OpenCourseWare

Digital Communications
Telecommunications Engineering

Chapter 2
Pulse amplitude (linear) modulations

Marcelino Lázaro

Departamento de Teorı́a de la Señal y Comunicaciones

Universidad Carlos III de Madrid

Creative Commons License

1 / 113

Index of contents
Baseband PAM modulations
I Constellations and transmission filters
I Power spectral density
I Equivalent discrete channel
F Transmission through Gaussian channels
F Transmission through linear channels
Bandpass PAM modulations
I Generation of bandpass modulated signals
I Constellations
I Power spectral density
I Equivalent discrete channel
F Transmission through Gaussian channels
F Transmission through linear channels

c Marcelino Lázaro, 2013 Digital Communications Linear modulations 2 / 113

Pulse amplitude (linear) modulations
Linear modulation in a N-dimensional signal space
X NX1
s(t) = Aj [n] · j (t nT)
n j=0

I Information is linearly conveyed

F In the amplitude of the set of functions { j (t)}Nj=01
I Encoder: A[n]
F Constellation in a space of dimension N
F Designed considering energy (Es ) and performance (Pe ,
BER)
- Es : mean energy per symbol (Es = E[|A[n]|2 ])
- Pe : probability of symbol error
- BER: bit error rate
I Modulator: { j (t)}Nj=01
F Designed considering channel characteristics
F Ideally: the only distortion appearing after the transmission is
additive noise (white and Gaussian)
c Marcelino Lázaro, 2013 Digital Communications Linear modulations 3 / 113

Baseband PAM modulation

One-dimensional modulation: N = 1
X
s(t) = A[n] · g(t nT)
n

PAM (Pulse Amplitude Modulation)

ASK (Amplitude Shift Keying)

Constellations - Normalized: A[n] 2 {±1, ±3, · · · , ±(M 1)}

Examples: 2-PAM (a), 4-PAM (b), 8-PAM (c)

r1 +1 r r3 r1 +1 r +3
r
0 A[n] 0 A[n]
(a) (b)
r7 +5
r r3 r1 +1 r +3
r +5
r +7
r
0 A[n]
(c)

c Marcelino Lázaro, 2013 Digital Communications Linear modulations 4 / 113

PAM modulation as a filtering process
Signal of symbols: impulses with amplitudes A[n]
X
a(t) = A[n] · (t nT)
n

Generation of PAM signal

s(t) = a(t) ⇤ g(t)
A[1] A[3]
6 6
A[0] A[2]
6 6 a(t)
- - s(t) = a(t) ⇤ g(t)
Modulator
-
0 T 2T 3T
B[`] - A[n] - a(t) - s(t) -
Encoder g(t)

c Marcelino Lázaro, 2013 Digital Communications Linear modulations 5 / 113

Selection of g(t) waveforms

Waveform g(t) tipically receives two names:
I Transmitter filter
I Shaping pulse (although it is not necessarily a pulse)
Selection to be able of identify sequence A[n] from s(t)
I Pulses with duration limited to symbol period T
F No overlapping between waveforms delayed nT seconds
1 ⇣t⌘
ga (t) = p · ⇧
T T
F Symbol A[n] determines signal amplitude in its associated
symbol interval
I Pulses with higher length
F Overlapping: non-destructive interference at some point each
T seconds ⇣t⌘
1
gb (t) = p · sinc
T T
F Symbol A[n] determines signal amplitude at the
nondestructive point associated to its symbol interval
c Marcelino Lázaro, 2013 Digital Communications Linear modulations 6 / 113
Sinc pulses - Pulses shifted T seconds
1 ..... ......... ......... ......... ......... ......... ......... ......... ......... .........
... .. ... .. ... .. ... .. ... .. ... .. ... .. ... .. ... .. ...
... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ...
0.8 .... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ...
...... ..... ..... ..... ..... ..... ..... ..... ..... ...
. .
. . .
. . .
. . .
. . .
. . .
. . .
. . .
. . ...
0.6 ........ ....... ....... ....... ....... ....... ....... ....... ....... ...
nT)

.... ... .... .... .... .... .... .... .... .... .... .... .... .... .... .... .... .... ...
. .... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ...
..
0.4 .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. ...
... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ...
...
T ⇥ g(t

.. ... ... .... ... .... ... .... ... .... ... .... ... .... ... .... ... ....
. ..
0.2 .. ... .. ... .. ... .. ... .. ... .. ... .. ... .. ... .. ...
.. .............. .. ............. .. .............. .. .............. .. .............. .. .............. .. .............. .. .......... .. .. .......... .. .......... .....
..u.......................................u...................................u......................................u...............................u..............................u...............................u........................................u.....................................u.......................................u.....................................u
.
.
0 ................................................................................................................................... ...........................................................................................................................................................
p

............ .. ................... ................... .................. ................... ................... .................... .................... ... ........... ...............
.... ... ....... ...... ....... ...... ....... ...... ....... ...... ....... ...... ....... ...... ....... ...... .... ... ... ...
-0.2 ..... ........ ........ ....... ........ ....... ....... ....... ..... ......
u ⌘ zeros at nT (sampling instants)
-0.4
0 1 2 3 4 5 6 7 8 9 10
t/T

c Marcelino Lázaro, 2013 Digital Communications Linear modulations 7 / 113

Sinc pulses - Contribution of each symbol

n 0 1 2 3 4 5 6 7 8 9
Sequence: A[n] +1 1 1 +1 +1 +1 1 +1 1 +1

1 ...e.... ...e... ...e... ...e... ...e... ...e...

... .... .... .... .... .... .... .... .... .... ....
0.8 .... ... .... .... .... .... ....
. ... ....
. . .. ....
.
.. . ... ... ... ... ...
0.6 ..... ..... ........ ........ ... .... .
.... .... ...
... ... .. ...
nT)

... ... ... ... ... ... . . .. .. ...

0.4 ... . . ... .. . ... .. .
... .... .... ..... .... ..... ... .. ... .. ...
. . .
0.2 ........................ ........ ........................ .................... ........................... ........ .... ........................ ......................... ... . .
........................
T ⇥ A[n] · g(t

. u. .....................................u..... ......u..................................u.. ........u..............................u.........................u..............................u................................u............................u...........................................u

0 ............................................................ ............................................... ..............................................................................................................................................................
... . .. . ..
.......... ...... .... .... ............. ............. ....................... .... .. ... ....... .. ............ ...
-0.2 ... . ............. ......... ..... ..... .......... ..... ... ... ...... ..... ...
... ... .. .. ... .. .... ..
-0.4 .. . .
... .. .
. .
. .
... .
. ..
... ..... ...
. ... ... .....
. ...
.
-0.6 .... .... .. e ... .. ... ..
p

.. .
. .. .
.
. ⌘ A[n] at nT ... .. .
... . .
-0.8 ..... ..... ..... ..... u ⌘ zeros at nT.... .....
. ... ....
... ...
...e. ..e. ....e. .....e..
... .. .... .. ... ..
-1
0 1 2 3 4 5 6 7 8 9 10
t/T
c Marcelino Lázaro, 2013 Digital Communications Linear modulations 8 / 113
Sinc pulses - Waveform
n 0 1 2 3 4 5 6 7 8 9
Sequence: A[n] +1 1 1 +1 +1 +1 1 +1 1 +1

2
......
1.5 .... .....
.. A[5] .........
1 ....s.............. .............s....................s......................s........... ............s......... ..........s..........
A[0] A[3] A[4]
. ... A[7]
.
A[9]
. .
...... ........ .............. ... .. ... .... ..... ....
. ... .... .........
.
.... .
. .
.
... ... ..... ... ...
.
. .. ... ... ....
.
.
0.5 ... .. ... ... . ..... ... ..
....... ..... ... .....
T ⇥ s(t)

...... . ... .. . ...... . ... ...

.
... ... . .
. .... .
... .... . .
.... .... .
.
.... . ...............
0 ... .... .. ... ... .. ..... . .
.
... ...
..
..... ..... ..... ... ...
. . ... .....
.
.... ... ... . .
... .. .....
-0.5 .. ...
p

... .. .
.
. ... .
. .
..... .. .. . . ... .... ... ...
-1 ................s....................s.........
. .
. ................s.........
. . ...............s..........
.
. .
. .
A[1] .. . .. . . ........... ga (t)
... ... A[2] A[6] A[8]

-1.5 .
... .. . ............... gb (t)
... ...
...
-2
0 1 2 3 4 5 6 7 8 9 10
t/T
c Marcelino Lázaro, 2013 Digital Communications Linear modulations 9 / 113

Waveforms - Another example

n 0 1 2 3 4 5 6 7 8 9
Sequence: A[n] 1 +1 +1 1 +1 1 1 1 +1 1

2 ....
.... ..... ............... ga (t)
... ...
.
1.5 . .... ............... gb (t)
..... ...
..................s.......................s............. ...................s............... ....................s................
A[1]. ..A[2] A[4] A[8]
1
. ... ... ..... .... ... ...... .... ....
..... ..... . . .
. . . .
T ⇥ s(t)

... ... . . ... ... . ......

0.5 .. ..
... ... ... ... .... .
.. .
.
.....
.
. ...
. ... ... .. . .
. .
.
.... .
....
. .
..... ...
0 .
. .
... .. . .
. ..... .
. ....... ................
........ ... .. . .
.
. .
.
. ...... .
.
. .
.
. .
... .. .
. .
.
.. ... ..... .. ....
p

........ ... ... .... .... ... .... . . ..

... ... .. .. ..
-0.5 ... ...... ... ... ... ... .
... ... ... ... ... ...
........ ..... .. .. .
... ... .
. .
. .
. ... ... ......
-1 ....s...... .............s........ ..............s......................s.....................s........ ...............s............
. .... ... ... .. ...
. .
... .......... ..... .. . .......
A[0] A[3] A[5] A[6] ..... A[7] A[9]

-1.5
0 1 2 3 4 5 6 7 8 9 10
t/T
c Marcelino Lázaro, 2013 Digital Communications Linear modulations 10 / 113
Spectrum of a baseband PAM
PAM baseband signal
X
s(t) = A[n] · g(t nT)
n

Let {A[n]}1
n= 1 be a sequence of random variables (stationary
random process):
I E[A[n]] = m
I E[|A[n]|2 ] = Es
I E[A[k] · A⇤ [j]] = RA [k j] = RA [j k]
I Power spectral density function of A[n] is
1
X
SA (ej! ) = RA [n] · e j!n

n= 1

Let g(t) be any deterministic function with Fourier transform G(jw)

c Marcelino Lázaro, 2013 Digital Communications Linear modulations 11 / 113

Review: Wiener-Khinchin theorem

Power spectral density

def |XT (j!)|2 E[|XT (j!)|2 ]
SX (j!) = E lı́m = lı́m .
T!1 T T!1 T

Wiener-Khinchin theorem
If for any finite value ⌧ ans any interval A, of length |⌧ |, the
autocorrelationof random process fulfills
Z
RX (t + ⌧, t)dt < 1,
A

power spectral density of X(t) is given by the Fourier transform of

Z T/2
1 def
< RX (t + ⌧, t) > = lı́m RX (t + ⌧, t) · dt.
T!1 T T/2

c Marcelino Lázaro, 2013 Digital Communications Linear modulations 12 / 113

Corollary of Wiener-Khinchin theorem
Corollary 1: If X(t) is an stationary process and ⌧ RX (⌧ ) < 1 for all
⌧ < 1, then
SX (j!) = TF[RX (⌧ )].

