EE411

Principles of Communication

1. Introduction and Basic terminology of communication

In this course, we are going to learn the basic signal processing used in communication. But
before doing so, we must have an appropriate understanding of what is the purpose of
communication.
Communication can be defined as the process of transmitting "reliably" (or as reliably as
possible) information from one point (the source of information) to another (the destination).
This can be a transmission from one point to another distinct point (such as microwave link)
or the source and destination can be located at the same place. This is what happens when a
processor writes data in memory.
A typical block diagram of a communication system is shown in Figure 1- 1. The word
"Channel" means the totality of the media and apparatus used for the transmission of the
information. We will encounter this word quite often in this course and in future courses. It is
a catch-all word used to describe many different things.

Figure 1- 1 Block diagram of a communication system

In information theory, this word (channel) means the mathematical mapping between the
source and the destination. It is also used to describe the allocated band of frequencies given
to a particular transmitting station. We will discover other meanings as we progress in the
course. In many cases, we can use the word channel to describe the physical mean of
transmission such as coaxial cables, fiber optic cables, etc.
The system shown in Figure 1- 1 is called a "simplex" communication system. In many cases,
a transmitter and a receiver are located at the same place. If two such systems share the same
communication mean, the system is called "half-duplex". In this case, we must have some
kind of switching in order to connect the transmitter at one location with the receiver at the
other location. An example of such system is the analogue telephone system. The handset
contains a microphone and an earphone. This later is disconnected while the phone user is
talking. There are also many "push to talk" systems that are half-duplex. On the other hand, if
they use two different means, the system is called "full-duplex".

Figure 1- 2 Simplex communication system

 Figure 1- 3 Half-duplex communication system

Figure 1- 4 Full duplex communication system

The above communication systems are point to point ones. There exist also "broadcast"
systems where there is only one transmitter and many receivers. This is the case of the actual
radio and television broadcast. Finally, there are communication systems where the
transmitters and receivers are located at nodes of a network. The network can be local (LAN:
Local Area Network) or it can cover a wide geographic region (WAN: Wide Area Network
such as the Internet). Networks can be also classified according to the rules of connection.
The telephone network connects two users for the whole duration of the communication while
a packet switching network shares a communication link between many users (When the link
is free, it is allocated to the users).
2. Review of signals and systems

2.1 Basic definitions

When we want to transmit information, we have to carry it using signals. Signals can
represent the atmospheric pressure variations picked by a microphone in the case of speech
communication. This signal is further transformed to an electrical signal. The microphone is
the transducer that makes this transformation. Other type of signals can represent the light
intensity variation in a black and white picture. So, we are going to call a "signal" a function
that carries information. This function can be a mapping between a continuous variable (time)
and a real or complex number.
x : t ∈ ℝ ֏ x(t ) ∈ ℂ (0.1)
Equation (0.1) can represent a signal like the speech signal. A signal can also be
represented by a sequence of real or complex numbers.
x : n ∈ ℤ ֏ x ( n) ∈ ℂ (0.2)
The above sequence can represent a signal generated inside a computer or a sequence of
samples from an analogue signal. A black and white picture can be represented by a real
valued function of two variables (x, y) that represent the position of the pixel in the image. In
this course, we will consider essentially signals of the type represented by (0.1). Some
important signals are:
The Heaviside unit step function:
0 t < 0
1

u (t ) =  t =0 (0.3)
 2
 1 t > 0
The rotating phasor:
φ (t ) = A exp (ω0t + θ ) (0.4)
This signal can be represented graphically in the complex plane by the rotating arrow shown
in Figure 2- 1. A is the amplitude (modulus) of the phasor, ω0 is the speed of rotation in
radians/s and θ is the initial phase (at t = 0). The sine wave signal is its projection on the real
axis of the phasor.
x(t ) = Re [φ (t )] = A cos (ω0t + θ ) (0.5)
 Figure 2- 1 Rotating phasor and sine wave in the complex plane
6

-0.5 -0.4 -0.3 -0.2 -0.1 0 0.1 0.2 0.3 0.4

-2

-4

-6

Figure 2- 2 Sine wave in the time domain

The sine wave is also related to the phasor by the following relation:
φ (t ) + φ ∗ (t )
x(t ) = (0.6)
2
This is represented graphically by Figure 2- 3. The signal φ*(t) is the complex conjugate of the
signal φ(t). It has the same amplitude, an opposite initial phase and it rotates in the opposite
direction: we say that it has a negative frequency.
It is evident that three numbers are sufficient to characterize completely the phasor. They are:
ω0, A and θ. We can represent this information graphically. Figure 2- 4 represents the
"spectrum" of the phasor φ(t) or the single sided spectrum of the sine wave x(t).
 Figure 2- 3

