EE 477

Introduction to Digital Communications

Digital Modulation
Boğaziçi University
Fall 2024
 August filter
3, 2001tuned 13:9 i Figure 3.4 Use of98a pilot tone to
to fc 2π dt Analog Signal
where the message waveform is constrained
demodulate
Transmission
to satisfy the and Reception that
condition Chapter
|m(t3)| ≤ 1.
We observea DSB
that AM signal.
A c m(t ) cos(2π f c t + φc ) is a double-sideband AM signal and
Since u(t) is a bandpass signal, it can be represented as
A cos(2π f t + φ ) is the carrier component. Figure 3.5 illustrates an AM signal in the
time domain.
c c c
FM Modulator

As long as |m(t )| ≤ 1, the

m(t) amplitude A c [1 + m(t )] is always positive. This is the
FM
desired condition for conventional DSB AM that makes it easy to demodulate, as
rrier from the received signalu(t)
r (t= Ac cos(2π the
) without φ(t)) of a pilot signal. (3.3.2)
f c t +need This can be Modulator
described next. On the other hand, if m(t ) < −1 for some t , the AM signal is said to
be overmodulated and its demodulation is rendered more complex. In practice, m(t ) is
complished
Section 3.3
byAngle
use ofModulation
a phase-locked loop (PLL) as described in Section 5.2. scaled so that its magnitude is always 97less than unity.
and, therefore, It is sometimes convenient to express m(t ) as
m(t) PM
2.2 Conventional Amplitude Modulation Integrator

Review: Analog1 Modulation

m(t ) = am n (t )
Modulator
d
3.3.1 Representation
conventional AM signal of FM and
consistsfi (t)PM
of= aSignals
f c large
+ (3.3.3) to the
where m n (t ) is normalized such that its minimum value is −1. This can be done, for
carrier
φ(t) component in addition
2π dt
uble-sideband AM modulated signal. The transmitted signal is expressed mathemat- PM Modulator
An angle-modulated signal in general can be written as Envelope
ally as If m(t) is the message signal, then in a PM system we have u(t)
m(t) PM
A(t)
Modulator

• AM u(t ) = Au(t) = Ac cos(θ (t))

c [1 + m(t )] cos(2π f c t + φc )
φ(t) = k p m(t) (3.3.4) (3.2.7) t

θ (t) is
here thethemessage
phase ofwaveform
the signal, is andconstrained
its instantaneous to satisfyfrequency f i (t) is given
the condition m(t)
by
thatDifferentiator
|m(t )| ≤ 1.ModulatorFM
Figure 3.25 A comparison of FM and
and in an FM system we have
e observe that Ac m(t ) cos(2π f c t + φc ) is a double-sideband AM signal and
Figure PM 3.5
the time domain.
A conventional AM signal in
modulators.

cos(2π f c t + φc ) is the carrier component.

1 Figure
d
f i (t) =
2π dt 1 3.5
θ (t) d illustrates an AM signal (3.3.1)in the
me domain. fi (t) − f c = k f m(t) = φ(t) (3.3.5)
As long as |m(t )| ≤ 1, the amplitude
Since u(t) is a bandpass signal, it can be represented A c [1 + 2π dt
m(t
as )] is always positive. 1
This is the 2
sired condition for conventional DSB AM that makes it easy to demodulate, as
where k and k f are phase and frequency deviation constants. From the above relation-
scribed next. pOn the is said2 to 4 t 1
ships •wePM/FM
other u(t) hand, = ifAm(t ) < −1
cos(2π f t +for some t , the AM signal0 (3.3.2)
φ(t))
overmodulated have c c
and its demodulation is rendered more complex. In practice, m(t !1
) is 0
0 1 2 3 4 t

aled
and, so that its magnitude is always less than unity.
therefore, k p m(t), m(t ) as PM
!
1 1
It is sometimes convenient φ(t) = to express "t 1 d FM(3.3.6)
0 0