Corollary 2: If X(t) is cyclostationary and

Z To
RX (t + ⌧, t)dt < 1,
0

then
eX (⌧ )],
SX (j!) = TF[R
where Z To /2
eX (⌧ ) = 1
R RX (t + ⌧, t) · dt,
To To /2

and To is the period of the cyclostationary process.

c Marcelino Lázaro, 2013 Digital Communications Linear modulations 13 / 113

Mean and autocorrelation of a baseband PAM

Definition of random process for PAM signal
1
X
X(t) = A[n]g(t nT)
n= 1

Mean of random process X(t)

" #
X X X
mX (t) = E A[n]g(t nT) = E[A[n]]g(t nT) = m g(t nT)
n n n

Autocorrelation function of random process X(t)

RX (t, t + ⌧ ) = E[X(t)X ⇤ (t + ⌧ )]
2 !0 13
X X
= E4 A[k]g(t kT) @ A⇤ [j]g⇤ (t + ⌧ jT)A5
k j
XX
= E[A[k]A⇤ [j]]g(t kT)g⇤ (t + ⌧ jT)
k j
XX
= RA [k j]g(t kT)g⇤ (t + ⌧ jT)
k j
c Marcelino Lázaro, 2013 Digital Communications Linear modulations 14 / 113
Cyclostationarity
Mean is a periodical function of t (period T)
X X
mX (t + T) = m g(t + T nT) = m g(t (n 1)T)
n n
X
=m g(t jT) = mX (t)
j

Autocorrelation is a periodical function of t (period T)

RX (t + T, t + ⌧ + T) =
XX
= RA [k j]g(t + T kT)g⇤ (t + T + ⌧ jT)
k j
XX
= RA [k j]g(t (k 1)T)g⇤ (t (j 1)T + ⌧ )
k j
XX
= RA [m + 1 (n + 1)]g(t mT)g⇤ (t nT + ⌧ )
m n
XX
= RA [m n]g(t mT)g⇤ (t nT + ⌧ ) = RX (t, t + ⌧ )
m n

c Marcelino Lázaro, 2013 Digital Communications Linear modulations 15 / 113

Time average of autocorrelation function

Z T
1
R̃X (⌧ ) = RX (t, t + ⌧ )dt
T 0
Z T XX
1
= RA [k j]g(t kT)g⇤ (t + ⌧ jT)dt
T 0 j
k
1 1 Z T
1 X X
= RA [ m] g(t kT)g⇤ (t + ⌧ (k + m)T)dt
T 0
k= 1 m= 1
X1 1 Z
X (k 1)T
1
= RA [m] g(u)g⇤ (u + ⌧ mT)du
T m= 1 kT
k= 1
X1 Z 1
1
= RA [m] g(u)g⇤ (u + ⌧ mT)du
T m= 1 1
1
X
1
= RA [n]rg (nT ⌧ ),
T n= 1

rg (t) = g(t) ⇤ g⇤ ( t)
c Marcelino Lázaro, 2013 Digital Communications Linear modulations 16 / 113
Power spectral density (PSD)

1
1 X
R̃X (⌧ ) = RA [n] · rg (nT ⌧)
T n= 1
X1
1
= rg (⌧ ) ⇤ RA [n] · (⌧ nT)
T n= 1
X1
1 ⇤
= · g(⌧ ) ⇤ g ( ⌧ ) ⇤ RA [n] · (⌧ nT)
T n= 1

SX (j!) = FT R̃X (⌧ )
X1
1 ⇤ j!nT
= · G(j!) · G (j!) · RA [n] · e
T n= 1
1
= · |G(j!)|2 · SA (ej!T )
T

c Marcelino Lázaro, 2013 Digital Communications Linear modulations 17 / 113

Power spectral density (II)

1
Ss (j!) = · SA (ej!T ) · |G(j!)|2
T

Three contributions:
I A constant factor scale: T1 = Rs bauds
I A deterministic component given by g(t): |G(j!)|2
I A statistical component given by A[n]: SA (ej! )
For white sequences A[n] (the most typical case)

RA [n] = Es · [n], SA (ej! ) = Es = E |A[n]|2

Es
Ss (j!) = · |G(j!)|2
T

I g(t): Shaping pulse (determines the shape of spectrum)

c Marcelino Lázaro, 2013 Digital Communications Linear modulations 18 / 113
Spectrum for white sequences - PSDs
✓ ◆ ✓ ◆
p !T p !T
Ga (j!) = T · sinc , Gb (j!) = T ·⇧
2⇡ 2⇡
Es
..........................................
... ... ... ...
... ... .... ...
... ... ... ...
... ... . ... ... ............. ga (t)
... .. .
... .. . ............. gb (t)
... ... ... ...
... ... ... ...
... .. . ... ..
..... ......
.... ....
..
......... .....
.....
... .... ... ...
.... ... ... ....
... ... ... ...
... ... ... ....
.
....
.
.
....
.. .
...................................................... .................... .. . ... ..... ................................ ......................
....... .....................
5⇡ 4⇡ 3⇡ 2⇡ ⇡ ⇡ 2⇡ 3⇡ 4⇡ 5⇡
T T T T T 0 T T T T T !
c Marcelino Lázaro, 2013 Digital Communications Linear modulations 19 / 113

PSD for coloured data sequence

PSD shape can be modified by introducing correlation in
the sequence
B[n]
-g - g(t) s(t)-
A[n]- 6
A[n 1]
-z 1

White sequence A[n]: 2-PAM (A[n] 2h {±1})

i
I Mean energy per symbol: Es = E |A[n]|2 = 1
Coloured sequence
B[n] = A[n] + A[n 1]
X1
s(t) = B[n] · g(t nT)
n= 1
c Marcelino Lázaro, 2013 Digital Communications Linear modulations 20 / 113
Autocorrelation function of B[n]
Autocorrelation of A[n]: RA [k] = Es · [k] = [k]
Autocorrelation function of B[n]
RB [k] =E [B[n]B⇤ [n + k]]
=E [(A[n] + A[n 1]) · (A[n + k] + A[n + k 1])]
=E [A[n]A[n + k]] + E [A[n]A[n + k 1]]
+ E [A[n 1]A[n + k]] + E [A[n 1]A[n + k 1]]
=RA [k] + RA [k 1] + RA [k + 1] + RA [k]
=2RA [k] + RA [k 1] + RA [k + 1]

2 s
6RB [k]

s 1 s

-
-1 0 1 k
c Marcelino Lázaro, 2013 Digital Communications Linear modulations 21 / 113

Power spectral density

PSD for sequence B[n]
X
j! j!k
SB e =FT {RB [k]} = RB [k] · e
k
j!·0 j!·1 j!·1
=2 · e + e + e
=2 · [1 + cos(!)]

PSD for baseband PAM signal s(t)

This system transmits data sequence B[n]

1
SS (j!) = · SB (ej!T ) · |G(j!)|2
T
Evaluating the previously obtained expression for SB (ej!) ) in !T we have

2
SS (j!) = [1 + cos(!T)] · |G(j!)|2
T
c Marcelino Lázaro, 2013 Digital Communications Linear modulations 22 / 113
Power spectral density with ga (t)

..... ..... ........

.......... .........
...
................... |Ga (j!)|....2....
...
.... .... .... .... . .. . ..
. . ..
..... ..... ..... ..... .......... ..... .... ....................... SB (ej!T.....) .....
. .. . .. . .. .... .. . .. . ..
..... .... ..... .... ..... ......... ........ ..... ...................... SS (j!)..... ....
. ... . ... .. . ... .. ....
.... ... .... ... .... .......... ........ ....
. ....
. .
... .
. ...
.... ... ..
. ... .. .......
.
...... ...
. . ..
. . .
. .. ...
... ... .
.. ... ... ..... ....... ... ... ...
. .
.
. ...
... ... .
. ... .. .... ...... ... .. ... .
. ...
... ... ..
. ... ... ...... ..... ... ... .
. .
.
. ...
... ... ..
. ... ... ...... ...... ... ... ... .
.
. ...
... ... ...
. ... ... ...... ...... ... ... ...
.
.
. ...
... ... .. ... ... ...... ...... ... ... ... .
.. ...
... ... ... ...... ...... ....... ....... ... ..
. . ...
.
.
... .. .
. .
. . ...
. .. ..
...... .... .
.. .
.
... .. .
. ...
..... . . .... ...... . . ...
... .. . . ... ...... ...... .. ... ... .. ...
... ... .. . .
. ... .... ...... .. .. . . ... .. .
. . .
............... . .......... . . .. .............................................................................................................. ...........................................
. . . . .
................................................................................................................................... .......... . .... ....... ....
5⇡ 4⇡ 3⇡ 2⇡ ⇡ ⇡ 2⇡ 3⇡ 4⇡ 5⇡
T T T T T 0 T T T T T
!

c Marcelino Lázaro, 2013 Digital Communications Linear modulations 23 / 113

Power spectral density with ga (t)

.... 2Es
..... ....
. ..
..... .....
. . ............... SS (j!)
..... .....
. ..
..... ....
. ...
..... ...
...
... ...
..... ...
... ...
... ...
... ...
... ...
. ...
.... ...
.
..
..
. ... . ... ............
............................................................................................................ ........ ...... ...........................................................................................................
5⇡ 4⇡ 3⇡ 2⇡ ⇡ ⇡ 2⇡ 3⇡ 4⇡ 5⇡
T T T T T 0 T T T T T
!

c Marcelino Lázaro, 2013 Digital Communications Linear modulations 24 / 113

Power spectral density with gb (t)

..... ..... .............................................. ... ...

.... .... .... .... ... .... .... ... ................... |Gb (j!)|....2....
. . . . . . .. . ..
..... ..... ..... ..... .... ..... .... ..... ....................... SB (ej!T.....) .....
. .. . .. .. . .. .. . .. . ..
..... .... ..... .... .... ..... ..... .... ...................... SS (j!)..... ....
. ... . ... ... ... ... ... ... ... ....
...
.... ... .... ... ... .. ... .. .. ... . ...
... ... ... ... ... ... ... ... ... ... ..
. ...
.... ... ..
. ... ... ... ... ... ... ... ..
. ...
.
... . .
... .. .. . . ... .. .. . . .
... . ...
.... ... .
.
.
. ... .. .. . . ... .. .. . . ... .
.
.. ...
... ... .
.
. ... .. .. . . .
... ... .. . . ... .
.
. ...
.... ... .
.
.
. ... .. .. . . ... .. .. . ... .
.
.
. ...
... ... .. . ... .. .. . . ... .. .. . . ... .. . ...
... ... ... ... ... ... ... .. .. . . .
... ... . ...
... ... .. ... ... .. ... .. .. . . . . ...
... .. . ... .. .. . .
... .. . ...
..... ... .. . ... .. .. . . ... .. ...
.
... .. .. ... .. . ...
.. . . .. . . .
....... . . . ..
................................................................................................................................................................................. .............................................................................................................................................................................
5⇡ 4⇡ 3⇡ 2⇡ ⇡ ⇡ 2⇡ 3⇡ 4⇡ 5⇡
T T T T T 0 T T T T T
!

c Marcelino Lázaro, 2013 Digital Communications Linear modulations 25 / 113

Power spectral density with gb (t)

..... 2Es
.... .....
. ..
..... ....
. .. ............... SS (j!)
..... .....
. ...
..... ...
... ...
... ...
... ...
... ...
... ...
...
..... ...
... ...
... ...
... ...
. ...
.... ...
..................................................................................................................................................................
................................................................................................................................................................
5⇡ 4⇡ 3⇡ 2⇡ ⇡ ⇡ 2⇡ 3⇡ 4⇡ 5⇡
T T T T T 0 T T T T T
!

c Marcelino Lázaro, 2013 Digital Communications Linear modulations 26 / 113

Power of a baseband PAM modulation
Power can be obtained from Ss (j!)
Z 1
1
PS = Ss (j!) d!
2⇡ 1