Figure 2- 4 Spectrum of a phasor

The above representation becomes interesting when a signal is the sum of many phasors.
N
x(t ) = ∑ Ak exp (ωk t + θ k ) (0.7)
k =1

The signal represented by (0.7) has N different amplitudes Ak and N different phases θk
located at the frequencies ωk.
 Figure 2- 5 Two sided spectrum of the signal x(t)
This representation of the signal x(t) presupposes that the amplitudes Ak and the phases θk are
functions of the frequency ω. In this case, for the signal x(t) to be real, its amplitude spectrum
must be an even function of the frequency and its phase spectrum must be an odd function of
the frequency. In other words, if the amplitude has the value A1 at the frequency ω = ω1, it
must have the same amplitude at the frequency ω = −ω1 and if the phase has the value θ = θ1
at the frequency ω = ω1, the phase at ω = −ω1 must be such that θ = −θ1. An example is the
spectrum of a sine wave x(t) = Acos(ω0t + θ) displayed in Figure 2- 6.

Figure 2- 6 two sided spectrum of a sine wave

It shows that the sine wave is the sum of two phasors of equal length rotating at the same
speed but in opposite direction and starting with opposite initial phases.
2.2 Some signal properties

Periodicity:
We say that the signal x(t) is periodic if it satisfies the following property:
∀t ∈ ℝ ∃T > 0 such that x(t + T ) = x(t )
The number T is called the period of the signal. It is evident that if T is a period of the signal
then kT (k ∈N*) is also a period. For a periodic signal, there exist an infinite set of positive
periods. The smallest one (we denote it T0) is called the fundamental period. The phasor
defined by equation (0.4) and the sine wave defined by equation (0.5) are two examples of
periodic signals. Their fundamental period is:
2π 1
T0 = = (0.8)
ω0 f 0
and f0 is the fundamental frequency (measured in Hertz). It indicates the number of turns per
second (cycles/s) of the rotating phasor.

Energy and power:

If the signal x(t) represents a current, then the instantaneous power that will be dissipated by a
resistance R will be p(t) = Rx2(t). In signal theory, we want a definition that will be
independent on the value of R. So, we define the instantaneous power of the signal x(t) as:
p (t ) = x(t )
2
(0.9)
i.e. it represents the power dissipated by a 1 Ω resistance if the signal corresponds to either
current or voltage. We use the absolute value because the signal might be complex. At that
time, the energy that will be dissipated in the interval of time [t1, t2] is:
E [t1 , t2 ] = ∫ p (t )dt = ∫ x(t ) dt
t2 t2 2
(0.10)
t1 t1
So, the total energy of the signal is:
+∞ 2
E=∫ x(t ) dt (0.11)
−∞
Of course, the above integral must converge. If it diverges, then the signal will have an
infinite energy. We can also define the average power in the interval [−T/2, T/2] as:
T
 T T 1  T T 1 2
P  − ,  = E  − ,  = ∫ T x(t ) dt
2
(0.12)
 2 2  T  2 2  T −2
And the total average power (that will be called simply power) is the limit when T goes to
infinity of the above power.
1 T2
P = lim ∫ T x(t ) dt
2
(0.13)
T →+∞ T −
2
The convergence properties of E and P impose a distinction or a classification of signals. If E
is finite, then P will be zero. The signal is classified as an energy signal. However, for P to be
different from zero, we must have E infinite. In this case, the signal is classified as a power
signal.
Energy signal: 0 < E < ∞ ; P = 0 .

Power signal: E = ∞ ; 0 < P < ∞ .

Energy signals must decay to zero at t = ±∝. Any bounded finite time signal is then an energy
signal.
For example, consider the signal Π(t).
 1 1
1 − ≤ t ≤
Π (t ) =  2 2
0 elsewhere
Its energy is of course E = 1. Consider now the following exponential signal.
x(t ) = A exp ( −α t ) u (t ) ; α > 0
+∞
+∞ A2 A2
Its energy is: E = ∫ A exp ( −2α t )dt = −
2
exp ( −2α t ) =
0 2α 0
2α
However, any periodic signal must be at least a power signal. In the case of periodic signals,
the computation of the total average power is simplified if we use the periodicity. If we
decompose the integral over T as a sum of integrals over the fundamental period T0, we obtain
the following formula:
1
P = ∫ x(t ) dt
2
(0.14)
T0 T0
The notation ∫
T0
means an integral over any interval of length T0, e.g. [α, α+T0[.
Using (0.14), the power of the phasor is: P = A2 and the power of the corresponding sine wave
is: P = A2/2.