f i (t)m(t 2πk
= )f c=f +−∞ m(τ) dτ, FM signal
(3.3.3)
am
2πndt (t )φ(t) !1
0 1 2 3
E
4 quiv
!1
0 1 2 3 4
ale
nt

here m nEquation
If m(t) is
(t the (3.3.6) showssuch
) is message
normalized the close
signal, then that and interesting
in aitsPM minimum
system we
relation
value between FM and PM sys-
haveis −1. This can be1 done, for 1

tems. This close relationship makes it possible to analyze these systems insignal PMparallel
0 0

and only emphasize their main differences. φ(t) = k pThe m(t)first interesting result observed !1 from (3.3.4) !1
0 1 2 3 4 0 1 2 3 4
Equation (3.3.6)u(t) is that if we phase modulate Envelope
A(t)
the carrier with the integral of a message, Figure 3.26 FM and PM of square and sawtooth waves.
and in anit isFM systemtowe
equivalent have modulation of
frequency the carrier with the original message. On
the other hand, in Equation (3.3.6) the relation can1be expressed d as
f i (t) − f c = k ! f m(t) = φ(t) (3.3.5)
d d
k p dt m(t), 2π dt
PM
t
φ(t) =
where k and k are phase and frequency deviation constants. From the above (3.3.7)
relation-
Why do we need Carrier Modulation?
• Altitude+ Antenna height is proportional to the wavelength, e.g., in most
applications /4 is chosen. The higher the carrier frequency the smaller
the wavelength, and thus the antenna height. High frequency carriers are
preferred to send signals with practical antenna sizes.
• Most of the spectrum is not free. It is licensed and assigned for specific
applications.
• FM Radio: 87.5-108 MHz. Each station having 200KHz bandwitdh.
• Cellular Communication, 4G in Turkey: 800MHz, 900MHz, 1.8GHz, 2.1GHz, 2.600GHz
(20-40MHz resource blocks)
• Wi-Fi: 2.5GHz, 5GHz (20, 40, 80MHz user signals)

• How does the spectrum look like?

The Electromagnetic Spectrum
The RF Spectrum
Digital Modulation
September 25, 2007 13:13

(Memoryless) Digital Modulation Schemes

Digital Communications

k k k
1

10...1 00...1 01...0... Modulator sm(t) 1

In a memoryless modulation scheme, the binary sequence is parsed into subsequences each of length k.
InThen
a memoryless modulation
the modulator scheme,
maps each theofbinary
sequence sequence
k binary symbolsis onto
parsed
oneinto subsequences
of M = 2k signal each of length k.
waveforms,
FIGURE 3.1–1Then
sm (t)the
(mmodulator maps
= 1, . . . , M). Weeach sequence
assume of ksignals
that these binary
aresymbols ontoat one
transmitted everyofTM = 2k signal
s seconds, where waveforms,
Ts is
Block diagram of a memoryless digital modulation scheme.
s (t) (m = 1, . . . , M ). We assume that these signals are transmitted at every T seconds, where T is
called
m the signaling interval. s s

called the signaling

This modulation interval.
scheme is equivalent to a mapping from M = 2k messages to M possible signals.
k
1≤m≤ 2 This
, signals regardless
modulation of the previously
scheme is equivalent transmitted
to a mapping from signals.
M = 2k messages to MThis modulation
possible signals.
k
scheme is equivalent to a mapping from M = 2 messages to M possible signals, as
1
shown in Figure 3.1–1. Rs = symbols/s. Symbol rate
Ts
In a modulation scheme with memory, the mappingBitis rate
Rb = kRs bits/s.
from the set of the current
k bits and the past (L − 1)k bits to X the
M set of1 possible
XM M = 2k messages. In this case
the transmitted signal dependsEavg. on
= the pcurrent
m Em = k bits Averageas
Em . as well Energy
the most recent L − 1
M
blocks of k bits. This defines a finite-state machine with 2(L−1)k states. The mapping that
m=1 m=1

defines the modulation scheme can be viewed as a mapping from the current state and
the current input of the modulator to the set of output signals resulting in a new state of
 1
Rs = symbols/s. Symbol rate
Ts
Rb = kRs bits/s. Bit rate
XM M
1 X
Eavg. = pm E m = Em . Average Energy
Pulse Amplitude Modulation (PAM)
m=1
M m=1

sm (t) = Am g(t) cos(2⇡fc t) 0  t  Ts

N
fc = (N is a very large integer)
Ts
Am = (2m 1 M) , m = 1, 2, . . . , M.