For white symbol sequences A[n]

Z 1
Es 1
PS = · |G(j!)|2 d!
T 2⇡ 1
| {z }
E{g(t)}

I If g(t) is normalized, by applying Parseval’s relationship

Es
PS = = Es ⇥ Rs Watts
T

c Marcelino Lázaro, 2013 Digital Communications Linear modulations 27 / 113

Equivalent discrete channel

A- s(t)- r(t)- q-
Modulator Channel Demodulator

.................
.. ... ..
Equivalent q-
A-
Discrete
Channel
Provides the discrete time expression for observations at the
output of the demodulator q[n] as a function of the transmitted
sequence A[n]
I In ideal systems: q[n] = A[n] + n[n]
Gaussian distributions for observations (conditioned to A[n] = ai )
ai ||2
1 ||q
fq[n]|A[n] (q|ai ) = e N0
(⇡No )N/2

Expressions will now be obtained for two channel models

I Gaussian channel
I Linear channel
c Marcelino Lázaro, 2013 Digital Communications Linear modulations 28 / 113
Transmission of PAM signals over Gaussian channels

A[n]
- s(t) - i r(t) - q(t) q[n]
-
g(t) f (t)
6 ?
n(t) t = nT
Gaussian Channel

Gaussian channel model

I Distortion during transmission is limited to noise addition
r(t) = s(t) + n(t)
n(t): stationary random process, white, Gaussian, zero mean, Sn (j!) = N0 /2

Receiver filter f (t)

I Typical set up: matched filter
f (t) = g⇤ ( t) = g( t), because g(t) is real
Signal at the input of the sampler
q(t) = s(t) ⇤ f (t) + n(t) ⇤ f (t)
c Marcelino Lázaro, 2013 Digital Communications Linear modulations 29 / 113

Equivalent discrete channel for Gaussian channels

Signal before sampling
0 1
s(t)
z
BX }| {C
q(t) = B
@ A[k] · g(t kT)C
A ⇤ f (t) + n(t) ⇤ f (t)
| {z }
k
Filtered noise z(t)
| {z }
Noiseless output o(t)
X ✓ ◆ X
o(t) = A[k] · g(t kT) ⇤ f (t) = A[k] · p(t kT)
k k

p(t) = g(t) ⇤ f (t): joint transmitter-receiver response

I This joint response determines the noiseless output at the receiver

Observation at demodulator output

X
q[n] = q(t)|t=nT = A[k] · p ((n k)T) + z(nT)
k

c Marcelino Lázaro, 2013 Digital Communications Linear modulations 30 / 113

Equivalent discrete channel for Gaussian channels (II)
Definition of equivalent discrete channel p[n]
p[n] = p(t) t=nT
= o[n] + z[n]
X
Noiseless output o[n] = A[k] · p[n k] = A[n] ⇤ p[n]
k

A[n]-
p[n] - hq[n]
-
6
z[n]
Ideal: p[n] = [n] ! q[n] = A[n] + z[n]
Real: Intersymbol interference (ISI)
X
q[n] = A[n] · p[0] + A[k] · p[n k] + z[n]
k
k6=n
X
ISI = A[k] · p[n k]
k
k6=n
c Marcelino Lázaro, 2013 Digital Communications Linear modulations 31 / 113

Effect of ISI - Extended constellation

ISI produces an extended constellation at the receiver side
Values of noiseless discrete output o[n] = A[n] ⇤ p[n]

Example: 2-PAM modulation

Channel A Channel B
p[n] = [n] + 14 [n 1] p[n] = [n] + 12 [n 1] + 14 [n 2]
o[n] = A[n] + 14 A[n 1] o[n] = A[n] + 12 A[n 1] + 14 A[n 2]

A[n] A[n 1] A[n 2] o[n]

+1 +1 +1 + 74
A[n] A[n 1] o[n] +1 +1 1 + 54
+1 +1 + 54 +1 1 +1 + 34
+1 1 + 34 +1 1 1 + 14
3 1 1
1 +1 4
+1 +1 4
1 +1 1 3
5
1 1 4
4
5
1 1 +1 4
1 1 1 7
4

u A[n] = +1
u u
2-PAM constellation
u A[n] = 1
2 1 0 +1 +2

ub u ub u u ub u u u ub u u
Extended constellation (Channel A) Extended constellation (Channel B)

2 1 0 +1 +2 2 1 0 +1 +2
c Marcelino Lázaro, 2013 Digital Communications Linear modulations 32 / 113
Joint transmitter receiver response p(t)
Response p(t) determines the ISI behavior
I Noiseless output depends on the value of p[n], which is
obtained by sampling at symbol rate the joint
transmitter-receiver response p(t)
Usual receiver: matched filter f (t) = g⇤ ( t)

p(t) = g(t) ⇤ g⇤ ( t) ⌘ rg (t)

rg (t): continuous autocorrelation of g(t) (or temporal ambiguity function of g(t))

c Marcelino Lázaro, 2013 Digital Communications Linear modulations 33 / 113

Some properties of the continuous autocorrelation function

Definition for deterministic finite energy function x(t)
rx (t) = x(t) ⇤ x⇤ ( t)
Informally: measure of similarity between a function and itself with a delay t
Expresión in the frequency domain
Rx (j!) = FT {rx (t)} = FT {x(t)} ⇥ F T {x⇤ ( t)}
= X(j!) · X ⇤ (j!) = |X(j!)|2
Maximum value is at t = 0: |rx (0)| |rx (t)|
Energy of the signal
Z 1 Z 1
1
Parseval: E{x(t)} = |x(t)|2 dt = |X(j!)|2 d!
1 2⇡ 1
Using the continuous autocorrelation function (temporal ambiguity func.)
Z 1
1
E{x(t)} = Rx (j!) d! ! E{x(t)} = rx (0)
2⇡ 1

c Marcelino Lázaro, 2013 Digital Communications Linear modulations 34 / 113

Nyquist criterion for zero ISI
Conditions for avoiding ISI expressed in the time domain

p[n] = p(t) = [n]

t=nT

Equivalent condition in the frequency domain

P ej! = 1
Equivalent continuous-time expressions
1
X
p(t) · (t nT) = (t)
n= 1
1 ✓ ◆
2⇡ X 2⇡
P(j!) ⇤ j! j k = 1
T T
k= 1
1 ✓ ◆
1 X 2⇡
P j! j k = 1
T T
k= 1

2⇡
I Replicas of P(j!) displaced multiples of T sum a constant
c Marcelino Lázaro, 2013 Digital Communications Linear modulations 35 / 113

Application: band-limited pulses

1 Rs
Example using a badwidth W < ⇡
T rad/s (or B < 2T = 2 Hz)
I Simplest choice for P(j!): squared pulse
(
⇣ ! ⌘ 1 |!| < W = 2⇡B
P(j!) = ⇧ =
2W 0 |!| > W = 2⇡B
1 ✓ ◆
1 X 2⇡
P j! j k
........................ ........................ ........................
T k= 1 T

... .. ... .. ... ..

... .. ... .. ... ..
... .
. ... .
. ... ..
.. .
. ... .
. ... ..
... .. ... .. ... ..
... .
. ... .
. ... ..
... .. ... .. ... ..
... .. ... .. ... ..
... .
. ... .
. ... ..
....... .
........⇡... .
.....⇡...... .......
2⇡ 2⇡
T T 0 T T
I Impossible to fulfill Nyquist criterion for W < ⇡
T rad/s
c Marcelino Lázaro, 2013 Digital Communications Linear modulations 36 / 113
Application: band-limited pulses (II)
Nyquist ISI criterion for squared pulses: only pulses with
✓ ◆
⇡ Rs
W = n · = n · ⇡ · Rs rad/s B = n · Hz
T 2
Which in the time domain means that
⇣ t⌘
p(t) = sinc n
T
Relationship bandwidth / transmission rate: optimal p(t)
1
I Minimum bandwidth to transmit without ISI at rate Rs = T bauds
✓ ◆
⇡ Rs
Wmin = = ⇡ · Rs rad/s Bmin = Hz
T 2
I Maximun rate without ISI through a bandwidth W rad/s (B Hz)
W
Rs max = = 2 · B bauds (symbols/s)
⇡
I Optimal pulses
⇣t⌘ ✓ ◆
!T
p(t) = sinc , P(j!) = T · ⇧
T 2⇡
c Marcelino Lázaro, 2013 Digital Communications Linear modulations 37 / 113

Raised cosine pulses

Family of bandlimited pulses : parameter ↵ 2 [0, 1]: roll-off factor
I Particular case ↵ = 0 is a sinc function
Expression for pulses in time domain
✓ ◆✓ ◆
sin(⇡t/T) cos(↵⇡t/T)
h↵,T
RC (t) =
⇡t/T 1 (2↵t/T)2
Fourier transform
8
> ⇡
>
> T 0  |!| < (1 ↵)
>
> " !!# T
>
<T
↵,T T ⇡ ⇡ ⇡
HRC (j!) = 1 sin |!| (1 ↵)  |!|  (1 + ↵)
>
> 2 2↵ T T T
>
>
>
> ⇡
:0 |!| > (1 + ↵)
T
Bandwidth for a transmission rate depends on the roll-off factor
⇡ Rs
W = (1 + ↵) · rad/s, B = (1 + ↵) · Hz
T 2
c Marcelino Lázaro, 2013 Digital Communications Linear modulations 38 / 113
Raised cosine pulses: h↵,T
RC (t)

h↵,T
RC (t)
............
......... ............
.... .......
............... .........
.........
.. . ............... ↵ = 1
................. ...............
... ............ ............... ↵ = 0,75
.................. ............
.... ............ ............... ↵ = 0,5
.................... ............
.. ............ ............... ↵ = 0
................. ............
........... ...........
. . . . ...
. ......... . . . ..
. .. . ........ ......................
.... . .
.
..
.. ...
.
. . .
. .
.
............................................................................................................ ...... .................................................................................
.......................................... t
.......... .................. . . . ... .
.. .............. ............... .
3T 2T ................... T 0 T .................. 2T 3T

c Marcelino Lázaro, 2013 Digital Communications Linear modulations 39 / 113

↵,T
Raised cosine pulses: HRC (j!)
↵,T
HRC (j!)
T .....................................................................................................
... ... ... .... .... .... ... ...
... ... ... ... . . . .... .. .. ..
... ... ... ..
... ... ... ... ... ... .. ..
... ... ...... ... ... ... ... ............... ↵ = 1
... .. .... ......... ..
... ...... . . . ......... ... ............... ↵ = 0,75
... ......... ........ ..
........ ... ............... ↵ = 0,5
............ ..........
....... ...... ............... ↵ = 0
...... ...
........... ..........
... . ............
............ .... ... ..........
......... ... ... ..........
.. ..... ... ... ...........
.. ... .. ..
. .. ... .......
.. .. . ...
. ... .. .. ...
.. ..... ... ... ... .... .... .....
. . .
. .... ... ... . . . ... ... .... .....
.... ...............................................................
......................................................
2⇡ ⇡ ⇡ 2⇡
T T 0 T T !
c Marcelino Lázaro, 2013 Digital Communications Linear modulations 40 / 113
Root-raised cosine pulses
Filter whose convolution is a raised cosine
h↵,T ↵,T ↵,T ↵,T ↵,T ↵,T
RRC (t) ⇤ hRRC (t) = hRC (t), HRRC (j!) · HRRC (j!) = HRC (j!)

