: Concept of Bandwidth

BW is a measure of the extent of significant spectral components of the signal for

positive frequencies. Following example shows strictly band-limited signal and not
strictly band-limited signal.

Figure 13: Examples of signals: (a) strictly band-limited signal; (b) not strictly band-
limited signal.

53
 Baseband and Passband

• Baseband / low-pass: A baseband/LP waveform has a spectral magnitude for

frequencies in the vicinity of the origin (f = 0) and negligible elsewhere; example
voice signal.
• Bandpass: a bandpass waveform has a spectral magnitude that is non-zero for
frequencies in some band concentrated at ±fc , where fc >> 0. The spectral
magnitude is negligible elsewhere.

Xbp (f)
Xlp (f)

f f
0 −fc 0 fc

54
 Definition of BW

• If the signal is strictly band-limited then the definition of BW is obvious.

• If the signal is not strictly band-limited (as it is often the case), then we have to
be more careful about how we define the BW. One way: use the main-lobe
measure, for example:

55
– baseband waveform: BW = B
( )
t
x(t) = rect → X(f ) = τ sinc(πf τ )
τ
– bandpass waveform: BW = 2B
( )
t τ
y(t) = rect cos(2πfc t) → Y (f ) = {sinc [π(f − fc )τ ] + sinc[π(f + fc )τ ]}
τ 2

56
 3 dB Bandwidth

• An alternative measure of BW is the so-called 3 dB bandwidth, also known as

half-power bandwidth. Typical (but not all) lowpass signals have magnitude
spectra which are:
– maximum at the origin, and
– decrease monotonically with increasing frequency.
Similarly, typical bandpass signals have
– magnitude spectra maximum at the center frequency and
– decrease monotonically as the frequency variable moves away from fc

For such signals we measure the 3-dB bandwidth as the frequency where the
power spectrum drops 3-dB relative to its maximum (which is halve the
maximum signal power).

57
 |X(0)| |X (f )|
√
|X (0)|/ 2
0

|X (0)|2
3 dB |X (f )|2 measured in [dB]
|X (0)|2 /2

3-dB Bandwidth = f0

58
 |X (f )|2 measured in [dB]

|X (0)|2
3 dB

2
|X (0)| /2

0
fL fc fU

3-dB Bandwidth = fu − fL

[ ]
|X(0)| 2
10 log10 = 10 log10 [|X(0)|2 ] − 10 log10 2 (27)
2
= 10 log10 [|X(0)|2 ] − 3 dB (28)

• There are other definitions of BW. We will define and explain them as required.

59
 Bandwidth Examples

• A typical speech signal covers the frequency band [0, 3.5] kHz.
• A typical CD quality audio signal covers the frequency band [0, 20] kHz.
• Live orchestral music (including the overtones / higher order harmonics) covers
the frequency band well beyond [0, 20] kHz.
• Human auditory system (when young) can perceive signals in the frequency
range [0, 20] kHz.

Bandwidth of the product of two signals :x1 (t) and x2 (t) with bandwidth B1 Hz and
B2 Hz respectively,
• Bandwidth of x1 (t)x2 (t) is B1 + B2 Hz.
• Easy to validate using the frequency convolution property:

F[x1 (t)x2 (t] = X1 (f ) ∗ X2 (f )

60
 Bandwidth for AM/FM Radios

AM Radio:
• Program material for AM radio cover the frequency band [0, 5] kHz
• AM radio broadcast signals cover the frequency band [ fc − 5, fc + 5 ] kHz, where
fc is the carrier frequency assigned to the radio station.
• CRTC assigns AM carrier frequencies from [540 − 1600] kHz in 10 kHz
increments.

FM Radio:
• Program material for FM radio cover the frequency band [0, 15] kHz
• FM radio broadcast signals cover the frequency band [fc − 100, fc + 100] kHz,
where fc is the carrier frequency assigned to the radio station.
• CRTC assigns FM carrier frequencies from [88.1 − 107.9] MHz in 200 kHz
increments.

61
 Section 1.8: LTI Systems and Signal Transmission

The LTI (Linear, Time-Invariant) system is described by:

• Unit impulse response function: h(t);
• Frequency response function: H(f ) = F[h(t)]

System output can be computed using either time-domain or frequency-domain

techniques:
y(t) = x(t) ∗ h(t) = F −1 [X(f )H(f )]

LTI
System
x(t) y(t)
Input System Output

62
• We assume that the communication channel is modeled as a linear,
time-invariant (LTI) system. This assumption is usually made to simplify and to
introduce a degree of mathematical tractability to the analysis of the underlying
communications problem.
• It is empirically shown that indeed a large class of communication system can be
very accurately represented by appropriate LTI models.
• In this course we will mostly use the Fourier transform. The Fourier transform
will allow steady-state, sinusoidal analysis of the underlying signals and systems.

63
 Section 1.8.1. Transfer Functions H(f )

Input: sinusoid at frequency f0

Output: sinusoid at the same frequency f0 with amplitude and phase altered by
H(f ) evaluated at input frequency f0

LTI
x(t) System y(t)
A sin(ωo t + θ) A|H (fo )| sin(ωo t + θ + arg[H (fo )])

64
 Response of an Ideal Lowpass Filter (LPF)

Let’s consider an ideal lowpass filter (LPF) with

|H(f )| = 1 (|f | ≤ B); ∠H(f ) = −2πf t0

Therefore, the frequency response is then given by

( )
f
HLP (f ) = rect e−j2πf td (29)
2B

65
 Impulse Response of an Ideal Lowpass Filter

We can determine the corresponding impulse response function as

[ ( ) ]
f
hLP (t) = F −1 rect e−j2πf td = 2Bsinc[2πB(t − t0 )] (30)
2B

66
Observations:
• Noncausual: impulse response hLP (t) begins before the input is applied (at
t = 0). One practical approach is ĥLP (t) = hLP (t)u(t). If t0 is sufficiently large,
ĥLP (t) will be a good approximation of hLP (t).
• As B → ∞, |HLP (f )| → 1 and hLP (t) → δ(t − t0 ), such that y(t) = x(t − t0 ).
The output will be identical to the input delayed by t0 . However, for a signal
with B < ∞, the input pulse will be spread-out as a result of transmission
through a lowpass filter with finite B.