g(t) is the pulse shaping filter to limit the duration of the waveform as there would be another symbol to
be transmitted in the next Ts seconds. Rectangular window is one solution:
8
✓ ◆ <
t Ts /2 1 0  t  Ts
g(t) = ⇧ =
Ts :0 elsewhere

Subsequent symbols:

K
X
s(t) = Am,k p(t kTs )
k=0
 Am = 2m 1 M, m = 1, 2, . . . , M.

shaping filter
g(t)to islimit
the the duration
pulse shapingoffilter
the waveform
to limit the asduration
there would bewaveform
of the another symbol to would be another symbol to
as there
Proakis-27466 book September 25, 2007 13:13
the next Tbe
s seconds. Rectangular
transmitted window
in the next is one Rectangular
Ts seconds. solution: window is one solution:
8 8
✓ ◆ < ✓ ◆
t Ts /2 1 0  tt  TTss /2 < 1 0  t  Ts
Pulse Amplitude Modulation (PAM)
g(t) = ⇧ = g(t) = ⇧ =
Ts : 0 elsewhere
Chapter
Ts Three: Digital
: 0 Modulation Schemes
elsewhere

bols: Subsequent symbols:

K
X K
X
s(t) = Am,k p(t kTss(t)
) = Am,k p(t kTs )
k=0 k=0
XK K
X
= Am,k g(t kTs ) cos(2⇡f
= c (t AkT ))
m,ks g(t kTs ) cos(2⇡fc (t kTs ))
k=0 k=0
✓X
K ◆ ✓X K ◆ (a)

= Am,k g(t kTs ) cos(2⇡f

= c t)Am,k g(t kTs ) cos(2⇡fc t)
k=0 k=0
| {z } | {z }
Baseband PAM Signal Baseband PAM Signal (1) (1)

(b)

FIGURE 3.2–2
Example of (a) baseband and (b) carrier-modulated PAM signals.

where ĝ(t) is the Hilbert transform of g(t). Thus, the bandwidth of the SSB signal i
 | {z }
is-27466 book September 25, 2007 13:13 A2m Eg M
= 1 ZX T + Eg X ZMT 2 0 ✓
Eavg. = 2 E2m = Am
Em2 =M sm (t)dt2M = A2m g 2 (t) cos2 (2⇡fc t)dt co
Am Eg m=1
0
m=1
0
= Z
M T Z T ✓ ◆
2 Eg 1X 1 E g M (M 2
1)
= = A2m g 2 (t)dt
(2m 1 M +)2 = A2m g 2 (t) cos(4⇡fc t)dt
2M2m=1 2 2M 3
Pulse Amplitude Modulation (PAM)
0
| 0 {z }
2
Eg (M2 1)M
100 = = Am Eg1 X +E XM
0 Digita
Eavg. =62
g
Em = A2m
• Geometrical Representation/ Constellation A2m EM g m=1
2M m=1
=
2 Eg X M FIGURE ✓ 3.2–1 2
◆
E M (M 1)
1 c t)MConstellation for PAM
g
= / g(t)(2m cos(2⇡f )2 =
2M m=1 2M 3
Eg (M 2 1) M M
(a)
= 1 X Eg X 2
Eavg. 6= Em = A
M m=1 2M m=1 m
M ✓
Eg X E g M (M
(2m c1t) M )2 =
= / g(t) cos(2⇡f
2M m=1 2M
Eg (M 2 1)
(b) =
6