General procedure to obtain transmission filter hRRC (t)

↵,T
1 Design in frequency domain from HRC (j!)
q
↵,T ↵,T
2 Divide in two contributions: HRRC (j!) = HRC (j!)
n o
h↵,T 1 H ↵,T (j!)
RRC (t) = TF
3
RRC

Root-raised cosine pulses

! !
⇡t 4↵t ⇡t
sin (1 ↵) · cos (1 + ↵)
T T T
h↵,T
RRC (t) =
2 !2 3
⇡t 4↵t
· 41 5
T T

c Marcelino Lázaro, 2013 Digital Communications Linear modulations 41 / 113

Root-raised cosine pulses: h↵,T

RRC (t)

1,5
....
............................
..... .... .......
. ............... ↵ = 1
1 ......... ........... ............... ↵ = 0,75
.. .....
........ ....... ............... ↵ = 0,5
. ...........
.................. ............ ............... ↵ = 0
0,5 ................ ............
............
.
................... .............
.. ....... ............
. . . ......... ...
.
...
..
..
.
..
..
...
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
... ...
. .
. . .. .
.
.. ................................... .. ......
.
.
. .
. . . .. .. .. .. ........
......
...
.
...
....
.
..
..
..
..
.........................................................
.
0 . ... ....................... ............................... ........................ ..... ................... .. ... .
..... .................... ....................... ....
.............. .................

0,5
3T 2T T 0 T 2T 3T
t

c Marcelino Lázaro, 2013 Digital Communications Linear modulations 42 / 113

 ↵,T
Root-raised cosine pulses: HRRC (j!)
↵,T
HRRC (j!)
p
.......................................................................................
T
... ... ... .... .... ... ... ..
... .. .. .. . . . ... ... .. ..
... .. ... ..
... ... ...... ...... ... ...
... .. .... . . . ...... .. .. ............... ↵ = 1
... ...... . . . ......... ...
... ......... ......... .. ............... ↵ = 0,75
. . . ......... ...
... ...... .......... ............... ↵ = 0,5
.......... .........
.... .... ............... ↵ = 0
....... ......
..... ..........
................. ...........
......... ... ... .......
... ... ..........
............. .... ... ...........
... .. ... ... .. .........
... ....... ... ... .. ... ...
... .... ... ....
. ...... ... .... ... ... ..... .....
.. ... ... ... .. ...................................................
.............. ...... ...................
2⇡
T
⇡
T 0 + ⇡T + 2⇡
T
!

c Marcelino Lázaro, 2013 Digital Communications Linear modulations 43 / 113

Raised cosines - side lobe attenuation

.
..........
. ..
................
... ....
.....................
...........
............. ..............
.............................
.
.
. . ... .
...... ............. ............ ...... ... .
......................................................................................... ....... ... ... ... ............................... .... ... .... .......................................................↵ . . . .. . ..= ........0...........................
.. . ... ....................... .. . ..
.. ..................... .................... ...
.
.. ..
. ..
. . .
.. ..
. . . . . .. . . . .. . .. . ..
................................................................................................................................................... ........... ...... .... ..... ..... ... ............. .............. ... ... ... ............................................................................................................↵ ..................=
.
.. .
.. .................0,1
.......................................................
. ... ........ ........... .. ..
.. ...... ...... ..
........................................................................................................................................................................................................................................................... ................... .................. .........................................................................................................................................................↵ .......................0,25
..................= .......................................................
.. .
......... ........ .. . .
.... ....
........................................................................................................................................................................................................................................................................... .........................................................................................................................................................................↵ .......................0,5
..................= .......................................................
. .
...... ...... . .
. .
...................................................................................................................................................................................................................................................................... ....................................................................................................................................................................↵ .......................1.......................................................
..................=
25T 20T 15T 10T 5T 0 +5T +10T +15T +20T +25T

c Marcelino Lázaro, 2013 Digital Communications Linear modulations 44 / 113

Raised
Raised cosines
cosines -- implementation
implementationdelay
delay
AA raised
raised cosine has
has aa number
numberofof“relevant”
“relevant”side
sidelobes
lobesthat
that
is is decreasing
decreasing with
with roll-off
roll-off factor
factor
I
I Non-relevant lobes could
Non-relevant lobes couldbe
betruncated
truncatedtotomake
makeeasier
easier the
the implementation
implementation

For implemententing
For implemententing the
the modulated
modulatedwaveform,
waveform,a adelay
delayis is
necessary
necessary
I Delay is related with the number of relevant side lobes that have to be cosidered before truncation
I Delay is related with the number of relevant side lobes that have to be cosidered before truncation
I Delay is lower for higher values of ↵ (higher bandwidth requirement)
I Delay is lower for higher values of ↵ (higher bandwidth requirement)
Example: generation of a 4-PAM waveform with ↵ = 0
Example: generation
I In the of alobes
example, 25 side 4-PAM waveform
are considered with
relevant (and = 0 25 side lobes are depicted)
↵therefore
I
I In the example,
A delay of 25 ⇥ 25 side lobes
T seconds are considered
is necessary relevant
to compute the (and therefore 25 side lobes are depicted)
addition
I
I A delay
Black of 25is⇥the
signal seconds
T last is necessary
one with to computeatthe
relevant contribution t =addition
0 (related with A[25])
I Black signal is the last one with relevant contribution at t = 0 (related with A[25])

..... ..... ..... .....

........ ........ ........ ........
.. .. .. .. .. .. rrr .. ..
... ..... ... ..... ... ..... ... .....
......... ..... ......... ..... ......... ..... .. ... ....
.
. . .... . ...... . ...... ... .. ..
........................................................................................................................................................................................................................................................................................................................................................................................................................................
.................... .............. .......... ... . .... ....
..... .. ..... .. .... ..
... ... ... ... .... ...
... ... ... ... .. ...
...... ...... .......
..... ..... .....
. . .
20T 10T 0 +10T +20T +30T +40T +50T

Marcelino Lázaro Digital Communications Linear modulations 44 / 78

c Marcelino Lázaro, 2013 Digital Communications Linear modulations 45 / 113

Raised
Raised cosines -- implementation
implementationdelay
delay(II)
(II)
Lower
Lower delays can
can be
be achieved
achievedby byusing
usinghigher
higherroll-off
roll-offfactors
factors
I The price
I price to
to be
be paid
paidisisaahigher
higherrequired
requiredbandwidth
bandwidth
Example: generation
Example: of
of aa4-PAM
4-PAMwaveform
waveformwith with↵↵==0,5
I In thegeneration
example, 4 side lobes are considered relevant
0,5
I In the example,
I A delay of 4 ⇥ T4seconds
side lobes are considered
is necessary relevant
to compute the addition
I A delay of 4 is T seconds
I Black signal ⇥the last one is necessary
with to compute at
relevant contribution the
t= addition
0 (related with A[4])
I
I Black
Delay signal is the last
is decreased fromone with
25 ⇥ relevant
T to 4 ⇥ T contribution
in this exampleat t(more
= 0 (related with lower)
than 6 times A[4])
I
I Delay is decreased
Required bandwidth from
is 50 25 ⇥ T to 4 ⇥ T in this example (more than 6 times lower)
% higher
I Required
NOTE: thebandwidth
number ofis“relevant”
50 % higher
lobes depends on required accuracy, this is just a simple example (numbers
NOTE:
can not the number
be taken as of “relevant”
a precise lobes depends on required accuracy, this is just a simple example (numbers
reference)
can not be taken as a precise reference)

.. ..
......... .........
.. . .. .
.... .... .... ....
.. .. .. ..
......... ..... ... ......
. .. . . .
.................................................................................................................................................................................................
. ........ .. .
.... ...
... ..
... ...
......
.....
..
20T 10T 0 +10T +20T +30T +40T +50T

Marcelino Lázaro Digital Communications Linear modulations 45 / 78

c Marcelino Lázaro, 2013 Digital Communications Linear modulations 46 / 113
Review: spectrum of continuous/discrete time signals
Continuous signal x(t) and discretized x[n] sampled at T seg.

x[n] = x(t) t=nT

= x(nT)

Usual notation
I X(j!): spectrum (Fourier transform) of x(t)
I X ej! : spectrum of x[n]

Relationship between both spectral responses

I To obtain discrete from continuous
1 X ✓! 2⇡
◆
X ej! = · X j j k
T T T
k

I To obtain continuous from discrete

X(j!) = T · X ej!T , |!|  ⇡

c Marcelino Lázaro, 2013 Digital Communications Linear modulations 47 / 113

Review: random processes and linear systems

X(t) - Y(t) -
h(t)

Theorem: X(t) is stationary, mean mX and autocorrelation

function RX (⌧ ). The process is the input of a time-invariant linear
system with impulse response h(t). In this case, input and output
processes, X(t) and Y(t), are jointly stationary, being
Z 1
mY = mX h(t) · dt
1

RY (⌧ ) = RX (⌧ ) ⇤ h(⌧ ) ⇤ h( ⌧ )
RXY (⌧ ) = RX (⌧ ) ⇤ h( ⌧ )
Moreover, it can be seen that
RY (⌧ ) = RXY (⌧ ) ⇤ h(⌧ )
c Marcelino Lázaro, 2013 Digital Communications Linear modulations 48 / 113
Review: expressions in the frequency domain
Mean for output process

mY = mX · H(0)

Power spectral density of the output process

SY (j!) = SX (j!) · |H(j!)|2

Crossed power spectral densities

def
SXY (j!) = TF[RXY (⌧ )]

SXY (j!) = SX (j!)H ⇤ (j!)

⇤
SYX (j!) = SXY (j!) = SX (j!)H(j!)

c Marcelino Lázaro, 2013 Digital Communications Linear modulations 49 / 113

Properties of the noise at the receiver

Noise n(t) is filtered by receiver filter f (t) (producing filtered noise
z(t)) and then sampled (z[n])
Analysis in the frequency domain
I PSD of filtered noise z(t)
N0
Sz (j!) = Sn (j!) · |F ⇤ (j!)|2 = · |F(j!)|2
2
F Coloured noise (non-flat PSD)
I PSD of sampled noise z[n]
✓ ◆2
N0 1 X
j! ! 2⇡
Sz (e ) = · F j j k
2 T T T
k | {z }
2⇡
Rf ( j T j T k )
!

F Sampled noise can be white !!!!