Time average
For power signals, we can define the time average of the signal (It is the dc value of the
signal).
1 T2
< x(t ) >= lim ∫ T x(t )dt (0.15)
T →+∞ T −
2
For periodic signals, (0.15) reduces to
1
< x(t ) >= ∫ x(t )dt (0.16)
T0 T0
Time average can be defined for energy signals. However, the dc value of energy signals is
zero.

2.3 Fourier series

The theory of Fourier series basically states that periodic signals can be expressed as a linear
combination of phasors. We will see later that phasors form an essential set of signals and
practically almost all signals of interest in communication are formed by such linear
combinations. Let us consider a periodic signal x(t) with fundamental period T0. If this signal
satisfies the following sufficient conditions:
Dirichlet conditions:
1. The signal x(t) is bounded in the interval [0, T0].
2. It has a finite number of discontinuities in the interval [0, T0].
3. It has a finite number of extrema in the interval [0, T0].
Then it can be developed in the following series:
+∞
x(t ) = ∑c n exp ( jnω0t ) (0.17)
n =−∞

where ω0 = 2π f 0 = 2π
T0
We note that the Dirichlet conditions are sufficient and not necessary. This means that we can
find functions that possess Fourier series without satisfying the above conditions. Even
though the Fourier series have been defined for periodic signals, the development (0.17) can
be applied to a finite time signal (i.e. a signal that is zero outside an interval of length T0). In
this case, the finite time signal will be equal to a periodic signal with period T0 inside the
interval. In this course, we are not going to study the convergence of the series. Simply, we
can state its convergence at the discontinuities of x(t).
Consider a signal x(t) having a discontinuity at t0 ∈ [0, T0]. Let x(t0) = x0 and
lim x(t ) = x(t0 + ) ; lim x(t ) = x(t0− ) then:
t →t0 ,t >t0 t →t0 ,t <t0
N
x(t0 + ) + x(t0 − )
lim ∑c n exp ( jnω0t ) = (0.18)
N →+∞ 2
n =− N
Relation (0.18) explains why we have defined the value of the step function at the origin as
u(0) = 1/2 and not 1 as it is usually defined in signal and system courses. The computation of
the coefficients cn is quite instructive:
Let us multiply both sides of equation (0.17) by exp(−jkω0t) and integrate over T0.
 +∞ 
T0 T0

∫−T20
2
x (t ) exp ( − jk ω 0 t ) dt = ∫−T20 n∑
2
cn exp ( jnω0t )  exp ( − jkω0t ) dt
=−∞ 
If the series converges, we can interchange the integration and the summation.

T0 +∞ T0 +∞ T0

∫ 2
T x(t ) exp ( − jkω0t ) dt = ∑ cn ∫ 2T0 exp ( jnω0t ) exp ( − jkω0t ) dt = ∑ cn ∫ 2T0 exp ( j ( n − k ) ω0t ) dt
− 0 − −
2 n =−∞ 2 n =−∞ 2
The integral is not hard to evaluate and its value is:

T0
sin π ( n − k )
∫ 2
exp  j ( n − k ) ω0t  dt = T0 = T0 sinc ( n − k ) (0.19)
π (n − k )
T
− 0
2

Equation (0.19) defines the function sinc(x) which is plotted below.

 1.2
sinc(x)

0.8

0.6

0.4

0.2

-5 -4 -3 -2 -1 0 1 2 3 4 5

-0.2

-0.4

Figure 2- 7 The sinc function

This function has the following property:

∀m ∈ ℤ ; sinc(m) = 1 if m = 0 and sinc(m) = 0 if m ≠ 0
Using the above property, we obtain:
T0 +∞