/ g(t) cos(2⇡fc t)

(c)

signal amplitudes differ by one binary digit as illustrated in Figure 3.2

 2

Pulse Amplitude Modulation (PAM)

Z T Z T ✓ ◆
1 + cos 2x
Em = s2m (t)dt = A2m g 2 (t) cos2 (2⇡fc t)dt cos2 x =
Z T Z T ✓ 2 ◆
• Signal Energy: 0
Z T
0
Z
Em1 = 2 2 s2m (t)dt1 = T 2 A22m g 2 (t) cos2 (2⇡fc t)dt cos2 x =
1 + cos 2x
= A g (t)dt + A g (t) cos(4⇡fc t)dt 2
2 0 m0Z T 2 0 0 m Z T
1 2 2
| 1 {z 2 2 }
=
A2m Eg 2 0 Am g (t)dt + A m g (t) cos(4⇡f c t)dt
= + 2
2 | 00 {z }
2
A2m Eg Am Eg
= = + 0
2 2
A2m Eg
=
2
M M
1 X Eg X 2
Eavg. = Em = A
• Average Energy: M m=1 2M m=1 m
M ✓ ◆
Eg 2 X 2 Eg 2 ( M (M 2 1)
= (2m 1 M) =
2M m=1 2M 3
2
Eg (M 2 1)
=
6

/ g(t) cos(2⇡fc t) (2)

 Eg (M 1)
=
6

/ g(t) cos(2⇡fc t) (2)

Phase Shift Keying (PSK)

sm (t) = g(t) cos(2⇡fc t + ✓m + ✓0 ) 0  t  Ts
2⇡
✓m = (m 1), m = 1, 2, . . . , M
M

where ✓0 is the offset phase which can be selected as anything. For convenience let us assume ✓0 = 0 for
now.

✓ ◆
2⇡
sm (t) = g(t) cos 2⇡fc t + (m 1)
M
✓ ◆ ✓ ◆
2⇡ 2⇡
= cos (m 1) g(t) cos(2⇡fc t) sin (m 1) g(t) sin(2⇡fc t)
M M

which means that PSK can be seen as the combination of two PAM’s on two orthogonal carriers. Why?
«A Little Paranthesis on Signals» 3

The inner product of two signals x(t) and y(t) is denoted by

Z 1
hx(t), y(t)i = x(t)y(t)dt
1

if the signals are real valued and

Z 1
hx(t), y(t)i = x(t)y ⇤ (t)dt
1

if they are complex-valued. Also:

Z 1
hx(t), x(t)i = |x(t)|2 dt = kx(t)k2 = Ex .
1

If hx(t), y(t)i = 0, x(t) and y(t) are said to be orthogonal. If in addition, Ex = Ey = 1 are, x(t) and y(t)
are said to be orthonormal.
The “Euclidean" distance between two signals d(x(t), y(t)) = dx,y is defined as:
✓Z 1 ◆2
p
d(x(t), y(t)) = dx,y = kx(t) y(t)k = |x(t) y(t)|2 dt = = Ex y .
1
if they are complex-valued. Also:

Z 1
hx(t), x(t)i = |x(t)|2 dt = kx(t)k2 = Ex .
1

If hx(t), y(t)i = 0, x(t) and y(t) are said to be orthogonal. If in addition, Ex = Ey = 1 are, x(t) and y(t)
«A
are saidLittle Paranthesis on Signals»
to be orthonormal.
The “Euclidean" distance between two signals d(x(t), y(t)) = dx,y is defined as:
✓Z ◆1/2
1 p
d(x(t), y(t)) = dx,y = kx(t) y(t)k = |x(t) y(t)|2 dt = Ex y .
1

Example: In PAM, the distance between two modulated waveforms sm (t) and sn (t):

✓ Z 1 ◆1/2 r r
Eg Eg
d(sm (t), sn (t)) = |Am An |2 g 2 (t) cos2 (2⇡fc t)dt = |Am An | = 2|m n| .
1 2 2
p
dmin = 2Eg .