✓ ◆
1X ! 2⇡
Condition: Rf j j k = constant
T T T
k

c Marcelino Lázaro, 2013 Digital Communications Linear modulations 50 / 113

Conditions for sampled noise z[n] being white
Sampled noise z[n] is white if
✓ ◆
1X ! 2⇡k
Rf j j = C is equivalent to Rf (ej! ) = C
T k T T

I Equivalent condition in the time domain

rf [n] = rf (t) t=nT

= C · [n], which implies C = rf (0)

Equivalent statement for z[n] being white

I z[n] is white if the continuous autocorrelation function of
receiver filter fulfills Nyquist condition for zero ISI
REMARK
I Condition for z[n] being white only depends on the shape of
receiver filter f (t) !!!

c Marcelino Lázaro, 2013 Digital Communications Linear modulations 51 / 113

Consequences of Nyquist criterion for Gaussian channels

A matched filter is assumed at the receiver

f (t) = g( t) since g(t) is a real function

Condition to avoid ISI

I Joint response p(t) = g(t) ⇤ f (t) fulfills Nyquist criterion
F Using matched filters p(t) = rg (t)

Condition for z[n] being white

I Continuous autocorrelation of the receiver filter, rf (t), fulfills
Nyquist criterion
F Using matched filters rf (t) = rg (t)

Conclusion: both conditions are equivalent

I Transmitting through a Gaussian channel using matched
filters, if ISI is avoided, sampled noise z[n] is white

c Marcelino Lázaro, 2013 Digital Communications Linear modulations 52 / 113

Signal to noise relationship
If Nyquist ISI criterion is satisfied, the received observation is
q[n] = A[n] + z[n]

In this case, signal to noise ratio is

✓ ◆
S E{|A[n]|2 } Es
= 2
= 2
N q z s

2 is the varianze of noise sequence z[n]

z
Z ⇡
2 1
z = Sz (ej! ) d!
2⇡ ⇡

I If Nyquist ISI criterion is fulfilled

F For a normalized receiver filter z2 = N20
F For a non-normalized receiver filter
2 N0 N0
z = ⇥ E{f (t)} = ⇥ rf (0)
2 2
c Marcelino Lázaro, 2013 Digital Communications Linear modulations 53 / 113

Evaluation of Probability of Symbol Error (Pe )

Definition
Pe = P(Â[n] 6= A[n])

Evaluation - Averaging of probability of symbol error for each

symbol in the constellation
M
X1
Pe = pA (ai ) · Pe|ai
i=0

Calculation of contitional probabilities of symbol error (conditional

probabilities of error)
Z
Pe|ai = fq|A (q|ai ) dq
q2I
/i

Conditional distribution of observations conditioned to transmission of

the symbol ai is integrated out of its decision region Ii

c Marcelino Lázaro, 2013 Digital Communications Linear modulations 54 / 113

Calculation of Bit Error Rate (BER)
Conditional BER for each symbol ai are averaged
M
X1
BER = pA (ai ) · BERai
i=0

Calculation of conditional BER for ai

M
X1 me|ai !aj
BERai = Pe|ai !aj ·
m
j=0
j6=i

I Pe|ai !aj : probability of deciding  = aj when A = ai was

transmitted Z
Pe|ai !aj = fq|A (q0 |ai ) dq0
q0 2Ij
I me|ai !aj : number of bit errors associated to that decision
I m: number of bits per symbol in the constellation
c Marcelino Lázaro, 2013 Digital Communications Linear modulations 55 / 113

Example - 1-D M-ary constellation

Example:
I M = 4, equiprobable symbols pA (ai ) = 14
I Constellation: a0 = 3, a1 = 1, a2 = +1, a3 = +3
I Decision regions: thresholds qu1 = 2, qu2 = 0, qu3 = +2

I0 = ( 1, 2], I1 = ( 2, 0], I2 = (0, +2], I3 = (+2, +1)

I Binary assignment

a0 ⌘ 01, a1 ⌘ 00, a2 ⌘ 10, a3 ⌘ 11

I0 I1 I2 I3
- - - -
t t t t
a0 ⌘ 01 a1 ⌘ 00 a2 ⌘ 10 a3 ⌘ 11
q
3 2 1 0 +1 +2 +3

c Marcelino Lázaro, 2013 Digital Communications Linear modulations 56 / 113

Example - 1-D M-ary constellation (II)
Probability of error
M
!
1 X1 3 1
Pe = Pe|ai = Q p
4 i=0
2 N0 /2

Bit error rate (BER)

! ! !
3 1 1 3 1 5
BER = Q p + Q p Q p
4 No /2 2 No /2 4 No /2

Analytic developments are detailed in Annex

c Marcelino Lázaro, 2013 Digital Communications Linear modulations 57 / 113

Calculation of Pe|a0.
.............
. ...
..... ...
...
..... ...
......
.. ... .......
...
.. ... ... .......
.
. .. ... ... ... .........
... ... ... ... ... ... ...........
u . . . . .. ..u. .................................................u......................................................................u......................................................................................................
.....
.. a0
........
3 2 1 0 +1 +2 +3 q
-
I0

Distribution fq|A (q|a0 )

I Gaussian with mean a0 = 3 and variance N0 /2
Conditional probability of error
I Integration of fq|A (q|a0 ) out of I0
Z !
1
Pe|a0 = fq|A (q|a0 ) dq = Q p
q2I
/0 N0 /2
c Marcelino Lázaro, 2013 Digital Communications Linear modulations 58 / 113
Calculation of Pe|a1
................
.. ...
..... ...
...
.... ......
.
........... ... .....
. .
.. ... ... ... . ... ... .....
. .
... .. .. .. ... ... ... ......
.. .
. ... ... ... ... ... a1 ... ... ... ... .........
...................................................................................... ..u.. .. .. .. .. u
.. .. ... ... ... ...u.........................................................u......................................................................................................
.. .
.
3 2 1 0 +1 +2 +3 q
-
I1

Distribution fq|A (q|a1 )

I Gaussian with mean a1 = 1 and variance N0 /2
Conditional probability of error
I Integration of fq|A (q|a1 ) out of I1
Z !
1
Pe|a1 = fq|A (q|a1 ) dq = 2Q p
q2I
/1 N0 /2
c Marcelino Lázaro, 2013 Digital Communications Linear modulations 59 / 113

Calculation of Pe|a2
..............
. ...
..... ...
...
....... ....
.... ... .......
. .. ... .......
.. ...... .... .... ... ... .......
.. .
. . . . ... ... ... ..........
. ...... ... ... ... ... a ... ... ... ... ... ............
u ..
..
. .
..
.
.. ..
..........
...... ...u... ... ... ... ...
. .
u
2
. . . . . .u. . ...............................................................................
................................................................................................................................ .
3 2 1 0 +1 +2 +3 q
-
I2

Distribution fq|A (q|a2 )

I Gaussian with mean a2 = +1 and variance N0 /2
Probability of error
I Integration of fq|A (q|a2 ) out of I2
Z !
1
Pe|a2 = fq|A (q|a2 ) dq = 2Q p
q2I
/2 N0 /2
c Marcelino Lázaro, 2013 Digital Communications Linear modulations 60 / 113
Calculation of Pe|a3
................
.. ...
..... ...
...
.... ...
.
........... ...
. .
.... ... ... . ...
. .
. ... ...
. .
. .
.
. .
.
. .
.
. ....
.
. . . . . . . ......
............................................................................................................u......................................................................u.................................. . .u u
......
........ ... ... ... ... ... a3
. .............
... .
3 2 1 0 +1 +2 +3 q
-
I3

Distribution fq|A (q|a3 )

I Gaussian with mean a3 = 3 and variance N0 /2
Probability of error
I Integration of fq|A (q|a3 ) out of I3
Z !
1
Pe|a3 = fq|A (q|a3 ) dq = Q p
q2I
/3 N0 /2
c Marcelino Lázaro, 2013 Digital Communications Linear modulations 61 / 113

Calculation of BERa0
..
. .
....................................................
.... ..................
..... ...................................
..... .................................................
.... ..............................................................
..
..............................................................................
..........................................................................................
................................. ....
a0 ...........................................................................................................................................................................................................
t .................................................................................................. ..................................................
3 2 1 0 +1 +2 +3 q
- - -
I1 I2 I3
Binary assignment: a0 ⌘ 01, a1 ⌘ 00, a2 ⌘ 10, a3 ⌘ 11
Distribution fq|A (q|a0 ): Gaussian with mean a0 and variance N0 /2
" ! !#
1 3 1
BERa0 = Q p Q p ⇥
N0 /2 N0 /2 2
|{z}
| {z } me|a !a
Pe|a !a 0 1
0 1 m
" ! !# " !#
3 5 2 5 1
+ Q p Q p ⇥ + Q p ⇥
N0 /2 N0 /2 2 N0 /2 2
|{z} |{z}
| {z } me|a !a | {z } me|a !a
Pe|a !a 0 2 Pe|a !a 0 3
0 2 m 0 3 m

c Marcelino Lázaro, 2013 Digital Communications Linear modulations 62 / 113

Cálculation of BERa1
. .. ..
...............................................
.
. ............................ .......................
. . .
. . ....................................
. . .
.. ...................................... ...................................................
. .
.. . .
. . .
. . .
. .
. .
. .
. .
. .
. .
. .
. . .....
. . .
.
. . .
........................................................... . .................................................................
. . .. .
. .
........................................................................... ..................................................................................
. . .. .
. . .
........................................................................................... ...................................................................................................
. . . .
................................................................................................................................................................. ...............................................................................................................................
........................................... t ................................................................................................................................................................
3 2 1 0 +1 +2 +3 q
- - -
I0 I2 I3
Binary assignment: a0 ⌘ 01, a1 ⌘ 00, a2 ⌘ 10, a3 ⌘ 11
Distribution fq|A (q|a1 ): Gaussian with mean a1 and variance N0 /2
" !# " ! !#
1 1 1 3 1
BERa1 = Q p ⇥ + Q p Q p ⇥
N0 /2 2 N0 /2 N0 /2 2
|{z} |{z}
| {z } me|a !a | {z } me|a !a
Pe|a !a 1 0 Pe|a !a 1 2
1 0 m 1 2 m
" !#
3 2
+ Q p ⇥
N0 /2 2
|{z}
| {z } me|a !a
Pe|a !a 1 3
1 3 m

c Marcelino Lázaro, 2013 Digital Communications Linear modulations 63 / 113

Calculation of BERa2
. . . .
.. ..................................................
.
. . . .
. ............................
.
.
....................
....................................
. . .
. . .
. . . .
. . .
. . .
. . .
.
. . . . . . . . . .
.............................................. . ...................................................
.. .
. . .
............................................................. ................................................................
.. . . . . .
............................................................................ .................................................................................
. . .. .
. . . .
.............................................................................................. ....................................................................................................
. ....................................................................................................................... a2 ............................................................................................................................................
. . . .
t
..... ....
...
... .
.............................................................................................................................................. ..................................................................................
3 2 1 0 +1 +2 +3 q
- - -
I0 I1 I3
Binary assignment: a0 ⌘ 01, a1 ⌘ 00, a2 ⌘ 10, a3 ⌘ 11
Distribution fq|A (q|a2 ): Gaussian with mean a2 and variance N0 /2
" !# " ! !#
3 2 1 3 1
BERa2 = Q p ⇥ + Q p Q p ⇥
N0 /2 2 N0 /2 N0 /2 2
|{z} |{z}
| {z } me|a !a | {z } me|a !a
Pe|a !a 2 0 Pe|a !a 2 1
2 0 m 2 1 m
" !#
1 1
+ Q p ⇥
N0 /2 2
|{z}
| {z } me|a !a
Pe|a !a 2 3
2 3 m

c Marcelino Lázaro, 2013 Digital Communications Linear modulations 64 / 113

Calculation of BERa3
. .. . .
...............................................
.
............................ ......
. . .
.. . .... ......
.. . .
. .
. .
. .
. .
.
.. ................................ .
. .
. .
. .
. .
. .....
. . .
.. . .
. . . . . . . . . . . .....
.. . . . . . . . . . . . .
................................................................. . . . . . ....
.. . .
. .
.
.. . .
.. .. ................................................................................................................
.
. . ........................................................................................................................................................
t
. ..... . .
...
. . .
...........................................................................................................................................................................................................................
3 2 1 0 +1 +2 +3
- - -
I0 I1 I2
Binary assignment: a0 ⌘ 01, a1 ⌘ 00, a2 ⌘ 10, a3 ⌘ 11
Distribution fq|A (q|a3 ): Gaussian with mean a3 and variance N0 /2
" !# " ! !#
5 1 3 5 2
BERa3 = Q p ⇥ + Q p Q p ⇥
N0 /2 2 N0 /2 N0 /2 2
|{z} |{z}
| {z } me|a !a | {z } me|a !a
Pe|a !a 3 0 Pe|a !a 3 1
3 0 m 3 1 m
" ! !#
1 3 1
+ Q p Q p ⇥
N0 /2 N0 /2 2
|{z}
| {z } me|a !a
Pe|a !a 3 2
3 2 m

c Marcelino Lázaro, 2013 Digital Communications Linear modulations 65 / 113

Modification of the binary assignment

Final result for previous binary assignment

01 00 10 11
s s s s
a0 a1 a2 a3
q
-3 -2 -1 0 1 2 3
! ! !
3 1 1 3 1 5
BER = Q p + Q p Q p
4 No /2 2 No /2 4 No /2

