McGill University

Department of Electrical and Computer Engineering

Communication Systems (304-411)

These notes do not replace class notes.

These notes are designed as a teaching aid for self review of the material, and they should be used
together with the class notes as well as the text book. The students are encouraged to read these
notes and follow the instructions.

1 Introduction
1.1 The communication process
Transport of information between points separated in space and/or time.

Source Information User

Figure 1: A communication process

Examples :

• Points separated in space :

– telephone (point to point)

– broadcast radio and TV (point to multi-point)
– cellular systems (point to multi-point in down link, multi-point to point in the up link).

• Points separated in time :

– magnetic recording systems

– compact disk systems.

1
 Messages Channel input signals
SOURCE TRANSMITTER

CHANNEL

USER RECEIVER
Destination messages Channel output signals

Figure 2: General model for a communication system

Transmitter : Source messages → Channel input signals

Channel : Channel input signals → Channel output signals
Receiver : Channel output signals → User messages

Communication systems design :

Given : Source, Channel and User, find a Transmitter and Receiver such that the user messages are
”as close as possible” to source messages.
Difference between source message and user message is called distortion.
Objective : small distortion
Resources : transmission power, bandwidth, complexity, and possible others depending on specific
systems.

2
1.2 Detailed model of a communication system

SOURCE CHANNEL
SOURCE MODULATOR
ENCODER ENCODER waveform
(modulated signal)

Message signal Source code word Channel code word

CHANNEL

Received signal
SOURCE CHANNEL
USER DEMODULATOR
DECODER DECODER

Estimate of Estimate of Estimate of

Message signal source code word channel code word

Figure 3: General model for a communication system

Source encoder :
Removes unnecessary details (redundancies, ets.) from source messages, and keeps only the infor-
mation that needs to be transmitted.

Channel encoder :
Adds redundancy to the information that has to be transmitted, in order to reduce the effects of
channel noise.

Modulator :
Maps the transmitted information into channel input signals.

Demodulator + Channel decoder :

Attempt to recover the transmitted information from the channel output (distorted) signal.

Source decoder :
Converts the recovered information to a form suitable for user.

Source or Channel encoder may or may not be needed. Modulators are always needed.

3
1.3 Analog and digital systems
Analog modulation: The set of all possible channel input signals is continuous (infinite).
Digital modulation:The set of all possible channel input signals is discrete and finite.

Examples :
Analog Communication Systems :
Suppose that the value of a variable A has to be transmitted. It is known that 0 < A < 1.
The modulation procedure could be x(t) = A cos(ωc t) where ωc = 2πfc . In this case the waveform
x(t) is transmitted over the channel. Other modulation scheme could employ x(t) = cos(ω c t +
2πA), or in general x(t) = s(t, A) where s(t, A) is a time function that depends on the parameter
A.
Digital Communication Systems:
Suppose that it is known that A = 1/4 or A = 1/2. The modulation procedure could be :

1/4 cos(ωc t) , if A = 1/4
x(t) =
1/2 cos(ωc t) , if A = 1/2
or :


cos[(ωc − ∆)t] , if A = 1/4
x(t) =
cos[(ωc + ∆)t] , if A = 1/2
or :


s1 (t) , if A = 1/4
x(t) =
s2 (t) , if A = 1/2

where s1 (t) 6= s2 (t).

4
2 Representation of signals and systems
2.1 Time and frequency domains (review)
a) Fourier transform
Give the definition of the Fourier transform, and state the conditions for its existance.

b) The Dirac delta function

Definitions:

• The delta function δ(t) is defined as the generalized function or distribution that satisfies
Z t2 (
f (0) if t1 < 0 < t2
For any continuous function f (t) at t = 0 f (t)δ(t)dt =
t1 0 else.

Therefore δ(t) satisfies

Z ∞
• δ(t)dt = 1, suggesting that δ(t) = 0 ∀ t 6= 0. Note that this latter definition is to be
−∞
used only in the context of integration.