Two more properties which are good to know!!!

kx(t) + y(t)k  kx(t)k + ky(t)k Triangle inequality

|hx(t), y(t)i|  kx(t)k · ky(t)k Cauchy-Schwartz inequality

Z 1
hg(t) cos(2⇡fc t), g(t) sin(2⇡fc t)i = g 2 (t) cos(2⇡fc t) sin(2⇡fc t)dt
Two more properties which are good to know!!!

kx(t) + y(t)k  kx(t)k + ky(t)k Triangle inequality

|hx(t), y(t)i|  kx(t)k · ky(t)k Cauchy-Schwartz inequality

Back to PSK!!!
Z 1
hg(t) cos(2⇡fc t), g(t) sin(2⇡fc t)i = g 2 (t) cos(2⇡fc t) sin(2⇡fc t)dt
1
Z T
1
= g 2 (t) sin(4⇡fc t)dt (sin 2x = 2 sin x cos x)
2 0
= 0

which implies g(t) cos(2⇡fc t) and g(t) sin(2⇡fc t) are orthogonal to each other.
 1 Z 1properties whichkx(t)
Two more are good to know!!! ✓
+ y(t)k  kx(t)k + ky(t)k ✓Z 1 T
hg(t) cos(2⇡fc t), g(t)Zsin(2⇡f t)i = g 2
(t) cos(2⇡f t) sin(2⇡f t)dt d(s (t), s (t)) = |A A
1 T 2c
m n m
c y(t)) = d
d(x(t), c
x,y = kx(t) y(t)k = |x(t)n
Z
1 |hx(t), y(t)i|  kx(t)k · ky(t)k 1 Ca
= g (t) sin(4⇡f
1 T 2c t)dt kx(t) + y(t)k  kx(t)k + (sin
ky(t)k 2x = 2Triangle sin xinc
2 0 = Example:
2 0
g (t) sin(4⇡f
In PAM, the t)dt (sin 2x =
c distance between two modulated waveform
|hx(t), y(t)i|  kx(t)k · ky(t)k
2 sin x cos x)
Cauchy-Sch
= 0 = 0 Z 1
Two more properties which are go
Chapter Three:hg(t) Digital Modulation
✓
cos(2⇡fc t), g(t) sin(2⇡fc t)i =Z Schemes
1
2
g (t) cos(2⇡f ◆2 t)
c

PSK Constellation
2 2 12 (2⇡fc t)dt
which implies g(t) cos(2⇡fc t) and g(t) sin(2⇡f aresnorthogonal
d(scmt)(t), (t)) = |Ato m eachAn |other.
Z 1 g (t) cos
Z Tkx(t) + y(t)k =
which implies g(t) cos(2⇡fc t) and g(t) sin(2⇡fc t) are orthogonal
cos(2⇡fc t), g(t)to each c t)i other.
1
1
hg(t) sin(2⇡f = g=2 (t) cos(2⇡f g 2p FIG
c t)sin(4⇡f
(t) sin(2⇡f cct
2 |hx(t), y(t)i| 
g Sign
1 d 0 = 2E .
4 Z T min
= 0
2-D constellation! =
1
g 2 (t) sin(4⇡fc t)dtand 8
Two more properties which are good2 to0 know!!!
which implies g(t) cos(2⇡fc= t) and
0 g(t) sin(2⇡fc t) are orthogo
✓Z T ◆1/2 kx(t) + y(t)k hg(t) cos(2⇡f
kx(t)k c t), g(t) sin(2⇡f
+ ky(t)k c t)
Triang
2
d(sm (t), sn (t)) = g 2 (t) cos(2⇡fc t + ✓m ) cos(2⇡fc t + ✓n dt which implies g(t) cos(2⇡fc t) and g(t) sin(2⇡fc t) are orthogonal to ea
0 |hx(t), y(t)i|  kx(t)k · ky(t)k Cauchy
q
= Eg [1 cos(✓m ✓n )].
s  ✓ ◆ Z 1
2⇡
dmin = Eg 1 cos hg(t) cos(2⇡fc t), g(t) sin(2⇡fcwhich
t)i = impliesgg(t)
2 cos(2⇡fcct)
(t) cos(2⇡f and
t) sin(
M 1
✓ ◆ Z T
p ⇡ 1
= g 2 (t) sin(4⇡fc t)dt
= 2Eg sin (cos 2x = 1 2 sin2 x). 2 0
M
= 0