If binary assignment is modified

11 00 10 01
s s s s
a0 a1 a2 a3
q
-3 -2 -1 0 1 2 3
I Terms Pe|ai !aj do not vary
I Terms me|ai !aj do vary ) BER is modified !!!
! !
5 1 1 3
BER = Q p Q p
4 No /2 4 No /2
c Marcelino Lázaro, 2013 Digital Communications Linear modulations 66 / 113
Gray Coding
Blocks of m bits assigned to symbols at minimum distance
differ in only a single bit
01 00 10 11
s s s s
a0 a1 a2 a3
q
-3 -2 -1 0 1 2 3
I This assignment minimizes BER for a given constellation
Terms Pe|ai !aj depend on the constellation
I Values depend on distance between ai and aj
I Highest values for symbols at minimum distance
me|ai !aj
Terms m
depend on bit assignment
I These terms weight the contribution of Pe|ai !aj
F Gray coding: minimizes impact of highest values of Pe|ai !aj
F For high values of signal to noise ratio (SNR), in most cases,
a symbol error produces a single erroneous bit
1
BER ⇡ · Pe
m
c Marcelino Lázaro, 2013 Digital Communications Linear modulations 67 / 113

Transmission of PAM through linear channels

- i
A[n] s(t) r(t) q[n]
- g(t) - h(t) - f (t)
6 ?
t = nT
n(t)
Linear Channel

Linear channel model

I PAM signal s(t) suffers a linear distortion during transmission
I Gaussian noise is also added
r(t) = s(t) ⇤ h(t) + n(t)
h(t): linear system impulse response modeling linear distortion
n(t): stationary random process, white, Gaussian, zero mean, Sn (j!) = N0 /2

Receiver filter f (t)

I Typical set up: matched filter f (t) = g⇤ ( t) = g( t)

Signal at the input of the sampler

q(t) = r(t) ⇤ f (t) = s(t) ⇤ h(t) ⇤ f (t) + n(t) ⇤ f (t)
c Marcelino Lázaro, 2013 Digital Communications Linear modulations 68 / 113

A[n]- - hq[n]
-
Equivalent discrete channel for linear channels
Signal before sampling
!
X
q(t) = A[k] · g(t kT) ⇤ h(t) ⇤ f (t) + n(t) ⇤ f (t)
k
X ✓ ◆
= A[k] · g(t kT) ⇤ h(t) ⇤ f (t) + n(t) ⇤ f (t)
k
X
= A[k] · p(t kT) + z(t)
k

p(t) = g(t) ⇤ h(t) ⇤ f (t): joint transmitter-channel-receiver response

I For a matched filter at the receiver
p(t) = g(t) ⇤ h(t) ⇤ g⇤ ( t) = rg (t) ⇤ h(t)
rg (t): continuous autocorrelation of g(t) (or temporal ambiguity function of g(t))

Observation at demodulator output

X
q[n] = q(t)|t=nT = A[k] · p ((n k)T) + z(nT)
k
c Marcelino Lázaro, 2013 Digital Communications Linear modulations 69 / 113

Equivalent discrete channel for linear channels (II)

Definition of equivalent discrete channel p[n]
p[n] = p(t) t=nT
X
q[n] = A[k] · p[n k] + z[n] = A[n] ⇤ p[n] + z[n]
k
A[n]-
p[n] - hq[n]
-
6
z[n]

Same basic model as for Gaussian channels but with a new

definition por joint response p(t)
I Now definition includes the effect of h(t)
p(t) = g(t) ⇤ h(t) ⇤ f (t), P(j!) = G(j!) · H(j!) · F(j!)
I Using matched filters: f (t) = g( t), F(j!) = G⇤ (j!)
p(t) = rg (t) ⇤ h(t), P(j!) = |G(j!)|2 · H(j!)
c Marcelino Lázaro, 2013 Digital Communications Linear modulations 70 / 113
Avoidance of ISI
Nyquist ISI criterion must be fulfilled for p[n] (or P(j!))
I Definition of p(t) includes now the effect of linear channel h(t)
Design of p(t)|P(j!) to fulfill Nyquist at symbol period T
Design usign matched filters in the receiver
Response of transmitter filter in the frequency domain
I P(j!) = H(j!) · |G(j!)|2
I Therefore (q
P(j!)
H(j!)
, if H(j!) 6= 0
G(j!) =
0, in other case
If the receiver filter is matched to the transmitter filter, this
choice for the transmitter filter eliminates ISI
I P(j!) is a design option
F Tipically, a raised-cosine response is selected
↵,T
P(j!) = HRC (j!)

c Marcelino Lázaro, 2013 Digital Communications Linear modulations 71 / 113

Avoidance of ISI - Problems

Channel response, H(j!), must be known
I It can be difficult to know it
I Channel can be time variant
Discrete noise sequence, z[n], is not white

j! N0 X P j !T j 2⇡
T
k
Sz e =
2T k H j !T j 2⇡
T
k

I Memoryless symbol by symbol detector is not optimal

I All sequence q[n] has to be used to estimate the symbol at a
given discrete instant n0 , A[n0 ]
I Noise can be amplified
F Channels with deep attenuation at some frequencies in the
band

c Marcelino Lázaro, 2013 Digital Communications Linear modulations 72 / 113

Using a generic receiver filter
Generic receiver, not necesarily a matched filter

r(t) q(t) q[n]

- f (t) -
?
t = nT
Definition of joint response p(t)
p(t) = g(t) ⇤ h(t) ⇤ f (t), P(j!) = G(j!) · H(j!) · F(j!)

Equivalent discrete channel at symbol rate p[n]

p[n] = p(nT) = (g(t) ⇤ h(t) ⇤ f (t)) t=nT

Filtered noise
z(t) = n(t) ⇤ f (t), z[n] = z(nT)

I Power spectral density for discrete noise z[n]

✓ ◆ 2
j! N0 1X ! 2⇡
Sz e = ⇥ F j j k
2 T k T T
c Marcelino Lázaro, 2013 Digital Communications Linear modulations 73 / 113

Criteria to design f (t)

Filter matched to the joint transmitter-channel response
f (t) = gh ( t), with gh (t) = g(t) ⇤ h(t)
I Maximizes the signal to noise ratio
I Does not provides zero ISI and noise z[n] is not white
Minimum mean squared error criterion: to maximize
n o
E (A[n]p[0])2
8 !2 9
< P =
E k A[k]p[n k] + z[n]
: k6=n ;

Simultaneously avoidance of ISI and white noise

I Selection of P(j!) fulfilling Nyquist
I Selection of F(j!) with Rf (j!) = |F(j!)|2 fulfilling Nyquist
P(j!)
G(j!) =
H(j!) · F(j!)
F Usually presents serious implementation problems
c Marcelino Lázaro, 2013 Digital Communications Linear modulations 74 / 113
Typical set up for linear channels
Receiver uses a matched filter f (t) = g( t) with rf (t) = rg (t)
fulfilling Nyquist
I This ensures discrete filtered noise z[n] is white
Joint response p(t) then does not fulfills Nyquist
I ISI is present in the system
F Receivers can be specifically designed to deal with ISI (as it
will be seen in chapter 4)

c Marcelino Lázaro, 2013 Digital Communications Linear modulations 75 / 113

Eye diagram
Visualization tool for a digital communication system
I Superposition of waveform pieces around a sampling point
I Duration of each piece: 2T
Main features
I In the middle and in both sides (horizontaly), there are
sampling instants
F Traces should have to go through values of the constellation
I Diversity of transition between sampling instants depend on
the shape of transmitter and receiver filters
It allows to detect several problems:
I Problems/sensitivity to synchronism
I Level of noise
I Presence (and level) of ISI

c Marcelino Lázaro, 2013 Digital Communications Linear modulations 76 / 113

Eye diagram - Examples

2-PAM ↵ = 0 4-PAM ↵ = 0

2-PAM ↵ = 0,25 Noisy 2-PAM ↵ = 0, ISI

c Marcelino Lázaro, 2013 Digital Communications Linear modulations 77 / 113

Eye diagram - Examples (II)

2-PAM ↵ = 0,25 2-PAM ↵ = 0,5

2-PAM ↵ = 0,75 2-PAM ↵ = 1

c Marcelino Lázaro, 2013 Digital Communications Linear modulations 78 / 113
Variability
Variability of of signal signal transitions transitionsrelated relatedwith with↵↵value value
Pictures
Pictures illustrateillustrate the the different differentvariability variabilityofofsignals signalsusing using
= 00 (above)
↵↵ = (above) and and ↵↵ = = 0,5 0,5 (below) (below)
... ... ... ...
....... ....... ....... .......
. . . . . . q q q .. ...
..... ..... ..... ..... ..... ..... ... ...
. . . . . . . . . . . . . . . .
..................................................................................................................................................................................................................................................................................................
.. ......... ......... ...... .. . .
... .. ... .. ... ..
...... ...... ......
...... ...... ......
.. .. ..
20T 10T 0 +10T +20T +30T +40T +50T

... ... ... ...

....... ....... ....... .......
... ... ... q q q ......
..... ..... ..... ..... ..... ..... ... .....
....... ....... ....... .. .
...............................................................................................................................................................................................................
..... .. ..... .. ..... ..
... .. ... .. ... ..
...... ...... ......
...... ...... ......
... ... ...
20T 10T 0 +10T +20T +30T +40T +50T

Marcelino Lázaro Digital Communications Linear modulations 78 / 78

c Marcelino Lázaro, 2013 Digital Communications Linear modulations 79 / 113

Band pass PAM - Generation by AM modulación

A baseband PAM is initially generated
X
s(t) = A[n] · g(t nT)
n

Then, this baseband PAM signal is modulated with an

amplitude modulation. Several options are available
I Conventional AM (double sided band with carrier)
I Double sided band PAM (DSB-PAM)
I Single sided band PAM (SSB-PAM)
F Lower sided band
F Upper sided band
I Vestigial sided band PAM (VSB-PAM)
F Lower sided band
F Upper sided band

c Marcelino Lázaro, 2013 Digital Communications Linear modulations 80 / 113

Drawbacks of using a AM modulation
Conventional AM and double sided band PAM (DSB-PAM)
I Spectral efficiency is reduced to the half (bandwidth is
doubled)
Single sided band PAM (SSB-PAM)
I Ideal analog side band filters are required
F Real filters introduce a distortion
Vestigial sided band PAM (VSB-PAM)
I Analog vestigial band filters are required
F Strong constraints

c Marcelino Lázaro, 2013 Digital Communications Linear modulations 81 / 113

Modulation by using quadrature carriers

Two sequences of symbols (not necessarily independent) are
simultaneously transmitted (rate Rs = T1 in both cases)
AI [n] and AQ [n]
Two baseband PAM signals are generated using g(t)
X X
sI (t) = AI [n] · g(t nT) sQ (t) = AQ [n] · g(t nT)
n n

sI (t): in-phase component, sQ (t): quadrature component

Generation of the band pass signal, x(t), from sI (t) and sQ (t)
p
2 cos(!c t)
AI [n] sI (t) ✏?
+✏
- g(t) -@
@ ? x(t)
sQ (t) ✏
-
AQ [n]
- g(t) -@ 6 p
@ x(t) = 2 · sI (t) · cos(!c t)
p
p 6 2 · sQ (t) · sin(!c t)
2 sin(!c t)
c Marcelino Lázaro, 2013 Digital Communications Linear modulations 82 / 113
Complex notation for band pass PAM
Complex sequence of symbols

