Wireless Communications Group

Digital modulations

Francois Horlin

1
 Outline

Introduction
Orthogonal expansion of signals
Pulse-amplitude modulation (PAM)
Phase-shift keying (PSK)
Quadrature-amplitude modulation (QAM)
Frequency-shift keying (FSK)
Exercises

2
 References

Digital Communications, Fourth Edition, J. G. Proakis

3
 Outline

Introduction
Orthogonal expansion of signals
Pulse-amplitude modulation (PAM)
Phase-shift keying (PSK)
Quadrature-amplitude modulation (QAM)
Frequency-shift keying (FSK)
Exercises

4
 Introduction

The modulator maps the digital information into analog waveforms

that match the characteristics of the channel:
Blocks of k bits M = 2k waveforms sm (t) tai sao phai biet so
symbol/bit?

khi truyen block k bit thi

Different schemes exist: phai luon luon co 2^k
waveform hay sao
waveform co phai symbol
Memoryless modulations/modulations with memory ko

Linear/non-linear modulations
The course focuses on the widely used memoryless linear modulations

5
 Outline

Introduction
Orthogonal expansion of signals
Pulse-amplitude modulation (PAM)
Phase-shift keying (PSK)
Quadrature-amplitude modulation (QAM)
Frequency-shift keying (FSK)
Exercises

6
 Expansion of digitally modulated signals

Goal: develop a vector representation of the signals

s1
..

s(t)


.

sK

Method: project the signals onto an orthonormal set of basis functions

7
 Expansion of digitally modulated signals

Example: digitally modulated signals can often be expanded in terms

of two orthonormal basis functions:

s(t) = s1 f1 (t) + s2 f2 (t)

with

r
2
f1 (t) = cos (2fc t)
T
r
2
f2 (t) = sin (2fc t)
T

8
 Preliminary concepts

Inner product of two complex signals x1 (t) and x2 (t):

Z +
< x1 (t), x2 (t) > := x1 (t) x2 (t) dt

Norm of a signal:
Z + 1/2
1/2
kx(t)k := (< x(t), x(t) >) = |x(t)|2 dt

Signals are orthonormal if they are orthogonal (inner product equal to

zero) and their norm are all unitary
Signals are independent if no signal can be represented as a linear
combination of the other signals

9
 Orthogonal signal expansion

Let s(t) be a deterministic, real-valued signal of finite energy

Z +
Es = |s(t)|2 dt

Let fk (t) be a set of orthonormal functions (k = 1, , K)

The signal s(t) can be approximated by a weighted linear combination

of the orthonormal functions:
K
X
s(t) = sk fk (t)
k=1

10
 Coefficient sk optimization

Approximation error:

e(t) := s(t) s(t)

Energy of the approximation error:

Z +
2
Ee = [s(t) s(t)] dt

" K
#2
Z + X
= s(t) sk fk (t) dt
k=1

11
 Coefficient sk optimization

The coefficients sk are selected to minimize Ee :

Z + " K
#
Ee X
= s(t) sk fk (t) fn (t) dt = 0 (n = 1, , K)
sn k=1

Since the functions fk (t) are orthonormal:

Z +
sk = s(t) fk (t) dt

The approximation s(t) is the projection of s(t) onto the

K-dimensional signal space spanned by the functions fk (t)

12
 Resulting approximation error energy

" K
#2
Z + X
Emin = s(t) sk fk (t) dt
k=1
Z +
2
= [s(t)]

K
X Z +
2 sk s(t)fk (t) dt
k=1

K X
X K Z +
+ sk sk fk (t)fk (t) dt
k=1 k =1

K
X
= Es s2k
k=1

13
 Complete set of orthonormal functions

If Emin = 0 (the approximation error has a zero energy):

K
X
s(t) = s(t) = sk fk (t)
k=1

and
K
X
Es = s2k
k=1

The set of orthonormal functions is complete when every finite energy

signal can be represented by a series expansion

14
 Euclidian distance

Euclidian distance between a pair of signals:

Z + 1/2
dmn := (sm (t) sn (t))2 dt
t=
= ksm sn k

It measures the dissimilarity between two signals signal co phai la symbol va waveform
luon ko?