which implies g(t) cos(2⇡fc t) and g(t) sin(2⇡fc t) are orthogonal t

and the minimum distance corresponding to |m −

! " #
 dmin = cos=
Eg 1 Eg [1 cos(✓m ✓n )]. sm (t) = Ami g(t) cos(2⇡fc t) A
s 
M ✓ ◆
✓ ◆ 2⇡
p d ⇡= E 1 cos
= 2Eg sinmin g (cos 2x = 1
M 2 sin2 x). = rm g(t) cos(2⇡fc t + ✓m
M ✓ ◆
p ⇡
= 2Eg sin (cos 2x = 1 2 sin2 x).
Quadrature Amplitude Modulation (QAM) M

t) =
-Joint Amplitude and
A g(t) cos(2⇡f t) A g(t) sin(2⇡f t)
mi c mq
Phase
0tT
c
Modulation- s
q ✓ ◆
Amq Ami = (2i
= rm g(t) cos(2⇡f
sm (t) =c t +A✓mimg(t) where c t)rm =
) cos(2⇡f A2misin(2⇡f
Amq g(t) + A2mqc t)
, ✓m 0=tan 1
t  Ts Ami . (3)
q ✓ ◆
Amq
= rm g(t) cos(2⇡fc t + ✓m ) where rm = A2mi + A2mq , ✓m = tan 1
. A
(3)mq = (2q
Ami

M = M1 M
k
Ami = (2i 1 M1 ) i = 1, . . . , M1 = 2 ,

Amq = (2q 1 M2 ) q = 1, . . . , M2 = 2n ,
Rectangular QAM
M = M1 M2 = 2n+k . (4)
 1
M 2 m=1 n=1 ✓ Z 1 ◆2 r
if they are complex-valued. Also: Eg
If hx(t), y(t)i = 0, x(t) and y(t) are said to be orthogonal. If in addition, are, and
g m (t), sn2M(M − A1)n| 1 g (t) cos (2⇡fct)dt = |Am
E = E = 1 x(t) y(t)
Ed(s x (t)) y= |Am 2 2 2
An |
2
= 2|m
are said to be orthonormal. = × (3.2–42)
2M
The “Euclidean" distance between two signals d(x(t), y(t)) = dx,y is defined as: Z 1 3
dmin
p
= 2Eg .
✓Z 1 M −1
hx(t),
◆2x(t)i = |x(t)|2 dt = kx(t)k2 = Ex .
d(x(t), y(t)) = dx,y = kx(t) y(t)k = = E
Two more properties pwhich are good to know!!!
|x(t) y(t)|2 dt g= Ex y . 1
Eavg. =
3
Rectangular QAM
Example: In PAM, the distance between
If hx(t),
twoy(t)i
modulated
1

x(t) and sy(t)