A[n] = AI [n] + jAQ [n]

I AI [n] = Re{A[n]}, AQ [n] = Im{A[n]}

Complex baseband signal, s(t):
X
s(t) = sI (t) + jsQ (t) = A[n] · g(t nT)
n

The band pass PAM signal can be written as follows

( )
p p X
x(t) = 2 · Re s(t) · ej!c t = 2 · Re A[n] · g(t nT) · ej!c t
n

c Marcelino Lázaro, 2013 Digital Communications Linear modulations 83 / 113

Bandpass PAM modulator

B[`] A[n] s(t) ✏ x(t)

- - g(t) -@ -
ENCODER
@
6
p j! t
2e c

p
2 cos(!c t)
Re{A[n]} sI (t) ✏?
- g(t) -@
@
✏
AI [n] +
B[`] ? x(t)
- ENCODER -

Im{A[n]} sQ (t) ✏ 6
- g(t) -@
@
AQ [n]
p 6
2 sin(!c t)

c Marcelino Lázaro, 2013 Digital Communications Linear modulations 84 / 113

Relationship with a 2D signal space
Signal in a 2D signal space can be written as
X X
x(t) = A0 [n] · 0 (t nT) + A1 [n] · 1 (t nT)
n n

I
0 (t) and 1 (t) are orthonormal signals
In this case, this only happens if
2⇡
!c = ⇥ k, with k 2 Z
T
In this case
A0 [n] = AI [n], AQ [n] = A1 [n]
0 (t) = g(t) · cos(!c t), 1 (t) = g(t) · sin(!c t)
0 (t nT) = g(t nT) · cos(!c (t nT)) = g(t nT) · cos(!c t)
1 (t nT) = g(t nT) · sin(!c (t nT)) = g(t nT) · sin(!c t)

c Marcelino Lázaro, 2013 Digital Communications Linear modulations 85 / 113

Modulator 2D signal space

Re{A[n]}
- 0 (t)

◆⇣
+
B[`] ? x(t)
-
✓⌘
ENCODER -

6
Im{A[n]}
- 1 (t)

c Marcelino Lázaro, 2013 Digital Communications Linear modulations 86 / 113

Bandpass PAM constellations
2D plotting of possible combinations for AI [n] and AQ [n]
Typical constellations
I QAM (Quadrature Amplitude Modulation) constellations
F M = 2m symbols, with m even
F Symbols arranged in a full squared lattice (2m/2 ⇥ 2m/2 levels)
- Both AI [n] and AQ [n] use baseband PAM constellations
- Independent symbol mapping, bit assignment, and
definition or decision regions are possible
I Crossed QAM constellations
F M = 2m symbols, with m odd
F Symbols arranged in a non-full squared lattice
- Independent symbol mapping, bit assignment, and
definition of decision regions are not possible
I PSK (Phase Shift Keying) constellations
F Symbols are drawn as points in a circle
- Constant energy for all symbols

c Marcelino Lázaro, 2013 Digital Communications Linear modulations 87 / 113

QAM constellations

r r r r r r r r
I m{A[n]} I m{A[n]} I m{A[n]}

r r r r
r r r r r r r r
r r
r r r r r r r r
r r r r
r r r r r r r r
r r r r r r r r
r r r r
r r r r r r r r
r r
r r r r r r r r
Re{A[n]} Re{A[n]} R { [ ]}
e A n

r r r r
r r r r r r r r

Constelaciones 4-QAM (QPSK), 16-QAM y 64-QAM

c Marcelino Lázaro, 2013 Digital Communications Linear modulations 88 / 113

Gray coding for QAM

00t01 01t01 11t01 10t01

01

00t00 01t00 11t00 10t00

00

00t10 01t10 11t10 10t10

10

00t11 01t11 11t11 10t11

11

00 01 11 10

c Marcelino Lázaro, 2013 Digital Communications Linear modulations 89 / 113

Crossed QAM constellations

q q q q q q q q
I m{A[n]} I m{A[n]} I m{A[n]}

r r r r r q q q q q q q q
r r r r r r q q q q q q q q q q q q
q q q q q q q q q q q q
r r r r r r r r q q q q q q q q q q q q
r r q q q q q q q q q q q q
r r r r r r q q q q q q q q q q q q
r r q q q q q q q q q q q q
r r r r r r q q q q q q q q q q q q
q q q q q q q q q q q q
Re{A[n]} Re{A[n]} Re{A[n]}

r r r r r q q q q q q q q
q q q q q q q q

Constellations: 8-QAM, 32-QAM y 128-QAM

c Marcelino Lázaro, 2013 Digital Communications Linear modulations 90 / 113

Phase shift keying (PSK) modulation
PSK constellation
p
A[n] = Es · ej'[n]

I Constant modulus
I Information is conveyed in the symbol phase
Waveform for PSK modulations
( )
p X
x(t) = 2Es Re g(t nT) · ej(!c t+'[n])
n
p X
= 2Es g(t nT) cos(!c t + '[n])
n

I Phase shifts in transitions from symbol to symbol

c Marcelino Lázaro, 2013 Digital Communications Linear modulations 91 / 113

PSK constellations

I m{A[n]} I m{A[n]} I m{A[n]}

r r r r r
r r r r
r r
r r r r r r
r r
r r r r
Re{A[n]} Re{A[n]} Re{A[n]}

r r r r r

Constellations: 4-PSK (QPSK), 8-PSK y 16-PSK

c Marcelino Lázaro, 2013 Digital Communications Linear modulations 92 / 113

Gray coding for PSK

Im{A[n]}

u 000

100 u u 001

u u
101 011

Re{A[n]}

111 u u 010

110 u

c Marcelino Lázaro, 2013 Digital Communications Linear modulations 93 / 113

Other constellations

s s
I m{A[n]} I m{A[n]}

s
s s s s s s
s s s s s s s
s s s s s s s s
s Re{A[n]}
s s s s s sRe{A[n]}
s s s s s s
s
s s

Constellations 1-7-AM-PM y 32-hexagonal

c Marcelino Lázaro, 2013 Digital Communications Linear modulations 94 / 113

Spectrum of a band pass PAM
Condition for cyclostationarity of signal x(t):

E {A[k + m] · A[k]} = 0, for all k, m, m 6= 0

I Conditions for QAM constellations

F Symbol sequences AI [n] and AQ [n] are mutually independent
F Autocorrelation functions of AI [n] and AQ [n] are identical
I Conditions for PSK constellations
F Samples of '[n] are independent

Under cyclostationarity the power spectral density function is

1
Sx (j!) = [Ss (j! j!c ) + Ss⇤ ( j! j!c )]
2
1
Ss (j!) = · SA ej!T · |G(j!)|2
T

c Marcelino Lázaro, 2013 Digital Communications Linear modulations 95 / 113

Spectrum of a band pass PAM (II)

For white sequences of symbols: SA ej! = Es
Es
Ss (j!) = · |G(j!)|2
T
The shaping pulse is responsible of the shape of the spectrum
1 Es h 2 2
i
Sx (j!) = |G(j! j!c )| + |G(j! + j!c )|
2T

I Example using pulses of raised cosine family

W-eq
...................
..... ... W-
... ... .................. ..................
... ...
... .... ... .... ...
. ... .. ... ..
0 ! !c 0 +!c !
Bandpass bandwidth W is double of equivalent baseband bandwidth Weq
Spectral efficiency is the same because now two sequences are transmitted

c Marcelino Lázaro, 2013 Digital Communications Linear modulations 96 / 113

Transmitted power
The mean transmitted power is
Z 1
1
PX = Sx (j!) d!
2⇡ 1

If symbol sequence A[n] is white

SA ej! = Es

I Power for a white symbol sequence

Z 1
Es 1 Es
PX = · |G(j!)|2 d! = · E{g(t)}
T 2⇡ 1 T

F For normalized pulses (with unitary energy)

Es
PX = = Es ⇥ Rs Watts
T
c Marcelino Lázaro, 2013 Digital Communications Linear modulations 97 / 113

Demodulator for band pass PAM

Demodulation and a baseband filter structure can be used
I Complex notation and implementation by components can
be seen in the following pictures

y(t) ✏ q[n] Â[n]

-@ - f (t) - -
@ DETECTOR

6 ?
p t = nT
2e j!c t
✏ Re{q[n]}
- @ - f (t) -
@
y(t) p 6 Â[n]
- 2 cos(!c t) DETECTOR -

✏ Im{q[n]}
- @ - f (t) -
@
?
p 6 t = nT
2 sin(!c t)
c Marcelino Lázaro, 2013 Digital Communications Linear modulations 98 / 113
Equivalent alternative demodulator
Signal at the input of the sampler (using complex notation)
⇣p ⌘
j!c t
q(t) = y(t) · e ⇤ 2 · f (t)

Expression for the convolution

p Z 1
q(t) = 2 f (⌧ ) · y(t ⌧ ) · ej!c ⌧ · e j!c t
d⌧
1

Rearranging terms, an equivalent demodulation scheme is

obtained
Z 1p
q(t) = e j!c t · 2 · f (⌧ ) · ej!c ⌧ · y(t ⌧ ) d⌧
1
⇣ ⇣p ⌘⌘
j!c t j!c t
q(t) = e · y(t) ⇤ 2 · f (t) · e
Bandpass filtering and then demodulation

c Marcelino Lázaro, 2013 Digital Communications Linear modulations 99 / 113

Equivalent alternative demodulator (II)

y(t) p ✏ q[n] Â[n]

- 2f (t)ej!c t -@ - -
@ DETECTOR

6 ?
t = nT
e j!c t
sin(!c t)
✏? ✏ Im{q(t)}
-@ - -
np o @
✏
- Re 2f (t)ej!c t 6
-@
@
y(t) cos(!c t) 6
✏
-
?
-@
np o @
✏ ✏
- Im 2f (t)ej!c t
? Re{q(t)}
-@ - -
@
6
sin(!c t)
c Marcelino Lázaro, 2013 Digital Communications Linear modulations 100 / 113
Noise characteristics at the receiver

n(t) p `(t) ◆⇣
z(t) z[n]
- -@
✓⌘
2f (t)ej!c t -
@
6 ?
t = nT
e j!c t
Some definitions:
p p
fc (t) = 2 · f (t) · ej!c t , Fc (j!) = 2 · F(j! j!c )

Properties:
1 z(t) is strict sense stationary only if `(t) es circularly
symmetric
NOTE: A complex process X(t) is circularly symmetric if real and imaginary parts, Xr (t) and Xi (t), are jointly
stationary, and their correlations satisfy

RXr (⌧ ) = RXi (⌧ ), RXr ,Xi (⌧ ) = RXi ,Xr (⌧ )

2 `(t) is circularly symmetric if !c is higher than bandwidth of

filter fc (t) (narrow band system)
S` (j!) = 2 · Sn (j!) · |F(j! j!c )|2
c Marcelino Lázaro, 2013 Digital Communications Linear modulations 101 / 113

Noise signal z(t) at the receiver

z(t) is circularly symmetric and its power spectral density is

Sz (j!) = 2 · Sn (j! + j!c ) · |F(j!)|2

I In the process is symmetric, its real and imaginary parts,

zI (t) and zQ (t), have the same variance and are independer
for any time instant t
I In general, zI (t1 ) and zQ (t2 ), for t1 6= t2 are not independent
I If spectrum is hermitic, Sz (j!) = Sz⇤ ( j!), zI (t1 ) and zQ (t2 ), for
t1 6= t2 are also independent
F If n(t) is white, this is fulfilled when f (t) is real

c Marcelino Lázaro, 2013 Digital Communications Linear modulations 102 / 113

Discrete noise sequence z[n] at the receiver
z[n] is circularly symmetric
✓ ◆ ✓ ◆ 2
j! 2 X ! !c 2⇡k ! 2⇡k
Sz e = · Sn j + j j · F j j
T T T T T T
k