• The delta function can also be considered as the limit of sequences of functions

1, for |v| ≤ 1 ,
2
δ(t) = lim a rect(at) where rect(v) =
a→∞ 0, for |v| > 1
2
sin(πv)
= lim a sinc(at) where sin(v) =
a→∞ πv
2 t2
= lim ae−πa
a→∞

Properties:

• δ(t) is an even function

• δ(t) satisfies the sifting or sampling property:

Z ∞
For any continuous function f (t) at t0 f (t)δ(t − t0 )dt = f (t0 )
−∞

• δ(t) satisfies the replication property:

For any continuous function f (t) at t0 f (t) ∗ δ(t − t0 ) = f (t − t0 )

5
 a a
a
1
a

u u u
-1 1 -1 1
2a 2a a a
2 2
(a) graph of a rect(at) (b) graph of a sinc(at) (c) graph of ae−πa t

• The delta function satisfies

f (t)δ(t − t0 ) = f (t0 )δ(t − t0 )

1
δ(at) = δ(t) a 6= 0
|a|
du(t)
δ(t) =
dt
where u(t) is the unit step function. (To be rigorous these equations are valid only if they
appear inside integrals.)

• Fourier transform relationships:

Z ∞
F  
δ(t) −→ 1 since F [δ(t)] = δ(t)e−j2πf t dt = e−j2πf t t=0 = 1
−∞
F
1 −→ δ(−f ) = δ(f ) by the frequency duality property of the Fourier transform
R∞
Therefore −∞ e−j2πf t dt = δ(f ) and

F
ej2πfc t −→ δ(f − fc )
F 1 
cos 2πfc t −→ δ(f − fc ) + δ(f + fc )
2
F 1 
sin 2πfc t −→ δ(f − fc ) − δ(f + fc )
2j

6
2.2 Hilbert transform and complex envelope representation of signals
a) Hilbert transform
Let x(t) be a real signal, the Hilbert transform of x(t) is defined by
Z Z t− Z A 
1 ∞ x(τ ) 1 x(τ ) x(τ )
x̂(t) = dτ = lim dτ + dτ
π −∞ t − τ →0 π
A→∞ −A t − τ t+ t − τ
Cauchy’s principal value of the integral
1
x̂(t) = x(t) ∗
πt
Fourier transform of x̂(t) :
 
1
F {x̂(t)} = X̂(f ) = F {x(t)} · F = X(f ) (−jsgn(f )) = −jsgn(f )X(f )
πt

where sgn(f ) is given by 

1,
 f >0
sgn(f ) = 0, f =0


−1, f <0
Let H(f ) = −jsgn(f ) (Hilbert transformer)
Positive frequencies: phase shift of −90o (90o phase lag).
Negative frequencies: phase shift of 90o (90o phase lead).

Inverse Hilbert transform :

Let us assume that X(0) = 0,

ˆ  2
X̂(f ) = −j sgn(f )X̂(f ) = −j sgn(f ) X(f ) = −X(f )

Thus

x̂(t) = −x(t)
Z ∞
1 1 x̂(τ )
x(t) = − x̂(t) = −x̂(t) ∗ =− dτ
πt π −∞ t−τ

7
Hilbert transform of cosine and sine functions
 
Let x(t) = cos 2πf0 t, then X(f ) = 21 δ(f − f0 ) + δ(f + f0 ) .

j 
X̂(f ) = −j sgn(f )X(f ) = − sgn(f0 )δ(f − f0 ) + sgn(−f0 )δ(f + f0 )
2
j 
= − δ(f − f0 ) − δ(f + f0 )
2
Thus
x̂(t) = sin 2πf0 t
Similarly show that if x(t) = sin 2πf0 t, then x̂(t) = − cos 2πf0 t.

Theorem 2.1 ( Hilbert transform of the product of a low pass and a high pass signal). Suppose
that x(t) = f (t)g(t) where f (t) is a low pass (i.e. F (f ) = 0, |f | > W ) and g(t) is a high pass
signal (i.e. G(f ) = 0, |f | < W ), and f (f ) and G(f ) do not overlap, then x̂(t) = f (t)ĝ(t).