= 0, waveforms m (t) are
kx(t) + y(t)k  kx(t)k
and said EgM1X
M1 X
sn (t):to be orthogonal. M+ 2 ky(t)k
If in
2 addition,
2
Triangle inequality
Ex = Ey = 1 are, x(t) = a
EFIGURE
avg. = g 3.2–5
|hx(t), y(t)i|E2M
X XM 2
(A + A )
(A2mi + A2mq ) Cauchy-Schwartz
kx(t)k · ky(t)k inequality
mi mq M1 X M2
Eavg. = Eg X
r Several2M signal space diagrams
Eavg. = for rectangular (A2mi + A
i=1 q=1
✓ are
Z 1said to be orthonormal. ◆2 ri=1 q=1 !
Eg M! 2 2M i=1 q=1
Eg E2g . X
M1
d(sm (t), sn (t)) = |Am An |2 g 2 (t) cos2 (2⇡fc t)dt = |Am An | QAM. = 2|m En| M1 X 2 M
X 2
X
2
=
The “Euclidean" distance between two signals 2 = g 2 M A Mis + M A
1 definedAmq Eas:
Z 1 A=
2 2 1 2 M
X
1 = d(x(t),
2MM2y(t)) mi d
mi
+x,y g
mq2 1
2
p 2M 2
i=1
hg(t) cos(2⇡fc t), g(t) sin(2⇡fc t)i = ✓ i=1 g (t) cos(2⇡f =i=1 M A mi +
c t) sin(2⇡fc t)dt
i=1 2 ◆
1) =
dmin = 2Eg . ✓2 2 2M 2◆
EgEg2 ✓Z M Z 1TM
1
(M11(M 2
11) 1) M ◆ 2 (M
M222 (M
✓ 1)
2
i=1
== M 1M1 2 + M+ 1 ME
1/2
1 g 2 p M1 (M 2
2M2M = 2 g 23(t) sin(4⇡f
3 t)dt
2 = 3= M 3E2(sin .2x =1 2 s
Two more properties which are good to know!!! d(x(t), y(t)) = dx,y = kx(t) y(t)k = 2 0
|x(t) y(t)|c dt 2M x y 3 p
EgEg2 = 2
1
== (M(M02 2
1 1 1) +
For
1)(M
2square
2
+2(M21) 1) Eg QAM, 2 M 1 = M 2 = M
kx(t) + y(t)k  kx(t)k + ky(t)k Triangle inequality 6 6 = 2
(M1 1) + (M
6
|hx(t), y(t)i| Example:
kx(t)k · ky(t)k thewhich impliesbetween
In PAM, Cauchy-Schwartz
distance two
g(t) cos(2⇡f
inequality
t) and g(t) sin(2⇡fc t) are orthogonal
pcpmodulated waveforms sm (t) and to eachsnother.
(t):
For square QAM, M = M = M and
For square QAM, M11 = M2 2 = For and QAM, M = M = pM and
Msquare
1 2

Z 1
Eg 2 2
Eavg. = ◆1/2 1) polar r
(M the
EUses Eg for2 r
hg(t) cos(2⇡fc t), g(t) sin(2⇡fc t)i = g 2 (t) cos(2⇡fc t)✓sin(2⇡fc t)dt Z 1
E 3 g representation Q
2 2 2 avg. = (M 1) EgEavg. = (M
d(sZm (t),
1 sn (t)) = |Am An | g (t) cos (2⇡fc t)dt 3 |Am An |
= = 2|m 3n|
1 T
1 2
PSK a different offset phase angle
= g 2 (t) sin(4⇡f
Uses thec t)dt (sin 2x
polar representation for=QAM.
2 sin x Concentric
cos x) PSK constellations for each PAM amplit
2 0
= 0 p
PSK a different offset phase angledcan = introduced.
min be 2Eg . Same as PAM.
Same as PAM.Why?
Why?

which implies g(t) cos(2⇡fc t) and g(t) sin(2⇡fc t) are orthogonal to each other.

Two more properties whichsmare

(t) good
= g(t) cos(2⇡(fc + m f )t)
to know!!! m = 1, 2, . . . , M, 0  t  Ts

or m = ±1, ±2, . . . , ±M/2

 6

Chapter Three: Digital Modulation

E
Schemes
M X
Eg X
M
(A2 + A2
1 2

= avg. mq )

Circular QAM
mi
2M i=1 q=1
M1 M2
!
Eg 2 X X
= M2 A2mi + M1 A2mq
2M i=1 i=1
2
✓ ◆
Eg M1 (M12 1) M2 (M22 1)
= M2 + M1
2M 3 3
Eg 2
= (M12 1) + (M22 1)
6

p
For square QAM, M1 = M2 = M and

Eg 2
Eavg. = (M 1)
3

Uses3.2–4
FIGURE the polar representation for QAM. Concentric PSK constellations for each PAM amplitude. For each

Examples
PSK of combined
a different PAM-PSK
offset phase angle canconstellations.
be introduced.