For white noise n(t)

N0
Sn (j!) =
2
Now
I zI [n] and zQ [n] are independent for any instant n
I zI [n1 ] and zQ [n2 ], for n1 6= n2 , are only independent if Sz ej!
is a symmetric function
F This happens for white noise if the ambiguity function of f (t),
rf (t) = f (t) ⇤ f ⇤ ( t), satisfies the Nyquist ISI criterion at
symbol period T

c Marcelino Lázaro, 2013 Digital Communications Linear modulations 103 / 113

Variance and distribution of z[n]

The variance of complex discrete noise is
Z ⇡
2 1
z = Sz ej! d!
2⇡ ⇡

In noise n(t) is white, with Sn (j!) = N0 /2 W/Hz, and if rf (t) is

normalized and satisfies the Nyquist ISI criterion
2
z = N0

If noise is circularly symmetric

I Real and imaginary parts (zI [n] and zQ [n]) are independent and
both have variance N0 /2
I Probability density function of noise level is
1 |z|2
fZ (z) = e N0
⇡N0
NOTE: If receiver filter is not normalized, noise variance is multiplied by E{f (t)}

c Marcelino Lázaro, 2013 Digital Communications Linear modulations 104 / 113

Baseband equivalent discrete channel
Definition of the complex equivalent baseband channel, heq (t)
j!c t
heq (t) = e · h(t) $ Heq (j!) = H(j! + j!c )
The behavior of the channel around central frequency !c is shifted down to baseband

Signal at the output of the matched filter

X
q(t) = A[n] · p(t nT) + z(t)
n

I p(t) = g(t) ⇤ heq (t) ⇤ f (t), P(j!) = G(j!) · Heq (j!) · F(j!)
Baseband equivalent discrete channel:
p[n] = p(t) t=nT
= p(nT)
✓ ◆
j! 1X ! 2⇡k
P e = P j j
T T T
k
✓ ◆ ✓ ◆ ✓ ◆
1X ! 2⇡k ! 2⇡k ! 2⇡k
= G j j · Heq j j ·F j j
T T T T T T T
k
c Marcelino Lázaro, 2013 Digital Communications Linear modulations 105 / 113

Equivalent discrete channels - baseband and band pass PAM

- j
A[n] - q[n]-
p[n]
6
z[n]
Identification of baseband and band pass PAM
I Symbols A[n]
I Equivalent discrete channel p[n]
I Discrete noise z[n]
F Are real in baseband PAM
F Are complex in band pass PAM

c Marcelino Lázaro, 2013 Digital Communications Linear modulations 106 / 113

Scattering diagram
Monitoring tool for band pass system
Plotting of Re(q[n]) versus Im(q[n])
I

I Ideally: the transmitted constellation must be plotted

I Allows to monitor noise level, ISI level, synchronism errors

2 2
qqqqqq q qqqqqq q
qqqqqqqqqqqqqq q q q qqq
q q qqq q qq q q q q qqqqqqqqqqqqq qqqqqqqqqqq
q
q q
qqqqqqqqqqqqqqqqqqqqqqqqqqqqqqq q
qqqqqqqqqq qqqqqqqqq q q
qqq qqqq qqqqqqqqqqq q
1.5 1.5
qqqqqqqqqqqqqqqqqqqqqqqq q q q qq q q qqqq qqq qqqqqqq q
q q
qqqqqqqqqq q q qqq qq q q q q qq q q qq q
q q qqq
q q q q q
qqqqqqqqqqqqq qqqqqqqqqqqqqqq q
q
q q qqqqqqqqqqqqq qqqqqqq
1 1
q q q qq q q q qq
Im{q([n])}

Im{q([n])}
0.5 0.5
0 0
q q q qqqqqqqqqqqqq q qqqqqqqqqqqqqqqq qqqqqqqqqqqqqqq qqqqqqqqqqqqqqq
qqqqqqqqqqqqqqqqqqqqqqq qq q q qq q q
qqqqqqqqqqqqqqqqqqqqqqqq qq q qq q q q q qqqq
qq
q q
qq qqq q qq qqqqqqqq qqqqqqqqqqq
-0.5 -0.5
qqqqqqqqqqqqqqqqqq q qqq qqq qq q qq
qqq qq q q q q qqq
qq qqq q q q q qq q q q q q q
qqqqqqqqqqqqqq qqqqqqqqqqqqqqqqq q qqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqq q
q qq q q q qq q q q q
-1 -1
-1.5 q -1.5
-2 -2
-2 -1.5 -1 -0.5 0 0.5 1 1.5 2 -2 -1.5 -1 -0.5 0 0.5 1 1.5 2
Re{q([n])} Re{q([n])}
p[n] = [n] p[n] = [n] 0,25 [n 1]
c Marcelino Lázaro, 2013 Digital Communications Linear modulations 107 / 113

Reminder - AM modulations
Product by a sinusoid of frequency !c shifts spectrum !c

...........A...
.... ....
... ..
.
...... .....
.... ..... ...... ..........................
!

...... .............. ....

. ..... ..
...
..... ..... ..... ... . ... ....
..... ...............................
.
.
........
.... .....
... .... ..... ..... ..... . .... .... .... ....
...
...

... ..... ..... ..... ..... ...... ..... ...... ...... ..... ..... ...... .... .... .. .....
... ... ...
. ... ........................ ... ..

@n
.
... .... . ... ..
m(t) - x(t)- .......... .......................... .....
...... ...... ...... ........... ....... ..... ..... ...... ........... ...........................................................
..........

.....
..... ...... ...... ...... ................................... ....................... ...... ............. ...

6 .....
..
.. .. .
. .
. .
.
. .
. . ..
... ..
.....
.
.... .. ..
...
.
.
... ....................... .....
. .
. ....
.

c(t) = cos(!c t) .
... ....
.
.........

... .... ... .... .... ... .... .... ... .... .... ... .... .... ... ..
.................... ....................
A
... .... ..... ..... .... ..... ..... .... ..... ..... .... ..... ..... .... ..... .. 2
.................. .................. .................. .................. .................. .... ... .... ...
..... ...... ...... ..... ...... ...... ..... ...... ...... ..... ...... ...... ..... ...... ......
. . . . . !c +!c !

6 6
!c +!c !

c Marcelino Lázaro, 2013 Digital Communications Linear modulations 108 / 113

Analytic analysis of modulation / demodulation
Block diagram for transmitter and receiver
p p
2 cos(!c t) 2 cos(!c t)

AI [n] sI (t) ✏? ✏ ? qI [n]

@ ✏
- g(t) -@ - @ - f (t) -
+ x(t)
? y(t)
@

sQ (t) ✏ ✏
- -
AQ [n] qQ [n]
- g(t) -@ 6 - @ - f (t) -
@ @
?
6 6 t = nT
p p
2 sin(!c t) 2 sin(!c t)

Transmitter multiplies two baseband signals by two orthogonal

carriers
Receiver demodulates each component and then filters with f (t)
I Receiver filter f (t) has a baseband characteristic
I Typical set-up: root-raised cosine filter

c Marcelino Lázaro, 2013 Digital Communications Linear modulations 109 / 113

Analytic analysis of modulation / demodulation (II)

Undistorted received signal (modulated signal) has the shape
y(t) = A · cos(!c t) + B · sin(!c t)
At the receiver, signal processing is splitted in two components
yA (t) = [A · cos(!c t) + B · sin(!c t)] ⇥ cos(!c t)
yB (t) = [A · cos(!c t) + B · sin(!c t)] ⇥ sin(!c t)
Trigonometric identities and removing (filtering) of bandpass terms
X X
X · cos(!c t) · cos(!c t) = + · cos(2!c t)
2 2
|{z} | {z } X
Desired Bandpass at 2!c X · sin(!c t) · cos(!c t) = · sin(2!c t)
2
X X | {z }
X · sin(!c t) · sin(!c t) = · cos(2!c t) Bandpass at 2!c
2 2
|{z} | {z }
Desired Bandpass at 2!c

..............
..... .....
................ . ... ................
..... ...
. .... ...
. ..... ...
.
.
2!c !c 0 +!c +2!c !
c Marcelino Lázaro, 2013 Digital Communications Linear modulations 110 / 113
Analytic analysis of modulation / demodulation (III)
The product of two carriers allows to recover the transmitted baseband
signals
I Products cos(!c t) ⇥ cos(!c t) or sin(!c t) ⇥ sin(!c t) introduce a 12 factor
p
F Factors 2 are introduced at transmiter and receiver to compensate it
I Complex notation fails to represent this scaling
F This has to be taken into account

Modulator Demodulator Joint effect (for p[n])

- ........j - ........j - ........j

s(t) x(t) y(t) r(t) s(t) r(t)
. . ..
.... - . . ....
.. - . . ....
.. -
6 6 6
p p
2 · ej(!c t+✓T ) 2·e j(!c t+✓R )
ej(✓T ✓R )

Non-coherent receivers
I Receiver whose demodulator has a phase that is different than phase at modulator
I Produces a rotation in the received constellation

c Marcelino Lázaro, 2013 Digital Communications Linear modulations 111 / 113

Binary transmission rate (Rb bits/s)

Binary transmission rate is obtained as Rs = m ⇥ Rs
I Symbol rate (Rs bauds)
I Number of bits per symbol in the constellation (m)
m = log2 (M)
M: number of symbols of the constellation
Limitation in the achievable binary rate
I Limitation in Rs : available bandwidth (B Hz)
Using filters of the raised cosine family

BASEBAND BAND PASS

2B B
Rs|max = 1+↵ Rs|max = 1+↵

I Limitation on the number of symbols M (and therefore

h ini m)
F Power limitation limits mean energy per symbol Es = E |A[n]|2
- This limits the maximum modulus of the constellation
F Performance requirements limit the minimum distance between symbols
!
dmin
Pe ⇡ k · Q p
2 N0 /2
F Es and Pe determine a maximum constellation density
c Marcelino Lázaro, 2013 Digital Communications Linear modulations 112 / 113
Constellation density - Example - QAM
4-QAM (2 bits) 16-QAM (4 bits)
r r r r r r

r r r r
01 11 00 11 01 11 11 11 10 11

r r r r
00 11 01 11 11 11 10 11

r r r r r r
00 01 01 01 11 01 10 01

00 10 00 00 01 00 11 00 10 00

64-QAM (8 bits) 256-QAM (16 bits)

r r r r r r r r rr rr rr rr rr rr rr rr rr rr rr rr rr rr rr rr
r r r r r r r r rr rr rr rr rr rr rr rr rr rr rr rr rr rr rr rr
r r r r r r r r rr rr rr rr rr rr rr rr rr rr rr rr rr rr rr rr
r r r r r r r r rr rr rr rr rr rr rr rr rr rr rr rr rr rr rr rr
r r r r r r r r rr rr rr rr rr rr rr rr rr rr rr rr rr rr rr rr
r r r r r r r r rr rr rr rr rr rr rr rr rr rr rr rr rr rr rr rr
r r r r r r r r rr rr rr rr rr rr rr rr rr rr rr rr rr rr rr rr
r r r r r r r r rr rr rr rr rr rr rr rr rr rr rr rr rr rr rr rr

c Marcelino Lázaro, 2013 Digital Communications Linear modulations 113 / 113