Exercise: Draw an example of the spectrum of a low pass and high pass signals.

Proof:

Z ∞
X(f ) = F (f − u)G(u)du
−∞
Z ∞
X̂(f ) = −j sgn(f ) F (f − u)G(u)du
−∞

Z ∞
 
F f (t)ĝ(t) = F (f ) ∗ Ĝ(f ) = F (f − u) · −j sgn(u)G(u)du
−∞
Z ∞
 
X̂(f ) − F [f (t)ĝ(t)] = −j F (f − u)G(u) sgn(f ) − sgn(u) du = 0
−∞

since F (u) and G(u) do not overlap.

8
b) Pre-envelope or complex analytic (CA) signal
The pre-envelope or complex analytic (CA) signal associated with a real-valued signal x(t) is
defined by
x+ (t) = x(t) + j x̂(t)
where x̂(t) is the Hilbert transform of x(t). The pre-envelope satisfies

x(t) = < {x+ (t)}

The Fourier transform of x+ (t) is given by

F {x+ (t)} = X+ (f ) = X(f ) + j X̂(f )

= X(f ) + j(−j sgn(f ))X(f )
= X(f ) [1 + sgn(f )]


 f >0
= f =0 exercise: complete the blanks


f <0

Exercise: Draw the Fourier transform of the pre-envelope of a low pass signal.

b) Complex envelope (CE) of x(t)

Let x(t) be a real signal then the complex envelope (CE) of x(t) with respect to the carrier f 0 is
defined as

x̃(t) = x+ (t)e−j2πf0 t = x(t) + j x̂(t) e−j2πf0 t
and we have 
x(t) = < x̃(t)ej2πf0 t


Example: Find the complex envelope with respect to f0 of an RF pulse x(t) = Arect t
T
cos 2πfc t,
where fc 6= f0 .
Frequency domain relations

 1 1
x(t) = < x̃(t)ej2πf0 t = x̃(t)ej2πf0 t + x̃∗ (t)e−j2πf0 t
2 2

9
Let X̃(f ) = F {x̃(t)}, then
1 1
X(f ) = X̃(f ) ∗ δ(f − f0 ) + X̃ ∗ (−f ) ∗ δ(f + f0 )
2 2
1 1 ∗
= X̃(f − f0 ) + X̃ (−f − f0 )
2 2
Since x̃(t) = x+ (t)e−j2πf0 t ,

 2X(f + f0 ) , if f > −f0
X̃(f ) = X+ (f ) ∗ δ(f + f0 ) = X+ (f + f0 ) = X(0) , if f = −f0
0 , if f < −f0


Exercise: Assume that x(t) is a low pass signal, draw the Fourier transform of its complex envelope
with respect to f0 . Repeat assuming that x(t) is a bandpass signal centered around ±f c . Assume
fc 6= f0 .

10
d) bandpass signals
Let x(t) be a real signal such that X(f ) is non-zero (or non-negligible) only in some frequency
band centered about ±fc and fc  0, then x(t) is a bandpass signal.
Exercise: Draw an example of the spectrum of a bandpass signal of bandwidth 2W .

If 2W  fc then x(t) is called a narrow-band signal.

Exercise: Show that the complex envelope with respect to fc of a bandpass signal centered
about ±fc is a low-pass signal.

Since x̃(t) is a complex-valued function

x̃(t) = xI (t) + jxQ (t)

where xI (t) and xQ (t) are real-valued low-pass functions.

Representation of a bandpass signal:
By definition of the complex envelope

x(t) = < x̃(t)ej2πfc t

= < [xI (t) + jxQ (t)] ej2πfc t
= xI (t) cos 2πfc t − xQ (t) sin 2πfc t

where xI (t) and xQ (t) are low-pass and real, i.e.