For each modulation, the bit error rate (BER) is dominated by the
minimum Euclidian distance between all possible transmitted signals

15
 Outline

Introduction
Orthogonal expansion of signals
Pulse-amplitude modulation (PAM)
Phase-shift keying (PSK)
Quadrature-amplitude modulation (QAM)
Frequency-shift keying (FSK)
Exercises

16
 Pulse-amplitude modulation (PAM) signal

j2fc t
 
sm (t) = Am g(t) e (m = 1, , M and 0 t T )
= Am g(t) cos (2fc t)

Amplitudes Am take the values Am = (2m 1 M )d where 2d is

the distance between adjacent symbols
Real-valued waveform g(t) shapes the spectrum of the transmitted
signal (assume first that g(t) is a rectangular pulse)
Symbol interval T determines the bit rate R = log2 (M )/T

17
Baseband representation

18
 Constellation symbols

bit/symbol lon hon 1, van dung

PAM?

Gray encoding: adjacent amplitudes differ only by one bit such that
most likely errors caused by noise lead to a single bit error

19
 Signal energy

Z T
Em = s2m (t) dt
t=0
T
A2m
Z
= g 2 (t) dt
2 t=0
A2m
= Eg
2

where Eg is the energy of the pulse g(t)

20
 Signal orthogonal expansion

sm (t) = sm f (t)

Unit-energy function:
s
2
f (t) := g(t) cos (2fc t)
Eg

Coefficient (m = 1, , M ):
r
Eg
sm := Am
2

21
 Euclidian distance

Euclidian distance between the symbols m and n:

dmn = |sm sn |
r
Eg
= |Am An |
2
p
= d 2Eg |m n|

Minimum Euclidian distance:

p
dmin = d 2Eg

BER is improved if the distance between the symbols or the energy of

the pulse is increased (but the signal energy is higher!)

22
 Outline

Introduction
Orthogonal expansion of signals
Pulse-amplitude modulation (PAM)

Phase-shift keying (PSK)

Quadrature-amplitude modulation (QAM)
Frequency-shift keying (FSK)
Exercises

23
 Phase-shift keying (PSK) signal

jm j2fc t
 
sm (t) = g(t) e e (m = 1, , M and 0 t T )
= g(t) cos (2fc t + m )
= g(t) cos (m ) cos (2fc t) g(t) sin (m ) sin (2fc t)

2
Phases m take the values m = M (m 1)
Real-valued waveform g(t) shapes the spectrum of the transmitted
signal
Symbol interval T determines the bit rate R = log2 (M )/T

24
 Constellation symbols

Gray encoding: adjacent phases differ only by one bit such that most
likely errors caused by noise lead to a single bit error

25
 Signal energy

Z T
E = s2m (t) dt
t=0
1 T 2
Z
= g (t) dt
2 t=0
1
= Eg
2

The signal waveforms have equal energy

26
 Signal orthogonal expansion

sm (t) = sm1 f1 (t) + sm2 f2 (t)

Orthonormal functions:
s
2
f1 (t) := g(t) cos (2fc t)
Eg
s
2
f2 (t) := g(t) sin (2fc t)
Eg

Vector of coefficients (m = 1, , M ):
h i h q q i
Eg Eg
sm := sm1 sm2 = 2 cos (m ) 2 sin (m )

27
 Euclidian distance
Euclidian distance between the symbols m and n:

dmn = ksm sn k
s  
2(m n)
= Eg 1 cos
M

Minimum Euclidian distance:

s  
2
dmin = Eg 1 cos
M

BER is improved if the energy of the pulse is increased (but the signal
energy is higher!) or if the number of phase states is decreased (but
the number of transmitted bits is lower!)
28
 Outline

Introduction
Orthogonal expansion of signals
Pulse-amplitude modulation (PAM)
Phase-shift keying (PSK)