67
Empirically Measuring H(f )

68
69
70
In our analysis the most important elements are:
• Frequency response H(f )
• System bandwidth Bh
• Bandwidth of the input signal Bx , and
• Bx vs. Bh

y1 (t)

Bh1
x(t) t

y2 (t)
t H(f ) Bh2
τ
t

Adjustable Bandwidth Bh y3 (t)

Bh1 > Bh2 > Bh3
Bh3
t

71
 Section 1.8.2. Ideal Distortionless Transmission System

Distortionless Transmission
: is an idealized case. However, by studying the
properties of such an ideal system, we will have an important benchmark. It has
following input-output relation:

y(t) = K · x(t − td ) (31)

where K is the gain factor, and td is system delay.

Distortionless requires the output y(t) is a scaled (K) and time-shifted version
(0 → td ) of the input x(t). From above equation, we can have

h(t) = K · δ(t − td ) (32)

72
The corresponding Fourier transform is

H(f ) = F[h(t)] = K · e−j2πf td (33)

System output spectrum Y (f ) can be written as

Y (f ) = F[y(t)] = F[Kx(t − td )]
= KF[x(t − td )] = KX(f )e−j2πf td

73
H(f ) in (33) implies that

|H(f )| = K θh (f ) = −2πf td

The time delay resulting from the signal transmission through a system is the
negative of the slope of the system phase response; that is
1 dθh (f )
tdelay (f ) = td (f ) = − (34)
2π df
If the slope of θh (which is td (f )) is constant (a.k.a., θh is linear with respect to f ,
also called linear phase), all the frequency components are delayed by the same time
interval td , leading to “distortionless transmission”. Otherwise, the components of
different frequencies undergo different amounts of time delay, and consequently the
output waveform will not be a replica of the input waveform. For a signal
transmission to be distortionless, td (f ) should be a constant td over the frequency
band of interest.

74
 An Illustrative Example

The time delay of td of a sinusoid g(t) = sin(2πf t) is given by

sin[2πf (t − td )] = sin(2πf t − 2πf td )
Therefore, a delay of td manifests as a phase delay of 2πf t0 . This is a linear function
of f , meaning that higher frequency components must undergo proportionately larger
phase shifts to achieve the same time delay.

g1 (t) = sin(2πf1 t)

g2 (t) = sin(2πf2 t)
f2 = 2f1 θ2 = 2θ1

75
 Section 1.8.3: Linear and Nonlinear Distortion

Linear distortion
is defined as a change in amplitude or phase with no new
frequencies added.
Any frequency spectrum that passes through a LTI system can undergo only linear
distortions. In practice these distortions may emerge by a change in signal amplitude
and/or a change in signal phase for each frequency component.

76
 Non-Linear Distortion

For non-linear distortion, new frequency component appears. The output can be
written as
y(t) = g[x(t)] = a0 + a1 x(t) + a2 x2 (t) + · · · (35)

Example: Let x(t) = A cos(ω1 t) + B cos(ω2 t), and y(t) = g[x(t)] = x(t) + 2x2 (t)
such that

y(t) = [A cos(ω1 t) + B cos(ω2 t)] + 2[A cos(ω1 t) + B cos(ω2 t)]2

= A cos(ω1 t) + B cos(ω2 t) + 2A2 cos2 (ω1 t) + 2B 2 cos2 (ω2 t) + 4AB cos(ω1 t) cos(ω2 t)
= A cos(ω1 t) + B cos(ω2 t) + A2 + A2 cos(2ω1 t) + B 2 + B 2 cos(2ω2 t)
+2AB cos(ω1 + ω2 )t) + 2AB cos(ω1 − ω2 )t

77
 Non-linear distortion Example: Intermodulation distortion

From the result of y(t), we can see its frequency compnents:

f = 0, f1 , f2 , 2f1 , 2f2 , f1 ± f2
Let f1 = 4 kHz and f2 = 3 kHz, we can observe “spectrum-spreading distortion”!

78
 Non-Linear Distortion: Intermodulation distortion

• Input contains a single tone component at frequency f1

• If this single-tone signal is applied to a non-linear system, the output signal
contains harmonics of f1 , i.e., 2f1 , 3f1 , 4f1 , ...
• If the input to the non-linear system contains two separate single tone
components at frequencies f1 and f2 with f2 > f1 , sum (f2 + f1 ) and difference
(f2 − f1 ) frequencies are also produced in addition to the harmonics.
• However, it doesn’t stop there. Since there are harmonics of f1 and f2 , then
there will be sum and difference products between all of the harmonics and the
fundamentals and between each other. These are the intermodulation products
which are frequency components distinct from the harmonic components.

79
80
Odd Order Components

81
 Odd Order Components

• Of all the harmonics and intermodulation components produced, we are often

interested only in those which fall in the passband of our equipment.
• In the case of the intermodulation components, those closet to the fundamental
frequencies have a larger impact
• The third order components are the closest and also usually the highest in
amplitude. Because of this, they are usually of most concern. These components
are measured and their relative power is used in performance specifications.

82
83