XI (f ) = 0, |f | > W and W < fc

XQ (f ) = 0, |f | > W

xI (t) is referred as the in-phase component of x(t), xQ (t) is referred as the quadrature component
of x(t).

11
 x+ (t) = x(t) + j x̂(t)

= xI (t) cos 2πfc t − xQ (t) sin 2πfc t + j xI (t) cos 2πfc t − j xQ (t) sin 2πfc t
= xI (t) cos 2πfc t − xQ (t) sin 2πfc t + jxI (t) cos 2πfc t − jxQ (t) sin 2πfc t
= xI (t) cos 2πfc t − xQ (t) sin 2πfc t + jxI (t) sin 2πfc t + jxQ (t) cos 2πfc t
= [xI (t) + jxQ (t)] ej2πfc t
= x̃(t)ej2πfc t

which is consistent with the definition of x̃(t).

complex envelope of x(t) (with respect to fc ):

x̃(t) = xI (t) + jxQ (t)

(natural) envelope of x(t):

q
r(t) = |x̃(t)| = x2I (t) + x2Q (t)

phase of x(t):  
−1 xQ (t)
φ(t) = arg [x̃(t)] = tan
xI (t)
Therefore

x̃(t) = r(t)ejφ(t)

x(t) = < x̃(t)ej2πfc t

= < r(t)ej(2πfc t+φ(t))


= r(t) cos 2πfc t + φ(t)

Exercise: Give an interpretation of the complex envelope of x(t) as a time-varying phasor.

12
Fourier transform of bandpass signals

XI (f ) = F {xI (t)} XQ (f ) = F {xQ (t)} X(f ) = F {x(t)}

x(t) = xI (t) cos 2πfc t − xQ (t) sin 2πfc t

1 1
X(f ) = XI (f ) ∗ [δ(f − fc ) + δ(f + fc )] − XQ (f ) ∗ [δ(f − fc ) − δ(f + fc )]
2 2j
1 1 j j
= XI (f − fc ) + XI (f + fc ) + XQ (f − fc ) − XQ (f + fc )
2 2 2 2
1 1
= [XI (f − fc ) + jXQ (f − fc )] + [XI (f + fc ) − jXQ (f + fc )]
2 2
since xI (t) and xQ (t) are real. Hence

1 1
X(f ) = X̃(f − fc ) + X̃ ∗ (−f − fc )
2 2
n o
Note that although X̃(f ) = XI (f ) + jXQ (f ), we have XI (f ) 6= < X̃(f ) and XQ (f ) 6=
n o
= X̃(f ) , since XI (f ) and XQ (f ) are not necessarily real.
Exercise: Show that
1 
XI (f ) = X+ (f + fc ) + X+∗ (−f + fc )
2
1  
XQ (f ) = X+ (f + fc ) − X+∗ (−f + fc )
2j

Generation of a bandpass signal from xI (t) and xQ (t): Transformation up frequency conver-
sion (UFC)
Draw a block diagram that generates a bandpass signal from xI (t) and xQ (t).

13
Generation of the complex envelope of a signal x(t): Transformation down frequency con-
version (DFC)
Draw a block diagram that generates the complex envelope of a signal x(t).

e) Bandpass linear time-invariant systems

A linear time-invariant system is called bandpass if its impulse response is a bandpass signal.
Let x(t) be a bandpass signal applied to the input of a linear time-invariant bandpass system
(around the same carrier frequency), then the output y(t) = h(t) ∗ x(t) is also a bandpass signal
and
ỹ(t) = 21 h̃(t) ∗ x̃(t)
where x̃(t), h̃(t) and ỹ(t) are the complex envelopes of x(t), h(t) and y(t) with respect to the same
carrier.
Exercise : Prove this result by first showing that y(t) is a bandpass signal. Then prove the
relationship by manipulating y(t) using pre-envelopes so that it is in the form

y(t) = < XXej2πfc t

where XX is to be found. By definition XX is the complex envelope of y(t), so the result follows.

14