In the special case where the signal amplitudes take the set of discrete va
{(2m − 1 − M), m = 1, 2, . . . , M}, the signal space diagram is rectangular, as sh
in Figure 3.2–5. In this case, the Euclidean distance between adjacent points, i.e.,
 ✓ ◆p
p Eg 2
Eg 2For square
M1 (M12 QAM,
1) M (M 2
1) and
For
2 square QAM, + M1 M1 = M2 =2 M
M = M = MEand
avg. = (M 1)
1 2 2 2
= M
2M 3 3 Eg
Eavg. = (M 1) 3
Eg 2 2 2 3
= (M1 1) + (M2 1)
6 Egg 22
EEavg.
avg. =
1)
1) = (M
(M
3
eses pthe
polar polar representation
representation
M = M and
for QAM.
for QAM. Concentric Concentric
PSK constellations for PSK constellations
each PAM amplitude. For for
eacheach PAM amplitude. For each
2

different offset
Frequency Shift Keying
Uses
SK a different
phase
Uses
the angle
the
Egpolar
offset
Eavg. =
3
can
2polar
be introduced.
representation
representation
phase
(M
angle for
1)
Same
for
canQAM.as Concentric
QAM. PAM. Why? PSK constellations
Concentric
be introduced. constellations
Same as PAM. for each
eachPAM
forWhy? PAMamplitude.
amplitude.ForFor
each
each

PSK
PSK a differentoffset
a different offsetphase
phaseangle
anglecan
can be
be introduced.
introduced. Same
Same as
as PAM.
PAM.Why?
Why?
sm (t) = g(t) cos(2⇡(fc + m f )t) m = 1, 2, . . . , M, 0  t  Ts
on for QAM. Concentric PSK constellations for each PAM amplitude. For each
sm (t) = asg(t)
se angle can be introduced. Samesm
cos(2⇡(forc +
PAM.=Why?
m±1,f±2,
m =c +
)t). . . , ±M/2 m = 1, 2, . . . ,(5) M, 0t  Ts
s (t) = g(t) cos(2⇡(f + m ff)t)
(t) g(t) cos(2⇡(f m )t) m = 1, 2, . . . , M,
m = 1, 2, . . . , M, 0  
0  t TsT
t
m c s

(t) cos(2⇡(f
hould c + m f )t)
f be? m = 1, 2, . . . , M, 0  t  Ts or
or m ±1,
m ==±1, ±2,±2, . . , ±M/2
. . . ,.±M/2
or m = ±1, ±2, . . . , ±M/2
(5)
(5)
(5)
or m = ±1, ±2, . . . , ±M/2 (5)
t f to be as small as possible and also such that all sm (t)’s are orthogonal to each other 8m.
What should f be?
What should f be?
= 2Tk s where k is an integer satisfy the orthogonality (Proove yourself!!). So f = 2T1 s is sufficient.
We want f to be as small as possible and also such that all sm (t)’s are orthogonal to each other 8m.
all as possible We wantsuch fthattoallbesmas
and also (t)’ssmall as possible
are orthogonal to eachand
otheralso
8m.such that all sm (t)’s are orthogonal to each other 8m.

f = k where
integer satisfy the orthogonality (Proove is an integer
k yourself!!).
Eavg. = ESo f E=g 2T1 s isthe
m =satisfy orthogonality
sufficient. (Proove yourself!!). So f= 1
is sufficient.
2Ts 2Ts
2

Eg p
Eavg. = Em = dm,n = Eg
2