∫ 2
T x(t ) exp ( − jkω0t ) dt = ∑ c T sinc(n − k ) = c T
n 0 k 0
− 0
2 n =−∞

and finally, the coefficients of the Fourier series are given by:
1
cn = x(t ) exp ( − jnω0t ) dt
∫ (0.20)
T0 T0
If the signal x(t) is real, then all positive frequency phasors must be compensated by a
negative frequency with same amplitude and opposite phase. In other words, we must have:
cn = c−* n (0.21)
or
cn = c− n
(0.22)
arg [ cn ] = − arg [ c− n ]
In this case, we can use these symmetries to obtain the following alternate expressions of the
Fourier series.
∞
x(t ) = c0 + ∑ 2 cn cos ( nω0t + arg [ cn ]) (0.23)
n =1

By using the trigonometric identity: cos(a + b) = cos a cos b − sin a sin b , we obtain the
following result.
∞ ∞
x(t ) = c0 + ∑ an cos nω0t + ∑ bn sin nω0t (0.24)
n =1 n =1
where
an = cn cos ( arg [ cn ])
 bn = − cn sin ( arg [ cn ]) (0.25)
In (0.23), the term corresponding to n = 1 is called the fundamental, the terms corresponding
to n >1 are called the nth harmonics and of course c0 is the average or dc value.

Symmetry properties
In the case of real signals, if the function x(t) is even, i.e. x(t) = x(−t), then the series (0.24)
will not contain any sine term: bn = 0 for all n. Alternatively, if the function is odd, i.e.
x(t) = −x(−t) then (0.24) will contain only sine terms, i.e. c0 = 0 and an = 0.
If the real signal x(t) has a half wave symmetry: x(t ± T0/2) = −x(t), then its Fourier series will
contain only odd harmonics, i.e. cn = 0 for n = 0, ±2, ±4, …

Example of Fourier series computation:

t T T
Consider the pulse train defined by: x(t ) = AΠ   ; 0 < τ < T0 for − 0 ≤ t ≤ 0 and
τ  2 2
x(t ) = x(t + T0 ) for all the other periods.
τ
τ 2
1 1 A
cn = ∫−τ2
2
A exp ( − jnω t )dt = exp ( − jnω t )
T0 − jnω0
0 0
T0 −
τ
2

τ  nτ 
=A sinc  
T0  T0 
magnitude

0.28

0.26

0.24

0.22

0.2

0.18

0.16

0.14

0.12

0.1

0.08

0.06

0.04

0.02

harmonic
-9 -8 -7 -6 -5 -4 -3 -2 -1 1 2 3 4 5 6 7 8 9

-0.02

-0.04

Figure 2- 8 Amplitude spectrum of the pulse train

2π
The above spectrum corresponds to harmonics located at frequencies nω0 = n along with a
T0
τ
duty cycle d = .
T0
A n A nπ
If now τ = T0/2 (square wave), we obtain: cn = sinc   . This means that cn = sin ,
2 2 πn 2
giving
  A
 n=0
2

cn =  0 n even
A A A
 , − , ,... n = ±1, ±3, ±5,...
 π 3π 5π
Grouping the terms corresponding to + and − n, we obtain the following series:
A 2A  1 1 
x(t ) = +  cos ω0t − cos ω0t + cos 5ω0t − ⋯ 
2 π  3 5 

Parseval's relation:
Let us consider two signals x(t) and y(t), both periodic with period T0. Their Fourier
development is:
+∞
x(t ) = ∑X n exp ( jω0t )
n =−∞
+∞
y (t ) = ∑Y n exp ( jω0t )
n =−∞
Then
+∞
1
T0 ∫T0
x(t ) y* (t )dt = ∑XY
n =−∞
n n
*
(0.26)

Proof:
1 1  +∞ +∞

∫ x(t ) y* (t )dt = ∫ ∑ n X exp ( 0 ) ∑ Ym* exp ( − jmω0t )  dt
jnω t
T0 T0 T0 T0  n =−∞ m =−∞ 
We interchange the integral with the sums.
* 1 
+∞ +∞
1
∑ ∑ ( ω )
T0 ∫T0 ∫
x (t ) y *
(t ) dt = X n Y m  exp j ( n − m ) 0 t dt 
n =−∞ m =−∞  T0 T0 
Using the properties of the sinc function, we finally obtain:
+∞
1
T0 ∫T0
x (t ) y *
(t ) dt = ∑
n =−∞
X nYn*

(q.e.d.)
Relation (0.26) can be applied for the case y(t) = x(t). We obtain
+∞
1
P = ∫ x(t ) dt = ∑ X n
2 2
(0.27)
T0 0 T
−∞

In other words, (0.27) means that the total average power of the waveform x(t) is the sum of
the powers of all the phasors constituting x(t).