Quadrature-amplitude modulation (QAM)

Frequency-shift keying (FSK)
Exercises

29
 Quadrature amplitude modulation (QAM) signal

Modulate two quadrature carriers:

 R I j2fc t

sm (t) = (Am + jAm ) g(t) e (m = 1, , M and 0 t T )
= AR
m g(t) cos (2fc t) A I
m g(t) sin (2fc t)

or, equivalently,

Modulate jointly amplitude and phase:

jm j2fc t
 
sm (t) = Am e g(t) e (m = 1, , M and 0 t T )
= Am g(t) cos (2fc t + m )

30
 Constellation symbols

amplitude and phase modulation

2-amplitude modulation

31
 Signal orthogonal expansion

sm (t) = sm1 f1 (t) + sm2 f2 (t)

Orthonormal functions:
s
2
f1 (t) := g(t) cos (2fc t)
Eg
s
2
f2 (t) := g(t) sin (2fc t)
Eg

Vector of coefficients (m = 1, , M ):
h i h q q i
Eg Eg
sm := sm1 sm2 = AR m 2 AIm 2

32
 Euclidian distance

Euclidian distance between the symbols m and n:

dmn = ksm sn k
r
Eg
= [(AR R 2 I I 2
m An ) + (Am An ) ]
2

Minimum Euclidian distance (rectangular constellation corresponding

to 2 PAM signals):
p
dmin = d 2Eg

BER is improved if the distance between the symbols or the energy of

the pulse is increased (but the signal energy is higher!)

33
 Outline

Introduction
Orthogonal expansion of signals
Pulse-amplitude modulation (PAM)
Phase-shift keying (PSK)
Quadrature-amplitude modulation (QAM)

Frequency-shift keying (FSK)

Exercises

34
 Frequency-shift keying (FSK) signal

M equal-energy orthogonal waveforms that differ in frequency:

"r #
2E j2mf t j2fc t
sm (t) = e e (m = 1, , M and 0 t T )
T
r
2E
= cos (2(fc + mf )t)
T

Orthogonal? Study of the cross-correlation...

35
 Signal cross-correlation

T
1
Z
kl := sk (t) sl (t) dt
E t=0
Z T
2
= cos (2(fc + kf )t) cos (2(fc + lf )t) dt
T t=0
Z T
1
= cos (2(k l)f t) dt
T t=0
1
= sin (2(k l)f T )
2(k l)f T

36
 Signal orthogonality

1
Two signals on adjacent carriers
0.8
(|k l| = 1) are orthogonal if:
0.6

2f T = n, n 6= 0
sin(x)/x

0.4

0.2

The minimum frequency separa-

0

tion for orthogonality is:

0.2

0.4 1
10 5 0 5 10
f =
x 2T

37
 Signal orthogonal expansion
M
X
sm (t) = smk fk (t)
k=1

Orthonormal functions:
r
2
fk (t) := cos (2(fc + kf )t)
T
Vectors of coefficients (size M ):
h i
s1 := E 0 0
.. ..
. .
h i
sM := 0 0 E

38
 Euclidian distance

Euclidian distance between the symbols m and n:

dmn = ksm sn k

= 2E

Minimum Euclidian distance:

dmin = 2E

BER is improved if the energy of the signal is increased

39
 Outline

Introduction
Orthogonal expansion of signals
Pulse-amplitude modulation (PAM)
Phase-shift keying (PSK)
Quadrature-amplitude modulation (QAM)
Frequency-shift keying (FSK)
Exercises

40
 Exercise 1

Demonstrate that the functions f1 (t), f2 (t) and f3 (t) are

orthonormal.
Express the signal s(t) as a weighted linear combination of the
functions f1 (t), f2 (t) and f3 (t).

41
 Exercise 2

Draw the transmitted signal for all linear modulations (PAM, PSK,
QAM, FSK) when two bits are mapped on each symbol and a
rectangular pulse is used to shape the symbols.

42